Synthesis, conformational analysis and biological activity of cyclic analogs of the octadecaneuropeptide ODN 

Design of a potent endozepine antagonist

Authors

  • Jérôme Leprince,

    1. Institut Fédératif de Recherches Multidisciplinaires sur les Peptides (IFRMP 23), Laboratoire de Neuroendocrinologie Cellulaire et Moléculaire, Institut National de la Santé et de la Recherche Médicale Unité 413, CNRS, Université de Rouen, Mont-Saint-Aignan, France;
    Search for more papers by this author
  • Hassan Oulyadi,

    1. IFRMP 23, Laboratoire de Résonance Magnétique Nucléaire, Institut de Recherche en Chimie Organique Fine, CNRS Unité Mixte de Recherches 6014, Université de Rouen, Mont-Saint-Aignan, France;
    Search for more papers by this author
  • David Vaudry,

    1. Institut Fédératif de Recherches Multidisciplinaires sur les Peptides (IFRMP 23), Laboratoire de Neuroendocrinologie Cellulaire et Moléculaire, Institut National de la Santé et de la Recherche Médicale Unité 413, CNRS, Université de Rouen, Mont-Saint-Aignan, France;
    Search for more papers by this author
  • Olfa Masmoudi,

    1. Institut Fédératif de Recherches Multidisciplinaires sur les Peptides (IFRMP 23), Laboratoire de Neuroendocrinologie Cellulaire et Moléculaire, Institut National de la Santé et de la Recherche Médicale Unité 413, CNRS, Université de Rouen, Mont-Saint-Aignan, France;
    Search for more papers by this author
  • Pierrick Gandolfo,

    1. Institut Fédératif de Recherches Multidisciplinaires sur les Peptides (IFRMP 23), Laboratoire de Neuroendocrinologie Cellulaire et Moléculaire, Institut National de la Santé et de la Recherche Médicale Unité 413, CNRS, Université de Rouen, Mont-Saint-Aignan, France;
    Search for more papers by this author
  • Christine Patte,

    1. Institut Fédératif de Recherches Multidisciplinaires sur les Peptides (IFRMP 23), Laboratoire de Neuroendocrinologie Cellulaire et Moléculaire, Institut National de la Santé et de la Recherche Médicale Unité 413, CNRS, Université de Rouen, Mont-Saint-Aignan, France;
    Search for more papers by this author
  • Jean Costentin,

    1. IFRMP 23, Laboratoire de Neuropsychopharmacologie, CNRS Unité Mixte de Recherches 6036, Université de Rouen, Rouen, France;
    Search for more papers by this author
  • Jean-Luc Fauchère,

    1. Institut de Recherches SERVIER, Suresnes, France
    Search for more papers by this author
  • Daniel Davoust,

    1. IFRMP 23, Laboratoire de Résonance Magnétique Nucléaire, Institut de Recherche en Chimie Organique Fine, CNRS Unité Mixte de Recherches 6014, Université de Rouen, Mont-Saint-Aignan, France;
    Search for more papers by this author
  • Hubert Vaudry,

    1. Institut Fédératif de Recherches Multidisciplinaires sur les Peptides (IFRMP 23), Laboratoire de Neuroendocrinologie Cellulaire et Moléculaire, Institut National de la Santé et de la Recherche Médicale Unité 413, CNRS, Université de Rouen, Mont-Saint-Aignan, France;
    Search for more papers by this author
  • Marie-Christine Tonon

    1. Institut Fédératif de Recherches Multidisciplinaires sur les Peptides (IFRMP 23), Laboratoire de Neuroendocrinologie Cellulaire et Moléculaire, Institut National de la Santé et de la Recherche Médicale Unité 413, CNRS, Université de Rouen, Mont-Saint-Aignan, France;
    Search for more papers by this author

  • Note: all optically active amino acids are of the l configuration unless otherwise noted.

H. Vaudry, European Institute for Peptide Research, Laboratory of Cellular and Molecular Neuroendocrinology, Institut National de la Santé et de la Recherche Médicale Unité 413, Unité Associée au Centre National de la Recherche Scientifique, University of Rouen, 76821 Mont-Saint-Aignan, France. Fax: + 33 235 14 6946, Tel.: + 33 235 14 6624, E-mail: hubert.vaudry@univ-rouen.fr

Abstract

The octadecaneuropeptide (ODN; QATVGDVNTDRPGLLDLK) and its C-terminal octapeptide (OP; RPGLLDLK), which exert anxiogenic activity, have been previously shown to increase intracellular calcium concentration ([Ca2+]i) in cultured rat astrocytes through activation of a metabotropic receptor positively coupled to phospholipase C. It has also been found that the [d-Leu5]OP analog possesses a weak antagonistic activity. The aim of the present study was to synthesize and characterize cyclic analogs of OP and [d-Leu5]OP. On-resin homodetic backbone cyclization of OP yielded an analog, cyclo1−8 OP, which was three times more potent and 1.4-times more efficacious than OP to increase [Ca2+]i in cultured rat astrocytes. Cyclo1−8 OP also mimicked the effect of both OP and ODN on polyphosphoinositide turnover. Conversely, the cyclo1−8[d-Leu5]OP analog was totally devoid of agonistic activity but suppressed the effect of OP and ODN on [Ca2+]i and phosphoinositide metabolism in astrocytes. The structure of these cyclic analogs has been determined by two-dimensional 1H-NMR and molecular dynamics. Cyclo1−8 OP exhibited a single conformation characterized by a γ turn comprising residues Pro2–Leu4 and a type III β turn encompassing residues Leu5–Lys8. Cyclo1−8[d-Leu5]OP was present as two equimolar conformers resulting from cis/trans isomerization of the Arg–Pro peptide bond. These pharmacological and structural data should prove useful for the rational design of non peptidic ODN analogs.

