One of the most critical points in PSD is the presentation of the target material to the bacteriophages. For affinity selection, the Eu³⁺ ions, intended to serve as target elements, had to be immobilized on nitrilotriacetic acid (NTA), which in turn is coupled to agarose beads (PureCube 100 NTA Agarose, Cube Biotech, Monheim, Germany), forming a stable complex with Eu³⁺. To remove the ethanol, NTA-agarose beads were washed six times with Millipore water and sedimented by centrifugation. The beads were mixed with europium chloride solution (10 mM EuCl₃, 100 mM KCl, pH 6.5) and incubated overnight at 4°C in an overhead shaker (99 rpm). The next day, Eu-NTA agarose beads were washed 10 times with TBST buffer (TRIS-buffered-saline with Tween 20; 50 mM TRIS-HCl, 150 mM NaCl, 0.1% (v/v) Tween 20, pH 7.5), centrifuged, and refilled to the initial volume (10 mL) with the same buffer and served as target for the first biopanning. To investigate the number of immobilized Eu³⁺ ions, three reference samples were taken at 200 μL each (100 μL settling volume) and the metal ions were eluted with 800 μL 1 M nitric acid (HNO₃) each for 1 h at 70°C. The Eu³⁺ concentrations of the samples were then determined by inductively coupled plasma mass spectrometry (ICP-MS).

All biopanning experiments were performed using the commercially available pIII-displayed peptide library Ph.D.-12 (Ph.D.™-12 Phage Display Peptide Library Kit, New England Biolabs GmbH, Frankfurt am Main, Germany (NEB)]. A detailed description of the library can be found in the manufacturer’s instructions.

For the amplification of the M13 phage of the peptide library Ph.D.-12, the bacterial strain Escherichia coli K12 ER2738 (F’proA+B+ lacIq Δ(lacZ)M15 zzf::Tn10(TetR)/fhuA2 glnV Δ(lac-proAB) thi-1 Δ(hsdS-mcrB)5; NEB) served as host. Host cells were used for both titration and amplification of the infectious phage particles.

After each round of selection, the phage particle concentration (phage titer) in the eluate and amplicon was determined. For that, a fresh E. coli culture was first prepared by inoculating LB medium (10 g/L bacto-tryptone, 5 g/L NaCl, 5 g/L yeast extract, pH 7.5) with the host strain and culturing it on a shaker at 37°C to log-phase with an optical density (λ = 600 nm; OD₆₀₀) of 0.5. Appropriate dilution series of each of the phage fractions were prepared in TBST buffer containing 0.5% (v/v) Tween 20 and 10 μL of each dilution were incubated with 200 μL fresh cells each for 5 min at room temperature. The infected E. coli cells (15 µL) were mixed with 200 μL of liquid top agarose (LB medium containing 7 g/L agarose) and placed in one well each of a pre-warmed LB/IPTG/X-Gal 24-well plate (LB medium containing 15 g/L agar, 0.05 mg/mL IPTG (isopropyl-β-D-thiogalactopyranoside) and 0.04 mg/mL X-Gal (5-bromo-4-chloro-3-indoxyl-β-D-galactopyranoside). The plate was incubated at 37°C overnight. By counting the blue appearing plaques, the phage titer could be determined as plaque forming units (pfu).

Depending on the elution protocol, slightly different phage amplification methods were used. For the propagation of phage particles, the entire eluate (minus 10 μL for titration) was added to either 30 mL of LB medium (biological elution protocol) or to 30 mL of a freshly grown E. coli culture (OD₆₀₀ ∼ 0.5; glycine-HCl elution protocol). Phage-infected host cells were incubated for 4.5 h at 37°C and 210 rpm. To separate the phages secreted into the medium from the cells, the culture was centrifuged twice (10,000 × g, 4°C, 10 min). For precipitation of phage particles, the supernatant was mixed with 1/6 of the volume of PEG/NaCl solution (20% (w/v) polyethylene glycol 8,000, 2.5 M NaCl) and incubated at 4°C overnight. The next day, phages were sedimented by centrifugation (10,000 × g, 4°C, 15 min) and resuspended in 1 mL TBS buffer (pH 7.5). In a subsequent purification step, a second precipitation was performed by adding 1/6 of the volume of PEG/NaCl solution to the suspension again and incubating it on ice for 1 hour. Phages were repeatedly spun down (14,000 × g, 4°C, 15 min) and the phage pellet was carefully resuspended in a final volume of 200 μL TBS buffer. Remaining contaminants as well as denatured residues could be removed by a final 1-min centrifugation. The amplicons were stored at 4°C.

