The plasma membrane represents an impermeable barrier for most macromolecules. Still some proteins and so-called cell-penetrating peptides enter cells efficiently. It has been shown that endocytosis contributes to the import of these molecules. However, conflicting results have been obtained concerning the nature of the endocytic process. In addition, there have been new findings for an endocytosis-independent cellular entry. In this study, we provide evidence that the Antennapedia-homeodomain-derived antennapedia (Antp) peptide, nona-arginine and the HIV-1 Tat-protein-derived Tat peptide simultaneously use three endocytic pathways: macropinocytosis, clathrin-mediated endocytosis and caveolae/lipid-raft-mediated endocytosis. Antennapedia differs from Tat and R9 by the extent by which the different import mechanisms contribute to uptake. Moreover, at higher concentrations, uptake occurs by a mechanism that originates from spatially restricted sites of the plasma membrane and leads to a rapid cytoplasmic distribution of the peptides. Endocytic vesicles could not be detected, suggesting an endocytosis-independent mode of uptake. Heparinase treatment of cells negatively affects this import, as does the protein kinase C inhibitor rottlerin, expression of dominant-negative dynamin and chlorpromazine. This mechanism of uptake was observed for a panel of different cell lines. For Antp, significantly higher peptide concentrations and inhibition of endocytosis were required to induce its uptake. The relevance of these findings for import of biologically active cargos is shown.
For most polar molecules, the plasma membrane represents an impermeable barrier. It is therefore highly remarkable that some proteins and peptides possess the ability to cross this border and reach the cytoplasm. Among these are transcription factors belonging to the homeodomain family, such as the Drosophila melanogaster Antennapedia homeodomain protein (1), the HIV-1 Tat protein (2) and fibroblast growth factors 1 and 2 (3). Peptides that possess this ability were identified as the protein transduction domains (PTDs) of aforementioned proteins such as the Antennapedia-homeodomain-derived antennapedia (Antp) peptide (4) and the HIV-1 Tat-derived Tat peptide, denoted here as Tat (5). Alternatively, peptides, among them oligo-arginine peptides (6), were designed de novo based on structure activity relationships of PTDs. The attractivity of these peptides, generally defined as cell-penetrating peptides (CPPs), in biomedical research is a consequence of their ability to mediate the import of membrane-impermeable, biologically active molecules such as small interfering RNA, DNA, peptides or entire proteins into the cells ex vivo and in whole organisms [for reviews, see Dietz and Bähr (7) and Snyder and Dowdy (8)].
Initially, the import of these cell-penetrating molecules, proteins and peptides alike, was considered to occur by direct permeation of the plasma membrane (9). This model was based on evidence obtained from cell biological as well as biochemical and biophysical experiments. More recently, it was shown that endocytosis plays a major role in the import of the basic and amphiphilic Antp peptide, the highly basic and arginine-rich R9 and Tat peptides and in the import of the HIV Tat protein itself (10–12). No specific receptor has been implicated in the uptake of these molecules. Instead, in some cases, the initial association with the plasma membrane was attributed to multivalent interactions with cell surface heparan sulfate proteoglycans (13,14).
To this point, however, conflicting results were obtained for the involvement of specific endocytic pathways. Data were presented that supported a role of macropinocytosis (15), clathrin-mediated endocytosis (CME) (11) and caveolae/lipid-raft-mediated endocytosis (16,17).
Contrary to this uptake through endocytosis, for a fluorescein-labeled Tat peptide, Ziegler et al. observed a rapid cellular import into fibroblasts that was heparan sulfate dependent (18). Similarly, Tunnemann et al. reported that the Tat peptide conjugated to a small peptide cargo enters cells by a rapid and endocytosis-independent process, with resemblance to the one described by Ziegler et al., while larger conjugates enter cells by endocytosis (19). They concluded that endocytosis is restricted to high molecular weight complexes. However, this conclusion is in conflict with a number of reports observing a vesicular mode of uptake for the Tat peptide as well (11,20–22).
As an explanation for these apparent discrepancies, in this article, we provide evidence that the peptides use the three endocytic pathways simultaneously. Antennapedia differs from R9 and Tat by the extent that individual processes contribute to import. Moreover, for R9 and Tat above a concentration threshold in the lower micromolar range, both peptides are internalized predominantly through a process that leads to a rapid distribution of peptides into the cytoplasm and nucleus. Our observations suggest that this mechanism is endocytosis independent. Blockage of macropinocytosis and lipid-raft-dependent endocytosis lowers this threshold. This rapid cytoplasmic entry originates from spatially confined zones of the plasma membrane. For Antp, this uptake was observed only for cells treated with the inhibitor of lipid-raft-dependent endocytosis methyl-β-cyclodextrin (MβCD) at significantly higher peptide concentrations. The analysis of the mechanistic basis showed that this uptake is heparan sulfate dependent and sensitive to the protein kinase C (PKC) inhibitor rottlerin, the expression of dominant-negative (DN) dynamin and chlorpromazine (CPZ). The relevance of these results for the application of CPPs as tools in cell biological research is shown using a Smac-derived peptide (23) that potentiates the Fas-dependent induction of apoptosis.
The intracellular distribution of R9 and Tat is concentration dependent
For the three cationic CPPs Antp, R9 and Tat (Table 1), there is a consensus that endocytosis at least strongly contributes to cellular uptake (24). However, conflicting evidence has been presented concerning the role of the individual endocytic pathways. Moreover, some researchers observed a rapid entry of peptides into the cytosol for which endocytosis did not seem to be involved. We reasoned that the apparent discrepancies were a consequence of the different experimental conditions with respect to cell lines, incubation times and peptide concentrations. For this reason, we performed experiments in the presence of different peptide concentrations. Experiments investigating the uptake mechanism of CPP have mostly used these peptides at lower micromolar concentrations, while the applications of CPP–peptide conjugates for interfering with molecular interactions inside the cell have frequently used these molecules in the mean to upper micromolar range (25–28). The selected concentration range for our experiments therefore covered peptide concentrations used in both types of experiments. As a cellular model system, we selected HeLa cells because this cell line has been used in a significant number of previous reports (11,13,17,20,21,29). All three peptides were synthesized as fluorescein-labeled analogues. The cellular uptake was quantified in living cells by flow cytometry. Trypsinization of cells, required to detach the cells from the tissue culture plate prior to flow cytometry, removed peptides merely adsorbed to the outer plasma membrane (11). In addition, the intracellular peptide distribution was investigated by live cell confocal laser scanning microscopy.
