The BAR (Bin/amphiphysin/Rvs) domain is the most conserved feature in amphiphysins from yeast to human and is also found

1. BAR Domains as Sensors of Membrane Curvature: The Amphiphysin BAR Structure Brian J. Peter,* Helen M. Kent,* Ian G. Mills, Yvonne Vallis, P. Jonathan G. Butler, Philip R. Evans, Harvey T. McMahon • The BAR (Bin/amphiphysin/Rvs) domain is the most conserved feature in amphiphysins from yeast to human and is also found in endophilins and nadrins. • We solved the structure of the Drosophila amphiphysin BAR domain. It is a crescent-shaped dimer that binds preferentially to highly curved negatively charged membranes. • With its N-terminal amphipathic helix and BAR domain (N-BAR), amphiphysin can drive membrane curvature in vitro and in vivo. • The universal and minimal BAR domain is a dimerization, membrane-binding, and curvature-sensing module. • From this, we predict that BAR domains are in many protein families, including sorting nexins, centaurins, and oligophrenins.

2. Structure of a BAR domain • The BAR (Bin/Amphiphysin/Rvs; pfam03114 and SMART00721) domain is the most highly conserved feature of amphiphysins from yeast to humans. It is a curved dimer with positive charge distributed along the membrane-interacting concave face.

3. Structure of the Drosophila amphiphysin BAR domain. • (A) Ribbon representation of the antiparallel homodimer (purple and green monomers). The principal side chains that make up the positive patches on the convex surface are marked. Protein Data Bank (PDB) identification (ID): 1URU. • (B) Same view as in (A), with the surface colored by electrostatic potential (red, –10 kTe–1, blue, +10 kTe–1). • (C) Ribbon diagram viewed along the dyad axis (concave face) perpendicular to (A). • (D) Electrostatic surface viewed as in (C). • (E) Sequence alignment of the BAR domains in amphiphysins and other proteins (53). Black triangles indicate mutations, and ovals indicate dimer contacts in the structures.

4. The amphiphysin BAR domain in association with low-curvature and high-curvature membranes. The BAR domain binds better to the more highly curved membranes because there is more opportunity for electrostatic interactions across the complete membrane-binding surface of the BAR. • BAR domains are banana-shaped lipid-binding domains found in a wide variety of proteins, which bind to membranes through their concave surface (http://www.endocytosis.org/). They are dimers, and given that the dimer interface and the membrane-binding region overlap, membrane binding may stabilize the dimer formation. • The BAR interaction with membranes is largely electrostatic and binds to negatively charged membranes. Thus the domain on its own is a sensor of high positive curvature. Given a high concentration of a curvature sensor it is clearly possible that a sensor will become an inducer. • An additional feature of some BAR domains is the presence of an N-terminal amphipathic helix (an N-BAR domain). This amphipathic helix will lead to membrane bending. Thus it is interesting to find these two curvature modules side by side in many proteins.

5. Amphipathic helices are stretches of -helix, one side of which is polar (charged) and the other hydrophobic. • These helices are frequently unstructured until they insert into membranes, when the helices are predicted to sit flat on the membrane surface with the hydrophobic residues dipping into the hydrophobic phase of the membrane. The result will be a displacement of lipid headgroups and a reorientation of acyl chains, giving an orientation more favourable to higher curvature. • The most important feature of the amphipathic helix for this review is its effect on membrane curvature. Given the asymmetric insertion (see figure) it acts like a wedge inserted into one leaflet of the membrane. All the amphipathic helices we have studied effect membrane curvature given a high local concentration. Thus it makes sense that epsins bind and promote clathrin polymerization, concentrating the curvature into a local membrane area.

6. Curvature sensing by BAR domains. • Brain liposomes of different intrinsic membrane curvatures were prepared and sequentially extruded through polycarbonate membranes with pores of 0.8, 0.4, 0.1, and 0.05 µm (56). (A) Electron micrographs of the liposomes after extrusion. The large vesicle preparations also contained many smaller vesicles; the smaller vesicle preparations were more homogeneous. Scale bar, 300 nm. (B) Liposome sedimentation assays were normalized by setting the amount of binding to the largest liposomes to 100%. The PtdIns(4,5)P2-binding protein Dab2 does not effect curvature and serves as a control for the available surface area of the liposomes. Inset: an example gel showing the liposome sedimentation results for oligophrenin BAR + PH. P, pellet; S, supernatant. (C) Liposome sedimentation assays as in (B) showing the effect of deleting the N-terminal amphipathic helix in amphiphysin1.

