Critical evaluation of models developed for monitoring an industrial submerged bioprocess for antibiotic production using near-infrared spectroscopy

Near-infrared spectroscopy (NIRS) is known to have potential for cost-effective monitoring of bioprocesses. Although this has been demonstrated in many instances and several models have been reported, information regarding the complexity of models required and their utility over extended periods of time is lacking. In the present study, the complexity of the models required for the NIRS prediction of substrate (oil) and product (tylosin) concentration in an industrial bioprocess that employs a physicochemically heterogeneous medium for antibiotic production was assessed. Measurements made by both the diffuse reflectance and transmittance modes were investigated. SEP values for the prediction of the analytes averaged 5% or less, for the successful models, when evaluated using an external validation set, 2 years after the initial model development exercise. Diffuse reflectance measurements showed poorer results, compared to transmittance measurements, especially for monitoring tylosin. In general, this investigation provides evidence to support the fact that models built for the prediction of analytes in a commercial bioprocess that employs a physicochemically complex production medium can be robust in performance over an extended period of time and that simple models based on fewer terms or latent variables can perform well, even in the context of matrices that are relatively complex. It also indicates that sample presentation is likely to be a critical factor in the successful application of NIRS in bioprocess monitoring, which merits further detailed investigation.     

Monitoring a high cell density recombinant Pichia pastoris fed-batch bioprocess using transmission and reflectance near infrared spectroscopy

The use of near infrared spectroscopy (NIRS) was investigated in the context of an efficient high cell density fed-batch industrial Pichia pastoris bioprocess for the production of a therapeutic mammalian protein. This process represented a considerable challenge from the viewpoint of using NIRS to model key analytes because it involved two carbon sources (glycerol and methanol) added at differing rates and times, used a chemically complex medium, and showed a change in liquid phase behaviour due to cell growth. Models for biomass, glycerol, methanol and product were constructed. Different methods of spectral collection and mathematical procedures were used relative to which analyte in the fermentation matrix was being modelled and the rationale behind the model building is clearly described. Regardless of the mode of spectral collection it was essential to consider the changes in modelled analyte concentration relative to changes in other spectral contributors (analytes). The study considerably extends the use of NIRS in fermentation processes to high cell density complex industrial production processes, and comments on how this further developments the technology towards routine in situ NIRS monitoring of bioprocesses.     

In-situ near infrared spectroscopy to monitor key analytes in mammalian cell cultivation

The use of in-situ near infrared spectroscopy (NIRS) as a tool for monitoring four key analytes in a CHO-K1 animal cell culture was investigated. Previous work using on-line NIRS to monitor bioprocesses has involved its application ex-situ where the analyzer is physically outside the fermentor, or to microbial bioprocesses. This novel application of NIRS to monitor analytes within an animal cell culture using a steam sterilizable in-situ fiber optic probe is very important for furthering the use of NIRS within the bioprocessing industry. The method of calibration used to develop the models involved the use of large data sets so that all likely variation in stoichiometry was incorporated within the models. Successful models for glucose, lactate, glutamine, and ammonia were built with Standard Error of Predictions (SEP's) of 0.072 (g/L), 0.0144 (g/L), 0.308 (mM), and 0.036 (mM), respectively of the total concentration range.     

Temporally segmented modelling: a route to improved bioprocess monitoring using near infrared spectroscopy?

Near infrared spectroscopy (NIRS) was used to monitor an industrial bioprocess for the production of the antibiotic, tylosin, using a segmented modelling approach. Models were built over the entire time course of the fermentation from 0 to 150 h, and also in two distinct phases or segments of the bioprocess from 50 to 100 h (synthetic phase) and from 100 to 150 h (stationary phase). All models were validated externally and the performance of the full range and segmented models compared. The standard error of prediction (SEP) of the segmented models was less in both 50–100 h and 100–150 h and the correlation highest in the 50–100 h range. This would suggest that data segmentation is potentially a useful method of accommodating the impact of the pronounced matrix changes which occur in some bioprocesses in NIRS models for key analytes. While there are many reports on bioprocess monitoring using NIRS, there have been no previous studies on the use of segmented NIR models within a bioprocess as a means of accommodating matrix change.     

