505
2007, 111, 505-507 Published on Web 12/23/2006
Origin of the Anomalous Two-Photon Absorption in Fluorescent Protein DsRed Riccardo Nifosı`*,† and Yi Luo‡ NEST-CNR INFM and Scuola Normale Superiore, Piazza dei CaValieri 7, I-56126 Pisa, Italy, and Department of Theoretical Chemistry, The Royal Institute of Technology, AlbaNoVa UniVersity Center, SE-106 91 Stockholm, Sweden ReceiVed: December 6, 2006
The red fluorescent protein DsRed displays a two-photon excitation band around 760 nm which is not accompanied by any feature in the corresponding one-photon spectral region (380 nm). By means of timedependent density functional theory, we are able to explain such an effect, as arising from an electronic excitation of the DsRed chromophore with ability to couple with a charge-transfer state, through an effective two-photon absorption channel.
Multiphoton fluorescence microscopy stands at the frontier of biological imaging technology.1 Though currently rather expensive, multiphoton excitation offers important advantages over confocal microscopy, such as three-dimensionally localized excitation and confined photobleaching and photodamage. Intrinsically fluorescent proteins (FPs), among which the wellknown green fluorescent protein (GFP), have rather large twophoton action cross sections and are hence now increasingly used in combination with multiphoton microscopy for studies of tissue explants and live animals (see ref 1 and references therein) as well as in protein trafficking studies.2,3 Two-photon imaging of single FP was also reported, together with multiphoton-induced on/off switching of fluorescence in a yellow GFP mutant.4 An optimal two-photon excitation wavelength can often be obtained by doubling the maximum one-photon excitation, and normally, the two-photon excitation profile of FPs at halved wavelength is found to mimic the one-photon excitation.5,6 A noticeable exception to this rule is DsRed, a bright red fluorescent protein from a Discosoma coral. Parker and coworkers2 and Schmidt and co-workers5 measured an unexpected increase of the two-photon excitation of DsRed at λ < 760 nm, not supported by any one-photon band at the corresponding wavelength (390 nm). Interestingly, femtosecond-pulsed irradiation with λ < 760 nm rapidly changes the fluorescence of DsRed from red to green, possibly by bleaching the mature, red-emitting chromophore within the DsRed tetramer.2 Thanks to the stability of color change and the three-dimensional localization of multiphoton excitation, DsRed “greening” can be exploited to optically highlight subcellular compartments. Despite such potential interest, the presence of the two-photon band at λ < 760 nm has not yet been explained. Indeed, the lack of symmetry of the chromophore within the protein matrix prevents one from invoking simple selection rules to understand why this transition is a unique feature of the two-photon spectrum. By means of time-dependent density functional theory * To whom correspondence should be addressed. E-mail:
[email protected]. † NEST-CNR INFM and Scuola Normale Superiore. ‡ The Royal Institute of Technology.
10.1021/jp068380j CCC: $37.00
(TD-DFT), we are able to explain such an effect, as arising from an electronic excitation of the DsRed chromophore with ability to couple with a charge-transfer state, through an effective twophoton absorption channel. The model chromophore (Figure 1) was obtained by substituting the Gln65 side chain with a methyl group and cutting the connections to the protein backbone and saturating with hydrogen atoms. We considered only the anionic protonation form, which is presumably the stable state inside the protein matrix.7 Neglecting the influence of surrounding amino acids is certainly a major approximation when calculating the optical properties of fluorescent proteins. Deviations of some tens of nanometers in absorption/fluorescence peaks are measured in different mutants of GFP8 and of model chromophores in various solvents.9 However, as shown by theoretical and experimental studies, the gas-phase chromophore model is able to capture the main features of the absorption spectrum for GFP.10,11 Additionally, it is certainly the first step to understand the more complex system (i.e., chromophore within the protein environment). To our knowledge, this is the first calculation of the twophoton cross section for a chromophore of the fluorescent protein family. Excitation energies, one-photon cross sections (σOPA), and two-photon cross sections (σTPA) are accessible to TD-DFT12 through linear13 and quadratic response theory.14,15 The reference quantity for two-photon calculations is the transition moment δTPA ) 1/15ΣijSiiS*jj + 2SijS*ij, where ij are Cartesian indexes (throughout the paper, it is assumed that polarization of incident light is linear). Sij is the two-photon transition matrix, expressed in terms of the expectation values of the dipole moment operator µi between the ground and the excited states as
Sij )
∑n
[
nf µ0n i µj
ω0n - ω0f /2
+
nf µ0n j µi
ω0n - ω0f /2
]
(1)
where µab ) 〈a|µ|b〉, f is the final state, and the sum runs over all excited states n. ω0n and ω0f are the excitation energies of the intermediate state n and the final two-photon state f, © 2007 American Chemical Society
506 J. Phys. Chem. B, Vol. 111, No. 3, 2007
Letters
Figure 1. DsRed model chromophore.
Figure 3. Isodensity surfaces of the relevant molecular orbitals.
