It is commonly believed that the assembly of a polarized actin cytoskeleton is achieved by concentrating regulators of actin assembly in subdomains at the cell cortex. A complex and poorly understood interplay between these regulators leads to actin assembly and the recruitment of factors that stabilize and specialize the network. One such regulator is Aip3p (Amberg et al., 1997
). Aip3p is critical to the establishment and/or maintenance of a polarized actin cytoskeleton. Spatially and temporally, Aip3p localization precedes actin polarization during the budding cell cycle of S. cerevisiae,
and loss of AIP3
leads to gross defects in many (if not all) aspects of cell polarity (Amberg et al., 1997
). Because little is known about mechanisms underlying the polarized localization of regulators of the actin cytoskeleton, we have focused on understanding the mechanism behind the regulation of Aip3p localization. Here we present genetic and biochemical data that strongly support the model that Aip3p is delivered to sights of polarized actin assembly by the secretory pathway.
Aip3p localization was scrutinized in strains mutant for proteins whose localization behavior mimicked that of Aip3p. We examined several such mutants and those that affected Aip3p localization were all tied directly to the function of the secretory pathway. These included the bona fide “sec”
mutants, sec2, sec3, sec4,
(see Figure ), as well as mutants in the type V myosin Myo2p and its regulatory light-chain calmodulin (see Figure ). Some of the secretory pathway components we examined possess a polarized localization pattern like Aip3p, such as components of the “exocyst” (TerBush and Novick, 1995
), the small GTPase Sec4p (Brennwald and Novick, 1993
), and Sec3p (Finger et al., 1998
); however, we found that early acting Sec proteins, such as Sec14p and Sec17p, are also essential for Aip3p localization. This argues that the entire secretory pathway must be operational for normal Aip3p delivery. Mutations in VRP1, SPA2, LAS17,
only mildly affected GFP-Aip3p localization, suggesting that they are not directly involved in the Aip3p localization pathway.
was first identified as a mutant defective in cell morphology (Johnston et al., 1991
), but the protein it encodes has since been determined to be involved in late secretory vesicle delivery (Lillie and Brown, 1994
; Govindan et al., 1995
), probably by transporting late secretory vesicles on actin cables that terminate in the bud (Pruyne et al., 1998
). Our discovery that a MYO2
mutant is defective for Aip3p localization indicates that Myo2p must act on Aip3p-bearing secretory vesicles. A mutation in the actin-binding motor domain of Myo2p (myo2-66
) severely affected Aip3p localization. However, a mutation in the tail domain of Myo2p known to cause vacuolar inheritance defects (Catlett and Weisman, 1998
) had a much less disruptive effect on Aip3p localization (see Figure ). We interpret these data to indicate that Myo2p delivers Aip3p-bearing vesicles by translocating these vesicles on actin cables.
Our data also demonstrate a requirement for calmodulin in Aip3p delivery. We believe the calmodulin connection to Aip3p localization is most likely through Myo2p. It is known that calmodulin is bound to Myo2p (Brockerhoff et al., 1994
) and that calmodulin in other organisms acts as a regulatory light chain for many different species of myosin motors. It is interesting that we observed strong allele specificity in calmodulin mutants with respect to defects in Aip3p localization. The construction and genetic analysis of a set of calmodulin phenylalanine-to-alanine mutants previously led to the identification of four distinct functions for yeast calmodulin (Ohya and Botstein, 1994a
). Intragenic complementation between the mutants was used to assign them to four functional groups. We found that mutants in one of the four complementation groups were most defective for Aip3p localization and that this defect could largely be attributed to a F12A mutation in calmodulin (see Figure ). These data indicate that the function affected in this complementation group is one that serves to activate the Myo2p motor to move late secretory vesicles (Aip3p-bearing vesicles included) on actin cables.
