Accessing Improbable Foldamer Shapes with Strained Macrocycles

Abstract

The alkylation of some secondary amide functions with a dimethoxybenzyl (DMB) group in oligomers of 8‐amino‐2‐quinolinecarboxylic acid destabilizes the otherwise favored helical conformations, and allows for cyclization to take place. A cyclic hexamer and a cyclic heptamer were produced in this manner. After DMB removal, X‐ray crystallography and NMR show that the macrocycles adopt strained conformations that would be improbable in noncyclic species. The high helix folding propensity of the main chain is partly expressed in these conformations, but it remains frustrated by macrocyclization. Despite being homomeric, the macrocycles possess inequivalent monomer units. Experimental and computational studies highlight specific fluxional pathways within these structures. Extensive simulated annealing molecular dynamics allow for the prediction of the conformations for larger macrocycles with up to sixteen monomers.

Frustrating folding propensity by imposing a macrocyclic structure to an otherwise helical backbone produces improbable yet stable strained molecular shapes. Both experiments and computational studies show how aromatic oligoamide foldamers constrained within a macrocycle try to fulfill their strong propensity to adopt helical conformations by forming local loops and unfolded sites that connect helical segments. Defined fluxional conformational pathways can be evidenced in such molecules.

Introduction

Macrocyclization is intimately linked to the stabilization of molecular shape.^[1] In flexible backbones, it helps to reduce conformational space and thus promotes conformations that are accessible to noncyclic precursors, but that may be disfavored for entropic reasons because they compete with multiple alternate stable states. Disulfide bridge formation in native protein structures and stapling of α‐helical peptides^[2] or proteins^[3] typically illustrate this effect. Folded macrocycles with multiple aromatic rings are another illustration.^[4] In contrast with the above, the so‐called shape‐persistent macrocycles consist of rigid subunits and generally have structures close to those of their related noncyclic precursors (Scheme 1a).[¹, ⁵, ⁶, ⁷] Their synthesis is often high yielding, even when it entails the oligomerization of multiple units. Shape‐persistent macrocycles are frequently flat, rigid, aryl‐containing rings, and often display interesting aggregation properties.^[8] A third category of macrocycles are those that are inherently strained. Strain generates constraints that reduce conformational freedom. It thus also stabilizes macrocycle shape but these shapes are spring loaded and by definition enthalpically disfavored in noncyclic precursors. Interest for strained rings has for long focused on small cycles with less than ten atoms because they display unusual reactivity and constitute desirable yet challenging synthetic targets.^[9] In recent years, various strained macrocycles have been produced as well. Typical examples include cycloparaphenylenes,^[10] cycles bearing angle strain alkynes,^[11] and hydrocarbon belts.^[12] In most cases, the synthetic approach involves the cyclization of a non‐strained shape‐persistent precursor, and subsequent irreversible transformations that change ring shape and introduce the strain. The resulting strain arises from the bending of rigid building blocks that would not be curved if the ring was cut open. With very large, less strained rings, using templates produces the desirable ring size.^[13]

Scheme 1.

(a) The ends of a crescent‐shape precursor may react to produce a shape‐persistent macrocycle. (b) When the precursor is a helix spanning not much more than one turn, cyclization may still occur and entail some distortion, for example, a saddle‐shape structure. (c) The ends of a longer helix may not meet, but the introduction of disruptors of helicity (P groups) allow for cyclization. A constrained macrocycle is obtained after removal of the P groups.

Here, we extend the concept of using strained macrocycles to generate otherwise inaccessible molecular shapes, by taking advantage of folding frustration. Considering a molecule that adopts a well‐defined folded conformation, the creation of a short covalent linkage between two parts of the molecule that are remote in space in the native fold will forbid this particular fold. If the molecule's folding propensity is weak, we anticipate that this destabilization of the native fold is unlikely to produce a new structurally well‐defined object. However, in a system where folding propensity is strong, we expect that the conformational frustration arising from the cyclization will result in one, possibly more, alternate stable conformations that could not be produced in a noncyclic structure. Because of their particularly strong folding propensity, aromatic amide foldamers^[14] are ideal candidates to explore this concept.

Many aromatic oligoamides possess an inherent curvature that has been exploited to produce shape persistent macrocycles,^[15] and that results in helical structures if a noncyclic sequence is extended beyond one turn. Oligomers of 8‐aminoquinolinecarboxylic acid Q (Figure 1a) adopt a particularly stable helical conformation in essentially any solvent.^[16] These conformations are promoted by the pseudo‐conjugation between sp²‐hybridized amide groups and quinoline rings, by electrostatic repulsions between amide oxygen atoms and neighboring quinoline endocyclic nitrogen atoms, and by bifurcated hydrogen bonds between amide NH and adjacent quinoline nitrogen atoms. These interactions give rise to a natural curvature of 2.5 units per turn. With minimal distortion, shape‐persistent trimeric macrocycle 1 can be obtained (Figure 1b,c).^[7] Tetramer 2 naturally spans more than 1.5 helix turns. Under conditions that destabilize the helix (heat and concentrated LiCl), it can nevertheless be cyclized into a saddle‐shape object with S ₄ symmetry (Scheme 1b, Figure 1b,c), which stood as a first illustration of our concept.^[7]

Figure 1.

