1.1.

Monomerization of RFPs

Through the use of protein engineering, tetrameric DsRed was converted into monomeric RFP 1 (mRFP1; ${\lambda }_{\mathrm{ex}}=584\mathrm{nm}$; ${\lambda }_{\mathrm{em}}=607\mathrm{nm}$).¹⁵ In the DsRed tetramer, each subunit is engaged in distinct contacts with two of the other three subunits via two different interaction surfaces (Fig. 1). In order to monomerize DsRed, the protein–protein contacts at each interface were destabilized through mutations to charged residues such as lysine and arginine. Disruption of one interface yielded a dimeric intermediate and subsequent disruption of the remaining interface produced the monomeric FP (Fig. 1). While this process had the desirable outcome of producing monomeric variants, it also had the undesirable outcome of decreasing the intrinsic fluorescent brightness. A total of 33 mutations were introduced during the course of engineering mRFP1, including 13 interface-disrupting mutations and 20 fluorescence-rescuing mutations.¹⁶

Fig. 1

Conversion of the wild-type tetrameric red fluorescent protein (RFP) DsRed to an engineered monomeric RFP. (a) Cartoon representation of the structure of wild-type tetrameric RFP DsRed. (b) Disruption of the first A–B interface produces an A–C dimer intermediate and subsequent disruption of the A–C interface produces a monomeric RFP. Interface-disrupting mutations are typically detrimental to the proper folding and chromophore maturation of the intermediate dimer or monomer; therefore, these variants must be rescued by directed evolution. Cartoon structures are based on PDB ID 1G7K.¹⁴

The monomeric nature of mRFP1 addresses the most critical shortcoming associated with tetrameric DsRed. Other favorable properties of mRFP1 include a much shorter maturation time ($t_{0.5}<1\mathrm{h}$) and a 25-nm red-shifted fluorescence emission at 607 nm. These advantages make mRFP1 suitable for the construction of fusion proteins for live cell fluorescence imaging, as well as in the multicolor fluorescence imaging with avGFP variants.¹⁵ Unfortunately, mRFP1 also exhibits disadvantages such as reduced fluorescence brightness and photostability. Efforts to further improve mRFP1 focused on higher brightness, color diversification, and improved photostability. These efforts eventually produced a number of useful RFP variants that are now known as the mFruit series.^16,17 The prototypical RFP in the mFruit series is mCherry, which is generally considered to be the successor of mRFP1.

1.2.

Structure and Chromophore Formation in RFPs

The x-ray crystal structures of DsRed and mCherry (as examples of prototypical RFPs) reveal that these proteins have a cylindrical shape created by eleven $\beta$-strands wrapped around a central helix.¹⁸ This distinctive tertiary structure, which is shared with avGFP,^13,14,19,20 is often referred to as a $\beta$-can or a $\beta$-barrel. The $\beta$-barrel is $\sim 4\mathrm{nm}$ in height and $\sim 3\mathrm{nm}$ in diameter (Fig. 2). The chromophore is located near the middle of the central helix and is protected from the external environment by the eleven $\beta$-strands that surround it.

Fig. 2

Structure of a representative monomeric RFP, mCherry. The secondary structure is shown in a cartoon representation with the helix colored in yellow, $\beta$-strands colored in red, and loops colored in orange. The chromophore is shown in a stick representation with carbon atoms colored in gray, nitrogen atoms colored in blue, and oxygen atoms colored in red (PDB ID 2H5Q).

DsRed forms its chromophore from three sequential amino acids: Gln65, Tyr66, and Gly67.¹¹ There have been several reports of investigations into the mechanism of chromophore formation.^11,21,22 The currently preferred proposed mechanism invokes a branched pathway that can lead to either green or red chromophores.²³ In this branched mechanism, the formation of the chromophore starts with the cyclization of the main chain to form a five-membered ring intermediate. This five-membered ring intermediate undergoes an initial step of oxidation to form a hydroxylated cyclic imine, which equilibrates with a cyclic imine. At this point, the mechanism branches. One branch results from the dehydration of the hydroxylated cyclic imine and leads to formation of the green fluorescent chromophore. On the other branch, irreversible oxidation of the cyclic imine leads to an intermediate with blue fluorescence. Further dehydroxylation and dehydration lead to formation of the red fluorescent chromophore of DsRed (Fig. 3).^23,24 Insight into the mechanism of formation of the RFP chromophore, and the influence of its local environment on its spectral properties, opens new avenues for engineering FPs. Accordingly, a variety of RFP-derived colors have been engineered through engineering of the chromophore structure and its immediate environment, as will be described below.

