2.1 Ln-binding peptides inspired from Ca²⁺ binding loops

Common Ca²⁺ binding loops in proteins such as calmodulin have 12 amino acids and the metal ion is coordinated by the O-donors of natural amino acid side-chains such as carboxylates of aspartates (Asp) or glutamates (Glu), phenolates of tyrosines (Tyr), amides of asparagines (Asn) and glutamines (Gln) or amides of the backbone peptide linkages. Small peptides based on these Ca²⁺ binding loops have been the starting point for the design of Ln³⁺–peptide complexes [30–32]. Szabo et al. designed 14-mer synthetic peptide loops from proteins of the calmodulin family to complex Tb³⁺ [33,34]. The consensus sequence DKDGDGYIDFEE was obtained by examination of more than 200 species variant loops. It was demonstrated that peptides bearing a too large number of negatively charged amino acid may induce electrostatic repulsion and decrease the affinity for the metal ion. Therefore, one charged Asp (D) residue was replaced by a neutral Asn (N) residue (in position 3) to give the peptide P^N, which shows a micromolar affinity for La³⁺ (Table 1) [32]. Aromatic residues such as tryptophan or tyrosine were introduced in the binding loop to benefit from the antenna effect and to sensitize Tb³⁺ luminescence (Fig. 1). Position 7 for the aromatic Tb³⁺ sensitizer in the peptide sequence is optimal for luminescence enhancement since the aromatic ring in the closest to the metal ion, only 5 Å [33].

Since then, many short peptide sequences derived from Ca²⁺ binding proteins have been studied [35–39]. Among them, chimeric peptides such as P3W (Table 1), which have a transcription factor helix-turn-helix (HTH) DNA-binding domain and a topologically equivalent Ca-binding EF-hand motif, may be considered as artificial endonucleases [40]. P3W was shown to bind Ln³⁺ ions with a micromolar affinity [41] and to promote DNA hydrolysis thanks to the presence of the metal ion [42,43]. Moreover, the Gd³⁺ complex of P3W showed a good MRI contrast efficiency, named relaxivity, which is even highly enhanced when the complex binds to DNA due to the larger molecular weight of the peptide-DNA adduct [44].

All Ln-binding peptides derived from Ca²⁺ binding loops demonstrated micromolar affinity for Ln³⁺. This is why these short peptide sequences and structures needed further optimization, specifically for intense luminescence and higher binding affinity.

2.2 Optimized Ln-binding tags

Ln-binding tags made up entirely of natural and encodable amino acids have been optimized for Tb³⁺ binding and luminescence by Imperiali et al. [45,46]. Prototype sequences were based on calcium-binding motifs of EF-hand proteins, which provide six coordinating residues [47]. Libraries of peptides were designed starting from the 14-mer LBT1 (Table 1) by varying C-terminal ligating residues, adding flanking hydrophobic residues and randomizing sensitizing residue and non-coordinating residue position [45]. The linear sequence of LBT (Table 1) was then selected by screening luminescent Tb³⁺ complexes. It demonstrates an affinity for Tb³⁺ in the nanomolar range, which represents a significant stabilization with respect to the starting prototype LBT1 [46]. The crystallographic structure of the Tb³⁺ complex of LBT reveals that Tb³⁺ is bound in an eight-coordinate complex without any inner-sphere water molecule, which leads to intense luminescence properties of the latter ion. Besides introduction of a disulfide bond in optimal positions in the cyclic peptide, LBT^SS (Table 1) provides an impressive 30-fold increase in Tb³⁺ affinity compared to the best linear sequence of LBT. The expression of LBT in proteins has proven its efficiency for the determination of protein concentration by luminescence, [47] the structural resolution of biomolecules by solution paramagnetic NMR or X-ray diffraction [48–50] and the study of protein interactions thanks to luminescence resonance energy transfer [51]. More recently, this sequence could also be introduced into loop regions of proteins without significantly changing the protein structure [52].

