Dysregulation Of Cell Signaling Pathways

Published Date: 02 Nov 2017

Dysregulation of cell signaling pathways that mediate proliferation, survival and migration are an underlying cause of many

cancers. In particular, dysregulation and over-expression of the vascular endothelial growth factor-2 (VEGFR2), αvβ3 integrin

and prostate specific membrane antigen (PSMA) correlate with poor prognosis in many human cancers, making these agents

attractive candidates for therapeutic intervention. Numerous papers have demonstrated cross-talk between biological

processes mediated by VEGFR2, αvβ3 integrins, PSMA, and their ligands, particularly pathways responsible for angiogenesis

(1, 2). Dual-specific proteins that can target and inhibit the activity of multiple receptors will have greater potential than singletargeted

agents to improve binding affinity, avidity, potency, and selectivity. In addition, agents that have the potential to

modulate multiple receptors would be beneficial due to differential expression of disease markers in different patients and the

ability of this expression to change over time. In parallel, it is critical to develop noninvasive molecular imaging agents that can

identify the patients who would benefit most from these targeted treatments and that can be used to monitor treatment and

disease progression. In particular, the number of imaging probes in clinical development for non-invasive visualization of

VEGFR2, αvβ3 integrin and PSMA expression is extremely limited. The availability of such dual-specific imaging agents would

aid both in cancer diagnostics, staging and disease management and in identifying patients who would be good candidates for

VEGFR2-, αvβ3 integrin- and PSMA-targeted therapies.

Most currently available bispecific protein therapeutics are antibodies (Abs) or antibody fragments that are assembled

through associating domains or a flexible tether. To overcome the many problems associated with the therapeutic and

diagnostic uses of antibodies or antibody fragments (3-5), we recently used an approach that entails engineering natural

protein ligands to function as non-immunoglobulin alternatives to antibodies. In particular, we engineered single-chain dualspecific

protein variants (scVEGFrgd) that antagonize both VEGFR2 and αvβ3 integrin and inhibit in vitro angiogenic

processes more effectively than variants that target VEGFR2 or αvβ3 integrin alone (6). We then used a random mutagenesis

and an affinity maturation process to identify mutations that significantly increase the affinity of the scVEGF variants for

VEGFR2. The identification of the mutations that significantly increase the affinity of the scVEGF variants to both αvβ3 integrin

and VEGFR2 independently allowed us to tailor a combination scVEGF variant (scVEGFcombo) with a significant enhanced

affinity to cells that express both receptors.

In this proposal, we will perform pre-clinical studies on these dual-specific proteins to determine their potential as in vivo

molecular imaging agents and cancer therapeutics. We will also extend the platform we have developed to engineer

bispecific scVEGF proteins that bind to both VEGFR2 and PSMA. Upon completion of these aims, we will have identified new

classes of engineered multispecific scVEGF proteins against several important tumor vasculature receptors for future testing

as molecular imaging agents and in tumor targeting applications. Below I describe the state of the art for the different building

blocks of my system, thereby shedding light on the rationale underlying my approach.

Role of integrin receptors in human cancers: Integrins are a family of cell surface adhesion receptors that noncovalently

associate into α/β heterodimers with distinct ligand binding specificities and cell signaling properties (7). Integrins

bind to extracellular matrix (ECM) proteins, such as vitronectin, fibronectin, collagens, and laminins, thereby promoting cell

adhesion to the ECM and activation of the signaling pathways involved in cell cycle progression (8), angiogenesis, tumor

growth and metastasis (9). In particular, αvβ3 integrins are highly expressed on activated endothelial cells in the tumor

neovasculature but are weakly expressed in resting endothelial cells and in most normal tissues and organs (10). Therefore,

there is a critical need for compounds that will target αvβ3 integrins with high binding affinity and specificity for cancer

diagnosis and therapy (11). Many integrins, including those containing the αv subunit, recognize an Arg-Gly-Asp (RGD)

tripeptide motif typically found in flexible solvent-exposed loops within ECM ligands. In natural ligands, the RGD motif is

typically found in flexible solvent-exposed loops; however, the amino acid residues flanking the RGD sequence and hence the

three-dimensional orientation of these loops varies substantially across integrin ligands and dictates their integrin binding

affinity and specificity. There are currently a few monoclonal antibodies (mAb) or peptide-based compounds that target

integrins in clinical trials as anticancer agents—CNTO95 (Centocor) and the cyclic peptide Cilengitide (Merck) target αvβ3 and

αvβ5, the mAb Vitaxin (MedImmune) targets αvβ3, the peptide ATN-161 (Attenuon) targets αvβ3 and α5β1, and the mAb

Volociximab (PDL) targets α5β1 [reviewed in ref (12)].

