MICROBIOLOGY AND MOLECULAR BIOLOGY REVIEWS, Mar. 2006, p. 37–120 Vol. 70, No. 1 1092-2172/06/$08.00ϩ0 doi:10.1128/MMBR.70.1.37–120.2006 Copyright © 2006, American Society for Microbiology. All Rights Reserved.

The BAR Domain Proteins: Molding Membranes in Fission, Fusion, and Phagy Gang Ren,1,4 Parimala Vajjhala,1 Janet S. Lee,1 Barbara Winsor,4 and Alan L. Munn1,2,3* 1 2 Institute for Molecular Bioscience, ARC Special Research Centre for Functional and Applied Genomics, and School of Downloaded from Biomedical Sciences,3 University of Queensland, St. Lucia, Queensland 4072, Australia, and UMR7156, Centre National Recherche Scientifique, Universite´Louis Pasteur, Strasbourg 67084, France4

INTRODUCTION ...... 39 BUDDING YEAST Rvs PROTEINS...... 40 Nutrient Availability and the Control of Cell Proliferation ...... 40 Rvs161p and Rvs167p Proteins and Their Common BAR Domain...... 40 BAR Domains of Rvs161p and Rvs167p Assemble into Heterodimers ...... 40 Loss of Rvs161p or Rvs167p Causes a Similar and Diverse Spectrum of Phenotypes ...... 42 http://mmbr.asm.org/ Reduced viability upon starvation (Rvs؊)...... 42 Growth sensitivity to salt (especially Na؉)...... 42 Growth sensitivity to cytotoxic compounds ...... 42 Growth sensitivity to elevated temperature...... 43 Growth on nonfermentable carbon sources ...... 43 Meiosis and sporulation ...... 43 Heterogeneous cell size and morphology upon starvation or exposure to Na؉ ...... 43 Loss of actin cytoskeleton polarization to sites of polarized growth...... 43 Delocalized cell wall chitin deposition...... 44

Loss of bipolar bud site selection...... 44 on October 26, 2015 by University of Queensland Library Defective fluid-phase and receptor-mediated endocytosis...... 44 Inefficient cell-cell fusion during mating...... 44 Rvs161p and Rvs167p Structure-Function Relationships...... 45 Rvs Proteins Localize to the Cortical Actin Cytoskeleton ...... 45 Subcellular localization of Rvs161p and Rvs167p...... 45 Interactions between Rvs167p and actin patch proteins...... 47 Roles of Rvs161p and Rvs167p in Endocytosis ...... 49 Rvs161p and Rvs167p in receptor-mediated internalization of ␣-factor...... 49 Role for Rvs161p and Rvs167p in postinternalization trafficking through endosomes?...... 57 Roles of Rvs161p and Rvs167p in the Secretory Pathway ...... 58 Rvs161p and Rvs167p are not essential for all secretory membrane traffic...... 58 Rvs proteins may be required for polarized secretion during cell division ...... 58 Rvs proteins and post-Golgi apparatus traffic ...... 58 Rvs proteins and ER-to-Golgi apparatus traffic...... 59 Rvs proteins and mating...... 59 rvs Mutations Interact with Actin and Myosin Mutations ...... 63 Rvs؊ phenotypes are associated with mutations affecting other actin cytoskeleton proteins ...... 63 Genetic interactions between Rvs proteins and actin...... 67 Genetic interactions between Rvs proteins and yeast myosins ...... 67 Rvs Proteins and Membranes...... 68 Association of cortical actin cytoskeleton with membranes...... 68 Fractionation of Rvs proteins with membranes ...... 68 Rvs proteins and lipid rafts ...... 68 Suppression of Rvs؊ phenotypes by mutations affecting glycosphingolipid biosynthesis...... 69 Regulation of Rvs167p by Phosphorylation ...... 72 Regulation of Rvs167p by Ubiquitination ...... 73 Genomic and Proteomic Approaches and the Diverse Network of Rvs Interactions...... 74 Large-scale two-hybrid screens for Rvs161p- and Rvs167p-interacting proteins ...... 74 Identification of Rvs167p SH3 domain-interacting proteins by phage display...... 74 Identification of Rvs161p- and Rvs167p-interacting proteins by high-throughput proteomics...... 74 Mapping the genetic interactions of Rvs161p and Rvs167p by synthetic genetic array...... 74 Rvs161p and Rvs167p physical and genetic interactions suggest multiple functions in vivo...... 74

* Corresponding author. Mailing address: Institute for Molecular Bioscience, University of Queensland, St. Lucia, Queensland 4072, Australia. Phone: 61-7-3346 2017. Fax: 61-7-3346 2101. E-mail: a.munn @imb.uq.edu.au.

37 38 REN ET AL. MICROBIOL.MOL.BIOL.REV.

FISSION YEAST BAR DOMAIN PROTEINS Hob1p AND Hob3p...... 75 Hob3p, Fission Yeast Ortholog of Rvs161p ...... 75 Hob3p has a domain structure similar to that of Rvs161p...... 75 Hob3p functions in polarization of the cortical actin cytoskeleton but not endocytosis...... 75 Functional conservation between Hob3p, Rvs161p, and human Bin3...... 75 Hob1p, Fission Yeast Ortholog of Rvs167p ...... 75 Hob1p has a domain structure similar to that of Rvs167p...... 75 Hob1p is not required for polarization of the cortical actin cytoskeleton or endocytosis ...... 75

Hob1p interacts with actin assembly proteins and protein kinases that regulate cell polarity...... 76 Downloaded from Hob1p functions in regulating cell cycle progression in response to nutrient availability...... 76 Functional Conservation of Hob1p and Bin1 and Divergence of Rvs167p ...... 76 Hob1p functions in regulation of cell cycle progression in response to DNA damage...... 76 Hob1p functions in a cell stress signal transduction pathway upstream of Hob3p ...... 77 VERTEBRATE BAR DOMAIN PROTEIN AMPHIPHYSIN 1 ...... 77 Endocytic Recycling of Synaptic Vesicles ...... 77 Role of dynamin 1 in synaptic vesicle recycling ...... 77 Discovery of Amphiphysin 1 in Chicken Brain ...... 77

Chicken amphiphysin 1 has a domain structure similar to that of budding yeast Rvs167p...... 77 http://mmbr.asm.org/ Chicken amphiphysin 1 is expressed in neurons ...... 77 Minor pool of chicken amphiphysin 1 is tightly associated with synaptic vesicles...... 78 Discovery of Human Amphiphysin 1...... 78 Human amphiphysin 1 is the autoantigen in stiff-man syndrome with breast cancer...... 78 Human amphiphysin 1 has a domain structure similar to that of yeast Rvs167p ...... 78 Mammalian amphiphysin 1 is highly expressed in the brain but is also expressed in other tissues...... 78 Amphiphysin 1 interacts with the endocytic proteins dynamin 1 and synaptojanin 1...... 78 Amphiphysin 1 Interacts with Clathrin and the AP-2 Adaptor...... 80 Regulation of Amphiphysin 1 Interactions by Phosphorylation ...... 80

Amphiphysin 1 Subcellular Localization in Presynaptic Terminals ...... 80 on October 26, 2015 by University of Queensland Library Amphiphysin 1 localizes to endocytic intermediates ...... 80 Amphiphysin 1 also localizes to the actin-rich cytomatrix...... 81 Amphiphysin 1 functions in polarized growth...... 81 Amphiphysin 1 Functions in Endocytosis ...... 81 Amphiphysin 1 interaction with dynamin 1 is essential for endocytosis in neurons...... 81 Amphiphysin interaction with dynamin is also essential for endocytosis in nonneuronal cells ...... 83 Amphiphysin 1 interacts with clathrin and AP-2...... 83 Appendage domain of ␣-adaptin mediates interaction with amphiphysin 1 ...... 84 Amphiphysin 1 interaction with clathrin and AP-2 is essential for endocytosis...... 84 Amphiphysin 1 interacts with endophilin A1/SH3p4/SH3GL2 ...... 84 Amphiphysin 1 forms homodimers ...... 84 Regulation of amphiphysin 1 complex formation by phosphorylation and dephosphorylation ...... 84 Protein phosphatase 2B/calcineurin is responsible for dephosphorylation of amphiphysin 1, dynamin 1, and synaptojanin 1 upon nerve stimulation...... 85 Proline-rich central insert domain of amphiphysin 1 is phosphorylated by Cdk5/p35 in vivo...... 85 Amphiphysin 1 Knockout Mice...... 86 Amphiphysin 1 is important but not essential for synaptic vesicle recycling in vivo...... 86 Essential role for amphiphysin 1 in learning...... 87 Amphiphysin 1 and Molding of Membranes ...... 87 Dynamin 1 binds membranes and evaginates tubules in vitro ...... 87 Amphiphysin 1 binds membranes and evaginates tubules in vitro...... 89 Amphiphysin 1 binding to membranes is dependent on lipid composition ...... 89 Amphiphysin 1 and dynamin 1 coassemble into rings on membrane tubules...... 89 Amphiphysin 1 stimulates assembly of dynamin 1 into thick rings in solution...... 89 Amphiphysin 1 coordinates clathrin bud and dynamin 1 tubule formation...... 89 Amphiphysin 1 stimulates the activity of dynamin 1 in membrane tubule fission ...... 89 In vivo role for amphiphysin 1 in membrane tubule formation ...... 90 Membrane curvature and amphiphysin 1 binding...... 90 Amphiphysin 1 tubulation of membranes is dependent on lipid composition...... 90 Domains of amphiphysin 1 required for membrane tubulation ...... 90 Domains of amphiphysin 1 required for coassembly with dynamin 1 into rings in solution...... 91 SECOND ISOFORM OF AMPHIPHYSIN, Bin1...... 91 Discovery of Bin1 ...... 91 Bin1؉6a؉12؉13/Amphiphysin 2/Amphiphysin II/BRAMP2 in the Brain...... 93 Distribution of Bin1؉12؉13 in the brain...... 93 Localization of Bin1؉6a؉12 on purified plasma membranes...... 93 Association of Bin1 with membranes...... 93 VOL. 70, 2006 BAR DOMAIN PROTEINS 39

Recruitment of Bin1؉6a؉12 to the plasma membrane...... 94 Interaction of Bin1؉12 with Other Brain Proteins...... 94 Bin1؉12 (with or lacking exons 6a and 13) interacts with dynamin 1 and synaptojanin 1 in the brain...... 94 Bin1؉12 (with or lacking exons 6a and 13) interacts with the ␣-adaptin subunit of AP-2 in the brain...... 94 Bin1؉12 (with or lacking exons 6a and 13) interacts with clathrin heavy chain in the brain...... 94 Interaction of Bin1؉12 with endophilin A1/SH3p4/SH3GL2 in the brain...... 94 ؉ ؉ Bin1 6a 12 forms homodimers and heterodimers in the brain...... 95 Downloaded from Interaction of amphiphysin 1/Bin1؉6a؉12 heterodimers with dynamin 1 in the brain ...... 95 Bin1؉6a؉12 is regulated by phosphorylation and dephosphorylation...... 95 Bin1؉6a؉12 functions in synaptic vesicle recycling/endocytosis ...... 95 Bin1؉12 may function in postinternalization transport through endosomes ...... 96 Bin1؉10؉13/Bin1...... 96 Identification of Bin1؉10؉13/Bin1...... 96 Domain structure of Bin1؉10؉13...... 96 Interaction of Bin1؉10؉13 with c-Myc ...... 97

/Bin1؉10؉13 is a tumor suppressor ...... 97 http://mmbr.asm.org Bin1؉10؉13 induces apoptosis specifically in tumor cells...... 98 Subcellular localization of Bin1؉10؉13 ...... 98 Stability of Bin1؉10؉13 in vivo...... 99 Phosphorylation of Bin1؉10؉13...... 99 Role for Bin1؉10؉13 in muscle differentiation...... 99 Bin1؉10؉13 localization in adult skeletal muscle...... 100 Bin1؉10؉13 and T-tubule formation in muscle...... 100 Bin1؉10؉13 and lipid rafts...... 101 Bin1؉10؉13 binds and tubulates liposomes in vitro ...... 101 ؊ ؊ ؉ Bin1 10 12 13/SH3p9 ...... 101 on October 26, 2015 by University of Queensland Library Bin1؊10؊12؊13/ALP1/Amphiphysin IIm ...... 101 Bin1؊10؊12؊13 in regulation of the actin cytoskeleton ...... 102 Bin1؊10؊12؊13 functions in phagocytosis in macrophages ...... 102 AMPHIPHYSIN-RELATED PROTEIN Bin2...... 103 AMPHIPHYSIN-RELATED PROTEIN Bin3...... 104 ENDOPHILIN FAMILY OF PROTEINS...... 104 Identification of Endophilin A1/SH3p4/SH3GL2 ...... 104 Endophilin A and B Family Proteins Bind and Tubulate Membranes ...... 105 Endophilin A Family Proteins Function in Receptor-Mediated Endocytosis ...... 106 OTHER BAR DOMAIN PROTEINS ...... 107 Rich-1/Nadrin 1...... 107 SNX1...... 107 Other BAR Domain Proteins ...... 107 CRYSTAL STRUCTURE OF THE AMPHIPHYSIN BAR DOMAIN ...... 107 Structure of the BAR Domain...... 107 Coupling of the BAR Domain with Additional Lipid-Binding Sequences ...... 109 Membrane Binding and GTPase Binding of BAR Domains: Independent or Associated?...... 109 BAR Domain and IMD Structure ...... 109 MEMBRANE TUBULATION AND BAR DOMAIN PROTEIN FUNCTION IN VIVO ...... 109 CONCLUSIONS AND FUTURE PERSPECTIVES...... 112 ACKNOWLEDGMENTS ...... 112 REFERENCES ...... 112

INTRODUCTION conserved features of the BAR domain suggest there may exist an underlying common molecular mechanism that is provided The BAR domain proteins form a rapidly expanding protein by the BAR domain and that has been adapted for use in these family defined by the presence of a homologous ␣-helical do- different physiological processes. main of 250 to 280 amino acids named after the founding Insight into what this common molecular mechanism may be members of this family: Bin1, Amphiphysin, and Rvs167 (BAR) has come from recent key discoveries. The first discovery was (278, 298). BAR domain proteins have been implicated in an that BAR domains bind liposomes in vitro and convert low- extraordinary diversity of cellular processes, including fission curvature spheres to high-curvature tubules (316). The second of synaptic vesicles, cell polarity, endocytosis, regulation of the discovery was that the BAR domain is itself curved (banana- actin cytoskeleton, transcriptional repression, cell-cell fusion, shaped) and is therefore exquisitely designed both to sense signal transduction, apoptosis, secretory vesicle fusion, excita- membrane curvature and to actively influence membrane cur- tion-contraction coupling, learning/memory, tissue differentia- vature (237). Furthermore, subsequent structure comparisons tion, ion flux across membranes, and tumor suppression. The revealed that many proteins that bind GTPases do so through 40 REN ET AL. MICROBIOL.MOL.BIOL.REV. a similar protein fold (109). These discoveries have created is vital for yeast cells to survive under adverse environmental great excitement, since it may now be possible to explain the conditions. diverse cellular roles of BAR domain proteins in terms of To identify genes important for cell cycle arrest in response sensing membrane curvature, binding GTPases, and actively to starvation, Michel Aigle and colleagues screened a random molding cellular membranes. collection of UV-induced mutants for those specifically defec- BAR domain proteins are encoded by most, if not all, eu- tive in their ability to adapt to starvation conditions. Mutants karyotic genomes. They are found in organisms from lower that were fully viable in the presence of nutrients but that lost unicellular eukaryotes such as budding yeast and fission yeast viability more rapidly than wild-type cells when starved of Downloaded from to insects, plants, and vertebrates. The phylogeny of BAR glucose (carbon), ammonium (nitrogen), or sulfate (sulfur) domain proteins has been the subject of an excellent recent were retained. Two mutants found to exhibit a Reduced Via- review (109). Database homology searches with known BAR bility upon Starvation (RvsϪ) phenotype and to carry mu- domains have not yet identified an obvious BAR domain en- tations in distinct genes were named rvs161 and rvs167 the coded by the genome of any prokaryote. Sequence homology corresponding wild-type genes are RVS161 and RVS167, re- between known BAR domains is, however, relatively modest. spectively). In cell cycle terminology, the proteins Rvs161p and The sequence features that confer folding of a polypeptide into Rvs167p are negative cell cycle regulators that link nutrient a structure that can bind and bend membranes are still not fully availability to cell cycle progression (12, 42). understood. It is possible that proteins with similar properties http://mmbr.asm.org/ exist also in prokaryotes but that low sequence homology has made them difficult to recognize. The apparent origin of BAR Rvs161p and Rvs167p Proteins and Their domain proteins in eukaryotes suggests a function(s) unique to Common BAR Domain eukaryotes. Some of the roles of BAR domain proteins (e.g., in Comparison of Saccharomyces cerevisiae Rvs161p and membrane traffic) would be relevant only in eukaryotic cells. Rvs167p revealed strong amino acid sequence homology be- Hence, the evolution of BAR domain proteins may parallel the tween the two proteins (Fig. 1). The homology extended over evolution of cellular compartmentation and complexity. the total length (265 residues) of Rvs161p and over the N- This article will first review the physiological roles and in- terminal 281 residues of Rvs167p. Over this region the two teractions of the BAR domain proteins, focusing primarily on proteins exhibit 27% amino acid sequence identity and 52% on October 26, 2015 by University of Queensland Library the amphiphysins (including the Bin1-3 proteins) and the en- amino acid sequence similarity. The homologous domain was dophilins. Other BAR domain proteins, such as the sorting predicted to contain two regions with predominantly ␣-helical nexins (SNXs), will be discussed only briefly. In particular, this structure (12, 47, 220, 298). The N-terminal homologous do- article focuses on the roles of BAR domain proteins in yeasts main was initially named the Rvs domain (299). Because ver- (budding yeast and fission yeast) and in mammals (with occa- tebrate amphiphysin 1 and Bin1 (see below) also feature a sional reference to their roles in flies). This article then reviews homologous domain, Sakamuro et al. renamed the Rvs domain the three-dimensional structure of the BAR domain. The ar- the Bin1/Amphiphysin/Rvs167 (BAR) domain (Fig. 1) (278). ticle concludes with a somewhat speculative attempt to explain The designation BAR domain is now commonly used to de- the known physiological roles of BAR domain proteins in scribe this homologous domain in both yeast and nonyeast yeasts and in mammals in terms of binding membranes and proteins. More recently, this domain has been subdivided into generating membrane tubules in vivo. a short N-terminal amphipathic ␣-helix (ϳ40 to 45 residues) and the BAR domain itself, as some BAR domain proteins BUDDING YEAST Rvs PROTEINS lack the N-terminal amphipathic ␣-helix (Fig. 1). Rvs167p has two additional C-terminal domains that are not Nutrient Availability and the Control of Cell Proliferation present in Rvs161p. Following the BAR domain is the glycine-, Yeasts as free-living unicellular organisms are at the mercy of proline-, and alanine-rich (GPA-rich) region which contains no their environment and experience a variety of stresses, including charged amino acids, includes a hydrophobic sequence, and is changes in osmolarity, temperature, pH, redox potential, and not predicted to adopt a defined secondary structure (Fig. 1). nutrient deprivation. Therefore, yeast cells must continuously The GPA-rich region is followed by a Src Homology 3 (SH3) monitor their extracellular environment and respond to changes domain (Fig. 1) (12). SH3 domains are short, 50- to 70-amino- in ways that ensure survival. A common response of yeast cells acid modules found in a diverse range of signal transduction when confronted by stress is to arrest progression through the cell and actin cytoskeletal proteins (234). They mediate protein- division cycle and become quiescent. For example, when starved protein interactions by binding primarily (but not exclusively) of essential nutrients, dividing yeast cells arrest uniformly in late to short linear proline-rich target motifs (261, 382).

G1 of the cell division cycle prior to initiation of chromosome replication and formation of a daughter cell (bud) (185, 186). This BAR Domains of Rvs161p and Rvs167p Assemble arrest point in yeast is analogous to the “restriction point” in into Heterodimers mammalian cells. Following cell division cycle arrest the starved yeast cells enter a quiescent phase (G0) (135). Yeast cells quies- Secondary-structure prediction identified two regions of ␣ cent in G0 survive for long periods and are resistant to stress high -helical potential in both the Rvs161p BAR domain (247). However, if yeast cells cannot enter this quiescent state, (residues 22 to 65 and 127 to 183) and the Rvs167p BAR their futile attempts to transit the cell division cycle without an domain (residues 30 to 57 and 144 to 191) (47, 220). Moreover, adequate nutrient supply have dire consequences and the cells the COILS algorithm revealed that the BAR domain ␣-helices perish. Hence, the ability to stop dividing and become quiescent are amphipathic and have a propensity to form coiled-coil VOL. 70, 2006 BAR DOMAIN PROTEINS 41 Downloaded from http://mmbr.asm.org/ on October 26, 2015 by University of Queensland Library

FIG. 1. Domain structure of yeast and human amphiphysin family proteins. Schematic representation of the domain organization of the budding yeast amphiphysin family proteins Rvs161p and Rvs167p, the fission yeast amphiphysin family proteins Hob1 and Hob3, and the human amphiphysin family proteins amphiphysin 1, Bin1/amphiphysin 2 (several tissue-specific and ubiquitous splice variants), Bin2, and Bin3, and endophilins A1, A2, A3, B1, and B2. 42 REN ET AL. MICROBIOL.MOL.BIOL.REV. structures that may mediate interaction of the BAR domain cycle arrest was found for rvs167 cells. Hence, under starvation with other coiled-coil proteins. Two-hybrid analysis revealed conditions rvs mutants still attempt to transit the cell cycle and that Rvs167p and Rvs161p interact and that the interaction is likely die as a consequence of their failure to arrest (12, 42). mediated by their BAR domains. In this study, neither Small GTPases of the Ras family regulate response to star- Rvs161p nor Rvs167p interacted with itself, suggesting that the vation in yeast. They do this by regulating adenylyl cyclase Rvs proteins form obligate heterodimers. The in vivo associa- activity and production of the second messenger cyclic AMP tion of Rvs161p with Rvs167p was confirmed by coimmuno- (282). Yeast mutants with constitutively activated Ras exhibit precipitation of Rvs161p with Rvs167p from yeast lysates in

reduced viability upon starvation and an inability to mount a Downloaded from vitro (220). normal physiological response to nutrient deprivation similar Subsequently, no fewer than eight other two-hybrid studies to rvs mutants. Wild-type yeast cells accumulate the storage have analyzed interactions between Rvs161p and Rvs167p carbohydrate glycogen when starved (177). One characteristic and each protein with itself (18, 38, 59, 84a, 97, 136, 180, of the constitutively activated ras phenotype is the failure to 339). All these studies confirmed interaction between Rvs161p accumulate glycogen upon glucose starvation (185, 325, 329). and Rvs167p via the respective BAR domains. In contrast to Can rvs mutant cells accumulate glycogen when nutrients be- the initial study (220), two of these later studies found that come limiting, e.g., when cultures reach stationary phase and Rvs167p also forms homodimers via the BAR domain (38,

the cells cease growth? When they reach stationary phase both http://mmbr.asm.org/ 180). Furthermore, two-hybrid interaction of Rvs167p with rvs161 and rvs167 mutant cells do indeed accumulate glycogen itself persists when RVS161 is deleted, indicating that Rvs167p- normally (12, 42, 53). This suggests the defect in rvs mutant Rvs167p interaction is either direct or mediated by a protein cells is distinct from that in hyperactivated ras mutants or other than Rvs161p (180). Two later studies, however, con- mutants in which adenylyl cyclase activity is constitutive. The cluded that Rvs167p does not form homodimers in vegetatively molecular basis of the defect in response to starvation in rvs growing cells (i.e., it only forms heterodimers) (84a, 97). In- mutants is not yet known. terestingly, in cells lacking Rvs161p the steady-state level of Growth sensitivity to salt (especially Na؉). rvs mutants are Rvs167p is considerably reduced due to accelerated proteolysis not only sensitive to starvation. They are also more sensitive to and vice versa (180). These results represent strong evidence a range of other stresses that include the presence of high that in vivo Rvs161p and Rvs167p form heterodimers and on October 26, 2015 by University of Queensland Library concentrations of various salts (e.g., NaCl, KCl, MgCl , and function in concert. 2 Na2SO4) in the medium. This sensitivity is to salt rather than to osmotic strength, because rvs mutants are not sensitive to sor- Loss of Rvs161p and Loss of Rvs167p Cause Similar and bitol at high osmotic strength. Moreover, there is selectivity for Diverse Spectra of Phenotypes certain cations. For example, rvs161 mutants exhibit morpho- logical defects at fivefold lower concentrations of NaCl (0.25 The yeast rvs161 and rvs167 mutations are highly pleiotropic M) than of KCl (1.2 M) while wild-type cells are unaffected by and the reported phenotypes are summarized below. For many of either salt. Furthermore, rvs161 mutants are no more sensitive the phenotypes tested, rvs161 and rvs167 mutants have identical to LiCl than wild-type cells (although both are sensitive) (42). phenotypes and the rvs161 rvs167 double mutant is not more ϩ severely affected than either single mutant. A recent genome- In this review the rvs salt sensitivity will be referred to as Na sensitivity to reflect this. The molecular basis of the enhanced wide analysis of genetic interactions of Rvs161p and Rvs167p ϩ revealed that loss of each protein is lethal in pairwise combination sensitivity of rvs mutants to Na is not known. with loss of the same set of 49 other yeast proteins. This indicates Growth sensitivity to cytotoxic compounds. The growth of that the roles of Rvs161p and Rvs167p in vegetative growth are both rvs161 and rvs167 mutants is hypersensitive to the pres- identical. The rvs161-1 and rvs167-1 mutations identified in the ence of various cytotoxic compounds, including 3-amino-1,2,4- initial mutant screen appear to be complete loss-of-function al- triazole (3-AT) and canavanine (12, 42). 3-AT is a histidine leles because deletion or disruption of the RVS161 gene (rvs161⌬) biosynthesis inhibitor that induces an artificial starvation re- or the RVS167 gene (rvs167⌬) in almost every case reveals a sponse when added to cells. Canavanine is an arginine analog phenotype identical to that with the original mutant allele (12, 42, that can be incorporated into proteins in place of arginine and 53, 84a). produce nonfunctional proteins. Yeast mutants defective in -Reduced viability upon starvation (Rvs؊). As introduced ubiquitin-mediated proteolysis are especially sensitive to cana above, rvs161 and rvs167 mutants were originally isolated based vanine and overexpression of ubiquitin confers canavanine re- on their inability to maintain viability when starved of a source sistance (32, 115). In addition, rvs161 cells have been reported of nitrogen, carbon, or sulfur. For example, rvs161 cells grown to be hypersensitive to the effects of sinefungin, which com- to stationary phase in glucose-limited or nitrogen-limited me- pletely blocks growth of rvs161 cells at 0.1 ␮M(rvs167 cells dium and then left in the depleted medium for 60 h exhibit only were not tested) (42). Sinefungin is a toxic analog of S-adenosyl- 20 and 65% viability, respectively. In contrast, wild-type cells methionine that inhibits methylation reactions. In yeast, sine- subjected in the same way to either glucose or nitrogen star- fungin has recently been shown to inhibit methylation of vation maintain close to 100% viability. A similar loss of via- guanosine to form m7GpppN, which is used to cap the 5Ј end bility under starvation conditions has been shown for rvs167 of mRNAs. Capping of mRNAs is in turn important for cells. In nitrogen- or glucose-limited minimal medium, 90 to mRNA stability and efficient translation (36). Whether hyper- 100% of wild-type cells arrest without buds after 50 h. In sensitivity to these compounds is specific or reflects a general contrast, under each condition ϳ20% of rvs161 cells still ex- hypersensitivity of rvs161 and rvs167 cells to all cytotoxic com- hibit a bud. Again, a similar defect in starvation-induced cell pounds is not known. VOL. 70, 2006 BAR DOMAIN PROTEINS 43

Growth sensitivity to elevated temperature. The standard loid cells do not sporulate upon nitrogen starvation (53). Sub- growth temperature for budding yeast is 30°C. Elevated tem- sequently, Colwill et al. showed that rvs167/rvs167 homozygous perature presents a stress to yeast cells and while wild-type diploid cells, although not entirely compromised for sporula- cells continue to grow well at 37°C they stop dividing at 42°C. tion, form spores with only 10% the frequency of wild-type Yeast mutants with constitutively activated ras mutations ex- diploids (38). rvs/rvs homozygous diploids may not sporulate hibit not only reduced viability upon starvation, but also an because they are unable to utilize the nonfermentable carbon inability to grow at normal temperature after a short heat source provided or because they lose viability upon nitrogen shock at 55°C. In contrast, rvs161 and rvs167 mutant cells both starvation. The sporulation defect in these mutants could be Downloaded from survive heat shock at 55°C. In two reports, rvs161 cells did not indirect. However, genetic interaction studies have suggested exhibit defects in growth at low (e.g., 15°C) or normal (28°C) that the Rvs proteins play a direct role in sporulation (53) temperature and neither rvs161 nor rvs167 cells exhibit defects (discussed below). The molecular mechanisms that underlie in growth at elevated (e.g., 36°C) temperature (12, 42). this defect have yet to be elucidated. The ability of rvs161 and rvs167 cells to grow at elevated Heterogeneous cell size and morphology upon starvation or temperature may be dependent on genetic background or the exposure to Na؉. Although rvs161 and rvs167 mutant cells accumulation of second-site suppressor mutations. In a some- exhibit the relatively uniform size and ellipsoid shape of wild- what different genetic background rvs161 and rvs167 cells did type cells under optimal growth conditions, when starved or not grow at 37°C (215). Exposure to elevated temperature exposed to Naϩ (or the cytotoxic compounds referred to http://mmbr.asm.org/ causes induction of heat shock proteins (HSP) which aid in the above) cultures of both rvs161 and rvs167 cells accumulate a refolding of damaged proteins. However, HSP induction upon high proportion of cells that are either grossly enlarged with exposure to elevated temperature was found to be normal in swollen vacuoles or abnormally tiny. Some mother cells also rvs167 mutant cells (12). rvs mutants are not sensitive to all have multiple buds. Each bud has a nucleus but is apparently environmental stresses, e.g., rvs161 and wild-type yeast cells unable to complete cytokinesis and separate from the mother cell are equally resistant to extremes of pH (12, 42). (12, 42). In wild-type yeast, there is a minimum cell size that is Growth on nonfermentable carbon sources. Both rvs161 and required at Start for commitment to a new cell division cycle rvs167 mutants grow well on a range of fermentable carbon (137). The rvs mutations appear to compromise this cell cycle sources (e.g., glucose, galactose, mannose, and sucrose), but regulation with under-sized mother cells continuing to divide. on October 26, 2015 by University of Queensland Library are unable to grow on nonfermentable carbon sources (e.g., Loss of actin cytoskeleton polarization to sites of polarized glycerol, lactate, or acetate) (12, 42). An inability to utilize growth. Yeast cells possess an actin cytoskeleton comprising nonfermentable carbon sources suggests a possible mitochon- filamentous actin (F-actin) and a diverse set of actin-associated drial defect. These carbon sources are metabolized in mito- proteins (71, 213, 371). Yeasts express orthologs of many, chondria (370). rvs161 mutant cells possess a normal spectrum although not all, vertebrate actin-associated proteins, including of mitochondrial cytochromes, but respire threefold more the Arp2/3 complex and its activators, both conventional (fil- slowly than wild-type cells as measured by oxygen consump- ament forming) and unconventional myosins, profilin, tropo- tion. It is not clear, however, whether this level of respiratory myosin, fimbrin, capping protein, and cofilin. When yeast cells defect is sufficient to fully account for the observed failure to are stained with fluorophore-conjugated phalloidin (an F- utilize nonfermentable carbon sources. Defects in glycerol uti- actin-specific reagent) several distinct structures are visible. lization are a common phenotype of mutations that affect the Actin patches are small highly motile spots located at or near actin cytoskeleton, but the molecular basis for the defect is still the cortex. Actin cables are long thick fibers comprising bun- unclear. Mitochondrial defects alone cannot account for the dled actin filaments that extend through the cortical cytoplasm. RvsϪ phenotype since deletions in the mitochondrial genome Although individual actin patches and actin cables turn over, ([rhoϪ]) that abolish growth on nonfermentable carbon patches and cables are visible throughout the cell cycle. sources do not give rise to the other RvsϪ phenotypes such as Dividing yeast cells display a third F-actin structure known as reduced viability upon starvation (12, 42). the contractile actomyosin ring (16, 178). This ring structure lo- Meiosis and sporulation. In the laboratory, budding yeast calizes at the neck between the mother cell and bud. Contraction cells can be maintained as either stable diploids or stable of the actomyosin ring occurs during cytokinesis and is accompa- haploids. Diploid yeast cells can be induced to undergo sporu- nied by deposition of new cell wall material (septum). Septum lation (meiosis) by nitrogen starvation on nonfermentable car- deposition eventually separates the cytoplasm of the mother cell bon sources. Sporulation has been well studied in yeast be- and bud and cleavage of the septum allows cell separation. cause it represents a cellular differentiation process that can be During the cell division cycle, the distribution of actin patches studied in budding yeast and that may have molecular mech- and actin cables changes. Immediately prior to bud emergence, anisms in common with more complex cellular differentiation cables align with their tips focused at the nascent bud site and pathways used in vertebrate development. Sporulation involves patches concentrate at this site. When the bud starts to emerge, switching off the expression of blocks of genes required for the cables align with their tips inside the growing bud and the vegetative growth and initiating a developmental program in patches localize at the bud tip. In G2 the actin patches remain which spore-specific genes are expressed in a highly regulated polarized to the bud but switch from a polarized to an isotropic pattern. In addition to meiosis, sporulation also involves de distribution within the bud. After the switch the bud expands novo bilayer membrane biogenesis to form the plasma mem- laterally as well as at the tip. During mitosis actin is recruited to brane of the resultant haploid cell and also cell wall biogenesis the actomyosin ring, cables become randomly oriented, and to form the spore wall. patches distribute randomly throughout the mother cell and bud. Desfarge et al. reported that homozygous rvs161/rvs161 dip- Finally, upon exit from mitosis the patches in the mother cell and 44 REN ET AL. MICROBIOL.MOL.BIOL.REV. bud repolarize to either side of the bud neck and cables in the results in random budding (65). mother cell and bud realign with their tips focused to the bud It is possible that spatial landmarks at the cell poles are not neck. At this stage of the cell cycle the actomyosin ring contracts formed properly in rvs161 or rvs167 mutant cells. This would to a dot, septum is deposited and cleaved, and the cells divide. mean that when the next division ensues the cell can no longer The newly divided mother and daughter cells transiently retain “remember” where the previous bud formed to place the next polarized actin patches and cables, and then polarity is lost until bud site appropriately. Alternatively, spatial landmarks may be a new bud site is selected (16, 178, 371). formed in rvs mutant cells, but not recognized or interpreted When rvs161 and rvs167 mutant cells were first stained to correctly. Subsequently, a close correlation has been estab- Downloaded from visualize F-actin it was apparent that the actin cytoskeleton in lished between those mutations that affect actin patch polar- these cells was abnormal. Actin patches were not as polarized ization and those that abolish bipolar bud site selection (375). at nascent bud sites and growing buds. Actin cables were more Defective fluid-phase and receptor-mediated endocytosis. difficult to visualize and appeared less well aligned than those Endocytosis is the process by which cells internalize plasma in wild-type cells. When rvs161 or rvs167 mutant cells were membrane material as well as ligands, particles, and fluid from exposed to high levels of salt (e.g., NaCl) or starved, the loss of the extracellular environment via invagination of the plasma actin cytoskeleton polarity became complete. Under both membrane and formation of endocytic vesicles. To identify stress conditions actin cables disappeared completely and actin genes important for endocytosis, Riezman and colleagues con- patches depolarized fully (12, 298). ducted a screen for yeast mutants unable to internalize plasma http://mmbr.asm.org/ Delocalized cell wall chitin deposition. In wild-type yeast membrane receptor-ligand complexes. A bank of random yeast cells the cell wall polysaccharide chitin is specifically depos- mutants was screened using an assay for receptor-mediated ited at the site of bud formation during both bud emergence internalization of ␣-factor, the peptide ligand secreted by hap- and bud growth, and to seal the scar after separation of the loid yeasts of the ␣ mating type. One of several mutants re- mother cell and the bud. In rvs167 mutant cells, however, covered from this screen carried a mutation in ENDocytosis- staining of chitin with Calcofluor and microscopic examina- defective 6 (END6). Isolation and characterization of the tion reveal that chitin is not restricted to sites of active bud END6 gene showed that END6 is identical to RVS161 and that formation and at scars from previous budding events. There the end6 mutation affects polarization of the actin cytoskeleton appears to be an accumulation of chitin distributed evenly as observed for rvs161. A previously isolated rvs167⌬ mutant on October 26, 2015 by University of Queensland Library throughout the cell wall and this defect is exacerbated by the was also blocked in internalization of ␣-factor. Hence, efficient presence of a sublethal concentration of Naϩ in the growth receptor-mediated endocytosis in budding yeast requires both medium (12, 84a). Rvs161p and Rvs167p. Uptake and accumulation in the vacu- Loss of bipolar bud site selection. Yeast cells do not bud ole of the membrane-impermeant fluid-phase endocytic dye randomly. After a bud has emerged from a mother cell, a lucifer yellow are also blocked in both rvs161 and rvs167 mu- scar made of chitin is left on the surface of both mother (bud tant cells (215). scar) and daughter (birth scar). These scars are permanent Inefficient cell-cell fusion during mating. Haploid budding and can be visualized by use of the fluoresent stain Cal- yeast cells exist in two mating types, a (MATa) and ␣ (MAT␣), cofluor. Bud scars on mother cells accurately record all sites that have an identical physical appearance. MATa cells se- where previous buds have emerged. In wild-type haploid crete a peptide pheromone known as a-factor that binds to cells, the bud sites are visualized as a cluster at one pole of the a-factor receptor (a G-protein-coupled receptor called the cell in an “axial” pattern, which is generated when bud- Ste3p) expressed only by MAT␣ cells. Conversely, MAT␣ ding occurs in new mothers at a site adjacent to the birth cells secrete a peptide pheromone known as ␣-factor that scar and in old mothers at a site adjacent to the previous bud binds to the ␣-factor receptor (a G-protein-coupled recep- scar. In diploid cells the bud sites form clusters at both poles tor called Ste2p) expressed only by MATa cells. Binding of of the cell in a “bipolar” pattern, which is generated when each pheromone to its receptor activates a mitogen-acti- budding occurs in new mothers at a site on the opposite pole vated protein kinase signal transduction cascade that has to the birth scar and in old mothers at a site that alternates two main readouts. First, cell cycle progression is arrested in between opposite cell poles. G1 to ensure that each mating cell has a normal 1C DNA Mutant studies have shown that the requirements for axial content in preparation for nuclear fusion. Second, expression and bipolar bud site selection are distinct. Mutations in the of various proteins specifically required for mating is induced, genes BUD1, BUD2, and BUD5 cause haploid cells to bud at e.g., proteins that promote cell-cell adhesion and fusion (41, random sites, while mutations in BUD3 and BUD4 cause hap- 365). loid cells to bud in a bipolar pattern (29, 30). So some genes During mating, each haploid cell chooses a mate from the are required for axial budding and some for all patterns of surrounding cells based on the level of pheromone each cell budding. Interestingly, RVS161 and RVS167 were among the secretes. Yeast cells are able to detect gradients of phero- first genes discovered that are specifically required for bipolar mone with extraordinary sensitivity and form a tube-like bud site selection. Haploid rvs161 and rvs167 mutant cells bud projection at their surface known as a mating projection or in an axial pattern like wild-type cells. However, rvs161/rvs161 shmoo. This projection grows up the gradient to the source and rvs167/rvs167 homozygous mutant diploids bud at random of pheromone, i.e., towards the cell that produces the high- sites (12, 53, 65, 298). The defect is not specific to diploid cells, est level. When mating yeast cells are treated with synthetic but to the process of bipolar bud site selection. When haploid mating pheromone they mate at random with cells of the cells are induced to undergo bipolar budding (e.g., by mutation opposite mating type even if those cells produce no phero- of BUD3 or BUD4) additional mutation of RVS161 or RVS167 mone. This behavior, known as “default” mating, is approx- VOL. 70, 2006 BAR DOMAIN PROTEINS 45 imately 10-fold less efficient than normal pheromone gradi- In a later study, various Rvs167p constructs were tested for ent mating. Supersensitive 2 (sst2) mutants are hypersensitive their ability to restore growth in the presence of Naϩ, bipolar to endogenous pheromone and are consequently unable to bud site selection, fluid-phase endocytosis, and sporulation to accurately detect subtle pheromone gradients. They mate by cells lacking Rvs167p (rvs167⌬). In addition to a truncated default with random partners even in the presence of endog- Rvs167p construct lacking an SH3 domain (BAR-GPA), this enous pheromone gradients (56). study also employed a full-length Rvs167p construct featuring In a search for mutants defective in default mating, candi- a P473L substitution in the SH3 domain that abolishes binding date mutants were tested for their ability to mate with a to proline-rich motifs. Unlike the BAR-GPA fragment, the Downloaded from “pheromoneless” partner in the presence of synthetic phero- full-length P473L mutant construct was expressed at the same mone. One of the mutants tested was rvs161⌬. Earlier work steady-state level as full-length Rvs167p and fully rescued all had shown rvs161⌬ sst2 double mutants (which mate by default rvs167⌬ defects. This study concluded that Rvs167p is fully due to sst2) mate extremely inefficiently. Indeed, the rvs161⌬ functional without a functional SH3 domain if expressed at mutant was specifically defective in default mating since normal levels. rvs161⌬ and wild-type cells mated with efficiency similar to that Do the GPA-rich and SH3 domains contribute to Rvs167p of wild-type cells of the opposite mating type (56). function? The GPA-rich and SH3 fragment (GPA-SH3) did In an independent study, a screen was performed for mu- not rescue growth in the presence of Naϩ, bipolar bud site tants defective in mating (using endogenous pheromone) but selection, or endocytosis, but unexpectedly was able to fully http://mmbr.asm.org/ only when both parents carry the mutation (i.e., bilateral mat- rescue the sporulation defect of rvs167⌬. The SH3 domain ing defect) (158). One such mutation blocked mating at the alone was also able to rescue the sporulation defect, showing stage of cell-cell fusion and was named fusion 7 (fus7) (91). that the SH3 domain was sufficient for this function (38). Further work showed fus7 is a mutation in RVS161 (21). The The BAR domains of Rvs161p and Rvs167p are highly mating defect of rvs161 cells was not apparent in the study of homologous. Sivadon et al. tested the effect of swapping the Dorer et al. (56) because only mating of rvs161⌬ cells to wild- BAR domains. The Rvs167p GPA-rich and SH3 domains were type cells was tested (which only detects unilateral mating fused to Rvs161p to create an artificial Rvs167p-like protein defects). Interestingly, rvs167 cells do not exhibit a mating (Rvs161p-GPA-SH3). The fragment of Rvs167p comprising defect in either test, one case where rvs161 and rvs167 cells the BAR domain only was used as the corresponding Rvs161p- on October 26, 2015 by University of Queensland Library differ in phenotype. like protein (Rvs167p-BAR). Rvs161p-GPA-SH3 retained full ability to rescue the growth, actin cytoskeleton, and bud site ⌬ Rvs161p and Rvs167p Structure-Function Relationships selection defects of rvs161 , but did not rescue these defects in rvs167 cells. Rvs167p-BAR was partially functional in rescuing The diverse range of phenotypes displayed by rvs161 and the phenotypes of rvs167 cells, but it lacked the ability to rescue rvs167 mutants suggested that Rvs161p and Rvs167p are mul- the defects of rvs161 cells. Nor could Rvs167p-BAR and tifunctional proteins. In the case of Rvs167p, which comprises Rvs161p-GPA-SH3 when coexpressed rescue the growth, actin three distinct domains, each domain may confer different bio- cytoskeleton, or bud site selection defects in rvs161 rvs167 logical activities. Two studies investigated structure-function double mutant cells (299). Clearly, the two BAR domains have relationships in Rvs167p (38, 299). In the first study the ability important differences. of various truncated Rvs167p constructs, each lacking one or Enforced overexpression of Rvs167p from a strong pro- more Rvs167p domains, to complement the phenotypes of moter induces lethality at normal growth temperature. In con- rvs167 was examined, i.e., viability upon glucose starvation, trast, overexpression of the GPA-SH3 construct does not affect growth in the presence of Naϩ or 3-AT, and utilization of a growth. However, the SH3 domain is important for the over- nonfermentable carbon source. The BAR domain alone was expression phenotype of full-length Rvs167p. Overexpression sufficient to rescue each phenotype tested, although less effi- of the full-length P473L mutant has milder deleterious effects ciently than full-length Rvs167p. In contrast, a fragment com- than wild-type Rvs167p and these only become apparent at prising only the GPA-rich and SH3 domains did not rescue any elevated temperature (38). The mechanism by which Rvs167p of the phenotypes (299). overexpression inhibits growth is not known. Consistent with The ability of the various truncated Rvs167p fragments to an important function for the Rvs167p SH3 domain, a recent rescue the loss of actin cables and depolarization of actin study identified conditions under which the Rvs167p SH3 do- patches in rvs167 mutant cells exposed to a sublethal concen- main becomes important for growth (84a). tration of Naϩ was also examined. In general, the same Rvs167p fragments that were functional in growth assays were Rvs Proteins Localize to the Cortical Actin Cytoskeleton also able to rescue the actin cytoskeleton defects, i.e., the BAR domain alone was able to correct the actin cytoskeleton de- Subcellular localization of Rvs161p and Rvs167p. Where do fects, but rescue by the BAR domain was not as complete as Rvs161p and Rvs167p localize in the cell? Rvs161p and rescue by full-lengthRvs167p. The fragment comprising the Rvs167p interact with each other and there is compelling GPA-rich and SH3 domains was not able to correct the actin evidence that these two proteins function as a heterodimer cytoskeleton defects. Expression of the BAR domain alone in most, if not all, of their various cellular functions (18, 38, restored bipolar bud site selection in 75% of rvs167/rvs167 97, 180, 220). Indeed, as discussed above, in the absence of homozygous diploid cells but full rescue of bipolar budding either Rvs protein the other is unstable and is degraded required full-lengthRvs167p. The GPA-rich and SH3 domain (180). Paradoxically, Rvs161p and Rvs167p appear to ex- fragment did not rescue bipolar bud site selection (299). hibit somewhat distinct subcellular localizations in live cells 46 REN ET AL. MICROBIOL.MOL.BIOL.REV. Downloaded from http://mmbr.asm.org/ on October 26, 2015 by University of Queensland Library

FIG. 2. Subcellular localization of Rvs161p in growing and mating yeast cells. A. Shown are budding yeast cells that express fusion proteins that comprise full-length Rvs161p fused to the fluorescent reporter Aequoria victoria green fluorescent protein (GFP). The same fields of cells were viewed by fluorescence optics (left) to visualize GFP and by differential interference contrast optics (right) to visualize the cell profiles. In panels A to J, M, and N the cells express Rvs161p-GFP with the reporter fused at the Rvs161pC terminus. In panels K and L the cells express VOL. 70, 2006 BAR DOMAIN PROTEINS 47 when fused to the green fluorescent protein (GFP) for vi- gives one confidence that this fusion protein displays an au- sualization by fluorescent imaging. It should be noted, how- thentic Rvs167p subcellular localization. Double labeling of ever, that fusion to GFP alters the subcellular distribution of F-actin and Rvs167p-GFP shows that most patches that con- some proteins, so differences may reflect differences in the tain F-actin also contain Rvs167p-GFP and vice versa, al- ability of Rvs161p and Rvs167p to tolerate fusion to GFP though the relative signal intensity of the two proteins varies without perturbing their subcellular localization rather than from one patch to another (11). This difference may reflect the differences between Rvs161p and Rvs167p in subcellular age of the patch, since a number of studies have shown that the localization per se. protein composition of an individual patch can vary during its Downloaded from In growing cells a fusion protein comprising Rvs161p and lifetime (132, 139, 143, 144). GFP (Rvs161p-GFP) was reported to exhibit a predominantly The distribution of Rvs167p-GFP patches changes as cells diffuse cytoplasmic distribution in cells without buds (Fig. 2A). progress through the cell cycle exactly as for actin patches. In cells with small buds Rvs161p-GFP localized to patches at Rvs167p-GFP specifically associates with actin patches and has the bud neck, but these patches were no longer apparent in not been observed to localize to other F-actin structures such large buds. Expression of Rvs161p-GFP fully rescued the cell as actin cables. Localization of Rvs167p is predominantly in- fusion defect of an rvs161⌬ mutant (see below), suggesting this dependent of F-actin, as disassembly of all F-actin in cells by fusion protein is functional (at least in cell fusion). This study treatment with the actin polymerization inhibitor latrunculin A focused on the role of Rvs161p in cell fusion during mating, so does not abolish Rvs167p localization to cortical patches (11). http://mmbr.asm.org/ the ability of this fusion protein to rescue other rvs161⌬ defects A recent study reported, however, that loss of F-actin appears was not tested. Both N- and C-terminal Rvs161p fusions to to cause a partial redistribution of Rvs167p from cortical GFP exhibited the same subcellular distribution (21). A more patches, as there is an apparent increase in diffuse cytoplasmic recent report demonstrated localization of a similar Rvs161p- Rvs167p after latrunculin A treatment (144). GFP fusion protein to numerous small cortical patches (Fig. Interactions between Rvs167p and actin patch proteins. 2B). These patches exhibited polarization to nascent bud sites How does Rvs167p localize to actin patches? As loss of and small buds during bud emergence and to the bud neck in Rvs161p does not perturb actin patch localization of Rvs167p, dividing cells. However, this particular Rvs161p-GFP fusion it seems unlikely that interaction with Rvs161p mediates actin protein was not able to rescue the defects of an rvs161⌬ mu- patch localization of Rvs167p (11). on October 26, 2015 by University of Queensland Library tant, so it is nonfunctional (10). Does Rvs167p associate with actin? In one study a two- A very recent report described the subcellular localization hybrid screen was performed to identify actin-interacting of an apparently fully functional Rvs161p-GFP fusion pro- proteins and a fragment encoding the Rvs167p SH3 domain tein (144). This Rvs161p-GFP fusion localized to cortical was recovered. Hence, there is some evidence Rvs167p as- actin patches but also exhibited a strong diffuse cytoplasmic sociates with actin. The Rvs167p-actin interaction requires distribution. Cortical actin patches are larger and less nu- the Rvs167p SH3 domain since deletion of the SH3 domain- merous than the patches to which the nonfunctional encoding sequence abolishes the two-hybrid interaction (4). Rvs161p-GFP fusion used by Balguerie et al. localizes (10). SH3 domains interact with short proline-rich motifs (e.g., Localization of Rvs161p-GFP to cortical actin patches is PXXP) (331). The actin sequence contains only a single F-actin dependent, since localization to patches is abolished motif (PMNP) that might mediate SH3 domain interaction. by depolymerization of all F-actin by treatment with latrun- Subsequently, it was proposed that the Rvs167p-actin inter- culin A (144). Confirmation of the subcellular distribution action is indirect (38, 176, 180). A third protein might bind of native (untagged) Rvs161p by immunofluorescence stain- actin directly and contain proline-rich motifs that then bind ing with Rvs161p-specific antisera has not yet been reported. the Rvs167p SH3 domain. This may be due to the extreme sensitivity of Rvs161p an- Systematic “charged-to-alanine” scanning mutagenesis of the tigenicity and/or subcellular localization to chemical fixation yeast ACTin gene (ACT1) has been performed (363). Charged (our unpublished data). residues predicted to be surface exposed and potentially able to An Rvs167p-GFP fusion protein localizes in vegetatively engage in interactions with other proteins were replaced singly growing cells to cortical actin patches (Fig. 3). The Rvs167p- and in clusters with uncharged alanine residues and a collection GFP fusion protein was able to fully rescue the various defects of 35 act1 mutations was generated. These act1 mutants vary in of rvs167⌬ cells and hence appears to be functional (11). This phenotype (e.g., some are inviable and others are without obvious

GFP-Rvs161p with the reporter fused at the Rvs161pN terminus. Panels A to F show vegetatively growing cells, panels G and H show cells arrested in G1 and forming mating projections after pheromone treatment, and panels I to N show mating cells. In pheromone-treated cells Rvs161p-GFP concentrates at the tip of the mating projection (shmoo) and in mating cells Rvs161p-GFP and GFP-Rvs161p concentrate at the site of cell-cell fusion. In vegetatively growing cells Rvs161p-GFP exhibits a diffuse cytoplasmic distribution and is excluded from the vacuole (large indentation apparent in the cell profiles). (Reproduced from reference 21 by copyright permission of The Rockefeller University Press.) B. Shown are vegetatively growing budding yeast cells that express a fusion protein (Rvs161p-GFP) that comprises full-length Rvs161p with GFP fused at the Rvs161pC terminus. The same fields of cells were viewed by fluorescence optics (center and right columns) to visualize GFP and by differential interference contrast (left column) to visualize the cell profiles. For each field of cells two focal planes were viewed by fluorescence optics: an equatorial view (center column) and a top view (right column). A, an unbudded cell; B, a cell at an early stage of bud emergence; C, cells undergoing cell division (cytokinesis). This Rvs161p-GFP fusion protein localizes to numerous small cortical patches. (Reprinted with permission from reference 10.) 48 REN ET AL. MICROBIOL.MOL.BIOL.REV. Downloaded from http://mmbr.asm.org/ on October 26, 2015 by University of Queensland Library

FIG. 3. Subcellular localization of Rvs167p in growing and mating yeast cells. Shown are budding yeast cells that express a fusion protein (Rvs167p-GFP) that comprises full-length Rvs167p with GFP inserted between the BAR and GPA-rich domains of Rvs167p. The same fields of cells were viewed by fluorescence optics (right) to visualize GFP and by differential interference contrast (left) to visualize the cell profiles. A, unbudded cell; B, cell at an early stage of bud emergence; C, cell with small bud; D, cell with large bud; E, cells undergoing division (cytokinesis);

F, cells arrested in G1 and forming a mating projection (shmoo) after pheromone treatment (only one cell expressing Rvs167p-GFP is depicted in this panel). In vegetatively growing cells Rvs167p-GFP localizes to large cortical patches that polarize to nascent bud sites, small buds, and the bud neck during cell division (cytokinesis). In mating cells Rvs167p-GFP concentrates at the tip of the mating projection (shmoo). (Reprinted from reference 11 with permission of the Company of Biologists Ltd.) phenotype). Two-hybrid analysis using these 35 mutant act1 genes Abp1p contributes to Rvs167p function in vivo, loss of Abp1p and a panel of actin-interacting proteins revealed that different results in defects in sporulation, growth in the presence of Naϩ, interactions are affected by different act1 mutations. Unexpectedly, and growth on nonfermentable carbon sources, similar to but all those act1 mutations that abolish interaction with Rvs167p also weaker than defects associated with loss of Rvs167p. Abp1p and abolish interaction with another actin-binding protein, profilin. Rvs167p also exhibit functional redundancy with the same set of Conversely, the act1 mutations that do not affect interaction with proteins (i.e., with Sla1p, Sla2p, and Sac6p). Moreover, loss of profilin do not affect interaction with Rvs167p (4). So either Rvs167p reduces the deleterious effect of Abp1p overexpression Rvs167p binds to actin in the same way as profilin (i.e., makes the on growth and cell morphology, which suggests that the deleteri- same contacts) or Rvs167p binds to profilin and the Rvs167p-actin ous effect of Abp1p overexpression requires an intact Rvs167p- interaction is indirect and mediated by profilin. To date, interaction Abp1p complex (38, 176). A recent study identified conditions between Rvs167p and profilin has not been demonstrated. under which Rvs167p-Abp1p interaction becomes important for One actin patch component that has been proposed to mediate vegetative growth (84a). actin-Rvs167p interaction is Abp1p, encoded by the Actin Binding Despite the appeal of models in which Abp1p recruits Protein 1 (ABP1) gene. Abp1p was first identified biochemically Rvs167p into actin patches via interaction with the Rvs167p as a protein that binds with high affinity to F-actin in vitro (60). SH3 domain, experimental evidence suggests other mecha- The Rvs167p SH3 domain recognizes Abp1p in a Far-Western nisms must also exist: deletion of ABP1 (abp1⌬) does not affect blot and in two-hybrid screens. Consistent with a model that Rvs167p localization to actin patches (11), two-hybrid interac- VOL. 70, 2006 BAR DOMAIN PROTEINS 49 tion between the Rvs167p SH3 domain and actin does not cally affects the growth of vrp1⌬ mutants (las17⌬ was not require Abp1p (180), the act1 two-hybrid data suggest that tested). Interestingly, additional loss of Rvs161p (rvs161⌬) did Rvs167p interaction with actin is mediated by a protein that not have any obvious effect on the growth of vrp1⌬ cells binds monomeric actin (e.g., profilin) while Abp1p is an F- (las17⌬ was not tested) (275). Hence, Vrp1p may be required actin binding protein, localization of Rvs167p to cortical for all cellular functions of Rvs161p, which would be consistent patches does not require F-actin, the SH3 domain of Rvs167p with a possible role in Rvs161p localization to actin patches. alone does not associate with actin patches in vivo, and the Interestingly, in the fission yeast Schizosaccharomyces pombe SH3 domain of Rvs167p is not essential for localization to actin there is recent evidence that the Las17p ortholog (Wsp1p) is Downloaded from patches (11, 38). Abp1p may function in Rvs167p localization required for actin patch localization of the Rvs167p ortholog to actin patches, but its role may be redundant with that of (Hob1p) (131). other proteins (see Table 1 for a list of other cortical actin An interesting possibility that has not yet been explored is patch proteins known to interact with Rvs167p). that Vrp1p and/or Las17p mediates the two-hybrid interaction There is evidence that the BAR and SH3 domains of of Rvs167p with actin (4). Rvs167p interacts with both Las17p Rvs167p both contribute to its actin patch localization. For and Vrp1p via its C-terminal SH3 domain (38, 181, 331; our example, a fragment comprising only the Rvs167p BAR do- unpublished data). Both Vrp1p and Las17p possess actin main localizes poorly to cortical actin patches, while the monomer binding WASP Homology 2 (WH2) domains. Al- Rvs167p SH3 domain shows no localization to actin patches. In though the contacts that WH2 domains make with actin are http://mmbr.asm.org/ contrast, the full-length protein containing the BAR and SH3 not yet known, they may resemble the contacts that profilin domains localizes efficiently to actin patches (11). Perhaps makes with actin, since both WH2 domains and profilin bind other actin patch components mediate recruitment of Rvs167p specifically to actin monomers. If correct, this would account to actin patches via interactions with the BAR domain. Possi- for the observation that actin mutations that perturb profilin ble candidates are the products of three genes, synthetic lethal binding also perturb Rvs167p binding (4). with abp1 1 (SLA1) and SLA2 and suppressor of ras Val19 2 (SRV2), which localize to actin patches and have functions Roles of Rvs161p and Rvs167p in Endocytosis redundant with that of Abp1p (126, 176). Sla1p associates with Rvs167p in vivo and directly binds via Rvs161p and Rvs167p in receptor-mediated internalization on October 26, 2015 by University of Queensland Library multiple domains to recombinant Rvs167p in vitro, but the of ␣-factor. In vertebrates, receptor endocytosis occurs at spe- Rvs167p domain that mediates these interactions is not yet cific sites on the plasma membrane that bear a cytoplasmic known (59, 123, 307). Interestingly, Sla2p possesses a coiled- protein coat. This coat comprises the coat protein clathrin coil domain (coil1) that has been reported to interact with the (comprising heavy and light chains) and the clathrin coat as- Rvs167p BAR domain. Furthermore, the Sla2p coil1 domain sembly factor known as the clathrin Associated (or Assembly) has a function important for growth, actin patch polarization, Protein 2 (AP-2) adaptor. The assembly of clathrin coats on and endocytosis. Intriguingly, however, this function of coil1 is the cytoplasmic face of the plasma membrane by the AP-2 fully redundant with a function mediated by Abp1p (or Srv2p), adaptor generates clathrin-coated endocytic pits. Binding of i.e., the Sla2p coil1 mutant phenotypes only become apparent the AP-2 adaptor to the cytoplasmic tails of plasma membrane in cells that lack Abp1p (or Srv2p) (364). Interactions of the receptors to be endocytosed concentrates the receptors in the Rvs167p BAR domain with Sla2p coil1 and the Rvs167p SH3 forming clathrin-coated pits. The conversion of an invaginated domain with Abp1p may work in concert to facilitate Rvs167p clathrin-coated pit into a clathrin-coated vesicle is mediated by localization to actin patches. the GTPase dynamin. Dynamin assembles into rings at the The Rvs167p SH3 domain may also recruit actin monomers neck of invaginated clathrin-coated pits, and hydrolysis of GTP to patches via interaction with other actin patch components is accompanied by fission of the neck of the pit and the release such as Local Anaesthetic Sensitive 17 (Las17p, also known of a clathrin-coated vesicle into the cytoplasm. In vertebrates, as Bee1p) and Very Rich in Proline 1 (Vrp1p, also known as the Rvs161p and Rvs167p homolog amphiphysin has been End5p or verprolin) (Table 1). Las17p is the yeast ortholog of shown to function in endocytosis via recruitment of dynamin human Wiskott-Aldrich syndrome protein (WASp) that is mu- from the cytoplasm to plasma membrane clathrin-coated pits. tated in patients with the inherited immunodeficiency (8). Both Could rvs161 and rvs167 mutants perturb the recruitment of a WASp in mammals and Las17p in yeast are implicated in de yeast dynamin to clathrin-coated pits? novo assembly of actin monomers into actin filaments by a In yeast cells, in contrast to vertebrates, receptor endocy- highly conserved seven-subunit complex known as the Arp2/3p tosis is predominantly independent of clathrin. Yeast cells in complex (119, 372). Vrp1p is the yeast ortholog of human which the single genes encoding clathrin heavy chain or WASp-interacting protein (WIP), a protein identified by two- clathrin light chain are deleted continue to internalize ␣-fac- hybrid screens with WASp. Vrp1p and Las17p in yeast interact tor, although at a reduced rate (37, 235). A temperature- analogous to WASp and WIP in mammals (181, 219). sensitive clathrin mutant (chc1-ts) was constructed that can Consistent with a possible role for Vrp1p in recruitment of be inactivated by a shift to elevated temperature. This mu- Rvs167p to actin patches, loss of VRP1 (vrp1⌬) displays exactly tant was used to examine the acute effects of clathrin inac- the same pattern of negative genetic interactions with the yeast tivation on ␣-factor endocytosis. Even in this study, only a myosins (Myo1p-Myo5p) as loss of Rvs167p (see below) (19, slower kinetics of ␣-factor internalization and not a complete 275). However, Rvs167p must retain significant function in block was observed (318). Furthermore, single and combined cells lacking Vrp1p even if its localization to actin patches is deletions of the various genes encoding the subunits of the puta- affected, since additional loss of Rvs167p (rvs167⌬) dramati- tive yeast equivalent of the AP-2 adaptor and related proteins 50 REN ET AL. MICROBIOL.MOL.BIOL.REV.

