Design of a Disulfide-less, Pharmacologically Inert, and Chemically Competent Analog of Maurocalcine for the Efficient Transport of Impermeant Compounds into Cells*

Maurocalcine is a 33-mer peptide initially isolated from the venom of a Tunisian scorpion. It has proved itself valuable as a pharmacological activator of the ryanodine receptor and has helped the understanding of the molecular basis underlying excitation-contraction coupling in skeletal muscles. Because of its positively charged nature, it is also an innovative vector for the cell penetration of various compounds. We report a novel maurocalcine analog with improved properties: (i) the complete loss of pharmacological activity, (ii) preservation of the potent ability to carry cargo molecules into cells, and (iii) coupling chemistries not affected by the presence of internal cysteine residues of maurocalcine. We did this by replacing the six internal cysteine residues of maurocalcine by isosteric 2-aminobutyric acid residues and by adding an additional N-terminal biotinylated lysine (for a proof of concept analog) or an N-terminal cysteine residue (for a chemically competent coupling analogue). Additional replacement of a glutamate residue by alanyl at position 12 further improves the potency of these analogues. Coupling to several cargo molecules or nanoparticles are presented to illustrate the cell penetration potency and usefulness of these pharmacologically inactive analogs.

Maurocalcine is a 33-mer peptide initially isolated from the venom of a Tunisian scorpion. It has proved itself valuable as a pharmacological activator of the ryanodine receptor, and has helped the understanding the molecular basis underlying excitationcontraction coupling in skeletal muscles. Because of its positively charged nature, it is also an innovative vector for the cell penetration of various compounds. We report a novel maurocalcine analog with improved properties: (i) the complete loss of pharmacological activity, (ii) preservation of the potent ability to carry cargo molecules into cells, and (iii) with coupling chemistries not affected by the presence of internal cysteine residues of maurocalcine. We did this by replacing the six internal cysteine residues of maurocalcine by isosteric 2-aminobutyric acid residues and by adding an additional Nterminal biotinylated lysine (for a proof of concept analog) or an N-terminal cysteine residue (for a chemically-competent coupling analogue). Additional replacement of a glutamate residue by alanyl at position 12 further improves the potency of these analogues. Coupling to several cargo molecules or nanoparticles are presented to illustrate the cell penetration potency and usefulness of these pharmacologically inactive analogs. Maurocalcine (MCa) is a highly basic 33mer peptide isolated from the venom of the scorpion Scorpio maurus palmatus. It efficiently binds to the ryanodine receptor in skeletal muscles (RyR1 isoform) (1) and promotes channel opening to promote calcium release from the sarcoplasmic reticulum (SR). This pharmacological effect of MCa can be indirectly monitored through the stimulation it exerts on [ 3 H]-ryanodine binding (2). In muscle fibers, MCa produces a transient loss of voltage control of Ca 2+ release from RyR1 channels. This effect is due to an alteration of repolarization-induced closure of RyR1 channels, a process normally under the control of voltagedependent dihydropyridine (DHP)-sensitive calcium channels (3). This function of MCa is due to a partial sequence homology between MCa and a cytoplasmic loop of the DHP-sensitive channel (2). These observations explain why MCa, along with other members of the same family of toxins such as imperatoxin 1A (4), hemicalcin (5), and opicalcin 1 and 2 (6), are useful both for their pharmacological properties and for deciphering fine molecular details of the excitation-contraction coupling process.