Table 1. Primary structures of the peptides used in this studya
All peptides were synthesized as C-terminal peptide amides (-CONH2). Fluo represents 5(6)-carboxyfluorescein.
For Antp, the cellular fluorescence was proportional to the concentration of peptide in the medium (Figures 1A, 2A). With respect to the intracellular fluorescence, the cell population was fully homogeneous for each peptide concentration. Neither saturation nor changes in the intracellular distribution of peptide was observed. Antennapedia was localized predominantly within vesicular structures. Fluorescence could also be detected in the cytoplasm, and this fluorescence also increased with peptide concentration. In contrast, the cellular uptake of R9 and Tat did not increase linearly with peptide concentration (Figures 1B,C, 2A). Both flow cytometry and fluorescence microscopy clearly showed that at concentrations higher than 10 μm, cytoplasmic delivery of peptides was strongly enhanced. Simultaneously, the cell population became very heterogeneous with respect to peptide uptake. Some cells were fluorescent to the point that they were outside the detection range. Other cells showed a moderate uptake, and finally, some cells exhibited only a weak intracellular fluorescence. This heterogeneity in the level of the cell population was reflected by a heterogeneity in the distribution of fluorescence on the subcellular level. With increasing peptide concentration, an increasing number of cells exhibited a clear enrichment of peptide in the nucleus and cytoplasm, with little vesicular fluorescence, resembling the distribution reported by Tunnemann et al. (19).
To exclude that the highly efficient peptide import was a result of damage of the plasma membrane, we tested the membrane integrity using a trypan blue exclusion test (data not shown) and incubation with peptide in the presence of propidium iodide (PI) (Figure 1D). Cells coincubated with PI and R9 at a peptide concentration of 20 μm exhibited efficient peptide entry but no PI staining. However, the subsequent treatment of the same cells with detergent resulted in a strong PI signal within the nucleus and in the simultaneous exit of peptide already 2 min after addition of detergent. These two independent assays clearly show the integrity of the plasma membrane during peptide incubation. To further exclude toxic effects, the viability of cells was also determined after incubation with peptides for 6 h. Using the 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) test, no toxicity was observed for peptide concentrations up to 40 μm (Figure 1E).
R9 and Tat differ from Antp in the concentration dependence of the uptake mechanism
The surprising observation that for R9 and Tat, different peptide concentrations led to clearly distinct phenotypes with respect to intracellular peptide distribution motivated us to ask whether different import mechanisms may be responsible for this effect. To test this hypothesis, we investigated the involvement of the three major endocytic pathways in peptide uptake in dependence on peptide concentration. Pharmacological inhibitors were used to interfere with individual endocytic pathways (30–32). Chlorpromazine was used for the inhibition of clathrin-mediated internalization, 5-(n-ethyl-n-isopropyl)amiloride (EIPA) for inhibition of macropinocytosis and MβCD for disruption of import through caveolae/lipid rafts. The specificity of the inhibitors is an issue of debate, also within the field of CPP uptake (33). For fluorescently labeled tracer molecules [transferrin, dextran and cholera toxin beta subunit (CTB)], each of the inhibitors showed the expected effect (Figure S1). However, for none of the tracer molecules, import could be fully blocked. All three inhibitors alone or in combination, applied at concentrations used in this study, had no toxic effect (Figure S2). The effects of all three inhibitors on the delivery of Antp, R9 and Tat within a concentration range of 2–40 μm were tested by flow cytometry (Figure 2). The concentration-dependent peptide import measured in cells not treated with inhibitor was used as a 100% reference for determining the reduction in uptake caused by the different inhibitors (Figure 2A). For Antp, CPZ and EIPA reduced the cellular fluorescence by about 20% for the entire concentration range. Preincubation of cells with MβCD had no detectable effect. In contrast, for R9 and Tat, the effect exerted by each inhibitor was strongly dependent on peptide concentration. At peptide concentrations <5 μm, CPZ was without effect. However, with increasing peptide concentration, the inhibitory activity of CPZ increased, amounting to about 80% at a peptide concentration of 40 μm. At low peptide concentrations, both MβCD and EIPA inhibited the import of R9 and Tat only slightly if at all. Interestingly, at higher peptide concentrations, the same compounds did not reduce but dramatically enhanced the uptake of these CPPs. We observed a four- to sevenfold increase of cellular fluorescence compared with the fluorescence in the control, not treated with inhibitors. At a concentration of 40 μm, the fluorescence was outside the detection range. These results again confirm a difference between the cellular import of Antp and one of the other two peptides. Moreover, the dependence of the inhibitor effects on peptide concentration strongly suggests that the contribution of the individual uptake mechanisms may in fact depend on peptide concentration.
To test this hypothesis in more detail, the cellular peptide distribution was analyzed at low (2 μm) and high (20 μm) peptide concentrations. For Antp, all three inhibitors led to distinct phenotypes with respect to cellular peptide distribution (Figure 3A). Preincubation of cells with CPZ strongly decreased the small fraction of peptide present in the cytoplasm. The peptide was localized in vesicular structures that were predominantly localized at one side of the nucleus. The inhibition of caveolae/lipid-raft-mediated endocytosis by MβCD resulted in the reduction of cytoplasmic and vesicular fluorescence but simultaneously resulted in the enrichment of peptide at the plasma membrane. The major part of this fluorescence was resistant to trypsin treatment (not shown), explaining why in flow cytometry, no effect of MβCD on cellular fluorescence had been detected. The molecular basis for this observation, that is close association with the plasma membrane or incorporation, could not be resolved. Inhibition of macropinocytosis by EIPA also reduced the internalization of Antp; however, a cytoplasmic as well as a vesicular staining were detected. In this case, intense peptide-loaded vesicles were localized predominantly around the nucleus. Consistent with the results obtained by flow cytometry, the effects of the three inhibitors on uptake were independent on peptide concentration. To support the observation that the three endocytic pathways contribute to the uptake of Antp, we incubated cells with medium containing peptide and the corresponding tracer: transferrin for CME, CTB for caveolae/lipid-raft-mediated endocytosis and dextran for macropinocytosis (Figure 3B). We observed a partial colocalization of Antp with each tracer. However, the level of colocalization with dextran and CTB was strongly time dependent (Figure 3C). At the early stage of endocytosis, more than 50% of Antp-loaded vesicles were positive for transferrin and only 20% showed a colocalization with dextran or CTB, respectively. These observations indicate that CME is the preferred mode of internalization but not the exclusive one. However, the internalization of Antp and transferrin was not fully synchronized. Transferrin uptake was faster than the one of Antp and less than 20% of the transferrin vesicles contained detectable levels of peptide. In contrast, the internalization of dextran and CTB was slower and 60% of dextran- and CTB-loaded vesicles contained Antp. With the progression of endocytosis, the fraction of vesicles containing peptide and dextran/CTB continuously increased. The fusion of vesicles of different origin and the release of their contents into sorting/late endosomes is a likely explanation for this observation. At the early time-point, when one may assume that no merging of vesicles had occurred, the percentages of the Antp-containing vesicle populations added up to about 100%, validating the accuracy of the protocol used for image processing. The results obtained for Antp at a concentration of 10 μm were representative for the whole concentration range.
