Click display: a rapid and efficient in vitro protein display method for directed evolution

Abstract We describe a novel method for in vitro protein display—click display—that does not depend on maintaining RNA integrity during biopanning and yields covalently linked protein–cDNA complexes from double-stranded input DNA within 2 h. The display is achieved in a one-pot format encompassing transcription, translation and reverse transcription reactions in series. Stable linkage between proteins and the encoding cDNA is mediated by a modified DNA linker—ML—generated via a click chemistry reaction between a puromycin-containing oligo and a cDNA synthesis primer. Biopanning of a click-displayed mock library coupled with next-generation sequencing analysis revealed >600-fold enrichment of target binders within a single round of panning. A synthetic library of Designed Ankyrin Repeat Proteins (DARPins) with ∼1012 individual members was generated using click display in a 25-μl reaction and six rounds of library panning against a model protein yielded a panel of nanomolar binders. This study establishes click display as a powerful tool for protein binder discovery/engineering and provides a convenient platform for in vitro biopanning selection even in RNase-rich environments such as on whole cells.


INTRODUCTION
Rational design of proteins with desirable properties is challenging due to the complexity of a protein's structurefunction r elationship.Dir ected evolution, which mimics natural evolution and entails the screening of large libraries of variants to identify those with desired functions, offers a powerful alternati v e to rational protein design as it does not r equir e a priori knowledge of a protein's structurefunction r elationship.A pr er equisite for dir ected evolution is a link between the protein and its coding DN A / RN A and the choice of display method dictates the size / di v ersity of the protein library accessible for discovery.Over the y ears, many cell-based (e.g.y east / bacterial / phage-display) and cell-free ( in vitro ) protein display methods have been de v eloped.Since cell-free protein display methods are typically able to access a library size (10 12 -10 14 ) orders of magnitude larger than cell-based methods (10 6 -10 9 ), in vitro protein display methods are often pr eferr ed for drug de v elopment.
e89 Nucleic Acids Research, 2023, Vol. 51, No. 16 PAGE 2 OF 11 One of the earliest cell-free display technolo gies, mRN A display, takes advantage of the ability of puromycin to enter the ribosomal A site and attach to the nascent peptide.Specificall y, a single-stranded DN A ( ∼30 nt) harboring a puromycin molecule at its 3 end is ligated to the coding mRNA to enable the formation of protein-mRNA conjuga tes upon transla tion ( 1 , 2 ).This method, for the first time, enabled cell-free protein engineering and greatly expanded the library size accessible to directed ev olution ( 1 , 3 ).Ho we v er, major drawbacks of this system are the poor ligation efficiency, making the method highly labor-intensi v e, and dependence of RNase-free conditions to preserve RNA integrity.
Ribosome display, another in vitro protein display technology, utilizes a spacer sequence and a stem loop structure at the 3 end of the mRNA to 'trap' the ribosome, thus bypassing the need for mRN A-DN A ligation ( 4 ).Upon translation, this spacer sequence remains attached to the peptidyl tRNA and occupies the ribosomal tunnel, allowing the protein of interest to protrude out of the ribosome.Although ribosome display is procedurally simpler than mRNA-display (5)(6)(7)(8)(9)(10)(11), its dependence on mRNA and ribosomes for display confers significant limitations ( 12 ): (i) since the ribosome is a large protein-RNA complex, unpredictable interactions between the ribosome and the displayed protein can bias the selection of binder proteins ( 13 ); (ii) the peptidyl-tRN A / mRN A / ribosome complex is intrinsically unstable, requiring the selection steps to be perf ormed bef or e r e v erse transcription (RT), under a high magnesium concentration and at low temperatures (e.g.4˚C) ( 4 , 7 , 14 ), potentially fostering the selection of variants with restricted biological relevance; (iii) a strictly RNasefree environment is necessary to preserve the integrity of the mRNA during panning, making the success of ribosome display experiments highly sensitive to even small differences in experimental practices and reagents / consumables.
TRAP technology (TRanscription-translation coupled with Association of Puromycin linker) is another in vitro display approach de v eloped by Murakami et al. that couples mRNA tr anscription, protein tr anslation and re v erse transcription to dramatically reduce the time needed for protein display ( 12 ).In TRAP display, a single-stranded DNA linker functionalized with a puromycin molecule at its 3 end is hybridized to the mRNA in the tr anscription / tr anslation r eaction mixtur e that lacks Release Factor 1 (RF1).The mRNA contains a UAG STOP codon immediately preceding the mRN A / DN A hybridization region w hich stalls the ribosome and enables the puromycin to be incorporated into the nascent protein.The display efficiency was later increased by using 2 -Ome-modified DNA linker to suppress promoter-independent transcription and unwanted DN A / RN A duplex es ( 15 ).Most r ecently, Murakami and co-w ork ers created a branched DNA to covalently link the nascent protein with the re v erse transcribed first strand cDNA to obviate the need for RNase-free conditions during biopanning ( 16 , 17 ).Although conceptually elegant and simple, it is a non-trivial task to synthesize the branched DN A linker w hich is not commerciall y available at the time of this writing and suffers from very low synthesis yield ( < 0.5-4%) ( 18 ).
In this study, we describe a simple method for in vitro protein display --termed click display --that employs only r eagents that ar e r eadily available from commer cial sour ces and produces covalently linked pr otein-cDNA pr oduct from coding DNA in a one-pot reaction within 2 h.Covalent linkage of the protein variants and their encoding cDNA molecules is achie v ed using a modified DNA linker --ML --generated by copper-free click reaction between two oligos (Figure 1 A): oligo #1 contains a puromycin molecule at the 3 end and a click chemistry moiety (i.e.dibenzocyclooctyne (DBCO)) at the 5 end; oligo #2 is complementary to the 3 end of the mRNA and harbors a matching click chemistry moiety (i.e.azide (N 3 )) at its 5 end.The highly efficient and specific click chemistry reaction between DBCO and N 3 enables the modified oligo ML to be produced with very high efficiency under physiological conditions ( 19 ), obviating the need for purification.Oligo #2 also contains a 3-cyanovinylcarbzaole nucleotide (CnvK, red line in Figure 1 A), enabling ML to be covalently crosslinked to the mRNA with nearly 100% efficiency upon brief exposure to long wavelength UV ( > 360 nm) ( 20 , 21 ).Both oligos #1 and #2 are commercially available and the crosslinked mRNA-ML can be efficiently produced using a lo w-cost UV lamp, allo wing click display to be readily performed in any molecular biology lab.
Although in principle any in vitro protein synthesis mixture can be used for click display, we have optimized the reaction based on the PURExpress In V itr o Protein Synthesis Kit (NEB) ( 22 ) which contains all the components needed for in vitro transcription and transla tion.Templa te DNA harboring a T7 promoter and ribosome binding site (RBS) is added directly to the in vitro protein synthesis reaction mixture (1 g DNA per 25 l reaction) alongside ML (final 2 M, Figure 1 B).Transcription, ML-mRNA crosslinking, translation, and re v erse transcription are carried out sequentially in a one-pot reaction to yield nascent protein-cDNA complexes that can be used directly for biopanning (Figure 1 C).Synthesis of protein-cDNA complex from the input DNA template is complete within 2 h (Table 1 ).
The ability of click-display ed libr aries to support the selection of rare binders to a desired target protein was demonstrated using two complementary approaches.First, a click-displayed mock designed ankyrin repeat (DARPin) library containing 0.01% of a known enhanced green fluorescent protein (eGFP)-binding DARPin --D 3G86 -- ( 23 ) was panned against biotinylated eGFP.This approach re v ealed that 625-fold enrichment of the eGFP-binding DARPin was achie v ed in a single-round of selection, representing a 6-12-fold enhancement in enrichment efficiency ov er conv entional mRNA-display ( 24 ).Ne xt, we displayed a na ïve DARPin library with ∼10 12 variants in a 25l click-display reaction and used it to identify a panel of binders to a model target protein --clumping factor A (ClfA)( 25 ) --following six rounds of panning.Se v eral of the selected ClfA binders showed nanomolar target binding affinity as determined by ELISA (Figure 7 A).