7. Amphiphysins are lipid tubulation, dynamin-, synaptojanin-, adaptor • Endophilins are lipid tubulation, synaptojanin- and dynamin-binding proteins previously implicated in clathrin-mediated endocytosis; endophilin B1 is associated with intracellular membranes. (Shuske et al 2003, Verstreken et al 2003, Modregger et al 2003, Farsad et al 2001) • Tuba is a protein which binds dynamin and actin regulatory proteins, and activates Cdc42 (Salazar et al, 2003). • Pacsins interact with dynamin, synaptojanin, and N-WASP and are involved in vesicle trafficking (Modregger et al 2000). • Oligophrenin is a GTPase regulator in which mutations lead to mental retardation. (Fauchereau et al, 2003, Billuart et al 1998) • Centaurins/ACAPs/ASAP1/AGAPs are regulators of cytoskeletal and membrane trafficking events. (Nie et al 2002, Randazzo et al 2000, Jackson et al 2000) • Sorting nexins comprise a large family of proteins involved in vesicle trafficking. (Lundmark and Carlsson 2003, Worby and Dixon, 2002, Phillips et al 2001 • ICA69 is involved in endocrine secretion; it has a very extended BAR domain. (Spitzenburger et al, 2003, Pilon et al, 2000) • PICK1 is a neuronal phosphorylation substrate which binds AMPA receptors and acid-sensing ion channels, and is involved in long-term depression (Perez et al, 2001, Baron et al 2002, Xia et al 2000). • Arfaptins are putative cytosolic targets of the Arf and Rac GTPases. (Tarricone et al, 2001, Shin and Exton 2001) • Often BAR domains are combined with GEF (guanine nucleotide exchange factor) and GAP (GTP hydrolysis–activating protein) domains, which regulate the activity of small GTPases.

8. Coincidence detection • In sorting nexin-1 there is a dimerized BAR domain (blue), which recognizes membrane curvature, and an additional lipid specificity domain (dark green, PX domain). Binding of both domains — that is, coincidence detection of both membrane composition and curvature — is required for recruitment of the protein and for stabilization of membrane curvature. Hexagons represent phosphatidylinositol phosphate headgroups to which PX domains bind.

9. Gross-section of the “zone of contact”: 1AR(red), synapsin 1(green)

10. In striated muscle, the plasma membrane forms tubular invaginations (transverse tubules or T-tubules) that function in depolarization-contraction coupling. Caveolin-3 and amphiphysin were implicated in their biogenesis. • In vitro studies have shown that the BAR domain of amphiphysin binds and evaginates lipid membranes into narrow tubules suggesting that M-amphiphysin 2 may generate membrane curvature in vivo and perhaps contribute to the biogenesis of T-tubules. • Ultrastructural observations have suggested that T-tubules of striated muscle develop from beaded tubular invaginations ofthe plasma membrane that resemble strings of caveolae. • Amphiphysin proteinsfunction as adaptors between the plasma membrane and submembranouscytosolic scaffolds

11. In vivo and in vitro tubulation of lipid membranes by M-amphiphysin 2. •  (A and B) Electron microscopic views of transfected CHO cells expressing either M-amphiphysin 2 (A) or its BAR* domain (B) reveal the presence of narrow tubules continuous with the plasma membrane [inset of (A)]. (C) Electron micrograph demonstrating the massive tubulation of liposomes induced by recombinant M-amphiphysin 2. (D and E) GFP-dynamin 2 is recruited to tubules when coexpressed with untagged full-length M-amphiphysin 2 (D), but not when coexpressed with the BAR* domain, which lacks the SH3 domain (E). In these two fields, M-amphiphysin 2 and BAR* domain were detected by immunofluorescence. Insets of (D) and (E) show endogenous dynamin immunoreactivity in cells transfected with GFP-M-amphiphysin 2 full-length and GFP-BAR*, respectively. (F) Double immunofluorescence for caveolin-1 and amphiphysin of M-amphiphysin 2-transfected CHO cells. Caveolin-1 immunoreactivity (red) is detectable in a punctate pattern along M-amphiphysin 2-positive tubules (green).