To determine possible amplification advantages of individual sequences via NGS, the naïve phage library was also amplified. Two 1 L flasks, each containing 100 mL terrific broth medium (12 g/L bacto-tryptone, 24 g/L yeast extract, 4 g/L glycerol, 8.5 mM KH₂PO₄, 36 mM K₂HPO₄), were inoculated with 1 mL E coli preculture. The cultures were incubated with vigorous shaking (250 rpm) at 37°C to an OD₆₀₀ of around 2.0. The shaking speed was then reduced for five minutes (100 rpm) to allow pili regeneration. This was followed by infection with approximately 10¹² pfu to each batch and a 15 min incubation at 37°C with slow shaking. The two cultures were then divided into six pre-warmed 1 L flasks, each containing 300 mL LB medium, and incubated at 37°C for 35 min with vigorous shaking. Tetracycline (18 μg/mL) was added and 7 µL samples were taken for titer determination, for which they were diluted with E. coli culture and then plated. The six amplification cultures were incubated overnight at 37°C with vigorous shaking before the phage clones were precipitated as described above, combined from the six batches and stored at 4°C.

The identification of novel target binding peptides requires for every target a well-defined biopanning protocol. In this biopanning set-up, the naïve Ph.D.-12 peptide library was incubated with Eu³⁺ as target ions, which were previously immobilized to a NTA agarose matrix (see 2.1). While the complexes of bound phages and target element were retained on the solid phase, the unbound phages remaining in solution could be removed by washing several times. For the following elution step of the bound phages from the target material, we used E. coli cells for biological elution as well as glycine-HCl or glycine-HCl in combination with ultrasound for chemical or chemical-physical elution in several independent biopanning experiments. In a final step, the eluted phages were amplified in the host organism E. coli, titered and used at a concentration of 10¹¹ pfu per mL for the next round of selection. Additionally, the stringency was increased by using a higher Tween 20 concentration in the TBST binding and washing buffer with each round. The process was repeated two to three times, until strong-binding bacteriophages have enriched (Figure 1).

To validate the success of each biopanning experiment discussed in this study individual clones from the last round of selection were analysed by Sanger sequencing. Single plaques of infected host bacteria were picked from titer plates and transferred to 50 μL TBS, allowing phages to diffuse overnight at 4°C. Residual agar and cells were removed by centrifugation (10,000 × g, 10 min, 4°C). Phage particles were stored at 4°C for further characterization. To identify the displayed peptide sequence of individual bacteriophage clones, the phage particle solution was used as the template for a polymerase chain reaction (PCR) with the oligonucleotide primer pair 5′-GCA​ACT​ATC​GGT​ATC​AAG​CT-3′ (forward) and 5′-CCC​TCA​TAG​TTA​GCG​TAA​CG-3′ (reverse). PCR was performed using Taq DNA Polymerase with ThermoPol^® Buffer (New England Biolabs GmbH, Frankfurt am Main, Germany) following the manufacturer’s protocol: initial denaturation (120 s, 95°C), 35 cycles of denaturation (30 s, 95°C), annealing (30 s, 55°C), elongation (45 s, 72°C), and a final elongation (120 s, 72°C). The resulting PCR product was sequenced via Sanger sequencing (Eurofins Genomics, Köln, Germany) using the primer 5′-CCC​TCA​TAG​TTA​GCG​TAA​CG-3′.

To avoid bias in the peptide sequences due to amplification advantages of some phages as well as random selection of phage clones from clone plates, the phage eluates from the last round of selection were analysed using Illumina-NGS. This sequencing method has a higher throughput and thus allows a more holistic view of a large number of strongly bound and eluted phage variants. In preparation for NGS samples, the peptide-coding inserts of the genomes of the eluted phages were first amplified by PCR using the oligonucleotide primer pair 5′ATT​ATT​CGC​AAT​TCC​TTT​AGT​G-3′ (forward) and 5′CTT​TCC​AGA​CGT​TAG​TAA​ATG-3′ (reverse) at a concentration of 0.5 µM each and Phusion Hot Start II High-Fidelity DNA Polymerase (Thermo Fisher Scientific GmbH, Dreieich, Germany) at 0.015 U/µL. The remaining components were used according to the manufacturer’s instructions, with an additional 3% Dimethylsulfoxid (DMSO) added. DNA template (phage eluate) was used at 1.25 µL per 50 µL-reaction volume. For difficult templates, the 5X Phusion HF Buffer was replaced with the 5X Phusion GC Buffer (Thermo Fisher Scientific GmbH, Dreieich, Germany). The following temperature-time program was used: initial denaturation of 180 s at 98°C; 40 cycles of 10 s and 98°C denaturation, 60 s and 60.5°C annealing, and 30 s and 72°C elongation; and a final elongation of 300 s at 72°C. The PCR products were separated by agarose gel electrophoresis (2%), the dominant fragments of the correct size (161 bp) were cut out (see Supplementary Figure S1) and purified using the Wizard SV Gel and PCR Clean-Up System kit (Promega GmbH, Walldorf, Germany). DNA concentration was determined using the Qubit 4 Fluorometer with the Qubit 1x dsDNA HS Assay Kit (Thermo Fisher Scientific GmbH). NGSelect paired-end Illumina sequencing with 2 × 150 bp was performed externally by Eurofins Genomics Germany GmbH, Ebersberg, Germany. The 5 million read pairs per sample in format of FASTQ files were analysed using an in-house script. Known biases in the library (e.g., overrepresentation of wild-type clones, stop codons) were addressed by filtering out sequences that did not match the expected peptide library format. This ensured that only valid peptides were considered for frequency calculation. For this, the size of the fraction, which is either the eluate, original or amplified phage library, was first determined as the total number of reads without read count 1, using this value to calculate proportions. For each peptide sequence, its relative frequency within in the original library was determined using the formula: Relative frequency in original library  Rel .  freq .  in ori .  lib . = number   of   reads   for   a   specific   sequence total   number   of   reads   in   original   library