DN A / mRN A synthesis and purification
The template DNA sequence used for the click display of D 3G86 is shown in Supplementary Figure S3.For optimizing the reaction condition using mRNA-ML, this DNA template was PCR-amplified using primers 1938 and 2643   S1) and use as the template for mRNA synthesis (1 g per 20 l reaction, HiScribe ™ T7 High-Yield RNA Synthesis Kit (New England Biolabs, E2040S)).At the end of the reaction, 70 l nuclease-free water, 10 l 10 × DNase I Buffer, and 2 l DNase I (2 U / l, New England Biolabs, M0303S) were added and the mixture was incubated at 37 • C for another 15 min to degrade the template DNA.The synthesized mRNA was purified using EZNA MicroElute RNA Cleanup Kit (VWR, 101319-190), eluted in nuclease-free water and stored at -80 • C in aliquots until use.
To construct a na ïve DARPin library for click display, synthetic oligos AR-F2 and AR-R2 (Supplementary Table S1, final 1.6 M), encoding all the variable r egions, wer e in-cubated in 1 × Isothermal Amplification Buffer (New England Biolabs, B0537S) in the presence of MgSO 4 (final 6 mM) and dNTP (final 1.4 mM each) at 94 • C f or 2 min f ollo wed by slo w cooling (-0.1 • C / s) to 65 • C before the addition of Bst2.0 (final 320 U / ml).The mixture was incubated at 65 • C for ∼30 min to generate double-stranded DARPin variable region (DVR).The synthesized DNA product was purified from agarose gel.The N-cap and C-cap region of D 3G86 were amplified by PCR using primers 2617 / 2701 and 2702 / 2581, respecti v ely, digested with BsaI and ligated to BsaI-digested DVR.The final ligation product (700 bp) was loaded on an agarose gel, and, after gel purification, 650 ng DNA product was r ecover ed.This amount of DNA corresponds to 1.4  and Taq polymerase in a total of 3.2 ml of PCR reaction to preserve the library diversity.The final DNA product was purified again on agarose gel and used for click display (1 g per 25 l reaction).

