Partitioning of Proteins into Plasma Membrane Microdomains

Internalization of membrane proteins involves their recruitment into plasma membrane clathrin-coated pits, with which they are thought to interact by binding to AP-2 adaptor protein complexes. To investigate the interactions of membrane proteins with coated pits at the cell surface, we applied image correlation spectroscopy to measure directly and quantitatively the clustering of influenza hemagglutinin (HA) protein mutants carrying specific cytoplasmic internalization signals. The HA system enables direct comparison between isolated internalization signals, because HA itself is excluded from coated pits. The studies presented here provide, for the first time, a direct quantitative measure for the degree of clustering of membrane proteins in coated pits at the cell surface. The degree of clustering depended on the strength of the internalization signal and on the integrity of the clathrin lattices and correlated with the internalization rates of the mutants. The clustering of the HA mutants fully correlated with their ability to co-precipitate α-adaptin from whole cells, the first such demonstration for a membrane protein that is not a member of the epidermal growth factor receptor family. Furthermore, both the clustering in coated pits and the co-precipitation with α-adaptin were dramatically reduced in the cold, suggesting that low temperature can interfere with the sorting of proteins into coated pits. In addition to the specific results reported here, the general applicability of the image correlation spectroscopy approach to study any process involving the clustering or oligomerization of membrane receptors at the cell surface is discussed.

Receptor-mediated endocytosis constitutes a crucial part in the life cycle of cell surface proteins and provides a major mechanism for receptor down-regulation and signal termination (1)(2)(3)(4). The clathrin-coated pits, which are the major port of entry into the endocytic pathway in many cell types, represent the best studied example of a sorting domain where a coat of proteins at the cytosolic face of the membrane selectively binds membrane proteins containing specific motifs that serve as internalization signals (5)(6)(7). Two classes of clathrin-associated assembly proteins (APs) 1 have been identified, one specific for plasma membranes (AP-2) and one for the trans-Golgi network (AP-1) (5-10). AP-2 is known to function both in assembly of the clathrin lattice (10 -13) and in binding to membrane proteins carrying internalization signals, which are concentrated in clathrin-coated pits for endocytosis (5, 14 -22).
Recent evidence identifies several distinct classes of cytoplasmic internalization signals (reviewed in Refs. 3, 4, 23, and 24). The best characterized are those containing an essential aromatic residue, typically a tyrosine. They conform mostly to one of two major subclasses: NPXY, where Tyr is preceded by an asparagine-proline dipeptide and a random amino acid, or YXXZ, where Z is an amino acid with a hydrophobic side chain (3,(23)(24)(25). It was suggested that a type 1 tight turn conformation might constitute a general feature of tyrosine-based internalization signals (3, 23, 26 -30). However, a recent study employing combinatorial selection methods on the binding of YXXZ-containing peptides to the 2 chain of AP-2 indicated no requirement for a prefolded structure around the tetrapeptide signal (31). A different class of internalization signals contains a di-leucine motif (LL or LI) (24,(32)(33)(34).
Although much has been learned about clathrin-mediated endocytosis in recent years, currently it is not well understood how membrane proteins bind to coated pits at the surface of the intact cell. In vivo endocytosis studies generally measured the sequestration of proteins into vesicles or deeply invaginated pits and did not allow direct measurement of many events that occur earlier during coated vesicle formation (e.g. the selection of certain proteins for inclusion in coated pits and the exclusion of others). Furthermore, experimental evidence for interactions of the internalization signals of membrane receptors with AP-2 subunits has been limited mostly to in vitro assays employing solubilized and immobilized proteins (16 -19, 21, 31, 35). These assays have detected only a subset of the sequences known or suspected to function as internalization signals in vivo (36,37). In particular, association of AP-2 with receptors by co-immunoprecipitation from cells has thus far been demonstrated only for receptors from the epidermal growth factor (EGF) family (14,15,20,21,38). Even for these receptors, it is not clear whether this in vitro association is directly related to the efficiency of their internalization via coated pits (21,39,40). It is therefore important to explore the relationships between binding to AP-2, clustering in coated pits, and the efficiency of the internalization process as a function of the internalization signal; this manuscript presents measurements to that end.
Our studies were performed on a series of influenza hemagglutinin (HA) mutants carrying specific cytoplasmic internalization signals (37, 41,42). This system is advantageous, because wild-type HA (HA wt) lacks internalization signals and serves as a natural control, enabling investigation of specific internalization sequences introduced into its cytoplasmic tail (41,43,44). Using this system, we employed comparative studies of the lateral mobility of these mutants to characterize the mode of their interactions with coated pits at the surface of intact cells (43,44). However, in these studies, measurement of the binding of internalization-competent HA mutants to immobile structures presumed to represent coated pits was indirect and was inferred from the reduction in the lateral diffusion rate or mobile fraction of the mutants relative to HA wt. Because numerous factors (and not only interactions with coated pits) may inhibit the lateral mobility of membrane proteins, it was important to measure the association of the HA mutants with coated pits at the cell surface in a direct and independent manner.
