1 Biophysical Enhancement of Protein Therapeutics and Diagnostics

Biophysical Enhancement of Protein Therapeutics and Diagnostics Through Engineered

Linkers

Dissertation

Presented in Partial Fulfillment of the Requirements for the Degree Doctor of Philosophy

in the Graduate School of The Ohio State University

Nicholas Emerson Long

Graduate in The Ohio State University Biochemistry Program

The Ohio State University

2018

Dissertation Committee

Dr. Thomas Magliery, Advisor

Dr. Christopher Jaroniec

Dr. Edward Martin Jr.

Dr. Richard Swenson

Copyrighted by

Nicholas Emerson Long

2018

Abstract

Proteins play a major role in virtually every biological process. Thus, proteins are an ideal platform for the next generation of therapeutics. Over the last few decades, technological and scientific advances in protein production and engineering have led to a new wave of protein-based biologics used in clinical settings. In this body of work, we have engineered both a protein-based cancer diagnostic and an immunotherapeutic.

Antibody-based biologics are becoming one of the most widely approved drug platforms and owe their success to their versatility in binding targets, high stability, and low toxicity.

The anti-TAG-72 cancer-targeting antibody, 3E8, is one such molecule that shows great potential as a diagnostic. We have designed and biophysically characterized a library of

3E8 single chain antibody fragments (scFV) with varying linker composition and length as well as domain orientations. In this library, we have found substantial variation in protein stability, binding affinity, and oligomeric states. Surprisingly, a drastic difference in the oligomeric state of these constructs was seen between conventional IMAC purification and

Protein L purification. Therefore, the literature rules for scFV linker design must be updated to include the dependencies on purification method.

A single antibody construct with optimal biophysical properties (3E8.G4S) was further characterized and subjected to in vivo pharmacokinetic studies. Due to its multimeric composition, 3E8.G4S showed a longer and more favorable clearance time compared to that of a fast clearing scFV. Xenograft mouse imaging and biodistribution studies revealed successful targeting of a colorectal tumor by 3E8.G4S with little accumulation in normal tissues.

To determine the versatility of 3E8-based diagnostics and therapeutic agents, an expansive immunohistochemical analysis of TAG-72 expression was performed in over 1,500 tumors spanning 18 different cancer types. The results of this study showed enhanced staining of engineered 3E8.scFV.FLAG compared to commercially available anti-TAG-72 IgG, B72.3.

We found statistically significant TAG-72 expression differences in colon, prostate, lung, cervical, ovarian, pancreatic, gastrointestinal, rectal, and breast cancers when compared to corresponding normal tissues. The intense staining of these diseased tissues by

3E8.scFV.FLAG suggests wide applicability for its clinical use as a cancer detection agent and consequently the viability of 3E8.G4S as a state of the art cancer imaging agent.

Manipulation of the Notch signaling pathway is another promising therapeutic strategy for a variety of diseases. Specifically, activation of the Notch1 receptor by a clustered DLL1 ligand has been shown to enhance T-cell maturation and promote cancer cell death. This makes the engineering of a multimeric and discrete form of DLL1 a viable cancer therapy.

We have identified and validated the critical binding domain of the DLL1 ligand through

iii in vitro cell based and in vivo Notch inhibition studies. Additionally, the minimal multivalency for Notch activation was determined to be four DLL1 repeats. Through E. coli transgenic expression and protein refolding, we have successfully created a 4x and 6x tandem repeat of the DLL1 ligand capable of in vitro cell activation of the Notch1 pathway.

Through the activation of host immune cells, this novel Notch activator has the potential to be a broadly used cancer immunotherapeutic.

Dedication

This dissertation is dedicated to my Grandfather, Earl Emerson Schneider.

Acknowledgments

I would first like to thank the entire Magliery Lab for my training and making everyday lab life enjoyable. Without such a great lab, this would have been a much more difficult endeavor. Specifically, I want to thank Dr. Bandon Sullivan for monitoring me for my first few years in graduate school and allowing me to be the scientist I am today.

Thank you to my advisor, Dr. Thomas Magliery, for allowing me into his lab and giving me challenging and exciting projects. Without his initial trust in my abilities, none of this work would have been possible.

Thank you to Shubham Mangla, Cynthia Campbell, Callie Moore, and all the undergraduates I had the opportunity to mentor. Their young enthusiasm inspired many graduate students and made science “fun”, despite the seemingly constant frustrations that being a scientist entails.

A special thank you to my undergraduate advisor Dr. Vanessa McCaffrey. With no graduate students at Albion College, it becomes reliant on the advisor to assess if an undergraduate student will be successful in graduate school. Dr. McCaffrey encouraged me to conduct research, write a senior thesis, and apply to graduate school. Here I am, five vi years later, obtaining my PhD. Thank you so much for guiding me into this career path. I owe much of my success to you.

I would like to thank my friends in graduate school and from back home who pushed for me to have a social life despite my busy schedule. I look forward to the many beer fests in the future.

I would like to thank my mother, father, and sister for their continued support throughout graduate school. A lot has changed in five years but your love and encouragement were a constant for which I am truly grateful.

Finally, I would like to thank Kristi for her love, for her advice, for her constant support, and for being there to vent to when experiments did not go as planned. I look forward to the many bright things our future holds.

vii

Vita

2008…………………………………………………….…Rochester Adams High School

2012……………………………….…………….……B.A. Biochemistry, Albion College

2017………………...…Graduate Research Associate, Ohio State Biochemistry Program,

The Ohio State University

Publications

A 3E8.scFV.Cys-IR800 Conjugate Targeting TAG-72 in an Orthotopic Colorectal Cancer Model: Gong, L., Ding, H., Long, N.E. et al. Mol Imaging Biol (2017). https://doi.org/10.1007/s11307-017-1096-4

Fields of Study

Major Field: Biochemistry

viii

Table of Contents

Abstract ...... ii Dedication ...... v Acknowledgments...... vi Vita ...... viii List of Tables ...... xi List of Figures ...... xii Chapter 1. Introduction ...... 1 1.1 Preface: Protein Engineering ...... 1 1.2 The Tumor Microenvironment: A Complex Problem...... 7 1.3 Cancer Immunology and Immunotherapy...... 11 1.4 Tumor Associated Antigens ...... 17 1.5 Cancer Mucins and Tumor-Associated Glycoprotein 72 ...... 21 1.6 Advancements in Antibody Engineering ...... 28 1.7 Anti-TAG72 Antibodies ...... 37 1.8 Single Chain Antibody Fragments ...... 42 1.9 The Notch Signaling Pathway ...... 47 1.10 Notch Implication in Cancer Immunology ...... 51 1.11 Thesis Summary ...... 55 Chapter 2. Engineering linkers in anti-TAG-72 antibody fragments to optimize biophysical properties, serum half-life, and high-specificity tumor imaging ...... 56 2.1 Contributions ...... 56 2.2 Abstract ...... 57 2.3 Introduction ...... 59 2.4 Results ...... 63 2.5 Discussion ...... 81 ix

Chapter 3. Wide-scope Immunohistochemical analysis of TAG-72 expression in solid tumors by 3E8 and B72.3 antibodies ...... 84 3.1 Contributions ...... 84 3.2 Abstract ...... 85 3.3 Introduction ...... 86 3.4 Results ...... 89 3.5 Discussion ...... 102 Chapter 4. Engineering of a multivalent DLL1 Notch activator for cancer immunotherapy ...... 106 4.1 Contributions ...... 106 4.2 Abstract ...... 107 4.3 Introduction ...... 109 4.4 Results ...... 113 4.5 Discussion ...... 130 Chapter 5. Materials and Methodology ...... 134 5.1 Chapter 2 Materials and Methods ...... 134 5.2 Chapter 3 Materials and Methods ...... 142 5.3 Chapter 4 Materials and Methods ...... 148 Bibliography ...... 159

List of Tables

Table 1: List of non-repeat linkers found in the literature...... 46 Table 2: Biophysical characterization summary of 3E8.scFV linker and domain library. 68 Table 3: T-Tests conducted comparing multiple staining data sets...... 95

List of Figures

Figure 1: The size of biologically relevant molecules ...... 2 Figure 2: The reductionist and heterotypic view of a tumor...... 10 Figure 3: The immune system and cancer...... 13 Figure 4: The formation of tumor associated antigens(TAAs)...... 18 Figure 5: Typical forms of protein glycosylation...... 23 Figure 6: Detection of Sialyl-Tn in common cancers ...... 26 Figure 7: Mouse Hybridoma Technology ...... 30 Figure 8: Antibody and Antibody Fragments ...... 32 Figure 9: Schematic of phage display and panning...... 35 Figure 10: Antibody variable domain sequences of CC49, DPK24 and aka...... 40 Figure 11: Antibody domains swapping quaternary structure...... 45 Figure 12: Notch receptors, ligands, and activation mechanism ...... 48 Figure 13: Notch activation in cancer patients...... 53 Figure 14: DLL1 expression mouse tumor study...... 54 Figure 15: Library design and construction...... 64 Figure 16: Biophysical characterization of library constructs by domain orientation...... 65 Figure 17: Binding affinity of 3E8 library members ...... 66 Figure 18: Comparison of Ni-NTA IMAC purification with Protein L purification...... 70 Figure 19: Cloning schematic of tagless linker library...... 72 Figure 20: Tagless gel filtration using prep grade superdex 75 column...... 73 Figure 21: Analytical ultracentrifugation (AUC) of select tagless antibodies ...... 74 Figure 22: Exploration of purification discrepancies ...... 75 Figure 23: Oligomeric state reproducibility...... 76 Figure 24: Biophysical characterization summary of tagless 3E8.G4S...... 77 Figure 25: Separation and characterization of 3E8.G4S oligomeric states...... 78 Figure 26: In vivo studies of 3E8 construct VH-G4S-VL...... 80 Figure 27: Design and Biophysical characterization of 3E8.scFV.FLAG...... 90 Figure 28: Dilution optimization of 3E8.scFV.FLAG for comparative staining to B72.3 91 Figure 29: Selection of secondary antibody for immunohistochemical study...... 92 Figure 30: Analysis of select TAG72(+) tissue cores...... 94 Figure 31: Heat map data of 3E8.scFV and B72.3 staining - normal and diseased tissue. 97 Figure 32: Comparative staining using 3E8.scFV.FLAG and B72.3 of TAG72(+) tissue 98 Figure 33: FDA Slide Analysis of 3E8.scFV.FLAG staining in normal tissue...... 99 rd Figure 34: 3 Party Validation of 3E8.scFV.FLAG and B72.3 ...... 101 Figure 35: Dikov Lab Tumor Growth Study...... 111 xii

Figure 36: SDS-PAGE gels and cartoon of monomers and 2x Tandem constructs...... 114 Figure 37: In Vitro cell assay for validation of DLL1 ligand functionality...... 115 Figure 38: DLL1 Construct #42 in vivo mouse tumor growth study...... 117 Figure 39: Strategies for oligomerization of DLL1 ligand DSL-EGF1-EGF2...... 118 Figure 40: Maleimide modification of DLL1 construct #52 ...... 120 Figure 41: SDS-PAGE gel and graphical cartoon of tandem DLL1 constructs...... 122 Figure 42: In vitro cell based assay results of 4x (#83), 3x (#82), 2x (#55) and 1x (#42) mouse tandem DLL1 constructs...... 123 Figure 43: ELISPOT assay ...... 124 Figure 44: Tandem DLL1 constructs with varying GGGGS repeat linker...... 125 Figure 45: In vitro 4x tandem linker variant results...... 126 Figure 46: Concentration dependency of construct #87...... 127 Figure 47: In vitro analysis of Notch activating constructs...... 129 Figure 48: Design of monomers and 2xTandem constructs...... 148 Figure 49: Gene design of large DLL1 tandem constructs...... 153

xiii

Chapter 1. Introduction

1.1 Preface: Protein Engineering

It is truly remarkable that all varieties of life rely on the same twenty amino acids. And simpler still, these compounds are coded from just four nucleotides; adenine, cytosine, guanine, and thymine. Our understanding of the processes that give us fully formed, functional, and stable proteins from transcription and translation is vast. However, before

I begin the full discussion of proteins, their role in biology and medicine, and their applicability to be engineering into useful research and clinical tools, a consideration of their relative size is helpful. A comparative analysis of relevant biological molecules, organelles, and organisms can be seen in Figure 1.

Figure 1: The size of biologically relevant molecules compared to that of larger, organized forms of life Protein size (10nm) is shown in relation to a single atom and an adult female. Sized are relative to one another on a logarithmic scale. The above figure has been unmodified from its original source and can be accessed at Concepts of Biology. Authored by: Open Stax. Located at: http://cnx.org/contents/[email protected]:1/Concepts_of_Biology. License: Creative Commons

Starting from a single atom, a protein, and its amino acid building block, are composed of carbon, hydrogen, oxygen, nitrogen, and sulfur. The average amino acid is 110 Daltons and the average protein in the human genome is ~500 amino acids or 53 kilodaltons (kDa).

Involved in virtually every biological function in the cell, these proteins can function as enzymes, trafficking molecules, structural support. Unable to be resolved by light microscopes, these molecules must be studied using a variety of other techniques such as

NMR, X-ray crystallography, and electron microscopy. Development of these techniques has paved the way for scientists to study these biological molecules.

One of the first fundamental experiments in protein science, and what arguably gave rise to the field of protein engineering, was performed by Christian Anfinsen in 1954 (1). His exploration of ribonuclease (RNase), an enzyme that catalyzes the degradation of RNA, lead to the conclusion that a proteins structure and function is dependent on the amino acid sequence. The resulting protein structure-function hypothesis is, to this day, used in protein design and engineering.

The consequences of the Anfinsen experiment gave rise to a fundamental question: If the function of a protein is dictated by the amino acid sequence, can we design de novo proteins by changing the amino acid sequence? In short, the answer is no, or more accurately, not yet. We are limited by the complexities each new amino acid adds to a protein structure.

This is best described in 1968 by Cyrus Levinthal who proposed a thought experiment which encapsulates the impossibly many configurations a protein of modest size can have

(2). The Levinthal Paradox is as follows:

1. Given a protein 100 amino acids in length, this protein would have 99 peptide

bonds, each with phi and psi angles, totaling 198 different angles.

2. Assume there exists three different phi/psi angles to which a protein backbone can

adopt.

3. This gives a potential of 3198 different possibilities in which the protein can fold.

4. If all conformations were to be sampled before the proper fold is found, it would

require a longer time than the existence of the universe.

In reality, a protein can usually fold within milliseconds of translation and therefore the process of protein folding by confirmation sampling cannot be true. Instead biophysicists suggest a more intuitive mechanism known as the protein folding funnel, where proteins collapse into their functional fold through a variety of external forces (3).

Still today, the protein folding problem exists and this prevents us from the potential of rapid progress in medicine. We cannot identify a biological need and rationally satisfy it with a novel protein (keeping in mind that nature has accomplished this with time and evolutionary pressures over billions of years). Whether it be an enzyme to remove nerve toxins from the blood, a vaccine that effectively targets HIV, or a peptide that destabilized plaque formation in Alzheimer’s patients, the need for protein therapeutics is easy to imagine – but still we fall short.

However, with the knowledge the protein science field has gained in the past half-century, where much progress solving the protein folding problem and have developed many engineering strategies and tools for predicting the correct protein fold. These strategies can be broken down into three categories: directed evolution, rational design, and semi-rational design (4).

The first strategy involves the use of large libraries in which random mutations have been introduced into a gene of interest. Similar to the natural process of evolution itself, but where mutations are introduced over longer periods of time, the fittest proteins that are generated from the library are selected for improvements over the parent construct(5).

Library size can vary substantially from a small pool of variants to tens of billions. The success of these procedures depends on the probability of the protein to successfully adopt a mutation, or the hit rate. Low hit rates with small libraries may not yield a desired evolved protein, whereas a larger library and an efficient screening method will. On the other hand, large libraries may not be necessary if the experiment were to have a high hit rate (6). This strategy has had many successes including screening for antibody binders, improving thermal stability of proteins, and enhanced enzymatic activity (7-9).

The second tool in modern protein engineering is rational design. Mutations are made, in this case, from previously determined data. These data, for example, can be protein sequences found in nature, computationally assessed mutations, or even previous published mutagenic work (10,11). By necessity, rationally designed proteins have a higher success rate but are produced in much smaller quantities. Success with this method can be found in many forms. For example, analysis of a thermophilic homolog can give insight into mutations that stabilize a mesophilic protein (12). In an ideal case, a simple point mutation may be all that is required to engineer a stabilized protein.

Semi-rational design, arguably the most successful form of protein engineering, is a hybrid between the two previous forms. This process takes the best aspects of each engineering technique and combines them to allow for larger library sized that produces higher hit rates

(13). For instance, computationally determined mutations for stabilizing a protein are calculated using a model and energy functions. This model, although beneficial in identifying potential locations of mutations, may not reflect the actual biophysical properties of the protein. In this case, semi-rational libraries can be made to incorporate the computationally identified mutation as well as other random mutations. Libraries are then screened and produce better results than if only computational (rational) or random strategies were performed (14).

The experimental techniques used throughout this thesis are heavily dependent on the fundamental science of protein engineering. Without the significant body of work produced by structural biologists and protein engineers, translational projects, such as my own, would not be possible.

1.2 The Tumor Microenvironment: A Complex Problem

Decades of cancer research have given scientist new insight into the fundamental question of what cancer really is. It is understood, and this is perhaps what makes cancer seemingly impossible to cure, that this disease does not arise from an external source that can be protected against or even eliminated (e.g. bacterial infections in rats carrying the Bubonic

Plague, sexually transmitted viruses that spread HPV, or the mercury levels in fish causing consumer Erethism (15-17)). Cancer is a disease that arises in the very genes which are fundamental in its victim’s identity. Elimination of cancer by a method akin to other diseases is not an option.

Patient diagnosis of cancer can be complex and problematic. To help classify cancers and decipher their differences from surrounding normal tissue, scientists and physicians have classified six attributes of this disease in which they term the Hallmarks of Cancer. These include: limitless replicative potential, tissue invasion and metastasis, sustained angiogenesis, evading apoptosis, self-sufficiency in growth signals, and insensitivity to anti-growth signals (18,19). It is thought that most cancers, over time, develop all six hallmarks. This may seem like a significant deviation from normal tissue, but the commonalities between the cancerous tissue and their surrounding cells is far greater than their differences. Finding a proper treatment solution or cure for cancer, one that does not affect the normal tissues of the host, is a difficult task which modern medicine has yet to accomplish. 7

The most common types of cancer treatment are far from ideal. They include radiation therapy, chemotherapy, and surgery. In many cases, more than one treatment form is used concurrently.

Radiation therapy involves the use of high energy particles directed at areas of disease.

These particles, which include X-rays, gamma-rays, and other charged particles, damage cellular DNA beyond repair. While this treatment type is directed at the tumor, or in some cases a particle emitting material is placed within the tumor, cellular damage occurs in both diseased and surrounding normal tissue. Prolonged radiation exposure can lead to many side effects: recurring cancers can become less or non-responsive to successive treatments, long term damage and necrosis can occur in surrounding tissues, or new cancers can arise from once healthy tissue whose DNA has been damaged by the radiation (20).

The second form of treatment, chemotherapy, is constantly changing as new drugs become approved and readily available. The basic procedure, however, remains a constant.

Chemotherapy consists of a cocktail of toxins delivered intravenously which attack both diseased and normal tissue at a varying degree. The overall objective of this treatment method is to kill tumor cells while sparing non-diseased cells. Often, chemotherapy comes at the cost of damaging normal tissues. This can lead to severe side effects which may ultimately prevent the patient from continuing treatment (21).

The third form of cancer treatment, surgery, is applicable in cases of solid tumors only.

Cancers of the blood such as leukemia and myelomas have more success when treated with radiation and chemotherapy. Surgeries to remove tumors date back millennia where the one of the first cases of breast cancer removal was recorded in ancient Greece (22,23). In more modern times, surgical procedures were the first form of cancer treatment. The success of surgery is highly dependent on the doctor’s ability to remove the entirety of the tumor. If the cancerous cells are in close proximity to sensitive tissues or vital organs, a surgeon cannot operate. Additionally, the area in which the tumor meets normal tissue, or tumor margin, in many cases is not abundantly clear. Doctors use palpation and knowledge of tumor growth patterns to determine where the cancer and any metastasis may lie. Still, surgery leaves room for error and lengthy recovery times make this treatment option less than ideal.

With so many years of clinical and basic cancer research, why are we still mainly treating cancers with the same, flawed strategies? To answer this question, one must have a good understanding that cancer is heterogeneous in nature. A reductionist’s view of cancer could be a tumor as a group of cells which grow, multiply, and metastasize within surrounding normal tissue (Figure 2, left) (18). Before modern medical advances seen in the last few decades, this was the perception of cancer in which radiation therapy, chemotherapy, and surgeries were designed. In truth, a cancerous tumor is collection of cancer cells, immune cells, fibroblasts, and endothelial cells. More complex still, each subdivision of cell types can be heterogenous (e.g. immune cells can be a complex mixture of T lymphocytes, B

9 lymphocytes, neutrophils, and macrophages) (Figure 2, right). The term for this modern view of cancer has been coined by scientists as the tumor microenvironment.

Figure 2: The reductionist and heterotypic view of a tumor. The classical view of cancer as a group of cells growing amongst normal tissue (left) is an oversimplification of the heterogeneous nature of the tumor micro environment (right). Tumors contains endothelial cell vasculature, fibroblasts, and immune cells in addition to cancer cells (which also are heterogenous in nature) (18).

We can use this knowledge of tumor heterogeneity and the tumor microenvironment to create new alternatives in cancer therapy. The next two sections will discuss cancer immunotherapy and antigen targeted therapy which have already shown promise in the clinic.

1.3 Cancer Immunology and Immunotherapy

The manifestation of cancer does not happen with merely one or even half a dozen mutations. In fact, there are many check points in normal cell growth that prevent atypical cells from developing into cancer (24). Therefore, a tumor must develop a host of mutations that allow for physiologic changes and display the previously mentioned the hallmarks of cancer. One of the most effective ways our bodies naturally defend against cancer is through our own immune systems.

The current hypothesis of immunosurveillance describes a process where circulating immune cells recognize cancerous growth as foreign targets. In the field of cancer immunology, the term “immunoediting” describes a process in which a tumor has one of three outcomes when confronted with the host immune system (25). In the best case, the cancer will be fully eliminated (26). This may happen by immune cells, such as tumor associated macrophages, which arise from activated monocyte white blood cells, infiltrate cancerous tumors and kill neoplastic growth (27). Alternatively, the tumor may have a balanced immune response where the immune system attacks only the most immunogenic cells in a heterogeneous tumor, while other cells go undetected. Finally, the tumor may develop mutations or employ systems to effectively hide from immune detection, called tumor escape.

In tumor escape, cancerous growth can develop several approaches to suppress the immune system. One such approach found in certain colorectal cancers, for example, is the production and secretion of a soluble Fas ligand (FasL/CD95L) (26). The Fas ligand can trigger apoptotic events when bound to its receptor Fas (CD95/APO-1). The secretion of the Fas ligand is initially performed by tumor infiltrating lymphocytes (TILs) as a mechanism for destruction of cancer cells. Interestingly, the TILs also are susceptible to

FasL mediated apoptosis. When cancers develop mutations rendering them insensitive to this pathway, FasL secretion by cancer cells becomes an effective strategy for killing off invading immune cells (28).

Another example of cancer defense involves the production of a mucinous barrier, physically blocking interaction of host immune cells with the growing tumor (29). This process is especially prevalent in adenocarcinomas which can arise from epithelial cell. It is believed that much like normal epithelial cell tissue, production of secreted mucinous glycoprotein creates a protective barrier from the environment. In the case of a solid tumor, dense networks of abnormally glycosylated proteins can make passage of immune cells difficult as well as disrupt various signaling processes. More details on one such glycoprotein, Tumor-Associated Glycoprotein 72 (TAG-72), will be discussed later in this chapter.

Within the tumor microenvironment, there exist many more example of mechanisms which the immune system engages against invasive tumors as well as strategies these tumors use to avoid such engagement. Overall, these strategies can be partitioned into either immunostimulatory and immunosuppressive forces and is highlighted in Figure 3 (26).

Figure 3: The immune system and cancer. The tumor microenvironment interacts with host immune system in bother stimulatory and suppressive manners. Signaling molecules produces by the tumor can affect dendritic cell maturation. Immunotherapies stimulate immune system response, preventing tumor escape. Reproduced with permission from (26), Copyright Massachusetts Medical Society.

In many cases with immunostimulatory response, tumors will display or excrete antigens which will be detected by the host immune system. These antigens can attract dendritic cells which will accumulate in the tumor microenvironment. Upon interaction, these cells can mature and begin producing interleukin-12 which in turn stimulates Th1-type CD4 T- cells (30). These cells produce a soluble cytokine, interferon- γ, which is a critical activator of macrophages and helps increase population of CD8 T-lymphocytes, both of which lead to tumor cell destruction (26). Like many of these established cancer immunostimulitory pathways, new evidence shows in some cases that interferon- γ also plays a role in tumor escape (31).

Known immunosuppression pathways include the production of different sets of tumor antigens which do not have the same dendritic cell maturation effect. Instead, immature dendritic cells produce tumor necrosis factor α (TNF- α) which stimulates the production of ineffective tumor rejection agents, interleukin-4 and -13, by type 2 helper T-cells

(26,32). The immunosuppressed tumor microenvironment does stimulate regulatory T- cells as well as myloid-derived suppressor cells (MDSC). At this time, the tumor has, to some extent, evaded the immune system but is at an equilibrium between rejection and tumor escape.

