Lanthanide(III)-binding peptides and proteins: coordination properties and applications

Abstract

Lanthanides play a crucial role in modern medicine and technology as well as in the metabolism of methylotrophic bacteria. In this context, the research on lanthanide-binding peptides and proteins is an active and rapidly developing field. This comprehensive and critical review focuses on the structural, thermodynamic (affinity and selectivity) and kinetic parameters governing the interaction of Ln³⁺ ions with different peptides and proteins, including both naturally occurring and de novo-designed scaffolds. It thus provides guidelines and future directions for the rational design of Ln-binding peptides and proteins with suitable features for the main applications explored to date, including luminescent sensing, magnetic resonance imaging, Ln separation and recovery and Ln-based (photo)-catalysis.

---

1. Introduction

1.1 Context and scope

Lanthanides (Ln) are f-block elements known for their unique magnetism and luminescence properties. They are used in a wide variety of fields: as contrast agents in medicine, in laser technology, in superconductors and in permanent magnets.¹ Rare earth elements (REE), which include Ln, lanthanum (La), yttrium (Y) and scandium (Sc), have been identified as strategic resources by the European Union (EU).² Indeed, many advanced technologies depend on them, and China remains the main exporter of these resources, supplying over 98% of REE importation to the EU. Of note, the mining and extraction of REE have a deleterious impact on the environment.³

Even before the emergence of interest in Ln biochemistry due to the discovery of the first Ln-utilizing bacteria in the 2010s,^4–7 the design of Ln³⁺-binding peptides was an intensive research field aimed at exploiting the unique physical properties of Ln³⁺ ions for applications in sensing or imaging.⁸

The discovery of Ln-utilizing bacteria has been accompanied by the identification of natural Ln³⁺-binding proteins and Ln³⁺–enzymes, paving the way for a better description of the lanthanome, which corresponds to the proteins involved in Ln³⁺-uptake, trafficking and utilisation.^9–12 These findings have inspired a large community of biologists and chemists, resulting in a proliferation of studies on the rational design of Ln-binding peptides and proteins, and their applications in biomedicine,¹³ biohydrometallurgy,^14,15 and catalysis.^16–21

Previous reviews in the field focused on the rational design of lanthanide-binding peptides,⁸ and a general presentation of the coordination of f-block elements to bio-relevant ligands, including peptides and proteins.²² Here, we first discuss some critical parameters of Ln³⁺ aqueous chemistry and the design of peptides and proteins with a high affinity for Ln³⁺ ions. Then, we provide guidelines on how selectivity against metal ions (Ca²⁺, actinides (An³⁺), d-block metal ions) or among Ln³⁺ can be achieved, and on how to include kinetic considerations in the description of these systems. Last, we highlight recent applications for sensing, MRI, Ln separation and catalysis. In this review, we focus on Ln³⁺-binding peptides and proteins using natural and/or unnatural amino acids, excluding peptides and proteins functionalized with macrocycles such as DOTA, some of which were reviewed in ref. 23. This choice was driven by the focus on the interplay between the structure and the coordination properties (affinity, selectivity and kinetics) of Ln³⁺-binding peptides and proteins scaffold, which do not apply to macrocycles. Nevertheless, small chelators will be occasionally discussed to introduce key features and concepts (e.g. selectivity) useful for Ln-binding peptides and proteins, and peptides with appended macrocycles will be briefly mentioned among the applications, as they represent useful models for the design of luminescent and MRI probes.

1.2 Chemistry of Ln ions in aqueous solution

The most stable oxidation state for Ln in aqueous solution is +3 (electronic configuration [Xe]4fⁿ, with n = 0–14 from La³⁺ to Lu³⁺). Other oxidation states are also accessible for some Ln ions in aqueous solution, such as +2 for Eu and +4 for Ce, due to the enhanced stability associated with a half-filled 4f sub-shell ([Xe]4f⁷ for Eu²⁺) and with closed shell ([Xe]4f⁰ for Ce⁴⁺) configuration (Fig. 1A). Of note, in organic solvents, the entire series of Ln (except for the radioactive Pm) could be stabilised in the +2 oxidation state, while for the +4 oxidation state only Ce⁴⁺, Tb⁴⁺ and Pr⁴⁺ complexes were isolated.²⁴

Fig. 1 Trends of key parameters along the Ln series: (A) reduction potentials;²⁶ (B) ionic radii in aqueous solution²⁷ and in crystals (CN = 9);²⁸ (C) acidity constants (pK_a) for Ln³⁺-bound water molecule; and (D) solubility constants (log K_sp).²⁹ The values plotted are reported in Table 14.

Ln³⁺ ions are hard Lewis acids. Their Lewis acidity slightly increases along the series as a consequence of the decrease of the ionic radius, known as lanthanide contraction (Fig. 1B). Noteworthy, Ln³⁺ ions have a similar size to Ca²⁺ (∼1.0 Å), but are larger than d-block metal ions. As a consequence, Ln³⁺ have higher coordination numbers (CN) than d-block elements.

Due to their hard character, according to Pearson's HSAB principle,²⁵ Ln³⁺ ions prefer hard negatively charged ligands (e.g. carboxylate and phosphate). As a result of the core nature of the f orbitals, the nature of the Ln³⁺–ligand bond is mainly ionic, and coordination geometries are hence mostly dictated by electrostatic and steric repulsions between the ligands.

In water, the larger Ln³⁺ ions (La³⁺–Nd³⁺) are coordinated by 9 water molecules and adopt a tri-capped trigonal prismatic coordination geometry, whereas smaller Ln³⁺ (Gd³⁺–Lu³⁺) are bound to 8 water molecules in a square-antiprismatic arrangement. For Pm³⁺ to Eu³⁺, [Ln(H₂O)₈]³⁺ and [Ln(H₂O)₉]³⁺ are in equilibrium.³⁰

Importantly, as Ln³⁺ aqua ions display Brønsted acidity in aqueous solution, Ln³⁺ speciation depends on the pH, with the acidity constant pK_a (eqn (1)) decreasing along the series (Fig. 1C):

The stepwise deprotonation of bound water molecules leads to the formation of poorly soluble Ln³⁺ hydroxides, Ln(OH)₃. In particular, the solubility (expressed by the solubility product K_sp, eqn (2)) decreases along the Ln series (Fig. 1D).

Ln(OH)₃ (s) ⇄ Ln³⁺(aq) + 3HO⁻

K_sp = [Ln³⁺][HO⁻]³ (2)

In light of the trends described above, it appears that Ln³⁺ share some properties: they are all predominantly found in their +3 redox state and have similar Lewis acidities due to their similar ionic radii. At a more detailed level, Ln³⁺ display differences in their chemical properties: along the series, Ln³⁺ Lewis acidity increases, which can modulate their reactivity, the pK_a of the Ln³⁺-bound water molecule decreases by approximately 2 units (Fig. 1C) and the solubility of Ln(OH)₃ drops from almost 5 order of magnitude (Fig. 1D); moreover Ln ions have different redox properties (Fig. 1A), which have been capitalized on for catalysis (see 6.4 Catalysis).

2. Scaffolds structure & Ln³⁺ coordination

2.1 Generalities

2.1.1 Type of scaffolds. The binding of Ln³⁺ ions to peptides and proteins has been accomplished either by using existing proteins with adequate coordination environment (e.g. 2.2 EF-hand peptides and proteins and 2.5.1 Naturally occurring proteins) or by the rational design of a Ln³⁺-coordination site using natural and/or unnatural amino-acids (e.g. 2.3 Short peptides and 2.4 Peptide scaffolds of intermediate size). These scaffolds range from sequences of a few amino acids (AAs) to proteins, and can be divided into two main categories (Fig. 2): (i) those relying on an “EF-hand-like” motif for Ln-binding (2.2 EF-hand peptides and proteins); and (ii) other types of scaffolds.

Fig. 2 Categories of scaffolds by size (number of amino acids).

These scaffolds can be also differentiated based on their structure in the apo-form, i.e. in the absence of metal ions (Fig. 3). Whereas some are disordered in the apo-form and fold upon metal binding, others are pre-folded even in the absence of metal ions. In this latter case, the binding site can be (i) pre-disposed, i.e. the coordinating residues are placed at appropriate positions for metal-binding, or (ii) pre-organised, i.e. the coordinating residues are not only at an appropriate position but also with a good orientation that requires minimal reorganisation upon metal-binding.³¹ To our knowledge, there is no example of a pre-organised metal binding site in Ln³⁺-binding peptides and proteins, and hence, in the following, we will only describe unfolded scaffolds, or folded ones with a predisposed binding site. In order to discuss the scaffold structure and the Ln³⁺-coordination, we grouped the scaffolds into four sub-categories: (i) scaffolds relying on the well-described EF-hand motifs, which are found in short peptides such as lanthanide-binding tags (LBT), in helix-loop-helix (HLH) peptides, and naturally-occurring proteins such as calmodulin and lanmodulin; (ii) short sequences of ∼10 amino acids; (iii) peptides of intermediate size, spanning from Ln fingers (LF) to coiled coils (CC); (iv) Ln³⁺-binding proteins either naturally occurring or engineered, in which the Ln³⁺ ion is not coordinated in an EF-hand motif.

Fig. 3 Scaffolds that fold upon metal-binding (A), or are folded in the apo-form with a pre-disposed (B) or a pre-organized (C) binding site.

2.1.2 Characterization of Ln³⁺ coordination sphere and scaffold stability. Several techniques give access to information on the Ln³⁺ coordination sphere and on the stability of the apo-peptide and Ln–peptides, which are shortly presented below.
Ln³⁺ coordination sphere. X-ray crystallography is the most accurate way to determine the Ln³⁺ coordination sphere, especially the CN and the mode of binding of certain amino acids (e.g. mono- or bi-dentate for carboxylates in Asp and Glu residues). However, it requires the formation of crystals, which are often difficult to obtain, in particular when working with peptides.

Nevertheless, information on the mono- or bi-dentate binding mode of carboxylates can also be obtained in solution through FT-IR measurements.^32,33 Moreover, the CN may be also investigated in solution through X-ray absorption spectroscopy, specifically EXAFS (extended X-ray absorption fine structure),^34–37 which has not been applied to Ln³⁺-binding peptides and proteins so far.

In addition, insights into the Ln³⁺ coordination sphere can be gained by luminescence measurements. Eu³⁺-hypersensitive transitions can report on the geometry of the coordination environment.³⁸ The hydration number q, which corresponds to the number of Ln³⁺-bound water molecules, is also easily determined by measuring Ln³⁺ luminescence lifetimes (Ln = Eu³⁺, Tb³⁺, Yb³⁺) for a given complex in H₂O and D₂O and using empirical relationships,^39,40 which have been established based on a library of complexes with q ranging from 0 to 6. This is the most solid method and should be favoured compared to other empirical equations that rely only on lifetimes measured in water. Importantly, when working with peptides and proteins it may be difficult to work in H₂O-free conditions. Instead, measurements can be performed for different H₂O : D₂O ratios in order to extrapolate the value in 100% D₂O.⁴¹

Information beyond the number of Ln³⁺-bound water molecules can be gained from NMR measurements, by taking advantage of the magnetic properties of lanthanides.^42–46 Comparison of spectra obtained with diamagnetic Ln³⁺ (La, Lu) and paramagnetic ones (all the other) can be used to identify residues in the first coordination sphere or further away from the Ln³⁺-binding site. For well-folded scaffolds, paramagnetic Ln can even help to refine the protein solution structure. This has been accomplished either by incorporating Ln³⁺ in an intrinsic metal-binding site, such as in Ca²⁺-binding proteins,^47,48 or by introducing Ln-binding tags (LBT, see Section 2.2.2) in the sequence of the targeted protein.⁴⁶

Scaffolds stability. Information on scaffold stability can be gained from denaturation experiments, either using a temperature ramp or chemical denaturants, followed by circular dichroism. By comparison to a “native” scaffold, these measurements can inform on the impact of engineering a Ln³⁺-binding site into the scaffold and the range of temperatures for which the peptide/protein is folded. It is also helpful to quantitatively assess the stabilisation induced by rational optimisations of the scaffold and by Ln³⁺ binding.

2.2 EF-hand motifs

EF-hand motifs represent the largest and best-known family of Ln³⁺-binding sites in peptides/proteins. They are found in most Ca²⁺-binding proteins, including calmodulin (CaM), parvalbumin and troponin C (TnC), and in a recently discovered selective Ln-binding protein, lanmodulin. The EF-hand consists of two α-helices (almost perpendicular to each other, a topology that reminds the spread thumb and forefinger of human hands) linked by a 12-residues loop, which is responsible for Ca²⁺ and Ln³⁺ binding through the amino acids in positions 1, 3, 5, 7, 9 and 12 (Fig. 4).^49–53 In the following paragraphs, EF-hand proteins and peptides are described starting from long-known Ca²⁺-binding proteins (Section 2.2.1), which have inspired the design of optimized Ln³⁺-binding peptides (see 2.2.2 LBTs and 2.2.3 HLHs), and finishing by a recently discovered family of Ln³⁺-selective proteins (2.2.4 Lanmodulins).

Fig. 4 EF-hand peptides and proteins. Top: EF-loop sequence showing the most common Ln³⁺-binding amino acids (in bold and highlighted in pink) in each position. Bottom: Protein structures and Ln³⁺ coordination spheres in: left, troponin C (TnC) bound to two Tb³⁺ ions (PDB 1NCZ); middle, a lanthanide-binding tag (LBT) bound to Tb³⁺ (PDB 1TJB); right, Mex-LanM bound to four Nd³⁺ ions (PDB 8FNS). Figures were generated using Pymol.

2.2.1 Ca²⁺-binding proteins. In most EF-hand proteins, Ca²⁺ is coordinated by 7 oxygen ligands with a pentagonal bipyramidal geometry: residues in positions 1, 3, and 5 serve as monodentate ligands, while the Glu₁₂ residue binds always in a bidentate fashion (Fig. 4 and entry 1 in Table 1). The amino acid in position 7 (Thr in most natural scaffolds) binds instead with its backbone carbonyl. Finally, the 9th residue, which is most commonly (∼30%) Asp, does not bind directly to the metal ion, but it forms a hydrogen bond with a water molecule that completes the coordination sphere.

Table 1 Structural features of Ln³⁺ bound to EF-hand motifs. EF-loop patterns report the amino acids found in positions 1, 3, 5, 9 and 12 (note that the residue in position 7, not shown, always binds Ln³⁺via a backbone carbonyl donor) and their binding mode (m = monodentate, b = bidentate, hb = hydrogen-bonded to a coordinating water molecule, nb = non-bound). Coordination (CN) and hydration (q) numbers are derived from X-ray crystal structures (for which PDB code is reported) or luminescence data obtained in solution using Tb³⁺ or Eu³⁺

---

Since Ln³⁺ ions have similar ionic radius to Ca²⁺, the luminescent properties of some Ln³⁺ ions, mostly Tb³⁺ and Eu³⁺, have been exploited to probe the metal binding sites in Ca²⁺-binding proteins,^54–57 showing that the same coordination sphere with CN = 7 and q = 1 is observed for Ln³⁺ ions bound to Ca²⁺–proteins such as TnC (Fig. 4 and entry 1 in Table 1).^52,58,59

2.2.2 Lanthanide-binding tags (LBT). Following the seminal work of Szabo and coworkers,⁶⁶ the group of Imperiali has widely exploited the Ln³⁺-binding ability of EF-hand loops found in Ca²⁺-binding proteins to develop so-called lanthanide-binding tags (LBT) for various applications (see 6. Applications).^60,67–71

LBTs are 15- to 20-residue-long peptides that were designed by elongating the core 12-residue EF-loop at both termini with apolar amino acids that stabilize the Ln³⁺-bound conformation via hydrophobic interactions. The primary sequence of LBTs was optimized to feature high affinity for Tb³⁺ (see 3.2 EF-hand motifs),^72,73 and to control the hydration number for specific applications. For instance, LBT featuring a Ln³⁺-bound water molecule (entries 2–4, in Table 1) were explored for applications as MRI contrast agents (see 6.2 MRI). The crystal structure of Gd³⁺ bound to a double LBT-ubiquitin fusion construct showed a CN = 7, which is as low as that observed in Ca²⁺-binding proteins.⁶⁰

As Ln³⁺-bound water molecules quench its luminescence emission (see 6.1 Luminescent tags and probes), the coordination sphere was further optimized to afford efficient luminescent LBT by replacing the hydrogen-bonded Asp₉ with bidentate Glu₉. Thus, Tb³⁺–LBT complexes with q = 0 and CN = 8 were obtained (Fig. 4 and entries 9–11 in Table 1).⁶⁴ Moreover, in order to sensitize Tb³⁺ emission a Trp residue serving as the antenna was introduced in the 7th position of the EF-hand motif (see 6.1 Luminescent tags and probes).^66,74

2.2.3 Chimeric helix-loop-helix (HLH) motifs. The group of Franklin designed chimeric helix-loop-helix (HLH) motifs by engineering an EF-hand loop in a transcription factor helix-turn-helix DNA-binding domain, which were investigated as artificial nucleases (6.4 Catalysis) and DNA-responsive MRI probes (6.2. MRI).^16,75–80 The NMR solution structure of the La³⁺–P3 complex revealed a coordination site similar to natural EF-hand loops. However, Asp₉ points away from the metal ion, leaving an open face that could easily accommodate two inner-sphere water molecules, which is in agreement with luminescence data recorded on the Eu³⁺–P3 complex (entry 5 in Table 1).

2.2.4 Lanmodulins. In 2018, a novel family of EF-hand proteins, called lanmodulins (LanM), was discovered in Ln³⁺-utilising bacteria.⁸¹ Compared to previously known Ca²⁺-binding proteins, LanM shows considerably higher selectivity for Ln³⁺vs. Ca²⁺ (see 4. Selectivity). Recently, X-ray crystal structures of two LanM proteins from M. extorquens AM1 (Mex-LanM, Fig. 4) and H. quercus (Hans-LanM) were reported, thus facilitating a thorough comparison of their binding sites with those found in Ca²⁺-binding proteins.⁶² Of note, while CaM is folded in the apo-state and undergoes a conformational change upon metal-binding,⁸² LanM is disordered and only upon binding four Ln³⁺ ions, it forms a compact three-helix bundle fold that is unique among EF-hand proteins.⁸³ The Ln³⁺ coordination spheres in LanMs feature higher CN (9 or 10) than Ln³⁺ bound to Ca²⁺-binding loops and LBT (CN = 7–8). In Mex-LanM, this stems from a monodentate-to-bidentate switch of Asp₅ (entries 1 vs. 6 in Table 1) and the presence of an additional water molecule.

Compared to Mex-LanM, no Ln³⁺-bound water molecules are found in Hans-LanM binding sites due to replacement of Asp₉ with Glu₉ (entries 6–8 vs. 12–14 in Table 1), as already observed for LBT (entries 9 vs. 2 in Table 1). In Hans-LanM, the increased CN relative to Ca²⁺-binding proteins results from the presence of a bidentate Asp₃ (entries 12 and 14) rather than monodentate Asn₃ (entries 1, 9–11) and a monodentate-to-bidentate shift of Asp₅ (entries 12 and 14 vs. 9–11 in Table 1). Furthermore, a decrease in the CN is observed when La³⁺ is replaced by the smaller Dy³⁺ stemming from a bidentate-to-monodentate switch of Glu₉ (entries 12 vs. 14 in Table 1).⁶²

2.3 Short peptides

2.3.1 Cyclic decapeptide (PA). Following the works of Franklin and Imperiali on the design of Ln³⁺-binding peptides inspired by the Ca²⁺ coordination site in EF-hand motifs, Delangle and coworkers explored a different approach to improve the affinity of small peptides for Ln³⁺. First, they selected a regioselectively addressable functionalized templates (RAFT) type of cyclic decapeptide (noted PA) in order to design a Ln³⁺-binding site made of two Asp and two Glu residues (Fig. 5).⁴¹ The apo-peptide adopted an antiparallel β-sheet structure with a predisposed Ln-binding site at pH 3 with the side chains of the four coordinating residues on the same face of the cycle. Upon increasing the pH to 7, coulombic repulsions due to the deprotonation of the carboxylates led to an apparently disordered apo-peptide. Nevertheless, the addition of Ln³⁺ (La³⁺, Tb³⁺) compensated for this repulsion. The authors showed the presence of two water molecules in Tb³⁺ first coordination sphere (q = 2). By comparison with earlier work by Franklin^16,61,76 and Imperiali,⁶⁴ they proposed that the Ln³⁺ coordination sphere includes the four carboxylates (monodentate or bidentate) and two water molecules.

Fig. 5 Schematic structures of short cyclic (A) and linear (B) Ln-binding peptides. In (B) the structures of unnatural amino acids Xxx = Ada_n and Ed3a₂, and shapes adopted by linear peptides are shown. Adapted from ref. 8.

2.3.2 Linear peptides. Next, Delangle and coworkers investigated the use of unnatural amino acids in short linear hexapeptides for improving the affinity for Ln³⁺ (see 3. Thermodynamic stability). All the peptides have the sequence Ac–W–X–P–G–X–G–CONH₂, where X is either Ada_n with an aminodiacetate side chain (n = 1–3), or Ed3a₂ with an ethylenediamine triacetate side chain (Fig. 5), and Pro-Gly is a β-turn-inducer motif.^84–87 These peptides are disordered in the apo-form and fold upon metal binding. The speciation, coordination and solution structure of Ln–peptides were analysed in details and the authors showed that it was impacted by (i) side chain length; (ii) position of the Ln-binding residues in the sequence; and (iii) the type of residues (Ada_n or Ed3a₂, Table 2).

Table 2 Linear peptides speciation, coordination mode, structure and hydration number. The peptides have the sequence Ac–W–X–P–G–X–G–CONH₂, where X is either Ada_n with an aminodiacetate side chain (n = 1–3) or Ed3a₂

Peptide	Speciation	Ln^3+-binding AAs	Coordination mode	Structure	q	Ref.
P^11	1 : 1 and multimetallic species	Ada1	Tri- and tetra-dentate	S-shape	1	84
P^22	1 : 1	Ada2	Tridentate	U-shape	3	85
P^33	1 : 1 and multimetallic species	Ada3	Tridentate	n.d.	3	85
P^12	1 : 1	Ada1 and Ada2	Ada1: tetradentate	S-shape	0	86
P^21	1 : 1 and multimetallic species	Ada2 and Ada1	Ada1: tridentate	U-shape	3	86
P^HD2	1 : 1	Ada2 and Ed3a2	Ed3a2: pentadentate	n.d.	0	87
P^HD5	1 : 1 and multimetallic species	Ed3a2 and Ada2	Ed3a2: pentadentate	n.d.	0	87

---

All the peptides formed 1 : 1 ligand : metal complexes below 1 equivalent of Ln³⁺ relative to the peptide, however, in the presence of more equivalents of Ln³⁺ the formation of multimetallic species was also observed (Table 2).

Changes in Ada_n side chain length impacted its coordination mode to Ln³⁺ (Table 2). Ada₂ and Ada₃ bind Ln³⁺ in a tridentate manner,⁸⁵ whereas Ada₁ is tri- or tetra-dentate depending on its position in the sequence (fifth or second, respectively).⁸⁴ The additional chelating group was proposed to be the backbone carbonyl of the Ada₁ main chain, which can form an additional chelate ring. This change of coordination mode impacted the number of water molecules in the first coordination sphere of Tb³⁺, which dropped from 3 (P²², P²¹ and P³³) to 1 and 0 in peptides where Ada₁ is tetradentate (P¹¹ and P¹², respectively). It also impacted the structure of Ln³⁺–peptide complexes in solution: in the presence of Ln³⁺, P²² and P²¹ adopted a U-shape conformation, likely further stabilised by the formation of a type-II β-turn,⁸⁵ while coordination of Ada₁ backbone carbonyl to Ln³⁺ did not allow for the formation of a β-turn and resulted instead in an S-shape conformation of P¹¹ and P¹² (Fig. 5).⁸⁴

Building upon this knowledge, Delangle and co-workers combined the pentadentate Ed3a₂ unnatural amino acid with Ada₂ either by placing Ed3a₂ in the second (P^HD2) or fifth (P^HD5) position in the sequence (Table 2).⁸⁷ Again, the speciation of the complexes depended on the position of the Ln³⁺-binding amino acids in the sequence. For both peptides, the pentadentate Ed3a₂ excluded water molecules from the Tb³⁺ coordination sphere (q = 0).

More recently, short Ln³⁺-binding peptides rich in Asp or Glu residues were reported. Veliscek-Carolan and co-workers studied peptides made of two to four Glu residues, either with an L- or D-stereochemistry, and coupled to a naphthalene antenna.⁸⁸ Simoni and co-workers studied the structure–affinity relationships of three Asp-rich pentapeptides (ADPDA, DPDPD, DGDGD) with actinide ions and Eu³⁺ and they predicted that the Eu³⁺ coordination sphere contained two carboxylates, one backbone carbonyl, and 4–5 water molecules.⁸⁹

2.4 Peptide scaffolds of intermediate size

In the late 1980s, chemists began to dissect the complexity of proteins' fold and function, with the ultimate goal of being able to predict the primary sequence–structure–activity relationship.⁹⁰ This led to the development of de novo proteins, which refers to proteins designed from first principles. The design of α-helices was soon mastered, and the study of α-helical coiled coil made of the assembly of several α-helices has helped unravel design principles to obtain well-folded coiled coil assemblies. Since then, de novo proteins have been implemented with metal-binding site(s) and functions such as catalytic properties or electron transfer capacities.⁹¹ Regarding Ln-binding peptides, a large amount of work has been dedicated to the design of Ln-binding coiled coils (2.4.1 Coiled coils),^92–102 and a seminal work described the design of Ln-fingers¹⁰³ (2.4.2 Ln fingers).

2.4.1 Coiled coils. The sequences of α-helices in coiled coils are constituted by a repetition of a heptad of amino acids (abcdefg) (Fig. 6 and 8).⁹³ Their assembly is driven by hydrophobic interactions of core amino acids in positions a and d. Residues in positions e and g are used to provide stabilizing electrostatic interactions, which also control the number of helices that assemble as well as their orientations (parallel or anti-parallel). Last, amino acids in position f are chosen for their propensity to induce α-helix folding, whereas those in positions b and c are used to improve solubility.

Fig. 6 Two-stranded coiled coils scaffolds designed by Hodges and coworkers: (A) side view, (B) top view, (C) Ln-binding sites are placed at the helical interface using Glu or Gla AAs in e and g positions. In (B) stabilising non-covalent interactions (hydrophobic, electrostatic) responsible for the folding and assembly of coiled coils are highlighted.

Three strategies were employed to insert an Ln³⁺-binding site in such scaffolds: (i) in two-stranded coiled coils (2SCC, Fig. 6 and Table 3), positions e and g were used to introduce chelating moieties and place the Ln³⁺-binding site at the helical interface; in three-stranded coiled coils (3SCC, Fig. 8), the binding site was (ii) either placed at the N-terminal (Table 4) or (iii) buried in the hydrophobic core by replacing core amino acids in position a and d by Ln³⁺-binding residues (Table 5).

Table 3 Two-stranded coiled coil sequences designed by Hodges and coworkers

---

Table 4 Three-stranded coiled coil sequences designed by Kashiwada and Ashkenazy

---

Table 5 Three-stranded coiled coil sequences designed by Peacock and coworkers

---

Two-stranded coiled coils (2SCC). Pioneering works on Ln³⁺-binding 2SCCs were done by Hodges and coworkers.^94,95 They used as a starting point a de novo α-helical peptide made of the repetition of 5 heptads (35 amino acids, Table 3). The formation of a disulphide bridge ensured the obtention of a parallel 2SCC (Fig. 6). The native sequence was modified to include two (E₂) or three (E₃) Glu residues in e or g positions to form two Ln³⁺-binding sites per coiled coil.⁹⁴ The stability of the apo-peptides in their oxidized form was quantitatively investigated by urea denaturation followed by circular dichroism (Fig. 7). The destabilisation induced by the mutation of Gln to Glu was the greatest for the E₂(15,20)_ox construct, in which Glu in positions 15 and 20 generates the highest interchain repulsion. The introduction of a third Glu residue in the sequence (E₃(13,15,20) and E₃(15,20,22)) further destabilised the coiled coil.

