Biology Reference
In-Depth Information
Figure 4 shows that the predictions (in red) generally agree well with the measure-
ments (in green) throughout the network, as most of the
13
C values fall within the FVA
intervals, where intervals were predicted, or both values are close to each other (in
absolute values), when a single value was predicted. As
P. putida
lacks phosphofruc-
tokinase, glucose can be converted to pyruvate (the entry metabolite of TCA cycle) via
the pentose phosphate (PP) or the Entner-Doudoroff (ED) pathways. The ED pathway
is energetically more effi cient and the
13
C measurements indicate that KT2440 uses it
preferentially over the PP pathway. Therefore, the FVA yields locally single fl ux val-
ues rather than intervals, which refl ects the relative rigidity of this part of the network.
In contrast, the energy generating part of the central metabolic network (the TCA cycle
and its vicinity) exhibits greater fl exibility, as illustrated by the broad fl ux intervals.
First, the conversion of phosphoenylpyruvate into pyruvate can proceed either directly
or via oxaloacetate, although the bacterium appears to use the direct route (the
13
C
-model assumes, in fact, only the direct route. Second, the conversion of malate to
oxaloacetate may also occur directly or via pyruvate. The
13
C fl ux measurements in-
dicate that the bacterium uses the indirect route in addition to the direct one although,
according to the FVA, the indirect route is energetically less effi cient. Interestingly,
our model suggests also that the glyoxylate shunt could be used interchangeably with
full TCA-cycle without any penalty on growth yield. However, as the glyoxylate shunt
is inactivated in many bacterial species via catabolite repression upon glucose growth
[51], it is possible that this alternative is not used in
P. putida
.
Discrepancies Between Model Predictions and Measurements
Despite the general agreement between
in silico
predictions and
13
C measurements,
there still exist a number of discrepancies. For instance, the
13
C -experiments suggest
that the bacterium utilizes the portion of glycolysis between triose-3-phosphate and D-
fructose-6-phosphate in the gluconeogenic direction, which is not energetically opti-
mal and as such is not captured in standard FBA (or FVA) simulations. This illustrates
one of the possible pitfalls of FBA, which per definition assumes perfect optimality
despite the fact that microorganisms might not necessarily allocate their resources
towards the optimization function assumed in analysis, and in some cases may not
operate optimally at all [52, 53]. Another group of differences concentrates around the
pentose phosphate pathway (PPP), although these are relatively minor and are likely
due to differences in the quantities of sugar diverted toward biomass in the
13
C model
versus iJP815. A third group of differences revolves around pyruvate and oxaloacetate,
whereby the
in vivo
conversion of malate to oxaloacetate shuttles through a pyruvate
intermediate rather than directly converting between the two. The last area where dis-
crepancies exist between
in silico
and
13
C data is in the TCA cycle, around which the
flux is lower in FVA simulations than in the experiment. This suggests that the
in
silico
energetic requirements for growth (maintenance values) are still too low when
compared to
in vivo
ones, as the main purpose of the TCA cycle is energy production.
Suboptimal FVA
To investigate further these differences, we carried out a suboptimal FVA (Figure 4,
blue values), allowing the production of biomass to range between 90 and 100% of