Clinical importance of PSMA: Recent advances in expression profiling have identified PSMA as an ideal target for

prostate cancer diagnostic agents and therapeutics, especially in advanced disease. PSMA is a type 2 transmembrane

glycoprotein homodimer expressed almost exclusively in prostatic epithelial cells (13). PSMA has glutamate carboxypeptidase

II activity, and both the expression and enzymatic activities of PSMA are elevated in prostate cancer, with expression levels

being closely correlated with disease grade (14). The highest levels of PSMA expression are associated with hormonerefractory

metastatic prostate cancer (15). In addition, endothelial cells of the neovasculature of almost all solid tumors – but

not normal tissues – express PSMA (16). There are several examples of antibodies directed against the extracellular domain

of PSMA that have been used to detect and treat primary and metastatic prostate cancer, including phase I and II clinical trials

of the humanized mAb HuJ591 for both therapy and diagnosis and a phase I trial of the PSMA-directed immunoconjugate

MLN2704 (which includes the PSMA-targeted monoclonal antibody MLN591) in patients with progressive metastatic

castration-resistant prostate cancer. Yet, none of these antibodies is, as yet, FDA approved.

Therapeutic importance of VEGFR2: VEGF is a homodimeric protein that mediates angiogenesis in endothelial cells

through binding and activation of the receptor tyrosine kinase VEGFR2 (17). Binding of VEGF homodimers to two molecules of

VEGFR2 activates the receptor, resulting in autophosphorylation and activation of a number of signaling pathways, including

mitogen-activated protein kinase, Src, Akt, and focal adhesion kinase (18). The net results of these signaling events include

cell proliferation and migration and angiogenesis. Much like αvβ3 integrin and PSMA, VEGFR2 is overexpressed in the

2

endothelium of the tumor vasculature, yet is almost undetectable in the endothelium of adjacent normal tissue (19). The

development of agents that prevent activation of VEGFR2 have attracted a great deal of interest as anti-cancer therapeutics.

The anti-VEGF antibody bevacizumab (Avastin) is FDA approved for use in combination therapy in metastatic colorectal

cancer, non-small cell lung cancer, and metastatic breast cancer and is undergoing clinical trials for other applications (20).

VEGF-Trap (Regeneron) is a soluble version of VEGFR that targets the ligand as a dominant negative antagonist. While these

two agents target VEGF, IMC-ICII (Imclone) is a mAb against VEGFR2 that is in phase I clinical trials (21).

Protein-based targeting agents identified by phage display bind with weak affinities: In addition to mAbs, there have

been efforts to develop peptide-based antagonists to VEGFR2, αvβ3 integrin and PSMA. A number of peptides that inhibit

VEGFR2 at high nanomolar to micromolar concentrations have been identified by phage display, but the modest affinity of

these compounds makes it unlikely that they will move forward to clinical development (22). A more promising antagonist

developed by Adnexus is a VEGFR2-binding protein based on engineered fibronectin domains that is currently in phase I and

II clinical trials for glioblastoma (23). In another approach, mutations were introduced into the VEGF ligand, which changed its

activity from agonistic to antagonistic (24). Wild-type VEGF, which exists as a number of splice variants of different lengths, is

a homodimer with two VEGFR1 or VEGFR2 binding sites, one at each pole of the protein (Fig. 1A). The VEGF variants

developed by Siemeister et al. (25) and Boesen et al. (24) have mutations that block interactions between one of the VEGF

binding sites and VEGFR2, while leaving the other binding site intact (Fig. 1B). In particular, the antagonist VEGF mutant

described by Boesen et al. is a single-chain variant with the

two VEGF subunits separated by a 14-amino acid linker

(24). Although this mutant specifically antagonized the

mitogenic effects of the wild-type ligand on endothelial cells,

its potency was low, probably due to diminished affinity for

VEGFR2 relative to wild-type VEGF.

To date, there are no specific and effective inhibitors

against PSMA. For example, a random phage-displayed

library identified a linear peptide dimer containing a dihistidine

(HH) motif with relatively low affinity (KD in the low

micromolar range) and ability to inhibit PSMA enzymatic

activity (at concentrations in the micromolar range) (26).

Importantly, the HH peptide motif had also emerged as part

of a consensus PSMA-binding sequence identified in

another phage display-based screening of a cyclic sixamino-

acid peptide library (27). This cyclic peptide was also

able to bind purified PSMA (in the 10 μM range) and to

target phage to prostate cancer cells, but in contrast to the

linear peptide, it stabilized the PSMA to enhance its

enzymatic activity.