For the uptake of R9 and Tat at low peptide concentrations, CPZ had no effect (Figure 4). In contrast, at high peptide concentrations, the same inhibitor completely abolished the cytoplasmic and nuclear localization of the peptides. Under the assumption that CPZ is a specific inhibitor of clathrin-dependent endocytosis, this observation provided strong evidence that at high concentrations, this endocytic pathway is a key internalization mechanism used by these peptides, which leads to their cytoplasmic and nuclear enrichment. The inhibition of transferrin uptake confirmed that inhibition of CME is at least one activity exerted by CPZ on HeLa cells. However, CPZ originally was identified as an antipsychotic drug (34), acting as an antagonist of dopamine D2 receptors (35). Moreover, CPZ directly interacts with calmodulin, thereby also interfering with a number of Ca2+-dependent signaling pathways (36). Given the fact that our further experiments (see below) supported a nonendocytic uptake mechanism, in the following we refer to this uptake as CPZ sensitive rather than taking these results as direct evidence for clathrin-dependent endocytosis.
At low concentrations, the CPZ-sensitive pathway did not seem to be involved in the uptake of R9 and Tat. MβCD and EIPA exerted a slight inhibitory effect, leading to phenotypes similar to the ones observed for Antp. However, at high peptide concentrations, the same inhibitors strongly promoted the cytoplasmic and nuclear distribution of peptides (Figure 4). This result was surprising but completely consistent with the data obtained by flow cytometry (Figure 2C,D). Evidently, at high peptide concentrations, the inhibition of endocytosis exerted by these inhibitors directed the peptides toward a more efficient compensatory mechanism.
Having observed that for R9 and Tat, inhibitors of endocytosis lower the threshold for the rapid entry of peptide into the cytoplasm, we decided to investigate the uptake of Antp at even higher concentrations. Antennapedia was applied to HeLa cells at a concentration of 100 μm. In the absence of inhibitors, in spite of a further increase of cytoplasmic fluorescence, vesicular staining was still prominent (Figure 4C). In contrast, in the presence of MβCD, some cells showed a homogeneous cytoplasmic and nuclear fluorescence. However, after about 20-min incubation, morphological changes of the cells were indicative of cell damage. In contrast to R9 and Tat, EIPA was without effect. Nevertheless, these data indicate that the differences between Antp and R9 and Tat may be rather quantitative than qualitative.
To provide direct evidence that application of EIPA and MβCD promoted the uptake of R9 and Tat through the CPZ-sensitive rapid uptake mechanism, HeLa cells were incubated either with MβCD or with EIPA alone or in combination with CPZ (Figure 5A). R9 was applied at a concentration of 10 μm. In control cells, at this concentration, the peptide was predominantly localized within vesicles. Consistent with our hypothesis, at this peptide concentration, CPZ had no effect on peptide uptake. As before, preincubation of cells with MβCD or EIPA led to a pronounced localization of peptides in the cytoplasm and nucleus. However, this phenotype was completely reversed when cells were treated with either MβCD or EIPA in combination with CPZ. The primary inhibitory effects of both MβCD (membrane enrichment) and EIPA (vesicles distributed around the nucleus) were detected again. The MβCD-induced promotion of peptide import and the reversal of this effect by CPZ were confirmed by flow cytometry (Figure 5B). For R9 and Tat, applied at concentrations higher than 5 μm, the inhibition of caveolae/lipid-raft-mediated endocytosis by MβCD enhanced peptide uptake by up to 800 (R9) and 400% (Tat) of controls not treated with inhibitor. This increase was almost completely inhibited when cells were treated with a combination of MβCD and CPZ.
The uptake of R9 at high peptide concentrations originates from spatially restricted membrane regions
Having shown the remarkable efficiency of the CPZ-sensitive uptake mechanism used by R9 and Tat at high peptide concentrations, we aimed at obtaining information about the time–course of this uptake mechanism. Cells were incubated with R9 (20 μm), and the internalization of peptide was imaged by time-lapse confocal fluorescence microscopy, with the peptide still in the medium (Figure 6A). Image acquisition was started about 60 seconds after addition of peptide. Frames were recorded every 30 seconds.
In some cells, areas of higher fluorescence intensity were observed that seemed to emerge from spatially restricted membrane regions. During the next 10 min, further highly fluorescent zones (one or two per cell) developed that were the origin of a rapid spreading of fluorescence throughout the cytoplasm and nucleus. We chose to call these highly efficient internalization platforms nucleation zones (NZs). Almost immediately after formation of an NZ, the peptide was enriched in the nucleus. In agreement with our previous observation, the cell population became heterogeneous with regards to the intracellular fluorescence. The efficient peptide entry strictly depended on the formation of the NZs. Cells lacking these structures did not exhibit any cytoplasmic or nuclear fluorescence. At microscope settings at which usually endocytic uptake was followed, for many cells, fluorescence went into saturation. Therefore, further time-lapse recordings were performed in which saturation was carefully avoided (Figure 6B; Video S1). The uptake kinetics was analyzed by determining the fluorescence over time for circular regions of interest (ROI). For nearly all cells showing the rapid import, the intracellular fluorescence reached a plateau (Figure 6C). However, individual cells strongly varied in the level of fluorescence at which this plateau was reached. In addition, there was no correlation between the uptake kinetics and the increase in fluorescence. In one case, a strong entry of peptide occurred within 4 min, in another case, it was within 10 min. A similar variation in time was observed for cells with a much weaker uptake. Very remarkably in some cases, import was not continuous. Instead, the increase in cellular fluorescence occurred in sequential steps, each having similar kinetics and height. Such a discontinuous stepwise import would be highly unusual for endocytosis.
