Characterizing and engineering antibodies against the epidermal receptor

by

Ginger Chao

B.S. Chemical Engineering Rice University, 2002

Submitted to the Department of Chemical Engineering in partial fulfillment of the requirements for the degree of

DOCTOR OF PHILOSOPHY IN CHEMICAL ENGINEERING

at the

Massachusetts Institute of Technology February 2008

© 2007 Massachusetts Institute of Technology All rights reserved

Signature of Author: ______Department of Chemical Engineering October 15, 2007

Certified by: ______K. Dane Wittrup C.P. Dubbs Professor of Chemical Engineering and Bioengineering Thesis Supervisor

Accepted by: ______William M. Deen Professor of Chemical Engineering Chairman, Committee for Graduate Students Characterizing and engineering antibodies against the epidermal

by Ginger Chao

Submitted to the Department of Chemical Engineering on October 15, 2007 in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Chemical Engineering

ABSTRACT

Epidermal growth factor receptor (EGFR) signaling leads to cellular proliferation and migration, and thus EGFR dysregulation can significantly contribute to the survival of tumor cells. Aberrant EGFR signaling due to receptor overexpression, mutation, or autocrine ligand has been observed in a wide variety of malignancies, and antibody drugs which inhibit EGFR signaling have been developed. However, the epitopes of most EGFR antibodies have not been characterized, and the marginal efficacies of current antibodies underscore the need for improved therapeutics. In this thesis work, we have created a novel method of epitope mapping, which is the determination of antigen residues responsible for mediating an antibody-antigen interaction. In our technique, a library of random mutants of the EGFR antigen is displayed on the surface of yeast, and the library is combinatorially selected for loss of binding to the antibody being mapped. If a mutant shows loss of binding to an antibody, then that residue is a potential contact residue. In addition, we found that many mutants caused a global misfolding of the antigen, requiring the use of high-throughput sorting to remove the misfolded mutants. The development of our epitope mapping method using random mutagenesis and yeast surface display enabled the successful mapping of four different antibodies and three designed ankyrin repeat proteins binding to EGFR. In addition, we continued work on engineering novel antibodies against EGFR domains II and IV. Antibodies against these domains are hypothesized to directly inhibit receptor dimerization and subsequent activation, as opposed to traditional anti-EGFR antibodies which block ligand binding. To accomplish this, peptide mimics of EGFR loops were used as antigens; however, antibodies generated using both yeast surface display and rabbit monoclonal technology were peptide-specific, but did not bind to EGFR protein. Finally, we developed a mathematical model to describe equilibrium EGFR ligand binding and dimerization. Based on observations that polyclonal antibodies against EGFR domain II or IV eliminate high affinity EGF binding normally observed on the surface of cells, we hypothesized that preformed inactive dimers were the high affinity component. Our model incorporating this hypothesis successfully reproduced experimental data, resulting in characteristic concave-up Scatchard plots.

Thesis Supervisor: K. Dane Wittrup Title: C.P. Dubbs Professor of Chemical Engineering and Bioengineering

2

Dedicated to my parents, Jim and Grace, and my brother Eric.

Thanks for your love and support.

3 Acknowledgements

I would like to thank my research advisor Dane Wittrup, who has been an ideal thesis advisor these four years. He was always there with great advice and guidance while allowing me the perfect amount of independence and autonomy. I am grateful for the way which he helped my project evolve into several successful side projects and for his contributions to the development of my abilities as a researcher. I also want to thank my thesis committee, Doug Lauffenburger and Harvey Lodish, who were always willing to take time out of their extremely busy schedules to offer great ideas and their unique perspectives.

Dane has created a great research environment, and I would like to thank everyone I’ve worked with in the Wittrup lab. In particular, Jennifer Cochran originally trained me during my first summer, getting me started with my project and providing invaluable assistance and ideas. Other people who had been working on the EGFR project before I joined were extremely helpful: Yong-Sung Kim, Mark Olsen, and Alejandro Wolf-Yadlin. I learned a lot from the people who were in the lab before me: Katarina Midelfort, Jeffrey Swers, Bala Rao, Andy Yeung, Dave Colby, Shaun Lippow, and Stefan Zajic. Of course, I also want to thank Andy Rakestraw for his assistance in the lab and moral support outside the lab. Thanks to Steve Sazinsky, for his insightful discussions and very detailed protocols; Ben Hackel, who has helped me learn about library design and the intricacies of Microsoft Word; and Shanshan Howland, a great benchmate and inventor of many ingenious lab tricks. Thank you to Dasa Lipovsek, who provided me with excellent advice and got me a job! I also want to thank everyone else who has joined the lab over the years: Wai Lau, Andrea Piatesi, Greg Thurber, Annie Gai, Mike Schmidt, Margie Ackerman, Pankaj Karande, Eileen Higham, Kelly Davis, David Liu, and the new kids Jamie Spangler, Jordi Mata-Fink, and Chris Pirie. Thanks everyone for letting me borrow reagents, answering questions, and providing insight at group meetings!

I would like to thank my UROPs Heather Pressler and Nick Nguyen. Also thanks to everyone at the flow cytometry core facility: Glen Paradis, Mike Doire, Mike Jennings, Yong Xie, and Michelle Perry. I am grateful to the various people with whom I have had the opportunity to collaborate: Arvind Sivasubramanian and Jeff Gray at Johns Hopkins, Jeff Way at EMD Lexigen, and Daniel Steiner and Andreas Plueckthun at the University of Zurich.

Thank you Amy Lewis, Mark Styczynski, and Jared Johnson for being great friends and helping me make it through my first semester at MIT. Thanks to my parents, Jim and Grace Chao, and my brother Eric for being supportive through all these years.

My research was supported by a National Defense Science and Engineering Graduate Fellowship, NIH grant CA96504, and a David Koch Graduate Fellowship from the MIT Center for Cancer Research.

4 Table of Contents

Chapter 1: Introduction and background ...... 7 1.1 EGFR signaling and trafficking ...... 7 1.2 Biochemical and structural characterization of EGFR ...... 11 1.3 EGFR in cancer ...... 15 1.4 Current EGFR-targeted therapeutics ...... 18 1.5 Overview of thesis work ...... 25 1.6 Works cited ...... 27 Chapter 2: Epitope mapping of proteins binding to EGFR ...... 36 2.1 Background ...... 36 2.2 Materials and methods ...... 39 2.3 Results ...... 42 2.3.1 Creation of epitope mapping library ...... 42 2.3.2 Epitope mapping of mAb 806 ...... 44 2.3.3 Effects of EGFR mutations on binding of mAbs 806 and 175 ...... 52 2.3.4 Epitope mapping of mAbs 225 and 13A9 ...... 54 2.3.5 Epitope mapping of mAb EMD72000 ...... 63 2.3.6 Epitope mapping of DARPins ...... 68 2.4 Discussion ...... 75 2.5 Works cited ...... 80 Chapter 3: Engineering antibodies against EGFR ...... 84 3.1 Background ...... 84 3.2 Materials and methods ...... 94 3.3 Results ...... 100 3.3.1 Characterization of antibody clones against EGFR peptides ...... 100 3.3.2 Initial round of selection for scFvs against peptides ...... 103 3.3.3 Second round of affinity maturation ...... 107 3.3.4 Third round of affinity maturation ...... 110 3.3.5 Fab display libraries ...... 112 3.3.6 Rabbit monoclonal antibodies ...... 115 3.4 Discussion ...... 118 3.5 Works cited ...... 120 Chapter 4: Mathematical modeling of high affinity EGF binding ...... 123 4.1 Background ...... 123 4.2 Results ...... 126 4.2.1 Model formulation and output ...... 126 4.2.2 Model characterization ...... 134 4.3 Discussion ...... 137 4.4 Works cited ...... 142 Appendix 1: Sequences ...... 145 Appendix 2: Protocol ...... 153 Curriculum vitae ...... 154

5 Abbreviations Used

ADCC Antibody-dependent cellular cytotoxicity APC Allophycocyanin CDR Complementarity determining region DARPin Designed ankyrin repeat protein EGF EGFR Epidermal growth factor receptor Fab Fragment antigen binding portion of an antibody FACS Fluorescence-activated cell sorting FITC Fluorescein isothiocyanate IgG Immunoglobulin G Kd Apparent dissociation constant koff Kinetic off-rate kon Kinetic on-rate mAb MACS Magnetic cell sorting pAb Polyclonal antibody PBS Phosphate-buffered saline PCR Polymerase chain reaction PE Phycoerythrin scFv Single-chain variable fragment of an antibody TGF-α Transforming growth factor α TKI inhibitor VH Variable heavy chain of an antibody VL Variable light chain of an antibody

6 Chapter 1: Introduction and background

1.1 EGFR signaling and trafficking

The epidermal growth factor receptor (EGFR/Her1/ErbB1) is a type I transmembrane

and a member of the ErbB receptor family. This family includes three

other members: Her2 (ErbB2/Neu), Her3 (ErbB3), and Her4 (ErbB4). The receptors are

activated by a number of different ligands, such as epidermal growth factor (EGF) and

transforming growth factor alpha (TGF-α) which bind to EGFR, and and

which bind to both EGFR and Her4 (1). The exception to this is Her2, for which no natural

ligand has been discovered. The binding of ligand drives receptor dimerization, and the various receptors in the EGFR family are capable of both homo- and heterodimerization with each other.

However, it does not appear that Her2 homodimerizes at physiological expression levels, and homodimerization of Her3 receptors would not lead to receptor phosphorylation since Her3 is kinase dead and thus must be trans-activated by heterodimerization with other ErbB receptors.

The receptor dimerization event allows for kinase activation and trans-phosphorylation of

residues in the intracellular domain. A number of different EGFR tyrosine residues can become

phosphorylated, depending on the stimulating ligand and dimerization partner. Following

stimulation of EGFR by EGF, the major sites of phosphorylation include tyrosines 1086, 1148,

and 1173 (2). Phosphotyrosine-binding proteins then associate with these residues in the ErbB receptor tail and become activated, initiating various signaling cascades. Many of these interactions have been characterized in detail, such as Grb2 binding to EGFR tyrosine 1068 and

SHC binding to EGFR tyrosines 1148 and 1173 (3,4). EGFR has been shown to activate a number of signaling pathways, including the mitogen-activated protein kinase (MAPK),

7 phosphatidylinositol 3-kinase (PI3K), and protein kinase C (PKC) pathways. In addition, EGFR

can be trans-activated by other proteins, such as G-protein-coupled receptors (GPCRs) (5). The

various signaling cascades associated with EGFR activation are summarized in Figure 1-1. It has

also been observed that activated EGFR may undergo nuclear translocalization and subsequently

regulate gene expression, although the contribution of this pathway compared to more traditional

signal transduction is unclear (6,7). The activation of EGFR thus leads to multiple cell responses,

including cellular growth, differentiation, and migration.

Figure 1-1: EGFR signaling network. (a) EGFR family ligands are shown with their binding specificities. For simplicity, arrows are only drawn for EGF and 4 (NRG4), although other ligand specificities are shown in parentheses. (b) Signaling to the adaptor protein/kinase layer is only shown for two receptor dimers. (c) Possible cellular outcomes based on EGFR signaling. Figure has been reproduced from (5) with permission from Elsevier.

In the absence of ligand, EGFR is constitutively internalized with a half life of ~30

minutes and quickly recycled back to the cell surface, leading to a total receptor distribution of

~80-90% on the cell surface at any given time (Figure 1-2) (8). The metabolic half-life of EGFR

8 in a tumor cell line is 20 hours (9); thus, a single receptor will cycle through the endocytic

pathway many times during its lifetime. Ligand activation increases the rate of EGFR

endocytosis 5-10 fold, but does not increase the internalization rates of other ErbB receptors

(10). In addition, activated EGFR dimers bound to EGF are preferentially retained in the endosomes and not recycled (11), and these endosomes mature into multivesicular bodies and are targeted to lysosomes (12). This leads to the specific degradation of activated receptors and overall receptor downregulation. Earlier reports indicated that ligand-accelerated internalization required EGFR kinase activity (13,14), but a recent study found that receptor dimerization was sufficient to induce rapid internalization and that kinase activity was unnecessary (15). Although receptor ubiquitination does not affect EGFR internalization rates (16,17), ubiquitination by c-Cbl targets receptors to the lysosome and is dependent on kinase activity and an intact

C-terminal region (18).

While receptor internalization has often been viewed as a means of signaling attenuation,

it has been shown that activated EGFR continues to signal while in the endosome through certain

signaling pathways (19,20). Although Her2 has often been described as the preferred

heterodimerization partner for ErbB receptors (21), mathematical modeling and experiments

have shown that EGFR homodimerization and EGFR/Her2 heterodimerization occur with

comparable affinities (22), and EGFR/Her2 heterodimers appear to predominate because they

have a reduced rate of internalization and increased fraction of recycling (23). EGFR trafficking

has been the subject of much mathematical modeling, and the system serves as a paradigm for

other ligand-binding receptors (24). The constitutive trafficking patterns of EGFR can even

affect the global pharmacokinetics of antibody drugs which bind to the receptor due to antibody

9 internalization and subsequent degradation (25). Additional experiments on EGFR trafficking are reviewed in (8) and (26).

Figure 1-2: Trafficking of EGFR. Activated EGFR homo- and heterodimers are internalized through clathrin-coated pits and preferentially degraded. Times represent approximate mean time of the processes indicated. Figure has been reproduced from (8) with permission from Elsevier.

10 1.2 Biochemical and structural characterization of EGFR

EGFR is a 170 kDa protein that is heavily N-glycosylated but not O-glycosylated (27), with carbohydrate accounting for ~20% of the molecular mass. Mass spectrometry characterization has determined that eight of the eleven canonical N-linked sites are glycosylated in the A431 cell line, with two sites unglycosylated and one site glycosylated in a fraction of the receptors (28). The EGFR extracellular region is divided into four domains, numbered I-IV

(Figure 1-3). Domains I and III (DI and DIII), also designated L1 and L2, are members of the leucine rich repeat family and are similar to domains found in the receptor family (29).

Prior to receptor crystallization, it was known that both DI and DIII contribute to ligand binding

(30). However, domain III appears to contribute most of the binding energy, since the affinity of

DIII alone for EGF is ~400 nM as measured solubly by isothermal titration calorimetry (31), while soluble EGFR ectodomain binds immobilized EGF with an affinity of ~200 nM (32).

Domains II and IV (DII and IV), or the cysteine-rich (CR) domains, contain multiple small disulfide-bonded modules similar to those in laminin (33).

Figure 1-3: Domains of EGFR. The extracellular region consists of domains I-IV, followed by a transmembrane region (TM) and the intracellular portion. The regions inside the cell include the juxtamembrane portion (JM), tyrosine kinase (TK), and C-terminal tail containing tyrosine residues which become phosphorylated upon receptor activation.

11 The EGFR ectodomain has been crystallized in both monomer and dimer forms, and a

review of these structures can be found in (34). The monomer exists in a tethered, closed, or

autoinhibited state (35), with DII and DIV forming a tether contact (Figure 1-4, left). However,

in the dimer structures of EGFR bound to EGF (36) or TGF-α (37) ligand, the receptor is in an

untethered, open, or extended conformation, with ligand binding between DI and DIII (Figure

1-4, right). Although most cytokine receptors dimerize through ligand bridging the two receptor

monomers (38), all dimerization contacts in the EGFR dimer are fully receptor-mediated. The

dimer exists in a 2:2 receptor to ligand ratio, with multiple contacts through the DII dimerization

arm (residues 242-259). An additional DII loop, residues 271-283, also contributes important

dimer contacts (32). The dimerization interface may include DIV; however, in the dimer

structures, DIV is either unresolved (36) or was not present in the crystallized protein (37).

Peptides mimicking portions of DIV can inhibit EGFR heterodimerization with Her3 (39), and

DIV mutations can impair the ability of ligand to bind and induce EGFR phosphorylation (40).

However, these potential DIV contacts contribute <9% of the free energy of dimerization as

demonstrated by studies on soluble EGFR dimer formation (32). For illustrative purposes, DIV

has been added to the dimer structure in Figure 1-4 in the same relative orientation to DIII as

seen in the tethered monomer. A low-resolution molecular envelope structure of the full EGFR dimer in solution has been obtained from small-angle X-ray scattering (SAXS), and it is consistent with this dimer structure (41).

12

Figure 1-4: Crystal structures of monomeric and dimeric EGFR. Left, autoinhibited EGFR monomer (1NQL), with domain I (DI), red; DII, orange; DIII, blue; DIV, green. A 130o rotation of DI and DII allows DI and DIII to come together to bind EGF ligand. Right, dimeric EGFR bound to EGF ligand in yellow (1IVO). DIV is not resolved in this structure, but has been added in the same relative orientation to DIII as seen in the monomer. This figure was adapted from (34).

From the crystal structures of the monomer and dimer, a model for EGFR activation has been proposed (34). In this model, the monomer mostly exists in a tethered conformation, in which the DII dimerization arm and other residues are obscured, thus preventing receptor dimerization. The monomer equilibrates between the tethered conformation and an extended conformation similar to that seen in the dimer. The binding of ligand between DI and DIII stabilizes the extended conformation, which exposes the DII dimerization arm and other residues, thus allowing the formation of the dimer. However, the breaking of the tether is necessary but not sufficient for dimer formation, as demonstrated through mutations and deletions of DIV (37,42,43). It also appears that certain types of receptor glycosylation can shift the equilibrium toward the extended form (44,45). It has subsequently been found that the DII-

DIV tether interaction is not necessarily responsible for maintaining the receptor in the autoinhibited conformation; in fact, mutating every tether interaction in soluble EGFR still leads

13 to an autoinhibited conformation (41). In addition, it appears that ligand binding is necessary to

stabilize the conformation of the DII loop 271-283, which is an important dimerization contact

(32).

Ectodomain crystal structures are also available for the EGFR family members Her2 (46),

Her3 (47), and Her4 (48). The Her2 monomer, in contrast to EGFR, adopts an extended

conformation similar to that of EGFR in the dimer structure. Since Her2 has no activating ligand,

it is extended and poised to interact with other ErbB receptors. However, since ligand binding is

necessary to activate Her3 and Her4, these monomers adopt a tethered conformation.

EGFR has a single transmembrane domain, and the intracellular portion contains a short

juxtamembrane region, the tyrosine kinase, and a C-terminal tail containing tyrosine residues

which become phosphorylated upon receptor activation (Figure 1-3). The EGFR kinase domain

has been independently crystallized, revealing an intrinsically autoinhibited domain resembling

Src and cyclin-dependent kinases (CDKs) (49). In contrast to most kinases, phosphorylation of the EGFR activation loop is not necessary for its activation (50). Instead, the EGFR kinase is activated by an asymmetric dimer in which the C-terminal lobe of one kinase domain binds to the second kinase domain in a manner analogous to cyclin in activated CDK/cyclin complexes.

Thus, ligand binding brings two receptor monomers together and allows for the dimerization and subsequent activation of the kinase domains. In this model, the extent of EGFR kinase activation depends on the effective local concentration of the receptor, since the probability of dimerization will increase with increasing kinase domain density. This phenomenon has been demonstrated experimentally using increasing concentrations of EGFR kinase domains tethered to the surface of vesicles (49).

14 1.3 EGFR in cancer

EGFR was the first receptor to be directly linked to human cancer (51), and because

EGFR activation often leads to cellular growth, its signaling can provide tumor cells with substantial survival advantages. In addition, EGFR signaling has been implicated in tumor cell production of pro-angiogenic factors and cellular migration and invasion (52). EGFR dysregulation has been observed in a wide variety of carcinomas, including head and neck, breast, bladder, and non-small-cell (5). Excessive EGFR signaling can arise from receptor overexpression, autocrine signaling, or mutation. Normal cells usually express 4x104 to

1x105 EGF receptors per cell, but tumor cells can express as many as 2x106 (53). Receptor

overexpression commonly develops due to gene amplification, although it was recently reported

that the hypoxic microenvironment of tumors can also induce overexpression of EGFR by

increasing EGFR mRNA translation (54). In head and neck cancer, at least 80% of tumors

overexpress EGFR (55), and the observed percentages of tumors overexpressing EGFR in

various types of cancer are listed in Table 1-1. Furthermore, elevated levels of EGFR serve as a

prognostic indicator for poor survival rates (56).

Tumor type Percentage of tumors overexpressing EGFR Colon 22-75% Head and neck 80-100% Pancreatic 30-50% Non-small-cell lung 40-80% Breast 14-91% Renal 50-90% Ovarian 35-70% Glioma 40-63% Bladder 31-48% Table 1-1: EGFR overexpression in different tumor types. Table has been adapted from (53).

15 In addition, EGFR ligands can be overexpressed, resulting in high levels of autocrine signaling when the receptor is present (57). Autocrine production of TGF-α or EGF is also associated with reduced cancer survival (21). Finally, various activating EGFR mutations have been observed in tumor samples. Gene rearrangements in glioblastoma often lead to truncation mutants, the most common of which is EGFRvIII (EGFRΔ2-7), where amino acids 6-273

(encoded by exons 2-7) are deleted from the gene (58-60). This leads to a mutant which is missing DI and most of DII, with EGFR amino acids 1-5 fused to the rest of the receptor through a novel glycine residue resulting from the gene fusion. Although most frequently seen in glioblastoma, EGFRvIII has also been observed in other types of tumors, including non-small- cell lung, breast, and ovarian cancers (61-63). EGFRvIII is not activated by ligand, but signals constitutively at low levels (64) and thus confers enhanced tumorigenicity to cancer cells (65).

There is also evidence that EGFRvIII is downregulation defective, leading to prolonged signaling (66). In addition, point mutants of the EGFR ectodomain have recently been observed in glioblastoma, and select mutants showed high basal phosphorylation and conferred tumorigenicity to NIH-3T3 cells (67).

In non-small-cell lung cancer (NSCLC), mutations in the EGFR kinase domain, which are clustered around the ATP-binding pocket of the enzyme, have been observed (68). These mutations appear to be restricted to a subset of NSCLC patients, particularly non-smokers, women, and those of Asian ethnicity (69). Exon 19 in-frame deletions of some or all of residues

746-750 account for 45% of reported mutations, exon 21 substitutions (especially L858R) account for 45%, and the remaining 10% involve exons 18 and 20, such as G719S and V765A

(68). Most of these mutations have been shown to hyperactivate the kinase and have oncogenic

16 activity. Although predominately seen in NSCLC, very rare kinase domain mutations have also been observed in other types of tumors, including ovarian, colorectal, and head and neck (68).

Other EGFR family members have been shown to be dysregulated in cancer. Her2 overexpression due to gene amplification has been observed in various tumors, including breast, ovarian (70), and bladder cancer (71). In particular, Her2 overexpression is an indicator of poor prognosis in breast cancer (72). Her3 and Her4 signaling can also contribute to a malignant phenotype (5).

17 1.4 Current EGFR-targeted therapeutics

Because of the importance of the EGFR family in cancer, the receptors are promising

targets for rational oncology therapies designed to inhibit receptor signaling. The two main

classes of such therapeutics are tyrosine kinase inhibitors (TKIs) and monoclonal antibodies.

TKIs are small molecule drugs which are designed to bind the ATP pocket of the kinase domain

and inhibit EGFR phosphorylation (73). Two EGFR-targeted TKIs have been approved by the

U.S. Food and Drug Administration (FDA) for the treatment of non-small-cell lung cancer: (Iressa, ZD1839; AstraZeneca) (74) and (Tarceva, OSI-774;

OSI/Genentech/Roche) (75), which has been co-crystallized with the EGFR kinase domain (76).

Although these two TKIs are targeted to the EGFR tyrosine kinase, with IC50 values

(concentration required for 50% inhibition of activity) in the nanomolar range, they also will

inhibit the Her2 tyrosine kinase with IC50 values in the micromolar range (73). A small subset of

NSCLC patients dramatically respond to treatment with gefitinib (73) or erlotinib, and the

characterization of these patient tumors led to the discovery of activating EGFR kinase domain

mutations (described in Section 1.3). In fact, 77% of NSCLC patients who respond to gefitinib or erlotinib harbor certain kinase domain mutations, as opposed to a 7% incidence of mutation in non-responders. It has been hypothesized that these activating EGFR kinase domain mutations lead to a state of oncogene addiction for tumor cells, thus explaining the extreme sensitivity of mutation-bearing tumors to EGFR TKIs (68). Unexpectedly, in cells expressing the ectodomain truncation mutant EGFRvIII, gefitinib induces signaling at low concentrations (77). Gefitinib and erlotinib are reversible kinase inhibitors; however, irreversible TKIs which bind the kinase domain covalently, such as HKI-357 (78), have also been developed. These have been shown to

18 be effective against reversible-TKI resistant cells expressing a secondary kinase mutation

T790M. Additional TKI drugs in development are reviewed in (79) and (68).

An alternative to TKI therapeutics is treatment with anti-EGFR monoclonal antibodies

(mAbs; for a general discussion on antibody structure, see Section 3.1). While TKIs can be

conveniently delivered orally and have a reduced chance of inducing allergic reactions,

monoclonal antibodies have superior stability in vivo and lower gastrointestinal toxicity (79).

Cetuximab (Erbitux, IMC-C225; ImClone Systems) was the first anti-EGFR monoclonal

antibody to be approved by the FDA. Originally approved for the treatment of EGFR-expressing

metastatic , it has also recently been approved for head and neck cancer. The antibody was initially isolated by immunization of mice with EGFR, and the murine monoclonal antibody was designated mAb 225 (80). In vitro experiments revealed that this antibody inhibited the binding of ligand to EGFR, resulting in decreased tyrosine phosphorylation and reduced cell proliferation (80,81). The crystal structure of the Fab bound to the EGFR ectodomain reveals that the antibody directly competes with EGF ligand by binding to DIII on an epitope which overlaps the EGF binding site (82). It was also found that mAb 225 can induce EGFR internalization (83), which may reduce the number of receptors available to bind ligand. The bivalency of the mAb 225 molecule was shown to be important for its ability to internalize and downregulate EGFR and inhibit subsequent growth (84). In addition, cetuximab has been shown to exhibit anti-angiogenic properties, leading to even more effective treatment of tumor xenografts in vivo (85). Cetuximab may also induce antibody-dependent cellular toxicity

(ADCC) in vivo (86), and the antibody’s various mechanisms of action are further discussed in

(87). To decrease its potential immunogenicity, the mouse mAb 225 was chimerized with human

IgG1 by adding the mouse antibody variable regions to the human constant IgG regions. This

19 chimeric antibody, designated C225, was shown to have a higher affinity for EGFR and was

effective against human tumor xenografts (88). Additionally, cetuximab is able to inhibit the growth of lung cancer xenografts expressing the L858R/T790M EGFR kinase mutants, which are resistant to reversible TKI treatment (89). In clinical trials, cetuximab was well-tolerated with mild toxicities, which included acneiform rash, fatigue, and hypersensitivity (90). The effects to normal tissue appear to minimal, since many healthy cells expressing EGFR are not rapidly turned over, with the intestinal epithelium as an exception (53). In a Phase III clinical trial with

329 irinotecan-refractory metastatic colorectal cancer patients, 10.8% responded to cetuximab therapy alone while 22.9% responded to cetuximab plus irinotecan (90), leading to the FDA approval of this drug. Interestingly, cetuximab treatment was able to reverse tumor cell resistance to the chemotherapeutic irinotecan.

Other antibodies have been developed which similarly inhibit ligand binding to EGFR as

cetuximab does. (Vectibix, ABX-EGF; Amgen/Abgenix) has also been FDA-

approved for the treatment of metastatic colorectal cancer and blocks the binding of ligands to

EGFR (91). Panitumumab has one potential advantage over cetuximab; it is a fully human

antibody developed from the XenoMouse technology, in which human antibody genes were

cloned into mice (91). However, panitumumab is a class IgG2 antibody and thus incapable of

activating ADCC, although it is unclear how important this is for tumor inhibition (92). Data

from clinical trials shows that panitumumab produces an 8% overall response rate in metastatic

colorectal cancer, with progression-free survival extended from 60 days to 96 days (92). Another

cetuximab-like antibody is matuzumab (EMD72000; EMD Pharmaceuticals/Merck KgaA),

which is currently in various phase II clinical trials. The antibody was originally isolated from

mice (mAb 425/EMD55900) and inhibits the binding of ligands to EGFR (93). The murine

20 version of the antibody was humanized by grafting the murine complementarity-determining regions (CDRs) onto a human IgG1 framework (94). A phase I study of matuzumab treatment of patients with advanced pancreatic showed partial responses and stable disease in eight of twelve patients (95). However, in a recent phase II trial, matuzumab showed no evidence of significant clinical activity in ovarian tumors (96). In addition, a bispecific version of the humanized mAb 425, named MDX-447 (Medarex/Merck KgaA), has been developed, with the other arm of the antibody binding to the high affinity Fc receptor (FcγRI) (97). MDX-447 is currently undergoing further clinical trials in combination with activated monocytes in an effort to direct these immune cells to tumor sites. Finally, (hR-3, TheraCIM; YM

Biosciences/Center of Molecular Immunology, Cuba) is a which also blocks ligand binding to EGFR (98). Phase I/II trials of glioblastoma and anaplastic astrocytoma patients showed a 38% objective response rate (99), and the antibody is currently undergoing further trials in the U.S.

Antibodies which target the truncation mutant EGFRvIII have also been developed. Since

EGFRvIII contains a novel glycine residue at the junction between two EGFR fragments, this junctional peptide is a potential target for antibody binding and represents an epitope specific to tumor cells. One antibody targeting this peptide is MR1-1, which is an affinity-matured single chain variable fragment (scFv) specific for EGFRvIII (100). However, antibodies against the junctional peptide have no intrinsic anti-EGFRvIII activity, and thus must be conjugated to cytotoxic agents such as immunotoxins in order to kill tumor cells (101). Another antibody, mAb

806, was isolated from immunization of mice with EGFRvIII, and it was shown to bind and inhibit the growth of cells expressing the mutant protein. However, it was discovered that mAb

806 also bound to a small portion of wild-type EGFR (~5-10%) on cells overexpressing the

21 receptor and inhibited cell growth, but it did not bind to or inhibit cells expressing normal EGFR levels (102-104). Therefore, mAb 806 has the unique ability to recognize EGFR expressed by

tumor cells but not on normal tissues. Characterization of mAb 806 revealed that it binds a

peptide epitope shared by both EGFRvIII and EGFR (105). However, this epitope is only

exposed in certain situations: in EGFRvIII due to the deletion of DI and DII, when wild-type

EGFR transitions from the autoinhibited to open conformation, or by high-mannose forms of the

receptor in cells overexpressing EGFR (44). The mAb 806 epitope is discussed in further detail

in Chapter 2. mAb 806 has also been chimerized to reduce murine sequences, generating

antibody ch806 (106). In mouse models of cancer, 806 was effective in inducing regression of

lung tumors expressing either EGFR harboring an activating kinase domain mutation or

EGFRvIII (107). However, 806 is ineffective against kinase-dead EGFRvIII, suggesting that it

functions by blocking receptor trans-phosphorylation (108). Phase I clinical trial data showed

that ch806 was well-tolerated by patients, with highly specific targeting to tumors versus normal

tissues (109), and further trials are currently ongoing.

The only FDA-approved antibody which targets Her2, (Herceptin, 4D5;

Genentech), was isolated by mouse immunization (110) and humanized (111). Trastuzumab

inhibited the proliferation of breast tumor-derived cell lines (110) and also supported ADCC killing against the SK-Br-3 tumor line (111). Trastuzumab was shown to be effective against human gastric carcinoma xenografts (112), as well as breast cancer BT-474 xenografts (113). In phase III trials, the addition of trastuzumab treatment to chemotherapy was associated with a longer time to disease progression (7.4 vs. 4.6 months) and a higher objective response rate (50% vs. 32%) (114), leading to the FDA approval of this drug for the treatment of Her2-positive metastatic breast cancer. The main toxicities were cardiac adverse events, since Her2 signaling

22 appears to be important for adult heart function (115). Recent trials have shown that adjuvant

trastuzumab therapy reduces the 3-year risk of recurrence by half in early breast cancer patients

(116). Despite the success of trastuzumab, its mechanism of action is not immediately apparent.

A crystal structure of trastuzumab bound to Her2 reveals that it binds the C-terminal portion of

domain IV (117); however, it does not block Her2 heterodimerization with Her3 (118). Although

initial reports suggested that trastuzumab could inhibit tumors through downregulation of the

Her2 receptor (119,120), a more recent study demonstrated that the antibody is internalized and recycled back to the cell surface without mediating receptor degradation (121). The Her2 ectodomain undergoes cleavage, and the resulting membrane-bound kinase fragment shows increased tyrosine kinase activity (122). Trastuzumab blocks this Her2 shedding (123), presumably by restricting access to the protease cleavage site, thus preventing the formation of the activated kinase fragment (119). Finally, trastuzumab can induce ADCC killing of tumor cells (124), but this cannot be its sole mechanism since ADCC cannot explain the antibody’s efficacy in vitro. Additional experiments with trastuzumab are reviewed in (125). Another anti-

Her2 antibody named (Omnitarg, 2C4; Genentech) binds to domain II of Her2

(demonstrated by co-crystallization (126)), and thus it inhibits Her2/Her3 heterodimerization and

breast and prostate xenograft tumor growth (118). The murine version, mAb 2C4, was

humanized (127), and initial results from a phase II trial treating prostate cancer showed no

objective responses (128). However, a clinical trial evaluating pertuzumab treatment plus

trastuzumab is currently ongoing, motivated by preclinical studies where the two antibodies synergistically inhibited the survival of breast cancer cells (129).

Although therapeutics directed at the EGFR receptor family have been somewhat

effective in inhibiting the growth of tumor cells, optimal cancer treatments will likely include

23 some form of combination therapy. Since multiple genetic alterations contributing to a malignant phenotype normally arise during the course of tumor development (130), it may be difficult to

inhibit these tumors by blocking only one receptor. Synergistic inhibition effects of combining

anti-EGFR antibodies with chemotherapeutic agents in vitro were initially reported by (131).

Treatments of trastuzumab plus chemotherapy (114) or cetuximab plus irinotecan (90) both improved patient outcome in clinical trials. Additional trials examining combinations of anti-

EGFR antibodies with other chemotherapeutics have been and are currently being investigated

(3,53,132). However, not all combinations are synergistic or even additive, as demonstrated by failed clinical trials combining gefitinib or erlotinib with chemotherapy (133). Using two different classes of anti-EGFR inhibitors may be effective, since tumor xenograft experiments have shown synergy between mAb 806 and the TKI AG1478 (134). In addition, the combination of two anti-EGFR antibodies directed against different epitopes has been shown to stimulate receptor downregulation and inhibit signaling (135,136). Other successful combinations include simultaneously inhibiting EGFRvIII and c-Met (137), and rationales for treatment using ErbB plus mTOR inhibitors or ErbB plus estrogen inhibitors are discussed in (132).

24 1.5 Overview of thesis work

In this thesis work, we have sought to characterize and engineer antibodies against the

epidermal growth factor receptor. Multiple anti-EGFR antibodies have been developed, but their

binding epitopes had not been determined due to a lack of convenient epitope mapping methods.

To alleviate this, we developed a novel method of epitope mapping using random mutagenesis and yeast surface display. In our approach, we disrupted the antibody-antigen interaction through

point mutations. However, since we did not know the epitopes of the antibodies, we made a

library of random mutations, and then combinatorially screened this library using yeast surface

display. EGFR mutants which displayed a loss of binding to the antibody being mapped were

thus potential epitope residues. It was also necessary to screen for retention of binding to a

conformation-specific antibody to remove misfolded EGFR mutants. The development of our novel method enabled the epitope mapping of a total of 4 antibodies: mAbs 806, 225, 13A9, and

EMD72000. The determination of the antibody epitopes enabled further understanding of their mechanisms of action. Since mAbs 806, 225, and EMD72000 have high therapeutic relevance, this information may have clinical impact as well. In addition, the epitope mapping method was extended to mapping the binding of three designed ankyrin repeat proteins (DARPins) against

EGFR.

Instead of using the common inhibition strategy of using antibodies to block ligand

binding, we opted to engineer antibodies to directly inhibit receptor dimerization. Such

antibodies could have advantages in tumors driven by autocrine signaling or the EGFRvIII

mutant. We selected epitopes in EGFR domain II and domain IV to target with antibodies, and

synthesized peptide mimics of these loops to use as antigen. Starting from a nonimmune human

scFv library in yeast, candidate binders to the peptide epitopes were isolated and characterized.

