Royal Institue of TechnologySchool of Biotechnology

Stockholm 2017

Ken G Andersson

Combinatorial protein engineering of Affibody molecules using E. coli display and rational design

of Affibody-based tracers for medical imaging

© Ken AnderssonStockholm 2017

Royal Institute of TecnologySchool of BiotechnologyAlbaNova University CenterSE-106 91 StockholmSweden

Printed By: US-AB

ISBN 978-91-7729-504-4TRITA-Bio Report 2017:17ISSN 1654-2312

Ken G Andersson

III

Education is important but big biceps is importanter

IV

Public defense of dissertation

This thesis will be defended Oct 6th, 2017 at 13:00, in F3 Lindstedtsvägen 26, Sing-Sing, floor 2, KTH Campus for the degree of “Teknologie doctor” (Doctor of philosophy, PhD) in biotechnology.

RespondentKen Andersson, MScDivision of Protein Technology, School of Biotechnology, KTH Royal Institute of Technology, Stockholm, Sweden

Faculty opponentProf. Harald KolmarInstitute for Organic Chemistry and Biochemistry, Technische Universität Darmstadt, Darmstadt, Germany

Evaluation committeeDocent Ylva IvarssonDepartment of Chemistry, BMC Biomedicinskt centrum, Uppsala University, Uppsala, Sweden

Dr. Anders OlssonProtein expression and characterization facility, Science for Life Laboratory, Stockholm, Sweden

Docent Marika NestorDepartment of Immunology, Genetics and Pathology, Uppsala University, Uppsala, Sweden

ChairmanProf. Per-Åke Nygren Division of Protein Technology, School of Biotechnology, KTH Royal Institute of Technology, Stockholm, Sweden

Respondent’s main supervisorDocent John LöfblomDivision of Protein Technology, School of Biotechnology, KTH Royal Institute of Technology, Stockholm, Sweden

Respondent’s co-supervisor Prof. Stefan StåhlDivision of Protein Technology, School of Biotechnology, KTH Royal Institute of Technology, Stockholm, Sweden

Ken G Andersson

V

AbstractDirected evolution of affinity proteins is today an established method to generate binders against therapeutically and diagnostically important proteins. Thus, robust and high-throughput selection platforms are needed. This thesis describes the development of a cell display technology using Escherichia coli to engineer Affibody molecules. Further, the thesis describes rational design of Affibody-based tracers for target-specific medical imaging with potential benefits for patient stratification.

It is generally accepted that larger combinatorial protein libraries increase the probability of isolating binders with desired features. Typically, transformation efficiency is the limiting factor for library size. The Gram-negative bacterium E. coli has one of the highest reported transformation efficiencies, making it an attractive host for display of protein libraries. However, E. coli display is relatively challenging. A promising approach for surface display is based on the secretion pathway of the autotransporter family. In paper I, we engineered an expression vector for surface display of Affibody molecules using the autotransporter. This display platform showed high viability and enrichment in mock selections, indicating that it has good potential for combinatorial protein engineering. Following this study, the expression vector was reduced in size and a large combinatorial Affibody molecule library of 2.3x109 variants was generated. Selection against tumor-related biomarkers resulted in specific binders in the nanomolar range (Paper II). Additionally, an alternative strategy for secretion and conjugation of various reporters using Sortase A directly from the E. coli surface was developed, facilitating fast parallel downstream analysis of labeled binders (Paper III).

The three following studies describe rational design and development of Affibody-based tracers against two cancer-associated targets (EGFR and HER3) for molecular imaging. First, two anti-HER3 Affibody molecules were conjugated with a NOTA chelator and labelled with 111In. As one of the first HER3 imaging agents, both conjugates could specifically bind to HER3-expressing BT-474 xenografts in mice using SPECT (Paper IV). Furthermore, labeling with 68Ga for PET imaging showed that tumor uptake correlated with HER3 expression in vivo, suggesting that anti-HER3 tracers could potentially be used for patient stratification (Paper V). The last study describes the development and investigation of an anti-EGFR Affibody-based imaging agent. Labeled with 89Zr, the Affibody tracer demonstrated higher tumor uptake 3 hours post injection than the anti-EGFR antibody cetuximab 48 hours post injection (Paper VI).

In conclusion, this thesis describes new tools and results that will hopefully contribute to the development of affinity proteins for biotechnology, therapy or medical imaging in the future.

Keywords: Directed evolution, microbial display, E. coli, Affibody molecule, autotransporter, medical imaging, HER receptor family, Sortase A

VI

List of appended publications

This thesis is based on the work described in the following publications or manuscripts that is referred to in this thesis in roman numerals (I-VI). The papers presented here can be found in the appendix.

Paper I Fleetwood F.*, Andersson K.G.*, Ståhl S., Löfblom J. An engineered autotransporter-based surface expression vector enables efficient display of Affibody molecules on OmpT-negative E. coli as well as protease-mediated secretion in OmpT-positive strains. Microbial Cell Factories 2014

Paper II Andersson K.G., Persson J., Ståhl S., Löfblom, J. Autotransporter-mediated display of a naïve Affibody library on the outer membrane of E. coli. Manuscript

Paper III Andersson K.G., Sjöstrand N., Löfblom, J. Coupled release and site-specific conjugation of Affibody molecules from the surface of E. coli using Sortase A. Manuscript

Paper IV Andersson K.G., Rosestedt M., Varasteh Z., Malm M., Sandström M., Tolmachev V., Löfblom J., Ståhl S., Orlova A. Comparative evaluation of 111In-labeled NOTA-conjugated affibody molecules for visualization of HER3 expression in malignant tumors. Oncology Reports 2015

Paper V Rosestedt M., Andersson K.G., Mitran B., Tolmachev V., Löfblom J., Orlova A., Ståhl S. Affibody-mediated PET imaging of HER3 expression in malignant tumours#. Scientific Reports 2015

Paper VI Garousi J.*, Andersson K.G.*, Mitran B., Pichl M.L., Ståhl S., Orlova A., Löfblom J., Tolmachev V. PET imaging of epidermal growth factor receptor expression in tumours# using 89Zr-labelled ZEGFR:2377 Affibody molecules. International Journal of Oncology 2016

*These authors contributed equally

#These papers use British spelling of the word tumor

All papers are reproduced with permission from the copyright holders.

Ken G Andersson

VII

Contribution to appended publications

Paper I Performed all experimental work together with Filippa Fleetwood. Planned the experiments, analysed the results, prepared all the figures and prepared the manuscript together with co-authors.

Paper II Performed the engineering of the vector, construction of the library and selections against HER2, HER3 and IL3RA. Performed the SPR analysis together with Jonas Persson. Planned the experiments, analysed the results, prepared all the figures and prepared the manuscript together with co-authors.

Paper III Performed all experimental work together with Nanna Sjöstrand. Supervised the master’s thesis student Nanna Sjöstrand. Planned the experiments, analysed the results, prepared all the figures and prepared the manuscript together with co-authors.

Paper IV Performed all experimental work together with co-authors. Analysed the results, prepared figures and wrote the manuscript together with co-authors.

Paper V Planned and performed the production and characterization of the imaging agent. Prepared the manuscript together with co-authors.

Paper VI Supervised, planned and performed production and characterization of the imaging agent. Prepared figures and the manuscript together with co-authors.

VIII

Related publications

1. Orlova A., Malm M., Rosestedt M., Varasteh Z., Andersson K.G., Selvaraju R.K., Altai M., Honarvar H., Strand J., Ståhl S., Tolmachev V., Löfblom J. Imaging of HER3-expressing xenografts in mice using a (99m)Tc(CO) 3-HEHEHE-Z HER3:08699 affibody molecule. European Journal of Nuclear Medicine and Molecular Imaging 2014

2. Garousi J., Honarvar H., Andersson K.G., Mitran B., Orlova A., Buijs J., Löfblom J., Frejd F.Y., Tolmachev V. Comparative evaluation of Affibody molecules for radionuclide imaging of in vivo expression of carbonic anhydrase IX. Molecular Pharmacology 2016

3. Andersson K.G., Oroujeni M., Garousi J., Mitran B., Ståhl S., Orlova A., Löfblom J., Tolmachev V. Feasibility of imaging of epidermal growth factor receptor expression with ZEGFR:2377 affibody molecule labeled with 99mTc using a peptide-based cysteine-containing chelator. International Journal of Oncology 2016

4. Nosrati M., Solbak S., Nordesjö O., Nissbeck M., Dourado D.FAR, Andersson K.G., Housaindokht M.R., Löfblom J., Virtanen A., Danielson U.H., Flores S.C. Insights from engineering the Affibody-Fc interaction with a computational-experimental method. Protein Engineering, Design and Selection 2017

5. Garousi J., Andersson K.G., Dam J.H., Olsen B.B., Mitran B., Orlova A., Buijs J., Ståhl S., Löfblom J., Thisgaard H., Tolmachev V. The use of radiocobalt as a label improves imaging of EGFR using DOTA-conjugated Affibody molecule. Scientific Reports 2017

6. Rosestedt, M.*, Andersson K.G.*, Mitran B., Rinne S.S., Tolmachev V., Löfblom J., Orlova A., Ståhl S. Evaluation of a radiocobalt-labelled affibody molecule for imaging of human epidermal growth factor receptor 3 (HER3) expression in malignant tumours. International Journal of Oncology 2017

Ken G Andersson

IX

Abbreviations

ABD Albumin-binding domainABP Albumin-binding proteinAIDA-I Adhesin involved in diffuse adherenceCDRs Complementary determining regionDa DaltonDFO DesferoxamineDNA DeoxyribonucleotideDOTA 1,4,7,10-tetraazacyclododecane-1,4,7,10-tetraacetic acidEGF Epidermal growth factorEGFR Epidermal growth factor receptorELISA Enzyme-linked immunosorbent assayEPR Enhanced permeability retentioneSrtA Ehanced Sortase AFab Fragment, antigen bindingFc Fragment, crystallizableFACS Fluorescence-activated cell sortingFDA Food and drug administrationGFP Green fluorescent proteinHNSCC Head and neck squamous-cell carcinomaHSA Human serum albuminIL3RA Interleukin 3 receptor subunit alphaIM Inner membraneka Association rate constantkd Dissociation rate constantKD Equilibrium dissociation constantmAbs Monoclonal antibodyNOTA 1,4,7,10-tetraazacyclododecane-1,4,7,10-tetraacetic acidNSCLC Non-small cell lung cancerOM Outer membraneOmpT Outer membrane protease TPET Positron emission tomographypi Post-injection RTK Recetor tyrosine kinasescFvs Single chain fragment variableSPECT Single-photon emission computed tomographyT5SS Type V secretion systemsZwt Affibody molecule binding IgG

X

Populärvetenskaplig sammanfattning

Proteiner är mångfacetterade biomolekyler som kan ha mängder av olika egenskaper, som till exempel att transportera syre i kroppen eller bryta ner stärkelse när man tvättar kläder. Proteiner består av en sammanlänkad kedja av aminosyror och aminosyrornas inbördes ordning bestämmer hur de veckas, vilken struktur de får och därmed också deras egenskaper. Olika aminosyrasekvenser och veckningar kan vara väldigt varierande vilket möjliggör den stora spännvidden av olika funktioner.

Vår forskning syftar till att skapa och förändra proteiner och därmed förse dem med nya egenskaper. Ett tillämpningsområde är att generera bindarproteiner som kan binda till biomarkörer på cancerceller och därmed målsöka tumörer i syfte att skapa proteinläkemedel eller hjälpmedel för avbildningsdiagnostik. I avbildningsdiagnostik så förses det bindande proteinet med en radioaktiv atom, som kan injiceras så att proteinet målsöker och ackumuleras i tumören. Hög ackumulering i tumören gentemot andra organ medför att tumören kan avbildas med antingen positronemissionstomografi (PET) eller single-photon emission computed tomography (SPECT) tekniker. Dessa tekniker möjliggör för läkaren att utröna var tumören är lokaliserad och var eventuell spridning (metastaser) föreligger.

Denna avhandling beskriver utvecklandet av en metod för att kunna mutera och identifiera nya proteiner som kan binda till olika typer av biomarkörer för cancer (Artikel I och III). Sådana muterade proteiner mot biomarkörer som är associerade med bland annat bröstcancer och hjärntumörer har identifierats med hjälp av den nya metoden (Artikel II). De identifierade bindarna har karaktäriserats och har visats kunna binda biomarkörerna vilket inger förhoppningar om att dessa nya proteiner skulle kunna användas för avbildningsdiagnostik i framtiden. Utöver studierna kring den nya metoden i arbete I-III, så har vi även vidareutvecklat två olika bindarproteiner mot bland annat bröstcancer (Artikel IV och V) och skivepitelcancer (Artikel VI) som båda har visat sig effektiva för att lokalisera och identifiera tumörer i möss. I jämförelse med andra vanliga proteiner som används för avbildningsteknik så ger våra proteiner signifikant bättre kontrast i signal mellan tumör och frisk vävnad redan efter 3 timmar. Detta innebär att en patient kan avbildas kort efter injektionen vilket förenklar diagnostiken både för patienten och sjukvården. Den höga kontrasten skulle även kunna leda till en högre noggrannhet med avseende på att lokalisera huvudtumören och alla metastaser.

Sammanfattningsvis så har metoderna beskrivna i denna avhandling bidragit med nya verktyg för att utveckla nya proteiner som kan användas till avbildningsteknik och således förhoppningsvis inom en relativt snar framtid kunna förbättra diagnostiken av cancerpatienter.

Ken G Andersson

XI

Populärwissenschaftliche ZusammenfassungProteine, auch Eiweiße genannt, stellen eine Gruppe von Biomolekülen dar, die zahlreiche verschiedene Eigenschaften haben können. Dazu zählen das Vermögen, Sauerstoff im Körper zu transportieren oder beim Wäschewaschen Schmutzstoffe wie Stärke zu zersetzen. Proteine bestehen aus einer zusammenhängenden Kette von Eiweißbausteinen, sogenannt-en Aminosäuren. Die Reihenfolge der Aminosäuren – auch Sequenz genannt - innerhalb dieser Kette bestimmt, wie sich die Proteine falten, d.h. welche dreidimensionale Struktur sie annehmen, und somit ihre Eigenschaften. Je nach Protein kann die Aminosäuresequenz und dreidimensionale Struktur sehr unterschiedlich sein, was die große Vielfalt an Funk-tionen ermöglicht.

Ziel unserer Forschung ist es, Proteine zu herzustellen oder zu verändern und sie mit neuen Eigenschaften zu versehen. Ein Anwendungsgebiet ist das Hervorbringen sogen-annter Bindeproteine, die Krebszellen anhand bestimmter Merkmale (Biomarker) erkennen und binden können. Damit können sie Tumoren aufspüren und als Proteinpharmazeutika oder Hilfsmittel für die diagnostische Bildgebung verwendet werden. Für die diagnostische Bildgebung wird das Bindeprotein mit einem radioaktiven Baustein versehen und in den Körper eingespritzt. Dort sucht es den Tumor auf und sammelt sich darin an. Eine Anre-icherung des Bindeproteins im Tumor ermöglicht das Sichtbarmachen des Krebses mithilfe bildgebender Verfahren wie Positronen-Emissions-Tomographie (PET) oder Einzelpho-tonen-Emissionscomputertomographie (SPECT). Dies ermöglicht es Ärzten, herauszu-finden, wo genau sich der Tumor befindet und ob er sich im Körper durch Metastasen verbreitet hat.

Diese Doktorarbeit beschreibt die Entwicklung eines Verfahrens zur Veränderung und Identifizierung neuer Proteine, die verschiedene Arten von Krebs-Biomarkern binden können (Artikel I und III). Mithilfe dieser neuen Methode, konnten wir Proteine identi-fizieren und charakterisieren, die u.a. bei Brustkrebs und Gehirntumoren entsprechende Biomarker binden können (Artikel II). Dies gibt Anlass zur Hoffnung, dass diese neuen Proteine möglicherweise in Zukunft für die diagnostische Bildgebung verwendet werden können. Zusätzlich zur Entwicklung des in Artikel I-III beschriebenen Verfahrens haben wir zwei Bindeproteine weiterentwickelt, die unter anderem Brustkrebs (Artikel IV und V) und Stachelzellkrebs erkennen können. In Mausstudien konnten wir zeigen, dass beide Bindeproteine die Tumore aufspüren und identifizieren können.

Im Vergleich zu anderen Proteinen, die herkömmlicherweise für die bildgebende Diag-nostik verwendet werden, können unsere Proteine wesentlich besser zwischen Kreb-sgewebe und gesundem Gewebe unterscheiden, und das kann schon drei Stunden nach der Injektion gemessen werden. Dies bedeutet, dass ein Patient schon relativ kurze Zeit nach der Injektion untersucht werden kann. Dadurch wird das diagnostische Verfahren sowohl für den Patienten als auch für Ärzte und Pflegepersonal vereinfacht. Das Vermögen, besser zwischen Krebs und gesundem Gewebe unterscheiden zu können, könnte zudem zu einer besseren Genauigkeit bei der Lokalisierung von Ursprungstumor und Metastasen beitragen.

Zusammenfassend lässt sich sagen, dass die in dieser Doktorarbeit beschriebenen Methoden neue Werkzeuge für die Entwicklung von Proteinen liefern, die für bildgebende diagnos-tische Verfahren eingesetzt werden können. Dies kann hoffentlich in naher Zukunft zu Verbesserungen bei der Diagnostik für Krebspatienten führen.

XII

ContentsIntroduction ............................................................................................ 1

Proteins in therapy and diagnostics .......................................................... 21.1 What is a protein? ............................................................................................... 21.2 Protein therapeutics and diagnostics ................................................................... 31.3 Companion diagnostics ....................................................................................... 5

Cancer .........................................................................................................................72.1 What is cancer? ................................................................................................... 72.2 Biomarkers for cancer-targeted therapeutics and diagnostics ............................... 82.3 The epidermal growth factor receptor family ...................................................... 9

Imaging of cancer ......................................................................................................133.1 Molecular imaging ............................................................................................ 133.1.1 Molecular imaging using small proteins or peptides .......................................... 15

Affinity molecules ....................................................................................................164.1 Antibodies ........................................................................................................ 164.2 Alternative scaffolds ......................................................................................... 184.2.1 Affibody molecule ............................................................................................. 21

Engineering approaches ............................................................................................225.1 Directed evolution ............................................................................................. 225.2 Rational design.................................................................................................. 235.3 Random mutagenesis ........................................................................................ 245.4 Site-directional saturation mutagenesis ............................................................ 24

Selection methods .....................................................................................................266.1 Phage display ..................................................................................................... 266.2 Cell display........................................................................................................ 276.3 Yeast display ...................................................................................................... 296.4 E. coli display..................................................................................................... 306.4.1 Type V secretion systems .................................................................................. 31

Ken G Andersson

XIII

Present investigation .............................................................................. 35

Short summary of the papers in the thesis .............................................. 36

7.1 Paper I - An engineered autotransporter-based surface .......................... expression vector enables efficient display of Affibody .......................... molecules on OmpT-negative E. coli as well as ...................................... protease-mediated secretion in OmpT-positive strains ............................ 38

7.2 Paper II - Autotransporter-mediated display of a naïve .......................... Affibody library on the outer membrane of E. coli ............................. 44

7.3 Paper III – Coupled release and site-specific ........................................ conjugation of Affibody molecules from the surface of .......................... E. coli using Sortase A ..................................................................... 49

7.4 Paper IV - Comparative evaluation of 111In-labeled ............................... NOTA-conjugated Affibody molecules for visualization ........................ of HER3 expression in malignant tumors .......................................... 53

7.5 Paper V - Affibody-mediated PET imaging of HER3 ............................ expression in malignant tumours ...................................................... 57

7.6 Paper VI - PET imaging of epidermal growth ....................................... factor receptor expression in tumours using 89Zr- .................................. labelled ZEGFR:2377 Affibody molecules ......................................... 60

7.7 Concluding remarks and future perspectives ...................................... 63

Bibliography ......................................................................................... 66

Acknowledgement .................................................................................. 79

Introduction

2

Chapter1

Proteins in therapy and diagnostics

1.1 What is a protein?Our body consists of approximately 4x1013 individual cells that together ultimately define us as the mammal Homo sapiens1. Each of these cells contains DNA, encoding roughly 20 000 proteins, defined by UniProt, allowing the cell to not only sustain itself but also to enable a myriad of other functions, such as communication with other cells and helping you recover after an intoxication or engulfment of a virulent particle that has entered the body2–4. Without proteins, life as we know it, would not be possible.

But how can one class of molecules be essential for life and at the same time be so versatile? Proteins are high-molecular weight organic molecules that on a chemical level consist of a linear chain of amino acid residues covalently connected by peptide bonds5. While there are 20 unique, naturally occurring amino acids within our genetic code, they share common denominators, containing an amino acid-specific side chain (R group) as well as an amine- (NH2) and a carboxyl group (-COOH). Peptide bonds between amino acids are formed by condensation reactions of an amine moiety with another amino acid’s carboxyl moiety, generating a polypeptide. Multiple condensation reactions of amino acids increase the size of the polypeptide to a point when it is thermodynamically unfavorable to remain as a linear, unstructured polypeptide. The protein can fold into structures classified by Kaj Ulrik Linderstrøm-Lang in 1952 as primary, secondary, tertiary and quaternary protein structures6. According to Linderstrøm-Lang, the protein’s amino acid sequence is determining its primary structure. Similarly, the secondary and tertiary structure defines how the protein is folding within a three-dimensional space. The quaternary structure refers to how multiple subunits of the protein are arranged; however, this is only applicable for larger proteins.

Because these 20 different amino acids contain functional groups with distinct chemical properties, the sequence of which these are located within the protein vastly influences the folding of the protein. Protein folding determines its function, and this can be directly observed in nature. For instance, oxygen transport and storage in mammals is achieved by heme proteins such as myoglobin. Myoglobin is a protein consisting of 153 amino acids and folds into a globular protein that adopts eight alpha helices (Fig.1). Folding allows the protein to coordinate a heme group which in turn permits binding of oxygen molecules. Without folding, the coordination of the heme group, thus binding of oxygen, would be impossible7. Folding into different structures is one of the reason that proteins are the most plastic macromolecule in our cells with all the abilities essential for life8.

3

Ken G Andersson

1.2 Protein therapeutics and diagnosticsProteins as sort of drugs for curing diseases have been utilized since the beginning of rudimentary medicine. Perhaps not fully understanding that the effector molecule was a protein at that time, doctors in ancient Greece wrote about the healing effect of snake venom9. Throughout time, this has developed into a field named biopharmacology that studies all pharmaceutical drugs that have been manufactured or extracted from biological sources. One of the most prominent sub-categories is biologics or biopharmaceuticals. Biologics refers to recombinant production of a protein drug by a host such as bacteria, yeast or mammalian cells10. One of the first examples of therapeutic use – the recombinant production of human insulin in a bacterial host – was reported in 197911. Until then, insulin was extracted from porcine or bovine pancreases, which not only was time-consuming, but the extracted insulin could give immunogenic side-effects12. Only three years later, the

Figure 1. Amino acids can through condensation reactions build large globular proteins with complex functions. A) All amino acids contain an amine group, a carboxyl group and a specific side chain. Amino acids can form peptide bonds with each other by linking the amine group to another amino acid’s carboxyl group. B) Through such condensation reactions, single amino acids assemble into peptides, and the sequence deter-mines its primary structure. C) If energetically favored, the primary structure can fold into different secondary structure elements, such as alpha helices (upper) and beta-sheet (lower). D) Myoglobin with 153 amino acids folds into eight alpha helices connected by loops which allow the protein to coordinate a heme group and supply muscle tissue with oxygen in mammals.

4

US food and drug administration (FDA) approved recombinant insulin as a drug under the product name Humulin, allowing the drug to be sold as a prescription drug within the United States, eventually benefitting millions of patients. The success of recombinant insulin production created an enthusiastic and hopeful atmosphere in the early 1980s, which unfortunately was followed by a decade of many unsuccessful clinical trials using non-humanized monoclonal antibodies against cancer13.

