Directed evolution using combinatorial protein libraries is a tremendously powerful technique for the generation of proteins with new or improved properties. A key aspect in such techniques is the link between individual protein variants and their corresponding genetic information. To provide this link, the most successful combinatorial protein engineering methods employ microorganisms, such as bacteriophages, bacteria or yeast for the production and display of libraries. This thesis focuses on the development and application of new directed evolution methods utilizing the bacterium Escherichia coli (E. coli), for the engineering of affinity proteins and proteases.

The first study aimed to engineer the substrate specificity of a protease from tobacco etch virus (TEV). For this purpose, a novel method was devised based on expression of intracellular protease libraries, and employed a reporter fusion protein consisting of amyloid β peptide fused to the N-terminus of enhanced green fluorescent protein (EGFP). Variants were screened for proteolytic activity on co-expressed target substrate by means of fluorescence-activated cell sorting (FACS). After three rounds of FACS, a set of TEV protease variants were enriched that exhibited improved proteolytic activity on the novel substrate.

Studies two to four describe the development of an E. coli surface expression system that was explored for directed evolution applications. The method is based on display of recombinant proteins on the outer membrane via fusion to a bacterial autotransporter, adhesin involved in diffuse adherence I (AIDA-I). The second study focused on the optimization of the surface display system and its application to directed evolution. In this effort, several affinity protein classes were evaluated for surface display via AIDA-I in a panel of E. coli strains. Results showed that smaller and less complicated affibody molecules were displayed at high levels, while more complex proteins, such as antibody fragments, varied in their performance and functioned best in certain engineered strains. A mock affibody library was used to develop a high-throughput magnetic-assisted cell sorting (MACS) protocol for enrichment of binders from very large libraries.

In the third and fourth study, the new E. coli display method combined with the MACS protocol was evaluated for generation of new affibody molecules.

In the third study, a large naïve affibody library (>1.5×10¹¹ members) was constructed, displayed on E. coli and characterized. The performance of the method and library was evaluated by selection of binders against two cancer-associated targets, tumor-associated calcium signal transducer 2 (TROP-2) and lymphocyte-activation gene 3 (LAG-3). MACS and FACS were performed, with flow cytometry assessment between rounds to monitor enrichment. Both selections produced high affinity binders to their respective targets.

In the fourth study, a maturation library was constructed for improving the properties of an affibody molecule toward the renal cell carcinoma biomarker carbonic anhydrase IX (CAIX). Selections included stringent off-rate procedures and yielded variants with improved affinities and folding stability compared to previously reported binders.

In summary, the work in this thesis demonstrate the potential of E. coli-based directed evolution methods for selection of new proteins with altered or improved properties.

Riktad evolution med hjälp av kombinatoriska proteinbibliotek är en oerhört kraftfull teknik för att ta fram proteiner med nya eller förbättrade egenskaper. En viktig del i sådana tekniker är att det finns en länk mellan individuella proteinvarianter och deras motsvarande genetiska information. För att tillhandahålla denna länk nyttjar de mest framgångsrika metoderna mikroorganismer, såsom bakteriofager, bakterier eller jästceller för produktionen och uttryck av proteinbibliotek. Denna avhandling fokuserar på utveckling och tillämpning av nya riktade evolutionsmetoder som använder bakterien Escherichia coli (E. coli) för selektion av nya affinitetsproteiner och proteaser.

Den första studien syftade till att förändra substratspecificiteten för ett proteas från tobacco etch virus (TEV). För detta ändamål utvecklades en metod baserad på expression av intracellulära proteasbibliotek, och använde ett reporterfusionsprotein bestående av amyloid β-peptid fuserad till N-terminalen av förstärkt grönt fluorescerande protein (EGFP). Varianter screenades för proteolytisk aktivitet på samuttryckt målsubstrat med hjälp av fluorescensaktiverad cellsortering (FACS). Efter tre rundor av FACS selekterades ett antal varianter av TEV-proteas som uppvisade förbättrad proteolytisk aktivitet på det nya substratet.

Studierna två till fyra beskriver utvecklingen av en annan E. coli-metod som utforskades för riktad evolution av affinitetsproteiner. Metoden är baserad på expression av rekombinanta proteiner på det yttre membranet av E. coli via fusion till en bakteriell autotransporter, adhesin involved in diffuse adherence I (AIDA-I). Den andra studien fokuserade på optimering av metoden och dess tillämpning inom riktad evolution. I studien utvärderades flera olika affinitetsproteiner för expression via AIDA-I i en panel av E. coli-stammar. Resultaten visade att relativt små affibodymolekyler uttrycktes i höga nivåer, medan större och mer komplexa proteiner, såsom antikroppsfragment, varierade i uttryck och bara fungerade i vissa modifierade stammar. I studien utvecklades även ett protokoll för magnet-assisterad cellsortering (MACS) med syftet att möjliggöra selektion av bindare från mycket stora bibliotek.

I den tredje och fjärde studien utvärderades den nya E. coli-metoden i kombination med MACS-protokollet för selektion av nya affibodymolekyler.