The current studies, which are potentially applicable for studying any type of clustering or oligomerization at the cell surface, provide a direct demonstration and measure for internalization signal-dependent clustering of HA mutants in cell surface coated pits. Together with the findings on signal-dependent co-precipitation of AP-2 with the various HAs in strict correlation with the ICS data, our results suggest that the degree of clustering in coated pits for a given protein depends on the strength of the association of its internalization signal with the clathrin-associated adaptor complexes. In turn, this clustering plays an important role in determining the endocytosis rate and restricts the lateral mobility of the internalization-competent proteins at the cell surface.
Antibodies-Polyclonal rabbit antiserum that recognized all the mutants of the Japan HA (A/Japan/305/57 strain) was employed throughout (43,44). Fluorescein-labeled affinity purified F(abЈ) 2 of goat IgG directed against rabbit F(abЈ) 2 (FITC-GAR F(abЈ) 2 ) and normal goat IgG were purchased from Jackson ImmunoResearch. Monovalent FabЈ fragments were prepared from the IgG fraction of the anti-HA antibodies and from normal goat IgG as described (45). The FITC-GAR F(abЈ) 2 were converted to monovalent FabЈ fragments (FITC-GAR FabЈ) by reduction with mercaptoethanol followed by alkylation with iodoacetamide (45). To eliminate any possible IgG traces, all FabЈ preparations were treated with protein A-Sepharose. The resulting FabЈ were free of contamination by IgG or F(abЈ) 2 as judged by SDS-polyacrylamide gel electrophoresis under nonreducing conditions. AC1-M11 mouse monoclonal antibodies specific for the ␣a and ␣c chains of AP-2 (46) were donated by Dr. Margaret S. Robinson (University of Cambridge, UK).
Recombinant Virus Vectors and Cell Culture-The cDNA encoding the HA wt protein from the A/Japan/305/57 strain or several HA mutants were introduced into the SV40-based vector pkSVE (47). The mutants are depicted in Table I. The derivation of the mutants HAϩ8 (containing a carboxyl-terminal extension of eight amino acids, which generates a strong internalization signal based on the motif YKSF) and HAϩ4 (which is HAϩ8 minus the last four residues, with the internalization motif truncated after YK) is detailed elsewhere (42). HA-Y543, which is a point mutant of HA wt (cysteine 543 replaced by tyrosine), was described previously (41). Recombinant SV40 virus stocks were prepared in CV-1 monkey fibroblasts (American Type Culture Collec-tion, Rockville, MD) using the appropriate pkSVE vector together with the dl1055 helper virus as described (48). The CV-1 cells were grown in Dulbecco's modified Eagle's medium containing 10% fetal calf serum (from Biological Industries, Beth Haemek, Israel, or from Life Technologies, Inc.), 100 units/ml penicillin, and 100 g/ml streptomycin (Biological Industries or Life Technologies, Inc.). For ICS experiments on the degree of aggregation and for co-immunoprecipitation studies, subconfluent CV-1 cells were infected in suspension with second or third passage recombinant virus stocks as described (37,48). The cells were plated either on glass coverslips (for immunofluorescent labeling and ICS studies) or in 150-mm dishes (for co-immunoprecipitation studies). Experiments were performed 36 -38 h post infection (for HA wt, HAϩ4, and HA-Y543) or after 44 -46 h (for HAϩ8). The longer post infection time for HAϩ8 was selected to achieve cell surface density closer to that of the other mutants, because the percentage of HAϩ8 at the cell surface is significantly lower (42).
Immunofluorescent Labeling for Image Correlation Spectroscopy-CV-1 cells infected and grown as above were washed twice with cold HBSS containing 20 mM HEPES and 2% BSA (HBSS/HEPES/BSA, pH 7.2) and labeled successively in this buffer at 4°C (washing three times after each incubation) with the following antibodies: (a) anti-HA FabЈ (100 g/ml, 30 min); (b) FabЈ of normal goat IgG (to block nonspecific staining; 200 g/ml, 20 min); and (c) FITC-GAR FabЈ (30 g/ml, 30 min). The labeling was done in the cold to allow only surface labeling and to avoid endocytosis. In cases where the cells were preincubated in specific buffers to alter coated pit structure (see below), the specific buffers employed for each treatment were used throughout all antibody incubation and subsequent steps. In most cases (except where indicated), the labeled cells were warmed to 22°C for 10 min prior to fixation in methanol (Ϫ20°C, 5 min) and acetone (Ϫ20°C, 2 min) to enable stronger interactions with coated pits while avoiding significant internalization of some mutants at 37°C (43,44). The fixed cells were mounted in Slowfade solution (Molecular Probes, Eugene, OR) or in Airvol 205 containing n-propylgallate and taken for the ICS studies.