Fig. 7 Urea denaturation of 2SCC at pH 7 followed by circular dichroism.⁹⁴ Peptide sequences are given in Table 3. Experiments performed at 20 °C in 50 mM TRIS, 100 mM KCl, [pep] = 70–140 μM, without or with 50 mM LaCl₃. [urea]_1/2 corresponds to the concentration at which 50% of the peptide is unfolded.

A stabilising effect of La³⁺ on peptide unfolding was noted for E₂(15,20) and E₃ constructs, in which the Glu residues are well positioned to bind La³⁺ at the interhelical interface, the resulting interhelical bridges being responsible for the stability enhancement. The higher stabilisation measured for the E₃ sites may also indicate that the third Glu could participate in binding.

In addition, the authors evaluated the stability of the reduced peptides. These were less stable than their oxidised counterparts, highlighting the importance of the disulphide bridge for the stability of the scaffold. The addition of LaCl₃ resulted in an important stabilisation, which was attributed not only to La³⁺-binding but also to a La³⁺-driven dimerization. Hence, the most stable Ln³⁺-binding constructs were achieved thanks to the presence of a disulphide bridge to form exclusively a 2SCC, and to Ln³⁺-binding residues at the helical interface.

In a following work, Hodges and co-workers turned to γ-carboxyglutamic acid (Gla, Fig. 6), with two carboxylate functions, in order to (i) enhance the Ln³⁺-affinity of the coiled coil thanks to the higher denticity of Gla vs. Glu, and (ii) design a system that folds upon metal-binding as the higher electrostatic repulsion induced by Gla was expected to lead an unfolded apo-peptide.⁹⁵ In order to further destabilise the apo-peptides, they mutated a Val residue (V) in position 23 by an Asn (N) (Table 3). The resulting peptide in its oxidised form, Gla₂(15,20)N, was fully unfolded at 20 °C pH 7. The peptide folded upon the addition of 0.5 mM LaCl₃, the amount of coiled coil formation increasing from ∼3 to 100% to match the one of the native Asn.

Three-stranded coiled coils (3SCC). A few years later, Kashiwada and coworkers used the same Gla residue for Ln³⁺-binding, but inserted it in a three-stranded coiled coil (3SCC) scaffold.⁹⁶ In their design, the Gla residue was placed in an a layer of Pep1 (Table 4) so that the Ln³⁺-binding site would be buried in the hydrophobic core rather than placed at the helical interface. Due to the insertion of the Gla residue in the sequences, the resulting peptide Pep3 was unfolded at 20 °C, pH 7. Circular dichroism experiments showed that the addition of 1 equivalent of Ln³⁺ (Ce³⁺, Eu³⁺, Tb³⁺) per trimer drove the assembly and folding of the coiled coil.

Ashkenazy and coworkers investigated an alternative strategy, by placing the Ln³⁺-binding site at the N-terminal of the 3SCC (Fig. 8 and Table 4), which has the advantage of not perturbing the folding and assembly of 3SCC.⁹⁷ They coupled a hydroxyphenol oxazoline moiety (HPO) to the peptide through a triazole linker. HPO acted both as a chelating moiety and as a sensitizing antenna.

Fig. 8 Three-stranded coiled coils scaffolds: (A) side view, (B) top view, (C) Ln-binding site placed at the N-terminal with a hydroxyphenol oxazoline (HPO) ligand, (D) Ln-binding site in the hydrophobic core (a and d positions). In (B) stabilising non-covalent interactions responsible for the folding and assembly of coiled coils are highlighted.

The group of Peacock contributed to the design of Ln³⁺-binding 3SCCs and provided a detailed understanding of the influence of Ln³⁺-binding site position, Ln³⁺-binding residues, second sphere effects, and scaffold size on Ln³⁺-affinity, selectivity, and hydration number (Table 5).^98–102,104 They also reported the design of a heterobimetallic 3SCC bearing an Ln³⁺-binding site and a Hg-binding site.⁹²

The first peptide designed by the Peacock group, MB1-2, was a parallel 3SCC made of the repetition of 5 heptads and with an (Asn)₃(Asp)₃ metal binding site with Asn residues in a d layer, and Asp residues in the following a layer (Table 5).⁹⁸ The peptide also contains Trp residues to sensitize Tb³⁺ emission (see 6.1 Luminescent tags and probes).

The positioning of the Ln³⁺-binding site within the hydrophobic core of the 3SCC destabilised the peptide, which was unfolded in the apo-state but assembled and folded upon Ln³⁺-binding (Fig. 9). The hydrophobic environment and the adequate coordination sphere ((Asn)₃(Asp)₃) provided a suitable CN to the Ln³⁺ ion, preventing water from binding directly to the metal ion (q = 0).

Fig. 9 Comparison of the percentage of folding of 3SCC.^99,102 Peptide sequences are given in Table 5. Experiments were performed in 5 mM HEPES, pH 7, 293 K, [pep] = 30 μM, without or with 10 μM GdCl₃. The percentage folding was calculated based on: , with [θ]_222nm the molar ellipticity at 222 nm, [θ]_coil the molar ellipticity for a random coil, and the theoretical maximum molar ellipticity, where n is the number of residues in the sequence.¹⁰⁵

Building on this work, the group of Peacock investigated how the position of the binding site in the 3SCC scaffold and the type of binding site ((Asn)₃(Asp)₃vs. (Asp)₃) would impact Ln³⁺ coordination and hydration number.⁹⁹ With this aim, they extended the MB1 series containing an (Asn)₃(Asp)₃ binding site with peptides MB1-1, MB1-3 and MB1-4, and designed the CS1 series containing an (Asp)₃ binding site with peptides CS1-1, CS1-2 and CS1-4 (Table 5). In the MB1 series, the position of the metal-binding site impacted differently the stability of the apo-peptides (Fig. 9). Whereas positioning the metal-binding site in the middle of the scaffold destabilized the apo-peptides (21% and 15% folded for MB1-2 and MB1-3, respectively), positioning at the C-terminal was better tolerated (55% folded), and when placed close to the N-terminal the apo-peptide seemed almost unaffected (80% folded). For the CS1 series, in which there is no top (Asn)₃ layer, the apo-peptides were always more stable than their MB1 analogues (Fig. 9). As was previously observed for MB1-2,⁹⁸ Gd³⁺ stabilised the 3SCC assembly and led to the formation of 1 : 3 Ln³⁺ : peptide complex. However, the authors noted that for CS1-2 and CS1-4, there was only a small improvement of folding upon Tb³⁺-binding compared to their MB analogues (Fig. 9), and only a small increase in Tb³⁺-emission which suggested that these peptides may bind Tb³⁺ through non-specific interactions.

Depending on the position of the binding site, the number of Tb³⁺-bound water molecules varied from 4 (CS1-1) and 3 (MB1-1) for the more solvent-exposed sites at the N-terminal, to 2 (MB1-4) for the site at the C-terminal, to 0 (MB1-2 and MB1-3) for sites more buried into the scaffold (Table 5). These results demonstrated that it is possible to control the number of Ln³⁺-bound water molecules by controlling the position of the metal binding site in the 3SCC. The more solvent-exposed sites at the N- and C-termini had the higher hydration number, which suggested a change in Asp coordination mode (monodentate vs. bidentate) or that not all Asp and Asn were involved in Ln³⁺-coordination. On the other hand, the sites that were more buried within the hydrophobic core of the 3SCC were shielded from the solvent. This is consistent with the (Asn)₃(Asp)₃ site fulfilling the preference of Ln³⁺ for a high coordination number. The recent publication of the crystal structure of HC02 (PDB 7P3H), a 3SCC similar to MB1-2, confirmed that Tb³⁺ is nona-coordinated by three Asn (monodentate) and three Asp (bidentate) as shown in Fig. 8.¹⁰¹

In later works, Peacock and coworkers investigated both the effect of the peptide length on the overall stability of the assembly¹⁰² and of second-sphere effects on the hydration number (Table 5).¹⁰⁰ Whereas the shorter MB1-1S was too destabilised to bind GdCl₃, there was no marked difference in peptide folding or hydration number between MB1-1 and the longer MB1-1L (Fig. 9), and only a qualitative enhancement of stability for MB1-1L compared to MB1-1. One explanation proposed was that the binding site is too far (4–5 nm) from the sixth heptad to detect any positive effect on folding, Ln³⁺ affinity and hydration number.

In order to investigate 2^nd sphere effects on the hydration number, core Ile (I) residues placed in the a layer above the metal-binding site of MB1-1 were mutated to Ala (A), Phe (F), Tyr (Y) or Trp (W) (MB1-1(2X) series, Table 5). Tuning of the steric hindrance provided control of the hydration number of Tb³⁺ that varied from q = 3 (MB1-1, MB1-1(2A)), to 2 (MB1-1(2F)), to 1 (MB1-1(2Y)), to 0 (MB1-1(2W)). This strategy was also successfully applied at the C-terminal for MB1-4 site, which was mutated to MB1-4(37W) in order to shield Tb³⁺ from the solvent, which decreased the hydration number from 2 to 0.

The results obtained by Peacock and coworkers highlight that with the same set of coordinating AAs, the number of Ln³⁺-bound water molecules can be controlled either by controlling the position of the Ln³⁺-binding site within the scaffold or by tuning second sphere residues thanks to steric effects. The replacement of core amino acids by Ln³⁺-binding residues resulted in the destabilisation of peptide folding and assembly that depended on the position of the site within the scaffold and could be in part compensated for by Ln³⁺-binding.

2.4.2 Ln fingers. A seminal work by Zondlo and coworkers described the rational design of lanthanide fingers (LF, Fig. 10 and Table 6).¹⁰³ Their goal was to obtain a peptide that adopts the same fold as the naturally occurring zinc fingers, whose structure is constituted of an α-helix placed above a β-hairpin.

Fig. 10 Lanthanide fingers. A model of the hypothetic structure and binding site of Ln-fingers was generated from a closely related zinc finger structure (PDB 1ZAA) using Pymol. Zn is represented as a grey sphere.

Table 6 Ln fingers sequences designed by Zondlo and coworkers. ^DP = D-proline

---

The (His)₂(Cys)₂ Zn²⁺-binding site of Zn-fingers was replaced by an (Asp)₂(Glu)₂ binding site, more suited for Ln³⁺-binding. The LF scaffold was redesigned to favour the formation of secondary structures, leading to LF1. Modifications included notably the use of α-helix or β-sheet inducer amino acids and of a type II′ β-turn (^DPro-Ser) to favour the formation of the β-hairpin. The addition of an N-cap (LF2) and a C-cap (LF3), which are motifs favouring the formation of α-helices,^106–108 improved the folding. The extent of stabilisation of peptides fold following these modifications was qualitatively assessed by comparing CD spectra obtained for apo-peptides. The influence of the size of the binding site on Ln³⁺-binding was investigated by opening up the site thanks to the deletion of two AAs in the sequence (LF4), as well as the type of Ln³⁺-coordinating residues (LF7). Genetically encodable LFs were designed by replacing the type II′ β-turn (^DPro-Ser) with Thr-Ile (LF5 and LF6).

A detailed NMR analysis of LF4 without and with Lu³⁺ showed rearrangements throughout the peptide upon Lu³⁺-binding, which suggested that, as for Zn-fingers, Ln-fingers fold upon metal binding. For all the sequences, Ln³⁺ binding improved the folding of the peptides and the data obtained were consistent with a 1 : 1 peptide : Ln³⁺ ratio. Hydration numbers were not determined, but H₂O molecules are likely to participate in Ln³⁺ coordination sphere in addition to the Ln³⁺-binding AAs.

2.5 Other Ln-binding protein scaffolds

2.5.1 Naturally occurring proteins.
Ln-ADH. The discovery of Ln-utilizing bacteria has sparked a great interest in the identification of the lanthanome.^9–12 The first Ln-binding protein discovered in the 2010s was an enzyme: a pyrroloquinoline quinone (PQQ)-dependent methanol dehydrogenases (MDH) encoded by the gene xoxF, whose active site also contains an Ln³⁺ taking part in the catalysis.^4–7 A characteristic of this enzyme is that it is directly involved in the metabolism of the organisms that produce it, mostly methylotrophic bacteria, which are hence the sole organisms known so far for which Ln³⁺ can be essential elements.⁹ Following this discovery, several Ln-MDH were isolated and their crystal structures were reported with either La³⁺,^109,110 Ce³⁺,⁷ Nd³⁺,¹¹¹ or Eu^3+ 112 (ref. 112) in the active site (Fig. 11). In addition, the family of Ln-dependent enzymes was extended to ethanol dehydrogenases (EDH) following the identification of ExaF¹¹³ and PedH¹¹⁴ in methylotrophs and non-methylotrophs.¹¹⁵ The crystal structure obtained for a Pr³⁺-EDH showed that the active site of this enzyme shared a lot of similarities with the one of Ln-MDH, including conserved residues involved in Ln³⁺- and PQQ-binding.¹¹⁶ Due to these similarities, Ln-MDH and Ln-EDH are often grouped under the term Ln-alcohol dehydrogenases (Ln-ADH).

Fig. 11 Ln-ADH structure and active site. Zoom on the active site and Ln-coordination sphere of crystals obtained with the PQQ cofactor (blue) in the active site (left) and without it (right). In addition to Ln-binding AAs and PQQ, a methanol molecule is also bound to Nd³⁺.

The active site of Ln-ADH is closely related to other PQQ-dependent alcohol dehydrogenases, whose active site contains a Ca²⁺ ion, but which are encoded by different genes (mxaF for Ca-MDH, exaA and pedE for Ca-EDH).⁹ The main difference between the active sites of Ca- and Ln-ADH is the presence of an additional Asp residue in the Ln³⁺ coordination sphere. Crystal structures showed that in Ln-ADH, the Ln³⁺ ion is nona-coordinated by two Asp (one monodentate, one bidentate), a Glu (bidentate), an Asn (monodentate), and the PQQ cofactor (tridentate), as shown in Fig. 11. For some structures, an additional ligand (PEG,⁷ MeOH¹¹¹) is also found coordinated to the Ln³⁺, which suggests that there is a free site for substrate coordination to the metal ion, giving a total coordination number of 10.

It is not clear how the mature Ln-ADHs are formed, especially how they acquire their two cofactors, PQQ and Ln³⁺, and how these impact the stability of the scaffold. Based on the obtention of crystal structures with only Ln³⁺ (and not PQQ) in the active site,^110,111 it has been proposed that Ln³⁺ could be loaded independently of PQQ. However, it is also possible that PQQ leaks out of the enzyme pocket during the crystallization steps. The comparison of the holo-MDH (with Ln and PQQ) and Ln-only-MDH (no PQQ) shows nearly identical structures (Fig. 11).^110,111 Ln is coordinated by the same residues in both cases. Residues in interaction with PQQ show minimal rearrangement in the holo-MDH relative to the Ln-only MDH, which suggests that the MDH active site may be predisposed for PQQ binding.

On the other hand, ADH enzymes can also be obtained with PQQ-only in their active site. Indeed, several papers report on the purification of Ln-ADH with 0.4 to 1 equivalent of Ln per protein.^{6,7,109–113,117,118} Moreover, Daumann and coworkers demonstrated that upon incubation of partially metallated MDH with Ln³⁺ the activity of the enzyme could be improved, which suggests the in situ reconstitution of a functional enzyme in which PQQ was already present.^112,119 Martinez-Gomez and coworkers also observed the partial metallation of a Nd-MDH, without change in the PQQ content of the enzyme.¹¹⁷ However, with their conditions, the reconstitution of a holo-enzyme by incubation with Nd³⁺ was unsuccessful.

Working on a parent enzyme, the PQQ- and Ca²⁺-dependent soluble glucose dehydrogenase (sGDH), Stines-Chaumeil, Limoges and coworkers demonstrated that reconstitution of the holo-enzyme could follow either a path where PQQ is added first to the active site or one where Ca²⁺ is bound first.¹²⁰ Their results evidenced that the kinetic of reconstitution depended on the order of addition of the two cofactors. The reconstitution was fast when the protein was metallated with Ca²⁺ before the binding of PQQ, and slow if PQQ binding happened first. Although little is known about the mechanism and kinetics of Ln-ADH reconstitution, the two reconstitution paths proposed for sGDH could also be considered for Ln-ADH.

The data on Ln-ADH stability are scarce. Some data show that Ln-MDHs are sensitive to temperature and denature at temperatures higher than 50 °C.^111,112 However, the exact value at which denaturation occurs may be influenced, among other factors, by the organism Ln-MDH is purified from, as the enzyme obtained from the extremophile M. fumariolicum SolV has an optimum temperature for catalytic activity at 60 °C.⁷ In addition, the thermal stability of Ln-MDH could be Ln-sensitive, as suggested by Nakagawa and coworkers who compared the thermal stability of a La- and a Nd-MDH and found that the latter denatured at lower temperatures although it is not clear what could cause such a difference.¹¹⁸

Lanpepsy (LanP). Following the discovery of ADH, other Ln³⁺-binding proteins were identified: lanmodulins (LanM, see above), lanpepsy (LanP), and landiscernin (LanD also referred to as LutD, see below).¹²¹ LanP is composed of two PepSY domain.¹²² Studies on deletion strains suggested that LanP function is not directly linked to Ln³⁺ sensing or uptake. The 3D structure predicted by AlphaFold suggested that the folded protein forms a cavity with several negatively charged residues. Competitive titration experiments determined that LanP could bind up to 3–4 Ln³⁺ ions, which was confirmed by ICP-MS, although a higher number of sites was determined by ITC titration (∼6). The exact binding sites and coordination environments of Ln³⁺ in LanP still need to be uncovered.
Landiscernin (LanD). A role for LanD in the acquisition of Ln³⁺ ions was first suggested in 2019.^123,124 Following these preliminary works, the protein was isolated and crystallized by Cotruvo and coworkers.¹²⁵ It forms a three-helix bundle stabilised by a disulphide bridge and hydrophobic interactions at its core. Several negatively charged residues are found at its surface, close to the N-terminal. Crystal structures obtained with 0.5 equivalents of Ln³⁺ (La, Ce, Eu, and Ho) per protein showed the formation of a Ln³⁺-binding site at the interface between two proteins made of three Glu residues from each protein (Glu70, Glu73, Glu75), which except for this coordination site do not interact with each other (Fig. 12). Interestingly, as the Ln³⁺ ionic radius decreases, the coordination number decreases from 9 (La³⁺, Ce³⁺) to 8 (Eu³⁺, Ho³⁺), due to the loss of a Ln³⁺-bound water molecule for the latter (Fig. 12).

Fig. 12 Landiscernin (LanD) structure and active site. Zoom on the Ln³⁺ coordination sphere (left). Overlay of Ln³⁺–LanD structures (Ln³⁺ = La, Ce, Eu, Ho) was generated with Pymol (right). A Ln³⁺-bound water molecule is only present in the structure of La³⁺–LanD and Ce³⁺–LanD. Two glutamate residues adopt different orientations in the crystal structures and their rotamers are represented. They do not participate in Ln³⁺-binding but are important for dimer formation.

In solution, the protein was shown to be monomeric in the apo-state with a dissociation constant for the dimer of K_dimer = 610 μM but was differently influenced by metal ions. Whereas La³⁺, Ce³⁺ and Nd³⁺ seemed to favour the dimerization of the protein (K_dimer = 117, 200 and 253 μM, respectively), this was not the case with Eu³⁺ and Ho³⁺ (K_dimer = 1400 and 700 μM, respectively). Based on the crystal structures obtained, the authors proposed that as the ionic radius of the metal ion decreases, electrostatic repulsions due to the presence of negative charges on the surface of the proteins increase and disfavour the dimerization.

2.5.2 Proteins designed/engineered for Ln-binding.
PqqT. The discovery of Ln-utilising bacteria has also inspired the field of artificial enzyme design, with two lead examples recently published. Olshansky and coworkers aimed at designing a Ln-ADH mimic, which could serve as a model to study the mechanism of the enzymes.²⁰ Analysis of the X-ray structure of PqqT, a PQQ-binding protein, showed that several residues near PQQ may be predisposed for Ln-binding, including Glu₁₁₁, Asp₁₄₁ and Asn₁₄₃ (Fig. 13). In addition, a lysine (Lys₁₄₂) interacts through electrostatic interactions with PQQ with its side chain positioned in the PQQ pocket where Ln³⁺ binds. Mutation of Lys₁₄₂ (K₁₄₂) to an Asp afforded the mutant PqqT–K₁₄₂D that binds 1 PQQ (model structure in Fig. 13) and 1 Ln³⁺ ion. In the engineered active site, PQQ was required for the binding of La³⁺ and hence participated in its coordination sphere. In addition to Asp₁₄₂, other residues (Glu₁₁₁, Asp₁₄₁, Asn₁₄₃) may also participate in Ln-binding, which still needs to be confirmed. The mutant containing La³⁺ and PQQ, noted La³⁺–PQQ⊂PqqT–K₁₄₂D, was able to oxidize benzyl alcohol under aqueous conditions (see 6.4 Catalysis).

Fig. 13 Ln–PqqT–K₁₄₂D mutant bound to PQQ and its possible Ln³⁺-binding site (zoom). Model structure was generated from the crystallographic structure of PQQ-bound PqqT (PDB 9B1U) using the mutagenesis tool in PyMOL.

TIM barrel. The second example of an Ln artificial enzyme came from Zeymer and coworkers. Starting from a computationally designed de novo TIM barrel scaffold,¹²⁶ they aimed to design an enzyme for Ce⁴⁺ photoredox catalysis (see 6.4 Catalysis).¹⁹

The initial scaffold was designed by combining a TIM domain made of eight parallel β-strands surrounded by eight α-helices, and a ferredoxin (FD) insert domain, either with a (His)₂(Glu)₂ or with a (Glu)₄ binding site.¹²⁶ The stability of the scaffolds with these two binding sites was assessed by thermal and chemical denaturation. The two scaffolds showed a high stability with temperature (T_m > 95 °C), and unfolded in two steps with increasing concentration of guanidinium chloride, each event corresponding to the unfolding of one domain (TIM and FD). X-ray structure of the apo-protein scaffold showed a pre-disposed (Glu)₄ binding site. The hydration number of Tb³⁺ was determined using a third mutant with a Trp for Tb³⁺ sensitisation. In this mutant, Tb³⁺ was bound by one water molecule. X-ray structure showed a CN = 9 for Tb³⁺, with the four Glu acting as bidentate ligands (Fig. 14).

Fig. 14 TIM barrel structure and active site. Zoom on Ln-coordination sphere showing the presence of a Ln³⁺-bound water molecule.

2.5.3 Ln-binding to other metalloproteins for specific applications. Investigation of Ln-binding was also conducted with other metalloproteins (RTX, PTE-R18) to exploit these scaffolds for specific applications.
RTX. RTX corresponds to the block V of the repeats-in-toxins (RTX) domain of adenylate cyclase.^127–129 It is a Ca²⁺-binding protein that can bind up to eight Ca²⁺ ions. In the apo-state, the protein is intrinsically disordered and folds upon metal binding. The folded protein adopts a parallel β-helix structure. Ca²⁺-binding sites are located between turns of successive layers and are composed of one Asp residue and other donor atoms from the main chain to provide CN = 7. Recently, the capacity of RTX to bind Ln³⁺ was investigated more thoroughly for potential applications in Ln³⁺ recycling (see 6.3 Ln recovery and separation).¹²⁹ Circular dichroism experiments demonstrated that Ln³⁺ induced the folding of the protein. However, it also suggested that the protein might adopt a different structure relative to the Ca²⁺-induced folding, and the exact Ln-coordination sphere in this protein is still unknown.
PTE-R18. PTE-R18 is a mutant from a Zn²⁺-dependent phosphotriesterase enzyme (PTE) designed for the hydrolysis of 2-naphthyl hexanoate. Jackson and coworkers investigated the use of this scaffold for Ln³⁺-binding to design an Ln–enzyme and exploit the Lewis acid properties of Ln³⁺ (see Catalysis).²¹ Crystal structures obtained by co-incubation of the apo-protein and Ln³⁺ (Eu, Gd) showed the positioning of the Ln³⁺ ion in the Zn²⁺-binding site coordinated by two His, one Asp (bidentate), a water molecule and a chloride ion (CN = 6). This low CN is surprising given the preference of Ln³⁺ for a higher coordination number, and could explain the low occupancy (∼20%) of the metal-site by Ln³⁺.

2.6 Discussion and guidelines

A large variety of scaffolds were reported for Ln³⁺-binding, which range from peptide sequences of a few AAs to proteins, that either fold upon metal-binding or that are pre-folded. For unfolded scaffolds or pre-folded scaffolds of intermediate size, additional stabilising interactions beyond the binding site are required to obtain a stable assembly, which in turn may improve Ln affinity (see Thermodynamic stability). In particular, the design of pre-folded scaffolds with a predisposed Ln³⁺-binding site using short or intermediate-size peptides requires finding a balance between the stability of the structure, which relies mostly on hydrophobic interactions, and the introduction of metal-binding residues that can destabilise the scaffold. With this aim, two strategies were employed, sometimes in combination: (i) to put the metal binding site at a position that ensures a minimal perturbation of the scaffold (e.g. on the surface vs. inside the hydrophobic core of coiled coils, or at different positions within the hydrophobic core of coiled coils); (ii) to select amino acids known for their propensity to favour the formation of specific secondary structures. Overall, only a few studies investigated in detail the stability of their constructs and the impact of Ln³⁺-binding site incorporation and Ln³⁺ binding. These data could be helpful to guide the optimisation of scaffolds and validate the strategies followed.

All Ln³⁺ coordination sites reported meet Ln³⁺ preference for a high coordination number (CN 8–10) and hard Lewis bases. This includes either natural (Asp, Glu, Gla) or unnatural (Ada_n, Ed3a₂) negatively charged AAs and other AAs with O-donor atoms either in their side chains (Asn, Gln) or main chain (carbonyl of peptide bond). When such high CN cannot be achieved solely by the peptide or protein residues, water molecules are found coordinated to Ln³⁺.

Control over the hydration number is important for applications in imaging (see 6. Applications). A few general principles can be drawn upon the analysis of the CN and q of EF-hand motifs with distinct amino acid patterns (identity and position of residues 1, 3, 5, 9 and 12), which constitute the larger database on this topic. First, in EF-hand the presence of Glu₉ (rather than Asp₉) warrants a coordination sphere devoid of water molecules (q = 0), which, as mentioned above, can be sought to optimize luminescent emission (see 6.1 Luminescent tags and probes). Instead, Asp₉ is never directly involved in Ln³⁺-binding, but instead hydrogen-bonded to an Ln³⁺-bound water molecule. Furthermore, higher CN (9–10) are observed in LanM, which are proteins evolved by Nature to bind Ln³⁺ in physiological conditions, relative to Ca²⁺-binding proteins or engineered LBTs (CN = 7–8). This does not stem merely from a change in the pattern of coordinating amino acids, but rather from second-sphere factors including a fine-tuning of backbone conformation and orientation of amino acid side chains. From the analysis of the data obtained with non-EF-hand scaffolds, two additional general principles can be underlined. First, the denticity of Ln-binding residues impacts q, as is seen with short peptides (e.g. Ada_nvs. Ed3a₂) and coiled coils (e.g. Glu vs. Gla). Second, the environment in which the Ln-binding site is placed participates in the tuning of q, as evidenced in 3SCC, for which Ln³⁺-hydration state depends on the position of Ln³⁺-binding site in different regions (e.g. the more solvent-exposed termini vs. the buried hydrophobic core of the coiled coils), and on second-sphere effects (e.g. tuning of steric hindrance of second sphere residues). Thus, to summarize, control over the hydration number can be achieved by tuning three key parameters: (i) the denticity of Ln-binding residues and the length of their side chain; (ii) the position of the binding site in the scaffold; and (iii) the steric hindrance of second sphere residues. In addition, works on EF-hand motifs underline the influence of a fourth key parameter, more difficult to control, which is the spatial positioning of AAs side chains that allows AAs to bind in a mono- or bi-dentate fashion, and so participate in the fine-tuning of Ln³⁺ hydration number.

