Linear covariance models to examine the determinants of protein levels in Lactococcus lactis

When confronted to environmental changes, microorganisms adjust protein levels in order to adapt their growth and metabolic performances. Biological mechanisms involved in protein regulation are extremely complex and still poorly understood. This study aims at the identification, by statistical modelling, of the major determinants of protein concentrations in a bacterial model Lactococcus lactis. Protein concentrations were predicted by covariance models taking into account various quantitative and qualitative parameters. Best models were selected thanks to Akaike Information Criterion. For protein estimation, we found that the sequence-related feature Codon Adaptative Index was a more influential parameter than the transcript amount, suggesting the control by genetic determinism is stronger than by metabolic adaptation. In addition, protein length, aromaticity but also their biological functions, were proved to have unexpected influences on protein concentrations. These protein determinants were for the first time demonstrated to be not constant and depended on the adaptation process, the main difference between permanent and transient adaptations being detected for regulatory protein concentrations. With the growing accumulation of omics data this statistical method appears to be a valuable tool to understand biological networks and their regulations. This approach was applied to study the translation of proteins but can be extended to other metabolic processes and is also adaptable to other microorganisms.