Abbreviations
ODN

octadecaneuropeptide

OP

octapeptide

DBI

diazepam-binding inhibitor

BZ

benzodiazepine

TTN

triakontatetraneuropeptide

HMP

4-hydroxymethyl-phenoxymethyl-copolystyrene-1%-divinylbenzene resin

HBTU

O-benzotriazol-1-yl-N,N,N′,N′-tetramethyluronium hexafluorophosphate

HOBt

1-hydroxybenzotriazole

DIEA

N,N-diisopropylethylamine

PEG-PS

poly(ethylene glycol)–polystyrene resin

NMP

N-methylpyrrolidin-2-one

DMF

N,N-dimethylformamide

[Ph3P]4Pd

tetrakis(triphenylphosphine)palladium(0)

NMM

N-methylmorpholine

DMEM

Dulbecco's modified Eagle's medium

DSS

sodium 2,2-dimethyl-2-silapentane-5-sulfonate

IP

inositol phosphate

PIP

polyphosphoinositide

Diazepam-binding inhibitor (DBI) is an 86-amino-acid polypeptide that has been originally isolated from rat brain extracts as an endogenous ligand of benzodiazepine (BZ) receptors [1]. Proteolytic cleavage of DBI generates several biologically active peptides including the triakontatetraneuropeptide TTN (DBI17−50) [2] and the octadecaneuropeptide ODN (DBI33−50) [3] which are collectively designated by the term endozepines [4]. Intracerebroventricular injection of endozepines provokes anxiogenic effects [5], induces proconflict behavior [1,6], reverses the anticonflict action of diazepam [1] and inhibits food intake [7].

The mechanism of action of endozepines is not fully understood. It has been initially proposed that these peptides act as inverse agonists of central-type benzodiazepine receptors [6] thus inhibiting the activity of the GABAA-chloride channel complex [8]. Subsequently, endozepines were found to interact with peripheral-type BZ receptors and to stimulate cholesterol transport into mitochondria [9]. More recently, it has been shown that, in rat astrocytes, ODN activates a metabotropic receptor positively coupled to phospholipase C, leading to an increase in cytosolic calcium concentration [10,11]. Structure-activity relationship studies have shown that the C-terminal octapeptide of ODN (OP; ODN11−18) is the minimum sequence retaining full calcium-mobilizing activity [12]. The Ala-scan of OP has revealed that replacement of the Leu6 residue suppresses the activity of the peptide. It has also been found that the [d-Leu5]OP analog exhibits a weak antagonistic activity [12].

On the basis of these observations, we have undertaken the design of selective ODN analogs that would exhibit high affinity for the rat astrocyte endozepine receptor. Backbone cyclization is an efficient approach which has been widely used to stabilize the spatial conformation of peptides without altering the side chain motifs that are often involved in their biological activity [13,14]. In several cases, cyclization has been found to enhance the potency of peptides on their receptors [15–18]. The aim of the present study was to prepare head-to-tail cyclic analogs of OP in order to generate possible agonists and antagonists of ODN. We have determined the secondary structure of the cyclic OP analogs by two-dimensional 1H-NMR and molecular dynamic simulation, and we have investigated the biological activity of these analogs by measuring their ability to modify cytosolic calcium concentrations and polyphosphoinositide turnover in cultured rat astrocytes.

Materials and methods

Materials

All amino-acid residues, preloaded 4-hydroxymethyl-phenoxymethyl-copolystyrene-1%-divinylbenzene resin [Fmoc-Lys(Boc)-HMP], O-benzotriazol-1-yl-N,N,N′,N′-tetramethyluronium hexafluorophosphate (HBTU), 1-hydroxybenzotriazole (HOBt), piperidine and N,N-diisopropylethylamine (DIEA) were purchased from Applied Biosystems (St Quentin en Yvelines, France). Preloaded poly(ethylene glycol)–polystyrene resin [Fmoc-Asp(PEG-PS)-OAl] was obtained from PerSeptive Biosystems (Voisins-le-Bretonneux, France). Trifluoroacetic acid, trichloroacetic acid, phenol, thioanisol, ethanedithiol, N-methylpyrrolidin-2-one (NMP), N,N-dimethylformamide (DMF), tetrakis(triphenylphosphine)palladium(0) ([Ph3P]4Pd), sodium diethyldithiocarbamate, N-methylmorpholine (NMM), U73122 (1-[6-([(17β)-3-methoxyestra-1,3,5-(10)-trien-17-yl]amino)hexyl]1H-pyrrole-2,5-dione), Dulbecco's modified Eagle's medium (DMEM), F12 culture medium, insulin and d(+)-glucose were from Sigma-Aldrich Chimie (St Quentin Fallavier, France). Glutamine, the antibiotic–antimycotic solution and Hepes were from Bioproducts (Gagny, France). Fetal bovine serum was from Biosys (Compiègne, France). BSA (fraction V) was from Roche Molecular Biochemicals (Mannheim, Germany). Indo-1-acetoxymethyl ester and fluo-4-acetoxymethyl ester were from Molecular Probes Europe (Leiden, The Netherlands). Myo-[3H]inositol (100 Ci·mmol−1) was from Amersham International (Les Ulis, France). Sodium 2,2-dimethyl-2-silapentane-5-sulfonate (DSS) and D2O were from Euriso-top (CEA, Saclay, France).

Peptide synthesis

ODN (QATVGDVNTDRPGLLDLK) and OP (RPGLLDLK) were synthesized (0.25 mmol scale for ODN; 0.1 mmol scales for OP) on a Fmoc-Lys(Boc)-HMP resin using an Applied Biosystems model 433A peptide synthesizer using the standard FastMocΩMonPrevPK® procedure as previously described [12]. The synthesis of linear precursors of cyclo1−8 OP and cyclo1−8[d-Leu5]OP (0.25 mmol scale each) was performed on a Pioneer™ PerSeptive Biosystems peptide synthesizer on a Fmoc-Asp(PEG-PS)-OAl resin using similar coupling procedure, extended to a capping cycle of the nonacylated primary amine. After completion of the chain assembly, deprotection of the allyl ester was performed manually by Pd(0) under Ar as previously described [19,20]. The catalyst (3 eq., 0.75 mmol), dissolved in 26 mL Ac2O/CHCl3/NMM mixture (2 : 37 : 1; v/v/v) was transferred to a sealed tube containing the Fmoc-peptidyl(resin)-OAl using an Ar flushed gas-tight syringe and gently agitated for 2.5 h at room temperature. The resin was then washed sequentially with fresh catalyst dissolving mixture, 0.5% DIEA in DMF, 0.5% sodium diethyldithiocarbamate in DMF, 0.5% HOBt in DMF and DCM, and dried in vacuo. The N-terminal Fmoc group was removed by treatment with 20% piperidine in NMP. Prior to each manual step, part of the X-peptidyl(resin)-Y (X = Fmoc or H; Y = OAl or OH) was completely deprotected by trifluoroacetic acid to generate linear peptide sequences for reversed-phase HPLC (RP-HPLC) analysis. On-resin head-to-tail cyclization of the peptide was performed twice by addition of HBTU/DIEA (8 eq., 2 mmol; 1 : 1, mol/mol) in 15 mL of DMF for 2 × 3.5 h with occasional gentle agitation. Cyclization was monitored by the Kaiser's test. Peptides were deprotected and cleaved from the resin as previously described [12].

