Conformation and anion binding properties of cyclic hexapeptides containing l-4-hydroxyproline and 6-aminopicolinic acid subunits

Abstract

Two cyclic hexapeptides containing alternating all R and all S configured l-(4R/S)-hydroxyproline and 6-aminopicolinic acid subunits are presented, and the influence of the hydroxyl groups on the solubility, conformation, and receptor properties is investigated. Cyclopeptide 2, containing the natural 4R configured hydroxyproline, adopts a conformation similar to that of the unsubstituted peptide 1, which is able to bind anions such as halides and sulfate in aqueous solution. 2 also interacts with these anions, but whereas 1 forms sandwich type 2:1 complexes, in which the anion is bound by two cyclopeptide moieties, 2 forms 1:1 complexes. The stabilities of the halide and sulfate complexes of 2 range between 10⁰ and 10² M⁻¹ in 80% D₂O/CD₃OD. Complex formation is detectable even in water, but with slightly smaller stability constants. Using this information a quantitative evaluation of the stability of the 2:1 complexes of 1, for which overall stability constants in the order 10⁴ to 10⁵ M⁻² in 80% D₂O/CD₃OD were observed, was made. In contrast to 2, the conformation of 3, containing the non-natural 4S configured hydroxyproline, is strongly affected by the presence of the hydroxyl groups. In d₆-DMSO and methanol/water mixtures a slow conformational equilibrium between two C₃-symmetrical conformers is observed, and 3 is thus much less preorganized for anion binding than either 1 or 2.

The crystal structures of anion binding proteins with the substrate resting in the active site have provided valuable information on the principles nature uses for the recognition of negatively charged compounds (1). Important binding interactions are, for example, multiple hydrogen bonds formed between the substrate and NH groups of the protein backbone or hydrogen bond donors in amino acid side chains (2, 3). A way of mimicking these interactions in an artificial system is to arrange amide or urea NH groups around the cavity of a suitable host molecule. This strategy has led to the development of many artificial hosts for anions, some of which possess remarkable substrate affinity, even in highly competitive solvents such as DMSO or acetonitrile (4–12). The ultimate goal has, of course, been the design of receptors that bind anions by hydrogen bonds in water. However, despite the recent advances in the development of anion receptors (4–7), most systems described so far require strong electrostatic or coordinative interactions to be effective in this solvent. One of the few exceptions is the cyclic hexapeptide 1 that has recently been described by us (13). This neutral macrocyclic receptor interacts with anions such as halides or sulfate even in water/methanol mixtures, a behavior that can largely be attributed to the special geometry of the complexes formed. We have found that the presence of a suitable anion in solution induces an aggregation of two cyclopeptide molecules, whereby a cavity is formed into which the anion is included (Fig. 1). In this complex, the anion is shielded from the surrounding solvent and can thus interact with the six NH groups of the two peptide moieties by hydrogen bonding. In addition, entropic effects, i.e., the release of solvent molecules from the hydration spheres of the host and the guest, contribute to complex stability.

Figure 1.

Side view of the 2:1 iodide complex of 1, showing the six N-H⋅⋅⋅I⁻ hydrogen-bonding interactions between the cyclopeptide rings and the I⁻ anion. The hydrogen atoms that are substituted by hydroxyl groups in 2 are marked in yellow.

Such a complex formation resembles the anion-induced self-association that has also been described for other systems (14–18). Furthermore, the structure of the complexes formed are comparable to those of the molecular capsules described by the groups of Rebek (19–22) or Böhmer (23–26), but because both cyclopeptide subunits in the complex are held together only by the anion and not by additional intermolecular interactions, the term sandwich complex is more appropriate. Two other sandwich-type anion complexes of artificial receptors have been described quite recently (27, 28). Only 1 binds anions even in aqueous solution, however, and therefore it represents a promising basis for further investigations on biomimetic anion complexation. A major advantage of this peptide is that its structure and consequently also its binding properties can be influenced simply by varying the amino acid subunits from which it is composed. In principle, even a combinatorial receptor optimization is conceivable.

In this article, we describe two derivatives of 1, 2 and 3, which contain hydroxyproline subunits instead of prolines (Structure T1). These peptides were conceived to increase the water solubility of 1, and thus allow investigations on receptor properties in water. Moreover, the additional OH groups could lead to the stabilization of a dimeric anion complex by intramolecular hydrogen bonds, a strategy that has been used by Rebek and Kang (29) for the stabilization of a molecular softball. Here we show that the configuration of the C atoms carrying the hydroxyl substituents in the proline rings has a profound effect on the receptor properties of the cyclopeptides. Furthermore, one of these peptides proved to be useful for a quantitative estimation of the anion affinity of 1.

