Journal of Cell Science 102, 717-722 (1992)Printed in Great Britain © The Company of Biologists Limited 1992

717

Antennapedia homeobox as a signal for the cellular internalization and

nuclear addressing of a small exogenous peptide

F. PEREZ1, A. JOLIOT1, E. BLOCH-GALLEGO2, A. ZAHRAOUI3, A. TRILLER4 and

A. PROCHIANTZ1 •lCNRS URA 1414, Ecole Normale Supirieurc, 46 rue d'Ulm, 75230 Paris Cedex 05, France2CNRS UPR 8402, BP 5051, Route de Mende, 34033 Montpellier, France3INSERM U 248, Faculty de Midecine Lariboisiere Saint-Louis, 10 avenue de Verdun, 75010 Paris, France4INSERM U 261, Institut Pasteur, 26 rue du Docteur Roux, 75724 Paris Cedex 15, France

•Author for correspondence

Summary

In a previous study we demonstrated that a homeoboxpeptide corresponding to the 60 amino acid long DNA-binding region of the DrosophUa antennapedia homeo-protein was capable of crossing the plasma membrane ofcells in culture. This finding has led us to investigatewhether chimeric molecules encompassing the homeo-box would behave in a similar manner. We demonstratehere that a peptide of 93 amino acids composed of thehomeobox and of the C terminus of Rab3, a small GTP-binding protein, crosses the membrane of myoblasts,myotubes and neurons and is conveyed to their nuclei.

This transport is highly efficient, is observed in all thecells present in the culture and occurs at 37°C and 12°Cwithout quantitative peptide degradation. Beyond itstheoretical implications for our current views on cellularinteractions, this finding could have technical repercus-sions on the development of drugs with intracellulartargets.

Key words: antennapedia homeobox, peptides,transmembrane carriers.

Introduction

In a previous study we made the unexpected obser-vation that pAntp, the 60 amino acid homeobox of theDrosophUa homeoprotein antennapedia, is able topenetrate nerve cells in culture and to accumulate intheir nuclei through an energy-independent mechanism(Joliot et al., 1991a,b). Internalization of pAntp was atleast four times more efficient in cells expressing neuralcell adhesion molecule (N-CAM)-linked a^2,8-poly-sialic acid chains (PSA). Structural comparisons sugges-ted that the enhancing effect of PSA is due to itsstructural similarities to the motif that, in the DNAbackbone, is recognized by homeoboxes (Joliot et al.,1991b).

A first implication of this unexpected finding is thatlarger molecules encompassing homeobox sequencesmight also be able to cross the cell membrane throughan unconventional pathway and, in consequence, thatthe homeobox peptide could serve as a "signal" for theinternalization of other polypeptides. Among suchpolypeptides encompassing the homeobox are homeo-proteins or homeoprotein isoforms; thus raising thepossibility that such molecules might, in some circum-

stances, behave as paracrine or autocrine factors(Prochiantz et al., 1992a,b).

A second point of interest is the mechanism oftranslocation of the homeodomain through the mem-brane. Indeed, it is rather unconventional that 60 aminoacid long peptides should travel across the plasmamembrane. As already mentioned, previous studieshave shown that the attachment to PSA at the surface ofthe cell accelerates the penetration of the peptide butthe ability of pAntp to penetrate into a large number ofcells devoid of PSA suggests the existence of a moregeneral mechanism allowing peptide penetration.

Finally and more pragmatically, this finding mightprovide a non-traumatic protocol for the introductionof a large number of molecules into living cells. Indeed,injection of peptides into cells is a direct means ofgaining certain types of information in cell biology, andhas many potential applications for therapeutic devel-opments. However, this approach has been limited tocell types with specific physical characteristics, inparticular large cytoplasmic volume, and several popu-lations of fragile cells such as mammalian neurons arerefractory to injection protocols. In this latter case,peptides have been "injected" by trituration, a variant

718 F. Perez and others

of the scrape loading technique, which, although useful,is quite traumatic and retricted to a short time windowduring development (Borasio et al., 1989; Ayala et al.,1990).

Furthermore, even in the case of cells with therequired physical characteristics, injection of peptidesinto single cells is time consuming and the analysis islimited to a few hundred cells at a time. Therefore, thedevelopment of molecules able to serve as transmem-brane carriers in order to introduce macromoleculesinto the cytoplasm or the nucleus of large numbers ofliving cells would be of great interest.

