Huntingtin-interacting Protein 1 (Hip1) and Hip1-related Protein (Hip1R) Bind the Conserved Sequence of Clathrin Light Chains and Thereby Influence Clathrin Assembly in Vitro and Actin Distribution in Vivo *

Clathrin heavy and light chains form triskelia, which assemble into polyhedral coats of membrane vesicles that mediate transport for endocytosis and organelle biogenesis. Light chain subunits regulate clathrin assembly in vitro by suppressing spontaneous self-assem-bly of the heavy chains. The residues that play this regulatory role are at the N terminus of a conserved 22-amino acid sequence that is shared by all vertebrate light chains. Here we show that these regulatory residues and others in the conserved sequence mediate light chain interaction with Hip1 and Hip1R. These related proteins were previously found to be enriched in clath-rin-coated vesicles and to promote clathrin assembly in vitro . We demonstrate Hip1R binding preference for light chains associated with clathrin heavy chain and show that Hip1R stimulation of clathrin assembly in vitro is blocked by mutations in the conserved sequence of light chains that abolish interaction with Hip1 and Hip1R. In vivo overexpression of a fragment of clathrin light chain comprising the Hip1R-binding region affected cellular actin distribution. Together these results suggest that the roles of Hip1 and Hip1R in affecting clathrin assembly and actin distribution are mediated by their interaction with the conserved sequence of clathrin light chains.

Clathrin-coated vesicles (CCVs) 1 perform selective transport for receptor-mediated endocytosis and protein sorting during organelle biogenesis. The triskelion-shaped clathrin molecule is composed of trimerized clathrin heavy chains (HCs), each with a bound light chain (LC) subunit. Triskelia assemble into polyhedral lattice-coated cellular membranes. Most known regulatory and adaptor proteins influence coat assembly, membrane association, membrane fission, and uncoating by binding to sites on the HCs (1, 2), which themselves can self-assemble into a lattice. The LC subunits suppress spontaneous lattice assembly in vitro and thus allow cellular clathrin assembly to be controlled and subjected to regulation by additional proteins (3,4). Other roles for the LCs have not yet been clearly defined, despite the fact that LCs appear to be composed of multiple functional domains (5). In this study, we establish that the function of a conserved sequence in vertebrate LCs is to bind to the CCV-enriched proteins, Hip1 and Hip1R (referred to henceforth as Hip1/R). The potential roles for this LC-Hip1/R interaction in the network of CCV regulation are characterized.
In vertebrate clathrin, there are two types of LCs, LCa and LCb, encoded by different genes. These LCs are expressed in all tissues at varying relative levels (6) and are heterogeneously distributed in clathrin triskelia (7). Near the N termini of LCs there is a 22-residue sequence, shared by LCa and LCb and completely conserved in all vertebrate LCs, which is also highly conserved in LCs from non-vertebrate species (5). The first three residues of this conserved sequence are a triplet of negatively charged residues (EED) that regulate clathrin assembly (4). If these residues are neutralized by mutation, the LCs no longer suppress in vitro clathrin assembly (4). Residues to the N-terminal side of the conserved sequence in LCb are the target for casein kinase phosphorylation and residues to the C-terminal side in LCa constitute a site that can be bound by Hsc70 (5), a protein involved in CCV uncoating. Central to both LCs is a helical region that associates with HCs (8). C-terminal to this region are sites for short inserted sequences that are present because of alternate splicing of LC RNA in neurons (5). These sequences influence calmodulin binding to the LC C termini (9). LCs bind to the triskelion hub, a domain formed by the C-terminal third of the HCs through trimerization near their C termini. Thus LCs have the potential to regulate molecular interactions in the central region of the triskelion, whereas the majority of known clathrin-binding proteins interact with HC terminal domains at the ends of the triskelion legs.
Hip1/R proteins were recently shown to bind LCs (10 -13) and were therefore candidates for interacting proteins that might regulate CCV formation and function through the hub domain. Hip1 was identified as a protein that loses binding to mutant huntingtin and has thus been implicated in Huntington's disease (14,15). Hip1 and the related Hip1R protein (47% sequence identity) are highly expressed in neuronal tissue. They have different developmental expression patterns (13) and are also expressed in non-neuronal cells (15). Hip1/R proteins belong to the yeast Sla2/End4 family and all these proteins are implicated in actin binding and membrane traffic (16). They have an N-terminal ANTH domain for phospholipid interaction (17), a central helical domain for dimerization and clathrin interaction, and a C-terminal talin-like actin binding module (10,13). Both Hip1/R are enriched in CCVs and promote clathrin assembly in vitro (10, 13, 18 -20). We had previously shown that transfection of cells with the hub fragment caused both LC and Hip1R dissociation from clathrin, suggesting that Hip1R was binding to LCs associated with the overexpressed hubs (11). Furthermore, hub overexpression disrupted the characteristic offset alignment of clathrin-coated pits with the actin cytoskeleton (11). We consequently hypothesized that the CCV link to actin might be mediated by Hip1R binding to the LC subunits of clathrin. Recent Hip1R RNA silencing experiments showed that decreasing Hip1R levels caused clathrin-coated pits to become more tightly associated with actin instead of being offset from actin filaments (21), further corroborating a role for Hip1R in linking CCVs to actin dynamics.
Here we map Hip1/R binding to the conserved sequence of clathrin LCs. Mutations in this LC sequence abolished interaction with Hip1/R and abolished the ability of Hip1R to promote clathrin assembly in vitro, suggesting that the Hip1R effects on assembly result from binding the regulatory residues of LCs. In vivo overexpression a LC fragment comprising the conserved Hip1/R binding sequence affected the cellular distribution of actin. These results indicate that the conserved sequence of LCs bind Hip1/R and coordinate Hip1/R action on clathrin and the actin cytoskeleton.

