University of Southern California Dissertations and Theses

Developing peptide and antibody-mimetic ligands for the cell surface receptors β2AR and DC-SIGN

(USC Thesis Other)

Developing peptide and antibody-mimetic ligands for the cell surface receptors β2AR and DC-SIGN

PDF

Download a page range

Download transcript

Copy asset link

Request this asset

Transcript (if available)

Content

DEVELOPING PEPTIDE AND ANTIBODY-MIMETIC LIGANDS FOR
THE CELL SURFACE RECEPTOR !
2
AR AND DC-SIGN

by

Kuo-Chan Hung

A Dissertation Presented to the
FACULTY OF THE USC GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirement for the Degree
DOCTOR OF PHILOSOPHY
(GENETIC, MOLECULAR & CELLULAR BIOLOGY)

August 2012

Copyright 2012 Kuo-Chan Hung
ii
Acknowledgements
“Life will find a way” – The Lost World: Jurassic Park (1997)
Although fifteen years have passed, this phrase still reverberates in my
mind. It has remained especially poignant during the tougher times. Traveling
abroad to pursue my dream to become a scientist has been just like a selection
process. I’ve always believed that I would always find a way and, just as with
evolution, I would develop into a rigorous scientist fit to join the scientific
community. Evolution is exactly what I learned from my Ph.D. experience. If I
have “evolved” into a more independent and collaborative person in my research
or personal life, this is due to the generous guidance of my advisor, Professor
Richard Roberts.
I am indebted to him for his patience as well as the encouragement and
motivation he has instilled in me. This has given me the strength to overcome the
tougher challenges I have faced. The freedom he has allowed me as well his
constructive criticism has made me into a more independently minded scientist.
Professor Roberts made possible the great opportunity to collaborate with
Professor Brian Kobilka and Professor Pin Wang’s outstanding research groups.
Three colleagues have proved instrumental to my research. Dr. Soren G. F.
Rasmussen and Dr. Cheng Zhang in the Kobilka Laboratory who kindly provided
me with the !
2
AR proteins and all the necessary information. Liang Xiao of the
iii
Wang Laboratory prepared DC-SIGN protein and performed cell functional
assay. Collaborating with them has greatly broadened my scientific perspective.
I would also like to thank my colleagues for providing a stimulating
learning environment — particularly Dr. Anders Olson, Dr. Ryan Austin, and Dr.
Terry Takahashi. Dr. Olson and Dr. Austin helped me get started in the
laboratory by offering valuable advice. Dr. Takahashi consistently helped me
solve problems by pointing me in the right direction. I am grateful for the counsel
given by my distinguished advisory committee, including Professors Ite Laird-
Offringa, Donald Arnold, and Janos Peti-Peterdi.
Finally, I am grateful to my parents for their unconditional support and
belief in me during these past nine years. Their love and encouragement has
given me the strength to come this far. Without them my dreams would have
certainly remained out of reach. I hope that my Ph.D. degree will honor their
steadfast faith in me.

iv
Table of Contents
Acknowledgements............................................................................................... ii ii
List of Figures........................................................................................................v v
Abstract............................................................................................................... vii vii
Chapter 1: Introduction and Overview...................................................................1
Chapter 2: Developing Antibody-Mimetic Ligands Targeting Different Activity
States of !
2
AR Using mRNA Display.................................................22
Abstract ........................................................................................................22
Introduction...................................................................................................24
Materials and Methods .................................................................................28
Results..........................................................................................................40
Discussion ....................................................................................................60

Chapter 3: Developing G"s-Like Peptide Ligands Targeting Active !
2
AR Using
mRNA Display ...................................................................................69
Abstract ........................................................................................................69
Introduction...................................................................................................71
Materials and Methods .................................................................................77
Results..........................................................................................................87
Discussion ..................................................................................................102

Chapter 4: Novel Vaccines via Dual-Specific Antibody-Mimetics Against
Mouse and Human DC-SIGNs.........................................................110
Abstract ......................................................................................................110
Introduction.................................................................................................112
Materials and Methods ...............................................................................121
Results........................................................................................................132
Discussion ..................................................................................................152

Bibliography ......................................................................................................161

v
List of Figures
Figure 1.1: mRNA display ................................................................................4

Figure 1.2: Comparison of the solution structures of the Llama VHH domain
and wild-type human 10FnIII domain (wt 10FnIII), and the
sequence of wt 10FnIII...................................................................7

Figure 1.3: Drug efficacy for GPCRs and the structures of active and
inactive !
2
AR................................................................................13

Figure 1.4: The intracellular processing of antigen bound to DC-SIGN
receptor and the domains of human and mouse DC-SIGN..........18

Figure 2.1: Sequence, construction, and possible structure of e10FnIII
library ...........................................................................................29

Figure 2.2: Structure and immbolization of !
2
adrenergic receptor (!
2
AR).....32

Figure 2.3: Different selection methods..........................................................35

Figure 2.4: THRX-!
2
AR (active) and ICI-!
2
AR (inactive) selections...............43

Figure 2.5: Sequences and clone bindings of THRX pool 8 and 10...............45

Figure 2.6: The capturing method reduces background binding to M1
beads ...........................................................................................47

Figure 2.7: cDNA/mRNA is responsible for non-specific electrostatic
interaction that can be blocked by MgCl
2
.....................................50

Figure 2.8: Monobromobimane on !
2
AR contributes to non-specific
electrostatic interaction, and MgCl
2
and dsDNA can block
electrostatic interaction.................................................................52

Figure 2.9: BI-!
2
AR pool binding and cation exchange..................................54

Figure 2.10: BI-!
2
AR binding of selected clones..............................................56

Figure 2.11: Fn BI 9-5 is not functional ............................................................59

Figure 2.12: Possible reasons for the electrostatic interaction and the
background binding to M1 beads .................................................66
vi

Figure 3.1: !
2
AR structure and C-termini of G" subunits...............................72
Figure 3.2: Design of a doped peptide library based on the C-terminus of
G"s subunit and the selection against active BI-bound !
2
AR ......79

Figure 3.3: Pool binding of mRNA display against BI-!
2
AR and the
sequencing results of pool 4 and 5...............................................90

Figure 3.4: The C-terminal FLAG issue and solution .....................................94

Figure 3.5: Pool binding using different formats of fusions.............................97

Figure 3.6: Sequening results of pool 6 and the pool 6 clone binding..........100

Figure 3.7: Function of peptide ligands in stabilizing agonist-bound !
2
AR...103

Figure 4.1: Domains and the alignment of amino acid sequences of human
DC-SIGN (hDC-SIGN) and mouse DC-SIGN (mDC-SIGN)........116

Figure 4.2: Selection scheme.......................................................................118

Figure 4.3: Results of mRNA display targeting mDC-SIGN..........................133

Figure 4.4: Results of mRNA display against hDC-SIGN.............................135

Figure 4.5: Binding and internalization of mouse clones in mDC-SIGN
expressing 293T cells.................................................................137

Figure 4.6: Binding of dual-specific ligands to hDC-SIGN expressing 293T
cells............................................................................................140

Figure 4.7: Internalization of M3H4-18-HA into hDC-SIGN expressing 293T
cells............................................................................................144

Figure 4.8: Binding and internalization of M3H4-18-HA in human
peripheral blood mononuclear cells (hPBMCs)-derived
immature dendritic cells (iDCs)...................................................147

Figure 4.9: Antigen-based immunity induced by M3H4-18-antigen..............151

vii
Abstract
mRNA display is an in vitro selection technique that can evolve novel ligands to
modulating protein-protein interactions and regulate crucial biological functions.
Previous efforts have focused on soluble protein targets. However, here we use mRNA
display to target cell surface receptors that represent 60% of drug targets but remain the
most challenging targets. Two cell surface receptors, beta-2 adrenergic receptor (!
2
AR)
and Dendritic Cell-Specific ICAM-3-Grabbing Non-integrin (DC-SIGN), were chosen as
our targets in order to develop novel ligands for structural studies, drug development, or
vaccine design. mRNA display using a doped G"s C-terminal peptide library was
capable of targeting the active state of !
2
AR, resulting in active state-specific peptide
ligands with function similar to Gs protein. Selections against both mouse and human
DC-SIGN by mRNA display using an antibody-mimetic library resulted in ligands with
dual specificity to both mouse and human DC-SIGN. One selected DC-SIGN specific
ligand could induce antigen-specific immune responses in human dendritic cells and has
the potential for developing DC-based cancer vaccines. In these proof of concept studies,
we demonstrated a general approach for the development of novel functional ligands
capable of targeting cell surface receptor.
1
Chapter 1
Introduction and Overview
In vitro display techniques
In vitro display techniques are selection-based technologies for isolating peptides
or proteins with high affinity and specificity to virtually any biopolymer target. These
ligands can serve as powerful tools to study protein-protein interactions, regulate
biological processes, and for novel therapeutic applications. The basic principle of in
vitro display is molecular selection analogous to Darwinian natural selection. A diverse
population of displayed molecules (a pool) is exposed to a selective pressure, which is
coupled to reproduction (molecules retained on target are amplified), resulting in the
fittest individuals dominating the population after successive generations of selection.
Polypeptide molecular libraries present a unique selection challenge in that these
molecules must be linked or “displayed” on an amplifiable unit, because polypeptide
cannot be replicated and reproduced as oligonucleotide can. Early in vitro display
techniques utilized biological carriers such as phage (Smith, 1985), yeast (Boder and
Wittrup, 1997), bacteria (Fuchs et al., 1991), and eukaryotic virus (Kasahara et al., 1994)
to display such peptide libraries. The library complexity of these cellular-linkage display
methods is, however, limited by transformation efficiency into a cellular host and are
generally limited to <10
10
molecules. As library diversity and complexity increases, so
does the probability to access high-affinity ligands for a target (Ling, 2003; Perelson and
Oster, 1979). Subsequently ribosome display (Hanes and Pluckthun, 1997; Mattheakis et
al., 1994), mRNA display (Roberts and Szostak, 1997), DNA-polypeptide fusion display
2
(CIS display and CAD display) (Odegrip et al., 2004; Reiersen et al., 2005), and in vitro
compartmentalization (IVC) (Tawfik and Griffiths, 1998) techniques were invented to
increase the library sizes (10
12
to 10
15
) by directly coupling an individual genotype to its
displayed protein in a cell-free environment (acellular-linkage).
mRNA display achieves this coupling through covalent fusion of a translated
polypeptide to its own coding mRNA via a chemically ligated puromycin moiety,
generating a highly stable mRNA-peptide fusion (Takahashi et al., 2003). Puromycin is a
chemically stable small molecule and structural analog of aminoacyl-tRNA. The stability
and chemical simplicity of mRNA display have made it a versatile technology and
successful selections have been reported using naïve randomized peptides and proteins
(Barrick et al., 2001; Ja and Roberts, 2004; Olson et al., 2008; Xia et al., 2003; Xu et al.,
2002) as well as libraries incorporating unnatural amino acid chemistries (Frankel et al.,
2003) and post-translational peptide cyclization (Millward et al., 2007; Millward et al.,
2005). While previous efforts have focused on RNA and globular protein targets, we
reasoned that mRNA display could be used for selections against recalcitrant cell surface
receptor targets, which require detergent-rich and more stringent selection conditions.

mRNA display and antibody-mimetic library based on the 10th fibronectin type III
domain of human fibronectin (10FnIII)
mRNA display is an in vitro selection technique which enables peptide and
protein design by directed molecular evolution. In order for this approach to work, an
mRNA bearing a pendant 3’ puromycin is fused with its encoded peptide or protein to
3
generate an mRNA-peptide or mRNA-protein fusions (Figure 1.1A). Directed molecular
evolution is achieved by several in vitro selection cycles (Figure 1.1B). An initial library
of double-stranded DNA templates is created using PCR. The library diversity can be
tuned by the incorporation of randomized DNA cassettes (e.g. NNS codons; N = A, T, C,
G; S = C, G) (Barrick et al., 2001; Ja and Roberts, 2004; Olson et al., 2008; Xu et al.,
2002), doped cassettes (Austin et al., 2008; Ja et al., 2006), mutagenic PCR (Olson et al.,
2008; Tsang and Joyce, 1996), or a combination of these. mRNA is synthesized by in
vitro transcription of a dsDNA library template, and is then ligated to a synthetic DNA
linker containing a 3’ puromycin. The ligation product is translated in vitro using
reticulocyte lysate to generate mRNA-peptide or mRNA-protein fusions in which each
peptide or protein is covalently attached to its own coding mRNA. The fusion is
subjected to reverse transcription to mitigate RNA secondary structure (Barrick et al.,
2001). cDNA/mRNA-peptide or cDNA/mRNA-protein fusions are selected by panning
the library over the target. Molecules with target affinity are specifically eluted and
amplified via PCR (Keefe, 2001; Takahashi and Roberts, 2009).
mRNA display has several important advantages relative to either cellular-linkage
displays or acellular-linkage displays. In comparison with cellular-linkage displays (e.g.
phage display), mRNA display provides the ability to scrutinize very large libraries
(>10
13
unique sequences) in the absence of a living cell, and tight experimental control
over binding and stringency (Takahashi et al., 2003). mRNA display is also monovalent
and no expression bias. In contrast to other accellular-linkage displays, mRNA-peptide
fusions present a simpler, more stable display platform. Ribosome display is mediated via
4

Figure 1.1 mRNA display
(A) The library mRNA (black) is translated in vitro to generate peptide or protein.
At the end of in vitro translation, the puromycin (blue) linked to RNA via a DNA
linker (dark gray) is fused to the last amino acid. Puromycin is used as a bridge
to form an mRNA-peptide or protein fusion by covalently attaching an mRNA to
its newly synthesized peptide or protein. (B) The experimental steps for one
round of mRNA display. ssDNA library is chemical synthesized and is then PCR
amplifed to generate dsDNA library. dsDNA library is transcribed into mRNA,
ligated to a DNA linker-puromycin, and translated in vitro. The fusion library is
reverse transcribed, FLAG purified, and pre-cleared before interacting with
target. Immobilized target is used to affinity purify target binders from the non-
functional members of the library. Affinity purified fusion is then enriched by
PCR, and the dsDNA products are subjected to the next round of selection.

5
association with the ribosome, which could present a significant steric hindrance and
requires non-physiological selection conditions to maintain the stability of the display
(Diaz et al., 2003; Lipovsek and Pluckthun, 2004). Coupling of a DNA binding protein
with its encoding DNA (cis-activity) in CIS display and CAD display also leads to steric
hindrance and possible protein-folding issues (RepA is 33kDa and P2A is 86kDa)
(Ullman et al., 2011). In vitro compartmentalization (IVC) utilizes water-in-oil emulsions
that could be disturbed by detergents used to solubilize and stabilize cell surface receptor
(Griffiths and Tawfik, 2006). The library complexity of mRNA display is arguably
unmatched. For instance, a nominal 10% mRNA-peptide fusion efficiency results in
>10
12
unique library molecules in a 2.5 mL in vitro translation volume. Thus, mRNA
display is the best strategy to develop specific, high-affinity ligands that bind cell surface
receptors.
Previous mRNA display libraries can be divided into two basic types: peptide and
protein libraries. Doped or randomized linear peptide libraries can be further modified for
versatile functions by conjugating with drugs, incorporating unnatural amino acids by
suppressor tRNA, and post-translational cyclization. Thus, peptide libraries can be further
divided to four subtypes: 1) linear peptide libraries (Austin et al., 2008; Barrick et al.,
2001; Ja et al., 2005; Ja and Roberts, 2004; Ja et al., 2007; Ja et al., 2006; Xia et al.,
2003), 2) peptide-drug conjugate libraries (Li and Roberts, 2003), 3) peptide libraries
containing unnatural amino acids (Frankel et al., 2003; Frankel and Roberts, 2003; Li et
al., 2002), and 4) cyclic peptide libraries (Millward et al., 2007; Millward et al., 2005).

6
Protein libraries consists of random (Keefe and Szostak, 2001), patterned (Wilson
et al., 2001), and scaffold-based (Xu et al., 2002) designs. Randomized protein libraries
are linear 109mer proteins containing a randomized region of 80 amino acids. Patterned
protein libraries comprise distinct short 11a.a. cassettes with either polar or nonpolar
amino acids that are patterned to form amphipathic !-helix or "-strands within an 88
residue randomized region. Rather than flexible unstructured libraries, we are particularly
interested in scaffold-based protein libraries, due to their superior and promise for the
generation of stable protein domains with high affinity (K
D
= 10
-9
– 10
-12
) to targets
(Boder et al., 2000; Koide et al., 1998).
There are a variety of protein scaffolds exploited for the construction of
combinatorial libraries, including antibody-derived (Fab, scFvs, and single variable
domain/V-domain) (Caravella and Lugovskoy, 2010; Gill and Damle, 2006) and novel
non-immunoglobin-based (Adnectin/monobody, Affibody, anticalin, knottin/microprotein,
and designed ankyrin repeat protein/DARPin) (Hosse et al., 2006; Lofblom et al., 2011).
Non-immunoglobin-based scaffolds have been used for selection of specific, high-affinity
proteins that serve as promising alternatives to antibodies in therapeutic applications (Binz
et al., 2005; Gebauer and Skerra, 2009; Nuttall and Walsh, 2008).
Among these scaffolds, the human fibronectin-based 10FnIII has several
advantages for selections. The 10FnIII library is an antibody mimetic as the !-fold and
binding loops of 10FnIII structurally resembles an immunoglobulin (Ig) VHH fold with its
complementarity determining regions (CDRs) (Figure 1.2) (Dickinson et al., 1994). The
10FnIII scaffold presents a better alternative than other scaffolds for antibody. The 10FnIII
7

Figure 1.2 Comparison of the solution structures of the Llama VHH
domain and wild-type human 10FnIII domain (wt 10FnIII), and the
sequence of wt 10FnIII
(A) The solution structures of Llama VHH (Left, PDB 1G9E) and wt 10FnIII
(Right, PDB 1TTG) both have an immunoglobulin fold. The disulfide bond
between Cys 22 and Cys 96 of Llama VHH is shown in yellow and three
complementarity determining regions of Llama VHH are color-coded (CDR-H1,
red; CDR-H2, green; CDR-H3, blue). Three loops in wt 10FnIII are labeled the
same color as their structurally corresponding loops in Llama VHH (BC, red; DE,
green; FG, blue). The three flexible loops are important for the molecular
recognition and specificity. (B) The amino-acid sequence of wt 10FnIII is shown
here. Secondary structures are indicated (7 beta sheets and 3 loops) and the
three loops are color-coded corresponding to the tertiary structure.

8
scaffold maintains similar structure as an antibody while containing beneficial features
such as a small size, lack of disulfide bonds, high thermostability, and excellent bacterial
expression properties (Bloom and Calabro, 2009). Its small size (94-amino acid) and
lacking disulfide bonds enable the 10FnIII scaffold to be used inside cells as an intrabody
(Chen et al., 1994). High thermostability (T
m
~88
o
C) from its hydrophobic core of the !-
fold makes it a stable scaffold for library design (Batori et al., 2002; Cota et al., 2001;
Litvinovich and Ingham, 1995). High levels of expression in bacteria also make protein
production more economical. Lastly, the 10FnIII scaffold originates from human
fibronectin, minimizing the potential for immunogenicity when used in humans.
Koide et al. developed the first 10FnIII scaffold using a phage display library with
10
8
unique members and with 10 randomized residues; 5 in the BC loop and 5 in the FG
loop (Koide et al., 1998). This library likely contains too few randomized residues for an
adequate interaction surface, since their experiments only resulted in low-affinity ubiquitin
binders. The second generation of the 10FnIII-based library achieved higher diversities of
10
12
unique molecules that contained 21 randomized residues (7 in BC; 4 in DE; 10 in
FG). This library was used in an mRNA display selection and resulted in a picomolar
ligand (Xu et al., 2002). However, the 21 randomized positions increase the possibility of
stop codons and may destabilize the scaffold by altering structurally critical DE loop
(Getmanova et al., 2006; Parker et al., 2005). The third generation of the 10FnIII scaffold
has been optimized for maximization of library diversity (~30 trillion unique sequences),
minimization of stop codons and frame-shifts, and maintenance of stability (Olson and
Roberts, 2007). In this library, 17 positions are randomized over the BC (7 residues) and
9
FG (10 residues) loops. The unstructured N terminus was also removed because it may
interfere with binding in this scaffold (Getmanova et al., 2006; Main et al., 1992). This
10FnIII-based library has been characterized for expression and folding by a GFP
folding/expression reporter assay (Waldo et al., 1999). Several members of the library
have also been characterized by guanidinium chloride denaturation (Santoro and Bolen,
1988), resulting in chemical stabilities within 2-3 kcal of the wild type 10FnIII
["G(wt)
unfolding
= 7.7 kcal/mol]. Finally, the sequencing of this library confirmed that the
codon usage and amino acid representation is similar to well-expressed proteins. This third
generation of 10FnIII-based library has been successfully utilized in mRNA display to
select for phospho-I#B$ and SARS N protein (Liao et al., 2009; Olson et al., 2008).

Cell surface receptors
A crucial property of living cells is the ability to receive the signals from their
environment and promptly respond by changing gene expression. Signal transduction
involves extracelluar signaling molecules interacting with cell surface receptors, which
relay the signals to the downstream molecules that elicit a physiological response
(Rodbell, 1980). A variety of cell surface receptors embedded in the cell membrane
trigger specific cellular responses to distinct signaling molecules. In general, cell surface
receptors contain extracellular, transmembrane, and intracellular domains. Extracellular
domains protrude outside the cell membrane and receive signals by interacting with
signaling molecules. Transmembrane domains are embeded in the membrane mostly by
their $ helices (White, 2009). Intracellular domains translate signals to intracellular
10
responses by physical or chemical modulation of downstream molecules. Analysis of the
genes in human genome suggests that 27% of the human proteome are $-helical
transmembrane proteins. 23% of $-helical transmembrane proteins are cell surface
receptors that compose largest functional group of transmembrane proteins (Almen et al.,
2009). Cell surface receptors account for 60% of drug targets, indicating their important
role in regulation of essential physiological functions (Overington et al., 2006). However,
membrane protein structures only occupy ~2% of crystal structures deposited in the
protein data bank (Arinaminpathy et al., 2009). The difficulties in working with cell
surface receptors are poor expression, insolubility and instability in aqueous solution, and
conformational and functional heterogeneity (Seddon et al., 2004).
These difficulties can be partially solved with several recent advances in
expression and purification of functional receptors (Grisshammer, 2009), improvement of
solubility and stability in aqueous solution by detergents (Seddon et al., 2004), and
increase of homogeneity with antibody binding (Hunte and Michel, 2002). However, a
general systematic method to directly target cell surface receptors to design functional
ligands for academic research or therapeutic drugs is still very challenging. Although
many efforts have been made to target cell surface receptors using phage display, there
are still limitations. In phage display, targets had been domains of receptor or full-length
receptors over-expressed in cells, but not full-length receptors dissolved in detergent-rich
buffer (Molek et al., 2011). There are several advantages in using mRNA display to
directly target full-length cell surface receptors: 1) no structural information is needed, 2)
no cell transformation is required, and 3) the detergent-rich conditions for receptors can
11
be accommodated. Beta-2 adrenergic receptor (!
2
AR) and dendritic cell-specific ICAM-3
grabbing non-integrin (DC-SIGN) were chosen as targets for our mRNA display
selections targeting cell surface receptors. !
2
AR is a member of G protein-coupled
receptors (GPCRs) that compose the largest family of cell surface receptors and are
targeted by 27% of FDA-approved drugs (Almen et al., 2009; Overington et al., 2006).
We would like to develop a general strategy to evolve ligands targeting GPCRs for
structural studies and drug development using !
2
AR as a model system. DC-SIGN is a C-
type lectin receptor, which is mainly expressed on dentritic cells (DCs) and is responsible
for antigen-specific immune responses (Geijtenbeek et al., 2000). We aim to develop
dual-specific ligands capable of binding both mouse and human DC-SIGN to develop
novel cancer vaccines.

G protein-coupled receptor and beta-2 adrenergic receptor
G protein-coupled receptors (GPCRs) constitute the largest family of cell surface
receptors (Almen et al., 2009). GPCRs are responsible for important physiological
functions such as vision, olfaction, and taste. They transmit information from hormones
and neurotransmitters into cellular responses. GPCRs share a common structural feature
and are composed of seven transmembrane $-helices and span through the membrane to
form extracellular, transmembrane, and intracellular parts (Grigorieff et al., 1996;
Henderson and Unwin, 1975; Nathans and Hogness, 1983; Pebay-Peyroula et al., 1997).
The extracellular part starts with an N-terminal peptide followed by three loops that
separate the transmembrane helices. The transmembrane part comprises seven $-helices
12
that are linked one by one with six loops. The intracellular part consists of three loops to
separate helices and ends up with a C-terminal peptide. Functionally, the extracellular
part interacts with signaling molecules, enabling the conformational change of GPCR.
This conformational change transfers the signal through transmembrane $-helices to
intracellular part, which activates the coupled heterotrimeric G protein by inducing
exchange of GDP to GTP coupled with G protein. The activated G protein dissociates
from GPCR to modulate downstream effector proteins, ultimately leading to cellular
response. In vertebrates, GPCRs are divided into five classes based on their sequence
homology and structural similarity, Class A (rhodopsin), Class B (secretin), Class C
(glutamate), adhesion, and Frizzled/Taste2 (Pierce et al., 2002; Rosenbaum et al., 2009).
Due to their important role in regulating essential physiological functions, GPCRs
are by far the most prominent targets for drug development (Overington et al., 2006).
Thus, developing ligands to bind and modulate GPCRs with different efficacies is a
major focus in drug discovery. Natural and synthetic ligands have specific efficacies that
are defined by the effect of a ligand on the structural and functional properties of a
receptor. Based on the efficacy, ligands of GPCRs can be grouped to 4 classes, including
full agonist, partial agonist, neutral antagonist, and inverse agonist (Figure 1.3)
(Rosenbaum et al., 2009). Full agonists are able to maximize the biological response of a
receptor, but partial agonists never reach the full activity of a receptor even at saturating
concentration. Neutral antagonists only keep the basal activity of a receptor by preventing
other ligands from interacting with a receptor. Inverse agonists inhibit the basal activity
of a receptor and reduce the biological responses below unliganded receptors. The wide
13

Figure 1.3 Drug efficacy for GPCRs and the structures of active and
inactive !
2
AR
(A) Different ligands have different efficacies for GPCRs. Full and partial
agonists activate GPCRs to increase their biological responses above the basal
activity. Neutral antagonists keep GPCRs in native state with only basal
activities. Inverse agonists inactivate GPCRs and inhibit the basal activities.
[Diagram courtesy from D. M. Rosenbaum, S. G. F. Rasmussen, B. K. Kobilka,
Nature 459, 356-363 (2009)] (B) The active x-ray crystal structure of the full
agonist (BI-167107)-bound !
2
AR (PDB 3SN6) and inactive x-ray crystal
structure of inverse agonist (Carazolol)-bound !
2
AR (PDB 2RH1) are shown
here. On the top, the full agonist (BI-167107; indigo) binds to !
2
AR (green) and
the binding increases biological response of !
2
AR. On the bottom, the basal
activity of !
2
AR (blue) was inhibited with inverse agonist (Carazolol; red)
binding. Different ligands binding to !
2
ARs lead to the different conformational
changes, which result in different activities of !
2
ARs.

14
range of ligand efficacies indicates that different ligands might induce and stabilize
distinct conformations of GPCRs to elicit the various biological activities (Kobilka and
Deupi, 2007). Indeed, biophysical studies of fluorescently labeled !
2
AR interacting with
various ligands demonstrated that partial and full agonists stabilize distinct
conformational states, implying that GPCRs are no longer thought to simply switch
conformation between active and inactive states. Instead of two-state switches, there are
an almost infinite variety of conformations between these two extreme states and specific
ligands can stabilize particular sets of conformations to achieve varying efficacies for
different signaling pathways (Yao et al., 2006). This concept leads to our strategy for
ligand design – ligands can be developed or screened to induce or target certain
conformation of a GPCR to elicit its specific biological activity.
Among 900 members of GPCRs, beta-2 adrenergic receptor (!
2
AR) is an
excellent model system for our studies due to its several important features. First, !
2
AR is
a member of Class A, the largest and most diverse of GPCR classes. Thus, the study of
!
2
AR could provide a general understanding of the majority of GPCRs, and the methods
developed to study !
2
AR could also be applied for other GPCRs (Attwood and Findlay,
1994). Second, !
2
AR is the first human GPCR that was crystallized, suggesting that high-
quality functional !
2
AR can be routinely produced for selection. Third, because the
structure is known, detailed structural information is available for data interpretation
(Cherezov et al., 2007; Rasmussen et al., 2007; Rosenbaum et al., 2007). Fourth, several
clinically valuable agonists and inverse agonists for this receptor are known, and these
ligands can stabilize it in distinct conformations that can be used for selecting structural
15
and functional specific ligands (Yao et al., 2006). Fifth, !
2
AR is an important therapeutic
target for treatment of asthma, chronic obstructive pulmonary disease, premature labor,
childhood migraines, hypertension, and heart failure (Liggett, 2000).
To better understand the structural basis for different activities of !
2
AR and
modulate the functions of !
2
AR for therapeutic applications, mRNA display was applied
for the development of protein and peptide ligands targeting different states of !
2
AR. The
key to our strategy of designing ligands to !
2
AR is our ability to target full-length !
2
AR
in detergent-rich aqueous conditions. mRNA display is uniquely suited to this, because it
utilizes puromycin to link genotype to phenotype rather than cells in phage display,
emulsions in IVC, or large proteins in ribosome display and DNA display. Cells and
emulsions could be disrupted by detergents, and large proteins might have steric
hindrances or non-specific target binding to interfere affinity purification. mRNA display
also has the most chemically stable linkage between genotype and phenotype.
Furthermore, through collaboration with Professor Brian Kobilka, we can obtain high-
quality, functional !
2
AR.
First, we aimed to develop novel antibody-mimetic protein ligands to stabilize
active or inactive states of !
2
AR. These ligands could be useful to obtain high-resolution
structures of different states of !
2
AR as well as possibly manipulating !
2
AR function.
Previous work has demonstrated that antibodies are a powerful tool to solve x-ray crystal
structures of receptors by co-crystallizing and stabilizing their flexible structures (Day et
al., 2007; Hunte and Michel, 2002). We aim to show that mRNA display using human
16
10FnIII-scaffolded library can evolve antibody-mimetic ligands to serve the same
purpose as antibodies. This work is discussed in detail in Chapter 2.
We have also targeted the intracellular portion of activated !
2
AR. Using a doped
peptide library based on the C-terminus of G$s protein, the natural binding partner of
!
2
AR, we have evolved peptide ligands with G$s-like function. These ligands can
specifically modulate the !
2
AR-Gs protein interaction. This strategy has several
advantages. First, structural and functional studies of !
2
AR have shown that the
extracellular ligand-binding pocket has higher structural homology relative to
intracellular interface between the !
2
AR and G$s. Thus, targeting this intracellular G
protein-binding surface of a GPCR may lead to ligands with higher specificity for that
particular GPCR (Rasmussen et al., 2011b). Second, the extreme C-termini of G protein
$ subunits (G$) have been shown to bind the cytosolic face of their target GPCRs. This
binding stimulates formation of active GPCR (Dratz et al., 1993; Hamm et al., 1988;
Rasenick et al., 1994). Thus, using a peptide library based on this C-terminal G$s peptide
could lead to peptide ligands with G$s-like function, and the ability to stabilize active
!
2
AR and block the wild type !
2
AR-Gs protein interaction. Third, small peptide ligands
containing G$s protein function could be beneficial in drug development, since peptides
can be easily modified for higher protease resistance and cell permeability (Covic et al.,
2002a; Millward et al., 2007). This selection is extensively covered in Chapter 3.

17
Dendritic cells and dendritic cell-specific intercellular adhesion molecule-3 (ICAM-
3)-Grabbing Non-integrin
Dendritic cells (DCs) are the most potent antigen-presenting cells (APCs) and
serve as crucial regulators in both the innate and the adaptive immunity (Clark et al.,
2000; Mellman and Steinman, 2001). DCs recognize, process, and present antigens to
naïve T cells via either the MHC class I pathway or the MHC class II pathway for
antigen-specific immune responses (Only the receptor-mediated endocytosis pathway is
shown in Figure 1.4A) (Banchereau and Steinman, 1998; Tacken et al., 2007). A
dendritic cell captures antigens through its cell surface receptors, resulting in endocytosis
of antigens. Endocytosed antigens either remain in the endosome or escape from the
endosome to the cytosol. Antigens in the cytosol are processed by the MHC I pathway.
Antigens are first digested by immunoproteasome to small peptides, which are
transported into endoplasmic reticulum (ER) by transporter associated with antigen
processing (TAP) and loaded onto MHC class I receptors in the ER. The antigenic
peptide-loaded MHC class I molecules are presented on the surface of DCs for the
activation of CD8
+
cells. Antigens in the endosome are processed through the MHC class
II pathway where the endosome fuses with protease-containing lysosomes, leading to the
degradation of antigens into small peptides. Antigenic peptides are loaded onto MHC
class II molecules in the MHC class II compartment (MIIC) and the complexes are
transported to cell surface to activate CD4
+
cells (Banchereau et al., 2000; Banchereau
and Steinman, 1998).

