Noncovalent interactions in biocomplexes

2.4.1 Through-metal ligand–ligand interactions

Through-metal ligand–ligand interactions are regarded as electronic interactions between ligands mediated by the central metal ion. These interactions have been concluded for ternary Cu(II) complexes containing DA (= 2,2′-bipyridine (bpy), 1,10 phenanthroline (phen), etc.) and a negatively charged oxygen ligand such as catecholate (cat), Cu(DA)(cat), where the combination of an electron-deficient DA with an electron-rich oxygen ligand, cat, in the Cu(II) coordination plane is favoured [21–23] (cf. 2.5.1.). The stabilising effect of such a ligand combination is explained by electron donation from ligands such as –O^– and Cl^– with filled π orbitals to the metal ion (π-donation) and from the metal ion to ligands such as pyridine and CN^–, which have empty π* orbitals at a relatively low energy level (π-back donation). The former ligands are called π-donors or π-bases and the latter are called π-acceptors or π-acids.

Tanaka presented an equation for estimating the stability constants of ternary Ni(II) and Cu(II) complexes of nitrogen-and/or oxygen-donor ligands from mechanistic considerations by introducing the ligand interaction terms, δ_ij, which were calculated from the reported stability constants and allow for the effect of the donor atom X_i of a ligand on the donor atom Y_j of the other ligand [47, 48]. The δ_ij values may be considered to reflect the through-metal ligand-ligand interactions, and the equation gave estimates of the stability constants for mixed ligand complexes such as His-containing ternary Cu(II) complexes, which were in excellent agreement with the experimental values [49].

2.4.2 Through-space ligand‒ligand and metal‒ligand interactions

In ternary (mixed ligand) complexes there can be steric repulsion between the coordinated ligands due to bulky side chain groups, but here we will consider attractive ligand‒ligand interactions, such as hydrogen bonds, electrostatic interactions, and interactions involving aromatic rings, which favour ternary complex formation. Most L-a-amino acids are effective biological N,O-donor ligands, and those with a metal binding side chain group, especially His, Cys, methionine (Met), aspartate (Asp), glutamate (Glu), and Tyr, are important metal binding sites in proteins. α-Amino acids with the side chain group (X) at neutral pH as shown below are capable of interaction and thus of interest for their possible interactions in and around metal complexes:

1. Negatively charged group: Asp, Glu (X = ‒COO^–)

2. Positively charged groups: Arg (X = ‒NHC(NH₂)₂⁺); lysine (Lys) (X = ‒NH₃⁺)

3. Polar groups: serine (Ser), Thr (X = ‒OH); Asn, Gln (X = ‒CONH₂)

4. Aromatic rings: phenylalanine (Phe) (X = phenyl); Tyr (X = p-hydroxyphenyl); tryptophan (Trp) (X = 3-indolyl); His (X = 4(5)-imidazolyl)

At higher pH or at the metal site, some X groups and Cys (X = ‒SH) dissociate to give a negative charge and may be involved in noncovalent interactions and/or metal binding. In addition, the derivatives such as phosphotyrosine (PTyr) and phosphoserine (PSer) (X = ‒OPO₃^–) may be involved in hydrogen bonds or electrostatic interactions, and biological amino acids such as cysteic acid (CySO₃H; X = ‒SO₃^–), ornithine (Orn; X = ‒NH₃⁺), and citrulline (Cit; X = ‒CONH₂) also have interacting side chains. Bulky alkyl side chain groups may be involved in CH‒π and hydrophobic interactions in and around the metal centre [27, 43, 50].

2.5.1 Stability constants

The existence of intramolecular ligand‒ligand interactions has been concluded for various complexes in solution by stability constant measurements. The stability enhancement of ternary complexes can be evaluated by using the values such as log K_m, ∆log K, and log K calculated from the relevant stability constants [22, 29, 51]. Here the constants will be expressed as the stepwise stability constants (K values) for 1:1 complexes and as the overall stability constants (β values) for complexes with more than one ligand. The equilibrium constant, K_m, is defined as follows (charges are omitted for simplicity) [22, 51]:

where M is a metal ion and A and B are bidentate ligands such as α-amino acids. The statistical value of log K_m is 0.6 for square-planar complexes, and therefore the log K_m value greater than 0.6 indicates that the ternary complex MAB is favoured over the binary complexes MA₂ and MB₂. The ∆log K value corresponds to the logarithm of the equilibrium constant K′ for eq. (3) and indicates the preference for binding of A to MB or binding of B to MA rather than binding of A or B to the solvated metal ion M [22, 51]:

However, the constant K is defined for the following equilibrium [29]:

where A and B are ligands with an interacting group and A′ and B′ are corresponding ligands without it. This is an equilibrium showing preferred formation of complex MAB with ligand‒ligand interactions, with MA′B′ serving as the standard. While the log K_m and ∆log K values indicate the preferential formation of ternary complexes due to various factors, the log K value reflects the stability increase of MAB relative to MA’B’ mainly due to ligand‒ligand interactions [29, 40].

Stabilisation of ternary complexes by through-metal ligand‒ligand interactions has been known for complexes with DA and a phenolate ligand such as cat (L) as seen in 2.4.1. The log K_m and ∆log K values for Cu(DA)(L) with DA = bpy or phen and L = cat or oxalate are greater than those for Cu(en)(L) and Cu(DA)(en) (en = ethylenediamine) due to the above mentioned electron flow, which does not occur with en [21–23].

