key: cord-0051372-jzx5u7s5 authors: Grindel, Brian J.; Engel, Brian J.; Hall, Carolyn G.; Kelderhouse, Lindsay E.; Lucci, Anthony; Zacharias, Niki M.; Takahashi, Terry T.; Millward, Steven W. title: Mammalian Expression and In Situ Biotinylation of Extracellular Protein Targets for Directed Evolution date: 2020-09-22 journal: ACS Omega DOI: 10.1021/acsomega.0c03990 sha: 2c0b470e7162e5a27895f5f60fc425f7a031e410 doc_id: 51372 cord_uid: jzx5u7s5 [Image: see text] Directed evolution is a powerful tool for the selection of functional ligands from molecular libraries. Extracellular domains (ECDs) of cell surface receptors are common selection targets for therapeutic and imaging agent development. Unfortunately, these proteins are often post-translationally modified and are therefore unsuitable for expression in bacterial systems. Directional immobilization of these targets is further hampered by the absence of biorthogonal groups for site-specific chemical conjugation. We have developed a nonadherent mammalian expression system for rapid, high-yield expression of biotinylated ECDs. ECDs from EGFR, HER2, and HER3 were site-specifically biotinylated in situ and recovered from the cell culture supernatant with yields of up to 10 mg/L at >90% purity. Biotinylated ECDs also contained a protease cleavage site for rapid and selective release of the ECD after immobilization on avidin/streptavidin resins and library binding. A model mRNA display selection round was carried out against the HER2 ECD with the HER2 affibody expressed as an mRNA–protein fusion. HER2 affibody–mRNA fusions were selectively released by thrombin and quantitative PCR revealed substantial improvements in the enrichment of functional affibody–mRNA fusions relative to direct PCR amplification of the resin-bound target. This methodology allows rapid purification of high-quality targets for directed evolution and selective elution of functional sequences at the conclusion of each selection round. Efficient directed evolution screens and selections are critical for the development of novel peptide-and protein-based affinity ligands, therapeutics, and imaging agents. 1,2 Directed evolution of peptide, peptidomimetics, and protein libraries can be carried out using phage display, 3 bacterial display, 4 yeast display, 5 ribosome display, 6 and mRNA display. 7 These display technologies are based on the creation of a library of compounds directly linked to the genetic information used to encode them. In a typical selection experiment, libraries are panned against an immobilized target, subjected to washing to remove nonfunctional sequences, enriched for functional sequences in an elution step, and amplified to generate an enriched library for the next selection round. 8−12 This iterative screening strategy has been used to select high-affinity ligands for the development of novel imaging agents and therapeutics. 13 −18 The success of a selection experiment is dependent upon many parameters including library design, selection pressure stringency, and target presentation. In the latter case, failure to present the protein target in its folded state with correct posttranslational modifications may result in a final pool of ligands with poor affinity and selectivity for the biologically relevant form of the protein. Many directed evolution screens or selections have been carried out against the extracellular domains (ECDs) of cancer-associated receptor tyrosine kinases (RTKs) 11, 19 or immune checkpoint receptors. 20, 21 Unfortunately, many ECD targets are prohibitively expensive, are of unknown quality even if an ECD is commercially available, or challenging to express in quantities sufficient for selection and downstream validation experiments. Moreover, mammalian cell expression is preferred for ECD targets because they are chaperone-dependent, 22 disulfide-rich, 23 and often decorated with N-and O-glycosylation. 24, 25 These features make it very challenging to produce biologically relevant mammalian ECDs in bacterial expression systems. 26 Unfortunately, ECD expression in adherent mammalian cells requires transfection, stable clone selection, and cost-prohibitive cell factory production. 25 In addition to expression and purification, site-specific immobilization of ECD proteins remains challenging. Common nonspecific immobilization strategies include direct chemical immobilization on reactive resins by NHS, maleimide, and other conjugation chemistries. 27, 28 Alternatively, nonselective biotinylation can be similarly performed, followed by immobilization on avidin/streptavidin resins. Although this strategy is often the most facile and straightforward, nonspecific biotinylation may occlude potential binding surfaces and result in nonbiologically relevant target presentation. A third challenge in directed evolution experiments is the selective recovery of high-affinity library members that remain bound to the immobilized target protein. A variety of solutions have been proposed to address this problem including competitive elution with a known ligand, 29 the use of an "acid bump" elution, 30 the use of detergents (e.g., SDS 31 ), passive adsorption and recovery from immuno tubes, 32 and onresin amplification of the genetic component of the bound library. Although highly selective, elution with a competitor ligand biases selected sequences to those that share a common epitope with the competitor and may not efficiently elute the highest affinity library members. Moreover, in the case of a novel or poorly characterized target, a high-affinity competitor ligand may be unavailable. Elution with acid or detergent is a nonspecific process and may result in amplification of lowaffinity/selectivity sequences, sequences that bind the capture resin, and/or loss of library members that are not acid-elutable. Detergent-based elution also typically requires downstream processing to remove the detergent in order to perform genetic amplification. Failure to remove the detergent may dramatically reduce the efficiency of the subsequent amplification step. Protein conformation may also be altered via surface unfolding using passive adsorption, which may create unnatural epitopes and occlude desired targeting interfaces. 33 Similar to nonspecific elution, direct amplification of the resin-bound library is also likely to result in amplification of nonfunctional sequences, particularly if the library contains a significant amount of unfused genetic material or there is significant binding to the solid-phase resin itself. In contrast, selective elution of the immobilized target could result in the recovery of target-specific library members and eliminate the need for downstream cleanup procedures. To address these issues, we built upon previously described expression systems for production and site-specific in situ biotinylation of ECD proteins ( Figure 1 ) for use in directed evolution experiments. Several in situ biotinylation technologies have been reported, 34−36 and the expression system described in this manuscript adapts them to a nonadherent, commercially available, high-density mammalian cell culture. This technology allows production of properly folded and post-translationally modified ECDs of epidermal growth factor receptor family members (EGFR, HER2, and HER3) 37 in the cell culture supernatant. These proteins are enzymatically biotinylated at the C-terminus in situ during expression for directional immobilization on avidin/streptavidin resins. In contrast to other in situ biotinylation systems, 35 ECD and BirA biotinylation expression is split between two vectors allowing production of nonbiotinylated protein, if necessary. To enable selective elution of the resin-bound target, a protease cleavage site was installed immediately N-terminal to the biotinylation site. Biotinylated ECDs were found to be properly folded and could be expressed and purified in multimilligram quantities in less than 2 weeks. Mock selection with an affibody−mRNA fusion followed by protease elution demonstrated significant signal-to-noise enhancement relative to direct on-resin amplification and specific enrichment of the affibody over the scrambled counterpart. Cell Culture. 