TY - JOUR
T1 - Improving metabolic pathway efficiency by statistical model-based multivariate regulatory metabolic engineering
AU - Xu, Peng
AU - Rizzoni, Elizabeth Anne
AU - Sul, Se Yeong
AU - Stephanopoulos, Gregory
N1 - Publisher Copyright:
© 2016 American Chemical Society.
PY - 2017/1/20
Y1 - 2017/1/20
N2 - Metabolic engineering entails target modification of cell metabolism to maximize the production of a specific compound. For empowering combinatorial optimization in strain engineering, tools and algorithms are needed to efficiently sample the multidimensional gene expression space and locate the desirable overproduction phenotype. We addressed this challenge by employing design of experiment (DoE) models to quantitatively correlate gene expression with strain performance. By fractionally sampling the gene expression landscape, we statistically screened the dominant enzyme targets that determine metabolic pathway efficiency. An empirical quadratic regression model was subsequently used to identify the optimal gene expression patterns of the investigated pathway. As a proof of concept, our approach yielded the natural product violacein at 525.4 mg/L in shake flasks, a 3.2-fold increase from the baseline strain. Violacein production was further increased to 1.31 g/L in a controlled benchtop bioreactor. We found that formulating discretized gene expression levels into logarithmic variables (Linlog transformation) was essential for implementing this DoE-based optimization procedure. The reported methodology can aid multivariate combinatorial pathway engineering and may be generalized as a standard procedure for accelerating strain engineering and improving metabolic pathway efficiency.
AB - Metabolic engineering entails target modification of cell metabolism to maximize the production of a specific compound. For empowering combinatorial optimization in strain engineering, tools and algorithms are needed to efficiently sample the multidimensional gene expression space and locate the desirable overproduction phenotype. We addressed this challenge by employing design of experiment (DoE) models to quantitatively correlate gene expression with strain performance. By fractionally sampling the gene expression landscape, we statistically screened the dominant enzyme targets that determine metabolic pathway efficiency. An empirical quadratic regression model was subsequently used to identify the optimal gene expression patterns of the investigated pathway. As a proof of concept, our approach yielded the natural product violacein at 525.4 mg/L in shake flasks, a 3.2-fold increase from the baseline strain. Violacein production was further increased to 1.31 g/L in a controlled benchtop bioreactor. We found that formulating discretized gene expression levels into logarithmic variables (Linlog transformation) was essential for implementing this DoE-based optimization procedure. The reported methodology can aid multivariate combinatorial pathway engineering and may be generalized as a standard procedure for accelerating strain engineering and improving metabolic pathway efficiency.
KW - Combinatorial optimization
KW - Metabolic engineering
KW - Promoter library
KW - Statistical models and response surface methodology
KW - Synthetic biology
UR - http://www.scopus.com/inward/record.url?scp=85012838041&partnerID=8YFLogxK
U2 - 10.1021/acssynbio.6b00187
DO - 10.1021/acssynbio.6b00187
M3 - 文章
C2 - 27490704
AN - SCOPUS:85012838041
SN - 2161-5063
VL - 6
SP - 148
EP - 158
JO - ACS Synthetic Biology
JF - ACS Synthetic Biology
IS - 1
ER -