Currently, most therapeutically used molecules are small chemically manufactured compounds, e.g. Aspirin, containing acetylsalicylic acid, an effector molecule of 180 Dalton (Da). These small molecules can typically be administered orally and reach intracellular targets in comparison to biologics. The much more complex biologics is on the other hand more specific and allow extended circulation times, and has furthermore the potential to deliver payloads and/or recruit immune cells for additional effector functions14. These features have been the driving force for biologics to become the fastest growing class of pharmaceutical molecules, with the antibody Humira being the top grossing pharmaceutical with 13 billion dollar revenue in 2016 in United States alone15,16. The generally high specificity and affinity of biologics is one of the main reasons for the high clinical success rate in oncology with 18-24% depending on the antibody’s origin, as compared with success rates of merely 5% that have been reported for small molecules17–19.

Proteins binding other molecules are common in nature. In fact, one major characteristic of proteins is their ability to recognize (specificity) and interact (affinity) with other molecules. The interaction rate is determined by intermolecular forces defined via ionic bonds, hydrogen bonds and Van der Waals forces between the protein and the molecule. Generally, high affinity between the protein and the molecule correlates with stronger intermolecular forces permitting longer retention time or faster binding for the complex8. When two molecules [A] and [B] interact [AB], the kinetics of the reaction is typically expressed as two constants, (i) the on-rate, association rate constant (ka; M

-1s-1), and (ii) the off-rate, dissociation rate constant (kd; s

-1). Together, the ratio between the two constants (ka/kd) is expressed as the dissociation equilibrium constant KD (M). The constant KD describes the state where the rate of forming the interaction [AB] equals the rate of dissociation into molecule [A] and [B]. Another definition of KD is the concentration of the binder at which 50 % of the two molecules interact with each other. The dissociation equilibrium constant can be identical between two interactions while the ka and kd are vastly different. The constant is regularly considered very useful when ranking and evaluating the strength of the interactions. Proteins normally interact with a KD in the millimolar to femtomolar range. Since the affinity determines the concentration required to obtain 50% binding, a higher affinity is usually very favorable when the concentration of the specific binding partner is low20.

5

Ken G Andersson

For proteins aimed for therapeutic purposes, high affinity and specificity are undoubtedly desireable features. These properties are also very attractive for proteins used as diagnostics. Harnessing the features of protein interaction to determine concentrations of molecules or detecting dangerous microorganisms is extremely valuable. According to estimations, the molecular diagnostic market was worth 4 billion US dollars in 201121. The largest contributor to this market is ‘over the counter’ products such as glucose and pregnancy tests that can be readily tested at home. In 1976, the pregnancy test was introduced as one of the first immunoassays using antibodies specific to hCG. Released before FDA approval, the test reported unprecedented 98 % true positive diagnoses22. The selectivity arises from having two different antibodies both targeting hCG at different regions, named epitopes. One antibody is immobilized onto a surface and the other is conjugated with a reporter molecule. Signal can only arise in a pregnancy test if both the immobilized and the reporter antibody bind the same hCG molecule in a sandwich manner 23. This method named enzyme-linked immunosorbent assay (ELISA) was invented at Stockholm University in 1971 and is now routinely used in hospital diagnostics24. Since ELISA can evaluate the presence of a protein with a very high sensitivity it has been applied in a multitude of assays, e.g. HIV-testing25, to measure inflammation26, malaria-testing27and, most impoartanyl, to determine gluten intolerance28.

1.3 Companion diagnosticsA companion diagnostic is a medical device that can provide information that is essential for safe and effective use of a certain drug. For instance, a glucose meter determines the concentration of glucose in the blood, making it possible for diabetic patients to regulate injection of insulin and thus minimizing the side-effects. Although diagnostic medical devices that can be purchased ‘over the counter’ are invaluable to determine simpler conditions, limited success has been achieved when commercializing detection of proteins, nucleic acids or key metabolites at home29. More complex measurements are routinely performed in hospitals and require an expert for correct analysis. Treating and diagnosing a patient today involves numerous tests to assign the patient to a particular treatment category based on the predicted response. This is termed precision medicine, combining sophisticated diagnostic methods to tailor an evidence-based stratification of selected patients. The high need for diagnostic tools that can evaluate evidence-based stratification gave rise to the field of companion diagnostics. Before, pharmacotherapy was mainly characterized by ‘trial and error’ where success rates after treatment were consequently lower30. Using companion diagnostics to define an appropriate therapeutic approach and to determine if the patient is likely to have an increased risk for serious side-effects has increased response rates in clinical testing31. Data evaluating integration of molecular diagnostics with therapeutics indicates that having a companion diagnostic developed to the therapeutic gives a response rate of 41-80.2 % compared to 6.8-45 % for a therapeutic without a companion diagnostic32. Although companion diagnostics is a relatively novel field, first mentioned in the literature in 2006, many researchers developing drugs are currently also developing companion diagnostics for increased success rate in the clinics33.

6

An example of where companion diagnostics are applied is in breast cancer patients. Approximately 25% of breast tumors are overexpressing the protein receptor tyrosine kinase erbB-2 (HER2), which is associated with poor prognosis34. Treatment using the HER2-targeting antibody trastuzumab has shown reduction in risk of death by 30% and tumor recurrence by 50% compared to treatment without Traztuzumab34,35. However, therapy targeting the HER2 only works for tumors with HER2 pathway dependency and increased expression36. Therefore, companion diagnostics have been developed to measure HER2 expression in a tumor biopsy enabling the physician to determine which treatment is most suitable for the individual patient37.

7

Ken G Andersson

Chapter2

Cancer

2.1 What is cancer?Cancer is a generic term for a group of diseases involving abnormal cell growth that can affect almost any part of the body. According to World Health Organization (WHO), cancer is the second leading cause of mortality in the world causing nearly 1 in 6 deaths38. Currently, prostate and breast cancer are the types of cancer with the highest incidence in Sweden (23% and 13%, respectively), while highest mortality is observed for the more aggressive lung and colorectal cancer39. Cancer begins with genetic changes in one or a small group of cells. Normally, the human body is regulating growth and proliferation of each individual cell within the body. A neoplasm, more commonly called tumor, is occurring when a cell temporally grows unrestrictedly over their normal counterparts. While there are many groups of benign neoplasms which lack the ability to metastasize, cancer constitutes malignant neoplasms capable of invading neighboring tissue40. This invasion results in about 90% of the cancer-associated mortality41. Interestingly, in contrast to normal cells, cancer cells share many features with unicellular organisms. Many tumorigenic cells meet their energy needs by anaerobic fermentation and by acquiring fuel by phagocytosis of other cells, surprisingly similar to how amoeba acquire energy42.

In 2000, Hanahan and Weinberg released a beautiful paper providing a logical framework for understanding the incredible diversity of the neoplasm diseases called cancer43. They proposed that every cancer requires six unique capabilities that allow for limitless growth and tumorigenesis. About a decade later, in 2011, the increasing body of research in the field meant that the authors could report four additional capabilities. The first six capabilities or hallmarks of cancer were focused on how single cells allow themselves limitless replication by evading apoptosis and growth suppressors, self-signaling, oxygen supply through vessels, and invading other tissues. Later research indicated that cancers are even more complex, illustrated by the typical ability to alter energetic systems, genome instability, creating tumor promoting micro-environments and avoiding the immune system43,44. Together, the ten hallmarks proposed in the two papers enabled researchers and physicians to identify what differentiates a normal cell from a tumorigenic cell in a straightforward manner (Fig 2). However, the paths that cells take to become malignant are highly variable. Even within a given cancer type, such as the common invasive ductal carcinoma, responsible for nearly 80 % of all diagnosed preinvasive breast cancers, there is great heterogeneity within the group45,46. Additionally, due to high mutation rates and clonal diversity, heterogeneity can also be found intratumorally, further diversifying the disease47,48.

8

2.2 Biomarkers for cancer-targeted therapeutics and diagnosticsDeveloping targeted therapeutics for cancer is tremendously challenging. Due to tumor heterogeneity, target-specific therapy attempting to exploit a cancer’s dependence on critical signaling pathways often render a part of the tumor resistant to the drug49. It has been reported in clinics that the use of trastuzumab that blocks HER2 signaling often results in upregulation of other signaling pathways resulting in resistance to therapy50. Additionally, tumors share essential pathways with normal tissue, increasing the risk of side-effects when targeting an important mechanism for cell survival or growth. However, treating cancer is not impossible. Since tumorigenic cells depend on limitless replication, mutations and overexpression of proteins are common51,52. If these proteins are expressed in- or outside the tumorigenic cell and can identify a tumorigenic process in a reproducible manner they are called a cancer biomarker53.

Figure 2. The ten hallmarks necessary for tumorigenic progression of cells as suggested by Hanahan and Weinberg43

9

Ken G Andersson

The discovery of novel cancer biomarkers is an expanding research field that can be somewhat simplistically divided into two main subgroups: predictive and prognostic. A good prognostic biomarker enables monitoring of the cancer, being able to determine potential malignancy as well as disease prognosis. Predictive biomarkers provides information supporting the therapeutic decision and probability obtaining a response to treatment54. As mentioned earlier, a relevant biomarker in breast cancer patients is HER2 which can be used both as a predictive and prognostic biomarker. Tumors overexpressing HER2 have increased angiogenesis, proliferation and inhibition of apoptosis signifying a worse prognosis compared to the negative tumors55. Using companion diagnostics and predictively determining expression of HER2 allows for targeting therapy using for example trastuzumab56.

2.3 The epidermal growth factor receptor familyPerhaps one of the most important biomarker families is the epidermal growth factor (EGF) receptor(R) family and a large part of this thesis is based on studies, targeting these receptors. This family consists of four receptor tyrosine kinases (RTK), all structurally related to the first discovered member, EGFR. All four family members, EGFR, HER2, HER3 and HER4, are structurally comprised of a single glycosylated polypeptide. Similar to most RTKs, the receptors are characterized by an extracellular binding region, a transmembrane-spanning surface anchoring region and a cytoplasmic region mediating downstream signaling. Somewhat similar to an allosteric enzyme, each EGFR family member, except HER2, contains a regulatory domain that binds an allosteric regulator, typically an EGF-like structured growth factor ligand peptide. Ligand binding induces pronounced conformational changes of the receptor and enables the formation of heterodimers, homodimers or even oligomers57 (Fig 3).

Together, the family of receptors can be activated by eleven different growth factors, serving as specific agonists, where EGFR and HER4 can homodimerize upon binding of specific signaling molecules. Upon ligand binding , the receptors dimerize and generate intracellular signals culminating in for example cell migration, proliferation, growth, adhesion and differentiation58. In contrast to the other receptors, HER2 does not bind any known ligand with high affinity but can generate downstream signaling through homodimerization or heterodimeric complexes with other, ligand-bound, family members. Interestingly, the lack of a ligand for HER2 results in that the receptor is constitutively in an activated conformation59.

10

Furthermore, the HER2 receptor has the strongest catalytic kinase activity of all the family members and heterodimers with other receptors yield the strongest downstream signaling compared to other heterodimeric complexes within the family60. This was best exemplified with the elgegant paper by Sliwkowski et al in 1994, discovering that the non-autonomous receptors HER2 and HER3 together with the ligand heregulin dramatically increase downstream signaling in tumorigenic cells61. The study indicated that HER3 and heregulin cannot signal without the presence of a HER2 receptor; meanwhile only HER2 receptors cause little or no stimulation of tyrosine phosphorylation in response to heregulin. This indicates that a heterodimeric complex between the heregulin-activated HER3 and the constitutively active HER2 is important for increased downstream signaling.

Figure 3. Crosstalk between receptors in the EGFR family. Similar to all RTKs, the EGFR family receptors consist of an extracellular binding region, a transmembrane-spanning surface anchoring region and a cytoplas-mic region for downstream signaling. Family members such as EGFR, HER3 and HER4 require ligand binding to induce hetero or homodimerization. Dimerization leads to phosphorylation of the cytoplasmic kinase domain followed by specific signal transduction depending on receptor complex. While the complexes depicted here have other additional signaling pathways, they all share the RAS-RAF pathway that leads to nuclear transcrip-tion and increased migration, proliferation and survival of the cell.

11

Ken G Andersson

All four EGFR family members are heavily involved in signaling throughout development and adulthood. The plasticity derived from homo- and heterodimerization in addition to the different allosteric ligands makes it an important signaling pathway for many different processes such as organogenesis and cell-to-cell interaction62. However, the high versatility and potent signaling allows tumors to exploit these pathways for tumorigenic purposes such as increased growth and cell migration. Due to the extensive cross-talk between each receptor, resistance against monotherapies targeting one of the receptors is possible by increasing or decreasing expression of other receptor family members or ligands.

Overexpression of EGFR by gene amplification is frequent in human malignancies such as head and neck cancer and squamous cell lung carcinomas63. The increased expression in primary tumors has shown potential as a prognostic marker since it is associated with higher cell proliferation and reduced survival64. An overexpression of EGFR has been shown to correlate with trastuzumab resistance in breast cancer cells65. In addition to membrane-associated signaling, EGFR in tumors has been shown to function as transcription factor by physically interacting with the signal transducer and activators of transcription 3 (STAT3) inside the nucleus66. Several therapeutic monoclonal antibodies have been developed or are currently in clinical development against EGFR, such as cetuximab and panitumumab approved by FDA in 2004 and 2006, respectively67. To date, the only targeted agent against EGFR-expressing head and neck squamous cell carcinoma (HNSCC) is cetixumab68. Additionally, the EGFR-targeting zalutumumab is being developed against HNSCC and is currently in phase III clinical trial awaiting FDA approval69.

One of the best studied receptors associated with overexpression in cancer is HER2. This receptor can be found in several types of cancers, such as breast cancer, lung cancer, pancreas cancer, colon cancer, and many more. Overexpression correlates with increased risk of metastases, tumor size, aneuploidy and lack of steroid hormone receptors. Interestingly, high HER2 expression levels are more commonly reported for early stages of breast cancer, indicating that HER2 alone is not sufficient for malignant progression70. Trastuzumab was first approved by FDA against breast cancer in 1998, followed by a second therapeutic antibody named pertuzumab, which was approved in 2012. While trastuzumab is targeting subdomain IV on HER2, pertuzumab can prevent HER2 dimerization by binding to subdomain II and thus inactivates HER2-dependent signaling pathways71. Additionally, a so-called antibody-drug conjugate (ADC), based on trastuzumab conjugated to the cytotoxic drug emtansine (DM1), was recently approved by FDA and show increased overall survival amoung patients with HER2-positive metastatic breast cancer72.

The neuregulin-binding receptor HER3 is becoming an increasingly interesting prognostic and predictive biomarker. Expression of HER3 can occur in breast, colon, gastric, prostate and oral squamous cell cancer73. While not overexpressed to the same extent as EGFR or HER2, elevated expression levels have been shown to correlate with poor overall survival74. Additionally, loss of HER3 expression reportedly decreases the viability of HER2-overexpressing breast cancer cells, indicating that the receptor may be a good prognostic marker75. Resistance against the EGFR-specific monoclonal antibody

12

cetuximab has been shown to be mediated by overexpression of heregulin, resulting in increased dimerization between HER2 and HER3 and subsequent rescue of downstream signaling76. Targeted therapies against HER3 are currently under clinical development, comprising both monoclonal antibodies and tyrosine kinase inhibitors (TKIs). Antibodies such as patritumab, targeting HER3 on non-small cell lung cancer (NSCLC), have reached phase III where it has demonstrated increased progression free survival in combination with the TKI erlotinib. Another HER3-specific monoclonal antibody in development, called seribantumab, has reached phase II for treatment of breast cancer and has shown to be more effective on high heregulin/low HER2 expressing tumors77.

Overexpression of HER4 is a potential predictive marker for and has also been shown to mediate resistance against trastuzumab by activation and overexpression of the receptor, resulting in poor patient prognosis78. Overexpression of HER4 has been correlated with resistance to chemotherapy and short disease free survival in limb soft-tissue sarcoma. HER4 is frequently present in several different cancer types such as prostate, ovarian, endometrial cancers. However, published clinical data about the role of HER4 in tumorigenesis and tumor progression is somewhat unclear79. Currently there are no FDA-approved HER4 targeting antibodies and the development of antibodies against this target is somewhat limited compared to the other receptor family members. Nevertheless, a monoclonal antibody mAb 1479 has recently been described and can specifically target tumor-associated HER4 and stimulate degradation80.

13

Ken G Andersson

Chapter3

Imaging of cancer

3.1 Molecular imagingPapers II, IV, V and VI in this thesis are all on development of new probes for so-called molecular imaging. The field of molecular imaging is the concept of administering a radiopharmaceutical to the patient that later can be detected using external detectors such as gamma cameras or PET scanners for localization of for example tumor lesions within the body. In vivo molecular imaging has the potential to improve diagnosis of many diseases by quantitatively measuring abnormal cellular and molecular alterations. Moreover, depending on the biomarker, this method could provide valuable prognostic or predictive information. Currently, biopsies are the most common diagnostic method to determine biomarker prevalence in tumor cells; however repeated biopsies increase the risk of inducing metastasis in the unstable tumor environment81. Additionally, some biopsies are considerably harder to perform, such as in the brain, liver and lung and could lead to severe discomfort or even be harmful for the patient. Injection of radiopharmaceuticals, followed by imaging of therapeutically-relevant targets can be repeated multiple times without substantial side-effects and provides a global and anatomically correct view of potential metastatic lesions in the body. Additionally, monitoring changes in expression level of the biomarkers during therapy could help avoid over- and undertreatment82. Compared to biopsies, this method is usually far more expensive, requiring access to radionuclides and expensive detection instrumentations, such as positron emission tomography (PET) and single photon emission computer tomography (SPECT). While SPECT is more widely available and can use more conventionally available radionuclides, PET has generally better sensitivity, higher resolution and can more accurately quantify radioactivity concentration in vivo.

Molecular imaging using proteins or peptides (called probes) requires three distinct properties: i) a protein or peptide that can specifically bind a relevant biomarker, ii) a chelator, covalent linkage or radionuclide binding motif that physically links the radionuclide to the protein or peptide, Iii) a radionuclide of choice that can be accommodated by the chelator and measured with currently available imaging modalities (Fig 4). The choice of radionuclide will for example influence which type of chelator that can be used and which type of detection method that is suitable. Furthermore, factors such as physical half-life, availability at the hospitals, photon or particle energy and the selective deposition of energy in tissue also need to be considered.

The residualizing property of the radionuclide is perhaps one of the most important aspects of a radionuclide if the probe is rapidly internalized. Rapid internalization and lysosomal degradation are associated with some biomarkers and probes. A residualizing radionuclide will not pass through the membrane of the cell after degradation, which leads

14

to accumulation and increased retention inside the tumor cell, resulting in better contrast. However, it is important to note that residualizing radionuclides also increase retention in healthy tissues that has internalized the probe or biomarker. High kidney uptake is often an issue for many small probes, such as Single chain variable fragments (scFv), nanobodies, Affibody molecules and DARPins, which have a molecular weight of less than 60 kDa and can therefore pass the glomerular membrane83. Currently, the mechanism for reabsorption of the probes in the kidneys is not fully understood, but “scavenger” receptors such as megalin, residing in the brush border of the kidneys have been shown to confer reabsorption to some extent84. Using a non-residualizing radionuclide will significantly decrease kidney uptake, but at a cost of lower tumor uptake if the probe is degraded inside the tumor.

The choice of type of protein or peptide for development of a probe is also enormously important for molecular imaging. Using the large repertoire of available antibodies would be a straightforward way to develop imaging agents. However, while larger proteins, such as antibodies, have significantly lower kidney uptake, the long circulation time results in poor imaging contrast and usually the analysis has to be performed several days post injection (pi). Furthermore, due to the enhanced permeability and retention effect (EPR), tumors accumulate macromolecules as a result of their highly permeable vasculature85. Although this effect is a valuable feature in a therapeutic setting, a non-target expressing tumor can have up to 20-30% of injected antibody residing around the tumor compared to a target-expressing tumor86. Smaller proteins or peptides are not equally affected by the EPR effect, resulting in a generally higher specificity compared with antibodies. Additionally, due to the faster blood clearance of smaller probes, the resulting contrast is typically much higher and analysis can be performed within a couple of hours82.

Figure 4. Molecular imaging using proteins. Exemplified here using a tumor targeting Affibody molecule with a recombinantly added unique cysteine residue in the C-terminus of the protein. In this example, maleimide chemistry is used to site-specifically covalently link a DOTA chelator to the protein, which allows subsequent radionuclide chelation before injection.

15

Ken G Andersson

3.1.1 Molecular imaging using small proteins or peptidesMolecular imaging using small proteins and peptides with different radionuclides has in general demonstrated higher contrast and specificity compared with antibodies. Short peptides (<2 kDa) derived from nature, such as the RGD-peptide and bombesin analogs, have shown promising results with high specificity and excellent contrast87,88. However, directed evolution of new peptides with sufficiently high affinity against novel targets is challenging. In contrast and as will be discussed in more detail in section 4.2, small affinity proteins are relatively easy to generate with similar high affinities and specificities as for monoclonal antibodies. In addition, using the same so-called protein scaffold for different targets might facilitate optimization, since accumulated knowledge regarding for example optimal injection route, in vivo half-life and dosing from previous probes can be used when investigating new imaging agents.

One of the most well investigated protein scaffolds for molecular imaging is the Affibody molecule (which has been used in all studies in this thesis and will be further described in chapter 4). Site-specific conjugation of macrocyclic chelators such as 1,4,7,10-tetraazacyclododecane-1,4,7,10-tetraacetic acid (DOTA) can be achieved by genetically incorporating a cysteine residue into the scaffold and subsequent thiol-specific maleimide coupling to that residue89. Good refolding capabilities and excellent pH tolerance of the Affibody molecule permit chelation of radiometals under harsh conditions that many other probes would not tolerate. The rapid in vivo pharmacokinetics of this scaffold are compatible with both long-lived radionuclide 111In (t½ 2.8 days) and 99mTc (t½ 6h) using SPECT imaging, as well as short-lived radionuclides 68Ga (t½ 68min) and 18F (t½ 110min) for PET90. Affibody-based tracers have been developed against a number of tumor-specific targets such as EGFR, HER2, HER3, Insulin-like growth factor 1 (IGF-1R) and carbonic anhydrase 9 (CAIX)91. Additionally, clinical molecular imaging studies have been performed on breast cancer patients with HER2 positive tumors using Affibody molecules. An Affibody molecule recognizing HER2 was site-specifically conjugated with DOTA and subsequently chelated with 111In or 68Ga. This tracer was the first non-immunoglobulin-based protein used for clinical imaging and could successfully detect primary tumors and localize metastatic lesions92. This Affibody molecule have shown very promising results in clinical imaging and is currently investigated using 68Ga labeling and PET/CT imaging for increased sensitivity compared to SPECT for quantification of HER2 expression93,94.

Further, an anti-EGFR Affibody molecule is currently being investigated as a fluorescence-guided brain cancer surgery imaging agent95. Due to the high target specificity and favorable tissue distribution properties, the tumor can be rapidly identified and characterized which may allow image-guided surgery as a potential option for improved brain tumor treatment. The agent is currently investigated in a phase 0 microdosing study by an academic/industry partnership to establish an efficient pipeline for development of Affibody molecules as imaging agents for guided surgery.

16

Chapter4

Affinity molecules

4.1 Antibodies Among the most predominant molecules in our body’s immune system are glycoproteins from the immunoglobulin superfamily called antibodies. Antibodies are a class of proteins mainly produced by plasma cells that have evolved to recognize and bind in principle any molecule capable of inducing an immune response (so-called antigens). The rather large protein of 150 kDa is a tetramer of two light chains and two heavy chains, forming a structure resembling the letter Y. The antigen-binding region of the antibody (so-called paratope) is located in the variable region of the fragment antigen-binding (Fab) arms. Binding is facilitated by three loops on the variable region of both the light and the heavy chain, called complementary determining regions (CDRs). The constant region of the antibody contains the fragment crystalizable (Fc) in the heavy chain which gives the protein many additional features besides binding such as half-life extension and effector functions via activation of different immunological responses (Fig 5). Two different light chains, lambda (λ) and kappa (κ), can be paired with five distinct heavy chains defining the isotype of the antibody. The five different isotypes in mammals IgA, IgD, IgE, IgG and IgM are used by different immunological responses, where IgG constitutes the majority of the antibody-based immunity against foreign molecules and pathogens. Further, IgG isotypes can be divided in four subclasses in order of decreasing abundance in the body: IgG1, IgG2, IgG3 and IgG4. Even though all subclasses share >90% sequence similarity, each subclass has a unique profile with respect to immune complex formation, complement activation, half-life, placental transport and antigen binding. This is a significant factor that has to be taken into account when engineering an antibody for in vivo therapeutics or diagnostics96–98.