Recently, MCa has also proven of interest for its property of efficiently crossing the plasma membrane, either alone or when coupled to a membrane-impermeant cargo protein (2,7). MCa is a highly charged peptide with 12 basic residues out of 33, and a net global positive charge of +8. Most of these residues are on one face of the molecule, the opposite face being mostly hydrophobic in nature. The rich content in basic amino acid residues of MCa is reminiscent of that of all cell penetrating peptides (CPPs) characterized so far (Tat, penetratin and poly-R). Hence, MCa can be classified within an emerging family of toxin CPPs 2 that have no structural homologies apart from their content in basic residues. Recently, a new toxin, crotamin, has been purified from the venom of a South American snake. It also behaves as a CPP but apparently with a preference for dividing cells (8). The great diversity in CPP sequences observed so far suggests that designing new MCa CPP analogues should be easy. This tolerance in sequence variation is possibly due to the diverse nature of membrane receptors implicated in CPP cell translocation. These receptors nevertheless possess a single point in common: they all appear to interact with CPPs on the basis of electrostatic interactions. On the contrary, the structural characteristics of the interaction between MCa and RyR1 appear far more constrained. A single mutation within MCa, the replacement of an Arg residue by an alanyl, abolishes the pharmacological effect of MCa but has only mild effects on its cell penetration efficacy (9). Nevertheless, segregating the pharmacological properties from the cell penetrating properties proved more complex than expected on the sole basis of amino acid substitutions of MCa (9). This appears to be due to structural imperatives of the molecule since it has four functions to fulfill: first, possess the attributes of a CPP, second, conserve sequence homology with the DHP-sensitive calcium channel, third, bind to RyR1 and fourth, activate this latter channel. Many residues contribute both to the pharmacological effects and the cell penetration properties, namely some basic residues that make the functional segregation difficult with simple amino acid substitutions.
To both circumvent these difficulties and take advantage of the flexibility in structural constraints required for MCa cell penetration, we sought a novel strategy for the design of an MCa analogue that would lose its pharmacological activity while retaining most of its cell penetration efficacy. As determined by 1 H-NMR, MCa folds along an inhibitor cystine knot motif with a disulfide bridge pattern of Cys 3 -Cys 17 , Cys 10 -Cys 21 , and Cys 16 -Cys 32 . It contains three β-strands that run from amino acid residues 9-11 (strand 1), 20-23 (strand 2), and 30-33 (strand 3), with βstrands 2 and 3 forming an antiparallel β-sheet (10). In earlier studies on another toxin, maurotoxin (MTX), active on voltage-gated potassium channels, it was found that the disulfide bridges of the peptide play an essential role in the three-dimensional structure of the toxin and thus on its activity (11). We therefore adopted a similar strategy for the design of a novel analogue of MCa in which we replaced all native cysteine residues, engaged in the three disulfide bridges, by isosteric 2-aminobutyric acid residues. The goal was to obtain a structurally-altered analogue displaying a complete loss of pharmacological activity but preserving the positively charged nature of the peptide, a property required for the efficient cell penetration of a CPP. To further validate this analogue, three derivatives were produced that comprise either an N-terminal biotinylated lysine residue with or without an alanyl substitution of Glu 12 , previously shown to favor cell penetration (9), or an N-terminal cysteine residue for coupling chemistries to various cargoes. This chemical synthesis was accompanied by pharmacological assays, toxicity experiments, and proof of concept that cargo penetration is fully conserved. The data indicate that cysteine replacement within MCa can produce potent cell-penetrating analogues of MCa devoid of pharmacological activity.

