With the rapid progresses being made in exploring diverse applications of
fluorescent proteins in biosensors, one of the major considerations is improving
the “brightness” of the fluorophore to achieve higher sensitivity. The brightness of
fluorescent protein depends on how well a molecule absorbs light and how fast it
emits light. Light absorption by a fluorophore is quantified in terms of molecular
extinction coefficient, whereas emission of light intensity is quantified by quantum
yield. A promising fluorescent protein for designing a biosensor is identified based
on the high quantum yield of the protein. The quantum yield relates the efficiency at
which a fluorescent molecule converts absorbed photons into emitted photons, i.e.,
number of photons emitted divided by the number of photons absorbed, with an
efficiency of 1.0 being the maximum possible value. Since it is difficult to know the
precise number of photons absorbed without specialized instrumentation, the typical
practice of measuring quantum yield depends on comparing the unknown to a
known standard [19, 20]. In most of the fluorescent proteins, the quantum yield of
the fluorophore is not solely determined by absorbed photons but also by other
environmental factors such as pH, temperature, polarity, etc.
The construction of fluorescent biosensors generally relies on the rational design
of the strategy, which begins with an effort to find a macromolecular receptor with
appropriate affinity and specificity to the target. The second step integrates the
receptor molecular recognition event into suitable fluorescent signal transduction,
which involves foreign reporter moieties such as an engineered autofluorescent
protein (AFP). The resultant possibilities for the engineered biosensor are then
screened based on biosensor quantum yield, sensitivity, and measurement dynamics
[21]. In spite of a seemingly simple procedure, researchers are attempting to fabricate
a novel fluorescent biosensor for a given target that would inevitably struggle with
unexpected labor intensity in screening such large possibilities of random mutagenesis. The field of computational biology and machine learning offers an exceptional
brute force approach in screening the such large sea of variants, by placing a
“virtual” molecule of interest in the binding site of a virtual receptor [22]. The
program sequentially mutates the receptor amino acids involved in binding the
ligand, searching for sequences that form a surface complementary to the ligand.
Typically, even with 12–18 amino acids mutations, around 10
23 variants arise which
excludes the possibility of in vitro screening. Finding productive biosensors with
high quantum yield and sensitivity in such a large number of possibilities requires
powerful computational algorithms. The success of such algorithms depends on how
precisely the model recapitulates the energy (or “fitness”) of the interacting groups.
However, when approaching this problem computationally, not only the amino acid
sequence of the receptor must be specified but also the orientation of the ligand, as
well as the various conformations that might be adopted by the side chains of the
mutated amino acids. Different models have evolved over the course of time for
protein biosensor engineering, for instance, Rangefinder, a computational algorithm
developed by Mitchell et al., which performs in-silico screening of dye attachment
sites in a ligand-binding protein for the conjugation of a dye molecule to act as a
Förster acceptor for a fused fluorescent protein [23]. Such computational protein
designs have been successfully used to precisely arrive at efficient protein models; a
Applications of Fluorescent Protein-Based Sensors in Bioimaging
153
fluorescent proteins in biosensors, one of the major considerations is improving
the “brightness” of the fluorophore to achieve higher sensitivity. The brightness of
fluorescent protein depends on how well a molecule absorbs light and how fast it
emits light. Light absorption by a fluorophore is quantified in terms of molecular
extinction coefficient, whereas emission of light intensity is quantified by quantum
yield. A promising fluorescent protein for designing a biosensor is identified based
on the high quantum yield of the protein. The quantum yield relates the efficiency at
which a fluorescent molecule converts absorbed photons into emitted photons, i.e.,
number of photons emitted divided by the number of photons absorbed, with an
efficiency of 1.0 being the maximum possible value. Since it is difficult to know the
precise number of photons absorbed without specialized instrumentation, the typical
practice of measuring quantum yield depends on comparing the unknown to a
known standard [19, 20]. In most of the fluorescent proteins, the quantum yield of
the fluorophore is not solely determined by absorbed photons but also by other
environmental factors such as pH, temperature, polarity, etc.
The construction of fluorescent biosensors generally relies on the rational design
of the strategy, which begins with an effort to find a macromolecular receptor with
appropriate affinity and specificity to the target. The second step integrates the
receptor molecular recognition event into suitable fluorescent signal transduction,
which involves foreign reporter moieties such as an engineered autofluorescent
protein (AFP). The resultant possibilities for the engineered biosensor are then
screened based on biosensor quantum yield, sensitivity, and measurement dynamics
[21]. In spite of a seemingly simple procedure, researchers are attempting to fabricate
a novel fluorescent biosensor for a given target that would inevitably struggle with
unexpected labor intensity in screening such large possibilities of random mutagenesis. The field of computational biology and machine learning offers an exceptional
brute force approach in screening the such large sea of variants, by placing a
“virtual” molecule of interest in the binding site of a virtual receptor [22]. The
program sequentially mutates the receptor amino acids involved in binding the
ligand, searching for sequences that form a surface complementary to the ligand.
Typically, even with 12–18 amino acids mutations, around 10
23 variants arise which
excludes the possibility of in vitro screening. Finding productive biosensors with
high quantum yield and sensitivity in such a large number of possibilities requires
powerful computational algorithms. The success of such algorithms depends on how
precisely the model recapitulates the energy (or “fitness”) of the interacting groups.
However, when approaching this problem computationally, not only the amino acid
sequence of the receptor must be specified but also the orientation of the ligand, as
well as the various conformations that might be adopted by the side chains of the
mutated amino acids. Different models have evolved over the course of time for
protein biosensor engineering, for instance, Rangefinder, a computational algorithm
developed by Mitchell et al., which performs in-silico screening of dye attachment
sites in a ligand-binding protein for the conjugation of a dye molecule to act as a
Förster acceptor for a fused fluorescent protein [23]. Such computational protein
designs have been successfully used to precisely arrive at efficient protein models; a
Applications of Fluorescent Protein-Based Sensors in Bioimaging
153