Antibody engineering for generation of new binders with desired specificities is today an important tool in medicine and other areas of life science. Earlier, new specific antibodies could only be generated by injecting an antigen into an animal and collecting the antiserum via bleeding, which is called immunization. The adaptive immune system of the animal responds to the antigen and generates polyclonal antibodies (pAbs) against the antigen. In comparison to monoclonal antibodies (mAbs) coming from a single cell lineage, the polyclonal antibody are a mix of many different antibodies that will typically recognize multiple epitopes on the antigen99. While pAbs are inexpensive, rapid to generate and highly versatile, heterogeneity and batch-to-batch variation make them unsuitable for some biotechnological and therapeutical applications100. In contrast, the homogeneity, renewable and reproducible production and single-epitope recognition of mAbs are essential features in many immunological assays and in particular for development of immunotherapeutics.

17

Ken G Andersson

However, long and expensive development and sensitivity to mutations on the antigens are some of the drawbacks of monoclonal antibodies98.

Generation of monoclonal antibodies was first invented by Céasar Milstein and Georges J. F. Köhler in 1975 and they were subsequently rewarded with a Nobel Prize in 1984. They found that isolated antibody-producing plasma cells from immunized mice could be fused with myeloma cells. These so-called hybridoma cells, allowed continuous antibody production101. While early production of mAbs was invaluable for research, production using hybridomas was genetically unstable with relatively low yields, and severe side-effects from mouse-derived antibodies were reported in the clinics 102,103. One of the solutions, developed to circumvent these challenges, was to mimic the immune system in a test tube using recombinant libraries of antibody derivatives (e.g. scFvs) and selection of new binders in vitro (described in more detail in chapter 5), followed by reformatting into human full-length antibodies after selection 104–108. Using fully human antibodies as drugs instead of murine, dramatically reduced the immunogenicity when administered into patients. Furthermore, the pharmacokinetic profile (e.g. half-life in blood, placental transport) and effector functions can be tailored by careful selection of IgG isotype as well as Fc engineering109.

Figure 5. An antibody and its derivatives. An IgG antibody weighs 150 kDa and consists of two heavy and two light chains. The heavy chain encloses the Fc region that is responsible for half-life extension and effector functions. The antigen binding region is located in the fragment variable (Fv) of the Fab arms and by proteo-lytic cleavage or recombinant production, the Fab can be used as an antibody derivative. The single chain fragment variable (scFv) is a smaller derivative from the antibody and is commonly used for antibody engineer-ing.

18

4.2 Alternative scaffolds Although antibodies and their derivatives are invaluable reagents in research, diagnostics and therapeutics, high affinity and specificity can also be found in non-immunoglobulin-based proteins. Many other proteins in nature can structurally accommodate binding with similar affinities as reported for antibodies. Naturally occurring protein interactions are not engineered by a dedicated immune system, instead, spontaneous mutations and natural selection has made it possible to generate extremely high affinity binders, such as streptavidin (binds vitamin B7 (biotin) with femtomolar affinity110).

There are many strategies to develop new binding proteins. One strategy is based on mutating a protein of interest that already have binding capacity to the intended target in order to improve the affinity and/or specificity. An example is the approach reported by Cochran et al, where they mutated EGF that has a natural affinity for EGFR, and selected for improved variants111. Another more recent approach is based on computer-based in silico modelling, for example as reported by David Baker and co-workers, where they grafted binding motifs onto different designed scaffolds using computational methods112. However, the most common approach is to select a suitable protein scaffold, mutate it and generate novel binding to a new target. Although this approach might appear straightforward, it is still challenging and requires relatively sophisticated methods. The term protein scaffold is referring to the parental protein that is mutated, while retaining hopefully the original structure and favorable properties. Usually, the mutated positions are defined in the scaffold and can either be in loops or distributed over a suitable molecular interface of the protein113. However, alternative strategies using a more random mutation approach with techniques such as error-prone PCR are also sometimes used. A small protein scaffold has some advantageous features such as production by chemical peptide synthesis, potentially reduced unspecific binding, inexpensive production in prokaryotic hosts, flexible engineering and multimerization, rapid biodistribution and option for alternative administration routes 114. Reducing the size also differentiate the alternative scaffolds from monoclonal antibodies, which might allow for other applications where the use of antibodies is currently limited. Further, continuously using the same scaffold will generate increasing information about potential issues, such as detrimental sequences for solubility, protease susceptibility and immunogenic properties. However, engineering the same scaffold could speculatively generate biases towards certain types of epitopes or antigens that structurally fit with the selected scaffold.

Just as choosing between Billy Ocean and Metallica, selecting the ideal scaffold is a matter of preference. Since every scaffold has different properties, scaffold selection should be application-driven. Popular traits for reported scaffolds are usually high thermostability and easily produced monodispersed molecules115. Some applications require the scaffold to permit exotic features that can only be possible using certain proteins, such as single domain bispecific binding116, super-high structural rigidity117 or binding small molecules118. Some examples of different scaffolds and their unique features is discussed in the following paragraph.

19

Ken G Andersson

Presently, over 50 different scaffolds have been reported119. Though most are currently only used in an academic setting, some scaffolds are reporting interesting features that in comparison to antibodies could be advantageous in a therapeutic or diagnostic setting and could therefore be of interest for the clinics. One of the smallest reported so-called bispecific scaffolds is the ADAPT, derived from an albumin-binding domain (~5 kDa) of streptococcal protein G. Randomization on positions directly opposite of the surface that is interacting with human serum albumin (HSA) and subsequent selection has yielded a single-domain bispecific binders towards a number of different targets, such as HER3 and HSA with a KD of 10 nM and 0.4 nM, respectively120,121. Another small scaffold that has shown very promising results in both pre-clinical therapeutic settings and diagnostic application are the 4 kDa cystine-knot proteins, also known as knottins122. The highly rigid scaffold contains three intertwined disulfide bridges conferring proteolytic resistance and high thermal stability. The high structural stability in this class of proteins allows for mutations in the loop region between the cysteine residues which has generated many different binders to various targets, such as a subnanomolar binder to human matriptase-1123.

One of the most structurally investigated alternative scaffold is the designed Ankyrin repeat proteins (DARPins). Over 20 published X-ray crystallography structures of DARPins with and without ligand are currently deposited in the protein data bank. DARPins consist of four or five repeated domains (14-22 kDa in total), where the first and the last repeat serve as a hydrophilic surface124. Randomization within the loop region of the non-flanking repeats, followed by selections have yielded picomolar binders against several targets after affinity maturations125. The most widely-used domain for generation of affinity proteins is probably the fibronectin type III (FN3) domain, spawning many different scaffolds such as Adnectins126, Centryins127, Monobodies128 and TN3129. The FN3 domain is an evolutionary conserved protein domain with a cysteine-free beta-sandwich structure (10 kDa) that has successfully been used for selection of binders against many targets130. The first reported use was by Koide and co-workers in 1998 and it has recently also been successfully delivered into cells as intracellular inhibitors128,131. Originally, mutations on the Monobody were directed to the loop region, mimicking an antibody-binding region. Later, randomization in side-and-loop compared with loop-region showed that the two classes of libraries are capable of recognizing different epitopes with distinct topography132 (Fig 6). This further suggests that the type of scaffold or library design influences which epitopes or even antigens that are compatible.

20

Many of these scaffolds are currently entering clinical trials against cancer, inflammation disorders, and other diseases133. The low molecular weight, generally high stability and high specificity permits these scaffold to take advantage of features such as high tissue penetrations and fast clearance. Currently, the only FDA-approved alternative scaffold is the kunitz domain DX-77 against hereditary angioedema. However, many other scaffold are in early phases of clinical trial such as DARPins against macular degeneration in phase III, adnectins regulating angiogenesis in cancer in phase II. The Affibody molecule has gone furthest in clinical trial with its medical imaging agent against HER2 expressing breast cancer. While, cystine-knots for pain-relieving by targeting Nav1.7 is in pre-clinical trial and the ADAPT scaffold has not been reported yet for clinical trials94,119.

Figure 6. Examples of alternative scaffolds and different randomization sites for directed evolution. A) The small scaffold ADAPT (~5 kDa) that is derived from an albumin-binding domain of streptococcal protein G. Randomization on positions opposite of the HSA-binding surface allows for bispecific binding. B) The 46 amino acids large agouti-related cystine knot is randomized in the loop region between Cys 22 and Cys 30 to confer binding against targets. Cysteine disulfide bridges in the scaffold are depicted in yellow C) The DARPin scaffold, depicted here as a five repeat protein, containing 169 amino acids with randomization in three of the repeats. D-E) One of the first publications on alternating randomization position to recognize different epitopes were on the Monobody129. Randomization in the loop region (D) resulted in binders with similar epitopes to antibodies, while randomization on the beta sheets (E) resulted in binders with a concave binding surface with the possibility to bind more flat epitopes. F) The Affibody molecule (~6 kDa) is conventionally randomized in 13 positions and with a flat binding surface has been shown to be able to generate specific binders down to picomolar affinities.

21

Ken G Andersson

4.2.1 Affibody moleculeThe Affibody molecule is a versatile non-immunoglobulin folded scaffold protein, originally isolated from domain B in staphylococcal protein A. The three-helix bundle of 58 amino acids (~6 kDa) is a robust scaffold that is rapidly folding, highly soluble, easily expressed in prokaryotes and can refold after thermal denaturation134–136. Using combinatorial protein engineering, libraries have been constructed with typically 13 randomized positions on helix 1 and 2, comprising a relatively flat binding surface of about 800 Å2 137. Typically, 18 different amino acids are included in the randomization where cysteine and proline are excluded, avoiding unwanted helix breakage and dimerization. Successful selections against various targets have been achieved using a variety of different display methods such as phage, ribosomal, cell and mRNA display (described in more details in chapter 6)138. The small scaffold can be used as a modular binding unit by readily fusing it to other domains or proteins, resulting in for example avidity effects or bispecific binding. Examples of fusions include: other affibody molecules139, full-length monoclonal antibodies as well as to Fc140 or an albumin binding domain for in vivo half-life extension 141.

Recombinant, site-specific labeling is straightforward by genetic insertion of a unique cysteine residue and subsequent malemide chemistry142. Additionally, peptide synthesis of the Affibody molecule enables additional modifications, such as multiple fluorescent tagging143, inter-molecular crosslinking144 and incorporation of unnatural amino acids145. The Affibody molecule is this year celebrating its 20th year since first publication146. During these two decades, over 400 studies have been published and the scaffold has entered clinical development for both therapy and molecular imaging. Affibody molecules against over 40 different targets, with affinities ranging between low micromolar (from the earliest studies) to low picomolar have been reported in literature138. While selection from a naïve (unbiased) library usually results in affinities in the nanomolar range, affinity maturations using second-generation libraries have yielded binders with low picomolar affinities. For instance, first naïve selections against the cancer-associated targets HER2147, HER3148 and EGFR149 yielded binders in the nanomolar range, where a subsequent affinity maturation resulted in picomolar binders. Direct comparison between first and second-generation Affibody molecules against HER2 indicated a 4-fold enhanced tumor uptake in vivo for the affinity-matured variant150. Currently, the company Affibody AB (commercializing the Affibody technology) is developing Affibody molecules in four proprietary programs for therapeutic and in vivo diagnostic purposes151.

22

Chapter5

Engineering approachesEngineering proteins to improve or create desired traits is a large field in biotechnology that includes anything from small truncations to multiple mutations. When substituting amino acids in protein engineering to alter binding interactions, there are usually three mutation strategies that are employed: rational design, random mutagenesis and site-directed saturation mutagenesis. Whether it is to enhance the affinity or specificity of the protein of interest or to engineer a completely novel interaction, there is one thing they all have in common, it requires time. The following chapter will discuss different methods for protein interaction engineering and selection methodologies that can be used to generate new binders.

5.1 Directed evolutionDirected evolution is based on the principle of construction of extremely large libraries (typically 107 – 1012 variants) of different mutants of the protein, followed by selection or high-throughput screening for isolation of new binders. The large libraries that are typically generated in directed-evolution efforts require relatively sophisticated screening or selection methods (examples are described in chapter 6). Although it is possible to screen a handful or even hundreds of variants using ELISA, already after randomizing two or three positions this becomes extremely time-consuming. Many different selection methods have thus been developed over the years to enable fast and robust screening of the desired trait. Usually divided into two groups, cell-free and cell dependent, both methods rely on a physical link connecting the proteins with the genetic information that encodes them. Which could be regarded as a protein ID-tag, the linked and later co-selected DNA permits straightforward amplification by PCR or replication in the host as well as identification of the amino acid sequence by DNA sequencing. Amplification of the isolated proteins can hence be achieved in a subsequent step by transcription and translation from the encoding DNA.

Incubating the protein library with the antigen allows for selection on the specific interaction and subsequent retrieval of binding proteins. The antigen is typically immobilized to a solid support, allowing for capturing of binders and washing away the large background of non-binders. Ideally, washing also eliminates most of the non-specific and low affinity binders and is typically varied for controlling the stringency. Elution or isolation, depending on method, recovers the proteins of interest and the co-selected DNA for amplification, and the library is thereafter usually subjected to another round of selection. Decreasing the antigen concentration throughout each selection round typically allows for enrichment of high-affinity binders. High affinity is generally considered to increase specificity of the interaction by permitting more residues to specifically interact, further minimizing the

23

Ken G Andersson

off-target interaction169. Northrup and Erickson summarized several experiments in 1992 and proposed that association kinetics for protein-protein interaction in mammalian salt conditions does not exceed ~107 M-1s-1 170, suggesting that off-rate is a key parameter for further increasing affinity. Off-rate selection was first reported by Greg Winter’s group in 1992 using phage display of antibodies171. They suggested that preloading a library with biotinylated antigens, followed by relatively long incubations in an excess of unlabeled antigen should favor variants with slow dissociation. Since then, off-rate selections have been proven to be an extremely valuable tool for selecting binders with high affinity172,173.

A variety of selection systems are available today for screening and selecting binders with high affinity. For instance, cell-free display fuses the protein directly to the genetic information and exploits the translation machinery of bacterial or eukaryotic cell lysates to express the proteins in the library in complex with each corresponding gene. Generally, these systems are considered to cover a larger library size with up to 1011-1013 different members, which is generally higher compared to the cell-dependent systems that requires DNA to be transformed into the hosts174. A potential concern with these methods is that the negatively charged genetic code could interfere with the function of the displayed protein. Additionally, expression of complex proteins might be challenging using these systems175. While cell-free systems such as ribosomal display176, mRNA display174, cis display177 and CSR display178 are all relatively established selection systems, this thesis will not focus more on these methods.

5.2 Rational designRationally designing a protein interaction is one of the most challenging tasks in biotechnology. As little as a single amino acid substitution can have great impact on affinity, expression, activity or stability152,153. The use of X-ray crystallography and nuclear magnetic resonance (NMR) on proteins has greatly accelerated the field by allowing high-resolution structural models of proteins and protein/protein interactions. Solving the crystal structure for the protein interacting with its target gives in-depth information about which amino acids in both the ligand and the binder that are essential for the interaction. However, obtaining high-resolution structures of protein and complexes is typically very laborious and some proteins, such as trans-membrane proteins like G protein-coupled receptors, are even more challenging to crystalize154. Further, a static image of the protein interacting with its ligand does not necessarily give information about the conformational changes that might be required to support binding.

In this evolving field, a larger change is occurring and currently one of the biggest contributors to rational design is computational prediction of protein folding and interaction155. Although in perhaps its infancy, some groups have managed to theoretically predict structures followed by double blind testing aligning their predicted structure against known structures, confirming that the in silico modulated structure overlaps with the empirical counterpart156. Recently, powerful computational design techniques have emerged which augments the rational design field further. Recently, David Baker’s

24

group managed to reengineer a protein-protein interaction by computational modelling, generating a pH-dependent FC binder. Determining hot-spots on a minimized protein A molecule in combination with phage display, they managed to identify a binder with 500-fold higher affinity at pH 8.2 compared to pH 5.5 157. Perhaps even more impressively, Flieshmann et al managed to predict and generate a scFv against the conserved stem region of influenza hemagglutinin158. Although both projects required combining the computational predictions with directed evolution to generate high affinity binders, computational design has shown potential to be a valuable alternative or complement for the identification of novel binders in the future.

5.3 Random mutagenesisRandom mutagenesis is one approach for generating the libraries used in directed evolution. As the name of the method suggests, random mutagenesis is a strategy for diversifying a protein in a random manner and does not require any detailed knowledge of the protein. Notably, this approach resembles somewhat how a protein is naturally evolving and could speculatively generate unexpected and interesting features159,160. However, a major drawback of the method is the incorporation of stop codons and biases for certain amino acids that are more prevalent in the genetic code. Mutations will also occur in positions that are important for the structural integrity of the protein, hence typically decreasing the functional portion of the library much more compared with site-directed saturation mutagenesis, which is described in the following section 5.4.

One method to induce random mutagenesis is error-prone PCR (ep-PCR). This method, first described by Leung et al, exploits the low-fidelity of DNA polymerase together with manganese and magnesium ions161. Replication of the gene using low-fidelity DNA polymerase or mutagenic dNTP analogues increases the mutation rate to around 10-4 ~ 10-3

per replicated base162,163. The average number of mutations per clone can be estimated by calculating backwards from the number of reaction cycles, estimated error rate and size of the gene164.

5.4 Site-directional saturation mutagenesis The site-directional saturation mutagenesis approach is somewhat semi-rational, combining the rational design and random mutagenesis approach by selecting which positions are suitable for randomization. In comparison to random mutagenesis, this approach allows for guiding the protein interaction site and potentially generates more functional mutants. Somewhat similar to how the immune system diversifies the antibody’s hyper variable region (HVR), this method is an excellent approach for generating naïve libraries in scaffold proteins. However, similar to random mutagenesis, this method requires efficient selection systems that permit functional screening or selection of the generated libraries (described in detail in chapter 6). Indeed, fully randomizing for example seven positions in a protein generates a diversity of roughly one billion and can therefore not be screened using low throughput conventional methods such as ELISA or SPR.

25

Ken G Andersson

Generating diversity and creating a library of mutants can be achieved with multiple methods. Traditionally, introduction of diversity has been achieved by synthesis of DNA with position-specific degenerate NNN codons. The term N represents any of the four DNA bases (G/A/T/C) and NNN thus signifies randomization using all codons. Due to degeneracy in the genetic code, 64 codons are encoding the 20 naturally occurring amino acids. Therefore, randomization using NNN gives a high genomic diversity but the degeneracy in codon results in biases as well as introduction of stop codons. Other methods to randomize codons such as NNB, NNK, NNS (where B =C/G/T; K=G/T and S=C/G) have been developed, minimizing the degeneracy and biases by allowing 48, 32 and 32 codons respectively165. Although synthesis with degeneracy is the most inexpensive way to generate a combinatorial approach, the biases and random stop codons reduce the quality of the library. A more sophisticated method is the use of trinucleotide phosphoramidites in solid-phase DNA synthesis. The method synthesizes DNA as trinucleotide codons covering all 20 amino acid without the addition of a stop codon166. Removing or altering the frequency of amino acids in each position allows a tailor-made library to be constructed. This is especially attractive for scaffolds with a binding surface located in secondary structures, since some amino acids, such as proline and glycines are known to disrupt secondary structure. Additionally, in the CDR loops of antibodies, tyrosine residues have been shown to be important for binding in certain cases167,168. Therefore, a CDR loop randomization permitting a surplus of tyrosine codons in this region, could generate a more functional library, which some studies suggest168.

26

Chapter6

Selection methods

6.1 Phage displayThe most commonly used selection method is phage display. First described by George P. Smith in 1985, he demonstrated that peptides could be displayed as fused to the gene III product on filamentous phage179. By inserting recombinant DNA fragments into bacteriophage M13, the peptide could be displayed on the phage surface, while the phage retained its infectivity to E. coli. The 1-2 microns long M13 phage is 6 nm in diameter and mainly composed of the coat protein pVIII, encapsulating the DNA that permits replication after infection180. Fusing the protein of interest to either the pIII or the pVIII gene facilitates surface display on the particle, while containing the gene encoding the protein inside the phage, resulting in a phenotype-genotype linkage. Selection from phage-displayed libraries is typically done by i) adding the library-displaying phages to antigen immobilized on a surface or capturing the phages on target-coated magnetic beads, ii) washing to remove unspecific binding and non-binding phages, and iii) elution of the binding phages that subsequently can be re-infected into E. coli and amplified. A common strategy to retrieve high affinity binders is repeating cycles with increased washing stringency and lowering the concentration of antigen throughout each round (Fig 7). The use of this method for screening large protein- or peptide-based libraries has truly changed the field of antibody and protein engineering, enabling generation of binders to almost any conceivable target such as DNA181, proteins182, peptides183 and carbohydrates184. Phage display technology is today routinely used for antibody drug discovery. Studies in the early 1990s indicated that display of antibody-based scaffolds on phages was possible and Barbas et al showed that specific Fab fragments that bound tetanus toxin could be isolated using phage display105,185. Currently, this technique is an important method for antibody drug discovery and phage-displayed antibody libraries are constructed and screened on a daily basis all over the world.

The phage display method has been improved and modified over the years since it was first reported, and today there exist a number of different variants of the technology. An example is a variant that was first suggested by Smith and Petrenko, involving a more iterative strategy that is analogous to natural selection. Stepwise randomizations between each round are introduced, which would slowly alter the selected binders and might be more efficient for generating high-affinity variants186. Although stepwise or continuous randomization between each round in phage display (and other methods like ribosome display) has been successfully shown to generate novel features and excellent binders, the method itself is relatively time-consuming187,188.

27

Ken G Andersson

6.2 Cell displayUtilizing cells for displaying the protein of interest is an alternative strategy analogous to phage display. Similar to the phage particle, the cells link the phenotype (protein) displayed on the surface to the genotype (DNA) encapsulated inside the cell. Due to the large particle size and multivalent display, high-throughput screening and isolation of binders from the libraries is typically performed using fluorescence-activated cell sorting (FACS). Sorting using flow cytometry enables quantitative analysis of single cells and separation of the heterogeneous library into two or more containers. The flow cytometer measures the light scattering as well as the fluorescence signals of the cells and subsequently encapsulates the cells in droplets. Analysis of more than 70 000 cells per second allows for high-throughput sorting. Moreover, conversion of the detected light to digital signals permits gating and selection on which droplets should be retrieved. The gated droplet is subsequently charged and electrostatic deflection diverts the selected droplet into a predefiened container. Multiparameter laser measurement and charge variance allows for multiplexed sorting into different containers (Fig 8).

Figure 7. Phage display using M13 filamentous bacteriophage. A) The major coat protein PVIII encapsulates the phage DNA that can be propagated in new cells after selection. Five copies of the minor coat protein PVI and PIII are presented on the end of the phage and are the first proteins to interact with E. coli during host infection. Most commonly, the protein of interest is fused to the PIII protein of the phage particle but both PVIII and PVI has been studied and shown functional for phage display. B) The affinity selection technique using phages is called biopanning where phages are typically captured on a surface-immobilized antigen, washed multiple times and eluted with for example pH or proteases. The output can be amplified for another round of selection or analyzed by reinfection and sequencing of the DNA.

28

Fluorophore-labeled antigen can bind to the cells in solution, allowing for fluorescence-based quantification of the affinity, since the fractional occupancy on the cell is directly correlated to the equilibrium dissociation constant. Sorting a cell-displayed combinatorial library on a pre-defined fluorescent threshold, results in quantitative isolation of the strongest binder and permits monitoring of binding throughout the selection process. This also allows for optimization of sorting parameters and conditions since the enrichment can be accurately quantified during and between each selection round. Furthermore, fusing the randomized scaffold to a reporter tag and subsequent co-incubation with a fluorescent agent towards the tag, allows for normalization of the surface expression level189. Another advantage of using multiparameter sorting is for engineering species-specific cross-reactivity in binders. By labeling animal and human antigen orthologues with distinct fluorophores and using them as antigens in selection, subsequent sorting and ranking of cross-reactivity can be achieved. For example, using cell display with two-color FACS could select antibodies that were cross-specific to two different botulinum neurotoxin subtypes190.