Chemical structures of (a) non cyclic 8‐amino‐2‐quinolinecarboxylic acid oligoamides, and (b) macrocycles 1–5. (c) Crystal structures of 1–3 shown at the same scale.[⁷, ¹⁷] Included solvent molecules, isobutyl groups, and protons other than amide NHs have been removed for clarity. (d) Cis to trans amide conformational transition upon removing a dimethoxybenzyl group (DMB) from an aromatic amide bond.

However, longer Q_n helices are extremely stable (e. g. even in TFA) and no denaturing conditions have been reported until now. Joining the diverging ends of these longer helices to form larger homomeric macrocycles does not occur spontaneously. In a recent report,^[17] we showed that the introduction of removable 2,4‐dimethoxybenzyl (DMB) substituents on some amide nitrogen atoms disrupted the helix, allowing for the cyclization of a pentamer in the same series (Scheme 1c). After DMB removal, pentamer 3 was produced and characterized (Figure 1c,d). Conformational frustration in 3 is expressed by the occurrence of a disfavored cis secondary amide.^[18] In the following we show that the same strategy may be applied to produce hexamer 4 and heptamer 5. In both cases, original, compact structures are obtained that differ much from the noncyclic helices and that undergo complex fluxional behaviors. In addition, computational investigations allowed us to make plausible hypotheses about the conformational behavior of longer sequences with up to sixteen units. Given the variety of monomers and linkages that can be introduced in aromatic foldamer backbones, the breadth of systems to which this approach may be applied is considerable. Innumerable original molecular shapes may thus be accessed, which would not be significantly populated in noncyclic molecules.

Results and Discussion

Macrocycle synthesis

DMB groups have been used as removable amide conformational disruptors in solid phase peptide synthesis to prevent aggregation or folding on the solid support.^[19] Similarly, they have been used to disrupt aggregation and thus enhance the solubility of Kevlar‐like oligomers of 4‐amino benzoic acid, thus allowing for their functionalization prior to DMB removal.^[20] The alkylation of amides derived from an aromatic acid and an aromatic amine favors an unusual cis orientation of the two aryl rings, in sharp contrast with the trans conformation of the secondary amide precursors,^[21] a property that has been exploited to generate original tertiary amide and tertiary urea foldamers.^[22] If the alkyl group is removable, like the acid labile DMB, its insertion or removal results in a radical change of conformational preference. We have previously taken advantage of this conformation change to produce cyclic pentamer 3 from noncyclic precursors containing two DMB groups.^[17] In order to produce the equivalent hexameric and heptameric macrocycles, a series of noncyclic precursors 6 and 7 was synthesized (Scheme 2, see Supporting Information for synthetic Schemes and experimental procedures). We found that cyclization failed for all sequences containing a single DMB group and for 6 b which contains two consecutive DMB groups at the C‐terminus. In contrast, all precursors containing two DMB groups within their sequence could be converted into the corresponding macrocycles 8 c, 8 d, and 9 b in moderate to good yield.

Scheme 2.

Macrocyclization of quinolinecarboxamide oligomers of different lengths bearing DMB groups at various amides, and subsequent DMB deprotection. Conditions: (i) PPh₃, trichloroacetonitrile, DIPEA, CHCl₃; (ii) TFA (neat), 60 °C. DMB: 2,4‐dimethoxybenzyl.

The solid state structures of all three products were obtained by X‐ray diffraction analysis on single crystals (Figure 2a–2c). The three macrocycles contain in total six DMB‐substituted amides (two per product), of which five adopt the expected cis conformation (shown in purple in Figure 2), and one is trans (Figure 2a). The structures also reveal two flipped amides per structure (shown in green in Figure 2), that is, amide functions that deviate from the preferred conformation shown in Figure 1a. These amide groups may for example be perpendicular to an adjacent quinoline or with the carbonyl oxygen atom pointing towards – instead of away from – a neighboring endocyclic quinoline nitrogen atom.

Figure 2.

