Structurebased Drug Discovery Leslie William

Tari download

https://ebookbell.com/product/structurebased-drug-discovery-
leslie-william-tari-4108162

Explore and download more ebooks at ebookbell.com

Here are some recommended products that we believe you will be
interested in. You can click the link to download.

Structurebased Drug Discovery 1st Edition Harren Jhoti Editor

https://ebookbell.com/product/structurebased-drug-discovery-1st-
edition-harren-jhoti-editor-2164816

Structurebased Drug Discovery Methods In Molecular Biology V841 2012th

Edition Leslie W Tari

https://ebookbell.com/product/structurebased-drug-discovery-methods-
in-molecular-biology-v841-2012th-edition-leslie-w-tari-2494624

Structurebased Drug Discovery Roderick E Hubbard

https://ebookbell.com/product/structurebased-drug-discovery-roderick-
e-hubbard-4336128

Biomolecular Simulations In Structurebased Drug Discovery Francesco L

Gervasio

https://ebookbell.com/product/biomolecular-simulations-in-
structurebased-drug-discovery-francesco-l-gervasio-7302608
Computer Aided Drug Design Cadd From Ligandbased Methods To
Structurebased Approaches Mithun Rudrapal

https://ebookbell.com/product/computer-aided-drug-design-cadd-from-
ligandbased-methods-to-structurebased-approaches-mithun-
rudrapal-48777796

Drug Design Structure And Ligandbased Approaches 1st Edition Kenneth M

Merz Editor

https://ebookbell.com/product/drug-design-structure-and-ligandbased-
approaches-1st-edition-kenneth-m-merz-editor-2541488

Drug Design Structure And Ligandbased Approaches 1st Edition Kenneth M

Merz

https://ebookbell.com/product/drug-design-structure-and-ligandbased-
approaches-1st-edition-kenneth-m-merz-5557152

Structurebased Design Of Drugs And Other Bioactive Molecules Tools And

Strategies 1st Edition Arun K Ghosh

https://ebookbell.com/product/structurebased-design-of-drugs-and-
other-bioactive-molecules-tools-and-strategies-1st-edition-arun-k-
ghosh-4726226

Structurebased Design Of Drugs And Other Bioactive Molecules Tools And

Strategies 2014th Edition Arun K Ghosh

https://ebookbell.com/product/structurebased-design-of-drugs-and-
other-bioactive-molecules-tools-and-strategies-2014th-edition-arun-k-
ghosh-77873396
METHODS IN MOLECULAR BIOLOGY™

Series Editor
John M. Walker
School of Life Sciences
University of Hertfordshire
Hatfield, Hertfordshire, AL10 9AB, UK

For further volumes:

http://www.springer.com/series/7651
Structure-Based Drug Discovery

Edited by

Leslie W. Tari
Trius Therapeutics, San Diego, CA, USA
Editor
Leslie W. Tari
Trius Therapeutics
San Diego, CA, USA
[email protected]

ISSN 1064-3745 e-ISSN 1940-6029

ISBN 978-1-61779-519-0 e-ISBN 978-1-61779-520-6
DOI 10.1007/978-1-61779-520-6
Springer New York Dordrecht Heidelberg London

Library of Congress Control Number: 2011944430

© Springer Science+Business Media, LLC 2012

All rights reserved. This work may not be translated or copied in whole or in part without the written permission of the
publisher (Humana Press, c/o Springer Science+Business Media, LLC, 233 Spring Street, New York, NY 10013, USA),
except for brief excerpts in connection with reviews or scholarly analysis. Use in connection with any form of information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology now known or
hereafter developed is forbidden.
The use in this publication of trade names, trademarks, service marks, and similar terms, even if they are not identified
as such, is not to be taken as an expression of opinion as to whether or not they are subject to proprietary rights.

Printed on acid-free paper

Humana Press is part of Springer Science+Business Media (www.springer.com)

Preface

The potential utility of atomic resolution structures of protein drug targets in drug discovery
has long been acknowledged. Without structure, medicinal chemists must rely on the costly,
time-consuming endeavor of screening large libraries of compounds for hits, and are often
forced to live with high molecular weight, non-ligand-efficient inhibitor scaffolds that must
be blindly decorated with thousands of groups to generate SAR, improve potency and
properties. With knowledge of the shape and chemical composition of the ligand-binding
pocket of the drug target, the de novo design of ligand efficient inhibitor scaffolds is
enabled. Also, iterative-structure-guided ligand optimization can be used to rationally
improve early leads in a few steps rather than with thousands of analogs. However, despite
its promise, structure-based drug design (SBDD) did not live up to expectations in its early
days: only a limited range of protein targets were tractable to crystallographic studies, crystal
structures took months or years to solve, and limitations in computing power and unrealistic
expectations of the capabilities of molecular modeling methods reduced the scope and
effectiveness of SBDD.
The last decade has seen the confluence of several enabling technologies that have
allowed protein crystallographic methods to live up to their true potential. Off-the-shelf
systems exist that allow the rapid cloning, and recombinant expression and isolation of large
quantities of protein in a wide range of prokaryotic or eukaryotic hosts. Low-cost nanovolume
liquid-handling robotic systems are available for the automated screening of vast arrays of
diverse solution conditions to find crystallization conditions for a protein target using mini-
mal quantities of protein. Latest generation synchrotron radiation sources allow for the
collection of high-resolution X-ray diffraction data on microcrystals in minutes. Continuing
improvements in computing power and advances in crystallographic software have made it
possible to go from X-ray dataset to refined crystal structure in less than an hour on a laptop
computer. Taken together, these advances have made it possible to tackle difficult biological
targets with a high probability of success: intact bacterial ribosomes have been structurally
elucidated, as well as eukaryotic trans-membrane proteins like the potassium channel and
GPCRs. Of additional importance is the impact the above mentioned advances have had on
the throughput of crystallographic structure determinations: it is now possible for medicinal
chemists to have access to structural information on their latest small molecule candidates
bound to the therapeutic target within days of compound synthesis, allowing structure-
guided ligand optimization to occur in “real time.” Also, using fragment screening, crystal
structures of hundreds of small molecule cores complexed with the protein target can be
utilized to construct novel inhibitor scaffolds.
The goal of this book is to provide scientists interested in adding SBDD to their arsenal
of drug discovery methods with a practical guide to the methods used to generate crystal
structures of biological macromolecules, how to leverage the structural information to
design new inhibitor classes de novo, and how to iteratively optimize hits and convert them
to leads. Where possible, specific protocols are described. Some examples highlighting the
utility of structural biology in the discovery and development of small molecule and protein
therapeutic agents are provided in the later chapters.

v
vi Preface

I am deeply grateful to all contributors who agreed to share their experiences in the
development and application of methodologies that support SBDD. I believe their patience
and hard work will be rewarded by the impact this volume has on scientists involved in drug
discovery. I would like to extend special thanks to John Walker for his guidance, inspiration
and patience in the preparation of this volume. Also, I am grateful to Les Tari Sr. for his
critical evaluation of this volume and sharp editorial eye.

San Diego, CA, USA Leslie W. Tari

Contents

Preface. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . v
Contributors. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . ix

1 The Utility of Structural Biology in Drug Discovery . . . . . . . . . . . . . . . . . . . . 1

Leslie W. Tari
2 Genetic Construct Design and Recombinant Protein Expression
for Structural Biology . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 29
Suzanne C. Edavettal, Michael J. Hunter, and Ronald V. Swanson
3 Purification of Proteins for Crystallographic Applications. . . . . . . . . . . . . . . . . 49
Daniel C. Bensen
4 Protein Crystallization for Structure-Based Drug Design. . . . . . . . . . . . . . . . . 67
Isaac D. Hoffman
5 X-Ray Sources and High-Throughput Data Collection Methods . . . . . . . . . . . 93
Gyorgy Snell
6 The Use of Molecular Graphics in Structure-Based Drug Design. . . . . . . . . . . 143
Paul Emsley and Judit É. Debreczeni
7 Crystallographic Fragment Screening . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
John Badger
8 The Role of Enzymology in a Structure-Based Drug
Discovery Program: Bacterial DNA Gyrase . . . . . . . . . . . . . . . . . . . . . . . . . . . 179
Mark L. Cunningham
9 Leveraging Structural Information for the Discovery
of New Drugs: Computational Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 209
Toan B. Nguyen, Sergio E. Wong, and Felice C. Lightstone
10 Chemical Informatics: Using Molecular Shape Descriptors
in Structure-Based Drug Design . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 235
Andy Jennings
11 Accounting for Solvent in Structure-Based Drug Design . . . . . . . . . . . . . . . . . 251
Leslie W. Tari
12 Structure-Based Drug Design on Membrane Protein Targets: Human
Integral Membrane Protein 5-Lipoxygenase-Activating Protein . . . . . . . . . . . . 267
Andrew D. Ferguson
13 Application of SBDD to the Discovery of New Antibacterial Drugs . . . . . . . . . 291
John Finn

vii
viii Contents

14 Leveraging SBDD in Protein Therapeutic Development:

Antibody Engineering. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 321
Gary L. Gilliland, Jinquan Luo, Omid Vafa,
and Juan Carlos Almagro
15 A Medicinal Chemistry Perspective on Structure-Based
Drug Design and Development. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 351
Shawn P. Maddaford

Index . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 383
Contributors

JUAN CARLOS ALMAGRO • Centocor R&D Inc., Radnor, PA, USA

JOHN BADGER • Zenobia Therapeutics, San Diego, CA, USA
DANIEL C. BENSEN • Trius Therapeutics, San Diego, CA, USA
MARK L. CUNNINGHAM • Trius Therapeutics, San Diego, CA, USA
JUDIT É. DEBRECZENI • Structure and Biophysics, Discovery Sciences, AstraZeneca,
Alderley Park, Macclesfield, UK
SUZANNE C. EDAVETTAL • Centocor R&D Inc., San Diego, CA, USA
PAUL EMSLEY • Department of Biochemistry, University of Oxford, Oxford, UK
ANDREW D. FERGUSON • Discovery Sciences, AstraZeneca Pharmaceuticals, Waltham,
MA, USA
JOHN FINN • Trius Therapeutics, San Diego, CA, USA
GARY L. GILLILAND • Centocor R&D Inc., Radnor, PA, USA
ISAAC D. HOFFMAN • Takeda San Diego, San Diego, CA, USA
MICHAEL J. HUNTER • Centocor R&D Inc., San Diego, CA, USA
ANDY JENNINGS • Takeda San Diego, San Diego, CA, USA
FELICE C. LIGHTSTONE • Lawrence Livermore National Laboratory, Physical and Life
Sciences Directorate, Livermore, CA, USA
JINQUAN LUO • Centocor R&D Inc., Radnor, PA, USA
SHAWN P. MADDAFORD • NeurAxonInc, Mississauga, ON, Canada L5K 1B3
TOAN B. NGUYEN • Lawrence Livermore National Laboratory, Physical and Life
Sciences Directorate, Livermore, CA, USA
GYORGY SNELL • Takeda San Diego, San Diego, CA, USA
RONALD V. SWANSON • Centocor R&D Inc., San Diego, CA, USA
LESLIE W. TARI • Trius Therapeutics, San Diego, CA, USA
OMID VAFA • Centocor R&D Inc., Radnor, PA, USA
SERGIO E. WONG • Lawrence Livermore National Laboratory, Physical and Life
Sciences Directorate, Livermore, CA, USA

ix
 Chapter 1

The Utility of Structural Biology in Drug Discovery

Leslie W. Tari

Abstract
Access to detailed three-dimensional structural information on protein drug targets can streamline many
aspects of drug discovery, from target selection and target product profile determination, to the discovery
of novel molecular scaffolds that form the basis of potential drugs, to lead optimization. The information
content of X-ray crystal structures, as well as the utility of structural methods in supporting the different
phases of the drug discovery process, are described in this chapter.

Key words: X-ray crystallography, Structure-based drug design, Fragment screening, Structural bio-
informatics, Lead optimization

1. Introduction

The discovery of new drugs is a time and labor-intensive process.

On average, the discovery of a new drug requires the preparation
and evaluation of approximately 10,000 compounds over 12 years
at a cost of more than $350 million (1). Once in the marketplace,
many drugs fail to recover their development costs (as many as
30%, according to data from the 1980s (2)), and many others are
ultimately withdrawn from the market. These facts coupled with
limits on patent lifetime, escalating global competition, and increas-
ingly stringent government regulations for drug approval have
demanded more efficient and accelerated approaches to drug dis-
covery. Conventional “brute force” methods of lead discovery via
high-throughput screening (HTS) of proprietary synthetic, com-
binatorial, or natural product libraries, while effective in many
cases, are expensive and have limitations; they require access to
large compound libraries (sometimes over 1,000,000 compounds),
often yield hits with high molecular weight, poor ligand efficiency,

Leslie W. Tari (ed.), Structure-Based Drug Discovery, Methods in Molecular Biology, vol. 841,
DOI 10.1007/978-1-61779-520-6_1, © Springer Science+Business Media, LLC 2012

1
2 L.W. Tari

limited or no potential for optimization, and provide no information

to guide ligand optimization.
Advances in crystallographic methods, computational power,
molecular biology, and recombinant protein expression systems
over the last 30 years have provided researchers with rapid and reli-
able access to three-dimensional structural information on a wide
variety of protein drug targets. Structural information on protein–
ligand complexes can eliminate much of the complexity involved in
the discovery and optimization of prospective drug leads. Indeed,
structure-guided drug design efforts have led to the discovery of
high profile drugs in multiple therapeutic areas, including the pep-
tidomimetic HIV protease inhibitors for the treatment of HIV, the
neuraminidase inhibitor Tamiflu™ for the treatment of influenza,
the carbonic anhydrase inhibitor dorzolamide for the treatment of
glaucoma, and the thrombin inhibitor ximelagatran, an oral anti-
coagulant (3). Access to structural information on the target of
interest can streamline all aspects of drug discovery, from target
selection to lead discovery and optimization, using methods that
are summarized in this chapter.