Although cell display using FACS offers affinity discrimination and sorting of up to 70 000 cells per second, the selection process is typically more time-consuming than phage display and screening high complexity libraries beyond 108 members is generally not feasible191. Therefore, pre-selection using antigen-coated magnetic beads (so-called MACS) can be used to enrich and decrease the library size before subjecting it to FACS-based selections192. Many different cell types and anchoring strategies have been investigated for directed evolution using cell-display. While yeast193 is the most commonly used host, others, such as gram positive148, gram negative194 and mammalian hosts195 each have their respective advantages and disadvantages.

Figure 8. Cell display using the eukaryotic S. cerevisiae. A) Yeast can express a foreign protein on the surface using the a-agglutinin system where Aga1p constitutes surface anchoring and Aga2p is fused to the recombinant protein and linked to Aga1p by two thiol bonds. B) Similar to biopanning, affinity selections can be achieved by labeling an antigen with a fluorescent agent. Cells can be quantitatively separated based on fluorescent signal using FACS that allows selection of binders from non-binders. The isolated binders can be analyzed using sequenc-ing or amplified for another round of selection. C) The FACS instrument can individually and quantitatively measure the fluorescence of each cell and sort based on desired fluorescence or light scattering which permits sorting cells of interest. After measuring fluorescence, cells are encapsulated into droplets using mechanical force and the droplets can be electrostatically deflected into different capsules depending on the fluorescence.

29

Ken G Andersson

6.3 Yeast displayUsing yeast as host for cell display was first described by Boder and Wittrup in 1997193. Fusing the protein of interest C-terminally to α-agglutinin under control of a galactose promoter allowed for anchoring on the surface of S. cerevisiae. The anchoring protein α -agglutinin consists of two subunits where the larger Aga1p constitutes surface anchoring via β-glucan covalent linkage, and the Aga2p subunit, containing the fusion protein, is linked to the Aga1p via two thiol-bonds196,197. The galactose promoter is repressed when grown in glucose medium and induction of expression is achieved by transferring to medium containing galactose. This avoids counter-selection against sequences that negatively influence yeast growth193. Additionally, the yeast microbe possesses a relatively sophisticated DNA homologous recombination machinery, allowing for transformation of linear plasmids together with overlapping fragments that the yeast recombines into a circular plasmid198. However, yeast has lower transformation efficiency compared to E. coli, leading to smaller libraries than in phage display199,200. Regardless of this, Wittrups group managed in 2003 to generate a non-immune library of 109 and to express the repertoire on the surface of yeast192. The authors succeeded to format the human repertoire into a scFv scaffold using overlap extension PCR, and selection against multiple targets was successful, generating binders with single- or two-digit nanomolar affinity. Recent improvements in transformation efficiency reported impressively around 108 transformants per µg of vector and using this method, the authors generated a large naïve human spleen antibody library with approximately 1010 variants201. Although this method has an increased transformation efficiency, other groups have to our knowledge not been able to achieve equally high efficiencies202,203.

About 70% of all approved therapeutic proteins are glycoproteins204. In comparison to prokaryotic hosts, the eukaryotic S. cerevisiae contains an endoplasmic reticulum (ER) in which different oligosaccharides are transferred to asparagine residues, resulting in N-linked glycosylation, or to the hydroxyl group of serine and threonine for O-linked glycosylation. While N-glycosylation has been shown to be important for protein function in protein therapeutics, there is currently insufficient information regarding the function and biological role of O-glycosylation 96,205. Additionally, it has been shown that glycosylation can influence antibody epitope recognition206. Yeast glycosylation has high-mannose content and consequently dissimilar to human. Though efforts have been made to humanize the glycosylation pathway, currently only ‘hybrid’ proteins produced in S. cerevisiae containing both yeast and human glycosylation have been reported207.

Since first described 20 years ago, the yeast display field has exploded in terms of number of publications208. Although it exhibits only a fraction of the publications compared to phage display, a comparative study between the two selection techniques found that yeast display could sample the library repertoire more completely and, according to the authors, was “less labor intensive”. The same study also showed that only 5/12 of the selected scFv derived from phage display, compared to yeast 11/12, could be functionally expressed in E. coli 209. The increasing popularity of this method has resulted in surface expression of

30

many different proteins and peptides. Whereas combinatorial protein engineering on yeast surface has been achieved using almost anything between enzymes210 to shark antibodies116 and full length antibodies211, other fields have also been explored such as oral vaccine delivery by expression of pathogenic proteins212 and as an antibiotic particle by expression of antimicrobial peptides213.

6.4 E. coli displayThe most widely studied prokaryotic model organism and perhaps also the most important species in the field of biotechnology, is the Gram-negative bacterium Escherichia coli. After phage display, E. coli is one of the earliest developed and most well-investigated hosts for protein and peptide display. Features such as short doubling time, high transformation efficiency, high expression level, no unwanted post-translational modifications, choice of titratable promoters and small size that permits high cell densities have all been desirable for display purposes. However, anchoring a recombinant protein, termed passenger protein, on the surface of the dual-membraned bacterium has been a challenge.

The first study on directed evolution using E. coli was published in the early 1990s, where Stanley Brown managed to engineer an iron oxide adhesion motif into a phage λ receptor (LamB)214. The possibility to use the phage λ receptor as a surface anchoring protein was suggested already in 1988, allowing up to 60 amino acids to be anchored215. However, this anchoring method requires the library to be both C and N-terminally translationally fused to LamB, making it difficult for more complex proteins to fold correctly.

Many alternative anchoring methods have been developed over the years. Francisco et al presented one important milestone by the proof-of-principle demonstration of functionally displaying scFv and subsequent FACS sorting of a spiked library216. Later, another important milestone was the introduction of a technique called APEx, fusing a scFv to an anchoring protein on the inner membrane (IM), such as lipoproteins or the phage minor coat protein III of M13. However, a downside with IM anchoring is that the outer membrane (OM) will impede antigen access, which is problematic when labeling for FACS. Permeabilizing the OM before labeling is therefore necessary. This abolishes self-replication of the cell and the genotype of the selected binders needs therefore to be recovered using PCR. Expression, followed by capture or entrapment inside the periplasm of full-length antibodies has also been done on E. coli. Georgiou and co-workers managed to generate full length antibodies from immunized animals against protective antigen of Bacillus antracis by utilizing APEx and tethering a protein A derivate on the inner membrane, subsequently capturing the aglycosylated full-length IgGs217. Genentech approached this differently, and instead of capturing the antibody and removing the OM, deletion of the major outer membrane prolipoprotein (Lpp) and mild permeabilizing conditions allowed the outer membrane to act as a filter, permitting antigens up to 100 kDa to enter and be retained while the antibody resided inside the periplasm. Randomization inside the antibody framework and selection for high expression revealed that three mutations increased the antibody expression yields up to 4- or 6-fold in HEK293T and E. coli, respectively218.

31

Ken G Andersson

Although a larger library can be obtained with E. coli as host, using FACS for selection from libraries >108 is in practice not feasible. Adequately covering (usually >10-fold) such a large library would require the sorting (at 70 000 events per second) to be maintained for at least 8 hours, which would result in dissociation of most binders. Therefore, pre-selection using MACS is required, where biotinylated target is typically captured onto streptavidin-coated magnetic beads. Adding the coated beads to the library can easily cover libraries >1010 and is only limited by the volumes required to cover the amount of bacterial cells. Magnetic selection is generally not as quantitative compared to FACS. However, using four rounds of MACS sorting on a nanobody library expressed on the surface of E. coli resulted in binders with nanomolar affinities against TirM (EHEC) indicating that this method could also be used throughout the whole selection, if necessary219.

To anchor a passenger protein on the outer membrane of E. coli requires the protein to traverse two membranes. First, the transcription and translation occurs in the reducing cytoplasmic environment occurs. Two major translocons exist in prokaryotes to secrete proteins across the IM, the general secretion route (Sec-pathway) and the twin-arginine translocation pathway (Tat-pathway). The Sec-pathway generally transports unfolded proteins either co-translational or post-translational, while the Tat system has been speculated to be used only in cases where cytoplasmic folding excludes the Sec secretory system220,221. It should be noted that both periplasmic anchoring and outer membrane anchoring is usually denominated as surface display. Traversing the outer membrane is done by various highly specialized secretion systems classified by roman numerals from type I to type V222. Although there are multitudes of alternatives for outer membrane translocation in E. coli this thesis will focus on the type V secretion systems.

6.4.1 Type V secretion systems The type V secretion system (T5SS) is different compared to most other prokaryotic secretion systems as it involves one, two or three polypeptide chains for a self-sufficient export mechanism using a β-barrel domain. This rather large group of secretion systems can be sub-categorized into five distinct groups ranging from type Va to Ve, where Vb and Vc are more complex, consisting of two or more polypeptide chains, while Va, Vd and Ve only consist of a single polypeptide chain223. Common to all T5SS, is utilizing the sec-apparatus for unfolded IM translocation. A translocation domain, folding into a β-barrel is chaperoned and integrated into the OM, acting as an amphipathic pore allowing the unfolded passenger domain to be secreted or surface anchored depending on the translocation domain224. Typically, the passenger sequence remains chaperoned and unfolded until OM translocation223.

32

The first elucidated translocation pathway for T5SS was for type Va, also known as autotransporters. In 1987, Pohlner et al investigated the immunoglobulin A1 (IgA1) protease secretion in Neisseria gonorrhoeae and found that the polypeptide consists of three functional domains: i) an N-terminal signaling peptide that is cleaved upon translocation over the IM; ii) a subsequent N-terminally free passenger sequence for outer membrane function; and iii) a C-terminal translocation domain, catalyzing transfer over the outer membrane225 (Fig 9). Autotransporters are naturally found in pathogenic gram negative bacteria and usually encodes for large β-solenoid structures over 100 kDa226. Interestingly, the group type Ve, termed inverse autotransporters or intimins, is almost identical to type Va, though the passenger sequence is C-terminally located while the N-terminal part forms the translocation pore227. The passenger domain is typically immunoglobulin- or lectin-like, which is unique to type V secretion systems.

Using type V secretion systems for E. coli display and directed evolution of protein binding is a relatively new field and not so well investigated. However, one of the larger contributions to this field is the expression and directed evolution of anticalins on the surface of E. coli. Here, the authors evaluated expression levels of an anti-CTLA-4 anticalin fused to five different bacterial autotransporters under tet-promoter control: IgA1 protease from Neisseria gonorrhoeae, AIDA-I adhesion from enteropathogenic E. coli O127:H27, EspP from enteohemorrhagic E. coli O157:H7, YpjA autotransporter from K-12 strain and esterase EstA from Pseudomonas aeruginosa. The results indicated that functional expression of anticalins was highest using EspP and combinatorial protein engineering together with off-rate selection using FACS yielded affinity-maturated binders with 8-fold slower dissociation rate constant228. Another group has reported on expression of nanobodies on the surface of E. coli using both the autotransporter EheA and the reverse autotransporter intimin. They showed that nanobodies were more stably expressed as fusion proteins to the intimin translocation domain rather than EheA. However, the natural passenger protein of intimin has an immunoglobulin-like fold, which is an indication that some secretion systems could be potentially better for certain types of scaffolds. Additionally, they generated a relatively small immune library of 3 x 106 nanobodies that were screened against translocated intimin receptor from EHEC (TirMEHEC) using MACS. Four rounds of MACS identified a nanobody with single digit nanomolar affinity estimated using both surface plasmon resonance and apparent affinity determination analysis on the surface of E. coli219. In the early 2000, Kolmar and coworkers used an intimin-based method to both map the epitopes of monospecific antibodies as well as to express enzymes, cytokines, knottins and calcium binding protein on the surface of E. coli229–232. By expressing many different passenger proteins they revealed that passenger proteins fused to intimin require folding into a ‘compatible fold’ for tranlocation through the bacterial outer membrane230.

33

Ken G Andersson

Relatively few articles have been published using type V secretion systems for directed evolution and the full potential and limitations are yet not completely investigated. Although not established as a method for directed evolution, many research groups have successfully displayed different proteins, ending up using the microbe for a wide range of biotechnological and biomedical applications, such as whole-cell biocatalysis, live bacterial vaccine and biosorbents233. Surface-displayed molecules such as streptavidin234, scFv235, melanin236, epitopes237 and many more, suggest that this method is a robust and efficient for displaying recombinant proteins on the surface of E. coli.

Figure 9. Expression on E. coli surface using type V secretion systems. Surface expression of recombinant proteins on the Gram-negative E. coli is relatively difficult since the protein of interest has to traverse two membranes. Many virulent strains of E. coli express type V secretions systems that permits surface expression. Surface expression is achieved by hijacking the Sec-apparatus for unfolded secretion into the periplasm followed by insertion of the translocation domain into the outer membrane that allows the passenger sequence to be translocated and subsequently folded onto the surface of E. coli.

Present investigation

36

Chapter7

Short Summary of the paperS in the theSiS

The overall aim of the work presented in this thesis was to develop and establish novel methods for combinatorial protein engineering of affinity proteins and development of tracers for medical imaging. The development and evaluation of an E. coli outer-membrane display system is described, which was employed for generation of Affibody molecules to be investigated in biotechnology, therapy and medical imaging applications. Further, this thesis aims at establishing rational design of Affibody-based tracers for medical imaging to increase contrast of tumors expressing EGFR and HER3.

Paper I

The aim of this study was to engineer a dual expression vector that would enable expression of Affibody molecules on the surface of E. coli as well as protease-mediated secretion using the endogenous OmpT.

Paper II

The aim of this study was to further improve the expression vector for E. coli display and generate a large combinatorial Affibody molecule library. To investigate the performance of the new method, selections were performed against a panel of different biomarkers: HER2, HER3, CD69, DGCR2 and IL3RA

Paper III

The aim of this study was to investigate if Sortase A could directly conjugate and release Affibody molecules from the surface of E. coli. In this study, conjugation and release directly from the surface using GFP, fluorescein and biotin substrates were evaluated. The investigated approach has the potential to increase the throughput of downstream production and characterization of candidates after directed evolution.

Paper IV

The aim of this study was to design and investigate an anti-HER3 Affibody molecule for medical imaging. Two HER3 binding Affibody molecules were site-specifically conjugated with the chelator NOTA and labeled with 111In. The two conjugates demonstrated specific binding to HER3-expressing BT-474 xenografts in mice. The results indicated that the imaging agent is appropriate for repeatable visualization of HER3 expression in metastatic cancer using SPECT.

37

Ken G Andersson

Paper V

The aim of this study was to further optimize the developed anti-HER3 Affibody molecule intended for stratification of patients for anti-HER3 immunotherapy. The Affibody molecule that showed best biodistribution in the previous study was labeled with 68Ga and used for PET imaging in mice. Tumor uptake correlated with HER3 expression in vivo, where uptake was significantly higher for the high HER3-expressing BT474 and BxPC-3 xenografts compared with the low HER3-expressing A431.

Paper VI

The aim of this study was to design and investigate an anti-EGFR Affibody for medical imaging. The anti-EGFR Affibody molecule ZEGFR:2377 was site-specifically conjugated with the chelator deforoxamine (DFO) and labeled with 89Zr. The long half-life of 89Zr (t½ 78.4 h) allows for good clinical translation and comparison with the anti-EGFR antibody cetuximab. Mice bearing EGFR-expressing A431 xenografts demonstrated higher tumor uptake 3 hours post injection compared to cetuximab 48 hours post injection.

38

7.1 Paper I - An engineered autotransporter-based surface expression vector enables efficient display of Affibody molecules on OmpT-negative E. coli as well as protease-mediated secretion in OmpT-positive strains

Directed evolution using cell-display is one of the fastest growing methods for protein engineering. In contrast to phage display, cell surface display has several features that are well-suited for library applications, such as the use of self-amplifying living cells, large particle size and multivalent display, allowing quantitative isolation of high-affinity binders using FACS191. Additionally, the highest affinity binders today, for both Affibody molecules and scFvs, have been isolated from cell-display using off-rate selection182,238. Although a promising method, cell-display, using yeast as example, does have some limitations. First, it requires expensive instruments to quantitatively isolate binders, and the secondly, the lower transformation efficiencies of yeast, as compared with E. coli, leads to smaller libraries than for phage display. While large libraries are potentially not equally important for affinity maturations, the affinity of the selected binders is still proportional to the size of the library239.

Transformation of DNA into E. coli generally results in much higher transformation efficiencies compared to other hosts195,201,239. Recently, it has been demonstrated that fusing the protein of interest to an autotransporter (type Va) enables secretion and subsequent display of recombinant proteins on the surface of E. coli. Thus, using autotransporters for display of protein libraries has the potential of combining the benefits of high transformation efficiency with the advantage of being able to quantitatively isolate high-affinity binders using FACS.

The aim of paper I was to modify and evaluate an E. coli-based system that could functionally display Affibody molecules on the outer membrane of the bacteria. Additionally, a secondary function was implemented into the expression vector, allowing secretion upon retransformation into OmpT containing strains. This was intended for i) surface display of Affibody molecule libraries for directed evolution ii) fast, post-selection secretion of isolated candidates for downstream characterization. We used the autotransporter Adhesin Involved in Diffuse Adherence (AIDA-I) in this study, which has previously been successfully used for display of various proteins on the surface, such as streptavidin and β-lactamase 240,241. To our knowledge, this autotransporter has not been used for large combinatorial libraries for naïve selection of affinity proteins.

We designed a dual function expression vector for surface display and secretion by introducing an IgG-binding Affibody molecule (Zwt) into the passenger sequence of AIDA-I. To enable normalization of the expression level in flow cytometry, an albumin binding protein (ABP), derived from Streptococcal Protein G with affinity for HSA was translationally fused to the IgG-binding motif. Between these two proteins, a hexa-histidine tag and an OmpT site were inserted to facilitate secretion and allow subsequent purification when transformed into an OmpT-expressing strain (Fig 10).

39

Ken G Andersson

Figure 10 Schematic illustration of the expression cassette A) The designed expression cassette (EC) contains a promoter region, signal peptide for IM translocation, an Affibody molecule binding IgG (Zwt), a his tag for purification, an OmpT protease substrate sequence, an albumin binding protein for surface expression normalization and lastly a β-domain for OM translocation of the passenger sequence. B) Surface display of the recombinant passenger sequence is only possible in OmpT negative strains. Secretion of the passenger is possible upon transformation of the expression cassette into an OmpT positive strain. C) Schematic drawing of a potential workflow for selection Affibody molecule and a subsequent secretion using OmpT positive strain.

40

Initial experiments using the naturally constitutive AIDA-I promoter (aidA) showed functional expression of both Zwt and ABP in OmpT-negative strains. However, the constitutive promoter could potentially result in problems with growth bias in future selections and therefore three alternative inducible promoters (T7, araBAD and rhaBAD) were tested (Fig 11). Evaluation using flow cytometry revealed that expressions using these promoters were significantly higher than the constitutive counterpart and that araBAD and rhaBAD were both highly titratable. Longer cultivation of cells containing the T7 promoter showed a decrease in cell viability and surface expression (data not shown) and was therefore not included in further studies.

Figure 11 Comparisons of different constitutive and inducible promoters for display of recombinant proteins on the surface of E. coli. A) Representative dot plot of the native constitutive AIDA-I promoter, Y-axis is target binding, while X-axis is representing albumin binding proteins interaction with labeled HSA. C) Flow-cytometric analysis of rhamnose and arabinose induced promoter show high expression and binding for both promoters.

41

Ken G Andersson

Next, we evaluated the range for suitable cultivation conditions under control of the two promoters araBAD and rhaBAD. Temperature, induction time and inductor concentration were evaluated by flow-cytometric analysis. The time points 1h, 3h, 5h and 16h were investigated at 25 °C, 30 °C and 37 °C, respectively. Additionally, the effects of the inductors (L-arabinose and L-rhamnose) were investigated at different concentrations (0.1%, 0.2%, 0.4%, 0.6%, 0.8% and 1.0%, respectively). Results from the studies on different cultivation conditions showed that induction for at least 3 hours was necessary for sufficient surface expression. Higher temperature yielded in general increased expression levels; however, overexpression-induced toxicity could be observed for the araBAD promoter at higher temperatures.

Surface expression of three unrelated Affibody molecules, targeting HER2, HER3 and TNF-α, respectively, were analyzed. High display levels and specific target binding could be observed using flow-cytometry, suggesting that the display system could be suitable for future library applications.

A flow-cytometric mock sorting was performed to simulate directed evolution and estimate the enrichment factor between each round. Bacteria expressing Zwt on the surface were mixed at a ratio of 1:100 000 with cells displaying HER2-specific Affibody molecules. One round of highly stringent sorting was performed on cells labeled with fluorescent IgG and the enrichment factor was determined by sequencing as well as flow cytometric analysis to approximately 20 000-fold per round. Additionally, by sorting out onto semi-solid agar plates containing selective antibiotics, it was determined that 90% (87± 9%) of the cells were viable (Fig 12).

Lastly, the developed secretion system was evaluated for small-scale production. This was an effort to decrease the time for downstream characterization of candidates selected using directed evolution on the microbial surface. Transformation into OmpT-positive strains enabled release of the surface-tethered Affibody molecule into the medium and the recombinantly added his-tag allowed straightforward purification using metal ion affinity chromatography (IMAC). The purified, secreted Affibody molecule showed no contaminations on SDS-PAGE and retained its binding affinity as determined with SPR. Mass spectrometry identified one peak at 9711, corresponding to the OmpT-cleaved theoretical weight of 9712 (Fig 13). Approximately 3 mg of released Affibody molecule could be purified per liter of culture.

In conclusion, our results suggest that the dual function-expression vector based on the AIDA-I autotransporter is a promising system for directed evolution of Affibody molecules. Transformation into OmpT positive strains allowed small-scale protein production for downstream characterization of the isolated candidates. The combination of high enrichment, high transformation efficiency, high cell viability and fast downstream characterization using a single expression vector without the need for sub-cloning shows the potential of the method to increase the throughput for engineering of Affibody molecules.

42

Figure 12 Surface expression of different Affibody molecules and fluorescence-activated cell sorting from spiked library. A) Representative dot-plots of grafted Affibody molecules against various targets. B) FACS of a spiked library 1:100 000 Zwt in ZHER2 was enriched in one round. The leftmost dot-plot represents spiked library before selection and the right dot-plot represents the library after one round of selection. Enrichment factor was deter-mined to roughly 20 000-fold per round. CFU:s after FACS onto LB-plate determined the viability to around 90%.

43

Ken G Andersson

Figure 13 Evaluation of OmpT-mediated secretion. A) Flow-cytometric analysis of Zwt on OmpT negative BL21 compared to OmpT positive Dh5α indicated decrease of signal and potential release of Zwt. B) Purification using IMAC and SDS-page analysis showed a single band for the induced Dh5-α strain. C) Detected size using ESI-MS correlated with theoretical 9712 Da. D) SPR-based biosensor analysis on the IMAC-purified protein showed retained binding to the immobilized IgG.

44

7.2 Paper II - Autotransporter-mediated display of a naïve Affibody library on the outer membrane of E. coliFollowing the proof-of-principle study in paper I, another study was conducted to select novel Affibody molecules from a large combinatorial library. Selecting Affibody molecules using cell display and FACS has been performed previously using the Gram-positive host Staphylococcus carnosus242. However, the relatively modest transformation frequency of Gram-positive bacteria makes construction of larger libraries for naïve selection very challenging and therefore pre-selections using phage display have mostly been employed182.

The aim of paper II was to further evaluate if microbial display on the surface of E. coli using autotransporters could be used for directed evolution of Affibody molecules. Here, a naïve library was created and five different biomarkers were used as targets in MACS and FACS. Three out of the five biomarkers named HER2, HER3 and interleukin 3 receptor subunit alpha (IL3RA, CD123) are involved in metastatic cancer progression56,243,244. The EGFR receptor family members, HER2 and HER3, have been previously used in selections of affibody molecules with phage or staphylococcal display. The target IL3RA, which is a diagnostic and therapeutic biomarker in many hematological malignancies, has not been evaluated as target for the selection of Affibody molecules. Similarly, the targets CD69 and DGCR2 have previously not been selected against. Selection against two previously selected targets and three novel, permits better understanding of the performance of the selection system.