Figure T1.

Structures of compounds 1–3.

Materials and Methods

General Methods.

Analyses were carried out as follows: melting points, Büchi (Flawil, Switzerland) 510 apparatus; optical rotation, Perkin–Elmer 241 MC digital polarimeter (d = 10 cm); CD, Jasco (Tokyo) J-600 (d = 0.05 cm); NMR, Varian VXR 300, Bruker (Billerica, MA) DRX 500 equipped with an automatic sampler; fast atom bombardment-MS, Finnigan-MAT (San Jose, CA) 8200; elemental analysis, Pharmaceutical Institute of the Heinrich-Heine-University; RP chromatography, Merck LiChroprep RP-8 (40–63 μm) prepacked column size B (310–25).

Job Plot.

Equimolar solutions of 2 (2 mM) and the guest in D₂O were prepared and mixed in various ratios. ¹H NMR spectra of the solutions were recorded, and the change in chemical shifts of the H(α), APAH(3), and APAH(5) cyclopeptide protons were analyzed.

NMR Titrations.

Stock solutions of the guest (50 μmol/800 μl) in D₂O and a cyclopeptide (1 μmol/200 μl) in 0.002% tetramethylsilane (TMS)/CD₃OD (for titrations in 80% D₂O/CD₃OD) or 0.002% CD₃OD/D₂O (for titrations in D₂O) were prepared. In total, 11 NMR tubes were set up for titrations with 2 and 21 for titrations with 1 by adding increasing amounts of the guest solution (0–800 μl) to 200-μl aliquots of the host solution. All samples were made up to a volume of 1 ml with D₂O, and the respective ¹H NMR spectra were recorded. The chemical shifts of the H(α), APAH(3), or APAH(5) protons of the cyclopeptide were referenced against the internal standard used and plotted against the ratio of guest/host concentration. From the resulting saturation curves, K_a and Δδ_max were calculated by the suitable nonlinear least-squares fitting method described for 1:1 (30) or 1:2 (31) complexes with SIGMA PLOT 3.0 (SPSS, Chicago) software package or the program eqnmr (32). Both methods gave essentially the same results, although eqnmr proved to be somewhat more sensitive to the starting conditions chosen.

Dipeptide tert-Butoxy Carbonyl (BOC)-(4R)-l-Hydroxyproline (Hyp)-6-Aminopicolinic Acid (APA)-OBn.

APA benzyl ester (1.71 g, 7.50 mmol) (13), BOC-(4R)-Hyp (2.61 g, 11.3 mmol), and chlorotripyrrolidinophosphonium hexafluorophosphate (4.74 g, 11.3 mmol) were dissolved in CH₂Cl₂ (150 ml). At room temperature, N-ethyldiisopropylamine (DIEA) (3.9 ml, 22.6 mmol) was added dropwise, and then the reaction mixture was stirred for 5 d. The solvent was subsequently evaporated in vacuo, and the residue was subjected to a silica gel column (hexane/ethyl acetate, 3:1; ethyl acetate). All fractions containing the product were collected and evaporated to dryness in vacuo. The remaining residue was dissolved in CH₂Cl₂ (100 ml), and this solution was poured into diethyl ether (800 ml) under stirring. After 10 min, the precipitate was filtered off, and the filtrate was again evaporated to dryness. From the residue, pure product was isolated by another chromatographic purification step on a silica gel column (hexane/ethyl acetate, 1:15). Yield: 2.41 g (73%); mp. 78–82°C; [α]_D²⁵ = −39.0 (c = 2, MeOH); ¹H-NMR (300 MHz, [d₆]DMSO, 100°C, TMS) δ 1.32 (s, 9H), 1.97 (m, 1H), 2.15 (m, 1H), 3.30 (d, ²J = 11.0 Hz, 1H), 3.49 (dd, ²J = 11.0 Hz, ³J = 4.8 Hz, 1H), 4.30 (m, 1H), 4.55 (t, ³J = 7.6 Hz, 1H), 4.71 (d, ³J = 3.8 Hz, 1H), 5.38 (s, 2H), 7.39 (m, b, 5H), 7.76 (dd, ³J = 7.5 Hz, ⁴J = 1.0 Hz, 1H), 7.94 (t, ³J = 7.7 Hz, 1H), 8.26 (dd, ³J = 7.7 Hz, ⁴J = 1.0 Hz, 1H), 10.47 (s, 1H); C₂₃H₂₇N₃O₆⋅2 H₂O (477.5): calculated, C 57.85, H 6.54, N 8.80; found, C 58.16, H 6.21, N 8.64.