A correlation of this increase in cellular fluorescence with the image sequences showed that for at least some cells, each step was accompanied with the formation of an NZ and the dissipation of fluorescence from this NZ into the cytoplasm (Figure 6B,C). If no NZ was visible during uptake, this was probably a result of the confocal nature of the image acquisition that restricted fluorescence detection to one section through the cell.
In addition, for some cells, after local enrichment of fluorescence, a bleb-like membrane protrusion was formed that culminated in the pinching-off of a fluorescent vesicular structure. An example for a large bleb is shown in Figure 6D. Concomitant with bleb formation, the rapid uptake of fluorescence and distribution of fluorescence into the cytoplasm and nucleus occurred. In some cases, a fluorescent vesicle was suddenly visible next to an NZ in only one image frame. It is therefore very likely that in many of the cases in which no bleb was detected, this was because of the timing of the image acquisition and/or the confocality of the detection volume.
The compensatory uptake of R9 in the presence of MβCD exhibited the same characteristics (not shown). To exclude (i) that the formation of NZ and rapid cellular uptake were because of the presence of fluorophore and (ii) that the cytoplasmic fluorescence was a result of a concentration-dependent release of proteolytic breakdown products into the cytoplasm, unlabeled R9 (15 μm) was mixed with fluorescein-labeled R9 (5 μm) peptide. The formation of NZs, the import kinetics and the cellular distribution fully corresponded to the one observed for the fluorescein-labeled analogue alone (not shown). In addition, the same formation of NZ, uptake kinetics and distribution were observed for a tetramethylrhodamine-labeled Tat peptide, further excluding a role of the fluorophore in our observations. For MβCD-treated HeLa cells, on incubation with 50 μm Antp, NZ were also observed.
In the NZ, peptides are transiently confined
Fluorescence microscopy had failed to resolve distinct vesicles at the NZs. As an alternative means to obtain information about a possible vesicular confinement of the peptide within the NZ, we compared the mobility of the peptide within and outside the NZs using fluorescence loss in photobleaching [FLIP (37)]. The peptide-associated fluorescein fluorescence was bleached in ROI either inside or outside the NZ (Figure 7). Bleaching inside the NZ resulted in the complete destruction of fluorescence not only in the bleached region but also in the whole cell. In contrast, bleaching outside the NZ depleted all fluorescein fluorescence, except the NZ-associated signal, providing strong evidence for a confinement or association of peptide within these structures. These FLIP experiments had been conducted with peptide present in the medium. To decide whether the NZ-associated fluorescence after bleaching of fluorescence outside the NZ represented peptide that was permanently associated with a subcellular structure or only transiently confined peptide, cells were washed prior to FLIP inside the nucleus. In this case, FLIP resulted in the depletion of all fluorescence (not shown). This result confirms that fluorescence inside the NZs represents a highly dynamic trafficking state of the peptides. This observation also explains why in previous experiments, in which cells were routinely washed prior to microscopy, it had been impossible to detect the NZ.
The formation of NZs is CPZ sensitive and dynamin and heparan sulfate dependent
Having shown that cytoplasmic and nuclear peptide delivery of R9 and Tat could be inhibited by CPZ, we asked whether this inhibitor directly affected the development of NZ. Pretreatment of cells with CPZ abolished the development of NZ and the efficient internalization of peptide (Figure 8A). The rapid import kinetics of the NZ-mediated uptake and the failure to detect vesicles questioned whether the interference of CPZ with this import pathway was in fact a result of the inhibition of clathrin-dependent endocytosis. To address this point by an independent experimental approach, HeLa cells were used in which endocytosis by clathrin-coated pits and caveolae/lipid rafts could be disrupted through the inducible expression of a DN mutant dynamin-2. Dynamin is a guanosine triphosphatase, which is critically involved in the scission of vesicles in these endocytic pathways (38,39). As expected, the expression of DN dynamin inhibited the internalization of transferrin, a classical marker for CME (Figure 8B, lower panels), leading to the enrichment of receptor-bound transferrin at the plasma membrane. In addition, the induction of DN dynamin abolished the development of NZs and consequently the delivery of peptide into the cytoplasm (Figure 8B, upper panels).
For the full-length Tat protein and CPPs, it had been shown that acidification of the endosomal content is required for the release of CPPs into the cytoplasm (12,20,21). We therefore probed for a role of acidification in the NZ-mediated rapid peptide import into the cytoplasm. Contradictory results were obtained for the two well-established inhibitors of endosomal acidification, ammonium chloride and chloroquine (Figure 8C). Treatment of cells with ammonium chloride completely abolished the development of NZs. In contrast, a clear vesicular staining could be detected. In contrast, chloroquine did not interfere with the rapid cellular uptake through NZ.
Moreover, we probed for an involvement of heparan sulfate in peptide uptake. For the uptake of the Tat protein, heparan sulfate proteoglycans were shown to be involved in uptake (40–42). In addition, the interaction of R9 with these molecules was directly shown by biochemical experiments (43). Heparan sulfate was selectively removed from the cell surface by treatment of cells with heparinases. Interestingly, different effects on uptake were observed at high and low concentrations of R9 (Figure 8D). At a low peptide concentration, heparinase treatment was without effect. Neither the number of peptide-loaded vesicles nor their cellular localization was altered. In contrast, at high peptide concentration, the number of cells showing the development of NZs as well as the rapid delivery of R9 into the cytoplasm and nucleus was reduced. Therefore, the contribution of heparan sulfate to peptide uptake is restricted to the highly efficient uptake through NZs.
By electron microscopy, NZs are indistinguishable from the rest of the cell
Light microscopy had failed to provide evidence for an involvement of vesicular structures in the NZ-mediated cellular uptake of nona-arginine and Tat. However, the abolition of NZ formation in cells expressing DN dynamin and the inhibitory effect of CPZ suggested a clathrin-dependent endocytic process. To address the presence of densely packed small vesicles that could not be resolved by fluorescence light microscopy, a fine structural analysis of NZ was performed by electron microscopy. Growth of cells on coverslips structured with co-ordinates enabled a matching of fluorescence and electron microscopy images (Figure 9). Paraformaldehyde fixation required for sample preparation for electron microscopy maintained the distribution of fluorescence typical for NZ. By electron microscopy, the NZs were indistinguishable from the remainder of the cell. Individual clathrin-coated vesicles could be identified within the cells (not shown). However, there was no indication of an enrichment of clathrin-containing structures either inside the cytoplasm or at the plasma membrane.
