From fluorescent proteins to fluorogenic RNAs: Tools for imaging cellular macromolecules

Abstract

The explosion in genome‐wide sequencing has revealed that noncoding RNAs are ubiquitous and highly conserved in biology. New molecular tools are needed for their study in live cells. Fluorescent RNA–small molecule complexes have emerged as powerful counterparts to fluorescent proteins, which are well established, universal tools in the study of proteins in cell biology. No naturally fluorescent RNAs are known; all current fluorescent RNA tags are in vitro evolved or engineered molecules that bind a conditionally fluorescent small molecule and turn on its fluorescence by up to 5000‐fold. Structural analyses of several such fluorescence turn‐on aptamers show that these compact (30–100 nucleotides) RNAs have diverse molecular architectures that can restrain their photoexcited fluorophores in their maximally fluorescent states, typically by stacking between planar nucleotide arrangements, such as G‐quadruplexes, base triples, or base pairs. The diversity of fluorogenic RNAs as well as fluorophores that are cell permeable and bind weakly to endogenous cellular macromolecules has already produced RNA–fluorophore complexes that span the visual spectrum and are useful for tagging and visualizing RNAs in cells. Because the ligand binding sites of fluorogenic RNAs are not constrained by the need to autocatalytically generate fluorophores as are fluorescent proteins, they may offer more flexibility in molecular engineering to generate photophysical properties that are tailored to experimental needs.

Short abstract

PDB Code(s): 6E8S, 6E8T and 6E8U

Introduction

The discovery and subsequent application of green fluorescent protein (GFP) from the jellyfish Aequorea victoria revolutionized the study of biology by providing visual access to the cellular environment.1, 2, 3 Genetically encoded fluorescent protein (FP) tags allowed cellular processes to be monitored and analyzed in real time. Since their initial discovery, considerable effort has been dedicated to engineering the biophysical and biochemical properties of FPs for improved stability, enhanced brightness, and an extended fluorescent spectrum. This has led to the development of blue FP, cyan FP, yellow FP (YFP), and red FP (RFP) in addition to superfolding FPs, far‐red FP probes, split FPs, and photoactivated FPs, among others.4, 5, 6, 7, 8, 9 More recently, these FP tags have inspired a new branch of molecular imaging in the realm of RNA.

No RNAs with bright intrinsic fluorophores have been discovered. Instead, RNA mimics of FPs are molecules that bind to cognate small molecules, that is, aptamers.10 These aptamers recognize compounds that are minimally fluorescent on their own and increase their fluorescence by as much as 5000 times upon binding.11, 12 All such described fluorescence turn‐on aptamers are the result of in vitro evolution and design experiments. In principle, the biophysical properties of fluorogenic RNA–fluorophore complexes can be tuned more broadly than those of FPs, since neither the chemical nature of the conditionally fluorescent small molecules nor the structure of the turn‐on aptamer RNAs is constrained by evolutionary history. Consistent with this, and although the imaging potential of fluorescence turn‐on aptamers as tags for cellular RNAs was recognized only recently, the spectral range covered by aptamer RNA–fluorophore complexes is already comparable to that of the FPs (Fig. 1).

Figure 1.

Comparison of spectral range for chromophores found in FPs (top) and fluorogenic RNAs (bottom). Depicted colors reflect the visual appearance of chromophores in complex with their respective FP/RNA rather than their excitation or emission wavelengths.

Fluorescent Proteins

GFP adopts the now classic 11‐stranded β‐barrel fold (Fig. 2), which provides the environment for spontaneous formation of 4‐(p‐hydroxybenzylidene)‐5‐imidazolinone (p‐HBI; Fig. 1) from three amino acid residues (Ser65, Tyr66, and Gly67) within the protein core.13, 18 Amino acid residues surrounding the intrinsic p‐HBI restrain the rotational and vibrational motion of its photoexcited state, reducing nonradiative decay and enhancing fluorescence (Fig. 3).19, 20 Consistent with this, denaturation of GFP destroys fluorescence, and free p‐HBI in solution is nonfluorescent.21, 22 Wild‐type GFP displays two distinct absorption bands that arise from the neutral (phenol) and anionic (phenolate) forms of p‐HBI (395 and 475 nm, respectively). Mutations stabilizing the phenolate form of p‐HBI, such as those in enhanced GFP (EGFP), result in improved brightness.6, 23, 24

Figure 2.

Comparison of overall structures of GFP and fluorogenic RNA aptamers. (a) Characteristic 11 strand β‐barrel of GFP containing buried fluorophore.13 (b) Structure of Spinach aptamer with bound DFHBI ligand, featuring a G‐quadruplex fluorophore binding core flanked by two helices.14 (c) Homodimer structure of Corn aptamer with DFHO ligand bound between the G‐quadruplex faces of each Corn protomer.15 (d) Mango‐I structure with TO1–biotin fluorophore bound in the G‐quadruplex core, flanked by a short helix.16 (e) DIR2s structure with bound OTB fluorophore between base triple and single unpaired nucleotide in the terminal region of aptamer.17 Compared to GFP, which has a buried fluorophore, fluorogenic RNA aptamers feature solvent‐exposed binding pockets.