For sequences detected in the original library, their relative frequency is directly reported as the calculated ratio. For sequences not detected in the original library, the relative frequency is marked as “n.a.” (not available), since it cannot be determined from the data. For the calculation of enrichment factors, the relative frequencies of both the main biopanning elution fraction and the corresponding negative selection fraction were determined, using the formula: Relative frequency in respective fraction = number   of   reads   for   a   specific   sequence total   number   of   reads   in   respective   fraction

Relating these two values to each other gives the enrichment factor for one specific sequence, as shown in the following formula: enrichment factor = relative   frequency   in   main   biopanning   elution   fraction relative   frequency   in   negative   selection   fraction

The amplification advantage factors were calculated in the same way as the enrichment factors, using the ratio of the relative frequencies of amplified to original library instead of eluate to negative selection fraction: amplification advantage factor = relative   frequency   in   amplified   library   relative   frequency   in   original   library

Relative frequencies are expressed as small fractions, e.g., 2.5 × 10⁻⁶, due to the high diversity of the library or elution fraction, respectively. For this example, it means that for every 1,000,000 reads, approximately 2.5 reads correspond to that specific peptide sequence.

In order to analyse the affinity of the selected peptides towards Eu³⁺, each peptide was titrated into a solution of EuCl₃ and excited using a laser, the emitted light of which was recorded in 21 time-shifted spectra. For each peptide, three titration series (=triplicate) consisting of 20 titration steps with increasing peptide concentration were measured.

Peptides (Tables 1, 2) as well as CaM-EF4 (DIDGDQVNYEE) that served as reference peptide were synthesized via solid-phase peptide synthesis (SPPS) and obtained from DGpeptides (Co., Ltd.; Hangzhou, CN) as lyophilized TFA salts with a purity of ≥85%. To prepare peptide stock solutions, 4–6 mg of lyophilized peptide powder was dissolved in an appropriate volume of ultrapure water (Merck KGaA, Darmstadt, Germany) to achieve a stock concentration of 5 mM. The peptide concentration was measured using a NanoDrop spectrophotometer (Life Technologies GmbH, Darmstadt, Germany). Potassium chloride was added to a final concentration of 100 mM, and the pH was adjusted to 5.2 using NaOH. The stock solution was then diluted to a final peptide concentration of 2.5 mM.

For each titration series, 500 µL of a 10 μM EuCl₃ solution (in 100 mM KCl, pH 5.2) was placed in a quartz glass cuvette. A 2.5 mM peptide stock solution (in 100 mM KCl, pH 5.2) including 10 μM EuCl₃ (to avoid dilution effects) was titrated to the EuCl₃ solution with a dynamic increase in concentration (measurement 1 was without any peptide, see Supplementary Table S1).

For TRLFS, a Nd:YAG laser-pumped optical parametric oscillator (OPO, NT230, Ekspla) with ∼5 ns pulse and a wavelength of 394 nm was used to excite the sample in a quartz glass cuvette. An optical fibre was used to transmit the emitted light to the spectrometer (SR-303i-A, Andor Technology). A grating with 300 lines per mm generated monochromatic light, which was detected with a cooled (−30°C) ICCD camera (Andor iStar, DH320T-18U-63). The initial delay was set to 5 μs, gate width to 200 µs, slit width to 200 µm and gain to 4,095. Eight accumulations were gathered. Baseline correction was performed by Andor Solis software (Andor Technology). The step size for 21 spectra was calculated from the formula 3e⁺⁶+(3e^+6x) μs. A Peltier element (TC1 temperature controller, Quantum Northwest) controlled the sample temperature of 25°C and the mixing of the sample (1,200 rpm) using a magnetic stir bar. The three-dimensional fluorescence data obtained were analysed using PARAllel FACtor analysis (PARAFAC).

The goal of the biopanning experiments described was to identify peptide sequence motifs with an affinity for Eu³⁺ ions. This was achieved by screening the commercially available Ph.D.-12 Phage Display Peptide Library (NEB) displaying randomized dodecamer peptides on the pIII phage minor coat protein. In order to retain the Eu³⁺ binding phages, it was necessary to immobilize our target Eu³⁺ on a solid matrix such as NTA agarose beads. The Eu³⁺ content of the NTA agarose was determined by ICP-MS after elution with 1 M HNO₃ in three reference samples. This resulted in a loading of 120 μg Eu³⁺/100 μL bead volume. Thus, 240 μg Eu³⁺ were available as target material for each biopanning round, in which approximately 200 μL beads were provided.