Molecular imaging of integrins, PSMA and VEGFR: Despite intensive efforts in the field there are currently no clinically

approved molecular imaging agents for αvβ3 integrin, PSMA extracellular domain (ECD) or VEGFR2. The PSMA antibody,

capromab pendetide, labeled with 111In is marketed as ProstaScint, an FDA-approved preparation for the detection of nodal

metastases in prostate cancer patients (28). However, this antibody is directed against an intracellular epitope of PSMA, which

is considered a suboptimal target for antibody imaging. A number of studies have been conducted with the mAb J591, which is

directed against an epitope on the extracellular domain of PSMA (29). One such study has shown that J591 accumulated in

metastatic prostate cancer lesions (30). In a recent phase I trial, the feasibility of targeting the neovasculature of a wide range

of adenocarcinomas with 111In-labeled humanized J591 was investigated in patients with melanoma and cancers of the breast,

colon, liver, and kidney. In these patients the antibody accreted in all known tumor sites. It was also found that of a group of

patients shown by standard scans to have soft tissue disease, almost all also showed antibody uptake in the soft tissues, as

was the case for all patients with bone disease. These data show selective targeting of PSMA expressed on tumor

endothelium (31). More recently, the first positron emission tomography (PET) imaging agent for PSMA was synthesized. In

extension of their previous work with the 11C-labeled DCMC, which binds to the active carboxy peptidase site of PSMA (32),

Mease et al. synthesized 18F-DCFBC, which was demonstrated to localize in PSMA-expressing tumors in mice, thereby

facilitating imaging by small-animal PET (33).

Rational drug design and phage display have identified small peptides and peptidomimetics that target αvβ3 (and αvβ5)

integrins or α5β1 integrin (34). However, chemical modification to improve the receptor binding affinity, tumor uptake, and in

vivo pharmacokinetics of these small peptides has not been successful. Moreover, covalent attachment of imaging probes

affects their integrin binding-properties (35). Suboptimal tumor targeting efficacy and pharmacokinetics have therefore limited

the clinical translation of these probes as molecular imaging agents (36). Only one compound, a glycosylated cyclic RGD

pentapeptide, 18F-galacto-c(RGDfK), has advanced to the clinical level for molecular imaging in human subjects (37). While

this agent was able to identify integrin-positive lesions in human subjects and image intensity correlated with αvβ3 integrin

expression, its relatively poor tumor uptake and high background (e.g., in the liver) indicates that there is room for substantial

improvement. In a human melanoma mouse xenograft model, 18F-galacto-c(RGDfK) also exhibited low tumor uptake values

(35), suggesting that imaging agents that have weak tumor contrast in small animal models will also exhibit weak imaging

signals in humans.

Figure 1. (A) VEGFwt is a homodimer that binds to two VEGFR2

molecules to activate signaling. (B) Single-chain mutant VEGF

(scmVEGF) has one of the VEGFR2 binding sites disrupted, preventing

a second receptor molecule from binding and thereby antagonizing

signaling. (C) The variants we plan to engineer here have one VEGFR

binding site mutated with a αvβ3 integrin or PSMA recognition loop,

making them capable of binding VEGFR2 and/or αvβ3 integrin and/or

PSMA. We will test whether these unique mutants can antagonize

signaling events mediated by VEGFR2, αvβ3 integrin and PSMA.

3

To date, very little work has been reported on developing VEGFR2 imaging agents. A small number of studies have used

radiolabeled VEGF ligands to image VEGFR2 expression in vivo (reviewed in ref. (38)): Small-animal PET imaging revealed

rapid, specific, and prominent uptake of 64Cu-DOTA-VEGF121 in highly vascularized small U87MG human glioblastoma tumors,

but significantly lower and sporadic uptake in large U87MG tumors (39). In addition, strong liver and kidney uptake was

observed, due to high VEGFR1 expression, in the kidney, with a clearance route primarily via the hepatic pathway. Modified

VEGF mutants with 20-fold decreased binding affinity for VEGFR1 showed decreased kidney uptake (40) and PEGylated

VEGF showed enhanced pharmacokinetics (41). However, vast improvements and further development of VEGFR2 imaging

agents is desperately needed.

Protein engineering through yeast surface display: Yeast surface display is a powerful directed evolution technology

that has been used to engineer proteins for enhanced binding affinity (42), proper folding (43), and improved stability (44).

Previously, this technology was applied to rapidly characterize protein-protein interactions (45), to engineer the epidermal

growth factor receptor for soluble expression in yeast (43), and to engineer growth factor ligands with significantly enhanced

receptor binding affinity for use in wound healing applications (46). We have also recently applied this technology to engineer a

single-chain vascular endothelial growth factor (scVEGF) to bind simultaneously to VEGFR2 and αvβ3 integrin with antibodylike

affinity (6). The yeast surface display system takes advantage of the chaperone-assisted folding, the disulfide bond

formation, and the quality control mechanisms of the eukaryotic secretory pathway (47). A schematic representation of yeast

display is shown in Fig. 2. Proteins are displayed on the yeast surface through genetic fusion to the yeast mating agglutinin

protein Aga2p (48). Aga2p is disulfide bonded to Aga1p, which is covalently linked to the yeast cell wall. The protein of interest

is flanked by N-terminal hemagglutinin (HA) and C-terminal c-myc epitope tags, which

are used to confirm expression of the construct on the yeast cell surface and to

quantify surface expression levels (48). High-throughput screening of tens of millions

of yeast-displayed mutants allows for the rapid isolation of proteins with altered

properties. Combinatorial libraries of up to 108 transformants can routinely be created,

with each yeast cell displaying approximately 50,000 identical copies of a particular

mutant protein on its surface. Yeast-displayed protein libraries are then stained with a

fluorescently labeled ligand or receptor. Fluorescence activated cell sorting (FACS) is

used to screen the yeast libraries for mutants with a desired phenotype, usually

increased binding affinity to a target protein.

