Catenation and encapsulation induce distinct reconstitutions within a dynamic library of mixed-ligand Zn₄L₆ cages

Subcomponent self-assembly generates dynamic combinatorial libraries of Zn₄L₆ cages whose composition is strongly affected by catenation and encapsulation.

Abstract

Two new Zn₄L₆ cages composed of diamine subcomponents containing either naphthalene diimide (NDI) or porphyrin moieties are described. Their structural differences allow these cages to exhibit distinct interactions with different chemical stimuli, yielding different supramolecular products. The electron-poor NDI subunits of the first cage were observed to thread through electron-rich aromatic crown-ether macrocycles, forming mechanically-interlocked species up to a [3]catenane, whereas the porphyrin ligands of the second cage interacted favourably with C₇₀, causing it to be bound as a guest. When mixed, the two cages were observed to form a dynamic combinatorial library (DCL) of seven constitutionally distinct mixed-ligand Zn₄L₆ cages. The DCL was observed to reconstitute in opposing ways when treated with either the crown ether or C₇₀: the electron-rich macrocycle templated the formation of heteroleptic catenanes, whereas C₇₀ caused the DCL to self-sort into homoleptic structures.

Introduction

Living systems respond in complex ways to different stimuli. The study of synthetic chemical systems that reconstitute and respond to different stimuli can offer insight into the signalling events that underpin biological systems,1 as well as identifying new means to synthesize useful stimuli-responsive systems2 and materials.3

One approach to investigating stimuli-responsive behaviour is to design dynamic combinatorial libraries (DCLs)4 of molecules that exchange components under thermodynamic control, and thus are inherently responsive to changes in physical conditions and the addition of different stimuli.5 This approach has been used to identify protein inhibitors,6 design mixtures of compounds that self-sort into different host–guest complexes7 and to engineer separations of mixtures of molecules held together by dynamic covalent bonds.8 In some cases, the response of a DCL to a stimulus has led, through reconstitution of the system, to the promotion of library members that may otherwise not be observed.9

As part of our investigations into stimuli responsive systems, we sought to design an adaptive library of different metallo-supramolecular cages that could respond to templates that are either bound inside the central cavity of a cage,10 or interact specifically with the edges of an assembly (i.e. catenation).11 To this end, it was necessary to design a library comprised of M₄L₆ cages containing two different kinds of diamine subcomponent, each of which engendered a response to a distinct kind of stimulus. It was also essential that the different ligands in such a library possess near-identical lengths to prevent sterically-driven self-sorting behaviour.12

Our DCL is based upon two similarly-sized Zn₄L₆ tetrahedral cages, built using ligands that contain naphthalene diimide (NDI) or Zn–porphyrin moieties, respectively. Combining these cages in solution yielded a DCL of seven different Zn₄L₆ cages, differing in their NDI : porphyrin ligand ratios. By utilizing different molecular recognition events, either the threading of the NDI components through a crown ether, or the encapsulation of the fullerene C₇₀, we explored the responsive nature of this DCL. The formation of either homoleptic or heteroleptic M₄L₆ structures was favoured, depending on the chemical stimulus that was employed. Our study thus represents a singular example of a multi-stimuli-responsive DCL of three-dimensional metallo-supramolecular assemblies. The challenge of analysing and characterising the members of this complex DCL was met by employing electrospray ionization Fourier-transform ion cyclotron mass spectrometry (ESI-FTICR MS).13

Results and discussion

NDI cage N₆ was synthesized through the subcomponent self-assembly14 of NDI-containing diamine 1 (6 equiv.), 2-formylpyridine (12 equiv.) and zinc triflimide (Zn(NTf₂)₂, 4 equiv.) in MeCN (Fig. 1). In similar fashion, Zn–porphyrin cage P₆ was prepared using Zn–porphyrin diamine 2 (6 equiv.), 2-formylpyridine (12 equiv.) and zinc triflimide (4 equiv.) in MeCN : CHCl₃ (7 : 3). Both assemblies consisted of Zn₄L₆ tetrahedra, with the bis(imine) condensation products of 2-formylpyridine and the respective diamine acting as bis-bidentate ligands linking adjacent metal centres.

Fig. 1. Synthesis of the Zn₄L₆ cages N₆ (top) from NDI-containing diamine 1, and P₆ (bottom) from Zn–porphyrin containing diamine 2.