Figure 3.

Comparison of fluorophore binding sites of (a) HBI within GFP,13 with (b) DFHBI in Spinach,14 (c) DFHO in Corn,15 (d) TO1–Biotin in Mango‐I,16 and (e) OTB in DIR2s.17 The fluorophore pockets for all molecules are highlighted, and hydrogen bonds are depicted by purple dashed lines. Stacking interactions in the fluorogenic RNA aptamers are highlighted by colored van der Waals surfaces.

Formation of p‐HBI is initiated by nucleophilic attack of the amide of Gly67 on the carbonyl of Ser65. This intramolecular cyclization is followed by dehydration and oxidation to yield the conjugated fluorophore.25, 26, 27 Modification of the reactive amino acids, as well as their environment, gives rise to the diverse intrinsic fluorophores and fluorescent properties of related FPs. The blue‐shifted enhanced blue FP and enhanced cyan FP arise from mutation of Tyr66 to histidine or tryptophan, respectively.27, 28, 29 Mutating an adjacent amino acid (Thr203Tyr) promotes π‐π stacking between Tyr66 and Tyr203. This causes a 20 nm bathochromic shift, as seen in YFP.30 The intrinsic fluorophores of RFPs, isolated from corals, arise from a process similar to that of GFP. After intramolecular cyclization and an initial oxidation, a second oxidation step kinetically competes with a dehydration step (which would form the green chromophore in GFP) in order to form the acylimine seen in the extended conjugation of the RFP chromophore (similar to the chromophore found in mCherry; Fig. 1).31, 32

Considerable effort has been devoted to engineering improved folding, stability, and oligomerization state of FPs for imaging applications. The improved variant EGFP retains its brightness at 37°C and is commonly used for imaging and microscopy.33 Many FPs form dimers or tetramers in solution.34 Studies that employ FP fusion proteins can be disrupted by oligomerization. Monomeric variants of FPs have been developed by mutating residues at the oligomerization interface.35 In some instances, these mutations abolish the FPs' intrinsic fluorescence and require further mutational engineering to recover fluorescence.36 Psudeomonomeric FPs have also been developed to address the challenges of oligomerization and aggregation, where two monomer FPs are connected using a flexible peptide linker. This intentional dimerization reduces the likelihood of further oligomerization during use.37

Fluorescence Activating Proteins

Although FPs with intrinsic fluorophores have long been the gold standard for cellular imaging, proteins that bind and induce fluorescence of conditionally fluorescent small molecules (fluorogen‐activating proteins [FAPs]) have also been engineered. FAPs differ from FPs resulting from tagging approaches (e.g., Halo, SNAP, CLIP, and smURFP tags)38, 39, 40, 41 in that the fluorophores (either endogenous or exogenous to cells) are noncovalently bound. Endogenous fluorophores that reversibly bind to FAPs include flavin mononucleotide (FMN) and bilirubin.42, 43 Bacterial photoreceptors comprised of light, oxygen, and voltage (LOV) sensing domains that naturally bind FMN form a covalent adduct between FMN and a cysteine residue within the protein, ultimately quenching fluorescence. LOV‐based FPs were engineered to prevent the covalent bond formation, maintaining the natural‐induced fluorescence of FMN. LOV‐based FAPs excite at 450 nm and emit at 495 nm. UnaG is another protein that binds an endogenous chromophore, bilirubin, to induce fluorescence excitation and emission at 498 and 527 nm, respectively.42 Bilirubin is naturally occurring in mammalian cells, but use of UnaG in bacterial cells which do not produce it requires the chromophore to be exogenously introduced. FAPs that rely on synthetic fluorophores such as malachite green (MG), thiazole orange (TO), and dimethylindole red (DIR) derivatives have been described.44, 45 These FAPs are single‐chain antibodies selected using FACS. A fluorescence‐activating and absorption‐shifting tag was developed through directed evolution of the bacterial photoreceptor photoactive yellow protein, which noncovalently binds to and induces the fluorescence of hydroxybenzylidene rhodanine derivatives.46