Figure 2. Comparison of the calculated σTPA and σOPA values of the DsRed model chromophore with the experimental absorption spectrum7 (green squares) and two-photon action cross section5 (blue triangles) of DsRed. The units on the y-axis are arbitrary. a, b, and c denote the three peaks discussed in the text. The theoretical peaks were broadened using a Lorentzian line shape with a line width of 0.03 eV.
TABLE 1: Excitation Wavelengths (w.l.), Oscillator Strengths (o.s.), Two-Photon Transition Moments (δTPA), and Cross Sections (σTPA; 1 GM ) 10-50 cm4‚s/Photon) exc.
w.l. (nm)
o.s.
δTPA (au)
σTPA (GM)
1 2 3 4 5 6
524 452 382 377 370 336
0.96 0.00 0.02 0.00 0.00 0.23
2.8 × 103 3.5 × 10-1 1.6 × 103 4.6 × 10 6.8 × 10
22.7 0.0 24.9 0.7 1.1
a
exp. (nm) 558a (a) 380/760b (b) 335a (c)
From ref 7. b From ref 5.
respectively. Rather than using eq 1, Sij can be obtained from the residues of the quadratic response function.16 The latter technique also provides the transition dipole moments between excited states. From δTPA, σTPA can be calculated by introducing the phenomenological line width,17 which was set to 0.1 eV (a smaller line width was used for the spectra in Figure 2, in order to make the peaks more distinguisheable). Such a value is commonly employed in comparisons with experiments and roughly corresponds to the line width of the experimental absorption spectrum. All optical properties were calculated using the B3LYP18 exchange-correlation functional and split valence basis set with addition of diffuse functions (6-31+G*), while geometry optimization of the model chromophore was performed with the 6-31G* basis set. As in a previous study,19 the sCdNs CdO dihedral angle in the conjugated tail deviates from planarity by about 30°. Excitation energies/oscillator strength calculations were carried out with Gaussian 03,20 while twophoton cross sections and excited-state transition dipole moments were computed with Dalton.21 The first six excitations were considered for one-photon cross section calculation, and only five in the case of σTPA because of convergence issues (see the Supporting Information for results using more states
with a smaller basis set). The good performance of this method (B3LYP/TDDFT quadratic response) for TPA properties of organic chromophores was assessed in previous studies.14,22 Table 1 reports the excitation wavelengths, oscillator strengths, and two-photon cross sections of the calculated electronic transitions, and Figure 2 compares the resulting spectra with the experiments.2,5,7 The first excitation has a relatively large oscillator strength and is mainly HOMO-LUMO with partial charge-transfer character. The excitation wavelength (524 nm) compares fairly well with the absorption peak in the protein (558 nm, peak a in Figure 2). The ∼30 nm deviation can be attributed to the effect of the environment as well as to systematic overestimation of excitation energy in B3LYP calculations in the anionic chromophores of fluorescent proteins.10,23 The first excitation has also relatively strong twophoton character, and the resulting σTPA value is of the same order of magnitude as that found in the experiments (11.01 GM5). The limited detection window of the two-photon spectrum (up to 1050 nm) prevents one from following the behavior of this first excitation completely. We presume that the two-photon peak at ∼970 nm belongs to the vibronic progression clearly observed for the first excitation in the onephoton spectrum. While it is expected that the charge-transfer state is both onephoton and two-photon allowed,17 more interesting is the case of the third excitation, |0〉 f |3〉, which has a rather low oscillator strength (0.02) but large σTPA value (22.7 GM). Such a characteristic together with its excitation wavelength, 382 nm, prompts us to assign this transition to the two-photon excitation band at 730-760 nm2,5 (peak b in Figure 2). The decomposition of |0〉 f |3〉 in Kohn-Sham single particle excitations contains HOMO-to-LUMO+1 and HOMO-2-to-LUMO transitions (see Figure 3). Interestingly, these transitions involve the conjugated “tail” of the chromophore. Indeed, GFP and its mutants, where the conjugated tail is absent, do not show any enhancement of σTPA at λ < 760 nm.5 The structure of eq 1 for the two-photon transition moment allows one to understand the origin of the large σTPA value of |0〉 f |3〉. Indeed, while the transition dipole moment µ03, and hence the one-photon transition probability, is small, the twophoton transition moment can couple to excited states |i〉 with large µ0iµi3, and with excitation energy ω0i close to half ω03 ) 3.25 eV. One such state is |1〉, with ω01 ) 2.37 eV, for which |µ01| and |µ13| are both large (4.53 and 0.92 au, respectively). The other possible channels are suppressed either by the weak |0〉 f |i〉 and |i〉 f |3〉 transition moments (see Table 1 and the Supporting Information), and/or by the large ω0i - ω03/2 energy difference. The calculated permanent dipole moment of state |3〉, µ33, is larger than that of the ground state, µ00 (4.8 vs 3.2 au). Such increased charge separation upon excitation to |3〉 can reduce
Letters the stability of the chromophore and possibly lead to irreversible photochemical reaction. Photobleaching by