2. The Information
Content of Protein
Crystal Structures
Protein crystals, like any crystalline substance, are regular, three
dimensionally periodic arrays of identical molecules or molecular
complexes (see Fig. 1). A common misconception regarding pro-
tein crystal structures is that they are not representative of the pro-
tein in solution due to the influence of extensive intermolecular
interactions present in the crystalline state. The idea that protein
crystal structures are heavily biased by “solid state” artifacts arises
from inaccurate comparisons made between protein crystals and
crystals of small molecular weight compounds. Crystals of small
molecules and proteins differ in ways that extend beyond the prop-
erties of their component molecules. Small-molecule crystals typi-
cally only comprise the small molecule, while protein crystals
contain 25–90% solvent by volume, depending on the protein. The
remaining volume in protein crystals is occupied by protein mole-
cules, and is analogous to an ordered gel with large interstitial
spaces between protein molecules. By comparison, the number of
contacts made in relation to the molecular mass of the protein in
protein crystals is smaller by orders of magnitude than it is for
small-molecule crystals. This causes the mechanical stability and
integrity of protein crystals to be much worse than it is for crystals
of small molecules. The high solvent content and tenuous thermo-
dynamic stability of protein crystals complicate the subsequent
steps in X-ray diffraction experiments, since these properties result
in crystal handling difficulties, susceptibility to temperature changes
 1 The Utility of Structural Biology in Drug Discovery 3

Fig. 1. A view of crystal packing in a Haemophilus influenzae dihydrofolate reductase crystal. Boundaries for a single unit
cell within the crystal are shown. The view is perpendicular to the c-axis of the unit cell. The unit cell is the fundamental
building block of the crystal, a translationally periodic substance comprising trillions of unit cells that extend in three
dimensions. The unit cell is an arbitrary construction that describes the smallest “box” with the highest metric symmetry.

and dehydration, weaker diffraction, and greater sensitivity to radiation

damage. However, the key role played by solvent in protein crys-
tallization is a double-edged sword; while it adversely affects dif-
fraction, it is the very element that makes protein crystal structures
valuable. The high solvent content of protein crystals is essential
for maintaining the structures of the macromolecules in their solu-
tion states. Therefore, to a large extent, proteins in crystals possess
the structural, enzymatic, and functional properties of their coun-
terparts in solution. Protein crystal structures must be regarded
with care, however. In the hands of the uninformed, the danger
exists that crystallographic structural data will be misinterpreted,
or overreaching conclusions drawn. An understanding of the
parameters derived from crystallographic experiments is essential if
structural information from crystallographic experiments is to be
used effectively to support drug discovery.
X-ray crystallography and light microscopy share the same
basic principle; electromagnetic radiation scattered by the object to
be imaged is recombined and focused by a lens to reform the image
of the object. Theoretically, the resolving power of any imaging
technique is equal to one half of the wavelength of the radiation
used for imaging. To resolve the atomic details of protein struc-
tures, crystallographic experiments involve the exposure of protein
crystals to high-energy monochromatic X-rays (wavelengths on
the order of 1 Å). Imaging using X-rays is complicated by the fact
4 L.W. Tari

Fig. 2. A schematic outlining the steps in a crystallographic structure determination. Crystals are systematically exposed to
monochromatic X-rays in multiple orientations, and the diffraction patterns are captured with electronic detectors. Since
crystals are three-dimensionally periodic substances, the diffraction pattern comprises a series of spots rather than a
continuous function. Each spot represents a family of diffracted waves that map to discrete spatial periodicities in the unit
cell of the crystal. The diffraction pattern is a summation of waves of electromagnetic radiation and can thus be described
by a Fourier series, and the diffraction pattern and disposition of the atomic contents of the unit cell are related mathemati-
cally by a Fourier transform. An image of the atomic contents of the unit cell of the crystal is derived by applying a math-
ematical lens (inverse Fourier transform, equation shown on the lower left ) to the diffracted X-rays. The image reconstruction
process is complicated by the fact that only intensities of the diffracted X-rays are measurable (F (h) terms in the equation
shown), but not the relative phase shifts between each family of diffracted waves. The missing information is referred to
as the crystallographic phase problem. The missing phases are obtained using other experimental or computational meth-
ods described in the text. Since the diffraction of X-rays is caused by the interaction of the X-rays with electrons, the
resulting image obtained in a crystallographic experiment is of the electron density distribution in the unit cell of the crystal.
Interactive model building software is used to build the final atomic model into electron density.

that X-rays interact very weakly with matter, so that no lenses exist
which are able to reconstruct the image from the scattered X-rays.
Hence, the scattered X-rays from crystals must be captured with
electronic detectors and the function of a lens must be simulated
mathematically. A schematic describing the steps involved in the
solution of a crystal structure is shown in Fig. 2.
Mathematical reconstruction of the structure of the atomic
contents of the crystal is complicated by the fact that one of the
two key pieces of information describing the diffracted X-ray waves,
the relative phase shifts between the different families of diffracted
 1 The Utility of Structural Biology in Drug Discovery 5

waves, cannot directly be measured (see Fig. 2). Three methods

are commonly employed to overcome the phase problem, as sum-
marized below.
(a) Molecular replacement. When an approximate model of the
unknown crystal structure is available, it can be used to over-
come the phase problem. The principle is simple; the model is
first oriented and then positioned in the unit cell of the target
crystal structure using rotation and translation functions. The
correctly oriented model is subsequently used to calculate
approximate phases and electron density maps. Alternate cycles
of interactive correction and rebuilding of the model into elec-
tron density and model refinement are used to improve the
quality of the phases and to transform the model structure into
the real structure. The success of molecular replacement
depends critically on two factors: the fraction of the asymmet-
ric unit for which suitable models exist, and the r.m.s. devia-
tion (after optimal superposition) between the model and
target structures. Generally, r.m.s. deviation increases with
decreasing sequence identity, or in cases where the target struc-
ture undergoes significant conformational changes with respect
to the model structure (e.g., movement of protein domains).
In the latter case, the model structure can be separated into
individual fragments that are sequentially oriented and posi-
tioned in the unit cell. Newer maximum-likelihood molecular
replacement algorithms, such as those implemented in the pro-
gram Phaser (4) are more discriminating, and have been suc-
cessful in solving difficult molecular replacement problems that
were previously intractable.
(b) Isomorphous replacement methods. This is a classical approach
used to solve protein structures with unknown folds. Crystals
are soaked in multiple solutions containing salts of heavy atoms
such as Hg, Pt, Pb, Au, etc., until conditions are found where
a small number of heavy atoms incorporate in well-defined
positions on the crystallized protein molecule (without alter-
ing the structure of the underlying protein). By analyzing the
differences in the intensities of diffraction patterns from the
native and heavy atom derivatized protein crystals, it is possible
to determine the locations of the heavy atoms in the unit cell
and to use the scattering “signal” from the heavy atoms to
calculate phases and an electron density map (reviewed in refs.
(5–7)).
(c) Anomalous scattering methods. For heavier elements, some
inner shell electrons have absorption edges in the range of the
X-ray wavelengths used in diffraction experiments. The heavy
atoms in the protein crystal cause absorption of the impinging
radiation, and impart small phase shifts on the radiation scat-
tered from the crystal. This phenomenon is used to determine
6 L.W. Tari

the positions of the heavy atoms in the unit cell, and subsequently
to extract phase information to allow electron density map
generation. Anomalous scattering can be used to supplement
the phase information obtained from isomorphous heavy atom
derivatives, or to independently obtain complete phase infor-
mation. A very powerful de novo phase determination method
utilizes anomalous scattering from proteins that are homoge-
neously labeled with selenomethionine (incorporated during
recombinant expression of the protein in Escherichia coli), a
derivatized selenium-containing amino acid. Independent dif-
fraction experiments are carried out (on the same crystal, if
possible) at multiple X-ray wavelengths on the high and low
energy sides of the selenium absorption edge that maximize
the anomalous diffraction signal. This method requires a tun-
able X-ray source, which is present only at synchrotrons
(reviewed in refs. (5–7)).
X-ray diffraction is caused by the interaction of the electric
field vector of monochromatic X-rays with electrons in a protein
crystal. These details, coupled with the fact that crystals are made
up of three-dimensionally periodic lattices of molecules, have sev-
eral important consequences (for excellent reviews see refs. (5–7)):
(1) X-ray diffraction experiments generate three-dimensional
images of the electron density distribution of the molecular com-
ponents of the crystal. So heavier atoms generate a proportionally
stronger signal, and hydrogen atoms are generally not discernable
in protein crystal structures. (2) The short wavelength radiation
used in X-ray diffraction experiments allows for the resolution of
macromolecular structures at an exquisite level of detail (typical
protein crystal structures are determined at resolutions between
1.5 and 3.0 Å resolution). (3) In a crystallographic experiment, the
structure of the molecular contents of the unique portion of a crys-
tal (called the asymmetric unit of the unit cell, which is the micro-
scopic building block of the crystal) are obtained, and the resulting
crystal can be built by the application of crystallographic symmetry
operators to the contents of the asymmetric unit, as shown in
Fig. 1. Since the diffraction signal from a crystal arises from con-
structive interference from trillions of crystallographic asymmetric
units, the resulting crystal structure comprises a time- and space-
averaged picture of the contents of the copies of asymmetric units
that are sampled. Hence, components of the asymmetric unit with
a large degree of random spatial heterogeneity, i.e., disordered
protein loops or side chains and the bulk solvent occupying the
spaces between protein molecules, fade into the background and
cannot be modeled. However, in cases where a molecular compo-
nent of a crystal, such as a protein side chain, occupies a finite
number of distinct, low energy conformations in different asym-
metric units, it is possible to simultaneously characterize each alter-
native conformation.
 1 The Utility of Structural Biology in Drug Discovery 7

Examination of the equation relating diffracted X-rays to the

crystal structure provides insight into the structural parameters
that are modeled in a crystallographic experiment (see Eq. 1).
N
Fhkl = ∑ f j e − (B sin
2
θ)/ λ 2
e2 πi (hx + ky +lz ) . (1)
j =1

Equation 1 is one of the explicit forms of the structure factor

equation (8). Each Fhkl term represents a unique family of diffracted
X-ray waves from the crystal (diffracted waves from crystals con-
structively interfere to form patterns of spots, as shown in Fig. 2,
which can each be assigned integer indices h, k, and l ), which cor-
respond to discrete spatial periodicities in the crystal lattice. The
intensity and phase of each family of diffracted waves is derived via
a summation of the scattering contributions from all of the atoms
in the asymmetric unit of the crystal. The second exponential term
in Eq. 1 computes the net phase shift relative to an arbitrary origin
of the scattered wave with index h, k, l due to the relative positions
of the individual atoms in the unit cell (with fractional coordinates
x, y and z). The fj term corresponds to the scattering factor for each
atom in the summation, and is directly proportional to the number
of electrons in the atom in question. The first exponential B sin2 θ/λ2
term (θ is the angle of the scattered radiation with respect to the
source X-ray beam, and λ is the wavelength of the X-rays) accounts
for the reduction in the intensity of the scattered radiation with
scattering angle due to interference between scattered waves from
different parts of the electron cloud surrounding each atom. X-ray
scattering is attenuated further by smearing of the electron clouds
surrounding each atom due to thermal motion of the atoms.
Atomic thermal motion is modeled using the extra B term in the
structure factor equation. As a first approximation it is assumed
that the thermal motion of atoms is isotropic (spherically symmet-
ric), with B = 8π2μ2, where μ is the root mean square amplitude of
atomic vibration. Using the calculation above, for a B-factor of
15 Å2, the displacement of an atom from its equilibrium position is
approximately 0.44 Å, and it is as much as 0.87 Å for a B-factor of
60 Å2. Thus, analysis of B-factors is very important during any
structural analysis to provide insight into the dynamics and struc-
tural integrity of different regions of a protein molecule. However,
one must exercise caution before interpreting B-factors too quan-
titatively. In addition to measuring dynamic disorder caused by
temperature dependent vibration of atoms, the B-factor is also
influenced by subtle structural differences between protein mole-
cules in different unit cells throughout the crystal (which spatially
smears the atom positions), steric constraints from intermolecular
lattice contacts, and certain systematic experimental errors, such as
absorption of the X-ray beam during X-ray data collection.
Advanced mathematical models can be used to provide more
8 L.W. Tari

detailed information on atomic thermal motions. For example, the

relative motions of entire protein domains can be characterized
using TLS refinement (9). Also, when high-quality X-ray data are
available from crystals that diffract to high resolution (typically
better than 1.2 Å, rare in protein structure determinations), the
isotropic thermal correction can be replaced by a tensor, which
corrects not only for the extent of thermal motion of the atoms but
also for spatial anisotropy in their motions (10).
Based on the mathematical description of X-ray diffraction
provided above, four parameters are optimized in a single crystal
X-ray diffraction experiment for each atom in a protein crystal
structure: the x, y, and z coordinates of each atom and the B-factor
describing the thermal motion of each atom. The quality of
resulting electron density maps and the accuracy of refined para-
meters in protein crystal structures are largely dependent on the
resolution of the X-ray diffraction data (equivalent to the pixel size
of electron density sections). Examples of the effects of diffraction
resolution on electron density map quality are shown in Fig. 3.
The model is generally manually built (or refit) into electron density
by a crystallographer, using two types of electron density maps,
|2Fo − Fc|αc maps, and |Fo − Fc|αc difference maps, described below.

Fig. 3. Representative electron density maps contoured around tyrosine residues (using |2Fo − Fc|αc coefficients) from three
refined crystal structures: (a) A 2.8 Å resolution structure of Francisella tularensis topoisomerase IV, (b) A 2.2 Å structure of
Escherichia coli topoisomerase IV, and (c) A 1.4 Å structure of Enterococcus faecalis DNA gyrase B (all from D. Bensen and
L. Tari, unpublished results). The electron density maps were contoured using the electron density visualization software COOT
(see ref. (11), Chapter 6). At better than 3.0 Å resolution, amino-acid side chains can be recognized with the help of protein
sequence information, while at better than 2.5 Å resolution solvent molecules can be observed and added to the structural
model with some confidence. As the resolution improves to better than 2.0 Å resolution, fitting of individual atoms may be
possible and most of the amino-acid side chains can be readily assigned even in the absence of sequence information.
 1 The Utility of Structural Biology in Drug Discovery 9

|Fo − Fc|αc maps. |Fo − Fc|αc maps, or difference maps, are generated
by subtracting the calculated structure factor amplitudes (Fc, from
the best current model structure) from the observed structure
factor amplitudes (Fo), using phase information (αc) calculated
from the available model structure. To a good approximation, this
operation is equivalent to subtracting the electron density calcu-
lated from the model from the “real” electron density in the crystal.
What is left behind is the electron density for ordered components
of the crystal structure that have not been accounted for by the
model, or that have not been modeled correctly. Features that are
present in the true structure that have not been accounted for in
the model structure appear as positive peaks, while atoms that have
been incorrectly placed in the model structure (i.e., that do not
exist in the real structure) appear as holes or negative peaks. These
maps are used to fix improperly modeled side-chains and/or entire
polypeptide chains, as well to fit substrates, inhibitors, and ordered
solvent molecules into the structure. A special type of difference
map called an omit map can be used to confirm the presence of
important features in a protein structure. An omit map is calcu-
lated by removing the feature of interest (say, an inhibitor) from
the model, refining the structure in the absence of that feature, and
calculating a new difference map. If the feature of interest is still
observed in a difference density map, then it is real, and not an
artifact caused by model bias present in the calculated phases. An
example of a difference map is shown in Fig. 4.
|2Fo − Fc|αc maps. |2Fo − Fc|αc maps are the maps most commonly
used for model fitting. They are used instead of |Fo|αc maps, which
suffer from model bias, and tend to show only electron density that
is associated with the model. As described above, |Fo − Fc|αc maps
reveal everything in the |Fo|αc map that has not been modeled. The
|2Fo − Fc|αc map essentially superposes an |Fo|αc map over an
|Fo − Fc|αc difference map, so that it simultaneously shows both the
electron density for the model and the electron density for features
that have not been accounted for by the model. Several weighting
schemes are used to further diminish the effects of model bias,
including figure-of-merit and σA weighting schemes (reviewed in
refs. (5–7)). An example of a |2Fo − Fc|αc electron density map is
shown in Fig. 4.
In addition to providing a more detailed picture of the elec-
tron density, higher resolution X-ray data correlates with a greater
number of experimental observations to support structure refine-
ment. For a typical protein structure from a crystal with a solvent
content of about 50%, the number of experimental observations
and refinement parameters will be about the same at 2.8 Å resolu-
tion. The paucity of experimental data compared with the number
of parameters that need to be defined make least squares model
optimization methods intractable. Additionally, at resolutions
lower than 2.8 Å, individual atomic B-factors have a very limited
10 L.W. Tari

Fig. 4. Examples of |Fo − Fc|αc and |2Fo − Fc|αc electron density maps. The electron density maps in all panels are drawn as
thin chicken-wire representations. In (a) an |Fo − Fc|αc map contoured at 3σ is used to fit an incorrectly modeled glutamic
acid side chain in an E. faecalis GyrB crystal structure. In the model structure, part of the side chain is in a negative electron
density peak, while a positive difference density peak on the left-hand side of the figure reveals the correct position for the
side chain from the experimental data. The correctly positioned glutamic acid side chain is shown in (b). In (c), an |Fo − Fc|αc
difference electron density map contoured at 3.5σ was used to fit a small-molecule inhibitor into the substrate-binding
pocket of E. faecalis gyrase B. The difference map was calculated in the absence of the inhibitor, indicating that the differ-
ence density shown arises entirely from the experimental X-ray data. Panel (d) shows a representative section of a
|2Fo − Fc|αc electron density map contoured at 1σ for an E. faecalis GyrB crystal structure. The map displays electron
density for both regions of the model that have been correctly fit, as well as regions that have not been accounted for by
the model. Because it comprises a superposition of an |Fo|αc map and a |Fo − Fc|αc map, |2Fo − Fc|αc maps are less subject
to the effects of model bias than |Fo|αc maps. During model fitting, crystallographers generally utilize |2Fo − Fc|αc and
|Fo − Fc|αc simultaneously to trace the polypeptide chain and correct errors in the existing model.