Views of the X‐ray crystal structures of: (a) 8 c, (b) 8 d and (c) 9 b. DMB groups are shown in orange ball and stick representation. The cis and flipped amide groups are marked with magenta and green balls, respectively. Hydrogen atoms other than NH, isobutyl side chains, and solvent molecules have been removed for clarity. Part of the 600 MHz ¹H NMR spectra at 298 K of (d) 8 c, (e) 8 d and (f) 9 b in CDCl₃. Amide NH protons, major and minor conformers are marked with filled squares (▪) and filled circles (•), respectively.

The ¹H NMR spectra of 8 c and 9 b both show two sets of signals that indicate that at least one conformer exists in solution in addition to the conformer observed in the solid state. Of note is the simple NMR spectrum of 8 d (Figure 2e) which matches with its C ₂‐symmetrical solid state structure conformation. These various conformers were not investigated further. They presumably represent conformations close to the conformers conducive to cyclization. The presence of flipped amides within these conformers suggests that some conformational distortion was necessary for cyclization, in addition to the DMB‐induced cis amide bonds. This makes it difficult to accurately predict which precursor may be prone to cyclization and which may not. Nevertheless, the results shown in Scheme 2 provide useful empirical guidelines.

The final removal of the DMB groups in neat TFA yielded 4 and 5, the structure and dynamics of which were investigated as follows.

Structure analysis and conformational dynamics of cyclic hexamer 4

The solid state structure of 4 could also be obtained by single crystal X‐ray diffraction (Figure 3a). The conformation has an average – non crystallographic – center of symmetry. It contains two helix segments with opposite P and M handedness linked “head to tail” by two extended (E) segments. Each E segment has two adjacent flipped amides (in green in Figure 3a), so there is a total of four flipped amides out of six in the structure. This highlights how far this object is from a canonical Q_n helix. Its conformation would be extremely unstable if not constrained in a macrocycle. We denote this conformation as ME₂PE₂, in which each letter indicates the M (red), P (blue), or E (green) folding pattern of a Q−Q linkage.

Figure 3.

(a) Views of the ME₂PE₂ X‐ray crystal structure of 4 and (b) snapshot of another stable species (M₂EM₂E) found in MD simulations of 4. Hydrogen atoms other than NH, isobutyl chains, and solvent molecules have been removed for clarity. From top to bottom are side views, top views and tube representations of the conformers. The vertices of the tube representation connect centers of mass of the quinoline and the amide units. Red, blue and green tubes/balls represent left‐handed (M), right‐handed (P) and extended sections (flipped amide, E), respectively.

The ¹H NMR spectrum indicates that hexameric 4 has a symmetrical structure with three inequivalent rings in solution as well (Figure 4c), in agreement with its solid state structure. This partial symmetry contrasts with the ¹H NMR spectrum of 2, which shows that all rings are equivalent in the tetramer (Figure 4a), and with that of 3 which shows that that all five rings are inequivalent in the pentamer (Figure 4b). In parallel, the conformation of 4 was assessed using molecular dynamics (MD) simulations (see Supporting Information for details). The simulations identified the solid state ME₂PE₂ structure as a populated species (Figure 3a), but they also revealed a second major conformer with an M₂EM₂E (or P₂EP₂E) structure (Figure 3b). The M₂EM₂E conformer possesses an average C ₂ symmetry axis and thus three inequivalent rings. Remarkably, it contains only two energetically disfavored flipped amides (dihedral N_Ar‐C_Ar‐C_Am‐N_Am angle ≈180°) and thus also stands as a plausible candidate for a stable conformer. This observation called for a validation of the solution structure of 4.

Figure 4.

Part of the 600 MHz ¹H NMR spectra at 298 K of (a) 2, (b) 3, (c) 4 and (d) 5 in CDCl₃ and of (e) 5 in DMSO‐d₆, showing amide and aromatic resonances. Black circles and squares indicate NH resonances of two distinct conformers in the spectra of 5.

The elucidation of the folded structure of 4 (in CD₂Cl₂ at 258 K where signals were particularly well resolved) was made possible by using a combination of 2D NMR techniques including ¹H‐¹H COSY, ¹H‐¹³C HSQC, ¹H‐¹³C HMBC and ¹H‐¹H ROESY (Figures S1‐S5). The assignment of a large part of the ¹H and ¹³C spectra (i. e. excluding some side chain protons) is reported in the Supporting Information (Table S1). NOE correlations, including strong correlations between the NH of flipped amides and neighboring aromatic protons (Figure S5) support the prevalence in solution of the structure observed in the solid state and exclude the computed conformer found only in MD simulations (Figure 3b).