Fig. 3

Branched pathway mechanism for red chromophore formation (and dead-end green chromophore formation) in DsRed.

1.3.

Classification of RFPs

FPs are now available in a wide range of colors spanning the visible spectrum.^25–27 Relative to the more blue-shifted FPs, RFPs have a number of inherent advantages. Specifically, they are spectrally distinct from the commonly used avGFP variants, which makes them particularly useful for multicolor imaging applications. In addition, red-shifted fluorescence is associated with reduced background autofluorescence, lower phototoxicity, and better tissue penetration due to lower absorption.^15,16,28 All other factors being the same, these properties should make RFPs superior probes for fluorescence imaging, particularly for in vivo applications.

Before further discussing RFPs and RFP-based biosensors, it is important to mention that all of the widely used FPs are the products of a combination of rational design and directed evolution. Engineering by rational design involves making changes to the amino acid sequence using insights derived from inspection of high-resolution protein structures, possibly supplemented with additional insights obtained from computer modeling. In practice, rational design alone rarely results in useful new FPs due to unanticipated negative effects, such as diminished protein folding efficiency. By contrast, directed evolution does not require prior information on the protein structure. Instead, random mutagenesis is carried out on the gene encoding the protein-of-interest to produce large libraries of mutants, which are then screened for variants with the desired properties. The power of directed evolution for protein engineering is well established but suffers from being relatively labor intensive and requiring an effective screening protocol. This approach may also lead to the accumulation of multiple “silent” mutations in addition to the beneficial ones. Nonetheless, the thoughtful combination of both strategies has significantly benefited not only the engineering of RFPs and RFP-based biosensors but also essentially all of the FP variants currently available.

For the sake of this review, we have categorized RFPs into three classes based on their fluorescence spectral profiles. These three classes are standard RFPs (i.e., short Stokes shift with emission maxima in the 550 to 620 nm range), far RFPs (i.e., emission maxima at $>620\mathrm{nm}$), and long Stokes shift (LSS) RFPs (Table 1). Phototransformable RFPs including photoactivatable RFPs,²⁹ photoswitchable RFPs,^30,31 and photoconvertible RFPs³² will not be further discussed in this review.³³ It is important to note that the blanket designation of all of these proteins as “red” is misleading, since many of them emit wavelengths of light that would appear orange to the naked eye.

Table 1

Properties of selected red fluorescent protein (RFPs).

1.4.

Standard RFPs

The class of standard RFPs can be further subdivided into the “orange” RFPs with emission maximum at 550 to 580 nm and the “red” RFPs with emission maximum at 580 to 620 nm. One of the most important advances in generating standard RFPs in the orange and red spectral regions occurred during further evolution of mRFP1. Gln66, the first amino acid in the chromophore-forming tripeptide, is a critical determinant of the spectral profile of mRFP1 derivatives. For example, the Gln66Met mutation of mRFP1.1 and mCherry (${\lambda }_{\mathrm{ex}}=587\mathrm{nm}$; ${\lambda }_{\mathrm{em}}=610\mathrm{nm}$) causes a slight red shift in the excitation and emission relative to mRFP1. Mutation of Gln66 to Cys or Thr was found to cause a blue shift in fluorescence. This observation inspired the development of mTangerine (${\lambda }_{\mathrm{em}}=585\mathrm{nm}$; Gln66Cys), mOrange (${\lambda }_{\mathrm{em}}=562\mathrm{nm}$; Gln66Thr) and mStrawberry (${\lambda }_{\mathrm{em}}=596\mathrm{nm}$; Gln66Thr).

The gene for the brightly fluorescent dimeric intermediate created during the process of DsRed monomerization was fused to a second copy of itself to create a “tandem dimer” RFP termed tdTomato (${\lambda }_{\mathrm{ex}}=554\mathrm{nm}$ and ${\lambda }_{\mathrm{em}}=581\mathrm{nm}$). Due to the formation of an intramolecular pseudodimer, tdTomato behaves like an exceptionally bright monomeric RFP, making this a popular tool for many applications.¹⁶

Later RFP engineering efforts focused on improving the photostability of the mFruit variants. Such efforts led to the production of mOrange2 (${\lambda }_{\mathrm{ex}}=549\mathrm{nm}$ and ${\lambda }_{\mathrm{em}}=565\mathrm{nm}$) with a 25-fold increase in photostability relative to mOrange.¹⁷ Due to their high brightness and photostability, mOrange and mCherry are generally the mFruit FPs of choice for live cell fluorescence imaging experiments when orange or red fluorescence is required. Several of the other mFruit FPs, including mTangerine and mStrawberry, are not often used for imaging as they suffer from low intrinsic brightness and poor photostability.

