Metabolic fluxes (i.e. in vivo reaction rates) represent the properties of cellular regulation and are
most suitable for characterizing specific phenotypes of microorganisms. As intracellular fluxes are
not directly measurable they have to be estimated from measured quantities through model-based
evaluation with the aid of computational routines based on13C labeling experiments (13C-MFA).
However, the question might be raised about the biological significance and statistical reliability of
those flux estimations based on single experiments. In addition the medium throughput approach
of13C-MFA leads to a multiplied amount of data, so the used algorithms for data pre-processing
and evaluation of results has to be optimized to increase the efficiency.
Seeking to answer the question and to adopt the challenge of medium throughput13C-MFA a se-
ries of standardized13C-labeling experiments with the model organism Corynebacterium glutami-
cum wildtype (WT) and a lysine producer (LP) was carried out under well-controlled conditions:
continuous cultivation mode with dilution rates of 0.20h−1, 0.15h−1, 0.10h−1and 0.05h−1was
chosen to ensure metabolic and isotopic stationarity. For increasing throughput and minimizing
13C labeling costs a parallel bioreactor setup at small scale (300mL working volume) was uti-
lized. Hence, all experiments were performed in fourfold biological replicates for calculation of
extracellular rates, i.e. substrate uptake and (by-) product formation, including corresponding stan-
dard error. By online FT-IR spectrometry discrimination between13C and12C carbon dioxide was
realized. From each single experiment six technical samples were taken and labeling patterns of
intracellular metabolites were analyzed by LC-MS/MS technology.
To realise data evaluation, processing and storage of such experiments an application-oriented
software package 13CFLUX2-Essentials was developed. The package offers an environment
to handle the huge amount of data, e.g. the graphically aided adjustment and management of
analytical raw data or different functions for comparing results. Calculation of extracellular rates
as well as carbon balances, network modeling, parallel multi-start simulations, evaluation and
visualization of results is automatically performed. Thus,13C-MFA in a medium throughput is
possible now and was developed and used in this work.
From the 20 datasets intracellular fluxes were estimated using the 13CFLUX2 software package.
Comprehensive statistical analysis of error propagation was performed to investigate the influence
of analytical and experimental errors on the determinacy of all fluxes. Label measurements of
intracellular metabolites showed different reproducibility between technical and biological repli-
cations. Carbon balances were closed between 73–102% with an average about 86%. Beside the
extracellulare rates ~66 labeling fragments were included for parameter fitting. 5–7 metabolites
were rejected due to analytical errors. With each dataset 10000 single fittings were performed in
an multi start optimization (MSO). Differences between biological replicas were caused by va-
riations in the fractional labeling of intermediates. 12 out of 20 datasets showed unique global
optimal solutions in the MSO results.
The studies in a small-scale system discover bottle necks and points for improvement for further
investigations. It is demonstrated that multiple experimental runs distinctively increases reliability
and, in turn, the potential to generate meaningful fluxome data. Hence, these results strengthen the
position of metabolic flux analysis to be an effective diagnostic tool for metabolic engineering.