---
pmid: '20826345'
title: Structure of the PTEN-like region of auxilin, a detector of clathrin-coated
  vesicle budding.
authors:
- Guan R
- Dai H
- Harrison SC
- Kirchhausen T
journal: Structure
year: '2010'
full_text_available: true
full_text_extraction_method: xml
pmcid: PMC2955424
doi: 10.1016/j.str.2010.06.016
pubmed_publication_types:
- Journal Article
- Research Support, N.I.H., Extramural
- Research Support, Non-U.S. Gov't
publication_type: PRIMARY_RESEARCH
---

# Structure of the PTEN-like region of auxilin, a detector of clathrin-coated vesicle budding.
**Authors:** Guan R, Dai H, Harrison SC, Kirchhausen T
**Journal:** Structure (2010)
**DOI:** [10.1016/j.str.2010.06.016](https://doi.org/10.1016/j.str.2010.06.016)
**PMC:** [PMC2955424](https://www.ncbi.nlm.nih.gov/pmc/articles/PMC2955424/)

## Abstract

1. Structure. 2010 Sep 8;18(9):1191-8. doi: 10.1016/j.str.2010.06.016.

Structure of the PTEN-like region of auxilin, a detector of clathrin-coated
vesicle budding.

Guan R(1), Dai H, Harrison SC, Kirchhausen T.

Author information:
(1)Department of Cell Biology, Harvard Medical School, Boston, MA 02115, USA.

Erratum in
    Structure. 2010 Dec 8;18(12):1688. Han, Dai [corrected to Dai, Han].

Auxilin, a J-domain containing protein, recruits the Hsc70 uncoating ATPase to
newly budded clathrin-coated vesicles. The timing of auxilin arrival determines
that uncoating will commence only after the clathrin lattice has fully assembled
and after membrane fission is complete. Auxilin has a region resembling PTEN, a
PI3P phosphatase. We have determined the crystal structure of this region of
bovine auxilin 1; it indeed resembles PTEN closely. A change in the structure of
the P loop accounts for the lack of phosphatase activity. Inclusion of
phosphatidylinositol phosphates substantially enhances liposome binding by
wild-type auxilin, but not by various mutants bearing changes in loops of the C2
domain. Nearly all these mutations also prevent recruitment of auxilin to newly
budded coated vesicles. We propose a specific geometry for auxilin association
with a membrane bilayer and discuss implications of this model for the mechanism
by which auxilin detects separation of a vesicle from its parent membrane.

Copyright © 2010 Elsevier Ltd. All rights reserved.

DOI: 10.1016/j.str.2010.06.016
PMCID: PMC2955424
PMID: 20826345 [Indexed for MEDLINE]

## Full Text

Clathrin mediated endocytosis carries a large part of the membrane traffic that originates at the cell surface and terminates at membrane-bound internal compartments. Assembly of a clathrin lattice around an invaginating vesicle engulfs membrane-associated components, including receptors and their bound ligands (Brett and Traub, 2006; Kirchhausen, 2009). Once the vesicle has pinched off from the parent membrane, the coat is lost, and its components are recycled (Lee et al., 2006; Massol et al., 2006). The vesicle itself can then fuse with the membrane of a suitable target compartment. The uncoating step is the point at which ATP hydrolysis drives the cycle of clathrin assembly and disassembly in a unique direction (Barouch et al., 1994; Greene and Eisenberg, 1990; Schmid and Rothman, 1985). The uncoating ATPase is Hsc70, a member of the heat-shock protein 70 (Hsp70) family. Like all Hsp70 homologs, Hsc70 has a J-domain containing partner that determines the specificity of its target. In the case of clathrin coats, the relevant J-domain protein is known as auxilin (Ungewickell et al., 1995).

Endocytic clathrin-coated vesicles shed their coats almost immediately after budding from the plasma membrane (Lee et al., 2006; Massol et al., 2006). The timing is determined by the arrival of auxilin, which in turn recruits Hsc70. Recruitment of auxilin only after membrane budding is complete ensures that Hsc70-driven uncoating does not compete with coat assembly (Massol et al., 2006). Mammalian species have two auxilin paralogs: auxilin 1(Ahle and Ungewickell, 1990), which is neuron-specific, and auxilin 2, which is ubiquitous. The organization of the auxilin 1 polypeptide chain is shown in Fig. 1A. The elements essential for binding clathrin coats and recruiting Hsc70 are in the C-terminal half of the molecule. The N-terminal part has an amino-acid sequence that can be aligned with the phosphatase and C2 domains of the phosphoinositide phosphatase, PTEN. Differences at key residues in the catalytic cleft suggest that the PTEN-like region of auxilin will not have phosphatase activity, however, and indeed, no activity can be detected. Auxilin 2, also known as GAK (cyclin G associated kinase (Kanaoka et al., 1997)), has an additional, N-terminal kinase domain.