EXPERIMENTAL PROCEDURES
Yeast Two-hybrid Assays-LCb fragments were amplified from bovine brain cDNAs (22) and cloned into the BamHI site of pACT2 (Clontech). Hip1-(124 -1037) was amplified from a human brain cDNA library (Clontech, provided by Suwon Kim, University of California, San Francisco) and cloned into the SalI/SmaI sites of pGBT9 (Clontech) or SmaI/XhoI sites of pACT2. Mouse Hip1R cDNA, provided by David Drubin, University of California, Berkeley, was amplified and cloned into the SalI site of pGBT9 or the BamHI/XhoI sites of pACT2. HC fragments were amplified from bovine brain cDNAs (23) and cloned into the BamHI-SalI sites of pGBT9 (Clontech). Rat AP2 ␤-hinge ear (616 -937) was amplified (24) and cloned into the SmaI/XhoI sites of pACT2. Reverse yeast two-hybrid screens were carried out as previously described (8) and the reporter assays were performed following the Clontech yeast protocols handbook. Single residue mutations were generated with a QuikChange site-directed mutagenesis kit (Stratagene).
In Vitro Binding Assays-Mouse Hip1R-(346 -655) was amplified and cloned into the SalI site of pGEX5X-1 (Amersham Biosciences) and expressed as a GST fusion protein following the manufacturer's protocol. Bovine hub fragment with an N-terminal His tag and bovine brain LCb were prepared as described previously (23). Protein concentration was determined by Bradford assays (Bio-Rad). The in vitro binding assay was carried out as follows. About 10 g of GST-Hip1R-(346 -655) or GST alone were bound to 20 l of 80% glutathione-Sepharose resin (Amersham Biosciences) in binding buffer (50 mM Tris, pH 8.0, 200 mM KCl, 1 mM EDTA, 1% Triton X-100, 50 mM imidazole, 0.5 mg/ml bovine serum albumin) at 4°C for 1 h. After washing once, indicated molar ratios of LCb and/or Hub were added and incubated at 4°C for 1 h. In sequential binding experiments, Hub or LCb were added and incubated for another hour. After washing 2-3 times with the same buffer with 10 mM extra imidazole, the bound proteins were resolved by SDS-PAGE and immunoblotted with a rabbit polyclonal antiserum against the conserved region of LCs (25), and mouse monoclonal antibodies against the His tag (His 6 , Clontech) and against the GST tag (GST, Covance). Following incubation with horseradish peroxidase-conjugated goat anti-mouse/rabbit antiserum (ZyMed), the binding was detected by enhanced chemiluminescence (Amersham Biosciences). In the peptide competition experiments, 3.5 g (ϳ50 pmol) of GST-Hip1R-(346 -655) were bound to resins as described above. The peptide of the conserved region (either wild-type, EEDPAAAFLAQQESEIAGIEND, or mutant EEVPAAAFLAQQESEAAGIAND, 80% purity, SynPep) was added in a 500-l binding reaction and incubated for 30 min. 9 g (ϳ300pmol) of recombinant LCb were added directly into the mixture and incubated for 1 h. The bound fractions were assayed by immunoblotting as described above.
Clathrin Assembly Assays-Hip1R-(346 -655) was prepared by cleaving the GST tag off from the fusion construct using factor Xa following the manufacturer's protocol (Novagen). The assays were modified ac-cording to Ref. 24. His tag hub was preincubated with excess LCb at 4°C for at least 1 h to saturate the hub with the LC subunits. The mixture was loaded onto Nanosep 30,000 concentrators (Pall Corp.) and equilibrated with 10 mM Tris, pH 7.7, to wash away the unbound LCb. After a 5-min spin to remove aggregates, indicated proteins were loaded in mini-dialysis units (Slide-A-Lyzer, 10,000, Pierce) and dialyzed overnight at 4°C against assembly buffer (100 mM MES, pH 6.7, 1 mM EGTA, 0.5 mM MgCl 2 , 1 mM Tris(2-carboxyethyl)phosphine hydrochloride). The mixtures were then centrifuged at 109,000 ϫ g for 45 min to separate the assembled lattices (pellets) from the free forms (supernatant). Equivalent proportions of each fraction were analyzed by SDS-PAGE and Coomassie Blue staining. For quantification of band intensities, gels were scanned and analyzed using NIH Image software.
Immunoprecipitation-Bovine brain LCa was cloned into the EcoRI/ XhoI sites of pcDNA3 (Invitrogen) with the HA sequence incorporated immediately subsequent to the ATG codon. Mouse Hip1R-(1-655) was cloned into the EcoRI/SalI sites of pFlag-CMV2 (Sigma) for expression as an N-terminal FLAG-tagged fusion protein. Human 293T cells were transfected by Lipofectamine 2000 (Invitrogen) with both HA-tagged LCa (or mutants) and FLAG-tagged Hip1R (or the empty vector) for 2 days. The lysates were collected in lysis buffer (phosphate-buffered saline, 0.75% Nonidet P-40, 5 mM EDTA, 1 mM dithiothreitol). Immunoprecipitation was carried out with a monoclonal antibody to the FLAG tag (M2, Sigma) and protein G-Sepharose (Amersham Biosciences). After a 2-3 h incubation, the resins were washed 3 times with the same buffer. The bound proteins were resolved by SDS-PAGE and immunoblotted with monoclonal antibodies to the FLAG and HA (HA.11, Covance) tags.