Use of at-line and in-situ near-infrared spectroscopy to monitor biomass in an industrial fed-batch Escherichia coli process

One of the key goals in bioprocess monitoring is to achieve real-time knowledge of conditions within the bioreactor, i.e., in-situ. Near-infrared spectroscopy (NIRS), with its ability to carry out multi-analyte quantification rapidly with little sample presentation, is potentially applicable in this role. In the present study, the application of NIRS to a complex, fed-batch industrial E. coli (RV308/PHKY531) process was investigated. This process undergoes a series of temperature changes and is vigorously agitated and aerated. These conditions can pose added challenges to in-situ NIRS. Using the measurement of a key analyte (biomass) as an illustration, the details of the relationship between the at-line and in-situ use of NIRS are considered from the viewpoint of both theory and practical application. This study shows that NIRS can be used both at-line and in-situ in order to achieve good predictive models for biomass. There are particular challenges imposed by in-situ operation (loss of wavelength regions and noise) which meant the need for signal optimisation studies. This showed that whilst the at-line modelling process may provide some useful information for the in-situ process, there were distinct differences. This study shows that the in-situ use of NIRS in a highly challenging matrix (similar to those encountered in current industrial practice) is possible, and thus extends previous works in the area.     

Monitoring complex media fermentations with near-infrared spectroscopy: comparison of different variable selection methods

Near-infrared spectroscopy (NIRS) is known to be a suitable technique for rapid fermentation monitoring. Industrial fermentation media are complex, both chemically (ill-defined composition) and physically (multiphase sample matrix), which poses an additional challenge to the development of robust NIRS calibration models. We investigated the use of NIRS for at-line monitoring of the concentration of clavulanic acid during an industrial fermentation. An industrial strain of Streptomyces clavuligerus was cultivated at 200-L scale for the production of clavulanic acid. Partial least squares (PLS) regression was used to develop calibration models between spectral and analytical data. In this work, two different variable selection methods, genetic algorithms (GA) and PLS-bootstrap, were studied and compared with models built using all the spectral variables. Calibration models for clavulanic acid concentration performed well both on internal and external validation. The two variable selection methods improved the predictive ability of the models up to 20%, relative to the calibration model built using the whole spectra.     

Hybrid-EKF: Hybrid model coupled with extended Kalman filter for real-time monitoring and control of mammalian cell culture

In a decade when Industry 4.0 and quality by design are major technology drivers of biopharma, automated and adaptive process monitoring and control are inevitable requirements and model-based solutions are key enablers in fulfilling these goals. Despite strong advancement in process digitalization, in most cases, the generated datasets are not sufficient for relying on purely data-driven methods, whereas the underlying complex bioprocesses are still not completely understood. In this regard, hybrid models are emerging as a timely pragmatic solution to synergistically combine available process data and mechanistic understanding. In this study, we show a novel application of the hybrid-EKF framework, that is, hybrid models coupled with an extended Kalman filter for real-time monitoring, control, and automated decision-making in mammalian cell culture processing. We show that, in the considered application, the predictive monitoring accuracy of such a framework improves by at least 35% when developed with hybrid models with respect to industrial benchmark tools based on PLS models. In addition, we also highlight the advantages of this approach in industrial applications related to conditional process feeding and process monitoring. With regard to the latter, for an industrial use case, we demonstrate that the application of hybrid-EKF as a soft sensor for titer shows a 50% improvement in prediction accuracy compared with state-of-the-art soft sensor tools.     

Real-time monitoring and control of microbial bioprocesses with focus on the specific growth rate: current state and perspectives

Understanding the growth characteristics of microorganisms is an essential step in bioprocessing, not only because product formation may be growth-associated but also because they might influence cell physiology and thereby product quality. The specific growth rate, a key variable of many bioprocesses, cannot be measured directly and relies on the estimation through other measurable variables such as biomass, substrate, or product concentrations. Techniques for real-time estimation of the specific growth rate in microbial fed-batch cultures are discussed in the present paper. The advantages and limitations of different models and various monitoring techniques are discussed, highlighting the importance of the specific growth rate in the development of fast, reliable, and robust processes for the production of high-value products such as recombinant proteins.     