Dipeptide BOC-(4R)-O-Acetyl-Hyp-APA-OBn.

BOC-(4R)-Hyp-APA-OBn (2.21 g, 5.00 mmol) was dissolved in acetic anhydride (20 ml). After addition of DIEA (1.30 ml, 7.50 mmol), the reaction mixture was stirred for 18 h at room temperature. Afterward, it was cooled with an ice bath, and the acetic anhydride was hydrolyzed by addition of water (100 ml). After stirring for 1 h, the aqueous layer was extracted three times with ethyl acetate. The combined organic layers were washed with water once, twice with 10% aqueous Na₂CO₃, and again three times with water. The solvent was removed in vacuo, and the product was isolated from the residue chromatographically (hexane/ethyl acetate, 1:1). The product was triturated with water to afford a white solid. Yield: 2.23 g (92%); mp. 113–117°C; [α]_D²⁵ = −28.6 (c = 2, MeOH); ¹H-NMR (300 MHz, [d₆]DMSO, 100°C, TMS) δ 1.33 (s, 9H), 2.01 (s, 3H), 2.22 (m, 1H), 2.34 (m, 1H), 3.45 (dt, ²J = 12.1 Hz, ³J = 1.8 Hz, 1H), 3.68 (dd, ²J = 12.1 Hz, ³J = 4.9 Hz, 1H), 4.60 (t, ³J = 7.6 Hz, 1H), 5.21 (m, 1H), 5.38 (s, 2H), 7.39 (m, b, 5H), 7.78 (dd, ³J = 7.5 Hz, ⁴J = 1.0 Hz, 1H), 7.95 (t, ³J = 7.9 Hz, 1H), 8.26 (dd, ³J = 8.3 Hz, ⁴J = 1.0 Hz, 1H), 10.56 (s, 1H); C₂₅H₂₉N₃O₇ (483.5): calculated, C 62.10, H 6.05, N 8.69; found, C 61.91, H 6.16, N 8.67.

Dipeptide BOC-(4S)-O-Acetyl-Hyp-APA-OBn.

BOC-(4R)-Hyp-APA-OBn (2.21 g, 5.00 mmol), triphenyl phosphine (1.97 g, 7.50 mmol), and acetic acid (0.43 ml, 7.50 mmol) were dissolved in tetrahydrofurane (30 ml). The reaction mixture was cooled with an ice bath, and diisopropyl azodicarboxylate (1.47 ml, 7.50 mmol) was added under stirring. After 30 min the ice bath was removed, and stirring was continued for 4 h at room temperature. The solvent was then removed in vacuo, and the product was isolated from the remaining residue by chromatographic workup (hexane/ethyl acetate, 1:1). The product solidified upon drying in vacuo. Yield: 1.86 g (77%); mp. 54–57°C; [α]_D²⁵ = −44.6 (c = 2, MeOH); ¹H-NMR (300 MHz, [d₆]DMSO, 100°C, TMS) δ 1.36 (s, 9H), 1.90 (s, 3H), 2.09 (m, 1H), 2.60 (m, 1H), 3.45 (d, ²J = 11.9 Hz, 1H), 3.74 (dd, ²J = 11.7 Hz, ³J = 5.6 Hz, 1H), 4.50 (dd, ³J(H_ax,ax) = 9.2 Hz, ³J(H_ax,eq) = 3.7 Hz, 1H), 5.13 (m, 1H), 5.37 (s, 2H), 7.44 (m, b, 5H), 7.78 (d, ³J = 7.7 Hz, 1H), 7.96 (t, ³J = 8.0 Hz, 1H), 8.24 (d, ³J = 8.4 Hz, 1H), 10.22 (s, 1H); C₂₅H₂₉N₃O₇ (483.5): calculated, C 62.10, H 6.05, N 8.69; found, C 61.99, H 6.13, N 8.46.

Cyclopeptides.

