10
3 Building Synthetic Protein Switches
Considering computational approaches can only optimize a limited number of biophysical and functional properties in a protein;
this means the construction of synthetic protein switches relies, to
a significant extent, on empirical optimization based on mediumto high-throughput screening assays. As a general rule of thumb,
synthetic protein switches generated by means of random domain
insertion rely on higher throughput screening approaches due to
the less predictable effect of recombining two structurally welldefined protein domains on fold, structure, and function. In contrast, modularly organized synthetic protein switches can be
engineered in a more rational manner solely focusing on the length
and structure of the linkers connecting individual domains.
Consequently, each of the individual design strategies imposes different challenges on the underlying DNA assembly process.
Before the advent of highly affordable synthetic DNA, random
domain insertions were created following a limited endonuclease
digest of a circular DNA construct coding for an actuator and subsequent fusion with a linear DNA construct coding for an allosteric
receptor. The latter may also be circularly permutated resulting in
a set of new N- and C-termini which potentially enhances the
transmission of conformational changes between the receptor and
the actuator; these are not necessarily confined to the original Nand C-termini, but most pronounced at internal sites [96]. The
resulting libraries are then empirically screened for domain insertion mutants that are functionally recombined in allosteric hotspots
(c.f. SCA that aims to predict allosteric hotspots as opposed to
experimentally screen for them). This strategy has, for instance,
been successfully applied to engineer a number of allosterically
regulated enzymes, including maltose regulated β-lactamase [25,
97, 98], xylose regulated xylanase [99], and HIF1-binding domain
cytosine deaminase [100]. Considering only 1 in 6 constructs are
in frame and the unpredictable effect of domain insertion of protein structure and function, a large number of domain insertion
mutants need to be screened using a suitable high-throughput
screening assay. These can either be directly screened for functional
protein switches, e.g., based on antibiotic resistance conferring
β-lactamases [97] or in case of more technically challenging enzyme
assays fused with GFP to identify in frame, non-homologously
recombined genes before assaying for the relevant enzyme function in a multiwell plate assay format [99].
Alternatively, more focused DNA insertion libraries can be created
by means of homology-dependent DNA cloning methods overcoming the limitations associated with out-of-frame insertions:
e.g., overlap extension PCR (OE-PCR) constitutes one of the
3.1 Non-HomologyDependent
Recombination
Strategies
3.2 HomologyDependent
Recombination
Strategies
Viktor Stein
3 Building Synthetic Protein Switches
Considering computational approaches can only optimize a limited number of biophysical and functional properties in a protein;
this means the construction of synthetic protein switches relies, to
a significant extent, on empirical optimization based on mediumto high-throughput screening assays. As a general rule of thumb,
synthetic protein switches generated by means of random domain
insertion rely on higher throughput screening approaches due to
the less predictable effect of recombining two structurally welldefined protein domains on fold, structure, and function. In contrast, modularly organized synthetic protein switches can be
engineered in a more rational manner solely focusing on the length
and structure of the linkers connecting individual domains.
Consequently, each of the individual design strategies imposes different challenges on the underlying DNA assembly process.
Before the advent of highly affordable synthetic DNA, random
domain insertions were created following a limited endonuclease
digest of a circular DNA construct coding for an actuator and subsequent fusion with a linear DNA construct coding for an allosteric
receptor. The latter may also be circularly permutated resulting in
a set of new N- and C-termini which potentially enhances the
transmission of conformational changes between the receptor and
the actuator; these are not necessarily confined to the original Nand C-termini, but most pronounced at internal sites [96]. The
resulting libraries are then empirically screened for domain insertion mutants that are functionally recombined in allosteric hotspots
(c.f. SCA that aims to predict allosteric hotspots as opposed to
experimentally screen for them). This strategy has, for instance,
been successfully applied to engineer a number of allosterically
regulated enzymes, including maltose regulated β-lactamase [25,
97, 98], xylose regulated xylanase [99], and HIF1-binding domain
cytosine deaminase [100]. Considering only 1 in 6 constructs are
in frame and the unpredictable effect of domain insertion of protein structure and function, a large number of domain insertion
mutants need to be screened using a suitable high-throughput
screening assay. These can either be directly screened for functional
protein switches, e.g., based on antibiotic resistance conferring
β-lactamases [97] or in case of more technically challenging enzyme
assays fused with GFP to identify in frame, non-homologously
recombined genes before assaying for the relevant enzyme function in a multiwell plate assay format [99].
Alternatively, more focused DNA insertion libraries can be created
by means of homology-dependent DNA cloning methods overcoming the limitations associated with out-of-frame insertions:
e.g., overlap extension PCR (OE-PCR) constitutes one of the
3.1 Non-HomologyDependent
Recombination
Strategies
3.2 HomologyDependent
Recombination
Strategies
Viktor Stein