Quantitative fluorescence imaging reveals point of release for lipoproteins during LDLR-dependent uptake.

The LDL receptor (LDLR) supports efficient uptake of both LDL and VLDL remnants by binding lipoprotein at the cell surface, internalizing lipoprotein through coated pits, and releasing lipoprotein in endocytic compartments before returning to the surface for further rounds of uptake. While many aspects of lipoprotein binding and receptor entry are well understood, it is less clear where, when, and how the LDLR releases lipoprotein. To address these questions, the current study employed quantitative fluorescence imaging to visualize the uptake and endosomal processing of LDL and the VLDL remnant β-VLDL. We find that lipoprotein release is rapid, with most release occurring prior to entry of lipoprotein into early endosomes. Published biochemical studies have identified two mechanisms of lipoprotein release: one that involves the β-propeller module of the LDLR and a second that is independent of this module. Quantitative imaging comparing uptake supported by the normal LDLR or by an LDLR variant incapable of β-propeller-dependent release shows that the β-propeller-independent process is sufficient for release for both lipoproteins but that the β-propeller process accelerates both LDL and β-VLDL release. Together these findings define where, when, and how lipoprotein release occurs and provide a generalizable methodology for visualizing endocytic handling in situ.

adding 39°C FLPPS medium and fl oating the 24-well dish in a 37°C water bath within a 37°C CO 2 incubator. Uptake was halted by rapidly replacing media with ice-cold PBS.

Immunofl uorescence
Cells were fi xed with 3% paraformaldehyde in PBS on ice for 20 min. Cells were then washed with ice-cold PBS and permeabilized with 10 g/ml digitonin in PBS for 2 min on ice. Cells were blocked with 1% normal goat serum (NGS) in PBS for 1 h at room temperature and incubated with primary antibody in PBS + 1% NGS for 1 h at room temperature. Cells were then washed with PBS + 1% NGS and incubated with Alexa-labeled secondary antibodies (Invitrogen) in PBS + 1% NGS for 1 h at room temperature. Following secondary incubation, cells were washed, DAPI stained (3 g/ml in PBS) for 10 min at room temperature, and then washed and mounted using Aqua poly/mount (Polysciences).

Image acquisition and colocalization
Four-channel fl uorescence z-stacks were taken at each time point of each experiment using a Deltavision pDV microscope with a 60×/1.42NA objective (Applied Precision). Acquisition settings were identical for all datasets to allow results from different time points and conditions to be compared quantitatively. Z-stacks were deconvolved using the blind deconvolution algorithm of Autoquant X (Media Cybernetics). Colocalization was evaluated quantitatively using the coloc module of Imaris (Bitplane/Andor). Intensity thresholds were set automatically using the method of Costes ( 21 ). Colocalization statistics (MC and PC) for each twochannel pair were output as csv fi les and analyzed with Excel. Voxel dimensions were 0.21526 × 0.21526 × 0.2 m. Typical experiments generated 50,000 to 500,000 doubly fl uorescent voxels per colocalization channel per 3D reconstruction. All reported quantitative imaging data are means ± SEM using image datasets acquired from at least three independent experiments performed on different days. An example of primary data is pre-

Lipoprotein internalization rate
Rates of lipoprotein internalization were determined as previously described ( 22,23 ). Briefl y, cells were incubated with 10 g/ml 125 I-LDL or 5 g/ml 125 I-␤ -VLDL for 1 h at 4°C in MEM media. Media were changed for the times indicated with warm FLPPS media also containing 10 g/ml 125 I-LDL or 5 g/ml 125 I-␤ -VLDL. Cells were extensively washed with ice-cold PBS and incubated with 1 mg/ml protease K in protease buffer (PBS + 1 mM EDTA) for 2 h at 4°C. The cell suspension was then centrifuged at 5,000 g for 10 min over a cushion of 10% sucrose in PBS. The tubes were frozen in liquid nitrogen, cut to separate the cells (internal) from the solution (surface-bound material released by protease K), and counted on a ␥ counter. Nonspecifi c activity was assessed in parallel experiments in the presence of 250 g/ml unlabeled lipoprotein. Nonspecifi c activities were subtracted from mean values for each data point. Data are means ± SD from four replicate trials and are representative of three independent experiments.

