Cell internalization of the third helix of the Antennapedia homeodomain is receptor-independent.

We have recently reported that a 16-amino acid long polypeptide corresponding to the third helix of the DNA binding domain (homeodomain) of Antennapedia, a Drosophila transcription factor, is internalized by cells in culture (Derossi, D., Joliot, A. H., Chassaing, G., and Prochiantz, A. (1994) J. Biol. Chem. 269, 10444-10450). The capture of the homeodomain and of its third helix at temperatures below 10°C raised the problem of the mechanism of internalization. The present demonstration, that a reverse helix and a helix composed of D-enantiomers still translocate across biological membranes at 4 and 37°C strongly suggests that the third helix of the homeodomain is internalized by a receptor-independent mechanism. The finding that introducing 1 or 3 prolines in the structure does not hamper internalization also demonstrates that the α-helical structure is not necessary. The data presented are compatible with a translocation process based on the establishment of direct interactions with the membrane phospholipids. The third helix of the homeodomain has been used successfully to address biologically active substances to the cytoplasm and nucleus of cells in culture (Théodore, L., Derossi, D., Chassaing, G., Llirbat, B., Kubes, M., Jordan, P., Chneiweiss, H., Godement, P., and Prochiantz, A. (1995) J. Neurosci. 15, 7158-7167). Therefore, in addition to their physiological implications (Prochiantz, A., and Théodore, L. (1995) BioEssays 17, 39-45), the present results open the way to the molecular design of cellular vectors.

Homeoproteins are transcription factors involved in several important biological processes occurring primarily, but not exclusively, during development. The DNA binding domain of these transcription factors is highly conserved and is called the homeodomain. It consists of 60 amino acids arranged in 3 ␣-helices. Helix 3 is separated from helix 2 by a ␤ turn and is called the recognition helix, because it is involved in the interaction of the homeodomain with specific sites in the large groove of double-stranded DNA (Gehring et al., 1994).
In the course of our studies on the role of homeoproteins in neuronal development we observed that the homeodomain of Antennapedia, a Drosophila homeoprotein, is internalized by cells in culture and, following internalization, is conveyed to cell nuclei where it can directly and specifically interfere with transcription Joliot et al., 1991aJoliot et al., , 1991bLe Roux et al., 1993. Internalization occurs at both 4 and 37°C and thus does not seem to involve classical receptor-mediated endocytosis.
The region of the homeodomain (AntpHD) 1 of the Drosophila transcription factor responsible for its internalization by cells in culture, in particular nerve cells, has been mapped to its third helix . This finding has led to the demonstration that a 16-amino acid long peptide corresponding to the third helix of the homeodomain (minus an N-terminal glutamate residue) translocates at 37 and 4°C across biological membranes, reaches the cytoplasm, and is eventually conveyed to the nucleus of cells in culture (Derossi et al., 1994).
The successful use of this peptide as a vector to address biologically active compounds inside live cells Troy et al., 1996) was a strong incentive to start experiments aimed at understanding its mechanism of translocation. Our previous studies (Derossi et al., 1994) had led to the conclusion that the peptide is poorly structured in water but adopts an amphipathic ␣-helical structure in a hydrophobic environment. We also observed a tendency of the peptide to form dimers and even multimers in the presence of SDS, a phenomenon tentatively attributed to the formation of and association with detergent micelles. In the present study we report new data demonstrating that internalization does not require specific interactions with a chiral receptor (binding site or transporter) or the formation of a charged pore by an ␣-helical conformation of the peptide.

