Geldanamycin stimulates internalization of ErbB2 in a proteasome-dependent way

It is well-established that the receptor tyrosine kinase ErbB2 is overexpressed in some cancers and has a long half-life at the plasma membrane, where it appears to be the preferred heterodimerization partner for the other ErbB family members (Baulida et al., 1996; Graus-Porta et al., 1997; Wang et al., 1999; Yarden, 2001; Yarden and Sliwkowski, 2001). ErbB2 is associated with a more aggressive disease, and it is a validated drug target in breast cancer (Hynes and Stern, 1994; Klapper et al., 2000; McCann et al., 1991; Yarden, 2001). Thus, it is important to improve our understanding of the mechanism(s) that underlie possible downregulation of ErbB2.

Inhibition of the chaperone HSP-90 with geldanamycin leads to downregulation of many HSP-90 client proteins, of which ErbB2 is among the most affected (Neckers and Ivy, 2003). HSP-90 inhibition causes ErbB2 to become ubiquitinylated by the co-chaperone CHIP (Xu et al., 2002; Zhou et al., 2003), however the subsequent events have remained elusive and several different cellular mechanisms have been suggested to take part in the downregulation of ErbB2. Assuming that ErbB2 is a recycling receptor (Austin et al., 2004; Yarden, 2001), geldanamycin could cause a shift in the traffic of ErbB2 from a recycling pathway to a degradative, lysosomal pathway as suggested by Austin and co-workers (Austin et al., 2004). However, if ErbB2 is an internalization-resistant receptor (Hommelgaard et al., 2004; Longva et al., 2005; Wang et al., 1999), then it is more plausible that geldanamycin stimulates the internalization of ErbB2 from the plasma membrane, resulting in delivery to lysosomes. Additionally, Tikhomirov and Carpenter observed that ErbB2 becomes fragmented after geldanamycin stimulation and they proposed a model involving cleavage of the ErbB2 kinase domain by cytoplasmic proteases, followed by lysosomal and proteasomal degradation of the N-terminal and C-terminal parts of ErbB2, respectively (Tikhomirov and Carpenter, 2000). Additionally, proteasomal inhibitors can efficiently rescue ErbB2 from geldanamycin-induced downregulation (Mimnaugh et al., 1996), and this has led to a wide acceptance of proteasomal proteolysis as the mechanism degrading ErbB2 (Citri et al., 2002; Hong et al., 1999; Way et al., 2004; Xu et al., 2001; Zheng et al., 2000). The following questions arise from this. (1) Does geldanamycin induce lysosomal degradation of ErbB2 by reducing its recycling or by increasing its internalization? (2) Is cleavage of the ErbB2 kinase domain required for internalization? (3) What is the role of proteasomal activity in the internalization of ErbB2?

To answer these questions, we have used a non-invasive receptor internalization assay based on confocal microscopy, allowing us to analyze receptor internalization, recycling, transport to lysosomes, as well as cytoplasmic receptor cleavage in both individual cells and cell populations. We show that without geldanamycin treatment ErbB2 is not internalized to any significant level, and that the primary action of geldanamycin is to stimulate internalization of ErbB2 rather than redirecting internalized ErbB2 from a recycling pathway to lysosomes. Moreover, proteasomal activity is required for the internalization of ErbB2 and therefore indirectly for lysosomal degradation of ErbB2. Finally, we show that, although cytoplasmic cleavage of ErbB2 is observed in some cells, this cleavage is not required for ErbB2 internalization.

ErbB2 does not bind any ligands with high affinity, and the ligands that can activate ErbB2 bind to other ErbB family members with higher affinity. Thus, classical receptor internalization assays based on the binding and uptake of iodinated or fluorescent ligand are of little use, as they might measure internalization of other ErbB family members instead of ErbB2. Alternative strategies, based on labeled antibody binding and internalization followed by acid stripping or quenching of surface-bound antibody, require harsh conditions for efficient stripping, and importantly the assays may themselves induce crosslinking and internalization of the receptors. We therefore used a noninvasive ErbB2 internalization assay with no preparative steps before fixation. First, SK-BR-3 cells are fixed, and surface ErbB2 is labeled (termed ErbB2_surface hereafter). Next, cells are permeabilized and ErbB2 labeled again with the same primary antibody (termed ErbB2_total hereafter) (Fig. 1A). Confocal microscopy of cells stained in this manner revealed that ErbB2_surface and ErbB2_total colocalize at exposed surfaces (Fig. 1B, left panel), confirming that internalized ErbB2 does not normally accumulate intracellularly to any detectable degree, and that the amount of ErbB2 in the biosynthetic pathway is very small. Internalization of ErbB2 induced by extensive crosslinking (Guillemard et al., 2005; Hommelgaard et al., 2004) caused a fraction of the receptor population to become internalized and inaccessible to ErbB2_surface staining (Fig. 1B, right panel). Following geldanamycin stimulation of SK-BR-3, T47D or MCF7 cells, a part of ErbB2 relocated from the surface to intracellular vesicles (Fig. 1C, supplementary material Fig. S2, not shown). It is noteworthy that the surface staining often appeared punctuated after geldanamycin stimulation, suggesting that the receptor aggregated prior to internalization (Fig. 1C).