The fluxional behavior of 4 was also investigated by MD simulations. Twelve simulations with different initial structures bearing structural characteristics of either ME₂PE₂ or M₂EM₂E were carried out for a total simulation time of ∼3 μs. Each simulation consisted of multiple heating/cooling cycles at up to 800 K. The observation of the ME₂PE₂ or M₂EM₂E conformer in each MD trajectory is exclusive, i. e. no interconversion between the two was observed, which precluded an assessment of their relative stability by means of sampling different populations of conformers. Ab initio optimizations in the gas phase were undertaken at the B3LYP/6‐31 g(d) level taking two snapshots from MD as initial structures, but they also remained inconclusive in this respect, because they are not representative of the multiple local minima observed during MD simulations, and because they do not provide solvation energy and entropic terms.

The MD simulations showed that the six degenerate conformations of ME₂PE₂ form two groups of three conformers that are only convertible within the group as two independent fluxional processes. We denote the degenerate states as M_iP_i+3 (or P_iM_i+3) in which i and i+3 indicate the central position of the two helical segments with M or P labelling handedness of the segment, respectively. The interconversion follows the M_iP_i+3↔P_i+1M_i+4↔M_i+2P_i+5↔P_i+3M_i+6 (which is M_iP_i+3) path or the mirroring P_iM_i+3↔M_i+1P_i+4↔P_i+2M_i+5↔M_i+3P_i+6 (which is P_iM_i+3) path. On average, four interconversions occur within ten heating/cooling cycles and the transition is quick and clean with no lingering intermediates. As a result, the ME₂PE₂ conformers account for ∼96 % of the total ensemble of MD trajectories at T=300 K (920 ns). The frequency of interconversion indicates that the degenerate states are separated by a modest barrier and are close to each other in the conformational space. Furthermore, the ∼96 % population indicates steep minima, and thus stiff conformers.

The absence of interconversion between ME₂PE₂ and M₂EM₂E conformers during the MD trajectories despite the high in silico temperatures involved hint at a high kinetic barrier between them. One could thus hypothesize that the observed ME₂PE₂ conformer is in fact a kinetically trapped conformation. Though this hypothesis cannot be formally excluded, we note that the TFA treatment of the two different DMB‐protected precursors 8 c and 8 d yielded the same ME₂PE₂ conformer of 4, i. e. that this conformer can be produced by different conformation pathways.

Structure analysis of cyclic heptamer 5

The crystal structure of heptamer 5 reveals an unprecedented miniature helix‐loop motif (Figure 5a). The structure contains only two flipped amides. The macrocycle is thus large enough to allow a sufficient number of monomers to fulfill their potential for noncovalent interactions and thus form two helical turns. This helix‐loop conformation resembles that of the DMB‐substituted precursor 9 b (Figure 2c), except that the latter contained two cis amides in addition to two flipped amides. The ¹H NMR spectrum of 5 in DMSO‐d₆ shows a single species with seven inequivalent quinoline rings, which indicates that fluxional behavior is slow on the NMR time scale (Figure 4e). In contrast, in CDCl₃, two species in comparable proportions were observed (Figure 4d), both with seven inequivalent monomers. Spectra in solvent mixtures allowed us to assign one of the two CDCl₃ species to the one that prevails in DMSO (Figure S7). A ¹H‐¹⁵N HSQC spectrum shows that the second species in CDCl₃ has an unusually upfield shifted amide at 7.20 ppm. This value is much lower than what is expected for hydrogen bonded amides, and even for flipped amides (Figure 2a, 2c), making it in unlikely that this conformer corresponds to that observed in the solid state. Thus, the solid state structure more likely corresponds to the species that prevails in DMSO.

Figure 5.

(a) Views of the M₄EME X‐ray crystal structures of 5 and (b) snapshot of a higher energy M₃CM₂E conformation found in MD simulations of 5. Hydrogen atoms other than NH, isobutyl chains, and solvent molecules are removed for clarity. From top to bottom are side view, top view and tube representation. The vertices of the tube representation connect centers of mass of the quinoline and the amide units. Red, blue, green and magenta tubes/balls represent left‐handed (M), right‐handed (P), extended sections (flipped amides, E), and cis amide (C), respectively.

Twenty‐five MD simulations of macrocycle 5 were then carried out, starting from different initial structures, each with multiple heating/cooling cycles (Table 1). The total ensemble of MD trajectories amounted to 8 μs. A major conformer, M₄EME (or P₄EPE), was found; it corresponds to the helix‐loop structure observed in the solid state and accounts for 70 % of the total ensemble of 300 K trajectories. Two other conformers, each accounting for less than 10 % of the total ensemble, were also identified. The first one, M₃CM₂E, is again a helix‐loop conformer (Figure 5b). It differs from the major conformer notably by a cis‐amide misfolding, noted as C. The second one, ME₂P₂E₂, resembles the centro‐symmetrical ME₂PE₂ conformation of 4 (Figure 3a), with one extra monomer inserted in a helical section. The pattern and frequency of occurrence of these two conformers suggest that M₃CM₂E represents a stable (albeit minor) conformation, whereas ME₂P₂E₂ is more likely an intermediate along the conversion pathways. Indeed, M₃CM₂E occurs about ten times less frequently but each occurrence lasts seven times longer than those of ME₂P₂E₂. Furthermore, one or more ME₂P₂E₂ conformers always appears along any of the interconversion pathways between two helix‐loop conformers, whereas M₃CM₂E does not. In addition, conformational fluctuation is much greater in ME₂P₂E₂, which is a characteristic of intermediate species.

