2.1 Isolation and characteristics

The seeds of Annona muricata were successively extracted with cyclohexane and MeOH and the MeOH extract after concentration was dissolved in EtOAc. The EtOAc soluble fraction was purified by exclusion chromatography, silica-gel column chromatography and C₁₈ reversed-phase-HPLC to yield a new cyclopeptide termed annomuricatin C (1). Positive reaction with chlore/o-tolidine reagent suggested it was a peptide and the absence of coloration of its TLC spot with ninhydrin that it was cyclic. Analysis of the total acidic hydrolysate, after derivatization, indicated the presence of Ala (1), Gly (1), Phe (1), Pro (1), Ser (1) and Val (1). The amino acids were converted into the n-propyl esters of their N-trifluoroacetyl derivatives, analysed by gas chromatography on a chiral capillary column ; their retention times compared with those of standards indicated that all the chiral amino acids were l.

2.2 Mass spectral analysis

The molecular weight 558 for annomuricatin C (1) was deduced from the positive ESI-QTOF spectrum, which displayed the [M+N_a]⁺ adduct ion at m/z 581 and the protonated molecular [M+H]⁺ ion at m/z 559. According to the amino acid composition, the molecular formula C₂₇H₃₈N₆O₇ was assigned to 1.

Cyclopeptides are not easily sequenced even by mass spectrometry. The reason is that multiple and indiscriminate ring-opening reactions occur during the CID of cyclic peptides, resulting in the superimposition of random fragment ions and making the interpretation difficult [7,8]. However we have shown that when a proline is present in the sequence, a specific fragmentation occurs at the peptidyl–prolyl (Xaa–Pro) amide bond, leading to a linear peptide C-ended by an acylium ion (b_n), which undergoes further fragmentation, generating series of acylium ions from which the sequence could be deduced [2,9]. This specific fragmentation is explained by the more basic nature of the proline nitrogen relative to the other peptide-bond nitrogen atoms. In this way, the more basic site at proline level strongly directs the protonation, making the fragmentation less complex.

The protonated molecular ion [M+H]⁺ of 1 at m/z 559 was subjected to CID experiment (Fig. 1). The ring opening began at the Ala-Pro amide bond level and a series of adjacent acylium ions (b_n) at m/z 488, 401, 302 and 155 was generated, from which the sequence could be deduced: amino acid residues were lost sequentially from the C-terminus to the N-terminus and for annomuricatin C (1) was observed the successive losses of Ala, Ser, Val, and Phe, yielding the N-terminal dipeptide Pro-Gly (Fig. 2). A second significant series of ions was observed at m/z 531, 460, 373, 274, 127 and 70, which were assigned to adjacent a_n ions related to the above b_n ion series. At m/z 541 was observed an abundant ion originating from the loss of a molecule of water from the protonated molecular ion [M+H]⁺. It underwent fragmentation, yielding a third series of abundant ions here termed h_n (Fig. 1B), corresponding to the successive loss of Val, Phe, Gly, Pro and Ala, giving a final ion at m/z 70, which corresponded to a dehydrated Ser residue. This latter series of adjacent fragments confirmed the above sequence and was explained by a favoured cleavage of the Val–Ser amide bond after Ser dehydration; this cleavage could be due to the formation of a cyclic amino acid residue from Ser.

When analysing the fragmentation of the cationated adduct [M+Na]⁺ ion at m/z 581, it was observed that the energy collision needed for fragmentation should be higher (60 eV) and that the corresponding b_n series was not observed. The main observed fragmentation yielded an abundant series of a_n ion fragments at m/z 553, 482, 395, 296 and 149, in agreement with the above sequence. The base peak at m/z 563 was due to the loss of a water molecule from the [M+Na]⁺ ion. These results suggested the sequence [H–Pro¹–Gly²–Phe³–Val⁴– Ser⁵–Ala⁶]⁺ for the linearised peptide ion derived from annomuricatin C and thus the structure 1 for the natural cyclohexapeptide.