2.3 Phosphorylated peptides

The influence of peptide's phosphorylation on Ln³⁺ complexation was studied for α-synuclein (α-syn) fragment 119-132 because its sequence, in particular the carboxylate groups arrangement, is similar to Ca²⁺ binding loops [53,54]. The type and location of the phosphorylated amino acid (Tyr or Ser) have a dramatic influence on the metal-binding properties of the peptides. The largest stability constant for 1:1 Tb³⁺ binding was obtained for the peptide pY125 (Table 1), with a phosphorylated tyrosine. Besides, comparing the phosphorylated peptide with the unphosphorylated version demonstrates that phosphorylation increases the affinity for metal ions.

2.4 Lanthanide-fingers

Other native sequences than Ca²⁺ binding loops may be modified to generate Ln-binding peptides. A “lanthanide-finger” was recently designed by replacing the cysteine and histidine residues involved in the metal coordination of a compact zinc finger (Cys₂His₂) by amino acids with O-donors. The cysteines and histidines were substituted by aspartic acids and glutamic acids, respectively, to generate a tetracarboxylate peptide ligand (Asp₂Glu₂). LF4 demonstrate the largest affinity for Ln³⁺ with a micromolar affinity for Eu³⁺ (Table 1) [55].

2.5 De novo designed peptides

Another approach consists in preorganizing the side-chains of coordinating amino acids in de novo designed peptides to afford well-defined Ln³⁺ coordination sites. This methodology was applied in de novo designed helical coiled coils: metal ion binding sites namely glutamates were incorporated into heptad positions which commonly contain hydrophilic side-chains and binding of La³⁺ was demonstrated to enhance the stability of the overall supramolecular peptide assembly, even though the binding of La³⁺ appeared to be weak (in the mM range) [56].

More recently, the cyclodecapeptide P^A (Table 1) was tested as a novel peptide architecture to coordinate Ln³⁺. Indeed P^A is particularly adapted to metal coordination since its β-sheet structure allows the preorientation of four amino acid side-chains of glutamic or aspartic acids to coordinate the Ln³⁺ ion. The solution NMR structure of the free peptide demonstrates that the four carboxylic acid groups are oriented in the same half-space of the β-sheet peptide scaffold, which is perfectly suitable to coordinate metal ions [57]. A mononuclear Ln³⁺ complex with two water molecule in the Ln³⁺ coordination sphere is specifically formed with an affinity constant in the same range or even larger than EF-hand motifs, emphasizing the relevance of this de novo approach using cyclic templates. Interestingly, the corresponding Gd³⁺ complex combines the relaxation properties of Gd³⁺ with the high hydrophilicity of the peptide scaffold to provide large second-sphere contributions to water relaxation [57,58].

Peptide sequences built exclusively from natural amino acids exhibit affinities for Ln³⁺ in the micromolar range or in the low nanomolar range for the best Ln-binding tags, which structure is perfectly suited to Ln³⁺ coordination. Indeed, these natural residues bind metal ions with carboxylates, phenolates, amides or phosphates, which show a moderate affinity for Ln³⁺ ions. Therefore, all these peptides with O-donors have a modest affinity for lanthanide ions in comparison with synthetic multidentate ligands such as polyaminocarboxylates [15]. As mentioned before, the stability of the lanthanide complex is a key-parameter for the design of bioprobes in particular to avoid release of the cation and its coordination in other metal binding-sites. Therefore, non standard amino acids are increasingly used for the design of peptide ligands to introduce strong donor groups for metal complexation or to improve the sensitization of metal ion emission.

3.1 Unnatural amino acids for Ln³⁺ sensitization

Time-resolved luminescence of lanthanides is particularly attractive for applications of these ions as luminescent bioprobes, since it allows to eliminate the background natural fluorescence. The aromatic groups of tyrosine and tryptophan (Fig. 2) are Tb³⁺ sensitizers as exemplified by lanthanide–peptide complexes presented in the previous section. The overall quantum yield, ΦLnL, is a key-parameter since it characterizes the luminescence efficiency of the complex: it is equal to the ratio between the number of emitted photons divided by the number of absorbed photons. Two important factors contribute to the overall quantum yield: the metal-centered light emission or intrinsic quantum yield, ΦLnLn, and the sensitization efficiency, ηsens (Eq. 1) [59–62].