In the present report, and as a first step in answeringthe questions raised by our previous observation, wetested the ability of a chimeric molecule, encompassingpAntp, to translocate through the cell membrane and tobe conveyed to the cell nucleus. The C-terminal regionof a small GTP-binding protein rab3 was linked topAntp and we followed the fate of the chimericmolecule in cultures of myoblasts and neurons. Wedemonstrate that pAntp mediates the penetration ofthe foreign polypeptide and conveys it to the nucleus.

Materials and methods

Polypeptide productionAntennapedia homeobox and rab3 were cloned, producedand purified as described previously (Joliot et al., 1991a;Zahraoui et al., 1989). To clone the chimeric molecule AR3C,plasmid pAHl containing the coding sequence of anten-napedia homeobox was modified in that the TGA terminationcodon was replaced by a Ser codon within an Xhol restrictionsite (pAH61S). To this end, pAHl was linearized at theBamHl site and amplified by the polymerase chain reaction(PCR) between oligonucleotides CGGGATCCGGCTG-CTAACAAAGCCC (annealing to pET-3a sequences down-stream from the homeobox) and CGGGATCCTC-GAGTTCTCCTTCiTCCAC (the 16 last nucleotides ofwhich anneal to the 3'end of the homeobox sequence).Conditions for PCR were: 94°C, 1 min; 60.5°C, 1 min; 72°C,12 min (25 cycles).

The sequence coding for the 33 C-terminal amino acids ofrab3A was amplified by PCR between oligonucleotidesCAAACTCGAGGTCCGAGTCGTTGGACAC and TTGG-ATCCTCAGCAGGCGCAGTCCT, allowing for the intro-duction of BamHl and Xhol sites at 3' and 5', respectively.Conditions for PCR were: 94°C, 1 min; 54°C, 1 min; 72°C, 10min (25 cycles). The amplified 121 bp fragment was cloned inbluescript between Xhol and BamHl. After sequenceverification, the fragment was inserted in pAH61S. Theresulting plasmid pAR3C was expressed in Escherichia coliBL21(DE3)LysS and the proteins were produced andextracted (Joliot et al., 1991a).

AR3C was purified by loading clarified bacterial extracts onheparin-Ultrogel equilibrated in 0.1 M KC1. AR3C was elutedbetween 0.3 and 0.8 M KC1. Metabolic labelling of theproteins was performed in IPTG (isopropyl-/3-D-thiogalacto-side)-induced bacteria by the addition of rifampicin (200/ig/ml) for 15 min and S-methionine (1000 Ci/mmol, 15/iCi/ml) for 5 min. After a 10 min chase with 2 fjM methionine,the cells were pelleted and the proteins extracted as describedby Joliot et al. (1991a).

ImmunoblotsProteins separated by SDS-PAGE (12%) were electrotrans-ferred to nitrocellose. Filters were saturated (30 min) with 1%gelatine and 0.5% bovine serum albumin (BSA) in TBS (Tris-buffered saline) before further incubation (30 min) in thesame buffer plus 0.2% Tween 20 (buffer A) and incubatedwith the affinity-purified anti-rab3A polyclonal antibody(1/1000 in buffer A). After 4 washes in the same bufferwithout BSA (buffer B), filters were incubated with 125I-Protein A (0.5 /uCi/ml, Amersham), washed 4 times in bufferB and exposed for autoradiogTaphy.

Cell cultureChick embryo myoblasts were cultured as described byHenderson et al. (1981), except that they were plated on apolyornithine-laminin substratum at a density of 5X104 cellsper cm2. Neurons from the cortex of embryonic rat wereprepared as already described and were plated at a density of5 x 104 cells/cm2 (Joliot et al., 1991b). Following 4 washes withculture medium, the cells were fixed with ethanol/acetic acid(5 min, —20°C). Immunocytochemical procedures using theanti-rab3 antibody at a dilution of 1 to 100 were as describedearlier (Joliot et al., 1991b). Confocal data were obtained witha Phoibos* 1000 (Molecular Dynamics) (Joliot et al., 1991b).