Table 1.

Macrocycle conformations at the start and end of MD annealing simulations.

Qn macrocycle size	Starting conformation	Number of simulations	Total simulation time (ns)	Final conformational distributions[a]
6mer (4)	M2EM2E	7[b]	1120	M2EM2E (71 %[c]+26 %[d])
	ME2PE2	5	1840	ME2PE2 (96 %)
7mer (5)	M4EME	25[e]	8000	M4EME (70 %);[c] M3CM2E (4 %); ME2P2E2 (8 %)
8mer	M4E4	1	160	M4EM2E (61 %)
	M4EM2E	1	160	M4EM2E (66 %)
	ME3ME3	1	160	MEP2EPE2 (99.9 %)
9mer	M6E3	1	80	M4E2PE2 (99 %)
	M4E2PE2	1	160	M4E2PE2 (85 %)
10mer	M6E4	1	80	No stable structure
	M4EM4E	1	80	No stable structure
11mer	M6E5	1	80	M6E2PE2 (56 %); M6E5 (39 %)
	M6E2PE2	1	160	M6E2PE2 (56 %); M6E5 (40 %)
12mer	M4E2P4E2	1	320	M4E2P4E2 (>99.9 %)
	M8E4	1	80	M7E2PE2 (53 %)
13mer	M6E3ME3	1	80	M6E3ME3 (75 %)
14mer	M6E2P4E2	1	160	M6E2P4E2 (>99.9 %)
	M4E3M4E3	1	80	M4E3M4E3 (>99.9 %)
	M9E2PE2	1	160	M9E5 (99 %)
	M9E5	1	80	M9E5 (99 %)
16mer	M6E2P6E2	1	320	M6E2P6E2 (94 %)
	M6E3M4E3	1	200	M6E3M4E3 (>99.9 %)
	M9E3ME3	1	80	M9E3ME3 (75 %)

[a] M, P, C and E indicate left‐handed, right‐handed, cis‐amide, and extended Q−Q linkages, respectively. Conformers in italics contain strongly fluctuating sections typical of intermediate structures. Conformers in bold are thought to be prevalent. Percent values are based on 300 K data only. Percentage includes all degenerate states of each conformer. [b] Seven different starting structures with various combinations of aryl‐CO, aryl‐NH and cis‐amide misfolds at the two E sites. [c] Only aryl‐CO misfolds. [d] With one or two aryl‐NH misfolds. [e] Twenty‐five starting structures with varying misfold types and degenerate states.

Using a density functional theory (DFT) method (B3LYP/6‐31G(d)//WP04/6‐31G++(d,p)) and a continuum solvent model (chloroform), we calculated NMR chemical shift values of the different conformers of 5 after geometry optimization (see Table S2 and Figure S10). These calculations have limited predictive accuracy in the case of these large folded molecules involving extensive ring current effects, in contrast with other systems where they have been used effectively.[^4b, ^4c] For instance, tentatively matching unassigned signals between the calculated and observed spectra requires accommodation of chemical shift differences in the 0.5 to 1 ppm range. Nevertheless, one and only one high‐field shifted amide NH signal at 8.05 ppm is predicted for the cis amide NH of the M₃CM₂E conformer. Since this is at least 3 ppm upfield‐shifted from a normal value, we think that comparing observed and calculated signal is valid in this case. Thus, this makes it reasonable to propose that this conformer corresponds to the second species observed in NMR spectra in CDCl₃.

Fluxional behavior of cyclic heptamer 5

The main helix‐loop conformer of 5 has fourteen degenerate states, seven for M₄EME and seven for the enantiomer P₄EPE. During MD simulations, interconversions between these degenerate conformers occurred via two pathways. A M_i↔P_i±1 chirality inversion pathway dominated. A M_i↔M_i±2 (or P_i↔P_i±2) chirality preservation pathway occurred about half as frequently. On average, 0.6 conversions were observed for every 10 heating/cooling cycles. The transition typically involved various intermediates.

