2.1.

Ensemble Spectroscopy

UV-VIS absorption spectra were recorded with a Lambda 6 spectrophotometer (Perkin-Elmer). Fluorescence spectra were recorded with a Spex Fluorolog 1500 fluorimeter (Spex Industries). The fluorescence quantum yield of HcRed in PBS was measured relative to that of Texas Red (Sigma Aldrich) in water with using $590\text{-}\mathrm{nm}$ excitation. Time-correlated single-photon counting (TCSPC) experiments were performed using either 470-nm or 570-nm pulsed excitation from the frequency-double output of a Titanium:Sapphire laser (Mira 900 from Coherent, pulse width $150\mathrm{fs}$ , repetition rate $76\mathrm{MHz}$ ) pumped by an $\mathrm{Ar}$ -ion laser (Spectra Physics 2080 Beamlok). For these experiments, a pulse picker (Conoptics, Inc.) was used to reduce the repetition rate of the laser to $7.6\mathrm{MHz}$ . Isotropic and anisotropic fluorescence decays were detected using a microchannel plate photomultiplier (Hamamatsu, response time of $23\mathrm{ps}$ ) coupled to a monochromator (Acton Research). Fluorescence decays were collected with a TCSPC board (SPC630, Becker & Hickl GmbH) in 4096 channels with 10,000 photon counts in the peak and analyzed by a reweighted iterative reconvolution method based on the Levenberg-Marquardt algorithm. Details on the TCSPC setup and the fluorescence lifetime analysis have been published previously.⁹

2.2.

Fluorescence Correlation Spectroscopy

For FCS experiments, we excited HcRed with the 594-nm line of a $\mathrm{He}\text{-}\mathrm{Ne}$ laser (Uniphase). The excitation power at the sample was adjusted using a polarizer combined with a half-wave plate. The laser light was steered into an inverted confocal microscope (IX-71, Olympus-Japan) by a dichroic mirror (590DRLP, Chroma Technologies) and focused into the sample by a water-immersion objective (Olympus Japan, 1.2 NA, $60\times$ ). Fluorescence was collected from the sample through the same lens, spectrally filtered by the dichroic mirror and a bandpass filter (650/50HQBP, Chroma Technologies), spatially filtered by a $75\mu \mathrm{m}$ pinhole, split by a 50/50 beamsplitter, and last imaged onto two single-photon avalanche photodiodes (SPAD, SPCM AQR 14, Perkin-Elmer). The signals from the SPADs were cross-correlated by a fast hardware correlator (ALV5000/E-FAST, ALV-Laser $\mathrm{GmbH}$ ) whose lag-time resolution is $12.5\mathrm{ps}$ . Curve fitting of the FCS data was performed using Igor Pro 5.2 (WaveMetrics, Inc.).

2.3.

Single-Molecule Spectroscopy

Samples containing individual immobilized HcRed or GFP proteins were imaged with the inverted confocal microscope described earlier, now fitted with a closed-loop piezo-electric scanning stage (NanoH100, Mad City Labs, Inc.). The same objective lens as earlier was used for optical excitation and fluorescence collection. For HcRed, we used the 594-nm line of a $\mathrm{He}\text{-}\mathrm{Ne}$ laser and a similar detection scheme as for the FCS experiments, except that the beamsplitter was removed so that the collected fluorescence was imaged into a single SPAD. For GFP, optical excitation was accomplished with the 470-nm frequency-doubled output of our Ti:Sapphire laser, with the collected fluorescence filtered in this case by a combination of a dichroic mirror (DRLP505, Chroma Technologies) and a bandpass filter (HQ535/50, Chroma Technologies). Time trajectories for both HcRed and GFP proteins were recorded with the ALV5000 hardware correlator, with a dwell time of $1\mathrm{ms}$ .

3.1.