Lipoprotein accumulation assay
Rates of lipoprotein accumulation were determined as previously described ( 22 ). Briefl y, cells were incubated with 10 g/ml Alexa546-LDL or 5 g/ml Alexa546-␤ -VLDL for 1, 2, 3, or 4 h at 37°C, washed, scraped, fi xed, and washed. Mean cell fl uorescence was determined by fl ow cytometry from 10,000 gated events. Events were gated based upon side and forward scatter consistent with intact cells. Mean values were normalized to WT LDLR cell uptake at 4 h. Three separate experiments were process in which mildly acidic pH drives the ␤ -propeller into contact with LA4 and LA5 ( 16,17 ). This intramolecular contact accelerates lipoprotein release, most likely through an induced structural change in LA5 ( 18 ). The ␤ -propellerdependent release process has been proposed to be necessary for LDL release in endosomes because deletion of the EGF-B, ␤ -propeller, and EGF-C modules (BC region) cripples cellular accumulation of LDL ( 8 ). By contrast, VLDL remnant accumulation is not affected by this deletion, suggesting that the calcium release process supports remnant release. Why different release processes might have different roles during uptake of different lipoproteins is not clear.
To address this question, we determined where, when, and how cells drive lipoprotein release during endocytosis. Previous efforts to characterize LDLR release during endocytosis have been hindered by the inability to visualize the release process. Here, we developed a quantitative 3D fl uorescence imaging approach that captures the release process. We fi nd that the majority of both LDL and VLDL remnants dissociate from the LDLR prior to entry of lipoprotein into early endosomes and that both release mechanisms operate during uptake of both lipoproteins. We also fi nd that receptors lacking the BC region fail to support LDL accumulation not because release fails to occur but because LDL internalized by these receptors is ineffi ciently delivered to early endosomes and is instead resecreted.

Materials
All cell culture reagents were purchased from Gibco/Invitrogen (Carlsbad, CA). Rabbit anti-LDLR IgG was a gift from Joachim Herz (Department of Molecular Genetics, UT Southwestern Medical Center, Dallas, TX). Formaldehyde was from Fluka (Buchs, Switzerland). All Alexa probes were from Molecular Probes/Invitrogen (Carlsbad, CA). All other chemicals were from Sigma (St. Louis, MO). LDL was prepared from freshly drawn human plasma as previously described ( 19 ). Beta migrating VLDL ( ␤ -VLDL) was prepared from the serum of cholesterolfed rabbits as previously described ( 20 ).

Pulse-chase uptake of fl uorescent lipoprotein
Cells on coverslips in 24-well dishes were placed on ice and incubated with ice-cold MEM medium (MEM supplemented with 10% FLPPS) for 10 min to halt endocytosis. Cells were then incubated with either 10 g/ml Alexa-labeled LDL or 5 g/ml Alexalabeled ␤ -VLDL in MEM medium for 90 min on ice. Cells were washed with ice-cold MEM medium and shifted to 37°C by microscopy and deconvolved to generate 3D images. The approach generated intensity information for three fl uorescent components with spatial resolution of 0.2 × 0.21526 × 0.21526 m per voxel for four million voxels per image stack. Use of pulse-chase methodology allowed visualization of lipoprotein progression through the endocytic system.
We employed two measures of colocalization to assess lipoprotein transit through the endocytic system. The fi rst measure is the Manders overlap coeffi cient (MC), which is simply the number of voxels with above-threshold fl uorescence in two channels divided by the number of voxels with above-threshold fl uorescence for one of the two individual channels ( 27 ). For example, the MC of the LDLR with clathrin provides the fraction of LDLR-positive voxels that also contain clathrin. The second metric is the Pearson product-moment correlation coeffi cient (PC), which was calculated for the population of voxels that contain above-threshold fl uorescence in both channels (doubly fl uorescent voxels). PC is the normalized covariance in fl uorescent intensities between two channels. Values range from Ϫ 1 to 1, with 1 indicating a one-to-one correspondence in intensities, 0 indicating that the two intensities are uncorrelated, and Ϫ 1 indicating that the intensities are inversely correlated. Because fl uorescent intensity is a function of the abundance of a component within a voxel, the fl uorescent intensity of two components will covary if the two components interact stoichiometrically. For this reason, PC is widely used to distinguish true association from random overlap ( 28 ). The MC and PC values are the means ± standard errors of at least 15 image stacks of randomly chosen fi elds of view from three independent trials (minimum of 5 image stacks per experimental condition per trial). Use of 5 images was suffi cient sampling because the mean PC and MC values of different trials were not signifi cantly different ( P > 0.05). Additional method details and example images are provided in supplementary Fig. I. Additional discussion of the use of MC and PC for colocalization analysis has been detailed in recent reviews ( 24,29 ).
