Huishan Lia,
Sally Lok-Wan Ng
a,
Dominic J. Blackb,
Wei Han
a,
Robert Palb and
David Parker
*a
aDepartment of Chemistry, Hong Kong Baptist University, Kowloon Tong, 999077, Hong Kong, China. E-mail: davidparker@hkbu.edu.hk
bDepartment of Chemistry, Durham University, South Road, Durham DH1 3LE, UK
First published on 8th July 2025
Combined luminescence studies and stochastic molecular dynamics simulations have revealed for the first time how cooperative hydrophobic binding and reversible metal ion coordination to protein glutamate residues occurs. A combined experimental and theoretical approach has been used to explain the very different free energies of binding observed between three structurally analogous chiral europium(III) complexes and common variants of serum albumin. In particular, reversible binding of a carboxylate from a glutamate residue was observed; this residue is found in human serum albumin but not the other variants. The binding free energy is exquisitely sensitive to the europium probe structure and charge, and favours complexation of a right handed stereoisomer in the chiral binding pocket. Each process has been visualised by short movies, revealing probe conformational exchange dynamics and the pathway to the protein binding site, on a sub-microsecond timescale.
Early examples of strongly emissive rare earth probes that enhance metal emission on protein binding include a charge neutral terbium DTPA-diamide complex bearing nitroimidazole groups, in which HSA was observed to enhance emission, probably resulting from suppression of oxygen quenching of the chromophore triplet excited state, following protein binding.9 In addition, certain Eu complexes have been shown to exhibit protein binding in which displacement occurs of a ligand donor or coordinated and/or second sphere water molecules that quench the metal excited state by vibrational energy transfer, leading to emission lifetime and intensity enhancement.10–12 Of particular interest in this set, are examples in which protein binding (e.g. to HSA13,14 or the important acute phase protein alpha-1-acid glycoprotein (α1-AGP)11) switches on or triggers enhanced circularly polarised luminescence (CPL) from the chiral Eu centre.15 In this respect, a few well-defined examples of the reversible appearance of CPL have been reported, involving a dynamically racemic Eu probe, in which preferred formation of a non-covalently bound diastereomer occurs, ‘switching on’ the CPL.11,13,14,16
The creation of selective CPL probes is beginning to attract increasing attention in solution, thin films and in the solid-state, as the benefits of encoding the ‘signature’ chiroptical information become more widely appreciated.17 For example, perovskite nanoparticles have been defined in lanthanide MOFs in which the CPL can be turned on18 and various examples with chiral catenanes and rotaxanes are emerging, albeit with modest emission dissymmetry factors.19 Most recently, ‘solvent or medium responsive’ CPL probes have been reported with highly emissive Eu complexes, allowing the tailoring of CPL sign sequences for possible use in security tag applications.20 Chiral lanthanide complexes with extended π conjugation for efficient CPL via two-photon excitation have also been reported.21
[EuL1] | [EuL2] | [EuL3]+ | |
---|---|---|---|
λmax/nm | 323 | 337 | 337 |
k (H2O)/ms−1 | 3.70 | 3.57 | 3.75 |
k (D2O)/ms−1 | 1.86 | 1.62 | 1.73 |
q16 | 1.9 | 2.0 | 2.0 |
[EuL1] | [EuL2] | [EuL3]+ | ||||
---|---|---|---|---|---|---|
τ (ms) | Log![]() |
τ (ms) | Log![]() |
τ (ms) | Log![]() |
|
a Human serum albumin and bovine serum albumin (10 μM), α1-AGP (15 μM).b O-phospho-L-serine/threonine/tyrosine (1 mM), NaHCO3 (2 mM), ammonium acetate (30 mM).c The concentration of the europium complex was determined from the chromophore absorbance using the estimated extinction coefficient.d Binding constants were determined by monitoring changes in the intensity of the main hypersensitive ΔJ = 2 transition at 614 nm. With added HSA and bicarbonate only, the intensity ratio of this electric dipole-allowed transition to the ΔJ = 1 magnetic-dipole allowed transition at 591 nm changed significantly and was also used giving log![]() |