physical meaning. The problem of statistical under determination

is overcome by augmenting the X-ray diffraction data with struc-
tural parameters of proteins and peptides derived from small-mol-
ecule crystallography and spectroscopic data. The resulting function
that is minimized in a crystallographic structure refinement incor-
porates the experimental X-ray data and a molecular mechanics
function (which restrains bond lengths, angles, stereochemistry,
planarity of peptide bonds and aromatic groups, etc. to reasonable
values). The quality of structures refined in this fashion is excellent,
even for structures determined at modest resolutions. Properly
refined protein crystal structures generated from carefully mea-
sured X-ray data yield atomic positions that are precise to within
one fifth to one tenth of the stated experimental resolution. Once
a structure is fully refined, multiple criteria are used to judge the
quality of the model, as described below.
 1 The Utility of Structural Biology in Drug Discovery 11

R-factor. The R-factor is the averaged error (in percent) between

the observed structure-factor amplitudes (the experimentally mea-
sured Fhkl values) and the calculated structure-factor amplitudes
(Fhklcalc) from the refined model of the contents of the crystal. The
ultimate value of the R-factor in a well-refined structure depends
on a number of variables, including the proportion of the contents
of the unit cell that can be correctly modeled, the relative weights
assigned to the molecular mechanics restraints vs. the experimental
X-ray data during refinement, the experimental resolution of the
diffraction experiment and the accuracy and overall quality of the
measured experimental X-ray intensities. In protein structures with
numerous dynamically disordered loops or domains that cannot be
modeled, the R-factor will not converge to low values. However,
as a general rule of thumb a correctly refined protein structure
should have an R-factor around 20%.
Free R-factor (Rfree). The function that is minimized during a
protein structure refinement is extremely complex, with multiple
false minima. Hence, when not used with care, modern refinement
algorithms can converge on convincing R-factors for incorrect
structures. The Rfree (12) statistic is an extremely simple and pow-
erful independent validation tool used in modern protein structure
refinement. The Rfree function is identical to the R-factor; the only
difference is that it is calculated using a small (5–10%) randomly
sampled subset of the X-ray diffraction data that is excluded from
structure refinement throughout the refinement process. In a cor-
rectly refined structure, Rfree will track with the R-factor to within
5–10%. For incorrect structures, Rfree will remain at a value near the
limit observed for random atomic models fit to an X-ray dataset
(~57%). In addition to Rfree, the geometric quality of the refined
protein structure should be used to evaluate the model. The aver-
aged bond lengths and angles of the final model should not deviate
much from ideal values (r.m.s. deviations from ideality should be
within 0.02 Å for bond lengths and 3° for bond angles), and the
majority of the protein residues should possess “allowed” combi-
nations of φ, ψ main-chain dihedral angles. It is important to note
that protein folding can force some residues into disallowed φ, ψ
values, which can have important functional significance (13). All
residues in disallowed regions must be carefully checked to ensure
that they are well described by experimental electron density.
Identification and refinement of ordered solvent molecules
becomes more reliable when data are available to at least 2.5 Å
resolution. Even then, before a water molecule is used in mecha-
nistic or computational analysis, it is always wise to check its
B-factor and to see if there exists at least one hydrogen bond to
hold the water to the protein or a nearby solvent molecule.
Unless the structure has been determined at very high resolu-
tion, electron density and refinement do not discriminate between
the oxygen and nitrogen atoms of asparagines and glutamines, or
12 L.W. Tari

the alternative conformations of histidine side chains. In a detailed

structural analysis, it is always necessary to check alternative con-
formations of Asn, Gln, or His side chains to decide which one
makes more sense chemically (i.e., by analyzing available H-bonding
networks). Also, great care has to be exercised when fitting dynam-
ically disordered protein side chains that are not fully described by
electron density. The crystallographer knows they are present from
the amino-acid sequence, and incorporates them in conformations
commonly observed for that side chain from databases of high-
resolution structures. The final refined conformation of the side
chain must ultimately be decided using the crystallographer’s
knowledge of chemistry and side-chain conformational prefer-
ences, in conjunction with the refinement program’s force field. In
many structures, entire loops or even domains are too disordered
to show any observable electron density. In such cases, the offend-
ing loops/domains are not included in the final model. When ana-
lyzing crystal structures, an additional point of caution that must
be noted regarding potential artifacts that can arise from contacts
between adjacent molecules in a crystal lattice. In the ideal sce-
nario, the protein of interest crystallizes in a lattice that leaves the
active-site/receptor pocket solvent exposed, with no lattice con-
tacts preventing the motion of functionally important mobile
structural elements surrounding the drug-binding site (i.e., the lat-
tice should not impede ligand-induced conformational changes in
the protein). However, protein crystallization does not allow for
control of lattice contacts, and the ideal situation does not always
exist. Hence, before a new protein crystal form is nominated as a
potential candidate for supporting structure-based drug design, a
careful analysis of the crystal lattice contacts between neighboring
molecules related by crystallographic or noncrystallographic sym-
metry must be carried out to assess the steric accessibility of the
receptor pocket and the solvent space around it, as well as the
nature and quantity of lattice contacts in the vicinity. This sort of
analysis is particularly important if the crystals are produced for the
purpose of ligand soaking experiments to support fragment screen-
ing or high throughput structure determination. If multiple crystal
forms are available, the crystal forms that approach the ideal crite-
ria should be chosen. Cocrystallization experiments usually cir-
cumvent problems related to lattice constraints, since the protein
and ligand are mixed in solution, allowing the system to reach a
low energy conformational state before crystallization occurs.
Additional important parameters to consider when analyzing crys-
tal structures are the solution conditions used in crystallization.
Some proteins undergo significant structural changes in different
solution conditions. A classic example is ribonuclease A, which
undergoes large, pH-dependent conformational changes that have
been characterized crystallographically (14).
 1 The Utility of Structural Biology in Drug Discovery 13

3. Using Structure
in Target Selection
and Product
Profile In addition to supporting lead discovery and lead optimization,
Development structural information can be used at a very early stage in a drug
discovery program to evaluate the viability of a protein as a drug
target. Does the protein possess a binding pocket with suitable
properties for potent inhibitor development? In a large, structur-
ally related protein family, such as eukaryotic protein kinases, is it
possible to develop selective inhibitors against a kinase of interest?
More generally, what are the prospects for the development of spe-
cific inhibitors against a protein target while avoiding off-target
binding? In an antibiotic program, do the protein orthologs
encompassed by the proposed target product profile possess suffi-
cient structural homology to allow for the development of a small-
molecule agent with the desired spectrum? Careful analysis of the
structures of the protein target(s) of interest coupled with struc-
tural bioinformatics and molecular modeling can be used to address
questions such as those posed above. Such an analysis is important
to expose liabilities in target selection or the proposed drug prod-
uct profile early in a drug discovery program, before a substantial
investment of time, money and manpower has been made to pur-
sue a flawed hypothesis.
For example, in the antibacterial arena, the emergence of
genomics and proteomics has profoundly changed the approach
used for the identification of new targets essential for the survival
of bacteria (15). To highlight how this information is used to facili-
tate target selection, the analysis that led to the selection of bacte-
rial topoisomerases as prospective drug targets at the author’s
company is summarized below. To pursue a drug discovery pro-
gram, we sought essential bacterial targets with the following prop-
erties: (1) Novel proteins that are not targets of marketed antibiotics,
to avoid issues of cross-resistance with existing antibiotics. (2)
Targets possessing recessed ligand-binding pockets with mixed
polar/lipophilic character, the potential for solvent sheltered
“anchoring interactions” and no closely related human counter-
parts. (3) A high degree of sequence/structure conservation in the
ligand-binding pockets of the protein target(s) across bacterial spe-
cies commonly implicated in bacterial infections. (4) If possible,
the option to inhibit multiple bacterial targets with a single
therapeutic agent to minimize the threat of resistance emergence.
A detailed structural bioinformatics analysis of proteins in several
key bacterial pathways revealed the bacterial topoisomerases DNA
gyrase and topoisomerase IV as prospective drug targets that met
the criteria listed above. DNA gyrase is a type II topoisomerase
that plays an essential role in bacterial DNA replication with no
direct mammalian counterpart. The enzyme catalyzes the intro-
duction of negative supercoils into DNA using the free energy of
14 L.W. Tari

ATP hydrolysis (16). DNA gyrase consists of two subunits, GyrA

and GyrB that form a functional heterodimer A2B2. GyrA is involved
in DNA cleavage and religation, while the GyrB domain contains
the ATP-binding site and mediates the passage of the uncut DNA
strand through the strand that is cleaved by GyrA (16). A closely
related bacterial enzyme from the topoisomerase II family is topoi-
somerase IV (topo IV), which also forms a heterodimer C2E2 con-
sisting of two ParC subunits and two ParE subunits (17). Despite
possessing a high degree of sequence identity with DNA gyrase,
topo IV is involved in different aspects of DNA replication than
gyrase. The two topoisomerase complexes are well established drug
targets. Fluoroquinolone antibiotics, such as ciprofloxacin, exert
their antimicrobial activity via inhibition of the GyrA and ParC
subunits (18). However, no commercial antibiotics have yet
reached the market which target the ATP binding domains of the
respective topoisomerase complexes (GyrB and ParE), despite the
fact that GyrB and/or ParE inhibition has been shown to effec-
tively kill bacteria (19). A sequence alignment of the ATP-binding
domains of DNA gyrase and topo IV from key pathogens involved
in community acquired pneumonia mapped on to the crystal structure
of one of the enzymes (see Fig. 5), suggests that the development

Fig. 5. A solvent accessible surface representation of the ATP-binding pocket of GyrB from
the crystal structure of E. faecalis GyrB complexed with a benzimidazole inhibitor
(D. Bensen and L. Tari, unpublished results). The surface is colored by the degree of sequence
conservation observed in the underlying residues for GyrB and ParE enzymes from the
major pathogens implicated in community acquired pneumonia. Amino-acid sequences
for the relevant proteins were extracted from the KEGG database (20) and sequence align-
ments were performed with CLUSTALW (21). The high degree of overall sequence conser-
vation (not shown) and the remarkable degree of sequence conservation in the ATP-binding
pockets of the selected GyrB and ParE orthologs suggest that the geometries and compo-
sitions of the active sites of the enzymes from the different pathogens possess sufficient
similarity to allow for the development of dual targeting, broad spectrum inhibitors.
Subsequent generation of homology models and crystal structures of several of the
orthologs listed on the figure confirmed this hypothesis.
 1 The Utility of Structural Biology in Drug Discovery 15

of broad spectrum, dual-targeting inhibitors against these enzymes

is feasible. As the above example demonstrates, structural bioinfor-
matics can be an important component in the target selection pro-
cess and drug product profile determination early in a drug
discovery program.

4. Using
Crystallographic
Methods to Initiate
a Drug Discovery The likelihood of success in a small-molecule drug discovery pro-
Program gram is greatly enhanced by the availability of multiple molecular
scaffolds that bind to and elicit the desired effects on the protein
target, while offering prospects for optimization into drug leads.
However, the discovery of viable molecular scaffolds for SBDD
and medicinal chemistry optimization is not trivial. HTS, when
successful, often delivers hits with high molecular weights and poor
potential for optimization. The probability of a small-molecule
ligand matching the shape and chemistry of a protein target
decreases as the complexity and size of the ligand increases, since
there exists a greater chance that some part of the ligand will pos-
sess features that do not complement those of the protein target.
Theoretically, the probability that a small molecule will bind to a
protein target decreases exponentially with increasing ligand com-
plexity (22). Thus, there is an advantage to screening for hits using
less complex, lower molecular weight compounds (called frag-
ments, with molecular weights ranging from 100 to 250 Da),
which interact with only a small number of sites on the protein and
possess a greater chance of achieving favorable steric and chemical
complimentarity with the protein target. However, the advantage
of screening with fragments is offset by the fact that fragments
generally bind with much lower affinities than the larger com-
pounds typically screened in HTS. Most biophysical techniques
perform poorly at detecting weak binding, limiting their utility in
screening fragment libraries. X-ray crystallography, however, is an
extremely sensitive technique, capable of detecting compounds
with binding constants in the low millimolar range. The extension
of crystallographic methods into the high-throughput realm over
the past decade has led to the adoption of crystallographic frag-
ment screening in many industrial and academic centers as a drug
discovery tool. In this section, the two flavors of crystallographic
fragment screening are reviewed: random fragment screening and
pharmacophore-based fragment screening.

4.1. Random The basic premise of crystallographic fragment screening is simple.

and Pharmacophore- A protein target is screened against a small library (typically <1,000
Based Fragment molecules) of structurally diverse, highly soluble low molecular weight
Screening compounds. The library is screened in one of two ways: pregrown
16 L.W. Tari

crystals of the protein of interest are soaked with concentrated

solutions (in aqueous solution or dimethyl sulfoxide) of individual
compounds or mixtures of compounds, or, the protein is crystal-
lized in the presence of compounds/compound mixtures. The lat-
ter method has the advantage of allowing the protein to undergo
compound induced conformational changes that may be precluded
in a preformed crystal lattice. However, cocrystallization generally
involves a scan of multiple crystallization conditions to generate
usable crystals and can lead to multiple crystal forms, so it is more
labor intensive, and requires much larger quantities of protein and
fragment solutions for screening. Once putative protein-fragment
complex crystals are created, X-ray data collection, crystal structure
solution and electron density map interpretation can proceed in a
high throughput manner, using high-flux laboratory or synchro-
tron X-ray sources equipped with sample handling robotics
(described in Chapter 5), and automated software for structure
solution, refinement and electron density map generation (described
in Chapter 6). A schematic representation of random fragment
screening and optimization paths from initial fragment hits is
shown in Fig. 6. Fragment screening methods are described in
more detail in Chapter 7.
An absolute requirement for the application of crystallographic
fragment screening is a target protein that is amenable to crystal-
lization (in its apo-form if crystal soaking experiments are used to
introduce fragments to the target). Moreover, the crystals must
routinely diffract to an adequate resolution (beyond 2.5 Å) to pro-
vide a detailed picture of the targeted binding pocket in the pro-
tein and the binding modes of bound fragments, and optimally, to
resolve ordered waters. When crystal soaking methods are
employed, the crystals must possess sufficient mechanical stability
to withstand the osmotic pressure generated during exposure to
concentrated fragment solutions, and provide unblocked access to
the target site. Once these prerequisites are met, crystallographic
fragment screening provides insights not offered by other screen-
ing techniques. Crystallographic experiments generate electron
density maps showing the binding mode of the fragment to the
protein target in three dimensions. A detailed knowledge of the

