Construction and Validation of a New Naïve Sequestrin Library for Directed Evolution of Binders against Aggregation-Prone Peptides

Affibody molecules are small affinity proteins that have excellent properties for many different applications, ranging from biotechnology to diagnostics and therapy. The relatively flat binding surface is typically resulting in high affinity and specificity when developing binding reagents for globular target proteins. For smaller unstructured peptides, the paratope of affibody molecules makes it more challenging to achieve a sufficiently large binding surface for high-affinity interactions. Here, we describe the development of a new type of protein scaffold based on a dimeric form of affibodies with a secondary structure content and mode of binding that is distinct from conventional affibody molecules. The interaction is characterized by encapsulation of the target peptide in a tunnel-like cavity upon binding. The new scaffold was used for construction of a high-complexity phage-displayed library and selections from the library against the amyloid beta peptide resulted in identification of high-affinity binders that effectively inhibited amyloid aggregation.


Introduction
Monoclonal antibody drugs have been transformative for the treatment of many different diseases during the past decades, most notably in the areas of autoimmune disorders and oncology. Although the real breakthrough in neurodegenerative disorders is still yet to be demonstrated, the recent results from the pivotal Clarity AD Phase 3 study of lecanemab is encouraging [1].
In parallel to the continued development and refinement of antibodies as drugs, research has also been focused on exploring non-antibody-based affinity protein as biopharmaceuticals [2,3]. Affibody molecules is a class of small (58 amino acids) affinity proteins with a three-helical structure that has been evaluated relatively extensively for in vivo diagnostics in oncology and as therapeutic drugs in both oncology and autoimmune diseases. The most advanced affibody molecule is currently tested in clinical phase III trials for treatment of different IL17A-driven autoimmune conditions and the results so far demonstrate excellent efficacy and safety [4].
In a previous effort to develop affibody molecules for amyloid beta (Aβ), involved in the pathogenesis of Alzheimer's disease (AD), phage display technology was used for selection of binders for the peptide. Screening of hits from the selection showed that the effort was successful, and analysis of the top candidate (denoted Z Aβ3 ) revealed that the directed evolution had resulted in a new type of binder that was distinct from the original affibody protein [5]. Although the genetic sequence was similar, the threedimensional structure and mode of binding differed quite dramatically from other reported affibody molecules. Most notably, the binder was shown to fold as a disulphide-stabilized homodimer formed by an evolved cysteine at a specific position in each subunit. Moreover, enrichment of helix-destabilizing prolines and glycines in the N-terminal region of the