||||||
HEPES buffer | 0.27 | — | 0.28 | — | 0.34 | — |
HSA | 0.30 | — | 0.38 | 6.67d | 0.37 | 5.64 |
BSA | 0.28 | — | 0.28 | — | 0.35 | 5.44 |
Rabbit serum | 0.35 | — | 0.40 | — | 0.38 | — |
Goat serum | 0.38 | — | 0.41 | — | 0.38 | — |
α1-AGP | 0.27 | 0.28 | 0.44 | 6.13 | ||
O-P-Ser | 0.27 | — | 0.29 | 4.54 | 0.35 | 4.67 |
O-P-Thr | 0.27 | — | 0.29 | 4.53 | 0.36 | 4.41 |
O-P-Tyr | 0.27 | — | 0.29 | 4.72 | 0.35 | 4.60 |
NaHCO3 | 0.42 | 3.01 | 0.42 | 3.34 | 0.42 | 4.40 |
NH4OAc | 0.33 | — | 0.32 | 3.13 | 0.34 | 4.42 |
The Eu excited state lifetime was also found to increase from 0.28 to 0.38 ms, and the ΔJ = 2/ΔJ = 1 intensity ratio increased by 87% in the HSA bound complex. These changes in emission spectral form, intensity and lifetime strongly suggest that anion binding to the metal occurs, displacing at least one Eu-coordinated water molecule, by more polarisable donor atoms.13 In contrast, adding BSA to [EuL2] gave rise to no significant spectral intensity changes and only a 10% increase in emission lifetime. An accurate binding constant could not be determined but a very rough estimate of overall intensity change with BSA gave a logK value of below 4, at least two orders of magnitude less than with HSA.
When, CSA (caprine/goat) or LSA (rabbit) (in serum form) were added to [EuL2], no significant changes in either Eu emission spectral form or lifetime were observed (Fig. S1–S3†) and only a reduction in intensity was seen. With [EuL1], no significant spectral changes were observed when adding any of the serum albumins, suggesting that binding in this case is even weaker and does not involve a change in the primary Eu coordination environment. All that was observed was a spectrum that was identical to the europium aqua complex, with little change in overall emission intensity with added protein, in every case, (Fig. S6†). With the mono-cationic complex, [EuL3]+ and HSA, a three-fold increase in emission intensity was observed (logK = 5.64) and the lifetime increased by 10%.
No significant spectral or lifetime changes were observed when adding α1-AGP to [EuL1] and [EuL2] (Fig. S2†), notwithstanding its established ability to bind to a lanthanide ion in structurally related complexes to [EuL1] and [EuL2].12 However, with the cationic complex, [EuL3]+, a seven fold increase in emission intensity and notable changes in emission spectral form were observed, (e.g., the ΔJ = 2/ΔJ = 1 ratio increased by 30%, Fig. 1F) and the Eu lifetime increased by 30%. A binding constant of logK = 6.13(05) was measured, consistent with strong binding into the hydrophobic pocket. This value is slightly higher than that found with the mono-cationic complex [EuL4]+, (log K = 5.82 (05), in which protein binding was shown to involve displacement of one of the azaxanthone N atoms and a Eu coordinated water molecule; chelation of a Glu-64 side chain carboxylate was implicated.12
Such behaviour is consistent with formation of a chelated anionic ternary complex, in which a four-membered Eu–O–C–O ring is formed via carbonate or carboxylate chelation.26 The measured affinity constants for bicarbonate with [EuL1], [EuL2] and [EuL3]+ were logK = 3.01, 3.34 and 4.40 respectively, and for binding to acetate, with [EuL2] and [EuL3]+, log
K = 3.13, and 4.42 (Fig. 1, S5, S7, S8† and Table 2). With [EuL1], observed changes were much less well-defined, and the binding constant was estimated to be log
K < 3. The formation of bidentate chelates between carbonate and acetate and a rare earth metal ion centre has been established by systematic X-ray diffraction and solution-state NMR analyses.27,28 The higher binding constants with [EuL3]+ are consistent with the greater gain in free energy arising from coulombic attraction.