Changes in MC and PC indicate relative change in association. When two components begin to associate, MC rises as the number of doubly fl uorescent voxels increases. PC also increases because the new population of doubly fl uorescent voxels has covariant fl uorescent intensities. During dissociation, MC decreases as the number of doubly fl uorescent voxels decreases. PC also declines because partial loss of interacting components occurs stochastically within the population of doubly fl uorescent voxels, thereby generating variability in the ratio of fl uorescent intensities. This loss of stoichiometry occurs prior to complete loss of fl uorescence and provides a truer measure of dissociation than MC. We have employed MC and PC to follow LDLR interactions with lipoprotein and the transit of lipoprotein through EEA1-positive early endosomes during lipoprotein uptake.
We began our analysis of lipoprotein uptake by visualizing lipoprotein binding at the cell surface. Fibroblasts expressing normal LDLR (WT LDLR cells) were surface loaded or not with fl uorescent lipoprotein at 4°C and then performed, and reported data are the means of the mean values from each of the three experiments ± SD.

Lipoprotein degradation and excretion assay
Lipoproteins were incubated with cells at 4°C at a concentration of 10 g/ml 125 I-LDL, 10 g/ml 125 I-LDL + 1.25 g/ml unlabeled ␤ -VLDL, 5 g/ml 125 I-␤ -VLDL, or 1.25 g/ml 125 I-␤ -VLDL ± 10 g/ml unlabeled LDL in MEM medium for 90 min. Cells were then extensively washed with ice-cold MEM medium and either shifted to 37°C by medium replacement with warm FLPPS medium or washed, hydrolyzed, and assayed for labeled lipoprotein binding per milligram of cell protein. At the indicated times, conditioned media from 37°C-shifted cells were isolated, and intact lipoproteins were precipitated using 5% trichloroacetic acid (TCA). TCA-insoluble counts represent retro-endocytosed LDL. TCA-soluble counts represent degraded LDL. Counts were converted to picograms of LDL using the specifi c activity of LDL. TCA-insoluble and TCA-soluble values were normalized to initial surface-bound lipoprotein. Reported data are mean values ± SD (n = 4) and are representative of three independent experiments.

Lipoprotein isosurface construction and analysis
Lipoprotein-containing isosurface objects were created for each deconvolved z-stack using Imaris 3D rendering and measurement software (Bitplane/Andor). Creation of surfaces was automated with Imaris Batch Coordinator, so that the same criteria were applied to all datasets. Minimum intensity thresholds were determined interactively for each color channel, and the same thresholds were used for all datasets that were compared. The data were not smoothed, and the minimum voxel size fi lter was set at 1.0. Channel intensity statistics for each resulting surface set were output as csv fi les and analyzed with Excel. For each time point and condition, lipoprotein-containing surface objects were sorted into four classes on the basis of their mean intensity in the EEA1 and LDLR channels. Minimum intensity thresholds were determined for each channel, and the same thresholds were used for all datasets that were compared. Objects with both EEA1 and LDLR mean intensity below their respective thresholds were assigned to the LDL/VLDL-only class. Objects with EEA1 below threshold and LDLR above threshold were assigned to the LDL/ VLDL + LDLR class, etc. The sum of the total lipoprotein fl uorescence contained within the surface set was computed for each class and normalized by the total overall lipoprotein fl uorescence within the unsorted surface set to obtain the fraction of lipoprotein in each class for each time point and condition. These procedures were performed separately on three independent experiments. The reported data is the means of the percentages of each class at each time point and condition ± SEM. Additional discussion of this method is provided in supplementary Fig. II. The use of isosurfaces for colocalization analysis has been described previously (24)(25)(26).

RESULTS
The progression of lipoprotein through the endocytic system was visualized by quantitative 3D fl uorescence imaging of the LDLR, clathrin heavy chain, EEA1, LDL, and the VLDL remnant ␤ -VLDL. Lipoproteins were directly labeled with Alexafl uor dyes, whereas the LDLR, clathrin, and EEA1 were labeled by indirect immunofl uorescence using Alexafl uor-conjugated antibodies. Image stacks of each visualized fi eld were acquired by wide-fi eld epifl uorescence MC for clathrin with LDLR increased in the presence of ␤ -VLDL but not LDL ( Fig. 1B ). PC for doubly fl uorescent voxels containing both clathrin and LDLR were similar irrespective of lipoprotein, suggesting that coated pits with either induced or constitutively targeted LDLRs have similar LDLR density. The concordance between published EM data and the current quantitative imaging indicates that the imaging approach accurately visualizes associations between lipoprotein, the LDLR, and clathrin.
