Self-organization by selection: Generation of a metallosupramolecular grid architecture by selection of components in a dynamic library of ligands

Abstract

Self-organization by selection is implemented in the generation of a tetranuclear [2 × 2] grid-type metallosupramolecular architecture from its components. It occurs through a two-level self-assembly involving two dynamic processes: reversible covalent bound connection and reversible metal ion coordination. Thus, mixing the aminophenol 3, the dialdehyde 4, and zinc acetate generates the grid complex 1a(Zn) via the assembly of the ligand 2a by imine formation and of the grid by zinc(II) binding. When the same process is conducted in a solution containing a mixture of different aminophenol and carbonyl components, the generation of the grid 1a(Zn) drives the selection of the correct components in a virtual dynamic library of ligands, displaying an amplification factor of >100 and a selectivity of >99%. Component exchange as well as reversible protonic modulation of the assembly/disassembly process display the dynamic character of the system and its ability to respond/adapt to changes in environmental conditions. The processes described demonstrate the implementation of a two-level self-organization by selection operating on the dynamic diversity generated by a set of reversibly connected components and driven by the formation of a specific product in a “self-design” fashion.

Self-organization (1–7) of chemical entities of higher complexity from their components may be directed by the design of both these components and their mode of assembly, i.e., by the molecular information stored in the components and by the supramolecular processing of this information through the interactional algorithm (the interaction pattern) involved. Its sensitivity to changes in internal or external parameters defines the robustness, or conversely the adaptability, of the organizational program. The implementation of self-organization processes represents a major goal of supramolecular chemistry and rests on the design of adequately programmed chemical systems.

The dynamic nature of supramolecular entities furthermore endows the system with selection, adaptation, and evolution features. Thus, in addition to self-organization by design, self-organization by selection may take place, provided a dynamic diversity of constituents is produced in the system on which selection may operate (7). Such diversity generation results when a collection of components undergoes continuous recombination of their connections to yield a set of interconverting constituents. It defines a constitutional dynamic chemistry operating on both the molecular and supramolecular levels (7).

The latter involves noncovalent interactions, which are reversible by nature, and is therefore dynamic by essence. It is illustrated, for instance, in the self-selection of ligand strands in helicates (8) and building blocks in supramolecular cages, capsules (9–11), and polymers (12).

The former covalent constitutional dynamic chemistry is dynamic by design because it requires the deliberate introduction of reversible covalent bonds to allow for constitutional variation through exchange and rearrangement of components, as occurs in various types of species (polymers, molecular capsules, macrocycles, etc.; for a recent comprehensive review see ref. 13), and is implemented in particular in the developing field of dynamic combinatorial chemistry (14–17). Selection occurs under the pressure of either internal [intrinsic relative stabilities of different combinations as in helicate self-recognition (8); self-stabilization in the dynamic formation of imine-linked foldamers (18)] or external [interaction with species in the environment, as in anion binding by circular helicates (19, 20)] factors. It may be kinetic as well as thermodynamic and incorporates statistical factors, whereby heterocombinations are favored over homo-combinations for multicomponent assemblies constructed from a mixture of different building blocks.

Ligand selection on self-assembly of a coordination architecture takes place, for instance, in the formation of a double and a triple helicate side by side [double selection of both ligand strand and metal ion (8)], multicomponent cages [from two different ligands (21, 22)], receptor cages under guest inclusion (10, 23–25) of diimine macrocycles (26), combinatorial Schiff-base ligands for ion extraction (27), etc.

Concept: Design of the System

To explore processes of self-organization by selection further, we describe here a two-level (organic molecular and inorganic supramolecular) self-assembly that is driven by the formation of a specific coordination architecture and demonstrates dynamic selection from a set of ligand components, exchange of components within the final architecture, as well as reversible medium control of the assembly reaction.

This process involves the generation of a [2 × 2] grid-type array of metal ions M of octahedral coordination geometry, incorporating ligands, that present two tridentate coordination sites formed from three components that are linked by formation of reversible imine connections. Octahedral [2 × 2] grid-type architectures based on predesigned ligands were investigated extensively in our laboratory and exhibit remarkable electronic and magnetic properties (27–29). The present species 1(M) (Fig. 1) incorporate ligands 2 containing two coordination subunits [of a type described earlier (30)] built from a central pyrimidine bridging unit connected via imine bonds to two lateral phenoxy groups (see Scheme 1).

Fig. 1.

Two-component, two-level self-assembly of the [2 × 2] grid complexes 1a(M) in the case M = Zn.

