An interaction between Sla1p and Sla2p plays a role in regulating actin dynamics and endocytosis in budding yeast

The importance of a dynamic actin cytoskeleton for facilitating endocytosis has been recognised for many years in budding yeast and is increasingly recognised in mammalian cells. However, the mechanism for actin recruitment and the role it plays in endocytosis is unclear. Here we show the importance of two yeast proteins in this process. We demonstrate that Sla1p and Sla2p interact in vitro and in vivo and that this interaction is mediated by the central domain of Sla2p, which includes its coiled-coil region, and by a domain of Sla1p between residues 118 and 361. Overexpression of the interacting fragment of Sla1p causes reduced fluid-phase endocytosis and, interestingly, defects in subsequent trafficking to vacuoles. We show that Sla2p is required for the polarised localisation of Sla1p in cells but not for its cortical localisation or for its overlapping localisation with actin. Generation of an Δsla1Δsla2 double mutant demonstrates that Sla2p is likely to act upstream of Sla1p in endocytosis, whereas sensitivity to latrunculin-A suggests that the proteins have opposite effects on actin dynamics. We propose that Sla2p recruits Sla1p to endocytic sites. Sla1p and its associated protein Pan1p then regulate actin assembly through interactions with Arp2/3 and Arp2/3-activating proteins Abp1p and Las17/Bee1p.

potentially due to the ENTH domain acting to localise the protein to regions of clathrin coats. Secondly, a region upstream of the talin homology region may mask its ability to bind to actin (McCann and Craig, 1997;Wesp et al., 1997;Yang et al., 1999). Thus Sla2p may only bind to actin via its talin-like domain when it is modified (e.g. phosphorylated) or when associated with other proteins that allow appropriate unmasking of the site.
Here we show that Sla1p and Sla2p associate both in vivo using a two-hybrid approach and in vitro using purified proteins. We also define the interacting domain as a stretch of 243 amino acids in Sla1p and of 458 amino acids in the coiledcoil region of Sla2p. The role of this interaction is investigated further using genetic and cell biological approaches. In addition, deletion of either or both genes is detrimental to the ability of cells to undergo endocytosis. Furthermore, in cells lacking both Sla1p and Sla2p, the cortical actin cytoskeleton is unable to polarise, and the majority of actin is localised to the distal end of the cell. On the basis of the experimental data shown here and of results from other published reports we propose a model in which Sla1p acts as a link between Sla2p, clathrin and other endocytic machinery and the cortical actin cytoskeleton. Within this complex Sla1p destabilises and facilitates depolymerisation of F-actin. Sla2p and other proteins may then promote formation of new F-actin structures, which are important for driving endocytosis.

Materials
Unless stated otherwise, chemicals were obtained from BDH/Merck or Sigma. Media was from Melford Laboratories (yeast extract, peptone, agar) or Sigma (minimal synthetic medium and amino acids). Latrunculin-A was a gift from Phil Crews (UC Santa Cruz, CA).

Yeast strains and growth conditions
Yeast strains and plasmids used in this study are listed in Table 1 and  Table 2, respectively. Unless stated otherwise, cells were grown in a rotary shaker at 30°C in liquid YPD medium (1% yeast extract, 2% Bacto-peptone, 2% glucose supplemented with 40 µg/ml adenine), except for KAY419, KAY439, KAY508 and KAY657 cells, which were grown under selection in synthetic media lacking uracil. Halo assays were performed as previously described (Ayscough et al., 1997). The mutant Sla1∆118-511 was myc-tagged using oligonucleotides and methods previously described .
Pulldown assays were carried out with purified GST-Sla1p bound to glutathione Sepharose beads as described previously . Whole cell lysates were spun at 2000 g for 4 minutes to remove unbroken cells, and the supernatant spun at 300,000 g for 20 minutes at 4°C. The final supernatant was removed and 100 µl incubated with either 20 µl of GST-Sla1p coated glutathione beads or with 20 µl of glutathione sepharose for 3 hours at 4°C with rotation. The beads were spun down at 500 g for 4 minutes and washed three times in 10 bed volumes of binding buffer. Bound proteins were eluted in sample buffer and separated on a 10% SDS polyacrylamide gel before transfer to PVDF for analysis.
Purified 6×His-tagged Sla2p was eluted from Ni-NTA beads in binding buffer (50 mM K-HEPES pH 7.5, 100 mM KCl, 1 mM EDTA, 1 mM EGTA + protease inhibitors as above) + 0.01% Triton X-100. Proteins were eluted from the beads in sample buffer and separated on a 10% SDS polyacrylamide gel before being transferred to PVDF for analysis. Western blotting of purified 6×His-Sla2p revealed that no Sla1p was co-purified (data not shown).
Yeast whole cell lysates were prepared as previously described  and samples were run on 10% SDS polyacrylamide gels. For immunoblots, affinity-purified rabbit antiyeast Sla2 antibodies (a gift from D. Drubin, UC Berkeley) were diluted 1:500. GST and GST-Sla1p were detected using affinitypurified polyclonal rabbit antiserum against the GST epitope at a 1:1000 dilution. Anti mouse and rabbit HRP-conjugated secondary antibodies (Sigma) were used at 1:5000 and detected using ECL chemiluminescence.