2. Preliminary results. We have started the development of a dual-specific agent

that antagonizes both VEGFR2 and αvβ3 integrin and effectively inhibits angiogenic

processes in vitro compared to a variant that targets only VEGFR2 (6). To achieve this

goal, we first determined the utility of VEGF as a molecular scaffold for protein

engineering. Specifically, mutations were introduced into a single-chain VEGF

(scVEGF) ligand that retained VEGFR2 binding, but prevented receptor

dimerization and activation (Fig. 1B). We created 32 different yeast-displayed

scVEGF mutant (scmVEGF) libraries containing randomized loops to determine

regions of the protein that could tolerate insertion of an additional epitope without

disrupting VEGFR2 binding. We then used this information to create and screen

combinatorial scVEGF libraries to identify dual-specific proteins (scVEGFrgd) that

bound to both VEGFR2 and αvβ3 integrin (Fig. 1C). Using surface plasmon resonance

(SPR) and cell binding experiments, we showed that the purified dual-specific proteins have the capability to simultaneously

engage both VEGFR2 and αvβ3 integrin with antibody-like affinities. We also showed that the engineered proteins bind to both

human and murine versions of these receptors. Dual-specific proteins, but not a variant that targets VEGFR2 alone,

significantly inhibited adhesion and VEGF-mediated receptor phosphorylation and proliferation of endothelial cells cultured on

vitronectin-coated surfaces. Moreover, dual-specificity conferred nearly complete inhibition of VEGF-mediated blood vessel

formation in Matrigel plugs in vivo, while a protein variant that bound only VEGFR2 was marginally effective. Collectively, this

work provides proof of concept for an approach to creating dual-specific proteins where additional functionality is introduced

into a natural protein ligand to complement its existing biological properties. Our in vitro data provides validation for moving

forward with pre-clinical testing and further development as in vivo imaging agents and therapeutics, and supports further

study of the dual-specific proteins sequence-structure relationships that confer their enhanced activity.

We further explored the concept of converting a VEGF ligand from an agonist to an antagonist by using an affinity

maturation process. To enhance the VEGFR2 binding of the antagonist, random mutations were introduced into the entire

scmVEGF gene, followed by screening the obtained random mutagenesis libraries against VEGFR2 and selecting the

mutants with highest affinity in increasingly stringent sorts. Using SPR spectroscopy, we identified variants with significantly

higher affinity for recombinant VEGFR2 than the parent scmVEGF (Fig. 1B) antagonist and the wild-type scVEGF

(scVEGFwt, Fig. 1A) agonist, even though the variants possessed only a single receptor binding site.

The identification of the mutations that significantly increase the affinity of the scVEGF variants to both αvβ3 integrin and

VEGFR2 independently allowed us to tailor a combination scVEGF variant (scVEGFcombo) with a significant enhanced

affinity to cells that express both receptors. The combo variant contains the engineered integrin-binding loop from the most

potent dual-specific scVEGF integrin binder and the amino acid mutations that are derived from the most potent affinity

matured VEGFR2 binder. Due to its expected tight binding to αvβ3 integrin and VEGFR2, this mutant was expected to disrupt

Figure 2. Yeast surface display

system. Protein expression on the

yeast cell surface can be measured

by flow cytometry using antibodies

against the C-terminal c-myc tag,

and protein-protein interactions can

be measured with antibodies against

a ligand of interest, or through

fluorescent-labeled ligands.

Ag a1p

Ag a2p

HA

C -myc

protein to be

eng ineered

s oluble

lig and

yeas t

mating

proteins

Yeast mating

proteins

Soluble

ligand

Protein to be

engineered

Yeast cell wall

~50,000 copies

4

VEGFR recognition at one pole of the molecule and integrin recognition on the other pole. Our most recent studies show that

scVEGFcombo significantly inhibits adhesion and VEGF-mediated receptor phosphorylation and proliferation of endothelial

cells cultured on vitronectin-coated surfaces. Moreover, the combo mutant confers nearly complete inhibition of VEGFmediated

blood vessel formation in Matrigel plugs in vivo.

