INTRODUCTION

TOP ABSTRACT INTRODUCTION THEORETICAL BACKGROUND MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

Two-photon excitation (TPE) has developed as an important alternative to traditional one-photon excitation (OPE) in fluorescence microscopy and spectroscopy (Denk et al., 1990; Bennett et al., 1996; Malak et al., 1997; Svoboda et al., 1997; Bewersdorf et al., 1998). The intrinsic advantages of the two-photon process include reduced background fluorescence from fluorophores outside the focal volume, decreased photobleaching, inherent optical sectioning capability, and lower photodamage of sensitive biological samples (Denk et al., 1995). TPE is also directly applicable to a new class of fluorophore that is revolutionizing the visualization of biological systems and molecular interactions, the Green Fluorescent Protein (GFP) from the jellyfish Aequoria victoria, and its mutants (Xu et al., 1996; Tsien, 1998).

The various GFPs have attracted enormous attention in recent years as important reporters in cell, developmental, and molecular biology (for a comprehensive overview see Tsien (1998) and references therein). When fused to proteins of interest and expressed in vivo, GFP acts as a versatile indicator of structure and function within cells (Misteli and Spector, 1997) and can be visualized using standard microscopy techniques (Niswender et al., 1995; Presley et al., 1997) and multiphoton excitation (MPE) microscopy (Kohler et al., 1997). GFP and its constructs are being increasingly used in fluorescence lifetime imaging microscopy (FLIM) and fluorescence resonance energy transfer (FRET) modes of microspectroscopy for the visualization of protein-protein interactions, signaling, and trafficking in cellular systems (Bastiaens and Jovin, 1996; Miyawaki et al., 1997; Mahajan et al., 1998; Ng et al., 1999). The measurement of excited-state fluorescence lifetimes allows the discrimination of fluorophores with similar spectral emission properties. It is therefore of primary importance to understand the fundamental photophysical characteristics of GFP and its mutants at the typical excitation wavelengths used in two-photon microscopy systems (e.g., 800 nm excitation from titanium:sapphire lasers). Early work by Ward (Perozzo et al., 1988) and Prendergast (Nageswara Rao et al., 1980) established the fluorescence lifetime and anisotropy of GFP, and recent experiments have characterized in greater detail the OPE photophysical properties of some GFPs (Chattoraj et al., 1996; Lossau et al., 1996; Patterson et al., 1997; Kummer et al., 1998; Striker et al., 1999), including an extensive study of pH-dependent photophysics (Haupts et al., 1998). However, there have been no detailed comparisons of the OPE and TPE fluorescence properties of GFPs in the literature.

In this report we focus on the time-resolved detection of the TPE fluorescence intensity and anisotropy decays, and compare OPE and TPE at 400 nm and 800 nm, respectively, of the wild-type GFP and two mutant forms of the protein, the S65T mutant (in which Ser-65 is replaced by Thr) and a mutant exhibiting a red-shifted absorption peak (RSGFP; bearing the substitutions F64M, S65G, Q69L). The different selection rules for one-photon and two-photon excitation provide insight into the properties of excited states not accessible by one-photon spectroscopy. Furthermore, time-resolved multiphoton techniques permit the monitoring of the rotational diffusion and population decay of the excited molecule under unusual conditions, and thus constitute a new tool for structural and dynamic studies.

	THEORETICAL BACKGROUND

TOP ABSTRACT INTRODUCTION THEORETICAL BACKGROUND MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

TPE of a fluorophore involves the absorption of two photons in the same quantum event generating an electronically excited state, followed by the subsequent spontaneous emission of another (generally higher-energy) photon at the characteristic wavelength of fluorophore emission. This induced fluorescence signal displays a squared dependence on the exciting optical power. The basic equation relating the number of fluorescence photons emitted per molecule and unit time I⁽²⁾(t) (in the absence of saturation, self-quenching, photobleaching, or stimulated emission) to the experimental parameters for TPE is given by Xu and Webb (1996)

(1)

where  is the fluorescence quantum yield of the molecule, ₂ the two-photon absorption cross-section, and  the incident photon flux density. The factor 2 in the denominator reflects the fact that two photons are required for each absorption event. Since TPE is essentially an instantaneous process, the peak incident photon flux density of a pulsed laser source _peak determines the excitation rate.