Fluorogenic RNA Aptamers

The invention of RNA in vitro selection (SELEX) led to the identification of numerous small‐molecule‐binding aptamers.10, 47, 48 The MG aptamer (MGA) was originally selected in order to generate a light‐activated RNA‐cleaving agent, as illumination of MG produces hydroxyl radicals.49 Subsequently, it was found that upon binding, the triphenylmethane dye exhibited a 2300‐fold increase in fluorescence.50 Structure determination demonstrated that the dye is sandwiched between a base pair and base quadruple.51 By restricting rotation around the central carbon bond connecting the phenyl rings, the aptamer reduces the fluorophore's ability to undergo nonradiative decay and thus enhances its fluorescence. The MGA provided proof of principle for fluorescent turn‐on aptamers, but its photoactivated RNA cleaving properties (and attendant light‐induced cytotoxicity) limited imaging applications. Turn‐on aptamers for derivatives of DIR and Hoechst were also identified.52, 53 Although these did not exhibit photoactivated RNA‐cleavage, their application to live‐cell imaging was limited, either because of modest fluorescence turn‐on (DIR, 60‐fold)53 or nonspecific nucleic‐acid binding by the fluorophores (Hoechst derivative).52

Spinach is an RNA Mimic of GFP

In 2011, the Jaffrey group reported the isolation and characterization of a variety of turn‐on aptamers selected for binding to p‐HBI derivatives.54 A 98 nt aptamer in complex with 3,5‐difluoro‐4‐hydroxybenzylidene (DFHBI), was the brightest, achieving a ~1000‐fold turn‐on and a quantum yield of 0.72, thus being ~50% as bright as EGFP (Table 1). This RNA, termed Spinach, was selected to bind preferentially to DFHBI (Fig. 1), whose two fluorine substituents lower the pK _a of the fluorophore, which thus exists in the brighter phenolate form (analogous to the ionization state of p‐HBI within EGFP). Spinach was used successfully as a live cell RNA‐imaging tag by expressing ribosomal RNA–Spinach fusions in live cells that were soaked in the cell‐permeable and noncytotoxic DFHBI, demonstrating the potential of this aptamer–fluorophore complex as a partially genetically encoded in vivo fluorescent tag.

Table 1.

Summary of Spectral Properties of Fluorogenic RNA Aptamers in Complex with Their Cognate Ligands

RNA aptamer	Ligand	λ ex (nm)	λ em (nm)	ε (M −1 cm−1)	Φ	Brightness	K D (nm)	Length (nt)	Ref.
Spinach	DFHBI	466	503	24,300	0.72	17,000	537	98	54
Spinach2	DFHBI	445	501	26,100	0.7	18,000	1450	95	55
	DFHBI‐1 T	482	505	31,000	0.94	29,000	560	95	56
iSpinach	DFHBI	442	503	26,100	0.98	25,000	920	69	57
Broccoli	DFHBI‐1 T	472	507	29,600	0.94	28,000	360	49	56
Chili	DMHBI‐Imd	413	542	20,000	0.08	1,600	71	52	58
	DMHBI+	456	592	21,000	0.4	8,400	63	52	58
	DMHBO+	463	594	22,000	0.1	2,200	12	52	58
Corn	DFHO	505	545	29,000	0.25	7,300	70	36(x2)	59
Orange Broccoli	DFHO	513	562	34,000	0.28	9,500	230	60	59
Red Broccoli	DFHO	518	582	35,000	0.34	12,00	206	60	59
Mango‐I	TO1	510	535	77,500	0.14	11,000	3.2	29	60
	TO3	637	658	9,300	‐	‐	5.1	29	60
Mango‐II	TO1	510	535	77,500	0.22	16,000	0.7	30	61
Mango‐III	TO1	510	535	77,500	0.56	43,000	5.6	31	61
Mango‐IV	TO1	510	535	77,500	0.41	32,000	11.1	30	61
DIR2s	OTB	380	421	73,000	0.51	37,000	662	57	62
	DIR‐Pro	600	658	164,000	0.33	54,000	252	57	62

Crystallographic structure determination of Spinach–DFHBI complexes revealed an RNA comprised of a single coaxial helical stack organized around a G‐quadruplex motif at its core [Fig. 2(b)].14, 63 The presence of a G‐quadruplex (Box 1) in Spinach was unexpected, because the aptamer lacks any of the canonical sequence signatures of G‐quadruplexes (such as consecutive stretches of G‐rich sequences) and computational secondary structure prediction had suggested only the presence of duplexes.54 The bound DFHBI is sandwiched between the top tetrad (quartet) of the G‐quadruplex and a base triple and is further restrained by an unpaired guanine nucleobase that is coplanar with the two heterocycles of the fluorophore [Fig. 3(b)]. Mutagenesis demonstrated the importance of the G‐quadruplex, base triple and unpaired guanine for fluorescence turn‐on, and NMR analysis confirmed that the G‐quadruplex is present in solution and that it is stabilized by DFHBI binding.14, 63 A crystal structure of the fluorophore‐free RNA indicated that the overall structure of the aptamer is preformed (as did SAXS analysis),14, 63 and that a nucleotide that is extruded from the helical stack in the DFHBI‐bound state occupies the fluorophore binding pocket in the absence of the small molecule.

Box 1.