A fragment of auxilin 1 comprising residues 547-910 supports Hsc70 and ATP-dependent uncoating in vitro as well as does the full-length protein (Holstein et al., 1996). In cells, however, auxilin that lacks the PTEN-like region fails to associate with coated vesicles (Massol et al., 2006). The PTEN-like part of auxilin thus appears to be the detector that reports when the vesicle within the coat has separated from the plasma membrane (Massol et al., 2006). Electron cryomicroscopy (cryoEM) and three-dimensional image reconstruction of clathrin coats with bound auxilin (547-910) show that the clathrin-binding region (residues 550-750, approximately) contacts a surface of the clathrin terminal domain, on the internal face of the clathrin lattice (Fotin et al., 2004). The PTEN-like region would then project inward and presumably contact the lipid vesicle. Because the amino-acid sequence of this region implies that it closely resembles a phosphoinositide phosphatase (Haynie and Ponting, 1996), we have proposed that it is likely to recognize a phosphoinositide headgroup (Massol et al., 2006). We have further suggested that stable recruitment at the concentration of auxilin in a cell requires both the contacts with clathrin seen by cryoEM and a membrane interaction of the PTEN-like domain. If an enzymatic activity that generates the species recognized by that domain is present in the coat, then its product will diffuse rapidly away as long as the vesicle bud remains continuous with the adjacent plasma membrane. As soon as dynamin-driven fission is complete, however, the product will accumulate in the vesicle. One candidate for this postulated enzymatic activity is synaptojanin, a lipid phosphatase that remove the 5-phosphate from PI (4,5) P2 (Perera et al., 2006; Woscholski et al., 1997).

We describe here the structure of the PTEN-like region of auxilin 1, which we have determined to help test the mechanism outlined in the preceding paragraph. The structure is, as expected, nearly identical to that of PTEN (Lee et al., 1999). The “catalytic cleft”, although devoid of enzymatic activity, has essentially the same geometry as in PTEN, and it could therefore accommodate an inositol phosphate group. We show that the fragment we have crystallized, auxilin (40-400), binds liposomes containing PI4P and PI (4,5) P2 and that mutations in the C2 domain, in loops known to contact membranes in C2 domains of other proteins, eliminate this binding. We also tested the effects of the mutations on recruitment of full-length auxilin to coated pits, using previously described live-cell imaging methods and found good concordance of lipid-binding and recruitment. We propose a model for the orientation of the protein on a membrane bilayer and discuss its implications for the mechanism by which auxilin detects separation of a vesicle from its parent membrane.