Dimerization of HcRed

According to the crystal structure,⁵ HcRed is a dimer comprised of two 25-kDa GFP-like protomers. Each protomer contains a chromophore formed from a Glu64-Tyr65-Gly66 tri-peptide. HcRed absorbs maximally at $590\mathrm{nm}$ [Fig. 1a dashed line], with an extinction coefficient of $\mathrm{120,000}{\mathrm{M}}^{-1}{\mathrm{cm}}^{-1}$ . HcRed fluorescence emission has its peak at $645\mathrm{nm}$ [Fig. 1a, line, excitation at $590\mathrm{nm}$ ], with a quantum yield of 0.05. This fluorescence decays bi-exponentially with lifetimes of $1.47\mathrm{ns}$ (75% contribution) and $0.51\mathrm{ns}$ (25%) [Fig. 1b]. HcRed owes its far-red absorption and emission properties to an interaction⁵ between the aromatic ring of the chromophore and a nearby amino acid residue, Hisl74. Increased $\pi$ -stacking interactions with the chromophore are known to red-shift the spectral properties of GFP.¹

Fig. 1

(a) Absorption (dashed line), fluorescence emission (solid line, excitation at $590\mathrm{nm}$ , and solid line plus dots, excitation at $470\mathrm{nm}$ ), and fluorescence excitation (dashed line plus dots, detection at $530\mathrm{nm}$ ) spectra of HcRed in phosphate buffered saline (PBS). (b) Fluorescence decay of HcRed in PBS excited at $570\mathrm{nm}$ and detected at $645\mathrm{nm}$ . Shown in gray is a biexponential fit with decay time constants of 1.47 and $0.51\mathrm{ns}$ . The inset is the residuals of the fit from the main panel. (c) Fluorescence decays of HcRed in PBS excited at $470\mathrm{nm}$ and detected at $645\mathrm{nm}$ (curve 1) and $530\mathrm{nm}$ (curve 2). Shown in gray are multiexponential fits (see main text). The inset is the residuals for the fits from the main panel. (d) Decay of the fluorescence anisotropy of HcRed ( $570\text{-}\mathrm{nm}$ excitation) and biexponential fit with time constants of 0.3 and $50\mathrm{ns}$ .

We probed the oligomerization state of HcRed with two methods: fluorescence correlation spectroscopy (FCS) and single-molecule fluorescence spectroscopy (SMS). Both methods require a low protein concentration (1 to $100\mathrm{pM}$ ) at which aggregation may be excluded.

FCS measures the fluorescence fluctuations arising from molecules or molecular complexes that diffuse freely into and out of a diffraction-limited laser focus. The time dependence of these fluctuations can yield information about the translational and rotational diffusion coefficients of the molecules and, therefore, about their size.¹⁰ These coefficients are obtained from the autocorrelation of the fluorescence intensity:

Here $N$ is the average number of molecules in the detection volume, ${\tau }_D$ is the translational diffusion time, $z_0$ and ${\omega }_0$ are, respectively, the axial and radial dimensions of the probe volume, and $F_i$ and ${\tau }_i$ are the fraction contribution and lifetime of the fluorescence flickers present in addition to translational diffusion. Figure 2 shows the $AC$ of HcRed in phosphate buffered saline (PBS, pH 7.4, excitation at $594\mathrm{nm}$ , with an average power at the sample of $50\mu \mathrm{W}$ ). In addition to a translational diffusion component $({\tau }_D=285\mu \mathrm{s})$ , the $AC$ of HcRed contains flicker lifetimes of $0.05\mu \mathrm{s}$ (57% contribution), $3\mu \mathrm{s}$ (18%), and $23\mu \mathrm{s}$ (25%).¹¹ Monomeric GFP in PBS excited at $470\mathrm{nm}$ has a translational diffusion time of $140\mu \mathrm{s}$ (FCS data not shown here: see Ref. 12). The translational diffusion time, ${\tau }_D$ , and the diffusion coefficient, $D$ , of a protein relate to each other by ${\tau }_D={\omega }_0^2∕4D$ . $D$ is dependent on the hydrodynamic radius of the protein through the Stokes-Einstein equation: $D=k_B\mathrm{T}∕6\pi \eta R_h$ , where $\eta$ is the viscosity of the solvent, and $R_h$ is the hydrodynamic radius of the protein. Assuming diffraction-limited probe volumes in these FCS experiments, the ratio ${\omega }_0\left(594\mathrm{nm}\right)∕{\omega }_0\left(470\mathrm{nm}\right)$ is $\sim 1.3$ , and the ratio of the hydrodynamic radii of HcRed and monomeric GFP is $R_h\left(\mathrm{HcRed}\right)∕R_h\left(\mathrm{GFP}\right)\sim 1.3$ . If we assume that the molecular weight (MW) of the protein scales with the third power of its molecular hydrodynamic radius, then MW(HcRed)/MW(GFP) is approximately 2.0, indicating that HcRed is a dimer in solution.