TABLE 1. Comprehensive list of Rvs161p and Rvs167p interactors in budding yeasta

Type(s) of interaction Group Gene Encoded protein Experimental evidence (reference[s])b

Cytoskeleton ABP1c Actin Binding Protein involved in actin cytoskeleton 7 (38, 59, 160, 176) Two-hybrid, peptide scanning and polarity assembly and cell polarity establishmentc

ACF2/PCA1 Actin Assembly Complementing Factor, intracellular 7 (59, 97, 160, 331) Two-hybrid, peptide scanning ␤-1,3-endoglucanase Downloaded from ACF4 Actin Assembly Complementing Factor, molecular 7 (59, 160, 331) Two-hybrid, peptide scanning function unknown

ACT1 Yeast ACTin 1 (19, 215), 7 (4, 19, 59, 180) Two-hybrid, synthetic lethal, nonallelic noncomplementation

APP1 Actin Patch Protein, molecular function unknown 1 (18), 7 (18, 59, 160, 331) Two-hybrid, peptide scanning

ARP2 Actin-Related Protein, essential component of the 1 (332), 7 (160, 332) Two-hybrid, synthetic lethal (rvs161), Arp2/3 complex synthetic sick (rvs167) http://mmbr.asm.org/ BBC1 Protein possibly involved in assembly of actin patches; 1 (332), 7 (332) Synthetic sick interacts with actin assembly factor Las17p and with the SH3 domains of type 1 myosins Myo3p and Myo5p

BSP1 Binding protein of Synaptojanin Polyphosphoinositide 7 (59, 160, 331) Two-hybrid, peptide scanning phosphatase domain, adaptor that links synaptojanins Inp52p (Sjl2p) and Inp53p (Sjl3p) to the cortical actin cytoskeleton

CAP1 CAPping protein, binds to barbed ends of actin 1 (332), 7 (332) Synthetic lethal filaments, preventing further polymerization on October 26, 2015 by University of Queensland Library

CAP2 CAPping protein, binds to barbed ends of actin 1 (332), 7 (332) Synthetic lethal filaments, preventing further polymerization

CDC24 Cell Division Cycle, guanine nucleotide exchange factor 1 (136) Two-hybrid for Cdc42p

CLA4 Protein serine/threonine kinase, homologous to Ste20p 1 (332), 7 (332) Synthetic sick

END3 ENDocytosis defect, EH domain-containing protein 1 (332), 7 (85, 332) Synthetic sick involved in endocytosis, actin cytoskeletal organization, and cell wall morphogenesis

EXO70 EXOcyst, 70-kDa subunit of exocyst complex 7 (18) Two-hybrid

GIM3 Gene Involved in Microtubule biogenesis, subunit of the 1 (332), 7 (332) Synthetic sick heterohexameric cochaperone prefoldin complex

GIM4 Gene Involved in Microtubule biogenesis, subunit of the 1 (332), 7 (332) Synthetic sick heterohexameric cochaperone prefoldin complex

GIM5 Gene Involved in Microtubule biogenesis, subunit of the 1 (332), 7 (332) Synthetic lethal heterohexameric cochaperone prefoldin complex

LAS17 Yeast mutant that is Local Anesthetic Sensitive, yeast 1 (123), 7 (18, 38, 59, 97, 123, Two-hybrid, peptide scanning, WASp 136, 160, 181, 331) ELISA, complex isolation, and mass spectrometry

MYO1 Yeast type II MYOsin 1 (19), 7 (19) Synthetic lethal

MYO2 Yeast type V MYOsin 1 (19), 7 (19) Synthetic sick

MYO3 Yeast type I MYOsin 7 (331) Two-hybrid

MYO5 Yeast type I MYOsin 1 (332), 7 (331, 332) Two-hybrid, synthetic sick

PAC10 Protein required in the Absence of Cin8p, part of 1 (332), 7 (332) Synthetic lethal heteromeric cochaperone GimC/prefoldin complex, also called GIM2

RVS161 Reduced Viability upon Starvation 7 (18, 38, 59, 97, 123, 136, 180, Two-hybrid, affinity purification, 220, 339) complex isolation, and mass spectrometry Continued on facing page VOL. 70, 2006 BAR DOMAIN PROTEINS 51

TABLE 1—Continued

Type(s) of interaction Group Gene Encoded protein Experimental evidence (reference[s])b

RVS167 Reduced Viability upon Starvation 1 (18, 38, 59, 97, 123, 136, 180, Two-hybrid, affinity purification, 220, 339), 7 (18, 180) complex isolation, and mass spectrometry

SAC6 Suppressor of ACtin mutations, yeast fimbrin 1 (97), 7 (97, 176) Two-hybrid, synthetic lethal

SEC8 SECretory, essential 121-kDa subunit of the exocyst 7 (18) Two-hybrid Downloaded from complex

SLA1 Synthetic Lethal with ABP1, protein required for 1 (123, 332), 7 (59, 85, 97, 123, Two-hybrid, Co-IP, affinity assembly of the cortical actin cytoskeleton 176, 307, 332) purification, synthetic lethal, complex isolation, and mass spectrometry

SLA2 Synthetic Lethal with ABP1, protein required for 1 (97), 7 (59, 97, 176, 364) Two-hybrid, synthetic lethal assembly of the cortical actin cytoskeleton

SUR7 SUppressor of Rvs167 1 (300, 381), 7 (300, 381) Genetic suppression http://mmbr.asm.org/

VRP1 Very Rich in Proline, yeast WASp-interacting protein 1 (123), 7 (123, 275, 331) Two-hybrid, synthetic sick, complex isolation, and mass spectrometry

YKE2 Yeast nuclear gene encoding a protein showing 1 (332), 7 (332) Synthetic sick homology to mouse KE2 and containing a putative leucine zipper motif

YSC84 SH3 domain-containing protein with function unknown, 7 (59) Two-hybrid also called LSB4 (LaS17p-Binding)

Transcription BCK1 Bypass of CKinase, mitogen-activated protein kinase 1 (332), 7 (332) Synthetic sick and kinase kinase on October 26, 2015 by University of Queensland Library signaling

CTI6 Cyc8-Tup1-Interacting protein, transcription factor 7 (18) Two-hybrid binding protein

EAP1 EIF4E-Associated Protein, implicated in TOR signaling 1 (332), 7 (332) Synthetic lethal (rvs161), synthetic sick (rvs167)

GTS1 Glycine Threonine Serine repeat protein, transcription 7 (160) Peptide scanning activator activity

RIM101 Regulator of IME2, transcriptional repressor involved in 7 (160) Peptide scanning the response to pH

RVB2 RuVB-like, involved in transcription regulation 7 (123) Affinity purification

SAP30 SIT4 protein phosphatase-Associated Protein, subunit of 1 (332), 7 (332) Synthetic sick a histone deacetylase complex

SDS3 Suppressor of Defective Silencing, involved in 1 (332), 7 (332) Synthetic lethal transcriptional silencing and required for sporulation

SIN3 Switch INdependent, DNA binding subunit of 1 (332), 7 (332) Synthetic sick Sin3p-Rpd3p histone deacetylase complex

SLN1 Synthetic Lethal of N-end rule, histidine kinase 7 (18) Two-hybrid osmosensor that regulates a MAP kinase cascade

SLT2 Suppression at Low Temperature, serine/threonine MAP 1 (20, 332), 7 (20, 332) Synthetic lethal kinase

SUM1 SUppresor of Mar1-1, nuclear protein involved in 1 (332), 7 (332) Synthetic sick silencing

SWI4 SWItching deficient, involved in cell cycle-dependent 1 (332), 7 (332) Synthetic lethal gene expression

TFC6 Transcription Factor C, subunit of RNA polymerase III 7 (18) Two-hybrid transcription initiation factor

TY1B Transposon Ty1 protein B 7 (18) Two-hybrid

Cell cycle BBP1 Bfr1p Binding Protein, required for the spindle pole 7 (160) Peptide scanning body (SPB) duplication Continued on following page 52 REN ET AL. MICROBIOL.MOL.BIOL.REV.

TABLE 1—Continued

Type(s) of interaction Group Gene Encoded protein Experimental evidence (reference[s])b

CYK3 CYtoKinesis, SH3 domain protein located in the mother- 1 (332), 7 (332) Synthetic sick bud neck, molecular function unknown

ESP1 Extra Spindle Pole bodies, sister chromatid separase 7 (18) Two-hybrid

PCL2 PHO85 CycLin, forms a functional kinase complex with 7 (163) Two-hybrid, direct binding, Co-IP Pho85p, activated by Swi5p Downloaded from

PCL9 PHO85 CycLin, forms a functional kinase complex with 7 (163) Two-hybrid Pho85p, activated by Swi5p

RED1 REDuctional division, involved in chromosome 1 (114) Two-hybrid segregation during the first meiotic division

Transport FEN1/ELO2, FENpropimorph resistance, fatty acid elongase, involved 1 (262) Genetic suppression ISUR5/VBM2 in sphingolipid biosynthesis

GRD19 Golgi Retention Deficient, sorting nexin required to 7 (350) Two-hybrid http://mmbr.asm.org/ maintain late-Golgi-resident enzymes

GUP1 Glycerol Uptake, plasma membrane protein with a 1 (332), 7 (332) Synthetic lethal possible role in proton symport of glycerol

GYL1 GYp-Like, putative GTPase-activating protein, stimulates 1 (18), 7 (18, 84, 317, 331) Two-hybrid, affinity purification Gyp5p GAP activity on Ypt1p

GYP5 GTpase-activating protein for Ypt Proteins, GAP for 1 (18), 7 (18, 84, 160, 317) Two-hybrid, peptide scanning, yeast rab1 affinity purification

HSE1 Has Symptoms of class E mutants, subunit of the 7 (18) Two-hybrid on October 26, 2015 by University of Queensland Library endosomal Vps27p-Hse1p complex

IVY1 Phospholipid-binding protein that Interacts with both 7 (18) Two-hybrid Ypt7p and Vps33p

KAP122 KAryoPherin, responsible for import of the 7 (18) Two-hybrid Toa1p-Toa2p complex into the nucleus

MGE1 Mitochondrial GrpE, involved in protein import into 1 (123) Affinity purification mitochondria

MNN2 MaNNosyltransferase 1 (332), 7 (332) Synthetic sick

MNN9 MaNNosyltransferase 1 (332), 7 (332) Synthetic lethal

MNN10 MaNNosyltransferase 1 (332), 7 (332) Synthetic lethal

MRS3 Mitochondrial RNA Splicing, mitochondrial iron 7 (160) Peptide scanning transporter

PEX14 PEroXisome related, peroxisomal membrane protein, a 7 (385) Peptide scanning central component of the peroxisomal protein import machinery

PHO84 PHOsphate metabolism, phosphate transporter and 7 (123) Affinity purification low-affinity manganese transporter

POR1 PORin, mitochondrial porin (voltage-dependent anion 1 (123) Affinity purification channel)

PSE1 Protein Secretion Enhancer, karyopherin/importin that 7 (18) Two-hybrid interacts with the nuclear pore complex

RSP5 Reversal of SptϪ Phenotype 5, yeast NEDD4 ubiquitin 7 (123, 307) Two-hybrid, co-immunoprecipitation, ligase affinity purification, complex isolation, and mass spectrometry

RUD3 Relieves Uso1-1 transport Defect, Golgi matrix protein 1 (332), 7 (332) Synthetic sick involved in the structural organization of the cis-Golgi

SEC21 SECretory, COP1 complex component 7 (18) Two-hybrid

SEC22 SECretory, R-SNARE protein, cycles between ER and 1 (332), 7 (332) Synthetic sick Golgi Continued on facing page VOL. 70, 2006 BAR DOMAIN PROTEINS 53

TABLE 1—Continued

Type(s) of interaction Group Gene Encoded protein Experimental evidence (reference[s])b

SEC27 SECretory, essential beta-coat protein of the COPI 1 (123) Affinity purification coatomer

SRP54 Signal Recognition Particle subunit 7 (18) Two-hybrid

SPF1 Sensitivity to Pichia farinosa killer toxin, P-type ATPase, 1 (332), 7 (332) Synthetic lethal ion transporter of the ER membrane involved in ER Downloaded from function and Ca2ϩ homeostasis

SUR4/ELO3, SUppressor of rvs161 and rvs167 mutations, elongase III 1 (53) Genetic suppression /VBM1 synthesizes 20–26-carbon fatty acids from C18-CoA primers

SXM1 Suppressor of mRNA eXport Mutant, nuclear transport 1 (18) Two-hybrid factor (karyopherin)

VPS21/YPT51 Vacuolar Protein Sorting, Rab5-like GTPase 1 (97, 332), 7 (97, 296, 332) Synthetic sick http://mmbr.asm.org/ Metabolism ARG1 ARGinine requiring, acetylglutamate synthase 7 (123) Affinity purification and biogenesis

BNI4 Bud Neck Involved, required for localization of chitin 1 (332), 7 (332) Synthetic sick (rvs161), synthetic synthase III to the bud neck lethal (rvs167)

CCW12 Covalently linked Cell Wall protein, expression 1 (332), 7 (332) Synthetic lethal down-regulated by alpha factor

CHS3 CHitin Synthase-related, chitin synthase III 1 (332), 7 (332) Synthetic sick on October 26, 2015 by University of Queensland Library CHS5 CHitin Synthase-related, molecular function unknown 1 (332), 7 (332) Synthetic sick

CHS6 CHitin Synthase-related, molecular function unknown 1 (332), 7 (332) Synthetic sick

CHS7 CHitin Synthase-related, molecular function unknown 1 (332), 7 (332) Synthetic sick

COR1 CORe protein of QH2 cytochrome c reductase 7 (123) Affinity purification

CSF1 Cold Sensitive for Fermentation 1 (332), 7 (332) Synthetic lethal

DEP1 Disability in regulation of Expression of genes involved 1 (332), 7 (332) Synthetic sick in Phospholipid biosynthesis

DOA1 Degradation Of Alpha2, regulatory component of the 1 (332), 7 (332) Synthetic sick proteasome pathway

ECM29 ExtraCellular Mutant, major component of the 7 (123) Affinity purification proteasome

FKS1 FK506 Sensitivity, catalytic subunit of 1,3-␤-D-glucan 7 (332) Synthetic lethal synthase

FUS2 cell FUSion, required for the alignment of parental 1 (18, 21, 97, 136, 221) Two-hybrid, Co-IP nuclei before nuclear fusion during mating

GDH3 Glutamate DeHydrogenase 3 7 (97) Two-hybrid

HOC1 Homologous to OCh1, ␣-1,6-mannosyltransferase 1 (332), 7 (332) Synthetic sick (rvs161) synthetic involved in cell wall mannan biosynthesis lethal (rvs167)

HOM6 HOMoserine requiring, homoserine dehydrogenase 7 (123) Affinity purification

HSP90 Heat Shock Protein, cytoplasmic chaperone 1 (389), 7 (332) Synthetic lethal

IDH1 Isocitrate DeHydrogenase 7 (123) Affinity purification

ILV5 IsoLeucine-plus-Valine requiring, acetohydroxy acid 7 (123) Affinity purification reductoisomerase

IPT1 Inositol PhosphoTransferase 1 (10) Genetic suppression

LPD1 LiPoamide Dehydrogenase 7 (123) Affinity purification

KGD2 ␣-KetoGlutarate Dehydrogenase 7 (123) Affinity purification Continued on following page 54 REN ET AL. MICROBIOL.MOL.BIOL.REV.

TABLE 1—Continued

Type(s) of interaction Group Gene Encoded protein Experimental evidence (reference[s])b

KRE1 Killer toxin REsistant, cell wall glycoprotein involved in 1 (332), 7 (332) Synthetic sick ␤-glucan assembly

KRE6 Killer toxin REsistant, cell wall glycoprotein involved in 7 (20) Synthetic sick ␤-glucan assembly

LYS12 LYSine requiring 7 (123) Affinity purification Downloaded from

MET18 METhionine requiring 7 (123) Affinity purification

MUS81 MMS and UV Sensitive, Helix-hairpin-helix protein 7 (18) Two-hybrid

PDI1 Protein Disulfide Isomerase 1 (114) Two-hybrid

PHO23 PHOsphate metabolism, component of the Rpd3 histone 1 (332), 7 (332) Synthetic sick deacetylase complex

PMI40 PhosphoMannose Isomerase 7 (123) Affinity purification http://mmbr.asm.org/

PRE10 PRoteinase yscE, 20S proteasome ␣-type subunit 7 (123) Affinity purification

PSK1 Pas domain-containing Serine/threonine protein Kinase 7 (160) Peptide scanning

SKT5 Activator of Chs3p (chitin synthase III) 1 (332), 7 (332) Synthetic sick (rvs161), synthetic lethal (rvs167)

SMI1 Suppressor of MAR Inhibitor, involved in (1,3)-␤-glucan 7 (332) Synthetic sick synthesis

SML1 Suppressor of Mec Lethality, ribonucleotide reductase 7 (160) Peptide scanning on October 26, 2015 by University of Queensland Library inhibitor

SRV2 Suppressor of RasVal19, adenylyl cyclase-associated 7 (59, 176) Two-hybrid, synthetic sick protein

SUR1 SUppressor of rvs161 and rvs167 mutations, catalytic 1 (53) Genetic suppression subunit of a mannosylinositol phosphorylceramide synthase

SUR2 SUppressor of rvs161 and rvs167 mutations, sphingosine 1 (53) Genetic suppression hydroxylase

TPS1 Trehalose-6-Phosphate Synthase, regulator of glucose 1 (332), 7 (332) Synthetic sick influx into the cell and into glycolytic pathway

TRP5 TRyPtophan requiring 7 (123) Affinity purification

UBI4 UBIquitin 7 (123) Affinity purification

UBP7 UBIquitin-specific Protease 7 (123) Affinity purification

URA7 URAcil requiring 7 (18, 123) Affinity purification

RNA and CDC33 Cell Division Cycle, cytoplasmic mRNA cap binding 7 (123) Affinity purification translational protein regulation

DED1 DEaD-box protein, ATP-dependent DEAD (Asp-Glu- 7 (123) Affinity purification Ala-Asp)-box RNA helicase

KRS1 Lysyl (K)tRNA Synthetase 7 (123) Affinity purification

LCP5 Lethal with Conditional Pap1, essential protein involved 7 (18) Two-hybrid in maturation of 18S rRNA

MSU1 Degradosome associates with the ribosome and mediates 7 (18, 331) Two-hybrid turnover of RNAs

PRP40 Pre-mRNA Processing, U1 snRNP protein involved in 7 (114) Two-hybrid splicing

RPC40 RNA Polymerase C subunit 7 (123) Affinity purification

SES1 SEryl-tRNA Synthetase 7 (123) Affinity purification Continued on facing page VOL. 70, 2006 BAR DOMAIN PROTEINS 55

TABLE 1—Continued

Type(s) of interaction Group Gene Encoded protein Experimental evidence (reference[s])b

SLF1 Associates with translating ribosomes 7 (18) Two-hybrid

SNP1 U1snRNP 70K protein homolog 7 (160) Peptide scanning

Others and CUE5 Coupling of Ubiquitin conjugation to ER degradation 7 (160) Peptide scanning unknown Downloaded from

DEM1 Differentially Expressed in Malignancies, molecular 7 (160) Peptide scanning function unknown

ECM30 ExtraCellular Mutant, function unknown 7 (18) Two-hybrid

GFD2 Great for Full DEAD box protein activity, molecular 7 (18) Two-hybrid function unknown

HUA1 Unknown function 7 (59, 136) Two-hybrid

HUA2 Unknown function 7 (59) Two-hybrid http://mmbr.asm.org/

RXT2 Molecular function unknown 1 (332), 7 (332) Synthetic sick

SDS23 Homolog of S. pombe SDS23, implicated in 1 (114) Two-hybrid APC/cyclosome regulation

YBP2 Yap1 Bing Protein, protein with a role in resistance to 7 (18, 136) Two-hybrid, affinity purification oxidative stress

YBR108W Hypothetical ORF 1 (339), 7 (59, 97, 136, 160) Two-hybrid, peptide scanning

YBR239C Hypothetical ORF 7 (160) Peptide scanning on October 26, 2015 by University of Queensland Library

YBR255W Hypothetical ORF 1 (332), 7(332) Synthetic sick

YDL156W Hypothetical ORF 7 (160) Two-hybrid, peptide scanning

YDR154C Hypothetical ORF 1 (136) Two-hybrid

YDR239C Hypothetical ORF 7 (160) Peptide scanning

YGL085W Hypothetical ORF 7 (160) Peptide scanning

YLR111W Hypothetical ORF 1 (332), 7 (332) Synthetic sick

YLR243W Hypothetical ORF 7 (123) Affinity purification

YNL086W Hypothetical ORF 7 (59) Two-hybrid

YNL152W Hypothetical ORF 7 (160) Peptide scanning

YPL009C Hypothetical ORF 7 (160) Peptide scanning

YPR091C Hypothetical ORF 7 (160) Peptide scanning

a This table contains a comprehensive listing of all reported genetic and physical interactions. Many interactions require further confirmation. Note that some interactions of Rvs proteins are with proteins whose predominant subcellular localization seems incompatible with interaction (e.g., Pdi1p is a protein of the ER lumen). The existence of additional (perhaps minor) cytoplasmic pools of these proteins cannot be formally excluded and so these putative interactions are included here. Abbreviations: ELISA, enzyme-linked immunosorbent assay; Co-Ip, coimmunoprecipitation; MAP, mitogen-activated protein; SPB, spindle pole body; R-SNARE, R-type soluble N-ethylmaleimide-sensitive fusion protein attachment protein receptor; CoA, coenzyme A; ORF, open reading frame; APC, anaphase promoting complex; MMS, methyl methanesulfonate. This table does not include all of the synthetic lethal interactions recently reported in reference 84a. b 1, interaction with Rvs161; 7, interaction with Rvs167. c Information on gene/gene product nomenclature and function was obtained from the Saccharomyces Genome Database (R. Balakrishnan, K. R. Christie, M.C. Costanzo, K. Dolinski, S. S. Dwight, S. R. Engel, D. G. Fisk, J. E. Hirschman, E. L. Hong, R. Nash, R. Oughtred, M. Skrzypek, C. L. Theesfeld, G. Binkley, C. Lane, M. Schroeder, A. Sethuraman, S. Dong, S. Weng, S. Miyasato, R. Andrada, D. Botstein, and J. M. Cherry [http://www.yeastgenome.org/ {November 2005}]). failed to block the formation of clathrin-coated vesicles and clathrin heavy chain does associate with the yeast epsins had less impact on ␣-factor endocytosis even than deletion of (Ent1p and Ent2p) (362) and AP180 adaptors (Yap1801 and clathrin heavy or light chain. Clathrin does not even appear to Yap1802) (361). The Ent1p and Ent2p adaptors are required associate with the AP-2 adaptor in yeast (130, 377). for endocytosis in yeast (362), however, Yap1801 and Yap1802 Other adaptors are known to bind clathrin heavy chain and are not required individually or collectively (361). Yap1801 to function in clathrin-mediated endocytosis in vertebrates. and Yap1802 are also not required for clathrin coat assembly Mammalian epsin (57, 58) and AP180 (2, 216) are both known or clathrin-coated vesicle formation in yeast (130). to bind clathrin heavy chain, promote assembly of clathrin For some time it was not clear whether clathrin-coated pits coats, and function in clathrin-dependent endocytosis. In yeast, exist on the yeast plasma membrane as they do in vertebrates. 56 REN ET AL. MICROBIOL.MOL.BIOL.REV.

Two very recent studies, discussed below, were the first to to form cortical patches that are initially stationary and do not demonstrate that a pool of clathrin exists on (or near) the yet contain F-actin. Next, Abp1p, the Arp2/3p complex itself, plasma membrane in yeast cells and colocalizes with cortical and actin are recruited to the patch. The Arp2/3p complex is actin patches (144, 223). responsible for de novo nucleation of actin filaments. Las17p, Is dynamin required for receptor endocytosis in yeast? Yeast Pan1p, and Abp1p are believed to promote actin filament cells possess three dynamin-like proteins (Vps1p, Dnm1p, and assembly by interacting with and stimulating the Arp2/3p com- Mgm1p). Vps1p localizes to the Golgi apparatus and/or per- plex. Following recruitment of actin and the Arp2/3p complex oxisomes (124, 274). Deletion of the VPS1 gene that encodes an actin cloud is formed via Arp2/3p-dependent de novo actin Downloaded from Vps1p (vps1⌬) does not abolish endocytic internalization (226). filament assembly. Finally, type I myosins are transiently re- A recent study showed that deletion or mutation of Vps1p cruited to the stationary patch. Type I myosins are a class of perturbs actin patch polarization and abolishes endocytic in- actin-dependent motor that also have the potential to interact ternalization, but only at elevated temperature (383). Dnm1p with and activate the Arp2/3p complex and have been pro- localizes to the surface of mitochondria (86, 166). Deletion of posed to function in endocytic vesicle fission. Las17p, Pan1p, the DNM1 gene that encodes Dnm1p did not affect the kinetics of and type I myosins remain in a stationary patch on the mem- ␣-factor internalization, although subsequent trafficking of ␣-fac- brane, while the cortical patch initiates slow movement for a tor to the vacuole was delayed (92). Mgm1p localizes to the distance of ϳ200 nm perpendicular to the plane of the mem- intramembrane space of mitochondria and therefore a role in brane. At first End3p, Pan1p, Sla1p, and Sla2p move with the http://mmbr.asm.org/ endocytosis seems unlikely (although it has not been directly actin patch but then they gradually dissociate. In contrast, tested) (373). Dnm1p and Mgm1p regulate mitochondrial mor- Abp1p, Arp2/3p, and actin move slowly into the cell with the phology (17, 86, 229, 289). Interestingly, none of the yeast dy- actin patch and remain associated as the actin patch undergoes namin-like proteins have been shown to interact physically with a subsequent fast long-range movement (132, 139, 143, 144). clathrin, AP-2 adaptor, or either Rvs protein. Important questions remain concerning the role of Arp2/3p- In yeast, internalization of receptors by endocytosis is de- activating proteins such as Las17p, Pan1p, Abp1p, and type pendent on prior covalent attachment of the protein ubiquitin I myosins in endocytosis. In vitro evidence has suggested that to the receptor cytoplasmic tail (117). Ubiquitin is attached to the yeast Arp2/3p complex is inactive unless activated by specific lysine residues in the receptor tail by the ubiquitin these proteins (372). This is consistent with the requirement on October 26, 2015 by University of Queensland Library protein ligase Rsp5p (63, 64). Ubiquitin-dependent endocyto- of WASp (mammalian ortholog of Las17p) for mammalian sis requires the attachment of only a single ubiquitin molecule Arp2/3 activity in actin filament nucleation. A recent study, (monoubiquitination) (327). This is in contrast to targeting of however, showed that the dependence of Arp2/3p activity on cytoplasmic proteins for degradation by the 26S proteasome, activating proteins in vitro is affected by the source of actin. which requires attachment of chains containing multiple ubiq- The original assays that showed the Las17p dependence of uitin molecules (polyubiquitination). The requirement for mono- Arp2/3p activity used mammalian muscle actin. When yeast ubiquitination for receptor endocytosis can be bypassed by the actin was used instead there was strong Arp2/3p-dependent use of recombinant receptors in which the C terminus of the actin polymerization even in the absence of Las17p (although receptor tail is fused to ubiquitin (327). Are rvs161 and rvs167 the addition of Las17p further enhanced Arp2/3p activity). mutants defective in receptor-mediated endocytosis of ␣-factor This study concluded that Arp2/3p-dependent polymerization because they cannot ubiquitinate the ␣-factor receptor cyto- of yeast actin does not absolutely require an activating protein plasmic tail? This has not yet been thoroughly investigated, but (at least in vitro) (360). However, other studies found a strict a recent study indicated that the ␣-factor receptor tail is still requirement for Arp2/3p-activating proteins for Arp2/3p-de- ubiquitinated in rvs167⌬ cells (307). pendent actin assembly even when the assay was performed The polymerization of actin monomers into actin filaments using only yeast actin (101). Whether activating proteins are is a process that can produce mechanical force. It has been required for Arp2/3p activity in yeast is an important question. proposed that the requirement for actin cytoskeletal proteins If they are indeed dispensable for Arp2/3p activation in vitro for endocytic internalization in yeast may reflect a role for and this is also true in vivo, then the role of Arp2/3p complex actin polymerization in producing the force necessary to sever “activators” in actin-dependent movement and endocytosis endocytic pits and release endocytic vesicles (71, 213, 214). may need to be revisited. Live cell imaging has been used to visualize endocytosis in Interestingly, two recent studies found that there is a cortical yeast and to look for colocalization of endocytosed plasma pool of clathrin in yeast in the form of small motile patches membrane receptors with proteins important for actin poly- (144, 223). Inhibition of actin patch assembly by treatment of merization during internalization (143). Transient colocaliza- wild-type cells with the actin polymerization inhibitor latrun- tion of endocytosed receptors with proteins implicated in de culin A or in cells lacking the actin assembly protein Sla2p novo actin filament assembly occurs at sites of endocytic inter- (sla2⌬) or End3p (end3⌬) arrests cortical clathrin patch move- nalization. Moreover, correlations have been observed be- ment, extends clathrin patch lifetime, and results in a dramatic tween the movement of actin filament assembly proteins and accumulation of clathrin at the cortex (144, 223). Interestingly, endocytic vesicles (132, 139, 143, 144, 223). in cells lacking Rvs161p or Rvs167p clathrin patch lifetime is During endocytic internalization clathrin first assembles to not extended and clathrin does not accumulate at the cortex. form a cortical patch. Clathrin is soon joined in the patch by Hence, not all mutations that block endocytic internalization Las17p and a second Arp2/3p activator protein, Pan1p. Other have this effect on clathrin. It is not yet clear why loss of Sla2p actin filament assembly proteins are then recruited, including or End3p has such different effects from loss of Rvs161p or End3p, Sla1p, and Sla2p. These proteins assemble sequentially Rvs167p. VOL. 70, 2006 BAR DOMAIN PROTEINS 57

Several lines of evidence suggest that cortical clathrin in back to the plasma membrane driven by membrane tension. yeast, while not essential, is involved in endocytic internaliza- The actin assembly complex defined by Abp1p that normally tion. First, after treatment of cells with latrunculin A to induce drives the subsequent rapid long-range patch movement in clathrin accumulation at the cortex, some cortical clathrin wild-type cells falls off the invaginated membrane as the invag- patches colocalize with endocytic cargo (e.g., ␣-factor recep- ination retracts in rvs161⌬ and rvs167⌬ mutant cells (144). tors). In untreated yeast cells some cortical clathrin patches Rvs161p and Rvs167p function in endocytosis as a het- colocalize with Arp2/3p-activating proteins and are recruited erodimer. The ability of Rvs protein overexpression to perturb to patches at the cortex just prior to recruitment of Arp2/3p- receptor-mediated endocytosis of ␣-factor in wild-type cells Downloaded from activating proteins to the patch. Moreover, accumulation of requires simultaneous overexpression of both proteins. The clathrin at the cortex upon latrunculin A treatment requires level of overexpression achieved in these experiments (which either the yeast AP180 adaptor Yap1801p or Yap1802p or a used elevated gene dosage rather than a strong promoter) did different type of yeast clathrin adaptor related to vertebrate not affect the growth of wild-type cells or exacerbate the epsin (Ent1p or Ent2p). Finally, loss of clathrin heavy chain or growth defect of rvs167⌬ mutant cells, but overexpression of clathrin light chain reduces the number and the lifetime of Rvs167p in rvs161⌬ cells caused mild inhibition of growth. cortical patches containing Las17p and Sla1p, although the Rvs161p overexpression in rvs167⌬ cells or, conversely, Rvs167p movement of the remaining Las17p and Sla1p patches and overexpression in rvs161⌬ cells enhances the defect in recep- disassembly of these proteins during movement appears unaf- tor-mediated endocytosis of ␣-factor. Interestingly, this exacer- http://mmbr.asm.org/ fected. This evidence directly implicates yeast clathrin in actin- bation of the endocytic defect occurs without obvious exacerba- dependent receptor-mediated endocytosis (144, 223). tion of the actin patch polarization defect. Clearly, endocytosis is Although Rvs161p and Rvs167p have not been shown to very sensitive to the total amount and ratio of Rvs161p and activate the Arp2/3p complex for actin filament assembly in Rvs167p, but actin patch polarization and cell growth are less vitro, they do interact with several other proteins that are sensitive (180). known Arp2/3p activators (see below). This, together with Role for Rvs161p and Rvs167p in postinternalization traf- their ability to directly bind membranes (see below), makes the ficking through endosomes? The strong defect in receptor Rvs proteins ideal candidates for linking actin filament assem- internalization in rvs161 and rvs167 mutants has made it diffi- bly to membrane dynamics during endocytic internalization. Is cult to test possible defects in postinternalization traffic to the on October 26, 2015 by University of Queensland Library actin patch movement off the plasma membrane or the cou- vacuole (213, 215). Defects in postinternalization traffic are pling of this actin patch movement to internalization of endo- often (although not always) associated with defects in biosyn- cytic cargo such as plasma membrane receptors dependent on thetic traffic through endosomes to the vacuole. This is because Rvs161p and/or Rvs167p? Rvs161p and Rvs167p have recently the endocytic pathway and the vacuole biosynthetic pathway been shown to assemble into cortical patches at sites of endo- meet in the prevacuolar compartment, which is a type of en- cytosis. Rvs161p and Rvs167p are recruited after Abp1p to dosome (214). Vacuole biogenesis appears normal in rvs161 patches during the initial actin cloud formation but while and rvs167 mutant cells. Neither end6-1 (Rvs161p-R59K mu- the patches are still stationary on the plasma membrane. When tation) nor rvs167⌬ affects the appearance of the vacuole. Traf- the patches exhibit the initial slow movement over a distance fic of newly synthesized soluble vacuolar proteins from the late of ϳ200 nm perpendicular to the plane of the membrane Golgi apparatus via endosomes to the vacuole can be moni- Rvs161p and Rvs167p undergo a very rapid movement over a tored based on the kinetics of endoplasmic reticulum (ER)- distance of ϳ100 nm and then immediately dissociate. While dependent and Golgi apparatus-dependent glycosylation and treatment of cells with latrunculin A to disassemble all F-actin the proteolytic processing event that occurs upon arrival in the does not prevent Rvs167p localization to cortical patches it vacuole. These events appear normal in rvs161 and rvs167 does block movement of Rvs167p patches off the membrane mutant cells even at elevated temperature (37°C). Neither (144). rvs161 nor rvs167 cells are defective in vacuolar protein sorting The effect of rvs161⌬ and rvs167⌬ mutations on actin patch because they do not missort soluble vacuolar proteins into the dynamics during endocytosis has been examined recently and is extracellular medium (215). particularly interesting (144). In the absence of Rvs161p and It is possible that Rvs161p and Rvs167p function in the late Rvs167p the actin patch (as defined by Sla1p) still assembles endocytic pathway but redundantly with other proteins. In sup- on the plasma membrane. Upon recruitment of Abp1p the port of this possibility, combining mutations in rvs167 with patch undergoes the initial slow movement perpendicular to mutations in two genes encoding proteins known to function in the plane of the plasma membrane and Sla1p gradually disso- the late endocytic pathway, YPT51 (296) and VPS20 (97), re- ciates as in wild-type cells. However, in cells lacking Rvs161p sults in double mutant cells that either have severely reduced or Rvs167p these patches do not undergo the second phase of viability compared to either single mutant or are totally invia- rapid long-range movement. Instead, many of the actin patches ble, respectively (Table 1). YPT51 (also known as VPS21) en- appear to stop and then exhibit an abberant retrograde move- codes Ypt51p, which is one of three yeast orthologs of human ment back to the plasma membrane. Interestingly, Abp1p, early endosomal Rab5 and is important for vacuolar protein which normally is retained on the patch during rapid long- sorting and traffic of internalized ␣-factor from early endo- range movement in wild-type cells, dissociates prior to the somes to late endosomes. The interaction with Ypt51p is spe- retrograde movement in rvs161⌬ and rvs167⌬ mutants. It has cific because Rvs167p does not genetically interact with Ypt7p, been proposed that invagination proceeds in the absence of the ortholog of human late endosomal Rab7, which functions Rvs161p and Rvs167p but membrane scission fails. As a con- in trafficking of internalized ␣-factor from late endosomes to sequence the coat complex defined by Sla1p is then retracted the vacuole (297). VPS20 encodes a small coiled-coil protein 58 REN ET AL. MICROBIOL.MOL.BIOL.REV.

(Vps20p) also required for postinternalization endocytic traf- tified interactions between Rvs167p and components of the ficking and vacuolar protein sorting (376). Interestingly, vps20 exocyst complex, notably Sec8p and Exo70p (Table 1) (18). mutants exhibit severely reduced viability upon entry into sta- The exocyst complex is required for fusion of Golgi apparatus- tionary phase (i.e., nutrient starvation) (7). This further sup- derived transport vesicles with the plasma membrane. More- ports a functional link between Vps20p and the Rvs161p and over, a recent genome-wide genetic interaction study of Rvs167p proteins. Rvs161p and Rvs167p revealed that loss of either protein is lethal in pairwise combination with deletions affecting the same set of (nonessential) secretory proteins. This indepen- Roles of Rvs161p and Rvs167p in the Secretory Pathway Downloaded from dent evidence is consistent with the results of the earlier two- Rvs161p and Rvs167p are not essential for all secretory hybrid study and further supports the view that the Rvs pro- membrane traffic. A functional secretory pathway is essential for teins function in secretion (84a). viability of all cells, including budding yeast. In the absence of a Rvs161p and Rvs167p also interact with two proteins that functional secretory pathway cells are unable to deliver newly regulate Rab GTPases: Gyp5p (GYP5/YPL249c gene product) synthesized plasma membrane components to the cell surface and and Gyl1p (GYL1/YMR192w gene product) (Table 1). The SH3 the cell can neither grow in size nor divide. The viability of both domain of Rvs167p binds Gyp5p, whereas full-length Rvs161p rvs161 and rvs167 single mutants and the rvs161 rvs167 double appears to be required for its interaction with Gyp5p. The mutant shows that the secretory pathway is not blocked (at least SH3 domain of Rvs167p binds Gyl1p and phosphorylation of http://mmbr.asm.org/ under normal growth conditions) in the absence of Rvs proteins Rvs167p inhibits this interaction (18, 84, 85, 123, 317, 331). (12, 42, 53). Trafficking of newly synthesized soluble vacuolar Rab GTPases cycle between an active GTP-bound form and an proteins from the ER to the Golgi apparatus was also unaffected inactive GDP-bound form. Rab GTPase-activating proteins in rvs161 or rvs167 mutants even at elevated temperature (37°C) (Rab-GAPs) stimulate GTP hydrolysis and promote inactiva- (215). Transport of the soluble secreted enzyme invertase from tion, while Rab GDP/GTP exchange proteins (Rab-GEFs) the ER via the Golgi apparatus to the cell surface was not affected stimulate GDP/GTP exchange and promote activation (287, in either rvs161 or rvs167 mutant cells (298). Hence, Rvs proteins 309). Gyp5p has strong Rab-GAP activity in vitro and is most are not strictly essential for a functional secretory pathway. active on the Rab GTPase Ypt1p, which functions in ER-to- Rvs proteins may be required for polarized secretion during Golgi apparatus traffic (see below), although it also has signif- on October 26, 2015 by University of Queensland Library cell division. Several lines of evidence suggest the rvs mutants icant activity on the Rab GTPase Sec4p, which functions in have a secretory pathway that is not fully functional. rvs167 mu- post-Golgi apparatus traffic (49). During subcellular fraction- tant cells exhibit delocalization of cell wall chitin (12). This sug- ation Gyp5p and Gyl1p copurified with post-Golgi apparatus gests a defect in polarized secretion of chitin synthases to the bud vesicles and plasma membrane. Gyp5p was shown to physi- site. Defects in cell wall biogenesis are further supported by an cally interact with Sec4p. gyp5⌬ interacts genetically with electron microscopy study of cell wall (septum) deposition at the sec2 mutations affecting the Sec4p GEF. At low tempera- neck of dividing rvs161 and rvs167 mutant cells (20). Wild-type ture, gyp5⌬ gyl1⌬ double mutant cells exhibit a slight defect yeast cells deposit septum at the bud neck and then cleave the in growth and polarized secretion and accumulate secretory septum and divide so rapidly that electron micrographs reveal few vesicles at sites of polarized growth (33). cells in the process of forming a septum. In contrast, a much In wild-type cells, transport vesicles bud from donor com- greater proportion of rvs161 and rvs167 mutant cells are seen in partments and efficiently dock and fuse with acceptor compart- electron micrographs in the process of depositing septa, strongly, ments. Hence, transport vesicles are transient and rarely visible suggesting that deposition of septum is considerably slowed (20). in electron micrographs. In contrast, both rvs161 and rvs167 Defects in cell wall deposition may arise because of inefficient mutant cells accumulate vesicles at sites of polarized growth. delivery of cell wall biosynthetic enzymes to the division site. rvs161 mutants accumulate vesicles mainly at the bud neck Rvs proteins and post-Golgi apparatus traffic. Further evi- during cell division, however, some vesicle accumulation at the dence for a role of Rvs161p and Rvs167p in the secretory pathway nascent bud site and in the small growing buds is also observed. comes from genetic studies. rvs161 and rvs167 mutations exhibit In contrast, rvs167 mutants accumulate vesicles mainly at the negative genetic interactions with mutations affecting the uncon- nascent bud site and in small buds, however, some vesicle ventional type V myosin Myo2p (myo2). rvs161 myo2 and rvs167 accumulation at the bud neck during cell division is also ap- myo2 double mutants exhibit a lethal phenotype (19). Myo2p parent. The vesicle accumulation in both rvs mutants is appar- plays a role in motor-driven polarized transport of Golgi appara- ent even under optimal growth conditions, but becomes dra- tus-derived vesicles along actin cables to the bud (104, 138, 150, matic under conditions of stress, e.g., in the presence of sublethal 253, 273, 286). myo2 mutations are also lethal in combination with levels of Naϩ. In both mutants, exposure to Naϩ initially re- a subset of secretion (sec) mutations that affect delivery of Golgi sults in vesicle accumulate at sites of growth, but later vesicles apparatus-derived transport vesicles to the plasma membrane distribute throughout the mother cell and bud (20). myo2 mu- (104). This suggests rvs161 and rvs167 may also have defects in tants (see above) accumulate what appear to be similar vesicles this late step of the secretory pathway. Moreover, as discussed (104, 138). further below, mutations in two genes, SUR4 and SUR5/FEN1, If Rvs proteins transport vesicles to the plasma membrane, that suppress the defects of rvs161 and rvs167 mutants also sup- what cargo is transported in these vesicles? The content of the press the defects in snc1 snc2 double mutants, which are known to vesicles that accumulate in rvs mutants has not been examined. be defective in post-Golgi apparatus traffic (48). Several other actin cytoskeletal mutations also cause accumu- Further hints for a role in vesicle traffic for Rvs161p and lation of vesicles and in some cases the contents have been Rvs167p comes from large-scale two-hybrid screens that iden- analyzed. For example, the vesicles that accumulate in act1 and VOL. 70, 2006 BAR DOMAIN PROTEINS 59 sla2 mutant cells contain Ypt1p. Ypt1p is a Rab family GTPase strong mutant phenotypes. Gyl1p overexpression is slightly in- equivalent to mammalian Rab1 that localizes to the Golgi hibitory for growth of wild-type cells. Interestingly, simul- apparatus and functions in ER-Golgi apparatus traffic in yeast taneous overexpression of both Gyl1p and Gyp5p is lethal (206). It has been proposed that the Ypt1p vesicles are not specifically in sec22⌬ and rud3⌬ cells, which are partially com- transport vesicles but rather Golgi apparatus fragments. promised in ER-to-Golgi apparatus transport (although the Interestingly, there are emerging links between Rvs167p and effect of overexpression of Gyl1p and Gyp5p on ER-to-Golgi regulators of Ypt1p GTPase activity (see below) and it is possible apparatus transport has not yet been directly tested). In this that altered activity of Ypt1p may lead either to accumulation of way, overexpression of Gyp5p and Gyl1p mimics loss of Downloaded from transport vesicles or perhaps to organelle fragmentation. Insight Rvs167p (84). Another study, however, found that Gyp5p and into the nature of the cargo transported in the vesicles that accu- Gyl1p function in post-Golgi apparatus traffic (see above) (33). mulate in rvs mutant cells came from a recent genomewide mu- Perhaps Gyp5p and Gyl1p function in early steps of the secre- tant screen (250). RVS161 was one of 137 genes (including sterol tory pathway with Ypt1p and in late steps of the secretory and sphingolipid biosynthesis genes) that, when deleted, resulted pathway with Sec4p. in reduced cell surface delivery and intracellular accumulation of Neither Gyp5p nor Gyl1p localizes to cortical actin patches; a lipid raft-associated integral membrane protein marker but are instead they exhibit diffuse cytoplasmic staining (49). Some not required for secretion of the soluble protein invertase into the reports also find cortical staining at nascent bud sites, the extracellular medium. tips of small growing buds, and the bud neck during cytoki- http://mmbr.asm.org/ Rvs proteins and ER-to-Golgi apparatus traffic. Recent ev- nesis. These are also areas where Rvs167p patches localize, idence suggests that the Rvs proteins may also function earlier but the distributions only partially overlap. Interestingly, in the secretory pathway. Rvs167p exhibits two-hybrid interac- Gyp5p-GFP localization to the bud tip and bud neck is lost in tion with the COP-I component Sec21p (␥-COP) (18). COP-I both rvs161⌬/rvs161⌬ and rvs167⌬/rvs167⌬ homozygous dip- is a coat complex that functions early in the secretory pathway loid cells, suggesting both Rvs proteins may be important for and mediates the budding of transport vesicles that shuttle Gyp5p localization to these sites. Loss of Gyp1p and Gyl1p, cargo between the ER and Golgi apparatus. A genomewide either singly or in combination, did not result in Rvs pheno- mass spectrometry analysis of yeast protein complexes identi- types with the possible exception of bipolar bud site selec- fied Sec27p in a complex with Rvs161p and Rvs167p (123). tion. Furthermore, overexpression of Gyp5p and Gyl1p did on October 26, 2015 by University of Queensland Library Sec27p is the ␤-COP subunit of the COP-I coat complex. not perturb actin patch polarization or endocytosis, so the These interactions are thought-provoking but await more de- inhibitory effect seems to be restricted to ER-to-Golgi ap- tailed analysis. paratus transport (33, 84, 317). Negative genetic interactions have been observed with two A model has been proposed in which Gyp5p, Gyl1p, and mutations that affect ER-to-Golgi apparatus membrane traffic, Rvs167p function together in a complex whose role in ER-to- sec22 and rud3 (84). Sec22p is a vesicle-associated soluble Golgi apparatus transport overlaps that of Sec22p and Rud3p. N-ethylmaleimide-sensitive fusion protein attachment protein In this model Gyp5p and Gyl1p negatively regulate Rvs167p receptor that functions specifically in ER-to-Golgi apparatus function (84). Another possibility to consider is that Rvs167p transport. Rud3p (also known as Grp1p) is a matrix protein of negatively regulates Gyp5p and Gyl1p. Loss of Rvs167p would the early Golgi apparatus. Genetic interactions suggest that lead to increased Gyp5p Rab-GAP activity and a reduction in Sec22p, present on ER-derived transport vesicles, and Rud3p, the pool of active GTP-bound Ypt1p. As GTP-bound Ypt1p is present on an early Golgi apparatus compartment, function in required for vesicle fusion, it may be interesting to test if the concert to promote fusion of ER-derived transport vesicles vesicles that accumulate in rvs161 and rvs167 mutants bear with the early Golgi apparatus. sec22 and rud3 exhibit negative Ypt1p and furthermore if hyperactivation of Gyp5p Rab-GAP genetic interactions and overexpression of Rud3p rescues the activity in rvs161 and rvs167 mutant cells plays an important phenotype of some sec22 mutations (153). role in the observed vesicle accumulation. Further evidence for a role of Rvs proteins in ER-to-Golgi Rvs proteins and mating. Unlike animal cells, yeast cells are apparatus traffic came from the discovery that both Rvs161p surrounded by a thick polysaccharide cell wall. When the tips and Rvs167p interact with regulators of the ER-to-Golgi ap- of each mating projection come into contact the two mating paratus Rab GTPase Ypt1p. Ypt1p, like other Rab GTPases, cells are still separated by intervening cell wall material. Ex- cycles between an active GTP-bound form and an inactive amination of wild-type cells in the process of mating reveals GDP-bound form. Rvs161p and Rvs167p both directly bind that after mating projections come into contact a seal is made Gyp5p (which as mentioned above has strong Rab-GAP activ- at the outer perimeter of the initial contact zone. This outer ity in vitro on Ypt1p) and Gyl1p (Table 1) (18, 84, 85, 123, 317, perimeter seal remains intact, but intervening cell wall material 331). Consistent with Gyp5p’s being a Ypt1p Rab-GAP, ge- within the initial zone of contact is rapidly removed. This netic interaction data are consistent with an in vivo role for process of cell wall removal continues until the plasma mem- Gyp5p in negatively regulating Ypt1p (49). Gyl1p shares ex- brane of each mating cell is exposed. The two plasma mem- tensive amino acid sequence homology with Gyp5p and was branes then undergo a fusion event and cytoplasmic continuity initially proposed to also be a Rab-GAP for Ypt1p (317). Unex- is established. The nucleus of each mating cell migrates to the pectedly, Gyl1p lacks detectable Rab-GAP activity on Ypt1p in site of cell-cell fusion, and these haploid nuclei then also fuse, vitro. However, Gyl1p binds to Gyp5p, possibly via coiled-coil resulting in one diploid nucleus. interactions, and stimulates its GAP activity on Ypt1p (84). An inability to mate can result from defective cell cycle There is some evidence that Gyp5p and Gyl1p function in arrest, mating projection formation, cell fusion, or nuclear ER-to-Golgi apparatus traffic with Rvs167p, despite the lack of fusion. As mentioned above, rvs161 (fus7) was identified as a 60 REN ET AL. MICROBIOL.MOL.BIOL.REV.

cells, the intervening cell wall at sites of contact appears to be rapidly removed. Electron micrographs that show wild-type cells in the process of removing the intervening cell wall are less common. In contrast, rvs161 mutant cells are often seen with wide contacts and at intermediate stages of cell wall re- moval (11, 20, 21, 91). Although the frequency of successful matings of rvs161 cells with wild-type cells is normal, electron micrographs reveal Downloaded from apparent delays and abnormalities even in these matings (21). Matings of rvs167 cells (rvs167 ϫ rvs167) also show apparent delays in cell wall removal at contact sites similar to that of rvs161 ϫ rvs161 matings despite the fact that the frequency of successful rvs167 ϫ rvs167 matings is similar to that of matings between wild-type cells (Fig. 6) (20). Evidence of a role for both Rvs161p and Rvs167p in mating was also found in a recent genome-wide genetic interaction study (84a). The ap- parent delay in localized removal of cell wall at sites of cell-cell http://mmbr.asm.org/ contact in rvs161 and rvs167 cells may be due to the larger area of wall that has to be removed. Alternatively, polarized deliv- ery of transport vesicles carrying cell wall-degrading enzymes to the site of contact may be affected. Electron micrographs show that in wild-type cells vesicles appear at sites where mating projections achieve cell-cell con- tact. These vesicles do not spread out evenly across the entire area of the contact, but instead form a tight cluster with a diameter of 0.6 ␮m, which is considerably less than the diam- on October 26, 2015 by University of Queensland Library eter of the zone of cell-cell contact (1 ␮m in wild-type and 2 to 3 ␮minrvs161 cells). Cell wall is degraded not throughout the region of cell-cell contact, but rather at a small zone within this region known as the fusion pore. Plasma membrane contact FIG. 4. Rvs161p functions in cell-cell fusion during mating. Shown are and fusion likely occur at the fusion pore. Intriguingly, the site matings between two wild-type haploids (panels A and B) or between two where vesicles cluster is precisely the site of the fusion pore. ⌬ rvs161 mutant haploids (panels C to F) visualized by fluorescence/dif- The diameter of the fusion pore is ϳ0.5 ␮m, which is similar ferential interference contrast optics. Cell nuclei are stained with the fluorescent dye 4Ј,6Ј-diamidino-2-phenylindole. Cell-cell fusion allows cy- to the diameter of the vesicle cluster. The vesicles that cluster toplasmic mixing and the nuclei of each mating partner meet at the site of at the fusion pore may carry enzymes that degrade cell wall cell fusion to fuse and create a single diploid nucleus. In the case of carbohydrates. Interestingly, some cell fusion mutants do not rvs161⌬ϫrvs161⌬ matings, delays in cell-cell fusion (apparent by differ- accumulate any vesicles at the site of cell contact (e.g., fus1 ential interference contrast) result in delays in nuclear fusion. This leads to the persistence of two nuclei. In panels C and D nuclear fusion is taking cells). Other cell fusion mutants accumulate vesicles but they place despite incomplete removal of the intervening cell wall (partial FusϪ do not form tight clusters at the fusion pore (e.g., spa2 mu- phenotype) and in panels E and F nuclear fusion has been completely tants). In contrast, both vesicle accumulation and clustering at Ϫ blocked by the inability to remove intervening cell wall (full Fus pheno- the fusion pore appear normal in rvs161 mutants (91). type). (Reproduced from reference 21 by copyright permission of the Electron micrographs reveal that mating rvs161 cells exhibit Rockefeller University Press.) an electron-dense plaque at the site of cell-cell contact (Fig. 5). This plaque may represent a fusion intermediate comprising newly deposited membrane material. This plaque is not ob- bilateral cell fusion mutation. In matings between rvs161 mu- served in cell fusion mutants that do not accumulate or cluster tants (rvs161 ϫ rvs161) mating projections are formed and vesicles, suggesting that this plaque material is transported by cell-cell contact is established, but cell-cell fusion is compro- the vesicles observed in clusters at the fusion pore. Interest- mised (Fig. 4) (20, 21, 91, 158). ingly, the electron-dense plaques observed in rvs161 cells are Consistent with a novel function in mating cells, an Rvs161p- major sites of plasma membrane invagination (Fig. 5) (91). GFP fusion was observed to redistribute to the tip of the Perhaps rvs161 cells have defects in fission of the necks of these mating projection upon treatment of haploid cells with ␣-fac- plasma membrane invaginations resulting in accumulation of tor mating pheromone. Interestingly, Rvs167p-GFP has also plaque material at the fusion pore. been reported to redistribute to the tip of the mating projec- In crosses to wild-type cells, why do rvs161 cells exhibit a defect tion upon pheromone treatment. Upon contact between the in default mating but not in mating in response to a pheromone tips of the two mating projections, Rvs161p-GFP remains con- gradient? Comparison of several mutants defective in default centrated at the site of cell contact throughout the processes of mating revealed a close correlation between defects in cell fusion cell wall removal and plasma membrane fusion. Electron mi- and defects in default mating (56). It was proposed that default croscopy showed that rvs161 mutant cells form cell-cell con- mating requires a much higher level of cell fusion activity than tacts that are abnormally large (Fig. 5). In mating wild-type mating in response to pheromone gradients. Hence, cell fusion VOL. 70, 2006 BAR DOMAIN PROTEINS 61 Downloaded from http://mmbr.asm.org/ on October 26, 2015 by University of Queensland Library

FIG. 5. Ultrastructure of wild-type, fus2, rvs161, and rvs161 fus2 cells during cell-cell fusion. Shown are electron micrographs of wild-type (WT), fus2⌬, rvs161⌬, and rvs161⌬ fus2⌬ cells undergoing cell-cell fusion during mating. Note that wild-type cells exhibit restricted zones of cell-cell contact during fusion while fus2⌬, rvs161⌬, and rvs161⌬ fus2⌬ mutants exhibit very wide zones of cell-cell contact during fusion. Wild-type and mutant cells all exhibit numerous vesicles that cluster at the zone of cell-cell fusion. Note that plasma membrane invaginations/tubules accumulate at the zone of fusion in the mutant cells, but are less common in wild-type cells. Labels indicate plasma membrane invaginations (black arrows), vesicle clusters (black arrowheads), electron-dense plaques (white arrows), and nonremoved cell wall remnants (asterisks). Bars ϭ 1 ␮m. (These previously unpublished images are shown here with the kind permission of Alison Gammie.) defects so mild that they barely affect mating using pheromone Could the cell fusion defect in rvs161 cells be attributable to the gradients still cause strong defects in default mating. When mat- defect in actin patch polarization or endocytosis? To address this ing in response to a pheromone gradient, yeast cells may use the question a collection of rvs161 mutant alleles was generated. gradient to more accurately localize the cell fusion machinery at rvs161 mutations confer defects in growth on media containing 1 the projection tip. Yeast cells mating in response to a pheromone MNaϩ or nonfermentable carbon sources such as glycerol (12, gradient are also more persistent in their attempts to mate. Either 42). Mutations in RVS161 that specifically affect sensitivity to Naϩ of these factors may compensate for a lowered cell fusion activity and utilization of glycerol (these mutations affect actin cytoskel- during normal pheromone gradient mating but not during default etal organization and endocytosis [ACE]) or that specifically mating. affect cell-cell fusion during mating (CF) were isolated. 62 REN ET AL. MICROBIOL.MOL.BIOL.REV. Downloaded from http://mmbr.asm.org/ on October 26, 2015 by University of Queensland Library

FIG. 6. Ultrastructure of wild-type, rvs161⌬, and rvs167⌬ cells during cell-cell fusion. Shown are electron micrographs of wild-type, rvs161⌬, and rvs167⌬ cells undergoing cell-cell fusion during mating. Panels a to c, matings between two wild-type cells; panels d to f, matings between two rvs161⌬ cells; panels g and h, matings between two rvs167⌬ cells. Bar, 1 ␮m. Labeled organelles: n, nucleus; v, vacuole; vs, vesicles; g, Golgi apparatus. (Reproduced from reference 20 with permission of the publisher. Copyright John Wiley and Sons Ltd.) VOL. 70, 2006 BAR DOMAIN PROTEINS 63

Intriguingly, rvs161 CF mutations do not affect actin cytoskel- Analysis of fus1⌬ rvs161⌬ and fus2⌬ rvs161⌬ double mutants etal organization or endocytosis. Conversely, rvs161 ACE mu- places Rvs161p in the Fus2p pathway of cell-cell fusion. Indeed, tations do not affect cell-cell fusion during mating. For exam- Rvs161p and Fus2p coimmunoprecipitate from yeast extracts. ple, the rvs161 mutant originally identified in the screen for Moreover, CF mutations in Rvs161p abolish interaction with endocytosis mutants (end6) (Rvs161p-R59K) is not defective Fus2p. The Rvs161p BAR domain and Fus2p are both predicted in cell-cell fusion during mating. Hence, Rvs161p has at least to form coiled-coil structures so it is likely that Fus2p interacts via two independent cellular roles, endocytosis/actin cytoskeleton coiled-coil interactions with Rvs161p. Why is Rvs161p-Fus2p in- teraction important for mating? Further work revealed that when

(ACE) and cell-cell fusion during mating (CF) (21). Downloaded from Rvs161p consists only of a BAR domain, and hence both Fus2p is bound to Rvs161p it is stable, but when Rvs161p-Fus2p ACE and CF functions lie within this domain (Fig. 1). Strik- interaction is abrogated by CF mutations Fus2p becomes unstable ingly, all ACE mutations map to the N-terminal 65% of and is degraded. Overexpression of Fus2p compensates for an Rvs161p, while all CF mutations map to the C-terminal 35% of inability of a CF mutant form of Rvs161p to bind Fus2p and Rvs161p. Thus, the Rvs161p BAR domain can be subdivided restores efficient mating. However, the converse is not true, i.e., into an N-terminal ACE domain regulating actin cytoskeleton overexpression of Rvs161p cannot compensate for loss of Fus2p. and endocytosis and a C-terminal CF domain regulating cell Hence, a major role of Rvs161p is to stabilize Fus2p (21). An understanding of Fus2p function in cell-cell fusion is

fusion. Rvs161p transcript levels increase fourfold and the http://mmbr.asm.org/ level of Rvs161p protein also increases following exposure of starting to emerge. A key finding was that mating yeast cells 2ϩ haploid cells to mating pheromone. This strongly supports the exhibit a transient flux of Ca ions across the plasma mem- view that Rvs161p provides some function specific for CF de- brane immediately prior to cell-cell fusion. This results in a 2ϩ spite the fact that Rvs161p clearly has functions other than CF transient increase in the level of free intracellular Ca and activation of various downstream signaling proteins. There are and is still expressed in both haploid cells not treated with 2ϩ mating pheromone and in diploid cells which cannot mate (21). two Ca influx systems, one of high affinity and one of low affinity. The latter has been named the low-affinity Ca2ϩ influx Vesicles and plasma membrane invaginations accumulate at 2ϩ the site of cell-cell fusion in cells lacking Rvs161p (rvs161⌬). Are system. Induction of the low-affinity Ca influx system (but not of the high-affinity system) is dependent on an integral these defects linked to the CF defect or the ACE defect? The plasma membrane protein known as Fig1p and on several on October 26, 2015 by University of Queensland Library rvs161 CF mutants were used to address this important question. known cell-cell fusion proteins, including Rvs161p, Fus1p, and Interestingly, rvs161 CF mutants still accumulate both vesicles and Fus2p. Loss of Fig1p (fig1⌬) results in a mating defect that can plasma membrane invaginations at the site of cell-cell fusion, ϩ be rescued by elevated Ca2 in the growth medium. Rvs161p although the total number of vesicles and invaginations per cell ϩ and Fus2p may be essential for the low-affinity Ca2 influx was somewhat lower than in cells lacking Rvs161p. Therefore, system because they play an important role in delivery of Fig1p while some of the vesicles and plasma membrane invaginations ϩ to the plasma membrane or in the activation of Fig1p Ca2 are probably attributable to the ACE defect, a subset certainly channel activity at the plasma membrane (210). appear to be specific for the CF defect. Rvs161p and Fus2p have at least one additional role in Electron micrographs reveal that rvs161⌬ cells accumulate two 2ϩ Ϯ mating. Unlike Fig1p (or the low-affinity Ca influx syste), types of vesicle. One type of vesicle has a diameter of 86 5nm Rvs161p and Fus2p are required for mating pheromone-in- and is found in vegetative cells predominantly at sites of polarized duced activation of the cell wall integrity Mpk1p/Slt2p mito- growth (Fig. 7). These are presumably the vesicles that accumu- gen-activated protein kinase cascade important for remodeling late when the ACE function is perturbed. A second type of vesicle of the cell wall during mating (210). In fus2⌬ cells, cell wall Ϯ with a diameter of 54 nm was observed in mating cells at the tip material, in particular ␤(1,3)-glucan, is improperly deposited of the mating projection and at the site of cell-cell fusion (Fig. 7). at the tip of the mating projection, where removal of cell wall These vesicles may be a novel type of secretory vesicle that de- is required for cell-cell fusion. The mating defects caused by livers cell wall-degrading enzymes to the tips of mating projection rvs161⌬ or fus2⌬ are partially suppressed by overexpression of and that accumulates when the CF function of Rvs161p is per- the protein Lrg1p, which is a Rho-GAP specific for Rho1p. turbed. The diameters of the vesicles that accumulate and form Rho1p in turn is an activator of the cell wall biosynthetic clusters at sites of cell-cell contact in the rvs161 CF mutant fus7 enzyme ␤(1,3)-glucan synthase (52). Loss of Lrg1p (lrg1⌬) are also small (ϳ50 nm) (Fig. 5) (20, 21, 91). appears to increase the activity of ␤(1,3)-glucan synthase and What is the role of Rvs161p in cell-cell fusion during mating? enhance cell wall deposition. This in turn inhibits cell-cell fu- Two genetically defined pathways important for fusion during sion. Lrg1p localizes in wild-type cells to the tip of the mating mating have been identified. One pathway is dependent on the projection, where it may act to locally inhibit Rho1p and Fus1p protein and the other on the Fus2p protein. Both Fus1p ␤(1,3)-glucan synthase to restrict cell wall deposition. Over- and Fus2p are important for cell-cell fusion during both default expression of Lrg1p in fus2⌬ mutants reduces the improper mating and mating in response to pheromone gradients, and the ␤(1,3)-glucan deposition at the tip of the mating projection expression of both proteins is induced by treatment with mating and corrects the mating defect (76). pheromone. Fus1p is an O-glycosylated integral plasma mem- brane protein, while Fus2p is a cytoplasmic coiled-coil protein rvs Mutations Interact with Actin and Myosin Mutations that localizes to patches at the cortex of the mating projection. The functions of Fus1p and Fus2p are partially redundant, since Rvs؊ phenotypes are associated with mutations affecting loss of both proteins has more severe effects on cell-cell fusion other actin cytoskeleton proteins. In the original screens for rvs than loss of each individual protein (56, 67, 335, 336). mutants, only a very few of the many mutations screened af- 64 REN ET AL. MICROBIOL.MOL.BIOL.REV. Downloaded from http://mmbr.asm.org/ on October 26, 2015 by University of Queensland Library