Subcellular Fractionation of Assembled Clathrin-293T cells transfected for 2 days were homogenized in buffer D (10 mM HEPES, pH 7.5, 150 mM NaCl, 1 mM EGTA, 0.5 mM MgCl 2 , 0.02% NaN 3 , 0.1 mM phenylmethylsulfonyl fluoride, and protease inhibitors) as described previously (25). Briefly, the cells were washed twice with the buffer, and lysed using a Dounce homogenizer (10 strokes) and by passaging twice through 28.5-gauge needles. Unbroken cells were spun away at 1,000 ϫ g for 1 min. The lysates were fractionated at 1,000 ϫ g for 15 min into low-speed pellets (P1, enriched with nuclei, plasma membrane) and supernatant, which were further spun at 100,000 ϫ g for 30 min to high-speed pellets (P2, enriched with Golgi, endocytic compartments and CCVs), and supernatants (S, cytosolic components). Equivalent proportions of each fraction were resolved by SDS-PAGE and immunoblotted with monoclonal antibodies to HC (TD.1), Hip1R (BD Biosciences), and actin (AC-40, Sigma) or rabbit antiserum against green fluorescent protein (GFP) (BD Biosciences).
Indirect Immunofluorescence-Bovine brain LCb-(1-44) (or mutant) was amplified and cloned into the BglII/EcoRI sites of pEGFP-C2 (Clontech) so that the LC fragment was expressed at the C terminus of GFP. Human 293T cells grown on coverslips were transfected with the GFP constructs for 2 days, and prepared as described (26). Hip1R and cortactin were labeled, respectively, with monoclonal anti-Hip1R (BD Biosciences) and anti-cortactin (4F11, Upstate Biotechnology) antibodies, followed by Alexa Fluor 647-conjugated goat anti-mouse IgG (Molecular Probes). Alexa Fluor 568 phalloidin (Molecular Probes) was used to stain F-actin. Images were collected by an API DeltaVision DV3 Restoration microscope using a MicroMax 5 MHz cooled CCD camera (Roper Scientific), and deconvolution was carried out using the API SoftWoRx software.

Hip1/R Binding to Clathrin Light
Chains-To characterize the interaction of Hip1/R with clathrin, a yeast two-hybrid assay was used to map the binding sites for Hip1/R in clathrin components. Consistent with reported properties of the yeast orthologue Sla2p (27,28) and studies of recombinant proteins (13), Hip1 and Hip1R interacted as homo-and heterodimers, although heterodimerization was harder to detect ( Fig. 1). In the assays neither Hip1 nor Hip1R interacted with any portion of the HC, terminal domain (residues 1-495), distal leg (546 -1073), or hub (1074 -1675). Hip1 was previously shown to interact with the HC terminal domain (13,19,20). Terminal domain constructs (1-495 and 1-546) were tested in the study, but we were restricted to data for the shorter fragment because the other gave positive signals regardless of the presence of any partner. We could not rule out the possibility that the interaction of Hip1/R with HC exists, but based on these assays HC did not seem to have a strong binding site. In contrast, Hip1/R interacted robustly with both LCs (LCb in Fig. 1 and not shown for LCa), consistent with conclusions of earlier protein binding and genetic studies of Hip1R, Hip1, and yeast Sla2p (10,12,13). Dimerization and LC interaction were detected for the central region of Hip1R (residues 346 -655, Fig. 1B). Hip1 bound weakly to the hinge ear fragment (residues 616 -937) of the ␤-subunit of the AP2 adaptor molecule, but Hip1R did not (Fig. 1B). Presumably, this is because of the presence of the ␤-ear-binding DPF motif in Hip1 and its absence in Hip1R (13,19,29).
Hip1/R Associates with Clathrin Light Chains via Their Conserved Sequence-To locate the minimal LC region for binding Hip1/R, LC fragments were paired with Hip1 or Hip1R in the yeast two-hybrid assay. Hip1 or Hip1R bound the N-terminal portion of both LCa and LCb, but not their central region where HC binds, nor the C terminus ( Fig. 2A and not shown for LCa). To identify essential interaction residues, random PCR mutagenesis of the N-terminal region of LCb (residues 1-77) was used to generate mutations that abolished the interaction with Hip1. Hip1 was chosen as a testing partner because its combination with LC gave a stronger signal in the yeast twohybrid assay than Hip1R ( Fig. 2A). Most mutations in the loss-of-binding LC mutants were scattered throughout the first 40 residues, including the conserved LC sequence (residues 20 -41, Fig. 2B). This was consistent with the quantitative yeast two-hybrid data, in which the fragment comprising LCb residues 1-44 showed the strongest binding to Hip1 or Hip1R of all the truncated LC fragments tested ( Fig. 2A).
Single mutations derived from the sequences of the mutants found in the screen were constructed by site-directed mutagen-esis of full-length LCb to determine which of the mutations were responsible for loss of binding to Hip1/R. Many LCb mutants with a single mutation within the LC conserved sequence exhibited weak (E20V), extremely weak (Q31R, I38M), or complete loss of interaction (D22V, I35A, E39A) with Hip1R, whereas mutations within the most N-terminal 20 residues of LCb did not significantly affect interactions (Fig. 2C). 2 Because these LCb mutant constructs all interacted with HC (not shown), the negative effects of mutation were not ascribable to low expression levels or misfolding. Interestingly, three single mutations (E32A, S33A, E34A) enhanced the binding signal between LCb and Hip1R. Whereas we cannot rule out that these three mutants had unusually high expression levels, the clustering of the mutations suggests that they too represent a sequence of residues that play a role in the extended interaction between the conserved LC sequences and Hip1/R.