18

Figure 1.4 The intracellular processing of antigen bound to DC-SIGN
receptor and the domains of human and mouse DC-SIGN
(A) Schematic mechanism of DC-SIGN. DC-SIGN is a dendritic cell-specific
receptor that can internalize the antigen bound to it. The internalized antigen will
be further processed by two different cellular pathways, and is loaded onto MHC
class I or MHC class II receptors. The antigen presented via MHC class I will
lead to CD8
+
cell maturation and the antigen-MHC class II complex will cause
proliferation of CD4
+
cells. [Courtesy from Tacken, P.J. et al. Nat. Rev. Immunol.
7, 790-802 (2007)]. (B) The domains of human DC-SIGN (CD209) and mouse
DC-SIGN (CD209a) are shown. The extracellular domain contains CRD
(Carbohydrate recognition domain), Neck Repeats, TM (Transmembrane
domain), and cytoplasmic domain. Human DC-SIGN has more neck repeats
than mouse DC-SIGN. [Courtesy from Powlesland, A. S. et al. JBC. Vol.
281(29), 20440-20449 (2006)]
19
Due to their importance in modulating immunity, DCs have been harnessed to
develop more effective vaccines, especially for cancer immunotherapy (Banchereau and
Palucka, 2005; Rosenberg et al., 2004; Schuler et al., 2003). Two strategies are applied
for development of DC-based vaccines for cancer treatment: ex vivo loading (Buchsel and
DeMeyer, 2006; Figdor et al., 2004) and in vivo targeting (Tacken et al., 2007). In ex vivo
loading, autologous DCs are harvested from patients, loaded with tumor antigen ex vivo,
and readministered to the patients. ex vivo loading successfully resulted in Sipuleucel-T
(Provenge) – the first FDA-approved therapeutic cancer vaccine for prostate cancer
(Plosker, 2011). However, this procedure is extremely laborious and expensive where the
three-course treatment will cost $93,000 USD.
An alternative is to load the tumor antigens onto DCs in vivo by targeting DC
surface receptors that are capable to internalize antigens bound to them, leading to
antigen presentation on MHC class I or II molecules. Among these DC receptors, C-type
lectin receptors (CLRs) are the best targets for vaccine development due to their specific
expression on APCs (Figdor et al., 2002). Three CLRs, the mannose receptor (CD206),
DEC-205 (CD205), and dendritic cell-specific ICAM-3-grabbing non-integrin (DC-
SIGN; CD209), have been reported to load endocytosed antigens onto MHC class I or II
molecules, inducing effective CD8
+
or CD4
+
T-cell responses (Bonifaz et al., 2002;
Ramakrishna et al., 2004; Tacken et al., 2005). However, the mannose receptor and
CD205 are also expressed on other types of APCs in humans, which could lead to lower
targeting efficiencies and possible undesired side effects (Kato et al., 2006; Linehan,
20
2005). In humans, DC-SIGN has become the main focus for vaccine development,
because it is the most DC-specific receptor (Geijtenbeek et al., 2000).
Human DC-SIGN (hDC-SIGN) is a member of type II transmembrane CLR
family. The structure of hDC-SIGN consists of an extracellular domain that contains a
carbohydrate recognition domain (CRD) and a neck region of seven and a half 23 amino-
acid repeats, a transmembrane region, and a cytoplasmic tail containing recycling and
internalization motifs (Figure 1.4 B) (Feinberg et al., 2005; Guo et al., 2004; Powlesland
et al., 2006). hDC-SIGN has been successfully targeted using a humanized anti-DC-
SIGN IgG2/IgG4 composite antibody derived from AZN-D1.This antibody was capable
of loading antigens onto DCs, resulting in enhanced T cell presentation of both naïve and
memory T cell responses (Tacken et al., 2005). However, generating humanized
antibodies is still laborious and costly. Another obstacle for using DC-SIGN is the lack of
an exact functional ortholog of hDC-SIGN expressed on mouse DCs. This is a
disadvantage since it prevents mice from being used as preclinical models for the
evaluation of DC-SIGN targeting. Although multiple mouse homologs of hDC-SIGN
were identified, these proteins exert different cell expression patterns or have no the same
functions as hDC-SIGN (Nagaoka et al., 2010; Powlesland et al., 2006). hDC-SIGN
transgenic mice that express human DC-SIGN under the control of CD11c promoter have
been generated to solve this issue (Singh et al., 2009), but the cost of obtaining and
maintaining the mice would be high.
We designed a strategy to develop effective antibody-mimetic ligands targeting
both mouse and human DC-SIGN. These ligands could be used to present antigen on
21
DCs with higher specificity and lower cost. mRNA display using a naïve 10FnIII library
was performed to evolve ligands with affinity to mouse DC-SIGN (mDC-SIGN). This
selected library was subsequently selected against hDC-SIGN to develop dual specificity.
The closest mouse homologue of hDC-SIGN, mDC-SIGN (CD209A/CIRE), was chosen
as the target for mDC-SIGN selection (Figure 1.4B) (Park et al., 2001; Takahara et al.,
2004). Extracellular domains of both mDC-SIGN and hDC-SIGN were constructed,
expressed, and used as targets for mRNA display selections. These dual-specific ligands
could then be used in mice as good preclinical animal model for human cancer vaccine
development. Details of the DC-SIGN project are discussed in Chapter 4.

22
Chapter 2
Developing Antibody-Mimetic Ligands Targeting
Different Activity States of !
2
AR Using mRNA Display

In collaboration with Dr. Soren G. F. Rasmussen and Professor Brian Kobilka,
Department of Molecular and Cellular Physiology, Stanford University

Abstract
G protein-coupled receptors (GPCRs) are the largest family of cell surface
receptors in eukaryotes and regulate various physiological functions. GPCRs are also
among the most heavily investigated drug targets in the pharmaceutical industry.
However, the development of GPCR-targeted drugs is not routine, so the single-GPCR
specificity of drugs is always not achievable. To overcome these, we attempted to
develop a consistent method to generate ligands that bind not only single GPCR but also
active or inactive state of GPCR. !
2
adrenergic receptor (!
2
AR) was used as a model
system, because it is the most studied human class A GPCR and first solved crystal
structure of human GPCR. In addition, the highly homogenous and functional !
2
AR can
be made routinely as stable selection target.
mRNA display was exploited to develop antibody-mimetic ligands with state-
specificity to !
2
AR using the tenth human fibronectin type three domain (10FnIII)-
scaffold library. Since an antibody was used to solve the first human !
2
AR structure,
antibody-mimetic ligands based on 10FnIII-scaffold could serve the same purpose.
Four attempts were made to develop inactive (inverse agonist-bound) or active
(agonist-bound) !
2
AR-specific antibody-mimetic ligands by mRNA display using
e10FnIII-scaffold naïve library. However, there were two problems, the background
23
binding to M1 beads and non-specific electrostatic interaction with !
2
AR immobilized on
M1 beads. The background binding to M1 beads can be solved by the capturing method
that involves first allowing e10FnIII library and !
2
AR to interact, and subsequently
capturing the complex by a M1 affinity column. The electrostatic interaction could be
partially blocked with MgCl
2
, but pool 6 in the forth attempt is full of positively charged
e10FnIII clones. Cation exchange was applied to remove most positively charged
e10FnIII clones. Although several neutral e10FnIII clones had good affinity and low
background binding to !
2
AR, one e10FnIII ligand that was tested in monobromobimane
fluorescent assay showed no function in stabilizing active !
2
AR. The lesson learned here
could pave the way for future success in targeting !
2
AR.

24
Introduction
G protein-coupled receptors (GPCRs) are the most abundant and important cell
surface receptors in the human genome and regulate essential physiological activities,
such as sensory, neurotransmission, and hormone responses (Almen et al., 2009). Due to
their importance in cell regulation, 27% of FDA-approved drugs target GPCRs
(Overington et al., 2006). These drugs are mostly small molecules that interact with the
extracellular region of GPCRs (Schlyer and Horuk, 2006). All GPCRs adopt a common
structure containing seven transmembrane helices that comprise an extracellular N-
terminus and three extracellular loops, and an intracellular C-terminus and three
intracellular loops (Nathans and Hogness, 1983; Pebay-Peyroula et al., 1997). The
accumulated evidence from previous structural and functional studies on GPCRs suggests
a strong correlation between conformational changes and activity (Rasmussen et al.,
2011b; Swaminath et al., 2004). The different conformations of a GPCR determine its
various activities, indicating that there is a spectrum of conformations for a GPCR in
between the active and inactive states (Kobilka and Deupi, 2007). Therefore, the better
strategy to develop a drug to modulate a specific function of GPCRs is to design a drug
based on the specific conformation of GPCRs for that function. It is important to target
distinct conformations of different activity states of a GPCR, since this could lead to a
better understanding of GPCR structures in different states and potential drug
development.
Despite their importance, only very few crystal structures of GPCRs have been
solved due to their challenging properties (Kolb and Klebe, 2011). First, GPCRs are
25
membrane proteins, so their structural integrity can only be retained when embedded in a
membrane-like environment. Second, it is very difficult to express GPCRs and purify
functional ones. Third, the extracellular and intracellular loops of GPCRs are flexible.
However, some recent breakthroughs led to the first structure of a human and non-
rhodopsin GPCR, beta-2 adrenergic receptor (!
2
AR) that was solved (Rasmussen et al.,
2007). These breakthroughs include that !
2
AR can be solubilized in a neutral detergent-
rich (N-dodecyl-!-D-maltoside/DDM) aqueous buffer and retain its structural integrity.
Second, high-quality functional !
2
AR can be routinely produced by Sf9 insect cells and a
sequential ligand and affinity chromatography (Kobilka, 1995). Finally, an antibody or
nanobody against !
2
AR was used to stabilize the flexible loops, resulting in co-
crystallization of an antibody or nanobody with !
2
AR (Rasmussen et al., 2011a;
Rasmussen et al., 2007; Rasmussen et al., 2011b). Since we could obtain high-quality,
functional !
2
AR through the unique opportunity to collaborate with Professor Brian
Kobilka, we were able to use !
2
AR as our model system for GPCR targeting.
Our goal was to develop novel antibody-mimetic ligands to stabilize active or
inactive states of !
2
AR. These antibody-mimetic ligands could be useful to obtain high-
resolution structures of different states of !
2
AR as well as possibly manipulating !
2
AR
function. To achieve our goal, the distinct states of !
2
AR can be induced and stabilized
using different small-molecule ligands. These GPCR ligands are classified based on their
effect on the GPCR basal activity (Rosenbaum et al., 2009). Agonists that increase the
GPCR basal activity can active a GPCR and inverse agonists that decrease that GPCR
basal activity can inactivate a GPCR. Antagonists bind to a GPCR and retain the basal
26
activity. Therefore, we used agonists to generate the active state of !
2
AR and inverse
agonist to make the inactive state of !
2
AR.
We utilized an in vitro selection technique, mRNA display along with antibody-
mimetic scaffold, the tenth human fibronectin type III domain (10FnIII) to target
!
2
AR. mRNA display provides the ability to scan very large libraries (>10
13
unique
molecules), increasing the probability of identifying high-affinity ligands for a target
(Ling, 2003; Perelson and Oster, 1979; Takahashi et al., 2003). In addition, a puromycin
covalently couples a peptide or a protein to its encoded mRNA, which makes mRNA
display a simple, more stable selection method (Roberts and Szostak, 1997). The stability
of mRNA display makes it uniquely suitable to target !
2
AR in detergent-rich conditions.
Since antibodies and nanobodies have been demonstrated to be powerful tools to
crystallize !
2
AR, we aimed to show that mRNA display using human e10FnIII-
scaffolded library could evolve antibody-mimetic ligands to serve the same purpose as
antibodies and nanobodies. e10FnIII was derived from wild type human 10FnIII to
enhance expression and stability in an E.coli and rabbit reticulocyte lysate (Olson et al.,
2011). 10FnIII is structurally similar to an antibody and more stable than antibody
without having any disulfide bond (Koide et al., 1998). mRNA display using 10FnIII-
scaffold library were exploited to evolve antibody-mimetic ligands with high specificity
and affinity to TNF-$, phospho-I#B$, SARS N protein, and IL-6 (Liao et al., 2009;
Olson et al., 2011; Olson et al., 2008; Xu et al., 2002). Thus, our strategy can potentially
reduce the cost and labor in generating antibodies or nanobodies, and develop antibody-
mimetic ligands with higher affinity and specificity than antibodies or nanobodies.
27
Several attempts were made to target agonist-bound or inverse agoinst-bound
!
2
AR by mRNA display using an e10FnIII-scaffold library. However, these selections
resulted in high background binding to no-target beads, positively charged proteins, and
nonfunctional neutral proteins. We then incorporated a different method, capturing in the
affinity maturation step to reduce the background binding to beads. To this end, fusions
are incubated with FLAG-tagged !
2
ARs, subsequently captured by a M2 affinity column,
and finally eluted. Capturing is different from the previous method, pull-down where
fusions were incubated with !
2
ARs that had been immobilized on M2 beads. Capturing
reduced the time for fusions to contact M2 affinity beads, resulting in significant
reduction of background binding to beads.
We also tried using high concentration of MgCl
2
in affinity maturation and cation
exchange as a pre-clear step, which efficiently removed positively charged proteins.
However, the selected neutral protein showed no function in inducing active !
2
AR
formation or stabilizing agonist-bound active !
2
AR. Although we were not able to find
out the exact reason for all the issues, we learned that the method and the beads used to
immobilize !
2
AR are critical. Moreover, !
2
AR itself or surrounding phospholipids could
present negatively charged surfaces to enrich positively charged proteins in our selection.
A previous study has reported that membrane proteins solubilized in detergent-rich
solution still had residual phospholipids attached to them (Guan et al., 2006). This
indicates that the heterogeneity of purified !
2
AR could lead to the positively charged
protein via non-specific electrostatic interaction. This strong electrostatic interaction
could eliminate the real binding events in the beginning of the selection process.
28
For future attempts to target !
2
AR, immobilization of !
2
AR on beads, library
design, and minimization of non-specific electrostatic interaction are the key components
to be taken into consideration.

Materials and Methods
Library construction
Construction of antibody-mimetic library using human 10FnIII as the scaffold has
been described previously (Olson and Roberts, 2007). An additional five mutations were
introduced into the N-terminus of wild-type 10FnIII scaffold and the last residue of BC
loop was doped to generate the e10FnIII library (Figure 2.1A and B) (Olson et al., 2011;
Olson et al., 2008). V5K, A6E, T8S, L12I, and L13Q mutations were cloned into
e10FnIII library to universally enhance expression and solubility (numbering based on the
start Met in our #1-7 truncated library). The last residue of BC loop was doped as
Leucine, Isoleucine, or Valine to retain the structural stability of e10FnIII library. Eight
oligonucleotides (Yale Keck Oligonucleotide Synthesis Facility or Integrated DNA
Technologies) were used to construct 10FnIII library as described previously, except that
three of eight oligonucleotides were redesigned to construct e10FnIII library (Figure
2.1C) (Olson and Roberts, 2007).
One new oligonucleotide, eFnoligo3 (5’-ACC AGC ATC CAG ATC AGC TGG
55S 55S 55S 55S 55S 55S VTT CGC TAC TAC CGC ATC ACC TAC G-3’; 5 indicated
the dNTP mixture of 20%T, 30%C, 30%A, and 20%C; S were mixed with 60%C and
40%G; V denoted the mixture of C, A, and G equally), containing the randomized BC
29

Figure 2.1 Sequence, construction, and possible structure of e10FnIII
library
(A) The sequence alignment of wild-type 10FnIII and e10FnIII shows the
difference between them. Five mutations, K, E, S, I, Q are highlighted in blue
and they were engineered into e10FnIII to increase protein expression. The last
position in BC loop is doped (1=Leu, Ile, and Val) and shown in green. (B) The
structure of wild-type 10FnIII is modified to show possible e10FnIII structure
(PDB: 1TTG). In e10FnIII construct, the first 7 amino acids are truncated, the
five mutations are labeled in blue, and the last postion of BC loop is shown in
green. Randomized BC and FG loops are in red. (C) The DNA template of
e10FnIII was generated using eight oligonucleotides as shown here.
Oligonucleotides in blue were newly designed to introduce five mutations in N-
terminus and one doped residue in the BC loops. The details were described in
materials and methods. [Modified from Olson, C.A. and Roberts, W.R. Protein
Sci. 16, 476-484 (2007)]

30
loop was used to introduce two mutations (L12I and L13Q) and one doped residue.
eFnoligo3 was annealed to Fnoligo4, extend by Klenow DNA polymerase, and purified
by agarose gel electrophoresis. The purified product was PCR amplified using the second
new primer, eFnoligo2 (5’-CAA TTA CAA TGC TCG AGG TCA AGG AAG CAT
CAC CAA CCA GCA TCC AGA TCA GCT GG-3’) and Fnoligo5. eFnoligo2 was used
to insert three mutations, V5K, A6E, and T8S, into the N-terminus of e10FnIII library.
Fnoligo6 and Fnoligo7 containing the randomized FG loop were annealed and extend by
Klenow DNA polymerase. All PCR and Klenow products were purified by agarose gel
eletrophoresis.
Both BC and FG fragments were digested with Bsa I and purified by agarose gel
electrophoresis. 1500 ng of each purified fragment were ligated together using T4 DNA
ligase in a 100 µL reaction. The ligated product was purified by agarose gel
electrophoresis and 13 ng/µL of purified product was recovered. The approximate
complexity of e10FnIII library is 10
12
(1 trillion unique sequences). The library was
extended and amplified by the third new oligonucleotide, eFnoligo1 (5’-TTC TAA TAC
GAC TCA CTA TAG GGA CAA TTA CTA TTT ACA ATT ACA ATG CTC GAG
GTC AAG G-3’), and Fnoligo9 (5’-GGA GCC GCT ACC CTT ATC GTC GTC ATC
CTT GTA ATC GGA TCC GGT GCG GTA GTT GAT GGA GAT CG-3’) in a 10-mL
PCR reaction. The final PCR products were aliquot into each tube containing 5 copies of
the 10
12
independent sequences (1 pmol/µL; 10 µL/aliquot).

31
!
2
AR protein expression and purification
!
2
adrenergic receptor was kindly provided by Dr. Soren G. F. Rasmussen from
Dr. Brian Kobilka lab. The cloning, expression, and purification of !
2
AR were described
previously (Kobilka, 1995; Rasmussen et al., 2007). In brief, a human !
2
AR construct
was truncated after residue 365 (!
2
AR365N) and expressed in Sf9 insect cells infected
with recombinant baculovirus (Figure 2.2A). !
2
AR was solubilized in n-dodecyl-!-D-
maltoside (DDM) aqueous buffer and purified by M1 FLAG affinity chromatography
(Sigma) before and after an alprenolol-sepharose chromatography. Pure and functional
!
2
AR was either left on the M1 beads or eluted into HLS buffer (20 mM HEPES-KOH
pH 7.5, 100 mM NaCl, and 0.1% DDM) (Figure 2.2B). !
2
AR was coupled with or
without ligands (agonists, THRX-144877 and BI-167107; or an inverse agonist, ICI-
118551) to generate active state, inactive state, or native state of !
2
AR (Figure 2.3).
!
2
ARs immobilized on M1 beads were stored at 4
o
C and used within a month. !
2
ARs
were concentrated in HLS buffer with 10% glycerol and stored at -80
o
C in small aliquots.

mRNA display selection
In the first round of mRNA display, 7.5 µL of the naïve e10FnIII library (1
pmol/µL) was PCR amplified by Taq DNA polymerase in a 1.5 mL PCR reaction. After
4-6 PCR cycles, the PCR product was purified by phenol-chloroform extraction and
ethanol precipitation to get total ~150 pmol (estimated to be ~90 copies of each 10
12

DNA molecules). The purified PCR product was transcribed to mRNA using T7 RNA
polymerase in 1.5 mL in vitro transcription reaction at 37
o
C for two hours. 1/10th volume
32

Figure 2.2 Structure and immbolization of !
2
adrenergic receptor (!
2
AR)
(A) Schematic secondary structure (left) and the crystal structure of !
2
AR
(right)(PDB: 3DQB) are shown. N-terminus (NT) and three extracellular loops
(ECL1-3) face outside of the cell. C-terminus (CT) and three intercellular loops
(ICL1-3) are in the cytosolic side. Seven transmembrane helices (TM1-7) reside
inside the cell membrane (grey). Two disulfide bonds are shown as yellow
dashed lines. Cysteine 265 (red) can be either modified by iodoacetamide,
monobromobimane, or maleimide-PEG11-Biotin. In the crystal structure, TM3
and TM4 are not labeled, because they are behind TM5, TM6, and TM7. A full
agonist, BI-167107 (blue sphere) is in the ligand-binding pocket that is in the
interface between the extracellular loops and transmembrane helices. The N-
terminal FLAG-tag (black) was used to purify and immobilize !
2
AR. (B) N-
terminal FLAG-tagged !
2
AR (red extracellular loops, black transmembrane
helices, and blue intracellular loops) is couled with a ligand. The coupled ligand
can be full agonist, partial agonist, antagonist, or inverse agonist. Ligand-bound
!
2
AR is immoblizied on anti-FLAG M1 antibody (M1)-conjugated cyanogen
bromide (CNBr) activated sepharose (M1 bead) via its N-terminal FLAG tag.

33
of 0.5 M EDTA pH 8.0 was added to dissolve the precipitated phosphate salt and the
mRNA product was purified by Urea PAGE electrophoresis, electroelution, and ethanol
precipitation. Purified mRNA was ligated to a puromycin attached to a DNA linker,
pF30P (8(A)
21
(7)
3
ACC6; 8 is chemical phosphorylation reagent, 7 is spacer
phosphoramidite 9, and 6 is puromycin CPG; Keck Oligonucleotide Synthesis Facility).
The mRNA, pF30P, and the splint oligonucleotide (FN-pF30P-Splint, 5’-
TTTTTTTTTTTTGGAGCCGCTACC-3’, is complementary to 3’-mRNA and 5’-
pF30P.) were mixed in 1:1.1:1.2 molar ratio and ligated by T4 DNA ligase in 1.0 mL
reaction at room temperature for an hour. The ligated product was purified by Urea
PAGE electrophoresis, electroelution, and ethanol precipitation.
Purified mRNA-pF30P was in vitro translated in 2.5 mL reaction using rabbit
reticular lysate (reticular lysate, Green Hectares; salts and buffers, Novagen) at 30
o
C for
an hour. The translation reaction was quenched and fusion formation of mRNA-pF30P-
e10FnIII protein was improved by adding salt mix (2 µL of 1 M MgCl
2
and 7 µL of 2.5
M KCl) per 25 µL of translation reaction at room temperature for 15 min. mRNA-pF30P-
e10FnIII fusions were purified by pull-down using an equal volume of 50% (v/v)
oligo(dT) cellulose slurry (GE Heathcare) in dT buffer (100 mM Tris-HCl pH 8.0, 1 M
NaCl, 0.2% Triton X-100, and 1 mM EDTA pH 8.0) at 4
o
C for an hour. dT cellulose
beads were washed four times with dT buffer and fusions were eluted with room
temperature ddH
2
O. The elution was desalted and exchanged with the first strand buffer
(50 mM Tris-HCl pH 8.3, 75 mM KCl, and 3 mM MgCl
2
) by pre-packed 5 mL NAP-25
spin column. The resulted fusions were reverse transcribed with reverse primer Fnoligo10
34
(5’-GGAGCCGCTACCCTTATCGTCG-3’) using Superscript II enzyme (Invitrogen) at
42
o
C for an hour.
Two different methods were used to do affinity maturation, pull-down and
capturing (Figure 2.3). In pull-down, N-terminal FLAG-tagged !
2
ARs were first
immobilized on M1-antibody conjugated-cyanogen bromide activated sepharoses (GE
Healthcare) at room temperature for an hour prior to the affinty maturation. !
2
AR-
immobilized beads were wash by selection buffer (20 mM HEPES-KOH pH 7.5, 100
mM NaCl, 0.1% DDM, 2 mM CaCl
2
, 0.5 mg/mL BSA, and 0.1mg/mL tRNA) three times
and the final concentration of !
2
AR per µL beads was 111.11 pmol. Fusions were
incubated with 20 µL of !
2
AR-immbilized beads at room temperature for an hour and
then washed 4 times with 1x selection buffer. Bound fusions were PCR ampified from
the beads and the PCR product were extracted by phenol-chloroform, and precipitated by
ethanol. Purified DNA pool was labeled as pool 1 and used for the next round of
selection. After the first round, the translation reaction was adjusted to smaller amount
(100 µL in vitro translation). In addition, the FLAG purification and pre-clear were
applied to each following round. C-terminal FLAG-tagged fusions were purified using
monoclonal anti-FLAG M2 agarose beads (Sigma) and eluted twice with 100 µL of 0.1
mg/ml 3X FLAG peptide (Sigma) at room temperature. A pre-clear step was introduced
to remove fusions that bound to no-target beads by flowing the fusions through the 100
µL empty M1 beads packed in a column prior to the affinity maturation. Instead of PCR
amplifying bound fusions directly from beads, !
2
AR-bound fusions were eluted by 5 mM

35

Figure 2.3 Different selection methods.
(A) !
2
AR was first coupled with a full agonist, THRX-144877 to induce active
!
2
AR. THRX-bound !
2
AR was immobilized on M1 bead, which was used in the
active !
2
AR selection to pull down the fusions. (B) An inverse agonist, ICI-
118551 was bound to !
2
AR to inactivate !
2
AR. ICI-bound !
2
AR was immobilize
on M1 beads and subsequently incubated with e10FnIII naïve library to pull
down the fusions with higher affinity to inactivate !
2
AR. (C) Capturing is applied
in affinity maturation to minimize the non-specific binding to M1 beads. A full
agonist, BI-167107 activated !
2
AR and subsequently incubated with naïve
e10FnIII library. After incubation, BI-bound !
2
AR and fusion-BI-bound !
2
AR
complexes were captured on M1 beads via the N-terminal FLAG in !
2
AR.
Fusion is shown as a cDNA/mRNA (grey/black)-pF30P (dark blue/grey)-
e10FnIII.

36
EGTA and 0.2 mg/ml FLAG peptide before PCR reaction in round 9 and 10. Pull-down
was performed to target THRX-bound and ICI-bound !
2
ARs.
In capturing, N-terminal FLAG-tagged !
2
AR (400 pmol) was incubated with
fusions (0.085 pmol) in binding buffer (20 mM HEPES-KOH pH 7.5, 100 mM NaCl,
0.1% DDM, 2 mM CaCl
2
, 50 mM MgCl
2
) at room temperature for 1 hour. After binding,
the mixture was passed through a pre-packed column containing 20 µL M1 beads
column. The !
2
AR-fusion complexes were captured by M1 beads and eluted with elution
buffer (HLS buffer, 5 mM EGTA and 0.2 mg/ml FLAG peptide). DNA of bound fusiton
was PCR amplified, phenol-chloroform extracted, and ethanol precipitated. Purified DNA
was subjected to the next round of selection. The FLAG purification was applied to every
round of this selection. There was no pre-clear and only 200 µL translation was
performed in the first round of selection, leading to lower complexity and copy number
of library (~1 X 10
12
). Sequentially, six rounds of selection using capturing and two
rounds of cation exchange were done. In the final round of selection (round 9), fusions
were pre-cleared by cation exchange and subsequently went through affinity maturation
using capturing. Capturing and cation exchange were used to target BI-bound !
2
AR.
Cation exchange was performed to remove highly-positively charged fusions.
Cation exchange was done in round 7 by incubating BI pool 6 fusions with 200 µL
HiTrap SP HP beads (GE Healthcare) in cation exchange buffer (20 mM HEPES-KOH
pH 7.5, 10 mM NaCl, and 0.1% DDM) at room temperature for 10 min. After incubation,
the mixture was put into a small column and the flowthrough was collected. 200 µL
cation exchange buffer was loaded into the column and the flowthrough was collected.
37
The flowthrough was pooled and subjected to PCR amplification. The BI pool 7
proceeded to another round of cation exchange (round 8). After 2 rounds of cation
exchange, another round of selection (round 9) was done using cation exchange to pre-
clear fusion before fusion was incubated with !
2
AR

In vitro radiolabeled binding assay
To verify the affinity of each pool to !
2
AR, radiolabeled fusions were generated
using
L
-[
35
S]methionines (MP Biomedicals) instead of cold methionine in the translation
reaction. Radiolabeled e10FnIII fusion was treated with ribonuclease A (Roche Applied
Science) to remove large mRNA. RNase-treated samples (pF30P-e10FnIII) were
incubated with !
2
AR-immobilized beads or empty beads at room temperature for 1 hour.
The beads were then wash 4 times with selection buffer and their radiation signals were
counted in a scintillation counter (Beckman Coulter, Inc.). The total input of radiolabeled
fusion was the sum of radiation counts in flowthrough, washes, and bead. The percentage
binding was estimated by diving bead counts with total counts.
The percentage pool binding of the capturing method was monitored in sync with
the selection. The radiolabeled fusions (hot) were mixed with the cold fusion in each
round of selection (100 µL of hot translation and 100 µL of cold translation for the first
and second rounds; after the first two rounds, 25 µL of hot translation was mixed with
100 µL of cold translation.). After capturing and elution, 1/10th of elution was counted
for radiation counts using a scintillation counter. The calculation of the percentage
binding was the same, except the radiation counts of elution were multiplied by 10.
38
The final pools were PCR amplified and two restriction sites were introduced
(Xho I and Bam HI) using the 5’-eFnoligo1 primer and the 3’-Fnoligo11 primer (5’-GGA
GCC GCT ACC CTT ATC GTC GTC ATC CTT GTA ATC GGA TCC GGT G-3’). The
PCR products were cloned into vector pAO5 (Olson and Roberts, 2007) using restriction
digestion and ligation. Ligated products were transformed into E. coli. XL-10. The
purified plasmids were sequenced (Laragen, Inc.) and the sequences were analyzed using
ClustalW2 (EMBL-EBI) (Goujon et al., 2010; Larkin et al., 2007). Some clones were
PCR amplified using eFnoligo1 and Fnoligo9 and in vitro transcribed to generate mRNAs
for in vitro translation. These mRNAs were translated into radiolabeled e10FnIII proteins
using rabbit reticular lysate supplied with
L
-[
35
S]methionines. Radiolabeled e10FnIII
proteins were purified using anti-FLAG M2 beads. The binding of these selected
e10FnIII proteins to !
2
ARs were either tested in pull-down or by capturing. The binding
percentages of the selected e10FnIII proteins to !
2
AR were estimated as that of pool
binding.

Selected e10FnIII protein expression and purification
Selected e10FnIII clones were transformed into E. coli. BL21(DE3). The
transformed cells were plated on Amp
+
LB agar plates and grown overnight at 37
o
C. For
each clone, one colony was picked up and grown in 5 mL LB with 50 µg/mL ampiclillin
at 37
o
C overnight. A 1 mL overnight culture was inoculated with 100 mL LB
supplemented with 50 µg/mL ampiclillin, and grown until the O.D. 600 reached 0.4~0.5.
Protein expression was induced by addition of 1 mM IPTG and the cells were grown
39
overnight at room temperature. Cells were harvested by centrifugation and the cell pellets
were frozen at -80
o
C.
The frozen cell pellet was thawed on ice, resuspended, and lysed in B-PER
reagent (Thermo Scientific) supplemented with protease inhibitor cocktails (Thermo
Scientific). The cell lysate was pelleted at 14,000 rpm for 20 min twice. 500 µL HisPur
Cobalt Resin (Thermo Scientific) was packed in a disposable 5 mL column (Thermo
Scientific) and equilibrated with 6-column volume of binding buffer (20 mM Tris-HCl
pH 8.0, 500 mM NaCl, 10 mM imidazole). The supernatant of the cell lysate was loaded
into the pre-packed column and subsequently the column was washed with 12-column
volume of wash buffer (20 mM Tris-HCl pH 8.0, 500 mM NaCl, 20 mM imidazole). The
proteins were eluted with 4 column volumes of elution buffer (20 mM Tris-HCl pH 8.0,
and 100 mM NaCl) containing 250 mM imidazole and followed by 4 column volumes of
elution buffer containing 500 mM imidazole. Different elutions were verified using SDS-
PAGE gel.
The elutions containing the pure e10FnIII protein were pooled together,
exchanged with HLS buffer, and concentrated using Amicon Ultra-15 centrifugal filter
units (Millipore). 10% glycerol was added to the concentrated protein and the
concentration of a protein was determined using O.D. 260 absorption and extinction
coefficient. The purified protein was aliquot to small volume, flash-frozen in liquid
nitrogen, and stored at -80
o
C.

40
Monobromobimane (mBBr) fluorescence assay
Purified !
2
AR was labeled with the environmentally sensitive fluorophore,
monobromobimane (mBBr, Invitrogen) at cysteine 265 (Cys265) located in the
cytoplasmic end of TM6 as described previously (Figure 2.2A) (Rosenbaum et al., 2007;
Yao et al., 2006). Purified !
2
AR and monobromobimane were mixed at same molarity
and incubated overnight on ice in the dark. The mBBr-labeled !
2
AR was purified by gel
filtration, concentrated, and stored at -80
o
C. The mBBr-labeled !
2
AR were diluted with
HLS buffer to 400 nM and the fluorescence spectroscopy experiments were performed on
spectrofluorometer (Shimadzu corp.) at room temperature. The excitation and emission
slit width were both 20 nm. The sample was excited by 370 nm and the emission spectra
were scanned from 417 to 570 nm for an interval time of 0.5 s/nm. To test the function of
selected proteins, the spectra were taken after the selected proteins (20 µM) were
incubated with !
2
ARs (0.4 µM) for 30 min at room temperature. After taking the spectra,
the agonist (BI, 10 µM) and antagonist (Alprenolol, 400 µM) were added to the mixture
and incubated for 30 min, sequentially. These spectra were taken to confirm the function
of purified !
2
AR. The spectra were corrected by subtracting the background fluorescence
that resulted from buffer, ligands, and selected proteins.