12. Amphiphysin 2 and caveolin in C2C12 cells. • (B and C) Immunofluorescence microscopy of differentiated C2C12 myotubes demonstrating localization of endogenous amphiphysin 2 on tubular elements and partial overlap of amphiphysin 2 with DHPR and caveolin-3. The insets of (B) and (C) show that puncta of DHPR and caveolin-3 immunoreactivity are often aligned with amphiphysin 2-positive tubules. The images of (C) were obtained by confocal miscroscopy. • (D) Electron micrograph of differentiated C2C12 myotubes after incubation with ruthenium red demonstrates the presence of deep, tubulovesicular plasma membrane invaginations (arrow). Localization of amphiphysin 2 (E) and caveolin-3 (F) in ultrathin frozen section of differentiated C2C12 myotubes as revealed by single immunogold labeling. (G to I) Samples prepared as in (E) and (F), but double-labeled for M-amphiphysin 2 (small gold) and caveolin-3 (large gold). In (E to I), amphiphysin 2 and caveolin-3 are concentrated on the tubular and vesicular portion, respectively, of the HRP-labeled network.

13. Amphiphysins are lipid tubulation, dynamin-, synaptojanin-, adaptor • Endophilins are lipid tubulation, synaptojanin- and dynamin-binding proteins previously implicated in clathrin-mediated endocytosis; endophilin B1 is associated with intracellular membranes. (Shuske et al 2003, Verstreken et al 2003, Modregger et al 2003, Farsad et al 2001) • Tuba is a protein which binds dynamin and actin regulatory proteins, and activates Cdc42 (Salazar et al, 2003). • Pacsins interact with dynamin, synaptojanin, and N-WASP and are involved in vesicle trafficking (Modregger et al 2000). • Oligophrenin is a GTPase regulator in which mutations lead to mental retardation. (Fauchereau et al, 2003, Billuart et al 1998) • Centaurins/ACAPs/ASAP1/AGAPs are regulators of cytoskeletal and membrane trafficking events. (Nie et al 2002, Randazzo et al 2000, Jackson et al 2000) • Sorting nexins comprise a large family of proteins involved in vesicle trafficking. (Lundmark and Carlsson 2003, Worby and Dixon, 2002, Phillips et al 2001 • ICA69 is involved in endocrine secretion; it has a very extended BAR domain. (Spitzenburger et al, 2003, Pilon et al, 2000) • PICK1 is a neuronal phosphorylation substrate which binds AMPA receptors and acid-sensing ion channels, and is involved in long-term depression (Perez et al, 2001, Baron et al 2002, Xia et al 2000). • Arfaptins are putative cytosolic targets of the Arf and Rac GTPases. (Tarricone et al, 2001, Shin and Exton 2001)

14. PICK1 is a calcium-sensor for NMDA-induced AMPA receptor traffickingJonathan G Hanley and Jeremy M Henley • Regulation of AMPA receptor (AMPAR) trafficking results in changes in receptor number at the postsynaptic membrane, and hence modifications in synaptic strength, which are proposed to underlie learning and memory. NMDA receptor-mediated postsynaptic Ca2+ influx enhances AMPAR internalisation, but the molecular mechanisms that trigger such trafficking are not well understood. We investigated whether AMPAR-associated protein-protein interactions known to regulate receptor surface expression may be directly regulated by Ca2+. PICK1 binds the AMPAR GluR2 subunit and is involved in AMPAR internalisation and LTD. We show that PICK1 is a Ca2+-binding protein, and that PICK1-GluR2 interactions are enhanced by the presence of 15 M Ca2+. Deletion of an N-terminal acidic domain in PICK1 reduces its ability to bind Ca2+, and renders the GluR2-PICK1 interaction insensitive to Ca2+. Overexpression of this Ca2+-insensitive mutant occludes NMDA-induced AMPAR internalisation in hippocampal neurons. This work reveals a novel postsynaptic Ca2+-binding protein that provides a direct mechanistic link between NMDAR-mediated Ca2+ influx and AMPAR endocytosis.