Structural overview of conventional G‐quadruplexes. (a) The G‐tetrad or G‐quartet is comprised of four Hoogsten‐bonded guanines on a plane. A cation, typically K⁺, is coordinated in the axial pore. (b) Each glyosidic bond in the G‐quadruplex can adopt either anti or syn conformations. (c) G‐quadruplexes are comprised of two or more tiers of G‐tetrads and can be oligomolecular or unimolecular. Strand orientations can be locally parallel or antiparallel, and loops connecting the stacked guanines can form propeller, lateral, or diagonal connections. Note how the connectivity restrains the possible glycosidic bond angles of the component guanine residues.

The presence of a G‐quadruplex in the fluorophore binding site of a turn‐on aptamer can be rationalized on the basis of three properties of this nucleic acid structural motif. First, by virtue of the cyclic symmetry of G‐quartets, as well as the axial coordination of a cation (Box 1), the exposed nucleobases of G‐quadruplexes tend to be coplanar. This is in contrast to the ends of Watson–Crick duplexes or triplexes that often exhibit pronounced propeller twist or buckling. As maximal fluorescence activation requires that the two heterocycles of the bound DFHBI (and similar fluorophores) be restrained on one plane, coplanarity of the exposed nucleobases of G‐quadruplexes could be advantageous for the binding site of turn‐on aptamers. Second, the exposed planar faces of G‐quadruplexes are larger than those of duplexes and triplexes, and this allows positioning of additional structural elements on the plane of the fluorophore that is stacking on the quadruplex. This is the case for the unpaired guanine in Spinach, which also hydrogen bonds to the bound DFHBI using a ribose from the RNA backbone. Third, axial cation coordination provides increased thermodynamic stability to G‐quadruplexes over simple duplexes. This is especially the case under physiological conditions, because most G‐quadruplexes (including Spinach) preferentially bind K⁺, which is the major intracellular monovalent cation.

Despite its large fluorescence turn‐on and overall brightness, Spinach has several shortcomings as a fluorescent tag for in vivo imaging of cellular RNAs. Spinach exhibits low thermostability and a propensity to misfold. These two properties may arise from the complex, nonstandard nature of its central G‐quadruplex. The structure of Spinach coaxially stacks an A‐form duplexes on each side of a G‐quadruplex. This necessarily implies complex connectivity of the RNA main chain, as it must transition from an antiparallel duplex (A‐form), to a four‐stranded structure (G‐quadruplex), back to an antiparallel duplex. Indeed, the G‐quadruplex of Spinach is among the most structurally complex described to date, comprised of two G‐tetrads, one noncanonical quartet, several nonconsecutive guanines, a combination of parallel and antiparallel connectivity, and variable loop lengths, the longest of which is 34 nucleotides. In addition, Spinach‐DFHBI exhibits poor photostability, losing fluorescence rapidly upon illumination (however, in the dark and in the presence of DFHBI, Spinach can rapidly recover fluorescence).64 This may arise from relatively weak binding to the fluorophore (K _d = 537 nM; Table 1) and the fact that the fluorophore is exposed to bulk solvent [Fig. 2(b)]. Lastly, as the autofluorescence of cells is in the blue–green range of the visible spectrum, fluorescent tags that are more red‐shifted would be preferable.

Several experiments to improve Spinach properties through further SELEX as well as functional reselection have been reported. The melting temperature (T _m) of Spinach is 34°C, limiting its application in live cells much as the limited thermal stability of wild‐type GFP does. Mutational optimization by reducing bulges and mismatches resulted in Spinach2.55 Although Spinach2 has a T _m of 38°C and was 2.8‐fold brighter than Spinach at 37°C, still only 37% of the improved aptamer was folded at this temperature. To improve its in vivo performance, Spinach2 was fused to tRNA^Lys3, using it as a folding scaffold (tSpinach2). This increased the fraction of folded aptamer to 60% at 37°C and overall brightness by 3‐ to 20‐fold relative to Spinach. Broccoli is an improved variant isolated by further in vitro and in vivo selection,56 and a selection experiment based solely on in vivo fluorescence also yielded another variant.65 A selection experiment employing microfluidic‐assisted in vitro compartmentalization yielded another variant termed iSpinach that is optimized for in vitro applications.57 Despite these efforts, the improved variants retain the fundamental structure of Spinach, with its complex fold and modest affinity for fluorophore. In order to achieve drastic improvements, several groups have turned to independent RNA scaffolds and different conditional fluorophores.