Scheme 1.

Structures of the ligands 2 a–c.

The suitability of phenoxy units for the generation of [2 × 2] grid-type structures is substantiated by earlier results on [2 × 2] grids incorporating such units (31) and related ones (32) as bridging groups. It has also been shown that the imine functionality formed from carbonyl and amine groups is a good candidate for the dynamic connection of components. It has been implemented in the generation of an inhibitor for carbonic anhydrase (33) as well as in a two-level dynamic system (34) based on dynamic combinatorial libraries of aldehydes and amines. It is easily and rapidly reversible but may be more or less complete in aqueous solution depending on the condensing partners (C. Godoy-Alcantar and J.-M.L., unpublished work). In addition, it provides a nitrogen site to which metal ions may bind and thus drive imine formation to completion.

Materials and Methods

NMR Spectrometric Measurements. Compound 3, 2-phenylpyrimidine-4,6-dicarboxaldehyde, was prepared as described (35). Compound 4, 3-amino-4-hydroxybenzenesulfonic acid-4-hydroxybenzenesulfonic acid (AHBS), was obtained commercially and recrystallized from hot water before use. All other compounds were obtained commercially and used without further purification.

Synthesis and Isolation of the [2 × 2] Grid Complexes 1a(Zn) and 1a(Co). The procedure described here for 1a(Zn) is representative for all sulfonate and carboxylate grids and has resulted in isolated yields >90% in all cases. In a typical experiment, solid 4 monohydrate (86.2 mg, 0.416 mmol), 3 (44.2 mg, 0.208 mmol), and zinc oxide (16.9 mg, 0.208 mmol) were loaded into a 10-ml centrifuge tube together with a small Teflon-coated stir bar. The addition of water (2 ml) to this mixture immediately caused a very dark-blue color to develop. The reaction was stirred for 2 h at room temperature, after which a solution of sodium bicarbonate (35.0 mg, 0.416 mmol) in water (1 ml) was added, causing CO₂ evolution. The reaction was stirred for 1 h longer and then filtered through a plug of glass wool in a Pasteur pipette into a 10-ml centrifuge tube. Dioxane (7 ml) then was added, and the reaction mixture was shaken and allowed to stand for 5 min. The solid product that had precipitated was separated by centrifugation and dried overnight under dynamic vacuum over P₄O₁₀, giving 130 mg of grid 1a(Zn) (as the ocatasodium salt) (yield 96%). ¹H NMR (300 MHz, ²H₂O, 298 K, referenced to 2-methyl-2-propanol at 1.24 ppm): δ = 5.78 (br m, 1 H, pyrimidine phenyl), 6.27 (br m, 1 H, pyrimidine phenyl), 6.38 (br m, 1 H, pyrimidine phenyl), 6.51 (d, J = 9 Hz, 2 H, AHBS H5), 7.31 (br m, 1 H, pyrimidine phenyl), 7.52 (dd, J = 9, 2 Hz, 2 H, AHBS H6), 7.69 (d, J = 2 Hz, 2 H, AHBS H2), 7.78 (t, J = 7.5 Hz, 1 H, pyrimidine phenyl H4), 8.19 (s, 1 H, pyrimidine), and 8.50 (s, 2 H, imine). ¹³C{¹H} NMR (75 MHz, ²H₂O, 298 K, referenced to the methyl groups of 2-methyl-2-propanol at 30.29 ppm): δ = 117.33, 121.56, 123.35, 123.64, 128.50, 128.64, 129.61, 130.63, 129.61, 130.63, 132.81, 136.79, 143.15, 159.16, 167.21, and 167.46. Electrospray ionization MS m/z 1,299 . Anal. Calcd for C₉₆H₈₈N₂₄O₃₂S₈Zn₄: ^•10(H₂O): C, 40.52; H, 2.33; N, 7.88. Found: C, 40.12; H, 2.50; N, 7.72. The corresponding cobalt(II) [2 × 2] grid 1a(Co) was prepared in a similar fashion by using CoCO₃ (yield 91%) and characterized accordingly.