Scaffold Design and Functionality of Sequestrin Library
The sequestrin (Sq) library scaffold was designed to comprise (1) a disulphide bond between the two subunits, (2) an eleven amino acid truncation of the unstructured N-terminus, as it has been shown to increase binding to the amyloid beta (Aβ) peptide [10,14,16], as well as (3) a flexible (S 4 G) 2 linker between the two subunits to allow for correct folding with minimal steric hindrance (Figure 1).
The positions that were targeted for randomization constitute the binding surface of previous dimeric affibody molecules in complex with aggregation-prone peptides (Figure 1). In total, 22 positions (eleven in each subunit) were subjected to diversification. In principle, each randomized position was designed to contain the codon for the wildtype amino acid spiked with a mix of codons for a pre-determined proportion of the remaining amino acids. In 18 of the diversified positions (nine in each subunit), an average proportion of 61% of the original codon was combined with a mix of approximately 39% of mutation codons yielding on average seven mutations in the scaffold. Diversification of residues that contribute to the hydrophobic core of the complex was restricted to codons for hydrophobic amino acids in order to preserve the hydrophobic interface of the subunits. Residues that were subjected to randomization in the beta strand area were allowed to vary between all remaining amino acids except for cysteine, and thus included beta-strand stabilizing proline and glycine. The remaining diversified residues were in principle allowed to vary between any amino acid except for proline, glycine, or cysteine. In addition to these 18 randomized positions, four residues (two in each subunit) were subjected to partial randomization and allowed to vary between the original amino acid and the codon for one other amino acid to a 50:50 proportion ( Figure 1D). allowed to vary between any amino acid except for proline, glycine, or cysteine. In addition to these 18 randomized positions, four residues (two in each subunit) were subjected to partial randomization and allowed to vary between the original amino acid and the codon for one other amino acid to a 50:50 proportion ( Figure 1D). The two library subunits were synthesized as separate DNA oligos and amplified to contain approximately 60 overlapping bases for hybridization into one long assembled library oligo. The synthesis of the DNA was performed using a semi-conductor-based method, which provided complete freedom when selecting codons for the library and allowed randomizations with minimal biases compared to degenerate codons and errorprone PCR based approaches. The design resulted in a theoretical library size of 1.27 × 10 22 individual candidates.
The library oligo was subcloned to a modified version of the phagemid vector pAffi1 [5] as N-terminal fusion to an albumin-binding domain (ABD) ( Figure S1A) [17]. The resulting library (Sqlib) yielded 5 × 10 9 individual clones as assessed by serial dilutions of transformed E. coli and had 72% sequestrins (dimers) and 24% monomers as assessed by PCR. However, a decrease in sequestrins (64%) and monomers (16%) in favour for dummy sequences (increased from 4% to 20%) was observed after phage amplification, possibly due to growth bias. To maintain a high proportion of sequestrins, the cultures were started at a higher OD600 (0.2 AU) than what is used in phage display selections of affibody molecules, resulting in fewer cell division cycles in the log-phase and higher proportion of phage displaying sequestrins compared to dummy phage. Moreover, likely due to the increased complexity of the sequestrin (compared to affibody molecules), a longer cultivation time was needed to reach sufficient optical density before superinfection The amino acid distribution in the library was verified by DNA sequencing (Figures 2 and  S2A,B). About 26% of the sequestrin sequences did not contain mutations other than from the design.
To assess the functionality of the produced phage expressing the Sqlib, a monoclonal sandwich ELISA was performed. Here, the C-terminal ABD [17] that is co-expressed with  I  II  III IV  IV   I  II  III IV  IV   23   12-13  16 19-20 The two library subunits were synthesized as separate DNA oligos and amplified to contain approximately 60 overlapping bases for hybridization into one long assembled library oligo. The synthesis of the DNA was performed using a semi-conductor-based method, which provided complete freedom when selecting codons for the library and allowed randomizations with minimal biases compared to degenerate codons and errorprone PCR based approaches. The design resulted in a theoretical library size of 1.27 × 10 22 individual candidates.
The library oligo was subcloned to a modified version of the phagemid vector pAffi1 [5] as N-terminal fusion to an albumin-binding domain (ABD) ( Figure S1A) [17]. The resulting library (Sq lib ) yielded 5 × 10 9 individual clones as assessed by serial dilutions of transformed E. coli and had 72% sequestrins (dimers) and 24% monomers as assessed by PCR. However, a decrease in sequestrins (64%) and monomers (16%) in favour for dummy sequences (increased from 4% to 20%) was observed after phage amplification, possibly due to growth bias. To maintain a high proportion of sequestrins, the cultures were started at a higher OD 600 (0.2 AU) than what is used in phage display selections of affibody molecules, resulting in fewer cell division cycles in the log-phase and higher proportion of phage displaying sequestrins compared to dummy phage. Moreover, likely due to the increased complexity of the sequestrin (compared to affibody molecules), a longer cultivation time was needed to reach sufficient optical density before superinfection The amino acid distribution in the library was verified by DNA sequencing (Figures 2 and S2A,B). About 26% of the sequestrin sequences did not contain mutations other than from the design. the sequestrins ( Figure S1A) on the phage allowed for detection of correctly expressed construct by assessing binding of HSA in an ELISA. Correlation between a correct sequestrin sequence and an ELISA positive signal was observed.

Phage Selection of Sequestrins towards Amyloid Beta Peptide
After verifying the compatibility of the phage display system and sequestrins, the functionality and library design was evaluated by preforming a selection towards the Aβ1-40 peptide ( Figure S1). The selection process was monitored by titrations of phage inputs and outputs for each cycle, showing enrichment of full sequestrin expressing clones and increase in phage titres with increased number of cycles in the selection. In the third selection cycle, an enrichment was seen in which 50% of the input phage were recovered after extensive washing. In this cycle, only full-length sequestrins were found in PCR screening of the phage eluate, indicating a successful selection.
Enrichment of target-binding clones during the selection was assessed by polyclonal phage ELISA. The phage pool showed an increased signal towards Aβ1-40 peptide with increasing selection rounds, correlating to phage enrichment in selection. Moreover, a shift in sequence distribution from the naïve library could be observed ( Figure S2), supporting the results from titrations and the polyclonal ELISA. The clones from the output had an average of four mutations per sequestrin, compared to seven in the naïve library. Mutations were found predominantly located in the second helix in each subunit. Thirteen sequestrins among the most reoccurring clones were selected for further characterization (Table S1).

Expression and Cloning of Sequestrins
The thirteen clones from the selection towards the Aβ1-40 peptide were subcloned into a pET26b(+)-vector with a C-terminal hexa-histidine tag (further denoted as SqAβ-His6) for expression in E. coli (Table S1). The clones obtained from the fourth and fifth selection round (clones 10-23) showed somewhat improved yields during production compared to the clones from the third round. Protein purity and size was analysed by gel electrophoresis ( Figure 3). A single band was observed at 14.4 kDa corresponding with the expected weight of approximately 13 kDa for all sequestrins (Table S2). The yield for SqAβ8 was very low, seen as a faint band of the expected size in the Sodium Dodecyl Sulphate-Polyacrylamide Gel (SDS-PAGE). The mass of the proteins was analysed using mass spectrometry  To assess the functionality of the produced phage expressing the Sq lib , a monoclonal sandwich ELISA was performed. Here, the C-terminal ABD [17] that is co-expressed with the sequestrins ( Figure S1A) on the phage allowed for detection of correctly expressed construct by assessing binding of HSA in an ELISA. Correlation between a correct sequestrin sequence and an ELISA positive signal was observed.