3. Hypothesis and specific aims. This research is based on the following hypothesis: Improved VEGFR2, αvβ3 integrin and

PSMA receptor binding affinity and protein stability will result in enhanced in vivo biological activity and specificity of bispecific

antagonistic scVEGF ligands. To achieve our overall aim of further developing and characterizing the engineered scVEGF

mutants discussed above and extending the bispecific scVEGF platform to other ligand-receptor systems and biological

processes, we will address the following specific aims: Aim 1: Test engineered bispecific scVEGF proteins as VEGFR2 and

αvβ3 integrin receptor imaging agents in mouse tumor models. Aim 2: Measure the effects of engineered bispecific scVEGF

proteins on tumor progression, angiogenesis, and metastasis in mouse tumor models. Aim 3: Engineer bispecific scVEGF

proteins that bind with high affinity to both VEGFR2 and PSMA. Aims 1 and 2 address further development of scVEGF

mutants, scVEGFrgd and scVEGFcombo, and testing their potential as in vivo imaging agents and cancer therapeutics, while

Aim 3 focuses on engineering and characterization of another dual-specific protein that targets both VEGFR2 and PSMA.

4. Experimental plan

Aim 1: Test engineered bispecific scVEGF proteins as VEGFR2 and αvβ3 integrin receptor imaging agents in mouse

tumor models. We will test the ability of labeled bispecific scVEGF proteins (i.e., scVEGFrgd and scVEGFcombo) to target

tumors in mouse models for development and use as optical and PET imaging agents. The PET imaging experiments will be

carried out in collaboration with Prof. Eyal Mishani, at the cyclotron radiochemistry unit at Hadassah Medical Center in

Jerusalem (see attached support letter). All the animal experiments in this proposal will be conducted using protocols approved

by BGU’s Committee of Use and Care of Animals. Fluorescent labeling of scVEGF proteins: Using molecular cloning, a

cysteine residue will be introduced at the N- or C-terminus of scVEGF proteins. This free thiol will be used for site-specific

conjugation of a maleimide derivative of the near-infrared dye Cy5.5. In addition, random conjugation of Cy5.5 to the ε-amino

group of lysine residues will be performed using an N-hydroxysuccinimide (NHS) ester derivative of Cy5.5. Recombinant

proteins will be expressed in the yeast strain Pichia pastoris with an enzymatically cleavable hexahistidine residue, treated with

mild reducing agent if necessary to liberate free thiol, and purified by both metal chelating chromatography and gel filtration

chromatography. Radiolabeling of scVEGF proteins: DOTA will be site-specifically conjugated to the N- or C-terminus of

scVEGF proteins through an introduced cysteine residue or randomly conjugated to the ε-amino group of lysine residues. For

thiol conjugation, a heterobifunctional crosslinker (e.g., sulfo-SMCC), which contains both maleimide and NHS ester, will first

be reacted with a molar excess of 2-(4-aminobenzyl)-DOTA. After incubation, unreacted NHS ester will be quenched, and

cysteine-containing scVEGF proteins will be added to the mixture to generate scVEGF proteins with DOTA conjugated at the

N- or C-terminus. In the second approach, a sulfosuccinimide ester derivative of DOTA will be synthesized by activation with 1-

ethyl-3-[3-(dimethylamino)propyl]carbodiimide (EDC) and N-hydroxysulfonosuccinimide. A molar excess of that compound will

be reacted with scVEGF proteins to randomly conjugate DOTA groups to lysine side chains. The resulting DOTA-scVEGF

conjugates will be purified by gel filtration chromatography. DOTA-conjugated scVEGF proteins (∼25 μg) will be radiolabeled

with 64Cu by incubating them with 5 mCi 64CuCl2 in 0.1 N sodium acetate (pH 6.3) for 1 h at 45 °C. The reaction will be

terminated by the addition of EDTA. The radiolabeled complexes will be purified by radio-HPLC using a gamma detector.

Receptor binding and internalization assays: Binding of Cy5.5 or DOTA conjugated scVEGF proteins to U87MG, A549, and

PC3 tumor cells that express both αvβ3 integrin and VEGFR2 (49) will be performed and compared to the binding of unlabeled

proteins. The level of cell surface-bound and internalized scVEGF proteins will be measured using direct binding assays.

Blocking studies will be performed on all cell lines by adding an excess of unlabeled scVEGF to measure non-specific binding

of labeled compounds and to determine if internalization is mediated by VEGFR2 and αvβ3 integrin receptor binding. Optical

and microPET imaging and biodistribution studies: U87MG, A549, and PC3 human tumor xenograft models will be

created as described below in Aim 2. In addition, an experimental metastasis mouse model will be generated by injecting