2.5.2 Spectral data

NMR spectra provide information on ligand‒ligand interactions. The interactions expected for ternary M‒amino acid complexes such as those containing an acidic and a basic amino acid [28] and those containing His and a polar amino acid [37] can affect the side chain conformation of amino acids, which may be studied by ¹H NMR spectral measurements of Pd(II) and low-spin Ni(II) complexes with a coordination structure similar to that of Cu(II) complexes. Changes in the side chain conformations of coordinated amino acids due to the intramolecular interactions in complexes can be detected by the changes in the populations (P) of their staggered rotamers (Fig. 3(a)) calculated from the coupling constants, J, of the ¹H NMR spectra according to the following equations [52, 53]:

Fig. 3

(a) Staggered rotamers of α-amino acids. (b) Structure of [Pd(L-CySO₃H)(L-Lys)] showing that rotamers III of the amino acids are necessary for intramolecular interactions [54].

where J_AB and J_AC are coupling constants between the protons shown in Fig. 3(a) and J_t= 13.56 Hz and J_g = 2,60 Hz [52]. The population of rotamer III (P_III), which enables the intramolecular interactions in complexes as shown for Pd(Lys)(CySO₃H) in Fig. 3(b)), increases with the temperature decrease and the solvent polarity decrease as compared with rotamers I and II, indicating the effect of the electrostatic side chain interactions [54]. Use of rotamer populations for determining the side chain conformations has been reported, for example, for metal‒peptide systems such as Ni(II)‒TyrGlyGly and Pd(II)‒AlaTyr (Gly = glycine; Ala = alanine), which exhibited the P_III increase due to the metal ion‒side chain aromatic ring interaction [55–57].

Aromatic ring stacking in complexes may be detected by the upfield shifts of the proton NMR signals due to the ring current effect [58]. The upfield shifts have been used for evaluating the stacking interactions in ternary complexes involving DA (= bpy, phen, etc.) and nucleotides, amino acids, or peptides [27, 59, 60]. The adduct formation between Pt(II) complexes such as Pt(phen)(en) and nucleotides such as AMP caused upfield shifts of the ¹H NMR signals and downfield shifts of the ¹⁹⁵Pt NMR signals, ∆δ_n, which have been shown to increase with the increase of the enthalpy changes, –Σ∆H_n^º (n = 1, 2 ), for the 1:1 and 1:2 adduct formations (Fig. 4) [61]. The downfield shifts indicate that the electron density of Pt(II) decreases with the increasing stability of the stacked form probably due to delocalisation of the electrons over the stacked structure.

Fig. 4

¹⁹⁵Pt NMR downfield shifts (∆δ_n/ppm) plotted against enthalpy changes (–Σ∆H_n°/kJ mol^‒1; n = 1 (●) and 2 (○)). The ionic strength (l) is indicated for each nucleotide as a suffix [61].

Absorption and CD spectra are important sources of information on ligand‒ligand interactions. Stacking interactions give rise to charge transfer (CT) bands in the near ultraviolet region as observed, for example, for ternary Cu(II)‒bpy‒nucleotide systems involving bpy‒purine base stacking [62]. CT bands were also observed for Cu(II)‒DA‒amino acid complexes; a Trp-containing complex Cu(bpy) (Trp), which has an average bpy···indole distance of 3.67 Å, exhibited a broad band centred at 320 nm in the difference spectrum [63], and similarly Cu(phen)(XPhe) (XPhe = Phe, Tyr, or p-aminophenylalanine (NH₂Phe)) and related complexes gave weak peaks assigned to CT at 320–400 nm [64, 65].

Optically active ligands such as L-α-amino acids and their peptides coordinated to transition metal ions give CD bands in the d-d region due to the vicinal effect of the asymmetric carbon. The CD magnitude (∆∊) of the Cu(II) complexes of oligopeptides are known to be an additive function of the magnitudes due to the component amino acid residues [66]. In the absence of ligand‒ligand interactions the CD magnitudes of ternary amino acid‒Cu(II) complexes have also been found to be an additive function of the contributions from each amino acid. The magnitude for M(A)(B) (M = Cu(II), Pd(II); A and B = amino acids), ∆∊_calcd, can be estimated from the magnitudes ∆∊_Cu(A)2 and ∆∊_Cu(B)2 observed for the binary complexes M(A)₂ and M(B)₂, respectively, by the following equation [28]:

When there exist interactions between the side chains of A and B in M(A)(B), the observed magnitude, ∆∊, deviates from the additivity, i.e. ∆∊/∆∊_calcd≠ 1, due to increased asymmetry. The magnitude anomaly was first observed for ternary complexes containing an acidic and a basic amino acid such as Cu(edma)(Arg) (edma = ethylenediamine-N-monoacetate) and Cu(Asp)(Arg) and the corresponding Pd(II) complexes (Table 2) [28, 67, 68] and has been assigned to the intramolecular electrostatic ligand‒ligand interactions between the oppositely charged side chains as shown in Fig. 2(b). For the complexes with stacking interaction such as Cu(phen)(AA) (AA = aromatic amino acids), the CD magnitude has been found to be much larger than that for Cu(en)(AA) and depend on the aromatic ring of AA in the order Phe < TyrO < Tyr < Trp (TyrO = Tyr with the deprotonated phenol moiety) [29].

Various other methods such as ESR and resonance Raman spectroscopies are also useful for dectection of structural changes due to interactions.