MATERIALS AND METHODS
Peptides Synthesis-Peptide synthesis was carried out on a 0.1 mmol scale (ABI Model 431A synthesizer) starting from an ␣-p-methylbenzhydrylamine resin (typical substitution, 0.77). All N-␣-t-butoxycarbonyl amino acids, in 10-fold excess, were assembled using dicyclohexylcarbodiimide and 1-hydroxybenzotiazol as coupling agents. After removal of the last N-t-butoxycarbonyl protecting group of 5-aminopentanoic acid, the N-hydroxysuccinimido ester of biotin was coupled manually in dimethylformamide. Peptides were cleaved from the resin by anhydrous fluorhydic acid and purified by preparative reverse phase HPLC (Applied Biosystems), using a 10 ϫ 250-nm Bronwlee column (Aquapore Octyl, 300 Å pore size) using various acetonitrile gradients in aqueous 0.1% trifluoroacetic acid. The purity of collected fractions was established by analytical isocratic separation (HPLC, Waters Associate) on lichrosphere 10 RP-8 columns (Merck) and in 0.25 M triethylammonium phosphate, pH 3.0 (buffer A), and acetonitrile. Peptide molecular weights were determined by electrospray ionization mass spectrometry in the laboratory of Pr. J. C. Tabet (Université P. et M. Curie, Paris). All peptides are presented in Fig. 1 and as follows: AntpHD(43-58) abbreviated 43-58; purification (gradient: 12-22, 8% CH 3 CN, linear, 30 min); purity Ͼ95%; HPLC (iso 19.2% CH 3 CN in buffer A), 17.3 min. Ant-* This study was supported by CNRS, Ecole Normale Supérieure and grants from Agence Nationale de la Recherche sur le SIDA and Fondation pour la Recherche Médicale (Sidaction). The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
Cell Cultures-The cell culture conditions have been described in earlier reports (Lafont et al., 1992). Briefly, small fragments from the cortical-striatal region of the embryonic rat brain between E13 and E15 were incubated in trypsin (0.25%; Life Technologies, Inc.) for 5 min at 37°C, washed twice in phosphate buffer containing 33 mM glucose (PBS), and incubated for 10 min at 37°C with 15 g/ml DNase I (Sigma) in Dulbecco's modified Eagle's medium/F12 supplemented with 33 mM glucose, 2 mM glutamine, 10 mM Hepes, pH 7.4, 3 mM NaHCO 3 , penicillin 5 units/ml, and streptomycin 5 g/ml (SFM). The cells were dissociated mechanically, washed three times in SFM, and plated at a concentration of 25,000 cells/cm 2 on glass coverslips (16-mm diameter) coated with 15 g/ml DL-polyornithine (Sigma). All cultures were in chemically defined medium consisting of SFM supplemented with 0.1% ovalbumin, 100 g/ml transferrin, 20 nM progesterone, 20 M putrescine, and 30 nM selenium.
Peptide Internalization and Visualization-When added to cells in culture, peptides were diluted in the chemically defined medium supplemented with 10% fetal calf serum and the following mixture of protease inhibitors: 0.5 mM Pefablock, 1 g/ml ␣ 2 -macroglobulin, and 10 g/ml leupeptin. 2-4 h after the addition of a volume of peptide equal to the volume present in the dishes, the cells were washed three times with the chemically defined medium, fixed for 5 min in 4% paraformaldehyde at room temperature, and then for 5 min at Ϫ20°C in ethanol/ CH 3 COOH (95/5), washed three times in PBS, and incubated for 30 min with fluorescein-linked streptavidin (Amersham Corp.) diluted 1000fold into PBS plus 10% newborn calf serum (Life Technologies, Inc.). At the end of the incubation the cells were washed three times in PBS, once in water, dried, and mounted in Mowiol for examination.
In some experiments freshly dissociated cells were resuspended in 0.6 ml of PBS or chemically defined medium containing appropriate peptide concentrations plus protease inhibitors. After 2-4 h of incubation with regular gentle shaking, the cells were centrifuged and washed once with 1 ml of PBS plus protease inhibitors and a second time with PBS adjusted to 0.5 M NaCl. The final cellular pellet was directly resuspended in Laemmli buffer or sonicated to allow for a brief purification of the peptides on streptavidin-agarose (Life Technologies, Inc.). The cells or the beads were boiled and frozen for storage at Ϫ80°C or immediately loaded onto a 12 to 22% polyacrylamide SDS gradient gel. Following migration, peptides were electrotransferred onto Immobilon (Amersham Corp.) in 25 mM Tris, 192 mM glycine, pH 8.8, the filter was fixed in 4% glutaraldehyde for 15 min, and nonspecific binding sites were blocked by a 2-h incubation in 20 mM Tris-HCl pH 7.5, 150 mM NaCl, 0.2% Tween 20, and 4% bovine serum albumin. The blots were incubated 1 h at room temperature with streptavidin-peroxidase (Amersham Corp.) diluted 1 to 1000 in the blocking solution with 0.5% bovine serum albumin, washed several times with the same solution, and revealed with luminol, in accordance with the instructions of the manufacturer (Amersham Corp.).