A549 cells via the tail vein for lung colonization, and an orthotopic model will be generated by implanting PC3 cells in the

prostate. Near-infrared fluorescence (NIRF) imaging will be performed with an IVIS 200 imaging system. Tumor-bearing mice

will be injected via tail vein with 1.5 nmol of Cy5.5-labeled scVEGF proteins, and images will be acquired at 0.5, 1, 2, 4, and 24

h. Tumor contrast will be quantified by drawing identically sized regions of interest (ROI) around the tumor and background

tissue located in the mouse's flank. For biodistribution studies, mice will be euthanized at 0.5, 1, and 24 h post injection of

labeled proteins. Organs and tumor tissue will be removed, and the fluorescence flux will be measured by fluorescence

imaging and reported as flux/gram of tissue. MicroPET imaging will be performed with a microPET R4 rodent model scanner

using 5-10 static scans. Tumor-bearing mice will be injected with ∼50-100 μCi of 64Cu-DOTA-labeled scVEGF proteins via the

tail vein and images will be acquired at 0.5, 1, 2, 4, and 24 h. ROIs will be drawn over the tumor on decay-corrected whole

body images. For biodistribution studies, mice will be euthanized at 0.5, 1, and 24 h post injection of radiolabeled proteins.

Organs and tumor tissues will be removed, and the radioactivity will be measured using gamma counting and reported as the

%ID/g of tissue. Dynamic scanning experiments will be performed over 35 min to measure the pharmacokinetics and

radiolabeled scVEGF proteins in the blood circulation (heart), kidneys, liver, tumor, and muscle. Serum and metabolic

stability of labeled scVEGF proteins: To test serum stability, Cy5.5 or 64Cu-DOTA-labeled scVEGF proteins will be

incubated in mouse serum for 1, 4, and 24 h at 37 °C. After incubation, the mixture will be analyzed by reverse phase HPLC

for the presence of intact protein, fragments, or free 64Cu or Cy5.5 using a radiodetector or photodiode array detector,

5

respectively. For in vivo metabolite analysis, tumor-bearing mice will be injected with 200-400 μCi of 64Cu-DOTA-labeled

scVEGF proteins via the tail vein. Mice will be euthanized at 1 and 4 h post injection, and blood, kidney, liver, and tumor tissue

will be removed, homogenized, and filtered. Each filtrate will be analyzed by reverse phase HPLC, and fractions will be

collected at 30 s intervals and analyzed by gamma counting or by using a photodiode array detector to assay for the presence

of intact protein, fragments, or free 64Cu or Cy5.5, respectively. Potential problems and alternatives: We will also use

oncogene-inducible mouse models (e.g. Kras, myc) since scVEGF proteins, although being engineered against the human

VEGFR2 and αvβ3 integrin receptors, are found to also bind to mouse VEGFR2 and αvβ3 integrin with high affinity (6). We will

also explore the use of other PET imaging labels, including 18F conjugation trough the introduced N- or C-terminal cysteine

residue. The short half-life of 18F, which is FDA-approved for clinical imaging, should be compatible with the fast clearance

expected for a relatively small scVEGF protein. Furthermore, although we expect scVEGF mutants to be tolerable to

incubation at the elevated temperatures required for 64Cu radiolabeling through chelation (i.e., a minimum of 1 h at 42 ºC), the

mild 18F coupling chemistry is a good alternative if scVEGF protein integrity is not retained. If low tumor uptake results from fast

blood clearance or if significant non-target tissue uptake (e.g., kidney or liver) is observed with engineered scVEGF proteins,

we will conjugate them to polyethylene glycol (PEG) in an attempt to obtain more desirable pharmacokinetic and in vivo

biodistribution. In previous experience with using Cy5.5-labeled proteins as αvβ3 integrin imaging agents, Cy5.5 resulted in

high levels of kidney uptake and moderate liver uptake (50). Therefore, we will also pursue the use of other near-infrared dyes

including Alexa 680 or Alexa 750 (Invitrogen), Dylight 680, 750, or 800 (Pierce Chemical Company) and IRdye 800 (Li-Cor) to

determine if lower levels of non-target tissue can be achieved compared to Cy5.5.

Aim 2: Measure the effects of engineered bi-specific scVEGF proteins on tumor progression, angiogenesis, and

metastasis in mouse tumor models. Engineered scVEGF proteins will be tested for their ability to inhibit tumor growth,

metastasis, and angiogenesis in subcutaneous xenograft models, and orthotopic or injection models generated from PC3

prostate and A549 lung carcinoma cells. These systems were chosen because they were previously used to test the

therapeutic efficacy of other VEGFR2 and αvβ3 integrin receptor targeting agents (49). We will use bioluminescence imaging

to non-invasively follow tumor growth and metastasis in response to treatment with wild-type scVEGF, engineered scVEGF

protein, or a vehicle (negative) control. A549 and PC3 cells containing luciferase or enhanced luciferase (luc2) vectors

(commercially available from Caliper Life Sciences) will be subcutaneously implanted into the shoulders of nude or SCID-bg

mice to form xenografts tumors. In addition, PC3 cells will be orthotopically implanted into the prostate, while A549 cells will be

injected intravenously for lung colonization. Although tumor growth will be monitored by electronic calipers, bioluminescence

performed in parallel will allow us to monitor early tumor development and to detect metastasis earlier and with higher

sensitivity. Ex vivo bioluminescence imaging will be used to confirm the presence and localization of metastasis.

