|Home | About | Journals | Submit | Contact Us | Français|
Sorting of transmembrane cargo into clathrin-coated vesicles requires endocytic adaptors, yet RNA interference (RNAi)-mediated gene silencing of the AP-2 adaptor complex only disrupts internalization of a subset of clathrin-dependent cargo. This suggests alternate clathrin-associated sorting proteins participate in cargo capture at the cell surface, and a provocative recent proposal is that discrete endocytic cargo are sorted into compositionally and functionally distinct clathrin coats. We show here that the FXNPXY-type internalization signal within cytosolic domain of the LDL receptor is recognized redundantly by two phosphotyrosine-binding domain proteins, Dab2 and ARH; diminishing both proteins by RNAi leads to conspicuous LDL receptor accumulation at the cell surface. AP-2–dependent uptake of transferrin ensues relatively normally in the absence of Dab2 and ARH, clearly revealing delegation of sorting operations at the bud site. AP-2, Dab2, ARH, transferrin, and LDL receptors are all present within the vast majority of clathrin structures at the surface, challenging the general existence of specialized clathrin coats for segregated internalization of constitutively internalized cargo. However, Dab2 expression is exceptionally low in hepatocytes, likely accounting for the pathological hypercholesterolemia that accompanies ARH loss.
When the clathrin coat was first discovered in mosquito oocytes (Roth and Porter, 1964 ), it was astutely speculated that the “bristle” coat provided cargo-selective properties that govern the process of yolk internalization. Mosquito yolk precursor receptors, the vitellogenin and lipophorin receptors, are members of the low-density lipoprotein (LDL) receptor superfamily. The first endocytic sorting signal was identified in this superfamily, when it was discovered that the 802FDNPVY sequence within the carboxy-terminal cytosolic domain of human LDL receptor promotes rapid clathrin-mediated uptake of LDL (Davis et al., 1986 ; Chen et al., 1990 ). Other internalization signals have subsequently been identified, including the widespread YXXØ motif, found in transferrin and mannose 6-phosphate receptors, the [DE]XXXL[LI]-type dileucine motif, and reversible poly/multiubiquitination (Bonifacino and Traub, 2003 ).
To link these sorting signals to an assembling clathrin lattice, adaptor proteins are required because clathrin triskelia, the protomers of the characteristic polyhedral coat, contain no direct membrane-binding information. The first endocytic adaptor characterized was AP-2, a multifunctional heterotetramer comprised of a core of small ς2, medium μ2, and large α and β2 subunits. An independently folded appendage projects off each large subunit, connected to the core by a flexible polypeptide hinge (Owen et al., 2004 ). The β2 subunit hinge and appendage bind physically to triskelia, allowing AP-2 to couple clathrin to the membrane, because the adaptor core binds phosphatidylinositol 4,5-bisphosphate (PtdIns(4,5)P2) directly, and also interacts with YXXØ and [DE]XXXL[LI] motifs (Ohno et al., 1995 ; Owen et al., 2004 ; Höning et al., 2005 ). Accordingly, AP-2 has a well-accepted role as the major cargo-selective component for endocytic coated vesicles. Nevertheless, distinct internalization signals neither saturate internalization at the same surface density nor compete directly with one another (Marks et al., 1996 ; Santini et al., 1998 ; Warren et al., 1998 ), which is not expected if AP-2 performs all cargo-selective steps, and AP-2 does not readily bind to alternative internalization signals, including polyubiquitin and the FXNPXY sequence. Moreover, when small interfering RNA (siRNA) is used to ablate AP-2, transferrin uptake halts but cargo utilizing alternative endocytic signals, such as the LDL receptor, still internalize efficiently (Hinrichsen et al., 2003 ; Motley et al., 2003 ).
Endocytic clathrin coats must therefore contain alternative sorting adaptors. These clathrin-associated sorting proteins (CLASPs) should synchronously bind clathrin, the plasma membrane, cargo, and other lattice assembly proteins, such as AP-2 (Traub, 2003 ). Disabled-2 (Dab2) and the autosomal recessive hypercholesterolemia (ARH) protein are potential FXNPXY-signal–sorting CLASPs; both contain an aminoterminal phosphotyrosine-binding (PTB) domain that simultaneously interacts with nonphosphorylated FXNPXY motifs and PtdIns(4,5)P2, along with an unstructured carboxy-terminal segment that contains tandemly arrayed clathrin- and AP-2–binding information (Morris and Cooper, 2001 ; He et al., 2002 ; Mishra et al., 2002a , 2002b ; Yun et al., 2003 ). Huge overexpression of a tandem Dab2 PTB domain construct selectively disrupts internalization of LDL receptor (Mishra et al., 2002a ) and the type-2 apolipoprotein E receptor (Cuitino et al., 2005 ). However, this does not unambiguously prove the involvement of Dab2 in clathrin-mediated endocytosis; rather it shows a binary interaction between the PTB domain and the FXNPXY sorting signal. That Dab2-nullizygous mice exhibit a (mild) proteinuria, indicative of megalin (a hepatic scavenger receptor of the LDL receptor superfamily) dysfunction in the renal proximal tubule, supports a role for Dab2 in endocytosis. There are two major splice isoforms of Dab2, so-called p96 (full length) and p67, with the central AP-2 and clathrin binding region spliced out (Xu et al., 1995 ). The poor ability of the p67 isoform to promote megalin-dependent transferrin uptake in developing embryos (Maurer and Cooper, 2005 ) also argues for Dab2 coupling cargo selection to clathrin coat assembly. Yet, conditionally-null Dab2−/− mice are surprisingly viable and fertile (Morris et al., 2002 ). Conversely, targeted ARH gene disruption in mice, just as in ARH patients (Zuliani et al., 1999 ; Garcia et al., 2001 ; Arca et al., 2002 ; Naoumova et al., 2004 ), leads to elevated circulating LDL and LDL internalization defects in the liver (Jones et al., 2003 ), but there is no evidence of proteinuria. These experiments hint at possible functional overlap between Dab2 and ARH activity in decoding FXNPXY sorting signals, but also uncover a seemingly complex interrelationship of the two proteins in vivo.
In this study, we show that Dab2 and ARH sort the LDL receptor in a functionally redundant manner; knocking down both proteins with siRNA causes extensive accumulation of LDL receptors at the cell surface. Assigning the CLASPs responsible for FXNPXY signal recognition is significant because LDL receptor superfamily members regulate numerous cellular processes; for instance, megalin appears to participate in steroid hormone action by promoting internalization of sex hormone–binding globulin complexes (Hammes et al., 2005 ). In Drosophila, where the FXNPXY signal is conserved, lipoproteins play an important role transporting lipid-linked morphogens, such as Wingless and Hedgehog, and contribute to establishment of appropriate morphogen gradients during development (Panakova et al., 2005 ), and this process might be conserved in chordates. In addition, clear appreciation of how Dab2 and ARH orchestrate FXNPXY sorting allows assessment of whether these CLASPs and cognate cargo are confined to functionally discrete subsets of clathrin structures at the plasma membrane. Despite recent propositions (Cao et al., 1998 ; Tosoni et al., 2005 ; Lakadamyali et al., 2006 ), we do not find evidence of major compositional heterogeneity within clathrin coats at the surface of nonpolarized cells, questioning the general existence of specialized vesicles dedicated to internalization of select subsets of cargo.