Synthesis of ML and mRNA-ML
Oligo #1 and #2 (Supplementary Table S1) were customsynthesized by IDT (Integrated DN A Technolo gies, Inc., IA) and resuspended in DNA duplex buffer (Integrated DN A Technolo gies, Inc., 11-01-03-01) to achie v e a final concentration of 0.5 mM.An equal volume of oligo #1 and oligo #2 wer e mix ed in a thin-wall PCR tube (Fisher Scientific, 14230225) and incuba ted a t room temperature in the dark.The reaction product --ML --was diluted 10-fold in DNA duplex buffer and stored at -80 • C until use.To estimate the reaction efficiency, an aliquot of the reaction mix was removed at 10 min, 1 day or 2 days, diluted 10-fold immedia tely in wa ter (final 25 M of each oligo) and stored at -20 • C until analysis on a 15% Urea-PAGE gel (Nov e x, EC6885BOX) and visualized under UV exposure both before and after staining with SYBR Green dye.mRNA-ML was pr epar ed by incubating purified mRNA (final 1-8 M) with ML (1:1.2 molar ratio) in DNA duplex buffer in the presence of RNasin Plus RNase inhibitor (final 1 U / l, Promega, PRN2611) in a thin-wall PCR tube.To maximize the hybridiza tion ef ficiency between mRNA and ML, the mixture was heated at 94 • C for 1 min, slowly cooled to 70 • C at -0.1 • C / s, incuba ted a t 70 • C for 5 min and finall y slowl y cooled to 25 • C at the same rate and held at 25 • C for an additional 15 min.After hybridization, the mixture was irradiated under a nail spa UV lamp (SUNUV Gel Nail Light, Model SUNONEWT, Amazon) for 30-90 s.The samples were resuspended in 2xRNA buffer (TBE-Urea Sample Buffer 2 ×, G-Biosciences, 786-474) and incuba ted a t 70 • C for 10 min to dena ture the mRNA.The crosslinking efficiency was analyzed using 15% and 6% urea-PAGE gels (2 pmol mRNA per lane, Nov e x ™ TBEurea gels, EC6885BOX, EC6865BOX) in warm 1 × TBE buffer (55 • C) under a constant voltage of 180 and 100 V, respecti v ely.

Click display translation optimization
To compare the amount of click displayed product (Figure 3 A), different amount of input mRNA-ML was added to the translation reaction mixture (New England Biolabs, PURExpress in vitro protein synthesis, E6800).The mixture was incuba ted a t 37 • C for 30 min with shaking followed by 20 min incubation at room temperature without shaking.After translation, RNase A (final 1 g / l, Omega Bio-tek, AC118) was added to degrade the mRNA, and the mixture was incubated at 37 • C for another 30 min.For gel analysis, the samples were first diluted 2-fold in 2 × gel shift buffer (50%(v / v) formamide, 125 mM Tris-HCl pH 6.8, 10 mM DTT, 20 mM EDTA and 2% (w / v) SDS) ( 15 ), heat inactiva ted a t 95 • C for 5 min, and then analyzed on 12% SDS-PAGE gels (Bio-Rad, Cat# 4561043).After electrophoresis, the gel was visualized under UV exposure without staining.
To evaluate the click display efficiency from DNA template (Figure 3 B), 1 g (600 nM) template DNA encoding D 3G86 , T7 promoter and terminator (Supplementary Figure S3) was added directly to 1 reaction volume of PURExpress in vitro protein synthesis reaction together with 2 or 3 M of ML.The reaction mixture was incubated at 37 • C for 20 min to allow for mRNA transcription before the addition of BsaI-HF (final 1.6 U / l, New England Biolabs, R3535) to cleave the template DNA and terminate the transcription.The mixture was then incubated at 37 • C for 10 min without shaking, for DNA digestion, and 30 min with vigorous shaking for protein translation, and was finally incubated for another 20-30 min at room temperature to allow for protein transfer.RNase A (Omega Bio-Tek, AC118, final 1 g / l) was then added to degrade the mRNA and the mixture was diluted 2-fold in 2 × gel shift buffer before being analyzed on 12% SDS-PAGE gels (Bio-Rad, 4561043).After electrophoresis, the gels were visualized under UV directly without staining.