Several independent biopannings were carried out using different elution methods. In the first panning strategy (Figure 1), the phage library was only incubated with the Eu-NTA agarose beads. In contrast, in the second panning strategy, the phage library was incubated with the Eu-NTA agarose beads as target material (main biopanning) and with the NTA agarose beads as negative control (negative selection panning) in parallel. The negative selection panning involves the binding of phages to plastic and/or the immobilisation material (NTA agarose beads), so that the phage library is cleared of non-specific binders (Nian et al., 2010; New England Biolabs, 2024). However, a negative selection step before target binding was deliberately avoided in order to prevent the removal of potentially good binders and thus premature depletion of the library. Instead, a parallel approach was used to screen for target-unspecific peptide variants (TUP), which were then compared with the peptide sequences from the main biopanning and eliminated. In a third panning strategy (Figure 2), the order of the classic biopanning process was changed slightly. In a first step, the bacteriophage library was incubated with the EuCl₃ solution before the Eu³⁺-phage complexes were immobilized on the matrix. The amplified eluate was divided into two fractions, one of which was subjected to a negative selection step against unloaded NTA agarose beads before being used for the next round. The other fraction was used directly as input for the next round of selection.

NGS was applied to identify the peptide inserts of the bacteriophages contained in the eluate of the final selection round. An in-house bioinformatics script was used to process the raw sequencing data (FASTQ files) so that enriched peptide sequences and their absolute abundances could be extracted. These absolute frequencies were set in relation to the reads per fraction in order to derive the relative abundances, which finally allow a comparison of the sequences with each other. The enrichment factor, which assesses whether a peptide variant is target-specific or interacts non-specifically with the NTA agarose beads, was determined by comparing the relative abundances in the elution fraction to those in the corresponding negative control. This negative control could either involve a parallel negative selection panning (ratio main biopanning/negative selection panning) or no subsequent negative selection step after elution (ratio with/without subsequent negative selection step), depending on the experimental setup (see visualization in Supplementary Figures S5, S6). To evaluate whether a sequence motif accumulated due to its affinity for the target or was influenced by amplification advantages during the selection rounds, the ratio of the relative abundances in the amplified library to those in the original library was calculated. Ratios greater than one indicate an amplification advantage, which should be considered when interpreting the enrichment of specific peptide sequences.

Since the parallel biopanning against both Eu³⁺-loaded and unloaded NTA agarose beads as well as the modified biopanning with and without a negative selection step on the amplified eluate were each performed with the same phage library (same lot number) but completely independently and with different elution types, the identified peptide sequences were screened for repeated occurrence in several different eluate fractions.

For the synthesis and affinity analysis of the peptides without phage construct, a small selection of interesting peptides was chosen from the huge amount of enriched and NGS-sequenced peptides (Table 1). The selection criteria included: (Schüler et al., 2011) strong enrichment in multiple independent biopannings or eluate fractions, as determined by sequencing data (Adler and Müller, 2014); consistent amplification across rounds, indicating robust phage growth (≤1); and (Fricker, 2006) minimal non-specific binding to plastic surfaces (≤0.5), calculated with SAROTUP (HLAB Co, 2025). The peptides identified by Sanger sequencing and selected for further analysis did not undergo the same selection criteria. While the high data density of NGS data allows selection based on specific criteria, the limited clone size in Sanger sequencing results is insufficient for such an approach. Some of the NGS-sequenced peptide variants are quite similar in their amino acid sequence, such as the two possible progeny sequences of VHHVAGNALQPF: VHHVAGNVLQPF and VHHVAGNALQPS (the latter two are not shown in Table 1 due to space constraints; only synthesized and tested sequences are included). The original sequence VHHVAGNALQPF shows an increased enrichment inthree independent eluates [Table 1, columns (C), (D1) and (D2)] compared to the respective negative control, either directly in parallel or as in the modified panning afterwards. In ultrasound elution it is even the most enriched peptide variant. Many of the other identified peptide sequences are rich in basic and neutral polar amino acids like serine, asparagine or glutamine and often contain at least one proline and/or acidic amino acid. Also notable are short motifs found in more than one sequence, such as Sx_1-2HS (where x is any amino acid) and VxHV. Moreover, the sequence SIHSVTKGRYPV (GC3) was already found in a previous biopanning against arsenic oxyanions (Braun et al., 2020). Sequence SIHSVTKGWYPV (RB-Ar-8), also enriched against arsenic oxyanions (Braun et al., 2020), differs from GC3 in only one amino acid at position 9 (tryptophan instead of arginine), which is why this peptide was also investigated in more detail. Identified and promising peptides shown in Table 1 were synthesized as linear peptides in a solid phase synthesis and dissolved in distilled water. Peptide GC2 was not soluble in water but only in 20% acetic acid and could therefore not be tested, as acetic acid itself interacts with Eu³⁺.