3. Thermodynamic stability

3.1 Generalities

3.1.1 Stability constants. The thermodynamic stability, or affinity, of metal–ligand complexes ML_n can be expressed by the overall stability constant β_n, which is the product of stepwise stability constants K_n (eqn (3)):The measurement of such a pH-independent formation constant is accessible through, for instance, potentiometric measurements of both apo-ligand and metal–ligand complex pK_as, which represent the most accurate method to determine metal-binding affinities. However, for practical reasons (e.g. poor stability over a wide pH range and a large number of protonable sites), potentiometric titrations can be rarely performed on long peptides and proteins. Moreover, K and β values have poor significance under biological conditions, as they correspond to the affinity of a fully deprotonated ligand (i.e. at very high pH).

Nevertheless, conditional stability constants, ^cK, at a given pH can be derived from the stability constant K taking into account the protonation constants (K_a,n) of the ligand (eqn (4)).

Conditional stability constants can be also determined through either direct metal–ligand titrations or competition experiments using a second chelator of known affinity in the conditions used.^130,131 In this case, an apparent stability constant, K_app, is measured in the presence of the competitor in a given buffer, from which ^cK can be derived by taking into account the concentrations (C) and the conditional stability constants of the competitor (comp) and the buffer (eqn (5)).

^cK = K_app(1 + ^cK_compC_comp + ^cK_bufferC_buffer) (5)

The stability constants of peptide and protein scaffolds for Ln³⁺ ions found in the literature and reported in this review have not been corrected for the buffer contribution, and hence they are apparent (K_app). This is also due to the lack of reliable stability constants for Ln³⁺–buffer systems (see Section 3.1.2). In addition, it is noteworthy that the buffer concentration has a considerable impact on the K_app of the complex under investigation: for instance, based on eqn (5), in the presence of a buffer with ^cK_buffer = 10³, the K_app will be ∼10-fold or ∼100-fold lower than ^cK when the buffer concentration is 10 or 100 mM, respectively.

Another important point to keep in mind is that when working with peptides and proteins, the techniques used to determine the apparent stability constant can monitor the metal binding and/or the induced folding of the scaffold. For instance, CD (circular dichroism) can only inform indirectly on metal binding, since this technique measures metal-induced changes in the scaffold folding. Variations of chemical shifts in NMR spectroscopy are due to both metal-binding and subsequent conformational changes; similarly, ITC (isothermal titration calorimetry) measures the heat exchanged for all events happening during the titration, including conformational changes and metal-binding. To some extent, luminescence titration is also dependent on folding and metal binding. Ln³⁺–luminescence increases upon Ln³⁺-binding due to the substitution of water molecules. Thus, in general, Ln³⁺–luminescence is also sensitive to induced local folding events (i.e. the formation of the binding site), but not to conformational changes that are more distant.

It is important to underscore that K_app values should be compared only if they were determined in the same conditions (metal and ligand concentration, buffer, pH, etc.) and through the same technique, and refers to complexes with the same metal–ligand stoichiometry. For a more accurate comparison of complexes with different metal–ligand stoichiometry or different metal/ligand concentrations, the use of pM (eqn (6)) would be recommended.^132,133

In the following, the thermodynamic stability of different Ln³⁺-binding peptide and protein scaffolds is discussed. The reported apparent stability constants have been determined with a variety of methods and conditions, preventing rigorous comparisons but allowing discussion and speculation about trends and binding mechanisms.

3.1.2 Ln³⁺ coordination by pH buffers. As mentioned in the former paragraph, knowledge of the stability constant of Ln³⁺ complexes with pH buffers would be beneficial to take their contribution into account and derive buffer-independent binding constants. Only a few data are available in the literature on the affinity of the most common buffer for Ln³⁺ (Table 7). Azab and coworkers studied the thermodynamic stability and speciation of Ln³⁺–buffers systems using potentiometry.^134–136 They suggested the formation of 1 : 1, 1 : 2 Ln³⁺–buffer complexes as well as [Ln(buffer)(OH)_n] species. The stability constants (log β) reported for 1 : 1 Ln³⁺–buffer complexes range between 3 and 4, but contrasting and hence no reliable values have been reported for the same Ln³⁺–buffer system.

Table 7 Stability constant (log β) of 1 : 1 Ln³⁺–buffers systems

Ln^3+	Buffer
	MOPSO	MES	PIPES	HEPES	POPSO	MOPS
a Ref. 134. b Ref. 135. c Ref. 136. d Values have been calculated by correcting the log β reported in ref. 137 for the buffer pKas (log β + pKa): MES, 6.27; PIPES, 7.14; HEPES, 7.56; MOPS, 7.18.^138
La^3+	3.34^a	3.40^a	n.d.	3.33^a	n.d.	n.d.
Ce^3+	3.31^a	3.36^a	n.d.	3.40^a	n.d.	n.d.
Pr^3+	3.36^a	3.39^a	4.11^b	3.44^a	n.d.	n.d.
	4.18^b	4.26^b
Eu^3+	3.39^a	3.38^a	4.22^b	3.43^a	3.28^c	3.68^d
	4.24^b	4.27^b	2.51^c	3.04^c	3.68^d
		2.57^d	3.04^d	3.4^d
Gd^3+	3.27^b	3.8^b	3.57^b	n.d.	n.d.	n.d.
Dy^3+	4.09^b	4.19^b	4.03^b	n.d.	n.d.	n.d.

---

More recently, the interaction of some buffers with Eu³⁺ was investigated employing NMR and time-resolved laser-induced fluorescence spectroscopy (TRLFS).¹³⁷ However, it is not clear how the values found in this study were determined. Overall, there is little agreement on the stability constants of Ln³⁺–buffers systems, for which reliable values need to be determined. Notwithstanding, given the range of their stability constants, buffers could compete with peptides and proteins with low affinity for Ln³⁺ ions. Therefore, the choice of the pH buffer and its concentration should be considered carefully.

Furthermore, buffer molecules could replace loosely bound water in the coordination sphere of Ln³⁺ complexes, leading to the formation of ternary species, which has been investigated only in a few cases. For instance, Delangle and coworkers compared the number of Ln³⁺-bound water molecules (hydration number, q) of a Tb³⁺-cyclic decapeptide complex in HEPES buffer (10 mM, pH 7) and in the absence of buffer at pH 6.5.⁴¹ In both cases, they obtained similar Tb³⁺ luminescence lifetimes and hydration numbers (τ_H2O = 1.65 ms, τ_D2O = 6.99 ms, and q = 2.0 in HEPES buffer; τ_H2O = 1.68 ms, τ_D2O = 6.28 ms and q = 1.9 in absence of buffer), suggesting the absence of buffer molecules in the first coordination sphere of Tb³⁺.

3.1.3 Ln³⁺ coordination by peptides and proteins. The affinity of peptides and protein scaffolds for Ln³⁺ ions spans over a very wide range (log K_app ∼ 2–13; Fig. 15). Table 8 gives some examples of log K_app for Tb³⁺ around neutral pH. Most of the scaffolds, regardless of their origin (natural or artificial) and size, have an affinity in the micro- to nanomolar range (log K_app = 5–9). The highest affinity is exhibited by large scaffolds, such as the natural LanM⁸¹ and the de novo designed TIM barrel protein (TFD-EE N6W),¹²⁶ and by short peptides (6 amino acids) featuring multidentate unnatural amino acids.⁸⁷ In the following paragraphs, several factors influencing the affinity of each scaffold will be described and discussed.

Fig. 15 Affinity range of different Ln³⁺-binding peptide and protein scaffolds around neutral pH (6–7.5). Colour code is based on the scaffold size (as in Fig. 2). Abbreviations: nAA, natural amino acids; unAA, unnatural amino acids; CC, coiled coils; LBT: Ln-binding tag; HLH, helix-loop-helix.

Table 8 Apparent stability constants of different Tb³⁺ complexes with peptide and protein scaffolds

Scaffold (peptide/protein)	log Kapp (Tb^3+)	Ref.
a Bis-Tris 20 mM pH 6, NaCl 25 mM. b HEPES 10 mM pH 7.5, NaCl 10 mM. c HEPES 10 mM pH 7, NaCl 100 mM. d HEPES 10 mM pH 7. e HEPES 10 mM pH 6.9. f HEPES 10 mM pH 7. g PIPES 25 mM pH 6.8, KCI 100 mM. h HEPES 10 mM pH 7, NaCl 100 mM. i MOPS 30 mM pH 7.2, KCl 100 mM. j HEPES 10 mM pH 7, NaCl 100 mM. k HEPES 25 mM pH 7.5, NaCl 100 mM.
RTX	4.6^a	129
Ln-fingers (LF4)	4.9^b	103
EF-hand (LBT1)	5.1^c	139
3SCC (MB1-1)	5.3^d	99
HLH (P3W)	5.4^e	80
Cyclic decapeptide (PA)	6.3^f	41
CaM (bovine, site I)	8.8^g	82
LBT–protein conjugate (ILB1β-L3)	8.9^h	65
LanM (Mex-LanM)	11.1^i	63
unAA short peptides (P^HD2)	12.7^j	87
TIM barrel (TFD-EE N6W)	13.1^k	126

---

3.2 EF-hand motifs

Thanks to the large variety of natural and engineered EF-hand motifs, it is possible to evaluate the effect of several structural factors influencing their affinity for Ln³⁺ ions.

3.2.1 Influence of the protein scaffold. All naturally-occurring 12-residue EF-hand motifs isolated from their either Ca²⁺- or Ln³⁺-binding native protein scaffolds show similar affinities in the high micromolar range (log K_app ≈ 5). Natural Ca²⁺-binding proteins and LBT–protein conjugates bind Ln³⁺ ions with at best log K_app ≈ 8–9 at pH ∼ 7. LanMs display the highest affinity (log K_app ≈ 10–12 at pH ∼ 7) among naturally-occurring Ln³⁺-binding scaffolds. The structural determinants of such enhanced affinity are not clear. Based on the current literature,^62,81,83 it seems that it results from second-sphere effects, including different loop backbone conformation and hydrogen bonding networks that ensure an optimal orientation of coordinating amino acids within the loop.

The loss of 5 orders of magnitude in Tb³⁺ affinity when EF1-3 loops are isolated from their native Mex-LanM scaffold (Δlog K_app ≈ 5, see entries 1 and 2, Table 9) highlights the momentous impact of a suitably folded, but not necessarily pre-disposed (note the LanM is disordered in the apo-state) scaffold on the affinity.

The important role of the core helical bundle in LanM has been further underscored by the recent study of a Mex-LanM fragment encompassing the highest-affinity and cooperative EF2-3 domains devoid of flanking helices, which showed much weaker affinity (log K_app < 5) than the whole protein.¹⁴⁰ Besides, another recent study has explored the impact of replacing EF-loops in CaM with those found in Mex-LanM.¹⁴¹ Interestingly, this chimaera (LanM-GCaMP) showed a weak conformational response to Ln³⁺ (from the micro- to the milli-molar range, depending on the Ln³⁺). Nevertheless, the authors were able to obtain a modified CaM scaffold (LanTERN) with a conformational transition similar to LanM (log K_app = 10.3 for La³⁺) by introducing key mutations in CaM loops that are considered responsible for enhanced Ln³⁺ selectivity in LanM (i.e. Pro₂ and Asp₉, see Section 3.2). Altogether, this evidence suggests that to further boost the Ln³⁺-binding affinity of EF-hand proteins, the loop sequence, including non-coordinating amino acids, and the protein folding must be optimised simultaneously.

3.2.2 Influence of primary sequence and post-translational modifications. The primary sequences of LBT have been artificially evolved mainly to optimize Tb³⁺ luminescence and affinity. Hence, they represent an outstanding dataset to derive structure–affinity relationships. In particular, the impact of specific modification can be quantified by comparing structurally analogous peptides (Table 9). For instance, the incorporation of hydrophobic amino acids (Phe, Tyr, Ile) at both N- and C-terminal to constrain the loop contributes to almost 1 order of magnitude of affinity (entries 4 vs. 3 in Table 9).⁷² A similar gain in affinity (about 1.5 order of magnitude) was accomplished by the cyclization of the EF-loop termini through the formation of disulfide bonds between Cys suitably located upstream and downstream of the loop (entries 8 vs. 7 in Table 9).^139,142

Table 9 Apparent stability constants of Tb³⁺ complexes with distinct EF-hand motifs

---

Furthermore, as already mentioned for natural proteins, the insertion of LBT within a protein scaffold increases the affinity up to log K_app ≈ 8–9 (entries 8–9 in Table 9). Of note, the rigidity of the regions flanking the loops is also crucial to achieving such an enhancement of the affinity, which is not accomplished, for instance, in the chimeric helix-loop-helix domains (log K_app ≈ 5–6) likely due to the high flexibility of the flanking helices (entry 11 in Table 9).⁸⁰

The pattern of coordinating amino acids in the EF loop also influences the affinity. Notably, the affinity increases of more than 1 order of magnitude (entries 5–6 vs. 4 in Table 9) when (i) Asn₃ is mutated to Asp (entries 5 vs. 4 in Table 9), likely due to the more negative charge of Asp relative to Asn, and (ii) Asp₉ is replaced by Glu (entries 6 vs. 4 in Table 9), which serves as bidentate ligands displacing the Ln³⁺-bound water hydrogen-bonded to Asp₉.

In order to develop kinase/phosphatase- and nitration-responsive probes (see 6.1.5 Probe design and selected applications), the group of Zondlo has also widely explored EF-hand motifs featuring phosphorylated and nitrated amino acids, including phosphotyrosine (pTyr), phosphoserine (pSer), phosphothreonine (pThr) and 3-nitrotyrosine (nTyr). These studies have shown that Glu₉ can be replaced by pSer, pTyr and nTyr with minor impact (Δlog K_app < ±1) of the EF-loop affinity (entries 1–5 in Table 10).^144–146 A similar effect was also observed when Asp residues in positions 1, 3 or 5 were replaced by Cys sulfinic acid (entries 7–9 in Table 10).¹⁴⁷ Remarkably, an affinity gain of about 1 order of magnitude was reported when Asp₉ was replaced by pThr (entry 6 in Table 10).¹⁴⁸ Interestingly, a similar affinity gain was also observed for C-terminally truncated 9-residues loops bearing pThr₉ or pSer₉ (entries 10–12 in Table 10).¹⁴⁸ Unfortunately, the lack of structural studies prevents a detailed description of the Ln³⁺ coordination sphere in these modified EF-hand motifs. Noteworthy, the impact of phosphorylated or nitrated amino acids in natural EF-hand proteins has not been investigated yet.

Table 10 Apparent stability constants of Tb³⁺ complexes with different EF-hand motifs bearing modified amino acids (in red). pSer = phosphoserine; pTyr = phosphotyrosine; nTyr = 3-nitrotyrosine; pThr = phosphothreonine

---

With respect to Ln³⁺ binding to phosphorylated peptides, it is also worth pointing out that a micromolar affinity for Tb³⁺ (log K_app = 6.5 at pH 7) has been reported for a fragment of the protein α-synuclein (119–132) with an EF-hand-like amino acid pattern and bearing a pTyr residue (Asp₁, Asp₃, Glu₅, pTyr₇, Glu₁₂).¹⁴⁹

3.3 Other scaffolds

3.3.1 Short and intermediate-size scaffolds. In short and intermediate-size peptides, three main parameters influenced the affinity for Ln³⁺: (i) the set of coordinating amino acids; (ii) their denticity and (iii) scaffold stability.

Among the short peptides reported in the literature (cyclic or linear), the one with the best affinity was obtained when using unnatural amino acids for Ln³⁺-binding (P^HD2 Ada₂ and Ed3a₂, log K_app = 12.7, Fig. 5).⁸⁷ Compared to Glu and Asp, these unnatural amino acids have a higher denticity, which is more favourable for Ln³⁺ binding. Furthermore, the peptide backbone designed to adopt a β-turn-fold upon metal binding provides additional stabilising interactions that make the peptide a better ligand than aminodiacetate groups separated by long alkyl chains (n > 2), but not as good as ligands such as EDTA for which the formation of 5- and 6-membered chelate rings result in a large stabilising chelate effect.⁸⁵

The influence of the coordination sphere on the apparent affinity constants is evidenced for LF scaffolds, in which modification of the coordination sphere from (Asp)₂(Glu)₂ (LF4) to AspAsn(Glu)₂ (LF7) decreased the apparent affinity constant for Eu³⁺ from log K_app = 4.6 to log K_app = 4.0. It is also apparent in the 3SCC designed by Peacock and coworkers when comparing the affinity of MB1-1 and CS1-1 for Tb³⁺ (log K_app = 5.3 vs. 4.6, respectively).⁹⁹ The additional (Asn)₃ layer in MB1-1 contributes to the slightly higher affinity observed relative to CS1-1. However, a change of hydration number for a given coordination sphere does not impact the apparent affinity constant, as can be seen in the 3SCC MB1-n series (n = 1–4), which have comparable affinities for Tb³⁺ (log K_app = 5.2–5.5) and hydration number ranging from q = 0 (MB1-2 and MB1-3) to q = 3 (MB1-1).⁹⁹

The influence of denticity on Ln-affinity is well illustrated by the 2SCC described by Hodges when using Gla instead of Glu.⁹⁵ The affinity of E₂(15,20)_ox has a log K_app ≈ 2 for La³⁺, whereas the one of Gla₂(15,20)N_ox has a log K_app = 6.2 for La³⁺, which can be linked to the higher denticity of the latter.

Finally, the link between scaffold stability and Ln-affinity was investigated in the design of the Ln-finger series. Changes in the sequence to improve the formation of secondary structures resulted in improved apparent affinity constants from log K_app < 2 for Eu³⁺ of LF1, to log K_app = 3.7 of LF3.¹⁰³ Increasing the size of the binding site by deletion of two AAs (LF4) gave the best apparent affinity constant for Eu³⁺ with log K_app = 4.6. In the MB1 series shifting the position of the binding site within the 3SCC scaffold impacted the stability of the assembly (see 2.4.1 Coiled coils).⁹⁹ However, this did not significantly impact the apparent affinity constants for Tb³⁺, which were all in the range log K_app = 5.2–5.5. The addition of an additional heptad in MB1-1L also did not impact the apparent affinity constant for Tb³⁺ (log K_app = 5.0).¹⁰²

3.3.2 Proteins. Proteins without EF-hand motifs cover a wide range of affinities for Ln³⁺ with log K_app ranging from 4 to 13. This large range of affinities can be explained by Ln³⁺-binding sites that are either not fully optimised for Ln³⁺-binding (e.g. in Ca²⁺–proteins), or found in natural proteins whose biological function may require affinity in an intermediate range (e.g. LanP, LanD).

The Ca²⁺-binding protein RTX has an apparent affinity constant for Ln³⁺ in the range log K_app = 4.1–4.6 at pH 6, and may bind 4 to 7 Ln³⁺ per protein.¹²⁹ Note that these values were obtained for a fusion protein, in which RTX is fused with two fluorescent proteins for detection of folding upon metal binding by FRET. The affinity of LanP for Ln³⁺ is log K_app = 6.0 and this protein binds 3 to 4 Ln³⁺.¹²² It was identified in Ln-utilizing organisms and due to its Ln³⁺-binding ability, it was hypothesized that LanP could be of importance in Ln-dependent methylotrophy. Similarly, the apparent affinity constant of the monomeric protein LanD for Ln³⁺ ranges from log K_app(La³⁺) = 5.7 to log K_app(Nd³⁺) = 6.6.¹²⁵ Because LanD and LanM are part of the same gene cluster involved in Ln³⁺ transport and utilisation, it was proposed that LanD could be responsible for transferring Ln³⁺ to LanM. This hypothesis was supported by the metalation of the modified fluorescent LaMP1¹²³ in the presence of La³⁺–LanD, and by ITC titration experiments that showed a specific 1 : 1 interaction between apo-Mex-LanD and apo-Mex-LanM with a dissociation constant of 4 μM of physiological relevance. As of now, little is known about the coordination environment of Ln³⁺ in these proteins, except for LanD, and the stability of these scaffolds. This does not allow the identification of specific features that could explain the affinity range observed.

When the protein scaffold and Ln³⁺-binding site are optimised, high affinity can be obtained by design, as is the case for the computationally designed TIM barrel that displayed the highest affinity for Ln³⁺.^126,150 The affinity of the mutant with a (Glu)₄ binding site and Trp for Tb³⁺-sensitisation (TFD-EE N6W) was first suggested to be log K_app ≈ 15 by competitive titration with EGTA.¹²⁶ However, when the same experiments were reproduced with a higher concentration of Tb³⁺–EGTA relative to TFD-EE N6W and longer equilibration times, the apparent affinity constant for Tb³⁺ was rather found to be log K_app = 13.1.¹⁵⁰ Of note, this stressed the importance of taking into account kinetic considerations (see 5. Kinetic stability), since when protein metalation, or metal ion exchange with a given competitor, is slow, reaching the thermodynamic equilibrium may take days as it is the case for this example. Nevertheless, it also shows that in a rigid scaffold (T_m > 95 °C) with a predisposed binding site suitable for Ln³⁺-binding, high affinity can be achieved by rational design.

Ln-ADH and its mimics stand apart since Ln-binding depends not only on the protein but also on the presence of the cofactor PQQ. Early work on Ln-ADH extrapolated apparent affinity constant based on activity assays and predicted an affinity in the low μM range for Ln³⁺ in the presence of PQQ (log K_app ∼ 6).^112,119 More recently, Zeymer and coworkers studied the affinity of a PedH mutant for Ln³⁺ in the absence of PQQ in the active site.¹⁹ They determined an apparent affinity constant log K_app = 7.0 for Tb³⁺ by luminescence titration, and log K_app = 6.2 for Ce³⁺ based on competition with Tb³⁺. Similarly, for the engineered PqqT scaffold, a log K_app = 6.2 was obtained for the mutant PqqT–K₁₄₂D containing PQQ (noted PQQ⊂PqqT–K₁₄₂D).²⁰ This represents a 10-fold improvement compared to the wild-type scaffold (PQQ⊂PqqT) and to the PQQ⊂PqqT–K₁₄₂A mutant. The presence of PQQ inside the protein was required for Ln³⁺-binding, as demonstrated by controlled experiments performed with the apo-scaffolds (PqqT, PqqT–K₁₄₂A, and PqqT–K₁₄₂D).

3.4 Effect of the pH

Ln recovery requires ligands with high affinity and selectivity for Ln³⁺ ions at pH ranging from pH < 1 to pH = 5–6 depending on the source (see 6. Applications).^151,152 The capacity to bind Ln³⁺ ions in acidic conditions is an overlooked feature of peptides and proteins, mainly because (i) protein folding is generally sensitive to acidic pH and (ii) typical Ln³⁺-binding residues (Asp, Glu) become protonated below pH 3–4, weakening Ln³⁺ binding.

Nonetheless, some works investigated the pH sensitivity of their scaffolds. 3SCC loose their Ln³⁺-binding ability below pH ∼ 4–5.⁹² CaM and LBTs start releasing Ln³⁺ below pH 6 and show no binding below pH 4.^153,154 Similarly to small chelators such as EDTA and DTPA, LanMs are able to bind Ln³⁺ ions down to pH ∼ 2.5, retaining approximately nanomolar affinity at pH 4 (log K_app ≈ 8–9) and 5 (log K_app ≈ 9–10).¹⁵⁵ The reason behind the higher pH stability of LanM relative to CaM and LBTs is not understood. Interestingly, the RTX scaffold has been recently reported to retain partial Ln³⁺-binding below pH 2. The protein could recover up to 20% of Nd³⁺ from a synthetic NdFeB magnet solution containing Nd, Dy, Fe and Co at pH < 1.¹²⁹

On the other end of the pH scale, studies on 3SCC scaffolds also showed that such moderate-affinity ligands undergo partial de-complexation of Ln³⁺ at basic pH (> ∼8.5), likely due to competing formation of lanthanide hydroxides.⁹²

More systematic studies will be necessary to understand the different pH-dependence of Ln³⁺ binding to different peptide and protein scaffolds.

3.5 Discussion and guidelines

Among natural peptide and protein scaffolds, LanM stands out for its log K_app ≈ 10–12. Although the structural factors for this remarkably high affinity have not been fully elucidated yet, the constraint imposed by the global protein scaffold, even if not pre-folded, is paramount. Indeed, isolated EF-loops with the same potential coordinating residues are much weaker ligands.

All rationally designed scaffolds with natural amino acids (i.e. short peptides, LFs and CCs), except for the TIM barrel (log K_app(Tb³⁺) = 13.1), have lower or at best comparable, affinity than natural scaffolds. As was noted earlier, these small to intermediate scaffolds are sensitive to the insertion of a metal-binding site, which destabilises the structure and in turn, may decrease Ln³⁺-affinity. Peptides tend to be flexible and explore a large conformational space. Ln³⁺-binding may improve the folding of the peptide; however, it is generally not sufficient by itself to compensate for the entropic cost of restricting the peptide conformational space. Optimisations of the peptide sequence to improve the formation of secondary structures can help fold the peptide, and hence, increase the affinity for Ln³⁺. This is observed qualitatively by the successive optimisations performed on Ln-fingers. It is also the case for the de novo designed TIM barrel, whose packing was optimised to rigidify the structure, and which displays a high stability (T_m > 95 °C) and affinity for Ln³⁺. Similarly, even though LanMs are not pre-folded, their affinity for Ln³⁺ is coupled to the formation of a thermally stable (T_m > 95 °C) folded protein. Nonetheless, there seem to be limitations to this, as increasing the size and stability of 3SCC (MB1-1 vs. MB1-1L) did not result in an impact on Ln³⁺-affinity.¹⁰² This could be due to the position of the binding site at the N-terminal, whereas the elongation was at the C-terminal. Interestingly, the correlation between the stability of the scaffold and metal-binding affinity has been observed for other metal-binding peptides. Studies performed on β-sheet WW-domain-like metal binding mini-proteins showed an improved affinity for Zn²⁺, Cu²⁺ and Ni²⁺ by improving the stability of the scaffold.¹⁵⁶ Moreover, a study on Hg- and Cd-binding coiled coils demonstrated a direct correlation between the stability of the coiled coil scaffold and metal ion affinity.¹⁵⁷ Altogether, these data validate the strategy of optimising the scaffold structure in order to improve Ln³⁺-affinity.

An alternative strategy is to use unnatural amino acids. Indeed, affinities comparable to that of LanM or TIM barrel were achieved in very short peptides (6 AAs) by combining Ada_n and Ed3a₂ (P^HD2, log K_app(Tb³⁺) = 12.7). Even in this case, the primary sequence played an important part since residues that favour the formation of a β-turn upon metal-binding were chosen to bring additional stabilising interactions.

4. Selectivity

The selectivity of peptide and protein scaffolds for Ln³⁺vs. other metal ions is a key aspect for their applications in both environmental and biological samples, as well as for understanding Ln biochemistry. In particular, the selectivity among Ln³⁺ and against actinides (An³⁺) is of interest for Ln³⁺ mining and separation, while the selectivity for Ln³⁺ over Ca²⁺ and other physiological cations (e.g. Zn²⁺ and Cu²⁺) is particularly relevant in the biological context.

The selectivity between two different metal ions can be quantified by calculating the selectivity factor, Δlog K (or Δlog β), which corresponds to the difference between the affinities of a certain ligand for distinct metal ions. In the following, we will express the Ln³⁺-selectivity, Δlog K^Ln (or Δlog β^Ln) as the difference between the lowest and the highest log K^Ln along the series (eqn (7)).

