From Protein Structure to Function with Bioinformatics.pdf

282 J.D. Watson and J.M. Thorntonin Pfam (PF01979) which contains a number of deaminases and is part of a wideramidohydrolase superfamily clan. As an unknown function protein it was selectedfor analysis using ProFunc and the results indicated a probable match to one of the189 known enzyme active site templates used by the server.The enzyme match (with an e-value of 2.45 × 10 −4 ) is to the active site templatefor adenosine deaminase (EC 3.5.4.4) derived from PDB entry 1a4l, which is involvedin purine metabolism. The overall sequence identity between Tm0936 and this structure,based on a pairwise FASTA alignment, is only 24%, yet the structural similarity(calculated by ProFunc as the number of residue-pairs in the structural alignment thatlie in one or more fittable segments, as a percentage of the number of residues equivalencedby the alignment) is 95% indicating a high level of structural similarity. Withina 10 Å sphere of the template match the local sequence identity is 27.7% which indicatesthere is a higher level of similarity around the active site than throughout theentire protein. In addition to this strong match there are a number of reverse templatematches to other deaminases and amidohydrolases.The authors of the structure of Tm0936 had published a predicted function for thisprotein as an adenine deaminase (Hermann et al. 2007). The approach they usedinvolved structure-based targeted docking using metabolic intermediates which are ina high energy state, the principle being that docking initial substrates or products toenzymes may not be as effective as docking the intermediates that are stabilised by theenzyme. The resultant hit list of putative ligands was dominated by adenine analogues,appearing well-positioned to undergo C6-deamination. Four were tested as substrates,with three showing substantial catalytic rate constants. The structure of the complexbetween Tm0936 and the product (S-inosylhomocysteine) resulting from the deaminationof S-adenosylhomocysteine was determined. The enzyme template match showeda very close overlap between this bound ligand and the DCF ligand (deoxycoformycin,an inosine analogue) bound in the template structure of 1a4l (Fig. 11.1).Fold analysis using MSDfold picks up various amidohydrolases and guanine/cytosine deaminases. The reason that MSDfold did not identify the adenosinedeaminase structure as the best hit is that the Tm0936 structure has some additionalembellishments outside the core match. This resulted in the structural similarityfalling below the default cutoff of 70% of secondary structure elements required fora match (Fig. 11.2). This highlights the power of local comparisons and is an illustrativecase where function can be accurately identified by strong enzyme templatematches that could not have been identified by sequence methods alone.Another interesting case comes from the MCSG with the recent publication of thestructures of the open (R) and closed (T) states of prephenate dehydratase (PDT) (Tanet al. 2008), which have implications for our understanding of its allosteric regulationby L-phenylalanine and other amino acids. This enzyme (EC 4.2.1.51) converts prephenateto phenylpyruvate in l-phenylalanine biosynthesis. Its key role in the biosynthesisof l-Phe in organisms that utilize the shikimate pathway makes it an essentialenzyme in microorganisms. The fact that this enzyme is not found in humans meansit could serve as a possible drug target against microbial pathogens.The structures deposited in the PDB (2qmw and 2qmx) come from two differentorganisms and represent the first crystal structures of PDT in relaxed (R)