25 The majority of binders were heavy chain only constructs, many of which were likely mediating

binding through the VH/VL interface instead of the antibody CDRs. In addition, rabbit monoclonal antibodies against a domain IV peptide were developed, but appeared to only bind the peptide and not the EGFR ectodomain.

Finally, previous experiments had shown that polyclonal antibodies (pAbs) against either

EGFR domain II or domain IV could eliminate the high affinity facet of EGF binding normally

observed on the surface of cells. Based on these results, we hypothesized that the pAbs inhibited

receptor dimerization, and thus inactive preformed EGFR dimers could represent the high

affinity binding component. We developed a mathematical model incorporating these hypotheses

to determine whether such a model could be consistent with experimental observations. The

equilibrium ligand binding and receptor dimerization model was fit to the experimental data

using three parameters: the high affinity binding constant, the dimerization constant, and the

maximum binding signal. The model was able to successfully reproduce the experimental data,

and it could produce characteristic concave-up Scatchard plots for a wide range of values.

Overall, we have increased understanding of the mechanisms of action of a number of anti-

EGFR antibodies.

26 1.6 Works cited

1. Harris, R. C., Chung, E., and Coffey, R. J. (2003) EGF receptor ligands, Exp Cell Res 284, 2-13. 2. Erba, E. B., Matthiesen, R., Bunkenborg, J., Schulze, W. X., Di Stefano, P., et al. (2007) Quantitation of multisite EGF receptor phosphorylation using mass spectrometry and a novel normalization approach, J Proteome Res 6, 2768-2785. 3. Sebastian, S., Settleman, J., Reshkin, S. J., Azzariti, A., Bellizzi, A., et al. (2006) The complexity of targeting EGFR signalling in cancer: from expression to turnover, Biochim Biophys Acta 1766, 120-139. 4. Schulze, W. X., Deng, L., and Mann, M. (2005) Phosphotyrosine interactome of the ErbB-receptor kinase family, Mol Syst Biol 1, 2005 0008. 5. Yarden, Y., and Sliwkowski, M. X. (2001) Untangling the ErbB signalling network, Nat Rev Mol Cell Biol 2, 127-137. 6. Wells, A., and Marti, U. (2002) Signalling shortcuts: cell-surface receptors in the nucleus?, Nat Rev Mol Cell Biol 3, 697-702. 7. Lo, H. W., and Hung, M. C. (2006) Nuclear EGFR signalling network in cancers: linking EGFR pathway to cell cycle progression, nitric oxide pathway and patient survival, Br J Cancer 94, 184-188. 8. Wiley, H. S. (2003) Trafficking of the ErbB receptors and its influence on signaling, Exp Cell Res 284, 78-88. 9. Stoscheck, C. M., and Carpenter, G. (1984) Characterization of the metabolic turnover of epidermal growth factor receptor protein in A-431 cells, J Cell Physiol 120, 296-302. 10. Waterman, H., Sabanai, I., Geiger, B., and Yarden, Y. (1998) Alternative intracellular routing of ErbB receptors may determine signaling potency, J Biol Chem 273, 13819- 13827. 11. Herbst, J. J., Opresko, L. K., Walsh, B. J., Lauffenburger, D. A., and Wiley, H. S. (1994) Regulation of postendocytic trafficking of the epidermal growth factor receptor through endosomal retention, J Biol Chem 269, 12865-12873. 12. Futter, C. E., Pearse, A., Hewlett, L. J., and Hopkins, C. R. (1996) Multivesicular endosomes containing internalized EGF-EGF receptor complexes mature and then fuse directly with lysosomes, J Cell Biol 132, 1011-1023. 13. Chen, W. S., Lazar, C. S., Poenie, M., Tsien, R. Y., Gill, G. N., et al. (1987) Requirement for intrinsic protein tyrosine kinase in the immediate and late actions of the EGF receptor, Nature 328, 820-823. 14. Schmidt, M. H., Furnari, F. B., Cavenee, W. K., and Bogler, O. (2003) Epidermal growth factor receptor signaling intensity determines intracellular protein interactions, ubiquitination, and internalization, Proc Natl Acad Sci U S A 100, 6505-6510. 15. Wang, Q., Villeneuve, G., and Wang, Z. (2005) Control of epidermal growth factor receptor endocytosis by receptor dimerization, rather than receptor kinase activation, EMBO Rep 6, 942-948. 16. Duan, L., Miura, Y., Dimri, M., Majumder, B., Dodge, I. L., et al. (2003) Cbl-mediated ubiquitinylation is required for lysosomal sorting of epidermal growth factor receptor but is dispensable for endocytosis, J Biol Chem 278, 28950-28960.

27 17. Huang, F., Kirkpatrick, D., Jiang, X., Gygi, S., and Sorkin, A. (2006) Differential regulation of EGF receptor internalization and degradation by multiubiquitination within the kinase domain, Mol Cell 21, 737-748. 18. Levkowitz, G., Waterman, H., Zamir, E., Kam, Z., Oved, S., et al. (1998) c-Cbl/Sli-1 regulates endocytic sorting and ubiquitination of the epidermal growth factor receptor, Genes Dev 12, 3663-3674. 19. Haugh, J. M., Huang, A. C., Wiley, H. S., Wells, A., and Lauffenburger, D. A. (1999) Internalized epidermal growth factor receptors participate in the activation of p21(ras) in fibroblasts, J Biol Chem 274, 34350-34360. 20. Wang, Y., Pennock, S., Chen, X., and Wang, Z. (2002) Endosomal signaling of epidermal growth factor receptor stimulates signal transduction pathways leading to cell survival, Mol Cell Biol 22, 7279-7290. 21. Citri, A., and Yarden, Y. (2006) EGF-ERBB signalling: towards the systems level, Nat Rev Mol Cell Biol 7, 505-516. 22. Hendriks, B. S., Opresko, L. K., Wiley, H. S., and Lauffenburger, D. (2003) Quantitative analysis of HER2-mediated effects on HER2 and epidermal growth factor receptor endocytosis: distribution of homo- and heterodimers depends on relative HER2 levels, J Biol Chem 278, 23343-23351. 23. Hendriks, B. S., Opresko, L. K., Wiley, H. S., and Lauffenburger, D. (2003) Coregulation of epidermal growth factor receptor/human epidermal growth factor receptor 2 (HER2) levels and locations: quantitative analysis of HER2 overexpression effects, Cancer Res 63, 1130-1137. 24. Wiley, H. S., Shvartsman, S. Y., and Lauffenburger, D. A. (2003) Computational modeling of the EGF-receptor system: a paradigm for systems biology, Trends Cell Biol 13, 43-50. 25. Lammerts van Bueren, J. J., Bleeker, W. K., Bogh, H. O., Houtkamp, M., Schuurman, J., et al. (2006) Effect of target dynamics on pharmacokinetics of a novel therapeutic antibody against the epidermal growth factor receptor: implications for the mechanisms of action, Cancer Res 66, 7630-7638. 26. Carpenter, G. (2000) The EGF receptor: a nexus for trafficking and signaling, Bioessays 22, 697-707. 27. Cummings, R. D., Soderquist, A. M., and Carpenter, G. (1985) The oligosaccharide moieties of the epidermal growth factor receptor in A-431 cells. Presence of complex- type N-linked chains that contain terminal N-acetylgalactosamine residues, J Biol Chem 260, 11944-11952. 28. Zhen, Y., Caprioli, R. M., and Staros, J. V. (2003) Characterization of glycosylation sites of the epidermal growth factor receptor, Biochemistry 42, 5478-5492. 29. Ward, C. W., and Garrett, T. P. (2001) The relationship between the L1 and L2 domains of the insulin and epidermal growth factor receptors and leucine-rich repeat modules, BMC Bioinformatics 2, 4. 30. Lax, I., Fischer, R., Ng, C., Segre, J., Ullrich, A., et al. (1991) Noncontiguous regions in the extracellular domain of EGF receptor define ligand-binding specificity, Cell Regul 2, 337-345. 31. Lemmon, M. A., Bu, Z., Ladbury, J. E., Zhou, M., Pinchasi, D., et al. (1997) Two EGF molecules contribute additively to stabilization of the EGFR dimer, Embo J 16, 281-294.

28 32. Dawson, J. P., Berger, M. B., Lin, C. C., Schlessinger, J., Lemmon, M. A., et al. (2005) Epidermal growth factor receptor dimerization and activation require ligand-induced conformational changes in the dimer interface, Mol Cell Biol 25, 7734-7742. 33. Ward, C. W., Hoyne, P. A., and Flegg, R. H. (1995) Insulin and epidermal growth factor receptors contain the cysteine repeat motif found in the tumor necrosis factor receptor, Proteins 22, 141-153. 34. Burgess, A. W., Cho, H. S., Eigenbrot, C., Ferguson, K. M., Garrett, T. P., et al. (2003) An open-and-shut case? Recent insights into the activation of EGF/ErbB receptors, Mol Cell 12, 541-552. 35. Ferguson, K. M., Berger, M. B., Mendrola, J. M., Cho, H. S., Leahy, D. J., et al. (2003) EGF activates its receptor by removing interactions that autoinhibit ectodomain dimerization, Mol Cell 11, 507-517. 36. Ogiso, H., Ishitani, R., Nureki, O., Fukai, S., Yamanaka, M., et al. (2002) Crystal structure of the complex of human epidermal growth factor and receptor extracellular domains, Cell 110, 775-787. 37. Garrett, T. P., McKern, N. M., Lou, M., Elleman, T. C., Adams, T. E., et al. (2002) Crystal structure of a truncated epidermal growth factor receptor extracellular domain bound to transforming growth factor alpha, Cell 110, 763-773. 38. Stroud, R. M., and Wells, J. A. (2004) Mechanistic diversity of cytokine receptor signaling across cell membranes, Sci STKE 2004, re7. 39. Berezov, A., Chen, J., Liu, Q., Zhang, H. T., Greene, M. I., et al. (2002) Disabling receptor ensembles with rationally designed interface peptidomimetics, J Biol Chem 277, 28330-28339. 40. Saxon, M. L., and Lee, D. C. (1999) Mutagenesis reveals a role for epidermal growth factor receptor extracellular subdomain IV in ligand binding, J Biol Chem 274, 28356- 28362. 41. Dawson, J. P., Bu, Z., and Lemmon, M. A. (2007) Ligand-Induced Structural Transitions in ErbB Receptor Extracellular Domains, Structure 15, 942-954. 42. Mattoon, D., Klein, P., Lemmon, M. A., Lax, I., and Schlessinger, J. (2004) The tethered configuration of the EGF receptor extracellular domain exerts only a limited control of receptor function, Proc Natl Acad Sci U S A 101, 923-928. 43. Walker, F., Orchard, S. G., Jorissen, R. N., Hall, N. E., Zhang, H. H., et al. (2004) CR1/CR2 interactions modulate the functions of the cell surface epidermal growth factor receptor, J Biol Chem 279, 22387-22398. 44. Johns, T. G., Mellman, I., Cartwright, G. A., Ritter, G., Old, L. J., et al. (2005) The antitumor monoclonal antibody 806 recognizes a high-mannose form of the EGF receptor that reaches the cell surface when cells over-express the receptor, Faseb J 19, 780-782. 45. Whitson, K. B., Whitson, S. R., Red-Brewer, M. L., McCoy, A. J., Vitali, A. A., et al. (2005) Functional effects of glycosylation at Asn-579 of the epidermal growth factor receptor, Biochemistry 44, 14920-14931. 46. Garrett, T. P., McKern, N. M., Lou, M., Elleman, T. C., Adams, T. E., et al. (2003) The crystal structure of a truncated ErbB2 ectodomain reveals an active conformation, poised to interact with other ErbB receptors, Mol Cell 11, 495-505. 47. Cho, H. S., and Leahy, D. J. (2002) Structure of the extracellular region of HER3 reveals an interdomain tether, Science 297, 1330-1333.

29 48. Bouyain, S., Longo, P. A., Li, S., Ferguson, K. M., and Leahy, D. J. (2005) The extracellular region of ErbB4 adopts a tethered conformation in the absence of ligand, Proc Natl Acad Sci U S A 102, 15024-15029. 49. Zhang, X., Gureasko, J., Shen, K., Cole, P. A., and Kuriyan, J. (2006) An allosteric mechanism for activation of the kinase domain of epidermal growth factor receptor, Cell 125, 1137-1149. 50. Gotoh, N., Tojo, A., Hino, M., Yazaki, Y., and Shibuya, M. (1992) A highly conserved tyrosine residue at codon 845 within the kinase domain is not required for the transforming activity of human epidermal growth factor receptor, Biochem Biophys Res Commun 186, 768-774. 51. de Larco, J. E., and Todaro, G. J. (1978) Epithelioid and fibroblastic rat kidney cell clones: epidermal growth factor (EGF) receptors and the effect of mouse sarcoma virus transformation, J Cell Physiol 94, 335-342. 52. Holbro, T., Civenni, G., and Hynes, N. E. (2003) The ErbB receptors and their role in cancer progression, Exp Cell Res 284, 99-110. 53. Herbst, R. S., and Shin, D. M. (2002) Monoclonal antibodies to target epidermal growth factor receptor-positive tumors: a new paradigm for cancer therapy, Cancer 94, 1593- 1611. 54. Franovic, A., Gunaratnam, L., Smith, K., Robert, I., Patten, D., et al. (2007) Translational up-regulation of the EGFR by tumor hypoxia provides a nonmutational explanation for its overexpression in human cancer, Proc Natl Acad Sci U S A 104, 13092-13097. 55. Ford, A. C., and Grandis, J. R. (2003) Targeting epidermal growth factor receptor in head and neck cancer, Head Neck 25, 67-73. 56. Nicholson, R. I., Gee, J. M., and Harper, M. E. (2001) EGFR and cancer prognosis, Eur J Cancer 37 Suppl 4, S9-15. 57. Sizeland, A. M., and Burgess, A. W. (1992) Anti-sense transforming growth factor alpha oligonucleotides inhibit autocrine stimulated proliferation of a colon carcinoma cell line, Mol Biol Cell 3, 1235-1243. 58. Sugawa, N., Ekstrand, A. J., James, C. D., and Collins, V. P. (1990) Identical splicing of aberrant epidermal growth factor receptor transcripts from amplified rearranged genes in human glioblastomas, Proc Natl Acad Sci U S A 87, 8602-8606. 59. Wong, A. J., Ruppert, J. M., Bigner, S. H., Grzeschik, C. H., Humphrey, P. A., et al. (1992) Structural alterations of the epidermal growth factor receptor gene in human gliomas, Proc Natl Acad Sci U S A 89, 2965-2969. 60. Frederick, L., Wang, X. Y., Eley, G., and James, C. D. (2000) Diversity and frequency of epidermal growth factor receptor mutations in human glioblastomas, Cancer Res 60, 1383-1387. 61. Garcia de Palazzo, I. E., Adams, G. P., Sundareshan, P., Wong, A. J., Testa, J. R., et al. (1993) Expression of mutated epidermal growth factor receptor by non-small cell lung carcinomas, Cancer Res 53, 3217-3220. 62. Wikstrand, C. J., Hale, L. P., Batra, S. K., Hill, M. L., Humphrey, P. A., et al. (1995) Monoclonal antibodies against EGFRvIII are tumor specific and react with breast and lung carcinomas and malignant gliomas, Cancer Res 55, 3140-3148. 63. Moscatello, D. K., Holgado-Madruga, M., Godwin, A. K., Ramirez, G., Gunn, G., et al. (1995) Frequent expression of a mutant epidermal growth factor receptor in multiple human tumors, Cancer Res 55, 5536-5539.

30 64. Huang, H. S., Nagane, M., Klingbeil, C. K., Lin, H., Nishikawa, R., et al. (1997) The enhanced tumorigenic activity of a mutant epidermal growth factor receptor common in human cancers is mediated by threshold levels of constitutive tyrosine phosphorylation and unattenuated signaling, J Biol Chem 272, 2927-2935. 65. Nishikawa, R., Ji, X. D., Harmon, R. C., Lazar, C. S., Gill, G. N., et al. (1994) A mutant epidermal growth factor receptor common in human glioma confers enhanced tumorigenicity, Proc Natl Acad Sci U S A 91, 7727-7731. 66. Han, W., Zhang, T., Yu, H., Foulke, J. G., and Tang, C. K. (2006) Hypophosphorylation of residue Y1045 leads to defective downregulation of EGFRvIII, Cancer Biol Ther 5, 1361-1368. 67. Lee, J. C., Vivanco, I., Beroukhim, R., Huang, J. H., Feng, W. L., et al. (2006) Epidermal growth factor receptor activation in glioblastoma through novel missense mutations in the extracellular domain, PLoS Med 3, e485. 68. Sharma, S. V., Bell, D. W., Settleman, J., and Haber, D. A. (2007) Epidermal growth factor receptor mutations in lung cancer, Nat Rev Cancer 7, 169-181. 69. Sequist, L. V., Joshi, V. A., Janne, P. A., Bell, D. W., Fidias, P., et al. (2006) Epidermal growth factor receptor mutation testing in the care of lung cancer patients, Clin Cancer Res 12, 4403s-4408s. 70. Slamon, D. J., Godolphin, W., Jones, L. A., Holt, J. A., Wong, S. G., et al. (1989) Studies of the HER-2/neu proto-oncogene in human breast and ovarian cancer, Science 244, 707- 712. 71. Sauter, G., Moch, H., Moore, D., Carroll, P., Kerschmann, R., et al. (1993) Heterogeneity of erbB-2 gene amplification in bladder cancer, Cancer Res 53, 2199-2203. 72. Slamon, D. J., Clark, G. M., Wong, S. G., Levin, W. J., Ullrich, A., et al. (1987) Human breast cancer: correlation of relapse and survival with amplification of the HER-2/neu oncogene, Science 235, 177-182. 73. Fry, D. W. (2003) Mechanism of action of erbB tyrosine kinase inhibitors, Exp Cell Res 284, 131-139. 74. Barker, A. J., Gibson, K. H., Grundy, W., Godfrey, A. A., Barlow, J. J., et al. (2001) Studies leading to the identification of ZD1839 (IRESSA): an orally active, selective epidermal growth factor receptor tyrosine kinase inhibitor targeted to the treatment of cancer, Bioorg Med Chem Lett 11, 1911-1914. 75. Moyer, J. D., Barbacci, E. G., Iwata, K. K., Arnold, L., Boman, B., et al. (1997) Induction of apoptosis and cell cycle arrest by CP-358,774, an inhibitor of epidermal growth factor receptor tyrosine kinase, Cancer Res 57, 4838-4848. 76. Stamos, J., Sliwkowski, M. X., and Eigenbrot, C. (2002) Structure of the epidermal growth factor receptor kinase domain alone and in complex with a 4-anilinoquinazoline inhibitor, J Biol Chem 277, 46265-46272. 77. Pedersen, M. W., Pedersen, N., Ottesen, L. H., and Poulsen, H. S. (2005) Differential response to gefitinib of cells expressing normal EGFR and the mutant EGFRvIII, Br J Cancer 93, 915-923. 78. Kwak, E. L., Sordella, R., Bell, D. W., Godin-Heymann, N., Okimoto, R. A., et al. (2005) Irreversible inhibitors of the EGF receptor may circumvent acquired resistance to gefitinib, Proc Natl Acad Sci U S A 102, 7665-7670. 79. Arteaga, C. L. (2003) ErbB-targeted therapeutic approaches in human cancer, Exp Cell Res 284, 122-130.

31 80. Sato, J. D., Kawamoto, T., Le, A. D., Mendelsohn, J., Polikoff, J., et al. (1983) Biological effects in vitro of monoclonal antibodies to human epidermal growth factor receptors, Mol Biol Med 1, 511-529. 81. Gill, G. N., Kawamoto, T., Cochet, C., Le, A., Sato, J. D., et al. (1984) Monoclonal anti- epidermal growth factor receptor antibodies which are inhibitors of epidermal growth factor binding and antagonists of epidermal growth factor binding and antagonists of epidermal growth factor-stimulated tyrosine protein kinase activity, J Biol Chem 259, 7755-7760. 82. Li, S., Schmitz, K. R., Jeffrey, P. D., Wiltzius, J. J., Kussie, P., et al. (2005) Structural basis for inhibition of the epidermal growth factor receptor by cetuximab, Cancer Cell 7, 301-311. 83. Sunada, H., Magun, B. E., Mendelsohn, J., and MacLeod, C. L. (1986) Monoclonal antibody against epidermal growth factor receptor is internalized without stimulating receptor phosphorylation, Proc Natl Acad Sci U S A 83, 3825-3829. 84. Fan, Z., Lu, Y., Wu, X., and Mendelsohn, J. (1994) Antibody-induced epidermal growth factor receptor dimerization mediates inhibition of autocrine proliferation of A431 squamous carcinoma cells, J Biol Chem 269, 27595-27602. 85. Petit, A. M., Rak, J., Hung, M. C., Rockwell, P., Goldstein, N., et al. (1997) Neutralizing antibodies against epidermal growth factor and ErbB-2/neu receptor tyrosine kinases down-regulate vascular endothelial growth factor production by tumor cells in vitro and in vivo: angiogenic implications for signal transduction therapy of solid tumors, Am J Pathol 151, 1523-1530. 86. Zhang, W., Gordon, M., Schultheis, A. M., Yang, D. Y., Nagashima, F., et al. (2007) FCGR2A and FCGR3A polymorphisms associated with clinical outcome of epidermal growth factor receptor expressing metastatic colorectal cancer patients treated with single-agent cetuximab, J Clin Oncol 25, 3712-3718. 87. Baselga, J. (2001) The EGFR as a target for anticancer therapy--focus on cetuximab, Eur J Cancer 37 Suppl 4, S16-22. 88. Goldstein, N. I., Prewett, M., Zuklys, K., Rockwell, P., and Mendelsohn, J. (1995) Biological efficacy of a chimeric antibody to the epidermal growth factor receptor in a human tumor xenograft model, Clin Cancer Res 1, 1311-1318. 89. Perez-Torres, M., Guix, M., Gonzalez, A., and Arteaga, C. L. (2006) Epidermal growth factor receptor (EGFR) antibody down-regulates mutant receptors and inhibits tumors expressing EGFR mutations, J Biol Chem 281, 40183-40192. 90. Cunningham, D., Humblet, Y., Siena, S., Khayat, D., Bleiberg, H., et al. (2004) Cetuximab monotherapy and cetuximab plus irinotecan in irinotecan-refractory metastatic colorectal cancer, N Engl J Med 351, 337-345. 91. Yang, X. D., Jia, X. C., Corvalan, J. R., Wang, P., and Davis, C. G. (2001) Development of ABX-EGF, a fully human anti-EGF receptor monoclonal antibody, for cancer therapy, Crit Rev Oncol Hematol 38, 17-23. 92. Messersmith, W. A., and Hidalgo, M. (2007) Panitumumab, a monoclonal anti epidermal growth factor receptor antibody in colorectal cancer: another one or the one?, Clin Cancer Res 13, 4664-4666. 93. Murthy, U., Basu, A., Rodeck, U., Herlyn, M., Ross, A. H., et al. (1987) Binding of an antagonistic monoclonal antibody to an intact and fragmented EGF-receptor polypeptide, Arch Biochem Biophys 252, 549-560.

32 94. Kettleborough, C. A., Saldanha, J., Heath, V. J., Morrison, C. J., and Bendig, M. M. (1991) Humanization of a mouse monoclonal antibody by CDR-grafting: the importance of framework residues on loop conformation, Protein Eng 4, 773-783. 95. Graeven, U., Kremer, B., Sudhoff, T., Killing, B., Rojo, F., et al. (2006) Phase I study of the humanised anti-EGFR monoclonal antibody matuzumab (EMD 72000) combined with gemcitabine in advanced pancreatic cancer, Br J Cancer 94, 1293-1299. 96. Seiden, M. V., Burris, H. A., Matulonis, U., Hall, J. B., Armstrong, D. K., et al. (2007) A phase II trial of EMD72000 (matuzumab), a humanized anti-EGFR monoclonal antibody, in patients with platinum-resistant ovarian and primary peritoneal malignancies, Gynecol Oncol 104, 727-731. 97. Fury, M. G., Lipton, A., Smith, K. M., Winston, C. B., and Pfister, D. G. (2007) A phase- I trial of the epidermal growth factor receptor directed bispecific antibody MDX-447 without and with recombinant human granulocyte-colony stimulating factor in patients with advanced solid tumors, Cancer Immunol Immunother. 98. Mateo, C., Moreno, E., Amour, K., Lombardero, J., Harris, W., et al. (1997) Humanization of a mouse monoclonal antibody that blocks the epidermal growth factor receptor: recovery of antagonistic activity, Immunotechnology 3, 71-81. 99. Ramos, T. C., Figueredo, J., Catala, M., Gonzalez, S., Selva, J. C., et al. (2006) Treatment of high-grade glioma patients with the humanized anti-epidermal growth factor receptor (EGFR) antibody h-R3: report from a phase I/II trial, Cancer Biol Ther 5, 375-379. 100. Kuan, C. T., Wikstrand, C. J., Archer, G., Beers, R., Pastan, I., et al. (2000) Increased binding affinity enhances targeting of glioma xenografts by EGFRvIII-specific scFv, Int J Cancer 88, 962-969. 101. Ochiai, H., Archer, G. E., Herndon, J. E., 2nd, Kuan, C. T., Mitchell, D. A., et al. (2007) EGFRvIII-targeted immunotoxin induces antitumor immunity that is inhibited in the absence of CD4+ and CD8+ T cells, Cancer Immunol Immunother. 102. Luwor, R. B., Johns, T. G., Murone, C., Huang, H. J., Cavenee, W. K., et al. (2001) Monoclonal antibody 806 inhibits the growth of tumor xenografts expressing either the de2-7 or amplified epidermal growth factor receptor (EGFR) but not wild-type EGFR, Cancer Res 61, 5355-5361. 103. Johns, T. G., Stockert, E., Ritter, G., Jungbluth, A. A., Huang, H. J., et al. (2002) Novel monoclonal antibody specific for the de2-7 epidermal growth factor receptor (EGFR) that also recognizes the EGFR expressed in cells containing amplification of the EGFR gene, Int J Cancer 98, 398-408. 104. Jungbluth, A. A., Stockert, E., Huang, H. J., Collins, V. P., Coplan, K., et al. (2003) A monoclonal antibody recognizing human cancers with amplification/overexpression of the human epidermal growth factor receptor, Proc Natl Acad Sci U S A 100, 639-644. 105. Johns, T. G., Adams, T. E., Cochran, J. R., Hall, N. E., Hoyne, P. A., et al. (2004) Identification of the epitope for the epidermal growth factor receptor-specific monoclonal antibody 806 reveals that it preferentially recognizes an untethered form of the receptor, J Biol Chem 279, 30375-30384. 106. Panousis, C., Rayzman, V. M., Johns, T. G., Renner, C., Liu, Z., et al. (2005) Engineering and characterisation of chimeric monoclonal antibody 806 (ch806) for targeted immunotherapy of tumours expressing de2-7 EGFR or amplified EGFR, Br J Cancer 92, 1069-1077.

33 107. Li, D., Ji, H., Zaghlul, S., McNamara, K., Liang, M. C., et al. (2007) Therapeutic anti- EGFR antibody 806 generates responses in murine de novo EGFR mutant-dependent lung carcinomas, J Clin Invest 117, 346-352. 108. Johns, T. G., Perera, R. M., Vernes, S. C., Vitali, A. A., Cao, D. X., et al. (2007) The efficacy of epidermal growth factor receptor-specific antibodies against glioma xenografts is influenced by receptor levels, activation status, and heterodimerization, Clin Cancer Res 13, 1911-1925. 109. Scott, A. M., Lee, F. T., Tebbutt, N., Herbertson, R., Gill, S. S., et al. (2007) A phase I clinical trial with monoclonal antibody ch806 targeting transitional state and mutant epidermal growth factor receptors, Proc Natl Acad Sci U S A 104, 4071-4076. 110. Hudziak, R. M., Lewis, G. D., Winget, M., Fendly, B. M., Shepard, H. M., et al. (1989) p185HER2 monoclonal antibody has antiproliferative effects in vitro and sensitizes human breast tumor cells to tumor necrosis factor, Mol Cell Biol 9, 1165-1172. 111. Carter, P., Presta, L., Gorman, C. M., Ridgway, J. B., Henner, D., et al. (1992) Humanization of an anti-p185HER2 antibody for human cancer therapy, Proc Natl Acad Sci U S A 89, 4285-4289. 112. Tokuda, Y., Ohnishi, Y., Shimamura, K., Iwasawa, M., Yoshimura, M., et al. (1996) In vitro and in vivo anti-tumour effects of a humanised monoclonal antibody against c-erbB- 2 product, Br J Cancer 73, 1362-1365. 113. Baselga, J., Norton, L., Albanell, J., Kim, Y. M., and Mendelsohn, J. (1998) Recombinant humanized anti-HER2 antibody (Herceptin) enhances the antitumor activity of and doxorubicin against HER2/neu overexpressing human breast cancer xenografts, Cancer Res 58, 2825-2831. 114. Slamon, D. J., Leyland-Jones, B., Shak, S., Fuchs, H., Paton, V., et al. (2001) Use of chemotherapy plus a monoclonal antibody against HER2 for metastatic breast cancer that overexpresses HER2, N Engl J Med 344, 783-792. 115. Citri, A., Skaria, K. B., and Yarden, Y. (2003) The deaf and the dumb: the biology of ErbB-2 and ErbB-3, Exp Cell Res 284, 54-65. 116. Baselga, J., Perez, E. A., Pienkowski, T., and Bell, R. (2006) Adjuvant trastuzumab: a milestone in the treatment of HER-2-positive early breast cancer, Oncologist 11 Suppl 1, 4-12. 117. Cho, H. S., Mason, K., Ramyar, K. X., Stanley, A. M., Gabelli, S. B., et al. (2003) Structure of the extracellular region of HER2 alone and in complex with the Herceptin Fab, Nature 421, 756-760. 118. Agus, D. B., Akita, R. W., Fox, W. D., Lewis, G. D., Higgins, B., et al. (2002) Targeting ligand-activated ErbB2 signaling inhibits breast and prostate tumor growth, Cancer Cell 2, 127-137. 119. Baselga, J., Albanell, J., Molina, M. A., and Arribas, J. (2001) Mechanism of action of trastuzumab and scientific update, Semin Oncol 28, 4-11. 120. Cuello, M., Ettenberg, S. A., Clark, A. S., Keane, M. M., Posner, R. H., et al. (2001) Down-regulation of the erbB-2 receptor by trastuzumab (herceptin) enhances tumor necrosis factor-related apoptosis-inducing ligand-mediated apoptosis in breast and ovarian cancer cell lines that overexpress erbB-2, Cancer Res 61, 4892-4900. 121. Austin, C. D., De Maziere, A. M., Pisacane, P. I., van Dijk, S. M., Eigenbrot, C., et al. (2004) Endocytosis and sorting of ErbB2 and the site of action of cancer therapeutics trastuzumab and geldanamycin, Mol Biol Cell 15, 5268-5282.

34 122. Segatto, O., King, C. R., Pierce, J. H., Di Fiore, P. P., and Aaronson, S. A. (1988) Different structural alterations upregulate in vitro tyrosine kinase activity and transforming potency of the erbB-2 gene, Mol Cell Biol 8, 5570-5574. 123. Molina, M. A., Codony-Servat, J., Albanell, J., Rojo, F., Arribas, J., et al. (2001) Trastuzumab (herceptin), a humanized anti-Her2 receptor monoclonal antibody, inhibits basal and activated Her2 ectodomain cleavage in breast cancer cells, Cancer Res 61, 4744-4749. 124. Clynes, R. A., Towers, T. L., Presta, L. G., and Ravetch, J. V. (2000) Inhibitory Fc receptors modulate in vivo cytoxicity against tumor targets, Nat Med 6, 443-446. 125. Mendelsohn, J., and Baselga, J. (2000) The EGF receptor family as targets for cancer therapy, Oncogene 19, 6550-6565. 126. Franklin, M. C., Carey, K. D., Vajdos, F. F., Leahy, D. J., de Vos, A. M., et al. (2004) Insights into ErbB signaling from the structure of the ErbB2-pertuzumab complex, Cancer Cell 5, 317-328. 127. Adams, C. W., Allison, D. E., Flagella, K., Presta, L., Clarke, J., et al. (2006) Humanization of a recombinant monoclonal antibody to produce a therapeutic HER dimerization inhibitor, pertuzumab, Cancer Immunol Immunother 55, 717-727. 128. Agus, D. B., Sweeney, C. J., Morris, M. J., Mendelson, D. S., McNeel, D. G., et al. (2007) Efficacy and safety of single-agent pertuzumab (rhuMAb 2C4), a human epidermal growth factor receptor dimerization inhibitor, in castration-resistant prostate cancer after progression from taxane-based therapy, J Clin Oncol 25, 675-681. 129. Nahta, R., Hung, M. C., and Esteva, F. J. (2004) The HER-2-targeting antibodies trastuzumab and pertuzumab synergistically inhibit the survival of breast cancer cells, Cancer Res 64, 2343-2346. 130. Hanahan, D., and Weinberg, R. A. (2000) The hallmarks of cancer, Cell 100, 57-70. 131. Aboud-Pirak, E., Hurwitz, E., Pirak, M. E., Bellot, F., Schlessinger, J., et al. (1988) Efficacy of antibodies to epidermal growth factor receptor against KB carcinoma in vitro and in nude mice, J Natl Cancer Inst 80, 1605-1611. 132. Hynes, N. E., and Lane, H. A. (2005) ERBB receptors and cancer: the complexity of targeted inhibitors, Nat Rev Cancer 5, 341-354. 133. Herbst, R. S., Fukuoka, M., and Baselga, J. (2004) Gefitinib--a novel targeted approach to treating cancer, Nat Rev Cancer 4, 956-965. 134. Johns, T. G., Luwor, R. B., Murone, C., Walker, F., Weinstock, J., et al. (2003) Antitumor efficacy of cytotoxic drugs and the monoclonal antibody 806 is enhanced by the EGF receptor inhibitor AG1478, Proc Natl Acad Sci U S A 100, 15871-15876. 135. Friedman, L. M., Rinon, A., Schechter, B., Lyass, L., Lavi, S., et al. (2005) Synergistic down-regulation of receptor tyrosine kinases by combinations of mAbs: implications for cancer immunotherapy, Proc Natl Acad Sci U S A 102, 1915-1920. 136. Perera, R. M., Narita, Y., Furnari, F. B., Gan, H. K., Murone, C., et al. (2005) Treatment of human tumor xenografts with monoclonal antibody 806 in combination with a prototypical epidermal growth factor receptor-specific antibody generates enhanced antitumor activity, Clin Cancer Res 11, 6390-6399. 137. Huang, P. H., Mukasa, A., Bonavia, R., Flynn, R. A., Brewer, Z. E., et al. (2007) Quantitative analysis of EGFRvIII cellular signaling networks reveals a combinatorial therapeutic strategy for glioblastoma, Proc Natl Acad Sci U S A 104, 12867-12872.

35 Chapter 2: Epitope mapping of proteins binding to EGFR

2.1 Background

Epitope mapping is the process of determining which antigen residues are responsible for mediating antibody-antigen interactions. This information can be important in understanding the mechanism of action of the antibody and can aid in the design of novel antibodies against the same antigen. One method of determining the epitope of an antibody is through co-crystallization

of the antibody in complex with its antigen. However, this can be very time-consuming and

technically difficult, especially for a complex antigen such as EGFR. Only within the past few

years was the EGFR monomer itself crystallized (1), and the only antibody-EGFR complex

structure was published recently (2). While co-crystallization is the “gold standard” of

characterizing the epitope of an antibody, other methods exist to determine which residues of the

antigen are important for antibody binding.

One class of epitope mapping methods utilizes peptide fragments of the antigen.

Antibody reactivity to each of these fragments is tested, and the epitope can be localized to

certain stretches of amino acids. SPOT synthesis involves the creation of synthetic peptides,

which are spotted on cellulose membranes and assayed for antibody binding (3). Alternately, peptides can be expressed on the surface of bacteriophage (4), Escherichia coli (5), or yeast (6), with subsequent antibody binding analysis. Mass spectrometry has also be used to identify digested peptides of the antigen which retain binding to an antibody (7). Phage display and

SPOT techniques have been utilized to determine the epitopes of various antibodies against ErbB receptor family members (8-10). However, peptide-based methods are limited in that they can only identify continuous, non-conformational epitopes. If the binding of the antibody depends on

36 the proper tertiary fold of the antigen, other methods must be used to determine its epitope.