Results and DiscussionStructure determinationWe expressed a set of nested constructs, spanning the predicted PTEN homology region within the N-terminal half of bovine auxilin 1. We could recover auxilin (1-400) with good yield, but the product showed evidence of proteolytic degradation to a smaller, stable species. Mass spectrometric analysis identified the end product as auxilin (40-400), which we expressed, purified, and crystallized as outlined in Methods. The crystals are in space group P3221 (a= b=137.9, c=51.3), with a single polypeptide chain per asymmetric unit. We recorded diffraction data to a minimum Bragg spacing of 2.18 Å and determined the structure by molecular replacement with a modified PTEN structure as the search model. Table 1 summarizes the data collection and refinement statistics. The final model contains residues 60-75 and 88-399; the first 20 residues and the loop spanning residues 76-87 are disordered (Fig. 2).Molecular structureThe molecule has a roughly crescent shaped outline, about 60x35x35 Å in external dimensions. Just as does PTEN itself (Lee et al., 1999), auxilin (40-399) comprises a phosphatase-like domain (residues 54-216) and a C2 domain (residues 226-363). Residues 363-399 extend back against the phosphatase-like domain, so that the N- and C-termini of the fragment polypeptide chain are at the same end of the crescent. The interface between the two domains contains both hydrogen-bond networks and a hydrophobic core, so that the orientation of one domain with respect to the other is probably fixed. In PTEN, the interface is also a tight one, and mutations at that interface compromise function.The structure of the phosphatase-like domain follows very closely that of the PTEN phosphatase domain (rmsd = 1.60 Å for 130 of the 144 ordered alpha carbons between residues 62 and 217), but with some altered active-site residues. The “P-loop” (residues 160-167; Fig. 3A) differs in conformation from its PTEN counterpart, so that the tip of the loop covers what would be the active-site cysteine at its base. In auxilin, the HCxxGxxR motif, conserved in P-loops of protein tyrosine phosphatases and dual specificity phosphatases as well as in PTEN and its orthologs, lacks the final arginine, and various other residues believed to participate in substrate recognition or in catalysis are also different (e.g., auxilin Ile129, which corresponds to PTEN Asp92). Nonetheless, except for absence of the cavity presumably occupied by the D3 phosphate of PtdIns-3,4,5-P3, the inactive “catalytic cleft” in auxilin has roughly the same shape as the active cleft in PTEN.The C2 domain, a sandwich of two, four-stranded, antiparallel β-sheets, has a so-called topology II C2 structure and lacks the canonical Ca2+ liganding residues in loops 1 and 3 that characterize those C2 domains dependent on Ca2+ for lipid-binding activity (Sutton et al., 1995; Sutton and Sprang, 1998). Its framework of eight β-strands is essentially identical to that of PTEN (rmsd=1.09 Å for 80 residues between 227 and 364), but the loops deviate more substantially, especially loops β1-β2 and β5-β6, on the surface of the domain that probably faces the membrane. The β5-β6 loop projects away from the body of the domain as an unusually long, somewhat twisted, β-hairpin. It contains three positively charged residues, and two exposed hydrophobic side chains project at its tip. In PTEN, there is a disordered loop (β6-β7) on the side of the C2 domain likely to face away from the membrane; this loop, excised when the PTEN structure was determined, is a short, well-ordered turn in auxilin.Membrane bindingWe examined association of auxilin (40-400) with large, multilamellar PC liposomes containing the negatively charged phospholipids, PS, PI, PI4P and PI (4,5) P2 at 5%, 10%, and 15%, with pure PC liposomes as a control. The liposomes were sedimented and examined by SDS-PAGE (Fig. 3B). Only the phosphoinositide-containing liposomes bound the auxilin fragment more strongly than PC alone, with an approximate affinity ranking, PI4P > PI (4,5) P2, in agreement with previous lipid-overlay data (Lee et al., 2006; Massol et al., 2006). We also tested the effects of mutations in loops 1 (β1- β2) and 3 (β5-β6) of the C2 domain, likely from studies of other C2 domains to be important membrane contacts (Bai et al., 2004) (Fig. 1B and 3C). We changed positively charged residues in these loops into negatively charged residues, in various combinations, and we mutated two hydrophobic residues in the β5-β6 loop individually to aspartic acid. The positively charged residues do not correspond to the canonical Ca2+ liganding residues in loops 1 and 3 required for the Ca2+ dependent lipid-binding activity of C2 domains (Sutton et al., 1995; Sutton and Sprang, 1998), as they are absent in the C2 domain of auxilin. We also made two mutants with changes in the “P-loop” of the phosphatase-like domain: His159 and Cys160 to Ala in one construct and Gly164 to Ala in the other. The mutations in the C2 domain all reduced binding to liposomes containing phosphoinositides but had no discernable effect on binding to liposomes that contained PC with PI or PS; the mutations in the P-loop had little or no effect. We conclude that membrane affinity and possibly a contribution to headgroup specificity depend strongly on the C2-domain surface that includes the β1-β2 and β5-β6 loops (Fig. 3A). The corresponding loops on one of the two C2 domains of synaptotagmin likewise appear to interact with a phosphoinositide-containing membrane (Bai et al., 2004). There is evidence that Ca2+ binding induces a reorientation of domain C2B, causing its β1-β2 and β5-β6 loops to insert into the bilayer, but there is no indication that these loops participate in specific interactions with phosphoinositides. In auxilin, we expect that the phosphatase-like domain -- and in particular, the region that corresponds to the PTEN catalytic cleft -- is likely to be the principal specificity determinant. Definitive exploration of that region will require a more exhaustive mutational analysis. We have failed to obtain crystals with a short-chain phosphoinositide, which could guide choice of mutations. The published structure of PTEN does not have a bound substrate: a model for PTEN substrate binding is based on a tartrate from the crystallization medium at the active site of the enzyme (Lee et al., 1999).Recruitment of auxilin to coated pitsWe tested the effects of the mutations just described on recruitment of auxilin to coated pits, using previously described live-cell imaging methods (Massol et al., 2006; Saffarian and Kirchhausen, 2008). In the earlier work, we had shown that auxilin appears in coated pits only after a burst of dynamin has led to release of the vesicle from the parent (plasma) membrane (Massol et al., 2006) (see also (Lee et al., 2006)). In Fig. 4A, we show as kymograph displays, the fluorescence from clathrin (labeled with a light-chain-mCherry chimera) and full length auxilin (labeled with EGFP). The abrupt appearance of auxilin, just before uncoating (loss of the clathrin signal), is evident in Figs. 4A (construct #1, Aux1 wt; wild-type control). Its absence is equally clear in Fig. 4A, which shows the effect of one particular mutation (construct #3, Aux1 R301E/R307E/K311E) and confirms that deletion of the PTEN domain (construct #0, ΔPTEN Aux1) also eliminates recruitment (Lee et al., 2006; Massol et al., 2006). We note that in the case of the mutant auxilin, the clathrin still uncoats, as expression of the mutated, fluorescent auxilin does not eliminate recruitment of endogenous auxilin. Figure 4B compares the number of coated-pit associated auxilin spots in cells expressing wt auxilin and each of the mutants. The concordance of lipid-binding and recruitment is good, except for the R245E mutation (construct #16), which diminishes lipid binding but has no detectable effect on coated-pit association.Model for membrane bindingThe various C2-domain mutations that diminish membrane association in the experiments reported in Fig. 3C are likely to map direct contacts with the lipid bilayer. The changes included charge reversal at several positively charged residues, which we propose will interact with negatively charged lipid headgroups, and introduction of aspartic acid instead of hydrophobic residues at the tip of loop 3. Assuming that the catalytic cleft of the phosphatase-like domain also interacts with a lipid headgroup (presumably from a phosphoinositide), then only the orientation of the C2 and phosphatase-like domains shown in Fig. 3A will allow all the relevant residues to face the membrane. Depending on how far loop 3 of the C2 domain dips into the membrane (for example, whether the hydrophobic side-chains at the tip of the loop penetrate as far as the proximal methylene groups of the fatty-acyl chains), it is possible that an inositol phosphate headgroup could be accommodated in the cleft of the phosphatase-like domain without displacing the phosphoinositide from its normal position within the bilayer. The C-terminus of the entire fragment is positioned so that the polypeptide chain that follows could project away from the membrane and toward the clathrin coat. The 150 amino-acid residues between the end of the PTEN-like region and the N-terminus of the clathrin-binding region could easily span the ~100Å gap between the location of the auxilin J-domain in a clathrin coat and its vesicle membrane. That segment is probably also quite extended in solution, as the Stokes radius of intact auxilin (62 Å) is much larger than expected for a 900 residue globular protein (Ahle and Ungewickell, 1990).Potential role of specific lipid binding in recruiting auxilin to newly budded coated vesiclesThe only reasonably abundant phosphoinositide in the plasma membrane is PI (4,5) P2 (about 0.5% of total lipid). Indeed, interaction with the headgroup of this lipid assists recruitment to the plasma membrane of proteins such as the AP-2 clathrin adaptor (Cremona et al., 1999; Gaidarov and Keen, 1999). As auxilin recruitment occurs only after a vesicle has separated from the bulk membrane (Lee et al., 2006; Massol et al., 2006), PI (4,5) P2 in unlikely to determine specific binding to the pinched-off vesicle. PI-4-P is a better candidate, as it could be produced from PI (4,5) P2 by an enzyme such as synaptojanin or similar lipid phosphatase recruited to the coated pit or to the coated vesicle (Perera et al., 2006). Even if that enzyme were present before membrane fission, the product would diffuse out of the bud and PI {4,5} P2 would maintain its equilibrium concentration. A vesicle 700 Å in diameter (about the size the vesicle within a typical endocytic clathrin coat containing 60-80 triskelions) would contain about 100 PI (4,5) P2 molecules, and even partial conversion to PI4P would recruit sufficient auxilin to drive uncoating (e.g., one auxilin per vertex). The strongest binding in the experiments in Fig. 3 is indeed to liposomes with PI4P, a result that is in agreement with earlier observations using lipid-strips (Lee et al., 2006; Massol et al., 2006). A definitive conclusion will require both a more sensitive and quantitative assay and a structure with a suitable ligand.

