Specific Covalent Labeling of Recombinant Protein Molecules Inside Live Cells

See allHide authors and affiliations

Science  10 Jul 1998:
Vol. 281, Issue 5374, pp. 269-272
DOI: 10.1126/science.281.5374.269


Recombinant proteins containing four cysteines at the i,i + 1, i + 4, and i + 5 positions of an α helix were fluorescently labeled in living cells by extracellular administration of 4′,5′-bis(1,3,2-dithioarsolan-2-yl)fluorescein. This designed small ligand is membrane-permeant and nonfluorescent until it binds with high affinity and specificity to the tetracysteine domain. Such in situ labeling adds much less mass than does green fluorescent protein and offers greater versatility in attachment sites as well as potential spectroscopic and chemical properties. This system provides a recipe for slightly modifying a target protein so that it can be singled out from the many other proteins inside live cells and fluorescently stained by small nonfluorescent dye molecules added from outside the cells.

Attachment of fluorescent or other useful labels onto proteins has traditionally been accomplished by in vitro chemical modification after purification (1). Green fluorescent protein (GFP) from the jellyfish Aequorea victoria can be genetically fused with many host proteins to produce fluorescent chimeras in situ (2, 3). However, GFP is potentially perturbative because of its size (238 amino acids), can usually only be fused at the NH2- or COOH-terminus of the host protein, offers a limited variety of colors, and is of no assistance for spectroscopic readouts other than fluorescence. We therefore designed and synthesized a tight-binding pair of molecular components: a small receptor domain composed of as few as six natural amino acids that could be genetically incorporated into proteins of interest, and a small (<700-dalton), synthetic, membrane-permeant ligand that could be linked to various spectroscopic probes or crosslinks. The ligand has relatively few binding sites in nontransfected mammalian cells but binds to the designed peptide domain with a nanomolar or lower dissociation constant. An unexpected bonus is that the ligand is nonfluorescent until it binds its target, whereupon it becomes strongly fluorescent.

Our approach exploits the facile and reversible covalent bond formation between organoarsenicals and pairs of thiols. Trivalent arsenic compounds bind to the paired thiol groups of proteins containing closely spaced pairs of cysteines or the cofactor lipoic acid (4, 5). Such binding, which is responsible for much of the toxicity of arsenic compounds, is completely reversed by small vicinal dithiols such as 2,3-dimercaptopropanol [British anti-Lewisite (BAL)] or 1,2-ethanedithiol (EDT), which form tighter complexes with the organoarsenical than do cellular dithiols (6, 7). If a peptide domain could be designed with even higher affinity than that of the antidotes for an organoarsenical ligand, the ligand could be administered in the presence of excess antidote and specifically bind the desired peptide domain without poisoning other proteins. To achieve this unusual affinity, we designed a peptide domain with four cysteines already organized to bind an organic molecule containing two appropriately spaced trivalent arsenics (Fig. 1). If the distance between the two pairs of cysteines matched the spacing between the arsenics, the two dithiol-arsenic interactions should be highly cooperative and entropically favorable. The four cysteines were placed at the i, i + 1, i + 4, and i + 5 positions of an α helix, so that the four thiol groups would form a parallelogram on one side of the helix. We chose acetyl-WEAAAREACCRECCARA-amide (8) as a model peptide for in vitro tests, on the basis of the known propensity of peptides of the form acetyl-W(EAAAR)nA-amide (9) to form α helices.

Figure 1

Synthesis of FLASH (20) and proposed structure of its complex with an α-helical tetracysteine-containing peptide or protein domain. Although the structure is drawn with thei and i + 4 thiols bridged by one arsenic and thei + 1 and i + 5 thiols bridged by the other, we cannot rule out the isomeric complex in which one arsenic links thei and i + 1 thiols while the other links thei + 4 and i + 5 thiols.

