Automated Measurement and Monitoring of Bioprocesses: Key Elements of the M3C Strategy - Measurement, Monitoring, Modelling, and Control of Bioprocesses - page 15

change very rapidly, in seconds to minutes for the entire population in a bioreactor.
Since only the reduced forms fluoresce, the sensor signal reflects the (inner) redox
state of a population rather than the biomass concentration. If this fact is not
absolutely clear, the user of this sensor will eventually be completely misled. The
commercial solution to this dilemma was withdrawal of the sensor from the
market.
Meanwhile, there are other commercially available sensors that allow the
monitoring of other fluorophores inside the cells, for instance, pyridoxins, flavines,
or various fluorescent proteins [ 48 - 51 ]. Skibsted et al. [ 52 ] scanned P. fluorescens
cultivations (270-550 nm excitation, 310-590 nm emission) and related the
spectral information to other process variables using partial least-squares regres-
sion models. From that, they concluded that this sensor could also predict—with
the model, of course—the states of nonfluorescent compounds such as nitrate and
succinate. Haack et al. [ 53 ] estimated the cell mass concentration of baker's yeast
grown in a defined medium from 2D fluorescence spectra including tryptophan,
NAD(P)H, and riboflavin using a chemometric model. The method was also
reported to be used online during downstream processing [ 54 ].
Since the in situ sensors always pick up the excitation light as scattered (or
reflected) light, it is worth considering the use of fluorescent reporter proteins that
have a maximal separation between excitation and emission wavelength, for
instance, wild-type GFP (395/509 nm) rather than the enhanced variant (eGFP:
484/507 nm), in order to achieve a reasonable separation of the desired signal from
the unwanted stray light band, a notorious disturbance.
3.2.4 Interfacing Axenic Bioprocesses to Process Analyzers
The techniques described thus far either exploit (heat or mass) balancing techniques
or consist of in situ sensors that tolerate sterilization. Hence, no problem of com-
promising the sterile barrier is to be expected. The use of process analyzers, how-
ever, requires an interface to the process that must—in most cases of scientifically
and industrially exploited processes with pure cultures—secure the axenic state of
the process. In many cases, this interface will also have the task of defoaming and
degassing the sample aliquot removed to feed the process analyzer, simply because
this is practically always a volumetric dosage. Defoaming and degassing require
some nonnegligible hydraulic residence time, independent of whether a small settler
(the sample being sucked out of its core volume after an appropriate batch time for
degassing) or a gas-permeable membrane (over which the sample flows) is used. An
extremely decisive aspect is the transport of the sample through the interface into the
process analyzer: the biocatalysts will continue to metabolize en route and can so
falsify the results obtained after the finite transfer time, even if the analyzer is error
free. In this context, the use of an autosampler (and a multifunctional analyzer) is
questionable, since the connection to the bioprocess is quite long (see, e.g., [ 55 ],
where the 1-mm capillary is 3 m long).