Allosterically Regulated Guest Binding Determines Framework Symmetry for an Fe^II ₄L₄ Cage

Abstract

Self‐assembly of a flexible tritopic aniline and 3‐substituted 2‐formylpyridine subcomponents around iron(II) templates gave rise to a low‐spin Fe^II ₄L₄ capsule, whereas a high‐spin Fe^II ₃L₂ sandwich species formed when a sterically hindered 6‐methyl‐2‐formylpyridine was used. The Fe^II ₄L₄ cage adopted a new structure type with S ₄ symmetry, having two mer‐Δ and two mer‐Ʌ metal vertices, as confirmed by NMR and X‐ray crystallographic analysis. The flexibility of the face‐capping ligand endows the resulting Fe^II ₄L₄ framework with conformational plasticity, enabling it to adapt structurally from S ₄ to T or C ₃ symmetry upon guest binding. The cage also displayed negative allosteric cooperativity in simultaneously binding different guests within its cavity and at the apertures between its faces.

Metal‐organic cages are an important class of synthetic hosts, which are able to mimic the structural adaptability and allosteric regulation of naturally occurring systems. A low‐symmetry Fe^II ₄L₄ tetrahedral cage reported here was able to adapt its geometry from S ₄ to C ₃ or T symmetry upon guest binding; multiple binding sites allowed the host to simultaneously bind distinct guests with negative allostery.

Biological receptors can dynamically adapt to optimize binding affinity, thus enhancing or inhibiting signal transduction in living systems. ^[1] Allosteric regulation, whereby biological receptors transmit the effect of binding a substrate at one site to another at a distant site, is an essential process by which natural systems process information. ^[2] The development of artificial allosteric systems, which are capable of emulating the biologically occurring processes of activity regulation, can enable new applications, ^[3] but these systems are challenging to design. ^[4]

Metal‐organic cages formed by coordination‐driven self‐assembly, with well‐defined cavities and binding sites, have emerged as a powerful platform for the modular design of biomimetic supramolecular systems. These cages have also proven useful across diverse areas, including molecular recognition and sensing, ^[5] chemical separation, ^[6] stabilization of otherwise unstable species, ^[7] and catalytic transformations. ^[8] The reversible linkages between metal vertices and coordination sites can enable these cages to disassemble and reassemble in response to external stimuli, ^[9] for instance, temperature, ^[10] light, ^[11] redox, ^[12] and pH. ^[13]

Most cages contain symmetric and rigid ligands, which produce high‐symmetry structures resembling Platonic and Archimedean solids, such as tetrahedra, ^[14] octahedra, ^[15] cubes, ^[16] cuboctahedra ^[17] and other higher‐order structures. ^[18] The high symmetry of such synthetic hosts differentiates them from biological systems, as natural receptors are rarely isotropic, highly symmetric species. ^[19] To this end, the rational design of low‐symmetry host systems with inherent conformational adaptability is needed, in order to study intricate binding behaviors in adaptable chemical systems, which may led to the development of bioinspired applications.

Here we present the synthesis of a S ₄‐symmetric Fe^II ₄L₄ cage 1 by subcomponent self‐assembly (Figure 1). This cage dynamically adapted upon guest encapsulation to form host–guest complexes with either lower (C ₃) or higher (T) symmetry. The cage also simultaneously bound different guests at two distinct sites, with negative allosteric modulation observed between them.

Figure 1.

a) Subcomponent self‐assembly of Fe^II ₄L₄ capsule 1 and Fe^II ₃L₂ sandwich 2 from A, Fe^II(NTf₂)₂ and 2‐formylpyridines B and C, respectively. All reactions were performed at 343 K in CD₃CN. b) Partial ¹H NMR and DOSY spectra of S ₄‐symmetric 1 (400 MHz, CD₃CN, 298 K), showing three sets of magnetically‐distinct peaks.

Tritopic aniline subcomponent A, with a 1,3,5‐tris(2‐furyl)‐2,4,6‐trimethoxybenzene central core, was synthesized from commercially available 1,3,5‐trimethoxybenzene over four steps (Figure S1). Steric clash between the furan rings and methoxy groups engenders a non‐planar conformation of the central core. The three‐dimensionality and structural flexibility of the core of A were elegantly exploited by Fujita, Takezawa et al. in a knotted cage prepared from a related ligand. ^[20]