Experimental Procedures
Reagents -Streptavidin-Cy5 (Strep-Cy5) was from Amersham Biosciences (USA), dihydroethidium (DHE) from Molecular Probes (USA), and rhodamine-and FITC-conjugated concanavalin A were from AbCys and Sigma, respectively. Doxorubicin was from Alexis Biochemicals (USA). The TK705-amino-(polyethyleneglycol) quantum dots (QD) with surface activated amine groups were purchased from InVitrogen. [ 3 H]ryanodine was from Perkin-Elmer (USA). Peptide synthesis -MCa b was synthesized as previously described (7). MCa b -Abu, MCa b -Abu E12A and FITC-Gpep-Cys were purchased from the Department of Pharmaceutical Sciences, University of Ferrara (Italy). Cys-MCa-Abu was synthesized by NeoMPS (France). MTX b -Abu was assembled by the group of Dr. JM Sabatier (France). Conjugation of Cys-MCa-Abu to various cargoes -Cys-MCa-Abu was conjugated to a FITC-Gpep-Cys molecule (sequence derived from Gβ 1 : FITCbeta-AGITSVAFSRSGRLLLAGYDDFN-Abu-NIWDAMKGDRAC-OH) according to the method used by Davidson et al. (12). Briefly, an equimolar mixture of Cys-MCa-Abu and FITC-Gpep-Cys was heated to 65°C for 15 min and then incubated at 37°C for 1 hr. The complex was purified by fast protein liquid chromatography. Cys-MCa-Abu was also conjugated to doxorubicin. Briefly, doxorubicin.HCl (1 mg/ml) was suspended in PBS pH 8.0 and conjugated to Cys-MCa-Abu using succinimidyl-4-(Nmaleimidomethyl) cyclohexane-1-carboxylate (Pierce) according to the protocol of Liang et al. (13). Successful coupling was followed by a 16.5% SDS-PAGE and UV detection of the resulting conjugated doxorubicin-linker-Cys-MCa-Abu peptide. QD were also conjugated to Cys-MCa-Abu. The amino groups of the QD (≈ 100 to 120 functions per particle) were first converted into a maleimide coating as described (14), thereby yielding QD M . Briefly, QD (1 nmol) were incubated for 4 hrs in the dark at room temperature in PBS in the presence of 4-maleimidobutyric acid Nhydrosuccinimide ester (1.8 µmol, in anhydrous dimethylsulfoxide (DMSO)), then purified using NAP-5 columns (Amersham Biosciences). 20 µL of tris-(2-carboxyethyl)phosphine hydrochloride solution 0.5 M was added to 100 nmol of the Cys-MCa-Abu in 80 µL of water for 30 min, and then incubated overnight at room temperature with QD M in the presence of 1 mM EDTA pH 7.4. Non-reacted maleimide groups were quenched for 20 min by adding 2-mercapthoethanol in excess (500 nmol). The QD M -Cys-MCa-Abu conjugates were purified using NAP-5 columns (Amersham Biosciences). Surface modifications of QD were detected by 1% agarose gel electrophoresis in a TAE buffer pH 8, and imaged by fluorescence using a 633 nm excitation wavelength and collecting emitted light above 700 nm. The hydrodynamic diameter of the QD M -Cys-MCa-Abu nanoparticles measured using a Zetasizer Nano (Malvern Instruments) were 25 nm in PBS (20.5 nm for non-modified QD). The QD M -Cys-MCa-Abu concentration was calculated using 532 nm absorbance measurements (QD molar extinction coefficient ε = 2.1 × 10 6 M -1 cm -1 at 532 nm). Cell culture -Wild-type Chinese Hamster Ovary (CHO-K1) cells (from ATCC) were maintained at 37°C in 5% CO 2 in F-12K nutrient medium (Invitrogen) supplemented with 10% (v/v) heatinactivated fetal bovine serum (InVitrogen) and 10,000 units/ml streptomycin and penicillin (Invitrogen). MDA-MB-231 cells from ATCC were grown in Leibovitz L15 medium supplemented with 10% (v/v) heat-inactivated fetal bovine serum and 10,000 units/ml streptomycin and penicillin. Media used for the culture of cerebellar granule neurons was based on Dulbecco's modified Eagle's medium (DMEM, Invitrogen) containing 10 unit/ml penicillin, 10 µg/ml streptomycin, 2 mM L-glutamine, and 10 mM HEPES, 25 mM KCl and 10% fetal bovine serum. Primary cultures were prepared from 6-dayold S/IOPS NMRI mice (Charles River Laboratories), as described previously (15). Preparation of heavy SR vesicles -Heavy SR vesicles were prepared following the method of Kim et al. (16). Protein concentration was measured by the Biuret method. Isolation and preparation of flexor digitorum brevis muscle fibers -Experiments were performed on single skeletal fibers isolated from the flexor digitorum brevis (FDB) muscles from 4to 8-week old OF1 mice (Charles River Laboratories, France) in accordance with the guidelines of the French Ministry of Agriculture (87/848) and of the European Community (86/609/EEC). Procedures for enzymatic isolation of single fibers and partial insulation of the fibers with silicone grease were as previously described (17,18). In brief, mice were killed by cervical dislocation after halothane (Sigma-Aldrich, France) inhalation before removal of the muscles. Muscles were treated with 2 mg/ml collagenase type I (Sigma-Aldrich) in Tyrode solution for 60 min at 37°C. Single fibers were then isolated by triturating the muscles in the experimental chamber. The major part of a single fiber was electrically insulated with silicone grease (Rhodia Siliconi Italia, Treviolo, Italia) so that whole-cell voltage-clamp could be achieved on a short portion of the fiber extremity. All experiments were performed at room temperature (20-22°C).