Many cancers, upon diagnosis where the tumor proliferation occurs but inflammation is present, lie in the balance between rejection and tumor escape. Here, intervention with

14 immunotherapies can counterbalance the evasive mechanisms of the growing tumor. One such method for treating cancers is the use of immune stimulating monoclonal antibodies

(mAbs). Immunotherapeutic mAbs contain a hypervariable region (HV) engineered to bind to a specific targeted antigen and a crystallizable region (FC) which binds to FC receptors located on a variety of immune cells. These antibodies can behave as a molecular bridge, guiding immune cells to cancer cell targets.

The described method can be classified into four different strategies: 1 – mAbs that are designed to bind immune repressing ligands, preventing them from carrying out their function. 2 – mAbs that bind and activate lymphocyte receptor involved in downstream immune activation. 3 – mAbs that bind and inhibit lymphocyte receptors involved in immune inhibition. 4 – mAbs that activate responses in antigen presenting cells, which in turn activate lymphocytes (33).

In addition to mAbs, other forms of treatment include cellular immunotherapy and therapeutic cancer vaccines. The former approach generally involves the harvesting of patient T-cells and ex vivo genetic engineering. T-cells containing a novel receptor engineered to further activate other T-cells through signal transduction machinery are returned back into the patient (34). Increased immune responses have been shown to work in clinical trials, and in some cases produce long term cancer remission (34). Alternatively, cancer vaccines involve the injection of known antigens specific to the patient’s cancer.

This has been practiced in the clinic using HER2 antigen in breast cancer and MUC1 (a cancer associated mucinous glycoprotein) antigen for lung cancers (35,36).

Overabundance of these antigens create a heightened immune sensitivity which is imparted upon the antigen expressing cancer.

Immunotherapy has shown a lot of promise for future therapeutic use. As of now, this form of therapy is most effective when used in cohort with chemotherapy (26). Other obstacles such as high drug prices (fueled by difficulty in production and complex intellectual property issues) and unknown adverse side effects (autoimmune disorders are common but not accurately predictable in mouse models) can limit the widespread use of these therapies

(33). But as clinical research endures, the cost of drug production decreases and development of more reliable immunoassays continues, paving way for a hopeful outlook in tomorrow’s cancer immunotherapy.

1.4 Tumor Associated Antigens

Although tumors are heterogeneous in nature and share near identical genetic make-up to the host, mutations can arise which allow for internal (immune system) and external

(targeted therapy) identification. These genetic differences create alterations in the antigens displayed on the cancer cell membrane. These tumor associated antigens, or TAAs, can typically be classified into two subtypes: antigens with high tumor specificity and antigens with low tumor specificity (37). High specificity antigens are those that are unique to the cancer cell and are not found in normal adjacent tissue. These can arise from mutations in genes that code for surface exposed peptides or protein. Additionally, methylation deregulation can lead to the expression of a membrane protein that is not expressed in normal surrounding tissue (Figure 4A) (37). Low specificity antigens include mutations that change expression levels of membrane proteins. This can cause absence of an antigen as well as overexpression of an antigen (Figure 4B) (37). In this case, immune and therapeutic targeting is difficult because there exists no unique surface marker that is not present on normal tissue.

Figure 4: The formation of tumor associated antigens (TAAs). Tumor antigen expression occur by mutations and/or changes in DNA methylation. This produces two classes of TAAs and can be classified into (A) high tumor specificity and (B) low tumor specificity (37).

Some therapeutics targeting TAAs have already found success in the clinic. It was once thought that targeted therapies, the “magic bullet” approach to treating cancer, would eliminate the need for the harsh treatments of radiation and chemotherapy. Unfortunately, due to cancer’s heterogeneity and innate ability to mutate, many targeted therapies become less effective after first rounds of treatment (38,39). This is due to the tumor evolving new protective responses or eliminating the TAA which is being targeted (38). Therefore, it is common to see this class therapy used in cohort with radiation, chemotherapy, and surgery.

One of the most successful examples of targeted therapy is the anti-human epidermal growth factor receptor 2 (HER2) antibody, Herceptin® or trastuzumab. Trastuzumab is a humanized Immunoglobulin G (IgG) (40). Administered intravenously in HER2+ breast cancer patients, this IgG binds to the growth factor. In doing so, cell signaling which stimulate division is disrupted and the tumor becomes less aggressive. Similar targeted antibody therapeutics are such as Genentech’s Avastin, Xolair, Rituxan, Rapitva, and

Lucentis are being widely used around the country and combined produce over $5 billion in sales (41). With their success, it is easy to imagine a paradigm shift in treatment strategies, and radiation and chemotherapy treatments will be reserved only for extreme, last case scenarios.

In addition to cancer therapies which involve activation or inhibition from cell target binding, another class of molecules is gaining interest in both the research and clinical setting: Antibody Drug Conjugates or ADCs. Because some TAA’s do not elicit any cellular response when bound by the targeted therapy, a drug payload may be conjugated to a targeting agent for delivery. The FDA approved drug brentuximab vedotin or Adcetris

®, is an anti-CD30 monoclonal antibody conjugated to monomethyl auristatin E, MMAE, a synthetic antineoplastic molecule (42). Alone, the high toxicity of MMAE prevents it from being a viable treatment option. However, with the help of a targeting agent, smaller doses, corresponding to lower side effects, can be administered while maintaining high local tumor concentration.

There are many complications and engineering hurdles involved in constructing an ADC.

This is, in part, responsible for their limited success in achieving FDA approval (43-45).

Detail of these engineering challenges will be discussed later in this chapter.

1.5 Cancer Mucins and Tumor-Associated Glycoprotein 72

The presence of highly glycosylated membrane proteins is commonly found in a subtype of cancers called adenocarcinomas. In normal epithelial cell, mucins protect the cell from environmental contact in areas such as the lungs, gastrointestinal tract, and secretory organs

(i.e. liver, pancreas, kidneys, and salivary glands) (46). It is in these tissues, among others, where adenocarcinomas form and can continue to excrete their native glycoproteins as well as abnormal variations. The following section will cover the physiological role of mucins in cancer and highlight one of therapeutic and diagnostic interest, Tumor-Associated

Glycoprotein 72 or TAG-72.

The protein family classified as mucins is known to contain a high propensity for the amino acids proline, serine, and threonine. Named form their amino acid make-up, the PTS domains found in these mucinous are highly modified via O-linked glycosylation of either the serine or threonine residues (29). The mucin family, classified by the nomenclature

MUC1 through MUC21, can be divided into two subgroups – excreted mucins and transmembrane mucins. The appearance of secretory mucins can be seen in early metazoan evolution and contain the mucins MUC2, MUC5, and MUC6. Transmembrane mucins, such as MUC1, MUC13, and MUC16, are believed to be derived the secretory mucin

MUC5B (47). This class of mucins, in addition to being membrane bound, share SEA (sea urchin enterokinase and agrin) domains flanked by two highly O-glycosylated PTS domains.

In general, protein glycosylation is a post-translational enzymatic process facilitated by glycotransferases. With over 200 known glycotransferases in human genome, these enzymes use sugars (e.g., galactose, fucose, and mannose) and proteins (both backbone and sidechains) as substrates to create a complex yet highly regulated form of protein modifications (48). The most common glycoforms found in modified proteins are classified as either N-linked (nitrogen of asparagine or arginine side-chain), O-linked (hydroxyl

22 oxygen of serine, threonine, or tyrosine side-chains), C-linked, and phospho-glycans. A more a detailed schematic of protein glycosylation can be seen in Figure 5.

Figure 5: Typical forms of protein glycosylation.Common types of protein post translational glycosylation include branching, core and terminal fucosylation, sialylation, and O-glycosylation (49).

Protein glycosylation plays an important physiological role in a host of different cellular processes. Once of the main functions of surface protein glycosylation is cell to cell communication and recognition. Once disrupted, this loss in communication can trigger

23 immune responses, not much unlike an immune response to pathogen with unfamiliar glycosylation isoforms (50). Cancers displaying this disrupted phenotype can either be isolated and eliminated by the host immune system, or masked from immune cells by the complexity of signaling molecules in which they are surrounded (46).

One physiological characteristic of cancers is the loss in cell polarity. This is very evident in cases of epithelial cells as their basement membrane and surface exposed mucinous layer are highly regulated. Under stress conditions, such as cases of pre-cancerous neoplastic growth, mucins begin forming nondirectionally. This creates mucinous barriers not only between the outside environment and luminal spaces, but also between normal tissue and the aberrant epithelial growth. Eventually, secreted mucins unable to be recycled or drained by normal processes, begin to pool and form what are known as mucinous lakes.

The expression and directionality of glycoproteins are not the only effects found in mucinous cancers. In addition, differences in glycosylation sites and sugar moieties are also known trademarks of cancer growth. This can be caused by several different physiological changes. For example, fluctuations in the pH of organelles, such as the endoplasmic reticulum and Golgi bodies, can alter glycosylation kinetics. Similarly, shortage or abundance of sugar substrates can alter the relative glycosylation rates of proteins. In addition to sugar substrates, mutations to protein chaperones or mutations in proteins to be glycosylated may decrease the availability of potential glycosylation’s sites.

Finally, dysregulation in glycotransferase expression levels may lead to abnormal glycosylation of proteins (48).

The glycoprotein TAG-72 was first associated with cancer after its partial, three-step, purification in the mid-1980’s. Extracted from homogenized xenograft of a human carcinoma cell line, LS-174T, the large glycoprotein was separated from other proteins and small molecules by size exclusion chromatography (Sepharose CL-4B) (51). The separated fraction was then subjected to two additional affinity columns which yielded a protein which ranged in size from 220 – 400 kDa. This variance in size could be explained by partial sheering which may have led to truncated versions. Still today, the gene which codes for TAG-72 has yet to be identified.

One characteristic of TAG-72 that has been well studied is the glycosylation moieties that are found on this glycoprotein. A sugar of particular interest is Sialyl-Tn. This sialic acid residue is a α2,6-linked glycan found covalently O-linked to Ser and Thr residues. What is unique about Sialyl-Tn is it relatively low abundance in normal tissue (52). The presence of Sialyl-Tn on TAG-72, however, makes it a tumor associated antigen, allowing for a potential target as a cancer therapy or diagnostic.

Many studies implicate the importance of TAG-72 and Sialyl-Tn in various cancers – specifically mucinous solid tumors (i.e. adenocarcinomas). Figure 6 shows the prevalence

25 these tumor markers in an array of common adenocarcinomas including pancreatic, ovarian, and colorectal cancers.

Figure 6: Detection of Sialyl-Tn in common cancers The points on the graph represent individual studies and the percent of patients with tumors expressing Sialyl-Tn. The red line represents the average percentage of those studies. The above figure is unmodified from an open access article distributed under the Creative Commons Attribution License (52).

The presence of Sialyl-Tn in cancer has also been linked to negative prognosis and can be correlated to tumor invasiveness (52). Taking advantage of this unique molecule in cancers, therapeutic vaccines used to heighten patient immune sensitivity were developed. The

Canadian biotech company Biomira fused Sialyl-Tn sugars to keyhole limpet hemocyanin

(KLH – a clinically safe carrier protein known to activate T-cells and initiate antibody production) (53). This fusion created a larger molecule capable of staying in the blood

26 stream longer than a small molecule, which are known to have short hour-long serum half- lives (54). Despite promising research that showed strong immune response, this vaccine therapeutic, named Theratope, failed to show statistically significant patient response in a phase III clinical trial (55). This, in part was, due to lack of patient screening before the vaccine was administered (52).

Despite the failure of Theratope in the clinic, Sialyl-Tn remains a promising cancer target for future diagnostic and therapeutic agents. Perhaps the approach of using Sialyl-Tn as a cancer vaccine was not the correct strategy. Alternatively, using an antibody based approach to target tumors in TAG-72(+) patients is a more viable option. Current antibody based drug research described in the literature and experiments illustrated later in this thesis will strongly argue for the clinical re-assessment of this target.

1.6 Advancements in Antibody Engineering

Antibodies have been used as a vehicle for therapeutic, diagnostic, and research agents for the past five decades. These immune related proteins possess optimal biophysical properties that are not common to other proteins. These properties include their ability to tightly bind to a large diversity of antigens, thermal stability, protease resistance, and non- immunogenicity.

In the mammalian immune system there are five different classes of antibodies: IgA, IgD,

IgE, IgG, and IgM. These immunoglobulins are Y-shaped protein produced by the immune system. Specifically, antibodies are created in B-cells and can exist as either soluble, free- floating protein or attached to the membrane of the B-cell in which it was produced. These proteins can recognize, bind, and neutralize harmful pathogens. Pathogens are commonly thought of as bacteria and viruses, but aberrant, pre-cancerous cells are also common B- cell targets. Once the foreign bodies are neutralized, the memory B-cells store the genetic code for the particular antibody and the host becomes immune to future exposures.

The most prevalent form of antibody found in human serum is the IgG. Composed of four proteins, the IgG contains two identical sets of light (~25 kDa) and heavy (~50 kDa) chains totaling about 150 kDa in size. The light chain has two homologues domains which includes the variable light (VL) and constant light (CL) domains. The variable heavy chains

28 contain four homologous domains which include the variable heavy (VH) and three constant heavy domains (CH1, CH2, and CH3). The variable light and variable heavy domains each contain complementarity-determining regions (CDRs) in the form of three unstructured loops. These loops are also known as the hypervariable region.

In today’s medicine, monoclonal antibodies, or antibodies which were transcribed and translated from identical DNA sequence, are the standard practice. Monoclonal antibodies differ from polyclonal antibodies in that polyclonal may contain a wide diversity of antibodies, most of which bind to the antigen, but derived from different B cells which differ in genetic makeup. Polyclonal antibodies can be easily purified from animal serum that has been previously exposed to a desired antigen. Monoclonal antibodies prove more difficult to obtain but yield a homogeneous and reproducible product, apt for FDA approval.

The first monoclonal antibody was produced in 1975 but was not licensed until the mid-

1980’s (56). Monoclonal antibodies were most commonly created using a hybridoma technique. Similar to polyclonal antibodies, this technique involves injection of a desired target antigen into an animal, in this case, a mouse (Figure 7). After sufficient time for immune response, the splenocytes (includes T- and B- lymphocytes, dendritic cells, as well as macrophages) are harvested from the mouse spleen and are either chemically or electrically fused with immortal myeloma cells. It is essential that single hybridoma cells

29 be isolated to ensure only one antibody gene is present – the alternative will lead to polyclonal antibodies. Successful fusions are isolated by selective hypoxanthine- aminopterin-thymidine (HAT) medium (57). Each new immortal cell line contains the unique gene coding for an IgG that is specific to the antigen. Once antibodies are purified from the cell line, validation of binding and affinity measurements must be performed experimentally.

Figure 7: Mouse Hybridoma Technology After a mouse is injected with desired antigen, the spleen cells are isolated and fused with Myeloma cells. The resulting Hybridomas can be selected on HAT medium and tested for expression of monoclonal antibodies that bind to target antigen. This is an open access article distributed under the Creative Commons Attribution License. Source: https://en.wikipedia.org/wiki/Monoclonal_antibody#/media/File:Monoclonals.png

Although hybridoma technology using immortal cells gave researches a way to reproducible make unlimited amounts of antibodies, this technique was limited in its clinical applicability. The first clinically approves mAb, Orthoclone, was approved in 1986 and targeted CD3 expressed on T-lymphocytes. Designed to decrease host immune rejection of kidney transplant, many patients experienced severe side effects (58). The most common side effect seen in early mAb clinical trials was the human anti-mouse antibody response, or HAMA, response (59). Because hybridoma technology yielded a genetically and therefore structurally mouse IgG, the human immune system would recognize this foreign protein and elicit adverse immune responses. As the body responds, the effects become more drastic as patients receive subsequent doses. The failure for many mAbs to reach clinical success at the end of the 20th century is in large part due to the HAMA response (56).

The development of antibody production in heterologous expression systems was a major advance in antibody engineering. Much cheaper to produce, scientists used recombinant

DNA technology to insert functional antibody genes into host expression organisms such as E. coli, yeast, insect cell culture, and mammalian cell culture (60). This expression technique allowed for host organisms to produce not only the mouse IgGs from genes sequenced after hybridoma experiments but also human IgGs. Production of human IgGs in heterologous expression systems does have its limits. The prokaryotic expression system 31 of E. coli, for example, lacks various folding pathways which prevent expression of large multidomain protein such as IgGs. Alternatively, smaller antibody fragments such as antigen-binding (Fab) fragment and single chain variable fragments (scFV) can be refolded from inclusion bodies or exported to the E. coli periplasm (61,62).

Figure 8: Antibody and Antibody Fragments The above cartoon shown the twelve domains in an IgG antibody. The Fab fragment contains the two variable, the CH1, and the CL1 domains. A single chain variable fragment contains only the variable domains connected by a peptide linker. The VHH/HV antibody fragment is a single domain antibody derived from the heavy chain only expression found camelids (63).

Using a eukaryotic expression system such as yeast has several advantages over E. coli.

Yeast have more advanced protein folding machinery that allow for production of larger proteins like IgGs as well as the higher production yields of antibody fragments (64). In addition, yeast contain the necessary glycosylation machinery that prokaryotes lack. This is of importance as the glycosylation of the CH2 in IgGs has shown to be important in preventing unwanted host immune response (65). Unfortunately, yeast glycosylation of

IgGs are slightly different than that of mammalian cells and do not wholly prevent a host immune response (60).

The insect and mammalian expression systems, although higher in cost, do not have many of the drawbacks of the E. coli and yeast systems. Both insect and mammalian cell expression have homologous glycosylation patterns to natural human IgGs and therefore allow production of non-immunogenic therapies (66). Insect cells have the advantage over mammalian expression system with the potential for higher yields. The obvious advantage of mammalian expression system is its effective ability to synthesize, transport, and secret a mammalian protein(60).

With the optimization of both technologies (hybridoma and heterologous expression) antibody engineering has seen rapid onset of biotechnological advancements. One of those advancements is through the process of antibody humanization via CDR grafting. The process of CDR grafting involves the genetic identification and transplantation of critical

33 binding CDR loops from the mouse IgG in which it was discover onto a human IgG scaffold. In theory, this would yield an non-immunogenic human antibody which retains binding affinity for the original target (67). This was first accomplished in Sir Gregory

Winter’s lab at the University of Cambridge in 1986 (68). In practice, this process contained two complications. 1 – The grafted CDR regions contained some mouse specific sequences that could illicit a human immune response. 2 – The variance between the human and mouse scaffolds caused alterations in CDR confirmations and tended to lower binding affinity (69). It would then appear that purely rational mutations are not sufficient for development of a biophysically favorable humanized antibody. Instead, a combination of rational and empirically derived mutations (semi-rational design) would be required.

The pursuit of a functional, fully humanized antibodies came to fruition with the development of antibody display systems. The first of which, and still widely used today, is that of phage display (70). Phage display is a high-throughput technique that involves the fusion of an antibody fragment, most commonly a scFV, to a coat protein of an M13 filamentous bacteriophage. The most common antibody fusions are with the pIII coat protein. This minor coat protein plays a physiological role in bacterial infection by the phage particle. A single antibody-pIII fusion can be displayed on a phage particle with the use of a helper phage. By either cloning in the antibody cDNA from the human genome or incorporating degenerate codons at the variable loops, this process mimics the natural hypermutability of the immune system. This in vitro technology allows for panning of

34 libraries as large as 1010 with a human scaffold. Details of the phage display technique can be seen in Figure 9.

Figure 9: Schematic of phage display and panning. The cartoon schematic showed the cyclic process of phage library panning. Large libraries of phage are grown in E. coli and purified. Incubation and washing of phage on immobilized antigen removed unwanted library members without antigen affinity. Eluted phage binders from immobilized surface are used to re-infect bacteria and enrich libraries. Subsequence rounds of panning are performed with increase stringency to isolate optimal binders. Antibody sequence can be elucidated from purified phage plasmid This article is distributed under the terms of the Creative Commons Attribution 4.0 International License. Source: (71).

Other display methods such as mRNA, ribosome, yeast, and bacterial display have shown successful display of antibodies and could be an alternative to phage display in development of humanized antibodies (72-75). Ribosome and mRNA display methods are advantages with their ability to create very diverse libraries. Cellular based display methods

(bacteria and yeast) can be used with other advanced techniques such as fluorescence- activated cell sorting (FACS) but are limited in library size.

Today, over 30 monoclonal antibodies have been approved by the FDA (56). With mAbs being one of the fastest growing platforms in clinical trials, this number is sure to increase.

Many new technologies are immerging in the antibody engineering field. These technologies include the use of small, 12 kDa single domain antibodies (VHH/VH in Figure

8 derived from the convergently evolved immune systems of camelids and sharks), bispecific antibodies capable of binding two unique antigens, and targeted ADCs capable of killing cancer cells with minimal side effects (43,76,77).

The promising future of antibody engineering is changing fast. We are now engaging a new era of alternative cancer therapies that are more diverse, effective, personalized, and above all safe. Hopefully, with the presence of more treatment options, the harsh and unforgiving practices of chemotherapy and radiation therapy will begin to phase out.

1.7 Anti-TAG72 Antibodies

The creation of the first anti-TAG-72 antibody was performed at the National Cancer

Institute by David Colcher and Jeffrey Schlom. Although the target was not known at the time, the generation of several monoclonal antibodies showed selective binding to breast cancer metastasis without significant interaction of normal tissues. Colcher and Schlom created this antibody library using hybridoma technology. Cellular extracts, which had been membrane enriched, from patient breast cancer liver metastasis were used to challenge the mice. Live cell immunoassays were used to assess antibody targeting sensitivity and selectivity. The most selective antibody, the first known anti-TAG-72 antibody, was an IgG called B72.3 (78). Later studies including the enzymatic digestion of

TAG-72 by neuraminidase, an enzyme naturally found in the influenza virus which cleaves sialic acid groups from glycoproteins, strongly suggested that the antigen in which B72.3 binds is the earlier described sugar moiety, Sialyl-Tn (78,79).

The next adaptation in anti-TAG-72 antibodies was with the generation of CC49. This mAb

IgG was one of twenty-eight second generation antibodies created by hybridoma technology using purified TAG-72 (similar three step purification method previously described) from homogenized human colon cancer cell lines, thus giving the library nomenclature CC (80). These experiments were conducted under the belief that because of the lack of other anti-TAG-72 mAbs in which to compare, there was no evidence that B72.3 was superior. Using a purified TAG-72 for mouse immune response rather than 37 homogenized cell membranes, would yield a library of TAG-72 binder which could be assessed for optimal biophysical properties. The Ka binding constants were determined to be 2.54 x 109 M-1 and 16.18 x 109 M-1 for B72.3 and CC49 respectively, showing an improved binding with CC49 (80).

The use of anti-TAG-72 antibodies B72.3 and CC49, within the realm of medical research, has been relatively extensive. Today, one of the most widely used application of B72.3 IgG is in the diagnosis and differentiation between pleural mesothelioma and lung adenocarcinoma (81). These two forms of cancer share similar anatomical characteristics that pathologist have difficulty in deciphering through traditional methods of immunohistochemistry (82). The use of B72.3 as a marker for TAG-72 in these cases allows for classification of the TAG-72 positive adenocarcinoma versus the TAG-72 negative mesothelioma. The mAb CC49 has shown clinical applicability in over 15 radioimmunotherapies at five institutions. Labeling of this this molecule has also been shown to be compatible with a variety of radionuclides including 131Iodine, 90 Yttrium, 177

Lutetium, and 111 Indium (83). In one such study, 131Iodine-CC49 was administered with interferon (IFN) which increased expression of TAG-72 and consequently increased tumor localization and penetration of CC49 (84).

As previously noted, mouse IgGs acquired from hybridoma technology induce the HAMA response in humans. This side effect limits the clinical applicability of mouse mAbs such as CC49. Because of this obvious drawback, a humanized version of CC49 was constructed

38 using CDR grafting. This was accomplished by polymerase chain reaction (PCR) mutagenesis to clone the CDRs from CC49 VL and VH domains onto their respective human scaffolds of LEN and 21/28' CL (85). The resulting antibody hybrid decreased in binding to TAG-72 by 2-3 fold but retained its ability for in vivo tumor binding in xenograft mice (85). Following studies with humanized, CH2 domain deleted CC49 reported no

HAMA response in any of the 21 humans tested (86).

The loss in binding affinity seen in the humanized CC49 created a desire for biophysical optimization. This was accomplished using an alternate strategy of SDR (specificity determining residues) grafting. SDR grafting involves the process of sequence alignments to determine the critical residues for antigen binding (87). The germline sequences DP25-

JH4 and DPK24-JK4 were chosen for VH and VL domains respectively based on their sequence homology (88). Residues within the CDRs of CC49 that deviated from the germline sequence were assigned as critical residues for binding and individually mutated into the germline sequences. This new construct (termed AKA and can be reference in

Figure 10) contained fewer sequence similarities to the mouse IgG than the previously engineered humanized CC49 and actively bound TAG-72 with a KA = 10.5 nM binding affinity (88).

Figure 10: Antibody variable domain sequences of CC49, DPK24, and aka. Sequence alignment shows mutations in the (A) variable light chain and (B) variable heavy chain (88).

