Hepatocytes Internalize Trophic Receptors at Large Endocytic “Hot Spots”

Abstract

Clathrin-mediated endocytosis in mammalian epithelial cells is believed to require the synergistic action of structural coat proteins and mechanochemical enzymes to deform and sever the plasma membrane (PM) into discreet vesicles. It is generally believed that the formation of clathrin-coated pits in epithelial cells occurs randomly along the apical and basolateral plasma membranes. In this study we have visualized the endocytic machinery in living hepatocytes using green fluorescent protein GFP–tagged dynamin, a large mechanochemical GTPase implicated in the liberation of nascent vesicles from the plasma membrane and a variety of internal membrane compartments. Confocal microscopy of living cells expressing the epithelial isoform of GFP-tagged dynamin [Dyn2-GFP] revealed a distribution along the ventral PM in discreet vesicle-like puncta or in large (2–10 μm) tubulo-reticular plaques. Remarkably, these large structures are dynamic as they form and then disappear, while generating large numbers of motile endocytic vesicles with which dynamin associates. Inhibiting dynamin function by microinjection of purified dynamin antibodies increases the number and size of the tubulo-reticular plaques. Importantly, these “hot spots” sequester specific trophic receptors and cognate ligands such as the TfR1, but not the TfR2. These findings suggest that hepatocytes sequester or pre-recruit both structural and enzymatic components of the clathrin-based endocytic machinery to functional “hot spots,” from which large numbers of coated pits form and vesicles are generated. This process may mimic the endocytic organization found at the synapse in neuronal cells.

INTRODUCTION

Clathrin-mediated endocytosis in hepatocytes and other epithelial cells involves the assembly of a structural scaffold along the PM composed of clathrin, adaptors, and other proteins that sequester surface receptors and deform the lipid bilayer into an invaginating bud (1–4). These coated invaginations are believed to be liberated from the PM by the combined efforts of lipid-modifying enzymes (5, 6), the actin-myosin cytoskeleton (7, 8), and the large GTPase dynamin (9–12). It is unclear how the location, interaction, and function of dynamin and other proteins are regulated. Further, whether these endocytic proteins assemble randomly along the PM or at discrete, pre-defined membrane areas similar to what occurs in the synapse is undefined.

To better understand how the endocytic machinery in hepatocytes is spatially organized and temporally regulated, we utilized confocal microscopy to observe these processes directly in living cultured epithelial cells expressing dynamin 2 (Dyn2) coupled to GFP. Dynamin is a large GTPase that has been implicated in the final stages of clathrin-mediated endocytosis (9–12). In defined in vitro systems, recombinant dynamin alone can sever or deform lipid tubules, indicating that this enzyme has mechanochemical properties that could pinch forming vesicle buds from donor membrane compartments in cells.

Because dynamin is considered a major component of the clathrin-coated pit-generating machinery, we predicted that recording the distribution of the labeled enzyme in cells over time would provide useful information about the function and distribution of the endocytic machinery in hepatocytes. The dynamins encompass a broad family of at least three distinct conventional gene products encoding multiple splice forms, which exhibit tissue-specific expression and reside at different cytoplasmic locations. In this study we tagged and expressed the Dyn2(aa) form, found predominantly in epithelial cells, in a non-transformed hepatocyte cell line derived from rat (Clone 9). Our findings were confirmed in primary rat hepatocytes isolated in culture. As for most epithelial cells, these cell lines do not express Dyn1 or Dyn3, which are found in brain, lung, testis, and heart (13).

We report here that hepatocytes expressing Dyn2(aa)-GFP display a distribution identical to untransfected cells stained with a Dyn2 antibody. Both tagged and endogenous Dyn2 localize to a punctate “lawn” of vesicular structures along the basal PM. Interestingly, interspersed among individual Dyn2 spots are large tubulovesicular structures that sequester the transferrin receptor 1 (TfR1) and stain positive for the endocytic coat proteins clathrin and AP2. Most remarkable is the highly dynamic nature of these Dyn2, clathrin, and AP2 structures. Time-lapse movies of transfected cells revealed that these endocytic regions generate large numbers of discrete endosomal vesicles. These findings suggest that the clathrin-based endocytic machinery maintains a dorsal/ventral distribution even in “non-polarized” cells. Further, hepatocytes, like the synaptic terminal in polarized neurons, can sequester components of the endocytic machinery into functional “hot spots” that act synergistically to tubulate and constrict the PM, thereby facilitating the generation of large numbers of endocytic vesicles.