The internalization of R9 at high peptide concentrations is sensitive to the PKC inhibitor rottlerin
Electron microscopy strongly indicated that uptake through NZ occurs by an endocytosis-independent mechanism. In this case, the activity of CPZ should be because of interference with processes other than assembly of the AP-2 coat on nascent clathrin-coated vesicles (44). As CPZ was shown to interfere with calcium-dependent signaling, we therefore tested a panel of inhibitors reported to also interfere with calcium-dependent and calcium-independent signaling processes. Ro-318220, which inhibits the calcium-dependent PKCα (45), was without effect on the internalization of 20 μm R9 through NZ. The PI-3 kinase inhibitor wortmannin, which had been shown to interfere with several steps in endosomal trafficking including clathrin-dependent processes (46,47), was also without effect. In contrast, preincubation of cells with rottlerin, a compound that had been reported to be a preferential inhibitor of PKCδ (48), albeit with specificity that was questioned later (45), abolished the development of the NZ (Figure 10A). R9 uptake at low peptide concentrations remained unaffected (Figure 10B). Rottlerin had no effect on the endocytosis of transferrin, providing further evidence that the activity of CPZ on NZ-dependent import was not a result of interference with clathrin-dependent endocytosis. In cells not treated with rottlerin, the formation of NZ also had no effect on the endocytosis of transferrin, showing that both clathrin-dependent processes were spatially and functionally distinct (Figure 10A, top panels). Even though these further inhibitor-based analyses do not provide a conclusive picture on the molecular mechanism involved in the NZ-mediated uptake, the inhibition by both CPZ and rottlerin constitutes a distinct pharmacological profile of this process, which may guide the further elucidation of this uptake mechanism.
The formation of NZs is a characteristic of different cell types
Having shown distinct characteristics of the uptake of Tat and R9 through NZ, we next wanted to learn whether this uptake mechanism is a particular characteristic of HeLa cells. For this reason, we selected a panel of cells derived from different species, different tissues and with different growth characteristics: MC57, mouse fibrosarcoma cells; Chinese hamster ovary cells; primary human dendritic cells and human Jurkat T-cell leukemia cells. In all cell lines, we observed a concentration-dependent formation of NZs (Figure 11). In addition, in all cases, the cell population was heterogeneous with respect to the formation of NZ. These results show that the formation of NZ is not only a mechanism present in cancer cells and furthermore that cell attachment is not a prerequisite for this uptake mechanism. For Jurkat cells, we tested whether incubation with MβCD could also induce cytoplasmic delivery of Antp. In correspondence to the observation with HeLa cells, Jurkat cells showed a homogeneous cytoplasmic fluorescence, albeit for a higher fraction of cells.
Relevance of the import mechanism for the cellular activity of cargo peptides
Antennapedia, R9 and Tat have been used as vectors for the cellular delivery of bioactive cargos. Having shown that the cellular uptake and cytoplasmic delivery of Antp are different from the one of the other two peptides, we finally investigated the relevance of these import characteristics for the bioactivity of a peptide delivered through conjugation either to Antp or to R9. We have shown previously that after endocytosis, a large fraction of imported peptides is broken down by endolysosomal proteases (49). One should therefore expect that an uptake mechanism that bypasses the endolysosomal compartment should lead to an increased activity of the cargo peptide. As bioactive cargo, we selected a peptide corresponding to the N-terminal seven amino acids of the proapoptotic protein Smac (23) (Table 1). Cytoplasmic delivery of the Smac peptide enhances apoptosis by promotion of caspase-3 activation (25). Because the free N-terminus of this peptide is required for bioactivity, the peptide was C-terminally elongated by the respective CPP, carrying a fluorescein moiety at its C-terminus (50). Analysis of the uptake of both conjugates confirmed that the internalization mechanisms observed for the CPPs alone remained valid for the CPP–cargo conjugates (Figure 12A). Over the tested concentration range, internalized Smac–Antp exhibited a concentration-independent vesicular staining. In contrast, Smac–R9 at 5 μm showed an exclusive vesicular staining. At 10 μm, the cell population was heterogeneous, with some cells showing bright cytoplasmic and nuclear fluorescence. At 20 μm, the peptide was internalized through NZ and delivered efficiently into the cytoplasm and nucleus. Surprisingly, for this peptide, we observed an enrichment of fluorescence within the nucleoli, an observation that we cannot explain in the moment, however, which has been reported before (19).
To compare the bioactivity of the Smac peptide delivered through either Antp or R9, caspase-3 activation and cell viability after stimulation of the death receptor Fas were selected as readouts. We had shown previously that in contrast to the tumour necrosis factor (TNF) receptors, Fas was not subject to CPP-dependent downregulation from the plasma membrane (50). Stimulation of Fas with an agonistic antibody in the presence of cycloheximide resulted in a 10-fold induction of caspase-3 activity. Pretreatment of cells with the Smac–R9 conjugate potentiated the induction of caspase-3 activation in a concentration-dependent manner by up to 100% (Figure 12B). In contrast, when the Smac-derived peptide was delivered into the cells through the Antp peptide, the enhancement of Fas-mediated caspase-3 activation was less efficient (only up to 30%). A similar difference in efficiencies exerted by Smac–R9 and Smac–Antp was observed when cell viability after 24 h was analyzed as a parameter for the efficiency of apoptosis induction (Figure 12C). While the treatment of cells with the Smac–R9 peptide reduced the cell viability by up to 50%, Smac–Antp had no detectable effect on viability. This experiment supports the intracellular integrity of the peptides after rapid cytoplasmic delivery, which is a requirement for the expected biological effect.
Concentration dependence of the uptake of Antp, R9 and Tat
This work shows that Antp on one hand and R9 and Tat on the other hand differ with respect to the contributions of individual endocytic processes to the uptake of these peptides. Moreover, the mechanism of entry depends on peptide concentration and the availability of alternative endocytic routes. Above a critical threshold, internalization occurs through a highly efficient nonendocytic pathway. This pathway originates from spatially confined areas of the plasma membrane and leads to a rapid release of peptides into the cytoplasm. The integrity of the plasma membrane is fully maintained. Antennapedia differs from the arginine-rich CPPs nona-arginine and Tat in that much higher concentrations and at least partial blockade of endocytic uptake are required for the induction of this pathway.
