There is much interest in the design of Ru(Ⅱ) arene complexes as catalysts and anticancer complexes [1–7]. The recognition of substrates and target sites can be tailored by appropriate choice of the ligands. In complexes of the type [(η⁶-arene)Ru(N, N)X], the presence of a labile ligand X, e.g., Cl, allows monofunctional targeting of G bases on DNA [8,9]. With a sulfonyl group on a chelated N, the 6^th coordination site becomes even more labile and a hydride acceptor/donor site in transfer hydrogenation catalysis [10,11]. In both applications, outer-sphere interactions play important roles in recognition. The NH protons on the chelated diamine H-bond with C6O of G [12–14], and bulky N-phenyl rings control asymmetric hydride transfer to substrates along with interactions with the arene ring [15,16]. In DNA recognition, extended arenes (e.g. biphenyl) can intercalate and stack with DNA bases [17,18]. Extension to di-Ru(Ⅱ) complexes bridged by a flexible linker gives rise to site-specific induced-fit recognition of DNA, cross-linking between distant DNA bases, and potential protein targets [3,19,20].

Bridged multinuclear complexes offer the design of 3D architectures such as rectangles, prisms and cubes, with cavities, a variety of accessible configurations and more sophisticated recognition sites, especially involving the bridging ligands [21–23]. Stang et al., for example, have reported that tetranuclear rectangular and hexanuclear trigonal prismatic metallacycles and cages containing dioxido-naphthoquinonato spacers and bis-imidazole bridging ligands self-assemble into single symmetrical and stable metallacycles and cages despite the possibility of forming isomeric conformers [24–26]. Here we use two flexible 1,3,5-trimethyl-2,4-di(imidazole-1-ylmethyl)benzene (m-bitmb) ligands [27], containing N-methylene imidazole ligands in the 2,6-positions of a central 1,3,5-trimethylbenzene unit to bridge terminal p-cymene (p-cym) Ru(Ⅱ) chloride fragments (Scheme 1). We have explored the configurational space available to the 24-membered macrocyclic ring by a combination of X-ray crystallography, NMR spectroscopy, HPLC, mass spectrometry, ligand field molecular mechanics (LFMM) and density functional theory (DFT) calculations which elucidated the mechanism of configurational interconversion. We show that both counter anions and solvents play crucial roles in controlling the configuration of the macrocyclic ring. These studies of the configurational dynamics of organo-ruthenium macrocycles add a new dimension to their design features.

Metallamacrocyclic complexes [Ru(η⁶-p-cymene)(µ₂-m-bitmb)Cl]₂·2X, m-bitmb = 1,3,5-trimethyl-2,4-di(imidazole-1-ylmethyl)benzene, containing 5 different counter anions X = Cl^- (1·2Cl), NO₃^- (1·2NO₃), CF₃SO₃^- (1·2CF₃SO₃), PF₆^- (1·2PF₆), or BF₄^- (1·2BF₄) were synthesized. Complex 1·2Cl was obtained by reaction of m-bitmb directly with [RuCl₂(η⁶-p-cym)]₂ in CH₂Cl₂/ethyl acetate, while complex 1·2CF₃SO₃ was formed by a two-step reaction. [RuCl(η⁶-p-cym)]₂²⁺, obtained from [RuCl₂(η⁶-p-cymene)]₂ by adding AgCF₃SO₃, was reacted with the m-bitmb ligand. Complexes 1·2PF₆, 1·2NO₃ and 1·2BF₄ were synthesized by counter-anion exchange from 1·2Cl using the corresponding silver salts (AgPF₆, AgNO₃, AgBF₄, Scheme 2a).

Single crystals suitable for X-ray diffraction were obtained for 1·2Cl from CH₂Cl₂/benzene and for 1·2NO₃ and 1·2CF₃SO₃ from CH₂Cl₂/toluene at ambient temperature. Yellow-block and rod-like crystals of 1·2Cl and 1·2NO₃ were each hand-separated from their mother liquids. The block crystals of 1·2Cl·0.5(CH₃COOCH₂CH₃)·7(H₂O), and 1·2NO₃·0.5(C₇H₈) have boat and twist-boat configurations, respectively (Fig. 1, Fig. 2a). The rod-like crystals of 1·2Cl·3(C₆H₆)·2(H₂O), 1·2NO₃·(CH₂Cl₂)·H₂O·(C₇H₈), and 1·2CF₃SO₃·2(CH₂Cl₂) have chair configurations for the macrocycle (Fig. 1, Fig. 2b–d). Crystallographic data, selected bond lengths and angles are in Tables S1–S4 (Supporting information).