Results
Construction of e10FnIII library
Wild-type 10FnIII library was constructed and successfully applied to mRNA
display in our lab (Liao et al., 2009; Olson et al., 2008; Olson and Roberts, 2007).
41
However, the selected 10FnIII proteins suffered from insufficient expression in E. coli,
low solubility, and instability. A pervious study had reported that V5K, A6E, T8S, L12I,
and L13Q mutations in the N-terminus of e10FnIII library enhanced protein expression
when expressed in E.coli and in rabbit reticulocyte lysate (Figure 2.1A) (Olson et al.,
2011). The paper also suggested that the last position of BC loop should be limited to
Leucine, Isoleucine, or Valine to retain the structural stability. This position is a buried
core residue in the wild-type 10FnIII, so it may interact in the transition-state folding
nucleus (Figure 2.1B) (Cota et al., 2001). Therefore, e10FnIII was constructed using the
same strategy as previously described (Olson and Roberts, 2007), but three re-designed
new oligonucleotides were exploited to insert five mutations in the N-terminal constant
region and the last position of the BC loop was doped to Leucine, Isoleucine, or Valine
(Figure 2.1C). This e10FnIII library was sequenced and the result showed that five
mutations and the last position of BC loop were successfully cloned into e10FnIII library.
The frequency of stop codon and frame-shift were in a reasonable range when compared
to the theoretical value (data not shown). Therefore, the e10FnIII library was constructed
and used in this study to develop antibody-mimetic ligands targeting !
2
AR.

mRNA display against THRX-bound and ICI-bound !
2
ARs using pull down
The !
2
AR construct that we used as selection target is truncated after residue 365
(!
2
AR365N), and is the fragment that was used to obtain the first human !
2
AR crystal
structure (Figure 2.2A) (Rasmussen et al., 2007). This suggests that !
2
AR is high-quality
and functional. The FLAG tag in N-terminus of !
2
AR was exploited to immobilize !
2
AR
42
on M1 beads (Figure 2.2B). To develop antibody-mimetic ligands with specificity to
different states of !
2
AR, a full agonist, THRX or an inverse agonist, ICI was coupled
with !
2
AR to form an active or an inactive state !
2
AR, respectively (Figure 2.3A and B).
THRX-bound !
2
AR was used as the target for active !
2
AR selection and ICI-bound
!
2
AR was the target for inactive !
2
AR selection.
Before the actual selection, a mock !
2
AR selection using small-scale e10FnIII
library was performed and showed that !
2
AR was intact after selection procedure (data
not shown). Ten rounds of selection were done for active and inactive !
2
AR, and the pool
binding shows that both selections converged the low target binding to higher (Figure
2.4A; THRX-bound !
2
AR selection: 5.2% of pool 0 to 47.7% of pool 10; ICI-bound
!
2
AR selection: 5.6% of pool 0 to 32% of pool 10). This data also shows that the binding
of two final pools to their targets is higher than background (Figure 2.4A). In addition,
the pool 0 representing naïve e10FnIII library binds to M1 beads, !
2
AR, and ligand-
bound !
2
AR with the same affinity, indicating that naïve e10FnIII library does not bias
!
2
AR binding.
However, there are several issues. First, the background binding of the pool 0 to
M1 beads was much higher (5%) than the reasonable value (< 1% from other successful
selections using mRNA display). This indicates that there must be some non-specific
interactions between M1 beads and naïve e10FnIII library, which could interfere the
specific interaction between e10FnIII ligands and targets. This could lead to non-specific
e10FnIII clones outcompeting specific e10FnIII clones, so there would be no specific
e10FnIII ligands for targets. Second, the background binding to M1 beads increased in
43

Figure 2.4 THRX-!
2
AR (active) and ICI-!
2
AR (inactive) selections
(A) The pool binding of THRX-!
2
AR selection(left) and ICI-!
2
AR(right) selections
shows the target binding increased. However, the background binding to beads
is too high and the pools after the selections have no specificity to their targets
(lack of state-specificity). (B) The analysis of pool amino acid composition
indicates that both selections converge the randomized neutral amino acid
sequences to positively charged amino acid sequences.
44
sync with the target binding in the selections, suggesting that the selections evolve
e10FnIII ligands that bind to M1 beads. Although the target binding is higher than M1
bead binding, the selected pools could have a mix of e10FnIII ligands binding to M1
beads, to targets, or to complex of M1 beads and targets. This could make the
identification of the target-specific e10FnIII ligands very difficult. Third, the similar
affinity of the selected pools to all three states of !
2
AR (native, active, and inactive) in
active and inactive !
2
AR selections indicate that no state-specificity was evolved.
The sequencing results of pool 0, 8, 9, 10 from both selections show the tendency
of selections toward highly positively charged amino-acid sequences. We analyzed the
probability of positively charged amino acids (Arginine, Lysine, and Histidine) in
different pools based on sequencing result using the normal distribution formula (Figure
2.4B). The normal distribution curves of the different pools show that the number of
positively charged amino acids increased from pool 0 to pool 10 (Figure 2.4B). In
addition, the curve of theoretical positively charged residues in naïve e10FnIII library and
that of the sequencing data of naïve e10FnIII library overlap well. This suggests that the
naïve e10FnIII library is not biased in favor of more positively charged amino acids.
Five selected e10FnIII ligands in THRX pool 8 containing less than three
positively charged amino acids and two highly positively charged e10FnIII ligands in
THRX pool 10 were tested for their binding to M1 beads and different states of !
2
AR
(Figure 2.5). The results demonstrate that the selected e10FnIII clones can be categorized
into three groups. The first group contains highly positively charged e10FnIII clones that
bind strongly to both M1 beads and !
2
ARs (THRX10-8 and 10-2). The second group
45

Figure 2.5 Sequences and clone bindings of THRX pool 8 and 10.
The pool 8 and 10 in THRX-!
2
AR selection were cloned and sequenced. Clones
were tested for target binding. In the sequence alignment, X indicates any 20
amino acids in BC and FG loops. 1 represents the residue that was doped to
Val, Leu, and Ile. Five clones in THRX pool 8 are less positively charged and
two clones in THRX pool 10 are positively charged. The positively charged
amino acids are shown in blue. The positively charged amino acids increased
from top to bottom in the sequence alignment and from left to right in the bar
graph of clone binding. Clone binding shows that positively charged clones have
high bead and target binding. Neutral clones could not bind to beads and target,
except THRX8-3 and THRX8-4. However, the background binding to beads is
still too high for THRX8-4.
46
includes the low positively charged e10FnIII clones that have some degree of affinity to
!
2
ARs and lower affinity to M1 beads, which is what we want (THRX8-3 and 8-4). Third
group consists of low positively charged e10FnIII clones that have almost no binding to
!
2
ARs and M1 beads. This supports that the high background binding to M1 beads
results in the mix of specific ligands and non-specific ligands, leading the great difficulty
to identify specific ligands. A preliminary screening of another 20 clones was attempted
and the results show that the distributions of the three groups are uneven, in which the
second group is especially less than two other groups and no positive clone was identified
(data not shown).

Capturing method reduces background binding to M1 beads
To solve the issue of the background binding to M1 beads, a pre-packed column
with M1 beads was used to capture !
2
AR that was pre-incubated with the e10FnIII
fusions for 1 hour. This M1 column would capture !
2
AR or the complex of !
2
AR and
fusion via the N-terminal FLAG tag of !
2
AR. The bound molecules were eluted by
EGTA and FLAG peptides (Figure 2.6A). There are two advantages of this capturing
method in reducing background binding to M1 beads. One is that the capturing reduces
the time that e10FnIII fusions interact with M1 beads to 1-2 min (1 hour in pull down).
Another is that elution of the bound e10FnIII fusions from M1 beads can prevent PCR
amplification of the e10FnIII fusions that bind to M1 beads. This problem results from
PCR amplification of fusions using resuspended BI-!
2
AR-immobilized M1 beads in the
pull down method.
47

Figure 2.6. The capturing method reduces background binding to M1
beads
(A) The capturing method is used to reduce the background binding to M1
beads. BI-!
2
AR is incubated with naïve e10FnIII library for 1 hour at room
temperature and subsequently captured by a column pre-packed with M1 beads
(M1 column). After the capturing, the bound complexes of BI-!
2
AR and e10FnIII
(indicated in the red box) are eluted by EGTA and FLAG peptide. The elution is
PCR amplified and subjected to next round of selection. (B) The bar graph (top-
right) shows that fusion of e10FnIII naïve library binds to M1 beads more
strongly in the pull-down than in the capturing. This indicates that the capturing
reduces the background binding of e10FnIII naïve library to M1 beads. However,
the non-specific binding of e10FnIII naïve library to !
2
AR is still high. The bar
graph of different pools in THRX-!
2
AR selection binding to !
2
AR is shown on
the bottom right. This demonstrated that the capturing decreases the positively
charged pool binding to !
2
AR, suggesting that the capturing can reduce
electrostatic interaction between positively charged e10FnIII and !
2
AR.

48
Fusion of naïve e10FnIII library (pool 0) binding to M1 beads decreased
significantly using the capturing method when compared to that using the pull down
method (Figure 2.6B top; pull down: 5.3% binding and capturing: 0.4% binding). The
binding of THRX pool 0, 8, and 10 fusions to !
2
ARs also decreased significantly using
capturing in comparison to using the pull down method (Figure 2.6B bottom). The later
pool has a greater decrease in !
2
AR binding using the capturing than using the pull down
method. In addition, the later pool contains a higher number of positively charged
e10FnIII clones. Together, this indicates that the capturing method could reduce the
background binding to M1 beads and the binding of highly positively charged e10FnIII
clones to !
2
AR, or the complex of M1 bead and !
2
AR. This may also indirectly imply
that the background binding to M1 beads may be due to electrostatic interaction.
The first attempt at using the capturing method in HLS buffer with CaCl
2
to do
selection against BI-!
2
AR did not converge the pools. BI is a full-agonist with superior
affinity to !
2
AR (K
D
=0.1 nM), so BI was used to replace THRX in the selection. The
pool binding shows low background binding to M1 beads (< 0.5%), suggesting that the
capturing method indeed reduced this background binding (data no shown). However, the
low PCR cycles (~10 cycles) and the low binding to BI-!
2
AR (1% - 4%) in every pool
are contradicted (data not shown). The low PCR cycles indicate that plenty of e10FnIII
clones survived the selection, so their binding to BI-!
2
AR should be high. The possible
reasons would be that either template DNA leaked through after mRNA was purified by
PAGE gel or cDNA/mRNA in fusion. Since the fusion used in the binding assay was
RNase treated (pF30P-e10FnIII), the binding data might not represent the real situation in
49
selection where the full fusion used is cDNA/mRNA-pF30P-e10FnIII. Therefore, the full
fusion of BI pool 1 was tested for the binding to M1 beads and !
2
AR (Figure 2.7A, black
boxed) and the results showed the strong binding of full fusion to !
2
AR but no binding to
M1 beads (56% and 0.2%). Other formats of BI pool 1 fusion were also tested for binding
to !
2
AR and the results demonstrated that cDNA/mRNA is responsible for high non-
specific binding of BI pool 1 (close to naïve e10FnIII library) to !
2
AR. This indicates that
!
2
AR may contain a positively charged region that interacts with negatively charged
cDNA/mRNA. This was not seen in the THRX-!
2
AR and ICI-!
2
AR seletions using the
pull down, because the buffer contains large amounts of tRNA and BSA to block this
interaction. However, this does not explain why the two selections resulted in positively
charged e10FnIII clones.
To avoid this non-specific electrostatic interaction, cations were introduced in the
BI-!
2
AR selection using capturing method. NaCl (+1), MgCl
2
(+2), spermidine (+3), and
spermine (+4) were added in the HLS buffer separately and the buffer with each cation
was tested for its effect in the e10FnIII naïve library binding to !
2
AR using the capturing
(Figure 2.7B, left). All the cations tested, except spermine, could efficiently decrease the
non-specific binding of e10FnIII naïve library to !
2
AR from > 50% in HLS buffer to
~1% in HLS buffer with high concentration of each cation. However, MgCl
2
is the only
cation that did not alter the function or structure of !
2
AR at a concentration that can block
non-specific electrostatic interaction, when tested in the monobromobimane (mBBr)
fluorescence assay (Figure 2.7B, right). Therefore, MgCl
2
was the optimal choice to

50

Figure 2.7 cDNA/mRNA is responsible for non-specific electrostatic
interaction that can be blocked by MgCl
2

(A) Different formats of naïve e10FnIII fusions have significantly different
affinities to !
2
AR. The box shows the fusion that we used in the selection. BI
pool 1 (close to naïve e10FnIII library) should not bind to !
2
AR, but the result
shows that fusions containing cDNA/mRNA hybrid or mRNA have high affinity to
!
2
AR. This high affinity can be significantly decreased by unfused cDNA/mRNA.
Fusion containing pF30P only has no abnormally high affinity to !
2
AR. All of
these imply that cDNA/mRNA causes non-specific electrostatic interaction with
!
2
AR. The different formats of fusions were shown in the cartoon. (B) Spermine,
Spermidine, MgCl
2
, and NaCl were tested to block this non-specific electrostatic
interaction. 50 mM MgCl
2
resulted in low naïve e10FnIII library binding to BI-
!
2
AR as to M1 beads (left graph). 50 mM MgCl
2
did not change the spectra of
BI-!
2
AR in bimane assay, but other cation did (right figure), suggesting that
MgCl
2
would not change the conformation of BI-!
2
AR. Therefore, MgCl
2
was
introduced into the BI-!
2
AR selection to block non-specific electrostatic
interaction.

51
prevent the non-specific electrostatic interaction and still keep the !
2
AR functional and
structurally intact.
It is reasonable to expect that there is positively charged-rich region in
!
2
AR. Since !
2
AR contains two different modifications on Cys265, either iodoacetamide
or monobromobimane, to avoid the formation of disulfide bond, we tested their binding
to BI pool 1 (close to naïve e10FnIII library ) full fusions (cDNA/mRNA-pF30P-
e10FnIII) (Figure 2.8A). The result shows that !
2
AR with iodoacetamide has
significantly lower non-specific binding to BI pool 1 full fusion (4%) than !
2
AR with
monobromobimane (50%) (Figure 2.8A). This suggests that !
2
AR is the key contributor
for non-specific electrostatic interaction between cDNA/mRNA in fusion and mBBr-
modified !
2
AR. This electrostatic interaction can be significantly decreased by removing
mBBr from !
2
AR and using buffer containing MgCl
2
and dsDNA (Figure 2.8B).
In summary, these binding data indicate that background binding to M1 can be
avoided using the capturing method. The capturing method may partially minimize non-
specific electrostatic interaction between positively charged e10FnIII pool and !
2
AR.
Using !
2
AR without mBBr modification and MgCl
2
can significantly decrease the
electrostatic interaction between cDNA/mRNA and !
2
AR (from >50% to < 1%).
Therefore, iodoacetamide-modified !
2
AR (no mBBr) and MgCl
2
were applied to the new
BI-!
2
AR selection using the capturing method to prevent the non-specific electrostatic
interaction.

52

Figure 2.8 Monobromobimane on !
2
AR contributes to non-specific
electrostatic interaction, and MgCl
2
and dsDNA can block electrostatic
interaction.
(A) The capturing method was used to test !
2
AR that was labeled with
monobromobimane (mBBr) or none in HLS buffer with CaCl
2
. There is no other
cation in the buffer to block electrostatic interaction. The result shows that !
2
AR
with mBBr has high affinity to naïve e10FnIII fusion, suggesting that mBBr
causes the electrostatic interaction with cDNA/mRNA. (B) Efficiency of different
buffers in blocking electrostatic interaction. The binding of BI-!
2
AR without mBBr
to naïve e10FnIII fusion was tested using the capturing method in different
buffers. The HLS buffer is HLS+CaCl
2
. The result suggests that MgCl
2
and
dsDNA can block non-specific electrostatic interaction between BI-!
2
AR and
naïve e10FnIII fusion.

53
Selection against BI-bound !
2
AR using the capturing method and cation exchange
The result of six rounds of selection shows that the binding for BI-!
2
AR increased
gradually from ~0.6% of BI pool 0 to 17% of BI pool 6 with relatively low background
binding to M1 beads (Figure 2.9A). However, sequencing results show that the BI pool 6
contains mainly positively charged amino aicds (Figure 2.9B, right). This result is
consistent with the previous observation that the electrostatic interaction results mainly
from a positively charged fusion interacting with !
2
AR, not with M1 beads.
To remove these positively charged e10FnIII clones from BI pool 6, cation
exchange was exploited (Figure 2.9B, left). After two rounds of cation exchange, the pool
binding to BI-!
2
AR decreased, indicating the removal of positively charged clones that
interact with BI-!
2
AR. In addition, one round of BI-!
2
AR selection using cation
exchange as the pre-clear step was subsequently performed. The binding data
demonstrate that BI pool 9 regained the binding to BI-!
2
AR, suggesting that non-
positively charged e10FnIII clones were enriched. The difference between two pool 6
bindings done in different dates indicates the inconsistency of the radiolabeled binding
assay, so this binding assay is only used as a preliminary test to monitor the selection and
determine the approximate binding, not the exact number.
Analysis of BI pool 9 sequencing results provides evidence that cation exchange
efficiently removed the positively charged e10FnIII clones, because the number of
positively charged amino acids decreased. However, 11 of 20 sequenced clones showed a
frame-shift in which the BC loop remains intact but residues after the FG loop are

54

Figure 2.9 BI-!
2
AR pool binding and cation exchange
(A) BI-!
2
AR selection using the capturing method. Pool binding indicates low
background binding and increasing affinity to BI-!
2
AR. Two rounds of cation
exchange were done to remove positively charged fusions, which reduced BI-
!
2
AR binding. Finally, one more round of selection using the capturing with pre-
clear by cation exchange was performed. This increased the pool binding to BI-
!
2
AR (B) A cation exchange column was used to remove positively charged
e10Fn fusions as shown at left. The red box indicates the neutral e10FnIII
fusions. The probabilities of positively charged amino acids (Arg, Lys, and His)
at 16 randomized residues (6 in BC loop and 10 in FG loop) in different pools
are shown in the right. This result indicates that positively charged amino acids
increased after selection. However, two rounds of cation exchange and one
round of BI-!
2
AR selection using cation exchange as a pre-clear step caused
the number of positively charged amino acids to drop significantly in BI pool 9.
55
mutated to different amino acids (data not shown). These frame-shift products are very
similar, but their binding and function are not tested. Six clones containing less than five
positively charged amino acids and two highly positively charged clones (Fn BI 6-4 and
6-9) containing more than 5 positively charged amino acids were tested for their binding
to BI-!
2
AR (Figure 2.10A). The radiolabeled binding assay was done in two different
ways (Figure 2.10B and C). Capturing was used to test the binding and the result reveals
that highly positively charged clones show better binding to BI-!
2
AR and minimal
binding to M1 beads (Figure 2.10B). However, binding using biotinylated BI-!
2
AR
immobilized on neutravidin beads shows that one low positively charged clone (Fn BI 9-
5) has the highest affinity to BI-!
2
AR and low background binding to BI-!
2
AR (Figure
2.10C). The binding of highly positively charged clones is similar between two different
binding assays, but that of low positively charged clones is inconsistent between two
binding assays. This may indicate that the binding of low positively charged clones to BI-
!
2
AR depends on the method of immobilization of BI-!
2
AR. Based on these results, this
may indicate that electrostatic interaction is specific for BI-!
2
AR binding. However, it is
hard to understand how an electrostatic interaction can be specific for protein-protein
interaction.

Functional test of selected e10FnIII ligands using monobromobimane (mBBr)
fluorescence assay
Two clones, Fn BI 6-6 and Fn BI 9-5, were chosen to test their function in
stabilizing active !
2
AR using the mBBr fluorescence assay. These two clones were
56
Figure 2.10 BI-!
2
AR binding of selected clones.
(A) Sequences of clones that were tested for BI-!
2
AR binding. The positively
charged amino acids are shown in blue. The copy number of individual clones
and the total number of clones are indicated inside parenthesis. (B) Binding was
tested using the capturing method. Free BI-!
2
AR was incubated with an e10FnIII
clone at room temperature for 1 hour. After incubation, the complexes of BI-!
2
AR
and e10FnIII clone were captured by M1 beads. After washing with buffer, the
beads were resuspended in water and counted. The binding results show that Fn
BI 6-4 and 6-9 have higher affinity to BI-!
2
AR and lower background binding to
beads, indicating that positively charged amino acids help BI-!
2
AR binding. (C)
Binding was tested using the pull-down method with biotinylated BI-!
2
AR.
Biotinylated (yellow) BI-!
2
AR is immobilized on neutravidin (green) agarose
beads (shown as NA). An e10FnIII clone was incubated with immobilized BI-
!
2
AR on NA beads at room temperature for 1 hour. After washing with buffer, the
beads were resuspended in water and counted. The binding result also suggests
that more positively charged amino acids help binding and have low background
binding. However, Fn BI 9-5 shows the highest binding, while showing the lowest
binding in the capturing method.

57

Figure 2.10 BI-!
2
AR binding of selected clones.

58
chosen, because they show a lower background binding to M1 beads in two binding
assays and high BI-!
2
AR binding in one assay (Figure 2.10). Fn BI 6-6 and Fn BI 9-5
were both expressed in E.coli, but only Fn BI 9-5 was expressed well, purified, and tested
using the mBBr fluorescence assay.
In the mBBr fluorescence assay, the Cys265 that labeled with mBBr in TM6 of an
inactive !
2
AR is buried in a hydrophobic environment, so the environmental sensitive
fluorescent probe, mBBr is active (Figure 2.11A, left). Active mBBr has higher %max
and the emission peak is 447 nm. However, if !
2
AR is activated by an agonist, the
conformational change leads to the exposure of Cys265 to hydrophilic environment.
Thus, mBBr fluorescence would be quenched and the %max decreases and the emission
peak shifts to 462 nm. The spectra of !
2
AR interacting with Gs protein provide the
positive control and model for this assay (Figure 2.11A, right). Unfortunately, the spectra
show that Fn BI 9-5 binding to !
2
AR cannot induce a conformational change of !
2
AR to
the active state as Gs protein can (Figure2.11B). In addition, the spectra without and with
Fn BI 9-5 are the same, indicating that Fn BI 9-5 binding cannot stabilize THRX-bound
!
2
AR by preventing Alprenolol (inverse agonist) from competing with THRX for !
2
AR.
However, this result may suggest that Fn BI 9-5 could bind to a different site compared to
ligands in !
2
AR and that the binding of Fn BI 9-5 does not confer the function to either
activate !
2
AR or stabilize agonist-bound !
2
AR.

59

Figure 2.11 Fn BI 9-5 is not functional
(A) Cys 265 on the C-terminus of transmembrane helix 6 (TM6) was labeled by
a fluorophore, monobromobimane (mBBr). After !
2
AR is activated, the
fluorescence of mBBr will be quenched by the aqueous environment due to the
conformational change of !
2
AR. Thus, active !
2
AR has a lower % max and the
peak of active !
2
AR shifts to a higher wavelength. Unliganded indicates the
native !
2
AR without ligands. After adding Gs protein, isoproterenol (ISO, a full
agonist), or both, the spectrum shifts to the right and the % max drops. This
shows that Gs protein, ISO, or both can activate !
2
AR in different degrees.
[courtesy of S. G. F. Rasmussen et al., Nature 469, 175-181 (2011)] (B) The
spectra at left are the negative control with no Fn BI 9-5 and the spectra at right
are the result with Fn BI 9-5. THRX is a full agonist and Alp (Alprenolol) is an
inverse agonist. The Fn BI 9-5 shows no effect in !
2
AR conformational change.
60
Discussion
The non-specific binding between naïve e10FnIII fusions and M1 beads and !
2
AR
Two major issues for the !
2
AR selection is background binding to M1 beads and
the electrostatic interaction that leads to the positively charged e10FnIII clones that
denominate the final pools. In the first two selections that used THRX-bound and ICI-
bound !
2
AR as targets, the e10FnIII fusions that bound to M1 beads were also enriched
by selections. Although the e10FnIII fusions were pre-cleared by following through a
pre-packed M1 bead column, the background binding to M1 beads couldn’t be removed.
Even though the binding of e10FnIII fusion to !
2
AR is always 2-fold higher than to M1
beads, the non-specific e10FnIII fusion binding to M1 beads could still compete with the
potential specific e10FnIII ligands for !
2
AR binding. This could result in the loss of
specific ligands for !
2
AR.
Sequencing results of the final pools from the first two selections contain mostly
positively charged amino-acid clones. This indicates that positively charged clones had
been selected possibly due to the non-specific pull down by M1 beads. These non-
specific e10FnIII fusions that bound to M1 beads could also be PCR amplified, when the
e10FnIII clones that bound to the target were PCR amplified directly from resuspended
beads.
To minimize this non-specific background binding to M1 beads, capturing and
elution were exploited in selections. Capturing and elution efficiently eliminated most
non-specific background binding to M1 beads, but not non-specific binding of the naïve
e10FnIII to !
2
AR. However, the capturing and elution seem to decrease the binding of
61
the positively charged e10FnIII fusions to !
2
AR. Thus, we have shown that the capturing
and elution can efficiently reduce the non-specific binding between e10FnIII naïve
library and M1 beads.
The first attempt to do BI-!
2
AR selection using capturing and elution ended up no
converge for BI-!
2
AR binding. We observed low PCR cycles (~10) and low BI-!
2
AR
binding (2%) that conflict with each other. Theoretically, a lower binding means less
target-bound materials for PCR amplification, so the less materials should require higher
PCR cycles to yield enough DNA. The cause for this problem was discovered using the
different formats of fusions in !
2
AR binding assay. The full fusion (cDNA/mRNA-
pF30P-e10FnIII protein) used in the selection is different from the fusion (RNase treated)
used in the binding assay (pF30P-e10FnIII protein). The assumption to do so is that full
fusion should behave the same as RNase treated fusion, which means that cDNA/mRNA
should not affect the binding between fusions and targets. This should be the working
hypothesis for mRNA display or any other displays to work, because the attached DNA
or RNA should be only used to enrich selected peptides or proteins. Unfortunately, the
weak BI-!
2
AR binding to RNased treated fusion (2%) but the strong binding between full
fusions and BI-!
2
AR (>50%) indicate that cDNA/mRNA binds to !
2
AR non-specifically
in this selection.
We also found that !
2
AR that was labeled with monobromobimane (mBBr)
resulted in strong interaction between the naïve e10FnIII fusions and !
2
AR. Thus, the
possible non-specific electrostatic interaction results from mBBr and cDNA/mRNA.
However, this interaction can be easily minimized by removing mBBr from !
2
AR and
62
adding MgCl
2
in the binding buffer. This issue was not observed in first two selections
using the pull down method, because tRNA and BSA were added in the binding buffer to
block the non-specific interaction. tRNA and BSA are the most common blocking agents
that were used in a mRNA display. They were taken out in the first BI-!
2
AR selection
using the capturing method due to their possible adverse effects in disrupting !
2
AR
integrity.

Electrostatic interaction is the major issue for in vitro display techniques
Most of in vitro display techniques suffer from non-specific electrostatic
interaction because of the negatively charge molecules in the library designs. For
example, mRNA display and other displays using oligonucleotide-peptide or
oligonucleotide-protein fusion have the same limitation to target proteins with high or
low pI. These proteins would become either negatively charged or positively charged in
neutral pH condition. Positively charged targets could interact with the negatively
charged oligonucleotides in fusions, leading to strong non-specific electrostatic
interaction. Negatively charged targets could repel the negatively charged
oligonucleotides in fusions, preventing the real ligands binding to targets. The same
limitation was also observed in the phage display, because the library peptides or proteins
are embedded in negatively charged membranes. A previous study had shown that using
cationic polymers could efficiently block the non-specific electrostatic interaction
between M13 phages and high-PI target proteins (Lamboy et al., 2009). Therefore, the
addition of cations in selections would be a feasible method to block the non-specific
63
electrostatic interaction. We decided to use MgCl
2
in the final BI-!
2
AR selection along
with the capturing method, because MgCl
2
could block the electrostatic interaction and
did not alter the conformation and function of BI-!
2
AR.
We also showed that MgCl
2
and dsDNA together could block the electrostatic
interaction more efficiently than MgCl
2
alone. However, dsDNA could decrease %max of
mBBr-modified !
2
AR in the mBBr fluorescence assay (data not shown), indicating that
dsDNA could affect the structure of !
2
AR. Moreover, using both cations and anions in a
selection could complicate the experimental conditions. The final optimized condition in
minimizing the electrostatic interaction is to use MgCl
2
in binding buffer and
iodoacetamide-modified !
2
AR to replace mBBr-modified !
2
AR as a selection target.

Positively charged e10FnIII fusions have the highest fitness to dominate !
2
AR
selections.
Although the optimal conditions described above were adopted in the final BI-
!
2
AR selection, the sequencing results of BI pool 6 still contain the positively charged
e10FnIII clones that dominate the pool. This implies that the non-specific electrostatic
interaction between e10FnIII fusions and &'(!
2
AR could not be totally blocked by
MgCl
2
, capturing, and elution.
We then utilized a cation exchange to remove highly positively charged e10FnIII
clones and the sequencing results of BI pool 9 demonstrated that a cation exchange
worked. However, the majority of e10FnIII clones in BI pool 9 is frame-shift and the
reason remains unexplored.
64
The clone binding shows that the positively charged e10FnIII clones bound to BI-
!
2
AR better than non-positively charged e10FnIII clones did. The positively charged
e10FnIII clones also had lower background binding to M1 beads. This indicates that the
positively charged e10FnIII clones have the highest fitness in the !
2
AR selections and are
very difficult to be removed by the conditions we applied.
Fn BI 9-5 was the only clone that has different affinity when tested in two sets of
binding assays that used different methods to immobilize BI-!
2
AR on different beads. In
addition, Fn BI 9-5 has a lower background binding to two different beads than other
non-positively charged e10FnIII clones. Fn BI 9-5 was also expressed well in E.coli.
Therefore, we tested the function of Fn BI 9-5 using mBBr fluorescence assay.
Unfortunately, the result showed that Fn BI 9-5 was unable to induce the conformational
change of !
2
AR to active state and to stabilize the agonist-bound BI-!
2
AR as Gs protein
is. This indicates that the selected non-positively charged e10FnIII clones might bind to a
site on !
2
AR that could not affect the structure and function of !
2
AR.

Possible reasons for selecting positively charged e10FnIII clones in the !
2
AR
selections
It is difficult to address that the electrostatic interaction between positively
charged e10FnIII clones and !
2
AR is a specific interaction. It is also difficult to
understand that the charged molecules, tRNA, BSA, or MgCl
2
cannot block this
electrostatic interaction. It could be that we do not find the optimal conditions to totally
block this interaction. In addition, the electrostatic interaction between mBBr and
65
cDNA/mRNA can be easily blocked by these charged molecules, indicating that the
source of this electrostatic interaction might not be only from the interaction between
positively charged e10FnIII clones and !
2
AR.
For !
2
AR that is able to bind negatively charged cDNA/mRNA and positively
charged e10FnIII clones, we could assume that !
2
AR contains both negatively charged
and positively charged surfaces. The interaction between cDNA/mRNA and !
2
AR could
be weaker than that between positively charged e10FnIII clones and !
2
AR. Therefore,
MgCl
2
was able to block the interaction between cDNA/mRNA and !
2
AR. However,
after BI-!
2
AR selection, MgCl
2
was unable to block the electrostatic interaction between
positively charged e10FnIII fusions and BI-!
2
AR. The binding of e10FnIII proteins
without oligonucleotide to BI-!
2
AR was efficiently blocked by MgCl
2
. These results
suggest that 50 mM MgCl
2
can efficiently block the interaction between cDNA/mRNA
and BI-!
2
AR, but cannot block the interaction between positively charged e10FnIII
proteins and BI-!
2
AR. Using different combinations and concentrations of various
charged molecules might be able to block both electrostatic interaction, which could lead
to non-positively charged e10FnIII ligands that can bind BI-!
2
AR.
A possible model for !
2
AR to contain negatively charged and positively charged
surfaces is that some phospholipids attach to !
2
AR (Figure 2.12A). A previous study
indicated that some phospholipids could still attach to membrane proteins even after
stringent purifications (Guan et al., 2006). In addition, by inspecting the amino-acid
composition and distribution of !
2
AR, the positively charged amino acids cluster in
intercellular surface and the pI of !
2
AR is a bit high (8.58). Thus, !
2
AR should be
66

Figure 2.12 Possible reasons for the electrostatic interaction and the
background binding to M1 beads
(A) There might be charges on !
2
AR or phospholipids. A positively charged
e10FnIII fusion can bind to a positively charged !
2
AR via sandwiching with its
own cDNA/mRNA. The !
2
AR structure shows that the intercellular loops are
positively charged. Another possibility is that a positively charged e10FnIII
fusion binds to negatively charged phospholipids attached to !
2
AR. (B) CNBr
sepharose might be positively charged. After chemically activated and
inactivated, CNBr is positively charged at the neutral pH. Therefore, CNBr may
pull down a positively charged e10FnIII fusion by sandwiching with a negatively
charged cDNA/mRNA. (C) Using FLAG to immobilize !
2
AR could be
problematic. First, the affinity of anti-FLAG M1 antibody to FLAG is not high
enough to immobilize and keep N-terminal FLAG-tagged !
2
AR on M1 beads
(K
D
=0.4µM). This could result in not enough !
2
AR on M1 beads for a selection.
Secondly, anti-FLAG M1 antibody may interact with a positively charged
e10FnIII fusion. Thirdly, a positively charged e10FnIII fusion may bind to the
complex of anti-FLAG M1 antibody and !
2
AR.