Fig. 2

Autocorrelations of the fluorescence intensity of HcRed and Texas Red in PBS ( $570\text{-}\mathrm{nm}$ excitation). Also shown are fits (gray lines, main panel; see text for details) and residuals of the fits (inset panels; upper, HcRed; lower, Texas Red).

Rotational diffusion of the protein induces the fast fluorescence flicker in the $AC$ of HcRed (here ${\tau }_{rot}=0.05\mu \mathrm{s}$ ). For a comparison, Fig. 2 shows the $AC$ of Texas Red in PBS (excitation at $594\mathrm{nm}$ with an average power of $50\mu \mathrm{W}$ ), which is flat at short time delays. This organic dye has a rotational correlation time of $\sim 250\mathrm{ps}$ , which is below the minimum lag-time resolution of the FCS experiment. The rotational correlation time ${\tau }_{rot}$ of a protein and its molecular hydrodynamic volume $V_h$ relate to each other¹³ by ${\tau }_{rot}=1/(6D_{\mathit{rot}}=k_{B}T∕(36V_h)$ , where $D_{rot}$ is the rotational diffusion coefficient. Monomeric GFP has a rotational correlation time of $16\mathrm{ns}$ .¹³ The ratio of the hydrodynamic volumes for HcRed and monomeric GFP will then be $V_h\left(\mathrm{HcRed}\right)∕V_h\left(\mathrm{GFP}\right)={\tau }_{rot}\left(\mathrm{HcRed}\right)∕{\tau }_{rot}\left(\mathrm{GFP}\right)\sim 3.1$ , once again suggesting the presence of dimeric HcRed.

We show evidence for the dimerization of HcRed by fluorescence spectroscopy of immobilized single molecules. Figure 3 is a confocal fluorescence image $(5\times 5\mu \mathrm{m})$ of individual HcRed proteins immobilized in water-filled pores of polyacrylamide gel. The immobilized proteins were excited with a $2\text{-}\mu \mathrm{W}$ , $594\text{-}\mathrm{nm}$ laser light, and their fluorescence was detected for wavelengths above $610\mathrm{nm}$ . The diffraction-limited fluorescence spots in this image correspond to single proteins and are comprised of as many as 700 photon counts per spot. Figure 4a shows the fluorescence intensity trajectory recorded from an individual HcRed molecule located by confocal imaging. This particular trajectory shows two discernible intensity levels, with stepwise intensity changes prior to photobleaching [Fig. 4a]. This behavior is characteristic of multichromophoric entities when observed individually.^{14, 15, 16} For this particular molecule, the background corrected histogram of the detected photon counts [Fig. 4b] is bimodal with peaks at $49\mathrm{counts}∕\mathrm{ms}$ and $29\mathrm{counts}∕\mathrm{ms}$ . As a comparison, Fig. 4c shows a fluorescence intensity trajectory recorded from a single monomeric GFP immobilized in polyacrylamide gel, under a $1\text{-}\mu \mathrm{W}$ excitation at $470\mathrm{nm}$ . The GFP trajectory shows one intensity level above the dark state, with a background-corrected histogram of the counts [Fig. 4d] displaying a single-mode distribution. The GFP trajectory is interrupted by dark periods, a phenomenon known as “blinking.”^{1, 15, 16} The combination of blinking and of a single-mode intensity histogram is a fingerprint of a single-chromophore single protein. Among the population of individual HcRed proteins we have probed, the majority of the recorded single-molecule trajectories (43 out of 60) show two levels of intensity [as seen in Fig. 4b], while the rest show one level. The observation of two levels of intensity in single-molecule fluorescence signals is indicative of the presence of two absorbing chromophores, that is, of the presence of dimeric HcRed.

