Selection of Peptides that Associate with Dye-Conjugated Solid Surfaces in a pH-Dependent Manner Using cDNA Display.

Peptides that recognize artificial materials including synthetic polymers and small molecules are drawing attention in the fields of biotechnology and chemical biology. In particular, reversible peptide aptamers that associate with the target molecules only under specific conditions are interesting. In this work, peptide aptamers that recognize a phenolphthalein derivative (PhP: a pH-sensitive organic dye) immobilized on a solid surface in a pH-dependent manner were selected using an in vitro display method (cDNA display). Considering the hydrophobic and aromatic nature of PhP, we prepared a biased DNA library (3A library) that encodes more aromatic amino acids than the standard random codon and performed seven rounds of selection from >1010 peptide species. The selected peptides including LVFLIWWM (LV59) associated with PhP-modified solid support (sepharose resin and magnetic beads) in neutral buffer but readily dissociated under basic conditions where PhP undergoes large structural change from lactone to quinoid, which is accompanied by increase of hydrophilicity and anionic charge. Control experiments suggested that LV59 recognized both phenol and lactone moieties, and the association under neutral pH is mainly driven by π-stacking and hydrophobic interaction between the peptide and PhP. Notably, however, total hydrophobicity and number of aromatic rings did not completely explain the affinity, and sequence specificity was observed to some extent. After further optimization, this interaction pair would be practically useful for protein purification.

Introduction

Polypeptides play fundamental roles in molecular recognition in biology: they recognize not only other proteins but also small molecules including lipids,[1] saccharides,[2] nucleic acids,[3,4] and others[5,6] with high specificity. To optimize or create such peptides, directed evolution technologies have been established that can generate and evaluate large numbers of peptides in vitro, such as phage display,[7] ribosome display,[8] and mRNA display.[9,10] So far, many peptides that bind to non-natural substances such as synthetic polymers, organic dyes, and explosives have been successfully identified.[11] Pioneering works in this field include peptides that bind to semiconductor surfaces[12] and carbon nanotubes[13] selected by phage display. Furthermore, stimuli-responsive peptides, which recognize target substances only under certain conditions, are lately gaining attention in terms of molecular recognition and practical applications. For example, Serizawa and others performed phage display screening to find peptides that preferentially (2–5 fold) bind to cis (i.e., UV-irradiated) forms of azobenzene on polymer–film surfaces.[14,15] Ito and others also selected a photo-responsive peptide aptamer that incorporated an azobenzene moiety using ribosome display.[16]

In addition to the display techniques above, cDNA display[17−23] is another promising technology to identify peptide aptamers from a large library. In cDNA display, a special DNA linker (Figure S1) featuring puromycin (an antibiotic that is incorporated at the C-terminal of a nascent peptide in ribosomes)[18] is ligated to an mRNA library, yielding mRNA–linker complex. In vitro translation and mRNA–protein linkage is then performed, and the mRNA moiety is reverse-transcribed to yield cDNA–protein conjugates for use in selection processes (Figure S2, summarized scheme is also shown in Figure c). Compared with other techniques, cDNA display has advantages in terms of library size (1012–13) and stability under harsh conditions such as organic solvents, strong acid/base, and heat. We and others have used cDNA display to identify peptides that associate with proteins,[18,19] nucleic acids,[20] solid phases,[21] and lipid membranes.[22] However, though a paper was published on a binder for a small biomolecule (ATP),[23] there is no report on peptide aptamers for artificial small molecules with this technique. Further, none of the targets or selected peptides so far underwent stimuli-responsive structural change.

Figure 3

Selection of PhP-binding peptides. (a) Expected amino acid appearance rate of 3A (filled bars) and NNN (open bars) libraries. Red bars indicate aromatic amino acids. (b) Overview of the DNA library. LHR: linker hybridization region. For a detail of the random region, see Figure S4. (c) Schematic representation of the selection cycle. (d) Progress of selection monitored by qPCR. (e) qPCR results of eluents at the seventh round. Peptide-driven interaction was confirmed, with almost no binding to the control beads.

Here, we designed and performed the first cDNA display-based selection of peptide aptamers that recognize the titration status of a small molecule, at least in part. We focused on pH indicators as ligands because some of these dyes show drastic, quick, and reversible structural change in response to pH (Figure ).[24] We also envisioned that such pH-dependent binding systems would be applicable for protein purification or smart materials if specific and high-affinity peptides were selected. In particular, we prepared a cDNA display library consisting of eight amino acids and performed selection against a phenolphthalein derivative (PhP, Figure a) immobilized on a solid phase (sepharose resin/magnetic beads). We then analyzed the selected sequences and found LVFLIWWM (LV59), which associated with PhP under neutral conditions and rapidly dissociated under basic conditions. Evaluation of analogous peptides and control target beads was also performed to elucidate the binding mechanism.