Click display reverse transcription condition optimization
Click display was carried out as described above using mRNA-ML (final 0.5 M, 25 l reaction) as the template.After translation, EDTA (final 30 mM, ThermoFisher Scientific, AM9260G) was added to the reaction to inactivate the ribosome ( 15 ), followed by MgCl 2 (300 mM × 3 l) to neutralize the excess EDTA.The reaction volume was then expanded 3-5-fold before the addition of dNTP (final 0.5mM), SuperScript II (SS-II, final 2.7 U / l, Invitrogen) or SuperScript IV (SS-IV, final 2.7 U / l, Invitrogen) reverse transcriptase and the appropriate reaction buffer.The mixture was incubated at 42 • C (for SS-II) or 50 • C (for SS-IV) for 30 min, diluted 100-fold in ddH2O and the amount of the synthesized cDNA was quantified using Forget-Me-Not ™ EvaGreen ® qPCR Master Mix (Biotium, 31045) with primers 1892 and 2488 (Supplementary Table S1) in a qPCR instrument (Bio-Rad, C1000 Touch ™ Thermal Cycler and CFX 96 Real-Time PCR Detection Systems).After the reaction, the PCR product was also analyzed on 1% agarose gels to ensure the amplification of the correct target DNA.

Protein expression and purification
ClfA, eGFP (amino acid sequences in Supplementary Figure S3) and all the DARPin molecules were expressed in Esc heric hia coli Bl21 (DE3) cells in LB medium and purified using gravity Ni-NTA agarose beads following the standard pr ocedure.Pr otein purity was determined using 12% SDS-PAGE gels.

Strepta vidin pull-do wn assay
eGFP and ClfA were biotinylated using EZ-Link Sulfo-NHS-LC-Biotin (1:4 molar ratio, Fisher Scientific, PI21335).Click-displayed D 3G86 was diluted 10-fold in blocking buffer SBTD0.3 (StartingBlock ™ Blocking Buffer (Thermo Fisher Scientific, 37578) supplemented with 0.05% Tween-20, and 0.3 mg / ml ssDNA (deoxyribonucleic acid sodium salt from salmon testes, Sigma, D1626)) in the absence or presence of 200 nM biotinylated eGFP.The mixture was incubated at room temperature for 1 h before the addition of streptavidin-coated magnetic beads (Dynabeads ™ MyOne ™ Streptavidin T1, Thermo Fisher Scientific, 65-601) to pull-down the biotinylated eGFP and the associated cDNA.The beads were washed 8 times with SBTD0.3.After the final wash, the beads were resuspended 1x Taq ThermoPol ® Reaction Buffer (New England Bio Labs, B9004S) and the amount of cDNA associated with the beads was quantified using Forget-Me-Not qPCR kit with primers 1892 and 2488.

Library enrichment
To estimate the enrichment efficiency, 1 g of DNA encoding a na ïve N3C DARPin library was mixed with 0.1 ng of DNA encoding D 3G86 and used as template for click display (Table 1 ).After reverse transcription, the mixture was purified by one-step IMAC and incubated with 200 nM biotinyla ted eGFP a t room tempera tur e for 1 hour.Str eptavidin T1 or neutravidin-coated magnetic beads (Cytiva SpeedBeads Magnetic Neutravidin Coated Particles, Fisher Scientific, 09 981 154) were added to the mixture to pull down biotinylated eGFP and the associated click-displayed product.The beads were washed 3-8 times for rounds 1 to 3.After the final wash, the beads were resuspended in 1x Standard Taq buffer and the cDNA was amplified by PCR using Taq polymerase (NEB, M0267S) and primers 1938 and 2643, which contain the T7 promoter and terminator r egion, r especti v ely.The PCR product was used as the template for the subsequent round of panning after gel purification.
To enrich binders to ClfA, click-displayed na ïve DARPin library made from 1 g of DNA was used.This library was purified by IMAC and incubated with biotinylated ClfA (final 200 nM).ClfA-binders were pulled down using streptavidin-or neutravidin-coated beads in alternati v e rounds.

Next-generation sequencing (NGS)
NGS samples wer e pr epar ed by amplifying the PCR product from different rounds of panning (R0-R3) using NGS primers containing the Illumina adapter sequences ( Supplementary Table S1).The different rounds were distinguished using a two-nucleotide barcode inserted between the Illumina adapter and the complementary region of the primers.The selected cDNA from each round was amplified with its respecti v e primer pair using Q5 High-Fidelity Polymerase (New England Biolabs, M0491).Each round was then purified using the ZymoClean Gel DNA Recovery Kit (Zymo Research), after which 150 ng of the DNA from each round was combined.Samples were sequenced using a Mi-Seq Next-Generation Sequencing system from Amplicon-EZ (Azenta Life Sciences).P air ed-end r eads wer e merged using FLASH ( 27 ).Sequences were converted to FASTA forma t, transla ted, and analyzed using SeqKit ( 28 ).