Results and Discussion

Structure determinationWe expressed a set of nested constructs, spanning the predicted PTEN homology region within the N-terminal half of bovine auxilin 1. We could recover auxilin (1-400) with good yield, but the product showed evidence of proteolytic degradation to a smaller, stable species. Mass spectrometric analysis identified the end product as auxilin (40-400), which we expressed, purified, and crystallized as outlined in Methods. The crystals are in space group P3221 (a= b=137.9, c=51.3), with a single polypeptide chain per asymmetric unit. We recorded diffraction data to a minimum Bragg spacing of 2.18 Å and determined the structure by molecular replacement with a modified PTEN structure as the search model. Table 1 summarizes the data collection and refinement statistics. The final model contains residues 60-75 and 88-399; the first 20 residues and the loop spanning residues 76-87 are disordered (Fig. 2).

Structure determination

We expressed a set of nested constructs, spanning the predicted PTEN homology region within the N-terminal half of bovine auxilin 1. We could recover auxilin (1-400) with good yield, but the product showed evidence of proteolytic degradation to a smaller, stable species. Mass spectrometric analysis identified the end product as auxilin (40-400), which we expressed, purified, and crystallized as outlined in Methods. The crystals are in space group P3221 (a= b=137.9, c=51.3), with a single polypeptide chain per asymmetric unit. We recorded diffraction data to a minimum Bragg spacing of 2.18 Å and determined the structure by molecular replacement with a modified PTEN structure as the search model. Table 1 summarizes the data collection and refinement statistics. The final model contains residues 60-75 and 88-399; the first 20 residues and the loop spanning residues 76-87 are disordered (Fig. 2).