15. PICK1-GluR2 binding is Ca2+-sensitive. • (A) Endogenous PICK1-GluR2 complexes from hippocampal neurons show maximal interaction at 15 M Ca2+. Extracts of rat cultured hippocampal neurons were immunoprecipitated in different [Ca2+]free with rabbit anti-PICK1 antibody and protein A sepharose. After washing beads in the same [Ca2+]free buffers, bound GluR2 was analysed by Western blotting using anti-GluR2 antibody. Graph shows pooled data, • (B) Total levels of PICK1 and GluR2 are stable in the 0-30 M [Ca2+]free range. (Top panel) Immunoprecipitates identical to those analysed in (A), above,were probed for PICK1 with goat anti-PICK1 antibody. (Bottom panel) Extracts of rat hippocampal neurons were prepared as above, and incubated on ice in the presence of different [Ca2+]free, in the absence of antibody. The levels of GluR2 in the extracts were analysed by Western blotting using anti-GluR2 antibody. (C) Endogenous GRIP-GluR2 complexes are not Ca2+-sensitive. (D) GluR2 C-terminus and PICK1 are sufficient for the Ca2+-sensitive interaction. GST-R2C immobilised on glutathione beads was incubated with his6PICK1flag in the presence of different [Ca2+]free. After washing in the same [Ca2+]free buffer, bound PICK1 was analysed by Western blotting using anti-flag antibody. (E) Total levels of his6PICK1flag and GST-R2C are stable in the 0-30 M [Ca2+]free range. (F) PICK1-GluR2 binding is not sensitive to [Mg2+].

16. Deletion of the N-terminal acidic region in PICK1 removes Ca2+-sensitivity of GluR2-PICK1 interaction. • To analyse the role of PICK1 as a Ca2+-sensor in AMPAR trafficking, we aimed to isolate a Ca2+-insensitive mutant of PICK1 that could be overexpressed in cells as a dominant mutant. We focused on regions of acidic amino acids as potential regions of PICK1 that may be involved in sensing Ca2+. Deletion of an N-terminal acidic region, D4LDYDIEED12, to make NT PICK1 (Figure 2A) removes the Ca2+-sensitivity of PICK1-GluR2 interaction in pull-down assays (Figure 2B). • (A) WT PICK1 contains a region of acidic amino acids at the N-terminus, from residues 4-12. A mutant was constructed that lacks this region, termed NT PICK1. • (B) NT PICK1 binds GluR2 C-terminus at maximal level in all [Ca2+]free tested. GST-R2C immobilised on glutathione beads was incubated with his6WT PICK1flag or his6NT PICK1flag in the presence of different [Ca2+]free as shown. After washing in the same [Ca2+]free buffer, bound PICK1 was analysed by Western blotting using anti-flag antibody. (Bottom panel) Graph shows pooled data for WT PICK1 (left) and NT PICK1 (right), n=4. • (C) Total levels of his6WT PICK1flag, his6NT PICK1flag and GST-R2C are stable in the 0-30 M [Ca2+]free range. His6PICK1flag and his6NT PICK1flag were incubated in buffer A in the presence of different [Ca2+]free as shown. GST-R2C immobilised on beads was treated in the same way in separate tubes. The levels of PICK1 were analysed by Western blotting using anti-flag and GST-R2C was analysed by Coomassie staining. To investigate whether D4LDYDIEED12 is sufficient for Ca2+ binding, we fused this region to the N-terminus of a protein that does not bind Ca2+. We chose -SNAP, a protein involved in the regulation of the GluR2-PICK1 complex (Hanley et al, 2002). The chimeric protein DLDYDIEED-SNAP did not bind significantly more Ca2+ than  -SNAP (data not shown), demonstrating that although DLDYDIEED is necessary, it is not sufficient for Ca2+ binding.

