[C–H···anion] interactions mediate the templation and anion binding properties of topologically non-trivial metal–organic structures in aqueous solutions

We report the anion-recognition properties and anion-mediated templation of Metal-Organic knots and links in aqueous solutions.


Introduction
During the past 15 years, many small synthetic molecules displaying a variety of structural motifs have been developed for binding anions in organic or aqueous solvents. [1][2][3][4][5][6] The most challenging goal, and the one with the greatest potential rewards in terms of practical applications, is the selective recognition of anions in water. 1,2,7 Not surprisingly, a survey of natural anion receptors provides impressive benchmarks for comparison and emulation. For example, the sulfate-binding and transport protein of Salmonella typhimurium sequesters sulfate in water with a dissociation constant, K D , of 20 mM. 8 The phosphate binding protein of Escherichia coli binds phosphate selectively with a K D of 0.7 mM. [9][10][11][12] Recent achievements involving the recognition of anions in water by synthetic receptors include the sensitive detection of pollutants, 13-18 the transportation of ions across membranes, and the sensing of biologically relevant anions in vivo. [19][20][21][22][23][24][25] Gale and co-workers have published comprehensive reviews of these applications and other recent highlights in the eld. [26][27][28] Previously, we reported a one-pot synthesis of a set of topologically non-trivial, Zn(II)-templated complexes that were isolated as triuoroacetate (TFA) salts: a [2]catenane, [2]C(TFA) 4 ; a trefoil knot, TK(TFA) 6 , and a Solomon link, SL(TFA) 8 . 29 By relying on reversible imine and metal-ligand bond formation (ref. 30) we were able to form all three complexes simultaneously from a simple pair of chelating ligands: diformylpyridine (DFP) and a diamino-2,2 0 -bipyridine (DAB). 30 The [2] catenane was fully characterized by NMR spectroscopy, mass spectrometry and X-ray crystallography. The more complex structures, TK(TFA) 6 and SL(TFA) 8 , initially resisted full characterization. We could not grow X-ray quality crystals of TK(TFA) 6 and could only detect SL species by mass spectrometry at early stages of the reaction. We now report (i) the solid state characterization of a bromide containing trefoil knot complex, TK(TFA) 4 Br 2 , (ii) quantitative studies in D 2 O of the exchange of two TFA anions of the TK(TFA) 6 complex for various other monovalent anions and (iii) the effects of temperature and anion size and shape on the product distribution of the templation reaction itself. We demonstrate that different monovalent anions can be used to favor formation of either TK 6+ or SL 8+ in mixed aqueous solvents. A notable feature of this system is the cooperative effect of both cationic and anionic templates. The zinc(II) cation is necessary for complex formation, whereas the anion template inuences complex topology.
We would also like to call attention to the relatively rare structural motif by which TK 6+ and SL 8+ bind anions within their central cavities: multiple weak but cooperative C sp 2 -H hydrogen bonds. This motif is present in small anion binders such as the bisimidazoliums of the Maeda group, 31 as well as the triazole-containing macrocycles and podands reported by Flood and coworkers. [32][33][34] Previous examples of topologically interesting complexes in which this feature is present include Leigh's pentafoil knot, [35][36][37] which is templated, in part, by a central chloride ion; and the chloride and nitrate-binding rotaxanes 38,39 of the Beer group. Nevertheless, measurements of the anion binding affinities of molecular links and knots have rarely been reported. 38,40 It is particularly remarkable that the C-H hydrogen bonding of the TK 6+ and SL 8+ complexes are effective in D 2 O, one of the most competitive solvents. We believe that the results described below, in particular, the anion binding studies of TK 6+ are a unique contribution to the eld of aqueous anion receptor chemistry.