Peptides have been used to mimic epitopes discontinuous in sequence (11,12), although it is not guaranteed that for any given epitope, a peptide “mimotope” exists.

To identify a discontinuous epitope, H/2H-exchange mass spectrometry has been used to localize an epitope to discontinuous proteolytic fragments (13). Antibody epitopes can also be resolved to protein fragments or domains displayed on the surface of phage (14) or yeast (15).

Another useful tool in dissecting protein-protein interactions is site-directed mutagenesis, in which residues of interest are mutated and subsequent changes in binding are measured. This allows for characterization of antibody binding in the context of the entire protein antigen.

Initially, only conservative mutations to amino acid homologs (16) or alanine (17) were made to minimize disruption to the protein secondary and tertiary structure. This method requires soluble protein expression and characterization of each mutant to ensure proper folding, which can be time consuming. Also, some knowledge of the general location of the epitope is necessary to indicate which site-directed mutants should be constructed. Alternately, shotgun alanine or homolog scanning mutagenesis is a high-throughput method using phage display libraries and can incorporate saturation mutagenesis so that no knowledge of the epitope is necessary. Shotgun scanning has been used to map paratopes and protein-protein interactions (18-20). However, the prokaryotic expression system used by this method may not be amenable to epitope mapping of complex eukaryotic proteins such as EGFR ectodomain, which contains 25 disulfide bonds and

10 N-linked glycosylation sites (21). Finally, other methods have utilized random mutations to all amino acids in order to disrupt antibody-antigen interactions. Random mutagenesis of the antigen followed by secretion by yeast in 96-well plates enabled epitope mapping in a semi-high- throughput manner (22). Libraries of randomly mutagenized antigen have also been displayed on

37 the surface of filamentous phage (23) and mammalian cells (24), and selection of nonbinding

mutants was performed in a combinatorial manner in these studies.

To overcome some of the shortcomings of existing techniques, we developed a novel method of epitope mapping using random mutagenesis and yeast surface display. In this method, a library of mutated antigen was displayed on the surface of yeast and screened using fluorescence-activated cell sorting (FACS). We applied this method to antibodies and designed ankyrin repeat proteins (DARPins) against EGFR, and we successfully determined the epitopes of several proteins, many of which bind in a conformation-specific manner.

38 2.2 Materials and methods

Construction and expression of epitope mapping libraries

The epitope mapping libraries were constructed using the Stratagene GeneMorph random

mutagenesis to give a low mutagenesis rate. The template used for library construction was a

pCT302 yeast-display plasmid backbone containing either the EGFR fragment 273-621 with a

C283A mutation to prevent disulfide mispairing (15) (Appendix 1.2) or the 404SG EGFR

ectodomain mutant (25) (Appendix 1.3). The PCR products were gel purified and extracted using

a Qiagen Qiaquick gel extraction kit. The library was transformed into Saccharomyces cerevisiae

strain EBY100 (26) by electroporation and homologous recombination using a Bio-Rad

(Richmond, CA) Gene Pulser Transfection Apparatus (27). The final library contained a roughly

Poisson distribution of amino acid changes to the EGFR fragment as demonstrated by plasmid recovery using a Zymoprep kit (Zymo Research) and sequencing of 100 library clones (MIT

Biopolymers Laboratory). Yeast were grown in SDCAA media overnight at 30 oC and were

o induced to express protein on the cell surface by induction at a starting OD600 of 0.5 at 30 C in

SGCAA media (27) with a 1/25 volume of SDCAA.

Labeling and sorting of yeast

The anti-human EGFR mouse monoclonal antibodies 806 and 175 were generously provided by the Ludwig Institute for Cancer Research, mAb 13A9 was provided by Genentech

(San Francisco, California), and mAb EMD72000 was provided by EMD-Lexigen

Pharmaceuticals. mAb 225 was purchased from LabVision (Fremont, CA), anti-c-myc chicken

IgY fraction was purchased from Invitrogen (Carlsbad, CA), and mAbs 528 (28), 199.12,

EGFR1 (29), and ICR10 (30) were purchased from Abcam (Cambridge, UK).

39 Yeast cells were labeled as described (27). Briefly, yeast cells (at least 10X library size)

were washed with PBSA buffer (phosphate buffered saline containing 1 mg/ml bovine serum

albumin). The cells were incubated with 1:250 anti-c-myc chicken IgY and the appropriate

concentration of mAb for 30 min at 25 oC. The cells were then washed with PBSA and incubated

with 1:50 phycoerythrin (PE)-labeled goat anti-mouse IgG (Sigma) or 1:100 goat anti-human-PE

IgG (Rockland) and 1:100 goat anti-chicken-Alexa488 IgG (Invitrogen) for 20 min at 4 oC. The labeled cells were washed with 1 ml PBSA, and cell libraries were sorted using either a MoFlo

(DakoCytomation) or Aria (BD Biosciences) FACS machine at the MIT flow cytometry core facility. In addition, a positive control of antibody binding to wild-type antigen was labeled to assist in drawing the selection window.

Identification and testing of single clones

Plasmids from the sorted library populations were recovered using a Zymoprep kit and

sequenced at the MIT Biopolymers Laboratory. Site-directed mutants were made using

QuickChange mutagenesis (Stratagene). Single clones were transformed into yeast using EZ

Yeast Transformation (Zymo Research) and grown and induced as above. For each clone, 106 cells were labeled as before in a final volume of 50 μL. Fluorescence data was obtained using a

Coulter Epics XL flow cytometer (Beckman-Coulter) and was analyzed using either

DakoCytomation Summit or FloJo (Tree Star, Inc.) software.

Titration of EGFR fragments against mAb806

Cells were grown, induced, and labeled as above. Mean fluorescence data of c-myc positive yeast were obtained using a Coulter Epics XL flow cytometer and were normalized by

40 maximal and minimal mean fluorescence intensities. The binding interaction was assumed to be

a single site binding model with no ligand depletion. Titration data was fit to the equation:

[mAb] fmAb = (1) [mAb]+ Kd

where fmAb is the fractional binding of mAb 806 to yeast-surface displayed EGFR 273-621,

[mAb] is the concentration of mAb 806, and Kd is the apparent dissociation constant. A global fit

of three data sets was performed using Microsoft Excel, and 68% confidence intervals were calculated according to (31).

Protein images and surface area calculations

All EGFR protein images were generated using PyMOL software (DeLano Scientific

LLC, at http://www.pymol.org). The solvent accessible surface area of each residue of EGFR

(using structure 1NQL) was calculated using Getarea 1.1 (Sealy Center for Structural Biology,

University of Texas Medical Branch, at http://www.scsb.utmb.edu/cgi-bin/get_a_form.tcl). A

water probe size of 1.0 was used to allow for correct identification of EGF contact residues as

being on the surface. Residues with a value of 20 or above were considered surface residues.

DARPin epitope mapping

Designed ankyrin repeat protein (DARPin) clones E1, E67, E68, E69, and E72 containing

C-terminal His tags were provided by Dr. Andreas Plueckthun (University of Zurich). DARPin

binding to 404SG on the surface of yeast was detected using 1:100 biotinylated mouse Penta-His

antibody (Qiagen) followed by 1:50 streptavidin-PE (Invitrogen). All other cell growth, sorting,

and clone analysis was performed as above.

41 2.3 Results

2.3.1 Creation of epitope mapping library

In order to determine the epitopes of antibodies binding to EGFR, we created a novel

method of epitope mapping using random mutagenesis and yeast surface display. Previously,

Cochran et al. reported domain-level epitope mapping through yeast surface-displayed fragments

of EGFR (15). Large polypeptide fragments of EGFR were expressed on the surface of yeast

cells and were used to localize antibody binding to particular regions of the receptor for both

continuous and discontinuous epitopes. Yeast surface display is a tool originally developed for

protein engineering (32), in which the gene for a protein of interest is fused the yeast native Aga2

protein. The yeast cells are then induced to display this protein fusion on the cell wall, where it is

accessible for binding to other soluble proteins (Figure 2-1).

Figure 2-1: Schematic of yeast surface display. The yeast native proteins Aga1 and Aga2 are displayed on the yeast cell wall, allowing the Aga2p-EGFR to be displayed. Fluorescent antibody against either the HA or c-myc epitope tag allows for the detection of EGFR on the yeast surface, while antibody binding to the EGFR can be detected through a differently colored fluorescent probe.

42 In creating our novel fine epitope mapping method, we sought to further resolve antibody

epitopes to individual residues involved in antibody-antigen contacts. To do so, we specifically

disrupted the interaction between the antibody being mapped and the EGFR antigen through

point mutations. If a single mutant displayed of a loss of binding to the antibody, then that particular residue was potentially a contact residue. However, mutating each individual residue of the EGFR ectodomain (621 amino acids) would be prohibitively time-consuming. To make the process high-throughput, we employed random mutagenesis of a fragment of the EGFR ectodomain (residues 273-621), and transformed this library into yeast. Then, using yeast surface display, we could induce each yeast cell to display a different EGFR mutant on its surface. High- throughput screening of this library for EGFR mutants with a loss of antibody binding was accomplished using fluorescence-activated cell sorting (FACS). Once the nonbinding population was enriched, the individual clones were sequenced to identify potential epitope residues. The process of epitope mapping using yeast display is summarized in Figure 2-2.

A low mutation-rate library was constructed by error-prone PCR of the EGFR fragment

273-621. The initial library size was 5x105, and sequencing of 100 clones from the unscreened library indicated that 72% were wild-type EGFR, 17% contained single amino acid mutations, and 11% contained multiple mutations or a frameshift. This corresponds to an effective library size of 8.5x104, which is an order of magnitude higher than the largest theoretical diversity of

single amino acid mutants of this 349-residue protein fragment (6.6x103), and nearly two orders

of magnitude higher than the 1.0x103 diversity of amino acids accessible by a single nucleotide

mutation. Yeast were transformed with the DNA library, and the cells were induced the express

the EGFR mutants on their surfaces. Using this epitope mapping method, we sought to determine

the epitopes of several therapeutically relevant antibodies against EGFR.

43

Figure 2-2: Summary of epitope mapping method. A mutant library of the antigen is transformed into yeast, and the yeast cells are induced to express the mutant antigens on their surfaces. The antibody to be mapped is added to the yeast mixture, and cells with reduced antibody binding are isolated using FACS.

2.3.2 Epitope mapping of mAb 806

mAb 806 against EGFR served as a positive control since its epitope had previously been

shown to reside in the disulfide-bonded loop between EGFR residues 287-302 (33). In order to map the mAb 806 epitope, the library of EGFR mutants displayed on the surface of yeast was labeled with chicken anti-c-myc IgY to detect full-length EGFR 273-621 expression. The cells were also labeled with a high concentration of mAb 806 to allow for differentiation between cells possessing wild-type binding and those with a decrease in binding affinity. The library was sorted using successive rounds of FACS to isolate those cells displaying mutants with a decrease in affinity to mAb 806 (Figure 2-3).

44

Figure 2-3: Flow cytometry data for mAb 806 binding to epitope mapping library. EGFR expression as detected through the c-myc tag is shown on the abscissa, and mAb 806 binding is shown on the ordinate. The yeast EGFR mutant library was labeled with 10 nM mAb 806 for sort 1 and 75 nM mAb 806 for sort 2. The yeast were also labeled with chicken anti-c-myc IgY, followed by goat anti-chicken-Alexa488 IgG and goat anti-mouse-PE IgG. The cell collection gates are indicated by the dotted lines, and the double negative population results from yeast plasmid loss.

Three sorts were performed for mAb 806, with round 1 labeled at 10 nM mAb 806, and rounds 2 and 3 labeled at 75 nM mAb 806. Clone isolation and DNA sequencing were performed after sorts 2 and 3. Out of 100 sequenced clones, ~20% contained multiple amino acid mutations and were omitted from subsequent analysis to facilitate interpretation of the data and to eliminate spurious neutral mutations. The single mutants isolated from the library for loss of binding to mAb 806 are listed in Table 2-1.

45

EGFR mutant mAb 806 binding C287G,R,S,W,Y + E293D,G + E293K − D297Y + G298D,S − R300C + R300P − K301E + C302F,R,Y + C302G,S − Table 2-1: EGFR mutants isolated for loss of binding to mAb 806. Binding scores were determined by testing for binding at 75 nM mAb 806 to mutants displayed on the surface of yeast. + indicates a partial loss of binding and − indicates a large loss of binding.

After the mutations were sequenced, the plasmids containing the EGFR mutants were re- transformed into yeast, and each mutant was retested to ensure that all mutations were correctly identified as having reduced binding to mAb 806. The mutations were also scored according to the degree of binding loss at a concentration of 75 nM mAb 806 (Table 2-1). Example flow cytometry data for various EGFR mutants binding to mAb 806 is shown in Figure 2-4.

46

Figure 2-4: Flow cytometry data for representative EGFR mutants and their binding scores to 75 nM mAb 806. Yeast displaying the mutants were also labeled with chicken anti-c-myc IgY, followed by goat anti-chicken- Alexa488 IgG and goat anti-mouse-PE IgG. ++ indicates wild-type binding, + indicates a partial loss of binding, and − indicates a large loss of binding (equal to negative control at this concentration).

To quantitatively correlate mAb 806 labeling scores with particular affinity constants,

binding titrations on the yeast surface were performed to determine the apparent dissociation

constant (Kd) of mAb 806 for wild-type EGFR 273-621 (++), mutant C287R (+), and mutant

E293K (−). The titration curves are shown in Figure 2-5. The data were fit to a single site

binding model with no depletion. Although the bivalent mAbs could potentially result in avidity

effects, these apparent dissociation constants could still be used for comparison between EGFR

mutants. The apparent dissociation constant of mAb 806 for yeast surface-displayed wild-type

EGFR 273–621 is 2.13 nM (68% confidence interval of 1.83–2.50 nM), which is consistent with

47 the affinity of mAb 806 binding to mammalian cells expressing EGFR vIII (34). The C287R

substitution increases the Kd to 127 nM (68% confidence interval of 103–160 nM), which gives a

ΔΔG value of +2.4 kcal/mol when compared to wild-type and, in alanine scanning terms, is a significant loss of binding (35). The E293K substitution leads to a Kd value of at least 30 μM,

corresponding to a ΔΔG of at least +5.7 kcal/mol. This large loss of binding is consistent with the

charge-reversing nature of the amino acid substitution. The above ΔΔG values demonstrate the

relative energetic importance of these mutations, and the results from these three representative

clones were used to roughly estimate the energetic importance of the other untitrated mutants

based on their binding scores.

1

0.5 , fractional binding , fractional mAb f

0 0.001 0.01 0.1 1 10 100 1000 10000 [mAb 806], nM

Figure 2-5: Titration of mAb 806 against yeast surface-displayed EGFR 273-621 and mutants. Red, wild-type (++); green, C287R (+); blue, E293K (−). A global fit to a single site binding model was performed with three independent sets of data. Square, triangles, and diamonds represent separate data sets.

To verify the results from the fine epitope mapping method, we performed alanine

scanning through site-directed mutagenesis on the residues comprising the EGFR disulfide-

bonded loop 287-302 (Table 2-2). All amino acid sites with a loss of mAb 806 binding by

48 substitution by alanine corresponded to mutants identified using our yeast-displayed library method (287, 293, 298, and 302). Conversely, residues 297, 300, and 301 were not identified as important for mAb 806 binding by alanine scanning, yet form an energetically important component of the epitope as determined by our method. Clearly, using mutations to residues other than alanine allows for greater discrimination of the epitope of an antibody.

EGFR mutant mAb 806 binding C287A + G288A ++ A289K ++ D290A ++ S291A ++ Y292A ++ E293A + M294A ++ E295A ++ E296A ++ D297A ++ G298A + V299A,K ++ V299D − R300A ++ K301A ++ C302A − Table 2-2: EGFR mutants created by site-directed mutagenesis. Binding was tested at a concentration of 75 nM mAb 806 against mutants displayed on the surface of yeast. ++ indicates wild-type binding, + indicates a partial loss of binding, and − indicates a large loss of binding.

All of the mutations leading to a decrease in mAb 806 binding were localized to the disulfide loop between cysteine residues 287 and 302, consistent with published results (33).

However, the residue-level resolution afforded by our method provides additional insight into the mAb 806 binding epitope. The residues identified as energetically important for mAb 806 binding are shown in terms of the EGFR structure are shown in Figure 2-6. These residues are

49 clustered on one face of the disulfide loop between cysteines 287 and 302, indicating that this region is the mAb 806 binding epitope. Interestingly, Val299 is located in the middle of these residues, but was not identified in the library or alanine scan. Thus, the V299K and V299D site-

directed mutants were created and tested, and although the epitope can accommodate a lysine

residue without effect, an aspartic acid substitution ablates detectable binding (Table 2-2). This further indicates that mAb 806 is likely to contact this face of loop 287–302.

Although mAb 806 binds to heat and SDS-denatured EGFR (33), implying a linear binding epitope, the library analysis performed here suggests that the epitope is not entirely continuous in sequence. Examination of the structure of the loop shows that the sidechain of

Glu293 projects from one side of the loop to the putative contact face. In addition, the mAb 806 epitope is constrained by a disulfide bond between cysteines 287 and 302, and two salt bridges

(Glu293-Arg300 and Asp297-Lys301) (Figure 2-6c). All six residues involved in these

interactions were isolated from the library for loss of binding to mAb 806, highlighting the

importance of the constrained nature of the epitope for high affinity binding. It is unclear

whether these residues directly make contact with the antibody, or if they are only important for

constraining the peptide and reducing its entropy. The cysteine at position 287 tolerated a wide

variety of substitutions that led to an intermediate loss of binding (Table 2-1), indicating that its

energetic contribution to the epitope may arise more from constraining the loop rather than

directly contacting the antibody. In contrast, the cysteine at position 302 may be a contact

residue since it only tolerated substitutions to larger residues with aromatic character.

50

Figure 2-6: mAb 806 binding epitope on EGFR. (a) and (b) Front and back views of residues 287-302 in chain a of the EGF-bound dimerized EGFR structure (1IVO), which was used because Glu293 is not resolved in the monomer structure (1NQL). Residues shown in color are those which were identified as important for mAb 806 binding using our yeast library method. Red, residues that also caused loss of binding upon alanine substitution; orange, residues that did not; gray, residues that were not identified as important for binding by both library and alanine scanning methods. (c) The epitope is constrained by a disulfide bond and two salt bridges. Negatively charged residues, red; positively charged residues, blue; cysteine residues, yellow. Image includes residues 287-302 on both EGFR molecules in dimer structure (1IVO). (c) mAb 806 epitope in context of full tethered EGFR monomer (1NQL), colored as in (a), with the rest of EGFR in blue.

51 The mAb 806 epitope in context of the tethered EGFR monomer structure (1) is shown in Figure

2-6d. From viewing the structure, it appears that in this conformation, an antibody would be

partially blocked from accessing the epitope. This is consistent with the observation that mAb

806 does not bind soluble EGFR, but does bind to an untethered EGFR mutant perturbed away

from the autoinhibited conformation (33).

2.3.3 Effects of EGFR mutations on binding of mAbs 806 and 175

Because of the convenience of performing titrations on the surface of yeast, the binding

of mAb 806 to additional EGFR mutants was characterized in further studies. This work was

conducted in collaboration with Arvind Sivasubramanian and Jeff Gray at Johns Hopkins

University. We combined computational protein-protein docking with computational and

experimental mutagenesis to predict the structure of the complex formed by mAb 806 and

EGFR. Initially, Arvind docked mAb 806 to its peptide epitope of EGFR residues 287-302.

Potential mAb 806-EGFR structures, or decoys, were generated, and computational mutagenesis

was used to filter them according to their agreement with the previously generated experimental

mutagenesis data in Tables 2-1 and 2-2. Further computational mutagenesis suggested additional

mutations, which we tested to differentiate between three potential docked models. The site- directed mutants were created and transformed into yeast, and mAb 806 binding to these mutants was scored as before in Section 2.3.2. The results are listed in Table 2-3, and they allowed

Arvind to arrive at a final structure with the highest agreement between computational and

experimental mutagenesis data, which is described in (36).

52

EGFR mutant mAb 806 binding E293Y − E293F − E293W − M294Y ++ M294F ++ M294W ++ E295Y ++ E295F ++ E295W ++ E296W ++ D297W + V299Y ++ R300Y ++ R300F ++ R300W ++ K301W ++ Table 2-3: mAb 806 binding to EGFR mutants used to discriminate between computational models. Binding was tested at a concentration of 75 nM mAb 806 against mutants displayed on the surface of yeast. ++ indicates wild- type binding, + indicates a partial loss of binding, and − indicates a large loss of binding.

Another antibody, mAb 175, was also characterized using binding to EGFR mutants

displayed on the surface of yeast. mAb 175 was isolated at the same time as mAb 806, and

displays similar reactivity to the peptide of EGFR residues 287-302 (T.G. Johns et al.,

manuscript in preparation). Therefore, the mutants originally tested for binding to mAb 806 were

tested for mAb 175 reactivity. The binding scores for mAb 175 compared to those for mAb 806

are shown in Table 2-4. The results show only minor binding differences between mAbs 175 and

806, with mAb 175 showing slightly reduced binding to mutants D297A, V299A, and V299K.

This is consistent with the crystal structures of mAbs 175 and 806 each bound to the EGFR

peptide 287-302, which show identical contact residues for the two antibodies (T.G. Johns et al.,

manuscript in preparation).

53

EGFR mutant mAb 175 mAb 806 binding binding C287A,G,R,S,W,Y + + G288A ++ ++ A289K ++ ++ D290A ++ ++ S291A ++ ++ Y292A ++ ++ E293A,D,G + + E293K − − M294A ++ ++ E295A ++ ++ E296A ++ ++ D297A + ++ D297Y + + G298A + + G298D,S − − V299A,K + ++ V299D − − R300A ++ ++ R300C + + R300P − − K301A ++ ++ K301E + + C302A,G,S − − C302F,R,Y + + Table 2-4: Binding of mAb 175 to EGFR mutants compared to mAb 806. Binding was tested at a concentration of 75 nM mAb 175 or mAb 806 against mutants displayed on the surface of yeast. ++ indicates wild-type binding, + indicates a partial loss of binding, and − indicates a large loss of binding.

2.3.4 Epitope mapping of mAbs 225 and 13A9

The binding of mAbs 806 and 175 to EGFR did not depend on the correct tertiary fold of the protein; therefore, their epitopes were not strictly conformational. However, epitope mapping of mAbs 225 and 13A9 presented a greater challenge because these two antibodies have conformational epitopes (15). mAb 225 is the murine precursor to the FDA-approved antibody drug cetuximab (Section 1.4). Both antibodies bind to EGFR domain III, compete with each other (data not shown), and compete with TGF-α ligand binding (37,38). However, mAb 225

54 competes with EGF ligand binding, while mAb 13A9 does not (38,39). This is unexpected,

because it appears from the crystal structures of EGF and TGF-α bound to EGFR that the two

ligands bind in the same location (40,41). We sought to determine the epitopes of these

antibodies to better understand their behavior and mechanisms of action.

For epitope mapping of mAb 225, the EGFR mutant library described in Section 2.3.1

was screened for loss of binding to 50 nM mAb 225 for sort 1 and 100 nM for sorts 2-4, with clone sequencing after sorts 3 and 4. Out of 185 clones sequenced, ~40% contained multiple amino acid mutations and were omitted from subsequent analysis. In total, 84 unique single amino acid substitutions were isolated from the library and were re-transformed and retested in yeast to confirm loss of mAb 225 binding. These mutations are shown in Figure 2-7a. The majority of the mutations are localized or adjacent to domain III, consistent with the observation that mAb 225 blocks ligand binding to this domain. Since mAb 225 recognizes a conformational epitope (15), a properly folded domain III is required for antibody binding. Clearly, all of the residues shown in Figure 2-7a cannot comprise the mAb 225 epitope; therefore, it seems that

most of these mutations are instead causing a perturbation of the domain III conformation,

leading to the observed loss of mAb 225 binding. As shown by this analysis, it appears that

domain III is very sensitive to conformational perturbation through mutation. Although it was

expected that the yeast secretory quality control apparatus would not allow misfolded mutants to

be displayed on the cell surface (42,43), it appears that many of these conformationally altered

EGFR mutants are able to circumvent quality control. In fact, the intertwined disulfides Cys482-

Cys491 and Cys486-Cys499 of EGFR domain IV are required for the proper folding of domain

III (44), and, accordingly, all of these mutations were isolated from the library for loss of binding

to mAb 225.

55

Figure 2-7: (a) All single amino acid mutation sites isolated from the library that exhibited a loss of binding to mAb 225, yellow. EGFR fragment 273-621 (1NQL) is shown in surface representation in grey. (b) and (c) Front and back view of surface only mutation sites that were not mutated to/from proline, glycine, or cysteine or from a glycosylation site, yellow. (d) Residues identified after antibody cross-testing; mAb 225 residues, yellow; mAb 13A9 residues, red. (Note: Lys465 and Glu472 side chains are not resolved in this structure.)

56 In an attempt to extract useful contact residue information from these mutational data, we

developed two criteria to eliminate those mutations which seemed particularly disruptive to

protein folding. First, we only considered mutations of surface residues (based on solvent- accessible surface area), since mutations to the protein core could be expected to disrupt protein

conformation. The mutations with loss of binding to mAb 225 and disregarded from analysis in

this step are listed in Table 2-5, top. However, the remaining mutations were still dispersed

across the surface of EGFR domain III in a fashion that would not allow simultaneous contact

with a single antibody-binding surface. Mutations to or from a proline, glycine, or cysteine

residue could potentially perturb the EGFR domain III secondary structure, and such mutations

were found to be abundantly represented amongst the pool of non-binding mutants. Therefore,

we applied a second criterion to disregard mutations of a potentially disruptive nature, and

residues that were mutated to or from a proline, glycine, or cysteine or from a glycosylation site

were eliminated from the mAb 225 epitope analysis (Table 2-5, middle). Since elimination of all

of these mutations is overly restrictive, the sites eventually identified by this procedure can

formally only be a subset of the energetically significant contact residues. While the use of these

two criteria reduced the number of potential mAb 225 contact residues, the remaining mutations

still were not localized to a plausible contact surface (Figure 2-7b,c and Table 2-5, bottom).

57

Table 2-5: EGFR mutants with loss of binding to mAb 225 Surface inaccessible: C313Y F352Y A385V V437A G315R H356Y W386C I439N S326Y T360N A395S C475Y N331Y L363Q F396S,V,Y C486S F335S L368M G404D,S G490D I341S L371M S413F C491R,Y G343D T378I A415T W492C L345H F380L L429F,H G493D I347N L381F,S I432K C499Y A351E I383N G435R T546I Mutated to/from Pro, Gly, Cys, glycosylation site: V312G P361H R427C C482Y D344G P362T S428P S487P P349L,T P365L G441E,V N544I G354S R403C G471R Remaining mutants: V312A V350M K465E I467M S324L T358I I466F S487F A329T T373A,I

mAb 13A9, another antibody that binds to a conformational epitope, was also screened against the EGFR mutant library. The library was labeled at 50 nM mAb 13A9 for sort 1 and 100 nM for sorts 2-5, with individual clone analysis after sort 5. Out of 170 clones sequenced, ~40% contained multiple amino acid mutations and were omitted from subsequent analysis. A total of

73 unique single amino acid mutations were identified as leading to a loss of binding to mAb

13A9. As seen with mAb 225, the majority of mutations were located in or adjacent to domain

III but were not localized to an epitope. The same criteria used to eliminate potentially misfolded mAb 225 mutants were applied to the mAb 13A9 mutants (Table 2-6). Notably, many of the

eliminated mutants were isolated for loss of binding to both mAb 13A9 and mAb 225. This further supports the hypothesis that these mutants are conformationally altered such that they can

58 no longer bind conformation-specific antibodies and justifies their elimination from the epitope mapping analysis.

Table 2-6: EGFR mutants with loss of binding to mAb 13A9 Surface inaccessible: G315E,R T360A G404D C475Y F321V L363Q,R L414I,P C486Y S326P V374A S423F C491Y I327K G379W L429F C499G F335I L381S I432M T546I C338Y A385V V437M C567R T339I P387L,S N442D P572L I341N L393H,P C446S K585N L345H F396C,V F457L A351E,S,V L399I N469I Mutated to/from Pro, Gly, Cys, glycosylation site: G317D P365L,S G441R C482Y G354A,C R403C G471A,S N544I P362H Remaining mutants: K322I D344V K443N E472D,K A329V T358I S468I S487F E388K H359Y

The recurrence of the same mutations in the mAb 225 and mAb 13A9 sorts led to the concept of antibody cross-tests, in which the conformation-specific antibodies were used as probes for the proper folding of EGFR domain III. We hypothesized that loss of binding to only one but not the other antibody indicated a specific, energetically important contact residue for the antibody that showed loss of binding to the mutant. Representative antibody cross-tests are shown in Figure 2-8. For example, the EGFR mutant I467M does not bind to mAb 225 but shows wild-type binding for mAb 13A9. Therefore, the loss of binding to mAb 225 is most likely due to the mutation of a specific contact residue for the antibody, and not due to protein misfolding. Although unlikely, the possibility exists that a mutation might cause a

59 conformational change in domain III that permits binding of one antibody but not the other, therefore generating a false positive. For the EGFR mutant A329V, loss of some degree of binding to both mAb 225 and mAb 13A9 indicates that it is either altered in domain III conformation or the residue is in an overlapping portion of the two antibody epitopes. Such mutants with loss of binding to both antibodies were eliminated from the analysis, although this will also eliminate any residues that form important energetic contacts with both antibodies.

Using the antibody cross-tests, we were able to identify two epitope residues each as energetically important for the binding of mAbs 225 and 13A9. EGFR mutants K465E and

I467M showed loss of binding to mAb 225, but retained binding to mAb 13A9; therefore,

Lys465 and Ile467 were implicated as energetically significant contact residues for mAb 225.

Likewise, S468I and E472K showed loss of binding to mAb 13A9, but retained binding to mAb

225; therefore, Ser468 and Glu472 were identified as the mAb 13A9 epitope. The epitopes in context of EGF and TGF-α ligand binding are shown in Figure 2-9.

60

Figure 2-8: Flow cytometry data for antibody cross-tests of mAbs 225 and 13A9 binding to yeast surface-displayed EGFR 273-621 mutants at 50 nM mAb. EGFR display fluorescence is shown on the abscissa, and mAb binding is shown on the ordinate. Negative controls and wild-type data are shown along with mutants isolated from the library.

61

Figure 2-9: mAb 225 and mAb 13A9 epitopes in context of ligand binding. Blue, mAb 225 epitope; red, mAb 13A9 epitope; green, ligands. (a) Domain III of EGF-bound EGFR dimer (chain a of 1IVO). (b) Domain III of TGF-α- bound EGFR dimer (chain a of 1MOX).

The experimental work leading to the identification of the epitopes of mAbs 806, 225, and 13A9

(described in Sections 2.3.1-2.3.2 and 2.3.4) was published in (45). After the publication of this paper, the crystal structure of the Fab fragment of mAb 225 bound to EGFR was solved (2). The residues Lys465 and Ile467, which were identified by our library method, were found to be in the

62 center of the mAb 225/EGFR complex interface; thus, the results of the epitope mapping method

have been independently verified by co-crystallization.

2.3.5 Epitope mapping of mAb EMD72000

In a partnership with EMD-Lexigen Pharmaceuticals, the epitope of mAb EMD72000

was determined. mAb EMD72000 is a humanized antibody, and like mAb 225, inhibits the binding of EGF to EGFR (46). However, initial experiments using EGFR fragments on the

surface of yeast revealed that mAb EMD72000 did not bind EGFR 273-621. Although mAbs

225 and 13A9 were able to detectably bind EGFR 273-621, at least a portion of the protein

displayed on the yeast surface was misfolded and thus unreactive to the antibodies, as evidenced

by the lower fluorescence signal at saturating binding conditions compared to mAb 806 binding

(Figure 2-8 vs. Figure 2-4). mAb EMD72000 binding may be extremely sensitive to

conformational perturbation or yeast hyperglycosylation (O-linked or hypermannosylation of N-

linked sites). Although the full EGFR ectodomain is not correctly expressed by yeast, previous

work has identified an EGFR ectodomain mutant, 404SG, which is properly folded on the

surface of yeast (25). 404SG contains two point mutations in domain I (A62T and L69H) and

two mutations in domain III (F380S and S418G). This receptor is able to bind the conformation-

specific antibodies 199.12, ICR10, EGFR1, 225, and 13A9 (25). mAb EMD72000 also bound

the 404SG mutant displayed on the surface of yeast and did not compete for binding with mAb

225 or mAb 806 (Figure 2-10). However, the binding of mAb 225 to EGFR appears to slightly

decrease the affinity of mAb EMD72000 for the receptor. One possibility is that mAb 225

perturbs the conformation of EGFR domain III upon binding, thus decreasing the binding of

mAb EMD72000 to the receptor.

63

Figure 2-10: mAb EMD72000 binding to yeast surface-displayed 404SG EGFR ectodomain mutant. (a) Binding histograms to yeast labeled with either 40 nM mAb 225 or 0.5 μM mAb EMD72000. (b) Binding histograms to yeast labeled with mAb 225 and mAb EMD72000 simultaneously. (c) Binding histograms to yeast labeled with either 75 nM mAb 806 or 0.5 μM mAb EMD72000. (d) Binding histograms to yeast labeled with mAb 806 and mAb EMD72000 simultaneously.

The epitope of mAb EMD72000 was determined as before using random mutagenesis

and yeast surface display (Section 2.3.1). However, a new low mutation rate epitope mapping

library was created using 404SG as the DNA template. The library had a diversity of 6.5x105, and sequencing of 10 random clones from the library yielded 5 wild-type clones, 1 single mutant, and 4 multiple mutants and/or frameshifts. The library was sorted to enrich for mutants with loss of binding to mAb EMD72000 with four rounds of FACS at 0.5 µM EMD72000 (Figure 2-11a).

Then, instead of antibody cross-tests as were performed in Section 2.3.4, high-throughput FACS

selection was then used to remove the misfolded 404SG mutants from the analysis. The library

was sorted for retention of binding to mAb 225 with 2 rounds of positive FACS (Figure 2-11b).

Since mAb 225 binding is dependent on the correct conformation of EGFR domain III, this

64 ensured that only properly folded EGFR mutants were subsequently sequenced and analyzed.

These selections also abrogated the need to apply filtering criteria to remove potentially

misfolded mutations from the epitope analysis, as was necessary for mAbs 225 and 13A9.

Figure 2-11: FACS selections for mAb EMD72000 epitope mapping. 404SG expression as detected through the c- myc tag is shown on the abscissa, and antibody binding is shown on the ordinate. Selection windows are shown in red. (a) 404SG mutant library labeled with 0.5 μM mAb EMD72000, sorted for a loss of binding to the antibody. The yeast were also labeled with chicken anti-c-myc IgY, followed by goat anti-chicken-Alexa488 IgG and goat anti-human-PE IgG. (b) Enriched mutants from (a) labeled with 40 nM mAb 225, sorted for retention of binding to mAb 225. The yeast were also labeled with chicken anti-c-myc IgY, followed by goat anti-chicken-Alexa488 IgG and goat anti-mouse-PE IgG.

After the selections using FACS were completed, the plasmids from the yeast were

rescued, and a total of 100 clones were sequenced. All unique sequences were then re-

transformed into yeast to ensure their loss of binding to mAb EMD72000 and retention of

binding to mAb 225, and these mutants are listed in Table 2-7. For this selection, a total of seven

residues were identified as being energetically important for mAb EMD72000 binding, instead

of only two residues as before with mAbs 225 and 13A9. One reason was that the elimination of

misfolded mutants was high-throughput through the use of FACS, thus enabling the sequencing

and identification of many more properly folded mutations than before. Also, since the probe for

proper EGFR mutant folding, mAb 225, did not compete with mAb EMD72000, more epitope

residues could be identified. The mutants showed different degrees of loss of binding to mAb

65 EMD72000 (Figure 2-12), and these residues thus represent the mAb EMD72000 epitope. The epitope is compared to other antibody epitopes on EGFR domain III in Figure 2-13 and the

EGFR dimer in Figure 2-14.

N 452 I K 454 E*, Q, N F 457 I* S 460 A, F, P, T, Y G 461 D Q 462 K*, L K 463 E, I, N, T Table 2-7: Mutants with loss of binding to mA b EMD 72000 and retention of binding to mAb 225. Mutants marked with an asterisk were only found in the library as clon es with additional irrelevant mutations, which were mutated back to wild-type sequence, and the single mutant was then re-tested to confirm its behavior.

