thereby the major cause of acute kidney failure in 
children. 
The structure of the B subunit of VT-1 , a member 
of the Shiga toxin family, has been determined by 
multiple isomorphous replacement (MIR). The B 
subunit forms a pentamer that binds to the carbohy- 
drate component of a cell-surface glycolipid, glo- 
botriaosylceramide (Gbj). Comparison of the 
amino acid sequences of all the toxins in this family 
indicates that a surface cleft between B-subunit 
monomers is likely to be the carbohydrate-binding 
site. Crystallographic binding studies are under way 
to test this prediction. 
The VT- 1 B subunit bears a striking resemblance 
to the B subunit of the E. co// heat-labile enterotoxin 
(LT), a member of the cholera toxin family. This is 
unexpected, because the associated A subunits are 
completely unrelated, the B subunits are very differ- 
ent in size (69 residues for VT-1 vs. 103 for LT), and 
there is no detectable sequence homology. In col- 
laboration with Drs. Titia Sixma and Wim Hoi (Uni- 
versity of Groningen) , the two B subunits have been 
compared. Fifty-two amino acids superimpose very 
closely in the structural alignment, yet only three 
are identical. 
Pertussis toxin (PT) is produced by Bordetella 
pertussis, the bacterium that causes whooping 
cough. This toxin's role in improved vaccines is a 
major interest. Killed whole-cell vaccines are rea- 
sonably effective but have an undesirable level of 
toxicity that could be reduced or eliminated in a 
defined vaccine produced from genetically engi- 
neered proteins. PT has been shown to be a neces- 
sary component of effective whooping cough vac- 
cines. A three-dimensional structure would help 
show how to remove its toxic activities while pre- 
serving the antigenic determinants necessary to 
evoke protective immunity. In addition, it would 
lead to a more complete understanding of the bio- 
chemistry of PT action. Considerable progress has 
been made on this structure. MIR, using synchro- 
tron data collected at the Photon Factory in Japan, 
has given useful phase information. This has been 
improved by a combination of electron density- 
averaging and solvent-flattening techniques. 
Psetidomonas aeruginosa Pilin 
Pseudomonas aeruginosa is an opportunistic 
pathogen that infects burn victims and immuno- 
compromised patients. It is also one of the major 
pathogens infecting the lungs of cystic fibrosis pa- 
tients. Colonization of epithelial cell surfaces is 
promoted by pili, filaments that are formed from a 
helical array of identical pilin subunits. PAK pilin 
(from Pseudomonas aeruginosa, strain K), ob- 
tained in collaboration with Dr. William Paranchych 
(University of Alberta), has been crystallized. Prob- 
lems with crystal quality that have hindered prog- 
ress are being overcome by improving the purity of 
protein preparations. The three-dimensional struc- 
ture would aid in understanding the details of cell- 
surface binding, which could be used to devise strat- 
egies to interfere with colonization. 
Fab (antigen-binding fragment of immunoglobu- 
lin) fragments from two monoclonal antibodies 
raised against PAK pilin have been crystallized (with 
Dr. Randall Irvin, University of Alberta). Pilin-Fab 
complexes can be formed in solution, and attempts 
to crystallize them are under way. In addition, a 
17-residue peptide from the carboxyl terminus of 
PAK pilin, obtained from Dr. Robert Hodges (Univer- 
sity of Alberta), has been shown to bind to one of 
these antibodies, PK99H. Attempts to crystallize a 
complex with this peptide are in progress. The 
structures of antibody complexes could be relevant 
to the development of peptide vaccines, since the 
monoclonal antibodies have been shown effective 
in passive immunization studies. 
Computer-aided Drug Design 
Most drugs with known mechanisms of action 
bind specifically to a drug receptor, often a protein 
molecule. Since crystallography allows one to exam- 
ine the structures of receptors and the details of 
their interactions with drugs, it has the promise of 
helping to improve existing drugs, or even to invent 
new ones. Progress has been made on improving ex- 
isting drugs using structural information, but the 
problem of exploiting such information to design 
new drugs from scratch is still extremely difficult. 
There are probably billions of compounds that 
might be used as drugs; determining which of these 
might bind to the target protein is far from trivial. 
A "divide and conquer" approach to drug design 
should reduce the magnitude of this problem. The 
vast set of possible compounds is made up of various 
combinations of a much smaller set of molecular 
fragments. It is proposed that drugs can be designed 
by using a computer first to dock members of a li- 
brary of fragments to the region of the desired bind- 
ing site, then to combine docked fragments to form 
chemically sensible molecules. Some success in the 
fragment-docking problem has been achieved by us- 
ing a Monte Carlo procedure to find favorable inter- 
actions. Multiple random docking trials are per- 
formed to ensure that most of the favorable 
interaction sites will be found. In test calculations. 
528 
