A method for prediction of the locations of linker regions within large multifunctional proteins, and application to a type I polyketide synthase

Abstract

Multifunctional proteins often appear to result from fusion of smaller proteins and in such cases typically can be separated into their ancestral components simply by cleaving the linker regions that separate the domains. Though possibly guided by sequence alignment, structural evidence, or light proteolysis, determination of the locations of linker regions remains empirical. We have developed an algorithm, named UMA, to predict the locations of linker regions in multifunctional proteins by quantification of the conservation of several properties within protein families, and the results agree well with structurally characterized proteins. This technique has been applied to a family of fungal type I iterative polyketide synthases (PKS), allowing prediction of the locations of all of the standard PKS domains, as well as two previously unidentified domains. Using these predictions, we report the cloning of the first fragment from the PKS norsolorinic acid synthase, responsible for biosynthesis of the first isolatable intermediate in aflatoxin production. The expression, light proteolysis and catalytic abilities of this acyl carrier protein-thioesterase didomain are discussed.

Figure 1

The estimated probability distributions of UMA scores S(i,j ), calculated with the BLO-SUM62 sequence homology matrix, and the secondary structure homology matrix from Table 1. The weighting factors are: K_α = K_β = K_C = 5, Q_α = Q_β = Q_C = 1, and γ = 20. As the number of sequences included in the alignment increases, the most commonly occurring S(i,j ) generated from an alignment of unrelated sequences decreases, demonstrating the lack of conservation between these sequences.

Figure 2

(a) UMA prediction graph for E. coli MetH. (b) Crystal structure of the C-terminal domains of E. coli MetH.

Figure 3

UMA prediction graph for E. coli DNA polymerase I. The region shaded blue corresponds to the Klenow fragment.

Figure 4

UMA prediction graph for E. coli ThiI with the known trypsin-preferred cut site as indicated.

Figure 5

UMA prediction graph for human FAS. I, II, and III indicate Wakil’s domains I, II, and III, respectively. ID indicates the predicted linker-like interdomain region.

Figure 6

UMA prediction graph for A. parasiticus NA synthase PKS with corresponding predicted domains.

Figure 7

Light proteolysis of TA1 by proteinase K. Lanes: (1) Benchmark protein ladder (Invitrogen); (2) 0 minute, no proteinase K; (3) 0 minute, added proteinase K; (4) one minute; (5) two minutes; (6) five minutes; (7) ten minutes; (8) Benchmark protein ladder.

Figure 8

Components of the prediction of E. coli DNA polymerase I. (a) The sequence similarity component: Q_α = 1, Q_β = Q_C = 0; (b) secondary structure component: Q_α = Q_C = 0, Q_β = 1; (c) hydrophobicity component: Q_α = Q_β = 0, Q_C = 1; (d) final score Q_α = Q_β = Q_C = 1.