For a non-biased study of a larger cell population, we used a semiquantitative approach based on 2D scatter diagrams of the pixel intensities of individual cells. First, we analyzed ErbB2_surface compared with ErbB2_total staining (exemplified in Fig. 1D) in untreated cells, which because of the high degree of colocalization yielded a linear correlation with a high Pearson's correlation coefficient (r²) in the 2D pixel scatter diagram (Fig. 1D). In cells in which a part of ErbB2 was internalized due to geldanamycin treatment, the staining of intracellular structures with only ErbB2_total resulted in reduced correlation and a lower r² (Fig. 1D, right panel). Thus, reduced r², when comparing ErbB2_surface and ErbB2_total staining, is a quantitative measure of ErbB2 internalization and termed r²_i. (For validation, see Fig. S1 in supplementary material.)

When measuring the r²_i values after geldanamycin stimulation over a 4 hour time course, we observed a gradual decrease in average r²_i (Fig. 1E). However, at later time points the amount of intracellular ErbB2 varied widely from cell to cell resulting in a broad range of r²_i values. The variation in the r²_i values did reflect actual differences in the cells rather than uncertainties from the quantification, and could also be observed in cells where internalization had been induced by crosslinking antibodies (Fig. 1F).

To confirm that geldanamycin induces delivery of ErbB2 to lysosomes as previously reported (Austin et al., 2004; Longva et al., 2005; Tikhomirov and Carpenter, 2000), we showed that ErbB2 colocalizes both with transferrin in endosomes as well as cathepsin D in late endosomes/lysosomes (Fig. 2A,B). To further establish the nature of the ErbB2-labeled intracellular structures seen after geldanamycin stimulation, immunogold labeling of ultracryosections was used. After 2 hours of geldanamycin treatment, typical late endosomes/lysosomes of the multivesicular body (MVB) type were ErbB2-labeled, and the gold particles were confined to the inner vesicles of the structures rather than to the limiting membrane, as expected for proteins traveling along the degradative pathway (Fig. 2C). Double-labeling experiments confirmed that these ErbB2-labeled late endosomes/lysosomes also contained cathepsin D (M1 & M3 in Fig. 2D). In contrast, very dense (terminal) lysosomes that were heavily labeled for cathepsin D only contained little ErbB2, presumably because internalized ErbB2 had been degraded (not shown). In control cells not treated with geldanamycin, the intracellular labeling for ErbB2 was very low and labeled MVBs very rarely seen (not shown).

To confirm that degradation of ErbB2 actually took place in lysosomes we co-treated the cells with 200 nM bafilomycin, an inhibitor of lysosomal degradation (Van Deurs et al., 1996). SK-BR-3 or T47D cells that were both treated with bafilomycin and geldanamycin did indeed accumulate more internalized ErbB2 than cells treated with geldanamycin alone (Fig. 2E, supplementary material Fig. S2).

To examine whether the lack of intracellular ErbB2 revealed by our non-invasive internalization assay was due to efficient recycling from endosomes (Austin et al., 2004) or to impaired internalization (Hommelgaard et al., 2004; Longva et al., 2005; Wang et al., 1999), we treated the cells with 10 μM monensin, a well-established inhibitor of recycling (Basu et al., 1981; Stein et al., 1984). In SK-BR-3 or T47D cells incubated with monensin for 2 hours to block receptor recycling, only very little ErbB2 was seen intracellularly (Fig. 3A, left panel, supplementary material Fig. S2) as reported by others (Longva et al., 2005). Although a few cells in the population had increased amounts of intracellular ErbB2 and decreased r²_i values, the two populations did not differ significantly (P=0.24, two-sided Mann-Whitney U-test) (Fig. 3B, compare black bars in the upper and lower graph). This makes the possibility of efficient recycling less likely as a general mechanism. Importantly, when geldanamycin was added to SK-BR-3 or T47D cells where recycling was blocked with monensin the amount of intracellular ErbB2 increased a lot (Fig. 3A, right panel, supplementary material Fig. S2). This was reflected in the r²_i values of the population (Fig. 3B, lower graph), showing that the main effect of geldanamycin is to increase internalization of ErbB2.