Molecular structureThe molecule has a roughly crescent shaped outline, about 60x35x35 Å in external dimensions. Just as does PTEN itself (Lee et al., 1999), auxilin (40-399) comprises a phosphatase-like domain (residues 54-216) and a C2 domain (residues 226-363). Residues 363-399 extend back against the phosphatase-like domain, so that the N- and C-termini of the fragment polypeptide chain are at the same end of the crescent. The interface between the two domains contains both hydrogen-bond networks and a hydrophobic core, so that the orientation of one domain with respect to the other is probably fixed. In PTEN, the interface is also a tight one, and mutations at that interface compromise function.The structure of the phosphatase-like domain follows very closely that of the PTEN phosphatase domain (rmsd = 1.60 Å for 130 of the 144 ordered alpha carbons between residues 62 and 217), but with some altered active-site residues. The “P-loop” (residues 160-167; Fig. 3A) differs in conformation from its PTEN counterpart, so that the tip of the loop covers what would be the active-site cysteine at its base. In auxilin, the HCxxGxxR motif, conserved in P-loops of protein tyrosine phosphatases and dual specificity phosphatases as well as in PTEN and its orthologs, lacks the final arginine, and various other residues believed to participate in substrate recognition or in catalysis are also different (e.g., auxilin Ile129, which corresponds to PTEN Asp92). Nonetheless, except for absence of the cavity presumably occupied by the D3 phosphate of PtdIns-3,4,5-P3, the inactive “catalytic cleft” in auxilin has roughly the same shape as the active cleft in PTEN.The C2 domain, a sandwich of two, four-stranded, antiparallel β-sheets, has a so-called topology II C2 structure and lacks the canonical Ca2+ liganding residues in loops 1 and 3 that characterize those C2 domains dependent on Ca2+ for lipid-binding activity (Sutton et al., 1995; Sutton and Sprang, 1998). Its framework of eight β-strands is essentially identical to that of PTEN (rmsd=1.09 Å for 80 residues between 227 and 364), but the loops deviate more substantially, especially loops β1-β2 and β5-β6, on the surface of the domain that probably faces the membrane. The β5-β6 loop projects away from the body of the domain as an unusually long, somewhat twisted, β-hairpin. It contains three positively charged residues, and two exposed hydrophobic side chains project at its tip. In PTEN, there is a disordered loop (β6-β7) on the side of the C2 domain likely to face away from the membrane; this loop, excised when the PTEN structure was determined, is a short, well-ordered turn in auxilin.

Molecular structure

The molecule has a roughly crescent shaped outline, about 60x35x35 Å in external dimensions. Just as does PTEN itself (Lee et al., 1999), auxilin (40-399) comprises a phosphatase-like domain (residues 54-216) and a C2 domain (residues 226-363). Residues 363-399 extend back against the phosphatase-like domain, so that the N- and C-termini of the fragment polypeptide chain are at the same end of the crescent. The interface between the two domains contains both hydrogen-bond networks and a hydrophobic core, so that the orientation of one domain with respect to the other is probably fixed. In PTEN, the interface is also a tight one, and mutations at that interface compromise function.

The structure of the phosphatase-like domain follows very closely that of the PTEN phosphatase domain (rmsd = 1.60 Å for 130 of the 144 ordered alpha carbons between residues 62 and 217), but with some altered active-site residues. The “P-loop” (residues 160-167; Fig. 3A) differs in conformation from its PTEN counterpart, so that the tip of the loop covers what would be the active-site cysteine at its base. In auxilin, the HCxxGxxR motif, conserved in P-loops of protein tyrosine phosphatases and dual specificity phosphatases as well as in PTEN and its orthologs, lacks the final arginine, and various other residues believed to participate in substrate recognition or in catalysis are also different (e.g., auxilin Ile129, which corresponds to PTEN Asp92). Nonetheless, except for absence of the cavity presumably occupied by the D3 phosphate of PtdIns-3,4,5-P3, the inactive “catalytic cleft” in auxilin has roughly the same shape as the active cleft in PTEN.

The C2 domain, a sandwich of two, four-stranded, antiparallel β-sheets, has a so-called topology II C2 structure and lacks the canonical Ca2+ liganding residues in loops 1 and 3 that characterize those C2 domains dependent on Ca2+ for lipid-binding activity (Sutton et al., 1995; Sutton and Sprang, 1998). Its framework of eight β-strands is essentially identical to that of PTEN (rmsd=1.09 Å for 80 residues between 227 and 364), but the loops deviate more substantially, especially loops β1-β2 and β5-β6, on the surface of the domain that probably faces the membrane. The β5-β6 loop projects away from the body of the domain as an unusually long, somewhat twisted, β-hairpin. It contains three positively charged residues, and two exposed hydrophobic side chains project at its tip. In PTEN, there is a disordered loop (β6-β7) on the side of the C2 domain likely to face away from the membrane; this loop, excised when the PTEN structure was determined, is a short, well-ordered turn in auxilin.