Figure 2-12: Examples of single-clone analysis of 404SG and mutants displayed on the surface of yeast. 404SG expression as detected through the c-myc tag is shown on the abscissa, and antibody binding is shown on the ordinate. Yeast were labeled with 0.5 μM mAb EMD72000 or 40 nM mAb 225 and chicken anti-c-myc IgY, followed by goat anti-chicken-Alexa488 IgG and goat anti-human-PE or goat anti-mouse-PE IgG.

66

Figure 2-13: Space-filled representation of EGFR domain III bound to EGF ligand (1IVO). Green, EGF; red, mAb EMD72000 epitope; Yellow, mAb 13A9 epitope; Blue, mAb 225 epitope residues identified by this work; cyan, additional mAb 225 epitope residues from the mAb 225/EGFR crystal structure (2).

Figure 2-14: Space-filled representation of EGFR dimer bound to EGF (1IVO). EGFR residues 1-272 (the residues absent from EGFRvIII) are shown in beige and brown, while the remaining EGFR residues are shown in grey and olive. EMD72000 epitope, red; mAb 806 epitope, violet; all other coloring as in Figure 2-13.

67 2.3.6 Epitope mapping of DARPins

Designed ankyrin repeat proteins (DARPins) are a novel class of proteins which, like

antibodies, can be engineered for binding specificity to other proteins (47). DARPins that bind to a number of proteins have been isolated, including binders against Her2 (48), maltose binding protein (49), and caspase-2 (50). Originally, the molecules were designed from structure and sequence alignments of natural ankyrin repeats to generate a module with fixed framework residues and randomized residues for interaction with other proteins. For the full molecule, various numbers of this module are assembled (2 to 6) between N- and C-terminal capping repeats, and thus the randomized positions on several adjacent repeats comprise the binding interface. DARPins are extremely stable, with midpoints of thermal denaturation in the range of

65 to 85 oC (47) and thus are easily expressed and purified from E. coli. Five DARPin clones

(E1, E67, E68, E69 and E72) binding to EGFR were isolated by Daniel Steiner of the Andreas

Plueckthun laboratory at the University of Zurich. Initial characterization of the DARPin clones performed by Daniel is detailed in Table 2-8. It was determined that E1, E67, and E68 competed with mAb 225 for binding to EGFR, while E69 did not. In addition, E67 and E68 competed for

binding with E1.

Repeats Apparent mAb 225 E1 kon koff Kd MW competition competition [104 M-1s-1] [10-3 s-1] [nM] E1 3 28.0 yes yes 43.00 0.21 0.48 E67 3 26.8 yes yes 2.95 0.22 7.30 E68 3 58.8/26.8 yes yes 314.00 1.89 0.60 E69 4 33.4 no no 7.25 1.05 14.52 E72 3 26.8 - - - - >100 Table 2-8: Summary of characterization of DARPin clones performed by Daniel Steiner. MW, molecular weight; kon, on-rate; koff, off-rate; Kd, apparent dissociation constant. Because of a low binding signal for clone E72, competition data was inconclusive.

68 First, domain-level mapping of the DARPins was accomplished using EGFR fragments

on the surface of yeast as before (15). E1 was found to bind yeast-surface displayed 404SG but

did not bind EGFR 273-621 or 294-543, most likely for the same reasons that mAb EMD72000

binding to these fragments was not observed (Section 2.3.5). The same binding pattern was seen

for E67 and E68. Based on the previous experimental results, it was inferred that these three

clones bound to EGFR domain III, and thus binding to the other EGFR fragments not containing

this domain was not tested. E69 showed binding to 404SG and EGFR 1-294, very slight binding

to EGFR 1-176, and no binding to EGFR 1-124. Therefore, it appears that E69 is binding either

domain I or domain II, very possibly to residues spanning from just before amino acid 176 to

amino acid 294. However, we cannot exclude the possibility of misfolding of the fragments

which do not show E69 binding; therefore, formally, the epitope is located between amino acids

1-294. E72 only showed binding to 404SG and not to any of the other EGFR fragments tested.

These binding results are summarized in Table 2-9.

EGFR E1 E69 E72 fragment E67 E68 1-124 Ν − − 1-176 Ν ~ − 1-294 Ν + − 273-621 − − − 294-543 − − − 475-621 Ν − − 1-621 (404SG) + + + Table 2-9: Binding of DARPins to yeast surface-displayed EGFR fragments. +, binding; −, no binding; ~, very slight binding; N, not tested.

Binding of the DARPins to yeast surface-displayed 404SG is shown in Figure 2-15.

Although each DARPin was labeled at a concentration to give >90% receptor occupancy

69 according to its affinity, appreciable differences in binding signal were observed for the different

DARPins. Since 404SG has two mutations in domain III, these mutations may be affecting the binding of the DARPin clones to the yeast-displayed protein, with binding of clone E67

especially affected.

Figure 2-15: Labeling of yeast surface-displayed 404SG with DARPin clones. 404SG expression as detected through the c-myc tag is shown on the abscissa, and DARPin binding is shown on the ordinate. The yeast were labeled with each of the DARPins and detected using biotinylated Penta-His antibody followed by streptavidin-PE. Yeast were also labeled with chicken anti-c-myc IgY and goat anti-chicken-Alexa488 IgG to detect display level of 404SG. Cells were labeled with (a) 50 nM E1 (b) 150 nM E67 (c) 50 nM E68 (d) 150 nM E69 (e) 500 nM E72.

Competition binding experiments were conducted for DARPins and conformation-

specific anti-EGFR antibodies binding to yeast-displayed 404SG. The DARPin and antibody

were added simultaneously to the yeast mixture, and the change in DARPin binding was

measured using FACS (Figure 2-16). E1 competed with mAb 225 and mAb EMD72000 for

binding to 404SG, but not with mAb 528. Likewise, E67, E68, and E72 all displayed similar

behavior to E1. E69 competed with ICR10 and 199.12 binding, but not with EGFR1.

70

Figure 2-16: DARPin competition with anti-EGFR mAbs binding to 404SG on the surface of yeast. Cells were labeled with DARPins at the same concentrations as Figure 2-15. mAbs were added at a concentration of 50 nM simultaneously with the DARPins. DARPin binding fluorescence as detected using biotinylated Penta-His antibody followed by streptavidin-PE is shown on the x-axis.

DARPins E1, E68, and E69 were selected for further fine epitope mapping. The 404SG mutant library used for mAb EMD72000 epitope mapping (Section 2.3.5) was also utilized for the DARPin mapping. First, misfolded EGFR mutants were removed from the library through one FACS selection for retention of binding to either 50 nM mAb 528 or mAb EGFR1. Then, the mAb 528-positive library was sorted for loss of binding to E1 (sort 1 at 50 nM and sorts 2-3 at

100 nM) or E68 (sort 1 at 50 nM and sorts 2-3 at 100 nM). The mAb EGFR1-positive library

71 was sorted for loss of binding to E69, with sort 1 at 150 nM and sorts 2-3 at 300 nM. The enriched populations from the sorts were sequenced, and individual mutants were transformed into yeast and tested for proper folding and loss of binding to the DARPins. From 98 clones sequenced for DARPin E69, the epitope was localized to ten residues in EGFR domain I (listed in Table 2-10). The E69 epitope on the EGFR ectodomain is shown in Figure 2-17.

Y 101 H D 102 E A 103 P N 104 I K 105 E, I, N T 106 P P 130 H, L, S H 159 Y L 160 P G 161 D Table 2-10: Mutants with loss of binding to DARPin E69 and retention of binding to mAb EGFR1.

Figure 2-17: Epitope of E69. Epitope residues are shown in red, with the rest of the EGFR ectodomain (from 1IVO, chain a) shown in grey. EGF ligand is shown in stick representation in green.

72 For E1 and E68, 48 clones were sequenced from each enriched population, yielding 14

residues for both epitopes. As seen in Table 2-11, many residues were shared between the two

DARPins. The mutants Q408H, H409Y, Q411K, K465I, and G471D, which led to loss of E1

binding, were tested against E68 and retained binding to this DARPin. The mutant Q384H

showed loss of binding to E68, but retained binding to E1 (data not shown). However, some

mutants isolated for loss of binding to one DARPin (H409Y, F412V, A415E, I438K,

K463E/I/N/T, K465E, I467M/T, N469D) were not assayed for loss of binding to the other. Thus,

these mutants may not be unique to either the E1 or E68 epitopes, but perhaps were not sampled

in the clones sequenced. Amino acid 418 is an epitope residue for both E1 and E68, and it is mutated from serine to glycine in the 404SG EGFR ectodomain mutant, likely causing the lower than expected binding of these DARPins to 404SG (as seen in Figure 2-15).

Residue E1 E68 L 381 V V Q 384 H Q 408 H H 409 D, P, Q, Y D, P, Q Q 411 K, R R F 412 V A 415 T, V E, T, V V 417 A A G 418 D D I 438 K G 441 E, R E, R K 463 E, I, N, T K 465 E, I, N N I 467 M T S 468 G, R G, R N 469 H H, D G 471 D Table 2-11: Mutants with loss of binding to DARPins E1 and E68 and retention of binding to mAb 528. 404SG wild-type residues are shown in the left hand column, and mutations isolated for loss of binding to each DARPin are shown in the right columns.

73 The E1 and E68 epitopes are shown in Figure 2-18. Both DARPins bind to EGFR domain

III and are expected to directly compete with ligand binding to the receptor. In addition, the

DARPin epitopes share many residues with the mAb 225 epitope, consistent with these two

DARPins competing with mAb 225 binding (as seen in Table 2-8 and Figure 2-16).

Figure 2-18: (a) E1 and (b) E68 epitope residues (red) shown on EGFR domain III (chain a of 1IVO) bound to EGF ligand (green), compared to (c) mAb 225 epitope residues as identified by the crystal structure (2).

74 2.4 Discussion

This work describes a novel method of fine epitope mapping using screening of randomly

mutagenized antigen displayed on the yeast cell surface. The method is capable of identifying conformational epitopes of antibodies binding to complex eukaryotic proteins without prior knowledge of potential contact residues, which are several advantages over peptide epitope

mapping and alanine scanning. The yeast surface display system enables convenient protein

expression without the need to solubly express and purify each individual mutant. Clone

characterization and titration are also efficiently performed on the surface of yeast. Initially, we

identified the epitope of mAb 806, whose binding is not dependent on antigen tertiary structure.

Because of this, all single mutations initially isolated from the library for loss of binding to mAb

806 were localized to a single plausible antibody contact surface. In contrast, in order to identify

the epitopes of mAb 225 and mAb 13A9, it was necessary to apply filtering criteria and antibody

cross-tests to remove misfolded mutants from the analysis. Later, while mapping mAb

EMD72000, we improved our epitope mapping method through the use of high-throughput

selection of mutants which retained binding to a different conformation-specific antibody, thus

ensuring that all sequenced clones were properly folded.

This method can be generalized to map any protein-protein interaction. Yeast must be

able to surface display one binding partner, and the other binding partner must be made solubly.

Finally, another conformation-specific binder to the yeast-displayed protein is necessary to

eliminate misfolded mutants. Preferably, this protein should not compete with the interaction

being mapped, although useful information can still be gained if they do compete (as in the case

of mAbs 225 and 13A9). For a multi-domain protein such as EGFR, the probe for proper

conformation should preferably bind the same domain as the interaction being mapped.

75 Additionally, if yeast are unable to properly surface display a protein of interest, a small number

of mutations can often be found through directed evolution that enable the correct folding of the

protein, as in the case of EGFR and the 404SG mutant (25). Our method of random mutagenesis

and yeast surface display has since been used to map the paratope of a single-chain antibody (51)

and to determine the epitope of antibodies against botulinum toxin (52) and the West Nile virus

envelope protein (53).

A potential shortcoming of the fine epitope mapping method is that it is impossible to

isolate all possible mutants which lead to a loss of antibody binding. For example, C302A is not

an amino acid change accessible by a single nucleotide mutation; therefore, with a low mutation

rate library, it is extremely unlikely that this mutation will be seen. In addition, although the

epitope mapping library size was theoretically large enough to accommodate all amino acid

changes accessible by a single nucleotide mutation, the mutant V299D was not isolated from the

library. However, when this site-directed mutant was constructed, it led to a large loss of binding

to mAb 806. Since the generation of the library is random, by chance, the initial library may not

have contained this mutant. Alternately, the mutant may have been lost during growth or sorting of the library prior to single clone analysis, or it may have been necessary to sequence more clones to identify this mutant. To avoid such events, a larger library could be made to increase

the number of mutations, or a larger number of cells could be grown, sorted, and sequenced.

Despite the potential for loss of relevant mutants, the significantly reduced effort required in this approach outweighs the alternatives of site-directed mutagenesis or saturation mutagenesis of the

entire gene. In any case, the fine epitope mapping method can be used to localize antibody

binding to certain residues, followed by the creation of site-directed mutants to confirm and/or

expand the epitope.

76 In this method, substitutions other than those to alanine can be sampled. As demonstrated

here, some energetically important residues are not significantly impacted by mutation to alanine

(e.g. Asp297, Arg300, and Lys301 for mAb 806), and consequently would lead to false negatives

in an alanine scan. In some instances, the mutations in the library led to reversals of amino acid

charge. Such mutations at some distance from the antibody contact interface might conceivably

contribute to a loss of binding through electrostatic steering effects. However, this formal

possibility did not appear to pertain in the cases of the mAb 806, mAb 225, and mAb 13A9

epitopes, given the clustering of mutants in contiguous locations.

Using this epitope mapping method, the binding site for mAb 806 was localized to one

face of the constrained disulfide loop of EGFR 287-302, with Asp297-Cys302, Glu293, and

possibly Cys287 acting as contact residues. The mAb 806 epitope is an example of a cystine

noose, which is a disulfide-constrained, surface-exposed loop important in binding specificity

(54). Cystine nooses have also been identified as major antigenic epitopes on various proteins, including protein G of bovine respiratory syncytial virus (55) and measles virus hemagglutinin protein (56). This suggests that a disulfide-constrained loop is a favorable antigenic structure;

since it is already constrained, there is a smaller entropic cost upon antibody binding. Thus, a number of other disulfide loops on EGFR are potential epitope targets for antibody binding. mAb

806 has been shown to exhibit increased binding to EGFR on cells lacking the domain II dimerization arm (33). Therefore, it has been hypothesized that mAb 806 binds to a transitional form of the receptor as it changes from a closed to extended monomer (33). It is thought that upon mAb 806 binding, the EGFR monomer can no longer dimerize and activate the receptor, accounting for its anti-tumor activity. The mAb 806 epitope presented here and as a result of protein docking simulations (36) is consistent with this hypothesis. The epitope is only partially

77 accessible in the tethered monomer structure, with residues Glu293 and Cys302 obscured (Figure

2-6d). These residues could become exposed upon a conformational transition and allow binding

of mAb 806. The mAb 806 epitope of one EGFR monomer is adjacent to the other monomer in

the EGFR dimer structures, and antibody binding to this epitope could sterically prevent EGFR

dimerization. Likewise, the mAb 175 epitope was found to be highly similar to that of mAb 806,

and the two antibodies are expected to possess the same characteristics.

Two energetically important residues for each of the mAb 225 and mAb 13A9 epitopes

were also identified. Because of a lower fluorescence binding signal to noise ratio for these

conformation-specific antibodies, only mutants with a complete loss of binding could be

differentiated from wild-type. The location of these energetically important contact residues

might provide insight into the mechanisms of these antibodies (Figure 2-9). The mAb 225

epitope is positioned adjacent to the binding sites of both EGF and TGF-α, and thus mAb 225

most likely blocks binding of both ligands sterically. The epitope of mAb 13A9 appears to be

located further away from the ligand binding site; although Ser468 is adjacent to the EGF

binding site, Glu472 is farther towards EGFR domain IV. Therefore, it is possible to envision

both mAb 13A9 and EGF simultaneously binding to EGFR, consistent with the experimental

observation that mAb 13A9 does not compete with EGF binding (38). As viewed in the crystal

structures of EGF- and TGF-α-bound EGFR, it does not appear possible to sterically block

TGF-α binding without also sterically preventing EGF binding. However, mAb 13A9 prevents

TGF-α binding while allowing EGF binding (38). Based on the location of these residues and the

ligand binding site on EGFR, mAb 13A9 may prevent TGF-α binding through mediating

conformational changes to domain III that are sufficient to prevent TGF-α, but not EGF binding.

We have already shown that domain III is highly sensitive to conformational perturbation

78 through mutation. Conceivably, therapeutic antibodies could bind to a number of epitopes on

EGFR domain III to mediate such a conformational change without the need to sterically occlude ligand binding. However, it also possible that these crystal structures do not give a complete picture of ligand binding sterics, and mAb 13A9 may have a different mechanism of action.

Seven residues in EGFR domain III were identified as the epitope of mAb EMD72000.

Based on its location on the crystal structure, it does not appear that mAb EMD72000 directly inhibits EGF binding (Figure 2-13). However, mAb EMD72000 may inhibit EGF binding by preventing EGFR from adopting the fully extended conformation necessary to sandwich the ligand between domains I and III (Figure 2-14). This has also been suggested as an additional mechanism of mAb 225 (2). Alternately, mAb EMD72000 may perturb the conformation of domain III and prevent ligand binding in the same manner as we hypothesize mAb 13A9 does.

Finally, the yeast surface display epitope mapping method was extended to the mapping of

DARPin molecules binding to EGFR. Clone E69 bound to EGFR domain I and is not expected to inhibit ligand binding. However, clones E1 and E68 bound to domain III and are expected to inhibit ligand binding in a manner similar to mAb 225. The E1 and E68 DARPin epitopes highly overlap with the mAb 225 epitope, and all of these molecules also share contact residues with ligands which bind to EGFR. This portion of domain III represents a hot spot for protein recognition, with three very different types of proteins binding to the same location on EGFR.

In conclusion, we have developed a novel method of fine epitope mapping using random mutagenesis and yeast surface display. The technique was used to identify the binding epitopes of a total of four anti-EGFR antibodies and three DARPin molecules. The locations of these epitopes may aid in the understanding of EGFR antibody mechanisms of action and suggests new targets for EGFR inhibition.

79 2.5 Works cited

1. Ferguson, K. M., Berger, M. B., Mendrola, J. M., Cho, H. S., Leahy, D. J., et al. (2003) EGF activates its receptor by removing interactions that autoinhibit ectodomain dimerization, Mol Cell 11, 507-517. 2. Li, S., Schmitz, K. R., Jeffrey, P. D., Wiltzius, J. J., Kussie, P., et al. (2005) Structural basis for inhibition of the epidermal growth factor receptor by cetuximab, Cancer Cell 7, 301-311. 3. Frank, R., and Overwin, H. (1996) SPOT synthesis. Epitope analysis with arrays of synthetic peptides prepared on cellulose membranes, Methods Mol Biol 66, 149-169. 4. Mehra, V., Sweetser, D., and Young, R. A. (1986) Efficient mapping of protein antigenic determinants, Proc Natl Acad Sci U S A 83, 7013-7017. 5. Christmann, A., Wentzel, A., Meyer, C., Meyers, G., and Kolmar, H. (2001) Epitope mapping and affinity purification of monospecific antibodies by Escherichia coli cell surface display of gene-derived random peptide libraries, J Immunol Methods 257, 163- 173. 6. Benichou, S., and Inchauspe, G. (1996) Random fragment libraries using yeast expression plasmid, Methods Mol Biol 66, 241-255. 7. Hager-Braun, C., and Tomer, K. B. (2004) Determination of epitopes by mass spectrometry, Methods Mol Med 94, 109-120. 8. Wu, D. G., Wang, L. H., Sato, G. H., West, K. A., Harris, W. R., et al. (1989) Human epidermal growth factor (EGF) receptor sequence recognized by EGF competitive monoclonal antibodies. Evidence for the localization of the EGF-binding site, J Biol Chem 264, 17469-17475. 9. Yip, Y. L., Smith, G., Koch, J., Dubel, S., and Ward, R. L. (2001) Identification of epitope regions recognized by tumor inhibitory and stimulatory anti-ErbB-2 monoclonal antibodies: implications for vaccine design, J Immunol 166, 5271-5278. 10. Yip, Y. L., Novotny, J., Edwards, M., and Ward, R. L. (2003) Structural analysis of the ErbB-2 receptor using monoclonal antibodies: Implications for receptor signalling, Int J Cancer 104, 303-309. 11. Birkenmeier, G., Osman, A. A., Kopperschlager, G., and Mothes, T. (1997) Epitope mapping by screening of phage display libraries of a monoclonal antibody directed against the receptor binding domain of human alpha2-macroglobulin, FEBS Lett 416, 193-196. 12. Enshell-Seijffers, D., Denisov, D., Groisman, B., Smelyanski, L., Meyuhas, R., et al. (2003) The mapping and reconstitution of a conformational discontinuous B-cell epitope of HIV-1, J Mol Biol 334, 87-101. 13. Baerga-Ortiz, A., Hughes, C. A., Mandell, J. G., and Komives, E. A. (2002) Epitope mapping of a monoclonal antibody against human thrombin by H/D-exchange mass spectrometry reveals selection of a diverse sequence in a highly conserved protein, Protein Sci 11, 1300-1308. 14. Fack, F., Hugle-Dorr, B., Song, D., Queitsch, I., Petersen, G., et al. (1997) Epitope mapping by phage display: random versus gene-fragment libraries, J Immunol Methods 206, 43-52.

80 15. Cochran, J. R., Kim, Y. S., Olsen, M. J., Bhandari, R., and Wittrup, K. D. (2004) Domain-level antibody epitope mapping through yeast surface display of epidermal growth factor receptor fragments, J Immunol Methods 287, 147-158. 16. Cunningham, B. C., Jhurani, P., Ng, P., and Wells, J. A. (1989) Receptor and antibody epitopes in human growth hormone identified by homolog-scanning mutagenesis, Science 243, 1330-1336. 17. Cunningham, B. C., and Wells, J. A. (1989) High-resolution epitope mapping of hGH- receptor interactions by alanine-scanning mutagenesis, Science 244, 1081-1085. 18. Weiss, G. A., Watanabe, C. K., Zhong, A., Goddard, A., and Sidhu, S. S. (2000) Rapid mapping of protein functional epitopes by combinatorial alanine scanning, Proc Natl Acad Sci U S A 97, 8950-8954. 19. Vajdos, F. F., Adams, C. W., Breece, T. N., Presta, L. G., de Vos, A. M., et al. (2002) Comprehensive functional maps of the antigen-binding site of an anti-ErbB2 antibody obtained with shotgun scanning mutagenesis, J Mol Biol 320, 415-428. 20. Pal, G., Fong, S. Y., Kossiakoff, A. A., and Sidhu, S. S. (2005) Alternative views of functional protein binding epitopes obtained by combinatorial shotgun scanning mutagenesis, Protein Sci 14, 2405-2413. 21. Zhen, Y., Caprioli, R. M., and Staros, J. V. (2003) Characterization of glycosylation sites of the epidermal growth factor receptor, Biochemistry 42, 5478-5492. 22. Wan, Y., Zheng, Y. Z., Harris, J. M., Brown, R., and Waters, M. J. (2003) Epitope map for a growth hormone receptor agonist monoclonal antibody, MAb 263, Mol Endocrinol 17, 2240-2250. 23. Jespers, L., Jenne, S., Lasters, I., and Collen, D. (1997) Epitope mapping by negative selection of randomized antigen libraries displayed on filamentous phage, J Mol Biol 269, 704-718. 24. Rice, G. C., Goeddel, D. V., Cachianes, G., Woronicz, J., Chen, E. Y., et al. (1992) Random PCR mutagenesis screening of secreted proteins by direct expression in mammalian cells, Proc Natl Acad Sci U S A 89, 5467-5471. 25. Kim, Y. S., Bhandari, R., Cochran, J. R., Kuriyan, J., and Wittrup, K. D. (2006) Directed evolution of the epidermal growth factor receptor extracellular domain for expression in yeast, Proteins 62, 1026-1035. 26. Boder, E. T., and Wittrup, K. D. (2000) Yeast surface display for directed evolution of protein expression, affinity, and stability, Methods Enzymol 328, 430-444. 27. Chao, G., Lau, W. L., Hackel, B. J., Sazinsky, S. L., Lippow, S. M., et al. (2006) Isolating and engineering human antibodies using yeast surface display, Nat Protoc 1, 755-768. 28. Kawamoto, T., Sato, J. D., Le, A., Polikoff, J., Sato, G. H., et al. (1983) Growth stimulation of A431 cells by epidermal growth factor: identification of high-affinity receptors for epidermal growth factor by an anti-receptor monoclonal antibody, Proc Natl Acad Sci U S A 80, 1337-1341. 29. Waterfield, M. D., Mayes, E. L., Stroobant, P., Bennet, P. L., Young, S., et al. (1982) A monoclonal antibody to the human epidermal growth factor receptor, J Cell Biochem 20, 149-161. 30. Modjtahedi, H., Styles, J., and Dean, C. (1993) The Growth-Response of Human Tumor- Cell Lines Expressing the Egf Receptor to Treatment with Egf and or Mabs That Block Ligand-Binding, International Journal of Oncology 3, 237-243.

81 31. Lakowicz, J. R. (1999) Principles of fluorescence spectroscopy, 2nd ed., pp. 121-128, Kluwer Academic/Plenum, New York. 32. Boder, E. T., and Wittrup, K. D. (1997) Yeast surface display for screening combinatorial polypeptide libraries, Nat Biotechnol 15, 553-557. 33. Johns, T. G., Adams, T. E., Cochran, J. R., Hall, N. E., Hoyne, P. A., et al. (2004) Identification of the epitope for the epidermal growth factor receptor-specific monoclonal antibody 806 reveals that it preferentially recognizes an untethered form of the receptor, J Biol Chem 279, 30375-30384. 34. Johns, T. G., Stockert, E., Ritter, G., Jungbluth, A. A., Huang, H. J., et al. (2002) Novel monoclonal antibody specific for the de2-7 epidermal growth factor receptor (EGFR) that also recognizes the EGFR expressed in cells containing amplification of the EGFR gene, Int J Cancer 98, 398-408. 35. Bogan, A. A., and Thorn, K. S. (1998) Anatomy of hot spots in protein interfaces, J Mol Biol 280, 1-9. 36. Sivasubramanian, A., Chao, G., Pressler, H. M., Wittrup, K. D., and Gray, J. J. (2006) Structural model of the mAb 806-EGFR complex using computational docking followed by computational and experimental mutagenesis, Structure 14, 401-414. 37. Fan, Z., Masui, H., Altas, I., and Mendelsohn, J. (1993) Blockade of epidermal growth factor receptor function by bivalent and monovalent fragments of 225 anti-epidermal growth factor receptor monoclonal antibodies, Cancer Res 53, 4322-4328. 38. Winkler, M. E., O'Connor, L., Winget, M., and Fendly, B. (1989) Epidermal growth factor and transforming growth factor alpha bind differently to the epidermal growth factor receptor, Biochemistry 28, 6373-6378. 39. Sato, J. D., Kawamoto, T., Le, A. D., Mendelsohn, J., Polikoff, J., et al. (1983) Biological effects in vitro of monoclonal antibodies to human epidermal growth factor receptors, Mol Biol Med 1, 511-529. 40. Ogiso, H., Ishitani, R., Nureki, O., Fukai, S., Yamanaka, M., et al. (2002) Crystal structure of the complex of human epidermal growth factor and receptor extracellular domains, Cell 110, 775-787. 41. Garrett, T. P., McKern, N. M., Lou, M., Elleman, T. C., Adams, T. E., et al. (2002) Crystal structure of a truncated epidermal growth factor receptor extracellular domain bound to transforming growth factor alpha, Cell 110, 763-773. 42. Shusta, E. V., Holler, P. D., Kieke, M. C., Kranz, D. M., and Wittrup, K. D. (2000) Directed evolution of a stable scaffold for T-cell receptor engineering, Nat Biotechnol 18, 754-759. 43. Ellgaard, L., and Helenius, A. (2003) Quality control in the endoplasmic reticulum, Nat Rev Mol Cell Biol 4, 181-191. 44. Burgess, A. W., Cho, H. S., Eigenbrot, C., Ferguson, K. M., Garrett, T. P., et al. (2003) An open-and-shut case? Recent insights into the activation of EGF/ErbB receptors, Mol Cell 12, 541-552. 45. Chao, G., Cochran, J. R., and Wittrup, K. D. (2004) Fine epitope mapping of anti- epidermal growth factor receptor antibodies through random mutagenesis and yeast surface display, J Mol Biol 342, 539-550. 46. Murthy, U., Basu, A., Rodeck, U., Herlyn, M., Ross, A. H., et al. (1987) Binding of an antagonistic monoclonal antibody to an intact and fragmented EGF-receptor polypeptide, Arch Biochem Biophys 252, 549-560.

82 47. Binz, H. K., Stumpp, M. T., Forrer, P., Amstutz, P., and Pluckthun, A. (2003) Designing repeat proteins: well-expressed, soluble and stable proteins from combinatorial libraries of consensus ankyrin repeat proteins, J Mol Biol 332, 489-503. 48. Zahnd, C., Wyler, E., Schwenk, J. M., Steiner, D., Lawrence, M. C., et al. (2007) A designed ankyrin repeat protein evolved to picomolar affinity to Her2, J Mol Biol 369, 1015-1028. 49. Binz, H. K., Amstutz, P., Kohl, A., Stumpp, M. T., Briand, C., et al. (2004) High-affinity binders selected from designed ankyrin repeat protein libraries, Nat Biotechnol 22, 575- 582. 50. Schweizer, A., Roschitzki-Voser, H., Amstutz, P., Briand, C., Gulotti-Georgieva, M., et al. (2007) Inhibition of caspase-2 by a designed ankyrin repeat protein: specificity, structure, and inhibition mechanism, Structure 15, 625-636. 51. Colby, D. W., Garg, P., Holden, T., Chao, G., Webster, J. M., et al. (2004) Development of a human light chain variable domain (V(L)) intracellular antibody specific for the amino terminus of huntingtin via yeast surface display, J Mol Biol 342, 901-912. 52. Levy, R., Forsyth, C. M., LaPorte, S. L., Geren, I. N., Smith, L. A., et al. (2007) Fine and domain-level epitope mapping of botulinum neurotoxin type A neutralizing antibodies by yeast surface display, J Mol Biol 365, 196-210. 53. Oliphant, T., Engle, M., Nybakken, G. E., Doane, C., Johnson, S., et al. (2005) Development of a humanized monoclonal antibody with therapeutic potential against West Nile virus, Nat Med 11, 522-530. 54. Lapthorn, A. J., Janes, R. W., Isaacs, N. W., and Wallace, B. A. (1995) Cystine nooses and protein specificity, Nat Struct Biol 2, 266-268. 55. Langedijk, J. P., Meloen, R. H., Taylor, G., Furze, J. M., and van Oirschot, J. T. (1997) Antigenic structure of the central conserved region of protein G of bovine respiratory syncytial virus, J Virol 71, 4055-4061. 56. Putz, M. M., Hoebeke, J., Ammerlaan, W., Schneider, S., and Muller, C. P. (2003) Functional fine-mapping and molecular modeling of a conserved loop epitope of the measles virus hemagglutinin protein, Eur J Biochem 270, 1515-1527.

83 Chapter 3: Engineering antibodies against EGFR

3.1 Background

Directed evolution is a common procedure for engineering antibodies against an antigen of interest. In directed evolution, mutations of the antibody are made at the DNA level, and these mutants are then expressed as proteins. The mutants with improved binding to an antigen of interest are isolated, and these improved clones can be further mutated for increased antigen binding. The entire process is repeated multiple times until the desired antibody properties are attained (Figure 3-1).

Figure 3-1: Process of directed evolution.

In order to engineer antibodies through directed evolution, a display system must be used to link the genotype of an antibody to its phenotype. Tools to accomplish this include phage

84 display, mRNA display, and yeast surface display. Since it is difficult for these systems to

display or express full antibody immunoglobulin (IgG) molecules, smaller fragments of the IgG

are often used instead (Figure 3-2). The IgG molecule contains two identical antigen-binding sites at the end of each variable arm, and a Fab consists of one set of variable domains and CH1 and CL domains. A single chain variable fragment (scFv) molecule contains the minimal

sequences necessary for antigen binding, with the variable heavy and light chains connected by a

flexible polypeptide linker, usually (Gly4Ser)3. The complementarity-determining regions

(CDRs) are the antibody residues responsible for mediating antigen binding and consist of six

loops in the variable domain (Figure 3-2c).

Figure 3-2: Immunoglobulin (IgG) protein and its fragments. (a) Schematic of an IgG. Ag, antigen; scFv, single chain variable fragment; VH, variable heavy region; VL, variable light region; CH, constant heavy region; CL, constant light region. (b) Crystal structure of an immunoglobulin molecule (1HZH). Heavy chain, blue; light chain, yellow. (c) scFv of the antibody in (b), with CDRs shown in yellow and blue for the light and heavy chains, respectively.

85 Phage display is the most widely used method of antibody isolation and affinity

maturation. scFvs or Fabs are displayed on the surface of either fd or M13 bacteriophage and

selected for binding to an antigen of interest (1). Alternately, mRNA or ribosome display is an in

vitro method where the DNA encoding an scFv is stably linked to the protein through a

puromycin linker or ribosome, respectively (2). In yeast surface display (3), the scFv is displayed

on the yeast cell wall as a fusion to the native yeast Aga2 protein (Figure 3-3a). Expression of

the protein on the yeast surface can be detected through the use of the C-terminal c-myc epitope

tag, and binding to a biotinylated antigen of interest is detected through a different colored

secondary reagent. Expression of the Aga2p-scFv is under the control of a galactose-inducible promoter on the yeast display plasmid, which is maintained in yeast episomally with a nutritional

marker. Each yeast cell typically displays 104 to 105 copies of the scFv.

86

Figure 3-3: (a) Schematic of yeast surface display. The scFv (cyan) is displayed as an Aga2p (pink) fusion on the yeast surface. Expression can be detected through fluorescent antibodies binding to the epitope tags (tan), and binding of the scFv to a biotinylated antigen (orange) can be detected using fluorescent avidin (violet). HA, hemagglutinin; VL, variable light chain; VH, variable heavy chain. (b) Yeast cells labeled as in (a) for surface expression (abscissa) and antigen binding (ordinate). The double negative population of yeast cells is due to plasmid loss. As yeast surface display increases, so does the antigen binding signal, giving rise to a diagonal binding curve. The fluorescence data for two different clones are overlaid, with a mutant with two-fold higher affinity in red and wild-type in blue. Because of the expression normalization in yeast surface display, a diagonal sort window can be drawn to capture even slightly improved mutants efficiently. (c) The populations in (b) are difficult to differentiate based on the antigen-binding histogram alone. PE, phycoerythrin. Figures were reproduced from (4).

In addition, Fab fragments have been displayed on the surface of yeast (5,6). In the system described in (6), the heavy chain is expressed as an Aga2p fusion, with the light chain expressed from a separate plasmid (Figure 3-4). The two antibody chains associate in the yeast endoplasmic reticulum (ER) and are displayed as a full Fab construct on the yeast surface.

Separate epitope tags on each chain can be used to ensure the presence of both heavy and light regions.

87

Figure 3-4: Schematic of Fab display on the surface of yeast (6). The heavy chain variable and constant regions (cyan) are expressed as an Aga2p fusion and can be detected using an Alexa488-labeled antibody binding to the c- myc epitope tag (violet). The light chain (yellow) is captured by the heavy chain in the yeast ER and can be detected through a PE-labeled antibody binding the V5 epitope tag. Binding to an antigen such as biotinylated peptide (orange) can be detected using avidin (magenta) labeled with a third fluorophore such as allophycocyanin (APC).

Yeast surface display offers several advantages for the directed evolution of proteins. It

enables quantitative screens through the use of fluorescence-activated cell sorting (FACS),

allowing the equilibrium activity and statistics of the sample to be observed directly during the screening process. Furthermore, the antigen-binding signal is normalized for expression, eliminating artifacts due to host expression bias and allowing for fine discrimination between mutants (Figure 3-3b,c). Antibodies can be engineered for improved stability, as expression is

measured directly and has been shown to correlate with the stability of the displayed protein (7).