YFP-Dab2 was generated by first cloning EYFP into pECFP-N1 at the NheI and BglII sites, followed by replacing the in-frame CFP with full-length murine Dab2, inserted between the SalI and NotI sites. The plasmid was made RNA interference (RNAi)-resistant by introducing silent mutations (S249, L250, and R251) using QuikChange mutagenesis (Stratagene, La Jolla, CA). The GFP-epsin was created by cloning rat epsin 1 into pEGFP-C1 between the XhoI and KpnI sites, and the ARH-GFP was previously described (Mishra et al., 2005 ). Appropriate mutations were introduced into either YFP-Dab2 or ARH-GFP using QuikChange mutagenesis with the required primers, the sequences of which are available on request. The siRNA oligonucleotides used, which target base pairs 1384–1402 of the AP-2 β2-subunit, base pairs 472–490 of dynamin 2, base pairs 1325–1343 of epsin 1, and base pairs 3313–3331 of the clathrin heavy chain (HC) nucleotide sequence, have been described (Hinrichsen et al., 2003 ; Huang et al., 2004 ). The Dab2 siRNA selected targets base pairs 745–763 of the Dab2 nucleotide sequence (UUCUUUAAGAGAAAAUCCA), the ARH siRNA targets base pairs 633–651 of the ARH nucleotide sequence (CCUGCUGGACUUAGAGGAG; Qiagen, Valencia, CA or Dharmacon, Lafayette, CO), and the β1 siRNA oligo targets base pairs 1529–1547 of the human β1 subunit nucleotide sequence (CCACUCAGGACUCAGAUAA; Invitrogen, Carlsbad, CA or Dharmacon). The ARH-specific siGENOME SMARTpool (M-013025–00) was also obtained from Dharmacon, as was the AP-2 α-subunit siRNA oligo that is essentially that previously described (Hinrichsen et al., 2003 ) directed against base pairs 1052–1070 of the α subunit, although the final base pair of this sequence was omitted, yielding the sequence GCAUGUGCACGCUGGCCA. All siRNA oligos were synthesized with dTdT overhangs and directed against human sequences, unless otherwise noted.
The monoclonal antibodies (mAbs) against the clathrin HC TD.1 (Nathke et al., 1992 ) and X22 (Brodsky, 1985 ); the anti-AP-1/2 β1/β2-subunit mAb 100/1 (Ahle et al., 1988 ) and affinity-purified GD/1 antibodies (Traub et al., 1995 ); the anti-AP-1 μ1-subunit antibody RY/1 (Traub et al., 1996 ); the AP-2 α-subunit mAb AP.6 (Chin et al., 1989 ), and affinity-purified anti-epsin 1 polyclonal (Drake et al., 2000 ) have been described. Affinity-purified anti-eps15 polyclonal and anti-clathrin light chain polyclonal R461 were generous gifts from Ernst Ungewickell (Medizinische Hochschule Hannover, Hannover, Germany) and rabbit anti-AP-2 μ2-subunit serum was kindly provided by Juan Bonifacino (NIH, Bethesda, MD). Affinity-purified anti-LRP1 antibodies were kindly provided by Guojun Bu (Washington University School of Medicine, St. Louis, MO), the rat-isoform specific anti-LDL receptor polyclonal antibody was a kind gift from Gene Ness (University of South Florida, Tampa, FL), and the anti-α-mannosidase II antiserum was kindly provided by Kelley Moremen (University of Georgia, Athens, GA). A second mAb (clone 8) against the AP-2 α-subunit was purchased from BD Transduction Laboratories (San Jose, CA), and the anti-CD71/transferrin receptor mAb RVS-10 was from Chemicon (Temecula, CA). A goat anti-Dab2 polyclonal antibody was purchased from Santa Cruz Biotechnology (Santa Cruz, CA). The hybridoma secreting the anti-LDL receptor mAb IgG-C7 was obtained from the ATCC (Manassas, VA). The rodent-specific anti-asialoglycoprotein receptor mAb 8D7 was from HyCult Biotechnologies (Uden, The Netherlands). The anti-tubulin mAb E7 was purchased from the Developmental Studies Hybridoma Bank (Iowa City, IA). The affinity-purified ARH antibody (Mishra et al., 2002b ) was further cross-adsorbed with fixed and permeabilized ARH−/− fibroblasts, followed by removal of Hsc70 cross-reactivity by incubation with GST-DnaK bound to glutathione-Sepharose. The affinity-purified Dab2 (Mishra et al., 2002a ) antibody was similarly cross-adsorbed with immobilized GST-DnaK. Goat anti-mouse or anti-rabbit secondary antibodies conjugated to either AlexaFluor488 or AlexaFluor568 were purchased from Invitrogen, whereas goat anti-mouse or anti-rabbit antibodies conjugated to Cy5, HRP-conjugated donkey anti-mouse and anti-rabbit secondary antibodies, and colloidal gold-conjugated goat anti-rabbit secondary antibodies were purchased from Jackson ImmunoResearch (West Grove, PA). HRP-conjugated donkey anti-goat secondary antibody was from Santa Cruz.
All cell lines were cultured at 37°C in a humidified atmosphere containing 5% CO2. MCF-7 and MDA-MB-231 cells were grown in RPMI supplemented with 10% fetal calf serum (FCS; HyClone, Logan, UT) and 2 mM l-glutamine (Invitrogen), whereas primary control (GM01386) and ARH−/− (GM06697) fibroblasts and BS-C-1 cells were cultured in EMEM supplemented with 10% FCS, 2 mM l-glutamine and 1× nonessential amino acids (Mediatech, Herndon, VA). HeLa SS6 and HepG2 cells were cultured in DMEM supplemented with 10% FCS and 2 mM l-glutamine. BS-C-1 cells stably transfected with either the ς2-GFP or clathrin LCa-GFP were a generous gift from Tomas Kirchhausen (Harvard Medical School, Boston, MA) and cultured in DMEM supplemented with 10% FCS, 2 mM l-glutamine and 0.4 mg/ml G418 (Sigma, St. Louis, MO). For HeLa SS6 cell RNAi, the cells were passaged to yield 50–70% confluence in 24-well plates on the day of transfection. Transfections were performed with Transit-TKO reagent (Mirus Bio, Madison, WI) according to manufacturer’s protocol using 100 nM of each siRNA. The culture media was replaced with fresh medium the following day, and cells were prepared for analysis 48–50 h after transfection with siRNA duplexes. For rescue experiments, 100 ng of rescue plasmid DNA was included along with the siRNA oligos at the time of transfection, and did not block RNAi-mediated knockdown of the proteins. Transfection of primary fibroblasts was performed by electroporating 2 × 106 cells using a Nucleofector II (Amaxa, Gaithersburg, MD), primary fibroblast electroporation solution (Amaxa), and 1 μM siRNA oligo according to the manufacturer’s protocol with program U-023. Cells were passaged the following day and harvested ~72 h after electroporation. Plasmid transfections were performed as previously described (Keyel et al., 2004 ).