293-F cells (Life Technologies, A14527) were grown in suspension in a humidified incubator at 8% (v/v) CO 2 on a shaking platform at 125 rotation per minute (RPM) in Expi293 expression media (Life Technologies, A1435101). Polypropylene, 0.22 μm filter vent capped, nonbaffled flasks were used for the culture. Cells were diluted from a maximum density of 4 × 10 6 to 0.3 × 10 6 cells/mL for passaging. Protein expression was performed in these cells only up to passage 20. For frozen stocks, cells were centrifuged at 200 × g, gently resuspended in 10% (v/v) dimethyl sulfoxide in culturing media at 10 7 cells/mL, and frozen at a rate of −1°C/min to −80°C. Stocks were kept in liquid nitrogen for long-term storage. Thawed cells were only further cultured if immediate viability was near 100% according to trypan blue staining. Plasmids and Cloning. All mammalian expression system constructs used the cytomegalovirus promoter-driven pcDNA3.4 TOPO vector (Fisher, A14697). As outlined in Figure 1 , the vector includes a Kozak sequence, Gaussia princeps luciferase secretion signal peptide (SP) followed by the encoded ECD (either HER2/HER3/EGFR). The Cterminus includes an encoded thrombin (LVPRGS), TEV (ENLYFQG), or no protease site, followed by a BirA biotinylation site (GLNDIFEAQKIEWHE) and 6x polyhistidine tag (His-tag). Glycine-serine spacers are encoded between the ECD, protease site, BirA tag, and His-tag. In the DNA vector, flanking HindIII and XhoI restriction enzyme (RE) sites allow substitution of other ECDs. A NheI site follows the TEV/thrombin site to allow for protease site exchange. DNA encoding inserts for TOPO TA pcDNA3.4 cloning were purchased from GeneArt (Fisher) . Exchanging ECDs was performed using standard RE cloning and confirmed by Sanger sequencing. The vector for expression of secreted BirA-Flag enzyme was a gift from Gavin Wright (Addgene, 64395). 38 BirA for in vitro biotinylation was produced in Escherichia coli similar to previous methods. 39 BL21(DE3) cells were transformed with BirA-encoding plasmid and diluted into an overnight culture of Luria broth (LB). The culture was grown to ∼0.6 OD 600 and induced with 0.1 μM isopropyl-β-D-thiogalactoside (IPTG). After 3 h at 30°C, the culture was harvested by centrifugation. Cell pellets were resuspended in buffer A (25 mM Tris−HCl pH 8.0, 0.5 M NaCl, 0.1 mM dithiothreitol (DTT), 5% (v/v) glycerol, 20 mM imidazole) and lysed using a French press. After clarification by centrifugation, the supernatant was loaded onto a 5 mL Ni 2+ -NTA column (GE Healthcare) using DuoFlow fast protein liquid chromatography (Bio-Rad), washed with buffer A, and eluted with a linear gradient to 100% buffer B (25 mM Tris−HCl pH 8.0, 0.5 M NaCl, 0.1 mM DTT, 5% (v/v) glycerol, 400 mM imidazole). The protein was concentrated with a 10 kiloDalton (kDa) molecular weight cutoff (MWCO) filter (EMD Millipore, UFC901024) and buffer-exchanged into 50 mM imidazole pH 6.8, 50 mM NaCl, 5 mM β-mercaptoethanol, and 5% (v/v) glycerol. The concentration of BirA protein was adjusted to 1 mg/mL, frozen in liquid nitrogen, and stored at −80°C. Heregulin-β1 (Hrgβ1) DNA sequence (based on Uniprot Isoform 6, identifier Q02297-6; TSHLVKCAEKEKTFCVNG-GECFMVKDLSNPSRYLCKCPNEFTGDRCQNYVMAS-FYKHLGIEFMEAEELYQK) with flanking RE sites was ordered from IDT as a gBlock gene fragment. Using flanking BamHI and NotI sites, the sequence was inserted into a pQTEV bacterial expression vector, a gift from Konrad Buessow (Addgene, 34824). 40 Transfection and In Situ Biotinylation. Transient ExpiFectamine (Fisher, A14524) transfection of 293-F cells was performed according to manufacturer's directions. Overall transfection time, protocol deviation, and any additives (e.g., biotin) are noted in the results. For in situ biotinylation, 10% by weight of the BirA expression plasmid was added for cotransfection (e.g., 6 μg of BirA plasmid, 54 μg of ECD expressing plasmid for a 60 mL transfection). Mammalian Cell Protein Ourification and In Vitro Biotinylation. Following transient transfection and expression for 5 days (unless specified otherwise), the conditioned medium (CM) was centrifuged at 3500 × g for 10 min, supernatant recovered, and recentrifuged to remove cells. CM was passed through a 0.22 μm sterile polyethersulfone filter and treated with Pierce EDTA-free protease inhibitor tablets (A32965). CM was concentrated and partially bufferexchanged using an Amicon stirred-cell concentrator (EMD Millipore, UFSC20001) with a 30 kDa MWCO filter (EMD Millipore, PLTK06210) at 4°C. Concentrated CM was dialyzed into exchange buffer (25 mM HEPES pH 8.0, 200 mM NaCl, 5% (v/v) glycerol, 0.05% (w/v) sodium azide) with a 12−14 kDa MWCO membrane. Concentrated and dialyzed CM was subject to Ni 2+ -NTA chromatography to purify Histag-containing ECDs. After resin was washed with equilibration buffer (50 mM phosphate buffer pH 8.0, 300 mM NaCl, 5 mM imidazole, 0.05% (v/v) Tween 20, 0.05% (w/v) sodium azide), the CM was added and allowed to settle in a glass fritted column. Resin was washed with equilibration buffer followed by wash buffer (50 mM phosphate buffer pH 8.0, 300 mM NaCl, 10 mM imidazole, 0.05% (w/v) sodium azide). Resin-bound ECDs were released with elution buffer (50 mM phosphate buffer pH 8.0, 300 mM NaCl, 250 mM imidazole, 0.05% (w/v) sodium azide) and buffer-exchanged with Amicon 50 kDa MWCO centrifugal filters (Fisher, UFC905024) into exchange buffer. In vitro ECD biotinylation was performed with the recombinant BirA enzyme expressed in E. coli. ECD was diluted to 40 μM in biotinylation buffer (50 mM Tris−HCl, pH 8.3, 10 mM adenosine triphosphate, 10 mM Mg(OAc) 2 , and 50 μM d-biotin). BirA enzyme (2.5 μg) per 10 nmoles ECD was added and incubated overnight at 4°C. Following the reaction, the mixture was buffer-exchanged into exchange buffer with a centrifugal MWCO filter. Assaying In Situ Biotinylation Efficiency and Protease-Mediated Release. Both streptavidin (SA)-agarose (Pierce, 30249) and NeutrAvidin UltraLink resin (Pierce, 53150) were used to bind ECDs with similar results. To assess fraction biotinylated, SA resins were preblocked with 0.5% (w/ v) bovine serum albumin (BSA) in PBS (Corning, 21-031-CVR) for 30 min and washed with PBS. ECDs (in vitro biotinylated, in situ biotinylated, or not biotinylated) were incubated with an excess of resin for at least an hour at room temperature diluted in PBS. Centrifuged supernatant was collected as the unbound fraction. The resin was washed three times with PBS and resuspended in PBS. Unbound and bound fractions were volume-matched for direct comparison. SA resin-bound ECDs and unbound fractions were incubated in Laemmli sample buffer with β-mercaptoethanol at 95°C. Protein was separated on a 4−12% (w/v) acrylamide gel in tris-glycine SDS buffer. Gels were stained with Bio-Safe Coomassie (Pierce, 161-0786). Bands were analyzed by densitometry with ImageJ and percentage bound was calculated by comparing both input and eluted fractions to the unbound flow through. The mean of these two values are reported as the final percentage bound. To confirm protease release, the same procedure was followed but in tris-buffered saline (TBS, 50 mM Tris−HCl pH 7.5, 150 mM NaCl). Following ECD binding and washing, the resin was incubated for specific release in equal volume of TBS with recombinant