67
positively charged under the selection conditions, so the negatively charged
phospholipids might be the most possible cause for selecting positively charged e10FnIII
proteins. A positively charge e10FnIII fusion may bind to negatively charged
phospholipids or sandwich with its own oligonucleotide or oligonucleotides of other
clones and positively charged !
2
AR (Figure 2.12A).
Another possibility is that a positively charge e10FnIII fusion can sandwich with
its own oligonucleotide or other oligonucleotides of other fusions and positively charged
CNBr beads (Figure 2.12B). CNBr beads were chemically activated to immobilize anti-
FLAG M1 antibodies and inactivated by ethanolamine to generate M1 beads. This
immobilization involves the attachment of M1 antibodies to sepharose by an isourea
bond, which is positively charged at the neutral pH. This suggests that using other
methods to immobilize !
2
AR on different kinds of beads could be another solution for the
non-specific electrostatic interaction. Other issue results from this immobilization could
be not enough !
2
ARs that were immobilized on M1 beads due to the low affinity between
FLAG and M1 antibody (Figure 2.12C). In addition, free anti-FLAG M1 antibodies on
M1 beads may interact with e10FnIII fusions, resulting in the e10FnIII clones that bind to
M1 antibodies. The complex of M1 antobody and !
2
AR could be another source for off-
target binding.
We had learned that the major problem in !
2
AR selections is the electrostatic
interaction and figured out some solutions to minimize this problem. We also provide
several suggestions for the !
2
AR selections in the future. The first suggest is to use a
different method to immobilize !
2
AR on a different bead. Secondly, charged molecules
68
and a cation exchange could minimize the electrostatic interaction. The third one is to use
a doped library based on known protein ligands, such as Gs proteins, rather than a
randomized naïve library. A doped library could provide some molecules with some
affinities to !
2
AR in the beginning, which could outcompete the positively charged
e10FnIII fusions.

69
Chapter 3
Developing G$s-Like Peptide Ligands Targeting Active !
2
AR
Using mRNA Display

In collaboration with Dr. Cheng Zhang and Professor Brian Kobilka, Department of
Molecular and Cellular Physiology, Stanford University

Abstract
G protein-coupled receptors (GPCRs) are the most common drug targets for the
treatment of diverse diseases, such as cardiac dysfunction, neurological disorders,
diabetes, and cancers. The limitation of current drug development efforts is targeting only
extracellular binding pocket of GPCRs, leading to the lack of specificity to either
individual GPCRs or different activity states of a single GPCR. Comparison of active and
inactive !
2
adrenergic receptor (!
2
AR) structures show that the intracellular interface
between the !
2
AR and G$s has low structural homology relative to extracellular binding
pocket. This indicates that targeting the interface between the !
2
AR and G$s may
provide a better opportunity to develop highly selective drugs for specific GPCRs, even
specific activity states.
As a proof of principle, our goal is to target the interface of active !
2
AR and G$s
for the development of active !
2
AR-specific peptide ligands. The C-terminus of G$ has
been shown to directly interact with GPCRs and determine the specificity of G proteins to
their cognate receptors. We thus generated a doped peptide library based on the C-
terminus of G$s and used mRNA display to target active !
2
AR. Active !
2
AR was
induced using high-affinity small molecule agonist BI-167107.
70
Several high copy-number clones from selection were in vitro translated and
screened by in vitro binding assay. Peptides that had high affinity in the screen were
further verified for their function in stabilizing active !
2
AR using competition assay.
Three of four peptides (G$s BI 4-2, G$s BI 6-4, and G$s BI 6-22) were shown to
stabilize active !
2
AR as Gs heterotrimer does, but they are not as effective as Gs.
Here, we show that mRNA display using a doped G$ C-terminal peptide library
is a powerful method to develop GPCR state-specific peptide ligands that can stabilize
active state of GPCRs. The method is not only useful to develop functional peptide
ligands with state-specificity to other GPCRs, but the selected peptide ligands can also be
applied in GPCR functional and structural studies or in pharmaceutical drug development
to treat various GPCR-related diseases.

71
Introduction
G protein-coupled receptors (GPCRs) are a class of important transmembrane
receptors that regulate various physiological functions of eukaryotes in response to the
environmental stimulus. GPCRs can be activated by diverse ligands, such as photons of
light, odorants, ions, hormones, and neurotransmitters, and trigger a wide range of
signaling pathways inside cells. GPCRs are the largest family of transmembrane
receptors in human genome. The GPCR family generally adopts structures containing
seven transmembrane $ helices separated by three extracellular loops and three
intracellular loops, and N-terminal and C-terminal flexible fragments (Palczewski et al.,
2000) (Figure 3.1A). Extracellular loops and the part of seven transmembrane $ helices
form ligand-binding pocket to interact with ligands outside the cell (Figure 3.1A). Upon
the ligand binding, a GPCR undergoes a conformational change to transmit a signal
inside the cell by interacting with GDP-bound G$!) heterotrimer. The interaction leads
to the exchange of GDP for GTP on G$ and the dissociation of G$-GTP and G!)
dimmer from the GPCR, activating different downstream effectors and inducing
corresponding cell responses. Malfunctions of GPCRs have been reported to result in
many diseases, such as heart disease, hypertension, obesity, asthma, cancer, and more.
The importance of GPCRs makes the GPCR family the most popular targets of drugs,
accounting for approximately 40% of all modern prescription drugs (Filmore, 2004;
Pierce et al., 2002).

72

Figure 3.1 !
2
AR structure and C-termini of G$ subunits
(A) Schematic !
2
AR strcuture is shown in the left. Three extracellular and three
intracellular loops are shown as ECL1-3 and ICL 1-3. Seven transmembrane
domains reside inside the cell membrane (TM1-7). Two disulfide bonds are
shown as yellow dashed lines. Cys265 shown in red is modified by maleimide-
PEG11-Biotin. X-ray crystal strcuture of agonist-!
2
AR-Gs ternary complex (PDB:
3DQB) in the right shows that !
2
AR (orange) bound to an agonist (blue) and
interacted with C-terminus of G$s subunit (green). (B) X-ray crystal structures of
two GPCRs (orange) coupled with their C-termini of G$ subunits (green) were
shown. The left is active opsin-G$CT peptide [ILENLKDCGLF, G$t (340-
350)K341L] complex (PDB: 3SN6) and the right is agonist-!
2
AR-Gs protein
ternary complex (Only C-terminus of G$s subunit is shown here) (PDB: 3DQB).
These two structures suggest that the C-terminus of a G$ subunit plays a
pivotal role for the interaction between GPCRs and G$ subunits by directly
contacting with GPCR.
73
Majority of marked drugs are small-molecule compounds targeting GPCR
extracellular ligand-binding sites, which mostly lack high selectivity for specific GPCR
subtypes and some could lead to unwanted side effects (Schlyer and Horuk, 2006). The
accumulated !
2
adrenergic receptor (!
2
AR) structural evidence suggests that the
conformation of extracelluar ligand-binding pocket are highly conserved, so it is difficult
to develop highly specific drugs by targeting extracellular sites of GPCRs. In contrast, the
relatively low structural homology at intracellular G$-binding site reveals that targeting
intracellular sites of GPCRs is a potentially more promising strategy to develop highly
selective drugs (Rasmussen et al., 2011b).
Targeting intracellular sites of GPCRs has been reported to be effective by
pepducin technology (Anchor Therapeutics) in vitro and in preclinical in vivo animal
models (Covic et al., 2002b). Pepducins are lipopeptides comprising N-terminal lipid
moieties (usually palmitate) and synthetic peptides (typically 10-20 amino acids) derived
from GPCR intracellular loops (Covic et al., 2002a). A pepducin flips inside the cell
through its lipid moiety, binds with its cognate GPCR, and stabilizes bound GPCR in an
active or inactive conformation by an allosteric mechanism (Covic et al., 2002a).
Although pepducins can target GPCRs with higher selectivity, they can act as agonists,
inverse agonists or both, which cannot be designed in advance. The development of
pepducins bases on trial-and-error procedure, so this strategy may not be able to result in
drugs for all GPCRs and the function of drugs could not be designed purposely.
Our goal is to develop a general strategy to evolve active GPCR-specific peptide
ligands by targeting a GPCR intracellularly, leading to selective inhibition of an active
74
GPCR. To achieve our goal, we first aim to target the less structurally conserved interface
between the GPCR and G$, and chose BI-167107 (high-affinity full agonist; Boehringer
Ingelheim)-bound !
2
AR (BI-!
2
AR) as our model active GPCR target (Rasmussen et al.,
2011a). !
2
AR is one of the most studied GPCRs, and high-quality and functional !
2
AR
can be obtained using an optimized purification procedure (Rasmussen et al., 2007).
Second, we generated a doped library based on the C-terminus of G$s to
specifically target the G$s-binding site. The C-termini of G$ subunits are critical for the
interaction between GPCRs and G proteins (Dratz et al., 1993; Hamm et al., 1988), and
make direct contacts with GPCRs. This has been observed in the co-crystal structure of
opsin with a mutated C-terminal peptide of transducin G$t subunit, and the ternary
structure of agonist-!
2
AR-Gs complex (Rasmussen et al., 2011b; Scheerer et al., 2008)
(Figure 3.1B). The C-termini of G$ subunits are also important for the specificity of
GPCR-G protein interactions (Conklin et al., 1993). Previous studies have also shown
that synthetic G$ C-terminal peptides contain G$-like function in stabilizing the active,
agonist-bound GPCRs and competitively block G proteins binding to GPCRs (Dratz et
al., 1993; Hamm et al., 1988; Rasenick et al., 1994). However, these synthetic G$ C-
terminal peptides have much lower potency and affinity than their parental G$ proteins,
indicating that wild type G$ C-terminal peptides retain specificity but lose affinity.
The C-terminus of transducin G$t has been doped and selected against its active
cognate GPCR, metarhodopsin II using “peptides on-plasmids” display to develop
functional peptide ligands with high affinity to rhodopsin (Martin et al., 1996). This
resulted in small, linear peptides (11 residues) that were more potent than the wild type
75
G$t C-terminal peptide in binding to and stabilization of active rhodopsin,
metarhodopsin II. In addition, another group was able to mutate one amino acid in the C-
terminal peptide of transducin G$t [G$CT, an 11 amino acid synthetic peptide, G$t
(340-350) K341L] and obtained a co-crystal structure of opsin and G$CT (Scheerer et
al., 2008). These two studies provided strong evidence that the C-terminus of a G$
subunit could be evolved for higher specificity and affinity to its active cognate GPCR,
leading to G$-like function in stabilization of an active GPCR. The evolved peptides
could be used as research tools and potential therapeutic agents.
However, targeting other GPCRs is still challenging due to the flexibility and
heterogeneity of non-sensory GPCRs, such as !
2
AR. Thus, in order to evolve peptide
ligands possessing high specificity and affinity to active BI-!
2
AR, we decided to use
mRNA display. mRNA display is an in vitro selection technology that has been used to
successfully develop functional peptide or protein ligands with high affinity and
specificity for various targets, such as G$ protein, Methuselah GPCR, I$B!, and more in
our lab (Austin et al., 2008; Ja et al., 2007; Olson et al., 2008). One advantage of mRNA
display is the ability to construct high complexity of libraries (>10
12
unique sequences)
because higher library complexity increases the probability to access high-affinity ligands
for a target (Ling, 2003; Perelson and Oster, 1979). Another advantage is that mRNA
display presents a simpler, more stable display platform and is free of cell transformation,
because highly stable mRNA-peptide fusions are generated via a chemically ligated
puromycin moiety.
76
Using mRNA display, we constructed a doped G$s C-terminal library and
selected for binding against active BI-!
2
AR (BI is a high-affinity full agonist). The
selection resulted in a final pool that contained peptides with high specificity and affinity
to BI-!
2
AR. Sequencing analysis of the final pool confirmed the selection convergence to
a few sequences and several potential positive clones appeared in the final pool with high
copy number. In vitro binding tests of high-copy number peptides showed that the C-
terminal FLAG sequence, cloned into peptide library for purification, enhances the
binding of peptide ligands to BI-!
2
AR. By removing the C-terminal FLAG tags from the
peptide ligands, our results showed two groups of peptide ligands: The C-terminal
FLAG-dependent and the non-C-terminal FLAG-dependent. The C-terminal FLAG-
dependent peptide ligands required a C-terminal FLAG tag to bind BI-!
2
AR, as removal
of the FLAG tags there is no binding at all. However, the non-C-terminal FLAG-
dependent peptide ligands still bound when the C-terminal FLAG tags were removed, but
the binding dropped ~3-5 fold.
The non-C-terminal FLAG-dependent peptides were synthesized by solid phase
peptide synthesis and screened for G$s-like function by the isoproterenol (ISO, an
agonistof !
2
AR) competition assay (Swaminath et al., 2005; Whorton et al., 2007). Of the
four peptides tested in the ISO competition assay, three peptide ligands have G$s-like
function in stabilizing active ISO-bound !
2
AR. While these peptides are small (only 22
amino acids), they have activity comparable to the full-length G$s subunit (394 amino
acids) and are only ~5-fold less effective.
77
In summary, we have demonstrated a strategy to evolve functional peptides with
high affinity and specificity to the active state of an agonist-bound GPCR. Our strategy
includes securing active conformation of GPCR with high-affinity agonist, generating a
doped peptide library based on the C-terminus of G$ subunit, and performing mRNA
display. This resulted in peptides that were first screened by in vitro binding assay for
high-affinity peptides that were further tested in ISO competition assay for G$s-like
function. Finally, the high-affinity peptides having G$s-like function can finally be
verified in cellular functional assays for the ability to manipulate specific cell activities
by binding to active !
2
AR. Here, we are not only able to develop G$s-like functional
peptides that can be a lead for a potential drug targeting active !
2
AR or used for
structural studies, but also demonstrate a valid and promising strategy to develop
functional peptides with high specificity and affinity targeting GPCRs.

Materials and Methods
*xpression, purification, and biotinylation of !
2
AR
Biotinylated !
2
adrenergic receptors (!
2
ARs) were kindly produced by Dr. Cheng
Zhang in Dr. Brian Kobilka’s lab. The cloning, expression and purification of !
2
AR was
previously described (Kobilka, 1995; Rasmussen et al., 2007). Briefly, a human !
2
AR
construct was truncated after residue 365 (!
2
AR365N), expressed in Sf9 insect cells
infected with recombinant baculovirus, and solubilized in n-dodecyl-!-D-maltoside
(DDM). Pure and functional !
2
AR was obtained by the first M1 FLAG affinity
chromatography (Sigma) and followed by alprenolol-sepharose chromatography.
78
Subsequently, !
2
AR-bound alprenolol was exchanged for high-affinity agonist BI-
167107 by the second M1 FLAG affinity chromatography. BI-bound !
2
AR was eluted in
HLS buffer (20 mM HEPES-KOH pH 7.5, 100 mM NaCl, and 0.1% DDM) with 10 µM
BI-167107. BI-!
2
AR was biotinylated at cysteine 265 (Cys265) using maleimide-
activated reagent maleimide-PEG11-Biotin (Thermo Scientific) and purified using a
desalting column (Figure 3.1A). Biotinylated BI-!
2
ARs was concentrated in HLS buffer
(20 mM HEPES-KOH pH 7.5, 100 mM NaCl, and 0.1% DDM) with 10% glycerol and
stored at -80
o
C in small aliquots.

Construction of the doped G$s C-terminal peptide library
The doped G$s C-terminal peptide library was designed as MG-X
5
-
QRMHLRQYELL-X
4
-GSGSGSSDYKDDDDK (where X is any of the 20 amino acids)
with ~40-50% conservation in the core 11 G$s C-terminal amino acids (Figure 3.2A).
The antisense DNA template [5’-GCT GGA GCC ACT GCC AGA TCC (566)
4
512 512
343 241 342 532 512 242 714 532 342 (566)
5
GCC CAT TGT AAT TGT AAA TAG
TAA TTG TCC C-3’; 1=70% A: 10% G: 10% C: 10% T; 2=10% A: 70% G: 10% C:
10% T; 3=10% A: 10% G: 70% C: 10% T; 4=10% A: 10% G: 10% C: 70% T; 5=0% A:
50% G: 50% C: 0% T; 6=25% A: 25% G: 25% C: 25% T; 7=3.33% A: 3.33% G: 90% C:
3.33% T] was synthesized by Keck Oligonucleotide Synthesis facility and PCR amplified
with 5’-primer 47T7FP (5’-GGA TTC TAA TAC GAC TCA CTA TAG GGA CAA
TTA CTA TTT ACA ATT AC-3’) and 3’-primer 3GsDopeFLAG (3X9GsFLAG)(5’-
GCC CTT GTC ATC GTC GTC TTT GTA GTC GCT GGA GCC ACT GCC AGA
79
Figure 3.2 Design of a doped peptide library based on the C-terminus of
G$s subunit and the selection against active BI-bound !
2
AR
(A) The alignment of C-terminal sequences of different G$ proteins (left). Red X
letter indicates the randomized flanking region. Blue letters represent the doped
residues. The structure of G$CT and opsin in the right shows that the possible
position of our G$s doped peptide library interacts with the target. The G$s
doped peptide is color-coded same as sequences. Puromycin, DNA linker, and
cDNA/mRNA are indicated in purple sphere, grey single line, and grey/black
double strand, respectively. [The figure of G$CT-opsin structure is adopted from
Scheerer P. et al., Nature 455, 497-502 (2008)]. The finalized library design
contains 5 randomized N-terminal residues (red), 11 doped residues (blue), 4 C-
terminal randomized residues (red), GS linker (black), and FLAG tag
(underlined). The percentage of wt G$s amino acids for each position is indicated
in the lower table. FLAG tag is inserted for purification of fusion (Library was
designed and constructed by Dr. Terry Takahashi). (B) First 4 rounds of the
selection used streptavidin (orange) tosyl magnetic beads (brown)(labeled as
SA) to immobilize biotinylated BI-!
2
AR. Biotin is shown in yellow and labeled as
B. Doped G$s C-terminal peptide library is shown as a fusion in which RNA is in
black, cDNA is in dark gray, DNA linker is in light gray, and peptide with FLAG
are shown in red (FLAG is indicated.). Pre-clear by cation exchange was
introduced to the third round of the selection for excluding positively charged
fusions. (C) Fifth round of the selection switched the beads to neutravidin
80
agarose beads. Neutravidin is shown in green and an agarose bead is shown in
gray and labeled as NA.

Figure 3.2 Design of a doped peptide library based on the C-terminus of
G$s subunit and the selection against active BI-bound !
2
AR
81
TCC-5’) (Integrated DNA Technologies, Inc.) to generate the double-strand DNA
(dsDNA) template for the doped library. The purified dsDNA was comprised with a T7
promoter, a GGG transcription start sequence, "TMV, and open reading frame (ORF) of
MG-X
5
-QRMHLRQYELL-X
4
-GSGSGSSDYKDDDDK for in vitro transcription and
translation. The purified dsDNA was then subjected to in vitro transcription to produce
mRNA and the purified mRNA was ligated with pF30P [DNA-puromycin linker,
8(A)
21
(7)
3
ACC6; 8 is chemical phosphorylation reagent, 7 is spacer phosphoramidite 9,
and 6 is puromycin CPG; Keck Oligonucleotide Synthesis facility] to create mRNA-
pF30P as described previously (Takahashi and Roberts, 2009), except that the splint
BakFLAGSplint (5’-TTTTTTTTTTTTNGCCCTTGTCATC-3’) was used in ligation.

mRNA display selection
The first round of selection was started with 100 µL of cold in vitro translation
supplemented with methionine and 125 µL of hot in vitro translation supplemented with
L
-[
35
S]methionine (MP Biomedicals) to produce mRNA-pF30P-protein fusions of the
doped G$s C-terminal peptide library using reticulocyte lysates (Green Hectares). Cold
and hot fusions were then dT purified and reverse transcribed (RT) separately as
previously described (Takahashi and Roberts, 2009). A subsequent FLAG purification
was used to purify cold and hot fusions with C-terminal FLAG-tag using anti-FLAG M2
affinity agarose gel (Sigma), and eluted twice with 100 µL and 125 µL of 0.1 mg/ml 3X
FLAG peptide (Sigma) in HLS buffer, individually. FLAG purification can remove free
oligonucleotides to prevent their non-specific electrostatic interaction with BI-!
2
AR to
82
fail the selection (see Chapter 2). 100 µL of purified hot fusions were mixed with 100 µL
purified cold fusions and subsequently subjected to BI-!
2
AR selection. The remaining 25
µL of purified hot fusions were used for monitoring the non-specific binding of G$s BI
pool to beads without targets to be a no target control.
Biotinylated BI-!
2
ARs were immobilized on homemade streptavidin (SA)-tosyl
magnetic beads (Dynabeads M-280 tosylactivated, Invitrogen) freshly prior to BI-!
2
AR
selection (Figure 3.2B). SA-tosyl magnetic beads were prepared as manufacturer’s
instructions, in which streptavidin (SA) was incubated with tosyl-activated magnetic
beads at room temperature overnight, washed, and stored at 4
o
C. The concentration of BI-
!
2
AR immobilized on SA-tosyl magnetic beads was around 4.7 µM.
Cold and hot mixed library fusions were incubated with BI-!
2
AR immobilized
SA-tosyl magnetic beads in selection buffer (20 mM HEPES-KOH pH 7.5, 100 mM
NaCl, 0.1% DDM, 0.5 mg/mL BSA, and 0.1mg/mL tRNA) at room temperature for 1
hour. The beads were then washed with selection buffer for 4 times at room temperature
and the bound fusions were eluted with 0.15% SDS in ddH
2
O twice. 1/100th of the
elution was counted in a scintillation counter to estimate the binding, while the rest was
treated with SDS-Out SDS Precipitation Reagent (Thermo Scientific) following the
manufacturer's instructions to remove SDS for optimal PCR amplification. Treated
elution was PCR amplified and the purified PCR product was designated as G$s BI pool
1. In parallel, hot library fusion was subjected to the same procedure, except that this
sample was incubated with empty beads to monitor the non-specific background binding
as a no target control. The flowthrough, washes, elution, and beads from both BI-!
2
AR
83
and empty bead binding were counted in a scintillation counter. The binding percentage
was calculated by the ratio of each individual elution and bead counts divided by the total
counts added (the sum of flowthrough, wash, elution, and bead counts). This binding
percentage represented the binding of G$s BI pool 0 (original doped G$s C-terminal
peptide library) to BI-!
2
AR.
The same selection protocol was applied for the following four rounds, except
that the volume of hot translation was reduced to 50 µL (25 µL mixed with cold and
another 25 µL used for a control empty bead binding). Cation exchange was introduced
in third round (Figure 3.2B), and was used as a pre-clear step to eliminate any positively
charged peptides that could contribute to background binding to !
2
AR (see Chapter 2).
Cation exchange was performed by flowing FLAG-purified library fusion through a
column that was composed of HiTrap SP HP beads (GE Healthcare) and collecting the
flowthrough for the selection. In round five, the matrix was switched from SA-tosyl
magnetic beads to neutravidin (NA) agarose beads (Thermo Scientific) (Figure 3.2C).
Switching different matrices for target immobilization between different rounds of a
selection has been shown to significantly eliminate the background binding to matrix in
our lab (Takahashi and Roberts, 2009).
The G$s BI pool 4 and G$s BI pool 5 were cloned using TOPO TA cloning kit
(Invitrogen) and clones were verified to contain inserts by EcoRI (NEB) digestion and
running digested products on a 2% agarose gel. The positive clones with inserts were
sequenced (Laragen, Inc.) and the sequences were analyzed using ClustalW2 (EMBL-
EBI) (Goujon et al., 2010; Larkin et al., 2007) and weblogo (Crooks et al., 2004;
84
Schneider and Stephens, 1990). The clones with high sequence homology were grouped
(only 1 or 2 mutations allowed) and each high homologous group was only count on one
input in weblogo analysis. The percentage copy number of each major clones and total
singleton in pool 4 and 5 was calculated and plotted. Three major clones with higher copy
number in pool 4 were tested for their binding to BI-!
2
AR in an in vitro binding assay
(see below). This assay showed one clone to be C-terminal FLAG-dependent, since it lost
binding without C-terminal FLAG.
To get rid of C-terminal FLAG-dependent peptides, one more round of selection
was performed using the pool 5 library where the FLAG tag was removed (Figure 3.4B).
This was done by PCR amplifying the library with a 3’-primer 3GSNoFLAG (5’-
gCTggAgCCACTgCCAgATCC-3’). Selection steps were the same as previous round,
except that FLAG purification was omitted. The G$s BI pool 6 was cloned, sequenced,
and analyzed as with pool 4 and 5, except that the 3GSNoFLAG primer was used instead
of the 3GsDopeFLAG primer. The five most dominant clones in pool 6 were tested for
the binding by in vitro binding assay as described below.

In vitro binding assay
Selected clones in pool 4, G$s BI 4-1, 4-2, 4-3, were PCR amplified with 5’-
primer 47T7FP and 3’-primer 3GsDopeFLAG for transcription and translation. Purified
PCR products were used as DNA templates for in vitro transcription to make mRNAs,
and the mRNAs were subsequently gel purified. The purified mRNA of each clone was
subjected to in vitro translation supplemented with
L
-[
35
S]methionine and the products
85
were then FLAG purified. Purified peptides were incubated with BI-!
2
AR immobilized
on neutravidin (NA) agarose beads in selection buffer at room temperature for 1 hour,
washed 4 times, and resuspended in ddH
2
O. Flowthrough, washes, and beads were
counted in a scintillation counter (Beckman Coulter, Inc.) and the percentage binding was
estimated by diving bead counts with total counts (sum of flowthrough, washes, and
beads). G$s BI 4-2, 4-3, 6-4, 6-13, 6-22 were five major high-copy number clones in the
pool 6, which were verified for their specificity and affinity to BI-!
2
AR using similar in
vitro binding assay, except that 3GSNoFLAG primer was used for PCR amplification and
the FLAG purification was omitted. After in vitro translation, the reaction was directly
incubated with BI-!
2
AR immobilized on NA agarose beads and analyzed as previously.
The C-terminal FLAG-dependent G$s BI 4-1 was used as a negative control.

Peptide synthesis
Peptides with similar or higher percentage binding to BI-!
2
AR than negative
control G$s BI 4-1 and no background bead binding in the in vitro binding assay were
synthesized using PS3 Peptide Synthesizer (Protein Technologies, Inc.).
Wang resin (AnaSpec, Inc.) pre-loaded with the last Fmoc-amino acid of the
peptide was weighed to synthesize peptide at the 1.0 mmol scale, and swelled with
dichloromethane (DCM) for at least 10 min. Pre-packed vials containing 1.0 mmol of
each different Fmoc-amino acid supplemented with 1 mmol of HBTU (Protein
Technologies, Inc.) or 1 mmol of HATU (AnaSpec, Inc. and AAPPTec) were placed on
the peptide synthesizer. Swelled Wang resin was transferred into a reaction vessel
86
mounted on peptide synthesizer and washed 3 times with N-Methyl-2-pyrrolidone
(NMP). The program for peptide synthesis was set as described in the manufacturer’s
instructions and the amino acid coupling time was set 30 min.
After peptide synthesis, the resin was removed from the reaction vessel and
transferred to a 15 mL centrifuge tube and dried on a lyophilizer. Peptides were cleaved
from resin by incubating dried resin with 94% trifluoroacetic acid (TFA) and scavengers
[2.5% 1,2-Ethanedithiol (EDT), 1% triisopropylsilane (TIS), and 2.5% ddH
2
O] at room
temperature for 2 hours. Ice-cold ether was first added to precipitate the peptide and the
white peptide pellet was then washed with ice-cold ether twice. The peptide pellet was
dried on a lyophilizer and dissolved in DMSO. Resin and dissolved peptide were
separated by filtration and the peptide was purified via HPLC (Beckman Coulter, Inc.)
using a C18 Vydac column (W. R. Grace & Co.-Conn.). After HPLC purification, the
fractions collected from the peak were evaluated with matrix assisted laser desorption
ionization-time of flight (MALDI-TOF) mass spectrometry using Voyager DE-STR
MALDI-TOF mass spectrometer (Applied Biosystems). The fractions containing correct
molecular weight/number of charges on ion (m/z) were pooled together and dried with
lyophilizer. Purified peptides were dissolved in DMSO to make highly concentrated
stocks and the concentration was determined with O.D.280 measured by
spectrophotometer (Nanodrop). The peptide stocks were kept in -20
o
C to prevent
oxidation.

87
Isoprenol (ISO) competition assay
High-density lipoprotein (HDL) particles with reconstituted !
2
AR365N were
generated and purified as described previously (Whorton et al., 2007). Gs was
reconstituted into !
2
AR-HDL to serve as a positive control for this assay as described
previously (Whorton et al., 2007). !
2
AR-HDL with Gs, with 100 µM peptide, or without
peptide were incubated with 10 nM [
3
H]-labeled Alprenolol ([
3
H]DHA; an inverse
agonist) and various concentrations of Isoprenol (ISO; an agonist) (from 10
-10
M to 10
-3

M) at room temperature for 30 min. The samples were then harvested by passing through
Whatman GF/B filter paper. The amount of [
3
H]DHA bound to !
2
AR-HDL was
measured in a scintillation counter. The data were normalized and fit to a two-site
competition binding model using Prism (GraphPad).

Results
Targeting active BI-!
2
AR by mRNA display using a doped G$s C-terminal library
Functional !
2
AR was coupled with high-affinity agonist, BI-167107 to generated
active conformation of !
2
AR as our selection target (Figure 3.1A). !
2
AR is labile and
heterogeneous, but !
2
AR bound with BI would become more stable and have higher
conformational homogeneity, allowing the selection to target only active conformation of
!
2
AR. BI-!
2
AR was also biotinylated at Cys265 in the cytoplasmic end of TM6 using
sulfhydryl-reactive maleimide-PEG11-biotin for immobilization of BI-!
2
AR (Figure
3.1A). Cys265 has been reported to be one of the most reactive cysteines and the
specificity of modification at Cys265 by a sulfhydryl-reactive agent was confirmed
88
(Ghanouni et al., 2001). Due to the truncation of other two reactive cysteines in C-
terminus of full-length !
2
AR, the biotinylation of Cys265 in our !
2
AR365N should be
specific. As shown in Chapter 2, the method used for immobilization of targets on
matrices is extremely important for the selection targeting !
2
AR to success. In addition,
utilizing biotin to immobilize targets has been shown to consistently give low background
binding and lead to success in various selections in our lab, so we generated biotinylated
BI-!
2
AR to be immobilized on matrices for the selection.
To develop ligands targeting BI-!
2
AR with specificity and functionality, we
designed a doped peptide library based on the C-terminus of G$s subunit, providing
several advantages. First, a doped G$s C-terminal library allows us to specifically target
the interaction between BI-!
2
AR and G$s (Figure 3.1B), indicating that the ligands are
evolved to modulate this particular signaling pathway. Since GPCR activation can induce
different signaling pathways, legands that are capable to target only one signaling
pathway would serve as powerful research tools and potential drugs. Secondly, the C-
terminus of G$ subunit is important for specific GPCR binding and directly interacts
with GPCR to stabilize active conformation of GPCR (Figure 3.2A). Therefore, a doped
G$s C-terminal peptide library may contain potential peptide ligands with specificity and
function as G$s subunit, increasing the possibility to obtain peptide ligands targeting BI-
!
2
AR. This doped strategy is also indirectly confirmed by our unsuccessful attempts to
develop protein ligands targeting BI-!
2
AR using naïve library (see Chapter 2).
Thirdly, the crystal structure of C-terminus of G$ subunit shows that the end of
C-terminus is not spatial restricted (Figure 3.2A). Thus, it may be feasible to put extra
89
amino acids and oligonueleotides in C-terminus of G$ subunit for mRNA display and
would not affect the interaction between GPCR and G$ subunit. We thus designed a
doped G$s C-terminal peptide library by doping ~ 40-50% of 11 amino acids in the C-
terminus of G$s subunit (Figure 3.2A). The doped residues were flanked with five
randomized amino-acid residues in N-terminus and four randomized amino-acid residues
in C-terminus to increase the complexity, which may lead to selected peptide ligands with
high specificity and affinity (Austin et al., 2008). A Gly-Ser (GS) linker was introduced
to provide space between a FLAG tag and a doped sequence. A FLAG tag was
engineered in C-terminus of the doped library to remove the large quantity of excess
uncoupled oligonucleotides (cDNA/mRNA), since cDNA/mRNA had been shown to
have non-specific electrostatic interaction with !
2
AR in Chapter 2. This non-specific
electrostatic interaction from !
2
AR also resulted in obtaining the majority of positively
charged proteins in the !
2
AR selection (see Chapter 2). Therefore, we also introduced
cation exchange in the third round of BI-!
2
AR selection to remove positively charged
peptides in the pool (Figure 3.2B).