The reaction of trianiline A (4 equiv) and 3‐nitro‐2‐formylpyridine B (12 equiv) with iron(II) bis(trifluoromethanesulfonyl)imide (Fe^II(NTf₂)₂, 4 equiv) in CD₃CN at 343 K for 2 h produced cage 1 (Figure 1a), with its Fe^II ₄L₄ composition confirmed by electrospray ionization mass spectrometry (ESI‐MS, Figure S11). The ¹H NMR spectrum of 1 exhibited three sets of magnetically‐distinct proton signals, with all signals displaying the same diffusion coefficient in the ¹H diffusion‐ordered spectroscopy (DOSY) spectrum (Figures 1b and S6). ¹H‐¹H nuclear Overhauser effect (NOE) correlations between protons of pyridyl and phenylene rings were consistent with low‐symmetry meridional (mer) configurations at the iron(II) centers (Figure S8). ^[21] We therefore inferred Fe^II ₄L₄ cage 1 to possess S ₄ symmetry, ^[22] in which all iron(II) centers adopt a mer stereochemical configuration, with two exhibiting Δ handedness, and the other two Λ.

In place of subcomponent B, other 2‐formylpyridine derivatives were also attempted. Changing the substituent at the 3‐position of the 2‐formylpyridine did not affect the stereochemical outcome during self‐assembly, with the formation of a series of Fe^II ₄L₄ cages (3–5) possessing S ₄ symmetry (Figures S14–S23).

Intriguingly, high‐spin complex 2 was formed by the reaction of 6‐methyl‐2‐formylpyridine C with A and Fe^II(NTf₂)₂ (Figure 1a), as confirmed by the wide‐sweep ¹H NMR spectrum, with signals in the range −42 to 208 ppm (Figure S12). The ESI‐MS spectrum confirmed the Fe^II ₃L₂ composition of 2 (Figure S13). In accordance with the 18‐electron rule, ^[23] we inferred that each iron(II) vertex of 2 might be surrounded by two bidentate chelating units and two extra solvent molecules, with the high‐spin character of 2 being a consequence of steric clash between methyl groups and the adjacent pyridyl rings around the iron(II) centers, as observed previously. ^[24] By elongating the Fe^II−N bonds, such steric repulsion also results in the destabilization of Fe^II ₃L₂ 2 relative to Fe^II ₄L₄ 1 species. The formation of shorter, stronger Fe^II−N bonds in 1 was thus inferred to drive subcomponent exchange during the conversion of 2 to 1, following the addition of 2‐formylpyridine B (Figures 1a and S24).

Although many attempts to grow single crystals of 1 were unsuccessful, X‐ray quality crystals of Fe^II ₄L₄ analog 3, constructed from 3‐bromo‐2‐formylpyridine, were obtained by slow vapor diffusion of diethyl ether into an acetonitrile solution of 3 (Figure 2a). The solid‐state structure of 3 adopts idealized S ₄ symmetry, consistent with the solution‐state NMR data (Figures 2b and S15).

Figure 2.

Different views of the crystal structure of Fe^II ₄L₄ cage 3, assembled from 3‐bromo‐2‐formylpyridine, A and Fe^II(NTf₂)₂. a) Front views from ligand walls, with clockwise (type I) and anticlockwise (type II) oriented ligands colored brown and green, respectively. b) Front view and view down the pseudo‐S ₄ axis, with the ligands shown in space‐filling mode and the three magnetically‐distinct environments shown in blue, red, and black. Disorder, anions, solvents and hydrogen atoms are omitted for clarity.

The structure of 3 consists of four face‐capping ligands, with two ligands adopting a clockwise orientation and the other two anticlockwise. These ligands bridge four iron(II) vertices, two mer‐Δ and two mer‐Ʌ. The mean Fe^II⋅⋅⋅Fe^II distance along the four edges between Δ and Ʌ metal centers is 18.1±0.3 Å; the Fe^II⋅⋅⋅Fe^II distance between the two mer‐Λ metal centers (15.4 Å) is slightly longer than between the two mer‐Δ metal centers (14.4 Å). A cavity volume of 986 ų for 3 was calculated using the MoloVol program. ^[25]

In the structure of 3, the three dihedral angles between the central phenyl ring and the furan rings within the same clockwise face‐capping ligand were observed to differ (40.1°, 42.8°, 44.5°), as were the three torsion angles (52.2°, 58.2°, 62.2°) between the phenylene rings and the N−Fe^II−N chelate planes around each mer‐Ʌ metal vertex. These differing angles thus reflect the structural flexibility and plasticity of the Fe^II ₄L₄ framework to adopt different ligand configurations as required to minimize the energy of the system.

Encouraged by the flexible nature of its S ₄‐symmetric Fe^II ₄L₄ framework and enclosed cavity, we next investigated the structural and stereochemical adaptability of 1 upon guest binding. Cage 1 was observed to bind both neutral and anionic guests (Table 1), with binding affinities quantified by ¹H NMR titrations, and binding stoichiometries gauged by NMR or ESI‐MS, as detailed in Supporting Information Section 5.

Table 1.

Host‐Guest Properties of 1.