Cages N₆ and P₆ were characterised in solution using NMR spectroscopy and ESI-FTICR MS. In both cases, ¹H NMR spectra showed multiple resonances for each proton environment, consistent with the presence of the three possible diastereomers of tetrahedral M₄L₆ assemblies: homochiral T (ΔΔΔΔ/ΛΛΛΛ), heterochiral C₃ (ΔΔΔΛ/ΛΛΛΔ), and achiral S₄ (ΛΛΔΔ).15 Analysis of the integrated signal intensities of the ¹H NMR spectrum of P₆ indicated a diastereomer distribution of 32 : 48 : 20 (T : C₃ : S₄) at 298 K (see ESI, Fig. S10†). The diastereomer distribution of N₆ could not be determined from its ¹H NMR spectrum (Fig. S3†) due to the greater degree of signal overlap.

The structures of N₆ and P₆ were confirmed in the solid state by single-crystal X-ray analysis (Fig. 2). Crystals of N₆ were obtained by vapour diffusion of diethyl ether into a solution of N₆ in MeCN containing excess KPF₆ (10 equiv.). The crystals were found to contain only the S₄-symmetric diastereomer. Of the six ligands that bridge the four octahedral zinc centres, four thus displayed a syn-conformation, bridging zinc centres of opposing handedness, and two adopted an anti-conformation, linking zinc centres of identical handedness. The metal–metal separations are 20.6–20.8 Å and 21.0 Å for the syn- and anti-ligands respectively.

Fig. 2. Crystal structures of Zn₄L₆ tetrahedra (a) N₆ and (b) P₆. Yellow lines connect Zn^II centres to highlight the tetrahedral frameworks. (c) Comparison of the two diamine residues, highlighting their similar lengths. Hydrogen atoms, disorder, solvent molecules and anions are omitted for clarity. Zn = yellow, C = grey, H = white, N = blue, O = red.

Suitable crystals of P₆ were obtained through vapour diffusion of diethyl ether into a solution of its triflate salt in MeCN. In this case, only the T-symmetric diastereomer was observed, with both enantiomers (ΔΔΔΔ/ΛΛΛΛ) present in the unit cell. The metal–metal separations range from 20.7–21.4 Å, similar to those observed for N₆. Axial water ligands were observed to bind to three porphyrin–Zn^II centres, with all three water ligands directed inside the tetrahedral cage (Fig. S32†). The remaining three porphyrin–Zn^II axial ligands were MeCN molecules that point outwards from the edges of the tetrahedron. Three of the faces of the Zn₄L₆ tetrahedron are almost completely enclosed by the Zn–porphyrin ligands that define the edges of the assembly. The remaining face is more open due to crystal packing effects, where a corner of each tetrahedron is observed to protrude slightly into the face of another.

Based upon the known affinity of electron-poor NDI moieties for the electron-rich macrocycle bis-1,5-(dinaphtho)-38-crown-10 (C),16 we anticipated that the NDI edges of cage N₆ would thread through C to yield a library of catenated cages. An excess of C (10 equiv.) was thus added to a CD₃CN/CDCl₃ (1 : 1) solution of N₆, and the mixture left to equilibrate at 298 K for 24 h. The resultant ¹H NMR spectrum showed the presence of new resonances consistent with C threaded around the NDI groups of N₆ (Fig. S7†), and the ESI-FTICR mass spectrum exhibited signals consistent with N₆ associated with up to four molecules of C (Fig. 3b). In our previously studied system based on a larger NDI-edged tetrahedron, adducts of the cage with up to 12 molecules of C were detected by ESI-FTICR MS. However, infrared multiphoton dissociation (IRMPD) tandem MS analysis demonstrated that the higher adducts were the result of non-specific interactions in the gas phase.16

Fig. 3. ESI-MS spectra of (a) Zn₄L₆ NDI cage N₆, (b) N₆ + C (4 equiv.), yielding N₆C₂, (c) Zn₄L₆ porphyrin cage P₆ and (d) P₆ + C₇₀ (5 equiv.), yielding C₇₀ ⊂ P₆.