Figure 1

Schematic representation of the purpose of this work. We aimed to screen peptides that associate with a phenolphthalein derivative (PhP) at pH 7 (i.e., closed lactone form) and dissociate from the molecule at pH 12 (i.e., open quinoid form).

Figure 2

(a) Chemical structure of PhP. (b) pH-dependent structural change of PhP. Only the main part of the molecule is shown. (c) Color change of PhP-immobilized resin from pH 7.0 (left) to pH 12.0 (right).

Results and Discussion

Design and Synthesis of the Target Compound

As mentioned above, we chose pH-dependent structural change of a dye as the target of in vitro selection. Compared with light irradiation,[14−16] pH change is more easily and reproducibly performed, and in most cases, titration status can be checked with the naked eye. In contrast, it is often difficult to completely shift the whole population of azobenzene molecules to one isomer because of the thermal reverse process.[25] Moreover, immobilized azobenzene molecules seem to yield more mixed cis/trans isomers at the photostationary state.[25] Although a few DNA aptamers that show pH-dependent binding to the targets have been developed recently,[26,27] there is no report of the selection of pH-dependent peptide aptamers.

Phenolphthalein is one of the most widely used pH indicators, which changes its color from colorless (pH ≤ 9) to red (pH ≥ 9).[24] This is due to the structural change from a closed lactone form to an open quinoid form upon deprotonation (Figure b).[28] First, we synthesized amino-modified phenolphthalein and coupled that to a short poly(ethylene glycol) (PEG) linker (Scheme S1). The compound, which we named PhP, was immobilized to the sepharose resin having the carboxy group and showed distinct color change in response to pH (Figure c). PhP-immobilized magnetic beads were also prepared accordingly. Interestingly, apparent pKa of immobilized PhP was about 11, which was shifted to more basic than that in solution (pKa = 9.6, Figure S3). Although the exact mechanism is not known, similar shift had been suggested in a previous paper.[29] As negative control, resin/beads were also prepared using ethanolamine or phenol (Scheme S2) as the immobilized molecule.

Library Design and cDNA Display Formation

We then designed a new peptide library of eight residues using a biased mixed codon (3A library, library diversity = 3 × 1010, Figure S4). This library should present more aromatic amino acids (>2 fold) than a standard NNN random codon (Figure a), and we considered it desirable because PhP is a hydrophobic compound with three aromatic rings under neutral conditions. The random moiety was inserted at the N-terminal of a scaffold protein (BDA: B domain of protein A)[30] to avoid potential steric interference of the cDNA moiety, and other DNA sequences for preparing cDNA display such as the T7 promoter and linker hybridization region (LHR) were attached to make the full construct (Figures b and S4). The efficiency of cDNA display formation (i.e., the ratio of a successfully purified display molecule to input mRNA–linker) of the construct was estimated to be about 2% (Figure S5).

In Vitro Selection

Next, we performed an in vitro selection consisting of seven rounds, starting from 120 pmol of mRNA–linker (Figure c). Considering the formation efficiency, the number of obtained cDNA display molecules should be 1.5 × 1011. This would cover most of the sequence space of the 8-mer peptide library. In the first and second rounds, PhP-immobilized sepharose resin was used as the binding target, and elution with 1% sodium dodecyl sulfate (SDS) and 80 mM NaOH aq was performed, respectively (Figures S6 and S7). Notably, cDNA display is tolerant to such strong alkaline solution, unlike mRNA display and ribosome display. In this sense, this selection takes an advantage of the stability of cDNA display. From the third to sixth rounds, selections using PhP-immobilized magnetic beads and aq NaOH were performed. Binding, washing, and eluting conditions were gradually changed during the process (see the Supporting Information for details). For each round, control selection with negative (i.e., PhP-non-immobilized) beads was also performed using the same display libraries. cDNA in the eluents were then quantified by quantitative polymerase chain reaction (qPCR) (Figure d) to monitor enrichment and selectivity. After the sixth round, peptides having affinity to PhP (but not to the beads themselves) seemed to be sufficiently concentrated from the initial library. We then performed the final round with proteinase K elution, which hydrolyzes peptides before the elution with NaOH aq, to remove molecules that associated with PhP at their cDNA part (Figure e). With this elution, almost no DNA was recovered from the control beads, which suggested that collected DNA from the PhP beads should encode peptides that had affinity to PhP. Notably, recovery efficiency of the seventh round was much reduced from the previous round, probably because of the overnight incubation at 37 °C in the presence of proteinase K at the elution step (for detailed protocols, see the Supporting Information), which might decompose some DNA or promote physical adsorption of the cDNA display molecule to the plastic tubes.