The intrinsic quantum yield, ΦLnLn, reflects the extent of non-radiative deactivation processes, which compete with luminescence of the lanthanide ion. These deactivation processes may occur both in the inner- and outer-coordination sphere of the metal ion and are for instance vibration-induced processes or photo-induced electron transfer deactivations. Vibrations of bound ligands such as OH, NH or CH are particularly efficient non-radiative deactivation processes. Hence, to get efficient Ln³⁺ emission, water molecules should be excluded from the coordination sphere to get low hydration numbers, which may be evaluated from Ln³⁺ luminescence lifetimes in light water and heavy water according to empirical equations derived from series of complexes [63–65].

The sensitization efficiency is highly dependent on the energy gap between the sensitizer excited state (donor) and the Ln³⁺ emitting level (acceptor). It is generally considered that a difference of 2500–3500 cm⁻¹ between these two energy levels is ideal to optimize the energy transfer and minimize potential back-transfer [10].

Another key-parameter is the donor-acceptor distance, with a 1/r⁶ relationship for a dipole-dipole mechanism [26,29,66]. The distance between the Trp indole sensitizer and the accepting Ln³⁺ ion measured in the X-ray structure of the Tb³⁺-LBT complex is 7 Å [46]. Since the aromatic indole group is not directly coordinated to the metal, the sensitizer to metal distance is quite long and the energy transfer is clearly a limiting parameter in the quantum yields of these complexes. But, even though the energy transfer from Trp to Tb³⁺ is not optimum, the latter ion is significantly sensitized upon binding to the peptides, which Tb³⁺ complexes may be detected in the low nanomolar range.

The sensitization of Eu³⁺ by Trp and Tyr is far less efficient because of photo-induced electron transfer, which highly contributes with this readily reduced Ln³⁺ ion [27,67]. This mechanism quenches the aromatic group fluorescence without sensitizing the red-emitting Eu³⁺ ion.

To expand the potential applications of Ln-binding peptides, in particular to foresee their in vivo use, artificial chromophores (Fig. 2) with higher excitation wavelengths were introduced in Ln-binding tags [68] or in peptide sequences derived from the parvalbumin Ca²⁺ binding loop [38]. Carbostyril 124 (cs124) was demonstrated to sensitize Tb³⁺ luminescence upon excitation at 337 nm, with a better efficiency than Trp, without affecting the affinity and structure of the complex [68]. Moreover, whereas Trp-containing peptides are not able to sensitize Eu³⁺, the two lower energy chromophores acridone (Acd) [68] and naphthalimide (Naph) [38] introduced in short peptide sequences were successfully used to populate the Eu³⁺ excited state.

Unnatural amino acids may also be used to increase the peptide's affinity for Ln³⁺ ions as illustrated in the next sections.

3.2 Unnatural amino acids to enhance peptide affinity for Ln³⁺ ions

Some metal ions such as Zn²⁺, Cu⁺, Cu²⁺, Pb²⁺ or Hg²⁺ display large affinities for natural binding groups in proteins, such as the soft sulfur donors of cysteines or methionines or the borderline nitrogen donor of histidines. Therefore, efficient binding peptides of these metal ions can readily be obtained by incorporating cysteines, histidines or methionines in the sequences. Some of these metal-binding peptides are inspired from natural binding sites like metal transporters [69–77] or zinc fingers, [78–80] others are de novo designed highly-organized peptide architectures such as triple coiled coil assemblies [81,82] or pseudopeptides with a chemical scaffold to append several peptide units [83–86].

