E. Vives, Institut de Génétique Moléculaire de Montpellier, CNRS UMR 5124, BP5051, 1919 route de Mende, 34033 Montpellier cedex 1, France. Fax: + 33 467 040231, Tel.: + 33 467 613661, E-mail: firstname.lastname@example.org
Translocation through the plasma membrane is a major limiting step for the cellular delivery of macromolecules. A␣promising strategy to overcome this problem consists in the chemical conjugation (or fusion) to cell penetrating peptides (CPP) derived from proteins able to cross the plasma membrane. A large number of different cargo molecules such as oligonucleotides, peptides, peptide nucleic acids, proteins or even nanoparticles have been internalized in cells by this strategy. One of these translocating peptides was derived from the HIV-1 Tat protein. The mechanisms by which CPP enter cells remain unknown. Recently, convincing biochemical and genetic findings has established that the full-length Tat protein was internalized in cells via the ubiquitous heparan sulfate (HS) proteoglycans. We demonstrate here that the short Tat CPP is taken up by a route that does not involve the HS proteoglycans.
Several cell-penetrating peptides (CPP) allowing the efficient internalization of various nonpermeant drugs in different cell lines have been recently described. A covalent link had to be created between the CPP and the cargo molecule to promote efficient membrane translocation of the chimera [1–7]. A 16-mer peptide derived from the Antennapedia protein homeodomain  and a 13-mer peptide derived from the HIV-1 Tat protein  have been extensively studied. In our initial experiments using the short Tat basic domain, we demonstrated the uptake of chemically conjugated nonpermeant peptides . Then, several peptides showing a cellular activity were successfully vectorized either with the Antennapedia peptide  or the Tat peptide [12–14]. Along the same lines, antisense oligonucleotides (ON) were coupled chemically to the Antennapedia peptide , or to the short Tat peptide [2,15]. Efficient internalization and biological activity of the ONs were observed. Peptide nucleic acids (PNAs) were also taken up by cells after coupling to Transportan or to the Antennapedia peptide , or to the Tat peptide (E. Vivès & B. Lebleu, unpublished observations). Regulation of the galanin receptor expression by a sequence specific antisense activity was observed after incubation of cells with the chimera . The cellular internalization of proteins such as β-galactosidase, horseradish peroxidase or Fab antibody fragment was also reported. In these cases, the carrier Tat peptide and the transported protein were associated either by chemical coupling [4,5,16] or by genetic construction leading to a fusion protein expressing the 13-amino-acid CPP moiety at its N-terminus [6,7].
We have focused on the short HIV-1 Tat derived peptide. Indeed it was initially shown that the maximum rate of internalization was reached when three to four molecules of a 35-amino-acid Tat peptide were chemically coupled to the transported protein . In this case, the use of shorter peptides appeared to reduce the uptake process. A structure–function relationship study of the peptide encompassing this 35-amino-acid region then allowed delineation of the translocating activity domain to a 13-mer amino-acid sequence . This sequence contains six arginine residues and two lysine residues within a linear sequence of 13 amino acids, conferring a highly cationic character on this peptide . It was later shown that arginine residues were essential for translocation as deletion (or replacement by alanine) of a single arginine severely reduced internalization [10,17].
The mechanism by which these cell penetrating peptides (and their conjugates) enter cells is not yet determined, although endocytosis does not seem to be required [9,18]. First, it was shown for the Antennapedia peptide that structural requirements were not involved in the uptake process as the inverso d-isomer form of the peptide  or insertion of proline residues within the primary sequence  did not impair cell uptake. Tat behaviour is very similar to Antennapedia as the Tat peptide with all d-amino acids (48GRKKRRQRRRPPQ60C) still enters cells  and the retro-inverso form of the Tat peptide (57RRRQRRKKR49 with all d-amino acids) is even more efficiently translocated than the corresponding native peptide . Second, both peptides are internalized at 4 °C [9,18], a temperature which abolishes active transport mechanisms involving endocytosis. Third, both peptides were found to be taken up in various tissue types suggesting an ubiquitous process of internalization which strongly suggests binding to conserved cell membrane determinants. Recently, convincing biochemical and genetic evidence suggested that the cell surface heparan sulfate (HS) proteoglycans, which are expressed in most cell types, are responsible for the internalization of the full-length Tat protein fused to glutathione S-transferase (GST) and/or green fluorescent protein (GFP) . Moreover, mutations in the basic domain of Tat abolished uptake of these constructions  thus indicating that this domain is essential for binding to the receptor. The present work aimed at defining whether membrane translocation of the full-length Tat protein and cellular uptake of its basic domain make use of the same mechanism. Both genetic and biological evidence indicates that the cellular uptake of the Tat basic peptide does not involve binding to HS proteoglycans and endocytosis.