EXPERIMENTAL PROCEDURES

Reagents

MiniPrep Express™ Matrix and Luria-Bertani medium were from BIO 101, Inc. (Vista, CA). Restriction enzymes were from New England Biolabs (Beverly, MA). Both anti-clathrin (X22) and anti-α-AP2 antibodies were collected from the supernatant of the X22 hybridoma and AP.6 hybridoma cell lines (ATCC, Rockville, MD). TfR1 antibody was purchased from Zymed Laboratories (San Francisco, CA). Anti-TfR1-N, pan-dynamin (MC63 and MC65), and Dyn2 antibodies as described previously (14, 15). An anti-TfR2 antibody was raised against the peptide sequence QWSPRPSQTIYRRVEGPQLENLEEEDREEGE, corresponding to amino acids 13–43 in full-length rat TfR2 (Accession numbers XM_222022). Transferrin and secondary antibodies conjugated to Alexa Fluor 594 or 488 were from Invitrogen (Eugene, OR). Unless otherwise stated, all other chemicals and reagents were from Sigma (St. Louis, MO).

PCR for Plasmid Construction

Two TfR isoform cDNA sequences from GenBank (Accession numbers TfR1 NM_022712; TfR2 XM_222022). The primers designed for TfR1 were TfR1 5′, GCCGCTGCATTGCGGACAGAGGAGGTGCTT and TfR1 3′, GCACAACCAGCTCAAGTCTAGAAACAGACTACCC; and for TfR2 were TfR2 5′, ATGGTCCAAGAAATCCAGAGACCTGTTGCTGAG and TfR2 3′, TCAAAAGTTATTGTCGATGTTCCAAACGTCGCCACT. Full-length cDNA was amplified using the XL PCR kit (Applied Biosystems, Branchburg, NJ). The PCR cycle conditions were as follows: for TfR1, 94 °C for 1 min and 62 °C for 5 min for 28 cycles, followed by 72 ºC for 7 min; for TfR2, 94 ºC for 1 min and 65 °C for 5 min for 28 cycles, followed by 72 °C for 7 min. The PCR fragments were ligated into TA vector pCR3.1 (Invitrogen, Carlsbad, CA). WT Dyn2(aa) -GFP described previously(13).

Cell Culture and Transfection

Clone 9 and Hep3b2 cells (ATCC CRL-1439, Rockville, MD) was described previously (13); HepG2 and HuH-7 cells were described previously (16, 17). These cells are widely utilized to study hepatocyte function, although they are not polarized and do not express most hepatocyte specific proteins. Primary rat hepatocytes (18) in Williams medium E supplemented with ITS, dexamethasone, 10% FBS, 100 U/mL penicillin, and 100 μg/mL streptomycin were incubated at 5% CO₂/95% air at 37 °C. Cells were cultured on 22-mm microscope cover slips for transfections and immunocytochemistry. Transfections were using the LIPOFECTAMINE 2000 Reagent kit (Invitrogen, Carlsbad, CA) with 1 μg plasmid DNA per transfection. The conditions for stable expression of TfR2 -pCR3.1 in Clone 9 cells are as described previously (13).

Endocytosis Assays

Transferrin endocytosis assays were performed as described previously (14, 15).

RESULTS

Dynamin and Other Components of the Endocytic Machinery are Sequestered at Large Discoidal Structures along the Hepatocyte Base

To define the localization and function of GFP-tagged Dyn2 in living cells, we expressed the Dyn2(aa) form, which is expressed in all epithelial cell types examined and associates with clathrin-coated pits at the PM and trans-Golgi network (TGN) (13). Confocal microscopy of fixed, unpermeabilized Clone 9 cells expressing Dyn2(aa)-GFP revealed a modest localization of fluorescence along the dorsal membrane, with most of the labeled Dyn2 situated along the ventral PM. Optical sections along the base of expressing cells (Fig. 1) displayed “lawns” of Dyn2(aa)-GFP spots reminiscent of clathrin-coated pits. In addition to these puncta, we observed that Dyn2(aa)-GFP was also organized into large, flat, tubulovesicular plaques. These structures were observed in many but not all cells, ranged in size from 2–10 μm, and appeared to vary in the number of associated vesicles (Fig.1a,b).

Figure One. Dyn2 is Clustered into Prominent Plaques or “Hot Spots” along the Cell Base.