In chair-1·2Cl, the two pairs of imidazole rings of the m-bitmb ligands and Ru-Cl bonds face away from each other, all in trans arrangements (Fig. 1c). The chair structures of 1·2Cl, 1·2NO₃ and 1·2CF₃SO₃ are very similar (Figs. 2b–d), with similar Ru1-Ru2 distances (12.78 Å, 12.62 Å, and 12.88 Å, respectively). The macrocyclic structure is further linked through the H-bonds between the counter anions (Cl^-) and backbone, together with solvent water molecules (Table S5 in Supporting information) involving particularly H-2′ protons of the imidazole rings (Fig. 2c).

In boat-1·2Cl, the two pairs of imidazole rings and Ru-Cl bonds face each other in a cis arrangement (Fig. 1b), and the Ru1-Ru2 distance shortens to 10.8 Å. The two p-cymene ligands are twisted from co-planarity by only 4.9°, and in both fragments, the dihedral angles between the imidazole and arene ring are 84.9°. Intermolecular π-π stacking interactions between the two Ru₂L₂ backbones induce formation of a "bottom-to-bottom" dimer, which is further connected by two Cl^- anions via C—H···Cl H-bonds, again involving imidazole H-2 (2.64–2.81 Å; Fig. S1 and Table S5 in Supporting information).

Twist-boat-1·2NO₃ (block crystals) has a distinct configuration (Fig. 2a). The m-bitmb ligands and Ru-Cl bonds still have a cis arrangement, but the central phenyl rings are no longer parallel, with dihedral angles between phenyl rings and imidazole rings of 64.6° and 84.9°, respectively. The Ru1-Ru2 distance shortens to 8.9 Å (c.f. 10.8 Å in boat-1·2Cl). H-bonds are also observed between NO₃^- counter anions, and backbone imidazole rings with distances of 2.47–2.55 Å (Fig. 2a and Table S5).

Thus as illustrated in Fig. 2 and Fig. S1, the non-covalent electrostatic, van de Waals, and H-bonding interactions between the cationic complexes themselves and with the counter anions and solvent of crystallization, as well as the face-to-face π-π interactions between the two boat form Ru₂L₂ metallamacrocycle, can play important roles in the stalilization of these conformers in the solid, crystalline state.

Next we studied the behaviour of these complexes in solution. We dissolved 1·2Cl, 1·2NO₃ and 1·2CF₃SO₃ in 95% H₂O/5% DMSO solutions and after 24 h separated six species by reverse-phase C8 HPLC with CH₃CN/0.1% TFA gradient elution. Remarkably three pairs of peaks were identified by ESI-HRMS (Fig. S2 and Table S6 in Supporting information), assignable as isomers of [Ru(p-cym)(m-bitmb)Cl]₂²⁺ (peaks 4, 5), [Ru(p-cym)(m-bitmb)Cl]₂²⁺ DMSO (peaks 2, 6), and [Ru(p-cym)(m-bitmb)Cl]₂²⁺ 2DMSO (peaks 1, 3), with the help of a combinatorial approach integrating NMR, single-crystal structural analysis and high-performance liquid chromatography (HPLC) methods. These DMSO molecules appear to be bound non-covalently and not direcly coordinated to Ru since MS peaks assignable to Ru-DMSO fragments could be identified and no NMR evidence for direct binding could be obtained. As shown in Fig. S3 (Supporting information), a NMR titration of 1·2Cl with DMSO in DMF-d₇ revealed no new ¹H NMR signals, even after addition of 20 equiv. DMSO, only small upfield shifts of Hd/Hd' protons, suggesting weak non-coordinative interaction, confirming solvation rather than coordination.