In order to investigate this fluxional behavior experimentally, an analogue of Q bearing a fluorine atom in position 6 was synthesized and incorporated into 5F, an analogue of 5 containing only one such unit (Figure 6a). Macrocycle 5F thus differs from 5 by the replacement of one proton by a fluorine atom. Its conformers are no longer degenerate, save that each exists as a pair of enantiomers. As for 5, only one conformer family is observed in DMSO, and two in CDCl₃. The ¹⁹F NMR spectrum of 5F in DMSO‐d₆ shows seven peaks noted from A to G in Figure 6c, that correspond to the seven conformers in which the fluorine is at one of the seven possible different sites of the folded conformation. Remarkably, the peaks do not have the same intensity, indicating that subtle energy changes result from placing the fluorine at one site or another.

Figure 6.

(a) Views of the X‐ray crystal structures of 5F. The flipped amide groups and fluorine atoms are marked with green and yellow balls, respectively. The fluorine atom is distributed over two sites in a 7/3 ratio. Hydrogen atoms other than NH, isobutoxy side chains, and solvent molecules have been removed for clarity. Part of the: (b) ¹H NMR (400 MHz) and (c) ¹⁹F NMR (376 MHz) spectra of 5F in DMSO‐d₆ at 323 K. (d) ¹⁹F EXSY spectrum (323 K, 376 MHz, mixing time 50 ms) showing chemical exchange between the different conformers. (d,e) Arrows denote the direction of the fluxional motion.

The crystal structure of 5F was solved (Figure 6a), and found to be similar to that of 5. Structure refinement failed to converge when trying to model the electron density of the fluorine distributed over seven sites. This could be attributed to weak residual density that would amount to less than 10–20 % of a fluorine atom. The structure converged when fluorine disorder was modeled over two sites in a 3/7 ratio; X‐ray crystallography is less accurate than NMR in this matter and some positions with a low occupancy factors may have been overlooked. Moreover, the prevalence of conformers in the crystal lattice is influenced by crystal packing, not just by their proportions in solution. Nevertheless, the crystal structure corroborates the observation in solution that the seven conformers of 5F are not equally populated.

Fast exchange on the NMR timescale between the conformers was not reached even at 110 °C. EXSY spectroscopy at 50 °C and variable mixing times revealed the existence of a dominant exchange pathway between the conformers (Figure 6d,e). Indeed each ¹⁹F signal was found to exchange predominantly with two others in the following order: A↔B↔F↔C↔E↔D↔G↔A. Whether this pathway matches with one of the two observed in MD simulations could not be ascertained. Nevertheless, we find that MD simulations make quite reliable predictions of the macrocycle structures and thus think this is the case for the prediction of fluxional behavior.

Prediction of the conformations of larger macrocycles

The simulated annealing MD of cyclic hexamer 4 and heptamer 5 were carried out in parallel with different initial conformers or different degenerate states within one conformer (Table 1). Each simulation had an extensive time span. The analysis of structure and dynamics of 4 and 5 was thus based on a sampling of the conformational space sufficient to obtain meaningful information on conformational distributions and interconversion pathways.^[23] Nevertheless, as sequence length increases, the occurrence of interconversion between degenerate conformers becomes less frequent: from four transitions per 100 ns for 4 to less than one transition per 100 ns for 5. For macrocycles larger than a heptamer, sufficient sampling of the conformational space would require an impractical computational time. Fortunately, the study of the penta‐,^[17] hexa‐ and heptamer allowed us to establish certain rules and patterns of macrocycle conformations. We thus endeavored to predict reasonably stable structures for macrocyclic sequences from the octamer to the hexadecamer by designing different initial structures that follow these rules and patterns.

For each macrocycle, one or more starting structures were built with a minimal number of flipped amides. In the above hexa‐ and heptamer simulations, initial structures were built using three types of misfolds, at aryl‐CO, aryl‐NH, or amide bonds. However, all initial structures converged quickly to show that flipped aryl‐CO bonds are the most energetically accessible misfolds, in agreement with earlier findings on the helix handedness inversion pathway of noncyclic Q_n oligomers.^[24] Therefore, only aryl‐CO misfolds were used to build the starting structures of longer macrocycles. No other misfolds were subsequently observed in the MD simulations. In the following, we use M, P, or E notations as defined above. We found that ring size constrains the number of units that can be involved in each P, M, or E segment in a given macrocycle. In addition, the number of units in a given helical section determines the relative orientation of its N and C‐termini and thus constrains the geometry through which these termini may be connected. Table 1 summarizes the starting conformations and the conformations after MD annealing for the different macrocycle sizes.