Δlog K^Ln = log K^Ln_max − log K^Ln_min (7)

Similarly, the selectivity for Ln³⁺vs. Ca²⁺, Δlog K^Ca (Δlog β^Ca), is expressed as the difference between the log K^Ln for a certain Ln³⁺ and log K^Ca (eqn (8)).

Δlog K^Ca = log K^Ln − log K^Ca (8)

4.1 Selectivity among Ln³⁺

Prior to peptides and proteins, it is useful to discuss the selectivity of small chelators (Fig. 16 and 17A), which can be classified into four types.^158–161

Fig. 16 Chemical structure of the Ln³⁺ chelators mentioned in the text.

Fig. 17 Selectivity of different ligands across the Ln³⁺ series. (A) Small chelators and (B) peptides and protein scaffolds. Note that log β and log K_app are reported in (A) and (B), respectively. Values for LanM are referred to (i) Mex-LanM at pH 7.2 measured by CD spectroscopy; (ii) Mex-LanM at pH 5 measured by CD spectroscopy; (iii) Hans-LanM at pH 5 measured by CD spectroscopy; (iv) Mex-LanM at pH 5 measured by UV-vis spectroscopy. The values plotted are reported in Table 15.

Most ligands, including EDTA (Fig. 16), show increasing stability constants across the Ln series (type I, black dots, Fig. 17A), commensurate with the increase of Ln³⁺ Lewis acidity. For other ligands, such as DTPA and DOTA (Fig. 16), the affinity increases along the first part of the series, reaches a maximum and then remains constant or even decreases for late Ln³⁺ ions (type II, red dots, Fig. 17A). This is the result of the ligand steric hindrance, which impairs to accommodate the smaller Ln³⁺ ions.^158,159,162

An uncommon behaviour is shown by a few ligands, including Macropa (Fig. 16), for which a reverse-size selectivity is observed due to a fairly rigid scaffold better suited for larger than smaller Ln³⁺ ions (type III, blue dots, Fig. 17A).¹⁶⁰ Finally, a biphasic selectivity trend (type IV, green dots, Fig. 17A) has been recently reported with ligands such as Macrotripa (Fig. 16) that are better suited for both early and late Ln³⁺ due to a switch between a 10-coordinated conformation accommodating larger Ln³⁺ ions and an 8-coordinated conformation stabilizing smaller Ln³⁺ ions.^161,163

Overall, the selectivity trend among Ln³⁺ ions results from a compromise between electronic and steric effects, whose balance varies depending on the properties of the ligand, such as its denticity and rigidity. Moreover, the behaviour of these small chelators points out a common misconception: Ln³⁺ are not all the same, especially when bound to a chelator. Indeed, a remarkable difference can be observed in the stability constants of certain ligands for distinct Ln³⁺ ions: for instance, EDTA and Macropa show selectivity for Lu³⁺vs. La³⁺ (Δlog β^Ln) of ∼5 and ∼7 orders of magnitude, respectively.

The Ln³⁺-selectivity of only a few peptides and protein scaffolds has been reported (Fig. 17B). LBTs, Ln-fingers, RTX and LanD display a type II selectivity (Fig. 17B). The affinity of LBTs increases by almost 2 orders of magnitude from La³⁺ up to Ln³⁺ in the middle of the series (Eu³⁺–Tb³⁺) and then slightly decreases towards the end of the series.⁶⁴ Since LBTs were derived from Ca²⁺-binding loops, their preference for middle-sized Ln³⁺ ions (1.04–1.07 Å, CN = 8 in LBTs) could be explained by the closer similarity to Ca²⁺ ionic radius (1.06 Å, CN = 7 in EF loops).

Ishida and coworkers employed molecular dynamics simulations, as well as ITC and NMR measurements, in order to elucidate the selectivity trend of LBTs.¹⁶⁴ The authors found that one or two water molecules can be accommodated in the coordination sphere of large Ln³⁺ ions (La³⁺–Nd³⁺); this weakens the binding between the Ln³⁺ and Asn₅ and enhances the flexibility of the complex, resulting in reduced affinity. For Ln³⁺ from Sm³⁺ to Lu³⁺, water binding is rarely observed, correlating with the higher affinity for these smaller ions.

For LF4, the affinity increases by nearly 1 order of magnitude from La³⁺ to Er³⁺, then decreases back towards Lu³⁺, which was found to have the same affinity as La³⁺.¹⁰³ A similar trend is shown by RTX and LanD proteins, but in this case the Δlog K^Ln measured is much lower (<1).^125,129

The relative selectivity of 3SCC among Ln³⁺ has been also reported based on luminescence measurements of Tb³⁺ displacement by other Ln³⁺ ions.¹⁰¹ A bell-shaped selectivity trend was found for several 3SCC scaffolds that differ for the location of the Ln³⁺-binding site along the helices (MB1 series, Fig. 18 and Table 5). In particular, no significant discrimination was observed among medium-sized Ln³⁺ ions (Nd³⁺–Tb³⁺) for all scaffolds. For competing ions smaller than Tb³⁺, size-dependent discrimination was observed with scaffolds where the binding site is located around the centre or the C-terminal of the coiled coil. The higher promiscuity of the N-terminal binding site was attributed by the authors to its greater flexibility, which hence allows it to better accommodate also smaller Ln³⁺ ions.¹⁰¹

Fig. 18 Tb³⁺–displacement experiments followed by luminescence in three-stranded coiled coils showing a bell-shape selectivity within the Ln³⁺ series (MB1 series, Table 5). (A) Luminescence displacement, (B) Comparison of CS1-1 and MB1-1, (C) Comparison of MB1-2, MB1-3, and MB1-4. Reproduced from ref. 101.

LanP¹²² and PqqT²⁰ do not display significant selectivity among early Ln³⁺ ions (La³⁺–Gd³⁺; late Ln³⁺ were not studied). The TIM barrel TDF-EE N6W binds Eu³⁺, Gd³⁺, and Tb³⁺ with comparable affinity, while ∼10-fold weaker binding was observed with Ce³⁺.¹²⁶

Among Ln³⁺-binding peptides and proteins, LanM represents a peculiar and controversial case. In most reports by the groups of Cotruvo and Daumann, an unusual preference of LanM for larger (La³⁺–Eu³⁺) over smaller Ln³⁺ ions has been underscored based on CD and luminescence measurements,^81,123,165 which are responsive to both Ln³⁺ binding and Ln³⁺-induced conformational change (Fig. 17B, LanM(i) and LanM(ii)). It must be noted again, that similarly to RTX and 3SCC, Mex-LanM shows a very modest Ln³⁺-discrimination (Δlog K^Ln ∼ 0.5) relative to small chelators and LBTs (Fig. 17B). An enhanced selectivity (Δlog K^Ln ≈ 1.6) for early (La³⁺, Nd³⁺) vs. late Ln³⁺ (Dy³⁺) was found for Hans-LanM (Fig. 17B, LanM(iii)), thanks to a Ln³⁺ size-dependent dimerization.⁶²

Curiously, the opposite trend, i.e. an affinity increase across the end of the Ln³⁺ series, was observed when the intrinsic Ln³⁺-affinity of Mex-LanM was determined via UV-vis-NIR spectrophotometric competition experiments (Fig. 17B, LanM(iv)).¹⁵⁵ As suggested by the authors,¹⁶⁶ this could highlight a decoupling between Ln³⁺-binding and conformational change in LanM, which warrants further investigations, and underscores the importance of the method and the conditions chosen to determine and compare affinity values. Nonetheless, higher retention of early vs. late Ln³⁺ ions was observed for immobilized LanM upon pH-induced desorption.^62,167,168 Such (at least apparent) inconsistency between the affinity values measured via different techniques and in different conditions is worth further systematic investigations.

Finally, it is also worth mentioning that LanM showed higher affinity for the rare-earth element Sc³⁺ relative to Ln³⁺ ions (3-fold higher relative to Nd³⁺),¹⁶⁷ commensurate with its smaller ionic radius and higher Lewis acidity.

4.2 Selectivity over Ca²⁺

Peptide and protein scaffolds generally show selectivity for Ln³⁺ relative to Ca²⁺ (and Mg²⁺) which is classically attributed to their higher charge and coordination numbers (Fig. 19). In addition, as Ln³⁺ have higher CN than Ca²⁺ in water, Ln³⁺-binding releases a higher number of water molecules, resulting in a more favourable entropic change.

Fig. 19 Selectivity of small chelators and some peptide/protein scaffolds for Ln³⁺ over Ca²⁺. Δlog β and Δlog K_app are reported for small chelators and peptide/protein scaffolds, respectively. Values plotted were calculated from values reported in Table 16.

Among the EF-hand motifs, LanMs show a remarkably higher conformational selectivity against Ca²⁺ (Δlog K^Ca ≈ 8 for Mex-LanM)^81,155 relative to both natural and engineered EF loops (Δlog K^Ca ≈1–3).^59,82,169 In this respect, it is worth noting that Ca²⁺ seems to bind to LanM with log K^Ca = 5.5 without inducing a conformational change, which occurs upon further Ca²⁺ binding events with log K^Ca ≈ 3.⁸¹

The Pro₂ residue found in all EF loops of Mex-LanM is crucial for the conformational selectivity of LanM for Ln³⁺ over Ca²⁺. Notably, mutation of all four Pro₂ residues to Ala reduces the selectivity of Mex-LanM for Ln³⁺ against Ca²⁺ by strengthening the conformational response of Ca²⁺-binding to LanM of more than 2 order of magnitudes.⁸¹ Moreover, Pro₂ also had a key role in the Ln³⁺-specific conformational response of LanTERN chimera already discussed above (see 3.2.1 Influence of the protein scaffold).¹⁴¹ However, it is noteworthy that high selectivity over Ca²⁺ (Δlog K^Ca_a ≈ 4–6) is also observed in Hans-LanM, in which a Pro₂ residue is found only in one of the four EF-hand loops. Hence, it is likely that the unique selectivity of LanM for Ln³⁺vs. Ca²⁺ stems from minor differences in the second coordination sphere and backbone conformation.⁶²

The main difference between Ca-dependent and Ln-dependent alcohol dehydrogenases is the presence of an additional Asp in the coordination sphere of Ln³⁺. This residue appears essential for Ln³⁺-coordination and selectivity relative to Ca²⁺. Indeed, mutation of this AA to Ala in MDH and EDH enzymes led to the production of enzymes metallated only with Ca²⁺.¹¹⁰ However, the mutation did not lead to catalytically competent Ca-enzymes, which underscores that differences between Ca-dependent and Ln-dependent ADH beyond the metal coordination sphere are important for catalysis.

Among the de novo designed Ln³⁺-binding proteins, 2SCC scaffolds bearing non-canonical multidentate carboxyglutamic (Gla, Fig. 6) residues show a selectivity over Ca²⁺ comparable to that of small chelators and LanMs.⁹⁵

4.3 Selectivity over An³⁺

The groups of Deblonde and Daumann have explored the capacity of Mex-LanM to bind An³⁺ ions.^{165,170–172} LanM exhibits a higher affinity for An³⁺ over Ln³⁺ ions and Y³⁺. In particular, a ∼5-fold stronger affinity was found for Am³⁺ and Cm³⁺ relative to Ln³⁺ of similar size (Pr³⁺, Nd³⁺, Sm³⁺),¹⁷⁰ and for Ac³⁺ over Y³⁺.¹⁷¹ This trend is similar to that of small chelators such as EDTA and DTPA, which show slightly higher selectivity for Am³⁺ over Nd³⁺ (Δlog β ∼ 1.5).¹⁷³

For small complexes, the non-negligible covalency of An³⁺–ligand bonds has been generally considered accountable for the higher stability of An³⁺-complexes over their Ln³⁺-analogues, but this remains to be ascertained for An³⁺–protein complexes.¹⁷⁰

An improved selectivity for An³⁺ over Ln³⁺ ions at low pH (∼3) was obtained by Deblonde and coworkers via Mex-LanM engineering with softer Asn ligands in place of native Asp₉ in each loop (LanM 3D₉N).¹⁷⁴

A slightly higher affinity for Cm³⁺ over Eu³⁺ has also been reported for the EF-hand loops isolated from Mex-LanM.¹⁴³ Am³⁺ binding to an LBT was shown to be ∼6-fold stronger than Nd³⁺ and comparable to smaller Ln³⁺ (Tb³⁺, Eu³⁺). A higher (∼10-fold) selectivity for Am³⁺ over Nd³⁺ could be obtained by a modified LBT bearing a Cys₅ residue functionalized with a soft 2-methylene-pyridine group in place of Asp₅.¹⁷⁵

Daumann and coworkers also demonstrated that An³⁺ ions such as Am³⁺ and Cm³⁺ can replace early Ln³⁺ ions in the catalytic activity of the Ln³⁺-dependent MDH enzyme and ensure the growth of Ln³⁺-dependent bacteria in the absence of Ln³⁺ ions.¹⁷⁶

4.4 Selectivity over d-block metal ions

Despite its relevance for biomedical and bio-metallurgic applications, the selectivity of Ln³⁺-binding peptides and proteins over d-block metal ions is a largely neglected aspect. Generally, a higher selectivity for Ln³⁺ over d-block metal ions is obtained by leveraging their preference towards O- (carboxylates, carbonyls) over N-donors (e.g. histidines), and higher coordination numbers (8–10 vs. 4–6). For instance, LanP displayed no binding to Fe²⁺, Fe³⁺, or Zn²⁺.¹²² A 3SCC scaffold (MB1-2) engineered with Asn and Asp residues for Ln³⁺ binding, shows ∼6-fold higher affinity for Tb³⁺ over Cu²⁺.^98,177 A remarkably high selectivity for Nd³⁺ ions over Cu²⁺ (Δlog K ∼ 6) and Zn²⁺ (Δlog K ∼ 8) has been reported for Mex-LanM at pH 5.¹⁵⁵ On the contrary, EDTA shows higher selectivity for Cu²⁺ over Nd³⁺ (i.e. 100-fold higher affinity for Cu²⁺ over Nd²⁺) and no selectivity between Nd³⁺ and Zn²⁺, while DTPA has no selectivity between Nd³⁺ and Cu²⁺, but a ∼1000-fold higher affinity for Nd³⁺ relative to Zn²⁺.^155,173 It is worth noting that EDTA has a higher affinity for Cu²⁺ (log β^Cu = 18.8) than most Ln³⁺ ions (La³⁺–Ho³⁺) and DTPA has a higher affinity for Cu²⁺ (log β^Cu = 21.4) than the early Ln³⁺ (La³⁺–Pr³⁺). Altogether, these data point out that peptides and proteins benefit from a higher selectivity for the whole Ln³⁺ series against d-block metal ions relative to small chelators.

5. Kinetic stability

Beyond thermodynamic parameters, kinetic aspects are also important for a complete characterization of Ln^3+-binding peptides and proteins, especially for further applications, such as MRI or Ln recovery (see 6. Applications).

5.1 Lability and inertness: exchange reaction

Due to the shielding by 5s and 5p orbitals, 4f orbitals do not undergo strong ligand field stabilisation, making Ln³⁺ ions kinetically labile.

The lability/inertness is witnessed by the rate of the self-exchange reaction. Such a reaction can proceed according to two extreme pathways: associative or dissociative, corresponding to an increase or a decrease in the coordination number (CN) of the intermediate species, respectively. In turn, this depends on the CN of the starting complexes: for instance, Ln³⁺ complexes with low CN (CN ≤ 8, Gd³⁺–Lu³⁺) are expected to proceed through an associative pathway with a nine-coordinate intermediate, while those with high CN (CN ≥ 9, La³⁺–Eu³⁺) by a dissociative pathway with an eight-coordinated intermediate (Fig. 20A).³⁰ The rates of self-exchange for a certain Ln³⁺ ion depend on the relative stabilities of the eight- and nine-coordinated species for that Ln³⁺.

Fig. 20 (A) Self-exchange reaction characterized by the rate constant k_ex and related pathways. P and P* are two molecules of the same peptide. In the case of the dissociative pathway, P′ corresponds to peptide P which has some of its binding residues de-coordinated from the Ln. Hence, Ln³⁺ has a lower CN when bound to peptide P′ than P. (B) Association and dissociation reactions with their respective rate constants k_on and k_off. (C) Simplified scheme of proton-, ligand- and metal-assisted dissociation pathways (rate constants k_d).

Recently, Peacock and coworkers measured the activation parameters for water exchange in the Gd³⁺ complex with a 3SCC scaffold (MB1-1, q = 3) and proposed an associative mechanism.¹⁰² Furthermore, key insight can be deduced from the comparative NMR study of La³⁺vs. Lu³⁺ binding to the LBT3 peptide.¹⁶⁴ In particular, Ln³⁺ binding to LBT3 showed a fast and slow exchange regime for La³⁺ and Lu³⁺, respectively,¹⁶⁴ which may be explained through the different structural features of the complexes. Indeed, the La³⁺–LBT3 complex has CN = 9, with q = 1 and is structurally flexible, hence the de-coordination of the weakly-bound (relative to the peptide ligands) H₂O molecule can promote the self-exchange reaction through a dissociative pathway. Instead, the Lu³⁺–LBT3 complex has CN = 8, with q = 0, and poor solvent accessibility, hindering the possibility of undergoing an associative mechanism.

As a small number of data is available, it is difficult to generalize whether the exchange reaction proceeds via a dissociative or an associative mechanism in other Ln³⁺-binding peptides and proteins, and which parameters would influence the preference between one or the other pathways.

Furthermore, the self-exchange rate k_ex can be easily inferred by NMR experiments. For most unfolded or poorly folded peptides, a fast exchange regime characterized by a broadening and shift of the signals was reported at sub-stoichiometric Ln³⁺ : P ratio^{76,164,169,178} except for a 2SCC scaffold⁹⁵ and an HLH scaffold¹⁸ for which a slow exchange regime was observed. This is in line with the intrinsic lability of Ln³⁺ ions and the absence of constraints induced by the peptide scaffolds.

5.2 Association and dissociation rates

Beyond the exchange rate k_ex, key parameters to be defined are the association and dissociation rates, noted k_on and k_off, respectively. These two parameters are related to the thermodynamic stability constants K according to eqn (9):The k_on value can be measured experimentally by monitoring the appearance of spectroscopic signatures corresponding to the Ln³⁺ complex when formed within the right time scale (milliseconds to hours). Note that such k_on values reflect both Ln³⁺ binding to the protein and the induced folding unless the protein is pre-folded in the apo-state. Qualitative estimation of the k_on can be deduced for Ln-ADH¹⁹ and TDF-EE N6W,^126,130 for which a long incubation time (hours) was required to load the protein with Ln³⁺, indicating a slow association rate.

The k_off can be then calculated from the k_on and stability constant K values, using eqn (9). Indeed, the k_off cannot be directly determined experimentally since the release of the Ln³⁺ ion from a polydentate ligand, such as peptides and proteins, to form the fully solvated species is unlikely. Nevertheless, the complete dissociation of a Ln³⁺ from a polydentate ligand can be assisted by (i) protonation of the ligands (proton-assisted dissociation), (ii) capture by another ligand(s) (ligand-assisted dissociation), which can also be the buffer, (iii) trans-metalation by another metal ion (metal-assisted dissociation), or any combination of these three processes. It is worth noting that in these processes, the otherwise disfavoured Ln³⁺ release occurs through the alteration of the system thermodynamics, e.g. by weakening the ligand affinity (acid-assisted dissociation) or by inducing metal-transfer/exchange reaction with competing ligand or cations. Thus, although the concept of complex inertness is often equated with that of “assisted dissociation” by language abuse, they refer to intrinsically distinct processes.

Ligand-assisted dissociation rate constants (k_d), which differs from k_off, were reported by Falke and coworkers for the complex of Tb³⁺ with an EF-hand protein (galactose-binding protein, GBP) using excess EDTA as the competing ligand.¹⁷⁹ Interestingly, by measuring the dissociation rates of several mutants, these authors showed that the 9th residue of the EF-hand motif acts as a gateway that tunes the kinetics of Ln³⁺ dissociation. Notably, the Glu₉ mutant displayed the slowest dissociation rate (∼0.006 s⁻¹) compared to Asp₉ (∼0.002 s⁻¹) and Asn₉ (∼1 s⁻¹).

The chelator-free k_off of Tb³⁺-bound to Mex-LanM (∼0.02–0.05 s⁻¹) was estimated by extrapolation from k_obs values obtained by stopped-flow spectrofluorimetry in the presence of different concentrations of EGTA.⁶³ However, this approach is questionable, since the k_off value relative to a dissociative mechanism is inferred from ligand-assisted dissociation rates, for which a different, namely associative, mechanism is at play. Moreover, it is worth noting that, as mentioned for k_on values, dissociation rates may also take into account the contribution of protein unfolding induced by Ln³⁺ release.

Daumann and coworkers monitored over time the metal exchange reactions between a pre-formed Eu³⁺–LanM complex and the other Ln³⁺ ions. Interestingly, equilibration times varied from minutes for the late Ln³⁺ to hours for the early Ln³⁺, correlating with the size and Lewis acidity of the competing ions.¹⁶⁵

5.3 Discussion and guidelines

The kinetic aspects of Ln³⁺ coordination by peptides and proteins have been largely neglected and overlooked even though they are important in applications for imaging and Ln recovery. NMR is the method of choice to get a better picture of the dynamics of the system. A simple ¹H NMR titration experiment can help gain qualitative insights into the exchange rates using the diamagnetic La³⁺ and Lu³⁺, and thus determine the exchange regime on the NMR timescale. Comparison with paramagnetic Ln³⁺ can bring additional information on Ln³⁺ coordination site, and even peptides/protein solution structure.

Although it is difficult to predict the kinetic behaviour, based on the results and concepts described above, it is anticipated that the spontaneous dissociation of Ln³⁺ ions from polydentate peptides and proteins is negligible, especially in the case of buried sites within well-folded scaffolds. Notwithstanding, Ln³⁺ dissociation can be triggered upon acidification or the addition of competing ligands or cations. In the case of ligand-assisted dissociation, the presence of weakly bound water molecules in the Ln³⁺ coordination sphere is expected to promote dissociation since its replacement facilitates the formation of ternary intermediate complexes with the competing ligand.

6. Applications

Ln³⁺-binding peptides and proteins have been employed for a number of applications spanning from luminescence sensing and magnetic resonance imaging to metal ion separation and catalysis (Table 11). In the following paragraphs, applications in these four major fields will be discussed with respect to the peculiar properties of each type of Ln³⁺–peptide and protein scaffold.

Table 11 Non-exhaustive overview of applications reported for the main types of Ln³⁺-binding peptide and protein scaffolds

	Scaffold	Luminescence imaging/sensing	MRI	Metal ion separation	Catalysis (reaction)
EF-hand motifs	LBT	X	X	X
	HLH		X		X (hydrolysis)
	LanM	X	X	X
Other Ln-binding sites	Cyclic decapeptide		X
	3SCC		X
	LanD			X
	RTX			X
	TIM barrel				X (alcohols photooxidation)
	ADH or ADH mimics				X (alcohols oxidation)

---

6.1 Luminescent tags and probes

6.1.1 Ln³⁺ luminescence. Most Ln³⁺ ions display luminescent emission with several advantageous features relative to organic fluorophores and have hence gained interest for applications in optical bioassays, bioimaging and luminescence-guided surgery, as reviewed in ref. 180–185. In particular, they benefit from narrow emission bands and long luminescence lifetimes (μs–ms) at relatively invariant wavelengths, in a range spanning from the visible to the near-infrared (NIR) depending on each Ln³⁺, and they are not sensitive to photobleaching. Nevertheless, due to the low extinction coefficients (ε = 1–10 M⁻¹ cm⁻¹) of Laporte-forbidden f–f transitions, the direct excitation of Ln³⁺ emission is quite inefficient.^30,186,187 This downside is generally circumvented through the so-called antenna effect (Fig. 21).^30,186,187 This is the indirect sensitization of Ln³⁺via (i) the excitation of a suited chromophore – the antenna – to an excited singlet state (S₁), (ii) the energy transfer to an excited triplet state (T₁) via intersystem crossing, (iii) the energy transfer to an excited state of Ln³⁺, from which emission takes place. Even if it is common to observe sensitisation from T₁, the S₁ state can also be the sensitising level.¹⁸⁸ The use of an antenna also warrants larger pseudo-Stokes shifts (difference between the antenna absorption and Ln³⁺ emission wavelengths) than accessible with organic fluorophores.

Fig. 21 Scheme of the “antenna” effect: upon irradiation, a chromophore called “antenna” (blue, which can be part of or conjugated to a Ln-chelator, black) is excited to a singlet state (S₁); then, the energy is transferred to an excited triplet state (T₁) via intersystem crossing and finally transferred to an excited level of the Ln³⁺ (orange) from which emission takes place. Vibrational de-excitation by high-energy oscillators such as O–H groups is also depicted.

The efficiency of the energy transfer from the antenna to the Ln³⁺ depends on (i) the energy gap between an antenna triplet state and the Ln³⁺ excited acceptor state, which depends on the antenna–Ln³⁺ couple, and (ii) the antenna–Ln³⁺ distance. Several non-radiative pathways compete with Ln³⁺ luminescence:¹⁸⁹ (i) the antenna fluorescence and phosphorescence emissions, (ii) quenching of the antenna T₁ state by ³O₂, (iii) back energy transfer from Ln³⁺ excited state to the antenna triplet state when the energy levels are close, (iv) photo-induced electron transfer (PeT) from the excited antenna to the Ln³⁺, to which the most reducible Ln³⁺ (Eu³⁺, Yb³⁺, Sm³⁺) are particularly sensitive,¹⁹⁰ (v) non-radiative vibrational de-excitation of Ln³⁺ excited state by high-energy oscillators such as O–H, N–H and C–H groups.^39,40 As a result, Ln³⁺-bound H₂O molecules quench Ln³⁺ emission. Thus, the type and position of the antenna as well as the hydration number are key parameters influencing the luminescence of Ln³⁺ complexes, including those with peptide and protein ligands.

6.1.2 Type of antenna. Early studies using Tb³⁺ luminescence to investigate Ca²⁺-binding sites in proteins^54,191 have established that the amino acid tryptophan (Trp) is a suitable antenna for this Ln³⁺, also showing its superiority relative to Phe and Tyr. Since then, Trp has been by far the most common antenna incorporated in peptide and protein scaffolds to sensitize Tb³⁺ luminescence. Indeed, Trp (λ_ex = 280–290 nm) is characterized by a triplet state located 3700 cm⁻¹ above the emissive ⁵D₄ state of Tb³⁺, which makes it an efficient antenna for this Ln³⁺ ion.¹⁹² Trp could also be suitable for Eu³⁺-sensitisation; however, PeT from Trp quenches Eu³⁺-emission. PeT has also been described by Horrocks between Trp and Yb³⁺ in the EF-hand protein parvalbumin.¹⁹³ While for Eu³⁺ this process competes with Eu³⁺-emission, it was hypothesized that PeT could be in part responsible for the population of Yb³⁺ excited state and its subsequent emission.

Despite its straightforward incorporation in peptide and protein sequences, Trp is not an ideal antenna for applications in bioimaging, as its excitation wavelength in the UV spectral region challenges applications in samples with high absorption and fluorescence background, such as biological media.¹⁹⁴

To overcome such limitations, different unnatural antenna amino acids have been introduced. Gunnlaugsson and coworkers incorporated at the N-terminal of an EF-loop isolated from parvalbumin a 1,8-naphthalimide chromophore (λ_ex = 345 nm) able to sensitise Tb³⁺ and Eu³⁺ emission (Fig. 22).^169,178 The HPO ligand introduced into 3SCCs scaffolds (see 2.4.1 Coiled coils) was able to sensitize Tb³⁺ and Eu³⁺ upon excitation at 380 nm. Imperiali and co-workers have developed LBTs with unnatural amino acids featuring a carbostyryl-124 (Cs124) or an acridone (Acd) antenna able to sensitize Tb³⁺ (Cs124) and Eu³⁺ (Cs124 and Acd) at longer wavelengths, namely ∼340 and 390 nm, respectively (Fig. 22).¹⁹⁵ In particular, the Cs124 antenna was able to sensitise Tb³⁺ better than Trp, while Acd sensitised Eu³⁺ better than Cs124 (Fig. 22).