Identification of LV59

The final selected DNA sample was then analyzed by deep sequencing. There remained >103 peptides, suggesting relatively low selection pressure, so we cloned several top sequences and prepared cDNA display. Their pH-dependent interaction with PhP-bound beads was evaluated using SDS-polyacrylamide gel electrophoresis (PAGE) (Figures a,b and S8). As an index of peptide affinity, recovery efficiency (i.e., ratio of NaOH-eluted peptide to the input) was calculated (Table ). A few peptides, including LVFLIWWM (LV59), bound to PhP at pH 7 and dissociated under alkaline conditions as we expected, and it did not bind to the control beads (Table , Figure b). LV59 consists of hydrophobic amino acids only and have three aromatic residues. Currently, we do not know why the most abundant sequences (RYIRYYCP and VRRWWYPC) did not show detectable association to PhP beads, but one possibility is disulfide bond formation at a Cys residue with other peptides in the selection process. Notably, although the other peptides that showed affinity to PhP consisted of hydrophobic amino acids only (and some had more aromatic residues), LV59 showed highest recovery efficiency. This result indicated that total hydrophobicity of the peptide is an important but not the sole parameter for the interaction with PhP. Enrichment of aromatic residues except phenylalanine during the selection process was also seen at the library level (Figure S9). We believe these data supported our design and usage of 3A library for the selection.

Figure 4

pH-dependent interaction of LV59 and PhP. (a) Schematic representation of the experiment. Mixture of cDNA display (blue circle) and cDNA–linker complex (orange square) of LV59-BDA was reacted with PhP-immobilized (or control) beads, and only cDNA display was released from the beads by NaOH elution. IP: input, FT: flow through, EL: eluent. (b) SDS-PAGE analysis of fractions. Association of cDNA display to PhP(+) beads was supported by the decrease of the band of FT, and its NaOH-dependent recovery was confirmed from the band at EL. (c) Elution of cDNA display under different pH levels. Except for the second lane from the left, PhP beads were used. Samples were fluorescently detected using the fluorescein attached to the linker (see Figure S1).

Table 1

Sequences That Were Found by Next Generation Sequencing and Analyzed Individuallya

amino acid sequence	number of hydrophobic residues	number of aromatic residues	reads (PhP+, 7th)	recovery efficiency (%)b
RYIRYYCP	5	3	117	N.D.
VRRWWYPC	5	3	108	N.D.
LVFLIWWM (LV59)	8	3	59	3.5
VHYLYIDP	6	2	52	N.D.
YLYVWLWF	8	5	46	2.0
GHMHHHVG	4	0	43	N.D.c
YKMRHHWQ	3	2	43	N.D.c
WVIWILFF	8	4	22	1.6
MWFFMFWM	8	5	20	1.7

Number of reads of the peptides (in the final elute) are shown, together with their binding ability to PhP.

Recovery efficiency was evaluated from the intensity of the eluted cDNA display band of SDS-PAGE. N.D. indicates that recovery was not detected at all.

For these two peptides, recovery efficiency was evaluated at the peptide–linker forms.

Importantly, elution of LV59 was observed from pH 11 and gradually increased up to pH 12.5 (Figure c), which matched the color change of the immobilized PhP (Figure c). We also fused LV59 to another protein, PDO (POU-specific DNA binding domain of Oct-1)[31] and performed the same experiment (Figure S10). This confirmed that LV59 could reversibly associate to PhP regardless of the conjugated protein.

Analysis of LV59 Derivatives

We also prepared a peptide–linker complex of LV59-BDA by hydrolysis of mRNA during the formation of cDNA display (Figures a and S2) and used it for association/dissociation experiments. The obtained results not only proved that the molecule was bound to PhP in the absence of the cDNA (or mRNA) moiety but also indicated that recovery efficiency (i.e., ratio of eluted peptide to the input) was higher when these moieties were removed (compare Figures b and 4b). Apparent increased affinity is reasonable considering the bulky size and highly negative charge of the nucleic acid moieties, as demonstrated in our previous paper.[32] Taken together, the results indicated that the LV59 peptide (and not its nucleic acid) had some affinity to the closed lactone form of PhP but not to the open quinoid form. Importantly, both PhP beads and the eluted peptide could be used for the capture/release experiment again (Figure S11), which eliminated the possibility of decomposition- or irreversible denaturation-induced elution of the peptide during NaOH treatment.