NMR Studies of Grid Complexes 1a(M). A 100 mM acetate buffer was prepared by the addition of sodium acetate (106 mg, 1.29 mmol) and acetic acid (40.0 mg, 0.667 mmol) to a 20-ml volumetric flask and adding sufficient deuterium oxide to obtain 20 ml of solution. The measured p²H (pD) of this solution was 5.03. To make the ligand component solution, 3 (10.6 mg, 50.0 μmol) and 4 monohydrate (20.7 mg, 100 μmol) were dissolved in a sufficient amount of this same buffer to make up 5 ml of solution, which thus was 20 mM in sulfonic acid and 10 mM in dialdehyde. Metal salt solutions were prepared by the addition of the quantity of metal(II) acetate required to give a 20 mM solution to a measured quantity of acetate buffer. Experiments thus were carried out by adding 0.25 ml of metal(II) salt solution (5 μmol) to 0.5 ml of ligand component solution (5 μmol of 3 and 10 μmol of 4) in an NMR tube.

NMR Studies of the Formation of 1a(Zn) by Selection from a Combinatorial Library of Ligands. To 0.5 ml of the buffered ligand component solution (containing 5 μmol of 3 and 10 μmol of 4) described above was added 4-hydroxybenzaldehyde (1.25 mg, 10 μmol), 2-acetylpyridine (1.2 mg, 10 μmol), 3-amino-4-hydroxybenzenesulfonic acid-4-methoxybenzenesulfonic acid (2.03 mg, 10 μmol), 3-amino-4-hydroxybenzenesulfonic acid-5-hydroxybenzoic acid (1.53 mg, 10 μmol), 4-aminobenzenesulfonic acid (1.73 mg, 10 μmol), and 2-aminoethanesulfonic acid (1.25 mg, 10 μmol). An additional 0.5 ml of acetate buffer was added to this mixture to give a clear, faintly yellow solution. Proton NMR revealed only one very small peak in the imine region at 8.41 ppm, the integrated intensity of which corresponded to ≈1% of the total carbonyl-containing product. The addition of 0.25 ml of buffered zinc(II) acetate solution (5 μmol) caused a dark-blue color to develop instantly. Immediate (within 1 min) examination of the proton NMR spectrum revealed that the signals corresponding to 3 and 4 had diminished in intensity and that additional, broad peaks had grown in. The signals corresponding to the other components of the virtual combinatorial library (VCL) had not changed. Within 2 h at 25°C the peaks corresponding to 1a(Zn) had appeared, whereas those corresponding to its ligand components had disappeared. No other peaks had changed in integrated intensity.

Crystal Structure Data for Complex 1a(Zn). Crystals of complex 1a(Zn) were grown from an aqueous solution of barium perchlorate. Data were acquired on a NONIUS (Delft, The Netherlands) Kappa charge-coupled device diffractometer. Because of severe disorder among the Ba²⁺ ions, only a poor-quality structure (R ≈ 25%) could be obtained at this stage (A. DeCian, N. Kyritsakas, J.N., and J.-M.L., unpublished work).

Results and Discussion

Two-Component, Two-Level Self-Assembly of the [2 × 2] Grid 1a(Zn). To test the suitability of the system, the self-assembly of the grid 1a(Zn) from the required ligand components 2-phenylpyrimidine-4,6-dicarboxaldehyde 3 and AHBS 4 and from metal ions was first investigated. Because imine formation between these two ligand components is expected to generate a tridentate (O, N, N) binding subunit, metal ions of octahedral coordination geometry are required.

Indeed, mixing an aqueous solution containing 1 equivalent of 3 (10 mM) and 2 equivalents of 4 (20 mM) with an aqueous solution of 1 equivalent of Zn(OAc)₂ (both solutions buffered at pD ≈ 5) gives quantitatively the grid complex 1a(Zn) within ≈1 h. (Fig. 1). At this pD value, imine formation from 3 and 4 in the absence of metal ions was found to be ≤4% by ¹H NMR. The self-assembly process may be followed by ¹H NMR spectroscopy, showing the initial formation of a mixture of presently unidentified compounds, and ending up within ≈2 h in the generation of 1a(Zn) as the sole product, with no other species being detectable. Nuclear Overhauser effect spectroscopy experiments display an Overhauser effect between the imine protons and the internal protons on the terminal aminophenol rings, in agreement with the formation of the bis-tridentate ligand in 1a(Zn). A partially resolved crystal structure of the grid 1a(Zn) (recrystallized from an aqueous solution of barium perchlorate) clearly demonstrates the connectivity in agreement with the [2 × 2] grid-type species 1a(Zn), but severe disorder among the barium cations linking grids together through the sulfonate groups precluded successful refinement of the structure. A representation of the structure is shown in Fig. 2; no bond lengths are given due to the poor quality of the model. Cobalt(II) and iron(II) salts produce the corresponding tetra-cobalt(II) and tetra-iron(II) grids, 1a(Co) and 1a(Fe), in similar fashion. The ¹H NMR spectrum of both complexes are strongly paramagnetically shifted, indicating in particular that 1a(Fe) contains iron centers in the high-spin state.