Fig. 22 Antenna amino acids found in Ln³⁺-binding peptide scaffolds discussed in this review (top) and in peptides with appended polydentate Ln³⁺-chelators (bottom). The excitation wavelength and the Ln³⁺ ions that each antenna is able to sensitize are indicated. The antenna moieties are highlighted in blue. Abbreviations: Trp, tryptophan; Glu, glutamate; Cs124, carbostyryl-124; Naph, 1,8-naphthalimide; HPO, hydroxyphenol oxazoline; Acd, acridone; Nal, naphthyl-alanine; Phen, phenanthroline; Anthra, anthryl-alanine; NBD, nitrobenzodiazole.

Lately, Acd and its sulfur-substituted SAcd (Fig. 22) were incorporated into Mex-LanM via genetic code expansion, achieving long-lived sensitized Eu³⁺ emission upon excitation at 390 nm.¹⁹⁶ Interestingly, SAcd showed to be a more efficient antenna than Acd due to a prolonged triplet state lifetime.

These unnatural antenna amino acids were not able to sensitise the emission from other Ln³⁺ ions including Nd³⁺, Er³⁺, Tm³⁺ and Yb³⁺, which are NIR emitters. Actually, no Ln³⁺-binding peptide or protein scaffolds have been reported bearing suitable antenna amino acids for NIR-emitting Ln³⁺ ions, which are particularly sensitive to quenching by O–H, N–H and even C–H oscillators. Thus, in addition to a suitable antenna, the design of Ln–peptides and Ln–proteins as NIR emitters will also require avoiding O–H oscillators in their first coordination sphere, and likely also N–H oscillators. To the best of our knowledge, these antenna amino acids are the sole used in Ln³⁺-binding peptide and protein scaffolds.⁸ Other antennae were described in peptide and proteins functionalised with polydentate Ln³⁺-chelators such as EDTA and DOTA as reviewed in part in ref. 23. In these cases, the antenna has been introduced through either (i) ligand derivatization, often with conjugated pyridine or picolinate groups,^197–199 or (ii) unnaturall amino acids (Fig. 22), including naphthalene (for Eu³⁺, λ_ex = 280 nm),²⁰⁰ NBD (for Nd³⁺, λ_ex = 480 nm),^201,202 anthracene (for Nd³⁺ and Yb³⁺, λ_ex = 330–400 nm)²⁰³ and phenanthroline (for Tb³⁺ and Eu³⁺, λ_ex = 300 nm).²⁰⁴ Similarly to HPO, phenanthroline served as both Ln³⁺ ligand and antenna. Such an Ln³⁺-coordinating antenna is generally preferable because its proximity to the ion enhances the energy transfer efficiency.

6.1.3 Position of the antenna. As the Ln³⁺–antenna distance is critical to achieve efficient sensitization, antennas are introduced in close proximity to the metal-binding site. For instance, in 3SCC scaffolds (Table 5) Trp has been mostly introduced within the two Ln³⁺ binding Asn and Asp residues of each strand (–NEWKD–).⁹⁸

Szabo and coworkers have thoroughly investigated how the position of aromatic amino acids (Trp, Phe or Tyr) within the EF-hand motif sequence influences Tb³⁺ emission.⁶⁶ In particular, they compared the excitation and emission spectra of Tb³⁺ bound to EF-hand peptides containing aromatic amino acids in positions 2, 4, 7 or 10. Very interestingly, they found that, regardless of which amino acid occupied positions 2, 4 and 10, the amino acid in the 7th position is responsible for Tb³⁺ sensitization due to its closer distance to the ion (∼5 Å). For instance, despite Phe and Tyr having a lower extinction coefficient than Trp, Tb³⁺ was mainly sensitized by either Phe or Tyr in the 7th position even when Trp was simultaneously present in the 2nd position. Moreover, the authors reported a favourable impact of Tyr in the 2^nd and 4^th, but not 10^th, position on the Tb³⁺ emission in peptides having Trp in the 7^th position. In particular, the results suggested that this occurs thanks to an energy transfer between excited Tyr and Trp, which in turn sensitises Tb³⁺ emission. Based on these studies, Trp₇ has been incorporated as the Tb³⁺ sensitizer in many EF-hand motif scaffolds, including LBTs^72,73 and Mex-LanM mutants.⁶³ The 7th loop position has also been used to incorporate unnatural Cs124 and Acd antenna amino acids in LBTs¹⁹⁵ and Mex-LanM (in the latter case after screening of 42 incorporation sites).¹⁹⁶ Zondlo and coworkers also showed that Tb³⁺ emission can be sensitized using Trp at the 8^th position of EF-loop-like peptides.¹⁴⁸

6.1.4 Hydration number. As already mentioned, Ln³⁺-bound water molecules quench Ln³⁺–luminescence, which is used to determine Ln³⁺ hydration number. This is why, for instance, the coordination sphere of luminescent Tb³⁺–LBT complexes was optimized to achieve q = 0 (see 2.2.2 Lanthanide-binding tags). The effect of the hydration number is also proven by the correlation between the emission intensity of Ln³⁺ bound to different 3SCC scaffolds (MB1-n series, n = 1–4) and their q values (0, 2 or 3) (see 2.4.1 Coiled coils).⁹⁹

6.1.5 Probe design and selected applications. A simple application of emissive Ln³⁺-complexes is as luminescent tags attached to a (bio)molecule or receptor. In this regard, LBTs have been developed as encodable luminescent tags that can be easily conjugated to proteins via recombinant expression and applied, for instance, to the study of protein–protein interactions by luminescence resonance energy transfer (LRET).^139,205 Recently, LanM-Acd mutants conjugated with Affibodies and complexed with Eu³⁺ were used as a secondary antibody reagent for immunofluorescence staining.¹⁹⁶

Besides, peptide and protein scaffolds with suitable antennae may serve as luminescent probes for Ln³⁺ ions. For instance, a Mex-LanM Trp mutant (T90W) has been successfully employed to measure Tb³⁺ in acid mine drainage at very low pH (∼3),⁶³ while LanM-Acd mutants were applied for Eu³⁺ detection in cell culture.¹⁹⁶

Alternatively, Walker and coworkers reported on the design of fluorescent proteins with a high concentration of negative charges at their surface for the detection of Ln³⁺ (Ln = Sm, Eu, Tb, Dy, Tm, Yb) by LRET in the micro- to millimolar concentration range.²⁰⁶ Instead of exciting an antenna, it is the Ln³⁺ that is excited and transfers its energy to the fluorescent proteins, which resulted in a twofold higher fluorescence intensity and longer emission lifetimes.

As reviewed in ref. 23 and 180, Ln³⁺-based responsive probes, whose luminescent signal changes upon the interaction with a given target, can be conceived based on the above-mentioned factors influencing Ln³⁺ luminescence, such as (i) chemical modifications of the antenna, (ii) variation of the Ln³⁺–antenna distance, for instance through conformational changes induced upon target-binding or via proteolytic cleavage, and (iii) modification of the Ln³⁺ coordination sphere and hydration number.

These principles have been used to develop several responsive probes based on peptides with appended Ln³⁺–chelator complexes.²³ For instance, Vazquez and coworkers designed a peptide probe that folds upon RNA binding, inducing the coordination of a phenanthroline antenna to an appended Tb³⁺–EDTA complex.²⁰⁴ Similarly, Sénèque and coworkers developed Zn²⁺-²⁰¹ and RNA-responsive²⁰⁷ peptide probes based on conformational changes that decrease the distance between the antenna and a Ln³⁺–DOTA complex attached to the peptide. Exploiting the distinct emission bands of Ln³⁺ couples such as Tb³⁺–Eu³⁺ in the visible and Nd³⁺–Yb³⁺ in the NIR, the group of Sénèque also developed ratiometric probes (i.e. based on the ratio between two signals) introducing two suitably positioned Ln³⁺–chelator complexes.^200,203 In order to synthesize such regio-selectively hetero-metallated complexes, they used native chemical ligation to assemble two peptides functionalized with different Ln³⁺ complexes. Of note, this strategy relies on the kinetic inertness of the Ln³⁺-complexes grafted on the peptides and has never been applied so far on peptide scaffolds with “intrinsic” Ln³⁺–binding sites.

Lately, LanM-Acd mutants have been developed as protease sensors based on the impairment of the antenna effect upon protein cleavage. To this end, protease recognitions sequences were incorporated in suited position with the LanM scaffold.¹⁹⁶

Pazos and coworkers developed peptides with appended Tb³⁺–DO3A complex (Fig. 15) responsive to (i) Ser phosphorylation, based on a decrease of the hydration number,²⁰⁸ and (ii) to Tyr nitration, relying on the quenching of Tb³⁺ emission by 3-nitro-Tyr.²⁰⁹ Quenching of Tb³⁺ emission was also exploited in the design of Cu²⁺-selective peptide probes.^194,210

Zondlo and coworkers also developed Tb³⁺ complexes with EF-hand-like peptides (Table 10) responsive to (de)phosphorylation of Tyr, Ser and Thr,^144,145,148 Tyr nitration¹⁴⁶ and Cys oxidation.^147,211 In particular, Cys, Tyr, Ser or Thr residues were introduced in place of Asp or Glu residues within EF-loops (Table 10), drastically reducing Tb³⁺-binding (log K_app < 4), while their derivatization with phosphate or sulphinyl groups restored a log K_app ≈ 4–5. Hence, in this case, the response was mainly due to a change in the metallopeptide affinity upon amino acid modification.

6.2 Magnetic resonance imaging

Clinical contrast agents (CAs) for magnetic resonance imaging (MRI), reviewed by Caravan and coworkers in ref. 212, are mostly Gd³⁺-complexes with polyaminocarboxylate chelators, such as DOTA and DTPA (Fig. 16). Gd³⁺-based CAs (GBCAs) enhance image contrast by shortening the longitudinal (T₁) and transverse (T₂) relaxation time of proximal water protons, to an extent that depends on their longitudinal (r₁) and transverse (r₂) relaxivity, as defined by eqn (10):where 1/T⁰_i is the inherent relaxation rate of the tissue and 1/T^obs_i is the relaxation rate in the presence of the GBCA. Of note, the relaxivity (r) depends on the external magnetic field strength (Table 12). Clinical MRI scanners currently use low magnetic fields (∼0.5–3 T), while higher fields (up to ∼11.7 T) are used for research purposes. Nevertheless, in order to improve signal-to-noise ratio and spatial resolution, clinical MRI is progressively moving towards higher field strength.²¹³

Table 12 Properties of Gd³⁺-complexes relevant for MRI applications

Scaffold	Peptide/protein	MW (kDa)	q	τ R (ns)	τ m (ns)	r 1 (mM^−1 s^−1)	r 2 (mM^−1 s^−1)	Magnetic field (T)	Ref.
a Values obtained by the measurement of Tb^3+ or Eu^3+ luminescence lifetimes. b Values measured at 298 K.
DOTA	n/a	∼0.5	1	∼0.08	244	4.3	5.7	7	215
				n/a	n/a	4.7	n/a	0.47	216
DTPA	n/a	∼0.5	1	∼0.06	300	n/a	n/a	n/a	215
				n/a	n/a	4.7	n/a	0.47	216
				n/a	n/a	3.5	n/a	1.4	77
				n/a	n/a	5.4	8	1.5	217
				n/a	n/a	4.2	6.8	3	217
				n/a	n/a	5.1	9.4	7	218
HLH	P3W	∼4	2	n/a	n/a	16.2	n/a	0.47	77
						21.2	n/a	1.4
LBT	m-sSE3	∼2	1	n/a	n/a	5.5	n/a	11.75	60
	sSE3		0	n/a	n/a	1.2	n/a	11.75	60
dLBT	dSE3	∼4	0.08	n/a	n/a	5.9	n/a	11.75	60
dLBT–protein	dSE3-Ubiquitin	∼12	0.08	n/a	n/a	2.3	n/a	11.75	60
	q-dSE3-Ubiquitin		1	n/a	n/a	4.2	n/a	11.75	60
	xq-dSE3-Ubiquitin		0; 1	n/a	n/a	5.0	n/a	11.75	60
Cyclic decapeptide	PA	∼1.3	2	0.4	n/a	∼30	∼40	4.7	41 and 219
						∼20	∼35	9.4
						∼20	∼35	11.75
3SCC	MB1-1	∼12	3.1	7	1.56	10	89.3	7	102
						64.3	87.9	1
	MB1-1 (2W)		0	n/a	n/a	3.9	24.2	7	100
	MB1-2		0	n/a	n/a	4.2	21.3	7	99 and 100
	MB1-3		0	n/a	n/a	4	20.9	7	99
	MB1-4		2	n/a	n/a	7.5	37.9	7	99 and 100
	MB1-1L	∼15	3.7	10	n/a	10.9	81.8	7	102
						67.4	96.5	1
Proteins	CA1.CA2	∼12	2	9.1	n/a	117	129	1.5	217
						48	88	3
						6	50	9.4
	ProCA32	∼12	0.5	n/a	n/a	33.4	44.6	1.4	218
						21.9	56.9	4.7
						18.9	48.6	7
	LanM	∼12	2	n/a	n/a	12.1	17	3	220
	LanND	∼12	2	n/a	n/a	13.2	25.1	3
						6.9	31.2	7

---

In order to increase MRI sensitivity, reduce the injected GBCA dose and hence limit toxic side-effects, GBCAs with higher relaxivity are currently sought.²¹² The approaches developed to achieve this goal have been reviewed in ref. 214. The main molecular factors affecting the inner-sphere relaxivity of GBCAs are the hydration number q, the mean residency time τ_m of Gd-bound water and the rotational correlation time τ_R (Fig. 23).²¹² Hereafter, these factors are discussed with respect to the Gd³⁺ complexes with peptides and proteins that have been investigated as GBCAs (Table 12).

Fig. 23 Scheme of a GBCA and graphical description of the parameters that influence its relaxivity, including inner-sphere (IS) contributions (hydration number, q; water residency time, τ_m; rotational correlation time, τ_R) as well as second (2S) and outer (OS) sphere effects.

6.2.1 Hydration number. Inner-sphere relaxivity is directly proportional to the hydration number q, and indeed commercial GBCAs feature one or two water molecules in the coordination sphere. To meet this requisite, Imperiali and coworkers developed LBTs with q = 1 (m-sSE3 and q-dSE3-Ubiquitin, Table 12), which showed similar relaxivity relative to commercial GBCAs at high magnetic field (11.75 T).⁶⁰ As already mentioned above (2.2.2 Lanthanide-binding tags), the increase of q from 0 to 1 was accomplished by mutating Glu₉ to Asp₉, and it was accompanied by a nearly 5-fold increase in longitudinal relaxivity (Table 12).

Peacock and co-workers have investigated the relaxivity of Gd³⁺–3SCC scaffolds, showing a significant dependence on the location of the Ln³⁺-binding site along the coiled coil and the second sphere layer, which both affect the hydration number (see 2.4.1 Coiled coils).^{98–100,102,104} Remarkably, scaffolds with high hydration numbers, such as MB1-1 (q = 3), MB1-4 and the recently reported LanM derivative LanND (q = 2), display significantly higher relaxivity compared to clinical GBCAs, especially at low clinical field (1 T, Table 12).

6.2.2 Rotational correlation time. The rotational correlation time, τ_R, reflects the rate of molecular tumbling. Of note, the impact of τ_R on the relaxivity depends on the strength of the magnetic field, as illustrated in Fig. 24 by the nuclear magnetic relaxation dispersion (NMRD) profiles (i.e. relaxivity profiles as a function of the magnetic field strength) for different τ_R values (A) short ∼0.1 ns; B) intermediate; C) long, ∼10 ns).²¹² In general, the relaxivity increases with increasing τ_R at low magnetic fields (<∼1 T in Fig. 24), whereas at higher fields, the relaxivity increases with increasing τ_R up to a threshold beyond which the relaxivity decreases upon further τ_R increase.

Fig. 24 Experimental ¹H NMRD profiles showing the dependence of the relaxivity (per Gd³⁺ ion) as a function of the magnetic field strength for Gd-DTPA derivatives with short (A, ∼0.1 ns), intermediate (B), and long (C, ∼10 ns) rotational correlation times. Reprinted with permission from ref. 212. Copyright 2019 American Chemical Society.

To improve the relaxivity, a τ_R increase can obtained by increasing the molecular weight and the rigidity of GBCAs.²¹⁴

For instance, the Gd³⁺ complex with the rigid cyclodecapeptide scaffold PA (see 2.3.1 Cyclic decapeptide) showed remarkably high relaxivity (20–40 mM⁻¹ s⁻¹, Table 12) at high magnetic fields (4.7–11.75 T) thanks to a medium-range molecular weight (∼1.3 kDa) implying an optimal τ_R ≈ 0.4 ns.⁴¹

Besides, Yang and co-workers have developed protein-based CAs by engineering protein scaffolds with Gd³⁺-binding sites, achieving much higher relaxivity relative to small GBCAs at both low and high fields (Table 12).^217,218 In CA1.CD2 (∼12 kDa), where a de novo designed rigid Gd³⁺-binding site was engineered into a compact CD2 (a cell adhesion protein) domain, the enhanced relaxivity relative to small GBCAs is mainly the result of a higher hydration number (q = 2) and a remarkably longer τ_R (∼9 ns) relative to commercial GBCAs.²¹⁷ It is worth noting that the local rigidity of the Gd³⁺ complex, and not only the molecular weight, is crucial to achieving high relaxivity. As an example, Yang and coworkers also showed that Gd³⁺ bound to a CD2 domain engineered with a flexibly conjugated EF-loop from CaM showed a relaxivity comparable to that of small commercial GBCAs (∼3.4 mM⁻¹ s⁻¹ at 3 T).²¹⁷ Analogously, the mere conjugation of double LBT (dLBT) on a protein scaffold (ubiquitin) did not improve their relaxivity (Table 12).⁶⁰

A long τ_R (7–10 ns) was also reported for Gd³⁺–3SCC complexes;¹⁰² however, it is noteworthy that lengthening the coiled coil from five (MB1-1, ∼12 kDa) to six heptads (MB1-1L, ∼15 kDa) and the consequent τ_R increase had negligible impact on the relaxivity (Table 12), suggesting that the tumbling rate is not a limiting factor for such scaffolds.

6.2.3 Water residency time. The relaxivity has roughly a bell-shaped dependence on the water residency time (τ_m), as the water exchange must be neither too fast nor too slow to maximize the relaxation of bulk water molecules. Similarly to the case of τ_R, the impact of τ_m on the relaxivity also depends on the strength of the magnetic field: at the higher fields, the optimal relaxivity can be attained by a broad range of τ_m, while at low fields the optimal relaxivity can be obtained only for a narrow range of τ_m.²¹² Of note, the impact of τ_m on the relaxivity tends to be negligible for small molecules with fast tumbling, but it can be important for complexes with slow tumbling such as bigger peptides and proteins.²¹² Peacock and coworkers have recently demonstrated that a short water residency time is the predominant factor by which Gd³⁺–3SCC complexes achieve high relaxivity.^102,104 In particular, the authors measured an extremely fast water exchange rate (τ_m = 1.56 ns), comparable to that of free Gd³⁺ in water (τ_m = 1.02 ns) and much faster relative to that of small GBCAs (τ_m ≈ 240–300 ns, Table 12).¹⁰²

6.2.4 Outer-sphere relaxivity. Besides the inner-sphere (IS) relaxivity, second- (2S) and outer-sphere (OS) contributions exist involving hydrogen-bonded water or exchangeable protons in the second coordination sphere, and water molecules that diffuse close to the Gd³⁺ centre, respectively. Remarkably, such OS effects account for about 40% of the overall relaxivity of commercial GBCAs.²¹² Moreover, note that the relaxivity of Gd³⁺-complexes with no inner-sphere water molecules (q = 0) basically stems from 2S/OS contributions.

This is the case, for instance, of some Gd³⁺–3SCC scaffolds (e.g. MB1-2 and MB1-3) that showed a relaxivity comparable to that of commercial GBCAs at high field (7 T), despite having q = 0 (Table 11). Indeed, the authors hypothesized that this could be mainly related to an OS mechanism involving hydrogen bonding and proton exchange networks between the peptide surface and bulk water.⁹⁹

Delangle and coworkers measured the IS and OS relaxivities of the Gd³⁺–PA complex and indirectly determined the 2S contribution, which accounts for about 25% of the total relaxivity.^41,219 By means of MD simulations and the fit of the experimental data with appropriate models, the authors provided a molecular interpretation of the 2S contribution, identifying three 2S water molecules with a high residence time (≈1 ns).

Important 2S and OS contributions were also evidenced for the protein-based GBCAs reported by Yang and coworkers.²¹⁸

6.2.5 Discussion and perspectives. Although an exhaustive comparison among the magnetic properties of different Ln³⁺-binding peptide and protein scaffolds is prevented by the different conditions in which relaxivity data were collected, the highest relaxivity is observed (i) at low fields (0.47–3 T), for large and compact protein scaffolds with long τ_R (∼10 ns), such as some 3SCCs and CA1.CA2, and (ii) at high fields (7–11.75 T), for small rigid scaffolds with intermediate τ_R (∼0.5–2 ns) and important 2S/OS contributions, such as PA. This trend, which is consistent with the dependence of relaxivity on τ_R and the magnetic field strength (Fig. 22), points to the fact that, while large protein scaffolds could be suitable for current clinical (low-field) MRI settings, smaller scaffolds will meet the future need (see above) for GCBAs applicable at higher field. The analysis of the relaxivity data reported indicates that GBCAs based on peptide and protein scaffolds can benefit from higher relaxivity than current clinical GBCAs. This is due to their longer τ_R (related to higher molecular weight and/or rigidity) as well as important second- and outer-sphere contributions mediated by the peptide surface (e.g. hydrogen bonding networks). It is also worth noting that protein-based GBCAs such as ProCA32 and LanND have enhanced molecular relaxivity (∼40–50 mM⁻¹ s⁻¹ at 7 T), as they bind two and four Gd³⁺ ions, respectively.

However, most peptide and protein scaffolds display a lower affinity for Gd³⁺ (log K_app ≤ ∼13, see 3. Thermodynamic stability) than currently used GBCAs (log ^cK ≈ 15–19 at pH 7.4²¹²) and their kinetic inertness has been poorly explored, challenging their in vivo applications. Nevertheless, some protein-based GBCAs, including ProCA32 and LanND-Gd benefit from higher selectivity against physiological metal ions (e.g. Ca²⁺, Cu²⁺, Zn²⁺) than small chelators (see 4.4 Selectivity over d-block metal ions), and showed good biocompatibility, resulting as a promising alternative as next-generation GBCAs. Indeed, ProCA32 showed a half-life of ∼4 hours in the blood plasma of mice, which increased to ∼10 hours upon PEGylation. At 14 days postinjection of ProCA32, a 3- to 10-fold lower amount of Gd³⁺ was detected in different organs compared to the administration of Gd bound to a DTPA derivative.²²¹ Hence, ProCA32 has been conjugated with different targeting moieties, including PSMA (prostate-specific membrane-associated antigen), CXCR4 (chemokine receptor 4) and a collagen-targeting peptide, to selectively target prostate cancer, liver metastasis or lung fibrosis.^222–226 Similarly to clinical GBCAs, LanND concentration in mice blood dropped below detection level within 1 hour, and renal clearance was observed within 3 hours after LanND injection. Moreover, LanND did not show cellular and neural toxicity and did not elicit an evident immune response. Based on these favourable properties, LanND allowed the high-resolution visualization of brain vessels and the monitoring of kidney dysfunction in mice.²²⁰

Beyond the research of safer and more sensitive GBCAs, there is a growing interest in the development of responsive GBCAs, which have been reviewed in ref. 227–231. For instance, GBCAs responsive to the pH, molecular targets (cations, anions, metabolites, nucleic acids), biomarkers, or enzyme activity can be designed based on a change of the relaxivity, that is on the variation of hydration number (q-based) or the tumbling rate (τ_R-based), or both.

In the early 2000s, the HLH scaffold P3W (see 2.2.3 Chimeric helix-loop-helix motifs) was the first, and currently unique, Ln³⁺-binding peptide investigated as a responsive GBCA. Gd³⁺–P3W showed a 4- to 6-fold higher relaxivity than Gd³⁺–DTPA at low clinical magnetic fields (Table 11), as a result of higher hydration number (q = 2) and slower tumbling rate, whose further decrease upon DNA binding yielded a ∼100% relaxivity increase.⁷⁷ Lately, Sénèque and coworkers have also developed Zn²⁺-responsive peptides with appended Gd³⁺–DOTA/DO3A derivatives that show a Zn²⁺-dependent relaxivity increase based on a change of the system rigidity and/or hydration number upon Zn²⁺ binding.^232,233 These examples illustrate how responsive peptide/protein-based GBCAs could be designed.

6.3 Ln recovery and separation using Ln-binding peptides and proteins

6.3.1 Challenges in REE separation. Rare earth elements, which include Ln, Y, and Sc, have become indispensable for medicine (Gd, Lu), and green energies as they are essential elements of permanent magnets and rechargeable batteries (La, Ce, Pr, Nd, Sm, Gd, Dy), lightings (La, Eu, Gd, Tb, Y), and catalysis and refinery (La, Ce) as reviewed in ref. 234 and 235 (Fig. 25). Even Er, Tm and Yb, which are not used on a large scale, find industrial applications. Er is used in silica-glass fibres for communications and in laser for medical applications; Tm is also used in laser for laser surgery, ¹⁷⁰Tm radionuclide is used in portable X-ray sources, and so is ¹⁶⁹Yb.²³⁵

Fig. 25 Overview of the most developed Industrial uses of Ln: rechargeable batteries, permanent magnets, lightings, catalysis and refinery, and medicine. Pm is a radioactive element and does not occur in Nature.

The REE market is dominated by China, which supplies 98% of REE used in the EU.² Due to the expected increasing demand for REE in the coming years and to the lack of high concentrated REE mining sites in Europe, REEs have been identified as critical elements by the EU.^2,235 Thus, there is a strong interest in developing REE recovery technologies from secondary sources, such as the enormous amounts of waste from electrical and electronic equipment generated each year (WEEE 1–30% wt REE).²³⁴ Additionally, sources with low concentrations of REE such as phosphogypsum (0.01–2% wt REE), red mud (500–2500 ppm REE), mine tailings (3–5% wt REE), coal ashes (0.1–1% wt REE), or acid mine drainages (200–900 μg L⁻¹ REE) are available in such amounts that the total content of REE they contained could be worth exploiting. The industrial and economic feasibility of the valorisation of such wastes was discussed in ref. 236.

Beyond economic and geopolitical aspects, it is urgent to develop new and more sustainable REE recovery technologies to limit damages caused by mining and reduce the greenhouse gas emissions and the detrimental impact on the environment of actual processes that rely heavily on pyrometallurgy (high energy consumption, release of toxic gases, volatile metals and dust) and hydrometallurgy (use of strong acids such as HCl, HNO₃, or H₂SO₄; solvents such as kerosene or hexane; large amount of liquid wastes and sludges; risk of heavy metal pollution, including by radionuclides).³ The separation of REE is an additional challenge, which is performed by multiple separation cycles involving liquid/liquid extraction, or precipitation, and relying on the use of organic ligands.^14,237

In this context, biometallurgy is emerging as a sustainable alternative that could reduce the energy consumption, waste production, and volumes of strong acids and solvents used, even though it has not yet been translated at the industrial scale.^14,238–243 This approach relies on the use of micro-organisms or biomaterials for the (selective) recovery of critical metal ions and is particularly suited for low-grade REE feedstocks with low REE concentrations.