Model generation
The 3D model of D 1F8 was generated using AlphaFold 2.2.0 ( 30 , 31 ) at Texas A&M High Performance Research Computing (HPRC) facility.The fiv e lowest ranked models were nearly identical, and the lowest ranked model was used in this study.

Synthesis of ML and mRNA-ML
The puromycin-containing linker ML was produced by reacting oligo #1 and oligo #2 (Figure 2 A).Oligo #1 contains a fluorescently labeled nucleotide (TAMRA) to facilitate the subsequent gel analysis.Oligo #2 is complementary to the mRNA and can be crosslinked to the mRNA via the cnvK (K) nucleotide upon UV irradiation.These oligos wer e mix ed a t an equimolar ra tio (250 M each oligo) in DNA duplex buffer (IDT) and incuba ted a t room temperature.After only 10 min of incubation, nearly all the input oligos have reacted to yield ML (Figure 2 B, Supplementary Figure S1).The reaction efficiency is estimated to be > 90% by densitometry analysis.To simplify the calculation of the ML input amounts, for subsequent steps, we assumed 100% ML synthesis efficiency.
Next, we confirmed the ability of ML to be specifically crosslinked to mRNA.Purified mRNA (final 1 M), whose 3 end is complementary to oligo #2 (Figure 2 A), was mixed with ML at 1:1.2 molar ratio in DNA duplex buffer in a thin-walled PCR tube (20 l total).The mixture was heated to 94 • C for 1 min to denature the mRNA and then slowly cooled to 25  Oligo #1 and oligo #2 were mixed in at an equimolar ratio (250 M of each oligo) and incubated at room temperature for 10 min.100 pmol of each oligo, either alone or following the 10-min click reaction, was analyzed on a 15% urea-PAGE gel that was first visualized under UV, and then stained with SYBR Green and visualized again.Only oligo #1 and ML are visible under UV without dye staining.( C ) ML can be efficiently crosslinked to mRNA.After hybridization, the mixtures of mRNA and ML were irradiated under UV for 30, 60 or 90s before being analyzed on a 6% urea-PAGE gel.Only ML-crosslinked mRNA is visible on unstained gels.Due to its small size, ML is not visible on the 6% Urea-PAGE gel but can be seen on a 15% urea-PAGE gel (Supplementary Figure S2).and analyzed on Urea-PAGE gels.Over 70% of the input mRNA was crosslinked to ML after 90 seconds UV irradiation as determined by densitometric analysis, with free ML being nearl y completel y depleted (Figure 2 C, Supplementary Figure S2).Although the crosslinking efficiency was marginally higher with 90 s UV irradiation, to maximally preserve the mRNA integrity, 30 seconds irradiation time was selected for all subsequent experiments.To simplify the calculation of mRNA-ML input amounts for subsequent steps, we assumed that all the input mRNA is converted to mRNA-ML.

Optimization of conditions for protein display
To demonstrate the ability for protein display, mRNA-ML was used as template for protein translation using the PUR-Express In V itr o Protein Synthesis Kit (NEB).Considering that each PURExpress in vitro protein synthesis reaction contains 2.4 M of ribosome (60 pmol per 25 l reaction), and that ribosomes in association with the mRNA are not acti v el y released due to mRN A-ML crosslinking at the cnvK site, we compared the yield of protein display with 0.5, 0.75 and 1 M mRNA-ML.The protein translation reaction was carried out at 37 • C for 30 min followed by another 20 min incubation at room temperature to facilitate the transfer of the nascent protein to puromycin ( 32 ).Due to the large size of mRNA, it is difficult to distinguish free mRNA and mRNA crosslinked to ML and protein.To simplify the gel analysis, RNase A was added to the reaction mixture to degrade the single-stranded mRNA.Similar amounts of protein-ML were observed in all reactions, suggesting that the display efficiency is likely sa tura ted a t 0.5 M mRNA-ML (Figure 3 A).Since mRNA in duplex with ML is partially protected from RNase A digestion, both ML and partiall y digested mRN A-ML (p.mRN A-ML) are visible on the gel.In the reaction containing 1 M mRNA-ML, a large amount of p.mRNA-ML is also visible on the gel, indica ting tha t excess of mRNA-ML is present a t this condition.Densitometry analysis showed that the intensity of the band corresponding to protein-ML in the 0.5 M lane is ∼20% of the total ML input, corresponding to 1.5 × 10 12 mRNA molecules per 25 l reaction (0.5 M × 25 l × 20% × N A , N A = 6 × 10 23 molecules / mol).This value r epr esents the upper limit of the library di v ersity per 25 l reaction.
In order to avoid handling purified mRN A w hich is highly sensiti v e to RNase contamination, we next optimized the reaction conditions to enable DNA to be used directly as the template for click display.DNA template (2 pmol per 25 l reaction or 80 nM) and ML (final 2 and 3 M) wer e added dir ectly to the in vitro protein synthesis reaction and the transcription reaction was carried out at 37 • C for 20 min before being UV irradiated for 30 s. Excess ML was used in this reaction to compensate for the anticipated reduced hybridization efficiency between ML and nascent mRNA due to the lack of the heat denaturation and slow cooling steps.Since mRNA synthesized after UV irradiation will not be crosslinked to ML but can potentially compete with the crosslinked mRNA-ML for ribosomes in the reaction, the restriction enzyme BsaI, which cleaves the DNA template immediately after the T7 promoter (Supplementary Figure S3), was added to the mixture to terminate the transcription reaction (Figure 3 C).The mixture was then incubated at 37 • C for 40 min to allow both BsaI digestion and protein translation, and then treated with RNase A before electrophoresis.The amount of click-displayed protein appeared to be similar regardless of the amount of the ML (Figure 3 B) and is comparable to the control reaction using 0.5 M mRNA-ML as the templa te, indica ting tha t DNA templa te can be used directly for click display without compromising the protein display efficiency.A large ML band was visible on the gel, especially in the reaction containing 3 M ML, suggesting that the concentration of ML is likely in excess at this condition.