Clone 9 cells were transfected to express Dyn2(aa)WT-GFP (a) or fixed and stained with a pan-dynamin antibody (MC63) (b). In both images, plaques or “hot spots” 2–10 μm in diameter are observed along the basal PM (arrows). Bar = 10 μm. (c) Electron microscopy of Dyn2-associated plaques labeled with HRP, a fluid marker. Low-magnification image showing an en face view of an HRP-positive cell that displays small endocytic vesicles (arrow heads) and a large, dark tubulovesicular structure with similar dimensions to the plaques observed in living cells expressing Dyn2-GFP. A higher-magnification view of this structure shows that it is composed of numerous anastomosing tubules from which many coated and uncoated vesicle buds (arrows) are formed. Bar = 1.0 μm.

To test if these Dyn2-rich structures were present only in transfected cells as a result of dynamin over-expression or the GFP tag, untransfected cells were fixed and stained with a purified, pan-polyclonal antibody (MC63) to dynamin. These cells displayed a dynamin distribution identical to that of transfected cells. Antibody staining confirmed that endogenous Dyn2, like Dyn2(aa)-GFP, was incorporated into discrete puncta or larger, flat plaques along the cell base (images not shown). This result indicates that the localization and organization of the expressed protein mimics that of endogenous Dyn2.

To define the shape and organization of these structures at the ultrastructural level, electron microscopy was performed on Clone 9 cells. Cells were exposed to 10 μg/ml HRP in culture medium for 45 min before fixation and reaction of the HRP with DAB and H₂O₂. Cells were then fixed, dehydrated, embedded, and sectioned en face to the substrate to ensure that the hot spots were viewed in the same orientation as in the confocal microscopic images (Fig. 1 a,b). Electron micrographs of the HRP-treated cells revealed many spherical, densely labeled endosomes distributed throughout the cytoplasm, similar to what has been observed by others. Most striking were the large tubulovesicular, HRP-positive structures along the ventral PM at the cell base. These structures were comprised of many anastomosing tubules of a remarkably uniform thickness. In many areas these tubules were deformed and compressed to form vesicle-like buds. From these morphological and functional criteria, the tubulovesicular endocytic structures appear to be the hot spots observed in living cells by confocal microscopy.

Because Dyn2 participates in the formation of secretory vesicles from the TGN, caveolae, and clathrin-coated (9, 19) endocytic vesicles from the PM, we needed to define the coat proteins that comprise the large structures. Transfected and untransfected cells were fixed, permeabilized, and stained with antibodies to a variety of endocytic proteins such as clathrin, mannosidase II, AP1, AP2, TGN38, and caveolin-1. Whereas antibodies to mannosidase II, AP1, TGN-38, and caveolin-1 did not stain the large Dyn2 structures (data not shown), antibodies to clathrin and AP2 exhibited a nearly exact colocalization in Clone 9 cells (Fig 2. a′,b′) and primary rat hepatocytes (Fig. 2c′,d′), suggesting that the “hot spots” are endocytic rather than secretory structures.

Figure Two. Dyn2(aa)-GFP “Hot Spots” Colocalize with Clathrin and AP2.

Confocal images of Dyn2(aa)WT-GFP-expressing cells (a,b) that were co-stained with antibodies to either clathrin (a′) or to the adaptor protein AP2 (b′), which associates with endocytic clathrin-coated pits. Strong co-staining of these endocytic markers suggests that the hot spots are endocytic structures (see arrows). Bar = 10 μm. To test if “hot spot” structures form in primary rat hepatocytes, fixed cells were labeled with antibodies to Dyn2 (c,d) and co-stained for either clathrin or AP2 (c′,d′). The “hot spots” were nearly identical in size and shape to those observed in Clone 9 cells. (e,f) Number of “hot spot” structures observed in a variety of different hepatocellular models in the presence or absence of 10% serum. Bar = 10 μm.

To gain an appreciation for the prevalence of these “hot spot” structures, Clone 9 cells were either transfected to express Dyn2-GFP, or fixed and stained with Dyn2 antibodies, and the number of cells exhibiting large Dyn2 positive structures at the cell base were counted. Under normal culture conditions, using 10% fetal bovine serum (FBS), 5% of cells exhibited large Dyn2-positive structures. Interestingly, by eliminating FBS from the media, the number of cells forming these structures increased to near 20% (Fig. 2e,f) and suggests that nutritional status can regulate “hot spot” formation. It is important to note that as “hot spots” appear and disappear over time, longer term monitoring (3 hours) reveals that 50% of these cells form “hot spots” (25/50 cells observed). Thus, it is likely that all cells over extended periods of time form these structures. To test if Dyn2 “hot spots” are formed in other widely studied hepatocyte cell lines, HepG2, Hep3b, and HuH-7 cells were fixed and stained for Dyn2, or transfected to express Dyn2-GFP. Counting of more than 300 cells from each line indicated that these cells displayed “hot spots” similar to that in number to Clone 9 cells (Fig.2 e,f) when cultured under either normal or low serum conditions. Counts of more than 300 primary rat hepatocytes showed a similar percentage of cells with “hot spots” under normal serum conditions (4%). The primary cells did not do well under serum starvation so “hot spot” counts were not performed under these conditions.