The syntheses of the acetylated linear hexapeptide precursors of peptides 2 and 3 were performed analogously to the procedure described for 1 by starting from the corresponding dipeptides (see supplementary material of ref. 13). For the cyclization, the hexapeptide deprotected at both ends of the peptide chain (1.98 g, 2 mmol) was dissolved in a mixture of N, N-dimethylformaide (DMF) (200 ml) and DIEA (2.16 ml, 12.4 mmol), and the solution was heated to 80°C. A solution of O-(1H-benzotriazol-1-yl)-N,N,N′,N′-tetramethyluronium tetrafluoroborate (0.71 g, 2.2 mmol) in DMF (20 ml) was slowly added dropwise. Stirring was continued for 2 h at 80°C, and then the solvent was evaporated in vacuo. The product was isolated from the mixture by chromatographic work-up. An initial purification step was carried out by using a silica gel column (CH₂Cl₂/MeOH, 10:1). The material recovered was further purified on an RP-8 column. For this, it was dissolved in a small amount of DMF and applied to a column conditioned with 1,4-dioxane/H₂O (1:10). The eluent composition was gradually changed until the pure product eluted (1,4-dioxane/H₂O, 1:2). It was finally recrystallized from acetone.

The acetylated cyclopeptide was dissolved in methanol/methylene chloride (1:1) (200 ml/mmol). After addition of DIEA (2 equiv.), the reaction mixture was stirred for 7 d at room temperature. Afterward, the solvent was removed in vacuo, and the remaining crude product was purified on a RP-8 column. For this, it was dissolved in a small amount of DMF and applied to a column conditioned with 1,4-dioxane/H₂O (1:10). The eluent composition was gradually changed until the pure product eluted (1,4-dioxane/H₂O, 1:4).

Cyclo[(4R)-Hyp-APA]₃ (2).

The yield was 0.42 g (30%); mp. > 250°C; [α]_D²⁵ = −553.7 (c = 2, DMF); ¹H-NMR (500 MHz, [d₆]DMSO, 25°C, TMS) δ 1.96 (m, 3H), 2.53 (m, 3H), 3.58 (d, ²J = 12.3 Hz, 3H), 3.63 (dd, ²J = 12.5 Hz, ³J = 3.3 Hz, 3H), 4.35 (m, 3H), 5.26 (d, ³J = 2.8 Hz, 3H), 5.55 (t, ³J = 8.5 Hz, 3H), 7.08 (d, ³J = 8.2 Hz, 3H), 7.41 (d, ³J = 7.6 Hz, 3H), 7.74 (t, ³J = 8.0 Hz, 3H), 9.51 (s, 3H); ¹³C-NMR (125 MHz, [d₆]DMSO, 25°C, TMS) δ 40.6, 56.3, 60.8, 66.9, 116.1, 119.6, 138.8, 148.1, 151.9, 166.1, 171.0; C₃₃H₃₃N₉O₉⋅2H₂O (735.7): calculated, C 53.88, H 5.07, N 17.13; found, C 54.16, H 5.00, N 17.41; fast atom bombardment-MS: m/z (relative intensity): 700(15) [M+H⁺].

Cyclo[(4S)-Hyp-APA]₃ (3).

The yield was 0.55 g (39%); mp. > 250°C; [α]_D²⁵ = −471.7 (c = 2, DMF); ¹H-NMR (500 MHz, [d₆]DMSO, 25°C, TMS, only the signals of the major conformer are considered) δ 2.13 (m, 3H), 2.67 (m, 3H), 3.51 (dd, ²J = 12.0 Hz, ³J = 4.6 Hz, 3H), 3.72 (dd, ²J = 12.0 Hz, ³J = 5.5 Hz, 3H), 4.32 (m, 3H), 5.03 (d, ³J = 4.0 Hz, 3H), 5.71 (dd, ³J(H_ax,ax) = 8.7 Hz, ³J(H_ax,eq) = 5.3 Hz, 3H), 7.40 (d, ³J = 6.4 Hz, 3H), 7.44 (d, ³J = 9.1 Hz, 3H), 7.72 (t, ³J = 8.1 Hz, 3H), 9.67 (s, 3H); ¹³C-NMR (125 MHz, [d₆]DMSO, 25°C, TMS) δ 40.0, 55.1, 60.1, 66.4, 115.0, 119.3, 138.9, 148.7, 151.9, 166.0, 170.4; C₃₃H₃₃N₉O₉⋅2.5H₂O (744.7): calculated, C 53.22, H 5.14, N 16.93; found, C 53.09, H 4.94, N 16.75; fast atom bombardment-MS: m/z (relative intensity): 700(20) [M+H⁺], 682(5) [M-H₂O+H⁺].

X-Ray Crystallographic Structure Determination of 3⋅CH₂Cl₂⋅3CH₃OH.