The final development in anti-TAG-72 antibodies came with the construction of 3E8 using phage display. A phage display library was generated by randomizing CDR H3 of AKA and performing colony lift assay to screen for active TAG-72 binders. The resulting construct, 3E8, contained three mutations (L96W, N97I, and A99Q) increasing binding affinity by 20-fold (650 pM) (88). This latest progression in anti-TAG-72 antibody engineering has produced a molecule that is arguably superior by both biophysical and

40 clinical metrics. Perhaps this molecule will one day see successful clinical application as either a cancer diagnostic tool or, with further engineering, a targeted cancer therapy.

1.8 Single Chain Antibody Fragments

Antibodies are useful platforms for engineering of research, therapeutic, and diagnostic agents. Despite their favorable biophysical properties, the use of full IgG does have its drawbacks. One of these negative aspects is their size and complexity (89). A full IgG is around 150 kDa, and is too large to be easily produced by recombinant expression in yeast or E. coli. Additionally, the serum half-life of IgGs are incredibly long due to the constant domains regions interaction with FC receptors and, again, their large size. This causes complications in diagnostic antibodies, where fast blood clearance is desired. It has also been suggested that the large size of IgGs will reduce its potential for tissue penetration.

By and large, the most useful aspects of an antibody are its ability to tightly and specifically bind a target. The obvious exception is in the development of immunotherapies where effector functions of the FC domains are desired (90). In most other applications, only the binding domains, or FV region is necessary. These domains make up only four of the total twelve domains in an IgG. Isolating the FV region does have complications, however, and these issues will be discussed later in this section.

The earliest forms of antibody fragments which have seen success in the clinic are the Fab and the F(ab’)2 fragments. This includes the FDA approved Fab drugs ReoPro, Lucentis, and Cimzia produced by Eli Lilly, Genentech, and Union Chimique Belge (UCB) respectively (91). The Fab fragment contains four-domains: variable light, variable heavy, constant heavy and constant light domains. The F(ab’)2 fragment is a dimer of Fab 42 fragments linked by a disulfide bond. Traditionally, the Fab and the F(ab’)2 fragments were produced from full IgG’s by papain enzymatic digestion and acid incubation respectively

(92). More recently, Fab fragments have been transgenically expressed in both prokaryotic and eukaryotic expression systems (64,93,94). The resulting antibody fragments are stable, have increased tissue penetration, and do not elicit unwanted immunological responses.

Producing antibody fragments from full IgG’s as well as complications in Fab fragment production via E. coli does not allow for ease of engineering. This leaves room for an even smaller antibody fragment to be produced, one which can be efficiently expressed in a prokaryotic system. The development of the FV fragment allowed for the expression of the binding fragment in E. coli which could be readily mutated for enhanced biophysical properties.

The earliest production of the FV fragment involved the co-expression of two separate genes, one which coded for the variable light domain and one for the variable heavy(95).

These domains would associate with one another and form a functional dimer.

Unfortunately, the elimination of the constant domains severely affected the affinity for the domains to associate. As there would be no natural selection for the variable light and variable heavy domains to tightly associate, given they are expressed with the constant

-6 domains, the affinity for these two domains is low and on the scale of 10 M (96). The FV fragments disassociate under dilute conditions and consequently aggregate. One of the first strategies to correct this loss of domain affinity was to engineer an additional disulfide

43 bond between the two antibody domains, creating a disulfide FV fragment or dsFV (96-98).

Although the dsFV showed increased stability as a result of the covalently linked domains, engineering complications with correct disulfide placement and sensitivity to buffer oxidative/reduction conditions limited the use of dsFV in wide research and clinical use.

The next breakthrough in stabilization of the FV fragment was the genetic engineering of a peptide linker between the two variable domains producing a single chain variable fragment (scFV). These fragments were produced as a single gene and relied on the reduced entropic effects of the covalently linker domains to promote association and increase stability. Of note, a sc-dsFV has been engineered in attempt for additive stability effects, but tendencies in the unfolding of the variable domain interface yielded a less than desirable product (99). The scFV fragments have the interesting characteristic of domain swapping where the variable light chain of one scFV would associate with another scFV variable heavy chain (100). This would produce dimeric, trimeric and tetrameric species which retained their ability to bind antigen. It was also shown that adjusting the length of the linker between the two variable domains could dictate the oligomeric state of these proteins (101,102). The longer the linker (upwards of 25 amino acids), the more favorable monomeric formation of a scFV would become. Shorter linkers (less than 12 amino acids), however, restricted proper monomer formation and resulted in dimer formation. Smaller still (less than 4 amino acids), produced trimeric and tetrameric species (Figure 11).

Figure 11: Antibody domains swapping quaternary structure. The cartoon schematic of (A) a single chain antibody fragment and how domain swapping can lead to (B) diabody, (C) triabody, and (D) tetrabody formation. The variable heavy domain is shown in grey, the variable light domain is shown in white, and antigen is shown as a white circle (98).

Typically, the linkers used in the fusion of the variable light and the variable heavy domain are rich in glycine and serine. These two amino acids have small sidechains which allows a variety of torsion phi/psi angles. The polar nature of the serine residue also helps with solubility. These properties give Gly-Ser linkers flexibility and solubility (103).

In addition to Gly-Ser repeat linkers, there have been several other non-repeat linkers used in the literature. These are typically designed from known linkers found naturally in proteins or engineered to possess a certain feature (e.g. alpha helical). A list of non-repeat linkers found in the literature can be referenced in Table 1.

Table 1: List of non-repeat linkers found in the literature. The linker name, length in amino acids, sequence, quaternary structure (if any), and reference are listed.

Linker Size Protein Sequence Monomer: Reference Name (AA) Dimer

202’ 12 GKSSGSGSESKS 47:53 (104)

205 25 SSADDAKKDAAKKDDAKKDDAKKDG - (105)

212 14 GSTSGSGKSSEGKG 66:34 (105-107) (Proteolytic Site) 216 18 GSTSGSGKSSEGKGSTKG 90:10 (106,107)

217 12 GSTSKPSEEGKG - (106)

218 18 GSTSGSGKPGSGEGSTKG - (107)

205C 25 LSADDAKKDAAKKDDAKKDDAKKDL Monomer (108)

(Helical) 27 DQSNSEEAKKEEAKKEEAKKSNSLESL - (109)

CH1 23 EFAKTTAPSVYPLAPVLESSGSG - (109) Domain

1.9 The Notch Signaling Pathway

The Notch signaling pathway is a conserved mechanism of cell to cell communication used in multicellular organisms. This ancient machinery was first evolved into metazoan life hundreds of millions of years ago (110). In more modern times, the Notch gene was first discovered in 1919 and was studied as a loss-of-function mutation in embryonic development within the model system Drosophila melanogaster (111,112). It was later noted that the Notch gene coded for a 300 kDa transmembrane protein (113). The complexities of the Notch signaling pathway have been studied for almost a century and have been linked to a host of biological functions. Still today, scientists are learning new intricacies in the way the Notch protein interacts within its biological setting.

Figure 12: Notch receptors, ligands, and activation mechanism (A) In mammalian cells, the Notch signaling pathway consists of four receptors Notch1, Notch2, Notch3, and Notch4 and five ligands DLL1, DLL3, DLL4, Jagged1, and Jagged2. (B) Interaction of ligand and receptor expose a proteolytic cleavage site which allows the Notch receptor intracellular domains to migrate to the cell nucleus and promote transcription of Hes, Nrarp, and cMyc genes (114).

The Notch signaling pathway consists of four transmembrane receptors: Notch1, Notch2,

Notch3, and Notch4. These receptors consist of three regions named the Notch

Extracellular Domains (NECD), Notch Transmembrane region (TM), and the Notch

Intracellular Domains (NICD). The NECD are composed of EGF and LIN repeats. The

NICD contains a RAM domain, Ankyrin repeats (ANK) typically flanked by NLS domains, a TAD (in cases of Notch1 and Notch2), and a PEST domain. The NECD can interact with the Notch ligands expressed on adjacent cells. These ligands are also transmembrane proteins and include DLL1, DLL3, DLL4, Jagged1, and Jagged2. Unlike the Notch receptors, however, the ligands only have an extracellular region consisting of and MNNL, DSL, and EGF repeats (114). A detailed schematic of Notch activation can be seen in Figure 12: Notch receptors, ligands, and activation mechanism

The mechanisms for Notch activation involves the interaction of a Notch receptor to its ligand. This binding interaction exposes a protease close to the TM region on the NECD.

Upon protease cleavage, the NICD is then transported to the nucleus where it associates with the DNA-binding protein CSL. This interaction promotes the transcription of several downstream genes including Hes, Nrarp, and cMyc (115). These genes control many factors in cell fate, the regulation of which is critical for biological development

The Notch (1-4) receptor and ligand (DSL1,3,4 and Jagged1,2) interactions are complex and can yield a host of different downstream effects depending on the specific ligand or receptor involved. For example, development and growth of blood vessels, or angiogenesis, has been linked to the interaction between Notch1 the ligand DLL4. Normally, vascular endothelial growth factor (VEGF) can upregulate the number of DLL4 ligands expressed

49 in epithelial tip cells. This. in turn, activates Notch receptors and inhibits surrounding cells

(stalk cells) into developing new vessels (116). This process allows for spatially controlled vasculature production. Conversely, the overexpression of Jagged1 in the same cell types was shown to compete with DLL4 and inhibit the process of angiogenesis (116). It is curious cases such as these, and many similar studies have been published, that lead scientists to wonder exactly how homologous ligands and receptor interaction can have such diverging phenotypic downstream effects.

The same DLL4 – Notch1 interaction and signaling pathway has been implicated in cancer angiogenesis(117). This, however, is not the only link between the notch signaling pathway and cancer. Other connection to cancer include the initiation of epithelial-mesenchymal transition (EMT) after hypoxic shock, the resistance of anoikis (a form of programmed cell death), and the proliferation cancer stem cells (while inhibiting proper differentiation)

(117). The notch signaling pathway is so integrated in multicellular life that it becomes difficult to not find a connection between a particular cell type (diseased or normal) and its expressed notch receptors and ligands.

1.10 Notch Implication in Cancer Immunology

In addition to the well-studied connections of the Notch signaling pathway and angiogenesis, another major implication of Notch is its role in the immune system.

Recent studies of the notch signaling pathway revealed inhibition of T-cell population maturation in Notch1 deficient bone marrow. Development is halted at the immature thymocytes (CD44+/CD25−). Within this mouse model, it was shown that proper T-cell function is dependent on the expression of notch at the source of production, without which, B-cell accumulation was seen to occur instead (118,119). As expected, in addition to the Notch receptor, the Notch ligands DLL1 and DLL4 are also involved in T-cell development. Experiments involving overexpression of these ligands in immune deficient athymic nude mice showed recovery of T-cell differentiation and proliferation (120).

Interestingly, the Delta-like ligand family is known to function in a dose and density sensitive manner. The presence of high concentration of DLL1, for example, can promote

T-cell maturation while inhibiting the production of B-cells. However, at relatively low concentrations of DLL1, both the development of T- and B-cells are promoted (121).

One of the avenues already mentioned for cancer proliferation is through tumor escape, or the evasion of the immune system. Mature T-cells play a key role in the detection of cancerous growth, while other forms of T-cells such as regulatory T-cells (Tregs) can have

51 immunosuppressive characteristics (122). This leads to the question of what mechanisms these cancer cells use to effect host immune cells. With the close regulation of T-cell development being dependent on proper Notch1 and DLL1 functionality, it is not hard to imagine a problem in this pathway may be responsible for tumor escape from the host immune system.

Evidence of the Notch signaling pathway and T-cell development involved in the tumor microenvironment was noted recently in the analysis of bone marrow from patients with lung cancer (Figure 13). When compared with a control group of healthy patients, the expression level of DLL1 as well as a downstream Notch target gene, Hes1, was severely dampened in the cancer bearing patients. This difference in expression was also found experimentally in tumor bearing mice versus non-tumor bearing mice (123).

Figure 13: Notch activation in cancer patients. Comparative analysis of DLL1 (left) and downstream transcription factors of the Notch1 signaling pathway (right) show patients with cancer have decreased DLL1 expression as well as decreased Notch1 activation (123).

The phenotypic link of DLL1 expression to enhanced tumor progression was tested by introducing a DLL1-carrying retrovirus into lethally irradiated mice carrying lung cancer transplants. The recovered DLL1 gene in the mice infected with the retrovirus showed a decrease in tumor growth over a 35-day period when compared to the control group as shown in Figure 14 (123). The increased concentration of T-cells presence in the mouse spleen indicated a link between DLL1 expression, T-cell production and maturation, and reduction in tumor size.

Figure 14: DLL1 expression mouse tumor study. DLL1 suppressed mice show increased tumor growth than that of DLL1 expressing mice showing a correlation between cancer progression and the DLL1 mediated Notch1 signaling pathway (123).

This evidence suggests that the Notch signaling pathway could become an important drug target for cancer immunotherapy. A protein therapeutic which stimulates or represses either

Notch activation or inhibition may be a strategy for reversing immune escape of tumors by redirecting T-cells to attack neoplastic cancer growth.

1.11 Thesis Summary

The future approach to cancer diagnostics and therapeutics will be one of many platforms.

Within the field of protein science, researchers are strategically using the well characterized formats of antibodies in addition to other biologically relevant proteins to tackle this global health problem. With the technological advances and improved strategies found in recombinant protein expression, computational analysis, and high-throughput cellular display, identifying clinically relevant proteins for potential therapeutic development becomes increasingly less challenging.

The following body of work describes two translational protein engineering projects which aim to produce cancer diagnostic and therapeutic agents. Chapters 2 and 3 of this thesis describe the design, production, and characterization of 3E8 single chain antibody fragments (scFV) to be used as a pre-, intra-, and postoperative cancer imaging tool

(Chapter 2) as well as an ex vivo immunohistochemical staining agent (Chapter 3). Chapter

4 will discuss the development of a Notch activator by multimerization of the Notch ligand,

DLL1, to create a novel cancer immunotherapy. Materials and methodology of all experiments can be referenced in Chapter 5.

Chapter 2. Engineering linkers in anti-TAG-72 antibody fragments to optimize

biophysical properties, serum half-life, and high-specificity tumor imaging

2.1 Contributions

Nicholas E. Long – The design, cloning, purifications, and in vitro data of library members was performed by myself.

Brandon J. Sullivan – Brandon was responsible for the original design of the VL-VH and

VH-VL templates for cloning. He also assisted in large batch production of 3E8.G4S for in vivo imaging and helped further characterize 3E8.G4S oligomeric states.

The Tweedle Lab – Dr. Tweedle and his lab performed all pharmacokinetic, biodistribution, and in vivo imaging studies.

Charles Hitchcock – Dr. Hitchcock was responsible for the analysis of 3E8.scFV and

3E8.G4S immunohistochemical staining.

2.2 Abstract

Antibody fragments have great potential for clinical application as cancer therapeutics and diagnostics. Their small size compared to full-length IgGs allows for faster clearance from blood, low immunoreactivity for human-derived fragments, better tumor penetration, and simpler engineering and production. The smallest fragment of an IgG that retains binding to its antigen, the single-chain variable fragment (scFV), can be created by fusing the variable light and variable heavy domains together with a peptide linker. Along with switching domain orientations, altering the length and amino acid sequence of the linker can significantly change the biophysical characteristics such as binding, stability, and quaternary structure. Comprehensive studies of these attributes in a single scaffold have not been reported in the literature, making design and optimization of antibody fragments challenging. Here we constructed linker and orientation libraries of 3E8, an antibody specific to TAG-72, a mucinous glycoprotein overexpressed in 80% of adenocarcinomas.

We observe dramatic differences based on linker and orientation choices in biophysical properties and in vivo imaging characteristics. We have cloned, expressed, and characterized scFVs, diabodies, and higher order multimer constructs with varying linker compositions and sizes and domain orientations. These constructs were characterized by surface plasmon resonance (SPR) to test for antigen binding, by differential scanning fluorimetry (DSF) to test for thermal stability, and by gel chromatography to test for quaternary structure and homogeneity. We then optimized expression and purification of two biophysically favorable constructs, 3E8.scFV and 3E8.G4S. Both constructs were 57 subjected to biodistribution and pharmacokinetic studies in mice. From this analysis, we selected 3E8.G4S as a lead candidate agent for cancer imaging and detection. This hypothesis was confirmed with successful PET and SPECT imaging of adenocarcinoma xenograft mice 24 hours after injection.

2.3 Introduction

Despite decades of research, cancer is the second leading cause of death worldwide and is predicted to surpass heart disease as early as 2030. Last year alone, 1,685,210 new cancer cases and 595,690 cancer deaths were projected to occur in the United States (124). The inherent diversity and complexity of the disease makes the design of a single drug or cure impossible. The most widely used therapies for cancer include radiation and chemotherapy, both of which are aimed to decrease cancer viability as well as increase the patient’s longevity and quality of life. Increasingly, these are augmented with some form of immunotherapy. While these therapies alone, in some instances, cure patients, many are performed conjointly with surgery in cases of solid tumors. Today, the most successful instances of treatments depend on the complete resection of tumors and any metastases during surgery (125).

There are a handful of methods used in surgical oncology to ensure that a solid tumor has been completely removed. Knowledge of tumor progression pathways and physical palpation of patients is used alongside a variety of imaging instruments such as PET/CT and SPECT/CT for greater specificity of tumor and tumor metastases locations. These imaging techniques involve the injection of a radioactive element into a patient which is then detected by PET or SPECT and overlaid onto a 3D rendered image produced by CT.

The effectiveness of these techniques is limited by the specificity of the radioactive agent.

The most widely used agent, FDG, for example, is uptaken by rapidly metabolizing cells, 59 which is a characteristic of some tumors (126). But FDG has limited use in that it targets only highly metabolic tumors and also targets other non-cancerous metabolic cells such as the brain, liver and damaged or healing tissue. These shortcomings are in part what has inspired the use of molecular targeting agents, such as antibodies and antibody fragments, for better specificity.

Radiolabeled antibody fragments have the potential to produce better images, which can be taken pre-, intra-, and post-operatively. Unlike a full IgG, antibody fragments are small and do not contain constant domains, allowing for better tumor penetration and less severe host immune response (127,128). Most importantly, their small size allows for fast clearance from the blood, producing little background and higher signal to noise ratios, and therefore higher quality images. With the chemical addition of different size PEG molecules, the half-lives of these molecules can be specifically tuned to the ideal radio nuclide for radiologists or desired imaging time of patients (129). Our study also explores the possibility of half-life tuning without chemical modifications, but rather through designed manipulation of oligomeric states.

For antibodies to function as targeting agents, there must first be a Tumor Associated

Antigen (TAA) to which the antibody will bind (130). There have been many reported

TAAs, some of which have roles in cancer growth and metastasis that have the potential as therapeutic targets. Some antigens, on the other hand, have no known physiological function. One such TAA, TAG-72, is a mucinous glycoprotein expressed on the surface of

80% of adenocarcinomas (52). The antibody 3E8 is a humanized derivative of CC49, specific to a disaccharide, sialyl-Tn, which is found on TAG-72 (88). Our lab has created a scFV antibody fragment of 3E8 (3E8.scFV) which we use as our model protein (Sullivan

– unpublished).

Antibody fragments, despite their promising applicability in cancer diagnostic imaging, often suffer from instability, aggregation, reduced binding affinity, and decreased serum half-life (95). In the case of the variable fragment (FV), the loss in stability is due to the removal of the constant domains which have inter-domain disulfide bonds. This leaves only the weak noncovalent interaction between the VL and VH domains. Engineering in disulfide bridges (dsFV) or, more commonly, the genetic fusion of a flexible peptide linker between domains (scFV) has been shown to imrpove this interaction as well as overall stability (95,105). These linkers tend to be simple, rich in glycine and serine repeats, biologically inspired (e.g. the interdomain linker region of the fungal cellulase CBHI), or discovered via high throughput methods such as phage display (131,132). The reduction in binding affinity is often related to stability, but also can be correlated to antibody fragments monovalency (133).

An interesting feature of scFV antibodies is their ability to form dimeric, trimeric, and even higher order oligomeric species based on the size of their engineered linkers. The tunability of quaternary structure is due to the added constraints on intramolecular domain interactions, which can then only be satisfied intermolecularly. The rule of thumb is that

61 monomeric species form with twelve or longer amino acids in the linker, while 5-12 amino acids will produce a dimeric species, or diabody (134). Shorter linkers, or the direct fusion of variable domains, will produce triabodies in some studies, but tetramers in others (102).

In general, the studies leading to these general rules include small data sets and limited biophysical characterization.

In this study, through an expansive series of constructs, we elucidate the importance of antibody fragment linkers and domain orientation. Specifically, we have cloned, expressed, and characterized scFVs, diabodies, and higher order multimeric constructs with varying linker compositions, sizes and domain orientations. These constructs have been characterized by surface plasmon resonance (SPR) to test for antigen binding, by differential scanning fluorimetry (DSF) to test for thermal stability, and by gel chromatography to test for quaternary structure and homogeneity. Linker length and composition and domain orientation have great influence on all three of these biophysical characteristics, and that there exists a preferred construct for further engineering.

In addition, we have selected a single construct (VH-G4S-VL) from the linker library with ideal biophysical characteristics and pharmacokinetic (PK) properties to be used as a possible lead candidate as a cancer imaging and diagnostic agent. After further characterization, production optimization, and in vitro tissue staining, we have successfully produced xenograft mouse images with a stark 29:1 tumor to background contrast.

2.4 Results

The amino acid composition of linkers has been shown to be significant. In designing our library we chose primarily to include repeat glycine and serine linkers as their small side chains allow flexibility (135). It is also believed that serine’s polarity increases solubility and decreases immunotoxicity. We also included a non-Gly-Ser repeat linker, 205C, which is thought to be helical and advantageous over non-structured linkers, which can be more susceptible to proteolysis (107,136).

Construction of 6x-His Tagged 3E8 Library – The seventeen-member 3E8 antibody fragment library was cloned using overlap PCR and two different templates (Figure 15A).

For all VL-VH domain oriented constructs, our original 3E8.scFV (Sullivan - unpublished) was used as the template. For the cloning of VH-VL orientation constructs, a gene for a single construct was ordered and was used as the template. For all cloning and purification, our lab’s pHLIC vector was used and contained the features shown in Figure 15D (137).

Two separate PCR products were created for each construct which contained altered, but overlapping linker regions between domains, and joined together to create a full-length gene (Figure 15B, C). Constructs were purified from cell lysate by IMAC purification.

The concentration and purity of these constructs were determined by SDS-PAGE (Figure

15E), typically in the range of 0.2-0.3 mg/mL at 95% purity.

Figure 15: Library design and construction. (A) List of library members with varying linker lengths from 25 amino acids (205C) to a single glycine. (B) Detailed schematic of 3E8 gene coding region along with location of amplification and mutagenic primers. (C) Agarose gel displaying DNA fragments created during initial mutagenic PCR and full length 3E8 gene when those products are amplified together. (D) Vector map of T7 expression vector, pHLIC, containing ColE1 replication origin, Ampicillin resistance gene, Lac repressor and 3E8 cloning region between NdeI (red) and BamHI (green) restriction sites. (E) Representative SDS-PAGE gel showing a subset of purified library members purified with IMAC purification.

Gel Filtration of 6x-His Tagged 3E8 Library – The gel filtration analysis shows a distribution of oligomeric states of monomer, dimer, trimer, tetramer, to higher order oligomer (Figure 16A, B). The primary oligomeric state of our constructs is either a monomer or a dimer and is dependent on the linker length. Constructs with 15-25 amino acids produced monomers and constructs with 1-5 produced dimers (Table 2: Biophysical characterization summary of 3E8.scFV linker and domain library.). The overall pattern of quaternary structure is independent of domain orientation.

Figure 16: Biophysical characterization of library constructs by domain orientation. (A and B) Gel filtration chromatography of library members for determination of oligomeric state using analytical grade Superdex 75 column. Antibody fragment library members will elute from the analytical Superdex 75 column according to size in solution. Soluble aggregates will elute from the column followed by tetramer, trimer, dimer, and monomer. Antibody library members are grouped based on domain orientation with VH-VL orientation on the left and VL-VH domain orientation on the right. (C and D) Normalized differential scanning fluorimetry (DSF) of 3E8 library members. The hydrophobic dye, SYPRO orange increases fluorescence when binding unfolded 3E8. Relative stability can be assessed by melting temperature which is determined from fluorescence inflection points as 3E8 is thermally denatured. The constructs are grouped based on orientation where VH- VL orientation (bottom) appears to be generally less stable than VL-VH (top).

Stability Characterization – The thermostability of the seventeen antibody fragment constructs was analyzed by differential scanning fluorimetry (Figure 16: Biophysical characterization of library constructs by domain orientation.C, D). The melting points for each fragment were calculated by finding the maxima of the first derivative of each curve

(Table 2). The fragments exhibited a range in stability of 17 ºC. We have found that constructs in the VH-VL orientation displayed a lower stability of about 5 ºC in comparison

65 to their VL-VH counterparts. We also noted a non-two-state unfolding pattern which was more prevalent in the VH-VL orientation.

Binding Affinity Characterization – All antibody fragments bound to immobilized bovine submaxillary mucin (a source of sialyl-Tn) with low nM affinity (Table 2). Specifically, these constructs ranged in apparent affinity from 3-15 nM. We observed that antibody fragments in the VH-VL orientation were tighter binders. (Figure 17)

Figure 17: Binding affinity of 3E8 library members was determined by surface plasmon resonance (SPR). (A) Representative Association and dissociation curves of 3E8 VL-205C-VH construct produced by BIACORE software. Concentrations of antibodies were run at 0nM, 5nM, 10nM, 25 nM, 50 nM, 75 nM (in duplicate), 100 nM, 150 nM, and 200 nM. (B) Correlative scatter plot showing calculated KD in relation to Linker length in both VL-VH orientation (blue) and VH-VL orientation (orange).