Confocal Microscopy-Data were obtained with a confocal laser scanning microscope Sarastro 2000 (Molecular Dynamics). Excitation was with an argon ion laser set at 488 nm for fluorescein isothiocyanate excitation, and the emitted light was filtered with an appropriate long pass filter (510 nm). Sections presented were taken approximately at the mid-height level of the cells. Photomultiplier gain and laser power were identical within each experiment.
ELISA Quantification-The cells were incubated with the different peptides as described previously. After several washes with PBS, approximately 5 ϫ 10 4 cells per condition were plated on ELISA wells. Cells were left to attach for 2 h, fixed overnight in 5% paraformaldehyde at 4°C, permeabilized in methanol, and rinsed three times in PBS. Endogenous alkaline phosphatase was neutralized by incubating the ELISA plate at 65°C for 1 h. After blocking in 100 mM Tris-HCl, pH 7.4, NaCl 0.3 M, 0.2% Tween, and 5% milk, cells were incubated for 30 min, at room temperature, with alkaline phosphatase-streptavidin (Vectastain ABC kit). Alkaline phosphatase activity was measured spectrophotometrically at 405 nm using a freshly made solution of 1 mg/ml nitrophenyl phosphate (Sigma) in 10 mM diethanolamine (pH 9.5), 0.5 mM MgCl 2 .
Electron Microscopy-The cells (2 ϫ 10 6 /ml) were incubated for 2 h at 37°C with the peptides. After several washes with culture medium, the cellular pellets were included in a collagen gel (50% collagen in culture medium), fixed for 30 min at 4°C in PBS containing 4% paraformaldehyde and 0.1% glutaraldehyde, washed in PBS, incubated for 30 min at 4°C in PBS containing 0.3% NH 4 Cl, and rinsed overnight at 4°C with PBS. Progressive lowering of temperature was performed in an AFS Reichert (Leica) apparatus. This procedure involves stepwise reductions in temperature (20 to Ϫ35°C) as the concentration of the dehydrating agents is increased (30 to 95% ethanol). The infiltration in resin (Lowicryl K4M, Boiziau distribution) was performed at Ϫ35°C by progressively increasing the resin concentration and decreasing the percentage of absolute ethanol. Polymerization was carried out at Ϫ35°C for 48 h using reflected light from an ultraviolet source. The temperature was slowly increased to reach 20°C. Ultrathin sections from the different blocks were prepared and mounted on 400 mesh Formvarcoated nickel grids. Immunodetection was performed as follows: pretreatment in PBS containing 5% bovine serum albumin for 30 min, washes in 0.1% bovine serum albumin-PBS, incubation for 1 h in PBS containing 0.01% gelatin, 1/200 colloidal gold streptavidin conjugate (15 nm, British Bio Cell International), washes in PBS and H 2 O, fixation for 5 min in 1% glutaraldehyde, and washes in H 2 O. After air drying, sections were contrasted using 4% uranyl acetate in water and Millonig's lead acetate-tartrate. The samples were analyzed with a Philips 400 electron microscope.

Peptide Internalization Is Not
Receptor-mediated-To investigate whether 43-58 internalization requires a chiral receptor, we synthesized peptides 43-58D and 58-43. Peptide 43-58D has a primary sequence identical to that of 43-58 but is entirely composed of D-enantiomers. Peptide 58-43 is composed of the same amino acids as peptide 43-58, but the order of the amino acids has been reversed, thus leading to a totally different primary structure. Fig. 2, in which the peptides are represented in a ␣-helical conformation, illustrates that the modifications of the original peptide modify the orientation of the helix but not its amphipathic profile.