Image Correlation Spectroscopy-ICS (described in detail in Refs. 49 and 50) is an adaptation of the fluorescence correlation spectroscopy method (51-53) used to analyze fluorescence images collected on a confocal laser scanning microscope. It is sensitive to and capable of quantifying differences in the aggregation state and distribution of fluorescently labeled components at the cell surface. In fluorescence correlation spectroscopy, one examines the volume or area illuminated by a laser beam, usually in a microscope. In ICS, instead of observing fluorescent particles as they diffuse in and out of a fixed laser beam, one generates an image of the distribution of the fluorescence intensity on the cell surface by scanning a fluorescently labeled cell with a laser beam, recording fluctuations in fluorescence intensity as a function of position rather than of time (49,53). The signal analysis is based on calculating the correlation function of the fluorescence intensity fluctuations. When a single fluorescently labeled component is analyzed (as in the current measurements), the autocorrelation function, g(,), is derived and fitted to a two-dimensional gaussian function determined by the intensity profile of the laser beam (49), where and are the position coordinates (for the x and y axes, respectively) of the autocorrelation function, is the gaussian radius of the laser beam, and g(0,0) is the value of the autocorrelation function upon extrapolation of and to zero; g 0 is an offset introduced to account for the finite sample size, which may result in a decay of g(,) to a nonzero value. g(0,0) and are extracted from the fitting procedure (49). Importantly, 1/g(0,0) is equal to N p , the average number of independent fluorescently labeled particles in the area illuminated by the beam (49,54). The term "particles" refers to any fluorescently labeled species, i.e. a fluorescent monomer is one particle, and an aggregate containing many monomers is also one particle. Therefore, aggregation leads to a reduction in N p (larger and fewer particles), which is measured by the 1/g(0,0) value (49). After labeling and fixation as described under "immunofluorescent labeling", fluorescent images were obtained using a Bio-Rad MRC-600 confocal microscope, illuminating at 488 nm with a 25 mW argon ion laser attenuated to 1%. Images were collected in the photon-counting mode (accumulating 25 scans) to ensure linear scaling. Zoom-10 images of flat 16.3 ϫ 16.3-m 2 areas on the cell surface (avoiding regions around the nuclei, which are not in the same focal plane) were submitted for autocorrelation analysis (49) on a massively parallel computer (MP-2, MasPar Computer Corporation, Sunnyvale, CA). S.E. values were calculated at the 99% confidence level using the formula S.E. ϭ t(S.D./n 0.5 ), where S.D. is the sample standard deviation, n is the number of measurements (cells imaged), and t is the t statistic for n Ϫ 1 degrees of freedom (equal to 2.576 at the 99% confidence level in view of the high n in the experiments). p values were determined based on a t test for two independent samples with different variance (55), and the null hypothesis was rejected for p Ͻ 0.01.
Treatments That Alter Coated Pit Structure-The treatments employed were incubation in a hypertonic medium (56 -58) and acidification of the cytosol (58 -60). Hypertonic treatment, which blocks endocytosis via coated pits by dispersing the underlying clathrin lattices (56,58), was performed by a 15-min incubation (37°C) in HBSS/HEPES/ BSA supplemented with 0.45 M sucrose. Cytosol acidification, which alters the coated pit structure and eliminates endocytosis by blocking the pinching-off of clathrin-coated vesicles, was performed as described earlier (43,60). Briefly, cells were incubated in HEPES-buffered Dulbecco's modified Eagle's medium, pH 7.2, containing 30 mM NH 4 Cl (30 min, 37°C), followed by 5 min (37°C) in potassium-amiloride (KA) buffer (0.14 M KCl, 2 mM CaCl 2 , 1 mM MgCl 2 , 1 mM amiloride HCl, 20 mM HEPES, pH 7.2) containing 2% BSA. For both hypertonic and cytosol acidification treatments, the treated cells were subjected to immunofluorescent labeling at 4°C in the appropriate buffer (sucrosecontaining hypertonic buffer or KA), incubated 10 min at 22°C, fixed, and taken to the ICS studies. For co-immunoprecipitation, the cells were treated exactly the same, except that no antibodies were added during the 4°C incubation prior to detergent solubilization and immunoprecipitation.
Co-immunoprecipitation and Western Blotting-After infection in suspension by the appropriate SV40 recombinant viruses, the infected CV-1 cells were plated in parallel in 150-mm dishes (for co-immunoprecipitation) and in 30-mm dishes (to calibrate the level of expression of the different mutants at the cell surface by surface biotinylation; see below).
After incubation at 4°C (as for immunofluorescence, but without antibodies), the cells were warmed to 22°C for 10 min and subjected to co-immunoprecipitation based on an earlier protocol (adapted from Refs. 15 and 20). Cells in one to three 150-mm dishes were washed with HBSS/HEPES and solubilized in 0.5 ml/dish of cold lysis buffer (50 mM HEPES, pH 7.5, 150 mM NaCl, 1% (w/v) Triton X-100, 10% glycerol, 1.5 mM MgCl 2 , 1 mM EGTA, 10 g/ml leupeptin, 10 g/ml aprotinin, 1 mM phenylmethylsulfonyl fluoride). Extracts were precleared by incubation (1 h, 4°C) with 80 g/ml protein A-Sepharose. After centrifugation, the supernatant was incubated (2 h, 4°C) with anti-HA IgG (80 g/ml), followed by the addition of protein A-Sepharose at a level high enough to bind all the IgG (160 g/ml beads; 2 h, 4°C). Immunocomplexes were washed three times with 20 mM HEPES, pH 7.5, 150 mM NaCl, 0.1% (w/v) Triton X-100, and 10% glycerol. The pellets were dissolved in 40 l of Laemmli gel electrophoresis loading buffer (containing SDS and mercaptoethanol), boiled for 5 min, and subjected to SDS-polyacrylamide gel electrophoresis on 10% acrylamide gels. The amount loaded in each lane was calibrated to include similar levels of surface HA molecules using a surface biotinylation assay (see below). The proteins were electrophoretically transferred onto nitrocellulose filters. For incubation of the blots with antibodies, they were presaturated (1 h, 4°C) with TBST buffer (50 mM Tris, pH 7.4, 100 mM NaCl, 0.1% Tween 20) supplemented with 5% dry milk (low fat) and 2% BSA. AC1-M11 anti-␣-adaptin ascites (1:100 dilution) were added; the filters were incubated overnight at 4°C and washed extensively with TBST. For detection, AC1-M11 antibodies were labeled with protein A-peroxidase (1 g/ml, 1 h, 4°C). After washing with TBST three times, the filters were reacted 1 min with the ECL reagent and exposed to autoradiography film. Densitometry was performed on a BioImaging System 202D (Dynco-Renium, Jerusalem) using TINA 2.0 software by Ray Test (Straubenhardt, Germany).