human thrombin (>2800 units/mg protein, Sigma-Aldrich, SAE0006) or EZCut Tobacco Etch Virus (TEV) recombinant protease (>10,000 units/mg protein, BioVision, 7847-100). TEV protease incubation included 2.5 mM DTT in some experiments. Protease eluted fractions were collected and analyzed by SDS-PAGE as above to calculate percent released. Western blot was performed to verify protein identity. Following SDS-PAGE and 20% (v/v) methanol tris-glycine transfer to nitrocellulose, blots were blocked in 4% (w/v) BSA in TBS supplemented with 0.05% (v/v) Tween 20 (TBS-T) for 2 h at room temperature. Blots were incubated overnight at 4°C with blocking buffer diluted antibody (rabbit anti-HER2 (Cell Signaling Technologies, 4290S); mouse anti-HER3 clone 1B4C3 (BioLegend, 324710); mouse anti-EGFR clone H11 (Invitrogen, MA5-13070)) or SA-horse radish peroxidase (HRP, Pierce, 21130). After three times 5 min TBS-T washes, the blots were incubated with goat anti-rabbit or anti-mouse HRP (Azure Biosystems, AC2114/AC2115) for 2 h at room temperature, washed again, and developed with the chemiluminescent substrate (Pierce, 34080). Purification of Hrgβ1. Hrgβ1 expression plasmid was transformed into chemically competent One Shot BL21 (DE3) E. coli cells (Invitrogen, C600003). After selection and growth of a single clone in LB, expression of Hrgβ1 was induced by 40 μM IPTG (Teknova, I3305). After growth at 225 RPM for 3 h at 37°C, the centrifuged pellet was frozen at −80°C, thawed at 4°C, resuspended in lysis buffer (50 mM Tris HCl, pH 8.0, 500 mM NaCl, 5% (v/v) glycerol, 5 mM imidazole, 0.025% (v/v) Tween 20, 0.01% (w/v) sodium azide) with an EDTAfree protease inhibitor tablet, sonicated, and clarified at 19,000 × g for 60 min at 4°C. The pellet was further extracted with lysis buffer including 6 M guanidine-HCl. The solution was frozen at −80°C and then thawed rotating end-over-end at 4°C . Ni 2+ -NTA purification was used to bind the Hrgβ1 denatured lysate. To refold on the resin, bound protein was sequentially washed with lower amounts of guanidine-HCl (6 to 0 M) and higher amounts of imidazole (10 to 50 mM). After elution with 250 mM imidazole, recovered Hrgβ1 was buffer-exchanged into 25 mM HEPES pH 8.0, 300 mM NaCl, 10% (v/v) glycerol, and 0.01% (w/v) sodium azide. ECD-Ligand Enzyme-Linked Immunosorbent Assays (ELISAs). HER2 association with anti-HER2 affibody was assessed through an indirect ELISA. Biotinylated affibody (Abcam, ab31890) was immobilized on a preblocked SAcoated plate (Pierce, 15500) in assay buffer (TBS, pH 7.6, 0.1% (w/v) BSA, 0.05% (v/v) Tween 20). Plates were washed in assay buffer, and HER2-ECD was serially diluted over the SA-bound affibody. Plates were washed and incubated with a rabbit anti-HER2 antibody (Cell Signaling Technologies, 4290S). Following washing, goat anti-rabbit HRP was allowed to bind, washed, and incubated with the TMB (3,3′,5,5′tetramethylbenzidine) substrate (Pierce, 34028). Hrgβ1 association with biotinylated HER3-ECD was assessed by first directly adsorbing purified Hrgβ1 to a Nunc-Immuno MaxiSorp plate (Sigma-Aldrich, M5785) diluted in PBS overnight at 4°C. The surface was blocked with TBS with 2% (w/v) BSA, washed in TBS-T, and incubated with serially diluted HER3-biotin (diluted in TBS, 0.5% (w/v) BSA). After washing, SA-HRP (Pierce, 21130) was incubated, washed, and allowed to react with TMB. To measure the binding of EGFR with its ligand, human recombinant EGF (Gibco Fisher, PHG0313) was adsorbed to an ELISA plate overnight at 4°C in PBS. Following blocking with 2% (w/v) BSA in TBS, EGFR-ECD was serially diluted and allowed to bind. After washes with TBS-T, a rabbit anti-His-tag antibody (Cell Signaling Technologies, 12698S) was incubated to bind the C-terminal His-tag on EGFR. Wells were washed and incubated with goat anti-rabbit HRP, washed, and reacted with TMB. All ELISA steps except initial adsorption were performed at room temperature on a microtiter plate. Sulfuric acid (2 M) was used to stop color development. All conditions were measured in triplicate. GraphPad Prism 8 was used to calculate approximate dissociation constant (K D ) values using the specific binding with the Hill slope model. Anti-HER2 Affibody Peptide Synthesis and ECD Binding Assay. The anti-HER2 affibody was synthesized using standard Fluorenylmethoxycarbonyl chloride (Fmoc)protected solid-phase peptide synthesis (SPPS) similar to other chemically synthesized Z domain peptides. 41 The affibody sequence is based on ZHER2:477. 42 Briefly, N-methyl-2pyrrolidone (NMP) was used as the solvent and wash, and 20% (v/v) piperidine in NMP was used to remove Fmoc. N,N-Diisopropylethylamine and HBTU (2-(1H-benzotriazol-1-yl)-1,1,3,3-tetramethyluronium hexafluorophosphate) were used as the base and coupling reagent, respectively. Fmoc-Lpropargylglycine (Chem-Impex, 05138) was the first amino acid coupled to the rink amide SS resin (Advanced ChemTech, SA5030), which enabled the affibody to be modified by alkyne/azide click chemistry. Following double amino acid coupling SPPS, the peptide/resin was dried and cleaved from the resin with 2.5% (v/v) triisopropylsilane, 2.5% (v/v) water, and 95% (v/v) trifluoroacetic acid, precipitated with diethyl ether, dried, and purified by C-18 reversed-phase (RP) highperformance liquid chromatography (HPLC). Cu(I)-catalyzed azide-alkyne click chemistry was utilized to incorporate a fluorescein azide onto the C-terminal propargylglycine to track in HER2/3 binding assays. Samples were RP HPLC-purified, lyophilized, and resuspended in PBS. Biotinylated HER2/3 thrombin/TEV was immobilized on a preblocked SA resin (as above), washed, incubated with fluorescein-labeled affibody, and released with the corresponding protease, as described above. Sample fluorescent intensity (488 nm excitation, 512 nm emission) was read in a BioTek Synergy H4 plate reader. Anti-HER2 Affibody mRNA−Peptide Fusion Production and HER2 Thrombin Screening. mRNA display techniques were used to produce an mRNA−peptide fusion. 43 DNA encoding the affibody and scrambled affibody were purchased from IDT as gBlock gene fragments. Following PCR amplification with primers designed to add a 5′ T7 polymerase sequence and a 3′ linker region, the purified DNA was transcribed with T7 RNA polymerase and urea PAGE-purified. The DNA linker PF30P (dA 21 -(3 x Spacer 9)-dAdCdCpuromycin) was ligated to the mRNA in the presence of a DNA splint with T4 DNA ligase and urea PAGE-purified. The ligated product was translated using a rabbit reticulocyte lysate in vitro translation kit without methionine (Fisher, AM1200). The affibody/scrambled peptide was translated in the presence of [ 35 S]-methionine (PerkinElmer, Neg772002MC) for incorporation and peptide/fusion tracking. Following oligo dT-resin 44 purification via the dA 21 sequence, the mRNA− affibody fusions were reverse-transcribed with M-MuLV reverse transcriptase (Fisher, NC9950976), ethanol-precipitated, and resuspended in selection buffer (25 mM HEPES-KOH pH 7.5, 150 mM NaCl, 0.05% (v/v) Tween 20, 1 mM EDTA, 5 mM MgCl 2 , 0.5 mg/mL BSA, 10 μg/mL calf liver tRNA). An excess of NeutrAvidin resin was preblocked in selection buffer for 30 min at 4°C on an end-over-end rotator. Following washes, fusions were precleared with half of the preblocked resin (1 h at 4°C) to remove resin-binding sequences and then centrifuged through a Corning Spin-X 0.45 μm filter to recover precleared fusions in the flow-through. The other half of the resin, which contained at least a 10× molar excess of biotin-binding sites as compared to the input fusions, was then incubated with the HER2-thrombin cleavage sitebiotin protein. Unbound HER2 was washed away, and the HER2-loaded resin was incubated with the precleared fusions for 2 h at 4°C on an