Cells were prepared for immunofluorescence essentially as previously described (Mishra et al., 2005 ). For experiments involving LDL receptors, the cells were grown in DMEM (for HeLa and BS-C-1 cells) or EMEM (for fibroblasts) supplemented with 10% lipoprotein-deficient serum (LPDS, Cocalico Biologicals, Reamstown, PA), 2 mM l-glutamine for 24–48 h before to up-regulate LDL receptors, whereas for experiments involving transferrin, cells were transferred to DMEM supplemented with 25 mM HEPES, pH 7.2, 0.5% BSA (starvation medium) for 1 h before experiments. To measure surface levels of receptors, cells were incubated with either the anti-LDL receptor mAb IgG-C7, anti-transferrin receptor mAb RVS-10, 4 μg/ml DiI-LDL (Biomedical Technologies, Stoughton, MA), 25 μg/ml transferrin conjugated to AlexaFluor488 (Tfn488; Invitrogen), or 50 μg/ml transferrin conjugated to AlexaFluor633 (Tfn633; Invitrogen) in starvation medium for 1 h on ice at 4°C, fixed, and processed for immunofluorescence. To measure the LDL and transferrin receptor internalization for RNAi and rescue experiments, after 1 h in starvation medium, cells were incubated in the continuous presence of either 4 μg/ml DiI-LDL alone, 4 μg/ml DiI-LDL with 50 μg/ml Tfn633, or 25 μg/ml Tfn conjugated to AlexaFluor568 (Tfn568) with mAb IgG-C7 for 15 min at 37°C before fixation and processing for immunofluorescence. To measure LDL receptor colocalization simultaneously with Dab2 and ARH, HeLa cells were transfected with YFP-Dab2, cultured in DMEM containing 10% LPDS and 2 mM l-glutamine for ~24 h, incubated with mAb IgG-C7 for 1 or 5 min at 37°C, permeabilized on ice with 0.3% saponin, 25 mM HEPES-KOH, pH 7.2, 125 mM potassium acetate, 5 mM magnesium acetate, 2 mM EDTA, 2 mM EGTA for 1 min, fixed, and processed for immunofluorescence.
Images were acquired either on an Olympus Fluoview 500 or Olympus 1000 confocal microscope (Melville, NY) by sequentially scanning by line with an argon laser for 488-nm excitation, and two helium-neon lasers for 543- and 633-nm excitation through a PlanApomat 60 × 1.40 NA objective with a 488/543/633 dichroic. Emitted light was separated with beam splitters SDM560 and SDM640 and passed through either bandpass filters BA505–525 (green), BA560–600 (red) and longpass filter BA660IF (cy5) on the Fluoview 500 or diffraction gratings that pass 500–530-nm (green) or 555–625-nm (red) light or a BA650IF filter (cy5) on the Fluoview 1000 before detection. Linescan analysis of Dab2 and surface LDL receptor fluorescence along arbitrary 76-pixel lines using the average integrated intensity of three perpendicular pixels for each point along the line was measured in Metamorph (Molecular Devices, Downingtown, PA). Quantitation of the colocalization between various clathrin coat components was performed as described (Hawryluk et al., 2006 ). Quantitation of the rescue of the RNAi phenotype was measured by counting the number of transfected cells showing either normal, impaired, or no LDL receptor internalization and displaying each as a percentage of the total cells counted. A small portion of plasmid-transfected cells displayed an overexpression phenotype (Mishra et al., 2005 ); these were excluded from the analysis. For fluorescence recovery after photobleaching (FRAP) experiments, HeLa cells grown on chambered glass coverslips (Nalge Nunc International, Rochester, NY) were transfected with either YFP-Dab2 or ARH-GFP and imaged every 5.2 s with the argon 488-nm laser line on the Fluoview 1000. A 405-nm laser diode was used at maximum intensity for 5 s to bleach YFP-Dab2 or ARH-GFP fluorescence during the acquisition. The intensity of each bleached region was measured using Metamorph, corrected for background, and each plane normalized to the intensity of an unbleached cell to account for any fluctuations in focus or intensity and then normalized to the average intensity of the frames before the bleach. To follow LDL and transferrin internalization in the context of the knockdown, RNAi-treated, serum-starved HeLa cells were incubated with 25 μg/ml Tfn488 and 4 μg/ml DiI-LDL and imaged every 3–6 s. To visualize LDL internalization in real time, HeLa cells in chambered glass coverslips transfected with YFP-Dab2 were imaged every 3–4 s in starvation medium. During image acquisition, starvation medium containing DiI-LDL was added such that the final concentration of DiI-LDL was 4 μg/ml. Measurement of DiI-LDL and YFP-Dab2 intensity in Dab2 structures that accumulated LDL within 250 s of LDL addition was performed using Metamorph; 17 structures were measured and averaged. For all live-cell experiments, at least five separate cells were imaged, and experiments were repeated at least twice.
Cells cultured in LPDS were surface labeled with 50 μg/ml LDL (Biomedical Technologies) or not, before preparing “unroofed” cell cortices for rapid-freeze deep-etch electron microscopy (EM) essentially as previously described (Heuser, 2000 ; Edeling et al., 2006 ). Briefly, cells grown on small oriented, carbon-coated coverslips were disrupted by sonication directly in 2% paraformaldehyde, 0.025% glutaraldehyde in 30 mM HEPES-KOH, pH 7.3, 70 mM potassium chloride, 5 mM magnesium chloride, and 3 mM EGTA (KHMgE buffer). After washing, preparations were quenched with 50 mM ammonium chloride, 50 mM l-lysine in KHMgE and blocked with 1% BSA in KHMgE, followed by incubation with affinity-purified anti-Dab2 or anti-ARH antibodies and then 15-nm colloidal gold-conjugated anti-rabbit antibody. Finally, the membranes were fixed in 2% glutaraldehyde before freezing. For surface labeling, washed cells were fixed directly in 2% glutaraldehyde without sonication.