In addition to the DsRed-derived monomeric variants, a second lineage of standard monomeric RFPs was engineered from the sea anemone Entacmaea quadricolor RFPs eqFP578 and eqFP611.^48,34 For example, TagRFP (${\lambda }_{\mathrm{ex}}=555\mathrm{nm}$; ${\lambda }_{\mathrm{em}}=584\mathrm{nm}$) is a bright monomeric RFP engineered from the dimeric RFP eqFP578.³⁴ The Ser158Thr variant of TagRFP, designated as TagRFP-T, has improved photostability by approximately ninefold.¹⁷ Another member of the eqFP578 variant family is FusionRed (${\lambda }_{\mathrm{ex}}=580\mathrm{nm}$ and ${\lambda }_{\mathrm{em}}=608\mathrm{nm}$), which exhibits decreased cytotoxicity when expressed in mammalian cells.³⁷ The eqFP611 lineage has also yielded mRuby (${\lambda }_{\mathrm{ex}}=558\mathrm{nm}$ and ${\lambda }_{\mathrm{em}}=605\mathrm{nm}$) and the brighter mRuby2 variant (${\lambda }_{\mathrm{ex}}=559\mathrm{nm}$ and ${\lambda }_{\mathrm{em}}=600\mathrm{nm}$), which exhibit a relatively LSS ($\sim 50\mathrm{nm}$) between the excitation and emission maxima.^35,36

Although a growing number of engineered standard, far, and LSS RFPs have an “m” (as an abbreviation for monomeric) in front of their name, many do not behave as monomers when expressed in cells.⁴⁹ In some cases, the protein may form dimers or higher order oligomers, which can lead to aggregation and/or cytotoxicity.³⁷ Yet other RFPs, many of which are unambiguously monomeric, can form bright puncta in certain cells types due to accumulation in lysosomes or autophagosomes.^50,51

1.5.

Far RFPs

Far RFPs with an emission maximum over 620 nm are of particular importance for in vivo and deep-tissue imaging in small animal models such as mice and rats. The spectral region between 600 and 1200 nm, bounded at the low wavelength by hemoglobin absorption and at the long wavelength by the increasing absorption of water, is known as the near-infrared “optical window.” This “optical window” has motivated FP engineers to push the excitation and emission wavelengths of RFP into the far-red and even the near infrared region.²⁸

Early efforts in creating further red-shifted RFPs from mRFP1 yielded mPlum (${\lambda }_{\mathrm{ex}}=590\mathrm{nm}$ and ${\lambda }_{\mathrm{em}}=649\mathrm{nm}$), mRaspberry (${\lambda }_{\mathrm{ex}}=598\mathrm{nm}$ and ${\lambda }_{\mathrm{em}}=625\mathrm{nm}$),⁴⁰ and mGrape (${\lambda }_{\mathrm{ex}}=608\mathrm{nm}$ and ${\lambda }_{\mathrm{em}}=646\mathrm{nm}$).²⁸ mCherry has also served as a template for the engineering of longer wavelength emission with the aid of a computationally designed library.⁵² This effort led to the creation of mRouge with a relatively long wavelength emission (maximum at 637 nm) but relatively low brightness (quantum yield of 0.03). Generally speaking, the far RFPs derived from DsRed are relatively dim and have not proven particularly useful for in vivo applications.

Entacmaea quadricolor lineage has served as a more promising source of far RFPs than the DsRed lineage. For example, the eqFP578-derived mKate (${\lambda }_{\mathrm{ex}}=588\mathrm{nm}$ and ${\lambda }_{\mathrm{em}}=635\mathrm{nm}$), mKate2 (${\lambda }_{\mathrm{ex}}=588\mathrm{nm}$ and ${\lambda }_{\mathrm{em}}=633\mathrm{nm}$), and mNeptune (${\lambda }_{\mathrm{ex}}=600\mathrm{nm}$ and ${\lambda }_{\mathrm{em}}=650\mathrm{nm}$) variants were engineered to exhibit bright far-red emission above 630 nm.^28,38,39 Further efforts led to the development of two additional bright far-red mKate derivatives, mCardinal⁴³ and TagRFP657.⁴¹ These variants have excitation maxima of 604 and 611 nm, respectively, and 659-nm emission maxima for both. To date, the most red-shifted emission maximum for an RFP is 675 nm for the mKate variant TagRFP675.⁴²

1.6.