Peptide synthesis and labeling
Peptide synthesis was performed by solid phase on a Pioneer synthesizer (Applied Biosystems, Forster City, CA, USA) following the Fmoc chemistry protocol. The Tat peptide sequence was Gly-Arg-Lys-Lys-Arg-Arg-Gln-Arg-Arg-Arg-Pro-Pro-Gln-Cys as previously described . The Cys residue was added to the C-terminal end of the 13-amino-acid peptide corresponding to the primary sequence of the Tat protein to provide a sulfhydryl group for further ligation to a fluorochrome or to a cargo molecule. The peptide was purified by semipreparative HPLC and characterized by analytical HPLC, amino-acid analysis and MALDI-TOF analysis. Results were in full agreement with the expected criteria (data not shown). Labeling with the fluorochrome was performed on the purified Tat peptide through its cysteine side chain by conjugation with a 10-fold molar excess of fluorescein or rhodamine-maleimide derivatives (Molecular Probes Europe BV, Leiden, the Netherlands) in 50 mm Tris/HCl buffer pH 7.2 for 4 h in the dark. Labeled peptides were purified by semipreparative HPLC, freeze-dried, and resuspended in NaCl/Pi at 1 mg·mL−1. Peptide concentration was assessed by quantitative amino-acid analysis. Peptides were stored frozen at −20 °C until further use.
Cells and cell cultures
HeLa cells were cultured as exponentially growing subconfluent monolayers on 90-mm plates in RPMI 1640 medium (Gibco) supplemented with 10% (v/v) fetal bovine serum and 2 mm glutamine. Wild-type CHO K1 cells and CHO mutants deficient in proteoglycan biosynthesis  were obtained from ATCC (Manassas, VA). The A-745 and D-677 mutant cells were fully defective in proteoglycans. The B-618 mutant produces about 15% of the normal level of the proteoglycans synthesized in wild-type. The E-606 mutant produces an undersulfated form of HS proteoglycan. Finally, the C-605 mutant has also a defect in sulfate uptake leading to low expression of wild-type HS proteoglycans. CHO cell lines were grown in Dulbecco's modified Eagle's medium (Gibco) supplemented with 10% (v/v) fetal bovine serum.
Tat peptide internalization
Exponentially growing cells were dissociated with a nonenzymatic cell dissociation medium (Sigma). Cells (15 × 103 per well) were plated on eight-well LabTek coverslips (Nunc Inc.) and cultured overnight. The culture medium was discarded and the cells were washed with NaCl/Pi (pH 7.3). Cells were preincubated in 100 µL of Opti-MEM (Gibco) at 37 °C for 30 min before incubation with the peptide. Opti-MEM was discarded from the coverslips and the cell monolayers were incubated at 37 °C with Tat peptide dissolved in Opti-MEM at the appropriate concentration. Subsequently, cells were rinsed three times for 5 min with NaCl/Pi (pH 7.3) and fixed in 3.7% (v/v) formaldehyde in NaCl/Pi for 5 min at room temperature. For experiments at 4 °C, the protocol was the same except that all incubations were performed at 4 °C until the end of the fixation procedure. For direct detection of fluorescein-labeled or rhodamine-labeled peptides, cells were washed three times after the fixation, then incubated with 50 ng·mL−1 of Hoechst 33258 in NaCl/Pi at room temperature, and washed again with NaCl/Pi before being processed in Vectashield™ mounting solution (Vector Laboratories Inc., Burlingame, CA, USA).