Results & discussion
Solid state structure of TK 6+ Single crystals containing TK(TFA) 4 Br 2 and suitable for X-ray diffraction were isolated by slow vapor diffusion of n-butylether into a triuoroethanol solution of TK(TFA) 6 that contained a small amount of tetrabutyl ammonium bromide. Tri-bladed propeller-shaped cationic complexes of TK 6+ crystallized as a racemic mixture in the trigonal P 3 space group. The crystal structure presented in Fig. 1b, depicts the C 3 symmetry of the knotted Zn II 3 L 3 complex (where L represents the condensed DAB + DFP organic ligands). Three equivalent L strands are held together by three Zn(II) ions located 13.3Å apart. Each zinc cation is coordinated to ve nitrogen atoms (two from the 2,2 0bipyridine and three from the 2,5-diiminopyridine) and has a distorted octahedral geometry, with its coordination sphere being completed by one triuoroacetate anion. The bipyridines of the ligand strand are located between phenoxy substituents that are attached to the imine moieties. The shortest distance between phenoxy and bipyridyl rings is $3.5Å, which is at the upper limit associated with effective p-p stacking interactions. An adduct composed of two tetrahedral [ZnBr(CF 3 COO) 3 ] 2À complexes that are hydrogen-bonded by water molecules is co-crystallized with the cationic knot (Fig. 1c). The presence of bromide anions was conrmed (Fig. S1 †) by EDAX analysis of the single crystal used in the XRD experiment as well as by mass spectrometry in the gas phase. In the solid state, two bromide ions were found to occupy the central cavity of TK 6+ and seem to be essential for crystallization, as all attempts to crystallize TK 6+ from bromide-free solutions failed.
TK 6+ cations and adducts alternate regularly in such a way ( Fig. 1c) that the bromides of two adducts reside in the center of TK 6+ and are each xed by at least three CH/Br À chargeassisted hydrogen bonds that range in length from 2.92 to 2.98 A. The strength of binding between the TK 6+ cation and the adduct also likely facilitates crystallization of the knot.
The average diameter and average depth of the TK 6+ cavity, as deduced from the crystal structure, are 5.86Å and 4.06Å, respectively. In the solid state, this cylindrical pocket hosts two bromide anions by establishing CH$$$anion interactions. 32,[41][42] Computational modeling of TK 6+ Due to the large size of the TK 6+ complex, semi-empirical PM6 calculations were used to determine a theoretically optimal geometry. The PM6 algorithm had reproduced fairly well (see ESI †) the experimentally determined geometry of [2]C 4+ . 29 Subsequent single-point calculations on the PM6-generated TK 6+ structure were performed using density functional theory (DFT) at the B3LYP/6-31G(d) level to determine the electrostatic potential at the surface of the knot. The electrostatic potential over the surface of TK 6+ , depicted in Fig. 3a, was calculated in aqueous solution as dened by an isodensity surface of 0.001 electrons bohr À3 . 43 Regions with the highest positive electrostatic potential are located on the aromatic C sp 2 -H hydrogens that point toward the center of the cavity.
The optimized geometry of TK 6+ presents a nearly undistorted C 3 symmetry, where the symmetry axis passes through the center of the central cavity. However, the helical arrangement of the three pyridyl units of the supramolecular assembly results in a sizeable dipole moment directed along the C 3 axis, which at the B3LYP/6-31G(d) level amounts to 2.77 D in the gasphase and 4.97 D in aqueous solution (Fig. 3a).
Two centroids are dened by the three methylene protons that point to each side of the central cavity. These centroids delimit the cavity's height, which measures 6.57Å. The radius of the base is 2.93Å and is estimated from the distances between the methylene protons and the centroid that they dene. Thus, assuming the simultaneous binding of two anions, the size of the TK 6+ cavity ($130Å 3 ) is suited for the recognition of two anions of relatively small radius (#2.5Å).
The favorable distribution of positive charge in TK 6+ and its demonstrated ability to rmly host two bromide anions within its central cavity in the solid state inspired us to explore the knot's anion recognition properties in solution. There are few examples of molecular links that have pre-organized cavities capable of aromatic hydrogen bonding, 40,44,45 and to the best of our knowledge quantitative anion binding studies involving knotted structures in aqueous media have not been reported.