Fig. 6. A schematic representation of crystallographic fragment screening with random fragments. Fragment screening
libraries typically contain <1,000 structurally diverse compounds that meet the following criteria: (1) Molecular
weights <300 Da, (2) Less than three H–bond donors/acceptors, polar surface area <60 Å2, less than 3 rotatable bonds, (3)
c Log P < 3. The libraries can be screened using mixtures of 3–5 compounds, or using individual fragment solutions. (a) Two
fragments (the square and triangle) are shown binding to distinct regions of a target receptor binding-pocket. Based on
crystal structures of the fragment-receptor complexes, several optimization strategies can be employed. Fragments bound
to spatially distinct regions of the receptor can be linked as in (b) to form a more potent inhibitor, with an inhibition constant
(Ki ) proportional to the products of the Kids of the individual fragments. Or, as in (c), individual fragments can be optimized
to improve the steric and chemical fit with the receptor pocket. Additionally, fragments can be elaborated or “grown” into
adjacent pockets in the receptor site, as shown in (d).
1 The Utility of Structural Biology in Drug Discovery 17
18 L.W. Tari

key binding interactions, available analoging vectors off the bound

fragment(s) and accessible space in the protein defines the spatial
and chemical constraints on fragment optimization, streamlining
the optimization process. Additionally, crystallographic fragment
screening facilitates the identification of false positive hits and frag-
ments that bind to nonproductive sites on the protein target.
Conversely, fragment screening can identify novel binding sites
that impact protein function with therapeutic development poten-
tial. The weakness of X-ray crystallography as a screening method
is that it does not provide information on binding affinity and this
data must be obtained with a different technique (i.e., a solution
based assay, as described in Chapter 8). However, obtaining the
binding affinity of initial small fragment hits may be intractable and
testing for potency may only become feasible once elaborated
follow-on compounds are available.
By utilizing custom designed chemical fragment libraries based
on a known target pharmacophore model (as summarized in
Fig. 7), pharmacophore-based fragment screening differs philo-
sophically from the use of random chemical fragments in screen-
ing. Small molecules are designed or selected from commercial
libraries, to key in on specific H-bonding, electrostatic, lipophilic,
or π–stacking interactions in the receptor pocket of the target. The
same criteria used in random fragment library design for fragment

Fig. 7. Schematic outline of pharmacophore-based fragment screening. In this example, a simplified receptor-binding
pocket is shown that contains a closely spaced pair of H-bond donor/acceptor moieties comprising the pharmacophore
used to guide fragment library design. (a) Owing to steric incompatibilities, larger molecules frequently cannot bind to the
target, despite containing a potentially useful core. (b) To circumvent the problems observed in (a), and to find novel inhibi-
tor cores to form the basis for potent inhibitors, pharmacophore-based screening methods are effective. Using cheminfor-
matics software such as MOE™ (23), novel molecules can be designed and synthesized that contain the desired
pharmacophore, or commercial libraries can be searched for small-molecule entities with the target pharmacophore.
Crystallographic screening methods are then used to screen potential candidates for hits. (c) Using the three-dimensional
structural information describing the binding modes of fragment hits, the fragments are modified and elaborated with new
chemical groups to improve the fit between inhibitor and receptor, and to engage additional pockets in the receptor with
potency increasing interactions.
 1 The Utility of Structural Biology in Drug Discovery 19

size, solubility, etc. are applied to the custom designed libraries

used in pharmacophore-based screening. Starting with small frag-
ment units with known binding interactions, chemical lead series
can be rapidly discovered and optimized for drug-like properties.
The main advantages of designing fragments around a defined
pharmacophore are the creation of highly ligand efficient molecu-
lar scaffolds that target the most energetically rewarding interac-
tions (i.e., nonsolvent exposed, available polar interactions) of the
target receptor pocket, and the ability to design molecules that
achieve the desired selectivity profile by targeting a selected region
of the target receptor pocket. For example, if the product of inter-
est is a broad-spectrum antibiotic against a specific bacterial pro-
tein target, fragment libraries can be designed to engage only the
most conserved regions of the protein target across the different
bacterial species.

5. Using X-Ray
Crystallography in
Lead Optimization
The simplistic view of structure-guided lead optimization is that
structural information from crystallographic structures of com-
plexes of lead candidates with the protein target are used as an
in vitro assay of sorts, to optimize the potency of the lead against
the drug target. However, drug discovery requires optimization of
a number of properties, including solubility, intestinal absorption,
tissue distribution, metabolic stability, plasma protein binding,
elimination, toxicology, and cost of synthesis. To highlight the
importance of using structure-based methods in the broader con-
text of a drug discovery program, it is instructive to follow the
trajectory of neuraminidase inhibitor development that led to the
discovery of the Tamiflu™. Influenza virus neuraminidase has long
been recognized as a potential target in the treatment of influenza.
Molecular modeling studies based on crystal structures of
neuraminidase inhibitor complexes suggested that substitution of
the 4-hydroxyl group (structure 1 in Fig. 8) in a compelling lead
molecule with a charged basic group would yield a more potent
inhibitor (25). Indeed, replacement of the hydroxyl group by a
basic guanidine (structure 2 in Fig. 8) resulted in a 5,000-fold
increase in potency. Ultimately, this compound (zanamivir) was
developed by GlaxoSmithKline and led to the first marketed
neuraminidase inhibitor used in the treatment of influenza,
Relenza™. However, Relenza™ is not absorbed orally due to its
high polarity and basicity, necessitating the development of a dry
powder inhaler to topically dose the compound in the lung (26).
Substitution of the dihydro-2H-pyran scaffold by cyclohexene and
replacement of the polar glycerol and basic guanidinyl moieties
with a 1-ethylpropoxy and a primary amine moiety, respectively,
20 L.W. Tari

Fig. 8. Structures of neuraminidase inhibitors: (1), lead molecule, (2), Zanamivir (Relenza™), (3), Oseltamivir (Tamiflu™).
The ester prodrug is cleaved hepatically to form the carboxylic acid. Semitransparent solvent accessible surface represen-
tations of the structures of both drugs bound to H5N1 avian influenza virus neuraminidase (24) (PDB codes 2HTQ and
2HT8) are shown to illustrate the interaction between the zanamivir guanidine and the active-site pocket, and to highlight
the conformational changes induced by the carbocyclic scaffold and ethylpropoxy group in oseltamivir upon drug binding
to the enzyme.

generate oseltamivir (sold as Tamiflu™, structure 3 in Fig. 8), a

smaller, less polar inhibitor than Relenza™ that retains sufficient
potency for efficacy. The improved ligand efficiency observed for
Tamiflu™ is due in part to the ethylpropoxy group, which induces
a conformational change in key active-site pocket residues and par-
ticipates in lipophilic interactions (24, 27). Conversion of the zwit-
terionic parent compound to the ethyl ester pro-drug allow the
compound to be administered orally, making Tamiflu™ the first
neuraminidase inhibitor used as an oral anti-influenza drug. The
improved physical property profile of Tamiflu™ vs. Relenza™
translated to considerable commercial success. In 2008, Tamiflu™
outsold Relenza™ by a factor of 5:1 (www.marketresearchmedia.
com). This example highlights the difference between good inhibi-
tors and good drugs. When applying structure-based methods,
absorption, distribution, metabolism, and elimination (ADME)
properties need to be addressed during the quest for potency to
 1 The Utility of Structural Biology in Drug Discovery 21

avoid complications later in the drug discovery process. In the case

of the neuraminidase example above, crystallographic studies
revealed a ligand-induced active-site conformation that allowed for
the design of a small, moderately polar, less charged molecule with
superior drug-like properties to the first generation drug. When
used in such a manner, structural information can play a powerful
role in drug discovery. X-ray crystallographic methods can provide
information about active-site or ligand-binding pocket architec-
ture and its relationship to the functional state of a protein, bind-
ing pocket plasticity and small-molecule binding modes that can
dramatically streamline lead optimization. Additionally, crystal
structures can play a key role in resolving unexpected structure
activity relationships (SAR) arising from incorrect small-molecule
structure assignments, unanticipated small-molecule binding
modes or receptor plasticity. Additional examples highlighting
the utility of X-ray crystallography in lead optimization are pro-
vided below.
Once an experimental atomic structure of a protein–small mol-
ecule lead complex is in hand, available analoging vectors off the
lead molecule can be identified by the medicinal chemist and used
to guide the synthesis of the next molecule, as described in Fig. 7.
Vectors pointing toward empty pockets in the receptor can be filled
by complimentary groups to increase potency, while solvent facing
vectors can be used to generate analogs with improved bulk prop-
erties or metabolic stability. Knowledge of the structural and
chemical landscape of the receptor pocket also focuses optimiza-
tion efforts on analogs that add potency mainly via enthalpic (i.e.,
polar) interactions vs. analogs that add potency via entropic (i.e.,
lipophilic) interactions, improving the prospects for selective inhib-
itor binding and reducing the probability of off-target mediated
toxicity. Structure-based methods can also play a key role in more
complex systems, where the protein target must be captured in a
specific functional and structural state to achieve efficacy or to
improve prospects for designing selective inhibitors. For example,
stem cell factor receptor, c-Kit, is a receptor protein-tyrosine kinase
that initiates cell growth and proliferation signal transduction cas-
cades in response to stem cell factor binding (28). The kinase is
activated and transphosphorylates via dimer formation mediated
by the binding of stem cell factor dimers to its extracellular domain.
Mutations that constitutively activate c-Kit in the absence of the
stem cell factor are implicated in several highly malignant human
cancers, making it a validated target for the development of anti-
cancer drugs (29). Detailed analysis of the crystal structures of
c-Kit in multiple functional states (30), including an autoinhibited
form, an activated form, and a drug-bound form, reveal that the
kinase adopts discrete structural states when transitioning from an
autoinhibited to an activated state (see Fig. 9). The structural
results provide a detailed molecular basis for understanding the
22 L.W. Tari

Fig. 9. Ribbon representations of crystal structures of c-Kit kinase in three forms; (a) activated, in complex with ADP and
Mg2+, (b) unphosphorylated, containing the entire juxtamembrane region (autoinhibited state), and (c) in complex with the
anticancer drug Gleevec™. The mobile activation loop is colored black in each panel. Gleevec™ is a fairly selective inhibi-
tor that binds to few kinases, including Abl kinase and platelet-derived growth factor kinase (31, 32). The basis for this
selectivity stems from the fact that the inhibitor targets a kinase conformation that resembles the inactive state (compare
the activation loop conformations in (b) and (c)). However, Gleevec™ binding disrupts the fully autoinhibited state by pre-
venting the association of the juxtamembrane domain with the kinase domain. These results demonstrate that selective
inhibitors of type III protein-tyrosine kinases can be developed to exploit the unique autoinhibited conformations of these
kinases.

mechanism of c-Kit kinase autoinhibition, and snapshots of unique

structural states that are exploitable for the structure-based design
of specific and potent inhibitors targeting the activated or autoin-
hibited conformations of c-Kit kinase, as exemplified by the struc-
ture of c-Kit bound to the anticancer drug Gleevec™ described in
the study.
Many examples exist in the literature that demonstrate the
power of crystallographic methods in revealing receptor plasticity
in protein drug targets resulting in surprising ligand-induced con-
formational changes and inhibitor SAR. A case in point is a mem-
ber of the human histone deacetylase (HDAC) protein family, a
series of validated oncology targets. In eukaryotes, HDACs modu-
late the acetylation of histones and hence play a key role in the
regulation of gene expression (33). HDAC deregulation has been
linked to several types of cancer, and recently, the HDAC inhibitor
suberoylanilide hydroxamic acid (SAHA) was approved for the
treatment of cutaneous T-cell lymphoma (34). Crystal structures
of several inhibitor bound complexes of human HDAC8 (35)
reveal that the surface of the active-site pocket contains flexible
 1 The Utility of Structural Biology in Drug Discovery 23

Fig. 10. Solvent accessible surface representations around the active-site pockets of the structures of complexes of HDAC8
with (a) trichostatin A (TSA) and (b) the anticancer drug suberoylanilide hydroxamic acid (SAHA). TSA induces dramatic
conformational changes in several surface elements of HDAC8, creating a second deep pocket adjacent to the active-site
pocket. A second TSA molecule occupies the newly formed pocket.

elements that can adopt diverse conformations in response to

inhibitor binding. In one of the complexes (see Fig. 10), a loop on
the protein surface moves, revealing a deep pocket adjacent to the
active-site pocket. This work suggests that HDAC8 inhibitors
could be designed with isoform selectivity, despite the highly con-
served nature of HDAC active-site pockets (35). The HDAC
example highlights the importance of using crystallographic meth-
ods for the characterization of novel, low energy conformational
states of protein drug targets that can be exploited for the design
of selective inhibitors. Such insights would not be possible without
the detailed information provided by X-ray crystallography.
In addition to the characterization of receptor plasticity and
the correlation of protein functional states with their underlying
structures, careful application of crystallographic methods can be
used to resolve very detailed questions relating to small-molecule
inhibitor structure, binding mode, and, in many cases, ionization
state. When high-resolution (typically <2.2 Å) X-ray data are avail-
able, cases of mistaken ligand identity can be resolved, or the exact
stereochemistry of a protein-bound small molecule (from a mix-
ture of isomers) can be determined unambiguously, revealing the
stereochemical preferences of the receptor pocket. In favorable
cases, small-molecule ligands can even be fit to electron density
without prior knowledge of the structure of the small molecule.
Additionally, the experimentally observed conformations of inhibi-
tors bound to protein targets can be subjected to in silico confor-
mational analysis to reveal cases where ligand binding to the target
incurs a significant energetic penalty, resulting in reduced inhibitor
potency. Based on the results, molecular modeling can be used to
design optimized inhibitors that “preorganize” into competent
binding conformations in solution, allowing for the development
24 L.W. Tari

Fig. 11. Demonstration of how crystallographic structural information can be used to deduce the protonation states of
ionizable moieties and the hydrogen bonding networks between inhibitors and their protein targets. (a) A close-up view of
the binding of a napthylsulphonyl-amidino-phenylalanine inhibitor to bovine β-trypsin from the high-resolution crystal
structure (36). For clarity, only the interactions between the side chains for the serine protease catalytic triad and the inhibi-
tor are shown. Potential hydrogen bonds are depicted as dotted lines. As shown in (b), the carboxylic acid moiety of the
inhibitor must be protonated, since the hydroxyl proton of Ser195 engages the imidazole ring of His57. The protonation state
of His57 is locked as shown by its interaction with Asp102.

of inhibitors with improved potency often without increasing

molecular weight.
Small-molecule lead optimization is hampered if the protona-
tion states of key acidic and basic amino-acid side chains and/or
the tautomeric states of ionizable groups on protein-bound small-
molecule inhibitors are not understood. Assuming that the param-
eters for amino-acid side chains in a folded protein are similar to
their counterparts in solution is not always correct; the pKa values
for basic and acidic amino-acid side chains on a protein interior can
shift dramatically (>2 units) from their value in aqueous solution,
 1 The Utility of Structural Biology in Drug Discovery 25

based on the microenvironment created around the residue by the

protein structure (36). Furthermore, the pKa values of amino-acid
side chains can change upon complexation by an inhibitor, as can
the pKa values of the ionizable groups on the inhibitor. Although
impossible to determine experimentally from a protein crystal
structure, the positions (or presence, in the case of groups with
exchangeable protons) of hydrogen atoms can usually be inferred
from the molecular mechanics force fields used in structure refine-
ment and a careful analysis of the environment surrounding the
group in question. An example of how X-ray structural informa-
tion can be used to deconvolute complex hydrogen-bonding
networks and assign protonation states to ionizable groups is
shown in Fig. 11. Crystallographic structures and isothermal titra-
tion calorimetry were used to map the network of hydrogen bonds
of several thrombin and trypsin inhibitors, as well as the protona-
tion states of ionizable inhibitor moieties and active-site side
chains (37).