In the concentration range up to 40 μm, for the Antp peptide, there is strong evidence that at least three endocytic pathways, that is macropinocytosis, CME and caveolae/lipid-raft-mediated endocytosis contribute to uptake. Neither of these three pathways dominates. Pharmacological interference with any one pathway yields a distinct phenotype with respect to the subcellular distribution of fluorescence and is without effect on the other two pathways. The simultaneous inhibition of two endocytic pathways leads to the addition of the respective inhibitory effects.
For R9 and Tat at concentrations below 10 μm, entry of these peptides is sensitive to MβCD and EIPA, strongly indicative of entry through caveolae/lipid-raft-mediated endocytosis and macropinocytosis. Remarkably, there was no evidence for an involvement of CME at this concentration. At a concentration of 5 μm, inhibition of caveolae/lipid-raft-mediated endocytosis or macropinocytosis induces the rapid NZ-dependent uptake mechanism. The induction of this mechanism at a concentration of peptide at which, in the absence of inhibitors, uptake occurs through endocytosis indicates that a critical concentration of peptide associated with the plasma membrane is required. Because of the redirection of peptides toward this pathway, the effect of EIPA and MβCD is somewhat of a paradox: instead of inhibiting the uptake of peptides into the cells, at this concentration, these inhibitors actually promote the uptake. This pathway mediates a more rapid peptide uptake than the other two pathways, one indication for a nonendocytic nature of this import mechanism. When the rapid uptake mechanism is blocked by CPZ, the phenotypes observed for cells treated with MβCD or EIPA are recovered.
At higher concentrations, uptake occurs primarily by a mechanism leading to rapid entry of the peptides into the cytoplasm and nucleus. The EIPA- and MβCD-sensitive mechanisms still contribute to peptide uptake as shown by the fact that the inhibition of one of these pathways still increases the cellular fluorescence even more. Considering our previous observations on the CPP-induced endocytosis of TNF receptors by these three cationic CPPs (50), one should note that Antp was more effective in inducing the internalization of the receptors than R9 and the Tat peptide. At low peptide concentrations, Antp was in fact the only peptide inducing receptor internalization. The internalization of EGF receptors suggested that this internalization occurred through a clathrin-dependent mechanism. Consistently, at low concentrations, only entry of Antp was CPZ sensitive. The internalization induced by Tat and R9 at higher concentrations suggests that a fraction of these peptides, in addition to the rapid NZ-mediated uptake, was also internalized by clathrin-dependent endocytosis, albeit with a much slower kinetics. For Antp, this rapid cytoplasmic uptake was only observed at significantly higher peptide concentrations for cells treated with MβCD.
Uptake through NZs
The highly efficient uptake of R9 through NZs was heparin sulfate and dynamin dependent; sensitive to ammonium chloride, CPZ and rottlerin and insensitive to chloroquine. The dynamin dependence and sensitivity to CPZ, when considered in isolation, are strongly indicative of clathrin-dependent endocytosis. However, the failure to detect vesicles by electron microscopy; the rapid, stepwise and spatially confined uptake kinetics and finally the concomitant formation of membrane blebs all point toward a nonendocytic mechanism.
The ability to abolish this uptake by pharmacological interventions that inhibit molecular processes inside the cell indicates that association of peptide with the plasma membrane alone is insufficient to initiate this process. Instead, accumulation of peptide at the plasma membrane induces an active process inside the cell. Apparently, this process requires a certain minimum concentration of membrane-associated peptide. At medium peptide concentrations, peptides are steadily removed from the plasma membrane by endocytosis. Different import rates of Antp and the other two peptides by the various endocytic processes may therefore be one factor contributing to the lower propensity of Antp to enter along this pathway – a hypothesis that is supported by the ability to induce this uptake by MβCD in HeLa and Jurkat cells.
Once endocytosis is blocked, peptides accumulate and the threshold is reached. The initial formation of the NZ, bleb formation and rapid release into the cytoplasm are tightly coupled processes. All interventions that blocked this uptake also blocked the formation of a NZ. It will be interesting to see whether an intervention can be identified that only blocks the second step.
Considering the absence of vesicles in the electron microscopy images, the blockade of the uptake by expression of DN dynamin is somewhat surprising. However, the formation of blebs is a process involving rapid changes of membrane morphology. Dynamins couple to a large number of cellular signaling processes (51), and even though it is not clear why the peptides induce the formation of outward protrusions, dynamins could well be involved in such a process.
Interestingly, for any given cell for which uptake occurred by distinctive steps, these steps frequently were similar in size. One may hypothesize that the peptides are enriched on specific domains of the plasma membrane and that these domains have a certain capacity to bind peptide that varies from cell to cell. At this point, neither the nature of these domains nor the details of the uptake mechanism is clear. At least, our observations favor an ‘accumulate and discharge’ mechanism over a mechanism involving a transient formation of pores or continuous transfer of peptides across the lipid bilayer. The FLIP experiments also strongly support a transient association of the peptides with molecules inside and on the surface of the cell.
Heparan sulfates are candidate molecules for mediating this association. Cell surface heparan sulfates have been considered to play a role as multivalent, low-affinity receptors of cationic CPPs. Remarkably, however, at low peptide concentrations, removal of cell surface heparan sulfates was without effect on uptake, at least as detectable by fluorescence microscopy. Consistent with our results, for the Tat peptide at 100 μm, a requirement for cell surface heparan sulfate for peptide was shown (14), while at 1 μm, heparan sulfates were not required (52). It is difficult to understand why these molecules should functionally act as receptors only at high concentrations. One possible mechanism may be through cross-linking by polyvalent binding of peptides. Guanidinium groups have been proposed to engage in bidentate interactions, possibly with heparan sulfates on the cell surface (53,54). The higher fraction of arginine side chains in the R9 and Tat oligopeptides in comparison with Antp may explain the higher propensity of the former to induce this uptake mechanism. Moreover, such a mechanism would be consistent with ammonium chloride acting as an agent that shields off negative charges on the cell surface.