Following coated-pit targeting, LDLR-lipoprotein complexes are internalized into coated vesicles. These vesicles uncoat to form nascent endocytic vesicles and acquire early EEA1, a component of the molecular machinery that drives fusion of endocytic vesicles with each other and with early endosomes. Where in this endocytic process lipoprotein release occurs has not been formally demonstrated; however, early endosomes have been the presumed location because these endocytic compartments have lumenal pH and calcium concentrations that are suffi cient to drive both ␤ -propeller-dependent and calcium-releasedependent mechanisms ( 5,8 ). To visualize when and where release occurs, we performed pulse-chase uptake of fl uorescently labeled lipoproteins in conjunction with immunostaining for the LDLR and EEA1. Surface LDLRs of WT cells were loaded with fl uorescent lipoprotein at 4°C, washed to remove unbound lipoprotein, and then chased at 37°C. MC and PC were determined for LDLR with LDL and for LDLR with ␤ -VLDL to assess release of lipoprotein from the LDLR. Entry into early endosomes was assessed using the MC and PC for EEA1 with LDL and for EEA1 with ␤ -VLDL. We observed that both MC and PC for LDLR with lipoprotein fell rapidly during the chase ( Fig. 2A , B ), indicating in situ lipoprotein release and loss of LDLR from endocytic compartments bearing lipoprotein. Curve fi tting of the decrease in PC to a single-site dissociation model yielded a half-life of 0.89 min for LDLR-LDL complexes and 5.1 min for LDLR-␤ -VLDL complexes. The rapidity of the LDL dissociation is consistent with the ability of even mildly acidic pH to effi ciently drive LDL release ( 17 ). The slower release of ␤ -VLDL is consistent with the observation that ␤ -VLDL binding is less sensitive to both acidic pH and low calcium than is LDL binding ( 8 ). The fall in PC values for LDLR with lipoprotein coincided with increases in the MC and PC for EEA1 with lipoprotein ( Fig. 2C, D ). Importantly, the rises in PC and MC for EEA1 with LDL preceded the rises in PC and MC for EEA1 with ␤ -VLDL. The correlation of slower lipoprotein release with slower lipoprotein entry into early endosomes suggested that lipoprotein release is coupled to entry into early endosomes.
To distinguish this possibility from the formal alternative that ␤ -VLDL internalization is slower than LDL internalization, we assayed lipoprotein internalization rates. In these experiments, LDLRs on the surface of WT LDLR cells were loaded with 125 I-labeled LDL or ␤ -VLDL at 4°C followed by a 15 min chase at 37°C. At time points during the chase, internalized and surface-bound lipoprotein were assayed and used to calculate the ratio of internal versus surface lipoprotein. Rate constants derived from the immunostained for clathrin and the LDLR. Lipoprotein binding requires the presence of LDLRs on the surface of these cells ( 17 ). As a consequence, all surface-bound lipoprotein should have an associated LDLR, and the MC for LDL with LDLR and the MC for ␤ -VLDL with LDLR were both exceptionally high ( Fig. 1 ). In fi broblasts at steady state, approximately half of all LDLRs are surface exposed, with the remainder in transit through the endocytic system ( 30 ). As a consequence, MC values for LDLR with lipoprotein were lower than MC values for lipoprotein with LDLR. While PC for LDLR with LDL and PC for LDLR with ␤ -VLDL were both strongly positive, the PC for LDLR with LDL was signifi cantly higher than the PC for LDLR with ␤ -VLDL. This difference may result from differences in lipoprotein binding stoichiometry in that LDL has one copy of apoB100 and binds to the LDLR with a fi xed 1:1 stoichiometry, whereas ␤ -VLDL has multiple copies of apoE and can thus bind the LDLR with a range of stoichiometries. Prior thin-section electron microscopy (EM) has shown that, although LDLRs are constitutively targeted to coated pits, only 25-50% of all surface LDLRs are present in coated pits at steady state ( 22,31 ). The MC values of 0.4-0.5 for LDLR with clathrin in the presence and absence of lipoprotein is consistent with the EM quantifi cation. MC values for clathrin with LDLR were lower than MC values of LDLR with clathrin because clathrin is abundant in the cytosol and on internal membrane compartments, such as the trans-Golgi network, and because not all clathrin structures at the surface contained LDLRs. Prior thin-section EM has shown that binding of ␤ -VLDL, but not LDL, induces LDLR localization to coated pits ( 22 ). These EM observations suggest that ␤ -VLDL binding allows LDLRs to access or to induce the de novo formation of a novel population of coated pits. Consistent with the EM observations, quantifi ed the fraction of total surfaces that contained only lipoprotein during the course of lipoprotein uptake. We observed that the population of lipoprotein-only isosurfaces rose rapidly as isosurfaces with LDLR declined. The rise in lipoprotein-only surfaces preceded the rise in isosurfaces with EEA1 voxels, indicating that LDLRs leave endocytic vesicles prior to the entry of lipoprotein into early endosomes ( Fig. 3E, F ).