end-over-end rotator. Unbound fusions were removed by centrifuging the resin and removing the supernatant. After six washes in selection buffer, the resin was split in half and incubated with or without thrombin (control) to elute HER2-bound fusions. The resin was washed and resuspended in selection buffer (final volume equal to thrombin elution). Flow-through, elution, and resin fractions were analyzed by quantitative real-time PCR (qPCR) using the iTaq-Universal SYBR Green One-Step Kit (Bio-Rad) on a Bio-Rad CFX96 instrument. mRNA serial dilution standards were used to calculate the cycle threshold of known picograms of mRNA input. Fractions were also assessed for [ 35 S]methionine radioactivity by liquid scintillation to provide counts per minute (CPM) values. Model Affibody Selection. Affibody and scrambled fusions were produced separately as previously described and combined in a 100:1 (scrambled/affibody) ratio. The mixed fusions were incubated with HER2-bound neutravidin resin for 1 h at 4°C and washed six times with selection buffer to remove unbound fusions. The resin was split into two and either kept in selection buffer (control resin) or incubated with thrombin as outlined above to release HER2 with any bound product. Post-thrombin resin was also kept for analysis. Volumes were kept consistent between control resin, thrombin-eluted, and post-thrombin resin. After PCR amplification of all fractions, the DNA product was digested with the restriction enzyme HaeIII (NEB). The affibody cDNA does not contain the HaeIII restriction site, but the scrambled version does and is efficiently degraded. Densitometry of ethidium-stained agarose gels allowed a fraction affibody value to be calculated (intensity of the 240 base pair band in the digest lane divided by the 240 base pair band in the nondigested lane) with a value of 1 corresponding to 100% affibody cDNA and a value of 0 corresponding to 100% scrambled cDNA. Each condition was carried out in triplicate (n = 3) and statistical significance was assessed by Student's t test (GraphPad Prism). ■ RESULTS AND DISCUSSION ECD Expression and Purification. In our initial validation studies, we chose to focus on EGFR family members, HER2, HER3, and EGFR. These are common cancer-associated RTKs that are exploited in multiple diagnostic and therapeutic strategies. 45 The DNA constructs included an SP derived from G. princeps luciferase 46 to direct the ECDs into the secretory system where they are posttranslationally modified (e.g., disulfide shuffling, glycosylation). This sequence was previously found to increase the secretion of mammalian extracellular proteins. 46 The C-terminus of each ECD contained a BirA tag for sitespecific biotinylation followed by a 6x His-tag for Ni 2+ -NTA purification. A protease site (thrombin or TEV) was included immediately N-terminal to the biotinylation site allowing the release of immobilized ECD by digestion with the appropriate protease. The expression system uses commercially available, high-density, nonadherent 293-F cells grown on a cell incubator shaker platform in flasks. The cells are transiently transfected with the ECD expression plasmid either alone or in tandem with a construct that expresses and secretes BirA. Cotransfection allows for the biotinylation of the ECD Cterminal BirA tag in the secretory route and/or cell culture media (in situ biotinylation). 38 Validation and Characterization of ECD Purification and Biotinylation. The ECDs of HER2, HER3, and EGFR were expressed in the transient nonadherent mammalian cell system and analyzed by SDS-PAGE and western blotting (Figure 2) . A time-course analysis of HER2 ECD expression in CM showed peak expression at 5 days post-transfection (Figure 2A) . Expression of immunoglobulin-G (IgG) was carried out as a positive control and showed high levels of protein in the CM at day 6. We note that IgG appears in the anti-HER2 western blot as a result of reactivity with the HRPconjugated secondary antibody. As expected, the negative control (no transfection, lane 5) shows no signal at 75−100 kDa. SDS-PAGE and Coomassie staining of the CM ( Figure 2B ) shows a strong band at ∼100 kDa, which is higher than the predicted MW of the HER2 ECD (∼72 kDa). The higher MW and smearing suggest post translational glycosylation, which is common among extracellular RTKs. 47 Following Ni 2+ -NTA purification, the identity of the expressed protein was confirmed by western blot ( Figure 2C ) and the purity was estimated to be >90% by densitometry. The absence of protein in the flow-through (FT, lane 2) shows efficient binding to the Ni 2+ -NTA resin. Purification of EGFR ( Figure 2D ) and HER3 ( Figure 2E ) ECDs by Ni 2+ -NTA showed similar results with streaking likely caused by heterogeneous glycosylation. Having confirmed robust ECD expression in CM, we sought to determine the efficiency of in vitro and in situ biotinylation. In vitro biotinylation of HER2 ECD with BirA was highly efficient, as shown by western blotting with streptavidin-HRP ( Figure 3A) . The efficiency of in vitro biotinylation was semiquantitatively measured by incubation with streptavidinagarose (SA-agarose) followed by SDS elution and SDS-PAGE analysis ( Figure 3B ). As expected, significant levels of untreated ECD are seen in the flow-through, while BirA biotinylated HER2 remains bound to the resin until it is eluted with SDS, reducing agent, and heat. HER2 not treated with BirA is largely absent in the elution. Densitometry analysis indicates an in vitro biotinylation efficiency of 94%. We note that BSA and SA monomers are present in the elution because the resin was preblocked with BSA and the SA monomers are released by reducing agent and detergent. Over multiple independent experiments with different ERBB constructs, the mean in vitro biotinylation efficiency was found to be 92% ± 2.9% (n = 10). We next assessed the requirement for d-biotin supplementation in the media during cotransfection and in situ biotinylation ( Figure 3C ). CM was supplemented with 100 μM d-biotin either during initial transfection (lane 1) or after the addition of transfection enhancers (1 day post-transfection, lane 2). No exogenous d-biotin added (lane 3) and no transfection controls (lane 4) were also included. The addition of exogenous d-biotin did not significantly increase the amount of biotinylated HER2, indicating that biotin is present in sufficient levels in CM to support biotinylation during expression. To determine the effect of free biotin on purification by SA-agarose, we purified in situ biotinylated EGFR ECD with SA-agarose with and without prior dialysis of the CM ( Figure 3D ). The undialyzed EGFR-biotin shows no resin binding (lane 4), while the dialyzed EGFR-biotin protein is efficiently bound (lane 7). This assay demonstrates that dialysis of the CM is essential if streptavidin/avidin-mediated capture of biotinylated ECD is performed prior to subsequent purification steps. The approximate efficiency of in situ biotinylation for HER3 and EGFR was found to be 72 and 80%, respectively, in these samples ( Figure 3E ). Over multiple independent experiments with different ERBB constructs, the mean in situ biotinylation efficiency was found to be 75% ± 3.9% (n = 5), which was lower than the efficiency observed for in vitro biotinylation. However, the in situ strategy offers considerable time savings and the possibility of one-step protein purification and immobilization. ECDs Are Properly Folded As Determined by ELISA. The biochemical function of each ECD was confirmed by ELISA. For each assay, we tested the binding of a ligand that requires proper ECD post-translational folding and orientation. Each ELISA schematic can be found under the associated binding curve in Figure 4 . To assess HER2, the anti-HER2 affibody was used, which binds only to properly folded HER2. 