Intracellular distributionCells (2X106 per 35 mm dish) incubated with radioactivepeptides were cooled to 4CC before removing the medium(M), washed with PBS and incubated at 4°C with 700 1̂ NP40,0.5% in PBS, containing PEFA block (0.5 mM), a2-macroglobulin (1 /ig/ml) and leupeptin (10 /ig/ml). The 700 /Jwere centrifuged at 2000 g through a cushion of 0.3 M sucrose(in PBS plus protease inhibitors); 50% of the nuclear pelletand 10% of the supernatant or of the medium were analysedby PAGE.

Results

To examine whether pAntp could serve as a carriersequence for the penetration of peptides into the cellcytoplasm and nucleus we prepared a fusion proteinbetween pAntp and 33 amino acids corresponding tothe C terminus of human rab3A, a small GTP-bindingprotein (Zahraoui et al., 1989). The choice of rab3 wasarbitrary and made in consideration of ongoing projectsin the laboratory. The chimeric molecule was namedpAntp-rab3-Cter and abbreviated AR3C.

The expression in E. coli of a chimeric constructunder the transcriptional control of bacteriophage T7polymerase permitted the synthesis of large amounts ofthe chimeric molecule. Following induction in thepresence of rifampicin, added [ SJmethionine wasalmost exclusively incorporated into the exogenousproteins, as demonstrated in Fig. 1A for rab3 andAR3C. The induced chimeric molecule was expressedvery efficiently and recognized by an affinity-purifiedanti-rab3 antibody (Fig. IB).

Thanks to the DNA-binding properties of thehomeobox the chimera could be further purified onheparin-Ultrogel columns (Fig. 1C). The presence ofpAntp is also demonstrated by the capability of thechimeric molecule to retard the migration of a fragmentof the Hox-1.3 promoter (pHox-1.3) which encom-

Internalization and nuclear addressing of a polypeptide 719

97.4 —66.2 —

45.0 —

rab3NI I

AR3CNI I

B

97.4 —66.2 —

45.0 —

AR3CNI I

•ft

AR3CNI I

31.0 —

21.5 —

14.3-

31.0 —

21.5 —

14.3 —

1 2 3 4

97.4 —

66.2 —

45.0 —

31.0 —

21.5 —

14.3 —

1 2 3 4

AR3CpAntp C 200S 200NS 100NS 0

ir

-AR3C

^—pAntp

1 2 3 4 5 6

Fig. 1. Production andcharacterization of peptides.(A) PAGE analysis ofradioactive rab3 (lanes 1 and2) and AR3C proteins (lanes 3and 4) in extracts from non-induced (NI) and induced (I)bacteria. (B) Coomassiestaining (lanes 1 and 2) andcorresponding immunoblotsusing the anti-rab3 antibody(lanes 3 and 4) of extractsfrom non-induced (NI) orinduced (I) AR3C-producingbacteria. (C) PAGE analysis(Coomassie blue staining) ofthe proteins extracted frombacteria producing AR3C andeluting from heparin-Ultrogelat 0.1 M (lane 1) or 0.8 MKC1 (lane 2). (D) Gelretardation of the Hox-1.3promoter fragment (pHox-1.3)with either the homeoboxpeptide (pAntp) or thechimeric peptide (AR3C). Thefree probe (F) was retarded bypAntp (lane 1) and by AR3C(lane 6). AR3C was entirelydisplaced by 200 ng of non-radioactive pHox-1.3 (200Slane 3), but not by 100 ng(100NS lane 5) or 200 ng(200NS lane 4) of non-specificDNA. Values on the left in A-Care Mrxl0~3.

passes the cognate sequence for homeobox binding(Fig. ID) (Joliot et al., 1991a).

The chimeric peptide or rab3 at 2 /xg/ml was added for12 hours to 105 chick myoblasts and to mouse corticalneurons in culture. The results obtained with the twocell types (both contain polysialylated NCAM) areillustrated in Fig. 2. Immunocytochemical staining withthe affinity-purified anti-rab3 antibody demonstratesthe accumulation of AR3C in myoblast nuclei (Fig.2A). No staining was obtained when rab3 (not linked topAntp) was added to the cells (Fig. 2B).

Confocal microscopy was used to verify the intra-cellular and intranuclear localization of AR3C inmyoblasts and myotubes (Fig. 2C), and in neurons (Fig.2D). The chimera accumulated in the nuclei of both celltypes and of every single cell present in the culture.Non-nuclear staining was also observed, in particular inneurons, where it was due both to the presence of somechimeric peptide at the level of the plasma membrane(arrowheads) and to endogenous rab3 normally presentin this cell type (Olofsson et al., 1988). The internaliz-

ation of AR3C was observed at all concentrationstested, even when tracer quantities were used, asdemonstrated below for radioactive peptides.