The increase in ErbB2 internalization suggests that the properties of ErbB2 residing in the plasma membrane change after geldanamycin treatment. We therefore performed FRAP analysis of ErbB2-CFP before and after geldanamycin treatment. Live-cell imaging revealed that in SK-BR-3 cells a high amount of ErbB2-CFP resides in microspikes and irregular ruffles of the plasma membrane, but not in intracellular vesicles (Fig. 4A, upper panel). After geldanamycin treatment ErbB2 association with these membrane protrusions became much less pronounced (Fig. 4A lower panel), and ErbB2 had a significantly increased mobile fraction (Fig. 4B). The cells with the highest endocytosis of ErbB2 also had the highest mobile ErbB2-CFP fractions (Fig. 4C). This supports the notion that ErbB2 is released from a retention mechanism prior to internalization.

To evaluate whether geldanamycin treatment also influenced the distribution of ErbB2 in the plasma membrane, we used a pre-embedding electron microscopy approach (Hommelgaard et al., 2004). Whereas ErbB2 is preferentially associated with membrane protrusions in control SK-BR-3 cells (Fig. 4D), it frequently had a more even distribution in the plasma membrane after geldanamycin treatment (Fig. 4E,F). Importantly, in contrast to the control situation where ErbB2 was almost completely excluded from clathrin-coated pits (0.05% in coated pits), ErbB2 was much more frequently found in clathrin-coated pits and various membrane invaginations after geldanamycin treatment (0.8% in coated pits) (a total of 4393 and 1596 gold particles were counted from 20 and 42 different cell profiles in control or geldanamycin-treated cells, respectively). However, this redistribution of ErbB2 was less pronounced than after antibody crosslinking of ErbB2 where most ErbB2 was found in the bulk membrane between the protrusions, in clathrin-coated pits and other invaginations (Fig. 4G) (Hommelgaard et al., 2004). Taken together, the studies with FRAP and electron microscopy show that geldanamycin influences the distribution of ErbB2 in the plasma membrane in a way that favors ErbB2 internalization.

To study if ErbB2 became cleaved before being internalized, we extended the internalization assay described in Fig. 1 with a simultaneous labeling of the cytoplasmic C-terminus of ErbB2 (termed ErbB2_Cterminal hereafter) (Fig. 1A). Untreated SK-BR-3 cells exhibited a high degree of overlap between ErbB2_total and ErbB2_Cterminal stainings at the cell surface (Fig. 5A). Importantly, the geldanamycin-treated cells exhibited more diffuse blue ErbB2_Cterminal staining of the cytoplasm, showing that some cleavage takes place (Fig. 5A). This cleavage was confirmed with another antibody against the C-terminal part of ErbB2 (Fig. 5B), and in SK-BR-3 cells transfected with ErbB2-CFP (Fig. 5C).

Triple labeling that simultaneously visualized internalization and cleavage of ErbB2 in unstimulated cells revealed colocalization of all three stains at the cell surface (represented in white in Fig. 5D). Although some geldanamycin-treated cells had diffuse cytoplasmic ErbB2_Cterminal staining, most cells had ErbB2-containing intracellular vesicles that also stained positive for ErbB2_Cterminal (Fig. 5D). This suggests that a significant portion of the receptor remains uncleaved before and during internalization, and that cleavage of the kinase domain is not a requirement for ErbB2 internalization (Fig. 5D). To substantiate this finding with a biochemical assay, we affinity-precipitated biotinylated, internalized ErbB2, and used the purified protein for western blotting. As expected, this showed that geldanamycin induced a large increase in the amount of internalized ErbB2 while having no effect on internalization of the transferrin receptor. Importantly, the majority of internalized ErbB2 remained full length, supporting the idea that cleavage is not a requirement for ErbB2 internalization (Fig. 5E). To further support this notion, we performed immunogold double-labeling of ultracryosections with antibodies against the extracellular portion (the N-terminus) and intracellular portion (the C-terminus) of ErbB2. The plasma membrane as well as MVBs contained both the intra- and extracellular parts of ErbB2 (Fig. 5F,G). Moreover, the ratio between N- and C-terminal ErbB2 labeling seen at the cell surface and in MVBs was largely the same.

The amount of ErbB2 cleavage could be quantified from ErbB2_total and ErbB2_Cterminal colocalization in a manner similar to r²_i (Fig. 1). The resulting Pearson's correlation coefficient was termed r²_c, and a reduced r²_c indicated increased cleavage. After geldanamycin treatment a large part of the cell population had low r²_c values and the average r²_c was significantly lower than in the control cell population (Fig. 6A, upper panel).

To study if geldanamycin-induced lack of colocalization between the intra- and extracellular parts of ErbB2 could be inhibited by the pancaspase inhibitor zVAD-fmk as reported for cleavage of the ErbB2 kinase domain (Tikhomirov and Carpenter, 2001; Tikhomirov and Carpenter, 2003), SK-BR-3 cells were either treated with 50 μM zVAD-fmk and geldanamycin or geldanamycin alone. Treatment with zVAD-fmk could suppress the reduced r²_c values observed after geldanamycin treatment to the level of cells only treated with zVAD-fmk, suggesting that zVAD-fmk could inhibit cytoplasmic cleavage of ErbB2 in this assay too (Fig. 6A, lower panel). It was noteworthy that zVAD-fmk did not inhibit internalization, supporting the idea that cleavage is not needed for internalization (Fig. 6B).