Since Eu³⁺ is a fluorescent element, the binding affinity of the selected peptides towards Eu³⁺ was assessed by TRLFS. For this, changes in the Eu spectrum, in particular in the ⁷F₁/⁷F₂ peak intensity ratio, were determined. A titration series with a peptide excess of up to 100-fold relative to Eu³⁺ was used to detect even weak affinities. The sequences of GC1 and GC3 were subsequently modified at specific positions to improve binding affinity. These modifications were based on two criteria: the peptides’ high enrichment levels during biopanning and insights from initial binding tests (data not shown), which indicated potential interactions. By mimicking natural occurring Ln-binding molecules, the modifications targeted specific acidic amino acid residues believed to enhance Eu³⁺ binding by facilitating ionic interactions and taking into account structural considerations as discussed before. The goal was to form binding pockets and/or to replace alanine or valine, which were unexpected to interact with Eu³⁺. Derivates of the peptides were synthesized to assess the role of specific amino acids in Eu³⁺ binding (Table 2). The original peptides GC1 and GC3 as well as their modified versions were synthesized with a GGGS linker in order to have the possibility to easier immobilize them on a matrix for adsorption experiments in case of good affinity. We tested GC1 and GC3 with and without the GGGS linker but found no significant improvement or reduction in binding affinity (data not shown).

For eight of the twelve peptides tested (GC1, GC1-A/D, GC1+D + E, GC3, GC3+D + E, GC4, GC8, GC9), changes were observed in the emission spectrum of the Eu³⁺ aquo ion. Specifically, the intensity of the ⁷F₂ peak increased relative to the ⁷F₁ peak, which is characteristic of Eu³⁺ binding to the peptides and indicative of complexation (Figure 3).

Peptide GC1 and its modified variants demonstrate how the substitution of specific residues with acidic amino acids can significantly enhance binding affinity. The originally selected GC1 showed very low binding, characterized by the only slightly increased intensity of the ⁷F₂ peak compared to that in the spectrum of the Eu³⁺ aquo ion. The modified variant GC1-A/D, in which the alanines were replaced by aspartic acids, already showed improved Eu³⁺ binding. In the variant with additional N- and C-terminal aspartic and glutamic acids even a reversal of the peak ratio could be observed. When examining the distribution of different species in the system with increasing peptide concentration, only a small decrease in the intensity of the free Eu³⁺ ion and a slight increase in the 1:1 complex can be observed for GC1. For the two modified variants of GC1, in contrast, a strong decrease or increase of the corresponding species can be observed, indicating an enhanced complexation of Eu³⁺.

This can only be applied to a limited extent to the GC3 peptide and its modified versions. Although GC3 exhibited a higher affinity for Eu³⁺ than GC1, replacing the valine at position five with aspartic acid (GC3-V5/D) did not improve binding. On the contrary, it resulted in an affinity reduction. The addition of acidic amino acids in GC3+D+E led to a higher Eu³⁺ complexation, but not as high as in GC1+D+E.

Among the unmodified peptides enriched in the biopanning experiments, GC9 was the one with the highest affinity for Eu³⁺. Its t₀ spectra and species distribution diagram are similar to those of GC3+D+E, although it has one less acidic residue than the modified variant of GC3.

Peptide RB-Ar-8, already selected in a previous biopanning against arsenic oxyanions (Braun et al., 2020) and differing from GC3 only in one amino acid (tryptophan instead of arginine), showed no change in the spectrum and thus no binding to Eu³⁺, suggesting that the arginine at position nine in peptide GC3 plays an essential role in the binding. No binding affinity was found for GC5 and GC7 either.

For the biopanning selection process, bacteriophages that bind strongly to the target must be separated from weak or non-binding ones. While insoluble targets like inorganic particles can be directly used, soluble targets such as ions require reliable immobilization on a matrix. This ensures that strongly binding phages or phage-target complexes remain on the solid phase, while weak and non-specific binders are removed through repeated washing. Finally, the most strongly bound phages are eluted from the matrix. The challenge here is not to immobilize a sufficiently high number of ions. With 240 μg Eu³⁺ or approximately 9.5 × 10¹⁷ Eu³⁺ available per selection round to bind 1 × 10¹¹ phage particles, there was a clear excess of Eu³⁺, so that there should have been sufficient target ions for each phage clone. However, during metal ion immobilization, some of the metal binding sites are already occupied. In the case of Eu³⁺ and NTA theoretically four of the nine binding sites are occupied and thus not available for potential interaction with peptides. In addition, target immobilization in combination with affinity selection such as biopanning carries the risk that potential superbinders strip the Eu³⁺ from the NTA and are thus lost in the washing step. All these reasons justify a well-planned biopanning for Eu³⁺ as well as the combination of various protocols with respect to the target presentation, interaction environment as well as the type of binding.