The respective concentrations of radioactive fusionproteins in the culture medium, in cytoplasm plusmembrane and in nuclei were quantified as follows:2X105 ct^/min of radioactive proteins identical to thoseof Fig. 1A were added to 2xlO6 cortical neurons thatwere plated at embryonic day 16 and previouslycultured for 3 days in 1 ml of a medium chemicallydefined to promote neuron selective survival anddevelopment (Rousselet et al., 1990). After 12 hours at37°C, 12°C or 6°C, the cells were carefully washed andnuclei were separated from the membrane plus cytosolfraction.

Portions of each sample were analysed by polyacryl-amide gel electrophoresis and autoradiography. Asdemonstrated in the autoradiogram of Fig. 3A and inTable 1, most AR3C was recovered in the culturemedium (75%) and in the nuclei (20%). In addition, theamounts and distribution of the chimera in the three

720 F. Perez and others

Fig. 2. Intemalization of AR3C demonstrated by indirect immunofluorescence. (A) Immunostaining of chick myoblastspreincubated with AR3C. Note the presence of the peptide in every cell. Bar, 50 jxm. (B) Immunostaining of chick embryomyoblasts preincubated with rab3. Bar, 50 /an. (C) Confocal section showing an uneven nuclear distribution of the AR3Cimmunoreactivity in the nuclei of fused myoblasts. Bar, 5 /an. (D) Confocal section showing the presence of AR3C in aneuronal nucleus. Fluorescent clusters are present at the periphery of the cell (arrowheads). Bar, 5 ^m.

BAR3C,12° AK3C.370

N C/M M N C/M Mrah3,12° rab3,37°

N C/M M N C/M M

—32.5—

—27.5—

—18.5—

Fig. 3. Intemalization of AR3C and rab3 at 12°C and 37°C.(A) E. coli extracts containing radioactive AR3C (Fig. 1A)were incubated overnight with embryonic neurons. Theextracellular medium (M), cytoplasmic plus membranefraction (C/M) and nuclear fraction (N) were prepared foranalysis by PAGE. Incubation at 12°C did not block theintranuclear accumulation of AR3C. (B) Same experimentas in A but with rab3 instead of AR3C. Intemalization ofrab3 took place at 37°C but was entirely blocked at 12°C.

Table 1. Quantification of AR3C and rab3 in themedium (M), cytoplasmic plus membrane (C/M) and

nuclear pellet (N) fractions

AR3C rab 3A

37°C 12°C 37°C 12°C

MC/MN

77.0±1.94.4±0.7

18.6±2.3*

83.6±4.02.6±0.6

13.8±3.6*

92.4±0.32.0±0.35.6±0.6t

97.0±1.42.1±1.40.9±0.1t

Fractions were prepared as indicated in Materials and methodsfor 4 independent experiments at 37°C and 12°C. Values are %.

*Not significant (37°C versus 12°C).tP<0.01 (Student's /-test).

compartments were identical at 12°C and 37°C, sugges-ting that the penetration cannot be explained byclassical endocytosis (Table 1).

In contrast, the very limited intemalization of rab3

Internalization and nuclear addressing of a polypeptide 721

observed at 37°C was probably due to some endocytosisand to some contamination of the nuclear fraction withendosomes. Its complete inhibition at 12°C indicatedthat endocytosis was strongly reduced at this lattertemperature (Fig. 3B, Table 1).

Discussion

Introducing foreign peptides into living cells is com-monly achieved either by transfection of expressionvectors or by injection and scrape loading protocols(Lemons et al., 1988; Borasio et al., 1989; Ayala et al.,1990; Geller and Freese, 1990; Holt et al., 1990;Loeffler et al., 1990). These approaches, although veryuseful, suffer severe limitations, in particular in theyield of modified cells and the disruption of cellmembranes.

A less-invasive way to introduce peptides into cells isto link them with molecules normally internalized, suchas folate (Leamon and Low, 1991) or bacterial toxins(Prior et al., 1991; Stenmark et al., 1991). Comparedwith these models, our system has the advantage of notbeing based on the use of bacterial toxins and notdepending upon classical endocytosis, therefore limit-ing peptide loss by endosomal degradation.

