Systems Biology and Metabolic Engineering in Bacteria - Systems Biology of Metabolic and Signaling Networks - page 353

applied to maximize respiration rates in Geobacter sulfurreducens (Burgard
et al. 2003 ).
The construction of synthetic pathways is another class of applications that arise
as we build upon knowledge gained from understanding the function of metabolic
networks (Xu et al. 2012 ). The preferred chassis organisms for these synthetic
biology projects remain E. coli and S. cerevisiae (Na et al. 2010 ). Utilizing
databases of metabolic reactions similar to the ones employed for the reconstruction
of metabolic networks from genome annotations, Parkya and collaborators pro-
posed the OptStrain algorithm for pathway optimization by eliminating superfluous
reactions or constructing novel pathways in E. coli (Pharkya et al. 2004 ). The
additional dimension offered by synthetic biology is the possibility to explore
many variants of a particular metabolic pathway. A particularly clever method
has been pioneered by the Church group (Wang et al. 2009 ), according to which
several billion variants of a given pathway can be explored in parallel within a
couple of days. This strategy of combining synthetic biology with accelerated
evolution has considerably improved the efficiency of the lycopen production
pathway in E. coli . The question of which strategy—rational design of a pathway
and fine-tuning of intermediate reaction steps or combinatorial exploration of a
large number of variants of a particular pathway—is more efficient for the produc-
tion of new chemicals remains open (Yadav and Stephanopoulos 2010 ). A combi-
nation of both strategies may prove the most promising (see following paragraph).
No matter how sophisticated the rational design of a genetic-metabolic network
may be, there will always be “bugs” when the circuit is constructed in the host cell.
A first remedy would be to devise a method for easily detecting the problems. For
example, the imbalance of metabolic pathways often induces stress responses. The
signature of these responses could be used in future diagnostic tools for strain
optimization (Keasling 2012 ). Even though nonlinear models of complex systems
are essential for designing gene-metabolic systems, experimental strategies will be
needed for the final optimization of the construct. Recent experimental advances
allow a combinatorial exploration of diverse expression levels of the constituent
enzymes. A high throughput screen is used to select the “best” strain. This strategy
has been used to maximize xylose and cellobiose utilization in yeast
(Du et al. 2012 ). Xylose is an abundant pentose, but is inefficiently utilized by
ethanol producing yeast strains. Optimal xylose assimilation relies on the balancing
of enzymatic activities and cofactor usage. The best expression levels of the three
key enzymes constituting the xylose assimilation pathway was obtained by screen-
ing strains (based on colony size), each one containing different combinations of
about ten promoters of different strengths placed upstream of each of the three
genes comprising the pathway. The optimized strain improved ethanol yield by
more than 60 % with respect to the reference strain. A decisive advantage of the
experimental approach to optimization over a purely computational approach is the
possibility to adapt to varying behaviors of different strains. Indeed, the transcrip-
tional profile of two different strains optimized for the same pathway is different
(Du et al. 2012 ). Such fine adjustments are difficult, if not impossible, to predict
from modeling alone. A combination of modeling and combinatorial exploration