To calibrate the level of each HA mutant at the cell surface, intact cells growing in 30-mm dishes and preincubated exactly as in the specific co-precipitation experiment were biotinylated at the surface by the membrane-impermeable reagent sulfo-NHS-LC-biotin (61). The biotinylation (0.5 mg/ml sulfo-NHS-LC-biotin, 30 min, 4°C) was performed in cold (to stop HA transport to and from the membrane, as was done for the labeling with fluorescent antibodies) HBSS/HEPES (for untreated cells), HBSS/HEPES supplemented with 0.45 M sucrose (hypertonically treated cells), or KA buffer minus amiloride (which contains free amino groups and is not required at this stage due to the low temperature) for the cytosol-acidified cells. The cells were subjected to solubilization, immunoprecipitation with anti-HA IgG, SDS-polyacrylamide gel electrophoresis, and electroblotting as described above. For detection of the biotinylated HA proteins, the protocol described by Sargiacomo et al. (62) was employed. The filters were preblocked (1 h, 4°C) with TGG buffer (50 mM Tris, pH 7.4, 100 mM NaCl, 1 M glucose, 10% glycerol, 0.5% Tween 20) supplemented with 1% dry milk and 3% BSA and labeled with streptavidin-peroxidase (1 g/ml, 1 h, 4°C). After washing three times with TBST, they were visualized by ECL. Densitometry measurements were as described for co-immunoprecipitation.

ICS Demonstrates Selective Increase in the Clustering of Internalization-competent HA Mutants at the Cell Surface-ICS
measures the spatial fluorescence intensity fluctuations arising from fluorescently labeled membrane components distributed across the surface of the cell and employs autocorrelation analysis to derive a quantitative measure of the density of independent fluorescent particles from which the degree of aggregation can be determined (Ref. 49; see "Experimental Procedures"). Proteins internalized via the clathrin-coated pit pathway have internalization signals that are thought to interact with AP-2 adaptors (5, 14 -22, 31, 35, 38) and thus lead to preferential localization in plasma membrane coated pits. The resulting change in surface distribution of such internalization-competent proteins is essentially a clustering phenomenon, which could be measured by ICS. To demonstrate the capability of ICS to measure association with coated pits at the cell surface, we compared the surface distribution and clustering of HA wt and several HA mutants that carry different cytoplasmic internalization signals and are internalized at highly different rates. CV-1 cells expressing different HA mutants (Table I) were fluorescently labeled in the cold by monovalent FabЈ fragments to avoid any possibility for IgG-mediated cross-linking that might affect the clustering. The labeled cells were incubated (10 min) at 22°C prior to fixation, because earlier work (43,44) has shown that interactions of HA mutants with coated pits were greatly reduced at lower temperatures; warming to 37°C was avoided due to the fast internalization of HAϩ8 at this temperature (42). The fixed cells were taken for ICS experiments. Typical confocal images of a protein that is excluded from coated pits (HA wt) and a mutant that is internalized at a moderate rate (HA-Y543; Refs. 41 and 42) are depicted in Fig.  1. Qualitative differences can be observed between the distribution of the two proteins at the cell surface, with HA-Y543 appearing more clustered (Fig. 1, compare A and B). This clustering disappears in cells preincubated under hypertonic conditions ( Fig. 1C; this effect is discussed later). To quantify these differences, images of flat 16.3 ϫ 16.3-m 2 areas on the cell surface were processed by autocorrelation analysis to determine the g(0,0) values, from which 1/g(0,0) (which gives a direct measure of N p ; see "Experimental Procedures") values were derived.
The frequency distributions of the 1/g(0,0) values measured in a number of independent experiments for each of the four HA proteins studied are shown in Fig. 2. Although it is apparent that the 1/g(0,0) distribution patterns and average values of the mutants that undergo coated pit-mediated internalization, HAϩ8 (Fig. 2C) and HA-Y543 (Fig. 2D), are different from  (Fig. 2, A and B), the frequency distributions of these values among the cells are fairly broad. This could be caused, at least in part, by cell-to-cell differences in the expression levels of the HA proteins at the cell surface. The reason for the dependence of 1/g(0,0) on the surface density of the labeled protein is that a higher density can increase N p (and thereby 1/g(0,0)) by elevating the total number of fluorescently labeled molecules in the observation area (49). To eliminate this dependence on HA surface density and measure exclusively alterations in the state of aggregation, we define a new parameter, termed the degree of aggregation (DA): where ͗i͘ is the average fluorescence intensity of the same image used to generate the specific autocorrelation function.