Pool’s BI-!
2
AR binding was evolved by selection
After four rounds of selection, the BI-!
2
AR binding increased from 1.5% (G$s BI
pool 0) to 31.2% (G$s BI pool 4). The selection started with low background binding
(Figure 3.3A), however, the background bead binding also increased (from 1.1% of G$s
BI pool 0 to 3.8% of G$s BI pool 4). To correct this, the beads used for immobilization
were switched from SA-tosyl magnetic beads to NA-agarose beads in fifth round of
90
Figure 3.3 Pool binding of mRNA display against BI-!
2
AR and the
sequencing results of pool 4 and 5.
(A) Pool binding was monitored by in vitro binding assay using
L
-[
35
S]-labeled
fusion. BI-!
2
AR immobilized on two different matrices are labeled. Pool 5 resulted
from switching immobilized matrix from streptavidin magnetic bead (SA) to
neutravidin agarose bead (NA) as described in Figure 3.2C. The binding of pool
5 to BI-!
2
AR was tested on SA and NA as indicated here. (B) Sequencing results
of pool 4 and pool 5. Only sequences of high-frequency (show up twice in
sequencing results) clones are shown in bold followed with their mutants where
the mutations are in red. The doped region is underlined. There is the %
frequency of each major clone in the right of the sequences. The decrease of
singleton’s % frequency indicates the convergence of selection. The
conservation of amino acids in each position was analyzed by weblogo and
doped amino acid residues are boxed. Q8, M10, L12, Y15, E16, L17, L18 are
conserved to wild type G$s C-terminus.
91

Figure 3.3 Pool binding of mRNA display against BI-!
2
AR and the
sequencing results of pool 4 and 5.
92
selection. This not only decreased the background binding of the library to SA-tosyl
magnetic beads to 1.3%, close to the level of the starting pool, but also increased the BI-
!
2
AR binding to 47.5%. The background bead binding was even lower (0.5%) and BI-
!
2
AR binding was even higher (79.8%) when NA-agarose beads were used instead of
SA-tosyl magnetic beads (Figure 3.3A). The in vitro pool binding data demonstrated that
mRNA display against BI-!
2
AR using a doped G$s C-terminal library was able to
successfully evolve the pool to bind BI-!
2
AR.
G$s BI pool 4 and G$s BI pool 5 were sequenced and the results revealed three
things (Figure 3.3B). First, new clones that arose in G$s BI pool 5 were not present in
G$s BI pool 4. One explanation of this might be due to the small sample size of
sequenced clones (16 clones from G$s BI pool 4 and 39 clones from G$s BI pool 5).
Secondly, the fraction of singletons dropped from 62.5% (G$s BI pool 4) to 18% (G$s
BI pool 5). The increase of high-copy number clones and the decrease of singletons
indicate that the selection reduced the pool complexity and enriched specific clones with
higher affinity. Thirdly, several amino acids in the doped region are conserved, but many
are not. Among the doped amino acids, Q8, M10, L12, Y15, E16, L17, and L18 are
highly conserved, implying that these amino acids are important for the C-terminus of the
G$s subunit interacting with BI-!
2
AR. Pervious studies had reported that L12 (L388 in
the full-length G$s protein) and L17 (L393 in the full-length G$s protein) are conserved
in 12 G$ proteins (Dursi et al., 2002). Therefore, L12 and L17 might be conserved across
all G$ proteins for GPCR binding, but other amino acids remained conserved after
selection might contribute to the specificity of G$s interacting with BI-!
2
AR.
93
C-terminal FLAG tag significantly enhances clone binding to BI-!
2
AR
G$s BI 4-1, G$s BI 4-2, and G$s BI 4-3 peptides with C-terminal FLAG tags
were tested in an in vitro binding assay against BI-!
2
AR immobilized on NA-agarose
beads. The results show that all three have excellent affinities to BI-!
2
AR (Figure 3.4A).
G$s BI 4-3 without FLAG (G$s BI 4-3) was synthesized by solid phase peptide
synthesis and added to an in vitro binding assay to compete
35
S-labeled G$s BI 4-3 with
FLAG (G$s BI 4-3-FLAG), but no competition was observed (data not shown). We
hypothesized that this difference was due to C-terminal FLAG tag, and these data
suggested that G$s BI 4-3-FLAG has much stronger affinity than G$s BI 4-3. To test this
hypothesis, a N-terminal FLAG was cloned on to G$s BI 4-1, G$s BI 4-2, and G$s BI 4-
3 peptides. These N-terminal FLAG tagged peptides were compared to their C-terminal
FLAG tagged counterparts in an in vitro binding assay. The data showed that the N-
terminal FLAG tagged peptides bind significantly worst to BI-!
2
AR, compared to the
same peptides with a C-terminal FLAG tag (Figure 3.4A), supporting the hypothesis that
a C-terminal FLAG tag indeed dramatically enhances selected peptide binding to BI-
!
2
AR.
In addition, G$s BI 4-1 lost almost all binding to BI-!
2
AR when the C-terminal
FLAG was removed. This indicates that G$s BI 4-1 binding is C-terminal FLAG-
dependent or that FLAG can bind to BI-!
2
AR.However, FLAG itself was tested in the
same binding assay, but showed no binding to BI-!
2
AR (data not shown). Thus, a
selection using a C-terminal FLAG-tagged peptide library might enrich the peptides with
C-terminal FLAG (peptide-FLAG) with high affinity to BI-!
2
AR. It is surprising that
94

Figure 3.4 The C-terminal FLAG issue and solution
(A) Three clones in the pool 4 with either a N-terminal or C-terminal FLAG are
picked to test their binding to BI-!
2
AR. The sequences of three clones and wild
type (wt) are shown in the top. The bold amino acids are conserved to wild type
G$s C-terminus. (B) The sixth round of selection was done using the library
without a FLAG-tag as shown here. BI-!
2
AR was immobilized on neutravidin
agarose beads. After dT purification, ~5% is fusion and ~95% is pF30P-
cDNA/mRNA as shown.

95
C-terminal FLAG plays such an important role in BI-!
2
AR binding, because FLAG has
been introduced to C-terminus of peptide library and protein library for purification of
fusions without causing any problems (Cho et al., 2000).
Originally, the C-terminal FLAG was introduced to the doped G$s C-terminal
library for purification of fusions and removal of cDNA/mRNA to prevent non-specific
electrostatic interaction (This is the major issue in !
2
AR selection that was discussed in
Chapter 2). Even through C-terminal FLAG might solve the issue of electrostatic
interactions between cDNA/mRNA to BI-!
2
AR, there are two other adverse effects
caused by C-terminal FLAG in the BI-!
2
AR selection. One is to select for peptides that
require C-terminal FLAG tags for BI-!
2
AR binding, but that have almost no binding by
themselves (the C-terminal FLAG-dependent). This may result in several of the C-
terminal FLAG-dependent peptides to outcompete the high-affinity peptide ligands that
could bind without C-terminal FLAG. Another is that the C-terminal FLAG tag was co-
evolved with peptides to have additive effect for all the peptides, even those can bind to
BI-!
2
AR without a C-terminal FLAG tag (the non-C-terminal FLAG-dependent). This
selection might potentially exclude specific peptides with high affinity to BI-!
2
AR that
bind without a C-terminal FLAG tag due to the C-terminal FLAG-dependent peptides.
However, non-C-terminal FLAG-dependent peptides might contain specific and high-
affinity peptides for BI-!
2
AR. Therefore, it is important to adjust selection condition to
enrich peptides that bind to BI-!
2
AR without a C-terminal FLAG tag.
There are still non-C-terminal FLAG-dependent peptides in G$s pool 5, such as
G$s BI 4-2 and G$s BI 4-3 that bind to BI-!
2
AR without a C-terminal FLAG tag. This
96
suggests that G$s BI pool 5 may contain two species: the C-terminal FLAG-dependent
and non-C-terminal FLAG-dependent. Thus, we hypothesized that it might be possible to
enrich peptides that bind to BI-!
2
AR without a C-terminal FLAG tag from the non-C-
terminal FLAG-dependent peptides. In addition, because FLAG tags (DYKDDDDK) are
highly charged, it seems likely that any C-terminal FLAG association is due to a non-
specific electrostatic interaction. For example, an interaction between a FLAG tag and the
intracellular loop 3 (ICL3, Figure 3.1A) of !
2
AR containing positively charged residues
seems possible. Therefore, we performed an additional round of selection where we
removed the FLAG tag.

C-terminal FLAG-dependent peptides were removed by selection using pool without
the FLAG tag
One more round of selection against BI-!
2
AR was done using G$s BI pool 5
without FLAG tag. The FLAG tag was removed by PCR amplifying pool 5 using a 3’-
primer that has no FLAG tag. dT purified G$s BI pool 5 without FLAG tag was directly
subjected to selection (Figure 3.4B). The binding of dT purified G$s BI pool 6 without
FLAG tag increased, indicating that this selection enriched peptides that bind to BI-!
2
AR
without FLAG tag (Figure 3.5A). However, RNase treated and dT purified clones in G$s
BI pool 6 without FLAG tag bound ~ 90% to BI-!
2
AR, including the negative control,
G$s BI 4-1 without FLAG tag (data not shown). G$s BI 4-1 is a C-terminal FLAG-
dependent peptide, so it should not bind to BI-!
2
AR without FLAG tag. RNase treated
and dT purified G$s BI 4-1 without FLAG tag had pF30P that attached to the C-terminus
97

Figure 3.5 Pool binding using different formats of fusions
(A) Pool 5 and pool 6 without FLAG were tested for BI-!
2
AR binding. The three
different formats of pool 5 were used as indicated. Pool 6 incresed binding to BI-
!
2
AR indicates that selection works. pF30P-mRNA/cDNA blocks fusion of pool 5
binding to BI-!
2
AR. pF30P-mRNA/cDNA interfers the binding of fusion with
FLAG more than fusion without FLAG. (B) Binding of different formats of pool 5
to BI-!
2
AR. First, this result supports the last two ideas from A. Secondly,
peptide without FLAG binds to BI-!
2
AR stronger than peptide with FLAG. This
suggests that the residues after wild type G$s C-terminus are important for
electrostatic interaction. In both experiments, BI-!
2
ARs were immobilized on NA
beads.

98
of the peptide and replaced the C-terminal FLAG tag in dT and FLAG purified G$s BI 4-
1 with FLAG. pF30P (oligonucleotides) contains higher negative charge than FLAG tag,
so we hypothesized that the particular space adjacent to C-terminus of G$s BI peptides
are important for the non-specific electrostatic interaction.
To test our hypothesis, we compared the BI-!
2
AR binding of dT purified G$s BI
pool 5 without a FLAG tag (Bar 1 from the left) to that with a FLAG tag (Bar 3 from the
left) (Figure 3.5A). The results show that dT purified G$s BI pool 5 without a FLAG tag
binds to BI-!
2
AR better than that with a FLAG tag, supporting our hypothesis that highly
negatively charged pF30P in this particular space increases the non-specific electrostatic
interaction. In addition, dT and FLAG purified G$s BI pool 5 with a FLAG tag (Bar 4
from the left) has highest affinity to BI-!
2
AR, which suggests that excess amount of
unfused pF30P-mRNA/cDNA (95% vs 5% fusion) could block the non-specific
electrostatic interaction. The distinct products after different purifications or treatments of
fusion reaction are shown as cartoons in the bar graph (Figure 3.5).
To confirm the importance of the particular position adjacent to C-terminus of
selected peptides, G$s BI pool 5 without and with FLAG tags were treated with RNase
and dT purified before binding to BI-!
2
AR. The results show that G$s BI pool 5-pF30P
has higher affinity (81.5% of pool 5, RNase) than G$s BI pool 5-FLAG-pF30P (64.9% of
pool 5-FLAG, RNase) (Figure 3.5B). This provides strong evidences that replacing
FLAG with higher negatively charged pF30P in the same particular position adjacent to
C-terminus of selected improve the binding to BI-!
2
AR via non-specific electrostatic
interaction.
99
Taken together, the exact C-terminal position that contributes to electrostatic
interaction starts in the 12th residue after C-terminus of wt G$s protein (YELL) and
approximately lasts for 8-residue long (as shown in bold and underlined region;
YELLXXXXGSGSGSSDYKDDDDK). The first five rounds of the BI-!
2
AR selection
using peptide fusion with a C-terminal FLAG tag in this position could have two
advantages. One is to avoid non-specific electrostatic interaction from a stronger pF30P
in the same position that might outcompete peptide ligands that bind to BI-!
2
AR without
pF30P. Another is that FLAG purification can remove excess amount of pF30P-
mRNA/cDNA that might also compete with peptide ligands for BI-!
2
AR binding via
non-specific electrostatic interaction (This is the key obstacle for !
2
AR selection that was
discussed in Chapter 2). Therefore, G$s BI pool 5 contains the non-C-terminal FLAG-
dependent peptides that would be the potential peptide ligands for BI-!
2
AR.
The sixth round of selection using G$s BI pool 5 without the FLAG tag could
have a positive effect to remove the C-terminal FLAG-dependent peptides by blocking
the non-specific electrostatic interaction with pF30P-mRNA/cDNA. This was supported
by that the binding of dT purified G$s BI pool 6 is higher than the binding of dT purified
G$s BI pool 5 (Figure 3.5A). In addition, this can be confirmed with the sequencing
results of G$s BI pool 6 where the C-terminal FLAG-dependent peptide, G$s BI 4-1,
was eliminated (Figure 3.6A). The sample sizes are almost the same (39 sequences in
G$s BI pool 5 and 37 sequences in G$s BI pool 6), so the chance of G$s BI 4-1 in both
pools should be the same if it can survive the selection (This should exclude the bias from
the sample size). Since we hypothesized that the non-C-terminal FLAG-dependent
100

Figure 3.6 Sequening results of pool 6 and the pool 6 clone binding
(A) The sequences of pool 6 were aligned. The clones having higher copy
number in the pool 6 are shown in bold, and the % frequency of each major
clones and singleton is indicated. The mutants of each major clone are shown
adjacent to their related clones and the mutations are shown in red. Underlined
residues were doped. Weblogo analysis shows the conservation of amino acids
in each position. Q8, M10, L12, Y15, E16, L17, L18 are conserved to wild type
G$s C-terminus, but they are not as conserved as pool 4 and 5, except E16. (B)
Raw radiation counts for the binding of selected peptides to BI-!
2
AR
immobilized on NA beads. The data shows decent binding for selected clones.

101
peptides could include high-affinity and functional peptide ligands to BI-!
2
AR, the
clones with high copy number in G$s BI pool 6 (Figure 3.6A) were screened for their
binding to BI-!
2
AR by an in vitro binding assay.

Three G$s-like functional peptides were developed
Due to additive effects of the C-terminal FLAG tag and pF30P increasing peptide
binding to BI-!
2
AR, we verified the binding of selected peptides in G$s BI pool 6 using
an in vitro binding assay (Figure 3.6B). We translated and
35
S-labeled several clones in
vitro (G$s BI 4-2, 4-3, 6-4, 6-13, and 6-22) and measured their binding to BI-!
2
AR.
These peptides bind better than a wt G$s C-terminal peptide
(MGCRDIIQRMHLRQYELLGSGSGSS) based on the C-terminus of G$s. However, the
binding was raw radiation counts of radiolabeled peptides pulled down by beads, not the
percentage binding of pull-down/total input peptide counts. These data may not represent
the actual percent binding, because the more radiolabeled peptides pulled down may
result from a higher input of total
35
S-labeled peptides. It would be more accurate to
estimate the binding by knowing the total input and normalizing for the amount of added
peptides. We had difficulty quantifying the amount of input peptides since we are unable
to purify these peptides using FLAG or pF30P and the translated reaction is full of
unreacted hot
L
-[
35
S]methionine. Alternatively, there are two ways to roughly estimate
the input of translated peptides, TCA precipitation or densitometry analysis of translated
peptide bands in tricine gel. We attempted to use both methods to get the total input of
translated peptides, but the TCA precipitation always gave lower counts than pull-down
102
bead counts (input < output) and we were unable to resolve the translated peptides via
tricine gel.
However, based on the raw in vitro binding data, G$s BI 4-2, 4-3, 6-4, 6-22 had
higher or equal affinity when compared to G$s BI 4-1 (a C-terminal FLAG-dependent
peptide). Thus, these four peptides were synthesized without a C-terminal FLAG tag and
tested in an ISO competition assay (Figure 3.7A). All selected peptides except G$s BI 4-
3 efficiently increased the exchange rate of Alprenolol ([
3
H]DHA; an inverse agonist)
with ISO (Isoprenol; an agonist). This indicates that our peptides selectively bind and
stabilize active ISO-bound !
2
AR rather than the inactive Alp-bound !
2
AR (Figure 3.7B).
Although these G$s-like functional peptides are not as effective as Gs protein, the most
potent peptide, G$s BI 6-22 is about 5-fold less effective than Gs protein. This is
impressive that a small linear peptides can function similarly to Gs protein and much
better than non-functional wt G$s C-terminal peptide (data not shown). This supports the
concept that mRNA display is a powerful tool able to evolve G$s-like functional peptide
from non-functional wt G$s C-terminal peptide.

Discussion
mRNA display targeting BI-!
2
AR resulted in functional G$s-like peptides
After six rounds of selection, we isolated several peptides that bind BI-!
2
AR with
high affinity. Several clones (G$s BI 4-2, 4-3, 6-4, 6-13, and 6-22) were observed to
have higher copy number in the final pool (Gas BI pool 6), and bound to BI-!
2
AR with
better affinity than the wild type G$s C-terminal peptide. These results suggest that we
103

Figure 3.7 Function of peptide ligands in stabilizing agonist-bound !
2
AR
(A) The cartoon depicts the ISO competition assay used to test the function of
peptide ligands. !
2
AR (blue) was bound by radiolabeled inverse agonist
([
3
H]DHA) and followed by competition using agonist (ISO) under three different
conditions: with Gs (red), with peptide (purple), or with nothing (control). Gs and
peptide stabilized ISO-bound !
2
AR could drive the equilibrium toward ISO-
bound !
2
AR, resulting in that competition curve shifts to low ISO concentration.
(B) ISO competition assay shows that G$s BI 6-22, G$s BI 4-2, and G$s BI 6-4
were capable to stabilize agonist-bound !
2
AR in HDL, but not as potent as Gs
protein. G$s BI 4-3 was unable to exert the same function as Gs protein. G$s BI
6-22 is the most potent peptide and only ~5-fold less effective than Gs protein.
The data for each sample is color-coded and the legends are on the side. Error
bars represent the standard errors.

104
successfully evolved high-affinity peptides that bind BI-!
2
AR after six rounds of mRNA
display. However, we also observed that improvement of affinity does not guarantee
G$s-like function, so a second screen, based on ISO competition assay, must be
performed to verify G$s-like function. In the ISO competition assay, G$s BI 4-2, 6-4,
and 6-22 were able to stabilize isoprenol-bound !
2
AR, resulting in that isoprenol
outcompeted alprenolol for the binding to !
2
AR. In contrast, G$s BI 4-3 had no G$s-like
function, supporting that the affinity does not guarantee the function. Although the
function of three positive peptides is the same as full-length G$s subunit, they are 5-fold
less potent than G$s subunit in stabilizing agonist-bound !
2
AR. In comparison, a peptide
based on the wild type G$s C-terminal peptide has low affinity to !
2
AR and is non-
functional (data not shown).
Using a doped peptide library based on C-terminus of G$ subunit to target GPCR
was also shown here to be valid strategy to develop peptide ligands that may function as
G$ subunits. The C-terminus of G$ subunits have been reported to be crucial for the
specificity of a G$ subunit to its cognate GPCR and stabilize active conformation of its
bound GPCR. However, the wild type peptides based on C-terminuns of G$ subunits (wt
G$ C-terminal peptide) could lose the affinity and function. Therefore, we demonstrated
that mRNA display using a doped G$s C-terminal library against BI-!
2
AR is a powerful
strategy to evolve peptide ligands with high affinity and G$s-like function for BI-!
2
AR.
This overcomes the problems of low-affinity and nonfunctional wt G$ C-terminal
peptides and expends the application of wt G$ C-terminal peptides in targeting GPCRs.

105
Several key elements to minimize non-specific electrostatic interaction in the !
2
AR
selection
Previous unsuccessful attempts to isolate ligands against !
2
AR using in vitro
selection suggests that the major difficulty is due to the high net charge of !
2
AR, either of
the receptor itself or surrounding phospholipids (discussed in Chapter 2). This resulted in
isolation of highly-positively charged clones in several unsuccessful !
2
AR selections.
Therefore, we applied some modifications we learned from previous !
2
AR selections and
introduced new methods to minimize this non-specific electrostatic interaction between
target and library.
First, we could minimize the highly-positively charged peptides in the library
using a doped library and cation exchange. A doped peptide library should have less
highly-positively charged peptides when compared to naïve library used in previous
!
2
AR selections (randomized library, see Chapter 2). Cation exchange was shown to be
effective in removing the positively charged clones in the selected pool (Chapter 2), so
we included cation exchange in our selection here. Secondly, a FLAG tag was engineered
into the C-terminus of our doped G$s C-terminal peptide library to exclude unfused
pF30P-mRNA/cDNA and leave fusions in the selection after FLAG purification. This
could prevent an excess amount of pF30P-mRNA/cDNA interacting with BI-!
2
AR
electrostatically that may interfere the binding of relatively few potential ligands to BI-
!
2
AR. This adverse effect has been seen in previous !
2
AR selections (Chapter 2).
Thirdly, the matrices and the immobilization methods are also important for a !
2
AR
selection. Previously, the N-terminal FLAG tagged !
2
ARs were immobilized by M2
106
antibodies that were chemically coupled on CNBr-activated sepharose, which is
problematic (Chapter 2). Thus, we used biotinylated BI-!
2
AR that can be easily and
stably immobilized on any matrices with streptavidin or neutravidin, enabling the
methods to remove background matrix binding by switching matrices.

The C-terminal FLAG tag increases selected peptide binding to BI-!
2
AR but no
effect in peptide function
All of above avoided the isolation of positively charged peptides and high
background matrix binding, leading to the neutral peptide ligands with high affinity to
!
2
AR. However, the C-terminal FLAG tag was found to significantly increase its tagged
peptide ligands binding to BI-!
2
AR. The binding test separated the selected peptide
ligands into two groups, the C-terminal FLAG-dependent and non-C-terminal FLAG-
dependent. The C-terminal FLAG-dependent peptides required a C-terminal FLAG tag
for BI-!
2
AR, and the non-C-terminal FLAG-dependent can still bind to BI-!
2
AR but
worst without a C-terminal FLAG tag. This suggests that this selection strategy evolved
the peptide with C-terminal FLAG (peptide-FLAG) for BI-!
2
AR binding, since FLAG
alone could not bind to BI-!
2
AR. However, in ISO competition assay, neither G$s BI 4-3
without a FLAG tag nor G$s BI 4-3 with a FLAG tag had any effect, suggesting that the
C-terminal FLAG tag does not have function (data not shown). This also was confirmed
by that G$s BI 4-2 without a FLAG tag is more potent in stabilizing ISO-bound !
2
AR
than G$s BI 4-2 with a FLAG tag is (data not shown). This indicates that the FLAG tag
107
has an additive effect in selected peptides binding to !
2
AR, but has no significant effect
on the function of selected peptides.

Residues after the C-terminus of peptide ligands are sensitive to electrostatic
interaction
The interaction between a peptide ligand with a C-terminal FLAG tag and BI-
!
2
AR can be partially blocked by negatively charged pF30P-mRNA/cDNA, so this
suggests that the C-terminal FLAG tag binding is a non-specific electrostatic interaction.
In G$s BI pool 5 without a FLAG tag, the C-terminal pF30P occupied the same position
as a C-terminal FLAG tag in G$s BI pool 5 with a FLAG tag. This construct showed
stronger binding than that with a FLAG tag, supporting that pF30P has stronger
electrostatic interaction than a FLAG tag in C-terminus due to its stronger negatively
charged than a FLAG tag. This also suggests that the electrostatic interaction is non-
specific and the particular residues in the C-terminus after wild type G$s C-terminal
peptide is sensitive for the electrostatic interaction with !
2
AR.
Since a C-terminal FLAG tag has weaker electrostatic effect than a C-terminal
pF30P in this charge-sensitive region, so it might be beneficial to have a FLAG tag
instead of a pF30P in the beginning of BI-!
2
AR selection. This could lead to the selection
of the non-C-terminal FLAG-dependent peptides by avoiding them to be outcompeted by
stronger C-terminal pF30P-dependent peptides. However, in sixth round of selection, the
C-terminal pF30P binding to BI-!
2
AR can be blocked by excess amount of pF30P-
mRNA/cDNA, increasing the peptide ligands that do not required any tag for BI-!
2
AR
108
binding. The positive effect of this strategy was supported by the increase of G$s BI pool
6 binding to BI-!
2
AR. The elimination of G$s BI 4-1 (a C-terminal FLAG-dependent
peptide) in G$s BI pool 6 also confirmed that our strategy is useful to select the peptide
ligands with high affinity to !
2
AR.

A new strategy was developed to targeting GPCRs
mRNA display using doped G$s C-terminal peptide library, in vitro binding assay
screening, and ISO competition verification were demonstrated to be a valid strategy to
develop G$s-like functional peptides with high affinity for agonist-bound !
2
AR. It is
challenging to study GPCRs, because of their flexible structures and transient dynamics
among different activity states. Our strategy provides a possible solution to develop
peptide ligands targeting all the GPCRs. In addition, the G$s-like functional peptides
were evolved from a doped G$s C-terminal peptide library using mRNA display.
For structural studies, these functional peptides can be exploited for co-
crystallization of active !
2
AR to understand the dynamics of !
2
AR conformational
change upon Gs protein binding. The assumption is that these peptides might induce a
transition state of !
2
AR that is different from Gs protein-bound !
2
AR and free !
2
AR.
This could decode the mechanism of interaction between Gs protein and !
2
AR. These
peptides can also be re-engineered back to G$s subunit to replace original wild type G$s
C-terminus for the stronger interaction between G$s subunit and !
2
AR. This C-terminal
modified G$s subunit could be utilized to solve the co-crystal structure of nucleotide-
bound G$s subunit and !
2
AR that has never been solved.
109
These peptides can be potential drugs to manipulate the function of active !
2
AR
with high selectively. The major advantage of these peptides as drugs is their specificity
and affinity to target active state of !
2
AR and Gs signaling pathway, which would be
more effective and have less side effect by off-target interaction. These peptides are being
tested in cellular cAMP assay for their specificity and biological effect.

110
Chapter 4
Novel Vaccines via Dual-Specific Antibody-Mimetics
Against Mouse and Human DC-SIGNs

In collaboration with Liang Xiao and Professor Pin Wang, Department of Chemical
Engineering and Materials Science, University of Southern California

Abstract
A dendritic cell (DC)-based vaccine has been approved by the US FDA and
shows promise as an immunotherapy for the prostate cancer. However, this DC-based
vaccine involves the adoptive transfer of autologous DCs that are loaded with cancer
antigens ex vivo, which is tedious and expensive. Alternatively, direct loading of antigens
on DCs via the DC surface receptors in vivo can revolutionize the current design of DC-
based vaccines and provide a new therapeutic opportunity.
DC-Specific Intercellular adhesion molecule 3 (ICAM3)-Grabbing Non-integrin
(DC-SIGN) is the most potent DC-specific receptor for antigen uptake and transport. The
extracellular domains of mouse and human DC-SIGN (mDC-SIGN and hDC-SIGN) were
generated as targets to evolve DC-SIGN specific ligands. These ligands were generated
using mRNA display of an expression enhanced 10th human fibronectin type III domain
(e10FnIII)-scaffold library.
A new strategy was implemented to evolve dual-specific ligands for mDC-SIGN
and hDC-SIGN. First, three rounds of selection were performed on mDC-SIGN followed
by four rounds of selection against hDC-SIGN. This resulted in one ligand, M3H4-18,
with high affinity to hDC-SIGN expressing cells but very low affinity to mDC-SIGN
expressing cells. M3H4-18 was also shown to facilitate internalization of an antigenic
111
peptide, which resulted in immunization of human peripheral blood mononuclear cell
(hPBMC)-derived immature dendritic cells (iDCs).
The function of mouse DC-SIGN has not been fully studied due to the lack of
good antibodies. We have developed two mDC-SIGN-specific ligands, M7-7 and M7-18.
These two ligands were able to bind mDC-SIGN with high specificity and affinity,
leading to internalization of bound ligands. These two ligands are possible alternatives
for an antibody to study mDC-SIGN in details where there are still some controversies
about its function, especially internalization.
Here, we developed a new strategy to select functional ligands targeting
membrane receptors and evolved an antibody-mimetic ligand for the development of an
in vivo DC-based cancer vaccine in humans.

112
Introduction
Active immunization by prophylactic vaccines has been a great successful
medicine to prevent infectious diseases (Plotkin, 2005). Recent success in developing
prophylactic vaccines against viral pathogens known to cause cancer opens the door for
the development of cancer vaccines (Harper et al., 2006). In cancer, the immune system
unable to recognize malignant cells, thus the goal of a cancer vaccine is to induce the
immune system to mount an anticancer immune response (Nauts, 1989). Cancer vaccines
are an intriguing strategy for cancer immunotherapy due to their high specificity, low
toxicity, and low levels of recurrence (Emens, 2008). Despite the success of prophylactic
cancer vaccines, the development of therapeutic cancer vaccines to treat existing cancer
still remains limited (Lesterhuis et al., 2011; Rosenberg et al., 2004). However, because
of great advances in understating the role of dendritic cells (DCs) in innate and adaptive
immunities, DCs have become the major target for the development of novel vaccines,
including potential therapeutic cancer vaccines (Banchereau and Palucka, 2005;
Banchereau and Steinman, 1998; Gilboa, 2007)
DCs are specialized antigen presenting cells (APCs) that can uptake antigens and
process them for presentation on the major histocompatibility complex (MHC) and
activate naïve T cells (Mellman and Steinman, 2001). DCs have been successfully
exploited to develop DC-based vaccines for protective immunity against bacterial, viral,
and fungal infections (Fajardo-Moser et al., 2008). The development of DC-based
vaccines has been the major focus for cancer immunotherapy. DCs have been loaded with
tumor antigens ex vivo or in vivo to develop therapeutic DC-based cancer vaccines,
113
resulting in adaptive antitumor immunity (Gilboa, 2007; Tacken et al., 2007). With Ex
vivo loading, the tumor antigen is loaded onto autologous monocyte-derived DCs and
readministered back to the patient. This strategy resulted in the first FDA-approved
cancer vaccine, sipuleucel-T. However, the tedious procedure for generating the vaccine,
high cost ($93,000 USD), and modest increases in survival (an average of 4.1 months)
could limit its broad adoption (Hammerstrom et al., 2011; Plosker, 2011). An alternative
strategy that could be economic and more potent is to load tumor antigens directly on
DCs in vivo. This can be achieved by targeting DC-specific surface receptors that
facilitate internalization of the bound antigens for antigen presentation (Tacken et al.,
2007).
C-type lectin receptors (CLR) are the most well-studied family of proteins that
have the specific expression on APCs and have the unique capability to capture and
endocytose antigens (Figdor et al., 2002). Although several CLRs had been reported as
promising targets for loading antigens on DCs in vivo, dendritic cell-specific ICAM3-
grabbing non-integrin (DC-SIGN/CD209) is the most promising target due to two
advantages. First, of all the CLRs, DC-SIGN is the most DC-specific receptor, and
therefore targeting DC-SIGN will likely result in higher targeting efficiencies and less
side effects (Geijtenbeek et al., 2000). Secondly, antigen loading using an anti-DC-SIGN
antibody elicited a response of both naïve and memory T cells in vitro (Tacken et al.,
2005). The fact that both naïve and memory T cells were activated indicates that DC-
SIGN has multiple functions in the regulation of DCs for innate and adaptive immunities.
114
While targeting DC-SIGN is a promising approach to a cancer vaccine, some
issues must still be resolved. First, although antibody-mediated antigen loading is
efficient and safe, but the cost would be high and the constructions of various antigens on
humanized antibody would be extremely laborious (Dakappagari et al., 2006; Tacken et
al., 2005). Secondly, no exact functional orthologue of human DC-SIGN (hDC-SIGN)
has been validated in mouse DCs. This makes the evaluation of a DC-SIGN specific
vaccine more difficult since mice cannot be used as an animal model for preclinical
experiments. Although multiple mouse homologs of hDC-SIGN have been identified by
their sequence similarities and genetic loci, either these homologs exert different cell
expression patterns or do not show the same functionality as hDC-SIGN (Nagaoka et al.,
2010; Park et al., 2001; Powlesland et al., 2006). However, these observations need to be
examined in more detail due to the lack of good antibodies, especially against mouse DC-
SIGN (CD209a/CIRE) (Cheong et al., 2010). This particular hDC-SIGN homologue was
named mouse DC-SIGN (mDC-SIGN), because it has the highest homology to hDC-
SIGN. mDC-SIGN localizes syntenically to hDC-SIGN and expresses on a similar subset
of DCs as hDC-SIGN does (Park et al., 2001). However, previous studies have reported
conflicting results of mDC-SIGN’s ability to internalize a bound ligand (Powlesland et
al., 2006; Takahara et al., 2004). Thus, dual-specific antibody-mimetic ligands against
mDC-SIGN and hDC-SIGN are needed to provide cheap and less laborious alternatives
for an antibody. These dual-specific ligands also could be tested in mice for pre-clinical
evaluation, which makes mouse a suitable animal model for DC-SIGN-based vaccines.
115
Our strategy is to evolve DC-SIGN specific ligands using mRNA display of an
antibody-mimetic e10FnIII library. mRNA display is a powerful in vitro selection
technique that can evolve antibody-mimetic ligands with high specificity and affinity to
targets of interest. Fibronectin libraries (e.g., the 10FnIII library) of >10
12
molecules can
be generated using mRNA display (Olson et al., 2008; Xu et al., 2002). In order to
improve the expression and stability of the 10FnIII library, a new e10FnIII library was
derived from wild type 10FnIII (Olson et al., 2011). This scaffold is mostly of human
origin, which could decrease the likelihood of undesirable immune responses versus
fibronectin ligands during in vivo vaccination in humans. The extracellular domains of
mDC-SIGN and hDC-SIGN were chosen as targets to evolve functional ligands with
higher dual-specificity for DC-based vaccination (Figure 4.1).
To develop dual-specific ligands targeting mDC-SIGN and hDC-SIGN, we
decided to do selection against mDC-SIGN first for two reasons. One is that mDC-SIGN
contains one carbohydrate recognition domain (CRD) and only half repeat, which is
structurally simpler than hDC-SIGN and could be used as a better starting target for proof
of concept. Another is that hDC-SIGN contains one CRD as mDC-SIGN, but seven and
half 23 amino acid repeats. Thus, targeting hDC-SIGN first may result in ligands that
bind to long-repeat region rather than CRD, so these ligands may not be able to bind
mDC-SIGN due to its lack of long-repeat region. The better strategy would select the
pool with low affinity to mDC-SIGN from naïve library, which may contain some dual-
specific ligands via targeting the high homologous CRD. This pool is subsequently used
for the selection against hDC-SIGN to evolve dual-specificity.
116

Figure 4.1 Domains and the alignment of amino acid sequences of human
DC-SIGN (hDC-SIGN) and mouse DC-SIGN (mDC-SIGN)
The cartoon on top represents the different domains of hDC-SIGN and mDC-
SIGN proteins where individual domains are indicated and color-coded. TM:
transmembrane domain. CRD: Carbohydrate recognition domain. Amino acid
sequences of hDC-SIGN and mDC-SIGN that were used as selection targets
are aligned and the different domains are color-coded as indicated in the
cartoon. Neck repeats of hDC-SIGN are aligned and assigned with numbers
listed along with repeats. hDC-SIGN: CD209 (Uniprot: Q9NNX6). mDC-SIGN:
CD209a/CIRE (Uniprot: Q2TA59).