The M_xE_z helix‐loop geometry was considered as a reasonable starting point for a number of macrocycles (M₄E₄, M₆E₃, M₆E₅, M₈E₄, and M₉E₅). Except for M₉E₅, this conformation evolved during MD simulations to introduce one or two P or M folded elements within the loop: M₄EM₂E (Figure 7a), M₄E₂PE₂ (Figure 7b), M₆E₂PE₂, M₇E₂PE₂. This sometimes went along with a shortening of the initial helix segment. Yet, M_x or P_x helical sections with x=1,2 (i. e. that do not span one turn) sometimes stretch to retain an extended loop character, in order to match the height of the other helical section with two or more turns, as highlighted by the M₄EME conformation of 5 in the solid state (Figure 5a). Furthermore, these short helical sections may unfold as well: the two main 11mer conformations, M₆E₅ and M₆E₂PE₂, were found to interconvert frequently. The helix‐loop motif thus emerged as a viable conformation for many of the ring sizes.

Figure 7.

Conformations of larger Qn macrocycles as predicted by MD simulations. (a) 8mer. (b) 9mer. (c) 12mer. (d) 14mer top and side views. (e) 16mer. M, P, and E indicate left‐handed, right‐handed, and extended Q−Q linkages, respectively. For the structures shown in this figure, all E junctions entail a flipped secondary amide at an aryl‐CO bond. Atomistic representations of the molecular structures are shown in light grey(C)/red(O)/blue(N)/white(H) (hydrogen atoms other than NH and isobutyl side chains omitted for clarity). Overlaid tube representations depict the overall geometry. The vertices of the tubes connect centers of mass of the quinoline and the amide units. Red, blue and green tubes represent left‐handed (M), right‐handed (P) and extended (E) sections, respectively.

As an exception, the all‐E‐loop of M₉E₅ did not evolve during the simulations. A reason may be its genuine stability within its local conformational space, i. e. through folding/unfolding of one, or less than one, helix turn. Figure 7d (right) shows that the lateral length of the four‐turn helix M₉ segment matches well with the length of the E₅ loop. Such genuine stability is supported by the fact that a simulation starting with M₉E₂PE₂, whose loop contains a P section, quickly converges back to M₉E₅. Two helix‐helix conformers M₆E₂P₄E₂ or M₄E₃M₄E₃ are also found to be very stable by MD simulations (Table 1, Figure 7d). However, since no interconversion is observed, we cannot assess the relative stability of the three conformers via sampling.

Also considered as starting structures were helix‐helix conformations, i. e. conformations that contain two helical sections of three or more units. Annealing simulations showed an incredible stability of these conformers, that account for >99.9 % of the populations at 300 K in several cases (Figure 7c,d,e). It should be noted that the observed conformational changes in simulations starting with M_xE_z (helix‐loop) were limited to less than one helix turn (one or two units), whether it is folding of one or two extended units, or unfolding of one or two helical units. This is likely because the time span of each simulation was insufficient for folding and unfolding of longer sections. For example, the M₈E₄ starting structure evolved to M₇E₂PE₂ (53 % population in the time ensemble) in 80 ns, but was not able to evolve to the M₄E₂P₄E₂helix‐helix conformer. Though no direct conversion between M₇E₂PE₂ and M₄E₂P₄E₂ was observed, the fact that a simulation starting with M₄E₂P₄E₂ produces >99.9 % population of M₄E₂P₄E₂ suggests that it is more stable.

These results point to several simple structural rules. First, when z is odd, an E_z segment connects two helices of identical handedness; when z is even, the segment connects two helices of opposite handedness. In a macrocycle with two E_z segments, z values must then both be odd or both be even. Since most stable conformers are obtained with low values of z, two main classes of reasonable conformers having monomeric E or dimeric E₂ segments were found by MD to be stable: M_xEM_yE (and its enantiomer) and M_xE₂P_yE₂ (and its enantiomer), marked here as Type I and Type II conformers, respectively. As seen above, hexamer 4 adopts a Type II conformation (ME₂PE₂) while a Type I conformation (M₂EM₂E) was also found as a reasonable option by calculations but not observed experimentally (Figure 3b). Conversely, the prevalent conformation of heptamer 5 (M₄EME) is of Type I and is related to a helix‐loop structure. A Type I conformation is also predicted for the 8mer (M₄EM₂E, Figure 7a), but with large fluctuations in the M₂E segment, indicating its instability. We believe this instability comes from the mismatch of orientations of the C and N‐termini of helix segment M₄ to the N and C‐termini of helix segment M₂.

Secondly, in the macrocycle, the termini of the two helices are properly oriented with respect to each other. This orientation between the two termini in the helical plane is determined by the number of units and the intrinsic curvature angles of the Q units. Our study shows that M₁, M₄, M₆ and M₉ provide good termini orientation matches between helical segments,^[25] whereas M₂ requires a distortion of the intrinsic curvature and is thus energetically unfavorable. This is possibly the reason why the M₂EM₂E conformation of 4 is less favorable than the ME₂PE₂ even though M₂EM₂E has fewer misfolds. In summary, the second structural rule limits the number of units involved in each helix segment to a few favorable cases that do not constrain the intrinsic curvature of Q units.