2.3 NMR study

The ¹H-NMR spectrum of annomuricatin C (1) in DMSO-d₆ solution showed a main stable conformational state (> 85%) for which the five amide protons were clearly depicted, as well as the presence of six carbonyl groups in the ¹³C-NMR spectrum, in agreement with a hexapeptide structure including a proline (Table 1). The peptide sequence determination was based on the HMBC experiment. This heteronuclear methodology was preferred to the homonuclear method described by Wüthtrich and based on d_NN(i,i+1) and d_αN(i,i+1) connectivities from the ROESY/NOESY spectra [10,11], because for small size cyclic peptides, conformational information can interfere with sequential ones. All the amino acid spin systems were identified using scalar spin–spin couplings determined from the ¹H–¹H COSY and TOCSY experiments [12]. The ¹³C-NMR assignments of the protonated carbons were obtained from the proton detected heteronuclear HSQC spectrum and combined with the HMBC experiment optimised for a long-range J-value of 7 Hz, for the non-protonated carbons. This experiment allowed the carbonyl groups to be assigned. By this way, the sequence determination was done from the connectivities between the carbonyl of residue i with the amide and/or α protons of residue i + 1 (Fig. 3). The ³J_CH, CO (i) to NH (i+1) correlations depicted on the HMBC spectrum are shown in Fig. 4. As examples, the CO group of Gly² at δ 168.0 was correlated to both αH and α'H of Gly² and NH of Phe³, the CO of Ser⁵ at δ 168.8 to the αH and NH protons of both Ser⁵ and Ala⁶, and the CO of Pro¹ at δ 171.8 to the α, α'-H and NH of Gly² (Figs. 3 and 4).

The ROESY spectrum clearly showed only two d_{NN(i, i+1)} interactions between Gly² and Phe³, and between Ser⁵ and Ala⁶. A stretch of d_{αN(i, i+1)} sequential connectivities was depicted from Pro¹ to Ala⁶ (Fig. 4). The α proton of Ala⁶ gave a strong NOE correlation with the δ proton (3.77) of Pro¹ and a weaker with the geminal δ' proton (3.49), closing the cyclohexapeptide ring and indicating that the Ala⁶-Pro¹ amide bond was trans. Chemical shifts of the β and γ carbons of Pro¹ at 28.7 and 25.2 ppm, respectively, gave further evidence for the presence of a trans-Pro amide bond [13]. The prochiral assignment of the β, γ and δ protons of Pro resulted of the observed NOEs in the ROESY spectrum. In addition to the sequential NOEs, some other interactions with conformational significance were depicted: the NH proton of Ser⁵ was strongly correlated to the β-proton of Val⁴, and the β-methyl of Ala⁶ to the δ'-proton of Pro. The whole data agreed with the cyclic structure 1 for annomuricatin A, the sequence of which was thus determined as cyclo(Pro¹-Gly²-Phe³-Val⁴-Ser⁵-Ala⁶-).

Cyclic hexapeptide backbone rings are constrained with two β-turns generally stabilised by two transannular hydrogen bonds involving the two residues joining the β-turns and forming a short antiparallel β-sheet arrangement. Analysis of the NMR data of annomuricatin C suggested a constrained structure, leading to a main conformation.

All the ³J_NH–Hα coupling constants had high values, between 7.3 and 9.5 Hz, in agreement with β-turn and β-sheet structures. In a β-turn consisting of four residues numbered i to i+3, a trans-Pro¹ can only be located at the i+1 position, and as a result the following residue (Gly²) at the i+2 position. This is corroborated by the NOEs observed between NH–Phe³ (i+3 position) and both NH- and αH-Gly², and between NH–Gly² and αH–Pro¹. This first β-turn with Pro¹ and Gly² at the corners has thus the characteristics of a type II β-turn (Fig. 5). The second β-turn has Val⁴ and Ser⁵ at the corners and can be defined as a type-I β-turn. This is deduced from the strong d_{Ni, Ni+1} and d_{αi, Ni+1} interactions observed between Ser⁵ and Ala⁶ and the NOEs observed between NH-Ser⁵ and α-H and β-H of Val⁴. Confirmation of the conformation shown in Fig. 5, arose from the NOESY correlations between NH–Val⁴ and α-H and β-H of Phe³, and between NH–Ala⁶ and β-H of Phe³.

The temperature coefficients Δδ/ΔT were measured between 298 and 328 K in DMSO-d₆ (Table 1). The observed variations were linear and indicated that two amide groups (Phe³ and Ala⁶) with temperature coefficients between –0,9 and +0,5 × 10^–3 ppm K^–1 were involved in intramolecular hydrogen bonding, whereas the three other, especially that of Gly² (Δδ/ΔT: 10^–3 ppm K^–1) were solvent exposed, in agreement with two hydrogen bonds between CO–Ala⁶ and NH–Phe³ and between CO–Phe³ and NH–Ala⁶. The conformation of annomuricatin C, as typically observed for the cyclohexapeptide rings, appeared to be formed by a short antiparallel β sheet, stabilised by two hydrogen bonding and two β-turns, a type-II turn depicted for Pro¹-Gly² and a type-I turn for Val⁴–Ser⁵ (Fig. 5).