117
First, we performed seven rounds of selection against mDC-SIGN using a naïve
e10FnIII library. This selection resulted in four clones that appeared multiple times in
final pool (M7 pool). These four ligands were tested for binding and showed high
specificity and affinity to mDC-SIGN with no binding to beads without target or hDC-
SIGN. Three of these four ligands bound to mDC-SIGN overexpressed on 293T cells and
induced internalization of the bound ligands. We have thus demonstrated the evolution of
antibody-mimetic protein ligands that bind mDC-SIGN with high specificity and affinity,
and the binding leads to internalization. Furthermore, we have shown that mDC-SIGN is
capable of capturing and uptaking the antigen. Our results support the previous
observations that anti-V5 tag antibody could bind to V5 C-terminal tagged mDC-SIGN
expressed on HEK293T cells and internalize into cells, but contradict the results of no
internalization when mDC-SIGN expressed on Rat-6 fibroblasts was bound by sugar
ligands (Powlesland et al., 2006; Takahara et al., 2004). Thus, these mDC-SIGN-specific
ligands could be a powerful tool to further study the function and expression of mDC-
SIGN.
To evolve dual-specific ligands for mDC-SIGN and hDC-SIGN, we performed
four rounds of hDC-SIGN selection using the pool three from the mDC-SIGN selection
(M3 pool) as a starting pool (Figure 4.2). We chose to start with the M3 pool in order to
preserve mDC-SIGN ligands because the pool binding shows this pool with some level of
affinity to mDC-SIGN. The selection for ligands that bind both mDC-SIGN and hDC-
SIGN resulted in two clones with high-copy numbers and three singletons that bound to
both proteins with high affinity. Interestingly, all five ligands bind much tighter to
118

Figure 4.2 Selection scheme
(A) mDC-SIGNs immobilized on streptavidin acrylamide (SA) beads were used
for 2 rounds of mRNA display starting with a naïve e10FnIII library. mDC-SIGN
pool 2 (M2) was subjected to mDC-SIGN immobilized on neutravidin agarose
(NA) for another 5 rounds of mRNA display to get mDC-SIGN specific ligands.
The acryamide bead is shown in dark gray while the streptavidin is shown in
orange. Biotinylated mDC-SIGN is represented by a yellow biotin-tag, brown
neck region, and blue CRD. (B) mRNA display targeting hDC-SIGN utilized
mDC-SIGN pool 3 (M3) for further selection against hDC-SIGNs immobilized on
neutravidin agarose beads (NA). Agarose is shown in light gray and neutravidin
is shown in green. Four more rounds of mRNA display were completed against
human DC-SIGN for evolving dual-specific ligands of mDC-SIGN and hDC-
SIGN.

119
mDC-SIGN than hDC-SIGN, indicating that the ligands binding to mDC-SIGN could be
enriched by selection against hDC-SIGN. This might suggest that the hDC-SIGN and
mDC-SIGN have a similar domain for dual-specific ligands to bind. The fact that our
ligands bind both human and mouse DC-SIGN supports our strategy to evolve dual-
specific ligands for homologous proteins from different species.
However, when we tested these dual-specific ligands for binding to either mDC-
SIGN or hDC-SIGN overexpressed on 293T cells, only one of five ligands, M3H4-18,
was functional. M3H4-18 showed high affinity to hDC-SIGN overexpressed on 293T
cells (99%) and weak affinity to mDC-SIGN overexpressd on 293T cells (4.8%). One
reason why we only observed one functional clone is that the soluble extracellular DC-
SIGN targets were expressed in E.coli for selection, which may differ in conformation or
modification from the full-length DC-SIGNs expressed on 293T cells. Another
possibility is that e10FnIII ligands were expressed and folded better in reticulocyte lysate
than in E.coli. Thus, we may not have enough functional e10FnIII ligands that were used
in cell binding assay when compared to those used in radiolabeled binding assay. Directly
targeting DC-SIGN overexpressed on 293T cells coupled with selection steps for
evolving better expressed and folded e10FnIII library in E.coli would solve both issues.
This may lead to the dual-specific ligands with high affinity to mDC-SIGN and hDC-
SIGN.
M3H4-18 was further tested in human peripheral blood mononuclear cell
(hPBMC)–derived immature dendritic cells (iDCs) for their capability in binding,
internalization, and immunization. Our data demonstrated that M3H4-18 is able to
120
harness hDC-SIGN for internalization via binding to hDC-SIGN. The immunization test
showed that M3H4-18 coupled with a C-terminal antigen (HLA-A*0201 influenza
peptide, a MHC class I-restricted peptide) successfully increased IFN-) levels over the
controls. Since activated CD8
+
cells could produce IFN-), these data suggest that CD8
+

cells were activated by iDCs containing MHC class I receptors that were loaded with
MHC I-specific antigen via the interaction between M3H4-18 and hDC-SIGN.
Thus, we were able to evolve a dual-specific clone (M3H4-18) from dual
selections against mDC-SIGN and then hDC-SIGN. M3H4-18 elicited dual-specificity
and good affinity to mDC-SIGN and hDC-SIGN in in vitro binding assays but was only
able to bind to hDC-SIGN in a cell-binding assay. By binding to hDC-SIGN, M3H4-18
could deliver an antigen to DCs resulting in internalization and antigen presentation. We
had fused an antigen to the C-terminus of M3H4-18 and shown that the antigen could be
loaded onto MHC molecules of DCs and activate naïve T cells for the induction of an
antigen-specific immune response. As a control, we tested M3H4-18 that was not
conjugated to an antigen, and we do not observe an immune response. This suggests that
the M3H4-18 fibronectin itself would not likely induce an immune response.
Taken together, we have developed three mDC-SIGN-specific ligands (M7-7,
M7-18, and M7-19) and one hDC-SIGN-specific ligand (M3H4-18). Ligands only
specific for mDC-SIGN can be exploited as markers to study mDC-SIGN. M3H4-18 can
serve as a marker and an immunization tool. M3H4-18 could efficiently target hDC-
SIGN for antigen presentation, leading to naïve T cell activation and antigen-specific
immune response without a non-specific immune response toward M3H4-18. M3H4-18,
121
an hDC-SIGN-specific ligand, provides a revolutionary tool for immunization, which
could lead to novel vaccines for variety of diseases, especially cancer.

Materials and Methods
e10FnIII library construction
Construction of an antibody-mimetic e10FnIII library using eight oligonucleotides
synthesized by the Yale Keck Oligonucleotide Synthesis facility or Integrated DNA
Technologies has been described previously in the Chapter 2 of this thesis. Briefly, the
e10FnIII library differs from the human 10th fibronectin type III domain as the e10FnIII
library has an N-terminal truncation, 5 mutations in the 10FnIII scaffold, and the last
position of BC loop doped at a lower frequency. The V5K, A6E, T8S, L12I, and L13Q
mutations were introduced in the N-terminal constant region for better expression and
stability. The doped last position of BC loop was limited to Leucine, Isoleucine, and
Valine for higher structural stability (Olson et al., 2011).
eFnoligo1 (5’-TTC TAA TAC GAC TCA CTA TAG GGA CAA TTA CTA TTT
ACA ATT ACA ATG CTC GAG GTC AAG G-3’), eFnoligo2 (5’-CA ATT ACA ATG
CTC GAG GTC AAG GAA GCA TCA CCA ACC AGC ATC CAG ATC AGC TGG-
3’), and eFnoligo3 (5’-ACC AGC ATC CAG ATC AGC TGG 55S 55S 55S 55S 55S 55S
VTT CGC TAC TAC CGC ATC ACC TAC G-3’; where 5 indicated the dNTP mixture
of 20%T, 30%C, 30%A, and 20%C; S indicates 60%C and 40%G; and V denoted the
mixture of C, A, and G), were newly designed for introducing the mutations and the
doped last position of BC loop into e10FnIII library. A Klenow reaction, followed by
122
PCR by T7 polymerase, restriction digestion by Bsa I, and ligation by T4 DNA ligase
were performed to construct the e10FnIII library. All products were purified by agarose
gel electrophoresis. The complexity of ligated e10FnIII library is ~10
12
(1 trillion unique
molecules). The library was further PCR amplified to increase the copy number of each
unique sequence and then aliquoted into small fractions. Each aliquot contains 5 copies
of the 10
12
independent sequences.

Cloning, expression, and purification of mDC-SIGN and hDC-SIGN proteins
The cDNAs for the extracellular domain of mDC-SIGN and hDC-SIGN were
PCR amplified from plasmids FUW-mDCSIGN and FUW-hDCSIGN (Yang et al.,
2008). The PCR reaction also introduced a 5’ Eco RI restriction site, a 3’ Bam HI
restriction site, and a BirA biotinylation recognition sequence to the 5’ end after the
restriction site. The cDNAs were cloned into the pET302/NT-His vector by restriction
digestions and ligations to create an N-terminal His-BirA biotinylation recognition
sequence followed by the DC-SIGN extracellular domain proteins. Both mDC-SIGN and
hDC-SIGN clones were confirmed by DNA sequencing and freshly transformed into
BL21/DE3 bacteria for expression. The transformed BL21/DE3 bacteria were inoculated
in 1 liter of Luria

Broth supplemented with 100 %g/mL of ampicillin and 10 %g/mL of
chloramphenicol and grown at 37 °C with shaking. When the growth of bacteria reached
the mid-log phase at 37 °C, protein expression was induced

with isopropyl- -D-
thiogalactoside at a final concentration of

1 mM. In order to enhance biotinylation, free
biotin was added to a final concentration of 50 µM. The induced cultures

were incubated
123
for another 4 hours before the cells were harvested

by centrifuging at 4000 x g for 10 min
at 4 °C. The resulting

cell pellet was resuspended in 15 mL of 100 mM NaH
2
P
4
, 10 mM
Tris-HCl,

and 6 M guanidine HCl pH 8, and then lysed by French press. The lysate was

supplemented with 0.01% -mercaptoethanol and incubated at 4

°C for 2 hours and then
centrifuged at 150,000 x

g for 30 min at 4 °C in a Beckman JA-25 rotor. The

supernatant
was incubated with 800 %L of nickel-nitrilotriacetic

acid-agarose resins (Qiagen,
Valencia, CA) (pre-equilibrated

with denaturing buffer) at 4 °C overnight. The resins
were loaded onto a chromatography column and all subsequent

washes were done with a
10-fold resin-volume excess of wash

buffer (30 mM Tris-HCl pH 8, 0.5 M NaCl, 1 mM

CaCl
2
, 6 M urea, and 10 mM imidazole). The column was washed

again with same buffer
except 15 mM imidazole was used instead of 10 mM imidazole. Successive

washes using
30 mM Tris-HCl pH 8, 0.5 M NaCl with decreasing concentrations

of urea starting from
5 M urea were performed to renature the

proteins. The renatured proteins were eluted
with 30 mM Tris-HCl pH 8, 0.5

M NaCl, 1 mM CaCl
2
, and 1 M imidazole and exchanged
into selection buffer using a PD-10 column. The full-length mDC-SIGN and hDC-SIGN
proteins, and amino-acid sequences of mDC-SIGN and hDC-SIGN used for mRNA
display are shown in Figure 4.1.

mRNA display
The first round of selection was started with a 1.5 mL PCR of the naïve e10FnIII
library to obtain a library complexity of 10 copies of 10
12
independent sequences. The
PCR product was used as the template in a 1.5 mL in vitro transcription reaction using T7
124
RNA polymerase at 37
o
C for 2 hours to generate mRNA. The transcription reaction was
terminated with the addition of 1/10th of the reaction volume of 0.5M EDTA pH 8.0. The
RNA was purified by Urea PAGE electrophoresis, electroelution, and ethanol
precipitation. Purified RNA was ligated with the DNA linker-puromycin, pF30P (5’-
phospho-A
21
-9
3
-ACC-Pu; where 9 is phosphoramidite spacer 9, Pu is puromycin CPG,
and the 5’ end was phosphated using chemical phosphorylation reagent; Glen Research
Corp.), with a splint oligo, FN-pF30P-Splint (5’-TTTTTTTTTTTTGGAGCCGCTACC-
3’, which is complementary to 3’-library RNA and 5’-pF30P), and T4 DNA ligase in a
1.0 mL reaction at room temperature for 1 hour. The library of mRNA-protein fusions
was created via a 2.5 mL in vitro translation reaction in rabbit reticulocyte lysate (Green
Hectares; salts and buffers from Novagen) where purified RNA-pF30P was translated at
30
o
C for an hour. To enhance fusion formation, 10 µL of salt mix (1M MgCl : 2.5M KCl
= 2:7) was added to each 25 µL of translation and incubated at room temperature for 15
min. Fusions were purified with Oligo(dT) Cellulose Type 7 (GE Healthcare Life
Science) in dT buffer (100 mM Tris-HCl pH 8.0, 1M NaCl, 0.2% Triton X-100, and 1
mM EDTA) at 4
o
C for an hour and then eluted with room temperature ddH
2
O. The
elution, containing purified fusions, was desalted and exchanged with 1X first strand
buffer (50 mM Tris-HCl pH 8.3, 75 mM KCl, and 3 mM MgCl
2
) using a 5 mL NAP-25
column (GE Healthcare Life Science). The purified fusions were then reverse transcribed
with reverse primer Fnoligo10 (5’-GGA GCC GCT ACC CTT ATC GTC G-3’) using
Superscript II enzyme (Invitrogen) at 42
o
C for an hour to generate cDNA/mRNA-protein
fusions for selection.
125
N-terminal biotin-tagged mDC-SIGN proteins were immobilized on acrylamide-
streptavidin beads (Pierce) at room temperature for 1 hour immediately prior to the
selection. mDC-SIGN immobilized beads were resuspended in selection buffer [50mM
Tris-HCl pH8.0, 150 mM NaCl, 0.02%(v/v) Tween-20, 0.5 mg/mL BSA, and 0.1 mg/mL
tRNA] and incubated with the cDNA/mRNA-protein fusion library at 4
o
C for 1 hour.
After incubation, the beads were washed 4 times with selection buffer and subjected to
PCR to amplify bound fusions. The PCR product was labeled as pool 1 (M1), and used to
make the fusions for the next round of selection. From the second round of selection, the
volume of translation reaction was reduced to 100 µL and FLAG purification was
introduced to purify full-length fusions. After reverse transcription, C-terminal FLAG-
tagged fusions were pulled down with anti-FLAG M2 agarose beads (Sigma) at 4
o
C for
1 hour and eluted with 3X FLAG peptide (Sigma) twice at room temperature. A pre-clear
step was also introduced to exclude non-specific bead binders immediately before
selection. The FLAG-purified fusions were flowed through 100 µL of D-biotin treated
beads that were used for the selection packed in small column. After two rounds of
selection, the beads were switched to agarose-neutravidin beads (Pierce) to avoid bead
background binding (Figure 4.2A). A total of seven rounds of selection for mDC-SIGN
were performed.
To evolve dual-specific ligands for mDC-SIGN and hDC-SIGN, the pool 3 (M3
pool) of mDC-SIGN selection was used as the starting pool for the hDC-SIGN selection
(Figure 4.2B). This selection was the same as mDC-SIGN selection except that the target
was hDC-SIGN. A total of four rounds of selection were performed against hDC-SIGN,
126
for a total of seven (three rounds versus mouse DC-SIGN and four rounds versus human
DC-SIGN).

In vitro radiolabeled binding assay
The affinity of each pool was monitored by pull down of radiolabeled fusions
against mDC-SIGN or hDC-SIGN immobilized on beads.
L
-[
35
S]methionine (MP
Biomedicals) was used to replace cold methionine in the translation reaction. After
translation, the reaction was treated with ribonuclease A (Roche Applied Science) to
generate pF30P-linked to fibronectins. The radiolabeled fusions were incubated with
mDC-SIGN immobilized beads, hDC-SIGN immobilized beads, or beads without target
at 4
o
C for 1 hour. The beads were washed 4 times with selection buffer and resuspended
with ddH
2
O. The radioactivity present in the flowthrough, four washes, and beads were
detected by scintillation counting. The total input of radiolabeled fusion added to the
binding reaction was the total of flowthrough, wash, and bead counts. The percent
binding of each pool was calculated by the ratio of bead counts/total input counts.
The final pools from the mDC-SIGN selection (M7) or dual selection targeting
mDC-SIGN and hDC-SIGN (M3H4) were PCR amplified. Two restriction sites (Xho I
and Bam HI, NEB Inc.) were introduced using eFnoligo1 and Fnoligo11 during PCR.
The PCR products were cloned into the vector pAO5 by restriction digestion and ligation.
The clones were sequenced and grouped by the similarity of amino-acid sequences,
resulting in 4 groups with multiple clones from the M7 pool, while 2 high-copy number
groups were obtained from M3H4. There were also some singletons in M7 and M3H4
127
pools. Four representative clones from each of the four groups of the M7 pool and 7
clones from the M3H4 pool (2 clones from major groups and 5 singletons) were screened
for clones with the highest affinity using the in vitro radiolabeled binding assay. For in
vitro radiolabeled binding, selected clones were first PCR amplified using eFnoligo1 and
Fnoligo9 and then subjected to in vitro transcription to make mRNA. mRNA from each
clone was translated in rabbit reticulocyte lysate supplemented with
L
-[
35
S]methionine to
generate radiolabeled proteins. The radiolabeled proteins were purified by pull down with
anti-FLAG M2 beads and elution with 3X FLAG peptides. The binding of purified
radiolabeled clones to both mouse and human DC-SIGN, and the analysis of binding data
were performed as described above.

Cloning, expression, and purification of wt e10FnIII and selected M3H4 clones with
C-terminal HA tag or HLA-A*0201 influenza peptide
For the cell functional assay, a HA tag (YPYDVPDYA) was cloned on to the C-
terminus of wt e10FnIII and selected clones from the M7 and M3H4 pools (M7-4, M7-7,
M7-18, M7-19, M3H4-1, M3H4-7, M3H4-12, M3H4-13, and M3H4-18). The HA tag
was introduced via PCR using eFnoligo 1 and 3FnHA (5’-GGA GCC GGA TCC TGC
ATA ATC TGG CAC ATC ATA TGG ATA GCT ACC GGT GCG GTA GTT GAT
GGA GAT CG-3’). eFnoligo 1 introduced a 5’ Xho I site while 3FnHA inserted a HA tag
and Bam HI site onto the 3’ end. PCR products were double digested by Xho I and Bam
HI following the manufacturer’s instructions and the digested products were purified by
agarose gel. Vector pAO5 was also digested using Xho I and Bam HI and purified by an
128
agarose gel. Purified digested clones were ligated into vector pAO5 using T4 DNA ligase
at room temperature for 1 hour. Ligated products were transformed into XL-10
competent cells by heat shock and grown on amplicillin plates at 37
o
C overnight.
Colonies were picked, cultured, and mini-prepped to purify plasmids for sequencing.
Correct clones with HA tags were transformed into BL21/DE3 cells for protein
expression. Each transformed clone was cultured in 5 mL of Luria

Broth (LB)
supplemented with final concentration of 100 %g/ml ampicillin at 37 °C in a shaker
overnight. One mL of overnight culture for each clone was added to 100 mL LB with 100
µg/mL ampicillin and grown at 37 °C with shaking until the O.D.
600
reached ~0.4.
Isopropyl- -D-thiogalactoside (IPTG) at a final concentration of

1 mM was added to each
culture to induce protein expression and all cultures were incubated at 25
o
C with shaking
overnight. Cells were harvested

by centrifugation at 4000x g for 20 min at 4 °C and
frozen at -80
o
C.
The frozen cell

pellets were thawed, resuspended, and lysed with B-PER reagent
(Thermo Scientific) supplemented with cOmplete Protease Inhibitor Cocktail Tablets
(Roche Applied Science). The cell lysates were pelleted at 14,000 rpm for 20 minutes
twice to clear the supernatant. Proteins in supernatant were purified by a column packed
with HisPur Ni-NTA (Thermo Scientific). First, the Ni-NTA column was equilibrated
with 5-column volumes of binding buffer (20 mM Tris-HCl pH 8.0, 500 mM NaCl, 10
mM imidazole) and the supernatant was loaded onto the packed column. The column was
then washed with >10-column volumes of wash buffer (20 mM Tris-HCl pH 8.0, 500
mM NaCl, 20 mM imidazole) and proteins were eluted using elution buffer with an
129
imidazole gradient (20mM Tris-HCl pH 8.0 and 100 mM NaCl containing imidazole at
50, 100, 150, 200, 250, or 500 mM). Elutions were run on an SDS-PAGE gel and the
elutions containing the correct size of proteins were pooled together. The pooled elution
for each clone was buffer exchanged with HLS buffer and concentrated using Amicon
Ultra-15 centrifugal filter units (Millipore) by centrifugation at 4000 x g at 4
o
C. After
concentrated, glycerol was added to a final concentration of 10%. The concentration of
protein was determined by O.D.
280
and the extinction coefficient for each clone. Each
purified protein was divided to small aliquots, flash-frozen in liquid nitrogen, and stored
at -80
o
C. Purified proteins with HA tags of selected clones were tested in the cell
functional assay for binding and internalization.
The same cloning strategy was applied for insertion of HLA-A*0201 influenza
peptide (GILGFVFTL) onto the C-terminus of wt e10FnIII and M3H4-18, which were
used to test immunization. The peptide was added using the eFnoligo 1 5’-primer and the
3FnMP 3’-primer, (5’-GGA GCC GGA TCC GCG GCG GCG CAG GGT AAA CAC
AAA GCC CAG AAT GCC GCG GCG GCT ACC GGT GCG GTA GTT GAT GGA
GAT CG-3’). 3FnMP was designed to code for the HLA-A*0201 influenza peptide
followed by arginines and a His tag [(M3H4-18)-RR-GILGFVFTL-RRR-HHHHH]. The
HLA-A*0201 influenza peptide is derived from the influenza A matrix protein,
representing the HLA-A*0201 (human MHC class I molecule) restricted epitope (Nijman
et al., 1993). The arginines were added to increase the proteasomal digestion during
processing of peptides that will be presented via MHC class I (Dakappagari et al., 2006;
130
Sundaram et al., 2003). These constructs were expressed and purified using the same
protocols as HA-tagged protein (see above).

Cell functional assay – binding, internalization, and immunization
293T, 293T.mDCSIGN (a stable 293T cell line overexpressing mDC-SIGN), and
293T.hDCSIGN (a stable 293T cell line overexpressing hDC-SIGN) cells were used to
verify the binding and internalization of our selected fibronectins (Yang et al., 2008).
Four HA-tagged M7 proteins (M7-4-HA, M7-7-HA M7-18-HA M7-19-HA), five HA-
tagged M3H4 proteins (M3H4-1-HA, M3H4-7-HA, M3H4-12-HA, M3H4-13-HA, and
M3H4-18-HA), and HA-tagged wt e10FnIII (wt e10FnIII-HA) were tested. For the
binding test, cells were treated with 10 nM of HA-tagged fibronectins at 4
o
C for 30
minutes. The treated cells were then washed with PBS and incubated with rabbit anti-HA
antibodies (Abcam) at 4
o
C for 10 min. After incubation, the resulting cells were
immunostained with allophycocyanin (APC) anti-rabbit IgG (Invitrogen) and analyzed
via flow cytometry (BD; data are analyzed using flowjo). The binding of M3H4-18-HA
to 293T.hDCSIGN cells was further confirmed via confocal microscopy (Olympus) using
the same protocol as immunostaining except 100 nM M3H4-18 was used.
293T.hDCSIGN cells were immunostained with anti-human DC-SIGN antibody (9E9A8;
BioLegend, Inc.) and DAPI. The approximate K
D
of M3H4-18-HA was determined by
treating 293T.hDCSIGN cells with 0.4 nM to 200 nM of M3H4-18-HA followed by
analysis using flow cytometry at 4
o
C as described previously (Benedict et al., 1997). The
internalization induced by M3H4-18-HA or HA-tagged mDC-SIGN specific fibronectins
131
was tested by first incubating the 293T.hDCSIGN cells with 100 nM of HA-tagged
fibronectins at 4
o
C for 30 minutes in duplicate. After the unbound fibronectins were
washed off, one sample was cultured at 37
o
C for another 2 hours to induce
internalization while the duplicate was kept at 4
o
C as a non-internalizing control. After 2
hours, anti-HA immunostaining of cell surfaces was performed and the difference of
fluorescence intensity between 4
o
C and 37
o
C was calculated to obtain the internalization
efficiency (Dakappagari et al., 2006; Takahara et al., 2004).
The binding and internalization of M3H4-18-HA were also confirmed in human
peripheral blood mononuclear cell (hPBMC)–derived immature dendritic cells (iDCs).
hPBMC-derived iDCs were generated and purified as previously described (Romani et
al., 1994; Sallusto and Lanzavecchia, 1994), except that hPBMCs were from HLA-A2
+
patients (AllCells, LLC.). iDCs were treated with M3H4-18-HA, stained with anti-HA
antibody, and analyzed by either flow cytometry or microscopy same as 293T cells,
except that the incubation time and M3H4-18 concentration were different.
To understand the mechanism of hDC-SIGN dependent internalization, we
stained clathrin, caveolin-1 and early endosomal antigen 1 (EEA-1). iDCs were incubated
with 100 nM M3H4-18-HA for 30 minutes at 4
o
C, washed and then incubated at 37
o
C
for 30 minutes for visualization of clathrin or caveolin-1, and 1 hour for visualization of
EEA-1. Mander's overlap coefficient was calculated using more than 40 cells for each
experiment.
M3H4-18-HLA-A*0201 influenza peptide (M3H4-18-antigen) and wt e10FnIII-
HLA-A*0201 influenza peptide (wt e10FnIII-antigen) were used to immunize hPBMCs.
132
First, hPBMC-derived iDCs were incubated with M3H4-18-HA, wt e10Fn III-antigen, or
M3H4-18-antigen at 37
o
C for 2 hours. Treated iDCs were cocultured with hPBMCs for
60 h and IFN-) was monitored as an indicator of immune response. To do this, we used
the human IFN-! ELISPOT kit as previously described (Hoffmann et al., 2000).

Results
Novel e10FnIII-scaffold proteins with high specificity and affinity to mDC-SIGN
After seven rounds of selection against mDC-SIGN, we tested the pool for
binding against immobilized mDC-SIGN using radiolabeled fibronectin. This assay
showed that the radiolabeled pool binding increases as the selection progressed, up to
40% of the final M7 pool. This pool is highly specific for mDC-SIGN as it showed <
0.7% binding to hDC-SIGN and < 0.15% binding to beads without target (Figure 4.3A).
These data imply that this selection was successful in evolving fibronectins with high
specificity and affinity to mDC-SIGN. The M7 pool was sequenced, analyzed, and the
resulting clones were grouped by sequence similarity. There are four major groups
containing clones with multiple copies while the remaining sequences are singletons. The
first two major groups each contain clones with identical BC and FG loops. These two
groups are the M7-4 (6 copies) and M7-18 (2 copies) groups. A third group is composed
of sequences similar to M7-7, and contains three clones that share the same FG loop. The
last group contains sequences similar to M7-19, which contains two clones with a similar
FG loop. Six other unique clones that appear only once were also observed (Figure 4.3B).
Clones from each of the four major groups, M7-4, M7-18, M7-7, and M7-19, were tested
133

Figure 4.3 Results of
mRNA display targeting
mDC-SIGN
(A) An in vitro radiolabeled
binding assay was
performed to monitor pool
binding. The binding of M3
and M4 to hDC-SIGN are
not available. (B)
Sequencing results for the
mDC-SIGN pool 7 (M7).
The sequences of wild-type
10FnIII and e10FnIII are
also shown here. The
domains are indicated.
Blue-colored amino acids
represent the mutations
and amino acids shown in
red are randomized region.
The last letter of BC loop is
in green and labeled as 1
(1= Val, Leu, and Ile). Four
major groups are boxed
and four selected clones for
further study are shown in
bold. (C) The percentage
binding of four selected
mDC-SIGN clones and
their sequences. The
sequences are shown on
top. All four selected clones
show high binding to mDC-
SIGN, but almost no
binding to beads without
target or with hDC-SIGN.

134
for the binding in an in vitro radiolabeled binding assay. The binding data show that all
four proteins have high binding to mDC-SIGN (~60% to 80%) with no binding to hDC-
SIGN (< 0.8%) or beads without target (< 0.4%). These results suggest that the mDC-
SIGN selection successfully evolved novel fibronectins with high specificity and affinity
to mDC-SIGN (Figure 4.3C).

Dual-specific e10FnIII-scaffold proteins with high affinity to both mDC-SIGN and
hDC-SIGN
We wished to develop dual-specific protein ligands with high affinity to mDC-
SIGN and hDC-SIGN. Starting with the M3 pool, which has some affinity to mDC-SIGN
(10% binding), we performed further selection against hDC-SIGN (Figure 4.2). Four
additional rounds against hDC-SIGN were performed. The pool shows increased affinity
to both mDC-SIGN (~40% binding using the M3H4 pool) and hDC-SIGN (~20%
binding using the M3H4 pool). These data indicate that the final M3H4 pool contains
fibronectins that bind to both mDC-SIGN and hDC-SIGN (Figure 4.4A). The final M3H4
pool was sequenced and showed two dominant clones with multiple copies and six clones
that appear only once. The two dominant clones, M3H4-13 and M3H4-1, contain the
same BC and FG loops, except that one amino acid mutates from E21 (M3H4-13) to K21
(M3H4-1) in BC loop (Figure 4.4B).
The two dominant clones (M3H4-13 and M3H4-1) and five singletons (M3H4-7,
M3H4-12, M3H4-16, M3H4-18, and M3H4-19) were tested for binding using an in vitro
radiolabled binding assay (Figure 4.4C). One singleton, M3H4-9, was not included,
135

Figure 4.4 Results of
mRNA display against
hDC-SIGN
(A) An in vitro radiolabeled
binding assay was
performed to monitor pool
binding. M3H0 is M3 pool
from the mDC-SIGN
selection and the binding of
this pool to hDC-SIGN is
not available. (B) Mouse-
human pool 7 (M3H4)
sequences are shown here.
The color codes are the
same as Figure 4.3. The
bold sequences were
further characterized.
M3H4-9 is a member of the
most dominant group in
mouse pool 7 (M7-4). (C)
The percentage binding of
selected mouse-human
DC-SIGN clones and their
sequences. The sequences
of selected clones are
shown on top. Five of
seven clones from mDC-
SIGN and hDC-SIGN
selection show dual-
specificity for both mDC-
SIGN and hDC-SIGN, and
very low binding to beads
without target.
136
because it is a member of M7-4 group, which had previously been shown to not bind
hDC-SIGN. The clone binding shows that five of seven clones (M3H4-1, M3H4-7,
M3H4-12, M3H4-13, and M3H4-18) bound to both mDC-SIGN and hDC-SIGN. These
clones possessed higher affinity to mDC-SIGN and very little background binding to
beads without target (< 1.4%). Interestingly, the mutation of amino acid E21 (M3H4-13)
to K21 (M3H4-1) contributed to the shift of the binding from mDC-SIGN (52% of
M3H4-13 to 36% of M3H4-1) to hDC-SIGN (22% of M3H4-13 to 26% of M3H4-1). The
last two clones tested for binding were not DC-SIGN specific ligands as M3H4-16 has
high bead background binding (~35%) and does not bind to DC-SIGN, while M3H4-19
has almost no target binding. Here, we were able to develop a new strategy to evolve
e10FnIII-scaffold proteins ligands with dual specificity and affinity for both mDC-SIGN
and hDC-SIGN.

mDC-SIGN-specific ligands bind to mDC-SIGN overexpressed on 293T cells
(293T.mDC-SIGN cells) and internalize.
M7-4, M7-18, M7-7, and M7-19 bind to mDC-SIGN as shown via an in vitro
radiolabeled binding assay. We wished to test if these fibronectins bind mDC-SIGN that
was expressed on cells. We constructed HA-tagged versions and tested these clones for
their binding to mDC-SIGN overexpressed on 293T cells (293T.mDC-SIGN cells), and
detected binding using immunostaining or flow cytometry (Figure 4.5A). First, these four
clones were expressed and purified with a C-terminal HA tag for immunostaining.
293T.mDC-SIGN cells were treated with the same concentration of each fibronectin-HA,
137

Figure 4.5 Binding and internalization of mouse clones in mDC-SIGN
expressing 293T cells
(A) The binding of selected mDC-SIGN ligands to mDC-SIGN expressed on
293T cells was analyzed by flow cytometry. The same concentration of ligands
was used to incubate with cells. Therefore, the higher fluorescence indicates the
better binding. Ligands are color-labeled. Gray solid is a negative control where
cells were treated with wt e10FnIII protein. (B) The internalization efficiency of
ligands was estimated by the loss of fluorescence on cell surface at 37
o
C in
comparison to 4
o
C. The larger loss of cell-surface fluoresence suggests more
internalization of the ligands.

138
individually. M7-18-HA, M7-7-HA, and M7-19-HA shifted the fluorescence showing
that they bound to 293T.mDC-SIGN cells. M7-4-HA is the only clone that showed no
binding to 293T.mDC-SIGN cells. M7-18-HA treated cells showed the most shift in
fluorescent signal, indicating that M7-18-HA binds to mDC-SIGN with highest affinity.
M7-7-HA and M7-19-HA shift the fluorescence signals not as much as M7-18, but M7-7-
HA shifts more than M7-19-HA. This suggests that M7-19-HA has the lowest affinity.
We then tested the ability of M7-7-HA, M7-18-HA, and M7-19-HA to bind
mDC-SIGN and facilitate internalization into 293T.mDC-SIGN cells using flow
cytometry (Figure 4.5B). In this experiment, we can use flow cytometry to detect the
fluorescent signal of fibronectins remaining on cell surface, while a loss of fluorescence
indicates internalization. Internalization was inhibited by incubating the cells at 4
o
C.
Thus by comparing the fluorescent signal of cells incubated at 4
o
C versus 37
o
C, we can
estimate the percentage of fibronectin internalized (Dakappagari et al., 2006). The
fluorescent signal of cells incubated at 4
o
C again shows that M7-18-HA gives the highest
signal followed by M7-7-HA, and M7-19-HA. However, when comparing the fluorescent
signal loss for cells incubated at 37
o
C, we observed that M7-7-HA results in the largest
loss, followed by M7-18-HA then M7-19-HA. These data suggest that all three
fibronectins internalize into cells. Since M7-7-HA has the largest change in signal
between 37
o
C and 4
o
C, our data suggest that M7-7-HA is the most efficient in inducing
internalization. Further experiments (such as imaging by confocal microscopy) to
confirm that the fibronectins internalize via binding mDC-SIGN are on the way. The
method we used here did not account for the weaker affinity of the fibronectin ligands
139
due to increasing temperature. Therefore, we may overestimate the percentage of
internalization since the loss of fluorescence signal could result from the more
dissociation of the fibronectin ligands at 37
o
C than 4
o
C.