17. PICK1 is a Ca2+-binding protein. • (A) PICK1 is a Ca2+-binding protein, and Ca2+ binding is reduced in NT PICK1. Equilibrium dialysis using 45Ca2+ was employed to determine the Ca2+ binding to known amounts of his6WT PICK1flag (solid line) and his6NT PICK1flag (dashed line) in a range of [Ca2+]free buffers. Graph shows nanomoles of Ca2+ bound per milligram of protein at each [Ca2+]free. Pooled data, n=4. Diagrams demonstrate domain mutations made in PICK1 protein. • (B) Alanine substitutions in the N-terminal acidic region of PICK1 reduce Ca2+-binding capacity. His6PICK1 mutants as shown were subjected to equilibrium dialysis in buffer containing 15 M [Ca2+]free. Graph shows pooled data relative to Ca2+-binding values for his6WT PICK1, n=4, *P<0.05. Diagrams demonstrate mutations made in PICK1 protein. • (C) A C-terminal acidic domain is also involved in Ca2+ binding to PICK1. His6CT PICK1flag (solid line) and his6NT/CT PICK1flag (dashed line) were subjected to equilibrium dialysis as in (A). Graph shows nanomoles of Ca2+ bound per milligram of protein at each [Ca2+]free. Pooled data, n=4. Diagrams demonstrate domain mutations made in PICK1 protein.

18. PICK1 Ca2+ sensor regulates GluR2 internalisation in heterologous cells. • (A) COS cells coexpressing GluR2myc and PICK1flag exhibit Ca2+-sensitive GluR2 endocytosis. COS cells cotransfected with mycGluR2 and WT PICK1flag were live-labelled with rabbit anti-myc antibody. After 5 min ionomycin (1 M) treatment in media containing different [Ca2+]O as shown, cells were incubated in medium lacking Ca2+ for 10 min. GluR2 endocytosis was then assayed by acid-stripping immunocytochemistry. Internalised mycGluR2 was visualised with Alexa 488 and total PICK1 staining by anti-flag and Alexa 568. Representative confocal images of COS cells are shown for each condition. • (B) Quantification of internalised mycGluR2. Values represent total internalised GluR2 (anti-myc immunoreactivity) normalised for cell area (arbitrary units • (C) Expression levels of GluR2myc and PICK1flag are unaffected by ionomycin and [Ca2+]O. • (D) mycGluR2 expressed in the absence of PICK1 does not exhibit Ca2+-dependent increases in endocytosis. • (E) Ca2+-insensitive NT PICK1 stimulates maximal GluR2 endocytosis even in the absence of Ca2+ and at the same level for all [Ca2+]O tested. COS cells transfected with mycGluR2 and NT PICK1flag were subjected to the same treatment as in (A). • (F) COS cells coexpressing GluR1myc and PICK1flag show minimal GluR1 endocytosis, which is not Ca2+-dependent. Cells cotransfected with mycGluR1 and WT PICK1flag were subjected to the same treatment as in (A). Graph shows quantification of internalised GluR1myc. n=15-20 cells per condition.

19. PICK1 is a Ca2+-sensor for NMDA-induced endocytosis of GluR2-containing AMPARs in hippocampal neurons. • (A) Characterisation of NMDA-induced GluR2 internalisation in dissociated hippocampal neurons. Low-density hippocampal cultures, live-labelled with anti-GluR2 antibody, were exposed to drugs. After acid stripping of remaining surface antibody, endocytosis of GluR2-containing AMPAR was assayed by immunocytochemistry (visualised by Alexa 568 secondary). Transmission images and internalised GluR2, including a magnified dendritic segment, are shown from representative neurons. Graph shows quantification of internalised AMPAR. • MK-801 and APV are antagonists for NMDA • (B) Overexpression of WT PICK1 enhances and NT PICK1 occludes NMDA-induced endocytosis of GluR2 in hippocampal neurons.

20. Ca2+-sensitive PICK1 interactions influence GluR1 trafficking to the same extent as GluR2. • (A) Overexpression of WT PICK1 in hippocampal neurons enhances NMDA-induced removal of surface GluR1. Dissociated hippocampal cultures were infected with either empty -IRES-EGFP or WT PICK1-IRES-EGFP virus and exposed to drugs as in Figure 5E. Western blots were probed with anti-GluR1 antibody. n=5, *P<0.05. • (B) Overexpression of NT PICK1 in hippocampal neurons occludes NMDA-induced removal of surface GluR1. As above, except cultures were infected with WT PICK1-IRES-EGFP or NT PICK1-IRES-EGFP virus. n=5, *P<0.05.