2.6.1 Stability enhancement

In ternary complexes such as Cu(DA)(AA) the side chain group of aromatic amino acids (AA = Phe, Tyr, Trp) can be involved in intramolecular stacking interactions with coordinated DA (DA = bpy, phen, histamine (hista), etc.), and the observed stability enhancement has indicated the existence of such interactions. Table 3 shows a comparison of the log K values calculated for Cu(DA)(AA) by eq. (7) based on eq. (4),

which revealed the stability sequences due to DA and the side chain groups of AA as follows:

DA: phen > bpy > hista

Side chain aromatic group of AA:

Scheme 1

The sequences (Scheme 1) clearly show that the stabilisation due to ligand‒ligand stacking interactions depends on the size and electron density of the aromatic rings of DA and AA involved. The enhanced stability observed for I₂Tyr (= 3,5-diiodotyrosine) and I₂TyrO (I₂Tyr deprotonated from the phenol moiety) suggests the stabilising effect of the interaction between the iodine atom and the aromatic ring [69, 70]. In contrast to this, phosphorylation of Tyr to PTyr reduced the log K values by ca. 1 log unit for DA = bpy or phen (Table 3), which indicates that the stacking interaction is virtually lost [71]. The dependence of the complex stabilisation on the Hammett σ values (σ_p) of various p-substituents of Phe for the complexes Cu(DA)(XPhe) (DA = 4,4′-disubstituted-2,2′-bipyridines; XPhe = Phe, NH₂Phe, Tyr, NO₂Phe, FPhe) indicated that the stability enhancement due to stacking is larger for the systems with a larger electron density difference between the stacked rings [70]. For example, for DA = (NEt₂)₂bpy with two diethylamino groups, the ternary complex with NO₂Phe containing an electron-deficient ring showed a larger log K value than that for the complex with NH₂Phe with an electron-rich ring, while for DA = (COOEt)₂bpy with two electron-attracting ester substituents the log K value was larger for NH₂Phe than for NO₂Phe.

A small stability difference was previously reported for the D-and L-His-containing ternary Cu(II) complexes with basic amino acids (B) such as L-Lys and L-Arg with a protonated side chain at neutral pH. In Cu(D/L-His)(L-B), L-B is considered to be coordinated as seen for L-Asn in Fig. 1(a), and thus its side chain can interact with the axially bound carboxylate oxygen of L-His but not of D-His, resulting in stabilisation of Cu(L-His)(L-B); this stability difference disappeared upon deprotonation from the side chain of B, indicating that the interaction is electrostatic [72, 73]. The difference in the equilibrium constants due to the interactions in ternary complexes with an acidic amino acid (A) and B, Cu(A)(B), was detected in the deprotonation process of the protonated side chain group of B, whose pK_a value was 0.57–0.87 log units higher for Cu(A)(B) as compared with Cu(Ala)(B), which is devoid of the interacting group [74]. The result indicates that the proton of the side chain group of B is necessary for the interaction with A as shown in Fig. 2(b). For the ternary Cu(II) complexes with A = PTyr, PSer, etc. and B = Lys, Arg, etc., the log K values calculated from eq. (4) with A′ = B′ = Ala were found to be larger at I = 0.1 M (KNO₃) than at 1 M (KNO₃), which supports that the stabilising effect is due to the intramolecular electrostatic interactions involving the phosphorylated phenol moiety [71]. As compared with the low stability of stacking in Cu(DA)(PTyr) described above, the phosphoester moiety is effectively involved in electrostatic interactions with Arg and Lys. The result may suggest a possible conversion of the Tyr phenol moiety from stacking to electrostatic interactions upon protein phosphorylation [29].

Stabilisation due to noncovalent interactions may occur in binary complexes when ligands have a set of side chain groups as in peptides. The reactions of Cu(II) with a pentapeptide, NSFRY (= AsnSerPheArgTyr-NH₂), which is a fragment from the 28-peptide atrial natriuretic factor, and its analogs have been studied by pH titrations and spectroscopic methods, and this particular peptide was revealed to form an exceptionally stable 4N-donor complex with Cu(II) compared with the analogs [75, 76]. From comparative studies of the stability constants, the stabilisation has been attributed to various interactions between the side chain groups; the Asn side chain is considered to interact with the amino group to lower its pKa and make it react easily with Cu(II) and further form a fence with Phe. In the 4N-donor complex, hydrogen bonding between the polar atoms of Asn and Tyr is considered to be formed. An interesting stabilising effect comes from the Arg and Tyr residues, which is explained to be due to formation of an additional fence over the one formed by Asn and Phe [75–79]. The results suggest that non-coordinating side chain groups located close together in the coordination sphere interact with each other and cover the space around the metal centre, thus protecting it from the attack of solvent molecules, etc. The observations suggest a situation that is comparable to that at the metal site in proteins.

2.6.2 Structure and selectivity

As described in the foregoing sections, noncovalent ligand‒ligand interactions in complexes M(DA)(AA) and M(A)(B) (M = Cu(II), Pd(II)) have been concluded from solution and spectral studies. The X-ray structures of the complexes isolated as crystals have provided the details of the interactions, although the mode of interactions in the solid state may not always be the same as that in solution. Crystal growth requires interactions between molecules, and intramolecular interactions may be converted to intermolecular interactions to form a polymeric chain. The electrostatic ligand‒ligand interactions concluded for Cu(edma)(Arg) in solution [28] were found to be intermolecular in the solid state as shown in Fig. 5 [80].

Fig. 5

Schematic presentation of intermolecular guanidinium‒carboxylate interactions in [Cu(edma)(L-Arg)]⁺ in the solid state [80].