To confirm the mutagenesis results at the protein level, the LC-binding fragment of Hip1R (residues 346 -655) was expressed as a glutathione S-transferase fusion protein (GST-Hip1R) and was tested in vitro for binding to recombinantly expressed wild-type or mutant LCb protein. GST-Hip1R bound wild-type LCb but did not bind LCb mutants I35A, E39A (not shown), or the double mutant comprising I35A and E39A (Fig.  2D). The interaction between wild-type and mutant LCa proteins and Hip1R was then tested in a cellular context. FLAGtagged Hip1R-(1-655) and HA-tagged LCa (wild-type or mutant) were co-transfected into human 293T cells, and they were tested for co-immunoprecipitation. Antibody to the FLAG epitope isolated wild-type HA-tagged LCa bound to FLAGtagged Hip1R, confirming that Hip1R interacts with LCa (Fig.  2E). Single mutations in HA-tagged LCa, D25V, I38A, or E42A (corresponding to D22V, I35A, or E39A in LCb) disrupted the interaction with FLAG-Hip1R and the mutants were not coimmunoprecipitated ( Fig. 2E and not shown). These results confirm that several residues in the conserved sequences of both LCa and LCb are critical for interacting with Hip1R.
Because the residues identified were spread over the conserved sequence of the LCs, we investigated the possibility that the whole conserved sequence was involved in Hip1R binding. A peptide comprising the conserved LC sequence (LCb residues 20 -41 and LCa residues 23-44, EEDPAAAFLAQQESE-IAGIEND, 22-mer) was tested for its ability to compete with full-length LCb for GST-Hip1R binding in vitro. As increasing concentration of the peptide was added, the amount of LCb binding to GST-Hip1R declined (Fig. 2D). A peptide with mutations affecting LC-Hip1/R interactions (EEVPAAAFLAQQ-ESEAAGIAND) failed to compete with LCb for binding to GST-Hip1R (Fig. 2D). Competition between a peptide and a fulllength protein is impressive and strongly suggests that the peptide comprises the extended binding site on LCs for Hip1/R. Furthermore, these results confirm that the three mutated residues are critical for the interaction of these two proteins.
Hip1R Interacts with Clathrin Light Chains Bound to Heavy Chains in Preference to Unbound Light Chains-Hip1/R stimulates clathrin assembly in vitro (10,13,19,20). Because earlier evidence suggested they bind LCs (10 -13), the possibility was raised that Hip1/R might displace LCs and thereby alleviate LC suppression of assembly. Our mapping of the Hip1/R-binding site to the N-terminal LC conserved sequence here and our earlier mapping of the HC-binding site to the central helical region of LCs (8) then raised another possibility that Hip1/R and HC could simultaneously interact with LCs. We therefore tested if the binding of the two partners to LCs is exclusive, using recombinant proteins for in vitro binding as-2 C.-Y. Chen and F. M. Brodsky, unpublished data.

FIG. 1. Hip1 and Hip1R interact with clathrin light chain.
A, plate growth assay. Yeast AH109 cells were co-transformed with the indicated bait (in pGBT9) and prey (in pACT2). Transformants were spotted onto plates lacking leucine and tryptophan (to select for the vectors), with (ϩ) or without (Ϫ) histidine (His) and adenine (Ade). Transformants expressing interacting constructs grew in the absence of histidine and adenine after 4 -6 days. B, ␤-galactosidase filter assay. Yeast SFY526 cells were co-transformed with the indicated bait (in pGBT9) and prey (in pACT2). Transformants were spotted onto filters directly in contact with media and incubated for 2 days. Filter assays were performed following the manufacturer's protocol (Clontech). The blue color indicates a positive interaction, developed within 12 h following substrate application. The assays shown are typical of three independent experiments.

FIG. 2. Hip1 and Hip1R bind to the conserved sequence of clathrin light chains LCa and LCb.