The third rule concerns the length match between the two helical segments. Two helical segments with same or similar number of units result in more stable structures, be they helix‐loop or helix‐helix structures. M₄E₂PE₂, M₄E₂P₄E₂, M₆E₂P₄E_2, and M₆E₂P₆E₂ (Figure 7b,c,d,e) are examples of this kind. Otherwise, a mismatch in length results in less prevalent populations, as for M₆E₂PE₂, M₇E₂PE₂, M₆E₃ME₃ and M₉E₃ME₃.

In helix‐helix structures, i. e. when both helix segments span more than one turn, steric clashes would occur if a single E unit separates the helices because it is not long enough.^[26] M_xE₃M_yE₃ (x,y>2) conformations related to Type I are then encountered as, for example, M₄E₃M₄E₃ and M₆E₃M₄E₃. These sequences can also adopt Type II conformations M₆E₂P₄E₂ and M₆E₂P₆E₂, respectively (Figure 7d,e), i. e. conformations with fewer extended units, which is in principle more favorable. Nevertheless, the conformational analysis of 4 showed that the number of E units is not always the prevailing criterion to determine the most stable conformation. Type II conformers remain the most reasonable and common for a majority of macrocycles.

A few other situations worth mentioning were encountered in the context of these MD simulations. A reasonable conformation with three E segments (MEP₂EPE₂) was found for the 8mer. Most remarkably, MD annealing yielded no prevalent conformation for the 10mer. Multiple conformers were obtained of which none account for more than 33 %. This ring size thus constitutes an intriguing candidate for experimental studies. The 15mer was not investigated. Being that it has five more units than the 10mer (exactly two full helix turns), it might run into the same type of conformational conflict.

Conclusion

We have extended our trimer,^[7] tetramer^[7] and pentamer,^[17] series of synthesized and experimentally characterized cyclic Q_n oligomers to the hexamer and heptamer. Simulated annealing molecular dynamics allowed us to formulate general rules and to make reasonable predictions about the conformation behavior of longer homomeric macrocycles up to the 16mer. The conformations express the natural helicogenic behavior of the monomers and include flipped aryl‐CO bonds to accommodate the frustration that arises from constraining a helix in a short macrocycle. Two ensembles of plausible conformations dominate, namely helix‐helix and helix‐loop conformations. The prevalent hexamer and heptamer conformations represent prototypes of the helix‐helix and helix‐loop motifs, respectively. Altogether, the combination of macrocyclization and strong folding propensity emerges as a powerful approach to access shapes that would be improbable in noncyclic precursors. Given the great variety of aromatic amino acid monomers that have been developed in the context of foldamer research, there is considerable potential to extend our approach and diversify macrocycle structures. Furthermore, fluxional behavior within the macrocycles was shown experimentally and computationally to follow privileged pathways. The introduction of a single fluorine atom in the heptameric macrocycle showed some bias between the seven conformers that would otherwise be degenerate. This hints at the possibility that different side chains or aryl substitutions may prevent fluxional behavior, favor one conformer only and allow for the display of a defined array of functional groups. Progress along this line is currently being made and will be reported in due course.

Experimental Section

Molecular dynamics simulation setup: The general AMBER force field (GAFF),^[27] with improved torsional parameters,^[28] was used in all simulations. Each simulated system contains a macrocycle solvated by explicit chloroform molecules in a periodic box measuring 50 to 70 Å along each side. The systems were initially equilibrated for 1 ns at a constant temperature of 300 K and a pressure of 1 atm. This was followed by simulated annealing, in which systems go through multiple, 10 ns long cycles of heating, cooling, and constant temperature periods, at constant volume. To ensure sufficient conformational sampling, multiple simulations with different starting conformers for each macrocycle were carried out (Table 1). The systems were heated up to 800 K and cooled back down to 300 K for data collection. All simulations were carried out using the AMBER‐18 package.^[29]

Crystallographic data: Deposition numbers 2059866, 2059868, 2059869, 2059870, 2059871, and 2063705 contain the supplementary crystallographic data for this paper. These data are provided free of charge by the joint Cambridge Crystallographic Data Centre and Fachinformationszentrum Karlsruhe Access Structures service.

Conflict of interest

The authors declare no conflict of interest.

Supporting information

As a service to our authors and readers, this journal provides supporting information supplied by the authors. Such materials are peer reviewed and may be re‐organized for online delivery, but are not copy‐edited or typeset. Technical support issues arising from supporting information (other than missing files) should be addressed to the authors.

Supporting Information

K. Urushibara, Y. Ferrand, Z. Liu, K. Katagiri, M. Kawahata, E. Morvan, R. D'Elia, V. Pophristic, A. Tanatani, I. Huc, Chem. Eur. J. 2021, 27, 11205.