Internalization and detection of recombinant proteins
Recombinant GST–Tat protein and GST–Tat–GFP were prepared as already described . For direct detection of the GFP recombinant protein by fluorescence microscopy the protocol was identical to Tat peptide internalization. Incubation was performed at a protein concentration of 1 µg·mL−1 in the presence of 100 µm chloroquine in the cell culture medium. For FACS analysis, the concentration of the recombinant protein was increased to 5 µg·mL−1.
The internalization of the GST–Tat construct was monitored by immunodetection as described previously . After incubation with the recombinant construct for 4 h, cells were incubated with a monoclonal murine antibody directed against the Tat 49–58 epitope (Hybridolab, Institut Pasteur, Paris) at a final concentration of 10 ng·µL−1 for 1 h at room temperature. Cells were then washed five times for 5 min with warm NaCl/Pi (25–28 °C) before incubation with a rhodamine-conjugated anti-(mouse IgG) Ig (Sigma) for 30 min. The distribution of the fluorescence was analysed by microscopy on a Zeiss Axiophot fluorescence microscope .
To analyze the internalization of fluorochrome-labeled Tat peptides or GFP-Tat by cell cytometry, 5 × 105 cells per well were plated and cultured overnight. The culture medium was discarded, the cells were washed with NaCl/Pi (pH 7.3) and preincubated in 1 mL Opti-MEM at 37 °C for 30 min before incubation with the fluorescent constructs. Cells were washed three times with NaCl/Pi, dissociated with nonenzymatic cell dissociation medium, centrifuged at 250 g and resuspended in 500 µL NaCl/Pi. Fluorescence analysis was performed with a FACScan fluorescence-activated cell sorter (Becton Dickinson). A total of 10 000 events per sample were analyzed.
Cell treatment with heparinase III
Cell treatment with the heparinase III GAG lyase (Sigma) was performed as previously described . However for easier handling of the cells, treatment was performed on HeLa cells instead of CHO K1 cells. Cells were then incubated with of 5 µg·mL−1 Tat–GFP fusion protein or with 1 µm fluorescein Tat peptide and analyzed by FACS.
Uptake and cellular localization of fluorescently labeled Tat peptides
Cellular uptake of the full-length Tat protein fused to GFP and/or GST involves an interaction with cell surface HS proteoglycans as recently demonstrated by biochemical and genetic experiments . To establish whether the short Tat CPP follows the same internalization process, the fluorescein-labeled Tat peptide was incubated with the same cell lines, namely wild-type (wt) CHO-K1 cells and A-745 mutant cells which are completely defective in HS sulfate expression . As a positive control, uptake of the Tat peptide in HeLa cells was also monitored, as performed in previous studies .
Uptake of the short fluorescein-labeled Tat peptide took place in wt-CHO cells and in the A-745 cell line (Fig. 1; top panels), thus indicating that internalization of this short Tat peptide does not require HS expression. The morphology of CHO cells and their weak adherence on the glass slide rendered subcellular localization more difficult to assess than in HeLa cells. However, a nucleolar concentration in both CHO cell lines clearly took place (as indicated by triangles in Fig. 1) in agreement with data previously reported by our laboratory . Incubation of the Tat peptide was performed over a wide time range (from 15 min to 24 h) and no major differences in intracellular distribution were observed (data no shown).
In order to exclude a possible influence of the conjugated fluorochrome on translocation and intracellular distribution, the same experiments were performed with a Tat peptide labeled with rhodamine maleimide on its C-terminal cysteine residue (Fig. 1; bottom panels) or on its N-terminal residue (data not shown). No difference in the intracellular distribution of the peptide was observed whether wild-type or mutants HS-deficient CHO cells were used. Moreover identical results showing internalization of fluorochrome labeled peptide were obtained with the other HS mutated cell lines described in Experimental procedures (data not shown).