When a 2D dot plot of the r²_i and r²_c values from each cell at different time points was made, the dots were markedly spread out at each time point (Fig. 6C). For instance, after 2 hours of geldanamycin treatment, the cells ranged from a high degree of ErbB2 internalization and cleavage to virtually unaffected cells with r²_i and r²_c values resembling those of untreated cells (Fig. 6C).

Oddly, when cells were co-treated with zVAD-fmk and geldanamycin ErbB2 tended to accumulate more in internal vesicular structures than when cells were treated with geldanamycin only (compare panels C and D in Fig. 6). However, 50 μM zVAD-fmk has been reported to inhibit cathepsins as well (Foghsgaard et al., 2001; Rozman-Pungercar et al., 2003), suggesting that this accumulation might be due to reduced lysosomal proteolysis. In cells treated with inhibitors of lysosomal proteolysis such as bafilomycin (Fig. 6E), Ca074-Me (not shown), or zVAD-fmk (Fig. 6D) an inverse correlation existed between r²_i and r²_c in those cells most affected by the geldanamycin treatment. Thus, the cells with most internalized ErbB2 also tended to have the least ErbB2 cleavage and vice versa, supporting the idea that these events are at least in part competing processes. However, such an inverse correlation was not evident when the lysosomal degradation was not inhibited (Fig. 6C) possibly because of degradation of internalized ErbB2.

Western blots of SK-BR-3 cells treated with geldanamycin with or without the proteasome inhibitor lactacystin showed that lactacystin could abolish ErbB2 degradation (Fig. 7A). To examine if proteasomal activity is needed for internalization of ErbB2 rather than for ErbB2 degradation as such, SK-BR-3 cells were either pre-treated with 10 μM lactacystin or control medium for 1 hour followed by addition of geldanamycin and incubation for 4 hours. Strikingly, the geldanamycin-induced internalization of ErbB2 was almost completely inhibited by lactacystin (Fig. 7B,C), suggesting that the proteasomal activity is involved in events upstream of ErbB2 internalization and lysosomal degradation. Similar results were found when T47D cells were used (supplementary material Fig. S2). In addition, no redistribution of ErbB2 was seen by electron microscopy when SK-BR-3 cells were incubated with lactacystin in addition to geldanamycin (data not shown). In order to study whether lactacystin also reduced the geldanamycin-induced lysosomal degradation of ErbB2, we co-treated the cells with 200 nM bafilomycin, geldanamycin and lactacystin. In cells co-treated with all three compounds very little ErbB2 accumulated in the intracellular compartments compared to cells co-treated with bafilomycin and geldanamycin only, confirming that lactacystin reduces the geldanamycin-induced lysosomal degradation of ErbB2 as a consequence of reduced internalization (Fig. 7B,C). Furthermore, lactacystin also inhibited the effect of geldanamycin in cells where recycling was inhibited with monensin, suggesting that proteasomal activity is required at the level of internalization rather than recycling (Fig. 7D).

In the present study we show that geldanamycin stimulates the internalization of ErbB2 rather than redirecting the receptor from a recycling to a lysosomal pathway. Importantly, the internalization occurs in a proteasome-dependent way and subsequently leads to degradation in lysosomes, whereas cleavage of the ErbB2 kinase domain is not required for internalization.

It has been reported that ErbB2 becomes internalized in unstimulated cells, but efficiently recycles to the plasma membrane (Austin et al., 2004). However, this conclusion was mainly based on the binding of labeled herceptin (a humanized monoclonal antibody against ErbB2, also known as trastuzumab) to live cells followed by removal of non-internalized herceptin. Using an identical setup, Longva et al. (Longva et al., 2005) found that herceptin did not become internalized, but that a significant amount of herceptin remained at the cell surface after washing and that this could affect the readout. In concordance with Longva and co-workers, we found that in cells where recycling was inhibited by monensin, very little ErbB2 accumulated in intracellular compartments, supporting the fact that ErbB2 is not internalized and recycled to any significant level. Austin et al. furthermore found that geldanamycin interferes with the proposed recycling of ErbB2 leading to accumulation of more intracellular ErbB2 along the degradative late-endosome-lysosome pathway (Austin et al., 2004). In opposition, we report here that the amount of ErbB2 that accumulates in intracellular structures was far higher in cells that had been co-treated with geldanamycin and monensin than in cells treated with monensin alone, suggesting that the effect of geldanamycin is mainly to increase ErbB2 internalization. This internalization depends on proteasomal activity, because lactacystin can inhibit the geldanamycin-induced accumulation in cells with monensin-blocked recycling.