Structural analyses of MCa b, MCa b -Abu, MCa b -
Abu E12A by circular dichroism -Circular dichroism (CD) spectra were recorded on a Jasco 810 dichrograph using 1 mm-thick quartz cells. Spectra were recorded between 180 and 260 nm at 0.2 nm/min and were averaged from three independent acquisitions. The spectra were corrected for water signal and smoothed by using a third-order least squares polynomial fit.
[ 3 H]-ryanodine binding assay -Heavy SR vesicles (1 mg/ml) were incubated at 37°C for 3 hrs in an assay buffer composed of 5 nM [ 3 H]-ryanodine, 150 mM NaCl, 2 mM EGTA, 2 mM CaCl 2 (pCa=5), and 20 mM HEPES, pH 7.4. 1 µM MCa b , MCa b -Abu or MCa b -Abu E12A was added to the assay buffer just prior to the addition of heavy SR vesicles. [ 3 H]-ryanodine bound to heavy SR vesicles was measured by filtration through Whatmann GF/B glass filters followed by three washes with 5 ml of ice-cold washing buffer composed of 150 mM NaCl, 20 mM HEPES, pH 7.4. Filters were then soaked overnight in 10 ml scintillation cocktail (Cybscint, ICN) and bound radioactivity determined by scintillation spectrometry. Non-specific binding was measured in the presence of 20 µM cold ryanodine. Each experiment was performed in triplicate and repeated three times. All data are presented as mean ± S.D. Fluorescent measurements under voltage-clamp conditions -An RK400 patch-clamp amplifier (BioLogic, France) was used in whole-cell configuration as previously described (18). Voltage-clamp was performed with a microelectrode filled with the intracellular-like solution (in mM: 120 K glutamate, 5 Na 2 -ATP, 5 Na 2 -phosphocreatine, 5.5 MgCl 2 , 5 D-glucose, 5 HEPES adjusted to pH 7.2 with KOH). Indo-1 (Molecular Probes, USA) was present in this solution at 0.2 mM for fluorescence measurements under voltage-clamp conditions. The extracellular solution contained (in mM): 140 TEAmethanesulphonate, 2.5 CaCl 2 , 2 MgCl 2 , 0.002 tetrodotoxin, 10 HEPES, pH 7.2. The tip of the microelectrode was inserted through the silicon, within the insulated part of the fiber. Membrane depolarizations were applied every 30 s from a holding command potential of -80 mV. For the present set of measurements, the cytoplasm was dialyzed with the microelectrode solution which contained the calcium dye Indo-1 and a given peptide to be tested (200 µM for MCa b -Abu and MCa b -Abu E12A, and 100 µM for MCa b ). In order to facilitate intracellular dialysis, the electrode tip was broken within the silicon-insulated portion of the fiber by pushing it back and forth a few times towards the bottom of the chamber. Under these conditions, intracellular equilibration of the solution was awaited for 30 min. Equilibration was followed from the time course of increase of indo-1 fluorescence in the tested portion of the fiber. Indo-1 fluorescence was measured on an inverted Nikon Diaphot epifluorescence microscope equipped with a commercial optical system allowing the simultaneous detection of fluorescence at 405 nm (F 405 ) and 485 nm (F 485 ) by two photomultipliers (IonOptix, Milton, MA) upon 360 nm excitation. Background fluorescence at both emission wavelengths was measured next to each fiber tested and was then subtracted from all measurements. In an earlier study, we showed that 10-100 µM levels of MCa were necessary to affect voltage-activated Ca 2+ release in intact mammalian skeletal muscle fibers (3). For this reason, and despite the cell penetration properties of MCa and of its derivatives, it is easier and less costly to apply these compounds intracellularly through the voltage-clamp electrode rather than in the extracellular medium. Calibration of the Indo-1 response and [Ca 2+ ] intra calculation -The standard ratio method was used with the parameters: R = F 405 / F 485 , with R min , R max , K D and β having their usual definitions.