Excitation of a randomly oriented molecular system with plane-polarized light creates an anisotropic orientational distribution of excited molecules that is markedly more different with TPE than with OPE. A complex analysis is necessary for a complete description of two-photon phenomena (Callis, 1993; Chen and VanDerMeer, 1993; Wan and Johnson, 1994). In the special case of a fluorescent probe with cylindrical symmetry and one dominating two-photon transition tensor element excited with plane-polarized identical photons, the initial anisotropy upon TPE, r₀⁽²⁾, (before any rotational diffusion) can be expressed by the general equation that also accommodates the OPE case:

(2)

where l is the number of photons involved in the excitation process and  is the angle between the dominant absorption and emission transition moment (Gryczynski et al., 1995). For the colinear case ( = 0) it is expected that increasing the number of photons leads to a more highly oriented excited-state population. Thus, the maximal anisotropies for OPE and TPE are 2/5 and 4/7, respectively, highlighting the potentially higher dynamic range of TPE in time-resolved anisotropy measurements. Values of r₀⁽¹⁾ reported by Partikian et al. (1998) and in this work approach the limiting value of 0.4, supporting the assumption that   0 for GFPs.

For this special TPE case (emission transition moment colinear with dominating two-photon transition tensor element) the total time-dependent fluorescence intensity can be measured under the same conditions as for OPE to suppress the effects of molecular rotation. The emission is monitored through a polarizer placed at 54.7° relative to the polarization direction of excitation (Lakowicz et al., 1992; Wan and Johnson, 1994). The fluorescence decay curves, F^(l)(t), are presumed to obey a multiexponential decay law

(3)

where _k is the kth fluorescence (lifetime) decay component with a corresponding amplitude _k, and ₀ is a constant background (generally negligible). The fractional steady-state intensity associated with component k is given by

(4)

The time-dependent emission anisotropy is determined for fluorescence decay curves recorded with the analyzing polarizer oriented parallel, F^(l) (t), and perpendicular, F^(l) (t), to the linearly polarized excitation light. The total anisotropy decay of a mixture of fluorophores is analyzed according to:

(5)

where ₀ has been omitted for clarity. The right-hand term of Eq. 5 corresponds to a kinetic model for n fluorescent species characterized by an excited-state lifetime, _k, rotational correlation time, _k, and initial anisotropy r_0k^(l) (Jovin et al., 1982). The individual rotational correlation times represent the combined global and local depolarizing motions of the particular domain of the macromolecule associated with the particular deactivation pathway. We assume that all species share the global rotational correlation time of the macromolecule such that the _k can be factored in the form

(6)

The fact that the rotational correlation times for GFP are much longer than the excited-state lifetimes does not justify a more sophisticated decomposition into eigenvector components.

	MATERIALS AND METHODS

TOP ABSTRACT INTRODUCTION THEORETICAL BACKGROUND MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

GFPs

Wild-type (WT), and mutants [S65T, bearing a single amino acid substitution; RSGFP, with the three substitutions F64M, S65G, Q69L (Delagrave et al., 1995)] of the cloned A. victoria GFP in the expression vector pRSETa (Invitrogen, Carlsbad, CA) were used. These plasmids, a gift of Dr. Rolando Rivera-Pomar (MPIbpc, Göttingen, Germany), code for expression of the GFP open reading frame with an additional six histidines at the amino terminus, under control of the T7 promoter inducible by isopropyl--D-thiogalactoside. The histidine tag permits purification of the recombinant protein using standard procedures with a Ni-chelating resin (Ni-NTA-Agarose, Qiagen, Hilden, Germany). Protein purity was confirmed by SDS-PAGE. Further purification was performed as required with Pharmacia Superdex-75 size-exclusion chromatography. The purified protein was dialyzed extensively against 10 mM sodium phosphate buffer, pH 7.0, concentrated, and stored. All spectroscopic measurements were carried out with ~18 µM protein dissolved in 10 mM Tris-HCl buffer, pH 8.0, 0.1 M NaCl at room temperature (~22°C).