“Hot Spots” are Continuous with the Plasma Membrane and Generate Numerous Motile Endosomes in a Dyn2-Dependent Fashion

Clone 9 cells were imaged over a 20–30-min period in an attempt to visualize endocytic vesicle formation as it occurs in living cells. Images of Dyn2(aa)-GFP-expressing cells were captured every 8 s with a Zeiss confocal microscope. Subsequently, these images were assembled into time-lapse movies using NIH Image software. The resulting movies (Fig. 3) were surprising: the endocytic “hot spots” observed in fixed cells were far more active and dynamic than anticipated. Indeed, while many individual Dyn2 punctate spots were static, “hot spots” appeared de novo and generated many vesicles, leading to a reduction in their size (Fig. 3). After 20–30 min, it was not uncommon for “hot spots” to become consumed and disappear entirely (Fig. 3, later time points). Many cells displayed one or multiple (3–5) “hot spots” at any given time and thus may form scores of these structures over the course of several hours. The cell shown in Figure 3b displays two adjacent, discoidal “hot spots.” Over the recording period of 13 min, one of the structures became consumed by vesiculation, leaving only a small residual patch of Dyn2(aa)–GFP spots. Subsequently, the adjacent “hot spot” initiated vesicle formation and spewed a linear track of endocytic vesicles over a 10-min period until it, too, disappeared. Several minutes later, the same cell assembled additional “hot spots” at different locations along the cell base that also generated many vesicles and then disappeared (images not shown). Many of these dynamin-associated vesicles were seen translocating in a linear path, as if along cytoskeletal filaments. This observation suggests that dynamin may remain on the motile vesicle, rather than immediately dissociating as previously predicted (13). We have counted the number of vesicles generated during a typical “hot spot” lifetime (10–15 min) and observe an average steady state release of 4–6 vesicles/min (4 cells counted). Most remarkable is that each “hot spot” undergoes a 2–3 min burst that generates 15–20 vesicles per min during which the structure physically “dissolves”. This is exceptionally rapid when compared to the release of multiple vesicles from conventional clathrin pits (1–2 vesicles/min) (20).

Figure Three. Dyn2-GFP “Hot Spots” Situated at the Basal Membrane are Dynamic Structures.

Time-lapse confocal microscopy of Clone 9 cells transfected with Dyn2(aa)WT-GFP (a, b). “Hot spots” [see (a) arrows and (b) arrow heads] were extremely dynamic, generating numerous vesicles that translocated rapidly to the peripheral cytoplasm. Interestingly, these “hot spots” were transient and repeatedly formed, vesiculated, and then disappeared over a 5–10-min period. See movies 1 and 2. We have calculated that peak vesicle generation from these sites exceeds 15–19/min.

The rapid vesicle formation from Dyn2/clathrin/AP2 “hot spots” suggests that these are endocytic structures that are either continuous with the PM or reside as an internal endocytic sorting compartment. Because the HRP marker used to label “hot spots” by EM (Fig.1c) was internalized by cells over a 45-min time period, it is possible that the endocytic hot spots, while in intimate proximity with the PM, represent internal endocytic sorting compartments. To test if these endocytic structures are distinct from or continuous with the PM, we utilized Ruthenium red (RR), an electron-dense dye, to label the cell surface and any invaginations continuous with the external environment. Because the dye is included in the primary fixative, thereby preventing its internalization, membrane compartments stained with the dye represent extensions of the PM. As expected, cells stained with RR, processed for EM, and sectioned transverse to the substrate showed very dark apical and basal PMs, while internal membrane systems were only lightly stained (Fig. 4a–b). In control cells (Fig. 4a), small patches of dark, tubular invaginations were observed extending from the basal, but not the apical, PM. These structures were nearly identical in dimensions to those observed in the HRP-labeled cells that were sectioned en face (Fig. 1) and appear as a “side view” of the endocytic “hot spots” shown in the previous figures. Most importantly, these endocytic structures were darkly stained with the RR dye, indicating that they are continuous invaginations of the PM.

Figure Four. “Hot Spots” Represent Tubulovesicular Plaques Continuous with the Basal PM.