The scVEGF dose amount and schedule will be determined by pharmacokinetic studies in Aim 1. During the course of

treatment, we will periodically use non-invasive imaging to monitor VEGFR2 and αvβ3 integrin expression with fluorescent or

radiolabeled proteins as described in Aim 1, to determine if VEGFR2 and αvβ3 integrin expression levels change over time. In

addition, we will use 18F-labeled 2-fluoro-2-deoxy-D-glucose (FDG), in collaboration with Prof. Eyal Mishani, at the cyclotron

radiochemistry unit at Hadassah Medical Center in Jerusalem, to monitor tumor metabolism over the course of treatment, as

has been done previously (51). FDG is a sugar that is taken up and retained in tissues with high metabolic activity, such as

malignant tumors. To move our work towards clinical trials, it will be important to monitor therapeutic response using noninvasive

imaging techniques. For early therapy monitoring, a decrease in 18F-FDG inhibition was detected several days after

treatment of U87MG tumors with a VEGFR2-targeting monoclonal antibody, where changes in tumor volume were not seen

until one week after treatment (52). Moreover, the use of 18F-FDG will allow assessment of the internal metabolic

characteristics of a tumor, which cannot be measured by calipers.

Immunohistochemistry will be performed on excised tumor sections to measure: (i) angiogenesis by quantification of von-

Willebrand factor and CD34 to measure total capillary density and newly formed microvessel tubes, respectively, (ii) apoptosis

(using TUNEL staining), (iii) VEGFR2 and αvβ3 integrin expression, and (iv) cell proliferation, using the Ki67 antigen, which is

a prototypic cell cycle related nuclear protein expressed by proliferating cells in all phases of the active cell cycle (G1, S, G2

and M phase), and co-localization with CD31 expression, an endothelial cell marker. Potential problems and alternatives:

We have extensive experience with non-invasive molecular imaging in animal models, and in primary and secondary

carcinoma models with tumor cells constitutively expressing either luciferase or red fluorescence protein (53). Here, we chose

to use bioluminescence imaging rather than fluorescence imaging, given the higher sensitivity of the former, especially for

detecting metastases, and lower background signals due to tissue autofluorescence. However, as an alternative we also have

experience with using fluorescence imaging to follow tumor growth and metastases in response to therapeutic treatment using

DsRed-transfected cells (53). Moreover, as an alternative to non-invasive metabolic assessment of tumors with 18F-FDG, 3-

deoxy-3-18F-fluorothymidine (18F-FLT) could be used to monitor tumor proliferation in response to scVEGF protein treatment

(54).

If we do not observe measurable biological effects on tumors in pilot experiments or if the effects are small, we will explore

other strategies, in which we administer the scVEGF proteins conjugated to cancer drugs, such as Doxorubicin (55) or taxol

(56). Alternatively, engineered scVEGF proteins could be tested in other tumor systems, such as MDA-MB-231 breast

carcinoma or U87MG glioblastoma models, which have also been used to study the effects of VEGFR2 and αvβ3 integrin

receptor targeting proteins (57), as variable effects have been seen with protein-based therapeutics amongst different animal

models (58).

6

Aim 3: Engineer bispecific scVEGF proteins that bind with high affinity to both VEGFR2 and PSMA. We will extend the

scVEGF engineering platform that we have developed beyond integrin-binding agents to engineer scVEGF mutants (Fig. 1C)

that bind with high affinity to another highly important tumor vasculature target, namely PSMA. In a previous study, we showed

that substitutions of amino acid sequence composition and length within scVEGF loop 2 (Fig 3, orange) and loop 3 (Fig 3,

green) were better tolerated than mutations in scVEGF loop 1 (Fig 3, purple). In Aim 3A, we will adopt several mutagenesis

approaches to introduce diversity into scVEGF loops 2 and 3 and different library

screening methods to obtain variants that bind to the desired PSMA target. In

these approaches, scVEGF loop 2 or loop 3, or both, will be mutated in a

randomized or biased manner to create combinatorial libraries that will be

displayed on the yeast cell surface for high-throughput screening against PSMA.

Aim 3A.1 Naïve libraries with randomized scVEGF loops 2 and 3: (i) Naive

libraries will be created in which scVEGF loops 2 and 3 are individually

substituted with randomized loops of 6 to 10 amino acids. With this approach,

once mutants that bind their target through scVEGF loop 2 or loop 3 are obtained,

the other loop can be randomized and screened for variants with increased

binding affinity. (ii) Naïve libraries will be created in which scVEGF loops 2 and 3

are both substituted with randomized loops of 6 to 10 amino acids. This approach

will allow co-evolution of two scVEGF loops that bind to the desired target.

Aim 3A.2: Biased libraries through introduction of epitopes previously shown

to bind the desired target. Biased libraries for PSMA targeting will be created by

introducing previously identified peptide epitopes into scVEGF loops 2 or 3 and

randomizing the flanking residues. This approach is similar to the strategy we

used for engineering high-affinity integrin-binding scVEGF proteins.