Yeast two-hybrid screen The yeast two-hybrid screen used bait and activation plasmids and a yeast strain pJ69-2A designed and constructed by Philip James (James et al., 1996). The region of Sla1p to be tested in the two-hybrid assay was subcloned from a plasmid carrying the entire SLA1 gene [pKA51 (Ayscough et al., 1999)]. Eco R1 digestion was used to cut out the fragment corresponding to amino acids 118 -511 from the gene, and this fragment was then cloned into the pGBDU-C1 bait vector. The bait plasmid was checked for self-activation before the library screen was carried out. For checking the requirement of the third SH3 domain, EcoR1 was used to digest plasmid pKA116, which had been constructed previously (Ayscough et al., 1999). This released the fragment as above but without the SH3 domain.
For the screen, 2 µg of bait plasmid (pKA237) was transformed into the two-hybrid yeast strain pJ69-2A to give a strain KAY500. 2×10 9 KAY500 cells, grown in liquid media lacking uracil, were transformed with 20 µg library DNA from each reading frame (library was a gift from Francis Barr, Munich, Germany). In total, the 202 colonies that grew within 7 days were picked from plates lacking histidine, uracil and leucine. Following re-streaking and selection for the inability to grow on media lacking adenine and containing 5fluororotic-acid (5-FOA), a total of 57 yeast strains remained and plasmids were extracted from these. The plasmids were then retransformed into pJ69-2A with the bait plasmid to ensure that activation of reporter genes still occurred. Twenty-six of the 57 yeast strains could not grow on activation plates. The remaining 31 plasmids were sequenced to identify the region responsible for the two-hybrid interaction, and the resulting sequence was compared with DNA sequences from S. cerevisiae using the BLAST alignment program found at the Saccharomyces Genome Database Website (http://genome-www.stanford.edu/Saccharomyces/). The interacting genes identified were YSC84 [17 isolates (Dewar et al., 2002)], SLA2 (7 isolates), SED5 (3 isolates), RAD52 (1 isolate) and non-coding DNA (3 isolates).

Fluorescence microscopy
Endocytosis of the fluid phase marker lucifer yellow was performed according to the published method (Dulic et al., 1991), and analysis of the uptake of lipophylic dye FM4-64 was performed as described previously (Vida and Emr, 1995). Rhodamine-phalloidin (Molecular Probes) staining of actin was performed as described elsewhere (Hagan and Ayscough, 2000). Cells were processed for immunofluorescence microscopy as described previously (Ayscough and Drubin, 1998). Antibodies were used to detect Sla2p, Abp1p, Sac6p (a gift from D. Drubin, UC Berkeley, CA), Rvs167p (a gift from H. Friesen, University of Toronto, CA) and the myc epitope (for Sla1p-myc; Santa Cruz Biotech, CA) for immunofluorescence microscopy at 1:100 dilution. Secondary antibodies used were FITCconjugated goat anti-rabbit (Vector Laboratories) at dilutions of 1:1000. Cells were viewed with an Olympus microscope BX-60 fluorescence microscope with a 100 W mercury lamp and an Olympus 100× Plan-NeoFluar oil-immersion objective. Images were captured using a Roper Scientific Micromax 1401E cooled CCD camera using IP lab software (Scanalytics, Fairfax, VA) on an Apple Macintosh G4 computer.

Electron microscopy
Log phase cells were fixed using freshly prepared 2% potassium permanganate for 45 minutes at room temperature. Following washing the pellets were processed by dehydration through a series of ethanol from 50-100% and then in propylene oxide. Samples were then incubated overnight in a resin (Durcapan):propylene oxide 1:1 mix before embedding in resin and curing. Sections were cut and stained with uranyl acetate and lead citrate before viewing them on a Zeiss 902 transmission electron microscope.

Results
The domain of Sla1p responsible for its role in actin organisation is not required for cortical localisation In previous studies we demonstrated that deletion of distinct domains of Sla1p affected certain functions of the protein. In particular, the region between amino acids 118 and 511 of the 1244 amino-acid protein was shown to be important for the functions of Sla1p in regulating cortical actin structures (Ayscough et al., 1999). The Sla1p 118-511 region encompasses a region of novel sequence between the second and third SH3 domains (amino acids 118-361), the third SH3 domain (amino acids 362-415) and a further 96 amino acids beyond the third SH3 domain. Deletion of the third SH3 domain did not have a detectable effect on actin structures in cells, indicating that a combination of this and the sequences either side of the SH3 domain are likely to be important for interacting with other cell proteins (Ayscough et al., 1999). To avoid concerns with slight variation in copy number from these studies on plasmid borne mutations, the sla1∆118-511 mutant for this study was integrated into the genome and expressed as the sole form of Sla1p in cells. Expression was detected using anti-Sla1p antibodies and was observed to be at similar levels to wild-type protein expression (Fig. 1B).
Our antibodies did not prove useful for indirect fluorescence microscopy so, in order to localise the mutant protein in cells, a myc tag was integrated at the C-terminus of the mutant Sla1∆118-511p. The myc-tagged protein was also detectable by western blotting and had no observable phenotypes in addition to those of the mutant itself (data not shown). By indirect fluorescence microscopy, Sla1∆118-511p was observed to localise to small punctate patches at the cell cortex, which indicates that the protein still contains signals required for its localisation to discrete cortical sites in the cell (Fig. 1C). The pattern of staining was not well polarised and does not show the same organisation as actin, which is in large cortical chunks in these cells (Ayscough et al., 1999) (Fig. 5B).