Membrane bindingWe examined association of auxilin (40-400) with large, multilamellar PC liposomes containing the negatively charged phospholipids, PS, PI, PI4P and PI (4,5) P2 at 5%, 10%, and 15%, with pure PC liposomes as a control. The liposomes were sedimented and examined by SDS-PAGE (Fig. 3B). Only the phosphoinositide-containing liposomes bound the auxilin fragment more strongly than PC alone, with an approximate affinity ranking, PI4P > PI (4,5) P2, in agreement with previous lipid-overlay data (Lee et al., 2006; Massol et al., 2006). We also tested the effects of mutations in loops 1 (β1- β2) and 3 (β5-β6) of the C2 domain, likely from studies of other C2 domains to be important membrane contacts (Bai et al., 2004) (Fig. 1B and 3C). We changed positively charged residues in these loops into negatively charged residues, in various combinations, and we mutated two hydrophobic residues in the β5-β6 loop individually to aspartic acid. The positively charged residues do not correspond to the canonical Ca2+ liganding residues in loops 1 and 3 required for the Ca2+ dependent lipid-binding activity of C2 domains (Sutton et al., 1995; Sutton and Sprang, 1998), as they are absent in the C2 domain of auxilin. We also made two mutants with changes in the “P-loop” of the phosphatase-like domain: His159 and Cys160 to Ala in one construct and Gly164 to Ala in the other. The mutations in the C2 domain all reduced binding to liposomes containing phosphoinositides but had no discernable effect on binding to liposomes that contained PC with PI or PS; the mutations in the P-loop had little or no effect. We conclude that membrane affinity and possibly a contribution to headgroup specificity depend strongly on the C2-domain surface that includes the β1-β2 and β5-β6 loops (Fig. 3A). The corresponding loops on one of the two C2 domains of synaptotagmin likewise appear to interact with a phosphoinositide-containing membrane (Bai et al., 2004). There is evidence that Ca2+ binding induces a reorientation of domain C2B, causing its β1-β2 and β5-β6 loops to insert into the bilayer, but there is no indication that these loops participate in specific interactions with phosphoinositides. In auxilin, we expect that the phosphatase-like domain -- and in particular, the region that corresponds to the PTEN catalytic cleft -- is likely to be the principal specificity determinant. Definitive exploration of that region will require a more exhaustive mutational analysis. We have failed to obtain crystals with a short-chain phosphoinositide, which could guide choice of mutations. The published structure of PTEN does not have a bound substrate: a model for PTEN substrate binding is based on a tartrate from the crystallization medium at the active site of the enzyme (Lee et al., 1999).

Membrane binding

We examined association of auxilin (40-400) with large, multilamellar PC liposomes containing the negatively charged phospholipids, PS, PI, PI4P and PI (4,5) P2 at 5%, 10%, and 15%, with pure PC liposomes as a control. The liposomes were sedimented and examined by SDS-PAGE (Fig. 3B). Only the phosphoinositide-containing liposomes bound the auxilin fragment more strongly than PC alone, with an approximate affinity ranking, PI4P > PI (4,5) P2, in agreement with previous lipid-overlay data (Lee et al., 2006; Massol et al., 2006). We also tested the effects of mutations in loops 1 (β1- β2) and 3 (β5-β6) of the C2 domain, likely from studies of other C2 domains to be important membrane contacts (Bai et al., 2004) (Fig. 1B and 3C). We changed positively charged residues in these loops into negatively charged residues, in various combinations, and we mutated two hydrophobic residues in the β5-β6 loop individually to aspartic acid. The positively charged residues do not correspond to the canonical Ca2+ liganding residues in loops 1 and 3 required for the Ca2+ dependent lipid-binding activity of C2 domains (Sutton et al., 1995; Sutton and Sprang, 1998), as they are absent in the C2 domain of auxilin. We also made two mutants with changes in the “P-loop” of the phosphatase-like domain: His159 and Cys160 to Ala in one construct and Gly164 to Ala in the other. The mutations in the C2 domain all reduced binding to liposomes containing phosphoinositides but had no discernable effect on binding to liposomes that contained PC with PI or PS; the mutations in the P-loop had little or no effect. We conclude that membrane affinity and possibly a contribution to headgroup specificity depend strongly on the C2-domain surface that includes the β1-β2 and β5-β6 loops (Fig. 3A). The corresponding loops on one of the two C2 domains of synaptotagmin likewise appear to interact with a phosphoinositide-containing membrane (Bai et al., 2004). There is evidence that Ca2+ binding induces a reorientation of domain C2B, causing its β1-β2 and β5-β6 loops to insert into the bilayer, but there is no indication that these loops participate in specific interactions with phosphoinositides. In auxilin, we expect that the phosphatase-like domain -- and in particular, the region that corresponds to the PTEN catalytic cleft -- is likely to be the principal specificity determinant. Definitive exploration of that region will require a more exhaustive mutational analysis. We have failed to obtain crystals with a short-chain phosphoinositide, which could guide choice of mutations. The published structure of PTEN does not have a bound substrate: a model for PTEN substrate binding is based on a tartrate from the crystallization medium at the active site of the enzyme (Lee et al., 1999).