Fluorescence spectroscopy

Details of the experimental configuration consisting of a femtosecond titanium:sapphire laser system, a time-correlated single-photon counting (TCSPC) arrangement for time-resolved detection, and an optical fiber-coupled CCD spectrophotometer (Oriel MS127i/INSTASPEC) for steady-state spectral measurements were described previously (Volkmer et al., 1997). Briefly, a regeneratively amplified titanium:sapphire mode-locked laser system (800 nm excitation, 250 kHz repetition rate, 180 fs pulsewidth; Coherent Mira 900/RegA 9000) was used to generate pulses of up to 4 µJ energy. OPE at the same effective energy as 800 nm TPE was achieved by frequency doubling the titanium:sapphire fundamental output.

For fluorescence decay measurements a Glan Taylor prism polarizer was oriented at the magic angle (54.7°) relative to the polarization vector of the excitation beam to suppress the contributions of rotational depolarization to the fluorescence decay signal (see Theory section). Fluorescence anisotropy decay measurements were carried out by recording the fluorescence with the polarizer alternating between parallel and perpendicular orientations to the excitation polarization, with a dwell time of 30 s in each position to correct for excitation intensity drifts. The emerging light was imaged onto a microchannel plate photomultiplier detector (MCP-PM; Hamamatsu R1712U-03). Fluorescence decays were collected using standard TCSPC electronics in 1K channels at 0.0675 ns/ch.

In the time-resolved measurements the emission was spectrally isolated using bandpass interference filters with center wavelengths at 456 nm (FWHM = 70 nm) and 514 nm (FWHM = 8 nm). Additional steady-state fluorescence measurements were performed on an SLM 8000C fluorimeter and used to determine the relative quantum yields.

Data analysis

Least-squares re-convolution fits of fluorescence intensity decay records were carried out using the IBH decay analysis software library (IBH Consultants Ltd., Glasgow, Scotland). Inasmuch as the instrumental response function of this setup was ~50 ps FWHM (compared to nanosecond decay times), the experimental anisotropy decay curves were fitted to the above model to give an estimate of the anisotropy decay parameter without a deconvolution correction for the finite instrumental response time. Errors are reported as ±3 standard deviations, as determined by the fits to the data. We note the inherently large inaccuracy in determining rotational correlation times that are an order of magnitude larger than the corresponding fluorescence lifetime. Errors for the RSGFP anisotropy decays were derived from least-square fit results using Kaleidagraph (Synergy Software, Reading, PA).

	RESULTS

TOP ABSTRACT INTRODUCTION THEORETICAL BACKGROUND MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

Steady-state spectra

WT-GFP and the S65T and RSGFP mutants excited at 800 nm yielded the fluorescence spectra displayed in Fig. 1. The characteristic GFP bands had peaks at ~511 nm in the case of WT-GFP and RSGFP, and ~514 nm for S65T-GFP, and were almost indistinguishable from those observed with OPE at 400 nm. However, small differences emerged upon more detailed analysis of the wavelength region around 550 nm and the weak emission band at ~440 nm. The relative intensities of both spectral features increased upon excitation at 800 nm compared to 400 nm excitation. For OPE at 400 nm, the relative fluorescence intensity of the 440 nm band of S65T-GFP was approximately twice that of WT-GFP or RSGFP. For TPE with 800 nm photons, however, the relative intensities of the blue emission bands of both S65T- and WT-GFP were similar, whereas RSGFP showed comparatively low levels of 440 nm fluorescence.