Peptide internalization was tested on nerve cells after 2 days of development in vitro. Peptides at a concentration of 22 M were added for 2 h at 4°C or 37°C. Peptide 41-55, which is not internalized (Derossi et al., 1994), was used as a control. After several washes in PBS, including a final wash in PBS adjusted to 0.5 M NaCl, the cells were fixed and the presence of biotin revealed with fluorescein isothiocyanate-labeled streptavidin. Fig. 3 (top panel) illustrates that the 3 peptides, in contrast to FIG. 3. Peptide visualization with confocal microscopy. Cortical-striatal cells dissociated from rat E15 cells were cultured for 2 days at a density of 25,000 cells/cm 2 before addition of the different peptides at a concentration of 22 M. After 2 h of incubation at 37 or 4°C, the cells were rinsed and fixed, and the presence of biotin was revealed with fluorescent streptavidin. Sections presented here demonstrate that all peptides except peptide 41-55 are internalized at both temperatures. Note that at 4°C proline-containing peptides (Pro50 and 3Pro) are not found in the nuclei. Laser power is identical within each experiment, photomultiplier gain is reduced 4-fold for peptide 58-43 at both temperatures.

41-55, are internalized by the cells at both temperatures.
To establish that the internalized peptides were not degraded, the peptides retrieved by incubating the cellular extracts on streptavidin-agarose were separated by SDS-PAGE and blotted, and their presence was revealed with streptavidinperoxidase (see "Materials and Methods" for details). Fig. 4 demonstrates that 58-43 and 43-58D are internalized at 37°C (Fig. 4A) and 4°C (Fig. 4B). All peptides run as monomers (asterisks), dimers, and multimers on an SDS gel. For size comparison, the 11-amino acid long substance P peptide (molecular mass, 1670 Da), aprotinin (molecular mass, 6500 Da), and lysozyme (molecular mass, 14400 Da) were run in parallel. Note that peptide 41-55 is not internalized and that peptide 43-58 is highly sensitive to degradation at 37°C but not at 4°C.
This first series of experiments strongly suggested that peptide internalization is not receptor-dependent. To better establish this point, we compared the localization of 43-58 and 43-58D at the ultrastructural level. Fig. 5 illustrates that the two peptides can be localized within the cells, both in the cytoplasm and in the nucleus. The presence of gold particles in endocytic figures was rarely observed.
Is the ␣-Helical Structure Important?-To test the importance of the helical structure for internalization, two peptides were synthesized into which were introduced one or three prolines. In Pro50, the glutamine in position 50 was replaced by a proline. In 3Pro, in addition to Gln-50, Ile-45, and Lys-55 were also replaced by proline residues. As illustrated in Fig. 3, introducing one or three prolines into the sequence did not hamper internalization at 4 or 37°C, suggesting that the positions modified are not important and that a helical structure is not mandatory for translocation.
That the peptides were intact after internalization was ver-ified by gel electrophoresis and Western blotting of the peptides retrieved from the cells and purified on streptavidin-agarose. As demonstrated in Fig. 4, the two peptides migrate on the gels with electrophoretic mobilities which indicate that they are intact and suggest that they can form dimeric and multimeric structures.
Although the translocation at 4 and 37°C allows for the addressing of the two peptides to the cytoplasmic compart- FIG. 4. Gel electrophoresis of cell extracts. E15 cortical-striatal cells at (1.2 ϫ 10 6 cells in 0.6 ml) were incubated with the different peptides for 2 h at 37°C (A) or 4°C (B). After several washes, the intracellular presence of the peptides was analyzed by gel electrophoresis and electrotransfer. Note that all peptides have a tendency to form multimers. Multimerization is in part due to the formation of micelles in SDS (Derossi et al., 1994). The degree of multimerization is not similar with all peptides probably because of differences between peptide-peptide interactions. For example, it was observed by NMR (G. Chassaing, unpublished results) that a prolin in 50 provokes the construction of the N-terminal domain of the peptide into a ␤-sheet, thus favoring peptide dimerization. SP, substance P. Because of their spontaneous degradation at 37°C (as seen on the gel) twice (2ϫ) or four times (4ϫ) the amount of 43-58 or of 41-55 were added to the cells (in A).