Precision fermentation with mass spectrometry-based spent media analysis

Optimization and monitoring of bioprocesses requires the measurement of several process parameters and quality attributes. Mass spectrometry (MS)-based techniques such as those coupled to gas chromatography (GCMS) and liquid Chromatography (LCMS) enable the simultaneous measurement of hundreds of metabolites with high sensitivity. When applied to spent media, such metabolome analysis can help determine the sequence of substrate uptake and metabolite secretion, consequently facilitating better design of initial media and feeding strategy. Furthermore, the analysis of metabolite diversity and abundance from spent media will aid the determination of metabolic phases of the culture and the identification of metabolites as surrogate markers for product titer and quality. This review covers the recent advances in metabolomics analysis applied to the development and monitoring of bioprocesses. In this regard, we recommend a stepwise workflow and guidelines that a bioprocesses engineer can adopt to develop and optimize a fermentation process using spent media analysis. Finally, we show examples of how the use of MS can revolutionize the design and monitoring of bioprocesses.     

A hybrid representation approach for modelling complex dynamic bioprocesses

This paper considers the use of hybrid models to represent the dynamic behaviour of biotechnological processes. Each hybrid model consists of a set of non linear differential equations and a neural model. The set of differential equations attempts to describe as much as possible the phenomenology of the process whereas neural networks model predict some key parameters that are an essential part of the phenomenological model. The neural model is obtained indirectly, that is, using the prediction errors of one or more state variables to adjust its weights instead of successive presentations of input-output data of the neural network. This approach allows to use actual measurements to derive a suitable neural model that not only represents the variation of some key parameters but it is also able to partly include dynamic behaviour unaccounted for by the phenomenological model. The approach is described in detail using three test cases: (1) the fermentation of glucose to gluconic acid by the micro-organism Pseudomonas ovalis, (2) the growth of filamentous fungi in a solid state fermenter, and (3) the propagation of filamentous fungi growing on a 2-D solid substrate. Results for the three applications clearly demon- strate that using a hybrid model is a viable alternative for modelling complex biotechnological bioprocesses.     

Switching industrial production processes from complex to defined media: method development and case study using the example of Penicillium chrysogenum

ABSTRACT: BACKGROUND: Filamentous fungi are versatile cell factories and widely used for the production of antibiotics, organic acids, enzymes and other industrially relevant compounds at large scale. As a fact, industrial production processes employing filamentous fungi are commonly based on complex raw materials. However, considerable lot-to-lot variability of complex media ingredients not only demands for exhaustive incoming components inspection and quality control, but unavoidably affects process stability and performance. Thus, switching bioprocesses from complex to defined media is highly desirable. RESULTS: This study presents a strategy for strain characterization of filamentous fungi on partly complex media using redundant mass balancing techniques. Applying the suggested method, interdependencies between specific biomass and side-product formation rates, production of fructooligosaccharides, specific complex media component uptake rates and fungal strains were revealed. A 2-fold increase of the overall penicillin space time yield and a 3-fold increase in the maximum specific penicillin formation rate were reached in defined media compared to complex media. CONCLUSIONS: The newly developed methodology enabled fast characterization of two different industrial Penicillium chrysogenum candidate strains on complex media based on specific complex media component uptake kinetics and identification of the most promising strain for switching the process from complex to defined conditions. Characterization at different complex/defined media ratios using only a limited number of analytical methods allowed maximizing the overall industrial objectives of increasing both, method throughput and the generation of scientific process understanding.     