Optimization of conditions for reverse transcription
We next optimized the final component of the click display reaction series --re v erse transcription.To avoid mistaking the input DNA template for the synthesized cDNA, click display was carried out using mRNA-ML (final 0.5 M) as the template in these studies.At the conclusion of translation, EDTA (final 30 mM) was added to inactivate the ribosomes ( 15 ) followed by an equimolar amount of MgCl 2 to neutralize the EDTA.The mixture was then diluted 3-fold before the re v erse transcription reaction.We compared the cDNA synthesis efficiency of SuperScript II (SS-II) and SuperScript IV (SS-IV) w hose optim um reaction temperatur es ar e 42 • C and 50 • C, respecti v ely.The amount of the synthesized cDNA was quantified using qPCR.To ensure that the measured cycle threshold (Ct) values correspond to the amplification of the target DNA, the r eaction mixtur e after qPCR was also analyzed via electrophoresis.The amplified PCR product was present only in samples incubated in the presence of re v erse transcriptase (RT) (Figure 4 ), confirming the synthesis of cDNA under these conditions.Similar amounts of cDNA were present in both reactions, with SS-IV offers a slight advantage in the synthesis efficiency (slightly lower cycle threshold (Ct) value).

Evaluation of target binding ability of a click-displa y ed protein
The final one-pot tr anscription, tr anslation, and re v erse transcription reaction protocol is summarized in Table 1 .
To evaluate the efficiency of click display, we displayed DARPin 3G86 (D 3G86 ), w hich strongl y binds eGFP ( K D ∼ 9 nM) ( 23 ), and determined its ability to be pulled down by biotinylated eGFP in a streptavidin pull-down assay (Figure 5 A).Briefly, click-displayed D 3G86 was diluted 10-fold in blocking buffer with or without biotinylated eGFP (final 200 nM).After 30 min incubation at room temperatur e, str eptavidin-coated magnetic beads were added to the mixture to pull-down biotinylated eGFP and the associated click-displayed D 3G86 .
A much larger amount of cDNA was pulled-down in samples incubated in the presence of eGFP than that in buffer alone, as manifested in the significantly lower cycle threshold (Ct) values in quantitati v e PCR (qPCR, Figure 5 B) analysis, confirming that click displayed D 3G86 is acti v e and retains the ability to associate with eGFP.To ensure tha t the amplifica tion is specific, the qPCR product was analyzed on agarose gels.A faint band of the amplified cDNA was visible in the sample incubated in the absence of eGFP, likely resulting from non-specific association of the parental template DNA with the beads.
To further reduce the background signal, we exploited the 6xHis tag at the N-terminus of D 3G86 and purified the click-displayed D 3G86 product via one-step immobilized metal affinity chromato gra phy (IMAC) prior to incubation with eGFP or buffer.IMAC purification of the input clickdisplayed protein significantly reduced the background signal resulting in further elevated Ct values for samples incuba ted in buf fer alone and the reduction of PCR product on agarose gels (Figure 5 B).Although the Ct values for samples incubated in the presence of eGFP also increased slightly, the Ct between samples incubated with and without eGFP increased significantly from 5.6 (original) to 11.7 (after IMAC purification).This result indicates that IMAC purifica tion can ef ficiently reduce the background signal from click display.The dsDNA template present in the click reaction is likely responsible for the background signal.