The binding of monovalent anions of different size and shape to TK 6+ was initially evaluated computationally, using semi-empirical PM6 calculations. This computational study provided insight into possible binding modes and stoichiometries. The geometry optimized [TKBr 2 ] 4+ complex ( Fig. S2 †) is in reasonably good agreement with the X-ray structure. Energyminimized models of complexes of TK 6+ with linear (thiocyanate), trigonal planar (nitrate) and tetrahedral (tetra-uoroborate) anions are presented in Fig. 2. The anions are held within the knot's central cavity by multiple non-classical aromatic CH$$$anion hydrogen bonds.

Measurement of anion exchange and binding
Initially, TK 6+ was prepared as its triuoroacetate salt in isopropanol and characterized in methanol. 29 Aer discovering that TK(TFA) 6 is soluble and stable in water, we designed experiments to assess the knot's anion binding ability in this highly competitive medium. Preliminary 1 H-19 F HOESY NMR spectroscopy experiments (SI) gave no indication of interactions between the uorines of TFA and the hydrogens of the organic framework of TK 6+ ; hence, the mode of association of TFA with TK 6+ could not be determined.
The binding of bromide ion in water was evaluated in titration experiments monitored by 1 H NMR spectroscopy. Incremental amounts of aqueous tetrabutylammonium bromide were added to a D 2 O solution of TK(TFA) 6 at room temperature. The spectra show (Fig. 3c) the gradual spectral shis of TK 6+ resonances that occur as the result of the knot's interaction with bromide ions in solution. These continuous spectral changes are characteristic of an exchange process that is fast on the NMR timescale. The signal that corresponds to the H j protons exhibits the largest downeld shi, from 8.18 ppm (before addition of Br À ) to 9.10 ppm (upon saturation with the anion). Its change as a function of bromide ion concentration is illustrated in Fig. 3c. The signals that correspond to the diastereotopic H g protons are signicantly split apart as the concentration of bromide increases. The signal of the H e protons displays a small shi, and the signals that correspond to the H a , H b , H c , and H d protons are only slightly affected by bromide ion binding.
These spectroscopic results are consistent with the relative positions of the TK 6+ protons and bromide ions in the crystal structure of the knot, which shows, for example, that the H a , H b , H c , and H d protons point away from the knot's cavity and that the H g and H j bipyridyl protons point toward its center and are involved in hydrogen bonding with bromide.
The shi data was processed and found to t with good agreement to a 1 : 2 (TK 6+ : Br À ) binding model, indicating that two bromides bind within the cavity of TK 6+ in D 2 O. (A 1 : 1 binding model, and others, were not consistent with the data.) Mass spectrometry experiments (see ESI †) provided strong evidence for the existence of the 1 : 2 complex in the gas phase. A series of m/z fragments corresponding to [TK(TFA) 2 (Br) 2 ] 2+ , [TK(TFA) 3 Br] 2+ , and [TK(TFA) 4 ] 2+ were detected by ESI-HRMS, with no evidence for complexes containing more than two bromide ions.
From the solid state and solution studies, we infer that two TFA anions located outside the TK 6+ cavity are exchanged for two bromides that lodge within the cavity. The process is facilitated by electrostatic attraction between the positively charged host cavity and the negatively charged bromides. Multiple CH$$$anion interactions are formed in the host-guest complex, even in the aqueous solvent. The process is dynamic, with bound bromides being exchanged continuously with bromides free in solution.
