CoChat

Metabolic Control in Mammalian Fed-Batch Cell Cultures for Reduced Lactic Acid Accumulation and Improved Process Robustness

Open access

The DOI matches its Crossref record; 100 of the 111 references it cites were found in a registry or an index.

Abstract

Biomass and cell-specific metabolic rates usually change dynamically over time, making the “feed according to need” strategy difficult to realize in a commercial fed-batch process. We here demonstrate a novel feeding strategy which is designed to hold a particular metabolic state in a fed-batch process by adaptive feeding in real time. The feed rate is calculated with a transferable biomass model based on capacitance, which changes the nutrient flow stoichiometrically in real time. A limited glucose environment was used to confine the cell in a particular metabolic state. In order to cope with uncertainty, two strategies were tested to change the adaptive feed rate and prevent starvation while in limitation: (i) inline pH and online glucose concentration measurement or (ii) inline pH alone, which was shown to be sufficient for the problem statement. In this contribution, we achieved metabolic control within a defined target range. The direct benefit was two-fold: the lactic acid profile was improved and pH could be kept stable. Multivariate Data Analysis (MVDA) has shown that pH influenced lactic acid production or consumption in historical data sets. We demonstrate that a low pH (around 6.8) is not required for our strategy, as glucose availability is already limiting the flux. On the contrary, we boosted glycolytic flux in glucose limitation by setting the pH to 7.4. This new approach led to a yield of lactic acid/glucose (Y L/G) around zero for the whole process time and high titers in our labs. We hypothesize that a higher carbon flux, resulting from a higher pH, may lead to more cells which produce more product. The relevance of this work aims at feeding mammalian cell cultures safely in limitation with a desired metabolic flux range. This resulted in extremely stable, low glucose levels, very robust pH profiles without acid/base interventions and a metabolic state in which lactic acid was consumed instead of being produced from day 1. With this contribution, we wish to extend the basic repertoire of available process control strategies, which will open up new avenues in automation technology and radically improve process robustness in both process development and manufacturing.

