Peptide ligase–mediated display: A cell-free platform for tunable selection of affinity peptides

TL;DR

A new cell-free method called PL display allows for efficient selection of peptides that bind to specific targets, using covalent linkage and minimal linkers.

Contribution

The novel contribution is a cell-free platform using peptidyl transferase for covalent genotype-phenotype linkage with a minimal linker.

Findings

HA-tag sequences at 0.01% frequency were fully isolated in one round of FACS selection.

Consensus binding sequences were enriched from a random library with 1.7 × 10⁶ diversity in two FACS rounds.

The platform is not limited by cellular physiology or linker protein constraints.

Abstract

Herein, we report a bead-surface protein display method based on a peptidyl transferase reaction, termed peptide ligase–mediated display (PL display). This technique enables the covalent linkage of genotypic DNA and phenotypic protein variants on beads via a minimal nine–amino acid linker in a fully cell-free system. Using this method, hemagglutinin (HA)-tag sequences introduced at a 0.01% frequency were completely isolated in a single round of selection via fluorescence-activated cell sorting (FACS) against an anti-HA-tag antibody. Furthermore, consensus sequences that bind to the anti-HA-tag antibody were enriched from a random peptide library with a sequence diversity of 1.7 × 106 in two rounds of selection using FACS. This quantitative affinity selection platform using PL display is applicable under diverse conditions, as it is not constrained by cellular physiological properties, fluctuations in gene expression, or the structural and functional limitations of linker proteins involved in genotype–phenotype linkage. These advantages arise from the use of a fully cell-free system and covalent linkage with a minimal linker.

Figures

Schematic representation of target-binding peptide selection by PL display. A) Preparation of a peptide-displaying bead library using PL display. An initial DNA library is compartmentalized into emulsion droplets, each containing a single DNA molecule encoding a peptide variant, a capture primer-coated bead, and PCR primers, including a pentaglycine-conjugated primer. PCR is performed in each droplet to generate DNA products that encode peptide variants and carry a terminal pentaglycine motif. These DNA molecules hybridize to the capture primers on the beads and are subsequently extended to generate a DNA-immobilized bead library. After demulsification, the beads are re-encapsulated in emulsions for cell-free translation of peptide–sortase fusion proteins. Sortase-mediated transpeptidation transfers each peptide variant to the pentaglycine motif on its corresponding bead. The second demulsification step yields a peptide-displaying bead library, in which each peptide is covalently linked to its encoding DNA via a minimal peptide linker. B) Selection of target-binding peptide by PL display. Following emulsion PCR, Cy5-modified oligonucleotides are hybridized to the DNA on beads to identify those carrying the amplified sequences. Peptide variants are displayed on the beads through cell-free translation and sortase-mediated conjugation. The beads are then incubated with PE-conjugated antibodies, which serve as model targets for peptide binding. Beads displaying target-binding peptides are isolated by FACS based on PE fluorescence. DNA from the sorted beads is amplified to identify peptide sequences and regenerate the bead library for subsequent selection rounds.

DNA constructs used in this study. Gray shading indicates the variable regions. T7prom, T7 promoter; RBS, ribosome binding site; IEGR, recognition sequence for factor Xa protease; GGS5, (Gly-Gly-Ser)5 spacer; SrtAmut, activity-enhanced mutant of sortase A; T7term, T7 terminator; XYZ8, eight codons synthesized via MLSDS. Arrows denote primer binding sites used to distinguish between constructs encoding HA tag and His tag. Although the IEGR sequence is included, factor Xa cleavage was not performed.

DNA immobilization and peptide display on beads via PL display. FACS analysis of bead samples prepared through the following steps: emulsion PCR using an HA-tag gene construct, hybridization with Cy5-modified single-stranded DNA, PL-display procedure, and staining with PE-conjugated anti-HA-tag antibody. A) Confirmation of DNA immobilization. Cy5 fluorescence dot plots of beads prepared using primers either lacking (left) or containing (right) a pentaglycine modification. In both cases, ∼10% of beads were Cy5 positive, indicating successful DNA immobilization through hybridization of Cy5-labeled DNA to bead-bound templates. B) Confirmation of HA-tag peptide display. PE fluorescence histogram gated on Cy5-positive beads. Strong PE signals were observed only in the sample prepared with pentaglycine-modified primers, confirming peptide display via sortase-mediated transpeptidation. The sample lacking pentaglycine served as a negative control, showing minimal PE signal despite Cy5 positivity, thereby confirming the requirement of pentaglycine for peptide display.

