Sensor domains have to meet several criteria to be considered as scaffolds for the engineering of FRET-based metabolite nanosensors. First, it is essential that such a sensor domain is able to undergo a conformational change large enough to transduce metabolite binding into a FRET change. The conformationalchange must be tightly coupled to substrate binding. Sensor domains with similar three-dimensional structures but different substrate binding specificities will be ideal for development of a wide spectrum of nanosensors for many different analytes. Suitable fusion sites must exist, preferably at a distance corresponding to the R0 value for the given fluorophore pair. Another important consideration is that the sensor has to have a binding constant that corresponds to the expected detection range. A high substrate binding affinity would provide an ideal starting point to engineer mutant nanosensors for different physiologic detection ranges.
In the search for candidates that meet these criteria, we have focused our attention on a class of proteins found mainly in gram-negative bacteria, the periplasmic binding proteins (PBPs). PBPs comprise a large number of diverse proteins that cover a wide range of metabolites. PBPs typically have a diameter of approximately 5 to 7 nm, an ideal range for the estimated R0
of 4.9 for the pair of enhanced cyan fluorescent protein (ECFP) and enhanced yellow fluorescent protein (EYFP) (30
). PBPs typically bind ligands with affinities in the high nanomolar to low micromolar range, with a fast, one-step reversible binding scheme (31
). Most importantly, PBPs undergo a significant conformational change when binding to their target ligands, thus meeting the criteria for the design of nanosensors describe above. Based on the design concept for cameleon
, we tested whether PBPs could be engineered into FRET-based sensor proteins. Although unrelated at the primary sequence level, most PBPs consist of two similar globular domains. The binding site is created by specific residues in the cleft between the domains, which engulfs the ligand through a Venus flytrap-like hinge-twist motion. Crystal structures of more than a dozen PBPs, several in bound and unbound states, provide us with a detailed understanding of the mechanism of binding and the hinge motion (32
To develop a prototype nanosensor, we fused the periplasmic maltose binding protein (MBP) from Escherichia coli
between ECFP and EYFP by using short linker sequences (). Subsequent site-directed mutagenesis of amino acid residues involved in maltose binding led to the development of nanosensor FLIPmal-25μ (fluorescent indicator protein for maltose with a Kd
of 25 μM) with a binding constant for maltose of 25 μM () (33
). FLIPmal-25μ nanosensor was characterized in vitro after expression in E.coli
and was subsequently used to visualize real-time maltose uptake into the cytosol of living yeast cells.
Fig. 1 Domain structure of the FLIP cassette used for construction of the nanosensors. Different PBPs were flanked by short linker sequences (green) and fused between ECFP and EYFP. GGBP, glucose/galactose binding protein. [Color figure can be viewed in the (more ...)
Properties of the Nanosensors
Although the bacterial periplasmic glucose/galactose binding protein and ribose binding protein (RBP) are unrelated to MBP at the primary sequence level, both have tertiary structures similar to that of MBP (34
). Although glucose/galactose binding protein and RBP had different relative positions of the N- and C-termini compared with MBP, both proteins could successfully be engineered into FRET-based nanosensors for glucose or ribose, respectively () (35
). As predicted from the relative change in position of the termini, FRET decreased with increasing substrate concentration for FLIPglu and FLIPrib sensors. Using site-directed mutagenesis, five ribose sensors with a range of binding constants between 400 and 11.7 mM were generated (, ) (37
). This proof of concept suggests that it will be feasible to develop a wide spectrum of nanosensors that exploit the myriad of different PBPs and their structural relatives available in nature.
Fig. 2 Ribose nanosensors. A: Substrate-induced FRET changes. Spectra of FLIPrib-250n with and without a saturating concentration of ribose share an isosbestic point at 503 nm. B: In vitro substrate titration of purified nanosensor FLIPrib-4μ. Saturation (more ...)
The FLIP nanosensors were applied for in vivo imaging, which provided new insights into cellular sugar homeostasis. The chimeric proteins were expressed in mammalian cell culture, and cytosolic glucose and ribose levels were monitored by determination of the relative emission ratio from EYFP and ECFP using a device capable of rapid emission filter switches or by parallel imaging of both emission wavelengths with an image splitter.
At least in yeast, it seems that glucose is metabolized as soon as it enters the cell; thus, little or no glucose was detected within the intracellular space (38
). In eukaryotic cells, the relative rates of uptake, phosphorylation, and release are thought to be the major factors that control the levels of free glucose in the cytosol (39
). Using FLIPglu-600μ, a mutagenized glucose nanosensor with a Kd
600 μM for glucose, free glucose was detected in the cytosol of animal cells (35
). Metabolism of glucose occurred at a rate that kept cytosolic glucose levels at approximately half of the externally supplied concentration in COS-7 cells. Similar results were obtained when analyzing ribose dynamics, which demonstrated that free ribose accumulates in the cytosol, where it is slowly metabolized () (37
). Cell lines that express the nanosensors are currently used to characterize glucose compartmentalization and intracellular transport.
Fig. 3 Ribose transport and detection of ribose in the cytosol of COS-7 cells. Ratio images are pseudocolored to demonstrate ribose-dependent ratio changes. Red indicates high ratio and blue indicates low ratio. Integration of the ratio over all cells was used (more ...)
The overall diameter (longitudinal elliptic axis) of PBPs is in the range of 5 to 7 nm. The distance between the N-and C-termini as taken from the crystal structure of the ligand-free form is approximately 4 to 5 nm for MBP and RBP (). The N- and C-termini move closer together after ligand binding by about 0.7 nm in case of MBP and farther apart by 0.2 nm in the case of RBP. Interestingly, the maximal change in ratio in all cases is approximately 0.25, suggesting that other parameters such as dipole orientation contribute to the observed ratio change (40
Fig. 4 Dimensions of MBP (A) and RBP (B). The overall diameter (longitudinal elliptic axis) and the distance between N- and C-termini (as measured with Deepview PDB Viewer software) in open and closed form are indicated. The crystal structures used were 1OMP (more ...)
The observed maximal change in ratio is relatively small, thus limiting the dynamic range for in vivo detection. Although a set of mutant nanosensors may be generated for each case that covers the full range as in the case of FLIPrib, it would be advantageous if the dynamic range of individual nanosensors could be further improved to cover a larger detection range by increasing the maximal ratio change as has been done over time for the cameleon
). Ways to obtain sensors with larger changes of ratio after substrate binding might be site-directed mutagenesis or insertions and deletions in the linker regions that connect the binding protein to the fluorophores.
One of the major advantages of genetically encoded nanosensors is their suitability for the analysis of compartmentalization by targeting to subcellular compartments, as exemplified by the construction of a glucose nanosensor targeted to the nucleus (42
). The nuclear sensor FLIPglunuc
was used in a comparative study to verify that nuclear and cytosolic glucose levels are tightly coupled. As anticipated, the sensors could be targeted to different compartments to study their homeostasis individually.