Flow cytometry analysis of the Tat peptide internalization
Fluorescence microscopy clearly indicated internalization of the fluorescent peptide in wild-type and mutant HS deficient CHO cell lines. We then monitored the internalization of the Tat peptide by flow cytometry analysis (Fig. 2), a technique allowing the evaluation of the homogeneity of the cellular population in terms of uptake efficiency. As previously observed by fluorescence microscopy, the internalization of the Tat peptide took place to the same extent in HeLa cells, in wt-CHO cells and in the A-745 (defective in HS proteoglycan) mutant cell line. Moreover internalization appeared to be homogeneous in the whole cell population as a single massif was observed for all cell lines (Fig. 2A). In order to minimize cell handling prior to FACS analysis, no fixation step was included. Avoiding cell fixation and working on living cells eliminates potential artefacts linked with cell processing. FACS analysis showed that cellular uptake and distribution of the peptide was identical in fixed cells or in living cells (data not shown) in agreement with previous data on other cell lines  and with fluorescence microscopy data reported above.
To avoid any possible artefactual data in handling the different cell lines and/or experimental conditions, we reproduced the published results on the internalization of the Tat protein fusion construct . In keeping with previous work , the full-length Tat protein tested as a fusion recombinant protein with GST and GFP was normally internalized in wild-type cell line while the uptake was markedly inhibited on A-745 HS proteoglycans deficient cells (Fig. 2B).
The uptake of Tat CPP was further examined by FACS analysis in dose–response experiments at peptide concentrations ranging from 100 nm to 10 µm for 15 min incubation time (Fig. 3). This was performed on HeLa cells in which uptake of the fused Tat protein has been shown to involve HS proteoglycans . A saturation of the fluorescent signal was observed for extracellular doses above 1 µm. Whether this could reflect a saturation of the potential cellular binding sites for the peptide was not fully investigated. Along the same lines, competition experiments between a fixed dose of fluorescein-Tat peptide (100 nm) and increasing doses of unlabeled Tat peptide (up to 100 µm) only led to a slight reduction of the intracellular signal (data not shown). Whether there is saturation of intracellular binding sites or competition at the level of membrane structures implicated in the Tat peptide uptake is under evaluation.
Comparative FACS analysis of the internalization of the full-length Tat protein construct and the Tat CPP
Differences in the mechanisms of internalization between the Tat peptide and the Tat fused protein was also established by adding the Tat–GFP construct with the rhodamine-labeled Tat CPP in competition. The internalization of the Tat protein fused to GFP was detected by recording the intensity of the GFP signal itself in the 440 nm wavelength range (Fig. 4). Rhodamine-labeled Tat peptide internalization was monitored in the 560 nm wavelength range (data not shown). The Tat–GFP was incubated with wt-CHO in the absence (bold line) or in the presence (dotted line) of a 12.5-fold molar excess of the rhodamine-Tat peptide competitor (80 nm and 1 µm, respectively). As shown in Fig. 4 (panel A), the internalization of the Tat–GFP fusion construct was not significantly reduced in the presence of the excess of the Tat peptide, in keeping with separate internalization pathways. Internalization of the Tat–GFP fusion construct in these conditions was poorly efficient in the A-745 clone (Fig. 4, Panel B) as previously described. A weak displacement of the peak detected in the fluorescein channel could be due to nonreceptor mediated endocytosis during the 24 h incubation time.
Differences in the uptake mechanism between the two Tat entities were also confirmed by the temperature dependence of the internalization process. As shown in Fig. 5, fluorescein-labeled Tat peptide internalization was not abolished by low temperature (dotted lines in Fig. 5, left and right panels) in keeping with our previous data . However a rightward shift of the signal was observed signifying a reduction of the uptake of the Tat peptide at low temperature. Likewise, a threefold reduction of the uptake at 4 °C has been reported for the Antennapedia peptide compared to its cellular uptake at 37 °C . At variance with the Tat–GFP fusion construct (bold line in Fig. 5 left and right), the fluorescent signal was completely inhibited as expected for HS proteoglycans-mediated endocytosis.
In order to confirm the involvment of HS receptors in the uptake of the Tat protein, HeLa cells were treated with heparinase III, an enzyme mostly active on HS proteoglycans . The uptake of full length Tat protein was abolished by such treatment on CHO K1 cells . Likewise, heparinase treatment of HeLa cells completely inhibits the uptake of the Tat–GFP fusion protein (Fig. 6A, dotted line). On the contrary, the internalization of the Tat peptide was not affected by the heparinase treatment (Fig. 6B, dotted line) as similar internalized fluorescence was quantified in heparinase-treated cells compared to untreated cells (Fig. 6B, bold line).
