Report

RNA Mimics of Green Fluorescent Protein

See allHide authors and affiliations

Science  29 Jul 2011:
Vol. 333, Issue 6042, pp. 642-646
DOI: 10.1126/science.1207339

Abstract

Green fluorescent protein (GFP) and its derivatives have transformed the use and analysis of proteins for diverse applications. Like proteins, RNA has complex roles in cellular function and is increasingly used for various applications, but a comparable approach for fluorescently tagging RNA is lacking. Here, we describe the generation of RNA aptamers that bind fluorophores resembling the fluorophore in GFP. These RNA-fluorophore complexes create a palette that spans the visible spectrum. An RNA-fluorophore complex, termed Spinach, resembles enhanced GFP and emits a green fluorescence comparable in brightness with fluorescent proteins. Spinach is markedly resistant to photobleaching, and Spinach fusion RNAs can be imaged in living cells. These RNA mimics of GFP provide an approach for genetic encoding of fluorescent RNAs.

The fluorophore in green fluorescent protein (GFP) is formed from three residues in the nascent protein, Ser65-Tyr66-Gly67, that undergo an autocatalytic intramolecular cyclization. The resulting fluorophore, 4-hydroxybenzlidene imidazolinone (HBI) (Fig. 1A), is encased within the protein, enabling its fluorescence (1). Chemically synthesized HBI is nonfluorescent (2), as is denatured GFP (3). However, upon refolding, the fluorescence of GFP is recovered (3). The folded GFP protein forms specific contacts with the fluorophore that prevent intramolecular motions, making fluorescence the major pathway available to dissipate the energy of the excited state fluorophore (4).

Fig. 1

RNA aptamers switch on the fluorescence of GFP-like fluorophores. (A) Structures of HBI (green), in the context of GFP, and DMHBI. (B) 13-2 enhances the fluorescence of DMHBI. Solutions containing DMHBI, 13-2 RNA, DMHBI with 13-2 RNA, or DMHBI with total HeLa cell RNA were photographed under illumination with 365 nm of light. The image is a montage obtained under identical image-acquisition conditions.

The ability to confer GFP-like functionality to RNA would facilitate studies of RNA biology and advance RNA-based applications. An RNA sequence with GFP-like properties should exhibit fluorescence upon binding a small-molecule fluorophore. However, in order to have fluorescence only associated with the RNA, the small molecule would need to be in a nonfluorescent form when not bound and switch to a fluorescent form only when bound. Antibodies and aptamers capable of eliciting the fluorescence of conditionally fluorescent dyes have been described (57). However, most conditional fluorophores can also be activated nonspecifically (4) or have other undesirable properties such as cytotoxicity (8) [supporting online material (SOM) text]. We therefore sought to identify a small molecule whose fluorescence could be activated by a specific RNA sequence yet not activated by other cellular constituents.

Because fluorescence enhancement of HBI requires suppression of subtle movements of the fluorophore (4), we reasoned that the fluorescence of HBI would not be induced by cellular constituents. To test this, we prepared several HBI derivatives (fig. S1A) and found that these compounds were not detectably fluorescent upon incubation with cells (fig. S1B) or cellular RNA or DNA (fig. S1C). Incubation of GFP-like fluorophores with cells illuminated for 10 min at 37°C resulted in negligible cell death, whereas another conditional fluorophore, malachite green, exhibited considerable cytotoxicity (fig. S1D).

We next sought to identify RNA sequences that bind and activate the fluorescence of GFP fluorophores, beginning with 3,5-dimethoxy-4-hydroxybenzylidene imidazolinone (DMHBI) (Fig. 1A). We performed systematic evolution of ligands by exponential enrichment (SELEX) (9, 10) with a library containing ~5 × 1013 RNA molecules and selected RNAs for their ability to bind DMHBI-agarose. After five rounds of selection, the pool of RNAs weakly activated DMHBI fluorescence, with further increases in fluorescence up to round 10 (fig. S2A).