48 In this assay ( Figure 4A ), the HER2 ECD was found to bind the HER2 affibody with a dissociation constant (K D ) of 2.00 nM (+/− 0.16 nM), which is in-line with published accounts of the bivalent affibody (3 nM). 49 Hrgβ1 is the natural growth factor for HER3 and requires a properly folded HER3 (domains I and II) for maximum affinity. 50 We found that biotinylated HER3 ECD binds Hrgβ1 with a K D of 12.6 nM (+/−0.82 nM) ( Figure 4B ), which is also consistent with literature values (9.1 to 68 nM). 50, 51 Similarly, EGFR ECD bound to immobilized EGF with a K D of 125 nM (+/− 6.7 nM), a value in-line with previously published values (177 nM) 52 ( Figure 4C ). We note that the Hill slopes for all three ligand−receptor interactions were greater than 1. This may be explained by the dependence of ERBB receptor affinity on the oligomerization state with evidence for multiple ligand binding sites in the dimeric state. 53−56 This effect may also arise from the assay setup where a high local concentration of ligand is present on the surface. This may perturb the free diffusion of receptor when it is in close proximity to the surface yielding apparent binding cooperativity. This phenomenon has been well documented in the biosensor literature. 57, 58 These assays provide strong evidence that our expression and in situ biotinylation methodology leads to the production of properly folded and biologically active ECDs that can be used in downstream biochemical assays and selection experiments. Assessing Protease-Specific Release. HER2 and HER3 ECDs were produced with both thrombin and TEV protease sites to enable selective release from streptavidin/avidin resins. Nonbiotinylated HER2/3 with TEV or thrombin cleavage sites were incubated with their respective proteases in solution and analyzed by SDS-PAGE ( Figure 5A ) and western blotting ( Figure 5B ). After Coomassie staining, a slight band downshift is observed after TEV and thrombin cleavage, indicating proteolysis of the C-terminal fragment containing the His-tag. To confirm C-terminal His-tag removal, we performed a western blot with an anti-His-tag antibody. Without protease, a strong band (∼90 kDa) is detected, indicating that the His-tag is intact. Following protease incubation, the C-terminal His-tag is lost, resulting in dramatically reduced band intensity. The TEV protease itself contains a His-tag and therefore is detected in the western blot at ∼25 kDa. Thrombin is ∼35 kDa but does not have a His-tag and is barely visible after Coomassie staining. To determine if the proteases are also active in the context of a resin, biotinylated HER3 with a thrombin or TEV cleavage site was bound to SA resin and incubated with or with a thrombin/TEV cleavage site was bound to streptavidin (SA) resin and incubated with or without thrombin/TEV. Densitometry indicates 90% release by thrombin and 71.5% release by TEV. To the right is a schematic of the proteolytic release from the SA resin. These data confirm that thrombin and TEV can release the ECD from the resin without ECD proteolysis. without TEV or thrombin ( Figure 5C ). In fractions analyzed by Coomassie, both proteases release the ECD (lane 2, ∼90 kDa), leaving very little bound to the resin (lane 4). Without protease, the ECD remains with the resin (lane 7). In both solution and on-resin reactions, thrombin is more effective than TEV in releasing the ECD from immobilized streptavidin. Over multiple independent experiments with multiple ERBB constructs, the mean release efficiency with thrombin was 91% ± 6.4% (n = 9) while the mean release efficiency of TEV was 77% ± 5.3% (n = 3). Because thrombin was more effective in this system, we used it in the remaining experiments. In order to demonstrate that ligands that bind to immobilized ECDs can be selectively released with thrombin, we performed the experiment shown in Figure 6A . The HER2 binding affibody was chemically synthesized by SPPS with a Cterminal propargylglycine in order to facilitate fluorescein-azide conjugation by click chemistry. HER2-biotin (gray bar) or HER3-biotin (black bar) ECDs with a thrombin cleavage site were immobilized on SA resin, washed, and incubated with the fluorescent affibody ( Figure 6B ). The resin was treated with thrombin, washed, and resuspended in selection buffer. Fluorescent signal in the flow-through shows tight binding of the HER2 affibody to HER2 but not to HER3. In the absence of thrombin, the affibody remains bound to the HER2 resin. Addition of thrombin releases the fluorescent affibody with high efficiency (n = 1) based on this experiment and an analogous experiment described in Figure S1 . Both HER2 and HER3 ECDs are released by thrombin and only ECDs with a thrombin site can be removed from the resin in this manner, confirming protease orthogonality ( Figure S1 ). In summation, the purification and release experiments verify that ECDs are properly secreted, folded, site-specifically biotinylated, and can be released intact even from solid-phase resins. Table 1 provides a summary of ECD proteins expressed in this system (single batch) along with their respective yields, biotinylation methods, and engineered protease sites. The described purification system provides sufficient product per 100 mL of CM to support a directed evolution screen (≥1 mg). Further optimization and upscaling will likely increase the final output since the reported purification values are based on relatively small-scale CM batches (30−60 mL). Generally, the system achieved greater than 90% ECD purity for all ECDs tested. If additional purity is required, capture on avidin/streptavidin resins provides a potential route for achieving purities >95% prior to initiation of screening. While we present this expression/purification scheme using RTK ECDs as exemplars, it can be expanded to other heavily glycosylated and disulfide-dependent proteins and modified for specific post-translational characteristics. For example, if a specific glyco-pattern is desired, other cell types (e.g., CHO) can be used produce different glycosylation patterns. 59, 60 Glycoproteins are generally folded and soluble, 61 but fusion partners or other tags can be added if target solubility is an issue (e.g., maltose-binding protein fusions) 62, 63 or dimeric presentation is desired. 64 While we used the G. princeps derived SP for enhanced secretion, 46 SP choice is likely to be an empirical parameter for each recombinant protein and cell a Each recovery value is based on a single purification. The asterisk indicates a grouping of yields based on the stated characteristic (protease site, biotinylation method, or ECD) and shows the mean yield (± SEM) for ECDs sharing that characteristic. system, 65 and alternative SPs may significantly increase secretion and recovery rates. 