As well as showing that Sla1∆118-511p is expressed and localises to appropriate regions of the cell, we were also able to show that cells expressing the mutant are able to rescue a phenotype associated with the complete sla1 deletion. Cells lacking sla1 cannot grow in the absence of abp1 expression (deletion of both genes is synthetically lethal). However, a combination of sla1∆118-511 and ∆abp1 is not lethal to cells, which demonstrates, firstly, that the Sla1∆118-511 protein must be partly functional and, secondly, that the overlapping function with Abp1p lies in another part of the protein. Other work in the laboratory has indicated that this redundancy is attributable to the presence of the C-terminal repeat region of Sla1p (Ayscough et al., 1999). Further analysis of the sla1∆118-511 strain showed that it has similar resistance to the effects of the actin-disrupting drug latrunculin-A to cells in which the entire sla1 gene is deleted. Again, this indicates that the links with the actin cytoskeleton are mediated through this domain of the protein (data not shown).
A two-hybrid screen with the Sla1 118-511 region detects Sla2p as a binding partner To investigate the role of the Sla1(118-511) region further, a two-hybrid screen was conducted using this sequence as bait.
Journal of Cell Science 116 (12) (118-511)p or cells in which ∆sla1 was deleted (KAY302, 351 or 97 respectively) were grown overnight and cell extracts made. Proteins were separated by SDS-PAGE and western blotted using antibodies against Sla1p. As shown, proteins of the appropriate size were found in cells expressing Sla1p. No band was observed in the extract from ∆sla1 cells. As a loading control the same blots were probed with antibodies to a nuclear protein Anc1p. This showed equivalent levels in all lanes. (C) Log phase KAY303 and KAY367 cells expressing Sla1-myc and Sla1∆(118-511)myc were grown to log phase and processed for immunofluorescence microscopy. Both Sla1p-myc and Sla1∆(118-511)-myc localise to punctate patches at the cell cortex demonstrating that the parts of Sla1p required for cortical localisation are located elsewhere in the protein. Bar, 5 µM. Sla1and Sla2p link actin to endocytosis Details of the constructs used are given in Materials and Methods. Of 31 positive interactors, seven were subsequently sequenced and shown to encode Sla2p. Within the seven isolates two different start sites at amino acids 204 and 310 were identified, but they all carried the same end site corresponding to amino acid 768 in Sla2p. The minimum interacting sequence was therefore the region between amino acids 310 and 768. This region of Sla2p corresponds to its central domain, which is predicted to be largely coiled-coil in nature (Fig. 1A, Fig. 2A,B) Yang et al., 1999).
To ascertain whether the SH3 domain of Sla1p is important for the interaction with Sla2p, a modified bait plasmid was generated in which the SH3#3 domain is deleted. The twohybrid transformations were repeated, and an interaction was still found to occur, which indicates that the sequence important for the Sla1p:Sla2p interaction lies outside the third SH3 domain of Sla1p (Fig. 2). Interestingly, this interaction appeared to be stronger than with the construct containing the SH3 domain, which suggests that the domain might normally partly interfere with, or regulate, an Sla1p:Sla2p interaction. In addition, an Sla1p bait plasmid containing the sequence from residue 1 to 361 also interacted with the Sla2p activation plasmid, indicating that the region of Sla1p involved in the interaction with Sla2p lies between residues 118 and 361. Further interactions between Sla1p and Sla2p central domain fragments were also tested ( Fig. 2A,B), but none of these yielded a clear positive interaction, which might suggest that the interaction sites require a tertiary structure that cannot be generated by any of the Sla2p fragments.

Sla1p and Sla2p interact in vitro
Although an interaction between Sla1p and Sla2p has been shown through a two-hybrid approach, it is possible that this could occur through an intermediary protein. To investigate this possibility a plasmid expressing GST-Sla1p was transformed into yeast cells and purified from whole cell lysates. Following purification of Sla1p, cell extracts were passed over the protein on a column, and proteins in bound and unbound fractions were separated by SDS-PAGE and transferred to PVDF for analysis. Western blots were probed with antibodies raised against several proteins known to associate with cortical patches. Fig.  3A shows that under these conditions, an interaction was detected between GST-Sla1p and Sla2p.
To establish whether the interaction between Sla1p and Sla2p is likely to be direct and not mediated by an accessory, we purified 6×His-tagged Sla2p and performed a direct in vitro binding assay. The GST-Sla1p beads were blotted prior to Sla2p binding to verify that no Sla2p co-purified in the GST-Sla1p preparation (lower panel Fig. 3B). Fig. 3B clearly demonstrates that Sla1p and Sla2p interact in vitro. The controls presented demonstrate that the pulldown of purified Sla2p by Sla1p-coated beads is not due to the presence of the associated His or GST tags nor is it attributable to the presence of glutathione beads. In both GST pulldown experiments, Coomassie-stained gels showed that proteins other than Sla1p and Sla2p were not present at detectable levels in the purified protein extracts (data not shown).
Finally, an interaction between bacterially expressed GST-Sla1 fragment (118-511) and His-tagged Sla2p was demonstrated (Fig. 3C)  detectable by western blotting using antibodies against GST (for the GST-Sla1 preparation) or against Sla2p (His-tagged Sla2p preparation), which indicates that the Sla1p:Sla2p interaction is direct.