6. Summary

Structure-based drug design is now a staple in the pharmaceutical

industry and has contributed to the discovery of many marketed
drugs and late-stage clinical candidates. Access to detailed three-
dimensional structural information on protein drug targets can
streamline many aspects of drug discovery, from target selection
and target product profile determination, to the discovery of
novel molecular scaffolds that form the basis of potential drugs,
to lead optimization. Structural biology is currently in its golden
era; the advent of high-throughput methods for all of the steps
involved in the generation of protein crystal structures allow
empirically derived structural information to drive iterative lead
optimization efforts in real time for a wide range of protein
targets, avoiding many of the limitations that plague molecular
modeling techniques. Crystallographic methods are useful for
characterizing the structures correlated with specific functional
states of protein targets or alternative conformations of receptor
pockets that lead to unique structural states. Such information
can be leveraged to develop exquisitely selective small-molecule
ligands that target specific proteins, even in closely related pro-
tein families. When used carefully in conjunction with ADME
data during the lead optimization process, X-ray crystallographic
methods are an extremely powerful tool in the drug discovery
arsenal that will continue to contribute to the invention of new
medicines in diverse therapeutic areas.
26 L.W. Tari

References
1. Pharmaceutical Manufacturers Association 17. Peng, H. and Marians, K. J. (1993) Escherichia
(1993) ‘Facts at a Glance’, Washington DC. coli topoisomerase IV. Purification, character-
2. Grabowski, H. J. G. and Vernon, J. M. (1994) ization, subunit structure and subunit interac-
Returns to R&D on new drug introductions in tions. J. Biol. Chem. 268, 24481–24490.
the 1980s. J. Health Econ. 13, 282–406. 18. Wolfson, J. S. and Hooper, D. C. (1985) The
3. Gustafsson, D., Byland, R., Antonsson, T., fluoroquinolones: structures, mechanisms of
Nilsson, I., Nystrom, J. –E, Eriksson, E., action and resistance, and spectra of activity
Bredberg, U. and Teger-Nilsson, A. –C. (2004) in vitro. Antimicrob. Agents Chemother. 28,
A new oral anticoagulant: the 50-year chal- 581–586.
lenge. Nature Rev. Drug. Discov. 3, 649–659. 19. Oblak, M., Kotnik, M. and Solmajer, T. (2007)
4. McCoy, A. J., Grosse-Kunstleve, R. W., Adams, Discovery and Development of ATPase
P. D., Winn, M. D., Storoni, L. C. and Read, Inhibitors of DNA Gyrase as Antibacterial
R. J. (2007) Phaser crystallographic software. Agents. Curr. Med. Chem. 14, 2033–2047.
J. Appl. Cryst. 40, 658–674. 20. Kanehisha, M., Goto, S., Kawashima, S.,
5. Blundell, T. L. and Johnson, L. N. (1976) In Okuno, Y. and Hattori, M. (2004) The KEGG
Protein Crystallography. Academic Press, New resource for deciphering the genome. Nucleic
York. Acids Res. 32, 277–280.
6. Stout, G. H. and Jensen, L. H. (1989) In X-ray 21. Thompson, J. D., Higgins, D. G. and Gibson,
Structure Determination: A Practical Guide. T. J. (1994) CLUSTALW: improving the
2nd ed. Wiley, New York. sensitivity of progressive multiple sequence
7. Drenth J. (1999) In Principles of protein x-ray alignments through sequence weighting, posi-
crystallography. 2nd ed. Springer, New York. tion specific gap penalties and weight matrix
choice. Nucleic Acids Res. 22, 4673–4680.
8. Stout, G. H. and Jensen, L. H. (1989) In X-ray
Structure Determination: A Practical Guide. 22. Hann, M.M., Leach, A.R. and Harper, G.
2nd ed. Wiley, New York, Chapters 7–9. (2001) Molecular complexity and its impact on
the probability of finding leads for drug discov-
9. Winn, M. D., Murshudov G. N. and Papiz, M. ery. J.Chem.Inf.Comput.Sci. 41, 856–864.
Z. (2003) Macromolecular TLS refinement in
REFMAC at moderate resolutions. Methods 23. Labute, P. and Clark A. M. (2007) 2D
Enzymol. 374, 300–321. Depiction of Protein-Ligand Complexes.
J. Chem. Inf. Model 47, 1933–1944.
10. Drenth J. (1999) In Principles of protein x-ray
crystallography. 2nd ed. Springer, New York, 24. Russell, R. J., Haire, L. F., Stevens, D. J.,
pp. 89–90. Collins, P. J., Lin, Y. P., Blackburn, G. M., Hay,
A. J., Gamblin, S. J. and Skehel, J. J. (2006)
11. Emsley, P. and Cowtan K. (2004) Coot: model- The structure of avian flu neuraminidase
building tools for molecular graphics Acta suggests new opportunities for drug design.
Cryst. D60, 2126–2132. Nature. 443, 45–49.
12. Brünger, A. T. (1992) Free R value: a novel 25. von-Itzstein, M., Wu, W. Y., Kok, G. B., Pegg,
statistical quantity for assessing the accuracy of M. S., Dyason, J. C., Jin, B., Van Phan, T.,
crystal structures. Nature 355, 472–475. Smythe, M. L., White, H. F., Oliver, S. W.,
13. Jia, Z., Vandonselaar, M., Quail, J. W. and Colman, P. M., Varghese, J. N., Ryan, D. M.,
Delbaere, L. T. J. (1993) Active-center torsion- Woods, R. C., Bethell, R. C., Hotham, V. J.,
angle strain revealed in 1.6 Å-resolution struc- Cameron, J. M and Penn, C. R. (1993) Rational
ture of histidine-containing phosphocarrier design of potent sialidase-based inhibitors of
protein. Nature 361, 94–97. influenza virus replication. Nature. 363,
14. Bersio, R., Lazmin, V. S., Sica, F., Wilson, K. 418–423.
S., Zagari, A. and Mazzarella, L. (1999) Protein 26. (2001) In Physicians’ Desk Reference. 55th ed.
titration in the crystal state. J. Mol. Biol. 292, Medical Economics Company Inc. Montvale,
845–854. NJ, p.1454.
15. Payne, D. J., Gwynn, M. N., Holmes, D. J. and 27. Kim, C. U. Lew, W., Williams, M. A., Liu, H.,
Pompliano, D. L. (2007) Drugs for bad bugs: Zhang, L., Swaminathan, S., Bischofberger,
confronting the challenges of antibacterial dis- N., Chen, M. S., Mendel, D. B., Tai, C. Y.,
covery. Nat. Rev. Drug. Discov. 6, 29–40. Laver, W. G. and Stevens, R. C. (1997)
16. Champoux, J. J. (2001) DNA topoisomerases: Influenza neuraminidase inhibitors possessing a
structure, function and mechanism. Annu. Rev. novel hydrophobic interaction in the enzyme
Biochem. 70, 369–413. active-site: design, synthesis, and structural
 1 The Utility of Structural Biology in Drug Discovery 27

analysis of carbocyclic sialic acid analogues with factor receptors. J. Pharmacol. Exp. Ther. 295,
potent anti-influenza activity. J. Am. Chem. Soc. 139–145.
119, 681–690. 33. Khochbin, S., Verdel, A., Lemercier, C. and
28. Linnekin, D. (1999) Early signaling pathways Seigneurin-Berny, D. (2001) Functional sig-
activated by c-Kit in hematopoietic cells. Int. J. nificance of histone deacetylase diversity. Curr.
of Biochem. Cell Biol. 31, 1053–1074. Opin. Genet. Dev. 11, 162–166.
29. Hirota, S., Isozaki, K., Moriyama, Y., Hashimoto, 34. Marks, P. A. and Xu, W. S. (2009) Histone-
K., Nishida, T., Ishiguro, S., Kawano, K., deacetylase inhibitors: potential in cancer ther-
Hanada, M., Kurata, A., Takeda, M., Tunio, G. apy. J. Cell Biochem. 107, 600–608.
M., Matsuzawa, Y., Kanakura, Y., Shinomura, Y. 35. Somoza, J. R., Skene, R. J., Katz, B. A., Mol,
and Kitamura, Y. (1998) Gain of function muta- C. D., Ho, J. D., Jennings, A. J., Luong, C.,
tions of c-Kit in human gastrointestinal stromal Arvai, A., Buggy, J. J., Chi, E., Tang, J., Sang,
tumors. Science. 279, 577–580. B. C., Verner, E., Wynands, R., Leahy, E. M.,
30. Mol, C. D., Dougan, D. R., Schneider, T. R., Dougan, D. R., Snell, G., Navre, M., Knuth,
Skene, R. J., Krause, M. L., Schiebe, D. N., M. W., Swanson, R. V., McRee, D. E. and Tari,
Snell, G. P., Zou, H., Sang, B. –C. and Wilson, L. W. (2004) Structural snapshots of human
K. P. (2004) Structural Basis for the autoinhibi- HDAC8 provide insights into the class I his-
tion and ST-571 inhibition of c-Kit tyrosine tone deacetylases. Structure 12, 1–20.
kinase. J. Biol. Chem. 279, 31655–31663. 36. Harris, T. K. and Turner, G. J. (2002) Structural
31. O’Dwyer, M. E., Mauro, M. J., and Druker, B. basis of perturbed pKa values of catalytic groups
J. (2003) STI571 as a targeted therapy for in enzyme active sites. IUBMB Life 53, 85–98.
CML. Cancer Investig. 3, 429–438. 37. Dullweber, F., Stubbs, M. T., Musil, D.,
32. Buchdunger, E., Cioffi, C. L., Law, N., Stover, Sturzebecher, J. and Klebe, G. (2001)
D., Ohno-Jones, S., Druker, B. J. and Lydon, Factorising ligand affinity: A combined ther-
N. B. (2000) Abl protein-tyrosine kinase inhib- modynamic and crystallographic study of
itor STI571 inhibits in vitro signal transduction trypsin and thrombin inhibition. J. Mol. Biol.
mediated by c-kit and platelet-derived growth 313, 593–614.
 Chapter 2

Genetic Construct Design and Recombinant Protein

Expression for Structural Biology
Suzanne C. Edavettal, Michael J. Hunter, and Ronald V. Swanson

Abstract
Obtaining diffraction quality crystals is frequently an iterative process which traditionally has involved
screening large numbers of crystallization conditions. Due to advances in high-throughput gene engi-
neering, recombinant expression, and purification, the protein of interest has now become one of the
many variables routinely investigated during crystallization trials. As such, construct design is a critical step
in the path toward successful crystallization. In this chapter will we address construct design strategies
frequently employed to improve the solution and crystallization behavior of proteins. Topics covered
include choosing a recombinant expression system and reducing disorder through truncations and surface
mutagenesis. Also covered are strategies to reduce heterogeneity from posttranslational modifications,
impurities, and aggregation.

Key words: Protein Expression Constructs, Recombinant Protein Expression, X-ray crystallography,
protein crystallization

1. Introduction

A protein crystal represents a homogeneous population of three-

dimensionally arrayed protein molecules. To maximize the likeli-
hood of crystallizing a protein from a solution, it is important to
minimize the heterogeneity of the sample. The common miscon-
ception is that homogeneity is synonymous with purity at the pro-
tein contaminant level. For protein crystallization, this idea that
purity and homogeneity are synonymous is an oversimplification.
While purity is fundamentally important, the advent of recombi-
nant methods for overexpression, coupled with affinity chroma-
tography tags, has made achieving highly pure protein samples less
problematic than in the past. In addition, there are other, perhaps

Leslie W. Tari (ed.), Structure-Based Drug Discovery, Methods in Molecular Biology, vol. 841,
DOI 10.1007/978-1-61779-520-6_2, © Springer Science+Business Media, LLC 2012

29
30 S.C. Edavettal et al.

underappreciated, protein characteristics where heterogeneity

arises. At the conformational level, loops and/or the termini of the
protein may exhibit disorder, and in larger multi-domain proteins
the individual domains may be flexible relative to one another lead-
ing to many conformational isomers of the protein in solution and
representing a barrier to crystallization due to structural heteroge-
neity. Poor crystal formation can also arise from aggregation or
changes in monodispersity representing higher order solution state
heterogeneity. Heterogeneity may also be present at the chemical
level of the protein from posttranslational modifications, such as
phosphorylation or glycosylation, which are often incomplete or
heterogeneous. Proteolysis either during purification or expression
can also introduce heterogeneity. In addition, nonenzymatic chem-
ical modifications such as oxidation can occur, creating subspecies
of closely related, difficult to distinguish, contaminants. These
sources of heterogeneity can be addressed by choosing an appro-
priate expression system, designing proper construct boundaries,
advantageous mutational changes and an effective purification
strategy, the latter of which will be addressed in Chapter 3.
Frequently, generating a truly homogeneous crystallizable protein
solution will require exploring several constructs and variables. In
this chapter we explore the considerations used in determining
construct design as well as vector–host combinations aimed at
minimizing heterogeneity and generating homogenous protein
solutions that will lead to well-diffracting protein crystals.

2. Construct
Design
Protein engineering is a powerful tool for improving protein phys-
iochemical properties leading to proteins that are more stable,
soluble, and have a higher propensity to crystallize. The concept of
the protein as a variable in protein crystallization (1) is important
and has been enabled by modern molecular biology techniques.
Securing downstream success hinges primarily on rational con-
struct design and can be undertaken independent of vector/host
choice. The overall goal in construct design is to produce large,
homogenous quantities of soluble proteins with a high likelihood
to crystallize. An important decision in this effort is the choice of
expression boundaries as all other changes take place within these
confines. The degree of difficulty in determining domain boundar-
ies depends on the degree of similarity of the protein of interest to
other proteins and, in particular, to proteins of known structure.
Side chain, loop, or termini flexibility (i.e., changes in entropy) can
also be addressed in construct design. Alterations of surface exposed
residues, both hydrophilic and hydrophobic, can lead to better
behaved protein solutions and protein crystals. Posttranslational
 2 Genetic Construct Design and Recombinant Protein… 31

modifications, both the addition and subtraction of, can have

dramatic results with respect to protein stability and good crystal
formation. Even the isolation of protein domains out of the con-
text of the larger protein can lead to better crystal formation when
inherent inter-domain flexibility hinders crystallization efforts. In
this section, we will address these considerations individually and
give the reader a better sense of how the researcher addresses, and
ultimately strives to overcome, these issues.

3. Boundaries

Boundary determination often represents the crucial choice for

producing sufficient quantities of soluble, high-quality protein for
crystallographic purposes. For small prokaryotic proteins the native
termini are often ideal. However, for more complex eukaryotic
proteins, crystallization of truncated protein or of individual
domains for multi-domain proteins is often more appropriate. It is
difficult to predict the boundaries of a domain of interest or the
proper truncation of a protein’s termini from the primary sequence
in isolation. Analysis of the similarity of the protein of interest to
other family members and in particular to proteins of known struc-
ture provides the template for proper boundary choices. The first
step is to query the sequence of interest against the Protein
Databank (http://www.pdb.org/pdb/home/home.do). This
allows a quick assessment of the closest homologs with published
structure. If reasonable hits are found, a close approximation of
proper boundaries is already at hand through design of a construct
with similar termini. Refinement of boundary choice can be com-
pleted by analyzing whether the termini used in the database struc-
tures are ordered. Multiple sequence alignments also represent an
important analytical approach, as primary sequence similarity is a
strong predictor of structural similarity. From a structural biology
viewpoint, their utility comes in the ability to visually illustrate
conserved and variable sites within a protein family that typically
correspond to structurally important and dispensable features,
respectively. Commonly used alignment tools include algorithms
such as ClustalW2, Tree-based Consistency Objective Function for
alignment Evaluation (T-Coffee), or multiple sequence compari-
son by log-expectation (MUSCLE) (2, 3).
A more sophisticated approach entails homology or compara-
tive protein structural modeling (4). This method produces an all-
atom model of a sequence based on its alignment to one or more
related protein structures. Either sequential or simultaneous mod-
eling of the core of the protein as well as loops and side chains can
greatly facilitate the identification of not only end terminal boundar-
ies, but also the identification of surface exposed side chain residues
32 S.C. Edavettal et al.