It is well established that divalent antibodies can induce internalization of membrane proteins (Huet et al., 1980; Schreiber et al., 1983; Schwartz et al., 1986), we therefore recommend that caution is taken when exposing living cells to multivalent probes, especially when the valence is increased with antibodies combined with gold particles or secondary antibodies. Even monovalent Fab-fragments could potentially induce conformational changes or affect protein interactions (Agus et al., 2002; Hipfner et al., 1999; Mijares et al., 2000). This emphasizes the need for non-invasive assays such as the one we have described here when studying receptor internalization.

In concordance with the geldanamycin-induced changes in the behavior of ErbB2 at the plasma membrane, we found by FRAP that the mobile fraction of ErbB2-CFP is increased in response to geldanamycin. The overall increase in the mobile fraction in response to geldanamycin treatment is limited; however this fits well with the fact that internalization of ErbB2 is a relatively slow process. Nonetheless, it was possible to observe a higher mobile ErbB2 fraction in the cells with most internalization, altogether suggesting that ErbB2 is released from retaining interactions. Despite this, immobilization cannot be the only mechanism retaining ErbB2 from endocytosis. If the retention from endocytosis relied on a stable interaction with immobile constituents of the cell, then ErbB2 should be more readily endocytosed in unstimulated cells, where roughly half of the ErbB2-CFP population is mobile. Accordingly, either the postulated retention mechanism might be mobile itself or its interaction with ErbB2 is dynamic.

Geldanamycin caused a redistribution of ErbB2 on the cell surface, from protrusions to the bulk membrane, although less pronounced than after ErbB2 crosslinking. This supports the notion that the preferential localization to protrusions plays an important role in keeping ErbB2 away from endocytic membrane domains (Hommelgaard et al., 2004). Importantly, 0.8% of the ErbB2 gold particles were present in coated pits after geldanamycin treatment (a 16-fold increase compared with the control situation). It is generally assumed that coated pits occupy 1-2% of the cell surface area. This means that if a receptor does not have an internalization signal mediating its concentration in coated pits, about 1-2% of the plasma membrane receptor population would be expected to be present in coated pits at any given time point. Thus 0.8% of ErbB2 found in coated pits after geldanamycin treatment is likely to represent a fraction of the total plasma membrane ErbB2 population able to diffuse without restraints in the membrane including coated pits. Coated pits can be formed and internalized within a few minutes (Marsh and McMahon, 1999), therefore the fraction of coated-pit-localized ErbB2 can easily account for the amount of internalized receptor we found after geldanamycin treatment for 2 hours. However, more detailed calculations are not possible because the release from endocytic restraints and the uptake rate are not likely to be constant during the 2 hours of geldanamycin stimulation. Clathrin-independent endocytic mechanisms could play a role in ErbB2 uptake as well.

Tikhomirov and Carpenter (Tikhomirov and Carpenter, 2000; Tikhomirov and Carpenter, 2001; Tikhomirov and Carpenter, 2003) proposed that geldanamycin induces cleavage of the ErbB2 kinase domain in the cytosol. This could potentially release ErbB2 from restraining protein-protein interactions or expose cryptic motifs that stimulate internalization. However, 50 μM zVAD-fmk caused an intracellular accumulation of ErbB2 in geldanamycin-stimulated cells resembling that of cathepsin-inhibition or inhibition of lysosomal acidification, thus this concentration of zVAD-fmk might reduce ErbB2 degradation by inhibiting lysosomal proteolysis. Both extracellular and cytoplasmic parts of ErbB2 can be internalized and colocalized to endosomes. Additionally, purification of internalized biotinylated ErbB2 confirmed that the majority is uncleaved, further suggesting that ErbB2 is endocytosed as full-length protein and that cleavage is not a prerequirement for internalization. However, we cannot exclude the fact that cleavage of one ErbB2 in a oligomer complex can induce internalization of the entire complex, and therefore several uncleaved ErbB2 molecules. Moreover, some geldanamycin-stimulated cells actually had little colocalization between the intracellular and extracellular parts of ErbB2, and this cleavage was inversely related to internalization in cells co-treated with cathepsin inhibitors or bafilomycin, suggesting that these events are partially competing.