Recruitment of auxilin to coated pitsWe tested the effects of the mutations just described on recruitment of auxilin to coated pits, using previously described live-cell imaging methods (Massol et al., 2006; Saffarian and Kirchhausen, 2008). In the earlier work, we had shown that auxilin appears in coated pits only after a burst of dynamin has led to release of the vesicle from the parent (plasma) membrane (Massol et al., 2006) (see also (Lee et al., 2006)). In Fig. 4A, we show as kymograph displays, the fluorescence from clathrin (labeled with a light-chain-mCherry chimera) and full length auxilin (labeled with EGFP). The abrupt appearance of auxilin, just before uncoating (loss of the clathrin signal), is evident in Figs. 4A (construct #1, Aux1 wt; wild-type control). Its absence is equally clear in Fig. 4A, which shows the effect of one particular mutation (construct #3, Aux1 R301E/R307E/K311E) and confirms that deletion of the PTEN domain (construct #0, ΔPTEN Aux1) also eliminates recruitment (Lee et al., 2006; Massol et al., 2006). We note that in the case of the mutant auxilin, the clathrin still uncoats, as expression of the mutated, fluorescent auxilin does not eliminate recruitment of endogenous auxilin. Figure 4B compares the number of coated-pit associated auxilin spots in cells expressing wt auxilin and each of the mutants. The concordance of lipid-binding and recruitment is good, except for the R245E mutation (construct #16), which diminishes lipid binding but has no detectable effect on coated-pit association.

Recruitment of auxilin to coated pits

We tested the effects of the mutations just described on recruitment of auxilin to coated pits, using previously described live-cell imaging methods (Massol et al., 2006; Saffarian and Kirchhausen, 2008). In the earlier work, we had shown that auxilin appears in coated pits only after a burst of dynamin has led to release of the vesicle from the parent (plasma) membrane (Massol et al., 2006) (see also (Lee et al., 2006)). In Fig. 4A, we show as kymograph displays, the fluorescence from clathrin (labeled with a light-chain-mCherry chimera) and full length auxilin (labeled with EGFP). The abrupt appearance of auxilin, just before uncoating (loss of the clathrin signal), is evident in Figs. 4A (construct #1, Aux1 wt; wild-type control). Its absence is equally clear in Fig. 4A, which shows the effect of one particular mutation (construct #3, Aux1 R301E/R307E/K311E) and confirms that deletion of the PTEN domain (construct #0, ΔPTEN Aux1) also eliminates recruitment (Lee et al., 2006; Massol et al., 2006). We note that in the case of the mutant auxilin, the clathrin still uncoats, as expression of the mutated, fluorescent auxilin does not eliminate recruitment of endogenous auxilin. Figure 4B compares the number of coated-pit associated auxilin spots in cells expressing wt auxilin and each of the mutants. The concordance of lipid-binding and recruitment is good, except for the R245E mutation (construct #16), which diminishes lipid binding but has no detectable effect on coated-pit association.

Model for membrane bindingThe various C2-domain mutations that diminish membrane association in the experiments reported in Fig. 3C are likely to map direct contacts with the lipid bilayer. The changes included charge reversal at several positively charged residues, which we propose will interact with negatively charged lipid headgroups, and introduction of aspartic acid instead of hydrophobic residues at the tip of loop 3. Assuming that the catalytic cleft of the phosphatase-like domain also interacts with a lipid headgroup (presumably from a phosphoinositide), then only the orientation of the C2 and phosphatase-like domains shown in Fig. 3A will allow all the relevant residues to face the membrane. Depending on how far loop 3 of the C2 domain dips into the membrane (for example, whether the hydrophobic side-chains at the tip of the loop penetrate as far as the proximal methylene groups of the fatty-acyl chains), it is possible that an inositol phosphate headgroup could be accommodated in the cleft of the phosphatase-like domain without displacing the phosphoinositide from its normal position within the bilayer. The C-terminus of the entire fragment is positioned so that the polypeptide chain that follows could project away from the membrane and toward the clathrin coat. The 150 amino-acid residues between the end of the PTEN-like region and the N-terminus of the clathrin-binding region could easily span the ~100Å gap between the location of the auxilin J-domain in a clathrin coat and its vesicle membrane. That segment is probably also quite extended in solution, as the Stokes radius of intact auxilin (62 Å) is much larger than expected for a 900 residue globular protein (Ahle and Ungewickell, 1990).