The rst (K 1 ) and second (K 2 ) association constants (which can be considered to be binding constants) were calculated and found to be 4.4(0.5) Â 10 2 M À1 and 2.3(0.3) Â 10 3 M À1 , respectively. Interestingly, the TK(TFA) 4 Br 2 complex is signicantly more stable than the monobromide complex, TK(TFA) 5 -Br, which suggests that the binding of two anions is a cooperative process. The ratio of the calculated binding constants (K 2 /K 1 ) is approximately ve which indicates positive cooperativity and suggests that binding of the rst bromide ion causes conformational changes in the host's framework that facilitate binding of the second. [46][47][48][49][50][51] The same general pattern of spectral shis occurred during titrations of the knot with other monovalent anions of different size (ionic radii, r, of 1.7 to 2.4Å) and shape (Fig. S3-S7 †) including I À , N 3 À , SCN À , and NO 3 À (Fig. S3-S6 †). 1,32,52 The pattern was somewhat different with BF 4 À (Fig. S7 †). The H j signal was shied, indicating an interaction between the anion and the interior surface of the knot's cavity, but the direction of the shi was upeld rather than downeld. A sample containing the knot and BF 4 À was further analyzed in a 1 H-19 F HOESY experiment. The resulting spectrum (Fig. S8 †) shows NOE interactions between the uorine atoms of BF 4 À and the H g and H j protons, which conrms the close proximity of the anion and the walls of the knot's central cavity. Thus, all of the selected anions, including BF 4 À , bind within the cavity, though BF 4 À may extend beyond it, as suggested by computational modelling (Fig. 2, right). The shi of the H j signal that occurred during the titrations was used to calculate the association constants, K 1 and K 2 , for all anions except BF 4 À , whose association constants were deduced from the shi of the H h signal. A global binding constant, log b 2 , and a cooperativity parameter, K 2 / K 1 , were also calculated for each anion. The results are listed in Table 1. Regardless of size and shape, all of the selected anions were found to bind with the same 1 : 2 (TK 6+ : anion) stoichiometry. In the case of BF

anions.
As compared to other small-molecule receptors, TK 6+ exhibits high affinities for the selected anions in water. 1,2 However, the cooperativity of the two binding events associated with each anion type varied. For the spherical and linear anions, the K 2 /K 1 ratio was always signicantly larger (1.0 < K 2 /K 1 < 23.0) than 0.25, indicating positive cooperativity. 51 A comparison of the data for the two spherical anions investigated, Br À (r ¼ 1.82 A) and I À (r ¼ 2.06Å), revealed that I À binds with a slightly higher global binding affinity, but with lower cooperativity (log b 2 ¼ 6.1, K 2 /K 1 ¼ 3.3), than Br À (log b 2 ¼ 6.0, K 2 /K 1 ¼ 5.2). These results suggest that I À ts better in the cylindrical cavity of TK 6+ but that the smaller size of Br À minimizes anion-anion repulsion. Of the two linear anions, SCN À (length ¼ 2.13Å) and N 3 À (length ¼ 1.71Å), the larger one, SCN À , displays a comparable global binding constant but a higher cooperativity value (log b 2 ¼ 5.72, The unexpectedly higher cooperativity of SCN À is consistent with the negative charge of this anion being more localized on the nitrogen atom, a property that leads to a more directed interaction with TK 6+ and results in structural adaptation that facilitates the binding of the second SCN À anion. Considering the calculated structure of [TK(SCN) 2 ] 4+ illustrated in Fig. 2, we hypothesize that the two SCN À anions are bound in a staggered conguration that minimizes anion-anion repulsion and therefore increases binding affinity and cooperativity. Also, we suspect that resonance delocalization of the negative charge of N 3 À likely weakens this anion's interaction with TK 6+ and fails to induce the conformational changes that would enhance binding cooperativity. In contrast to the positively cooperative binding behavior of the spherical and linear anions, the binding of the larger trigonal planar (NO 3 À ) and tetrahedral (BF 4 À ) anions was negatively cooperative, with both K 2 /K 1 values being lower than 0.25. In these cases, the rst anion to bind hinders the second from binding due to increased steric and electronic repulsions. The K 2 /K 1 ratio of 0.21 for BF 4 À indicates statistically negative cooperation which can be attributed to the anion's relatively large ionic radius; however, BF 4 À was found to bind to TK 6+ with relatively high overall affinity (log b 2 ¼ 5.34). The larger triuoromethylsulfonate (triate, OTf À ) anion was titrated as its tetrabutylammonium salt, and no signal shiing in the knot's 1 H NMR spectrum was observed (Fig. S9 †). This result indicates that OTf À binding (to any part of the knot) is relatively weak under the experimental conditions and is likely due to the non-coordinating nature of the anion and its relatively large size which prevents entry into the knot's central cavity.