On the contrary, hard Ln³⁺ cations do not show such large affinities for natural binding sites as illustrated in Section 1. Therefore, high affinity Ln-binding groups were appended to synthetic unnatural amino acids to obtain more stable Ln-binding peptides. γ-carboxyglutamic acid (Gla), which contains two carboxylate substituents on the γ carbon (the side chain is a malonate type group, Fig. 3), was introduced in peptide chains to amplify the effects of interhelical repulsions and metal-binding on coiled coil structures. Whereas the apo form is destabilized by strong electrostatic repulsions, the folded form is stabilized by Ln³⁺ binding. Nevertheless, despite the presence of several bidentate Gla amino acids, the Ln³⁺ complex stability remains in the micromolar range [87].

This led us to consider using synthetic amino acids with a larger affinity for Ln³⁺, i.e. with appended polydentate chelating groups of the polyaminocarboxylate family [4]. A related strategy is to functionalize the peptide or the protein with a Ln³⁺ chelate to get peptide/protein conjugates [88–91]. The latter benefit simultaneously from the stability of the Ln³⁺ chelates and from peptide-based properties such as biomolecular recognition. In these conjugates, the binding site is appended at the periphery of the biomolecule and no efficient coupling between the properties of the chelating appendage and that of the peptide substructure is thus expected.

On the contrary, we chose to incorporate the Ln³⁺ binding site embedded in the peptide structure in order to intimately couple the metal ion and the peptide scaffold properties. Among metal-chelating amino acid side-chains, aminodiacetate groups have first attracted our attention [92–94]. The corresponding synthetic residues are referred to as Ada_n, n being the length of the alkyl chain separating the peptide backbone from the aminodiacetate nitrogen (Fig. 3). Ada₂ and Ada₃ are tridentate ligands, whereas the shortest analogue Ada₁ may act as a tetradentate donor as the coordination of the backbone carbonyl is favored due to the formation of a six-membered chelate ring. More recently the novel pentadentate amino acid Ed3a₂ (Fig. 3), which carries an ethylenediamine triacetate side-chain was synthesized to increase the denticity of the peptides.

The chelating unnatural amino acids Ada_n and Ed3a₂ were synthesized with protecting groups compatible with Fmoc/tBu strategy in solid phase peptide synthesis, i.e. in the form Fmoc-aa(tBu)_x-OH. The latter display normal reactivity in the on-resin synthetic steps, which allowed us to synthesize hexapeptides in the 50-mg scale with good yields.

4.1 Design of hexadentate peptides

To design peptides with a Ln-ligating site embedded in the peptide structure, we first inserted two Ada_n unnatural amino acids in the sequence to get potentially hexadentate ligands. However, the mere inclusion of chelating amino acids in an arbitrary peptide sequence is insufficient to achieve highly stable mononuclear complexes because the chelating side-chains would show independent behavior. To overcome this difficulty, the peptide backbone was used as a non-innocent spacer between the two chelating groups and chosen to favor a β-turn containing structure. In such a turn, a hydrogen bond is usually present and residues in position i and i+3 are brought in close proximity, which can be used to bring closer the two aminodiacetate side-chains in a disposition suitable for metal chelation. The XProGlyX sequence was chosen as a turn inducer between the two Ada_n residues. Indeed, being cyclic, l-proline is unique among amino acids to promote changes in protein backbone direction and to restrict the conformational space accessible to the peptide. In contrast, glycine possesses enough conformational freedom to comply with the backbone dihedral angle requirements of the type II β-turn structure and also provides sufficient conformational elasticity for efficient metal coordination. Finally, even though this sequence is not usually considered as a strong turn inducer in linear peptides, it is frequently encountered in type II β-turns in proteins [95].