For the enrichment and identification of Eu³⁺ binding phage clones in this preliminary study, a phage peptide library was screened that displays 12 amino acid long randomized peptides on the five pIII minor coat proteins of the M13 bacteriophage with a diversity of about 10⁹ single sequences. Considering our immobilized target, a pIII phage library seemed to be a better choice than a pVIII phage library. While pVIII phages display up to 270 peptide copies along the entire 1 µm length of the filamentous M13 phage (Kay and Hoess, 1996), this high density can lead to target clogging, making it difficult for peptides to effectively bind the immobilized Eu³⁺ ions. In contrast, the pIII phages display only five peptide copies at the proximal end, allowing more precise interactions with the target and reducing competition among phage variants. Although pVIII has more peptide copies and thus theoretically a multiple binding strength, the successful selection of phages with only five copies in the pIII system suggests strong and genuine target binding, making a pIII phage library more suitable for this study’s goal of identifying high-affinity Eu³⁺ binding peptides.

Furthermore, we refrained from blocking the target material with bovine serum albumin (BSA) prior to panning, which is often done in case of other targets (e.g., organic targets, insoluble targets) to avoid unspecific interactions with the target, as we wanted to prevent complete masking of the target and thus a lack of interaction with the phages. The selection pressure on the phages was increased with each round by rising the concentration of Tween 20 in the TBST buffer. As a surfactant, Tween 20 reduces the interfacial tension between two phases, which is why it becomes more difficult for the phages to attach to the target and to the resin and plastic surface as its concentration increases. As a result, only the strong-binding phages are still able to bind to the target material. In addition, negative selection prior to the binding step was avoided due to the risk of premature depletion of the library and thus premature exclusion of potentially strong-binding peptide variants. Instead, such a negative selection against the unloaded NTA agarose beads to remove TUPs was performed either in parallel or after the main biopanning following the propagation of the eluted phages.

For sequence identification of the peptide inserts of a large number (five million read pairs) of bacteriophages contained in the eluate fractions of the last selection round, we decided to use NGS in addition to Sanger sequencing (see Supplementary Tables S3, S4). NGS gives us a more holistic view of all strongly bound and eluted phage variants compared to Sanger sequencing of randomly isolated single phage clones. By comparing the relative sequence abundances of main biopanning and negative selection panning fractions, as well as of the naïve and amplified phage library, enrichment factors or amplification advantages of individual sequences can be calculated, and TUPs can be eliminated. The ratio of the amplified to naïve bacteriophage library should always be considered as an amplification advantage factor when evaluating frequent peptide sequences. Since certain phage clones have a propagation advantage and others a disadvantage and are therefore more or less strongly represented in the input phage solution for the next selection round, the amplification steps between rounds sometimes favour peptides that are not necessarily the best target binders (Derda et al., 2011). Phage clones with peptide sequences that have a growth advantage [propagation-advantaged phages (PAPs)] will always displace phage clones with peptide sequences that amplify more slowly, although both may have equally good binding properties to the target material (Rodi et al., 2002). The reasons for such an advantage or disadvantage are manifold and can occur in any amplification step. On the one hand, some peptide structures are obstructive for the infection of the host with the phage or for the packaging of the fusion protein (Rodi et al., 1999), so that this phage variant is less strongly amplified and other phage clones therefore occur more frequently in the amplificate. However, other metabolic and structural effects are also a limiting factor, such as the fact that each of the 20 amino acids is encoded by a different number of codons (Cwirla et al., 1990) or that advantageous mutations may be present in the regulatory region of the phage genes (Brammer et al., 2008). Derda et al. (2011) reported that about 70% of phage clones can be eliminated in a single amplification step. A biopanning with many rounds of selection is therefore not necessarily the best solution for selecting strong binders, as the sequence diversity is reduced enormously. The calculated enrichment factor, as well as the amplification advantage factor and the associated reduction of TUPs and PAPs, considerably increase the theoretical probability of finding peptide variants with a real affinity to the target.