Overview

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.3390/bioengineering3010005

Cites

100 of 100 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 1 Automatic control strategy using exclusively real-time available signals. Constants (safety factors and the biomass model) were defined a priori which led to either a rather loose or very strict adherence to a desired qs set-point.
Figure 1 Automatic control strategy using exclusively real-time available signals. Constants (safety factors and the biomass model) were defined a priori which led to either a rather loose or very strict adherence to a desired qs set-point.CC BY · PubMed Central
Figure 2 MVDA of the historical data set. (A) removal of outliers prior to analysis; (B) PLS-R model fit using two Principal Components, here Comp 1 and 2. R2Y(cum) indicates the explained variance, while Q2(cum) explains the predictive quality using one resp. two principal components; (C) VIP plot of the most important variables in the analysis sorted by relevance; (D) Loading scatter plot of data relationship between predictors (X) and predicted (Y) variable discriminated by VIP size.
Figure 2 MVDA of the historical data set. (A) removal of outliers prior to analysis; (B) PLS-R model fit using two Principal Components, here Comp 1 and 2. R2Y(cum) indicates the explained variance, while Q2(cum) explains the predictive quality using one resp. two principal components; (C) VIP plot of the most important variables in the analysis sorted by relevance; (D) Loading scatter plot of data relationship between predictors (X) and predicted (Y) variable discriminated by VIP size.CC BY · PubMed Central
Figure 3 Historical LAC profiles (A); Historical qLac profiles (B).
Figure 3 Historical LAC profiles (A); Historical qLac profiles (B).CC BY · PubMed Central
Figure 4 Historical qGlc profiles (A); On-line pH profile (B); Clustering of qGlc with a pH classifier (dark: low pH, bright: high pH) over time (C) and as box plot (D). A robust third-order polynomial function was applied in Graphpad Prism software to capture the general trend of a typical qGlc profile in (A) and qGlc in (C). No runs started with pH < 6.9 as pH initially falls and is then controlled.
Figure 4 Historical qGlc profiles (A); On-line pH profile (B); Clustering of qGlc with a pH classifier (dark: low pH, bright: high pH) over time (C) and as box plot (D). A robust third-order polynomial function was applied in Graphpad Prism software to capture the general trend of a typical qGlc profile in (A) and qGlc in (C). No runs started with pH < 6.9 as pH initially falls and is then controlled.CC BY · PubMed Central
Figure 5 Historical yield qLac/qGlc (Y L/G) versus qGlc showing the consequence of particular qGlc on the yield.
Figure 5 Historical yield qLac/qGlc (Y L/G) versus qGlc showing the consequence of particular qGlc on the yield.CC BY · PubMed Central
Figure 6 Experiment R-30, feeding profile with pH and GLC level (A); GLC and LAC concentrations (B); qGlc with trend and historical qGlc (C); overlay of resulting Y L/G profile with historical data (D).
Figure 6 Experiment R-30, feeding profile with pH and GLC level (A); GLC and LAC concentrations (B); qGlc with trend and historical qGlc (C); overlay of resulting Y L/G profile with historical data (D).CC BY · PubMed Central
Figure 7 Experiment R-31, feeding profile with pH and GLC level (A); GLC and LAC concentrations (note that leftover GLC is consumed and results in a small LAC until it is consumed, but qs changes immediately to the target value once GLC is limited by the stoichiometric feeding strategy) (B); qGlc (including dotted trend line) and historical qGlc (full line) (C); overlay of resulting Y L/G profile with historical data (D). Both targets (qGlc and resulting lower Y L/G) can be better achieved due to the tighter control specification, when compared with run R-30. A comparison follows in Table 3.
Figure 7 Experiment R-31, feeding profile with pH and GLC level (A); GLC and LAC concentrations (note that leftover GLC is consumed and results in a small LAC until it is consumed, but qs changes immediately to the target value once GLC is limited by the stoichiometric feeding strategy) (B); qGlc (including dotted trend line) and historical qGlc (full line) (C); overlay of resulting Y L/G profile with historical data (D). Both targets (qGlc and resulting lower Y L/G) can be better achieved due to the tighter control specification, when compared with run R-30. A comparison follows in Table 3.CC BY · PubMed Central
Figure 8 Benchmarking limited glucose control with historical data (black dots: historical data, squares: experiment R-30 with a broad qGlc set-point, triangles: experiment R-31 with a tight qGlc set-point. (A) and (B): metabolic profiles of glucose and lactic acid; (C) and (D): Specific glucose and lactic acid rates; (E) and (F): Viable cell concentration and product titer compared to historical data.
Figure 8 Benchmarking limited glucose control with historical data (black dots: historical data, squares: experiment R-30 with a broad qGlc set-point, triangles: experiment R-31 with a tight qGlc set-point. (A) and (B): metabolic profiles of glucose and lactic acid; (C) and (D): Specific glucose and lactic acid rates; (E) and (F): Viable cell concentration and product titer compared to historical data.CC BY · PubMed Central
Figure 9 Suggested flux dependency on productivity. A higher overall flux may lead to a higher flux in all metabolic nodes, including lactic acid secretion and the one responsible for product formation. Usually, the basic consumption of cells does not change much because pH is controlled tightly. However, by setting and holding physico-chemical conditions which act on this flux by a high pH, while still operating in a relatively low Y L/G range, the overall higher flux may be cumulatively translated into higher productivity.
Figure 9 Suggested flux dependency on productivity. A higher overall flux may lead to a higher flux in all metabolic nodes, including lactic acid secretion and the one responsible for product formation. Usually, the basic consumption of cells does not change much because pH is controlled tightly. However, by setting and holding physico-chemical conditions which act on this flux by a high pH, while still operating in a relatively low Y L/G range, the overall higher flux may be cumulatively translated into higher productivity.CC BY · PubMed Central
Figure 10 Metabolic control window with a suggested robust qGlc set-point. Data points represent both historical and experimental data. Depending on the task, the set-point range may be selected either conservatively (better overfeeding than starving, higher qGlc) or more courageously (improved metabolic state at the cost of reduced growth, lower qGlc). In general, the error of the online biomass estimation method may serve as a good first estimate for the range. It must be noted that the cells might not truly consume at a high qGlc just because it is desired by the operator and fed in this way, especially in the mid-phase of the culture—a fact which we suggest to solve by applying a high pH.
Figure 10 Metabolic control window with a suggested robust qGlc set-point. Data points represent both historical and experimental data. Depending on the task, the set-point range may be selected either conservatively (better overfeeding than starving, higher qGlc) or more courageously (improved metabolic state at the cost of reduced growth, lower qGlc). In general, the error of the online biomass estimation method may serve as a good first estimate for the range. It must be noted that the cells might not truly consume at a high qGlc just because it is desired by the operator and fed in this way, especially in the mid-phase of the culture—a fact which we suggest to solve by applying a high pH.CC BY · PubMed Central
Figure 11 Generic approach for the holistic improvement during mammalian cell culture development.
Figure 11 Generic approach for the holistic improvement during mammalian cell culture development.CC BY · PubMed Central
Figure A1 Historical GLC profiles. The GLC level itself was often inadequate to describe metabolic effects, such as the relation to qGlc in fed-batches. A robust third-order polynomial indicated an average trend of all runs, regardless of the individual trend in each cultivation.
Figure A1 Historical GLC profiles. The GLC level itself was often inadequate to describe metabolic effects, such as the relation to qGlc in fed-batches. A robust third-order polynomial indicated an average trend of all runs, regardless of the individual trend in each cultivation.CC BY · PubMed Central
Figure A2 MVDA coefficients of PC1 and PC2 capturing direction of parameter effect on the target variable (qLac). PC1 captures most variance in the following order: qGlc, H+, Lac, GLC, while PC2 captures some remaining variance in qGlc and Glc but not Lac and H+ as they are already well represented by PC1.
Figure A2 MVDA coefficients of PC1 and PC2 capturing direction of parameter effect on the target variable (qLac). PC1 captures most variance in the following order: qGlc, H+, Lac, GLC, while PC2 captures some remaining variance in qGlc and Glc but not Lac and H+ as they are already well represented by PC1.CC BY · PubMed Central
Figure A3 Viability trend in historical data, R-30 and R-31. The latter run had the highest viability at the process end. R-31 was tightly controlled throughout the process time in terms of metabolic state and featured a pH change from 7.4 to 7.2 between days 9 and 10.
Figure A3 Viability trend in historical data, R-30 and R-31. The latter run had the highest viability at the process end. R-31 was tightly controlled throughout the process time in terms of metabolic state and featured a pH change from 7.4 to 7.2 between days 9 and 10.CC BY · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 100 references