21. A model for the role of PICK1 Ca2+-sensor in NMDAR-dependent AMPAR endocytosis. • (A) Under basal conditions, when NMDARs are inactive, GluR2-containing AMPARs are either weakly bound to PICK1 or anchored by ABP/GRIP, resulting in a low level of AMPAR internalisation, representing constitutive cycling. • (B) Activation of NMDARs localised in close proximity to AMPARs results in a Ca2+ flux that raises the local [Ca2+] to around 15 M and therefore enhances binding of PICK1 to GluR2. • (C) This results in increased AMPAR internalisation and hence a decrease in surface AMPAR.

22. GluR2 binds to GRIP and PICK1. • At basal transmission, an equilibrium may exist between GluR2-ABP/GRIP (anchored) and GluR2-PICK1 (mobile). • An important issue is how this equilibrium is altered to regulate AMPAR trafficking. Phosphorylation of GluR2 C-terminus inhibits ABP/GRIP interactions, but allows PICK1-GluR2 binding (Chung et al, 2000; Hayashi and Huganir, 2004). • NSF disassembles GluR2-PICK1 complexes, and therefore favours AMPAR stabilisation at the plasma membrane (Hanley et al, 2002). • Here, we demonstrate a further level of regulation capable of responding directly and immediately to a Ca2+ signal, in contrast to enzymatic reactions that involve multistep pathways.

23. An important issue for synaptic plasticity is how can a [Ca2+] increase result in either LTD or LTP? • One theory for a mechanism to determine the polarity of change suggests that LTD requires a small Ca2+ transient, and LTP a larger one (Lisman, 1989; Artola and Singer, 1993; Cummings et al, 1996). • We propose that the biphasic Ca2+ dependence of PICK1-GluR2 binding could play an important role in such mechanisms; if 15 M Ca2+ is attained on LTD induction, PICK1 binding to GluR2 would mediate AMPAR endocytosis. At higher [Ca2+] reached during LTP, PICK1 would not bind GluR2 efficiently, and AMPAR internalisation would be limited. Spatial and temporal qualities of Ca2+ transients may also be important. • The source of Ca2+ has been shown to be crucial in determining the polarity of plasticity expressed (Nishiyama et al, 2000; Rose and Konnerth, 2001). A prolonged, modest rise in Ca2+ has been proposed to be an optimal LTD signal, in contrast to a short, high magnitude increase for LTP (Yang et al, 1999). It is possible that a Ca2+-sensing protein will respond to specific patterns of Ca2+ accumulation over time, depending on the affinity and capacity of Ca2+ binding.

24. Phylogenetic tree of proteins belonging to the BAR-domain family. • Six families can be distinguished: • the Arfaptin family, which contains the Arfaptins, Protein kinase C-binding protein 1 (PICK1) and the islet cell antigen Ica69; • The Amphiphysin family, which contains Amphiphysin I and II, as well as their paralogue BRAP1 and the RhoGEF Tuba; • the third family is composed of the GTPase-activating proteins Centaurins, Oligophrenins and the adaptor proteins APPL1 and APPL2; • the Rac-binding proteins Bap2 and Bap2-like seem to form a separate family; • the Sorting nexins (1, 2, 4, 5, 6, 7, 8, 9 and 18), which are the most divergent members of the BAR-domain family; • and finally the Endophilin and Nadrin family, which includes Endophilin I, II and III, Endophilin B, as well as Nadrin and SH3-BP1. • Those depicted in red indicate experimentally determined GTPase-binding, double circles indicate experimentally proven dimerization.

25. Coincidence detection • In sorting nexin-1 there is a dimerized BAR domain (blue), which recognizes membrane curvature, and an additional lipid specificity domain (dark green, PX domain). Binding of both domains — that is, coincidence detection of both membrane composition and curvature — is required for recruitment of the protein and for stabilization of membrane curvature. Hexagons represent phosphatidylinositol phosphate headgroups to which PX domains bind.