First, a minimization of the expression vector was initiated to potentially increase transformation efficiency and perhaps allow larger passenger proteins to be expressed on the surface. The crystal structure of AIDA-I solved in 2014 by Gawarenski et al revealed that the passenger sequence contained an additional linker region of 169 amino acids. The importance of the linker region for stability of the native protein has previously been reported245,246. Deletion of this linker region, a N-terminal multiple cloning site and exchange of the albumin binding protein to a smaller albumin binding domain (ABD) reduced the size of the encoded surface-expressed protein from 50.6 kDa 14.2 kDa (Fig 14). Additionally, the OmpT site was replaced with a Sortase A motif (LPXTG), intended for conjugation and release of potential isolated candidates. Surface expression using flow cytometry was analyzed using the IgG interaction. Flow cytometry revealed no noticeable difference in IgG binding between the original and the minimized construct, suggesting maintained expression level.

The Affibody library was synthesized using trinucleotide synthesis, creating an oligonucleotide with 14 randomized positions in helix 1 and 2 with a theoretical diversity of around 1x1017 (Fig 15). The oligonucleotide was kindly provided by Professor Per-Åke Nygren. Subcloning into the expression vector and subsequent electroporation into E. coli Bl-21 resulted in a library of around 2.3x109 clones and was named royal library of the mean snake (RLMS). Noticeably, flow-cytometric analysis of surface expression of the library revealed that 93 % of the clones were displayed on the surface.

45

Ken G Andersson

The generated library was subjected to a first round of selection using the five different biotinylated targets bound to streptavidin-coated magnetic beads, respectively. Pre-selection using MACS decreased the library size, which allowed for subsequent FACS. Three to four rounds of enrichment were required, depending on the target, to obtain DNA sequence convergence (Fig 16). The HER2 and HER3 tracks yielded four unique clones each, IL3RA yielded twelve unique clones and both CD69 and DGCR2 yielded one single clone from sequencing 96 colonies from each track.

Figure 14 Minimization of passenger sequence in the expression vector. A) Crystal structure of AIDA-I reported the amino acids Q962 in the alpha helix protruding out of the beta barrel as the last amino acid position for domain folding (PDB:4MEEE). B) Minimization of the passenger sequence reduced the size of the expression cassette with 338 amino acids. C) Comparative illustration of the parental and the new passenger, which is 36.4 kDa smaller. D) Reduction of passenger sequence did not reduce surface expression of E. coli when analyzed with the IgG binding interaction.

46

The isolated binders were transferred into an expression vector for soluble expression of dimers fused to an ABD motif, adding additional downstream characterization possibilities and the option to employ human serum albumin for capture assays in SPR. Purification using IMAC successfully yielded 19 pure constructs as analyzed by SDS-PAGE. Following this, the targets were immobilized on SPR as ligands and each binder was injected as analyte. The SPR data suggested that neither IL3RA nor HER3 could be functionally immobilized, since no signal was observed for any construct or the positive control for HER3. The HER2 binders yielded signals and fitting the data estimated the affinities to low nanomolar for the best binders. Capture assay on surface-bound HSA using HER3, CD69 and DGCR2 binders showed binding for two of the HER3 clones as well as for the binders against DGCR2 and CD69 (Fig 17).

To conclude, directed evolution of Affibody molecules on the surface of E. coli appears to be a robust method for protein engineering efforts. The high transformation efficiency, high expression level and fast doubling times makes this host promising for engineering proteins that can functionally fold on the surface of E. coli. Although generating a library in E. coli is fairly straightforward, to our knowledge, the obtained library size for a naïve combinatorial protein library on the outer membrane of E. coli has not been reported before. However, other groups have reported peptide libraries (109 - 1010 variants) on the E. coli247 and both Salema et al and Binder et al have reported on affinity maturation libraries (2x109 variants) as well as immune nanobody libraries (3x106 variants) using autodisplay219,228. We have shown that expression of a large naïve Affibody library is well tolerated by the host and can be enriched for binders against different targets using both FACS and MACS. Although additional characterization of the isolated candidates is warranted, the results acquired so far indicate that the method has potential for directed evolution of Affibody molecules.

Figure 15 Illustration of the 14 different position randomized using trinucleotide synthesis in the Affibody molecule library.

47

Ken G Andersson

Figure 16 Selection from naïve library against five targets IL3RA, HER2 HER3 DGCR2 and CD69 using MACS and FACS. Pre-selections using MACS indicated a small enrich-ment and subsequent FACS selection for two or three rounds could enrich target specific cells.

48

Figure 17 SPR-based biosensor analysis on isolated binders from HER2, HER3 and DGCR2 track. A) Surface immobilized HER2 were analyzed for binding interaction by flowing over a concentration dilution between 800nM to 100nM of binders isolated from the HER2 track. B) Surface immobilized HSA permits Affibody molecule binders to be captured using the translationally fused ABD motif. Injecting respective target over the chip permits analysis of binding interaction. C) Captured Affibody molecules isolated from the HER3 track indicated that clone 23 and 57 could bind the HER3 receptor. D) The isolated candidate from DGCR2 track was captured onto the HSA surface and analysis showed that the candidate could bind its target.

49

Ken G Andersson

7.3 Paper III – Coupled release and site-specific conjugation of Affibody molecules from the surface of E. coli using Sortase AOne of the more time-consuming parts of the directed evolution process is the downstream characterization of isolated candidates after selection. Typically, the isolated candidates are sub-cloned, expressed in soluble form, purified and subjected to downstream characterization of the desired properties104

An intrinsic feature using cell display is the multivalent expression on the surface, which allows for small-scale soluble production directly after isolation of binders from directed evolution without the need for subcloning. Multiple approaches have been developed for small-scale soluble production using either a secretion mechanism, surface-expressed proteases or supplemented proteases248–250. Although valuable, some downstream assays also require the candidates to be labeled or conjugated to reporters or other functional groups, which is often practically not feasible due to the typical low yields of on-cell small-scale soluble production.

The aim of paper III was to develop and evaluate a novel approach for sortase-mediated direct labeling and secretion of isolated candidates from the cells after selection. The natural function of Sortase A is to cleave and conjugate proteins that contain the motif LPXTG (Sortase A motif ) to a substrate with one or more N-terminal solvent-exposed glycines251. Recently, Liu and colluegues successfully increased the catalytic rate (140-fold) of wild type Sortase A using mutagenic PCR and directed evolution252. The enhanced Sortase A, named eSrtA, has proven valuable for different applications, allowing conjugation with considerably shorter incubation times to various different poly-glycine substrate, such as synthetic peptides253, macrocyclic chelators254, and reporter molecules such as fluorescein210.

In this project, we wanted to investigate if Sortase A could be used for conjugation of various substrates to Affibody molecules on the surface of E. coli, and subsequent release of the conjugated protein from the host. The approach is illustrated in Figure 18 and is based on adding eSrtA and poly-glycine substrates to the recombinant bacteria, displaying Affibody molecules in fusion to a C-terminal LPXTG motif (Fig 18).

To evaluate this strategy, enhanced Sortase A and the fluorescent substrate GGG-mNeonGreen were first produced and purified. A first test for conjugation and release was prepared by adding enzyme, GGG-mNeonGreen substrate and bacteria displaying an IgG-binding Affibody molecule that was C-terminally fused to a LPXTG motif. Analysis on SDS-PAGE revealed a relatively weak band corresponding to conjugated and released Affibody molecule (Fig 19).

50

Following this first experiment, we evaluated the influence of growth media on conjugation efficiency using a GGGYK-biotin and mNeongreen substrate, respectively. The result from the transpeptidase reaction was analyzed using Western blot. Staining with streptavidin detected the biotinylated Affibody molecule and the highest conjugation efficiency was observed for bacteria grown in LB or TSB+Y medium. An anti-Affibody primary antibody detected the mNeonGreen conjugated Affibody molecule (Fig 20).

Figure 18 Schematic drawing of Sortase A mediated secretion and conjugation from the surface of E. coli.

Figure 19 Production of reagents and proof-of-principle experiment for Sortase A-medi-ated secretion and conjugation. A) Production of GGG-mNeonGreen an eSrtA-mScarletI showed no additional bands on mNeonGreen and a degradation product for eSrtA-mScar-letI. Red arrow indicates purified protein with correct size, black arrows indicates impuri-ties B) Schematic drawing of Sortase A cleavage from the surface. C) Proof-of-principle experiment indicates a faint band containing conjugated and release Affibody molecule. Here, red arrows indicate conjugated and secreted affibody (est. size: 37002 Da) and black arrows indicate released Affibody molecule without conjugation (est. size: 8664 Da).

51

Ken G Andersson

Finally, a more large-scale transpeptidase reaction, using 200 ml of surface-expressing bacteria and GGGYK-biotin as substrate was conducted. In this experiment, the conjugated Affibody molecules were purified by a heat-treatment step and IMAC. The results demonstrated that around 0.5 mg/l culture of pure product could be obtained after purification. This is somewhat lower than the OmpT-mediated secretion used in paper I, yielding 3 mg/l of culture. Two assays were thereafter used for analyzing the biotinylation and the functionality of the conjugated anti-IgG Affibody molecules. First, an ELISA using wells coated with IgG and detection of bound Affibody by streptavidin-HRP, and secondly, a flow-cytometry analysis on HER2 expressing NCI-N87 cells, treated with the human HER2-specific antibody trastuzumab and detection of bound Affibody using streptavidin R-phycoerythrin (SAPE). The results from the ELISA and the flow cytometry showed that the Affibody molecules were biotinylated and retained the affinity for IgG. Additionally, a similar flow-cytometric assay with non-purified anti-IgG Affibody molecules conjugated to GGGYK-fluorescein and GGG-mNeonGreen, respectively, showed that the conjugated and released Affibody molecule could bind to IgG while conjugated with the fluorescent reporters even without purification from sortase (Fig 21).

In conclusion, the results from this study indicate that Affibody molecules can be conjugated and released directly from the bacterial surface using sortase A, which might be an attractive approach for increasing the throughput of the downstream analysis in the future. Moreover, the method has the potential to be used for conjugation to other functional groups, such as cell penetrating peptides, cytotoxins or other binding motifs. It is important to note that relatively high amount of substrate is currently required for secretion and conjugation from the surface of E. coli, and optimization of the conjugation reaction is thus warranted.

Figure 20 Western blot detection of surface cleaved and released Affibody molecule. Conjugation and release from the surface of E. coli using two substrates, GGGYK-biotin and GGG-mNeonGreen, could successfully be identified using western blot with strepta-vidin-HRP detection as well as anti-goat HRP. Red arrows indicate conjugated and release Affibody molecule.

52

Figure 21 Large-scale production and downstream characterization of sortase A released Affibody molecule. A) Production in 200ml, compared to previous 2ml, using the GG-GYK-biotin substrate released the Affibody molecule into the supernatant and could be purified using IMAC. B) ELISA coated with IgG and detected using streptavidin-HRP showed signal only for the Affibody molecule released with biotin substrate. C) Flow-cyto-metric analysis on NCI-N87 cells treated with Herceptin and the conjugated and released Affibody molecules as secondary reagents.

53

Ken G Andersson

7.4 Paper IV - Comparative evaluation of 111In-labeled NOTA-conjugated Affibody molecules for visualization of HER3 expression in malignant tumorsAs described in chapter 2, development of prognostic and predictive diagnostic devices for cancer biomarkers is an essential part of precision medicine. Using molecular diagnostics to analyze the biological markers for personalized medicine in cancer normally requires a biopsy sample of the tumor. However, repeated biopsies could increase the risk of inducing metastasis in the unstable tumor environment81. Additionally, molecular diagnostics relying on biopsies will only provide information based on the excised primary tumor and not give a “whole-patient” diagnosis. Administration of radiopharmaceuticals that recognize prognostic or predictive biomarkers and subsequent imaging could be advantageous. Molecular imaging allows a semi-quantitative and an anatomically correct view of the primary tumor and potential lesions in the body and can be applied multiple times without significant side effects. This permits both monitoring of expression levels of biomarkers throughout therapy as well as determining cancer development in both primary tumor and potential metastatic lesions.

An example of a promising prognostic biomarker for molecular imaging is the EGFR family member HER3. As described in chapter 2, the EGFR family is a very well-investigated receptor family with comprehensive cross-talk between each receptor within the family. The importance of HER3 signaling in various cancers marks its potential as an anti-cancer therapy target and has stimulated development of several pharmaceuticals using monoclonal antibodies or small molecule inhibitors255,256. A growing amount of evidence suggests that HER3 expression is involved in resistance to EGFR- and HER2-targeted therapies244,257,258. This suggests that continuously monitoring HER3 receptor expression levels could provide vital information for physicians and avoid under- or overtreatment82. Using molecular imaging for diagnosis of HER3 is challenging, since tumors generally have low receptor expression below 50 000 receptors per cell259. Additionally, HER3 is relatively highly expressed in many normal human tissues260. Particularly challenging is the, HER3 expression in the liver since the organ is large, well perfused and has a well-fenestrated vasculature, which potentially results in elevated uptake and poor contrast. Additionally, dense vascularization makes this organ a common metastasis site and it is therefore very important to obtain high contrast in this organ for molecular imaging261.

54

The aim of the study in paper IV was to develop and establish an anti-HER3 Affibody molecule for medical imaging. The Affibody molecules used in this study had low picomolar affinity as well as cross-species specificity to both murine and human HER3262. The first reported in vivo imaging study with one of the isolated anti-HER3 Affibody molecule was using the ZHER3:08699 label with 99mTc using a peptide chelator263. This proof-of-principle study showed that an N-terminal HEHEHE-tag significantly decreased undesired unspecific organ uptake and that the tracer could locate HER3-expressing tumor cells. Encouraged by that study, we wanted to compare two different HER3-specific binders differing by only two amino acids substitutions: HEHEHE-ZHER3:08698 with approximately 50 pM affinity and HEHEHE-ZHER3:08699 with 21 pM affinity.

The two HER3-binding Affibody molecules were expressed in E. coli and purified by heating the lysate to 90 °C and subsequently rapidly cooling it to precipitate endogenous host proteins. Both proteins were site-specifically conjugated with the macrocylic chelator NOTA using maleimide chemistry. The conjugated protein was separated using reverse-phase high pressure liquid chromatography (RP-HPLC) to a purity level >95 % determined with analytical RP-HPLC using absorbance at 220 nm. Circular dichroism analysis indicated that the proteins were folded into an alpha-helical bundle and variable temperature measurement (VTM) showed that both proteins could refold after heating to 90 °C and that the melting temperature was 63 °C for HEHEHE-ZHER3:08698 and 65 °C for HEHEHE-ZHER3:08699 (Fig 22).

Figure 22 Characterization of HEHEHE-HER3 binders A) RP-HPLC showed that purity was >95%. B) VTM measurement indicated a melting temperature of 63°C for HEHEHE-ZHER308698 and 65°C for HEHEHE-ZHER308699. C) CD measurement showed that the protein adopted an alpha-helical structure and that refolding was possible after heat-treat-ing the protein to 95°C. D) ESI-MS confirmed the expected Mw and that conjugation to NOTA was successful.

55

Ken G Andersson

The NOTA chelator provides stable labeling with many radionuclides such as 68Ga, 64Cu, or 18F (using aluminum monofloride chemistry) for PET imaging. In this study, labeling was performed using the radiometal 111In for both molecules, permitting imaging with SPECT. A radiolabel challenge was done by incubating the labeled protein in ethylenediaminetetraacetic acid (EDTA) and serum. Analysis by radio instant thin-layer chromatography (ITLC) showed no release of free radionuclide in both samples, indicating stable labeling.

Figure 23 Tumor-to-organ ratios and imaging of mice showed that tumor visualization was possible. A) Tumor-to-organ ratios in mice with BT-474 xenografts at 1 h, 4 h, 8 h and 24 h pi. Imaging using gamma camera 4 h pi showed that both imaging agents could recognize and visualize the tumor in the hind leg of the mouse.

56

Characterization in vitro, using the HER3-expressing BT-474 breast cancer cell-line, showed that both binders associated with the cells and that addition of unlabeled HER3-specific Affibody molecule could block the interaction. Administration of both binders to BT-474 xenografted mice and measurement at 1 h, 4 h, 8 h, and 24 h post injection showed no significant difference in biodistribution at earlier time points (1-8 h post injection (pi). However, HEHEHE-ZHER3:08698 had improved tumor retention of radioactivity after 24 h pi and fast blood clearance resulting in better contrast. Gamma camera imaging of mice bearing HER3-expressing BT-474 xenografts was done at 4 h pi and could clearly visualize tumor as well as kidneys and liver (Fig 23).

In conclusion, we have shown in this study that anti-HER3 Affibody molecules can demonstrate specific binding to HER3-expressing BT-474 xenografts in vivo. Additionally, injection of unlabeled Affibody moleculse showed that binding to HER3 was saturable in both murine and xenografted tissue. The only difference between the tracers is two amino acids substitutions where ZHER3:08698

contained Q11 and W25 and ZHER3:08699 N11 and Y25. The relatively small difference did however influence the biodistribution, where HEHEHE-ZHER3:08698 had higher tumor-to-blood ratio, which can be considered superior for imaging. Compared to the previous study using 99mTc and peptide chelation, this study provided a 2-fold increase in tumor uptake suggesting that HEHEHE-containing NOTA-conjugated Affibody molecule is preferable for development of HER3-imaging agents.

57

Ken G Andersson

7.5 Paper V - Affibody-mediated PET imaging of HER3 expression in malignant tumoursBeing able to determine receptor concentrations on the surface of cancer cells provides valuable information for diagnosis. The imaging modality PET provides better resolution, higher sensitivity and superior quantification accuracy than SPECT. In fact, an impressive study using 18F-labeled anti-HER2 Affibody molecules managed to semi-quantitatively assess expression in breast cancer xenografts using PET imaging and comparisons with ex vivo methods such as ELISA and Western blot correlated with the PET results264. PET imaging does however require more expensive equipment and the use of beta-decaying radionuclides such as 18F (t½ 110min) and 68Ga (t½ 68min).

The aim of the study in paper V was to further develop the HEHEHE-ZHER3:08698 and determine if PET imaging is possible when labeled with 68Ga. Further, we wanted to examine if this method could discriminate between tumors with high and low HER3 expression. Four cell lines with different HER3 expression levels were used in this study: BT-474 (breast cancer), BxPC-3 (pancreatic cancer) LS174T (colorectal cancer) and the negative control A431 (epidermoid cancer)

Figure 24 In vitro characterization of four cell lines: BT-474, BxPC-3 LS174T and A431. A) Estimated receptors per cell using radiolabeled Affibody molecule and gamma counter on the four cell lines revealed that BT-474 contained roughly 10-fold more HER3 receptors per cell compared to the A431 cell line. B) Analysis of cell binding and internali-zation of gallium labeled HEHEHE-ZHER308698 indicated a lower internalization rate for the LS174T cell line compared to BT-474 and BxPC3.

58

The probe HEHEHE-ZHER3:08698 was labeled with 68Ga at 95 °C, providing a labeling yield of >98% after purification. Challenge using free EDTA chelator did not show any release of radioactivity as analyzed by ITLC. Quantification of HER3 receptor expression was done by adding labeled Affibody molecules to the cell-line and compare to a non-labeled saturated cell-line using gamma counter. This quantification indicated highest HER3 expression in BT-474 (25 ± 2 × 103 receptors), followed by BxPC-3 (12 ± 2 × 103) and LS274T cells (4 ± 1 × 103) while A431 had lowest HER3 receptor expression on the surface. Additionally, determination of internalization rate of bound HER3 showed that BT-474 and BxPC-3 cells demonstrated high internalization of the radioactivity where roughly 20-30 % of associated radioactivity was internalized after 4 h. LS174T cell lines had a much slower internalization rate which did not exceed 5% of the cell-associated radioactivity at the end of the observation (Fig 24).

For the in vivo studies, mice carrying LS174T xenografted tumors, were injected with two doses (1 and 2 µg of protein) containing the same amount of radioactivity. Interestingly, this comparison showed a significant decrease of the accumulated radioactivity in normal organs while radioactivity in blood remained unaltered (Fig 25). Although not fully understood, this indicates that even small differences in protein dosage can lead to changes in imaging contrast and is therefore a very important parameter to optimize for clinical purposes. In this study however, we did not increase the protein dosage since previous studies with anti-HER2 Affibody molecules have shown that increase of dosage may cause reduction in tumor uptake265.

Figure 25 Tumor-to-organ ratios of two injected protein doses on LS174T tumor bearing mice. Injection of higher protein dose resulted in significantly higher tumor-to-organ ratios for HER3 expressing tissues such as salivary gland, liver and small intestine.

59

Ken G Andersson

In vivo specificity studies at 3 h pi on mice carrying xenografts demonstrated a saturable and specific uptake of the tracer in the tumor environment. Additionally, the tumor uptake from the specificity study correlated significantly with the estimated receptor expression (p=0.002, r=0.66) (Fig 26). This might provide discrimination between metastases with high and low HER3 expression which could be relevant for clinical development. Imaging using microPET/CT demonstrated clear visualization of tumors dependent on HER3 expression. High radioactivity was likewise visualized for kidneys, liver and excretory organs which was in agreement with the biodistribution data.

In conclusion, we have shown in this study that HEHEHE-ZHER3:08698 is a promising tracer for PET- imaging of malignant HER3 expressing tumors, and that it can provide a clear visualization of xenograft tumors 3 h pi and furthermore permit discrimination between high and low HER3 expression. Currently, radioactivity uptake in liver is higher than in the tumor due to endogenous HER3 expression and may decrease contrast for visualization of abdominal tumors using this tracer.

Figure 26 Tumor uptake of HER3 specific Affibody molecule showed correlation with estimated receptors per cell and imaging could visualize the tumor. A) Linear regression analysis on tumor uptake to the previously estimated receptors per cell showed correlation.B) microPET/CT could visualize mice bearing i) BT-474 ii) BxPC-3 iii) LS174T xenografts 3h pi. Coronal image using scale 0-3%ID/g could clearly visualize tumor in all xenografts.

60

7.6 Paper VI - PET imaging of epidermal growth factor receptor expression in tumours using 89Zr- labelled ZEGFR:2377 Affibody moleculesThe epidermal growth factor receptor (EGFR) discussed in chapter 2 is abnormally expressed in many epithelial tumors266. Several anti-EGFR therapeutics are today approved by FDA, such as monoclonal antibodies (cetuximab and panitumumab) as well as tyrosine kinase inhibitors (gefitinib and erlotinib), with many more in clinical development267. Overexpression of EGFR is associated with poor prognosis in many different types of cancer, e.g. breast cancer, head and neck squamous cell carcinoma (HNSCC) and non-small-cell lung cancer (NSCLC)64,268,269. Therefore, overexpression could be used for stratification of patients with advanced NSCLC to the small molecule gefitinib and to patients benefitting from chemotherapy in combination with cetuximab270,271. While patients with high EGFR expression in NSCLC could benefit from treatment with cetuximab, low expression of EGFR in tumor does not increase overall survival272. Thus detection of high EGFR expression in malignant tumors could provide important prognostic and predictive information for patient stratification.

As described in chapter 2, molecular imaging using antibodies should be a straightforward way for developing an imaging agent. However, the large molecule is not filtered by the glomeruli, which together with FcRn rescue results in very long circulation time and thus poor imaging contrast. Furthermore, the enhanced permeability and retention (EPR) effect leads to accumulation of large macromolecules, such as antibodies, in tumor, making quantification and sensitivity lower in a molecular imaging setting. Regardless, early pre-clinical and clinical studies using antibody 225 (murine predecessor of cetuximab) have shown that EGFR can be visualized in both mice and humans using SPECT273,274. Higher resolution and sensitivity could be achieved using positron-emission tomography (PET). While, positron-emitting radionuclides such as 64Cu (t½ 12.7 h) and 86Y (t½ 14.7 h) have shown promising results in murine models at 24-48 h, the half-life of these nuclides is too short for clinical translation. Using the more long-lived 89Zr (t½ 78.4 h) is probably the most viable option in a clinical setting.