and potential flexible loops. Although it should be noted that many

loops will be readily predictable from insertion/deletion gaps in
the multiple sequence alignments. Templates for comparative
model building are often found by sequence alignment methods
such as BLAST, PSI-BLAST, FASTA or SALIGN. Once identified,
the atomic coordinates of the templates and a short script file are
fed into a computer program for comparative protein structure
modeling such as MODELLER. This program implements com-
parative protein structure modeling by the satisfaction of spatial
restraints that are input by the user. MODELLER can also per-
form a number of auxiliary tasks including the calculation of phy-
logenetic trees, alignment of two protein sequences or their profiles,
multiple alignments of protein sequences or their profiles, multiple
alignments of protein sequence and/or structures, and de novo
modeling of loops in protein structures. This method remains the
most reliable method to predict the three dimensional structure of
a protein.
Recently, Mooij et al. presented a web-based tool, ProteinCCD
(CCD: Crystallographic Construct Design) which consolidates
common tools in structural biology into a single platform that
enables comparative analysis of the sequence and allows the design
of oligonucleotides for PCR amplification of the chosen protein
constructs (5). This suite is divided into four groups of sequence
analysis tools, predicting secondary structure, disordered regions,
structural motifs, and flexible domains. Secondary structure pre-
diction uses primary sequence information to predict stretches of
sequences that are likely to be beta-strands or alpha-helices in the
three-dimensional structure and, thus, should not be disrupted.
The second group of tools employs algorithms that aim to predict
disordered regions in the protein primary sequences. Rigidifying
or deleting loop structures predicted to be flexible can improve
crystal formation. Predicting specific features of a protein sequence
including transmembrane topology, signal peptides, or regions of
coiled-coils will also aid in construct design by identifying regions
which should have structural rigidity and therefore should not be
mutated. The Simple Model Architecture Research Tool (SMART)
and the domain Linker Predictor are used to analyze domain struc-
ture and identify genetically mobile domains. The information
output displays a condensed view of all results against the protein
sequence where the researcher can analyze the data and choose,
interactively, possible construct boundaries.
These computational methods normally yield multiple possible
termini for any given protein. Even in the best-case scenario
where a crystal structure of a close homolog exists as a guide, the
choice of the precise starting or ending residues may be difficult. In
some cases, one terminus may be better defined than the other. It
is often advisable to bring forward multiple constructs to test
different hypotheses. The expression system, the throughput of
 2 Genetic Construct Design and Recombinant Protein… 33

the lab, and the ambiguity of the alignment all impact the number
of clones that should be generated and evaluated. Ideally, high-
throughput expression and purification techniques can be
employed to efficiently assess the quality of each construct. High-
throughput melting temperature analysis has become a method of
choice for ranking the propensity for crystallization of such
constructs, as thermal stability has been correlated with crystalliz-
ability (6). However, expression level very often provides a good
surrogate assessment of the behavior of the protein; better expressers
often leading to better crystallizers. The number of constructs can
usually be limited by focusing on the most aggressively truncated
candidates, in most cases less is more.
In instances of truly novel sequences where computational or
comparative methods do not provide guidance, empirical
approaches may be employed. Boundaries can be based on infor-
mation from limited proteolysis/mass spectrometry (LPMS) where
a time course of digestion with a protease such as chymotrypsin,
subtilisin, or endo-Glu-C is followed by mass spectrometry to
identify stable proteolytic fragments. This powerful method can
identify exposed flexible termini or loop structures that can be
problematic for good crystal formation. Information about cleav-
age sites gained from limited proteolysis can be translated into new
construct design to eliminate inherently flexible regions creating a
more minimalist rigid structure and increasing the likelihood of
successful structure determination.
A second empirical method, based on enhanced hydrogen/
deuterium exchange mass spectrometry can be employed when
working with novel sequences (7). This method allows one to
identify regions of disorder in a protein through the enhanced
exchange rate of backbone amide hydrogens. Slower exchange
rates would suggest more highly structured regions whereas faster
exchange rates would be indicative of domain boundaries, flexible
loops, or disordered termini that could be adjusted or deleted in
subsequent construct design. Several examples have been pub-
lished outlining the utility of this approach (8). It is now routinely
employed by both the NESG and JCSG (9, 10).

4. Choosing an
Expression System
Several expression systems are routinely used to generate recombi-
nant protein suitable for crystallization purposes, including bacte-
ria, insect cells, yeast, and mammalian cells. Initially, the choice of
expression system may be a balance of cost, ease of use, or the
complexity of the system. Since most cDNAs can be expressed in
many different systems, choosing a host is generally based on
expressed protein yields, desired posttranslational modifications,
34 S.C. Edavettal et al.

Table 1
Protein classes

Protein classa Protein size (AA) Expression system (s)

Peptides <80 E. coli (generally as fusion

proteins)
Secreted proteins 80–500 All (proven track record in
yeast and mammalian)
Large secreted proteins/ >500 Mammalian
cell surface receptors
Non-secreted proteins >80 All, based on individual
nature or protein
a
Arbitrary classification of proteins

and relevant purity. The choice of expression system begins with

arbitrarily assigning the target protein to one of four broad classes
(see Table 1) (11). The first class is small proteins and peptides that
are less than ~80 amino acids in length. These are generally best
expressed in bacteria with fusion partners that are enzymatically
removed post-purification. The second class comprises secreted
proteins that range in size from approximately 80–500 amino acids.
This class of proteins can be expressed in all expression systems but
is generally targeted for yeast, insect cell, and mammalian cell
expression. The third class is composed of secreted proteins that
are large (>500 amino acids). These are best expressed in mam-
malian cells as these cells contain the complex machinery needed
for processing and adding posttranslational modifications. Finally,
the fourth class is cytosolic proteins that are generally larger than
80 amino acids. The choice of expression system for this class of
protein depends of the nature of the protein to be expressed as will
be discussed later where we address the advantages and disadvan-
tages of each system with respect to the specific class of protein.
While membrane proteins are not described here in detail, they are
rapidly becoming more common targets for crystallization studies.
The eukaryotic hosts are most commonly used for this purpose,
although several novel techniques have been described to express
membrane proteins in bacterial systems (12, 13). Finally, desired
posttranslational modifications, or lack thereof, can influence the
choice of expression system where the most limiting system is bac-
teria (see Table 2). However, the primary driver of choice of expres-
sion system should always be probability of success.

4.1. Prokaryotic Several factors may direct one toward a prokaryotic system includ-
Expression Systems ing target proteins which are cytosolic, prokaryotic in origin, or
lacking in relative complexity, as well as the desire for no or limited
 2 Genetic Construct Design and Recombinant Protein… 35

Table 2
Common posttranslational modifications in different host systems

Posttranslational modification E. coli Insect cells Yeast Mammalian cells

Disulfide bond formation Possiblea Yes Yes Yes

Proteolytic processing Signal sequence Yes Yes Yes
removal
Phosphorylation Yes Yes Yes Yes
N-linked glycosylation No Yes Yes Yes
O-linked glycosylation No Yes Yes Yes
N-terminal methionine removal Yes Yes Yes Yes
a
Possible when expressed in host cells with thioredoxin reductase (trxB) and glutathione reductase (gor)
mutations

posttranslational modifications. For most research labs, the choice

of bacterial cell host for recombinant protein expression is Escherichia
coli. This system is often chosen for simple economic considerations,
ease of use and a large selection of vector/host combinations allow-
ing one to tackle most protein expression situations. Hosts which
promote disulfide bond formation in the cytosol and the titration of
IPTG for more uniform expression are available as well as hosts that
provide rare codon tRNAs for non-codon optimized cDNAs.
A variety of expression vectors are available offering different
antibiotic resistances, induction protocols, and fusion partners.
Popular fusion partners include poly-histidine for immobilized
metal affinity chromatography purification, thioredoxin (Trx) for
disulfide bond formation in an oxidizing cytosol, signal peptides
or disulfide bond isomerase (Dsb) variants for periplasmic expres-
sion and disulfide bond formation and glutathione S-transferase
(GST), N utilization substance A (NusA), maltose-binding protein
(MBP), or small ubiquitin-related modifier (SUMO) for soluble
cytosolic expression of proteins and peptides (see Table 3). Also
offered, or engineered directly, are many choices for fusion partner
removal by proteolytic digestion. These include enterokinase,
thrombin, factor Xa, tobacco etch virus protease (TEV), and human
rhino virus 3C protease (HRV3C). The choice of which protease to
use is often dictated by the desired, mature N-termini; however
TEV usually represents a robust choice.
Direct, cytosolic, expression in bacteria is usually the method
of choice for a heterologous protein from the first and fourth
classes of proteins as long as this target protein does not contain an
inordinate number of cysteines involved in native disulfide bonds.
The reducing environment of the bacterial cytosol does not allow
disulfide bond formation and overexpression often leads to the
36 S.C. Edavettal et al.

Table 3
Fusion tags and partner proteins

Fusion partner Placement of tag Approx. size (AA) Advantages

Poly-histidine N, C, I 6–10 Affinity purification

Trx N 110 Disulfide formation
Signal peptide N 20 Periplasmic expression, native folding
Dsb N 220 Periplasmic expression, native
folding, disulfide formation
SUMO N 100 Cytoplasmic solubility, native
N-termini
GST N 220 Cytoplasmic solubility
NusA N, I 500 Cytoplasmic solubility
MBP N 400 Cytoplasmic solubility
N N-terminus; C C-terminus; I internal

formation of insoluble inclusion bodies that require solubilization

and refolding to yield the desired product. However, as mentioned
earlier, there are several vector choices that not only facilitate the
expression of a soluble fusion construct but also the formation of
disulfide bonds when used in conjunction with a host cell carrying
the thrioredoxin reductase (trxB) and glutathione reductase (gor)
mutations that result in an oxidizing cytosol. There are several
examples in the literature where these combinations, along with
refolding chaperones, led to native disulfide bond formation
(14–16) and the mature protein was acquired following enzymatic
removal of the fusion partner.
Intracellular expression in E. coli yields a protein containing
the initiating methionine residue. Endogenous proteins are nor-
mally processed by E. coli N-terminal methionine amino peptidase
(MAP). For highly expressed recombinant proteins, this process-
ing can be rate-limiting leading to N-terminal heterogeneity.
However, the “N-end rule” is often a good guide in construct
design when looking to enhance protein expression or to opti-
mize, or minimize, N-terminal methionine processing (17, 18).
Effective methionine processing has been shown to be directly
related to the radius of gyration of the penultimate residue.
Methionine cleavage decreases proportionally to the increase of
the minimal side chain length. Smaller residues such as Gly, Ala,
Ser, or Cys result in more efficient processing, intermediate resi-
dues such as Thr, Pro, Val, Gln, or Glu result in less efficient pro-
cessing and all other residues in the penultimate position result in
little to no methionine processing.
 2 Genetic Construct Design and Recombinant Protein… 37

4.2. Eukaryotic Another common approach to recombinant protein expression is

Expression Systems baculovirus-mediated expression (BEVS) using insect cells. This
system has proven itself for the rapid production of high levels of
4.2.1. Insect Cells
heterologous proteins. Insect cells are slower growing than both
bacteria and yeast, and the cost of the media is relatively high com-
pared to those two systems. However, recent advances with inter-
mediate-scale suspension systems (e.g., WAVE bioreactors) and the
newer baculovirus-mediated vector systems (e.g., FlashBAC
(NextGen Sciences) and BacMagic (Novagen)) have transformed
BEVS into the eukaryotic expression system of choice for struc-
tural biology. The most commonly used virus is Autographica cali-
fornica nuclear polyhedrosis virus (AcMNPV), which is preferred
due to the diversity in commercially available transfer plasmids.
Several insect cell lines for baculovirus expression are commonly
employed including Spodoptera frugiperda (fall armyworm) lines
Sf 9 and Sf 21. These cells double in ~20 h in both monolayer and
suspension cultures. Another frequent option that supports
AcMNPV replication is the Trichoplusia ni-derived BTI-Tn-5B1-4
(“High5”) cell line. These cells tend to be better for secreted pro-
teins but they do not grow as well in suspension.
Both intracellular and secreted expression is possible with
insect cells. The parameters for optimizing heterologous expres-
sion have been well studied and include the kinetics of infection
(multiplicity of infection (MOI), cell density, and time) (19). This
expression system is categorized as transient expression and is capa-
ble of providing milligram quantities of heterologous protein with
complex posttranslational modifications. These modifications
include correct proteolytic processing of signal peptides and inter-
nal cleavage sites, phosphorylation, N-terminal blocking, proper
folding and assembly, and glycosylation. The higher expression lev-
els and p10 and polh promoter activity in the very late phase of
infection can lead to incomplete posttranslational modifications
resulting in higher levels of protein heterogeneity. In addition,
BEVS has emerged as a useful host for membrane protein expres-
sion (20), especially for GPCRs with all currently published struc-
tures derived from BEVS expressed GPCRs.

4.2.2. Yeast For the past 25 years, Pichia pastoris has increased in popularity for
the production of heterologous proteins. P. pastoris is a methyl-
tropic yeast that uses the tightly regulated alcohol oxidase (AOX)
promoter enabling high levels of protein expression by the addi-
tion of methanol. Early biochemical studies demonstrated that
methanol utilization requires a metabolic pathway involving sev-
eral unique enzymes. The key enzyme, AOX catalyzes the first step
in the methanol utilization pathway. Two genes, AOX1 and AOX2,
encode AOX in P. pastoris where AOX1 is responsible for the
majority of AOX activity in the cell. The presence of methanol
is essential for the induction of high levels of protein expression.
38 S.C. Edavettal et al.

An attractive feature of this system is the simplicity of techniques

for genetic manipulation. P. pastoris has also been shown to express
foreign proteins at high levels, both intracellularly and extracellu-
larly. This system is also capable of performing many common
eukaryotic posttranslational modifications such as disulfide bond
formation, proteolytic processing, and glycosylation (see Table 2).
However, since P. pastoris has no native plasmids, the expression
cassette must be integrated into the chromosome and the selection
and optimization of the integrated gene can be a time consuming
process. An alternative strategy to employing gene integration in
Pichia is to exploit episomal expression in Saccharomyces cerevisiae
(21). This expression technique can provide more rapid evaluation
of the feasibility of yeast expression for a given protein.
Compared to a bacterial system, yeast generally produces pro-
teins at lower levels. Even so, yeast has proven its ability to produce
biologically active and, hence, biologically relevant proteins in ever
increasing cases. Yeast cell expression is a system that provides for
both intracellular expression as well as extracellular expression.
However, an extracellular expression approach remains the path of
choice, as yeast cells grown to maximum stationary phase possess
cell walls that are notoriously difficult to remove. Expression in a
yeast cell system has some very attractive features including rapid
cell growth, low cost of media reagents, high expression levels and
the generation of proteins with key posttranslational modifications.
Yeast has been shown to be a suitable host for the secreted expres-
sion of complex proteins like single chain antibody fragments, ser-
ine proteases such as DESC1 (22), matriptase (MTSP1) (23) or
urokinase plasminogen activator (24) and various growth factors
whereas human superoxide dismutase (25), fibroblast growth fac-
tors (26) and α-antitrypsin (27) have been successfully expressed
in the cytosol.

4.2.3. Mammalian Cells Complex proteins are best expressed in mammalian cell culture
where processing through the secretion pathway yields predomi-
nantly biologically active protein. Mammalian cell culture is still
the system of choice for complex, multi-domain proteins. Although,
not unlike insect cell culture, there are distinct drawbacks. Like
insect cells, mammalian cells grow at much slower rates than micro-
bial cells where doubling times are in the rage of 18–24 h. In addi-
tion there are several other considerations that must be evaluated
before employing a mammalian cell expression approach, such as
substantial time and cost investment, lower overall expression lev-
els, and culture sensitivity issues such as shearing, changes in tem-
perature, pH and oxygen levels and metabolites. The accumulated
benefits, including secretion into the media, proper folding, stabil-
ity and comprehensive posttranslational modifications support this
choice of expression system when producing proteins from the
 2 Genetic Construct Design and Recombinant Protein… 39

third protein class. Efficient protein expression from mammalian

systems can be approached in several ways.
Lower productivity has been overcome through the use of
stirred bioreactors or Wave systems for suspension cultures.
Suspension cultures, in combination with fed-batch approaches
allow for very high cell densities and overall increases in protein
production. Many commercially available vectors contain strong
promoters, such as simian vacuolating virus 40 (SV40) or cyto-
megalovirus (CMV) immediate early promoter, in addition to
dominant and recessive selection markers such as neomycin or
dihydrofolate reductase (DHFR), respectively. Homologous
recombination of the expression cassette into regions of high tran-
scription has been shown to increase the expression levels of recom-
binant proteins. However, for most glycoproteins, or proteins with
extensive posttranslational modifications, the rate-limiting step in
protein production is these posttranslational modifications.