Intracellular vectorization after chemical coupling or genetic fusion to the CPP derived from the HIV-1 Tat appears as a potent tool for the cellular delivery of various biomolecules. These include oligonucleotides , peptides [10–12,14], proteins [4,6,7], nanoparticles  or liposomes . The internalization process is not cell specific as a large number of cell lines tested so far entrapped the translocating peptide. These include cell types which were very poorly transfected by traditional methods as monocyte/macrophages progenitors . Moreover Tat peptide conjugated molecules also pass through the blood brain barrier .
Despite the large number of potential applications of these CPP, the mechanism by which translocation proceeds remains essentially unknown. Interestingly, HS proteoglycans were recently shown to be responsible for the uptake of the Tat protein in a large number of cell lines . The present studies were designed to test whether the short HIV-1 Tat peptide could enter cells via this receptor type. We first made use of CHO mutant cell lines deficient in the expression of HS proteoglycans . We clearly established that the fluorochrome labeled Tat CPP was taken up in these mutant cell lines as efficiently than in wt-CHO or in HeLa cells. The internalization of the peptide in these cell lines was monitored in parallel by fluorescence microscopy and by FACS scan analysis. The first technique confirmed the uptake of the peptide and its nucleolar concentration in CHO cells as previously observed in HeLa cells . The second technique showed that all the cells from a nonsynchronized population entrapped the peptide although the fluorescence intensity could be slightly variable among that population. In addition to these genetic findings, we treated cells with heparinase III prior to their incubation with the different Tat derived molecules in order to digest HS receptors. As previously described , such treatment abolished the internalization of the GFP-fused Tat protein but did not alter the uptake of the Tat CPP. These biochemical evidences confirmed a pathway for the entry of the Tat peptide unrelated to the HS proteoglycan receptors.
Internalization of the Tat CPP did not use a classic endocytosis pathway either, as low temperature incubation of the cells did not impair dramatically the Tat peptide uptake while it abolished the uptake of the GFP-fused Tat protein as expected. Translocation at low temperature was initially described for the Tat peptide  and for the Antennapedia peptide . However, a reduction of the Tat peptide uptake could be observed in our experiments when comparing FACS signal intensity at 4 and at 37 °C (Fig. 5). An identical reduction of the uptake at 4 °C was recently reported for the Antennapedia peptide as well . Even reduced, unambiguous internalization of both peptides at low temperature indicates the existence of an endocytosis independant process for cellular entry. Low temperature translocation of conjugated molecules was recently observed to be also effective as published for liposomes attached with the short Tat peptide . Moreover in our experiments, the rhodamine labeled Tat peptide was coincubated with the GFP-Tat fusion protein to assess the effective inhibition of the receptor mediated endocytosis. Despite a 12.5 molar excess of the Tat peptide, no detectable reduction of the uptake of the Tat–GFP fusion protein was observed when cells were incubated at 37 °C, thus providing additional evidences for separate entry routes for the Tat CPP and the Tat protein.
