Novel Drug Design Strategies for the Treatment of Human African Trypanosomiasis

 

Structural Analysis, Experimental Methods, and Drug Design

 

Structural features of T. brucei, E. coli, and human PFK
T. brucei PFK uses ATP as a high-energy phosphate donor, yet structurally, this enzyme belongs to the family of inorganic pyrophosphate-dependent PFK. Both human and E. coli PFK are energetically dependent on ATP. The three-dimensional structure of E. coli PFK is known, although the structure of human PFK is not. However, the amino acid sequence of the human enzyme is known. Therefore, comparison of T. brucei PFK to human PFK is largely based models derived from amino acid sequence alignment. Analyses by Martinez-Oyanedel et al. (2006) revealed the key primary and secondary structural differences between the T. brucei, E. coli, and human PFK.
Quartenary structure and global fold of T. brucei PFK. PFK is a tetramer (a pair of dimers) composed of identical 50kD subunits. It is considerably smaller than mammalian and some other ATP-dependent PFK monomers. T. brucei PFK has three domains, designated A, B, and C. Domains B and C are very compact, each containing 4 or 5-stranded, roughly antiparallel beta sheets flanked by alpha helicies. The domain boundaries were assigned by visual inspection. Each subunit contains a flexible loop in domain A that links the dimers. (in Figure 1, this loop is residues 62-81, between β strands b and c). Since this loop is highly mobile and thus invisible in the electron density map, its structure could not be precisely determined (Martinez-Oyanedel et al., 2006).	Figure 1. Structure of T. brucei PFK-1 apoenzyme subunit. Domains A, B, and C are shown in blue, green and red, respectively. A loop in domain A, between beta strand b and c contains residues that are invisible in the electron density map.
The N-terminal domain of T. brucei PFK is absent in E. coli PFK. A 94-residue N-terminal portion of domain A in T. brucei PFK is not present in the E. coli structure (Figure 2).  Within this N-terminal section of domain A, the loop with the mobile, invisible tip crosses from one domain to the adjacent one (not shown). These two crossing loops appear like "embracing arms" that link the two dimers. This feature may serve to confer additional stability to the quatenary structure of T. brucei PFK (Figure 3). The N-terminal loop may also serve to mediate interactions between PFK and other proteins in the glycosome. Glycolytic enzymes in the glycosome are known to form multi-enzyme complexes. These enzyme complexes are not observed in mammalian or other eukaryotic cells because these glycolytic enzymes are located in the cytosol (Martinez-Oyanedel et al., 2006).
Figure 2. The N-terminal portion of domain A from T. brucei subunit. The residues highlighted in red (residues 1-95) are absent in the E. coli and human PFK enzymes.	Figure 3. Space-filling representation of T. brucei PFK. The N-terminal loop of each dimer crosses over the other, forming "embracing arms."
The putative effector site of T. brucei PFK. Structural superimposition of T. brucei PFK and E. coli PFK was used to determine the putative effector site in the trypanosome structure. Structually, the major difference between the two effector sites is the loop length. However, the two enzymes differ significantly in amino acid composition at the effector site. Only two of the twelve residues in the effector site of the E. coli structure are conserved in the T. brucei enzyme. Additional structural studies of T. brucei with a bound ligand (such as a metabolite from the cititric acid cycle) in the effector site must be conducted to understand the function of these features (Martinez-Oyanedel et al., 2006).
The active site and loop 329-348 of T. brucei PFK. Many residues at the active site of T. brucei and E. coli PFK are conserved and the two enzymes have very similar conformations in this region. All the active site residues are conserved between the two PFKs, except Tyr380 in the trypanosome structure, which is His380 in the E.coli PFK. T. brucei PFK has an additional loop (residues 329-348) that is not present in the bacterial or human enzymes (Figure 4). The tip of this loop is immediately adjacent to the active site.  In both the bacterial and trypanosome PFKs, Asp229 (green in Figure 4) is opposite the 329-348 loop (red). In the E. coli structure, this residue serves as a general base. (For a description of the catalytic mechanism of bacterial ATP-dependent, see Evans 1992). The function of the 329-348 loop will be very important for mechanistic and drug design studies (Martinez-Oyanedel et al., 2006).	Figure 4. The active site of T. brucei PFK-1. The red residues represent the 329-348 loop. The green residue is Asp 229, which is a catalytic base in the E. coli structure.