Reassessment of previous results
To this point, a number of publications have addressed the mechanisms contributing to the cellular uptake of CPPs, with sometimes conflicting results. Observations of a rapid endocytosis-independent mechanism have been the latest addition in this field (18,19). Our model provides a framework to accommodate these previous observations. First of all, our data show that the individual CPPs, albeit all cationic in nature, possess remarkable differences in their import pathways. Moreover, it is now clear that peptide concentration is a key experimental condition when analyzing peptide uptake. Potocky et al. investigated the cellular delivery of the Tat peptide at a concentration of 7 μm. Three different phenotypes with respect to the distribution of the peptide in HeLa cells were observed (21). While some cells exclusively exhibited a vesicular staining, some cells showed a combination of vesicular and cytoplasmic fluorescence, and finally in some cells, fluorescence was only present in the cytoplasm and nucleus. We show that this heterogeneity in the distribution of the peptide is likely related to the fact that a concentration higher than 5 μm had been used. When cells were incubated with the Tat peptide at concentrations of 1–5 μm, the peptide was located only in vesicles (20). The partial colocalization of the Tat peptide with transferrin (11,21) as well as with dextran (20) may now readily be explained by the finding that the peptides simultaneously use several endocytic pathways.
To this point, we restricted ourselves to CPPs, two of which correspond to PTDs of the respective full-length proteins. It needs to be resolved to which degree the import mechanisms of the PTDs correspond to the ones of the full-length proteins. Still observations reported for the Antp and Tat peptide conjugated to protein cargos as well as for the full-length Tat protein are also fully consistent with our model. Antennapedia and Tat peptides differed substantially in their capacity to deliver avidin into HeLa cells (29). For a Tat-derived peptide (aa 11) and the Tat protein, conjugated to green fluorescent protein (GFP), evidence was presented that both conjugates are internalized through the same lipid-raft-dependent mechanism (17). The concentration of the peptide–GFP conjugate was less than 1 μm. At this concentration, lipid-raft-mediated endocytosis as a key uptake mechanism is fully consistent with our model. Moreover, for a Tat–Cre conjugate used at a concentration of less than 1 μm, both lipid-raft-mediated endocytosis and macropinocytosis were identified as mechanisms contributing to internalization (15). Finally, both the Tat peptide and the full-length Tat protein induce the internalization of TNF receptors from the plasma membrane (50), indicating the same mechanism of uptake.
The relevance of our findings for the application of CPPs is supported by the fact that the uptake mechanisms identified for CPPs alone retain their validity also for cargo peptides that are introduced into the cells through coupling to CPPs. Moreover, the demonstration that the uptake mechanism determines the cargo bioactivity stresses the importance of understanding in detail the mode of internalization of CPPs for their application as delivery vectors.
In conclusion, our results provide a comprehensive framework to encompass most previous observation on the cellular uptake of cationic CPPs. One given CPP may not only use different endocytic pathways for cellular entry but may also use nonendocytic entry routes. Considering the differences between Antp and the other two peptides, the fraction of arginine side chains may be a relevant structural characteristic, influencing the contribution of individual entry pathways to uptake, possibly by induction of cross-linking of cell surface heparan sulfates. Concerning the rapid cytoplasmic uptake, a distinctive pharmacological profile was defined. The data indicate that the uptake is a specific response of the cell to a local enrichment of cationic molecules at the plasma membrane. It is not clear so far whether the CPPs initiate the enrichment of the cellular molecules to which they associate or whether the NZ represent preformed membrane domains. Finally, it will be interesting to investigate whether a physiological role for this pathway exists.
Materials and Methods
Cells and reagents
The human cervical carcinoma cell line HeLa was obtained from the American Type Culture Collection (Manassas, VA, USA). HeLadynK44A expressing a DN form of dynamin-2 under the control of a tet on/off promoter, cultured in medium containing 1 μg/mL tetracycline, was a kind gift from Bo van Deurs (University of Copenhagen, Copenhagen, Denmark). The expression of mutant dynamin was induced by tetracycline deprivation for at least 24 h. Transferrin Alexa Fluor 633 conjugate, dextran Alexa Fluor 647 conjugate and CTB Alexa Fluor 555 conjugate and Zenon mouse IgG1 labeling kit (specific for the Fc part of immunoglobulin G1 antibodies) were purchased from Mobitech (Göttingen, Germany). Fluorogenic caspase-3 substrate (Ac-DEVD-AMC), wortmannin, Ro-318220, rottlerin and CPZ were from Calbiochem (Bad Soden, Germany); ammonium chloride, EIPA, MβCD, MTT and heparinase I, II and III were obtained from Sigma (Deisenhofen, Germany). The anti-Fas-activating antibody was obtained from Upstate (Hamburg, Germany). BSA was obtained from SIGMA (Steinheim, Germany). The anti-heparan sulfate proteoglycan (HSPG) vesicular stomatitis virus (VSV)-tagged single-chain antibody HS4C3 (55) and the mouse anti-VSV (clone P5D4) antibody were a kind gift of Toin van Kuppevelt (Radboud University Nijmegen Medical Centre, Nijmegen, The Netherlands).
Automated peptide synthesis was performed by solid-phase Fmoc/tBu-chemistry using an automated peptide synthesizer for multiple peptide synthesis (RSP5032; Tecan, Hombrechtlikon, Switzerland) in 2-mL syringes according to a protocol described elsewhere (20). Smac–Antp and Smac–R9 (Table 1) were synthesized using a nα-carboxyfluorescein-labeled lysyl-Rink amide resin (56). The purity of all peptides was determined by analytical high-performance liquid chromatography (HPLC). The identity of the peptides was confirmed by MALDI-TOF mass spectrometry. Peptides with a purity of less than 95% were purified by preparative HPLC.
HeLa cells were incubated with medium containing peptides at the indicated concentrations for 30 min at 37°C. After incubation, the cells were washed with medium, detached by trypsinization for 5 min, washed, suspended in PBS and measured immediately by flow cytometry (BD FACSCalibur System; Becton Dickinson, Heidelberg, Germany). In each case, the fluorescence of 10 000 vital cells was acquired. Vital cells were gated based on sideward and forward scatter.