The results reported here raise two important questions. By what means is ErbB2 retained from endocytosis in unstimulated cells? And how does proteasomal activity facilitate the geldanamycin-induced internalization of ErbB2? Internalization of the NMDA receptor GluR1 has been suggested to depend on preceding proteolytic breakdown of the PDZ-protein PSD-95 or an associated factor (Bingol and Schuman, 2004). PDZ proteins are important regulators of subcellular localization, and the PDZ proteins PICK1 and ERBIN have previously been described to bind ErbB2 and affect its signaling and localization in polarized cells (Borg et al., 2000; Jaulin-Bastard et al., 2001). However, in the case of ErbB2 it seems unlikely that the factor retaining ErbB2 from endocytosis is a PDZ protein, because ErbB2-CFP is internalized as reluctantly as wild-type ErbB2, despite its altered C-terminus that is incapable of PDZ-domain binding (our unpublished results). We previously found that the retention mechanism depended on neither the actin cytoskeleton nor lipid rafts (Hommelgaard et al., 2004). Thus, future studies will have to address the nature of the retention mechanism, and if proteasomal degradation of a responsible factor is needed for geldanamycin-induced internalization to occur. Alternatively, the effect of lactacystin on ErbB2 internalization could be due to the endocytosis machinery competing with the proteasome for free ubiquitin, which is then sequestered when the proteasome is inhibited with lactacystin. Similarly, Leithe and Rivedal have found that proteasomal activity is needed for monoubiquitinylation of connexin-43 and subsequent endocytosis (Leithe and Rivedal, 2004). However, Mimnaugh and co-workers have found that ErbB2 is still ubiquitinylated in cells treated with lactacystin and geldanamycin (Mimnaugh et al., 1996), suggesting that this is not the case.

In summary, we propose that geldanamycin-stimulated downregulation of ErbB2 consists of several phases: proteasome-dependent release of the receptor from its endocytic restraints; increased receptor mobility; redistribution of ErbB2 in the plasma membrane favoring endocytosis by random entrapment; internalization; and degradation in the lysosomes. In parallel to this, cleavage of the ErbB2 kinase domain occurs in some cells (Fig. 8).

The SK-BR-3 and T47D breast cancer cell lines were obtained from the American Type Culture Collection (Manassas, VA). The cells were grown in T25, T75 or T150 flasks (Nunc, Roskilde, Denmark) and incubated at 37°C, 5% CO₂ in DMEM (SK-BR-3) or RPMI-1640 (T47D) both without Phenol Red (Invitrogen, Carlsbad, CA) supplemented with 10% fetal calf serum (FCS) (Biochrom KG, Berlin), 2 mM glutamine, 10 U/ml penicillin and 10 μg/ml streptomycin (all from Invitrogen).

A construct encoding ErbB2 was kindly provided by Daniel Donoghue (Bell et al., 2000). ErbB2 was PCR amplified from this using the primers 5′-GGCGCGGCTA GCATGATCAT CATGGAGCTG-3′ and 5′-CCGCGCGGTA CCTCATACAG GTACATCCAG-3′ (TAG Copenhagen, Denmark) containing NheI and KpnI restriction sites. This product was digested with NheI and KpnI restriction enzymes (New England Biolabs) and subcloned into pcDNA3.1. The pcDNA3.1-ErbB2 construct was digested with HindIII and NheI restriction enzymes and ligated into the pECFP-N1 vector (Clontech, Palo Alto, CA), and the stop codon was removed by mutagenesis.

Cells were plated on eight-well chamber slices (Lab-Tek, Naperville, IL). Control cells were incubated with DMEM-HEPES buffer with 0.2% w/v BSA and 2 mM glutamine for the indicated time at 37°C, washed with PBS, and fixed in 2% paraformaldehyde in PBS (PFA) for 20 minutes at room temperature (RT). Where indicated, cells were incubated with 50 μg/ml Alexa Fluor 633-conjugated transferrin (Molecular Probes, Eugene, OR), 3 μM geldanamycin (Sigma-Aldrich, St Louis, MO), primary anti-ErbB2 mouse monoclonal antibody (Sc08, Santa Cruz Biotechnology, 1:100) plus secondary goat anti-mouse antibody (M8642, Sigma-Aldrich, 1:100), 10 μM lactacystin (Sigma-Aldrich), 100 nM bafilomycin (Sigma-Aldrich), 250 μM ALLN (Sigma-Aldrich), 50 μM z-Val-Ala-DL-Asp-CH₂F (zVAD-fmk, Bachem, Bubendorff, Switzerland), and/or 100 μM CA-074-Me (Peptides International, Louisville, KY) in DMEM-HEPES buffer with 0.2% w/v BSA and 2 mM glutamine for the indicated time at 37°C before fixation in 2% PFA. Where indicated, cells were immunostained before permeabilization the following way: non-specific binding was blocked with 5% goat serum (DAKO, Glostrup, Denmark) in PBS for 20 minutes, and cells were incubated with primary antibody (Sc08, 1:400) in 5% goat serum for 1 hour at RT, rinsed three times with PBS, incubated with secondary Alexa Fluor 488-conjugated goat anti-mouse antibody (GαM-488) (1:400, Molecular Probes) in 5% goat serum for 30 minutes at RT, and rinsed three times with PBS. All cells were permeabilized and blocked in blocking buffer (5% goat serum with 0.2% saponin) for 20 minutes at RT, incubated with primary antibodies in blocking buffer for 1 hour at RT, rinsed three times with PBS, incubated with secondary antibody in blocking buffer for 30 minutes at RT, rinsed three times with PBS and mounted with Fluoromount G (Southern Biotechnology Associates, Birmingham, AL). The primary antibodies used were Sc08 (recognizing an extracellular epitope, 1:400), 2242 (rabbit polyclonal anti-ErbB2 raised against a synthetic peptide corresponding to residues surrounding Tyr1222 of human ErbB2, 1:200, Cell Signaling Technology, Beverly, MA), Ab-1 (rabbit polyclonal anti-ErbB2 raised against a synthetic peptide corresponding to positions 1243-1255 (Gullick et al., 1987), 1:100, Neomarkers, Freemont, CA), A561 (rabbit polyclonal anti-cathepsin D, used at 1:100, DAKO), or A11122 (rabbit polyclonal anti-GFP, used 1:1000, Molecular Probes); the secondary antibodies used were Alexa Fluor 568-conjugated goat anti-mouse (GαM-568) and/or Alexa Fluor 633-conjugated goat anti-rabbit (both 1:400, Molecular Probes). The slides were examined with an LSM 510 Meta confocal microscope (Carl Zeiss, Jena, Germany) equipped with 40× and 63× objectives, and using the LSM software v. 3.2 (Carl Zeiss) and occasionally Adobe Photoshop 7.0.