65, 66 As we employed small-scale expressions to obtain multiple ERBB constructs, standard losses from each step (e.g., buffer exchange) may also have resulted in suboptimal yields. End-users of the purification system may expect improved final yields during scale-up where higher volumes of culture media are used. Given the amount of purified protein required for selection experiments (at least 1 Figure 7 . Directed evolution selection screen for increased target binding specificity. Site-specifically biotinylated HER2 ECD is immobilized on streptavidin resin and incubated with radiolabeled mRNA-affibody fusions. Following washing, a protease (TEV or thrombin) releases the protein target along with target-bound fusion molecules. Nonspecific or unfused (no peptide) fusions remain bound to the resin allowing for selective enrichment of HER2-binding fusions by PCR amplification. S-labeled mRNA−affibody fusion particle. DNA encoding the anti-HER2 affibody is transcribed into mRNA and ligated to a poly-dA oligo bearing puromycin at the 3′ end. This template is translated in rabbit reticulocyte lysate resulting in the formation of an affibody−mRNA fusion. The 35 S radiolabel is incorporated into the protein via [ 35 S]-methionine in the translation mixture. Following purification with oligo dT-resin, the affibody−mRNA fusion is reverse-transcribed and panned against the immobilized HER2 ECD. The mRNA templates contain a 5′ untranslated region used to initiate translation. (B) Composition of the affibody−mRNA fusion was confirmed by SDS-PAGE autoradiography. HER2 affibody−mRNA and scrambled HER2 affibody−mRNA fusions were translated and purified as described above and incubated with RNAse A to degrade the mRNA portion. Separation by SDS-PAGE followed by autoradiography shows the correct mass for the affibody−mRNA fusion as well as the affibody-poly-dA degradation product. (C) HER2 affibody and scrambled affibody fusions were panned against immobilized HER2 ECD followed by scintillation counting of the flow-through. CPM values were converted to a percent of input for respective fusions. The scrambled fusions appeared in the flow-through, while the affibody fusions were retained on the resin and the difference between the two was found to be statistically significant (p value < .0001, seven individual experiments). This indicates that the HER2 affibody was properly folded in the context of the HER2 affibody−mRNA fusion. mg) and the ∼0.7 to 3 mg yields from even 100 mL of CM, additional expression optimization may be unnecessary for directed evolution applications. However, these yields are likely to be prohibitively low for protein-intensive applications such as crystallography or isothermal titration calorimetry and would require significant scale-up and optimization. Improving Enrichment in mRNA Display Selections via Protease-Mediated Elution. Next, we determined if the addition of a specific elution step via site-specific protease digestion could be incorporated into an mRNA display selection (Figure 7) . mRNA display is a powerful biological display technology where translated peptides and proteins are covalently fused to their encoding mRNAs via a puromycin linker effectively linking the genotype (RNA) with its phenotype (peptide). 12 In order to track mRNA−protein fusions through purification and selection steps, translation is carried out in the presence of [ 35 S]-methionine resulting in incorporation of the 35 S radionuclide at all methionine positions in the fused protein. The process of mRNA display is shown in Figure 8A and described in more detail in the Experimental Procedures section. After panning the fusion library against an immobilized target, functional sequences can be recovered by PCR of the reverse-transcribed DNA. The recovered DNA is then transcribed and translated to generate an enriched protein library for the next round of selection. After multiple rounds of selection, the diversity of the mRNA display library is dramatically reduced, allowing identification of functional peptide/protein sequences through DNA sequencing. One of the challenges unique to mRNA display is unwanted amplification of mRNA that failed to fuse with a translated peptide/protein. Unfortunately, fusion efficiency is around 10%, 43 leaving greater than 90% of mRNA linker free to bind nonspecifically to the solid support. This is highly likely to increase noise during the selection experiment and accumulation of false-positive sequences in the final pool. In the following experiments, we generated mRNA−protein fusions consisting of a HER2-binding affibody and a nonfunctional scrambled affibody to test the specific release of HER2 ECD with thrombin treatment. This strategy allowed us to compare the target binding and elution of functional (mRNA− affibody) and nonfunctional (mRNA−scrambled) fusions and to compare the signal-to-noise ratio (SNR) of proteasemediated elution relative to direct on-bead amplification. SDS-PAGE autoradiography ( Figure 8B ) of the mRNA− affibody and mRNA−scrambled affibody fusion products shows a molecular weight consistent with the predicted value (∼87 kDa). Degrading the encoding RNA with RNAse A results in a shift of the band corresponding to only the affibody peptide and poly dA-puromycin DNA linker (∼16 kDa). These experiments confirmed the composition of the mRNA− affibody fusion. To confirm that the mRNA−affibody fusion binds tightly to the HER2 ECD, we carried out a radioligand binding experiment with mRNA−affibody fusions and SAagarose-immobilized HER2 ECD. We observed nearly quantitative binding of the mRNA−affibody fusion to the immobilized ECD, as evidenced by the low amount of radiolabeled affibody fusion in the flow-through fraction ( Figure 8C ). As expected, the scrambled affibody fusion fails to bind HER2 and is recovered mostly in the flow-through fraction. These data were obtained from seven independent experiments. We then tested the efficiency and selectivity of the sitespecific protease release of mRNA−affibody fusions from immobilized HER2. Both mRNA−affibody and scrambled fusions were incubated with the HER2-bound resin and incubated with or without thrombin ( Figure 9 ). Addition of thrombin results in nearly quantitative cleavage and release of the resin-bound affibody, with minimal activity remaining on the resin ( Figure 9A ). In contrast, merely washing the resin with buffer (no thrombin) fails to release the affibody fusion and results in retention of affibody on the resin. Subsequent analyses of HER2 ECD elution by SDS-PAGE and densitometry confirm that the ECD elution efficiency is similar between the scrambled and HER2 affibody, indicating that differences in affibody−mRNA elution efficiency are not the result of differential ECD elution efficiency ( Figure S2 ). To measure the selectivity of protease-mediated elution, we calculated the SNR by dividing CPM values of the affibody− mRNA fusions by the CPM values for the scrambled affibody− mRNA fusions ( Figure 9B ). The SNR obtained by protease elution was significantly higher than that achieved by direct recovery of resin-bound fusions. This was primarily driven by the higher background level of residual scrambled fusions that remained bound to the resin. Specific elution by protease should result in higher enrichment of target-binding mRNA−affibody fusions as any nonspecific resin-binding sequences will not be eluted by protease digestion. To quantify the effects of proteasemediated target elution on the effective SNR of a single selection round, we performed a mock selection experiment with and without thrombin elution and quantified the effects by qPCR. We performed qPCR on the eluent and resuspended resin, with or without thrombin treatment, and converted to input picograms (pg) DNA by extrapolation from a standard curve ( Figure 9C ). In line with the radioactive quantitation above, the qPCR data indicate the significant release of cDNA from anti-HER2 mRNA−affibody fusions immobilized on HER2 resins but not from resins incubated with scrambled fusions. The post-thrombin-treated resin showed some residual DNA for both scrambled and HER2-binding fusions. In the absence of thrombin treatment, direct PCR amplification of the resin-bound material yielded approximately half of the DNA obtained from the thrombin-mediated release. Quantification of the elution by qPCR in the no thrombin sample does result in some eluted cDNA despite lower corresponding levels of eluted 35 S activity ( Figure 9A ). This may indicate that unfused mRNA (without peptide) is being leached from the solid-phase resin. We compared the yield and SNR