Results were either expressed in terms of Indo-1 percent saturation or in actual free calcium concentration (18,19). In vivo values for R min , R max and β were measured using procedures previously described (17). No correction was made for Indo-1 Ca 2+ binding and dissociation kinetics. MTT assay -Primary cultures of cerebellar granule neurons were seeded into 96 well micro plates at a density of approximately 8 × 10 4 cells/well. After 4 days of culture, the cells were incubated for 24 hrs at 37°C with 10 µM MCa b, MCa b -Abu, Cys-MCa-Abu or MCa b E12A-Abu. Control wells containing cell culture medium alone or with cells, both without peptide addition, were included in each experiment. The cells were then incubated with 3-(4, 5-dimethylthiazol-2-yl)-2, 5-diphenyltetrazolium bromide (MTT) for 30 min. Conversion of MTT into purple colored MTT formazan by the living cells indicates the degree of 5 cell viability. The crystals were dissolved in DMSO and the optical density was measured at 540 nm using a microplate reader (Biotek ELx-800, Mandel Scientific Inc.) for quantification of cell viability. All assays were run in triplicate.  Fig.  1A illustrates the three-dimensional solution structure of MCa with three disulphide bridges. The aim of this study was to design a MCa analog for which convenient chemical coupling could be performed by using an N-terminal additional cysteine residue. As shown, MCa already contains six cysteine residues that contribute to the folding of the peptide to form an inhibitor cysteine knot motif. Adding an additional cysteine residue at the N-terminus may significantly change the normal folding of the peptide and the classical disulfide bridge arrangement (C 1 -C 4 , C 2 -C 5 , and C 3 -C 6 ) in an unpredicted manner. In turn, this could significantly affect the pharmacological activity and cell-penetration properties of the resulting molecule(s). To facilitate chemical coupling strategies of MCa to cargo molecules, we investigated the requirement of disulfide bridges on MCa's pharmacology and cell penetration properties. Several analogs were synthesized, each with an N-terminal biotinylated lysine for easy coupling to fluorescent streptavidin molecules (our reporter cargo for this study): MCa b , intact with the native disulfide bridges, MCa b -Abu in which we replaced all internal cysteine residues by isosteric 2-aminobutyric acid (Abu, Fig. 1B), and MCa b -Abu E12A, an analog of MCa b -Abu in which Glu 12 was replaced by alanyl, a substitution known to improve cell penetration efficacy of MCa b (9) (Fig.  1C). We would expect that removing the disulfide bridges of MCa might affect its pharmacology more than its cell penetration properties for two reasons. First, disulfide bridge patterns are known to contribute to the pharmacological activity of 6 toxins (20,21). Second, the structural requirements for the pharmacological activity of MCa were shown to be more stringent than those required for cell penetration as probed by alanine scanning of MCa (9). Finally, we synthesized MTX b -Abu, a biotinylated version of MTX that has no cell penetrating properties on its own and that acts on voltage-dependent potassium channels (Fig. 1D). This peptide, in which we also replaced six internal cysteine residues by Abu derivatives, as for the MCa b -Abu peptide, was used as a negative control in cell penetration experiments. Removing disulfide bridges in MCa disrupts the secondary structure of the peptide -Circular dichroism analyses were performed for MCa b , MCa b -Abu, and MCa b -Abu E12A (Fig. 2). As shown, the spectra for MCa b -Abu and MCa b -Abu E12A differed significantly from MCa b indicating the alteration of secondary structures of the peptides. These observations are coherent with a role of disulfide brides in the acquisition/stabilization of secondary structures of toxins (11). These data clearly indicate that blocking disulfide bridge formation in MCa is a successful