Using a two-color labeling scheme as in Figure 3-3a, stability and affinity engineering can be

accomplished simultaneously. Once maturation is complete, antibody affinity can be

conveniently titrated while displayed on the surface of the yeast, obviating the need for

expression and purification of each clone. In almost every case for dozens of different antibodies,

the binding properties on the yeast cell surface are in quantitative agreement with those measured

88 in solution or by biosensor methods (8). Finally, the displayed proteins are folded in the ER of

the eukaryotic yeast cells, taking advantage of ER chaperones and quality-control machinery.

A theoretical limitation of yeast surface display is the potentially smaller functional

library size (~107-109) than that of other selection methods (phage display, ~106-1011;

mRNA/ribosome display, ~1011-1013). However, it is difficult to determine the true functional

diversity of any display library, and bias-free propagation of yeast libraries has been confirmed

over an amplification of 1010-fold (9). Furthermore, all of these methods greatly undersample the theoretical sequence space of scFv CDR variation (~1080). To realize the advantages of kinetic

screening and expression normalization, yeast surface display also requires more complex

equipment than other display methods. Various technologies for screening recombinant antibody

libraries and their relative strengths and weaknesses have been reviewed in (10). However, in

parallel selections from the same DNA library, it has been shown that yeast display can yield

three times as many unique scFv clones as phage display does (11).

Yeast surface display has been used to engineer antibodies against various antigen

targets, including T cell receptors (12), huntingtin protein (13), carcinoembryonic antigen (14), and botulinum neurotoxin (15). In addition, an antibody to fluorescein has been engineered to

femtomolar affinity, the highest affinity yet reported (16). A nonimmune human library of 109 scFv clones displayed on the surface of yeast has been created (9) and can serve as an initial library for the isolation of antibodies against an antigen of interest.

Although there are currently two anti-EGFR antibody drugs (cetuximab and panitumumab) approved by the FDA for the treatment of cancer and many more in clinical trials

(see Section 1.4), these antibodies are only marginally effective. Cetuximab treatment leads to objective responses in only 10.8% of patients as monotherapy and 22.9% in combination with

89 irinotecan (17). Panitumumab produces an 8% overall response rate in metastatic colorectal

cancer, with progression-free survival extended from 60 days to 96 days (18). In a recent phase II

trial, another anti-EGFR antibody, matuzumab, showed no evidence of significant clinical

activity in patients with ovarian cancer (19). The low response rates are particularly noteworthy

since the clinical trials only involved patients who tested positive for EGFR expression in tumor

biopsies. Clearly, there is room for improved novel antibodies against EGFR.

In this antibody engineering work, we have opted to use an alternate inhibition strategy

from existing EGFR therapeutics. Traditional anti-EGFR antibodies such as cetuximab,

panitumumab, and matuzumab bind to EGFR domain III and inhibit ligand binding, thus

preventing receptor activation and downstream signaling (Figure 3-5a). However, we sought to

engineer antibodies against EGFR domains II and IV to directly inhibit receptor dimerization

(Figure 3-5b). Such antibodies could have advantages over ligand-competitive antibodies in

certain types of tumors. In particular, cancer cells driven by EGFR autocrine signaling (20) have

high local effective concentrations of ligand, and it may be difficult for an antibody diffusing into a tumor from the plasma to compete with ligand binding. In addition, ligand-competitive antibodies such as cetuximab are ineffective against EGFRvIII (21), a truncation mutant which signals constitutively in the absence of ligand (22). Finally, novel engineered antibodies could be used synergistically with existing EGFR antibody therapeutics.

Figure 3-5: EGFR inhibition strategies for antibodies. (a) Ligand-competitive antibodies such as cetuximab. (b) Signaling inhibition by blocking receptor dimerization. Blue, receptor; green, ligand; yellow, antibody.

90 Based on the EGFR dimer crystal structures, antibodies binding to either the domain II dimerization arm or portions of domain IV should inhibit receptor dimerization (see Figure 1-4).

Although domain IV is not resolved in the EGFR dimer crystal structures, it may contain receptor dimerization contacts for reasons discussed in Section 1.2. Even if domain IV does not mediate receptor dimerization, it is predicted that the two domains are sufficiently close in the dimerized form such that antibody binding to domain IV should sterically inhibit dimerization.

In order to target specific regions of EGFR domains II and IV with antibodies, a former postdoctoral associate in the laboratory, Mark Olsen, designed peptide mimics of loops in the receptor to use as antigen. Five peptides, one in domain II comprising the dimerization arm and four in domain IV, were selected and designed (Figure 3-6). In this thesis, the peptides will be referred to according to the first three amino acids in the corresponding EGFR sequence (i.e.

LYN, CRN, CAH, MGE, and LVW). All five peptides were synthesized with an N-terminal biotin molecule for detection, and a glycine amino acid to separate the peptide epitope from the biotin molecule. Because of the small size of the MGE peptide, a larger spacer molecule,

6-aminohexanoic acid, was used instead of glycine. The peptides were cyclized through a disulfide bond to mimic the structures of the loops in EGFR.

91

Figure 3-6: EGFR peptide epitopes targeted for antibody engineering. Peptides are shown in space-filled representation on the tethered EGFR monomer structure (1NQL). Peptides synthesized for use as antigen are shown on the right, with the corresponding EGFR residues shown below the peptide sequence. Aha, 6-aminohexanoic acid.

The peptide epitopes were chosen for their potential to inhibit EGFR dimerization. The

LYN peptide consists of the tip of the domain II dimerization arm (23,24), and proteins binding to this epitope would be analogous to pertuzumab, an antibody which binds to domain II of Her2

(25). The CRN peptide was chosen as a control peptide in domain IV. The other three domain IV peptides (CAH, MGE, and LVW) are analogous to the binding epitope of Herceptin on Her2

(26). The CAH peptide corresponds to part of the Herceptin epitope on Her2 residues 570-573, the LVW peptide corresponds to another part of the epitope on Her2 residues 593-603, and MGE corresponds to residues immediately N-terminal of the LVW peptide. In addition, the LVW peptide has previously been synthesized, and its binding to EGFR has been shown to inhibit receptor heterodimerization with Her3 (27). However, antibodies against this peptide have not been published. A monoclonal antibody (2D2) against the peptide 556IQCAHYIDGPHC567,

92 which includes the CAH peptide, has been generated for diagnostic purposes (28), but the effect

on EGFR dimerization of binding to this epitope has not been tested.

To isolate antibodies binding to these peptides, Mark Olsen performed the first round of

screening against the nonimmune human scFv library expressed on the surface of yeast (9). For each of the peptides, he first performed one magnetic bead isolation (MACS) at 1 μM peptide, followed by three selections using fluorescence-activated cell sorting (FACS). These screens yielded candidate clones binding to all of the peptides except the CAH peptide.

In this thesis work, antibody engineering against the above EGFR peptides was

continued. It was discovered that all the clones isolated previous to this thesis work were heavy

chain only binders. Full scFv molecules were engineered by adding random light chains to the

selected heavy chains. In addition, rounds of affinity maturation through random mutagenesis

using both peptide and soluble EGFR ectodomain as antigen were performed. Fab display on the surface of yeast and rabbit monoclonal antibody technology were utilized to further engineer

binding to peptides LYN and MGE. However, affinity maturation of Fabs to the MGE peptide

led again to binders consisting of only the heavy chain, and the rabbit monoclonal antibodies

binding to the MGE peptide did not bind the full EGFR ectodomain.

93 3.2 Materials and methods

Peptide and 404SG antigen

The peptides listed in Figure 3-6 were synthesized by Cell Essentials, Inc. (Boston, MA).

The molecular weights of the peptides were confirmed by mass spectrometry to 95% purity:

LYN, 1973; CRN, 1516; CAH, 1452; MGE, 1208; and LVW, 1565.

For the secretion of soluble 404SG, the S. cerevisiae strain YVH10 was used, which is

modified from BJα5464 to overexpress protein disulfide isomerase (29). The DNA for 404SG, a

mutant of the EGFR ectodomain (30), was cloned using NheI and BamHI into the secretion

plasmid pBala, which contains an N-terminal flag epitope tag (DYKDDDDK) and C-terminal

His tag (HHHHHH). YVH10 was transformed with this DNA in addition to an empty pRS314

vector (which enables tryptophan synthesis) using the EZ yeast transformation kit

(ZymoResearch). Transformed cells were plated on SDCAA plates (yeast media preparation is described in (4)), and individual colonies were grown in 5 ml SDCAA media at 30 oC overnight.

The seed culture was transferred to 1 L SDCAA media in a 2 L Tunair flask (Shelton Scientific)

and grown at 30 oC for 20-22 h. The cells were harvested and pelleted in 500 ml centrifuge tubes

for 15 min at 5000 rpm and 4 oC. The yeast were resuspended in 1 L SGCAA containing 0.5 g

lysozyme as a carrier protein and were induced in a 2 L glass shake flask with no baffles at 20 oC

for 3 d. The supernatant was separated from the yeast as above, and the pH was adjusted to 7.4.

404SG protein was initially purified according to the manufacturer’s instructions using a batch

protocol with TALON Superflow resin (BD Biosciences), which binds to the His6 tag.

Alternately, the yeast supernatant was concentrated to ~50 mL volume using an Amicon stirred cell concentrator (Millipore). The protein was then digested with EndoH (New England

Biosciences) to remove excess glycosylation and was purified using mAb 528 agarose resin

94 (Santa Cruz Biotechnology). The protein was eluted with 0.1 M glycine HCl, pH 3.5 and was

neutralized with 1 M Tris, pH 8. The protein concentration was determined using an extinction coefficient of 55,900 M-1cm-1 at 280 nm absorbance.

Yeast labeling for FACS and clone characterization

Yeast were labeled as in (4). Additional yeast surface display protocols are available in

(31) and (32). Briefly, EBY100 yeast transformed with the pCT2 yeast display plasmid were

grown in SDCAA media at 30 oC and induced in SGCAA at 20 oC. For FACS analysis and

sorting, 1x106 yeast (for individual clones) or 10X the library diversity were labeled for display using either mouse anti-HA 12CA5 IgG (Abcam; 1:100) or chicken anti-c-myc IgY (Invitrogen;

1:250), followed by goat anti-mouse Alexa488 or goat anti-chicken Alexa488 (Invitrogen;

1:100). Yeast were labeled with biotinylated peptides, followed by either streptavidin-PE (1:50), streptavidin-APC (1:100), or neutravidin-PE (1:50; all from Invitrogen). For 404SG labeling,

404SG was pre-incubated with up to 0.5 μM biotinylated M2 anti-flag antibody (Sigma) for at least 1 h on ice, followed by streptavidin-PE. All incubations were performed in PBS buffer with

0.1% BSA, with primary incubations for 30 min at room temperature and secondary incubations for 20 min on ice.

For characterization of individual clones, DNA was harvested from yeast cells using a

Zymoprep kit (ZymoResearch), transformed into XL-I Blue E. coli (Stratagene), and sequenced

by the MIT Biopolymers Laboratory. BLAST searching against germline genes was performed

using IgBlast at http://www.ncbi.nlm.nih.gov/projects/igblast. Fluorescence data was collected

using a Coulter Epics XL flow cytometer (Beckman-Coulter) and was analyzed using FloJo

software (Tree Star, Inc.).

95 Construction of scFv libraries

For the first round of affinity maturation pairing selected heavy chains with light chains,

plasmid DNA from the original scFv library construction (9) was amplified by transformation

and growth in E. coli strain XL-I Blue (Stratagene). Plating of transformed cells showed that the

maximum diversity of this amplified DNA was 8.5x105, which is slightly lower than the original

6 library’s reported VL diversity (1.2x10 ). However, in order to conserve the original library

DNA, it was determined that this was sufficient to proceed. The library DNA was digested with

NheI and BamHI to remove the original VH sequence. Then, the DNA containing the originally

selected VH clones (LYN-1, etc.) were digested with KpnI and NotI to generate the insert fragment. These two DNA fragments were transformed into EBY100 yeast as described to generate libraries of full scFvs (4). It was determined that this method yielded more full-length scFvs and thus was preferable to using the digested VH clone as backbone and a PCR product

from the library insert. Libraries were labeled as above and sorted using either a MoFlo or Aria

FACS machine at the MIT flow cytometry core facility. For the generation of subsequent

random mutagenesis libraries, the nucleotide analog mutagenesis protocol in (32) was used.

Yeast were transformed and libraries were sorted as above.

Fab display

Fab display vectors pPNL200 and pPNL300 (adapted from (6)) were generously

provided by Mike Feldhaus. For individual clone analysis, both plasmids were transformed into

the yeast strain JAR200 (Ura-, Trp-) and grown and induced as strain EBY100 above. To clone

scFv genes into the Fab display vectors, 5’ NheI and 3’ MluI restriction sites were added to the

variable heavy region using PCR, and for the light chain, 5’ DraIII and 3’ BsiWI sites were

96 added. Populations of yeast were cloned into the Fab display vectors by designing primers based

on sequenced clones and PCR using the zymoprep DNA as template. The primers used for

nucleotide analog mutagenesis of antibodies in the Fab vectors are listed in Table 3-1. For library

backbone generation, pPNL200 was digested with NheI and MluI and pPNL300 with DraIII and

BsiWI. Heavy and light chain libraries were constructed separately, with the heavy library transformed into JAR200 and grown in SDCAA+20 mg/L uracil (added from 0.2% stock) and the light library transformed into YVH10 and grown in SDCAA+20 mg/L tryptophan (added from 1% stock). The Fab library was created using yeast mating by resuspending 109 freshly

grown cells from each of the libraries in 10 mL YPD. The yeast were pipetted onto 4 YPD plates

(15 cm diameter) in a pool (not spread) and incubated at 30 oC overnight. The mated cells were

scraped off and resuspended in SDCAA, and dilutions were plated to determine mating

efficiency. This protocol was adapted from (6).

Table 3-1: Sequences of primers used in Fab display VH: 200for 5’-CTAAAGAAGAAGGGGTATCTCTCGAAAAAAGAGAGGCTGAAGCAACGCGT-3’ VH: 200rev 5’-CTCTTGGAGGAGGGTGCCAGGGGGAAGACCGATGGGCCCTTGGTGCTAGC-3’ VL: 300for 5’-TAAAGAAGAAGGGGTATCTCTCGAAAAAAGAGAGGCTGAAGCACGATGTG-3’ VL: 300rev 5’-TGCTCATCAGATGGCGGGAAGATGAAGACAGATGGTGCAGCCACCGTACG-3’ VH, VL sequencing: 5’-CTATTGCCAGCATTGCTGC-3’

Growth and labeling of mammalian cells

A431NS cells (A431 cells passaged to become less adherent) were obtained from

American Type Culture Collection (ATCC) and grown in DMEM (ATCC) supplemented with

10% fetal bovine serum (FBS). NR6 cells (33) were obtained from Doug Lauffenburger (MIT)

and grown in MEM-α supplemented with L-glutamine, sodium pyruvate, non-essential amino

acids, and 350 mg/L G418. Mammalian cell monolayer panning was conducted as described in

97 Appendix 2. For FACS labeling, A431 cells were lifted from the plate using versene (Invitrogen)

at 37 oC for 30 min. 2.5x105 cells were labeled in 200 μL/tube in PBS + 1% BSA on ice. Goat

anti-rabbit-PE (Invitrogen) was used to visualize rabbit antibody binding. 7-amino-

actinomycin D (7-AAD) for labeling dead cells was purchased from Calbiochem.

Production and growth of rabbit hybridomas

Rabbit hybridoma production was performed by Epitomics, Inc. (Burlingame, CA). LYN

and MGE peptides (lacking biotin) were conjugated to KLH and injected into rabbits. Rabbit

hybridoma clones were grown in RPMI 1640 supplemented with 0.05 mM 2-mercaptoethanol,

Epitomics RabMab supplement A, 10% FBS, and 1X HAT (Hypoxanthine, Aminopterin, and

Thymidine). Hybridoma cells were grown in 10 cm culture dishes and expanded to T-175 flasks

(60 mL media). For rabbit monoclonal antibody production, cells were adapted to serum-free

conditions by diluting growth media 1:4 in serum-free media (BD cell mAB medium (BD

Biosciences) supplemented with 1% glutamax (Gibco)). After 1-3 days of adapting, full serum- free media was used and changed every 3 days until antibody harvesting was complete (usually 3 media changes).

Purification and sequencing of rabbit monoclonal antibodies

Rabbit hybridoma supernatants were concentrated in an Amicon stirred cell concentrator

(Millipore) and diluted 1:1 in Ig binding buffer (Pierce). Antibodies were purified using Protein

A resin (Pierce) and eluted using Ig elution buffer (Pierce). Elutions were buffer-exchanged with

PBS using an Amicon ultra-15 centrifugal filter unit (Millipore). Typical yields were ~2 mg from

1 L of culture media. For rabbit monoclonal antibody sequencing, total hybridoma mRNA was

98 harvested using an RNeasy kit (Qiagen). The mRNA was reverse-transcribed using an

Omniscript RT kit (Qiagen) and primers IgGHrev (primes to heavy constant region) and

IgGKrev (primes to kappa constant region). The reverse transcription product was amplified using PCR with primers IgGHfor (primes to heavy leader sequence)/IgGHrev and

IgGKfor/IgGKrev for kappa light chains. Primers are listed in Table 3-2, with additional unused primers for rabbit IgGL (lambda light chain) sequences. For PCR of clone 31, the reverse primer

IgVHJrev was used to prime to the variable J region since this clone was not an IgG. PCR products were sequenced using the reverse primers, and cloned into an scFv format using overlap extension PCR.

Table 3-2: Primers used for sequencing rabbit monoclonal antibodies IgGHfor 5’-ATGGAGACTGGGCTGCGCTGGCTT-3’ IgGHrev 5’-GGTCACCGTGGAGCTGGGTG-3’ IgGKfor 5’-GGGCCCCCACTCAGCTGCTGGG-3’ IgGKrev 5’-CATCCACCTBCCAGGTGACGGTG-3’ IgGLfor1 5’-GGAGACTGGGCTGCGCTGGC-3’ IgGLfor2 5’-GGCCTGCACCCCGCTCCTCC-3’ IgGLrev 5’-GGGTAGAAGTCATTGATCAGAC-3’ IgVHJrev 5’-GGTGACCAGGGTGCCYKGGCCCCA-3’

B = C/G/T, Y = C/T, K = G/T

99 3.3 Results

3.3.1 Characterization of antibody clones against EGFR peptides

The antibody clones enriched for binding to peptides CRN, CAH, MGE, and LVW after

one round of selection by Mark Olsen were sequenced. Unexpectedly, all of the enriched clones

were not full-length scFvs, but only consisted of the heavy chain (VH) and not the light chain

(VL). Such clones may have arisen from errors in the PCR construction of the library or in yeast

homologous recombination, leading to a frameshift and eventual premature termination of

translation. The closest germline genes to the selected clones as determined by a BLAST search

engine IgBlast are listed in Table 3-3, and the amino acid sequences of the selected clones are

shown in Figure 3-7. Overall, each of the two clones isolated for binding to a particular peptide

were highly similar in sequence, with the exception of CRN-1 and CRN-2. The VH clones contain most of the (Gly4Ser)3 linker region and additional amino acids after the linker,

terminating in a stop codon. These additional amino acids are not light chain residues, as they are

frame-shifted from the original sequence. Each clone listed in Figure 3-7 was tested for binding

to its cognate peptide (Figure 3-8). Given the intermediate labeling of the clones at 1 μM peptide, it appeared that the affinity of these VH domains for the peptides was on the order of 10

μM or greater. It was confirmed that the VH clones did not bind any secondary reagents. In

addition, each VH was tested for binding to all of the other peptides. No appreciable cross-

binding was observed at 1 μM peptide except for clones LYN-1 and LYN-2 binding to CAH peptide (Figure 3-8(c-d)), although these clones do show a higher affinity for the LYN peptide.

100 VH clone Germline LYN-1, LYN-2 V3-30 CRN-1 V6-1 CRN-2 V4-34 MGE-1, MGE-2 V4-39 LVW-1, LVW-2 V2-5

Table 3-3: Heavy chain germline genes of VH clones selected for binding to peptides.

FR1 CDR1 FR2 CDR2 LYN-1 QVQLVQSEGGVVQPGRSLRLSCAAP GFTFSSYAMH WVRQAPGKGLEWVA VISYDGSNKY LYN-2 ------S ------

CRN-1 QVQLQQSGPGLVKPSQTLSLTCDIS GDSVSSDSAAWN WIRLSPSRGLEWLG RTYYWSKWYTD CRN-2 ------W----L---E------GV- -G-L-GYYWS ---Q--GKE---I- EINQGGSTN

MGE-1 QVQLQESGPGLVKPSETLSLTCTFS GGSIRSSSDYWG WVRQPPGKGLEWIG SISSSGSTY MGE-2 ------V- ----SN-DY--- -I------R--- --NYY-T--

LVW-1 QITLKESGPTVVKPTQTLTLTCTFS GFSLSTSGVGVG WIRQPPGKALEWLA LISWDDDKR LVW-2 ------

FR3 CDR3 LYN-1 YADSVKGRFTISRDNSKNTLYLQMNSLRAEDTAVYYCAR DEGYYGSGCIDY LYN-2 ------

CRN-1 YAVSVKSRITINPDTSKNQFSLQLNSVTPEDTAVYFCAR ERYCSSTSCYLDAFDI CRN-2 –NP-LR--V--SV----K-L--KV----AS-----Y--- -TFRG-NCFDS

MGE-1 YNPSLKSRVTISVDTSRNQFSLRLSSVTAADTAIYYCAR YNHYWYFDL MGE-2 ------AM-----K-----K------V----- RGANSWFFDL

LVW-1 YSPSLKSRLTITKDTSKNQVVLTMTNMDPVDTATYFCAH LNNFVDAIELRTGWCFDV LVW-2 ------I------

Linker LYN-1 WGQGTLVTVSSGILGSGGGGSGGGGSGGGGLQEF* LYN-2 ------*

CRN-1 WGQGTMVTVSSGILGSGGGGSGGGGSGGGGLQEF* CRN-2 -----L------I------*

MGE-1 WGRGTLVTVSSGILGSGGGGSGGGGSGGGGSKRHSRSLQHSSQRLQETKSTSPAKPAKTLMMI* MGE-2 ------LQEF*

LVW-1 WGRGTLVTVSSGILGSGGGGSGGGGSGGGVLKLC* LVW-2 ------R*

Figure 3-7: Amino acid sequences of VH clones selected for binding to peptides. FR, framework region; CDR, complementarity determining region; -, conserved residue; *, stop codon.

101

Figure 3-8: Characterization of individual VH clones isolated for binding to peptides. Yeast cells displaying the VH clones were labeled with anti-HA antibody 12CA5 (abscissa) and 1 μM peptide (ordinate), followed by goat anti- mouse-Alexa488 IgG and streptavidin-PE. (a-b) Clones LYN-1 and LYN-2 binding to LYN peptide. (c-d) LYN-1 and LYN-2 binding to CAH peptide. (e-f) CRN-1 and CRN-2 binding to CRN peptide. (g-h) MGE-1 and MGE-2 binding to MGE peptide. (i-j) LVW-1 and LVW-2 binding to LVW peptide.

102 3.3.2 Initial round of selection for scFvs against peptides

Single-domain antibodies have proven useful for some applications, such as intracellular

expression (13) or increased stability (34). However, for our purposes, single-domain antibodies would provide little advantage over full scFvs; therefore, we sought to pair our selected VH clones with VL domains. This was accomplished by creating libraries joining the VH DNA with

light chain DNA from the original library (9). The resulting scFv DNA was transformed into

yeast using electroporation, yielding libraries designated LYN-A (~107 diversity), CRN-A

(9x105), MGE-A (~107), and LVW-A (~107). In addition, the original nonimmune human scFv

library on the surface of yeast (9) was screened for binders against the CAH peptide using one round of MACS, since no clones were previously isolated against this peptide.

Selections of the above five populations of yeast were performed using FACS to screen

for binding to the peptides. For the LYN peptide, 4 FACS selections were performed at 1 μM

peptide, and clones from the enriched population were sequenced. However, only one clone

(LYN-A4-8, sequence in Appendix 1.4) was a full scFv, while 9/10 clones only had heavy chains

with no C-terminal c-myc tag. Since the anti-c-myc antibody used was a chicken IgY polyclonal,

it appeared that some molecules in this population were able to weakly bind a motif in the VH clones. The enriched cells after 3 FACS selections (LYN-A3) were re-sorted with an improved sorting window drawn to avoid these VH clones. Nine clones from the enriched population

(LYN-A5) were sequenced, revealing 7 unique clones (Appendix 1.4). However, tests of these

clones for binding to 1 μM LYN peptide did not show appreciable binding. For the CRN peptide,

3 FACS selections were performed at 1 μM peptide, and the enriched cell population (CRN-A3)

was sequenced. Out of 10 clones sequenced, all were unique (listed in Appendix 1.5) and showed

103 binding to 1 μM CRN peptide (Figure 3-9). It was further confirmed that these clones indeed

bound the peptide and did not bind the secondary streptavidin reagent.

Figure 3-9: Binding of CRN-A3 clones selected for the CRN peptide. Yeast were labeled with chicken anti-c-myc IgY (abscissa) and 1 μM CRN peptide (ordinate), followed by goat anti-chicken-Alexa488 IgG and streptavidin-PE. Clones CRN-A3-1 through CRN-A3-10 are shown in (a)-(j), respectively.

For the CAH peptide, the MACS-enriched population was subjected to 3 FACS

selections at 1 μM CAH peptide. At this point, it appeared that the sorted population contained

some binders to streptavidin instead of the CAH peptide. However, two more FACS selections

using alternating secondary reagents allowed the elimination of the streptavidin binders, resulting

in the CAH-A6 population. Out of 10 clones sequenced, 8 were unique and tested for binding,

and 4 of these clones bound to CAH peptide (Figure 3-10). It was confirmed that these clones did

not bind the streptavidin secondary reagent, and sequences for these clones are listed in

104 Appendix 1.6. For the MGE peptide, 4 FACS selections were performed at 1 μM peptide, and

the enriched population (MGE-A4) was sequenced. Out of 10 clones sequenced, 1 clone was a

VH only, with 8 unique scFv sequences. However, only 4 of these clones (sequences in Appendix

1.7) showed appreciable binding to 1 μM MGE peptide (Figure 3-11), and it was confirmed that

these clones did not bind streptavidin-PE. Finally, for the LVW peptide, 3 FACS selections were

performed at 1 μM peptide. 10 clones from the enriched population (LVW-A3) were sequenced,

revealing 1 VH only clone and 8 unique scFv sequences (Appendix 1.8). Of the 7 scFv clones

tested, all appeared to bind peptide to different degrees (Figure 3-12) and did not bind

streptavidin.

Figure 3-10: Binding of selected CAH-A6 clones. Yeast cells were labeled with 1 μM CAH peptide, followed by streptavidin-PE. Binding histograms for clones CAH-A6-2 (labeling similar to CAH-A6-1) and CAH-A6-5 (labeling similar to CAH-A6-4) were omitted for clarity.

Figure 3-11: Binding of MGE-A4 clones selected for the MGE peptide. Yeast were labeled with chicken anti-c-myc IgY (abscissa) and 1 μM MGE peptide (ordinate), followed by goat anti-chicken-Alexa488 IgG and streptavidin-PE. (a) MGE-A4-2 (b) MGE-A4-3 (c) MGE-A4-4 (d) MGE-A4-7.

105

Figure 3-12: Binding of LVW-A3 clones selected for LVW peptide. Yeast were labeled with chicken anti-c-myc IgY (abscissa) and 1 μM LVW peptide (ordinate), followed by goat anti-chicken-Alexa488 IgG and streptavidin-PE. (a) LVW-A3-2 (b) LVW-A3-3 (c) LVW-A3-4 (d) LVW-A3-5 (e) LVW-A3-7 (f) LVW-A3-8 (g) LVW-A3-9.

As shown above, full scFv clones binding to the various peptides were successfully

isolated in the first round of selections. However, based on binding data at 1 μM peptide, it appeared that these clones were of similar or worse affinity than the VH only clones. Since the

antigen consisted of only a small peptide, there may not have been enough surface area for the

light chain to also make contacts to the antigen. The closest germline genes for the light chains of

the selected clones listed in Appendix 1.4-1.8 are shown in Table 3-4. Selections for binding to

the five different peptides yielded only 7 unique germline genes out of 34 unique sequences.

These genes are from the most abundantly represented germline families in the original

unselected nonimmune library (9). The observation that many VL germline genes are shared between the scFv clones binding to different peptides lends further support to the hypothesis that the light chain is not significantly contributing to antigen peptide binding. In this round, it appears highly possible that we have selected for light chains that are merely compatible with the original VH only clones.

106 κ (Kappa) germline λ (Lambda) germline V1-39 V3-20 V5-2 V1-44 V2-14 V2-23 V6-57 LYN X X X CRN X X X X CAH X X MGE X X X X LVW X X X Table 3-4: Light chain germline genes represented by selected clones against peptides are indicated with an X.

3.3.3 Second round of affinity maturation

To improve the affinities of the clones from the first round of scFv selection, new antibody libraries were constructed. Random mutagenesis of the entire scFv gene was accomplished through PCR with nucleotide analogs, using DNA from the entire enriched population of yeast as a template. Therefore, this DNA population most likely contained additional clones which had been enriched, but had not been sequenced or characterized. The

DNA was transformed into yeast to create libraries with the following diversities: LYN-B,

6.4x105; CRN-B, 4.4x106; CAH-B, 1.7x107; MGE-B, 1.6x106; and LVW-B, 1.3x106.

Sequencing of random clones from the libraries showed a wide range of mutations (discernible by comparison of the heavy chains to the original VH clones), from 0-20 amino acid changes. At

this point, the antigen was changed from the peptides to the full EGFR ectodomain for library

screening. Since the light chain did not appear to be contacting the peptide antigen based on the previous round’s analysis, it was hypothesized that moving to a large protein antigen would provide more potential surface area for the light chain to bind, thus improving the affinity of the scFv. The specific antigen used was 404SG, an EGFR ectodomain mutant which has been engineered to be expressed solubly by yeast. 404SG contains four point mutations, none of which are in domains II or IV (30). Excess glycosylation was removed from the 404SG antigen using EndoH digestion prior to use.

107 The LYN-B library was enriched for binding to 404SG using 4 FACS selections. The

CRN-B library was similarly selected using 4 FACS sorts. Clones from the enriched population

(CRN-B4) were sequenced (listed in Appendix 1.9-1.11). The selection appeared to enrich for

some binding artifacts, including light chain only clones and peptides. The CAH-B library was

enriched for binding to 404SG through 4 FACS selections, yielding only one scFv clone due to library contamination with yeast-displayed EGF (CAH-B4-1 sequence in Appendix 1.12). The

MGE-B library was sorted using 6 FACS screens against soluble 404SG antigen, and the enriched MGE-B6 clones were sequenced (Appendix 1.13). Finally, the LVW-B library was selected using 4 FACS sorts, and the enriched yeast population (LVW-B4) was sequenced. In addition to scFvs, this library also yielded 2 clones consisting of a light chain only (Appendix

1.14).

Clones CAH-B4-1 and MGE-B6-2 from the second round of selections were tested for

their 404SG antigen-binding properties and compared to negative (D1.3, a lysozyme-binding

scFv) and positive (EGF) controls displayed on the surface of yeast (Figure 3-13). Initially,

binding could only be seen at very high 404SG concentrations, with cells labeled at 20 and 140

μM. The selected scFv clones also showed an unusual pattern of binding; instead of the normally

observed diagonal, the binding curves were more vertical. Such binding is normally indicative of

avidity effects, which may occur at high levels of yeast scFv expression and multimeric antigen.

This is a possibility since a bivalent molecule, biotinylated anti-flag IgG, was pre-incubated with

the 404SG to allow binding to the N-terminal flag tag. However, it is still unusual that there is a

population of cells which show no antigen labeling even though they express a high number of

scFv molecules.

108

Figure 3-13: Clone binding to soluble 404SG. Yeast were labeled with chicken anti-c-myc IgY (abscissa) and 404SG antigen which had been pre-incubated with biotinylated M2 anti-flag IgG (ordinate), followed by Alexa488 goat anti-chicken IgG and streptavidin-PE. (a-d) D1.3 negative control, EGF positive control, CAH-B4-1, and MGE-B6-2, respectively, displayed on the surface of yeast and labeled with 20 μM 404SG. (e-h) Yeast displaying proteins as in (a-d), labeled with 140 μM 404SG. (i) MGE-B6-2 labeled with 200 nM 404SG (mAb 528 affinity- purified). (j) MGE-B6-2 labeled as in (i) plus 3.6 μM soluble EGF.

For the labeling data in Figure 3-13 a-h, the 404SG antigen was affinity purified using the

C-terminal His6 tag and digested with EndoH to remove excess glycosylation. However, the high

concentration necessary to label even the positive control suggested that a portion of the 404SG

may have been misfolded. Therefore, the EndoH-digested protein was affinity purified using a

conformation-specific EGFR antibody, mAb 528. 404SG purified in this manner could label

cells at a much lower concentration of 200 nM 404SG (Figure 3-13i). Pre-incubation of the

404SG with soluble EGF ligand only slightly increased antigen binding (Figure 3-13j). In

addition, clones CAH-B4-1 and MGE-B6-2 were tested for binding to EGFR on the surface of

109 A431 cancer cells (expressing ~2x106 receptors/cell) using monolayer cell panning (35). In this

technique, yeast displaying antibodies are captured by adherent mammalian cells expressing an

antigen of interest (protocol in Appendix 2). No appreciable binding of these clones was seen in

this assay (data not shown), but their affinities may have been below the sensitivity of the method. scFvs selected against 404SG were tested for binding to their cognate peptides, showing that some clones still retained affinity for the peptides while other no longer did (Figure 3-14).

Figure 3-14: Binding of clones to peptides. Yeast were labeled with chicken anti-c-myc IgY (abscissa) and 1 μM peptide (ordinate), followed by goat anti-chicken-Alexa488 IgG and streptavidin-PE. (a) Clone CAH-B4-1 labeled with CAH peptide. (b-k) Clones MGE-B6-1 to MGE-B6-10 labeled with MGE peptide.

3.3.4 Third round of affinity maturation

In this round, the affinity maturation against 404SG antigen was narrowed down to libraries based on two peptides, LYN and MGE. Random mutagenesis libraries were again

110 constructed using nucleotide analog mutagenesis of the DNA from the population enriched in the

previous round. The libraries were transformed into yeast with diversities of 5.4x106 for LYN-C and 1.8x106 for MGE-C. For the LYN-C library, 2 FACS selections were performed, with the

first selection against 0.5 μM 404SG and the second at 200 nM 404SG. The enriched population

(LYN-C2) was sequenced, and representative clones are listed in Appendix 1.15. Individual

clones were also tested for their binding to the 404SG antigen (Figure 3-15). The vertical pattern

of binding seen in the previous round was again observed. The addition of soluble EGF did not

improve 404SG binding, and clones MGE-C2-1, 3, 5, 6, and 7 did not show appreciable binding to 1 μM LYN peptide (data not shown).

Figure 3-15: Binding of 404SG to LYN-C2 clones. Yeast displaying the scFvs were labeled with chicken anti-c- myc IgY (abscissa) and 200 nM 404SG antigen which had been pre-incubated with biotinylated M2 anti-flag IgG (ordinate), followed by goat anti-chicken-Alexa488 IgG and streptavidin-PE. (a) LYN-C2-1 (b) LYN-C2-3 (c) LYN-C2-5 (d) LYN-C2-6 (e) LYN-C2-7 (f) LYN-C2-8 (g) LYN-C2-10 (h) LYN-C2-11.

Library MGE-C was sorted against 500 nM 404SG using 4 FACS selections with no

enrichment for antigen binders. In addition, it was noted that the quality of the 404SG protein

varied between different preparations. Thus, we returned to using peptide as an antigen source,

111 and the CRN peptide was used for new selections. A library based on CRN-A from the first round was made, and this library (CRN-2B) had a diversity of 1.2x107. 6 FACS selections were performed, starting at 500 nM CRN peptide decreasing to 10 nM. Clones from this enriched population (CRN-2B6) were sequenced and are listed in Appendix 1.16. Based on the labeling of these clones, they appeared to have affinities on the order of 100 nM. To increase this affinity, the CRN-C library (6x107 diversity) was made based on random mutagenesis of the selected clones from the previous round. After 5 FACS selections at 10 nM peptide, the enriched CRN-

C5 population was sequenced. The two best clones, CRN-C5-1 and CRN-C-11 (sequences in

Appendix 1.16) were titrated on the surface of yeast and found to have ~50 nM affinity for the

CRN peptide (data not shown). However, clone CRN-C5-1 showed no appreciable binding to

400 nM 404SG or 1 μM Alexa488-labeled EGFR ectodomain produced by insect cells

(generously supplied by Jennifer R. Cochran). Likewise, clone CRN-C5-11 did not bind 200 nM

404SG or 400 nM insect cell EGFR.