Another advantage of the model is the high numberof copies that can be introduced, leading to concen-trations as high as 6xlO~7 M of the fusion proteinwithin a few hours (106 molecules per cell). Thepreferential localization of the peptide in the nucleusshould, indeed, allow targeting of drugs with potentialnuclear activities. However, the amount of protein inthe cytoplasm (one fourth of that in the nucleus, seeTable 1) is not negligible.

Furthermore, from the biotechnological point ofview, it is interesting to note that, in addition to nerveand muscle cells, AR3C has also been shown topenetrate into all cells so far tested, including fibro-blasts, the O-2A astrocyte oligodendrocyte progenitorand the pheochromocytoma cell line PC12 (not shown).The efficiency of penetration into cells devoid of PSAwas reduced between four- and tenfold as compared toneurons and muscle cells but we always found thatevery single cell in the culture was able to internalizethe peptide.

In the present series of experiments, we have nottried to inject the chimeric peptide AR3C in vivo.However, experiments with the homeobox alone havedemonstrated that it can be internalized by embryoniccells in vivo (Prochiantz et al., 1992b). The similarity ofresults obtained so far with the fusion-peptide and withthe homeobox favour the possibility that the formershould also be able to penetrate the cells in vivo.

The physical mechanism allowing pAntp penetrationis not understood. The finding that pAntp is soluble butcontains a high number of hydrophobic residuessuggests the following mechanism of translocation: thevicinity of the plasma membrane could be sufficientlyacidic to provoke a hydrophilic to hydrophobic tran-sition in the conformation of the peptide, thus trigger-

ing its insertion into the plasma membrane. Conversely,the neutral pH of the cytoplasm would create trieappropriate conditions for a reverse change of confor-mation, allowing for complete translocation across theplasma membrane.

This model presents several advantages, one of themis to explain the role of PSA in the fast translocationobserved in young muscle cells and neurons. Indeed,PSA might not only be able to concentrate the peptideat the surface of the cells but might also be responsible,thanks to its high number of carboxylic groups, for thelocal low pH environment of the external side of theplasma membrane.

Another interesting feature of the model is itssimilarity with that proposed by Stenmark et al. (1991)and Jiang et al. (1991) to explain the translocation ofdiphtheria toxin A subunit through the plasma mem-brane. It is interesting to note that in this latter case a 30amino acid long polypeptide linked to the C terminus ofdiphtheria toxin A subunit is translocated into the Verocell cytosol only when transient low pH conditions areimposed on the cells (Stenmark et al., 1991). However,the efficiency of translocation with diphtheria toxin as avector is quite low, since the maximal cytosolicconcentration of foreign peptides is only in the 10~9 Mrange.

The possible similarity between the process ofinternalization of AR3C and that of diphtheria toxinraises the possibility that the translocation occurs acrossthe membranes of early endosomes. This is veryunlikely in view of the internalization of AR3C (notshown) or pAntp (Joliot et al., 1991b) by neuronal cellsnot only at 12°C but also at 4°C. However, we cannotpreclude the possibility that internalization by other celltypes, in particular those deprived of PSA, requiresprior acidification in the early endosomal compartment.

It is worth underlining that "fusion-proteins" encom-passing homeoboxes exist naturally in the form ofhomeoproteins. Our results thus raise the possibilitythat homeoproteins or fragments of homeoproteinspassing from one cell to another may act as paracrineftww-activating factors. This suggestion certainly suf-fers from the absence of signal peptide in homeopro-teins. However, in addition to the fact that all isoformsof homeoproteins have not been deciphered^ it shouldbe noted that several proteins such as interleukin-1 andbasic FGF (fibroblast growth factor) are secreted in theabsence of signal peptide (Rubartelli et al., 1990;Mignatti et al., 1991). Basic FGF is of particular interestin the present context, since it has been demonstratedto be able to translocate into the nucleus where it itmay serve as a frww-activating factor (Baldin et al.,1990).

We should thus consider with some caution thedistinction between growth factors and frww-activatingfactors and, in fact, several recent studies havedemonstrated the presence of classical growth factors incell nuclei (Yeh et al., 1987; Renko et al., 1990;Wetmore et al., 1991). However, if we compare any ofthese growth factors, in particular bFGF, with homeo-proteins, a clear difference is that homeoproteins are