Fig. 2.

Schematic representation of a poorly resolved x-ray structure of the [2 × 2] grid complex 1a(Zn).

Furthermore, the [2 × 2] grid complexes of cadmium(II) and mercury(II), 1a(Cd) and 1a(Hg), are also obtained as sole products, displaying the expected ¹H NMR spectra with strong coupling of the metal nuclei to the imine protons [J(¹¹³Cd, ¹H) = 25 Hz; J(¹⁹⁹Hg, ¹H) = 127 Hz]. The formation kinetics of 1a(Hg), 1a(Cd), and 1a(Zn) are remarkably different with half-lives of the order of seconds for mercury(II), ≈2 min for cadmium(II), and ≈30 min for zinc(II).

1a(Zn) presents a deep-blue color with two very strong absorptions at λ_max (nm), ε(M^–1·cm^–1), 367 (5.1 × 10⁴) and 587 (7.8 × 10⁴), resulting from the highly delocalized deprotonated ligand. The corresponding bands for 1a(Co) are 367 (5.1 × 10⁴) and 572 (6.2 × 10⁴).

The formation of the [2 × 2] grid complexes 1a(M) represents a two-level self-organization process involving ligand formation through imine condensation between components 3 and 4, driven to completion by metal ion binding and subsequent assembly into the grid architecture. The build-up of 1a may involve several mechanisms such as (i) the initial formation of a mononuclear center, followed by growth into a “corner-type” unit, two of which could combine into a grid, by analogy to a process investigated earlier (36); (ii) preformation of a very small amount of the full ligand followed by assembly into a grid; or (iii) a combination of such steps. Indeed, several unidentified species are observed by ¹H NMR in the course of the assembly of 1a(Zn), as indicated above.

Selection from a VCL of Ligands. A dynamic combinatorial library is in its generality a VCL because its constituents need not be “real,” i.e., need not exist as preformed species, before selection is performed but must simply be potentially accessible by the continuously reversible connection between their components (7, 19). Selection within such a VCL represents a process of much conceptual significance, because it amounts to the emergence of a specific entity from a complex mixture under the pressure of the structural/information and interaction/processing features of the system. It represents a self-organization by selection, extending from the self-recognition of preformed ligands occurring in helicate formation (8) to the self-construction of the ligands themselves by the correct choice of the required components among all those present in the “instructed mixture” (8) in a sort of “self-design” (7).

Thus, when a zinc(II) salt is added to the VCL generated from a set of amines and aldehydes in aqueous solution at pD = 5.0 (see Materials and Methods), the [2 × 2] grid 1a(Zn) is formed quantitatively and exclusively through the selection of the required dialdehyde 3 and aminophenol 4 components driven by metal ion coordination (Fig. 3). At this pD and 10 mM concentration, <4% monoimine and no trace of bis-imine condensation products can be detected by ¹H NMR spectroscopy before addition of the metal ions to the mixture, which thus presents a genuinely “virtual” character. At higher pH, slightly larger amounts of condensation products are observed (8.0% monoimine and 0.7% bis-imine at pD 9.8).

Fig. 3.

Self-organization of the [2 × 2] grid 1a(Zn) by dynamic selection from a VCL of ligands.

Similarly, the correct 3-amino-4-hydroxybenzenesulfonic acid-4-hydroxybenzoic acid component is quantitatively selected after addition of 1 equivalent of an aqueous Zn(OAc)₂ solution (20 mM) to an equimolar mixture of this compound with its isomer 2-hydroxy-5-aminobenzoic acid (both 20 mM) and 3 (10 mM), yielding exclusively the expected grid complex 1b(Zn) of ligand 2b.

These processes may be considered to display a very high amplification with total selection, because no bis-imine condensation product can be detected in absence of zinc(II), and only 1a,b(Zn) are formed. Assuming that the ¹H NMR detection limit is <0.1 mM under the experimental conditions used, the zinc(II)-templated formation of bis-imine ligand from the 10 mM ligand component solution (which contains no detectable bis-imine at pD = 5) must display amplification and selection factors >100 and >99%, respectively.