View larger version (20K): [in this window] [in a new window]   FIGURE 1   Normalized steady-state OPE (exc = 400 nm; dashed lines) and TPE (exc = 800 nm; solid lines) fluorescence emission spectra of WT-GFP, S65T-GFP, and RSGFP.

For excitation at both 400 nm (_exc used in the time-resolved fluorescence experiments) and 490 nm (peak of the absorption), the emission quantum yield () of RSGFP was estimated to be 0.30 times that of S65T-GFP by comparing the integrated emission from samples of S65T-GFP and RSGFP normalized by the absorbance at the excitation wavelength. For RSGFP, this yields an absolute value of  ~ 0.19 based upon the published value of 0.64 for S65T-GFP (Patterson et al., 1997).

Excitation power dependence

The power dependence of the fluorescence detected at 514 nm was determined from a log-log plot of the fluorescence signal versus incident peak photon flux density (shown in Fig. 2 for RSGFP). For photon flux densities 1.6 × 10³⁰ photons cm² s¹, the induced fluorescence obeyed a power-squared intensity dependence as indicated by the measured slope of 2.02 ± 0.03, thereby confirming the existence of TPE. However, a decrease in the apparent power exponent was observed for larger irradiances. All time-resolved measurements were carried out at the lower intensity levels at which deviations from the second-order power law were absent.

View larger version (18K): [in this window] [in a new window]   FIGURE 2   Log-log plot of the fluorescence emission (514 nm) of RSGFP versus the incident photon flux density for TPE excitation at 800 nm. A linear regression fit for photon flux densities 1.6 × 1030 photons cm2 s1 yields a slope of 2.02 ± 0.03, confirming the two-photon nature of the excitation. Saturation of the fluorescence is observed for higher photon flux densities.

Fluorescence intensity decays

The analyses of the time-resolved fluorescence intensity decay measurements performed under magic angle conditions are summarized in Table 1. The measured fluorescence decays of WT- and S65T-GFP were predominantly monoexponential with a fluorescence lifetime of ~3 ns. With both OPE and TPE the emission at 514 nm was well-described by this single decay time, whereas better fits to the 456 nm emission data were obtained with a biexponential decay model yielding a small additional subnanosecond component of ~0.2 ns. While WT-GFP and S65T-GFP showed almost identical fluorescence decay kinetics, i.e., monoexponential at 514 nm and biexponential at 456 nm emission, RSGFP did not follow the same trend. The fluorescence decay curves of RSGFP were best described by a three-exponential decay model with lifetimes of ~0.4, ~1.1, and ~3.3 ns, and similar fractional intensities at both observation wavelengths.

Based on the quantum yields for S65T-GFP and RSGFP (see above), and fluorescence lifetimes measured at 514 nm for S65T-GFP and RSGFP (Table 1), and the radiative scheme of Fig. 4 we calculated the radiative (k_f) and nonradiative (k_NR) decay rates for both proteins according to the relations  = k_f ii, where _i = (k_f + k_NR,i)¹. For S65T-GFP we used the single lifetime  = 3.01 ns, while for RSGFP we used the two longer lifetime components (i = 2, 3; see Table 1), since the subnanosecond component associated with deactivation of the protonated species A* is not monitored at this emission wavelength and, hence, its contribution to the total quantum yield is negligible. These parameters yield k_f^S65T  2.1 × 10⁸ s¹, k_NR^S65T  1.2 × 10⁸ s¹, and k_f^RSGFP  1.4 × 10⁸ s¹. Assuming that the calculated k_f is unique for RSGFP (since it is associated with the chromophore in the I* configuration; see Fig. 4), we estimate that the nonradiative decay rates associated with the 1.1 ns and 3.3 ns lifetimes are k_NR^RSGFP,1.1ns  7.4 × 10⁸ s¹ and k_NR^RSGFP,3.3ns  1.5 × 10⁸ s¹, respectively.

Fluorescence anisotropy decays

Fluorescence anisotropy decays were monitored at 514 nm for all three types of GFP and are shown in Figure 3. The anisotropy parameters obtained from fitting the experimental data to Eq. 5 are summarized in Table 2.