Figure 5

Association and elution of LV59 peptide–linker and derivatives. (a) Peptide–linker conjugates (green triangle) were prepared by RNase treatment of mRNA display (see Figure S2), which were mostly composed of peptides. They were used for the experiment shown in Figure a. (b) Capture/elution of LV59-BDA. (c) Alanine replacement of aromatic residues (LV59A) completely suppressed interaction. (d) LV59R did not associate with PhP well. (e) LV59-BDA did not associate with phenol-immobilized beads well. IP: input, FT: flow through, EL: eluent. Samples were separated by SDS-PAGE and fluorescently detected using the fluorescein attached to the linker.

Next, to investigate the interaction of LV59 and PhP in more detail, peptide–linker complexes of several derivatives were synthesized and tested. Because LV59 contained three aromatic residues, we first replaced them for alanine (LV59A). This peptide did not show any detectable affinity (Figure c), clearly demonstrating the pivotal roles of π-stacking (and hydrophobic interaction) between these aromatic amino acids and phenolphthalein. The result also supported our working hypothesis in terms of the library design (vide supra). We also examined the effect of salt for this interaction, and a high concentration of salts gave a positive effect for the association (Figure S12), which further supported nonelectrostatic, hydrophobicity-driven interaction. Hence, the association mechanism of LV59 to PhP could be explained mainly by π-stacking of tryptophans and phenylalanine and in part by van der Waals interaction of all amino acids. In general, such interactions are not very specific and many (if not all) peptides consisting of hydrophobic and aromatic residues should have some affinity to PhP. This may explain why there were many sequences left after 7 rounds of selection and that even top sequences had relatively small read numbers (Table ).

However, when another derivative, LV59R, which had an inverted sequence of LV59, was tested, recovery efficiency was less than half that of LV59 (Figure d), which indicated that the total hydrophobicity and the number of tryptophans were important but not the only factors. In addition, when LV59 was attached at the C-terminal of BDA, a remarkable decrease of binding was observed, implying that the peptide–PhP interaction is position-specific (Figure S13). As another control, we made phenol-immobilized beads (Scheme S2) and examined if LV59 associated with the beads similarly. As shown in Figure e, recovery efficiency was only 40% compared with PhP-beads, indicating that these peptides do not associate well with simple (but also hydrophobic and pH-sensitive) phenol or the PEG linker for immobilization, and they recognize the lactone structure of phenolphthalein under neutral pH to some extent. All these experiments have been performed several times, and the results were reproducible (Table S1).

Conclusions

We designed and performed the first in vitro peptide selection against an immobilized pH-sensitive organic dye, PhP, using cDNA display technology. The selected peptide, LV59, which did not have any pH-sensitive side chains, associated with the dye-conjugated solid phase under neutral pH but apparently showed no affinity to the same material under basic conditions. After control experiments, it became clear that the dissociation is related to the pH-dependent structural change of PhP. Although the detailed binding mechanism is not clear, it is likely that π-stacking of tryptophans of the peptide and the three (protonated) aromatic rings of the dye play a pivotal role. LV59, and all other peptides in the library, has a pH-sensitive free amino group at the N-terminus, and we cannot eliminate the possibility that this group partially contributes to the interaction with the dye. Because only a small number of peptides showed detectable pH-dependent affinity with PhP (Table ), it is clear that the role of this group alone is small, if any. However, the synergetic effect of this amino group and the side chains of LV59 might be important for the interaction. Anyway, upon alkalization of the solution, PhP is deprotonated with concomitant color change, and the molecule becomes hydrophilic and anionic. This drastic structural change inhibits the above interaction and elutes the peptide from the beads. In contrast to previous studies using photo-isomerization,[14−16] the association/dissociation in this work also involves (de)protonation of a molecule and change of zeta potential of the solid phase. In this sense, there may be an argument that LV59 does not recognize pure conformation of the PhP at the single molecule level. However, considering a previous work,[14,15] the target organic molecule was used in the film; so, a change of macroscopic properties of the material could not be completely denied. Taking the results of phenol-immobilized control beads, we believe it is safe to say that our peptides recognize PhP that was immobilized on the solid material only at its lactone form.

Of course, there is more to study and optimize. For example, detailed characterization of the binding mechanism and affinity is yet to be performed. Also, a low recovery efficiency is a problem for practical applications, even though more target proteins could be obtained by repeated purification from the flow through (Figure S11). We believe further optimization of peptides is necessary, and we are moving in this direction now with newly designed libraries and selection protocols.