Fig. 3

Confocal fluorescence image of single HcRed proteins ( $5\times 5\mu \mathrm{m}$ , $125\times 125$ pixels, 10-ms integration time per pixel, average excitation power of $570\mathrm{nm}$ at the sample $2\mu \mathrm{W}$ ).

Fig. 4

(a) Fluorescence intensity signal detected from a single immobilized HcRed protein and featuring two levels. (b) histogram of photon count data from panel (a) and bimodal Gaussian fit (peaks at 290 and $490\mathrm{counts}∕10\mathrm{ms}$ ), clearly showing that two levels are present in the intensity signal from the left panel. (c) Fluorescence intensity signal detected from a single immobilized GFP and featuring one intensity level. (d) Histogram of photon count data from panel (c) and monomodal Gaussian fit (peak at $102\mathrm{counts}∕10\mathrm{ms}$ ).

3.2.

Intraprotein Energy Transfer in Dimers of HcRed

The center-to-center distance⁵ between chromophores within a dimer of HcRed is $\sim 3.2\mathrm{nm}$ . Assuming a molar extinction coefficient per red chromophore of $\mathrm{60,000}{\mathrm{M}}^{-1}{\mathrm{cm}}^1$ and a quantum yield of fluorescence of 0.05, we estimate¹⁷ a Förster radius for energy hopping (FRET between identical chromophores) of $R_0\sim 3.5\mathrm{nm}$ . This suggests that energy hopping between identical chromophores within a single HcRed dimer, illustrated in Fig. 5b , takes place with an efficiency¹⁷ $E=R_0^6∕(R_0^6+R^6)\sim 63\%$ .

Fig. 5

(a) Scheme accounting for directional intraprotein directional FRET in HcRed. Y, yellow chromophore; R, red chromophore; ${\mathrm{S}}_0$ , ground state; ${\mathrm{S}}_1$ , singlet excited state. (b) Scheme accounting for intraprotein energy hopping (homo-FRET) in HcRed.

We show evidence of intraprotein energy hopping in HcRed dimers by time-resolved fluorescence anisotropy experiments. Figure 1d shows the decay of the fluorescence anisotropy of HcRed in PBS (excitation at $590\mathrm{nm}$ , detection at $645\mathrm{nm}$ ). For a simple fluorophore, rotational diffusion is the dominant cause of fluorescence depolarization: following vertically polarized pulsed excitation, the fluorescence anisotropy $r\left(t\right)$ of a simple fluorophore decays as a single exponential,¹⁷ $r\left(t\right)=r_0\exp (-t∕{\tau }_{rot})$ , where ${\tau }_{rot}$ is the rotational correlation time of the fluorophore. For example, the fluorescence anisotropies of Texas Red and GFP decay as single exponentials with rotational correlation times of $\sim 0.25\mathrm{ns}$ and $16\mathrm{ns}$ , respectively, values reflecting their hydrodynamic volumes. For HcRed, the fluorescence anisotropy decays biexponentially, dominated by a fast $\left(0.3\mathrm{ns}\right)$ decay component in addition to a long $(>20\mathrm{ns})$ decay component that contributes more weakly. The long component is the rotational correlation time characteristic of the rotation of the whole HcRed dimer. Because of the relatively short fluorescence lifetime of this protein $(\sim 1.5\mathrm{ns})$ , an accurate determination of the long rotational correlation lifetime is difficult. We assign the short, dominant component to the energy hopping between identical (red) chromophores within an HcRed dimer. This process will induce additional depolarization in the fluorescence emitted by HcRed.¹⁷ We exclude the possibility that this fast component is due to rapid rotation of the chromophore inside the protein’s $\beta$ -can, because chromophores in GFP-like proteins are surrounded by a complex and relatively rigid hydrogen bond network involving several amino acids and water molecules.¹ This rigid network can be expected to forbid rotation of the chromophore on a short time scale similar to that of a small dye molecule rotating freely in solution (e.g., ${\tau }_{rot}\sim 250\mathrm{ps}$ for Texas Red).