FIG. 5. Peptides visualization by electron microscopy.
Corticalstriatal cells (2 ϫ 10 6 cells/ml) dissociated from E15 rat embryos were incubated with 44 M of peptides 43-58 or D43-58 for 2 h at 37°C. After incubation, cells were washed, inoculated in collagen gel, and analyzed by electron microscopy (see "Materials and Methods" for details). Note that both peptides accumulate primarily in the cytoplasm (C) and nucleus (N) and, within the nucleus, in the heterochromatin (h). Arrowheads underline the nuclear membrane. ment, we noted that peptides Pro50 and 3Pro were not conveyed to the nuclei to the same extent as were the other peptides. This latter point is illustrated in Fig. 3. It is shown that, compared to 58-43, which is present throughout the cytoplasm and nuclei of the cells, 3Pro is concentrated in the cytoplasm. Pro50 behaved similarly to 3Pro, whereas the distributions of 43-58 or 43-58D are identical to that of 58-43.
Quantification-The patterns of staining observed after internalization as well as the profiles of the Western blots suggested that all peptides were not internalized with the same efficiency. However, the small size of the peptides and the number of steps necessary to obtain a Western blot precluded the use of the latter approach to compare rigorously the internalization of the different peptides. To estimate better the amount of peptide internalized, an ELISA test was developed (see "Material and Methods" for details). In a preliminary experiment we used the 43-58D, which, because of the D-enantiomers is unlikely to be rapidly degraded, to define an optimal concentration to be used for this analysis. As shown in Fig. 6A, 43-58D internalization (2 h at 37°C) increases linearly with concentration and does not saturate between 10 and 80 M.
The comparison between the different peptides used in this study was thus done at a concentration of 40 M and both at 4 and 37°C. Similar results were obtained at the two temperatures except for 43-58 which, as already mentioned, is subject to proteolytic degradation at the highest temperature. Fig. 6B illustrates the results of such an experiment, done at 4°C. It confirms that 41-55 is not internalized. In addition, it suggests that all peptides are internalized with a similar efficiency, except for Pro50, for which the amount recovered inside the cells is significantly and reproducibly higher.

DISCUSSION
The strongest evidence against receptor-mediated endocytosis is the internalization of 43-58D and 58-43. The two peptides are very unlikely to interact specifically with a receptor that would recognize a precise sequence of amino acids (Wade et al., 1990). We cannot, however, entirely exclude the possibility that a receptor or a family of receptors would recognize a general structure conserved between the three polypeptides, in particular a defined organization of hydrophobic and charged amino acids. Even if such interactions exist, the formation of endocytotic coated vesicles is not compatible with internalization at 4°C. In addition, endocytotic figures were almost never observed at the ultrastructural level. NMR studies demonstrated (not shown) that in Pro50 the 43-50 domain is in a ␤-sheet conformation, the ␣-helix being maintained from residue 51 to residue 58. Although we do not have the corresponding NMR data, it is highly unlikely that the presence of three prolines would be compatible with an ␣-hel- ical structure. We thus propose that the ␣-helical structure observed in a hydrophobic environment (Derossi et al., 1994) is not necessary for the internalization of the peptides.
Peptides with prolines reach the cytoplasm but are poorly conveyed to the nucleus. This suggests that nuclear accumulation is not only due to the small size of the peptide but also to its sequence and/or structure. The notion that the highly basic third helix is part of a nuclear localization signal for homeoproteins is supported by homologies with the Mat-␣2 transcription factor the nuclear localization signal of which is in the homeodomain (Hall et al., 1990) and by results showing that deleting its homeodomain blocks the nuclear import of  In addition to its possible physiological significance regarding the structure of the nuclear localization signal, the primarily cytoplasmic accumulation of Pro50 and 3Pro might be useful for the development of vectors that would preferentially address drugs in the cytoplasm.