Phage Selection of Sequestrins towards Amyloid Beta Peptide
After verifying the compatibility of the phage display system and sequestrins, the functionality and library design was evaluated by preforming a selection towards the Aβ 1-40 peptide ( Figure S1). The selection process was monitored by titrations of phage inputs and outputs for each cycle, showing enrichment of full sequestrin expressing clones and increase in phage titres with increased number of cycles in the selection. In the third selection cycle, an enrichment was seen in which 50% of the input phage were recovered after extensive washing. In this cycle, only full-length sequestrins were found in PCR screening of the phage eluate, indicating a successful selection.
Enrichment of target-binding clones during the selection was assessed by polyclonal phage ELISA. The phage pool showed an increased signal towards Aβ 1-40 peptide with increasing selection rounds, correlating to phage enrichment in selection. Moreover, a shift in sequence distribution from the naïve library could be observed ( Figure S2), supporting the results from titrations and the polyclonal ELISA. The clones from the output had an average of four mutations per sequestrin, compared to seven in the naïve library. Mutations were found predominantly located in the second helix in each subunit. Thirteen sequestrins among the most reoccurring clones were selected for further characterization (Table S1).

Expression and Cloning of Sequestrins
The thirteen clones from the selection towards the Aβ 1-40 peptide were subcloned into a pET26b(+)-vector with a C-terminal hexa-histidine tag (further denoted as Sq Aβ -His 6 ) for expression in E. coli (Table S1). The clones obtained from the fourth and fifth selection round (clones 10-23) showed somewhat improved yields during production compared to the clones from the third round. Protein purity and size was analysed by gel electrophoresis (Figure 3). A single band was observed at 14.4 kDa corresponding with the expected weight of approximately 13 kDa for all sequestrins (Table S2). The yield for Sq Aβ8 was very low, seen as a faint band of the expected size in the Sodium Dodecyl Sulphate-Polyacrylamide Gel (SDS-PAGE). The mass of the proteins was analysed using mass spectrometry and correlated in general well to the respective molecular weights according to sequences. However, clone Sq Aβ8 , Sq Aβ9 and Sq Aβ14 showed a larger difference between expected and observed mass that is not explained by difference in cleavage of signal peptide during production and is likely a post translation modification (Table S2).
Mol. Sci. 2023, 24, x FOR PEER REVIEW 5 and correlated in general well to the respective molecular weights according to sequen However, clone SqAβ8, SqAβ9 and SqAβ14 showed a larger difference between expected observed mass that is not explained by difference in cleavage of signal peptide du production and is likely a post translation modification (Table S2).

Circular Dichroism for Secondary Structure Determination of Sequestrins
Circular dichroism (CD) spectroscopy was employed to verify the secondary st ture content of the new sequestrins. All the produced constructs except for SqAβ8 show the same characteristic alpha-helical secondary structure content ( Figure S3) as has pr ously been established for the other Aβ-binding affibody molecules such as ZAβ3 [7] ZSYM73 [10], indicating a similar three-dimensional structure The majority of the seq trins show an overlap between the CD spectra obtained before and after heat-indu denaturation, indicating full refolding capability. The thermal melting point (Tm) va between 35 and 48 °C (Table 1), which is in the same range as previously reported binding affibody molecules.

Circular Dichroism for Secondary Structure Determination of Sequestrins
Circular dichroism (CD) spectroscopy was employed to verify the secondary structure content of the new sequestrins. All the produced constructs except for Sq Aβ8 showed the same characteristic alpha-helical secondary structure content ( Figure S3) as has previously been established for the other Aβ-binding affibody molecules such as Z Aβ3 [7] and Z SYM73 [10], indicating a similar three-dimensional structure The majority of the sequestrins show an overlap between the CD spectra obtained before and after heat-induced denaturation, indicating full refolding capability. The thermal melting point (T m ) varied between 35 and 48 • C (Table 1), which is in the same range as previously reported Aβ-binding affibody molecules.

SPR-Based Biosensor Analysis of Binding to the Amyloid Beta Peptide
The affinity between the selected sequestrins and biotinylated Aβ 1-40 was evaluated by a surface plasmon resonance (SPR)-based biosensor assay where the target antigen was immobilized on a streptavidin chip. Eleven of the sequestrins demonstrated affinities in the range of K D 1-30 nM ( Figure S4 and Table 1). Generally, the kinetics of the interactions were characterized by slower association compared to what is typically observed for conventional three-helical affibody molecules, likely due to structural rearrangements upon target binding as has been demonstrated previously for related dimeric peptidebinding affibody molecules [7]. Dissociation was in general very slow, approximately to 10 −5 s −1 for the sequestrins with highest affinity, indicating a stable complex ( Figure S4 and Table 1). Sq Aβ8 that was produced with low yield demonstrated a relatively weak binding. The signal from the negative control Z Taq with specificity for Taq polymerase [18], was considerably lower compared to the sequestrins.
The sequestrins with the highest affinity (Sq Aβ22 and Sq Aβ23 with K D of 1 and 3 nM, respectively) were discovered in the output after the fifth selection cycle with more stringent washing, exemplified by clone Sq Aβ22 that was highly enriched in the fifth selection cycle ( Figure 4) and displaying the slowest dissociation among the new binders. As expected, the off rates increased when analysed at 37 • C. Still, the affinities for Sq Aβ22 and Sq Aβ23 were high for first-generation binders and in the range of K D 10-15 nM. were characterized by slower association compared to what is typically observed for conventional three-helical affibody molecules, likely due to structural rearrangements upon target binding as has been demonstrated previously for related dimeric peptide-binding affibody molecules [7]. Dissociation was in general very slow, approximately to 10 −5 s −1 for the sequestrins with highest affinity, indicating a stable complex ( Figure S4 and Table 1).
SqAβ8 that was produced with low yield demonstrated a relatively weak binding. The signal from the negative control ZTaq with specificity for Taq polymerase [18], was considerably lower compared to the sequestrins. The sequestrins with the highest affinity (SqAβ22 and SqAβ23 with KD of 1 and 3 nM, respectively) were discovered in the output after the fifth selection cycle with more stringent washing, exemplified by clone SqAβ22 that was highly enriched in the fifth selection cycle ( Figure 4) and displaying the slowest dissociation among the new binders. As expected, the off rates increased when analysed at 37 °C. Still, the affinities for SqAβ22 and SqAβ23 were high for first-generation binders and in the range of KD 10-15 nM.