Summary of 6x-His Tagged Library – The results of the library are a clear indication of the flexibility of the linker length, amino acid composition, and the domain orientation. All seventeen antibody fragments were well folded, stable, and functional binders. We did see

66 a tendency for instability relating to the VH-VL orientation. This decrease in stability is also correlated to constructs with higher oligomeric states. The advantages these molecules have over the monomeric species is their multivalences, increasing the binding avidity, which we can detect in the slower off rates and slightly lower KD (Table 2). The patterns we see in the oligomeric states of these molecules correlate with the linker length observations of Pluckthun (95,96,138).

Table 2: Biophysical characterization summary of 3E8.scFV linker and domain library. The list below shows each antibody construct and its measured melting temperature (degrees Celsius), binding affinity (nM), and oligomeric state (relative area calculated under each gel filtration peak). Oligomeric states are divided into monomer (Mono), dimer (Di), trimers (Tri), tetramers (Tet), and higher order oligomers (O).

Construct ID TM Ni-NTA Protein L VL-VH (˚C) KD (nM) (Tri – Di – Mono) (Oli – Tet – Tri – Di – Mono) V -(205C)-V 66.4 12 L H V -(G S) -V 68.6 11 L 4 4 H V -(G S) -V 67.4 7 L 4 3 H V -(G S) -V 70.4 2.5 N/A L 4 2 H V -(G S)-V 70 14 L 4 H V -(SG )-V 69.2 11.8 N/A L 4 H V -(GSSG)-V 71 11.9 L H V -(G )-V 71 14.9 L 4 H V -(GGG)-V 70 5.5 L H V -(GG)-V 68.6 4.6 L H V -(G)-V 70.8 3.7 L H

TM VH-VL (˚C) KD (nM) Ni-NTA Protein L V -(205C)-V 67.4 7.8 H L V -(G S) -V 67.6 5 H 4 4 L

V -(G S) -V 66 5 H 4 3 L V -(G S) -V 66.2 4.2 N/A H 4 2 L V -(G S)-V 66.8 2.6 H 4 L V -(SG )-V 54.2 5.5 H 4 L V -(GSSG)-V 67.6 5.6 H L V -(G )-V 56.8 5.3 N/A H 4 L

Design of Tagless 3E8 Library – After completion of the antibody linker library, a construct was selected for further in vivo characterization. The 3E8 construct VH-G4S-VL (3E8.G4S) showed enhanced binding and stability. In addition, the dimeric nature of this construct created a complex greater than 50 kDa and was predicted to have a longer serum half-life than its monomeric counterparts.

For future industrial production of 3E8.G4S, alterations in the expression plasmid and method of purifications were made. The resistance gene from pHLIC was changed from ampicillin to kanamycin to create the new expression vector pHLICK. Additionally, the

6x-His tag was removed from the sequence and an alternate method of purification by

Protein L was adapted. This new expression format and purification process increased our protein yields substantially. Most notably, the protein eluted from a single protein L column was equal, if not greater, in purity than our previously described two column Ni-NTA purification method (Figure 17).

Figure 18: Comparison of Ni-NTA IMAC purification with Protein L purification. SDS-PAGE gel shows purified protein yields of 3E8.G4S are increased with protein L (left) than that of IMAC purification (right). The gel filtration chromatograph shows substantial differences in oligomeric states between methods of purification. Protein L purification (red) produces two peaks corresponding to dimer and higher order oligomer trimer and tetramers. IMAC purification (blue) produces a homogeneous dimer peak.

A sample of 3E8.G4S purified using the new methodology was re-characterized and was shown to have comparable binding and stability. Surprisingly, gel filtration showed an increased presence of higher order oligomeric states (i.e. trimers, tetramers, and higher order oligomers). After further investigation, we determined that the reason for these differences was not an artifact of protein L purification or the presence of 6x-His tag, but rather is a more accurate representation of the oligomeric state. Additionally, we found that having the 6x-His tag on multimeric constructs prevented them from binding to Protein L.

For this reason, a library of tagless 3E8 linker and domain constructs was created for protein L purification and oligomeric state analysis.

Construction of Tagless 3E8 Library – The tagless 3E8 constructs were created using several different cloning methods. Two new linker constructs were created using overlap

PCR which contained a new ten amino acid linker (G4S)2 while removing the 6x-His tag and TEV site from the N-terminal portion of the gene. Once these constructs were made, all remaining VH-VL domain oriented sequences were created using a double digest of SpeI and EagI (Figure 19A). The fragments between the two cuts sites that contained the various linker sequences was ligated to the complementary fragment from the tagless construct. The VL-VH orientation sequences did not contain a SpeI so therefore had to be cloned using an alternate method. For these, PCR fragments were created using primers containing removable BsaI restriction sites (Figure 19B). The digested PCR products from each template, designed to have matching nucleotide overhangs, created a full plasmid with the desired features.

Figure 19: Cloning schematic of tagless linker library. (A) Tagless library members with VL-VH orientation were cloned using SpeI and EagI restriction sites. SpeI is located within the 3E8 gene between the TEV site and variable linker region. The small digested fragment of tagless construct 3E8.G4S was ligated to the large digested fragments from each VH-VL library member. (B) Mutagenic primers with flanking BsaI sites were used to PCR out a small fragment of the tagless 3E8 construct VL-205C-VH and a large fragment from all tagged VH-VL library members. BsaI sites were removed after digestion and complimentary overhangs between the two PCR fragments were ligated together.

Gel Filtration of Tagless 3E8 Library – Genetic removal of the 6x-His tag and purification using protein L affinity columns allowed for higher protein yield and purity. This produced a substantially different composition of oligomeric states (Figure 20A, B). Those constructs which were previously monomeric contained more dimeric and trimeric species.

Likewise, constructs which were previously dimeric contained more trimeric and tetrameric species. It also became clear that the VH-VL orientation contained larger amounts of soluble aggregate. 72

Figure 20: Tagless gel filtration using prep grade Superdex 75 column. Antibody fragment library members will elute from the column according to size in solution. Soluble aggregates will elute from the column followed by tetramer, trimer, dimer, and monomer. Antibody library members are grouped based on domain orientation with VH-VL orientation on the left and VL-VH domain orientation on the right. Substantial differences in higher order oligomers were seen between previous gel filtration data of library members.

Analytical Ultracentrifugation (AUC) – A subset of the tagless antibody constructs were analyzed using AUC. The results correlated to the gel filtration analysis in both predicted species molecular weights and relative distribution of oligomeric species (Figure 21).

Figure 21: Analytical ultracentrifugation (AUC) of select tagless antibodies was performed to corroborate gel filtration data. Sedfit analysis determined the peaks (left to right) to be monomer (2.5), dimer (3.75), trimer (4.8), and tetramer (6). Calculated ratios of oligomeric state complimented those determined by gel filtration.

Further Explorations Ni-NTA IMAC vs. Protein L Purification – To confirm the contrasting gel filtration results between Ni-NTA IMAC and Protein L purification, several additional experiments were performed. The first experiment was conducted to determine the effects of the 6x-His tag on the oligomeric states of the constructs. Gel filtration on construct VL-

(G4S)4-VL before and after TEV cleavage of the 6x-His tag revealed a small shift due to

74 the decrease in size but no change in the relative distribution of oligomeric states (Figure

22B).

Figure 22: Exploration of purification discrepancies (A) Normalized gel filtration of VH-205C- VL, VH-(G4S)4-VL, and VH-(G4S)3-VL antibody constructs at 0.2 mg/mL and 0.8 mg/mL concentrations. (B) 6x-His tag removal of VL-(G4S)4-VL antibody construct

Next, the concentration dependence of the oligomeric states was tested by concentrating three different antibody constructs and comparing their gel filtration chromatographs before and after concentration. No difference in the relative oligomeric distribution was seen (Figure 22A).

The reproducibility of the oligomeric state distribution was determined by 12 x 1 L purifications of 3E8.G4S. Gel filtration analysis of these small batch purifications showed only a small 10% deviation of dimer to higher oligomeric state ratio (Figure 23).

Figure 23: Oligomeric state reproducibility. The reproducibility of the oligomeric states of 3E8.G4S (VH-G4S-VL) was tested to determine if the various species were truly dependent on the specific linker or if variation could be seen. Our results showed a small fluctuation of trimer and tetrameter relative to the dimer species but overall the ratio fell within an average of ten percent.

Analysis of 3E8.G4S – After the initial analysis of the library members, the construct

3E8.G4S showed promising biophysical characteristics including a low KD of 3.6 nM, high stability, and a distribution of oligomeric states. Sed-fit analysis of the AUC data determined this construct to be 47% dimer, 42% trimer, and 11% tetramer. We selected this variant for further in vitro and in vivo testing. Chromatographs for 3E8.G4S analysis can be referenced in Figure 24.

Figure 24: Biophysical characterization summary of tagless 3E8.G4S. (A) SDS-PAGE gel of 3E8.G4S. Oligomeric states of tagless 3E8.G4S are determined by gel filtration (B) and analytical ultracentrifugation (C). (D) SPR binding curve shows tight binding of 3.6 nM to antigen. (E) Comparable stabilities were seen between tagged (light blue) and tagless (dark blue) 3E8.G4S constructs.

Analysis of 3E8.G4S, Enriched Diabody, and Enriched Oligomers – The enriched diabody and higher order oligomers were analyzed by gel filtration 24 hours after their original purification from 3E8.G4S. No change in oligomeric state was seen suggesting that domain swapping does not occur freely in solution (Figure 25). These enriched fractions were also analyzed by DSF and SPR. The data concluded that both diabody and oligomer enriched fractions were stable and active binders.

Figure 25: Separation and characterization of 3E8.G4S oligomeric states. (Left) Gel filtration chromatograph showing 3E8.G4S (blue). Dimer and oligomer peaks were collected and re-injected onto the column as a single dimer peak (grey) and oligomer peak (red). (Middle) Differential scanning fluorimetry of 3E8.G4S and separated dimer/oligomer samples. (Right) DSF and SPR melting temperatures and measured binding affinities of 3E8.G4S, diabody fraction, and oligomer fraction.

Immunohistochemistry – The application of targeting adenocarcinomas was tested by immunohistochemistry. For development purposes, a C-terminal FLAG tag was cloned into 3E8.G4S to create 3E8.G4S.FLAG. A single chain monomer version of the FLAG tagged 3E8 was also created, 3E8.scFV.FLAG, to see the effects of multivalence on detection sensitivity. The resulting staining showed enhanced detection ability in both antibody fragments over the commercially available anti-sialyl-Tn IgG, B-72.3 (Figure

26A) without an increase in background staining. Additionally, the multivalency of

3E8.G4S.FLAG allowed for greater dilution without loss in signal.

Pharmacokinetics – With the interest of seeing the effect of oligomeric state on the PK properties of 3E8.G4S, the trimeric and tetrameric species were separated from the dimeric species using prep scale gel filtration. These fractions were later analyzed using gel

78 filtration and no change in oligomeric species was seen (Figure 25). This suggests that the ratio of dimer, trimer, and tetramer is determined during initial periplasmic expression and does not freely exchange in solution.

Unaltered 3E8.G4S, 3E8.G4S purified dimer, and 3E8.G4S timer/tetramer were separately injected into mice. The resulting half-lives exposed the importance of oligomeric state in determining PK half-lives. The smaller molecules (100% dimer) cleared the blood faster than that of the higher order oligomers (trimer and tetramer). As expected, the half-life of

3E8.G4S was the average of the two previous half-lives (Figure 26B).

Biodistribution – In addition to PK studies, the biodistribution of the three distinct 3E8.G4S oligomeric samples were analyzed. Higher order oligomeric states of 3E8.G4S were found in higher concentration, dimeric 3E8.G4S was found in lower concentrations, and unaltered

3E8.G4S was found to fall in between. The biodistribution results correlated very well to the PK analysis and we did not see any more than expected accumulation of higher order oligomers in the kidney or liver (Figure 26C).

In Vivo imaging of xenograft mouse model – For final proof of concept, 3E8.G4S purified by Protein L was radiolabeled and injected into xenograft mice bearing human adenocarcinoma tissue (colorectal tumor). Imaging after 24 hours show low background and targeting of all tumors. The resulting SPECT image of protein L purified 3E8.G4S produced stark contrast with a tumor to background ratio of 29:1 (Figure 26D).

Figure 26: In vivo studies of 3E8 construct VH-G4S-VL. (A) Immunohistochemistry of human adenocarcinomas by two FLAG tagged versions of select library members. The far-left panel shows a monomeric FLAG-tagged 3E8 at 1:150 dilution from a 2mg/mL stock. Middle-left panel shows 3E8.G4S.FLAG (dimeric/trimeric/tetrameric mixture) at 1:2400 dilution of a 1.7mg/mL stock. The middle right panel shows commercially available B-72.3 IgG at 1:100 dilution of stock. The far- right panel shows a negative control in which the same staining protocol was used with exception to the removal of 3E8.FLAG antibody. (B) I125 activity change in blood post injection of I125- 3E8 DB, TB (12-15-14) and I125-3E8 mixture (10-13-14). Blood was collected at 0.5, 1, 5, 24, 48, and 72 h after injection for Mixture, and 0.5, 1.5, 3, 6, 24, 48, and 72h for DB and TB. Data was converted to %ID of whole blood pool and presented as average ± SD (n=5). (C) I125 activity change in mouse organs after 72 hours post injection. Unaltered 3E8.G4S (red) prevalence falls between the smaller molecular weight diabody fraction (grey) and the higher order oligomeric trimer and tetramer fraction (black). (D) Xenografted mouse images of radiolabeled 3E8.G4S. Images show iodinated 3E8.G4S antibodies accumulating at tumor sites on xenophraft mice with transplanted human colon cancer tissue. 3E8.G4S purified by Ni-NTA (containing only diabody - left) shows a smaller tumor to background ratio than 3E8.G4S purified by protein L (containing trimer and tetramer – right).

2.5 Discussion

Antibody fragments have a clear potential for clinical use as enhanced cancer imaging tools, but the widespread usage of engineered fragments is hindered by their decreased stability and binding affinity and by the difficulty of large-scale production.

The genetic fusion of the variable heavy and variable light domains via peptide linker not only allows for the expression of a single polypeptide but also reduces the entropic penalty of folding, causing increased domain interaction. Throughout the literature, single chain

FVs have been constructed with high variability in their linker orientation, length, and composition. Linker length has generally been shown to dictate multimerization where linkers greater than fifteen residues produce monomers, greater than five produce dimers, and less than five produce higher order mulitmers (101,102,134). This can be explained by the inability, due to short linker size, of the variable light and variable heavy domains to properly orient themselves for interaction. Instead, one antibody fragment’s variable light and heavy chains interact with another fragment’s respective domains in what is known as domain swapping (139). Additionally, the order in which the variable heavy and variable light domains are expressed varies throughout the literature. In some cases, VL-linker-VH rather than VH-linker-VL shows favorable biophysical characteristics, and in other cases the reverse is true (140). The complexity of these results motivated our comprehensive study in 3E8.

Our linker library study is consistent with observed linker length rules with our Ni-NTA purified 6x-His tag constructs. Larger linkers from 15-25 amino acids in length formed mostly monomers (with low amounts of dimer), whereas smaller linkers from 1-5 amino acids in length formed mostly dimers (with low amounts of trimer). However, we saw drastic changes in the quaternary structure that did not align with the established linker length rules when we analyzed protein L purified fragments. More heterogeneous compositions were observed with increased presence of higher order oligomers. We also see shifts in oligomeric distributions as the linker is decreased in size, rather than a sharp transition.

One explanation for this contrast, we believe, is the multivalency of the 6x-His tagged higher order oligomers with the Ni-NTA resin. The Ni-NTA column purification method, which is like that of most scFV linker studies, enriches monomeric and dimeric forms of

3E8 because it is easier to elute those species with imidazole. Protein L purification, alternatively, elutes all protein efficiently based on change in pH and total loss of binding.

This gives an unbiased view of the oligomeric states of the protein.

The biophysical properties of 3E8 are heavily dependent on the domain orientation. The most notable difference between the two sets of constructs is their oligomeric states. The

VH-VL domain orientation produces large amounts of higher order oligomers. In general, this explains the decrease in overall stability and the oligomers increased polyvalency is why we note lower apparent KDs in many of these constructs. The longest, and supposed

82 alpha helical, linker is of interest as it produced one of the most ideal constructs in the VL-

VH orientation but formed soluble aggregates in the VH-VL orientation. The two domains are not identical and there is no a priori reason that two possible linker connection topologies should be optimal at the same length. This optimal orientation for a particular peptide length may simply be idiosyncratic to the construct.

Immunohistochemistry with two FLAG-tagged library members complemented SPR data.

Qualitative analysis reveals darker staining of 3E8.G4S compared to 3E8.scFV at higher dilution, suggesting tighter binding of multivalent over monomeric constructs. These studies also prove the applicability of 3E8 antibody fragments for future IHC studies and the potential for antibody fragments as improved IHC agents.

Our lead molecule, 3E8.G4S, is a multivalent anti-TAG 72 antibody fragment with stability comparable to that of the full IgG. PK and biodistribution data reveal the mixture of dimer, trimer, and tetramer are ideal for serum half-life and do not bind nonspecific targets. The 48-hour image produced in a xenograft mouse model show the potential of this cancer diagnostic and the vast improvement over current technologies commonly used in the clinic.

Chapter 3. Wide-scope Immunohistochemical analysis of TAG-72 expression in solid

tumors by 3E8 and B72.3 antibodies

3.1 Contributions

Nicholas E. Long – The design, cloning, purification, and staining optimization of

3E8.scFV.FLAG was conducted by myself.

Brandon J. Sullivan – Brandon conducted proof of concept experiments and assisted in the optimization of secondary antibody.

Benjamin Swanson – Dr. Swanson was responsible for the analysis and grading of the 18 different core slides stained by either B72.3, 3E8.scFV, or no primary antibody.

Charles Hitchcock – Dr. Hitchcock assisted in the initial optimization and analysis of

3E8.scFV.FLAG immunohistochemical staining.

Kristin Miller and Kristin Schwartz – The Kristins performed the IHC staining of the slides and helped optimize the staining protocol. Specifically, Kristin Miller assisted with dilution and secondary antibody optimization. Kristin Schwartz performed the third-party validation.

3.2 Abstract

Immunohistochemistry (IHC) is a powerful tool for spatial detection of an antigen on a tissue of interest. In the clinical setting, IHC can give insight leading to accurate diagnosis, prognosis, and future mode of treatment. In cancer cases, the utility of IHC is heavily dependent on the presence of a tumor associated antigen or TAA to guide pathologist to areas of disease.

Our study is an expansive analysis of the tumor associated antigen, TAG-72. This antigen is expressed in 80% of adenocarcinomas and is relatively absent in normal tissue. Using the commercially available anti-TAG-72 IgG, B72.3, and a newly engineered

3E8.scFV.FLAG antibody fragment, we conducted a comparative analysis of TAG-72 expression in normal and diseased tissues. In addition, we stained a host of FDA relevant normal and diseased tissue to test non-specific and off target staining.

Our results showed a more sensitive staining of 3E8.scFV.FLAG…

3.3 Introduction

Last year, 1.5 million new cancer cases and over half a million cancer deaths were projected

(124). Dictated by its genetic make-up, each cancer case has variable prognoses and treatment options. In many instances, these genetic differences will give rise to changes in

RNA transcription and protein expression. This will ultimately lead to a deregulation in other biomolecules such as non-coding RNA’s and proteins which include micro-RNAs and surface exposed antigens respectively. Doctors can use tools, such as immunohistochemistry (IHC) to take advantage of tumor associated biomarkers to better identify and treat different cancers. This is already the standard in many types of current cancer diagnosis (e.g., HER2-), treatment, (e.g., miR-21) and prognosis (e.g., BRCA1)

(141).

One biomarker of importance is a mucinous tumor associated antigen (TAA), TAG-72, expressed in over 80% of adenocarcinomas which account for nearly half of all cancers diagnosed (124). This glycoprotein is commonly found in cases such as colonic, invasive ductal breast, epithelial ovarian, and non-small cell lung cancers (142). Previous clinical studies have shown, in some cancers, correlations of TAG-72 positive tissue with poor prognosis (143). This mucin is modified by an overexpressed sialyltransferase which adds a disaccharide, sialyl-Tn, that is not commonly found in normal tissue (52).

The effectiveness of IHC as a diagnostic tool relies heavily on the chosen antibody’s sensitivity and specificity, both primary and secondary, and the method of development. 86

There exist several types of anti-TAG-72 antibodies both in the literature and the clinic

(88). These antibodies have varied platforms, from full IgG to FAB and single chain antibody fragments (scFV), have ranging binding affinities, and include both mouse and humanized sequences. In vivo imaging and clinical work has been conducted using monoclonal IgG and scFV, CC49, whereas B72.3 is currently the standard in TAG-72 IHC

(144-146). A third antibody, 3E8, a humanized monoclonal antibody derived from CC49, has comparable stability and improved binding affinity (88). These characteristics make

3E8 a molecule of interest for in vivo cancer imaging and IHC with the potential for enhanced cancer detection and low immunotoxicity.

Previous work on 3E8 has yielded a stable scFV with favorable stability and low nM binding affinity (Sullivan - unpublished). The reduction in size permits fast clearance in vivo and addition of PEG and fluorescent tags allows for a highly tunable half-life and real- time detection (Sullivan - unpublished) (147). The full scope of 3E8.scFV as a cancer imaging and IHC tool is not yet known. Additionally, the reduced size of 3E8.scFV compared with B72.3 and CC49 may lead to deeper tissue penetration and increased antigen detection (148,149).

Using a FLAG-tagged version of 3E8 (3E8.scFV.FLAG), the following study is an expansive look at eighteen forms of cancer and the expression of TAG-72. In addition, we have stained a subset of normal tissues and have confirmed correlations of TAG-72 in cancers including colon, prostate, lung, cervical, ovarian, pancreatic, stomach, and breast.

For comparison to the IHC standard, B72.3 was used as a control for known TAG-72 positive tissues. We have noted drastic differences in both percent of tissues stained and degree of staining when compared to that of 3E8.scFV.FLAG. Through this study, we have identified applicable cancer types for 3E8.scFV.FLAG to be used preferentially over B72.3.

With its humanization, increased binding affinity, enhanced staining, and ease of production, 3E8.scFV and its derivatives prove to be the next generation of improved adenocarcinoma staining and imaging agents.

3.4 Results

IHC proof of concept experiments - A biotinlylated 3E8.scFV was used in a proof of concept experiment to determine if staining could be assessed in relation to B72.3. Staining of these colorectal adenocarcinoma slides showed comparable localizations of DAB but

3E8.scFV.Biotin was not as prominent as B72.3. In addition (Figure 27A). The effect of biotynlation on 3E8.scFV was not known and inconsistent reactions could result in lack of reproducibility and reduced binding affinity. To circumvent this problem, a tagged version of 3E8.scFV was designed.

Construction and characterization of 3E8.scFV.FLAG – A new staining format was designed using a FLAG-Tagged version of 3E8 (Figure 27C) which could be used in conjunction with a modified anti-FLAG IgG (Figure 27B). 3E8.scFV.FLAG was cloned by PCR that used a reverse primer to encode the C-terminal FLAG-Tag. This was ligated into a the pHLIC expression vector and grown in walker series C43 cells. Shaker flask expression yielded ~1mg/mL of 3E8.scFV.FLAG which was purified from cell lysate by

Ni-NTA affinity columns. Purified protein showed low nM binding to antigen when analyzed with surface plasmon resinance (SPR), high thermal stability, and no aggregation

(Figure 27D-F).

Figure 27: Design and Biophysical characterization of 3E8.scFV.FLAG. (A) Proof of concept Staining (B) Cartoon of staining procedure. (C) Gene schematic of 3E8.scFV.FLAG containing N- terminal PelB leader sequence, 6xHis Tag, TEV protease site, VL domain, 205C Linker, VH domain, and C-terminal FLAG-TAG. (D) Surface plasmon resonance sensogram curves and calculated KD (E) Differential scanning fluorimetry showing high melting temperature of 3E8.scFV.FLAG. (F) Gel filtration indicated that protein is majority monomeric with a small amount of dimer.

Dilution optimization and secondary antibody selection – A 2mg/mL stock of purified

3E8.scFV.FLAG was subjected to a series of dilutions ranging from 1:50 – 1:500 and used on core tissue slides for optimal staining intensity in diseased tissue, but minimal nonspecific blush in normal tissue (Figure 28). The dilution of 1:150 was chosen for the proper concentration of primary antibody to be used for further experimentation.

Figure 28: Dilution optimization of 3E8.scFV.FLAG for comparative staining to B72.3 (positive control).

Additionally, the FLAGBio secondary antibody was chosen over Genscript as it displayed no background staining while effectively staining cancerous tissue (Figure 29)

Figure 29: Selection of secondary antibody for immunohistochemical study.FLAGBio anti-FLAG, biotinylated IgG was showed dark staining of diseased tissue whereat Genscript was not compatible with 3E8.scFV.FLAG

Analysis of 18 diseased and normal tissues – After staining optimization was complete,

3E8.scFV.FLAG was used to detect the presence of TAG-72 in eighteen different tissue types. Each tissue slide contained 30-60 tissue cores with a majority of cancerous tissue cores and a small sample of normal tissue cores. Three slides of each tissue were stained with 3E8.scFV.FLAG, B72.3, and a negative control containing no primary antibody (3E8).