In contrast, relatively little change in the emission spectrum of each complex was observed following addition of O-P-Ser, O-P-Thr or O-P-Tyr. The changes in spectral form and overall intensity, however, were sufficiently distinctive to allow binding constants of logK = 4.54, 4.53 and 4.73 to be determined with [EuL2], for example (Fig. 1). These relatively high log
K values reflect the preference of lanthanide ions for a more polarisable phosphorus oxygen donor, and the slightly higher value with the Tyr example is consistent with its lower free energy of hydration, as it is a slightly more hydrophobic anion. With these phosphorus(V) oxy-anions, chelation does not occur and the oxy-anion binds as a unidentate ligand leaving the axial water intact in these ternary structures.28
It has been shown that the ligand field at Eu is exquisitely sensitive to the nature of the axial donor, rather than the equatorial donor29,30 in such mono-capped square-antiprismatic structures. Such an interpretation agrees with earlier conclusions regarding the constitution and configuration of a variety of ternary anion complexes.24–27
A comparison was made of the amino acid sequences in HSA and BSA, by examining the protein data base (PDB) X-ray structures of HSA and BSA, with and without a bound drug: (HSA: 7VRO, 1AO6 (free), 2BXD (warfarin); BSA: 3VO3, 4F5S (free) and 4JK4 (3,5-diodosalicylic acid); 4ORO (naproxen)). The side chain carboxylate groups in Glu-292, Glu-184 or Glu-188 may plausibly be involved in Eu coordination, when [EuL2] is bound to HSA, as they are relatively close in space to the drug binding site. Critically, Glu-188 and 184 are only found in HSA and are replaced by a Lys or Arg residue and by Thr or Ala, respectively, in the common albumin variants (Fig. S11†).
The next nearest Glu residue in the HSA structure is Glu-153, but it interacts strongly to Arg-256 via a stabilising H-bonded salt bridge. With BSA (and by analogy OSA and CSA) the corresponding Glu residue is interacting strongly with a proximate cationic lysine side chain group, (Lys-274 and/or Lys-276). Such favourable intramolecular coulombic and hydrogen bonding interactions are absent in the HSA structures, where Glu-292 and 188/184 are hydrogen bonded to (high energy) water molecules. The displacement of ‘high energy’ water molecules, following side chain Glu ligation to Eu, may be associated with a favourable free energy change in [EuL2] binding, as the water molecules return to bulk and a more effective hydrogen bonding environment, i.e. such a free energy change is likely to be enthalpically and entropically favourable, considering the local hydration structure.
The binding of [EuL3]+ to α1-AGP (vide supra) had been shown to be strongest (logK 6.13) amongst the three Eu probes examined. Chlorpromazine (used to treat severe anxiety and psychotic aggression) binds to this protein with a comparable affinity (log
K = 6.17).33 A competitive titration of the 1
:
1 α1-AGP/[EuL3]+ complex with chlorpromazine gave rise to displacement of the Eu probe from the protein adduct (Fig. 2C), and this equilibrium was characterised with an apparent binding constant of log
K = 5.90(04). The X-ray structure of chlorpromazine bound to this protein (PDB 3APX) reveals a favourable cation–π interaction between Tyr-36 and the protonated alkylamine group. Both Glu-64 and Glu-92 lie relatively close in space to the guest binding site near the p-Cl-phenyl ring, forming a hydrogen bonded array with Arg-90.
Thus, HSA binding of [EuL2] into the chiral binding pocket of drug site 1, can be hypothesised to favour complexation of the Δ complex more strongly than in the unbound complex, in a ternary structure stabilised by concomitant formation of a metal chelate involving the side chain carboxylate of Glu-184 or Glu-188. In this regard, changes in the relative intensity of the ΔJ = 1 manifold, notably the sign sequence are very evident, comparing the di-aqua complex of [EuL2] vs. its HSA-bound species. Such behaviour supports the hypothesis of a change in the Eu coordination environment that perturbs the local ligand field and determines the nature of the magnetic dipole allowed transitions in this manifold. Indeed, the inversion in sign was very apparent for the transition centred at 596 nm (A to E): three transitions were apparent for the protein bound complex, but only two in the aqua complex. By analysing the ΔJ = 1 manifold, a second order crystal field coefficient, i.e. B02 value of +380(20) cm−1 was measured in each case. The absence of change of the ligand field splitting and the fact that only small variations were observed for the g(em) values (Fig. 2) strongly suggest that the complex is adopting a square-antiprismatic (SAP) structure in both the di-aqua form and the protein adduct, as it is known that the twisted square antiprismatic (TSAP) isomers have a smaller ligand field splitting in these systems.30,34 No significant changes were observed in the total emission spectrum when adding either HSA or BSA to [EuL1], (Fig. S6†) and no CPL could be observed in any example with this europium complex. Furthermore, no change in CPL was observed when adding BSA/CSA or LSA to [EuL2], (Fig. S9†). With [EuL3]+, only addition of α1-AGP gave rise to a modified and more intense CPL signal, (Fig. S10†), and whilst the changes observed were less well-defined than with HSA and [EuL2], the overall sign sequence was the same, consistent with preferred complexation of the Δ stereoisomer.