An argument can be made that the secretory pathway involvement in Aip3p localization is indirect and is, for example, due to a failure to deliver a transmembrane receptor for Aip3p to the cell surface. This model could explain the Aip3p localization defects in secretory pathway mutants, but it does not account for the fractionation behavior of Aip3p. By crude fractionation and differential centrifugation (Figure ), velocity gradient analysis (Figure ), and floating and density gradient analysis (Figure ), a pool of Aip3p behaves like a secretory vesicle–associated protein. The sequence of Aip3p has no obvious signal peptide and no potential transmembrane domains and is in fact quite highly charged. Therefore, we believe that Aip3p is a peripherally associated membrane protein and that its association with vesicles must be through an interaction with another membrane-associated protein. Aip3p is not unique in this behavior; biochemically, Aip3p behaves similarly to the rab-like GTPase Sec4p. Sec4p cycles on and off late secretory vesicles in a mechanism that is believed to be tied to the GTPase cycle of the protein (Novick et al., 1993
). In crude fractionation experiments analogous to those we performed in Figure , Sec4p also partitions between cytosolic and microsome-associated pools (Goud et al., 1988
What are the proteins directly involved in Aip3p delivery and targeting? We have begun to address this question by defining the sequence elements within Aip3p that are involved in the interactions required for proper protein delivery and localization (Figure E). Mutational analysis has identified a region at the N terminus of Aip3p that is necessary and sufficient for Aip3p localization. Within this region of Aip3p is a sequence that is conserved with the N terminus of a protein of unknown function from S. pombe (Figure A). Exogenous overexpression of this domain of Aip3p interferes with Aip3p localization in trans (Figure B), and the N terminus of the S. pombe homologue can functionally replace the N terminus of Aip3p in the Aip3p localization pathway (Figure B). These data lead us to conclude that the sets of protein–protein interactions and the pathways required for Aip3p delivery and localization have been conserved and are mediated through a sequence motif we designate as the Aip3p “addressing domain.”
Several previous observations have suggested that Aip3p localization does not require the actin cytoskeleton. On the basis of the work presented here, it is clear that the Aip3p-actin interaction that led to our initial discovery of Aip3p (Amberg et al., 1997
) is unimportant for Aip3p localization (see Figure E). However, the experiments in this article strongly suggest that Aip3p is delivered to the bud site in association with secretory vesicles that are moved on actin cables. When cells are released from Go
in the presence of the actin disassembly drug latrunculin A, 30% of the cells are able to localize Aip3p to the bud site (Ayscough et al., 1997
). This actin-independent localization likely reflects diffusion of Aip3p to and capture by a receptor for Aip3p at the cell surface, and whether this is free Aip3p or vesicle-associated Aip3p we do not know. Similarly, the observation that Aip3p localization precedes actin polarization during the cell cycle may reflect a diffusion-based mechanism as well. Capture of a small amount of Aip3p may be sufficient to help seed actin assembly at these sites, leading to further polymerization events, including actin cable extension. Once cables have formed, then we expect that Myo2p begins the delivery of large amounts of Aip3p and associated proteins to the bud site so that a polarized actin cytoskeleton can be fully assembled and then maintained. Because actin filaments in cells are unstable and are continuously disassembling (Ayscough et al., 1997
), we expect that maintaining a polarized actin cytoskeleton requires the continual delivery of positive regulators of actin assembly. This may in fact be the major function of secretory pathway–mediated delivery of Aip3p: to deliver Aip3p-associated proteins (such as actin itself) to the bud site, acting like a molecular taxi cab.
This secretory pathway–mediated, Aip3p localization pathway is not used by all cell polarity effectors. Localization of the formin Bni1p, known to interact with Aip3p (Evangelista et al., 1997
), is unaffected by disruption of the secretory pathway. In addition, Aip3p and Bni1p are not dependent on each other for their localization. The phenotypic similarity between aip3Δ
strains, the fact that the two proteins interact with each other and show the same localization pattern, suggests they operate in the same pathway, and yet we have found that they are localized by different mechanisms. Collectively, these observations suggest a model for regulation of cell polarity in which different regulators are focused at the cell surface by independent methods and that perhaps it is the spatial overlap of these proteins that leads to the efficient assembly and maintenance of a polarized actin cytoskeleton.