Cited by

See all 47

Metadata

Authors
Konakovsky, V., Clemens, C., Müller, M., Bechmann, J., Berger, M., Schlatter, S., & Herwig, C.
Affiliations
  • TU Wien (AT)
  • Boehringer Ingelheim (Germany) (DE)
Corresponding author
Herwig, Christoph
Venue
Bioengineering
Publisher
MDPI AG
Published
2016-01-11
Volume
3
Issue
1
Pages
5
ISSN
2306-5354 (ISSN-L)
OpenAlex
W2231307160
Cited by
44
References
111
Citations by year
2025: 3 · 2024: 3 · 2023: 7 · 2022: 5 · 2021: 3 · 2020: 2 · 2019: 5 · 2018: 9 · 2017: 7
FWCI
3.64
Type
journal article
Language
en
Open access
Locations
  • Bioengineering (publishedVersion, cc-by)
  • Europe PMC (PubMed Central) (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Topics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Viral Infectious Diseases and Gene Expression in Insects
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Microbial Metabolic Engineering and Bioproduction
  • Physical Sciences › Chemical Engineering › Fluid Flow and Transfer Processes › Fluid Dynamics and Mixing
Keywords
metabolic control, lactate accumulation, glucose limitation, pH control, glycolytic flux, process robustness, multivariate analysis, mammalian cell culture, process automation

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.