A, mapping the binding region of Hip1 and Hip1R in bovine brain LCb. Yeast SFY526 cells were co-transformed with human Hip1-(124 -1037) or mouse Hip1R-(1-1068) (in pGBT9) and the indicated LCb fragments (in pACT2). Quantitative ␤-galactosidase assays were performed using o-nitrophenyl ␤-D-galactopyranoside as substrate and results are shown in ␤-galactosidase units. Units are defined as the amount that hydrolyzes 1 mol of o-nitrophenyl ␤-Dgalactopyranoside per min per cell, as described in the Clontech yeast protocols handbook. Above the fragments tested, a diagram of full-length LCb is delineated with the conserved and HC binding regions shaded. Similar results (not shown) were obtained with the equivalent fragments of bovine brain LCa. B, sequences of the mutant fragments of bovine brain LCb-(1-77) that were unable to bind Hip1. AH109 cells harboring Hip1-(124 -1037) (in pGBT9) were co-transformed with mutated PCR products of LCb-(1-77) and a gapped prey vector, pACT2. The inserts of the recombinant clones (numbered mutants) unable to interact with Hip1-(124 -1037) were sequenced. Mutated residues from the screens are in bold (black and color). Mutation sites tested individually by site-directed mutagenesis are in color: green, no effect on Hip1 binding; orange, a mild effect; and red, eliminated Hip1 binding. Residues of the conserved sequence (20 -41) are highlighted with a yellow background. C, interaction between the different LCb mutants and full-length Hip1R. The individual mutations generated in the LCb mutants were based on the screen results (Fig.  2B), except for changing residues 32-34 to alanine as part of an initial alanine mutation scan. Quantitative ␤-galactosidase (␤-gal) assays were performed in SFY526 cells harboring bovine brain LCb or mutants (in pACT2) and mouse Hip1R- (1-1068). The results are shown in ␤-galactosidase units as the mean, and error bars indicate the standard deviations of triplicate determinations. Asterisks (*) indicate that these mutants showed very weak interaction with Hip1R in a more sensitive filter assay. D, peptide competition assay. The GST Hip1R-(346 -655) fragment was first bound to glutathione-Sepharose resins. The indicated concentration of conserved peptide (wild-type, WT; or mutant with changes D22V, I35A, and E39A, MT) was added for 30 min, followed by incubating with recombinant LCb for 1 h. Proteins bound to the affinity resins were assayed by immunoblotting with a monoclonal antibody to GST to detect GST-Hip1R or a rabbit polyclonal antiserum raised against the conserved region of LCs. As a control, full-length LCb with mutations I35A and E39A were combined with GST-Hip1R and its binding was not detected. E, co-immunoprecipitation of LCa with Hip1R. Human 293T cells were transfected with FLAG-tagged Hip1R-(1-655) (or empty vector) and HA-tagged LCa (or mutants with changes I38A or D25V) and cell lysates were immunoprecipitated with antibody to the FLAG tag. Immunoprecipitates says. First, a GST fusion protein containing the Hip1R central region (346 -655) was incubated with recombinant LCb, hub, or LCb plus hub. As expected, this construct bound free LCb and only bound hub if LCb was present, because hub does not have an LC-independent Hip1R-binding site (Fig. 3A, lanes 4 -6). The control construct of GST alone pulled down neither LC nor hub (Fig. 3A, lanes 1-3). Next, hub and LCb were added sequentially to immobilized GST-Hip1R (washing in between). When LCb was added after hub, only LCb bound to the Hip1R (Fig. 3A, lane 7). If hub was added after LCb, neither construct was bound to Hip1R (Fig. 3A, lane 8). Thus hub appeared to bind LCb and compete it off the immobilized Hip1R, suggesting that the trimeric hub binds LC with stronger affinity than that between Hip1R and LC. This was not a result of the extra washes because treatment of Hip1R-bound LC with only buffer did not extract the bound LC from Hip1R (Fig. 3A, lane 9). When LCb and hub were introduced at the same time to immobilized Hip1R (Fig. 3A, lane 5), hub and LCb were bound by Hip1R, but less LCb was bound than in the absence of hub (Fig.  3A, lane 4). To establish the binding preferences of Hip1R for free LCb and hub-bound LCb further competition experiments were performed (Fig. 3B). Bound and free LCb were preincubated with GST-Hip1R followed by further binding of bound or free LCb. One, three, or six molar ratios of LCb were incubated with Hip1R and increasing amounts were bound (Fig. 3B, lanes  2-4). However, when these were exposed to subsequent incubation with a fixed amount of hub-bound LCb at three molar ratios to the Hip1R, equal amounts of hub and LCb were bound and the excess free LCb was displaced (Fig. 3B, lanes 5-7). On the other hand, when increasing molar ratios of hub-bound LCb were first bound to Hip1R and then incubated with three molar ratios of free LCb (lanes 11-13), the increasing amounts of hub and LCb bound to Hip1R remained unchanged (compared with lanes [8][9][10], showing no displacement of hub-bound LCb by free LCb. These results indicate that Hip1R interacts with LCs that are bound to HCs in preference to unbound LCs. This observation, however, appears to contradict the fact that less LCb is bound by Hip1R when hub is present (Fig. 3A, lane  5). One explanation is that free LCs binds Hip1/R stoichiometrically but weakly. Then, when bound to hub, LCs fold better and bind more strongly to Hip1/R but can no longer bind stoichiometrically. Taken together, these observations unequivocally indicate that Hip1R does not displace LC from its HC-binding site and demonstrate that Hip1R can bind the LC-hub complex, so both clathrin subunits can bind Hip1R simultaneously.
Hip1R Promotes the Assembly of Clathrin Hub through Light Chain Binding-Because LC displacement is not likely to play a role in Hip1/R effects on clathrin, we hypothesized that binding to the regulatory residues in the conserved LC sequence might account for stimulation of clathrin assembly in vitro. To address this, we analyzed the activity of a recombinant fragment of Hip1R (residues 346 -655) in promoting assembly of recombinant hub fragments with bound LCs (wild-type or with the Hip1R-binding residues mutated). Under weakly acidic conditions, pH 6.7, recombinant hub assembles into large pseudo-lattices (23,24), which can be fractionated from unassembled protein by ultracentrifugation (Fig. 4, lanes 1 and 2). Note that this assembly is reversible (data not shown). At pH 6.7, recombinant hub fully occupied by recombinant LCb does not extensively assemble because of the presence of the negatively charged residues at the N terminus of the LC conserved sequence (4), so most protein remains in the supernatant (14 -34% in the pellet versus 73% in the pellet without LCs) (Fig. 4,  lanes 3-8). When recombinant Hip1R (346 -655, cleaved from GST) was added to hub with bound wild-type LCb, it increased the amount of assembled protein in the pellet to 62% (Fig. 4,  lanes 11 and 12). Note that the Hip1R fragment did not appear in the pellet of assembled hub without LCb (lanes 9 and 10). Hip1R promoted hub assembly in the presence of LCb (lanes 11 and 12) but did not fully induce assembly to the levels observed in the absence of LCb (lanes 1 and 2 and 9 and 10). This could be explained if Hip1R is not very efficient at promoting assembly, which would be expected if Hip1R cannot bind stoichiometrically to hub-LCb complexes, as suggested by the results shown in Fig. 3. When the Hip1R fragment was added to hub occupied by mutant LCb (single I35A or double I35A,E39A mutants), it did not effectively promote assembly (only 22-43% in the pellet) (Fig. 4, lanes 13-16). The results were statistically significant (p Ͻ 0.05, for the comparison between lanes 11 and were analyzed for the presence of Hip1R and LCa by immunoblotting, respectively, with monoclonal antibodies to the FLAG and HA tags. The presence of the proteins in the transfected lysates was confirmed by immunoblotting the left-hand lanes. Note that LCa residue numbering is different from LCb, so that the mutations tested were equivalent to those shown in C and D.