While stacking within complexes such as M(DA)(AA) often remains localised, association of complexes or adducts with a stacked structure can occur, forming an infinite pile of stacked rings. In the crystal structure of [Pt(phen)(L-Ala)](AQS) (AQS = anthra-quinone-2-sulfonate), [Pt(phen)(L-Ala)]⁺ stacks with AQS^– (Fig. 6(a)), and this adduct unit stacks with the other units, resulting in an alternate pile of the coordinated phen and AQS rings (Fig. 6(b)) [81]. A similar solid state structure was disclosed, e.g. for [Pt(bpm)(L-Ala)](IA) (bpm = 2,2′-bipyrimidine; IA = indole-3-acetate) [82]. In contrast, the adduct of [Pt(bpm)(L-Arg)]²⁺ with a nucleotide (GMP^2– ) exhibited an interesting structure (Fig. 7), which is formed as a discrete unit by intramolecular interactions, a stacking interaction between coordinated bpm and the guanine ring of GMP and hydrogen bonds between the guanidinium and amino groups of Arg and the phosphate oxygen atoms of GMP [82]. This structural unit is connected to the neighbouring units through unique guanine‒guanine hydrogen bonds with the distances of 2.82‒2.86 Å.

Fig. 6

[Pt(phen)(L-Ala)]···AQS adduct. (a) Molecular structure and (b) stacking between the adducts [81].

Fig. 7

Structure of [Pt(bpm)(Arg)]²⁺···GMP^2– adduct [82].

A number of structures with π‒π stacking interactions have been reported for ternary complexes containing an aromatic amino acid (AA) or its peptide and DA, such as Cu(phen)(Trp) [83], Cu(bpy)(Trp) [63], Cu((CONH₂)₂bpy)(Phe) (Fig. 8(a)) [70], Cu(hista)(I₂TyrO) (hista = histamine) [84], Cu(phen)(TyrGlyH_-1) (Fig. 8(b)) [85], and Pd(bpy)(TyrGlyH_-1) [85], where the aromatic side chain of AA is involved in intramolecular stacking with coordinated DA. Figure 8(b) shows that stacking takes place between the Tyr phenol ring and phen bound perpendicular to the Cu(II) plane, which is different from the parallel stacking in Pd(bpy)(TyrGlyH_-1) on the Pd(II) plane. The interaction in M(DA)(AA) usually remains localised in the complex molecule with the shortest atomic distances of 3.0–3.5 Å. The stability sequence due to intramolecular stacking of Phe shows that Cu(DA)(Phe) is rather weakly stabilised by stacking compared with Tyr and Trp (cf. 2.6.1. and Table 3). Probably because of this, the side chain of Phe in [Cu(phen)(Phe)Cl] has been found to be both in the stacked structure and the extended structure without stacking [64], and other examples such as [Cu(bpy)(Phe)(H₂O)]⁺ [64] and [Cu(phen)(Phe)]ClO₄·H₂O [86] were also without intramolecular stacking.

Fig. 8

Structures of Cu(CONH₂)₂bpy)(L-Phe) (a) [70] and Cu(phen)(L-TyrGlyH_-1) (b) [85].

Stacking interactions in proteins are well known to contribute to the stability of proteins [3], where the modes of stacking are often edge-to-face and offset or parallel-displaced to avoid π‒π repulsion [87, 88] and possibly for steric reasons. In metal complexes with aromatic nitrogen donors such as M(DA)(AA), stacking is usually offset, showing limited overlapping due to the steric requirements of the coordination structure and the ligand side chain length. However, the distortion of the coordination plane observed in X-ray structures suggests that the aromatic rings involved in stacking tend to be close to each other to be in a parallel position. This may reflect the weakening of π‒π repulsion due to the electron density decrease of coordinated aromatic nitrogen heterocycles, which are already with a low π-electron density [89].

The stacking in M(DA)(AA) may be expressed as MLπ ‒L’π interactions shown in Fig. 9, where ML and L’ denote a metal-bound aromatic ligand and a pendent aromatic ring, respectively [70]. The stacking also implies MLπ ‒L’π or d‒π interactions and other interactions when there are ring substituents X and Y. Preference for stacking partners is seen from the structures of the Cu(II) and Pd(II) complexes of tridentate ligands containing a pyridine and a phenolate moiety as donor groups and a side chain indole ring as shown in Fig. 10, where the electron-rich indole ring stacks with the pyridine ring but not with the electron-rich phenolate ring [90]. It is interesting to note in this connection that a Trp residue was found to be in contact with the Cu-bound Tyr residue by stacking at the active site of a copper enzyme galactose oxidase; the coordinated Tyr272 phenolate ring with a thioether bridge stacks with the Trp290 indole ring, where the indole ring is considered to play important roles in stabilising the electron-deficient phenoxyl radical formed in the course of the reaction and in the functioning of the enzyme [91, 92].

Fig. 9

π‒π stacking between a metal-coordinated aromatic nitrogen ligand L and a pendent aromatic ring L′ [70]. X and Y are ring substituents.

Fig. 10

Stacking of the indole ring in a Pd(II) complex involving a pyridine and a phenolate moiety as coordinating groups [90].

The histidine imidazole ring is an important metal binding site in metalloproteins and can be involved in stacking with other aromatic rings. The X-ray analysis of ternary complexes, [Cu(hista)(AA)(ClO₄)] (hista = histamine; AA = Phe, Tyr), revealed intramolecular stacking interactions shown in Fig. 11 [93]. From the log K values listed in Table 3, the hista-containing ternary complexes are less stabilised by stacking than those containing bpy or phen probably due to the smaller ring size and higher electron density. The structures show that the shortest distances between the stacked rings are 3.45‒3.49 Å, which are within the normal range. Although ring overlapping is limited and the dihedral angles between the stacked rings are rather large (38.1‒38.5°) in the model complexes, the results suggested the possibility of His‒Phe and His‒Tyr stacking interactions at the metal sites in proteins. Later the coordinated His‒Phe stacking has been actually detected at the Cu site of a plastocyanin from fern (Dryopteris crassirhizoma) between coordinated His90 and nearby Phe12 (Fig. 12), which is the first observation of stacking interactions involving a coordinated His residue [94].