LSS RFPs

LSS typically refers to fluorophores for which the difference between the fluorescence excitation and emission maxima is larger than $\sim 100\mathrm{nm}$. The availability of LSS FPs provides researchers with a greater selection of spectrally resolvable colors for multicolor imaging application. LSS RFPs absorb blue light (usually in the range of $\sim 440$ to 460 nm) and fluoresce in the red region of the visible spectrum. They hold particular promise for two-photon fluorescence excitation due to the fact that they can be excited at the same two-photon wavelength as enhanced GFP (EGFP) using widely available pulsed laser systems.

The first reported LSS RFP, known as mKeima, was developed from a chromoprotein from the stony coral Montipora sp.⁴⁴ Later efforts were aimed at developing new LSS RFPs from standard and far RFPs. For example, three new LSS RFPs, LSSmKate1, LSSmKate2,⁴⁶ and mBeRFP⁴⁵ were engineered by providing an excited state proton transfer (ESPT) pathway for the mKate chromophore. Blue light excitation causes the chromophore to enter the excited state, which is associated with a decreased $\mathrm{p}{K}_a$ for the phenol group of the chromophore. Accordingly, following excitation of the neutral chromophore, the proton is transferred through the hydrogen bonds of the ESPT pathway to generate the lower energy excited state anionic state, which then emits red fluorescence.

LSSmKate1 and LSSmKate2 outperform mKeima in terms of pH-stability, photostability, and brightness. Unlike mKeima and mBeRFP, LSSmKates lack the additional excitation peak associated with the anionic ground state of the red chromophore (i.e., normal Stokes shift red fluorescence) at around 560 nm. The lack of this peak facilitates their combined use with standard RFPs in multicolor fluorescence imaging. The strategy of engineering ESPT pathways into standard RFPs has also been applied to some of the mFruit RFPs, including mOrange and mCherry, to generate variants with blue-shifted fluorescence excitation.⁵³ Further development produced LSSmOrange, which exhibits the highest brightness among all the LSS RFPs.⁴⁷

2.1.

Classification of RFP-Based Biosensors

Biosensors are generally composed of two parts: a molecular recognition/binding element that interacts with the target and a signal-transducing element that converts the interaction into a detectable signal, such as fluorescence. Several approaches have been employed to convert RFPs into effective signal-transducing elements for a variety of recognition events. Based on the design strategies employed, RFP-based biosensors can be categorized into four main classes: FP complementation-based biosensors, Förster resonance energy transfer (FRET)-based biosensors, dimerization-dependent FP (ddFP)-based biosensors, and single FP-based biosensors.

2.2.

RFP Biosensors Based on Complementation

Complementation-based biosensors are based on the interaction-induced reassembly of a complete and functional FP from two (or more) nonfunctional fragments. To use FP complementation to visualize a protein–protein interaction, one FP fragment is genetically fused to one gene of an interacting protein pair and the other fragment is fused to the other gene of the pair. Before complementation, both of the fragments are partially or fully unfolded and nonfluorescent. Interaction of the two fused partners brings the nonfluorescent FP fragments into close proximity and enables the formation of the functional FP (Fig. 4).^66,67 The first report of an RFP complementation system described one based on the Gln66Thr variant of mRFP1.⁶⁸ Later efforts used mCherry, mPlum, and mKate as the basis for FP complementation systems with longer emission wavelength and brighter fluorescence.^69–71 Recently, an mNeptune-based complementation system was introduced and successfully applied for in vivo imaging of RNA–protein and protein–protein interactions.⁷²

Fig. 4

Biosensor design based on RFP complementation. Two potentially interacting proteins are fused to the two fragments of a split RFP. Interaction between the two protein partners bring the RFP fragments in close proximity, leading to reconstitution of an intact RFP and a corresponding increase in red fluorescence.

All FP complementation systems are associated with some drawbacks and limitations, including background self-association and temperature sensitivity.⁷³ As FP complementation is effectively irreversible, it is useful for trapping both constitutive and transient protein–protein interactions. However, as formation of a mature chromophore in the reconstituted FP usually requires tens of minutes,^71,74 FP complementation is unsuitable for dynamic visualization of reversible protein–protein interactions. Fortunately, the limitations imposed by the irreversibility and slow kinetics can be overcome using alternative biosensing strategies such as FRET or ddFPs.

2.3.