Peptide purification

All peptides were purified by RP-HPLC on a semipreparative Vydac C18 column (1 × 25 cm; Touzart et Matignon, Courtaboeuf, France) using a linear gradient (10–50% over 40 min) of acetonitrile/trifluoroacetic acid (99.9 : 0.1, v/v) at a flow rate of 5 mL·min−1. Analytical RP-HPLC was performed on a Vydac C18 column (0.45 × 25 cm) using a linear gradient (10–40% over 30 min) of acetonitrile/trifluoroacetic acid at a flow rate of 1 mL·min−1. The purified peptides were characterized by FAB-MS on a conventional EB geometry mass spectrometer JEOL model AX-500 equipped with a DEC data system (JEOL-Europe SA, Croissy-sur-Seine, France) or by MALDI-TOF-MS on a Tofspec E (Micromass, Manchester, UK).

Cell culture

Primary cultures of rat astrocytes were performed as previously described [21]. Briefly, cerebral hemispheres from newborn Wistar rats were collected in DMEM/F12 culture medium (2 : 1, v/v) supplemented with 2 mm glutamine, 1% insulin, 5 mm Hepes, 0.4% d(+)-glucose and 1% of the antibiotic/antimicotic solution. The tissues were disaggregated mechanically using a Pasteur pipette, and filtered through a 100-µm nylon sieve (Poly Labo, Strasbourg, France). Dissociated cells were resuspended in culture medium supplemented with 10% fetal bovine serum and seeded on coverslips in 35-mm dishes (Dutscher, Brumath, France) at a density of 106 cells per dish. The cells were incubated at 37 °C in a moist atmosphere (5% CO2), and the medium was changed twice a week.

Confocal imaging

Five- to seven-day-old cells were loaded with 3 µm fluo-4-acetoxymethyl ester diluted in culture medium, at 37 °C for 30 min. Thereafter, the calcium-dye-probe was washed off and exchanged with 2 mL of fresh medium. The fluorescence emission of the fluo-4-loaded cells, induced by excitation at 488 nm (laser Ar/Kr) was recorded with a 500-nm long-pass filter on a Noran OZ confocal microscope (Noran Instruments, Middleton, WI, USA). Images were recorded as a time series (512 × 480 pixels at one image per 532 ms) and data processing was carried out using the intervision software (Noran Instruments). Cyclo1−8 OP (10−8 M) was ejected for 2 s in the vicinity of the cells by a pressure ejection system.

Measurement of cytosolic Ca2+ concentration

Five- to seven-day-old cells were loaded with 5 µm indo-1-acetoxymethyl ester diluted in culture medium, at 37 °C for 45 min. The cells were washed twice with 2 mL of fresh medium. The [Ca2+]i was monitored by a dual-emission microfluorimeter system constructed from a Nikon Diaphot inverted microscope, as previously described [12]. The fluorescence emission of indo-1, induced by excitation at 355 nm, was recorded at two wavelengths (405 nm and 480 nm) by separate photometers (Nikon, Champigny sur Marne, France). The 405/480 ratio was determined using an analogic divider (constructed by B. Dufy, University of Bordeaux, France) after conversion of single photon currents to voltage signals. All three signals (405 nm, 480 nm and the 405/480 ratio) were continuously recorded with the jad-fluo 1.2 software (Notocord Systems, Croissy-sur-Seine, France). The [Ca2+]i-values were calculated as previously described [22]. All secretagogues were ejected for 2 s in the vicinity of individual cells by a pressure ejection system. The indicated doses of peptides correspond to the concentration contained in the ejection pipette.

Measurement of polyphosphoinositide metabolism

Twelve- to 14-day-old cells were incubated with 10 µCi·mL−1myo-[3H]inositol (100 Ci·mmol−1) at 37 °C in glucose- and fetal bovine serum-free culture medium, in the absence or presence of ODN-related peptides. The incubation was stopped by removing the medium and adding 1 mL of ice-cold 10% trichloroacetic acid. The cells were homogenized and centrifuged (13 000 g; 15 min; 4 °C). The supernatant was washed three times with water-saturated diethylether, neutralized with 10 µL of 1 m NaHCO3. Free inositol and total tritiated inositol phosphates (IPs) were separated by anion exchange chromatography (AG1-X8 resin; 100–200 mesh; formate form; Bio-Rad Laboratories, Richmond, CA, USA) using distilled water and 0.8 m ammonium formate in 0.1 m formic acid, respectively, and the radioactivity contained in each fraction was counted in a β-counter (LKB 1217 Rack Beta, EG and G Wallace, Evry, France). Polyphosphoinositides (PIPs) were extracted from the pellet with 500 µL CHCl3/MeOH (2 : 1, v/v) and counted in a β-counter. The remaining pellets were used for measurement of protein concentration by the Lowry's method.

NMR spectroscopy

NMR experiments were carried out using an AVANCE DMX 600 MHz spectrometer (Bruker S.A., Wissenbourg, France) equipped with a SGI indigo 2 computer. One- and two-dimensional NMR spectra were obtained at temperatures of 275, 280, 285, 293 and 298 K. NMR samples were prepared by dissolving the peptides in 550 µL of H2O (10% D2O) or D2O. One- and two-dimensional spectra were recorded with carrier frequency in the middle of the spectrum coinciding with the water resonance which was suppressed either by a presaturation using continuous irradiation during relaxation delay or by using the gradient pulse watergate[23]. One- and two-dimensional 1H NMR spectra were calibrated using DSS as an external reference. Spin systems identification and sequential assignment were achieved by TOCSY, COSY and NOESY experiments. The TOCSY spectra were recorded with spin-lock time of 80 ms by using Mlev17 sequence for the isotropic mixing. Four mixing times (80, 150, 200 and 300 ms) were used for NOESY spectra in order to identify diffusion effects. All two-dimensional NMR experiments were acquired with a total of 2048 complex data points in F2, and 512 experiments in F1. Prior to Fourier transform, data matrices were zero filled in F1 dimension and a phase shifted sine-bell filter function was applied. Processing of NMR data was performed on a SGI Indigo 2 workstation using the manufacturer's programs xwinnmr 2.1 and aurelia 2.0.

