Imine-based [2]catenanes in water

The hydrophobic effect promotes the self-assembly of imine-based [2]catenanes in pure water.


Introduction
The synthesis of topologically complex molecules continues to challenge chemists, 1,2 prompting the need to develop new efficient synthetic strategies. In the literature, a few examples suggest that solvophobic effects could provide a powerful driving force to direct the assembly of catenanes and knots. [3][4][5][6][7] Most notably, it was serendipitously discovered that amphiphilic macrocycles can adopt non-trivial topologies in water [3][4][5] in order to minimize their hydrophobic surface area exposed to the solvent. The role of the hydrophobic effect in this process is poorly understood and it is not yet possible to predict the formation of complex topologies in such systems.
We describe here the dynamic combinatorial 8 synthesis of imine-based [2]catenanes in water from a dialdehyde (A, Fig. 1) and a series of homologous aliphatic diamines B n (n ¼ 4 to 9 denotes the number of CH 2 groups). Building block A is composed of two water-soluble, electron-decient aromatic surfaces (isoquinolinium bromide) connected by a p-xylylene linker and terminated by two aldehydes. Closing A with a hydrophobic aliphatic diamine provides amphiphilic macrocycles, wherein permanent positive charges alternate with lipophilic units. Our system was specically designed to probe whether the hydrophobicity of a purely aliphatic chain is sufficient to drive the formation of interlocked structures in the absence of motifs that favour more traditional supramolecular interactions (e.g. metal coordination, p-p donor-acceptor interactions or cooperative hydrogen bonds) and in spite of the charge repulsion between the isoquinonium moieties of A. In this context, the use of imine condensation presents several advantages. First, aliphatic diamines of various lengths are commercially available, allowing us to study easily the effect of chain length on the product distribution in the dynamic combinatorial libraries. In addition, aliphatic diamines display minimum hydrophilicity aer condensation. The reversibility of the imine bond has been extensively used to produce complex supramolecular architectures. 9 In water, the condensation of an aldehyde and an amine is generally disfavoured 10,11 but can be high yielding in some cases. In nature, a mechanism central to our visual cycle involves the formation of a Schiff base between the retinal chromophore and a lysine residue positioned in the hydrophobic cavity of opsin. 12 In the laboratory, imines can also be stabilized in water by metal coordination, 13 within the core of micelles, 14 or in the presence of an adjacent boronate. 15 In most other situations, chemists have preferred the related, more stable acyl hydrazones and oximes 16 over imines. With isoquinolinium-based aldehydes, the imine bond can form rather efficiently in water (condensation yield >60%, Fig. S3-S10 †). However, the equilibrium is very sensitive to the experimental conditions. In agreement with the literature, 11a we preliminarily assessed that imine condensation was most efficient at higher concentrations of building blocks, higher pH and lower temperatures.