A fourth slide was purchased and stained by hematoxylin and eosin (H&E) for comparison.

The slide staining was then rated by percent of cells stained (from 0-100%) and by staining intensity (0-3). Slide cores which stained the negative control were very rare and not included in the data analysis.

Analysis of stained representative cores slides – Six selected tissues that stained well by both 3E8.scFV.FLAG and B72.3 were selected as representative figures seen in Figure 30.

These include colon, prostate, lung, ovarian, pancreatic and breast cancers.

Figure 30: Analysis of select TAG72(+) tissue cores. 3E8.scFV.FLAG, B72.3, No Ab, and H&E stained core slides. (Top left) Colon – C10; (Top right) Prostate – B10; (Middle left) Lung – G3; (Middle right) Ovary – H10; (Bottom left) Pancreas – C3; (Bottom right) Breast – J4. 94

Using T-Tests for significance in staining differences (Table 3), the staining patterns can be partitioned into four distinct categories. (1) Cancer types that are stained by both

3E8.scFV.FLAG and B72.3; (2) Cancer types that are only stained by 3E8.scFV; (3) Cancer types where more data is needed for statistically significant data; (4) Cancer types that are not stained by 3E8.scFV.FLAG nor B72.3.

Table 3: T-Tests conducted comparing multiple staining data sets. (Far left) 3E8.scFV.FLAG staining of diseased tissue compared with normal tissue; (Middle left) B72.3 staining of diseased tissue compared with normal tissue; (Middle right) 3E8.scFV.FLAG staining compared with B72.3 staining of diseased tissue; (Far right) 3E8.scFV.FLAG staining compared with B72.3 staining of normal tissue. Each test was performed with both percent and intensity of cells stained.

3E8 DvH B72.3 DvH D: 3E8 vs B72.3 H: 3E8 vs B72.3 Percent Intensity Percent Intensity Percent Intensity Percent Intensity CO702a 6.69E-14 2.75E-20 1.51E-09 8.81E-17 6.52E-02 1.58E-01 1.00E+00 1.00E+00 PR8011a 1.11E-06 1.09E-08 4.66E-04 2.31E-05 8.42E-03 1.59E-03 3.37E-01 3.36E-01 LC813x 9.16E-07 4.06E-09 5.45E-03 5.97E-05 1.75E-03 2.39E-02 1.00E+00 1.00E+00 BR1505b 2.96E-27 5.56E-42 1.12E-03 2.50E-06 2.38E-22 4.81E-29 1.00E+00 1.00E+00 CR1001x 3.45E-09 1.88E-14 1.35E-05 1.77E-09 1.24E-04 1.63E-03 4.28E-01 7.01E-01 OV1005a 5.53E-11 4.26E-14 9.93E-07 9.39E-10 3.98E-04 2.74E-02 1.00E+00 1.00E+00 PA1001a 4.80E-18 8.30E-22 3.14E-12 2.73E-17 8.42E-07 3.56E-05 5.63E-02 4.21E-02 ST1021x 5.66E-14 2.14E-22 3.19E-06 6.84E-10 4.66E-05 9.26E-06 1.00E+00 1.00E+00 MS481a 5.67E-04 1.47E-04 8.32E-02 9.60E-02 7.93E-04 1.72E-03 1.00E+00 1.00E+00 LV1501x 1.12E-03 2.71E-05 3.18E-01 2.26E-01 7.16E-03 6.31E-03 1.68E-01 1.68E-01 EMC961x 4.21E-01 6.77E-02 9.03E-01 3.36E-01 1.95E-04 4.69E-05 8.69E-01 8.13E-01 HN481x 9.01E-01 7.54E-02 5.38E-01 5.61E-01 1.40E-03 8.26E-04 1.14E-01 8.84E-02 ES482x 4.05E-01 4.46E-02 8.66E-01 3.55E-01 1.42E-01 1.00E+00 1.00E+00 4.39E-01 SK721x 3.22E-01 3.22E-01 1.00E+00 1.00E+00 3.22E-01 3.22E-01 1.00E+00 1.00E+00 SO801a 3.21E-01 3.21E-01 1.00E+00 1.00E+00 3.21E-01 3.21E-01 1.00E+00 1.00E+00 LM801b 3.21E-01 3.21E-01 3.21E-01 3.21E-01 4.69E-01 5.29E-01 1.00E+00 1.00E+00 OS802a 8.64E-02 8.32E-02 1.00E+00 1.00E+00 8.64E-02 8.32E-02 1.00E+00 1.00E+00 BC06006x 6.50E-03 1.31E-02 5.50E-05 5.53E-08 3.13E-05 4.15E-04 2.91E-03 7.23E-04

Of the tissues in category 1, there is no statistically significant difference between cancer staining in colon. The rest of these tissue types, however, show an increased staining with

3E8.scFV.FLAG compared to B72.3. This is seen primarily in diseased tissue except for

95 rectal cancer, in which normal tissue shows low intensity staining with 3E8.scFV and no staining of B72.3. Liver and mesothelioma have low percentages of staining in diseased tissue for both 3E8.scFV.FLAG and B72.3. The increased staining of diseased tissue with

3E8.scFV.FLAG allows for a statistically significant difference between normal and disease, whereas B72.3 does not. The endometrium, esophagus, and head and neck core slides have positive staining in both normal an diseased tissue by both B72.3 and

3E8.scFV.FLAG and the staining is not statistically significant. More cores are needed to decipher the correlation of TAG-72 expression with these cancers. Category four tissues, including skin, sarcomas, and lymphomas, do not express TAG-72, with few exceptions, in either normal or diseased tissue. The summation of this data was compiled into a heat map and can be referenced in Figure 31.

Figure 31: Heat map data of 3E8.scFV and B72.3 staining - normal and diseased tissue. Each tissue is divided into four quadrants: (Upper left) 3E8.scFV.FLAG staining of diseased tissue; (Upper right) B72.3 staining of diseased tissue; (Lower left) 3E8.scFV.FLAG staining of normal tissue; (Lower right) B72.3 staining of normal tissue. Each quadrant contains both percent of cells stained (left) and intensity of cells stained (right). Stained tissue cores are organized by percent of cells stained from high to low.

Comparison of 3E8.scFV.FLAG and B72.3 Staining – Overall, the intensity and percent stained increased with the use of 3E8.scFV.FLAG over the commercially available IgG,

B72.3 (Figure 31). Representative slides were selected to show how drastic this staining difference can be in certain cancer types (Figure 32).

Figure 32: Comparative staining using 3E8.scFV.FLAG and B72.3 of TAG72(+) tissueThe selected core from various TAG-72 (+) tissue types were chosen to show differences in diseased tissue staining between 3E8.scFV.FLAG and B72.3. Each core is labeled by tissue type and shows 3E8.scFV.FLAG staining (top) compared to B72.3 (bottom). Tissue selected include breast, lung, prostate, liver, head and neck, mesothelium, and rectum.

FDA Slide Analysis – Two additional core slides were assessed which contained a series of

FDA relevant normal and diseased tissues. These were also stained with 3E8.scFV.FLAG,

B72.3, negative control, and H&E. The resulting data was analyzed on the same staining scale as previously. The summary of results can be seen in Figure 33.

Figure 33: FDA Slide Analysis of 3E8.scFV.FLAG staining in normal tissue. Three samples of each tissue type were tested, each representing one third of a circle. Percent of staining is indicated by relative size of circle (5-20% = small, 30-40% = medium, and 40-80% = large). Color indicates intensity of stain (green = no staining, yellow = 1, orange = 2, and red = 3).

Third Party Validation – The staining of a selected core tissue slide (CO702a) containing colorectal adenocarcinomas was validated by a third party. Analysis showed accurate

99 staining of diseased tissue. The intensity of 3E8.scFV.FLAG vs B72.3 was comparable to our initial IHC study (Figure 34).

100

rd Figure 34: 3 Party Validation of 3E8.scFV.FLAG and B72.3 Colorectal cancer slides were stained by a third party and correlate well to location of TAG-72, intensity, and percent stained. 101

3.5 Discussion

This expansive study has produced interesting data, giving insight into the presence of

TAG-72 in specific human cells and tissues. Additionally, the clinical application of a novel anti-TAG-72 antibody, 3E8.scFV.FLAG, shows promise over the current use of

B72.3.

The antibody fragment 3E8.scFV.FLAG is unique in its stability, which is comparable to a full IgG. This allows for month long storage in solution at 4˚C and lyophilization for extended room temperature storage and shipping. The efficient, reproducible, and cost- effective production from E. coli allows for potential widespread adaptation and use.

The ability to detect TAG-72 in cancerous tissues by 3E8.scFV.FLAG is greater than that of B72.3 in the majority of tissues studies. Specifically, this includes: prostate, lung, cervix, ovary, pancreas, stomach, rectum, and breast. In the case of colon cancer, however, there is no statistically significant difference in staining between the two antibodies, arguing that these detection difference are tissue dependent and not an artifact of the staining technique.

We found the most substantial staining difference in breast cancer core slides. The

3E8.scFV.FLAG stained slides showed 80% of breast cancer cases were positive for TAG-

72 expression, whereas B72.3 staining only indicated 20% TAG-72 positive. These slides consisted of invasive ductal carcinomas from patients ages 33 - 75 ranging in grade from

102

1 – 3 and stage from I – IV. However, we found no correlation between these factors and expression of TAG-72 in either 3E8.scFV.FLAG or B72.3 staining.

Interestingly, in the cases of liver cancer and mesothelioma, 3E8.scFV.FLAG staining is statistically correlated to these cancers whereas B72.3 is not. This can be explained by the increased sensitivity of 3E8.scFV.FLAG. It is important to note that this increase in staining of 3E8.scFV.FLAG was not statistically significant in normal tissue, suggesting that increased staining is a reflection on 3E8.scFV.FLAG’s ability to detect TAG-72 as well as

TAG-72 presence in cancerous tissue.

There are clear instances of cancers that do not contain TAG-72 such as skin cancers, lymphomas, and sarcomas. Although staining can be seen in some cores, these are extremely rare. This absence of TAG-72 expression has been previously documented in the literature in small scale expression analysis by B72.3 (142). These cancers do not have the propensity to express this tumor associated glycoprotein in neither diseased nor healthy tissue.

TAG-72 expression level analysis of the FDA core slides by 3E8.scFV.FLAG and B72.3 showed some staining of healthy tissue. This is either an indication of Sialyl-tn presence in normal tissue or nonspecific staining of 3E8.scFV.FLAG/B72.3. Previous studies have noted the presence of TAG-72 in human tissues including gastric system and salivary glands (150). Because these two tissues show a darker stain intensity than other normal

103 tissue staining (Figure 33) it is reasonable to conclude that there is in fact either TAG-72 or Sialyl-tn present in normal stomach and salivary tissue. In addition, the dark staining of gastric tissue may be an indication of inflammation or precancerous tissue that appears pathologically indistinguishable from normal tissue by traditional H & E staining (151).

For other positive staining tissue, for example, the normal esophagus tissue shown if

Figure 33, the intensity is low compared to that of diseased tissue, giving pathologist a better indication of false positives.

Another issue that may allow for the use of 3E8 as a platform for in vivo diagnostics is the metabolic turnover of TAG-72 (+) areas in normal tissue. In adenocarcinomas, mucinous lakes containing TAG-72 will pool in and around the tumor. This is in contrast with the stomach lining and glandular systems which have a constant turnover of mucin (152,153).

Binding of normal tissue by 3E8 may ultimately lead to the fast elimination of the antibody.

For an iodinated version of 3E8, this would leave only the diseased tissue to be detected and stomach background would become minimal.

In future studies, whole tumor analysis should be performed. This would allow for better understanding of the heterogeneity of TAG-72 expression. The heterogeneity of TAG-72 expression could be correlated with stage and grade of cancer, in which we could not elucidate from our core slide analysis. Expanding the data sets to include additional normal tissue cores would give better insight to the potential of false positives in future 3E8 clinical

104 studies. This would also assess the correlation of TAG-72 expression with cancerous tissue in lower frequency cancer types such as endometrium and head and neck cancers.

Although further data points are needed, specifically normal tissue, cases such as these allow for novel applications of 3E8.scFV.FLAG as an anti-TAG-72 and diseased tissue diagnostic agent in tissues less common for TAG-72 expression.

105

Chapter 4. Engineering of a multivalent DLL1 Notch activator for cancer immunotherapy

4.1 Contributions

Nicholas E. Long – The cloning and purifications of monomer DLL1 constructs was performed by myself. I also designed, cloned, and purified the tandem DLL1 constructs used in all experiments.

Brandon J. Sullivan – Brandon was responsible for the design of monomer DLL1 constructs as well as primer design for their cloning.

Mikhail Dikov and Lab – The Dikov lab performed all in vitro cell based and in vivo studies with the DLL1 ligands.

106

4.2 Abstract

Immunotherapy is currently one of the most promising forms of cancer therapy. This treatment method relies on the host immune system’s natural ability to fight aberrant cell development and growth (i.e. cancer). However, in many cases a tumor finds ways to evade immune cell detection. Undeterred, these cancers can continue to grow and metastasize, leading to progression of the disease and poor patient prognosis. Previous proof-of-concept experiments performed in the Dikov lab have shown promise in “re-energizing” immune

T-cell function even after cancer has formerly evaded detection. Through activation of the

Notch signaling pathway in T-cells using clustered DLL1 ligands, mouse model studies have shown overall decreased tumor growth and longer survival time. Unfortunately, this

DLL1 cluster does not satisfy requirements for FDA approval, cannot be patent protected, and is economically unfeasible for large-scaled industrial production. For these reasons, an alternative protein therapeutic is needed which has the same potency (mode of activation) as the DLL1 cluster, but without its restrictions.

In a collaborative effort directed by The Ohio State Drug Development Institute, the Dikov,

Poi, and Magliery Labs have successfully engineered and characterized a new multivalent

DLL1 ligand based Notch activator. This was accomplished by truncation studies of DLL1 which indicated the minimum domains required for function and various genetic and chemical methods to make multivalent conjugates thereof. This protein therapeutic has been shown in vitro to activate the Notch signaling pathway in T-cells, is patent protected,

107 and can be reliably and cheaply produced in E. coli. It is our hope that with the addition of in vivo tumor studies, we will have created a novel immunotherapy cancer drug with broad applications.

108

4.3 Introduction

The Notch signaling pathway is one of significant complexity and is involved in diverse biological systems throughout a wide range of cell types. Named after the notched wing phenotype in Drosophila melanogaster, the genetic model system in which it was discovered, the Notch signaling pathway involves the interaction of a transmembrane receptor to adjacent cells expressing membrane bound ligands (154).

Within this pathway used by all metazoan life, there exist multiple variations of both receptors and ligands. In mammals, there are four distinct Notch receptor genes (Notch1,

Notch2, Notch3, and Notch4) and five notch ligands (Jagged1, Jagged2, DLL1, DLL3, and

DLL4) (155). The activation and inhibition of the different Notch receptor by these ligands is what gives this pathway its wide diversity in a multitude of tissue types.

As previously discussed, some of the Notch receptor/ligand interactions control the maturation of T-cells through the activation of the Notch1 receptor and have the potential to be used as a cancer immunotherapeutic target. Specifically, studies in the Dikov Lab have shown a correlation between lower expression levels of DLL1 in patients with cancer.

Further tests revealed that this link could be repeated and studied within in a mouse model system. Mice which overexpressed DLL1 showed decrease in cancer growth when compared to those that did not express DLL1 (123). This lead to the idea that a solubilized

DLL1 ligand could rescue the reduced expression of cancer bearing patients, and consequently decrease cancer growth. 109

To test this hypothesis, a “Clustered DLL1” was developed. This large protein aggregate comprised of a DLL1-FC fusion. The FC portion of this protein contained a mouse sequence and could bind to a biotinylated donkey anti-mouse IgG antibody. This complex was then oligomerized by adding NeutrAvidin, a tetramer which binds tightly to biotin. The resulting product was a soluble multivalent DLL1 cluster which would mimic the multivalent membrane of a DLL1 expressing cell. Tumor growth studies, as shown in Figure 35, revealed that this cluster had the ability to activate the Notch signaling pathway in immune cells, reducing cancer growth substantially. This was not seen, however, in immunodeficient mice, confirming the Notch activation and reduction in tumor growth is

T-cell dependent.

The biophysical mechanism of Notch1 activation by DLL1 has yet to be fully explored.

Ligand binding does not necessarily translate to Notch activation as instances of Notch cis- inhibition have been well documented (156). It has therefor been suggested, and recent publications have also shown evidence, that pulling forces play a role in determination of

Notch inhibition versus activation (157). Regardless, the mechanism in which Notch is activated by its ligand is facilitated through a binding interaction. The determination of

DLL1 binding sites would then be critical for full understanding of the activation mechanism. Homologous binding studies on Jagged1 suggest the critical domains for

Notch binding are the DSL and first two EGF repeats (158). Other structural analysis on

110

DLL4 argue the N-terminal MNNL domain also plays a role in binding of Notch receptor

(159).

Figure 35: Dikov Lab Tumor Growth Study. Lewis lung carcinoma tumor growth studies in normal and immunodeficient mouse models shows a degrease of tumor size when treated with Clustered DLL1. This result in not seen in immunodeficient mice (Rag1-/-)

Although the strategy is one of promise, further developing Clustered DLL1 into a cancer immunotherapy would not be possible. Reproducibility of cluster size, cost of production, and potential negative immunological side effect limit its application as a wide spread therapeutic. Alternatively, a transgenically expressed, discrete, multivalent form of DLL1 could be engineered to similarly activate the Notch1 pathway. This would eliminate the obvious barriers in which Clustered DLL1 has for FDA approval.

111

The following chapter describes the development of a soluble multivalent DLL1 protein which was constructed to mimic the Clustered DLL1. The active binding region DSL-

EGF1-EGF2, which we have termed “construct #42”, has been successfully refolded from

E. coli out of inclusion bodies. We have determined that monomeric construct #42 is functional through in vitro cell assays and in vivo tumor growth studies as a Notch inhibitor.

Multimerization of construct #42 by genetic incorporation of flexible glycine serine rich linkers creates a Notch activator when repeats are greater than four. The reproducibility and low cost of production of these tandem constructs give it the potential for a cancer immunotherapeutic with wide clinical applications.

112

4.4 Results

Design and construction of DLL1 monomers – The initial set of DLL1 monomers consisted of the MNNL-DSL-EGF1-EGF2-EGF3 domains. These domains were selected based on previously published data on the DLL1 homolog, Jagged1. This publication suggested the critical domain for binding were the DSL, EGF1, and EGF2 domains, and was validated by positive surface plasmon resonance (SPR) binding of a DSL-EGF1-EGF2 Jagged1 ligand to the Notch receptor (158). The amino acids which make up the DSL-EGF1-EGF2 domains were determined from another publications solution structure of this binding interface (160). Because it was uncertain if the binding region was conserved over these different groups of Notch ligands, additional N- and C-terminal domain (MNNL and EGF3 respectively) were added in some construct. These first sets of monomers were designed with mouse and human sequences for their ability to function in both mouse and human cellular assays.

The transgenic expression of these constructs was performed using E. coli overexpression in the strain Origami B. Initially, a PelB leader sequence was added to some constructs for periplasmic expression and proper disulfide formation. Despite the presence of this leader sequence, the proteins could not be expressed solubly and therefore the PelB was not used in further construct design. Instead, the proteins were solubilized using 7M guanidine Tris buffer solution and refolded using Ni-NTA on column refolding. A subset of these constructs was also designed to contain a C-terminal cysteine for multimerization using

113 maleimide reactive PEG groups. The initial set of proteins constructed and tested can be seen in Figure 36.

Figure 36: SDS-PAGE gels and cartoon of monomers and 2x Tandem constructs. Larger constructs including the tandem repeats and MNNL-DSL-EGF1-EGF2-EGF3 did not express as well as the smaller constructs DSL-EGF1-EGF2(-EGF3).

In vitro cell-based studies of monomer constructs – The initial set of monomers were subjected to two different cell based studies. These included a surface bound assay and a solution based assay for activation of the Notch signaling pathway in EL4 T-cells. The activation of the T-cells was measured by presence of the downstream Notch activation transcription factors, Hes1 and Hay1. These transcription factors were detected using both

114 a real-time RT-PCR instrument and a western blot. The resulting data was consistent with literature results in Jagged1 binding and showed inhibition of DLL1 monomer DSL-EGF1-

EGF2 against the known Notch activator, the DLL1 Cluster. This suggested that the DSL-

EGF1-EGF2 domains (named construct #42) were sufficient in order to bind to the Notch receptor and prevent interaction of other ligands (Figure 37). Additionally, when construct

#42 was plated onto a cell culture dish, in a manner that mimicked a 2D cluster or the presence of many ligands on a cell surface, activation of T-cells was noted. This gave further evidence that construct #42 was functional and that multimerization of this ligand could prove to be a viable Notch activator.

Figure 37: In Vitro cell assay for validation of DLL1 ligand functionality. (Top) EL4 T-cell cultures in the presence of soluble DLL1 ligand DSL-EGF1-EGF2, construct #42, were inhibited

115 from activation by the Cluster DLL1. (Bottom) Surface bound construct #42 was shown to activate the T-cells.

In vivo mouse tumor growth studies of construct #42 – For further validation of construct

#42 as a functional Notch inhibitor and platform for creating a clustered activator, an in vivo mouse tumor study was performed. As shown in Figure 38, Lewis lung carcinoma cells were implanted into a mouse model and construct #42 intravenously injected over a period of 13 days. The resulting data showed an accelerated growth in mice with construct

#42 in comparison with the control group. This result is consistent with the expected Notch inhibition of construct #42 and would therefore suppress T-cell activation and maturation.

116

Figure 38: DLL1 Construct #42 in vivo mouse tumor growth study. Lewis lung carcinoma tumor growth of mice injected with Notch inhibitor, construct #42, in comparison to a control group.

Attempted chemical multimerization of DLL1 Monomer construct #42 – After both in vitro cell-based assays and in vivo mouse tumor growth studies showed promising results for the functionality of construct #42, chemical conjugation methods were used to create a Notch activating cluster. The summation of all clustering strategies can be found in Figure 39.

117

Figure 39: Strategies for oligomerization of DLL1 ligand DSL-EGF1-EGF2. DLLI ligand is naturally express on the outside of the cell membrane and activates Notch receptors of adjacent cells. The DLL1 domains DSL-EGF1-EGF2 inhibit this activation. These domains can be used to mimic DLL1 Cluster, a known Notch activator(123). This can be done by chemical conjugation using the N-termini of the protein with multi-functional PEG-aldehyde linkers post NaCNBH3 reduction, engineered C-terminal cysteines with multi-functional PEG-maleimide linkers, and biotinylated protein for avidin tetramerization. Additionally, genetic fusions can be engineered with varying tandem repeats. Linker lengths and the number of repeats must be taken into consideration.

118

Clustering was first attempted by using the mouse DLL1 ligand DSL-EGF1-EGF2-Cys

(known as construct #52) which had an engineered C-terminal cysteine. Like antibody

PEGylation techniques developed in the Magliery Lab, maleimide active groups could be used to conjugate the C-terminal cysteine of the DLL1 ligands to multi headed PEG molecules. Unfortunately, after exhaustive attempts were made to create these multiverse, several problems were elucidated with this objective. Highlighted in Figure 40, these include:

1) The limitations in protein concentration inhibits the completion of the bimolecular

reaction between the maleimide group (PEG) and the C-terminal cysteine

(construct #52). This problem is usually solved by increasing the amount of PEG

in the reaction to a 1:25 protein to PEG ratio. This cannot be done with multi-

functional PEGs because the molar ratio of protein to functional groups in the

solution must be 1:1. For a four headed maleimide PEG this would mean the protein

molarity must be four times greater than that of the PEG molarity or 4:1. Increasing

PEG concentration beyond this point yields singly modified PEGs and no clustering

effect.

2) Construct #52 must be reduced in only its single C-terminal cysteine for proper

conjugation. Formation of oxidative dimers was seen in non-reducing DSD-PAGE

gels blocking the access of the PEG linker maleimide groups from proper

formation. Furthermore, if too much reducing agent is present in the reaction buffer,

then the maleimide PEG will react with the reducing agent rather than the protein. 119

3) DLL1 construct #52 native disulfide bonds are delicate under optimized buffer

conditions and can be modified by maleimide PEG rather than the desired C-

terminal cysteines. This was seen when reactions were performed with excess of

maleimide PEG. Globular proteins, such as antibody fragments, were only singly

modified with 25x molar excess of PEG where DLL1 ligands were overmodified

Figure 40: Maleimide modification of DLL1 construct #52 in comparison with globular antibody construct 3E8.Cys. (Left) The 20kDa maleimide PEG singly modifies 3E8.Cys with 25x molar excess but appears to modify DLL1 construct #52 under identical conditions. (Right) A bifunctional PEG produces single modifications to 3E8.Cys at 4x molar excess as well as a faint small dimer peak. Under identical conditions, construct #52 appears to have no dimer peak and is highly modified, producing a smear.