To identify individual RNA aptamers that accounted for DMHBI fluorescence, we screened individual sequences and identified one RNA, 13-2 (fig. S2, B and C), which exhibited the highest degree of aptamer-induced fluorescence (Fig. 1B). The spectra of the 13-2-DMHBI complex contained a single emission peak at 529 nm and single excitation peak at 398 nm (fig. S2B). The brightness of 13-2-DMHBI was 12% relative to GFP (table S1), and the dissociation constant (Kd) for the 13-2-DMHBI complex is 464 nM (fig. S2D). Truncation and mutagenesis experiments supported the secondary structure predicted by Mfold (fig. S2E) (11) and resulted in a 60-nucleotide (nt) minimal domain with enhanced quantum yield (table S1 and fig. S2E).

To determine whether DMHBI could be spectrally tuned with RNA to exhibit a range of fluorescence properties, we performed further SELEX screens (12). Several aptamers were identified that exhibited markedly different spectral properties, including light blue (2-4), greenish-yellow (3-6), and yellow (17-3) fluorescence (Fig. 2, A to D, and table S1). Sequence alignment of these aptamers reveals little sequence similarity, and secondary structure analysis predicts a range of structures (fig. S3).

Fig. 2

Spectral tuning and fluorophore diversity produce a palette of RNA-fluorophore complexes. (A) Absorbance spectra of GFP-like fluorophores. Spectra were collected in the absence of RNA at pH 7.4. (B and C) Excitation (B) and emission (C) spectra of RNA-fluorophore complexes. Spectra were collected in the presence of excess fluorophore at pH 7.4 for RNAs binding to DMHBI (2-4, 13-2, 3-6, and 17-3), DFHBI (24-2), DMABI (11-3), and 2-HBI (6-8). Spectra are normalized to the excitation and emission peak for each complex. Arrows indicate the fluorophore from which each spectrum is derived. For emission spectra (C), DMHBI is indicated with blue shading. (D) RNA-fluorophore complexes were illuminated with ultraviolet light (365 nm) and photographed. From left to right, the tubes contain RNAs 2-4, 24-2, 11-3, 13-2, 3-6, 17-3, 6-8, and fluorophores as indicated above. The image is a montage obtained under identical image-acquisition conditions.

To further extend spectral properties, we generated two additional GFP-like fluorophores, 4-dimethylaminobenzylidene imidazolinone (DMABI) and 2-hydroxybenzlidene imidazolinone (2-HBI) (fig. S1A). An aptamer selected against DMABI exhibited green fluorescence, whereas an aptamer selected against 2-HBI exhibited orange-red fluorescence (Fig. 2D and fig. S3). Thus, a range of RNA-fluorophore complexes spanning the visible spectrum can be generated by using GFP-like fluorophores.

A major advance in GFP technology was the discovery of enhanced GFP (EGFP) (13). The HBI fluorophore in GFP/EGFP can exist in either the phenol (protonated) or phenolate (deprotonated) form. In GFP, the phenol form predominates at neutral pH because of the relatively high pKa (where Ka is the acid dissociation constant) of HBI (14), whereas in EGFP, HBI is almost exclusively in the phenolate form (13). The phenolate species exhibits a higher extinction coefficient, which contributes to the increased brightness of EGFP (13).

Characterization of the DMHBI-binding aptamers suggested that these complexes mimic GFP rather than EGFP. Like HBI, DMHBI at pH 7.4 is primarily in the phenol form, with a small portion in the phenolate form (Fig. 2A), which is consistent with its pKa of 8.0 (fig. S4A). When bound to 17-3, the excitation spectrum reveals both forms, indicating that this aptamer binds DMHBI irrespective of its protonation state (Fig. 2B). Thus, the 17-3–DMHBI complex most closely resembles GFP. In contrast, when DMHBI is bound to either 13-2, 2-4, or 3-6 the excitation spectra reveals that RNA binds exclusively to the phenolic form of the fluorophore (Fig. 2B).