View larger version (28K): [in this window] [in a new window]   FIGURE 3   Time-resolved fluorescence anisotropy decays of WT-GFP (A), S65T-GFP (B), and RSGFP (C) upon OPE at 400 nm (lower curve in each panel) and TPE at 800 nm. Fits to a monoexponential anisotropy decay model are shown for A and B. The decay model for RSGFP is discussed in the text. Residuals for all fits are indicated (OPE: lower curve).

As in the case of the fluorescence intensity decay measurements, WT- and S65T-GFP exhibited almost identical fluorescence anisotropy kinetics (Fig. 3, A and B). The initial anisotropy for 400 nm excitation was ~0.35. TPE at 800 nm, however, resulted in an r₀ of ~0.51, exceeding the one-photon value. The measured fluorescence anisotropy decays of WT-GFP and S65T-GFP were best described by an isotropic rotational motion of the chromophore with a monoexponential decay characterized by a rotational correlation time of ~16 ns.

In comparison to WT- and S65T-GFP, RSGFP exhibited markedly different fluorescence anisotropy decay at 514 nm. It was not possible to fit the OPE and TPE anisotropy decay curves with a monoexponential model. Two versions of Eq. 5 were explored further. In one, a unique correlation time _global was assumed, but the initial anisotropies (r_0k^(l)) of the three decay components were allowed to vary. No parameter set was found that could adequately represent the anisotropy curve in this manner. In the second approach, the ₂ = 1.1 ns decay component unique to RSGFP was assigned an additional rotational correlation term _local,2 (Eq. 6). The total decay function (denominator of Eq. 5) was fitted using the lifetimes from Table 1 and the resulting decay parameters (Eq. 3) were used to fit the difference curve (numerator of Eq. 5). The results were very satisfactory, yielding (for OPE) a _global of ~17 ns, a global r₀ value of 0.32, and an additional rotational correlation term (_local,2  13 ns) corresponding to the 1.1 ns lifetime. The same procedure, applied to the TPE data, yielded a higher _global of ~27 ns, a global r₀ value of 0.48, and _local,2  13 ns. Note that the determination of rotational correlation times that are significantly longer than the corresponding fluorescence lifetimes is inherently difficult and prone to errors. Fixing _global to a value <20 ns did not lead to an appreciable reduction in the quality of the fit. The resulting anisotropy functions (OPE/TPE) are shown in Fig. 3 C. For all three proteins, measurements with a fivefold increased time resolution did not yield evidence for a fast anisotropy decay in the subnanosecond range.

	DISCUSSION

TOP ABSTRACT INTRODUCTION THEORETICAL BACKGROUND MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

Decay scheme for GFPs

To interpret the data for RSGFP in terms of a model applicable for S65T- and WT-GFP, we invoke a modified excited-state proton transfer (ESPT) reaction scheme (Fig. 4). ESPT has been proposed on structural grounds (Brejc et al., 1997; Palm et al., 1997) and used to interpret the photophysical characteristics of GFP and its mutants in the frequency and time domain excited via OPE (Chattoraj et al., 1996; Lossau et al., 1996).

View larger version (20K): [in this window] [in a new window]   FIGURE 4   Decay scheme for GFPs, adapted from excited state proton transfer models (Chattoraj et al., 1996; Lossau et al., 1996). A, protonated species; B, deprotonated species; I, deprotonated intermediate species. The corresponding excited states (upon OPE at 400 nm or TPE at 800 nm) are denoted by asterisks. As found in this work, A* and B* transfer to I*; the chromophore environment in I* is assumed to reflect the excitation history of the state, and thus two substates are invoked, I*A and I*B, with different characteristic nonradiative decay rates. The B*  I* transition is not accessible in WT-GFP and S65T, but is favored in RSGFP, (dashed arrows; Creemers et al., 1999; Creemers et al., 2000). Interconversion between I*A and I*B is presumed to be negligible during the excited state lifetime. Approximate lifetimes and emission peaks are assigned to each decay path. In the case of WT- and S65T-GFP the green emission arises from both I* and B*. High-resolution experiments on WT-GFP have resolved the lifetimes associated with these states as ~3.3 ns and ~2.8 ns, respectively (Striker et al., 1999) (see text for discussion).