Selective enrichment of target-binding peptide genes from mixed gene pools. A) FACS histograms and dot plots of bead samples generated by PL display using a 1:1 mixture of HA- and His-tag gene constructs. Beads were incubated with PE-conjugated anti-HA-tag antibody (top) or anti-His-tag antibody (bottom). Two distinct populations were observed and distinguished by log-normal distribution fitting. Populations presumed to display HA- and His-tag peptides are indicated by the solid line and the dashed line, respectively. Boxed regions indicate the gating boundaries used for FACS sorting. B) Electrophoresis results of PCR products from DNA recovered from the sorted beads shown in (A). Reference lanes containing PCR products derived from HA- and His-tag gene constructs were included to indicate the expected band positions for each tag type. These serve as size markers and confirm the identity of the enriched gene products observed in the preselection lane and antibody-selected lanes. Although the initial gene ratio was 1:1, PCR products from antibody-bound bead populations showed selective enrichment of the peptide gene corresponding to the antibody used. C) FACS histograms and dot plots of bead samples prepared by PL display using a 1:10,000 mixture of HA- and His-tag gene constructs, following incubation with PE-conjugated anti-HA-tag antibody. A log-normal distribution was fitted to the fluorescence data. Boxed regions indicate gates used for FACS sorting. D) Electrophoresis results of PCR products from DNA recovered from beads sorted as shown in (C). Reference lanes containing PCR products derived from HA- and His-tag gene constructs were included, serving as size marker controls. “Pre” and “Post” lanes indicate samples before and after FACS sorting, respectively. Despite the initial low abundance of the HA-tag gene construct (1:10,000 relative to the His tag), it became predominant in the PCR products after selection with anti-HA-tag antibody. Fluorescence intensity data were collected from pre-FACS sorting measurements of several thousand beads from the same samples, as FACS sorting itself does not record fluorescence values.

Selection of target-binding peptides from a randomized library. A) FACS histograms and dot blots of bead samples incubated with PE-labeled anti-HA-tag antibody across successive selection rounds. The top, middle, and bottom panels correspond to the initial library, post-round 1, and post-round 2 samples, respectively. Fluorescence gating thresholds are indicated by lines, and beads exhibiting fluorescence above these thresholds were collected. After two rounds of selection, the proportion of beads exceeding the fluorescence threshold markedly increased. Fluorescence intensity data were obtained from pre-FACS sorting measurements of 20,000 beads per sample, as FACS sorting itself does not record fluorescence values. B) Eight-amino-acid randomized regions of the top 20 most frequent peptide sequences identified after rounds 1 and 2. The HA-tag sequence (YPYDVPDY) is shown as a reference, representing the epitope recognized by the anti-HA-tag antibody used for selection. Read counts for each sequence are indicated. Total sequencing reads were 248,731 (round 1) and 277,901 (round 2). Identical amino acid sequences derived from distinct nucleotide sequences are listed separately. C) Sequence logo representing the consensus motif derived from the top 16 peptide sequences obtained after the second selection round.

Full text

Introduction

Protein display technology is a powerful approach for identifying target-binding proteins, such as antibodies and peptides. It involves constructing a library of polypeptides physically linked to their encoding nucleic acids, thereby enabling the selection of genes that encode proteins with specific target interactions. This technology is widely applied in antibody discovery, protein interaction analysis, and proteomics research (1, 2).

Phage display, the earliest form of protein display, uses bacteriophages to link phenotype and genotype (3). Although phage display has yielded numerous achievements (4), it is limited by its inability to present cytotoxic proteins, restricted library size due to transformation efficiency, and expression bias resulting from uneven bacterial growth (1, 2). To address these issues, cell-free display methods such as ribosome display (5), mRNA display (6, 7), and cDNA display (8–10) have been developed. These systems overcome the key limitations of bacterial-based methods and support the incorporation of non-natural amino acids (11) and posttranslational modifications (8).

Both phage and cell-free display techniques rely on bulk affinity evaluation, in which all variants are screened simultaneously in a single tube. Although this enables ultra-high-throughput screening, it lacks precise control over selection thresholds due to washing steps with limited control. Consequently, gradual enrichment over multiple rounds is required, and higher-affinity variants tend to dominate, making it challenging to isolate moderate-affinity peptides. Nevertheless, moderate-affinity peptides have often been reported to play physiologically important roles (12–14), and their selective isolation requires not only recovery of intermediate binders but also exclusion of high-affinity variants.