As well as being able to select the correct ligand components, the present system is able to select the metal ions required to form grid complexes of type 1, as is demonstrated in the case of zinc(II)/lead(II). When 0.25 ml of a Pb(NO₃)₂ solution (20 mM) in acetate buffer (pD = 5) is added to 0.5 ml of buffered solution (pD = 5) of 3 (10 mM) and 4 (20 mM), a red precipitate is observed to form over the course of hours with concomitant disappearance of ¹H NMR signals corresponding to 3 and 4. If this experiment is conducted by using a Zn(OAc)₂ (20 mM)/Pb(NO₃)₂ (20 mM) mixture in place of the Pb(NO₂) solution, the formation of grid 1a(Zn) is favored over the formation of the red precipitate; ¹H NMR signals corresponding to 1a(Zn) grow in >2 h. These signals do not diminish in intensity in the course of 2 days at 25°C, indicating that this grid complex is stable to the presence of lead(II) over this time scale. Monovalent cations such as Na⁺, K⁺, and Tl⁺ likewise do not interfere with the self-assembly of 1a(Zn).

Theoretically, different cyclic oligomers might form besides 1a(Zn), as in the case of circular helicates (14, 19, 20), and in various cyclic inorganic (37–39) or organic (40) oligomers. None are detectable. Their absence may be attributed to the high geometrical definition and structural rigidity of the [2 × 2] grid complexes.

Dynamic Component Exchange in the [2 × 2] Grid 1a(Zn). Whereas under basic conditions (pD >10) the grid complex 1a(Zn) is stable in the presence of competing ligand components, in acidic medium exchange of components takes place, allowing the introduction of a different aminophenol. Thus, the preformed complex 1c(Zn) undergoes exchange of its 3-amino-4-hydroxybenzenesulfonic acid-4-hydroxy-5-nitrobenzene sulfonate (AHNBS) units at a pD 5.3 of after the addition of a 10-fold excess of 4 containing no nitro group (Fig. 4 Right). New ¹H NMR signals, which likely correspond to grid structures containing both nitrated and nonnitrated aminophenols, appear within 24 h of addition. After heating to 40°C for 48 h, complete conversion is observed, with only signals corresponding to 1a(Zn) and free AHNBS being observed. These signals are slightly broader than those observed for pure 1a(Zn), suggesting that a small amount of nitroaminophenol remains incorporated.

Fig. 4.

Dynamic processes of [2 × 2] grid complexes: protonic modulation of the breakdown and reconstruction of 1a(Zn) (Left) and dynamic exchange of its ligand components (Right) to give 1c(Zn).

Protonic Control of the Self-Assembly. The imine connections being easily dissociated, medium control of assembly/disassembly of grids 1(M) requires weakening of the metal ion coordination. This may be achieved by protonation of the phenolate sites followed by protonation of the amino groups resulting from imine disconnection. Indeed, the addition of acid to an aqueous solution of grid complex 1a(Zn) results in its breakdown to give a mixture of the hydrated dialdehyde 3 and the protonated aminophenol 4, which on basification regenerates the initial grid. This cycle represents a reversible pH control of the dynamic breakdown and reconstruction of 1a(Zn) (Fig. 4 Left). The fraction of 1a(Zn) as a function of pD is shown in Fig. 5; the equivalence point (50% complex remaining) is found at pD = 3.2. On the other hand, with 1c(Zn), the equivalence point is reached already at pD = 2.5. This difference may be attributed to the effect of the nitro substitution lowering the basicity of the phenolate group and suggests that it may be possible to achieve pD-dependent control of the components preferentially incorporated into the ligand. Indeed, when a mixture of AHBS (30 mM)/AHNBS (30 mM)/dialdehyde 3 (15 mM)/Zn(OAc)₂ (15 mM) is titrated with NaOD in ²H₂O solution, analysis of the ¹H NMR spectra gives a ratio of components incorporated into the grid structure of AHBS/AHNBS changing from ≈0.65 at pD = 3.37 to ≈1.3 at pD = 7.8. This effect represents a protonic modulation of the selection process. It also illustrates the ability of dynamic systems to respond and undergo adaptation to environmental conditions, expressed here by changes in acidity of the medium.

Fig. 5.

Distribution of the [2 × 2] grid species 1a(Zn) (○) and its components (•) as a function of the acidity of the medium (pD).

Conclusion

The present results demonstrate two-level self-assembly with component selection. A further step toward processes of increasing complexity could involve side-by-side formation of two or more self-assembled entities by correct selection of their specific components from an extended dynamic library containing a wide range of components undergoing reversible interconnection. Beyond single-level systems, one may envisage multilevel sequential self-assembly processes where the generation of a given entity sets the stage for the self-assembly of the next one with selection of specific partners at each level. Such developments extend further the implementation of constitutional dynamic chemistry toward self-organization by selection.