FIG. 7. Defects in vesicle fusion in rvs161 mutant cells result in vesicle accumulation. Shown are electron micrographs of wild-type (A), rvs161⌬ (B), and rvs167⌬ (C) yeast cells. The panels on the left depict cells grown under standard conditions,while those on the right depict cells grown in medium containing sublethal levels (3.4%, wt/vol) of Naϩ. Bar, 1 ␮m. Labeled organelles: n, nucleus; v, vacuole; vs, vesicles; g, Golgi apparatus apparatus; ps, primary septum; s, septum. The arrow indicates cell wall abnormalities. (Reproduced from reference 20 with permission of the publisher. Copyright John Wiley and Sons Ltd.) VOL. 70, 2006 BAR DOMAIN PROTEINS 65 Downloaded from http://mmbr.asm.org/ on October 26, 2015 by University of Queensland Library

FIG. 7—Continued. 66 REN ET AL. MICROBIOL.MOL.BIOL.REV. Downloaded from http://mmbr.asm.org/ on October 26, 2015 by University of Queensland Library

FIG. 7—Continued. fected survival upon starvation without also having nonspecific actin cytoskeletal and endocytosis proteins. Moreover, proteins effects on cell viability and thereby satisfied the criteria for a such as Whi2p that were originally thought to be specific reg- true RvsϪ phenotype (42). However, subsequent screens for ulators of the cell cycle in response to nutrients are now known mutations that affect arrest in G0 when cells are starved of to also be required for actin patch polarization and endocyto- nutrients identified not only the Rvs proteins, but also other sis. Hence, cell cycle arrest in response to starvation and long- VOL. 70, 2006 BAR DOMAIN PROTEINS 67 term viability upon starvation appear to require not only pro- found in the original single mutants. A striking result was that teins such as Whi2p, Rvs161p, and Rvs167p, but a functional genetic interaction is allele specific. Some actin mutations (e.g., actin cytoskeleton and/or endosomal system in general (24). act1-1) in combination with either rvs161⌬ or rvs167⌬ re- Essentially all the phenotypes associated with loss of sulted in lethality. In contrast, other actin mutations (e.g., Rvs161p or Rvs167p are also phenotypes associated with mu- act1-4) in combination with rvs161⌬ or rvs167⌬ yielded dou- tations in actin (10, 15, 19, 206, 215, 363). Among the collection ble mutants that were no more defective than the single of act1 charged-to-alanine mutant alleles described above mutants. The act1 alleles that are lethal in combination with (363), there is a strong correlation between act1 alleles that rvs mutations may affect the interaction of actin with a Downloaded from affect growth in the presence of Naϩ and those that affect protein that is functionally redundant with Rvs proteins in utilization of nonfermentable carbon sources. Interestingly, a some essential cellular process (19). number of act1 alleles confer reduced viability upon starvation. The three-dimensional structure of rabbit muscle actin has These alleles tend to be those that also affect growth in the been determined and the very high amino acid sequence ho- presence of Naϩ (19). These phenotypic similarities between mology of rabbit muscle actin to yeast actin has led to a plau- mutations affecting actin, Rvs161p, and Rvs167p suggest a sible predicted structure for yeast actin. The residues altered in close connection between Rvs161p and Rvs167p function and each act1 mutant have been mapped onto the predicted yeast actin cytoskeleton function. actin structure (125, 363). Several mutations that are lethal Genetic interactions between Rvs proteins and actin. Fur- when combined with rvs167⌬ map to a patch on the surface of http://mmbr.asm.org/ ther support for a functional connection between the Rvs pro- the actin molecule where myosins are known to bind. Although teins and the actin cytoskeleton comes from genetic interaction only a few act1 mutations were tested for lethality with data. The first genetic interaction was discovered by chance. rvs161⌬, those that exhibit lethality also map to the same patch Both the act1-1 mutation affecting actin and the end6 (later on the surface of actin (19). renamed rvs161-E6) R59K substitution in Rvs161p are reces- Genetic interactions between Rvs proteins and yeast myo- sive (i.e., act1-1/ACT1 and rvs161-E6/RVS161 heterozygous sins. The yeast genome encodes five MYOsin genes (MYO1 to diploids are without obvious phenotype). Remarkably, how- MYO5). Myosins all contain a globular domain known as the ever, a diploid doubly heterozygous for act1-1 and rvs161-E6 S1 fragment or head domain. This domain has ATPase activity (act1-1/ACT1 rvs161-E6/RVS161) displays the mutant pheno- and binds actin. The energy of ATP hydrolysis is coupled to on October 26, 2015 by University of Queensland Library type (i.e., is unable to grow at elevated temperature). This type association and dissociation of the myosin head domain with of genetic interaction is known as nonallelic noncomplemen- actin filaments and directed movement of the myosin along tation because rvs161-E6 and act1-1 affect different genes (i.e., actin filaments. Myosins in muscle have long coiled-coil tail are nonallelic), but nevertheless do not exhibit genetic comple- domains that associate in a tail-to-tail orientation to form mentation. This genetic interaction is specific for the rvs161-E6 dimers known as thick filaments. The myosin thick filaments allele, because rvs161⌬ and act1-1 fully complement, as one interdigitate actin thin filaments, and hydrolysis of ATP by would expect. Furthermore, rvs167⌬ and act1-1 also fully com- myosin is coupled to sliding of the myosin thick filaments over plement (215). A subsequent study showed rvs161-E6 exhibits the actin thin filaments during muscle fiber contraction. nonallelic noncomplementation with act1-1 for several other Not all myosins have coiled-coil tail domains, nor do all phenotypes, including bipolar bud site selection, growth in the myosins form dimers. The myosins that do have coiled-coil tail presence of Naϩ, and utilization of nonfermentable carbon domains and form dimers are known as type II myosins or sources (19). conventional myosins because they resemble muscle myosin. What does nonallelic noncomplementation tell us about Other myosins may have short tail domains that bind cargo, Rvs161p and actin? There are two ways in which nonallelic organelles, vesicles, mRNA, etc., that is then transported along noncomplementation can arise. One is based on gene dosage. actin filaments by the activity of the ATPase head domain (70, A doubly heterozygous diploid will have only half the normal 71, 95, 213, 214, 252, 358). level of each protein and this may be insufficient. This does not Genetic interactions between each yeast myosin and account for the case of rvs161-E6 and act1-1 because nonallelic Rvs161p and Rvs167p have been examined. The strongest ge- noncomplementation was not observed with rvs161⌬ and netic interaction resulted when loss of the only type II conven- act1-1 even though this diploid would have only half the nor- tional myosin in yeast, Myo1p (myo1⌬), was combined with mal level of each protein. The other way in which nonallelic rvs161⌬ or rvs167⌬; these double mutants were inviable (19). noncomplementation can arise is by the formation of a toxic Myo1p is the myosin that forms the actomyosin contractile ring complex between the two mutant proteins. This better ac- during cytokinesis (16, 178, 354). Hence, this genetic interac- counts for the nonallelic noncomplementation of rvs161-E6 tion suggests a role for Rvs161p and Rvs167p in cytokinesis. and act1-1 because it explains why rvs161⌬, which does not Neither rvs161 nor rvs167 mutants have been tested for defects encode a protein, would not show the same genetic interaction. in cytokinesis, however, loss of the Rvs167p-associated pro- Nonallelic noncomplementation suggests the mutant form of teins Las17p and Vrp1p causes defects in cytokinesis. In the Rvs161p (Rvs161p-R59K) binds the mutant form of actin in case of Vrp1p the cytokinesis and actin patch polarization vivo and this complex has deleterious effects. defects are genetically separable and probably represent A more extensive study of genetic interactions between actin distinct activities (173, 218, 260, 328). Loss of the Rvs161p and Rvs161p and Rvs167p was subsequently reported. Actin ortholog in the fission yeast Schizosaccharomyces pombe and Rvs161p or actin and Rvs167p mutations were combined (Hob3) affects cytokinesis (276). Like rvs161⌬ and rvs167⌬, in haploid cells and the double mutants created were examined both las17⌬ and vrp1⌬ mutations are lethal in combination for combinations that resulted in more severe defects than with myo1⌬ (275). A tight block in cytokinesis may explain 68 REN ET AL. MICROBIOL.MOL.BIOL.REV. why some rvs161 and rvs167 mutant cells continue to display other actin patches did not appear to be in contact with the buds with nuclei even under conditions of nutrient starva- plasma membrane. This may be due to an ability of actin tion (12, 42). patches to reversibly associate and dissociate from plasma The myo2-66 mutation is not lethal in combination with membranes. Alternatively, due to technical limitations, elec- rvs161⌬ or rvs167⌬, but the double mutants grow much more tron microscopy may not reveal all the plasma membrane con- poorly than either single mutant (19). MYO2 encodes Myo2p, tacts. which is a type V unconventional myosin and the only essential Fractionation of Rvs proteins with membranes. Subcellular myosin in yeast. Myo2p localizes to zones of polarized growth fractionation approaches initially revealed that the majority of Downloaded from where actin patches are also found, but does not strictly colo- Rvs167p and a significant fraction of Rvs161p are in tight calize with actin patches. A conditional allele, myo2-66, causes association with membranes in vivo (10, 98). Recently, purified temperature-sensitive defects in polarized secretion to the bud Rvs161p and Rvs167p were shown to directly bind liposomes in and accumulation of what appear to be secretory vesicles, vitro (84a). similar to rvs161 and rvs167 mutants. Moreover, myo2-66 is Rvs proteins and lipid rafts. Membranes both in mammalian lethal in combination with mutations that affect fusion of Golgi cells and in yeast have been shown to consist of a number of apparatus-derived secretory vesicles with the plasma mem- distinct “microdomains” characterized by distinct protein and brane (104, 138). Myo2p has been proposed to move secretory lipid compositions. The best characterized of these microdomains vesicles along actin cables to sites of surface growth, where are lipid rafts. Lipid rafts are formed by the lateral association of http://mmbr.asm.org/ they dock and fuse with the plasma membrane (273). The sphingolipids and sterols (cholesterol in mammalian cells and negative genetic interaction between myo2-66 suggests a role ergosterol in yeast) and have the biochemical property that, for Rvs161p and Rvs167p in polarized secretion. when maintained on ice, they resist solubilization by nonionic No genetic interactions between Rvs161p or Rvs167p and detergents such as Triton X-100. This property has led to lipid the other yeast myosins, Myo4p (type V myosin), Myo3p (type rafts’ being also referred to as detergent-resistant membranes I myosin), or Myo5p (type I myosin), were observed (19). or detergent-insoluble glycosphingolipid-enriched membranes Myo3p and Myo5p colocalize with Rvs167p in actin patches (22, 113, 159, 233). In wild-type cells, both Rvs167p and and have redundant functions in actin patch polarization and Rvs161p are associated with detergent-insoluble membranes. endocytosis (5, 96, 102). The type I myosins interact via an SH3 In mutant cells deficient in sphingolipid biosynthesis Rvs161p on October 26, 2015 by University of Queensland Library domain with Vrp1p and Las17p and via a C-terminal acidic tail becomes detergent soluble, suggesting the Rvs proteins may with the Arp2/3p complex and play a role in de novo actin partition into glycosphingolipid-enriched lipid raft micro- filament assembly (72, 161). Given the redundancy of Myo3p domains (10, 97). Further evidence that Rvs proteins are as- and Myo5p, it would be interesting to know the phenotype of sociated with glycosphingolipid-enriched lipid raft membranes myo3⌬ myo5⌬ rvs161⌬ and myo3⌬ myo5⌬ rvs167⌬ triple in vivo has come from genetic approaches. A key discovery was mutants. that mutations that affect glycosphingolipid biosynthesis sup- Actin mutations that map to the surface patch and affect press the full range of rvs phenotypes (10, 53, 97, 300, 381) (see binding of myosin head domains may genetically interact with below). rvs mutations because the myosin(s) whose binding is affected In yeast, the glycosylphosphatidylinositol (GPI)-anchored has redundant function with the Rvs proteins (19). If this protein Gas1p and the integral plasma membrane Hϩ-ATPase explanation is correct, the myosins whose binding is affected Pma1p are markers of lipid rafts (9). Lipid rafts form in the ER may be Myo1p and Myo2p. The Rvs proteins may help to (9). Export of lipid raft proteins such as GPI-anchored proteins stabilize actin-myosin interactions and become essential when from the ER has requirements distinct from those for nonraft actin-myosin interaction is affected by mutation of either pro- proteins, in particular a requirement for continuing glyco- tein. Alternatively, the Rvs proteins may act downstream to sphingolipid biosynthesis (127, 311). tether actin filaments to sites of vesicle fusion or play a role in A distinct class of transport vesicle that is lipid raft enriched vesicle fusion itself. mediates the transport of lipid rafts from the ER. The forma- tion of this type of transport vesicle has been reconstituted in vitro and requires the Rab GTPase Ypt1p. In reactions defi- Rvs Proteins and Membranes cient in Ypt1p lipid rafts are still exported from the ER but Association of cortical actin cytoskeleton with membranes. instead of being transported in lipid raft-enriched vesicles they The ability of yeast Rvs proteins to associate with lipid mem- are transported by default in vesicles containing nonraft pro- branes has been examined only recently. The diffuse cytoplas- teins (202). Mutations that affect sphingolipid biosynthesis per- mic distribution initially reported for Rvs161p did not suggest turb delivery of Pma1p to the cell surface (9). This suggests an association with membranes (21). In the case of Rvs167p, that sorting of Pma1p to the cell surface involves partitioning which localizes to cortical actin patches, a membrane associa- into lipid rafts. Interestingly, mutations affecting the actin tion seemed easier to envisage (11). Cortical actin patches have patch protein Sla2p reduce Pma1p levels at the plasma mem- been visualized by electron microscopy and at least some actin brane, suggesting that sorting of lipid raft proteins to the cell patches are found at sites where the plasma membrane forms surface requires a functional actin cytoskeleton (217). “finger-like” invaginations. The invaginations exhibit immuno- One of the important functions of lipid rafts is in providing reactivity for Abp1p and actin. Antibodies against actin deco- a platform for integration of signal transduction pathways. One rate transverse striations along the invagination that were pro- of the best examples is the activation of the mitogen-activated posed to be actin filaments. While some actin patches were protein kinase Raf by the small GTPase H-Ras, which only observed in contact with the plasma membrane invaginations, occurs when H-Ras partitions into lipid raft microdomains (27, VOL. 70, 2006 BAR DOMAIN PROTEINS 69

249). In yeast there exist five integral plasma membrane pro- that rescue the various rvs defects were isolated and charac- teins, Wsc1p to Wsc4p and Mid2p, that act as sensors of dif- terized. The rationale behind this approach is that by charac- ferent stress conditions (152, 179, 241, 346). Wsc1p is the terizing the mutations that rescue the rvs mutant defects we sensor responsible for signaling actin patch depolarization in may gain insight into what caused those rvs defects. Extragenic response to elevated temperature or exposure to Naϩ (10, 52). mutations were isolated that suppress rvs161⌬ (sur). Four sep- Wsc1p localizes in part to lipid rafts (179). Protein kinase C arate screens were performed to isolate sur mutations that (PKC) and mitogen-activated protein kinases regulate polar- restore different rvs phenotypes, i.e., viability upon glucose ization and depolarization of actin patches upon exposure of starvation, utilization of nonfermentable carbon sources (glyc- Downloaded from cells to stress and downstream of Wsc1p. Exposure of wild-type erol/ethanol), growth in the presence of 3-AT, and growth in yeast cells to high temperature or to high Naϩ induces depo- the presence of Naϩ. Since the starting strain had a deletion of larization of actin patches over a period of about 30 min. In the the RVS161 gene, all suppressor mutations isolated affect other continual presence of the stress, cells gradually adapt, and after genes (i.e., they are extragenic) (53). 90 min the actin patches have returned to a fully polarized Of Ͼ100 extragenic suppressor mutations isolated, some distribution (35). suppressed only the specific RvsϪ phenotype they were se- Protein kinase C (Pkc1p) is implicated both in depolariza- lected on (e.g., restored growth in the presence of Naϩ, but not tion of actin patches and in actin patch repolarization. In survival upon glucose starvation). Ten suppressor mutations, contrast, the PKC1-dependent mitogen-activated protein ki- however, that suppressed all RvsϪ phenotypes were recovered. http://mmbr.asm.org/ nase Slt2p/Mpk1p is not implicated in depolarization of actin These sur mutations suppressed the phenotypes not only of patches in response to heat stress, but is specifically required rvs161 but also rvs167 and rvs161 rvs167 double mutants. All sur for adaptation to heat stress and subsequent actin patch repo- mutations were recessive and genetic complementation analy- larization. The PKC effectors that mediate actin patch depo- sis showed they affect four separate genes called SUR1 to SUR4 larization are yet to be identified (52). Loss of PKC results in (53). Interestingly, sur1-4 mutations suppress vesicle accumu- loss of cell wall integrity and cell lysis during bud growth. In lation in rvs161 and rvs167 mutant cells (20). sur4 (vbm1) and comparison, loss of the downstream mitogen-activated protein fen1/sur5 (vbm2) were also isolated in an independent study as kinase Slt2p has less severe phenotypes and these mutants are mutations that suppress the growth and vesicle accumulation viable but exhibit depolarized actin patches, mild cell wall defects of snc1 snc2 double mutants (48). Snc1p and Snc2p are on October 26, 2015 by University of Queensland Library defects, and temperature sensitivity as do cells deficient in vesicle-associated soluble N-ethylmaleimide-sensitive fusion Rvs161p or Rvs167p. Intriguingly, rvs mutations exhibit strong protein attachment protein receptors that localize to post- genetic interactions with slt2 mutations: while both rvs and slt2 Golgi apparatus secretory vesicles and mediate docking and single mutants are viable, an rvs slt2 double mutation is lethal fusion of these vesicles with the plasma membrane (251). (10). The terminal phenotype of rvs slt2 mutants is reminiscent Snc1p interacts genetically with both Ras GTPases and Srv2p, of that of pkc1 mutants (arrested cells exhibit cell lysis during suggesting a role in response to nutrient availability (98). bud growth). The sur single mutants do not exhibit reduced viability upon Are rvs mutants defective in actin patch repolarization? In- starvation and they grow well on nonfermentable carbon deed, loss of either Rvs protein does not affect stress-induced sources. sur1-3 mutants sporulate well and exhibit a bipolar actin patch depolarization, but blocks repolarization. Depolar- budding pattern in diploids, but sur4 mutants have severe de- ization of actin patches is a physiological response to stress. fects in sporulation and bipolar bud site selection, like rvs Actin patch depolarization observed in untreated rvs mutant mutants. The sporulation defect exhibited by sur4 arises in cells may be due to stresses encountered normally during spite of the ability of sur4 mutants to utilize nonfermentable growth to which wild-type cells but not rvs mutant cells adapt carbon sources and survive nitrogen starvation. Moreover, sur4 easily. Restoration of actin patch polarity may require repair of rvs161 and sur4 rvs167 double mutant diploids are able to cell wall damage induced by elevated temperature or Naϩ. sporulate efficiently even though each single mutant diploid is Slt2p and Rvs proteins may function in distinct pathways down- not. These data suggest Rvs and Sur4p proteins have specific stream of PKC that are important for repair of cell wall dam- roles in sporulation. Where tested, deletion of each SUR gene age and actin patch repolarization. Slt2p may be required for suppresses rvs161 and rvs167 mutant phenotypes, similar to the induction of cell wall repair enzymes via phosphorylation of original sur mutation. This suggests the sur mutations isolated specific transcription factors. Rvs proteins may be required for in the mutant screen are loss-of-function mutations (53). polarized delivery of newly synthesized cell wall repair en- The genes affected by the sur1, sur2, and sur4 mutations have zymes to the cell surface. Consistent with this, mutation of the been identified (the gene affected by the sur3 mutation is still cell wall repair enzyme ␤-glucan synthase (kre6) causes only unknown). SUR1, SUR2, and SUR4 encode ER-localized inte- mild growth defects, but kre6 rvs167 double mutants are barely gral membrane proteins that function in biosynthesis of glyco- viable (10). Mapping of in vivo phosphorylation sites in sphingolipids (Fig. 8). Unlike glycerol-based phospholipids Rvs167p has not revealed phosphorylation of any site that that comprise a glycerol base to which one phosphate and two matches the PKC consensus, and it is not yet clear if Rvs167p long-chain fatty acids (C16 to C18) are esterified, glycosphin- is a PKC substrate. golipids contain a phytosphingosine base with one very long ؊ Suppression of Rvs phenotypes by mutations affecting gly- chain fatty acid (C26) attached via an amide linkage (54). The cosphingolipid biosynthesis. Loss of Rvs161p or Rvs167p re- phytosphingosine base is synthesized by the transfer of serine sults in diverse phenotypes. Several studies have sought to to palmitic acid (as palmitoyl-coenzyme A) to form 3-keto- determine if all these phenotypes result from one basic under- dihydrosphingosine followed by two reduction steps, first yield- lying deficiency. In one study, extragenic suppressor mutations ing dihydrosphingosine and then phytosphingosine (Fig. 8). 70 REN ET AL. MICROBIOL.MOL.BIOL.REV. Downloaded from http://mmbr.asm.org/ on October 26, 2015 by University of Queensland Library

FIG. 8. Sphingolipid biosynthesis in yeast. The yeast sphingolipid biosynthetic pathway indicating the steps blocked by sur1, sur2, sur4, fen1/sur5, and ipt1 mutations that suppress the phenotypes of rvs161 and rvs167 mutants. CoA, coenzyme A; LCB, long-chain base; PI, phosphatidylinositol;

Inos-P, inositol-1-P; Cer, ceramide; IPC, inositolphosphorylceramide; MIPC, mannosyl-inositolphoshorylceramide; M(IP)2C, mannosyl-diinosi- tolphoshorylceramide; Man-P-Inos, mannose(␣-1,2)inositol-1-P. (Reprinted from reference 54 with permission from Elsevier.)

The C26 fatty acid is then attached to phytosphingosine to yield SUR4/ELO3 also results in resistance to fenpropimorph. ceramide. Ceramide is converted into glycosphingolipids, which The mechanism by which these mutations confer resistance to in yeast are inositolphosphoceramide, mannosylinositolphospho- fenpropimorph has not been elucidated, however, this result does ceramide, and mannosyldiinositylphosphoceramide (54). suggest that Sur4p has a function closely related to that of The product of the SUR2 gene (Sur2p) is the reductase that Fen1p. Further characterization of Fen1p revealed that it, like converts dihydrosphingosine to phytosphingosine during syn- Sur4p, is involved in the elongation of fatty acids to give the thesis of the ceramide base. The product of SUR4 (Sur4p) is C26 fatty acid utilized in sphingolipid biosynthesis and FEN1 is required for elongation of fatty acid chains by fatty acid syn- also known as ELO2. fen1 mutations, like sur4 mutations, affect thase, e.g., conversion of palmitate (C16) to very long chain bipolar budding in homozygous diploids. Also like sur4 muta- fatty acids (C26). As very long chain fatty acids are incorpo- tions, fen1 mutations suppress the phenotypes of rvs161 and rated into sphingolipids, sur4 mutants are also defective in rvs167, and FEN1/ELO2 is also known as SUR5. sur4⌬ fen1⌬ sphingolipid biosynthesis. SUR4 is also known as ELOngation double mutants are inviable. This shows Sur4p and Fen1p have 3 (ELO3). The SUR1 gene product (Sur1p) acts downstream of redundant functions in an essential process, presumably syn- SUR2 and SUR4 and is required for conversion of ceramide to thesis of very long chain fatty acids (262, 300). glycosphingolipids. The mannosylation of inositolphosphocer- The other gene that encodes a protein homologous to Sur4p amide to yield mannosylinositolphosphoceramide is dependent on Sur1p (54). IPT1 is the gene required for conversion of is also required for fatty acid elongation and has been named mannosylinositolphosphoceramide to mannosyldiinositylphos- ELO1. Unlike FEN1/ELO2 and SUR4/ELO3, which are re- phoceramide (the step after the Sur1p step). Deletion of IPT1 quired for fatty acid elongation to yield very long chain fatty (ipt1⌬) suppresses the phenotypes of rvs161⌬ mutants (rvs167⌬ acids with a chain length of C26, ELO1 is involved in elongation was not tested), and hence IPT1 is also a SUR gene (10). of medium-chain C12 to C16 fatty acids to long-chain C16 to C18 When the amino acid sequence of SUR4/ELO3 was de- fatty acids. Deletion of ELO1 does not affect doubling time or termined it was noticed that two other yeast genes encode bipolar bud site selection. Moreover, elo1 mutations do not homologous proteins. One, FEN1, was originally identified suppress the phenotypes of rvs mutants (so ELO1 is not a SUR as the gene affected in a mutant resistant to the toxic sterol gene). This finding shows that suppression of rvs is specific to biosynthesis inhibitor fenpropimorph. Indeed, mutation of mutations that affect glycosphingolipid synthesis, which re- VOL. 70, 2006 BAR DOMAIN PROTEINS 71

quires C26 but not C16-18 fatty acids (262, 285). rvs mutants: sur1, sur2, sur4, fen1, and ipt1 also restore growth The phospholipid composition of rvs and sur single and of act1 mutant cells in the presence of Naϩ although effects on double mutants has been compared. The total amount of phos- actin patches in act1 mutant cells were not investigated (10). pholipid in rvs161⌬ was similar to that of the wild type, but the How do sphingolipids influence actin patch depolarization phospholipid content was reduced twofold in sur1, rvs161⌬ and repolarization? In mammalian cells, sphingolipids such as sur1, sur4, and rvs161⌬ sur4 mutant cells. The rvs161⌬ mutant ceramide act as important second messengers in a wide variety does not have an altered phospholipid composition. In all sur1 of signaling pathways that regulate differentiation, prolifera- and sur4 mutants the percentage of phosphatidylcholine and tion, apoptosis, and malignancy (111, 112, 156, 228). Sphingoid Downloaded from phosphatidylethanolamine increased, while the percentage of bases (e.g., phytosphingosine) are important second messen- phosphatidylglycerol plus cardiolipin and phosphatidylinositol gers in yeast and exogenous phytosphingosine is sufficient to decreased. The decrease in phosphatidylinositol was slight in restore endocytosis in yeast mutants unable to synthesize en- sur1 mutants but dramatic in sur4 mutants (approximately 70% dogenous sphingolipids. Restoration of endocytosis in sphin- reduction). The abundance of phosphatidylserine was relatively golipid-deficient mutants by exogenous phytosphingosine does normal in all mutants (53). It is more likely the suppression of rvs not require incorporation into ceramide or glycosphingolipids, phenotypes is associated with the altered glycosphingolipid com- so the phytosphingosine does not have to be incorporated into position rather than the altered phospholipid composition. lipid raft components. Instead, restoration of endocytosis by SUppressor of Rvs167 7 (SUR7) was identified in a screen to exogenous phytosphingosine depends on its ability as a second http://mmbr.asm.org/ identify wild-type genes that, when overexpressed, suppress the messenger to activate protein kinases Pkh1p/Pkh2p, which in RvsϪ phenotype of rvs167-1. Overexpression of SUR7 partially turn activate PKC, which results in activation of the down- rescued all the defects of rvs167 as well as rvs161 and rvs161 stream mitogen-activated protein kinase Mpk1p/Slt2p. These rvs167 double mutants. Deletion of SUR7 (sur7⌬) did not con- signaling pathways regulate many cellular responses, including fer any of the phenotypes of rvs mutants except for a somewhat actin patch repolarization (82, 83, 156, 384). reduced efficiency of sporulation. sur7⌬ did not suppress any There are two views about the importance of actin patch phenotypes when combined with the rvs161 or rvs167 mutation. repolarization for suppression. According to one view, actin The mechanism of SUR7 suppression therefore appears to patch repolarization is critical and suppression of RvsϪ phe- differ from that of sur1-5. SUR7 encodes an integral membrane notypes is interpreted according to the effects of both phyto- on October 26, 2015 by University of Queensland Library protein (Sur7p) with three or four predicted transmembrane sphingosine base and lipid raft-dependent Wsc1p on actin domains whose expression is regulated and peaks in late G2-M patch distribution. The levels of endogenous phytosphingosine of the cell cycle. Sur7p localizes to numerous small cortical and lipid rafts in sur mutants are proposed to influence the patches, which are distinct from actin patches (300, 381). phenotype of rvs mutants. According to the second view, sup- Sur7p is implicated in glycosphingolipid biosynthesis, but at pression still may involve phytosphingosine but can occur with- a different step than Sur1-5p. The phytosphingosine base of out actin patch repolarization. sphingolipids can be either C18 or C20 in length and is also Balguerie et al. reported data in support of the first view. variably hydroxylated. Glycosphingolipids are classified as class fen1 mutants, which repolarize more rapidly than the wild type, A, B, C, or D based on whether the phytosphingosine base has are blocked at an early step of glycosphingolipid synthesis and 1, 2, 3, or 4 hydroxyl groups, respectively. The plasma mem- accumulate phytosphingosine. Excess endogenous phytosphin- branes of sur7⌬ cells contain the normal ratio of sphingolipids gosine in fen1 mutants (and possibly also sur4 mutants, which with C18 and C20 phytosphingosine bases, but they exhibit a are also blocked at an early step) is proposed to induce repo- reduced amount of the more heavily hydroxylated class C and larization of actin patches to such an extent that Rvs161p is no D glycosphingolipids and a corresponding accumulation of longer necessary. sur1, sur2, and ipt1 mutant cells are blocked class B glycosphingolipids with only two hydroxyl groups. The later in the glycosphingolipid biosynthetic pathway and do not phospholipid composition was unaffected in plasma membranes accumulate phytosphingosine but may also have reduced levels from sur7⌬ cells (381). Given that SUR7 overexpression, but not of lipid rafts due to lowered glycosphingolipid levels. When deletion of SUR7, suppresses rvs161 and rvs167 mutant pheno- exposed to Naϩ these mutants are not able to fully depolarize types, it will be interesting to see the effect of SUR7 overexpres- their actin patches, so repolarization may not require Rvs161p. sion on glycosphingolipid and phospholipid composition. As mentioned above, one pool of Wsc1p localizes to lipid How do sur mutations suppress RvsϪ phenotypes? Does rafts (179). If lipid rafts are important for signaling actin altered glycosphingolipid composition in sur mutant cells affect patch depolarization, this may explain why mutants such as lipid raft-based signaling pathways that control depolarization sur1, sur2, and ipt1 mutants cannot fully depolarize their and/or repolarization of actin patches? In support of this idea, actin patches (10). sur4 and fen1 cells depolarize their actin patches somewhat Germann et al. reported data in support of the second view. faster than wild-type cells upon exposure to Naϩ. In contrast, This study confirmed that sur4⌬ suppresses the Naϩ sensitivity sur1, sur2, and ipt1 cells are not able to fully depolarize their of rvs161 and rvs167 mutants. It did so, however, without sig- actin patches upon exposure to Naϩ. Adaptation to the Naϩ nificantly restoring actin patch polarization. Furthermore, in stress and repolarization of actin patches were normal in sur4, sur4⌬ cells Rvs167p does not localize to actin patches but fen1, sur1, and sur2 cells, but slower in ipt1 cells. Importantly, redistributes to the cytoplasm, where it is proteolytically de- each rvs161 sur double mutant cell was able to adapt to stress graded. This reinforces the view that, despite improved growth, and repolarize actin patches after Naϩ-induced depolarization, the actin patch defect is not rescued by sur4⌬. This study whereas rvs161 single mutant cells were not. The suppressing identified several novel Rvs167p-interacting proteins and effect of sur mutations on RvsϪ phenotypes is not restricted to showed that some do not localize to actin patches. Intriguingly, 72 REN ET AL. MICROBIOL.MOL.BIOL.REV. the most critical Rvs167p-interacting proteins for sur4-medi- cycle-regulatory Cdk known as Cdc28p in budding yeast (Cdc2/ ated suppression were two proteins, Acf2p and Gdh3p (Table Cdk2 in fission yeast and mammalian cells) (200). The mam- 1), that do not localize to actin patches. In contrast, the malian ortholog of budding yeast Pho85p is Cdk5, and its Rvs167p-interacting proteins Abp1p and Sla1p, which do lo- activity is critical for neuronal function (303). Mammalian calize to actin patches, were not required for sur4-dependent Cdk5 has been shown to functionally substitute for Pho85p suppression of rvs161⌬ and Abp1p was not required for sur4- when expressed in budding yeast, so Cdk5 and Pho85p are dependent suppression of rvs167⌬ (note that Sla1p could not functional orthologs (129, 225). be tested because sla1⌬ rvs167⌬ is lethal [176]). Interestingly, As for all Cdks, the protein kinase activity of both Cdc28p Downloaded from Acf2p and Gdh3p fractionate with Rvs167p in detergent-insol- and Pho85p is dependent on their association with regulatory uble membranes, suggesting all three proteins localize to lipid subunits known as cyclins whose expression level often varies rafts such as Rvs161p (10, 97). through the cell cycle. Cdc28p is activated by one set of cyclins

These findings are in agreement with recent evidence from known as G1 cyclins (Cln1p to Cln3p) in late G1 for entry into several other studies that found actin patch polarization is not the cell cycle and by two other sets of cyclins known as S-phase strictly required for growth even at elevated temperatures. The and mitotic cyclins (Clb1p to Clb6p) at subsequent stages of idea that actin patch polarization may be important for growth the cell cycle. Pho85p has 10 known cyclins known as Pho85 came from two observations: no viable mutant that lacks actin cyclins (Pcls). The Pcls have been classified into two subfamilies patches has yet been reported (hence the existence of actin based on homology, the Pcl1p/2p subfamily (comprising Pcl1p, http://mmbr.asm.org/ patches appears to be essential) (149), and many mutants that Pcl2p, Clg1p, Pcl5p, and Pcl9p) and the Pho80p subfamily (com- exhibit depolarized actin patches also exhibit slow growth at prising Pho80p, Pcl6p-Pcl8p, and Pcl10p) (6, 170, 190). normal temperature and are unable to grow at elevated tem- In conjunction with different Pcls, Pho85p plays a role in perature. Examples include loss of Rvs167p, Rvs161p, Sla1p, phosphate and glycogen metabolism as well as a role in G1-S Sla2p/End4p, Sac6p, Srv2p, Vrp1p, Las17p, and point muta- cell cycle progression. Pcl1p and Pcl2p in association with tions in actin, Pan1p, and Arp2/3p complex subunits (15, 60, Pho85p are functionally redundant with Cdc28p and the G1 126, 173, 176, 181, 197, 215, 219, 321, 340, 363, 364). There are, cyclins Cln1p and Cln2p in G1-S cell cycle progression. This however, strains in which actin patch polarization is lost with- late G1 role is specific to Pcl1p and Pcl2p, as other Pcls are not out abolishing growth at elevated temperature, for example, in able to perform this function (in the case of the highly Pcl2p- on October 26, 2015 by University of Queensland Library the wild type or in vrp1⌬ mutants overexpressing Las17p (219), homologous Pcl9p, this may simply be due to insufficient ex- ⌬ vrp1 mutants expressing a fragment of Vrp1p (328), or mu- pression in late G1). Two-hybrid screens identified Rvs167p as tants deficient in capping protein Cap2p (cap2⌬) (3) or the a Pcl2p- and Pcl9p-associated protein and the interaction was actin-regulatory kinase Prk1p (prk1⌬) (387). Some mutations confirmed by coimmunoprecipitation of Rvs167p with Pcl2p in Ysl2p (ysl2-1) (296) and in actin (act1) (15) also severely from lysates and binding of recombinant Pcl2p and Rvs167p in perturb actin patch polarization without resulting in tempera- vitro. All Pcl1p/2p subfamily Pcls bind strongly to Rvs167p, but ture-sensitive growth. Pho80p subfamily Pcls bind Rvs167p only weakly (163). It is possible that alterations in cellular lipid composition Do pho85, pcl2,orpcl9 mutants exhibit RvsϪ phenotypes? induce growth defects in rvs mutants and suppression of growth As the Pcl1p/2p subfamily Pcls exhibit functional redundancy, defects in sur mutants via mechanisms unrelated to their effects a strain with deletions of all five genes encoding Pcl1p/2p on actin patch distribution. It is interesting that the sphingoid subfamily Pcls (referred to in reference 163 as quint⌬) was base component of sphingolipids can cause cell damage and constructed. Both quint⌬ and loss of Pho85p (pho85⌬) resulted induce apoptosis in mammalian cells (111, 112, 228). In S. cere- in essentially the full range of RvsϪ phenotypes. Consistent visiae the sphingoid base phytosphingosine regulates growth with a role for Pho85p/Pcl2p in regulation of Rvs167p, Pho85p/ via ceramide-activated protein phosphatase (CAPP) (75, 224). Pcl2p phosphorylates Rvs167p in vitro and phosphorylation of Lipids other than sphingolipids can also affect growth and Rvs167p in vivo is partially dependent on Pcl1p/Pcl2p subfam- viability. A recent study found that endogenous free fatty acids ily Pcls. Rvs167p is phosphorylated only in G1 phase, consistent and diacylglycerol induce apoptosis upon entry into stationary with in vivo phosphorylation by Pho85p in association with phase in S. pombe. This apoptosis pathway is caspase indepen- Pcl1p and Pcl2p (which are expressed specifically in G1). There dent, i.e., independent of the yeast caspase-like protease (388). are three major phosphorylation sites in vivo and all are lo- Finally, protein N-myristoylation is known to be essential for cated in the GPA-rich domain, S299, S379, and S321. When survival under starvation conditions in S. cerevisiae and re- S321 is mutated T323 is phosphorylated instead, so T323 is a quires the medium-chain fatty acid myristate (7). Perhaps lev- potential phosphorylation site but is not normally phosphory- els of myristate vary depending on the level of very long chain lated. Phosphorylation of S299 and S379 in vivo is partially fatty acids. The contribution of lipid-mediated signaling to the dependent on Pcl1p/Pcl2p, but phosphorylation of S321 is not various defects in rvs mutant cells is an area that has not yet (85, 163). been fully explored. Rvs167p is hyperphosphorylated upon exposure of cells to mating pheromone, suggesting a role for Rvs167p phosphory- Regulation of Rvs167p by Phosphorylation lation also in the mating response. Hyperphosphorylation re- quires Pho85p and specifically Pcl2p (whose expression is in- Insight into how Rvs proteins function to link nutrient avail- duced by mating pheromone). Hyperphosphorylation also ability to G1-S cell cycle progression came from the discovery requires the mitogen-activated protein kinase Fus3p, which is that Rvs167p is regulated by the kinase Pho85p (163). Pho85p induced by mating pheromone and functions in the mating is a yeast cyclin-dependent kinase (Cdk) like the major cell pheromone response pathway. In vitro, Fus3p phosphorylates VOL. 70, 2006 BAR DOMAIN PROTEINS 73 the same sites in the Rvs167p GPA-rich domain as Pho85p/ with the possibility that Rvs167p is ubiquitinated by Rsp5p, Pcl2p (i.e., S299, S321, S370, and, when S321 is missing, T323) mass spectrometry analysis of the yeast proteome detected (85). Hyperphosphorylation of Rvs167p by Fus3p supports a ubiquitin attached to K481 in the Rvs167p SH3 domain role for Rvs167p in mating, as proposed previously (11, 20). (122, 236). Is phosphorylation of Rvs167p important for its function? The Rvs167p-Rsp5p association is direct and mediated by An Rvs167p mutant protein in which S299, S321, T323, and the Rsp5p WW domains and the Rvs167p GPA-rich domain. S379 were all replaced with alanine (Rvs167-4A) did not result WW domains bind motifs of the consensus PPXY and the in RvsϪ phenotypes. This was perhaps not unexpected since Rvs167p GPA-rich domain includes one motif (PPAY) that Downloaded from these phosphorylation sites all lie within the GPA-rich domain, matches exactly and two motifs (PSY and PQY) that match which has been shown previously to be nonessential for known less closely. All three motifs contribute to the interaction with Rvs167p functions (38, 299). Under normal conditions, Rsp5p. The Sla1p-Rsp5p association is also direct and medi- rvs167-4A was also not lethal in combination with any muta- ated by residues 420 to 720 of Sla1p, which contains Sla1p ⌬ tions that are known to be lethal in combination with rvs167 Homology Domain 1 (SHD1) and a coiled-coil domain. It has (176). However, under conditions of high temperature and in ϩ been proposed that the Sla1p-Rvs167p complex may be the the presence of Na rvs167-4A was lethal in combination with yeast equivalent of the vertebrate CIN8-endophilin complex. sla1⌬. Moreover, testing of the set of systematic deletions

While Rvs167p is often compared to vertebrate amphiphysin, it http://mmbr.asm.org/ covering each nonessential S. cerevisiae open reading frame is also related to vertebrate endophilin, which has a domain ⌬ revealed that a mutant carrying the novel combination end3 structure similar to that of amphiphysin but was discovered ⌬ ⌬ rvs167 is inviable. Furthermore, end3 is also lethal in com- more recently (Fig. 1). Sla1p exhibits sequence homology to bination with rvs167-4A. Rvs167p phosphorylation may there- CIN8 and the two proteins have a similar domain structure. fore be important in the absence of Sla1p and essential in the Moreover, like the yeast Sla1p-Rvs167p complex, the verte- absence of End3p (85). brate CIN8-endophilin complex interacts with the ubiquitin End3p and Sla1p form a complex with the Arp2/3p complex- ligase Rsp5p (known as Nedd4 in vertebrates) and functions in activating protein Pan1p (62, 322). Activation of the Arp2/3p endocytosis (240, 305, 307, 313). complex by Pan1p may be essential in the absence of Las17p,

The ubiquitin ligase Rsp5p is responsible for Rvs167p mo- on October 26, 2015 by University of Queensland Library as loss of both Pan1p and Las17p results in lethality (62). noubiquitination, since an rsp5 mutation results in loss of the Given the deleterious interactions between rvs167-4A and mu- modified form of Rvs167p. Moreover, Rsp5p binding to tations affecting the Sla1p-End3p-Pan1p complex, loss of phos- Rvs167p is also important for monoubiquitination, since com- phorylation of Rvs167p may perturb Las17p function. This would result in dependency on the Sla1p-End3p-Pan1p com- bined mutation of the PPXY, PSY, and PQY motifs in the plex for Arp2/3p activation. Indeed, the Rvs167p SH3 domain Rvs167p GPA-rich domain or combined mutation of the three binds Las17p in vitro (38, 181). Binding of GPA-SH3 to Las17p Rsp5p WW domains not only reduce Rvs167p-Rsp5p associa- (and also Ymr192p/Gyl1p) in vitro is considerably reduced by tion but also reduce the level of monoubiquitinated Rvs167p in phosphorylation of the GPA-rich domain by Pho85p/Pcl2p. vivo. Interestingly, although the GPA-rich domain mediates Introduction of a charge via phosphorylation into the un- binding of Rvs167p to Rsp5p, it is not the site at which ubiq- charged GPA-rich domain may induce major conformational uitin is attached (indeed, the GPA-rich domain contains no changes in the GPA-rich domain that affect the neighboring lysine residues and hence cannot be ubiquitinated). K481 in SH3 domain and its ability to bind Las17p and other proteins. the SH3 domain is the only site ubiquitinated in Rvs167p. In this case Rvs167-4A (which cannot be phosphorylated) may Hence, binding of Rsp5p to the Rvs167p GPA-rich domain bind Las17p too strongly and perhaps inhibit Las17p-depen- induces monoubiquitination of K481 in the Rvs167p SH3 do- dent Arp2/3p activation, causing dependency on Sla1p-End3p- main (307). Pan1p (85). Is monoubiquitination of Rvs167p important for in vivo function? As discussed above, neither the GPA-rich nor SH3 domain is required for any of the known biological roles of Regulation of Rvs167p by Ubiquitination Rvs167p (38). Consistent with this, mutated forms of Rvs167p The role of ubiquitination in endocytosis extends beyond lacking PPXY and PXY motifs in the GPA-rich domain, with ubiquitination of the receptor tail by the ubiquitin protein a K481R substitution in the SH3 domain (to prevent ubiquitin ligase Rsp5p. Ubiquitination of the endocytic machinery may attachment), or lacking the SH3 domain entirely were fully func- also be essential for endocytosis (64). Genetic interactions tional for receptor-mediated endocytosis (307). An Rvs167p between Rsp5p and a number of actin patch proteins required mutant construct lacking the SH3 domain was previously found for endocytosis support an additional role for Rsp5p in regu- to be only partially functional for growth in the presence of ϩ lation of actin patches and the endocytic machinery (146). Na , possibly due to lowered expression (38, 299). Similarly, Substrates ubiquitinated by Rsp5p often associate with Rsp5p the mutant Rvs167p construct with a K481R substitution in the and can be identified by protein interaction screens. SH3 domain was also only partially functional for growth in the ϩ In a high-throughput mass spectrometry analysis of the presence of Na . However, this requirement is for K481 only, yeast proteome, a protein complex containing Rsp5p and not ubiquitination of K481. An Rvs167p mutant construct lack- Rvs167p was identified (123). Rvs167p also associates with ing a PPXY motif does not bind Rsp5p and is not ubiquitinated Sla1p (59, 123) and both Rvs167p and Sla1p immunopre- on K481, but nevertheless is fully functional for growth in the cipitate with Rsp5p from yeast lysate, showing Rvs167p, presence of Naϩ. Hence, K481, but not ubiquitination of K481, is Sla1p, and Rsp5p form a complex in vivo (307). Consistent important for some functions of Rvs167p (307). 74 REN ET AL. MICROBIOL.MOL.BIOL.REV.

Genomic and Proteomic Approaches and the Diverse the yeast proteome with the potential to bind to certain SH3 Network of Rvs Interactions domains, including the Rvs167p SH3 domain. Peptides that matched a relaxed consensus as deduced by phage display Large-scale two-hybrid screens for Rvs161p- and Rvs167p- experiments were identified. Then all these peptides were syn- interacting proteins. The first large-scale screen to identify all thesized at high density on a cellulose membrane and probed Rvs161p- and Rvs167p-interacting proteins was reported by with the SH3 domains. The SH3 domains were grouped by this Bon et al. Full-length RVS161 and RVS167 genes in both cen- approach into five classes with partially overlapping specificity. tromeric (low copy) and multicopy plasmids were used as bait.

This approach, named the whole interactome scanning exper- Downloaded from Each prey sequence isolated by either Rvs161p or Rvs167p was iment, permits identification of the partners of any peptide then tested against the other bait. Finally, selected preys were recognition module by peptide scanning of a proteome. The tested against each individual domain of both proteins. Nu- whole interactome scanning experiment found 24 potential merous potential interacting proteins were identified (Table partners for the Rvs167p SH3 domain (160) (Table 1). 1). Interestingly, five preys interacted with Rvs161p and Identification of Rvs161p- and Rvs167p-interacting proteins Rvs167p, specifically Las17p, Rvs167p itself, App1p, Gyp5p, by high-throughput proteomics. In the quest for comprehen- and Gyl1p (18). sive analysis of protein-protein interactions, high-throughput Subsequently, more global analyses of protein-protein inter- actions among all proteins encoded by the S. cerevisiae genome mass spectrometric analyses leading to protein complex iden- http://mmbr.asm.org/ were performed by Uetz et al. and by Ito et al. Uetz et al. used tification have been developed as a powerful proteomic ap- a genomewide set (6,000) of full-length genes as baits and proach. Selected yeast proteins were chosen as baits, epitope identified two interactors for Rvs161p and 10 for Rvs167p. Ito tagged, transiently overexpressed, and used for immunopre- et al. also used a genomewide set of 6,000 full-length genes as cipitation of protein complexes from cell extracts. Proteins in baits and preys and tested each combination systematically. each purified complex were resolved by polyacrylamide gel The interacting proteins identified in this screen overlapped electrophoresis, stained, excised from gels, and trypsin di- only slightly with those obtained by Uetz et al., however, only gested before mass spectrometric analysis. With 725 bait pro- three interactions for Rvs161p and five for Rvs167p were iden- teins, 3,617 associated proteins covering 25% of the yeast pro- teome were detected, including four Rvs161p and 27 Rvs167p tified in this screen (Table 1) (136, 339). on October 26, 2015 by University of Queensland Library A third study by Drees et al. aimed to produce a protein interactors (Table 1) (123). interaction map for proteins implicated in cell polarity devel- Mapping the genetic interactions of Rvs161p and Rvs167p opment; 68 selected baits relevant to cell polarity were chosen by synthetic genetic array. To identify functional relationships and used to screen an array of preys representing about 90% of between genes, a method for systematic construction of double the S. cerevisiae genome. A network of interactions was re- mutants, termed synthetic genetic array analysis, was devel- vealed that links signaling proteins, the actin cytoskeleton, and oped for S. cerevisiae. A haploid carrying a query mutation organelles to polarity cues. This network included Rvs161p and (conditional or deletion) linked to a dominant selectable Rvs167p homo- and heterointeractions as well as Rvs167p marker and expressed from a mating type-specific promoter is interactions with 14 other proteins (59) (Table 1). crossed to an array of approximately 4,700 haploid open read- Identification of Rvs167p SH3 domain-interacting proteins ing frame deletion mutants (nonessential genes) of the oppo- by phage display. Phage display is a technique in which librar- site mating type, where each open reading frame deletion is ies of ligand peptides are displayed on the surface of bacterio- marked with a drug resistance gene. The diploid progeny are phages and screened for the ability to interact with an immo- then sporulated to yield recombinant meiotic haploid progeny bilized receptor the bait). In phage display, sequencing the and those recombinant progeny expressing both deletion bacteriophage DNA is used to identify those proteins that markers (double mutants) are selected. The inviability or sick- confer interaction with the bait. This technique has the draw- ness of these meiotic double mutant progeny represents neg- back of low accuracy (false-positives). One study combined ative genetic interactions that may suggest the affected genes computational prediction of interactions using phage display are functionally related. Synthetic genetic array analysis of 132 ligand consensus sequences with large-scale two-hybrid inter- queries allowed the construction of a network of approximately action analysis to identify interactions involving yeast SH3 do- 1,000 genes and 4,000 functional interactions. Of these, 47 mains, including the Rvs167p SH3 domain. Phage display was involved Rvs161p and 51 involved Rvs167p (Table 1) (332). A used to define the binding consensus motifs of each SH3 do- more recent SGA analysis focusing specifically on Rvs161p and main (which were analyzed computationally by a position-spe- Rvs167p found 49 lethal combinations for each gene. Strik- cific scoring matrix) to identify potential binding partners for ingly, the genetic interactors identified for each Rvs protein each SH3 domain. Then the SH3 domain baits were screened were identical (84a). against yeast two-hybrid libraries and an ordered array of in- Rvs161p and Rvs167p physical and genetic interactions sug- teracting open reading frames was established. A phage display gest multiple functions in vivo. Interactions found by high- network containing 394 interactions among 206 proteins and a throughput approaches should be regarded as putative until two-hybrid network containing 233 interactions among 145 pro- further characterized, however, these studies have provided teins were established. Graph theoretical analysis identified 59 the first global view of the cellular processes in which Rvs161p highly likely interactions common to both networks. This revealed and Rvs167p are (or may be) implicated. What is interesting nine potential partners for the Rvs167p SH3 domain (331). is that the Rvs161p and Rvs167p interactors represent a Another study combined phage display and synthesis of pep- diversity of both organelles and structures as well as cellular tides on cellulose membranes to search for all the peptides in processes (Fig. 9). Some Rvs161p and Rvs167p physical VOL. 70, 2006 BAR DOMAIN PROTEINS 75

gated than wild-type cells and tend to form clumps, suggesting a defect in cytokinesis or cell separation. Wild-type fission yeast cells, unlike budding yeast cells, are cylindrical. A small percentage of hob3⌬ cells are spherical rather than cylindrical, indicating that in some Hob3p-deficient cells there is a loss of cell polarity. These morphological defects do not worsen upon a shift in temperature, osmotic strength, or nutrient availabil- ity, and loss of Hob3p does not affect the rate of cell prolifer- Downloaded from ation at high (37°C) or low (21°C) temperature or upon nitro- gen starvation. Furthermore, overexpression of Hob3p does not inhibit the growth of wild-type S. pombe cells (276). Hob3p functions in polarization of the cortical actin cy- toskeleton but not endocytosis. Loss of Hob3p perturbs the FIG. 9. Functional classification of Rvs161p and Rvs167p-interact- ing proteins. Pie chart depicting the proportions of the total set of actin cytoskeleton. Actin patches are not polarized to the grow- known Rvs161p- and Rvs167p-interacting proteins (including both ing cell tips in interphase hob3⌬ cells as is the case in wild-type physical and genetic) which belong to each functional category. cells but are often randomly distributed along the length of the cell. During cytokinesis, actin patches are not efficiently polar- http://mmbr.asm.org/ ized to the medial site of division in hob3⌬. Actin patches are interactors are cytoplasmic, while others are nuclear. Some often distributed randomly in the daughter cell, mother cell, or interactors are components of cortical actin patches, while both. Loss of actin cytoskeleton polarization often causes de- others are not. fects in cytokinesis and accumulation of multinucleate cells. One interesting finding from the large-scale two-hybrid and Indeed, hob3⌬ cells tend to become multinucleate, consistent peptide-scanning screens was the potential interaction of with an impairment of cytokinesis. Unlike S. cerevisiae rvs161 Rvs167p with Bsp1p (59, 160, 331). Bsp1p is a protein that mutants, S. pombe hob3⌬ cells endocytose the membrane-sol- interacts with two of the three yeast orthologs of vertebrate uble fluorescent dye FM4-64 with the same kinetics as wild-

synaptojanin (Inp52p/Sjl2p and Inp53p/Sjl3p). As will be dis- on October 26, 2015 by University of Queensland Library type cells, and the appearance of the endocytic compartments cussed further below, in vertebrates the function of BAR do- is normal. Moreover, hob3⌬ mutants mate with the same effi- main proteins is closely linked to the function of synaptojanin. ciency as wild-type cells and thus lack the cell-cell fusion defect In yeast, Bsp1p has been implicated in the recruitment of of S. cerevisiae rvs161 mutants (276). Inp52p/Sjl2p and Inp53p/Sjl3p to cortical actin patches (366). Functional conservation between Hob3p, Rvs161p, and human The diversity of potential Rvs161p and Rvs167p interactors Bin3. Hob3p function is fully conserved in Bin3 (see below), but is striking. Even among those interactors known to function in only partially conserved in Rvs161p. The cell morphology and membrane traffic, some function in endocytosis and others in actin patch polarization defects of hob3⌬ mutants are fully res- exocytosis. Many Rvs161p- and Rvs167p-interacting proteins cued by ectopic expression of the human Bin3 protein. In con- function in nucleocytoplasmic transport, some in translocation trast, ectopic expression of S. cerevisiae Rvs161p restored normal of secreted proteins into the ER, and others in mitochondrial cell morphology to hob3⌬ mutants but only partially restored the import. Some interactors are involved in transcription, others actin cytoskeleton. Rvs161p restored F-actin localization to the in translation, and still others in degradation of proteins and cell middle during cell division, but there was no significant im- mRNAs. Several are cell cycle regulators. Many are compo- provement in actin patch polarization to cell tips in interphase nents of the cytoskeleton. However, the largest number of cells. In contrast, expression of S. cerevisiae Rvs167p did not Rvs161p and Rvs167p interactors are metabolic enzymes. rescue any of the defects observed in hob3⌬ mutants. Interest- Hence, Rvs161p and Rvs167p should not be thought of as strictly ingly, ectopic expression of Hob3p in S. cerevisiae rvs161⌬ mu- endocytic proteins or as strictly actin patch components. They tants rescued growth in the presence of Naϩ, but ectopic expres- perhaps should be thought of as proteins that perform a basic sion of human Bin3 (or Bin1) in rvs161⌬ cells did not (276). This molecular function that is so important and fundamental that it shows that Rvs161p function is still conserved in Hob3p, but that has been utilized again and again by multiple cellular structures human Bin3 has diverged during evolution. for the provision of multiple cellular functions.