The proportionality constant ␣ depends on instrumental parameters and cancels out if relative DA values (ratios) are used. Because ͗i͘ is directly proportional to the average number of fluorescently labeled molecules in the area illuminated by the beam (N m ) and g(0,0) is equal to 1/N p , their product (DA) is proportional to N m /N p , the average number of fluorescently labeled molecules (in the current study, HA proteins) per particle. An increase in the surface density of labeled protein molecules with no changes in the aggregation state (i.e. the particle size is unaltered) will elevate N m and N p by the same factor, leaving DA constant. However, if more protein molecules are driven into aggregates, the average number of molecules per particle will increase, and the higher aggregation will lead to a higher DA value.
To internally calibrate the measurements for possible variations between separate experiments, each experiment included a "standard" sample of HA wt at 22°C. This sample (chosen due to HA wt being excluded from coated pits) was given the DA value of one, and the DA values of all the other samples were calibrated relative to this value. The results are depicted in Fig. 3. It is apparent that at 22°C, the internalization-competent HA mutants display significantly higher DA values. HAϩ8, which is internalized at a high rate and shows stable interactions with coated pits in lateral mobility studies (42,44), yields the highest DA value. HA-Y543, which is internalized at a moderate rate and appears to interact transiently with coated pits in lateral mobility experiments (42,43), shows an intermediate DA value; and HAϩ4, whose internalization is very slow and which does not interact appreciably with coated pits (42,44), has a DA value essentially identical to that of HA wt. The higher DA values of HAϩ8 and HA-Y543 suggest that they appear in aggregates containing, on the average, higher numbers of HA molecules than HA wt or HAϩ4. This phenomenon most likely reflects the tendency of HAϩ8 and HA-Y543 to cluster in coated pits (or bind to AP-2 aggregates) at the cell surface. This interpretation is strongly supported by the effects of disrupting the clathrin lattice structure on the DA values of the HA mutants and by the coprecipitation of ␣-adaptin with HAϩ8 and HA-Y543 (see Figs. 4A and 5). Interestingly, the increase in the DA values of HAϩ8 and HA-Y543 relative to HA wt was much weaker in the cold (Fig. 3), suggesting that the observed clustering and the interactions leading to it are temperature-dependent and are elevated at higher temperatures. This notion is supported by the significant reduction in the co-precipitation of ␣-adaptin with the internalizationcompetent mutants at 4°C (Fig. 5B). These findings are in accord with former observations based on lateral mobility studies (43,44), which demonstrated stronger mobility-restricting interactions (presumably with coated pits) at higher temperatures, and with the enhanced association of AP-2 with EGF receptors at 37°C (15,21).

Effects of Treatments That Disperse or Alter Coated Pit Structure on the Surface Distribution of HA Mutants-To obtain
further support for the notion that the increased DA of the internalization-competent HA proteins (HAϩ8 and HA-Y543) is due to their preferred localization in coated pits, we have tested the effects of treatments demonstrated previously to alter coated pit structure on the DA values of HAϩ8, HA-Y543, and HAϩ4 relative to HA wt. Two independent treatments that have drastic but different effects on the clathrin lattice structure were employed. The first was incubation in a hypertonic medium to disperse the clathrin lattices underlying coated pits (56,58). The second was acidification of the cytosol, which has been demonstrated to alter the morphology of the clathrin lattices, rendering them associated with the plasma membrane at a "frozen" state (56,58,60). Both treatments are highly effective in inhibiting coated pit-mediated internalization in CV-1 cells (42, 43).
The hypertonic treatment dispersed the clusters of the internalization-competent HA mutants (for example, see Fig. 1C).
The effects of this treatment on the clustered distribution of the HA proteins at 22°C as measured by ICS are depicted in Fig.  4A. The disruption of the clathrin lattices had a dramatic effect on the DA values of HAϩ8 and HA-Y543, which were reduced significantly as compared with untreated cells, becoming essentially equal to DA of HA wt on untreated cells. The decrease in the DA value is specific for the internalization-competent HA proteins and is not observed for HAϩ4 and HA wt, demonstrating that it is not due to a general change in cell shape or membrane properties. These results suggest that when the coated pits are dispersed, HA mutants that associate with them in untreated cells (HAϩ8 and HA-Y543) lose the clustered distribution on the cell surface and display degrees of aggregation similar to HA proteins that do not interact with coated pits to begin with.