Several bio-inspired strategies have been investigated for the recovery of REE from secondary sources which were discussed in recent reviews: (i) non-modified microorganisms: microorganisms were used either for biosorption of REE on their surface, or for their (active) bioaccumulation as reviewed in ref. 239–243. Notably, the Ln-dependent methylotrophic strain M. fumariolicum SolV was recently described for its ability to recover REE from various low-grade sources;²⁴⁴(ii) display of REE-binding motifs: micro-organisms were also modified to display Ln³⁺-binding peptides on their surface in order to improve REE biosorption, which was recently reviewed.^15,242 Examples include LBT displayed on phages,^245,246 or E. coli,^{154,247–251} and LanM displayed on Y. lipolytica and E. coli;^252,253(iii) Ln-binding peptides and proteins: Ln³⁺-binding peptides and proteins were also used directly for recycling (Table 13) as reviewed in ref. 15, 241, 242 and 254 and discussed in more detail below.

Table 13 Strategies based on Ln-binding peptides and proteins for REE recovery. Details are given regarding the scaffolds used, the pH range in which REE³⁺-binding was performed, the composition and pH of the solutions from which REE were recovered, the sorption/desorption strategies employed, and if intra-REE selectivity was achieved. n.d.: not determined; n/a: not applicable

Scaffold	pH range	REE^3+	Sorption (S)	Intra-REE selectivity	Ref.
		Competing metal ions	Desorption (D)
		(source, pH)
LBT	>5	Ln	S: filtration on functionalised fibres	Yes	255
		Al, Ca, Cu, Fe, Ni	D: HNO3, pH 2
		(mine tailings, pH 7)
	∼4	Tb	S: filtration on functionalised fibres	n.d.	256
		Ca or Cu or Fe or Zn	D: HCl, pH 1.7
		—
RTX	>0.5	Nd, Dy	S: filtration on 3 kDa cutoff ultrafilter	n.d.	129
		Fe, Co	D: n/a
		(NdFeB magnet solution, pH 6)
LanM	>2.5	Sc, Y, Ln	S: filtration on 3 kDa cutoff ultrafilter (Mex-LanM)	no	155
		Li, Be, Na, Mg, Ca, Sr, Al, Si, Mn, Fe, Co, Ni, Cu, Zn, U	D: n/a
		(lignite leachate, pH 3.7)
	>2.4	Sc, Y, Ln	S: filtration on functionalized agarose beads (Mex-LanM)	Yes	257
		Li, Na, K, Rb, Mg, Ca, Sr, Ba, Al, Si, Pb, V, Mn, Fe, Co, Ni, Cu, Zn, Se, Th, U	D: pH steps (HCl, pH 2.3, then pH 1.5)
		(coal fly ash leachate, pH 5)
	—	Sc, Y, Ln	S: filtration on functionalized agarose beads (Mex-LanM)	Yes	167
		—	D: two-cycle process: 1. Malonate, then pH ramp (2.3-2.1-1.9), 2. Citrate, then pH 1.5
		(coal combustion, mine tailings, pH 3)
	>2	Y, Ln	S: liquid–liquid phase separation (Mex-LanM)	Yes	258 and 259
		Mg, Na, Ca	D: phosphate-citrate buffer, pH 2.2
		(coal fly ash leachate, pH 4.5)
	>3	Sc, Y, Ln	S: functionalised magnetic nanoparticles (Mex-LanM)	n.d.	260
		Li, Na, K, Rb, Mg, Ca, Ba, Al, Si, Pb, V, Mn, Fe, Co, Ni, Cu, Zn, As, Se, Cs	D: HCl, pH 1.7
		(coal fly ash leachate, pH 5)
	—	Dy, Nd	S: filtration on functionalized agarose beads (Hans-LanM, Hans-LanM-R100K)	Yes	62
		—	D: 1. Malonate, 2. HCl, pH 1.5
		(pH 5)
LanD	n.d.	La, Ce, Pr, Nd	S: filtration on 10 kDa cutoff ultrafilter (LanD-E75Q/E78A)	Yes	125
		—	D: n/a
		(pH 6)

---

Table 14 Values from Fig. 1: trends of key parameters along the Ln series

Ln^3+	E^3+/2+ (V)^26	E^4+/3+ (V)^26	r (Å)^27	pKa^28	log Ksp^29
La^3+	−3.1		1.216	8.89	−22.3
Ce^3+	−3.2	1.7	1.196	8.31	−23.5
Pr^3+	−2.7	3.4	1.179	8.3	−23.4
Nd^3+	−2.6	4.6	1.163	8.13	−24.1
Pm^3+	−2.6	4.9	1.144	n/a	n/a
Sm^3+	−1.55	5.2	1.132	7.84	−24.8
Eu^3+	−0.34	6.4	1.12	7.66	−25.5
Gd^3+	−3.9	7.9	1.107	7.87	−24.8
Tb^3+	−3.7	3.3	1.095	7.6	−25.7
Dy^3+	−2.5	5	1.083	7.53	−25.7
Ho^3+	−2.9	6.2	1.072	7.43	−26.4
Er^3+	−3.1	6.1	1.062	7.46	−26.2
Tm^3+	−2.3	6.1	1.052	7.34	−26.4
Yb^3+	−1.05	7.1	1.042	7.31	−26.7
Lu^3+		8.5	1.032	7.33	−26.6

---

Table 15 Values from Fig. 17: selectivity of different ligands across the Ln³⁺ series

Ln^3+	EDTA^a	DTPA^a	Macropa^b	Macrotripa^c	LBT12^d	LF4^e	RTX^f	LanD^g	LanM^h	LanM^i	LanM^j	LanM^k
a log βML from ref. 173. b log βML from ref. 160. c log βML from ref. 161. d log Kapp in HEPES 10 mM pH 7, NaCl 100 mM, ref. 64. e log Kapp in HEPES 10 mM pH 7.5, NaCl 10 mM, ref. 103. f log Kapp in bis-Tris 20 mM pH 6, NaCl 25 mM, ref. 129. g log Kapp in MOPS 30 mM pH 7, KCl 100 mM, ref. —. h log Kapp for Mex-LanM in MOPS 30 mM pH 7.2, KCl 100 mM, measured by CD spectroscopy, ref. 81. i log Kapp for Mex-LanM in acetate buffer 20 mM pH 5, KCl 100 mM, measured by CD spectroscopy, ref. 155. j log Kapp for Hans-LanM acetate buffer 20 mM pH 5, KCl 100 mM, measured by CD spectroscopy, ref. 280. k log Kapp for Mex-LanM KCl/KCH3COO^− buffer 100 mM pH 5.0, measured by UV-vis spectroscopy, ref. 155.
La^3+	15.5	19.5	15.0	12.6	5.5	4.2	4.1	5.7	11.3		10.2
Ce^3+	16.1	20.5	15.1	12.8	6.0			6.0
Pr^3+	16.3	21.1	14.7	12.7			4.5	6.3		10.2		11.0
Nd^3+	16.6	21.6	14.4	12.3	6.6		4.5	6.6	11.3		10.0	10.7
Sm^3+	17.0	22.3	13.8	11.4				6.5	11.2			10.6
Eu^3+	17.2	22.4	13.0	10.9	7.2	4.6	4.6	6.5
Gd^3+	17.3	22.5	13.0	10.2	7.1	4.6		6.2	11.0	10.0
Tb^3+	17.8	22.7	11.8	10.2	7.2	4.9	4.6		10.7
Dy^3+	18.2	22.8	11.7	10.5	7.2		4.6			9.7	8.6	10.8
Ho^3+	18.4	22.8	10.6	10.7		4.9			10.6	9.6		11.1
Er^3+	18.9	22.7	10.1	11.1	7.1	5.0						11.9
Tm^3+	19.6	22.7	9.6	11.4								12.4
Yb^3+	19.7	22.6	8.9	11.8	7.0	4.4	4.6
Lu^3+	20.0	22.4	8.3	11.9	6.9	4.2	4.6

---

Table 16 Values from Fig. 19: selectivity of small chelators and some peptide/protein scaffolds for Ln³⁺ over Ca²⁺

Ln^3+	EDTA^a	DTPA^a	P3W^b	EF loop^c	CaM^d	Hans-LanM^e	Mex-LanM^f	RTX^g	PqqT^h	PA^i	2SCC^j
a log βML from ref. 173. b log Kapp in TRIS 5 mM pH 7.8, from ref. 76. c log Kapp in HEPES 10 mM pH 7, NaCl 100 mM, from ref. 169. d log Kapp in PIPES 25 mM pH 6.8, KCI 100 mM, from ref. 82. e log Kapp for Hans-LanM acetate buffer 20 mM pH 5, KCl 100 mM, measured by CD spectroscopy, ref. 280. f log Kapp for Mex-LanM in MOPS 30 mM pH 7.2, KCl 100 mM, measured by CD spectroscopy, ref. 81. g log Kapp in bis-Tris 20 mM pH 6, NaCl 25 mM, ref. 129. h log Kapp in HEPES 30 mM pH 7, NaCl 100 mM, from ref. 20. i log Kapp in HEPES 10 mM pH 7, from ref. 41. j log Kapp in Imidazole 30 mM pH 6.9, KCl 50 mM, from ref. 95.
La^3+	15.5	19.5	5.20			10.2	11.3	4.1	6.2		6.2
Ce^3+	16.1	20.5
Pr^3+	16.3	21.1						4.5
Nd^3+	16.6	21.6				10.0	11.3	4.5
Sm^3+	17.0	22.3					11.2
Eu^3+	17.2	22.4	5.2	6.8	8.00			4.6		6.8
Gd^3+	17.3	22.5	5.2				11.0
Tb^3+	17.8	22.7		6.8	8.8		10.7	4.6		6.3
Dy^3+	18.2	22.8				8.6		4.6
Ho^3+	18.4	22.8					10.6
Er^3+	18.9	22.7
Tm^3+	19.6	22.7
Yb^3+	19.7	22.6						4.6			6.4
Lu^3+	20.0	22.4						4.6
Ca^2+	11.0	10.7	4.5	4.6	6	4.2	3.2	3.3	3.8	3.7	1.7

---

6.3.2 REE recovery strategies. The development of recovery strategies based on Ln-binding peptides and proteins has been done using EF-hand scaffolds, as well as other types of scaffolds.

Joshi and coworkers used LBT to develop filters for REE recovery from dilute acid wash.²⁵⁵ They genetically engineered a Curli fibre produced in E. coli biofilm and known to form amyloid fibres extracellularly, to fuse it with two or four repeats of an LBT sequence. Following a similar strategy, Kaplan and co-workers developed filters based on silk nanofibrils (SNF) functionalized with LBT peptides (Fig. 26), either by chemical coupling or by genetically fusing LBT to a silk-elastin-like protein (SELPS).²⁵⁶

Fig. 26 Overview of the two approaches used to fabricate biobased filter membranes. Reprinted with permission from ref. 256. Copyright 2024 American Chemical Society.

Following the discovery of LanM, their potential for the recovery of REE (including Y³⁺ and Sc³⁺) was investigated mostly with Mex-LanM, and more recently with Hans-LanM (Table 13). The strategies employed relied on size-exclusion filtration, on LanM-functionalised agarose beads,²⁵⁷ on functionalized magnetic nanoparticles (MNP-LanM, Fig. 27),²⁶⁰ on polymeric membranes functionalized with LanM-derived peptides,²⁶¹ and on liquid–liquid phase separation made by fusing LanM to a thermo-responsive genetically encoded polypeptide RELP.²⁵⁸

Fig. 27 Use of LanM-functionalized magnetic nanoparticles for REE recovery.^258,259 Reprinted with permission from ref. 260. Copyright 2023 American Chemical Society.

Cotruvo and coworkers investigated the use of LanD for large Ln-separation.¹²⁵ They hypothesized that by combining mutations of LanD to lower the affinity of its monomer for Ln³⁺ (E75Q) and favour the dimer formation by removing repulsive electrostatic interactions (E78A), they could obtain a dimer that preferentially binds Pr³⁺ and Nd³⁺ over La³⁺ and Ce³⁺.

Still with the aim of developing a method for REE recovery, Banta and coworkers proposed to use the RTX protein, which can bind up to 8 lanthanides at pH 5.5.¹²⁹

6.3.3 pH sensitivity. As stressed above (see 3.4 Effect of the pH), the impact of pH on Ln³⁺-binding is generally an overlooked feature. Nonetheless, because applications for recovery require working with acidic REE sources, the pH sensitivity of some of the peptides/proteins used or their derived materials has been investigated.

LBT-based fibres displayed a similar pH-sensitivity than observed in solution.²⁵⁵ The fibres could bind Tb³⁺ down to pH 5, and their Tb³⁺-sorption capacity was decreased to the level of non-functionalized fibres at pH 3. Park and Cotruvo showed that Mex-LanM functionalised agarose beads retained their ability to bind REE at pH 3,²⁵⁷ which was consistent with Mex-LanM capacity to bind Ln³⁺ down to pH 2.5 in solution.¹⁵⁵ RTX is the protein that seems to be the more tolerant to pH as it retains some Ln³⁺-binding capacity at pH < 1.5. However, the proteins precipitate at low pH, which complicated the analysis.

Although the different scaffolds described to date present various pH-sensitivity, most of them were assayed for their ability to bind Ln on slightly acidic solutions (pH 4–7, Table 12) corresponding to REE sources obtained from mine tailings, or leachates. Desorption of REE was only investigated for LBT- and LanM-based materials and could be achieved with acidic solutions (HCl, HNO₃) at pH < 2. This is a consequence of the protonation of Ln³⁺-binding amino acids (e.g. Asp, Glu), which results in a lower affinity for Ln³⁺.

6.3.4 Selectivity.
REE vs. non-REE. Working with low-grade, secondary sources of REE that have a low content in REE (WEEE, phosphogypsum, bauxite, mine tailings, coal ashes, mine drainages) requires a good selectivity for REE vs. non-REE in order to minimize the purification steps. Usual competing metal ions consist notably of alkaline-earth (Ca, Mg), and d-block (Mn, Fe, Co, Ni, Cu, Zn) elements (Table 12). The ability of the different scaffolds or derived materials to retain selectively REE vs. non-REE has been assessed in most studies.

For the LBT-based fibres described by Joshi and coworkers,²⁵⁵ REE sorption fell in the presence of 10 to 100 times more concentrated non-REE metal ions (Al, Cu, Fe, Ni), whereas Ca only had a small impact. Similarly, Kaplan and coworkers reported that the Tb³⁺-binding capacity of silk nanofibrils functionalised with LBT was decreased in the presence of 10 to 100 times more concentrated Cu or Fe, but not by Ca and Zn.²⁵⁶ LBT has a selectivity against Ca²⁺ of Δlog K^Ca ≈ 2–3 (see 4. Selectivity), which appears sufficient to compensate for the difference in concentrations. However, from the observations made in these two studies, it is anticipated that LBT selectivity against d-block elements is quite low.

LanMs display a very high selectivity against Ca (Δlog K^Ca ≈ 8 for Mex-LanM,^81,155 Δlog K^Ca_a ≈ 4–6 for Hans-LanM⁶²) and d-block elements (Δlog K^Cu ≈ 6 and Δlog K^Zn ≈ 8 for Mex-LanM¹⁵⁵) (see 4. Selectivity). This can explain the impressive ability of LanM to selectively recover REE among other metal ions in complex solutions (Table 12) with high recovery yields and purity when using size-exclusion separation,¹⁵⁵ functionalised agarose beads,²⁵⁷ or liquid–liquid phase separation.^258,259 Moreover, Gao, Wei and coworkers demonstrated that even with concentrations of non-REE 10 to 1000 times higher than of REE, LanM-functionalized magnetic nanoparticles could selectively bind REE over competing metal ions and achieve a REE purity of 31% wt (compared to the 0.03% wt of the initial solution) corresponding to a 967-fold improvement.²⁶⁰

RTX is interesting due to its ability to bind several Ln per protein over a broad pH range.¹²⁹ However, one limitation to its use is the precipitation of the protein at low pH. To overcome precipitation issues, Banta and coworkers coupled the RTX protein to a maltose-binding protein (MBP). Working at pH 6, they showed that RTX-MBP could selectively recover up to ∼85% of Nd³⁺ and Dy³⁺ in a solution containing also Fe²⁺ and Co²⁺, which suggests a good selectivity of this modified protein for Ln over d-block metal ions.

Intra-REE. The design of Ln-binding peptides and proteins with good intra-REE selectivity would be highly appealing with the aim to reduce the number of steps required to obtain pure and isolated REE, but it is an outstanding challenge given the similarities of REE (see 1. Introduction) and the moderate intra-REE selectivity displayed so far by Ln-binding peptides and proteins (see 4. Selectivity). The LBT-fibers reported by Joshi and coworkers showed a moderate preference for the binding of Eu³⁺, Tb³⁺, and Dy³⁺ relative to other Ln³⁺.²⁵⁵ Cotruvo and coworkers noted an intra-REE selectivity between heavy (HREE: Tb³⁺–Lu³⁺, Y³⁺) and light (LREE: La³⁺–Gd³⁺) when using LanM-functionalised on agarose beads similar to the one observed in solution.²⁵⁷

To overcome the moderate intra-REE selectivity of Ln-binding peptides and proteins, some studies have proposed innovative strategies. The separation process based on LanM-functionalised on agarose beads was optimised by using mild competitors for Ln desorption (malonate, citrate).¹⁶⁷ A two-cycle process was proposed in order to recover five distinct fractions: Sc³⁺, Y³⁺, HREE (Gd³⁺–Lu³⁺), middle REE (MREE: Pr³⁺–Eu³⁺), and LREE (La³⁺, Ce³⁺). A solution reproducing metal ions concentrations found in coal combustion products or mine tailings was adsorbed onto the column. The first desorption cycle used malonate in order to desorb up to 99% of Sc with >99% purity. Next, pH desorption was performed by decreasing the pH successively to 2.3, 2.1 and 1.9, which enabled a first separation of REE into HREE and Y³⁺, LREE, and MREE. These fractions were further purified using citrate desorption, which led to the recovery of HREE with 88–100% purity, >95% recovery of Y³⁺ with 92% purity, recovery of MREE with ∼80% purity, and recovery of LREE with 98% purity.

As an alternative strategy, Cotruvo and coworkers investigated how intra-REE selectivity could be improved by controlling the dimerization state of Hans-LanM to separate Dy³⁺ from Nd³⁺ working with a synthetic leachate solution of NdFeB magnets.⁶² With Hans-LanM functionalised agarose beads, Dy³⁺ could be desorbed from the column with malonate and obtained with 83% purity without significant desorption of Nd³⁺, which was then desorbed at pH 1.5 and obtained at 99.8% purity. In a similar approach, the LanD-E75Q/E78A mutant optimised to form dimer with medium REE, was able to retain Pr³⁺ and Nd³⁺ better than the larger Ce³⁺ and La³⁺.

6.3.5 Concluding note. The development of biomaterials based on Ln-binding peptides and proteins is well suited to exploit secondary REE sources, which are more dilute and of lower grade. As such, it represents a complementary approach to the development of chemical strategies that aim to improve the extraction and separation processes used on primary REE sources.^14,237

REE recovery using Ln-binding peptides and proteins requires to meet several criteria. The first is the selectivity of the peptides and proteins for REE vs. non-REE, and ideally also for intra-REE separation. The discovery of LanM, which has a high selectivity vs. non-REE has represented a breakthrough compared to LBT-based strategies that are more sensitive to high concentrations of d-block metal ions. Nonetheless, Ln-binding peptides and proteins still present a moderate intra-REE selectivity (see 4. Selectivity) that will need to be improved. The second criterion is to improve the pH range in which the scaffold is stable and retains Ln-binding capacity. Last, different strategies were proposed for the sorption and recovery of REE. Scaling up to meet industrial needs will require a facile and low-cost production of the scaffold and its derived material. The strategy employed for functionalizing magnetic nanoparticles with LanM using Spy chemistry is interesting for this purpose, as it enabled the selective coupling of the protein directly from cell lysates, thus avoiding protein purification steps.²⁶²

Importantly, despite the considerable efforts made to develop innovative, more sustainable technologies to exploit secondary REE sources, the main challenge to their industrial implementation remains their economic viability. The economic feasibility of the biosorption process for the recovery of REE from different low-grade sources was evaluated and strongly depends on REE market prices. Only an increase of REE prices back to their level of 2011 would make actual biosorption processes profitable.²⁶³ The fluctuating demands in REE and its impact on the development at the industrial scale of more environmentally friendly technologies have been discussed in ref. 235.

6.4 Catalysis

The field of Ln–proteins and Ln–peptides catalysis is emerging, and a few examples have been published. Ln³⁺ Lewis acidity has been exploited for hydrolysis reactions and in alcohol dehydrogenase (ADH)-like reactions (Section 6.2.1). Ln³⁺ photoredox properties have been exploited for the homolytic cleavage and oxidation of diols (Section 6.2.2).

6.4.1 Lewis acidity. Before the discovery of Ln–enzymes, Franklin and coworkers used Ln³⁺ Lewis acidic properties for the design of Ln–peptides as chimeric nucleases.^17,18 Helix-turn-helix peptides (HTH) combined an Ln³⁺-binding loop and helices from engrailed homeodomain, the latter offering the possibility to recognize a specific DNA sequence. The resulting Ln–peptides (Ln = Eu³⁺, Ce⁴⁺) displayed DNA- and phosphate esters cleavage activities (Fig. 28).^264,265

Fig. 28 Phosphate ester and DNA cleavage activities of Ln-pep (Ln = Eu³⁺, Ce⁴⁺).

The Ce⁴⁺–peptides were more active than their Eu³⁺ counterparts, consistent with the increased Lewis acidity and nucleophilicity of the Ln³⁺-bound water molecules.¹⁶ The effect of the pH on Ln–peptides reactivity was evaluated in the range 6.9–9.0.²⁶⁶ The rate increased at pH 8.2, which is close to the pK_a of Ln³⁺-bound water molecule (Fig. 1), indicating that it is involved in DNA hydrolytic cleavage. In addition, the cleavage catalysed by Ln–peptides was site-selective, suggesting that DNA sequences could be recognized and cleaved preferentially.¹⁶

Ln³⁺ Lewis acidic properties were also used for the catalysis of the hydrolysis of 2-naphthyl hexanoate (Fig. 29). Jackson and coworkers used the PTE-R18 protein, which has a Zn²⁺-binding site, and showed that the demetallated protein could bind Ln³⁺.²¹ The activity of the resulting Ln–protein was pH-dependent, which was linked to the deprotonation of a Ln-bound water molecule.

Fig. 29 Hydrolysis of 2-naphthyl hexanoate by Ln-PTE-R18.

The first and sole known Ln–enzymes, Ln-alcohol dehydrogenases (Ln-ADH), are an example of how the fine-tuning of Ln³⁺ Lewis acid properties along the series may impact reactivity. In the enzyme, Ln³⁺ is essential for catalysis because it tunes the redox potential of the redox cofactor, the pyrroloquinoline quinone (PQQ). Interestingly, the activity of the XoxF-methanol dehydrogenase (MDH) is strongly dependent on the Ln³⁺ in the active site, the lighter and less Lewis acidic Ln³⁺ giving the higher activity while no activity is detected for Ln³⁺ heavier than Er³⁺ (Fig. 30).¹⁷⁶ A similar trend has been also recently observed for an Ln-ADH from Pseudomonas putida KT2440 (PedH).²⁶⁷ Several hypotheses have been proposed and are still under investigation to explain this dramatic impact on Ln-ADH reactivity. They consider not only PQQ redox properties, but also changes in the coordination number, or exchange rate constants of Ln³⁺-bound substrate/product.

Fig. 30 Ln-dependent activity of XoxF-MDH. The enzyme was reconstituted from the apo-enzyme by incubation with PQQ and metal ion for 72 h at 4 °C. Reproduced from ref. 176 (Creative Commons Attribution Non-Commercial License 2023).

Working towards bringing new insights into Ln-MDH mechanism, Olshansky and coworkers built a structural mimic of the enzyme.²⁰ Starting from the PQQ-binding protein PqqT, they engineered the binding site with a single mutation (K₁₄₂D) in order to introduce a Ln³⁺-binding site. They tested the ability of this structural mimic of Ln-MDH to oxidize benzyl alcohol, using a chemical trap (O-((perfluorophenyl)methyl)hydroxylamine (PFBHA)) to capture the benzaldehyde formed and identify it by LC-MS (Fig. 31). Only the La³⁺–PQQ⊂K₁₄₂D–PqqT mutant was able to perform this reaction, and not the WT protein or K₁₄₂A mutant both lacking an Ln³⁺-binding site, nor the mutants lacking La³⁺, or PQQ.

Fig. 31 Benzyl alcohol oxidation by La³⁺-bound PQQ⊂K₁₄₂D–PqqT.

The ability of ADH to oxidize alcohols has been exploited further by Bange, Klebensberger and coworkers for foreseen application in bio-based plastic production.¹¹⁶ Starting from the Pr-PedH from P. putida KT2440, they constructed libraries of mutants of the enzyme expressed in E. coli and screened each colony for in vivo enzymatic activity towards the oxidation of HMFA (Fig. 32). This compound is an important intermediate for the obtention of furan-2,5-dicarboxylic acid (FDCA), a synthon that could serve as a plant-based alternative in the production of polymers. Out of 30 000 clones, 3 mutants were identified and their activity was confirmed by testing the purified enzymes. They shared mutations of the same residues, F412 and W561, positioned in the substrate channel and on a lid loop, respectively. These mutations enable bulkier substrates, such as HMFA, to access the active site. The mutant with the best catalytic activity for HMFA oxidation, PedH^F412V/W561A, could also successfully oxidize ethanol, HMF and FFF (Fig. 32), which opens up the route to the whole-cell biocatalytic production of FDCA.

Fig. 32 Oxidation of 5-(hydroxymethyl)furoic acid (HMFA) and derivatives by mutants of PedH. FFA: 5-formylfuroic acid; HMF: 5-(hydroxymethyl)furfural; FFF: 5-formylfurfural; PMS: phenazine methosulfate; DCPIP: 2,6-dichlorophenol indophenol.

6.4.2 Photoredox properties. So far, only a single example reported on the use of Ln photoredox properties in artificial enzymes for catalysis. Zeymer and co-workers designed an Ln-binding protein based on a TIM-barrel scaffold with the aim of exploiting the photocatalytic properties of Ce.¹⁹

The scaffold TFD-EE N6W was the starting point for the following catalysis study (Fig. 33).¹⁹ Changes were made to the scaffold in order to introduce asymmetry in the binding site (single chain variant), to limit photodamage (removal of Trp residues), and to prevent non-specific metal binding to the protein surface (removal of clusters of negatively charged residues by mutating Asp to Asn and Glu to Gln), leading to the optimised scaffold PLZ1.4.

Fig. 33 Photoredox catalysis of aromatic diols homolytic cleavage by Ce-artificial enzymes.

Upon irradiation, the artificial enzyme Ce-PLZ1.4 was able to catalyse the homolytic cleavage of homobenzoin into benzaldehyde (Fig. 33). Compared to molecular catalysts, the interest of the protein scaffold is to be able to control the stereoselectivity of the reaction. The authors showed that Ce-PLZ1.4 catalysed better the cleavage of (R,R)- and (S,S)-hydrobenzoin compared to the meso-isomer, and so displayed some diastereoselectivity. Investigation of the substrate scope of the enzyme also evidenced chemoselectivity, with aromatic diols being better cleaved than aliphatic ones. Last, the PLZ1.1 protein was expressed at the surface of E. coli cells, which were subsequently treated with CeCl₃ and light and were able to cleave a lignin surrogate (2-phenoxy-1-phenylpropane-1,3-diol) opening the way to photocontrolled whole-cell catalysis.