(a,b) Cross-sectional EM images of Clone 9 cells fixed in the presence of the electron-dense dye Ruthenium red (RR), which stains all membranes continuous with the external milieu. Prior to fixation, cells were microinjected with either injection buffer alone as a control (a) or an affinity-purified, dynamin-inhibitory, polyclonal antibody (b). Control cells displayed several RR-positive membrane structures that extended from the basal plasma membrane (arrows). A higher-magnification view (insert) of one of these structures (boxed region) shows tubulovesicular structures nearly identical to the HRP-positive structures observed in Fig. 1c. Cells injected with the inhibitory antibody displayed a significant increase in the number and complexity of the “hot spot” structures (b, arrows). Microinjection of the same inhibitory antibodies into primary rat hepatocytes, which were fixed in the presence of RR, embedded, and sectioned en face, display similar structures of long plasma membrane invaginations continuous with the external environment (c,d). These membrane invaginations appeared as reticular tubules of uniform diameter (c) or as tubules constricted into attached vesicles (d). For 3-D images of these structures, see Fig. S1. e) Low magnification fluorescence image of Clone 9 cells expressing a phospho-mutant Dyn2Y231/597F that is known to attenuate dynamin function. A substantial number of both large and small “hot spots” have accumulated along the cell base (arrows). f) Graph depicting percentage of WT or Dyn2 mutant expressing Clone 9 cells with “hot spots.” Cells with impaired Dyn2 function exhibit 2–5 times the number of “hot spots” compared to Dyn2WT-expressing cells (300 cells counted for each condition). Bar = 1.0 μm (a–c) or 200 nm (d).

Because GFP-tagged Dyn2 associates with the endocytic “hot spots” (Figs. 1–3), we predicted that disruption of Dyn2 function would alter the number and complexity of the RR-positive PM invaginations. To test this hypothesis, cultured cells were microinjected with one of two purified polyclonal antibodies that we have used in previous studies to inhibit dynamin function. These antibodies were raised against synthetic peptides representing a domain conserved in the dynamin family (MC65) or against the isoform-specific tail domain of Dyn2. Injected cells were allowed to recover for 4–6 h and were then fixed with RR, processed for EM, and sectioned transverse to the substrate (Fig. 4b). Although endocytic invaginations were observed in uninjected cells and in cells injected with buffer or heat-inactivated antibodies (Fig. 4a), cells injected with the native dynamin antibodies displayed many more of these structures that were substantially larger in size and more extensive in length. Indeed, as shown in Figure 4b, the basal PM of cells in which Dyn2 function was inhibited was lined with numerous RR-positive membrane structures.

To further examine the structure of the hypertrophied endocytic invaginations in Dyn2-inhibited cells, MC65 antibody-injected primary rat hepatocytes were fixed, stained with RR, embedded, and thick sectioned for viewing with the high-voltage electron microscope. Our objective was to use the advantage of thick sections (0.2–0.4 μm) combined with RR to better define the effects of dynamin antibodies on hot spot morphology and the relationship of these structures with the PM. We found that inhibition of dynamin function induced several distinct changes in the PM (Figs. 4c,d). Consistent with our previous observations, the invaginated structures were not found uniformly along the PM but in distinct foci. The RR-positive endocytic PM was frequently tubulated in close proximity to the cell surface. These tubules often extended significant distances (5–10 μm) into the cell interior. Although some of these structures appeared to have spiked clathrin coats, many did not.

Interestingly, these tubules were often constricted with numerous varicosities, leading to the formation of a reticularized tubule network with many associated buds (Fig. 4c,d). This vesiculation was often so pronounced as to create endocytic structures that appeared as many vesicle buds attached to tubules, similar to grapes on a vine. These images are consistent with the prediction that dynamin functions at endocytic hot spots to constrict endocytic PM invaginations into discrete vesicles. Antibody-induced inhibition of dynamin function results in the accumulation of a spectrum of tubules and buds at various stages of the vesiculation process. Stereo 3-D images of these structures are provided as supplemental Fig. S1 and reveal the complexity of these very large endocytic structures in comparison to conventional clathrin-coated pits.