Multimerization using the native N-terminus of construct #42 and PEG-aldehyde linkers proved equally as difficult and produced similar problems. The third chemical conjugation strategy, using biotinylated DSL-EGF1-EGF2 showed some success. In this case, a C- terminal Avitag was engineered onto DLL1 ligand #42 and cloned into the expression vector pHLIC. Co-transformation of this construct with a vector containing the BirA gene into walker series C43 cells produced a biotinylated construct #42. This showed affinity for the streptavidin tetramer, but yields were too low for activation studies in vitro. 120

Design and construction of tandem DLL1 ligands – The second series of DLL1 constructs was based on construct #42 and contained repeats of DSL-EGF1-EGF2 domains genetically fused by a flexible linker with the sequence Gly-Ser-Ser-Gly-Ser-Ser-Gly

(Figure 41). These sets of constructs were inherently different from the chemically conjugated versions as they were aligned in tandem and not radially from the center as was the DLL1 Cluster, the PEG, and streptavidin tetramer designs. In addition, the size of these constructs and their readiness to express and refold are inversely related. Regardless of these potential complications these constructs were expressed and purified in the same manner as the monomer constructs.

121

Figure 41: SDS-PAGE gel and graphical cartoon of tandem DLL1 constructs.Purification yields of constructs are shown with SDS-PAGE gels.

In vitro cell studies on tandem DLL1 ligands – The solution based activation assay previously described was used to test the efficacy of the tandem series ligand. The resulting data showed no significant increase of activation over the control for monomeric, 2x, and

3x tandem DLL1 ligands. The 4x tandem DLL1 ligand (construct #83) showed increased presence of Hey1, a known downstream transcription factor of the Notch signaling pathway

(Figure 42).

122

Figure 42: In vitro cell based assay results of 4x (#83), 3x (#82), 2x (#55) and 1x (#42) mouse tandem DLL1 constructs. (Left) Real-time RT-PCR results showing transcription levels of HES1 by T-cells in the presence of various tandem DLL1 constructs. (Right) Western blot analysis shows expression of Hes1.

Enzyme-Linked ImmunoSpot (ELISPOT) Assay of tandem DLL1 ligands – In addition to the EL4 T-cell assays, an ELISPOT assay was performed on the monomeric, 2x, and 3x tandem DLL1 ligands. The results of this assay, as shown in Figure 43, showed Notch activation with the 4x tandem DLL1 ligand construct and no apparent activation with 2 or

3x tandem DLL1 ligand constructs.

123

Figure 43: ELISPOT assay using 4x (#83), 3x (#82), 2x (#55) and 1x (#42) mouse tandem DLL1 constructs. Mouse splenocytes were stimulated with CD3/CD28 antibodies without or with DLL1 constructs at indicated concentrations. Number of IFNγ-producing cells was determined by ELISPOT (per 105 cells)

Design and construction of tandem DLL1 ligand linker variants – The activation of the

Notch signaling pathway by the tandem DLL1 constructs is thought to be related to multiple interactions of DLL1 ligands and Notch receptors, as evident by the concentration dependence of DLL1 ligands on adjacent cell surfaces and the DLL1 Cluster (123). This being the case, the distance between the DLL1 repeats (i.e. linker size) in the tandem construct would also become an issue. To determine the importance of DLL1 tandem construct linker type, regarding Notch activation, a series of linker variants were designed.

These included a set of 2x tandem constructs with (G4S)2, (G4S)3, and (G4S)4 linkers as well as an additional three 4x tandem constructs containing the same linkers (Figure 44).

Although all constructs were successfully cloned, only the 4x tandem constructs expressed.

124

Figure 44: Tandem DLL1 constructs with varying GGGGS repeat linker. (Top from left to right) 2x tandem constructs with increasing middle linker sizes ranging from 10 – 20 amino acids in length. (Bottom from left to right) 4x tandem constructs with increasing middle linker sizes ranging from 10 – 20 amino acids in length

In vitro cell studies on tandem DLL1 ligand linker variants – The three 4x tandem DLL1 ligand linker variants were tested in the solution based in vitro EL4 T-cell assay. The results of this experiment showed a strong dependence on linker length with the downstream expression of Hes1 and a weak dependence on linker length with the downstream expression of Hey1. Constructs #87 with the shorter, 10 amino acid linker displayed the most activation of the Notch signaling pathway (Figure 45).

125

Figure 45: In vitro 4x tandem linker variant results. Real-time RT-PCR of transcription factors Hes1 and Hey1 of EL4 T-cells in the presence of 500nM 4x tandem (G4S)2 (construct #87), (G4S)3 (construct #88), and (G4S)4 (construct #89).

A separate experiment was designed to elucidate the concentration dependence of the

Notch activator, in this case, construct #87 (Figure 46). As expected, when the concentration of the tandem DLL1 ligand is increased, the downstream expression of Hes1 and Hey1 is also increased and presumably the Notch signaling pathway has been more abundantly activated.

126

Figure 46: Concentration dependency of construct #87.Real-time RT-PCR graph shows the concentration dependence of construct #87 with the expression Hes1 and Hey1 in EL4 T-cells.

Design, construction, and efficacy of 5, 6, 7, and 8x tandem DLL1 ligands – The results from the 4x tandem constructs, constructs #83 and #87, showed activation of the Notch signaling pathway in contrast to monomeric, 1, 2, and 3x tandem ligands. It was also noted that longer linkers decrease the capacity for Notch activation. The question remains: is there a specific number of DLL1 ligands required for Notch activation (i.e. a biological switch found between 3 and 4x tandem repeats)? If the answer is no, then perhaps adding more repeats would increase the potency of this molecule and activate the Notch pathway more effectively.

To further investigate this issue, mouse tandem constructs with 5-8x repeats were constructed by designing an 8x tandem gene and restriction digesting away portions of the gene to create the rest of the small library. These were then purified using the same on 127 column refolding method. The 7 and 8x tandem constructs were too large for efficient E. coli expression and therefore could not be validated in vitro. For a final validation of lead

Notch activating molecule, 4x tandem constructs #83 and #87, and 5 and 6x tandem constructs were all tested alongside known Notch activating DLL1 cluster. For these experiments, quantitative RT-PCR was used to measure Notch activation in 33L1

Fibroblasts cells. The results of these experiments showed, surprisingly, little to no activation of the Notch signaling pathway with construct #83. In addition, the 5x tandem

DLL1 ligand also showed little activation. However, both construct #87 and 6x tandem

DLL1 ligand showed activation of the Notch pathway and was comparable to Cluster

DLL1 (Figure 47).

128

Figure 47: In vitro analysis of Notch activating constructs. Quantitative RT-PCR of 33L1 fibroblasts in the incubated with decreasing amounts of 4x tandem construct #83 and #87, 5x and 6x tandem constructs, and Clustered DLL1. Protein concentration units are in ug/mL.

129

4.5 Discussion

The transgenic expression of the DLL1 ligands in E. coli allows for cost and time effective means of production. Protein production within a non-native cell line, however, can have complications. One such complication seen in prokaryotic expression systems (i.e. E. coli) is their lack of machinery for protein glycosylation (161). Studies have shown that glycosylation is critical for proper Notch1 expression, trafficking, and binding and therefore require eukaryotic, or, better yet, mammalian cell production (162). The Notch ligands can retain their binding ability (although decrease in binding affinity has been noted in non-glycosylated DLL4) with E. coli production despite their non-native glycosylation states (159,163,164). This potential loss in affinity of non-glycosylated forms of DLL1 may decrease its efficacy for therapeutic use. Mutations at the binding interface or production in a mammalian cell line may be a viable strategy to regain any loss in affinity and increase efficacy of the DLL1 immunotherapy.

Another complication involves the method of production. DLL1 ligand constructs produced in E. coli are insoluble and form inclusion bodies. These must be refolded using a reducing agent and high concentrations of guanidine HCl. Typical refolding strategies rely on the proteins ability to obtain its native fold without the assistance of chaperones.

Relative to their native state, misfolded or partially unfolded proteins are lower in stability.

These suboptimal proteins tend to aggregate and fall out of solution. It is under this assumption that we believe our soluble constructs have adapted their native fold and this is supported by in vitro and in vivo functionality. Still, the homogeneity of our constructs 130 is not well understood and refolding conditions could produce a variety of folds with varying degree of efficacy.

Our lead active molecule, DLL1 domains DSL-EGF1-EGF2, has been shown to be sufficient in binding to the Notch1 ligand in both an inhibitory manner with monomers, and activating manner with four and six tandem repeats. This correlates well with binding results found using the homologues ligand Jagged1. This data also suggested the DSL-

EGF1-EGF2 domains were efficient for binding. Subsequent publications, however, suggest the ligand MNNL domain also interacts with Notch1 receptor. The inclusion of the

MNNL in our molecules could therefore enhance binding and/or specificity and prove to be a more effective therapeutic.

The mechanism of Notch1 activation by our DLL1 tandem constructs is not fully understood. Furthermore, the transition between DLL1 monomeric Notch inhibition and activation by the addition of ligand repeats is a curiosity worth additional investigation. It can be reasoned that DLL1 monomers possess the ability to bind to Notch1 but lack an anchor point required to produce force and consequently activate. The DLL1 monomeric binding (i.e. construct #42) to Notch1 would, however, prevent other membrane anchored

DLL1 ligands from accessing the Notch1 binding site. Free DLL1 ligands would therefore be an effective Notch inhibitor as we see with construct #42.

131

The addition of subsequence DLL1 ligand repeats, although identical in sequence and structure, changes the biophysical properties of the molecule through several parameters.

The first is its multivalency, or its ability to bind multiple Notch1 receptors simultaneously.

Polyvalent molecules have increased observed binding affinities through avidity effects.

Alternatively, the multiple binding sites could allow for the tethering adjacent or neighboring cell Notch1 receptors. These interactions may facilitate activation through tugging forces. Finally, the addition of DLL1 ligand repeats increases the distance between terminal ligand binding sites. For example, 2 and 3x tandem DLL1 ligand repeats may not be effective Notch1 activators because the first DLL1 repeat is too spatially close to repeats

2 and 3. However, the fourth repeat, seen in 4x Tandem constructs #83 and #87, is perhaps far enough away from the first repeat to access two Notch1 binding sites at once and produce a force capable of activation. The reasoning behind Notch activation by our tandem DLL1 repeats is most likely do to a combination of increased affinity by avidity effects, its ability to cross link Notch1 receptors, and the distance between terminal bind sites.

Further scientific exploration is needed to address the questions I have thus far highlighted.

In regard to the development of a potential therapeutic from 4 and 6x DLL1 tandem constructs, in vivo tumor mouse studies must be performed to see decreased tumor growth.

In addition, dosing amounts and timing can be determined by pharmacokinetic analysis of our ligands. These studies are heavily dependent on the molecular weight of the injected

132 protein and could prove important when deciding between and ~50kDa protein (4x tandem) and a ~75 kDa protein (6x tandem).

For addressing fundamental question is DLL1/Notch1 activation mechanisms, NMR structural based studies using 15N and 13C labeled protein can be used and become feasible through E. coli growth in minimal media. In vitro binding analysis using surface plasmon resonance (SPR) would also give insight into binding affinities and avidity effects.

The lead molecules produced in this protein engineer project are a promising start to a viable immune therapeutic. With its mechanisms targeting T-cell maturation, the applicability for cancer beyond Lewis lung carcinomas is promising. Through further target validation and production optimization, the tandem DLL1 repeats find their way to the clinic, offering a novel strategy for cancer treatment.

133

Chapter 5. Materials and Methodology

5.1 Chapter 2 Materials and Methods

Design of linker library – The library of 3E8.scFV constructs was designed to include a variety of linkers. These include a long, structured linker, 205C, which was the native linker in 3E8.scFV and was known to form a monomer. Additionally, Gly-Ser repeat and non-repeat linkers composed of (G4S)4, (G4S)3, (G4S)2, G4S, SG4, G4, and GSSG, were included and were predicted to form monomers and dimers respectively. The order of the variable domains was also varied with these seven linkers. Lastly, smaller linkers G, GG, and GGG in the VL-Linker-VH orientation were chosen to test for higher order multimer formation.

Construction of linker library – 3E8.scFV constructs were cloned into a pHLIC plasmid with LacI cassette, ColE1 origin, and ampicillin resistance (Figure 15D). The original

3E8.scFV construct was designed to contain a PelB leader sequence for periplasmic expression, PAMA cleavage site, 6x-His tag, and a TEV protease cleavage site. This is followed by the variable light domain, 205C linker, and the variable heavy domain. The

3E8 VH-G4S-VL construct was used as a template in overlap PCR for VL-VH and VH-VL constructs respectively. Mutagenic primers were designed to include portions of the VH or

VL domain as well as the desired linker (primers II and III in Figure 15B). Mutagenic primers were paired with amplification primers which flanked the 3E8 gene on each side.

The two PCR products were then combined with amplification primers I and IV to produce 134 the full gene, coding for the new linker (Figure 15B, C). The gene was then ligated into a pHLIC vector between NdeI and BamHI sites, transformed into DH10B E.coli, and sequenced confirmed.

Cloning of VH-VL Tagless Variants – The constructs with VH-VL domain orientation were cloned using a series of PCR and restriction digest steps. Initially, a tagless version of VH-

G4S-VL was made by PCR using the tagged VH-G4S-VL library member. The forward primers contained the PelB leader sequence and ended with an 18-base pair overlap of the

VH domain, omitting the 6x-His tag and TEV protease site. This construct was then used to create the remaining tagless variants by restriction digest of the tagged and tagless plasmids with EagI and SpeI (Figure 19A). This digest separated all features upstream of the VH domain from the rest of the gene and all previously tagged plasmids could be ligated with a DNA sequence removed from an identical double digest of the tagless VH-G4S-VL.

The tagless construct VH-(G4S)2-VL was added to the library and cloned using the same overlap PCR methods as described in the construction of the tagged variants.

Cloning of VL-VH Tagless Variants – The constructs with VL-VH domain orientation did not contain a SpeI restriction site and therefore were cloned using PCR with primers containing flanking BsaI restriction sites (Figure 19B). Because BsaI restriction enzymes cut outside of their recognition sites, after primer digestion the restriction site is removed from the final product. The primers were designed to remove the 6x-His tag and TEV

135 protease site. The tagless construct VL-(G4S)2-VH was added to the library and cloned using the same overlap PCR methods as described in the tagged variants.

Expression of Linker Library – The sequence confirmed constructs were transformed into

C43(DE3) Walker Series expression strains. Seed cultures were grown and inoculated into

1.5 L cultures at 37 °C. Induction of cultures with 0.05 mM IPTG was performed when cell density measured 0.6 OD at 600 nm wavelength. Cultures were then cold shocked at 4

°C and expressed at 16 °C overnight. Cells were centrifuged at 9,000 g for 5 min and pellets were either stored at -80 °C or purified immediately.

IMAC Purification of 6x-His Tagged Antibodies – The cell pellets were re-suspended in 25 mL of a 50 mM Tris-HCl, 300 mM NaCl, 15 mM imidazole buffering solution (pH 8).

Additives to the cell suspension were then: 75 μL of 2 M MgCl2, 150 μL of 150 mM

CaCl2, 5 μL of 10 mg mL-1 RNase, 50 units of DNase (Roche Diagnostics), 30 mg of lysozyme, and 300 μL of 10% Triton X-100. The suspension was then lysed using and

Emulsiflex model, centrifuged at 40,000 g for 45 minutes and purified using Ni-NTA agarose resin. The imidazole eluted protein was then TEV protease cleaved overnight at room temperature, dialyzed out of imidazole, and incubated again with Ni-NTA resin to remove the affinity tag and TEV protease. The flow-through of the second Ni-NTA column was evaluated by SDS-PAGE for relative quantification and purity.

136

Protein L Purification of Tagless Antibodies – The cell pellets were re-suspended in 25 mL of protein L binding buffer (150 mM Sodium Phosphate Buffer, 150 mM NaCl, pH 7). The same additives as with Ni-NTA purification were mixed into the resuspension. The suspension was then lysed using Emulsiflex, centrifuged at 40,000 g for 2 x 45 minutes

(decanting supernatant into new centrifuges tubes between runs). The lysate was then hand pumped into pre-packed 1 mL GE Healthcare Lifesciences HiTrap Protein L columns at

~1 mL/min. Five column volumes of binding buffer were pumped through the column as a washing step. Protein was eluted from the column using a 100 mM glycine buffer (pH

3).

Differential Scanning Fluorimetry (DSF) – Purified antibodies were heated with a hydrophobic dye, SYPRO orange, and the resulting fluorescence was measured. A working stock of SYPRO orange was made by diluting in PBS from 5000X to 350X. 1 μL of working stock dye was added to 19 μL antibody at 0.2 mg/mL and 0.04 mg/mL concentration. Both dilutions of each antibody were measured in triplicate. Samples were heated and fluorescence was measured every 0.2 °C from 25-95 °C in a Bio Rad C1000

Thermocyler using the FRET channel.

Surface Plasmon Resonance – SPR experiments were performed on a Biacore T100.

Bovine submaxillary mucin (BSM) was conjugated to a CM5 censor chip by amine coupling. Antibody fragments were diluted in HEPES-buffered saline (HBS) to concentrations of 5 mM, 10 mM, 25 mM, 50 mM, 75 mM, 100 mM, 150 mM, and 200

137 mM. A buffer control and a replicate at 75 mM were also measured for binding. Chip regeneration was performed between each injection using a 6 M guanidine hydrochloride,

200 mM acetate solution.

Analytical Superdex 75 Column – IMAC purified antibody fragments were analyzed by

SuperdexTM 75 10/300 GL gel filtration column. 500 μL of 0.2 mg/mL sample was injected onto the column and run at 0.4 mL/min. The gel filtration running buffer contained

50 mM Tris-HCl, pH 8 and 100 mM NaCl.

Prep Grade Superdex 75 Column – Protein L purified antibody fragments were separated by HiLoadTM 16/60 SuperdexTM 75 prep grade gel filtration column. 500 μL of 0.2 mg/mL sample was injected onto the column and run at 0.5 mL/min. The gel filtration running buffer contained 50 mM Tris, pH 8 and 100 mM NaCl.

Analytical Ultracentrifugation (AUC) – A subset of the tagless antibody fragments were measured by AUC (model, location). Concentration of antibody fragments was normalized to 0.2 mg/mL. The oligomeric states and relative concentrations were analyzed by Sedfit software.

Gel filtration of 6x-His Tag Cleaved Antibody - To determine the effects of the 6x-His tag on quaternary states, the antibody construct VL-(G4S)4-VL, containing a TEV cleavable 6x-

His tag, was purified by Protein L. Purified protein was then separated into two aliquots.

One sample was fully digested by TEV protease, as determined by an apparent SDS-PAGE

138 molecular weight shift. Both samples were analyzed by the Prep Grade Superdex 75 column for changes in oligomeric states.

Concentration Dependence of Oligomeric States – The concentration dependence of oligomeric state distribution was determined by gel filtration analysis of three antibody constructs, VH-205C-VL, VH-(G4S)4-VL, and VH-(G4S)3-VL. Constructs were purified via

Protein L to a final concentration of 0.2 mg/mL (8 μM). A portion of each purified antibody was then further concentrated to 0.8 mg/mL (31 μM). All six samples were then injected onto the Prep Grade Superdex 75 column. Chromatographs were normalized to a maximum value of 1 to better illustrate quaternary structure distribution independent of variations in concentration.

Reproducibility of Protein L Purification – The reproducibility of the oligomeric distribution of 3E8.G4S was elucidated by purification of 12 x 1 L cultures. Samples were purified by protein L as previously described. Final concentration of the 2 samples in PBS was determined by SDS-PAGE gel to be 0.2 mg/mL (8 μM). Replicates were then injected onto the analytical Superdex 75 column. Chromatographs were normalized to a maximum value of 1 to better illustrate quaternary structure distribution independent of variations in concentration.

Separation of Oligomeric States in 3E8.G4S – Using the analytical Superdex 75 Column,

Protein L purified 3E8.G4S separated by oligomeric states. The resulting chromatograph

139 showed two distinct size distributions correlating to dimer and higher order oligomers (later determined as trimer and tetramer). Eluted fractions containing only dimer were collected separately from fractions containing only oligomers. Final concentrations of these fractions were determined by SDS-PAGE gel to be 0.7 mg/mL (27 μM) for enriched diabody and 1 mg/mL (38 μM) for enriched oligomers. After 24 hours, these samples were reanalyzed by gel filtration. A 1mg/mL (38 μM) sample of unaltered 3E8.G4S was reserved for later comparative biophysical analysis.

Pharmacokinetics (Tweedle Lab) – Normal 6- to 8-week old female Balb/c mice were used for evaluating blood clearance and biodistribution. 5 μCi of 125I-labeled proteins were injected in 100 μL of PBS in the tail veins. Blood samples (5 μL) were drawn from the saphenous vein by puncture using a 30 g syringe needle at 1, 5, 24, 48, and 72 h post injection, and blood samples were collected into a capillary tube. The radioactivity of the blood samples was counted using an automated well counter (Perkin-Elmer Wizard II,

Model 2480, Waltham, MA). A blood factor of 78 mL/kg was used to calculate % ID for each mouse based on the individual weight of the mouse. Mean %ID was determined for each dose group at each time point, and area under the mean %ID versus time curve was calculated between 1 and 72 h (AUC 1–72 h) using the trapezoid rule and expressed as

%ID*h. Blood clearance was calculated as dose/AUC and expressed as mL/h/kg.

Biodistribution (Tweedle Lab) – For tissue distribution, mice were sacrificed at 72 h after administration. Organs and tissues were dissected, including heart, lungs, spleen, liver,

140 kidneys, pancreas, gastrointestinal tract (GI), muscle, skin, blood, tail, and carcass. Organs and tissues were then weighed, and radioactivity was counted using well counter. The percentage of injected dose per gram (%ID/g) for each tissue and tumor / non-tumor tissue ratios were calculated from the %ID/g data.

141

5.2 Chapter 3 Materials and Methods

Expression of 3E8.scFV – The sequence confirmed construct was transformed into

C43(DE3) Walker Series expression strains. The expression protocol for 3E8.scFV which includes inoculation at 37°C, IPTG and cold-shock induction, and overnight expression at

16°C. Cells are spun down at 5000 RPM for 5 minutes and pellets were either stored at -

80C or purified immediately.

IMAC purification of 3E8.scFV – The cell pellets were re-suspended in 25mL of a 50mM

Tris, 300mM NaCl, 15mM IPTG Buffering solution. Additives to the cell suspension were then added as follows: The suspension was then lysed using Emulsiflex, centrifuges at

15000 RPM for 45 minutes and purified using Ni-NTA resin. The imidazole eluted protein will be TEV cleaved overnight at room temperature, dialyzed out of imidazole, and incubated again with Ni-NTA resin to remove the affinity tag and TEV. The flow-through of the second Ni-NTA column will then be run on an SDS-Page gel for relative quantification and purity.

3E8.scFV Biotinylation – The 3E8.scFV was labeled at surface exposed lysines with a 3 molar excess of NHS-biotin (H1759-Sigma). Excess labeling reagent was removed by dialysis into 50 mM potassium phosphate, 300 mM NaCl, pH 8. Slides were stained with

5 μM scFV.

142

B72.3 Immunohistochemistry Development – Paraffin-embedded tissue was cut at 4 µm and sections were placed on positively-charged slides. Slides were then placed in a 60 ˚C oven for 1 hour, cooled, deparaffinized and rehydrated through xylene and graded ethanol solutions to water. All slides were quenched for 5 minutes in a 3% hydrogen peroxide aqueous solution to block for endogenous peroxidase. Antigen retrieval was performed by

Heat-Induced Epitope Retrieval (HIER) in which the slides were placed in a 1X solution of Target Retrieval Solution for 25 minutes at 96 ˚C using a vegetable steamer (Black &

Decker) and cooled for 15 minutes in solution. Slides were stained with the Dako

Autostainer Immunostaining System. All incubations on the Autostainer were performed at room temperature. Monoclonal mouse antibody, B72.3, was ordered from BioCare

Medical (Catalog #CM002C) and a 1:100 dilution was applied to the slide for 30min. A

1:200 Goat anti-Mouse, Biotinylated, in 2% Normal Goat Serum, Vector, BA9200, 30 minutes, Vectastain Elite Standard ABC Kit-HRP, Vector, PK6100, 30 minutes. Liquid

DAB+ Chromogen (Dako K346811) was then added and developed for 5 minutes. Slides were counterstained in Richard Allen hematoxylin, dehydrated through graded ethanol solutions, cleared with xylene, and coverslipped.

Design and cloning of 3E8.scFV.FLAG – The FLAG-tag was cloned into 3E8.scFV at the

C-Terminus using forward primer:

attattattCATATGAAATATCTGTTACCTACTGCTGC and reverse primer:

aataatGGATCCTTActtatcgtcgtcatccttgtaatctCCaCTGCTCACGGTCACCAGGG.

143

The latter of which containing the sequence for the FLAG-tag. PCR fragments were digested and ligated into expression vector, pHLIC (137).

Expression and IMAC purification of 3E8.scFV.FLAG – The expression and purification of 3E8.scFV.FLAG was performed in the same manner as 3E8.scFV.

Differential Scanning Fluorimetry (DSF) of 3E8.scFV.FLAG – Purified antibodies were heated with a hydrophobic dye, SYPRO orange, and the resulting fluorescence was measured. A working stock of SYPRO orange was made by diluting in PBS from 5000X to 350X. 1uL of working stock dye was added to 19 uL antibody at 0.2mg/mL and

0.04mg/mL concentration. Both dilutions of each antibody were run in triplicate. Samples were heated and fluorescence was measured every 0.2 degrees C from 25-95 degrees

Celsius in a Bio Rad C1000 Thermocyler

Surface Plasmon Resonance (SPR) of 3E8.scFV.FLAG – Bovine submaxillary mucin

(BSM) was conjugated to a CM5 censor chip. Antibody fragments were diluted in HBS to concentrations of 5mM, 10mM, 25mM, 50mM, 75mM, 100mM, 150mM, and 200mM. A buffer control and a replicate at 75mM were also measured for binding.