It is remarkable that the peptides isolated from the Annonaceae family, hexa- to nonapeptides, enclose at least one proline residue. As usual, this residue induces, in such small-size cyclic peptides, conformational constraints. In addition, its presence is useful, as it favours the formation of one linearised peptide, the mass fragmentation of which allows the sequence to be unambiguously determined.

3.1 Plant material

Seeds of Annona muricata L. (Annonaceae) were collected near Dakar (Senegal) in January 2000. Samples were immediately washed with distilled water and dried at room temperature. A voucher has been deposed at the National Museum of Natural History (Paris).

3.2 Extraction and isolation

The dried and powdered seeds of A. muricata (1.3 kg) were macerated three times with cyclohexane (3 l) and the combined extracts yielded an oil (344 g) that was discarded. The seeds were then extracted three times with MeOH (3 l) at room temperature to give after concentration under reduced pressure the MeOH extract (89.7 g), which was partitioned between EtOAc and water. The organic phase was concentrated to dryness and the residue (15.8 g) was dissolved in MeOH and chromatographed on Sephadex LH-20 column with MeOH. The peptide fraction (15.8 g) was then repeatedly subjected to silica gel column chromatography (Kieselgel 60 H Merck) and eluted with CH₂Cl₂ containing increasing amount of MeOH from 5% to 20%.

The peptide purification was monitored by TLC (silica gel 60 F₂₅₄ Merck) with CH₂Cl₂/MeOH 9:1 as eluent system and the peptides were detected with the Cl₂/o-tolidine reagent, exhibiting a characteristic blue spot with R_f 0.38. The corresponding peptide mixture was finally purified by isochratic reversed phase HPLC (Kromasil C₁₈, 250 × 7.8 mm, 5 μm, AIT France; flow rate 2 ml min^–1, detection 220 nm) using MeOH/H₂O: 65/35 with 1% TFA to yield annomuricatin C (1, t_R 7.2 min, 23.1 mg).

3.3 Absolute configuration of amino acids

Solution of 1 containing 1 mg of peptide, in 6 N HCl (1 ml), was heated at 110 °C for 24 h in sealed tubes. After cooling, each solution was concentrated to dryness. The hydrolysate was dissolved in anhydrous solution of 3 N HCl in 2-propanol and heated at 110° C for 30 min. The reagents were evaporated under reduce pressure and the residue dissolved in CH₂Cl₂ (0.5 ml) and 0.5 ml trifluoracetic anhydride was added. The mixture was kept in a screw-capped at 110 °C for 20 min, the reagents were evaporated and the mixture analysed on a Chirasil-l-Val (N-propionyl-l-valine-tert-butylamide polysiloxane) quartz capillary column with helium (1.1 bar) as carrier gas and temperature program of 50–130 °C at 3 °C min^–1 and 130–190 °C at 10 °C min^–1, with a Hewlett Packard series 5890 apparatus. Comparison of R_t values with those of standards amino acids was used: l-Ala (11.6), l-Val (13.9), Gly (14.6), l-Thr (15.2), l-Pro (18.2), l-Ser (18.8), l-Leu 19.2) and l-Tyr (31.9).

3.4 Annomuricatin C (1)

Colourless solid, Mp 284–285 °C (MeOH), [α]_D²²–2.7° (c 0.1, MeOH) ;¹H and ¹³C NMR (see Table 1); ESI-QTOF, m/z: 597.23 [M+K]⁺, 581.26 [M+Na]⁺, 559.28 [M+H]⁺; ESI-QTOF MS/MS on m/z 559 [M+H]⁺ (ce 40 eV) m/z (%): 559 (28), 541 (34), 531 (12), 513 (9), 488 (7), 470 (4), 460 (9), 442 (61), 427 (4), 414 (28), 401 (33), 374 (11), 373 (44), 345 (10), 328 (13), 326 (9), 319 (10), 302 (100), 295 (96), 274 (44), 258 (11), 238 (27), 226 (7), 219 (12), 172 (10), 155 (61), 141 (4), 127 (5), 120 (26), 72 (13), 70 (30). ESI-QTOF MS/MS on [M+Na]⁺ (ce 60 eV) m/z (%): 581 (96), 563 (100), 553 (93), 535 (73), 523 (49), 507 (20), 492 (42), 482 (75), 466 (51), 464 (49), 454 (26), 436 (36), 422 (17), 405 (4), 396 (25), 395 (95), 368 (21), 367 (93), 324 (22), 306 (7), 297 (20), 296 (91), 276 (13), 234 (11), 220 (6), 202 (6), 163 (4), 149 (2).