PP­i and ATP-dependent PFK comparisons. The trypanosomatid PFK more closely resembles members of the PPi-dependent family of PFK enzymes, but uses ATP as a high-energy phosphate donor.   It is likely that the trypanosomatid PFKs have evolved from an ancestral PPi-dependent PFK, with "concomitant" phosphor-donor specificity (Martinez-Oyanedel et al., 2006). Amino acid sequence comparison. Sequence identity between T. brucei PFK and ATP or PPi-dependent PFK (Michels et al., 1997): 35% with PFK of bacterium Amycolatopsis methanolica (PPi-dependent) 30% with PFK of bacterium Borrelia burgdorferi, which causes Lyme disease (PPi-dependent) 25% with PFK of E. coli (ATP-dependent) 24% with catalytic domain of human muscle PFK (ATP dependent) Quartenary structure comparison of T. brucei and B. burgdorferi PFK. T. brucei PFK (ATP-dependent) and B. burgdorferi (PPi-dependent) PFK are more similar to each other than they are to the E.coli or human ATP-dependent PFKs.  The structural superimposition of T. brucei and B. burgdorferi structures shows a high degree of similarity.  For example, the B. burgdorferi enzyme posses the N-terminal loop found in domain A of the T. brucei structure (Figure 2). The T. brucei structure is 70 residues longer at this N-terminal region than are non-trypanosomatid ATP-dependent PFKs. In addition, the B. burgdorferi structure has a loop structure that interacts with the active site, which corresponds to the 329-348 loop in T. brucei PFK. This loop, however, is a hairpin in the B. burgdorferi PFK (Martinez-Oyanedel et al., 2006). Active stire structure comparison of T. brucei and B. burgdorferi PFK. Although the general folds of the B. burgdorferi and T. brucei PFKs are similar, the two structures differ significantly at the active site. These differences may confer specificity for ATP or pyrophosphate as the substrate. Moore et al. (2002) proposed that B. burgdorferi PFK binds PPi (instead of ATP) because of Asp177, which prevents the binding of nucleotides with an α-phosphoryl group. The corresponding residue in T. brucei (as in all other ATP-dependent PFKs) is glycine. Interestingly, site-directed mutagenesis on Asp177  is sufficient to alter the substrate specificity from PPi to ATP.
Global fold comparison of T. brucei and modelled human PFK. No crystal structure of human PFK has been solved. However, using the T. brucei structure (the only other eukaryotic PFK with known structure) and the primary sequence of human PFK, Martinez-Oyanedel et al. (2006) developed a model of human PFK using the protein structure modeling program MODELLER7v7 (Experimental Methods). This model was used for comparison with T. brucei PFK. The analysis yielded the following differences between T. brucei and human PFK-1 : 1. Subunit size. The trypanosome PFK-1 subunit is 50kDa, while that of the human PFK is 85kDa. This change in subunit size is likely due to gene duplication/fusion (Michels 1997). 2. Presence of 329-348 loop. The 329-348 loop (Figure 4) of the T. brucei structure is absent in the human muscle PFK-1. 3. N-terminal portion of domain A. The entire N-terminal region of domain A found in the T. brucei structure (Figure 2) is absent from human PFK-1. This includes helicies 1 and 2, strands a,b, and c, and the "embracing arm." These features provide excellent opportunites for the development of species-specific inhibitors of the parasitic PFK.