An alternative to antibodies is using the natural ligand of EGFR, named epidermal growth factor (EGF). Labeling EGF with 111In has demonstrated faster clearance, higher tumor-to-blood ratios and higher specificity compared to antibodies275. However, one major issue with the use of EGFR is its strong agonistic action that causes side effects, such as vomiting, nausea and hypertension when injected at higher doses276. Using a small tracer without agonistic properties would be ideal for molecular imaging. Therefore, a sub-nanomolar, high-affinity anti-EGFR Affibody molecule has been developed with cross-reactivity to both human and murine EGFR277.

61

Ken G Andersson

The aim of the study in paper VI was to evaluate and compare the imaging properties of an anti-EGFR Affibody molecule with the monoclonal antibody cetuximab. Both tracers were labeled with 89Zr and imaged using PET.

The anti-EGFR Affibody molecule ZEGFR:2377 was produced in E. coli and purified without affinity tag using heat treatment and ion-exchange chromatography. Site-specific conjugation to a C-terminal cysteine using maleimide chemistry allowed deferoxamine (DFO) to be conjugated to the Affibody molecule. The antibody cetuximab and the Affibody molecule ZEGFR:2377 were both conjugated with DFO and labeled with 89Zr to a purity of >99% as measured with ITLC. The affinity of ZEGFR:2377 for EGFR on an A431 tumor cell line was estimated using LigandTracer to around 160 pM. In vitro specificity studies showed that ZEGFR:2377 binding was saturable and cellular processing indicated a relatively slow internalization rate with approximately 40 % of the radioactivity residing within the cells after 24 h (Fig 27).

An in vivo saturation experiment for ZEGFR:2377 revealed a significant reduction of uptake in EGFR-expressing tissues, such as salivary gland, liver and lung. Additionally, there was a significant reduction of uptake in muscle and blood while renal uptake was increased. This suggests that renal uptake is not EGFR-mediated and that saturation increases re-absorption of tracers from the blood into kidneys. Comparing the in vivo biodistribution between 89Zr labeled ZEGFR:2377 and cetuximab showed that ZEGFR:2377 had significantly higher tumor-to-organ ratios in all studied organs except for kidneys compared to the antibody at both 3 h and 24 h pi. At 24 h pi, the tumor-to-blood ratio was 1.7-fold higher compared to 3 h pi, but at the cost of generally lowered tumor uptake and increased uptake in bone. In vivo imaging using PET-CT of A431 xenograft-carrying mice clearly visualized the tumor and the results were in good agreement with the biodistribution data (Fig 28).

Figure 27 in vitro characterization of ZEGFR2377. A) Representative LigandTracer sensor-gram of ZEGFR2377 binding to A431 using 0.33 nM and 1 nM Affibody molecule indicated an affinity of 160±60 pM. B) In vitro specificity against A431 cells showed that binding of Affibody molecule was blockable with both cetuximab as well as unlabeled ZEGFR2377. C) Cellular processing of ZEGFR2377 on A431 cells indicated a relatively slow internalization.

62

To conclude, in paper VI we have labeled an anti-EGFR Affibody molecule with the beta-decaying radionuclide 89Zr that allows for PET imaging of EGFR expression in malignant tumors. The long-lived radionuclide 89Zr permits PET imaging several days after injection. Furthermore, comparative studies showed that the anti-EGFR Affibody molecule had higher tumor retention and better biodistribution already 3 hours after injection compared to the antibody. The biodistribution of 89Zr-DFO-cetuximab was measured at 48 hours after injection since the literature suggests that this time point is optimal for imaging of A431 xenografts. Additionally, the analyzed data of cetuximab was in good agreement with the biodistribution described in a previously published study by Aerts et al278. This cumulative data suggests that the Affibody molecule is a more sensitive imaging agent and could be an interesting alternative for further clinical development.

Figure 28 Tumor-to-organ ratios and visualization of ZEGFR2377. A) Comparative study with mice bearing A431 xenografts between cetuximab and ZEGFR2377 showed better tu-mor-to-organ ratios for ZEGFR2377 4 h pi. B) Imaging of ZEGFR2377 using microPET/CT 3 h and 24 h pi indicated a clear visualization of the tumor as well as liver and kidney.

63

Ken G Andersson

7.7 Concluding remarks and future perspectivesThe work in paper I describes the use of the AIDA-I autotransporter for display of Affibody molecules on the surface of E. coli and evaluation of FACS-mediated enrichment from a spiked library. Further, the expression vector could be used for both surface expression as well as for small-scale production of soluble proteins by secretion into the culture medium. The high expression level, good transformation efficiency, high viability and high enrichment between selection rounds are all attractive features for use in combinatorial protein engineering. Although surface expression levels of Affibody molecules using AIDA-I are very high, it is noteworthy that the Affibody molecule is originally derived from a bacterial surface protein. Consequently, they could potentially be more easily expressed and displayed on bacteria compared to other proteins. However, other groups have expressed more complex proteins such as streptavidin and β-lactamase in fusion to AIDA-I indicating that this autotransporter might be used for directed evolution of more complex proteins as well234,241. Recombinant proteins containing multiple cysteines is a general challenge for bacterial display using autotransporters, since such proteins will most likely hamper outer membrane translocation by forming intermolecular disulfide bridges in the periplasm. Although challenging, some groups have managed to display cysteine-rich proteins, such as the human cytochrome p450 containing nine unique cysteine residues279. Even if some cysteine-rich proteins have been displayed on the surface, alternative E. coli strains, such as DsbA knock out strains, can be used to make it more efficient. These DsbA-negative strains reduce the formation of disulfide bonds in the periplasm, which would improve the translocation through the OM. One potential issue with these strains is that disulfide bond formation has to occur by spontaneous oxidation, which will not be equally efficient or accurate as the enzyme-catalyzed reaction, and could hamper the functionality recombinant proteins280.

The method that was developed in paper I was utilized in both paper II and III. Construction of a library and electroporation into E. coli resulted in a library size of around 2.3x109 members. MACS and subsequent FACS sorting allowed isolation of binders against five different targets (HER2, HER3, CD69, DGCR2 and IL3RA). Follow-up studies should be focused on more detailed characterization of the new candidates, and in particular on identification of the corresponding epitopes of each isolated binder. Speculatively, the binders against HER2 and HER3, which showed small or no homology to previously selected HER2 and HER3 binders that were isolated using phage display, might recognize new epitopes. The negative charge of the E. coli membrane might potentially influence target binding and orientation and therefore result in different binding epitopes on the target. Novel epitopes could for example be of great interest for construction of so-called biparatopic binders, as has been shown previously using other Affibody molecules242. Regardless of the binding epitope, the isolated binders could still be interesting for molecular imaging, since many studies have shown that even small changes in amino acid composition can have great influence on the biodistribution. The display platform using AIDA-I and E. coli as a host has the potential to be an attractive complement to other cell display methods and phage display. However, more studies are needed to fully understand its potential benefits and limitations, compared

64

to both yeast and phage display. In a more long-term perspective, this method can hopefully be an efficient tool for generation of binders against biomarkers or molecules that eventually result in novel therapies, biotechnological solutions or medical imaging agents.

In paper III, we investigated a novel method for conjugation and release of isolated binders from the bacterial surface using Sortase A. We demonstrated that it is possible to conjugate and release Affibody molecules using eSrtA together with reporter molecules ranging from around 1 kDa to ~30 kDa. This method could be valuable for straightforward conjugation of various reporter molecules to new isolated candidates without the need for subcloning. The approach should also be applicable for conjugation to other functional groups, such as for example cytotoxic drugs when evaluating the potential of different binders as targeting-domains for development of ADCs. The isolated binders could thus be rapidly screened for best cytotoxic effects on cancer cell lines. Additionally, the conjugation method with other types of substrates could be used to screen isolated binders for many different features, such as intracellular delivery, pro-inflammation in immune checkpoints, translocation over barriers or antimicrobial properties. As an example, a PD-1 binding motif, disabling the tumors T-cell supression , could be coupled to the Affibody molecule binding a biomarker. This conjugate could be straightforwardly screened in an immune assay containing T-cells and tumorigenic cells expressing the biomarker. Measuring apoptosis or T-cell activation could provide a fast indication on the efficacy of this Affibody molecule PD-1 conjugation. In the future, it would be very interesting to see if this strategy for conjugation could be implemented in the directed evolution process for increasing the throughput and in the end result in more information when performing the internal ranking of selected binders.

In paper IV, we described a comparative study between two Affibody molecules differing in only two amino acids. We showed that both constructs had high tumor uptake and fast clearance, suggesting that these imaging agents can enable a non-invasive and repeatable visualization of HER3 expression in metastatic cancer. This could be valuable for detection of overexpression in tumors and lesions which would suggest including anti-HER3 therapy in the treatment. Although no anti-HER3 therapies are currently approved, both petritumab and seribantumab have shown promising results in clinical trials. The minor differences in amino acids had an impact on biodistribution in vivo, suggesting that altering as little as 2% of the amino acids in this protein significantly changes renal blood clearance as well as liver and small intestinal uptake. Already at 4 h pi, both Affibody molecules demonstrated reasonable imaging of the tumor using 111In, indicating that this imaging agent can be used with more short-lived radionuclides such as 18F (t½ 110min) and 68Ga (t½ 68min) for PET imaging.

In paper V, we reported on the first PET imaging using anti-HER3 Affibody molecules, which could clearly visualize tumors in vivo. A clear increase in sensitivity was observed compared to SPECT, which permitted in vivo semi-quantification of HER3-receptor density in the xenografted tumors. The high sensitivity that PET imaging provides, together with the Affibody imaging agent, should have high potential for patient stratification in terms of HER3 expression in future personalized medicine.

65

Ken G Andersson

Lastly, the work in paper VI compared the antibody cetuximab with an anti-EGFR Affibody molecule, both labeled with 89Zr. Here, we showed that the Affibody molecule has better tumor uptake than the antibody already after 4 h pi compared to the antibody 48 h pi. High tumor retention and fast blood clearance indicate that this imaging agent is suitable for further clinical development. The EGFR receptor is expressed in many normal tissues, such as skin and placenta, which makes imaging more challenging. Regardless, imaging using the anti-EGFR Affibody molecule could clearly visualize tumors in mice and had very good contrast in almost all organs except for kidney and liver. Recently, we have shown in another paper using the same anti-EGFR Affibody molecule that labeling with 57Co significantly decreases liver uptake (3-fold) compared to labeling with 68Ga. The only difference between these two agents is electrostatical, where 57Co contains one C-terminal local negative net-charge which is not present in 68Ga281. The exact reason for this observed phenomenon is still not quite clear and it would therefore be interesting to further investigate the influence of charge differences in the C-terminus on biodistribution. One approach for such studies, could be addition of for example glutamic acids to the C-terminal, followed by analysis of biodistribution patterns. While 57Co currently shows the best biodistribution of the labeled EGFR binders, this radionuclide is probably not clinically translatable, since the half–life is too long (t½ 271.8d). One alternative could be to exchange this radionuclide for 55Co (t½ 17.5h) with much shorter half-life.

Speculatively, to maximize the prognostic and predictive capabilities in the EGFR receptor family, a multiplex approach measuring two or more receptors in the family simultaneously throughout tumor progression could be highly beneficial for patient outcome. Theoretically, this could be achieved using radioimaging with dual isotopes on both the Affibody molecule tracers described in this thesis. Using two different isopopes such as 57Co and 111In and measure non-overlapping emission spectra would allow semi-quantitative measurement of both HER3 and EGFR simultaneously. Hopefully, this multiplex analysis could be used in future clinics to further understand tumoral and metastatic progression and potential resistances against therapies.

In conclusion, the work presented in this thesis describes the expansion and evaluation of a bacterial display platform using E. coli to develop Affibody molecules against tumor targets for applications in therapy, biotechnology and medical imaging. Due to the short half-life and fast tumor penetration of the Affibody molecule, it is a very interesting type of affinity protein for development of imaging agents. Here, we have shown that rational site-specific conjugation of EGFR- and HER3-binding Affibody molecules permits identification and visualization of tumors in xenografted mice in vivo. However, improving contrast in medical imaging using protein-based probes is still a very complex field. In this thesis, we have shown that changes in amino acids, N-terminal purification tags, labels and protein dosing, all contribute to changes in contrast and have to be evaluated and addressed before clinical development.

66

Bibliography1. Bianconi, E. et al. An estimation of the number of cells in the human body. Ann Hum Biol 40,

301–4460 (2013).2. Wilhelm, M. et al. Mass-spectrometry-based draft of the human proteome. Nature 509, 582–7

(2014).3. Kim, M.-S. et al. A draft map of the human proteome. Nature 509, 575–581 (2014).4. Uhlen, M. et al. Towards a knowledge-based Human Protein Atlas. Nat. Biotechnol. 28,

1248–50 (2010).5. IUPAC. Compendium of Chemical Terminology. Gold Book (2014). doi:10.1351/goldbook.

I033526. Edsall, J. T. Kaj Ulrik Linderström-Lang 1896-1959. Adv. Protein Chem. 14, xii–xxiii (1959).7. Kendrew, J. C. et al. A Three-Dimensional Model of the Myoglobin Molecule Obtained by

X-Ray Analysis. Nature 181, 662–666 (1958).8. Petsko, G. a & Ringe, D. Protein Structure and Function. Protein Structure and Function (2004).

doi:10.1088/1674-1056/23/7/0787059. Utkin, Y. N. Animal venom studies: Current benefits and future developments. World J. Biol.

Chem. 6, 28 (2015).10. Sanchez-Garcia, L. et al. Recombinant pharmaceuticals from microbial cells: a 2015 update.

Microb. Cell Fact. 15, 33 (2016).11. Goeddel, D. V et al. Expression in Escherichia coli of chemically synthesized genes for human

insulin. Proc. Natl. Acad. Sci. U. S. A. 76, 106–110 (1979).12. The, M. J. Human insulin: DNA technology’s first drug. Am. J. Hosp. Pharm. 46, S9-11

(1989).13. Leader, B., Baca, Q. J. & Golan, D. E. Protein therapeutics: a summary and pharmacological

classification. Nat. Rev. Drug Discov. 7, 21–39 (2008).14. Imai, K. & Takaoka, A. Comparing antibody and small-molecule therapies for cancer. Nat.

Rev. Cancer 6, 714–727 (2006).15. Liu, J. K. H. The history of monoclonal antibody development - Progress, remaining challenges

and future innovations. Annals of Medicine and Surgery 3, 113–116 (2014).16. Schumock, G. T. et al. National trends in prescription drug expenditures and projections for

2016. Am. J. Heal. Pharm. 73, 1058–1075 (2016).17. Kola, I. & Landis, J. Can the pharmaceutical industry reduce attrition rates? Nat. Rev. Drug

Discov. 3, 1–5 (2004).18. Reichert, J. M., Rosensweig, C. J., Faden, L. B. & Dewitz, M. C. Monoclonal antibody

successes in the clinic. Nat. Biotechnol. 23, 1073–1078 (2005).19. Ecker, D. M., Jones, S. D. & Levine, H. L. The therapeutic monoclonal antibody market.

mAbs 7, 9–14 (2015).20. Schreiber, G. & Keating, A. E. Protein binding specificity versus promiscuity. Current Opinion

in Structural Biology 21, 50–61 (2011).21. St John, A. & Price, C. P. Existing and Emerging Technologies for Point-of-Care Testing. Clin.

Biochem. Rev. 35, 155–67 (2014).22. Chard, T. Pregnancy tests: a review. Hum. Reprod. 7, 701–10 (1992).23. Bandi, Z. L., Schoen, I. & DeLara, M. Enzyme-linked immunosorbent urine pregnancy tests.

Clinical specificity studies. Am. J. Clin. Pathol. 87, 236–242 (1987).24. Engvall, E. & Perlmann, P. Enzyme-linked immunosorbent assay (ELISA) quantitative assay

of immunoglobulin G. Immunochemistry 8, 871–874 (1971).25. Filice, G. et al. Sensitivity and specificity of anti-HIV ELISA employing recombinant (p24,

p66, gp120) and synthetic (gp41) viral antigenic peptides. Microbiologica 14, 185–94 (1991).

67

Ken G Andersson

26. Leng, S. X. et al. ELISA and Multiplex Technologies for Cytokine Measurement in Inflammation and Aging Research. Journals Gerontol. Ser. A Biol. Sci. Med. Sci. 63, 879–884 (2008).

27. Noedl, H. et al. Sensitivity and specificity of an antigen detection ELISA for malaria diagnosis. Am. J. Trop. Med. Hyg. 75, 1205–1208 (2006).

28. Sblattero, D. et al. Human recombinant tissue transglutaminase ELISA: An innovative diagnostic assay for celiac disease. Am. J. Gastroenterol. 95, 1253–1257 (2000).

29. Lee, T. M. H. Over-the-counter biosensors: Past, present, and future. Sensors 8, 5535–5559 (2008).

30. Spear, B. B., Heath-Chiozzi, M. & Huff, J. Clinical application of pharmacogenetics. Trends in Molecular Medicine 7, 201–204 (2001).

31. J.T., J. Companion diagnostics: The key to personalized medicine. Expert Rev. Mol. Diagn. 15, 153–156 (2015).

32. Jørgensen, J. T. & Hersom, M. Companion diagnostics—a tool to improve pharmacotherapy. Ann. Transl. Med. 4, 482–482 (2016).

33. Papadopoulos, N., Kinzler, K. W. & Vogelstein, B. The role of companion diagnostics in the development and use of mutation-targeted cancer therapies. Nat. Biotechnol. 24, 985–995 (2006).

34. Slamon, D. J. et al. Use of Chemotherapy plus a Monoclonal Antibody against HER2 for Metastatic Breast Cancer That Overexpresses HER2. N. Engl. J. Med. 344, 783–792 (2001).

35. Romond, E. H. et al. Trastuzumab plus Adjuvant Chemotherapy for Operable HER2-Positive Breast Cancer. N. Engl. J. Med. 353, 1673–1684 (2005).

36. Papaldo, P. et al. A phase II study on metastatic breast cancer patients treated with weekly vinorelbine with or without trastuzumab according to HER2 expression: Changing the natural history of HER2-positive disease. Ann. Oncol. 17, 630–636 (2006).

37. Press, M. F. et al. Evaluation of HER-2/neu gene amplification and overexpression: Comparison of frequently used assay methods in a molecularly characterized cohort of breast cancer specimens. J. Clin. Oncol. 20, 3095–3105 (2002).

38. Cardamone, L. et al. Chemistry, Toxicology and QSAR : an introduction, Politécnico de Milano, Instituto di Ricerche Farmacologiche ‘Mario Negri’. Chem. Inf. Model. 45, 1565–1554 (2010).

39. Ferlay, J. et al. Cancer incidence and mortality worldwide: Sources, methods and major patterns in GLOBOCAN 2012. Int. J. Cancer 136, E359–E386 (2015).

40. Clark, W. H. Tumour progression and the nature of cancer. Br J Cancer 64, 631–644 (1991).41. Li, J. & King, M. R. Adhesion receptors as therapeutic targets for circulating tumor cells.

Front. Oncol. 2, 79 (2012).42. Fais, S. Cannibalism: A way to feed on metastatic tumors. Cancer Letters 258, 155–164 (2007).43. Hanahan, D. & Weinberg, R. A. The hallmarks of cancer. Cell 100, 57–70 (2000).44. Hanahan, D. & Weinberg, R. A. Hallmarks of cancer: The next generation. Cell 144, 646–674

(2011).45. Weigelt, B., Geyer, F. C. & Reis-Filho, J. S. Histological types of breast cancer: How special

are they? Molecular Oncology 4, 192–208 (2010).46. Perou, C. M. et al. Molecular portraits of human breast tumours. Nature 406, 747–752 (2000).47. Marusyk, A. & Polyak, K. Tumor heterogeneity: Causes and consequences. Biochim. Biophys.

Acta - Rev. Cancer 1805, 105–117 (2010).48. Russnes, H. G., Navin, N., Hicks, J. & Borresen-Dale, A. L. Insight into the heterogeneity of

breast cancer through next-generation sequencing. J. Clin. Invest. 121, 3810–3818 (2011).49. Fisher, R., Pusztai, L. & Swanton, C. Cancer heterogeneity: implications for targeted

therapeutics. Br. J. Cancer 108, 479–485 (2013).

68

50. Vu, T. & Claret, F. X. Trastuzumab: updated mechanisms of action and resistance in breast cancer. Front. Oncol. 2, 62 (2012).

51. Ross, J. S. Targeted Therapy in Breast Cancer: The HER-2/neu Gene and Protein. Mol. Cell. Proteomics 3, 379–398 (2004).

52. Wang, Q. et al. Mutant proteins as cancer-specific biomarkers. Proc. Natl. Acad. Sci. 108, 2444–2449 (2011).

53. Strimbu, K. & Tavel, J. A. What are biomarkers? Curr. Opin. HIV AIDS 5, 463–466 (2010).54. Nalejska, E., Maczynska, E. & Lewandowska, M. A. Prognostic and predictive biomarkers:

Tools in personalized oncology. Mol. Diagnosis Ther. 18, 273–284 (2014).55. Ménard S, Fortis S, Castiglioni F, Agresti R, B. A. HER2 as a prognostic factor in breast

cancer. Oncology 61, 67–72 (2001).56. Piccart, M., Lohrisch, C., Di Leo, a & Larsimont, D. The predictive value of HER2 in breast

cancer. Oncology 61 Suppl 2, 73–82 (2001).57. Bublil, E. M. & Yarden, Y. The EGF receptor family: spearheading a merger of signaling and

therapeutics. Curr. Opin. Cell Biol. 19, 124–134 (2007).58. Lingi, B. & Carpenter, G. ErbB receptors: new insights on mechanisms and biology. Trends

Cell Biol. 16, 649–656 (2006).59. Garrett, T. P. J. et al. The crystal structure of a truncated ErbB2 ectodomain reveals an active

conformation, poised to interact with other ErbB receptors. Mol. Cell 11, 495–505 (2003).60. Tzahar, E. et al. A hierarchical network of interreceptor interactions determines signal

transduction by Neu differentiation factor/neuregulin and epidermal growth factor. Mol. Cell. Biol. 16, 5276–87 (1996).

61. Sliwkowski, M. X. et al. Coexpression of erbB2 and erbB3 proteins reconstitutes a high affinity receptor for heregulin. J. Biol. Chem. 269, 14661–14665 (1994).

62. Burden, S. & Yarden, Y. Neuregulins and Their Receptors: A Versatile Signaling Module in Organogenesis and Oncogenesis. Neuron 18, 847–855 (1997).

63. Irish, J. C. & Bernstein, A. Oncogenes in Head and Neck Cancer. Laryngoscope 103, 42(52) (1993).

64. Selvaggi, G. et al. Epidermal growth factor receptor overexpression correlates with a poor prognosis in completely resected non-small-cell lung cancer. Ann. Oncol. 15, 28–32 (2004).

65. Ritter, C. A. et al. Human breast cancer cells selected for resistance to trastuzumab in vivo overexpress epidermal growth factor receptor and ErbB ligands and remain dependent on the ErbB receptor network. Clin. Cancer Res. 13, 4909–19 (2007).

66. Lo, H.-W. et al. Nuclear interaction of EGFR and STAT3 in the activation of the iNOS/NO pathway. Cancer Cell 7, 575–589 (2005).

67. Ecker, D. M., Jones, S. D. & Levine, H. L. The therapeutic monoclonal antibody market. MAbs 7, 9–14 (2015).

68. Blaszczak, W., Barczak, W., Wegner, A., Golusinski, W. & Suchorska, W. M. Clinical value of monoclonal antibodies and tyrosine kinase inhibitors in the treatment of head and neck squamous cell carcinoma. Med. Oncol. 34, 60 (2017).

69. Schick, U., Gujral, D. M., Richards, T. M., Harrington, K. J. & Nutting, C. M. Zalutumumab in head and neck cancer. Expert Opin. Biol. Ther. 12, 119–125 (2012).