Hob1p, Fission Yeast Ortholog of Rvs167p FISSION YEAST BAR DOMAIN PROTEINS Hob1p AND Hob3p Hob1p has a domain structure similar to that of Rvs167p. An S. pombe gene encoding a protein with homology to S. Hob3p, Fission Yeast Ortholog of Rvs161p cerevisiae Rvs167p and mammalian Bin1 was identified by Hob3p has a domain structure similar to that of Rvs161p. A homology search. This gene was named homolog of Bin1 fission yeast (Schizosaccharomyces pombe) gene encoding a (hob1ϩ). Hob1p has a domain structure similar to Rvs167p protein that comprises a BAR domain only and is homologous with an N-terminal BAR domain, a central domain, and a to S. cerevisiae Rvs161p was identified by homology search with C-terminal SH3 domain. The BAR domain of Hob1p exhibits the Rvs161p sequence and named homologue of bin3 (hob3ϩ). 51% homology to the BAR domain of Rvs167p and only 27 to Hob3p exhibits 56 and 29% sequence identity to S. cerevisiae 28% homology to the BAR domains of mammalian amphiphy- Rvs161p and human Bin3 (see below), respectively. Hob3p is sin 1 and Bin1 (277). not essential, but Hob3p-deficient cells (hob3⌬) are more elon- Hob1p is not required for polarization of the cortical actin 76 REN ET AL. MICROBIOL.MOL.BIOL.REV. cytoskeleton or endocytosis. Hob1p is not an essential protein patches, but causes depolarization of Hob1p patches. Repres- and Hob1p-deficient cells (hob1⌬) exhibit relatively normal sion of Nak1p expression has mild effects on growth in the morphology, but are slightly more elongated than wild-type S. presence of Naϩ, but combination with loss of Hob1p causes pombe cells. Hob1p-deficient cells exhibit no defect in cell severe growth sensitivity to Naϩ, suggesting that Nak1p and proliferation or in mating and retain a fully polarized actin Hob1p have some redundant functions. Consistent with this, cytoskeleton. Endocytosis of the membrane-soluble fluores- overexpression of Hob1p or of the Hob1p BAR domain alone cent dye FM4-64 is also unaffected by loss of Hob1p (or even partially restores growth and normal elongated cell morphol- by loss of both Hob1p and Hob3p). Hob1p-deficient cells also ogy to cells in which Nak1p expression is repressed. In contrast, Downloaded from remain fully viable upon nitrogen starvation, at a range of overexpression of the Hob1p central and SH3 domains did not temperatures, and in the presence of 250 mM Naϩ. The ex- restore normal cell morphology to Nak1p-deficient cells, in- pression of S. pombe Hob1p in S. cerevisiae rvs167⌬ cells did deed, it induced even more severe growth defects, perhaps by not restore growth on medium containing Naϩ. This sug- interfering with endogenous Hob1p function (131). gests that Hob1p function has diverged from that of These results strongly implicate Hob1p in organization of Rvs167p during evolution. Overexpression of Hob1p weakly the actin cytoskeleton, in growth under stress conditions, and inhibits growth but does not affect cell morphology. In con- in cell polarity. It seems that the lack of effect of hob1⌬ on cell trast, overexpression of the Hob1p BAR domain alone does polarity, growth in the presence of Naϩ, or actin cytoskeleton not inhibit growth, while overexpression of the central and organization is due to functional redundancy with both Wsp1p http://mmbr.asm.org/ SH3 domains is inhibitory for growth only when combined and Nak1p. Hence, Hob1p plays a role in cell polarity, actin with exposure to high Naϩ (131, 277). cytoskeleton organization, and growth under stress conditions Hob1p interacts with actin assembly proteins and protein much like Rvs167p in budding yeast, but its role is more re- kinases that regulate cell polarity. Hob1p localizes to cortical dundant with Wsp1p/Las17p family proteins in fission yeast actin patches whose distribution is polarized to sites of surface than in budding yeast, and this obscures the hob1 phenotypes growth at cell tips (equivalent to buds in budding yeast) and in fission yeast. the medial region in cells undergoing division (equivalent to Hob1p functions in regulating cell cycle progression in re- the bud neck in budding yeast). This localization is similar to sponse to nutrient availability. hob1⌬ cells have a defect in cell that of Rvs167p in budding yeast. Unlike Rvs167p, however, cycle progression under nitrogen-limiting conditions. Unlike S. on October 26, 2015 by University of Queensland Library localization of Hob1p to patches is abolished by disassembly of cerevisiae cells, which spend most of their cell cycle in G1 phase F-actin by treatment with the actin polymerization inhibitor prior to DNA replication, S. pombe cells spend most of their latrunculin A (131). cell cycle in G2 phase after DNA replication. Hence, S. pombe Hob1p interacts with proteins important for actin filament cells normally exhibit a 2C DNA content. Upon nitrogen star- assembly, such as the fission yeast ortholog of budding yeast vation, however, S. pombe cells continue to divide and at each Las17p, Wsp1p. Interaction requires the Hob1p SH3 domain, division exhibit a smaller cell size than during the previous as in the case of Rvs167p-Las17p interaction in budding yeast, division. Eventually, the starved cells reach a minimum size at but differs in that it also requires the BAR domain. Wsp1p is which they are unable to satisfy the cell size requirement to essential for Hob1p localization to cortical patches since enter the next S phase. The starved cells therefore accumulate

Hob1p adopts a diffuse cytoplasmic distribution when Wsp1p in G1 with a 1C DNA content and eventually enter G0.In expression is repressed. Overexpression of Hob1p causes contrast to wild-type cells, hob1⌬ cells starved of nitrogen Wsp1p to redistribute from patches to the cytoplasm, and this maintain a 2C DNA content. The hob1⌬ cells do arrest and effect requires the Hob1p SH3 domain that interacts with presumably enter G0, but do so directly from G2. Despite their Wsp1p. Overexpression of Hob1p also reduces the ability of 2C DNA content, hob1⌬ cells remain viable even after pro- cell extracts to support Wsp1p-dependent actin filament as- longed periods of nitrogen starvation (277). sembly in vitro, suggesting that Hob1p may inhibit Wsp1p. Repression or complete loss of Wsp1p alone has mild effects Functional Conservation of Hob1p and Bin1 on cell polarity and growth, but repression of Wsp1p expres- and Divergence of Rvs167p sion in hob1⌬ mutants severely inhibits growth and induces aberrant round-cell morphology. Overexpression of Wsp1p Ectopic expression in hob1⌬ cells of the ubiquitous human also inhibits growth and causes cells to become somewhat Bin1 isoform that lacks exon 10, Bin1Ϫ10Ϫ12ϩ13 (SH3p9), roundish, and these defects are corrected by additional over- but not human amphiphysin 1 or Bin2, restored the ability to expression of Hob1p. Hence, Hob1p and Wsp1p appear to arrest in G1 and enter G0 with a 1C DNA content upon nitro- have multiple functions in growth and polarity, and some of gen starvation. This suggests that Hob1p function is most con- these are probably synergistic, while others are antagonistic served with human Bin1Ϫ10Ϫ12ϩ13/SH3p9. Ectopic expres- (131). sion of human Bin1Ϫ10Ϫ12ϩ13/SH3p9 in rvs167⌬ S. cerevisiae Hob1p also interacts with the protein kinase Nak1p. Nak1p cells did not rescue growth in the presence of Naϩ, again plays an essential role in polarized growth in fission yeast and suggesting that Rvs167p function has diverged from that of is a member of the germinal center kinase-like kinases that Bin1Ϫ10Ϫ12ϩ13/SH3p9 and Hob1p. Intriguingly, overexpres- form a subgroup of the p21-activated kinase family of kinases sion of hob3ϩ in hob1⌬ cells efficiently rescued the cell cycle that have a conserved role in cell polarity. The Hob1p BAR arrest defect. Hence, it has been proposed that Hob3p func- domain mediates interaction of Hob1p with Nak1p. Repres- tions in the same signaling pathway as Hob1p, but downstream sion of Nak1p expression causes S. pombe cells to become of Hob1p (277). round. Loss of Nak1p does not affect Hob1p localization to Hob1p functions in regulation of cell cycle progression in VOL. 70, 2006 BAR DOMAIN PROTEINS 77 response to DNA damage. Loss of Hob1p also causes hyper- of dynamin 1 block receptor-mediated endocytosis of trans- sensitivity to both chemical (phleomycin)- and UV-induced ferrin and epidermal growth factor, but do not affect fluid- DNA damage, but not to inhibitors of DNA replication (e.g., phase endocytosis (45, 116, 121, 344). Upon a shift to the hydroxyurea). hob1⌬ cells arrest cell cycle progression nor- restrictive temperature, Drosophila melanogaster cells carrying mally in response to DNA damage (so the checkpoint is still a temperature-sensitive mutation of dynamin (shibire) become functional), but presumably the ability of the mutant cells to paralyzed due to failure of nerve transmission. Examination of repair the DNA damage and then reenter the cell cycle is neurons from shibire mutant flies reveals the accumulation of impaired. Combination of hob1⌬ with rad3⌬ and chk1⌬ mu- deeply invaginated pits at the synaptic plasmalemma that ap- Downloaded from tations, which are also hypersensitive to DNA damage, re- pear to be arrested prior to fission (155, 184). In vitro treat- vealed that hob1⌬ rad3⌬ double mutants are considerably ment of permeabilized nerve terminals (synaptosomes) with more sensitive to phleomycin than either single mutant, i.e., nonhydrolyzable GTP analogs to inhibit dynamin 1 results Hob1p and Rad3p genetically interact. In contrast, hob1⌬ in accumulation of deeply invaginated clathrin-coated pits chk1⌬ double mutants are only as sensitive to phleomycin as blocked prior to fission. Moreover, electron-dense rings that the more sensitive single mutant (hob1⌬). It has been proposed contain dynamin 1 are visible at the necks of these deeply that Rad3p and Hob1p act in distinct pathways that each con- invaginated pits (121, 315). tribute to repair of DNA damage (277). http://mmbr.asm.org/ Hob1p functions in a cell stress signal transduction path- Discovery of Amphiphysin 1 in Chicken Brain way upstream of Hob3p. Unexpectedly, the lethality caused by DNA damage in hob1⌬ cells is dependent on hob3ϩ. hob1⌬ The function of amphiphysin 1 in the brain has been the hob3⌬ double mutants, unlike hob1⌬ single mutants, are not topic of an earlier review (368). hypersensitive to DNA damage. The mechanisms involved are Chicken amphiphysin 1 has a domain structure similar to still unclear, but the data seem consistent with a function for that of budding yeast Rvs167p. Amphiphysin 1 was originally Hob3p in cell cycle arrest downstream of Hob1p. Interestingly, identified by expression cloning in a large-scale immunoscreen the steady-state levels of both Hob1p and Hob3p increase for novel proteins of the chicken synaptic plasma membrane. dramatically upon exposure of cells to DNA damage. The level Amphiphysin 1 is a hydrophilic protein with the only hydro- of Hob1p increases within 15 min of phleomycin treatment, phobic segment at residues 478 to 499. Chicken amphiphysin 1 on October 26, 2015 by University of Queensland Library while the increase in Hob3p levels is slightly delayed and be- has a predicted size of 75 kDa, but its electrophoretic mobility comes apparent only after 30 min and the levels of both pro- suggests a protein of 115 kDa (175). Aberrant gel mobility is teins remain elevated for at least 24 h. The increase in Hob3p not specific to chicken amphiphysin and has been reported protein levels is absolutely dependent on Hob1p, as it is absent also for rat, mouse, and human amphiphysin 1 (47, 50), in hob1⌬ cells (277). Whether Rvs161p and Rvs167p levels Bin1ϩ10ϩ13 (278), and Bin1ϩ6aϩ12ϩ13 (amphiphysin 2) increase upon exposure of S. cerevisiae cells to DNA damage (23, 168, 256, 367) (Fig. 1). and if loss of Rvs proteins causes sensitivity to DNA damage The N terminus of amphiphysin 1 features a short amphi- has not been examined. pathic ␣-helix. This is followed by the BAR domain, which, as in the case of Rvs161p and Rvs167p, is also predicted to have a strongly ␣-helical secondary structure and has a high density VERTEBRATE BAR DOMAIN PROTEIN AMPHIPHYSIN 1 of charged residues (Fig. 1) (47, 175, 220). The BAR domain is Endocytic Recycling of Synaptic Vesicles followed by a highly basic region with a high content of glycine, proline, alanine, serine, and threonine residues, similar to the Nerve terminals are known to be sites highly active in en- Rvs167p GPA-rich region (Fig. 1). This region of amphiphysin docytosis. When synaptic vesicles fuse with the plasma mem- 1 is predicted to adopt extensive ␤-turn secondary structure. brane (plasmalemma) and release their cargo of neurotrans- The C-terminal half of amphiphysin 1 is highly acidic and mitters into the presynaptic cleft, synaptic vesicle membrane terminates, as in the case of Rvs167p, with a single SH3 do- material is rapidly recovered back into the cell via endocytosis. main (Fig. 1) (12, 175). These empty synaptic vesicles are then used for storage of Chicken amphiphysin 1 is expressed in neurons. A single newly synthesized neurotransmitters. Endocytosis of the syn- major chicken amphiphysin 1 transcript was shown to be most aptic vesicle membrane occurs at clathrin-coated pits. Clathrin abundant in chicken brain (both forebrain and cerebellum) assembles to form a coat on the cytoplasmic face of the plasma and also present in the adrenal gland. Initial studies reported membrane at subdomains where endocytosis occurs and this an absence of amphiphysin 1 transcripts in other tissues of assembly is driven by the plasma membrane-clathrin assembly chicken, including skeletal muscle, lung, testis, liver, spleen, complex AP-2. Clathrin assembly promotes curvature of the pancreas, and heart. Amphiphysin 1 protein is highly expressed plasma membrane, leading to the formation of an invagination in the chicken central nervous system, including forebrain, known as a clathrin-coated pit. Progressive invagination results cerebellum, hippocampus, olfactory bulb, and spinal cord. A in the formation of a deeply invaginated clathrin-coated pit cross-reacting protein in the rat is expressed in the brain and connected to the plasma membrane by a narrow neck of mem- adrenal gland but also present in the anterior and posterior brane (120, 121, 155, 283, 290, 293, 295, 315, 368). pituitary (175). A subsequent study showed a rat protein rec- Role of dynamin 1 in synaptic vesicle recycling. The GTPase ognized by antiserum to human amphiphysin 1 is also ex- dynamin 1 is required for the scission of the neck of deeply pressed in the testis (256). In the chicken forebrain, amphiphy- invaginated clathrin-coated pits to release free clathrin-coated sin 1 is present throughout the cortex and is highly enriched at vesicles into the cytoplasm. Dominant negative mutant forms punctate structures that represent presynaptic terminals. Elec- 78 REN ET AL. MICROBIOL.MOL.BIOL.REV. tron microscopy reveals that amphiphysin 1 immunoreactivity is apparent size of 128 kDa based on gel electrophoresis. Aber- concentrated around clusters of synaptic vesicles in the presynap- rant electrophoretic behavior was also observed for chicken tic cytoplasm, but is absent from postsynaptic neurons (175). amphiphysin 1. Human amphiphysin 1 has the same domain Minor pool of chicken amphiphysin 1 is tightly associated structure as chicken amphiphysin 1, with an N-terminal BAR with synaptic vesicles. During subcellular fractionation a small domain, a central alanine- and glutamate-rich domain contain- pool of amphiphysin 1 is present in purified chicken synaptic ing a proline-rich motif and hydrophobic region, and a C- vesicles, but it is also present in purified chicken synaptosomal terminal SH3 domain (Fig. 1). The autoimmune sera from SMS patients were predominantly directed against the am-

plasma membrane. The rat brain contains an antigen that Downloaded from cross-reacts with both chicken and human amphiphysin 1 an- phiphysin 1 SH3 domain (47, 50). Amphiphysin 1 does have tisera (see below). In subcellular fractionation of the rat brain, some tissue-specific splice variants, e.g., amphiphysin Ir, which amphiphysin 1 distributes partly in the soluble fraction and is retina specific (128), but few compared to Bin1 (Table 2) partly in the sedimentable fraction. Highly purified synaptic (see below). vesicles from rat and chicken also contain amphiphysin 1, but Human amphiphysin 1 has a domain structure similar to it is only a minor component. The cross-reacting rat protein that of yeast Rvs167p. Human amphiphysin 1 displays strong copurifies with rat synaptophysin-positive synaptic vesicles and, sequence similarity to Rvs161p and Rvs167p within the N- terminal amphipathic ␣-helix and BAR domain (domain A in

although peripheral, behaves almost like an integral mem- http://mmbr.asm.org/ brane component. As for the integral membrane protein syn- reference 47) (Fig. 1). Within this region, a stretch of 209 aptophysin, neither physiological salt (0.15 M KCl), high salt (1 residues in amphiphysin 1 shares 42% amino acid sequence M KCl), nor pH 3 buffer was able to efficiently strip the rat similarity and 25% identity with the corresponding 208 resi- amphiphysin 1-cross-reacting protein from the synaptic vesi- dues in yeast Rvs161p and 39% similarity and 22% identity cles, however, both proteins were solubilized from synaptic with the corresponding 223 residues in yeast Rvs167p. Am- vesicles by detergent (50, 80, 175, 256). phiphysin 1 also shares homology with Rvs167p within the Because only a small pool of the rat amphiphysin 1-cross- C-terminal SH3 domain (domain D in reference 47). Although reacting protein is associated with synaptic vesicles and there is the central insert domain of amphiphysin 1 (domains B and C a large cytoplasmic pool, it is not enriched in synaptic vesicles in reference 47) is not as conserved in Rvs167p, the Rvs167p on October 26, 2015 by University of Queensland Library during purification. Fractionation of crude rat synaptosome GPA-rich region and amphiphysin 1 central insert domain lysates reveals that approximately 75% of the amphiphysin both feature proline-rich motifs and a single hydrophobic 1-cross-reacting protein in the initial lysate remains in a high stretch (12, 47, 298). The latter are also conserved in chicken speed (260,000 ϫ g) supernatant, while only 25% pellets with amphiphysin 1 (175). Mammalian amphiphysin 1 is highly expressed in the brain membranes. When rat synaptosomal membranes are subjected but is also expressed in other tissues. Like chicken and rat to Triton X-114 extraction to separate hydrophilic peripheral amphiphysin 1, human amphiphysin 1 is highly expressed in membrane proteins (recovered in the aqueous phase) from the brain, where it localizes to punctate presynaptic termini hydrophobic integral membrane proteins (recovered in the (46, 175, 256). Human amphiphysin 1 and dynamin 1 colocalize detergent phase) the rat amphiphysin 1-cross-reacting protein extensively in the brain to presynaptic termini and both also is present exclusively in the aqueous phase. There is no appar- exhibit a diffuse cytoplasmic localization (46). Although highly ent difference in electrophoretic migration of the membrane- expressed in the brain, amphiphysin 1 transcripts and antigen associated or soluble form of the rat amphiphysin 1-cross- are also expressed in certain other tissues (e.g., testis, ovary, reacting protein, and the biochemical feature that distinguishes pituitary, pancreas, and adrenal gland) (50, 80, 256, 367). Am- these two forms is unknown. The name amphiphysin was cho- phiphysin 1 has also been reported to be overexpressed in sen to reflect the existence of both soluble and membrane- some breast tumors and breast cancer cell lines (77). associated pools of this protein (175). Amphiphysin 1 interacts with the endocytic proteins dy- namin 1 and synaptojanin 1. Consistent with the possibility Discovery of Human Amphiphysin 1 that amphiphysin 1 functions with dynamin 1 in brain, gel overlay experiments using recombinant fragments of am- Human amphiphysin 1 is the autoantigen in stiff-man syn- phiphysin 1 as probes revealed interaction between an SH3 drome with breast cancer. Subsequently, human amphiphysin domain-containing fragment of amphiphysin 1 and a brain 1 was discovered as the 128-kDa autoantigen in breast cancer protein with a size (100 kDa) characteristic of dynamin 1. patients with the rare neurological condition stiff-man syn- Affinity purification approaches using the amphiphysin 1 SH3 drome (SMS). SMS is characterized by severe muscle rigidity domain as affinity ligand confirmed that dynamin 1 is a major throughout the body and intermittent spasms and is caused by amphiphysin 1 SH3 binding protein in the brain. Dynamin 1 an autoimmune response against antigens in the central ner- and amphiphysin 1 also coimmunoprecipitate from brain ex- vous system. How breast cancer leads to an autoimmune re- tracts, suggesting they form complex in vivo (46). Binding of a sponse against amphiphysin 1 is not clear, but it has been recombinant glutathione S-transferase (GST) fusion protein proposed that elevated expression of amphiphysin 1 or a re- containing only the human amphiphysin 1 SH3 domain to lated protein in breast tumors may trigger an autoimmune endogenous dynamin 1 in rat brain extracts has also been reaction against amphiphysin 1. demonstrated (256). Cloning of the human amphiphysin 1 gene based on homol- Dynamin 1 is not the only major binding partner for the ogy to chicken amphiphysin 1 revealed that human amphiphy- amphiphysin 1 SH3 domain in the brain. A major 145-kDa sin 1 has a predicted size of 76 kDa, which is smaller than the protein was also affinity purified from brain extracts using the VOL. 70, 2006 BAR DOMAIN PROTEINS 79

TABLE 2. Nomenclature of various BAR domain proteins and their splice variantsa

Type Nameb (reference) Other name(s) Exon(s) missing Remarks

Different amphiphysin 1 Amphiphysin 1 (or I) Longer amphiphysin isoform expressed in isoforms isoform 1 (47) brain.

Amphiphysin 1 (or I) Shorter amphiphysin isoform expressed in isoform 2 (77) brain. Also detected in breast carcinoma. Downloaded from Amphiphysin 1 (or I) Retina-specific amphiphysin isoform, (326) contains two novel insertions.

Different BIN1-iso1 (245)1 Amphiphysin IIa (255); S11/R3-a; 10 Expressed in brain and concentrated in BIN1/amphiphysin 2/ BRAMP2 (168) nerve terminals. Localizes to the amphiphyin II/Amph2/ cytoplasm. BRAMP2/SH3p9/ALP1/ amphiphysin IIm BIN1-iso2 (245)1 Amphiphysin IId (255); S11/R3-b 10; 12a Expressed in brain. Localizes to the isoforms cytoplasm and nucleus.

BIN1-iso3 (245)1 Amphiphysin IIc1 (255) 10; 12 (a, b, c) Expressed in brain. Localizes to the cytoplasm and nucleus. http://mmbr.asm.org/

BIN1-iso4 (245)2 6a; 12 (b, c, d) Expressed in brain and possibly in muscle. Localizes to the cytoplasm and nucleus.

BIN1-iso5 (245)3 Amphiphysin IIb (255) 6a; 10; 12 (b, c) Expressed in brain. Localizes to the cytoplasm and nucleus.

BIN1-iso6 (245)3 AP II2 (338) 6a; 10; 12 (b, c, d) Expressed in brain. Localizes to the cytoplasm and nucleus. Aberrantly expressed in melanoma (93).

BIN1-iso7 (245)3 Amphiphysin IIc2 (255) 6a; 10; 12 (a, b, c) Expressed in brain. Localizes to the cytoplasm and nucleus. on October 26, 2015 by University of Queensland Library

BIN1-iso8 (245)2 Amph 2-7 (369) 6a; 12 (a, b, c, d) Expressed in muscle and localizes to the nucleus.

BIN1-iso9 (245)4 BIN1-10 (356); SH3P9 (306); 6a; 10; 12 (a, b, c, d) Ubiquitously expressed. AP II3 (338)

BIN1-iso10 (245)5 BIN1-10-13 (356); ALP1 (141); 6a; 10; 12 (a, b, c, d); 13 Ubiquitously expressed, the only Amph amphiphysin IIm (100); II/Bin1 isoform in macrophages. Amph 2-6 (369)

BIN1-iso11 Amph 2-1 (369) 10; 13 Expressed in brain.

BIN1-iso12 Amph 2-2 (369) 10; 12d; 13 Expressed in brain.

BIN1-iso13 Amph 2-3 (369) 10; 12 (a, b, c, d); 13; 14; Expressed in brain, lacks SH3 domain. 15; 16

BIN1-iso14 Amph 2-4 (369) 6a; 10; 11; 12 (a, b, c, d); Expressed in brain, lacks SH3 domain. 13; 14; 15; 16

BIN1-iso15 Amph 2-5 (369) 10; 12 (a, b, c, d); 13 Expressed in brain.

Endophilin family proteins Endophilin A1 (134) Endophilin 1 (323); SH3P4 (323); Expressed mainly in brain. SH3GL2 (99); EEN-B1 (304)

Endophilin A2 (134) Endophilin 2 (323); SH3P8 (323); Ubiquitously transcribed (270), but only SH3GL1 (99); EEN (304) translated in brain and testis (269).

Endophilin A3 (134) Endophilin 3 (323); SH3P13 (323); Expressed mainly in brain and testis. SH3GL3 (99); EEN-B2 (304)

Endophilin B1 (134) SH3GLB1 (244); Bif-1 (43) Expressed mainly in heart, placenta, and skeletal muscle, cytoplasmic protein.

Endophilin B2 (134) SH3GLB2 (244) Expressed mainly in heart, placenta, and skeletal muscle, cytoplasmic protein.

a This table adopts the Bin1 exon nomenclature proposed by Wechsler-Reya et al. (356); however, one additional missing exon located between exons 6 and 7is referred to here as exon 6a. In the recently introduced systematic nomenclature, exon 6a is renamed exon 7 and the following exons have been assigned new numbers (n ϩ 1) with respect to the exon numbers referred to here and in the previous literature reports. The new systematic Bin1 splice isoform nomenclature and some of the exon information shown in this table was obtained from the National Center for Biotechnology Information and the U.S. National Library of Medicine, Bethesda, Md., via the following internet URL: http://www.ncbi.nlm.nih.gov/entrez/query.fcgi?dbϭgene&cmdϭRetrieve&doptϭfull_report&list_uidsϭ274. b Superscript numbers are as follows: 1, Bin1ϩ6aϩ12ϩ13 in the text; 2, Bin1ϩ10ϩ13 in the text; 3, Bin1Ϫ6aϩ12ϩ13 in the text; 4, Bin1Ϫ10Ϫ12ϩ13 in the text; 5, Bin1Ϫ10Ϫ12Ϫ13 in the text. Note: Bin1ϩ12ϩ13 in the text refers collectively to Bin1Ϫ6aϩ12ϩ13 and Bin1ϩ6aϩ12ϩ13; Bin1ϩ12 in the text refers collectively to Bin1Ϫ6aϩ12ϩ13, Bin1ϩ6aϩ12ϩ13, Bin1Ϫ6aϩ12Ϫ13, and Bin1ϩ6aϩ12Ϫ13; Bin1ϩ10 in the text refers collectively to Bin1ϩ10Ϫ13 and Bin1ϩ10ϩ13. 80 REN ET AL. MICROBIOL.MOL.BIOL.REV. amphiphysin 1 SH3 domain (46). This protein was subse- Regulation of Amphiphysin 1 Interactions quently identified as a polyphosphoinositide 5-phosphatase by Phosphorylation and named synaptojanin 1. Amphiphysin 1 and synaptojanin 1 Amphiphysin 1, like dynamin 1 and synaptojanin 1, is con- extensively colocalize in brain sections to presynaptic termi- stitutively phosphorylated in resting neurons, but is rapidly nals. Like dynamin 1, synaptojanin 1 is implicated in synaptic dephosphorylated upon nerve terminal depolarization and syn- vesicle endocytosis via its regulation of phosphoinositide levels aptic vesicle exocytosis. Stimulated dephosphorylation affects (188). Synaptojanin 1 also is a major binding partner of the only certain residues in amphiphysin 1, because some basal SH3 domain of another BAR domain protein, endophilin A1 phosphorylation remains even after nerve depolarization. Downloaded from (also called SH3p4 and SH3GL2) (51, 192) (see below). Stimulated dephosphorylation is not dependent on exocytosis Intriguingly, the amphiphysin 1 SH3 domain is also able to of synaptic vesicles because it is not affected by prior treatment mediate an intramolecular interaction. This is predicted to of neurons with tetanus toxin to block synaptic vesicle exocy- involve the SH3 domain binding to highly conserved proline- tosis. Dephosphorylation is, however, dependent on extracel- rich sequences located within the central insert domain. In- lular Ca2ϩ. This suggests transmembrane Ca2ϩ ion flux may be tramolecular interactions of the amphiphysin 1 SH3 domain required (13, 189, 193, 222, 271, 302). may be in competition with intermolecular interactions and Stimulated dephosphorylation of amphiphysin 1 in nerve ϩ regulate amphiphysin 1 SH3 domain interactions with other terminals is also blocked by inhibitors of Ca2 /calmodulin- http://mmbr.asm.org/ proteins such as dynamin 1 and synaptojanin 1 (74). The three- dependent protein phosphatase 2B (calcineurin) (e.g., FK506 dimensional structure of the amphiphysin 1 SH3 domain has and cyclosporine A), but not by inhibitors of protein phospha- not yet been solved, but is predicted to be similar to that of the tase 1 or 2A (e.g., okadeic acid). This suggests a role for Bin1 SH3 domain. The Bin1 SH3 domain also interacts with calcineurin in stimulated dephosphorylation of amphiphysin 1. dynamin 1 and synaptojanin 1 (see below) and its three-dimen- Moreover, purified recombinant calcineurin (rendered consti- sional structure has been solved (232). tutively active by mutation) dephosphorylates amphiphysin 1 The amphiphysin 1 SH3 domain interacts with a single PSR- directly in vitro. Amphiphysin 1 can be induced to undergo PNR motif (residues 833 to 838) in the C-terminal proline-rich multiple rounds of phosphorylation and dephosphorylation in domain (PRD) of dynamin 1. Deletion of this motif or R835D/ synaptosomes in response to a series of induced nerve depolar- on October 26, 2015 by University of Queensland Library R838D/P836A substitutions within this motif abolish associa- izations separated by rest periods to restore polarization. This tion of a recombinant dynamin 1 PRD fragment with both suggests that cycles of constitutive phosphorylation and stimu- lated dephosphorylation are physiologically important for am- endogenous amphiphysin 1 in brain extracts and purified re- phiphysin 1 function in synaptic vesicle recycling in vivo (13). combinant amphiphysin 1 in overlay blots. The amphiphysin 1 Subcellular fractionation shows that while the sedimentable SH3 domain is critical for interaction with dynamin 1. The membrane-bound pool of amphiphysin 1 contains equal amounts double amino acid substitution G684R/P687L within the target of the fully phosphorylated and partially dephosphorylated forms, binding cleft of the amphiphysin 1 SH3 domain abolishes in- the cytoplasmic pool is predominantly dephosphorylated. Inter- teraction with both dynamin 1 and synaptojanin 1. The disso- estingly, this does not reflect the localization of the kinase that ciation constant for amphiphysin 1 SH3 domain interaction phosphorylates amphiphysin 1. Addition of ATP and incubation with the dynamin 1 PRD is approximately 190 nM, which is at 37°C resulted in phosphorylation of amphiphysin 1 present relatively high affinity for an interaction mediated by an SH3 in the cytosolic fraction but not amphiphysin 1 present in the domain. A similar motif, PXRPXR in the C-terminal PRD of membrane fraction. Interestingly, nonhydrolyzable GTP analogs synaptojanin 1, may mediate interaction of synaptojanin 1 with strongly stimulate amphiphysin 1 phosphorylation in the presence the amphiphysin 1 SH3 domain (46, 105, 302). of brain cytosol in vitro. Hence, amphiphysin 1 may be phosphory- lated by a kinase whose activity is regulated by a GTP-binding protein (13). Amphiphysin 1 Interacts with Clathrin and the AP-2 Adaptor Amphiphysin 1 Subcellular Localization Other domains of amphiphysin 1 also interact with proteins in Presynaptic Terminals implicated in synaptic vesicle endocytosis. The ␣ (and to a c Amphiphysin 1 localizes to endocytic intermediates. There lesser extent ␣ ) subunit of the plasma membrane clathrin a is a large cytoplasmic pool of amphiphysin 1 in rat brain (13, assembly complex AP-2 has been shown to associate with am- 50, 80, 175, 256). Permeabilization of rat brain synaptosomes ␣ phiphysin 1 (46, 352). Moreover, c adaptin and amphiphysin followed by treatment with a nonhydrolyzable GTP analog is ␣ ␣ 1 colocalize in the brain. However, although the a and c known to induce the formation of deeply invaginated clathrin- subunits of AP-2 bind SH3 domains of other proteins (e.g., the coated pits. Immunoelectron microscopy on synaptosomes SH2-SH3 adaptor protein Grb2) neither protein binds the treated in this way shows that amphiphysin 1 localizes to the amphiphysin 1 SH3 domain. The interaction of amphiphysin 1 accumulated clathrin-coated buds and also partially colocalizes ␣ ␣ with a and c is mediated by non-SH3 domains of amphiphy- with dynamin 1 at the necks of deeply invaginated clathrin- sin 1 (see below) (46). Interestingly, amphiphysin 1 (and Bin1, coated pits. Amphiphysin 1 also localized to the less abundant see below) also binds clathrin heavy chain. This interaction is clathrin-coated pits observed in untreated synaptosomes, also mediated by sequences in amphiphysin 1 that are distinct showing that localization of amphiphysin 1 to endocytic inter- from the SH3 domain (see below) (187, 301, 302). mediates is physiologically relevant and not simply an artifact VOL. 70, 2006 BAR DOMAIN PROTEINS 81 induced by nonhydrolyzable GTP analogs. Amphiphysin 1 is In contrast, there is no apparent effect of this treatment on the not present on all clathrin-coated pits and dynamin 1-coated microtubule cytoskeleton. It has been proposed that disorga- tubules, however, and its distribution on those clathrin-coated nization of the actin cytoskeleton may be responsible for the pits where it is found is irregular and patchy. This suggests that, block in neurite outgrowth (212). This loss of cell polarity in unlike clathrin or AP-2, amphiphysin 1 is not a major coat neurons is perhaps analogous to what is seen in yeast cells component (13). lacking Rvs161p or Rvs167p (12, 215, 298). No interaction of Amphiphysin 1 also localizes to the actin-rich cytomatrix. amphiphysin 1 with G-actin or F-actin has been observed, Amphiphysin 1 is also found in the actin-rich cortical cytoma- however, despite their colocalization in patches. In contrast, Downloaded from trix, which is distinct from endocytic intermediates but sur- yeast Rvs167p not only colocalizes with actin in patches, but rounds clusters of synaptic vesicles (13). This localization sug- also interacts genetically and in two-hybrid screens with actin gests amphiphysin 1 may have a role in the cortical actin (4, 11, 180, 212, 215). cytoskeleton, perhaps analogous to the role of Rvs167p in cortical actin patches in yeast (11). It is interesting that in Amphiphysin 1 Functions in Endocytosis Drosophila melanogaster, amphiphysin is expressed in a range of polarized cells (including epithelial cells), where it localizes Amphiphysin 1 interaction with dynamin 1 is essential for to the actin-rich apical plasma membrane domain (386). This endocytosis in neurons. The essential role of Rvs proteins in suggests a conserved role for amphiphysin family proteins in endocytosis in yeast together with the observed interactions of http://mmbr.asm.org/ organization of the cortical actin cytoskeleton. amphiphysin 1 with dynamin 1, clathrin, and AP-2 in neurons Amphiphysin 1 functions in polarized growth. The strongest suggested a possible role for amphiphysin 1 in synaptic vesicle evidence that amphiphysin 1 plays a role in organization of the endocytosis. The first evidence for a role of amphiphysin 1 in cortical actin cytoskeleton and cell polarity comes from a study endocytosis came from experiments in which a recombinant looking at neurite outgrowth in cultured hippocampal neurons. fusion protein containing the amphiphysin 1 SH3 domain was Hippocampal neurons plated at low cell density in culture form microinjected into the giant reticulospinal synapse of the lam- axonal and dendritic processes that exhibit dramatic polarized prey eel. Injection of the SH3 domain had no effect on the growth. As the cells approach confluence the axonal processes number of synaptic vesicles or plasmalemmal clathrin-coated eventually contact other cell bodies and form synapses in vitro. pits in resting neurons. However, electrical excitation of the on October 26, 2015 by University of Queensland Library When hippocampal neurons are initially plated, the amphiphy- injected neurons to induce synaptic vesicle exocytosis resulted sin 1 expression level is detectable but low, however, a dra- in persistent depletion of synaptic vesicle numbers and the matic increase in amphiphysin 1 expression accompanies neu- dramatic accumulation of deeply invaginated clathrin-coated rite outgrowth and synapse formation. Moreover, amphiphysin pits at the plasmalemma (Fig. 10). Injection of a synthetic 1 and dynamin 1 become highly concentrated at the tips of peptide corresponding to the motif in the dynamin 1 PRD that actively growing axons and dendrites and at high magnification binds the amphiphysin 1 SH3 domain had similar effects, while can be seen to localize to numerous patches containing F-actin a mutant amphiphysin 1 SH3 domain unable to bind dynamin distributed throughout the cortex of the growth cone. As 1 and an unrelated SH3 domain were without effect (293). The growth cones come into contact with other neurons in culture results obtained from such domain microinjection experiments and mature into synapses, amphiphysin 1 and dynamin 1 con- should be taken as suggestive of a cellular role rather than centrate at the synaptic termini, analogous to amphiphysin 1 proof, however, as the injected domain can have off-target and dynamin 1 localization in mature brain (212). effects, especially at these concentrations. Neurons derived from temperature-sensitive Drosophila dy- Amphiphysin 1 does not seem to play an essential role in namin mutants (shibire mutants) and cultured in vitro fail to exocytosis in neurons. Injection of the amphiphysin 1 SH3 form neurites upon a shift to the restrictive temperature (184). domain did not affect synaptic vesicle exocytosis and release of Also, down-regulation of dynamin 1 in cultured hippocampal neurotransmitters at low levels of nerve terminal stimulation. neurons prevents neurite outgrowth and causes growth cone However, at higher levels of stimulation, where the ability to collapse (333). Does amphiphysin 1 also play a role in neurite rapidly recycle synaptic membranes becomes more important, outgrowth and synaptogenesis? Consistent with a role for am- neurotransmitter release was affected (293). These data are phiphysin 1 in neurite outgrowth, down-regulation of am- consistent with the view that the amphiphysin 1 SH3 domain- phiphysin 1 expression using antisense oligonucleotides leads mediated interaction with a proline-rich motif in dynamin 1 is to a severe block in neurite outgrowth such that many cells fail essential for endocytosis but not for exocytosis of synaptic to form neurites and the cells that do form neurites exhibit only vesicles. This is consistent with the finding that Rvs161p and a few short neurites. Hence, loss of amphiphysin 1 causes Rvs167p are not essential for exocytosis in yeast but loss of collapse of growth cones on both dendrites and axons. Inter- these proteins does nevertheless perturb exocytosis, leading to estingly, this effect is reversible, as removal of the antisense vesicle accumulation (20, 298). oligonucleotides leads to reestablishment of growth cones and As dynamin 1 performs the scission event at the necks of recommencement of neurite outgrowth (212). invaginated clathrin-coated pits, it is possible that amphiphysin The defect in neurite outgrowth caused by amphiphysin 1 1-dynamin 1 interaction is important for dynamin 1 recruit- down-regulation is not due to effects on endocytosis, because ment to the necks of clathrin-coated pits. Consistent with this, the level of down-regulation achieved was insufficient to block the necks of the deeply invaginated clathrin-coated pits that receptor-mediated endocytosis. The cortical actin cytoskeleton accumulate in neurons injected with the amphiphysin 1 SH3 is grossly perturbed by amphiphysin 1 down-regulation, how- domain appear to lack the electron-dense rings characteristic ever, with loss of F-actin polarization to growing neurite tips. of dynamin 1 (121, 293). What was not clear was if this role of 82 REN ET AL. MICROBIOL.MOL.BIOL.REV. Downloaded from http://mmbr.asm.org/ on October 26, 2015 by University of Queensland Library

FIG. 10. Amphiphysin 1 SH3 domain interactions are required for fission of endocytic vesicles. Lamprey giant reticulospinal synapses were injected with a fusion protein comprising glutathione S-transferase and the human amphiphysin 1 SH3 domain (GST-amphSH3) and effects on synaptic vesicle recycling from the plasmalemma (plasma membrane) to regenerate the cytoplasmic vesicle pool were examined. Panel A, electron micrograph of a synapse in an axon injected with GST-amphSH3 and then maintained in low Ca2ϩ (0.1 mM Ca2ϩ,4mMMg2ϩ), without stimulation. Panel B, electron micrograph of a synapse in an axon injected with GST-amphSH3 and then electrically stimulated at 0.2 Hz for 30 VOL. 70, 2006 BAR DOMAIN PROTEINS 83 amphiphysin 1 in recruitment of dynamin 1 is specific to neu- Amphiphysin 1 interacts with clathrin and AP-2. Amphiphysin rons. A potential SH3 domain binding motif in the dynamin 1 1 also interacts with other components of clathrin coats. Am- PRD was shown to be essential for localization to clathrin- phiphysin 1 binds directly to clathrin heavy chain (187). Fur- coated pits when dynamin 1 was expressed ectopically in thermore, amphiphysin 1 binds the ␣-adaptin subunit of the COS-7 cells. However, this motif is distinct from the PSRPNR AP-2 adaptor in vitro (352). Subsequently, it was found that motif that has been shown to interact specifically with the the appendage domain of ␣-adaptin binds a non-SH3 domain amphiphysin 1 SH3 domain and likely mediates interaction of amphiphysin 1 (46). Binding of a recombinant central insert with another SH3 domain protein (292). domain of human amphiphysin 1 to endogenous clathrin heavy ␣ ␣ ␣ Downloaded from Amphiphysin interaction with dynamin is also essential for chain and -adaptin ( a and c) in rat brain extract was sub- endocytosis in nonneuronal cells. Amphiphysin 1 may have a sequently confirmed (256). more general function in recruiting dynamin family GTPases Does amphiphysin 1 act as an adaptor to link dynamin to to clathrin-coated pits during endocytosis in nonneuronal clathrin-coated pit components such as clathrin heavy chain cells. In support of this is the known role of Rvs proteins in and AP-2? Although an early report suggested that dynamin 1 endocytosis in yeast. Further support came from the finding binds directly to the ear domain of ␣-adaptin (352), subsequent that a 125-kDa isoform of amphiphysin is also expressed at studies found that the direct dynamin 1–␣-adaptin interaction lower but still significant levels in nonneuronal cells, includ- is very weak. When amphiphysin 1 and dynamin 1 are both ing muscle, testis, lung, and fibroblast. Moreover, the am- overexpressed in COS-7 cells ␣-adaptin forms complexes con- http://mmbr.asm.org/ phiphysin 1 SH3 domain interacts with a nonneuronal iso- taining both proteins but the ␣-adaptin appendage domain form of dynamin expressed endogenously in the fibroblast binds poorly to dynamin 1 compared to amphiphysin 1. The cell line COS-7 (369). ability of amphiphysin 1 to form heterotrimeric complexes with To investigate whether amphiphysin 1 functions in endocy- ␣-adaptin and dynamin 1 is consistent with the possibility that tosis in nonneuronal cells, the amphiphysin 1 SH3 domain was amphiphysin 1 associates with clathrin-coated pits via interac- transiently expressed in COS-7 cells and the effect on receptor- tion with ␣-adaptin and then recruits dynamin 1 to these pits mediated endocytosis was examined. Expression of the am- via its SH3 domain, a process that is likely to be perturbed by phiphysin 1 SH3 domain abrogated endocytic uptake of both overexpression of the isolated amphiphysin 1 SH3 domain. The transferrin and epidermal growth factor in COS-7 fibroblasts, effects of amphiphysin 1 SH3 domain overexpression on dy- on October 26, 2015 by University of Queensland Library but had no effect on fluid-phase uptake of fluorescein isothio- namin subcellular localization were not investigated directly, cyanate-labeled dextran. Overexpression of other SH3 do- however, subcellular fractionation revealed that dynamin lev- mains that bind to dynamin, but via motifs distinct from the els in the membrane fraction are reduced by overexpression of amphiphysin 1 SH3, did not inhibit transferrin endocytosis. the amphiphysin 1 SH3 domain (367, 369). The effect of the amphiphysin 1 SH3 domain on endocytosis is The clathrin heavy chain and ␣-adaptin binding sites on due to perturbation of dynamin, since transient overexpression amphiphysin 1 have been defined. Residues 262 to 405 of of dynamin neutralized the effect of the amphiphysin 1 SH3 human amphiphysin 1 mediate binding of both clathrin domain and restored endocytosis (369). Similarly, microinjec- heavy chain and ␣-adaptin. The sites that interact with clath- tion of the amphiphysin 1 SH3 domain into cultured 3T3-L1 rin heavy chain and ␣-adaptin are distinct, but adjacent. adipocytes inhibited receptor-mediated endocytosis of trans- Minimal ␣-adaptin binding requires residues 322 to 340, ferrin and caused accumulation of the Glut4 glucose trans- however, residues 340 to 363 are required for full-strength porter at the plasma membrane (349). These results are sug- interaction. The latter overlap with the clathrin heavy-chain gestive of a role for amphiphysin 1 in endocytosis, but the binding site and contain a DPF motif found in multiple overexpressed amphiphysin 1 SH3 domain may engage in off- copies in other ␣-adaptin binding proteins. Mutation of target interactions leading to nonspecific effects. residues 323 to 326 (FFED) abolishes ␣-adaptin binding As discussed above, there appears to be a role for am- without affecting clathrin heavy-chain binding. Binding to phiphyin 1 in regulation of the cortical actin cytoskeleton dur- clathrin heavy chain is conferred by residues 347 to 386 (256, ing polarized growth. Intriguing, links between dynamin and 301, 302). The three-dimensional structure of the ␣-adaptin the actin cytoskeleton are also emerging. Dynamin 2 in non- appendage domain has been determined and a model has neuronal cells modulates the activation of the Arp2/3 actin been proposed to account for its remarkable ability to me- filament nucleation complex and thereby de novo actin fila- diate binding not only to amphiphysin 1 but also to several ment assembly at the cortex (151, 157, 280, 281). Mechanical other proteins in clathrin coats via a single interaction in- force generated by de novo actin filament assembly may be terface (231, 334). important for dynamin-mediated scission of the neck of clath- Mutational analysis identified two motifs that are important rin-coated vesicles during endocytosis as well as in cortical for binding to clathrin heavy chain, LLDLD and WDLW. actin cytoskeletal dynamics during polarized growth. LLDLD fits the consensus for clathrin heavy chain binding,

min in high Ca2ϩ (2.6 mM Ca2ϩ and 1.8 mM Mg2ϩ) to induce exocytosis of synaptic vesicles. Panel C, electron micrograph of a synapse in an axon treated as in panel B showing the accumulation of invaginated endocytic pits at the plasmalemma. Panel D, electron micrograph of a synapse in an axon injected with GST-amphSH3 and then electrically stimulated at 5 Hz for 30 min in high Ca2ϩ (2.6 mM Ca2ϩ and 1.8 mM Mg2ϩ) to induce exocytosis of synaptic vesicles. Bars, 0.2 ␮m. The bar in panel A applies to panels A to C. (Reprinted from reference 293 with permission of the publisher. Copyright 1997 American Association for the Advancement of Science.) 84 REN ET AL. MICROBIOL.MOL.BIOL.REV.

L(L/I)(D/E/N)(L/F)(D/E), proposed based on analysis of 302). Again, however, these results from overexpression exper- clathrin binding sites in other clathrin coat components such as iments should be interpreted with caution due to the possibility the ␤-adaptin subunit of AP-2 and AP-180. The affinity of the of off-target interactions and nonspecific effects. recombinant clathrin-binding domain of amphiphysin 1 for Amphiphysin 1 interacts with endophilin A1/SH3p4/SH3GL2. native clathrin triskelia was estimated at 1 nM (255). Mutation As well as clathrin heavy chain and AP-2, the amphiphysin 1 of either LLDLD or WDLW motifs in amphiphysin 1 weakens central insert domain also binds the other synapse-enriched clathrin heavy-chain binding and the double mutation abol- BAR domain protein endophilin A1 (also called SH3p4 and ishes clathrin heavy-chain binding, but has no effect on ␣-adap- SH3GL2) (Fig. 1). Recombinant endophilin A1 binds endog- Downloaded from tin binding. Combination of the mutations in ␣-adaptin and enous amphiphysin 1 in rat brain extracts via its SH3 domain clathrin heavy-chain binding sites, as expected, abolish am- (the BAR domain does not interact). The recombinant central phiphysin 1 interaction with both ␣-adaptin and clathrin heavy insert domain of amphiphysin 1 binds endogenous endophilin chain. A1 in brain extracts and recombinant endophilin A1. These Amphiphysin 1 binds an N-terminal fragment of clathrin results show that the endophilin A1-amphiphysin 1 interaction heavy chain (residues 1 to 579 in rat clathrin heavy chain) in is direct and mediated by the endophilin A1 SH3 domain and vitro and this fragment contains the ␤-propeller “foot” or “ter- the amphiphysin 1 central insert domain. A similar interaction minal” domain of clathrin heavy chain. The foot domain is also has been reported between endophilin A1 and a neuronal the site in clathrin heavy chain that binds a range of other splice variant of Bin1 (Bin1ϩ6aϩ12ϩ13) (see below). The http://mmbr.asm.org/ clathrin-associated proteins, including the ␤-adaptin subunit of central insert domains of both amphiphysin 1 and Bin1ϩ AP-2, the adaptor AP-180, epsin, and ␤-arrestin. These pro- 6aϩ12ϩ13 include a proline-rich SH3 domain binding consen- teins all have motifs similar to the LLDLD and WDLW motifs sus motif that may mediate the interaction (193, 255). in amphiphysin 1 that mediate interaction with the foot do- Amphiphysin 1 forms homodimers. The recombinant am- main of clathrin heavy chain. In vitro, amphiphysin 1 competes phiphysin 1 BAR domain associates in overlay assays with with AP-2 and AP-180 for binding to clathrin heavy chain, full-length endogenous amphiphysin 1 in brain extract, suggest- suggesting all three proteins recognize the same site within the ing that amphiphysin 1 forms homodimers (302). Amphiphysin foot domain of clathrin heavy chain. As AP-2 plays a key role 1 also forms heterodimers with Bin1ϩ6aϩ12ϩ13 in the brain in clathrin assembly into plasma membrane clathrin-coated (see below). on October 26, 2015 by University of Queensland Library pits, competition with AP-2 for binding clathrin heavy chain Regulation of amphiphysin 1 complex formation by phos- may explain the loss of clathrin from plasma membrane-coated phorylation and dephosphorylation. Interestingly, amphiphy- pits observed upon overexpression of clathrin-binding frag- sin 1, dynamin 1, and synaptojanin 1 are phosphorylated in ments of amphiphysin 1 (57, 58, 187, 194, 255, 256, 301). resting synapses but are rapidly dephosphorylated upon syn- Appendage domain of ␣-adaptin mediates interaction with apse stimulation and exocytosis of synaptic vesicle neurotrans- amphiphysin 1. Amphiphysin 1 is required for efficient associ- mitters (they are referred to collectively as dephosphins) (13, ation of dynamin 1 in brain lysates with ␣-adaptin appendage 78, 193, 271, 272, 302, 330). Are associations of amphiphysin 1 domain or clathrin heavy chain in vitro. Similarly, the addition with clathrin heavy chain, AP-2, dynamin 1, and synaptojanin 1 of recombinant amphiphysin 1 to extracts increases the asso- regulated during the cycle of synaptic vesicle exocytosis and ciation of both clathrin heavy chain and AP-2 with the dynamin endocytosis at nerve terminals? Dynamin 1, synaptojanin 1, 1 PRD in vitro. Hence, amphiphysin 1 can bind clathrin heavy clathrin heavy chain, and AP-2 coimmunoprecipitate with am- chain, AP-2, and dynamin 1 simultaneously and thus has the phiphysin 1 from brain extracts. However, when brain extract is potential to recruit dynamin 1 to plasma membrane clathrin- preincubated with ATP and protein phosphatase inhibitors to coated pits (46, 231, 256, 301, 302, 334, 352). induce phosphorylation, amphiphysin 1, dynamin 1, and syn- Amphiphysin 1 interaction with clathrin and AP-2 is essen- aptojanin 1 undergo a shift to slower electrophoretic mobilities tial for endocytosis. Transient overexpression of amphiphysin consistent with phosphorylation. Under these conditions dy- 1 fragments that contain both the clathrin heavy chain- and namin 1, synaptojanin 1, clathrin heavy chain, and AP-2 no AP-2-binding domains in CHO cells potently blocks receptor- longer coimmunoprecipitate with amphiphysin 1. The presence mediated endocytosis of transferrin. Expression of these am- of a protein kinase inhibitor (K252a) prevents both mobility phiphysin 1 fragments causes loss of clathrin heavy chain shift and loss of coimmunoprecipitation with amphiphysin 1, from plasma membrane-coated pits and results in a diffuse consistent with regulation by phosphorylation (302). cytoplasmic clathrin heavy-chain distribution. It also causes Is phosphorylation of amphiphysin 1 or its binding partners relocalization of AP-2 adaptors from clathrin-coated pits into responsible for loss of complex formation? To address this, larger plasma membrane-associated aggregates. Transient over- studies were performed in which brain extracts were preincu- expression in CHO cells of a mutated amphiphysin 1 fragment bated with ATP and protein phosphatase inhibitors to induce specifically unable to bind AP-2 has similar effects on clathrin phosphorylation and then tested for binding to recombinant heavy chain and AP-2 subcellular distribution and also blocks amphiphysin 1 SH3 domain. This treatment reduced binding of transferrin endocytosis. In contrast, overexpression of a mu- both dynamin 1 and synaptojanin 1 to the amphiphysin 1 SH3 tated amphiphysin 1 fragment specifically unable to bind clath- domain in vitro. This shows that it is phosphorylation of dy- rin heavy chain does not cause redistribution of clathrin heavy namin 1 and synaptojanin 1 (not amphiphysin 1) that regulates chain but still induces aggregation of AP-2 adaptors and blocks these protein interactions. In contrast, this treatment had little receptor-mediated endocytosis of transferrin. Hence, interac- effect on the binding of clathrin heavy chain or AP-2 to a tion of amphiphysin 1 with both clathrin heavy chain and AP-2 recombinant fragment of amphiphysin 1 comprising the clath- may be important for clathrin-dependent endocytosis (301, rin heavy chain and AP-2 binding domains in vitro. This sug- VOL. 70, 2006 BAR DOMAIN PROTEINS 85 gests that phosphorylation of clathrin heavy chain and AP-2 such as amphiphysin 1. Amphiphysin 1 can be phosphorylated may not regulate association with amphiphysin 1. However, the by immunoprecipitates of the Cdk5/p35 complex from rat brain treatment did reduce amphiphysin 1 binding to a recombinant or by recombinant Cdk5/p35 in vitro. Amphiphysin 1 Cdk5/p35 fragment comprising the appendage domain of ␣-adaptin and in vitro phosphorylation sites have been mapped to S261, S272, the unphosphorylated form of amphiphysin 1 bound more ef- S276, S285, and T310 (78, 272, 330). These Cdk5/p35 phos- ficiently than the phosphorylated form (302). Hence, phosphor- phorylation sites are all within the central proline-rich domain ylation of amphiphysin 1 regulates association with ␣-adaptin. of amphiphysin 1. Protein phosphatase 2B/calcineurin is responsible for de- Two approaches have been used to determine whether Downloaded from phosphorylation of amphiphysin 1, dynamin 1, and synapto- Cdk5/p35 phosphorylates amphiphysin 1 in vivo. One study janin 1 upon nerve stimulation. The protein phosphatase re- compared the phosphorylation status of amphiphysin 1 in rest- sponsible for stimulation-dependent dephosphorylation of ing synaptosomes from wild-type and p35-deficient mice with amphiphysin 1, dynamin 1, and synaptojanin 1 in synapses is and without treatment with the calcineurin inhibitor FK506. In the Ca2ϩ-dependent protein phosphatase calcineurin (13, 189, this study FK506 induced a very slight decrease in electro- 271, 302, 367). The addition of phosphatase inhibitors is nec- phoretic mobility, corresponding to phosphorylation in wild- essary to observe a shift in electrophoretic mobility of dynamin type mouse synaptosomes. This slight mobility shift was not 1 and synaptojanin 1 upon treatment of brain extract with induced by FK506 in p35-deficient mice synaptosomes (330). ATP. The specific calcineurin inhibitor cyclosporine A alone is This suggests that Cdk5/p35 contributes to amphiphysin 1 http://mmbr.asm.org/ sufficient to observe this shift in electrophoretic mobility upon phosphorylation in presynaptic terminals, although other ki- ATP treatment. However, addition of okadaic acid and vana- nases may be required for full phosphorylation. date as well as cyclosporine A increases the shift in electro- The second study examined rephosphorylation of amphiphy- phoretic mobility compared to cyclosporine A alone. Protein sin 1 following dephosphorylation induced by nerve terminal phosphatases other than calcineurin are therefore also likely to stimulation. Purified synaptosomes were radiolabeled with 32P play a role in dephosphorylation of endocytic proteins in syn- and then stimulated with Kϩ to induce dephosphorylation of apses (302). amphiphysin 1. Then the labeled synaptosomes were washed to Proline-rich central insert domain of amphiphysin 1 is remove Kϩ and allow rephosphorylation. Finally, the synapto- phosphorylated by Cdk5/p35 in vivo. Protein kinase C and somes were stimulated again with Kϩ. Complexes containing on October 26, 2015 by University of Queensland Library cyclin-dependent kinase 5 (Cdk5) have both been reported to amphiphysin 1 and its associated proteins were isolated from mediate constitutive phosphorylation of dynamin 1 and synap- the labeled synaptosomes and analyzed for phosphorylation tojanin 1 in resting synapses (40, 165, 271, 319, 330). Different status after each treatment. Amphiphysin 1 was rephosphory- results have been obtained with respect to phosphorylation of lated following the removal of Kϩ and this was unaffected by amphiphysin 1 by PKC. One study found that, unlike dynamin theCdk5-specificinhibitorroscovitine.Incontrast,rephosphory- 1 and synaptojanin 1, an inhibitor of PKC (Ro 31-8220) did not lation of dynamin 1 and synaptojanin 1 was blocked by rosco- block amphiphysin 1 rephosphorylation after stimulation-in- vitine. This suggests that rephosphorylation of dynamin 1 and duced dephosphorylation in brain synaptosomes (40). A sub- synaptojanin 1 is dependent on Cdk5/p35, but rephosphoryla- sequent study found that Ro 31-8220 does block phosphoryla- tion of amphiphysin 1 is not (319). Hence, if Cdk5 does play a tion of amphiphysin 1 in brain synaptosomes and that activators role in amphiphysin 1 phosphorylation in synaptosomes, this of PKC such as phorbol 12-myristate 13-acetate (PMA) stimulate role is redundant with other kinases. phosphorylation of amphiphysin 1 (367). However, PMA non- This is reminiscent of yeast Rvs167p, which binds Pcl2p, a specifically activates many kinases. Hence, despite evidence Pho85p cyclin, and is phosphorylated by Pho85p kinase in for PKC phosphorylation of amphiphysin 1 in vitro, there is vitro. Yeast Pho85p and Pcl2p are functional orthologs of no consensus on whether PKC phosphorylates amphiphysin mammalian Cdk5 and p35/p39, respectively, as shown by the 1 in vivo. ability of mammalian Cdk5/p35 to functionally replace Pho85p/ More recently, attention has focused on the potential role of Pcl2p in yeast (129, 225). The central GPA-rich domain of Cdk5 in regulation of amphiphysin 1. Amphiphysin 1 binds the yeast Rvs167p, which contains the major in vivo phosphoryla- Cdk5 cyclin p35 in vitro and a recombinant fragment compris- tion sites for Pho85p/Pcl2p, is equivalent to the central proline- ing residues 1 to 306 of amphiphysin 1 is sufficient. This shows rich domain of amphiphysin 1 that is phosphorylated by Cdk5/ that binding is mediated by either the N-terminal BAR domain p35 in vitro. As in the case of amphiphysin 1, Rvs167p or the central proline-rich insert domain, but not by the C- phosphorylation in vivo is only partially dependent on Pho85p/ terminal SH3 domain. This may also be analogous to the situa- Pcl2p. Deletion of all five Pcl2p-like cyclins somewhat reduces, tion in yeast, where the central GPA-rich domain of Rvs167p but does not abolish, Rvs167p phosphorylation in vivo (85, mediates Pcl2p binding (163). The amphiphysin 1-p35 interaction 163). This suggests that in yeast as well as in mammalian cells is likely to occur in vivo, as p35 coimmunoprecipitates with am- other kinases act redundantly with Cdk5 family kinases to phiphysin 1 (but not Cdk5) from rat brain extract (78). phosphorylate amphiphysin family proteins. Interestingly, p35 and the related Cdk5 cyclin p39 are ex- Does Cdk5-dependent phosphorylation of amphiphysin 1 pressed only in neurons and developing muscle cells (171, 242, affect interaction with its partner proteins dynamin 1, clathrin 320, 337). Immunostaining of rat cortical neurons in culture heavy chain, and AP-2? In one study in vitro binding assays shows that amphiphysin 1 and p35 colocalize in growth cones. showed that phosphorylation of amphiphysin 1 in vitro by Consistent with a role for Cdk5 and p35 in regulation of am- Cdk5/p35 had no effect on binding to dynamin 1, however, it phiphysin 1 in vivo, both Cdk5 and p35 have been implicated in inhibited binding to AP-2. In the same study phosphorylation neurite outgrowth, synapse formation, and neuronal migration of dynamin 1 in vitro by Cdk5 inhibited its interaction with the 86 REN ET AL. MICROBIOL.MOL.BIOL.REV. isolated SH3 domain of amphiphysin 1 (330). In contrast, an- sin 1 can be phosphorylated in vitro by the major cell cycle other report showed that Cdk5/p25 (p25 is a truncated form of regulatory Cdk Cdc2/cyclin B. The Cdc2/cyclin B kinase sites p35) phosphorylation of dynamin 1 in vitro does not affect mapped to date correspond to the known Cdk5/p35 phosphor- dynamin 1 binding to full-length amphiphysin 1 (319). This ylation sites, S272, S276, and S285 (78). While amphiphysin 1 study did find, however, that preincubation of dynamin 1 with is predominantly expressed in neurons, it is also expressed at the amphiphysin 1 SH3 domain blocked subsequent phosphory- lower levels in some other cell types (see above). In these lation of dynamin 1 by Cdk5. This suggests that phosphoryla- nonneuronal cells phosphorylation by Cdc2/cyclin B may have tion of dynamin 1, instead of regulating binding to amphiphy- a physiological role. Downloaded from sin 1, may in fact be regulated by binding to amphiphysin 1. Maybe phosphorylation of dynamin 1 in vivo only occurs at the Amphiphysin 1 Knockout Mice termination of synaptic vesicle endocytosis after dynamin 1 and amphiphysin 1 dissociate. The initial evidence for a role of amphiphysin 1 in synaptic These studies showing effects of phosphorylation on the vesicle recycling and endocytosis came from cell-based assays interactions of dynamin 1 and amphiphysin 1 have made use of employing overexpression of dominant negative constructs. Is proteins that are phosphorylated in vitro by Cdk5. As it is not amphiphysin 1 essential for synaptic vesicle recycling and en- yet clear to what extent Cdk5 phosphorylates amphiphysin 1 in docytosis in the context of a whole animal? The amphiphysin 1 vivo and as other kinases that phosphorylate amphiphysin 1 in (AMPH1) gene has been knocked out in mice and the conse- http://mmbr.asm.org/ vivo may exist, the physiological relevance of these findings quence of amphiphysin 1 deficiency on endocytosis at the syn- remains uncertain. In some cases effects on binding were in- apse examined. Mice homozygous for the amphiphysin 1 vestigated using protein fragments that represent individual knockout are devoid of amphiphysin 1 protein but do not domains (e.g., SH3 domain). In these cases phosphorylation exhibit developmental or anatomical abnormalities, are phys- may affect the interactions of these domains but not necessarily ically robust, and reproduce to yield viable progeny. However, affect the interactions of the full-length proteins (which may the amphiphysin 1-deficient mice are more susceptible to sei- interact via multiple domains). Whether interaction per se is zures that resemble epilepsy upon reaching adulthood than the only aspect that is important (e.g., for holding proteins wild-type mice. The amphiphysin 1 knockout mice also exhibit together in a complex) or whether the precise way in which the severe learning deficiencies. Hence, amphiphysin 1 is impor- on October 26, 2015 by University of Queensland Library proteins interact is also important (e.g., in terms of conforma- tant for several aspects of brain function (55). tion and activity of the complex) is an interesting area that has Various endocytic proteins associate in vitro with liposomes barely been explored. There are examples in which interactions prepared from purified brain lipids. These associations occur involving SH3 domains appear to be transient and more im- via interaction of endocytic proteins with amphiphysin 1, which portant for regulation than for formation of stable multipro- directly binds to membranes. When brain extracts from am- tein complexes (260). phiphysin 1-deficient mice and wild-type mice were com- As mentioned above, dephosphorylation of amphiphysin 1 pared, all endocytic proteins tested except amphiphysin 1 and and its associated proteins following nerve terminal depolar- Bin1ϩ6aϩ12ϩ13 (see below) were present at similar levels, ization induces synaptic vesicle recycling via endocytosis (40, showing that endocytic proteins are stable in the absence of 183). To determine if Cdk5/p35-dependent phosphorylation is amphiphysin 1. However, the ability of clathrin heavy chain, also important for synaptic vesicle recycling, the effect of Cdk5- AP-2, and synaptojanin 1 present in brain extracts from am- specific kinase inhibitors was examined. In the study that found phiphysin 1-deficient mice to associate in vitro with brain lipid some Cdk5-dependent phosphorylation of amphiphysin 1 (see liposomes was severely reduced. In contrast, association of above), Cdk5 inhibitors were also found to increase the num- dynamin 1 with brain lipid liposomes was unaffected by loss of ber of synaptic vesicles formed during that round of synaptic amphiphysin 1, presumably because dynamin 1 can bind mem- vesicle endocytosis in cultured hippocampal neurons (330). branes directly via its pleckstrin homology domain, which is This suggests that dephosphorylation of amphiphysin 1 favors known to bind phosphoinositides (55). synaptic vesicle endocytosis. In contrast, the study that found Is recruitment of dynamin 1 to clathrin coats affected by loss Cdk5-dependent phosphorylation of dynamin 1 but not am- of amphiphysin 1? The recombinant amphiphysin 1-binding phiphysin 1 also found that addition of Cdk5 inhibitors after C-terminal PRD of dynamin 1 forms complexes in vitro with nerve depolarization (to prevent rephosphorylation) actually amphiphysin 1 and also with clathrin heavy chain and AP-2 in inhibited the subsequent rounds of synaptic vesicle endocytosis wild-type brain extracts. When amphiphysin 1 knockout brain (319). This study concluded that rephosphorylation of endo- extracts were analyzed, the absence of amphiphysin 1 resulted cytic proteins (although not amphiphysin 1) by Cdk5/p35 is in reduced association of clathrin heavy chain and AP-2 with required for the maintenance of the synaptic vesicle pool. The the dynamin 1 PRD in vitro. However, binding of clathrin difference between the results of the two studies may be due to heavy chain and AP-2 to the dynamin 1 PRD was restored one study’s observing acute and the other study’s observing when the brain extract was supplemented with physiological long-term effects of Cdk5/p35 inhibition. amounts of recombinant amphiphysin 1 (55). Hence, although When amphiphysin 1 is transiently expressed in CHO cells dynamin 1 binds liposomes directly in the absence of am- that lack p35 or p39 amphiphysin 1 is still phosphorylated, phiphysin 1, in vivo amphiphysin 1 is important for linking indicating the existence of other kinases that can potentially dynamin 1 to components of clathrin coats on membranes. phosphorylate amphiphysin 1 in vivo. Cell cycle synchroniza- Amphiphysin 1 is important but not essential for synaptic tion reveals that amphiphysin 1 in CHO cells specifically un- vesicle recycling in vivo. Are there defects in synaptic vesicle dergoes phosphorylation during mitosis. Moreover, amphiphy- exocytosis or endocytic recycling in the synapses of amphiphy- VOL. 70, 2006 BAR DOMAIN PROTEINS 87 sin 1 knockout mice? Electron micrographs showed that the their fluorescent dye with comparable efficiency during initial synapses of amphiphysin 1 knockout mice, either in situ or action potentials, however, after more action potentials the after culturing in vitro, exhibit normal domain organization ability of the mutant neurons to recycle fluorescent dye into and contain the same number of synaptic vesicles as wild-type the medium declines compared to that of wild-type neurons synapses. Significantly, there was no significant accumulation (55). Hence, amphiphysin 1 knockout neurons exhibit a re- of coated invaginations at the plasmalemma in amphiphysin 1 duced active vesicle pool. knockout synapses. This result was surprising since it con- Synaptic vesicles that have been reinternalized must be rep- trasted with the situation in neurons that had been microin- rimed before they can undergo another round of stimulated Downloaded from jected with the recombinant amphiphysin 1 SH3 domain (293). exocytosis. The reduction in fluorescent dye recycling into the It also contrasted with what had been observed in neurons of medium observed in the amphiphysin 1 knockout neurons is dynamin mutant (shibire) flies (155). Hence, thorough exami- consistent with a possible defect in synaptic vesicle repriming. nation did not reveal any apparent defect in resting nerve Repriming is assayed by labeling neurons with a fluorescent terminals of amphiphysin 1 knockout mice (55). lipophilic dye using an action potential and then examining the Could subtler defects in synaptic vesicle exocytosis or endo- effect of extending the duration of the action potential beyond cytic recycling be revealed under conditions of nerve stimula- the period of dye exposure. The prolonged action potential can tion? Primary synaptosomes were isolated from the cerebral result in immediate recycling of the internalized dye into the cortex of wild-type and amphiphysin 1 knockout mice and medium if repriming of endocytosed synaptic vesicles is rapid. http://mmbr.asm.org/ tested for high-Kϩ-stimulated release of the neurotransmitter If repriming is delayed, the recycling of the internalized dye glutamate. The kinetics of Kϩ-stimulated exocytosis and the into the medium is slower and only apparent if the duration of quantity of glutamate release are both unaffected in amphiphy- the action potential after dye removal is extended. Consistent sin 1 knockout compared to wild-type synaptosomes. Quanti- with a longer synaptic vesicle repriming time in the amphiphy- tative assays of high-Kϩ-stimulation-dependent endocytic in- sin 1 knockout synapses, the action potential has to be ex- ternalization of fluorescent lipophilic dyes were performed and tended for longer times after removal of dye before the inter- here it was observed that following stimulation, endocytosis nalized dye is reexocytosed compared to wild-type neurons was 40% less efficient in amphiphysin 1 knockout compared to (55). Defects in repriming may account for the reduced pool of wild-type synaptosomes. Moreover, the dye that was internal- active synaptic vesicles in amphiphysin 1 knockout neurons. on October 26, 2015 by University of Queensland Library ized after a brief high-Kϩ stimulation was less efficiently de- Essential role for amphiphysin 1 in learning. Particularly livered back to the plasmalemma after a second high-Kϩ stim- exciting is the finding that amphiphysin 1 knockout mice have ulation in amphiphysin 1 knockout compared to wild-type severe learning deficiencies as revealed by standardized behav- synaptosomes (55). This indicates that the efficiency of both ioral tests (55). Knockout of other genes in mice that encode endocytic recycling and the subsequent return of endocytosed synaptic vesicle proteins, e.g., synapsin 1, causes epileptic sei- synaptic vesicles to an “active exocytic pool” are reduced by zures like those caused by knockout of amphiphysin 1, but in loss of amphiphysin 1. contrast to amphiphysin 1 knockout these do not affect learn- Cerebral cortex neurons cultured from amphiphysin 1 ing (294). It may be relevant that a human gene mutated in a knockout mice also exhibit a reduced percentage of labeled mental retardation disorder encodes a neuronal Rab GDP synaptic vesicles and endosomes after high-Kϩ stimulation in dissociation inhibitor (44). In yeast, Rvs167p interacts with a the presence of the fluid-phase endocytic marker horseradish catalytically active Rab-GAP (Gyp5p) as well as a catalytically peroxidase compared to cultured neurons from wild-type mice. inactive Rab-GAP family member (Gyl1p) (33, 84, 317), rvs167 This difference in the percentage of horseradish peroxidase- genetically interacts with a gene encoding an endosomal Rab labeled synaptic vesicles and endosomes is dependent on Kϩ protein (Ypt51p/Vps21p) (296), and loss of Rvs161p or stimulation, however. In resting synapses the fraction of syn- Rvs167p cause accumulation of what appear to be exocytic aptic vesicles and endosomes labeled with horseradish perox- vesicles that likely carry the Rab protein Ypt1p (20). idase is equivalent in amphiphysin 1 knockout and wild-type Could amphiphysin 1 in brain function in connection with Rab- neurons. Hence, there are mild defects in synaptic vesicle re- GTPase regulators? Altered Rab-GTPase activity in amphiphysin cycling and exocytosis in amphiphysin 1 knockout mice that 1 knockout mice, if it indeed occurs, might explain the delayed only become apparent under conditions of strong nerve stim- repriming of endocytosed synaptic vesicles, since Rab-GTPases ulation (55). are key regulators of vesicle exocytosis and their association with Interestingly, while neurons cultured from amphiphysin 1 vesicles is regulated by their GTP/GDP status. Indeed, the appar- knockout and wild-type mice have the same total number of ent endocytic defects in amphiphysin 1 knockout synapses may be synaptic vesicles, the pool that participates in action potential a consequence of a failure of vesicles internalized in one round stimulated exocytosis differs. Measurements of fluorescent li- of endocytosis to undergo exocytosis and thus contribute mem- pophilic dye uptake showed that the percentage of the total branes for the next round of endocytosis. Although there is no pool of synaptic vesicles that undergo exocytosis (and become direct evidence for amphiphysin 1–Rab-GAP interactions in labeled with fluorescent dye at the plasma membrane) in- mammals, such interactions may be worth exploring. creases in proportion to the number and duration of stimula- tory action potentials applied. However, the percentage of the Amphiphysin 1 and Molding of Membranes total pool of vesicles that participate in exocytosis and can be labeled is 25 to 30% lower for the amphiphysin 1 knockout Dynamin 1 binds membranes and evaginates tubules in neurons than wild-type neurons at each level of stimulation. vitro. Binding of clathrin coat components to protein-free li- Labeled wild-type and amphiphysin 1 knockout neurons lose posomes results in deformation of the initially spherical lipo- 88 REN ET AL. MICROBIOL.MOL.BIOL.REV. Downloaded from http://mmbr.asm.org/ on October 26, 2015 by University of Queensland Library