To generate RNA-fluorophore complexes that exhibit the spectral properties of EGFP, we used a biomimetic strategy to obtain RNAs that bind the phenolate form of a GFP-like fluorophore. We designed a new HBI derivative, 3,5-difluoro-4-hydroxybenzylidene imidazolinone (DFHBI), which is exclusively in the phenolate form because of the introduction of fluorine residues that reduce the pKa (Fig. 2A). An aptamer selected against DFHBI (24-2) exhibited an excitation spectra that is consistent with exclusive binding to the phenolate form of DFHBI (Fig. 3A) and a markedly enhanced quantum yield of 0.72, which is 20% higher than EGFP (table S1). The molar brightness of 24-2–DFHBI is 53% of EGFP but brighter than many other fluorescent proteins (table S1) (15).

Fig. 3

RNA-fluorophore complexes with EGFP-like properties. (A) Normalized excitation (blue) and emission (green) spectra of 24-2–DFHBI complex. (B) Normalized absorbance spectra of DFHBI at pHs 5.0, 6.0, 7.0, and 8.0. (C and D) 24-2 incubated with excess DFHBI (C) or HBI (D) at pHs 6.0, 7.0, and 8.0. (E) Photobleaching curves for 24-2–DFHBI, EGFP, and fluorescein. Fluorophores were immobilized on glass slides and illuminated continuously with a 130 W mercury lamp. Total fluorescence was then plotted against exposure time and normalized to the maximum intensity of each fluorophore.

To determine whether 24-2 selectively recognizes the phenolate form of DFHBI, we measured its spectral properties at different pHs. When the pH is lowered to pH 6.0, both the phenolic and phenolate forms of DFHBI are detected (Fig. 3B). However, at pH 6.0, the 24-2–DFHBI complex exhibits only the phenolate excitation peak, indicating preferential binding of the RNA to the phenolate form of the fluorophore (Fig. 3C). We also examined the binding of 24-2 to HBI, which resembles DFHBI but lacks fluorines. 24-2 weakly binds and activates the fluorescence of HBI at pH 8.0, at which both the phenol and phenolate forms are present (16). However, at pH 7.0 and 6.0, at which only the phenol form is present, 24-2 fails to activate HBI fluorescence. Taken together, these data suggest that 24-2 selectively recognizes the phenolate form of the fluorophore.

We next examined the photobleaching properties of 24-2–DFHBI complexes. 24-2 was immobilized to a glass surface, and total fluorescence was measured with DFHBI in solution over 45 min of continuous illumination. Compared with fluorescein and EGFP, 24-2–DFHBI exhibited negligible photobleaching, suggesting that exchange of bound DFHBI with DFHBI in solution prevents the accumulation of photobleached complexes.

Because of the green fluorescence and useful spectral properties of 24-2–DFHBI complexes, we termed the 24-2 sequence “Spinach.” To examine Spinach fluorescence in cells, we transformed Escherichia coli with plasmids expressing Spinach fused to a short RNA-stabilizing element (17). After a brief incubation with DFHBI, we found that fluorescence was readily detectable in individual cells (fig. S5A) as well as in colonies on a plate (fig. S5B).

To determine whether Spinach could be used to tag RNAs in living mammalian cells, we fused Spinach to the 3′ end of 5S (fig. S6), a small noncoding RNA transcribed by RNA polymerase III (Pol III) that associates with the large ribosomal subunit, and transfected this construct into human embryonic kidney (HEK) 293 T cells. The 3′ end of 5S is solvent-exposed, and addition of short sequences to the 3′ end does not affect 5S localization (18). 5S-Spinach fluorescence was detected throughout cells (Fig. 4A), with a distribution similar to that of endogenous 5S in the same cell type (18). After application of 600 mM sucrose, a form of cellular stress that induces the formation of cytoplasmic RNA granules (19), 5S-Spinach relocalized to large (~2 to 3 μm) cytosolic foci, many of which colocalize with T-cell intracellular antigen-1–related protein (TIAR), a marker of stress granules (fig. S7) (20).