Corn and Chili Activate Red‐Shifted FP‐Related Fluorophores

Initial efforts to achieve red‐shifted emission relative to Spinach led to the DFHBI derivatives DFHBI‐1T and DFHBI‐2T (Fig. 1) that contain a trifluoromethyl moiety. In conjunction with the Spinach or Broccoli aptamers, these modified conditional fluorophores yielded 35 and 53 nm bathochromic shift in excitation and 4 and 22 nm bathochromic shift in emission wavelengths, respectively, which are better suited to existing filters designed for GFP fluorescence.66 When the cocrystal structures of Spinach were determined, it was hypothesized that the poor photostability of Spinach–DFHBI may be a consequence of the relatively sparse hydrogen bonding network between the RNA and the fluorophore and that if the fluorophore was to include additional hydrogen bonding “handles,” photostability may be enhanced. To this end, 3,5‐difluoro‐4‐hydroxybenzylidene imidazolinone‐2‐oxime (DFHO) was synthesized, which augments DFHBI by addition of an oxime moiety (Fig. 1). In vitro selection for aptamers that selectively bound DFHO yielded Corn, a ~30 nt aptamer that is red‐shifted relative to Spinach and Broccoli, exciting at 505 nm and emitting at 545 nm, and which is ~1000 times more photostable than Spinach.59 A parallel experiment in which variants of Spinach that recognize DFHO were selected yielded Orange Broccoli and Red Broccoli, which while being red‐shifted relative to Broccoli–DFHBI, exhibit rapid photobleaching, indicating that the photostability of Corn–DFHO is a property of the RNA aptamer, rather than the fluorophore.

Crystallographic structure determination revealed that, unexpectedly, Corn is a homodimer that binds one molecule of DFHO at the interprotomer interface [Fig. 2(c)].15 Subsequent biophysical analyses confirmed that the aptamer is a stable homodimer and that the fluorescent species has a 2:1 RNA:DFHO stoichiometry. Each Corn protomer folds into a mixed quadruplex, composed of two G‐quartets and two mixed sequence quartets. A G‐quartet from each protomer flanks the bound DFHO, restraining it into a planar conformation [Fig. 3(c)]. Notably, the homodimer interface lacks any interprotomer base pairing and is comprised only of three unpaired adenines from each RNA and the bound DFHO. One of these adenines makes a bidentate contact with the oxime moiety that characterizes DFHO, while another makes bifurcated hydrogen bonds bridging both heterocycles of DFHO. Because in the Corn‐bound DFHO its two heterocycles are coplanar, the fluorophore has a plane of mirror symmetry that bisects the (Corn)₂–DFHO complex. However, as Corn is a chiral molecule, its dimer cannot have a plane of mirror symmetry. This paradox is resolved by quasisymmetry,67 whereby the two RNA protomers are locally asymmetric at the DFHO binding interface, even though their overall structure is nearly indistinguishable.

In an independent study, analogs of HBI featuring extended aromatic and cationic substituents on the imidazolinone heterocycle were synthesized. When complexed with the compact aptamer “13‐2 min” that was originally isolated by Jaffrey and coworkers in the same selection experiment that yielded Spinach, three of these new fluorophores (DMHBI‐Imi, DMHBI⁺, and DMHBO⁺; Fig. 1) fluoresced green, yellow, and red, leading the authors to term the complexes Chili.58 In addition to providing access to a broad spectral range, Chili also has the largest Stokes shift reported for fluorogenic aptamers, ~130 nm for each of the DMHBI analogs. This is reminiscent of FPs such as LSSmOrange and LSSmKate,68, 69 which were engineered for their sizeable Stokes shift. While no structure has been reported for Chili, initial NMR characterization of the 13‐2 min aptamer in complex with DFHBI indicated that this aptamer also contains a G‐quadruplex, that its G‐quadruplex is different from that of Spinach, and that binding of fluorophore protects the G‐quadruplex from proton exchange, suggesting that the fluorophore is in direct contact with it.14

High Fluorophore Affinity of Compact Mango Aptamers

In addition to spectral range, fluorescence turn‐on, quantum yield, and photostability, an important property of turn‐on aptamer complexes of conditional fluorophores is the strength of the RNA–small molecule interaction. A high affinity allows more dilute fluorophore to be employed, and also makes it possible to investigate low copy‐number RNAs in live cells. Small size and high affinity were explicit considerations in selecting RNAs that would bind and activate the thiazole TO derivative TO1–Biotin. TO has a large extinction coefficient, but in its unmodified form, it binds nonspecifically to cellular nucleic acids yielding high background.70 Addition of a substituent to the benzothiazole of TO suppresses nonspecific binding, and incorporation of biotin allowed immobilization on streptavidin beads for in vitro selection. Initial selection experiments yielded the ~30 nt Mango‐I aptamer (also “RNA Mango”) which bound its cognate fluorophore with ~3 nM K _d, and induced orange fluorescence with a quantum yield of 0.14.60 Mango‐I also binds the variant fluorophore TO3–Biotin, which has extended conjugation, albeit with decreased affinity. The Mango‐I complex of TO3‐Biotin is strongly red‐shifted (Table 1).