That translocation is not due to an ␣-helical amphipathic conformation leading to the formation of a positively charged channel is further supported by the following facts. First, a 16-amino acid ␣-helix is not long enough to span the plasma membrane; second, we could never measure the passage of a current that would suggest the formation of a channel 3 ; third, replacing the two tryptophan residues by two phenylalanines does not modify the helical structure or the amphipathicity, but virtually abolishes translocation (Derossi et al., 1994). Finally, translocation of long polypeptides or oligonucleotides attached to the helix through a positively charged pore (Allinquant et al., 1995;Chatelin et al., 1995;Perez et al., 1994;Troy et al., 1996aTroy et al., 1996bFåhraeus et al., 1996) is very unlikely.
The two main possibilities that remain to explain peptide translocation are the formation of inverted micelles and fluid phase pinocytosis (de Kruijff et al., 1985;Illinger et al., 1991;Sandvig and van Deurs, 1994). Indeed, both phenomena are receptor-independent and can occur at 4°C. The reasons for favoring a model implying inverted micelles (Fig. 7) are as follows. First, endocytosis figures are rarely seen at the ultrastructural level; second, as opposed to results obtained with the glycophosphatidylinositol linked folate receptor (Leamon and Low, 1991), the size of the polypeptides that can be internalized after attachment to the third helix is limited and, for a similar molecular weight, depends on the structure of the peptide attached to the homeodomain or to its third helix (Chatelin et al., 1996;Perez et al., 1994); third, tryptophans, which are necessary for translocation (Derossi et al., 1994), are known inducers of inverted micelles (de Kruijff et al., 1985); fourth, the formation of peptide multimers in the presence of SDS could reflect micelle formation (Derossi et al., 1994). Finally, phosphorus NMR studies in which we studied the interactions of 43-58 (which is internalized) and of 43-58 in which tryptophans are replaced by phenylalanines (poorly internalized) (Derossi et al., 1994) with brain phospholipids demonstrated that only the former peptide provokes the formation of hexagonal phases or inverted micelles. 4 The interest of the present study is 2-fold. First, it is potentially important to the development of new vector systems for the intracellular addressing of pharmacological substances. The discovery that the homeodomain of Antennapedia, and within this structure its third helix (peptide 43-58), could be used to transport polypeptides and oligonucleotides into cells has now permitted the identification of structural features responsible for this unexpected property. Although we do not yet understand the details of the mechanism of translocation, the simple fact of being able to fabricate a D-helix unlikely to be rapidly degraded, or variants (Pro50 and 3Pro) that could address exogenous molecules exclusively to the cytoplasm is of obvious interest for the study of intracellular functions.
A second point of interest is the physiological significance of homeodomain and homeoprotein internalization. Although the internalization and even the nuclear addressing of exogenous proteins has been noted before Rubartelli et al., 1990), it is seldom that a systematic study of the structure and mechanisms involved in these processes has been undertaken. It is striking that, in the case of Antennapedia, the structure which is necessary for translocation corresponds to the third helix of the homeodomain. Because this helix is highly conserved among several homeodomains, it is possible that membrane translocation is a property shared by several homeodomains. This latter point is supported by experiments showing that the homeodomains of Engrailed, fushitarazu, and Hoxa-5 are internalized (Chatelin et al., 1996). More interesting is the finding that full-length Hoxa-5 (Chatelin et al., 1996) and Hoxc-8 3 are also internalized and targeted to the nucleus of cells in culture. Our proposal is that the third helix of a homeodomain could induce the formation of an inverted micelle and thus of an hydrophilic cavity capable to accommodate an entire homeoprotein. It has to be placed in the perspective of the hypothesis that cells may exchange positional information through the local trading of homeoproteins (Chatelin et al., 1995;Prochiantz and Théodore, 1995).