Structure calculations

The volumes of the cross-peaks of the NOESY spectra acquired with a mixing time of 200 ms were integrated using the aurelia 2.0 software from Bruker. Interproton distances were calculated using the isolated spin pair approximation and setting the average methylene cross peak volume at 0.18 nm. A range of 20% of the distance values was used for defining the upper and lower bonds of the constraints. For the methyl protons and protons that could not be stereospecifically assigned, pseudoatoms were generated during primary structure construction. For these protons the lower bond of the distance constraints was set to the sum of the Van der Waals radii. Backbone dihedral restraint residues were deduced from 3JNH-Hα coupling constant by using the empirically Karplus-type relations [24]. For coupling constant where more than a single Φ-value is possible, additional secondary information from NOE data was used to reduce the number of solutions. A range of ± 30° was used for defining the upper and lower angle of the constraints. The structures were generated from the experimental data with a standard dynamical simulated annealing protocol using the x-plor 3.1 program. A force field adapted for NMR structure determination (file parallhdg.pro and topallhdg.pro) was used, and an initial structure was built by randomly generating Φ and Ψ angles. In the calculations starting from random structure, we used higher force constants for the bond lengths, bond angles, improper and planarity angles. After a short energy minimization, the first stage started with scaling of the weights of the NOE, constrained improper dihedrals, and nonbonded terms initially small values to more realistic ones using in total 30 000 time steps each of 0.002 ps at 1000 K. The second stage performed a slow cooling of the system to 100 K in 50 stages each with 5000 time steps of 0.002 ps. The final stage involved 2000 cycles of constrained Powell energy minimization. Structures with the lowest distance and dihedral constraint energy were selected and refined by restrained Powell energy minimization using a more realistic force field based on the charmm 19 program. The Van der Waals energy was calculated with switched Lennard–Jones potential, and the electric energy was calculated with a shifted Coulomb potential with a dielectric constant ε = 80.7. Structures were displayed using the sybyl software package (Tripos Associates, Inc., St Louis, MO, USA). The whole structural analysis described in the text was performed using structures calculated without explicit H-bond restraints. The final structures were examined to obtain pairwise root mean square differences (rmsd) over the backbone heavy atoms (N, Cα and C). Structure calculations without dihedral constraints were also carried out. This did not produce any structural modification, and only a slight change of the rmsd values was observed for the backbone atoms.

Calculations and statistics

Data are expressed as mean ± SEM. Student's t-test was used to determine statistical differences between control and experimental values within the same set of experiments [25].

Results

Synthesis of cyclic peptides

The two cyclic peptides cyclo1−8 OP and cyclo1−8[d-Leu5]OP were synthesized directly on solid support and the head-to-tail cyclization was achieved using orthogonal allyl protection for the α-carboxylic function of aspartic acid (Fig. 1). After the last coupling cycle and removal of the allyl protecting group with Pd(0) and N-methylmorpholine, an aliquot of Fmoc-peptidyl(PEG-PS)-OH was treated with the cleavage mixture (reagent K) [26] and submitted to RP-HPLC analysis. The chromatograms corresponding to the different steps of the synthesis of cyclo1−8 OP are shown in Fig. 2. The HPLC profiles revealed the existence of two peaks which eluted at 17.0 and 26.2 min (Fig. 2A). The major one (retention time 26.2 min), which was assumed to be the Fmoc-peptidyl-OH derivative, completely vanished after treatment with piperidine (Fig. 2B). After two periods of 3.5 h of lactamization with intermediate reactivation, followed by side-chain deprotection and cleavage of the peptide from the resin, RP-HPLC analysis revealed the occurrence of a new major peak eluting at 18.6 min and the disappearance of the linear precursor form (Fig. 2C). Similar chromatograms were obtained during the synthesis of cyclo1−8[d-Leu5]OP (data not shown).

Figure 1.

Scheme for the stepwise approach in the synthesis of the cyclic peptides. (A) (1) piperidine 20% NMP, (2) Fmoc-AA-OH/HBTU/HOBt/DIEA, (3) Ac2O/CHCl3/NMM. (B) (1) [Ph3P]4Pd/AcOH/CHCl3/NMM, (2) AcOH/CHCl3/NMM, (3) DIEA 0.5% DMF, (4) sodium diethyldithiocarbamate 0.5% DMF, (5) HOBt 0.5% DMF, (6) DCM; (C) (1) piperidine 20% NMP, (2) HBTU/DIEA; (D) trifluoroacetic acid/phenol/thioanisol/ethanedithiol/H2O. Xaa = Leu (cyclo1−8 OP) or d-Leu (cyclo1−8[d-Leu5]OP).

Figure 2.

RP-HPLC monitoring of N- and C-terminal deprotections and on-resin cyclization of the linear precursor of cyclo1−8 OP. Aliquots of the reaction media after allyl ester deprotection (A), Fmoc removal (B) and cyclization of the linear precursor (C) were cleaved and injected onto a Vydac C18 analytical column. The dotted lines represent the profile of the elution gradient (% acetonitrile).

Effect of ODN analogs on [Ca2+]i

The spatial-temporal [Ca2+]i changes in rat astrocytes were visualized by means of a confocal laser scanning microscope. Ejection of 10−8M cyclo1−8 OP in the vicinity of cultured cells provoked a wave of calcium in the cytoplasm of astrocytes (Fig. 3A) while ejection of culture medium alone had no effect (Fig. 3B). Comparison of the amplitude of the response with that of ODN and OP revealed that, at a concentration of 10−8 m, cyclo1−8 OP was more efficient than ODN and OP (Fig. 3C). Administration of graded concentrations of OP and cyclo1−8 OP induced a bell-shaped [Ca2+]i response (Fig. 4). For concentrations ranging from 10−11 to 10−8 m, OP and cyclo1−8 OP provoked a dose-related increase in [Ca2+]i with maximum responses at concentrations of 10−8 m and 3.16 × 10−9 m, respectively. At higher concentrations (10−7−10−5 m), the effect of OP and cyclo1−8 OP gradually declined. The efficacy of cyclo1−8 OP in raising [Ca2+]i was 1.4-fold higher than that of OP. Repeated pulses of cyclo1−8 OP (3.16 × 10−9 m) resulted in sequential increases in [Ca2+]i with gradual attenuation of the response (Fig. 4).