RFP Biosensors Based on FRET

FRET is the phenomenon of radiationless energy transfer via dipole–dipole interaction between two chromophores that have compatible energy levels and are close in distance ($<10\mathrm{nm}$). The basic design principle of all FRET-based biosensors is to couple a specific binding event or covalent modification of a protein to a change of the energy transfer efficiency between the higher energy donor FP and the lower energy acceptor FP. A variety of FP FRET-based biosensors for detection of protein–protein interaction, ion concentrations, small molecule concentrations, and enzyme activities have been developed (Fig. 5).⁷⁵

Fig. 5

Representative Förster resonance energy transfer (FRET)-based biosensors with RFPs. (a) Intermolecular biosensors for protein–protein interaction. Unlike FP complementation-based biosensors, the FRET-based biosensors of protein–protein interactions are reversible. (b) Ion/small molecule biosensors. An intramolecular protein complex is formed, or a conformation changed, upon the binding of a specific ion or small molecule. (c) Protease biosensors where the two FPs are initially linked by a protease substrate sequence.

The design of intermolecular FRET-based biosensors for protein–protein interaction detection is similar to that of FP complementation. However, rather than having the interacting proteins of interest fused to the FP fragments, they are fused to the donor FP and acceptor FP. FRET efficiency increases when the two protein partners interact to form a complex. For biosensor designs intended for detection of a protein conformational change, an intramolecular FRET-based biosensor can be constructed by linking both donor and acceptor FPs in a single polypeptide. The intramolecular biosensor design offers a more consistent signal output due to the fixed ratio of donor and acceptor concentrations in different cells.

The cyan and yellow FP-based FRET donor and receptor pair is an excellent choice for the construction of genetically encoded FRET biosensors due to the large spectral overlap and their relatively high brightness. However, the development of various monomeric RFP variants has now provided new and exciting possibilities to construct red-shifted FRET pairs. For example, mRuby2, currently one of the brightest monomeric RFPs, has been paired with a bright GFP variant, Clover.³⁶ This new FRET pair confers a greater dynamic range and photostability compared to various existing cyan FP (CFP)- and yellow FP (YFP)-based FRET biosensors.

The availability of new RFPs also provides opportunities to construct new FRET pairs with novel spectral properties. One of the main justifications for such efforts is to achieve spectral compatibility with the CFP–YFP pair.^76,77 For example, the mOrange–mCherry pair is spectrally orthogonal to the CFP–YFP pair, though FRET biosensors based on this pair tend to have only modest signal changes.^78,79 The orange–red FRET pair was recently improved by developing self-associating variants of mOrange and mCherry by reversion of the hydrophobic dimeric interface breaking mutations.^80,81 LSS mOrange and mKate2 are yet another orange–red FRET pair that has been simultaneously imaged with a CFP–YFP FRET pair using a single laser excitation wavelength.⁴⁷

Generally speaking, the single most important advantage of FRET-based biosensors is that they provide ratiometric fluorescent changes that can typically be calibrated, making this class of indicators most appropriate for quantitative imaging. The single major disadvantage of FRET-based biosensors is that the fluorescent changes are often quite small (as low as a few percent, though there are some with much larger changes).⁸² A second disadvantage is that FRET-based biosensors require two distinct emission channels for ratiometric imaging, making it challenging to use more than one type of biosensor in a single experiment.

2.4.

RFP Biosensors Based on ddFPs

ddFPs are a relatively recent addition to the FP toolbox that provide an alternative platform for biosensor design.^83,84 The ddFP strategy is based on a pair of FPs, engineered from dTomato, which exhibit minimal to no fluorescence in their monomeric states. Upon heterodimerization, the chromophore environment of one FP is modified such that the anionic state of the chromophore is stabilized, leading to an increase in red fluorescence. The first ddFP to be engineered was a red variant (ddRFP) with a 10-fold increase in red fluorescence intensity upon heterodimer formation. The dimerization-dependent fluorescence change of ddRFP was used for detection of reversible ${\mathrm{Ca}}^{2+}$-dependent association of calmodulin (CaM) and M13 in live cells, as well as imaging of caspase-3 activity during apoptosis (Fig. 6).⁸⁴ Green and yellow ddFP pairs were later engineered and applied for detection of membrane–membrane contacts at the mitochondria associated membrane.⁸³

Fig. 6

ddRFPs and ddRFP-based biosensors. (a) Fluorescence intensity increase upon the formation of heterodimeric ddRFP pair. (b) ddRFP-based caspase-3 biosensor. (c) ddRFP-based ${\mathrm{Ca}}^{2+}$ biosensor.