Crystal data are C₃₃H₃₃N₉O₉⋅CH₂Cl₂⋅3CH₄O, colorless crystals from MeOH/CH₂Cl₂, Mr = 880.74, orthorhombic, space group P2₁2₁2₁ (no. 19), a = 9.5108(1), b = 12.5969(1), c = 34.2121(5) Å, V = 4098.83 (8) ų, T = 100 K, Z = 4, ρ_calcd = 1.43 g⋅cm⁻³, μ = 0.232 mm⁻¹, crystal size 0.05 × 0.10 × 0.35 mm, Nonius KappaCCD diffractometer, Mo_Kα radiation, 1.19 < θ < 27.08°, absorption correction (T_min = 0.98029, T_max = 0.99545), 18,646 measured reflections, 8,740 independent and 5,988 with I > 2σ(I), programs: SHELXS-97 and SHELXL-97; both programs are from G. M. Sheldrick, University of Göttingen (1997); 550 parameters, R1 = 0.093, wR2(all data) = 0.239, (Δ/σ)max = 0.0, the two C-Cl distances were restrained to be equal (standard uncertainty 0.02 Å). OH hydrogen atoms were found on a difference Fourier map and refined isotropically, otherwise H atoms were refined by using a riding model, maximum/minimum residual density 1.596/−1.728 e Å⁻³ (0.84/0.16 Å from Cl2).

Results and Discussion

Synthesis, Structure, and Receptor Properties of 2.

In cyclopeptide 2, all of the l-proline subunits of 1 are replaced with the natural l-(4R)-Hyp. This peptide was obtained according to the general strategy that has been established in our group for the synthesis of similar compounds (33, 34). First, APA benzyl ester was coupled with BOC-l-(4R)-Hyp, and then the resulting dipeptide was acetylated at the hydroxyl group (Scheme S1). This acetylation is not necessarily required for peptide synthesis, but it has the advantage that the solubility of longer oligomers of the dipeptide in organic solvents is increased, and the chromatographic purification of these compounds is thereby facilitated. The acetylated dipeptide was chain elongated up to the linear hexapeptide, which was cyclized under dilution conditions after cleavage of the protecting groups at both ends of the peptide chain. The resulting acetylated cyclic product was deprotected in the last step of the synthesis to afford 2.

Scheme 1.

Syntheses of the building blocks of 2 and 3.

Peptide 2 is well soluble in methanol, methanol/water mixtures, and pure water. Thus, as expected, the introduction of three additional hydroxyl groups results in an increase in water solubility in comparison to 1. The hydroxyl groups do not, however, appear to significantly influence peptide conformation, as the close resemblance of CD spectra of 1 and 2 in 80% H₂O/CH₃OH illustrates (Fig. 2). Also the two-dimensional nuclear Overhauser effect NMR spectra of the two peptides in d₆-DMSO are comparable. The most informative feature in both spectra is the crosspeak between the NH and the H(α) signals, which demonstrates the close spatial proximity of the corresponding protons in both compounds. The arrangement of these protons in 2 most probably resembles that found in the crystal structure of 1 (13).

Figure 2.

CD spectra of 1 (red), 2 (blue), and 3 (black) in 80% H₂O/CH₃OH.

Because the spectroscopic findings suggested a close conformational relationship between 1 and 2, we also expected similar effects in the ¹H NMR spectra upon interactions of both peptides with anions. For 1 we have shown that anion complexation causes a downfield shift of the signals of the NH and the H(α) peptide protons. The NH signal is shifted because of hydrogen bond formation between the anions and the amide protons. This interaction brings the anions into close proximity to the H(α) protons, so that their resonance is also affected (13). In protic solvents, in which the NH protons are in fast exchange with the solvent protons, the shift of the H(α) signal is still clearly detectable, thus allowing investigations on anion binding even in these solvents. In the case of 2, an addition of sodium phenylsulfonate to a 2 mM solution of 2 in d₆-DMSO also resulted in a downfield shift of the NH and the H(α) signals by 0.40 ppm and 0.23 ppm, respectively (Figs. 6 and 7, which are published as supporting information on the PNAS web site, www.pnas.org). Moreover, in 80% D₂O/CD₃OD, the solvent mixture we have used to study the receptor properties of 1, and even in water the characteristic downfield shift of the H(α) signal occurs upon addition of, e.g., NaI or Na₂SO₄. Thus, 2 is also able to bind anions in highly competitive media, and the properties of this peptide seem to resemble those of 1 in many aspects. Only quantitative investigations revealed that the additional hydroxyl groups in 2 do in fact have a profound effect on anion binding. In particular, peptide 2 does not form 2:1 complexes. The Job plot of the iodide complex of 2 is compared with that of the corresponding complex of 1 in Fig. 3. The maximum of the curve at x = 0.5 for 2 clearly illustrates the formation of a 1:1 complex (35, 36). A similar result was also obtained for the sulfate complex. Furthermore, NMR titrations carried out with a variety of anions showed that the same stoichiometry holds for all of the complexes investigated. Thus, under the same conditions that lead to the formation of 2:1 complexes in the case of peptide 1, only 1:1 complexes are observed for 2. It therefore appears that the presence of the additional hydroxyl groups in 2 prevents the aggregation of two molecules of this cyclopeptide. This significantly different binding behavior can have various causes. It is possible that 1:1 complexes are preferred in the case of 2 because hydroxyproline groups are better solvated than proline groups, and a desolvation that accompanies an aggregation of two cyclopeptide molecules occurs less readily. Another reason could be steric. The x-ray structure of the sandwich complex of 1 with I⁻ shows that the cyclopeptide subunits approach one another to within van der Waals distances (Fig. 1). It is therefore probable that the introduction of the hydroxyl groups in the 4R positions of the proline rings (indicated in yellow in Fig. 1) would prevent the two cyclopeptide units from coming together. Finally, we can currently also not rule out the possibility that 2 adopts a conformation in the anion complexes, in which the OH groups participate in hydrogen-bonding interactions with the guest.