Figure 3.

Effect of cyclo1−8 OP on [Ca2+]i in cultured rat astrocytes. Time series of pseudocolor images illustrating [Ca2+]i changes in astrocytes loaded with fluo-4-acetoxymethylester. (A) Intracellular Ca2+-wave following ejection of cyclo1−8 OP (10−8 m). (B) Intracellular calcium in cells from the same field after ejection of medium alone. Sampling rate, 1 image per 532 ms. The pseudocolor scale indicates the corresponding [Ca2+]i changes expressed in arbitrary units. (C) Effects of ODN, OP and cyclo1−8 OP (10−8 m each) on the amplitude of the calcium response measured by microfluorimetry. Each value represents the mean amplitude (± SEM) of the calcium response calculated from at least 10 different dishes from five independent cultures. The number of cells studied is indicated in parentheses. NS not statistically significant, *P < 0.05, **P < 0.01.

Figure 4.

Effect of graded concentrations of OP and cyclo1−8 OP on [Ca2+]i in cultured rat astrocytes. A 2-s pulse of OP (10−11−10−5 m) and cyclo1−8 OP (10−11−10−6 m) was administred in the vicinity of the cells and [Ca2+]i was measured by microfluorimetry. Each value represents the mean amplitude (± SEM) of the calcium response induced by OP (▪) and cyclo1−8 OP (●) calculated from at least four different dishes from two independent cultures. The number of cells studied is indicated in parentheses. The inset shows a typical profile of the calcium response to 2-s pulses of 3.16 × 10−9 m cyclo1−8 OP (arrows).

Administration of cyclo1−8[d-Leu5]OP, for concentrations ranging from 10−10 to 10−5 m, did not affect [Ca2+]i in rat astrocytes (Fig. 5). A 10-min preincubation of astrocytes with graded concentrations of cyclo1−8[d-Leu5]OP (10−10−10−6 m) provoked a dose-dependent inhibition of the ODN-induced [Ca2+]i increase with a pIC50 value of 7.17 ± 0.29 (Fig. 6). At a concentration of 10−6 m, cyclo1−8[d-Leu5]OP totally abolished the [Ca2+]i response to ODN. Cyclo1−8[d-Leu5]OP also suppressed the [Ca2+]i increase evoked by 10−8 m OP (Fig. 7A). On the other hand, cyclo1−8[d-Leu5]OP, at concentrations of 10−6 and 10−5 m, significantly reduced (P < 0.001) but did not abolish the [Ca2+]i increase induced by cyclo1−8 OP (Fig. 7B).

Figure 5.

Effect of graded concentrations of cyclo1−8[d-Leu5]OP on [Ca2+]i in cultured rat astrocytes. A 2-s pulse (arrow) of cyclo1−8[d-Leu5]OP (10−10−10−5 m) was administered in the vicinity of different cells. The number of cells studied is indicated in parentheses.

Figure 6.

Effect of graded concentrations of cyclo1−8[d-Leu5]OP on the ODN-evoked [Ca2+]i increase in cultured rat astrocytes. The cells were incubated for 15 min in the absence (○) or presence (●) of cyclo1−8[d-Leu5]OP (10−10−10−6 m) before administration of a 2-s pulse of ODN (10−8 m). Each value represents the mean amplitude (± SEM) of the ODN-evoked response calculated from at least four different dishes from two independent cultures. The number of cells studied is indicated in parentheses.

Figure 7.

Effect of cyclo1−8[d-Leu5]OP on the OP- and cyclo1−8 OP-evoked [Ca2+]i increase in cultured rat astrocytes. The cells were incubated for 15 min in the absence or presence of cyclo1−8[d-Leu5]OP (10−6 or 10−5M) before administration of a 2-s pulse (arrow) of OP (A) or cyclo1−8 OP (B) (10−8M, each). Each value represents the mean amplitude (± SΕΜ) of the ODN-evoked response calculated from at least 4 different dishes from two independent cultures. The number of cell studied is indicated in parentheses.

Effect of ODN analogs on polyphosphoinositide metabolism

Exposure of cultured astrocytes to ODN, OP, or cyclo1−8 OP (10−8 m each) caused a significant increase in the formation of [3H]IPs (Fig. 8A) and a concomitant decrease in the levels of [3H]PIPs (Fig. 8B). In contrast, cyclo1−8[d-Leu5]OP, even at a high concentration (10−6 m), did not affect the basal level of [3H]IPs and [3H]PIPs (Fig. 8A,B). In the presence of 10−6 m cyclo1−8[d-Leu5]OP, the effect of ODN, OP and cyclo1−8 OP (10−8 m each) on [3H]IPs and [3H]PIPs was totally abolished (Fig. 8C,D). Similarly, incubation of astrocytes for 10 min with the phospholipase C inhibitor U73122 (10−5 m) abrogated the effects of ODN, OP and cyclo1−8 OP (10−8 m each) on [3H]IPs and [3H]PIPs (Fig. 8E,F).

Figure 8.

Effect of cyclo1−8[d-Leu5]OP and U73122 on inositol phosphate formation and polyphosphoinositide breakdown induced by ODN, OP and cyclo1−8 OP in cultured rat astrocytes. The cells were incubated for 5 min with ODN, OP or cyclo1−8 OP (10−8M each) in the absence (A, B) or presence (C, D) of cyclo1−8[d-Leu5]OP (10−6M). In another set of experiments (E, F), the cells were preincubated with U73122 (10−5M) for 10 min and then incubated for 5 min with ODN, OP or cyclo1−8 OP (10−8M each). Each bar represents the mean (± SEM) value from at least three independent experiments. The number of determinations is indicated in parentheses. **P < 0.01, ***P < 0.001.

