Design and development of fluorescent biosensors
A focus on single-FP (single fluorescent protein) sensor designs with fluorescence lifetime readouts
There are numerous approaches to making genetically encoded biosensors based on fluorescent proteins (FPs), but the two main classes are "FRET sensors" and single-FP sensors.
FRET sensors are based on the principle of Förster resonance energy transfer (FRET), the process where one excited FP can transfer its energy directly (without emitting a photon) to another FP. The design requires two FP's, usually of different colors: for instance, the experimenter excites a cyan FP and measures how efficiently it transfers its energy to a yellow FP in the same biosensor protein. The principle of biosensing is that somehow the binding of a target analyte leads to a conformational change that brings the two FPs closer together (or further apart), changing the FRET efficiency and producing a change in the ratio of yellow to cyan emission. FRET sensors work quite well for engineered conformational changes (for instance, for kinase sensors containing a pseudosubstrate linked by a flexible linker to a binding protein that binds only the phosphorylated pseudosubstrate sequence), but often have more subtle responses to the smaller, naturally occurring conformational changes in analyte binding proteins.
For such smaller conformational changes, a single-FP approach can work better. Normally, simply splicing together an FP with an analyte binding protein does not produce a change in fluorescence, since the normal N- and C-termini of the FP are well separated from the chromophore in the interior of the protective β-barrel structure (in the cartoon below, the green FP shows the chromophore in a space-filling format, with the normal N- and C-termini at the lower right). However, in a "circularly permuted" cpFP, the coding sequence of the FP is rearranged to start with a new N-terminus in the middle of the protein, reading to the end (original C-terminus), connecting to the original N-terminus by a short protein linker, and then continuing around to the new N-terminus (the new N- and C-termini are at the lower left side of the green FP). Now this cpFP sequence can be inserted into a moving element of the binding protein (in the cartoon, the light gray linkers connect the termini of the cpFP to the blue mobile loop of the ATP-binding protein at the bottom). Permutations placing the linkers in the 7th strand of the β-barrel are commonly used, as these cpFPs typically fold well, and they often allow conformational changes in the binding protein to produce fluorescence changes.
The sensor depicted in the cartoon is Perceval, our sensor of the cellular ATP:ADP ratio, a fusion protein that consists of a bacterial regulatory protein, GlnK1, and a cpFP. GlnK1 is a small trimeric protein with three nucleotide binding sites. These sites can bind either ATP or ADP - but only ATP produces a local conformational change that involves closure of the "T-loop" over the nucleotide binding site. To convert this conformational change to a fluorescent signal, we spliced a "circularly permuted" GFP variant (cpmVenus) into the T-loop. This cartoon shows a single subunit of GlnK1 with Mg-ATP bound; the T-loop (dark blue) is linked to circularly permuted Venus (green).
We use directed evolution to tune the properties of a prototype sensor for use in living cells. Multiple rounds of semi-random mutagenesis - using either doped oligonucleotides or error-prone PCR - are combined with screening of the fluorescent response (in a programmable 96-well plate reader, or more recently in microfluidically produced objects called "gel-shell beads") to give the best response properties.
Most cpFP sensors allow for a quantitative "ratiometric" readout: the unoccupied sensor is mainly excited by one wavelength of light, while the peak of the occupied sensor is at a different wavelength. Measuring the ratio between the two excitation wavelengths allows the occupancy of the sensor to be inferred, independent of the total expression level of the biosensor — for any quantitative measurement of occupancy, it is essential to factor out variation in the expression level. Unfortunately, the two excitation peaks correspond to two different chemical species which are interconverted by protonation, and therefore the ratio can be quite sensitive to the ambient pH, which can often change slightly in the target cells where the sensor is deployed.
An alternative sensing modality is a change in "fluorescence lifetime", the average time that the fluorescent protein spends between absorption and emission of a photon. This time is typically in the range of 1-3 ns for most FPs, and it can sometimes change in response to analyte binding. Like the excitation ratio, the lifetime is independent of the expression level of the biosensor. A ratiometric-readout sensor can often be converted to a lifetime-readout sensor by replacing the original FP with an FP with very low pKa (protonation constant) that is outside the physiologic range, and then reoptimizing the sensor for a substantial lifetime change.
Stay tuned for our new research on the structural mechanism of the lifetime change in several biosensors!
As for other fluorescence parameters like ratio, fluorescence lifetime can be monitored in living cells, in vivo or ex vivo, using a properly equipped fluorescence microscope (in this case a fast-FLIM microscope).