Several images of randomly selected regions from each sample were acquired with similar scanning parameters (63× plan-apochromat/1.4 oil, triple track, 824×824 pixels, Zoom=1.8, 1.52 μsecond pixel dwell time, 4× averaging, pinholes 250 μm). Using the LSM software v. 3.2, individual cells were marked as regions of interest, and a fluorescence intensity threshold applied including all fluorescence from cells but not the space between. For each region of interest the software calculated the Pearson's correlation coefficient (r²) for Sc08/GαM-488 (ErbB2 extracellular epitope labeling before permeabilization) and Sc08/GαM-568 (ErbB2 extracellular epitope labeling after permeabilization), which we termed r²_i, as well as the Pearson's correlation coefficient for 2242/GαM-633 (ErbB2 intracellular/C-terminal epitope labeling after permeabilization) and Sc08/GαM-568 (ErbB2 extracellular epitope labeling after permeabilization), which we termed r²_c.

SK-BR-3 cells grown in T25 flasks, either control cells, cells treated with geldanamycin for 2 hours, or with geldanamycin and lactacystin for 2 and 3 hours, respectively, were washed in PBS and fixed in 0.1% glutaraldehyde and 2% PFA in 0.1 M phosphate buffer at RT for 30 minutes. After a wash in PBS the cells were incubated with Sc08 (Santa Cruz Biotechnology) against the extracellular (N-terminal) part of ErbB2 followed by 10-nm-gold-labeled goat anti-mouse antibody (GαM-10) (Amersham Bioscience AB) and further processed for Epon embedding and electron microscopy as previously described (Hommelgaard et al., 2004). Antibody crosslinking of ErbB2 with Sc08 followed by GαM-10 was done as previously described (Hommelgaard et al., 2004).

For detection of intracellular ErbB2, control cells and cells treated with geldanamycin for 2 hours were processed for immunogold labeling of ultracryosections as previously described (Hommelgaard et al., 2004). To detect ErbB2, Sc08 was used, followed by GαM-10. This was in some experiments preceded by incubation with a rabbit polyclonal anti-cathepsin D antibody (DAKO) followed by protein A conjugated to 5 nm gold (PAG-5). To detect the extracellular (N-terminal) and intracellular (C-terminal) parts of ErbB2 simultaneously, the C-terminal-specific rabbit polyclonal antibodies 2242 (Cell Signal Technology) or Ab- 1 (Neomarkers, Freemont, CA) were used followed by PAG-5 or PAG-10, and then the sections were incubated with Sc08 and GαM-10 or GαM-5, respectively.