FIG. 11. Amphiphysin 1 evaginates liposomes in vitro to generate coated membrane tubules. A, electron micrographs of negatively stained membrane tubules generated in vitro from spherical liposomes composed of brain lipids a) after incubation with recombinant full-length amphiphysin 1 (cleaved from a GST fusion protein) or b) after incubation with a recombinant fusion protein comprising GST and the amphiphysin 1 BAR domain only (residues 1 to 286). Bar, 500 nm. B, electron micrographs of negatively stained membrane tubules generated in vitro from spherical liposomes composed of brain lipids after incubation with clathrin coat components plus: a, recombinant dynamin 1; b, both recombinant dynamin 1 and recombinant amphiphysin 1; or c, recombinant amphiphysin 1. The insets at the top right of each panel show the ends of tubules at high magnification (a, negatively stained sample; b and c, positively stained embedded and thin-sectioned samples). For comparison, the bottom inset in panel a shows clathrin-coated buds at high magnification. Note that the presence of amphiphysin 1 in panels b and c increases the frequency of clathrin-coated bud formation at tubule ends compared to dynamin 1 alone in panel a. C, electron micrographs showing the appearance of the tubular coat that forms when spherical liposomes are incubated in vitro with recombinant amphiphysin 1 alone (a and b); recombinant dynamin 1 alone (c and d); both recombinant amphiphysin 1 and recombinant dynamin 1 (e and f), or total brain cytosol with ATP and GTP␥S (g). The VOL. 70, 2006 BAR DOMAIN PROTEINS 89 somes to create clathrin-coated buds. Furthermore, binding of purified dynamin 1 and amphiphysin 1 in combination results dynamin 1 to protein-free liposomes results in evagination of in thick striations along the evaginated tubules, as seen when the liposomes to form narrow membrane tubules that resemble total brain extract is used. This suggests the rings assembled on the necks of deeply invaginated clathrin-coated pits. These evaginated tubules in the presence of total brain extract con- tubules have a striated appearance and can be immunolabeled tain both dynamin 1 and amphiphysin 1. The formation of with antiserum to dynamin 1, suggesting that assembly of dy- these thicker rings requires not only the membrane binding namin 1 rings on the liposome drives membrane deformation BAR domain of amphiphysin 1, but also the dynamin 1-inter- to form tubules (314). Dynamin 1 spontaneously assembles acting amphiphysin 1 SH3 domain. The presence of the iso- Downloaded from into closed rings and open spirals in vitro even in the absence lated amphiphysin 1 SH3 domain does not alter the appear- of liposomes, and these structures have dimensions similar to ance of coats on tubules evaginated by dynamin 1. Similarly, the membrane tubules (121). This suggests dynamin 1 ring and the isolated amphiphysin 1 BAR domain did not alter the spiral assembly at the membrane surface may drive membrane appearance of coats on tubules evaginated by dynamin 1. evagination to form tubules (121, 315). Addition of GTP to Hence, the formation of thick rings on evaginated tubules by dynamin 1-coated tubules causes them to undergo fission to dynamin 1 requires full-length amphiphysin 1 (314, 316). yield small vesicles (312). This reaction in vitro may mimic dy- Amphiphysin 1 stimulates assembly of dynamin 1 into thick namin 1 function in cleavage of the necks of deeply invaginated rings in solution. Purified amphiphysin 1, unlike purified dy- clathrin-coated pits during clathrin-mediated endocytosis. namin 1, is unable to assemble into rings in the absence of http://mmbr.asm.org/ Amphiphysin 1 binds membranes and evaginates tubules in liposomes. However, amphiphysin 1 does stimulate the assem- vitro. Amphiphysin 1 also binds directly to liposomes and has bly of dynamin 1 into rings in solution. Moreover, the rings membrane-tubulating activity in vitro (Fig. 11A). The lipo- formed by dynamin 1 in the presence of amphiphysin 1 in some-tubulating activity of amphiphysin 1 is even greater than solution contain both proteins and resemble the thick rings that of dynamin 1. The tubules formed by amphiphysin 1 dis- formed when liposomes are also present (316). Ring assembly play a striated appearance under electron microscopy, suggest- by amphiphysin 1 and dynamin 1 is influenced by their phos- ing amphiphysin 1 assembles into rings that surround mem- phorylation state. Dephosphorylated forms of amphiphysin 1 brane tubules such as dynamin 1 (Fig. 11B). When clathrin and dynamin 1 are highly active for assembly and their mixture coat components are also present, the membrane tubules results in highly efficient thick-ring assembly in vitro. In con- on October 26, 2015 by University of Queensland Library formed by amphiphysin 1 are almost always associated with trast, phosphorylated amphiphysin 1 and dynamin 1 are less clathrin-coated buds (Fig. 11C). Other clathrin coat compo- active for ring assembly and form few thick rings (330). nents (e.g., clathrin and AP-180) are unable to evaginate lipo- Amphiphysin 1 coordinates clathrin bud and dynamin 1 somes into tubules in vitro (316). Similarly, other known lipid- tubule formation. It is important that budding and fission be binding domains such as the pleckstrin homology (PH) domain coordinated during clathrin-mediated endocytosis. However, of phospholipase C␦ are also unable to evaginate liposomes when both clathrin coat components and dynamin 1 are incu- into tubules in vitro (73). This suggests the liposome-tubulating bated with liposomes in vitro, although clathrin-coated buds activity of amphiphysin 1 in vitro is specific and requires more and dynamin 1-coated tubules both form on the liposomes, than the ability to bind membranes. Amphiphysin 1 activity in there is a lack of coordination between the two structures. Few liposome tubulation in vitro likely reflects an important in vivo clathrin-coated buds have tubular dynamin 1-coated necks and function of amphiphysin 1 in clathrin-dependent endocytosis. few dynamin 1-coated tubules have clathrin-coated buds. Be- Amphiphysin 1 binding to membranes is dependent on lipid cause amphiphysin 1 interacts via the central insert domain composition. The membrane tubulation activity of amphiphysin 1 with clathrin coat components and via its C-terminal SH3 do- is affected by the lipid composition of the membrane. Liposomes main with dynamin 1, a key question is whether amphiphysin 1 composed of crude mixtures of brain lipids or of pure phospha- links clathrin coat components to dynamin 1 and thereby co- tidylserine and phosphatidylcholine are efficiently tubulated by ordinates vesicle budding with vesicle fission. Indeed, if am- amphiphysin 1, but tubulation activity is proportional to the per- phiphysin 1 is added to liposomes together with clathrin coat centage of phosphatidylserine (316). Hence, similar to dynamin 1, components and dynamin 1, there is an increase in the per- amphiphysin 1 preferentially binds and tubulates membranes centage of dynamin 1-coated tubules that are associated with composed of acidic phospholipids. clathrin-coated buds and vice versa (316). Amphiphysin 1 and dynamin 1 coassemble into rings on Amphiphysin 1 stimulates the activity of dynamin 1 in mem- membrane tubules. Incubation of liposomes with total brain brane tubule fission. Although amphiphysin 1 alone is unable extract results in very obvious thick striations with regular to mediate membrane fission in vitro, the addition of purified spacing along the evaginated membrane tubules. When either amphiphysin 1 substantially increases the extent of liposome purified dynamin 1 or amphiphysin 1 is used instead, the evagi- tubulation and subsequent GTP-induced fission mediated by nated tubules display a continuous and tightly spaced arrange- purified dynamin 1. Under the conditions used in this study the ment of thin rings. Interestingly, incubation of liposomes with addition of amphiphysin 1 did not increase dynamin 1 associ-

images in panels a, c, and e represent negatively stained samples, while those in panels b, d, f, and g represent samples embedded, thin sectioned, and then positively stained. Note that recombinant amphiphysin 1 or dynamin 1 alone forms thin tightly packed rings on membrane tubules while recombinant amphiphysin 1 and dynamin 1 together or endogenous amphiphysin 1 and dynamin 1 in total brain extract form thick rings with regular spacing on membrane tubules. (Reprinted from reference 316 with permission of the publisher.) 90 REN ET AL. MICROBIOL.MOL.BIOL.REV. ation with membranes or stimulate dynamin 1 GTPase activity of small liposomes that, because of their size, cannot support (316). Interestingly, the activity of amphiphysin 1 and dynamin evagination of high-curvature tubules, the basal GTPase activ- 1 in liposome tubulation and fission is regulated by their phos- ity of dynamin 1 was higher than in the presence of large phorylation state. Dephosphorylated forms of both dynamin 1 liposomes. However, in contrast to what was seen with large and amphiphysin 1 were highly active in tubulation and fission, liposomes, addition of amphiphysin 1 in this case led to a and incubation with liposomes resulted in dramatic generation dose-dependent inhibition of dynamin 1 GTPase activity. Small of small vesicles. In contrast, phosphorylation by Cdk5/p35 of liposomes contain insufficient membrane material to support only dynamin 1 or amphiphysin 1 slightly reduced vesicle for- the formation of dynamin 1- and amphiphysin 1-coated high- Downloaded from mation, while phosphorylation of both proteins almost com- curvature tubules, and amphiphysin 1 under these conditions pletely inhibited the generation of small vesicles (330). may engage in inhibitory interactions with dynamin 1. How- In vivo role for amphiphysin 1 in membrane tubule forma- ever, in the case of large liposomes the conversion of low- tion. Is the ability of amphiphysin 1 to bind and tubulate curvature membranes into evaginated tubules with high curva- membranes important in vivo? The first evidence that am- ture allows extensive coassembly of dynamin 1 and amphiphysin phiphysin 1 generates membrane tubules in vivo came from 1 into rings. This promotes a different type of interaction between experiments in which a fragment comprising the amphiphysin amphiphysin 1 and dynamin 1 that in turn stimulates dynamin 1 1 N-terminal amphipathic ␣-helix and BAR domain (N-BAR) GTPase activity (380). was overexpressed in COS-7 cells. Overexpression of the N- Amphiphysin 1 tubulation of membranes is dependent on http://mmbr.asm.org/ BAR fragment induced massive tubulation of the plasma lipid composition. The influence of membrane lipid composi- membrane in vivo (237). The importance of the N-terminal tion on stimulation of dynamin 1 GTPase activity by amphiphy- amphipathic helix for high-affinity liposome binding and tubu- sin 1 was also investigated. In the absence of liposomes dy- lation in vitro was first demonstrated for the BAR domain namin 1 GTPase activity was low and only very weakly protein endophilin A1 (see below), but the amphiphysin 1 stimulated by amphiphysin 1. Large liposomes containing brain N-terminal amphipathic helix appears to play a similar role lipids and supplemented with cholesterol and phosphatidylino- (73). There is also evidence that amphiphysin 1 plays a physi- sitol-4,5-phosphatase [PtdIns(4,5)P2]-stimulated dynamin 1 ologically important role in dynamin 1-mediated membrane GTPase activity, even in the absence of amphiphysin 1. Addi- fission. Brain cytosol prepared from amphiphysin 1 knockout tion of amphiphysin 1 in the presence of these liposomes re- on October 26, 2015 by University of Queensland Library mice (and also deficient in Bin1ϩ6aϩ12ϩ13, see below) has sulted in further strong stimulation of dynamin 1 GTPase ac- considerably reduced ability to support dynamin 1-dependent tivity (380). This is consistent with the view that stimulation of fission of large unilamellar liposomes into small vesicles in dynamin 1 GTPase activity by amphiphysin 1 is not due to vitro. This deficiency can be rescued by addition of purified direct amphiphysin 1 interaction with dynamin 1 in solution, recombinant amphiphysin 1 (380). This shows that the mem- but instead related to the ability of these two proteins to brane fission defect in these extracts is specifically due to a lack coassemble into rings on high curvature membranes. of amphiphysin 1 (although Bin1ϩ6aϩ12ϩ13 is also deficient, PtdIns(4,5)P2 binds the PH domain of dynamin 1 and is it appears that amphiphysin 1 alone can stimulate dynamin known to not only recruit dynamin 1 to membranes but also 1-dependent fission). activate dynamin 1 GTPase activity in vitro. If PtdIns(4,5)P2 Membrane curvature and amphiphysin 1 binding. The N- was substituted with phosphatidylcholine in the large lipo- terminal amphipathic ␣-helix is critical for high-affinity binding somes, dynamin 1 GTPase activity was not as strongly stimu- to liposomes and tubulation, however, the BAR domain alone lated by liposomes alone, but addition of amphiphysin 1 still retains lower affinity binding to liposomes. A key finding was resulted in strong stimulation (380). Perhaps amphiphysin 1 is that binding of the amphiphysin 1 BAR domain alone to lipo- especially important for dynamin 1 recruitment to membranes somes in vitro is critically dependent on liposome size and low in PtdIns(4,5)P2. Dynamin 1 and amphiphysin 1 have been therefore membrane curvature. While the amphiphysin 1 BAR shown to tubulate liposomes containing high levels of phos- domain binds poorly to large liposomes (0.8-␮m diameter) that phatidylserine (314, 316). Furthermore, liposomes containing have low curvature, it binds strongly to small liposomes phosphatidylserine are subject to GTP-dependent fission by (0.05-␮m diameter) that have high curvature. This result is dynamin 1 in vitro (312, 316). When PtdIns(4,5)P2 was re- important because it shows the BAR domain not only gener- placed with phosphatidylserine, dynamin 1 GTPase activity was ates membrane curvature (in conjunction with the N-terminal strongly stimulated by liposomes alone and activity was directly amphipathic ␣-helix), but alone can also act as a sensor of proportional to liposome phosphatidylserine content. More- membrane curvature. An exciting possibility is that amphiphy- over, the addition of amphiphysin 1 led to a very dramatic sin 1 may recruit dynamin 1 specifically to high-curvature further stimulation of dynamin 1 GTPase activity that exhib- membranes such as those at the tubular neck of clathrin-coated ited positive cooperativity with liposome phosphatidylserine pits. Hence, amphiphysin 1 may potentiate dynamin 1-depen- content (380). dent membrane fission by localizing dynamin 1 within the mem- Domains of amphiphysin 1 required for membrane tubula- brane to sites where fission is most efficient (237, 316, 380). tion. Testing of various recombinant fragments of amphiphysin The effects of membrane curvature on amphiphysin 1- and 1 revealed that the N-terminal BAR domain (residues 1 to dynamin 1-dependent membrane fission were investigated re- 286) is sufficient for full liposome tubulation activity and that cently. In the presence of large liposomes that can be evagi- an N-terminal piece of the BAR domain (residues 1 to 161) nated into high-curvature tubules, addition of amphiphysin 1 retains low tubulation activity. In contrast, the C-terminal SH3 results in a dose-dependent stimulation of both dynamin 1 domain has no tubulation activity (316). Tubulation activity GTPase activity and recruitment to liposomes. In the presence correlates with stimulation of dynamin 1 GTPase activity. In VOL. 70, 2006 BAR DOMAIN PROTEINS 91 the presence of large liposomes containing cholesterol and requirement for the BAR domain even in the absence of mem- PtdIns(4,5)P2, the isolated amphiphysin 1 BAR domain stim- branes suggests that coassembly of amphiphysin 1 and dynamin ulates dynamin 1 GTPase activity as efficiently as full-length 1 in solution may require that amphiphysin 1 make contacts amphiphysin 1. This is probably due to the BAR domain’s with dynamin 1 via the BAR domain as well as the SH3 do- promoting high-curvature tubule formation, which in turn fa- main. However, direct evidence of amphiphysin 1 BAR do- cilitates dynamin 1 binding and assembly into rings. In con- main-dynamin 1 interaction is lacking. trast, the isolated amphiphysin 1 SH3 domain was not able to stimulate dynamin 1 GTPase activity in the presence of large Downloaded from liposomes (380). SECOND ISOFORM OF AMPHIPHYSIN, Bin1 The BAR domain of amphiphysin 1 is predicted to be highly Discovery of Bin1 acidic and hence to bear a strongly negative net charge. That the BAR domain is the part of amphiphysin 1 that mediates A number of findings suggested the existence of a second, binding to acidic lipids, including phosphatidylserine and more ubiquitously expressed amphiphysin isoform in verte- PtdIns(4,5)P2, initially seemed somewhat puzzling. However, brates. These included the ability of the amphiphysin 1 SH3 the three-dimensional structure of the amphiphysin BAR do- domain to powerfully inhibit endocytosis in nonneuronal cells main provides a simple explanation for this apparent paradox that express only very low levels of amphiphysin 1 (369), the because it reveals that the membrane binding surface of BAR existence of amphiphysin-related proteins in unicellular eu- http://mmbr.asm.org/ domains bears a net positive charge (see below). karyotes such as yeast (12, 42, 47, 298), and the finding that Interestingly, a construct containing both the BAR and SH3 amphiphysin 1-reactive sera from stiff-man syndrome patients domains but missing the central insert domain is considerably also recognize an additional protein distinct in size from am- more active in stimulation of dynamin 1 GTPase activity than phiphysin 1 (338). Various approaches led to the isolation of full-length amphiphysin 1. Hence, full stimulation of dynamin this second isoform of amphiphysin, which has resulted in its 1 GTPase activity requires both the BAR and SH3 domains, having several names, e.g., amphiphysin 2 (168), amphiphysin but not the central insert domain. The central insert domain of II (23, 256, 338), Amph2 (367), BRAMP2 (168), ALP1 (141), amphiphysin 1 is known to bind the SH3 domain via an in- SH3P9 (306), and Bin1 (278). The Human Genome Organi- tramolecular interaction and this inhibits interaction of the zation-approved designation for this family of amphiphysin- on October 26, 2015 by University of Queensland Library SH3 domain with dynamin 1 (74). Deletion of the proline-rich related proteins is Bin1, so Bin1 is the name that will be used motif in the central insert domain that binds the SH3 domain in this review. enhances the stimulatory effect of amphiphysin 1 on dynamin Sequence comparison revealed that Bin1 is the product of a 1 GTPase activity. The isolated amphiphysin 1 SH3 domain novel gene and that Bin1 transcripts are subject to extensive does not further enhance the ability of the isolated amphiphy- differential splicing, leading to a diversity of Bin1 forms. Table sin 1 BAR domain to stimulate dynamin 1 GTPase activity in 2 lists all the known splice variants of Bin1 and the various the presence of large liposomes (380). This shows that the names that have been assigned to them. Bin1 has a domain function of the SH3 domain is to link dynamin 1 to the am- structure similar to that of amphiphysin 1, featuring an N-termi- phiphysin 1 BAR domain and that the SH3 domain does not nal BAR domain with predicted coiled-coil structure and a C- independently stimulate dynamin 1 GTPase activity. terminal SH3 domain (Fig. 1) (23, 141, 168, 256, 278, 306, 367). The stimulatory effect of these amphiphysin 1 constructs on The central insert domain of Bin1 is more divergent than dynamin 1 GTPase activity correlates with their ability to re- that of amphiphysin 1. Some splice variants include a central cruit dynamin 1 to membranes in vitro, e.g., the amphiphysin 1 insert domain that is highly homologous to the central insert construct lacking the central insert domain was more active domain of amphiphysin 1 and that interacts with clathrin and than full-length amphiphysin 1 both in recruiting dynamin 1 to AP-2/␣-adaptin (sometimes referred to as the endocytosis do- liposomes and in stimulating dynamin 1 GTPase activity (380). main) (Fig. 1). Other splice variants lack this central insert This suggests that coassembly on membranes and stimulation domain or have alternative central insert domains that do not of dynamin 1 GTPase activity are linked. This is consistent with interact with clathrin or AP-2/␣-adaptin (Fig. 1) (168, 255, 278, results from other studies that show dynamin 1 assembly stim- 338, 356). The central insert domain contains two proline-rich ulates its GTPase activity (353). Stimulation is due to intermo- motifs (residues 297 to 306 and 340 to 346 in mouse Bin1) that lecular interactions between the GTPase effector domain of may bind SH3 domains (168). Although the hydrophobic re- one dynamin 1 monomer and the GTPase catalytic domain of gion in the central insert domain of human and chicken am- the adjacent dynamin 1 monomer that can take place only phiphysin 1 is missing in Bin1, there is a region of hydropho- within a multimeric assembly (203). bicity near the Bin1 SH3 domain (338). In some brain splice Domains of amphiphysin 1 required for coassembly with variants a 31-residue insert (N-terminal insert domain [NTID]) dynamin 1 into rings in solution. The domains of amphiphysin occurs within what was originally predicted to be a coiled-coil 1 required for coassembly with dynamin 1 into rings in solution motif in the BAR domain (338). Comparison with the recently have also been investigated. Visualization of negatively stained elucidated three-dimensional structure of Drosophila am- protein samples by electron microscopy reveals that full-length phiphysin now suggests this site of insertion is a disordered amphiphysin 1 and amphiphysin 1 constructs lacking either the loop between ␣-helices (237). proline-rich motif or the entire central insert domain are able Four studies identified Bin1 partial sequences simply by to coassemble with dynamin 1 into rings in solution. In con- searching expressed sequence tag databases and these were trast, the isolated BAR and SH3 domains lack the ability to then used to obtain full-length cDNAs by hybridization screens coassemble with dynamin 1 into rings in solution (380). This (23, 256, 338, 367). In one study, a mouse cDNA expression 92 REN ET AL. MICROBIOL.MOL.BIOL.REV. library was screened with a synthetic peptide known to be a which of the exon 12 variants is present, only that an exon 12 ligand of the Src SH3 domain in an attempt to identify novel is present. SH3 domain proteins. One of the novel murine SH3 domain Finally, exon 13 includes part of the Myc binding domain proteins identified was named SH3p9 and is a splice variant of and is differentially spliced in a tissue-independent manner Bin1 (Table 2) (306). Bin1 was also isolated in two-hybrid (100, 141, 338, 356, 357, 367). Bin1ϩ6aϩ12ϩ13 represents the screens for proteins that interact with the oncoproteins c-Myc neuronal isoforms of Bin1 (amphiphysin 2, amphiphysin II, (278) and c-Abl (141) (Table 2). Bin1 was independently iden- Amph2, and BRAMP2), and Bin1ϩ10ϩ13 represents the muscle isoform of Bin1 (commonly referred to as Bin1).

tified in a two-hybrid screen for mouse brain proteins that Downloaded from interact with the proline-rich C-terminal domain of the human Bin1Ϫ10Ϫ12ϩ13 represents the widely expressed Bin1 iso- Ras GDP/GTP exchange factor son of sevenless 1 (Sos-1) and form known as SH3p9. Finally, Bin1Ϫ10Ϫ12Ϫ13 is the most named BRain form of AMPhiphysin 2 (BRAMP2) (Table 2). ubiquitously expressed Bin1 isoform, known as ALP1 or am- Purified recombinant BRAMP2 bound in vitro-translated human phiphysin IIm. SOS-1 and -2, confirming the yeast two-hybrid interaction and Different Bin1 splice variants exhibit different tissue distri- showing that BRAMP2 binding to SOS-1 and -2 is direct (168). butions. In the rat, every tissue tested expressed transcripts ϩ Characterization of independent Bin1 cDNAs from different lacking exons 6a and 10 except the retina (only 6a) and skeletal muscle (only ϩ10). Exon 6a was present in at least

tissues and even from the same tissue revealed complete se- http://mmbr.asm.org/ quence identity over included domains but a high degree of some transcripts from all nerve tissues, but was not present in heterogeneity in domain structure, consistent with the possi- transcripts from any other tissues. In contrast, all skeletal mus- bility that the Bin1 transcript is subject to differential splicing cle transcripts included exon 10, but exon 10 was not present in (Fig. 1 and Table 2) (23, 168, 256, 278, 338, 356, 367). In two transcripts from any other tissue (Fig. 1 and Table 2). In the reports at least 10 different transcripts could be amplified by human, rat, and mouse, transcripts containing exon 12 that PCR from human or rat brain tissue using primers that recog- encodes the 127-residue central insert domain are expressed nize sequences encoding the N- and C-terminal domains of only in the brain (in particular the cerebral cortex and pons), Bin1 (338, 367). The longest Bin1 splice variant in human brain spinal cord, other neuronal tissue, and neuron-like pheochro- mocytoma (PC12) cells, similar to amphiphysin 1 (23, 168, 256, exhibits 71% amino acid sequence similarity and 55% amino 278, 338, 356, 367). Some nonneuronal cells misexpress tran- on October 26, 2015 by University of Queensland Library acid sequence identity with human amphiphysin 1 (23). Inter- scripts containing exon 12 following tumorigenesis (Table 2) estingly, in one study two of the Bin1 cDNAs appeared to (93, 356). In rat brain Bin1 represents 0.1% of the total pro- encode splice variants that lacked a C-terminal SH3 domain, tein, similar to rat brain amphiphysin 1 (367). similar to yeast Rvs161p (Table 2) (367). In the rat, forms of Bin1 protein lacking the large brain- The intron-exon boundaries of the 54-kb human BIN1 gene specific central insert domain encoded by exon 12 were ex- have been mapped by Wechsler-Reya et al. BIN1 encodes at pressed at high levels in skeletal muscle and could be detected least 20 exons, of which at least seven are differentially spliced, (at approximately 10-fold lower levels) in many tissues, includ- giving rise to a vast array of Bin1 splice variants (Fig. 1 and ing heart, brain, lung, liver, skeletal muscle, kidney, ovary, and Table 2). This review will adopt the splice variant nomencla- thymus (the skeletal muscle forms include the 15-residue mus- ture proposed by Wechsler-Reya et al., but with the addition of cle-specific domain encoded by exon 10 but the presence of this one originally overlooked exon located between exons 6 and 7, domain is difficult to distinguish by electrophoretic mobility) which will be referred to in this review as exon 6a (356). It (367). In the mouse, transcripts lacking exon 12 were expressed should be noted that in the new systematic exon nomenclature in multiple tissues, including brain, liver, lung, kidney, heart, for Bin1, exon 6a is renamed exon 7 and all following exons ovary, testis, spleen, and skeletal muscle (168, 278). In the ϩ have numbers that are n 1 with respect to those assigned by human, transcripts lacking exon 12 were present in the brain, Wechsler-Reya. The majority of published reports have used heart, placenta, lung, liver, kidney, pancreas, spleen, ovary, the exon nomenclature of Wechsler-Reya and not the new testis, and skeletal muscle (23, 141). They were expressed at systematic exon nomenclature. Our decision to retain the older equal levels in cultured normal human fibroblasts and in some nomenclature in this review ensures the exon numbers referred cultured human tumor cell lines (e.g., HeLa cervical carcinoma to here are consistent with those in most of the published cells), however, they were absent in other tumor cell lines (e.g., literature. HepG2 hepatocarcinoma cells and MCF7 breast carcinoma Although there exist many splice variants, the major forms cells) (278). vary in four exons: 6a, 10, 12, and 13 (Fig. 1 and Table 2). Exon The various Bin1 splice variants exhibit anomolous electro- 6a encodes the 31-residue brain-specific NTID (257). There is phoretic mobilities. In humans the brain isoform of Bin1 mi- no sequence homologous to the NTID in the reported clones grates at 85 kDa, while in rats it migrates at 92 kDa. Ectopic of amphiphysin 1 (23, 367). Exon 10 is muscle specific and expression of the equivalent mouse cDNA gave rise to a dou- encodes a 15-residue sequence that contains a putative nuclear blet at 88 and 96 kDa. In contrast, the longest Bin1 transcript localization sequence and lipid binding sequence (162, 278). in human, rat, or mouse brain is predicted to encode a protein Exon 12 includes a series of alternative brain-specific exons of only 65 kDa (23, 168, 256, 367). The aberrant gel mobility of (12A to -D) that encode the central insert domain. Depending brain Bin1 isoforms is similar to what has been observed for on the particular exon 12 variant this central insert domain can amphiphysin 1. The amino acid sequence encoded by exon 12 interact with ␣-adaptin/AP-2 and/or clathrin heavy chain corresponds to a sequence that in amphiphysin 1 has been and/or endophilin A1 (23, 168, 187, 255, 256, 338, 356, 367). shown to confer aberrant electrophoretic mobility (47, 175). For simplicity, except in Table 2 this review will not distinguish Indeed, in the rat the presence of exon 12 in Bin1 has been VOL. 70, 2006 BAR DOMAIN PROTEINS 93 shown to confer aberrant electrophoretic mobility (367). Inter- also localizes to actin-rich membrane domains in a variety of estingly, the muscle isoform containing the 15-residue domain polarized cell types, including the actin-rich apical domain in encoded by exon 10 migrates on gels at 60 to 70 kDa although intestinal epithelial cells and the actin-rich sensory microvilli of its predicted size is 50 kDa, so it also exhibits aberrant elec- photoreceptor neurons (386). trophoretic mobility (278). Interestingly, axon initial segments and nodes of Ranvier are active in endocytosis. Within neurons, axon initial seg- Bin1؉6a؉12؉13/Amphiphysin 2/Amphiphysin ments and nodes of Ranvier have a higher concentration of

II/BRAMP2 in the Brain clathrin-coated pits and vesicles than any other site except Downloaded from the presynaptic terminal. Moreover, they are also enriched The function of Bin1ϩ6aϩ12ϩ13 in the brain has been in a variety of integral membrane proteins that mediate ion the topic of an excellent earlier review (368). There are flux, including Naϩ channels, Naϩ/Kϩ-ATPase, and Naϩ/ several brain-specific isoforms of Bin1 that vary in exon 12 Ca2ϩ exchangers (23). This suggests axon initial segments subtype (12A to 12D) (encoding the central insert domain) and nodes of Ranvier are important sites within neurons for and the presence of exon 6a (encoding the NTID) (Fig. 1 the generation and propagation of action potentials. This and Table 2). is interesting, as yeast rvs161 and rvs167 mutants exhibit Distribution of Bin1؉12؉13 in the brain. Different local- ϩ

growth that is severely sensitive to Na . Regulation of ion http://mmbr.asm.org/ ϩ izations have been reported for Bin1 12 in the brain (the channel activity may turn out to be a conserved role of antisera used for these studies could not distinguish between amphiphysin family proteins. isoforms containing and lacking exons 6a and 13, so they will While the reason for the different subcellular localizations ϩ collectively be referred to here as Bin1 12). In one study of Bin1ϩ12 observed in different studies is not yet clear, it ϩ immunocytochemistry revealed a distribution of Bin1 12 in appears to be related to the antibody used for detection. ϩ the brain similar to that reported for amphiphysin 1. Bin1 12 Independent monoclonal antibodies raised to the SH3 do- was strongly enriched in a subset of nerve terminals within all main of Bin1ϩ12 preferentially label axon initial segments three layers of the cerebellum as well as the hippocampus and and nodes of Ranvier. In contrast, polyclonal antibodies pontine nucleus (367). Similarly, in a second study immuno- raised against Bin1ϩ12 preferentially label presynaptic ter- fluorescence staining of brain sections showed that Bin1ϩ12 on October 26, 2015 by University of Queensland Library mini. One possibility is that the SH3 domain epitopes rec- staining is punctate and colocalizes with known synaptic vesicle ognized by the monoclonal antibodies are masked at syn- markers and a similar localization to punctate synapses was apses due to conformational changes or posttranslational seen in spinal cord. The subcellular distribution of Bin1ϩ12 modification. was cytoplasmic and never nuclear, in contrast to what has -Localization of Bin1؉6a؉12 on purified plasma mem been reported for some other splice variants of Bin1 (see branes. Those isoforms of Bin1 in the brain that include the below) (256). Subcellular fractionation showed that Bin1ϩ12 NTID encoded by exon 6a specifically localize to the plasma cofractionates with isolated nerve terminals. Immunoelectron membrane upon transient expression in COS-7 cells. The dis- microscopy of rat brain sections shows that Bin1ϩ12 is present tribution of Bin1ϩ6aϩ12 (this includes isoforms with or lack- on the surface of synaptic vesicles and also in isolated patches on the plasma membrane (367). ing sequences encoded by exon 13) on plasma membrane is In contrast, another study found Bin1ϩ12 distribution in the punctate, and the puncta partially colocalize with clathrin- ϩ ϩ brain differs from that of amphiphysin 1. While amphiphysin 1 coated pits. Bin1 6a 12 is detected at 62% of plasma mem- is enriched at synapses in the cerebral cortex, hippocampus, brane clathrin-coated pits. This degree of colocalization with and cerebellum, this study found Bin1ϩ12 enriched at other clathrin-coated pits is similar to that reported for dynamin sites in the brain. In gray matter, which contains perikarya 1, but less than that reported for other clathrin-coated pit (neuronal cell bodies), dendrites, and axons, Bin1ϩ12 local- components, e.g., intersectin (257). There were puncta of ϩ ϩ ized to axon initial segments. Axon initial segments are short Bin1 6a 12 on the plasma membrane that did not colocalize unmyelinated tubular segments innervated by basket cells with clathrin-coated pits, but the identity of these structures is where the axon is attached to the perikaryon. Bin1ϩ12 also not known. Membranes other than plasma membrane were not ϩ ϩ localized to the nodes of Ranvier. The nodes of Ranvier are examined in this study, so Bin1 6a 12 may also associate with fine ring-like structures distributed at intervals along the length other intracellular membranes. of axons. In white matter, which contains mainly axons, the Association of Bin1 with membranes. Subcellular fraction- densely clustered nodes of Ranvier were found to be the main ation of purified nerve terminals revealed that Bin1ϩ6aϩ12 sites of Bin1ϩ12 localization (23). (including isoforms with or lacking sequences encoded by exon Both axon initial segments and nodes of Ranvier exhibit a 13) is present at approximately equal levels in a soluble pool dense submembranous cytoskeletal matrix. This matrix con- and in a membrane-associated sedimentable pool (256, 367). tains actin and a neuron-specific isoform of ankyrin 3 as major When Bin1ϩ6aϩ12 is expressed ectopically in COS-7 fibro- components. In high-magnification images of both axon initial blasts it is also found in association with a sedimentable frac- segments and nodes of Ranvier, Bin1ϩ12 localizes within the tion. Hence, association of Bin1ϩ6aϩ12 with membranes does submembranous cytomatrix (23). This subcellular localization not require brain-specific integral membrane proteins. As suggests a role for Bin1ϩ12 in organization or function of the mentioned above, the membrane association of Bin1ϩ6aϩ12 cortical actin cytoskeleton and/or in defining these specialized is very strong and resists washes with 0.5 M Tris-HCl, 1 M KCl, membrane domains within neurons. In this context it is inter- and 1 M NaCl that strip all clathrin heavy chain and most AP-2 esting that in fruit flies (Drosophila melanogaster) amphiphysin from membranes (257, 367). Tight and salt-resistant associa- 94 REN ET AL. MICROBIOL.MOL.BIOL.REV. tion with membranes is also a property of one pool of am- SH3 domain and the proline-rich motif GPPPQVPSRPNR in phiphysin 1 (175). the dynamin 1 PRD appear to mediate the interaction (23, Recruitment of Bin1؉6a؉12 to the plasma membrane. 168). Although both amphiphysin 1 and Bin1ϩ6aϩ12ϩ13 bind Bin1ϩ6aϩ12 contains the central insert domain encoded by dynamin 1 and synaptojanin 1, peptide competition experi- exon 12 which, depending on the subtype of exon 12 (A to D), ments suggest the two SH3 domains may differ in their affinity may bind both clathrin heavy chain and/or AP-2 (Fig. 1). Could for these proteins (e.g., the Bin1ϩ6aϩ12ϩ13 SH3 domain this domain mediate plasma membrane clathrin-coated pit as- binds synaptojanin 1 with much lower affinity than the am- sociation? The discovery that Bin1ϩ6aϩ12 also localizes to phiphysin 1 SH3 domain) (23, 302, 367). Interestingly, one Downloaded from regions of plasma membrane other than clathrin-coated pits study found that interaction of endogenous Bin1ϩ6aϩ12ϩ13 together with the finding that Bin1ϩ6aϩ12 remains associated with dynamin 1 in PC12 cells is regulated by nerve growth with isolated plasma membranes after clathrin heavy chain and factor (168). AP-2 are both removed by salt washes suggests that other Bin1؉12 (with or lacking exons 6a and 13) interacts with interactions may play an important role. Consistent with this the ␣-adaptin subunit of AP-2 in the brain. Recombinant view, ectopic expression in COS-7 fibroblasts of four different Bin1ϩ6aϩ12ϩ13 binds the ␣-adaptin subunit of AP-2 in splice variants of Bin1 that differ in the presence or absence of extracts of NIH-3T3 fibroblasts or rat brain (168, 187). The sequences encoded by exons 6a and 12 revealed that some BAR domain and central insert domain (residues 1–422) of isoforms that lack the central insert domain encoded by exon Bin1ϩ6aϩ12ϩ13 and the appendage domain of ␣-adaptin me- http://mmbr.asm.org/ 12 (and that therefore lack the ability to bind clathrin heavy diated this interaction (187). In contrast, another study found chain or AP-2) are still efficiently targeted to the plasma mem- that ␣-adaptin does not bind to the central insert domain of brane. Intriguingly, plasma membrane targeting of Bin1 in Bin1ϩ6aϩ12ϩ13 but rather to the SH3 domain (256). The neurons correlates with the presence of the NTID. Further- difference in ␣-adaptin binding is not due to different exon 12 more, deletion of the NTID from splice variants that include splice variants because a sequence motif (LFED) similar to the this sequence completely abolishes plasma membrane localiza- FFED ␣-adaptin binding motif of amphiphysin 1 is present in tion of all Bin1 isoforms examined, including those that also the Bin1ϩ6aϩ12ϩ13 constructs used in all three studies. contain clathrin heavy chain- and AP-2-binding domains (257). Bin1؉12 (with or lacking exons 6a and 13) interacts with The NTID also plays an important role in general associa- clathrin heavy chain in the brain. Clathrin heavy chain also on October 26, 2015 by University of Queensland Library tion of Bin1 with membranes. When expressed in COS-7 fi- binds some splice variants of Bin1ϩ12 depending on which of broblasts the BAR domain alone of each plasma membrane the four exons 12 (A to D) are present in the spliced transcript. isoform of Bin1 associates with membranes such as the full- Binding to clathrin heavy chain is direct and mediated by the length protein during subcellular fractionation. Hence, the Bin1ϩ12 central insert domain (187, 256). In one study it was SH3 domain is not essential for association of Bin1 with mem- observed that full-length Bin1ϩ12 is less efficient at binding branes. In contrast, deletion of the BAR domain from plasma clathrin heavy chain than a fragment comprising the BAR and membrane isoforms of Bin1 completely abolishes association central insert domains only and that the addition of dynamin 1 with membranes (257). These data indicate that the BAR inhibits binding of full-length Bin1ϩ12 to clathrin heavy chain domain of plasma membrane isoforms of Bin1 is both neces- (187). While the mechanism by which dynamin 1 binding to the sary and sufficient for association with membranes. This find- Bin1ϩ12 SH3 domain inhibits clathrin heavy chain binding to ing is consistent with the observation that the isolated BAR the central insert domain is not understood, one possibility is domain of yeast Rvs167p retains some ability to localize to that coassembly of Bin1ϩ12 and dynamin 1 into rings in solu- cortical actin patches and to functionally substitute for full- tion reduces the ability of Bin1ϩ12 to bind clathrin heavy length Rvs167p (11, 299). chain. Deletion of the NTID from either full-length Bin1ϩ6aϩ12 or Only those central insert domains that contain a 44-residue its isolated BAR domain significantly reduces, but does not abol- domain (residues 378 to 422 in human Bin1ϩ12) exhibit high- ish, association with the sedimentable fraction (257). Hence, the affinity clathrin heavy-chain binding in vitro. This domain is NTID contributes to membrane association but other sequences located at the C terminus of the central insert domain. Within in the BAR domain are sufficient to preserve membrane associ- this domain there are two motifs highly conserved in the se- ation of full-length Bin1ϩ6aϩ12 or its isolated BAR domain in quence of amphiphysin 1 (that also binds clathrin heavy chain), the absence of the NTID. Although yeast Rvs161p and Rvs167p LLDLDFDP and PWDLW (255). Fragments of Bin1ϩ12 lack- also associate with membranes they lack a BAR domain sequence ing either LLDLDFDP or PWDLW bind clathrin heavy chain that resembles the Bin1ϩ6aϩ12ϩ13 NTID. with approximately equal affinity, but deletion of both motifs abrogates clathrin heavy-chain binding. Clathrin triskelia bind to a human Bin1ϩ12 fragment containing residues 378 to 422 Interaction of Bin1؉12 with Other Brain Proteins with an affinity of 3 nM, but deletion of either LLDLDFDP or .(Bin1؉12 (with or lacking exons 6a and 13) interacts with PWDLW reduces the affinity to 9 nM (255 dynamin 1 and synaptojanin 1 in the brain. The first indication Interaction of Bin1؉12 with endophilin A1/SH3p4/SH3GL2 that the brain isoforms of Bin1 may function in endocytosis was in the brain. In addition to clathrin heavy chain and ␣-adaptin/ their ability to bind the same set of endocytic proteins as AP-2, the central insert domain of Bin1ϩ12 also binds the amphiphysin 1. Bin1ϩ6aϩ12ϩ13 binds several dynamin iso- other BAR domain protein endophilin A1 (also called SH3p4 forms, including brain dynamin 1 and also brain synaptojanin 1 and SH3GL2) (Fig. 1). This interaction is direct and mediated (23, 168, 187, 256). The interaction of Bin1ϩ6aϩ12ϩ13 with by the SH3 domain of endophilin A1 and residues 335-378 of ϩ dynamin 1 is direct and the Kd is 240 nM (168, 187). The Bin1 the central insert domain of Bin1 12. Residues 335 to 378 of VOL. 70, 2006 BAR DOMAIN PROTEINS 95 the central insert domain contain one proline-rich motif that Dephosphorylation occurs concomitant with synaptic vesicle fits the consensus for SH3 domain binding and may bind the exocytosis, suggesting that Bin1ϩ6aϩ12 may have to be de- endophilin A1 SH3 domain (193, 255). phosphorylated to become active and that Bin1ϩ6aϩ12 may Bin1؉6a؉12 forms homodimers and heterodimers in the play a role in synaptic vesicle recycling via endocytosis like brain. Bin1ϩ12 isoforms that contain exon 6a (Bin1ϩ6aϩ12) amphiphysin 1 and dynamin 1. bind amphiphysin 1 and fractionate on density gradients with The interaction between amphiphysin 1 and Bin1ϩ6aϩ12 is amphiphysin 1 during subcellular fractionation. Chemical not inhibited by phosphorylation of either protein, however, as cross-linking experiments confirmed the existence of 1:1 het- observed for amphiphysin 1, association of Bin1ϩ6aϩ12 with Downloaded from erodimers of amphiphysin 1 and Bin1ϩ6aϩ12 in the brain. dynamin 1, synaptojanin 1, clathrin heavy chain, and AP-2 is Heterodimer formation is mediated by residues 1 to 150 of the inhibited by phosphorylation (of either Bin1ϩ6aϩ12 or its amphiphysin 1 BAR domain and residues 1 to 328 of the partner) (302). Hyperphosphorylation of Bin1ϩ6aϩ12 occurs Bin1ϩ6aϩ12 BAR domain and the latter also mediate in response to treatment of neurons with the protein kinase C Bin1ϩ6aϩ12 homodimer formation. The formation of ho- activator phorbol 12-myristate 13-acetate. This hyperphosphor- modimers or heterodimers by the Bin1ϩ6aϩ12 BAR domain ylation is inhibited by pretreatment with the protein kinase C was considerably reduced, but not abolished, by deletion of the inhibitor Ro 31-8220, suggesting that protein kinase C is re- NTID. Hence, the NTID contributes to dimer formation but sponsible for Bin1ϩ6aϩ12 phosphorylation in synapses (367). /other sequences in the BAR domain are sufficient to maintain Bin1؉6a؉12 functions in synaptic vesicle recycling/endocy- http://mmbr.asm.org dimer formation when the NTID is absent (257, 302, 367). tosis. Does Bin1ϩ6aϩ12 function in endocytosis? A hint that Interaction of amphiphysin 1/Bin1؉6a؉12 heterodimers it may came from studies that tested the effect of transient with dynamin 1 in the brain. Amphiphysin 1/Bin1ϩ6aϩ12 overexpression of amphiphysin 1, Bin1ϩ6aϩ12, or both pro- heterodimers associate with dynamin 1 in a 1:2 stoichiometry teins on receptor-mediated endocytosis of transferrin in COS-7 in the brain, suggesting that each SH3 domain in the het- cells. Overexpression of either amphiphysin 1 or Bin1ϩ6aϩ12 erodimer binds one dynamin 1 molecule (367). The assembly had a similar deleterious effect on transferrin endocytosis, but of dynamin 1 monomers into oligomers in vitro stimulates in contrast coexpression of amphiphysin 1 and Bin1ϩ6aϩ12 dynamin 1 GTPase activity (353). Amphiphysin 1/Bin1ϩ6aϩ12 did not. Since amphiphysin 1 and Bin1ϩ6aϩ12 both bind dy- heterodimers stimulate dynamin 1 GTPase activity in vitro, namin 1 via their SH3 domains, this result suggests that over- on October 26, 2015 by University of Queensland Library suggesting that their ability to bind two dynamin 1 molecules expression of the individual proteins may lead to sequestration via their SH3 domains may promote dynamin 1 assembly into of dynamin 1 into amphiphysin 1-dynamin 1 or Bin1ϩ6aϩ12- rings in solution (367). Stimulation of dynamin 1 GTPase ac- dynamin 1 complexes that are inactive for endocytosis. Perhaps tivity by amphiphysin 1/Bin1ϩ6aϩ12 heterodimers is likely to only amphiphysin 1/Bin1ϩ6aϩ12 heterodimers are able to be important in vivo during dynamin 1-mediated severing of functionally interact with dynamin 1 and stimulate endocytosis. the necks of deeply invaginated clathrin-coated pits. A subsequent study showed an association of other clathrin The three-dimensional crystal structure of the Bin1ϩ6aϩ12 coat components, including dynamin 1, synaptojanin 1, Eps15, SH3 domain has been solved to 2.2 Å resolution (232). Unlike and AP-180, with the dynamin 1 PRD in vitro that is depen- most other SH3 domains, the Bin1ϩ6aϩ12 SH3 domain has a dent on binding of amphiphysin 1/Bin1ϩ6aϩ12 heterodimers highly acidic patch positioned close to the binding cleft. The to the dynamin 1 PRD (302). existence of this acidic patch explains the curious requirement Further evidence for a role of Bin1ϩ6aϩ12 in clathrin- for two basic (arginine) residues in the proline-rich motif dependent endocytosis came from studies similar to those of within the dynamin PRD recognized by this SH3 domain (168). Shupliakov et al. (293), Wigge et al. (369), and Volchuk et al. The Bin1ϩ6aϩ12 SH3 domain also has a long insert se- (349) that first showed amphiphysin 1 has a role in endocytosis. quence not found in other SH3 domains that is predicted to Transient overexpression of the Bin1ϩ6aϩ12 SH3 domain in- form an extended loop near the dynamin-binding site. The hibited receptor-mediated endocytosis of transferrin in intact Bin1ϩ6aϩ12 SH3 domain not only binds dynamin, but also cells. Moreover, the Bin1ϩ6aϩ12 SH3 domain caused potent interferes with dynamin assembly into rings in vitro. Deletion inhibition of in vitro assays for clathrin-dependent endocytosis of the insert loop abolishes the ability of the Bin1ϩ6aϩ12 SH3 (half-maximal inhibition at 6 ␮M). The Bin1ϩ6aϩ12 SH3 do- domain to inhibit dynamin assembly in vitro and receptor- main inhibited a target present in the membrane fraction and mediated endocytosis of transferrin in vivo. Moreover, transfer most potently inhibited the late step of membrane fission and of this insert loop to an unrelated dynamin-binding SH3 do- clathrin-coated vesicle release. The inhibition by the Bin1ϩ main (Grb2) confers the ability to inhibit dynamin assembly in 6aϩ12 SH3 domain is specific, because control SH3 domains vitro (232). did not inhibit endocytosis in this in vitro assay even at much .(Bin1؉6a؉12 is regulated by phosphorylation and dephos- higher concentrations (295 phorylation. One study did not observe significant phosphor- The role of Bin1ϩ12 in endocytosis and synaptic vesicle ylation of Bin1ϩ6aϩ12 in resting synapses, in contrast to am- recycling has also been examined using knockout mice. phiphysin 1 (193). Subsequent studies, however, showed that Bin1ϩ12 expression in the brain is critically dependent on when brain extract is incubated with ATP and protein phos- amphiphysin 1 expression. Brain extracts of amphiphysin 1-de- phatase inhibitors, Bin1ϩ6aϩ12, like dynamin 1, synaptojanin ficient mice contained normal levels of Bin1ϩ12 (with or lack- 1, clathrin heavy chain, and AP-2, undergoes a shift to lower ing exons 6a and 13) transcript, but were almost devoid of electrophoretic mobility suggestive of phosphorylation. As in Bin1ϩ12 protein. This suggests that in the brain only am- the case of the other endocytic proteins, Bin1ϩ6aϩ12 is rap- phiphysin 1/Bin1ϩ12 heterodimers are stable. It is likely that idly dephosphorylated upon nerve terminal depolarization. Bin1ϩ12 is subject to proteolysis when amphiphysin 1 is ab- 96 REN ET AL. MICROBIOL.MOL.BIOL.REV. sent. The loss of Bin1ϩ12 expression in the brain of amphiphy- Bin1ϩ12 isoforms may not only function in the internalization sin 1 knockout mice is specific. No reduction in expression step, but may also direct postinternalization traffic through level in the brain was observed for other synaptic proteins that endosomes (167). interact with amphiphysin 1. Moreover, the loss of Bin1 ex- pression is brain specific because Bin1ϩ10 was expressed nor- Bin1؉10؉13/Bin1 mally in the muscle of amphiphysin 1 knockout mice (am- phiphysin 1 is not expressed in muscle). Despite the loss of Identification of Bin1؉10؉13/Bin1. c-Myc is an important both amphiphysin 1 and Bin1ϩ12, amphiphysin 1 knockout regulator of cell proliferation versus terminal differentiation Downloaded from mice exhibit relatively mild defects in synaptic vesicle recycling decisions and is mutated in many human cancers. Nondividing that become apparent only under conditions of continuous cells do not express c-Myc, but after mitogenic stimulation nerve stimulation (55). c-Myc is one of the first proteins to be induced before cells Mice homozygous for a deletion of the BIN1 gene that enter the cell cycle. The role of c-Myc in cell proliferation is encodes Bin1 exhibit postnatal inviability. Although Bin1ϩ highlighted by recent studies using Drosophila c-Myc. Overex- 6aϩ12 is highly expressed in the brain, the brains of homozy- pression of Drosophila c-Myc in select cells in a tissue induces gous Bin1 knockout mice appear to possess normal structures. these cells to proliferate at the expense of surrounding low- Interestingly, amphiphysin 1 expression in the brain is not expressor cells in the same tissue, i.e., to become “supercom- dependent on Bin1. Hence, while in yeast each Rvs protein is petitors.” Indeed, in these competition experiments the low- http://mmbr.asm.org/ unstable in the absence of the other, in mouse brain only expressor cells are not only overgrown by the high-expressors Bin1ϩ6aϩ12ϩ13 is unstable in the absence of its partner. As but undergo apoptosis (199). However, high c-Myc expression expected from analysis of homozygous amphiphysin 1 knock- is not always advantageous. When mammalian cells in culture out mice that also lack Bin1 in the brain, homozygous Bin1 are serum starved c-Myc expression is down-regulated and knockout mice also show no severe endocytic defects in the cells stop dividing and become quiescent. Ectopic c-Myc ex- brain. Cultured Bin1-deficient neurons exhibit normal neurite pression in serum-starved cells prevents the cells from becom- outgrowth and form synapses in vitro. There is no obvious ing quiescent. c-Myc-overexpressing cells that continue to divide accumulation of clathrin-coated pits or vesicles at these syn- under serum-starved conditions eventually undergo c-Myc-de- apses. Bin1-deficient embryonic fibroblasts express no detect- pendent apoptosis (278). on October 26, 2015 by University of Queensland Library able Bin1 isoforms but grow and divide as rapidly in culture as c-Myc has been shown to shuttle between the cytoplasm and wild-type embryonic fibroblasts, endocytose fluorescein iso- the nucleus and can act as a transcriptional activator or repres- thiocyanate-conjugated transferrin with kinetics even more sor. The N-terminal domain of c-Myc is a transcription activa- rapid than that of wild-type embryonic fibroblasts, and exhibit tion/repression domain and the C-terminal domain is a DNA a normal actin cytoskeleton that includes normal actin stress binding domain. The activity of c-Myc both in oncogenic trans- fibers (207). formation and in apoptosis requires two conserved N-terminal The reason for the apparent discrepancy between the results domains known as Myc boxes 1 (MB1) and 2 (MB2). MB1 is obtained by transient overexpression or microinjection exper- positioned within the transcriptional activation domain of c- iments in cultured cells and the results obtained by gene Myc and is subject to cell cycle-dependent phosphorylation on knockout in mice is not yet clear. It is possible that loss of residues T58 and S62. MB1 is critically important for control of endocytic function is only observed transiently upon acute per- cell proliferation since most c-Myc mutations in primary hu- turbation of Bin1ϩ6aϩ12 function. In knockout cells that have man tumors and cultured tumor cell lines map to the MB1 been without Bin1 expression for some time, long-term adap- domain. The observation that MB1 is mutated in tumors sug- tation processes that compensate for loss of Bin1may have gests a negative regulator of c-Myc may interact via the MB1 been activated. Alternatively, in the transient overexpression domain and mutation of MB1 may result in loss of this inter- or microinjection experiments there may have been effects that action and deregulation of c-Myc activity (278). were not limited to perturbation of Bin1ϩ12 alone. These A Bin1 splice variant (referred to here as Bin1ϩ10ϩ13) was nonspecific effects, rather than loss of Bin1ϩ12 interactions, isolated in a two-hybrid screen with the c-Myc MB1 domain as may have been the cause of the observed endocytic defect. the bait and named Box-dependent Myc-interacting protein Bin1؉12 may function in postinternalization transport (or subsequently bridging integrator) 1 (Bin1) (Fig. 1 and through endosomes. While attention has mainly focused on the Table 2) (276, 278). As predicted for a negative regulator of possible role of Bin1ϩ12 isoforms in endocytosis at clathrin- c-Myc, expression of Bin1ϩ10ϩ13 inhibited the ability of c- coated pits on the plasma membrane, one study found that Myc to induce oncogenic transformation in culture (69, 278). Bin1ϩ6aϩ12ϩ13 binds sorting nexin 4 (SNX4), a protein that Domain structure of Bin1؉10؉13. In comparison to Bin1ϩ localizes predominantly to intracellular compartments. Inter- 6aϩ12 (which can either have or lack exon 13), Bin1ϩ10ϩ13 action is mediated by the C-terminal part of the Bin1ϩ lacks the NTID domain encoded by exon 6a that mediates the 6aϩ12ϩ13 BAR domain (BAR-C) and a short sequence at the formation of heterodimers with amphiphysin 1 and plasma extreme C terminusof SNX4 (167). Interestingly, the amino membrane targeting (Fig. 1 and Table 2). It also lacks the acid sequence of SNX4 suggests that SNX4 may possess a central insert domain encoded by exon 12 that in some iso- C-terminal BAR domain such as SNX1 (25). If so, the forms of Bin1ϩ12 (with or lacking exons 6a and 13), depending Bin1ϩ6aϩ12ϩ13-interacting sequence would lie within the on the choice of exon 12, binds clathrin heavy chain, ␣-adaptin, SNX4 BAR domain. A pool of Bin1ϩ6aϩ12ϩ13 localizes and/or endophilin A1 (Fig. 1 and Table 2). Bin1ϩ10ϩ13 con- to SNX4-containing compartments that contain internalized tains, however, an additional 15-residue sequence encoded by transferrin and are therefore a type of early endosome. Hence, exon 10 that includes a nuclear localization sequence and lipid VOL. 70, 2006 BAR DOMAIN PROTEINS 97 binding sequence that is absent in all variants of Bin1ϩ12 (with in turn important for prevention of c-Myc-induced apoptosis in or lacking exons 6a and 13). Bin1ϩ10ϩ13 and some splice mammalian cells and for survival under starvation conditions variants of Bin1ϩ12 (with or lacking exon 6a) contain a com- in yeast cells. Bin1ϩ10ϩ13 and Rvs167p are not functionally plete c-Myc binding sequence, a third of which is encoded by interchangeable, however, as Bin1ϩ10ϩ13 expression in exon 13 (Fig. 1 and Table 2). In this way Bin1ϩ10ϩ13 and rvs167⌬ yeast mutants does not rescue the phenotype (278). Bin1ϩ6aϩ12ϩ13 differ from ALP1/amphiphysin IIm, which Rvs167p also lacks the MBD found in Bin1ϩ10ϩ13 (Rvs167p lack the sequences encoded by exon 13 (see below) (Fig. 1 and has the GPA-rich region in place of the MBD) (Fig. 1). In Table 2) (23, 168, 256, 257, 278, 338, 356, 357, 367). Hence, addition, yeast does not have an ortholog of mammalian Downloaded from according to the nomenclature of Wechsler-Reya et al., this c-Myc. original Bin1 isoform is named Bin1ϩ10ϩ13 (356). Is interaction of Bin1ϩ10ϩ13 with c-Myc important for sup- Interaction of Bin1؉10؉13 with c-Myc. The interaction of pression of oncogenic transformation? Deletion of the MBD Bin1ϩ10ϩ13 with c-Myc was originally identified by two-hy- abolishes the ability of Bin1ϩ10ϩ13 to suppress oncogenic brid screens but has been confirmed by protein binding in vitro transformation by c-Myc. Moreover, expression of a frag- and is direct (278). c-Myc interacts with a central domain of ment of Bin1ϩ10ϩ13 comprising only the MBD interferes in Bin1ϩ10ϩ13 located between the N-terminal BAR domain a dominant negative manner with the ability of full-length and the C-terminal SH3 domain and this domain has been Bin1ϩ10ϩ13 to suppress oncogenic transformation by c-Myc. named the Myc-binding domain (MBD, residues 270 to 377). This suggests binding of Bin1ϩ10ϩ13 to c-Myc is physiologi- http://mmbr.asm.org/ A recent study, however, found that the SH3 domain of cally important. Interestingly, mutations in MB1 that do not Bin1ϩ10ϩ13 mediates binding to c-Myc (245). Bin1ϩ10ϩ13 appear to affect Bin1ϩ10ϩ13 interaction but that prevent binding to c-Myc in vitro requires both MB1 and MB2 on MB1 phosphorylation (e.g., T58M) also abolish suppression c-Myc, although in the two-hybrid assay MB1 is sufficient activity. This shows that Bin1ϩ10ϩ13 binding to MB1 is nec- (278). essary but not sufficient for c-Myc suppression (69, 278). An initial study did not observe coimmunoprecipitation of Another Bin1ϩ10ϩ13 domain essential for suppression of c-Myc and Bin1ϩ10ϩ13 from cell extracts, suggesting these oncogenic transformation by c-Myc is the C-terminal half of proteins may not exist in a stable complex in vivo (278). Prior the BAR domain (BAR-C) (Fig. 1). Deletion of the BAR-C to differentiation into myotubes, proliferating C2C12 cells ex- domain causes partial redistribution of Bin1ϩ10ϩ13 from the on October 26, 2015 by University of Queensland Library press both c-Myc and Bin1ϩ10ϩ13, however, after differenti- nucleus to the cytoplasm. Moreover, the BAR-C domain con- ation c-Myc expression is lost and Bin1ϩ10ϩ13 expression is tains a nuclear localization sequence. The effect of the BAR-C dramatically elevated (357). A subsequent study using prolif- deletion shows that this nuclear localization sequence, rather erating C2C12 cells found endogenous Bin1ϩ10ϩ13 and c- than that encoded by exon 10, is the critical sequence for Myc do coimmunoprecipitate from cell extracts and hence do Bin1ϩ10ϩ13 localization to the nucleus. Deletion of Bin1ϩ form stable complexes in vivo, at least in this cell type. 10ϩ13 domains other than the MBD and BAR-C domains had Bin1ϩ10ϩ13 also coimmunoprecipitates with c-Myc from ex- little or no effect on suppression of c-Myc transformation. The tracts prepared from insect Sf9 cells ectopically expressing ability of Bin1ϩ10ϩ13 to inhibit proliferation of tumor cells both proteins (69). that lack endogenous Bin1ϩ10ϩ13 (e.g., HepG2) is critically Bin1؉10؉13 is a tumor suppressor. Bin1ϩ10ϩ13 is a neg- dependent on the presence of BAR-C. In contrast, despite the ative regulator of cell cycle progression and plays a role in exit importance of the MBD in suppression of oncogenic transfor- from the cell cycle in response to serum deprivation. The mation by c-Myc, Bin1ϩ10ϩ13 lacking the MBD retains signif- ability of Bin1ϩ10ϩ13 to suppress oncogenic transformation icant ability to inhibit tumor cell proliferation, suggesting this by c-Myc demonstrates that Bin1ϩ10ϩ13 has properties char- activity may be partially independent of c-Myc binding (69). acteristic of a tumor suppressor. Moreover, Bin1ϩ10ϩ13 may Further insight into the role of Bin1ϩ10ϩ13 in suppression be part of an important mechanism preventing oncogenic of oncogenic transformation by c-Myc came from a study of a transformation and tumorigenesis in vivo, since the domain of Bin1 splice variant that lacks exon 10 but contains neuron- c-Myc that mediates Bin1ϩ10ϩ13 binding (MB1) is frequently specific exon 12 (Bin1ϩ12ϩ13) and whose expression in non- mutated in tumor cells. Moreover, the human and murine neuronal cells correlates with oncogenesis (93, 245, 356). A BIN1 genes map to chromosomal loci that are hotspots for highly sensitive protein interaction assay revealed that many deletion in tumors and Bin1ϩ10ϩ13 expression is absent in a Bin1 isoforms, including those that lack the MBD encoded in high percentage of tumor cell lines and primary breast and part by exon 13, do in fact interact with c-Myc. In this study the other tumors. Furthermore, ectopic expression of Bin1ϩ isolated MBD was not sufficient to bind c-Myc in vitro. The 10ϩ13 in tumor cells that have lost endogenous Bin1ϩ10ϩ13 novel c-Myc interaction involves the C-terminal SH3 domain expression causes arrest of cell proliferation. This strongly of various Bin1 isoforms interacting with a proline-rich motif supports the idea that loss of Bin1ϩ10ϩ13 expression contrib- positioned within the MB1 domain of c-Myc. This interaction utes significantly to tumorigenesis and is not simply a conse- is regulated by phosphorylation of S62 within MB1. quence of genetic instability in tumor cells (278). Intriguingly, Bin1ϩ12ϩ13 differs from all other Bin1 iso- This role of Bin1ϩ10ϩ13 as a negative regulator of cell cycle forms tested in that its SH3 domain is unable to bind c-Myc. progression is reminiscent of the role of the yeast Rvs proteins The explanation comes from the finding that exon 12 contains (12, 42). While Bin1ϩ10ϩ13 plays a role in exit of serum- a proline-rich motif. When exon 12 is present the SH3 domain starved mammalian cells from the cell cycle (278), the Rvs engages in an intramolecular interaction with exon 12 that proteins play a role in exit of nutrient-starved yeast cells from prevents its engagement with c-Myc (245). ϩ ϩ the cell cycle (12). Exit from the cell cycle and arrest in G0 are The effect of Bin1 10 13 may be on the transcriptional 98 REN ET AL. MICROBIOL.MOL.BIOL.REV. activity of c-Myc. c-Myc stimulates transcription of various Bin1 plays a critical role in tumor suppression in vivo (69, 245, genes, including those encoding ornithine decarboxylase and 278). Further support for an important role for Bin1ϩ10ϩ13 ␣-prothymosin and also stimulates transcription of minimal in tumor suppression in vivo comes from analysis of the neo- viral promoters that have been engineered to include upstream plastic potential of murine embryonic fibroblasts (MEFs) from c-Myc binding elements. The ability of c-Myc to activate tran- homozygous Bin1 knockout mice. Bin1 knockout MEFs, but scription from each promoter was inhibited by a construct not wild-type MEFs, cotransfected with c-Myc and a mutant expressing Bin1ϩ10ϩ13. Deletion of the Bin1ϩ10ϩ13 MBD form of Ras exhibited morphological features of oncogenically abolished its ability to inhibit c-Myc-dependent activation of transformed cells. Coexpression of c-Myc and mutant Ras in Downloaded from some promoters (e.g., ornithine decarboxylase) but not other Bin1 knockout MEFs, but not wild-type MEFs, also led to an promoters (e.g., ␣-prothymosin). Hence, Bin1ϩ10ϩ13 inhibits increased ability to form tumors in mice. This effect was spe- c-Myc-dependent transcriptional activation but does so by dis- cific to c-Myc, as Bin1 knockout MEFs transfected with viral tinct mechanisms, only one of which requires binding of c-Myc oncogenes did not exhibit increased tumor formation (209). via the MBD. Bin1ϩ10ϩ13 also prevents tumor cells from being recog- The effect of Bin1ϩ10ϩ13 on transcriptional activation is nized and killed by anti-tumor T cells. Bin1ϩ10ϩ13 has re- specific to c-Myc. Bin1ϩ10ϩ13 does not inhibit transcriptional cently been shown to be a key regulator of the immunoregula- activation by other proteins, e.g., herpesvirus VP16. The N- tory enzyme indoleamine 2,3-dioxygenase (IDO). Overexpression terminal domain of c-Myc that contains MB1 and MB2 confers of IDO in Bin1 knockout cells may be a major factor in im- http://mmbr.asm.org/ inhibition of transcriptional activation by Bin1ϩ10ϩ13, as fu- mune escape and accelerated progression of tumor cells, as sion of this domain to the yeast Gal4p transcriptional activator treatment of tumor-bearing mice with IDO inhibitors helps results in the ability of Bin1ϩ10ϩ13 to inhibit transcriptional induce tumor regression (208). .activation of Gal4p-specific promoters (69). Bin1؉10؉13 induces apoptosis specifically in tumor cells How does Bin1ϩ10ϩ13 inhibit transcriptional activation by One mechanism by which Bin1ϩ10ϩ13 may suppress tumors c-Myc? Initially, it was proposed that Bin1ϩ10ϩ13 binding to is by inducing apoptosis (programmed cell death). Ectopic MB1 and MB2 acts sterically to inhibit the ability of MB1 and expression of Bin1ϩ10ϩ13 in the cultured tumor cell line MB2 to activate transcription at c-Myc-dependent promoters HepG2 (which has lost endogenous Bin1ϩ10ϩ13 expression) (278). Subsequently, it was suggested that Bin1ϩ10ϩ13 re- induces cell death via activation of an apoptotic pathway. This on October 26, 2015 by University of Queensland Library cruits a transcriptional repressor to c-Myc. Fusion of Bin1ϩ apoptotic pathway involves reduction of cell volume, loss of 10ϩ13 to Gal4p is sufficient to repress transcription at a Gal4p- adherence, accumulation of cytoplasmic vacuoles, and DNA dependent promoter and this repression can be relieved by degradation. This apoptotic role of Bin1ϩ10ϩ13 in tumor cells coexpression of unfused Bin1ϩ10ϩ13, perhaps via titration of requires the BAR-C domain. Apoptosis and DNA degradation a transcriptional repressor. Interestingly, the Bin1ϩ10ϩ13 are induced independently of cell cycle stage. This apoptotic MBD is not required either for repression of Gal4p-dependent mechanism does not require the tumor suppressor p53 or ret- transcription or for relief of this repression. Hence, the puta- inoblastoma protein, caspases, Bcl-2, or Fas-dependent apo- tive repressor must bind a domain of Bin1ϩ10ϩ13 distinct ptotic signaling. Inhibition of the apoptotic pathway induced by from the MBD. The nature of the repressor recruited by c-Myc upon serum withdrawal also inhibits DNA degradation Bin1ϩ10ϩ13 is not known (69). induced by Bin1ϩ10ϩ13, suggesting some apoptotic effects of Bin1ϩ10ϩ13 activity in tumor suppression is not restricted Bin1ϩ10ϩ13 are mediated via mechanisms common to c-Myc- to effects on c-Myc. Bin1ϩ10ϩ13 also suppresses oncogenic dependent apoptosis (68). transformation by the unrelated adenovirus E1A oncoprotein The role of Bin1ϩ10ϩ13 in c-Myc-dependent apoptosis has and by a dominant negative mutated form of the tumor sup- been tested in vivo using homozygous Bin1 knockout mice. pressor p53. The mechanism by which Bin1ϩ10ϩ13 suppresses Ectopic expression of c-Myc in primary homozygous Bin1 transformation by these oncoproteins appears distinct, how- knockout MEFs still induced apoptosis upon serum withdrawal ever, as neither the MBD nor the BAR-C domain is essential. (207). This is perhaps not unexpected, since Bin1ϩ10ϩ13 has Moreover, the Bin1ϩ10ϩ13 MBD does not bind E1A in vitro. only been shown to be important for induction of apoptosis in A short sequence (U1) encoded by exon 9 is essential for oncogenically transformed cells (e.g., HepG2) (68). Ectopic suppression of both adenovirus E1A and mutant p53, while the expression of c-Myc in homozygous Bin1 knockout MEFs is SH3 domain is also essential for suppression of mutant p53. insufficient to induce oncogenic transformation (207). Adenovirus E1A and p53 both function in the nucleus, and Subcellular localization of Bin1؉10؉13. The subcellular hence the partial loss of nuclear localization of Bin1ϩ10ϩ13 localization of Bin1ϩ10ϩ13 has been an area in which differ- caused by deletion of the BAR-C domain may not fully ac- ent findings have been reported. Bin1ϩ10ϩ13 was initially count for loss of c-Myc suppression. Bin1ϩ10ϩ13 does not reported to exhibit a nuclear localization when ectopically ex- inhibit oncogenic transformation by all oncoproteins, however. pressed in HepG2 hepatocarcinoma cells (278). This localiza- For example, Bin1ϩ10ϩ13 has no effect on oncogenic trans- tion is consistent with the ability of Bin1ϩ10ϩ13 to interact formation by the simian virus 40 large T antigen (69, 278). both physically and functionally with the nuclear transcrip- Hence, the mechanisms by which Bin1ϩ10ϩ13 inhibits onco- tional activator c-Myc and the presence of a nuclear localiza- genic transformation and tumorigenesis are specific. tion sequence encoded by exon 10 (69, 278). Examination of The frequency with which human tumors either lack Bin1 endogenous Bin1ϩ10ϩ13 in a collection of human and rodent expression or misexpress Bin1ϩ12ϩ13 together with the ability Bin1ϩ10ϩ13-expressing tumor cell lines showed again that of ectopic Bin1ϩ10ϩ13 expression in tumor cells lacking en- Bin1ϩ10ϩ13 localizes to the nucleus and in part to a sub- dogenous Bin1ϩ10ϩ13 to inhibit proliferation suggests that nuclear compartment. In some tumor cells Bin1ϩ10ϩ13 local- VOL. 70, 2006 BAR DOMAIN PROTEINS 99 izes entirely to the subnuclear compartment (355). In another differentiation: Bin1ϩ10ϩ13 is a negative regulator of cell study the expression level and subcellular localization of cycle progression, it is encoded by a transcript that undergoes Bin1ϩ10ϩ13 in various normal human tissues were compared differential splicing during differentiation of proliferating and differences were observed between tissues (e.g., bone mar- C2C12 myoblasts into terminally differentiated myotubes, and row, breast, and intestine). Within intestinal epithelia, an ex- it is highly expressed in adult skeletal muscle. Moreover, pression gradient of Bin1ϩ10ϩ13 was observed, with strongest Bin1ϩ10ϩ13 was identified in a large-scale screen of muscle- expression in those cells located at the tips of intestinal pro- specific proteins as a protein required for muscle cell differ- jections (villi) that are committed to undergo apoptosis (61). entiation. The murine skeletal muscle-derived cell line C2C12 Downloaded from Subsequently, two studies found Bin1ϩ10ϩ13 is predomi- can be induced to differentiate into muscle in culture. Conflu- nantly nonnuclear in muscle cells, where this isoform is endo- ent C2C12 myoblasts subjected to serum withdrawal in culture genously expressed at high levels (23, 162). When transiently undergo dramatic elongation and then lateral cell-cell fusion to expressed in COS-7 cells, both brain Bin1ϩ6aϩ12ϩ13 and form huge multinucleate myotubes similar to those found in muscle Bin1ϩ10ϩ13 were found to localize exclusively to the muscle. When Bin1ϩ10ϩ13 is down-regulated differentiation cytoplasm. Even when ectopically expressed in HepG2 cells the of C2C12 myoblasts into myotubes upon serum withdrawal is localization of Bin1ϩ10ϩ13 was predominantly cytoplasmic. blocked (172). The reason for the different subcellular localizations reported To investigate the role of Bin1ϩ10ϩ13 in muscle cell dif- for Bin1ϩ10ϩ13 is not known. It has been suggested that ferentiation its expression during C2C12 differentiation in cul- http://mmbr.asm.org/ Bin1ϩ10ϩ13 may shuttle between the cytoplasm and the nu- ture was examined. Bin1ϩ10ϩ13 transcript and protein levels cleus (23). This process may be regulated in response to some were observed to increase upon serum withdrawal concomitant yet-to-be-identified environmental signal. with terminal differentiation and cell-cell fusion into myotubes. During differentiation of C2C12 myoblasts into myotubes in c-Myc is a marker of proliferating cells. As expected, c-Myc vitro there is a switch of splice variant from Bin1Ϫ10 to expression levels declined after serum withdrawal as cells ex- Bin1ϩ10 (each with or lacking exon 13) (357). Both the former ited the cell cycle and became committed to terminal differ- (including SH3p9 and ALP1/amphiphysin IIm) and the latter entiation. Interestingly, higher-molecular-weight Bin1 protein have been reported to be present strictly in the cytoplasm, species appear during differentiation. These forms arise by strictly in the nucleus, or distributed between the cytoplasm altered transcript splicing such that exon 10 is included. Bin1 in on October 26, 2015 by University of Queensland Library and nucleus in different studies (23, 69, 141, 162, 278, 355). undifferentiated C2C12 cells lacks sequences encoded by exon Hence, while exon 10 encodes a predicted nuclear localization 10 and localizes predominantly to the nucleus. In contrast, the sequence, this exon is not required for nuclear localization of form of Bin1 found in terminally differentiated C2C12 myo- Bin1. As mentioned above, a second nuclear localization se- tubes contains sequences encoded by exon 10 and forms a quence that is critical for nuclear localization is present in the filamentous network in the cytoplasm (357). BAR-C domain (69). It is worth noting that the antibodies Expression of Bin1ϩ10ϩ13 has also been examined during used for immunofluorescence localization of Bin1ϩ10 may not embryonic development in mice (182). By embryonic day 10.5 have distinguished between those splice variants that include expression of Bin1ϩ10ϩ13 has already been induced in myo- exon 10 sequences and those that lack exon 10 sequences. The tomes, which are the precursors of skeletal muscle cells. domain encoded by exon 10 is required for the plasma mem- Bin1ϩ10ϩ13 appears to regulate cell cycle progression via brane binding and tubulating activity of Bin1ϩ10ϩ13 (see be- induction of the Cdk inhibitor p21WAF1. Down-regulation of low) (162). Perhaps this property is responsible for a shift from Bin1ϩ10ϩ13 in C2C12 cells prevents induction of p21WAF1 a nuclear to a cytoplasmic distribution when this exon is incor- and hence interferes with cell cycle arrest upon serum with- porated. drawal. Hence, Bin1ϩ10ϩ13 functions at a very early stage in .(Stability of Bin1؉10؉13 in vivo. The in vivo stability of differentiation of C2C12 myoblasts into myotubes (182 Bin1ϩ10ϩ13 has been examined by pulse-chase radiolabeling Overexpression of human Bin1ϩ10ϩ13 slows the prolif- and Bin1ϩ10ϩ13 is rather unstable in cells, with a half-life of eration of C2C12 cells in culture and promotes differentia- only about 2 h (355). This is interesting in view of recent tion into myotubes. A decrease in the rate of cell prolifer- evidence that Rvs167p in yeast interacts with a ubiquitin pro- ation is observed even in the presence of serum, however, tein ligase (Rsp5p, the yeast ortholog of Nedd4) and is subject differentiation still requires serum withdrawal. Lowering to ubiquitination (307), and that Rvs167p is rapidly degraded Bin1ϩ10ϩ13 expression increases the rate of cell prolifer- in the absence of Rvs161p (180). ation in the presence of serum and also abolishes differen- -Phosphorylation of Bin1؉10؉13. Like amphiphysin 1 and tiation upon serum withdrawal. C2C12 cells in which expres other splice variants of Bin1, Bin1ϩ10ϩ13 is phosphorylated sion of Bin1ϩ10ϩ13 is down-regulated do not elongate, nor in vivo (355). The role of this phosphorylation, the sites phos- do they undergo cell-cell fusion to form multinucleate myo- phorylated, and the kinase(s) responsible are not yet known. tubes. Induction of muscle myosin II is normally observed Interestingly, Bin1ϩ10ϩ13 lacks a region of homology to the in C2C12 cells prior to cell-cell fusion, however, after proline-rich sequences in the central insert domain of am- Bin1ϩ10ϩ13 down-regulation serum withdrawal no longer phiphysin 1 and GPA-rich domain of Rvs167p that are subject induces myosin II expression. The ability of Bin1ϩ10ϩ13 to phosphorylation by Cdk5/Pho85p (78, 85, 330). Therefore, overexpression to slow cell proliferation is dependent on the phosphorylation of Bin1ϩ10ϩ13 may have some unique fea- presence of exon 10. In contrast, exon 10 is not essential for tures compared to that of amphiphysin 1 or yeast Rvs167p. differentiation. Accelerated differentiation of C2C12 myo- Role for Bin1؉10؉13 in muscle differentiation. Several blasts occurs upon serum withdrawal even in cells expressing findings suggested that Bin1ϩ10ϩ13 may play a role in muscle a form of Bin1ϩ10ϩ13 in which exon 10 is deleted. Hence, 100 REN ET AL. MICROBIOL.MOL.BIOL.REV. increased expression of Bin1ϩ10ϩ13 (although not exon between Bin1ϩ10ϩ13 and the Z-line. The insulin-regulated 10) is essential for differentiation of C2C12 myoblasts into plasma membrane glucose transporter Glut4 is a protein that myotubes in culture (162, 357). undergoes clathrin-dependent endocytosis. Glut4 localization A role for Bin1ϩ10ϩ13 in cell differentiation is consistent in skeletal muscle extensively overlaps that of both clathrin with aspects of the function of the Rvs161p and Rvs167p pro- heavy chain and Bin1ϩ10ϩ13 (23). teins in yeast. In budding yeast two cell differentiation path- Interestingly, in the fruit fly Drosophila melanogaster the ways have been described, sporulation and pseudohyphal single amphiphysin ortholog is enriched in postsynaptic neu- growth. The sporulation of vegetative diploid cells to produce romuscular junctions in larvae. In contrast, Drosophila am- Downloaded from haploid spores is dependent on the function of both Rvs161p phiphysin expression is low or absent in the central nervous and Rvs167p (10, 12, 38, 53). This suggests a possible evolu- system. Flies in which the amphiphysin gene is disrupted are tionarily conserved role in cell differentiation. Furthermore, viable and exhibit normal synaptic vesicle recycling and neu- the importance of Bin1ϩ10ϩ13 expression for the cell-cell rotransmission, but display severe locomotory defects in both fusion of myoblasts to form multinucleate myotubes is remi- larvae and adult flies (e.g., the adults cannot fly). The function niscent of the role of Rvs161p/Fus7p and Rvs167p in fusion of of the neuromuscular junction appears normal. Amphiphysin haploid yeast cells during mating (20, 21, 91). Mating leads localizes to T-tubules in muscle and amphiphysin-deficient flies initially to the formation of cells with two haploid nuclei have a disorganized T-tubule system. This suggests the loco- (multinucleate), however, subsequent nuclear fusion results in motory defects in amphiphysin-deficient flies result from inef- http://mmbr.asm.org/ a single diploid nucleus. Fusion of myoblasts requires elevated ficient excitation-contraction coupling in which the T-tubule expression of type II conventional muscle myosin and this is plays a critical role. Moreover, although Drosophila amphiphy- dependent on Bin1ϩ10ϩ13 (357). It is interesting that in yeast, sin has both an N-terminal BAR domain and C-terminal SH3 of all the myosins, the strongest genetic interactions of RVS161 domain highly conserved with mammalian amphiphysin 1 and and RVS167 are with MYO2, which encodes the sole conven- Bin1 isoforms, the central insert domain of mammalian am- tional type II myosin (Myo1p) (19). phiphysin 1 and Bin1ϩ12 that interacts with clathrin heavy -Bin1؉10؉13 localization in adult skeletal muscle. In skel- chain and AP-2 is not well conserved in Drosophila amphiphy etal muscle, Bin1ϩ10ϩ13 localizes to distinctive transverse sin. Indeed, Drosophila amphiphysin does not bind clathrin bands that appear with regular periodicity along the muscle heavy chain (169, 258, 259, 386). Hence, Drosophila amphiphy- on October 26, 2015 by University of Queensland Library fiber. The Bin1ϩ10ϩ13-specific bands were compared with sin resembles mammalian Bin1ϩ10ϩ13. known bands of muscle fiber and found to be within the broad Bin1؉10؉13 and T-tubule formation in muscle. Although actin filament-containing I-band and positioned on either side endogenous amphiphysin 1 and both neuronal Bin1ϩ6aϩ12 of the thinner desmin-containing Z-line. The submembranous (with or lacking exon 13) and muscle Bin1ϩ10ϩ13 localize to cytoskeletal protein ankyrin 3 localizes to what appear to be the cortical cytomatrix, they differ in their ability to localize the same bands as Bin1ϩ10ϩ13 within the muscle fiber (23). to the cortex when ectopically expressed in CHO cells. Am- Ankyrin 3 is a specific marker for invaginations of the muscle phiphysin 1 and neuronal Bin1ϩ6aϩ12ϩ13 both exhibit a pre- cell plasmalemma known as transverse (T)-tubules (79). That dominantly diffuse localization throughout the cytoplasm. In Bin1ϩ10ϩ13 localizes to T-tubules in muscle was confirmed by contrast, muscle Bin1ϩ10ϩ13 concentrates at the cytoplasmic colocalization with another T-tubule-specific marker, triadin face of the plasma membrane (162). (108), and immunoelectron microscopy of ultrathin skeletal Ectopic expression of muscle Bin1ϩ10ϩ13, but not am- muscle frozen sections (23). Interestingly, like axon initial seg- phiphysin 1 or neuronal Bin1ϩ6aϩ12ϩ13, in CHO cells that ments and nodes of Ranvier in the brain, T-tubules in muscle normally lack T-tubules leads to the dramatic appearance of bear on their cytoplasmic face a dense cytomatrix and it is to tubules that resemble T-tubules. These tubules are contiguous this cytomatrix to which ankyrin 3 localizes. Bin1ϩ10ϩ13 may with the plasma membrane and are enriched in PtdIns(4,5)P2. function within this cortical cytomatrix to regulate traffic to and The ability of Bin1ϩ10ϩ13 to induce membrane tubules in from T-tubules and may play a role in defining these special- nonmuscle cells requires the BAR domain and the 15-residue ized membrane subdomains (23, 79, 162). muscle-specific domain encoded by exon 10 (Fig. 1) (162). This T-tubules are membrane tubules that extend from the plas- domain is highly basic with the sequence RKKSKLFSRLR malemma to form an extensive intracellular membrane retic- RKKN and contains a putative nuclear localization sequence ulum. This T-tubule membrane reticulum is closely associated originally thought to confer nuclear localization on with the bundles of muscle fibers within muscle cells. T-tubules Bin1ϩ10ϩ13 (278). Full-length Bin1ϩ10ϩ13 is able to recruit have been proposed to play a key role in passing the Naϩ/Kϩ a muscle isoform of dynamin (dynamin 2) to the induced tu- action potential at the cell surface to each of the muscle fibers bules in CHO cells. Although the Bin1ϩ10ϩ13 BAR domain located in the cytoplasm. This propagation of the action po- and exon 10-encoded domain are sufficient to induce tubules in tential ensures that the response in muscle fibers is coordi- CHO cells, the SH3 domain must also be present for recruit- nated and synchronous and results in efficient depolarization- ment of dynamin 2 to the tubules. The tubules induced by the contraction coupling. Moreover, T-tubules are enriched in BAR domain and exon 10-encoded domain also appear to lack integral membrane proteins that function in ion transport, surrounding cytomatrix, in contrast to the tubules induced by which suggests a possible role for Bin1ϩ10ϩ13 in regulation of full-length Bin1ϩ10ϩ13 (162). ion fluxes across membranes. T-tubules may also be sites of The role of endogenous Bin1ϩ10ϩ13 in T-tubule forma- endocytosis in skeletal muscle. Clathrin heavy chain, like tion in muscle cells has been investigated using the myoblast Bin1ϩ10ϩ13, localizes to the I-band. The localizations are not cell line C2C12. During serum starvation-induced differen- identical, however, as clathrin heavy chain localizes to a region tiation of C2C12 cells, expression of both Bin1ϩ10ϩ13 and VOL. 70, 2006 BAR DOMAIN PROTEINS 101 caveolin 3 increases and both proteins localize to tubules As caveolae and T-tubules are both lipid-raft-enriched invagi- that are contiguous with the plasma membrane and enriched nations of the plasma membrane and newly forming T-tubules in PtdIns(4,5)P2. Mature T-tubules are narrow tubules con- contain caveolin-3, it has been proposed that T-tubules may taining Bin1ϩ10ϩ13 but not caveolin 3. However, newly form in the same way as caveolae (26). Interestingly, in CHO formed T-tubules resemble a series of vesicular structures con- cells the endogenous isoform of caveolin (caveolin 1) localizes nected by narrow tubules. Caveolin-3 is concentrated at the to Bin1ϩ10ϩ13-induced tubules (162). The localization of vesicular elements and Bin1ϩ10ϩ13 at the interconnecting Bin1ϩ10ϩ13 to lipid-raft-enriched T-tubules in muscle is in- tubular elements. Hence, Bin1ϩ10ϩ13 appears to play an im- teresting in light of the observation that Rvs161p and Rvs167p Downloaded from portant in vivo role in deforming membranes to generate nar- in yeast fractionate with detergent-insoluble lipid rafts (10, 97). -row tubules, consistent with its ability to bind and deform Bin1؉10؉13 binds and tubulates liposomes in vitro. Re membranes in vitro (see below) (162). combinant Bin1ϩ10ϩ13 has been shown to bind liposomes Further support for a role of Bin1ϩ10ϩ13 in muscle differ- in vitro via direct interaction with lipids. When incubated in entiation has come from studies of homozygous Bin1 knockout vitro with liposomes recombinant Bin1ϩ10ϩ13 generates mice that lack Bin1ϩ10ϩ13. Homozygous Bin1 knockout mice narrow membrane tubules by a process of evagination (162). appear to undergo normal embryonic development and are This is similar to what was shown earlier for recombinant represented at the expected Mendelian frequency at birth amphiphysin 1 in vitro (316). The lipid composition of the (207). However, the homozygous Bin1 knockout mice do not liposomes influences the efficiency of binding of Bin1ϩ10ϩ13. http://mmbr.asm.org/ feed and die in the first 24 h after birth. Anatomical charac- The presence of PtdIns(4,5)P2 strongly enhances binding and terization of homozygous Bin1 knockout mice revealed obvi- PtdIns(4)P has a similar but slightly weaker effect. While the ous cardiac myopathies, in particular a ventricular wall so se- Bin1ϩ10ϩ13 N-terminal amphipathic ␣-helix and BAR do- verely expanded that both ventricular chambers of the heart main are sufficient for binding to liposomes in vitro, enhance- were occluded. The ventricular cardiomyocytes are packed ment of binding by PtdIns(4,5)P2 or PtdIns(4)P is dependent more tightly than those in wild-type mice, but do not appear to on the domain encoded by exon 10. Neither PtdIns(4,5)P2 nor undergo abnormal proliferation. In ventricular cardiomyocytes PtdIns(4)P affect binding of fragments comprising the N-ter- of wild-type mice Bin1ϩ10ϩ13 is localized to the nucleus minal amphipathic ␣-helix and BAR domains of either am- (207). This is in contrast to skeletal muscle cells, where phiphysin 1 or Bin1ϩ6aϩ12ϩ13 to liposomes (162). Drosoph- on October 26, 2015 by University of Queensland Library Bin1ϩ10ϩ13 localizes to T-tubules that line myofibrils in the ila amphiphysin also has the ability to evaginate tubules from cytoplasm (23, 162). liposomes composed of brain lipids in vitro, showing that this Ultrastructural examination shows that the myofibrils in the property has been conserved during evolution (258). ventricular cardiomyocytes are not arranged in closely packed The importance of the Bin1ϩ10ϩ13 BAR domain for lipo- regular arrays in Bin1ϩ10ϩ13-deficient cardiomyocytes as some binding in vitro is consistent with the results of in vivo they are in normal cardiomyocytes. T-tubules are normally studies. The Bin1ϩ10ϩ13 BAR domain and the sequence en- positioned either side of the thin desmin-containing Z-line in coded by exon 10 are both essential for ectopically expressed muscle. In Bin1ϩ10ϩ13-deficient ventricular cardiomyocytes Bin1ϩ10ϩ13 to localize to the plasma membrane in CHO the Z-line is more diffuse than in normal cardiomyocytes, sug- cells. The plasma membrane is the major site of PtdIns(4,5)P2 gesting a possible defect in the T-tubule system. Although in cells. Hence, PtdIns(4,5)P2 binding may be important for skeletal muscle is present and gross abnormalities are not Bin1ϩ10ϩ13 localization and membrane tubulation in vivo apparent in homozygous Bin1 knockout mice, ultrastructural (162). examination reveals that the myofibrils in skeletal muscle cells also appear less well organized than in skeletal muscle from Bin1؊10؊12؉13/SH3p9 wild-type mice. In addition, the T-tubules appear to be shifted to the Z-line rather than being aligned on either side of it as in A splice variant of Bin1 that lacks the putative nuclear lo- normal skeletal muscle (207). Hence, Bin1ϩ10ϩ13 is not es- calization sequence and sequences required for lipid binding sential for muscle development, but is important for T-tubule encoded by exon 10 compared to muscle Bin1ϩ10ϩ13 but that organization in cardiomyocytes and, to a lesser extent, in skel- still retains the central third of the MBD encoded by exon 13 etal muscle. was identified in a screen for novel SH3 domain proteins and Bin1؉10؉13 and lipid rafts. The membranes of T-tubules named SH3p9 (Fig. 1) (306). SH3p9 is referred to here as have a distinctive lipid composition, being enriched in sterols Bin1Ϫ10Ϫ12ϩ13. This Bin1 isoform is expressed in many tis- and glycosphingolipids (26). Treatment of C2C12 cells with sues, although it may be absent in some specialized cell types agents such as amphotericin B and methyl-␤-cyclodextrin, such as macrophages (100) (see below). The function of which deplete cholesterol from membranes, interferes with Bin1Ϫ10Ϫ12ϩ13 remains unknown. Bin1ϩ10ϩ13 distribution and prevents T-tubule formation. The same treatments also prevent the induction of membrane Bin1؊10؊12؊13/ALP1/Amphiphysin IIm tubules in CHO cells upon ectopic expression of Bin1ϩ10ϩ13, suggesting these tubules are also enriched in cholesterol (162). While Bin1 was identified as a c-Myc-interacting protein, a During muscle differentiation the lipid raft marker caveolin-3 distinct splice variant was identified by its ability to bind an- localizes to T-tubules (26). Caveolins are a family of proteins other oncoprotein, c-Abl, and named Amphiphysin-Like Pro- that localize to flask-shaped, lipid-raft-enriched cell surface tein 1 (ALP1) (141). The same splice variant was identified as invaginations known as caveolae. Caveolae have been shown to the sole amphiphysin and Bin1 isoform in macrophages and function in endocytosis and in signal transduction (113, 159). named amphiphysin IIm (100). Bin1Ϫ10Ϫ12Ϫ13 is the most 102 REN ET AL. MICROBIOL.MOL.BIOL.REV. ubiquitous splice variant of Bin1. This splice variant is distinct SH3 domain. The mutant form of Bin1Ϫ10Ϫ12Ϫ13 that lacks from the original Bin1 isoform (Bin1ϩ10ϩ13) because it lacks an SH3 domain has dominant negative effects on receptor- the 15-residue sequence containing the putative nuclear local- mediated endocytosis of low-density lipoprotein (100). ization sequence and sequences required for lipid binding en- Is Bin1Ϫ10Ϫ12Ϫ13 also required for phagocytosis of parti- coded by exon 10 and furthermore is distinct from both cles? Phagocytosis of large particles (Ͼ0.5 ␮m) does not utilize Bin1Ϫ10Ϫ12ϩ13 (SH3p9) and Bin1ϩ10ϩ13 in that it lacks clathrin-coated pits regulated by clathrin, AP-2, and dynamin. exon 13, which encodes the central third of the MBD (note Instead, phagocytosis (literally, “cell eating”) involves the spreading of plasma membrane projections around a particle