Comparison of Expression Levels
Earlier studies on dimeric peptide-binding affibody molecules have indicated that the larger protein structure compared to monomeric three-helical affibody molecules tend to have a negative impact on the yield when recombinantly produced in E. coli. The sequestrins Sq Aβ22 and Sq Aβ23 were expressed in fusion to a C-terminal his-tag in E. coli and purified using Immobilized Metal Affinity Chromatography (IMAC). A previously described dimeric amyloid-beta binding affibody molecule, Z SYM73 [10], was included for comparison and the results showed around 2-fold higher yield for the new sequestrins compared to Z SYM73 (Table 2).

Analysis of Changes in Secondary Structure Content upon Target Binding
It has previously been demonstrated that similar homodimeric amyloid-beta binding affibody molecules undergo structural rearrangement upon peptide sequestration, forming a four-stranded beta-sheet comprised of one strand from each subunit and a beta hairpin from the amyloid-beta peptide [7]. CD spectroscopy was used to reveal indications of similar structural rearrangements in Sq Aβ22 and Sq Aβ23 upon binding Aβ. First, CD spectra for free Aβ 1-40 peptide and sequestrin were recorded, respectively, and thereafter compared to the spectra for co-incubated Aβ 1-40 peptide and sequestrin. The sum of the ellipticities for the free interactants were then subtracted from the ellipticities recorded for the co-incubated samples, demonstrating a change in signal thus indicating a change in secondary structure content upon binding ( Figure 5A,B). The secondary structure content was approximated using the BeStSel algorithm [19], revealing a loss of unstructured peptide content, a decrease in alpha helicity and gain of beta sheet content for the samples co-incubated with Aβ peptide and sequestrin, and the effect was more prominent for Sq Aβ23 . Next, the melting temperatures for the co-incubated samples were determined, demonstrating that binding to the Aβ peptide resulted in a substantial increase in T m , from 42 to 59 • C for Sq Aβ22 and 48 to 68 • C for Sq Aβ23 (Figure 5C,D). In a control experiment, the CD spectrum for Aβ 1-40 peptide was recorded, as well as the ellipticity for Aβ 1-40 peptide at 221 nm in a variable temperature measurement, showing negligible and relatively stable signal at 221 nm in that temperature range ( Figure S5).
Finally, as for free the free sequestrins, Sq Aβ22 and Sq Aβ23 co-incubated with Aβ 1-40 peptide demonstrated full refolding after heat-induced denaturation ( Figure 5E,F).

Aβ Aggregation
Potential inhibition of amyloid-beta aggregation was assessed using a thioflavin T (ThT)-based fluorescence-spectroscopy assay. ThT fluorescence, corresponding to Aβ peptide aggregation, was monitored for 92 h in a fluorescence microplate reader. In addition to Aβ 1-40 , the more aggregation-prone Aβ 1-42 was also included in the analysis. First, ThT fluorescence was measured at 92 h for 20 µM Aβ 1-40 and Aβ 1-42 , respectively, showing the expected aggregation, and higher signal for Aβ 1-42 ( Figure 6A,B). As a control, the sequestrins Sq Aβ22 and Sq Aβ23 were incubated with ThT fluorophore and the fluorescence was recorded for 92 h, demonstrating negligible signal compared to the Aβ peptides ( Figure 6A,B). Next, Aβ 1-40 and Aβ 1-42 , respectively, were co-incubated with equimolar concentrations (1:1) of respective sequestrin, showing very low signals comparable to background for all four samples ( Figure 6A,C). Moreover, ThT fluorescence was monitored for five different concentrations of Aβ 1-42 , ranging from 12 to 20 µM, showing the expected aggregation and a concentration-dependent maximum signal ( Figure 6D). The sequestrins are expected to bind to monomeric Aβ peptide with 1:1 stoichiometry. Nevertheless, the assay was also conducted with lower sequestrin-to-peptide ratios, corresponding to 4 µM sequestrin in 20 µM Aβ peptide (i.e., 1:5 ratio), as well as 2 µM sequestrin in 20 µM Aβ peptide (i.e., 1:10 ratio). The results showed lower and incomplete inhibition of the aggregation, with signals at 92 h that were similar to the signals corresponding to 16 µM free Aβ peptide for the 1:5 ratio and 18 µM free Aβ peptide for the 1:10 ratio ( Figure S6).