Comparison of physics-based and data-driven modelling techniques for dynamic optimisation of fed-batch bioprocesses

The development of digital bioprocessing technologies is critical to operate modern industrial bioprocesses. This study conducted the first investigation on the efficiency of using physics-based and data-driven models for the dynamic optimisation of long-term bioprocess. More specifically, this study exploits a predictive kinetic model and a cutting-edge data-driven model to compute open-loop optimisation strategies for the production of microalgal lutein during a fed-batch operation. Light intensity and nitrate inflow rate are used as control variables given their key impact on biomass growth and lutein synthesis. By employing different optimisation algorithms, several optimal control sequences were computed. Due to the distinct model construction principles and sophisticated process mechanisms, the physics-based and the data-driven models yielded contradictory optimisation strategies. The experimental verification confirms that the data-driven model predicted a closer result to the experiments than the physics-based model. Both models succeeded in improving lutein intracellular content by over 40% compared to the highest previous record; however, the data-driven model outperformed the kinetic model when optimising total lutein production and achieved an increase of 40–50%. This indicates the possible advantages of using data-driven modelling for optimisation and prediction of complex dynamic bioprocesses, and its potential in industrial bio-manufacturing systems.     

Coupled metabolic-hydrodynamic modeling enabling rational scale-up of industrial bioprocesses

Metabolomics aims to address what and how regulatory mechanisms are coordinated to achieve flux optimality, different metabolic objectives as well as appropriate adaptations to dynamic nutrient availability. Recent decades have witnessed that the integration of metabolomics and fluxomics within the goal of synthetic biology has arrived at generating the desired bioproducts with improved bioconversion efficiency. Absolute metabolite quantification by isotope dilution mass spectrometry represents a functional readout of cellular biochemistry and contributes to the establishment of metabolic (structured) models required in systems metabolic engineering. In industrial practices, population heterogeneity arising from fluctuating nutrient availability frequently leads to performance losses, that is reduced commercial metrics (titer, rate, and yield). Hence, the development of more stable producers and more predictable bioprocesses can benefit from a quantitative understanding of spatial and temporal cell-to-cell heterogeneity within industrial bioprocesses. Quantitative metabolomics analysis and metabolic modeling applied in computational fluid dynamics (CFD)-assisted scale-down simulators that mimic industrial heterogeneity such as fluctuations in nutrients, dissolved gases, and other stresses can procure informative clues for coping with issues during bioprocessing scale-up. In previous studies, only limited insights into the hydrodynamic conditions inside the industrial-scale bioreactor have been obtained, which makes case-by-case scale-up far from straightforward. Tracking the flow paths of cells circulating in large-scale bioreactors is a highly valuable tool for evaluating cellular performance in production tanks. The “lifelines” or “trajectories” of cells in industrial-scale bioreactors can be captured using Euler-Lagrange CFD simulation. This novel methodology can be further coupled with metabolic (structured) models to provide not only a statistical analysis of cell lifelines triggered by the environmental fluctuations but also a global assessment of the metabolic response to heterogeneity inside an industrial bioreactor. For the future, the industrial design should be dependent on the computational framework, and this integration work will allow bioprocess scale-up to the industrial scale with an end in mind.     

Assessment of near-infrared spectral information for rapid monitoring of bioprocess quality.

Access to real-time process information is desirable for consistent and efficient operation of bioprocesses. Near-infrared spectroscopy (NIRS) is known to have potential for providing real-time information on the quantitative levels of important bioprocess variables. However, given the fact that a typical NIR spectrum encompasses information regarding almost all the constituents of the sample matrix, there are few case studies that have investigated the spectral details for applications in bioprocess quality assessment or qualitative bioprocess monitoring. Such information would be invaluable in providing operator-level assistance on the progress of a bioprocess in industrial-scale productions. We investigated this aspect and report the results of our investigation. Near-infrared spectral information derived from scanning unprocessed culture fluid (broth) samples from a complex antibiotic production process was assessed for a data set that incorporated bioprocess variations. Principal component analysis was applied to the spectral data and the loadings and scores of the principal components studied. Changes in the spectral information that corresponded to variations in the bioprocess could be deciphered. Despite the complexity of the matrix, near-infrared spectra of the culture broth are shown to have valuable information that can be deconvoluted with the help of factor analysis techniques such as principal component analysis (PCA). Although complex to interpret, the loadings and score plots are shown to offer potential in process diagnosis that could be of value in the rapid assessment of process quality, and in data assessment prior to quantitative model development.     