The effects of cytosol acidification on the degree of aggregation of the HA proteins at the cell surface are shown in Fig. 4B. Unlike the hypertonic treatment, cytosol acidification does not disperse the clathrin lattices (58 -60). If the enhanced aggre- FIG. 3. DA values of HA؉4, HA-Y543, and HA؉8 relative to HA wt. CV-1 cells expressing various HAs were prepared and labeled as described under "Experimental Procedures" and in the legend to Fig. 1. After labeling, they were incubated for 10 min at either 22 or 4°C prior to fixation and ICS measurements. The DA values were calculated relative to the HA wt sample at 22°C, included as a standard in each independent experiment. Each bar is the mean Ϯ S.E. of five to seven independent experiments (for the 22°C data) or two experiments (4°C data), with 30 -40 cells measured in each experiment. The use of a t test for the means of two independent samples with different variance (Ref. 55; see "Experimental Procedures") indicates that at 22°C the differences between the pairs HAϩ8/HA wt and HA-Y543/HA wt are highly significant (p Ͻ 0.0005), whereas HAϩ4 is not significantly different from HA wt (p Ͻ 0.025). At 4°C, only HAϩ8 was significantly different from HA wt incubated at the same temperature (p Ͻ 0.005), and even this difference was only slightly above the 99% confidence level employed in the analysis of the ICS data. For the other pairs (HAϩ4/HA wt and HA-Y543/HA wt), the differences were not significant (p Ͻ 0.025 and p Ͻ 0.150, respectively).

FIG. 4. Effect of hypertonic treatment (A) or cytosol acidification (B) on the DA values of HA protein mutants.
CV-1 cells expressing the various HA proteins were subjected to hypertonic or to cytosol acidification treatments as described under "Experimental Procedures." After fluorescent labeling with FabЈ fragments in the appropriate medium (hypertonic medium or KA buffer, respectively), they were incubated 10 min at 22°C in the same medium, fixed, and taken for the ICS measurements. The DA values were calculated relative to untreated HA wt at 22°C, which was included as a standard in each experiment. Each bar is the mean Ϯ S.E. of several independent experiments (two for each HA mutant in A and for HAϩ4 in B; three for all other HAs in B), with 30 -40 cells measured in each. Data reproducibility between experiments is high, as demonstrated by the similarity between the DA values of the untreated cells in the current experiments and those from the different set of independent experiments shown in Fig. 3. Statistical analysis of the data in A (t test as in Fig. 3) indicates that the reduction in DA of hypertonically treated HAϩ8 and HA-Y543 (each relative to the same protein in untreated cells) is highly significant (p Ͻ 0.0005), whereas the DA values of HAϩ4 in treated and untreated cells are essentially similar (p Ͻ 0.4). The DA of HA wt shows a minor elevation following hypertonic treatment; this increase is marginally significant statistically (p Ͻ 0.005) but is clearly very small. A similar analysis of the data in B indicates that cytosol acidification does not significantly alter the DA values of HAϩ4, HA-Y543, and HAϩ8 (p Ͻ 0.2, p Ͻ 0.2, and p Ͻ 0.4, respectively). Only the change in DA of HA wt is significant (p Ͻ 0.0005).
gation of HAϩ8 and HA-Y543 is indeed due to preferential clustering in coated pits, it is expected to persist (possibly with some modifications) following cytosol acidification. Indeed, the DA values of both HAϩ8 and HA-Y543 continue to be higher than those of HA wt and HAϩ4 following this treatment. These findings are in accord with our former lateral mobility measurements, which suggested that the interactions of internalization-competent HA mutants with coated pits are not disrupted and in some cases are enhanced following cytosol acidification (43,44). The DA of HA wt, which is excluded from coated pits, increased somewhat following cytosol acidification. The reason for this elevation is unknown, and it could indicate an increase in its aggregation (unrelated to coated pits) under these conditions.
Co-immunoprecipitation of ␣-Adaptin with HA Mutants Carrying Cytoplasmic Internalization Signals-The results described above suggest that the clustering of internalizationcompetent HA mutants at the cell surface is due to their preferred association with coated pits. If this were the case, these mutants should interact with AP-2 adaptor complexes. To examine this issue, we measured the extent of coprecipitation of ␣-adaptin (which is specific to the plasma membrane AP-2) with the various HA proteins. After detergent solubilization and immunoprecipitation of the HA proteins, the cell extracts were analyzed for ␣-adaptin by SDS-polyacrylamide gel electrophoresis and immunoblotting (Fig. 5), calibrating the samples for similar levels of HA proteins at the cell surface by a surface biotinylation assay (based on HA immunoprecipitation followed by blotting with streptavidin; see "Experimental Procedures"). The calibration was performed according to the level of HA expression at the cell surface, because the percentage of the internalization-competent mutants (especially HAϩ8) at the cell surface is significantly lower (42). Because clathrin and AP-2 appear to dissociate rapidly from endocytic vesicles and remain soluble in the cytoplasm until being recruited again into coated pits (46, 63-65), one should compare the different mutants based on their level at the plasma membrane rather than on their total expression level. As can be seen in Fig. 5A, when the immunoprecipitation was performed after a 10-min incubation at 22°C, a high amount of ␣-adaptin co-precipitated with HAϩ8, and a moderate amount accompanied HA-Y543. Only trace amounts of ␣-adaptin co-precipitated with HA wt and HAϩ4. The co-precipitation of ␣-adaptin with HAϩ8 and HA-Y543 was significantly reduced when the cells were preincubated at 4°C, and the 10-min incubation at 22°C was omitted (Fig. 5B). These results are in full agreement with the ICS experiments, and the rank order for association with ␣-adaptin in the coprecipitation studies is similar to that of the clustering at the cell surface in the ICS studies.