What might be the reasons underlying the observed differences in cellular uptake between the Tat CPP and the GST–Tat–GFP protein, that both contain the same amino-acid sequence? It might be envisaged that the Tat basic domain is found in different molecular environments in the two molecular species. In the case of the short Tat CPP, the cluster of basic amino acids is likely to be fully accessible to cellular components inducing the translocation event, with particular reference to the arginine residues which appear to be the main determinants for the translocating activity [10,17]. Within the large recombinant protein, the exposure and/or the environment of this basic cluster of amino acids might be different, even if the high hydrophilic nature of this domain likely leads to its exposure at the surface of the GST–Tat fusion protein as it does in the Tat protein itself . Easy accessibility of this domain can be also inferred from the notion that both a GST–Tat and a GST–Tat–GFP fusion proteins are able to transactivate the HIV-1 LTR sequence, an event which requires binding of the Tat basic domain to the TAR sequence on nascent RNAs [21,28,29]. Accordingly, no transactivation was obtained when the arginine residues from the Tat basic domain were mutated to alanine in a HeLa derived cell line . These considerations indirectly reinforce the argument that the basic domain should be exposed at the surface of the Tat-containing recombinant proteins and call other reasons to explain the differences in the mechanism of internalization between the CPP peptide and the Tat-containing proteins. Along this line, it has been reported that chemical coupling of Tat peptides with different length to heterologous proteins resulted in variable efficiency of internalization . In particular, it was reported that the maximum rate of internalization was reached when three or four molecules of a 35-amino-acid Tat peptide (sequence 37–72) were chemically coupled to a large protein cargo. Despite the presence of the basic region (sequence 49–57), the use of shorter chemically-bound peptides (sequence 37–58 or 47–58) was described to be less effective than the Tat peptide 37–72 in the internalization process . Thus, steric hindrance of the heterologous protein itself could reduce the exposure of these shorter peptides to cellular structures, and therefore, reduce the efficiency of translocation. For recombinant fusion proteins, it has been clearly demonstrated that an 11-amino-acid peptide containing only the basic amino-acid cluster is highly efficient in mediating internalization of heterologous proteins when fused at the N-terminal domain of these proteins . Cellular internalization of this peptide fused to β-galactosidase was even observed in vivo in various tissues including the brain after intraperitoneal injection into the mouse . While comparative studies are still lacking, it can be speculated that fusion of the Tat peptide to the N-terminal region of proteins favors its steric accessibility to cellular structures involved in the translocation process, thus accounting for the more efficient cellular internalization of these fusion proteins as compared to their chemically linked counterparts. This would explain why a fusion construct containing only one Tat peptide sequence at its N-terminal end is taken up more efficiently than chemically linked β-galactosidase despite the higher number of peptides. As far as Tat␣peptides are concerned, the 13-amino-acid peptide encompassing the basic domain of Tat (Tat 48–60) was found to be more effective than longer peptides such as Tat 43–60 or Tat 37–60 . The primary sequence of the Tat peptide itself does not seem to be a key feature in cell uptake as several analogues were tested without noticeable variation of the cellular uptake intensity provided the total number of basic amino acids was left unchanged (E. Vivès & B. Lebleu et al. unpublished results). Likewise the retro-inverso form of the peptide did not impair the Tat translocating properties [17,20]. A receptor-mediated mechanism of cellular internalization of the peptide thus appears unlikely. The number of arginine residues within the Tat peptide appeared to be the main determinant for maintaining a high translocating activity as previously shown by alanine-arginine substitution scan [10,17]. Several other arginine-rich peptides, such as flock house virus (FHV) or Rev derived peptides, showed similar cell uptake properties . It was shown recently that short polyarginine peptides were even more potently internalized into cells [17,20]. Moreover the length of the polyarginine tract seems critical, as a maximal rate of internalization was observed for a peptide nine arginine residues in length. The d-form and the retro-inverso form of the polyarginine peptide were␣found to internalize more efficiently. However the higher stability in serum containing cell culture medium of the d-form or the peptides was proposed as the reason of this apparent increased uptake, as the rate of uptake was the same in serum-free medium . As already stated above, this Tat CPP peptide is able to vectorize various cargo molecules inside cells [2,4–7,10,12–16]. Strikingly, efficient internalization in vitro and in vivo of ferromagnetic particles (45 nm diameter) when three to four short Tat peptide molecules were conjugated to it  suggests a noncommon mechanism of entry. Whether binding to other cell surface determinants (as for instance to polar lipid heads) is involved is currently being investigated. Whatever the mechanism however, the possibility to deliver heterologous molecules into different tissues and even through the blood brain barrier has high potential in biotechnology.
We thank Dr Pierre Travo for his help in fluorescence imaging and computerized analysis of pictures. We are grateful to Dr Jean-Jacques Vasseur for performing MALDI-TOF analysis of peptides. We also thank I. Robbins for proofreading of the manuscript. This work was supported by grants from the Association pour la Recherche sur le Cancer to B. L. and E. V. and from MURST and Istituto Superiore di Sanita′, Rome, Italy to M. G.