Biology Reference
In-Depth Information
identifi cation of essential reactions and, through the GPRs, the genes responsible for
their catalysis. This enables identifi cation of vulnerable points in the metabolic
network.
Model Application—Production of Polyhydroxyalkanoates from
Nonalkanoates
To illustrate the utility of a genome-scale model for metabolic engineering, we used
iJP815 to predict possible improvements to an industrially relevant process; namely,
the production of PHAs from non-alkanoic substrates for biomedical purposes [61-
63]. As the production of PHAs uses resources that would be otherwise funneled to-
wards growth, increasing
in silico
PHA production would decrease the growth. Con-
sequently, in classic optimization-based approaches (e.g., FBA), no PHA production
would be predicted while optimizing for growth yield. The aim was thus to increase
the available pool of the main precursor of PHAs—Acetyl Coenzyme A (AcCoA).
This approach was based on the observation that inactivation of isocitrate lyase (ICL)
enhances the production of PHAs in
P. putida
due to increased availability of AcCoA
that is not consumed by ICL [64]. We therefore searched for other possible interven-
tion points (mutations) in the metabolic network that could lead to the accumulation of
AcCoA. This analysis was performed through application of a modified OptKnock ap-
proach [28], which allowed for parallel prediction of mutations and carbon source(s)
that together provide the highest production of the compound of interest.
Two main methods were employed to model a cellular pooling of AcCoA. The fi rst
was the maximization of AcCoA production by pyruvate dehydrogenase (PDH). In
the second, an auxiliary reaction was introduced that consumed AcCoA (concurrently
producing CoA, to avoid cofactor cycling artifacts) and that would represent the pool-
ing of AcCoA (Figures 6A and B, insets). It is noteworthy that the value of “AcCoA
production” predicted by the fi rst method includes AcCoA that is then consumed in
other reactions (some of which will lead towards biomass production for instance),
whereas the value of “AcCoA pooling” predicted by the second method includes only
AcCoA that is taken completely out of the system, and therefore made available for
PHA production but unusable for growth or other purposes. Therefore, only with the
fi rst method (AcCoA production) can AcCoA fl uxes and growth rates be compared
directly with the wild type AcCoA fl ux and growth rate, as the second method (AcCoA
pooling) will display lower values for AcCoA fl uxes and growth rates but will avoid
“double counting” AcCoA fl ux that is shuttled towards growth, and therefore is not
available for PHA production (see plots in Figures 6A and B).
To create the
in silico
mutants, we allowed the OptKnock procedure to block a
maximum of two reactions, which corresponds, experimentally, to the creation of a
double mutant. To avoid lethal
in silico
strains, the minimal growth yield was limited
to a value ranging between 0.83 and 6.67 g
DW
mol
C
−1
, corresponding to about 5 and
40% of maximum growth yield, respectively.
Six mutational strategies suggested by this approach are presented in Table 5. The
fi rst three were generated by the AcCoA production method, and the last three were
generated by the AcCoA pooling method. The results provide a range of options for