Using a similar approach, the number of catenations per cage between N₆ and C was investigated by IRMPD analysis, in which mass-selected parent ions were irradiated with a CO₂ laser to induce fragmentation (Fig. 4, S19–22†). The [Zn₄N₆C(NTf₂)]⁷⁺ and [Zn₄N₆C₂(NTf₂)]⁷⁺ ions both underwent symmetrical fragmentation into ‘half-cage’ species [Zn₂N₃]⁴⁺ and [Zn₂N₃C]⁴⁺, which further decomposed into smaller fragments (e.g. [ZnN]²⁺ and [ZnNC]²⁺, Fig. 4a–d). When [Zn₄N₆C₃(NTf₂)]⁷⁺ was isolated and irradiated, [Zn₄N₆C₂(NTf₂)]⁷⁺ (the product of the non-cage-destructive loss of C), [Zn₂N₃]⁴⁺ and [Zn₂N₃C]⁴⁺ were observed (Fig. 4e and f). Furthermore, the non-symmetric [Zn₂N₃C₂]⁴⁺ fragment was not detected, indicating that the [Zn₄N₆C₃(NTf₂)]⁷⁺ ion present in the ESI-FTICR mass spectrum does not correspond to a [4]catenane, but results from the non-specific gas-phase interactions between N₆C₂ and C. Similar conclusions can be drawn from the fragmentation of the [Zn₄N₆C₄(NTf₂)]⁷⁺ ion, which shows successive loss of two molecules of C without destruction of the cage framework (Fig. 4g and h).

Fig. 4. IRMPD ESI-MS spectra of [N₆C_1–4NTf₂]⁷⁺ ions. (a) N₆C₁, 30% laser intensity (b) N₆C₁, 70% (c) N₆C₂, 30% (d) N₆C₂, 50% (e) N₆C₃, 25% (f) N₆C₃, 35% (g) N₆C₄, 20% (h) N₆C₄, 28%.

The inhibition of further catenation once two equivalents of C have been threaded onto the N₆ framework contrasts with the statistical non-cooperative binding behaviour observed between C and our previously reported larger NDI-containing tetrahedron. In the larger system, a [7]catenane was obtained, in which up to six molecules of C were threaded onto a tetrahedral framework. The more limited catenation propensity of N₆ appears to arise from its smaller overall size, leading to a sterically-controlled, inhibitive binding process.17 We postulate that the degree of catenation is limited to two crown ethers, because the binding of a third molecule would require two macrocycles to thread around adjacent NDI ligands, generating steric strain within the resulting assembly.

C₇₀ was identified as a potential guest molecule for both N₆ and P₆ due to their large internal cavities and the potential for π-interactions between C₇₀ and the ligands of the hosts.18 C₇₀ (5 equiv.) was thus added to an MeCN solution of either N₆ or P₆ and the mixtures were heated to 323 K for 24 h. Analysis of the mixture of N₆ and C₇₀ provided no evidence for host–guest interaction by either ¹H NMR or ESI-MS. We infer that the large pores in the faces of N₆ provide limited screening of C₇₀ from the bulk solvent; an inner phase is thus not differentiated, in contrast with more enclosed NDI-containing assemblies that have been shown to bind fullerenes.18a In contrast, NMR analysis of P₆ after the addition of C₇₀ provided clear evidence of guest binding (Fig. S15†).

The ¹³C NMR spectrum of the solution showed resonances consistent with the presence of C₇₀, despite its limited solubility in MeCN. ESI-MS of the solution also showed signals corresponding to the C₇₀ adduct of P₆ (Fig. 3d). We infer from these observations that C₇₀ is encapsulated within the cavity of P₆ (Fig. 5). The behaviour of P₆ contrasted with the analogous porphyrinato–Ni^II assembly, which was found to rearrange into new architectures in the presence of C₇₀.19 We attribute this differing behaviour to the planar conformation adopted by Zn–porphyrin 2 (as observed in its crystal structure, Fig. 2c and S28†), whereas the Ni–porphyrin congener adopts a saddled conformation.

Fig. 5. Summary of the host–guest of the homoleptic cages N₆ and P₆: catenation of N₆ with crown C yielding N₆C₂ (top), encapsulation of fullerene C₇₀ in porphyrin cage P₆, yielding P₆ ⊂ C₇₀, (bottom).

Having established the responses of N₆ and P₆ to the chemical stimuli of crown C and fullerene C₇₀ (Fig. 5), the responses of a mixed-ligand DCL of tetrahedra were investigated. Separate solutions of pre-formed N₆ and P₆ (CD₃CN/CDCl₃, 1 : 1, 1.0 mM) were mixed in a 1 : 1 ratio at 298 K. Attempts to confirm the equilibrium state by an independent synthesis from a mixture of all building blocks failed due to incompatible solubilities. As no further changes were observed after 24 hours, we inferred that equilibrium had been reached.

