Biology Reference
In-Depth Information
the ability to activate gene expression in response to custom ligands that inhibit ribozyme
cleavage. Various instantiations of this method have created regulatory nodes displaying
logic functions, as well as higher-order forms such as sigmoidal transfer functions. The
control of mRNA stability, especially in eukaryotes where active processing of mRNA
requires a specific 3
-UTR, offers a promising strategy to control gene expression.
'
Together, common principles concerning the design and use of elements controlling
transcription initiation/elongation, translation initiation, and degradation are emerging.
Standard molecular platforms are being discovered and developed to create families of parts
with similar physics of function and mapping of sequence to activity. Each element class
controls different elements of gene expression dynamics, including steady-state levels, timing,
and noise, and each allows for modulation of their activity through a sensing layer.
43
Thus,
each may be used to sculpt complex expression control functions for a given set of genes.
ASSEMBLING PARTS
However, once parts controlling gene expression are developed, they are meant to be used
and reused in different applications. This means that they will be taken out of the context
in which they were initially developed and characterized
usually in standard strains with
GFP as the reporter gene and small culture volumes
and used to control other genes,
possibly in specialized host strains and varied culturing conditions. In addition, a part will
rarely be used in isolation; more often, it will be composed with other regulatory parts into
genetic switchboards, complex nodes, and regulatory circuits. Therefore, the success of the
parts-based approach to genetic circuit engineering depends on the ability of parts to function
predictably in novel arrangements. However, it has become clear that most biological parts
are not robust across different contexts.
6
As a result of these context effects, engineering of
synthetic biological systems from parts is often reduced to an ad hoc process wherein random
libraries circuits composed of parts and their variants are screened for desired function.
69
As an example, consider the simplest and most necessary combination of parts: for a gene
to be expressed, its coding sequence must be prefixed by a promoter and a 5
-UTR, which in
bacteria contains the RBS. This already leaves room for interactions among the promoter,
RBS, and coding sequence, from which unpredictable effects may emerge. To explore
these interactions, Mutalik et al. recently assembled and experimentally measured all the
combinations between seven promoters, eleven 5
'
-UTRs, and two genes (fluorescent
reporters).
44
They then used the analysis of variance (ANOVA) framework to accurately
quantify average part performance across different contexts and part composability, which
is the variability in performance that can be attributed to the interaction between functionally
different genetic elements. In this case, they found that promoters could be reliably composed
with 5
'
-UTRs presented
unreliable performance when combined with different genes. Though perhaps unsurprising
-UTRs and genes without significant emergent behavior. However, 5
'
'
it has been known for some time that 5
-untranslated RNA can, for example, form structures
with particular coding sequences that sequester ribosome binding sites
8,9
'
this example
shows that even the simple task of taking constitutive promoters and combining them with
unregulated 5
-UTRs and two fluorescent reporter genes is subject to emergent properties that
influence expression. Are there strategies for making parts and their assembly more robust?
'
Models of Function
If one has a complete model of a part
s behavior in different contexts, then the issue of
variability is moot as behavior in any context can be predicted in silico. Of course, achieving
accurate models that consider all relevant contextual variables is an ambitious goal; but for
several systems, it has become within reach. For instance, in the RBS calculator example
discussed above,
36
the algorithm used to generate RBS parts does not treat the RBS in
isolation but explicitly models its interaction with surrounding sequences, using the range of
'