The two hexapeptides P²² and P³³ that incorporate two Ada_n chelating amino acids were first designed with a ProGly peptide spacer (Fig. 4 and Table 2). A tryptophan residue was also inserted in the sequence to benefit from the antenna effect to sensitize Tb³⁺ luminescence. The comparison of the Ln³⁺ complexation properties of the two latter peptides evidences a dramatic effect of the Ada_n side-chain length [19]. Indeed, the most flexible peptide P³³ with propyl spacers between the peptide backbone and the chelating aminodiacetate groups forms a mononuclear hexadentate complex of moderate stability, in which the two chelating moieties behave as nearly independent subunits. This also leads to the formation of polymetallic species in excess of metal ion and therefore to uncontrolled speciation of metal complexes. On the contrary, the more compact peptide P²² acts as an efficient hexadentate ligand of Ln³⁺ ions with picomolar affinity in water, which does not give any polymetallic species. The TbP²² complex shows a significant stabilization of two orders of magnitude, in comparison with ligands bearing independent aminodiacetate moieties, even though the two coordinating nitrogen atoms are separated by 14 atoms. This substantial stabilization is assigned to peptide secondary structure elements evidenced in the solution NMR structure of LaP²²: an expected type II β-turn with a well-defined H-bond and an amazing hydrophobic Trp(indole)–Pro interaction. Therefore, P²² demonstrates that the peptide backbone can be used as a non-innocent scaffold with a defined conformation and therefore as an asset to the Ln³⁺ complex stability.

4.2 Higher denticity hexapeptides

Higher denticity Ln-binding peptides were designed in order to increase the complex stability and also to dehydrate the Ln³⁺ ion so as to enhance the luminescence properties by decreasing the vibrational non-radiative deactivations from water molecules O–H. To further enhance the denticity of Ln-binding peptides, we used the shorter side-chain analogue Ada₁ in peptides P¹¹, P¹² and P²¹ (Table 2). Indeed, Ada₁ may act as a tetradentate donor thanks to the coordination of its backbone carbonyl in a six-membered chelate ring. This attempt was successful in decreasing the hydration state of the Ln³⁺ ion (q_Tb = 0 in TbP¹² and q = 0-1 in TbP¹¹) but did not lead to greater stability of the Ln³⁺ complexes (Table 2) [20,22]. Even though the carbonyl function of Ada₁ in position 2 was involved in the metal coordination by the two peptides P¹² and P¹¹, no significant stabilization was observed due to this additional donor (K_d in Table 2). Indeed, supplementary carbonyl binding groups are neutral donors, which do not provide significant stabilization of Ln³⁺ complexes in water, as evidenced before with neutral tripodal N,O ligands [96]. But, interestingly Ada₁ is less basic than Ada₂, thanks to the withdrawing effect of the peptide backbone, which is greater via the short Ada₁ side-chain; this decreases the overall basicity of Ada₁-based peptides and reverses the order of the complexes stabilities at pH 7 (K_d^pH7, Table 2).

To get a major effect on both the Ln³⁺ ion dehydration and the complex stability, the novel chelating amino acid Ed3a₂, with a pentadentate polyaminocarboxylate side-chain and thus an extra-aminoacetate group, was inserted in place of one of the two tridentate residues in P²² [21]. The two hexapeptides P^HD2 and P^HD5 (Table 2) are only differing in the position of the pentadentate residue and are able to bind Ln³⁺ in an octadentate coordination mode. The mononuclear Tb³⁺ complexes demonstrate low femtomolar stability in water, i.e. a gain of four orders of magnitude with respect to the first generation hexapeptide P²². The octadentate coordination of Tb³⁺ by P^HD2 and P^HD5 leads to total dehydration of the metal ion in the mononuclear complexes with long luminescence lifetimes (> 2 ms). This results in an improvement of the luminescence properties with respect to the trisaquo TbP²² complex of a factor 4.

The position of the higher denticity amino acid Ada₁ or Ed3a₂ in the hexapeptide sequence appears to be critical for the control of the metal complex speciation. Indeed, the peptides P^HD2 and P^HD5 or P¹² and P²¹ only differ by the positions of the two chelating amino acids with respect to the Pro-Gly spacer and show different behaviors regarding speciation. Whereas P^HD5 and P²¹ promote the formation of polymetallic species in excess of Ln³⁺, P^HD2 and P¹² form exclusively the mononuclear complex. Hence, it appears that positioning the higher denticity amino acid, either Ada₁ or Ed3a₂ in position 5 after the Pro-Gly spacer favors an independent coordination of the two chelating moieties. Therefore, to control speciation and stabilize the higher coordination mode that involves the two chelating arms of the hexapeptides, the higher denticity amino acid has to be introduced in position 2, prior to the cyclic proline residue, which probably constrains more efficiently the peptide backbone. The position of the two chelating amino acids with respect to the Pro-Gly spacer is thus a key-parameter for the control of Ln³⁺ complexes speciation.