Rat liver Golgi membranes (Tabas and Kornfeld, 1979 ) and plasma membrane sheets (Hubbard et al., 1983 ) were prepared using published protocols. Freshly isolated rat hepatocyte and total rat liver nonhepatocyte cell fractions were kind gifts from William Bowen, Jr. For Western blots, cells were trypsinized, washed in PBS, lysed directly in boiling SDS-sample buffer, heated to 95°C for 5 min, sonicated to shear DNA, and centrifuged at 12,000 × gmax for 2 min to pellet insoluble material. Protein levels were standardized using a Coomassie blue–based filter paper binding assay (Minamide and Bamburg, 1990 ); 25 μg was loaded per lane. Rat liver samples were prepared by finely mincing the tissue with a razor blade at 0°C, homogenizing in a small volume 10 mM HEPES-KOH, pH 7.2, 0.3 M sucrose, and Complete protease inhibitor (Roche, Indianapolis, IN) with a Teflon homogenizer, and solubilizing directly in boiling SDS-sample buffer or in nonreducing SDS-sample buffer at 37°C. After incubation at either 95 or 37°C, respectively, for 5 min, the samples were sonicated and centrifuged at 12,000 × gmax for 2 min to pellet insoluble material.
Sedimentation binding assays utilizing either multilamellar control liposomes composed of 40% phosphatidylcholine, 40% phosphatidylethanolamine, 10% phosphatidylserine, and 10% cholesterol or PtdIns(4,5)P2-containing liposomes composed of 35% phosphatidylcholine, 35% phosphatidylethanolamine, 10% phosphatidylserine, 10% cholesterol, and 10% PtdIns(4,5)P2 were performed as described previously (Mishra et al., 2002a , 2002b ).
Samples were resolved on polyacrylamide gels prepared with an altered acrylamide:bis-acrylamide (30:0.4) ratio stock solution. The decreased cross-linking generally improves resolution but also affects the relative mobility of several proteins, most noticeably epsin 1. After SDS-PAGE, proteins were either stained with Coomassie blue or transferred to nitrocellulose in ice-cold 15.6 mM Tris, 120 mM glycine. Blots were usually blocked overnight in 5% skim milk in 10 mM Tris-HCl, pH 7.8, 150 mM NaCl, 0.1% Tween 20, and then portions were incubated with primary antibodies as indicated in the individual figure legends. After incubation with HRP-conjugated anti-mouse or anti-rabbit IgG, immunoreactive bands were visualized with enhanced chemiluminescence. Alternatively, donkey anti-rabbit IgG was iodinated with ICl utilizing an established method (Breitfeld et al., 1989 ) and used at 200,000 cpm/ml to detect the bands. Quantitation of the signal from blots was performed using a Phosphor-Imager (Bio-Rad, Hercules, CA).
Cultured fibroblasts derived from ARH patients do not display major defects in LDL uptake, unlike LDL receptor-defective fibroblasts from individuals with familial hypercholesterolemia (Garcia et al., 2001 ; Eden et al., 2002 ). Because ARH−/− fibroblasts do express Dab2 (Mishra et al., 2002b ), the common domain/sequence signatures and functional analogy between ARH and Dab2 suggest that Dab2 might compensate for the absence of ARH in these cells. Indeed, Dab2 is positioned in small puncta scattered apparently randomly over the surface of ARH-null cells, most of which colocalize with both AP-2 and the LDL receptor (Figure 1, A and B). Transiently knocking down Dab2 levels by transfection with siRNA oligonucleotides (Figure 1C) strongly diminishes Dab2 staining, but the AP-2 distribution is unaltered (Figure 1D). In ARH−/− cells with depressed Dab2 levels, the LDL receptor diffuses over the cell surface and weak foci of clustered receptor, compared with adjacent non–gene-silenced cells, correspond mostly to clathrin-coated regions harboring residual Dab2 (Figure 1E). Fluorescence intensity analysis of these images shows clearly that in ARH−/− fibroblasts, peaks of LDL receptor labeling coincide precisely with a Dab2 signal but that there is negligible correspondence with the more diffuse LDL receptor after diminishing Dab2 protein levels (Supplementary Figure S1). By contrast, the transferrin receptor remains concentrated at bud sites irrespective of the intracellular Dab2 concentration (Figure 1F). Internalization of DiI-labeled LDL and subsequent concentration in perinuclear endosomes is also obviously slowed in ARH−/− cells with decreased Dab2 levels (Figure 1, G and H). Together, these results suggest that Dab2 actively clusters LDL receptors at clathrin bud sites in ARH nullizygous fibroblasts.
In HeLa cells, RNAi-mediated gene silencing of either Dab2 or ARH transcripts (Figure 2A) results in only mild impairment of LDL uptake. Compared with control cells, LDL receptors accumulate at the cell surface, although substantial LDL internalization is evident (Figure 2, C and D). Again, transferrin endocytosis proceeds normally in both control and RNAi-treated cells, and the intracellular transferrin colocalizes, in part, with internalized LDL. Yet, simultaneously knocking down both ARH and Dab2 with siRNA duplexes has a dramatic effect on LDL receptor trafficking (Figure 2E). LDL accumulates diffusely over the surface, but transferrin uptake continues relatively unimpeded. Anti-LDL receptor antibody binding shows a surface density increase of at least fivefold after Dab2+ARH RNAi. Time-resolved imaging of cells incubated with fluorescent LDL and transferrin confirms the severe effect of the Dab2+ARH double knockdown. Although LDL and transferrin are efficiently internalized and some sorted into common endosomes in the mock-treated HeLa cells, the surface of the siRNA-treated cells is heavily labeled with LDL, also diffusing into filopodia despite normal transport of transferrin to the cell interior (Supplementary Movie 1). Strikingly, these results are the reciprocal of those obtained when AP-2 levels are reduced >75% by RNAi, where transferrin but not LDL accumulates at the cell surface (Hinrichsen et al., 2003 ; Motley et al., 2003 ; see also Figure 6, A and B).
Because Dab2 has multiple clathrin-binding sites like AP-2 (Mishra et al., 2002a ), it is possible that general clathrin lattice assembly is impaired in the Dab2+ARH double knockdowns and this, rather than sorting aberrations, is responsible for the surface LDL accumulation. However, RNAi of epsin 1, an abundant poly/multiubiquitin-selective CLASP with multiple clathrin-binding sites, does not produce a similar effect. Despite >80% decrease in epsin levels in siRNA-transfected HeLa cells (Figure 3A), LDL uptake is similar in control and epsin-depleted cells (Figure 3, B and F), and in ARH or epsin+ARH knocked-down cells (Figure 2, C and G). Still, tandem Dab2+ARH knockdown severely blocks LDL internalization, with characteristic accumulation of surface LDL (Figure 3E). Clathrin (and dynamin 2; unpublished data), but not AP-2, ablation likewise inhibits LDL internalization, though the LDL still accumulates in punctate structures on the surface, which, given the presence of LDL-selective CLASPs and the fact that adaptors/CLASPs retain appropriate intracellular positioning in the absence of clathrin (Hinrichsen et al., 2006 ), is expected (Figure 3, H and I, arrowheads). In these experiments, the extent of ARH depletion is somewhat variable (compare Figures 2A, A,3A,3A, and and6A).6A). Nevertheless, no substantial increase in the extent of ARH gene silencing is seen if a multiple siRNA-containing SMARTpool is used instead of the single duplex used for the majority of this study. In both cases, ~80% decrease in steady state ARH levels can be achieved (Supplementary Figure S2).