Although the EM approach described here revealed dramatic changes in the number and size of endocytic hot spot invaginations with dynamin inhibition, these observations are qualitative in nature. Accordingly, we transfected Clone 9 cells with either WT Dyn2 or Dyn2 bearing loss-of-function mutations (Dyn2K44A, Dyn2 Y231/597F) that inhibit activity. Then, to test if inhibiting Dyn2 function leads to an accumulation of “hot spots”, we examined cells by IF with either Dyn2 or clathrin antibody staining. Indeed, of the 100 cells examined, 5–6% of WT-expressing cells displayed two or more “hot spots” at the time of fixation. This number increased 2–5-fold in the mutant-expressing cells (Fig. 4f). Interestingly, live cell imaging of Clone 9 cells expressing the Dyn2K44A-GFP mutants (Fig. 4f) revealed many more “hot spots” that were substantially less dynamic, persisted for greater periods of time, and generated far fewer vesicles (0.3/min) than those cells expressing the WT Dyn2-GFP (15–20/min) viewed in Figure 3. Taken together, these studies suggest that Dyn2 is not only associated with endocytic “hot spots” but also participates in the morphogenesis of these structures. Thus, inhibition of Dyn2 function may prevent endocytic vesicle formation from these structures, resulting in a marked increase in their number, size, and complexity.

Dyn2 “Hot Spots” Are Endocytic Structures that Selectively Sequester the Type One Transferrin Receptor (TfR1) and Cognate Ligand

Our observations indicate that “hot spots” are composed of endocytic scaffolding proteins and generate large numbers of cytoplasmic vesicles in a Dyn2-dependent manner. Next, it was important to define the cargo internalized by these structures. To this end, Dyn2(aa)-GFP-expressing cells were either co-stained with TfR1antibody (Fig. 5a) or incubated with 5 μg/ml Alexa-594-transferrin (Tf) for 20 min, washed briefly, then fixed and viewed by confocal microscopy. Because “hot spots” appear to be continuous with the PM (Fig. 4), we did not include an acid wash of cells that is routinely used to provide cleaner, more attractive images through the removal of excess surface-bound ligand. Thus, although the absence of an acid wash resulted in more diffuse surface labeling of cells, it did allow staining of the “hot spots”. As shown in Figure 5, a substantial amount of Tf (red) can be seen either on the cell surface or within the cell. Most interesting is the clear colocalization of the membrane-bound Tf ligand with the Dyn2-associated tubulovesicular hot spots at the cell base. Because the delicate arrangement of Dyn2 puncta gives endocytic hot spots a unique appearance, one can easily discriminate the specific association of ligand with these structures from random, non-specific binding. These observations indicate that cell surface receptor/ligand complexes are specifically sequestered within the endocytic “hot spots.”

Figure Five. Selective Sequestration of the TfR1 and Tf ligand to Dyn2 “Hot Spots”.

(a,b) Fluorescence microscopy of Clone 9 cells expressing Dyn2(aa)WT-GFP revealed large, elaborate hot spots at the cell base. These cells were either fixed and co-stained with an antibody to the cytoplasmic tail (15) of the rat TfR1 (a,a′), or prior to fixation, they were incubated in 5 μg/ml Alexa-594-Tf for 20 min (b,b′). In both instances, there was remarkable colocalization of the TfR and its ligand to the Dyn2 hot spots, showing active sequestration of this receptor/ligand complex to the endocytic sites. Bar = 10 μm.

To examine the capacity of endocytic “hot spots” to sequester other receptor/ligand complexes, we tested if the TfR2 receptor, a distinct but related TfR member, also resides within these large endocytic structures. The TfR2 shares limited homology with the TfR1 (33.6% similarity) and is expressed mainly in primary hepatocytes (21). Because Clone 9 cells do not express this receptor form, we transfected these cells with exogenous TfR2 to determine if the receptor colocalizes with clathrin. In Clone 9 cells the TfR1 form showed remarkable incorporation into large, circular “hot spots” (Fig. 6a), whereas TfR2 distributed into small, individual foci that did not overlap with clathrin (Fig. 6b–d). This non-clathrin distribution is consistent with the findings of others, suggesting that the TfR2 is internalized by an endocytic pathway that is independent of clathrin (22). Further, this result supports the premise that the endocytic “hot spots” observed in hepatocytes are indeed selective for specific cargo, even discriminating between receptor/ligand complexes that share significant homology.

Figure Six. Differential Sequestration of TfR1 and TfR2 at the Base of Clone 9 Cells.

(a,a′) IF images of a Clone 9 cell transfected to express the rat TfR1, then fixed and stained with the TfR1 antibody used in Fig. 5. The cell displays numerous large “hot spots” that have sequestered substantial levels of both the expressed and endogenous TfR1 receptor. (a′) Higher-magnification view of boxed region in (a). In contrast, Clone 9 cells transfected to express an exogenous rat TfR2 (b,b′) and stained with a specific rat TfR2 antibody showed these receptors to be widely distributed into punctate spots that did not reside at clathrin “hot spots” or even individual clathrin pits (c,d). These findings suggest that the two related receptors are differentially sorted, sequestered, and internalized by distinct endocytic structures. (b′–d′) Higher-magnification views of the boxed regions in (b–d). Bar = 10 μm.