To determine how endocytic compartments drive lipoprotein release, we compared lipoprotein uptake in WT internal/surface ratios showed that LDL and ␤ -VLDL internalized at the same rate with a receptor cycling time of 14 min ( Fig. 3A , B ). The endocytic rates observed in WT LDLR cells are similar to rates of LDLR-dependent uptake previously observed in human fi broblasts ( 32 ). Similarity between endocytic rates of LDL and ␤ -VLDL internalization has also been observed previously ( 23 ). These fi ndings indicate that a slower rate of internalization is not responsible for the slower release and entry into early endosomes of ␤ -VLDL.
Quantitative imaging of LDLR colocalization with EEA1 provides further support for the hypothesis that lipoprotein release precedes lipoprotein entry into early endosomes. Because release of ␤ -VLDL from the LDLR is slower than release of LDL ( 8 ), if entry into early endosomes is independent of release, then LDLR colocalization with EEA1 should increase during the chase and ␤ -VLDL should promote more LDLR colocalization with EEA1 than LDL. By contrast, if early endosome entry is dependent upon release, then LDLR colocalization with EEA1 should remain constant during the chase irrespective of lipoprotein. Examination of MC for EEA1 with LDLR showed that at no time point with either lipoprotein did LDLR content in EEA1-positive voxels increase ( Fig. 3C ). PC values for EEA1 with LDLR were low and at no time rose above the 0.1 threshold for signifi cant colocalization in the presence of either lipoprotein ( Fig. 3D ). Quantitative imaging also afforded the ability to track lipoprotein-containing compartments that lack both LDLR and EEA1 as a function of time. For this purpose, we generated isosurfaces from lipoprotein-positive voxels (supplementary Fig. II) and  Isosurfaces were classifi ed into four bins: surfaces that lack both LDLR and EEA1; surfaces that contain LDLR but not EEA1; surfaces that contain both LDLR and EEA1; and surfaces that contain EEA1 but not LDLR. Percentages of total lipoprotein surfaces for each bin were calculated for each experiment and are reported as means ± SEM, n = 3.
MC for EEA1 with LDLR-⌬ BC also showed no change over the chase with either lipoprotein ( Fig. 4G, H ). These fi ndings show that, although the ␤ -propeller independent release mechanism is suffi cient for both LDL and ␤ -VLDL release, the ␤ -propeller accelerates both LDL and ␤ -VLDL release.