Although conceptually analogous to FP complementation, the advantage of ddFP lies in the reversibility of heterodimer formation. Accordingly, they can be used to visualize dynamic and reversible protein–protein interactions in live cells, similar to how FRET is used. Compared to FRET-based biosensors, ddFPs do have an inherent advantage for multiparameter imaging. Specifically, a ddFP occupies just one color channel (i.e., green, yellow, or red) while FRET-based biosensors occupies two (i.e., donor and acceptor). One drawback of ddFPs is that they will spontaneously dimerize at relatively high concentrations (above $10\mu \mathrm{M}$). By contrast, FRET pairs have only a weak tendency to dimerize and the dissociation constants are typically much higher ($>100\mu \mathrm{M}$).⁸¹

2.5.

RFP Biosensors Based on a Single FP

As their name implies, single FP-based biosensors contain only one engineered FP signal-transduction domain. The biosensor is engineered such that the FP responds to the biochemical stimulus of interest with a reversible change in fluorescence intensity (intensiometric), excitation spectral profile (excitation ratiometric), or emission spectral profile (emission ratiometric). The major advantage of single FP-based biosensors is that they typically exhibit a substantially larger intensity change at a single wavelength than a FRET-based biosensor. Furthermore, single FP-based biosensors have the benefit of using a smaller region of the visible spectrum window, enabling the simultaneous use of more than one fluorophore color. Yet another advantage relative to intramolecular FRET-based biosensors is the smaller protein size.

One way to create a single FP-based biosensor is to take advantage of the intrinsic sensitivities of certain FP variants [Fig. 7(a)]. For example, all FPs exhibit some pH dependence and some have apparent $\mathrm{p}{K}_a\mathrm{s}$ close to physiologically relevant pH values. Among the many examples of such FP-based pH biosensors, the most widely used are the pHlourin variants of avGFP.⁵⁶ The chloride ion sensitivity of YFP is another example of intrinsic FP sensitivity.^85,86 It has also been proven possible to rationally engineer intrinsic sensitivity into an FP by incorporating an analyte binding site directly on the exterior of the FP barrel. For example, the reduction/oxidation sensitive roGFP^87,88 and the calcium ion (${\mathrm{Ca}}^{2+}$) sensitive CatchER were engineered in this way.⁸⁹

Fig. 7

Single FP-based biosensors. (a) Single FP-based pH biosensor based on intrinsic sensitivity. (b) Single FP-based ${\mathrm{Ca}}^{2+}$ biosensor with an extrinsic ${\mathrm{Ca}}^{2+}$ binding domain.

The majority of intrinsic single FP-based biosensors are green or yellow fluorescent, and only a few intrinsic red single FP-based biosensors have been described. Examples include mNectarine, which was applied for detection of nucleoside transport,⁹⁰ and pHTomato, which was used to report synaptic neurotransmitter release at nerve terminals.⁵⁷ An excitation ratiometric pH biosensor, pHRed, was engineered from LSS RFP mKeima and used to image energy-dependent changes of cytosolic and mitochondrial pH.⁵⁸ Recently, a pH-sensitive RFP, known as pHuji, was engineered from mApple and used for imaging of endocytosis and exocytosis.⁵⁹

Another strategy for engineering a single FP-based biosensor is to genetically incorporate an extrinsic analyte recognition domain into the FP. The extrinsic domain is typically fused to one of the termini or inserted into a solvent-exposed region of the FP in order to minimize disruption of the protein structure. For the FP to work as an effective signal transducer, the extrinsic recognition domain must be in relatively close proximity to the chromophore to allosterically modulate the chromophore environment upon interaction with the target analyte. It is important to note that the FP chromophore is well protected in the center of the barrel structure, and the termini and loop region are relatively distant from the chromophore. An extrinsic recognition domain fused to one of the termini is unlikely to have much influence on the chromophore environment. To circumvent this problem, researchers rely on the strategy of circular permutation (Fig. 8). Circularly permutated FPs are generated by genetically linking the original N- and C-termini with a short polypeptide linker and introducing new N- and C-termini at a position elsewhere in the protein.^91,92 For FP-based biosensor construction, the new N- and C-termini are introduced close to the chromophore such that conformational changes in the extrinsic recognition element cause alterations in the chromophore environment and, correspondingly, in the fluorescence intensity or hue of the FP.^93–96

Fig. 8

Schematic presentation of FP circular permutation at both the DNA and protein levels.

3.1.