The model is comprised of a protonated state of GFP (absorption peak at ~400 nm, termed A), a deprotonated state (absorption peak at ~480-490 nm, B), and an intermediate deprotonated species I. Recent high-resolution spectral hole-burning experiments on wild-type GFP at cryotemperatures have directly identified the intermediate I in the ground state and established the 0-0 transition of all three species (Creemers et al., 1999). OPE at 400 nm and TPE at 800 nm result in the excitation of the protonated (A) and deprotonated (B) ground states of the chromophore. In WT-GFP at pH 8 the A ground-state absorption is dominant, whereas in S65T-GFP and RSGFP the equilibrium in the ground state is strongly shifted toward the deprotonated molecular forms.

Depopulation of A* occurs via 1) highly dispersive ESPT on the picosecond time scale (Chattoraj et al., 1996; Lossau et al., 1996) forming an excited deprotonated species in the A configuration (I*) [which cannot achieve an equilibrium conformation during its short lifetime (Lossau et al., 1996)]; 2) photoconversion between A* and B [photochromicity, (Striker et al., 1999)]; and 3) radiative and nonradiative deactivation processes to the ground state. The radiative transition manifests itself as a weak blue emission at ~440 nm, indicating that the dominant depopulation channel of A* is via ESPT to I*.

The green emission is thought to originate from both the I* and B* forms. The competing B*  I* transition following excitation into the B* state is dependent on the height of the corresponding energetic barrier. The energy schemes of various GFPs have been estimated by high-resolution hole-burning and excitation/emission spectroscopy (Creemers et al., 1999; Creemers et al., 2000); they depend strongly on the specific amino acid substitutions. Thus, the B*  I* transition does not occur in WT-GFP (Creemers et al., 1999) or S65T, but is energetically favored in RSGFP (Creemers et al., 2000; also see below). In the scheme of Fig. 4 we postulate that the excited electronic intermediate state I* may be distinguished by an excitation-history-dependent intramolecular (chromophore) environment; that is, whether I* is accessed via the A* or the B* manifolds. The terminology I^*_A and I^*_B denotes these substates, each of which is presumed to have a (different) characteristic nonradiative decay rate reflecting the subtle differences in chromophore environment. This modified kinetic scheme is the simplest model consistent with the time-resolved fluorescence data for GFPs (Striker et al., 1999; present work; see below) and the hole-burning spectroscopy data upon OPE (Creemers et al., 2000).

We note that in a recent theoretical treatment of GFP photophysics (Weber et al., 1999), a fourth state (Z/Z*) representing a zwitterionic GFP chromophore was postulated in addition to the standard A, I, and B species (Weber et al., 1999). Z* could be a "dark" nonfluorescent state for some GFPs (e.g., the WTGFP), but have a finite quantum yield in the case of the other mutants. We do not mean to imply a correspondence between Z* and I^*_B, however.

Steady-state OPE and TPE fluorescence

The green emission in all three GFPs observed upon 800 nm excitation was in general agreement with the fluorescence spectra excited by OPE at 400 nm. This fact, coupled to energy conservation considerations, indicates a simultaneous absorption of two 800 nm photons. Direct evidence for the TPE phenomenon was provided by the measured power-squared dependence of the induced fluorescence intensity on the incident 800 nm photon flux density and the zero-time anisotropies upon 800 nm excitation exceeding the OPE limit of 0.4. The independence of the fluorescence intensity and anisotropy decay kinetics on the multiplicity of photon excitation also suggests that the emission occurs from the same excited states.