In contrast, cell-surface display, protein variants are presented on living cells, such as yeast (15), bacteria (16), or mammalian cells (17), allowing quantitative assessment of individual binding affinities via fluorescence-activated cell sorting (FACS). This single-variant resolution minimizes inter-variant competition, preventing the loss of high-affinity variants and the co-fractionation of low-affinity variants. In addition, protein variants, including those with low or moderate affinities, can be selectively isolated by adjusting the FACS sorting gates (12). However, the practical library size is limited to ∼10^9^ due to cellular volume, and the method shares other limitations with phage display due to its reliance on living cells.

Bead-surface display methods that use cell-free expression systems have been developed to overcome these constraints. These approaches colocalize genotypic DNA and phenotypic protein on a bead by encapsulating them in shared emulsion compartments. In the earliest report of this method, beads preloaded with antibodies and DNA captured cell-free-expressed proteins via antibody binding, enabling genotype–phenotype linkage and ∼1,000-fold enrichment per round (18). Subsequent methods employed alternative strategies for protein immobilization within emulsions (19–22). Among these, covalent linkage between DNA and protein demonstrated superior performance (21). This approach employs SNAP display, in which DNA and protein are covalently linked via the reaction between benzylguanine and the SNAP tag, enabling selection under harsh conditions such as surfactants, organic solvents, or extreme pH (23).

Protein tags such as SNAP tag (∼20 kDa) and Halo tag (∼33 kDa) (24) enable conjugation of fusion proteins to DNA via small-molecule ligands; however, they leave large residual fragments. To minimize this, peptide ligation enzymes, such as sortase, asparaginyl endopeptidases, transglutaminase, split intein, SpyLigase, and SnoopLigase, have been used (25). Among these, sortase is the most widely used. Sortase recognizes the LPXTG motif (X: any amino acid), cleaves threonine and glycine, and ligates the threonine to an N-terminal glycine-bearing molecule. Engineered variants of sortase have shown improved performance (26, 27) and are compatible with cell-free systems (28).

In this study, we developed a bead-surface display platform that employs sortase-catalyzed peptide ligation to covalently link genotypic DNA and phenotypic protein variants via a minimal peptide linker, all within a fully cell-free system. The platform enabled >10,000-fold enrichment of target-binding peptides per round and achieved efficient selection from a randomized library of 1.7 × 10^6^ variants within two rounds of FACS selection, demonstrating its potential for rapid and controllable cell-free selection of functional binders.

Results

Preparation procedure of peptide ligase–mediated display

The workflow of peptide ligase–mediated display (PL display) is shown in Fig. 1A. The DNA constructs encoded fusion proteins comprising a peptide variant, an LPETG sortase substrate, a flexible spacer, and a mutant sortase A (SrtA_mut_). SrtAmut is an activity-enhanced variant of Staphylococcus aureus sortase A lacking the N-terminal 59 residues and carrying the substitutions P94S, D160N, D165A, and K196T (29). Each DNA molecule and PCR primer-coated magnetic bead was co-encapsulated in water-in-oil (w/o) emulsions containing PCR reagents and pentaglycine-modified primers, following a Poisson distribution. The emulsions were generated using a microfluidics-free pipette-based method that was optimized for reproducible droplet formation (Fig. S4). After encapsulation, PCR was performed to generate pentaglycine-modified double-stranded DNA (dsDNA) on the bead surfaces. Following demulsification, beads bearing pentaglycine-modified dsDNA were re-encapsulated with a cell-free expression system (PUREfrex 1.0, GeneFrontier). The expressed fusion proteins underwent intramolecular sortase-mediated transpeptidation (30), covalently linking the peptide variant to the pentaglycine moieties immobilized on the beads via dsDNA within the emulsion. After demulsification, a bead library was obtained in which each phenotypic protein variant was covalently linked to its corresponding genotypic DNA via a minimal peptide linker and immobilized on the bead surface.