3.3.5 Fab display libraries

At this point, all antibody selections were moved from scFv display to Fab display on the surface of yeast. Eventually, we envisioned reformatting our selected scFv clones into full IgG molecules for cancer therapy. However, it has been observed that antibody affinity may change during reformatting (36). Therefore, Fab display enabled testing of the effects of adding the antibody constant regions to the variable regions, while still being able to work conveniently with yeast. The scFv clones LYN-C2-6, LYN-C2-13, CRN-C5-1, and CRN-C5-4 were reformatted to Fabs displayed on the surface of yeast. The yeast displaying these clones were tested for the presence of both heavy and light chains (Figure 3-16a to d). The LYN-C2 clones

112 appeared to reformat well, with good association of the heavy and light chains, and the Fab

clones retained binding to the antigen (Figure 3-16e,f). However, the CRN-C5 clones did not

reformat well, with little light chain captured by the heavy chain and displayed on the surface.

Consequently, these clones also showed poor antigen binding, demonstrating that both heavy and light chains were necessary for peptide binding.

Figure 3-16: Properties of Fab-reformatted clones. For (a-d), yeast displaying Fabs were labeled with chicken anti- c-myc and mouse anti-V5 to detect the VL (ordinate), followed by goat anti-chicken-Alexa488 and goat anti-mouse- PE IgG. (a) Clone LYN-C2-6 (b) LYN-C2-13 (c) CRN-C5-1 (d) CRN-C5-4. For (e-h), the clones in (a-d) were labeled with chicken anti-c-myc to detect the VH (abscissa) and antigen (ordinate). For (e-f), yeast were labeled with 200 nM 404SG antigen which had been pre-incubated with biotinylated M2 anti-flag IgG, followed by streptavidin- PE. For (g-h), yeast were labeled with 1 μM CRN peptide followed by streptavidin-PE. Cells in the c-myc positive, V5 negative quadrant may represent yeast which contain the heavy chain plasmid but have lost the light chain plasmid.

scFv clones from the MGE-A4 and MGE-B6 populations were also reformatted into Fab

display vectors and showed good association between the heavy and light chains on the surface

of yeast. Therefore, a new library was made using nucleotide analog mutagenesis of the variable

regions. The heavy and light libraries were transformed separately into two different strains of yeast, and the Fab library was created using yeast mating of the heavy and light libraries.

Because of the uncertainty of the quality of the 404SG antigen, the Fab library was sorted against

113 MGE peptide using 6 FACS selections starting at 1 μM peptide decreasing to 250 nM. 3-color sorts (as shown in Figure 3-4) were performed to select clones that showed both heavy and light chain expression, as well as antigen binding. The enriched population (MGE-Fab-M6) was sequenced (listed in Appendix 1.17). Of the 5 sequenced light chains, 4 contained only a

fragment of the variable light chain consisting of the first 12 amino acids of a variable region

fused directly to the constant region (Figure 3-17). Sequencing of the original unselected library

demonstrated that this was a rare event in the starting library, and thus these clones had been

selected for better antigen binding. It appeared these clones preferred to be heavy chain only

binders since they were originally selected as such. Most likely, the MGE peptide was not

binding to the antibody CDRs, but instead to somewhere near the VH/VL interface. Another

single-domain antibody, a light chain clone, was previously selected using yeast surface display

(13) and has been shown to bind antigen at the VH/VL interface (Arne Skerra, personal

communication). During selections of light chains compatible with the original heavy chains, we

may have simply selected for light chains which did not associate well with the heavy chains (but

were still connected through the linker), thus allowing the peptide access to the binding site.

Clones MGE-Fab-M6-1, 2, and 3 were tested by mammalian cell panning for binding to EGFR

expressed on NR6 cells (~105 receptors/cell), and no appreciable binding was observed.

<----VL----><------CL------> DIQVTRARMETKRTVAAPSVFIFPPSDEQLKSGTASVVCLLNNFYPREAKVQWKVDNAL

Figure 3-17: Light chain sequence of clone MGE-Fab-M6-1. The variable light region (VL) is truncated to 12 amino acids, and the beginning portion of the constant region (CL) is shown.

114 3.3.6 Rabbit monoclonal antibodies

Since the leads isolated from the nonimmune human library in yeast did not appear

promising, we sought to obtain antibodies from rabbits against the LYN and MGE peptides.

Rabbits typically produce higher affinity antibodies than mice due to the use of both gene

conversion and somatic hypermutation for antibody diversification (37). In addition, previous

work had demonstrated that rabbits injected with these peptides were capable of generating

antibodies against EGFR (38). Thus, rabbits were immunized with the peptides, and the binding

of the rabbit antisera to A431 cancer cells was tested. The rabbits did not appear to produce a

response against the LYN peptide (data not shown), but one rabbit showed a response against the

MGE peptide (Figure 3-18). The binding of the rabbit antiserum to the A431 cells could also be

competed off by excess MGE peptide. In addition, the MGE antiserum appeared to inhibit EGFR

phosphorylation, although the pre-bleed showed inhibition of the receptor to some extent as well

(data not shown).

Figure 3-18: FACS data for rabbit polyclonal antiserum against MGE peptide bound to A431 cells. Cells were labeled overnight at 4 oC with rabbit pre-bleed or post-peptide immunization bleed, followed by goat anti-rabbit-PE. Antiserum binding was competed down to pre-bleed levels using 200 μM MGE peptide, but not LYN peptide.

The rabbit which produced a response against the MGE peptide was sacrificed, and its spleen

cells were fused with a plasmacytoma cell line to generate immortalized hybridoma cells (39).

115 The hybridoma supernatants were screened for their binding to A431 cells, and three hybridomas

were selected for further subcloning (Figure 3-19).

Figure 3-19: Binding of hybridoma supernatants to A431 cells. Cells were labeled with supernatant at 4 oC for (a) 2 h or (b) overnight, followed by goat anti-rabbit-PE.

The sequences of the rabbit monoclonal antibodies secreted by each of the subcloned

hybridomas was determined (Figure 3-20). All three clones contained kappa light chains, with

high sequence similarity between antibodies 13 and 32. During the course of sequencing, it was

determined that antibody 31 was not an IgG molecule, but some other class of immunoglobulin.

Thus, the purification of antibody 31 was unsuccessful, with the antibody precipitating out of

solution. However, antibodies 13 and 32 were purified for testing on A431 cells. Both antibodies

showed unusual labeling, with a large negative peak of cells even after overnight incubation with

the antibody (Figure 3-21a), although this binding could be competed by excess MGE peptide. It

was determined that antibody 13 only labeled dead cells, as shown by co-staining with 7-AAD

(Figure 3-21b). The antibody may be recognizing a different protein which is only accessible for

binding once the cell membrane has been compromised, or it may only be able to bind to slightly

degraded EGFR in dead cells.

116 VH-FR1 VH-CDR1 VH-FR2 VH-CDR2 13 QSVEESGGRLVTPGTPLTLTCTVS GFSLSTHGMI WVRQAPGEGLEWIG IINSNHYTA 31 ------SV--S ------K------NYGRAY 32 ------N-Y------Y-- VV-TD-T-Y

VH-FR3 VH-CDR3 13 YANWAKGRVTISKTSTTVDLKMTSLTTEDTAIYFCGR RGNFAFNL WGQGTLVTVSS 31 --S-----F---R------I--P------T---A- IYPDDSIGNL ------32 --S-V---F------AA---T------S------

VL-FR1 VL-CDR1 VL-FR2 VL-CDR2 13 AQVLTQTASSVSAAVGGTVTISC QSSQSVYNNNWLA WYQQKAGQPPNLLIY QASKLAS 31 DI-M---PA--E------K- -A-ERIG-ALA -----P----K---- G--T-T- 32 -R-----P-P------E---R------P----K------

VL-FR3 VL-CDR3 13 GVPSRFSGSGSGTHFTLTISGVQCDDAATYYC QGEFSCSIADCVA FGGGTEVVVK 31 ------K------Q------DLE-A------SYYASGSRIYGYD ------32 ------Q------RSG--I------Figure 3-20: Sequences of rabbit monoclonal antibodies.

Figure 3-21: Flow cytometry data for rabbit monoclonal antibody 13 binding to A431 cells. (a) Cells were labeled with 133 nM antibody 13 overnight at 4 oC, and binding was competed with 200 μM MGE peptide. (b) Cells were labeled with 133 nM antibody 13 overnight and then labeled with 7-AAD, which stains dead cells.

The antibodies were cloned into an scFv format and displayed on the surface of yeast.

Titrations with MGE peptide indicated that the clones bound the peptide with affinities of ~25 nM. However, the clones did not bind 404SG or EGFR domain IV secreted by yeast (provided by Ben Hackel). In addition, the clones were negative for binding to EGFR on the surface of

A431 cells as assayed by cell monolayer panning. Therefore, it appeared that the rabbit had produced antibodies which bound to the MGE peptide, but not to EGFR.

117 3.4 Discussion

In this work, we have continued efforts to engineer antibodies against epitopes in

domains II and IV of EGFR. To do so, we used previously designed peptide mimics of EGFR

loops as antigens. However, many of the antibodies we selected were specific for the peptide

mimcs, but not the full EGFR protein (such as the clones against peptides CRN and MGE). This

suggests that the peptides may not accurately represent the conformations of the EGFR loops.

However, it has been shown that rabbits are capable of producing antibodies against the MGE

peptide that also bind EGFR on the surface of cells (38). Another possibility is that the some of

the peptides are accurate mimics, but the antibody clones which we happened to select bound the

peptides in such a manner that would not allow EGFR protein binding.

Another issue that arose during antibody engineering was inconsistency in antigen

quality. The yeast 404SG EGFR ectodomain mutant, even when affinity purified against a conformation-specific antibody, showed variation between batches. In addition, it was later discovered that the EGFR ectodomain mostly adopts a closed or tethered conformation when in solution (40), thus masking many of the targeted domain II and IV epitopes. The addition of EGF to 404SG did not appear to improve binding, presumably because these epitopes were also masked in the dimerized EGFR. In an attempt to untether the EGFR ectodomain, mutants directed at disrupting the domain II/domain IV tether were secreted (Y246D, H566F,

Y251A/R285S, and K585A). However, these mutants did not show increased binding to the antibodies, and it was later published that the domain II/domain IV tether does not contribute significantly to maintaining the receptor in the autoinhibited state (40). Finally, since domain IV is unresolved in the EGFR dimer crystal structure (24), it may be inherently disordered and thus difficult to bind.

118 In order to overcome these problems for selections against EGFR domain IV in the

future, one possibility is to use the EGFR ectodomain mutant Y251A/R285S plus EGF as an antigen. This mutant has been shown to adopt the extended conformation, but not dimerize, upon addition of EGF ligand (40). An isolated EGFR domain IV could also be used, but it would need to be extensively characterized for proper folding. Finally, if the receptor could be constrained in such a way as to prevent dimerization, such as C-terminal immobilization on a surface, EGF could be added to untether the receptor.

119 3.5 Works cited

1. Hoogenboom, H. R. (2002) Overview of antibody phage-display technology and its applications, Methods Mol Biol 178, 1-37. 2. Lipovsek, D., and Pluckthun, A. (2004) In-vitro protein evolution by ribosome display and mRNA display, J Immunol Methods 290, 51-67. 3. Boder, E. T., and Wittrup, K. D. (1997) Yeast surface display for screening combinatorial polypeptide libraries, Nat Biotechnol 15, 553-557. 4. Chao, G., Lau, W. L., Hackel, B. J., Sazinsky, S. L., Lippow, S. M., et al. (2006) Isolating and engineering human antibodies using yeast surface display, Nat Protoc 1, 755-768. 5. van den Beucken, T., Pieters, H., Steukers, M., van der Vaart, M., Ladner, R. C., et al. (2003) Affinity maturation of Fab antibody fragments by fluorescent-activated cell sorting of yeast-displayed libraries, FEBS Lett 546, 288-294. 6. Weaver-Feldhaus, J. M., Lou, J., Coleman, J. R., Siegel, R. W., Marks, J. D., et al. (2004) Yeast mating for combinatorial Fab library generation and surface display, FEBS Lett 564, 24-34. 7. Shusta, E. V., Kieke, M. C., Parke, E., Kranz, D. M., and Wittrup, K. D. (1999) Yeast polypeptide fusion surface display levels predict thermal stability and soluble secretion efficiency, J Mol Biol 292, 949-956. 8. Lipovsek, D., Lippow, S. M., Hackel, B. J., Gregson, M. W., Cheng, P., et al. (2007) Evolution of an interloop disulfide bond in high-affinity antibody mimics based on fibronectin type III domain and selected by yeast surface display: molecular convergence with single-domain camelid and shark antibodies, J Mol Biol 368, 1024-1041. 9. Feldhaus, M. J., Siegel, R. W., Opresko, L. K., Coleman, J. R., Feldhaus, J. M., et al. (2003) Flow-cytometric isolation of human antibodies from a nonimmune Saccharomyces cerevisiae surface display library, Nat Biotechnol 21, 163-170. 10. Hoogenboom, H. R. (2005) Selecting and screening recombinant antibody libraries, Nat Biotechnol 23, 1105-1116. 11. Bowley, D. R., Labrijn, A. F., Zwick, M. B., and Burton, D. R. (2007) Antigen selection from an HIV-1 immune antibody library displayed on yeast yields many novel antibodies compared to selection from the same library displayed on phage, Protein Eng Des Sel 20, 81-90. 12. Kieke, M. C., Cho, B. K., Boder, E. T., Kranz, D. M., and Wittrup, K. D. (1997) Isolation of anti-T cell receptor scFv mutants by yeast surface display, Protein Eng 10, 1303-1310. 13. Colby, D. W., Garg, P., Holden, T., Chao, G., Webster, J. M., et al. (2004) Development of a human light chain variable domain (V(L)) intracellular antibody specific for the amino terminus of huntingtin via yeast surface display, J Mol Biol 342, 901-912. 14. Graff, C. P., Chester, K., Begent, R., and Wittrup, K. D. (2004) Directed evolution of an anti-carcinoembryonic antigen scFv with a 4-day monovalent dissociation half-time at 37 degrees C, Protein Eng Des Sel 17, 293-304. 15. Razai, A., Garcia-Rodriguez, C., Lou, J., Geren, I. N., Forsyth, C. M., et al. (2005) Molecular evolution of antibody affinity for sensitive detection of botulinum neurotoxin type A, J Mol Biol 351, 158-169.

120 16. Boder, E. T., Midelfort, K. S., and Wittrup, K. D. (2000) Directed evolution of antibody fragments with monovalent femtomolar antigen-binding affinity, Proc Natl Acad Sci U S A 97, 10701-10705. 17. Cunningham, D., Humblet, Y., Siena, S., Khayat, D., Bleiberg, H., et al. (2004) Cetuximab monotherapy and cetuximab plus irinotecan in irinotecan-refractory metastatic colorectal cancer, N Engl J Med 351, 337-345. 18. Messersmith, W. A., and Hidalgo, M. (2007) Panitumumab, a monoclonal anti epidermal growth factor receptor antibody in colorectal cancer: another one or the one?, Clin Cancer Res 13, 4664-4666. 19. Seiden, M. V., Burris, H. A., Matulonis, U., Hall, J. B., Armstrong, D. K., et al. (2007) A phase II trial of EMD72000 (matuzumab), a humanized anti-EGFR monoclonal antibody, in patients with platinum-resistant ovarian and primary peritoneal malignancies, Gynecol Oncol 104, 727-731. 20. Sizeland, A. M., and Burgess, A. W. (1992) Anti-sense transforming growth factor alpha oligonucleotides inhibit autocrine stimulated proliferation of a colon carcinoma cell line, Mol Biol Cell 3, 1235-1243. 21. Sok, J. C., Coppelli, F. M., Thomas, S. M., Lango, M. N., Xi, S., et al. (2006) Mutant epidermal growth factor receptor (EGFRvIII) contributes to head and neck cancer growth and resistance to EGFR targeting, Clin Cancer Res 12, 5064-5073. 22. Nishikawa, R., Ji, X. D., Harmon, R. C., Lazar, C. S., Gill, G. N., et al. (1994) A mutant epidermal growth factor receptor common in human glioma confers enhanced tumorigenicity, Proc Natl Acad Sci U S A 91, 7727-7731. 23. Garrett, T. P., McKern, N. M., Lou, M., Elleman, T. C., Adams, T. E., et al. (2002) Crystal structure of a truncated epidermal growth factor receptor extracellular domain bound to transforming growth factor alpha, Cell 110, 763-773. 24. Ogiso, H., Ishitani, R., Nureki, O., Fukai, S., Yamanaka, M., et al. (2002) Crystal structure of the complex of human epidermal growth factor and receptor extracellular domains, Cell 110, 775-787. 25. Franklin, M. C., Carey, K. D., Vajdos, F. F., Leahy, D. J., de Vos, A. M., et al. (2004) Insights into ErbB signaling from the structure of the ErbB2-pertuzumab complex, Cancer Cell 5, 317-328. 26. Cho, H. S., Mason, K., Ramyar, K. X., Stanley, A. M., Gabelli, S. B., et al. (2003) Structure of the extracellular region of HER2 alone and in complex with the Herceptin Fab, Nature 421, 756-760. 27. Berezov, A., Chen, J., Liu, Q., Zhang, H. T., Greene, M. I., et al. (2002) Disabling receptor ensembles with rationally designed interface peptidomimetics, J Biol Chem 277, 28330-28339. 28. Baron, A. T., Huntley, B. K., Lafky, J. M., Reiter, J. L., Liebenow, J., et al. (1997) Monoclonal antibodies specific for peptide epitopes of the epidermal growth factor receptor's extracellular domain, Hybridoma 16, 259-271. 29. Robinson, A. S., Hines, V., and Wittrup, K. D. (1994) Protein disulfide isomerase overexpression increases secretion of foreign proteins in Saccharomyces cerevisiae, Biotechnology (N Y) 12, 381-384. 30. Kim, Y. S., Bhandari, R., Cochran, J. R., Kuriyan, J., and Wittrup, K. D. (2006) Directed evolution of the epidermal growth factor receptor extracellular domain for expression in yeast, Proteins 62, 1026-1035.

121 31. Boder, E. T., and Wittrup, K. D. (2000) Yeast surface display for directed evolution of protein expression, affinity, and stability, Methods Enzymol 328, 430-444. 32. Colby, D. W., Kellogg, B. A., Graff, C. P., Yeung, Y. A., Swers, J. S., et al. (2004) Engineering antibody affinity by yeast surface display, Methods Enzymol 388, 348-358. 33. Chen, P., Gupta, K., and Wells, A. (1994) Cell movement elicited by epidermal growth factor receptor requires kinase and autophosphorylation but is separable from mitogenesis, J Cell Biol 124, 547-555. 34. Holt, L. J., Herring, C., Jespers, L. S., Woolven, B. P., and Tomlinson, I. M. (2003) Domain antibodies: proteins for therapy, Trends Biotechnol 21, 484-490. 35. Wang, X. X., and Shusta, E. V. (2005) The use of scFv-displaying yeast in mammalian cell surface selections, J Immunol Methods 304, 30-42. 36. Yeung, Y. A. (2005) Antibody engineering for cancer therapy, Ph.D. Thesis, Massachusetts Institute of Technology, MA. 37. Mage, R. G., Lanning, D., and Knight, K. L. (2006) B cell and antibody repertoire development in rabbits: the requirement of gut-associated lymphoid tissues, Dev Comp Immunol 30, 137-153. 38. Wolf Yadlin, A. (2006) Development of mass spectrometry based technologies for quantitative cell signaling phosphoproteomics: the epidermal growth factor receptor family as a model system, Ph.D. Thesis, Massachusetts Institute of Technology, MA. 39. Spieker-Polet, H., Sethupathi, P., Yam, P. C., and Knight, K. L. (1995) Rabbit monoclonal antibodies: generating a fusion partner to produce rabbit-rabbit hybridomas, Proc Natl Acad Sci U S A 92, 9348-9352. 40. Dawson, J. P., Bu, Z., and Lemmon, M. A. (2007) Ligand-Induced Structural Transitions in ErbB Receptor Extracellular Domains, Structure 15, 942-954.

122 Chapter 4: Mathematical modeling of high affinity EGF binding

4.1 Background

Equilibrium titrations of EGFR on the surface of cells reveal two populations of EGFR which appear to bind EGF with high and low affinity (1). The two populations become apparent

at low EGF concentrations, where the binding signal is higher than one would expect based on

simple receptor-ligand equilibrium. An alternate representation of equilibrium titration data is a

Scatchard plot, where bound ligand divided by free ligand is plotted versus bound ligand (2).

High affinity EGF binding thus manifests itself as a concave-up Scatchard plot. The high affinity

receptors comprise ~5-10% of the total population with an apparent EGF affinity in the range of

100-300 pM, while the remaining receptors bind EGF with an affinity of 2-15 nM (3).

Although ligand binding is the main driving force for receptor dimerization, a small

number of preformed dimers have been observed on the surface of cells in the absence of ligand.

These preformed dimers have been studied using imaging (4-6) and biochemical (7,8)

techniques, but they are only basally phosphorylated (7). This suggests that there may be inactive

and active conformations of the dimer, with ligand binding inducing a conformational change or

rotation leading to activation (4,9). In addition, early experiments suggested that purified EGFR dimers bound EGF with a higher affinity than isolated monomers (10,11). Introduction of a cysteine residue in the extracellular juxtamembrane region of EGFR led to the formation of covalent dimers and an increase in the proportion of high affinity receptors (12). Imaging techniques have also found that EGF bound EGFR dimers faster than monomers (13).

Taken together, this experimental evidence suggested that EGFR dimers may be responsible for the high affinity population of receptors seen on the cell surface. Subsequent

123 mathematical modeling of ligand binding and dimerization events using this experimental

information was able to produce concave-up Scatchard plots; however, the model was unable to

adequately fit experimental data (14). Expansions to this model which incorporate heterogeneity

in receptor density on the cell surface can sufficiently fit experimental data (14,15). Additional

models have been developed where receptor binding to a hypothetical third site leads to the

experimentally observed high affinity EGF binding. These models can either adequately fit

experimental data without modeling dimerization events (16) or produce concave-up Scatchard

curvature while modeling dimerization and monomer untethering events (17). Overall, there is

still a lack of consensus on the origin of the apparent high affinity population of EGFR.

In previous work by Alejandro Wolf-Yadlin, rabbit polyclonal antisera against the LYN

and MGE peptides (see Figure 3-6 and Section 3.1) were characterized. The LYN peptide

mimics the EGFR domain II (DII) dimerization arm, and the MGE peptide represents an EGFR

domain IV (DIV) epitope. Based on the epitopes of the rabbit polyclonal antibodies (pAbs), they

were expected to inhibit receptor homo- and heterodimerization. Accordingly, it was found that

both pAbs inhibited EGFR phosphorylation at tyrosine 1173 and trans-phosphorylation of Her2

(18). This inhibition was not caused by the blockade of ligand binding, as the pAbs and EGF could simultaneously bind mammalian cells expressing EGFR at saturating ligand concentrations. However, pretreatment of cells with the pAbs led to a blockade of EGF binding to only the high affinity receptor population, as revealed by equilibrium titrations (18).

124 In the present thesis work, we have used this experimental result to develop a mathematical model describing high affinity EGF binding. The equilibrium model for receptor dimerization and ligand binding assumes that the high affinity EGF binding component is due to inactive preformed dimers and that the antibodies inhibit receptor homo- and heterodimerization.

The model can successfully reproduce the previous experimental data with only three fit parameters and can produce a concave-up Scatchard plot for a large range of parameter values.

125 4.2 Results

4.2.1 Model formulation and output

Previous experiments revealed that pAbs against either EGFR DII or DIV eliminated

high affinity EGF binding (18). Based on the epitopes of the pAbs and their ability to inhibit

receptor phosphorylation, we hypothesized that the pAbs were blocking receptor dimerization.

Therefore, it appeared that receptor dimerization was involved in high affinity EGF binding;

specifically, it was hypothesized that preformed receptor dimers were the high affinity binding

component. We developed an equilibrium ligand binding and dimerization model incorporating these hypotheses to test whether such a model could be consistent with the experimental data.

The molecular events described by the model are indicated in Figure 4-1. Reactions 1-6 detail

EGF binding to and dimerization of the extended EGFR monomer, the scheme proposed in (14).

Reactions 7-8 indicate the tethered monomer converting to the untethered state and binding to

ligand as modeled in (17). Finally, reactions 9-11 are necessary to describe the analogous

heterodimerization events, since the cells used in the experiments, human mammary epithelial

cells (HMECs) (18), also express HER2.

126

Figure 4-1: Model schematic. The equilibrium binding of EGF ligand to monomeric extended EGFR is represented by reaction 1. Reactions 2 and 3 represent the binding of EGF to unoccupied and singly occupied preformed EGF dimers, respectively. Reactions 4-6 represent dimerization of occupied and unoccupied EGFR extended monomers. Reaction 7 represents the tethered EGFR monomer becoming untethered, and reaction 8 represents EGF binding to tethered EGFR. Reactions 9-10 represent heterodimerization with HER2, and reaction 11 represents EGF binding to a preformed heterodimer. Each reaction is shown with its relevant equilibrium dissociation constant value.

When the HMECs are pretreated with either the DII or DIV pAb, both dimerization and tether formation could potentially be inhibited. In this case, the reactions outlined in Figure 4-1 would simplify to only the first reaction, which corresponds to a single-site binding model characteristic of simple receptor-ligand binding equilibrium. Then, the fractional EGF binding yEGF is described by the equation

max Ly yEGF = , (1) + KL 1

127 where L is the free ligand concentration. The data for cells pretreated with pAb were first fit to this simple equation varying two parameters: the single-site equilibrium binding constant K1 and the maximum binding signal ymax. As seen in Figure 4-2a, the data fit the single-site binding

model well for both pAbs, giving K1 = 6.04 nM (68% confidence interval [4.95, 7.37] by support

plane analysis (19)) for the DII pAb and K1 = 7.34 nM (68% confidence interval [6.05, 8.90]) for

the DIV pAb. This value for the dissociation constant is consistent with previous estimates for

the low affinity EGFR population (3,16), and is also consistent with surface plasmon resonance

measurements for soluble untethered EGFR ectodomain (C-terminal truncation starting at

residue 502) binding to immobilized EGF (20). The pAb pretreatment titration data is also

plotted in Figure 4-2c,d as Scatchard plots. Although the data points below 100 pM bound EGF

may appear to give a slightly concave-down Scatchard plot, this is likely due to experimental

error at these low concentration points. While Scatchard representations are useful for

visualization of data, the transformation distorts errors, especially at low ligand concentrations

(21). It appears from the data that the signals from these low concentration points may be below

the sensitivity of the experimental binding measurement using flow cytometry. Therefore, we

believe that the data is best fit by a single-site binding model (Figure 4-2a), which predicts linear

Scatchard plots.

128 AB0.7 1 0.6

0.8 0.5

0.6 0.4

0.3 0.4 Bound/Free EGF Bound/Free Fractional binding Fractional 0.2 0.2 0.1

0 0 0.01 0.1 1 10 100 1000 0 0.1 0.2 0.3 0.4 0.5 0.6 EGF concentration (nM) Bound EGF (nM)

CD0.2 0.2

0.1 0.1

0 0 Bound/Free EGF Bound/Free EGF Bound/Free 0 0.1 0.2 0.3 0.4 0.5 0.6 0 0.1 0.2 0.3 0.4 0.5 0.6 Bound EGF (nM) Bound EGF (nM)

Figure 4-2: Fit of model results to experimental data from (18). (a) Data from HMECs titrated at 4 oC using the indicated concentrations of Alexa488-EGF without (filled diamonds) or with pretreatment of 2 μM DII (open squares) or DIV (open triangles) pAb. Best fits of the model to the experimental data are shown in solid lines for no pAb and DII pAb pretreatment and in dotted lines for DIV pAb pretreatment. (b) Scatchard representation of titration data for no pAb pretreatment (diamonds). Best fit of the model is indicated by the solid line. The data point indicated in grey in (a) and (b) was not used for fitting the model. (b, inset) Residual errors between the model and experimental data are plotted for no pAb pretreatment. Titration data for DII pAb pretreatment (c) and DIV pAb pretreatment (d) are shown in Scatchard representation. Best fit of the model is indicated by the solid line. All error bars indicate standard deviation of at least 3 replicates.

Next, the full model shown in Figure 4-1 was used to fit the EGF titration data without pAb pretreatment. All reactions were described by mass action relations through equilibrium dissociation constants, with the following reactions corresponding to those diagrammed in Figure

4-1:

129 ⎯→←+ CLR K1 ←+ ⎯→⎯ DLD 0 K2 2/ 1 ←+ ⎯→⎯ DLD 1 2K3 2 ←+ ⎯→ DRR K4 0 RC ←+ ⎯→⎯ D K5 2/ 1 ←+ ⎯→ DCC (2) K6 2 ←⎯→ RT K7 ←+ ⎯→CLT K8 T HR ←+ ⎯→⎯ H K9 2/ 0 HC ←+ ⎯→⎯ H K10 2/ 1 ←+ ⎯→⎯ HLH 0 K11 1

From these reactions, the following relations were derived:

RL RL K1 = C = (3) C K1

2 2 2 0 LD DK 12 2 2 LRK R KK 42 K 2 = D0 == ⇒= K5 = (4) D1 2L 2 51 LKK K 4 K1

22 22 1LD 1LD 2 LR LR KK 53 K3 = D2 == 2 ⇒= K6 = (5) 2D2 2K3 2 KKK 531 1 KK 6 K1

R 2 R 2 K 4 = D0 = (6) D0 K 4

2CR CR 22 2 LR K5 = D1 == (7) D1 K5 KK 51

C 2 C 2 LR 22 K6 = D2 == 2 (8) D2 K6 1 KK 6

T K = = RKT (9) 7 R 7

130 TL TL 7 RLK K8 = CT == (10) CT K8 K8

2RH 2RH K9 = H 0 = (11) H 0 K9

2CH CH 22 RHL K10 = H1 == (12) H1 K10 KK 101

0 LH 0 LH RHL 22 RHL KK 119 K11 = H1 ⇒=== K10 = (13) H1 K11 KK KK 101119 K1

Using a mass balance on total Her2 receptors (HT = 0.075 nM in the experiments) yields an expression for the concentration of Her2:

T ++= HHHH 10 (14)

RH 22 RHL T HH ++= (15) K KK 1019

KKKH H = T 1091 (16) 1091 101 ++ 22 9 RLKRKKKKK

Likewise, a total EGFR balance yields an equation relating the concentration of EGFR (RT = 0.5 nM in the experiments) to free ligand, L:

T 222 210 ++++++= T + + HHCTDDDCRR 10 (17)

2 2 22 RL R LR 242 LR 7 RLK T RR 2 7 RK +++++++= K K1 K 4 KK 51 1 KK 6 K8

2 KKKRH 2 KKKRLH T 1091 + T 1091 (18) ()10919 101 ++ 22 9 RLKRKKKKKK ()1091101 101 ++ 22 9 RLKRKKKKKKK

131 To solve for R, Equation 18 was transformed to the form:

3 2 DCRBRAR =+++ 0 (19)

where

2 2 4( 1 + 2 + 5464165 )( + 9101 LKKKLKKLKKKKKK ) A = 2 (20) 1 KKKK 654

⎛ L 242 L2 ⎞ ⎛ L LK ⎞ = KKKB ⎜ ++ ⎟ 2 ⎜1 KLKKK +++++ 7 ⎟ (21) 1091 ⎜ 2 ⎟ ()9101 ⎜ 7 ⎟ ⎝ K KK 514 1 KK 6 ⎠ ⎝ K1 K8 ⎠

⎛ L LK ⎞ ⎜ 7 ⎟ (22) = 1091 ⎜1 KKKKC 7 +++ ⎟ 2()++ 9101 (− RHLKKK TT ) ⎝ K1 K8 ⎠

−= T KKKRD 1091 (23)

From Equation 19, R can be calculated and then used to compute the total amount of bound ligand by summing the different ligand-bound species as a function of L:

2 21 T ++++= HCDDCbound 1 (24)

RL 2 LR 22 LR 22 RLK 2RL ⎛ KKKH ⎞ bound 7 ++++= ⎜ T 1091 ⎟ (25) 2 ⎜ ⎟ K1 KK 51 1 KK 6 K8 KK 101 ⎝ 1091 + 2 101 + 2 9 RLKRKKKKK ⎠

Nonlinear least squares fit of the model parameters to experimental data was performed

using MATLAB (Mathworks) function lsqnonlin. Despite what appears to be many free

parameters, thermodynamic and physical constraints reduce the number of fit parameters to

three. Representative values for model parameter K1 were determined earlier from fitting the

pAb pretreatment data to a single-site binding model, and therefore K1 was set equal to 7 nM. K2 represents the binding constant for the high affinity preformed dimers and was left as a free

132 parameter for data fitting. K3 represents the affinity of binding a second EGF to a singly

occupied EGFR dimer. Previously, it has been shown that for an equilibrium dimerization/ligand-binding model such as this one, K3 must be greater than K2 to produce a

concave-up Scatchard plot (14). Also, if K3 were equal to K2, the dimerization of bound and unbound monomers (reaction 5) would lead to the creation of many more high affinity sites than are experimentally observed. To reduce the number of fit parameters, as a simplifying

assumption K3 was therefore set equal to K1, the binding constant for the low affinity binding

site. K4 represents the dimerization or cross-linking coefficient, and was left as a free parameter.

Three of the equilibrium constants are not independent; K5, K6, and K10 can be derived as a

function of the other parameters (see Equations 4, 5, and 13). K7 was set to 30 according to the

estimated energy of untethering (17,20,22), and K8 was set to 400 nM as the affinity of EGF for

EGFR domain III (17,23). K9, the EGFR/HER2 heterodimerization constant, was set equal to the

EGFR homodimerization constant K4. Modeling and experiments have shown that homo- and

heterodimerization occur with comparable affinities, with the prevalence of heterodimers

observed in many cell types arising from a combination of relative expression levels and

differential endocytic trafficking properties (24). Finally, K11 was set equal to K2, the high

affinity interaction, since increasing expression of HER2 has been shown to increase the fraction

of receptors in the high affinity state (25). The EGFR and HER2 receptor concentrations, RT and

HT, respectively, were calculated as the total average concentration in the sample based on the average number of receptors per cell and number of cells per labeling volume. Therefore, for modeling of the EGF titration without pAb pretreatment, a total of three parameters were varied to fit the data: K2, K4, and the maximum binding signal.

133 4.2.2 Model characterization

As seen in Figure 4-2, the model fits the EGF titration data well. The residual errors

between the model and the data are randomly distributed around zero, indicating an adequate fit

(Figure 4-2b, inset). The model and experimental data are also shown in Scatchard format in

Figure 4-2b. One data point, indicated in gray, was determined to be an outlier and not used for

the model fitting. One possible explanation is that the reaction was under depleting ligand

conditions, leading to a lower signal than expected. Nevertheless, if this data point was used in

the analysis, the fit parameters were not affected by more than twofold, and the model fit was

only slightly worse (data not shown). The best-fit parameters from the model were K2 = 110 pM

for high affinity EGF binding to preformed dimers and K4 = 13 pM for the dimerization constant.

This K2 value is consistent with previous estimates for the high affinity EGFR population (3,16).

For EGFR on the surface of cells, the value of K4 depends on the total receptor concentration

used in the model and is normally within an order of magnitude of RT (17), as it is in this case.

Also, K4 and RT being within an order of magnitude is required if both receptor monomers and dimers are present in the absence of ligand, which has been experimentally observed (4-7).

Although K3 was fixed and set equal to the binding constant for the low affinity site K1 (7 nM), allowing K3 to vary as a free parameter yielded similar results, with K3 = 18 nM, indicating a low

affinity binding site.

For the HMECs, the model predicts that in the absence of ligand, the EGF receptors are

distributed as 81% tethered monomers, 6% preformed homodimers, 10% preformed

heterodimers with HER2, and 3% untethered monomers. Half of the homodimers plus the

heterodimers, ~13% of the receptors in this case, are thus predicted to be in a high affinity state.

134 While the percentage of receptors in the high affinity state will vary according to cell line and

EGFR and HER2 expression levels, this percentage is within the generally reported range (3,6).