Fig. 11

Structures of Cu(hista)(L-Phe) (a) and Cu(hista)(L-Try) (b) [93].

Fig. 12

Stacking of coordinated imidazole with Phe in plastocyanin from fern (PDB code: 1KDJ) [94].

While steric hindrance in ternary complex formation is well known as a source of stereoselectivity of ligands, noncovalent interactions are important for chiral and molecular recognition in systems involving complexes and other molecules [4, 43]. Selective incorporation of the L-enantiomer of DL-His via formation and isolation of Cu(His)(L-Asn or L-Cit) is in line with the intramolecular hydrogen bonds inferred from the syntheses and structures of the ternary Cu(II) complexes containing His and an amino acid with a polar side chain (Fig. 1(b)) [37, 95] (cf. 2.1 and 2.6.1). Stereoselectivities or chiral recognitions upon complex formation with His-or hista-containing ligands have been reported for complexes, such as Cu(II)‒His‒amino acid [96] and Cu(II)‒cyclo-HisHis‒amino acid complexes [97] by stacking, and for Cu(II)‒hista-functionalised β-cyclodextrin‒amino acid complexes by hydrophobic interactions [98]. Differences in the steric requirements for D-and L-Ala in a chiral Co(III) complex [99] and in the rate of complex formation by a chiral Co(III) complex of a leucine-containing ligand with D-and L-Phe due to π‒π stacking [100] and stereoselective binding of α-amino acids by a chiral cyclen‒Co(III) complex [101] serve as further examples of enantioselectivity arising from ligand‒ligand interactions.

3.2.1 Adduct formation and its effect

Pt(II) complexes such as Pt(phen)(Arg) bind with IA, AQS (Fig. 6(a)), GMP (Fig. 7), and other aromatic molecules by stacking and hydrogen bonding [81, 82]. FMN (riboflavin 5’-phosphate) is known as the prosthetic group of redox carrier proteins flavodoxins; in the flavodoxin from Anabaena, FMN is noncovalently bound to the protein through its isoalloxazine ring sandwiched between the Tyr phenol and Trp indole rings in addition to hydrogen bonds, and the molecular environment is known to control the semiquinone/hydroquinone redox potential of FMN [108, 109]. FMN was reported to form ternary Cu(II) complexes with DA = bpy or phen, where it is bound to Cu(II) through the phosphate moiety and stacking with DA [110]. The Pt(II) complexes were found to interact with FMN to form 1:1 adducts as seen for AQS with a similar three consecutive ring system (Fig. 6(a)), and the stability constants (log K values) were determined to be 2.83~3.42 by ¹H NMR spectra [81]. Upon adduct formation the redox potential (E_1/2) for the two-electron redox processes of FMN exhibited anodic shifts due to the electron density decrease, the shift differences indicating the effects of both stacking and hydrogen bonding.

Complex molecules can interact with the other molecules in the second coordination sphere. For example, crown ethers have been shown to surround Ru-ammine complexes in place of solvent molecules [111]; 18-crown-6 and other crown ethers form adducts with the Ru complexes by hydrogen bonds between the oxygen atoms and the coordinated ammine ligands and cause a cathodic shift of the Ru redox potential, which indicates that the electron density of the Ru centre increases due to binding with the oxygen donors. The factors affecting the adduct formation have been studied [112].

In view of the functions of polyamines such as putrescine and spermidine in genetic information transfer processes, Lomozik and collaborators have investigated the complex formations in metal-ligand systems involving polyamines and nucleosides or nucleotides and interactions between metal complexes and non-coordinated molecules or groups. In ternary systems involving amino acids, nucleotides etc., coordinated polyamines have some protonated amine nitrogens, which bind with the other ligand by hydrogen bonds [113, 114] (for details, see Chapter 1.3.3.2).

3.2.2 Protein‒small molecule interactions

Interactions of metal complexes with biological macromolecules are of current interest in view of the activities of metallodrugs [13–18]. Biological processes of enzyme catalysis and electron transfer require interactions between the enzyme and its substrate and between the electron donor and the acceptor, respectively. Structural studies have been performed, for example, for the enzyme‒substrate model complex of CPA (cf. 2.1) and the electron transfer complexes such as cytochrome c‒cytochrome c peroxidase [9] and amicyanin‒methylamine dehydrogenase‒cytochrome c_551i [115], where various interactions, notably electrostatic interactions between basic and acidic amino acid residues and hydrophobic interactions, have been observed within the complexes.

Plastocyanin (PC) is a mobile electron transfer protein involved in photosynthesis and accepts an electron from cytochrome f (cyt f) of photosystem II and transfers it to photosystem I [116]. Higher plant and green algae PCs have consecutive acidic amino acid residues at the solvent-accessible site near the Tyr residue (negative patch), while cyt f has a Lys residue-rich site exposed to solvents (positive patch) [117]. These oppositely charged sites are known to be involved in the recognition process. Subtle effects of the interaction between the proteins on the structural and electrochemical properties of Silene pratensis PC [118] have been studied by using positively charged Lys peptides such as tetralysine in place of cyt f at neutral pH (Fig. 14) [119, 120]. The absorption spectral changes of PC at around 600 nm caused by pentalysine indicated that the Cu site structure or the Cu‒S(Cys) bond was altered, and from the difference resonance Raman spectrum in the 200–600 cm^‒1 region (excitation wave-length, 591.0 nm), several bands at 375–475 cm^‒1 related with the Cu‒S bond were found to be slightly shifted to lower frequencies, indicating that the Cu‒S(Cys) bond was weakened by addition of pentalysine. The resonance Raman spectral changes of PC caused by lysine peptides were the same as those by cyt c, which is positively charged, and the fact that the peptides, especially tetra-and pentalysine peptides, competitively inhibited the electron transfer from reduced cyt c to oxidised PC indicates that lysine peptides serve as the PC interacting site models of cyt c, cyt f, etc. The PC‒lysine peptide association constants were found to increase with the peptide length, dilysine < trilysine < tetralysine < pentalysine, showing that PC binds more strongly with longer peptides or peptides with more positive charges. Interestingly, as the Cu‒S(Cys) bond became longer due to the PC‒lysine peptide interaction, the redox potential shifted to a higher value, facilitating the electron transfer from the redox partner to PC.