Schematic representation of target-binding peptide selection by PL display. A) Preparation of a peptide-displaying bead library using PL display. An initial DNA library is compartmentalized into emulsion droplets, each containing a single DNA molecule encoding a peptide variant, a capture primer-coated bead, and PCR primers, including a pentaglycine-conjugated primer. PCR is performed in each droplet to generate DNA products that encode peptide variants and carry a terminal pentaglycine motif. These DNA molecules hybridize to the capture primers on the beads and are subsequently extended to generate a DNA-immobilized bead library. After demulsification, the beads are re-encapsulated in emulsions for cell-free translation of peptide–sortase fusion proteins. Sortase-mediated transpeptidation transfers each peptide variant to the pentaglycine motif on its corresponding bead. The second demulsification step yields a peptide-displaying bead library, in which each peptide is covalently linked to its encoding DNA via a minimal peptide linker. B) Selection of target-binding peptide by PL display. Following emulsion PCR, Cy5-modified oligonucleotides are hybridized to the DNA on beads to identify those carrying the amplified sequences. Peptide variants are displayed on the beads through cell-free translation and sortase-mediated conjugation. The beads are then incubated with PE-conjugated antibodies, which serve as model targets for peptide binding. Beads displaying target-binding peptides are isolated by FACS based on PE fluorescence. DNA from the sorted beads is amplified to identify peptide sequences and regenerate the bead library for subsequent selection rounds.

Validation of peptide display on beads

To assess peptide display, hemagglutinin (HA)-tag peptides were expressed on DNA-bearing beads and detected using dual fluorescence labeling. The DNA constructs used for display are shown in Fig. 2. Cyanine5 (Cy5)-labeled oligonucleotides were hybridized to the single-stranded 5′ region of the pentaglycine-modified PCR primer, which remained unpaired after amplification (Fig. 1B). The displayed HA tag was stained with phycoerythrin (PE)-conjugated anti-HA antibodies and analyzed using flow cytometry. Cy5-positive beads, indicating successful DNA amplification, accounted for ∼10% of the total beads, consistent with the Poisson distribution estimates. This proportion was similar in control samples lacking the pentaglycine modification (Fig. 3A). However, PE fluorescence was markedly higher in the presence of pentaglycine, confirming transpeptidation-mediated HA-tag display (Fig. 3B).

DNA constructs used in this study. Gray shading indicates the variable regions. T7prom, T7 promoter; RBS, ribosome binding site; IEGR, recognition sequence for factor Xa protease; GGS5, (Gly-Gly-Ser)5 spacer; SrtAmut, activity-enhanced mutant of sortase A; T7term, T7 terminator; XYZ8, eight codons synthesized via MLSDS. Arrows denote primer binding sites used to distinguish between constructs encoding HA tag and His tag. Although the IEGR sequence is included, factor Xa cleavage was not performed.

DNA immobilization and peptide display on beads via PL display. FACS analysis of bead samples prepared through the following steps: emulsion PCR using an HA-tag gene construct, hybridization with Cy5-modified single-stranded DNA, PL-display procedure, and staining with PE-conjugated anti-HA-tag antibody. A) Confirmation of DNA immobilization. Cy5 fluorescence dot plots of beads prepared using primers either lacking (left) or containing (right) a pentaglycine modification. In both cases, ∼10% of beads were Cy5 positive, indicating successful DNA immobilization through hybridization of Cy5-labeled DNA to bead-bound templates. B) Confirmation of HA-tag peptide display. PE fluorescence histogram gated on Cy5-positive beads. Strong PE signals were observed only in the sample prepared with pentaglycine-modified primers, confirming peptide display via sortase-mediated transpeptidation. The sample lacking pentaglycine served as a negative control, showing minimal PE signal despite Cy5 positivity, thereby confirming the requirement of pentaglycine for peptide display.

PE-positive beads were isolated using FACS and subjected to single-particle quantitative PCR, which revealed ∼54,400 copies of immobilized dsDNA per bead (Fig. S5). The number of peptides displayed per bead was then estimated by measuring the enzymatic activity of horseradish peroxidase conjugated to the anti-HA antibodies and was found to be ∼29,700 molecules, corresponding to a display efficiency of 55% (Fig. S6).

FACS-based PL-display selection for gene-specific enrichment

To evaluate the gene-separation capability of PL display, we performed FACS-based selection (Fig. 1B) using a DNA library containing HA- and His-tag constructs mixed at equimolar ratios. The constructs were designed to yield PCR amplicons of distinguishable lengths, allowing relative quantification using electrophoresis (Fig. 2).

After cell-free expression and PL display in the emulsion droplets, the beads were stained with PE-conjugated anti-HA or anti-His antibodies. Approximately half of the beads exhibited high fluorescence with either antibody, confirming that each bead displayed only one tag (Fig. 4A). Beads with high fluorescence were sorted by FACS, and the immobilized DNA was amplified by PCR. Electrophoresis revealed two distinct bands in the preselection sample, corresponding to both genes, whereas only the target gene was detected after selection (Fig. 4B), indicating effective tag-specific separation.