In order to assess the parameter space of the model, a support plane analysis (19) was performed (Figure 4-3a). The analysis yielded a defined confidence interval when the product of

the high affinity binding constant and the receptor dimerization constant, K2K4, was used as a

lumped parameter. K2K4 was varied, and the remaining parameters, K4 and maximum signal,

were re-fit at each of the values of K2K4. This was repeated until the resulting error exceeded an acceptable value as judged by the appropriate F-statistic for the system, yielding a 68%

-4 -3 confidence interval of [4.6x10 , 2.4x10 ] for K2K4. Next, we sought to determine how well the

model can capture the experimentally observed concave-up Scatchard behavior of EGF binding.

K2 and K4 were varied over 9 orders of magnitude, and the concavity of the resulting Scatchard

plot was determined. For the model, a concave-up Scatchard plot is obtained for a wide range of values of K2 and K4 (Figure 4-3b).

135 (a)

χ 2 2 χ min

2 K2K4 (nM )

(b)

101

10-1

(nM) -3 4 10 K

10-5

10-7 10-5 10-3 10-1 101 K (nM) 2

Figure 4-3: Model characterization. (a) Support plane analysis of parameter space of the model. The product K2K4 was varied, and the remaining parameters K4 and maximum signal were re-fit at each value of K2K4. The y-axis shows the error at the corresponding value of K2K4 divided by the minimum error (when all 3 parameters are allowed to vary). The dotted line indicates the 68% confidence interval. (b) Parameter space giving a concave-up Scatchard plot. Parameters K2 and K4 were varied, and the shaded area indicates parameter values for which the resulting Scatchard plot was concave-up. Scatchard plots were considered concave-up if all calculated points were below the straight line connecting the two endpoints of the curve.

136 4.3 Discussion

Since pretreatment of cells with the DII and DIV pAbs eliminated the high affinity EGF-

binding component typically seen with cell-surface EGFR, we speculated that pAb binding

inhibited receptor homo- and heterodimerization and that dimers may represent that high affinity

binding mode. To probe this idea, we formulated an equilibrium model where preformed dimers were the high affinity component, and the model (Figure 4-1) was successfully fit to the experimental data (Figure 4-2). Further analysis of the model demonstrated its appropriateness to

the system, as a wide variety of parameter values could yield a concave-up Scatchard plot

(Figure 4-3b). The features of the model that produce concave-up Scatchard plots are: [a] high

affinity preformed dimers and [b] negative cooperativity of EGF binding to the dimer. Therefore,

in this model, the first EGF binds an unoccupied preformed dimer with high affinity, and the

second EGF binds with low affinity to the singly occupied dimer. There exists experimental

evidence that preformed dimers possess a higher affinity for EGF; dimers stabilized by a

juxtamembrane disulfide bond on the surface of living cells show an increased percentage of

high affinity receptors (12), and dimers observed by FRET also appear to be high affinity (4).

While there is mathematical support for negative cooperativity in EGF binding (14), there is also

experimental support in that disulfide-stabilized EGFR dimers still display both high and low

affinity binding to EGF (9,12). One possible structural interpretation of these model features is

that the preformed dimer holds EGFR domains I and III in an ideal orientation for ligand

binding, with correct positioning of the domains while allowing enough flexibility for the ligand

to enter the binding site, thus leading to high affinity binding. However, once the first EGF binds

to the preformed dimer, this could potentially change the positions of the two EGFR molecules with respect to each other, restricting the motion of the unoccupied EGFR site and making it

137 more difficult for the second EGF to bind. It has been shown that ligand binding stabilizes an

additional EGFR dimerization interface C-terminal to the DII dimerization arm that is not

present in the absence of ligand (20), and this interface could mediate such a change.

Although wild-type EGFR molecules form dimers on the surface of cells in the absence

of ligand, receptors lacking the cytoplasmic domain through partial truncation or substitution

with the EpoR cytoplasmic domain do not form these predimers (7). Mutants lacking part of the

cytoplasmic domain also give rise to linear Scatchard plots of a single low affinity binding site

(26), but can still dimerize in the presence of ligand (27). Taken together, these data are

consistent with our model of preformed receptor dimers as the high affinity component and

suggest that the cytoplasmic domain contributes to stabilizing the EGFR dimer, especially in the

absence of ligand. Recent modeling and experiments on EGFR with a cytoplasmic domain truncation suggest that the low affinity receptors represent interconverting tethered and extended receptor conformations (28), which is also the case in our model. Likewise, deletion of the

domain II dimerization arm (29) or mutations in this region (30) result in a loss of high affinity

binding. A V583D mutation designed to promote receptor untethering increases the percentage

of high affinity sites (30), which is consistent with our model. However, the same study found

that the mutation did not increase the amount of dimers in the absence of ligand, although the

sensitivity of the immunoblotting assay may have been limited. Further interpretation of

mutation results is difficult because many mutations designed to inhibit receptor tethering may also inhibit dimerization with opposing effects. In addition, fibroblasts in which α1,6- fucosyltransferase has been deleted (leading to a loss of protein core fucosylation) show no high affinity EGF binding. These cells also have substantially lower EGF-induced EGFR phosphorylation (31), suggesting a possible defect in receptor dimerization, and this data is

138 consistent with the hypothesis that receptor dimers are the high affinity component. Finally, an

N579Q mutant designed to eliminate glycosylation at this site leads to both an elevated level of

preformed dimers in the absence of ligand and a larger fraction of receptors in the high affinity

state (32).

Our model is similar to that of Wofsy et al., which also incorporates high affinity

preformed dimers with negative cooperativity (14). However, in that model both binding events

to the dimer were still higher affinity than binding to the monomer, and the best fit parameters

led to less than 1% of EGFR in preformed dimers in the absence of ligand. Although able to

produce concave-up Scatchard plots, the authors were unable to successfully fit this model to

their data, even though this model incorporated two essential elements of our model. The

experiments in Wofsy et al. used membrane vesicles with a C-terminal truncation of EGFR

starting at residue 1023 in their experiments, which may have affected the results. A separate

study showed that membrane vesicles only contained low affinity EGFR (33), demonstrating

variability in membrane vesicle composition. The parameters in the Wofsy et al. model could not

be uniquely determined, so certain parameters were fixed while others were fit. It is possible that

some relatively recent experimentally determined constants not available to the authors at the

time could have aided in the convergence of their model in a more restricted parameter space.

Wofsy et al. thus concluded that other factors such as heterogeneity in receptor density could be

responsible for the concavity of the Scatchard plot, and successfully modeled this hypothesis

(14). Likewise, Mayawala et al. also recently modeled heterogeneous EGFR density with success

(15). While receptor heterogeneity likely exists on the cell surface (34), it is unclear how the

binding of an antibody to either EGFR domain II or IV could lead to a loss in this heterogeneity

to produce a linear Scatchard plot as observed in previous experiments (18). Thus, heterogeneity

139 may not be the cause, or at least not the sole cause, of observed Scatchard concavity. Finally, our

model is not inconsistent with models which include a hypothetical third binding partner that leads to high affinity EGFR sites (16,17). Klein et al. modeled that only receptor dimers could

bind to the hypothetical partner, and thus our observation that putatively dimerization-blocking

antibodies result in loss of high affinity binding could be consistent with this model (17).

However, we believe that the utility of our model is in its ability to reproduce experimental data by incorporating known EGFR molecular events and parameters based on experiments, without the need to introduce any additional mechanisms not yet independently verified.

A ligand-binding and dimerization model for soluble EGFR ectodomain has also been developed in the past (23). It is difficult to compare receptor dimerization constants measured in

solution with those on the surface of cells, since receptors restricted to the membrane lose

rotational and translational degrees of freedom (35) and the cytoplasmic domain may also

contribute to dimerization (see above). Since we have modeled equilibrium binding only, kinetic

parameters and predictions are beyond the scope of the model. However, a spatially distributed

Monte Carlo-based simulation that utilized kinetic parameters determined that at low EGF

concentrations, the predominant ligand-binding mode is binding to preformed dimers (36), which is consistent with the model presented here. An additional kinetic model based on experimental imaging data found that preformed EGFR dimers bound EGF faster than monomeric sites, but their mathematical model suggested positive cooperitivity in EGF binding (37). Also beyond the

scope of the model are higher-order EGFR and HER2 oligomers, which experiments on the

surface of cells suggest may exist (38).

EGFR antibodies that lead to altered Scatchard plots have previously been described.

Monoclonal antibody 108 was shown to reduce the number of high affinity EGFR receptors, as

140 the pAbs did; however, unlike the pAbs, 108 both partially inhibited EGF binding and was

unable to block signaling at saturating EGF concentrations (39). Conversely, another monoclonal

antibody, 2E9, appeared to block binding to only low affinity EGFR; however, 2E9 competed

with EGF binding almost completely at saturating concentrations of EGF, and EGFR was still

able to signal when cells were treated with a 90% saturating amount of 2E9 (40). This is not

surprising, considering that a calcium signaling response has been demonstrated for a few as 300

molecules of bound EGF (13). While these antibodies have been reported to alter Scatchard

plots, they still appear to at least partially compete with EGF binding at saturating concentrations; thus, their mechanisms of action are unclear.

Physiologically, the formation of high affinity preformed dimers in the absence of ligand

could aid the cell in responding quickly to a ligand stimulus. These dimers on the cell surface

would be poised to signal once ligand is present, without needing to wait for two ligand-bound

monomers to diffuse together. We hypothesize that the function of the autoinhibited EGFR

conformation is to limit the number of extended monomers that can form dimers in the absence

of ligand, thus preventing a high level of basal signaling.

141 4.4 Works cited

1. Lax, I., Bellot, F., Howk, R., Ullrich, A., Givol, D., et al. (1989) Functional analysis of the ligand binding site of EGF-receptor utilizing chimeric chicken/human receptor molecules, EMBO J. 8, 421-427. 2. Scatchard, G. (1949) The attractions of proteins for small molecules and ions, Ann. NY Acad. Sci. 51, 660-672. 3. Schlessinger, J. (1988) The epidermal growth factor receptor as a multifunctional allosteric protein, Biochemistry 27, 3119-3123. 4. Gadella, T. W., Jr., and Jovin, T. M. (1995) Oligomerization of epidermal growth factor receptors on A431 cells studied by time-resolved fluorescence imaging microscopy. A stereochemical model for tyrosine kinase receptor activation, J. Cell. Biol. 129, 1543- 1558. 5. Sako, Y., Minoghchi, S., and Yanagida, T. (2000) Single-molecule imaging of EGFR signalling on the surface of living cells, Nat. Cell Biol. 2, 168-172. 6. Martin-Fernandez, M., Clarke, D. T., Tobin, M. J., Jones, S. V., and Jones, G. R. (2002) Preformed oligomeric epidermal growth factor receptors undergo an ectodomain structure change during signaling, Biophys. J. 82, 2415-2427. 7. Yu, X., Sharma, K. D., Takahashi, T., Iwamoto, R., and Mekada, E. (2002) Ligand- independent dimer formation of epidermal growth factor receptor (EGFR) is a step separable from ligand-induced EGFR signaling, Mol. Biol. Cell 13, 2547-2557. 8. Zhu, H. J., Iaria, J., Orchard, S., Walker, F., and Burgess, A. W. (2003) Epidermal growth factor receptor: association of extracellular domain negatively regulates intracellular kinase activation in the absence of ligand, Growth Factors 21, 15-30. 9. Moriki, T., Maruyama, H., and Maruyama, I. N. (2001) Activation of preformed EGF receptor dimers by ligand-induced rotation of the transmembrane domain, J. Mol. Biol. 311, 1011-1026. 10. Yarden, Y., and Schlessinger, J. (1987) Self-phosphorylation of epidermal growth factor receptor: evidence for a model of intermolecular allosteric activation, Biochemistry 26, 1434-1442. 11. Boni-Schnetzler, M., and Pilch, P. F. (1987) Mechanism of epidermal growth factor receptor autophosphorylation and high-affinity binding, Proc. Natl. Acad. Sci. USA 84, 7832-7836. 12. Sorokin, A., Lemmon, M. A., Ullrich, A., and Schlessinger, J. (1994) Stabilization of an active dimeric form of the epidermal growth factor receptor by introduction of an inter- receptor disulfide bond, J. Biol. Chem. 269, 9752-9759. 13. Uyemura, T., Takagi, H., Yanagida, T., and Sako, Y. (2005) Single-molecule analysis of epidermal growth factor signaling that leads to ultrasensitive calcium response, Biophys. J. 88, 3720-3730. 14. Wofsy, C., Goldstein, B., Lund, K., and Wiley, H. S. (1992) Implications of epidermal growth factor (EGF) induced egf receptor aggregation, Biophys. J. 63, 98-110. 15. Mayawala, K., Vlachos, D. G., and Edwards, J. S. (2005) Heterogeneities in EGF receptor density at the cell surface can lead to concave up scatchard plot of EGF binding, FEBS Lett. 579, 3043-3047.

142 16. Holbrook, M. R., Slakey, L. L., and Gross, D. J. (2000) Thermodynamic mixing of molecular states of the epidermal growth factor receptor modulates macroscopic ligand binding affinity, Biochem. J. 352 Pt 1, 99-108. 17. Klein, P., Mattoon, D., Lemmon, M. A., and Schlessinger, J. (2004) A structure-based model for ligand binding and dimerization of EGF receptors, Proc. Natl. Acad. Sci. USA 101, 929-934. 18. Wolf Yadlin, A. (2006) Development of mass spectrometry based technologies for quantitative cell signaling phosphoproteomics: the epidermal growth factor receptor family as a model system, Ph.D. Thesis, Massachusetts Institute of Technology, MA. 19. Lakowicz, J. R. (1999) Principles of fluorescence spectroscopy, 2nd ed., pp. 121-128, Kluwer Academic/Plenum, New York. 20. Dawson, J. P., Berger, M. B., Lin, C. C., Schlessinger, J., Lemmon, M. A., et al. (2005) Epidermal growth factor receptor dimerization and activation require ligand-induced conformational changes in the dimer interface, Mol. Cell. Biol. 25, 7734-7742. 21. Motulsky, H., and Christopoulos, A. (2004) Fitting models to biological data using linear and nonlinear regression : a practical guide to curve fitting, Oxford University Press, New York. 22. Burgess, A. W., Cho, H. S., Eigenbrot, C., Ferguson, K. M., Garrett, T. P., et al. (2003) An open-and-shut case? Recent insights into the activation of EGF/ErbB receptors, Mol. Cell 12, 541-552. 23. Lemmon, M. A., Bu, Z., Ladbury, J. E., Zhou, M., Pinchasi, D., et al. (1997) Two EGF molecules contribute additively to stabilization of the EGFR dimer, EMBO J. 16, 281- 294. 24. Hendriks, B. S., Opresko, L. K., Wiley, H. S., and Lauffenburger, D. (2003) Quantitative analysis of HER2-mediated effects on HER2 and epidermal growth factor receptor endocytosis: distribution of homo- and heterodimers depends on relative HER2 levels, J. Biol. Chem. 278, 23343-23351. 25. Wilkinson, J. C., and Staros, J. V. (2002) Effect of ErbB2 coexpression on the kinetic interactions of epidermal growth factor with its receptor in intact cells, Biochemistry 41, 8-14. 26. Livneh, E., Prywes, R., Kashles, O., Reiss, N., Sasson, I., et al. (1986) Reconstitution of human epidermal growth factor receptors and its deletion mutants in cultured hamster cells, J. Biol. Chem. 261, 12490-12497. 27. Kashles, O., Yarden, Y., Fischer, R., Ullrich, A., and Schlessinger, J. (1991) A dominant negative mutation suppresses the function of normal epidermal growth factor receptors by heterodimerization, Mol. Cell. Biol. 11, 1454-1463. 28. Ozcan, F., Klein, P., Lemmon, M. A., Lax, I., and Schlessinger, J. (2006) On the nature of low- and high-affinity EGF receptors on living cells, Proc. Natl. Acad. Sci. USA 103, 5735-5740. 29. Garrett, T. P., McKern, N. M., Lou, M., Elleman, T. C., Adams, T. E., et al. (2002) Crystal structure of a truncated epidermal growth factor receptor extracellular domain bound to transforming growth factor alpha, Cell 110, 763-773. 30. Walker, F., Orchard, S. G., Jorissen, R. N., Hall, N. E., Zhang, H. H., et al. (2004) CR1/CR2 interactions modulate the functions of the cell surface epidermal growth factor receptor, J. Biol. Chem. 279, 22387-22398.

143 31. Wang, X., Gu, J., Ihara, H., Miyoshi, E., Honke, K., et al. (2006) Core fucosylation regulates epidermal growth factor receptor-mediated intracellular signaling, J Biol Chem 281, 2572-2577. 32. Whitson, K. B., Whitson, S. R., Red-Brewer, M. L., McCoy, A. J., Vitali, A. A., et al. (2005) Functional effects of glycosylation at Asn-579 of the epidermal growth factor receptor, Biochemistry 44, 14920-14931. 33. Berkers, J. A., van Bergen en Henegouwen, P. M., Verkleij, A. J., and Boonstra, J. (1990) Membrane vesicles of A431 cells contain one class of epidermal growth factor binding sites, Biochim. Biophys. Acta 1052, 453-460. 34. Le Roy, C., and Wrana, J. L. (2005) Clathrin- and non-clathrin-mediated endocytic regulation of cell signalling, Nat. Rev. Mol. Cell Biol. 6, 112-126. 35. Qi, S., Krogsgaard, M., Davis, M. M., and Chakraborty, A. K. (2006) Molecular flexibility can influence the stimulatory ability of receptor-ligand interactions at cell-cell junctions, Proc. Natl. Acad. Sci. USA 103, 4416-4421. 36. Mayawala, K., Vlachos, D. G., and Edwards, J. S. (2005) Computational modeling reveals molecular details of epidermal growth factor binding, BMC Cell Biol. 6, 41. 37. Teramura, Y., Ichinose, J., Takagi, H., Nishida, K., Yanagida, T., et al. (2006) Single- molecule analysis of epidermal growth factor binding on the surface of living cells, Embo J 25, 4215-4222. 38. Clayton, A. H., Walker, F., Orchard, S. G., Henderson, C., Fuchs, D., et al. (2005) Ligand-induced dimer-tetramer transition during the activation of the cell surface epidermal growth factor receptor-A multidimensional microscopy analysis, J. Biol. Chem. 280, 30392-30399. 39. Bellot, F., Moolenaar, W., Kris, R., Mirakhur, B., Verlaan, I., et al. (1990) High-affinity epidermal growth factor binding is specifically reduced by a monoclonal antibody, and appears necessary for early responses, J. Cell Biol. 110, 491-502. 40. Defize, L. H., Boonstra, J., Meisenhelder, J., Kruijer, W., Tertoolen, L. G., et al. (1989) Signal transduction by epidermal growth factor occurs through the subclass of high affinity receptors, J. Cell Biol. 109, 2495-2507.

144 Appendix 1: Sequences

1.1 EGFR ectodomain amino acid sequence (1-621)

LEEKKVCQGTSNKLTQLGTFEDHFLSLQRMFNNCEVVLGNLEITYVQRNYDLSFLKTIQEVAGYVLIALNTVERIPLENLQIIRGNMYYENSYALAV LSNYDANKTGLKELPMRNLQEILHGAVRFSNNPALCNVESIQWRDIVSSDFLSNMSMDFQNHLGSCQKCDPSCPNGSCWGAGEENCQKLTKIICAQQ CSGRCRGKSPSDCCHNQCAAGCTGPRESDCLVCRKFRDEATCKDTCPPLMLYNPTTYQMDVNPEGKYSFGATCVKKCPRNYVVTDHGSCVRACGADS YEMEEDGVRKCKKCEGPCRKVCNGIGIGEFKDSLSINATNIKHFKNCTSISGDLHILPVAFRGDSFTHTPPLDPQELDILKTVKEITGFLLIQAWPE NRTDLHAFENLEIIRGRTKQHGQFSLAVVSLNITSLGLRSLKEISDGDVIISGNKNLCYANTINWKKLFGTSGQKTKIISNRGENSCKATGQVCHAL CSPEGCWGPEPRDCVSCRNVSRGRECVDKCNLLEGEPREFVENSECIQCHPECLPQAMNITCTGRGPDNCIQCAHYIDGPHCVKTCPAGVMGENNTL VWKYADAGHVCHLCHPNCTYGCTGPGLEGCPTNGPKIPS

1.2 EGFR 273-621 DNA sequence

CGTAATTATGTGGTGACAGATCACGGCTCGGCCGTCCGAGCCTGTGGGGCCGACAGCTATGAGATGGAGGAAGACGGCGTCCGCAAGTGTAAGAAGT GCGAAGGGCCTTGCCGCAAAGTGTGTAACGGAATAGGTATTGGTGAATTTAAAGACTCACTCTCCATAAATGCTACGAATATTAAACACTTCAAAAA CTGCACCTCCATCAGTGGCGATCTCCACATCCTGCCGGTGGCATTTAGGGGTGACTCCTTCACACATACTCCTCCTCTGGACCCACAGGAACTGGAT ATTCTGAAAACCGTAAAGGAAATCACAGGGTTTTTGCTGATTCAGGCTTGGCCTGAAAACAGGACGGACCTCCATGCCTTTGAGAACCTAGAAATCA TACGCGGCAGGACCAAGCAACATGGTCAGTTTTCTCTTGCAGTCGTCAGCCTGAACATAACATCCTTGGGATTACGCTCCCTCAAGGAGATAAGTGA TGGAGATGTGATAATTTCAGGAAACAAAAATTTGTGCTATGCAAATACAATAAACTGGAAAAAACTGTTTGGGACCTCCGGTCAGAAAACCAAAATT ATAAGCAACAGAGGTGAAAACAGCTGCAAGGCCACAGGCCAGGTCTGCCATGCCTTGTGCTCCCCCGAGGGCTGCTGGGGCCCGGAGCCCAGGGACT GCGTCTCTTGCCGGAATGTCAGCCGAGGCAGGGAATGCGTGGACAAGTGCAACCTTCTGGAGGGTGAGCCAAGGGAGTTTGTGGAGAACTCTGAGTG CATACAGTGCCACCCAGAGTGCCTGCCTCAGGCCATGAACATCACCTGCACAGGACGGGGACCAGACAACTGTATCCAGTGTGCCCACTACATTGAC GGCCCCCACTGCGTCAAGACCTGCCCGGCAGGAGTCATGGGAGAAAACAACACCCTGGTCTGGAAGTACGCAGACGCCGGCCATGTGTGCCACCTGT GCCATCCAAACTGCACCTACGGATGCACTGGGCCAGGTCTTGAAGGCTGTCCAACGAATGGGCCTAAGATCCCGTCC

Note: WT cysteine 283 mutated to alanine (truncation at 273 causes unpaired cysteine) N-terminal c-myc tag included in above sequence

1.3 404SG DNA sequence

CTGGAGGAAAAGAAAGTTTGCCAAGGCACGAGTAACAAGCTCACGCAGTTGGGCACTTTTGAAGATCATTTTCTCAGCCTCCAGAGGATGTTCAATA ACTGTGAGGTGGTCCTTGGGAATTTGGAAATTACCTATGTGCAGAGGAATTATGATCTTTCCTTCTTAAAGACCATCCAGGAGGTGACTGGTTATGT CCTCATTGCCCACAACACAGTGGAGCGAATTCCTTTGGAAAACCTGCAGATCATCAGAGGAAATATGTACTACGAAAATTCCTATGCCTTAGCAGTC TTATCTAACTATGATGCAAATAAAACCGGACTGAAGGAGCTGCCCATGAGAAATTTACAGGAAATCCTGCATGGCGCCGTGCGGTTCAGCAACAACC CTGCCCTGTGCAACGTGGAGAGCATCCAGTGGCGGGACATAGTCAGCAGTGACTTTCTCAGCAACATGTCGATGGACTTCCAGAACCACCTGGGCAG CTGCCAAAAGTGTGATCCAAGCTGTCCCAATGGGAGCTGCTGGGGTGCAGGAGAGGAGAACTGCCAGAAACTGACCAAAATCATCTGTGCCCAGCAG TGCTCCGGGCGCTGCCGTGGCAAGTCCCCCAGTGACTGCTGCCACAACCAGTGTGCTGCAGGCTGCACAGGCCCCCGGGAGAGCGACTGCCTGGTCT GCCGCAAATTCCGAGACGAAGCCACGTGCAAGGACACCTGCCCCCCACTCATGCTCTACAACCCCACCACGTACCAGATGGATGTGAACCCCGAGGG CAAATACAGCTTTGGTGCCACCTGCGTGAAGAAGTGTCCCCGTAATTATGTGGTGACAGATCACGGCTCGTGCGTCCGAGCCTGTGGGGCCGACAGC TATGAGATGGAGGAAGACGGCGTCCGCAAGTGTAAGAAGTGCGAAGGGCCTTGCCGCAAAGTGTGTAACGGAATAGGTATTGGTGAATTTAAAGACT CACTCTCCATAAATGCTACGAATATTAAACACTTCAAAAACTGCACCTCCATCAGTGGCGATCTCCACATCCTGCCGGTGGCATTTAGGGGTGACTC CTTCACACATACTCCTCCTCTGGACCCACAGGAACTGGATATTCTGAAAACCGTAAAGGAAATCACAGGGTCTTTGCTGATTCAGGCTTGGCCTGAA AACAGGACGGACCTCCATGCCTTTGAGAACCTAGAAATCATACGCGGCAGGACCAAGCAACATGGTCAGTTTTCTCTTGCAGTCGTCGGCCTGAACA TAACATCCTTGGGATTACGCTCCCTCAAGGAGATAAGTGATGGAGATGTGATAATTTCAGGAAACAAAAATTTGTGCTATGCAAATACAATAAACTG GAAAAAACTGTTTGGGACCTCCGGTCAGAAAACCAAAATTATAAGCAACAGAGGTGAAAACAGCTGCAAGGCCACAGGCCAGGTCTGCCATGCCTTG TGCTCCCCCGAGGGCTGCTGGGGCCCGGAGCCCAGGGACTGCGTCTCTTGCCGGAATGTCAGCCGAGGCAGGGAATGCGTGGACAAGTGCAACCTTC TGGAGGGTGAGCCAAGGGAGTTTGTGGAGAACTCTGAGTGCATACAGTGCCACCCAGAGTGCCTGCCTCAGGCCATGAACATCACCTGCACAGGACG GGGACCAGACAACTGTATCCAGTGTGCCCACTACATTGACGGCCCCCACTGCGTCAAGACCTGCCCGGCAGGAGTCATGGGAGAAAACAACACCCTG GTCTGGAAGTACGCAGACGCCGGCCATGTGTGCCACCTGTGCCATCCAAACTGCACCTACGGATGCACTGGGCCAGGTCTTGAAGGCTGTCCAACGA ATGGGCCTAAGATCCCGTCC

Note: Contains four mutations from wild-type EGFR to allow expression in yeast: A62T, L69H, F380S, S418G

145 1.4 VL sequences of selected clones from LYN-A4 and LYN-A5

VL-FR1 VL-CDR1 VL-FR2 A4-8 NFMLTQPHSVSESPGKTATISC TRSSGSIASNYVQ WYQQRPGGAPTTVIY A5-2 ETTLTQSPAFLSATPGDKVDISC KASQDIDDDLN WYQQKPGEAAVFIIQ A5-3 ETTFTQSPAFMSATPGDKVNIFC KASQDIDDDVS WYQQRPGEAPIFIIQ A5-4 ETTLTQSPAFMSATPGDKVNISC KASQDIDDDMN WYQQKPGEAAIFIIQ A5-5 QPVLTQSPSASGTPGQSVTISC SGSGSNIGNNKVN WYQQLPGTAPKLLIY A5-6 NFMLTQPHSVSESPGETVTISC TGSSGSIASNSVQ WYQQRPGSAPRTVIY A5-8 NFMLTQPHSVSESPGKTVTISC TGSSGSIATNYVQ WYQQRPGSAPTTVIY A5-9 ETTFTQSPAFMSATPGDKVNISC KASQDIDEDVN WYQQKPGEAPIFIIQ

VL-CDR2 VL-FR3 VL-CDR3 A4-8 EDNQRPS GVPDRFSGSIDSSSNSASLTISGLRTEDEADYYC QSYDRSNKAV FGGGTQLTILSGIL A5-2 EATTLVP GIPPRFSGSGYGTDFTLTINNIESEDAAYYFC LHHDDLPLT FGQGTKLEIKSGIL A5-3 EATTLVP GTPPRFSGSGYGTDFTLTIDNIESEDAAYYFC LQDDSFPLT FGQGTKLEIKSGIL A5-4 EATTLVP GIPPRFSGSGYGTDFTLTINNIESEDAAYYFC LQHDNFPWT FGQGTRVEIKSGIL A5-5 SNNQRPS GVPDRFSGSKSGTSASLAISGLQSEDEADYYC AAWDDSLNGYV FGTGTKLTVLSGIL A5-6 EDHQRPS GVPDRFSGSIDTSSNSASLTISGLRTEDEGDYYC RSYDSSHTV FGGGTQLTVLSGIL A5-8 ENNQRPS GVPDQFSGSIDSSSNSASLTISGLKTEDKADYYC QSYDSSXLXV XXXGTKXXVLXXIX A5-9 DATTLVP GIPPRFSGSGHGTDFTLTIDNIESEDAAYYFC LQHDNFPPYT FGQGTKLEIKSGIL

Note: A4-8, A5-2, A5-3, A5-4, A5-6, A5-8, A5-9: LYN-1 VH A5-5: LYN-2 VH

1.5 VL sequences of selected clones from CRN-A3

VL-FR1 VL-CDR1 VL-FR2 1 NFMLTQPHSVSESPGKTVTISC TGSSGSIATNYVQ WYQQRPGSAPTTVIY 2 EIVLTQSPGTLSLSPGERVTLSC RASQTVSSSYLA WYQQKTGQAPRLLMY 3 DIQVTQSPSSLSASVGDRVTITC RASQSISSYLN WYQQKPGTAPKLLIY 4 ETTLTQSPGTLSLSPGERATLSC RASQSVGGSYLA WYQKKTGQAPRLLIY 5 ETTLTQSPGTLSLSPGERATLSC RASQGVAGSDLA WYQQKPGQAPRLLIY 6 NFMLTQPRSVSESPGKTVTISC TGSSGSIASNYVQ WYQQRPGSAPTIVIH 7 ETTLTQSPAFMSATPGDKVNISC KASQDIDDDMN WYQQKPGEAPILILQ 8 EIVLTQSPATLSLSPGERATLSC RASQSFTGNYLA WYQQKPGQAPRLLIY 9 DIRVTQSPSSLSASVGDRVTITC RASQSINTYLN WYQQKPGKAPKLLIY 10 ETTLTQSPGTLSLSPGDRVTISC KASHDVDDDLN WYQQKPGKAPVLIIR

VL-CDR2 VL-FR3 VL-CDR3 1 ENNQRPS GVPDRFSGSIDSSSNSASLTISGLKTEDEADYYC QSYDSSNHAV FGGGTQLTVLSGIL 2 GASTRAT GIPDRFSGSGSGTDFTLTISRLEPEDFAVYYC QQYGSSRWT FGQGTKVEIKSGIL 3 DATRVES GVPSRFSGSGSGTDFTLTISGLQVEDVATYYC QRGYNTPRT FGQGTKVDIKSGIL 4 DASSRAT GIPDRFSGSGSGTDFTLTINRVEPEDFAMYYC QQYGSSPVT FGQGTRLKIKSGIL 5 GASSRAT GIPDRFSGSGSGTDFSLTISRLEPEDFAVYYC QQFDTVTWT FGQGTKVEIKSGIL 6 EDNQRPS GVPDRFSGSIDSSSNSASLTISGLKTEDEADYYC QSYDSSSRWV FGGGTKLTVLSGIL 7 ESTVLLP GIPPRFSGSGYGSDFTLTINNMQSEDAAYYFC LQHDSFPFT FGQGTKLEIKSGIL 8 GASSRAT GIPDRFSGSGSGTDFTLTISRLEPEDFAVYHC QQYGTSPT FGQ 9 AASSLQS GVPSRFSGSGSGTDFTLTISSLQPEDFATYYC QQSYSSPSL FTFGPGNKVDIKSG 10 EATTLVP GIPPRFSGSGYGTDFTLTINNMESEDAAYYFC LQHDSPPYS FGQGTKLEIRSGIL

146 1.6 Selected clones from CAH-A6

VH-FR1 VH-CDR1 VH-FR2 VH-CDR2 1 EVQLVESGGGLVQPGRSLRLSCAAS GFSFADYAMH WVRQVPGKGLEWVS SISWNSGIKG 2 QVQLKQSGPGLVNPSQTLSLTCAIS GDSVSSDSAA WNWIRQSPSRGLEWLG RTYYRSKWYNE 4 QVQLVESEGGVVQPGRSLRLSCAAS GFSFSRFAMH WVRQAPGKGLEWVA VISYDGSNKF 5 QVQLQQSGPGLVKPSQTLSLTCAIS GDSVSSNSAAWN WIRQSPSRGLEWLG RTAYRSKWNSD

VH-FR3 VH-CDR3 1 YADSVKGRFTISRDNAKNTVYLEVNNIRPEDTALYYCAK VRDPNIEAFDV WGQGTMVTVSS 2 YTESVKSRIIINPDTSKNQFSLQLNSVTPEDTALYYCAR FLRGTFDI WGQGSMVTVSS 4 YADSVKGRFTISRDNSKNTLYLQMDSLRAEDTAVYYCAR HYDSSGRDH WGQGTLVTVSS 5 YAASVRSRITINPDTSKNQFSLQLNSVTPDDTAMYYCAR SRSSSPIDY WGQGTLVTVSS

Linker VL-FR1 VL-CDR1 VL-FR2 1 GILGSGGGGSGGGGSGGGGS EIVLTQSPGTLSLSPGERATLSC RASQSVSSSYLG WYQQKPGQAPRLLIY 2 ASTKGPSGILGSGGGGSGGGGSGGGGS NFMLTQPHSVSESPGKTVTISC TGSSGSIATNYVQ WYQQRPGSAPTTVIY 4 GILGSGGGGSGGGGSGGGGS EIVMTQSPGTLSSSPGERATLSC RASKDVSSQFLA WYQQKPGQAPRLLIY 5 GILGSGGGGSGGGGSGGGGS EIVLTQSPGTLSLSPGERVTLSC RASQRVSSSYVA WYQQKPGQAPRLLIY

VL-CDR2 VL-FR3 VL-CDR3 1 GASSRAT GIPDRFSGSGSGTDFALTISRLEPEDFAVYYC QQHDRSSWT FGQGTKVEIKSGIL 2 ENNHRPS GVPDRFSGSIDSSSNSASLTISGLKTEDEAGYYC QSSDNDFHAV FGGGTQLTVLSGIL 4 ETSTRAT GIPDRFSGSGSGTDFTLTISRLEPEDFAVYYC QQYGRSPRT FGQGTKVEIKSGIL 5 GASSRAT GIPDRFSGSGSGTDFTLTISRLEPEDFAVYYC QQYDSSPRT FGQGTKLEIKSGIL

1.7 VL sequences of selected clones from MGE-A4

VL-FR1 VL-CDR1 VL-FR2 2 DIRVTQSPSSLSASVGDRVTITC RASQSISNYLN WYQQKPGKAPKLLIH 3 QSVLTQPASVSGSPGQSITISC TGTSSDVGGYDYVS WYQQHPGKTPKRMIY 4 QSVLTQPPSVSGSPGQSITISC TGTSSDVGNYNLVS WYQRHPGKAPKLMIY 7 NFMLTQPHSVSESPGKTVAISC TRSSGSIASNYVQ WYQQRPGSSPTTVIY

VL-CDR2 VL-FR3 VL-CDR3 2 AASSLQS GVPSRFSGSGSGTDFTLTISSLQPEDFATYYC QQSYSMPFN FGPGTK 3 EVNKRPS GVSNRFSGSKSGNTASLTISGLQAEDEADYYC TSYRSGSSYV FGTGTKLTVLSGIL 4 EVNKRPS GISHRFSGSKSGNTASLTISGLQAED 7 EDKERPS GVPDRFSGSIDSSSNSASLTISGLKTEDEADYYC QSYDGTNTV FGGGTQLTVLSGIL