5. Strategies
to Facilitate
Protein
Crystallization Attrition rates in the production of high quality protein crystals
remain high when taking into account the entire process from the
generation of high quality protein to the determination of a high
resolution crystal structure. One of the key prerequisites to gener-
ating high quality protein crystals is effectively addressing disorder
and protein instability. Disorder can manifest itself in the form of
surface residue dynamics where a high degree of side chain confor-
mational entropy, inter-domain flexibility, or unrestrained termini
make the formation of high quality crystal contacts prohibitively
difficult. Disorder also appears in the form of heterogenic post-
translational modifications such as glycosylation or phosphoryla-
tion. Screening of surface mutations to address hydrophobicity,
addressing solvent accessible cysteine residues and even abrogating
regions of high conformational flexibility can improve protein sta-
bility. In this section, we will address strategies to reduce surface
entropy, improve solubility and increase the overall propensity to
form high quality crystals capable of diffracting to high resolution.

5.1. Identifying The apparent driving force of the universe is the disposition to
Regions of Disorder move toward lower enthalpy (H) and higher entropy (S) according
to the Gibbs free energy equation. The actual driving force is the
tendency to move toward greater entropy. Crystallization is a
unique process that occurs at a high entropic cost. The confor-
mational entropy of surface residues and loops, which become
ordered in the crystallization process, are the main contributors
to this entropic loss. This would suggest that rationally engi-
neering out regions of amino acids with higher conformational
40 S.C. Edavettal et al.

entropy, as well as loops with lower flexibility, could lead to more

thermodynamically favored crystal formation.
An early example of using surface protein engineering to
enhance the propensity for crystal formation was work done by
Lawson et al. where surface mutations of human H ferritin pro-
moted crystal formation (28). This group used information taken
from the crystal structures of other homologues suggesting that
the incorporation of a single L86Q substitution was sufficient to
create favorable crystal contacts and led to crystals diffracting to
1.9 Å or better and subsequent three-dimensional structure deter-
mination. D’Arcy et al. later demonstrated that the surface muta-
genesis of DNA gyrase B led to the formation of new crystal forms
that diffracted to higher resolution and concluded that single point
mutations can have a marked effect on the crystallization proper-
ties of proteins resulting in both the number of crystal hits as well
as improvements in crystal quality (29).
Recently, Derewenda has addressed the idea of reducing sur-
face entropy by creating “low-entropy” surface patches through
rational site directed mutagenesis (30). The idea that most pro-
teins have evolved a “surface-entropy shield” that is comprised
predominantly of lysine and glutamate residues was proposed by
Doye et al. (31). The hypothesis is that this prevents nonspecific
aggregation and precipitation. The author points out that the
spontaneous crystallization of proteins in vivo is responsible for
several serious diseases. Thus, an effective entropic-shield makes
these interactions less than favorable. Several publications by
Derewenda and colleagues have described this concept in more
detail (32–34). This approach suggests that mutating residues with
high conformational mobility, lysine or glutamate, to smaller amino
acids, such as alanine or glycine, should be effective in reducing
overall surface conformational entropy. The fact that, together,
68% of these residues are found completely exposed while only 6%
are found buried supports the idea that these residues play a large
part in creating this entropic shield (35). These authors tested this
concept through a series of mutational experiments using the glob-
ular domain of the human regulatory protein RhoGDI. Although
this molecule is relatively small, it has a surface rich in lysine and
glutamate residues. The majority of these experiments involved
mutating these residues to alanine with dramatic results. Most of
the mutations resulted in proteins with a higher propensity to crys-
tallize. As further validation, in most cases the crystal contacts were
mediated by the mutated epitopes. Different sets of mutations also
led to novel crystal forms that exhibited superior diffraction. Taken
together, the authors suggest that it is the nature of the crystal
contacts that is the primary determinant in the quality of the crys-
tal formation. It is important to note however, that mutating polar
amino acids like lysine and glutamate to hydrophobic amino acids
like alanine can occasionally destabilize the target protein (36).
 2 Genetic Construct Design and Recombinant Protein… 41

In aqueous solution, the surface exposure of large hydrophobic

regions is energetically unfavorable and can promote aggregation,
especially at high protein concentrations typical of crystallization
experiments. It has been proposed, however, that increased side
chain entropy is more detrimental to crystal lattice formation than
decreased hydrophobicity (37). Therefore, a balance must be
struck between decreasing surface entropy enough to promote
crystallization and maintaining sufficient surface hydrophilicity to
deter aggregation when engineering surface mutations.
Limited proteolysis, as mentioned previously, is often used to
identify flexible regions at the terminal ends of the protein of inter-
est. Limited proteolysis is also used to identify flexible loops in
proteins that could have an adverse effect on crystal packing. This
method often results in a more rigid form of the molecule that has
a higher propensity for crystal formation. Using information from
limited proteolysis, constructs can be designed where offending
residues are either removed or substituted with residues that can
have a stabilizing effect. Rosenbaum et al. identified a poorly struc-
tured intracellular loop in β2-adrenergic receptor that contributed
to the relatively unrestricted movement of two transmembrane
helices (38). The conformational heterogeneity resulting from the
free movement led to crystallization problems. The authors were
able to replace the flexible loop with the well-folded protein, T4
lysozyme. The rational engineering effort added a polar surface
while restricting the movement of the transmembrane helices. This
chimeric protein was efficiently expressed, retained binding affini-
ties and resulted in crystals diffracting to 2.4 Å.

5.2. Reducing Heterogeneity and aggregation are frequently the main impedi-
Heterogeneity ments to the generation of good quality protein crystals. It is critical
therefore to ensure that the initial crystallization sample is homoge-
neous and monodisperse. Heterogeneity can arise from several
sources, including contaminating proteins, aggregates, and post-
translational modifications. Sample purity, normally assessed by
SDS-PAGE, is necessary to ensure the sample is homogenous with
minimal contaminating proteins present (normal good practice is
>95% purity by SDS-PAGE). Aggregation, a frequent source of
sample heterogeneity, is often detected by SEC (size exclusion
chromatography) during purification. However, given that samples
are typically concentrated after SEC, and prior to crystallization,
this method is not an accurate measure of aggregation in the pro-
tein sample as used for crystal screening. Dynamic light scattering
(DLS) instruments are now available with both high-sensitivity
and plate-readers for higher throughput and are becoming a more
common technique to assess aggregation. DLS is well suited to
determine protein aggregation as it is compatible with both the
concentrations and buffer conditions used in crystallization
(reviewed in ref. (39)). Posttranslational modifications can provide
Another Random Scribd Document
with Unrelated Content
 TO SIR, WITH LOVE.

Branch, William.

STILL A BROTHER, INSIDE THE NEGRO

MIDDLE CLASS.

Brandel, Marc.

DOUBLE TROUBLE.

Brauer Productions.

THE LAST MAN.

Braun Engineering Co.

THE 1-3/8" SLUG HEADER.

OPERATION OF THE SINGLE-ARM DIE.

Brauner (Artur)-CCC Film Productions,

G.M.B.H.

FANNY HILL: MEMOIRS OF A WOMAN OF

PLEASURE.

Brazier, Gary P.

BUILDING POLITICAL LEADERSHIP.

Breffort, Alexandre.

IRMA LA DOUCE.

Brennan Productions, Inc.

 BARB WIRE.

SHOOT OUT AT BIG SAG.

Brennan-Westgate Productions.

THE REAL MCCOYS.

Brenner (Joseph) Associates.

CARESSED.

HIGH.

THE SISTERS.

Brentwood Productions, Inc.

CAPTAIN NEWMAN, M.D.

TO KILL A MOCKINGBIRD.

Bresler (Jerry) Productions, Inc.

DIAMOND HEAD.

GIDGET GOES HAWAIIAN.

GIDGET GOES TO ROME.

LOVE HAS MANY FACES.

MAJOR DUNDEE.

Breslin, Howard.
 PLATINUM HIGH SCHOOL.

Brickhill, Paul.

THE GREAT ESCAPE.

Bricusse, Leslie.

STOP THE WORLD, I WANT TO GET OFF.

Brien Productions, Inc.

DUEL AT DIABLO.

Briggs, Lee.

PARISH LIFE OF THE FALLS CHURCH.

Brigham Young University. SEE Young

(Brigham) University, Provo, Utah.

Brigham Young University, Motion Picture

Dept. SEE Young (Brigham) University,
Provo, Utah. Motion Picture
Dept.

Brighton Pictures, Inc.

SAM WHISKEY.

Brinter-Brasil Internacional Films.

MACUMBA LOVE.

Briskin (Fred) Productions, Inc.

 WORLD CHAMPIONSHIP GOLF.

Briskin Productions, Inc.

AWARD THEATRE.

THE DONNA REED SHOW.

Brisson (Frederick) Production.

FIVE FINGER EXERCISE.

Bristol Pictures, Inc.

THE GLORY GUYS.

Bristol Productions, Inc.

FRANCIS ALBERT SINATRA DOES HIS THING.

FRANK SINATRA: A MAN AND HIS MUSIC.

A MAN AND HIS MUSIC + ELLA + JOBIM.

Bristol-Myers Co.

BUCKY BEAVER, ARTIST.

BUCKY BEAVER, DETECTIVE.

BUCKY BEAVER, SHERIFF.

BUCKY BEAVER, STAGECOACH DRIVER.

Britannia Film Distributors.

 JET STORM.

Britannia Film Distributors, Ltd.

FOXHOLE IN CAIRO.

NEARLY A NASTY ACCIDENT.

British Broadcasting Corp.

VICTORY IN EUROPE, 20 YEARS AFTER.

British Lion Films.

JET STORM.

British Lion Films, Ltd.

FOXHOLE IN CAIRO.

RING OF TREASON.

British Travel Assn.

SONG OF LONDON.

WONDERFUL SCOTLAND.

Broadcasting & Film Commission.

OPERATION DEVELOPMENT.

Broadcasting & Film Commission. SEE

Broadcasting Film Commission.

Broadcasting Film Commission.

THE NEW AGE IN JAPAN.

Broadman Films.

ANSWERING OBJECTIONS IN WITNESSING,

No. 1.

ANSWERING OBJECTIONS IN WITNESSING,

No. 2.

BITTER FRUIT.

THE COMMUNIST THREAT.

CONCEPT OF GOD.

CONCEPT OF LIFE.

CONCEPT OF MAN.

THE CONSTANT WITNESS.

DANIEL.

THE DOUBLE GUILT.

THE GREAT CHALLENGE.

HOW TO WITNESS.

IRON HANDS.

LET'S HAVE A PARTY.

 LIFELINE TO THE WORLD.

MAGNIFICENT HERITAGE.

MOSES AND THE MOUNTAIN OF FIRE.

MY WILL BE DONE.

NEHEMIAH.

PROPHET FROM TEKOA.

RECLAIMING THE SAVED.

ROAD TO EN-DOR.

TAKE A GIANT STEP.

TEEN QUEEN.

THE WAITING WORLD.

WHAT DIRECTION?

WHAT FIRST?

WHAT'S IMPORTANT?

WHAT'S LEFT?

Broekman (Robert) Productions.

EAT TO YOUR HEART'S CONTENT.

Broidy (William F.) Productions, Inc.

 WILD BILL HICKOK.

Bronston (Samuel) Production.

JULIUS CAESAR: THE RISE OF THE ROMAN

EMPIRE.

Bronston (Samuel) Productions, Inc.

EL CID.

55 DAYS AT PEKING.

KING OF KINGS.

Bronston-Midway Productions.

CIRCUS WORLD.

Bronston-Roma Productions.

THE FALL OF THE ROMAN EMPIRE.

Bronze Division, James H. Matthews

& Co. SEE Matthews (James H.) &
Co. Bronze Division.

Brooklyn College Television Center.

THE MANAGEMENT IN PURCHASING MANAGEMENT.

SYSTEMS CONTRACTING.

Brooks, Walter.
 MISTER ED.

Brophy, John.

THE DAY THEY ROBBED THE BANK OF

ENGLAND.

Brother (D. P.) & Co.

THE INSIDE STORY ON OLDSMOBILE AIR

CONDITIONING.

JETFIRE.

Brotherhood Co.

THE BROTHERHOOD.

Broussard, Elsie R.

MOTHERS AND BABIES.

Brower, Lincoln Pierson.

PATTERNS FOR SURVIVAL.

Brown, George Emerson, Jr.

STOLEN HOURS.

Brown, George H.

THE BOY WHO STOLE A MILLION.

Brown, Harry.
 EL DORADO.

Brown, Harry Joe, Jr.

DUFFY.

Brown, Helen Gurley.

SEX AND THE SINGLE GIRL.

Brown, James.

CHOOSING A CLASSROOM FILM.

CREATING INSTRUCTIONAL MATERIALS.

HOW TO USE CLASSROOM FILMS.

SELECTING AND USING READY-MADE

MATERIALS.

Brown, Peter S. SEE Brown (Peter S.)

Productions.

Brown (Peter S.) Productions.

THE WILL TO WIN.

Brown, V. W. SEE Brown, Victor W.

Brown, Victor W.

LA REATA.

Browning, Ricou.
 FLIPPER.

Bruce, Jean.

THE VISCOUNT.

Brunswick. Bowling Division.

AUTOMATIC LANE MAINTENANCE MACHINE.

Brunswick Corp.

GOLDEN OPPORTUNITY.

THE NEW MODEL A-2 PINSETTER.

Bruton Film Productions, Ltd.

HAVING A WILD WEEKEND.

Bryna-Quine Productions.

STRANGERS WHEN WE MEET.

Buchan, John.

THE 39 STEPS.

Buchman, Chris, Jr.

DANTINI THE MAGNIFICENT.

FROM BUCHANAN HOUSE.

Buchsbaum, Ralph.
 NATURAL SELECTION.

Buchwald, Art.

SURPRISE PACKAGE.

Buck, Pearl S.

THE GUIDE.

SATAN NEVER SLEEPS.

Buck, Pearl Sydenstricker.

THE BIG WAVE.

Buckner, Robert.

MOON PILOT.

Bucky, Peter A.

EINSTEIN, THE MAN.

Bucyrus-Erie Co.

SAFETEAM CAMPAIGN.

Budd Co.

A NEW CONCEPT IN TRANSIT TRUCKS.

Budweiser.

HAWAII—BIG GAME COUNTRY.

Buehler, Ltd.

COARSE GRINDING.

FINE GRINDING.

MODERN METALLOGRAPHY, APPARATUS,

METHODS.

MOUNTING.

PETROGRAPHY-CERAMOGRAPHY SAMPLE
PREPARATION.

ROUGH AND FINAL POLISHING.

SECTIONING.

Buell, Marjorie H.

ALVIN'S SOLO FLIGHT.

Buell, Marjorie Henderson.

FROG LEGS.

Buena Vista Distribution Co.

CHARLIE THE LONESOME COUGAR.

MARY POPPINS.

WINNIE THE POOH AND THE HONEY TREE.

Buena Vista Distribution Co., Inc.

THE ABSENT-MINDED PROFESSOR.

THE ADVENTURES OF BULLWHIP GRIFFIN.

ALMOST ANGELS.

AQUAMANIA.

BABES IN TOYLAND.

BIG RED.

BLACKBEARD'S GHOST.

BON VOYAGE.

THE COMPUTER WORE TENNIS SHOES.

A COUNTRY COYOTE GOES HOLLYWOOD.

THE DANUBE.

DONALD AND THE WHEEL.

DONALD'S FIRE SURVIVAL PLAN.

EMIL AND THE DETECTIVES.

THE FIGHTING PRINCE OF DONEGAL.