strategy to alter the structure of MCa. Cysteine replacement by Abu derivatives produces pharmacologically-inert MCa analogues -MCa is known to bind to RyR1 from skeletal muscles (2). Upon binding, it modifies the conformation of this intracellular calcium channel in such a way that it favours binding of [ 3 H]-ryanodine to its receptor, probably by converting low affinity binding sites to high affinity one. The effect of 1 µM MCa b on [ 3 H]-ryanodine binding to sarcoplasmic reticulum (SR) containing RyR1 was confirmed in this study (Fig. 3). An average stimulation in binding of 59.7-fold was measured (n=3). The value of this stimulation depends upon the basal level of binding and can occasionally be lower, down to 7fold (not shown). Importantly, the two analogs in which the cysteine residues were replaced by Abu derivatives, MCa b -Abu and MCa b -Abu E12A, had no significant effect on [ 3 H]-ryanodine binding indicating that the structural impact of these substitutions fully blocked the effect of MCa on RyR1. Slight reductions in [ 3 H]-ryanodine binding were observed although these effects were not significant. Higher concentrations of these two analogs were also without effect (not shown). The E12A substituted analogue thus proved no better than MCa b -Abu in spite of the fact that a similar mutation in the folded MCa has higher affinity for RyR1 (9).

Synthesis of disulphide-less analogs of MCa -
We also investigated the effect of these analogs on Ca 2+ homeostasis in muscles fibers. The effects of the three peptides MCa b , MCa b -Abu and MCa b -Abu E12A were tested on the free Ca 2+ transients elicited by voltage-clamp depolarizations in single skeletal muscle fibers from mouse. Fig. 4A shows representative Indo-1 saturation traces obtained in response to pulses from -80 mV to +10 mV of 5, 10, 20 and 50 ms duration in (from top to bottom) a control fiber, a fiber dialyzed with 100 µM MCa b , a fiber dialyzed with 200 µM MCa b -Abu or a fiber dialyzed with 200 µM MCa b -Abu E12A, respectively. These high concentrations of peptides were chosen to ensure that effective intracellular levels could be reached after equilibration (see Methods). As previously described (3) Conserved cell penetration properties of disulfideless MCa analogues -Biotinylated MCa, MCa b , favours the efficient penetration of fluorescentlylabelled streptavidine (Strep) (7,9,22), through macropinocytosis (23). The cell penetration of MCa b /Strep complexes involves one or multiple steps of attachment to the membrane through binding to cell surface glycosaminoglycans and negatively charged lipids. The complex can be found predominantly in endosomes, leading to a punctate cytoplasmic distribution, a pattern that is supposedly linked to the nature of the cargo rather than the vector itself. Here, we investigated whether the disulfide-less analogues of MCa b were capable of carrying Strep-Cy5 into CHO cells (Fig.  5A). As shown, 1 µM of Strep-Cy5 alone is unable to enter CHO cells. This was true for 100% of the cells examined by confocal microscopy (n=200). However, when coupled to either MCa b , MCa b -Abu or MCa b -Abu E12A in a 4 to 1 ratio and incubated for 2 hrs with living CHO cells, the resulting complexes gave punctate staining. Here again, 100% of the cells were positively stained by Strep-Cy5 (n=200 for each condition). This punctate staining is an indication of endosomal distribution further confirming that it may be linked to the nature of the cargo rather than the vector, since similar distributions are observed for the MCa analogs despite the structural changes in the vector. In contrast, no cell penetration of Strep-Cy5 was observed in CHO cells using a biotinylated analogue of MTX, MTX b -Abu, a voltage-gated K + channel blocker ( Fig. 5B; 0% of n=200 cells examined by confocal microscopy). This indicates that MCa b -Abu is unlikely to penetrate non-specifically into cells owing to its amphipatic nature.