A clear kinetic lag in the uptake of DiI-LDL occurs in Dab2+ARH-silenced cells (Figure 4and Supplementary Figure S3). Both transferrin and LDL populate common peripheral endosomes in control cells incubated for 5 min in the continuous presence of both ligands (Figure 4C and Supplementary Figure S3). Between 20 and 40 min marked concentration of LDL in juxtanuclear endosomes is apparent (Supplementary Figure S3B and C). A low level of LDL is present at the surface in the control cells between 1 and 5 min (Figure 4, A, C, and E, and Supplementary Figure S3, A and G), which is in accord with both the lower absolute rate of internalization and a less abundant rate-limiting component for LDL uptake compared with transferrin (Warren et al., 1998 ). In the Dab2+ARH-depleted cells, already between 1 and 5 min, very prominent appearance of a circumferential band of LDL at the plasma membrane occurs and is still evident at 40 min (Supplementary Figure S3F). Although LDL endocytosis is slowed (compare Figure 4, F to E), the ligand is nonetheless still internalized and traffics to the transferrin-positive early endosome compartment, but not as efficiently as in mock-treated cells (Supplementary Figure S4G). This overall slowed internalization of LDL in the FXNPXY-selective CLASP-depleted cells is fully consistent with the about fivefold reduction in internalization index characteristic of LDL receptor FXNPXY mutants devoid of an operational sorting signal (Davis et al., 1986 , 1987 ; Chen et al., 1990 ).
Importantly, efficient LDL uptake is restored upon transient expression of either an ARH-GFP or an siRNA-resistant YFP-Dab2 fusion protein in Dab2+ARH knockdown cells (Figure 5, A and E). This demonstrates that either of these proteins can individually restore LDL uptake and that the appended fluorescent tags do not significantly interfere with LDL endocytosis. We conclude that Dab2 and ARH appear to operate in a functionally redundant manner to drive the efficient internalization of the LDL receptor in nonpolarized cells and that the YXXØ (transferrin receptor) and FXNPXY (LDL receptor) sorting signals are decoded by distinct components of the clathrin coat.
Dab2 and ARH each contain cargo- (FXNPXY), PtdIns(4,5)P2-, AP-2-, and clathrin-binding regions, which allow direct coupling of bound cargo to the clathrin coat machinery (Morris and Cooper, 2001 ; He et al., 2002 ; Mishra et al., 2002a , 2002b ; Garuti et al., 2005 ). If either the AP-2– or the clathrin-binding information within ARH-GFP is singly compromised by mutation of either F259A or 212LLD→AAA, respectively, the expressed protein still promotes LDL internalization (Figure 5, B and C), though quantitation reveals that neither mutant rescues fully (Figure 5G; Garuti et al., 2005 ). These results are reminiscent of data obtained when the analogous binding sequences for clathrin or AP-2 are singly altered in β-arrestin 2 (Laporte et al., 2000 ; Kim and Benovic, 2002 ; Santini et al., 2002 ). When both these binding sites are disrupted, the ARH-GFP becomes diffusely distributed in the cytoplasm and fails to rescue the Dab2+ARH knockdown phenotype (Figure 5D). S122Y, a PTB domain mutation affecting the FXNPXY binding surface of YFP-Dab2 (Stolt et al., 2004 ), similarly prevents restoration of LDL internalization (Figure 5F), even with the expressed Dab2 still residing in puncta that colocalize with clathrin. By contrast, mutations to the phosphoinositide binding region (K51M and K150E) of the ARH PTB domain are still compatible with LDL internalization (Figure 5G), despite strongly interfering with ARH binding to PtdIns(4,5)P2 in vitro (Supplementary Figure S4). This affirms that the lipid-binding surface of ARH is juxtaposed to, but functionally separate from, the FXNPXY contact site (Mishra et al., 2002a , 2002b ; Stolt et al., 2004 ). Thus, Dab2 and ARH (when overexpressed) require engagement of both cargo and core coat components for proper LDL receptor sorting, whereas additional interactions with PtdIns(4,5)P2 and clathrin likely increase the efficiency of the process.
Despite the apparent functional redundancy of Dab2 and ARH, the two CLASPs clearly engage the AP-2 adaptor in molecularly distinct ways. Dab2 has several low-affinity (KD ~50–150 μM), tandemly arrayed α appendage–binding motifs (Morris and Cooper, 2001 ; Mishra et al., 2002a ), whereas ARH has only a single, moderate affinity (KD 1–2 μM) AP-2 interaction sequence with an absolute selectivity for the β2 appendage (He et al., 2002 ; Mishra et al., 2002b , 2005 ). β-arrestins have a related, α-helical [DE]nX1–2FXX[FL]XXXR motif that binds the platform subdomain of the β2 appendage (Mishra et al., 2005 ; Edeling et al., 2006 ; Schmid et al., 2006 ). Consequently, ARH might function similarly to β-arrestins, not participating directly in clathrin lattice assembly, but rather ushering LDL receptors through pre-existing coated buds. One prediction of this idea is that ARH requires AP-2 for localization to clathrin bud sites. Indeed, in AP-2+Dab2 knocked-down cells, but not with AP-2 or Dab2 knockdown alone, both transferrin and LDL receptor internalization is impaired despite normal ARH levels (Figure 6F). In these uptake experiments, fluorescent ligands are added directly to cells at 37°C, and in the AP-2–depleted cells (Figure 6B), transferrin, but not LDL, is poorly internalized and accumulates diffusely over the cell surface. Consistent with the ability of Dab2 to polymerize clathrin (Mishra et al., 2002a ) and drive LDL receptor internalization (Morris et al., 2002 ; Maurer and Cooper, 2005 ), Dab2 does not exhibit a requirement for AP-2; knockdown of ARH+AP-2 dramatically impairs transferrin but not LDL receptor internalization (Figure 6E). Taken altogether, these experiments clearly reveal important differences in the molecular mode of operation of these two CLASPs: like β-arrestin, ARH operates by directing LDL receptors to pre-existing clathrin-coated structures containing AP-2, whereas Dab2 can sustain clathrin-dependent LDL internalization independent of AP-2.