Gel filtration (Analytical Superdex 75 Column) of 3E8.scFV.FLAG – IMAC Purified antibody fragments were analyzed by SuperdexTM 75 10/300 GL gel filtration column.

144

500uL of 0.2mg/mL sample was injected onto the column and run at 0.4mL/min. The gel filtration running buffer contained 50mM Tris, pH = 8 and 100mM NaCl.

Selection of secondary Ab and 3E8.scFV.FLAG dilution: Four secondary antibodies were purchased from the following companies:

1) GenScript A01429-100 DYKDDDDK Tag Antibody

2) Cell Signaling 2908S “DYKDDDDK Tag Antibody”

3) Rockland/VWR 200-306-383 FLAG detection Antibody, Monoclonal

Antibody Biotin Conjugated

4) Thermo Fisher Scientific MA191878BTIN FLAG Tag mAB, Biotin

Conjugate

All secondary antibodies showed no background/non-specific staining when staining without primary antibody. Positive 3E8.scFV.FLAG staining was then tested with antibodies from GenScript and Thermo fisher. Only Thermo Fisher Scientific secondary

Ab showed comparable staining to B72.3 positive control.

Staining of US Biomax core tissue slides – Eighteen tissue slides purchased from

USBiomax were stained with the three following conditions. 1 – 3E8.scFV.FLAG. 2 –

B72.3. 3 – No primary antibody (negative control).

145

3E8.scFV.FLAG Immunohistochemistry Development – Paraffin-embedded tissue was cut at 4 µm and sections were placed on positively-charged slides. Slides were then placed in a 60 ˚C oven for 1 hour, cooled, deparaffinized and rehydrated through xylene and graded ethanol solutions to water. All slides were quenched for 5 minutes in a 3% hydrogen peroxide aqueous solution to block for endogenous peroxidase. Antigen retrieval was performed by Heat-Induced Epitope Retrieval (HIER) in which the slides were placed in a

1X solution of Target Retrieval Solution for 25 minutes at 96 ˚C using a vegetable steamer

(Black & Decker) and cooled for 15 minutes in solution. Slides were stained with the

Intellipath Autostainer Immunostaining System. All incubations on the Autostainer were performed at room temperature. A 1:150 dilution of a 2 mg/mL stock was added to the slide for 60 minutes. 1:200 FLAG Epitope Tag Antibody, Biotin conjugate (Thermo Fisher

MA1-91878-BTIN) 30 minutes. Vectastain Elite Standard ABC Kit- HRP (Vector

PK6100) 30 minutes. Liquid DAB+ Chromogen (Dako K346811) was then added and developed for 5 minutes. Slides were counterstained in Richard Allen hematoxylin, dehydrated through graded ethanol solutions, cleared with xylene, and coverslipped.

B72.3 (+): Protocol described above in “B72.3 Development”

No Ab (-): Some protocol as 3E8.scFV.FLAG but with PBS added instead of primary antibody

H&E: Pre-stained slide was purchased from USBiomax

146

Analysis of staining (Dr. Benjamin Swanson) – Stained IHC core slides were scanned at

20x (look up scanner) and sent to Dr. Ben Swanson. Cores were analyzed and assessed by percentage of cell staining (0-100%) and stain intensity (0-3). Additionally, it was noted whether luminal and/or cytosolic staining was present.

147

5.3 Chapter 4 Materials and Methods

Design and construction of DLL1 ligand monomers and 2xTandem constructs – Four genes were ordered from GENEWIZ, Inc. These constructs consisted of two human and two mouse dimer variants each with and without EGFR3 domains. Using PCR, features such as a PelB leader sequence, C-terminal cysteine, and an additional DLL1 repeat were selectively cloned into the expression vector, pHLIC, using designed primers highlighted in Figure 48. The various combinations of primers have the potential to produce eight different flavors of each gene ordered.

Figure 48: Design of monomers and 2xTandem constructs.

Human and mouse genes of the DLL1 ligand contained a PelB leader sequence, His6 tag, and a TEV protease site. This was followed by a DSL, EGF1, EGF2, and an EGF3 domain

(in two of the four genes). A flexible GSSGSSG linker separated the two repeats. Red arrows represent the potential oligonucleotides that can be used for PCR production of desired gene.

148

The following protein sequences are the genes ordered from GENEWIZ, Inc Appendix?

Mouse DLL1:

MKYLLPTAAAGLLLLAAQPAMAA HHHHHH GGENLYFQG FVCDEHYYGEGCSVFCRPRDDAFGHFTCGDRGEKMCDPGWKGQYC TDPICLPGCDDQHGYCDKPGECKCRVGWQGRYCD ECIRYPGCLHGTCQQPWQCNCQEGWGGLFCN (QDLNYCTHHKPCRNGATCTNTGQGSYTCSCRPGYTGANCEL) GSSGSSG FVCDEHYYGEGCSVFCRPRDDAFGHFTCGDRGEKMCDPGWKGQYC TDPICLPGCDDQHGYCDKPGECKCRVGWQGRYCD ECIRYPGCLHGTCQQPWQCNCQEGWGGLFCN (QDLNYCTHHKPCRNGATCTNTGQGSYTCSCRPGYTGANCEL) GSSGC

Human DLL1:

MKYLLPTAAAGLLLLAAQPAMAA HHHHHH GGENLYFQG FVCDEHYYGEGCSVFCRPRDDAFGHFTCGERGEKVCNPGWKGPYCT EPICLPGCDEQHGFCDKPGECKCRVGWQGRYCD ECIRYPGCLHGTCQQPWQCNCQEGWGGLFCN (QDLNYCTHHKPCKNGATCTNTGQGSYTCSCRPGYTGATCEL) GSSGSSG FVCDEHYYGEGCSVFCRPRDDAFGHFTCGERGEKVCNPGWKGPYCT EPICLPGCDEQHGFCDKPGECKCRVGWQGRYCD ECIRYPGCLHGTCQQPWQCNCQEGWGGLFCN (QDLNYCTHHKPCKNGATCTNTGQGSYTCSCRPGYTGATCEL) GSSGC

Design and construction of DLL1 Ligands containing MNNL Domains – Two additional genes were ordered from GENEWIZ, Inc which contained an additional N-Terminal

MNNL domain, both human and mouse. These genes did not contain tandem repeats.

Primers were designed to include all three EGF domains or only two EGF domains.

Additionally, the PelB leader sequence could be removed by PCR.

149

The following MNNL-DSL-EGF123 protein sequences are the genes ordered from

GENEWIZ, Inc Appendix?

Mouse DLL1:

MKYLLPTAAAGLLLLAAQPAMAA HHHHHH GGENLYFQG QVWSSGVFELKLQEFVNKKGLLGNRNCCRGGSGPPCACRTFFRVCLKHYQASVS PEPPCTYGSAVTPVLGVDSFSLPDGAGIDPAFSNPIRFPFGFTWPGTFSLIIEALHTD SPDDLATENPERLISRLTTQRHLTVGEEWSQDLHSSGRTDLRYSYR FVCDEHYYGEGCSVFCRPRDDAFGHFTCGDRGEKMCDPGWKGQYC TDPICLPGCDDQHGYCDKPGECKCRVGWQGRYCD ECIRYPGCLHGTCQQPWQCNCQEGWGGLFCN QDLNYCTHHKPCRNGATCTNTGQGSYTCSCRPGYTGANCEL

Human DLL1:

MKYLLPTAAAGLLLLAAQPAMA A HHHHHH GGENLYFQG QVWSSGVFELKLQEFVNKKGLLGNRNCCRGGAGPPPCACRTFFRVCLKHYQASV SPEPPCTYGSAVTPVLGVDSFSLPDGGGADSAFSNPIRFPFGFTWPGTFSLIIEALHT DSPDDLATENPERLISRLATQRHLTVGEEWSQDLHSSGRTDLKYSYR FVCDEHYYGEGCSVFCRPRDDAFGHFTCGERGEKVCNPGWKGPYCT EPICLPGCDEQHGFCDKPGECKCRVGWQGRYCD ECIRYPGCLHGTCQQPWQCNCQEGWGGLFCN QDLNYCTHHKPCKNGATCTNTGQGSYTCSCRPGYTGATCEL

Expression and on-column refolding of DLL1 ligands – The purification of DLL1 ligand monomers and tandem repeats was optimized using Ni-NTA on column refolding

(although successful refolding was seen using dialysis). 1μL of plasmid containing the

DLL1 ligand gene in the pHLIC expression vector was transformed into 40μL of Origami

B cells by electroporation. The cells were recovered for 1 hour with 1mL of 2YT media at

37˚C with 200 RPM agitation. A tertiary streak of 50μL recovered cell culture was

150 performed on agar plates containing ampicillin and grown overnight at 37˚C. The next day, one colony was picked and grown overnight in a 25mL 2YT culture supplemented with ampicillin (200 RPM). The saturated seed culture was then used to inoculate 2 Liters of

2YT media supplemented with ampicillin and grown at 200 RPM, 37˚C. The culture was grown until optical density of the media reached OD = 0.6 at 600nM wavelength. The culture was then induced with 200μL of IPTG and grown for an additional 4 hours at 200

RPM, 37˚C. Cell cultures were then pelleted at 6000 RPM for 10 minutes. Pellets were then purified or frozen at -80˚C for later purification.

Pellets were re-suspended in 25 mL of Lysis Buffer (50 mM Tris pH=8, 100 mM NaCl).

The suspension was supplemented with 75 μL 2 M MgCl2, 150 μL 150 mM CaCl2, 5 μL

Dnase, 5μL Rnase, and 300μL 10% Triton X-100. Finally, 100mM PMSF protease inhibitor was added to prevent degradation of proteins. The cell suspension was then lysed using an EmulsiFlex with three passes through the system. Cells were then spun down at

15,000 RPM for 1 hour and the supernatant was discarded.

The insoluble cell pellet was re-suspended in 25 mL of unfolding buffer (7M Guanidine

HCl, 50 mM Tris, 50 mM NaCl, and 1 mM DTT. This was incubated at 200 RPM, 37 ˚C overnight. The unfolded solution was then spun down at 15000 RPM for 1 hour and 2mL of Ni-NTA resin slurry (50% etOH solution) was incubated with the supernatant for 2 hours at 4 ˚C. The solution was then lightly pelleted at 1000 RPM for 15 minutes and the supernatant was discarded. The pelleted resin was then suspended in 5 mL of 2 M

151

Guanidine HCl, 50 mM Tris pH=8, 50 mM NaCl, 1 mM DTT, and 30 mM Imidazole. The suspension was transferred onto a fritted column and varying concentrations of Guanidine

HCl buffer, including 1.5, 1, 0.5, 0.25, 0.125 M Guanidine HCl was washed over the column. The refolded DLL1 ligand was then eluted off the column with a 10% Glycerol,

50 mM Tris pH=8, 50 mM NaCl, 1 mM DTT, and 300 mM imidazole elution buffer. The protein was then digested with TEV protease and dialyzed into PBS pH=7.4, 10% Glycerol, and 0.5mM DTT for cell and in vivo mouse studies.

Design and construction of lead mouse DSL-EGF1-EGF2 3-4x tandem constructs – The tandem constructs 3x and 4x of the mouse monomer DSL-EGF1-EGF2 were created by using PCR and a removable BsaI restriction digest site at the 3’ end of one PCR product containing two monomer repeats and the 5’ end of another PCR product containing one or two monomer repeats. Ligation of these two products created a 3 and 4x tandem construct separated by GSSGSSG linkers. This was then ligated into the expression vector pHLIC with the restriction sites NdeI and BamHI.

Design and construction of mouse DLL1 ligand 5-8x tandem constructs – A gene was designed which contained eight repeats of the functional monomer DSL-EGF1-EGF2. This was used to further clone smaller tandem repeats such as 5x, 6x, and 7x tandems. This was done by digesting the plasmid with either XhoI for 7x tandem DLL1, BsrGI for 6x tandem

DLL1, or XmaI for 5x Tandem DLLI. The digested vector was gel purified from the digested pop-out and ligated to itself. The resulting plasmid contained the desired number

152 of tandem repeats. These were then sequenced by multiple primers designed to anneal to the uniquely coded linker sequences (GSSGSSG) within the gene Figure 49.

Figure 49: Gene design of large DLL1 tandem constructs. Restriction site allow for cloning of smaller repeat constructs. Gly-Ser linkers are highlighted in orange and were designed with varying codon usage for future sequencing. Variation in codon usage (highlighted in blue) were also designed at the 3’ end of repeats 5, 6, and 7 for potential PCR annealing sites.

Eight repeats of the DSL-EGF1-EGF2 mouse construct were cloned into the expression vector pHLIC using NdeI and BamHI restriction sites. Altering codons usage for specific sequencing primer design was used to code for the GSSGSSG linkers (Orange).

Randomized codon usage at the 3’ end of repeats 5, 6, and 7 were used for potential PCR applications. Seven restriction sites were added to specific areas of each gene for the ability to remove whole repeats and construct smaller tandem repeats from the larger 8x tandem repeat.

The following sequence entire mouse 8x tandem DSL-EGF1-EGF2 Gene:

153

CATATGGCCGCACATCATCATCACCACCATGGTGGTGAAAACCTGTATTTTCA GGGCTTTGTGTGCGATGAGCATTATTACGGTGAAGGCTGTAGCGTGTTTTGTC GACCGCGTGATGACGCATTTGGCCATTTTACCTGCGGTGACCGCGGCGAGAA AATGTGCGATCCGGGCTGGAAAGGTCAGTATTGTACAGACCCTATTTGTCTGC CGGGCTGTGACGATCAGCACGGCTACTGTGACAAGCCGGGCGAATGTAAATG TCGCGTTGGCTGGCAGGGTCGCTATTGTGATGAATGCATCCGCTACCCGGGCT GCCTGCACGGTACCTGTCAGCAGCCGTGGCAATGCAATTGTCAAGAGGGCTG GGGAGGTTTATTTTGTAATCAAGGCTCGAGCGGTTCATCGGGATTTGTTTGCG ATGAACACTATTATGGTGAAGGTTGTAGTGTGTTTTGCCGCCCGCGCGACGAC GCTTTTGGTCATTTTACATGCGGCGATCGCGGTGAAAAAATGTGCGATCCGG GTTGGAAGGGCCAGTACTGTACCGATCCGATTTGTCTGCCGGGTTGCGATGAT CAGCATGGCTATTGTGACAAGCCGGGTGAGTGCAAGTGTCGTGTGGGTTGGC AGGGCCGCTACTGTGACGAGTGCATTCGCTATCCGGGCTGTTTACACGGCAC CTGTCAGCAGCCGTGGCAATGTAATTGTCAAGAAGGCTGGGGAGGTTTATTT TGTAATCAAGGCTCGAGTGGAAGCTCAGGCTTTGTGTGTGATGAGCATTATTA CGGCGAAGGCTGCAGCGTTTTCTGTCGCCCGCGCGATGACGCATTTGGCCAC TTCACATGTGGCGATCGCGGCGAAAAGATGTGCGACCCTGGCTGGAAAGGCC AATATTGTACAGATCCTATCTGCCTGCCGGGTTGTGACGATCAACACGGTTAT TGCGACAAGCCGGGTGAATGCAAGTGTCGTGTGGGCTGGCAGGGCCGTTATT GTGACGAATGTATTCGCTATCCGGGTTGCTTACACGGCACATGTCAACAGCC GTGGCAGTGTAATTGCCAGGAAGGTTGGGGAGGTTTATTTTGTAATCAAGGT TCTTCCGGATCGAGTGGCTTCGTGTGTGACGAGCACTACTATGGCGAGGGTTG TAGCGTTTTCTGTCGCCCGCGCGACGACGCTTTTGGCCACTTTACCTGTGGCG ATCGCGGTGAGAAAATGTGTGATCCTGGTTGGAAGGGCCAGTACTGCACAGA CCCGATTTGTCTGCCGGGTTGCGATGATCAGCATGGCTACTGCGACAAACCG GGTGAATGCAAATGTCGTGTGGGCTGGCAGGGCCGCTATTGCGACGAGTGTA TTCGCTACCCGGGTTGCTTACATGGCACATGCCAACAGCCTTGGCAATGTAAT TGTCAGGAAGGCTGGGGAGGTTTATTTTGTAATCAAGGCAGTTCAGGGTCTA GTGGTTTTGTGTGCGACGAACATTACTACGGCGAGGGCTGCAGCGTGTTCTGT CGACCTCGTGACGACGCTTTTGGCCACTTCACCTGCGGCGATCGCGGCGAGA AGATGTGCGATCCTGGCTGGAAAGGCCAGTACTGCACAGATCCGATTTGCCT GCCGGGCTGTGATGACCAACATGGCTATTGTGATAAGCCGGGCGAGTGTAAG TGCCGTGTTGGTTGGCAAGGCCGCTATTGCGATGAGTGCATCCGTTATCCTGG CTGTCTGCATGGCACTTGTCAGCAACCGTGGCAATGCAACTGCCAAGAGGGT TGGGGTGGGTTGTTCTGCAATCAGGGATCTTCGGGGTCCTCTGGTTTTGTGTG TGACGAACACTATTACGGCGAGGGCTGCAGTGTGTTCTGCCGCCCTCGTGAC GACGCCTTCGGTCATTTCACCTGTGGCGATCGTGGCGAAAAGATGTGTGATCC TGGTTGGAAGGGTCAATACTGTACCGACCCGATCTGTCTGCCTGGCTGCGAC GACCAGCATGGTTACTGTGACAAACCGGGCGAGTGTAAATGCCGTGTTGGTT GGCAAGGCCGCTACTGTGATGAGTGCATTCGCTATCCTGGTTGCCTGCACGGT ACCTGCCAGCAACCTTGGCAATGCAACTGCCAGGAAGGTTGGGGGGGCCTGT TCTGCAACCAGGGTAGCTCTGGCTCATCCGGGTTCGTTTGCGACGAGCATTAT 154

TACGGCGAAGGCTGTAGTGTTTTCTGCCGTCCTCGTGACGACGCATTTGGCCA TTTCACCTGTGGTGACCGCGGCGAAAAAATGTGTGACCCTGGCTGGAAGGGC CAATATTGCACTGATCCGATTTGTCTGCCGGGTTGTGACGATCAACACGGCTA CTGCGATAAGCCGGGCGAATGTAAGTGCCGTGTGGGTTGGCAAGGCCGCTAT TGTGATGAATGCATTCGCTATCCGGGCTGTCTGCACGGCACTTGCCAGCAGCC GTGGCAATGTAATTGCCAAGAAGGTTGGGGCGGACTCTTTTGTAACCAAGGA TCATCAGGTAGTAGCGGCTTTGTTTGTGACGAGCACTACTATGGTGAAGGCTG CAGCGTTTTTTGTCGTCCGCGCGATGATGCCTTTGGTCATTTCACCTGTGGCG ACCGCGGCGAGAAAATGTGCGATCCTGGTTGGAAAGGCCAGTATTGCACCGA TCCGATCTGCCTGCCGGGCTGCGACGATCAACACGGCTATTGTGACAAACCG GGTGAATGCAAATGTCGTGTGGGTTGGCAGGGCCGCTACTGCGATGAGTGTA TTCGTTATCCTGGTTGCCTGCATGGCACATGTCAACAGCCGTGGCAGTGCAAT TGTCAGGAAGGTTGGGGCGGCCTGTTCTGCAACCAGTAAGGATCC

Evaluation of Notch activation (Dikov Lab) – Notch activation was assessed by the expression of Notch downstream targets Hes1 and Hey1 using quantitative real time PCR

(qRT-PCR) and Western blot and accumulation of intracellular domain of Notch (ICD) by

Western blot.

3T3 or LLC cells were plated at 0.5-2 x 106 per well in 12- or 6-well RNA, EL4 cells were seeded at 2-4 x 106 cells. DLL1 constructs were added to cells at varying doses in the same volume of the construct buffer (PBS, 10% glycerol, 5 mM DTT). Buffer added at the same volume was used as a negative control. For positive control, clustered DLL1 was titered at concentrations between 0 and 2 g/ml (based on DLL1-Fc protein). When surface-bound constructs were used, they were absorbed on the plates by incubating their varying dilutions from 0.5 to 5 g/ml in PBS in cell culture plate overnight at 4 oC; plates the were washed with PBS and 10% BSA solution in PBS was then added for additional 2 hrs at room temperature. Plates were washed with PBS and cells seeded. Cells were cultured for 6-12 hrs for qRT-PCR, 16 hrs for Western blot analyses and harvested. 155

RNA was extracted with an RNeasy Mini kit and possible genomic DNA contamination was removed by on-column DNase digestion using the RNase–free DNase set (Qiagen;

Valencia, CA). cDNA was synthesized using SuperScript III Reverse Transcriptase kit

(Invitrogen, Grand Island, NY). cDNA, iQ SYBR green supermix (Bio-Rad, Hercules,

CA) and the gene-specific primers were used in 20 μl PCR reactions as recommended by the manufacturer and described previously(165). The sequence of the primers used is listed below. Amplification of endogenous β-actin or GAPDH was used as internal controls.

Hes1 forward: GCC AAT TTG CCT TTC TCA TC

Hes1 reverse: AGC CAC TGG AAG GTG ACA CT

Hey1 forward: CTC TCA GCC TTC CCC TTT TC

Hey1 reverse: ATC TCT GTC CCC CAA GGT CT

Western blot analysis (Dikov Lab) – Cells were lysed in a lysis buffer containing 20 mM

HEPES, 150 mM NaCl, 10% glycerol, 1% Triton X-100, 1 mM EGTA, and 1.5 mM MgCl2 with set of inhibitors, as described previously(166). Equal amounts of protein were mixed with SDS sample buffer and separated by 7.5 or 10% SDS-PAGE, and transferred to PVDF membrane (Amersham Biosciences, Piscataway, NJ). The following antibodies were used

156 for detection: Notch1 (Cell Signaling, MA) recognizing epitopes in intracellular domain

(ICD) of Notch; Hes1 (Cell Signaling Technology, MA).

Modulation of cytokine expression analysis by ELISPOT assay (Dikov Lab) – The ability of DLL1 constructs to modulate cytokine expression in splenocytes was evaluated by

ELISPOT assay. Splenocytes from normal mice were seeded at 2.5x105 cells per well in

IFN-γ ELISPOT plates (CTL, Shaker Heights, OH). Cells were stimulated with CD3/CD28 beads (Life Technologies, AS) at the ratio of 1:2 cells to beads for sub-optimal T cell activation. Varying doses of DLL1 constructs, clustered DLL1 or buffer control were added to cells, as described above. Splenocytes were cultured for 48 hrs and IFN-γ- producing cells were enumerated by ELISPOT assay according to the manufacturer’s protocol.

In vivo tumor growth study (Dikov Lab) – Female and male Balb/c, C57BL/6 mice (7 to 8- week-old) were purchased from The Jackson Laboratory. The animals were housed in pathogen-free units at the Vanderbilt University School of Medicine, in compliance with the Institutional Animal Care and Use Committee regulations. To induce tumor, mice were inoculated subcutaneously (s.c.) in flank with 0.3106 D459 or LLC cells, as described previously (123,167). Tumor volume was measured with calipers and tumor tissues were weighed at the endpoint of the experiments.

157

Mouse DLL1-Fc fusion protein is composed of the extracellular domain of mouse or human DLL1 and the Fc part of mouse IgG2A or human IgG1, respectively. To form DLL1 clusters, DLL1-Fc, biotinylated anti-IgG antibodies, and NeutrAvidin (Pierce, Rockford,

IL) were mixed at a molar ratio of 1:4:10 in PBS, as described earlier (123,168). As a control in all applications, Fc fragment of mouse IgG2 (Sigma-Aldrich, St. Louis, MO) was used instead of DLL1-Fc. Mouse DLL1-Fc and biotinylated donkey anti-mouse IgG antibodies were from R&D Systems (Minneapolis, MN).

Tumor-bearing mice received clustered DLL1 at doses of 0.15 μg/kg (4 μg per injection) of DLL1-Fc protein in 100 l of PBS intraperitoneally (i.p.) every other day for 3 weeks.

The control group received control clusters with Fc fragments instead of DLL1-Fc protein.

Twice higher doses of clustered DLL1 were used in some experiments with similar results suggesting dose saturation of the clustered DLL1 effects.

Data were analyzed using the GraphPad Prism 4.0 software (GraphPad Software Inc., San

Diego, CA) and presented as mean ± SEM. Comparisons between two groups were performed using two-tailed unpaired t tests. Values were considered statistically significant when P was less than 0.05.