that although the initial report by Kadlec et al. [141] described Downloaded from ALP1 as essentially identical to SH3p9 in lacking exon 13, to be internalized in a process dependent on the cortical actin ALP1 in fact differs from SH3p9 in lacking exon 13) (100, 141, cytoskeleton. Macrophages function in phagocytosis of patho- 278, 306, 338, 357). gens and other particles and express cell surface receptors that Despite lacking the putative nuclear localization sequence specifically recognize particles to be phagocytosed (1). Binding encoded by exon 10, when transiently expressed in NIH 3T3 of particles to phagocytic receptors is followed by recruitment Ϫ Ϫ Ϫ fibroblasts Bin1Ϫ10Ϫ12Ϫ13 localizes to the nucleus in a high of actin, Bin1 10 12 13, and dynamin 2 to the site of par- percentage of cells. However, in some transfected NIH 3T3 ticle binding. Actin filament assembly then generates an “actin Ϫ Ϫ Ϫ pedestal” beneath the bound particles. Finally, membrane ruf- cells Bin1 10 12 13 is found predominantly in the cyto- http://mmbr.asm.org/ plasm (141). Nuclear import of Bin1Ϫ10Ϫ12Ϫ13 is presum- fles and projections form that eventually engulf the receptor- Ϫ Ϫ Ϫ ably mediated by the alternative nuclear localization sequence bound particles. Both actin and Bin1 10 12 13 are gradu- present in the BAR-C domain, which is common to all Bin1 ally shed from the surface of phagosomes during maturation of splice variants (69). What cellular mechanism regulates the phagosomes into digestive compartments (100). nuclear/cytoplasmic distribution of Bin1Ϫ10Ϫ12Ϫ13 is un- Expression of a dominant negative mutant form of Ϫ Ϫ Ϫ known. Bin1 10 12 13 lacking the C-terminal SH3 domain Ϫ Ϫ Ϫ Ϫ c-Abl is a nonreceptor tyrosine kinase that localizes to both (Bin1 10 12 13SH3 ) prevents recruitment of dynamin 2 the nucleus and the actin cytoskeleton, and mutations in c-Abl to the site of particle binding and formation of membrane ruffles and projections, and as a consequence results in a severe have the ability to transform cells. c-Abl and ALP1 coimmu- block in particle engulfment and internalization (Fig. 12). on October 26, 2015 by University of Queensland Library noprecipitate from cotransfected cells, showing they associate However, expression of Bin1Ϫ10Ϫ12Ϫ13SH3Ϫ does not affect in vivo. Moreover, in vitro protein binding experiments showed cell viability, expression of the phagocytic receptors, particle that the C-terminal SH3 domain of ALP1 mediates interaction binding, cell spreading, actin pedestal formation, or secretion and binds at least two distinct proline-rich motifs within the to the cell surface (100). Interestingly, in these experiments the c-Abl C-terminal domain. The c-Abl SH3 does not bind ALP1. recruitment of Bin1Ϫ10Ϫ12Ϫ13 to the site of particle binding Unlike Bin1ϩ10ϩ13, Bin1Ϫ10Ϫ12Ϫ13 does not bind c-Myc in was not affected by deletion of the SH3 domain. This suggests vitro, presumably due to the absence of MBD sequences en- that the BAR domain mediates the subcellular localization of coded by exon 13 (141). Bin1Ϫ10Ϫ12Ϫ13. This is consistent with the finding that in .Bin1؊10؊12؊13 in regulation of the actin cytoskeleton Ϫ Ϫ Ϫ yeast the BAR domain mediates subcellular localization of Transient coexpression of c-Abl and Bin1 10 12 13 induces Rvs167p and that Rvs161p localizes similarly despite compris- a dramatic change in cell morphology in which actin stress ing only a BAR domain (11, 144). Subcellular localization of fibers are lost and the cells adopt a more rounded and “spin- Bin1Ϫ10Ϫ12Ϫ13 may be achieved by binding specific mem- dly” appearance. High-level expression of c-Abl inhibits cell brane lipids. In support of this, treatment of macrophages Ϫ Ϫ Ϫ proliferation and coexpression of Bin1 10 12 13 does not with the phosphatidylinositol 3-kinase inhibitor wortmannin Ϫ Ϫ Ϫ reverse this inhibition. Bin1 10 12 13 expression does not to deplete PtdIns(3)P abolishes the recruitment of Bin1Ϫ10Ϫ activate c-Abl tyrosine kinase activity, however, induction of 12Ϫ13 to the site of particle binding (100). Interestingly, the morphological change by coexpression requires an intact Bin1Ϫ10Ϫ12Ϫ13 lacks exon 10 that encodes the PtdIns(4,5)P2 Ϫ Ϫ Ϫ c-Abl kinase domain as well as the Bin1 10 12 13 SH3 binding domain of Bin1ϩ10ϩ13, so other domains in Bin1Ϫ domain and the proline-rich motifs in c-Abl with which it 10Ϫ12Ϫ13 may bind phosphoinositides (162). interacts. The change in cell morphology induced by coexpres- In these experiments an acute perturbation of Bin1Ϫ10Ϫ12Ϫ13 sion of c-Abl and Bin1Ϫ10Ϫ12Ϫ13 resembles those that ac- function was achieved using a dominant negative construct. Interest- company oncogenic transformation, but is not accompanied by ingly, however, macrophages derived from homozygous Bin1 oncogenic transformation. The altered BCR-Abl form of c-Abl knockout mice do not exhibit obvious defects in phagocytosis is associated with leukemia. Expression of Bin1Ϫ10Ϫ12Ϫ13 (207). The reason for the difference between the results ob- somewhat increases the ability of BCR-Abl to oncogenically tained by dominant negative constructs and gene knockout is transform cells in culture (141). This is in contrast to not yet clear. Expression of a mutant construct can have off- Bin1ϩ10ϩ13, whose high-level expression suppresses onco- target effects, so it is not certain that the defects in phagocy- genic transformation by the oncoprotein c-Myc (278). tosis seen using the Bin1Ϫ10Ϫ12Ϫ13SH3Ϫ construct arise -Bin1؊10؊12؊13 functions in phagocytosis in macro- from a role of amphiphysin IIm in phagocytosis. Another pos phages. Murine macrophages express Bin1Ϫ10Ϫ12Ϫ13 as sibility is that Bin1Ϫ10Ϫ12Ϫ13 function is critical for phago- their only Bin1 splice variant and this has been named am- cytosis but long-term compensatory mechanisms are induced phiphysin IIm but is identical to ALP1 (100, 141). Like am- during the development of homozygous knockout mice lacking phiphysin 1 and other splice variants of Bin1, Bin1Ϫ10Ϫ12Ϫ13 Bin1Ϫ10Ϫ12Ϫ13 that bypass the normal requirement for binds dynamin (dynamin 2 in macrophages) via its C-terminal Bin1Ϫ10Ϫ12Ϫ13 in phagocytosis. Further experiments will be VOL. 70, 2006 BAR DOMAIN PROTEINS 103 Downloaded from http://mmbr.asm.org/

FIG. 12. Bin1Ϫ10Ϫ12Ϫ13 functions in phagocytosis in macrophages. RAW-TT10 (murine macrophage-derived) cells were transiently cotrans- on October 26, 2015 by University of Queensland Library fected with a construct that expresses GFP and with either an empty vector (Control) or a construct that expresses a dominant negative form of Bin1Ϫ10Ϫ12Ϫ13 lacking the C-terminal SH3 domain (AmphIImSH3Ϫ). The cells expressing the highest levels of GFP were recovered by flow cytometry and incubated with immunoglobulin G-coated sheep red blood cells for 10 min (panels A and B) or for 1 h (panels C and D) before visualization by scanning electron microscopy. The control cells formed actin pedestals beneath the bound sheep red blood cells, then formed membrane ruffles (A), and then phagocytosed the bound sheep red blood cells (C). In contrast, cells expressing the dominant negative mutant Bin1Ϫ10Ϫ12Ϫ13 lacking the C-terminal SH3 domain formed actin pedestals beneath the bound sheep red blood cells (B), but did not form membrane ruffles, and did not phagocytose the bound sheep red blood cells which remained on the cell surface (D). The scale bar in panel B applies to each panel. (Reprinted from reference 100 with permission from Elsevier.)

necessary to distinguish between these possibilities. A possible yeast and phagocytosis in mammalian cells. Important parallels endocytic or phagocytic role is supported by the finding that include the dependence on a functional cortical actin cytoskel- Bin1Ϫ10Ϫ12Ϫ13 interacts via its BAR-C domain with SNX4, eton and the lack of strict dependence on cytoplasmic coat such as neuronal Bin1ϩ6aϩ12ϩ13. SNX4 plays a key role in proteins such as clathrin and AP-2 (95, 213, 266). A role for the membrane traffic through early endosomes (167). ER in yeast endocytosis is possible, but has not been fully Originally, it was assumed that the membrane that engulfs explored. Some mutations that block the secretory pathway at bound particles during phagocytosis is derived from the plasma the step of ER exit also compromise endocytosis (118, 265). membrane and that the process is one of plasma membrane However, it has been difficult to assess how direct these effects invagination. However, both in the slime mold Dictyostelium on endocytosis are (118). discoideum and in murine macrophages a role for the endo- Rvs161p and Rvs167p have recently been shown to physi- plasmic reticulum in phagocytosis has recently emerged. Dictyo- cally interact with proteins involved in both membrane traffic stelium mutations that affect the ER-resident proteins calnexin from the ER and also post-Golgi apparatus membrane traffic and calreticulin severely compromise phagocytosis (211). In to the cell surface (e.g., Gyp5p and Gyl1p) (33, 84, 317). Gyp5p murine macrophages, direct fusion between the ER and the and Gyl1p regulate the Rab GTPase Ypt1p, which has been plasma membrane occurs at sites of particle engulfment and implicated in export of lipid rafts from the ER as well as vesicle the membrane that surrounds the particles being phagocy- targeting (202). In mammals, the Ypt1p ortholog Rab1 has tosed is largely derived from the ER (14, 88). The role of also been implicated in export from the ER as well as vesicle Bin1Ϫ10Ϫ12Ϫ13 in phagocytosis is not yet well understood. targeting (227, 238, 239, 248). It would be interesting if these Bin1Ϫ10Ϫ12Ϫ13 lacks an NTID that has been shown to be Rab proteins have additional functions in endocytosis in yeast critical for plasma membrane targeting of Bin1ϩ6aϩ12ϩ13 and phagocytosis in mammalian cells. (257). Perhaps Bin1Ϫ10Ϫ12Ϫ13 localizes to an intracellular compartment such as the ER that moves to the plasma mem- AMPHIPHYSIN-RELATED PROTEIN Bin2 brane only after engagement of phagocytic receptors. A potential role for Bin1Ϫ10Ϫ12Ϫ13 in phagocytosis is par- Bin2 is perhaps the most enigmatic of the amphiphysin- ticularly interesting in light of parallels between endocytosis in related proteins. It was identified as a protein in cells not 104 REN ET AL. MICROBIOL.MOL.BIOL.REV. expressing amphiphysin 1 that is also distinct from Bin1 but AMPHIPHYSIN-RELATED PROTEIN Bin3 cross-reacts with polyclonal antisera to Bin1ϩ10ϩ13. The For a time it appeared that proteins comprising a BAR cross-reactive antibodies specifically recognize the BAR-C re- domain only are unique to yeast (42). However, a human gion of Bin1ϩ10ϩ13. A homology search of the expressed protein containing only a BAR domain was eventually identi- sequence tag database identified a B-lymphocyte cDNA en- fied by homology search of the expressed sequence tag data- coding a protein with high sequence identity to Bin1ϩ10ϩ13 in base using the amino acid sequence of S. cerevisiae Rvs161p as the BAR-C region and this protein was named Bridging INte- the query and named Bin3 (Fig. 1). Bin3 is encoded by a gene

grator 2 (Bin2). Bin2 is the product of a distinct gene known as Downloaded from distinct from those that encode amphiphysin 1 (AMPH1), Bin1 BIN2. Both a major 2.6-kb and a minor 3.5-kb transcript are (BIN1), and Bin2 (BIN2) and has been named BIN3. Human BIN2 encoded by the gene, suggesting that Bin2, like am- Bin3 exhibits 28 and 29% amino acid sequence identity to S. phiphysins 1 and Bin1, is differentially spliced (94). cerevisiae Rvs161p and S. pombe Hob3p, respectively, but ex- Bin2 has a predicted N-terminal amphipathic helix (residues hibits significantly lower homology to the BAR domains of S. 23 to 45) (see below) and BAR domain (residues 45 to 249) cerevisiae Rvs167p, S. pombe Hob1p, and human Bin1. Unlike homologous to those in amphiphysin 1 and Bin1 (residues 1 to Bin1, which generates a large number of distinct transcripts, 249 of Bin2 exhibit 61% amino acid sequence identity to Bin1) there appears to be only a single major Bin3 transcript. This

(Fig. 1). Bin2 has a large C-terminal acidic domain enriched in Bin3 transcript is expressed ubiquitously in embryos and all http://mmbr.asm.org/ serine and proline residues, but lacks a central insert domain adult tissues tested with the sole exception of the brain. Unlike ϩ homologous to those in amphiphysins 1 and Bin1 12 that Bin1, whose expression is lost in a significant percentage of ␣ bind clathrin heavy chain and AP-2/ -adaptin or to that in cultured cancer cell lines and primary tumors, all tumor cell ϩ ϩ Bin1 10 13 that binds c-Myc. Bin2 also lacks an SH3 domain. lines tested to date retain Bin3 expression, suggesting that Bin3 As found for other amphiphysin-like proteins it migrates is not a tumor suppressor (276). There has not yet been a anomalously during electrophoresis, giving an apparent size of thorough characterization of Bin3 and its cellular functions 80 kDa, while its sequence predicts a size of 61.7 kDa. When remain to be identified. transiently expressed at high levels in COS-7 fibroblasts, Bin2 was found to be cytosolic (94). It is possible that endogenous ENDOPHILIN FAMILY OF PROTEINS on October 26, 2015 by University of Queensland Library Bin2 associates with membranes in lymphoid cells. Bin2 is highly expressed in lymphoid cells and its expression Identification of Endophilin A1/SH3p4/SH3GL2 increases during induced differentiation of a granulocyte pre- Endophilin A1 (also called SH3p4 and SH3GL2) is a 40-kDa cursor cell line (HL60) in vitro (granulocytes are a class of BAR domain protein identified as a major synaptojanin 1 lymphoid cell that includes neutrophils, basophils, and eosin- binding protein in the brain (28, 51, 192, 269, 270), as a novel ophils). Bin2 transcripts are abundant in cultured lymphoid SH3 domain protein (SH3p4) that binds a synthetic peptide cell lines, but are not found in cell lines derived from brain, known to be a ligand of the Src SH3 domain (306), as a protein liver, lung, breast, prostate, connective tissue (fibroblast), or containing a GRB2-like SH3 domain (SH3GL2) (99), and as colon, suggesting that Bin2 may be lymphoid cell specific (94). ϩ ϩ an SH3 domain protein that binds a proline-rich motif in the Bin2 forms 1:1 heterodimers in vitro with Bin1 10 13. The ␤1-adrenergic receptor cytoplasmic tail (323) (Fig. 1 and Table association is specific because Bin2 cannot form heterodimers 2). The name endophilin derives from the affinity this protein in vitro with amphiphysin 1. Bin2 coimmunoprecipitates with displays for several different endocytic proteins (51, 192, 193). ϩ ϩ ϩ ϩ Bin1 10 13 (or with the neuron-specific isoform Bin1 6a Endophilin A1 possesses an N-terminal amphipathic ␣-helix, ϩ 12 13) from extracts of COS-7 cells transiently expressing a BAR domain that mediates homodimer formation, and a both proteins, suggesting that Bin2 may also form het- C-terminal SH3 domain that binds dynamin 1, synaptojanin 1, erodimers in vivo. Association requires the N-terminal part of amphiphysin 1, and Bin1ϩ12 (Fig. 1). Although the endophilin ϩ ϩ the Bin1 10 13 BAR domain (residues 1 to 122), but not the A1 SH3 domain binds the C-terminal PRDs of both synapto- BAR-C part (residues 124 to 207). Heterodimer formation janin 1 and dynamin 1, its affinity for the former is considerably ϩ ϩ does not require the Bin1 10 13 MBD or C-terminal SH3 greater (51, 192, 270). Unlike amphiphysin 1, endophilin A1 domain (94). However, heterodimer formation has not yet has only a short central domain and does not bind clathrin been demonstrated for endogenous Bin2 in lymphoid cells. heavy chain or AP-2/␣-adaptin (28, 51, 73, 192, 193, 255, 269) Unlike amphiphysin 1 and Bin1, Bin2 is not implicated in (note: Cestra et al. [28] refer to endophilin A1/SH3p4 as en- endocytosis. Transient high-level expression of Bin2 in COS-7 dophilin 2). Endophilin A1 has been reported to exhibit lyso- cells, unlike high-level expression of neuronal Bin1ϩ6aϩ12ϩ13, phosphatidic acid acyltransferase activity in vitro (284). How- does not inhibit receptor-mediated endocytosis of transferrin. ever, a very recent study found this activity is due to a BIN2 knockout mice have not yet been reported, so the effect contaminant (89). of loss of Bin2 expression on endocytosis is not yet known. The endophilin A1 SH3 domain binds the PRD of dynamin High-level expression of Bin2, unlike that of Bin1ϩ10ϩ13, 1 at a site distinct from the amphiphysin 1 SH3 domain (302). does not inhibit growth of the tumor cell lines HepG2 (hepa- Similarly, the endophilin A1 SH3 domain binds a motif in the toma), MCF-7 (breast carcinoma), A549 (lung carcinoma), or synaptojanin 1 PRD (PKRPPPPR) distinct from the two mo- DU145 and PC3 (prostate carcinoma), nor does Bin2 affect the tifs recognized by the amphiphysin 1 SH3 domain (PIRPSR ability of Bin1ϩ10ϩ13 to inhibit growth of these tumor cells in and PTIPPR) (28, 269). The endophilin A1 SH3 domain also coexpression experiments (94). The physiological role of Bin2 binds germinal center kinase-like kinase (GLK). Endophilin remains to be determined. A1 binding to GLK is implicated in Jun N-terminal kinase VOL. 70, 2006 BAR DOMAIN PROTEINS 105

(Jnk) activation (254). This interaction is interesting, as fission Endophilin A and B Family Proteins Bind yeast Hob1p also interacts via its SH3 domain with a kinase of and Tubulate Membranes the GLK family (Nak1p) (131). Like amphiphysin 1, endophilin A1 is highly enriched in the Recombinant endophilin A1 can bind directly to liposomes brain, where it colocalizes with synaptojanin 1 in presynaptic in vitro. Binding results in evagination of membrane tubules 20 nerve terminals. Endophilin A1 does not stably associate with to 100 nm in diameter similar to those formed by dynamin 1 or the plasmalemma or with synaptic vesicles but is predomi- amphiphysin 1. The tubules formed by endophilin A1 display a pattern of fine tightly packed transverse striations that resem-

nantly soluble in the nerve terminal cytoplasm. Endophilin A1 Downloaded from and synaptojanin 1 coimmunoprecipitate from brain extract, bles that seen on the tubules formed by amphiphysin 1. Mem- indicating that they form a stable complex in vivo. Synaptoja- brane binding and tubule formation by endophilin A1 are not nin 1 in the brain also forms complexes with amphiphysin dependent on the presence of the substrates for its reported 1/Bin1ϩ6aϩ12ϩ13 heterodimers that do not contain endophi- lysophosphatidic acid acyltransferase activity, consistent with a lin A1. Hence, synaptojanin 1 in the brain is present in two recent report that this activity is due to a contaminant enzyme separate complexes, one containing endophilin A1 and the (89). Indeed, liposome evagination appears to be nonenzy- matic. When clathrin coat proteins are also present, most of other containing amphiphysin 1 and Bin1ϩ6aϩ12ϩ13. Inter- the endophilin A1-coated tubules that form terminate in a

estingly, while amphiphysin 1, dynamin 1, and synaptojanin 1 http://mmbr.asm.org/ clathrin-coated bud (73). This is similar to what is observed for are all phosphorylated in resting neurons and rapidly dephos- incubation of liposomes with clathrin coat proteins and am- phorylated upon neuron depolarization, endophilin A1 is not phiphysin 1 (316). phosphorylated in resting neurons (51, 192, 193). Structure-function analysis reveals that the N-terminal do- Two very closely related proteins with the same domain main (residues 1 to 125) of endophilin A1 is both necessary structure as endophilin A1 that are also enriched in the brain and sufficient for in vitro liposome binding and tubulation are endophilin A2 (also called SH3p8, SH3GL1, and EEN) activity. This domain is predicted to have a high content of and endophilin A3 (also called SH3p13 and SH3GL3) (Fig. 1 ␣-helices and to form coiled coils. One endophilin A1 se- and Table 2) (28, 34, 99, 133, 269, 270, 306, 323). Endophilin quence in particular (residues 1 to 35) exhibits particularly high A2 is a 45-kDa protein that localizes to presynaptic nerve sequence homology to residues 1 to 41 of the BAR domain of on October 26, 2015 by University of Queensland Library terminals and forms heterodimers via BAR domain interac- amphiphysin 1. Deletion of this sequence in either endophilin tions with endophilin A1 (269). The endophilin A2 SH3 do- A1 or amphiphysin 1 abolishes binding to liposomes and mem- main, like the endophilin A1 SH3 domain, binds the PRD of brane tubulation. This sequence is predicted to contain an dynamin 1 via a site distinct from that recognized by the am- amphipathic helix with a hydrophobic face and a highly basic phiphysin 1 SH3 domain and also binds the synaptojanin 1 face and may interact with the phospholipid headgroups as PRD (270, 302). The endophilin A2 SH3 domain also binds the well as the hydrophobic interior of the bilayer. Consistent with short central domain of endophilin A2 and this homotypic this idea, replacement of hydrophobic residues with charged interaction may also occur in the case of endophilin A1 (31). residues within the hydrophobic face of this predicted helix The endophilin 2 SH3 domain also binds a proline-rich motif abolishes liposome binding and tubulation activity (73). ␤ in the 1-adrenergic receptor cytoplasmic tail (323). Although Chemical cross-linking shows that endophilin A1 assembles enriched in the brain, endophilin A3 is also expressed at low into high-molecular-weight oligomers in the presence of lipo- levels in the cervical carcinoma HeLa cells, where it localizes to somes in vitro. Moreover, endophilin A1 assembles with dy- the cytoplasm and also the nucleus (133). The endophilin A3 namin 1 to form a coat on membrane tubules that has a dif- SH3 domain, like those of endophilins A1 and A2, binds the ferent physical appearance from those coats formed by C-terminal PRDs of dynamin 1 and synaptojanin 1 (270) and a endophilin A1 alone. The coat formed by endophilin A1 and ␤ proline-rich motif in the 1-adrenergic receptor cytoplasmic dynamin 1 has thicker and more widely spaced striations and tail (323). Endophilins A1, A2, and A3 are each encoded by a resembles the coat formed by brain cytosol or by a mixture of distinct gene (99). purified dynamin 1 and amphiphysin 1. To form this distinctive A second subfamily of endophilins comprises endophilins B1 coat requires not only the BAR domain but also the C-terminal (also called SH3GLB1) and B2 (SH3GLB2), which are also SH3 domain of endophilin A1 that binds dynamin 1. Unex- encoded by distinct genes (196, 244) (Fig. 1 and Table 2). pectedly, coated membrane tubules formed by dynamin 1 and Endophilins B1 and B2 have a domain structure similar to that endophilin A1 do not vesiculate in vitro upon addition of GTP of endophilin A family members, with an N-terminal BAR (73). This is in contrast to tubules that are coated by dynamin domain and C-terminal SH3 domain (Fig. 1). Endophilins B1 1 and amphiphysin 1, which do vesiculate efficiently upon GTP and B2 form homo- and heterodimers via BAR domain inter- addition (316). Endophilin A1 may therefore prevent dynamin actions (244). Endophilin B1 interacts with amphiphysin 1, 1-mediated vesiculation of membrane tubules. Both endophi- Bin1ϩ12ϩ13, and dynamin 1, but not synaptojanin 1. En- lin A1 and amphiphysin 1 have been reported to inhibit dy- dophilin B1 was reported to possess lysophosphatidic acid acyl- namin 1 GTPase activity in vitro (73). However, stimulation of transferase activity (196), but as with endophilin A1 this is dynamin 1 GTPase activity was subsequently shown in the case probably due to a contaminant (89). Endophilins B1 and B2 of amphiphysin 1 and this stimulation is dependent on lipo- colocalize to structures in the cytoplasm and are not found in some size (380). Perhaps this will also be the case for endophi- the nucleus (244). Endophilin B1 has three known splice vari- lin A1, but this is not known. ants, of which one is ubiquitously expressed and two are ex- Interestingly, an amphipathic membrane binding and tubu- pressed mainly in the brain (196). lating domain homologous to that in endophilin A1 is also 106 REN ET AL. MICROBIOL.MOL.BIOL.REV. found in endophilin B1. In neurons a small pool of endophilin vaginated clathrin-coated pits, suggesting a dual role in mem- B1 localizes to presynaptic termini, similar to endophilin A1, brane fission and in uncoating of clathrin-coated vesicles. Ac- but there is also a large pool that localizes to the Golgi appa- cumulation of deeply invaginated clathrin-coated pits was also ratus. Recombinant endophilin B1 binds liposomes in vitro and induced by microinjection of the endophilin A1 SH3 domain evaginates membrane tubules such as endophilin A1. More- (87). Microinjection of antibodies to endophilin A1 into syn- over, deletion of the short amphipathic ␣-helix in the N-ter- apses has been reported to block the conversion of flat clathrin minal BAR domain abolishes membrane tubulation activity, lattices to invaginated clathrin-coated pits, suggesting a role for suggesting that endophilin B1 binds and tubulates membranes endophilin A1 in the earliest steps of clathrin-coated vesicle Downloaded from in a manner similar to endophilin A1 (73). formation (268). Using in vitro assays that reconstitute the sequential steps of Endophilin A Family Proteins Function in clathrin-dependent endocytosis in permeabilized 3T3-L1 adi- Receptor-Mediated Endocytosis pocytes and A431 adenocarcinoma cells, one study found that the endophilin A1 SH3 domain specifically inhibits the late Evidence in support of an endocytic role for endophilin A1 step of membrane fission and clathrin-coated vesicle release has come from studies similar to those that first implicated (295). In contrast, in another study using permeabilized A431 amphiphysin 1 in endocytosis (293, 369). In permeabilized neu- cells, the endophilin A1 SH3 domain was observed to inhibit rons treated with nonhydrolyzable guanosine 5Ј-O-(3-thio- the formation of deeply invaginated and constricted clathrin- http://mmbr.asm.org/ triphosphate) (GTP␥S) to inhibit dynamin 1 GTPase activity coated pits as well as inhibiting clathrin-coated vesicle fission and induce the accumulation of deeply invaginated clathrin- (120). Immunoelectron microscopy of synaptic membranes af- coated pits, endophilin A1 colocalizes with amphiphysin 1 and ter incubation with brain extract, ATP, and GTP␥S in vitro dynamin 1 in the electron-dense rings at the neck of each revealed that endophilin A1 is not a stoichiometric component clathrin-coated pit (270). Transient overexpression of the en- of clathrin coats. Instead, endophilin A1 specifically localizes dophilin A1 SH3 domain inhibits receptor-mediated endocy- to dynamin 1-coated tubules. Consistent with this (and in con- tosis of transferrin in intact cells (295). In the case of the trast to treatment with the endophilin A1 SH3 domain) immu- endophilin A1 SH3 domain the target is more likely to be nodepletion of endophilin A1 from brain extract does not synaptojanin 1 than dynamin 1 due to the higher affinity for affect the formation of deeply invaginated clathrin-coated pits on October 26, 2015 by University of Queensland Library synaptojanin 1. These results from SH3 domain overexpression on synaptic membranes incubated with ATP and GTP␥Sin experiments should be interpreted with caution, however, due vitro, but does reduce dynamin 1-coated tubule formation to the possibility of off-target interactions and nonspecific ef- (268). fects. As yet, there are no reports of endophilin A1 knockout Interestingly, in Drosophila flies endophilin A is highly ex- mice. It will be interesting to see how similar the phenotype pressed in neurons and localizes to the presynaptic side of associated with loss of endophilin A1 is to the phenotype neuromuscular junctions. Like mammalian endophilin A1, reported for amphiphysin 1 knockout mice (55). Drosophila endophilin A was reported to possess lysophospha- The endophilin A1 SH3 domain also inhibits transferrin tidic acid acyltransferase activity in vitro (106), although, as internalization in in vitro receptor-mediated endocytosis assays recently shown for mammalian endophilin A1, this activity may that utilize either permeabilized 3T3-L1 adipocytes or A431 be due to a contaminant (89). Drosophila endophilin A colo- adenocarcinoma cells (half-maximal inhibition in 3T3-L1 adi- calizes with dynamin, suggesting that it may function in endo- pocytes occurs at 11 ␮M) (295). The endophilin A1 SH3 do- cytosis (267). Mutation of endophilin A perturbs endocytosis main is therefore somewhat less potent an inhibitor than the of membrane-soluble endocytic dyes and blocks synaptic vesi- Bin1 SH3 domain. The inhibition may be specific because the cle recycling in neurons, resulting in lethality at an early larval c-Abl SH3 domain (which does not bind dynamin 1 or synap- stage (106, 267, 347). As in mammalian cells, Drosophila en- tojanin 1) does not inhibit transferrin internalization in this in dophilin A appears to function in early stages of clathrin- vitro assay, even at high concentrations. Treatment of perme- coated pit invagination as well as fission of invaginated pits abilized cells with the endophilin A1 SH3 domain prior to (106). Drosophila endophilin A also functions in concert with assay gave stronger inhibition than treatment of the cytosol synaptojanin in uncoating of clathrin-coated vesicles (348). (120, 295). This suggests that the endophilin A1 SH3 domain Hence, in Drosophila flies amphiphysin is specialized in target relevant to endocytosis is membrane associated. Unex- muscle function, while endophilin A is specialized in neu- pectedly, addition of neither dynamin 1 nor synaptojanin 1 to ronal function. the assay relieved inhibition by the endophilin A1 SH3 domain. Ubiquitination plays an important role in the sorting of This suggests that these known endophilin A1 SH3 domain plasma membrane proteins for endocytosis in yeast (117, 327). interactors may not be the targets whose inhibition blocks More recently it has become apparent that ubiquitination is endocytosis (120). These results with SH3 domain inhibition also critical for the sorting of plasma membrane receptors into should be considered suggestive rather than conclusive due to clathrin-coated pits and endocytosis in mammalian cells (103, possible off-target interactions and nonspecific effects. 308, 310, 345). Endophilins A1, A2, and A3 form complexes At what stage of endocytosis does endophilin A1 function? with CIN85 (an SH3 domain adaptor protein similar to yeast Endophilin A1 may have multiple roles in clathrin-dependent Sla1p) and function with the RING domain ubiquitin ligase endocytosis. Microinjection into living synapses of a peptide Cbl in ubiquitin-dependent sorting during endocytosis of that binds the endophilin A1 SH3 domain and inhibits its growth factor receptors in mammalian cells (240, 305, 313). interaction with dynamin 1 and synaptojanin 1 induces the A further role of endophilins A1 and A2 in synaptic vesicle accumulation of both clathrin-coated vesicles and deeply in- endocytosis is in regulating Ca2ϩ influx into neurons. Both VOL. 70, 2006 BAR DOMAIN PROTEINS 107 endophilin A1 and A2 exhibit direct binding to voltage-gated found to have GTPase-activating protein activity on Rho fam- Ca2ϩ channels (VGCCs). This interaction is regulated by Ca2ϩ ily GTPases (263). RICH-1 is also known as nadrin 1. RICH-1 binding to a site in the central domain of endophilins A1 and possesses an N-terminal amphipathic ␣-helix homologous to A2. The endophilin A1 and A2 SH3 domains interact with a that in endophilin A1 and B1 that mediates high-affinity lipo- proline-rich motif in the central domain adjacent to the Ca2ϩ- some binding and tubulation as well as a BAR domain and has binding site. Ca2ϩ binding to the central domain of endophilins been confirmed to have liposome-binding and tubulating ac- A1 and A2 breaks this intramolecular SH3 interaction and tivity in vitro (73, 237, 264). RICH-1 localizes to the ER: in allows interaction of the central domain with VGCCs. En- contrast to amphiphysin 1 and Bin1 isoforms, RICH-1 does not Downloaded from dophilin A2 association with VGCCs is important in vivo since localize to either the plasma membrane or endosomes. An transient expression in hippocampal neurons of a mutant en- interesting feature of RICH-1 is the ability of its isolated BAR dophilin A2 construct that constitutively binds VGCCs inhibits domain to directly bind phosphoinositides in vitro, including endocytosis (possibly by preventing Ca2ϩ influx) (31). This PtdIns, PtdIns(3)P, PtdIns(4)P, PtdIns(5)P, PtdIns(3,5)P2, and regulation of Ca2ϩ influx is somewhat reminiscent of the role PtdIns(4,5)P2 as well as other phospholipids, including phos- of Rvs161p in low-affinity Ca2ϩ influx during mating in yeast phatidylserine and phosphatidic acid on a solid support (264). (210). The specificity of binding to these lipids has not yet been tested In yeast, Rvs161p and Rvs167p are implicated in de novo using more physiological assays such as liposome binding, and http://mmbr.asm.org/ actin filament assembly. Rvs167p interacts via its SH3 domain it is possible that the BAR domain is recognizing predomi- with the WASp-related protein Las17p and with Abp1p, both nantly negative charge rather than these specific lipids. of which are activators of the Arp2/3p actin filament nucleation complex (18, 38, 101, 176, 181, 372). The neuronal isoform of SNX1 mammalian WASp (N-WASp) is a key activator of the mam- Sorting nexin 1 (SNX1) also has a BAR domain which has malian Arp2/3 actin filament nucleation complex and has been the ability to sense membrane curvature and mediate the for- implicated in clathrin-dependent endocytosis (119, 151, 280). mation of SNX1 dimers. SNX1 functions in membrane traffic Yet evidence of a direct role for amphiphysin 1 or Bin1 in actin of the cation-independent mannose 6-phosphate receptor polymerization in vertebrate cells is lacking. from the early endosome to the trans-Golgi network. SNX1 on October 26, 2015 by University of Queensland Library Interestingly, there is evidence that endophilin A3 may reg- has been shown to drive tubulation of the early endosome in ulate Arp2/3-dependent actin filament assembly during endo- vivo and the purified protein also efficiently tubulates lipo- cytosis in mammalian cells. The endophilin A3 SH3 domain somes in vitro (25). directly binds proline-rich motifs in N-WASp. Moreover, en- dophilin A3 enhances the ability of N-WASp to activate the Arp2/3 complex for actin polymerization in vitro. Endophilin Other BAR Domain Proteins A3 and N-WASp coimmunoprecipitate from cell extracts, Other BAR domain proteins recently shown to tubulate showing that they form complexes in vivo. Complex formation liposomes in vitro include arfaptin2, RICH-2/nadrin 2, cen- and accumulation of F-actin at intracellular sites where en- taurin ␤2, and oligophrenin (237). Other proteins have been dophilin A3 and N-WASp colocalize are both enhanced by predicted on the basis of sequence homology to possess BAR treatment of cells with epidermal growth factor (230). En- domains, e.g., Tuba (279), SNX2, SNX4-9, and SNX18 (25, dophilin A family members may therefore have a role in actin- 325a), Insulin Receptor Substrate protein of 53 kDa (IRSp53), dependent movement during endocytosis potentially analo- and Adaptor protein containing PH domain, PTB domain, and gous to that of Rvs161p and Rvs167p (which have been Leucine zipper motif 1 (APPL1) and 2 (APPL2) (109). These proposed to bear more resemblance to mammalian endophil- proteins are predicted to have liposome binding and tubulating ins than to amphiphysins (307). activity in vitro, but this has not yet been experimentally veri- These studies show that endophilin A family members have fied. Liposome binding and tubulation appear to be common multiple roles in endocytosis. They not only have the ability to features of BAR domain proteins. tubulate membranes, but also function in Ca2ϩ influx to trigger dephosphorylation and activation of the endocytic machinery, CRYSTAL STRUCTURE OF THE AMPHIPHYSIN in ubiquitin-dependent sorting of plasma membrane proteins BAR DOMAIN into forming vesicles, in actin filament assembly which may propel endocytic vesicles from the plasma membrane, and in Structure of the BAR Domain uncoating of clathrin-coated vesicles for fusion with endo- The crystal structure of the BAR domain of Drosophila am- somes in concert with synaptojanin family proteins. phiphysin was solved by Peter et al. (237) and the structure has been the topic of several excellent recent reviews (90, 109, 164, 390). Using the structural data together with liposome-binding OTHER BAR DOMAIN PROTEINS and tubulation experiments with wild-type and mutant BAR Rich-1/Nadrin 1 domain proteins, the authors propose a general mechanism for membrane binding and/or bending by BAR domains in differ- The RhoGAP Interacting with CIP4 Homologs (RICH, sub- ent proteins. Structural analysis revealed a banana-shaped sequently renamed RICH-1) protein was identified in a screen dimer formed by monomers, each comprising three ␣-helices for proteins that interact with the SH3 domain-containing in a coiled-coil arrangement (Fig. 13). The kinks in two of the adaptor protein CDC42-Interacting Protein 4 (CIP4) and ␣-helices together with the angled arrangement of the mono- 108 REN ET AL. MICROBIOL.MOL.BIOL.REV. Downloaded from http://mmbr.asm.org/ on October 26, 2015 by University of Queensland Library