If Dab2 can maintain LDL endocytosis independent of AP-2, it is possible that this CLASP nucleates discrete clathrin coats, allowing different cargo types to be internalized within different subpopulations of coated vesicles, a possibility already suggested for certain G protein–coupled receptors (GPCRs) (Cao et al., 1998 ), transferrin (Tosoni et al., 2005 ), and LDL (Lakadamyali et al., 2006 ) receptors. In this case, Dab2 should only be present within a select subpopulation of coated structures along with the respective cargo. However, in both normal and ARH−/− fibroblasts at steady state, AP-2 is present in 92% of surface-proximal clathrin structures, Dab2 in 82%, and Dab2 populates 92% of the AP-2–positive structures (Figure 1 and Supplementary Figure S5; Morris and Cooper, 2001 ). Essentially analogous results are observed with HeLa and CV-1 cells (Figure 7, A and B, and unpublished data), clearly discrepant with the reported distribution of Dab2 in BS-C-1 cells (Lakadamyali et al., 2006 ), which, like CV-1 cells, are derived from the African green monkey (see below). Freeze-etch EM reveals Dab2 positioned within the majority of clathrin structures at the plasma membrane. Frequently, the CLASP appears positioned near the edges of the lattice (Figure 8), like eps15 (Tebar et al., 1996 ; Edeling et al., 2006 ), a Dab2-binding partner, although immunogold Dab2 labeling is also seen within regions of assembled lattice and on buds, particularly in HeLa cells. The predominantly coat-associated distribution of Dab2 is similar to the near-perfect colocalization observed for either eps15 or epsin in HeLa, CHO, and A431 cells (Chen et al., 1998 ; Yim et al., 2005 ; Hawryluk et al., 2006 ).
Moreover, the distribution of Dab2 is paralleled by the LDL receptor; the vast majority (84–89%) of surface LDL (and transferrin; 89–93%) is found in Dab2-positive puncta (Figure 7 and Supplementary Figure S5). In fibroblasts, 77% of ARH-positive puncta are positive for the LDL receptor and, at steady state, LDL, transferrin and clathrin signals are also highly coincident in both HeLa cells and fibroblasts; 88% of clathrin light-chain–positive structures contain transferrin, 81% contain LDL, and 56% have both transferrin and LDL signals when fibroblasts are labeled simultaneously with both ligands on ice (Figure 7 and Supplementary Figure S5). On adding LDL to cells at 37°C, YFP-Dab2 and endogenous ARH colocalize with LDL receptors during the initial period of LDL uptake, but diverge within 5 min (Figure 7, G and H). Also, in HeLa cells, 95% of eps15-positive structures contain LDL receptors (Figure 7C), again attesting to the occurrence of numerous CLASPs within a single assembling coat.
EM views of replicas of the external surface of fibroblasts show oval, ~25-nm LDL particles clustered around and within invaginations (coated buds; Figure 8E; Heuser, 1980 ; Sanan et al., 1987 ; Heuser and Anderson, 1989 ). HeLa cells, too, have LDL massed at presumptive bud site invaginations (Figure 8F). We also note that the extent of LDL receptor up-regulation upon culture in lipoprotein-deficient serum is not uniform in all cells, so adjacent cells often display different LDL receptor densities at the surface (Figures 7B and 10D). Dab2 is nevertheless distributed throughout surface clathrin structures and, indeed, there is little difference in the subcellular distribution of Dab2 between cells grown in whole or lipoprotein-deficient serum (Supplementary Figure S5). In sum, at least for constitutively endocytosed receptors, we find little evidence of obvious compartmentalization of incoming cargo within functionally distinct surface clathrin-coated structures, and the localization of FXNPXY-recognizing CLASPs to endocytic coats appears independent of receptor level fluctuations.
The small dispersed surface puncta that transiently overexpressed YFP-Dab2 and ARH-GFP both localize to in HeLa cells (Figures 5 and and7)7) contain AP-2 and clathrin (unpublished data). In these structures, the recovery of Y/GFP fluorescence after photobleaching reinforces the adaptor-like properties of these CLASPs (Figure 9, A and B). YFP-Dab2 repopulates bleached puncta with a t1/2 of ~15 s, analogous to clathrin and AP-2 (Wu et al., 2001 , 2003 ), whereas ARH-GFP recovers slightly quicker (t1/2 = ~9 s; Figure 9C). Typically, >85% of the prebleach puncta rapidly reacquire up to ~70% of the initial fluorescence in these experiments, consistent with both Dab2 and ARH cycling between soluble and membrane-bound states. Yet, despite dynamic oscillation of Dab2 and ARH at bud sites, many surface structures appear comparatively immobile over time, a behavior typically observed for tagged clathrin, AP-2, and other endocytic components (Gaidarov et al., 1999 ; Rappoport et al., 2003 ; Keyel et al., 2004 ; Merrifield et al., 2005 ; Bellve et al., 2006 ) but which, nonetheless, correlate with endocytic uptake (Bellve et al., 2006 ). In time-resolved images of transiently transfected HeLa cells, added DiI-LDL initially clusters at YFP-Dab2–positive surface puncta, whereas 5–20 min later, internalized LDL within larger, mobile endosomes is not colocalized with Dab2 (Figure 9E; Supplementary Movie 2); instead, rapid lateral LDL translocations indicative of directed endosome movement are apparent (Supplementary Movie 2 and Figure 9E, kymograph). Consequently, we judge the small, relatively immobile spots to be coats assembling at the plasma membrane. Similar DiI-LDL-containing clusters are seen in HeLa cells expressing ARH-GFP, GFP-epsin 1, or GFP-clathrin light chain (LCa; unpublished data), verifying the surface positioning these structures and, again, showing LDL concentrates in clathrin regions containing various CLASPs. At these small peripheral structures, LDL can be seen to accumulate within a minute or two after addition of DiI-LDL (Figure 9, D and E) and, in general, LDL concentrates in these puncta (Figure 9D).
The temporal relationship between Dab2 and LDL in our data sets can be categorized into what appear to be distinct types of budding events. For example, occasions where both Dab2 and LDL signals disappear from the optical section simultaneously occur. More frequently, overlapping Dab2- and LDL-positive puncta separate, and the Dab2 signal is lost rapidly revealing an LDL-positive endocytic vesicle (Figure 9F), which moves locally and often fuses with an adjacent vesicle. Also noticeable are larger LDL-rich endosomes moving close to the plasma membrane surface that consume local cargo emerging from assembled Dab2 zones (Supplementary Movie 2). The behavior of these structures is strikingly reminiscent of the consumption of just internalized endocytic vesicles by actin-dependent directed-movement of cargo-laden endosomal vesicles recently reported in Saccharomyces cerevisiae (Toshima et al., 2006 ). Altogether, our imaging studies confirm that the spatiotemporal behavior of Dab2 and ARH is consistent with a proposed role as FXNPXY-selective CLASPs. Further, the data reveal a dense subsurface population of dynamic, LDL-positive transport vesicles and a substantial degree of rapid homogenization/intermingling of the internalized ligand within early sorting endosomes.