Methods

Preparation of the PhP-Immobilized Solid Phase

5-(NH2-PEG2-amide)-phenolphthalein, or PhP, was synthesized from 4-nitrophthalic anhydride and phenol in five steps. Detailed synthetic procedures and characterization of the compounds are described in the Supporting Information. The obtained compound was immobilized to the NHS-activated solid phase (resin or beads), and the unreacted functional group on the solid phase was passivated with ethanolamine. Control beads without PhP were also prepared.

DNA Construction

The DNA library used for selection was prepared by 2-step primer-extension PCR from the DNA coding BDA[18] and the primers shown in Table S2. The product was purified by preparative PAGE and amplified by PCR. The sequence of the final product was checked by Sanger sequencing. For more details, please refer the Supporting Information. Insertion of LV59 and other peptides at the N-terminus of BDA (or PDO) was also performed in a similar way.

Preparation of cDNA Display

Conversion of the DNA library to cDNA display was performed according to the method described in our previous paper,[33] with slight modifications. Briefly, DNA was transcribed to mRNA using T7 RiboMAX Large Scale RNA Production System (Promega) and photo-ligated to the cnvK-rG linker, which had been prepared as described in the Supporting Information. The obtained mRNA–linker complex was then in vitro translated with a rabbit reticulocyte lysate system (Promega, 6 pmol of mRNA–linker was added in a 50 μL reaction volume) at 30 °C for 20 min. After incubation in the presence of high K+ and Mg2+ to promote mRNA display formation, the mRNA display library was immobilized on streptavidin (SA)-coated magnetic beads (Dynabeads MyOne Streptavidin C1 magnetic beads, Invitrogen, 60 μL) at 25 °C for 30 min. The immobilized library was then reverse-transcribed by the ReverTra Ace reverse transcriptase (Toyobo, Japan, 200 U in a 50 μL reaction volume) at 42 °C for 30 min. RNase T1 (250 U in 50 μL, Thermo Fisher Scientific) was added to the beads, and the mixture was incubated for 15 min 37 °C to release the mRNA/cDNA–protein fusion molecules (i.e., cDNA display) from the beads. The eluted crude cDNA display library was purified by the His6 tag inserted at the C-terminus of the protein, using His Mag Sepharose Ni beads (20 μL, GE Healthcare). Detailed protocol is described in the Supporting Information.

Selection of Peptides against PhP

The purified cDNA display library was first incubated with the control solid phase without PhP to remove nonspecific binders. This preselection process was repeated for several times if necessary. In general, the flow-through was then incubated with the PhP-immobilized solid phase at 4 °C for 30 min, and the material was washed six times with selection buffer (50 mM Tris-HCl, 0.5 M NaCl, 1 mM EDTA, 0.05% Tween 20, pH 7.4). Bound cDNA display molecules were eluted twice with 80 mM NaOH aq at 25 °C for 15–30 min and the eluate was immediately neutralized with 1 M Tris (pH 6.6). From the second round, the amount of cDNA in each fraction was quantified by qPCR, and the results were normalized to the total input. At the first round, elution was performed by 1% SDS instead of NaOH aq, and at the seventh round, proteinase K was also used for elution. From the third round, control selections were also performed, where control resin or beads were used instead of PhP-immobilized resin or beads to obtain information on selectivity. After each round, PCR amplification of the eluate was performed and the obtained DNA was used for the preparation of the cDNA display library for the next round. Detailed protocol is described in the Supporting Information.

Formation of Peptide–Linker Complexes

The mRNA–linker complex (6 pmol) was converted to mRNA display as described above. Immobilization of mRNA display molecules on Dynabeads MyOne Streptavidin C1 beads (60 μL) was performed at 25 °C for 30 min, and then, RNase T1 (200 U) in selection buffer (40 μL) was added to decompose the mRNA moiety. The obtained supernatant solution contained the titled peptide–linker complex. Detailed protocol is described in the Supporting Information.

Calculation of Recovery Efficiency

cDNA display was prepared as described above, but purification using Ni-beads was not performed. Then, crude cDNA display (i.e., the input) was incubated with PhP-immobilized Dynabeads MyOne Carboxylic Acid at 25 °C for 30 min. The beads were washed with selection buffer and bound cDNA display was eluted with 50 mM NaOH aq at 25 °C for 15 min. The eluate was neutralized with Tris-HCl buffer (1 M, pH 6.6). Fractions were collected and analyzed by SDS-PAGE (4% stacking–6% separating, 20 mA, 120 min) containing 8 M urea, after the RNase H treatment. Band intensities were quantified using fluorescein, which was labeled at the puromycin linker. Recovery efficiency was calculated as the ratio of cDNA display in the eluate to that in the input. For peptide–linker complex, essentially the same protocol was used. Detailed protocol is described in the Supporting Information.