70. Ross, J. S. & Fletcher, J. A. The HER&amp;hyphen;2&amp;sol; neu Oncogene in Breast Cancer: Prognostic Factor, Predictive Factor, and Target for Therapy. Stem Cells 16, 413–428 (1998).

71. Mignot, F. et al. Concurrent administration of anti-HER2 therapy and radiotherapy: Systematic review. Radiother. Oncol. (2017). doi:10.1016/j.radonc.2017.07.006

72. Verma, S. et al. Trastuzumab Emtansine for HER2-Positive Advanced Breast Cancer. N. Engl. J. Med. 367, 1783–1791 (2012).

69

Ken G Andersson

73. Yarden, Y. & Sliwkowski, M. X. Untangling the ErbB signalling network. Nat. Rev. Mol. Cell Biol. 2, 127–137 (2001).

74. Qian, G. et al. Heregulin and HER3 are prognostic biomarkers in oropharyngeal squamous cell carcinoma. Cancer 121, 3600–3611 (2015).

75. Lee-Hoeflich, S. T. et al. A Central Role for HER3 in HER2-Amplified Breast Cancer: Implications for Targeted Therapy. Cancer Res. 68, 5878–5887 (2008).

76. Yonesaka, K. et al. Activation of ERBB2 Signaling Causes Resistance to the EGFR-Directed Therapeutic Antibody Cetuximab. Sci. Transl. Med. 3, (2011).

77. Zhang, N., Chang, Y., Rios, A. & An, Z. HER3/ErbB3, an emerging cancer therapeutic target. Acta Biochim. Biophys. Sin. (Shanghai). 7, gmv103 (2015).

78. Mohd Nafi, S. N. et al. Nuclear HER4 mediates acquired resistance to trastuzumab and is associated with poor outcome in HER2 positive breast cancer. Oncotarget 5, 59934–5949 (2014).

79. Hollmén, M. & Elenius, K. Potential of ErbB4 antibodies for cancer therapy. Future Oncol. 6, 37–53 (2010).

80. Hollmén, M., Määttä, J. A., Bald, L., Sliwkowski, M. X. & Elenius, K. Suppression of breast cancer cell growth by a monoclonal antibody targeting cleavable ErbB4 isoforms. Oncogene 28, 1309–1319 (2009).

81. Shyamala, K., Girish, H. C. & Murgod, S. Risk of tumor cell seeding through biopsy and aspiration cytology. J. Int. Soc. Prev. Community Dent. 4, 5–11 (2014).

82. Tolmachev, V., Stone-Elander, S. & Orlova, A. Radiolabelled receptor-tyrosine-kinase targeting drugs for patient stratification and monitoring of therapy response: prospects and pitfalls. Lancet Oncol. 11, 992–1000 (2010).

83. Stigbrand, T., Carlsson, J. & Adams, G. P. Targeted radionuclide tumor therapy : biological aspects. (Springer, 2008).

84. Vegt, E. et al. Renal uptake of different radiolabelled peptides is mediated by megalin: SPECT and biodistribution studies in megalin-deficient mice. Eur. J. Nucl. Med. Mol. Imaging 38, 623–32 (2011).

85. Maeda, H. The enhanced permeability and retention (EPR) effect in tumor vasculature: the key role of tumor-selective macromolecular drug targeting. Adv. Enzyme Regul. 41, 189–207 (2001).

86. Lub-de Hooge, M. N. et al. Preclinical characterisation of 111 In-DTPA-trastuzumab. Br. J. Pharmacol. 143, 99–106 (2004).

87. Zhang, X. et al. 18F-labeled bombesin analogs for targeting GRP receptor-expressing prostate cancer. J. Nucl. Med. 47, 492–501 (2006).

88. Chen, X. et al. 18F-labeled RGD peptide: initial evaluation for imaging brain tumor angiogenesis. Nucl. Med. Biol. 31, 179–189 (2004).

89. Lindberg, H. et al. Evaluation of a HER2-targeting affibody molecule combining an N-terminal HEHEHE-tag with a GGGC chelator for 99mTc-labelling at the C terminus. Tumour Biol. 33, 641–51 (2012).

90. Wållberg, H. & Ståhl, S. Design and evaluation of radiolabeled tracers for tumor imaging. Biotechnol. Appl. Biochem. 60, 365–383 (2013).

91. Ståhl, S. et al. Affibody Molecules in Biotechnological and Medical Applications. Trends Biotechnol. xx, 1–22 (2017).

92. Baum, R. P. et al. Molecular Imaging of HER2-Expressing Malignant Tumors in Breast Cancer Patients Using Synthetic 111In- or 68Ga-Labeled Affibody Molecules. J. Nucl. Med. 51, 892–897 (2010).

93. Sandström, M. et al. Biodistribution and Radiation Dosimetry of the Anti-HER2 Affibody Molecule 68Ga-ABY-025 in Breast Cancer Patients. J. Nucl. Med. 57, 867–71 (2016).

70

94. Sörensen, J. et al. Measuring HER2-Receptor Expression In Metastatic Breast Cancer Using [ 68 Ga]ABY-025 Affibody PET/CT. Theranostics 6, 262–271 (2016).

95. de Souza, A. L. R. et al. Fluorescent Affibody Molecule Administered In Vivo at a Microdose Level Labels EGFR Expressing Glioma Tumor Regions. Mol. Imaging Biol. 19, 41–48 (2017).

96. Vidarsson, G., Dekkers, G. & Rispens, T. IgG subclasses and allotypes: From structure to effector functions. Front. Immunol. 5, (2014).

97. Reich, N. C. Janeway’s Immunobiology . Seventh Edition. By Kenneth Murphy, Paul Travers, and , Mark Walport; contributions by, Michael Ehrenstein, Claudia Mauri, Allan Mowat, and , Andrey Shaw. Garland Science . New York: Taylor &amp. The Quarterly Review of Biology 83, (2008).

98. Leenaars, M. & Hendriksen, C. F. M. Critical Steps in the Production of Polyclonal and Monoclonal Antibodies: Evaluation and Recommendations. ILAR J. 46, 269–279 (2005).

99. Howard, G. & Kaser, M. Making and using antibodies: A practical handbook. Lavoisier.Fr (Taylor & Francis/CRC Press, 2007).

100. Busby, M. et al. Systematic comparison of monoclonal versus polyclonal antibodies for mapping histone modifications by ChIP-seq. Epigenetics Chromatin 9, 49 (2016).

101. KÖHLER, G. & MILSTEIN, C. Continuous cultures of fused cells secreting antibody of predefined specificity. Nature 256, 495–497 (1975).

102. Sgro, C. Side-effects of a monoclonal antibody, muromonab CD3/orthoclone OKT3: bibliographic review. Toxicology 105, 23–29 (1995).

103. Li, F., Vijayasankaran, N., Shen, A., Kiss, R. & Amanullah, A. Cell culture processes for monoclonal antibody production. MAbs 2, 466–479 (2010).

104. Winter, G., Griffiths, A. D., Hawkins, R. E. & Hoogenboom, H. R. Making Antibodies by Phage Display Technology. Annu. Rev. Immunol. 12, 433–455 (1994).

105. McCafferty, J., Griffiths, A. D., Winter, G. & Chiswell, D. J. Phage antibodies: filamentous phage displaying antibody variable domains. Nature 348, 552–554 (1990).

106. Marks, J. D. et al. By-passing immunization. Human antibodies from V-gene libraries displayed on phage. J. Mol. Biol. 222, 581–597 (1991).

107. Steinberger, P., Sutton, J. K., Rader, C., Elia, M. & Barbas, C. F. Generation and characterization of a recombinant human CCR5-specific antibody: A phage display approach for rabbit antibody humanization. J. Biol. Chem. 275, 36073–36078 (2000).

108. Baca, M., Presta, L. G., O’Connor, S. J. & Wells, J. A. Antibody humanization using monovalent phage display. J. Biol. Chem. 272, 10678–10684 (1997).

109. Lo, B. K. C. in Methods in molecular biology (Clifton, N.J.) 248, 135–59 (Humana Press, 2004).

110. Chaiet, L. & Wolf, F. J. The properties of streptavidin, a biotin-binding protein produced by Streptomycetes. Arch. Biochem. Biophys. 106, 1–5 (1964).

111. Cochran, J. R., Kim, Y. S., Lippow, S. M., Rao, B. & Wittrup, K. D. Improved mutants from directed evolution are biased to orthologous substitutions. Protein Eng. Des. Sel. 19, 245–253 (2006).

112. Azoitei, M. L. et al. Computational Design of High-Affinity Epitope Scaffolds by Backbone Grafting of a Linear Epitope. J. Mol. Biol. 415, 175–192 (2012).

113. Nygren, P.-Å. & Uhlén, M. Scaffolds for engineering novel binding sites in proteins. Curr. Opin. Struct. Biol. 7, 463–469 (1997).

114. Nygren, P. Å. & Skerra, A. Binding proteins from alternative scaffolds. J. Immunol. Methods 290, 3–28 (2004).

115. Skerra, A. Alternative non-antibody scaffolds for molecular recognition. Curr. Opin. Biotechnol. 18, 295–304 (2007).

116. Zielonka, S. et al. The Shark Strikes Twice: Hypervariable Loop 2 of Shark IgNAR Antibody

71

Ken G Andersson

Variable Domains and Its Potential to Function as an Autonomous Paratope. Mar. Biotechnol. 17, 386–392 (2015).

117. Heinis, C., Rutherford, T., Freund, S. & Winter, G. Phage-encoded combinatorial chemical libraries based on bicyclic peptides. Nat. Chem. Biol. 5, 502–507 (2009).

118. Kim, H. J., Eichinger, A. & Skerra, A. High-affinity recognition of lanthanide(iii) chelate complexes by a reprogrammed human lipocalin 2. J. Am. Chem. Soc. 131, 3565–3576 (2009).

119. Vazquez-Lombardi, R. et al. Challenges and opportunities for non-antibody scaffold drugs. Drug Discov. Today 20, 1271–1283 (2015).

120. Nilvebrant, J., Åstrand, M., Löfblom, J. & Hober, S. Development and characterization of small bispecific albumin-binding domains with high affinity for ErbB3. Cell. Mol. Life Sci. 70, 3973–3985 (2013).

121. Nilvebrant, J. & Hober, S. the Albumin-Binding Domain As a Scaffold for Protein Engineering. Comput. Struct. Biotechnol. J. 6, 1–8 (2013).

122. Kintzing, J. R. & Cochran, J. R. Engineered knottin peptides as diagnostics, therapeutics, and drug delivery vehicles. Curr. Opin. Chem. Biol. 34, 143–150 (2016).

123. Glotzbach, B. et al. Combinatorial Optimization of Cystine-Knot Peptides towards High-Affinity Inhibitors of Human Matriptase-1. PLoS One 8, e76956 (2013).

124. Plückthun, A. Designed Ankyrin Repeat Proteins (DARPins): Binding Proteins for Research, Diagnostics, and Therapy. Annu. Rev. Pharmacol. Toxicol. 55, 489–511 (2015).

125. Binz, H. K. et al. High-affinity binders selected from designed ankyrin repeat protein libraries. Nat. Biotechnol. 22, 575–582 (2004).

126. Sachdev, E., Gong, J., Rimel, B. & Mita, M. Adnectin-Targeted Inhibitors: Rationale and Results. Curr. Oncol. Rep. 17, 35 (2015).

127. Diem, M. D. et al. Selection of high-affinity Centyrin FN3 domains from a simple library diversified at a combination of strand and loop positions. Protein Eng. Des. Sel. 27, 419–429 (2014).

128. Sha, F. et al. Dissection of the BCR-ABL signaling network using highly specific monobody inhibitors to the SHP2 SH2 domains. Proc. Natl. Acad. Sci. 110, 14924–14929 (2013).

129. Oganesyan, V. et al. Fibronectin type III domains engineered to bind CD40L: cloning, expression, purification, crystallization and preliminary X-ray diffraction analysis of two complexes. Acta Crystallogr. Sect. F Struct. Biol. Cryst. Commun. 69, 1045–1048 (2013).

130. Škrlec, K., Štrukelj, B. & Berlec, A. Non-immunoglobulin scaffolds: a focus on their targets. Trends Biotechnol. 33, 408–418 (2015).

131. Koide, A., Bailey, C. W., Huang, X. & Koide, S. The fibronectin type III domain as a scaffold for novel binding proteins. J. Mol. Biol. 284, 1141–1151 (1998).

132. Koide, A., Wojcik, J., Gilbreth, R. N., Hoey, R. J. & Koide, S. Teaching an old scaffold new tricks: Monobodies constructed using alternative surfaces of the FN3 scaffold. J. Mol. Biol. 415, 393–405 (2012).

133. Owens, B. Faster, deeper, smaller—the rise of antibody-like scaffolds. Nat. Biotechnol. 35, 602–603 (2017).

134. Orlova, A. et al. Synthetic Affibody Molecules: A Novel Class of Affinity Ligands for Molecular Imaging of HER2-Expressing Malignant Tumors. Cancer Res. 67, (2007).

135. Nord Karin, Gunneriusson E, Ringdahl J, Ståhl S, Uhlén M, N. P. Binding proteins selected from combinatorial libraries of an alpha- helical bacterial receptor domain. Nat. Biotechnol. 15, 772–777 (1997).

136. Arora, P., Oas, T. G. & Myers, J. K. Fast and faster: a designed variant of the B-domain of protein A folds in 3 microsec. Protein Sci. 13, 847–53 (2004).

137. Tashiro, M. et al. High-resolution solution NMR structure of the Z domain of staphylococcal protein A. J. Mol. Biol. 272, 573–590 (1997).

72

138. Ståhl, S. et al. Affibody Molecules in Biotechnological and Medical Applications. Trends Biotechnol. (2017). doi:10.1016/j.tibtech.2017.04.007

139. Friedman, M. et al. Engineering and characterization of a bispecific HER2 × EGFR-binding affibody molecule. Biotechnol. Appl. Biochem. 54, 121–131 (2009).

140. Rönnmark, J. et al. Construction and characterization of affibody-Fc chimeras produced in Escherichia coli. J. Immunol. Methods 261, 199–211 (2002).

141. Malm, M. et al. Engineering of a bispecific affibody molecule towards HER2 and HER3 by addition of an albumin-binding domain allows for affinity purification and in vivo half-life extension. Biotechnol. J. 9, 1215–1222 (2014).

142. Ahlgren, S. et al. Evaluation of Maleimide Derivative of DOTA for Site-Specific Labeling of Recombinant Affibody Molecules. Bioconjug. Chem. 19, 235–243 (2008).

143. Engfeldt, T., Renberg, B., Brumer, H., Nygren, Per åke & Eriksson Karlström, A. Chemical Synthesis of Triple-Labelled Three-Helix Bundle Binding Proteins for Specific Fluorescent Detection of Unlabelled Protein. ChemBioChem 6, 1043–1050 (2005).

144. Ekblad, T. et al. Synthesis and chemoselective intramolecular crosslinking of a HER2-binding affibody. Biopolymers 92, 116–123 (2009).

145. Webster, J. M., Zhang, R., Gambhir, S. S., Cheng, Z. & Syud, F. A. Engineered Two-Helix Small Proteins for Molecular Recognition. ChemBioChem 10, 1293–1296 (2009).

146. Nord, K. et al. Binding proteins selected from combinatorial libraries of an alpha-helical bacterial receptor domain. Nat. Biotechnol. 15, 772–7 (1997).

147. Wikman, M. et al. Selection and characterization of HER2/neu-binding affibody ligands. Protein Eng. Des. Sel. 17, 455–462 (2004).

148. Kronqvist, N. et al. Staphylococcal surface display in combinatorial protein engineering and epitope mapping of antibodies. Recent Pat. Biotechnol. 4, 171–82 (2010).

149. Friedman, M. et al. Phage display selection of Affibody molecules with specific binding to the extracellular domain of the epidermal growth factor receptor. Protein Eng. Des. Sel. 20, 189–199 (2007).

150. Orlova, A. et al. Tumor Imaging Using a Picomolar Affinity HER2 Binding Affibody Molecule. Cancer Res. 66, (2006).

151. AffibodyAB. Pipeline for drug development. Available at: http://www.affibody.se/technology/pipeline/. (Accessed: 26th August 2017)

152. Rudikoff, S., Giustit, A. M., Cookt, W. D. & Scharfft, M. D. Single amino acid substitution altering antigen-binding specificity. Immunology 79, 1979–1983 (1982).

153. R R Franke, T P Sakmar, D. D. O. and H. G. K. A single amino acid substitution in rhodopsin (lysine 248----leucine) prevents activation of transducin. Journal of Biological Chemistry 187, (American Society for Biochemistry and Molecular Biology, 1950).

154. Parker, J. L. & Newstead, S. Membrane Protein Crystallisation: Current Trends and Future Perspectives. Adv. Exp. Med. Biol. 922, 61–72 (2016).

155. Tiwari, M. K., Singh, R., Singh, R. K., Kim, I.-W. & Lee, J.-K. Computational approaches for rational design of proteins with novel functionalities. Comput. Struct. Biotechnol. J. 2, e201209002 (2012).

156. Rohl, C. A., Strauss, C. E. M., Misura, K. M. S. & Baker, D. Protein structure prediction using Rosetta. Methods Enzymol. 383, 66–93 (2004).

157. Strauch, E.-M., Fleishman, S. J. & Baker, D. Computational design of a pH-sensitive IgG binding protein. Proc. Natl. Acad. Sci. U. S. A. 111, 675–80 (2014).

158. Fleishman, S. J. et al. Computational Design of Proteins Targeting the Conserved Stem Region of Influenza Hemagglutinin. Science (80-. ). 332, (2011).

159. Packer, M. S. & Liu, D. R. Methods for the directed evolution of proteins. Nat. Rev. Genet. 16, 379–394 (2015).

73

Ken G Andersson

160. Romero, P. A. & Arnold, F. H. Exploring protein fitness landscapes by directed evolution. Nat. Rev. Mol. Cell Biol. 10, 866–876 (2009).

161. Leung, D., Chen, E., Goeddel, D., Leung, D. W. & Goeddel, D. V. A method for random mutagenesis of a defined DNA segment using a modified polymerase chain reaction. (1989).

162. Eckert, K. A. & Kunkel, T. A. High fidelity DNA synthesis by the Thermus aquaticus DNA polymerase. Nucleic Acids Res. 18, 3739–44 (1990).

163. Zaccolo, M., Williams, D. M., Brown, D. M. & Gherardi, E. An Approach to Random Mutagenesis of DNA Using Mixtures of Triphosphate Derivatives of Nucleoside Analogues. J. Mol. Biol. 255, 589–603 (1996).

164. Pritchard, L., Corne, D., Kell, D., Rowland, J. & Winson, M. A general model of error-prone PCR. J. Theor. Biol. 234, 497–509 (2005).

165. Nov, Y. When second best is good enough: another probabilistic look at saturation mutagenesis. Appl. Environ. Microbiol. 78, 258–62 (2012).

166. Virnekäs, B. et al. Trinucleotide phosphoramidites: ideal reagents for the synthesis of mixed oligonucleotides for random mutagenesis. Nucleic Acids Res. 22, 5600–7 (1994).

167. González-Muñoz, A. et al. Tailored amino acid diversity for the evolution of antibody affinity. MAbs 4, 664–72 (2012).

168. Fellouse, F. A., Wiesmann, C. & Sidhu, S. S. Synthetic antibodies from a four-amino-acid code: A dominant role for tyrosine in antigen recognition. Proc. Natl. Acad. Sci. 101, 12467–12472 (2004).

169. Eaton, B. E., Gold, L. & Zichi, D. A. Let’s get specific: the relationship between specificity and affinity. Chem. Biol. 2, 633–638 (1995).

170. Northrup, S. H., And, T. & Ericksont, H. P. Kinetics of protein-protein association explained by Brownian dynamics computer simulation. Biochemistry 89, 3338–3342 (1992).

171. Hawkins, R. E., Russell, S. J. & Winter, G. Selection of phage antibodies by binding affinity. Mimicking affinity maturation. J. Mol. Biol. 226, 889–96 (1992).

172. Gustavsson, M. et al. Improved cell surface display of Salmonella enterica serovar Enteritidis antigens in Escherichia coli. Microb. Cell Fact. 14, 47 (2015).

173. Zahnd, C., Sarkar, C. A. & Plückthun, A. Computational analysis of off-rate selection experiments to optimize affinity maturation by directed evolution. Protein Eng. Des. Sel. 23, 175–184 (2010).

174. Roberts, R. W. Totally in vitro protein selection using mRNA-protein fusions and ribosome display. Curr. Opin. Chem. Biol. 3, 268–273 (1999).

175. Wang, H. & Liu, R. Advantages of mRNA display selections over other selection techniques for investigation of protein–protein interactions. Expert Rev. Proteomics 8, 335–346 (2011).

176. Zahnd, C., Amstutz, P. & Plückthun, A. Ribosome display: selecting and evolving proteins in vitro that specifically bind to a target. Nat. Methods 4, 269–279 (2007).

177. Odegrip, R. et al. CIS display: In vitro selection of peptides from libraries of protein-DNA complexes. Proc. Natl. Acad. Sci. U. S. A. 101, 2806–10 (2004).

178. Ghadessy, F. J., Ong, J. L. & Holliger, P. Directed evolution of polymerase function by compartmentalized self-replication. Proc. Natl. Acad. Sci. U. S. A. 98, 4552–7 (2001).

179. Smith, G. P. Filamentous fusion phage: novel expression vectors that display cloned antigens on the virion surface. Science 228, 1315–7 (1985).

180. Marvin, D. A., Hale, R. D., Nave, C. & Helmer-Citterich, M. Molecular models and structural comparisons of native and mutant class I filamentous bacteriophages Ff (fd, f1, M13), If1 and IKe. J. Mol. Biol. 235, 260–86 (1994).

181. Wölcke, J. & Weinhold, E. a Dna-Binding Peptide From a Phage Display Library. Nucleosides, Nucleotides and Nucleic Acids 20, 1239–1241 (2001).

182. Kronqvist, N. et al. Combining phage and staphylococcal surface display for generation of

74

ErbB3-specific. Protein Eng. Des. Sel. 24, 385–396 (2011).183. Lowman, H. B., Bass, S. H., Simpson, N. & Wells, J. A. Selecting high-affinity binding

proteins by monovalent phage display. Biochemistry 30, 10832–10838 (1991).184. Bazan, J., Całkosiński, I. & Gamian, A. Phage display--a powerful technique for

immunotherapy: 1. Introduction and potential of therapeutic applications. Hum. Vaccin. Immunother. 8, 1817–28 (2012).

185. Barbas, C. F., Kang, A. S., Lerner, R. A., Benkovic, S. J. & Benkovic, S. J. Assembly of combinatorial antibody libraries on phage surfaces: the gene III site. Proc. Natl. Acad. Sci. U. S. A. 88, 7978–82 (1991).

186. Smith, G. P. & Petrenko, V. A. Phage Display. Chem. Rev. 97, 391–410 (1997).187. Bloom, J. D. & Arnold, F. H. In the light of directed evolution: pathways of adaptive protein

evolution. Proc. Natl. Acad. Sci. U. S. A. 106 Suppl 1, 9995–10000 (2009).188. Hubbard, B. P. et al. Continuous directed evolution of DNA-binding proteins to improve

TALEN specificity. Nat. Methods 12, 939–42 (2015).189. Löfblom, J., Wernérus, H. & Ståhl, S. Fine affinity discrimination by normalized fluorescence

activated cell sorting in staphylococcal surface display. FEMS Microbiol. Lett. 248, 189–198 (2005).

190. Garcia-Rodriguez, C. et al. Molecular evolution of antibody cross-reactivity for two subtypes of type A botulinum neurotoxin. Nat. Biotechnol. 25, 107–116 (2007).

191. Löfblom, J. Bacterial display in combinatorial protein engineering. Biotechnol. J. 6, 1115–29 (2011).

192. Feldhaus, M. J. et al. Flow-cytometric isolation of human antibodies from a nonimmune Saccharomyces cerevisiae surface display library. (2003). doi:10.1038/nbt785

193. Boder, E. T. & Wittrup, K. D. Yeast surface display for screening combinatorial polypeptide libraries. Nat. Biotechnol. 15, 553–557 (1997).