Selective enrichment of target-binding peptide genes from mixed gene pools. A) FACS histograms and dot plots of bead samples generated by PL display using a 1:1 mixture of HA- and His-tag gene constructs. Beads were incubated with PE-conjugated anti-HA-tag antibody (top) or anti-His-tag antibody (bottom). Two distinct populations were observed and distinguished by log-normal distribution fitting. Populations presumed to display HA- and His-tag peptides are indicated by the solid line and the dashed line, respectively. Boxed regions indicate the gating boundaries used for FACS sorting. B) Electrophoresis results of PCR products from DNA recovered from the sorted beads shown in (A). Reference lanes containing PCR products derived from HA- and His-tag gene constructs were included to indicate the expected band positions for each tag type. These serve as size markers and confirm the identity of the enriched gene products observed in the preselection lane and antibody-selected lanes. Although the initial gene ratio was 1:1, PCR products from antibody-bound bead populations showed selective enrichment of the peptide gene corresponding to the antibody used. C) FACS histograms and dot plots of bead samples prepared by PL display using a 1:10,000 mixture of HA- and His-tag gene constructs, following incubation with PE-conjugated anti-HA-tag antibody. A log-normal distribution was fitted to the fluorescence data. Boxed regions indicate gates used for FACS sorting. D) Electrophoresis results of PCR products from DNA recovered from beads sorted as shown in (C). Reference lanes containing PCR products derived from HA- and His-tag gene constructs were included, serving as size marker controls. “Pre” and “Post” lanes indicate samples before and after FACS sorting, respectively. Despite the initial low abundance of the HA-tag gene construct (1:10,000 relative to the His tag), it became predominant in the PCR products after selection with anti-HA-tag antibody. Fluorescence intensity data were collected from pre-FACS sorting measurements of several thousand beads from the same samples, as FACS sorting itself does not record fluorescence values.

To assess enrichment under practical conditions, we next attempted to isolate a rare gene present at a 1:10,000 ratio. A mixture containing 3 × 10^7^ His- and 3 × 10^3^ HA-tag constructs was subjected to PL display. Beads were stained with PE-conjugated anti-HA antibodies, and those exceeding a predefined fluorescence threshold, excluding His-tag-displaying beads, were collected by FACS (Fig. 4C). Of the 5.9 × 10^6^ beads analyzed, 8% were Cy5 positive, and 47 beads (0.01% of the Cy5-positive population) displayed the HA-tag peptide. Among the 47 HA-tag-displaying beads, four were sorted within a predefined gating threshold. This count closely matched the predicted value of 7.8, which was estimated from a log-normal distribution constructed using previously measured fluorescence intensities. The DNA from the sorted beads was then amplified and analyzed by electrophoresis (Fig. 4D). Before selection, only the His-tag band was visible; after selection, it was replaced with the HA-tag band. These results demonstrate that PL display combined with FACS enables ≥10,000-fold gene enrichment in a single round when stringent gating is applied.

Proof-of-concept screening of a model epitope peptide from a randomized peptide gene library

The feasibility of PL display–based affinity screening was evaluated through enrichment of an epitope peptide from a randomized library composed of eight residues consisting of the primitive amino acids A/D/V/P/S and the target-binding-prone amino acid Y, with a sequence diversity of 1.7 × 10^6^ (Figs. 2 and S2). This corresponds to approximately six copies per variant when evaluating 1 × 10^7^ beads, near the practical upper limit for FACS screening. This library also included the HA-tag epitope peptide (YPYDVPDY), which was used as a model in this study. As a performance benchmark, epitope tag enrichment experiments are frequently employed in related library screening systems (18, 31, 32). A random peptide library with equal frequencies of the six amino acids was prepared using multi-line split DNA synthesis (MLSDS) (33), as described in SI Materials and Methods.

Emulsion PCR was performed using the library, and ∼1.2 × 10^7^ DNA-positive beads were isolated by FACS, as guided by the Cy5-labeled probes. These beads were subjected to cell-free expression in emulsion to generate a peptide-displaying bead library. The beads library was stained with PE-conjugated anti-HA-tag antibody, and the top ∼0.2% exceeding a predefined fluorescence threshold were collected by FACS (Fig. 5A). DNA from the selected beads was then amplified and used to regenerate peptide-displaying beads via PL display. Upon restaining, 0.2% of beads again exceeded the same fluorescence threshold, indicating no significant change from the first round. However, after the second round of selection, the proportion of high-fluorescence beads increased to 3.7% at the same threshold (Fig. 5A).