The stochastic simulations revealed that there are two conformational exchange processes occurring with [EuL2], operating on different timescales. The cooperative rotation of the three ligand arms, involving a switch of the N–C–C–O/N torsional angle of the arms between positive (Δ – clockwise) and negative values (Λ – anticlockwise), occurred on the sub-microsecond timescale (Fig. 3 and Movie A† (https://doi.org/10.5281/zenodo.15099015)). With an S configuration at C in the alanine moiety, the carboxylate groups at the stereogenic centre prefer to lie away from the π cloud of the pyridine ring: this situation occurs in the Δ isomer but not in the Λ isomer, where the carboxylate is closer to the electron rich aromatic ring. This electrostatic preference of about 2.7 kJ mol−1 favours the Δ isomer in [EuL2] and its congeners (Fig. 3), in agreement with the conclusion of the CPL study. The cooperative ring inversion of the tetra-aza macrocycle, involving the simultaneous sign switch of four N–C–C–N torsional angles on the macrocyclic 12-N-4 ring was much slower, and could not be observed during hundreds of nanoseconds of simulations (Fig. 3). Such behaviour is consistent with a large body of experimental data derived from VT-NMR experiments, that showed that the rate of ring inversion in 12-N-4 coordination complexes of lanthanide ions is relatively slow, with a value of around 50 s−1 at ambient temperature.37
The free energy profile for ring inversion was further probed through umbrella sampling simulations. As shown in Fig. 3, there is a large barrier of 25.5 kJ mol−1 that slows down the rate of ring inversion. The Δ-λλλλ isomer, where the values of all the ring N–C–C–N torsion angles are negative (λ, − sign), is lower in free energy than the Δ-δδδδ isomer (δ, + sign). Thus, in an aqueous solution of [EuL2] the Δ-λλλλ stereoisomer is lower in energy than the Λ-δδδδ (in 9-coordinate mono-capped square antiprismatic coordination geometries), in accord with the Δ absolute configuration deduced by CPL (Fig. 2).
The structures corresponding to this minimum were further analysed (Fig. 3 and Movie C†(https://doi.org/10.5281/zenodo.15099015)). The bound complex, [EuL2], was observed to insert the aryl–alkynyl–pyridyl moiety into DS-1, lean against loop region 289–295, and expose its metal centre to solution and the nearby helix spanning residues 173–206. Various types of interaction between HSA and [EuL2] were observed, including favourable electrostatic interactions between Arg-160/Arg-222 and the carboxylate groups of the two alanine-derived arms and between Lys-195/Lys-199/Arg-218 and the single carboxylate group of the chromophore moiety. In addition, there is a cation–π interaction between Arg-222 and the electron rich aryl ring (substituted by an ether oxygen and two methyl groups). Furthermore, favourable hydrophobic interactions occur between Leu-219/Ile-290/Val-293 and the two phenyl ring methyl groups.
Intriguingly, Glu-188 in helix 289–295 was found in proximity to the metal centre of [EuL2]. The distance between carboxyl oxygen atoms of Glu-188 and Eu fluctuated between 0.4 nm and 1.0 nm, indicating that although the sidechain of Glu-188 remained flexible, it had the potential to coordinate with Eu. However, the coordination of this sidechain with Eu was not observed at first, presumably because of the high energy barrier to replacement of water ligands in the coordination shell of Eu, that could not be overcome during the funnel metadynamics sampling process describing the overall positions and orientations of [EuL2]. The standard free energy of formation of this bound structure was calculated to be 25.8 kJ mol−1, corresponding to logK = 4.50, considerably smaller than the experimental value (log
K = 6.67). Therefore, the bound structure observed here can be considered to represent an intermediate state, before the coordination between Eu and HSA takes place with displacement of the two Eu-bound water molecules.