Guest[a]	K a [M−1][b]	Host Symmetry[c]
G1	(1.82±0.05)×102	S 4
G2	(1.21±0.05)×102	S 4
G3 [d]	(2.33±0.10)×103	T
G4	(1.17±0.15)×105	C 3

[a] ¹H NMR titrations were performed by the addition of aliquots of a CD₃CN solution of G1 or G2 (50 mM) into a stock solution of 1 in CD₃CN (0.5 mM); ¹H NMR titrations were performed by portionwise addition of G3 or G4 as a solid into a stock solution of 1 in CD₃CN (0.75 mM) using tert‐butyl acetate as an internal standard. [b] Binding constant determined by ¹H NMR titration. [c] Host point symmetry inferred from NMR spectroscopy. [d] The binding constant for PP‐CRY (G3′) of K _a=(2.40±0.10)×10³ M⁻¹ was identical within error.

Upon addition of tetraphenylborate G1 or G2, the proton signals of 1 were observed to shift, in line with fast guest exchange on the NMR time scale (Figures S25 and S29). The signals of the cage furan and phenylene rings as well as the OMe groups at the apertures of 1 shifted most. In addition, when the central cavity of 1 (ca. 986 ų) was occupied by G3 (596 ų) or G4 (471 ų), 1 was observed to bind G1 (321 ų) simultaneously, as shown in Figure 4. These observations led us to infer that G1 and G2 are bound peripherally at the edges of 1. ^[17c] The exact binding stoichiometries could not be gauged using ¹H NMR titrations (Figures S27 and S31), which fitted slightly, but not overwhelmingly better to a 1 : 2 model using BindFit. ^[26] Chemical shift changes for 1 were plotted and fitted to a Hill function (Figures S28 and S32), ^[27] with apparent association constants determined to be (1.82±0.05)×10² M⁻¹ and (1.21±0.05)×10² M⁻¹ for G1 and G2, respectively. In both cases, the Hill coefficients were approximately 1, indicating non‐cooperative binding of tetraphenylborates to 1. Putative peripheral binding of tetraphenylborates during the titration process did not alter the S ₄ symmetry of the framework of 1 (Figure 3b).

Figure 3.

a) Schematic diagram showing the symmetry and stereochemical adaptation of 1 driven by guest binding. b) Partial ¹H NMR spectra of the imine region of 1 and its host–guest complexes (400 MHz, CD₃CN, 298 K), with imine peaks marked by stars. For G3⊂1 and G4⊂1, red stars correspond to the major host–guest complex diastereomer (G3⊂ΔΔΔΔ‐1 or G4⊂ΔΔΔΛ‐1), whereas blue stars correspond to the minor diastereomer with an enantiomorphous host configuration. c) CD spectra of host–guest complexes in acetonitrile at equal concentrations.

By contrast, enantiopure guest G3, MM‐cryptophane (CRY),[ ²⁵ , ²⁸ ] with a calculated volume of 596 ų was observed to bind centrally within the cavity of 1 in slow exchange on the NMR time scale, with a binding constant of (2.33±0.10)×10³ M⁻¹ (Figures 3a and S33). During titration experiments, proton signals corresponding to G3⊂1 emerged as free 1 disappeared from the ¹H NMR spectrum. NMR integration revealed the formation of a 1 : 1 host–guest complex. All proton signals from the bound guest shifted significantly upfield as a result of shielding effects, consistent with central binding of G3 (Figure S35). Interestingly, cage 1 was not observed to encapsulate C₆₀ (549 ų), which we attribute to a poor fit between the bumpy inner surface of the host and the smooth curvature of the guest, precluding effective stacking interactions between them.

The ¹H NMR spectrum of G3⊂1 was consistent with the formation of a T‐symmetric species (Figures 3b and S35), with ligands in a threefold‐symmetric environment and all metal centers adopting facial (fac) stereochemical configurations with the same handedness, as observed in high‐symmetry face‐capped tetrahedra. ^[29] The presence of fac stereochemistry was also confirmed by the absence of characteristic NOE correlations between protons on the pyridyl and phenylene rings (Figure S38).

Two groups of proton peaks in a ratio of 4.8 : 1 were observed in the ¹H NMR spectrum, indicating that G3⊂1 existed as a pair of diastereomers, MM–CRY⊂ΔΔΔΔ‐1 and MM–CRY⊂ΛΛΛΛ‐1. The circular dichroism (CD) spectrum of G3⊂1 displayed strong Cotton effects in the ranges 280–400 nm and 410–680 nm, assigned to π–π* and metal‐to‐ligand charge transfer (MLCT) transitions, respectively (Figure 3c). The negative sign of the MLCT bands correlates with Δ handedness, ^[30] suggesting that the major diastereomer of MM–CRY⊂1 has four fac‐Δ metal vertices. Cage 1 also bound the enantiomer PP‐CRY to yield PP‐CRY⊂1 possessing four fac‐Ʌ metal centers as the major diastereomer, with an identical‐within‐error binding affinity of K _a=(2.40±0.10)×10³ M⁻¹ (Figures S34 and S40).