Aβ Aggregation
Potential inhibition of amyloid-beta aggregation was assessed using a thioflavin T (ThT)-based fluorescence-spectroscopy assay. ThT fluorescence, corresponding to Aβ   expected to bind to monomeric Aβ peptide with 1:1 stoichiometry. Nevertheless, the assay was also conducted with lower sequestrin-to-peptide ratios, corresponding to 4 μM sequestrin in 20 μM Aβ peptide (i.e., 1:5 ratio), as well as 2 μM sequestrin in 20 μM Aβ peptide (i.e., 1:10 ratio). The results showed lower and incomplete inhibition of the aggregation, with signals at 92 h that were similar to the signals corresponding to 16 μM free Aβ peptide for the 1:5 ratio and 18 μM free Aβ peptide for the 1:10 ratio ( Figure S6).

Discussion
Neurodegenerative disorders are a huge burden for society and the burden is expected to grow with the increasing global life expectancy [20]. A better understanding of the molecular mechanisms behind such diseases and finding suitable diagnostics and treatment strategies is thus essential.
In an effort to contribute with new molecular tools for studies and inhibition of peptide aggregation in neurodegenerative disorders, a new protein scaffold was designed, intended for generation of affinity reagents for peptides such as the amyloid beta (Aβ) peptide.
Based on data from previous research on related affibody-based peptide binders, a head-to-tail dimeric protein scaffold (denoted sequestrin) was designed and used for construction of a combinatorial protein library, containing around 5 × 10 9 variants. The sequestrin library was displayed on phage and used in biopanning against the Aβ peptide to assess the utility of the scaffold and the library design in respect of development of peptide-binding affinity reagents. The library design resulted in a theoretical diversity of 1.27 × 10 22 variants. The obtained phage library, containing 5 × 10 9 individual clones, is thus only covering a very small fraction of all possible combinations. However, the library was intentionally designed to include many randomized positions in the protein, given

Discussion
Neurodegenerative disorders are a huge burden for society and the burden is expected to grow with the increasing global life expectancy [20]. A better understanding of the molecular mechanisms behind such diseases and finding suitable diagnostics and treatment strategies is thus essential.
In an effort to contribute with new molecular tools for studies and inhibition of peptide aggregation in neurodegenerative disorders, a new protein scaffold was designed, intended for generation of affinity reagents for peptides such as the amyloid beta (Aβ) peptide.
Based on data from previous research on related affibody-based peptide binders, a head-to-tail dimeric protein scaffold (denoted sequestrin) was designed and used for construction of a combinatorial protein library, containing around 5 × 10 9 variants. The sequestrin library was displayed on phage and used in biopanning against the Aβ peptide to assess the utility of the scaffold and the library design in respect of development of peptide-binding affinity reagents. The library design resulted in a theoretical diversity of 1.27 × 10 22 variants. The obtained phage library, containing 5 × 10 9 individual clones, is thus only covering a very small fraction of all possible combinations. However, the library was intentionally designed to include many randomized positions in the protein, given the limited amount of information on this new type of protein scaffold and mode of binding. The long-term aim is that selections, followed by sequencing and characterization of binders will result in knowledge that will guide the design of improved next-generation sequestrin libraries in the future. Moreover, the recent substantial improvement in computational methods for predicting protein folding and protein-protein interactions, such as neural network-based models [21], paves the way for computer-aided library design based on deep-sequencing data on naïve, displayed and enriched library clones in the future.
A tendency of growth bias was observed during amplification of the phagemid library, which is likely due to the more complex protein structure compared to monomeric affibody molecules. Adjusting the amplification conditions during cultivation as described above was critical for ensuring a successful selection procedure.
Sequencing of the naïve library revealed that the average number of mutations per sequestrin was around seven, which after selection had shifted to around four among the enriched peptide-binding variants. Future selections against other targets will show if this trend continuous and could hopefully be valuable for further developments of the sequestrin scaffold for protein engineering.
Based on the most reoccurring sequences and taking into account representation from the different selection cycles (3, 4, and 5), 13 clones were selected for characterisation. The thirteen sequestrins were produced, purified, and initially analysed in terms of secondary structure content, thermal stability, refolding and affinity.
Eleven of the thirteen candidates showed high affinity for the amyloid-beta peptide with K D in the 1-30 nM range, characterized by relatively slow kinetics for both the association and the dissociation. The variants with highest affinity (Sq Aβ22 and Sq Aβ23 ) originated from the fifth selection cycle, indicating that the increased stringency in the later cycles indeed enriched for high-affinity binding. As expected, analysis of the interaction at 37 • C demonstrated faster dissociation, corresponding to around 5-to 10-fold higher K D . It should be noted that immobilizing the Aβ peptide on the chip probably results in an underestimation of the affinity. Previously reported SPR-based studies on structurally related dimeric affibody molecules for Aβ revealed a relatively large difference when the affinity was determined using SPR compared with analysis in solution [10]. This is likely due to the structural rearrangements of both interactants upon binding [7], which might be slower when the Aβ peptide is immobilized on the chip surface.
To get an indication whether similar structural rearrangement also occurs when the new sequestrins bind to the Aβ peptide, circular dichroism (CD) spectroscopy was used to detect potential changes in secondary structure content upon binding. The analysis showed that the Aβ peptide is unstructured in solution and the secondary structure content of the sequestrins is largely alpha helical. Co-incubation of Aβ peptide and respective sequestrin resulted in changes in the CD spectrum compared with the sum of CD spectra of the free interactants. Moreover, the thermal stability of the complex was increased 17-20 • C compared to the T m for the free sequestrins. In future studies, determining the three-dimensional structure of the complex would confirm whether these results are indeed due to structure rearrangements in the sequestrins and the Aβ peptide, similar to what has been observed for dimeric peptide-binding affibody molecules [7].
Finally, the effect from binding on the aggregation propensity of Aβ was assessed using a ThT fluorescence assay. The results showed that both sequestrins fully inhibited aggregation of Aβ when added at a 1:1 molar ratio, both for Aβ 1-40 and for the more aggregation-prone variant Aβ 1-42 . Future studies on the effects of the new sequestrins on Aβ pathology in AD models is hence warranted. A challenge when targeting neurodegenerative diseases with biologics is the limited uptake across the blood-brain-barrier. Hence, genetic fusion of the sequestrins with a domain for transferrin receptor (TfR)-mediated transcytosis transportation [22] would be interesting.
In summary, the results show that the design of the new sequestrin scaffold and library is suitable for selection of high-affinity peptide binders using phage display technology. Future selections against other aggregation-prone peptides, such as alpha synuclein, tau and TDP-43 will hopefully further validity the utility of this new class of affinity reagent.