Fig. 4

Live-cell imaging of Spinach fusion RNAs. (A) Live-cell imaging of Spinach-tagged 5S RNA. Fluorescence and phase images of HEK293 T cells expressing 5S tagged with either Spinach or Lambda, a control RNA. Fluorescence is detected in 5S-Spinach–expressing cells in the presence of 20 μM DFHBI, with granule formation present in cells treated with 600 mM sucrose for 30 min (↑Suc). White dashed lines indicate nuclear borders assessed by means of Hoescht 33342 staining. (B) 5S-Spinach RNA induction in response to stress. 5S-Spinach–expressing HEK293 T cells were pretreated with 30 nM ML-60128 for 16 hours and then treated with vehicle or 600 mM sucrose for 60 min. Treatment of cells with sucrose resulted in a rapid induction of 5S-Spinach RNA and an increase in total 5S-Spinach levels compared with control cells. (C) 5S-Spinach RNA localization into granules. 5S-Spinach–expressing HEK293 T cells were stimulated with 600 mM sucrose in order to monitor the rate of formation of 5S-Spinach–containing granules. Arrowheads indicate granules that formed earliest, and arrows indicate granules that developed later during the time course of treatment. Scale bar, 10 μm.

To monitor live-cell 5S dynamics, cells were treated with the Pol III inhibitor ML-60218 (21), which reduces 5S-Spinach fluorescence to baseline levels (movie S1). We first monitored nuclear export of 5S-Spinach. After washout of ML-60218, cells were incubated with leptomycin B, an inhibitor of nuclear export of 5S (22). Under these conditions, 5S-Spinach accumulates in the nucleus. Upon removal of leptomycin-B, 5S-Spinach rapidly appeared in the cytosol (movie S1), indicating highly efficient nucleocytoplasmic trafficking of 5S. We next monitored the induction of 5S-Spinach in response to sucrose. After washout of ML-60218, treatment of cells with sucrose resulted in rapid induction of 5S-Spinach over 60 min and a higher total level of 5S-Spinach than that in untreated cells (Fig. 4B and movie S2). Unlike control cells, sucrose treatment caused 5S-Spinach to accumulate to higher levels in the nucleus than in the cytosol (fig. S8), possibly reflecting saturation of the nuclear export machinery. Additionally, 5S-Spinach accumulated in cytoplasmic granular structures 30 min after sucrose treatment, which is consistent with stress granule formation (movie S2). We next examined the time course of relocalization of 5S-Spinach into granules in cells not treated with ML-60218. Before experimental treatment, 5S-Spinach exhibited diffuse nuclear and cytoplasmic localization (Fig. 4C). After sucrose treatment, 5S-Spinach clustered into granules in as little as 9 min (movie S3), with new granules continuing to form up to 30 min later. Together, these data indicate the ability of Spinach to reveal the intracellular dynamics of RNA in living cells.

We have described a palette of RNA-fluorophore complexes spanning much of the visible spectrum. RNAs can be tagged with Spinach, providing a simple strategy for introducing a compact fluorescent tag for live-cell imaging of RNAs while avoiding the problems associated with current methods for tagging RNAs (23). Spinach is different from GFP in that it exhibits considerable resistance to photobleaching, and fluorescence is observed shortly after Spinach transcription in cells, which contrasts with the delay in acquisition of fluorescence by nascent GFP because of the requirement for fluorophore maturation (24). The results described here raise the possibility of using genetically encoded RNA-fluorophore complexes for other applications, including RNA-RNA and RNA-protein fluorescence resonance energy transfer, and simultaneous imaging of multiple RNAs.

Supporting Online Material

www.sciencemag.org/cgi/content/full/333/6042/642/DC1

Materials and Methods

SOM Text

Figs. S1 to S8

Table S1

References (25–40)

Movies S1 to S3

References and Notes

  1. Materials and methods are available as supporting material on Science Online.
  2. Acknowledgments: We thank M. S. Cohen, A. Deglincerti, J. D. Warren, and S. C. Blanchard (Weill Cornell Medical College) and D. S. Tan (Sloan-Kettering Institute) for useful comments and suggestions; D. Engelke (University of Michigan) for providing plasmids containing the 5S sequence; and F. Dardel (Université Paris Descartes) for providing plasmids containing the tRNA scaffold sequence. This work was supported by the McKnight Neuroscience Technology Innovation Award, Weill Cornell Medical College, and National Institute of Neurological Disorders and Stroke grant NS064516 (S.R.J.) and training grant T32CA062948 (J.S.P.).
View Abstract

Navigate This Article