¹H NMR spectra of the mixture showed many resonances, corresponding to the numerous magnetically distinct environments of the mixed-ligand DCL members. The challenge of performing meaningful NMR analysis on this DCL was further complicated by the presence of resonances corresponding to the different regio- and stereo-isomers of the library members. The complicated nature of the spectra rendered DCL analysis by NMR impractical, even with the use of two-dimensional and diffusion-ordered experiments. Therefore, ESI-FTICR MS analysis was employed to gain insight as to the composition of the library. Mass spectra revealed signals which corresponded to all seven of the possible Zn₄L₆ assemblies, hereby labelled N_xP_(6–x) where N and P represent the NDI and porphyrin-based ligands (Fig. 6a).

Fig. 6. ESI-MS spectra (left) and corresponding charts (right) that show the relative abundances summed for all observed charge states: (a) 1 : 1 DCL of N₆ and P₆ after 24 h of equilibration, no template, (blue) and statistical distribution (orange) (b) 1 : 1 DCL of N₆ and P₆ containing C₇₀ (4 equiv.), (c) 1 : 1 DCL of N₆ and P₆ with C (0.5 equiv.), and (d) 1 : 1 DCL of N₆ and P₆ with C (4 equiv.).

Strictly speaking, a quantitative analysis of the changes in DCL composition using mass spectrometry requires that all species analysed possess similar ESI-response factors.20 As the homoleptic cages N₆ and P₆ appeared with similar intensities in the mass spectrum of the DCL (Fig. 6a), we infer that this requirement is met to such a degree that changes in peak intensity between mixed ligand species correlate to the degree of change in their solution concentrations in the DCL.21

All analysed DCLs were prepared under similar conditions, where solutions of N₆ and P₆ (CD₃CN/CDCl₃, 1 : 1, 1.0 mM) were allowed to equilibrate over 24 h at 298 K before template addition. The template (C or C₇₀) was added to DCLs in excess, at a ratio of 4 : 1 template/cage because this ratio was found to yield the strongest ESI-MS signals. Following template addition, the library was allowed to equilibrate at room temperature for a further 24 h, by which time no further changes to the library were observed. The MS signals observed in each experiment were processed to establish the relative abundance of each N_xP_(6–x) cage species (summing the detected m/z signal intensities of the +7 and +8 charge states for each ligand ratio, including those signals that comprise additional C or C₇₀, see the ESI†). In the case of the non-templated DCL (Fig. 6a), each one of the seven possible N_xP_(6–x) species is observed, with N₄P₂ and N₁P₅ occurring in highest abundance.

This distribution deviates from a statistical ratio of species which would follow the 7^th level of Pascal's triangle (1 : 6 : 15 : 20 : 15 : 6 : 1). We infer this deviation to result from a thermodynamic preference for the homoleptic cages N₆ and P₆. Minor differences between the lengths of subcomponents 1 and 2 could give rise to slight strain when both ligands are incorporated into the same assembly,22 thereby favouring the formation of homoleptic over heteroleptic assemblies.21 Given the high abundance of N₁P₅ observed from a 1 : 1 mixture of N₆ and P₆, attempts were made to form N₁P₅ exclusively in similar experiments using a 1 : 5 mixture of N₆ and P₆. These experiments also yielded a N_xP_(6–x) library, which was enriched with N₁P₅, rather than a single species (Fig. S28†).

Having generated a N_xP_(6–x) DCL, we sought to investigate its response to C and fullerene C₇₀, which interacted with the individual homoleptic cages N₆ and P₆, respectively. Following addition of C₇₀ to the DCL and equilibration for 24 h at 298 K (Fig. 6b), the composition shifted to give almost entirely N₆ and C₇₀ ⊂ P₆. ESI-MS analysis of the library showed that only P₆ was interacting with C₇₀ to form C₇₀ ⊂ P₆, indicating that this self-sorting behaviour can be attributed to the thermodynamic stability of the C₇₀ ⊂ P₆ host–guest complex, which removes the porphyrin ligands from the system.23 We observed no evidence for a mixed-ligand assembly that encapsulates C₇₀. We infer that the larger surface area of the porphyrin ligands relative to the NDIs (Fig. 2c) renders P₆, the most enclosed cage, uniquely capable of binding C₇₀.

While the addition of C₇₀ causes the library distribution to shift almost completely towards the formation of the species with an optimized binding pocket for the template, we anticipated that the addition of C would have a less pronounced effect on DCL composition. The recognition events of C are limited to interactions with the NDI components and thus can be expected to affect the library distribution in a way that will maximize the favourable NDI-napthalene π-interactions.