Finally, if Dab2 and ARH are both generally present in surface coated buds and can operate analogously to promote LDL receptor internalization, why is ARH so crucial for maintaining normal plasma levels of LDL in vivo? One potential explanation is tissue-specific expression patterns of Dab2 and ARH. Comparisons of rat liver homogenate, purified Golgi membranes, and isolated plasma membrane sheets show dramatically different overall protein compositions (Figure 10A, lanes a–c) and the expected enrichment of AP-1 at the Golgi and AP-2 on the plasma membrane (Figure 10A, lanes g–i). However, both Dab2 and ARH are similarly enriched within the isolated plasma membrane preparation. These sheets are purified from whole liver homogenates (Hubbard et al., 1983 ), but when separated hepatocyte and nonhepatocyte fractions are compared, Dab2 is strikingly absent from asialoglycoprotein receptor–positive hepatocytes, although the CLASP is well expressed in the nonparenchymal (Kupffer, sinusoidal endothelial, and hepatic stellate cells) fraction (Figure 10B). By contrast, ARH and epsin 1 are present in both populations. Purified rat hepatocytes express less Dab2 protein at steady state than previously characterized cultured cell lines with down-regulated Dab2 (MDA-MB-231 and MCF-7; Fazili et al., 1999 ), although, again, ARH levels are roughly comparable. Also of note, HepG2, a human hepatoma line and BS-C-1 cells also contain low relative levels of Dab2 (Figure 10B). Yet, in our experiments, BS-C-1 cells still cluster LDL receptors at the majority AP-2 (ς2-GFP)-containing structures (Figure 9C), and at many surface clathrin (LCa-GFP) puncta (Figure 10D). These results again reinforce the redundancy between ARH and Dab2 and differ from a recent report showing only ~30% of surface clathrin structures associated with labeled LDL in BS-C-1 cells (Lakadamyali et al., 2006 ). We conclude that since the vast majority of plasma LDL is cleared/degraded by hepatocytes in vivo (Osono et al., 1995 ), the exceedingly low level of Dab2 present in this cell type is fully consistent with the phenotype that results from inheritance of two mutant ARH alleles; although nonparenchymal cells may be able to internalize LDL in an ARH−/− background, hepatocytes are clearly unlikely to internalize FXNPXY signals normally.
Clathrin-coated vesicles are high-capacity carriers that efficiently shuttle numerous cargo types into the cell interior in a generally noncompetitive manner (Marks et al., 1996 ; Santini et al., 1998 ; Warren et al., 1998 ). Although it is well accepted that cargo selection depends upon adaptors, here we provide functional evidence that efficient internalization of the LDL receptor is only indirectly dependent on the AP-2 adaptor and can, if driven by Dab2, occur in the absence of the heterotetramer. Thus, our experiments formally validate the CLASP hypothesis (Traub, 2003 ; Robinson, 2004 ; Sorkin, 2004 ) and lead to three major conclusions. First, the FXNPXY-selective CLASPs ARH and Dab2 operate in a functionally redundant manner, despite engaging the clathrin coat machinery in molecularly distinct ways; these proteins operate analogously to, and in parallel with, AP-2–dependent sequestration of transferrin receptors (Hinrichsen et al., 2003 ; Motley et al., 2003 ) and epsin/eps15/eps15R-mediated sorting of poly/multiubiquitinated proteins (Barriere et al., 2006 ; Duncan et al., 2006 ; Hawryluk et al., 2006 ; Wang et al., 2006 ). Very similar results revealing both the cargo selectivity and functional redundancy of ARH and Dab2 have been obtained in two independent studies (Maurer and Cooper, 2006 ; Anne Soutar, personal communication). Second, when Dab2 is present at ”normal“ cellular levels, the CLASP generally populates most clathrin-coated structures along with AP-2, epsin, and eps15, independent of LDL receptor expression levels. Third, Dab2 expression is cell type specific and, consequently, appears to underlie the hypercholesterolemic phenotype observed in both ARH-nullizygous patients and mice.
Given the selectivity and ability of certain CLASPs to sustain clathrin-mediated endocytosis in the absence of AP-2, refined models positing biased sorting of different cargo classes into discrete subpopulations of surface clathrin coats have been proposed (Tosoni et al., 2005 ; Lakadamyali et al., 2006 ). We believe that CLASPs are unlikely to target/generate distinct populations of clathrin coats for several reasons. First, classic EM studies document the colocalization of various cargo within single budding profiles at the plasma membrane: for example, LDL and α2-macroglobulin (LRP1; Via et al., 1982 ), transferrin and asialoglycoprotein (Neutra et al., 1985 ; Stoorvogel et al., 1987 ), lysosomal hydrolases (mannose 6-phosphate receptor), and α2-macroglobulin (Willingham et al., 1981 ), LDL and EGF (Carpentier et al., 1982 ), EGF and transferrin (Hanover et al., 1984 ), and EGF and α2-macroglobulin or adenovirus (Willingham et al., 1983 ). Further, active segregation of CLASPs and cargo would certainly require energy but, as most cargo sorted by clathrin-mediated events converge in a common early endosome population (Ajioka and Kaplan, 1987 ; Dunn et al., 1989 ; Maxfield and McGraw, 2004 ), any advantage of presorting various cargo would evidently be eliminated. In nonpolarized cells, peripheral Rab5-positive LDL-containing early endosomes mature by fusing with other Rab5- and transferrin/LDL-positive structures; homotypic fusion is estimated to occur once or twice per minute as early endosomes expand in size over ~10–15 min, mingling incoming endocytic vesicle contents (Rink et al., 2005 ). Rab4-containing tubules emanate from these early endosomes, allowing transferrin to exit, whereas LDL (and other molecules destined for degradation) concentrates within the vacuolar portion of the sorting endosome. There is a replete literature showing that LDL, asialoglycoprotein, and transferrin initially enter common sorting endosomes and segregate with a t1/2 of ~2.5 min (Stoorvogel et al., 1987 ; Dunn et al., 1989 ; Ghosh et al., 1994 ; Maxfield and McGraw, 2004 ), and the combined data have been synthesized into a general model for the iterative processing of cargo types that accompanies endosome maturation (Maxfield and McGraw, 2004 ). The enlarged LDL-positive late endosomes that ensue (Figures 3 and and9;9; Supplementary Movie 2) must mean that LDL-carrying structures fuse together as the endosome transits toward the lysosome and converts to a Rab7-positive stage. The time frame of this transition from small Rab5-positive to large Rab7-positive endosomes is ~20–40 min (Ghosh et al., 1994 ; Rink et al., 2005 ), in excellent agreement with classic data showing that detection of LDL degradation products requires >30 min after initial internalization (Brown and Goldstein, 1986 ). With this general and extensively substantiated model for endosome maturation along the degradative pathway, the teleological benefit of targeting a population of LDL-laden coated vesicles to a Rab7-positive status within only a few minutes (Lakadamyali et al., 2006 ) is not at all obvious to us. In this regard, it is also worth noting that although LDL moves to the late endosome for catabolism, the LDL receptor has a long biological life (~20 h) and follows a different trajectory, recycling back to the surface roughly five or six times per hour in fibroblasts (Brown and Goldstein, 1986 ), similar to the transferrin receptor. Furthermore, the slowed progression of LDL toward the late endosome reported after EGF stimulation (Rink et al., 2005 ) tends again to argue against compositionally distinct partitioning of LDL into a subset of clathrin structures at the cell surface.