FIG. 3. Hierarchy of binding preference among recombinant hub, LC, and Hip1R.
A, LCs bind to hub more strongly than to Hip1R. GST alone or the GST Hip1R-(346 -655) fragment was first bound to glutathione-Sepharose resins. Recombinant LCb, His-tagged hub, or LCb and hub together were applied to the bound resins and incubated for 1 h. In sequential binding (lanes 7-10), the first fragment (1) was applied for 1 h. Subsequent to a wash, the second component (2) was applied for another hour. Proteins bound to the resins were assayed for GST Hip1R-(346 -655) (or GST), hub and LCb fragments by immunoblotting, respectively, with monoclonal antibodies to GST, to the His tag epitope, and a rabbit polyclonal antiserum to the conserved region of LCs. B, Hip1R prefers hub-bound LCs to free LCs. The GST Hip1R-(346 -655) fragment was bound to glutathione resins. Hub and LCb were preincubated for 1 h. Recombinant LCb, His-tagged hub, or LCbbound hub were applied to the bound resins and incubated for another hour. The numbers indicated refer to the molar ratios of LCb and HC fragments relative to the amount of Hip1R fragments applied to the resin. Note that three HC fragments form a hub. For sequential binding (lanes 5-7 and 11-13), the LCb or Hub ϩ LCb at the ratios indicated were incubated with the Hip1R on the resin, then following a wash, the second component (or buffer as mock) was applied for another hour. Proteins bound to the resins were assayed for the presence of hub and LCb fragments as described in A. 12 and lanes 15 and 16). Thus the mutant LCs retained the regulatory residues that inhibited hub assembly, but were defective in Hip1R binding through mutations in other residues of the conserved LC sequence. The implication is that Hip1R binding to the conserved sequence can reverse suppression of hub assembly by LCs.
Overexpression of the Hip1R Binding Fragment of Clathrin Light Chains Affects Actin Distribution in Vivo-To test the role of the Hip1R-LC interaction in vivo, we transfected 293T cells with a fragment of LCb (residues 1-44) comprising the Hip1R-binding region, expressed as a fusion protein with GFP. Protein expression and clathrin assembly state were analyzed in the transfected cells. The amount of Hip1R was noticeably decreased in the cells expressing GFP-LCb-(1-44) (Fig. 5A), suggesting that the LCb fragment was blocking the binding of Hip1R to cellular proteins and thereby decreasing its stability. This effect on Hip1R stability was not observed upon transfection with a GFP fusion protein comprising LCb-(1-44) bearing three mutations in the Hip1/R binding sequence (D22V, I35A, E39A) or upon transfection with GFP alone (Fig. 5A). The pools of assembled versus unassembled clathrin were then compared in cells transfected with the different GFP constructs. The amounts of HC in the assembled (P2) and cytosolic (S) fractions were similar for cells expressing any of the GFP constructs (Fig. 5B). However, the amount of Hip1R in cells expressing GFP-LCb-(1-44) was still significantly lower than the level in cells expressing GFP alone or the GFP construct with mutant LCb-(1-44) (Fig. 5B). Thus, overexpression of the LC fragment containing the Hip1/R-binding region does not appear to affect the degree of clathrin assembly, despite causing a reduction in Hip1R level.
The morphology of the transfected cells was examined by immunofluorescence. Hip1R staining appeared diminished in cells expressing GFP LCb-(1-44), compared with those expressing GFP alone (Fig. 5C). However, localization of clathrin HC or the AP2 adaptor or internalized transferrin seemed to be unchanged by transfection with either GFP construct, as assessed by immunofluorescence (not shown). Most of the actin staining around the cell cortex was the same in cells transfected with GFP, GFP-LCb-(1-44), or GFP-mutant LCb-(1-44). However, effects on distribution of cortical actin were observed at the interface region where cells were in contact with coverslips (Fig. 5, C and D). In cells expressing GFP-LCb-(1-44), abnormal actin projections in globular-spherical (Fig.  5D) or parallel-oblong (Fig. 5C) shapes were observed at focal contact sites. These actin structures were rarely seen in cells expressing GFP alone or the GFP mutant LCb-(1-44). The abnormal actin projections were then analyzed for the presence of other proteins. HC, AP2, or vinculin (a protein involved in focal adhesion) partially co-localized with the projections. However, cortactin, a protein recently reported to bind Hip1R (21), was consistently localized to the tips of the actin protrusions in cells expressing GFP-LCb-(1-44) (Fig. 5C). These observations suggest that disruption of Hip1R binding to LCs results in Hip1R instability and has a consequent effect on cellular actin dynamics.

