Proline degradation deficiency increases both hyp yield and formation rate
To determine how proline metabolism and its regulation affect the physiology of a
proline-hydroxylating biocatalyst, the impact of the putA deletion on exponential growth parameters and hyp synthesis was investigated. The
engineered strain, bearing either pET-24a (?putA_pET) or pET_p4h1of (?putA_p4h1of), was incubated aerobically in M9 minimal medium with glucose as carbon and
energy source in the presence or absence of proline and compared with the wildtype
strain cultivated under the same conditions (Table 1).
Table 1. Physiological comparison of aerobically growing recombinant E. coli BL21(DE3)(pLysS) strains
In case of pET-24a containing wildtype cells (wt_pET) grown in the presence of proline,
PutA associates with the membrane, catalyzing the first steps of proline degradation
to glutamate 12]. Glutamate is subsequently deaminated to the TCA cycle intermediate ?-KG, supplying additional carbon, nitrogen, and energy to the cells. This ability
of the wildtype to catabolize proline together with glucose led to an increase in
growth rate, final biomass titer, and biomass yield on glucose, compared to growth
on glucose alone. On the contrary, the mutant strain ?putA was, as expected, unable to degrade proline. The specific growth rate (?), final
biomass titer, and biomass yield on glucose of the ?putA_pET strain were virtually unchanged upon proline addition and resembled the wt_pET
strain grown on glucose (Table 1, see also Additional file 1: Figure S1). Similarly, proline addition had no effect on the growth rate of the
?putA_p4h1of strain. Interestingly, the presence of proline also did not significantly influence
the growth rate of the wildtype strain containing pET_p4h1of (wt_p4h1of). This can
be explained by the reduced amount of proline used for biomass formation due to hydroxylation,
and the decreased proline uptake rate induced by P4H activity 11]. Nevertheless, proline addition increased the final biomass titer of the wt_p4h1of
strain.
In all cases, both growth rate and biomass formed were decreased, when P4H was produced,
indicating a metabolic burden imposed by heterologous P4H synthesis 14]. The glucose uptake rate and the biomass yield on glucose were also reduced upon
P4H synthesis, independently of proline addition, indicating that less energy and
metabolite precursors are used for biomass formation. It should be noted that P4H
was successfully produced at constant levels under all conditions studied (see Additional
file 1: Figure S2) and that, in the absence of extracellular proline, both wt_p4h1of and
?putA_p4h1of could synthesize hyp from endogenous proline. Interestingly, upon proline
addition, the ?putA_p4h1of strain displayed the lowest glucose uptake rate (3.9 mmol g
?1
 h
?1
), but its biomass yield on glucose (0.45Â g
CDW
 g
glc?1
) was comparably high, suggesting energy-efficient biomass formation.
Acetate was secreted by all strains under the conditions studied, implying an excess
of available acetyl-CoA as a result of the TCA cycle not keeping pace with glycolysis
15], 16], and was assimilated after glucose depletion. The specific growth rate was not negatively
affected by the relatively low acetate concentrations (from 0.1 to 0.26 g L
?1
, see Additional file 1: Figures S1Â and S3), which were below the reported growth-inhibiting acetate concentrations
of 0.5–5 g L
?117], 18]. While the presence of proline resulted in an increased acetate formation rate in
the case of wt_pET and wt_p4h1of, no such effect was observed for the corresponding
?putA strains. Thus, proline-derived carbon influx into the central metabolism appears
to promote overflow metabolism. Interestingly, the ?putA_pET strain showed less acetate formation as the only difference to the wt_pET strain
indicating a higher carbon cost of maintenance. However, in the absence of PutA, P4H
synthesis (?putA_p4h1of) caused an increase in acetate yield on glucose, probably due to the metabolic
stress.
Overall, proline hydroxylation profited from the following beneficial effects of putA deletion: (1) quantitative transformation of proline into hyp, (2) doubling of the
specific hyp formation rate, and (3) a 2.3-fold higher molar hyp yield on glucose,
characterizing this strain as a favorable biocatalyst for proline hydroxylation (Table 1, see also Additional file 1: Figure S3).