Figure 3.

Job plots of the iodide complexes of 1 (Left) and 2 (Right).

The formation of 1:1 complexes in the case of 2 allows a straightforward evaluation of anion affinity by NMR titrations (30, 35). Not only could the shift of the H(α) proton be followed in these titrations, but also those of protons in the 3 and 5 positions of the aromatic peptide subunits, whose resonances are also affected by complex formation. The fact that the stability constants calculated from the saturation curves obtained were essentially independent of the proton that was followed demonstrates that the assumption of 1:1 complex stoichiometry is valid and that no higher complex equilibria have to be considered. In Table 1, the stability constants and maximum chemical shifts of the H(α) protons of 2 are summarized for different anion complexes. We have used only salts of strong acids as guests to avoid protonation equilibria of the anions during the titrations and the need to use buffers. Interactions of 2 with nitrate were too weak to be followed quantitatively.

Table 1.

Anion complex stabilities of 1 and 2 at 298 K [stability constants K₁ and K₂ in M⁻¹, K_a in M⁻², error limits of the stability constants of the complexes of 1 < 20% and 2 < 40%; Δδ_max maximum chemical shift of the peptide H(α) protons]

Salt		1 80% D2O/CD3OD		2 80% D2O/CD3OD		D2O
		K	Δδmax	K	Δδmax	K	Δδmax
NaCl	K1	5	0.58	8	0.32	5	0.31
	K2	6,770	0.23
	Ka	0.34 × 105
NaBr	K1	16	0.49	13	0.43	10	0.36
	K2	6,820	0.46
	Ka	1.09 × 105
NaI	K1	22	0.55	19	0.58	14	0.50
	K2	7,380	0.84
	Ka	1.62 × 105
KI	K1	24	0.50
	K2	6,660	0.87
	Ka	1.60 × 105
N(CH3)4I	K1	24	0.53
	K2	7,470	0.83
	Ka	1.79 × 105
Na2SO4	K1	96	0.44	95	0.51	52	0.44
	K2	1,270	0.77
	Ka	1.22 × 105

Table 1 shows that 2 forms the most stable complexes with the 2-fold charged sulfate anion, and that the interaction with the spherical halides increases from chloride through bromide to iodide. This order can most probably be rationalized in terms of cavity size available for anion complexation. The arrangement of the NH groups in the crystal structure of 1 indicates that these are better placed to interact with the larger iodide or sulfate anions rather than with smaller anions (average N⋅⋅⋅N distance 4.50 Å) (13). Table 1 also shows that interactions of 2 with anions are not only measurable in 80% D₂O/CD₃OD but also in water. Although the anion affinity in this solvent is small, the fact that peptide 2 represents a neutral receptor able to bind anions even in water makes the observed stability constants still noteworthy.