Fractionation of cell extracts indicates that Sla1p and Sla2p form a complex in vivo
Evidence that Sla1p and Sla2p may form a complex in vivo was produced by carrying out subcellular fractionation of yeast total protein extracts on a 3-30% sucrose gradient (Fig. 4). Fractions taken from the velocity gradient were probed with antibodies raised against Sla2p, Abp1p, actin, cofilin, clathrin heavy chain (Chc1p) and Sla1p. Identical fractionation was also performed on cell extracts from strains lacking sla1. A number of significant findings were made. Firstly, there are two peaks that are highly enriched for actin, Abp1p and cofilin (fractions 6-10 and 16-20), indicating these proteins are found in both small and large complexes in the cell. In extracts from wild-type cells, Sla1p and Sla2p are not found in the fraction 6-10 peak but do co-fractionate in the 16-20 peak, which represents larger protein complexes. Clathrin is also found in these same dense fractions. Interestingly, in the absence of Sla1p, the major Sla2p peak is shifted to a less dense fraction, indicating that it may have been part of an Sla1p complex, which is consequently smaller in the absence of this 140 kDa protein. A significant proportion of Sla2p has also moved out of the dense fractions into less dense fractions; 72% of Sla2p is in fractions 14-20 in wild-type cells compared with 43% in ∆sla1 cells. This indicates that Sla1p is important for maintaining the integrity of the Sla2p complex.
In the wild-type extracts 30% of actin was found in fractions 16-20, this falls to 20% in the absence of sla1, also suggesting a role for Sla1p in association of actin with the large complex. These proportional changes in actin and Sla2p fractionation patterns in the presence and absence of Sla1p were reproducible over the three experiments performed. Abp1p is found in two distinct peaks, both in the presence and absence of Sla1p, though in the absence of Sla1p the denser fraction is reduced in size, which indicates that it may have contained Sla1p in the wild-type cell. Therefore Abp1p localisation with both complexes appears to be independent of Sla1p. Interestingly, the small peak of cofilin that co-fractionates with the denser actin complexes completely disappears in the absence of Sla1p.
Localisation of Sla1p and Sla2p in wild-type and mutant strains Sla1p and Sla2p localise to cortical patches, a subset of which contain actin (Ayscough et al., 1999;Yang et al., 1999). In order to examine the distribution of Sla1p-myc and Sla2p, a co-immunofluorescence experiment was carried out (Fig. 5A). Cells are shown that represent different stages within the cell cycle of budding yeast. Both Sla1p-myc-and Sla2p-containing patches were polarised to the site of bud emergence, to the enlarging bud and to the bud neck (Fig. 5A). Furthermore, Sla1p and Sla2p showed substantial colocalisation throughout the cell cycle. However, it was also apparent that a number of cortical patches contained only Sla1p-myc and rather fewer appeared to contain Sla2p alone, which suggests that colocalisation was not always complete.
Deletion of the entire SLA1 coding region abrogated the interaction of Sla2p with the actin cytoskeleton (Ayscough et al., 1999), but our work detailed above shows that the interaction is mediated in vitro and in a two-hybrid assay by Journal of Cell Science 116 (12) Fig. 3. Sla1p and Sla2p interact in vitro. (A) GST-Sla1p was purified as described in Materials and Methods. Cell extracts were passed over the protein on a column and proteins in bound and unbound fractions were separated by SDS-PAGE and transferred to PVDF for analysis. Western blots were probed with antibodies raised against Sla2p. Sla2p bound only to Sla1p carrying beads and not to control beads alone. FT, flow through; W1,W3, washes 1 and 3; B, bound. (B) To investigate whether the interaction between Sla1p and Sla2p was direct, His-tagged Sla2p was purified from cells as described in Materials and Methods, and its binding to GST control beads or to GST-Sla1p beads was assessed. As shown in the upper panel Sla2p only binds to Sla1p-containing beads. As a control, the extracts from KAY419 cells from which GST-Sla1p was purified were probed to demonstrate the presence of Sla2p in the extract. Then, following GST-Sla1p purification the beads were probed to demonstrate that no Sla2p had co-purified with the Sla1p. Therefore, the Sla1p-Sla2p binding observed was only caused by the presence of the added Sla2p to the GST-Sla1p beads. FT, flow through; W1,W3, washes 1 and 3; B, bound. (C) Bacterially expressed GST or GST-Sla1(118-511∆SH3#3) was incubated with His-tagged Sla2p as described in Materials and Methods. The beads were washed and the input material (I), wash (W) and bound (B) fractions were run on a gel that was then stained using Coomassie dye. The three bands marked x are all detected by western blotting with GST antibodies indicating that they are all Sla1 degradation products; bands marked ‡ were detectable using Sla2 antibodies. Sla1and Sla2p link actin to endocytosis amino acids 118-511 of Sla1p. To demonstrate the importance of this interaction in vivo, Sla2p was localised in cells expressing mutant Sla1p (sla1∆118-511). As shown in Fig. 5B, while Sla2p is still able to localise to the cell cortex, it no longer shows an overlapping localisation with the cortical actin patches. Thus, both in vivo and in vitro, the 118-511 region of Sla1p mediates an interaction between Sla2p and the actin cytoskeleton.
Sla1p is required for the localisation of Sla2p to actincontaining cortical patches and for its correct polarisation during growth (Ayscough et al., 1999), and so here we examined the distribution of Sla1-myc in an ∆sla2-null background by immunofluorescence microscopy (Fig. 5C). We found that instead of Sla1p being localised to areas of cell growth as is seen in wildtype cells, its distribution was depolarised and Sla1p-containing patches appeared evenly distributed around the mother cell and bud. Co-staining with rhodamine-phalloidin revealed that Sla1p-myc still localises to a subset of cortical patches that contain actin (Fig. 5C, right). Co-staining for Sla1p-myc and actin in cells lacking sla2 also revealed the existence of novel cortical structures. Normally actin and Sla1p, when colocalised, show an apparently tight overlap . In the structures shown in Fig. 5C Sla1p and actin only partially overlap and Sla1p appears to be mostly localised at the cortex of the cell, whereas actin appears to localise on the cytoplasmic side of the structures (Fig. 5C, inset).