Model for membrane binding

The various C2-domain mutations that diminish membrane association in the experiments reported in Fig. 3C are likely to map direct contacts with the lipid bilayer. The changes included charge reversal at several positively charged residues, which we propose will interact with negatively charged lipid headgroups, and introduction of aspartic acid instead of hydrophobic residues at the tip of loop 3. Assuming that the catalytic cleft of the phosphatase-like domain also interacts with a lipid headgroup (presumably from a phosphoinositide), then only the orientation of the C2 and phosphatase-like domains shown in Fig. 3A will allow all the relevant residues to face the membrane. Depending on how far loop 3 of the C2 domain dips into the membrane (for example, whether the hydrophobic side-chains at the tip of the loop penetrate as far as the proximal methylene groups of the fatty-acyl chains), it is possible that an inositol phosphate headgroup could be accommodated in the cleft of the phosphatase-like domain without displacing the phosphoinositide from its normal position within the bilayer. The C-terminus of the entire fragment is positioned so that the polypeptide chain that follows could project away from the membrane and toward the clathrin coat. The 150 amino-acid residues between the end of the PTEN-like region and the N-terminus of the clathrin-binding region could easily span the ~100Å gap between the location of the auxilin J-domain in a clathrin coat and its vesicle membrane. That segment is probably also quite extended in solution, as the Stokes radius of intact auxilin (62 Å) is much larger than expected for a 900 residue globular protein (Ahle and Ungewickell, 1990).

Potential role of specific lipid binding in recruiting auxilin to newly budded coated vesiclesThe only reasonably abundant phosphoinositide in the plasma membrane is PI (4,5) P2 (about 0.5% of total lipid). Indeed, interaction with the headgroup of this lipid assists recruitment to the plasma membrane of proteins such as the AP-2 clathrin adaptor (Cremona et al., 1999; Gaidarov and Keen, 1999). As auxilin recruitment occurs only after a vesicle has separated from the bulk membrane (Lee et al., 2006; Massol et al., 2006), PI (4,5) P2 in unlikely to determine specific binding to the pinched-off vesicle. PI-4-P is a better candidate, as it could be produced from PI (4,5) P2 by an enzyme such as synaptojanin or similar lipid phosphatase recruited to the coated pit or to the coated vesicle (Perera et al., 2006). Even if that enzyme were present before membrane fission, the product would diffuse out of the bud and PI {4,5} P2 would maintain its equilibrium concentration. A vesicle 700 Å in diameter (about the size the vesicle within a typical endocytic clathrin coat containing 60-80 triskelions) would contain about 100 PI (4,5) P2 molecules, and even partial conversion to PI4P would recruit sufficient auxilin to drive uncoating (e.g., one auxilin per vertex). The strongest binding in the experiments in Fig. 3 is indeed to liposomes with PI4P, a result that is in agreement with earlier observations using lipid-strips (Lee et al., 2006; Massol et al., 2006). A definitive conclusion will require both a more sensitive and quantitative assay and a structure with a suitable ligand.

Potential role of specific lipid binding in recruiting auxilin to newly budded coated vesicles

The only reasonably abundant phosphoinositide in the plasma membrane is PI (4,5) P2 (about 0.5% of total lipid). Indeed, interaction with the headgroup of this lipid assists recruitment to the plasma membrane of proteins such as the AP-2 clathrin adaptor (Cremona et al., 1999; Gaidarov and Keen, 1999). As auxilin recruitment occurs only after a vesicle has separated from the bulk membrane (Lee et al., 2006; Massol et al., 2006), PI (4,5) P2 in unlikely to determine specific binding to the pinched-off vesicle. PI-4-P is a better candidate, as it could be produced from PI (4,5) P2 by an enzyme such as synaptojanin or similar lipid phosphatase recruited to the coated pit or to the coated vesicle (Perera et al., 2006). Even if that enzyme were present before membrane fission, the product would diffuse out of the bud and PI {4,5} P2 would maintain its equilibrium concentration. A vesicle 700 Å in diameter (about the size the vesicle within a typical endocytic clathrin coat containing 60-80 triskelions) would contain about 100 PI (4,5) P2 molecules, and even partial conversion to PI4P would recruit sufficient auxilin to drive uncoating (e.g., one auxilin per vertex). The strongest binding in the experiments in Fig. 3 is indeed to liposomes with PI4P, a result that is in agreement with earlier observations using lipid-strips (Lee et al., 2006; Massol et al., 2006). A definitive conclusion will require both a more sensitive and quantitative assay and a structure with a suitable ligand.