Host‐guest investigations also revealed that 1 was able to accommodate an equimolar amount of Δ‐tris(tetrachloro‐1,2‐benzenediolato)phosphate (Δ‐TRISPHAT, G4) ^[31] with a binding affinity of K _a=(1.17±0.15)×10⁵ M⁻¹. When more than one equivalent of Δ‐TRISPHAT was added during NMR titrations, no further changes were observed in ¹H and ³¹P NMR spectra (Figures S44 and S45), consistent with the formation of a 1 : 1 host–guest complex G4⊂1, as was also observed by ESI‐MS (Figure S51). The ¹H NMR spectrum had two groups of signals, with each group containing four sets of signals (Figures 3b and S46), consistent with the presence of two C ₃‐symmetric diastereomers of G4⊂1 in a 4.1 : 1 ratio. NOE correlations provided evidence for a 1 : 3 fac:mer configuration of metal centers for the major diastereomer (Figure S49), while the MLCT bands in the CD spectrum indicated an excess of Δ stereochemistry within G4⊂1 (Figure 3c). We thus inferred that the major diastereomer of G4⊂1 with C ₃ symmetry contains one fac‐Ʌ and three mer‐Δ metal centers, as in the elegant “sorting hat” structure reported by Hooley and co‐workers. ^[32]

The conversion of S ₄‐symmetric 1 into T‐ and C ₃‐symmetric host–guest complexes did not occur at low temperatures, requiring to be heated at 343 K for 2 h to complete these transformations. The high temperatures required to overcome the energy barriers for this symmetry breaking and rearranging suggested a mechanism involving extensive disassembly and reassembly. ^[33]

Having identified that 1 has two distinct binding sites that can bind different guests, we then explored allosteric effects by simultaneously treating 1 with two different guests (Figure 4). To consider cooperative effects, we determined the factor α=K′_a/K _a to quantify allosteric regulation, where K′_a and K _a are the binding constants of a guest bound by the host–guest complex and the empty host, respectively. A value of α>1 indicates positive cooperativity, whereas α<1 suggests negative cooperativity.[ ² , ^26a ]

Figure 4.

Schematic showing the allosteric effects and cooperative binding.

Both T‐symmetric G3⊂1 and C ₃‐symmetric G4⊂1 maintained their symmetries upon binding tetraphenylborate in fast exchange on the NMR time scale (Figure 4). Negative cooperativity (α=0.59 and 0.74) was observed in both cases (Figures S52–S57). When 1 bound BPh₄ ⁻ peripherally, the central binding of both G3 and G4 was likewise inhibited. Negative cooperative effects (α=0.60 and 0.81) were observed on the binding of these two guests, which gave rise to T‐symmetric G3 ⋅ (G1)_x ⊂1 and C ₃‐symmetric G4 ⋅ (G1)_x ⊂1, respectively (Figures S58 and S59). This negative cooperativity was inferred to be a consequence of changes in aperture size and cavity volumes that took place in order to optimize binding, but which do not favor the binding of two guests at once. The binding of G3 or G4 within the cavity of 1 may also physically block G1 from reaching inside the windows to bind. The presence of an electron‐rich anionic or neutral guests within electron‐deficient 1 may also weaken the binding of further electron‐rich guests by electrostatic repulsion.

The higher binding affinity of G4 over G3 allowed for guest displacement, thus enabling the conversion of the T‐symmetric cage framework into the C ₃‐symmetric one (Figures S60 and S61). The most strongly‐binding guest in the system thus dictated the framework symmetry.

The ability of 1 to adopt three distinct diastereomeric conformations in order to optimize guest binding, including the singular S ₄ framework, with two mer‐Δ and two mer‐Ʌ metal vertices, thus complements and builds usefully upon previous studies of cage stereochemistry. Its two allosterically active sites enabled the simultaneous binding of two distinct guests, peripherally and centrally. Future work may enable such allosteric effects to control reactions that are catalyzed within the cage cavity or at a peripheral site, potentially allowing up‐ or down‐regulation of catalysis in biomimetic fashion.

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

Supporting Information

Supporting Information

Xue W., Wu K., Ouyang N., Brotin T., Nitschke J. R., Angew. Chem. Int. Ed. 2023, 62, e202301319; Angew. Chem. 2023, 135, e202301319.

Data Availability Statement

The data that support the findings of this study are available in the Supporting Information of this article.