A dual-specific fibronectin binds hDC-SIGN and facilitates internalization
From our in vitro radiolabeled binding assay, we identified five dual-specific
fibronectins that could bind human and mouse DC-SIGN. We added a C-terminal HA tag
to each of these fibronectins to test if they bound to either mDC-SIGN (293T.mDCSIGN)
or hDC-SIGN (293T.hDCSIGN) overexpressed on 293T cells. After binding, we
immunostained and analyzed the cells by flow cytometry. Our data demonstrated that of
the five candidate fibronectins, only M3H4-18-HA bound to ~99% of 293T.hDCSIGN
cells and also bound weakly to 293T.mDCSIGN cells (4.6%). The other four fibronectins
did not bind to 293T.mDCSIGN or 293T.hDCSIGN cells (Figure 4.6A). These data are
somewhat surprising since all five fibronectins bound well in the in vitro radiolabeled
binding assay, but four of the five fibronectins did not bind and M3H4-18-HA bound
poorly to mDC-SIGN overexpressed on 293T cells. The different results between the in
vitro radiolabeled binding assay and the cell binding assay suggest that the in vitro
radiolabeled binding assay may not be a good indicator of the function of selected
fibronectins. Therefore, our data suggest that to better screen for functional fibronectins,
it is necessary to screen clones with both the in vitro radiolabeled binding and cell-based
assays.

140
Figure 4.6 Binding of dual-specific ligands to hDC-SIGN expressing 293T
cells
(A) Screening of selected dual-specific ligands for their binding to DC-SIGN
expressed on 293T cells. The binding of selected dual-specific ligands to mDC-
SIGN or hDC-SIGN expressed on 293T cells was analyzed by flow cytometry.
Almost 99% of hDC-SIGN expressing 293T cells was bound by M3H4-18-HA.
However, the binding of M3H4-HA to mDC-SIGN expressing 293T cells is only
4.6%. Other ligands show no binding. The grey area indicates the untreated cells
while the blue line represents the population of cells treated with dual-specific
ligands. (B) Confocal images of hDC-SIGN expressing on 293T cells treated with
M3H4-18-HA show that both M3H4-18-HA and DC-SIGN localize to the cell
membrane. However, they are partially overlapped, suggesting that M3H4-18
might compete with anti-hDC-SIGN antibodies. As a control, hDC-SIGN
expressed on 293T cells were treated with wt e10FnIII-HA. These control cells do
not show anti-HA staining, which indicates no binding. Immunostaining of M3H4-
18-HA by an anti-HA antibody is shown in red and hDC-SIGN immunostained
with anti-hDC-SIGN antibody is shown in green. The nucleus is blue due to DAPI
staining and the merged images of anti-hDC-SIGN and anti-HA are also shown.
(C) The K
D
of M3H4-18 was measured by treating hDC-SIGN expressed on 293T
cells with different concentrations of M3H4-18. These cells were analyzed by flow
cytometry. The left plot shows that the binding of M3H4-18-HA increased when
the concentration of M3H4-18 increased. However, there is no change in two
141
controls. One control is that hDC-SIGN expressed on 293T cells was treated with
wt e10FnIII-HA. Another control is that 293T cells were treated with M3H4-18-
HA. The right plot was made to calculate K
D
of M3H4-18 to hDC-SIGN on cell
surface. The calculated K
D
is 1.89 +/- 0.56 x 10
-8
M.

Figure 4.6 Binding of dual-specific ligands to hDC-SIGN expressing 293T
cells
142

Figure 4.6 Binding of dual-specific ligands to hDC-SIGN expressing 293T
cells
143
We first treated the DC-SIGN expressing cells with M3H4-18-HA followed by
immunostaining with anti-HA and anti-hDC-SIGN antibodies. Our data demonstrate that
M3H4-18-HA and hDC-SIGN localize on the cell surface but only partially overlap with
each other. As a control, we treated cells with wt e10FnIII-HA that lacks binding to
293T.hDCSIGN cells. Thus, the cell-imaging results indicate that M3H4-18-HA not only
binds to 293T.hDCSIGN cells via hDC-SIGN, but also may compete with the anti-hDC-
SIGN antibody for hDC-SIGN binding (Figure 4.6B). Using this assay, we were also able
to determine the approximate binding constant for M3H4-18-HA to 293T.hDCSIGN cells
which is 18.9±0.56 nM (Figure 4.6C). Thus, using a combination of flow cytometry and
confocal microscopy we have shown that M3H4-18-HA binds with high specificity and
affinity to hDC-SIGN overexpressed on 293T cells.
Using the internalization assay described above that detects the loss of
fluorescence resulting from fibronectin internalization, we determined the approximate
internalization efficiency of M3H4-18-HA. 293T.hDCSIGN cells were treated with
M3H4-18-HA at 37
o
C and compared to cells incubated at 4
o
C (Figure 4.7A). We
observed a 50% loss of fluorescence signal of cells incubated at 37
o
C versus cells
incubated at 4
o
C. This suggests that the approximate internalization efficiency is 50%.
However, as discussed earlier, it is possible that the loss of fluorescence signal could be
due to decreased fibronectin affinity at higher temperature, thus the internalization
efficiency could be less than 50%.
We also confirmed the internalization of M3H4-18-HA by confocal microscopy
and tested the mechanism by which M3H4-18 induced internalization (Figure 4.7B).
144

Figure 4.7 Internalization of M3H4-18-HA into hDC-SIGN expressing 293T
cells
(A) The internalization efficiency of M3H4-18-HA was estimated by measuring
the loss of fluorescence at 37
o
C as compared to 4
o
C with flow cytometry.
Almost 50% of fluorescence is lost at 37
o
C as compared to 4
o
C. This indicates
that at most 50% of M3H4-18 internalizes into hDC-SIGN expressing 293T cells.
(B) The internalization of M3H4-18-HA was imaged by confocal microscopy.
The upper two rows show that M3H4-18-HA internalizes and overlaps with
clathrin, but no overlap is observed with caveolin-1. This implies that
internalization induced by M3H4-18-HA binding to hDC-SIGN is via a clathrin-
dependent pathway. The bottom row shows that some M3H4-18-HA is localized
to early endosomes after internalization, since some overlap of M3H4-18 and
early endosome antigen 1 (EEA-1) is observed. Immunostaining of M3H4-18-
HA by the anti-HA antibody is shown in red and others are in green. The
nucleus is stained blue by DAPI. (C) Mander's overlap coefficient was plotted by
analysis of more than 40 cells for each sample. The overlap of M3H4-18-HA
and clathrin is statistically significant while the overlap of M3H4-18-HA and
caveolin-1 is not statistically significant.

145
Internalization via hDC-SIGN binding has been reported to function through a classical
clathrin-dependent mechanism (Cambi et al., 2009). Thus, we stained for clathrin, which
showed overlap with M3H4-18. As a control, we stained for caveolin-1, which is another
molecule for endocytosis. Figure 4.7C shows that staining for clathrin overlaps with
M3H4-18 more significantly than caveolin-1. This suggests that M3H4-18 internalizes
via a clathrin-dependent mechanism. We also calculated Mander's overlap coefficient to
demonstrate that the overlap of clathrin staining and M3H4-18 staining is significantly
higher than the overlap of caveolin-1. However, our results are not able to exclude other
mechanisms for endocytosis induced by M3H4-18 via its binding to hDC-SIGN, because
there are other endocytosis molecules that we haven’t tested.
After receptor-mediated endocytosis, the vesicle will fuse with early endosome.
EEA1 is the marker for early endosome, so the EEA1 staining shows that M3H4-18 is in
early endosome after endocytosed. This suggests that the endocytosis of M3H4-18 is a
receptor-mediated endocytosis and the receptor should be DC-SIGN based on our
binding data. The images of all channels show that M3H4-18-HA dose internalize into
the cells, especially obvious in the EEA1 column. The reason is that there was 1-hour
incubation in the EEA1 staining instead of 30-minute incubation in other two
experiments. This indicates that the M3H4-18 induced internalization takes time to
activate.

146
M3H4-18 immunizes hPBMCs by binding to hDC-SIGN and internalizing iDCs.
We then tested if M3H4-18 could bind and internalize on hPBMC-derived iDCs.
hPBMCs are derived from human blood and contain all the blood cells that have a round
nucleus and are responsible for immunity. These cells include dendritic cells and
lymphocytes in which we would like to target immature dendritic cells. hPBMCs have
also been used to generate immature dendritic cells (iDCs) for dendritic cell studies. We
first used flow cytometry to analyze hPBMC-derived iDCs treated with M3H4-18-HA or
wt e10FnIII-HA (Figure 4.8A). First, immunostained hPBMC-derived iDCs show that
the majority of cells express hDC-SIGN, confirming that were able to both generate and
purify hPBMC-derived iDCs. Secondly, hPBMC-derived iDCs treated with wt e10FnIII-
HA give a similar signal as an untreated isotype control when analyzed via flow
cytometry. This indicates that wt e10FnIII-HA does not bind to hDC-SIGN. Thirdly,
hPBMC-derived iDCs treated with M3H4-18-HA show high M3H4-18-HA and DC-
SIGN fluorescence, suggesting that M3H4-18-HA binds to hDC-SIGN.
We further confirmed that M3H4-18-HA binds to hPBMC-derived iDCs via the
confocal microscopy. We treated hPBMC-derived iDCs with either wt e10FnIII-HA or
M3H4-18-HA, and immunostained with anti-HA and anti-hDC-SIGN antibodies (Figure
4.8B). Our data show that HA staining and hDC-SIGN staining are overlapped on the cell
surface, which provides evidence that M3H4-18-HA binds to hDC-SIGN on the cell
surface of hPBMC-derived iDCs. No anti-HA fluorescence is observed in wt 10FnIII
treated hPBMC-derived iDCs, which excludes any signal from a non-specific interaction
between the e10FnIII scaffold and hDC-SIGN. The internalization of M3H4-18-HA into
147
Figure 4.8 Binding and internalization of M3H4-18-HA in human peripheral
blood mononuclear cell (hPBMC)-derived immature dendritic cells (iDCs).
(A) Flow cytometry analysis of M3H4-18-HA binding to hPBMC-derived iDCs
after stained with anti-HA and anti-hDC-SIGN antibodies (9E9A8). The cells
treated with different ligands are indicated. Two dot plots in the left show no
ligand-treated control cells. Cells treated with wt e10FnIII-HA show no anti-HA
staining, indicating no binding of wt e10FnIII-HA. M3H4-18-HA proteins bind to
hDC-SIGN on cells, which can be detected by anti-HA staining. The w/o staining
shows the background autofluorescence of untreated cells that were not stained
with any antibody. Isotype control is untreated cells that were stained with anti-
HA and anti-hDC-SIGN antibodies. Isotype control shows that there is no non-
specific binding between anti-HA antibody and untreated cells. (B) M3H4-18-HA
binds to hDC-SIGN expressed on hPBMC-derived iDCs. Images on the left are
iDCs treated with wt e10FnIII-HA while images on the right are iDCs incubated
with M3H4-18-HA. Cells were immunostained with anti-HA and anti-hDC-SIGN
antibodies. The red channel represents anti-HA, the green channel indicates anti-
hDC-SIGN, the blue channel is DAPI, which stains the nucleus, and the merged
images are also shown. There is clearly no wt e10FnIII-HA binding to hDC-SIGN
expressing iDCs, suggesting no background binding from the e10FnIII scaffold.
Colocalization of the anti-HA and anti-hDC-SIGN signals on the cell surface
indicates that M3H4-18-HA binds to hDC-SIGN expressed on iDCs. (C) M3H4-
18-HA internalizes into hDC-SIGN expressing iDCs. Red shows M3H4-18-HA,
148
green represents other staining, and blue is DAPI staining to show the nucleus.
M3H4-18-HA proteins are inside the iDCs shown by all three images, suggesting
that M3H4-18-HA is able to internalize into hDC-SIGN expressing iDCs. Clathrin
staining and EEA1 staining do partially overlap with M3H4-18-HA, indicating that
the internalization likely proceeds via a clathrin-dependent pathway. There is no
overlap between M3H4-18-HA and caveolin-1, thus it is unlikely that
internalization proceeds via a caveloin-1-dependent pathway.

Figure 4.8 Binding and internalization of M3H4-18-HA in human peripheral
blood mononuclear cells (hPBMCs)-derived immature dendritic cells
(iDCs).
149

Figure 4.8 Binding and internalization of M3H4-18-HA in human peripheral
blood mononuclear cells (hPBMCs)-derived immature dendritic cells
(iDCs).
150
hPBMC-derived iDCs and possible mechanism were also confirmed by intracellular
immunostaining and confocal microscopy (Figure 4.8C). The images show that the
M3H4-18-HA localizes to the cell surface and cytoplasm, which suggests the
internalization of M3H4-18-HA. The partial overlap of hDC-SIGN and M3H4-18
staining supports that M3H4-18-HA internalizes via the binding to hDC-SIGN on the
surface of hPBMC-derived iDCs. The hPBMC-derived iDCs were also immunostained
with anti-clathrin, anti-caveolin-1, and anti-EEA1 antibodies, which showed that clathrin
and EEA1 staining partially overlap with anti-HA staining, except caveolin-1. These data
confirm that the most likley mechanism of internalization is due to clathrin-dependent
endocytosis.
Lastly, we tested if our fibronectins could immunize hPBMC-derived iDCs. We
conjugated HLA-A*0201 influenza peptide to either wt fibronectin (wt 10FnIII-antigen)
or M3H4-18 (M3H4-18-antigen) (Figure 4.9A). In this assay, increasing IFN-)
production indicates that more CD8
+
T cells were activated. Since the peptide used here
is MHC I class specific antigenic peptide, we expected that the loading of this peptide on
MHC class I molecule via M3H4-18 binding to hDC-SIGN would activate CD8
+
T cells.
Thus, the increase of IFN-) production indicates that immune response was boosted.
When we treated hPBMC-derived iDCs with M3H4-18-antigen we observed a significant
increase of IFN-) when compared to cells treated with either wt 10FnIII-antigen or
M3H4-18-HA, which lacks the antigen. IFN-) is produced by natural kill (NK) cells
during the innate immune response, or by CD4
+
and CD8
+
T cells for antigen-specific
immunity (Schoenborn and Wilson, 2007). Our results suggest that treatment of cells
151

Figure 4.9 Antigen-based immunity induced by M3H4-18-antigen
(A) wt e10FnIII and M3H4-18 were cloned with a C-terminal GILGFVFTL
peptide (HLA-A2-restricted epitope). The GILGFVFTL epitope is shown in
brown. Arginine shown in blue indicates additional residues to enhance
intracellular proteasomal proteolysis. The BC and FG loops are shown in red.
(B) Immune response was monitored by measuring IFN-) production. IFN-) was
secreted by T cells when the immune response was induced by infection. Thus,
the higher IFN-) production indicates the stronger immune response. The
amount of IFN-) per million cells was plotted. The bar graph shows that the
immune response was induced only by M3H4-18-antigen.

152
with M3H4-18-antigen induces the increase of IFN-) production. Our proposed
mechanism is that M3H4-18-antigen binds to hDC-SIGN on the surface of the cell, the
fibronectins are internalized into hPBMC-derived iDCs, and the antigens are processed to
be presented on iDCs, which results in activation of CD8
+
T cells. We observed low IFN-
) production with wt e10Fn III-antigen treated cells, suggesting that binding to hDC-
SIGN is necessary for development of an immune response. Cells treated with M3H4-18-
HA, which lacks the antigen, also produce low IFN-), implying that M3H4-18-HA does
not induce an unwanted immune response through the presentation of peptides derived
from the digested M3HA-18-HA. Taken together, our data show that M3H4-18 can both
specifically target and deliver antigen to DCs expressing hDC-SIGN and induce an
antigen-specific immune response.

Discussion
Our strategy is successfully to develop antibody-mimetic ligands targeting DC-SIGN
for immunization
For mRNA display against mDC-SIGN, seven rounds of selection were able to
converge pool binding to mDC-SIGN from 10.4% (M3 pool) to 40% (M7 pool) with no
hDC-SIGN binding (0.7%) and binding to beads without targets (0.15%). Four selected
clones with high-copy numbers from the M7 pool were shown to have strong binding
(>60%) to mDC-SIGN and no binding to hDC-SIGN (<0.78%) or beads without target
(<0.35%) (Figure 4.3).
153
To overcome the fact that mice are not a suitable animal model for DC-SIGN-
based vaccine development (Tacken et al., 2007), we developed a new strategy to evolve
dual-specific ligands targeting both mDC-SIGN and hDC-SIGN. Dual-specific ligands
thus can be evaluated preclinically in a mouse model and subsequently transferred to
humans for clinical trials and therapeutic treatments. Our strategy is to exploit mRNA
display and a fibronectin library to evolve a pool specific for mouse DC-SIGN (the M3
pool) to also bind hDC-SIGN. Since both human and mouse DC-SIGN have a similar
carbohydrate recognition domain (CRD), it is likely that this pool contains clones with
high affinity to mDC-SIGN and hDC-SIGN (Figure 4.1). Using the M3 pool, we
performed four rounds of mRNA display against hDC-SIGN. The final pool bound 38%
to mDC-SIGN and 19% to hDC-SIGN (Figure 4.4A). The fact that binding increased to
mDC-SIGN even if the pool was solely selected for binding to hDC-SIGN indicates that
there is a similar binding site on both mDC-SIGN and hDC-SIGN. It is likely that the
CRD is the best target for developing multiple-specific ligands to a variety of mouse DC-
SIGNs and human DC-SIGN, because of the high conservation of the CRD among all
DC-SIGNs (Powlesland et al., 2006).
Five of seven clones tested in an in vitro radiolabeled binding assay show dual
specificity for mDC-SIGN (8.5%-52%) and hDC-SIGN (6.8%-26%) and do not bind
beads without target (<1.4%). Our results show that we are able to evolve dual-specific
antibody-mimetic ligands that bind proteins from similar species. All five clones bind
more tightly to mDC-SIGN versus hDC-SIGN, suggesting that theM3 pool excludes
exceptional ligands for hDC-SIGN. In the future, a possible solution to evolve ligands
154
with similar affinity to both mDC-SIGN and hDC-SIGN would be to use mDC-SIGN as
a target in one round and switch to hDC-SIGN in the following round.
The five positive clones form the dual-selection have different binding patterns
for mDC-SIGN and hDC-SIGN. Interestingly, the sequence difference between M3H4-
13 and M3H4-1 is only one amino acid, which is a Glu (M3H4-13) to Lys (M3H4-1)
mutation. However, this one mutation affects the fibronectin affinities to mDC-SIGN and
hDC-SIGN. This Glu to Lys mutation might contribute to the decrease in affinity of
M3H4-1 binding to mDC-SIGN, but increase its binding to hDC-SIGN. This results in
M3H4-1 having similar affinity for both DC-SIGNs. This also suggests that this position
could be important for hDC-SIGN specific binding and the binding could be driven by
electrostatic interactions. In summary, we are able to utilize mRNA display to evolve
antibody mimetic ligands binding to mDC-SIGN or with dual affinity to both mDC-SIGN
and hDC-SIGN.

The difference between in vitro radiolabeled binding assay and cell binding assay
All the positive clones tested in in vitro radiolabeled binding assay were further
tested for their function in DC-SIGN overexpressed on 293T cells. Three of four mDC-
SIGN fibronectins that bound in the radiolabeled assay retain binding to mDC-SIGN
overexpressed on 293T cells. However, the affinities observed in the cell-based
experiments are different from those observed in the in vitro radiolabeled binding assay.
We also observed this effect when testing the dual-specific ligands. Only one of
five positive dual-specific ligands, M3H4-18-HA, exerts dual affinity when tested in the
155
cell binding assay. However, M3H4-18-HA has excellent affinity to 293T.hDCSIGN
cells (99%), but low affinity to 293T.mDCSIGN cells (4.6%). The other four clones have
no binding to either 293T.mDCSIGN cells or 293T.hDCSIGN cells. This difference
between in vitro radiolabeled binding assay and cell binding assay may result from DC-
SIGN, ligands, or both.
One reason why this difference is observed is that DC-SIGN expressed on 293T
cells might be different from DC-SIGN expressed in E.coli. The extracellular domain of
DC-SIGN expressed in E.coli has been found in inclusion body. E. coli expressed DC-
SIGN must first be unfolded, followed by refolding to form the correct structure and
exert normal function. It is possible that the refolded DC-SIGN used as a target in mRNA
display might not be identical to that expressed on 293T cells. In addition, there might be
different modifications between DC-SIGN expressed in E.coli and in 293T cells. A
possible solution is to use 293T.mDCSIGN cells and 293T.hDCSIGN cells as targets for
mRNA display. However, this strategy is difficult since targeting whole cells might
increase non-specific interactions. Therefore, a better strategy is to target E.coli expressed
DC-SIGN to evolve a pool with good affinity to DC-SIGN where the non-specific
ligands have been excluded. This pool is subsequently used in a selection against DC-
SIGN expressed on 293T cells to evolve ligands binding to DC-SIGN on cell surface.
Another reason why there is a difference between the in vitro radiolabeled binding
and cell-based assays is the different expression systems used to produce the fibornectins.
In mRNA display, the proteins are translated in reticulocyte lysate, while the fibronectins
are expressed in E.coli for cell-based experiments. The different expression systems
156
could cause a difference in folding, post-translational modification, or expression. A
previous work had shown that selected 10FnIII-scaffold proteins were well expressed and
folded in reticulocyte lysate, but were poorly expressed and folded in E.coli (Olson and
Roberts, 2007). As a result of this observation, the e10FnIII-scaffold library was
constructed in order to maintain high expression and better folding of selected
fibronectins expressed in E.coli. The e10FnIII scaffold does seem to partially resolve this
issue, because two of four mDC-SIGN-specific ligands and four of five dual-specific
ligands were expressed in E.coli. in good amount. However, after purification, there were
less or equal amount of impurities when compared to products (data not shown). It is
possible that the impurities or fibronectin misfolding when expressed in E.coli might be
the cause for the different results between in vitro binding assay and cell binding assay
rather than expression. Another solution for folding or impurities is to engineer e10FnIII-
MBP fusions, but the addition of MBP (a moderately large protein) may interfere with
DC-SIGN binding or be processed and presented as antigens on DCs. The ultimate
solution for this could be to introduce selection steps for evolving better expression and
proper folding in E.coli.
Other than the issues discussed above, we partially achieved our goal to develop
dual-specific ligands with high affinity to mDC-SIGN and hDC-SIGN. M3H4-18 is a
strong binder for hDC-SIGN overexpressed on 293T cells and a weak binder for mDC-
SIGN overexpressed on 293T cells. In addition, M7-7, M7-18, and M7-19 are also good
ligands for mDC-SIGN overexpressed on 293T cells.

157
Three mDC-SIGN-specific protein ligands can be developed as markers and M3H4-
18 can be exploited as an immunization tool.
M7-7 and M7-18 can be developed as mDC-SIGN markers due to their high
affinity and specificity to mDC-SIGN. Neither the expression nor function of mouse DC-
SIGN has been studied in detail because of a lack of good antibodies (Cheong et al.,
2010). Although a new anti-mouse DC-SIGN antibody has recently been developed, the
manufacture of antibodies is still laborious and expensive (Cheong et al., 2010). Our
antibody-mimetic mDC-SIGN fibronectins provide a good alternative to antibodies.
mDC-SIGN specificity of M7-7 and M7-18 can be confirmed by binding to other
possible human DC-SIGN homologs in mouse, such as SIGNR1 and SIGNR3.
mDC-SIGN has been reported to internalize bound antibody, however mDC-
SIGN was unable to internalize bound mannose-BSA (Powlesland et al., 2006; Takahara
et al., 2004). Since these two findings directly contradict each other, further study of
mDC-SIGN is necessary. M7-7 and M7-18 have been shown here to bind mDC-SIGN
and internalize, which supports the idea that our ligands can be used to study the function
of mDC-SIGN. However, our method is the same as Takahara et al, which can be biased
by assuming that the loss of signal on cell surface only results from internalization. Our
future experiments involve confirming the internalization by immunostaining and
confocal microscopy. It would be more interesting to examine the antigen presenting
function of mDC-SIGN using our ligands, which can determine whether mDC-SIGN is a
functional orthologue of hDC-SIGN or not. Finding a functional orthologue of hDC-
158
SIGN will be a great advance in making mice a suitable model for the development of an
hDC-SIGN-based vaccine.
Finally, we tested the ability of M3H4-18 to facilitate immunization via hDC-
SIGN binding. Treatment of hPBMC-derived iDCs with M3H4-18 fused to the HLA-
A*0201 influenza peptide (M3H4-18-antigen) increases IFN-) levels significantly
compared to the wt e10FnIII-antigen or M3H4-18-HA controls. First, these results
indicate that M3H4-18 and antigen are both necessary for inducing immune response.
M3H4-18 binds to hDC-SIGN, acting as an antigen carrier to deliver the antigen to DCs
and trigger the antigen presenting process. Secondly, administration of M3H4-18 not
conjugated to an antigen did not induce an immune response, implying that M3H4-18 is
not processed for antigen presentation; antigen is required. This makes M3H4-18 a great
immunization tool since by itself, M3H4-18 will not be presented on DCs and induce an
unwanted M3H4-18-specific immune response. In addition, the antigen can easily be
attached to the C-terminus of M3H4-18, which induces an immune response. It is much
easier to construct and produce these fibronectin-antigen conjugates than antibody-
antigen conjugates. This simple design also avoids tedious optimization as to where the
antigen can be inserted into the antibody to get a maximal antigen-specific immune
response.
We also explored the mechanism of M3H4-18-induced immunization. M3H4-18
was localized to early endosomes after internalization via clathrin-dependent endocytosis.
This suggests that the M3H4-18 in early endosomes will be processed by the MHC class
II pathway and present the carried antigen for CD4
+
cell activation. However, this cannot
159
exclude the possibility that the M3H4-18 carried antigen could undergo cross-
presentation via the MHC class I pathway to activate CD8
+
T cells. The cross-
presentation was shown by using HLA-A*0201 influenza peptide to induce immune
response, which was demonstrated by the increase of IFN-) production. The HLA-
A*0201 influenza peptide is a substrate for MHC class I, thus the peptide could escape
the endosome into the cytosol and be cross-presented on MHC class I molecule,
activating CD8
+
cells that produce IFN-) (Nijman et al., 1993). Future experiments will
monitor the production of CD4
+
and CD8
+
cells to further address the mechanism of
immunization.
We have utilized mRNA display of an e10FnIII-scaffold library to develop both
mDC-SIGN-specific and dual human/mouse-specific DC-SIGN fibronectins. Since DC-
SIGN is a membrane protein, it is likely that we can use this strategy to generate
fibronectins capable of binding to other membrane receptors with high specificity and
affinity. We also propose several changes to improve our protocol to evolve ligands that
function in vitro and in cells. Two fibronectins that bind mDC-SIGN are good candidates
for further study and characterization in mDC-SIGN mediated immunization. These
fibronectins also can be exploited to further study the biology of mDC-SIGN, which may
lead to the discovery of a functional ortholog of hDC-SIGN. The discovery of the mouse
ortholog will enable the use of mice as animal models for the development of a human
DC-SIGN-based vaccine.
We have also generated a fibronectin, M3H4-18, that binds to both human and
mouse DC-SIGN. An antigen attached to this fibronectin can be delivered to DCs via
160
binding to DC-SIGN, resulting in an antigen-specific immune response. Thus, M3H4-18
is a candidate for facilitating the development of in vivo cancer vaccines. M3H4-18 has
several advantages over DC-SIGN specific antibodies including the ability to produce an
antigen-fibronectin fusion easily and at lower costs than antibodies. Additionally, M3H4-
18 is derived from the 10th domain of human fibronectin, which argues that little
unwanted immune response is likely to be observed. Lastly, M3H4-18 binds to both
human and mouse DC-SIGN, which will facilitate the use of mice to perform preclinical
experiments. Overall, our study not only develops a new strategy for evolving functional
ligands for membrane receptors, but also generates new ligands for the development of
new in vivo DC-based cancer vaccines.