Multilayer adaptive control of continuous bioprocesses using optimising control technique. Case study: Bakers' yeast culture

High costs associated with many fermentation processes in an increasingly competitive industry make any prompt application of modern control techniques to industrial bioprocesses very desirable. However, this is often hampered by the lack of adequate mathematical models, on the one hand, and by the absence of continuous, on-line measurement of the most relevant process variables, on the other hand. This paper addresses these problems and offers a new strategy to control continuous bioprocesses using a hierarchical structure such that neither structured process models nor continuous measurement of all relevant variables have to be available. The control system consists of two layers. The lower layer represents a dynamic adaptive follow-up control of a continuously measured output — in our case dissolved oxygen concentration. This variable is supposed to be strongly correlated with the key output variable — in our case cellular concentration which is not continuously available for measurement. The higher layer is then designed to maintain a desired profile of the process key output using a set-point optimising control technique. The Integrated System Optimisation and Parameter Estimation method used operates on an appropriately chosen steady-state performance criterion. A prerequisite for successful application of the proposed approach is an approximate steady-state model, describing the relationship between the measured output and the process key output variable. Furthermore, occasional in situ, off-line or laboratory measurement values of the key output variable are needed. Promising simulation results of the biomass concentration control, by manipulating the air flow-rate in the continuous bakers' yeast culture are presented.     

Bioprocess monitoring by marker gene analysis

The optimization and the scale up of industrial fermentation processes require an efficient and possibly comprehensive analysis of the physiology of the production system throughout the process development. Furthermore, to ensure a good quality control of established bioprocesses, on-line analysis techniques for the determination of marker gene expression are of interest to monitor the productivity and the safety of bioprocesses. A prerequisite for such analyses is the knowledge of genes, the expression of which is critical either for the productivity or for the performance of the bioprocess. This work reviews marker genes that are specific indicators for stress- and nutrient-limitation conditions or for the physiological status of the bacterial production hosts Bacillus subtilis, Bacillus licheniformis and Escherichia coli. The suitability of existing gene expression analysis techniques for bioprocess monitoring is discussed. Analytical approaches that enable a robust and sensitive determination of selected marker mRNAs or proteins are presented.     

Genome shuffling of Lactobacillus for improved acid tolerance

Fermentation-based bioprocesses rely extensively on strain improvement for commercialization. Whole-cell biocatalysts are commonly limited by low tolerance of extreme process conditions such as temperature, pH, and solute concentration. Rational approaches to improving such complex phenotypes lack good models and are especially difficult to implement without genetic tools. Here we describe the use of genome shuffling to improve the acid tolerance of a poorly characterized industrial strain of Lactobacillus. We used classical strain-improvement methods to generate populations with subtle improvements in pH tolerance, and then shuffled these populations by recursive pool-wise protoplast fusion. We identified new shuffled lactobacilli that grow at substantially lower pH than does the wild-type strain on both liquid and solid media. In addition, we identified shuffled strains that produced threefold more lactic acid than the wild type at pH 4.0. Genome shuffling seems broadly useful for the rapid evolution of tolerance and other complex phenotypes in industrial microorganisms.     

Process analytical technology tools for perfusion cell culture

During cell cultivation processes for the production of biopharmaceuticals, good process performance and good product quality can be ensured by online monitoring of critical process parameters (e.g. temperature, pH, or dissolved oxygen). These data can be used in real‐time for process control, as suggested by the process analytical technology (PAT) initiative. Today, solutions for real‐time monitoring of parameters such as concentrations of cells, main nutrients, and metabolism by‐products are developing, but applications of these more complex tools in industrial settings are still limited. Here, we evaluated the use of dielectric spectroscopy (DS) and near‐infrared spectroscopy (NIRS) as PAT tools for a perfusion PER.C6^® cultivation process. We showed that DS enabled predictions of viable cell density from the cultivation vessel, with a root mean square error of prediction (RMSEP) of 4.4% of the calibration range. Additionally, predictions of glucose and lactate concentrations from the cultivation vessel (RMSEP of 10 and 14%, respectively) and from the perfusion stream (RMSEP of 12 and 10%, respectively) were achieved with NIRS. We also showed that the perfusion stream offers great opportunities for noninvasive, yet frequent process monitoring. Accurate online monitoring of critical process parameters with PAT tools is the essential first step toward increased control of process output.