Fourteen biarsenical ligands were synthesized and tested for their ability to bind the tetracysteine peptide in the presence of a small excess of BAL or EDT (10), but the only successful ligand was 4′,5′-bis(1,3,2-dithioarsolan-2-yl)fluorescein, which may also be called FLASH-EDT2 [fluorescein arsenical helix binder, bis-EDT adduct (Fig. 1)]. FLASH-EDT2 was prepared in a single step by transmetallation (11,12) of commercially available fluorescein mercuric acetate, followed by addition of EDT to facilitate purification. The product was virtually nonfluorescent but became brightly fluorescent after addition of the tetracysteine peptide to displace the EDT. The excitation and emission peaks were 508 and 528 nm, respectively (Fig. 2), which are about 20 nm longer in wavelength than those of free fluorescein. The quantum yield of the FLASH-peptide complex was 0.49, whereas FLASH-EDT2 was ≤5 × 10–4 times as fluorescent. The small size of EDT probably permits rotation of the aryl-arsenic bond and excited state quenching by vibrational deactivation or photoinduced electron transfer, whereas the peptide complex may evade such quenching because its more rigid conformation should hinder conjugation of the arsenic lone pair electrons with the fluorescein orbitals. The equilibrium reaction FLASH-EDT2 + peptide ⇌ FLASH-peptide + 2 EDT favored FLASH-peptide at ≤10 μM EDT and FLASH-EDT2 at ≥1 mM EDT, so that labeling was reversed by millimolar concentrations of EDT (Fig. 2B). Monothiols such as 2-mercaptoethanol or glutathione were helpful to catalyze equilibration but did not compete themselves. The FLASH-peptide complex showed no sign of dissociation even when diluted to 1 nM in 5 mM 2-mercaptoethanol and left for weeks, which indicates that complex formation was essentially irreversible in the absence of excess EDT. The apparent pK a of the fluorescein chromophore in the FLASH-peptide complex was 5.4, so the fluorescence should not be sensitive to variations in cytosolic pH near 7.

Figure 2

Fluorescence of FLASH is induced by binding to a tetracysteine motif. (A) Fluorescence excitation (Exc) and emission (Em) spectra of 250 nM FLASH bound either to a model tetracysteine-containing peptide (20) in phosphate-buffered saline at pH 7.4 (solid lines) or to EDT at the same gain settings (dashed line, emission spectrum only). (B) Kinetics of binding of 1 μM FLASH to 10 μM peptide in the presence of 10 μM EDT and subsequent reversal by a higher concentration (5 mM) of EDT. The apparent fluorescence of FLASH-EDT2 ranged from 0.05% (A) to 0.5% (B) of that of the FLASH-peptide complex and might merely reflect trace impurities such as free fluorescein.

To test the membrane permeability of FLASH-EDT2 and the specificity of the FLASH-peptide interaction in live mammalian cells, we genetically fused the designed peptide (with the tryptophan changed to an alanine) to the COOH-terminus of a cyan mutant [enhanced cyan fluorescent protein (ECFP)] of GFP (13, 14) and transiently expressed this fusion protein in HeLa cells. ECFP was chosen as the host protein so that expressing cells could be distinguished from nonexpressing cells by ECFP fluorescence (Fig. 3) and because formation of the ECFP-Cys4 peptide-FLASH complex should be demonstrable by fluorescence resonance energy transfer (FRET) from ECFP to FLASH. Before addition of FLASH-EDT2, some cells were brightly fluorescent at the ECFP emission maximum (480 nm) and only very dimly fluorescent at 635 nm, at which ECFP barely emits; whereas the other cells did not express ECFP. When 1 μM FLASH-EDT2 was added with 10 μM EDT, cells already expressing ECFP increased their fluorescence at 635 nm more than threefold, because of the long-wavelength tail of FLASH emission, whereas the fluorescence of ECFP at 480 nm declined by >70%, indicating FRET from the ECFP to the bound FLASH. The high efficiency of FRET showed that >70% of the intracellular ECFP molecules had directly bound FLASH and that the distance between the chromophores was <5 nm as expected from the protein dimensions (15). Cells not expressing ECFP were unaffected by FLASH-EDT2. After binding was complete, which required about 1 hour (Fig. 3B), removal of FLASH from the medium while maintaining 10 μM EDT had little effect, but 1 mM EDT largely reversed the effects of FLASH-EDT2 (Fig. 3, A and B). The controllable onset and reversibility of binding and labeling should prove valuable in many applications.

