Biology Reference
In-Depth Information
dicted essential reactions sets (259 and 274 reactions for glucose and acetate growth,
respectively). However, it must first be noted that each essential reaction set contains
about 25 (26 and 27 for the glucose and acetate essential set, respectively) non-gene-
associated reactions, and that elucidating the genes catalyzing these reactions would
increase substantially the number of
in silico
essential genes. Further, approximately
20% of each minimal set (78 and 84 genes under glucose and acetate conditions, re-
spectively) consists of essential reactions that can be catalyzed by two or more iso-
zymes and thus are essential at the metabolic network level but not at the genetic level.
In contrast to non-essential reactions in the minimal set, these reactions essential at
the metabolic network level but not at the genetic level are not clustered in particular
metabolic pathways but are rather spread throughout the entire metabolic network. Al-
together, these results indicate that for about 40% of the reactions required to produce
biomass, there are alternative at either the genetic or the metabolic network level.
This analysis highlights the limitations of possible interventions aimed at reshap-
ing the fl ux distributions, because these can be applied only to reactions that are not
essential (since the inactivation of an essential reaction yields a lethal phenotype).
Identifi cation of reactions catalyzed by multiple enzymes shows which reactions may
be best avoided when planning mutational strategies as their inactivation may pose
additional technical problems, by requiring production of multiple knock-outs.
Flexibility of Flux Distributions
To further investigate these conclusions, we determined the flexibility of fluxes over
particular reactions as a measure of metabolic network flexibility during biomass pro-
duction. We found that the variability of fluxes is similar under either glucose or ac-
etate growth, but that acetate growth instills a slightly higher rigidity to the metabolic
network (as observed above). We observed also that the flux of more than a half of
the reactions can vary to some degree without influencing biomass output. We next
analyzed the pathway-distribution of reactions exhibiting variable flux, and found that
biosynthetic pathways are in general more rigid (i.e., the fraction of reactions with
flexible flux is relatively lower) than other pathways. This rigidity might reflect the
essentiality of these pathways modules for the survival of the cell. A further measure
to ascertain network flexibility was the assessment of pairwise couplings between the
reactions via flux coupling finder. This analysis indicated that for 90% of the reactions
that are unblocked in a given condition, at least one other reaction exists whose flux
is proportionally coupled to the flux of the first reaction, and therefore that the great
majority of reactions can be inactivated through inactivation of some other reaction.
This analysis is helpful in optimizing mutational strategies as it pin-points alternative
mutations that exhibit equivalent outcomes.
Prediction of Auxotrophic Mutations and Model Refinement
Assessment of network models through comparison of
in silico
growth-phenotypes
with the growth of knock-out strains is a powerful way to validate predictions. This
has been done in a number of studies for which knock-out mutant libraries were avail-
able [59, 60]. As there is currently no mutant library for
P. putida
, we tested gene
knock-out predictions with a set of
P. putida
auxotrophic mutant strains created in our