Metabolic network operation: ?-KG formation via the TCA cycle does not increase upon P4H catalysis
To investigate how the putA deletion and hyp synthesis affect the operation of the intracellular reaction network,
13
C-MFA was performed for the wildtype and the ?putA strain, bearing either pET-24a or pET_p4h1of, during exponential batch growth in
the presence and absence of proline in M9 medium containing labeled glucose (80% [1-
13
C] and 20% [U-
13
C]). The relative carbon flux distributions throughout the central metabolic pathways
for all strains and conditions studied, normalized to the glucose uptake rate, are
mapped in Fig. 2.
Fig. 2. Metabolic fluxes in recombinant E. coli BL21(DE3)(pLysS) and E. coli BL21?putA(DE3)(pLysS) strains containing pET-24a or pET_p4h1of during exponential aerobic growth
at 30°C in M9 medium containing 5 g L
?1
glucose (4:1 mixture of 1-
13
C-labeled and U-
13
C-labeled glucose) in the absence or presence of 5Â mM proline. All fluxes are given
as relative fluxes normalized to the specific glucose uptake rate of 100 for each
strain (given for the reaction of glucose to glucose-6-phosphate). For the extracellular
metabolite fluxes, the experimental error is given. Arrows indicate the main direction of reversible reactions.
Flux patterns in the upper part of glycolysis (until phosphoenolpyruvate) were similar
for all strains and were virtually unaffected by the presence of proline. When using
putA-positive strains, unlabeled proline-derived carbon only became evident in amino acids
synthesized from TCA cycle intermediates (Table 2). As expected for ?putA strains, which are unable to channel proline carbon into central metabolism, the
unlabeled carbon content in amino acids (except for proline) did not change upon proline
addition.
Table 2. Fractional abundance of unlabeled amino acid fragments (m0) during cultivation on labeled glucose with and without unlabeled proline
The largest differences among strains were observed in the reactions mediated by the
malic enzymes and the phosphoenolpyruvate carboxykinase, the TCA cycle, and the fluxes
related to proline metabolism and hydroxylation. As the ?putA_pET strain was not able to catabolize proline, proline addition did not have an impact
on the flux distribution, which resembled that of the wt_pET strain grown on glucose
only. However, the ?putA_pET strain virtually substituted endogenous proline biosynthesis with the uptake
of extracellular proline. In contrast, proline addition increased the acetate secretion
rate by almost 25%, in the wt_pET strain, reversed the flux between ?-KG and glutamate towards ?-KG synthesis, increased the TCA flux from ?-KG to succinate by almost 40%, and activated the conversion of malate to pyruvate.
In all other cases, independently of proline addition, this gluconeogenetic reaction
was not active. The oxidative decarboxylation of malate catalyzed by the malic enzymes
ScfA and/or MaeB 19] has previously been considered absent in E. coli cells growing on glucose 20]. Such a flux through malic enzymes, for which simulations gave evidence only in the
case of the wt_pET strain grown on glucose and proline, was not identified in our
earlier study, where the pools of malate and oxaloacetate in the TCA cycle and the
pools of pyruvate and phosphoenolpyruvate in glycolysis were lumped 11]. Accordingly, proline addition was suggested to reduce the anaplerotic fluxes. In
order to assess whether the malate decarboxylation indeed takes place in vivo, a tracer
experiment using 100% U-
13
C labeled glucose was performed and the fraction of unlabeled alanine derived from
proline via malate and pyruvate was quantified. When proline was added, the unlabeled
alanine m0 fraction of the 260, 232, and 158 [m/z] fragment ions doubled from 1.8, 1.8, and 2.2% to 3.8, 3.4, and 4.0%, respectively
(see Additional file 1: Table S4). This experimentally determined increase, reaching on average approximately
1.8%, is in agreement with the 2.1% calculated theoretically from the metabolic fluxes
presented in Fig. 2, confirming active flux via malic enzymes. The role of malic enzymes is not clear
during growth on glucose, since the combined activities of pyruvate carboxylase, malate
dehydrogenase, and malic enzymes result in net ATP consumption and can therefore be
regarded as parts of a futile cycle. However, it is generally accepted that the flux
from malate to pyruvate produces NADPH and thus functions as NADPH generator for biosynthetic
purposes when E. coli grows on substrates that do not make use of glycolysis to enter central metabolism
(e.g., acetate, C4-dicarboxylic acids, amino acids) 19], 21], 22]. Moreover, as phosphoenolpyruvate carboxykinase and the malic enzyme(s) may be responsible
for the withdrawal of C4- and C5-intermediates from the TCA cycle, they might fulfil
a cataplerotic function 23]. Thus, the addition of proline to the wt_pET strain, accompanied by the increased
NADH generating flux from a-KG towards succinate and malate, may have led to malic enzyme activation, pyruvate
surplus, and finally a higher acetate excretion rate.