Discussion

In this study we expressed GFP-tagged Dyn2 in a cultured hepatocyte cell line to better understand the mechanisms supporting the endocytic process. We discovered two distinct populations of dynamin-associated structures along the basal PM: small, static foci that appear to represent conventional clathrin-coated pits, and large (2–10 μm), Dyn2-positive structures that we have termed “hot spots.” Because both of these clathrin-based systems occupy the basal membrane domain, optical microscopy methods needed to be used exclusively in this study. “Hot spots” were found in both transfected and untransfected cells, are associated with clathrin and the adaptor protein AP2 but not AP1, and are functional endocytic structures that internalize Tf and fluid markers. Most strikingly, we found that “hot spots” are tubulovesicular invaginations of the basal PM that generate massive numbers of endocytic vesicles that translocate to the cell interior.

Importantly, “hot spots” appear to be selective for clathrin-based internalization processes because we observed a striking sequestration of the TfR1 to these structures compared to the TfR2. Although both receptors internalize the Tf ligand, the TfR1 is well known to utilize clathrin-coated pits during endocytosis, whereas TfR2 uptake appears to be clathrin independent and may utilize caveolae (21, 23). This finding, and the fact that we do not find caveolin proteins localized to the “hot spots” (data not shown), suggests that these structures are clathrin-based, are prevalent in most cells examined, and responsive to the nutritional conditions of their surroundings as serum starvation can increase the number of cells with “hot spots” by 4–5-fold (Fig. 2e,f). It is important to note that as “hot spots” are ephemeral structures, lasting only 15 to 60 min, a “snap shot” of fixed cells would identify only a portion of the cells forming these structures. Indeed, monitoring “hot spot” formation in cells over 3–4 hours in normal serum reveals that 50% of cells form and consume “hot spots” making these structure more prevalent than first thought.

Assembly of the Endocytic Machinery along the Basal Membrane of Cultured Cells

The incorporation of Dyn2-GFP within discrete clathrin-coated pits and large “hot spots” along the basal membrane of cultured cells was surprising in that we had assumed these structures would be distributed evenly along both the dorsal and ventral PM. It should be noted that most epithelial cells in situ, such as ductular kidney cells or hepatocytes, undergo substantial endocytic activity along the baso-lateral membrane, which is in intimate proximity to the nutrient-rich blood space or sinusoid. In these cells, plasma proteins are internalized for eventual degradation and/or transport and inserted or released into the apical membrane or space.

It is interesting that although the Clone 9 cells used in this study have lost their original sinusoidal and canalicular membrane domains, they have maintained some fundamental polarity. This polarity is expressed, at least in part, by the maintenance of the endocytic machinery along the basal membrane. To ensure that “hot spot” formation is not a peculiarity of this cultured hepatocyte cell line, we examined three widely used hepatocellular culture models (HepG2, Hep3b, HuH-7; Fig. 2f) as well as MDCK cells (data not shown), and primary rat hepatocytes (Fig. 2), by immunofluorescence staining and observed similar structures. As a large proportion of studies focused on the mechanisms of endocytic vesicle formation are performed in the polyploid neoplastic, and highly altered, HeLa cell model, this current study makes a strong attempt to observe endocytic dynamics in a variety of hepatocyte cell models. Most relevant was the observation that these structures can be resolved in 3–4% of primary rat hepatocytes that, when utilized in the first 24–48 hours post isolation, maintain many polarized characteristics (18). The concept that cellular polarity may in someway alter “hot spot” formation and function is something that we have not fully addressed. Although these structures are observed in primary rat hepatocytes, we have found that “hot spot” formation is markedly influenced by the concentration of serum in the culture media (5 fold increase; Fig. 2 e,f). As serum is also a significant factor in maintaining cell polarity we did not attempt to study the variables of polarity on “hot spot” formation.