Figure 3

FLASH labeling of a tetracysteine motif appended to a GFP mutant in living cells. (A) Reversible intracellular labeling of a tetracysteine-containing peptide, AEAAAREACCRECCARA (8), fused to the COOH-terminus of ECFP (13, 14) and expressed in HeLa cells (21). Fluorescence images (22) were recorded before and after incubation with 1 μM FLASH-EDT2 and after treatment with 1 mM EDT. Excitation was at 440 ± 10 nm, and emissions were collected at 480 ± 15 nm (top row) and 635 ± 25 nm (bottom row). Comparison with a transmitted-light view (far left) shows that cells not expressing the fusion protein showed negligible fluorescence at either emission wavelength. Fluorescence images were taken at 4, 150, and 195 min. (B) Time course of intensities at cyan (circles) and red (squares) emission wavelengths for the fluorescent cell at the upper right, compared to the same measurements (up- and down-pointing triangles) on a nonexpressing cell. Excess FLASH and EDT were removed as indicated (washout), then the cells were incubated with the indicated concentrations of EDT.

The tetracysteine motif also worked when introduced into an endogenous α helix. The gene for Xenopus calmodulin (14) was mutated to replace four residues of the NH2-terminal α helix (Glu6, Glu7, Ala10, and Glu11) with cysteines. The cytosol and nuclei of HeLa cells expressing this mutant calmodulin became brightly fluorescent when treated with 1 μM FLASH-EDT2and 10 μM EDT (Fig. 4). Control nontransfected cells in a different dish showed dimmer staining, which appeared to be mostly mitochondrial. This background labeling could be somewhat further reduced by higher EDT concentrations, which would not interfere with FLASH labeling if the affinity of the tetracysteine peptide for FLASH could be increased by combinatorial optimization of adjacent residues.

Figure 4

FLASH labeling of a tetracysteine motif inserted within a protein. (Left) Fluorescence images (excitation 480 ± 15 nm, emission 535 ± 12.5 nm) of HeLa cells transiently transfected with a gene for Cys6,7,10,11-calmodulin, labeled 36 hours later with 1 μM FLASH-EDT2 and 10 μM EDT in Hank's balanced salt solution for 1 hour at 25°C, then washed free of excess dye just before imaging. (Middle) Nontransfected HeLa cells labeled and imaged under identical conditions. (Right) Nontransfected controls reimaged at 4.5-fold higher gain.

FLASH-EDT2 (1 μM) administered with 10 μM EDT had no detectable effect for up to 4 hours on the viability of HeLa cells and Jurkat lymphocytes assayed by propidium iodide exclusion; reduction of methylthiazolyldiphenyltetrazolium to the colored formazan (16); or ability to respond to muscarinic stimulation, assessed by activation of the nuclear factor of activated T cells and monitoring of the expression of a β-lactamase reporter (17). However, 2 μM phenylarsine oxide, which contained an equivalent amount of arsenic but lacked the EDT antidote, was quite toxic. Therefore, EDT seems necessary and sufficient to prevent FLASH-EDT2 from exerting acutely toxic effects (6, 7).

It is unlikely that every native protein that contains the core motif CCXXCC (8) will bind FLASH. The thiols must be able to reach an α-helical or other conformation able to form two pairs that grip the arsenics like pincers, yet the thiols must not be disulfide-bonded or tightly chelated to a metal. Endogenous competing proteins or ligands are clearly rare enough in mammalian cells to permit easy detection of transfected proteins over background (Figs. 3and 4). FLASH confers temporal control of labeling, which is particularly helpful in quantifying FRET because it enables donor emission to be compared in situ before and after labeling by FLASH acting as an acceptor (Fig. 3). Derivatives of FLASH that incorporate luminescent or magnetic resonance reporters, environmentally sensitive fluorophores or indicators, or photochemically reactive moieties should label proteins containing the CCXXCC motif and confer the appropriate spectroscopic properties. Immobilized FLASH analogs and tetracysteine motifs might complement systems such as biotin:avidin, glutathione:glutathione S-transferase, and nickel:polyhistidine for protein attachment and affinity purification (5, 18). FLASH dimers with two xanthene nuclei and four arsenics should induce homodimerization of peptides or proteins containing the tetracysteine motif in analogy to methods that rely on immunosuppressants binding to immunophilins (19). The new system combines high affinity and specificity, easy reversibility, easy modification of the ligand, small size of the peptide domain, physiological compatibility, and fluorescence indication that binding has occurred.

  • * Present address: Aurora Biosciences, 11010 Torreyana Road, San Diego, CA 92121, USA.


Stay Connected to Science

Navigate This Article