Overexpression of Sla1(118-511) causes defects in endocytosis and membrane trafficking
The coiled-coil domain of Sla2p interacts with clathrin light chain (Henry et al., 2002). Because Sla1p also interacts with this region of Sla2p, we reasoned that overexpression of Fig. 4. Sucrose density gradient fractionation of wild-type and ∆sla1 cells. Extracts from KAY302 (wildtype) and KAY300 (∆sla1) cells were prepared and separated on 3-30% sucrose gradients as described in Materials and Methods. Fractions were then run on SDS-PAGE and blotted onto PVDF membranes. The membrane was cut and probed with individual antibodies. Antibodies were used as described in Materials and Methods. Quantitation of band intensity was performed using NIH Image 1.6.1 Software. The y-axis of the graphs is the band intensity in arbitrary units normalised in each case for background. The gradients were repeated three times. Shown here is a representative example. Sla1(118-511∆SH3#3) might interfere with other functions of Sla2p requiring this domain. Cells induced by growth in galactose to overexpress Sla1(118-511∆SH3#3) were assessed for their ability to endocytose the dye lucifer yellow. As shown in Fig. 6A,B, a very marked phenotype is observed. There is a Journal of Cell Science 116 (12) (A) KAY303 cells expressing Sla2p and Sla1-myc were grown to log phase and processed for immunofluorescence microscopy as described in Materials and Methods. Both Sla1p-myc-containing and Sla2p-containing patches were polarised to the site of bud emergence, to the enlarging bud and to the bud neck. Furthermore, Sla1p and Sla2p showed substantial co-localisation throughout the cell cycle. However, as indicated by arrows, it was also apparent that a number of cortical patches contained only Sla1p-myc, which suggests that the colocalisation was not always complete. Bar, 10 µM. (B) KAY302 (wildtype) and KAY351 (∆sla1∆118-511) cells were grown to log phase and then fixed and processed for immunofluorescence microscopy as described in Materials and Methods. Actin was stained using rhodamine-phalloidin, and Sla2p was localised using antibodies raised to the protein. Bar, 10 µM. (C) The distribution of Sla1-myc in an ∆sla2-null background was examined using immunofluorescence microscopy. In wild-type cells Sla1p was localised to areas of active cell growth, whereas in the absence of sla2 expression Sla1p distribution was depolarised and Sla1p-containing patches appeared evenly distributed around the mother cell and bud. Co-staining with rhodamine-phalloidin revealed that Sla1p-myc was still localised to a subset of cortical patches that contained actin. These images also suggested that in cells lacking Sla2p there are novel structures containing distinct regions of Sla1p and actin organisation (arrows). Bars, 10 µM. The inset shows part of the same image magnified two-fold, and the arrow points toward the same structure in both figures.  , lucifer yellow uptake was disrupted. These cells showed little vacuolar staining compared to those expressing GST alone and possessed more abundant brightly stained dots outside the vacuole. Lucifer yellow staining at the plasma membrane of these cells also appeared brighter than in the control cells. (C) A mutant form of Sla1 (118-511∆SH3#3)* was expressed that cannot bind to another strong interactor of the 118-511 domain (Ysc84p). The same defect in uptake and subsequent trafficking was seen with this mutants with the Sla1(118-511∆SH3#3) mutant, indicating that the defect is not because of the domain blocking normal Ysc84 function. (D) Overexpression of full-length Sla1p abrogated lucifer yellow uptake but did not result in accumulation of stained endosomes. Bar, 10 µM. Sla1and Sla2p link actin to endocytosis reduction in uptake of lucifer yellow to the vacuole coupled with an increase in staining of smaller organelles, potentially endosomes, as well as increased plasma membrane staining. Thus Sla1(118-511∆SH3#3) overexpression causes reductions in both endocytosis and, interestingly, defects in subsequent trafficking to the vacuole. In order to determine whether this trafficking phenotype is largely due to Sla1(118-511) binding Sla2p but no longer being open to the regulation afforded by the presence of the rest of the molecule, two controls were performed. Firstly, full-length Sla1p was overexpressed. This caused a complete inhibition of endocytosis, and no endosomal staining was seen at all (Fig. 6D). Second, a mutant Sla1(118-511∆SH3#3)* was expressed. This sequence contains a mutation rendering the domain unable to interact with Ysc84p, the other Sla1p-binding partner identified in the original twohybrid screen (Dewar et al., 2002) (see also Materials and Methods and two-hybrid screen results). Cells expressing this mutant form of Sla1p sequence exhibited the same trafficking defect as in the Sla1(118-511∆SH3) cells (Fig. 6C).
Over the time course of overexpression the actin phenotype was also monitored. Despite the endocytic defect, cells overexpressing either of the Sla1 fragments did not have an obviously disrupted actin phenotype, and the majority of actin was polarised in small punctate patches, which suggested that these cells are producing partially functional proteins. However, overexpression of the full-length sequence causes actin to aggregate into large cortical chunks, which produces an almost complete block in fluid phase endocytosis.