FIG. 13. Three-dimensional crystal structure of the Drosophila amphiphysin BAR domain. A, Two amphiphysin monomers (depicted in purple and green) associate in antiparallel orientation to form a banana-shaped homodimer represented here in ribbon format. The residues that form the basic patches on the concave membrane-binding surface of the dimer are highlighted. The Protein Data Bank identifier is 1URU. Note that the Drosophila amphiphysin BAR domain lacks an NTID, but sequence alignment with mammalian Bin1 suggests the NTID would be positioned within the unstructured loop between helices 2 and 3 as seen in the Drosophila BAR domain. This loop is positively charged and two conserved lysine residues have been shown to be important for membrane association of both Drosophila amphiphysin and rat amphiphysins 1 and 2 (237). The presence of the NTID would contribute extra positive charges to the unstructured loop and may further strengthen membrane association. B, The homodimer represented in electrostatic surface format and viewed with the same orientation as in A. Electrostatic potential is indicated: Ϫ10 kTeϪ1 in red; ϩ10 kTeϪ1 in blue). C, The homodimer represented in ribbon format but viewed at an angle perpendicular to that in panel A to highlight details of the concave face. D, The homodimer represented in electrostatic surface format and viewed with the same orientation as in panel C. (Reprinted from reference 237 with permission of the publisher. Copyright 2004 American Association for the Advancement of Science.)

mers generate the curved shape of the dimer. Basic residues in the BAR domain alone to liposomes is dependent on liposome the concave face of the dimer (K137 and R140) as well as basic size, displaying enhanced binding to smaller liposomes with residues at each end of the dimer (K161 and K163) were shown more curvature. In contrast, the full-length protein binds effi- to be critical for membrane binding and tubulation. ciently to liposomes independently of their size. Based on these The electrostatic interactions between the dimer and mem- two observations, Peter et al. propose that the BAR domain brane coupled with the rigidity of the BAR domain are pre- itself is primarily a sensor of membrane curvature (237). dicted to be sufficient for membrane bending (390). However, More recently, the crystal structure of the endophilin A1 the BAR domain alone does not tubulate liposomes as effi- BAR domain was solved (359). This BAR domain also forms a ciently as the full-length protein, which also contains the N- banana-shaped dimer. However, a striking difference between terminal amphipathic helix (residues 1 to 41). Also, binding of this structure and that of the Drosophila amphiphysin BAR VOL. 70, 2006 BAR DOMAIN PROTEINS 109 domain is the presence of a disordered region (amino acids 60 Membrane Binding and GTPase Binding of BAR Domains: to 88), within helix 1. The disordered regions in each monomer Independent or Associated? are adjacent and in the middle of the dimer, where they extend The structure of the BAR domain is highly homologous to out of the concave face of the dimer. It is not clear whether the structure of a fragment of Arfaptin2 GTPase, which binds they extend perpendicularly out of the concave face of the to Rac (341), and Arf GTPase (147). Arfaptin2 is proposed to dimer or to the side of the dimer or whether they become modulate membrane ruffling through interactions with these ordered upon membrane binding. The role of these regions GTPases (324). Arfaptin2 is now recognized to possess a BAR

and their effect on the membrane binding properties of the Downloaded from domain and the concave face of the Arfaptin2 BAR domain BAR domain have not been determined. In addition, the en- dimer contains the binding site for Rac. APPL1 and APPL2 dophilin A1 BAR domain was bound to 11 cadmium ions and bind the Rab5 endosomal GTPase via sequences predicted to is proposed to bind to Ca2ϩ in vivo and provide a basis for the 2ϩ form BAR domains. APPL1 and APPL2 colocalize and are role of intracellular Ca in endocytosis. associated with tubular and vesicular elements of a membrane compartment that is likely to be an endosome. The binding of these proteins to GTPases via their BAR domains has raised Coupling of the BAR Domain with Additional the question of whether membrane binding and GTPase bind-

Lipid-Binding Sequences ing are mutually exclusive or common functions of the BAR http://mmbr.asm.org/ domain (109). The BAR domain of Arfaptin2 is also able to BAR domains frequently exist together with a second mem- tubulate liposomes in vitro, suggesting that it retains the mem- brane binding sequence such as an amphipathic ␣-helix, a brane-binding function. However, an in vivo role for arfaptin2 pleckstrin homology (PH) domain, or a phox homology (PX) in membrane binding or bending has not yet been demon- domain (109, 237). The N-terminal amphipathic ␣-helix found strated. Further studies are required to establish whether other in amphiphysin 1, arfaptin, and RICH-1/nadrin 1 increases the BAR domain proteins bind small GTPases and whether mem- ability of the BAR domain protein to bend membranes. These brane binding and GTPase binding are mutually exclusive or proteins are efficient at tubulation, and binding of the N-ter- common features of BAR domains. minal amphipathic ␣-helix and BAR domain of these proteins to liposomes is independent of membrane curvature (237). on October 26, 2015 by University of Queensland Library This may be due to deformation of the membrane by the BAR Domain and IMD Structure amphipathic ␣-helix in a manner similar to the amphipathic IRSp53 is a Rac GTPase binding protein that was also pre- ␣-helix of epsin (81). In amphiphysin 1 the BAR domain and dicted to contain a BAR domain (109). The N-terminal 250 N-terminal amphipathic ␣-helix are thought to mediate mem- amino acids form a domain that is homologous to the B iso- brane invagination upon recruitment of the protein to the form of Missing In Metastasis (MIM). Thus, this region, which membrane. The BAR domain then drives accumulation of the resembles a BAR domain, has been independently defined as protein at the neck of the invagination, where the curvature is an IMD (IRSp53 MIM homology Domain) (374). The IMD optimal for binding. This is followed by recruitment of dy- alone has been shown to play a role in actin bundling and namin to the neck of the invagination, where it oligomerizes contains the Rac1 binding site. The recent crystal structure of and allows fission of the vesicle from the membrane (380). the IRSp53 IMD revealed that it is a cylindrically shaped dimer BAR domain proteins with adjacent PH domains, which lack with tapered ends (195). Nevertheless, there is a striking ho- ␣ the N-terminal amphipathic -helix, tubulate liposomes less mology between the IMD and BAR domain. The IMD mono- efficiently and binding of these BAR domains and their adja- mer is composed of a coiled-coil arrangement of ␣-helices cent PH domains to liposomes is sensitive to membrane cur- similar to that of the BAR domain monomers, however, an vature (237). The PH and PX domains, which bind to specific additional shorter fourth ␣-helix is present and the resulting phosphoinositides, are thought to increase the specificity of dimer is flat, unlike the curved BAR domain. Basic residues membrane targeting of BAR domain proteins since the BAR present at the ends of the IMD dimer have been shown to be domain together with the PH/PX domain would allow target- critical for the function of the IMD in actin filament bundling. ing to a membrane subdomain with a specific membrane cur- The structures of the BAR domain and IMD illustrate how the vature and lipid composition. coiled-coil motif can be used to build protein domains with Several members of the sorting nexin family of proteins, diverse functions, including membrane binding, binding to which are characterized by the presence of PX domains, are small GTPases, and actin filament bundling. predicted to contain BAR domains adjacent to PX domains. The BAR and PX domains of SNX1 have been shown to target MEMBRANE TUBULATION AND BAR DOMAIN SNX1 to a highly curved microdomain in early endosomes that PROTEIN FUNCTION IN VIVO contains PtdIns(3)P (25). SNX1 mediates tubulation of the early endosome and via association with the retromer complex We now know that BAR domains bind and tubulate mem- (107) drives transport from the early endosome to the trans- branes both in vitro and also in vivo (73, 264, 316, 380). Fur- Golgi network. A PtdIns(4,5)P2 binding site encoded by exon thermore, crystallography has revealed that BAR domains are 10 is thought to recruit Bin1ϩ10ϩ13 to plasma membrane sites banana-shaped dimers able to sense and (depending on the enriched in this phosphoinositide in muscle cells. The BAR presence of other membrane-binding domains) create mem- domain and N-terminal amphipathic ␣-helix then bend the brane curvature (237). The challenge now is to use this new membrane and drive the formation of T-tubules (162). knowledge to gain insight into the diverse roles of BAR do- 110 REN ET AL. MICROBIOL.MOL.BIOL.REV. main proteins in vivo. Can a role in membrane binding and the nucleus APPL1 interacts with the multisubunit histone tubulation help explain the roles of BAR domain proteins in deacetylase complex NuRD/MeCP1. NuRD/MeCP1 is a key clathrin-dependent endocytosis, muscle development, sporula- factor in chromatin remodeling and regulation of gene expres- tion, tumor suppression, phagocytosis, mitochondrial function, sion. Both APPL1 and APPL2 are known to be important for transmembrane ion flux, cell-cell fusion, and viability upon cell proliferation. Hence, APPL1 and APPL2 have been pro- starvation? posed to link events occurring at the cell surface to nuclear The ability of BAR domain proteins to tubulate the plasma changes in gene expression required for cell proliferation membrane readily fits with current understanding of clathrin- (191). Downloaded from dependent endocytosis. In clathrin-dependent endocytosis in- The role of Bin1ϩ10ϩ13 in the nucleus (if this splice variant vagination of the clathrin-coated pit leads to the formation of is indeed found in the nucleus) is still unclear. Bin1ϩ10ϩ13 a deeply invaginated pit in which the formed vesicle is still regulates c-Myc function, but whether this regulation is im- attached to the plasma membrane by a narrow neck comprising posed in the cytoplasm or nucleus is a question that remains a membrane tubule (121, 283, 290, 293, 315). Formation of the unresolved. A recent study suggests the role of Bin1ϩ10ϩ13 tubular neck by amphiphysin 1/Bin1ϩ12 heterodimers com- may be related to that of APPL1. As mentioned above, over- bined with SH3 domain interactions may facilitate the recruit- expression of Drosophila c-Myc in select cells within a tissue ment of dynamin 1 to the neck of deeply invaginated pits. induces the high-expressor cells to become supercompetitors Curvature sensing by amphiphysin 1/Bin1ϩ12 dimers may also and to proliferate at the expense of their neighboring cells in http://mmbr.asm.org/ play an important role in positioning the associated dynamin 1 the tissues that express lower levels of Drosophila c-Myc (199). correctly at the neck, which has the highest membrane curva- Intriguingly, the same study showed that overexpression of ture (380). Dynamin 1 mediates GTP-dependent fission of Rab5 also induces a supercompetitor state. This finding sup- membrane tubules in vitro (312). Moreover, dynamin 1 is im- ports the existence of links between Rab5 function in early plicated in fission of the necks of deeply invaginated clathrin- endosomes and c-Myc signaling pathways that regulate cell coated pits and release of clathrin-coated vesicles in vivo (45, proliferation and survival. Moreover, Bin1 has been shown to 116, 121, 155, 344). interact via its C-terminal SH3 domain with Rin3 (142). Rin3 The ability of Bin1ϩ10ϩ13 to bind and tubulate membranes displays Rab5-specific GTP/GDP exchange factor activity (i.e., via its BAR domain and PtdIns(4,5)P2-binding domain en- is a Rab5-GEF) and colocalizes with Rab5 to a subset of early on October 26, 2015 by University of Queensland Library coded by exon 10 also readily explains the role of this protein endosomes (142). This study implicates Bin1 in Rab5 GTPase in T-tubule biogenesis. T-tubules are believed to form by pro- activation via interaction with Rin3. gressive invagination of the plasma membrane (26, 162). In this An intriguing possibility is that the tumor suppression and case dynamin-dependent fission does not occur and the tubules apoptotic functions of Bin1ϩ10ϩ13 reflect an involvement in remain as a type of extensive network of deeply invaginated the process known as autophagy, in which membranes form plasma membrane pits. around internal organelles as a prelude to their degradation (see Membrane tubulation by BAR domain proteins is not a below). There are strong emerging links between autophagy in phenomenon specific to the plasma membrane. For example, mammalian cells and tumor suppression and apoptosis (66). RICH-1 has been shown to localize to the ER and to play a The picture that is emerging for the BAR domain proteins in role in tubule formation at this organelle (73, 264). Endophilin some ways resembles that which has emerged for dynamin. The B1 tubulates Golgi apparatus membranes (73) while SNX1 classical dynamin (dynamin 1) was first implicated in endocy- plays a role in tubule formation at the early endosome (25). tosis via clathrin-coated pits (45, 116, 121, 155, 290, 344). How- Bin1ϩ6aϩ12ϩ13 and Bin1Ϫ10Ϫ12Ϫ13 both interact with ever, the discovery and characterization of diverse dynamin- SNX4 and may therefore function with SNX4 in membrane like proteins has revealed that the membrane fission activity of tubulation events in endosomes (167). Endophilin A3 has been this family of proteins extends to other membrane compart- shown to localize to a fine network of cytoplasmic tubules as ments, including the ER, mitochondria, and peroxisomes (154, well as to sites of endocytosis at the plasma membrane (133). 174, 246, 378, 379). Different BAR domain proteins may func- Hence, the phenomenon of membrane tubulation by BAR tion in different organelles in processes related to membrane domain proteins may not be specific to clathrin-coated pits or binding, tubulation, fission, and fusion. It is also possible that T-tubules. the same BAR domain protein may function in more than one Is membrane tubulation a process that also occurs in the cellular location and process. This possibility is especially likely nucleus? The observation that Bin1ϩ10 and Bin1Ϫ10Ϫ12 in the case of Rvs161p and Rvs167p, since these are the only (each isoform with or lacking exon 13) localize to the nucleus BAR domain proteins in yeast and therefore expected to pro- in at least some cell types suggests a possible role for BAR vide a more diverse range of functions than the vertebrate domain proteins in nuclear events (69, 141, 278, 355). A role BAR domain proteins, of which there are many. for BAR domain proteins in the nucleus is further supported How can a role in membrane tubule formation explain the by studies of APPL1 and APPL2 (191). APPL1 and APPL2 are phenotypes of yeast rvs mutants? Loss of Rvs161p or Rvs167p predicted BAR domain proteins that bind the endosomal Rab5 reduces the efficiency of endocytic internalization in yeast (21, GTPase and localize with Rab5 to a subset of early endosomes 38, 180, 215). This endocytic defect may be explained by an (109, 191). APPL1 and APPL2 play a role in signaling to the inability of the rvs mutants to tubulate the plasma membrane nucleus in response to changes in the extracellular environ- to form deeply invaginated endocytic pits or to form endocytic ment. Upon exposure of cells to epidermal growth factor or vesicles by membrane fission. Rvs161p and Rvs167p het- oxidative stress, APPL1 has been shown to dissociate from erodimers may be sufficient for plasma membrane tubulation the early endosome and to be imported into the nucleus. In during endocytosis, since clathrin, AP-2, and dynamin are not VOL. 70, 2006 BAR DOMAIN PROTEINS 111 strictly required for endocytosis in yeast (37, 92, 130, 226, 235, brane tubule formation also plays a role in transfer of lipids to 318, 377, 383). Alternatively, by analogy to the putative role of mitochondria and defects in this process may perturb mito- amphiphysin IIm in phagocytosis in macrophages, Rvs161p chondrial function. Consistent with a role for BAR domain and Rvs167p may be required for fusion of the cortical ER proteins in lipid transfer to mitochondria, overexpression or with sites of endocytosis at the plasma membrane in order to down-regulation of endophilin B1 in mammalian cells has se- provide additional membrane material (14, 88, 100, 211). In- vere consequences for mitochondrial morphology (148). deed, some mutations that block traffic from the ER also lower It is interesting that the lipid transferred from the ER to the efficiency of endocytosis, although the molecular basis for mitochondria by the MAM, phosphatidylserine, is also a lipid Downloaded from this link remains unclear (118, 265). enriched in yeast plasma membrane-associated membranes Three lines of evidence suggest that a model in which Rvs (see above) (243). Moreover, phosphatidylserine strongly stim- proteins regulate direct fusion of cortical ER with plasma ulates dynamin GTPase activity and is essential for dynamin- membrane is worth consideration. First, electron microscopy dependent liposome fission in vitro (312, 380). Finally, given studies reveal regions of contact between the cortical ER and the roles of some Bin1 splice isoforms in caspase-independent plasma membrane in yeast known as plasma membrane-asso- apoptosis, it is interesting that elevated phosphatidylserine in ciated membranes (named by analogy to mitochondrion-asso- the outer leaflet of the plasma membrane is a hallmark of ciated membranes [MAMs]) (see below) (243). Second, apoptosis (342). It is possible that BAR domain proteins will Rvs161p and Rvs167p physically and genetically interact with be found to play a role in delivery of phosphatidylserine to the http://mmbr.asm.org/ proteins involved in membrane traffic from the ER (Gyp5p, outer leaflet of the plasma membrane during Myc-induced Gyl1p, Rud3p, Sec22p, Sec21p, and Sec27p) (84, 123, 317, 332) apoptosis and in phosphatidylserine delivery from the ER to and with proteins that function in vesicle fusion with the mitochondria via MAMs in concert with dynamin family pro- plasma membrane (Exo70p, Sec8p, Sur4p, Fen1p/Sur5p, and teins, although this is conjecture and experimental evidence for possibly also Gyp5p and Gyl1p) (10, 18, 33, 48, 53, 262). Third, such postulated roles is lacking. a very recent study found that at least one yeast integral mem- A role for membrane tubulation in lipid transfer between brane protein (Ist2p) is transported directly from the cortical the ER and other organelles might also contribute to sporula- ER to the plasma membrane without passing through inter- tion. After sporulation of a diploid cell to form four haploid mediate membrane compartments, suggesting the ER and spores, the spores are held together in a cluster (ascus) by the on October 26, 2015 by University of Queensland Library plasma membrane are able to undergo localized fusion (140). surrounding plasma membrane contributed by the diploid par- Can a role in membrane tubulation provide an explanation ent. Each haploid nucleus formed during meiosis requires de for why loss of Rvs161p and Rvs167p perturbs the polarized novo membrane biogenesis to construct a membrane (pros- distribution of cortical actin patches (12, 215, 298)? Cortical pore membrane) that will become the plasma membrane of the actin patches are associated with plasma membrane invagina- spore (198). The process of membrane assembly around hap- tions (205). Electron microscopy studies suggested that these loid nuclei is not well understood but involves recruitment of invaginations are distinct from the invaginations associated membrane material from elsewhere in the cell. A role for with receptor-mediated endocytosis (204). However, several another cortical actin patch component (End3p) in membrane subsequent real-time imaging studies have shown that mem- traffic to the prospore membrane has very recently been dem- brane structures containing endocytosed material colocalize onstrated (201). Perhaps the Rvs proteins play a role in mem- with cortical actin patches and that actin filament nucleation brane recruitment via in vivo tubule formation. Mutations in correlates with movement of endocytic cargo (71, 132, 139, the ER-to-Golgi apparatus Rab GTPase Ypt1p that confer 143). Therefore, a defect in plasma membrane tubulation reduced viability upon starvation and block sporulation as rvs might directly impact on the formation, movement, and distri- mutations do have been isolated (288). Rvs161p and Rvs167p bution of cortical actin patches. Alternatively, defects in lipid both interact with proteins that regulate Ypt1p, suggesting delivery to the plasma membrane from the ER may perturb the Rvs161p and Rvs167p may regulate Ypt1p function in sporu- membrane dynamics of the plasma membrane or perturb sig- lation (18, 33, 49, 84, 85, 317). nal transduction pathways involving lipid second messengers. Finally, it is interesting to consider the phenotype of reduced These changes in rvs mutants may then affect plasma mem- viability upon starvation exhibited by rvs mutants. During nu- brane tubulation and cortical actin patch formation, move- trient starvation, cells induce a process known as autophagy in ment, and distribution. which nonessential internal structures and organelles are de- Although the role of Rvs161p and Rvs167p in mitochondrial graded (145, 351). Autophagy is a way of providing nutrients function is poorly understood it is intriguing that some Rvs- for use in essential cellular processes when external nutrients interacting proteins are implicated directly or indirectly in mi- are no longer available. A classic phenotype of mutants defec- tochondrial import of hydrophobic proteins, e.g., Pse1p/ tive in autophagy is reduced viability upon starvation. Like Kap121p (18, 39). Moreover, it is interesting that the dynamin- sporulation, autophagy is a process that relies on de novo like protein 1 (DLP1) in mammalian cells has been shown to membrane biogenesis. During autophagy, internal structures localize to sites of contact between the ER and mitochondria and organelles are surrounded by a double-layer membrane known as MAMs and to play a role in regulation of mitochon- derived, at least in part, from the ER. The newly formed drial morphology (378). MAMs are thought to play a role in compartment is known as the autophagosome. The autopha- lipid transfer from the site of lipid synthesis in the ER to gosome eventually fuses with the vacuole, and its contents are mitochondria, in particular the transfer of phosphatidylserine degraded by resident hydrolases. (291, 342, 343, 378). The current view is that transfer is The formation of an autophagosome requires the recruit- achieved by lipid flipping at a site of contact. Perhaps mem- ment of membrane material from donor compartments such as 112 REN ET AL. MICROBIOL.MOL.BIOL.REV. the ER. This process is still very poorly understood. One pos- REFERENCES sibility is that membrane is recruited via membrane tubules 1. Aderem, A., and D. M. Underhill. 1999. Mechanisms of phagocytosis in that connect the donor compartment (e.g., the ER) to the macrophages. Annu. Rev. Immunol. 17:593–623. 2. Ahle, S., and E. Ungewickell. 1986. Purification and properties of a new autophagosome. If this is correct then BAR domain proteins clathrin assembly protein. EMBO J. 5:3143–3149. such as Rvs161p and Rvs167p may play a crucial role in gen- 3. Amatruda, J. F., J. F. Cannon, K. Tatchell, C. Hug, and J. A. Cooper. 1990. erating the membrane tubules that connect the donor com- Disruption of the actin cytoskeleton in yeast capping protein mutants. Nature 344:352–354. partment to the nascent autophagosome. According to this 4. Amberg, D. C., E. Basart, and D. Botstein. 1995. Defining protein interac- model, loss of Rvs161p or Rvs167p would block the formation tions with yeast actin in vivo. Nat. Struct. Biol. 2:28–35. Downloaded from of the autophagosome and thereby prevent efficient turnover 5. Anderson, B. L., I. Boldogh, M. Evangelista, C. Boone, L. A. Greene, and L. A. Pon. 1998. The Src homology domain 3 (SH3) of a yeast type I myosin, of nonessential structures and organelles. This would result in Myo5p, binds to verprolin and is required for targeting to sites of actin an acute shortage of nutrients for essential processes and re- polarization. J. Cell Biol. 141:1357–1370. 6. Andrews, B., and V. Measday. 1998. The cyclin family of budding yeast: duced viability upon starvation. Intriguingly, elements of the abundant use of a good idea. Trends Genet. 14:66–72. ER itself are turned over by autophagy in yeast. Although it is 7. Ashrafi, K., T. A. Farazi, and J. I. Gordon. 1998. A role for Saccharomyces not known if Rvs161p and Rvs167p specifically are required for cerevisiae fatty acid activation protein 4 in regulating protein N-myristoyl- ation during entry into stationary phase. J. Biol. Chem. 273:25864–25874. this process, the actin cytoskeleton has recently been demon- 8. Badour, K., J. Zhang, and K. A. Siminovitch. 2003. The Wiskott-Aldrich strated to play a critical role (110). Mutations that affect GTP syndrome protein: forging the link between actin and cell activation. Im- hydrolysis by Ypt1p and lock it in the active GTP-bound state munol. Rev. 192:98–112. http://mmbr.asm.org/ 9. Bagnat, M., S. Keranen, A. Shevchenko, and K. Simons. 2000. Lipid rafts appear to induce cellular changes characteristic of autophagy function in biosynthetic delivery of proteins to the cell surface in yeast. (49). The Rvs161p and Rvs167p interactors Gyp5p/Ypl249cp Proc. Natl. Acad. Sci. USA 97:3254–3259. and Gyl1p/Ymr192wp are important regulators of Ypt1p (18, 10. Balguerie, A., M. Bagnat, M. Bonneu, M. Aigle, and A. M. Breton. 2002. Rvs161p and sphingolipids are required for actin repolarization following 33, 49, 84, 85, 317). salt stress. Eukaryot. Cell 1:1021–1031. 11. Balguerie, A., P. Sivadon, M. Bonneu, and M. Aigle. 1999. Rvs167p, the budding yeast homolog of amphiphysin, colocalizes with actin patches. CONCLUSIONS AND FUTURE PERSPECTIVES J. Cell Sci. 112:2529–2537. 12. Bauer, F., M. Urdaci, M. Aigle, and M. Crouzet. 1993. Alteration of a yeast The BAR domain proteins provide a function so fundamen- SH3 protein leads to conditional viability with defects in cytoskeletal and tal and important it has been used repeatedly in different budding patterns. Mol. Cell. Biol. 13:5070–5084. on October 26, 2015 by University of Queensland Library 13. Bauerfeind, R., K. Takei, and P. De Camilli. 1997. Amphiphysin I is associated contexts, including synaptic vesicle recycling in neurons, fusion with coated endocytic intermediates and undergoes stimulation-dependent de- of myoblasts into myotubes in developing muscle, tumor sup- phosphorylation in nerve terminals. J. Biol. Chem. 272:30984–30992. pression, and possibly regulation of ion flux across membranes. 14. Becker, T., A. Volchuk, and J. E. Rothman. 2005. Differential use of endo- plasmic reticulum membrane for phagocytosis in J774 macrophages. Proc. Genetic studies in yeast have shown that even in a simple Natl. Acad. Sci. USA 102:4022–4026. unicellular eukaryote BAR domain proteins are important for 15. Belmont, L. D., G. M. Patterson, and D. G. Drubin. 1999. New actin mutants allow further characterization of the nucleotide binding cleft and a bewildering range of processes, including regulation of cell drug binding sites. J. Cell Sci. 112:1325–1336. cycle progression, cell polarity, endocytosis, mitochondrial 16. Bi, E., P. Maddox, D. J. Lew, E. D. Salmon, J. N. McMillan, E. Yeh, and function, response to nutrient starvation, sensitivity to Naϩ, J. R. Pringle. 1998. Involvement of an actomyosin contractile ring in Sac- charomyces cerevisiae cytokinesis. J. Cell Biol. 142:1301–1312. cell-cell fusion during mating, actin cytoskeletal organization, 17. Bleazard, W., J. M. McCaffery, E. J. King, S. Bale, A. Mozdy, Q. Tieu, J. exocytosis, and differentiation into spores. The underlying mo- Nunnari, and J. M. Shaw. 1999. The dynamin-related GTPase Dnm1 reg- lecular function of the BAR domain as a generator and a ulates mitochondrial fission in yeast. Nat. Cell Biol. 1:298–304. 18. Bon, E., P. Recordon Navarro, P. Durrens, M. Iwase, A. Toh, and M. Aigle. sensor of membrane curvature has been demonstrated. The 2000. A network of proteins around Rvs167p and Rvs161p, two proteins structure of the BAR domain has revealed the structural basis related to the yeast actin cytoskeleton. Yeast 16:1229–1241. for this molecular function. The major challenge now is to 19. Breton, A. M., and M. Aigle. 1998. Genetic and functional relationship between Rvsp, myosin and actin in Saccharomyces cerevisiae. Curr. Gen. define how the fundamental processes of membrane curvature 34:280–286. generation and sensing are adapted in vivo to support such a 20. Breton, A. M., J. Schaeffer, and M. Aigle. 2001. The yeast Rvs161 and Rvs167 proteins are involved in secretory vesicles targeting the plasma tremendous range of important physiological processes from membrane and in cell integrity. Yeast 18:1053–1068. yeast to humans. 21. Brizzio, V., A. E. Gammie, and M. D. Rose. 1998. Rvs161p interacts with Fus2p to promote cell fusion in Saccharomyces cerevisiae. J. Cell Biol. 141:567–584. ACKNOWLEDGMENTS 22. Brown, D. A., and E. London. 1998. Functions of lipid rafts in biological membranes. Annu. Rev. Cell Dev. Biol. 14:111–136. We apologize to those whose important contributions to this field 23. Butler, M. H., C. David, G. C. Ochoa, Z. Freyberg, L. Daniell, D. Grabs, O. unfortunately could not be included in this review because of space and Cremona, and P. De Camilli. 1997. Amphiphysin II (SH3P9; BIN1), a time constraints. We thank H. Riezman, B. Andrews, H. Friesen, A. member of the amphiphysin/Rvs family, is concentrated in the cortical cytomatrix of axon initial segments and nodes of ranvier in brain and Gammie, A. Breton, F. Bauer, J. Nickels, P. Robinson, G. Prendergast, around T tubules in skeletal muscle. J. Cell Biol. 137:1355–1367. P. De Camilli, R. Parton, and P. McPherson for invaluable discussions 24. Care, A., K. A. Vousden, K. M. Binley, P. Radcliffe, J. Trevethick, I. Man- and a special thanks to G. Prendergast, B. Andrews, P. Robinson, and nazzu, and P. E. Sudbery. 2004. A synthetic lethal screen identifies a role H. Riezman for sharing unpublished data and/or providing manu- for the cortical actin patch/endocytosis complex in the response to nutrient scripts prior to publication. We thank A. Gammie and M. Rose for deprivation in Saccharomyces cerevisiae. Genetics 166:707–719. permission to reproduce original images in this review. R. Teasdale is 25. Carlton, J., M. Bujny, B. J. Peter, V. M. Oorschot, A. Rutherford, H. kindly acknowledged for performing advanced bioinformatic searches Mellor, J. Klumperman, H. T. McMahon, and P. J. Cullen. 2004. Sorting of sequence databases for novel BAR domains. nexin-1 mediates tubular endosome-to-TGN transport through coincidence sensing of high-curvature membranes and 3-phosphoinositides. Curr. Biol. Financial support from the National Health and Medical Research 14:1791–1800. Council of Australia (Project Grant 252750) and the Queensland State 26. Carozzi, A. J., E. Ikonen, M. R. Lindsay, and R. G. Parton. 2000. Role of Government (to A.L.M.) as well as from UMR7156 and CNRS, cholesterol in developing T-tubules: analogous mechanisms for T-tubule France (to B.W.), is gratefully acknowledged. G.R. acknowledges the and caveolae biogenesis. Traffic 1:326–341. support of a UQ Graduate Student Scholarship. 27. Carozzi, A. J., S. Roy, I. C. Morrow, A. Pol, B. Wyse, J. Clyde Smith, I. A. VOL. 70, 2006 BAR DOMAIN PROTEINS 113

Prior, S. J. Nixon, J. F. Hancock, and R. G. Parton. 2002. Inhibition of lipid McPherson. 1997. Identification of the major synaptojanin-binding proteins raft-dependent signaling by a dystrophy-associated mutant of caveolin-3. in brain. J. Biol. Chem. 272:8710–8716. J. Biol. Chem. 277:17944–17949. 52. Delley, P. A., and M. N. Hall. 1999. Cell wall stress depolarizes cell growth 28. Cestra, G., L. Castagnoli, L. Dente, O. Minenkova, A. Petrelli, N. Migone, via hyperactivation of RHO1. J. Cell Biol. 147:163–174. U. Hoffmuller, J. Schneider Mergener, and G. Cesareni. 1999. The SH3 53. Desfarges, L., P. Durrens, H. Juguelin, C. Cassagne, M. Bonneu, and M. domains of endophilin and amphiphysin bind to the proline-rich region of Aigle. 1993. Yeast mutants affected in viability upon starvation have a synaptojanin 1 at distinct sites that display an unconventional binding spec- modified phospholipid composition. Yeast 9:267–277. ificity. J. Biol. Chem. 274:32001–32007. 54. Dickson, R. C., and R. L. Lester. 1999. Yeast sphingolipids. Biochim. 29. Chant, J., and J. R. Pringle. 1991. Budding and cell polarity in Saccharo- Biophys. Acta 1426:347–357. myces cerevisiae. Curr. Opin. Gen. Dev. 1:342–350. 55. Di Paolo, G., S. Sankaranarayanan, M. R. Wenk, L. Daniell, E. Perucco,

30. Chant, J., and J. R. Pringle. 1995. Patterns of bud-site selection in the yeast B. J. Caldarone, R. Flavell, M. R. Picciotto, T. A. Ryan, O. Cremona, and Downloaded from Saccharomyces cerevisiae. J. Cell Biol. 129:751–765. P. De Camilli. 2002. Decreased synaptic vesicle recycling efficiency and 31. Chen, Y., L. Deng, Y. Maeno Hikichi, M. Lai, S. Chang, G. Chen, and J. F. cognitive deficits in amphiphysin 1 knockout mice. Neuron 33:789–804. Zhang. 2003. Formation of an endophilin-Ca2ϩ channel complex is critical 56. Dorer, R., C. Boone, T. Kimbrough, J. Kim, and L. H. Hartwell. 1997. for clathrin-mediated synaptic vesicle endocytosis. Cell 115:37–48. Genetic analysis of default mating behavior in Saccharomyces cerevisiae. 32. Chen, Y., and P. W. Piper. 1995. Consequences of the overexpression of Genetics 146:39–55. ubiquitin in yeast: elevated tolerances of osmostress, ethanol and canava- 57. Drake, M. T., M. A. Downs, and L. M. Traub. 2000. Epsin binds to clathrin nine, yet reduced tolerances of cadmium, arsenite and paromomycin. Bio- by associating directly with the clathrin-terminal domain. Evidence for chim. Biophys. Acta 1268:59–64. cooperative binding through two discrete sites. J. Biol. Chem. 275:6479– 33. Chesneau, L., S. Dupre, A. Burdina, J. Roger, S. Le Panse, M. Jacquet, and 6489. M. H. Cuif. 2004. Gyp5p and Gyl1p are involved in the control of polarized 58. Drake, M. T., and L. M. Traub. 2001. Interaction of two structurally distinct exocytosis in budding yeast. J. Cell Sci. 117:4757–4767. sequence types with the clathrin terminal domain beta-propeller. J. Biol. http://mmbr.asm.org/ 34. Cheung, N., C. W. So, J. W. Yam, C. K. So, R. Y. Poon, D. Y. Jin, and L. C. Chem. 276:28700–28709. Chan. 2004. Subcellular localization of EEN/endophilin A2, a fusion part- 59. Drees, B. L., B. Sundin, E. Brazeau, J. P. Caviston, G. C. Chen, W. Guo, ner gene in leukaemia. Biochem. J. 383:27–35. K. G. Kozminski, M. W. Lau, J. J. Moskow, A. Tong, L. R. Schenkman, A. 35. Chowdhury, S., K. W. Smith, and M. C. Gustin. 1992. Osmotic stress and McKenzie, 3rd, P. Brennwald, M. Longtine, E. Bi, C. Chan, P. Novick, C. the yeast cytoskeleton: phenotype-specific suppression of an actin mutation. Boone, J. R. Pringle, T. N. Davis, S. Fields, and D. G. Drubin. 2001. A J. Cell Biol. 118:561–571. protein interaction map for cell polarity development. J. Cell Biol. 154:549– 36. Chrebet, G. L., D. Wisniewski, A. L. Perkins, Q. Deng, M. B. Kurtz, A. 571. Marcy, and S. A. Parent. 2005. Cell-based assays to detect inhibitors of 60. Drubin, D. G., K. G. Miller, and D. Botstein. 1988. Yeast actin-binding fungal mRNA capping enzymes and characterization of sinefungin as a cap proteins: evidence for a role in morphogenesis. J. Cell Biol. 107:2551–2561. methyltransferase inhibitor. J. Biomol. Screening 10:355–364. 61. DuHadaway, J. B., F. J. Lynch, S. Brisbay, C. Bueso Ramos, P. Troncoso, 37. Chu, D. S., B. Pishvaee, and G. S. Payne. 1996. The light chain subunit is T. McDonnell, and G. C. Prendergast. 2003. Immunohistochemical analysis required for clathrin function in Saccharomyces cerevisiae. J. Biol. Chem. of Bin1/Amphiphysin II in human tissues: diverse sites of nuclear expres-

271:33123–33130. sion and losses in prostate cancer. J. Cell. Biochem. 88:635–642. on October 26, 2015 by University of Queensland Library 38. Colwill, K., D. Field, L. Moore, J. Friesen, and B. Andrews. 1999. In vivo 62. Duncan, M. C., M. J. Cope, B. L. Goode, B. Wendland, and D. G. Drubin. analysis of the domains of yeast Rvs167p suggests Rvs167p function is 2001. Yeast Eps15-like endocytic protein, Pan1p, activates the Arp2/3 com- mediated through multiple protein interactions. Genetics 152:881–893. plex. Nat. Cell Biol. 3:687–690. 39. Corral Debrinski, M., N. Belgareh, C. Blugeon, M. G. Claros, V. Doye, and 63. Dunn, R., and L. Hicke. 2001. Domains of the Rsp5 ubiquitin-protein ligase C. Jacq. 1999. Overexpression of yeast karyopherin Pse1p/Kap121p stimu- required for receptor-mediated and fluid-phase endocytosis. Mol. Biol. Cell lates the mitochondrial import of hydrophobic proteins in vivo. Mol. Mi- 12:421–435. crobiol. 31:1499–1511. 64. Dunn, R., and L. Hicke. 2001. Multiple roles for Rsp5p-dependent ubiq- 40. Cousin, M. A., T. C. Tan, and P. J. Robinson. 2001. Protein phosphorylation uitination at the internalization step of endocytosis. J. Biol. Chem. 276: is required for endocytosis in nerve terminals: potential role for the de- 25974–25981. phosphins dynamin I and synaptojanin, but not AP180 or amphiphysin. 65. Durrens, P., E. Revardel, M. Bonneu, and M. Aigle. 1995. Evidence for a J. Neurochem. 76:105–116. branched pathway in the polarized cell division of Saccharomyces cerevi- 41. Cross, F., L. H. Hartwell, C. Jackson, and J. B. Konopka. 1988. Conjuga- siae. Curr. Genet. 27:213–216. tion in Saccharomyces cerevisiae. Annu. Rev. Cell Biol. 4:429–457. 66. Edinger, A. L., and C. B. Thompson. 2004. Death by design: apoptosis, 42. Crouzet, M., M. Urdaci, L. Dulau, and M. Aigle. 1991. Yeast mutant necrosis and autophagy. Curr. Opin. Cell Biol. 16:663–669. affected for viability upon nutrient starvation: characterization and cloning 67. Elion, E. A., J. Trueheart, and G. R. Fink. 1995. Fus2 localizes near the site of the RVS161 gene. Yeast 7:727–743. of cell fusion and is required for both cell fusion and nuclear alignment 43. Cuddeback, S. M., H. Yamaguchi, K. Komatsu, T. Miyashita, M. Yamada, during zygote formation. J. Cell Biol. 130:1283–1296. C. Wu, S. Singh, and H. G. Wang. 2001. Molecular cloning and character- 68. Elliott, K., K. Ge, W. Du, and G. C. Prendergast. 2000. The c-Myc-inter- ization of Bif-1. A novel Src homology 3 domain-containing protein that acting adaptor protein Bin1 activates a caspase-independent cell death associates with Bax. J. Biol. Chem. 276:20559–20565. program. Oncogene 19:4669–4684. 44. D’Adamo, P., A. Menegon, C. Lo Nigro, M. Grasso, M. Gulisano, F. 69. Elliott, K., D. Sakamuro, A. Basu, W. Du, W. Wunner, P. Staller, S. Tamanini, T. Bienvenu, A. K. Gedeon, B. Oostra, S. K. Wu, A. Tandon, F. Gaubatz, H. Zhang, E. Prochownik, M. Eilers, and G. C. Prendergast. 1999. Valtorta, W. E. Balch, J. Chelly, and D. Toniolo. 1998. Mutations in GDI1 Bin1 functionally interacts with Myc and inhibits cell proliferation via mul- are responsible for X-linked non-specific mental retardation. Nat. Genet. tiple mechanisms. Oncogene 18:3564–3573. 19:134–139. 70. Endow, S. A., and M. A. Titus. 1992. Genetic approaches to molecular 45. Damke, H., M. Gossen, S. Freundlieb, H. Bujard, and S. L. Schmid. 1995. motors. Annu. Rev. Cell Biol. 8:29–66. Tightly regulated and inducible expression of dominant interfering dynamin 71. Engqvist Goldstein, A. E., and D. G. Drubin. 2003. Actin assembly and mutant in stably transformed HeLa cells. Methods Enzymol. 257:209–220. endocytosis: from yeast to mammals. Annu. Rev. Cell Dev. Biol. 19:287– 46. David, C., P. S. McPherson, O. Mundigl, and P. de Camilli. 1996. A role of 332. amphiphysin in synaptic vesicle endocytosis suggested by its binding to 72. Evangelista, M., B. M. Klebl, A. H. Tong, B. A. Webb, T. Leeuw, E. Leberer, dynamin in nerve terminals. Proc. Natl. Acad. Sci. USA 93:331–335. M. Whiteway, D. Y. Thomas, and C. Boone. 2000. A role for myosin-I in 47. David, C., M. Solimena, and P. De Camilli. 1994. Autoimmunity in stiff- actin assembly through interactions with Vrp1p, Bee1p, and the Arp2/3 Man syndrome with breast cancer is targeted to the C-terminal region of complex. J. Cell Biol. 148:353–362. human amphiphysin, a protein similar to the yeast proteins, Rvs167 and 73. Farsad, K., N. Ringstad, K. Takei, S. R. Floyd, K. Rose, and P. De Camilli. Rvs161. FEBS Lett. 351:73–79. 2001. Generation of high curvature membranes mediated by direct en- 48. David, D., S. Sundarababu, and J. E. Gerst. 1998. Involvement of long dophilin bilayer interactions. J. Cell Biol. 155:193–200. chain fatty acid elongation in the trafficking of secretory vesicles in yeast. 74. Farsad, K., V. Slepnev, G. Ochoa, L. Daniell, V. Haucke, and P. De Camilli. J. Cell Biol. 143:1167–1182. 2003. A putative role for intramolecular regulatory mechanisms in the 49. De Antoni, A., J. Schmitzova, H. H. Trepte, D. Gallwitz, and S. Albert. 2002. adaptor function of amphiphysin in endocytosis. Neuropharmacology 45: Significance of GTP hydrolysis in Ypt1p-regulated endoplasmic reticulum 787–796. to Golgi transport revealed by the analysis of two novel Ypt1-GAPs. J. Biol. 75. Fishbein, J. D., R. T. Dobrowsky, A. Bielawska, S. Garrett, and Y. A. Chem. 277:41023–41031. Hannun. 1993. Ceramide-mediated growth inhibition and CAPP are con- 50. De Camilli, P., A. Thomas, R. Cofiell, F. Folli, B. Lichte, G. Piccolo, H. M. served in Saccharomyces cerevisiae. J. Biol. Chem. 268:9255–9261. Meinck, M. Austoni, G. Fassetta, G. Bottazzo, et al. 1993. The synaptic 76. Fitch, P. G., A. E. Gammie, D. J. Lee, V. B. de Candal, and M. D. Rose. vesicle-associated protein amphiphysin is the 128-kD autoantigen of Stiff- 2004. Lrg1p Is a Rho1 GTPase-activating protein required for efficient cell Man syndrome with breast cancer. J. Exp. Med. 178:2219–2223. fusion in yeast. Genetics 168:733–746. 51. de Heuvel, E., A. W. Bell, A. R. Ramjaun, K. Wong, W. S. Sossin, and P. S. 77. Floyd, S., M. H. Butler, O. Cremona, C. David, Z. Freyberg, X. Zhang, M. 114 REN ET AL. MICROBIOL.MOL.BIOL.REV.

Solimena, A. Tokunaga, H. Ishizu, K. Tsutsui, and P. De Camilli. 1998. the Arp2/3 complex by the actin filament binding protein Abp1p. J. Cell Expression of amphiphysin I, an autoantigen of paraneoplastic neurological Biol. 153:627–634. syndromes, in breast cancer. Mol. Med.. 4:29–39. 102. Goodson, H. V., B. L. Anderson, H. M. Warrick, L. A. Pon, and J. A. 78. Floyd, S. R., E. B. Porro, V. I. Slepnev, G. C. Ochoa, L. H. Tsai, and P. De Spudich. 1996. Synthetic lethality screen identifies a novel yeast myosin I Camilli. 2001. Amphiphysin 1 binds the cyclin-dependent kinase (cdk) 5 gene (MYO5): myosin I proteins are required for polarization of the actin regulatory subunit p35 and is phosphorylated by cdk5 and cdc2. J. Biol. cytoskeleton. J. Cell Biol. 133:1277–1291. Chem. 276:8104–8110. 103. Govers, R., P. van Kerkhof, A. L. Schwartz, and G. J. Strous. 1997. Linkage 79. Flucher, B. E., M. E. Morton, S. C. Froehner, and M. P. Daniels. 1990. of the ubiquitin-conjugating system and the endocytic pathway in ligand- Localization of the alpha 1 and alpha 2 subunits of the dihydropyridine induced internalization of the growth hormone receptor. EMBO J. 16: receptor and ankyrin in skeletal muscle triads. Neuron 5:339–351. 4851–4858.

80. Folli, F., M. Solimena, R. Cofiell, M. Austoni, G. Tallini, G. Fassetta, D. 104. Govindan, B., R. Bowser, and P. Novick. 1995. The role of Myo2, a yeast Downloaded from Bates, N. Cartlidge, G. F. Bottazzo, G. Piccolo, et al. 1993. Autoantibodies class V myosin, in vesicular transport. J. Cell Biol. 128:1055–1068. to a 128-kd synaptic protein in three women with the stiff-man syndrome 105. Grabs, D., V. I. Slepnev, Z. Songyang, C. David, M. Lynch, L. C. Cantley, and breast cancer. N. Engl. J. Med. 328:546–551. and P. De Camilli. 1997. The SH3 domain of amphiphysin binds the pro- 81. Ford, M. G., I. G. Mills, B. J. Peter, Y. Vallis, G. J. Praefcke, P. R. Evans, line-rich domain of dynamin at a single site that defines a new SH3 binding and H. T. McMahon. 2002. Curvature of clathrin-coated pits driven by consensus sequence. J. Biol. Chem. 272:13419–13425. epsin. Nature 419:361–366. 106. Guichet, A., T. Wucherpfennig, V. Dudu, S. Etter, M. Wilsch Brauniger, A. 82. Friant, S., R. Lombardi, T. Schmelzle, M. N. Hall, and H. Riezman. 2001. Hellwig, M. Gonzalez Gaitan, W. B. Huttner, and A. A. Schmidt. 2002. Sphingoid base signaling via Pkh kinases is required for endocytosis in Essential role of endophilin A in synaptic vesicle budding at the Drosophila yeast. EMBO J. 20:6783–6792. neuromuscular junction. EMBO J. 21:1661–1672. 83. Friant, S., B. Zanolari, and H. Riezman. 2000. Increased protein kinase or 107. Gullapalli, A., T. A. Garrett, M. M. Paing, C. T. Griffin, Y. Yang, and J. Trejo. 2004. A role for sorting nexin 2 in epidermal growth factor receptor decreased PP2A activity bypasses sphingoid base requirement in endocy- http://mmbr.asm.org/ tosis. EMBO J. 19:2834–2844. down-regulation: evidence for distinct functions of sorting nexin 1 and 2 in 84. Friesen, H., K. Colwill, K. Robertson, O. Schub, and B. Andrews. 2005. protein trafficking. Mol. Biol. Cell 15:2143–2155. Interaction of the Saccharomyces cerevisiae cortical actin patch protein 108. Guo, W., A. O. Jorgensen, and K. P. Campbell. 1994. Characterization and Rvs167p with proteins involved in ER to Golgi vesicle trafficking. Genetics ultrastructural localization of a novel 90-kDa protein unique to skeletal 170:555–568. muscle junctional sarcoplasmic reticulum. J. Biol. Chem. 269:28359–28365. 84a.Friesen, H., C. Humphries, Y. Ho, O. Schub, K. Colwill, and B. Andrews. 109. Habermann, B. 2004. The BAR-domain family of proteins: a case of bend- 2006. Characterization of the yeast amphiphysins Rvs161 and Rvs167 re- ing and binding? EMBO Rep. 5:250–255. veals roles for the Rvs heterodimer in vivo. Mol. Biol. Cell, in press. 110. Hamasaki, M., T. Noda, M. Baba, and Y. Ohsumi. 2005. Starvation triggers 85. Friesen, H., K. Murphy, A. Breitkreutz, M. Tyers, and B. Andrews. 2003. the delivery of the endoplasmic reticulum to the vacuole via autophagy in Regulation of the yeast amphiphysin homologue Rvs167p by phosphoryla- yeast. Traffic 6:56–65. tion. Mol. Biol. Cell 14:3027–3040. 111. Hannun, Y. A. 1996. Functions of ceramide in coordinating cellular re- 86. Fukushima, N. H., E. Brisch, B. R. Keegan, W. Bleazard, and J. M. Shaw. sponses to stress. Science 274:1855–1859. 2001. The GTPase effector domain sequence of the Dnm1p GTPase regu- 112. Hannun, Y. A., and R. M. Bell. 1989. Functions of sphingolipids and sphin- on October 26, 2015 by University of Queensland Library lates self-assembly and controls a rate-limiting step in mitochondrial fission. golipid breakdown products in cellular regulation. Science 243:500–507. Mol. Biol. Cell 12:2756–2766. 113. Harder, T., and K. Simons. 1997. Caveolae, DIGs, and the dynamics of 9: 87. Gad, H., N. Ringstad, P. Low, O. Kjaerulff, J. Gustafsson, M. Wenk, G. Di sphingolipid-cholesterol microdomains. Curr. Opin. Cell Biol. 534–542. Paolo, Y. Nemoto, J. Crun, M. H. Ellisman, P. De Camilli, O. Shupliakov, 114. Hazbun, T. R., L. Malmstrom, S. Anderson, B. J. Graczyk, B. Fox, M. Riffle, B. A. Sundin, J. D. Aranda, W. H. McDonald, C. H. Chiu, B. E. Snydsman, and L. Brodin. 2000. Fission and uncoating of synaptic clathrin-coated P. Bradley, E. G. Muller, S. Fields, D. Baker, J. R. Yates 3rd, and T. N. vesicles are perturbed by disruption of interactions with the SH3 domain of Davis. endophilin. Neuron 27:301–312. 2003. Assigning function to yeast proteins by integration of technol- ogies. Mol. cell 12:1353–1365. 88. Gagnon, E., S. Duclos, C. Rondeau, E. Chevet, P. H. Cameron, O. Steele 115. Heinemeyer, W., J. A. Kleinschmidt, J. Saidowsky, C. Escher, and D. H. Mortimer, J. Paiement, J. J. Bergeron, and M. Desjardins. 2002. Endo- Wolf. 1991. Proteinase yscE, the yeast proteasome/multicatalytic-multifunc- plasmic reticulum-mediated phagocytosis is a mechanism of entry into tional proteinase: mutants unravel its function in stress induced proteolysis macrophages. Cell 110:119–131. and uncover its necessity for cell survival. EMBO J. 10:555–562. 89. Gallop, J. L., P. J. Butler, and H. T. McMahon. 2005. Endophilin and 116. Herskovits, J. S., C. C. Burgess, R. A. Obar, and R. B. Vallee. 1993. Effects CtBP/BARS are not acyl transferases in endocytosis or Golgi fission. Na- of mutant rat dynamin on endocytosis. J. Cell Biol. 122:565–578. ture 438:675–678. 117. Hicke, L., and H. Riezman. 1996. Ubiquitination of a yeast plasma mem- 90. Gallop, J. L., and H. T. McMahon. 2005. BAR domains and membrane brane receptor signals its ligand-stimulated endocytosis. Cell 84:277–287. curvature: bringing your curves to the BAR. Biochem. Soc. Symp. 72:223–231. 118. Hicke, L., B. Zanolari, M. Pypaert, J. Rohrer, and H. Riezman. 1997. Gammie, A. E., V. Brizzio, and M. D. Rose. 91. 1998. Distinct morphological Transport through the yeast endocytic pathway occurs through morpholog- 9: phenotypes of cell fusion mutants. Mol. Biol. Cell 1395–1410. ically distinct compartments and requires an active secretory pathway and 92. Gammie, A. E., L. J. Kurihara, R. B. Vallee, and M. D. Rose. 1995. DNM1, Sec18p/N-ethylmaleimide-sensitive fusion protein. Mol. Biol. Cell 8:13–31. a dynamin-related gene, participates in endosomal trafficking in yeast. J. Cell 119. Higgs, H. N., and T. D. Pollard. 2001. Regulation of actin filament network Biol. 130:553–566. formation through ARP2/3 complex: activation by a diverse array of pro- 93. Ge, K., J. DuHadaway, W. Du, M. Herlyn, U. Rodeck, and G. C. Prendergast. teins. Annu. Rev. Biochem. 70:649–676. 1999. Mechanism for elimination of a tumor suppressor: aberrant splicing of 120. Hill, E., J. van Der Kaay, C. P. Downes, and E. Smythe. 2001. The role of a brain-specific exon causes loss of function of Bin1 in melanoma. Proc. dynamin and its binding partners in coated pit invagination and scission. Natl. Acad. Sci. USA 96:9689–9694. J. Cell Biol. 152:309–323. 94. Ge, K., and G. C. Prendergast. 2000. Bin2, a functionally nonredundant 121. Hinshaw, J. E., and S. L. Schmid. 1995. Dynamin self-assembles into rings member of the BAR adaptor gene family. Genomics 67:210–220. suggesting a mechanism for coated vesicle budding. Nature 374:190–192. 95. Geli, M. I., and H. Riezman. 1998. Endocytic internalization in yeast and 122. Hitchcock, A. L., K. Auld, S. P. Gygi, and P. A. Silver. 2003. A subset of animal cells: similar and different. J. Cell Sci. 111:1031–1037. membrane-associated proteins is ubiquitinated in response to mutations in 96. Geli, M. I., and H. Riezman. 1996. Role of type I myosins in receptor- the endoplasmic reticulum degradation machinery. Proc. Natl. Acad. Sci. mediated endocytosis in yeast. Science 272:533–535. USA 100:12735–12740. 97. Germann, M., E. Swain, L. Bergman, and J. T. Nickels, Jr. 2005. Charac- 123. Ho, Y., A. Gruhler, A. Heilbut, G. D. Bader, L. Moore, S. L. Adams, A. terizing the sphingolipid signaling pathway that remediates defects associ- Millar, P. Taylor, K. Bennett, K. Boutilier, L. Yang, C. Wolting, I. Donaldson, ated with loss of the yeast amphiphysin-like orthologs, Rvs161p and S. Schandorff, J. Shewnarane, M. Vo, J. Taggart, M. Goudreault, B. Rvs167p. J. Biol. Chem. 280:4270–4278. Muskat, C. Alfarano, D. Dewar, Z. Lin, K. Michalickova, A. R. Willems, 98. Gerst, J. E., L. Rodgers, M. Riggs, and M. Wigler. 1992. SNC1, a yeast H. Sassi, P. A. Nielsen, K. J. Rasmussen, J. R. Andersen, L. E. Johansen, homolog of the synaptic vesicle-associated membrane protein/synaptobre- L. H. Hansen, H. Jespersen, A. Podtelejnikov, E. Nielsen, J. Crawford, V. vin gene family: genetic interactions with the RAS and CAP genes. Proc. Poulsen, B. D. Sorensen, J. Matthiesen, R. C. Hendrickson, F. Gleeson, T. Natl. Acad. Sci. USA 89:4338–4342. Pawson, M. F. Moran, D. Durocher, M. Mann, C. W. Hogue, D. Figeys, and 99. Giachino, C., E. Lantelme, L. Lanzetti, S. Saccone, G. Bella Valle, and N. M. Tyers. 2002. Systematic identification of protein complexes in Saccha- Migone. 1997. A novel SH3-containing human gene family preferentially romyces cerevisiae by mass spectrometry. Nature 415:180–183. expressed in the central nervous system. Genomics 41:427–434. 124. Hoepfner, D., M. van den Berg, P. Philippsen, H. F. Tabak, and E. H. 100. Gold, E. S., N. S. Morrissette, D. M. Underhill, J. Guo, M. Bassetti, and A. Hettema. 2001. A role for Vps1p, actin, and the Myo2p motor in peroxi- Aderem. 2000. Amphiphysin IIm, a novel amphiphysin II isoform, is re- some abundance and inheritance in Saccharomyces cerevisiae. J. Cell Biol. quired for macrophage phagocytosis. Immunity 12:285–292. 155:979–990. 101. Goode, B. L., A. A. Rodal, G. Barnes, and D. G. Drubin. 2001. Activation of 125. Holtzman, D. A., K. F. Wertman, and D. G. Drubin. 1994. Mapping actin VOL. 70, 2006 BAR DOMAIN PROTEINS 115

surfaces required for functional interactions in-vivo. J. Cell Biol. 126:423– 151. Kessels, M. M., and B. Qualmann. 2004. The syndapin protein family: 432. linking membrane trafficking with the cytoskeleton. J. Cell Sci. 117:3077– 126. Holtzman, D. A., S. Yang, and D. G. Drubin. 1993. Synthetic-lethal inter- 3086. actions identify two novel genes, SLA1 and SLA2, that control membrane 152. Ketela, T., R. Green, and H. Bussey. 1999. Saccharomyces cerevisiae Mid2p cytoskeleton assembly in Saccharomyces cerevisiae. J. Cell Biol. 122:635– is a potential cell wall stress sensor and upstream activator of the PKC1- 644. MPK1 cell integrity pathway. J. Bacteriol. 181:3330–3340. 127. Horvath, A., C. Sutterlin, U. Manning Krieg, N. R. Movva, and H. Riezman. 153. Kim, D. W. 2003. Characterization of Grp1p, a novel cis-Golgi matrix 1994. Ceramide synthesis enhances transport of GPI-anchored proteins to protein. Biochem. Biophys. Res. Commun. 303:370–378. the Golgi apparatus in yeast. EMBO J. 13:3687–3695. 154. Koch, A., M. Thiemann, M. Grabenbauer, Y. Yoon, M. A. McNiven, and M. 128. Hosoya, O., and K. Tsutsui. 2004. Localized expression of amphiphysin Ir, Schrader. 2003. Dynamin-like protein 1 is involved in peroxisomal fission.