158

Bibliography

1. Anfinsen, C. B., Redfield, R. R., Choate, W. L., Page, J., and Carroll, W. R. (1954) J Biol Chem 207, 201-210 2. Dill, K. A., Ozkan, S. B., Shell, M. S., and Weikl, T. R. (2008) Annu Rev Biophys 37, 289-316 3. Onuchic, J. N., Wolynes, P. G., Luthey-Schulten, Z., and Socci, N. D. (1995) P Natl Acad Sci USA 92, 3626-3630 4. Lutz, S. (2010) Curr Opin Biotechnol 21, 734-743 5. Cherry, J. R., and Fidantsef, A. L. (2003) Curr Opin Biotechnol 14, 438-443 6. Breinbauer, R., Vetter, I. R., and Waldmann, H. (2002) Angew Chem Int Ed Engl 41, 2879-2890 7. Pedersen, H., Holder, S., Sutherlin, D. P., Schwitter, U., King, D. S., and Schultz, P. G. (1998) P Natl Acad Sci USA 95, 10523-10528 8. Giver, L., Gershenson, A., Freskgard, P. O., and Arnold, F. H. (1998) P Natl Acad Sci USA 95, 12809-12813 9. Boder, E. T., Midelfort, K. S., and Wittrup, K. D. (2000) P Natl Acad Sci USA 97, 10701-10705 10. Chianella, I., Lotierzo, M., Piletsky, S. A., Tothill, I. E., Chen, B., Karim, K., and Turner, A. P. (2002) Anal Chem 74, 1288-1293 11. Vallee-Belisle, A., and Plaxco, K. W. (2010) Curr Opin Struct Biol 20, 518-526 12. Lehmann, M., and Wyss, M. (2001) Curr Opin Biotechnol 12, 371-375 13. Chica, R. A., Doucet, N., and Pelletier, J. N. (2005) Curr Opin Biotechnol 16, 378-384 14. Lippow, S. M., and Tidor, B. (2007) Curr Opin Biotechnol 18, 305-311 15. Clarkson, T. W., Magos, L., and Myers, G. J. (2003) N Engl J Med 349, 1731- 1737 16. Harper, D. M., Franco, E. L., Wheeler, C. M., Moscicki, A. B., Romanowski, B., Roteli-Martins, C. M., Jenkins, D., Schuind, A., Costa Clemens, S. A., and Dubin, G. (2006) Lancet 367, 1247-1255 17. Titball, R. W., and Williamson, E. D. (2001) Vaccine 19, 4175-4184 18. Hanahan, D., and Weinberg, R. A. (2000) Cell 100, 57-70 19. Hanahan, D., and Weinberg, R. A. (2011) Cell 144, 646-674 20. Dearnaley, D. P., Khoo, V. S., Norman, A. R., Meyer, L., Nahum, A., Tait, D., Yarnold, J., and Horwich, A. (1999) Lancet 353, 267-272 21. Nie, X., Liu, D., Li, Q., and Bai, C. (2013) J Geriatr Oncol 4, 334-339 22. Weiss, L. (2000) Cancer Metastasis Rev 19, I-XI, 193-383 23. Mukherjee, S. (2010) 24. Loeb, L. A., Loeb, K. R., and Anderson, J. P. (2003) P Natl Acad Sci USA 100, 776-781 25. Dunn, G. P., Old, L. J., and Schreiber, R. D. (2004) Annu Rev Immunol 22, 329- 360 159

26. Finn, O. J. (2008) N Engl J Med 358, 2704-2715 27. Coussens, L. M., and Werb, Z. (2002) Nature 420, 860-867 28. Houston, A., Bennett, M. W., O'Sullivan, G. C., Shanahan, F., and O'Connell, J. (2003) Br J Cancer 89, 1345-1351 29. Kufe, D. W. (2009) Nat Rev Cancer 9, 874-885 30. Lasek, W., Zagozdzon, R., and Jakobisiak, M. (2014) Cancer Immunol Immunother 63, 419-435 31. Zaidi, M. R., and Merlino, G. (2011) Clinical cancer research : an official journal of the American Association for Cancer Research 17, 6118-6124 32. Todaro, M., Alea, M. P., Di Stefano, A. B., Cammareri, P., Vermeulen, L., Iovino, F., Tripodo, C., Russo, A., Gulotta, G., Medema, J. P., and Stassi, G. (2007) Cell Stem Cell 1, 389-402 33. Melero, I., Hervas-Stubbs, S., Glennie, M., Pardoll, D. M., and Chen, L. (2007) Nat Rev Cancer 7, 95-106 34. Klebanoff, C. A., Rosenberg, S. A., and Restifo, N. P. (2016) Nat Med 22, 26-36 35. Disis, M. L., Schiffman, K., Guthrie, K., Salazar, L. G., Knutson, K. L., Goodell, V., dela Rosa, C., and Cheever, M. A. (2004) J Clin Oncol 22, 1916-1925 36. Butts, C., Murray, N., Maksymiuk, A., Goss, G., Marshall, E., Soulieres, D., Cormier, Y., Ellis, P., Price, A., Sawhney, R., Davis, M., Mansi, J., Smith, C., Vergidis, D., MacNeil, M., and Palmer, M. (2005) J Clin Oncol 23, 6674-6681 37. Coulie, P. G., Van den Eynde, B. J., van der Bruggen, P., and Boon, T. (2014) Nat Rev Cancer 14, 135-146 38. Mellman, I., Coukos, G., and Dranoff, G. (2011) Nature 480, 480-489 39. Palucka, K., and Banchereau, J. (2012) Nat Rev Cancer 12, 265-277 40. Vogel, C. L., Cobleigh, M. A., Tripathy, D., Gutheil, J. C., Harris, L. N., Fehrenbacher, L., Slamon, D. J., Murphy, M., Novotny, W. F., Burchmore, M., Shak, S., Stewart, S. J., and Press, M. (2002) J Clin Oncol 20, 719-726 41. Aggarwal, S. (2007) Nat Biotechnol 25, 1097-1104 42. Sassoon, I., and Blanc, V. (2013) Methods Mol Biol 1045, 1-27 43. Zolot, R. S., Basu, S., and Million, R. P. (2013) Nat Rev Drug Discov 12, 259-260 44. Ducry, L. (2012) Methods Mol Biol 899, 489-497 45. Ducry, L., and Stump, B. (2010) Bioconjug Chem 21, 5-13 46. Hollingsworth, M. A., and Swanson, B. J. (2004) Nat Rev Cancer 4, 45-60 47. Duraisamy, S., Kufe, T., Ramasamy, S., and Kufe, D. (2007) Int J Oncol 31, 671- 677 48. Cummings, R. D., and Pierce, J. M. (2014) Chem Biol 21, 1-15 49. Christiansen, M. N., Chik, J., Lee, L., Anugraham, M., Abrahams, J. L., and Packer, N. H. (2014) Proteomics 14, 525-546 50. Szymanski, C. M., and Wren, B. W. (2005) Nat Rev Microbiol 3, 225-237 51. Johnson, V. G., Schlom, J., Paterson, A. J., Bennett, J., Magnani, J. L., and Colcher, D. (1986) Cancer Res 46, 850-857 52. Julien, S., Videira, P. A., and Delannoy, P. (2012) Biomolecules 2, 435-466

160

53. Ragupathi, G., Howard, L., Cappello, S., Koganty, R. R., Qiu, D., Longenecker, B. M., Reddish, M. A., Lloyd, K. O., and Livingston, P. O. (1999) Cancer Immunol Immunother 48, 1-8 54. Kontermann, R. E. (2011) Curr Opin Biotechnol 22, 868-876 55. Miles, D., Roche, H., Martin, M., Perren, T. J., Cameron, D. A., Glaspy, J., Dodwell, D., Parker, J., Mayordomo, J., Tres, A., Murray, J. L., and Ibrahim, N. K. (2011) Oncologist 16, 1092-1100 56. Liu, J. K. (2014) Ann Med Surg (Lond) 3, 113-116 57. Cianfriglia, M., Armellini, D., Massone, A., and Mariani, M. (1983) Hybridoma 2, 451-457 58. Leavy, O. (2010) Nat Rev Immunol 10, 297 59. Mirick, G. R., Bradt, B. M., Denardo, S. J., and Denardo, G. L. (2004) Q J Nucl Med Mol Imaging 48, 251-257 60. Verma, R., Boleti, E., and George, A. J. (1998) J Immunol Methods 216, 165-181 61. Bird, R. E., and Walker, B. W. (1991) Trends Biotechnol 9, 132-137 62. Lilie, H., Schwarz, E., and Rudolph, R. (1998) Curr Opin Biotechnol 9, 497-501 63. Nelson, A. L. (2010) Mabs-Austin 2, 77-83 64. Horwitz, A. H., Chang, C. P., Better, M., Hellstrom, K. E., and Robinson, R. R. (1988) P Natl Acad Sci USA 85, 8678-8682 65. Albert, H., Collin, M., Dudziak, D., Ravetch, J. V., and Nimmerjahn, F. (2008) P Natl Acad Sci USA 105, 15005-15009 66. Hasemann, C. A., and Capra, J. D. (1990) P Natl Acad Sci USA 87, 3942-3946 67. Almagro, J. C., and Fransson, J. (2008) Front Biosci 13, 1619-1633 68. Jones, P. T., Dear, P. H., Foote, J., Neuberger, M. S., and Winter, G. (1986) Nature 321, 522-525 69. Studnicka, G. M., Soares, S., Better, M., Williams, R. E., Nadell, R., and Horwitz, A. H. (1994) Protein Eng 7, 805-814 70. Clementi, N., Mancini, N., Solforosi, L., Castelli, M., Clementi, M., and Burioni, R. (2012) Int J Mol Sci 13, 8273-8292 71. Wu, C. H., Liu, I. J., Lu, R. M., and Wu, H. C. (2016) J Biomed Sci 23, 8 72. Tabata, N., Sakuma, Y., Honda, Y., Doi, N., Takashima, H., Miyamoto-Sato, E., and Yanagawa, H. (2009) Nucleic Acids Res 37, e64 73. Hanes, J., Schaffitzel, C., Knappik, A., and Pluckthun, A. (2000) Nat Biotechnol 18, 1287-1292 74. Chao, G., Lau, W. L., Hackel, B. J., Sazinsky, S. L., Lippow, S. M., and Wittrup, K. D. (2006) Nat Protoc 1, 755-768 75. Daugherty, P. S. (2007) Curr Opin Struct Biol 17, 474-480 76. Konning, D., Zielonka, S., Grzeschik, J., Empting, M., Valldorf, B., Krah, S., Schroter, C., Sellmann, C., Hock, B., and Kolmar, H. (2017) Curr Opin Struct Biol 45, 10-16 77. Kontermann, R. E., and Brinkmann, U. (2015) Drug Discov Today 20, 838-847 78. Colcher, D., Hand, P. H., Nuti, M., and Schlom, J. (1981) P Natl Acad Sci USA 78, 3199-3203

161

79. Aymard-Henry, M., Coleman, M. T., Dowdle, W. R., Laver, W. G., Schild, G. C., and Webster, R. G. (1973) Bull World Health Organ 48, 199-202 80. Muraro, R., Kuroki, M., Wunderlich, D., Poole, D. J., Colcher, D., Thor, A., Greiner, J. W., Simpson, J. F., Molinolo, A., Noguchi, P., and et al. (1988) Cancer Res 48, 4588-4596 81. Moch, H., Oberholzer, M., Christen, H., Buser, M., Dalquen, P., Wegmann, W., and Gudat, F. (1993) Virchows Arch A Pathol Anat Histopathol 423, 493-496 82. Ordonez, N. G. (2003) Am J Surg Pathol 27, 1031-1051 83. Meredith, R., Shen, S., Macey, D., Khazaeli, M. B., Carey, D., Robert, F., and LoBuglio, A. (2003) Cancer Biother Radiopharm 18, 393-404 84. Meredith, R. F., Khazaeli, M. B., Macey, D. J., Grizzle, W. E., Mayo, M., Schlom, J., Russell, C. D., and LoBuglio, A. F. (1999) Clinical cancer research : an official journal of the American Association for Cancer Research 5, 3254s- 3258s 85. Kashmiri, S. V. S., Shu, L., Padlan, E. A., Milenic, D. E., Schlom, J., and Hand, P. H. (1995) Hybridoma 14, 461-473 86. Xiao, J., Horst, S., Hinkle, G., Cao, X., Kocak, E., Fang, J., Young, D., Khazaeli, M., Agnese, D., Sun, D., and Martin, E., Jr. (2005) Cancer Biother Radiopharm 20, 16-26 87. Kashmiri, S. V., De Pascalis, R., Gonzales, N. R., and Schlom, J. (2005) Methods 36, 25-34 88. Yoon, S. O., Lee, T. S., Kim, S. J., Jang, M. H., Kang, Y. J., Park, J. H., Kim, K. S., Lee, H. S., Ryu, C. J., Gonzales, N. R., Kashmiri, S. V., Lim, S. M., Choi, C. W., and Hong, H. J. (2006) J Biol Chem 281, 6985-6992 89. Baneyx, F. (1999) Curr Opin Biotechnol 10, 411-421 90. Weiner, L. M., Surana, R., and Wang, S. (2010) Nat Rev Immunol 10, 317-327 91. Nelson, A. L., and Reichert, J. M. (2009) Nat Biotechnol 27, 331-337 92. Rousseaux, J., Rousseaux-Prevost, R., and Bazin, H. (1983) J Immunol Methods 64, 141-146 93. Buchner, J., and Rudolph, R. (1991) Biotechnology (N Y) 9, 157-162 94. Jeong, K. J., Jang, S. H., and Velmurugan, N. (2011) Biotechnol J 6, 16-27 95. Pluckthun, A. (1991) Bio-Technol 9, 545-551 96. Glockshuber, R., Malia, M., Pfitzinger, I., and Pluckthun, A. (1990) Biochemistry-Us 29, 1362-1367 97. Brinkmann, U., Reiter, Y., Jung, S. H., Lee, B., and Pastan, I. (1993) P Natl Acad Sci USA 90, 7538-7542 98. Gil, D., and Schrum, A. G. (2013) Adv Biosci Biotechnol 4, 73-84 99. Young, N. M., MacKenzie, C. R., Narang, S. A., Oomen, R. P., and Baenziger, J. E. (1995) FEBS Lett 377, 135-139 100. Kortt, A. A., Dolezal, O., Power, B. E., and Hudson, P. J. (2001) Biomol Eng 18, 95-108 101. Kortt, A. A., Lah, M., Oddie, G. W., Gruen, C. L., Burns, J. E., Pearce, L. A., Atwell, J. L., McCoy, A. J., Howlett, G. J., Metzger, D. W., Webster, R. G., and Hudson, P. J. (1997) Protein Eng 10, 423-433 162

102. Dolezal, O., Pearce, L. A., Lawrence, L. J., McCoy, A. J., Hudson, P. J., and Kortt, A. A. (2000) Protein Eng 13, 565-574 103. Pluckthun, A. (1991) Biotechnology (N Y) 9, 545-551 104. Bedzyk, W. D., Weidner, K. M., Denzin, L. K., Johnson, L. S., Hardman, K. D., Pantoliano, M. W., Asel, E. D., and Voss, E. W., Jr. (1990) J Biol Chem 265, 18615-18620 105. Pantoliano, M. W., Bird, R. E., Johnson, S., Asel, E. D., Dodd, S. W., Wood, J. F., and Hardman, K. D. (1991) Biochemistry-Us 30, 10117-10125 106. Whitlow, M., Filpula, D., Rollence, M. L., Feng, S. L., and Wood, J. F. (1994) Protein Eng 7, 1017-1026 107. Whitlow, M., Bell, B. A., Feng, S. L., Filpula, D., Hardman, K. D., Hubert, S. L., Rollence, M. L., Wood, J. F., Schott, M. E., Milenic, D. E., Yokota, T., and Schlom, J. (1993) Protein Eng 6, 989-995 108. Gruber, M., Schodin, B. A., Wilson, E. R., and Kranz, D. M. (1994) J Immunol 152, 5368-5374 109. Kriangkum, J., Xu, B., Nagata, L. P., Fulton, R. E., and Suresh, M. R. (2001) Biomol Eng 18, 31-40 110. Brocks, J. J., and Butterfield, N. J. (2009) Nature 457, 672-673 111. Poulson, D. F. (1937) P Natl Acad Sci USA 23, 133-137 112. Artavanis-Tsakonas, S., Rand, M. D., and Lake, R. J. (1999) Science 284, 770- 776 113. Mumm, J. S., and Kopan, R. (2000) Dev Biol 228, 151-165 114. Lobry, C., Oh, P., Mansour, M. R., Look, A. T., and Aifantis, I. (2014) Blood 123, 2451-2459 115. Kopan, R., and Ilagan, M. X. (2009) Cell 137, 216-233 116. Benedito, R., Roca, C., Sorensen, I., Adams, S., Gossler, A., Fruttiger, M., and Adams, R. H. (2009) Cell 137, 1124-1135 117. Hu, Y. Y., Zheng, M. H., Zhang, R., Liang, Y. M., and Han, H. (2012) Adv Exp Med Biol 727, 186-198 118. Radtke, F., Wilson, A., Stark, G., Bauer, M., van Meerwijk, J., MacDonald, H. R., and Aguet, M. (1999) Immunity 10, 547-558 119. Wilson, A., MacDonald, H. R., and Radtke, F. (2001) J Exp Med 194, 1003-1012 120. de La Coste, A., Six, E., Fazilleau, N., Mascarell, L., Legrand, N., Mailhe, M. P., Cumano, A., Laabi, Y., and Freitas, A. A. (2005) J Immunol 174, 2730-2737 121. Dallas, M. H., Varnum-Finney, B., Delaney, C., Kato, K., and Bernstein, I. D. (2005) J Exp Med 201, 1361-1366 122. Wang, H. Y., and Wang, R. F. (2007) Curr Opin Immunol 19, 217-223 123. Huang, Y., Lin, L., Shanker, A., Malhotra, A., Yang, L., Dikov, M. M., and Carbone, D. P. (2011) Cancer Res 71, 6122-6131 124. Siegel, R. L., Miller, K. D., and Jemal, A. (2016) CA Cancer J Clin 66, 7-30 125. Mammen, J. M., James, L. E., Molloy, M., Williams, A., Wray, C. J., and Sussman, J. J. (2007) Am J Surg 194, 349-354

163

126. Shankar, L. K., Hoffman, J. M., Bacharach, S., Graham, M. M., Karp, J., Lammertsma, A. A., Larson, S., Mankoff, D. A., Siegel, B. A., den Abbeele, A., Yap, J., and Sullivan, D. (2006) Journal of Nuclear Medicine 47, 1059-1066 127. Todorovska, A., Roovers, R. C., Dolezal, O., Kortt, A. A., Hoogenboom, H. R., and Hudson, P. J. (2001) J Immunol Methods 248, 47-66 128. Moran, N. (2011) Nat Biotechnol 29, 5-6 129. Chapman, A. P., Antoniw, P., Spitali, M., West, S., Stephens, S., and King, D. J. (1999) Nat Biotechnol 17, 780-783 130. Brannon-Peppas, L., and Blanchette, J. O. (2004) Adv Drug Deliver Rev 56, 1649- 1659 131. Alfthan, K., Takkinen, K., Sizmann, D., Soderlund, H., and Terri, T. T. (1995) Protein Eng 8, 725-731 132. Tang, Y., Jiang, N., Parakh, C., and Hilvert, D. (1996) Journal of Biological Chemistry 271, 15682-15686 133. Cuesta, A. M., Sanchez-Martin, D., Blanco-Toribio, A., Villate, M., Enciso- Alvarez, K., Alvarez-Cienfuegos, A., Sainz-Pastor, N., Sanz, L., Blanco, F. J., and Alvarez-Vallina, L. (2012) Mabs-Austin 4 134. Atwell, J. L., Breheney, K. A., Lawrence, L. J., McCoy, A. J., Kortt, A. A., and Hudson, P. J. (1999) Protein Eng 12, 597-604 135. Huston, J. S., Mudgetthunter, M., Tai, M. S., Mccartney, J., Warren, F., Haber, E., and Oppermann, H. (1991) Method Enzymol 203, 46-& 136. Pavlinkova, G., Beresford, G. W., Booth, B. J. M., Batra, S. K., and Colcher, D. (1999) Journal of Nuclear Medicine 40, 1536-1546 137. Durani, V., Sullivan, B. J., and Magliery, T. J. (2012) Protein Expr Purif 85, 9-17 138. Worn, A., and Pluckthun, A. (2001) J Mol Biol 305, 989-1010 139. Holliger, P., Prospero, T., and Winter, G. (1993) P Natl Acad Sci USA 90, 6444- 6448 140. Dorai, H., Mccartney, J. E., Hudziak, R. M., Tai, M. S., Laminet, A. A., Houston, L. L., Huston, J. S., and Oppermann, H. (1994) Bio-Technol 12, 890-897 141. Mittal, S., Kaur, H., Gautam, N., and Mantha, A. K. (2017) Biosens Bioelectron 88, 217-231 142. Thor, A., Ohuchi, N., Szpak, C. A., Johnston, W. W., and Schlom, J. (1986) Cancer Res 46, 3118-3124 143. Ponnusamy, M. P., Venkatraman, G., Singh, A. P., Chauhan, S. C., Johansson, S. L., Jain, M., Smith, L., Davis, J. S., Remmenga, S. W., and Batra, S. K. (2007) Cancer Lett 251, 247-257 144. Goel, A., Baranowska-Kortylewicz, J., Hinrichs, S. H., Wisecarver, J., Pavlinkova, G., Augustine, S., Colcher, D., Booth, B. J., and Batra, S. K. (2001) J Nucl Med 42, 1519-1527 145. Goel, A., Colcher, D., Baranowska-Kortylewicz, J., Augustine, S., Booth, B. J., Pavlinkova, G., and Batra, S. K. (2000) Cancer Res 60, 6964-6971 146. Hjerpe, A., Ascoli, V., Bedrossian, C. W., Boon, M. E., Creaney, J., Davidson, B., Dejmek, A., Dobra, K., Fassina, A., Field, A., Firat, P., Kamei, T., Kobayashi, T., Michael, C. W., Onder, S., Segal, A., and Vielh, P. (2015) Acta Cytol 59, 2-16 164

147. Gong, L., Ding, H., Long, N. E., Sullivan, B. J., Martin, E. W., Jr., Magliery, T. J., and Tweedle, M. F. (2017) Mol Imaging Biol 148. Holliger, P., and Hudson, P. J. (2005) Nat Biotechnol 23, 1126-1136 149. Kenanova, V., Olafsen, T., Crow, D. M., Sundaresan, G., Subbarayan, M., Carter, N. H., Ikle, D. N., Yazaki, P. J., Chatziioannou, A. F., Gambhir, S. S., Williams, L. E., Shively, J. E., Colcher, D., Raubitschek, A. A., and Wu, A. M. (2005) Cancer Res 65, 622-631 150. Epivatianos, A., Poulopoulos, A., Kayavis, I., and Papanayotou, P. (2000) Oral Dis 6, 112-117 151. Thor, A., Itzkowitz, S. H., Schlom, J., Kim, Y. S., and Hanauer, S. (1989) Int J Cancer 43, 810-815 152. Kararli, T. T. (1995) Biopharm Drug Dispos 16, 351-380 153. Ro, J. Y., Grignon, D. J., Ayala, A. G., Fernandez, P. L., Ordonez, N. G., and Wishnow, K. I. (1990) Hum Pathol 21, 593-600 154. Lai, E. C. (2004) Development 131, 965-973 155. Borggrefe, T., and Oswald, F. (2009) Cell Mol Life Sci 66, 1631-1646 156. del Alamo, D., Rouault, H., and Schweisguth, F. (2011) Curr Biol 21, R40-47 157. Luca, V. C., Kim, B. C., Ge, C., Kakuda, S., Wu, D., Roein-Peikar, M., Haltiwanger, R. S., Zhu, C., Ha, T., and Garcia, K. C. (2017) Science 355, 1320- 1324 158. Cordle, J., Johnson, S., Tay, J. Z., Roversi, P., Wilkin, M. B., de Madrid, B. H., Shimizu, H., Jensen, S., Whiteman, P., Jin, B., Redfield, C., Baron, M., Lea, S. M., and Handford, P. A. (2008) Nat Struct Mol Biol 15, 849-857 159. Luca, V. C., Jude, K. M., Pierce, N. W., Nachury, M. V., Fischer, S., and Garcia, K. C. (2015) Science 347, 847-853 160. Pintar, A., Guarnaccia, C., Dhir, S., and Pongor, S. (2009) BMC Struct Biol 9, 43 161. Chen, R. (2012) Biotechnol Adv 30, 1102-1107 162. Takeuchi, H., Yu, H., Hao, H., Takeuchi, M., Ito, A., Li, H., and Haltiwanger, R. S. (2017) J Biol Chem 292, 15964-15973 163. Panin, V. M., Shao, L., Lei, L., Moloney, D. J., Irvine, K. D., and Haltiwanger, R. S. (2002) J Biol Chem 277, 29945-29952 164. Koch, U., Yuan, J. S., Harper, J. A., and Guidos, C. J. (2003) Semin Immunol 15, 99-106 165. Biktasova, A. K., Dudimah, D. F., Uzhachenko, R. V., Park, K., Akhter, A., Arasada, R. R., Evans, J. V., Novitskiy, S. V., Tchekneva, E. E., Carbone, D. P., Shanker, A., and Dikov, M. M. (2015) Cancer Res 75, 4728-4741 166. Atochina-Vasserman, E. N., Biktasova, A., Abramova, E., Cheng, D. S., Polosukhin, V. V., Tanjore, H., Takahashi, S., Sonoda, H., Foye, L., Venkov, C., Ryzhov, S. V., Novitskiy, S., Shlonimskaya, N., Ikeda, M., Blackwell, T. S., Lawson, W. E., Gow, A. J., Harris, R. C., Dikov, M. M., and Tchekneva, E. E. (2013) American journal of physiology. Renal physiology 304, F1295-1307 167. Novitskiy, S. V., Ryzhov, S., Zaynagetdinov, R., Goldstein, A. E., Huang, Y., Tikhomirov, O. Y., Blackburn, M. R., Biaggioni, I., Carbone, D. P., Feoktistov, I., and Dikov, M. M. (2008) Blood 112, 1822-1831 165

168. Heinzel, K., Benz, C., and Bleul, C. C. (2007) Proc Natl Acad Sci U S A 104, 8421-8426

166