The Sla2p:Rvs167p interaction requires the presence of Sla1p
Work from other laboratories has demonstrated a two-hybrid interaction between Sla2p and one of the yeast amphiphysin homologues Rvs167p . This interaction was postulated to facilitate the role of Sla2p in endocytosis. Rvs167p has also been reported to be an actin-associated protein (Cid et al., 1995). To address whether Rvs167p association with Sla2p occurred in the absence of Sla1p, we observed the localisation of Rvs167p in wild-type cells and in cells expressing the sla1 deletion mutant (Fig. 7). In wild-type cells there was a partial overlap between Rvs167p and actin localisation, whereas in the ∆sla1 cells Rvs167p colocalises completely with F-actin structures. From our previous data (Ayscough et al., 1999) and the data in Fig. 5B, we have shown that Sla2p does not colocalise with actin in the sla1 deletion strain. Therefore, because Rvs167p colocalises with actin under these conditions, its association with Sla2p is likely to occur after the interaction between Sla1p and Sla2p.

Characterisation of ∆sla1∆sla2 double mutant cells
To characterise the combined role played by Sla1p and Sla2p further, yeast strains were generated that did not expression both proteins, and the growth and morphology of these cells was analysed. ∆sla1∆sla2 double mutant cells appeared to grow at a similar rate and possessed a similar temperaturesensitive phenotype to that of ∆sla2 cells. However, comparisons between ∆sla1∆sla2 double mutants with wildtype, ∆sla1 or ∆sla2 cells revealed a unique cold-sensitive phenotype only when both Sla1p and Sla2p are absent (results are summarised in Table 3). The significance of this double mutant phenotype is not known at present, although cold sensitivity has been associated with mutations in proteins involved in forming multi-subunit complexes (Moir et al., 1982). It was also noted that the ∆sla1∆sla2 cells were increased in cell volume compared with either ∆sla1 or ∆sla2 cells. The volume of cells grown at 29°C was measured using a Schärfe CASY™ cell counter and the mean volume of wildtype cells is 5.76 fl and ∆sla1∆sla2 cells is 13.8 fl. These data are summarised in Table 3.
We have shown previously that the deletion of sla1 and sla2 individually influences the sensitivity of yeast to the actindisrupting drug Latrunculin-A (LAT-A). Because LAT-A binds to actin monomers, sensitivity to the drug can be related to whether a protein increases or decreases the stability of Factin in cells. When LAT-A was administered to cells in culture, as described previously, deletion of sla1 reduces the sensitivity to LAT-A whereas ∆sla2 cells are more sensitive to the effects of the drug (Ayscough et al., 1999;Warren et al., 2002). However, the double mutant cells displayed an intermediate phenotype, being more sensitive than ∆sla1 cells but less so than the ∆sla2 mutant (data not shown). A similar result was obtained when LAT-A sensitivity was assessed by a halo assay approach. A summary of the halo assay data is shown in Fig. 8A.   Fig. 7. Rvs167p localises with actin and not Sla2p in ∆sla1 cells. KAY302 (wild-type) and KAY 97 (∆sla1) cells were processed for immunofluorescence microscopy as described in Materials and Methods. Actin was stained using rhodamine-phalloidin, and Rvs167p was localised using antibodies raised against the protein.
Although actin organisation is altered in the sla1∆ strain, Rvs167p still localises to these aberrant structures, which suggests that its association with actin is not mediated by Sla1p. Bar, 10 µM. Actin patch organisation in ∆sla1∆sla2 cells The most striking phenotype of ∆sla1∆sla2 double mutant cells is observed when the actin cytoskeleton is stained with rhodaminephalloidin. Fig. 8B shows F-actin staining in wildtype, cells with deletions of either ∆sla1 or ∆sla2 or both ∆sla1 and ∆sla2. Wild-type cells demonstrated polarisation of the actin cytoskeleton to sites of cell growth. In ∆sla1 cells, cortical actin patches appeared larger than those in wildtype, were fewer and were less well polarised. In ∆sla2 mutant cells the documented phenotype of smaller, more numerous, actin patches that were distributed evenly throughout the mother cell and bud was seen (Holtzman et al., 1993). In ∆sla1∆sla2 double mutant cells a novel phenotype was observed. Cortical actin patches were primarily polarised to the distal end of the mother cell. In these cells, small, less brightly stained patches could often be seen at polarised regions of the cell (arrows) and an actin ring could be seen at the site of cytokinesis (arrowheads), but this represented a small fraction of actin that was visualised (Fig. 8B lower panel). The organisation of actin in the ∆sla1∆sla2 cells was unexpected as it appears to be opposite from the cell polarity as defined by the sites of bud growth. However, in these cells, budding appears to occur relatively normally and tubulin staining of cytoplasmic microtubules and the mitotic spindle is not aberrant (data not shown).
To determine whether the distally localised F-actin patches contained a normal complement of actin-binding proteins we used immunofluorescence microscopy to determine the localisation of the yeast fimbrin homologue Sac6p and the actin-binding protein Abp1p (Fig. 8C,D). Both of these proteins normally localise to cortical actin patches, and their localisation to the distally located actin patches in the ∆sla1∆sla2 cells indicates that this actin is likely to be folded correctly.