RFP-Based Ca²⁺ Biosensors

${\mathrm{Ca}}^{2+}$ is the principal secondary messenger associated with neuronal signaling pathways and is reliably elevated during the firing of action potentials. Accordingly, FP-based ${\mathrm{Ca}}^{2+}$ biosensors are exceptionally useful for the imaging of neuronal activity in contexts ranging from in vitro cultured cells to in vivo brain activity in behaving animals.

Over the last decade, the GCaMP-type single FP-based biosensors have emerged as the predominant technology for in vivo imaging of neuronal activity.^100,101 GCaMP is composed of cpGFP with M13 and CaM fused to the N- and C- termini, respectively. Structural studies reveal that in its ${\mathrm{Ca}}^{2+}$ free state, the fluorescence is quenched because the chromophore is exposed to bulk solvent. In the presence of ${\mathrm{Ca}}^{2+}$, CaM wraps around M13 and forms a new interaction with the chromophore that stabilizes the phenolate (i.e., the fluorescent form) state.

Following the GCaMP-type design strategy, a red ${\mathrm{Ca}}^{2+}$ biosensor known as R-GECO1⁶⁰ was created by replacing the cpGFP in an improved GCaMP variant with a circularly permuted variant of mApple. R-GECO1 was further optimized and engineered into spectrally diversified and low-affinity variants, including an improved R-GECO1.2, a blue-shifted O-GECO, a red-shifted CAR-GECO,¹⁰² a highlightable GR-GECO,¹⁰³ an LSS REX-GECO¹⁰⁴ and low-affinity red-GECO variants.¹⁰⁵ RCaMP, a similar single RFP-based ${\mathrm{Ca}}^{2+}$ biosensor, was engineered from the cpmRuby template.¹⁰⁶ Further improved variants of R-GECO1, confusingly named R-CaMP1.07 and R-CaMP2, have recently been reported.^107,108

The development of mApple-based R-GECO1⁶⁰ and mRuby-based RCaMP,¹⁰⁶ has unlocked new opportunities for simultaneous multicolor optical imaging for neural activities as well as integration of optogenetics for orthogonal activation and measurement. For example, R-GECO1 has been used to report neural activity in vivo in the zebrafish retinotectal system, with comparable performance to the green ${\mathrm{Ca}}^{2+}$ biosensor GCaMP3.¹⁰⁹ CAR-GECO1, a red-shifted variant based on R-GECO1, was used for optogenetic activation and ${\mathrm{Ca}}^{2+}$ imaging concurrently in combination with channelrhodopsin-2(T159C)-EGFP in mouse neocortical slice culture.¹⁰² RCaMP, along with green glutamate sensor, was used for imaging synaptic input and output in Caenorhabditis elegans neurons.⁹³ The LSS REX-GECO1 was used in the eye and optic tectum of albino Xenopus laevis tadpoles for two-photon fluorescence imaging of ${\mathrm{Ca}}^{2+}$ dynamics in vivo.¹⁰⁴

Due to an inherent tendency of mApple to undergo photoactivation (i.e., a temporary increase in brightness that can be easily confused with a true ${\mathrm{Ca}}^{2+}$ elevation) with blue light, one must be cautious when using R-GECO series indicators with optogenetic tools requiring violet/blue activation light.¹⁰² In comparison, this photoactivation effect was not observed from mRuby-based RCaMP series indicators. Therefore, RCaMP should be a better-suited ${\mathrm{Ca}}^{2+}$ indicator for use with ChR2 or other violet/blue light activatable optogenetic tools.

3.2.

RFP-Based Voltage Biosensors

For imaging of neuronal activity, the signals obtained from an FP-based ${\mathrm{Ca}}^{2+}$ biosensor are, necessarily, only a surrogate for action potentials. Nevertheless, this indirect measure has proven to be very useful, largely because ${\mathrm{Ca}}^{2+}$ biosensors have traditionally been far superior to voltage biosensors in terms of their brightness and magnitude of fluorescence response. The tradeoffs associated with the reliance on ${\mathrm{Ca}}^{2+}$ signals (which have much slower temporal dynamics than voltage changes) are that neither fast series of spiking events, nor subthreshold voltage changes, can be visualized. In order to overcome these limitations, biosensors that directly report on membrane voltage are needed.¹¹⁰ Accordingly, the FP research community has been pursuing the development of voltage indicators for as long as they have been pursuing FP-based ${\mathrm{Ca}}^{2+}$ biosensors, though with more modest success to date.