Binding experiments with Cl À , CN À , OCN À and ClO 4 À were also attempted, but all of these anions caused TK 6+ to precipitate from solution, which prevented accurate measurements. In addition to binding to the central cavity of TK 6+ , these anions might also be replacing TFA anions that coordinate axially to the Zn(II) metal centers and in this way causing reduced solubility and precipitation of the complexes in aqueous media.
Using variable temperature NMR spectroscopy, we studied and compared the binding of Br À with that of BF 4 À . At 298 K, the H g and H j signals are sharp in a spectrum (Fig. S15 †) of a solution of the knot and Br À measured at 298 K, which indicates that exchange of Br À is relatively fast at room temperature. At 268 K, the H g and H j peaks are considerably broader, which reects a slower exchange with respect to room temperature. In contrast, at 298 K, the H g and H j signals are broad in a spectrum (Fig. S16 †) of a solution containing BF 4 À and the knot, whereas they are sharp when the temperature is 333 K. Thus, for any given temperature, exchange of BF 4 À is slower.

Controlling [2]C 4+ , TK 6+ and SL 8+ populations in solution
The triuoroacetate salt of the DAB ligand that was previously used 29 for the synthesis of the three links was replaced by the corresponding neutral DAB molecule (see ESI † for synthetic details). Use of the neutral ligand prevented precipitation of the complexes in the mixed aqueous solvent and allowed us to monitor their simultaneous formation under different reaction conditions. 53 Mixing neutral DAB with DFP and zinc(II) acetate in a 1 : 1, D 2 O : MeOD solvent mixture, at temperatures ranging from 50 C to 90 C, lead to the formation of TK 6+ and [2]C 4+ in various proportions. When the reaction was carried out at 90 C, [2]C 4+ was formed exclusively, whereas, at 50 C, a signicant amount (63%) of TK 6+ was formed. We found that the TK 6+ : [2]C 4+ ratio was dependent on both temperature and solvent. There were no signs of SL 8+ formation under any of these conditions. The fact that a greater proportion of [2]C 4+ formed at higher temperatures substantiates previous ndings 29 of ours that suggested that [2]C 4+ is the thermodynamic product of the reaction and that TK 6+ is a kinetic product.
We had postulated that bromide was responsible for significant stabilization of the X-ray structure of the knot. This hypothesis was supported by an NMR investigation of the ability of Br À to template the formation of TK 6+ in solution. In the absence of a Br À template, DFP, DAB and Zn(OAc) 2 afforded a mixture of 63% TK(TFA) 4 Br 2 and 37% [2]C(TFA) 4 in a 1 : 1, D 2 O : MeOD solution at 50 C. However, the mole fraction of the knot increased to 77% upon addition of one equivalent (relative to the stoichiometry of the starting materials) of tetrabutylammonium bromide, and addition of two equivalents of bromide resulted in an even greater proportion (85%) of the knot (Fig. S19 †).