The NMR spectra of these 1·2X complexes showed a dependence on solvent, counter anion, and time, Fig. 3 and Figs. S4–S22 (Supporting information). Peaks were assigned by a combination of the correlation between peaks for freshly-dissolved crystals of known configuration, their behavior over time, comparison with spectra of the as-synthesized complexes, and 2D NOESY studies (Fig. 3, Figs. S4 and S15, Tables S7–S11 in Supporting information). For example, the initial single set of peaks for crystals of chair-1·2NO₃ gave rise to a second set over the course of a few hours assignable to the boat configuration (Fig. 3c). Sensitivity to solvent (DMF-d₇, and CD₃CN) of the shifts of the imidazolium (Hd, He and Hf) and central phenyl (Hc) ¹H resonances at low field, is evident, Figs. 3a and b, and S4, Tables S7 and S8 (Supporting information).

Similarly we recorded ¹H NMR spectra of fresh solutions of the other complexes. The initial boat: chair ratios differed from 3:1 for as-synthesized 1·2NO₃, 1·2PF₆ and 1·2BF₄, 2:1 for 1·2Cl, and 1:1 for 1·2CF₃SO₃ (Figs. S5–S8). As can be seen in Fig. S9, over a period of about 120 h, all complexes reached an equilibrium ratio of ca. 1:1, as determined by integration of Hd(boat)/Hd'(chair) peaks, the same as for 1·2CF₃SO₃ which did not change. Treatment of the NMR data in terms of approach-to-equilibrium kinetics (Scheme 2b) gave rate constants for forward (k₁) and reverse (k_-1) reactions of ca. 3-6 µs^-1 (half-lives 32–58 h) in the order PF₆^- > NO₃^- > BF₄^- > Cl^- (Table 1 and Figs. S5–S8 in Supporting information).

As shown in Figs. S14 and S15 (Supporting information), two sets of signals were observed in the 2D [¹H, ¹H] NOESY NMR spectra of 1·2Cl or 1·2CF₃SO₃ one week after dissolution in DMSO‑d₆ solution, indicating the co-existence of both of chair- or boat-configurational isomers in each solution. Near-complete assignments of NOESY NMR spectra of 1·2Cl and 1·2CF₃SO₃ were achieved, details are listed in Tables S9 and S10 (Supporting information). Characteristic NOE cross-peaks allowed identification of the boat or chair isomers in solution (Table S11, Fig. S16 in Supporting information). For example, the intensities of NOE cross-peaks for Hi-Hf, Hf-Hl, He-Ha and He-Hb, or Hf-Hl, Hd-Hi, Hi-Hf, He-Ha, He-Hb and Hf-Hj differ for boat and chair isomers of 1·2Cl or 1·2CF₃SO₃.

Next we carried out NMR titration experiments to determine whether there are specific interactions between anions and the macrocyclic ring as observed in the X-ray crystal structures (Figs. S17–S23 in Supporting information). As seen in Figs. S17–S19 (Supporting information) gradual low field shifts for the imidazole Hd and Hd' protons of both boat and chair forms arose from addition of LiCl to 1·2NO₃, 1·2PF₆ or 1·2CF₃SO₃ in DMF-d₇, amounting to ca. 0.6 ppm with 5 equiv. of LiCl, giving shifts close to those for 1·2Cl, but very small shifts for other peaks. The titration did not reach the saturation point before using 40-fold excess of chloride anions (Figs. S24A-Cb in Supporting information). The most notable change was the chemical shift of the characterized imidazole C—H proton involved in the formation of C—H···Cl^- hydrogen-bonds. A Job's plot analysis showed a maximum change in chemical shift at a mole fraction of 0.333, thus indicating a 1:2 binding stoichiometry. (Figs. S24A-Ca in Supporting information). The apparent Cl^- binding constants (L/mol) were determined using WinEQNMR2 software [28] to be K₁₁ = 131.7 for 1·2NO₃, K₁₁ = 142.8 for 1·2PF₆ and K₁₁ = 508.3 for 1·2CF₃SO₃ (Table S12 in Supporting information). Second Cl^- complexation constants K₁₂ were much smaller, ca. 11 L/mol (Table S12). In the reverse titration of 1·2Cl with LiNO₃, > 200 equiv. were necessary to regenerate the higher field shifts of Hd and Hd' for 1·2NO₃ (Fig. S20). Similarly addition of ca. 70 equiv. of CF₃SO₃^-, 40 equiv. of PF₆^- to 1·2Cl in DMF-d₇ were required to convert the chemical shifts of Hd and Hd' to those of 1·2CF₃SO₃ and 1·2PF₆, respectively (Figs. S21 and S22). Their binding constants are listed in Table S13 (Supporting information). The best fit to the titration data using a 1:2 binding model (Figs. S24D-F) yielded association constants of K₁₁ = 55.8, 146.9 and 103.7 for NO₃^-, PF₆^-, and CF₃SO₃^- anions, respectively (Table S13). Similar to the Cl^- titration, the second complexation constant K₁₂ was much weaker. The trend for binding constants reflects the strength of anion binding to the macrocycle in the order Cl^- ~ PF₆^- > CF₃SO₃^- > NO₃^-, a dependency on their shape, size and H-bonding capability. Anion titrations with LiCl in Figs. S17–S19 under the same conditions as for 1·2Cl, led to gradual low-field shifts for the imidazole Hd and Hd' protons of both boat and chair forms on addition of LiCl to 1·2NO₃, 1·2PF₆ or 1·2CF₃SO₃ in DMF-d₇, amounting to ca. 0.6 ppm with 5 equiv. of LiCl, giving shifts close to those for 1·2Cl, but with very small shifts for other peaks.