194. Daugherty, P. S., Chen, G., Olsen, M. J., Iverson, B. L. & Georgiou, G. Antibody affinity maturation using bacterial surface display. Protein Eng. 11, 825–32 (1998).

195. Ho, M. & Pastan, I. in Methods in molecular biology (Clifton, N.J.) 525, 337–352 (2009).196. Roy, A., Lu, C. F., Marykwas, D. L., Lipke, P. N. & Kurjan, J. The AGA1 product is

involved in cell surface attachment of the Saccharomyces cerevisiae cell adhesion glycoprotein a-agglutinin. Mol. Cell. Biol. 11, 4196–206 (1991).

197. Lu, C. F. et al. Glycosyl phosphatidylinositol-dependent cross-linking of alpha-agglutinin and beta 1,6-glucan in the Saccharomyces cerevisiae cell wall. J. Cell Biol. 128, 333–40 (1995).

198. Benders, G. A. et al. Cloning whole bacterial genomes in yeast. Nucleic Acids Res. 38, 2558–2569 (2010).

199. Gietz, R. D., Schiestl, R. H., Willems, A. R. & Woods, R. A. Studies on the transformation of intact yeast cells by the LiAc/SS-DNA/PEG procedure. Yeast 11, 355–360 (1995).

200. Griffiths, A. D. et al. Isolation of high affinity human antibodies directly from large synthetic repertoires. EMBO J. 13, 3245–60 (1994).

201. Benatuil, L., Perez, J. M., Belk, J. & Hsieh, C.-M. An improved yeast transformation method for the generation of very large human antibody libraries. Protein Eng. Des. Sel. 23, 155–159 (2010).

202. Doerner, A., Rhiel, L., Zielonka, S. & Kolmar, H. Therapeutic antibody engineering by high efficiency cell screening. FEBS Lett. 588, 278–287 (2014).

203. Xu, L. et al. Rapid optimization and prototyping for therapeutic antibody-like molecules. MAbs 5, 237–254 (2013).

204. Wildt, S. & Gerngross, T. U. The humanization of N-glycosylation pathways in yeast. Nat. Rev. Microbiol. (2005). doi:10.1038/nrmicro1087

205. Shields, R. L. et al. Lack of Fucose on Human IgG1 N -Linked Oligosaccharide Improves

75

Ken G Andersson

Binding to Human FcαRIII and Antibody-dependent Cellular Toxicity. J. Biol. Chem. 277, 26733–26740 (2002).

206. Uray, K., Mizuno, M., Inazu, T., Goto, K. & Hudecz, F. The effect of glycosylation on the antibody recognition of a MUC2 mucin epitope. Biopolymers 102, 390–395 (2014).

207. Gerngross, T. U. Advances in the production of human therapeutic proteins in yeasts and filamentous fungi. Nat. Biotechnol. 22, 1409–1414 (2004).

208. Gai, S. A. & Wittrup, K. D. Yeast surface display for protein engineering and characterization. Curr. Opin. Struct. Biol. 17, 467–73 (2007).

209. Bowley, D. R., Labrijn, A. F., Zwick, M. B. & Burton, D. R. Antigen selection from an HIV-1 immune antibody library displayed on yeast yields many novel antibodies compared to selection from the same library displayed on phage. Protein Eng. Des. Sel. 20, 81–90 (2007).

210. Dorr, B. M., Ham, H. O., An, C., Chaikof, E. L. & Liu, D. R. Reprogramming the specificity of sortase enzymes. Proc. Natl. Acad. Sci. U. S. A. 111, 13343–8 (2014).

211. Rhiel, L. et al. REAL-Select: Full-Length Antibody Display and Library Screening by Surface Capture on Yeast Cells. PLoS One 9, e114887 (2014).

212. Shibasaki, S. et al. An oral vaccine against candidiasis generated by a yeast molecular display system. Pathog. Dis. 69, 262–268 (2013).

213. Ren, R. et al. Display of adenoregulin with a novel Pichia pastoris cell surface display system. Mol. Biotechnol. 35, 103–8 (2007).

214. Brown, S. Engineered iron oxide-adhesion mutants of the Escherichia coli phage lambda receptor. Proc. Natl. Acad. Sci. U. S. A. 89, 8651–5 (1992).

215. Charbit, A., Molla, A., Saurin, W. & Hofnung, M. Versatility of a vector for expressing foreign polypeptides at the surface of gram-negative bacteria. Gene 70, 181–9 (1988).

216. Francisco, J. A., Campbell, R., Iverson, B. L. & Georgiou, G. Production and fluorescence-activated cell sorting of Escherichia coli expressing a functional antibody fragment on the external surface. Proc. Natl. Acad. Sci. U. S. A. 90, 10444–8 (1993).

217. Mazor, Y., Blarcom, T. Van, Mabry, R., Iverson, B. L. & Georgiou, G. Isolation of engineered, full-length antibodies from libraries expressed in Escherichia coli. Nat. Biotechnol. 25, 563–565 (2007).

218. Lombana, T. N., Dillon, M., Bevers Iii, J. & Spiess, C. Optimizing antibody expression by using the naturally occurring framework diversity in a live bacterial antibody display system. Sci. Rep. 5, 17488 (2015).

219. Salema, V. et al. Selection of Single Domain Antibodies from Immune Libraries Displayed on the Surface of E. coli Cells with Two β-Domains of Opposite Topologies. PLoS One 8, e75126 (2013).

220. de Keyzer, J., van der Does, C. & Driessen, A. J. M. The bacterial translocase: a dynamic protein channel complex. Cell. Mol. Life Sci. 60, 2034–2052 (2003).

221. Rose, R. W., Brüser, T., Kissinger, J. C. & Pohlschröder, M. Adaptation of protein secretion to extremely high-salt conditions by extensive use of the twin-arginine translocation pathway. Mol. Microbiol. 45, 943–50 (2002).

222. Green, E. R. & Mecsas, J. Bacterial Secretion Systems: An Overview. Microbiol. Spectr. 4, (2016).

223. Leo, J. C., Grin, I. & Linke, D. Type V secretion: mechanism(s) of autotransport through the bacterial outer membrane. Philos. Trans. R. Soc. Lond. B. Biol. Sci. 367, 1088–101 (2012).

224. Fan, E., Chauhan, N., Udatha, D. B. R. K. G., Leo, J. C. & Linke, D. in Virulence Mechanisms of Bacterial Pathogens, Fifth Edition 4, 305–335 (American Society of Microbiology, 2016).

225. Pohlner, J., Halter, R., Beyreuther, K. & Meyer, T. F. Gene structure and extracellular secretion of Neisseria gonorrhoeae IgA protease. Nature 325, 458–62

226. Otto, B. R. et al. Crystal Structure of Hemoglobin Protease, a Heme Binding Autotransporter

76

Protein from Pathogenic Escherichia coli. J. Biol. Chem. 280, 17339–17345 (2005).227. Leo, J. C., Oberhettinger, P., Schütz, M. & Linke, D. The inverse autotransporter family:

Intimin, invasin and related proteins. Int. J. Med. Microbiol. 305, 276–282 (2015).228. Binder, U., Matschiner, G., Theobald, I. & Skerra, A. High-throughput sorting of an Anticalin

library via EspP-mediated functional display on the Escherichia coli cell surface. J. Mol. Biol. 400, 783–802 (2010).

229. Krause, S. et al. Bacteria Displaying Interleukin-4 Mutants Stimulate Mammalian Cells and Reflect the Biological Activities of Variant Soluble Cytokines. ChemBioChem 5, 804–810 (2004).

230. Adams, T. M., Wentzel, A. & Kolmar, H. Intimin-mediated export of passenger proteins requires maintenance of a translocation-competent conformation. J. Bacteriol. 187, 522–33 (2005).

231. Wentzel, A., Christmann, A., Adams, T. & Kolmar, H. Display of passenger proteins on the surface of Escherichia coli K-12 by the enterohemorrhagic E. coli intimin EaeA. J. Bacteriol. 183, 7273–84 (2001).

232. Christmann, A., Wentzel, A., Meyer, C., Meyers, G. & Kolmar, H. Epitope mapping and affinity purification of monospecific antibodies by Escherichia coli cell surface display of gene-derived random peptide libraries. J. Immunol. Methods 257, 163–173 (2001).

233. Jose, J. & Meyer, T. F. The autodisplay story, from discovery to biotechnical and biomedical applications. Microbiol. Mol. Biol. Rev. 71, 600–19 (2007).

234. Park, M. et al. Autodisplay of streptavidin. Enzyme Microb. Technol. 48, 307–311 (2011).235. Veiga, E., De Lorenzo, V. & Fernández, L. A. Structural tolerance of bacterial autotransporters

for folded passenger protein domains. Mol. Microbiol. 52, 1069–1080 (2004).236. Gustavsson, M., Hörnström, D., Lundh, S., Belotserkovsky, J. & Larsson, G. Biocatalysis on

the surface of Escherichia coli: melanin pigmentation of the cell exterior. Sci. Rep. 6, 36117 (2016).

237. Jose, J. & von Schwichow, S. ‘Cystope tagging’ for labeling and detection of recombinant protein expression. Anal. Biochem. 331, 267–274 (2004).

238. Boder, E. T., Midelfort, K. S. & Wittrup, K. D. Directed evolution of antibody fragments with monovalent femtomolar antigen-binding affinity. Proc. Natl. Acad. Sci. U. S. A. 97, 10701–5 (2000).

239. Bradbury, A. R. . & Marks, J. D. Antibodies from phage antibody libraries. J. Immunol. Methods 290, 29–49 (2004).

240. Park, M. et al. Autodisplay of streptavidin. Enzyme Microb. Technol. 48, 307–11 (2011).241. Lattemann, C. T., Maurer, J., Gerland, E. & Meyer, T. F. Autodisplay : Functional Display

of Active β -Lactamase on the Surface of Escherichia coli by the AIDA-I Autotransporter Autodisplay : Functional Display of Active α -Lactamase on the Surface of Escherichia coli by the AIDA-I Autotransporter. (2000). doi:10.1128/JB.182.13.3726-3733.2000.Updated

242. Fleetwood, F. et al. Simultaneous targeting of two ligand-binding sites on VEGFR2 using biparatopic Affibody molecules results in dramatically improved affinity. doi:10.1038/srep07518

243. Testa, U. et al. Interleukin-3 receptor in acute leukemia. Leukemia 18, 219–226 (2004).244. Garrett, J. T. et al. Transcriptional and posttranslational up-regulation of HER3 (ErbB3)

compensates for inhibition of the HER2 tyrosine kinase. Proc. Natl. Acad. Sci. 108, 5021–5026 (2011).

245. Maurer, J., Jose, J., Meyer, T. F., Maurer, J. & Jose, J. Autodisplay : one-component system for efficient surface display and release of soluble recombinant proteins from Escherichia coli . Autodisplay : One-Component System for Efficient Surface Display and Release of Soluble Recombinant Proteins from Escheric. J. Bacteriol. 179, 794–804 (1997).

77

Ken G Andersson

246. Gawarzewski, I. et al. Crystal structure of the transport unit of the autotransporter adhesin involved in diffuse adherence from Escherichia coli. J. Struct. Biol. 187, 20–29 (2014).

247. Rice, J. J., Schohn, A., Bessette, P. H., Boulware, K. T. & Daugherty, P. S. Bacterial display using circularly permuted outer membrane protein OmpX yields high affinity peptide ligands. Protein Sci. 15, 825–836 (2006).

248. Fleetwood, F., Andersson, K. G., Ståhl, S. & Löfblom, J. An engineered autotransporter-based surface expression vector enables efficient display of Affibody molecules on OmpT-negative E . coli as well as protease-mediated secretion in OmpT-positive strains An engineered autotransporter-based surface expression . (2014). doi:10.1186/s12934-014-0179-z

249. Van Deventer, J. A., Kelly, R. L., Rajan, S., Wittrup, K. D. & Sidhu, S. S. A switchable yeast display/secretion system. Protein Eng. Des. Sel. 28, 317–325 (2015).

250. Kronqvist, N., Löfblom, J., Severa, D., Ståhl, S. & Wernérus, H. Simplified characterization through site-specific protease-mediated release of affinity proteins selected by staphylococcal display. FEMS Microbiol. Lett. 278, 128–136 (2008).

251. Popp, M. W.-L. & Ploegh, H. L. Making and breaking peptide bonds: protein engineering using sortase. Angew. Chem. Int. Ed. Engl. 50, 5024–5032 (2011).

252. Chen, I., Dorr, B. M. & Liu, D. R. A general strategy for the evolution of bond-forming enzymes using yeast display. Proc. Natl. Acad. Sci. 108, 11399–11404 (2011).

253. Westerlund, K., Honarvar, H., Tolmachev, V. & Eriksson Karlström, A. Design, Preparation, and Characterization of PNA-Based Hybridization Probes for Affibody-Molecule-Mediated Pretargeting. Bioconjug. Chem. 26, 1724–1736 (2015).

254. Massa, S. et al. Sortase A-mediated site-specific labeling of camelid single-domain antibody-fragments: a versatile strategy for multiple molecular imaging modalities. Contrast Media Mol. Imaging 11, 328–339 (2016).

255. Ma, J., Lyu, H., Huang, J. & Liu, B. Targeting of erbB3 receptor to overcome resistance in cancer treatment. (2014). doi:10.1186/1476-4598-13-105

256. Xie, T. et al. Pharmacological targeting of the pseudokinase Her3. Nat. Chem. Biol. 10, 1006–12 (2014).

257. Baselga, J. & Swain, S. M. Novel anticancer targets: revisiting ERBB2 and discovering ERBB3. Nat. Rev. Cancer 9, 463–475 (2009).

258. Kruser, T. J. & Wheeler, D. L. Mechanisms of resistance to HER family targeting antibodies. Exp. Cell Res. 316, 1083–1100 (2010).

259. Robinson, M. K. et al. Targeting ErbB2 and ErbB3 with a bispecific single-chain Fv enhances targeting selectivity and induces a therapeutic effect in vitro. Br. J. Cancer 99, 1415–25 (2008).

260. The Human Protein Atlas database - ERBB3. Available at: http://www.proteinatlas.org/ENSG00000065361-ERBB3/tissue. (Accessed: 31st July 2017)

261. Gupta, G. P. & Massagué, J. Cancer Metastasis: Building a Framework. Cell 127, 679–695 (2006).

262. Malm, M. et al. Inhibiting HER3-Mediated Tumor Cell Growth with Affibody Molecules Engineered to Low Picomolar Affinity by Position-Directed Error-Prone PCR-Like Diversification. PLoS One 8, e62791 (2013).

263. Orlova, A. et al. Imaging of HER3-expressing xenografts in mice using a 99mTc(CO)3-HEHEHE-ZHER3:08699 affibody molecule. Eur. J. Nucl. Med. Mol. Imaging 41, 1450–1459 (2014).

264. Kramer-Marek, G., Kiesewetter, D. O. & Capala, J. Changes in HER2 expression in breast cancer xenografts after therapy can be quantified using PET and (18)F-labeled affibody molecules. J. Nucl. Med. 50, 1131–9 (2009).

265. Tolmachev, V. et al. Optimal specific radioactivity of anti-HER2 Affibody molecules enables

78

discrimination between xenografts with high and low HER2 expression levels. Eur. J. Nucl. Med. Mol. Imaging 38, 531–9 (2011).

266. Mendelsohn, J., Baselga, J., Dukes, M., al., et & al., et. Epidermal growth factor receptor targeting in cancer. Semin. Oncol. 33, 369–85 (2006).

267. Goffin, J. R., Zbuk, K., Thomas, S. M., al., et & Li, G. Epidermal growth factor receptor: pathway, therapies, and pipeline. Clin. Ther. 35, 1282–303 (2013).

268. Tsutsui, S. et al. Prognostic and Predictive Value of Epidermal Growth Factor Receptor in Recurrent Breast Cancer. Clin. Cancer Res. 8, (2002).

269. Ang, K. K., Andratschke, N. H. & Milas, L. Epidermal growth factor receptor and response of head-and-neck carcinoma to therapy. Int. J. Radiat. Oncol. 58, 959–965 (2004).

270. Pirker, R. et al. EGFR expression as a predictor of survival for first-line chemotherapy plus cetuximab in patients with advanced non-small-cell lung cancer: analysis of data from the phase 3 FLEX study. Lancet Oncol. 13, 33–42 (2012).

271. Cappuzzo, F. et al. Epidermal Growth Factor Receptor Gene and Protein and Gefitinib Sensitivity in Non-Small-Cell Lung Cancer. JNCI J. Natl. Cancer Inst. 97, 643–655 (2005).

272. Bradley, J. D. et al. Standard-dose versus high-dose conformal radiotherapy with concurrent and consolidation carboplatin plus paclitaxel with or without cetuximab for patients with stage IIIA or IIIB non-small-cell lung cancer (RTOG 0617): a randomised, two-by-two factorial phase 3 study. Lancet Oncol. 16, 187–199 (2015).

273. Divgi, C. R. et al. Phase I and imaging trial of indium 111-labeled anti-epidermal growth factor receptor monoclonal antibody 225 in patients with squamous cell lung carcinoma. J. Natl. Cancer Inst. 83, 97–104 (1991).

274. Goldenberg, A. et al. Imaging of human tumor xenografts with an indium-111-labeled anti-epidermal growth factor receptor monoclonal antibody. J. Natl. Cancer Inst. 81, 1616–25 (1989).

275. Reilly, R. M. et al. A comparison of EGF and MAb 528 labeled with 111In for imaging human breast cancer. J. Nucl. Med. 41, 903–11 (2000).

276. Cuartero-Plaza, A. et al. Radiolocalization of squamous lung carcinoma with 131I-labeled epidermal growth factor. Clin. Cancer Res. 2, 13–20 (1996).

277. Tolmachev, V. et al. Imaging of EGFR expression in murine xenografts using site-specifically labelled anti-EGFR 111In-DOTA-ZEGFR:2377 Affibody molecule: aspect of the injected tracer amount. Eur. J. Nucl. Med. Mol. Imaging 37, 613–622 (2010).

278. Aerts, H. J. W. L. et al. Disparity Between In Vivo EGFR Expression and 89Zr-Labeled Cetuximab Uptake Assessed with PET. J. Nucl. Med. 50, 123–131 (2008).

279. Schumacher, S. D. & Jose, J. Expression of active human P450 3A4 on the cell surface of Escherichia coli by Autodisplay. J. Biotechnol. 161, 113–120 (2012).

280. Jose, J. & Meyer, T. F. The autodisplay story, from discovery to biotechnical and biomedical applications. Microbiol. Mol. Biol. Rev. 71, 600–19 (2007).

281. Garousi, J. et al. The use of radiocobalt as a label improves imaging of EGFR using DOTA-conjugated Affibody molecule. Sci. Rep. 7, 5961 (2017).

79

Ken G Andersson

AcknowledgementsDet absolut störta tacket går till min handledare John löfblom! Inte nog med att jag är otroligt tacksam för att du riskerade liv och lem för att ta in mig som doktorand, du har verkligen pedagogiskt lärt mig kritiskt tänkande och förbättrat mitt forskade med magnituder. Du har konstant sporrat mig till att förbättra mig vilket jag är ytterst tacksam för! Du har dessutom lärt mig att man ska fuska så mycket som möjligt när staffgruppen har sina lekar :) (oklart om det är en hemlis)

Den kanske bästa bihandledaren man kan ha med stans bästa vaderna är Stefan Ståhl. Inte nog med att du kontinuerligt bidrar med entusiasm och glädje så är du alltid där ifall man knackar på och har frågor. Med en korrekt musiksmak och enligt mig en perfekt syn på livet så har du sedan jag börjat på KTH alltid varit en förebild för mig.

This work would not have been possible without the remarkable expertise possessed by two of the most dedicated scientist I know. Thank you, Anna Orlova and Vladimir Tolmachev for teaching me hands-on in the lab the exciting field of molecular imaging. I am truly honored that you have welcomed me into your research family.

A big thank you to all collaborators and co-authors: Rosestedt M, Oroujeni M, Garousi J, Mitran B, Buijs J, Frejd F, Pichl M, Malm M, Varasteh Z, Selvaraju R, Altai M, Honarvar H, Strand J.

However some co-authors from Uppsala have truly amplified my experiences and deserve additional appreciation: Tack Maria, för allting! Med din kunskap och mitt mod så kan ingen tumör stoppa oss! Thank you Javad showing me that experiment should work on the first try :). Big thanks to Maryam whom has done some incredible research and to Bogdan keep up the dank memeing!

Tack till Vetenskapsrådet VR och Pronova excellence center for protein technology (VINNOVA) för finansiering av projekten och resestipendier för att åka på konferenser

An extra thank you to the eldsjälar who have survived my supervision methodology Natalie, Cornelia, Therese, Xenia, Charles, Christoph, Nanna, Javad you are all great scientist but also, luckily for me, even better friends!

Thank you Affibody AB för att ni var en partner I Pronova projektet och gjorde E.coli display möjligt och tack Pelle för allting kring CD mätningar!

Ett stort tack till alla som är och har varit i staff gruppen: Tarek, Sebastian, Rezan, Charlie, Lisa, Hanna, Magdalenda, Jonas, Andreas, Johan N och ett speciellt tack till Filippa för att du handledde mig och få mig förälskad i E.coli display.

Alla PIs på plan 3 som gör det möjligt att ha en sådan fantastisk stämning! Speciellt tack till Per-Åke som inte bara har låst upp dörrar i labbet på nätterna men även mentala dörrar i ämnen som t.ex. singalpeptider från 90 talet. Hur bra är inte Billy Ocean egentligen?

Tack My för att du granskade min avhandling och kom med värdefulla kommentarer!

80

Ett extra stort tack till Stefan, John och Jennifer som har gjort ett hästjobb med att korrläsa denna avhandling!

Inget snack om saken, hade vi inte haft HPA på plan3 hade min forskning varit betydligt sämre! speciellt tack till: Anne-Sophie S, Siri, Lucas, Henric, Anna, Hanna, Marica, Mehri, Lan Lan, Emma, Melanie och Delaram

Stort tack till Inger som alltid är glad, otroligt hjälpsam och har en jättefin bil!

Det störta sociala bidraget och stöttningen på jobbet kommer från kontoret och där har vi väldigt många som egentligen behöver en större a4 med tack: Tarek, Anna, Daniel, Basia, Helena, Lovisa, Hanna, Marco, Anne-Sophie (super-french), Anders, Ulrika, Nancy och såklart den som kanske är oftast i kontoret Sebastian Wolfmeister

Otroligt roligt och fantastiskt lärorikt har det varit med dessa personer på konferens: Hanna, Lisa, Mikael, Rezan, Jonas, Sebastian, Tarek, Anna-luisa, Björn, Filippa och Andreas.

Riktigt bra forskning görs dock inte utan bicepscurls så väldigt stora tack till: Anna, Martin, Jonas, Siri, Lucas, Aman, Tarek, Nanna, Max, Anne-Sophie (ultra-french), Sebastian, Rezan, Charlie (när han saknar ursäkter), Mikael och Andreas B.

Ett extra stort tack till mina vänner som har hjälpt mig att förgylla min forskningstid på jobbet likväl som min fritid, utan er hade forskning varit tråkigt: Anna-luisa, Magnus, Josefine, Martin, Björn, Tarek, Rezan, Sebastian, Therese, Mikael, Lovisa, Charlie, Anders. För ständigt uppbackning med lägenheten, kylkonsult samt personlig coachning i den snedvridna trädlinjen hoppar pitegrabbarna till undsättning. Det vetenskapliga bidraget är inte att undervärder heller, då E. coli biblioteket är döpt till RLMS efter en av medlemmarna!

Verkligen tack till min älskade familj Mamma, Pappa, Sara, Jim och Morfar som alltid ställer upp och alltid finns där! Ni tillsammans med en till person är det absolut bästa jag har och jag skattar mig så lyckligt lottad att jag har er i mitt liv!

Och slutligen vill jag tacka den personen som betyder mest för mig. Min fästmö, Jennifer, som inte bara delar livet med mig och gör det fullständigt men också utmanar mig och får mig att nå nya mål jag inte trodde var möjliga. Tack för allting du har hjälpt mig med, det betyder så otroligt mycket för mig! Längtar inte till ett, utan två bröllop med dig!