Deletion of the BC region reduces LDL accumulation by 70%, but it has no effect on ␤ -VLDL accumulation ( Fig. 5A , B ) ( 8 ). Because the LDLR-⌬ BC binds both LDL and ␤ -VLDL with normal affi nity and supports normal rates of lipoprotein uptake ( 8 ), the reduction in LDL accumulation suggests that the majority of LDL that is internalized by the LDLR-⌬ BC is excreted. In support of this possibility, the total LDL fl uorescence intensity associated with LDLR-⌬ BC cells fell rapidly during the chase with LDLR cells with uptake in cells expressing the LDLR-⌬ BC, which lacks the EGF-B, ␤ -propeller, and EGF-C modules (BC region). Cell surface release assays have shown that deletion of the BC region sharply impairs acid-dependent release of both LDL and ␤ -VLDL ( 8 ) and that PC values for LDLR with LDL and LDLR with ␤ -VLDL decayed with slower kinetics in LDLR-⌬ BC cells compared with WT LDLR cells ( Fig. 4A , B ). Consistent with the conclusion that release precedes entry into early endosomes, PC values for EEA1 with LDL and EEA1 with ␤ -VLDL peaked with delayed kinetics in LDLR-⌬ BC cells compared with WT cells ( Fig. 4C, D ), and at no time did PC for EEA1 with LDLR rise to the 0.1 signifi cance threshold ( Fig. 4E, F ).   5. Deletion of the BC region increases LDL retro-endocytosis. The indicated cells were assayed for LDL and ␤ -VLDL accumulation (A and B), total fl uorescence intensity (C and D), or degradation and excretion of lipoprotein (E and F). A and B: Cells were incubated with saturating concentrations of Alexa546-LDL or ␤ -VLDL for the indicated times at 37°C, washed, fi xed and processed by fl ow cytometry. Data was normalized to cellular fl uorescence of WT cells following 4 h of uptake. Data are means ± SD from three independent experiments. C and D: Total integrated fl uorescence from each experiment described in Fig. 4 was quantifi ed. Data are means ± SD, n = 3. E and F: Cells were incubated with saturating concentrations of 125 I-LDL or 125 I-␤ -VLDL at 4°C, washed, and either assayed for surface-bound lipoprotein or incubated at 37°C for 4 h. Media from cells incubated at 37°C were assayed for degradation products of LDL (TCA soluble counts) or excreted LDL (TCA insoluble counts). Data are shown as a fraction of initially surface-bound lipoprotein and are means ± SEM, n = 9. * P < 0.05 for LDLR-⌬ BC compared with WT LDLR. releases both lipoproteins rapidly following internalization and that most release occurs prior to entry of lipoprotein into EEA1-positive early endosomes. Biochemical structure-function studies previously identifi ed two release mechanisms: a ␤ -propeller-dependent mechanism that is triggered by acidic pH and a ␤ -propeller-independent mechanism that is triggered by loss of calcium. Quantitative imaging shows that both mechanisms participate in the release of both lipoproteins.
Key features of the quantitative imaging approach were the size of the datasets and the use of both MC and PC to follow lipoprotein progression through the endosomal system. The 3D fl uorescence imaging of multiple fi elds across multiple experiments generated datasets composed of several million of fl uorescent voxels per dataset. Use of MC with these datasets provided a robust assessment of kinetics that mirrored the rate of LDL dissociation from the LDLR-⌬ BC ( Fig. 5C, D ). Published work has shown that a fraction of both LDL and ␤ -VLDL that is internalized by the LDLR is excreted through a process termed retro-endocytosis ( 23,33,34 ). We tested whether the BC deletion specifi cally augmented retro-endocytosis of LDL by comparing LDL and ␤ -VLDL degradation and excretion in pulse-chase experiments using 125 I-labeled LDL and ␤ -VLDL. LDLR-⌬ BC cells supported less degradation and more excretion of LDL compared with WT LDLR cells, while degradation and excretion of ␤ -VLDL was unchanged by the BC deletion ( Fig. 5E, F ). These observations indicate that deletion of the BC region compromises LDL uptake by facilitating loss of LDL through retroendocytosis.
Why the BC deletion compromises transfer of LDL but not of ␤ -VLDL into early endosomes is unclear. One possibility is that LDL release and ␤ -VLDL release occur in distinct compartments and only transfer of lipoprotein from the LDL release compartment to early endosomes is sensitive to the BC deletion. We tested whether LDL and ␤ -VLDL are segregated into distinct pre-early endosomal compartments using the pulse-chase uptake methodology with a combination of LDL and ␤ -VLDL. If cells segregate LDL from ␤ -VLDL into separate release compartments, MC and PC between LDL and ␤ -VLDL should decline prior to entry of lipoprotein into early endosomes. Instead, we observed that the MC and PC of LDL with ␤ -VLDL increased throughout the chase ( Fig. 6A ). Furthermore, while fusion between LDL-only and ␤ -VLDL-only vesicles was observed in live cell imaging experiments, we failed to detect fi ssion events generating LDL-only or ␤ -VLDL-only vesicles from endocytic compartments with both lipoproteins (data not shown). These fi ndings indicate that LDL and ␤ -VLDL are not segregated prior to entry into early endosomes and that a common pre-early endosomal compartment accommodates both LDL and ␤ -VLDL. This conclusion is consistent with thin-section EM experiments showing that LDL and ␤ -VLDL cotraffi c through the endocytic system ( 35 ). Compared with single lipoprotein uptake experiments, the release of lipoprotein from the LDLR was slowed and entry of lipoprotein into early endosomes was slightly delayed by the presence of a second lipoprotein ( Fig. 6B, C ); however, lipoprotein degradation was not affected ( Fig. 6D ). These fi ndings show that cells do not segregate LDL and ␤ -VLDL into separate release compartments but rather use a common preendosomal compartment for release. The presence of multiple lipoproteins slows release in this common compartment; however, this delay is small and does not infl uence fi nal degra dation rates.