NMR solution structure of cyclo1−8 OP

Amino-acid spin systems of cyclo1−8 OP were readily identified from two-dimensional COSY and two-dimensional TOCSY spectra starting from amide protons in the region of 9.5–7 p.p.m. and were confirmed by inspection of cross-peaks in the high field region corresponding to side-chain through-bond connectivities. NOESY data as those presented in Fig. 9 were then used to determine the sequential assignments and the chemical shifts reported in Table 1.

Figure 9.

Region of 600- MHz NOESY spectrum of cyclo1−8 OP obtained in H2O at 280 K. The spectrum was recorded with a mixing time of 200 ms.

Table 1. 1H-NMR assignments, 3JHN-Hα coupling constant and backbone angles Φ, Ψ for cyclo1−8 OP at 280 K. Chemical shifts are relative to DSS.
ResidueChemical shift (p.p.m.)3JHN-Hα
(Hz)
Backbone angles (°)
HNΦΨ
Arg17.564.741.911.64/1.503.227.267.9−164166
Pro2 4.272.31/1.882.14/1.953.82/3.57  −90128
Gly38.844.20/3.70    6.386−54
Leu47.874.681.68/1.581.580.88 8.9−114105
Leu58.794.141.71/1.601.710.94/0.89 3.47862
Asp68.524.493.10/2.88   −44−27
Leu77.854.241.84/1.671.670.90 7.9−72−30
Lys88.073.991.881.351.652.98 NH3 : 7.596.1−1293

Amide protons which were relatively slow to exchange with solvent were identified by dissolving cyclo1−8 OP in deuterated solvent D2O. Then, one-dimensional 1H-NMR spectra were recorded at regular interval (Fig. 10). Due to fast H/D exchange, the amide NH of residues Gly3, Leu5 and Asp6 disappeared rapidly (within one hour or less) compared to the other amide protons.

Figure 10.

Region of 600-MHz amide proton NMR spectra. (A) Spectrum of cyclo1−8 OP in H2O. (B–D) Spectra of cyclo1−8 OP in D2O recorded after 10, 75 and 255 min, respectively.

The 3JNH-Hα coupling constants for the amide protons of cyclo1−8 OP were measured from the 1D 1H-NMR spectrum (Table 1). The 3JNH-Hα coupling constants of the three leucine residues (Leu4, Leu5 and Leu7) differed from the averaged value usually observed for small peptides (≈ 7 Hz). The examination of short and medium range NOEs (Fig. 11), in particular NOE observed between Hα-Pro2 and Hα-Leu4, Hα-Leu5 and NH-Leu7, Hα-Leu6 and NH-Lys8, in combination with the coupling constants and slow H/D exchange of NH-Leu4, NH-Lys8, suggested a first turn centered on the Gly3 and Leu4 residues, and a second one in the region encompassing the Leu5, Asp6 and Leu7 residues. In order to better localize the different turns and to identify each type of turn, molecular modeling under experimental NMR restraints was performed.

Figure 11.

Summary of NOEs observed in a 600-MHz NOESY spectrum of cyclo1−8 OP obtained in H2O at 280 K. The sequence is displayed with the one-letter code. The heights of the bars indicate the intensities of the NOEs. Residues with exchanging times of amide protons larger than 1 h are indicated by black squares above the sequence.

NOEs data and coupling constants detected for cyclo1−8 OP were used to drive a set of 47 distance and six dihedral angle restraints. These restraints were used to generate a set of 30 structures by simulated annealing as described in Materials and methods. All the calculated structures fitted the experimental data quite well and converged with high precision. Analysis of the Φ and Ψ angles showed that all the residues were in the energetically favorable region of the Ramachandran diagramme. Ten structures with the lowest distance and dihedral constraint energies were selected, providing a well-defined shape of the backbone foldings of cyclo1−8 OP (Fig. 12). The rmsd value calculated relatively to the mean structures over all backbone atoms of the cycle was 0.009 nm.

Figure 12.

Lowest energy conformer of cyclo1−8 OP from simulated annealing. The dotted lines indicate hydrogen bonds consistent with NMR data.

NMR solution structure of cyclo1−8[d-Leu5]OP

The 1H-NMR spectrum of cyclo1−8[d-Leu5]OP in H2O at 280 K was assigned by using the same strategy as for cyclo1−8 OP. In a first step, two-dimensional spectra COSY and TOCSY were used for the identification of the amino-acid spin system. In these spectra the glycine residue was easily identified from its characteristic remote peaks (Hα1, Hα2) at the amide proton frequency. Two glycine-type remote peaks were clearly observed instead of one expected, suggesting the existence of two spectroscopically distinct molecular conformers. Analysis of other spectra confirmed the occurrence in the solution of two distinct species corresponding to the same primary structure. In a second step, sequential assignment was simultaneously conducted for the two conformers by using dαN, dβN, and dNN in the NOESY experiments (Fig. 13A). At this stage, the origin of the structural heterogeneity was identified as a peptidyl-prolyl cis-trans isomerism of the peptide bond Arg1-Pro2. Some of the spectral data are illustrated in Fig. 13B, that show a region of the NOESY spectrum containing the sequential dαα(Arg1, Pro2) and dαδδ′(Arg1, Pro2) connectivities, characteristic of the two isomeric forms. The proton chemical shift values of the cis and trans conformers at 280 K are given in Table 2. Amide protons that exchange with solvent were identified by dissolving cyclo1−8[d-Leu5]OP in D2O. Due to fast H/D exchange all amide protons of cyclo1−8[d-Leu5]OP disappeared rapidly (within one hour or less) (Fig. 14).

Figure 13.

Regions of a 600-MHz NOESY spectra of cyclo1−8[d-Leu5]OP recorded with a mixing time of 200 ms at 280 K. (A) NH-αCH cross peaks in H2O: cis isomer, bold letters; trans isomer, italic letters. (B) αCH-αCH region in D2O solution.

Table 2. 1H-NMR assignments, 3JHN-Hα coupling constant and backbone angles Φ, Ψ for cis and trans conformers of cyclo1−8[d-Leu5]OP at 280 K. Chemical shift is relative to DSS.
ResidueChemical shift (p.p.m.)3JHN–Hα
(Hz)
Backbone angles (°)
HNΦΨ
  1. aCis conformer. bTrans conformer.

