Biology Reference
In-Depth Information
laboratory that are incapable of growth on minimal medium with acetate as the sole
carbon source. First we compared whether the corresponding
in silico
mutants fol-
lowed the same behavior (lack of growth on minimal medium with acetate, where zero
biomass flux during FBA corresponded to a no-growth phenotype). This comparison
was performed only for strains whose knocked-out gene is included in iJP815. Thirty-
eight out of the 51 strains tested did not grow
in silico
. Of the remaining 13 false
positives (i.e., those growing
in silico
but not
in vivo
), four (PP1470, PP1471, PP4679,
and PP4680) are mutated in genes considered non-essential
in silico
due to “weakly
annotated” gene putatively encoding redundant isozymes. In the case of PP5185 (cod-
ing for N-acetylglutamate synthase), its essentiality is removed by PP1346 (coding for
bifunctional ornithine acetyltransferase/N-acetylglutamate synthase protein), which is
not only an isozyme of PP5185 (the N-acetylglutamate synthase function) but which
also catalyses a reaction (ornithine acetyltransferase) that produces N-acetyl-L-glu-
tamate (the product of N-acetylglutamate synthase) and thus renders the activity of
PP5185 redundant. It appears either that this is a mis-annotation or that the enzyme is
utilized only under different conditions.
In addition, PP0897 (
fum
C) seems to have two paralogues (PP0944, PP1755) cod-
ing for isoenzymes of fumarate hydratase, but since the mutant in PP0897 does not
grow auxotrophically, they are either non functional or mis-annotated. The enzyme
complex that is composed of proteins expressed from the genes knocked-out in the two
false positives PP4188 and PP4189 catalyzes the decarboxylation of α-ketoglutarate
to succinyl-CoA in the TCA cycle, concurrently producing succinyl-CoA for anabolic
purposes. In the model, this functionality is not needed as this part of the TCA cycle
can be circumvented by the glyoxylate shunt, whereas succinyl-CoA can be produced
by reverse operation of succinate-CoA ligase. Restricting this reaction to be irrevers-
ible renders both genes essential. This altogether suggests that either the succinate-
CoA ligase is irreversible or the glyoxylate shunt is inactive. The latter solution is,
however, impossible, due to the essentiality of the glyoxylate shunt upon growth on
acetate.
The false positive PP4782 is involved in thiamine biosynthesis. This cofactor is not
included in the biomass, which is why the gene is not
in silico
essential. This suggests
thus that the
in-silico P. putida
biomass reaction should be enriched with this cofactor.
The remaining false positives (PP1768, PP4909, PP5155) are involved in the serine
biosynthesis pathway. We found experimentally that mutants in these genes can grow
on acetate if the medium also contains L-serine. These genes can be rendered
in silico
essential by setting glycine hydroxymethyltransferase to operate only unidirection-
ally from L-serine to glycine. The operation of this enzyme, however, is required for
growth of the bacterium on glycine, which is possible; though very slow (results not
shown). One of these genes (PP5155) has also a weakly annotated isozyme (PP2335).
We found out as well that several of the mutants (PP1612, PP4188-9, PP4191-4) grow
in silico
on glucose, which we confi rmed experimentally (results not shown). Alto-
gether, these experimental results assisted us in improving the accuracy of the model.
Albeit limited to a relatively small mutant set, this analysis shows that while CB
models are not always able to predict exact fl ux values, they are very useful in the