The crystal structure of Mango‐I in complex with TO1–Biotin revealed a mostly parallel three‐tiered G‐quadruplex flexibly linked to an A‐form duplex [Fig. 2(d)].16 The fluorophore binds to one of the G‐quadruplex faces and is restrained by three flap‐like unpaired nucleotides [Fig. 3(d)]. Unexpectedly, the structure revealed that all elements of TO1–Biotin, including the fluorescent TO headgroup, the biotin, and the intervening PEG linker, are closely packed against the G‐quadruplex, adopting a circular arrangement where the biotin packs against the TO. A consequence of this arrangement is that the two heterocycles of TO (a benzothiazole and a methylquinoline) are bound by the aptamer making a 45° angle. As with the p‐HBI family of fluorophores, TO is maximally fluorescent when its two heterocycles and the connecting methine carbon are coplanar. Thus, the modest quantum yield of Mango‐I may be a result of this mode of fluorophore binding.

To improve the functional properties of Mango‐I, the selection pool from which this aptamer was isolated was subjected to functional (i.e., fluorescence‐based) microfluidic reselection.61 Three close variants of the parental aptamer were isolated: Mango‐II, Mango‐III, and Mango‐IV. These all had improved brightness, and it was estimated that four individual aptamer molecules were sufficient for visualization in cells. Structure determination of the Mango‐II aptamer in complex with TO1–Biotin and TO3–Biotin revealed a molecule with a simple overall architecture similar to that of Mango‐I (a three‐tiered G‐quadruplex and an A‐form duplex), but with a more open ligand binding pocket in which the TO headgroup of the fluorophore lies in a near‐planar conformation, consistent with an improved quantum yield of 0.21 (for TO1–Biotin). The open pocket also accommodates the TO3 headgroup, explaining the similar affinity of both fluorophores, 1.1 and 1.3 nM). Biotin is not present in the ligand binding pocket, and while most of the PEG linker appeared to be excluded from the pocket, it was shown that its interactions with the RNA contribute to affinity and selectivity.71

Thermal melt analysis of the four related Mango aptamers revealed that only Mango‐III exhibited a cooperative, biphasic melting profile, suggestive of tertiary structure. The cocrystal structure of the ~30 nt Mango‐III aptamer in complex with TO1–Biotin revealed a remarkably complex structure for an RNA of this size, distinctly different from those of Mango‐I and Mango‐II.72 Mango‐III is also organized around a G‐quadruplex (albeit of only two tiers), but incorporates two duplexes: one A‐form analogous to those of Mango‐I and Mango‐II and a second, noncanonical duplex formed between a propeller loop (Box 1) of the G‐quadruplex and nucleotides outside the G‐quadruplex. This connectivity, unprecedented for a G‐quadruplex, is reminiscent of how pseudoknots arise from the connectivity of two duplexes.73 The A‐form duplex of Mango‐III, rather than being flexibly linked to the G‐quadruplex as in Mango‐I and Mango‐II, stacks coaxially on a base triple that in turn stacks on the G‐quadruplex. Rather than unpaired flaps, the ligand binding site of Mango‐III is closed by a long‐range Watson–Crick base pair (an unusual trans pair), and the TO headgroup of the fluorophore is constrained into a planar conformation between this tertiary pair and the G‐quadruplex. Consistent with the near‐optimal conformation of the fluorophore, the quantum yield of the Mango‐III–TO1–Biotin complex is 0.55. Mutation of the trans‐Watson–Crick pair into a homopyrimidine pair further improved the quantum yield to 0.65. This mutant Mango‐III in complex with TO1–Biotin is 50% brighter than EGFP.

Fluorogenic Aptamers without G‐Quadruplexes

Just as the 11‐stranded β‐barrel is not the only way to achieve protein fluorescence, some fluorescence turn‐on RNAs are not organized around G‐quadruplexes. A promiscuous aptamer, DIR2s, which turns on both DIR and oxazole thiazole blue (OTB) allows imaging at both the blue and red edges of the visible spectrum.62 Due to the high extinction coefficients of both conditional fluorophores and the sizable fluorescence turn‐on, these are the brightest fluorogenic RNA aptamer–fluorophore complexes reported. Crystallographic structure determination demonstrated that the fluorophore binding site of DIR2s is not organized around a G‐quadruplex [Fig. 2(e)].17 Instead, the OTB fluorophore is bound between a two‐tiered stack of base triples and one unpaired adenosine [Fig. 3(e)]. Despite the potential for hydrogen bonding of OTB, its primary interaction with DIR2s is through relatively limited stacking, with hydrogen bonding limited to between the sulfonate group of OTB and a guanine in one of the base triples. The lack of extensive stacking or hydrogen bonding between the aptamer and OTB may explain the promiscuity of this turn‐on aptamer for chemically similar ligands, including DIR and even TO1–Biotin. However, as structures of DIR2s in complex with other ligands have not yet been determined, the possibility remains that other fluorophores bind to a different sites, and interfere allosterically with binding to a second fluorophore. Riboglow is the result not of in vitro selection from random RNA library, but of engineering of a bacterial riboswitch that evolved to bind cobalamin, primarily through shape complementarity in an interhelical junction.74, 75 Riboglow capitalizes on the fluorescence quenching properties of cobalamin. A series of fluorophores linked to cobalamin were synthesized that are poorly fluorescent in free form. Upon binding by the cobalamin riboswitch, the quencher is sequestered, thereby resulting in overall fluorescence turn‐on.