Figure 7. Comparison of T. brucei and human PFKs. (a) Superimposition of the crystal structure of T. brucei PFK-1 (blue) and modelled human PFK-1 (pink). Highlighted in cyan is the 329-248 loop, a structural feature unique to T. brucei PFK-1. There are two insertions in the human PFK-1 that were not included in the model. These insertions are located between helix 8 and strand h (residues 138-154, human PFK-1 numbering) and at the C-terminus (residues 368-388, human PFK-1 numbering). (b) Proposed model for the quartenary structure of human PFK-1. Each subunit is shown as two tones of the same color (blue, red, brown, and grey). The darker tones correspond to the N-terminal halves. The N and C-terminal halves of each subunit are are linked by alpha-helical peptides.
Experiemental Procedures
An introduction to X-ray crystallography. The T. brucei and E. coli PFK-1 structures were solved using x-ray crystallography. X-ray crystallography provides the highest resolution images of protein structures currently available. The technique can reveal the precise three-dimensional arrangement of all the atoms in a protein. X-rays are used because the wavelength of an x-ray is about the same length as a covalent bond. There are three major components in the process of x-ray crystallography: growing the protein crystal, exposing the crystal to x-rays, collecting data, and analyzing the data to determine the three-dimensional structure (Berg et al. 2005). Protein crystals are obtained by slowly adding salt to a concentrated solution of protein. The salt reduces the solubility of the protein and favors the formation of highly ordered crystals. Some proteins crystallize more readily than others (Berg et al. 2005). Next, a beam of x-rays with a wavelength of 1.54 A is directed at the crystals. Part of the beam goes through the crystal and the rest is scattered in several directions, producing a pattern of scattered x-rays on the film. This pattern reveals ample information about protein structure (Berg et al. 2005). The next step is to reconstruct the protein's atomic structure based on the diffraction pattern. To do this, a mathematical relationship called a Fourier transform is applied to the data. For each spot on the film, the Fourier transform calculates a wave whose amplitude is proportional to the square root of the observed intensity of the spot. After the Fourier transform calculations, the stage is set for calculation of the electron density map. The electron density can be used to calculate the final structure of the protein (Berg et al. 2005). The development of computer programs has facilitated the analysis of x-ray crystallography data. The Protein Data Bank (http://www.rcsb.org/pdb/home/home.do) catalogues thousands of structures solved by x-crystallography. These structures are readily visualized using modern molecular graphics programs
Figure 7. X-ray diffraction pattern for a protein crystal. Image courtesy of the Department of Biochemistry and the School of Medical Sciences, Otago University.	Figure 6. Protein crystals. Crystals of insulin. Image courtesy of NASA.
Protein model building. The comparative protein structure modeling program, MODELLER7v7 was used to develop the model of human PFK. The sequence of the N-terminal half of human muscle PFK and structural alignments between human and T. brucei PFK were used to build the model. MODELLER7v7 uses the principle of modeling by satisfaction of spatial restraints. The program develops a set of restraints derived from the alignment, and the model is then obtained by minimizing the violations to these restraints (Wallner 2005). The optimally satisfied spatial restraints are expressed as probability density functions for the features restrained. For example, the probabilities for main-chain conformation of a modeled residue may be restrained by residue type, conformation of an equivalent residue in the related protein, and local similarity between the two sequences (Sali and Blundell 1993). The quality of the model can be evaluated with another program called PROCHECK, which evaluates the residues based on a Ramachandran plot. Using PROCHECK, Martinez-Oyanedel et al. (2006) determined that 87% of the residues were in the most favored regions of a Ramachandran plot.
Drug design and future studies
The unique structural features of the trypanosome PFK provide the basis for the development of species-specific inhibitors. These targeted structural features are the 329-348 loop, and the N-terminal portion of domain A. A number of iInhibitors containing lead have been identified and are undergoing further development (Martinez-Oyanedel 2006). These inhibitors were developed by the research groups of Prof. Linda A. Fothergill-Gilmore, Prof. Malcom D. Walkinshaw, and Prof. Nick Turner at the University of Edinburgh in Edinburgh, Scotland. They have designed and written a database-mining program called LIDAEUS (LIgand Discovery At Edinburgh UniverSity) to search potential ligands for the binding pockets of T. brucei PFK. A number of novel ligands that bind with micromolar dissociation constants have been discovered and they have selected ligands suitable for combinatorial chemistry.