In the absence of proline, p4h1 expression and the resulting proline hydroxylation lead to a doubling of the proline
synthesis rate in both strains. Additionally, recombinant P4H production was associated
with a metabolic burden as reflected by higher relative TCA fluxes in both wt_p4h1of
and ?putA_p4h1of indicating increased biosynthetic and energy demands. When proline was added
to the medium, the same effect was observed for wt_p4h1of, but not for ?putA_p4h1of which retained similar relative TCA cycle fluxes as ?putA_pET. Another difference between wt_p4h1of and ?putA_p4h1of is the anaplerotic net flux between phosphoenolpyruvate and oxaloacetate.
Upon proline hydroxylation, this flux decreased by 30% in wt_p4h1of, implying an anaplerotic
role of proline metabolism, whereas it remained similar in ?putA_p4h1of, as can be expected from the inability of this strain to metabolize proline.
However, in wt_pET such an anaplerotic role of proline metabolism was not observed.
Instead, the flux through the malic enzymes was activated as described above. For
wt_p4h1of, the missing evidence for an activated malic enzyme flux can be explained
by a decreased rate of proline metabolism caused by proline hydroxylation and the
lower proline uptake rate.
In the presence of proline, the flux from ?-KG to succinate was decreased by 60% in ?putA_p4h1of compared to only 10% in wt_p4h1of, which can be ascribed to ?-KG withdrawal for hyp synthesis. Strikingly, even though putA deletion led to increased proline uptake and hydroxylation rates, it did not induce
a “driven-by-demand†increase of the ?-KG generating TCA flux, pointing towards a possible limitation of P4H catalysis by
the cosubstrate ?-KG. Moreover, the intracellular ?-KG concentrations determined for both strains (Fig. 3) show that even though the hyp formation rate is almost twice as high when using
the ?putA_p4h1of as compared to wt_p4h1of, P4H has to withdraw ?-KG from a pool of similar size competing for ?-KG with ?-KG dehydrogenase and glutamate dehydrogenase. These results indicate that the main
limitation of proline hydroxylation is shifted from intracellular proline to ?-KG availability and thus to reactions involved in ?-KG formation, such as oxidative isocitrate decarboxylation, rendering them promising
targets for future metabolic engineering efforts.
Fig. 3. Intracellular a-ketoglutarate (a-KG) and succinate (Succ) concentrations in the mid-exponential growth phase for recombinant
E. coli BL21(DE3)(pLysS) (wt) and E. coli BL21?putA(DE3)(pLysS) (?putA) containing pET-24a (pET) or pET_p4h1of (p4h1of) in M9 medium with 5 g L
?1
glucose (Glc) in the absence or presence of 5 mM proline (Pro) at 30°C. Concentrations
are given as average values in µmoles per gram cell dry weight from three different
samples with standard deviations displayed as error bars.