Endocytosis and membrane trafficking in ∆sla1∆sla2 cells
Defects in endocytosis have been reported in both ∆sla1 and ∆sla2 single mutant strains. The effect of sla1 deletion on fluid phase uptake of the fluorescent dye lucifer yellow is not absolute, with about 30% of cells showing no uptake and 60-70% of cells showing a reduction in uptake of the dye . Sla2p appears to play a more critical role in endocytosis with most cells showing only a low level of lucifer yellow uptake (Fig.  9A) (Raths et al., 1993). The ∆sla1∆sla2 cells show a phenotype that is more severe than the ∆sla2 strain, and we observed less than 1% of cells with any fluid phase uptake of the dye. The percentage of cells showing vacuolar staining was Journal of Cell Science 116 (12) Fig. 8. The actin phenotypes of ∆sla1∆sla2 cells. (A) Sensitivity to the actin monomer binding drug LAT-A is proposed to reflect the level of actin monomer in cells and therefore can be used to assess to whether a protein increases or decreases the stability of F-actin. Sensitivity to LAT-A was assessed by halo assay as described in Materials and Methods. These results are summarised graphically. Deletion of sla1 reduces the sensitivity to LAT-A whereas ∆sla2 cells are more sensitive to the effects of the drug. The double mutant strains displayed an intermediate phenotype. The data shown are the average of three halo assays. Strains tested were KAY302, KAY97, KAY136 and KAY128. (B) Rhodamine-phalloidin was used to stain the F-actin of wild-type (KAY302), ∆sla1 (KAY97), ∆sla2 (KAY136) and ∆sla1∆sla2 (KAY128) cells. In the absence of both sla1p and Sla2p, actin is largely localised to the distal pole of the cell. However, in these cells small, less bright patches could often be seen at polarised regions of the cell (arrows), and actin was also seen at the cytokinetic ring (arrowheads) but this represented a small fraction of the actin that was visualised. Bar, 10 µm. To ascertain whether this larger aggregation of actin patches was likely to be functional, the localisation of two F-actin-binding proteins was determined. Both Sac6p (C) and Abp1p (D) were also observed to localise to the distal regions of the cell, indicating that the actin in these structures is associated with a normal complement of actin-binding proteins. Bar, 10 µm. assessed for all strains; this is summarised graphically in Fig.  9B and all experiments were performed three times, counting 200 cells of each strain on every occasion.
Uptake of the dye FM4-64 was also monitored in wild-type and mutant cells. This dye can be used to study mutants that have defects in membrane trafficking in addition to in the initial internalisation step of endocytosis (Vida and Emr, 1995). All of the mutant strains showed a kinetic delay in FM4-64 uptake, although by 60 minutes, labelling was similar to that found in wild-type cells. Wild-type cells showed vacuolar staining after 15 minutes, ∆sla1 and ∆sla2 after 25 minutes and ∆sla1∆sla2 cells showed vacuolar staining after 35 minutes, although this was weaker than in the wild-type or single mutant strains.
Finally we used electron microscopy to investigate whether there were any marked changes in number or organisation of membrane structures. Previously, ∆sla2 cells were shown to accumulate vesicles in the bud (Mulholland et al., 1997). As shown in Fig. 9C-F, there is a clear increase in the number of vesicles in both ∆sla2 and ∆sla1∆sla2 cells. One difference however between these two mutants is that the vesicles in the double mutant appear to accumulate in both the mother and bud, whereas the vesicles in the ∆sla2 mutant appear mostly to localise to the bud.

Discussion
The work presented here provides new insights into the interactions between the endocytic machinery and the proteins of the cortical actin cytoskeleton. We show that Sla1p binds to Sla2p both in vivo and in vitro. We also demonstrate that this interaction is mediated by amino acids 118-361 of Sla1p and the region 310-768 of Sla2p, which encompasses its central coiled-coil domain. The interaction does not require the third SH3 domain of Sla1p. The central domain of Sla2p, which includes two regions with coiled-coil structure, has previously been shown to be important for its role in endocytosis  and has been reported to interact with the actinregulating kinase Ark1p . More recently this domain has been shown to interact with clathrin light chain (Henry et al., 2002).
On the basis of the data shown above, previous work in our laboratory and work from other laboratories, we believe that current data are consistent with the following model for the role of the interaction of Sla1p and Sla2p in coupling cortical actin to endocytosis. It should be noted, however, that this model does not aim to be all inclusive and that both proteins are also likely to make interactions with other proteins at the cell cortex that may also play a role in the endocytic process. Our model is summarised in Fig. 10. Numbers indicate potentially sequential interactions.
(1) Sla2p associates with clathrin light chain at sites marked for endocytosis. This localisation to the cortex may also be mediated by the Sla2p ENTH domain, which is likely to interact with PtdIns(4,5)P2 in the membrane. The demonstration that deletion of SLA2 results in loss of both actin and Sla1p polarity (Fig. 5) strongly implicates Sla2p localisation at the cortex as an early step in the process of endocytosis (Holtzman, 1993;Henry et al., 2002).
(2) Sla2p is then able to bind to Sla1p, which itself is already complexed to Pan1p and End3p (Zeng et al., 2001). Sla1p does not appear to be recruited from the cytosol to the Sla2p sites; rather it is found in pre-existing complexes at the plasma membrane that are distinct from both Sla2p (Fig. 5A) and actin complexes .
(3) Sla2p in association with Sla1p, End3p and Pan1p may then recruit actin patches. Sla1p binds to the Arp2/3-activating protein Abp1p  and Las17/Bee1p (Li, 1997;Warren et al., 2002). Pan1p has been shown to associate with Arp2/3 (Duncan et al., 2001). These interactions (and most probably others) then facilitate cycles of actin assembly and disassembly, which allow reorganisation of the actin cytoskeleton, which in turn promote endocytosis.