Applications of Fluorescence Turn‐On Aptamers

Fusions of FPs to RNA‐binding proteins, such as GFP‐MS2, have long been employed to visualize RNAs (tagged with the cognate binding site for the RNA‐binding protein) in vivo.76 Although successful, this approach suffers from low contrast as the fusion protein is intrinsically fluorescent, and thus any RNA‐unbound molecules produce background. To overcome this, the binding site is multimerized, which in turn results in the fluorescent tag often becoming as large or larger than the RNA whose cellular behavior is ostensibly being studied. In addition, these multimeric tags are recombinogenic and have also been shown to adversely affect post‐transcriptional processing of the tagged RNAs.77 Fluorescence turn‐on aptamers overcome these limitations by virtue of their small size, and because their fluorophores are conditional, offer in principle very low background (limited by nonspecific turn‐on through binding to other cellular components). Fluorogenic RNA aptamers have been used to image both coding and noncoding RNA transcripts in vivo and in vitro.78, 79, 80

Beyond their use in directly imaging RNA transcripts, fluorogenic RNA aptamers have also been developed to work as biosensors for small molecule metabolites, Förster resonance energy transfer (FRET) sensors, and split fluorescence detectors for RNA–RNA interactions. Reported small molecule biosensors are designed such that the fluorescent RNA aptamer remains unstructured until a fused metabolite‐sensing unit binds to its cognate ligand. Binding of the metabolite indirectly stabilizes the fluorophore binding site, triggering fluorescence. This approach has enabled monitoring in vivo of adenosine diphosphate (ADP), cyclic adenosine monophosphate‐guanosine monophosphate (cAG), cyclic dimeric guanosine monophosphate (ci‐di‐GMP), guanine, guanosine triphosphate (GTP), S‐adenosyl methionine (SAM), and thiamine pyrophosphate.81, 82, 83, 84

Many protein biosensors rely on FRET for readout. Recently, FRET between fluorescence turn‐on aptamers has been demonstrated, opening the way to this approach. A new analog of TO3–Biotin (oxazole yellow derivative [YO3]–Biotin) was used to develop a FRET system in which Spinach–DFHBI and Mango–YO3–Biotin to report on conformational changes to an RNA scaffold within the cell.85 The fluorescence of Spinach is also quenched when in close proximity to a protein, allowing the RNA–protein pair to be used to study RNA–protein interactions.86 As the library of available fluorogenic RNA aptamers expands, it is conceivable that new RNA FRET pairs will be developed, providing a more extensive toolbox for studying intermolecular RNA interactions. Beyond FRET, split fluorogenic RNA aptamers have also been developed as useful tools for studying RNA–RNA or RNA–DNA interactions. Split Spinach and Broccoli Florets have both been developed to monitor RNA assembly in vitro and in vivo, respectively.87, 88 Furthermore, Split Spinach has been developed to act as a sequence specific nucleic acid probe,89 to report on ribozyme self‐cleavage activity,90 and to study RNA–RNA assembly in vitro.91

Protein‐ and RNA‐Induced Fluorescence

At a global level, a striking difference between FPs and fluorescence turn‐on aptamers is that while the former mostly are variants of one fold (the 11‐stranded β‐barrel), the latter adopt diverse, apparently unrelated three‐dimensional architectures. This dichotomy could indicate that there are fewer ways for a protein scaffold to induce fluorescence of small molecules than for an RNA structure. Alternatively, this could reflect evolutionary constraints, either because the adaptive advantage of evolving FPs is small or because the evolution of FPs had to optimize both fluorophore maturation and fluorescence turn‐on. The latter hypothesis is supported by recent experiments92 in which a synthetic β‐barrel protein was designed to activate the fluorescence of DFHBI. Despite the use of state‐of‐the‐art methodology, the best that could be achieved with the protein was fluorescence that is ~6% of the fluorescence of Spinach (and ~2% of that of Mango‐III). As several FAPs can achieve fluorescence brightness exceeding that of EGFP (and Mango aptamer–fluorophore complexes), this does not indicate that proteins need to be bound covalently to their fluorophores to achieve efficient turn‐on. Instead, these results support the idea that the β‐barrel fold is a compromise between efficient autocatalytic formation of intrinsic fluorophores and fluorescence turn‐on.