Mock library enrichment
To evaluate the efficiency of click display to support the selection of rare target protein binders from a large library, we carried out a mock library enrichment experiment.DNA encoding the eGFP-binding DARPin D 3G86 (0.1 ng) was mixed with a na ïve DARPin library (1 g) at 1:10 000 molar ratio and used as the template for click display.The click displayed product was purified via IMAC and underwent standard panning / selection against biotinylated eGFP (final 200 nM).Strepta vidin and neutra vidin beads were used to capture the eGFP-binding DARPin(s) in alternati v e rounds.To quantify the amount of D 3G86 present in the output from different rounds of selection, approximately 150 ng of the final PCR product from each round was sequenced with a Mi-Seq Next-Generation Sequencing system yielding a dataset with 149768 full-length protein sequences.The frequency of D 3G86 increased from 0.0024% in R0 to 1.5% in R1 and peaked at 17.5% in R2 (Figure 6 ).The enrichment efficiency from R0 to R1 is estimated to be 625fold (1.5% / 0.0024%), which is 6-12-fold higher than conv entional mRNA-display ( 24 ).Some what une xpectedly, the fr equency of D 3G86 decr eased in R3 to 9.6%.We suspect this phenomenon may be due to the selection of other eGFPbinding DARPins from the library that may out-compete D 3G86 for eGFP binding.Ne v ertheless, these results confirm the ability of click display to ra pidl y enrich binders to a desired target protein.

Discovery of DARPin binders to a model protein using click display
Next, we used a click-displayed synthetic DARPin library to pan for binders to a model target protein --clumping factor A (ClfA), an S. aureus surface virulence protein ( 25 , 33 ).A click-display ed DARPin libr ary pr epar ed from 1 g of DNA templa te ( ∼10 12 dif ferent molecules) was purified via IMAC and underwent standard selection against biotinylated ClfA (final 200 nM).After fiv e rounds of panning, significantl y more DARPin-cDN A complexes were pulled down by streptavidin beads incubated in the presence of the target protein ClfA, which is manifested as a higher Ct (Supplementary Figure S4A).The PCR product from Round 6 was subcloned into the pET28a expression vector and used to transform BL21(DE3) cells.One hundred and eighty-eight (two 96-well plates minus 4 control wells per plate) clones were picked and grown in LB.The next day, the cell lysates after protein induction were harvested and their ability to bind ClfA was determined via ELISA.Overall, > 50% of the clones showed > 3 standard deviations above the background binding to ClfA, with > 30% showing > 10 standard deviations above the background binding (Supplementary Figure S4).
Se v en clones that showed strong ClfA-binding when applied as crude cell lysates were purified via IMAC and re-evaluated for binding to ClfA via ELISA (Figure 7 A,  B).The sequences of these clones were found to be distinct although they cluster into two large families, with DARPin 1E5 (D 1E5 ) adopting a different set of targetbinding residues than all the other clones.Clones D 1E10 , D 2D2 , D 2G3 and D 2E7 share the identical target-binding residues but contain different point mutations in the scaffold (Figure 7 C, D, Supplementary Figure S5).The very different ClfA-binding ability of D 1E10 and D 2E7 highlights the influence of frame wor k residues on target binding.To e valuate the di v ersity, the PCR product from R6 was subjected to next-generation sequencing (NGS).We obtained a total of 27026 full-length DARPin sequences, among which D 2G3 was the most abundant, r epr esenting 627 r eads (2.3%, Supplementary Figure S6).This result shows that significant poly clonal di v ersity remains in the click-display-selected pools e v en after 6 rounds of panning.Since the di v ersity of the na ïve library (estimated to be ∼10 12 ) greatly exceeds the capacity of NGS, we did not sequence this initial library.
The successful enrichment of ClfA-binding DARPins, as well as the preservation of high library di v ersity after 6 rounds of enrichment, underscores the power of click displa y f or protein binder discovery.

DISCUSSION
Current in vitro protein display methods are complex, laborintensi v e, and often r equir e strictly RNase-free conditions for purification and / or library panning, putting them outof-reach for many r esear chers.For example, RNA is used to link the displayed protein to its coding sequence in both mRNA-display and ribosome-display, limiting their applications to highly purified and RNase-free targets ( 1 , 4 , 7 , 14 ).Although a recently de v eloped in vitro display method, TRAP display, enables simultaneous mRNA transcription and protein translation, it r equir es a branched oligo that is not commercially available (Integrated DN A Technolo gies, Inc. does not offer synthesis of the branched oligos at the time of this manuscript preparation) and whose synthesis is non-trivial and inefficient ( < 0.5% to 4%,) ( 18 ), se v erely limiting the utility of this approach.
In this study, we de v eloped an alternati v e in vitro protein display technology --click display --that enables the production of co valently link ed protein-cDNA complex from double-stranded input DNA in a one-pot reaction within 2 h.A modified oligo linker (ML) with two 3 ends --one functionalized with puromycin and the other a primer for cDNA synthesis --was generated using copper-free click chemistry and was used to link the coding cDNA and the displayed protein.The constituents of ML, Oligo #1 and Oligo #2, ar e commer cially availa ble ( e .g. IDT) and no purification step is needed for ML synthesis.The estimated cost of ML is ∼$ 4 per 25 l in vitro protein synthesis reaction.Since the protein is co valently link ed to the cDNA, there is no need to maintain RNase-free conditions during biopanning,