In ∆sla2 cells, Sla1p cannot localise to the endocytic sites, which in turn means there is no actin recruitment to these sites. This results in a very significant reduction in levels of endocytosis. The loss of ∆sla1 is rather more subtle, possibly because there are proteins with roles that partially overlap with that of Sla1p. In the absence of Sla1p there is an increase in F-actin, visualised as fewer larger chunks of actin in cells (Holtzman et al., 1993). This correlates to a reduction but not total abrogation of endocytosis . The effect of sla1 deletion on F-actin organisation indicates that the Sla1p interaction with the Arp2/3 activators Abp1p and Las17/Bee1p may be inhibitory. In wild-type cells, Sla1p may downregulate actin polymerisation and thereby result in net depolymerisation. Thus, when Sla1p is absent, there is an imbalance in actin assembly and disassembly rates, and this leads to accumulation of actin chunks, which are relatively stable However, there is possibly still some level of actin dynamics that means that endocytosis is not completely prevented. It is interesting to speculate from the sucrose density gradient analysis (Fig. 4) that the absence of cofilin from large protein complexes may serve to partly explain the lack of depolymerisation in the absence of Sla1p.
The observation that Sla1p and Sla2p proteins have opposite effects on actin dynamics in cells even though they are found in overlapping complexes is likely to be crucial (Fig. 5). As mentioned above, deletion of sla1 causes cells to become resistant to the effects of LAT-A, which indicates that it normally has a destabilising effect on F-actin, whereas deletion of sla2 causes a marked increase in sensitivity to LAT-A, suggesting a role for Sla2p in stabilising F-actin (Ayscough et al., 1997) (Fig.  8A). Actin rearrangements require both disassembly of existing structures (therefore a destabilising/disassembling activity is required) and subsequently reassembly into new structures (requiring nucleating/stabilising properties). Talin has been shown to stabilise F-actin structures (Hemmings et al., 1996;Priddle et al., 1998) and so it is possible that the talin homology region of Sla2p may perform a similar role to this mammalian protein. The Sla2p actin-binding function appears to be masked by intramolecular interactions that have been demonstrated to be due to sequences within the central domain . This Sla2p domain, however, appears to be involved in a number of interactions. As well as interacting with itself to form a dimer, associations have also been reported with clathrin light chain (Henry et al., 2002), the actin-regulating kinase Ark1p  and, as described here, with Sla1p. It is possible that association with one or more of these proteins leads to unmasking of the actin-binding site, which then facilitates formation of actin structures that favour endocytosis. (1) Sla2p associates at the cell cortex with clathrin light chain and also potentially directly with the plasma membrane through its ENTH domain. (2) The central region of Sla2p is able to recruit Sla1p (through an interaction with a region encompassed by residues 118-361 of Sla1p) and the associated endocytic machinery Pan1p/End3p. (3) Sla1p also binds to the actin-associated proteins Abp1p and Las17p/Bee1p, allowing recruitment of actin patches to these endocytic sites These interactions then lead to the reorganisation of the actin cytoskeleton, which is likely to promote endocytosis. Evidence for these is from these studies: 1, Henry et al. (Henry et al., 2002); 2, this study; 3, Duncan et al., Li and Warren et al. (Duncan et al., 2001;Li, 1997;Warren et al., 2002).
We also show that overexpression of the Sla1-Sla2 interaction domain (Sla1-118-511) can block both endocytosis and membrane trafficking to the vacuole. This suggests that Sla1p and Sla2p must normally become dissociated for endocytosis to proceed. Work by Zeng and colleagues has indicated that phosphorylation of the C-terminal region of Sla1p by the actin-regulating kinase Prk1p causes it to dissociate from cortical complexes (Zeng et al., 2001). This may then be a pre-requisite for endocytosis. In our experiment the fragment Sla1(118-511) cannot be regulated by Prk1p because it lacks the C-terminal motifs that become phosphorylated. It may therefore continue to associate with the endocytic complex, presumably blocking interactions that are required for endocytosis to proceed. Interestingly, there appears to be a low level of fluid phase uptake of the dye, into small organelles, which may be endosomes. Alternatively, these structures could be similar to the previously reported finger-like invaginations, which may not have yet pinched off from the plasma membrane (Wendland et al., 1996). This result may suggest that proteins with which Sla1p/Sla2p interact also act downstream of the initial endocytosis events.
Finally, we report the consequences of combining the ∆sla1 and ∆sla2 mutations in cells. This leads to a complete block in fluid phase endocytosis and also to a novel actin phenotype in which the majority of F-actin is localised to the distal pole of the cell. The reason for this organisation of cortical actin is not clear. It does however illustrate that lack of polarised cortical actin does not abrogate polarised growth of cells and adds further support to the importance of actin cables rather than cortical actin patches in generating and maintaining cell polarity through directed cell growth (Evangelista et al., 2002;Sagot et al., 2002) The role of actin in the process of endocytosis is not yet clear. However, studies in mammalian cells have shown that depolymerisation of F-actin can increase the lateral movement of clathrin-coated pits, which suggests that the actin cytoskeleton provides a scaffold at the plasma membrane for the endocytic machinery (Gaidarov et al., 1999). Other ideas for the role of actin include a role in invagination of the membrane, fission of the vesicle or movement of the vesicle into the cell (reviewed in Qualmann et al., 2000). The data we present here suggest that actin is not required for the initial marking of the site of endocytosis. Rather it is required to be dynamic and is rearranged at sites of endocytosis. Potentially it is these rearrangements that drive formation of the vesicle or its inward movement into the cell.