Decreased maintenance energy demands upon hyp synthesis
Even though stoichiometric modelling provides a precious basis to elucidate how cellular
metabolism functions and responds to perturbations, it does not directly offer a deeper
insight in the way bacterial cells balance catabolic energy generation with anabolic
demands, including redox pools. Thus, based on the metabolic fluxes derived from
13
C-MFA, net formation rates of NTPs (ATP and GTP), NAD(P)H, and FADH
2
were determined (Table 3, See also Additional file 1: Table S5). The rate of NTP formation via substrate-level phosphorylation was calculated
based on glycolysis, TCA cycle, and acetic acid formation rates. The ATP generating
pathway from acetyl-CoA to acetate catalyzed by phosphotransacetylase and acetate
kinase was considered, assuming that the flux through pyruvate oxidase B (PoxB) plays
a minor role 24]. The growth-related demand for energy and redox equivalents was calculated using
a stoichiometric equation for biomass formation generated by the FiatFlux software
(see Additional file 2) 25]–27]. The equation contains the specific growth rate (µ) as a variable, thereby considering
the differing macromolecular biomass composition at different growth rates 28]. The rate of ATP formation via oxidative phosphorylation was calculated based on
the assumptions that (1) the NADH and FADH
2
not used for biomass formation are utilized for ATP formation at the maximum P/O ratio
(NADH ? 3ATP, FADH
2
 ? 2ATP) 29], and (2) only NADH and not NADPH is oxidized by the respiratory chain 30].
Table 3. Effect of putA deletion, p4h1of expression, proline addition, and proline hydroxylation on NTP and cofactor formation
and consumption rates
Substrate-level phosphorylation occurred at similar rates with all strains and under
all conditions tested except for ?putA_p4h1of which displayed a significantly lower rate in the presence of proline (Table 3). This low NTP formation rate can be explained by the decreased flux from ?-KG to succinate upon P4H catalysis at high rates. The NTP consumed by the pET containing
strains for cell growth was at similar levels, except for wt_pET in the presence of
proline which showed a higher NTP demand due to increased biomass formation resulting
from proline assimilation.
Upon P4H synthesis, the NTP amounts invested for cell growth on glucose decreased
by 26 and 30% for the wt_p4h1of and ?putA_p4h1of, respectively. The same phenomenon was observed when proline was added to
the medium, i.e., 33 and 30% decrease of cell growth-associated NTP consumption, respectively.
The metabolic burden of P4H production imposed on the cells when grown on glucose
as sole source of carbon was also manifested by an increased respiratory activity
that led to 18 and 10% higher NTP formation rates in wt_p4h1of and ?putA_p4h1of, respectively, and this at a lower growth rate 31], 32]. Accordingly, enhanced maintenance requirements were observed during recombinant
protein production as a non-growth associated function that is in agreement with other
published results 32], 33]. This situation changed when proline was added to the medium. In ?putA_p4h1of, the NADH generating flux from ?-KG to succinate was partially replaced by the biotransformation and, in wt_p4h1of,
less proline was converted to glutamate, both resulting in decreased NADH/FADH
2
formation and thus reduced oxidative phosphorylation. This effect was more prominent
for ?putA_p4h1of.
ATP produced and not consumed for growth-related production of cellular material typically
is consumed for maintenance processes such as maintenance of electrochemical gradients
across the plasma membrane that can reach up to 50% of the ATP produced, degradation
and regeneration of cellular macromolecules, futile cycles, energy spilling reactions,
proofreading, and cell motility 34], 35]. Based on the calculations performed for ?putA_p4h1of, the most noteworthy bioenergetic alteration caused by P4H activity consists
in the dramatic decrease of NADH available for oxidative phosphorylation. Most interestingly,
it seems that, upon hyp synthesis, the engineered strain responded by optimizing its
energetic efficiency by reducing its maintenance energy demand instead of increasing
the TCA activity. This is especially remarkable as recombinant p4h1of expression itself had the opposite effect increasing the energy demand. Taking into
account that the stoichiometry of energy transducing membranes is not fixed and that
the maximum P/O ratios assumed in Table 3 will not be reached in reality, the engineered strain may possibly utilize glucose
more efficiently by exhibiting a higher in vivo P/O ratio.