Figure 1 .
Figure 1.Click display of proteins on their encoding cDNA molecules.( A ) Design of the modified linker ML. ( B , C ) Ov ervie w of click display steps.Purified DNA template (1 g / 25 l reaction) is added directly to the in vitro protein synthesis reaction together with ML (final 2 M).In vitro transcription, ML-mRNA crosslinking, in vitro translation, and re v erse transcription reactions are carried out sequentially in a single tube to yield click-displayed protein-cDNA complexes.Black and green arrows represent single stranded DNA and mRNA, respecti v ely.All arrows point to the 3 end.

Table 1 .•
Ov ervie w of one-pot click display of proteins to their encoding cDNA Incubate at 37 • C for 20 min for transcription • UV irradiation (30 s) BsaI-HF v2 (20 U / l) 2 l • Incubate at 37 • C for 10 min without shaking.• Incubate at 37 • C for 30 min with shaking for translation.• Incubate at room temperature for 20 min without shaking to promote protein transfer.EDTA (300 mM) 3 l 30 mM • Incubate at room temperature for 5 mins to inactivate the ribosome.MgCl 2 Expand vol 3-5-fold • Incubate at 42 or 50 • C for 30 min (Supplementary Table

Figure 2 .
Figure 2. ML synthesis by copper-free click chemistry.( A ) Design of oligo #1 and oligo #2. ( B ) ML can be efficiently synthesized under ambient conditions.Oligo #1 and oligo #2 were mixed in at an equimolar ratio (250 M of each oligo) and incubated at room temperature for 10 min.100 pmol of each oligo, either alone or following the 10-min click reaction, was analyzed on a 15% urea-PAGE gel that was first visualized under UV, and then stained with SYBR Green and visualized again.Only oligo #1 and ML are visible under UV without dye staining.( C ) ML can be efficiently crosslinked to mRNA.After hybridization, the mixtures of mRNA and ML were irradiated under UV for 30, 60 or 90s before being analyzed on a 6% urea-PAGE gel.Only ML-crosslinked mRNA is visible on unstained gels.Due to its small size, ML is not visible on the 6% Urea-PAGE gel but can be seen on a 15% urea-PAGE gel (Supplementary FigureS2).

Figure 3 .
Figure 3. Protein display.( A ) Comparison of protein display efficiency of mRNA-ML at different concentrations.The product was analyzed on 12% SDS-PAGE gels and visualized under UV without staining.Only ML (containing the fluorescent oligo #1) and protein co valently link ed to ML are visible on the gel.p.mRN A-ML: partiall y digested mRN A-ML.( B ) Comparison of the protein display efficiency of reactions containing 1 g of DN A template and 2 or 3 M of ML to that with 0.5 M of mRNA-ML.( C ) Schematic for terminating the transcription reaction.Red triangle r epr esents the BsaI recognition site.Black and green arrows indicate single stranded DNA and mRNA, respecti v ely.

Figure 4 .
Figure 4. Confirmation of one-pot re v erse transcription reaction.RT: Reverse Transcriptase.SS-II: SuperScript II; SS-IV: SuperScript IV.Ct: cycle threshold.The qPCR product was further analyzed on a 1% agarose gel.

e89 16 PAGE 8 OF 11 Figure 5 .
Figure 5. Pull-down of click-displayed anti-GFP DARPin D 3G86 .( A ) Scheme of the pull-down experiment.Click displayed D 3G86 -cDNA were incubated with streptavidin beads either in the presence or absence of the biotinylated target GFP.( B ) qPCR quantification of either the crude (unpurified) or IMAC-purified D 3G86 -cDNA complex pulled down by the str eptavidin beads.One r epr esentati v e e xperiment with the mean value of two technical replicates ± standard error is shown.Agarose gel electrophoresis of the PCR products from one of the reactions is shown beneath the corresponding bar graph.

Figure 6 .
Figure 6.The frequency of D 3G86 sequence in each round during the mock enrichment.

Figure 7 .
Figure 7. Characterization of ClfA-binding DARPin clones.ClfA-binding affinity measurement by ELISA ( A ) and SDS-PAGE ( B ) of IMAC-purified DARPins selected from six rounds of panning on ClfA.Data points are the mean ± s.d.from one of two r epr esentati v e e xperiments carried out in duplica te. ( C ) Illustra tion of the sequences in the variable residues of the different DARPins.( D ) AlphaFold2-predicted structure of D 1F8 colored based on each residue's pLDDT value.Variable residues comprising the binding interface (randomized during library creation) are shown in stick form.The pLDDT values (indication of prediction confidence) for residues within the variable loops are much lower than those within the scaffold.