The TPE fluorescence clearly saturated at higher flux densities, thereby establishing an upper limit for quantitative TPE microscopy of ~1.6 × 10³⁰ photons cm² s¹. The apparent saturation may be attributed to a variety of nonlinear optical processes, such as stimulated emission (Kusba et al., 1994), excited-state absorption (Bradley et al., 1972), photolysis (Brand et al., 1997), and ground-state depletion (Xu et al., 1996). Nonperturbative nonlinear phenomena caused by a high instantaneous field strength become important only at intensities 10³¹ photons cm² s¹ (i.e., 10⁷ V cm¹) (Xu et al., 1996; Brand et al., 1997). Stimulated emission could also be excluded, since there was no overlap of the excitation wavelength of 800 nm with the fluorescence (also see Fig. 1). Fluorescence saturates at the limit of one transition per pulse per fluorophore. For a two-photon process, saturation occurs when _{2 peak}² _pulse  1 (Xu et al., 1996). Assuming a typical value of ₂ ~ 3 × 10⁵⁰ cm⁴ s photon¹ for GFP (Xu et al., 1996), _pulse = 180 fs, and a threshold value for the saturation peak intensity of _peak  1.6 × 10³⁰ photons cm² s¹ (Fig. 2), the above relation yields ~0.014. Since OPE at 800 nm can be neglected, this result indicates that <1.4% of molecules in the focal spot are excited per laser pulse in TPE. Therefore, ground-state depletion cannot have been the dominant contribution to the observed saturation. In contrast, an analogous estimation for the OPE case suggests that excited singlet-state absorption efficiently depopulates the excited state. Indeed, Lossau et al. (1996) recently reported excited-state absorption features of WT-GFP in the spectral range 630-950 nm, peaking at ~700 nm. To perform a quantitative analysis of the deviation of the fluorescence from the power-square law as shown in Fig. 2, a rate equation formalism has to be used, taking into account the two-photon cross-sections at 800 nm reported by Xu et al. (1996) and cross-sections not yet reported for one-photon excited-state absorption. In conclusion, quenching of the excited singlet state by excited-state absorption at high power levels was the most likely cause for the observed deviation from the simple power law.

The relative steady-state emission intensity at 440 nm of all three systems increased upon 800 nm TPE when compared to 400 nm OPE. This may indicate a decreased ESPT rate for A* depopulation upon 800 nm excitation or, alternatively, a change in the ratio of ground-state absorption of the protonated and deprotonated species. In the former case, the electronic excited states achieved with OPE and TPE may be different due to the different photoselection rules and may consequently have different ESPT susceptibilities. In this scenario, the two-photon allowed excited state would have a lower ESPT rate, thus favoring the radiative (fluorescent) deactivation pathway. A second approach to interpreting the enhanced blue fluorescence intensity is based on the observation that absorption of 400 or 800 nm photons results in simultaneous excitation of all ground-state species to an extent dependent on their corresponding absorption cross-sections. The enhanced blue emission intensity upon 800 nm TPE indicated an increased population of A* relative to B*, suggesting that the ratio ₂(A)/₂(B) at 800 nm exceeds ₁ (A)/₁(B) for 400 nm excitation. Since steady-state absorption and fluorescence excitation spectra (not shown), obtained with high-resolution spectroscopy techniques in particular (Creemers et al., 1999; Creemers et al., 2000), provide direct evidence for the simultaneous excitation of all ground-state species, we favor the second interpretation.

Fluorescence intensity decay kinetics

WT- and S65T-GFP

RSGFP

Fluorescence anisotropy decay kinetics

In all cases the initial anisotropy values measured with 800 and 400 nm excitation approached the theoretical limits of 0.57 and 0.40 for TPE and OPE, respectively. The measured OPE r₀ agreed well with the value of ~0.39 for humanized red-shifted GFP recently reported by Partikian et al. (1998). In conclusion, the one- and two-photon induced fluorescence anisotropies of the GFPs studied were found to obey the theory derived for polyene-like molecules (Eq. 2), according to which the 400 nm absorption dipole or correspondingly the dominant element of the 800 nm two-photon transition tensor is almost colinear with the 514 nm emission transition dipole (  0).