In a 1 : 1 : 1 mixture of CD 3 OD, D 2 O and CD 3 CN the bulky triate anion gave rise to an additional set of resonances in the spectrum of the reaction mixture (Fig. 4) that matched neither those of TK 6+ nor [2]C 4+ . Further NMR and mass spectrometric analysis allowed us to assign the new set of peaks to SL 8+ (Fig. S20 †). SL 8+ is much larger than TK 6+ and has a large central cavity that may be able to accommodate the triate anion. The size of the cavity was estimated with the aid of PM6 calculations. The optimized geometry of SL 8+ (Fig. S21 †) presents a nearly undistorted C 2 symmetry with a cylindrical cavity having a height of $5.90Å (6.57Å for the TK 6+ ) and bases of ca. 4.52Å Table 1 Successive (K 1 and K 2 ) and global (log b 2 ) binding constants determined from 1 H titrations of TK 6+ with monovalent anions at room temperature in D 2 O. Binding constants were deduced from 1 H chemical shifts measured as a function of anion concentration. H j protons were monitored in all cases except the BF 4 À titration, for which the H h protons were monitored. Standard deviations are shown in parentheses radius (versus 2.39Å for TK 6+ ). Thus, the central cavity of SL 8+ is signicantly wider than that of TK 6+ . The ESI-HRMS spectrum of the reaction mixture in which triate was used as a templating anion conrmed the presence of the three links. In addition to peaks characteristic of the catenane and the trefoil knot, the spectrum reveals two major peaks (Fig. S20 †)  Diffusion ordered 1 H NMR spectroscopy (DOSY) of reaction mixtures that included OTf À as a template produced spectra (Fig. 4b) that conrmed the presence of the three complexes. The diffusion coefficients of [2]C 4+ , TK 6+ and SL 8+ complexes were found to be 1.98(0.01) Â 10 À10 m 2 s À1 , 1.62(0.01) Â 10 À10 m 2 s À1 and 1.460(0.005) Â 10 À10 m 2 s À1 , respectively. The corresponding hydrodynamic radii of the complexes were calculated to be 1.35, 1.65 and 1.83 nm, respectively (see ESI † for calculations). Moreover, 1 H-19 F heteronuclear NOESY (HOESY) experiments revealed through-space interactions between the aromatic protons of SL 8+ 's cavity and the uorine nuclei of the triate anion. Fig. 4c shows three cross peaks that indicate correlations between the uorines and the H e , H g and H j protons of the Solomon link. These correlations conrm the close proximity of the protons and uorines and provide evidence for tri-ate's role as a template. The H g and H j protons seem to be involved in stronger coupling interactions, as indicated by the greater signal intensities of their cross peaks. PM6 calculations performed on the SL(OTf) 7+ system provided (Fig. S21 †) an optimized geometry in which triate is held inside the cavity of the host by CH/O and CH/F interactions involving H g and H j , and H e protons, respectively, and which is in good qualitative agreement with the experimental measurements.

Conclusions
Anion templation and temperature variation were used to control the product distribution of a dynamic library of zinc(II)based molecular knots and links ([2]C 4+ , TK 6+ and SL 8+ ). Electrostatic forces, including weak non-covalent CH$$$anion interactions that operated in the MeOD/D 2 O solvent mixtures mediated the topological outcome of the reaction. In the solid state, the electropositive central cavity of TK 6+ was found to accommodate two bromide anions with multiple CH hydrogenbonds. These CH$$$anion interactions occur between bipyridinyl units and bromide anions and are a major stabilizing feature in the packed crystal. In D 2 O, TK 6+ preserved its anion binding properties: monovalent anions of various shapes and sizes were found to bind to TK 6+ in 1 : 2 (TK 6+ : anion) stoichiometries and with high affinities, with log b 2 values typically in the range of 4 to 6.
Thermodynamic control over the library's product distribution was possible by varying the temperature of the reaction and/or by changing the anion template. Catenane [2]C 4+ , being the most thermodynamically stable complex, was formed exclusively when the reaction was carried at 90 C. Lowering the temperature to 50 C caused [2]C 4+ and TK 6+ to form simultaneously in 37% and 63% chemical yields, respectively. Addition of two equivalents of bromide ion to the reaction at 50 C resulted in a much greater proportion of TK 6+ (85%). Addition of the bulkier triate anion (OTf À ) allowed for formation and characterization of a Solomon link, SL 8+ . The presence of CH/F interactions inside the cavity of the SL 8+ were supported by 2D heteronuclear 19 F-1 H-HOSEY NMR experiments.
An analogous system in which both cations and anions inuence the distribution of several metallosupramolecular products by templation has been described by Nitschke and coworkers. 54 In that system a set of cages, helicates and prisms were formed. To our knowledge, ours is the rst such library involving knots and links.
With further development, anion binding within the topologically unique cavities of these molecular complexes could nd application in areas such as anion-sensing and anionassisted catalysis. For example, incorporation of uorogenic substituents into TK 6+ could allow for the sensing of specic anions. 41 Furthermore, it might be possible to fabricate ion selective electrodes by modifying the surfaces of the electrodes with molecular knots and links.