Note: All clones MGE-2 VH

1.8 VL sequences of selected clones from LVW-A3

VL-FR1 VL-CDR1 VL-FR2 1 DIRVTQSPSSLSASIGDRVTITC RASQTIGSYLN WYQHRPGKAPKLLIY 2 NFMLTQPHSVSESPGKTVTISC TGSSGSIASNYVQ WYQQRPGSAPTTVIY 3 NFMLTQLHSVSGSPGKTITISC TRSSGSIATNYVQ WYQQRPGRAPSTVIY 4 ETTLTQSPAFLSATPGDKVTISC KASLDIDDDVN WYQQKPGQAPIFIIQ 5 NFMLTQPHSVSESPGKTVTISC TRSSGSIASNYVQ WYQQRPGRAPTTIIF 7 NFMLTQPHSVSESPGKTVTISC TGSSGSIADKYVQ WYQQRPGSAPRNVIF 8 NFMLTQPHSVSESPGKTVTISC TGSSGSITINYVQ WYQQRPGSAPTTVIY 9 NFMLTQPHSVSESPGKTVTISC TGSSGSIASNYVQ WYQQRPGSAPTTVIY

VL-CDR2 VL-FR3 VL-CDR3 1 AASSLQD GVPSRFSGGGSGTDFTLTISNLQPEDFTTYYC QQSYGXXRDV RXRDQGGXSNPGIL 2 EDNQRPS GVPDRFSGSIDSSSNSASLTISGLKTEDEADYYC QSYDNSNPAV FGGGTQLTVLSGIL 3 EDKQRPS GVPDRFSGSIDSSSNSASLTISGLKTEDEADYYC QSYDSSNYV FGTGTKLTVLSGIL 4 EATTLVP GIPPRFSGSGYGTDFTLTINNIQPEDAAYYFC LQHDNFPYT FGQGTKLEVKSGIL 5 EDNQRPS GVPDRFSGSIDSSSNSASLTISGLKTEDEADYYC QSYDSSNAV FGGGTQLTVLSGIL 7 DDDQRPS GVPDRFSGSIDTSSNSASLIISGLKTEDEADYYC QSFDSSNKTV FGRSTQLTVLSGIL 8 EDNQRPS GVPDRFSGSIDSSSNSASLTISGLKTEDEADYYC QSYDSSNQGV FGRRTKLTVLSGIL 9 ENNQRPS GVPDRFSGSIDSSSNSASLTISGLKTEDEAEYYC QSYDSSNHAV FGGGTQLTVLSGIL

Note: All clones LWV-2 VH

147 1.9 Selected clones from CRN-B4

VH-FR1 VH-CDR1 VH-FR2 VH-CDR2 1 3 QVQLQQWVPGLLKPSETLSLTCGVS GGSLSGYYWS WIRQSPGKELEWIG EINQGGSTN 5 QVQLQQWGPGLLEPSETLSLTCGVS GGSLSGYYWS WIRQSPGKELEWIG EINQGGSTN 6 QVQLQQWGPGLLKPSETLSLTCGVS GGSLSGYYWS WIRQSPGKELEWIG EINQGGSTN 7 QVQRQQWGPGPLKPSETLFLTCGVS GGSLSGYYWS WVRQSPGKELEWIG DINQGGSTN 10 QLQQWGPGLLKPSETLSLTCGVS GGSLSGYYWS WIRQSPGKELEWIG EINQGGSTN

VH-FR3 VH-CDR3 1 3 YNPSLRSRVTISVDTSKKQLSLKVNSVTASDTAVYYCAR ETFRGSNCFDS WGQGTLVTVSS 5 YNPSLRSRATISVDTSKKQLSLKVNSVTASDTAVYYCAR ETFRGSNCFDS WGQGTLVTVSS 6 YNPSLRSRVTISVDTSKKQLSLKVNSVTASDTAVYYCAR ETFRGSNCFDS WGQGTLVTVSS 7 YDPSLRSRVATSAGTSKKQLSLKVDPATASDTAAYYCAR EIFRGSNRFDS WGQGILVAASS 10 YNPSLRSRVTISVDTSKKQLSLKVNSVTASDTAVYYCAG ETFKGSNCFDS WGQGTLVTVSS

Linker VL-FR1 VL-CDR1 VL-FR2 1 EIVLTQSPGTLSLSPGERVTLSC RASQTVSSSYLA WYQQKTGQAPRLLMY 3 GILGSGGGGSGGGGSGGGGS EIVLTQSPGTLSLSPGERVTLSC RASQTVSSSYLA WYQQKTGQAPRLLMY 5 GILGSGGGGSGGGGSGGGGS EIVLTQSPGTLSLSPGERVTLSC RASQTVSSSYLA WYQQKTGQAPRLLMY 6 GILGSGGGGSGGGGSGGGGS EIVLTQSPGTLSLSPGERVTLSC RASQTVSSSYLA WYQQKTGQAPRLLMY 7 GILGPGGGGGGGGGSGGGGP EIVLTQSPGALSLSPGERVTLSC RASQIVGSSYLA WCQRETGQAPRLLTY 10 GILGSGGGGSGGGGSGGGGS EIVLTQSPGTLSLSPGERVTLSC RASQTVSSSYLA WYQQKTGQAPRLLMY

VL-CDR2 VL-FR3 VL-CDR3 1 GASTRAT GIPDRFSGSGSGTDFTLTISRLEPEDFAVYYC QQYGSSRWT FGQGTKVEVKSGIL 3 GASTRAT GIPDRFSGSGSGTDFTLTISRLEPEDFAVYYC QQYGSSWWT FGQGTKVEIKSGIL 5 GASTRAT GIPDRFSGSGSGTDFTLTISRLEPEDLAVYYC QQYGSSRWT FGQGTKVEIKSGIL 6 GASTRAT GIPDRFSGSGSGTDFTLTISRLEPEDFAVYYC QQYGSSRWT FGQGTKVEIKSGIL 7 GASTRAT GIPDRFSGSGSGTDFTLTISRLEPENLAVYHC QQYGSSRWT LGQGTKVETKPGIL 10 GASTRAT GIPDRFSGSGSGTDFTLTISRLEPEDFAVYYC QQYGSSRWT FGQGTKVEIKSGIL

1.10 CRN-B4-2

QVQLQQWGPGLLKPSETLSLTTVIYEDNQRPSAVPDRFSGSIDSSSNSASLTISGLKTEDEADYYCQSYHSSIVVFGGGTKLTVLSGIL

1.11 CRN-B4-4

GGGSGGGGSAGILEQKLISEEDL

1.12 CAH-B4-1

VH-FR1 VH-CDR1 VH-FR2 VH-CDR2 EVQLVESGGGLVQPGRSLRLSCAAS GFSFADYAMH WVRQVPGKGLEWVS SISWNSGIKG

VH-FR3 VH-CDR3 YADSVKGRFTISRDNAKNTVYLEVTNIRPEDTALYYCAK VRDPNIEAFDV WGQGTMVTVSS

Linker VL-FR1 VL-CDR1 VL-FR2 GILGPGGGGSGGGGSGGGGS EIVLTQSPGTLSLSPGERATLSC RASQSVSSSYLG WYQQKPGQAPRLLIY

VL-CDR2 VL-FR3 VL-CDR3 GASSRAT GIPDRFSGSGSGTDFALTISRLEPEDFATYYC QQANSFPYT FGQGTKLEIKSGIL

148 1.13 Selected clones from MGE-B6

VH-FR1 VH-CDR1 VH-FR2 VH-CDR2 1 QVQLQEPGPGLARPSETLSLTCTVS GGPIGDSDYCWG WIRQPPGKGLRWVG GINCYGTTC 2 QVQLQESGPGLVKPSETLSLTCTVS GGSISNSDYYWG WIRQPPGKGLRWIG SINYYGTTY 3 QVQLQESGPGLVKPSETLSLTCTVS GGSISNSGYYWG WIRQPPGKGLRWIG SINYYGTTY 4 QVQLQESGPGLVKPSETLSLTCTVS GGSISNSDYYWG WIRQPPGKGLRWIG SINYYGTTY 5 QVQLQESGPGLVKPSETLSLTCTVS GGSISNSDYYWG WIRQPPGKGLRWIG SINYYGTTY 6 QVQLQESGPGLVKPSETLSLTCTVS GGSISNSDYYWG WIRQPPGKGLRWIG SINYYGTTY 7 QVQLQESGPGLVKPSETLSLTCTVS GGSISNSDYYWG WIRQPPGKGLRWIG SINYYGTTY 8 QVQLQESGPGLVKPSEALSLTCAVS GGSISNSDYYWG WVRQPPGKGLRWIG SINYYGTTY 9 QVQLQESGPGLVKPSETLSLTCTVS GGSISNSDYYWG WIRQPPGKGLRWIG SINYYGTTY 10 QVQLQESGPGLVKPSETLSLTCTVS GGSISNSDYYWG WIRQPPGKGLRWIG SINYYGTTY

VH-FR3 VH-CDR3 1 YNPSLKNRVAMSVGTSKDQFSLRLGPVAATDTAVYYCAR RGANSWFFDL WGRGTLVTVSS 2 YNPSLKSRVAMSVDTSKNQFSLKLSSVTAADTAVYYCAR RGANSWFFDL WGRGTLVTVSS 3 YNPSLKSRVAMSVDTSKNQFSLKLSSVTAADTAVYYCAR RGANSWFFDL WGRGTLVTVSS 4 YNPSLKSRVAMSVDTSKNQFSLKLSSVTAADTAVYYCAR RGANSWSFDL WGRGTLVTVSS 5 YNPSLKSRAAMSVDTSKNQFSLKLSSVTAADTAVYYCAR RGANSWFFDL WGRGTLVTVSS 6 YNPSLKSRVAMSVDTSKNQFSLKLSSVTAADTAVYYCAR RGANSWFFDL WGRGTLVTVSS 7 YNPSLKSRVAMSVDTSKNQFSLKLSSVTAADTAVYYCAR RGANSWFFDL WGRGTLVTVSS 8 YNPSLKSRVAMSVDTSKNQFSLKLSSVTAADTAVYYCAR RGANSWFFDL WGRGTLVTVSS 9 YNPSLKSRVAMSVDTSKNQFSLKLSSVTAADTAVYYCAR RGANSWFFDL WGRGTLVTVSS 10 YNPSLKSRVAMSVDTSKNQFSLKLSSVTAADTAVYYCAR RGANSWFFDL WGRGTLVTVSS

Linker VL-FR1 VL-CDR1 VL-FR2 1 GILGSGGGGSGGGGSGGGGS NFMLTQPHSVSESPGKMAAISC TRSSGSIASNYVQ WYQQRPGSSPTTVIY 2 GILGSGGGGGGGGGSGGGGS QSVLTQPPSVSGSPGQSITISC TGTSSDVGGYDYVS WYQQHPGKTPKRMIY 3 GILGSGGGGSGGGGSGGGGS QSVLTQPASVSGSPGQSITISC TGTSSDVGNYNLVS WYQRHPGKAPKLMIY 4 GILGSGGGGSGGGGSGGGGS QSVLTQPASVSGSPGQSITISC TGTGSDVGGYDYVS WYQQHPGKTPKRMIY 5 GILGSGGGGSGGGGSGGGGS DIRVTQSPSSLSASVGDRVTITC RASQSISSSLN WYQQKPGKAPNLLIY 6 GILGSGGGGSGGGGSGGGGS QSVLTQPASVSGSPGQSITISC TGTSSDVGGYDYVS WYQQHPGKTPKRMIY 7 GILGSGGGGSGGGGPGGGGS QSVLTQPASVSGSPGQSITISC TGTSSDVGGYDYVS WYQQHPGKTPKRMIY 8 GILGSGGGGSGGGGSGGGGP QSVLTQPPSVSGSPGQSITISC TGTSSDVGNYNLVS WYQRHPGKAPKLMIY 9 GILGSGGGGSGGGGSGGGGS DIQVTRSPSSPSASVGGRVTITC RASPSTSSSLS WYQQEPGKASDLLTY 10 GILGSGGGGSGGGGSGGGGS QSVLTQPASVSGSPGQSITISC TGTSSDVGGYDYVS WYQQHPGKTPKRMIY

VL-CDR2 VL-FR3 VL-CDR3 1 EDKERPS GVPDRLSGSTDSSSNSASLTISGLKPEDGADYYC QPYDRNNRAV FGGGTQLTVLSGIL 2 EVNKRPS GVSNRFSGSKSGNTASLTISGLQAEDEADYYC TSYRSGSSYV FGTGTKLTVLSGIL 3 EVNKRPS GISHRFSGSKSGNTASLTISGLQAKDEGDYYC CSYTSTGGTV LGGGTQLTVLSGIL 4 EVNKRPS GASNRFSGSKSGNTASLTISGLQAEDEADYYC TSYRSGSSYV FGTGTKLTVLPGIL 5 ATSSLQR GVPSRFSGSGSGTDFTLTISSLQPEDFATYYC QQSYNTPRT FGQGTKVEIKSGIL 6 EVNKRPS GVSNRFSGSKSGNTASLTISGLQAEDEGDYYC CSYTSTGGTV LGGGTQLTVLSGTL 7 EVNKRPS GVSNRFSGSKSGNTASLTISGLQAEDEADYYC TSYRSGSSYV FGTGTKLTVLSGIL 8 EVNKRPS GISHRFSGSKSGNTASLTISGLQAEDEGDYYC CSYTSTGGTV LGGGTQLTVLSGIL 9 ATPSLQR GVPSGSSGSGSGADSTLTASSLQPEDFATCHC QQSYSTPRT SGQGARMETKSGIL 10 EVNKRPS GVSNRFSGSKSGNTASLTISGLQAEDEADYYC TSYRSGSSYV FGTGTKLTVLSGIL

149 1.14 Selected clones from LVW-B4

VH-FR1 VH-CDR1 VH-FR2 VH-CDR2 1 QTTLEEPGPTVVKPAQTLTPTCTFP GSSLSTSGVGVG WIRQSPGKALEWLA PISWDGDKR 2 QTVLKEPSPTAARPTQALTLTCASS GLSLSTSGVGTD RAHQPPGRALEWPA LISRDDGKR 3 5 8 QIILKESGPTVVRPTQTLTLTCTSF GLSPSTSGVGVG WARQPPGRALEWLA LISWGDGKR 9 QTTLKESGSTTAKPIQTPTLACTFS GFSLSTSGVGAG WIRRLPGKAPEWPA LISWGDDKR

VH-FR3 VH-CDR3 1 YSPSLRSRPTITEDTSKNQVVLTMTNMDPVDTATYFCAH LTSFVDVIELRTGWCFDA WGRGTLVTVSS 2 HSPSPKGRLTITKGTSKNQAVLAVASMDPVGTAACPWAH PVSSVGAVELRTGWCSDT WGRGTLVTVPS 3 5 8 YSPSLKSRLPITKDTSKDQVALTATNMDPVDTATYFCAY PINFADAVELRTGWRFDV WGRGTLVTASS 9 YSPSLKSRPTITKDTSKSQVILTMTNMDPVGTATYFCTH LINFVDATERRTGWRSDV WGRGTLVTASS

Linker VL-FR1 VL-CDR1 VL-FR2 1 GILGSGGGGSGGDGSGGGGS NSTPTQPHSVSESPGKTVTISR TRSSGSIASDYVQ WPQQRPGSVPTTVTH 2 GTLGSGGGGSGGGSSGGGGP DFMLTQPHPVSESPRRTVTVSC TRNSDGIAGGRAQ RHQQRPGSSPTTTTR 3 DIRVTQSPSSLSASVGDRVTITC RPSRSISNYLN WYQQKPGKAPKLLIY 5 ETAFTQSPAFMSATPGDKVNIFC KASQDIDDDVS WYQQRPGEAPIFIIQ 8 GTLGSGGGGSGGGGPGGGGP NFMLTQPRAVSESPGKMATASC TGSSGSIASDPVQ WYQQRPGSAPTTVTH 9 GILGSGGGGSGGGGSGGGGS NFVLTQPHSVSESPGKAVTISY ARSGGSVATNYVR WCQQRPGGSPTTVVY

VL-CDR2 VL-FR3 VL-CDR3 1 ENNQRPS GVPDRFFGPIDSSSNSASPTISGLKTEDEADHYC QSYDSSNHVV FGGGTQPTVLSGTL 2 EGERRPP GVPGRLSGSADHPPNSASLTVFGLKAGDGAGHYC QSYDGNSRAV FGEGTQLTVLPGAP 3 AASSLQS GVPSRFSGSGSGTDFTLTISSLQPEDSATYYC LQHNSLPLT FGPGTKVDIKSGIL 5 EATTLVP GTPPRFSGSGYGTDFTLTIDNIESEDAAYYFC LQDDSFPLT FGQGTKLEIKSGIL 8 ETNHRPS GAPDRFSGSIDSSSNSASLTISGLKTEDEADYHR QSYDRSNHVV FGGGTRLAVLSGIP 9 GDSQRPS AAPDRLPGSADSSSDSASPPTSGLKAEDEADYYC QSHYSSTVV FDGGAKLTALSGIL

150 1.15 Selected clones from LYN-C2

VH-FR1 VH-CDR1 VH-FR2 VH-CDR2 1 QVQLVQSEGGVVQPGRSLGLSCAAS GFTFSSYAMH WVRQAPDKGLEWVA VISYDGSNKY 3 QVQLVQSEGGVVQPGRSLRLSCAAS GFTFSSYAMH WVRQAPGKGLEWVA VISCDGSNKY 5 QVQLVQSEGGVVQPGRSLRLSCAAS GFTFSSYAMH WVRQAPGKGLEWVA VISYDGSNKY 6 QVQLVQSEGGVVQPGRSLRLSCAAS GFTFSSYAMH WVRQAPGKGLEWAA AVSYDGSNKY 7 QVQLVQSEGGVVQPGRSLRLSCAAS GFTFSSYAMH WVRQAPGKGLEWVA VISYDGSNKY 8 QVQLVQSEGGVVQPGRSLRLSCAAS GFTFSSYAMH WVRQAPGKGLEWVA VISYDGSNKY 10 QVQLVQSEGGVVQPGRSLRLSCAAS GFTFSSYAMH WVRQAPGKGLEWVA VISYDGSNKY 11 QVQLVQSEGGVVQPGRSLRLSCAAS GFTFSSYAMH WVRQAPGKGLEWVA VISYDGSNKY 12 QVQLVQSEGGVVQPGRSLRLSCAAS GFTFSSYAMH WVRQAPGKGLEWVA VISYDGSNKY 13 QVQLVQSEGGVVQPGRSLRLSCAAS GFTFSSYAMH WVRQAPGKGLEWVA VISYDGSNKY

VH-FR3 VH-CDR3 1 YADSVKGRFTISRDNSKNTLYLQMNSLRAEDTAVYYCAR DEGYYGSGCIDY WGQGTLVTVSS 3 YADSVKGRFTISRDNSKNTLYLQMNSLRAEDTAVYYCAR DEGYYGSGCIDY WGQGTLVTVSS 5 YADSVKGRFTISRDTSKNTLYLQMNSLRAEDTAVYYCAR DEGYYGSGCIDY WGQGTLVTVSS 6 YADSVKGRFTISRDNSKNTLYLQMNSLRAEDTAVYYCAR DEGYYGSGCIDY WGQGTLVTASS 7 YADSVKGRFTISRDNSKNTLYLQMNSLRAEDTAVYYCAR DEGYYGSGCIDY WGQGTLVTVSS 8 YADSVKGRFTISRDNSKNTLYLQMNSLRAEDTAVYYCAR DEGYYGSGCIDY WGQGTLVTVSS 10 YADSVKGRFTIARDNSKNTLYLQMNSLRAEDTAVYYCAR DEGYYGSGCIDY WGQGTLVTVSS 11 YADSVKGRFTISRDNSKNTLYLQMNSLRAEDTAVYYCAR DEGYYGSGCIDY WGQGTLVTVSS 12 YADSVKGRFTISRDNSKNTLYLQMNSLRAEDTAVYYCAR DEGYYGSGCIDY WGQGTLVTVSS 13 YADSVKGRFTISRDNSKNTLYLQMNSLRAEDTAVYYCAR DEGYYGSGCIDY WGQGTLVTVSS

Linker VL-FR1 VL-CDR1 VL-FR2 1 GILGSGGGGSGGGGSGGGGS NFMLTQPHSVSESPGETVTISC TGSSGSTASNSVQ WYQQRPGSAPRTVIY 3 GILGSGGGGSGGGGSGGGGS NFMLTQPHSVSESPGKTVTISC TRSSGSIATNYVQ WYQQRPGSSPTTVIY 5 GILGSGGGGSGGGGPGGGGS NFMLAQLHSVSESPGKTVTTSC TRSSGSIATNYVQ WYQQRPGSSPTTVIY 6 GILGSGGGGSGGGGSGGGGS SFMLTQPHSAPESPGKTVTISC TRSSGSTATNYVQ WYQQRPGSSPTTVIY 7 GILGSGGGGSGGGGSGGGGS NFMLTQPHSVSESPGKTVTISC TGSSGSIATNYVQ WYQQRPGSSPTTVIY 8 GILGSGGGGSGGGGSGGGGS NFMLTQPHSVSESPGKTVTISC TRSSGSIATNYVQ WYQQRPGSSPTTVIY 10 GILGSGGGGSGGGGSGGGGS NFMLTQPHSVSESPGKTVTISC TRSSGSIATNYVQ WYQQRPGSSPTTVIY 11 GILGSGGGGSGGGGSGGGGS NFMLTQPHSVSESPGKAATISC TRSSGSIASNYVQ WYQQRPGGAPTTVIY 12 GILGSGGGGSGGGGSGGGGS NFMLTQPHSVSESPGKTATISC TRSSGSIASNYVQ WYQQRPGGAPTTVIY 13 GILGSGGGGSGGGGSGGGGS ETTFTQSPAFLSATPGDKVDISC KASQDIDEDVN WYQQKPGEAPIFIIQ

VL-CDR2 VL-FR3 VL-CDR3 1 EDHQRPS GVPDRFSGSIDTSSNSASLTISGLRTEDEGDYYC RSYDSSHTV FGGGTQLTVLSGIL 3 EDNQRPS AVPDRFSGSIDSSSNSASLTISGLKTEDEADYYC QSYHSSIVV FGGGTKLTVLSGIL 5 EDNQRPS AVPDRFSGSIDSSSNSASLTISGLKTEDEAGYYC QSYHSSIVV FGGGTKLTVLPGIP 6 EDNQRPS AVPDRFSGSIDSSSNSASLTISGLKTEDEADYYC QPYDSSNPWV FGGGTKLTVLSGIL 7 EDNQRPS AVPDRFSGSIDSSSNSASLTISGLKTEDEADYYC QSYHSSIVV FGGGTKLTVLSGIL 8 EDNQRPS AVPDRFSGSIDSSSNSASLTISGLKTEDEADYYC QSYHSSIVV FGGGTKLTVLSGIL 10 EDNQRPS AVPDRFSGSIDSSSNSASLTISGLKTEDEADYYC QSYHSSIVV FGGGTKLTVLSGIL 11 EDNQRPS GVPDRFSGSIDSSSNPASLTISGLRTEDEADYYC QFYHSSIVV FGGGTKLTVLSGIL 12 EDNQRPS GVPDRLSGSIDSSSNSASLTISGLRTEDEADYYC QSYDRSNKAV FGGGTQLTILSGIL 13 DATTLVP GIPPRFSGSGHGTDFTLTIDNIESEDAAYYFC LQHDNFPPYT FGQGTKLEIKSGIL

151 1.16 Selected clones from CRN-2B6 and CRN-C5

VH-FR1 VH-CDR1 VH-FR2 VH-CDR2 2B6-1 QVQLQQWGPGLLKPSETLSLTCGVS GGSLSGYYWS WIRQSPGEELEWIG EINRDGSTN 2B6-4 QVQLQQWGPGLLKPSETLSLTCGVS GGSLSGYYWS WIRQSPGEELEWIG EINQGGSTN 2B6-6 QVQLQQWGPGLLKPSETLSLTCGVS GGSLSGYYWS WIRQSPGEELEWIG EINQGGSTN 2B6-8 QVQLQQWGPGLLKPSETLSLTCGVP GGSLSGYYWS WIRQSPGEELEWIG EINQGGSTN C5-1 QVQLQQWGPGLLKPSETLSLTCGVS GGSLSGYYWS WIRQSPGEELEWIG EINRDGSTN C5-11 QVQLQQWGPGLLKPSETLSLTCGVS GGSLSGYYWS WIRQSPGEELEWIG EINQGGSTN

VH-FR3 VH-CDR3 2B6-1 YNPSLRSRVTISVDTSKKQLSLKVNSVTASDTAVYYCAR ETFRGSNCFDS WGQGTPVTVSS 2B6-4 YNPSLRSRVTISVDTSKKQLSLKVNSVTASDTAVYYCAR ETFRGSNCFDS WGQGTPVTVSS 2B6-6 YNPSLRSRVTISVDTSKRQLSLKVNSVTASDTAVYYCAR ETFRGSNCFDS WGQGTPVTVSS 2B6-8 YNPSLRSRVTISVDTSKKQLSLKVNSVTASDTAVYYCAR ETFRGSNCFDS WGQGTPVTVSS C5-1 YNPSLRSRVTISVDTSKKQLSLKVNSVTASDTAVYYCAR ETFRGSDCFDS WSQGTPVTVSS C5-11 YNPSLRSRVTISVDTSKKQLFLKVNSVTASDTAVYYCAR ETFRGSNCFDS WDQGTLVTVSS

Linker VL-FR1 VL-CDR1 VL-FR2 2B6-1 GILGSGGSGSGGGGSGGGGS EIVLTQSPGTLSLSPGERVTLSC RASRTVSSSYLA WYQQKTGQAPRLLMY 2B6-4 GILGSGGGGSGGGGSGGGGS EIVLTQSPGTLSLSPGGRVTLSC RASQTVSSSYLA WYQQKTGQAPRLLMY 2B6-6 GILGSGGGGSGGGGSGGGGS EIVLTQSPGTLSLSPGERVTLSC RASQTVSSSYLA WYQQKTGQAPRLLMY 2B6-8 GILGSGGGGSGGGGSGGGGS EIVLTQSPGTLSLSPGERVTLSC RASQTVSSSYLA WYQQKTGQAPRLLMY C5-1 GILGSGGGGSGGGGSGGGGS EIVLTQSPGTLSLSPGERVTLSC RASRTVSSSYLA WYQQKTGQAPRLLMY C5-11 GILGSGGGGSGGGGSGGGGP NFMLTQPHSVSESPGKTVTISC TRSSGSIASNYVQ WYQQRPGSFPTTVIY

VL-CDR2 VL-FR3 VL-CDR3 2B6-1 GASTRAT GIPDRFSGSGSGTDFTLTISRLEPEDFAVYYC QQYGSSRWT FGQGTKVGIKSGVL 2B6-4 GASTRAT GIPDRFSGSGSGTDFTLTISRLEPEDFAVYYC QQYGSSRWT FGQGTKVEIKSGIL 2B6-6 GASTRAT GIPDRFSGSGSGTDFTLTISRLEPEDFAVYYC QQYGSSRWT FGQGTKVEIKSGIL 2B6-8 GASTRAT GIPDRFSGSGSGTDFTLTISRLEPEDFAMYYC QQYGSSRWT FGQGTKVEIRSGIL C5-1 GASTRAT GIPDRFSGSGSGTDFTLTISRLEPEDFAVYYC QQYGSSRWT FGQGTKVGIKSGVL C5-11 EDDQRPS GVPERFSGSIDSSSNSASLTISGLKTEDEADYYC QSYDSSDGWV FGGGTKLAVLSGIL

1.17 Selected clones from MGE-Fab-M6

VH-FR1 VH-CDR1 VH-FR2 VH-CDR2 1 QVQLQESGPGLVKPLETLSLTCTVS GGSISNSGYYWG WIRQPPGKGLEWIG SINHYGTTY 2 QVQLQESGPGLVKPSETLSLTCTVS GGSISNSDYYWG WVRQPPGKGLEWIG SINYYGTTY 3 QVQLQESGPGLVKPSETLPLTCTVS GGSISNSDYYWG WVRQPPGKGLEWIG SINYYGTTY

VH-FR3 VH-CDR3 1 YNPSLKSRVAMSVDTSKSQFSLKLSSVTAADTAVYYCAR RRANSWSFDL WGRGTLVTVSS 2 YNPSLKSRVAMSADTSKNQFSLKLSSVTAADTAVYYCAR RGANSWFFDL WGRGTLVTVSS 3 YNPSLKSRVAMSVDTSKNQFSLKLSSATAADTAVYYCAR RGANSWLFDL WGRGTLVTVSS

VL-FR1 VL-CDR1 VL-FR2 1 DIQVTRARMETK 2 DIQVTRARMETK 3 DFMLTQPHSVSESPGKTVTISCTGS SGSIASNYVQ WYQQRPGSAPTTVIF

VL-CDR2 VL-FR3 VL-CDR3 1 2 3 EDNQRPS GVPDRFSGSIDSSSNSASLTISGLKTEDEADYYCQS YDRSNHAVFGRGTQLTVL

152 Appendix 2: Protocol

Yeast panning of mammalian cell monolayers (adapted from Wang & Shusta, J Immunol Methods, 304:30, 2005 and Steve Sazinsky)

Materials:

PBSCM (PBS w/ 1 mM CaCl2, 0.5 mM MgSO4, 0.5% BSA), sterile filtered SDCAA, pH 3.5 (20 g/L glucose, 6.7 g/L yeast nitrogen base, 5 g/L casamino acids, 14.7 g/L sodium citrate, 4.29 g/L citric acid monohydrate), pH to 3.5, sterile filtered

Note: keep PBSCM and 6-well plate on ice during Steps 3-7.

1. Grow mammalian cell monolayer to confluency. 2. Serum starve cell monolayers overnight before panning. 3. Wash monolayers 3x with 1 mL PBSCM. 4. Wash yeast cells 2x with 1mL PBSCM: a. Use 2x106 yeast/cm2 of monolayer cells. b. For 6-well plates, use 2x107 cells/well. c. Use 1 well for yeast displaying an irrelevant antibody, and 1 well for a yeast- displayed positive control. 5. Resuspend cells in 1 mL PBSCM (per well) and pipette yeast slurry onto monolayer dropwise to cover all surface area. 6. Gently swirl plate and incubate at 4 oC with gentle rocking for 2 h. 7. Wash wells 2x with PBSCM: a. Tilt plate and aspirate yeast slurry. b. Gently pipette 1 mL buffer onto side of well. c. Rock gently at 4 oC for 1 min. 8. Add 1 mL SDCAA, pH 3.5 to each well and rock vigorously at room temperature for 20 min to elute the bound yeast from the monolayer. 9. Measure OD600 to determine number of cells captured. After OD measurement, yeast can be removed from cuvette, and dilutions can be plated on SDCAA plates as an alternate measure of cells captured.

153 Office: Home: 77 Massachusetts Ave. E19-563 345 Franklin St. #105 Cambridge, MA 02139 Cambridge, MA 02139 (617) 258-5293 (409) 350-2984 [email protected]

Ginger Chao

Education Massachusetts Institute of Technology, Cambridge, MA Doctoral candidate in Chemical Engineering, Ph.D. anticipated October 2007 Minor in Biology, GPA: 4.9/5.0

Rice University, Houston, TX Bachelor of Science in Chemical Engineering with Bioengineering Focus Conferred Summa Cum Laude May 2002

Research Massachusetts Institute of Technology January 2003 – present Experience Graduate Research Assistant

Thesis Advisor: K. Dane Wittrup

Thesis Title: “Characterizing and engineering antibodies against the epidermal growth factor receptor (EGFR)”

• Developed a novel method for epitope mapping antibodies using random mutagenesis and yeast surface display • Utilized above method to determine the epitopes of three therapeutically relevant antibodies against EGFR • Continued work on engineering single-chain antibodies against dimerization epitopes of EGFR using yeast surface display and directed evolution • Created a mathematical model for equilibrium receptor-ligand binding and dimerization to describe the elimination of high affinity EGF binding to

EGFR through antibodies against dimerization epitopes

Pharmacia Summer 2001 Summer Engineering Intern Conducted laboratory experiments on catalyst immobilization and reaction optimization to yield high enantiomeric selectivity

Rice University Summer 1999 Undergraduate Research Assistant Conducted research sponsored by the BP Amoco Summer Research Program, with results included in publication:

Lee, J.T., Chou C.Y., Chao G., Pilaski D.R. & Robert M. (2005). Two- dimensional diffusion of colloids in polymer solutions. Mol. Physics 103, 2897-2902.

154 Publications Chao G., Cochran J.R. & Wittrup K.D. (2004). Fine epitope mapping of anti- epidermal growth factor receptor antibodies through random mutagenesis and yeast surface display. J. Mol. Biol. 342, 539-550.

Colby D.W., Garg P., Holden T., Chao G., Webster J.M., Messer A., Ingram V.M. & Wittrup K.D. (2004). Development of a human light chain variable domain (VL) intracellular antibody specific for the amino terminus of huntingtin via yeast surface display. J. Mol. Biol. 342, 901-912.

Sivasubramanian A., Chao G., Pressler H.M., Wittrup K.D. & Gray J.J. (2006). Prediction of the mAb 806-EGFR complex structure by computational docking followed by computational and experimental mutagenesis. Structure 14, 401-414.

Chao G., Lau W.L., Hackel B.J., Sazinsky S.L., Lippow S.M. & Wittrup K.D. (2006). Isolating and engineering human antibodies using yeast surface display. Nat. Protocols 1, 755-768.

Supervisory Supervision of MIT Undergraduate Research Opportunity Program Experience (UROP) students: Heather Pressler, “Mutation analysis of the antibody 806 epitope” (2004-2005) Nick Nguyen, “Measuring anti-EGFR antibody endocytosis rates” (2007)

Teaching Assistant Experience: Introduction to Experimental Biology, MIT (Spring 2006) Conducted daily recitation lectures, provided instruction and management during laboratory hours, and graded exams

Thermodynamics I, Rice University (2001-2002) Held weekly office hours and graded homework assignments

Skills of Recombinant DNA techniques Interest Protein biochemistry Protein engineering/directed evolution using yeast surface display Fluorescence activated cell sorting (FACS) Mammalian cell culture MATLAB programming

155

Honors National Defense Science and Engineering Graduate Fellowship National Science Foundation Graduate Fellowship (declined) David Koch Graduate Fellowship (MIT Center for Cancer Research) Robert T. Haslam Presidential Graduate Fellow, MIT Outstanding Chemical Engineering Seminar Award (Fall 2005) Rice University President’s Honor Roll Tau Beta Pi National Engineering Honor Society Ann and Joe Hightower Award in Chemical Engineering (Rice University)

Presentations, Oral Presentations: Posters, and Chao G., Olsen M.J., Wolf-Yadlin A. & Wittrup K.D. (2005). Engineering Patents antibodies against the epidermal growth factor receptor to block dimerization. American Institute of Chemical Engineers (AIChE) Annual Meeting.

Chao G., Cochran J.R. & Wittrup K.D. (Sept. 2006). Epitope mapping of antibodies against the epidermal growth factor receptor (EGFR). American Chemical Society (ACS) National Meeting.

Posters: Chao G., Olsen M.J., Wolf-Yadlin A. & Wittrup K.D. (2005). Engineering anti- epidermal growth factor receptor (EGFR) antibodies to block receptor dimerization. Keystone Symposium: Antibody-Based Therapeutics for Cancer.

Chao G., Wolf-Yadlin A., Olsen M.J., Lauffenburger D.A. & Wittrup K.D. (2007). Antibodies against domains II or IV of the epidermal growth factor receptor (EGFR) inhibit high affinity EGF binding. Keystone Symposium: Antibodies as Drugs.

Chao G., Wolf-Yadlin A., Olsen M.J., Lauffenburger D.A. & Wittrup K.D. (2007). Antibodies against domains II or IV of the epidermal growth factor receptor (EGFR) inhibit high affinity EGF binding. Protein Society Annual Symposium.

Patent Applications: Johns T.G., Scott A.M., Burgess A.W., Old L.J., Adams T.E., Wittrup K.D. & Chao G. (2005). EGF receptor epitope peptides and uses thereof.

Wittrup K.D., Lauffenburger D.A, Wolf-Yadlin A., Olsen M.J. & Chao G. (2005). Antibodies and peptides targeted to specific epidermal growth factor receptor epitopes for diagnostic and therapeutic applications in human disease. (Provisional).

Way J.C., Chao G., Cresson C., Lauffenburger D.A., Wittrup K.D. (2007). Treatment of tumors expressing mutant EGF receptors.

156