There was good agreement within experimental uncertainty between the rotational relaxation times in OPE and TPE, as expected since rotational rates are determined only by the size and shape of the probe molecule and the nature of its microenvironment. The finding also confirms the absence of significant heating by the intense laser beam in TPE. Such an effect would have caused a decrease of both the fluorescence intensity and rotational decay times. A mean rotational correlation time calculated from the data for all the GFPs studied was ~16 ns. Due to the absence of any measured fast (subnanosecond) anisotropy decay, this value corresponds to the rotation of the entire protein without significant contributions from segmental motions of the chromophore. The rotational correlation times obtained for S65T-GFP were comparable to the ~19 ns previously reported for this mutant in PBS (pH 7.4) using OPE at 492 nm (Swaminathan et al., 1997).

The theoretical value for the rotational correlation time of GFP, assuming the rotation of a 31-kDa isotropic rotator (Ormo et al., 1996) and a hydrated volume of a typical protein of ~1 cm³ g¹ (Cantor and Schimmel, 1980), would be _sp ~ 13 ns. Since the x-ray structure of GFP is known, one can readily compare the experimental value of  with the theoretical predictions in more detail. The plane of the chromophore group lies buried inside the cylinder-shaped protein oriented at an angle of ~60° to the symmetry axis of the cylinder of 2.4 nm diameter and 4.2 nm height (Ormo et al., 1996; Yang et al., 1996). Three anisotropy decay components are predicted theoretically using a prolate ellipsoid model with an axial ratio of 1.75 (Kawski, 1993), yielding correlation time components expressed as a function of the calculated isotropic correlation time _sp of ₁ = 1.34 × _sp, ₂ = 1.22 × _sp, ₃ = 0.95 × _sp, with fractional intensities of 2%, 62%, and 36%, respectively. In the experimental anisotropy decays of WT- and S65T-GFP, however, only a single rotational correlation time could be resolved, corresponding to the average of the three theoretically derived correlation time components weighted by their fractional intensities. The theoretical average is 14.5 ns, in good agreement with the experimental values (Table 2). These values assume that the protein is in monomeric form, which is expected for the concentrations of proteins used in these experiments (Ward, 1998).

In contrast to the monoexponential anisotropy decay of WT- and S65T-GFP, the red-shifted mutant RSGFP exhibited a behavior indicative of an apparent internal (local) motion of the chromophore, due to increased flexibility and/or passage between the different excited-state species. This interesting finding is under further investigation. The global rotational correlation time, however, was in agreement with that of WT- and S65T-GFP, reflecting the rotational motion of the entire protein and consistent with the expectation that these mutations do not substantially alter the tertiary structure of the protein. Such an internal motion has been recently invoked in the context of shortened excited-state lifetimes and fast ground-state recovery in specific red-shifted GFP mutants (Kummer et al., 1998).

	CONCLUSION

TOP ABSTRACT INTRODUCTION THEORETICAL BACKGROUND MATERIALS AND METHODS RESULTS DISCUSSION CONCLUSION REFERENCES

Time-resolved detection of multiphoton-induced anisotropic fluorescence using TCSPC techniques confirms that a higher degree of molecular orientation for TPE is obtained in comparison to OPE. TPE thus offers a more sensitive measurement of fluorescence anisotropy decays of GFPs when probing the dynamics and structure of nanometer-scale biological systems using its higher intrinsic anisotropy. In particular, this potential could be exploited in in vivo fluorescence microspectroscopy and diffusion studies using GFP or GFP-tagged molecules. Despite the complexity of the fluorescence intensity and anisotropy decay kinetics of RSGFP, its significantly shorter average lifetime (~1.7 ns) would allow its discrimination in fluorescence lifetime imaging microscopy from other spectrally similar GFPs (e.g., S65T-GFP) exhibiting the usual ~3 ns fluorescence lifetime.