Arg1a8.884.671.831.833.45/3.217.27−46135
Arg1b7.64.791.81.623.257.3−12096
Pro2a 4.492.39/1.952.14/2.033.68  −56140
Pro2b 4.462.38/2.251.95/1.773.56  −7268
Gly3a9.114.11/3.60    11.390153
Gly3b8.894.09/3.88    11.7−170−37
Leu4a8.244.471.721.720.96/0.89 7.3−129179
Leu4b8.194.571.71.570.92 8.7−118−51
Leu5a8.664.171.641.640.94 4.4104−117
Leu5b8.274.351.83/1.641.640.94/0.89 12475
Asp6a9.34.83.04/2.83   8.3−9833
Asp6b8.024.852.92   −132−163
Leu7a7.554.311.75/1.511.751.03/0.93 7.1−140−154
Leu7b8.494.221.80/1.651.80.96/0.91 4.943−34
Lys8a8.214.141.741.561.743.06 NH3 : 7.644.4−142137
Lys8b8.364.181.891.50/1.421.73.02 NH3 : 7.656.8−17854
Figure 14.

Summary of NOEs observed in a 600-MHz NOESY spectrum of cyclo1−8[d-Leu5]OP obtained in H2O at 280 K. (A) Cis isomer. (B) Trans isomer. The sequence is displayed with the one-letter code. The heights of the bars indicate the intensities of the NOEs.

In the NOESY spectrum, the cross-peaks of the cis and trans conformers could be clearly separated and analyzed for structure determination. In addition, no exchange of cross-peak could be observed between the resonances of the two conformers, indicating that the interconversion is very slow on the NMR time-scale. NOEs detected for trans and the cis isomers in combination with the coupling constant, measured from the one-dimensional spectrum, supported the existence of a well-defined structure and were used to drive a set of 27 distance and four dihedral angle constraints for the trans conformer and 31 distance and six dihedral angle constraints for the cis conformer. These restraints were used to generate a set of 30 structures for each conformer of cyclo1−8[d-Leu5]OP using the same protocol as for cyclo1−8 OP. Ten final best structures obtained for the trans and cis conformers were selected providing a well-defined shape of the backbone foldings (Fig. 15). The rmsd values calculated relatively to the mean structures over all backbone atoms of the cycle was 0.009 nm for the trans conformer and 0.007 nm for the cis conformer.

Figure 15.

Lowest energy conformers of cyclo1−8[d-Leu5]OP from simulated annealing. (A) Cis conformer. (B) Trans conformer.

Discussion

It has been previously reported that OP is the minimum active sequence of ODN that possesses full agonistic activity on [Ca2+]i in cultured rat astrocytes [12]. It has also been shown that the [d-Leu5]OP analog behaves as a weak antagonist in the same in vitro model [12]. The present study demonstrates that the cyclic analogs cyclo1−8 OP and cyclo1−8[d-Leu5]OP exhibit, respectively, potent agonistic and antagonistic activities both on calcium mobilization and on polyphosphoinositide metabolism in rat astroglial cells. The secondary structure of these two ODN analogs has been determined by combining two-dimensional 1H-NMR and molecular dynamics.

Introduction of conformational restraint through cyclization has become a standard strategy in medicinal chemistry for increasing the receptor affinity and selectivity of peptide ligands [13,14,27]. The Ala-scan of OP has revealed that the side chain of each residue is required for the full activity of the peptide [12]. These data led us to synthesize cyclic analogs of OP and to use the N- and C-terminus for cyclization in order to keep the side chains unmodified. We have taken advantage of the presence of an aspartic acid residue in the core sequence of OP and its analog to carry out head-to-tail cyclization on peptides bounded to the resin according to the strategy of Trzeciak & Bannwarth [19], rather than cyclization of a protected peptide in solution. With this procedure, aspartic acid was attached to the solid support via the β-carboxyl group whereas the α-carboxylic group was protected as allyl ester. Monitoring of peptide deprotection and lactamization processes by analytical RP-HPLC revealed that the Nα-Fmoc group was not totally stable in reductive media. This observation was at variance with the data reported by Carpino & Han [28], who found that Fmoc-derivatives are not sensitive to catalytic hydrogenolysis. Such a phenomenon, which has been already reported by others [29,30], can be ascribed to the occurrence of traces of dimethylamine in DMF or to the resonance stabilization properties of the fluorenyl system. HPLC analysis revealed that the peptides were entirely end-to-end cyclized indicating that the experimental conditions favored intramolecular rather than intermolecular cyclization. The low reticulation grade of the solid support and the PEG spacer used in this study produced a pseudodilution phenomenon [31] and complete solvatation of the reactive sites [32,33] which contributed to the efficacy of the intramolecular cyclization.

Using a video imaging confocal microscopy technique, we found that cyclo1−8 OP induced calcium waves in cultured rat astrocytes, suggesting that cyclization did not impair the agonistic activity of OP. Quantitative measurement of [Ca2+]i by microfluorimetry showed that cyclo1−8 OP induced a bell-shaped increase in [Ca2+]i which was reminiscent of those previously observed with ODN and OP [12]. However, we found that the potency of cyclo1−8 OP was increased by a factor of three and its efficacy by a factor of 1.4 compared to its linear counterpart in eliciting [Ca2+]i rise in rat astrocytes. As the dose–response curves obtained with ODN and OP are strictly superimposable, these data indicate that cyclo1−8 OP is a potent agonist of both ODN and OP.

Cyclo1−8[d-Leu5]OP did not modify [Ca2+]i in cultured rat astrocytes but totally blocked both the OP- and the ODN-evoked [Ca2+]i increase. The fact that the linear counterpart, [d-Leu5]OP, only partially reduced the stimulatory effect of ODN on [Ca2+]i[12] indicates that cyclization of the peptide strongly enhances its antagonistic activity. Concurrently, cyclo1−8[d-Leu5]OP did not completely abolish the cyclo1−8 OP-induced [Ca2+]i increase, confirming that cyclo1−8 OP is a more potent agonist than the linear peptide.