Mammalian tissues express three conventional isoforms of dynamin, although Dyn2 appears to be the primary, if not exclusive, form expressed in the hepatocyte (13). Detailed PCR analysis of Dyn2 expression in rat liver revealed that the Dyn2 form is expressed as 4–6 splice variants that appear to reside at different cellular locations and may perform distinct functions (9, 13). Indeed, we found that expression of the Dyn2(aa) variant induces the greatest number of endocytic “hot spots” compared to the Dyn2(ba) or Dyn2(bb) forms. To ensure that the GFP tag was not altering the subcellular distribution or function of the expressed dynamin protein, we performed several control experiments. First, we compared the localization of endogenous dynamin in untransfected cells with that of Dyn2(aa)-GFP in transfected cells (data not shown). The staining patterns for dynamin, as well as for clathrin and AP2, were identical between the control and transfected cells, and the fluorescent ligand was sequestered in the hot spots. Second, we demonstrated that endocytosis of Tf in transfected cells was equal to or greater than that observed in control, untransfected cells or in transfected cells expressing untagged Dyn2(aa). Finally, we observed that the GFP tag does not reduce the ability of a mutant Dyn2-GFP protein [Dyn 2(aa)K44A-GFP] to inhibit endocytosis of ligands (data not shown). Expression of a mutant dynamin protein in cells was equally effective in attenuating endocytosis with or without the GFP tag. Thus, these combined observations suggest that the GFP tag does not interfere with the distribution or function of the dynamin to which it is attached.

Advantages of a Sequestered Endocytic Machinery

Initial observations suggesting that clathrin-based endocytosis might occur at concentrated sites came from live mammalian cells that express GFP-tagged clathrin light chain. The formation of coated pits appeared to be restricted to discrete domains of the PM (20, 24, 25) that liberate several clathrin-coated vesicles over short times. Because these spots moved in temporal and spatial synchrony at the surface of cells treated with detergents, it was suggested that these sites are interconnected and positioned by an actin cytoskeletal network that might also act to sequester coat-forming components.

We have found that these sites in cultured hepatocytes are much more extensive than originally reported, represent exceptionally large (2–10 μm) tubuloreticular structures that may form hundreds of nascent vesicles, and are dependent on dynamin function. Thus, it appears that hepatocytes, like neurons, form specialized endocytic domains for the large-scale production of clathrin-coated vesicles. This sequestration and organization at predefined platforms in the hepatocyte is likely to increase endocytic efficiency substantially, as is well known to occur at the neuronal synapse. As depicted by the illustration in Fig. 7 the generation of endocytic vesicles is markedly increased at hotspots (15–20/min ) in comparison to the conventional internalization of clathrin coated pits (<1/min) by providing a site for large scale vesiculation of the PM. The location of these platforms is likely to be dictated by the enrichment of specific lipids into micro-domains and are highly dependent on actin and actin-binding proteins that recruit and stabilize many components of the endocytic machinery, from clathrin and dynamin to endophilin and intersectin to name just a few. In comparison, the formation of a single clathrin vesicle from an isolated site would require the time-consuming process of a sequential recruitment and assembly of many proteins from the cytosol.

Figure Seven. Hepatocyte “Hot Spots” Provide an Efficient Platform for Endocytic Vesicle Generation.

Illustration depicting a tubulated endocytic “hot spot” that has recruited large amounts of Dyn2 (rings), clathrin (spikes), actin (filaments) and many other endocytic proteins to generate many vesicles rapidly. In contrast, a single clathrin-coated pit may generate only a few vesicles and require reformation at other sites.

One might conclude that clathrin “hot spots” have been observed for some time from early electron micrographs taken by Palade, Porter and others. For example, high-magnification images hepatocytes in situ show the PM decorated with individual clathrin-coated pits and vesicles at different stages of maturation. Most striking is that within very small domains of the cell surface (<1.5 μm²) reside 7–8 clathrin-coated pits. This concentration of pits spread over the sinusoidal membrane of a hepatocyte, covering approximately 3800 μm² of smooth surface area, would represent over 26,000 clathrin invaginations! More likely, this image represents a clathrin “hot spot.” Clearly, not all clathrin-mediated endocytosis in the hepatocyte occurs from “hot spots,” and many questions remain regarding these structures. For example, what are the signals and structural cues that determine when and where “hot spots” form, and might these “hot spots” be “hijacked” by invading viruses or pathogens during infection? Future studies are needed to fully elucidate the properties of these intriguing structures.

Supplementary Material

Supp Fig S1. Figure S1. Inhibition of Dyn2 Function Induces the Accumulation of Complex Tubulovesicular Endocytic Structures at “Hot Spots”.

Stereo electron microscopic images of Dyn2 antibody injected primary rat hepatocytes as displayed in 2-D in Figure 4. These en face images of RR stained cells show two prevalent morphologies observed at hepatocyte “hot spots” along the cell base. (a). Large and complex reticular membrane invaginations comprised of cross-linked tubules of uniform diameter with numerous vesicle buds can be seen extending from the endocytic tubules. (b) Highly vesiculated tubules that appear as “beads on a string” suggesting that the final stages of pinching and vesiculation at these endocytic tubules has been inhibited. Color 3-D glasses are needed to view these images.