Library Design and Molecular Cloning
Two randomized double stranded DNA oligos, encoding each of the two sequestrin (Sq) subunits, were purchased from TWIST Bioscience (San Francisco, CA, USA) for assembly into one library oligo by hybridization of overlapping bases in each subunit: 5 -GCGGGTGGCGAANNNNNNNNNNNNCCGAACTTANNNNNNGACCAANNN-TGTGCCNNNNNNCGTAGTNNNGAAGATGATCCTAGTCAAAGCGCTAACTTG-NNNGCAGAAGCTAAAAAGCTAAATGATGCTCAGGCGCCGGCGAGCAGCAGCA GCGGGAGCAGCAGCAGCGGGCGCGCGAGTGCGGGTGGCGAGNNNNNN-NNNNNNCCGAACTTANNNNNNGACCAANNNTGTGCCNNNNNNCGTAGTNNN GAGGATGACCCTAGTCAAAGCGCTAACTTGNNNGCAGAAGCTAAAAAGCT-AAATGATGCTCAGGCGCCGAAA -3 (with randomized codons illustrated as NNN). The genes were flanked by BamHI and SalI restriction sites for subcloning in fusion to a gene for an albumin-binding domain (ABD; [17]) into the pAffi1 phagemid [5]. Each subunit of the library was amplified by polymerase chain reaction (PCR) in 12 cycles using Phusion DNA polymerase (New England Biolabs, Ipswich, MA, USA) and primers introducing 60 overlapping bases into the two oligo subunits for subsequent hybridization. The PCR products were purified using a PCR purification kit (Qiagen GmbH, Hilden, Germany). Equimolar amounts of the oligo subunits were hybridized into one long library gene of 321 bp, which was subsequently PCR-amplified in 10 cycles using Phusion DNA polymerase (New England Biolabs, Ipswich, MA, USA). PCR products of correct length were purified by preparative gel electrophoresis (2% agarose gel) followed by purification using a QIAquick gel purification kit (Qiagen GmbH, Nordrhein-Westfalen, Germany). Purified PCR products were digested by BamHI and SalI (New England Biolabs, Ipswich, MA, USA) enzymes. The modified pAffi1 vector was digested by the same enzymes and purified by preparative gel electrophoresis. Purified sequestrin fragments were ligated to the phagemid vector as N-terminal fusion to a gene encoding an albumin-binding domain (ABD) [17], using T4 DNA ligase (New England Biolabs, Ipswich, MA, USA) at a 1:6 molar ratio of vector to insert. Ligated vector was purified using a QIAquick PCR purification kit (Qiagen GmbH, Germany) before transformed into TG1 electrocompetent cells (Lucigen, Middleton, WI, USA).
Phage stocks were created by standard procedures in XL1Blue cells (Aglient, Santa Clara County, CA, USA). Briefly, superinfection was performed using 5-fold excess of M13K07 phage (New England Biolabs, Ipswich, MA, USA) and precipitation of phage particles was performed using polyethylene glycol (PEG)/NaCl to yield phage titres of approximately 10 13 pfu/mL.