of on-resin PCR versus PCR following thrombin elution of the HER2 target ( Figure 9D ). Thrombin-mediated elution results in >2-fold more (scrambled/affibody), selected for binding to HER2, recovered directly off the resin or with thrombin, and PCR-amplified. The PCR product is then incubated with or without the HaeIII restriction enzyme. The scrambled DNA is digested by HaeIII but the affibody DNA remains intact. (B) Agarose gel of PCR product digests of the mixed input, the unbound mix flow-through, and the pure affibody and scrambled fusions. The affibody DNA remains intact after HaeIII treatment but the scrambled DNA is efficiently digested. The mixed input is mostly scrambled affibody as expected. (C) After selection, elution with thrombin yields predominantly affibody DNA (intact in HaeIII). In contrast, the thrombin-treated resin retains only scrambled DNA. The resin control also yields predominantly affibody DNA. (D) Densitometry of PCR bands was used to calculate the fraction affibody (HaeIII digest lane)/(nondigest lane), with higher values corresponding to a higher fraction of affibody DNA. While the mixed input yields mostly scrambled DNA, the thrombin elution yields mostly affibody DNA. Affibody DNA is also enriched in the resin control but significantly less so than in the thrombin elution (*** = p value < 0.001). The post-thrombin resin yields mostly scrambled DNA, indicating that the majority of the affibody−mRNA fusion has been released by thrombin activity. selective DNA recovery relative to on-resin amplification. To measure the contribution of nonspecific recovery (recovery of DNA from the scrambled mRNA fusion), we calculated the SNR as the ratio between the affibody DNA (signal) and the scrambled affibody DNA (noise) (affibody/scrambled). We observed that the average SNR is approximately 1.9-fold higher for protease elution compared to on-resin PCR (250 versus 130). Addition of thrombin has no effect on the efficiency of the PCR reaction ( Figure S3 ). Taken together, these data indicate that selective, protease-mediated elution of target results in considerably higher yields and lower nonspecific background compared to direct PCR amplification of the solidphase resin. Specific Affibody Enrichment in Mixed Screen. To assess if the functional affibody is enriched in HER2 selections, we performed the experiment described in Figure 10 . Translated fusions were mixed in a ratio of 100:1 (scrambled/affibody), allowed to bind to HER2, and washed. The resin was split for resin-bound control or thrombin elution. After PCR amplification, fractions were subjected to digestion by HaeIII where the affibody DNA is left intact but the scrambled DNA is digested at a single cut site ( Figure 10A ). This digestion eliminates the 240 bp scrambled DNA, leaving only the 240 bp affibody DNA product. Figure 10B ,C displays the agarose gels of each fraction along with input controls. The affibody DNA remains intact on the gel in the presence of HaeIII but the scrambled affibody DNA is quantitatively digested (Figure 10B ), indicating that HaeIII digest works to completion. The mixed input (100:1) is mostly scrambled DNA as is the flow-through fraction. Figure 10C supports strong enrichment of the affibody over scrambled in the thrombin elution and resin control. In contrast, the DNA remaining on the thrombin-treated resin is derived almost entirely from the scrambled population. The band intensity (top band of 240 bp) was quantified by densitometry, allowing a simple calculation of affibody fraction (240 bp band intensity in the digested lane/240 bp band intensity in the nondigested lane) for each digest replicate (n = 3, Figure 10D ). DNA from affibody fusion PCR has a value close to 1, indicating that it is not digested and therefore is pure affibody DNA. Conversely, this value approaches 0 for the scrambled fusions. Interestingly, the value decreases in the unbound flow-through versus the mixed input, indicating that the small amount of affibody fusions in the input mix bound the HER2 resin (p = 0.012). The fraction affibody is high in thrombin-eluted samples (0.70) and control resin (0.37) but low in the resin after thrombin treatment (0.087). The fraction of affibody DNA after thrombin elution was nearly twice that of the on-resin sample and this difference was statistically significant (p value < 0.001). This 2-fold enhancement in SNR following thrombin elution is very much in-line with the results shown in Figure 9 . These data also confirm that our purified HER2 is properly folded, binds selectively to affibody fusions, and can be efficiently released by thrombin. Although mRNA−protein fusions (e.g., mRNA−fibronectin libraries) have been extensively described in the literature, 67, 68 the mRNA−affibody fusion has only been recently described. 44 In addition to increasing the probability of nonspecific binding, the large DNA/RNA carrier molecule (∼300 bp) may significantly increase nonspecific interaction with the solid phase resin. Indeed, the experiment described in Figure 9 required preclearing with a blocked resin to eliminate the high background of largely unfused genetic material. This observation also highlights the challenges associated with unfused genetic material, which binds nonspecifically to the solid-phase support during mRNA Display selections. If this material is amplified at the end of each round, it will introduce significant noise in the selection process, which may increase the number of false-positive (nonfunctional) sequences in the final pool. Aside from the time-consuming and costly process of identifying and discarding these sequences in the validation stage, increased noise also decreases specific enrichment of target binding sequences and extends the number of selection rounds required for library convergence. Fortunately, the data in Figure 10D suggest that the nonspecific fusion material bound to the streptavidin or the agarose matrix can be eliminated from the final eluent, providing a pathway to noise reduction throughout the selection. Given the SNR values in Figure 9D , a starting diversity of 10 13 unique sequences, and a final goal of 10 3 functional sequences, protease-mediated target release could potentially decrease the selection process by two rounds relative to on-resin PCR. Directed evolution remains a powerful, yet challenging technology for the discovery of functional peptides and proteins. Reducing noise in directed evolution screens is critical to a successful selection. This can be accomplished, in part, by generating high-quality protein targets and selectively immobilizing them on inert, biologically compatible resins. Our purification scheme is designed to decrease upfront time costs associated with each screen, allowing researchers to rapidly and inexpensively generate folded, site-specifically biotinylated targets for downstream selection applications. Indeed, target purification and immobilization can be carried out in a single step. Previous work has demonstrated the feasibility of in vitro biotinylation of proteins expressed in E. coli 29 and in situ biotinylation in adherent mammalian cell systems. 