Perhaps a more serious concern with the general idea of cargo-specific coats at the plasma membrane is that we believe there is currently no plausible mechanistic basis for their formation. Cargo-selective CLASPs, including β-arrestins, epsin, ARH, and Dab2, bind with relatively high-affinity to clathrin and PtdIns(4,5)P2, general components of all surface clathrin structures, as well as to AP-2. The membrane and cytosolic pools of these adaptors exchange rapidly, with a t1/2 of ~15 s (Figure 9; Wu et al., 2003 ; Yim et al., 2005 ), permitting soluble adaptors to interchange repeatedly at each clathrin bud during its 20–90-s lifetime (Gaidarov et al., 1999 ; Merrifield et al., 2002 ; Wu et al., 2003 ). Therefore, an ongoing and efficient segregation mechanism would be necessary to ensure the conversion of random CLASP encounters with assembling lattices into differential patterning of CLASPs within only a subset of coats, rather than in all of them. Yet, several time-resolved imaging studies demonstrate that fluorescently tagged β-arrestins populate virtually all detectable clathrin structures at the surface after ligand activation of a suitable GPCR (Santini et al., 2002 ; Scott et al., 2002 ). ARH accesses clathrin coats in a manner strictly analogous to β-arrestins, by binding to the AP-2 β2 appendage platform and the clathrin heavy chain terminal domain. Indeed, ectopically expressed ARH-GFP is found in the majority of clathrin structures on the surface (Figure 9B). Thus, our current understanding of the dynamics and points of contact between CLASPs and the core clathrin machinery makes the assembly and maintenance of compositionally distinct buds improbable in our view and is clearly inconsistent with the steady state distribution of Dab2 (Morris and Cooper, 2001 ), ARH, surface LDL receptors (Anderson et al., 1978 ; Heuser and Anderson, 1989 ; Sanan and Anderson, 1991 ), and epsin 1 (Chen et al., 1998 ; Yim et al., 2005 ; Hawryluk et al., 2006 ) in several cell types.
Instead, our data show specifically that Dab2 is present in the majority of clathrin structures at the surface, consequently allowing clustering of LDL. We cannot, however, rule out that under some circumstances, regulation of CLASP cargo-engagement properties, perhaps by phosphorylation as occurs with AP-2 (Olusanya et al., 2001 ; Höning et al., 2005 ), limits the composition of single clathrin buds by focally disabling certain CLASPs within the assembled coat. It is also conceivable that massed cargo could stabilize CLASPs in the assembly zone to generate cargo-selective buds, but this would apparently require preferential entry of transmembrane cargo into specific, but not all, coat-forming regions on the plasma membrane. Yet, LDL receptors are seemingly dispersed randomly over the surface membrane upon reinsertion from recycling endocytic vesicles, as are internalization-defective LDL receptors (Sanan et al., 1987 ). Our deep-etch images reveal that Dab2 is frequently positioned around the rims of the clathrin lattice. This may be due, in part, to association of Dab2 with eps15, which is almost exclusively positioned at lattice edges (Tebar et al., 1996 ; Edeling et al., 2006 ). Similar visualization of LDL particles on the extracellular plasma membrane surface (Heuser, 1980 ; Sanan et al., 1987 ; Heuser and Anderson, 1989 ; Sanan and Anderson, 1991 ), as well as thin section analysis (Orci et al., 1978 ; Handley et al., 1981 ; Paavola et al., 1985 ), often shows clustered peripheral LDL particles, positioned circumferentially around the involuting bud (Figure 8). Although we cannot rule out that epitope masking reduces detection of Dab2 within the lattice interior, it is probable this peripheral clustering of LDL is due to deposition of Dab2 near the edges. Our live cell data show some YFP-Dab2 does, on occasion, accompany LDL on a budding vesicle, and we also do observe some Dab2 labeling within the lattice, on invaginating buds, and in coated vesicle preparations (Hawryluk et al., 2006 ). ARH, on the other hand, appears to be found predominantly in the interior of clathrin lattices, occasionally on invaginating regions. Although the density of ARH labeling is low, this finding is consistent with ARH and β-arrestin playing similar roles and β-arrestin being enriched in preparations of purified coated vesicles (Blondeau et al., 2004 ). In time-resolved studies, the majority of the YFP-Dab2 signal at surface puncta is relatively immobile, in accord with other endocytic proteins (Rappoport et al., 2003 ; Keyel et al., 2004 ; Bellve et al., 2006 ) and recent work in primary adipocytes (Bellve et al., 2006 ). Indeed extensive arrays of flat hexagonal clathrin lattice are not only found on the ventral surface of normal human fibroblasts (Heuser, 1980 ; Heuser and Anderson, 1989 ), but on the dorsal surface as well (Damke et al., 1994 ), where LDL is similarly enriched around the circumference of the clathrin assemblies (Sanan and Anderson, 1991 ). Thus, in cells where extensive arrays of clathrin occur, Dab2 may locally concentrate FXNPXY cargo molecules, like the LDL receptor, for packaging into buds that form within or at the periphery these extended lattices.
There are at least two mammalian tissues where functional redundancy between Dab2 and ARH appears to have been selected against. Hepatocytes do not express Dab2 in sufficient quantity to bypass ARH loss and, reciprocally, during embryonic development the visceral endoderm does not appear to express ARH (Maurer and Cooper, 2005 ) and, consequently, targeted gene disruption of Dab2 is homozygous lethal (Morris et al., 2002 ). Hepatocytes characteristically respond to liver damage or partial hepatectomy with rapid proliferation and remodeling to restore the functional capability and mass of the liver. Because Dab2 acts downstream of TGF-β receptors to promote SMAD activation (Hocevar et al., 2001 ), and hepatocyte DNA synthesis and proliferation is potently suppressed by TGF-β (Wollenberg et al., 1987 ), perhaps the very low level of Dab2 in parenchymal cells is functionally related to the intrinsic proliferative capacity of hepatocytes.
We are extremely grateful to William Bowen, Jr. for generously providing rat hepatocyte fractions, to our numerous colleagues who provided important reagents, to Jon Cooper and Anne Soutar for sharing unpublished results, to members of the Traub lab for critical reading of the manuscript, and to Stuart Kornfeld for insightful advice and suggestions. This work was funded by National Institutes of Health Grant R01 DK53249 and American Heart Association (AHA) Established Investigator Award (0540007N) to L.M.T. P.A.K. was supported by AHA Predoctoral Fellowship Award 0415428U.
The online version of this article contains supplemental material at MBC Online (http://www.molbiolcell.org).
This article was published online ahead of print in MBC in Press (http://www.molbiolcell.org/cgi/doi/10.1091/mbc.E06-05-0421) on July 26, 2006.