DISCUSSION
Role of the Conserved LC Sequence-All vertebrate clathrin LCs share a conserved sequence of 22 residues, located about 20 residues from the N terminus. Apart from this sequence the vertebrate LCs (LCa and LCb) have 60% sequence identity, although they do share other common features including the central location of the helical HC-binding domain, a calciumbinding site just N-terminal to this domain, and a calmodulinbinding site at the C terminus. Non-vertebrate species that have only one clathrin LC have a considerable range of sequence divergence. 3 However, for many of these, the most conserved sequence besides the HC-binding domain corresponds to the conserved LC sequence in vertebrates. Thus the function of the conserved sequence must be fundamental to clathrin function. Previously we showed that the three negatively charged N-terminal residues of this sequence regulate clathrin assembly (4). Here we demonstrate that one of these residues and several in the remainder of the conserved sequence are critical for binding Hip1/R proteins, and that the entire 22-residue segment serves as a Hip1/R-binding site. Three hypotheses emerge from this observation. First, the coincidence of Hip1/R binding with the assembly regulating residues is likely to play some role in controlling clathrin assembly. Second, Hip1/R binding must be vital for CCV function because its binding site is so conserved. Third, a corollary of these first two hypotheses is that the regulation by Hip1/R of actin dynamics relative to clathrin dynamics is likely to be critical for effective CCV formation or transport.
Role of Hip1/R in Clathrin Assembly-The promotion of clathrin assembly by Hip1/R in vitro can be explained if the binding of Hip1/R to the conserved sequence blocks the inhibitory negative charges at the N terminus. Indeed, we show that the promotion of Hip1/R of the assembly depends on its ability to bind the conserved LC sequence. Mutation of either the LC conserved sequence as shown here or the corresponding binding region on Hip1/R (42) abrogates assembly activity. Furthermore, the binding site on Hip1 for LCs was found to contain some basic patches, such as 451 LLR (42) that could potentially contribute to neutralization of the overall acidic nature of the LC conserved sequence. It is also feasible, although not sup- ported by the studies reported here, that Hip1/R acts to promote clathrin assembly through dual binding to clathrin via the LC binding site plus a predicted clathrin box motif. This motif, which is supposed to interact with the N-terminal domain of HCs (30), is 5Ј to the central region of Hip1 (LMDMD) or Hip1R (LIEIS). Although, no interaction with the relevant domain of HCs was detected by the yeast two-hybrid assays performed here, in vitro studies showed that that Hip1 bound HCs, but Hip1R did not (10,13,19,20). Thus there could be some subtle interaction of HCs with the Hip1 clathrin box that is compromised in the yeast two-hybrid assay. Despite the apparent strength of the LC-Hip1R interaction, disruption of the LC-Hip1R interaction by transfection did not have a detectable effect on levels of intracellular clathrin assembly, although it did decrease the levels of Hip1R in cells. However, in a cell, clathrin assembly and disassembly is controlled by numerous accessory proteins. Whereas, LCs suppress the fundamental assembly reaction, this suppression is overcome by the binding of several different intracellular adaptor molecules, including AP1, AP2, GGAs, AP180/CALM, epsin, and Dab2 (2,31). These factors presumably orient triskelia into anti-parallel alignment for assembly, increase the curvature of the bound membrane, and/or alter the thermodynamics to favor lattice formation (24,(32)(33)(34). Thus even if Hip1/R contributes to reducing the threshold of clathrin assembly by neutralizing the intrinsic regulation by LCs, loss of its assembly promoting activity might not be easily detectable in the cellular context where promotion of clathrin assembly is a redundant activity of a number of CCVassociated proteins.
Role of Hip1/R-LC Binding in Connecting CCVs to Actin-Hip1R is an actin-binding protein, as well as a clathrin-binding protein, and several studies have demonstrated that Hip1R can influence the orientation of actin with respect to CCVs (11,21). Recently, reduction of Hip1R in mammalian cells by RNA in- and cultured for 2 days. Equal volumes of cell lysates were resolved by SDS-PAGE and immunoblotted with monoclonal antibodies to detect clathrin heavy chain (CHC), Hip1R, and actin, or a rabbit polyclonal antiserum against GFP to detect LCb-(1-44) and GFP. B, the cell lysates were fractionated at 1,000 ϫ g for 15 min into low-speed pellets (P1), and the supernatants were further spun at 100,000 ϫ g for 30 min to yield high-speed pellets (P2) and supernatants (S). Immunoblotting was carried out as described in A. C, 293T cells grown on coverslips were transfected with GFP, GFP-LCb-(1-44), or GFP-LCb-(1-44) mutant (with the triple mutations D22V,I35A,E39A). Hip1R and actin were, respectively, labeled with monoclonal antibody to Hip1R plus Alexa 647-goat anti-mouse IgG and Alexa 568-phalloidin. GFP and the other fluorescent markers were visualized using specific filters and the individual images are shown in the top three rows with each column representing the same field of cells transfected as indicated in the top row. The focal contact surfaces of the cells are shown here and bottom row images are digital enlargements of the boxes indicated above with the GFP, Hip1R, and actin images merged. D, similar experiments were performed to those in C, except that cortactin were stained instead of Hip1R. The bottom row images are digital enlargements of the boxes indicated above with the actin and cortactin staining merged. Bars, 10 m. terference or of sla2p protein in yeast sla2⌬ mutants was reported to enhance co-localization of actin with CCVs and to increase actin polymerization in the cell periphery (21,35). We observed a similar phenotype in cells expressing LCb-(1-44), which displayed reduced Hip1R levels. The loss of interaction of Hip1/R with LCs appeared to destabilize Hip1/R. Aberrant actin projections were present at the focal adhesion side of the plasma membrane. In some instances, clathrin-coated pits visualized by AP2 staining of transfected cells appeared to be tightly associated with actin structures. This contrasted with AP2 staining in wild-type cells that aligned with actin but did not have overlapping localization. However, no dramatic effects on transferrin uptake were observed in the LC fragment expressing cells (not shown), unlike inhibitory effects observed upon small interfering RNA reduction of Hip1R in cells (21). Whereas Hip1R was less stable in the presence of the interacting LC fragment, the reduction in Hip1R levels was not as great as observed when Hip1R was targeted by small interfering RNA, which could account for the differences in endocytic phenotype observed. Surprisingly, no effects on transferrin and epidermal growth factor uptake were observed when LC expression was inhibited by small interfering RNA (36). However, in these experiments the fate of Hip1/R was not established and could be different from its fate when bound to an LC fragment.