Selection of target-binding peptides from a randomized library. A) FACS histograms and dot blots of bead samples incubated with PE-labeled anti-HA-tag antibody across successive selection rounds. The top, middle, and bottom panels correspond to the initial library, post-round 1, and post-round 2 samples, respectively. Fluorescence gating thresholds are indicated by lines, and beads exhibiting fluorescence above these thresholds were collected. After two rounds of selection, the proportion of beads exceeding the fluorescence threshold markedly increased. Fluorescence intensity data were obtained from pre-FACS sorting measurements of 20,000 beads per sample, as FACS sorting itself does not record fluorescence values. B) Eight-amino-acid randomized regions of the top 20 most frequent peptide sequences identified after rounds 1 and 2. The HA-tag sequence (YPYDVPDY) is shown as a reference, representing the epitope recognized by the anti-HA-tag antibody used for selection. Read counts for each sequence are indicated. Total sequencing reads were 248,731 (round 1) and 277,901 (round 2). Identical amino acid sequences derived from distinct nucleotide sequences are listed separately. C) Sequence logo representing the consensus motif derived from the top 16 peptide sequences obtained after the second selection round.

Next-generation sequencing revealed no dominant sequences after the first round; however, enrichment of HA-tag-homologous sequences was observed after the second round (Fig. 5B). The top 16 sequences converged on the consensus motif (Y/A)DVPD(Y/A) and accounted for 5.6% of the total reads (277,901), consistent with the 3.7% high-fluorescence bead fraction. These sequences showed an average enrichment of 135-fold in the second round (ranging from 41- to 686-fold) and an estimated average enrichment of 65-fold in the first round (ranging from 7- to 128-fold) (Table S1).

Discussion

In this study, we developed a bead-surface protein display system that covalently links genotypic DNA and phenotypic protein via a minimal peptide linker in a fully cell-free format. This architecture enables the tunable selection of target-binding peptides using FACS under diverse binding conditions. Bulk selection methods, such as phage display and other cell-free systems, evaluate variant mixtures collectively and typically yield modest enrichment rates, ranging from a few to several thousand-fold (31, 32, 34, 35). In contrast, PL display enables single-variant resolution and allows precise modulation of selection stringency via gate-defined FACS parameters. As demonstrated in this study, this approach enabled gene enrichment exceeding 10,000-fold, representing a substantial advancement relative to previously reported cell-surface- and particle-based display systems that have laid the foundation for high-throughput selection technologies (18, 21, 29).

In the experiment targeting rare epitope-encoding genes from a random peptide library, the gate settings were adjusted to prioritize collection efficiency over target purity, resulting in moderate enrichment as a trade-off. The top 16 most abundant gene variants after round 2, all corresponding to consensus sequences, exhibited enrichment factors ranging from 7- to 128-fold in round 1 and from 41- to 686-fold in round 2, with up to ∼20-fold differences among individual variants (Table S1). Next-generation sequencing analysis of the initial library identified 57 consensus nucleotide variants among 257,306 unique sequences and 54 consensus amino acid variants among 243,497 unique sequences, with frequencies of 0.02215 and 0.02218%, respectively, consistent with theoretical expectations. Notably, 99.81% of the nucleotide sequences and 94.63% of the amino acid sequences appeared without duplications, indicating minimal bias in variant distribution. Over the two selection rounds, 16 of the 432 possible consensus amino acid variants were predominantly enriched, whereas a broader set of sequences progressively accumulated, suggesting that different variants exhibited distinct enrichment patterns across successive selection rounds (Table S2). In particular, some consensus sequences showed limited or delayed accumulation. Such cases compromise the advantage of strict selection stringency inherent to PL display. The variability in enrichment among the consensus sequences likely stems from bead-to-bead differences in DNA and protein immobilization, deviations from the expected variant identity per bead, and stochastic fluctuations in bead–target interactions. These sources of heterogeneity can be reduced through workflow refinements, which are expected to improve the reliability of affinity-based variant recovery, thereby making enrichment ratios more faithfully reflect underlying binding affinities.

