Berkeley Structural Genomics Center (Lawrence
Berkeley National Laboratory)
PI: Sung-Hou Kim [9/2004-Present]
Midwest Center of Structural Genomics
(Argonne National Laboratory)
PI: Andrzej Joachimiak [5/2004-8/2004]
1.6 Å Crystal Structure of YteR Protein from Bacillus subtilis, a Predicted Lyase (2004) - R. Zhang, T. Minh, L. Lezondra, S. Korolev, F. Collart, A. Joachimiak. Accepted for publication to Proteins: Structure, Function, and Genetics.
W.M. Keck Center
for Molecular Structure (California State University at Fullerton)
PI: Katherine
Kantardjieff [5/2001-8/2001]
Homology Modeling Of Botulinum Neurotoxins C-G And Type A Antibodies (2003) – T. Minh, J. Warfel, S. R. Herron, K. A. Kantardjieff; 38th ACS Western Regional Meeting Abstracts 2003: 199
Botulism is a serious paralytic disease caused by botulinum neurotoxin (BT), a zinc endopeptidase and the most
potent enzyme known. The aerosolized form is currently listed by the Centers for Disease Control as one of six
high-priority agents (cat. A) that could be used as a major biological weapon. There are seven antigenically distinct
forms of the toxin, which have a distinct geographic distribution in the United States. X-ray crystal structures exist
for toxin serotypes A and B. Although the overall three-dimensional structures of BT-A and BT-B are similar,
structural details are different, owing to the specificity for the substrate and the scissile bond being cleaved, as well
as for the receptors being bound. A pentavalent toxoid against toxins A-E has been developed as an investigational
agent, but it does not distinguish the serotypes. Equine botulinum antitoxin, which has antibodies to toxins A, B, and
E, will not reverse existing paralysis. We have used and evaluated several different homology modeling programs to
derive structurally valid models of each serotype, as well as 38 different antibodies (scFvs) that have been shown in
vivo to bind to BT-A. MODELLER and JACKAL produced complete (entire sequence) models with the best
stereochemistry and fold quality, as determined by PROCHECK and ICM. Simulated protein-protein docking
between the toxins and the antibodies will be used in structure-guided design of antibodies for identification and
inactivation/neutralization of BT serotypes in the event of a biological attack.
Simulated Rigid-Body Docking Of Antitoxins To Botulinum Toxin (2003) – J. Warfel, T. Minh, S. R. Herron, K. A. Kantardjieff; 38th ACS Western Regional Meeting Abstracts 2003: 193
Botulinum neurotoxin, produced by Clostridium botulinum, is responsible for the condition referred to as Botulism.
The condition results in flaccid paralysis and is most often life-threatening. There is a growing concern that in
aerosolized form it will be used as a bio-warfare agent, and it is listed by the Centers for Disease Control as one of
six high-priority agents. Within the botulinum toxin family, there are seven serotypes, A-G, which are found in
geographically distinct regions of the United States. Three-dimensional crystallographic structures for serotypes A
and B reveal conserved and non conserved regions that determine substrate specificity and receptor binding.
Typically, a penta- or heptavalent toxoid is given as treatment for exposure to toxin, but this can result in serum
sickness, anaphylaxis and urticaria. These side effects have led to the development of humoral and murine immune
response libraries to toxin serotype A. The 38 different antibodies (scFvs) have been modeled using MODELLER,
and toxins C-G have been modeled using JACKAL (see poster by Minh et al.) We have performed preliminary
rigid-body docking of validated model antibodies to the 3D structure of toxin serotype A using ICM-Docking, which
can predict binding geometry based on global energy optimization. Reasonable rigid-body dockings, consistent with
experimental binding data, have been selected for explicit global optimization of the surface side-chains. The
generated solutions will be used in structure-guided design of antibodies for identification and
inactivation/neutralization of toxin serotypes in the event of a biological attack.
Towards Statistical Testing of Protein Crystallization Parameters (2002) –A. Kataoka, E. Salinas, T. Minh, S. R. Herron, K. A. Kantardjieff, B. Segelke, and B. Rupp; Biophys. J. (Annual Meeting Abstracts) 2002 : 167a, San Francisco, CA
There are currently no a-priori predictive methods to guide the selection of conditions for crystallization of
proteins. In the absence of predictive methods and with no possibility for comprehensive screening due to
the multidimensional nature of crystallization space, efficient search routines are critical for successful
crystallization. Provided that efficient sampling is also statistically valid (such as random sampling
implemented in CRYSTOOL), correlations between protein properties and the crystallization conditions
can be established. The overall success rate of crystallization can then be optimized by more frequent
sampling of conditions shown to have a statistically significant higher likelihood for success.
Undergraduate and high school research students have begun to build a database from the sub-optimal
Biological Macromolecule Crystallization Database and the NASA Protein Crystal Growth Archive
containing information about crystallization, space group, sequences, tags, pI of the macromolecule, pH of
crystallization and other possible factors, and to search the data for correlations to crystallization success.
Analysis to date suggests that the correlation between pI and pH of successful crystallization is not zero,
and the correspondence is statistically significant using the 0.1 standard. K-mean cluster analysis suggests
that the data can be divided into two subsets, with pIs above and below 7. Correspondence between pI and
pH is somewhat stronger for the first group, but neither correspondence is statistically significant. We also
find that, in correlation with prevailing practice in protein purification, proteins crystallized with His-tags
are predominantly N-terminally tagged. N-terminal tags were visible in electron density in 41 % of cases,
whereas only 13% of C-terminal tags were visible in electron density.