Interestingly, cortactin was consistently observed at the tips of the aberrant actin projections formed following Hip1R reduction by transfection with GFP-LCb-(1-44). Cortactin was recently shown to be enriched in clathrin-coated pits and capable of binding Hip1R, dynamin, and F-actin (21,37). Normally, cortactin localizes to the periphery of cells that are polarized or migrating, playing a role in regulating actin nucleation and stabilizing actin filament branching (38). In non-polarized cells, like 293T, cortactin is not usually abundant at the cell cortex, but upon reduction of Hip1R, we observe that its colocalization with actin at cell-substrate adhesions is increased. Thus it appears that Hip1R is involved in regulating actin dynamics through cortactin. Whether it sequesters cortactin for activation or inactivation near CCVs cannot yet be determined, although others have suggested that Hip1/R negatively regulates actin dynamics (21,39).
Proposed Mechanism for Hip1/R Action Linked to Clathrin-Based on the in vitro binding data shown in Fig. 3, we propose a hierarchy of Hip1/R and clathrin interaction (Fig. 6). First, free LC has a preference for binding to hub over binding to Hip1/R (Fig. 6A), explaining how hub can "steal" LCs from Hip1R (Fig. 3A). Second, Hip1/R has a preference for LCs bound to hub over free LCs (Fig. 6B), explaining the fact that LCs are competed off Hip1R by hub-bound LCs (Fig. 3B). Third, the Hip1/R dimer binds more free LCs than hub-bound LCs so there must be a reduction in the stoichiometry because of steric hindrance from the presence of HCs (Fig. 6B), even though there is an increase in avidity for LCs when they are HC bound. This avidity increase is a likely result of the fact that LCs fold better when bound to HC (8). Thus, in the presence of hub, hub sequesters LCs away from Hip1/R. Hip1/R can rebind to folded hub-bound LCs but with a slower on rate, because of the reduced LC accessibility on hub. Furthermore, in this state a Hip1/R dimer does not appear to be able to bind two hub-LC complexes. This binding paradigm suggests that in a cell, Hip1/R dimer binding to LCs on a clathrin triskelion is substoichiometric (Fig. 6C). Substoichiometric binding between Hip1/R and HC-bound LCs is additionally consistent with the partial effect of Hip1R on assembly of hub-LC complexes in vitro (Fig. 4).
Binding of the central domain of Hip1/R to LCs localizes part of Hip1/R to the outside surface of CCVs, where it could easily regulate CCV interaction with the actin cytoskeleton. Images of Hip1R as an extended "dumbbell" (10) reveal that it could also project into the interior of a CCV as far as the internal membrane vesicle where its ANTH domain could influence lipid rearrangement (Fig. 6D). Lipid interactions and actin polymerization, as well as LC binding by Hip1/R could all contribute to CCV formation and function. Our analysis of the binding preference between clathrin subunits and Hip1R indicates that the basic structure of the clathrin triskelion is likely to form before Hip1/R binds. Furthermore, Hip1/R binding is unlikely to disrupt a triskelion by removal of LCs. Binding of Hip1/R to a triskelion may then lower the threshold for clathrin assembly by neutralizing LC suppression. Assembly will be further promoted by cooperative interactions from HC distal leg domains (40) brought into proper alignment by various adaptor molecules (24). Rearrangement of polygonal lattices readily occurs with little energy cost and spontaneously contributes to the basket curvature and membrane deformation (32). However, this latter process is facilitated by various lipidbinding adaptors (41) and Hip1/R may consequently contribute. One unusual feature of Hip1/R is that it is the first such adaptor to interact with the externally oriented LC-hub portion of the clathrin coat rather than the internally oriented HC terminal domain. This may facilitate the multifunctional roles of Hip1/R and allow it to regulate the relationship between FIG. 6. Proposed hierarchy, stoichiometry, and morphology of Hip1/R and clathrin interaction. A, LC was prebound to Hip1/R and then combined with hub as done in Fig. 3A. LCs have a preference for hub over Hip1/R. Eventually, the LC-hub complex can rebind to Hip1/R but with reduced stoichiometry and slower equilibrium than binding of free LCs to Hip1/R. B, LC was prebound to Hip1/R and then combined with LC-hub complexes as done in Fig. 3B. LC is competed off Hip1/R because hub-bound LC interacts more avidly, but the stoichiometry of LC-hub binding is reduced compared with free LC binding to Hip1/R. C, in vivo, LCs bind first to HCs forming a triskelion. Hip1/R binds to a triskelion and to an assembled clathrin lattice at substoichiometric amounts compared with the number of potential LC binding sites for Hip1R. D, the geometry of the Hip1/R-LC interaction could allow Hip1/R to influence actin interactions on the CCV surface and simultaneously bind to lipids invaginating into a CCV.
CCVs and the actin cytoskeleton as well as influence CCV assembly by interaction with LCs and membrane lipids.