In order to obtain half-maximal Penetration Concentration values (PC 50 ) and the extent of total penetration of the complex, quantification of cell penetration by fluorescence activated cell sorting (FACS) was performed. Quantification is a way to distinguish the cell penetration efficacies of various MCa analogs (9) or of various cell lines (23). As shown in Fig. 6, after 2 hrs incubation, both MCa b -Abu and MCa b -Abu E12A analogs were less effective than MCa b since the total Strep-Cy5 fluorescence entering CHO cells at maximally effective concentration of the peptides (10 µM) was reduced on average by 28% and 59% for MCa b -Abu E12A and MCa b -Abu, respectively (Fig. 6). It is of interest that the E12A mutant was more effective than the analog without mutation. A similar observation was made with the folded version of this mutation (9). This effect is most likely due to an increase in the basic content of the molecule. With regard to PC 50 values, only slight reductions in the efficacies of the analogs were observed. In these experiments, we measured PC 50 values of 669 ± 27 nM for MCa b /Strep-Cy5 (n=3), 910 ± 51 nM for MCa b -Abu/Strep-Cy5 (n=3) and 1042 ± 89 nM for MCa b -Abu E12A/Strep-Cy5 (n=3). Therefore, in the case of disulfide-less analogs, maximal cell penetration of Strep-Cy5 could be obtained by only using slightly increased peptide concentrations. Adding an N-terminal cysteine residue to a disulfide-less MCa analog produces a chemicallycompetent peptide for the cell penetration of various cargoes -Although the use of Strep brings proof of concept, it remains a useless cargo for many biological applications, especially if it concentrates into endosomes. We therefore produced a novel disulfide-less analogue of MCa in which we simply replaced the N-terminal biotinylated lysine by a cysteine residue in order to make various chemical coupling strategies possible (Cys-MCa-Abu whose sequence is shown in Fig.  7A). Next, this peptide was coupled to various cargoes using several chemistries and investigated for its ability to enter cells (Fig. 7B-D). Cys-MCa-Abu was first coupled to a FITC fluorescent peptide containing a C-terminal cysteine residue. The coupling consisted in favoring the formation of a disulfide bridge among the two peptides. At 1 µM, this complex entered CHO cells (100% of the cells by confocal microscopy, n=200) whereas incubation of CHO cells with 1 µM FITC-labeled peptide alone resulted in no cell penetration ( Fig.  7B; 0% of n=200 cells observed). The FITCpeptide complex stained diffusely the cells indicating that it reached both the cytoplasm and the nucleus either through direct plasma membrane translocation or through endosomal escape. Since the Strep and FITC complexes distribute differently in the cell, it appears that the cargo rather than the vector determines the cell distribution. Cys-MCa-Abu was also coupled to doxorubicin, an anti-tumor drug, which penetrates into MDA-MB-231 cells by itself, and concentrates in nuclei where it acts on DNA replication. The chemical coupling was performed 8 using a cross-linker, allowing a directional coupling with the unique amino group of doxorubicin and the SH function of the cysteine residue of Cys-MCa-Abu. As shown, incubation of MDA-MB-231 cells with either 3 µM doxorubicin or doxorubicin-linker-Cys-MCa-Abu resulted in significantly different cell distributions (Fig. 7C). Covalent linkage of doxorubicin with Cys-MCa-Abu produced a marked cytoplasmic localization, also diffuse, in marked contrast to the predominantly nuclear distribution of doxorubicin. In 96% of the cells examined (n=200), the coupling to Cys-MCa-Abu resulted in a cytoplasmic distribution of doxorubicin. Finally, near-infrared emitting quantum dots from Invitrogen, with amine surfaces, were linked to Cys-MCa-Abu using maleimide chemistry. Maleimide was added to QD yielding QD M , that could further be linked to the SH function of the Cys residue of