Date Published: June 7, 2018
Publisher: Public Library of Science
Author(s): Zachary P. Metz, Tong Ding, David J. Baumler, Garret Suen.
Listeria monocytogenes is a microorganism of great concern for the food industry and the cause of human foodborne disease. Therefore, novel methods of control are needed, and systems biology is one such approach to identify them. Using a combination of computational techniques and laboratory methods, genome-scale metabolic models (GEMs) can be created, validated, and used to simulate growth environments and discern metabolic capabilities of microbes of interest, including L. monocytogenes. The objective of the work presented here was to generate GEMs for six different strains of L. monocytogenes, and to both qualitatively and quantitatively validate these GEMs with experimental data to examine the diversity of metabolic capabilities of numerous strains from the three different serovar groups most associated with foodborne outbreaks and human disease. Following qualitative validation, 57 of the 95 carbon sources tested experimentally were present in the GEMs, and; therefore, these were the compounds from which comparisons could be drawn. Of these 57 compounds, agreement between in silico predictions and in vitro results for carbon source utilization ranged from 80.7% to 91.2% between strains. Nutrient utilization agreement between in silico predictions and in vitro results were also conducted for numerous nitrogen, phosphorous, and sulfur sources. Additionally, quantitative validation showed that the L. monocytogenes GEMs were able to generate in silico predictions for growth rate and growth yield that were strongly and significantly (p < 0.0013 and p < 0.0015, respectively) correlated with experimental results. These findings are significant because they show that these GEMs for L. monocytogenes are comparable to published GEMs of other organisms for agreement between in silico predictions and in vitro results. Therefore, as with the other GEMs, namely those for Escherichia coli, Staphylococcus aureus, Vibrio vulnificus, and Salmonella spp., they can be used to determine new methods of growth control and disease treatment.
Due to the importance of L. monocytogenes to the food industry, there is a continuous interest in new methods of control and treatment. The development of rapid, relatively cheap genome sequencing techniques in recent years has led to the emergence of genomic tools for identifying more properties of microorganisms through the field of systems biology and its associated computational techniques.
A numerical comparison of the genomes of the six strains was accomplished using the RAST (Rapid Annotation using Subsystem Technology) database (Table 1) . Each strain has approximately 3,000 genes, with 85%– 90% of those genes shared among all of the strains. Less than 3% of the genes for each strain are unique, and none of the three serovars chosen had more than 53 genes unique to that serovar.
When the content of predicted ORFs is compared, it is clear that much of the genome contents are shared by all six L. monocytogenes strains (Table 1). The genomes are all very similar in length, and greater than 85% of each genome is shared between all strains. This degree of conservation was expected, since most evidence suggests that the genomes of modern L. monocytogenes strains are highly conserved .
This study describes the creation, multi strain comparison, and qualitative experimental validation of six GEMs for L. monocytogenes. Nutrient utilization comparisons were made for sole sources of carbon, nitrogen, phosphorous, and sulfur. The models were then curated to better agree with the experimental results for carbon utilization, since that is the method most commonly used to validate GEMs. It was found that the final carbon source utilization agreement of the newly created models was comparable to the agreement seen by previously published models. Additionally, the nitrogen and phosphorus agreements were lower than previously published models, but the sulfur agreement was higher than previously published models. Quantitatively validating the GEMs as laboratory tools was done using batch growth experiments, and determination of scalar factor adjustments resulted in growth rate and growth yield predictions that were much more accurate, and were significantly correlated with the experimental results. The potential essential reactions identified by this study provide promising directions for future studies improving the safety of foods commonly contaminated by L. monocytogenes, including cheese, deli meats, salads, and fruit. The models could also be used for future iterations of GEMs for L. monocytogenes. These results indicate that GEMs generated by semi-automated tools provided by KBase, followed by manual curation and experimental validation, can also become successful predictive tools for use in the laboratory.