a retina-specific variant of amphiphysin I, in the ribbon synapse and its J. Biol. Chem. 278:8597–8605. Downloaded from functional implication. Eur. J. Neuroscie. 19:2179–2187. 155. Koenig, J. H., and K. Ikeda. 1989. Disappearance and reformation of 129. Huang, D., G. Patrick, J. Moffat, L. H. Tsai, and B. Andrews. 1999. Mam- synaptic vesicle membrane upon transmitter release observed under revers- malian Cdk5 is a functional homologue of the budding yeast Pho85 cyclin- ible blockage of membrane retrieval. J. Neurosci. the official J. Soc. Neu- dependent protein kinase. Proc. Natl. Acad. Sci. USA 96:14445–14450. rosci. 9:3844–3860. 130. Huang, K. M., K. D’Hondt, H. Riezman, and S. K. Lemmon. 1999. Clathrin 156. Kolesnick, R. N., and M. Kronke. 1998. Regulation of ceramide production functions in the absence of heterotetrameric adaptors and AP180-related and apoptosis. Annu. Rev. Physiol. 60:643–665. proteins in yeast. EMBO J. 18:3897–3908. 157. Krueger, E. W., J. D. Orth, H. Cao, and M. A. McNiven. 2003. A dynamin- 131. Huang, T. Y., M. Renaud Young, and D. Young. 2005. Nak1 interacts with cortactin-Arp2/3 complex mediates actin reorganization in growth factor- Hob1 and Wsp1 to regulate cell growth and polarity in Schizosaccharomy- stimulated cells. Mol. Biol. Cell 14:1085–1096. ces pombe. J. Cell Sci. 118:199–210. 158. Kurihara, L. J., C. T. Beh, M. Latterich, R. Schekman, and M. D. Rose. 132. Huckaba, T. M., A. C. Gay, L. F. Pantalena, H. C. Yang, and L. A. Pon. 1994. Nuclear congression and membrane fusion: two distinct events in the http://mmbr.asm.org/ 2004. Live cell imaging of the assembly, disassembly, and actin cable- yeast karyogamy pathway. J. Cell Biol. 126:911–923. dependent movement of endosomes and actin patches in the budding yeast, 159. Kurzchalia, T. V., and R. G. Parton. 1999. Membrane microdomains and Saccharomyces cerevisiae. J. Cell Biol. 167:519–530. caveolae. Curr. Opin. Cell Biol. 11:424–431. 133. Hughes, A. C., R. Errington, R. Fricker Gates, and L. Jones. 2004. En- 160. Landgraf, C., S. Panni, L. Montecchi Palazzi, L. Castagnoli, J. Schneider dophilin A3 forms filamentous structures that colocalise with microtubules Mergener, R. Volkmer Engert, and G. Cesareni. 2004. Protein interaction but not with actin filaments. Brain research. Mol. brain research. 128:182– networks by proteome peptide scanning. PLoS Biol. 2:E14. 192. 161. Lechler, T., A. Shevchenko, and R. Li. 2000. Direct involvement of yeast 134. Huttner, W. B., and A. Schmidt. 2000. Lipids, lipid modification and lipid- type I myosins in Cdc42-dependent actin polymerization. J. Cell Biol. 148: protein interaction in membrane budding and fission—insights from the 363–373. roles of endophilin A1 and synaptophysin in synaptic vesicle endocytosis. 162. Lee, E., M. Marcucci, L. Daniell, M. Pypaert, O. A. Weisz, G. C. Ochoa, K. Curr. Opin. Neurobiol. 10:543–551. Farsad, M. R. Wenk, and P. De Camilli. 2002. Amphiphysin 2 (Bin1) and 135. Iida, H., and I. Yahara. 1984. Specific early-G1 blocks accompanied with T-tubule biogenesis in muscle. Science 297:1193–1196.

stringent response in Saccharomyces cerevisiae lead to growth arrest in 163. Lee, J., K. Colwill, V. Aneliunas, C. Tennyson, L. Moore, Y. Ho, and B. on October 26, 2015 by University of Queensland Library resting state similar to the G0 of higher eucaryotes. J. Cell Biol. 98:1185– Andrews. 1998. Interaction of yeast Rvs167 and Pho85 cyclin-dependent 1193. kinase complexes may link the cell cycle to the actin cytoskeleton. Curr. 136. Ito, T., T. Chiba, R. Ozawa, M. Yoshida, M. Hattori, and Y. Sakaki. 2001. Biol. 8:1310–1321. A comprehensive two-hybrid analysis to explore the yeast protein interac- 164. Lee, M. C., and R. Schekman. 2004. BAR domains go on a bender. Science tome. Proc. Natl. Acad. Sci. USA 98:4569–4574. 303:479–480. 137. Johnston, G. C., C. W. Ehrhardt, A. Lorincz, and B. L. Carter. 1979. 165. Lee, S. Y., M. R. Wenk, Y. Kim, A. C. Nairn, and P. De Camilli. 2004. Regulation of cell size in the yeast Saccharomyces cerevisiae. J. Bacteriol. Regulation of synaptojanin 1 by cyclin-dependent kinase 5 at synapses. 137:1–5. Proc. Natl. Acad. Sci. USA 101:546–551. 138. Johnston, G. C., J. A. Prendergast, and R. A. Singer. 1991. The Saccharo- 166. Legesse Miller, A., R. H. Massol, and T. Kirchhausen. 2003. Constriction myces cerevisiae MYO2 gene encodes an essential myosin for vectorial and Dnm1p recruitment are distinct processes in mitochondrial fission. transport of vesicles. J. Cell Biol. 113:539–551. Mol. Biol. Cell 14:1953–1963. 139. Jonsdottir, G. A., and R. Li. 2004. Dynamics of yeast Myosin I: evidence for 167. Leprince, C., E. Le Scolan, B. Meunier, V. Fraisier, N. Brandon, J. De a possible role in scission of endocytic vesicles. Curr. Biol. 14:1604–1609. Gunzburg, and J. Camonis. 2003. Sorting nexin 4 and amphiphysin 2, a new 140. Juschke, C., A. Wachter, B. Schwappach, and M. Seedorf. 2005. SEC18/ partnership between endocytosis and intracellular trafficking. J. Cell Sci. NSF-independent, protein-sorting pathway from the yeast cortical ER to 116:1937–1948. the plasma membrane. J. Cell Biol. 169:613–622. 168. Leprince, C., F. Romero, D. Cussac, B. Vayssiere, R. Berger, A. Tavitian, 141. Kadlec, L., and A. M. Pendergast. 1997. The amphiphysin-like protein 1 and J. H. Camonis. 1997. A new member of the amphiphysin family con- (ALP1) interacts functionally with the cABL tyrosine kinase and may play necting endocytosis and signal transduction pathways. J. Biol. Chem. 272: a role in cytoskeletal regulation. Proc. Natl. Acad. Sci. USA 94:12390– 15101–15105. 12395. 169. Leventis, P. A., B. M. Chow, B. A. Stewart, B. Iyengar, A. R. Campos, and 142. Kajiho, H., K. Saito, K. Tsujita, K. Kontani, Y. Araki, H. Kurosu, and T. G. L. Boulianne. 2001. Drosophila amphiphysin is a post-synaptic protein Katada. 2003. RIN3: a novel Rab5 GEF interacting with amphiphysin II required for normal locomotion but not endocytosis. Traffic 2:839–850. involved in the early endocytic pathway. J. Cell Sci. 116:4159–4168. 170. Lew, D. J., and S. I. Reed. 1992. A proliferation of cyclins. Trends Cell Biol. 143. Kaksonen, M., Y. Sun, and D. G. Drubin. 2003. A pathway for association 2:77–81. of receptors, adaptors, and actin during endocytic internalization. Cell 171. Lew, J., Q. Q. Huang, Z. Qi, R. J. Winkfein, R. Aebersold, T. Hunt, and 115:475–487. J. H. Wang. 1994. A brain-specific activator of cyclin-dependent kinase 5. 144. Kaksonen, M., C. P. Toret, and D. G. Drubin. 2005. A modular design for Nature 371:423–426. the clathrin- and actin-mediated endocytosis machinery. Cell 123:305–320. 172. Li, F. Q., A. Coonrod, and M. Horwitz. 2000. Selection of a dominant 145. Kamada, Y., T. Sekito, and Y. Ohsumi. 2004. Autophagy in yeast: a TOR- negative retinoblastoma protein (RB) inhibiting satellite myoblast differ- mediated response to nutrient starvation. Curr. Top. Microbiol. Immunol. entiation implies an indirect interaction between MyoD and RB. Mol. Cell. 279:73–84. Biol. 20:5129–5139. 146. Kaminska, J., B. Gajewska, A. K. Hopper, and T. Zoladek. 2002. Rsp5p, a 173. Li, R. 1997. Bee1, a yeast protein with homology to Wiscott-Aldrich syn- new link between the actin cytoskeleton and endocytosis in the yeast Sac- drome protein, is critical for the assembly of cortical actin cytoskeleton. charomyces cerevisiae. Mol. Cell. Biol. 22:6946–6948. J. Cell Biol. 136:649–658. 147. Kanoh, H., B. T. Williger, and J. H. Exton. 1997. Arfaptin 1, a putative 174. Li, X., and S. J. Gould. 2003. The dynamin-like GTPase DLP1 is essential cytosolic target protein of ADP-ribosylation factor, is recruited to Golgi for peroxisome division and is recruited to peroxisomes in part by PEX11. membranes. J. Biol. Chem. 272:5421–5429. J. Biol. Chem. 278:17012–17020. 148. Karbowski, M., S. Y. Jeong, and R. J. Youle. 2004. Endophilin B1 is 175. Lichte, B., R. W. Veh, H. E. Meyer, and M. W. Kilimann. 1992. Amphiphy- required for the maintenance of mitochondrial morphology. J. Cell Biol. sin, a novel protein associated with synaptic vesicles. EMBO J. 11:2521– 166:1027–1039. 2530. 149. Karpova, T. S., J. G. McNally, S. L. Moltz, and J. A. Cooper. 1998. Assem- 176. Lila, T., and D. G. Drubin. 1997. Evidence for physical and functional bly and function of the actin cytoskeleton of yeast: relationships between interactions among two Saccharomyces cerevisiae SH3 domain proteins, an cables and patches. J. Cell Biol. 142:1501–1517. adenylyl cyclase-associated protein and the actin cytoskeleton. Mol. Biol. 150. Karpova, T. S., S. L. Reck Peterson, N. B. Elkind, M. S. Mooseker, P. J. Cell 8:367–385. Novick, and J. A. Cooper. 2000. Role of actin and Myo2p in polarized 177. Lillie, S. H., and J. R. Pringle. 1980. Reserve carbohydrate metabolism in secretion and growth of Saccharomyces cerevisiae. Mol. Biol. Cell 11:1727– Saccharomyces cerevisiae: responses to nutrient limitation. J. Bacteriol 143: 1737. 1384–1394. 116 REN ET AL. MICROBIOL.MOL.BIOL.REV.

178. Lippincott, J., and R. Li. 1998. Sequential assembly of myosin II, an IQGAP- 1999. Visualization of receptor-mediated endocytosis in yeast. Mol. Biol. like protein, and filamentous actin to a ring structure involved in budding yeast Cell 10:799–817. cytokinesis. J. Cell Biol. 140:355–366. 205. Mulholland, J., D. Preuss, A. Moon, A. Wong, D. Drubin, and D. Botstein. 179. Lodder, A. L., T. K. Lee, and R. Ballester. 1999. Characterization of the 1994. Ultrastructure of the yeast actin cytoskeleton and its association with Wsc1 protein, a putative receptor in the stress response of Saccharomyces the plasma membrane. J. Cell Biol. 125:381–391. cerevisiae. Genetics 152:1487–1499. 206. Mulholland, J., A. Wesp, H. Riezman, and D. Botstein. 1997. Yeast actin 180. Lombardi, R., and H. Riezman. 2001. Rvs161p and Rvs167p, the two yeast cytoskeleton mutants accumulate a new class of Golgi-derived secretary amphiphysin homologs, function together in vivo. J. Biol. Chem. 276:6016– vesicle. Mol. Biol. Cell 8:1481–1499. 6022. 207. Muller, A. J., J. F. Baker, J. B. DuHadaway, K. Ge, G. Farmer, P. S. 181. Madania, A., P. Dumoulin, S. Grava, H. Kitamoto, C. Scharer Brodbeck, A. Donover, R. Meade, C. Reid, R. Grzanna, A. H. Roach, N. Shah, A. P. Soler,

Soulard, V. Moreau, and B. Winsor. 1999. The Saccharomyces cerevisiae and G. C. Prendergast. 2003. Targeted disruption of the murine Bin1/ Downloaded from homologue of human Wiskott-Aldrich syndrome protein Las17p interacts amphiphysin II gene does not disable endocytosis but results in embryonic with the Arp2/3 complex. Mol. Biol. Cell 10:3521–3538. cardiomyopathy with aberrant myofibril formation. Mol. Cell. Biol. 23: 182. Mao, N. C., E. Steingrimsson, J. DuHadaway, W. Wasserman, J. C. Ruiz, 4295–4306. N. G. Copeland, N. A. Jenkins, and G. C. Prendergast. 1999. The murine 208. Muller, A. J., J. B. Duhadaway, P. S. Donover, E. Sutanto Ward, and G. C. Bin1 gene functions early in myogenesis and defines a new region of synteny Prendergast. 2005. Inhibition of indoleamine 2,3-dioxygenase, an immuno- between mouse chromosome 18 and human chromosome 2. Genomics regulatory target of the cancer suppression gene Bin1, potentiates cancer 56:51–58. chemotherapy. Nat. Med.. 11:312–319. 183. Marks, B., and H. T. McMahon. 1998. Calcium triggers calcineurin-depen- 209. Muller, A. J., J. B. DuHadaway, P. S. Donover, E. Sutanto Ward, and G. C. dent synaptic vesicle recycling in mammalian nerve terminals. Curr. Biol. Prendergast. 2004. Targeted deletion of the suppressor gene bin1/am- 8:740–749. phiphysin2 accentuates the neoplastic character of transformed mouse fi- broblasts. Cancer Biol. Ther. 3:1236–1242. 184. Masur, S. K., Y. T. Kim, and C. F. Wu. 1990. Reversible inhibition of http://mmbr.asm.org/ endocytosis in cultured neurons from the Drosophila temperature-sensitive 210. Muller, E. M., N. A. Mackin, S. E. Erdman, and K. W. Cunningham. 2003. mutant shibirets1. J. Neurogenet. 6:191–206. Fig1p facilitates Ca2ϩ influx and cell fusion during mating of Saccharomy- 185. Matsumoto, K., I. Uno, and T. Ishikawa. 1983. Control of cell division in ces cerevisiae. J. Biol. Chem. 278:38461–38469. Saccharomyces cerevisiae mutants defective in adenylate cyclase and 211. Muller Taubenberger, A., A. N. Lupas, H. Li, M. Ecke, E. Simmeth, and G. cAMP-dependent protein kinase. Exp. Cell Res. 146:151–161. Gerisch. 2001. Calreticulin and calnexin in the endoplasmic reticulum are 186. McFarlane, E. S. 1980. Ribonuclease activity during G1 arrest of the yeast important for phagocytosis. EMBO J. 20:6772–6782. Saccharomyces cerevisiae. Arch. Microbiol. 124:243–247. 212. Mundigl, O., G. C. Ochoa, C. David, V. I. Slepnev, A. Kabanov, and P. De 187. McMahon, H. T., P. Wigge, and C. Smith. 1997. Clathrin interacts specif- Camilli. 1998. Amphiphysin I antisense oligonucleotides inhibit neurite ically with amphiphysin and is displaced by dynamin. FEBS Lett. 413:319– outgrowth in cultured hippocampal neurons. J. Neurosci. 18:93–103. 322. 213. Munn, A. L. 2001. Molecular requirements for the internalisation step of 188. McPherson, P. S., E. P. Garcia, V. I. Slepnev, C. David, X. Zhang, D. Grabs, endocytosis: insights from yeast. Biochim. Biophys. Acta 1535:236–257. W. S. Sossin, R. Bauerfeind, Y. Nemoto, and P. De Camilli. 1996. A 214. Munn, A. L. 2000. The yeast endocytic membrane transport system. Mi- presynaptic inositol-5-phosphatase. Nature 379:353–357. croscopy research and technique 51:547–562. on October 26, 2015 by University of Queensland Library 189. McPherson, P. S., K. Takei, S. L. Schmid, and P. De Camilli. 1994. p145, 215. Munn, A. L., B. J. Stevenson, M. I. Geli, and H. Riezman. 1995. end5, end6, a major Grb2-binding protein in brain, is colocalized with dynamin in nerve and end7: mutations that cause actin delocalization and block the internal- terminals where it undergoes activity-dependent dephosphorylation. J. Biol. ization step of endocytosis in Saccharomyces cerevisiae. Mol. Biol. Cell Chem. 269:30132–30139. 6:1721–1742. 216. Murphy, J. E., I. T. Pleasure, S. Puszkin, K. Prasad, and J. H. Keen. 1991. 190. Measday, V., L. Moore, R. Retnakaran, J. Lee, M. Donoviel, A. M. Neiman, Clathrin assembly protein AP-3. The identity of the 155K protein, AP 180, and B. Andrews. 1997. A family of cyclin-like proteins that interact with the and NP185 and demonstration of a clathrin binding domain. J. Biol. Chem. Pho85 cyclin-dependent kinase. Mol. Cell. Biol. 17:1212–1223. 266:4401–4408. 191. Miaczynska, M., S. Christoforidis, A. Giner, A. Shevchenko, S. Uttenweiler 217. Na, S., M. Hincapie, J. H. McCusker, and J. E. Haber. 1995. MOP2 (SLA2) Joseph, B. Habermann, M. Wilm, R. G. Parton, and M. Zerial. 2004. APPL ϩ affects the abundance of the plasma membrane H -ATPase of Saccharo- proteins link Rab5 to nuclear signal transduction via an endosomal com- myces cerevisiae. J. Biol. Chem. 270:6815–6823. partment. Cell 116:445–456. 218. Naqvi, S. N., Q. Feng, V. J. Boulton, R. Zahn, and A. L. Munn. 2001. Vrp1p 192. Micheva, K. D., B. K. Kay, and P. S. McPherson. 1997. Synaptojanin forms functions in both actomyosin ring-dependent and Hof1p-dependent path- two separate complexes in the nerve terminal. Interactions with endophilin ways of cytokinesis. Traffic 2:189–201. and amphiphysin. J. Biol. Chem. 272:27239–27245. 219. Naqvi, S. N., R. Zahn, D. A. Mitchell, B. J. Stevenson, and A. L. Munn. 1998. Micheva, K. D., A. R. Ramjaun, B. K. Kay, and P. S. McPherson. 193. 1997. SH3 The WASp homologue Las17p functions with the WIP homologue End5p/ domain-dependent interactions of endophilin with amphiphysin. FEBS verprolin and is essential for endocytosis in yeast. Curr. Biol. 8:959–962. Lett. 414:308–312. 220. Navarro, P., P. Durrens, and M. Aigle. 1997. Protein-protein interaction 194. Miele, A. E., P. J. Watson, P. R. Evans, L. M. Traub, and D. J. Owen. 2004. between the RVS161 and RVS167 gene products of Saccharomyces cerevi- Two distinct interaction motifs in amphiphysin bind two independent sites siae. Biochim. Biophys. Acta 1343:187–192. on the clathrin terminal domain beta-propeller. 11:242–248. 221. Nelson, B., A. B. Parsons, M. Evangelista, K. Schaefer, K. Kennedy, S. 195. Millard, T. H., G. Bompard, M. Y. Heung, T. R. Dafforn, D. J. Scott, L. M. Ritchie, T. L. Petryshen, and C. Boone. 2004. Fus1p interacts with compo- Machesky, and K. Futterer. 2005. Structural basis of filopodia formation nents of the Hog1p mitogen-activated protein kinase and Cdc42p morpho- induced by the IRSp53/MIM homology domain of human IRSp53. EMBO genesis signaling pathways to control cell fusion during yeast mating. Ge- J. 24:240–250. netics 166:67–77. 196. Modregger, J., A. A. Schmidt, B. Ritter, W. B. Huttner, and M. Plomann. 222. Nemoto, Y., M. Arribas, C. Haffner, and P. DeCamilli. 1997. Synaptojanin 2003. Characterization of endophilin B1b, a brain-specific membrane-asso- 2, a novel synaptojanin isoform with a distinct targeting domain and ex- ciated lysophosphatidic acid acyl transferase with properties distinct from pression pattern. J. Biol. Chem. 272:30817–30821. endophilin A1. J. Biol. Chem. 278:4160–4167. 223. Newpher, T. M., R. P. Smith, V. Lemmon, and S. K. Lemmon. 2005. In vivo 197. Moreau, V., A. Madania, R. P. Martin, and B. Winson. 1996. The Saccha- dynamics of clathrin and its adaptor-dependent recruitment to the actin- romyces cerevisiae actin-related protein Arp2 is involved in the actin cy- based endocytic machinery in yeast. Dev. Cell 9:87–98. toskeleton. J. Cell Biol. 134:117–132. 224. Nickels, J. T., and J. R. Broach. 1996. A ceramide-activated protein phos- 198. Moreno Borchart, A. C., and M. Knop. 2003. Prospore membrane forma- phatase mediates ceramide-induced G1 arrest of Saccharomyces cerevisiae. tion: how budding yeast gets shaped in meiosis. Microbiol. Res. 158:83–90. Genes Dev. 10:382–394. 199. Moreno, E., and K. Basler. 2004. dMyc transforms cells into super-com- 225. Nishizawa, M., Y. Kanaya, and A. Toh. 1999. Mouse cyclin-dependent petitors. Cell 117:117–129. kinase (Cdk) 5 is a functional homologue of a yeast Cdk, pho85 kinase. 200. Morgan, D. O. 1995. Principles of CDK regulation. Nature 374:131–134. J. Biol. Chem. 274:33859–33862. 201. Morishita, M., and J. Engebrecht. 2005. End3p-mediated endocytosis is 226. Nothwehr, S. F., E. Conibear, and T. H. Stevens. 1995. Golgi and vacuolar required for spore wall formation in Saccharomyces cerevisiae. Genetics membrane proteins reach the vacuole in vps1 mutant yeast cells via the 170:1561–1574. plasma membrane. J. Cell Biol. 129:35–46. 202. Morsomme, P., and H. Riezman. 2002. The Rab GTPase Ypt1p and tethering 227. Nuoffer, C., H. W. Davidson, J. Matteson, J. Meinkoth, and W. E. Balch. factors couple protein sorting at the ER to vesicle targeting to the Golgi 1994. A GDP-bound of rab1 inhibits protein export from the endoplasmic apparatus. Dev. Cell 2:307–317. reticulum and transport between Golgi compartments. J. Cell Biol. 125: 203. Muhlberg, A. B., D. E. Warnock, and S. L. Schmid. 1997. Domain structure 225–237. and intramolecular regulation of dynamin GTPase. EMBO J. 16:6676– 228. Obeid, L. M., C. M. Linardic, L. A. Karolak, and Y. A. Hannun. 1993. 6683. Programmed cell death induced by ceramide. Science 259:1769–1771. 204. Mulholland, J., J. Konopka, B. Singer Kruger, M. Zerial, and D. Botstein. 229. Otsuga, D., B. R. Keegan, E. Brisch, J. W. Thatcher, G. J. Hermann, W. VOL. 70, 2006 BAR DOMAIN PROTEINS 117

Bleazard, and J. M. Shaw. 1998. The dynamin-related GTPase, Dnm1p, action with the germinal center kinase-like kinase. J. Biol. Chem. 276: controls mitochondrial morphology in yeast. J. Cell Biol. 143:333–349. 28913–28919. 230. Otsuki, M., T. Itoh, and T. Takenawa. 2003. Neural Wiskott-Aldrich syn- 255. Ramjaun, A. R., and P. S. McPherson. 1998. Multiple amphiphysin II splice drome protein is recruited to rafts and associates with endophilin A in variants display differential clathrin binding: identification of two distinct response to epidermal growth factor. J. Biol. Chem. 278:6461–6469. clathrin-binding sites. J. Neurochem. 70:2369–2376. 231. Owen, D. J., Y. Vallis, M. E. Noble, J. B. Hunter, T. R. Dafforn, P. R. Evans, 256. Ramjaun, A. R., K. D. Micheva, I. Bouchelet, and P. S. McPherson. 1997. and H. T. McMahon. 1999. A structural explanation for the binding of Identification and characterization of a nerve terminal-enriched amphiphy- multiple ligands by the alpha-adaptin appendage domain. Cell 97:805–815. sin isoform. J. Biol. Chem. 272:16700–16706. 232. Owen, D. J., P. Wigge, Y. Vallis, J. D. Moore, P. R. Evans, and H. T. 257. Ramjaun, A. R., J. Philie, E. de Heuvel, and P. S. McPherson. 1999. The N McMahon. 1998. Crystal structure of the amphiphysin-2 SH3 domain and terminus of amphiphysin II mediates dimerization and plasma membrane

its role in the prevention of dynamin ring formation. EMBO J. 17:5273– targeting. J. Biol. Chem. 274:19785–19791. Downloaded from 5285. 258. Razzaq, A., I. M. Robinson, H. T. McMahon, J. N. Skepper, Y. Su, A. C. 233. Parton, R. G., and J. F. Hancock. 2004. Lipid rafts and plasma membrane Zelhof, A. P. Jackson, N. J. Gay, and C. J. O’Kane. 2001. Amphiphysin is microorganization: insights from Ras. Trends Cell Biol. 14:141–147. necessary for organization of the excitation-contraction coupling machinery 234. Pawson, T., and G. D. Gish. 1992. SH2 and SH3 domains: from structure to of muscles, but not for synaptic vesicle endocytosis in Drosophila. Genes function. Cell 71:359–362. Dev. 15:2967–2979. 235. Payne, G. S., D. Baker, E. van Tuinen, and R. Schekman. 1988. Protein 259. Razzaq, A., Y. Su, J. E. Mehren, K. Mizuguchi, A. P. Jackson, N. J. Gay, transport to the vacuole and receptor-mediated endocytosis by clathrin and C. J. O’Kane. 2000. Characterisation of the gene for Drosophila am- heavy chain-deficient yeast. J. Cell Biol. 106:1453–1461. phiphysin. Gene 241:167–174. 236. Peng, J., D. Schwartz, J. E. Elias, C. C. Thoreen, D. Cheng, G. Marsischky, 260. Ren, G., J. Wang, R. Brinkworth, B. Winsor, B. Kobe, and A. L. Munn. J. Roelofs, D. Finley, and S. P. Gygi. 2003. A proteomics approach to 2005. Verprolin cytokinesis function mediated by the Hof one trap domain. understanding protein ubiquitination. Nat. Biotechnol. 21:921–926. Traffic 6:575–593. http://mmbr.asm.org/ 237. Peter, B. J., H. M. Kent, I. G. Mills, Y. Vallis, P. J. Butler, P. R. Evans, and 261. Ren, R., B. J. Mayer, P. Cicchetti, and D. Baltimore. 1993. Identification of H. T. McMahon. 2004. BAR domains as sensors of membrane curvature: a ten-amino acid proline-rich SH3 binding site. Science 259:1157–1161. the amphiphysin BAR structure. Science 303:495–499. 262. Revardel, E., M. Bonneau, P. Durrens, and M. Aigle. 1995. Characteriza- 238. Peter, F., C. Nuoffer, S. N. Pind, and W. E. Balch. 1994. Guanine nucleotide tion of a new gene family developing pleiotropic phenotypes upon mutation dissociation inhibitor is essential for Rab1 function in budding from the in Saccharomyces cerevisiae. Biochim. Biophys. Acta 1263:261–265. endoplasmic reticulum and transport through the Golgi stack. J. Cell Biol. 263. Richnau, N., and P. Aspenstrom. 2001. Rich, a rho GTPase-activating 126:1393–1406. protein domain-containing protein involved in signaling by Cdc42 and 239. Peter, F., H. Plutner, H. Zhu, T. E. Kreis, and W. E. Balch. 1993. Beta-COP Rac1. J. Biol. Chem. 276:35060–35070. is essential for transport of protein from the endoplasmic reticulum to the 264. Richnau, N., A. Fransson, K. Farsad, and P. Aspenstrom. 2004. RICH-1 Golgi in vitro. J. Cell Biol. 122:1155–1167. has a BIN/Amphiphysin/Rvsp domain responsible for binding to membrane 240. Petrelli, A., G. F. Gilestro, S. Lanzardo, P. M. Comoglio, N. Migone, and S. lipids and tubulation of liposomes. Biochem. Biophys. Res. Commun. 320: Giordano. 2002. The endophilin-CIN85-Cbl complex mediates ligand-de- 1034–1042.

pendent downregulation of c-Met. Nature 416:187–190. 265. Riezman, H. 1985. Endocytosis in yeast: several of the yeast secretory on October 26, 2015 by University of Queensland Library 241. Philip, B., and D. E. Levin. 2001. Wsc1 and Mid2 are cell surface sensors for mutants are defective in endocytosis. Cell 40:1001–1009. cell wall integrity signaling that act through Rom2, a guanine nucleotide 266. Riezman, H., A. Munn, M. I. Geli, and L. Hicke. 1996. Actin-, myosin- and exchange factor for Rho1. Mol. Cell. Biol. 21:271–280. ubiquitin-dependent endocytosis. Experientia 52:1033–1041. 242. Philpott, A., E. B. Porro, M. W. Kirschner, and L. H. Tsai. 1997. The role 267. Rikhy, R., V. Kumar, R. Mittal, and K. S. Krishnan. 2002. Endophilin is of cyclin-dependent kinase 5 and a novel regulatory subunit in regulating critically required for synapse formation and function in Drosophila mela- muscle differentiation and patterning. Genes Dev. 11:1409–1421. nogaster. J. Neurosci. 22:7478–7484. 243. Pichler, H., B. Gaigg, C. Hrastnik, G. Achleitner, S. D. Kohlwein, G. 268. Ringstad, N., H. Gad, P. Low, G. Di Paolo, L. Brodin, O. Shupliakov, and Zellnig, A. Perktold, and G. Daum. 2001. A subfraction of the yeast endo- P. De Camilli. 1999. Endophilin/SH3p4 is required for the transition from plasmic reticulum associates with the plasma membrane and has a high early to late stages in clathrin-mediated synaptic vesicle endocytosis. Neu- capacity to synthesize lipids. European J. Biochem. FEBS. 268:2351–2361. ron 24:143–154. 244. Pierrat, B., M. Simonen, M. Cueto, J. Mestan, P. Ferrigno, and J. Heim. 269. Ringstad, N., Y. Nemoto, and P. De Camilli. 2001. Differential expression of 2001. SH3GLB, a new endophilin-related protein family featuring an SH3 endophilin 1 and 2 dimers at central nervous system synapses. J. Biol. domain. Genomics 71:222–234. Chem. 276:40424–40430. 245. Pineda Lucena, A., C. S. Ho, D. Y. Mao, Y. Sheng, R. C. Laister, R. 270. Ringstad, N., Y. Nemoto, and P. De Camilli. 1997. The SH3p4/Sh3p8/ Muhandiram, Y. Lu, B. T. Seet, S. Katz, T. Szyperski, L. Z. Penn, and C. H. SH3p13 protein family: binding partners for synaptojanin and dynamin via Arrowsmith. 2005. A structure-based model of the c-Myc/Bin1 protein a Grb2-like Src homology 3 domain. Proc. Natl. Acad. Sci. USA 94:8569– interaction shows alternative splicing of Bin1 and c-Myc phosphorylation 8574. are key binding determinants. J. Mol. Biol. 351:182–194. 271. Robinson, P. J., J. M. Sontag, J. P. Liu, E. M. Fykse, C. Slaughter, H. 246. Pitts, K. R., Y. Yoon, E. W. Krueger, and M. A. McNiven. 1999. The McMahon, and T. C. Sudhof. 1993. Dynamin GTPase regulated by protein dynamin-like protein DLP1 is essential for normal distribution and mor- kinase C phosphorylation in nerve terminals. Nature 365:163–166. phology of the endoplasmic reticulum and mitochondria in mammalian 272. Rosales, J. L., M. J. Nodwell, R. N. Johnston, and K. Y. Lee. 2000. Cdk5/ cells. Mol. Biol. Cell 10:4403–4417. p25nck5a interaction with synaptic proteins in bovine brain. J. Cell. Biochem. 247. Plesset, J., J. R. Ludwig, B. S. Cox, and C. S. McLaughlin. 1987. Effect of 78:151–159. cell cycle position on thermotolerance in Saccharomyces cerevisiae. J. Bac- 273. Rossanese, O. W., C. A. Reinke, B. J. Bevis, A. T. Hammond, I. B. Sears, J. teriol. 169:779–784. O’Connor, and B. S. Glick. 2001. A role for actin, Cdc1p, and Myo2p in the 248. Plutner, H., A. D. Cox, S. Pind, R. Khosravi Far, J. R. Bourne, R. inheritance of late Golgi elements in Saccharomyces cerevisiae. J. Cell Biol. Schwaninger, C. J. Der, and W. E. Balch. 1991. Rab1b regulates vesicular 153:47–62. transport between the endoplasmic reticulum and successive Golgi com- 274. Rothman, J. H., C. K. Raymond, T. Gilbert, P. J. O’Hara, and T. H. partments. J. Cell Biol. 115:31–43. Stevens. 1990. A putative GTP binding protein homologous to interferon- 249. Prior, I. A., C. Muncke, R. G. Parton, and J. F. Hancock. 2003. Direct inducible Mx proteins performs an essential function in yeast protein sort- visualization of Ras proteins in spatially distinct cell surface microdomains. ing. Cell 61:1063–1074. J. Cell Biol. 160:165–170. 275. Roumanie, O., M. F. Peypouquet, M. Bonneu, D. Thoraval, F. Doignon, and 250. Proszynski, T. J., R. W. Klemm, M. Gravert, P. P. Hsu, Y. Gloor, J. Wagner, M. Crouzet. 2000. Evidence for the genetic interaction between the actin- K. Kozak, H. Grabner, K. Walzer, M. Bagnat, K. Simons, and C. Walch- binding protein Vrp1 and the RhoGAP Rgd1 mediated through Rho3p and Solimena. 2005. A genome-wide visual screen reveals a role for sphingo- Rho4p in Saccharomyces cerevisiae. Mol. Microbiol. 36:1403–1414. lipids and ergosterol in cell surface delivery in yeast. Proc. Natl. Acad. Sci. 276. Routhier, E. L., T. C. Burn, I. Abbaszade, M. Summers, C. F. Albright, and USA 102:17981–17986. G. C. Prendergast. 2001. Human BIN3 complements the F-actin localiza- 251. Protopopov, V., B. Govindan, P. Novick, and J. E. Gerst. 1993. Homologs of tion defects caused by loss of Hob3p, the fission yeast homolog of Rvs161p. the synaptobrevin/VAMP family of synaptic vesicle proteins function on the J. Biol. Chem. 276:21670–21677. late secretory pathway in S. cerevisiae. Cell 74:855–861. 277. Routhier, E. L., P. S. Donover, and G. C. Prendergast. 2003. hob1ϩ, the 252. Provance, D. W., and J. A. Mercer. 1999. Myosin-V: head to tail. Cell. Mol. fission yeast homolog of Bin1, is dispensable for endocytosis or actin orga- Life Sci. 56:233–242. nization, but required for the response to starvation or genotoxic stress. 253. Pruyne, D. W., D. H. Schott, and A. Bretscher. 1998. Tropomyosin-contain- Oncogene 22:637–648. ing actin cables direct the Myo2p-dependent polarized delivery of secretory 278. Sakamuro, D., K. J. Elliott, R. Wechsler Reya, and G. C. Prendergast. 1996. vesicles in budding yeast. J. Cell Biol. 143:1931–1945. BIN1 is a novel MYC-interacting protein with features of a tumour sup- 254. Ramjaun, A. R., A. Angers, V. Legendre Guillemin, X. K. Tong, and P. S. pressor. Nat. Gen.. 14:69–77. McPherson. 2001. Endophilin regulates JNK activation through its inter- 279. Salazar, M. A., A. V. Kwiatkowski, L. Pellegrini, G. Cestra, M. H. Butler, 118 REN ET AL. MICROBIOL.MOL.BIOL.REV.

K. L. Rossman, D. M. Serna, J. Sondek, F. B. Gertler, and P. De Camilli. binding studies of the EEN gene family, new interacting partners for dy- 2003. Tuba, a novel protein containing bin/amphiphysin/Rvs and Dbl ho- namin, synaptojanin and huntingtin proteins. Biochem. J. 348 Pt. 2:447–458. mology domains, links dynamin to regulation of the actin cytoskeleton. 305. Soubeyran, P., K. Kowanetz, I. Szymkiewicz, W. Y. Langdon, and I. Dikic. J. Biol. Chem. 278:49031–49043. 2002. Cbl-CIN85-endophilin complex mediates ligand-induced downregu- 280. Schafer, D. A. 2002. Coupling actin dynamics and membrane dynamics lation of EGF receptors. Nature 416:183–187. during endocytosis. Curr. Opin. Cell Biol. 14:76–81. 306. Sparks, A. B., N. G. Hoffman, S. J. McConnell, D. M. Fowlkes, and B. K. 281. Schafer, D. A., S. A. Weed, D. Binns, A. V. Karginov, J. T. Parsons, and J. A. Kay. 1996. Cloning of ligand targets: systematic isolation of SH3 domain- Cooper. 2002. Dynamin2 and cortactin regulate actin assembly and filament containing proteins. Nat. Biotechnol. 14:741–744. organization. Curr. Biol. 12:1852–1857. 307. Stamenova, S. D., R. Dunn, A. S. Adler, and L. Hicke. 2004. The Rsp5 282. Schmelzle, T., T. Beck, D. E. Martin, and M. N. Hall. 2004. Activation of ubiquitin ligase binds to and ubiquitinates members of the yeast CIN85-

the RAS/cyclic AMP pathway suppresses a TOR deficiency in yeast. Mol. endophilin complex, Sla1-Rvs167. J. Biol. Chem. 279:16017–16025. Downloaded from Cell. Biol. 24:338–351. 308. Stang, E., F. D. Blystad, M. Kazazic, V. Bertelsen, T. Brodahl, C. Raiborg, 283. Schmid, S. L., and E. Smythe. 1991. Stage-specific assays for coated pit H. Stenmark, and I. H. Madshus. 2004. Cbl-dependent ubiquitination is formation and coated vesicle budding in vitro. J. Cell Biol. 114:869–880. required for progression of EGF receptors into clathrin-coated pits. Mol. 284. Schmidt, A., M. Wolde, C. Thiele, W. Fest, H. Kratzin, A. V. Podtelejnikov, Biol. Cell 15:3591–3604. W. Witke, W. B. Huttner, and H. D. Soling. 1999. Endophilin I mediates 309. Strom, M., P. Vollmer, T. J. Tan, and D. Gallwitz. 1993. A yeast GTPase- synaptic vesicle formation by transfer of arachidonate to lysophosphatidic activating protein that interacts specifically with a member of the Ypt/Rab acid. Nature 401:133–141. family. Nature 361:736–739. 285. Schneiter, R., V. Tatzer, G. Gogg, E. Leitner, and S. D. Kohlwein. 2000. 310. Strous, G. J., P. van Kerkhof, R. Govers, A. Ciechanover, and A. L. Schwartz. 1996. The ubiquitin conjugation system is required for ligand- Elo1p-dependent carboxy-terminal elongation of C14:1⌬(9) to C16:1⌬(11) fatty acids in Saccharomyces cerevisiae. J. Bacteriol 182:3655–3660. induced endocytosis and degradation of the growth hormone receptor. EMBO J. 15:3806–3812. 286. Schott, D., J. Ho, D. Pruyne, and A. Bretscher. 1999. The COOH-terminal http://mmbr.asm.org/ domain of Myo2p, a yeast myosin V, has a direct role in secretory vesicle 311. Sutterlin, C., T. L. Doering, F. Schimmoller, S. Schroder, and H. Riezman. targeting. J. Cell Biol. 147:791–808. 1997. Specific requirements for the ER to Golgi transport of GPI-anchored 287. Segev, N. 2001. Ypt and Rab GTPases: insight into functions through novel proteins in yeast. J. Cell Sci. 110:2703–2714. interactions. Curr. Opin. Cell Biol. 13:500–511. 312. Sweitzer, S. M., and J. E. Hinshaw. 1998. Dynamin undergoes a GTP- 288. Segev, N., and D. Botstein. 1987. The ras-like yeast YPT1 gene is itself dependent conformational change causing vesiculation. Cell 93:1021–1029. essential for growth, sporulation, and starvation response. Mol. Cell. Biol. 313. Szymkiewicz, I., K. Kowanetz, P. Soubeyran, A. Dinarina, S. Lipkowitz, and 7:2367–2377. I. Dikic. 2002. CIN85 participates in Cbl-b-mediated down-regulation of 289. Sesaki, H., and R. E. Jensen. 1999. Division versus fusion: Dnm1p and receptor tyrosine kinases. J. Biol. Chem. 277:39666–39672. Fzo1p antagonistically regulate mitochondrial shape. J. Cell Biol. 147:699– 314. Takei, K., V. Haucke, V. Slepnev, K. Farsad, M. Salazar, H. Chen, and P. 706. De Camilli. 1998. Generation of coated intermediates of clathrin-mediated endocytosis on protein-free liposomes. Cell 94:131–141. 290. Sever, S., H. Damke, and S. L. Schmid. 2000. Garrotes, springs, ratchets, 315. Takei, K., P. S. McPherson, S. L. Schmid, and P. De Camilli. 1995. Tubular and whips: putting dynamin models to the test. Traffic 1:385–392. membrane invaginations coated by dynamin rings are induced by GTP- on October 26, 2015 by University of Queensland Library 291. Shiao, Y. J., G. Lupo, and J. E. Vance. 1995. Evidence that phosphatidyl- gamma S in nerve terminals. Nature 374:186–190. serine is imported into mitochondria via a mitochondria-associated mem- 316. Takei, K., V. I. Slepnev, V. Haucke, and P. De Camilli. 1999. Functional brane and that the majority of mitochondrial phosphatidylethanolamine is partnership between amphiphysin and dynamin in clathrin-mediated endo- derived from decarboxylation of phosphatidylserine. J. Biol. Chem. 270: cytosis. Nat. Cell Biol. 1:33–39. 11190–11198. 317. Talarek, N., A. Balguerie, M. Aigle, and P. Durrens. 2005. A novel link 292. Shpetner, H. S., J. S. Herskovits, and R. B. Vallee. 1996. A binding site for between a Rab GTPase and Rvs proteins: the yeast amphiphysin homo- SH3 domains targets dynamin to coated pits. J. Biol. Chem. 271:13–16. logues. Cell Biochemistry and Function. 23:253–266. Shupliakov, O., P. Low, D. Grabs, H. Gad, H. Chen, C. David, K. Takei, P. 293. 318. Tan, P. K., N. G. Davis, G. F. Sprague, and G. S. Payne. 1993. Clathrin De Camilli, and L. Brodin. 1997. Synaptic vesicle endocytosis impaired by facilitates the internalization of seven transmembrane segment receptors 276: disruption of dynamin-SH3 domain interactions. Science 259–263. for mating pheromones in yeast. J. Cell Biol. 123:1707–1716. 294. Silva, A. J., T. W. Rosahl, P. F. Chapman, Z. Marowitz, E. Friedman, P. W. 319. Tan, T. C., V. A. Valova, C. S. Malladi, M. E. Graham, L. A. Berven, O. J. Frankland, V. Cestari, D. Cioffi, T. C. Sudhof, and R. Bourtchuladze. 1996. Jupp, G. Hansra, S. J. McClure, B. Sarcevic, R. A. Boadle, M. R. Larsen, Impaired learning in mice with abnormal short-lived plasticity. Curr. Biol. M. A. Cousin, and P. J. Robinson. 2003. Cdk5 is essential for synaptic 6:1509–1518. vesicle endocytosis. Nat. Cell Biol. 5:701–710. 295. Simpson, F., N. K. Hussain, B. Qualmann, R. B. Kelly, B. K. Kay, P. S. 320. Tang, D., J. Yeung, K. Y. Lee, M. Matsushita, H. Matsui, K. Tomizawa, O. McPherson, and S. L. Schmid. 1999. SH3-domain-containing proteins func- Hatase, and J. H. Wang. 1995. An isoform of the neuronal cyclin-dependent tion at distinct steps in clathrin-coated vesicle formation. Nat. Cell Biol. kinase 5 (Cdk5) activator. J. Biol. Chem. 270:26897–26903. 1:119–124. 321. Tang, H. Y., and M. Cai. 1996. The EH-domain-containing protein Pan1 is 296. Singer Kruger, B., and S. Ferro Novick. 1997. Use of a synthetic lethal required for normal organization of the actin cytoskeleton in Saccharomy- screen to identify yeast mutants impaired in endocytosis, vacuolar protein ces cerevisiae. Mol. Cell. Biol. 16:4897–4914. sorting and the organization of the cytoskeleton. Eur. J. Cell Biol. 74:365– 322. Tang, H. Y., J. Xu, and M. Cai. 2000. Pan1p, End3p, and S1a1p, three yeast 375. proteins required for normal cortical actin cytoskeleton organization, asso- 297. Singer Kruger, B., H. Stenmark, A. Dusterhoft, P. Philippsen, J. S. Yoo, D. ciate with each other and play essential roles in cell wall morphogenesis. Gallwitz, and M. Zerial. 1994. Role of three rab5-like GTPases, Ypt51p, Mol. Cell. Biol. 20:12–25. Ypt52p, and Ypt53p, in the endocytic and vacuolar protein sorting path- 323. Tang, Y., L. A. Hu, W. E. Miller, N. Ringstad, R. A. Hall, J. A. Pitcher, P. ways of yeast. J. Cell Biol. 125:283–298. DeCamilli, and R. J. Lefkowitz. 1999. Identification of the endophilins 298. Sivadon, P., F. Bauer, M. Aigle, and M. Crouzet. 1995. Actin cytoskeleton (SH3p4/p8/p13) as novel binding partners for the beta1-adrenergic recep- and budding pattern are altered in the yeast rvs161 mutant: the Rvs161 tor. Proc. Natl. Acad. Sci. USA 96:12559–12564. protein shares common domains with the brain protein amphiphysin. Mol. 324. Tarricone, C., B. Xiao, N. Justin, P. A. Walker, K. Rittinger, S. J. Gamblin, Gen. Genet. 246:485–495. and S. J. Smerdon. 2001. The structural basis of Arfaptin-mediated cross- 299. Sivadon, P., M. Crouzet, and M. Aigle. 1997. Functional assessment of the talk between Rac and Arf signalling pathways. Nature 411:215–219. yeast Rvs161 and Rvs167 protein domains. FEBS Lett. 417:21–27. 325. Tatchell, K., L. C. Robinson, and M. Breitenbach. 1985. RAS2 of Saccha- 300. Sivadon, P., M. F. Peypouquet, F. Doignon, M. Aigle, and M. Crouzet. 1997. romyces cerevisiae is required for gluconeogenic growth and proper re- Cloning of the multicopy suppressor gene SUR7: evidence for a functional sponse to nutrient limitation. Proc. Natl. Acad. Sci. USA 82:3785–3789. relationship between the yeast actin-binding protein Rvs167 and a putative 325a.Teasdale, R. D., D. Loci, F. Houghton, L. Karlsson, and P. A. Gleeson. 2001. membranous protein. Yeast 13:747–761. A large family of endosome-localized proteins related to sorting nexin 1. 301. Slepnev, V. I., G. C. Ochoa, M. H. Butler, and P. De Camilli. 2000. Tandem Biochem. J. 358:7–16. arrangement of the clathrin and AP-2 binding domains in amphiphysin 1 326. Terada, Y., K. Tsutsui, K. Sano, O. Hosoya, H. Ohtsuki, and A. Tokunaga. and disruption of clathrin coat function by amphiphysin fragments com- 2002. Novel splice variants of amphiphysin I are expressed in retina. FEBS prising these sites. J. Biol. Chem. 275:17583–17589. Lett. 519:185–190. 302. Slepnev, V. I., G. C. Ochoa, M. H. Butler, D. Grabs, and P. De Camilli. 327. Terrell, J., S. Shih, R. Dunn, and L. Hicke. 1998. A function for monou- 1998. Role of phosphorylation in regulation of the assembly of endocytic biquitination in the internalization of a G protein-coupled receptor. Mol. coat complexes. Science 281:821–824. Cell 1:193–202. 303. Smith, D. S., and L. H. Tsai. 2002. Cdk5 behind the wheel: a role in 328. Thanabalu, T., and A. L. Munn. 2001. Functions of Vrp1p in cytokinesis trafficking and transport? Trends Cell Biol. 12:28–36. and actin patches are distinct and neither requires a WH2/V domain. 304. So, C. W., M. H. Sham, S. L. Chew, N. Cheung, C. K. So, S. K. Chung, C. EMBO J. 20:6979–6989. Caldas, L. M. Wiedemann, and L. C. Chan. 2000. Expression and protein- 329. Toda, T., I. Uno, T. Ishikawa, S. Powers, T. Kataoka, D. Broek, S. Cameron, VOL. 70, 2006 BAR DOMAIN PROTEINS 119

J. Broach, K. Matsumoto, and M. Wigler. 1985. In yeast, RAS proteins are 351. Wang, C. W., and D. J. Klionsky. 2003. The molecular mechanism of controlling elements of adenylate cyclase. Cell 40:27–36. autophagy. Mol. Med. 9:65–76. 330. Tomizawa, K., S. Sunada, Y. F. Lu, Y. Oda, M. Kinuta, T. Ohshima, T. 352. Wang, L. H., T. C. Sudhof, and R. G. W. Anderson. 1995. The appendage Saito, F. Y. Wei, M. Matsushita, S. T. Li, K. Tsutsui, S. Hisanaga, K. domain of alpha-adaptin is a high-affinity binding-site for dynamin. J. Biol. Mikoshiba, K. Takei, and H. Matsui. 2003. Cophosphorylation of am- Chem. 270:10079–10083. phiphysin I and dynamin I by Cdk5 regulates clathrin-mediated endocytosis 353. Warnock, D. E., J. E. Hinshaw, and S. L. Schmid. 1996. Dynamin self- of synaptic vesicles. J. Cell Biol. 163:813–824. assembly stimulates its GTPase activity. J. Biol. Chem. 271:22310–22314. 331. Tong, A. H., B. Drees, G. Nardelli, G. D. Bader, B. Brannetti, L. Castagnoli, 354. Watts, F. Z., G. Shiels, and E. Orr. 1987. The yeast MYO1 gene encoding M. Evangelista, S. Ferracuti, B. Nelson, S. Paoluzi, M. Quondam, A. a myosin-like protein required for cell division. EMBO J. 6:3499–3505. Zucconi, C. W. Hogue, S. Fields, C. Boone, and G. Cesareni. 2002. A 355. Wechsler-Reya, R., K. Elliott, M. Herlyn, and G. C. Prendergast. 1997. The

combined experimental and computational strategy to define protein inter- putative tumor suppressor BIN1 is a short-lived nuclear phosphoprotein, Downloaded from action networks for peptide recognition modules. Science 295:321–324. the localization of which is altered in malignant cells. Cancer Res. 57:3258– 332. Tong, A. H., G. Lesage, G. D. Bader, H. Ding, H. Xu, X. Xin, J. Young, G. F. 3263. Berriz, R. L. Brost, M. Chang, Y. Chen, X. Cheng, G. Chua, H. Friesen, 356. Wechsler-Reya, R., D. Sakamuro, J. Zhang, J. Duhadaway, and G. C. D. S. Goldberg, J. Haynes, C. Humphries, G. He, S. Hussein, L. Ke, N. Prendergast. 1997. Structural analysis of the human BIN1 gene. Evidence Krogan, Z. Li, J. N. Levinson, H. Lu, P. Menard, C. Munyana, A. B. for tissue-specific transcriptional regulation and alternate RNA splicing. Parsons, O. Ryan, R. Tonikian, T. Roberts, A. M. Sdicu, J. Shapiro, B. J. Biol. Chem. 272:31453–31458. Sheikh, B. Suter, S. L. Wong, L. V. Zhang, H. Zhu, C. G. Burd, S. Munro, 357. Wechsler-Reya, R. J., K. J. Elliott, and G. C. Prendergast. 1998. A role for C. Sander, J. Rine, J. Greenblatt, M. Peter, A. Bretscher, G. Bell, F. P. the putative tumor suppressor Bin1 in muscle cell differentiation. Mol. Cell. Roth, G. W. Brown, B. Andrews, H. Bussey, and C. Boone. 2004. Global Biol. 18:566–575. mapping of the yeast genetic interaction network. Science 303:808–813. 358. Weiss, A., and L. A. Leinwand. 1996. The mammalian myosin heavy chain 333. Torre, E., M. A. McNiven, and R. Urrutia. 1994. Dynamin 1 antisense gene family. Annu. Rev. Cell Dev. Biol. 12:417–439. http://mmbr.asm.org/ oligonucleotide treatment prevents neurite formation in cultured hip- 359. Weissenhorn, W. 2005. Crystal structure of the endophilin-A1 BAR do- pocampal neurons. J. Biol. Chem. 269:32411–32417. main. J. Mol. Biol. 351:653–661. 334. Traub, L. M., M. A. Downs, J. L. Westrich, and D. H. Fremont. 1999. 360. Wen, K. K., and P. A. Rubenstein. 2005. Acceleration of yeast actin poly- Crystal structure of the alpha appendage of AP-2 reveals a recruitment merization by yeast Arp2/3 complex does not require an Arp2/3-activating platform for clathrin-coat assembly. Proc. Natl. Acad. Sci. USA 96:8907– protein. J. Biol. Chem. 280:24168–24174. 8912. 361. Wendland, B., and S. D. Emr. 1998. Pan1p, yeast eps15, functions as a 335. Trueheart, J., J. D. Boeke, and G. R. Fink. 1987. Two genes required for cell multivalent adaptor that coordinates protein-protein interactions essential fusion during yeast conjugation: evidence for a pheromone-induced surface for endocytosis. J. Cell Biol. 141:71–84. protein. Mol. Cell. Biol. 7:2316–2328. 362. Wendland, B., K. E. Steece, and S. D. Emr. 1999. Yeast epsins contain an 336. Trueheart, J., and G. R. Fink. 1989. The yeast cell fusion protein FUS1 is essential N-terminal ENTH domain, bind clathrin and are required for O-glycosylated and spans the plasma membrane. Proc. Natl. Acad. Sci. endocytosis. EMBO J. 18:4383–4393. USA 86:9916–9920. 363. Wertman, K. F., D. G. Drubin, and D. Botstein. 1992. Systematic muta-

337. Tsai, L. H., I. Delalle, V. S. Caviness, Jr., T. Chae, and E. Harlow. 1994. p35 tional analysis of the yeast ACT1 gene. Genetics 132:337–350. on October 26, 2015 by University of Queensland Library is a neural-specific regulatory subunit of cyclin-dependent kinase 5. Nature 364. Wesp, A., L. Hicke, J. Palecek, R. Lombardi, T. Aust, A. L. Munn, and H. 371:419–423. Riezman. 1997. End4p/Sla2p interacts with actin-associated proteins for 338. Tsutsui, K., Y. Maeda, S. Seki, and A. Tokunaga. 1997. cDNA cloning of a endocytosis in Saccharomyces cerevisiae. Mol. Biol. Cell 8:2291–2306. novel amphiphysin isoform and tissue-specific expression of its multiple 365. White, J. M., and M. D. Rose. 2001. Yeast mating: getting close to mem- splice variants. Biochem. Biophys. Res. Commun. 236:178–183. brane merger. Curr. Biol. 11:R16–20. 339. Uetz, P., L. Giot, G. Cagney, T. A. Mansfield, R. S. Judson, J. R. Knight, D. 366. Wicky, S., S. Frischmuth, and B. Singer Kruger. 2003. Bsp1p/Ypr171p is an Lockshon, V. Narayan, M. Srinivasan, P. Pochart, A. Qureshi Emili, Y. Li, adapter that directly links some synaptojanin family members to the cortical B. Godwin, D. Conover, T. Kalbfleisch, G. Vijayadamodar, M. Yang, M. actin cytoskeleton in yeast. FEBS Lett. 537:35–41. Johnston, S. Fields, and J. M. Rothberg. 2000. A comprehensive analysis of 367. Wigge, P., K. Kohler, Y. Vallis, C. A. Doyle, D. Owen, S. P. Hunt, and H. T. protein-protein interactions in Saccharomyces cerevisiae. Nature 403:623– McMahon. 1997. Amphiphysin heterodimers: potential role in clathrin- 627. mediated endocytosis. Mol. Biol. Cell 8:2003–2015. 340. Vaduva, G., N. C. Martin, and A. K. Hopper. 1997. Actin-binding verprolin 368. Wigge, P., and H. T. McMahon. 1998. The amphiphysin family of proteins is a polarity development protein required for the morphogenesis and and their role in endocytosis at the synapse. Trends Neurosci. 21:339–344. function of the yeast actin cytoskeleton. J. Cell Biol. 139:1821–1833. 369. Wigge, P., Y. Vallis, and H. T. McMahon. 1997. Inhibition of receptor-medi- 341. Van Aelst, L., T. Joneson, and D. Bar Sagi. 1996. Identification of a novel ated endocytosis by the amphiphysin SH3 domain. Curr. Biol. 7:554–560. Rac1-interacting protein involved in membrane ruffling. EMBO J. 15:3778– 370. Wills, C. 1990. Regulation of sugar and ethanol metabolism in Saccharo- 3786. myces cerevisiae. Crit. Rev. Biochem. Mol. Biol. 25:245–280. 342. Vance, J. E. 2003. Molecular and cellular biology of phosphatidylserine and 371. Winsor, B., and E. Schiebel. 1997. Review: an overview of the Saccharo- phosphatidylethanolamine metabolism. Progress Nucleic Acid Res. Mol. myces cerevisiae microtubule and microfilament cytoskeleton. Yeast 13: Biol. 75:69–111. 399–434. 343. Vance, J. E. 1990. Phospholipid synthesis in a membrane fraction associated 372. Winter, D., T. Lechler, and R. Li. 1999. Activation of the yeast Arp2/3 with mitochondria. J. Biol. Chem. 265:7248–7256. complex by Bee1p, a WASP-family protein. Curr. Biol. 9:501–504. 344. van der Bliek, A. M., T. E. Redelmeier, H. Damke, E. J. Tisdale, E. M. 373. Wong, E. D., J. A. Wagner, S. V. Scott, V. Okreglak, T. J. Holewinske, A. Meyerowitz, and S. L. Schmid. 1993. Mutations in human dynamin block an Cassidy Stone, and J. Nunnari. 2003. The intramitochondrial dynamin- intermediate stage in coated vesicle formation. J. Cell Biol. 122:553–563. related GTPase, Mgm1p, is a component of a protein complex that medi- 345. van Kerkhof, P., M. Sachse, J. Klumperman, and G. J. Strous. 2001. ates mitochondrial fusion. J. Cell Biol. 160:303–311. Growth hormone receptor ubiquitination coincides with recruitment to 374. Yamagishi, A., M. Masuda, T. Ohki, H. Onishi, and N. Mochizuki. 2004. A clathrin-coated membrane domains. J. Biol. Chem. 276:3778–3784. novel actin bundling/filopodium-forming domain conserved in insulin re- 346. Verna, J., A. Lodder, K. Lee, A. Vagts, and R. Ballester. 1997. A family of ceptor tyrosine kinase substrate p53 and missing in metastasis protein. genes required for maintenance of cell wall integrity and for the stress J. Biol. Chem. 279:14929–14936. response in Saccharomyces cerevisiae. Proc. Natl. Acad. Sci. USA 94:13804– 375. Yang, S., K. R. Ayscough, and D. G. Drubin. 1997. A role for the actin 13809. cytoskeleton of Saccharomyces cerevisiae in bipolar bud-site selection. 347. Verstreken, P., O. Kjaerulff, T. E. Lloyd, R. Atkinson, Y. Zhou, I. A. J. Cell Biol. 136:111–123. Meinertzhagen, and H. J. Bellen. 2002. Endophilin mutations block clath- 376. Yeo, S. C., L. Xu, J. Ren, V. J. Boulton, M. D. Wagle, C. Liu, G. Ren, P. rin-mediated endocytosis but not neurotransmitter release. Cell 109:101– Wong, R. Zahn, P. Sasajala, H. Yang, R. C. Piper, and A. L. Munn. 2003. 112. Vps20p and Vta1p interact with Vps4p and function in multivesicular body 348. Verstreken, P., T. W. Koh, K. L. Schulze, R. G. Zhai, P. R. Hiesinger, Y. sorting and endosomal transport in Saccharomyces cerevisiae. J. Cell Sci. Zhou, S. Q. Mehta, Y. Cao, J. Roos, and H. J. Bellen. 2003. Synaptojanin is 116:3957–3970. recruited by endophilin to promote synaptic vesicle uncoating. Neuron 377. Yeung, B. G., H. L. Phan, and G. S. Payne. 1999. Adaptor complex-inde- 40:733–748. pendent clathrin function in yeast. Mol. Biol. Cell 10:3643–3659. 349. Volchuk, A., S. Narine, L. J. Foster, D. Grabs, P. De Camilli, and A. Klip. 378. Yoon, Y., K. R. Pitts, S. Dahan, and M. A. McNiven. 1998. A novel dy- 1998. Perturbation of dynamin II with an amphiphysin SH3 domain in- namin-like protein associates with cytoplasmic vesicles and tubules of the creases GLUT4 glucose transporters at the plasma membrane in 3T3-L1 endoplasmic reticulum in mammalian cells. J. Cell Biol. 140:779–793. adipocytes. Dynamin II participates in GLUT4 endocytosis. J. Biol. Chem. 379. Yoon, Y., K. R. Pitts, and M. A. McNiven. 2001. Mammalian dynamin-like 273:8169–8176. protein DLP1 tubulates membranes. Mol. Biol. Cell 12:2894–2905. 350. Vollert, C. S., and P. Uetz. 2004. The phox homology (PX) domain protein 380. Yoshida, Y., M. Kinuta, T. Abe, S. Liang, K. Araki, O. Cremona, G. Di interaction network in yeast. Mol. Cell. Proteomics 3:1053–1064. Paolo, Y. Moriyama, T. Yasuda, P. De Camilli, and K. Takei. 2004. The 120 REN ET AL. MICROBIOL.MOL.BIOL.REV.

stimulatory action of amphiphysin on dynamin function is dependent on 386. Zelhof, A. C., H. Bao, R. W. Hardy, A. Razzaq, B. Zhang, and C. Q. Doe. lipid bilayer curvature. EMBO J. 23:3483–3491. 2001. Drosophila amphiphysin is implicated in protein localization and 381. Young, M. E., T. S. Karpova, B. Brugger, D. M. Moschenross, G. K. Wang, membrane morphogenesis but not in synaptic vesicle endocytosis. Devel- R. Schneiter, F. T. Wieland, and J. A. Cooper. 2002. The Sur7p family defines opment 128:5005–5015. novel cortical domains in Saccharomyces cerevisiae, affects sphingolipid metab- 387. Zeng, G., and M. Cai. 1999. Regulation of the actin cytoskeleton organi- olism, and is involved in sporulation. Mol. Cell. Biol. 22:927–934. zation in yeast by a novel serine/threonine kinase Prk1p. J. Cell Biol. 382. Yu, H., J. K. Chen, S. Feng, D. C. Dalgarno, A. W. Brauer, and S. L. 144:71–82. Schreiber. 1994. Structural basis for the binding of proline-rich peptides to 388. Zhang, Q., H. K. Chieu, C. P. Low, S. Zhang, C. K. Heng, and H. Yang. SH3 domains. Cell 76:933–945. 2003. Schizosaccharomyces pombe cells deficient in triacylglycerols synthe- 383. Yu, X., and M. Cai. 2004. The yeast dynamin-related GTPase Vps1p func- sis undergo apoptosis upon entry into the stationary phase. J. Biol. Chem.

tions in the organization of the actin cytoskeleton via interaction with Sla1p. Downloaded from J. Cell Sci. 117:3839–3853. 278:47145–47155. 384. Zanolari, B., S. Friant, K. Funato, C. Sutterlin, B. J. Stevenson, and H. 389. Zhao, R., M. Davey, Y. C. Hsu, P. Kaplanek, A. Tong, A. B. Parsons, N. Riezman. 2000. Sphingoid base synthesis requirement for endocytosis in Krogan, G. Cagney, D. Mai, J. Greenblatt, C. Boone, A. Emili, and W. A. Saccharomyces cerevisiae. EMBO J. 19:2824–2833. Houry. 2005. Navigating the chaperone network: an integrative map of physical 385. Zarrinpar, A., S. H. Park, and W. A. Lim. 2003. Optimization of specificity and genetic interactions mediated by the hsp90 chaperone. Cell 120:715–727. in a cellular protein interaction network by negative selection. Nature 390. Zimmerberg, J., and S. McLaughlin. 2004. Membrane curvature: how BAR 426:676–680. domains bend bilayers. Curr. Biol. 14:R250–252. http://mmbr.asm.org/ on October 26, 2015 by University of Queensland Library