We have previously demonstrated that the effect of ODN on [Ca2+]i is mediated through activation of a membrane receptor coupled to a phopholipase C [10]. The present study shows that cyclo1−8 OP stimulates inositol phosphate formation and polyphosphoinositide breakdown in cultured rat astrocytes and that cyclo1−8[d-Leu5]OP totally abolishes the effects of ODN, OP and cyclo1−8 OP on inositol phosphate turnover. In addition, the effects of cyclo1−8 OP on polyphosphoinositide metabolism were totally blocked by U73122, a specific inhibitor of phospholipase C activity [34]. These results confirm that cyclo1−8 OP and cyclo1−8[d-Leu5]OP act, respectively, as an agonist and an antagonist of the metabotropic receptor which mediates the ODN-induced [Ca2+]i increase in rat astrocytes.

It has been previously shown that a vacant C-terminal carboxylic group is essential for ODN and OP to displace benzodiazepines from their binding site [3,35]. In contrast, the present data indicate that free N- and C-terminal extremities are not required for the agonistic and antagonistic activities of the two OP analogs on [Ca2+]i, indicating that cyclo1−8 OP and cyclo1−8[d-Leu5]OP are specific ligands for the metabotropic receptor that should not recognize central-type benzodiazepine receptors. This observation is of particular interest as the anxiogenic effects of ODN are mediated through central-type benzodiazepine receptors [5] while the anorexic action of ODN is likely to be mediated through the metabotropic receptor [7].

We have next investigated the conformation of cyclo1−8 OP and cyclo1−8[d-Leu5]OP in solution by 1H-NMR and restrained molecular dynamics. The structure calculations were obtained without application of hydrogen bonds that could be derived from amide proton exchange rate, using only the angle and distance restraints deduced from 3JNH-Nα coupling constant and NOE cross peak, respectively. Analysis of the (Φ, Ψ) angles, and the potential hydrogen bonding in calculated structures of cyclo1−8 OP revealed that the sequence consists of a γ turn centered on Pro2-Gly3-Leu4 and a type III β turn centered on Leu5-Asp6-Leu7-Lys8 as indicated by the following structural parameters: (a) the (Φ, Ψ) dihedral angles of residue Gly3 (86, −54); (b) the formation of intramolecular hydrogen bonds (i, i + 2) between the C = O of Pro2 and the NH of Leu4 to form a ring associated with a standard γ turn, and (c) a hydrogen bond (i, i + 3) between the C = O of Leu5 and NH of Lys8 associated with the average (Φ, Ψ) dihedral angles of residues Asp6 (−44, −27) and Leu7 (−72, −30) which correspond to almost canonical values for a type III β turn [36,37]. Furthermore, the analysis of the calculated structures revealed a short distances between NH of Leu7 and the side chain carboxyl group of the Asp6 residue (≈ 0.28 nm), and NH of Leu7 residue and C=O of Leu5 (≈ 0.24 nm). The proximity of these two C=O acceptors for one NH donor could explain the slow exchange H/D observed for NH of the Leu7 residue.

Analysis of NMR data of cyclo1−8[d-Leu5]OP, showed that two equimolar conformer populations corresponding to the same primary structure are present in solution. NOE cross-peaks dαα(Arg1, Pro2) and dαδδ′(Arg1, Pro2) clearly indicated a peptidyl-prolyl cis-trans isomerism of the peptide bond Arg1-Pro2. This observation suggests that the presence of a d-Leu residue in position 5 destabilizes the β, γ turn conformer and gives rise to a population containing a cis X-Pro bond. The occurrence of cis bonds has been reported in a number of small peptides, particularly in peptides containing amide bonds involving a proline residue [38,39]. Analysis of hydrogen bonds and dihedral angles (Φ, Ψ) for the average structures generated from members of cis and trans family of cyclo1−8[d-Leu5]OP showed that the residues of the cis and trans conformers of cyclo1−8[d-Leu5]OP do not fall into any classical turn. Cis and trans conformers generally cannot be isolated as stable compounds and therefore cannot be tested separately for biological activity. The cyclic tetrapeptide [Sar1]-tentoxin appears to be the only molecule known so far that has been isolated as two different conformers [40].

The solution structure of cyclo1−8 OP differs from that determined for either its cis and trans diastereoisomer cyclo1−8[d-Leu5]OP and also from that of the linear ODN peptide (data not shown) in that the structure of cyclo1−8 OP consists of a γ turn centered on Pro2-Gly3-Leu4 residues and a type III β turn involving Leu5-Asp6-Leu7-Lys8 residues. γ and β turns are typical elements of peptide and protein conformation that are involved in peptide folding and thus are intimately associated with biological activity [41]. For instance, a β turn is found in the core sequence of somatostatin [42], neurotensin [43] and gonadotropin-releasing hormone [44] while a γ turn is present in bradykinin [45]. The fact that cyclo1−8 OP exhibits a defined solution structure while ODN has a very low tendency to form secondary structures [46] suggests that the γ turn and/or the type III β turn may play a pivotal role in enhancing the calcium-mobilizing activity of the peptide on rat astrocytes.

In conclusion, we have shown that head-to-tail cyclization of the ODN agonist OP and the weak antagonist [d-Leu5]OP enhances their biological activity on rat astrocytes leading to a super-agonist and to the first efficacious ODN antagonist described so far. Furthermore, cyclo1−8 OP adopts a single conformation in solution encompassing a γ turn and a type III β turn whereas cyclo1−8[d-Leu5]OP is present as two equimolar conformers resulting from cis/trans isomerism. Characterization of the secondary structure of these two ligands of ODN receptors should provide useful information for the design of potent, stable and selective agonists and antagonists of these receptors. As ODN exerts anxiogenic effects and induces proconflict behavior [5,6], specific antagonists may have a therapeutic value for treatment of psychiatric disorders. In addition, the development of specific agonists that would selectively mimic the anorexigenic effect of ODN [7] might be of interest for the treatment of obesity and feeding disorders.

Acknowledgements

The authors wish to thank Mrs Catherine Buquet, Huguette Lemonnier and Gérard Cauchois for skillful technical assistance. This study was supported by INSERM (U413), the LARC-Neuroscience network and the Conseil Régional de Haute-Normandie. J. L. was the recipient of a scholarship from Oril-Servier Laboratories and the Conseil Régional de Haute-Normandie. O. M. was the recipient of a scholarship from the French Ministry of Research.

Ancillary