38 Although these methods work well for the largescale production of unmodified proteins or small-scale production of post-translationally modified proteins, they are unsuitable for the rapid production of multimilligram quantities of site-specifically biotinylated, post-translationally modified mammalian targets for directed evolution. Further, methods involving adherent cells require dual stable transfection, which is costly in upfront clonal selection time. The nonadherent system described here allows for rapid production of targets and control targets (i.e., mutations, domain deletions) to enhance selections without months of prior work. The selection system also can be easily adapted to multiwell formats by using streptavidin-coated plates. Specific protease release will simplify library recovery and decrease noise arising from nonspecific adsorption to hydrophobic passive adsorption surfaces. Selective elution of target-bound sequences is also a potent mechanism of noise reduction, as seen in the mock affibody selection. Although this experiment illustrates the potential gains in SNR for each round, it likely represents the upper bound of noise reduction as many sequences, unlike the scrambled affibody fusion, will show some affinity for the target. Nevertheless, the SNR improvements obtained by protease-mediated target release are significant and are likely to reduce the time required to achieve library convergence and eliminate confounding false positives in the final pool. Taken together, these improvements will facilitate more rapid, high-quality selection experiments against emerging targets identified by proteomic analyses. The Supporting Information is available free of charge at https://pubs.acs.org/doi/10.1021/acsomega.0c03990. Figure S1 , specific proteolytic release of ERBB2/HER2associated affibody; Figure S2 , quantitation of ERBB2/ HER2 release from the streptavidin resin; Figure S3 , effect of thrombin treatment on cDNA stability; Figure S4 , full western blots from Figure 2A ,C; Figure S5 , full western blots from Figure 2D ,E; Figure S6 , full western blots from Figure 3A ; Figure S7 , full western blots from Figure 3C (PDF) Selection platforms for directed evolution in synthetic biology ) Omidfar, K.; Daneshpour, M. Advances in phage display technology for drug discovery Bacterial surface display: Trends and progress Molecular evolution of peptides by yeast surface display technology Ribosome display: A perspective In vitro selection of protein and peptide libraries using mRNA display Macrocyclic Peptides as Drug Candidates: Recent Progress and Remaining Challenges RNA Display Methods for the Discovery of Bioactive Macrocycles In vitro selection of highly modified cyclic peptides that act as tight binding inhibitors Directed Evolution of Scanning Unnatural-Protease-Resistant (SUPR) Peptides for in Vivo Applications MRNA display: From basic principles to macrocycle drug discovery α1-FANGs: Protein Ligands Selective for the α-Bungarotoxin Site of the α1-Nicotinic Acetylcholine Receptor Allosteric Inhibition of a Semaphorin 4D Receptor Plexin B1 by a High-Affinity Macrocyclic Peptide FN3: a new protein scaffold reaches the clinic. Drug Discovery Today Peptide-based PET quantifies target engagement of PD-L1 therapeutics Rapid PD-L1 detection in tumors with PET using a highly specific peptide Anti-tumor effect of CT-322 as an adnectin inhibitor of vascular endothelial growth factor receptor-2. mAbs A current perspective on applications of macrocyclic-peptide-based high-affinity ligands Combination cancer immunotherapy and new immunomodulatory targets Bioactive Macrocyclic Inhibitors of the PD-1/PD-L1 Immune Checkpoint Vertebrate protein glycosylation: Diversity, synthesis and function Protein disulfide-isomerase, a folding catalyst and a redox-regulated chaperone Optimisation of the cellular metabolism of glycosylation for recombinant proteins produced by mammalian cell systems Better and faster: improvements and optimization for mammalian recombinant protein production Bacterial expression systems for recombinant protein production: E. coli and beyond Protein immobilization strategies for protein biochips Nonspecific amine immobilization of ligand can be a potential source of error in BIAcore binding experiments and may reduce binding affinities Cloning antifungal single chain fragment variable antibodies by phage display and competitive panning elution Filamentous fusion phage: Novel expression vectors that display cloned antigens on the virion surface Design of Cyclic Peptides That Bind Protein Surfaces with Antibody-Like Affinity Selection of high-affinity phage antibodies from phage display libraries Conformation and orientation of a protein folding intermediate trapped by adsorption Metabolic Biotinylation of Recombinant Proteins in Mammalian Cells and in Mice In vivo site-specific biotinylation of proteins within the secretory pathway using a single vector system Efficient biotinylation and single-step purification of tagged transcription factors in mammalian cells and transgenic mice The EGFR family: Not so prototypical receptor tyrosine kinases. Cold Spring Harbor Perspect Large-scale screening for novel low-affinity extracellular protein interactions Expression and purification of E. coli BirA biotin ligase for in vitro biotinylation Structural genomics of human proteins-target selection and generation of a public catalogue of expression clones Chemical synthesis of triple-labelled three-helix bundle binding proteins for specific fluorescent detection of unlabelled protein Tumor imaging using a picomolar affinity HER2 binding Affibody molecule Optimized synthesis of RNA-protein fusions for in vitro protein selection 2020, 30, 126934. (45) Seshacharyulu, P.; et al. Targeting the EGFR signaling pathway in cancer therapy The level of synthesis and secretion of Gaussia princeps luciferase in transfected CHO cells is heavily dependent on the choice of signal peptide Glycosylation in cancer: Mechanisms and clinical implications Structural basis for high-affinity HER2 receptor binding by an engineered protein Affibody-mediated tumour targeting of HER-2 expressing xenografts in mice Identification of a Heregulin Binding Site in HER3 Extracellular Domain Formation of a high affinity heregulin binding site using the soluble extracellular domains of ErbB2 with ErbB3 or ErbB4 Quantitative Analysis of Ligand-EGFR Interactions: A Platform for Screening Targeting Molecules Ligand-induced, receptor-mediated dimerization and activation of EGF receptor Oligomers of ERBB3 have two distinct interfaces that differ in their sensitivity to disruption by heregulin Quantitation of the effect of ErbB2 on epidermal growth factor receptor binding and dimerization Heterogeneity in EGF-binding affinities arises from negative cooperativity in an aggregating system Kinetics of ligand binding to receptor immobilized in a polymer matrix, as detected with an evanescent wave biosensor I. A computer simulation of the influence of mass transport Real-time biacore measurements of escherichia coli singlestranded dna binding (ssb) protein to polydeoxythymidylic acid reveal single-state kinetics with steric cooperativity Protein glycosylation in diverse cell systems: implications for modification and analysis of recombinant proteins Differences in the glycosylation of recombinant proteins expressed in HEK and CHO cells Enhancing the secretion of recombinant proteins by engineering N-glycosylation sites Recombinant protein expression and purification: A comprehensive review of affinity tags and microbial applications Maltose-binding protein (MBP), a secretionenhancing tag for mammalian protein expression systems Fcγ1 fragment of IgG1 as a powerful affinity tag in recombinant Fc-fusion proteins: immunological, biochemical and therapeutic properties Over-expression of secreted proteins from mammalian cell lines Overcoming the Refractory Expression of Secreted Recombinant Proteins in Mammalian Cells through Modification of the Signal Peptide and Adjacent Amino Acids mRNA display design of fibronectin-based intrabodies that detect and inhibit severe acute respiratory syndrome coronavirus nucleocapsid protein Design, expression, and stability of a diverse protein library based on the human fibronectin type III domain The authors declare no competing financial interest.