Results and discussion
Dynamic combinatorial libraries were prepared from A and each one of the diamines B n (20 mM total, 1 : 1) in water under optimum conditions for imine condensation (room temperature, pH 9.6). The libraries re-organised within a few minutes when conditions were changed ( Fig. S7-S10 †), preventing us from analysing the libraries by conventional chromatography and to isolate the products formed. Clean imine reduction also proved challenging, mostly due to the presence of the reactive isoquinolinium unit. Therefore, we characterised the products formed directly within the libraries, using 1 H NMR spectroscopy and electrospray ionization mass spectrometry (ESI-MS), including traveling wave ion-mobility spectrometry (TWIMS).
The libraries reached thermodynamic equilibrium within ve minutes and their composition did not evolve any further over a few days. At equilibrium, a signicant amount of the initial A and B n remained. Nevertheless, the total yield of iminebased products, measured from the integration of the NMR signals using hydroquinone as an internal standard, ranged between 49% and 74% (Fig. 1).
Overall, the yields of imine-based products displayed an oddeven effect 17 with respect to n. Imine yields were lower if n was even (61% and 49% for n ¼ 4 and 6) and higher if n was odd (74% and 65% for n ¼ 5 and 7). The odd-even effect was attenuated for longer diamines and the yields slightly decreased (56% and 49% for n ¼ 8 and 9). With these longer diamines, a ne precipitate was observed, leading to the disappearance of imine oligomers from the solution. This behaviour explains why conversion and total yield of imine diverge when n $ 7.
The odd-even effect was also apparent in the library compositions. If n was even, the 1 H NMR spectra were rather complex (Fig. S11 †) suggesting the formation of mixtures of multiple oligomers. Several species were observable in the ESI mass spectra of these libraries.
The main species corresponded to the closed [1 + 1] macrocycles (M n ). Only traces of the larger [2 + 2] Fig. 2a) were observed. In contrast, the 1 H NMR spectra were much simpler if n was odd and were clearly dominated by the presence of one major product. In the mass spectra, M n was still observed, but the abundance of C n had signicantly increased ( Fig. 2a). From an entropic point of view, the formation of the relatively large [2 + 2] species was intriguing. Therefore, we investigated whether the it could be a [2]catenane (C n ), composed of two interlocked macrocycles M n , rather than a trivial macrocycle (T n ). The insource formation of a non-specic complex of two M n was immediately ruled out, as charge repulsion would promote dissociation rather than unspecic association, especially at the concentration at which the mass spectra were recorded (20 mM). For clarity, the results of this investigation are only described here when n ¼ 7 as a representative example.
Narrow arrival time distributions (ATDs) for both M 7 and the [2 + 2] species were consistent with single molecular structures. Both species were independently subjected to collision-induced decay experiments resulting in virtually identical spectra (Fig. 3a). This result strongly suggested that the [2 + 2] species was the [2]catenane C 7 . Initial bond cleavage of C 7 led to chargeseparation-driven dissociation into one closed and one open [1 + 1] macrocycles, which subsequently fragmented like the  Overall conversion (grey dotted line), total yield of imine products (blue line) and of [2]catenanes C n (red line) plotted as a function of n. These values were measured from the integration of the 1 H NMR signals (internal standard: hydroquinone). Insert: zigzag representation of odd-and even-numbered aliphatic chains within a macrocycle. simpler M 7 and thus gave rise to identical spectra. T n should fragment in a non-symmetrical fashion and its collisioninduced decay spectrum should be markedly different from that of M n . The collision energy needed to induce the fragmentation of C 7 (Fig. S38 †) also conrmed its interlocked nature. The acceleration voltages at which 50% of the parent ions have fragmented is signicantly lower for C 7 (15.5 V) than for M 7 (23.0 V) even though the larger C 7 ions can store more internal energy before they fragment. This agrees only with a catenated structure, which merely requires a single bond cleavage for the initial fragmentation. Charge repulsion within the quadruply charged [2]catenane further reduces the energetic barrier that must be overcome for cleavage. T n would only fragment when two covalent bonds are broken and would require a higher voltage than M n .
Another experiment was independently conducted to conrm our assignment without ambiguity. On the mass spectra of the full library, we selected the molecular ion m/z ¼ 256, corresponding to both C 7 and M 7 . We rst performed on this specic molecular ion collision-induced decay (CID) at various collision energies. The species produced were subsequently analysed with TWIMS and MS. Three peaks were observed in the mobilogram (Fig. 3b). The rst two peaks corresponded to C 7 and M 7 . The third peak corresponded to another isomer of M 7 , as identied by its isotope pattern, which exactly matched that of M 7 . Its longer dri time indicated a less compact structure and it could thus be assigned to the open form of M 7 , produced by cleavage of one bond via in-source fragmentation. Only traces of this species were observed in the previous experiments. As the collision energy increased from 10 V to 20 V, we observed that C 7 disappeared while the amount of the open and closed M 7 increased ( Fig. 3b and S37 †). At higher collision energies, M 7 also underwent bond cleavage, forming the open M 7 and smaller fragments. The cumulated increase of both M 7 peaks was directly proportional to the decrease of C 7 peak, showing that the fragmentation of C 7 cleanly produced the open and closed M 7 .
Similarly, we demonstrated by ESI-MS that the main products formed from odd-numbered diamines were the [2]catenanes C 5 , C 7 and C 9 (Fig. S42 †). Detailed 1 H NMR analysis of the libraries conrmed the catenated structure of these main products and provided more detailed insight into their binding mode. The spectrum of [2]catenane C 5 (Fig. 4b) reected its interlocked nature. The signals of the aromatic protons b and d were substantially shied upeld compared to the parent dialdehyde (Dd $ 1.5 ppm). Proton c exhibited the largest shi (Dd $ 3 ppm) and was considerably broadened. These shis were consistent with a conformation in which the four isoquinolinium units stack in an antiparallel way in order to maximise stacking and minimise charge repulsion. The [2]catenane C 5 must be conned into this compact conformation by the short length of the aliphatic chain. The high symmetry re-ected in the 1 H NMR spectrum implied that the inner and outer isoquinolinium units rapidly exchange on the NMR timescale by circumrotation of the two wheels. NOE cross-peaks (Fig. 5a) between the stacking isoquinoliniums (g 4 d) and their neighbouring xylylene (b, c 4 i) and aliphatic (g 4 j) moieties correlated well with the proposed structure.
The [2]catenanes bearing longer aliphatic chains (n $ 7) displayed signicantly different spectra. The [2]catenane C 7 (Fig. 4c) is described here as a characteristic example. All the aromatic protons of C 7 resonated in the same range as that of the parent dialdehyde A (7.5-10.3 ppm). However, the aliphatic protons k, l and m, located around 1.3-1.5 ppm in the parent diamine, were signicantly shied upeld, even to negative ppm values (0 to À2 ppm). The [2]catenane C 7 must therefore adopt a different conformation, wherein the aliphatic chain is threaded between the isoquinolinium units of the other wheel. This conformation is thermodynamically more stable because the four permanent positive charges are located further apart in space, thus minimizing their repulsion. In agreement with this proposed structure, NOE correlations conrmed the close proximity in space of the inner aliphatic and the outer aromatic protons (Fig. 5b). In comparison, the topologically trivial [1 + 1] macrocycle M 7 , clearly identiable in the 1 H NMR of the same library (triangles, Fig. 4c), did not present any deviations from the expected chemical shi values. Diffusionordered spectroscopy (DOSY, Fig. 4e) further conrmed our assignment of the library members. The smallest species of the library was B 7 (diffusion coefficient D ¼ 5.4 Â 10 À6 cm 2 s À1 ), followed by A (3.9 Â 10 À6 cm 2 s À1 ), M 7 (3.3 Â 10 À6 cm 2 s À1 ), and nally the larger C 7 (3.0 Â 10 À6 cm 2 s À1 ). As expected, the size of C 7 was comparable to that of C 5 (3.0 Â 10 À6 cm 2 s À1 , Fig. S20 †).
With a purely aliphatic chain threaded between the aromatic units, C 7 exhibits a rather uncommon structure. 4 The formation of such a [2]catenane appears to be mostly driven by the hydrophobic effect, even if the contribution of other supramolecular interactions cannot be excluded. To conrm this hypothesis, we performed an incremental addition of CD 3 CN 3,5 from 0% to 30% to the aqueous library containing C 7 (Fig. 6). The presence of a co-solvent with a lower dielectric constant (D CH 3 CN ¼ 37.5) than water (D H 2 O ¼ 78.4) alters the hydrophobic effect and shis the equilibrium in favour of M 7 . The kinetics of equilibration were also affected by the presence of a co-solvent and equilibration time signicantly increased with the proportion of acetonitrile (>2 h in 30% CD 3 CN). Similarly, all the [2] catenanes disappeared upon incremental addition of CD 3 CN or CD 3 OD (Fig. S54-S60 †), a solvent closer to D 2 O in term of its hydrogen bonding ability (D CH 3 OH ¼ 32.7). This latter experiment highlights the specic properties of water in the context of this study. Finally, the choice of the counterion (Cl À , Br À or CF 3 SO 3 À ) did not affect the library distributions ( Fig. S15, S24, S26, S36, S43 and S52 †), showing that the assembly was not templated by a specic counterion, as observed for other catenanes and more complex interlocked structures. 18