The stability constants of the anion complexes of 2 in 80% D₂O/CD₃OD also provide valuable information about the complex stabilities of 1. As in 2, shifts of the H(α), APAH(3), or APAH(5) signals in the ¹H-NMR spectra of 1 during complex formation allow the equilibrium to be followed quantitatively by means of NMR titrations. However, the saturation curves obtained have to be fitted to four parameters K₁, K₂, Δδ_max1, and Δδ_max2, and although mathematical methods for such nonlinear regressions are described (31, 32), all have in common that the results often depend sensitively on the starting conditions chosen. This made a quantitative determination of the anion complex stability of 1 difficult despite the fact that we increased the number of data points in the NMR titrations by a factor of 2 to describe the saturation curves more exactly. The results obtained for peptide 2 provided us, however, with an estimation for K₁ of the anion complexes of 1. By using this information in the nonlinear regressions, we were able to describe all saturation curves satisfactory. Although the errors of K₁ and K₂ resulting from these calculations are relatively large (ca. 40%), the fact that we obtained comparable complex stabilities from the saturation curves of different receptor protons made us confident that our approach is valid (Figs. 8–11 and Tables 2 and 3, which are published as supporting information on the PNAS web site). The resulting stability constants determined for the complexes of 1 are also included in Table 1.

The good agreement between K₁ of the complexes of 1 and 2 is evident. K₂ is significantly larger than K₁ in every complex of 1, and overall complex stabilities in the order of 10⁴–10⁵ M⁻² result. Interestingly, K₂ is of the same order of magnitude for all three halide complexes of 1, which implies that the aggregation of two cyclopeptide subunits is essentially independent of the halide anion that is initially bound. The overall stability of the halide complexes is thus only determined by K₁ and increases from chloride to iodide, an order that is consistent with the one we have derived from electrospray ionization MS measurements (13). The stability constant K₂ of the sulfate complex of 1 is smaller than those of the halide complexes, but because of the larger K₁ of this complex, its overall stability is also high.

All investigated anion complexes of 1 thus possess remarkably high stability constants in aqueous solutions. Moreover, the countercation has, at least in the case of the iodide complex, no significant effect on complex stability, because by replacing sodium iodide with the corresponding potassium or tetramethylammonium salt in the titrations, practically identical values for K₁ and K₂ were obtained (Table 1). We will now carry out microcalorimetric measurements, which should provide information about the enthalpic and entropic contributions to complex formation. At this point it was also interesting to ask, however, whether 2:1 complex formation of 2 is only prevented by steric effects of the hydroxyl groups. If this is the case, a cyclopeptide 3 with the non-natural S configuration at hydroxyproline subunits should be able to form sandwich complexes like 1.

Synthesis, Structure, and Receptor Properties of 3.

Cyclopeptide 3 was synthesized analogously to 2 by starting from the acetylated dipeptide with a 4S configured hydroxyproline subunit, which was obtained from the product of the coupling reaction between BOC-l-(4R)-Hyp and APA benzyl ester under Mitsunobu conditions similarly to described procedures (37–39) (Scheme S1).

Surprisingly, 3 is almost insoluble in pure methanol or water. In water/methanol mixtures, its solubility is somewhat better but still significantly lower than that of 1 or 2. This property is an initial indication that the conformation of 3 differs more than originally expected from those of the other cyclopeptides. The CD spectrum of 3 supports this assumption because it clearly deviates from those of 1 or 2 (Fig. 2), and also the NMR spectrum indicates a different conformational behavior of 3. Whereas 1 and 2 possess simple ¹H NMR spectra consistent with an averaged C₃ symmetrical conformation, in the spectrum of 3 in d₆-DMSO at room temperature, two sets of signals in a ratio of ca. 1:10 are visible, each of which represents a C₃-symmetrical peptide conformation (Fig. 4). Increasing the temperature leads to a shift and a broadening of all signals, and at 150°C the two signal sets are essentially reduced to only one. However, even at this temperature the signals are broad, and the underlying conformational equilibrium of 3 is thus slow on the NMR time scale. The temperature-dependent shift of the NH resonance is −6.5 ppb/K for the major conformer of 3 and −4.7 ppb/K for the minor conformer. This finding indicates that in the minor conformer, the NH groups are involved in intramolecular hydrogen bonds (40).

Figure 4.

¹H-NMR spectrum of 3 in d₆-DMSO. The signals of the major conformer are marked with red dots and those of the minor conformer with blue dots.