The red chromophore is not the only species present in HcRed dimers. Under 470-nm excitation, the fluorescence spectrum of HcRed features a peak at $530\mathrm{nm}$ , apart from the main peak at $645\mathrm{nm}$ [Fig. 1a, line and dots spectrum]. When monitored at this wavelength $\left(530\mathrm{nm}\right)$ , the excitation spectrum of HcRed shows a main peak at $490\mathrm{nm}$ [Fig. 1a, dashed line and dots spectrum]. It is noteworthy that this 490-nm excitation peak has no corresponding peak in the absorption spectrum of HcRed. On $470\text{-}\mathrm{nm}$ excitation, the fluorescence of HcRed detected at $530\mathrm{nm}$ [Fig. 2c, curve 2] decays as a multiexponential with a major contributing lifetime of $0.04\mathrm{ns}$ (63% contribution) and two additional lifetimes of $0.14\mathrm{ns}$ (33%) and $0.51\mathrm{ns}$ (4%). For the same excitation wavelength $\left(470\mathrm{nm}\right)$ , the fluorescence decay detected at the main peak of the emission spectrum [ $645\mathrm{nm}$ ; Fig. 1c, curve 1] contains a rise time of $0.04\mathrm{ns}$ ( $-21\%$ contribution) and two lifetimes with similar values to those detected in the $640\text{-}\mathrm{nm}$ fluorescence decay recorded upon $570\text{-}\mathrm{nm}$ excitation. The stationary and time-resolved fluorescence data reported here [Figs. 1a, 1b, 1c] clearly show that the yellow species absorbing/emitting at 490/530-nm is connected with the red chromophore through an excited state pathway, as depicted in Fig. 5a. Given the overlap between the emission spectrum of the yellow chromophore [Fig. 1a, line plus dots] and the absorption spectrum of the red chromophore [Fig. 1a, dotted line], when both chromophores are part of the same HcRed dimer,⁵ conditions are satisfied for directional intraprotein energy transfer [hetero-FRET; Fig. 5a] to occur upon direct excitation of the yellow chromophore. FRET is demonstrated here by the match between the major decay time component detected at $530\mathrm{nm}$ (yellow emission, $0.04\mathrm{ns}$ ) and the rise time component detected at $645\mathrm{nm}$ (red emission, $0.04\mathrm{ns}$ ). The 0.04-ns decay time detected at $530\mathrm{nm}$ is the quenched fluorescence of the yellow chromophore due to FRET interaction with the red chromophore, while the 0.04-ns rise time detected at $645\mathrm{nm}$ is due to the delayed buildup in the excited state population of the red chromophore, as it is excited indirectly via FRET from the yellow chromophore.

The time-resolved fluorescence data we report here also show that a fraction of the yellow chromophores optically excited at $470\mathrm{nm}$ are not involved in directional intraprotein FRET toward red chromophores. A large subpopulation of yellow chromophores fluoresces with a lifetime of about $0.5\mathrm{ns}$ (37% of the total population of yellow species excited at $470\mathrm{nm}$ ). These chromophores are part of “yellow” HcRed dimers, i.e., dimers that contain a pair of yellow chromophores. Assuming that 0.51 and $0.04\mathrm{ns}$ are the fluorescence lifetimes of the yellow (donor) chromophore in the absence and presence of the red (acceptor) chromophore, directional intraprotein FRET from a yellow to a red chromophore takes place with an efficiency $E=R_0^6∕(R_0^6+R^6)=1-{\tau }_{DA}∕{\tau }_D\sim 92\%$ , thus yielding a Förster radius of $R_0=4.8\mathrm{nm}$ .

We believe that the yellow chromophore of HcRed is an immature form of the red chromophore. In another red fluorescent protein, DsRed from Discosoma genus, the red chromophore maturates through an intermediate, GFP-like green absorbing/emitting chromophore, a step that includes an additional oxidation step at one of the residues involved in the chromophore formation.^{9, 15} In DsRed, this maturation does not complete for every chromophore, leading to a distribution of green and red chromophores within individual oligomers, similar to what we see here in HcRed: yellow dimers (two yellow chromophores), mixed dimers (one yellow and one red chromophore), and fully matured dimers (two red chromophores). The absorption and emission properties of the yellow chromophore of HcRed are expected to be red-shifted in comparison to those of immature chromophores of DsRed because of the $\pi$ -stacking with the nearby His174 residue.⁵