Conclusions
The self-assembly of amphiphilic imine-based [2]catenanes in water is mostly driven by the need to minimize the hydrophobic surface area exposed to water. It does not require the complementary use of metal coordination 6 or donor-acceptor p-p stacking. 4,5,19 Even the charge repulsion between the positively charged motifs is overcome by the hydrophobic effect. The products of the libraries could not be isolated, but extensive NMR studies and tandem MS experiments, together with ion mobility spectrometry, allowed for the unambiguous identication and characterisation of the [2]catenanes within the libraries. More importantly, our system generated   unexpected results regarding the complex role played by the length of the aliphatic chain. Indeed, the length of the aliphatic diamine controls both the yield and the conformation of the [2] catenanes. Evidently, odd-numbered aliphatic chains favour, and even-numbered chains disfavour, [2]catenane formation. Odd-even effects are oen observed in solid phase, 17 but are much less common in solution. 4,20 Within the [2]catenanes, the aliphatic chains cannot adopt a fully relaxed zigzag conformation. Folding of odd-numbered chains (Fig. 1, insert), required to close the [2]catenanes, induces a chair-like turn that resembles most a relaxed zigzag conformation. On the other hand, folding of even-numbered chains generates unavoidable gauche interactions. In dynamic combinatorial libraries, the nal equilibrium distribution can reect clearly such small energetic differences 8 (3.9 kJ mol À1 per gauche interaction).
The formation of the [2]catenanes is rather remarkable considering that imine condensation is not favoured in pure water, especially under dilute conditions. 11 The presence of [2] catenanes systematically correlated with an increase in the overall yield of imine ( Fig. 1), because the higher thermodynamic stability of the [2]catenanes shis the equilibria in favour of imine condensation. The system presented here is still quite simple but allowed us to probe how solvophobic effects can direct the formation of interlocked molecules. The knowledge gained from this study will help constructing topologically more complex architectures, such as knots or interwoven materials, using the hydrophobic effect.

Conflicts of interest
There are no conicts to declare.