161
Bibliography
Almen, M.S., Nordstrom, K.J., Fredriksson, R., and Schioth, H.B. (2009). Mapping the
human membrane proteome: a majority of the human membrane proteins can be
classified according to function and evolutionary origin. BMC Biol 7, 50.
Arinaminpathy, Y., Khurana, E., Engelman, D.M., and Gerstein, M.B. (2009).
Computational analysis of membrane proteins: the largest class of drug targets. Drug
Discov Today 14, 1130-1135.
Attwood, T.K., and Findlay, J.B. (1994). Fingerprinting G-protein-coupled receptors.
Protein Eng 7, 195-203.
Austin, R.J., Ja, W.W., and Roberts, R.W. (2008). Evolution of class-specific peptides
targeting a hot spot of the Galphas subunit. J Mol Biol 377, 1406-1418.
Banchereau, J., Briere, F., Caux, C., Davoust, J., Lebecque, S., Liu, Y.J., Pulendran, B.,
and Palucka, K. (2000). Immunobiology of dendritic cells. Annu Rev Immunol 18, 767-
811.
Banchereau, J., and Palucka, A.K. (2005). Dendritic cells as therapeutic vaccines against
cancer. Nat Rev Immunol 5, 296-306.
Banchereau, J., and Steinman, R.M. (1998). Dendritic cells and the control of immunity.
Nature 392, 245-252.
Barrick, J.E., Takahashi, T.T., Ren, J., Xia, T., and Roberts, R.W. (2001). Large libraries
reveal diverse solutions to an RNA recognition problem. Proc Natl Acad Sci U S A 98,
12374-12378.
Batori, V., Koide, A., and Koide, S. (2002). Exploring the potential of the monobody
scaffold: effects of loop elongation on the stability of a fibronectin type III domain.
Protein Eng 15, 1015-1020.
Benedict, C.A., MacKrell, A.J., and Anderson, W.F. (1997). Determination of the
binding affinity of an anti-CD34 single-chain antibody using a novel, flow cytometry
based assay. J Immunol Methods 201, 223-231.
Binz, H.K., Amstutz, P., and Pluckthun, A. (2005). Engineering novel binding proteins
from nonimmunoglobulin domains. Nat Biotechnol 23, 1257-1268.
Bloom, L., and Calabro, V. (2009). FN3: a new protein scaffold reaches the clinic. Drug
Discov Today 14, 949-955.
162
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.
Boder, E.T., and Wittrup, K.D. (1997). Yeast surface display for screening combinatorial
polypeptide libraries. Nat Biotechnol 15, 553-557.
Bonifaz, L., Bonnyay, D., Mahnke, K., Rivera, M., Nussenzweig, M.C., and Steinman,
R.M. (2002). Efficient targeting of protein antigen to the dendritic cell receptor DEC-205
in the steady state leads to antigen presentation on major histocompatibility complex
class I products and peripheral CD8+ T cell tolerance. J Exp Med 196, 1627-1638.
Buchsel, P.C., and DeMeyer, E.S. (2006). Dendritic cells: emerging roles in tumor
immunotherapy. Clin J Oncol Nurs 10, 629-640.
Cambi, A., Beeren, I., Joosten, B., Fransen, J.A., and Figdor, C.G. (2009). The C-type
lectin DC-SIGN internalizes soluble antigens and HIV-1 virions via a clathrin-dependent
mechanism. Eur J Immunol 39, 1923-1928.
Caravella, J., and Lugovskoy, A. (2010). Design of next-generation protein therapeutics.
Curr Opin Chem Biol 14, 520-528.
Chen, S.Y., Bagley, J., and Marasco, W.A. (1994). Intracellular antibodies as a new class
of therapeutic molecules for gene therapy. Hum Gene Ther 5, 595-601.
Cheong, C., Matos, I., Choi, J.H., Schauer, J.D., Dandamudi, D.B., Shrestha, E.,
Makeyeva, J.A., Li, X., Li, P., Steinman, R.M., et al. (2010). New monoclonal anti-
mouse DC-SIGN antibodies reactive with acetone-fixed cells. J Immunol Methods 360,
66-75.
Cherezov, V., Rosenbaum, D.M., Hanson, M.A., Rasmussen, S.G., Thian, F.S., Kobilka,
T.S., Choi, H.J., Kuhn, P., Weis, W.I., Kobilka, B.K., et al. (2007). High-resolution
crystal structure of an engineered human beta2-adrenergic G protein-coupled receptor.
Science 318, 1258-1265.
Cho, G., Keefe, A.D., Liu, R., Wilson, D.S., and Szostak, J.W. (2000). Constructing high
complexity synthetic libraries of long ORFs using in vitro selection. J Mol Biol 297, 309-
319.
Clark, G.J., Angel, N., Kato, M., Lopez, J.A., MacDonald, K., Vuckovic, S., and Hart,
D.N. (2000). The role of dendritic cells in the innate immune system. Microbes Infect 2,
257-272.
163
Conklin, B.R., Farfel, Z., Lustig, K.D., Julius, D., and Bourne, H.R. (1993). Substitution
of three amino acids switches receptor specificity of Gq alpha to that of Gi alpha. Nature
363, 274-276.
Cota, E., Steward, A., Fowler, S.B., and Clarke, J. (2001). The folding nucleus of a
fibronectin type III domain is composed of core residues of the immunoglobulin-like fold.
J Mol Biol 305, 1185-1194.
Covic, L., Gresser, A.L., Talavera, J., Swift, S., and Kuliopulos, A. (2002a). Activation
and inhibition of G protein-coupled receptors by cell-penetrating membrane-tethered
peptides. Proc Natl Acad Sci U S A 99, 643-648.
Covic, L., Misra, M., Badar, J., Singh, C., and Kuliopulos, A. (2002b). Pepducin-based
intervention of thrombin-receptor signaling and systemic platelet activation. Nat Med 8,
1161-1165.
Crooks, G.E., Hon, G., Chandonia, J.M., and Brenner, S.E. (2004). WebLogo: a sequence
logo generator. Genome Res 14, 1188-1190.
Dakappagari, N., Maruyama, T., Renshaw, M., Tacken, P., Figdor, C., Torensma, R.,
Wild, M.A., Wu, D., Bowdish, K., and Kretz-Rommel, A. (2006). Internalizing
antibodies to the C-type lectins, L-SIGN and DC-SIGN, inhibit viral glycoprotein
binding and deliver antigen to human dendritic cells for the induction of T cell responses.
J Immunol 176, 426-440.
Day, P.W., Rasmussen, S.G., Parnot, C., Fung, J.J., Masood, A., Kobilka, T.S., Yao, X.J.,
Choi, H.J., Weis, W.I., Rohrer, D.K., et al. (2007). A monoclonal antibody for G protein-
coupled receptor crystallography. Nat Methods 4, 927-929.
Diaz, J.E., Howard, B.E., Neubauer, M.S., Olszewski, A., and Weiss, G.A. (2003).
Exploring biochemistry and cellular biology with protein libraries. Curr Issues Mol Biol
5, 129-145.
Dickinson, C.D., Veerapandian, B., Dai, X.P., Hamlin, R.C., Xuong, N.H., Ruoslahti, E.,
and Ely, K.R. (1994). Crystal structure of the tenth type III cell adhesion module of
human fibronectin. J Mol Biol 236, 1079-1092.
Dratz, E.A., Furstenau, J.E., Lambert, C.G., Thireault, D.L., Rarick, H., Schepers, T.,
Pakhlevaniants, S., and Hamm, H.E. (1993). NMR structure of a receptor-bound G-
protein peptide. Nature 363, 276-281.
Dursi, A.M., Albrizio, S., Greco, G., Mazzeo, S., Mazzoni, M.R., Novellino, E., and
Rovero, P. (2002). Conformational analysis of the Galpha(s) protein C-terminal region. J
Pept Sci 8, 476-488.
164
Emens, L. (2008). Cancer vaccines: on the threshold of success. Expert opinion on
emerging drugs 13, 295-603.
Fajardo-Moser, M., Berzel, S., and Moll, H. (2008). Mechanisms of dendritic cell-based
vaccination against infection. Int J Med Microbiol 298, 11-20.
Feinberg, H., Guo, Y., Mitchell, D.A., Drickamer, K., and Weis, W.I. (2005). Extended
neck regions stabilize tetramers of the receptors DC-SIGN and DC-SIGNR. J Biol Chem
280, 1327-1335.
Figdor, C.G., de Vries, I.J., Lesterhuis, W.J., and Melief, C.J. (2004). Dendritic cell
immunotherapy: mapping the way. Nat Med 10, 475-480.
Figdor, C.G., van Kooyk, Y., and Adema, G.J. (2002). C-type lectin receptors on
dendritic cells and Langerhans cells. Nat Rev Immunol 2, 77-84.
Filmore, D. (2004). It's a GPCR world. Modern Drug Discovery 7, 24-28.
Frankel, A., Millward, S.W., and Roberts, R.W. (2003). Encodamers: unnatural peptide
oligomers encoded in RNA. Chem Biol 10, 1043-1050.
Frankel, A., and Roberts, R.W. (2003). In vitro selection for sense codon suppression.
RNA 9, 780-786.
Fuchs, P., Breitling, F., Dubel, S., Seehaus, T., and Little, M. (1991). Targeting
recombinant antibodies to the surface of Escherichia coli: fusion to a peptidoglycan
associated lipoprotein. Biotechnology (N Y) 9, 1369-1372.
Gebauer, M., and Skerra, A. (2009). Engineered protein scaffolds as next-generation
antibody therapeutics. Curr Opin Chem Biol 13, 245-255.
Geijtenbeek, T.B., Torensma, R., van Vliet, S.J., van Duijnhoven, G.C., Adema, G.J., van
Kooyk, Y., and Figdor, C.G. (2000). Identification of DC-SIGN, a novel dendritic cell-
specific ICAM-3 receptor that supports primary immune responses. Cell 100, 575-585.
Getmanova, E.V., Chen, Y., Bloom, L., Gokemeijer, J., Shamah, S., Warikoo, V., Wang,
J., Ling, V., and Sun, L. (2006). Antagonists to human and mouse vascular endothelial
growth factor receptor 2 generated by directed protein evolution in vitro. Chem Biol 13,
549-556.
Ghanouni, P., Steenhuis, J.J., Farrens, D.L., and Kobilka, B.K. (2001). Agonist-induced
conformational changes in the G-protein-coupling domain of the beta 2 adrenergic
receptor. Proc Natl Acad Sci U S A 98, 5997-6002.
Gilboa, E. (2007). DC-based cancer vaccines. J Clin Invest 117, 1195-1203.
165
Gill, D.S., and Damle, N.K. (2006). Biopharmaceutical drug discovery using novel
protein scaffolds. Curr Opin Biotechnol 17, 653-658.
Goujon, M., McWilliam, H., Li, W., Valentin, F., Squizzato, S., Paern, J., and Lopez, R.
(2010). A new bioinformatics analysis tools framework at EMBL-EBI. Nucleic Acids
Res 38, W695-699.
Griffiths, A.D., and Tawfik, D.S. (2006). Miniaturising the laboratory in emulsion
droplets. Trends Biotechnol 24, 395-402.
Grigorieff, N., Ceska, T.A., Downing, K.H., Baldwin, J.M., and Henderson, R. (1996).
Electron-crystallographic refinement of the structure of bacteriorhodopsin. J Mol Biol
259, 393-421.
Grisshammer, R. (2009). Purification of recombinant G-protein-coupled receptors.
Methods Enzymol 463, 631-645.
Guan, L., Smirnova, I.N., Verner, G., Nagamori, S., and Kaback, H.R. (2006).
Manipulating phospholipids for crystallization of a membrane transport protein. Proc
Natl Acad Sci U S A 103, 1723-1726.
Guo, Y., Feinberg, H., Conroy, E., Mitchell, D.A., Alvarez, R., Blixt, O., Taylor, M.E.,
Weis, W.I., and Drickamer, K. (2004). Structural basis for distinct ligand-binding and
targeting properties of the receptors DC-SIGN and DC-SIGNR. Nat Struct Mol Biol 11,
591-598.
Hamm, H.E., Deretic, D., Arendt, A., Hargrave, P.A., Koenig, B., and Hofmann, K.P.
(1988). Site of G protein binding to rhodopsin mapped with synthetic peptides from the
alpha subunit. Science 241, 832-835.
Hammerstrom, A.E., Cauley, D.H., Atkinson, B.J., and Sharma, P. (2011). Cancer
immunotherapy: sipuleucel-T and beyond. Pharmacotherapy 31, 813-828.
Hanes, J., and Pluckthun, A. (1997). In vitro selection and evolution of functional
proteins by using ribosome display. Proc Natl Acad Sci U S A 94, 4937-4942.
Harper, D.M., Franco, E.L., Wheeler, C.M., Moscicki, A.B., Romanowski, B., Roteli-
Martins, C.M., Jenkins, D., Schuind, A., Costa Clemens, S.A., and Dubin, G. (2006).
Sustained efficacy up to 4.5 years of a bivalent L1 virus-like particle vaccine against
human papillomavirus types 16 and 18: follow-up from a randomised control trial. Lancet
367, 1247-1255.
Henderson, R., and Unwin, P.N. (1975). Three-dimensional model of purple membrane
obtained by electron microscopy. Nature 257, 28-32.
166
Hoffmann, T.K., Meidenbauer, N., Dworacki, G., Kanaya, H., and Whiteside, T.L.
(2000). Generation of tumor-specific T-lymphocytes by cross-priming with human
dendritic cells ingesting apoptotic tumor cells. Cancer Res 60, 3542-3549.
Hosse, R.J., Rothe, A., and Power, B.E. (2006). A new generation of protein display
scaffolds for molecular recognition. Protein Sci 15, 14-27.
Hunte, C., and Michel, H. (2002). Crystallisation of membrane proteins mediated by
antibody fragments. Curr Opin Struct Biol 12, 503-508.
Ja, W.W., Olsen, B.N., and Roberts, R.W. (2005). Epitope mapping using mRNA display
and a unidirectional nested deletion library. Protein Eng Des Sel 18, 309-319.
Ja, W.W., and Roberts, R.W. (2004). In vitro selection of state-specific peptide
modulators of G protein signaling using mRNA display. Biochemistry 43, 9265-9275.
Ja, W.W., West, A.P., Jr., Delker, S.L., Bjorkman, P.J., Benzer, S., and Roberts, R.W.
(2007). Extension of Drosophila melanogaster life span with a GPCR peptide inhibitor.
Nat Chem Biol 3, 415-419.
Ja, W.W., Wiser, O., Austin, R.J., Jan, L.Y., and Roberts, R.W. (2006). Turning G
proteins on and off using peptide ligands. ACS Chem Biol 1, 570-574.
Kasahara, N., Dozy, A.M., and Kan, Y.W. (1994). Tissue-specific targeting of retroviral
vectors through ligand-receptor interactions. Science 266, 1373-1376.
Kato, M., McDonald, K.J., Khan, S., Ross, I.L., Vuckovic, S., Chen, K., Munster, D.,
MacDonald, K.P., and Hart, D.N. (2006). Expression of human DEC-205 (CD205)
multilectin receptor on leukocytes. Int Immunol 18, 857-869.
Keefe, A.D. (2001). Protein selection using mRNA display. Curr Protoc Mol Biol
Chapter 24, Unit 24 25.
Keefe, A.D., and Szostak, J.W. (2001). Functional proteins from a random-sequence
library. Nature 410, 715-718.
Kobilka, B.K. (1995). Amino and carboxyl terminal modifications to facilitate the
production and purification of a G protein-coupled receptor. Anal Biochem 231, 269-271.
Kobilka, B.K., and Deupi, X. (2007). Conformational complexity of G-protein-coupled
receptors. Trends Pharmacol Sci 28, 397-406.
Koide, A., Bailey, C.W., Huang, X., and Koide, S. (1998). The fibronectin type III
domain as a scaffold for novel binding proteins. J Mol Biol 284, 1141-1151.
167
Kolb, P., and Klebe, G. (2011). The golden age of GPCR structural biology: any impact
on drug design? Angew Chem Int Ed Engl 50, 11573-11575.
Lamboy, J.A., Arter, J.A., Knopp, K.A., Der, D., Overstreet, C.M., Palermo, E.F.,
Urakami, H., Yu, T.B., Tezgel, O., Tew, G.N., et al. (2009). Phage wrapping with
cationic polymers eliminates nonspecific binding between M13 phage and high pI target
proteins. J Am Chem Soc 131, 16454-16460.
Larkin, M.A., Blackshields, G., Brown, N.P., Chenna, R., McGettigan, P.A., McWilliam,
H., Valentin, F., Wallace, I.M., Wilm, A., Lopez, R., et al. (2007). Clustal W and Clustal
X version 2.0. Bioinformatics 23, 2947-2948.
Lesterhuis, W., Haanen, J., and Punt, C. (2011). Cancer immunotherapy--revisited.
Nature reviews Drug discovery 10, 591-1191.
Li, S., Millward, S., and Roberts, R. (2002). In vitro selection of mRNA display libraries
containing an unnatural amino acid. J Am Chem Soc 124, 9972-9973.
Li, S., and Roberts, R.W. (2003). A novel strategy for in vitro selection of peptide-drug
conjugates. Chem Biol 10, 233-239.
Liao, H.I., Olson, C.A., Hwang, S., Deng, H., Wong, E., Baric, R.S., Roberts, R.W., and
Sun, R. (2009). mRNA display design of fibronectin-based intrabodies that detect and
inhibit severe acute respiratory syndrome coronavirus nucleocapsid protein. J Biol Chem
284, 17512-17520.
Liggett, S.B. (2000). Pharmacogenetics of beta-1- and beta-2-adrenergic receptors.
Pharmacology 61, 167-173.
Linehan, S.A. (2005). The mannose receptor is expressed by subsets of APC in non-
lymphoid organs. BMC Immunol 6, 4.
Ling, M.M. (2003). Large antibody display libraries for isolation of high-affinity
antibodies. Comb Chem High Throughput Screen 6, 421-432.
Lipovsek, D., and Pluckthun, A. (2004). In-vitro protein evolution by ribosome display
and mRNA display. J Immunol Methods 290, 51-67.
Litvinovich, S.V., and Ingham, K.C. (1995). Interactions between type III domains in the
110 kDa cell-binding fragment of fibronectin. J Mol Biol 248, 611-626.
Lofblom, J., Frejd, F.Y., and Stahl, S. (2011). Non-immunoglobulin based protein
scaffolds. Curr Opin Biotechnol 22, 843-848.
168
Main, A.L., Harvey, T.S., Baron, M., Boyd, J., and Campbell, I.D. (1992). The three-
dimensional structure of the tenth type III module of fibronectin: an insight into RGD-
mediated interactions. Cell 71, 671-678.
Martin, E.L., Rens-Domiano, S., Schatz, P.J., and Hamm, H.E. (1996). Potent peptide
analogues of a G protein receptor-binding region obtained with a combinatorial library. J
Biol Chem 271, 361-366.
Mattheakis, L.C., Bhatt, R.R., and Dower, W.J. (1994). An in vitro polysome display
system for identifying ligands from very large peptide libraries. Proc Natl Acad Sci U S
A 91, 9022-9026.
Mellman, I., and Steinman, R.M. (2001). Dendritic cells: specialized and regulated
antigen processing machines. Cell 106, 255-258.
Millward, S.W., Fiacco, S., Austin, R.J., and Roberts, R.W. (2007). Design of cyclic
peptides that bind protein surfaces with antibody-like affinity. ACS Chem Biol 2, 625-
634.
Millward, S.W., Takahashi, T.T., and Roberts, R.W. (2005). A general route for post-
translational cyclization of mRNA display libraries. J Am Chem Soc 127, 14142-14143.
Molek, P., Strukelj, B., and Bratkovic, T. (2011). Peptide phage display as a tool for drug
discovery: targeting membrane receptors. Molecules 16, 857-887.
Nagaoka, K., Takahara, K., Minamino, K., Takeda, T., Yoshida, Y., and Inaba, K. (2010).
Expression of C-type lectin, SIGNR3, on subsets of dendritic cells, macrophages, and
monocytes. J Leukoc Biol 88, 913-924.
Nathans, J., and Hogness, D.S. (1983). Isolation, sequence analysis, and intron-exon
arrangement of the gene encoding bovine rhodopsin. Cell 34, 807-814.
Nauts, H.C. (1989). Bacteria and cancer--antagonisms and benefits. Cancer Surv 8, 713-
723.
Nijman, H.W., Houbiers, J.G., Vierboom, M.P., van der Burg, S.H., Drijfhout, J.W.,
D'Amaro, J., Kenemans, P., Melief, C.J., and Kast, W.M. (1993). Identification of
peptide sequences that potentially trigger HLA-A2.1-restricted cytotoxic T lymphocytes.
Eur J Immunol 23, 1215-1219.
Nuttall, S.D., and Walsh, R.B. (2008). Display scaffolds: protein engineering for novel
therapeutics. Curr Opin Pharmacol 8, 609-615.
169
Odegrip, R., Coomber, D., Eldridge, B., Hederer, R., Kuhlman, P.A., Ullman, C.,
FitzGerald, K., and McGregor, D. (2004). CIS display: In vitro selection of peptides from
libraries of protein-DNA complexes. Proc Natl Acad Sci U S A 101, 2806-2810.
Olson, C.A., Adams, J.D., Takahashi, T.T., Qi, H., Howell, S.M., Wu, T.T., Roberts,
R.W., Sun, R., and Soh, H.T. (2011). Rapid mRNA-Display Selection of an IL-6
Inhibitor Using Continuous-Flow Magnetic Separation. Angew Chem Int Ed Engl.
Olson, C.A., Liao, H.I., Sun, R., and Roberts, R.W. (2008). mRNA display selection of a
high-affinity, modification-specific phospho-IkappaBalpha-binding fibronectin. ACS
Chem Biol 3, 480-485.
Olson, C.A., and Roberts, R.W. (2007). Design, expression, and stability of a diverse
protein library based on the human fibronectin type III domain. Protein Sci 16, 476-484.
Overington, J.P., Al-Lazikani, B., and Hopkins, A.L. (2006). How many drug targets are
there? Nat Rev Drug Discov 5, 993-996.
Palczewski, K., Kumasaka, T., Hori, T., Behnke, C.A., Motoshima, H., Fox, B.A., Le
Trong, I., Teller, D.C., Okada, T., Stenkamp, R.E., et al. (2000). Crystal structure of
rhodopsin: A G protein-coupled receptor. Science 289, 739-745.
Park, C.G., Takahara, K., Umemoto, E., Yashima, Y., Matsubara, K., Matsuda, Y.,
Clausen, B.E., Inaba, K., and Steinman, R.M. (2001). Five mouse homologues of the
human dendritic cell C-type lectin, DC-SIGN. Int Immunol 13, 1283-1290.
Parker, M.H., Chen, Y., Danehy, F., Dufu, K., Ekstrom, J., Getmanova, E., Gokemeijer,
J., Xu, L., and Lipovsek, D. (2005). Antibody mimics based on human fibronectin type
three domain engineered for thermostability and high-affinity binding to vascular
endothelial growth factor receptor two. Protein Eng Des Sel 18, 435-444.
Pebay-Peyroula, E., Rummel, G., Rosenbusch, J.P., and Landau, E.M. (1997). X-ray
structure of bacteriorhodopsin at 2.5 angstroms from microcrystals grown in lipidic cubic
phases. Science 277, 1676-1681.
Perelson, A.S., and Oster, G.F. (1979). Theoretical studies of clonal selection: minimal
antibody repertoire size and reliability of self-non-self discrimination. J Theor Biol 81,
645-670.
Pierce, K.L., Premont, R.T., and Lefkowitz, R.J. (2002). Seven-transmembrane receptors.
Nat Rev Mol Cell Biol 3, 639-650.
Plosker, G.L. (2011). Sipuleucel-T: in metastatic castration-resistant prostate cancer.
Drugs 71, 101-108.
170
Plotkin, S.A. (2005). Vaccines: past, present and future. Nat Med 11, S5-11.
Powlesland, A.S., Ward, E.M., Sadhu, S.K., Guo, Y., Taylor, M.E., and Drickamer, K.
(2006). Widely divergent biochemical properties of the complete set of mouse DC-SIGN-
related proteins. J Biol Chem 281, 20440-20449.
Ramakrishna, V., Treml, J.F., Vitale, L., Connolly, J.E., O'Neill, T., Smith, P.A., Jones,
C.L., He, L.Z., Goldstein, J., Wallace, P.K., et al. (2004). Mannose receptor targeting of
tumor antigen pmel17 to human dendritic cells directs anti-melanoma T cell responses
via multiple HLA molecules. J Immunol 172, 2845-2852.
Rasenick, M.M., Watanabe, M., Lazarevic, M.B., Hatta, S., and Hamm, H.E. (1994).
Synthetic peptides as probes for G protein function. Carboxyl-terminal G alpha s peptides
mimic Gs and evoke high affinity agonist binding to beta-adrenergic receptors. J Biol
Chem 269, 21519-21525.
Rasmussen, S.G., Choi, H.J., Fung, J.J., Pardon, E., Casarosa, P., Chae, P.S., Devree,
B.T., Rosenbaum, D.M., Thian, F.S., Kobilka, T.S., et al. (2011a). Structure of a
nanobody-stabilized active state of the beta(2) adrenoceptor. Nature 469, 175-180.
Rasmussen, S.G., Choi, H.J., Rosenbaum, D.M., Kobilka, T.S., Thian, F.S., Edwards,
P.C., Burghammer, M., Ratnala, V.R., Sanishvili, R., Fischetti, R.F., et al. (2007). Crystal
structure of the human beta2 adrenergic G-protein-coupled receptor. Nature 450, 383-387.
Rasmussen, S.G., DeVree, B.T., Zou, Y., Kruse, A.C., Chung, K.Y., Kobilka, T.S., Thian,
F.S., Chae, P.S., Pardon, E., Calinski, D., et al. (2011b). Crystal structure of the beta2
adrenergic receptor-Gs protein complex. Nature 477, 549-555.
Reiersen, H., Lobersli, I., Loset, G.A., Hvattum, E., Simonsen, B., Stacy, J.E., McGregor,
D., Fitzgerald, K., Welschof, M., Brekke, O.H., et al. (2005). Covalent antibody display--
an in vitro antibody-DNA library selection system. Nucleic Acids Res 33, e10.
Roberts, R.W., and Szostak, J.W. (1997). RNA-peptide fusions for the in vitro selection
of peptides and proteins. Proc Natl Acad Sci U S A 94, 12297-12302.
Rodbell, M. (1980). The role of hormone receptors and GTP-regulatory proteins in
membrane transduction. Nature 284, 17-22.
Romani, N., Gruner, S., Brang, D., Kampgen, E., Lenz, A., Trockenbacher, B.,
Konwalinka, G., Fritsch, P.O., Steinman, R.M., and Schuler, G. (1994). Proliferating
dendritic cell progenitors in human blood. J Exp Med 180, 83-93.
Rosenbaum, D.M., Cherezov, V., Hanson, M.A., Rasmussen, S.G., Thian, F.S., Kobilka,
T.S., Choi, H.J., Yao, X.J., Weis, W.I., Stevens, R.C., et al. (2007). GPCR engineering
171
yields high-resolution structural insights into beta2-adrenergic receptor function. Science
318, 1266-1273.
Rosenbaum, D.M., Rasmussen, S.G., and Kobilka, B.K. (2009). The structure and
function of G-protein-coupled receptors. Nature 459, 356-363.
Rosenberg, S.A., Yang, J.C., and Restifo, N.P. (2004). Cancer immunotherapy: moving
beyond current vaccines. Nat Med 10, 909-915.
Sallusto, F., and Lanzavecchia, A. (1994). Efficient presentation of soluble antigen by
cultured human dendritic cells is maintained by granulocyte/macrophage colony-
stimulating factor plus interleukin 4 and downregulated by tumor necrosis factor alpha. J
Exp Med 179, 1109-1118.
Santoro, M.M., and Bolen, D.W. (1988). Unfolding free energy changes determined by
the linear extrapolation method. 1. Unfolding of phenylmethanesulfonyl alpha-
chymotrypsin using different denaturants. Biochemistry 27, 8063-8068.
Scheerer, P., Park, J.H., Hildebrand, P.W., Kim, Y.J., Krauss, N., Choe, H.W., Hofmann,
K.P., and Ernst, O.P. (2008). Crystal structure of opsin in its G-protein-interacting
conformation. Nature 455, 497-502.
Schlyer, S., and Horuk, R. (2006). I want a new drug: G-protein-coupled receptors in
drug development. Drug Discov Today 11, 481-493.
Schneider, T.D., and Stephens, R.M. (1990). Sequence logos: a new way to display
consensus sequences. Nucleic Acids Res 18, 6097-6100.
Schoenborn, J.R., and Wilson, C.B. (2007). Regulation of interferon-gamma during
innate and adaptive immune responses. Adv Immunol 96, 41-101.
Schuler, G., Schuler-Thurner, B., and Steinman, R.M. (2003). The use of dendritic cells
in cancer immunotherapy. Curr Opin Immunol 15, 138-147.
Seddon, A.M., Curnow, P., and Booth, P.J. (2004). Membrane proteins, lipids and
detergents: not just a soap opera. Biochim Biophys Acta 1666, 105-117.
Singh, S.K., Stephani, J., Schaefer, M., Kalay, H., Garcia-Vallejo, J.J., den Haan, J.,
Saeland, E., Sparwasser, T., and van Kooyk, Y. (2009). Targeting glycan modified OVA
to murine DC-SIGN transgenic dendritic cells enhances MHC class I and II presentation.
Mol Immunol 47, 164-174.
Smith, G.P. (1985). Filamentous fusion phage: novel expression vectors that display
cloned antigens on the virion surface. Science 228, 1315-1317.
172
Sundaram, R., Sun, Y., Walker, C.M., Lemonnier, F.A., Jacobson, S., and Kaumaya, P.T.
(2003). A novel multivalent human CTL peptide construct elicits robust cellular immune
responses in HLA-A*0201 transgenic mice: implications for HTLV-1 vaccine design.
Vaccine 21, 2767-2781.
Swaminath, G., Deupi, X., Lee, T.W., Zhu, W., Thian, F.S., Kobilka, T.S., and Kobilka,
B. (2005). Probing the beta2 adrenoceptor binding site with catechol reveals differences
in binding and activation by agonists and partial agonists. J Biol Chem 280, 22165-22171.
Swaminath, G., Xiang, Y., Lee, T.W., Steenhuis, J., Parnot, C., and Kobilka, B.K. (2004).
Sequential binding of agonists to the beta2 adrenoceptor. Kinetic evidence for
intermediate conformational states. J Biol Chem 279, 686-691.
Tacken, P.J., de Vries, I.J., Gijzen, K., Joosten, B., Wu, D., Rother, R.P., Faas, S.J., Punt,
C.J., Torensma, R., Adema, G.J., et al. (2005). Effective induction of naive and recall T-
cell responses by targeting antigen to human dendritic cells via a humanized anti-DC-
SIGN antibody. Blood 106, 1278-1285.
Tacken, P.J., de Vries, I.J., Torensma, R., and Figdor, C.G. (2007). Dendritic-cell
immunotherapy: from ex vivo loading to in vivo targeting. Nat Rev Immunol 7, 790-802.
Takahara, K., Yashima, Y., Omatsu, Y., Yoshida, H., Kimura, Y., Kang, Y.S., Steinman,
R.M., Park, C.G., and Inaba, K. (2004). Functional comparison of the mouse DC-SIGN,
SIGNR1, SIGNR3 and Langerin, C-type lectins. Int Immunol 16, 819-829.
Takahashi, T.T., Austin, R.J., and Roberts, R.W. (2003). mRNA display: ligand
discovery, interaction analysis and beyond. Trends Biochem Sci 28, 159-165.
Takahashi, T.T., and Roberts, R.W. (2009). In vitro selection of protein and peptide
libraries using mRNA display. Methods Mol Biol 535, 293-314.
Tawfik, D.S., and Griffiths, A.D. (1998). Man-made cell-like compartments for
molecular evolution. Nat Biotechnol 16, 652-656.
Tsang, J., and Joyce, G.F. (1996). In vitro evolution of randomized ribozymes. Methods
Enzymol 267, 410-426.
Ullman, C.G., Frigotto, L., and Cooley, R.N. (2011). In vitro methods for peptide display
and their applications. Brief Funct Genomics 10, 125-134.
Waldo, G.S., Standish, B.M., Berendzen, J., and Terwilliger, T.C. (1999). Rapid protein-
folding assay using green fluorescent protein. Nat Biotechnol 17, 691-695.
White, S.H. (2009). Biophysical dissection of membrane proteins. Nature 459, 344-346.
173
Whorton, M.R., Bokoch, M.P., Rasmussen, S.G., Huang, B., Zare, R.N., Kobilka, B., and
Sunahara, R.K. (2007). A monomeric G protein-coupled receptor isolated in a high-
density lipoprotein particle efficiently activates its G protein. Proc Natl Acad Sci U S A
104, 7682-7687.
Wilson, D.S., Keefe, A.D., and Szostak, J.W. (2001). The use of mRNA display to select
high-affinity protein-binding peptides. Proc Natl Acad Sci U S A 98, 3750-3755.
Xia, T., Frankel, A., Takahashi, T.T., Ren, J., and Roberts, R.W. (2003). Context and
conformation dictate function of a transcription antitermination switch. Nat Struct Biol 10,
812-819.
Xu, L., Aha, P., Gu, K., Kuimelis, R.G., Kurz, M., Lam, T., Lim, A.C., Liu, H., Lohse,
P.A., Sun, L., et al. (2002). Directed evolution of high-affinity antibody mimics using
mRNA display. Chem Biol 9, 933-942.
Yang, L., Yang, H., Rideout, K., Cho, T., Joo, K.I., Ziegler, L., Elliot, A., Walls, A., Yu,
D., Baltimore, D., et al. (2008). Engineered lentivector targeting of dendritic cells for in
vivo immunization. Nat Biotechnol 26, 326-334.
Yao, X., Parnot, C., Deupi, X., Ratnala, V.R., Swaminath, G., Farrens, D., and Kobilka,
B. (2006). Coupling ligand structure to specific conformational switches in the beta2-
adrenoceptor. Nat Chem Biol 2, 417-422.

Abstract (if available)

Abstract mRNA display is an in vitro selection technique that can evolve novel ligands to modulating protein-protein interactions and regulate crucial biological functions. Previous efforts have focused on soluble protein targets. However, here we use mRNA display to target cell surface receptors that represent 60% of drug targets but remain the most challenging targets. Two cell surface receptors, beta-2 adrenergic receptor (β2AR) and Dendritic Cell-Specific ICAM-3-Grabbing Non-integrin (DC-SIGN), were chosen as our targets in order to develop novel ligands for structural studies, drug development, or vaccine design. mRNA display using a doped Gαs C-terminal peptide library was capable of targeting the active state of β2AR, resulting in active state-specific peptide ligands with function similar to Gs protein. Selections against both mouse and human DC-SIGN by mRNA display using an antibody-mimetic library resulted in ligands with dual specificity to both mouse and human DC-SIGN. One selected DC-SIGN specific ligand could induce antigen-specific immune responses in human dendritic cells and has the potential for developing DC-based cancer vaccines. In these proof of concept studies, we demonstrated a general approach for the development of novel functional ligands capable of targeting cell surface receptors.

Linked assets

University of Southern California Dissertations and Theses

Conceptually similar

PDF

Dendritic cell-specific vaccine utilizing antibody-mimetic ligand and lentivector system

PDF

Customization of the mRNA display cycle – from protease resistance to high throughput analysis of peptide ligands

PDF

The selective targeting of Ras: a tale of kinetics, mRNA display, and vaccines

PDF

mRNA display as a method to generate stable fibronectin ligands for the alpha 1 subunit of the nicotinic acetylcholine receptor

PDF

Generation and characterization of peptide theranostics by mRNA display

PDF

PD-L1-GM-CSF fusion protein-loaded DC vaccination activates PDL1-specific humoral and cellular immune responses

PDF

Discovery of mature microRNA sequences within the protein- coding regions of global HIV-1 genomes: Predictions of novel mechanisms for viral infection and pathogenicity

PDF

Development of stable cell lines expressing α1β2γ2 and α5β3γ2 GABAA receptors

PDF

The calcium-sensing receptor in the specification of normal and malignant hematopoietic cell localization in the bone marrow microenvironment

PDF

Mysm1, a histone de-ubiquitinase, is essential for dendritic cell development and function

PDF

Engineering immunotoxin and viral vectors for cancer therapy

PDF

Antibody discovery and structural studies of the viral oncogene ORF74

PDF

Engineering viral vectors for T-cell immunotherapy and HIV-1 vaccine

PDF

The histone H2A deubiquitinase MYSM1 regulates CCR9 expression on CD4⁺ T cells and thymocytes

PDF

The modulation of dynamin and receptor endocytosis machinery using elastin-like polypeptides

PDF

Generation and characterization of anti-CD138 chimeric antigen receptor T (CAR-T) cells for the treatment of hematologic malignancies

PDF

The role of neuregulin receptors in cell differentiation and the response to inflammatory cytokines in the intestinal epithelium

PDF

The role of endoplasmic reticulum chaperones in adipogenesis, liver cancer and mammary gland development

Asset Metadata

Creator Hung, Kuo-Chan (author)

Core Title Developing peptide and antibody-mimetic ligands for the cell surface receptors β2AR and DC-SIGN

School Keck School of Medicine

Degree Doctor of Philosophy

Degree Program Genetic, Molecular and Cellular Biology

Publication Date 07/18/2014

Defense Date 05/03/2012

Publisher University of Southern California (original), University of Southern California. Libraries (digital)

Tag cancer vaccine,DC-SIGN,Dendritic cells,fibronectin scaffold,G protein-coupled receptors,in vitro selection technique,mRNA display,OAI-PMH Harvest,β2AR

Language English

Contributor Electronically uploaded by the author (provenance)

Advisor Laird-Offringa, Ite A. (committee chair), Arnold, Donald B. (committee member), Peti-Peterdi, Janos (committee member), Roberts, Richard W. (committee member)

Creator Email kuochanh@gmail.com,kuochanh@usc.edu

Permanent Link (DOI) https://doi.org/10.25549/usctheses-c3-59567

Unique identifier UC11289873

Identifier usctheses-c3-59567 (legacy record id)

Legacy Identifier etd-HungKuoCha-965.pdf

Dmrecord 59567

Document Type Dissertation

Rights Hung, Kuo-Chan

Type texts

Source University of Southern California (contributing entity), University of Southern California Dissertations and Theses (collection)

Access Conditions The author retains rights to his/her dissertation, thesis or other graduate work according to U.S. copyright law. Electronic access is being provided by the USC Libraries in agreement with the a...

Repository Name University of Southern California Digital Library

Repository Location USC Digital Library, University of Southern California, University Park Campus MC 2810, 3434 South Grand Avenue, 2nd Floor, Los Angeles, California 90089-2810, USA