In this study, emulsion PCR was used to immobilize multiple copies of DNA on each bead (36). Real-time PCR quantified an average of 5.4 × 10^4^ ± 1.7 × 10^4^ copies per bead (Fig. S5), with variability attributed to the emulsion droplet size. Although microfluidic devices produce uniform droplets (37), they often cause bead sedimentation and uneven dispersion (38, 39). We refined the conventional method involving the gradual addition of the PCR solution to the stirred oil mixture (19, 20) and optimized the addition rate, stirring speed, and oil temperature. These refinements were primarily aimed at equalizing emulsion droplet size, which is expected to suppress variability in DNA immobilization on beads and to improve the correlation between binding affinity and the selection outcomes, such as variant recovery and enrichment ratios. Although the droplet size variation was largely controlled, the emulsion still exhibited a log-normal distribution with modal diameters of 4.0 µm (asymmetric spread: −0.6 to +1.6 µm, by droplet count) and 8.6 µm (asymmetric spread: −0.6 to +1.7 µm, by occupied volume), respectively. The latter directly influenced the Poisson loading probability (Fig. S4 and SI Materials and Methods). To infuse the bead-containing PCR solution, we used conventional plastic pipette tips with an orifice diameter of ∼400 µm, owing to their widespread availability and ease of handling. However, as previously reported (40), tips with narrower diameters may enable more precise emulsion generation, suggesting a potential avenue for further optimization.

The resolution of FACS-based selection in this study was constrained by relatively low fluorescence signals, which is likely due to suboptimal protein display levels on the beads. Based on qPCR and enzymatic measurements, the peptide display efficiency on the bead was estimated to be ∼55% relative to the amount of immobilized DNA. This indicates the incomplete occupancy of the binding sites and suggests that the absolute amount of the displayed peptide is not solely determined by the bead-surface area. Even with PUREfrex 2.0, which has a higher protein synthesis capacity than PUREfrex 1.0, the protein display did not appreciably increase, implying that synthesis capacity is not the limiting factor. Sortase A is typically Ca^2+^ dependent but retains partial activity in the presence of Mg^2+^ (41). Consistent with this, PL display was indeed successfully achieved in a Ca^2+^-free cell-free system. However, the low-display efficiency may partly stem from the absence of Ca^2+^ and could be improved by employing Ca^2+^-independent sortase variants (42, 43). Additionally, immobilizing protein-capturing molecules on bead surfaces may enhance the display levels, as previously reported (21).

To address the potential range of displayable peptide lengths, we note that green fluorescent protein (GFP, 238 amino acids), together with an upstream Myc tag and a downstream IEGR sequence, resulting in a total of 254 amino acids, was successfully displayed on beads via sortase-mediated conjugation, as confirmed by its fluorescence (Fig. S7). This result indicates that our platform can accommodate polypeptides of at least this length. While the ability of a protein to fold correctly and exhibit function does not depend solely on chain length, the GFP experiment provides clear evidence that proteins of ∼250 amino acids can be displayed using this system.

The trade-off between bulk selection using large libraries, which affords only limited control through washing conditions, and FACS-based selection using smaller libraries, which enables precise and tunable control, remains a topic of interest. Although bulk selection can yield high-affinity binders, its efficiency is theoretically limited by Kd-dependent competition, potentially excluding low-abundance, high-affinity variants. In contrast, FACS enables the individual assessment of each variant on bead surfaces, avoiding this limitation (44, 45).

Because FACS-based selection is inherently limited in the number of variants that can be individually evaluated, strategies to overcome this constraint are essential when working with highly diverse libraries (e.g. >10^9^ variants). A common approach is to perform bulk preselection followed by FACS-based refinement; however, this carries the risk of eliminating rare but functional variants. Alternatively, amino acid–restricted libraries can be designed to contain moderately functional sequences with high probability, thereby enabling direct FACS-based selection and subsequent diversification (46, 47). Although codon-specific phosphoramidites offer precise control over amino acid composition (48), their high cost limits practical application. A more feasible strategy involves the degenerate codons (49, 50), albeit with limited precision. In this study, we employed the MLSDS method, which integrates split DNA synthesis with a genetic algorithm-optimized phosphoramidite mixture to flexibly define amino acid species and frequencies (33). This approach yielded high-quality libraries that improved the efficiency of PL display–based selection.

Although our enrichment experiments targeted epitope sequences, PL display is not limited to binding molecules. FACS-based screening of enzymatic activity has been demonstrated using bead and cell-surface display systems (22, 29, 51–53). With its cell-free format and covalent, minimal genotype–phenotype linkage, PL display is considered applicable to enzymatic activity screening approaches designed to leverage these features.

