Model guided design of microbial consortia for mixed-culture fermentations

Lactococcus lactis and Lactococcus cremoris are among the best-studied lactic acid bacteria and are used extensively worldwide for the production of fermented dairy products. Starter cultures for dairy fermentations usually comprise of a consortium of different microorganisms and often contain several genetically different L. lactis and L. cremoris strains. L. lactis and L. cremoris strains play an important role in acidification, phage resilience, texture formation, gas production, and flavor development; therefore the strain balance in the culture is crucial. For the production of industrial microbial cultures for food fermentations, strains are usually produced individually and combined into the final culture. However, for cultures containing a larger number of different strains, this approach becomes unfeasible due to high production costs. If, instead, multiple strains could be produced together in a mixed-culture fermentation, while maintaining the strain balance, production costs can be reduced significantly. For this, it is crucial to identify consortia of several different strains that can be grown together in a stable mixed culture.

The aim of this thesis was to combine computational and experimental approaches to gain insights into the underlying mechanisms that are of key importance for understanding the behavior of L. lactis and cremoris strains in co-culture. Firstly, I have focused on the development of a method for generating strain-specific genome-scale metabolic models (GEMs) for a large number of Lactococcus strains and using the resulting models in a comparative study as a means to identify metabolic diversity among these strains. The GEMs predicted a difference in the ability to utilize pentose sugars xylose and ribose, between L. lactis and L. cremoris, indicating that over 90% of L. cremoris strains were unable to utilize xylose. Additionally, the workflow presented in this thesis enables the rapid generation of genome-scale metabolic models for a large number of strains, facilitating GEM-based comparisons to assess their suitability for various biotechnological applications.

Secondly, I aimed to gain a deeper understanding of the differences in growth physiology and pathway expression in Lactococcus strains by collecting and analyzing single strain growth and transcriptomics data for 12 dairy L. lactis and L. cremoris strains. By combining this data with pairwise co-cultivation interaction data and using a kinetic modeling approach, it was shown how substrate competition impacts strain balance as a result of differences in substrate affinity and the maximum specific growth rates. Furthermore, it was hypothesized that genetic diversity and variations in expression levels of the bacteriocin lactococcin could contribute to the dominance of one group of strains.

In this work, I have presented different hypotheses on the underlying mechanisms and strain-strain interactions that could impact the strain balance in mixed-culture fermentations. However, further research will be needed to combine these learnings and develop a method for accurately modeling co-culture behavior.