Similarly to peptides 1 and 2, a crosspeak between the NH and H(α) signals of the major conformer of 3 is visible in the nuclear Overhauser effect two-dimensional NMR spectrum in d₆-DMSO. There is none, however, between the same signals of the minor conformer. An addition of 1 equiv. of n-butyltrimethylammonium iodide to a 2 mM solution of 3 in d₆-DMSO causes a downfield shift of the H(α) signal of the major conformer by 0.11 ppm. In contrast, the NH signals of both conformers and the resonance of the H(α) protons of the minor conformer are almost unaffected by the presence of the salt. When the tosylate salt of the same cation is used instead of the iodide, essentially only one set of signals remains in the ¹H NMR spectrum, with both the NH and the H(α) signals shifted downfield by 0.67 and 0.51 ppm with respect to the corresponding signals of the major conformer, respectively. But even in this spectrum, small signals of the minor conformer are still visible (Figs. 12–16, which are published as supporting information on the PNAS web site). Changing the solvent from d₆-DMSO to water/methanol mixtures causes a shift of the conformational equilibrium of 3 toward the minor conformer. In 50% D₂O/CD₃OD, e.g., the ratio of the two conformers is ca. 1:1.6. Upon addition of NaI, the signals of the minor conformer decrease in favor of the ones of the major conformer, and the H(α) signal of the major conformer is simultaneously shifted downfield. All of these results strongly suggest that the structure of the major conformer of 3 resembles the preferred conformations of 1 and 2. At this point we had not enough information to assign a structure to the minor conformer but for this, x-ray crystallography proved to be helpful.

Peptide 3 crystallizes from methanol/methylene chloride with three molecules of methanol and one molecule of methylene chloride. The corresponding crystal structure is depicted in Fig. 5. It shows that all three proline amides of 3 adopt the trans conformation in the crystal and consequently, the structure in Fig. 5 closely resembles those we have determined for cyclic hexapeptides with 3-aminobenzoic acid subunits (33, 34). The unusual anion affinity of 1 is, however, associated with a conformation involving cis proline amides (13). Because we have already seen that 1 structurally corresponds to the major conformer of 3, it is reasonable to assume that the crystal structure is close to the minor conformer. This assumption is supported by the fact that temperature-dependent NMR spectroscopy indicates the presence of intramolecular hydrogen bonds in the minor conformer. Indeed, despite the presence of methanol during crystallization, two of the NH groups of 3 form intramolecular N-H⋅⋅⋅O hydrogen bonds to neighboring OH groups (N1⋅⋅⋅O7, N4⋅⋅⋅O8) of hydroxyproline subunits in the crystal (Fig. 5). The third NH group does not form such a hydrogen bond because it is interacting with a methanol molecule (N7⋅⋅⋅O10) included in the peptide cavity. Moreover, in accordance with the nuclear Overhauser effect NMR spectrum of the minor conformer of 3, the NH and H(α) protons are arranged in an orientation that should not give rise to a crosspeak.

Figure 5.

Crystal structure of 3⋅CH₂Cl₂⋅3CH₃OH, viewed onto the plane through the three amide N atoms, showing the main H-bonding interactions. Selected distances (Å): N4⋅⋅⋅O8 2.924(6), N7⋅⋅⋅O10 2.843(7), N9⋅⋅⋅O10 2.909(6), N1⋅⋅⋅O7 2.832(6), O9⋅⋅⋅O11 2.991(9), O7⋅⋅⋅O11 2.795(8), O8⋅⋅⋅O2* 2.755(6), O9⋅⋅⋅O11 2.991(9), and O7⋅⋅⋅O11 2.795(8).

The conformational equilibrium observed for 3 thus seems to be caused by a cis/trans isomerization at the proline amides. Because only C₃-symmetrical conformers are observed in the ¹H-NMR spectra, only the all-cis and the all-trans conformers of 3 would appear to be possible. The presence of significant amounts of the all-trans conformer of 3 in solution can be ascribed to the stabilizing effects of intramolecular hydrogen bonds, which cannot be formed in peptide 2 because of the different configuration at C(4) of the hydroxyproline subunits.

We have not been able to make a quantitative evaluation of the anion affinity of 3 because the conformational behavior of this peptide in combination with its low solubility in solvents used for 1 and 2 made such investigations difficult. It is reasonable to assume, however, that because of the unfavorable preorganization of 3, the anion affinity of this peptide is significantly lower than those of the other two derivatives studied.

In conclusion, our investigations show that subtle structural variations in a macrocyclic receptor can have a profound effect on conformation and binding properties. The hydroxyl groups in 2 prevent a formation of 2:1 complexes, whereas in 3 they induce a peptide conformation that is not well suited for anion complexation. Peptides 1 and 2 thus represent promising artificial receptors for anion recognition in aqueous solution. Whereas 2 possesses a weak anion affinity even in water, the anion affinity of 1 in methanol/water mixtures is high.

Abbreviations

BOC

tert-butoxy carbonyl

DIEA

N-ethyldiisopropylamine

Hyp

l-hydroxyproline

APA

6-aminopicolinic acid

TMS

tetramethylsilane

DMF

N,N-dimethylformamide