4.3 How does the sequence affect the structure of the complexes?

The influence of the sequence onto the complex's structure could be investigated by solution NMR for two complexes. Indeed, LaP²² and LuP¹¹ showed numerous nuclear Overhauser correlations in their NOESY spectra, which allowed their solution structure's calculation. The difference in the Ln³⁺ coordination mode by P²², which is hexadentate and P¹¹, which is heptadentate, modify deeply the structures of the peptides in their complexes (Fig. 5).

In LaP²², the Ada₂PGAda₂ motif forms a type II β-turn showing a perfect H-bond between the carbonyl of Ada₂2 and the NH group of Ada₂5 and a hydrophobic interaction between the aromatic group of the tryptophan residue and the aliphatic proline ring. The peptide backbone adopts a U-shape structure with an “upper face” devoted to La³⁺ coordination and a “lower face” where the hydrophobic interaction between the Trp indole group and the proline cycle takes place. The La³⁺ ion is located 8.7 Å from the center of the sentizing indole group [19]. These two structural stabilization elements due to the peptide scaffold are responsible for the high stability of LnP²² complexes as mentioned in Section 3.1.

On the contrary, LuP¹¹ revealed an extended S-shape structure due to the extra coordination of the amide carbonyl of Ada₁(2), which prevents the formation of the β-turn and drives the Ln³⁺ ion in the proline plane and closer to the Trp indole sensitizer (5.7 Å) [20].

Although no structure could be calculated for the other complexes, their photophysical data give insights into the backbone arrangements and the metal coordination. For instance, LnP²¹ and LnP²² complexes point to similar structures with a triply hydrated Ln³⁺ ion and a U-shape conformation of the peptide backbone. Indeed, both Tb³⁺ complexes show a metal hydration state of 3 and an increase of Trp fluorescence upon Tb binding, which indicate similar Trp environment in the Ln³⁺ complexes. On the contrary, LnP¹² and LnP¹¹ complexes exhibit dehydrated Ln³⁺ ions, a decrease of Trp fluorescence upon Tb binding and short Eu³⁺ luminescence lifetimes in D₂O, indicative of deactivation by CH groups of the ligand. All these characteristics allow us to conclude to the coordination of the carbonyl group of Ada₁(2) inducing a S-shape conformation of the peptide backbone, as evidenced previously in the NMR structure of LuP¹¹ [22].

Even though P¹¹ and P¹² are heptadentate Ln³⁺ ligands due to the extra-coordination of the backbone carbonyl of Ada₁(2), they do not benefit from the secondary structure elements found in the U-shape structure of LnP²². Therefore, TbP¹¹ and TbP¹² complexes are less stable than TbP²² (K_d in Table 2).

4.4 Role of the spacer between the two chelating amino acids

Finally, the role of the Pro-Gly spacer was studied by designing a high denticity peptide similar to the most efficient compound P^HD2, but with a Gly-Trp spacer, which cannot induce a turn in the peptide backbone (P^HD’ in Table 2). The synthesis and complexing properties of P^HD’ are reported in the supplementary material. A mononuclear complex TbP^HD’ of significantly lower stability than TbP^HD2 is formed. In addition, P^HD’ gives polymetallic complexes in excess of metal. Whereas P^HD2 and P^HD5 are able to promote the total dehydration of Tb³⁺ ions in the octadentate complexes (q = 0), coordination by P^HD’ leave some access of water to the metal center (q = 0.5).

These data evidence the crucial role of the turn inducing effect of the Pro-Gly spacer in the stabilization of the mononuclear complex with simultaneous coordination of the two chelating amino acid side chains and in promoting a perfect Tb³⁺ octadentate coordination with no access of water to the metal center.