Fig. 14

Structure of Silene pratensis plastocyanin (PDB code: 1BYO) [118] and schematic presentation of interaction with a lysine peptide [119]. Modified from ref [120].

Recently proteins have become important targets for metal complexes and metal-based pharmaceuticals, both for therapeutic and diagnostic purposes [16, 121]. The specificity of enzymes depends on the structural fitness of the substrate to the active site, where the substrate is bound to the enzyme by noncovalent interactions. Transition metal ions have the possibilities of forming diverse structures by complex formation and may fit into the protein’s crucial site, serving as metallodrugs. Interesting studies have been reported by Meggers and collaborators on fitting the structures of inert complexes to the protein kinase active site by mimicking a natural product staurosporine (Fig. 15(a)) known as an effective inhibitor of the enzyme [122]. They developed pyridocarbazole ligands resembling staurosporine and synthesised mixed ligand complexes, which were named as octasporines (Fig. 15(b)), and tailored them to fit into the ATP binding site of protein kinases, where the pyridocarbazole moiety occupies the adenine binding site. The inhibitory activities of the complexes thus prepared were studied, and the Ru complex shown in Fig. 15(b), for example, has been found to be a selective inhibitor of the a-form of glycogen synthase kinase 3 [123]. The structure of a protein kinase, human 8-oxo-dGPTase, with a bound Ru complex is shown in Fig. 16(a), where the complex (Fig. 16(b)) interacts with the Lys, Asn, and Asp residues of the kinase by hydrogen bonds in a hydrophobic environment with Phe, Trp, etc. [124]. The studies indicate the importance of steric fitness of the complex to the active site and noncovalent interactions between them.

Fig. 15

Structures of staurosporine (a) and an octasporine (b) [122].

Fig. 16

Structures of protein kinase‒inhibitor complex (PDB code: 3WHW) (a) and the inhibitor (b) [124].

The approach toward site specific binding of small complexes to proteins will be important for the elucidation of the active site structures and functions and developing new metallodrugs and other functional complexes that can pinpoint the target.

3.2.3 Interactions involving coordinated ligands at the metal site of proteins

Functions of metalloproteins depend on the coordination structure and properties of the central metal ion, donor atoms, and the effect of the molecular environment. Hydrogen bonds and stacking interactions of the donor groups with the second-sphere groups such as the peptide -NHCO-and the side chain groups of amino acid residues can affect the electron density of the coordinating groups and thus the redox properties of the central metal ion. We will see some examples of non-covalent interactions in blue copper proteins and iron-sulfur proteins and their models.

Blue copper proteins such as PC [116] (cf. 3.2.2.) have a unique Cu coordination structure with two His imidazole nitrogens and Cys thiolate and Met thioether sulfurs. As shown in Fig. 12, PC from fern has a Phe residue stacked with the coordinated imidazole and exhibits a higher redox potential than the other PCs from higher plants [94]. Further studies have been reported on the spectroscopic and electrochemical effects of mutation of Achromobacter cycloclastes pseudoazurin at Met16 located close to coordinated His81 to aromatic and aliphatic amino acids [125, 126]. Mutation to Tyr, Trp, and Phe caused shifts of the Cu‒S(Cys) stretching modes to a higher frequency region, indicating that a structural perturbation has occurred to make the Cu‒S(Cys) bond stronger. Comparison of the CysS-to-Cu(II) CT bands and ESR spectra indicated a trigonal disposition of the Cu site. The redox potentials were found to be higher for the mutants with the aromatic amino acids than the wild type and the mutants with aliphatic amino acids [125, 126].

At the metal centre of proteins, hydrogen bonds between the donor groups and the peptide bonds and/or side chain groups of the protein are known to modulate the redox activities of the metal centre. The Cu sites of a blue copper protein azurin and its mutant are shown in Fig. 17 [127, 128]. Wild type azurin has two hydrogen bonds involving coordinated Cys thiolate sulfur and two peptide groups [127], but upon mutation of Phe114 to Pro114 one of the NH moieties was lost, leaving a hydrogen bond with the peptide NH of Asn47 only [128]. This caused a redox potential lowering from 297 mV to 211 mV. A similar shift has been reported for amicyanin, where a hydrogen bond added at the Cu site by mutation caused an anodic shift of the redox potential [130]. Hydrogen bonds around the metal centre have also been known for iron-sulfur [131] and nonheme-iron proteins such as shown in Fig. 18 for superoxide reductase (SOR) [132]. Recent detailed ENDOR and DFT studies on a heme enzyme cd₁ nitrite reductase indicated dynamic formation of hydrogen bonds between the reduction product nitric oxide bound to the heme and the distal His and Tyr residues [133]. Among the approaches to the iron-sulfur proteins, the iron-sulfur clusters were synthesised by using a Cys-containing peptide ligand incorporating the characteristic CysGlyAla sequence of the metal site and related ligands [134]. Fe₄S₄(ZCysGlyAlaOMe)₄]^2‒, which is capable of N‒H···S hydrogen bonding when bound through –S^‒, exhibited a higher redox potential than [Fe₄S₄(ZCysGlyOMe)₄]^2‒ where hydrogen bonding is not possible. The result was supported by further studies [135, 136].