Cells were plated in T25 flasks. The cells were rinsed twice in ice-cold PBS with Ca²⁺ and Mg²⁺ (PBS-CM) for 10 minutes at 4°C. Sulfo-NHS-SS-biotin (0.5 mg/ml, Pierce) was dissolved in PBS-CM and added to the cells at 4°C on a shaking table. After 20 minutes new 0.5 mg/ml sulfo-NHS-SS-biotin was added to the cells and further incubated at 4°C for 20 minutes. The cells were washed with DMEM-HEPES buffer with 1% w/v BSA and 2 mM glutamine (DMEM-BSA) for 10 minutes at 4°C. Control cells were incubated with DMEM-BSA for 2 hours at 4 or 37°C. Some cells were incubated with either 30 μM geldanamycin or Sc08 diluted 1:100 in DMEM-BSA for 60 minutes at 37°C. The cells incubated with Sc08 were washed and further incubated for 60 minutes at 37°C with GαM-488 diluted 1:400. The treatment was stopped by transferring the tubes back to ice, and the cells were rinsed twice with ice-cold DMEM-BSA. The biotin on the membrane surface was cleaved by incubating the cells in reducing solution [50 mM 2-sodium 2-mercaptoethanesulfonate (Sigma), 100 mM NaCl, 50 mM Tris-HCl pH 8.7, 2.5 mM CaCl₂] for 20 minutes at 4°C, which was repeated twice. The cells were washed three times with PBS-CM. Then, the cells were scraped off in lysis buffer [1% Triton X-100, 50 mM Tris-HCl pH 7.4, 150 mM NaCl, 2 mM EDTA, 10 mM NaF, 1 mM vanadate, and phosphatase inhibitor cocktail 1:100 (Sigma)], lysed for 20 minutes at 4°C, sonicated twice, and centrifuged for 5 minutes at 6000 g at 4°C. Protein concentrations were determined (Bio-Rad) and the samples standardized. Streptavidin-coated beads (20 μl, Sigma) were added to the samples overnight at 4°C. The cells were centrifuged for 30 seconds at 4°C at 16,000 g, the pellet was washed four times in lysis buffer and centrifuged at high speed for 30 seconds at 4°C after each wash. The pellet was then dissolved in lysis buffer and Laemmli buffer with 50 mM DTT added, heated for 5 minutes at 95°C, and finally processed for western blotting.

The samples were resolved on 8% bis-tris-acrylamide gels, transferred to a polyvinylidene difluoride membrane (Amersham Biosciences AB, Uppsala, Sweden), and non-specific binding was blocked with 5% milk powder (Bio-Rad) in PBS containing 0.1% Tween 20 (blocking buffer). Blots were probed with primary antibody (mouse monoclonal anti-ErbB-2, 1:3000, Ab-17, Neomarkers; or mouse monoclonal anti-TfR, 1:1000, 13-6800, Zymed) in blocking buffer, followed by a horseradish peroxidase (HRP)-conjugated goat anti-mouse secondary antibody (1:2000, DAKO, Carpentaria, CA) in blocking buffer. The HRP signal was detected using ECL⁺ (Amersham Biosciences AB) chemiluminescence reagent and Hyper-ECL films (Amersham Biosciences AB).

SK-BR-3 cells were plated on custom-made 60 mm plates with a 0.2 mm cover glass inserted in the centre, allowed to grow for 24 hours, and transfected with 1.0 μg ErbB2-CFP using Fugene 6 (Roche, Basel, Switzerland). After 48 hours the medium was substituted with 37°C HEPES buffer (20 mM HEPES pH 7.5, 140 mM NaCl, 2 mM CaCl₂, 10 mM KCl, 1 mg/ml Glucose), and the cells were studied at 37°C using a heated microscope stage. CFP was excited and photobleached using a 458 nm Argon laser, and CFP fluorescence was separated from background using the Meta detector of a Zeiss LSM510-Meta and linear unmixing. Fluorescence recovery of bleached regions was recorded for ∼40 frames, and compensated for the loss of fluorescence in unbleached regions due to scanning. The data were exported to Excel, corrected for background fluorescence and bleaching, and exported to Origon 7.0 for non-linear analysis using the function for simple diffusion: F(t)=(F(0)+F_∞(t/T_1/2))/(1+t/T) (Yguerabide et al., 1982). The obtained F(0), F_∞ and T_1/2 were used to calculate the diffusion coefficient D and the mobile fraction M_f. The mobile fraction was calculated as: M_f=(F_∞-F(0))/(F_before-F(0)), where F_before is the fluorescence in the region of interest before bleaching. The diffusion coefficient D was calculated as: D=βr²/(4T_1/2), where r is the radius of the bleached region, and the β parameter depends on the percentage of fluorescence bleached and can be found in Table 1 of Yguerabide et al. (Yguerabide et al., 1982).

The degree of internalization was estimated by three people who independently assigned the values 0 (no internalization), 1, 2, 3 (most internalization), or X (not determined) to the images without knowing the identity of the cell images, and the average value was calculated for each cell disregarding Xs. The approach was validated by comparing the obtained values of control cells to geldanamycin-treated cells, which resulted in a highly significant increase in the degree of internalization (data not shown). The geldanamycin-treated cells were then divided into two populations, depending on whether their assigned value was above or below the average degree of internalization.

We appreciate the skilled technical assistance offered by Ulla Hjortenberg, Mette Ohlsen, and Izabela Rasmussen. We thank Silas Bruun, Kirstine Roepstorff and Frederik Vildhardt for valuable discussions. This study was supported by grants from the Danish Cancer Society, the Danish Medical Research Counsil, the Novo Nordic Foundation, and the John and Birthe Meyer Foundation.