Both FRET-based and single FP-based voltage sensors for imaging of membrane potential changes in neurons have been reported. These indicators are constructed by tethering an FP, or a FRET pair of FPs, to a voltage-sensitive membrane protein, such that a voltage-dependent conformation change alters either the brightness of the FP or the FRET efficiency, respectively. Some notable examples include FlaSh/Flare,^111,112 SPARC,¹¹³ and the voltage-sensitive FPs.^110,114,115 These indicators have undergone improvements resulting in variants with faster kinetics,¹¹⁶ improved cell surface targeting,¹¹⁷ and larger signal changes.¹¹³ Despite these improvements, voltage indicators have been notoriously challenging to apply in research applications, especially when judged against highly optimized and robust GCaMP-type ${\mathrm{Ca}}^{2+}$ indicators. One of the most pressing issues with FP-based voltage sensors was their relatively small signal changes, with all sensors reported prior to 2012 exhibiting maximal fluorescence changes of $<10\%$.^118,119

In 2012, Jin et al.¹²⁰ reported a GFP-based voltage biosensor, Arclight, with an unprecedented 35% decrease in fluorescence intensity in response to a 100 mV depolarization. Arclight provides sufficient brightness and signal change to enable detection of single action potentials and subthreshold activities in individual neurons and dendrites, although with relatively slow response kinetics. Further engineered Arclight variants provided faster kinetics but at the expense of reduced signal changes.¹²¹ Accelerated sensor of action potentials 1 (ASAP1) is another recently developed green fluorescent voltage sensor.⁹⁵ As its name implies, it offers faster kinetics relative to Arclight and enables continuous monitoring of membrane potential in neurons at kilohertz frame rates using standard epifluorescence microscopy.

Efforts to develop red fluorescent voltage indicators have lagged behind the efforts to develop green ones. VSFP_cpmKate, VSFP3.1_TagRFP, and VSFP3.1_mKate2 are some examples of voltage indicators that emit in the red region of the visible spectrum.^122,123 However, the fluorescence brightness, response amplitude, and kinetics of these red-shifted VSFPs are not comparable to that of Arclight or ASAP1. In unpublished work, our group has developed a red fluorescent voltage biosensor, designated FlicR1, which is based on the voltage-sensing domain of Arclight and the cpmApple of R-GECO1 (Ahmed Abdelfattah, unpublished results).

3.3.

RFP-Based Synaptic Transmission Biosensors

Yet another important application of FP-based biosensors is the detection of synaptic transmission. The first FP designed for the purpose of detecting vesicle fusion at the synapse was synapto-pHluorin.^56,124 To engineer synapto-pHluorin, a pH-sensitive variant of avGFP, known as superecliptic pHluorin (SEP),¹²⁵ was fused to the luminal side of the vesicular protein, synaptobrevin. SEP is initially quenched by the acidic conditions of the vesicle lumen but increases in fluorescence $\sim 20$-fold upon release of the vesicle contents following fusion with the plasma membrane. Fusing SEP to proteins highly localized to synaptic vesicles, such as synaptophysin¹²⁶ or the glutamate transporter VGlut1,¹²⁷ resulted in improved signal-to-noise ratios.

As with other classes of biosensors, efforts to develop an RFP-based biosensor of synaptic fusion lagged far behind the development of the GFP-based biosensor. By taking advantage of the pH-sensitive property of the orange FP mOrange2, a red-shifted biosensor was constructed by fusion to VGlut1.¹²⁸ Designated as VGlut1-mOrange2, this probe was used in conjunction with GCaMP3 to simultaneously image synaptic vesicle recycling and changes in cytosolic ${\mathrm{Ca}}^{2+}$. In a similar application, the pH-sensitive pHTomato RFP was coexpressed with GCaMP3 for concomitant imaging of neurotransmitter release and presynaptic ${\mathrm{Ca}}^{2+}$ transients at single nerve terminals.⁵⁷ Coexpression of pHTomato and ChR2 provided an all-optical approach for multiplex control and tracking of distinct circuit pathways.

Another approach to visualizing synaptic transmission is to detect the neurotransmitter itself. For example, the genetically encoded biosensor GluSnFR is a FRET-based biosensor that incorporates the periplasmic glutamate-binding protein GltI as a molecular recognition element.¹²⁹ A single FP-based green glutamate biosensor called iGluSnFR was also engineered by insertion of a cpGFP into the glutamate-binding domain GltI.⁹³ In unpublished work, our group has converted iGluSnFR⁹³ into a red fluorescent variant by substituting the cpGFP with the cpmApple domain from R-GECO1 (Jiahui Wu, unpublished results).