6.4.3 Concluding note. Lanthanides have versatile properties ranging from their Lewis acidity which offers a fine-tuning of the reactivity, to their redox properties that cover a broad range of potentials, and to their photochemical properties, which open the way to photocatalytic reactions. The similarities of Ln allow, with a single ligand, to modulate the reactivity of the catalyst simply by choosing the adequate Ln³⁺ ion, which is not possible with any other family of metal ions.

To date, the scope of Ln–peptides and Ln–proteins catalysts has been limited mainly to reactions exploiting Ln³⁺ Lewis acidity properties, with the exception of an original work demonstrating that photoredox catalysis is also achievable.¹⁹ In addition, the peptide scaffold of the catalysts has been used to select the substrate, whether by using DNA recognition sequence, or by engineering the catalytic pocket,¹¹⁶ or by favouring one stereoisomer over others.¹⁹

In the medium term, recent advances obtained with molecular catalysts should be transposable to Ln artificial enzymes. Ln-complexes have been used as Lewis acids catalysts,^268,269 in particular for C–C bond formation in aqueous media.²⁶⁸ Ln-complexes are also powerful photocatalysts, that can be used for photo-reduction or photo-oxidation.^270–274 Notably, the reduction of CO₂, N₂ and NO_x was achieved in mild aqueous conditions.^273,274 Similar reactions could be implemented in artificial Ln–enzymes, which operate in aqueous media and whose peptide scaffold allows modulation of the hydrophilicity of the active site. Combining the chemical properties of Ln with the modularity of the peptide scaffold will help control the reactivity of the metal centre, the selectivity of the reaction and fine-tuning of second sphere interactions, and is expected to broaden the reactivities accessible with artificial enzyme.²⁷⁵

7. Conclusions & perspectives

In the last decade, significant progress has been made in the study of natural Ln-dependent proteins as well as in the design and engineering of Ln-binding peptides and proteins for a variety of relevant applications. Notwithstanding, numerous challenges remain to be addressed for the advancement of this flourishing field, starting with a sound and quantitative understanding of Ln coordination by pH buffers to refine the accuracy of available data and allow reliable comparisons.

The current limitations in designing Ln-binding systems with high affinity, selectivity, and inertness underscore the complexity of controlling key factors such as the coordination sphere (i.e. coordination and hydration number), scaffold stability, and the influence of long-range effects.

The control over the hydration number q is well understood and achievable for EF-hand motifs, and it can be tuned for other scaffolds by adjusting the denticity and positioning of the binding site within a hydrophobic environment. However, certain long-range outer-sphere factors, such as the local conformation of the main chain and the orientation of amino acids can influence q in ways that are not yet fully controllable.

The majority of peptide and protein scaffolds exhibit an Ln-binding affinity in the micro- to nanomolar range (log K_app ∼ 6–9), with only a limited number extending to the picomolar range (log K_app ∼ 12), which is significantly lower than the affinity achievable with small chelators. Similarly to the tuning of the hydration number q, long-range outer-sphere factors, which are not yet understood, strongly influence the Ln affinity, as in the case of LanM compared to its isolated EF-hand motifs. It is worth noting that this gap between the Ln-affinity of peptide/proteins and small organic chelators is not observed for d-block metal ions, for which proteins and chelators of similar affinity, up to log K_app ∼ 21, are known. The underlying reason for this remains to be elucidated. One possible explanation is that, considering that we are at the dawn of lanthanome research, the tightest Ln-binding proteins have yet to be discovered. Besides, due to the ionic nature of Ln bonding and their preference for higher CN than d-block metal ions, peptides and proteins might need more demanding constraints to bind Ln³⁺ as efficiently as d-block metal ions.

Achieving higher affinities requires careful optimisation of both the Ln-binding site and scaffold stability by fine-tuning key parameters such as the type and denticity of the coordinating amino acids, as well as local and long-range structural constraints in the scaffold. To this end, computational design appears as a powerful tool to obtain a well-structured and stable protein scaffold, as is the case for the Ln-binding TIM barrel. In addition, the pH-dependency of Ln-binding by peptides and proteins has only been seldom looked at but is an important parameter for applications in REE recovery.

The selectivity relative to Ca²⁺ is generally favourable due to the higher coordination number and higher charge of Ln³⁺. Nevertheless, the remarkable selectivity exhibited by LanM remains to be fully elucidated. Again, this is likely attributable to the influence of outer sphere effects.

With respect to d-block elements, the selectivity of peptides and proteins for Ln³⁺ often surpasses that of small chelators, a phenomenon attributable to distinct preferences for ligand types, coordination number and geometry.

In comparison to Ln³⁺, peptide and protein scaffolds demonstrate a slightly higher affinity for An³⁺, presumably due to enhanced covalency. However, further investigation is warranted to fully elucidate the underlying mechanisms and to boost the selectivity between Ln³⁺ and An³⁺. Since An³⁺ can be (radioactive) contaminants of common Ln³⁺ sources, this will be crucial in the context of Ln recovery and radioactive decontamination.

The intra-Ln selectivity remains an outstanding challenge due to their similar ionic radii and Lewis acidity. The intra-Ln selectivity exhibited by peptide and protein scaffolds is not yet comparable to that observed for small chelators. This discrepancy may be attributed to the greater flexibility of peptides, which consequently enables a higher degree of adaptability to Ln³⁺ ions differences in ionic radius and coordination number. Further studies are required to ascertain whether size discrimination can be enhanced based on both electronic and steric considerations, for instance by modulating the rigidity of the coordination site and its position (buried or surface-exposed) within the scaffold.

The kinetic inertness of Ln coordinated to peptides and proteins is a largely neglected aspect that merits greater attention and research focus in the upcoming years, especially with respect to the development of efficient imaging agents and tools for Ln separation. In order to understand how to control the kinetic inertness of Ln–peptide and protein complexes, it is also essential to accumulate a substantial body of experimental data, which is currently lacking.

Overall, it appears that the elusive contributions beyond the first coordination sphere are of paramount importance in tuning properties like affinity, selectivity, and inertness, and must be studied further to achieve optimized designs. It is expected that emerging technologies based on machine learning and artificial intelligence will soon be beneficial in unravelling the influence of such long-range interactions, provided that solid experimental knowledge is established.^276–278 In this context, significant contributions are also expected from approaches beyond rational design, such as directed evolution.²⁷⁹

By deepening our understanding of these long-range effects and their impact on the coordination environment, we will be able to address the key challenges in the field: (i) developing high-affinity, intra-Ln selective peptides and proteins for enhanced Ln-recovery; (ii) achieving kinetically inert Ln-binding scaffolds for medical imaging and therapy; and (iii) optimizing the design of Ln-based (photo)catalysts. These advancements hold the potential to revolutionize a wide range of applications, from environmental sustainability to cutting-edge medical technologies.

Conflicts of interest

There are no conflicts to declare.

Data availability

No primary research results, software or code have been included and no new data were generated or analysed as part of this review.

Appendices

Acknowledgements

The authors acknowledge financial support from the French National Research Agency (ANR) through the ANR-23-CE07-0002 (LnZYME) and ANR-22-CE29-0026 (DREAMY) projects.