From the large data set of NGS-analysed enriched sequences, some interesting ones were selected and characterized with regard to their amino acid composition. For this, besides a strong enrichment factor in several different elution fractions, we took also into account amplification advantages (PAPs) as well as the probability of binding plastic and thus being a non-specific binder (TUPs). As mentioned previously, we identified the highly enriched sequence VHHVAGNALQPF, along with two very similar sequences: VHHVAGNVLQPF and VHHVAGNALQPS. It remains unclear whether these differences represent true sequence variants contributing to a conserved motif important for target binding or are simply the result of sequencing or PCR errors during the experiment. If these sequences are genuine variants, they may form a cluster of related motifs with enhanced binding potential. Alternatively, if the variations are due to errors, the reads from these sequences should be attributed to the original sequence, increasing its overall enrichment factor. Further analysis would be required to distinguish between these possibilities. Many of the peptide sequences that are most enriched in multiple independent biopannings have numerous basic as well as neutral polar amino acids like serine, asparagine and glutamine and at least one proline and/or one acidic residue. The sequences richest in acidic amino acids are HDPRMEHSLPKS and EALTVNIKREME. Short motifs that occurred in more than one sequence were identified as Sx_1-2HS (SIHSVTKGRYPV, TLAHHPVSISHS, ARSLEPAPSRHS) and VxHV (VHHVAGNALQPF, VAHVDGTPRLRN), where x represents any amino acid. Since the motif-bearing sequence SIHSVTKGRYPV (GC3), as well as SIHSVTKGWYPV (RB-Ar-8), which differs from GC3 in only one amino acid (W instead of R), could already be enriched against arsenic oxyanions in a previous biopanning (Braun et al., 2020), it was assumed that the motif SxHS contribute substantially to metal binding. However, this hypothesis could not be proven in this study. Our experiments showed a moderate binding of GC3, but no binding for RB-Ar-8, indicating that the SxHS motif that both sequences have in common is not responsible for Eu³⁺ binding. This suggests other factors, such as sequence context or peptide conformation, play a larger role. While acidic amino acids with their carboxyl groups play an important role in the interaction of Ln³⁺ ions (Hosokawa et al., 2022; Legendziewicz et al., 1980), basic amino acids, in particular histidine, but also lysine, are better known for their affinity to transition metals like nickel (Hochuli et al., 1988) or copper (Curtis et al., 2017), and interact with the amino groups of their side chain. The presence of basic amino acids was, therefore, unexpected, but can be justified by the presence of the NTA group in the resin. The bond between the carboxyl groups of NTA and Eu³⁺ ions is nitrogen-stabilized by the side chain amino group of the basic residues, which could explain why peptides containing basic amino acids have been enriched during the panning.

Generally, the experimental setup could serve the purpose to identify peptides for any metal ion target. In other studies (transition) metal ions such as Ni²⁺ (Pagano et al., 2015a; Kay and Hoess, 1996), Zn²⁺ (Derda et al., 2011), or Ga³⁺ (Gonzalez et al., 2014) were immobilized on solid matrices, and biopanning experiments demonstrated good sequence enrichment and strong metal affinities for the selected peptides. However, the ion radius and available binding sites of a metal ion effect the immobilization mechanism. The interaction of transition metal ions with NTA or IDA differs from that of Eu³⁺ ions. The high affinity of the carboxyl groups in NTA for Ln³⁺ ions means that Eu³⁺ may bind more tightly to NTA compared to Ni²⁺ or Zn²⁺, assuming that more binding sites are occupied. Following this theory, the enriched phages or peptides are then Eu³⁺-NTA-adapted variants, which could potentially act as moderate to weak binders for free Eu³⁺, as they utilize only a portion of the available binding sites for the interaction. This highlights the importance of considering how the immobilization method influences target presentation, as it can significantly affect the outcome of biopanning experiments, particularly when using NTA- or IDA-based matrices.

Nevertheless, we succeeded in enriching interesting sequences and since, as already mentioned, we wanted to use a randomized approach with regard to the amino acid composition, the selected sequences (Table 1) were synthesized and investigated for their affinity.

TRLFS was used to evaluate Eu³⁺ binding affinities by analysing changes in the emission spectrum, particularly the ⁷F₁/⁷F₂ peak intensity ratio. A 100-fold peptide excess ensured detection of weak complexes and enabled determination of europium species and affinity constants. Affinity to Eu³⁺ was detected in 8 of 12 tested peptides, highlighting the strength and efficiency of our PSD approach combined with NGS. From the original peptides selected by biopanning, GC9 (EALTVNIKREME) was found to have the highest affinity for Eu³⁺ with a K_D of 1.05 × 10⁻³ M (see Supplementary Table S2). Although with three acidic amino acids it has one less than GC3+D+E, the t₀ spectra and species distribution diagrams look similar for both peptides (Figure 3). However, as previously observed for the CaM-EF4 (DIDGDQVNYEE), acidic amino acids are not always involved in target binding (Mulqueen et al., 1985). This seems to be dependent also on the three-dimensional arrangement and distance between the acidic residues. Both the reversed sequence of GC9 (EMERKINVTLAE) and CaM-EF4 have acidic amino acids at positions 1, 3, and 12. Additionally, both peptides have a valine at position 8, and a possible oxygen donor at position 7 (asparagine in GC9 and glutamine in CaM-EF4). Polar amino acids, which could also act as oxygen donors, are present at position 9 (threonine in GC9 and asparagine in CaM-EF4).

Both peptides contain nearly the same amount of hydrophobic amino acids (GC9 = five, CaM-EF4 = four), which are distributed across the sequence. While isoleucine, glycine, valine and tyrosine occur at a distance of one amino acid in CaM-EF4, the hydrophobic residues alanine, leucine, valine, isoleucine and methionine are not as evenly distributed in GC9. Due to the short length of only 12 amino acids, the peptides should not be able to fold in a manner that the binding site consists of a hydrophilic smaller shell embedded in a hydrophobic larger shell, as is the case with metal-protein binding patterns (Yamashita et al., 1990). However, a folding in which the hydrophobic residues point outwards to mimic such an effect cannot be ruled out and should be further analysed by circular dichroism.