Fig. 17

Shift of the redox potential (midpoint potential E_m) due to loss of a N‒H···S hydrogen bond at the copper site in a blue copper protein azurin. (a) Pseudomonas aeruginosa azurin (PDB code: 4AZU) [127]; (b) Phe114Pro mutant of azurin (PDB code: 2GHZ) [128]. Hydrogen bonds are indicated by broken lines. Reproduced from ref [129] by the courtesy of Asakura Publishing, Tokyo, Japan.

Fig. 18

NH···S hydrogen bonding in a non-heme iron enzyme superoxide reductase (SOR) from Pyrococcus furiosus (PDB code: 1DQI) [132]. Hydrogen bonds are indicated by broken lines. Reproduced from ref. [129] by the courtesy of Asakura Publishing, Tokyo, Japan.

3.3.1 Metal‒aromatic ring interactions

Aromatic rings are known to contribute to the protein structure stabilisation [1, 3, 137] and molecular recognition [138–140] and are often located close to the coordination sphere. This is well known for Cu(II) complexes, where the side chain aromatic ring of amino acids and peptides have been shown to be bent over the Cu(II) coordination plane. As shown earlier (cf. 2.6.2), the aromatic rings of AA and DA in Cu(DA)(AA) undergo stacking interactions within the complex molecule, but at the same time the side chain aromatic ring is located close above the Cu(II) centre to be within the sum of the van der Waals radii, suggesting electronic interactions between them. Aromatic rings tend to occupy a space above the Cu(II) coordination plane even in the absence of aromatic‒aromatic stacking. In the solid state structure of Cu(II)‒GlyTrp complex [141], deprotonated GlyTrp (GlyTrpH_-1) coordinates to Cu(II) as a tridentate ligand in the planar positions with the side chain indole ring of Trp bent over Cu(II) with a rather large dihedral angle of 50°, which is probably due to the planarity of the dipeptide coordination with a deprotonated peptide nitrogen; on the other side of the Cu(II) plane, however, there is the indole ring from a neighbouring molecule at the distance of 3.12 Å and with an angle of 13°. The aromatic ring‒metal ion interaction was also proposed by NMR studies of the Tyr-containing dipeptide complexes of Pd(II) [56] and tripeptide complexes of Ni(II) and Pd(II) [55], and similar Cu(II)-aromatic ring contact was established by X-ray analysis of complexes such as Cu(L-Tyr)₂ [142]. These observations indicate that aromatic rings have a tendency to be close to the metal centre possibly due to electrostatic, electronic, and/or hydrophobic interactions.

The intramolecular stacking shown in the earlier sections is considered to involve both metal-coordinated ligand‒ligand p ‒ p (or MLp‒L’ p) and metal d‒L’ p interactions (Fig. 9). The presence of the electronic interaction between the metal ion and the aromatic ring is suggested by the ¹⁹⁵Pt NMR shift upon adduct formation with AMP, etc. (Fig. 4) considered to be due to delocalisation of the d electrons, and the relatively large enthalpy changes, -SDHo, indicate that the adduct is formed as a result of bonding interactions rather than the entropy effect [61].

Cation–π interactions are of current interest, and indeed they are well recognised in protein structures, ion channels, and enzyme reactions, where the interactions of alkali metal ions, the Arg guanidinium group, and the trimethylammounium group of acetylcholine have been shown to interact with the aromatic rings of Trp, etc. [143–147]. A striking example of transition metal‒aromatic ring interactions in proteins has been reported for a copper chaperone CusF, where the Cu(I) ion bound with two Met sulfurs and a His imidazole nitrogen interacts with a Trp indole ring located close to the Cu site (Fig. 19(a)) [148]. This is the first example of the interaction of the Trp indole ring with a metal ion in biological systems. Figure 19(a) shows that the indole C(4)-C(5) moiety of Trp44 is in contact with Cu(I) with rather long distances of 2.67 and 2.86 Å, and the interaction has been assigned to a cation‒π interaction [148, 149]. Subsequent studies indicated that the Cu(I) binding affinity (log K) of CusF is 14.3 ± 0.1 [150] and that the Cu(I)-Trp44 interaction is important for stabilising the complex and protecting Cu(I) from oxidation by water [151, 152].

Fig. 19

(a) Cu(I)‒p interaction in a copper chaperone CusF (PDB code: 2VB2) [148]. Reproduced from ref. [129] by the courtesy of Asakura Publishing, Tokyo, Japan. (b) Cu (I)‒indole h²-bonding with a 3N-donor ligand [154].

On the other hand, Cu(I) is known to form p-type bonds with alkenes [153]. The reaction of Cu(I) with a 3N-donor ligand containing a pendent indole ring has been shown to give a Cu(I) complex, whose X-ray structure revealed that Cu(I) bound to three nitrogen atoms forms an η²-bond with the C(2)‒C(3) moiety as shown in Fig. 19(b) [154]. The bond distances between Cu(I) and the carbon atoms (2.228 and 2.270 Å) are longer than those of the Cu(I)‒alkene complexes (1.943–2.028 Å), indicating that the bonds are rather weak; when acetonitrile is added to the solution of the complex dissolved in CH₂Cl₂, the carbon donors are replaced by acetonitrile with concomitant changes of the Cu(I)-to-indole CT band at 308 nm. In this connection structures and quantitative evaluation of similar η²-bond formation between Cu(I) and a side chain phenyl ring have been reported for ligands containing a p-substituted phenyl ring in place of the indole ring such as shown in Fig. 20 [155], where the Cu‒C bond distances were 2.336 and 2.211 Å and comparable with those for the indole ring. These Cu(I)-arene interactions were concluded to consist mainly of the interaction between the Cu(I) d_z² orbital and the phenyl ring π orbital. The cation‒ π interactions involving transition metal ions are considered to be somewhat different from those of alkali metal ions. A theoretical study on the Cu(I)‒benzene systems indicated that the electrostatic interaction is not important and that Cu(I) forms an η⁶ cation‒ p complex with the decrease of the Cu(I) 3d electron density as a result of a 3d→π* electron flow, i.e. back donation from Cu(I) d to π* of the aromatic ring [156]. The study also showed that Cu(I) tends to form η² complexes in the presence of counterions.