Therefore, starting with the bound structure, another stochastic metadynamics simulation was conducted to probe the free energy surface for coordination of Glu-188 and [EuL2]. In this simulation, the distance (dO–Eu) between the carboxylate oxygen and Eu and the water coordination number were employed as collective variables to enhance sampling. The resulting FES exhibits a flat minimum at dO–Eu > 0.4 nm, consistent with the results from the funnel metadynamics simulation, and a sharp minimum was located around dO–Eu = 0.22 nm, consistent with direct coordination of Glu-188 to Eu (Fig. 3). The two minima were separated by a barrier over 50 kJ mol−1, in accord with the energy barriers found for dissociative water exchange at a rare earth metal centre in complexes with charge neutral donor atoms.37
Both monodentate and bidentate coordination modes were observed. The latter occurred more often, consistent with their occurrence in related X-ray structural studies, studying binding of acetate to diaqua-lanthanide complexes.24,26 The coordination of Glu-188 with Eu further lowered the free energy by 13.8 kJ mol−1, enhancing the affinity of [EuL2] for HSA. Considering both initial binding and subsequent coordination, the overall affinity of [EuL2] for HSA, based on the calculation here, is estimated to be logK = 4.50 + 2.40 = 6.90, which agrees fairly well with the experimental result (6.67) from the luminescence titration.
At the pocket entrance to drug site-1 in HSA, Lys-195/199 and Arg-218 and 222 are found; in the species variants, these residues are also either Lys or Arg. At the base of DS-1, His-242, Arg-257 and Tyr-150 are present in HSA and in all variants (Fig. S11†).The overall free energy of protein binding with the Eu complexes is thus made up of three major contributions: the first involves favourable electrostatic interactions of the two alanine carboxylate groups and the basal carboxylate on the aryl ring in the complex to the proximate Lys, Arg and His residues in the protein (Fig. 3 and Movie C† (https://doi.org/10.5281/zenodo.15099015)); the second is a classical hydrophobic effect, enhanced by a cation–π interaction between Arg-222 and the electron rich aryl ring, where the binding pocket and the chromophore aromatic moiety are size-matched; the third involves reversible metal coordination of a proximate Glu carboxylate side chain, i.e. Glu-188 in HSA and Glu-64 with a1-AGP, displacing Eu-bound water molecules, accompanied by the return to bulk of ‘high energy’ water molecules that occupy the drug binding site or are hydrogen-bonded to the Glu residue side chain. Such an interaction has been hypothesised to occur in the past for a variety of diaqua lanthanide coordination complexes, notably gadolinium.39
This selective protein binding behaviour is not exhibited by [EuL1]. It lacks two methyl groups on the peripheral aryl ring and has a relatively weak affinity for every serum albumin studied and for α1-AGP. This very striking size complementarity between the drug-site 1 binding pocket and the aryl–alkynyl–pyridyl moiety is remarkably finely tuned, in a manner that has no obvious precedent. Given that protein–ligand binding rate constants for association are typically of the order of 105 to 106 M−1 s−1,40 then the binding of [EuL2] to HSA and of [EuL3]+ to α1-AGP is likely to be associated with a dissociation rate of 10−2 to 100 s−1, i.e. slow on the Eu emission timescale. A tentative explanation for the weaker binding of [EuL1] in each case, and the invisible Eu CPL, can then be traced to a faster dissociation rate and a lower free energy barrier to dissociation, owing to the smaller steric demand of the aryl moiety in this Eu complex. Indeed, the nearest residue to the aryl ortho positions (Me vs. H) are relatively bulky hydrophobic residues i.e. Leu-219/Ile-290/Val-293 residues (Movie C† (https://doi.org/10.5281/zenodo.15099015)). Thus, the dramatic impact of Me/H substitution in the aryl ring in these complexes may tentatively be ascribed both to a higher free energy of binding, when more favourable non-bonding interactions involving the methyl groups on the phenyl ring can occur, and to a concomitant larger steric barrier for dissociation, (i.e. for [EuL2]), that slows down the rate of dissociation of the more bulky complex from its protein bound state.
Footnote |
† Electronic supplementary information (ESI) available: The synthesis, purification and characterisation of the ligands and their Eu(III) complexes, together with a description of the equipment used for luminescence and CPL measurements and their analysis. Conventional molecular dynamics and metadynamics simulations, trajectory post-processing and data analysis were performed in water (0.15 M NaCl) and relevant references are given. See DOI: https://doi.org/10.1039/d5sc03017j. |
This journal is © The Royal Society of Chemistry 2025 |