Synthetic Biology | State of the Art
37
codon optimisation - that is the creation of a sequence encoding the same protein, but avoiding rare codons.
The longevity of mRNA transcripts can be controlled by modifying its secondary structure in untranslated
regions that make it more or less vulnerable to RNases (Arpino et al. 2013).
Similarly,
protein degradation can be tuned by addition of degradation tags. For enzymatic activities, mutants
with
altered substrate specificity or thermostability can be created (Arpino et al. 2013).
Random sequence variation in DNA units can be achieved by various techniques such as directed evolution
via
error-prone PCR (Arpino et al. 2013) or by synthesis with degenerate sequence (Ellis et al. 2009). The resulting
libraries of diversified parts can be characterised in test systems and versions with optimal input-output
characteristics and reaction rates can be selected for further applications. Mathematical
modelling can greatly
assist this process. For example, Ellis et al. (2009) created a library of promoters and a model describing their
varying expression and inhibition thresholds. Based on this model the authors were able to produce devices
with quantitatively predictable behaviour. For RBS-optimisation a universal and popular software tool is
available: The RBS-calculator (Salis et al. 2009) reliably predicts ribosome binding strength of prokaryotic
Shine-Dalgarno sequences.
Copy number
The gene copy number can increase or decrease the expression of a protein over a partially linear range
(Arpino et al. 2013). Similarly, multiplication of repressor or activator binding sites has been shown to
influence gene expression in a predictable way (Xu et al. 2012).
The copy number of plasmid-borne genes and constructs can be modified easily by changing the origin of
replication to modify the
plasmid copy number. However maintenance of increasing numbers of plasmids
adds to the metabolic burden of the host cell. Alternatively repetitive
tandem copies of genes or control
elements can be inserted into the same plasmid or chromosome (Arpino et al. 2013; Xu et al. 2012).
Modularisation of genetic systems (MMME)
Multigene pathways can be systematically improved by multivariate modular metabolic engineering
(MMME).
MMME groups genes with similar turnover into modules,
i.e. into synthetic operons. Such groups of genes can
be
tuned in their expression as a single unit, using the toolbox of promoter and RBS strength, copy number
etc. This simplifies the adjustment of multiple genes to each other in order to eliminate bottlenecks and
maximise pathway turnover. An excellent example of MMME is the study of Ajikumar et al. (2010), who
optimised a synthetic terpenoid pathway for the synthesis of the taxol-precursor taxadiene in
Escherichia coli
(Figure 9). Ten genes of bacterial and plant origin were arranged in two modules that could be harmonised in
their expression levels. Using this approach, the authors needed to construct only 32 strains to identify a
variant with 15000-fold increased taxadiene production, whereas varying and combinatorial screening of the
ten individual genes would have required construction of 10 000 strains.
Modular plasmid backbones that
enable parallel generation of differently configured pathway variants within a few cycles of cloning facilitate
MMME (Xu et al. 2012).
Synthetic Biology | State of the Art
38
Figure 9: Modularisation of a taxadiene biosynthesis pathway. Detail from a figure of Ajikumar et al. (2010)
Simultaneous modification and combinatorial screening of multiple genes (MAGE)
More frequently than not, several genes need to be modified simultaneously to obtain a superior phenotype.
The recently developed multiplex-automated genomic engineering (
MAGE) protocol produces genomic
diversity by simultaneously modifying several genomic locations on the chromosome of
a single cell or across a
cells population (Yadav et al. 2012). The capabilities of MAGE have been demonstrated in a study of Wang et
al. (2009) who targeted all 20 genes involved in the terpenoid pathway of
Escherichia coli to modify
biosynthesis of the tetraterpene pigment lycopene. Significant improvements were made within short time.
However, combinatorial screening of 20 genes required testing of 100000 mutants. MAGE is therefore
restricted to end-products with colorimetric properties or other properties that can be assessed in high
throughput screenings (Yadav et al. 2012).
Metabolic flux analysis
Empowered by the volume of gene, protein and metabolite data accumulated in biotechnology databases the
focus of metabolic engineering has shifted from perturbing individual pathways towards manipulating the
entire cell for optimised product formation. The target pathway is considered in the context of the entire
metabolic network of the host cell. This enables handling competition with native pathways, accumulation of
inhibitory side products and other interaction phenomena (Yadav et al. 2012).
Flux balance analysis (
FBA) is a popular simulation tool for integrative metabolic engineering of well described
organisms, such as
Escherichia coli. FBA describes the network of a cell's metabolic reactions in a
stoichiometric matrix. Stoichiometric information is retrieved from experimental biochemical data and from
comparative analysis of enzyme coding genes of different completely sequenced organisms. Based on this
matrix, FBA predicts metabolic fluxes,
i.e. turnover rates of metabolites, and helps to maximise desired
functions (Comba et al. 2012; Bilgin and Wagner 2012). For example Xu et al. (2011) used FBA to identify a
minimum set of genetic interventions required to redirect the carbon flux in
Escherichia coli towards malonyl-
CoA building blocks for flavonoid synthesis.
For
dynamic regulation of metabolic fluxes, regulatory devices can be employed to sense accumulation of
metabolites and subsequently up-regulate reactions that redirect fluxes towards the intended output (Zhang
et al. 2012).
Metabolic network analysis can also be used to
predict the minimum metabolic requirements for a cell under
given conditions. Applying FBA to sets of random networks that were automatically generated from
biochemical reaction databases, Bilgin and Wagner (2012) calculated that a cell synthesizing 20 biomass
molecules from 20 alternative carbon sources requires a minimum of 260 reactions. Every additional carbon
source requires two extra reactions; every additional product requires
three extra reactions, on average.