The effects of hypertonic and cytoplasmic acidification treatments on the co-precipitation of ␣-adaptin with the HA mutants are shown in Fig. 6. Dispersal of the clathrin lattices by hypertonic treatment markedly reduced the co-precipitation of ␣-adaptin with HAϩ8 or HA-Y543, in correlation with the loss of their clustered distribution on the cell surface (Fig. 4A) and the increase in their lateral mobility parameters (43,44) following this treatment. Thus, in the absence of clathrin lattices, the internalization-competent HA mutants are not limited to specific microdomains at the cell surface and can disperse laterally throughout the plasma membrane. On the other hand, and again in accord with the ICS (Fig. 4B) and lateral mobility results (43,44), the co-precipitation of ␣-adaptin with HAϩ8 or HA-Y543 was not reduced following cytosol acidification (which "freezes" but does not disperse clathrin lattices), and in the case of HA-Y543 it was even enhanced (Fig. 6B). DISCUSSION The first step in receptor-mediated endocytosis involves the recruitment of membrane proteins destined for internalization into coated pits. Their preferred localization in these domains should alter their surface distribution due to clustering. This  A and B), prior to immunoprecipitation with anti-HA as described under "Experimental Procedures." The level of each HA mutant at the cell surface was determined by surface biotinylation and immunoprecipitation (see "Experimental Procedures") with anti-HA (C and D) and quantified by densitometry. Based on this calibration, the lanes in A and B were loaded with equal amounts of each HA mutant at the cell surface. A, immunoprecipitation with anti-HA, blotting with AC1-M11 (anti ␣-adaptin) followed by protein A-peroxidase. The experiment is representative of four carried out. Densitometric analysis showed that the amount of ␣-adaptin co-precipitated with HA-Y543 at 22°C was 16% of that precipitated with HAϩ8. Essentially no coprecipitation was detected with HAϩ4 or HA wt in this experiment; in some cases (e.g. see Fig. 6) trace amounts of ␣-adaptin were precipitated with these proteins (up to 5% relative to the level precipitated with HAϩ8). B, immunoprecipitation and blotting were as in A. The pairs of 4 and 22°C were taken from the same infected batch of cells. The amounts of ␣-adaptin co-precipitated at 4°C were lower than for the same mutant at 22°C (18% for HAϩ8 and 30% for HA-Y543). C and D, immunoprecipitation with anti-HA and blotting with streptavidin-peroxidase.
FIG . 6. Effect of hypertonic treatment (A) and cytosol acidification (B) on the co-precipitation of ␣-adaptin with the HA protein mutants. Cells expressing specific HA mutants were subjected to hypertonic or cytosol acidification treatments (described under "Experimental Procedures"). They were incubated in the cold followed by 10 min at 22°C in the hypertonic medium or in the KA buffer as in Fig. 4, prior to detergent solubilization and immunoprecipitation with anti-HA (see "Experimental Procedures"). Using surface biotinylation and immunoprecipitation to determine the level of each HA mutant at the cell surface as described in the legend to Fig. 5 ("Experimental Procedures"), the amounts loaded on each lane were calibrated to include similar levels of cell surface HA proteins. The arrow indicates the doublet of ␣-adaptin. The results are representative of three experiments for each treatment. The pairs of treated (ϩ) and untreated (Ϫ) cells were taken from the same infected batch of cells. A, hypertonic treatment. The amounts of ␣-adaptin co-precipitated with the internalization-competent mutants following the treatment (relative to the same untreated mutant) were 30 and 16% for HAϩ8 and HA-Y543, respectively. The amount of ␣-adaptin precipitated with HAϩ4 and HA wt was very low both before and after treatment. B, cytosol acidification. The treatment did not significantly affect the amount of ␣-adaptin co-precipitated with HAϩ8 (90% relative to the untreated mutant) and mediated an increase in the co-precipitation with HA-Y543 (200%). No significant co-precipitation was observed with HAϩ4 and HA wt. can be measured directly and quantitatively by ICS, employed here to directly assess the clustering in coated pits of a series of influenza HA mutants carrying different cytoplasmic internalization signals. The fact that HA wt is devoid of any such signal provides an internal control of a protein that is excluded from coated pits and enables a direct comparison between specific internalization signals of variable strength.
The ICS data on untreated cells (Fig. 3) demonstrate a striking dependence of the degree of aggregation on the internalization signal introduced at the HA cytoplasmic tail. The rank order of the DA values at 22°C was similar to that of the internalization rates of the HA mutants (42), with HAϩ8 (which undergoes fast endocytosis) displaying a higher DA value than the moderately internalized HA-Y543, whereas HAϩ4 and HA wt (whose internalization is very slow) exhibit lower and equal DA values. These results are fully compatible with the notion that HAs bearing internalization signals cluster in specific regions on the cell surface (most likely coated pits) and that the level of clustering depends on the strength of the interactions with these domains. The identification of these domains as coated pits is supported by several observations: (a) the correlation between the DA values at 22°C (Fig. 3) and the internalization rates of the HA mutants (42); (b) the reduction in DA of HAϩ8 and HA-Y543 following hypertonic treatment to the level measured for the noninteracting HA wt and HAϩ4 (Fig. 4A); (c) the persistence of the selective clustering of the internalization-competent mutants under conditions that freeze the coated pits at the plasma membrane (cytosol acidification; Fig. 4B); (d) the selective co-precipitation of ␣-adaptin with the internalization-competent HAs at the same rank order of potency found for the DA values (Fig. 5); and (e) the specific effects of treatments that affect the structure of the coated pits on the above co-precipitation (Fig. 6).