Regardless of the fundamental reasons for the diversity of folds of fluorescence turn‐on RNA aptamers, the existence of diverse structural solutions opens the way to molecular engineering efforts aimed at optimizing a particular feature of interest to the researcher. This is most obviously the case in the choice of fluorophore. For use in vivo, efficient cellular uptake, lack of interaction with other cellular components (to provide low background fluorescence and molecular specificity), and low cytotoxicity (especially under illumination) are primary considerations. Some of the FP‐derived fluorophores such as DFHBI and DFHO meet all these criteria, although their extinction coefficients are only modest. Other fluorophores, such as those employed by Mango60 and Riboglow74 have higher extinction coefficients, but carry functional groups (biotin and cobalamin, respectively) that could interact in unexpected ways with cellular components. Of note, however, is that such bifunctional probes can allow two experimental modalities to be employed simultaneously. Thus, Mango–TO1–Biotin complexes can serve simultaneously for localization through fluorescence microscopy and pull‐down by streptavidin.60, 93

The specific functional groups that surround the fluorophores of FPs also must reflect the competing needs for efficient autocatalytic maturation and fluorescence turn‐on. In the case of fluorescence turn‐on RNA aptamers, the molecular environment of the fluorophore can be tuned to achieve the specific objectives of the researcher. Thus, while G‐quadruplexes are widespread among the aptamers, DIR2s and the MGA show that there are other ways of assembling binding sites that induce fluorescence. Indeed, Spinach flanks its fluorophore with a G‐quartet and a base triple on either side, and mutagenesis experiments demonstrate that the fluorophore and the base triple are closely coupled electronically, as changing the nature of the triple results in large absorption and emission wavelength changes.14, 63 G‐quadruplexes have been reported to quench fluorescence94, 95 but also have intrinsic fluorescence96, 97 so that in principle, they could function as antennas for intramolecular FRET. G‐quadruplexes are stabilized by binding a cation in their axial pore (Box 1), and experiments with Spinach revealed that changing the identity of the cation from the physiological K⁺ to Ba²⁺ and Sr²⁺ resulted in spectral shifts.14 How other RNA functional groups modulate fluorescence is just starting to be studied,98 but indications are that both fluorophore‐proximal and ‐distal groups can be important. Thus, in the case of Mango‐III, mutation of the lone Watson–Crick base pair that constrains the fluorophore against the core G‐quadruplex from A • U to U • U resulted in a ~18% increase in quantum yield. Functional reselection of Mango‐III produced iMango‐III, which achieves a comparable quantum yield but is blue‐shifted by 4–7 nm. Mutational analysis indicates that this hypsochromic shift is not a result of changes in the immediate environment of the fluorophore. Rather, it is a function of nucleotides that are at least 10 Å away.72 This is reminiscent of the effect of mutations away from the active sites of globular proteins on catalytic activity. The complex interplay of RNA structure and fluorescence properties of RNA aptamers represents an experimental challenge but also provides multiple avenues for fine‐tuning the properties of these powerful molecular tools.

The current body of work on developing and improving fluorogenic RNA aptamers highlights the often surprising structural and biophysical properties of these fluorescent tags. The discovery of a G‐quadruplex core in Spinach and subsequent structural characterizations of Corn and Mango aptamers suggested the ubiquity of G‐quadruplex motifs for binding and turning‐on small molecule fluorophores. Then, the structure determination of DIRs2 revealed an absence of a G‐quadruplex, perhaps being the exception that proves the rule. Regardless of the ultimate ubiquity of G‐quadruplexes in turn‐on aptamers, the diversity of RNA folds available to achieve binding and inducing the fluorescence of small molecule fluorophores showcases the structural versatility of RNA. Consistent with this view, aptamers selected for new conditional fluorophores have invariably had unanticipated structures. Structure‐guided design efforts have met with moderate success, yielding more conservative changes in the biophysical and biochemical properties of the aptamers. Upon the structural characterization of Spinach, the aptamer was successfully miniaturized (from 98 to 51 nt) by removing nonessential regions of the RNA molecule while retaining full fluorescence turn‐on.14 Following the structural characterization of Mango‐III, structure‐guided randomization and functional (i.e., fluorescence‐based) reselection of nucleotides surrounding the ligand led to a new variant (iMango‐III) with 13% fluorescence enhancement and unexpected hypsochromic shift and shorter fluorescence lifetime.72 The apparent disparity between the success of rational engineering of FPs, and the unexpected results of turn‐on RNA aptamer development probably reflect the latter efforts being in an earlier phase of development, and also the fact that in vitro selection has provided access to a much wider sequence space for RNAs than that explored by the canonical FPs.

This work was supported by the NIH‐Oxford‐Cambridge Research Scholars Program (L.T.); Intramural Program of the National Heart, Lung and Blood Institute, NIH.