Systems analysis of Mycoplasma hyopneumoniae to improve vaccine production

Mycoplasma hyopneumoniae (M. hyopneumoniae) is a bacterial pathogen that has evolved from a gram-positive ancestor and specifically colonizes the lower respiratory tract of pigs where it causes enzootic pneumonia and plays a major role in the development of respiratory disease in pigs. Whole-cell inactivated vaccines are available that lower the severity of disease and are widely applied in pig industry to prevent clinical signs and improve pig herd health. However, production of these vaccines is challenging because it is not known which bacterial components are needed for protection and complex cultivation media are needed because growth requirements are not completely understood. The aim of this thesis was to understand growth and survival strategies of M. hyopneumoniae during infection, to integrate this knowledge with metabolic modeling under conditions used for vaccine production and apply this knowledge to improve the current production process for M. hyopneumoniae vaccines. Chapter 1 provides a general introduction into the disease, treatment and prevention methods with a focus on vaccines. I then introduce the characteristics of the M. hyopneumoniae genome, transcriptome and review the current knowledge on infectious mechanisms and the response of the pig to infection and vaccination. Finally, I discuss the challenges related to vaccine production and introduce systems biology tools that will be applied in the thesis. In chapter 2 we define a strategy for risk-based process development of bacterial vaccines which provided the framework for future studies performed during this thesis. We propose to integrate the academic workflow for rational strain design with the industry standard for process design. Systems biology tools, especially genome-scale metabolic models, play an essential role in this strategy because application of these tools reduces process risks and increases process understanding. Therefore, in line with this strategy, we created a manually curated genome-scale metabolic model of M. hyopneumoniae which we applied to dynamically model the cultivation step in the vaccine production process (chapter 3). We found that only 16% of cellular energy in a standard fermentation was used for growth and 84% was used for non-growth associated maintenance. By model-driven experimentation we were able to increase the fraction of cellular energy used for growth by addition of pyruvate to the production medium, and showed in dedicated fermentor experiments that the improved process reached a 2.3 times higher biomass yield. Although the metabolic model helped to increase process yield, it did not allow prediction of a defined cultivation medium without components from porcine origin. Therefore, to better understand the dependency of M. hyopneumoniae on host derived components, we performed a functional comparison of 80 mycoplasma genomes and used multivariate and machine-learning algorithms to relate functional capability to the specific host and niche of mycoplasma species (chapter 4). This analysis allowed us to identify protein domains possibly needed for growth and survival in the pig lung. In addition, we found that protein domains expected to be essential for bacterial growth were not persistently present in mycoplasma genomes suggesting that alternative domain configurations exist that bypass their essentiality. To better understand whether the proteins we identified as possibly important for survival in pigs actually play a role during M. hyopneumoniae infection, we sequenced the bacterial mRNA during infection in chapter 5 and compared the in vivo transcriptome to that of broth grown mycoplasma. We found 22 up-regulated and 30 down-regulated genes during infection (FDR<0.01 and fold change >2LOG2) and identified differentially expressed ncRNAs. In chapter 6 we build upon our mycoplasma basis to further analyse the role of ncRNAs in bacterial genomes. We identified an exponential relationship between the AT content of genomes and the number of ncRNAs and propose that this relation is the result of spurious transcription, which is more likely to occur in AT rich genomes. This hypothesis is further substantiated by showing that spurious transcription demands minimal cellular energy and that overexpression of cis-binding ncRNAs in M. pneumoniae did not influence the level of proteins translated from their overlapping mRNAs. Finally, in chapter 7 I discuss four system strategies, identified in this thesis and derived from recent literature, and discuss how these strategies could be integrated in the metabolic model of M. hyopneumoniae. Lastly, I provide an outlook on the next steps needed for improvement of the production process for M. hyopneumoniae vaccines. In conclusion, this work provided novel insight in the metabolic capability of M. hyopneumoniae based on the proteome domain content, captured in a genome-scale metabolic model and studied under in vitro and in vivo conditions. Biomass yield of the cultivation step for vaccine production was increased and the basis was laid to further improve the production process for M. hyopneumoniae vaccines using model-based experimentation.