I den tredje studien konstruerades ett stort naivt affibody-bibliotek (>1,5×10¹¹ varianter). Potentialen hos metoden och det nya biblioteket utvärderades genom selektion av bindare mot två cancer-associerade receptorer, tumor-associated calcium signal transducer 2 (TROP-2) och lymphocyte-activation gene 3 (LAG-3). MACS och FACS användes i selektionsprocessen, och kombinerades med flödescytometrisk analys av biblioteket mellan rundorna, för att övervaka selektionen. Båda selektionerna resulterade i nya bindare med relativt hög affinitet till sina respektive receptorer.

I den fjärde studien konstruerades ett riktat bibliotek med syftet att förbättra egenskaperna hos en affibodymolekyl mot carbonic anhydrase IX (CAIX). Selektionen inkluderade stringenta förfaranden för att anrika varianter med långsam dissociation.  Analys efter FACS visade på varianter med förbättrad affinitet och stabilitet jämfört med tidigare rapporterade bindare.

Sammanfattningsvis visar arbetet i denna avhandling potentialen hos E. coli-baserade metoder för selektion av nya proteiner med förändrade eller förbättrade egenskaper.

Autotransporters constitute a large family of natural proteins that are essential for delivering many types of proteins and peptides across the outer membrane in Gram-negative bacteria. In biotechnology, autotransporters have been explored for display of recombinant proteins and peptides on the surface of Escherichia coli and have potential as tools for directed evolution of affinity proteins. Here, we investigate conditions for AIDA-I autotransporter-mediated display of recombinant proteins. A new expression vector was designed and engineered for this purpose, and a panel of proteins from different affinity-protein classes were subcloned to the vector, followed by evaluation of expression, surface display and functionality. Surface expression was explored in ten different E. coli strains together with assessment of transformation efficiencies. Furthermore, the most promising strain and expression vector combination was used in mock library selections for evaluation of magnetic-assisted cell sortings (MACS). The results demonstrated dramatically different performances depending on the type of affinity protein and choice of E. coli strain. The optimized MACS protocol showed efficient enrichment, and thus potential for the new AIDA-I display system to be used in methods for directed evolution of affinity proteins.

Due to their ability to catalytically cleave proteins and peptides, proteases present unique opportunities for the use in industrial, biotechnological, and therapeutic applications. Engineered proteases with redesigned substrate specificities have the potential to expand the scope of practical applications of this enzyme class. We here apply a combinatorial protease engineering-based screening method that links proteolytic activity to the solubility and correct folding of a fluorescent reporter protein to redesign the substrate specificity of tobacco etch virus (TEV) protease. The target substrate EKLVFQA differs at three out of seven positions from the TEV consensus substrate sequence. Flow cytometric sorting of a semi-rational TEV protease library, consisting of focused mutations of the substrate binding pockets as well as random mutations throughout the enzyme, led to the enrichment of a set of protease variants that recognize and cleave the novel target substrate.

Carbonic anhydrase IX (CAIX) is a transmembrane enzyme that plays an important role in pH regulation and cellular homeostasis. Overexpression of CAIX is commonly observed in various solid tumors, making it an attractive target for cancer therapy. Affibody molecules have previously been developed for CAIX and demonstrated potential as tracers for molecular imaging in murine xenografted tumor models. Here, we further develop such affibody molecules by displaying a designed affibody library on the surface of Escherichia coli, followed by isolation of CAIX-binding variants by a combination of magnetic-assisted cell sorting and fluorescence-activated cell sorting. The sortings were successful and screening of candidates after selection revealed several variants with improved properties, most notably an increase in affinity and CAIX cell binding at 37°C and 42°C. 

Due to their ability to catalytically cleave proteins and peptides, proteases present unique opportunities for the use in industrial, biotechnological, and therapeutic applications. The possibility to engineer proteases with redesigned substrate specificities has the potential to expand the scope of practical applications of this enzyme class. We here apply a combinatorial protease engineering screening method that links proteolytic activity to the solubility and correct folding of a fluorescent reporter protein to redesign the substrate specificity of Tobacco Etch Virus (TEV) protease. The target substrate EKLVFQA differs at three of seven positions from the TEV consensus substrate sequence and exhibits high sequence similarity to the aggregation-inducing hydrophobic core region of the amyloid beta (Aβ) peptide. Flow cytometric sorting of a semi-rational TEV protease library led to the enrichment of a set of protease variants that recognize and cleave the novel target substrate.

Within combinatorial protein engineering, the probability of finding variants with new or improved attributes correlates with the size of the library used for the selection. In this paper, we describe what we believe to be the largest cell-displayed affibody library reported to date. Leveraging the high transformation efficiency of Escherichia coli, we generated an affibody library containing over 100 billion different protein variants, which were displayed on the surface of bacteria. Through a series of selections using magnetic-assisted cell sorting (MACS) and fluorescent-activated cell sorting (FACS), we screened the library against two different cancer relevant targets: TROP-2 and LAG-3. Next generation sequencing (NGS) confirmed that the distribution of mutations within the library was according to design and showed target-specific enrichment of distinct sequence clusters. We employed flow cytometry for both monitoring target binding across selection cycles, and for subsequent single-clone analysis of affibody hits, which helped identify the most promising candidates for further development. Affinity characterizations of soluble affibody molecules revealed low-nanomolar affinities for both respective targets, aligning with the enrichment trends observed in both flow cytometry and MiSeq analyses.