Fig. 20

d‒π interaction in Cu(I) complexes [155].

As shown in Fig. 19, the distances for the Cu-indole interaction in CusF are longer than those for the Cu(I)‒indole complex, which suggests that the interaction is a more electrostatic cation‒π interaction rather than d‒ π interactions. Thus, depending on the metal ions involved and the distances between the interacting groups, cation‒π interactions may lie somewhere between covalent η-type bonds and electrostatic interactions. A spectroscopic study on Cu(II)‒Trp interactions using Gly-AsnHisTrp-NH₂ showed the UV and CD spectral changes which were ascribed to the cation‒π interaction, and the most stable structure obtained by molecular mechanics calculations indicated that the Cu(II) ion bound by four nitrogens of the peptide is located above the pyrrole moiety of the indole ring with a distance of 3.85 Å, which supported that the interaction is a cation‒π interaction [157]. The pyrrole NH moiety of the indole ring has a very weak acidity which may be compared with the phenol ring [158], and the metal binding ability of indole by the nitrogen or by the carbon atoms has been reviewed recently [159]. These results suggest further possibilities of Trp‒metal ion interactions in biological systems, for example, in prion‒Cu(II) binding.

3.3.2 Interactions involving a hydrogen atom

A hydrogen atom from a metal-coordinated ligand has attracted attention due to its ability to interact with an aromatic ring in the second coordination sphere. The active site of metalloproteins is often associated with aromatic rings from aromatic amino acid residues such as Trp and Phe. The hydrogen atom bound to a ligand may gain a positive charge from the metal ion and undergo an effective cation‒π type interaction. By detailed surveys of the crystallographic data of the Protein Data Bank (PDB), Zaric and collaborators revealed the existence of such interactions in metalloproteins and called them “metal ligand aromatic cation‒π (MLACp)” interactions [160]. MLACp interactions will be stronger when the number of bonds from the coordinating atom to the hydrogen atom is smaller. In metalloproteins the distances between the aromatic centroid and the nearest non-hydrogen ligand atom were found to be 3.09–4.41 Å, and the energies of interactions were calculated to be 4–120 kJ/mol [161]. Interestingly the structure of the alcohol dehydrogenase‒ethanol complex indicated the MLACp interaction of the CH₂ moiety of ethanol bound to Zn(II) with a Phe residue, which may suggest a possible pathway for the hydride transfer to NAD [160, 161].

Close proximity of an alkyl side chain at the type 1 copper site of laccases and domain 2 of ceruloplasmin may be interesting in view of the influence of the molecular environment on the metal site of proteins. The Cu site of these proteins is formed by two His imidazoles and a Cys thiolate, and a branched alkyl group of Leu occupies the position above the trigonal plane with the Cu···C distance of 3.71 Å for ceruloplasmin (Fig. 21) [162]. It has a very high redox potential as compared with that of the blue copper proteins and oxidases with a 2N2S-donor distorted tetrahedral structure. Although the alkyl side chain is not considered as a ligand to the Cu ion, the rather short Cu···alkyl distance suggests certain noncovalent interactions which could affect the properties of the central ion. As structural models for the active site, Cu(II) and Pd(II) complexes of 3N-donor ligands with a branched alkyl side chain have been synthesised and revealed to have a structure as shown for a Pd(II) complex in Fig. 22, where the methyl group (C(4)) is located above the coordination plane with the metal‒C(4) distances of 3.30–3.35 Å [163]. The ¹H NMR spectra indicated that the conformation with the methyl group above the Pd(II) plane is maintained in solution, and the down-field shifts of the signals were larger in solvents with a lower dielectric constant. Considering that the ligand side chain is flexible and can be extended outward in less polar solvents, these observations indicate that the metal‒alkyl contact is considered to be a weak bonding interaction of electrostatic nature. A similar Cu‒alkyl contact has been reported for the Cu(II) complex of an ephedrine derivative [164]. The M‒H‒C angles observed for these complexes are greater than 100°, and the interactions may be classified as hydrogen bonds and not agostic bonds [165]. An interesting property probably arising from the interaction is that the Cu redox potentials of the complexes with the Cu‒alkyl contact were more positive than those of the complexes without the contact, which suggests that the electron density is decreased due to the Cu···H‒C interaction [163]. These observations suggest that the Cu sites of laccase, etc., with an alkyl side chain in close proximity, may also have a decreased electron density.

Fig. 21

Cu···alkyl close contact at the type I Cu site of ceruloplasmin (PDB code: 2J5W) [162]. Reproduced from ref. [129] by the courtesy of Asakura Publishing, Tokyo, Japan.

Fig. 22

Behaviour of the branched alkyl side chain in a Pd(II) complex [163]. (a) ¹H NMR spectra: (a) ligand Mbu-L; (b) Pd(II)‒Mbu-L complex. (b) Structures of Mbu-L and its Pd(II) complex.