Cys-MCa-Abu. Incubation of 50 nM of QD M -Cys-MCa-Abu with CHO cells resulted in cell penetration with a diffuse distribution in both the cytoplasm and the nucleus ( Fig. 7D; 100% of cells observed; n=200). No penetration was observed with QD M incubation alone (0% of cells, n=200). The rather efficient penetration observed with such a low concentration of QD M -Cys-MCa-Abu might be explained by the high potential of Cys-MCa-Abu grafting at the surface of QD M (about 100 to 120 maleimide converted functions at the surface of a single QD). Cell toxicity of MCa analogues -To be considered as good cell penetrating vectors, peptides must have limited cell toxicity. We investigated cerebellar granule cell survival after incubation with 10 µM of free vector peptide for 24 hrs (Fig.  8). This cell system is generally used for the evaluation of neuronal survival. It is more stringent than the use of cell lines that are more resistant to toxic agents. The conditions used in this study were deliberately extreme considering the conditions required for the cell penetration of our vectors. First, incubation times in the range of an hour are largely sufficient; second, cell penetration is observed at lower concentrations than 10 µM; and third, we have evidence suggesting that intracellular concentrations of cell penetrating peptides increase markedly when cells are incubated with the vector alone rather than in complex with the cargo. Nevertheless, the experimental conditions used here indicate that only one of the analog had a significant effect on neuronal survival (MCa b -Abu, mean survival of 74.3 ± 4.9%, n=6), whereas all other analogs were without significant effects despite of the conditions used here (n=6 for each condition).

CONCLUDING REMARKS
By simple substitution of the six cysteine residues of MCa, we have produced a series of analogs that are devoid of secondary structure and therefore of pharmacological activity. These analogs nevertheless possess cell-penetrating properties that are closely related to those established for native MCa. This finding indicates that the proper folding of MCa is not essential for the membrane translocation. Similarly, the spatial separation of a basic face and a hydrophobic face of MCa seems of little importance for cell penetration. It cannot be excluded however that during the translocation process itself, when the peptide interacts with negatively charged lipids, it may adopt a conformation close to that of native MCa. The observation that MTX b -Abu, used as a negative control, does not penetrate cells indicates that MCa-Abu does not act as a detergent by means of its amphipathic nature. Slight increase in effective concentrations or reductions in total transport capacities of these analogs barely counterbalance the benefits of the loss of pharmacological effects or the ability to graft an extra-N-terminal cysteine residue to the sequence for versatile cargo coupling strategies. Our results further indicate that the Cys-MCa-Abu vector is valuable for the cell delivery of a variety of cargoes. The cytoplasmic localization of the delivered cargoes using this vector should be invaluable for many biological applications for which targeting to this compartment are an absolute requirement. For future applications, it might be recommended to assess the cell toxicity of the vector/cargo complexes of interest in the conditions of the application. The benefits of all cell penetrating peptides undoubtedly lie in the value of penetration/toxicity concentration ratio of the formed complexes, a value that needs to be determined for each application.

FOOTNOTES
We thank the team of Pr. Rémy Sadoul for help with cerebellar granule cell cultures. We also thank Dr. Jonathan Coles for critical reading of the manuscript. We acknowledge financial support from Inserm and the Fond de Valorisation of the CEA. Narendra Ram is a fellow of the Région Rhône-Alpes and is supported by an Emergence grant to M.D.W.