DISCUSSION
The current work defi nes where, when, and how the LDLR releases LDL and the VLDL remnant ␤ -VLDL during endocytosis using the fi broblast model system. The quantitative 3D fl uorescence imaging shows that the LDLR Fig. 6. Co-internalization of LDL and ␤ -VLDL reveals that LDL and ␤ -VLDL use common endocytic compartments. WT LDLR cells were incubated with 10 g/ml Alexa488-LDL, 5 g/ml Alexa546-␤ -VLDL, or a mixture of 10 g/ml Alexa488-LDL and 1.25 g/ml Alexa546-␤ -VLDL at 4°C, and then washed and incubated at 37°C for the indicated times. The 10:1.25 ratio resulted in a ‫ف‬ 50:50 ratio of LDL and ␤ -VLDL binding. At each time point, cells were rapidly chilled to 4°C, washed, fi xed and immunostained for either EEA1 (A and C) or LDLR (B). A: MC and PC for colocalization of LDL with ␤ -VLDL for cells incubated with a mixture of the two lipoproteins. B: PC for LDLR with LDL alone, with LDL in the presence of ␤ -VLDL, with ␤ -VLDL alone, or with ␤ -VLDL in the presence of LDL. C: PC for EEA1 with LDL alone, with LDL in the presence of ␤ -VLDL, with ␤ -VLDL alone, or with ␤ -VLDL in the presence of LDL. D: Degradation assays in which 125 I-LDL or ␤ -VLDL were incubated with WT LDLR cells at 4°C in the absence or presence of suffi cient unlabeled ␤ -VLDL or LDL, respectively, to decrease 125 I-lipoprotein binding in half. + ␤ -VLDL and +LDL indicate assays in which unlabeled ␤ -VLDL or LDL were present. Concentrations of unlabeled lipoprotein were titrated to reduce radiolabeled lipoprotein binding by 50%. Cells were then washed and incubated at 37°C for the indicated times and assayed for 125 Idegradation products. Data was normalized to initially bound 125 Ilipoprotein and are means ± SD, n = 4. membrane. It is at the limiting membrane where the Vpump and the TRPV2 channel inject protons and withdraw calcium ions, and this proximity may facilitate lipoprotein release from the LDLR.
In addition to lipoprotein release, LDLRs must recycle back to the cell surface, and the short transit time of LDLRs through internal compartments necessitates rapid recycling. This rapid recycling likely contributes ineffi ciency to the uptake process through premature commitment of LDLRs to the recycling pathway, thereby resulting in excretion (retro-endocytosis) of lipoprotein. Not surprisingly, inhibitors of endosome acidifi cation increase retroendocytosis ( 38 ); however, these inhibitors also hinder LDLR recycling, resulting in loss of LDLRs from the cell surface ( 39 ). This latter observation suggests that the LDLR has a mechanism that hinders entry of LDLRs into the recycling pathway prior to exposure of LDLRs to an acidifi ed compartment. Acidic pH drives lipoprotein release and this retention mechanism likely reduces the loss of internalized lipoprotein to retro-endocytosis. Recent work has shown that the intracellular retention of LDLRs that occurs in the presence of acidifi cation inhibitors requires the BC region ( 8 ). This region undergoes a massive conformational change in response to acidic pH ( 16,40 ) and this change may turn off the retention activity of the BC region. Why deletion of the BC region allows retroendocytosis of LDL but not of ␤ -VLDL is not clear. LDL and ␤ -VLDL do not separate following internalization ( Fig. 6 ), indicating that the two lipoproteins traffi c through common release compartments. One possibility is that an intrinsic feature of ␤ -VLDL hinders LDLR-␤ -VLDL complexes from entering the recycling pathway. ␤ -VLDL has been shown to augment intracellular retention of LDLRs when receptor recycling is impaired by endosomal acidification inhibitors or by deletions that remove the EGF-A and BC regions ( 7,8,39 ). Features that may hinder LDLR-␤ -VLDL access to the recycling pathway include size ( ␤ -VLDL is much larger than LDL) and valency ( ␤ -VLDL contains multiple binding sites for LDLRs whereas LDL contains only one LDLR binding site).