Next we used computational modelling to gain insight into the mechanism of the boat/chair-configuration interconversions. Exploring the configurational space of a moderately large, potentially flexible system like 1·2Cl or 1·2CF₃SO₃ using quantum chemistry is challenging. In this case, the LFMM method was employed by extending the force field developed for monometallic Ru-arene complexes [29–31] to include imidazole ligands (Fig. 4 and Fig. S25 in Supporting information). Remarkably, the energy difference among LFMM-optimized boat-1·2Cl and chair-1·2Cl, boat-1·2CF₃SO₃ and chair-1·2CF₃SO₃ configurations is 0.5 kJ/mol, in excellent agreement with the NMR experiments (Fig. S25, Table 1). As shown in Table S11 and Fig. S16, the NOE cross-peak distances of boat or chair isomers in 1·2Cl or 1·2CF₃SO₃, are in good agreement with the corresponding distances from the X-ray structures and models. In particular, characteristic NOE cross-peaks allow identification of the boat-1·2Cl and chair-1·2Cl isomers in solution, Table S11 and Fig. S16a. However, while broadly similar to the experimental X-ray structures, the LFMM geometries are different and hence the energetic agreement must be treated with caution. Nevertheless, the overall performance of the LFMM force field is sufficiently good to assert that LFMM calculations produce meaningful, although qualitative, results (Figs. S16, S25 and Table S11).

Inspection of the LFMM models revealed that any attempt to convert boat-form to chair-form structures without rupturing the macrocyclic ring would lead to very high energy intermediates. Essentially, as one end of the boat-form structure needs to rotate around a 'hinge' line between the two methylene groups, which connects imidazole and mesityl groups (Figs. 4b amd c), the carbon atoms at the hinge points need to invert. This requires passing through a planar configuration, which is clearly not viable. In this case, the macrocycle must be broken, most likely via fission of one of the Ru-N bonds.