Confocal laser scanning microscopy
Confocal laser scanning microscopy was performed on an inverted LSM510 laser scanning microscope (Carl Zeiss, Göttingen, Germany) using a Plan-Apochromat 63× 1.4 N.A. lens. All measurements of peptide uptake were performed with living, nonfixed cells grown in eight-well chambered cover glasses (Nunc, Wiesbaden, Germany). Cells were seeded at a density of 4 × 104/well 1 day before the experiment and cultured in RPMI-1640 supplemented with 10% fetal calf serum. For detection of fluorescein-labeled peptides, the 488-nm line of an argon ion laser was directed over an HFT UV/488 beam splitter, and fluorescence was detected with a BP 505–550 band pass filter. For the simultaneous detection of fluorescein-labeled peptides and Alexa Fluor 633 or -647 conjugates, the 488-nm line of an argon ion laser and the light of a 633-nm helium neon laser were directed over an HFT UV/488/633 beam splitter, and fluorescence was detected using an NFT 545 beam splitter in combination with a BP 505–530 band pass filter for fluorescein detection and an LP 650 long pass filter for Alexa Fluor 633 and Alexa Fluor 647 detection. Life cell microscopy was performed at room temperature.
Quantitative analysis of colocalization
First, color channels containing the signal of fluorescein-labeled Antp and the respective tracer were extracted from the multichannel confocal images. After low-pass filtering to reduce image noise, for both channels, binary masks corresponding to vesicle-associated fluorescence were generated (MaskAntp and MaskTracer). Threshold levels were selected by visual inspection. A third mask containing those pixels in which vesicle-associated fluorescence was present in both channels was generated by applying an AND-operation to the masks for peptide- and tracer-associated fluorescence (MaskColoc). Because of the presence of peptide and tracer in the medium, all three masks also included all pixels outside the cells. For this reason, individual cells were selected as ROI and the number of objects (NAntp, NTracer and NColoc) within each cell corresponding to individual vesicles or small groups of vesicles in all three binary masks was counted using the object analysis routines within Image Pro Plus 4.5 (Media Cybernetics Inc., Silver Spring, MD, USA). The fraction of tracer-positive vesicles colocalizing with Antp was calculated by dividing NColoc by NTracer, and the fraction of Antp-positive vesicles colocalizing with tracer was calculated by dividing NColoc by NAntp.
HeLa cells were grown on CELLocate 5245, with a square size of 55 μm (Eppendorf, Hamburg, Germany). After addition of medium containing R9 in a concentration of 20 μm, formation of NZ was followed by confocal laser scanning microscopy as described previously. Directly after appearance of NZ, cells were washed with ice-cold PBS and subsequently fixed with 4% formaldehyde in PBS for 30 min at room temperature. For electron microscopy, cells were stored for 1 h at 4°C, postfixed with 1% osmium tetroxide in 100 mm phosphate buffer at pH 7.2 for 1 h on ice, washed with H2O, treated with 1% aqueous uranyl acetate for 1 h at 4°C, dehydrated through a graded series of ethanol and embedded in Epon. Ultrathin sections were stained with uranyl acetate and lead citrate and viewed in a Philips CM10 electron microscope. The imprint of the cellocate coverslip in the Epon resin was used as an orientation for the matching of fluorescence and electron microscopy images.
Incubation with inhibitors
Cells were treated with the indicated inhibitors for 30 min at 37°C. Then, the medium was removed and medium containing peptide as well as the corresponding inhibitor was added. After 30 min of incubation at 37°C, the cells were washed twice with medium and analyzed by fluorescence microscopy or flow cytometry.
HeLa cells were seeded at a density of 4 × 104/well in eight-well chambered covered glasses and incubated with medium containing 10 U/mL of heparinase I, 5 U/mL heparinase II and 2 U/mL heparinase III for 6 h at 37°C. Then, the cells were washed with ice-cold PBS containing 0.1% (w/v) BSA and incubated with the anti-HSPG antibody HS4C3 on ice. After 1.5 h of incubation, cells were washed with ice-cold PBS/BSA. To visualize bound antibodies, the cells were incubated for 1 h with an anti-VSV antibody (P5D4)/Zenon Alexa Fluor 647 conjugate on ice. The antibody staining was analyzed by confocal laser scanning microscopy.
Caspase-3 activity assay
HeLa cells were incubated with medium containing peptides for 30 min at 37°C. After one washing step, cells were treated as indicated with agonistic Fas antibody and cycloheximide (CHX) for the induction of apoptosis, followed by incubation for further 3 h. Cells were harvested by scraping, washed with ice-cold PBS and lysed in lysis buffer [1% Triton, 150 mm NaCl, pH 7.7, supplemented with protease inhibitor cocktail tablets (Roche Diagnostics, Mannheim, Germany)] for 30 min on ice. The protein content in lysates was determined using a commercially available Bradford protein assay kit (Bio Rad Laboratories, München, Germany). Equivalents of 30 μg protein for each sample were diluted in caspase activity buffer (20 mm HEPES, 10 mm dithiothreitrol, 10% glycerol, 100 mm NaCl, pH 7.5). Caspase-3 substrate was added to the samples to a final concentration of 2 μm. The efficiency of the substrate cleavage by active caspase-3 was analyzed immediately after substrate addition and after 1 h of incubation at 37°C using a luminescence spectrometer LS50B (PerkinElmer, Norwalk, CT, USA).
HeLa cells were seeded in 96-well microtiter plates (1.5 × 104/well) and cultivated over night. The next day, cells were incubated with peptides for 6 h at 37°C. Cell viability was measured using the colorimetric MTT dye. Cells were incubated with MTT at a concentration of 1 mg/mL for 4 h. The formazan product was solubilized with SDS [10% (w/v) in 10 mm HCl]. Cell viability was determined by measuring the absorbance of each sample at 570 nm using a microplate reader (Molecular Devices SpectraMax 340; GMI, Ramsey, Minnesota, USA).
HeLa cells were treated with the indicated peptides for 30 min at 37°C, washed and stimulated with the agonistic Fas antibody (100 ng/mL) and CHX (2 μg/mL) for the induction of apoptosis. After an additional 24-h incubation at 37°C, cells in both groups were washed with PBS, followed by crystal violet staining [20% (v/v) methanol, 0.5% (w/v) crystal violet] for 15 min. The wells were washed with H2O and air-dried. The dye was dissolved in methanol and the optical density at 550 nm measured with an enzyme-linked immunosorbent assay plate reader.
We thank Viktoria Wolf for the isolation and cultivation of human dendritic cells, Brigitte Sailer for oriented cutting of ultrathin sections and Anne Spang and Mark Trautwein for helpful discussions. F. D. is a scholar of the Graduiertenkolleg 794. R. B. gratefully acknowledges financial support from the Volkswagen Foundation (“Nachwuchsgruppen an Universitäten” I/77 472).