In the experiment where excess C (4 equiv.) was added to the N_xP_(6–x) DCL (Fig. 6d), ESI-MS of the equilibrated library showed signals corresponding to mixed ligand assemblies N₅P₁, N₄P₂, N₃P₃, N₂P₄, and N₁P₅. These mixed-ligand cages were observed to interact with 1 or 2 units of C.

No signals for the homoleptic assemblies N₆ or P₆ were observed. Only one signal in the ESI-MS is consistent with a species interacting with three units of C (N₄P₂C₃). Given that this stoichiometry exceeds the degree of catenation previously observed for N₆ (Fig. 4), we infer that the third crown ether is not likely to be catenated.16,24

In the similar experiment conducted with a sub-stoichiometric amount of C (0.5 equiv., Fig. 6c), the ESI-MS shows a distribution of all possible tetrahedral library members except P₆. In these experiments a maximum of one equivalent of C was shown to interact with each tetrahedron, and signals were also observed for tetrahedra that are not associated with C, as expected given the limited amount of C added in the experiment.

Regardless of the amount of C added, catenated cages with ligand ratios of N₄P₂ and N₁P₅ were the most abundant species observed in the DCL. This observation follows the trend of speciation in the non-templated DCL, where N₄P₂ and N₁P₅ were also the most abundant library members. However, the addition of C significantly increases the preference for N₄P₂ and N₁P₅ at the expense of N₆ and P₆ (Fig. 6c and d).

The shift in library composition towards mixed-ligand tetrahedra upon addition of C appears to maximize the number of interactions between the NDI ligands and C. In N₆, a maximum of two molecules of C can thread the six NDI ligands that are available. Therefore, shifting the library towards mixed-ligand assemblies increases the total number of NDI sites within the system that are available to thread through C. This hypothesis is supported by the absence of N₆ in the DCL when an excess of C is present and the observation that all of the species observed by ESI-MS are catenated one or two times.

In order to achieve the maximum degree of catenation between the NDI ligands and C, the optimal tetrahedra to form are N₄P₂, N₃P₃ and N₂P₄ because these species are each able to accommodate two molecules of C and would thus maximize the number of NDI-napthalene π-interactions. That N₃P₃ and N₂P₄ are only minor species in the DCL suggests that there are other factors influencing the distribution of the library. This inference is supported by the observation that the non-templated DCL also deviates from the purely statistical distribution, suggesting that the strain leading to a non-statistical distribution of the non-templated DCL also plays a role in controlling the DCL composition in the presence of C.

Experiments in which both templates, C and C₇₀, were added to the DCL either simultaneously or sequentially yielded poor ESI-MS data. We infer that aggregation may be occurring between C and C₇₀, which has been observed in other systems,25 precluding us from probing possible synergistic effects of these templates upon the DCL.

Conclusions

The difference in the shape and functionality of the NDI and porphyrin moieties of the ligands of N₆ and P₆ manifests in the form of distinct host–guest interactions with either C or C₇₀. The similarity in lengths of the edges of N₆ and P₆ allows for ligand exchange between the two assemblies, yielding a dynamic library of seven different tetrahedral cages N_xP_(6–x). This DCL was shown to respond to the addition of C or C₇₀, with the different templates influencing the composition of the DCL in distinct ways. A point of interest in our study is the use of both encapsulation and catenation (Fig. 7) as templation mechanisms within a DCL of three-dimensional receptors. Unlike systems derived from dynamic covalent species, the labile nature of these Zn₄L₆ tetrahedra precluded analysis by standard chromatographic separation4b and instead a methodology was employed where information as to the distribution of the library could be extracted from ESI-MS. This approach allowed us to decipher the behaviour of a complex mixture of seven different cage species, each comprising 22 individual components, and each of which was shown to form additional host–guest interactions. The multiple-stimuli responsiveness thus engendered could be of use in modulating guest uptake within more complex chemical networks that include DCLs of cages.

Fig. 7. A schematic overview of the different templation effects observed in the 1 : 1 DCL formed upon mixing of N₆ and P₆. Addition of C₇₀ leads to the self-sorting of ligands into their homoleptic cages, as the library shifts to form the C₇₀ receptor porphyrin cage P₆. Addition of macrocycle C leads to the formation of heteroleptic cages, in which the total number of N–C interactions is maximized, given that only two such interactions per cage are possible.