The co-precipitation experiments on untreated cells (Fig. 5) provide direct evidence that the HA mutants bearing internalization signals interact with AP-2 adaptors. To our knowledge, such interactions were thus far demonstrated by co-immunoprecipitation from whole cells only for EGF receptor family members (14,15,20,21,38). The co-precipitation data (Fig. 5) are in accord with surface plasmon resonance studies on the binding of purified adaptors to immobilized peptides corresponding to the cytoplasmic tails of HA wt and HA-Y543, where significant binding was observed only for the latter (35). Although the interactions might also be with AP-2 complexes that are not associated with coated pits, as proposed for the EGF receptor (21), the dissolution of the clusters of the internalization-competent HAs following hypertonic treatment argues against this possibility. Under such conditions, AP-2 aggregates devoid of clathrin remain at the plasma membrane (58), but the selective clustering of the HA mutants disappears concomitantly with the dispersal of the clathrin lattices. The weaker interactions with AP-2 upon dispersal of the clathrin lattices differ from the reports that K ϩ depletion, which is analogous to the hypertonic treatment in dispersing the clathrin lattices underlying coated pits (56,58,66), enhanced the association of EGF receptors with AP-2 (15,21). The latter effect may therefore be specific to EGF family receptors, whose endocytosis may involve several signals and multiple pathways, as suggested by the ability of a mutant EGF receptor that lacks AP-2 binding to undergo internalization (39) and by the involvement of the GRB2 protein in EGF receptor endocytosis in association with dynamin (67). This notion is in line with the report on specific requirements for the recruitment of EGF receptors (versus transferrin receptors) into coated pits (68).
The interactions of the internalization-competent HAs with coated pits appear to be temperature-dependent, becoming weaker in the cold. This is evident for the clustering of HAϩ8 and HA-Y543 as measured by ICS (Fig. 3), which is dramatically reduced at 4°C, and for their reduced association with ␣-adaptin in the co-precipitation experiments (Fig. 5). This conclusion is in accord with the loss of the constraints on the lateral mobility of these HA mutants in the cold (43,44). Lower interactions with AP-2 in the cold were also reported for the EGF receptors, where warming to 37°C was required for efficient co-precipitation (15,21). The notion that low temperature affects the association of the internalization signal with AP-2 rather than the coated pit structure is supported by reports that clathrin-coated pits persist at 4°C (69,70) and by the similar DA values obtained for ␣-adaptin distribution at 37 and 4°C. 2 The ICS measurements described here and the lateral mobility approach to measure interactions with coated pits (43,44) measure different parameters and are complementary in many respects. The ICS data are in excellent correlation with lateral mobility studies on the same HA mutants (43,44). Specifically, the lateral mobility data (44) demonstrate that 20 -25% of the HAϩ8 population is stably entrapped in coated pits, whereas the remainder are free to diffuse at the cell surface. The ICS data show that HAϩ8, which has the strongest internalization signal, also yields the highest DA value (4-fold higher than HA wt; Fig. 3). From the lateral mobility data it is apparent that this average DA value consists of contributions from both the HAϩ8 molecules entrapped in coated pits (i.e. reside in clusters) and the 75-80% that are dispersed throughout the membrane. Therefore, the actual increase in HAϩ8 cluster size relative to HA wt is most likely significantly higher than the 4-fold elevation in the average DA value. HA-Y543, which has a weaker internalization signal, exhibits only a moderate increase in the DA value (Fig. 3), in accord with the lateral mobility studies (43) which demonstrated that HA-Y543 interacts transiently with coated pits. This in turn suggests that in the dynamic equilibrium characterizing HA-Y543 association with coated pits, the number of HA-Y543 molecules that reside at any given time in clusters is fewer than in the case of HAϩ8. The HAϩ4 mutant, whose internalization signal is truncated in half and which is internalized very slowly, has no detectable interactions with coated pits by either method (Fig. 3 and Ref. 44). The correlation also holds for treatments that affect the structure of coated pits; both methods indicate that the interactions with coated pits are disrupted by the hypertonic treatment that disperses the clathrin lattices, but not by the cytosol acidification which freezes them underneath the plasma membrane ( Fig. 4 and Refs. 43 and 44).
The ICS results provide insight into the factors that may affect the lateral mobility experiments. In the latter studies, reduction in the mobile fraction or in the lateral diffusion rate may arise from any of several types of interactions (71)(72)(73). The present correlation between reduced lateral mobility and enhanced clustering indicates that the internalization-competent HA mutants are trapped in coated pits and that this entrapment is the cause for their reduced mobility at the cell surface (43,44).
In summary, the studies presented here demonstrate directly and quantitatively the clustering of internalization-competent membrane proteins in coated pits at the cell surface. In combination with complementary approaches, these studies provide a deeper insight into the interactions that lead to the sorting of membrane receptors into coated pits. The ICS approach presented can be extended to explore various processes involving the aggregation and interactions of plasma membrane proteins, such as determining the extent to which signaling molecules at the cell surface are clustered in caveoli or in glycolipid-rich domains.