Failure of LDLRs to recycle results in eventual lysosomal degradation of receptors. Failure to recycle can result from mutations in the LDLR (class V FH mutations) ( 41 ) or from the action of two regulatory proteins, IDOL and PCSK9. IDOL is an E3 ligase that ubiquitinates the LDLR and drives LDLR traffi cking to lysosomes via the ESCRT complex ( 42 ). PCSK9 is a secreted protein that inhibits LDLR recycling by binding to the EGF-A module of the LDLR, which normally promotes LDLR recycling ( 8,43 ). Inhibitors of IDOL and PCSK9 hold promise as cholesterol-lowering therapies ( 44,45 ). Our data show that LDLRs in fi broblasts are normally excluded from early endosomes irrespective of lipoprotein release effi ciency ( Fig. 4 ), and future characterization of this activity may identify new therapeutic targets that can be exploited to prolong LDLR half-life and thereby facilitate LDL clearance.
In summary, the present work identifi es where, when, and how LDLRs release lipoprotein during endocytosis. Release is rapid and occurs prior to entry of lipoprotein colocalization that was stable across multiple experiments. Use of PC within the colocalized volumes provided information regarding the quality of association of lipoprotein with the LDLR and EEA1. The combination of MC and PC afforded the ability to compare colocalization as a function of experimental manipulation. The application of these statistical methods to time courses of LDLR-dependent lipoprotein uptake provided the means to quantify the kinetics of lipoprotein release and entry into early endosomes. Additionally, analysis of fl uorescence intensities inside lipoprotein isosurface objects allowed us to classify subpopulations of lipoprotein-containing compartments based upon their contents and to examine the changes in these subpopulations over time.
Our fi nding that lipoprotein release occurs rapidly is consistent with the speed of LDLR internalization cycles. The time required for the LDLR to complete a cycle of lipoprotein internalization is 12-15 min in fi broblasts ( Fig.  3A ) ( 32 ). At steady state, half of all LDLR are surface exposed in fi broblasts ( 36,37 ). Thus, the transit time for LDLRs through the endocytic system is 6-7 min. Lipoprotein release must therefore be fast; estimates of LDL and ␤ -VLDL release from the time-dependent change in PC suggest rates of 0.78 min Ϫ 1 for LDL release and 0.14 min Ϫ 1 for ␤ -VLDL release. Actual rates are likely faster because PC cannot distinguish between LDLR that is lipoproteinbound and LDLR that has released lipoprotein but has not yet traffi cked away from the endocytic compartment holding lipoprotein. The rapid rate of release is incompatible with release in early endosomes because lipoprotein colocalization with EEA1 does not peak until 8-15 min ( Fig. 2 ). Consistent with this conclusion, MC and PC for EEA1 with LDLR are small and remain small throughout uptake of both LDL and ␤ -VLDL ( Fig. 3 ). Furthermore, quantitative imaging identifi ed a population of lipoprotein voxels lacking both LDLR and EEA1, which appeared following the rapid loss in PC of LDLR with lipoprotein and prior to the increase in PC of EEA1 with lipoprotein. These observations indicate that release occurs prior to lipoprotein entry into early endosomes.
Rapid release satisfi es the kinetics of LDLR transit through the endosomal system, but it is problematic for the mechanism of release. The two known release processes, the ␤ -propeller-dependent process and the ␤ -propellerindependent process, cooperate in surface release assays ( 8 ); however, the pH and calcium concentrations necessary to achieve rapid release are not thought to exist early in the endocytic process. The rate of endocytic release inferred from the change in PC equates with rates of surface release observed for pH between 5.5 and 6.0 and calcium concentrations between 2.5 and 10 M ( 8 ). Ratiometric measurements of endosomal pH and calcium have shown that dextrans do not experience these conditions until 20 min after internalization ( 5 ). How then does the LDLR complete its endocytic cycle in the allotted 6-7 min? One possibility is that the pH and calcium measured in the bulk of the lumen by labeled dextrans is different from the pH and calcium experienced by the LDLR, which, as an integral membrane protein, is attached to the limiting