For peptide GC1 (VHHVAGNALQPF), it could be clearly shown that the replacement of amino acids by acidic amino acids (GC1-A/D) or the incorporation of additional carboxyl group-providing amino acids (GC1+D + E, DVHHVDGNDLQPFEGGGS) increases the binding strength, reducing the K_D by two to three orders of magnitude (4.03 × 10⁻¹ M > 1.11 × 10⁻³ M > 4.89 × 10⁻⁴ M, respectively). In contrast, this only applied in part to the peptide GC3 (SIHSVTKGRYPV) and its modified variants. Although GC3 already had a higher affinity (K_D = 1.47 × 10⁻²) than GC1 and the replacement of the residue at position five with a negatively charged amino acid was expected to result in a higher affinity, the opposite was observed. The binding could not be improved by replacing valine with aspartic acid (GC3-V5/D), suggesting that valine, as a hydrophobic amino acid at this position, probably plays an important role in the folding of GC3. The two serines in GC3 could also provide oxygen atoms for binding and could be one reason for the higher affinity of GC3 compared to GC1. For GC3+D+E (DSIHSDVTKDGRYPVEGGGS) with additionally inserted acidic amino acids, a high affinity with a K_D of 1.54 × 10⁻³ M could again be measured, but here the valine was kept and only an aspartic acid was inserted next to it, so that it retained its function at this position. It is also interesting that the peptide RB-Ar-8 (SIHSVTKGWYPV), which like GC3 was already identified as a possible binder in an earlier biopanning against the arsenic oxyanions and differs from GC3 only at position nine by a tryptophan instead of arginine, showed no Eu³⁺ complexation. Therefore, we infer that the basic amino acid arginine in GC3 plays a key role in binding. It can only be speculated whether tryptophan as a hydrophobic amino acid in the case of RB-Ar-8 prevents the peptide from wrapping around the Eu³⁺ ion or whether the arginine in the case of GC3 stabilises the formed complex together with the other basic amino acids histidine at position three and lysine at position seven.

In summary, eight of the 12 selected peptides showed a complexation of Eu³⁺, although in some cases with low affinity. Although none of the peptides found in the PSD approached the Eu³⁺ affinity of our reference peptide CaM-EF4, we have succeeded in identifying novel, previously unknown Eu³⁺-binding peptide sequences. These sequences and methods cannot yet exploit their full potential, as the synthetic peptides probably differ from the pIII-fused peptides in their folding. The peptides displayed by the bacteriophage are likely stabilized by the phage surface, which could improve their interaction with the environment. Thus, it may be possible that the peptides bound to the phage have a strong binding to Eu³⁺, while the synthetic, free peptides have little or no affinity–not because of the sequence itself, but because of the lack of stabilization. To overcome this, circularization of the peptide sequences could be considered. The addition of an N- and C-terminal cysteine forming a disulfide bridge can stabilize the peptide as well (Barkan et al., 2016) and thus improve the interaction with the target material (McLafferty et al., 1993). Other reasons for the reduced affinity of the peptides found compared to CaM-EF4 could lie in the way Eu³⁺ ions are immobilized, as already discussed in 4.2, and in the naïve Ph.D.-12 phage display peptide library itself. As Sloth et al. (2022) have shown, substantial deviations from randomness occur at the nucleotide, amino acid, and peptide levels in the Ph.D.-12 library. Additionally, the presence of stop codons and a high abundance of wild-type clones reduce the effective size of the naïve library. This suggests that the library may lack optimal Eu³⁺ binding motifs altogether, highlighting the need for selecting or designing a library better suited to the target.

Furthermore, an important factor to consider is the multivalency of phage-displayed peptides. In phage display, each phage presents five copies of the same peptide in a highly specific orientation, which can significantly enhance binding through cooperative interactions. In contrast, synthesized peptides are monovalent, which may lead to differences in binding affinity and behavior. This multivalency effect could explain discrepancies between phage-displayed and synthesized peptides in binding assays.

An application of the peptides for complex mining waters has not yet been tested in this study. It can be assumed that the affinity of the identified peptides is not currently sufficient for use in the recovery of REE from mining waters. For such an application, however, not only the affinity but in particular the selectivity is crucial. Furthermore, a high affinity is not necessarily an advantage for the regeneration of the peptides. The extent to which the peptides are suitable for real applications or how they can be improved will be the subject of future studies. Goal for our following research is the investigation of the identified peptides for their selectivity for Eu³⁺ in contrast to other Ln. Those peptides will be applied in biohybrid filter modules for the selective recovery of Eu³⁺ and, with the help of novel selected peptides, also of other Ln from mining wastewater.