Relationship between Metabolic Fluxes and Sequence-Derived Properties of Enzymes

Abstract

Metabolic fluxes are key parameters of metabolic pathways being closely related to the kinetic properties of enzymes, thereby could be dependent on. This study examines possible relationships between the metabolic fluxes and the physical-chemical/structural features of enzymes from the yeast Saccharomyces cerevisiae glycolysis pathway. Metabolic fluxes were quantified by the COPASI tool using the kinetic models of Hynne and Teusink at varied concentrations of external glucose. The enzyme sequences were taken from the UniProtKB and the average amino acid (AA) properties were computed using the set of Georgiev's uncorrelated scales that satisfy the VARIMAX criterion and specific AA indices that show the highest correlations with those. Multiple linear regressions (88.41% <R adjusted (2) < 93.32%; P < 0.00001) were found between the values of metabolic fluxes and the selected sets of the average AA properties. The hydrophobicity, α-helicity, and net charge were pointed out as the most influential characteristics of the sequences. The results provide an evidence that metabolic fluxes of the yeast glycolysis pathway are closely related to certain physical-chemical properties of relevant enzymes and support the view on the interdependence of catalytic, binding, and structural AA residues to ensure the efficiency of biocatalysts and, hence, physiologically adequate metabolic processes.

Figure 1

Linear and nonlinear pair correlations between the metabolic fluxes and the average AA properties of the yeast Saccharomyces cerevisiae enzyme sequences, as specified in Table 1; the data represent Teusink's (a) and Hynne's ((b), (c)) models I and II, respectively (Table 2). The correlations are significant at the nonparametric assessment (Kendall's τ, Spearman's ρ correlation coefficients).

Figure 2

The multiple linear regressions showing changes of the metabolic fluxes as dependent variables upon the two average AA properties of the yeast Saccharomyces cerevisiae enzyme sequences, as specified in the Table 1. The data ((a), (c)) represent models I and II, respectively (Table 2). The observed versus predicted plots ((b), (d)) for the values of dependent variables ((a) and (c), resp.). The predicted values were calculated from the regression equations: flux (model I) = Flux: 108.975 + 988.917∗P _aveWV7 − 553.390∗P _aveWV5 (R _adj. ² = 47.84%, P = 0.0000); flux (model II) = 47.576 − 0.757∗(P _aveWV2)² + 3.696∗P _aveWV7 (R _adj. ² = 35.71%, P = 0.0000). All the multiple and pair correlations ((a), (b), (c), (d)) are significant at the nonparametric assessment (Kendall's τ, Spearman's ρ correlation coefficients).

Figure 3

The changes in the percentage of explained variance (□) and the values of corrected Akaike's information criterion (AIC c) (▲) on the growing number of independent variables (the average AA properties of enzyme sequences) included in the multiple regression. Variables in the model (Tables 1 and 2, model II): 1—(P _aveWV2)², 2—(P _aveWV2)², P _aveWV7, 3—(P _aveWV2)², P _aveWV7, P _aveWV1, 4—(P _aveWV2)², P _aveWV7, P _aveWV1, (P _aveWV1)², 5—(P _aveWV2)², P _aveWV7, P _aveWV1, (P _aveWV1)², (P _aveWV5)², 6—(P _aveWV2)², P _aveWV7, P _aveWV1, (P _aveWV1)², (P _aveWV5)², P _aveWV5, 7—(P _aveWV2)², P _aveWV7, P _aveWV1, (P _aveWV1)², (P _aveWV5)², P _aveWV5, P _aveWV6, 8—(P _aveWV2)², P _aveWV7, P _aveWV1, (P _aveWV1)², (P _aveWV5)², P _aveWV5, P _aveWV6, (P _aveWV3)², 9—(P _aveWV2)², P _aveWV7, P _aveWV1, (P _aveWV1)², (P _aveWV5)², P _aveWV5, P _aveWV6, (P _aveWV3)², P _aveWV10^a, 10—(P _aveWV2)², P _aveWV7, P _aveWV1, (P _aveWV1)², (P _aveWV5)², P _aveWV5, P _aveWV6, (P _aveWV3)², P _aveWV10, P _aveWV9^b. (a) The scale WV10 correlates with the NMR parameters and pK values of AA [18]. (b) The scale WV9 correlates with the indices of protein backbone topography and relative mutability of AA [18].