Industrial Systems Biology and Metabolic Engineering of Saccharomyces cerevisiae
Doctoral thesis, 2009
Saccharomyces cerevisiae is the most well characterized eukaryote, the preferred microbial cell factory
for the largest industrial biotechnology product (bioethanol), and a robust commercially compatible scaffold to be exploited for diverse chemical production. Succinic acid is a highly sought after added-value chemical which
is not overproduced in native S. cerevisiae strains. The genome-scale metabolic network reconstruction of S.
cerevisiae enabled in silico gene deletion predictions. First, a multi-gene, non-intuitive, genetic engineering
strategy guided by an evolutionary programming method to couple biomass formation through glycine/serine
amino acid requirements to succinate production was proposed. Pursuing these targets, a multi-gene deletion
strain was constructed, and directed evolution with selection was used to identify a succinate producing mutant.
Physiological characterization coupled with integrated data analysis of transcriptome data in the metabolically
engineered strain were used to identify 2nd-round metabolic engineering targets – overexpression of ICL1. The
resulting strain represents a 30-fold improvement in succinate titer, and a 43-fold improvement in succinate yield on biomass, with only a 2.8-fold decrease in the specific growth rate compared to the reference strain. Further genome-scale metabolic modeling supplemented with pathway visualization, flux balance analysis, and model
modifications to better simulate batch glucose conditions was performed. Identification of the top single and
double gene deletion strategies, under aerobic and anaerobic conditions, resulted in three predictions with a 10-fold improvement in succinate yield on glucose compared to the reference: MDH1, OAC1, and DIC1. While
⊿mdh1 and ⊿oac1 strains failed to produce more succinate relative to the reference, ⊿dic1 produced 0.02 C-mol C-mol-glucose-1, in close agreement with model predictions (0.03 C-mol C-mol-glucose-1). Pathway
visualization, coupled with transcriptional profiling, suggested that succinate formation was coupled to
mitochondrial redox balancing, and more specifically, reductive TCA cycle activity. The aforementioned
metabolic engineering strategies were designed based on glucose supplementation and metabolism. Future S.
cerevisiae microbial cell factories capable of fast and efficient xylose consumption for biorefinery compatibility,
and succinic acid overproduction would be highly desirable. Metabolic engineering of S. cerevisiae for
consumption of xylose aerobically without redirection of some carbon flux to overflow metabolites (ethanol,
glycerol, acetate, xylitol) was accomplished by expression of PsXYL1, PsXYL2, and PsXYL3 from the native
xylose-metabolizing Pichia stipitis, and subsequent, directed evolution. The resulting S. cerevisiae strain showed xylose consumption at a specific rate of 0.31 g g-cell-1 h-1, a specific growth rate of 0.18 h-1, and a biomass yield
of 0.62 C-mol C-mol-xylose-1. Plasmid isolation and re-transformation confirmed the conferred phenotype
resulted from a chromosomal modification. Transcriptional profiling confirmed a strongly up-regulated
glyoxylate pathway enabling sustained respiratory metabolism. A proof-of-concept study was performed to
determine if whole high-throughput genome sequencing could be used as a tool in metabolic engineering for
direct identification of genotype to phenotype correlations. Therefore, whole genome sequencing of S.
cerevisiae S288C and CEN.PK113-7D resulted in identification of 13,787 filtered SNPs in CEN.PK113-7D,
with a total of 939 SNPs detected across 158 unique metabolic genes, 85 of which contained a total of 219 nonsilent
SNPs. S. cerevisiae CEN.PK113-7D exhibited significantly higher ergosterol content correlating with
non-silent SNPs identified in ERG8 and ERG9. The flux through the galactose uptake pathway was much lower
in S288C compared with CEN.PK113-7D, correlating with the non-silent SNP enrichment in GAL1 and GAL10.
Inspection of the significantly differentially expressed genes between strains did not reveal an obvious gene
cluster that would explain the significant physiological differences, strongly suggesting that genotype to
phenotype correlation is manifested post-transcriptionally or post-translationally.
industrial systems biology