Library Validation
Superinfected bacterial colonies were individually screened for (1) library size by titration, (2) insert length by PCR amplification using DreamTaq DNA Polymerase (Thermo Scientific, Waltham, MA, USA) and gel electrophoreses, and (3) DNA sequence by Sanger sequencing (Microsynth SeqLab, Gottingen, Germany). Monoclonal phage ELISA was used to validate expression of recombinant protein on the phage surface by ABD binding to human serum albumin (HSA). Individual clones were cultivated in 96-deep well format in tryptic soy broth medium supplemented with yeast extract (TSBY) and 100 µg/mL carbenicillin at 30 • C, 250 rpm, ON. After ON, cultures were re-inoculated and incubated at 37 • C, 250 rpm until OD 600 reached 0.5-0.8 AU. Superinfection with 0.3 × 10 6 pfu/clone M13K07 (New England Biolabs, Ipswich, MA, USA) at 37 • C without rotation for 30 min proceeded induction with 1 mM Isopropyl β-D-1-thiogalactopyranoside (IPTG; Chemtronica, Stockholm, Sweden) and further cultivation ON at 37 • C, 250 rpm, under additional antibiotic pressure from 30 µg/mL kanamycin. Harvested phage supernatants were incubated for 1 h in 384-microwell plates (Nunc, PS, Low binding, Hi-Edge, clear) precoated with HSA [5 µg/mL] or bovine serum albumin (BSA) [1 w/v%]. The ELISA plate was blocked with 1 w/v% BSA prior to phage incubation to minimize background, and plates were washed in phosphate-buffered saline + 0.05% Tween-20 (PBST) before incubating with a mouse monoclonal M13 Bacteriophage Antibody (HRP) (Sino biological, Beijing, China) according to supplier's instructions. Signal development was done with Pierce™ TMB Substrate Kit (Thermo Scientific, Waltham, MA, USA), as by instructions. After colorimetric development, the reaction was terminated by addition of 2 M H 2 SO 4 . Absorbance was measured at 450 nm using a CLARIOStar Plus plate reader (BMG Labtech, Ortenberg, Germany). DNA sequencing was performed by Sanger sequencing (Microsynth SeqLab, Göttingen, Germany), and sequences analysed with the Geneious software (version 11.2, Biomatters LTD, Auckland, New Zealand).

Selections of Sequestrins against Amyloid Beta
The sequestrin library phage stock (denoted Sq lib ) was used in selections against C-terminally biotinylated Aβ 1-40 peptide (AnaSpec, Fremont, CA, USA) in a total of five rounds with decreasing amount of soluble target antigen in each round (50 nM, 40 nM, 20 nM, 10 nM, 1 nM). The incubation temperature varied between the cycles (4 • C ON for cycle 1, 1 h RT for cycle 2-4, 1 h 45 • C for cycle 5) before antigen was captured by Dynabeads M-280 Streptavidin beads (Invitrogen, Waltham, MA, USA). All cycles were preceded with a negative selection round towards beads pre-blocked with 5 w/v% BSA. The stringency of PBSTB (0.1% Tween-20, 3 w/v% BSA) washes was increased with each selection round (2 × 1 min, 4 × 1 min, 5 × 3 min, 5 × 6 min, 4 × 6 min + 1 × 2 h + 1 × 6 min), where the last wash was done in PBS buffer. Phage eluates were obtained by incubation for 10 min in 0.1 M glycine-HCl, pH 3.0, followed by neutralization by Tris-HCl, 1 M, pH 8.0. Eluates were amplified by infection with 100× excess of XL1Blue E. coli cells, followed by plating on Aquare BioAssay Dish (Corning, Somerville, MA, USA) with blood agar base (BAB; Merck, Darmstadt, Germany) supplemented with 2% D-glucose, 100 µg/mL carbenicillin, and 10 µg/mL tetracycline and incubated at 37 • C ON. Bacterial colonies were recovered by addition of TSBY medium to the plates, followed by scraping off colonies from the plates and dissolving the cells in TSBY before continuing cultivation in a suspended format. The cultures containing infected bacteria with the eluted phage (>100× excess compared to eluate complexity) were reinoculated to OD 600 = 0.2 AU, followed by superinfection with five times excess of M13K07 phage at an OD 600 = 0.8 AU.

Phage ELISA
Amplified phage stock and individual clones after each selection round were cultured for polyclonal and monoclonal phage ELISA, respectively, as described above. All phage stocks were diluted to the same concentration before incubating with target antigen.
ELISA plates in 384-well format (Nunc, PS, Low binding, Hi-Edge, clear) were prepared with four wells per individual sample according to: (1) one well coated with 5 µg/mL HSA, (2) one well coated with 1 w/v% BSA, (3) one well pre-coated with 5 µg/mL streptavidin followed by 1 µg/mL biotinylated amyloid beta peptide (Aβ 1-40 ; AnaSpec, USA) and (4) one well pre-coated with 5 µg/mL streptavidin followed by 1 w/v% BSA. Development and washes were performed as described above. The signal, representing background, from wells with BSA (for 1) or streptavidin-BSA (for 3) were subtracted. The signal from wells with Aβ 1-40 was normalized with the signal from wells with HSA. DNA sequences were identified by Sanger sequencing (Microsynth SeqLab, Göttingen, Germany), and the sequences were analysed with the Geneious software (version 11.2, Biomatters LTD, Auckland, New Zealand).