In summary, we developed a PL-display system, which is a fully cell-free bead-surface display system that covalently links genotypic DNA and phenotypic protein via a minimal peptide linker. It enables tunable FACS selection with gate-defined stringency control at the single-variant level, and its simple architecture supports a flexible screening design. These features establish PL display as a robust and tunable platform for the discovery of functional molecules.

Materials and methods

Construct Preparation

The DNA constructs used in this study are schematically shown in Fig. 2, and full sequences are provided in Fig. S1. Constructs were designed to encode peptides for display, followed in order by the LPETG sortase substrate, a flexible (Gly-Gly-Ser)5 spacer, and the sortase A mutant (SrtAmut). SrtAmut is an activity-enhanced mutant of S. aureus sortase A lacking the N-terminal 59 residues and carrying the substitutions P94S, D160N, D165A, and K196T (29). A peptide gene library restricted to six amino acids (A, D, V, P, S, and Y) with uniform composition was synthesized using the MLSDS method (33). Further details are provided in SI Materials and Methods.

Primer preparation

A pentaglycine-modified primer was prepared by conjugating a DNA primer bearing a 5'-terminal azide group with a pentaglycine peptide bearing a C-terminal alkyne group. The DNA primer contains a region complementary to a sulfo-Cy5-modified oligonucleotide for fluorescent labeling. This labeling enables identification of DNA-positive beads by Cy5 fluorescence, as illustrated in Fig. 1B. Detailed procedures are provided in SI Materials and Methods.

Preparation of a library of gene-coding DNA-immobilized beads

A library of gene-coding DNA-immobilized beads was prepared based on a previously reported method (36), with slight modifications. Dual-biotinylated primers were immobilized onto streptavidin-coated magnetic beads and used for emulsion PCR with the template DNA–containing genes encoding the peptides to be displayed and the associated elements. The template and bead concentrations were optimized according to the Poisson distribution to ensure that over 95% of the DNA-positive beads harbored a single variant, comprising 8–10% of the total population. After thermal cycling, demulsification was performed, and the beads were recovered and washed. The beads bearing amplified DNA were fluorescently labeled by hybridization with a sulfo-Cy5-modified oligonucleotide targeting the single-stranded region derived from the pentaglycine-conjugated primer. For the selection using a random peptide gene library, Cy5-positive beads bearing amplified DNA were enriched by FACS prior to peptide display on beads via cell-free translation and PL display; otherwise, all beads were used directly. Detailed procedures are provided in SI Materials and Methods.

Preparation of a peptide-DNA-immobilized beads library

A library of peptide-displaying DNA-immobilized beads was constructed using a w/o emulsion-based cell-free expression system. Beads bearing gene-coding DNA were mixed with the reaction mixture (PUREfrex 1.0) and dispersed in stirred oil to form emulsions. Peptide expression and the PL display proceeded at 37 °C for 3 h. Bead concentrations were optimized according to the Poisson distribution to ensure that over 95% of the beads display a single peptide variant. After the reaction, demulsification was performed, and the beads were recovered and washed. Detailed procedures are described in SI Materials and Methods.

FACS-based affinity selection

Affinity selection of the peptide-displaying DNA-immobilized beads was performed using FACS. Beads were incubated with PE-conjugated anti-HA-tag or anti-His-tag antibodies, washed, and analyzed by flow cytometry. Fluorescently labeled beads were then sorted based on antibody-dependent fluorescence intensity, and the collected beads were subjected to PCR amplification. The resulting DNA was purified and used as a library for the next selection round. Fluorescence intensity thresholds were defined as follows: in mixtures of HA- and His-tag genes at 1:1 or 1:10,000 ratios, thresholds were set to minimize contamination from nonbinding beads; for selection from random peptide libraries, a predetermined intensity threshold corresponding to the HA-tag population was used to maximize recovery. To evaluate the relative abundance of HA- and His-tag coding constructs in the samples before and after selection, bead mixtures from pre- and postselection were used as templates to amplify gene-coding regions of distinct lengths corresponding to each construct, and the resulting products were separated and analyzed by microchip electrophoresis. Detailed procedures are described in SI Materials and Methods.

Sequence analysis of the peptide gene libraries

A 94-bp region, spanning from just after the start codon to the GGS spacer downstream of the random codon region, was amplified from peptide gene libraries prepared before and after selection using a primer set compatible with Illumina sequencing. Paired-end sequencing was then performed using an Illumina MiSeq platform at FASMAC. Detailed procedures are provided in SI Materials and Methods.

Supplementary Material

pgag031_Supplementary_Data