The His-His (HH) epitope, previously identified from both linear and cyclic phage display libraries to bind to PSMA with

micromolar affinity (26), will be introduced into scVEGF loops 2 or 3, or both, as a starting point for scVEGF engineering

studies (Fig. 4). Cyclic peptides like these, which are linked through a direct disulfide bond, have many conformational degrees

of freedom, often limiting their ability to bind a desired target with high affinity. Incorporating cyclic peptides into a scVEGF

scaffold will provide a format for obtaining peptides with high affinity target binding by introducing conformational constraints

and allowing additional mutations and contact residues to be introduced in other areas of the scVEGF protein. This approach is

possible because the loops of a scVEGF protein are not connected through a single direct disulfide bonds, but instead

anchored to the protein backbone.

Libraries will be created in which scVEGF loops 2 and 3 are individually substituted with the HH epitope (Fig. 4). The

residues flanking these motifs will be randomized using degenerate oligonucleotides. These libraries will be displayed on the

yeast cell surface and will be screened to identify scVEGF mutants that bind to PSMA. After

these mutants are identified, further engineering could be performed to increase their

binding affinity to target, if necessary. This will be performed by randomizing either loop 2 or

3 (the loop that was not mutated in the original clones), or by co-evolving scVEGF loops 2

and 3 by creating libraries in which one loop contains the HH motif and the second loop

contains completely randomized residues. In Aim 3B, we will use several complementary

high-throughput methods to screen yeast-displayed libraries to identify mutants that bind to

VEGFR2 and PSMA. Libraries will initially be screened by magnetic activated cell sorting

(MACS), using magnetic beads coated with PSMA. This method will allow isolation, through

avidity effects of rare binders with weak (micromolar) affinity, which are expected to be

present in naïve libraries. If necessary, further rounds of mutagenesis will be performed on

these initial hits and/or subsequent rounds of library screening will be performed using flow

cytometric sorting to identify mutants that bind to both fluorescently labeled VEGFR2 and

PSMA; this method will allow isolation of higher affinity binding clones, which will retain binding to soluble VEGFR2 and PSMA

after the wash steps needed for sample preparation and analysis. In Aim 3C, engineered scVEGF proteins with the highest

affinity for their targets will be recombinantly expressed in P. pastoris. We will include a C-terminal hexahistidine tag as a

handle for protein purification and for detection of binding to tumor cell surface receptors. In Aim 3D, we will measure the

binding of these scVEGF proteins to appropriate (VEGFR2 and PSMA expressing) tumor cell lines using flow cytometry.

Potential problems and alternatives: Aims 3A-B: The theoretical diversity required to exhaustively sample a randomized

loop of 7 amino acids is 207, a value above the 107–108 clones that we can efficiently screen with yeast surface display

methods. Therefore, it is likely that several rounds of mutagenesis and screening will be required to obtain scVEGF proteins

with low nanomolar binding affinity, as we and others have observed. Error-prone PCR and/or synthetic recombination will be

performed on the pool of mutants isolated from the first round of directed evolution to increase library diversity and recombine

favorable mutations, and screening will be performed as outlined above. Aim 3C: We have produced many scVEGF-based

proteins recombinantly in P. pastoris at yields of over 10 mg/L (6). Non-reduced SDS-PAGE analysis of these proteins

indicated that they were 80% monomeric after Ni-NTA purification, and they could be further purified by gel filtration

chromatography. Aim 3D: Based on our previous studies, we do not expect the C-terminal His tag to interfere with target

binding (6); however, we will confirm this by performing direct binding assays using recombinant purified scVEGF proteins that

are labeled with fluorescein at their N-terminus, which can then be used in competitive binding assays if necessary.

XXHHXXXX

XXXHHXXX PSMA

XXXXHHXX

Fig. 4: Epitopes to be

substituted into scVEGF loop 2

or 3 as a starting point for

scVEGF engineering studies.

X= any possible amino acid.

Longer loops could also be

explored if necessary.

Figure 3. Structure of wild-type VEGF

homodimer (pdb: 2VPF). VEGF Chain

1: dark blue; VEGF Chain 2: light blue.

Loop 1: purple; Loop 2: orange; Loop

3: green. Point mutations introduced

into scmVEGF are indicated in red.

7

The specific outcomes of the research will be: (i) evaluation of the therapeutic efficacy of engineered high-affinity

scVEGF proteins in pre-clinical models, an important step on the path to clinical translation; (ii) identification of new classes of

engineered scVEGF proteins against other important tumor vasculature receptors for future testing as molecular imaging

agents and in tumor targeting applications; and (iii) an improved understanding of the VEGFR/integrin/PSMA system, which

will provide the basis for the rational and combinatorial engineering of multiple ligand-receptor interactions.