Density functional theory (DFT) was used to examine this possibility because, as is common to any valence force field (FF), LFMM cannot handle bond-making/bond-breaking [32]. The Ru-N dissociation energy was computed for a model system [Ru(η⁶–1,4-dimethylbenzene)(methyl-imidazole)Cl]Cl (Ⅰ, Fig. 4d). A linear transit calculation using one of the Ru-N distances as the constrained variable resulted in an energy profile, which shows a maximum at ~3 Å followed by a local minimum at ~3.6 Å (Fig. S26 in Supporting information). A full transition state search starting from the maximum of the linear transit converged to a TS (single imaginary mode = −97 cm^-1, Fig. S27 in Supporting information) with a Ru-N distance of 2.62 Å at an energy 87 kJ/mol higher than Ⅰ (at the BP86D3/TZVP/COSMO (acetone) level of theory) [33–36]. The subsequent local minimum suggested by the linear transition displays a η²–type interaction from the C=C double bond of the imidazole to the Ru center, while the linear transit suggests a progressive increase in potential energy as the Ru-N bond is stretched to infinity leading to a dissociation energy of ~180 kJ/mol. There is a significant entropy contribution included (~70 kJ/mol) for the dissociative process, and the computed free energy barrier is ~110 kJ/mol (Ⅱ, Fig. 4e), which compares favorably with the NMR experimental measurements (102.8–104.5 kJ/mol) (Table 1). Once the Ru-N bond breaks, both imidazoles (disconnected and attached), are relatively free to rotate from their boat to chair forms. The C(phenyl)-C(H₂) rotational barrier is estimated at ~38 kJ/mol based on MMFF (Merck molecular force field) calculations using the Molecular Operating Environment implementation (Fig. S28 in Supporting information). The fragment can also freely rotate around the other C(phenyl)-C(H₂) bond, which orients the coordinatively unsaturated Ru center correctly for reforming the Ru-N bond and thus reconstituting the metallamacrocyclic ring (Ⅲ, Fig. 4f). These findings are likely to stimulate investigations of dynamic configurational changes which might occur via metal-ligand bond breaking and reformation in other second-row metallamacrocycles, or the even more kinetically-labile first-row transition metals.

Crystal structures revealed H-bonding between solvent (H₂O/CH₂Cl₂) and isomers. Hd/Hd' protons in DMF showed downfield shifts vs. acetonitrile (polarity-dependent), while DMSO titration caused upfield shifts (weak interactions, no coordination). Computed isomerization barriers (~110 kJ/mol) matched NMR data (103–105 kJ/mol), consistent with macrochelate ring-opening mechanisms. Solvent polarity influenced local electronic environments but did not alter isomer ratios, indicating H-bonding stabilizes conformations without driving interconversion.

This discovery of a new class of metallamacrocycles with stable, but interconvertible, boat and chair configurations capable of distinctly different dinuclear coordinative binding as well as hydrophobic and H-bonding interactions with the backbone, suggests applications for induced fit recognition of biological targets such as proteins and DNA, and therapeutic activity [24,37–39]. In comparison, literature reports show that other macrocyclic configuration changes can result from either of the metal centers [40–42]. Hydrolysis of both Ru-Cl bonds in 1·2Cl appears to be facile in aqueous media to give mono- and bis-aqua adducts (Figs. S29–S31 in Supporting information), and mono- and bis-9EtG (9-ethylguanine) adducts have been characterized by ESI-MS (Fig. S32 in Supporting information), which raises the possibility that boat and chair isomers might have selective GG recognition sites on DNA. This study focuses on the synthesis, characterization, and novel dynamic interconversion behavior of a new class of organo-metallomacrocycles. While preliminary experiments suggest there might be potential biological applications, systematic biological investigations remain to be conducted. In future work, we plan to further explore these through a series of biological experiments, such as cellular activity assays and target-binding validation.

Declaration of competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.

CRediT authorship contribution statement

Yaqiong Kong: Writing – original draft, Project administration, Methodology, Investigation, Formal analysis, Data curation. Yan Su: Writing – original draft, Methodology, Investigation, Formal analysis. Yong Qian: Visualization, Resources, Methodology. Robert J. Deeth: Methodology, Investigation, Formal analysis. Isolda Romero-Canelón: Validation, Software, Formal analysis, Data curation. Zhi Su: Validation, Supervision, Methodology, Investigation, Formal analysis. Hong-Ke Liu: Writing – review & editing, Supervision, Methodology, Funding acquisition, Conceptualization. Peter J. Sadler: Writing – review & editing, Supervision, Resources, Funding acquisition, Conceptualization.

Acknowledgments

This work was funded by the Original Exploration Program and General program of National Natural Science Foundation of China (Nos. 22350001, 21977052, 22077066). PJS thanks Anglo American and EPSRC (No. EP/P030572/1) for funding platinum group metals (PGM) research in his laboratory. Yaqiong Kong thanks the Natural Science Foundation of Anhui Higher Education Institutions (No. 2022AH051716).

Supplementary materials

Supplementary material associated with this article can be found, in the online version, at doi:10.1016/j.cclet.2025.111375.