Expression and Purification of Soluble Sequestrins
Sequences from the selections were chosen for further characterization. Sequestrinencoding DNA was amplified from phagemids with primers designed for the In-Fusion HD cloning kit (Takara Bio Europe, Göteborg, Sweden), according to manufacturer's recommendations into the pET-26b(+)-vector for periplasmic production with a C-terminal His 6 -tag. Sequence-verified clones (Sanger sequencing, Eurofins Genomics GmbH, Ebers-berg, Germany) were transformed using heat shock to E. coli BL21(DE3) for recombinant protein expression. Cultures were inoculated to an OD 600 of 0.1 AU and incubated in TSBY with 25 µg/mL kanamycin at 37 • C until OD 600 reached 1.0 AU before inducing with 1 mM IPTG and thereafter incubated at 25 • C for 16 h. Cell pellets from harvest were dissolved in running buffer [47 mM Na 2 HPO 4 , 3 mM NaH 2 PO 4 , 300 mM NaCl, 15 mM imidazole, pH 7.4] before proceeding with purification. Cells were sonicated using a Vibra-Cell VCX 130 sonicator (Sonics, Newtown, CT, USA) and cell debris was removed by centrifugation and filtration. Cell lysate was loaded on an equilibrated HisPur Cobalt Resin (Thermo Scientific, Waltham, MA, USA) and washed with running buffer. The sample was eluted with running buffer supplemented with 150 mM imidazole. Fractions of eluate containing protein according to analysis with the Pierce™ BCA Protein Assay Kit (Thermo Scientific, Waltham, MA, USA), performed according to manufacturer's instructions, were pooled and buffer exchanged to PBS on PD 10 desalting columns (Cytiva, Marlborough, MA, USA). The samples were analysed with Sodium Dodecyl Sulfate-PolyAcrylamide Gel (SDS-PAGE) (NuPAGE Bis-Tris 4-12%, Invitrogen, Waltham, MA, USA). Molecular mass was analysed by electrospray ionization mass spectrometry on a Thermo Ultimate3000 Bruker Impact II system connected to a ProSwift RP-4H, 1 × 50 mm column (Thermo Fisher, Waltham, MA, USA) using a linear gradient elution with acetonitrile (3 to 95%), supplemented with 0.1% formic acid. The instrument was run with settings for electrospray ionization with a positive ion polarity, spanning the mass range 500 to 50,000 m/z.

Circular Dichroism Spectroscopy
The secondary structure content of the sequestrins was analysed using circular dichroism spectroscopy on a Chirascan system (Applied Photophysics, Leatherhead, UK) with a 1 mm High precision cell (110-1P-40 cuvettes, Hellma Analytics, Munich, Germany). Five wavelength scans were recorded and averaged between 195 nm and 260 nm at 20 • C on 0.2 mg/mL protein in PBS. Melting point was determined by using a temperature gradient of 1 • C per minute at 221 nm for five average readings. The refolding capability was assessed by repeating the spectral scan after the sample had been subjected to heat treatment and cooled down to 20 • C. The spectra from before and after heating was compared to assess refolding. Respective sequestrin and Aβ 1-40 (AnaSpec, USA) were co-incubated at equimolar concentrations of 15.7 µM and analysed by circular dichroism for analysis of changes in secondary structure content upon interaction. Secondary structure content was approximated by BeStSel algorithm [19].

Aggregation Assay
Fibrillization inhibition assays were done for the Aβ 1-40 (#AS-24236, AnaSpec, Fremont, CA, USA), and Aβ 1-42 (#AS24224, AnaSpec, Fremont, CA, USA) using the sequestrins Sq Aβ22 and Sq Aβ23 as inhibitors for aggregation at decreasing molar ratios 1:1, 1:5, and 1:10 in relation to the 20 µM of Aβ. All proteins were thawed on ice and spun down at 10,000× g for 5 min at 4 • C to remove precipitates. Each sample was prepared in triplicate in a volume of 60 µL per reaction by adding PBS (Aβ 1-40 : 10 mM phosphate and 150 mM NaCl, pH 7.2; Aβ 1-42 : 10 mM phosphate and 150 mM NaCl, pH 8.0), 20 µM Thioflavin T (ThT) dye, and 20 µM sequestrin before vortexing for 3 s and kept on ice. When all samples were prepared the Aβ peptide was added at 20 µM concentration to all sequestrin samples, or as a standard at 12-20 µM, to the tubes and vortexed for 1 s. The sample was added to a 384-well plate (Nunc, PS, Low binding, Hi-Edge, clear) with surrounding wells filled with PBS, and sealed to minimize evaporation during analysis. The fluorescence intensity was immediately analysed in a CLARIOStar Plus (BMG Labtech, Ortenberg, Germany) with excitation at 440-10 nm and emission at 480-10 nm. The readings were done during 999 cycles with a cycle time of 325 s at 37 • C with 50 flashes with 0.1 s settling time per well, using 1500 gain adjustment. The plate was shaken in an orbital motion 500 rpm for 1 s before each measurement. Analysis was done in the Mars analysis software (version 3.0, BMG Labtech, Ortenberg, Germany) where replicates were averaged, and blank subtracted from well containing only PBS with ThT dye.