FOLLOW ME, BOYS!

FREEWAYPHOBIA, No. 1.

GALA DAY AT DISNEYLAND.

THE GNOME-MOBILE.

GOLIATH II.

GOOFY'S FREEWAY TROUBLES.

GREYFRIARS BOBBY.

HANG YOUR HAT ON THE WIND.

THE HAPPIEST MILLIONAIRE.

THE HORSE IN THE GRAY FLANNEL SUIT.

THE HORSE WITH THE FLYING TAIL.

THE HOUND THAT THOUGHT HE WAS A

RACCOON.

THE INCREDIBLE JOURNEY.

ISLANDS OF THE SEA.

IT'S TOUGH TO BE A BIRD.

JAPAN.

THE JUNGLE BOOK.

JUNGLE CAT.

KIDNAPPED.

THE LEGEND OF LOBO.

THE LEGEND OF THE BOY AND THE EAGLE.

LT. ROBIN CRUSOE, U. S. N.

THE LITTERBUG.

THE LOVE BUG.

THE MIRACLE OF THE WHITE STALLIONS.

THE MISADVENTURES OF MERLIN JONES.

MONKEYS, GO HOME!

THE MONKEY'S UNCLE.

MOON PILOT.

THE MOON-SPINNERS.

MYSTERIES OF THE DEEP.

NEVER A DULL MOMENT.

NIKKI, WILD DOG OF THE NORTH.

THE ONE AND ONLY, GENUINE, ORIGINAL

FAMILY BAND.

ONE HUNDRED AND ONE DALMATIANS.

THE PARENT TRAP.

POLLYANNA.

RASCAL.
RUN, APPALOOSA, RUN.

THE SAGA OF WINDWAGON SMITH.

SAVAGE SAM.

SCROOGE MCDUCK AND MONEY.

THE SEARCH OF THE CASTAWAYS.

THE SIGN OF ZORRO.

SMITH.

SON OF FLUBBER.

SUMMER MAGIC.

SWISS FAMILY ROBINSON.

THE SWORD IN THE STONE.

SYMPOSIUM ON POPULAR SONGS.

THE TATTOOED POLICE HORSE.

TEN WHO DARED.

THAT DARN CAT.

THOSE CALLOWAYS.

THE THREE LIVES OF THOMASINA.

A TIGER WALKS.
 TOBY TYLER.

THE UGLY DACHSHUND.

WINNIE THE POOH AND THE BLUSTERY DAY.

YELLOWSTONE CUBS.

Buford, Gordon.

THE LOVE BUG.

Burden, Shirley.

I WONDER WHY.

Burdick, Eugene.

FAIL-SAFE.

THE UGLY AMERICAN.

Bureau of National Affairs. BNA Films.

SEE BNA Films.

Bureau of National Affairs, Inc.

THE CASE OF THE MISSING MAGNETS.

THE CHALLENGE OF LEADERSHIP.

INSTRUCTIONS OR OBSTRUCTIONS.

LISTEN, PLEASE.
Bureau of Old-Age & Survivors Insurance.

SEE U.S. Bureau of Old-Age & Survivors

Insurance.

Burgess Pub. Co.

ACTION OF ANTIBIOTICS ON BACTERIA.

ANTHOCYANIN.

ASEPTIC TRANSFER OF BACTERIAL CULTURES.

BACTERIAL EXTRACELLULAR ENZYMES.

BUDS I: LEAF BUCKEYE.

BUDS II: FLOWER, PEACH, ELM, MAPLE.

GERMINATION I: CORN.

GERMINATION II: BEAN.

GERMINATION III: PEA.

GERMINATION IV: CASTOR BEAN.

HOW TO MAKE SLIDES.

HOW TO USE THE MICROSCOPE.

MEDICAL ASEPSIS PUTTING ON A PREVIOUSLY

WORN GOWN.

PIGMENTS: SUMMER RED LEAF.

 PIGMENTS I: GREEN LEAF.

PIGMENTS II: YELLOW LEAF.

PREPARATION OF FOLEY CATHETER TRAY.

PREPARATION OF NUTRIENT BROTH.

PREPARATION OF SMEAR OF BACTERIAL

CELLS ON MICROSCOPE SLIDE.

PREPARATION OF STAINED WET MOUNT FROM

THROAT SWAB.

PRINCIPLES OF INFANT FEEDING.

SKIN PREPARATION FOR DELIVERY.

STERILE GLOVE TECHNIQUE.

STREAKING AN AGAR PLATE.

TRANSFER OF STOCK CULTURE ON AGAR

SLANT.

Burke, Andrew.

REJOICE.

Burke (Billie) Production.

ANTEPARTUM PROBLEMS.

Burke-Hockman-Swain Productions.
 STARK FEAR.

Burlingame Productions, Inc.

ANGEL.

Burlington Industries, Inc.

IPHANSIS.

Burnett, W. R.

CAIRO.

Burnford (Paul) Film Productions.

DISCOVERING COMPOSITION IN ART.

DISCOVERING CREATIVE PATTERN.

DISCOVERING DARK AND LIGHT.

DISCOVERING HARMONY IN ART.

DISCOVERING IDEAS FOR ART.

FLIGHT OF BIRDS.

INSECTS THAT HELP US.

MONKEYS AND APES: AN INTRODUCTION TO

THE PRIMATES.

Burnford, Sheila.

THE INCREDIBLE JOURNEY.

Burngood, Inc.

CAROL AND COMPANY.

THE CAROL BURNETT SHOW.

CAROL + 2.

Burns, Bart.

WALT DISNEY'S WONDERFUL WORLD OF COLOR.

Burns, Ellen.

ADMINISTRATION OF AN INTRAMUSCULAR
INJECTION.

BACK RUB.

BED BATH.

CARE OF DENTURES.

DROP FOOT: SOME CAUSES, PREVENTION AND

CARE.

HOW TO MAKE AN OCCUPIED HOSPITAL BED.

MANIPULATING HOSPITAL BED LINEN.

POSITIONING A PATIENT: PREVENTION OF

EXTERNAL ROTATION USING A TROCHANTER
ROLL.

PREPARATION OF INJECTION FROM A TABLET.

 PREPARATION OF INJECTION FROM A VIAL.

PREPARATION OF INJECTION FROM AN AMPULE.

SELECTION OF SITE FOR INTRAMUSCULAR

INJECTION: DELTOID.

SELECTION OF SITE FOR INTRAMUSCULAR

INJECTION: DORSOGLUTEAL.

SELECTION OF SITE FOR INTRAMUSCULAR

INJECTION: LATERAL THIGH.

SELECTION OF SITE FOR INTRAMUSCULAR

INJECTION: VENTROGLUTEAL.

SUBCUTANEOUS INJECTION: SITE SELECTION

AND ADMINISTRATION.

Burns, William A.

HORSES AND THEIR ANCESTORS.

MAN AND HIS TOOLS.

A WORLD FULL OF HOMES.

Burroughs, Edgar Rice.

TARZAN AND THE GREAT RIVER.

TARZAN AND THE JUNGLE BOY.

TARZAN AND THE VALLEY OF GOLD.

 TARZAN GOES TO INDIA.

TARZAN THE MAGNIFICENT.

TARZAN'S THREE CHALLENGES.

Burrows, Abe.

CAN-CAN.

Burstyn (Joseph) Film Enterprises, Inc.

SKYSCRAPER.

Burstyn (Joseph) Releasing Corp.

EUROPEAN NIGHTS.

Bush-Fekete, Leslie.

PEPE.

Busher, Frederic H.

ONE HUNDRED YEARS AND FOREVER.

Business Equipment Manufacturers Assn.

THE GREEN LIGHT.

Butler, Jim.

THE SPIRIT OF INVENTION.

Byco.
 CHANNING.

Byrne, Patrick. SEE National Institute

of Modeling, Oakland, Calif.

Byrna Productions, Inc.

SPARTACUS.

Byrna Productions, S.A.

THE LAST SUNSET.

Byrnes, David L.

SUBURBAN AIRPORT, IDEAL SITE FOR

INDUSTRY.

Byway Productions.

SKATERDATER.

Bywin Productions.

BID-IT, THE TV AUCTION SHOW.

 C
C-B Educational Films, Inc.

BEGINNING SPANISH.

KEYS TO READING.

PATHWAYS TO READING.

C. B. Films, S.A.-Espana.

RETURN OF THE SEVEN.

CBS Enterprises, Inc. Terrytoons. SEE

Terrytoons.

CBS Films.

ANGEL.

CBS Films, Inc.

THE BROTHERS BRANNAGAN.

LARIAT SAM.

THE ROBERT HERRIDGE THEATER.

WHIRLYBIRDS.

CBS Films, Inc. Terrytoons. SEE

 Terrytoons.

CBS News.

ELECTION PREVIEW.

THE FOUR NAVY DESERTERS.

SANTO DOMINGO, WHY ARE WE THERE?

CBS-TV Film Sales, Inc.

THE BROTHERS BRANNAGAN.

CBS Television Network.

BAILEY'S OF BALBOA.

THE BEVERLY HILLBILLIES.

CAROL AND COMPANY.

THE CAROL BURNETT SHOW.

CAROL + 2.

CONQUEST.

DAKTARI.

THE DEFENDERS.

DENNIS THE MENACE.

EAST SIDE/WEST SIDE.

FOR THE PEOPLE.

GARRY MOORE SHOW.

GENTLE BEN.

GILLIGAN'S ISLAND.

GUNSLINGER.

THE JOEY BISHOP SHOW.

THE JONATHAN WINTERS SHOW.

THE JUDY GARLAND SHOW.

MY FAVORITE MARTIAN.

MY LIVING DOLL.

PERRY MASON.

PETTICOAT JUNCTION.

THE RED SKELTON HOUR.

THE RED SKELTON SHOW.

RICHARD DIAMOND, PRIVATE DETECTIVE.

RUN, BUDDY, RUN.

THE SID CAESAR, IMOGENE COCA, CARL

REINER, HOWARD MORRIS SPECIAL.

 WANTED: DEAD OR ALIVE.

WILD, WILD WEST.

CBS Television Network. Operations

Dept.

LIGHTING FOR TELEVISION.

CBS Television System Network. SEE

Columbia Broadcasting System, Inc.

CCC Film—Arthur Brauner.

JOURNEY TO THE LOST CITY.

CCC-Film Kunst G.M.B.H. & Co., K.G.

THE BOY CRIED MURDER.

CCC Production.

GENGHIS KHAN.

CDA.

THE FAT BLACK PUSSY CAT.

C. E. Productions, Inc.

RAYMIE.

CGA, Inc. SEE Guess (Cameron) &

Associates, Inc.

CIPRA.
 RIFIFI IN TOKYO.

CMW Productions.

GUNSLINGER.

COM Production.

LOST BATTALION.

C. P., Inc.

CISCO KID.

CR Enterprises, Ltd.

NUREMBERG.

C. R. Enterprises, Ltd. Rebfilms,

Ltd. SEE Rebfilms, Ltd.

CRT Productions, Inc.

THE HOSPITAL WITHOUT WALLS.

Caam Co.

RINGS AROUND THE WORLD.

Cable Springs, Ltd.

PEACE FOR A GUNFIGHTER.

Cacoyannis, Michael.
 ELECTRA.

Cacoyannis (Michael) Productions, Ltd.

THE DAY THE FISH CAME OUT.

Cagney-Montgomery Productions, Inc.

GALLANT HOURS.

Cahill (Charles) & Associates, Inc.

ALCO BEAT.

AMERICAN HISTORY: BIRTH OF A NATION,

1760-1790.

CITY DRIVING TACTICS.

DAIRY—FARM TO DOOR.

GOING PLACES.

HEALTH: YOU AND YOUR HELPERS.

LIGHT—ON THE SUBJECT OF LIGHT.

THE MAGNETISM OF MAGNETS.

THE MOON: ADVENTURE IN SPACE.

RED LIGHT RETURN.

SAFELY WALK TO SCHOOL.

SEA FEVER.
 SIMPLE MACHINES AT SEA, AN INTRODUCTION
TO THE SUBJECT.

THE STUDY SERIES.

TRUCK FARM TO STORE.

Cahill, Charles H.

AMERICAN HISTORY: BIRTH OF A NATION,

1760-1790.

BROKEN BUS.

SAFELY WALK TO SCHOOL.

SAFETY RULES FOR SCHOOL.

Cahill (Charles H.) & Associates, Inc.

BAKERY BEAT.

BROKEN BUS.

ECONOMICS—IT'S ELEMENTARY.

NARCOTICS—WHY NOT?

SAFETY RULES FOR SCHOOL.

WEATHER: AIR IN ACTION.

Caillou, Alan.

RAMPAGE.
Cain, Sugar, pseud. SEE McCain,
Constance.

Caldwell, Erskine.

CLAUDELLE INGLISH.

Caldwell, Lynton K.

BUILDING POLITICAL LEADERSHIP.

California. Resources Agency.

Department of Beaches and Parks.

THE NICKEL FERRY.

California. University. Academic

Communication Facility.

THE CHILD AMPUTEE FROM INFANCY TO

SCHOOL AGE: THE BELOW-ELBOW AMPUTEE.

EARLY DEVELOPMENT OF AMBULATION: THE

UNILATERAL BELOW-KNEE CHILD
AMPUTEE.

California. University. Dept. of

Anthropology.

BUCKEYE.

THE CALUMET: PIPE OF PEACE.

DREAM DANCES OF THE KASHIA POMO: THE

BOLE-MARU RELIGION WOMEN'S DANCES.
 KASHIA MEN'S DANCES: SOUTHWESTERN
POMO INDIANS.

OBSIDIAN POINT MAKING: TOLOWA

INDIANS.

PINE NUTS.

THE SINEW-BACKED BOW AND ITS ARROWS.

California. University. Dept. of

Chemistry.

NUCLEAR MAGNETIC RESONANCE.

California. University. Regents.

ACID-BASE INDICATORS.

ACORNS.

BASKETRY OF THE POMO.

BIOCHEMISTRY AND MOLECULAR STRUCTURE.

BROMINE, ELEMENT FROM THE SEA.

BUCKEYE.

THE CALUMET: PIPE OF PEACE.

CATALYSIS.

A CHANCE TO WONDER WHY.

CHEM STUDY: INFORMATION FOR EDUCATORS.

CHEM STUDY TEACHER TRAINING PROGRAM.

CHEMICAL BONDING.

CHEMICAL FAMILIES.

THE CHILD AMPUTEE FROM INFANCY TO

SCHOOL AGE: THE BELOW-ELBOW AMPUTEE.

CRYSTALS AND THEIR STRUCTURES.

DREAM DANCES OF THE KASHIA POMO: THE

BOLE-MARU RELIGION WOMEN'S DANCES.

EARLY DEVELOPMENT OF AMBULATION: THE

UNILATERAL BELOW-KNEE CHILD AMPUTEE.

ELECTRIC INTERACTIONS IN CHEMISTRY.

ELECTROCHEMICAL CELLS.

EQUILIBRIUM.

THE GAME OF STAVES.

GAS PRESSURE AND MOLECULAR COLLISIONS.

GASES AND HOW THEY COMBINE.

GURKHA COUNTRY.

HIGH TEMPERATURE RESEARCH.

HIMALAYAN FARMER.
Welcome to our website – the perfect destination for book lovers and
knowledge seekers. We believe that every book holds a new world,
offering opportunities for learning, discovery, and personal growth.
That’s why we are dedicated to bringing you a diverse collection of
books, ranging from classic literature and specialized publications to
self-development guides and children's books.

More than just a book-buying platform, we strive to be a bridge

connecting you with timeless cultural and intellectual values. With an
elegant, user-friendly interface and a smart search system, you can
quickly find the books that best suit your interests. Additionally,
our special promotions and home delivery services help you save time
and fully enjoy the joy of reading.

Join us on a journey of knowledge exploration, passion nurturing, and

personal growth every day!

ebookbell.com