References

1. S. V. Eliseeva and J.-C. G. Bünzli, New J. Chem., 2011, 35, 1165–1176.
2. Directorate-General for Internal Market, Industry, Entrepreneurship and SMEs (European Commission), M. Grohol and C. Veeh, Study on the critical raw materials for the EU 2023: final report, Publications Office of the European Union, 2023.
3. V. Balaram, Geosci. Front., 2019, 10, 1285–1303.
4. N. A. Fitriyanto, M. Fushimi, M. Matsunaga, A. Pertiwiningrum, T. Iwama and K. Kawai, J. Biosci. Bioeng., 2011, 111, 613–617.
5. Y. Hibi, K. Asai, H. Arafuka, M. Hamajima, T. Iwama and K. Kawai, J. Biosci. Bioeng., 2011, 111, 547–549.
6. T. Nakagawa, R. Mitsui, A. Tani, K. Sasa, S. Tashiro, T. Iwama, T. Hayakawa and K. Kawai, PLoS One, 2012, 7, e50480.
7. A. Pol, T. R. M. Barends, A. Dietl, A. F. Khadem, J. Eygensteyn, M. S. M. Jetten and H. J. M. O. den Camp, Environ. Microbiol., 2014, 16, 255–264.
8. L. Ancel, A. Niedźwiecka, C. Lebrun, C. Gateau and P. Delangle, C. R. Chim., 2013, 16, 515–523.
9. E. Skovran, C. Raghuraman and N. C. Martinez-Gomez, Curr. Issues Mol. Biol., 2019, 101–116.
10. L. J. Daumann, Angew. Chem., Int. Ed., 2019, 58, 12795–12802.
11. E. R. Featherston and J. A. Cotruvo, Biochim. Biophys. Acta, Mol. Cell Res., 2021, 1868, 118864.
12. L. J. Daumann, A. Pol, H. J. M. Op Den Camp and N. C. Martinez-Gomez, Advances in Microbial Physiology, Elsevier, 2022, vol. 81, pp. 1–24.
13. A. M. Webster and A. F. A. Peacock, Chem. Commun., 2021, 57, 6851–6862.
14. J. A. Mattocks and J. A. Cotruvo, Chem. Soc. Rev., 2020, 49, 8315–8334.
15. Q. Ye, D. Wang and N. Wei, Trends Biotechnol., 2024, 42, 575–590.
16. R. T. Kovacic, J. T. Welch and S. J. Franklin, J. Am. Chem. Soc., 2003, 125, 6656–6662.
17. S. Lim and S. J. Franklin, Cell. Mol. Life Sci., 2004, 61, 2184–2188.
18. S. J. Franklin and J. T. Welch, Comments Inorg. Chem., 2005, 26, 127–164.
19. A. S. Klein, F. Leiss-Maier, R. Mühlhofer, B. Boesen, G. Mustafa, H. Kugler and C. Zeymer, J. Am. Chem. Soc., 2024, 146(38), 25976–25985.
20. P. J. Thompson, D. G. Boggs, C. A. Wilson, A. T. Bruchs, U. Velidandla, J. Bridwell-Rabb and L. Olshansky, Proc. Natl. Acad. Sci. U. S. A., 2024, 121, e2405836121.
21. C. W. Breeze, Y. Nakano, E. C. Campbell, R. L. Frkic, D. W. Lupton and C. J. Jackson, Acta Crystallogr., Sect. D:Struct. Biol., 2024, 80, 289–298.
22. L. Götzke, G. Schaper, J. März, P. Kaden, N. Huittinen, T. Stumpf, K. K. K. Kammerlander, E. Brunner, P. Hahn, A. Mehnert, B. Kersting, T. Henle, L. F. Lindoy, G. Zanoni and J. J. Weigand, Coord. Chem. Rev., 2019, 386, 267–309.
23. R. Sánchez-Fernández, I. Obregon-Gomez, A. Sarmiento, M. E. Vázquez and E. Pazos, Chem. Commun., 2024, 60, 12650–12661.
24. M. A. Hay and C. Boskovic, Chem. – Eur. J., 2021, 27, 3608–3637.
25. R. G. Pearson, J. Am. Chem. Soc., 1963, 85, 3533–3539.
26. L. R. Morss, Chem. Rev., 1976, 76, 827–841.
27. P. D’Angelo, A. Zitolo, V. Migliorati, G. Chillemi, M. Duvail, P. Vitorge, S. Abadie and R. Spezia, Inorg. Chem., 2011, 50, 4572–4579.
28. R. D. Shannon, Acta Crystallogr., Sect. A, 1976, 32, 751–767.
29. C. Ekberg and P. L. Brown, Hydrolysis of Metal Ions, John Wiley & Sons, Ltd, 2016.
30. H. C. Aspinall, f-Block Chemistry, Oxford University Press, 2020.
31. L. Ruckthong, M. L. Zastrow, J. A. Stuckey and V. L. Pecoraro, J. Am. Chem. Soc., 2016, 138, 11979–11988.
32. S. C. Edington, S. Liu and C. R. Baiz, in Methods in Enzymology, ed. J. A. Cotruvo, Academic Press, 2021, vol. 651, pp. 157–191.
33. S. Liu, E. R. Featherston, J. A. Cotruvo and C. R. Baiz, Phys. Chem. Chem. Phys., 2021, 23, 21690–21700.
34. S. A. Cotton, C. R. Chim., 2005, 8, 129–145.
35. A. Smerigan, S. Biswas, F. D. Vila, J. Hong, J. Perez-Aguilar, A. S. Hoffman, L. Greenlee, R. B. Getman and S. R. Bare, Inorg. Chem., 2023, 62, 14523–14532.
36. T. Summers, D. Zhang, J. Sobrinho, A. de Bettencourt-Dias, R. Rousseau, V.-A. Glezakou and D. Cantu, ChemRxiv, 2024, preprint DOI:10.26434/chemrxiv-2024-lgkjh.
37. H. Lumpe, A. Menke, C. Haisch, P. Mayer, A. Kabelitz, K. V. Yusenko, A. Guilherme Buzanich, T. Block, R. Pöttgen, F. Emmerling and L. J. Daumann, Chem. – Eur. J., 2020, 26, 10133–10139.
38. K. Binnemans, Coord. Chem. Rev., 2015, 295, 1–45.
39. A. Beeby, I. M. Clarkson, R. S. Dickins, S. Faulkner, D. Parker, L. Royle, A. S. de Sousa, J. A. G. Williams and M. Woods, J. Chem. Soc., Perkin Trans. 2, 1999, 493–504.
40. R. M. Supkowski and W. D. W. Horrocks, Inorg. Chim. Acta, 2002, 340, 44–48.
41. C. S. Bonnet, P. H. Fries, S. Crouzy, O. Sénèque, F. Cisnetti, D. Boturyn, P. Dumy and P. Delangle, Chem. – Eur. J., 2009, 15, 7083–7093.
42. G. Pintacuda, M. John, X.-C. Su and G. Otting, Acc. Chem. Res., 2007, 40, 206–212.
43. K. N. Allen and B. Imperiali, Curr. Opin. Chem. Biol., 2010, 14, 247–254.
44. W.-M. Liu, M. Overhand and M. Ubbink, Coord. Chem. Rev., 2014, 273–274, 2–12.
45. C. Nitsche and G. Otting, Prog. Nucl. Magn. Reson. Spectrosc., 2017, 98–99, 20–49.
46. D. Joss and D. Häussinger, Prog. Nucl. Magn. Reson. Spectrosc., 2019, 114–115, 284–312.
47. I. Baig, I. Bertini, C. Del Bianco, Y. K. Gupta, Y.-M. Lee, C. Luchinat and A. Quattrone, Biochemistry, 2004, 43, 5562–5573.
48. I. Bertini, C. Del Bianco, I. Gelis, N. Katsaros, C. Luchinat, G. Parigi, M. Peana, A. Provenzani and M. A. Zoroddu, Proc. Natl. Acad. Sci. U. S. A., 2004, 101, 6841–6846.
49. N. C. J. Strynadka and M. N. G. James, Annu. Rev. Biochem., 1989, 58, 951–999.
50. A. Lewit-Bentley and S. Réty, Curr. Opin. Struct. Biol., 2000, 10, 637–643.
51. Z. Grabarek, J. Mol. Biol., 2006, 359, 509–525.
52. J. L. Gifford, M. P. Walsh and H. J. Vogel, Biochem. J., 2007, 405, 199–221.
53. D. B. Halling, B. J. Liebeskind, A. W. Hall and R. W. Aldrich, Proc. Natl. Acad. Sci. U. S. A., 2016, 113(9), 1216–1225.
54. H. G. Brittain, F. S. Richardson and R. B. Martin, J. Am. Chem. Soc., 1976, 98, 8255–8260.
55. B. Martin and F. S. Richardson, Q. Rev. Biophys., 1979, 12, 181–209.
56. W. D. Horrocks and D. R. Sudnick, Acc. Chem. Res., 1981, 14, 384–392.
57. M. A. Valentini and J. C. Wright, Anal. Biochem., 1985, 150, 47–57.
58. C.-L. A. Wang, P. C. Leavis, W. D. Horrocks and J. Gergely, Biochemistry, 1981, 20, 2439–2444.
59. J. Gariepy, B. D. Sykes and R. S. Hodges, Biochemistry, 1983, 22, 1765–1772.
60. K. D. Daughtry, L. J. Martin, A. Sarraju, B. Imperiali and K. N. Allen, ChemBioChem, 2012, 13, 2567–2574.
61. S. Jain, J. T. Welch, W. D. Horrocks and S. J. Franklin, Inorg. Chem., 2003, 42, 8098–8104.
62. J. A. Mattocks, J. J. Jung, C.-Y. Lin, Z. Dong, N. H. Yennawar, E. R. Featherston, C. S. Kang-Yun, T. A. Hamilton, D. M. Park, A. K. Boal and J. A. Cotruvo, Nature, 2023, 618, 87–93.
63. E. R. Featherston, E. J. Issertell and J. A. Cotruvo, J. Am. Chem. Soc., 2021, 143, 14287–14299.
64. M. Nitz, M. Sherawat, K. J. Franz, E. Peisach, K. N. Allen and B. Imperiali, Angew. Chem., Int. Ed., 2004, 43, 3682–3685.
65. K. Barthelmes, A. M. Reynolds, E. Peisach, H. R. A. Jonker, N. J. DeNunzio, K. N. Allen, B. Imperiali and H. Schwalbe, J. Am. Chem. Soc., 2011, 133, 808–819.
66. J. P. MacManus, C. W. Hogue, B. J. Marsden, M. Sikorska and A. G. Szabo, J. Biol. Chem., 1990, 265, 10358–10366.
67. J. Wöhnert, K. J. Franz, M. Nitz, B. Imperiali and H. Schwalbe, J. Am. Chem. Soc., 2003, 125, 13338–13339.
68. L. J. Martin, M. J. Hähnke, M. Nitz, J. Wöhnert, N. R. Silvaggi, K. N. Allen, H. Schwalbe and B. Imperiali, J. Am. Chem. Soc., 2007, 129, 7106–7113.
69. N. R. Silvaggi, L. J. Martin, H. Schwalbe, B. Imperiali and K. N. Allen, J. Am. Chem. Soc., 2007, 129, 7114–7120.
70. D. Barthelmes, M. Gränz, K. Barthelmes, K. N. Allen, B. Imperiali, T. Prisner and H. Schwalbe, J. Biomol. NMR, 2015, 63, 275–282.
71. T. W. Victor, K. H. O’Toole, L. M. Easthon, M. Ge, R. J. Smith, X. Huang, H. Yan, Y. S. Chu, S. Chen, D. Gursoy, M. Ralle, B. Imperiali, K. N. Allen and L. M. Miller, J. Am. Chem. Soc., 2020, 142, 2145–2149.
72. M. Nitz, K. J. Franz, R. L. Maglathlin and B. Imperiali, ChemBioChem, 2003, 4, 272–276.
73. L. J. Martin, B. R. Sculimbrene, M. Nitz and B. Imperiali, QSAR Comb. Sci., 2005, 24, 1149–1157.
74. L. J. Martin and B. Imperiali, in Peptide Libraries: Methods and Protocols, ed. R. Derda, Springer, New York, NY, 2015, pp. 201–220.
75. Y. Kim, J. T. Welch, K. M. Lindstrom and S. J. Franklin, JBIC, J. Biol. Inorg. Chem., 2001, 6, 173–181.
76. J. T. Welch, W. R. Kearney and S. J. Franklin, Proc. Natl. Acad. Sci. U. S. A., 2003, 100, 3725–3730.
77. P. Caravan, J. M. Greenwood, J. T. Welch and S. J. Franklin, Chem. Commun., 2003, 2574–2575.
78. S. Lim and S. J. Franklin, Protein Sci., 2006, 15, 2159–2165.
79. S. W. Wong-Deyrup, Y. Kim and S. J. Franklin, JBIC, J. Biol. Inorg. Chem., 2006, 11, 17–25.
80. K. L. Harris, S. Lim and S. J. Franklin, Inorg. Chem., 2006, 45, 10002–10012.
81. J. A. Cotruvo, E. R. Featherston, J. A. Mattocks, J. V. Ho and T. N. Laremore, J. Am. Chem. Soc., 2018, 140, 15056–15061.
82. C.-L. A. Wang, R. R. Aquaron, P. C. Leavis and J. Gergely, Eur. J. Biochem., 1982, 124, 7–12.
83. E. C. Cook, E. R. Featherston, S. A. Showalter and J. A. Cotruvo, Biochemistry, 2019, 58, 120–125.
84. F. Cisnetti, C. Lebrun and P. Delangle, Dalton Trans., 2010, 39, 3560–3562.
85. F. Cisnetti, C. Gateau, C. Lebrun and P. Delangle, Chem. – Eur. J., 2009, 15, 7456–7469.
86. A. Niedźwiecka, F. Cisnetti, C. Lebrun, C. Gateau and P. Delangle, Dalton Trans., 2012, 41, 3239–3247.
87. A. Niedźwiecka, F. Cisnetti, C. Lebrun and P. Delangle, Inorg. Chem., 2012, 51, 5458–5464.
88. J. Veliscek-Carolan, T. L. Hanley and K. A. Jolliffe, RSC Adv., 2016, 6, 75336–75346.
89. S. Safi, A. Jeanson, J. Roques, P. L. Solari, F. Charnay-Pouget, C. Den Auwer, G. Creff, D. J. Aitken and E. Simoni, Inorg. Chem., 2016, 55, 877–886.
90. I. V. Korendovych and W. F. DeGrado, Q. Rev. Biophys., 2020, 53, e3.
91. K. J. Koebke, T. B. J. Pinter, W. C. Pitts and V. L. Pecoraro, Chem. Rev., 2022, 122, 12046–12109.
92. P. Teare, C. F. Smith, S. J. Adams, S. Anbu, B. Ciani, L. J. C. Jeuken and A. F. A. Peacock, Dalton Trans., 2018, 47, 10784–10790.
93. D. N. Woolfson, Advances in Protein Chemistry, Academic Press, 2005, vol. 70, pp. 79–112.
94. W. D. Kohn, C. M. Kay and R. S. Hodges, J. Pept. Res., 1998, 51, 9–18.
95. W. D. Kohn, C. M. Kay, B. D. Sykes and R. S. Hodges, J. Am. Chem. Soc., 1998, 120, 1124–1132.
96. A. Kashiwada, K. Ishida and K. Matsuda, Bull. Chem. Soc. Jpn., 2007, 80, 2203–2207.
97. M. Samiappan, S. Alasibi, R. Cohen-Luria, A. Shanzer and G. Ashkenasy, Chem. Commun., 2012, 48, 9577–9579.
98. M. R. Berwick, D. J. Lewis, A. W. Jones, R. A. Parslow, T. R. Dafforn, H. J. Cooper, J. Wilkie, Z. Pikramenou, M. M. Britton and A. F. A. Peacock, J. Am. Chem. Soc., 2014, 136, 1166–1169.
99. M. R. Berwick, L. N. Slope, C. F. Smith, S. M. King, S. L. Newton, R. B. Gillis, G. G. Adams, A. J. Rowe, S. E. Harding, M. M. Britton and A. F. A. Peacock, Chem. Sci., 2016, 7, 2207–2216.
100. L. N. Slope, M. G. Hill, C. F. Smith, P. Teare, F. J. de Cogan, M. M. Britton and A. F. A. Peacock, Chem. Commun., 2020, 56, 3729–3732.
101. L. N. Slope, O. J. Daubney, H. Campbell, S. A. White and A. F. A. Peacock, Angew. Chem., Int. Ed., 2021, 60, 24473–24477.
102. S. L. Newton, A. Franke, A. Zahl, G. Molinaro, A. Kenwright, D. J. Smith, I. Ivanovic-Burmazovic, M. M. Britton and A. F. A. Peacock, Dalton Trans., 2023, 52, 15665–15668.
103. C. W. A. Ende, H. Y. Meng, M. Ye, A. K. Pandey and N. J. Zondlo, ChemBioChem, 2010, 11, 1738–1747.
104. A. F. A. Peacock, J. Inorg. Biochem., 2025, 268, 112903.
105. J. K. Myers, C. N. Pace and J. M. Scholtz, Proc. Natl. Acad. Sci. U. S. A., 1997, 94, 2833–2837.
106. J. S. Richardson and D. C. Richardson, Science, 1988, 240, 1648–1652.
107. L. G. Presta and G. D. Rose, Science, 1988, 240, 1632–1641.
108. L. Serrano and A. R. Fersht, Nature, 1989, 342, 296–299.
109. Y. W. Deng, S. Y. Ro and A. C. Rosenzweig, JBIC, J. Biol. Inorg. Chem., 2018, 23, 1037–1047.
110. N. M. Good, M. Fellner, K. Demirer, J. Hu, R. P. Hausinger and N. C. Martinez-Gomez, J. Biol. Chem., 2020, 295, 8272–8284.
111. R. A. Schmitz, N. Picone, H. Singer, A. Dietl, K.-A. Seifert, A. Pol, M. S. M. Jetten, T. R. M. Barends, L. J. Daumann and H. J. M. Op den Camp, mBio, 2021, 12, e01708–21.
112. B. Jahn, A. Pol, H. Lumpe, T. R. M. Barends, A. Dietl, C. Hogendoorn, H. J. M. Op den Camp and L. J. Daumann, ChemBioChem, 2018, 19, 1147–1153.
113. N. M. Good, H. N. Vu, C. J. Suriano, G. A. Subuyuj, E. Skovran and N. C. Martinez-Gomez, J. Bacteriol., 2016, 198, 3109–3118.
114. M. Wehrmann, P. Billard, A. Martin-Meriadec, A. Zegeye and J. Klebensberger, mBio, 2017, 8, e00570–17.
115. J. Huang, Z. Yu, J. Groom, J.-F. Cheng, A. Tarver, Y. Yoshikuni and L. Chistoserdova, ISME J., 2019, 13, 2005–2017.
116. M. Wehrmann, E. M. Elsayed, S. Köbbing, L. Bendz, A. Lepak, J. Schwabe, N. Wierckx, G. Bange and J. Klebensberger, ACS Catal., 2020, 10, 7836–7842.
117. N. M. Good, R. S. Moore, C. J. Suriano and N. C. Martinez-Gomez, Sci. Rep., 2019, 9, 1–12.
118. L. Wang, A. Hibino, S. Suganuma, A. Ebihara, S. Iwamoto, R. Mitsui, A. Tani, M. Shimada, T. Hayakawa and T. Nakagawa, Enzyme Microb. Technol., 2020, 136, 109518.
119. H. Lumpe, A. Pol, H. J. M. Op den Camp and L. J. Daumann, Dalton Trans., 2018, 47, 10463–10472.
120. C. Stines-Chaumeil, F. Mavré, B. Kauffmann, N. Mano and B. Limoges, ACS Omega, 2020, 5, 2015–2026.
121. P. Roszczenko-Jasińska, H. N. Vu, G. A. Subuyuj, R. V. Crisostomo, J. Cai, N. F. Lien, E. J. Clippard, E. M. Ayala, R. T. Ngo, F. Yarza, J. P. Wingett, C. Raghuraman, C. A. Hoeber, N. C. Martinez-Gomez and E. Skovran, Sci. Rep., 2020, 10, 12663.
122. J. L. Hemmann, P. Keller, L. Hemmerle, T. Vonderach, A. M. Ochsner, M. Bortfeld-Miller, D. Günther and J. A. Vorholt, J. Biol. Chem., 2023, 299, 102940.
123. J. A. Mattocks, J. V. Ho and J. A. Cotruvo, J. Am. Chem. Soc., 2019, 141, 2857–2861.
124. A. M. Ochsner, L. Hemmerle, T. Vonderach, R. Nüssli, M. Bortfeld-Miller, B. Hattendorf and J. A. Vorholt, Mol. Microbiol., 2019, 111, 1152–1166.
125. W. B. Larrinaga, J. J. Jung, C.-Y. Lin, A. K. Boal and J. A. Cotruvo, Proc. Natl. Acad. Sci. U. S. A., 2024, 121, e2410926121.
126. S. J. Caldwell, I. C. Haydon, N. Piperidou, P.-S. Huang, M. J. Bick, H. S. Sjöström, D. Hilvert, D. Baker and C. Zeymer, Proc. Natl. Acad. Sci. U. S. A., 2020, 117, 30362–30369.
127. M. A. Blenner, O. Shur, G. R. Szilvay, D. M. Cropek and S. Banta, J. Mol. Biol., 2010, 400, 244–256.
128. W. Abdallah, K. Solanki and S. Banta, ACS Catal., 2018, 8, 1602–1613.
129. F. Khoury, Z. Su and S. Banta, Inorg. Chem., 2024, 63(29), 13223–13230.
130. J. A. Mattocks, J. L. Tirsch and J. A. Cotruvo, in Methods in Enzymology, ed. J. A. Cotruvo, Academic Press, 2021, vol. 651, pp. 23–61.
131. G. Deblonde, Methods in Enzymology, Academic Press, 2021, vol. 651, pp. 1–22.
132. W. R. Harris, C. J. Carrano and K. N. Raymond, J. Am. Chem. Soc., 1979, 101, 2722–2727.
133. E. Falcone and P. Faller, Dalton Trans., 2023, 52, 2197–2208.
134. Z. M. Anwar and H. A. Azab, J. Chem. Eng. Data, 2001, 46, 613–618.
135. H. A. Azab, S. S. Al-Deyab, Z. M. Anwar, I. I. Abd El-Gawad and R. M. Kamel, J. Chem. Eng. Data, 2011, 56, 2613–2625.
136. H. A. Azab, Z. M. Anwar and R. M. Kamel, J. Solution Chem., 2016, 45, 1516–1528.
137. P. Mandal, J. Kretzschmar and B. Drobot, JBIC, J. Biol. Inorg. Chem., 2022, 27, 249–260.
138. R. N. Goldberg, N. Kishore and R. M. Lennen, J. Phys. Chem. Ref. Data, 2002, 31, 231–370.
139. K. J. Franz, M. Nitz and B. Imperiali, ChemBioChem, 2003, 4, 265–271.
140. E. A. Alasadi, W. Choi, X. Chen, J. A. J. Cotruvo and C. R. Baiz, ACS Chem. Biol., 2024, 19, 1056–1065.
141. E. M. Jones, Y. Su, C. Sander, Q. A. Justman, M. Springer and P. A. Silver, ACS Synth. Biol., 2024, 13, 958–962.
142. L. L. Clainche, M. Figuet, V. Montjardet-Bas, S. Blanchard and C. Vita, Biotechnol. Bioeng., 2006, 95, 29–36.
143. S. M. Gutenthaler, S. Tsushima, R. Steudtner, M. Gailer, A. Hoffmann-Röder, B. Drobot and L. J. Daumann, Inorg. Chem. Front., 2022, 9, 4009–4021.
144. S. C. Zondlo, F. Gao and N. J. Zondlo, J. Am. Chem. Soc., 2010, 132, 5619–5621.
145. S. Balakrishnan and N. J. Zondlo, J. Am. Chem. Soc., 2006, 128, 5590–5591.
146. A. R. Urmey and N. J. Zondlo, Biochemistry, 2019, 58, 2822–2833.
147. A. R. Urmey and N. J. Zondlo, Free Radical Biol. Med., 2020, 152, 166–174.
148. F. Gao, B. S. Thornley, C. M. Tressler, D. Naduthambi and N. J. Zondlo, Org. Biomol. Chem., 2019, 17, 3984–3995.
149. L. L. Liu and K. J. Franz, J. Am. Chem. Soc., 2005, 127, 9662–9663.
150. J. A. Mattocks, J. L. Tirsch and J. A. Cotruvo, Methods in Enzymology, Elsevier, 2021, vol. 651, pp. 23–61.
151. B. Mwewa, M. Tadie, S. Ndlovu, G. S. Simate and E. Matinde, J. Environ. Chem. Eng., 2022, 10, 107704.
152. S. Pramanik, S. Kaur, I. Popovs, A. S. Ivanov and S. Jansone-Popova, Eur. J. Inorg. Chem., 2024, e202400064.
153. J. Haiech, C. B. Klee, J. G. Demaille and J. Haiech, Biochemistry, 1981, 20, 3890–3897.
154. D. M. Park, D. W. Reed, M. C. Yung, A. Eslamimanesh, M. M. Lencka, A. Anderko, Y. Fujita, R. E. Riman, A. Navrotsky and Y. Jiao, Environ. Sci. Technol., 2016, 50, 2735–2742.
155. G. J.-P. Deblonde, J. A. Mattocks, D. M. Park, D. W. Reed, J. A. Cotruvo and Y. Jiao, Inorg. Chem., 2020, 59, 11855–11867.
156. T. L. Pham, M. R. Conde González, S. Fazliev, A. Kishore, P. Comba and F. Thomas, ChemBioChem, 2024, 25, e202300715.
157. D. Ghosh, K.-H. Lee, B. Demeler and V. L. Pecoraro, Biochemistry, 2005, 44, 10732–10740.
158. J. A. Peters, K. Djanashvili, C. F. G. C. Geraldes and C. Platas-Iglesias, Coord. Chem. Rev., 2020, 406, 213146.
159. R. K. Wilharm, S.-Y. Huang, I. J. Gugger and V. C. Pierre, Inorg. Chem., 2021, 60, 15808–15817.
160. A. Roca-Sabio, M. Mato-Iglesias, D. Esteban-Gómez, É. Tóth, A. de Blas, C. Platas-Iglesias and T. Rodríguez-Blas, J. Am. Chem. Soc., 2009, 131, 3331–3341.
161. A. Hu, S. N. MacMillan and J. J. Wilson, J. Am. Chem. Soc., 2020, 142, 13500–13506.
162. S. L. Wu and W. D. W. Horrocks, J. Chem. Soc., Dalton Trans., 1997, 1497–1502.
163. A. Hu, E. Aluicio-Sarduy, V. Brown, S. N. MacMillan, K. V. Becker, T. E. Barnhart, V. Radchenko, C. F. Ramogida, J. W. Engle and J. J. Wilson, J. Am. Chem. Soc., 2021, 143, 10429–10440.
164. T. Hatanaka, N. Kikkawa, A. Matsugami, Y. Hosokawa, F. Hayashi and N. Ishida, Sci. Rep., 2020, 10, 19468.
165. H. Singer, B. Drobot, C. Zeymer, R. Steudtner and L. J. Daumann, Chem. Sci., 2021, 12, 15581–15587.
166. J. A. Mattocks, J. L. Tirsch and J. A. Cotruvo, Methods in Enzymology, Elsevier, 2021, vol. 651, pp. 23–61.
167. Z. Dong, J. A. Mattocks, J. A. Seidel, J. A. Cotruvo and D. M. Park, Sep. Purif. Technol., 2024, 333, 125919.
168. J. Seidel, P. Diep, Z. Dong, J. A. Cotruvo and D. M. Park, JACS Au, 2024, 4(11), 4273–4284.
169. C. S. Bonnet, M. Devocelle and T. Gunnlaugsson, Org. Biomol. Chem., 2012, 10, 126–133.
170. G. J.-P. Deblonde, J. A. Mattocks, H. Wang, E. M. Gale, A. B. Kersting, M. Zavarin and J. A. Cotruvo, J. Am. Chem. Soc., 2021, 143, 15769–15783.
171. G. J.-P. Deblonde, J. A. Mattocks, Z. Dong, P. T. Wooddy, J. A. Cotruvo and M. Zavarin, Sci. Adv., 2021, 7, eabk0273.
172. G. J.-P. Deblonde, K. Morrison, J. A. Mattocks, J. A. Cotruvo, M. Zavarin and A. B. Kersting, Environ. Sci. Technol., 2023, 57, 20830–20843.
173. R. M. Smith and A. E. Martell, Sci. Total Environ, 1987, 64, 125–147.
174. J. A. Mattocks, J. A. Cotruvo and G. J.-P. Deblonde, Chem. Sci., 2022, 13, 6054–6066.
175. S. Özçubukçu, K. Mandal, S. Wegner, M. P. Jensen and C. He, Inorg. Chem., 2011, 50, 7937–7939.
176. H. Singer, R. Steudtner, A. S. Klein, C. Rulofs, C. Zeymer, B. Drobot, A. Pol, N. Cecilia Martinez-Gomez, H. J. M. Op den Camp and L. J. Daumann, Angew. Chem., Int. Ed., 2023, 62, e202303669.
177. A. Shah, M. J. Taylor, G. Molinaro, S. Anbu, M. Verdu, L. Jennings, I. Mikulska, S. Diaz-Moreno, H. E. L. Mkami, G. M. Smith, M. M. Britton, J. E. Lovett and A. F. A. Peacock, Proc. Natl. Acad. Sci. U. S. A., 2023, 120, e2219036120.
178. C. S. Bonnet, M. Devocelle and T. Gunnlaugsson, Chem. Commun., 2008, 4552–4554.
179. S. K. Drake and J. J. Falke, Biochemistry, 1996, 35, 1753–1760.
180. M. C. Heffern, L. M. Matosziuk and T. J. Meade, Chem. Rev., 2014, 114, 4496–4539.
181. A. J. Amoroso and S. J. A. Pope, Chem. Soc. Rev., 2015, 44, 4723–4742.
182. E. Mathieu, A. Sipos, E. Demeyere, D. Phipps, D. Sakaveli and K. E. Borbas, Chem. Commun., 2018, 54, 10021–10035.
183. U. Cho and J. K. Chen, Cell Chem. Biol., 2020, 27, 921–936.
184. J. H. S. K. Monteiro, Molecules, 2020, 25, 2089.
185. R. Lengacher, K. E. Martin, D. Śmiłowicz, H. Esseln, P. Lotlikar, A. Grichine, O. Maury and E. Boros, J. Am. Chem. Soc., 2023, 145, 24358–24366.
186. J.-C. G. Bünzli, Chem. Rev., 2010, 110, 2729–2755.
187. S. V. Eliseeva, Angew. Chem., Int. Ed., 2015, 54, 8598.
188. A. D’Aléo, F. Pointillart, L. Ouahab, C. Andraud and O. Maury, Coord. Chem. Rev., 2012, 256, 1604–1620.
189. J.-C. G. Bünzli, Coord. Chem. Rev., 2015, 293–294, 19–47.
190. D. Kovacs and K. E. Borbas, Coord. Chem. Rev., 2018, 364, 1–9.
191. J. De Jersey, P. J. Morley and R. B. Martin, Biophys. Chem., 1981, 13, 233–243.
192. M. Latva, H. Takalo, V.-M. Mukkala, C. Matachescu, J. C. Rodríguez-Ubis and J. Kankare, J. Lumin., 1997, 75, 149–169.
193. W. D. Horrocks, J. P. Bolender, W. D. Smith and R. M. Supkowski, J. Am. Chem. Soc., 1997, 119, 5972–5973.
194. E. Falcone, P. Gonzalez, L. Lorusso, O. Sénèque, P. Faller and L. Raibaut, Chem. Commun., 2020, 56, 4797–4800.
195. A. M. Reynolds, B. R. Sculimbrene and B. Imperiali, Bioconjugate Chem., 2008, 19, 588–591.
196. J. Wang, X. Liu, K. Li, T. Shi, Q. Xu, T. Peng, Q. Huang, Z. Gao, H. Zhou, W. Lu and J. Wang, J. Am. Chem. Soc., 2025, 147, 15205–15215.
197. M. Starck, J. D. Fradgley, S. Di Vita, J. A. Mosely, R. Pal and D. Parker, Bioconjugate Chem., 2020, 31, 229–240.
198. H.-F. Chau, Y. Wu, W.-Y. Fok, W. Thor, W. C.-S. Cho, P. Ma, J. Lin, N.-K. Mak, J.-C. G. Bünzli, L. Jiang, N. J. Long, H. L. Lung and K.-L. Wong, JACS Au, 2021, 1, 1034–1043.
199. J.-H. Choi, G. Fremy, T. Charnay, N. Fayad, J. Pécaut, S. Erbek, N. Hildebrandt, V. Martel-Frachet, A. Grichine and O. Sénèque, Inorg. Chem., 2022, 61, 20674–20689.
200. C. Cepeda, S. A. Denisov, D. Boturyn, N. D. McClenaghan and O. Sénèque, Inorg. Chem., 2021, 60, 17426–17434.
201. M. Isaac, L. Raibaut, C. Cepeda, A. Roux, D. Boturyn, S. V. Eliseeva, S. Petoud and O. Sénèque, Chem. – Eur. J., 2017, 23, 10992–10996.
202. Y.-H. Yeung, H.-F. Chau, H.-Y. Kai, W. Zhou, K. H.-Y. Chan, W. Thor, L. J. Charbonnière, F. Zhang, Y. Fan, Y. Wu and K.-L. Wong, Adv. Opt. Mater., 2024, 12, 2302070.
203. C. Cepeda, L. Raibaut, G. Fremy, S. V. Eliseeva, A. Romieu, J. Pécaut, D. Boturyn, S. Petoud and O. Sénèque, Chem. – Eur. J., 2020, 26, 13476–13483.
204. C. Penas, J. L. Mascareñas and M. E. Vázquez, Chem. Sci., 2016, 7, 2674–2678.
205. B. R. Sculimbrene and B. Imperiali, J. Am. Chem. Soc., 2006, 128, 7346–7352.
206. K. Y. Huang, L. Cardenas, A. D. Ellington and D. J. F. Walker, Nat. Commun., 2024, 15, 9200.
207. L. Raibaut, W. Vasseur, G. D. Shimberg, C. Saint-Pierre, J.-L. Ravanat, S. L. J. Michel and O. Sénèque, Chem. Sci., 2017, 8, 1658–1664.
208. E. Pazos, M. Goličnik, J. L. Mascareñas and M. E. Vázquez, Chem. Commun., 2012, 48, 9534–9536.
209. R. Sánchez-Fernández, A. Sánchez-Temprano, D. Esteban-Gómez and E. Pazos, ChemBioChem, 2023, 24, e202300072.
210. E. Falcone, B. Vileno, M. Hoang, L. Raibaut and P. Faller, J. Inorg. Biochem., 2021, 221, 111478.
211. M. J. Scheuermann, C. R. Forbes and N. J. Zondlo, Biochemistry, 2018, 57, 6956–6963.
212. J. Wahsner, E. M. Gale, A. Rodríguez-Rodríguez and P. Caravan, Chem. Rev., 2019, 119, 957–1057.
213. M. E. Ladd, P. Bachert, M. Meyerspeer, E. Moser, A. M. Nagel, D. G. Norris, S. Schmitter, O. Speck, S. Straub and M. Zaiss, Prog. Nucl. Magn. Reson. Spectrosc., 2018, 109, 1–50.
214. P. Caravan, Chem. Soc. Rev., 2006, 35, 512–523.
215. P. Caravan, J. J. Ellison, T. J. McMurry and R. B. Lauffer, Chem. Rev., 1999, 99, 2293–2352.
216. D. H. Powell, O. M. N. Dhubhghaill, D. Pubanz, L. Helm, Y. S. Lebedev, W. Schlaepfer and A. E. Merbach, J. Am. Chem. Soc., 1996, 118, 9333–9346.
217. J. J. Yang, J. Yang, L. Wei, O. Zurkiya, W. Yang, S. Li, J. Zou, Y. Zhou, A. L. W. Maniccia, H. Mao, F. Zhao, R. Malchow, S. Zhao, J. Johnson, X. Hu, E. Krogstad and Z.-R. Liu, J. Am. Chem. Soc., 2008, 130, 9260–9267.
218. S. Xue, H. Yang, J. Qiao, F. Pu, J. Jiang, K. Hubbard, K. Hekmatyar, J. Langley, M. Salarian, R. C. Long, R. G. Bryant, X. P. Hu, H. E. Grossniklaus, Z.-R. Liu and J. J. Yang, Proc. Natl. Acad. Sci. U. S. A., 2015, 112, 6607–6612.
219. C. S. Bonnet, P. H. Fries, S. Crouzy and P. Delangle, J. Phys. Chem. B, 2010, 114, 8770–8781.
220. Y. Liu, D. Gao, Y. He, J. Ma, S. Y. Chong, X. Qi, H. J. Ting, Z. Luo, Z. Yi, J. Tang, C. Chang, J. Wang, Z. Sheng, H. Zheng and X. Liu, Nat. Commun., 2024, 15, 9834.
221. D. Li, M. Kirberger, J. Qiao, Z. Gui, S. Xue, F. Pu, J. Jiang, Y. Xu, S. Tan, M. Salarian, O. Ibhagui, K. Hekmatyar and J. J. Yang, Invest. Radiol., 2024, 59, 170–186.
222. F. Pu, M. Salarian, S. Xue, J. Qiao, J. Feng, S. Tan, A. Patel, X. Li, K. Mamouni, K. Hekmatyar, J. Zou, D. Wu and J. J. Yang, Nanoscale, 2016, 8, 12668–12682.
223. M. Salarian, H. Yang, R. C. Turaga, S. Tan, J. Qiao, S. Xue, Z. Gui, G. Peng, H. Han, P. Mittal, H. E. Grossniklaus and J. J. Yang, Biomaterials, 2019, 224, 119478.
224. S. Tan, H. Yang, S. Xue, J. Qiao, M. Salarian, K. Hekmatyar, Y. Meng, R. Mukkavilli, F. Pu, O. Y. Odubade, W. Harris, Y. Hai, M. L. Yushak, V. M. Morales-Tirado, P. Mittal, P. Z. Sun, D. Lawson, H. E. Grossniklaus and J. J. Yang, Sci. Adv., 2020, 6, eaav7504.
225. H. Yang, S. Tan, J. Qiao, Y. Xu, Z. Gui, Y. Meng, B. Dong, G. Peng, O. Y. Ibhagui, W. Qian, J. Lu, Z. Li, G. Wang, J. Lai, L. Yang, H. E. Grossniklaus and J. J. Yang, Cancer Gene Ther., 2022, 29, 1827–1839.
226. O. Y. Ibhagui, D. Li, H. Han, G. Peng, M. L. Meister, Z. Gui, J. Qiao, M. Salarian, B. Dong, Y. Yuan, Y. Xu, H. Yang, S. Tan, G. Satyanarayana, S. Xue, R. C. Turaga, M. Sharma, Y. Hai, Y. Meng, K. Hekmatyar, P. Sun, G. Sica, X. Ji, Z.-R. Liu and J. J. Yang, Chem. Biomed. Imaging, 2023, 1, 268–285.
227. Q. N. Do, J. S. Ratnakar, Z. Kovács and A. D. Sherry, ChemMedChem, 2014, 9, 1116–1129.
228. G. Angelovski and É. Tóth, Chem. Soc. Rev., 2017, 46, 324–336.
229. J. Lux and A. D. Sherry, Curr. Opin. Chem. Biol., 2018, 45, 121–130.
230. C. S. Bonnet and É. Tóth, Curr. Opin. Chem. Biol., 2021, 61, 154–169.
231. H. Lu, A. Chen, X. Zhang, Z. Wei, R. Cao, Y. Zhu, J. Lu, Z. Wang and L. Tian, Nat. Commun., 2022, 13, 7948.
232. M. Isaac, A. Pallier, F. Szeremeta, P.-A. Bayle, L. Barantin, C. S. Bonnet and O. Sénèque, Chem. Commun., 2018, 54, 7350–7353.
233. K. P. Malikidogo, A. Pallier, F. Szeremeta, C. S. Bonnet and O. Sénèque, Dalton Trans., 2023, 52, 6260–6266.
234. M. Sethurajan, E. D. van Hullebusch, D. Fontana, A. Akcil, H. Deveci, B. Batinic, J. P. Leal, T. A. Gasche, M. Ali Kucuker, K. Kuchta, I. F. F. Neto, H. M. V. M. Soares and A. Chmielarz, Crit. Rev. Environ. Sci. Technol., 2019, 49, 212–275.
235. K. Binnemans, P. T. Jones, T. Müller and L. Yurramendi, J. Sustainable Metall., 2018, 4, 126–146.
236. K. Binnemans, P. T. Jones, B. Blanpain, T. Van Gerven and Y. Pontikes, J. Cleaner Prod., 2015, 99, 17–38.
237. S. Pramanik, S. Kaur, I. Popovs, A. S. Ivanov and S. Jansone-Popova, Eur. J. Inorg. Chem., 2024, e202400064.
238. N. C. Martinez-Gomez, H. N. Vu and E. Skovran, Inorg. Chem., 2016, 55, 10083–10089.
239. S. Dev, A. Sachan, F. Dehghani, T. Ghosh, B. R. Briggs and S. Aggarwal, Chem. Eng. J., 2020, 397, 124596.
240. R. M. Brown, A. Mirkouei, D. Reed and V. Thompson, Renewable Sustainable Energy Rev., 2023, 173, 113099.
241. M. Chakankar, F. Lederer, R. Jain, S. Matys, S. Kutschke and K. Pollmann, Management of Electronic Waste, John Wiley & Sons, Ltd, 2024, pp. 375–405.
242. X. Qian, C. Ma, H. Zhang and K. Liu, Bioorg. Chem., 2024, 143, 107040.
243. M. Vítová and D. Mezricky, World J. Microbiol. Biotechnol., 2024, 40, 189.
244. H. Singer, R. Steudtner, I. Sottorff, B. Drobot, A. Pol, H. J. M. O. den Camp and L. J. Daumann, Chem. Commun., 2023, 59, 9066–9069.
245. I. Chae, A. Shivkumar, F. M. Doyle and S.-W. Lee, Nano Lett., 2024, 24, 9946–9952.
246. I. Chae, F. M. Doyle and S.-W. Lee, TMS 2025 154th Annual Meeting & Exhibition Supplemental Proceedings, Springer, Cham, 2025, pp. 1139–1148.
247. D. M. Park, A. Brewer, D. W. Reed, L. N. Lammers and Y. Jiao, Environ. Sci. Technol., 2017, 51, 13471–13480.
248. A. Brewer, E. Chang, D. M. Park, T. Kou, Y. Li, L. N. Lammers and Y. Jiao, Environ. Sci. Technol., 2019, 53, 7714–7723.
249. A. Brewer, A. Dohnalkova, V. Shutthanandan, L. Kovarik, E. Chang, A. M. Sawvel, H. E. Mason, D. Reed, C. Ye, W. F. Hynes, L. N. Lammers, D. M. Park and Y. Jiao, Environ. Sci. Technol., 2019, 53, 13888–13897.
250. D. Park, A. Middleton, R. Smith, G. Deblonde, D. Laudal, N. Theaker, H. Hsu-Kim and Y. Jiao, Sep. Purif. Technol., 2020, 241, 116726.
251. E. Chang, A. W. Brewer, D. M. Park, Y. Jiao and L. N. Lammers, Appl. Geochem., 2020, 112, 104478.
252. X. Xie, K. Yang, Y. Lu, Y. Li, J. Yan, J. Huang, L. Xu, M. Yang and Y. Yan, J. Hazard. Mater., 2022, 438, 129561.
253. M. Gut, T. Wilhelm, O. Beniston, S. Ogundipe, C.-C. Kuo, K. Nguyen and A. Furst, Adv. Mater., 2025, 37, 2412607.
254. F. F. Khoury, B. S. Heater, D. R. Marzolf, S. S. Abeyrathna, J. W. Picking, P. Kumar, S. A. Higgins, R. Jones, A. T. Lewis, K. H. Kucharzyk and S. Banta, Chem. Sci., 2025, 16, 15333–15346.
255. P. K. R. Tay, A. Manjula-Basavanna and N. S. Joshi, Green Chem., 2018, 20, 3512–3520.
256. R. A. Scheel, J. K. Sahoo, L. D. Morton, Z. Xia, J. R. Blum, Z. Martin-Moldes and D. L. Kaplan, ACS ES&T Eng., 2024, 4(9), 2135–2144.
257. Z. Dong, J. A. Mattocks, G. J.-P. Deblonde, D. Hu, Y. Jiao, J. A. Cotruvo and D. M. Park, ACS Cent. Sci., 2021, 7, 1798–1808.
258. Z. Hussain, S. Kim, J. Cho, G. Sim, Y. Park and I. Kwon, Adv. Funct. Mater., 2022, 32, 2109158.
259. Z. Hussain, D. Dwivedi and I. Kwon, Front. Bioeng. Biotechnol., 2024, 12 DOI:10.3389/fbioe.2024.1385845.
260. Q. Ye, X. Jin, B. Zhu, H. Gao and N. Wei, Environ. Sci. Technol., 2023, 57, 4276–4285.
261. L. Johnson and C. E. Duval, Langmuir, 2025, 41, 9581–9589.
262. Q. Ye, X. Jin, B. Zhu, H. Gao and N. Wei, Environ. Sci. Technol., 2023, 57, 4276–4285.
263. H. Jin, D. M. Park, M. Gupta, A. W. Brewer, L. Ho, S. L. Singer, W. L. Bourcier, S. Woods, D. W. Reed, L. N. Lammers, J. W. Sutherland and Y. Jiao, ACS Sustainable Chem. Eng., 2017, 5, 10148–10155.
264. J. T. Welch, M. Sirish, K. M. Lindstrom and S. J. Franklin, Inorg. Chem., 2001, 40, 1982–1984.
265. M. Sirish and S. J. Franklin, J. Inorg. Biochem., 2002, 91, 253–258.
266. Y. Kim and S. J. Franklin, Inorg. Chim. Acta, 2002, 341, 107–112.
267. L. Wang, K. Liu, Z. Song, H. Do, L. Yang, J. Wu, L. Jiang and H. Yu, Colloids Surf., B, 2025, 251, 114596.
268. M. J. Allen, Synlett, 2016, 1310–1317.
269. H. Pellissier, Coord. Chem. Rev., 2017, 336, 96–151.
270. Y. Qiao and E. J. Schelter, Acc. Chem. Res., 2018, 51, 2926–2936.
271. A. Prieto and F. Jaroschik, Curr. Org. Chem., 2022, 26, 6–41.
272. N. Mahieu, J. Piątkowski, T. Simler and G. Nocton, Chem. Sci., 2023, 14, 443–457.
273. R. Bhimpuria, R. Charaf, K. Ye, A. Thapper, H. Sathyan, M. Ahlquist, L. Hammarström and K. E. Borbas, Chem, 2025, 102547.
274. R. Bhimpuria, M. Tomar, A. Thapper, M. Ahlquist and K. E. Borbas, Chem, 2025, 102450.
275. F. Schwizer, Y. Okamoto, T. Heinisch, Y. Gu, M. M. Pellizzoni, V. Lebrun, R. Reuter, V. Köhler, J. C. Lewis and T. R. Ward, Chem. Rev., 2018, 118, 142–231.
276. P. Kouba, P. Kohout, F. Haddadi, A. Bushuiev, R. Samusevich, J. Sedlar, J. Damborsky, T. Pluskal, J. Sivic and S. Mazurenko, ACS Catal., 2023, 13, 13863–13895.
277. P. Notin, N. Rollins, Y. Gal, C. Sander and D. Marks, Nat. Biotechnol., 2024, 42, 216–228.
278. A. Winnifrith, C. Outeiral and B. L. Hie, Curr. Opin. Struct. Biol., 2024, 86, 102794.
279. K. K. Yang, Z. Wu and F. H. Arnold, Nat. Methods, 2019, 16, 687–694.
280. J. A. Mattocks, J. J. Jung, C.-Y. Lin, Z. Dong, N. H. Yennawar, E. R. Featherston, C. S. Kang-Yun, T. A. Hamilton, D. M. Park, A. K. Boal and J. A. Cotruvo, Nature, 2023, 618, 87–93.

---

Footnote

† These authors contributed equally to this work.

---