Resolution of the Lifetimes and Correlation Times of the Intrinsic Tryptophan Fluorescence of Human Hemoglobin Solutions using 2 GHz Frequency-domain Fluorometry*

We used 2 GHz harmonic content frequency-domain fluorescence to measure the intensity and the anisotropy decays from the intrinsic tryptophan fluores- cence from human hemoglobin (Hb). The tryptophan intensity decays are dominated by a short-lived component which accounts for 35-60% of the total steady state intensity. The decay time of this short component varies from 9 to 27 ps and this component is sensitive to the ligation state of Hb. Our error analyses indicate the uncertainty is about k3 ps. The intensity decays also show two longer lived components near 0.7 and 8 ns, which are probably due either to impurities or to Hb molecules in conformations which do not permit energy transfer. The anisotropy decays indicate the tryptophan residues in Hb are highly mobile, with apparent correlation times near 55 ps.

the uncertainty is about k3 ps. The intensity decays also show two longer lived components near 0.7 and 8 ns, which are probably due either to impurities or to Hb molecules in conformations which do not permit energy transfer. The anisotropy decays indicate the tryptophan residues in Hb are highly mobile, with apparent correlation times near 55 ps.
There is a considerable interest in the structure and dynamics of hemoglobin which resulted in elucidation of its threedimensional structure allosteric properties and "static" models which try to explain the structure-function relationships of the system (1). There is now a desire to understand how the dynamics of the molecule contributes to its functional properties. This problem has been approached by computer simulations, probing the local motions of individual residues (in the picosecond and nanosecond range), and the correlation of these motions with protein conformation and ligand binding (2,3).
Processes which occur on the picosecond-nanosecond timescale can be studied by fluorescence spectroscopy. While such measurements have always been possible in principle, the time resolution and sensitivity of fluorescence methods have increased considerably within the past several years. In hemoglobin, fluorescence spectroscopic studies have been limited because the emission of either extrinsic or intrinsic probes is strongly quenched by resonance energy transfer to the heme (4,5). Nonetheless, several laboratories have published careful studies (6)(7)(8) which indicate that the intrinsic tryptophan emission from Hb' can be detected. Recent time-resolved studies on Hb (9)(10)(11) indicate the decay times are 100 picosecond or less, which are consistent with quenching by energy transfer to the heme. In all these studies the authors also detected longer lived components on the nanosecond timescale. The origin of these components is difficult to determine, and, in at least one case (Aplysia myoglobin), has been interpreted as the result of difficult to remove impurities (12). It should be stressed that picosecond timescale decay times are consistent with the values expected for the tryptophan residues, after consideration of the 100-fold quenching produced by the heme. In this report we emphasize interpretation of the picosecond decay times, which we believe are due to the intrinsic tryptophan emission.
In general it is difficult to obtain picosecond resolution of fluorescence decay times. If one uses time-correlated single photon counting, then the pulsed source must have a comparably short pulse width and the detector must have picosecond resolution. The more exotic probe-pump technique circumvents the time response of the detector by using the optical pulse as a gate. However, the sensitivity of this method is low, and the method has only occasionally been applied to protein fluorescence. In this report we used the frequency-domain method (13). The resolution of this technique was recently extended to picosecond processes by extending the frequency range to 2 GHz (14). This was possible by using the intrinsic high frequency harmonics of a picosecond laser source and a microchannel plate detector. Additionally, the sensitivity is high because the excitation source is not attenuated by a number of optical elements, which is necessary for intensity modulation of continuous light sources (13). The GHz instrument was shown to provide reliable measurements of 25 picosecond time delays (14) and has been used to measure correlation times as short as 8 picoseconds (15).
Using this instrument we have found that the emission of tryptophan in extensively purified samples of hemoglobin was largely dominated by lifetimes in the picosecond region which were sensitive to ligation. Longer lived components in the nanosecond range were also present, which accounted for less than 1% of the emitting tryptophans and were not sensitive to ligand binding. We suggest that only the lifetimes in the picosecond range belong to hemoglobin. The correction times detectable in the system indicate rapid mobility and extensive rotational freedom of the tryptophans.

MATERIALS AND METHODS
Human hemoglobin was prepared from washed red cells obtained from fresh blood samples donated by the local blood bank. They were hemolyzed in 0.005 M phosphate buffer at pH 7.0, and the stroma were eliminated by filtration through a 0.45 pm pellicon cassette. The protein was concentrated by ultrafiltration through a M. 10,000 pellicon cassette, dialyzed against water, recycled through a mixedbed resin cartridge for removing polyphosphates and other ions, and stored at -90 "C.
The extent of purification is critical to these experiments. Manipulation of hemoglobin solutions was avoided as much as possible. Pipetting the solutions from one container to another exposes the protein to surface denaturation, which even if minimal may influence the long-lived emission of the samples. For this reason we choose chromatographic procedures which automatically eliminated denatured products, producing samples at the proper concentration (a few milligrams per milliliters) for single dilutions with the necessary buffers. The solutions were stored in the cold and used within a few Oxyhemoglobin was purified by high pressure liquid chromatography through a DEAE-PW5 preparative column (Waters) using a gradient formed by 0.015 M Tris acetate at pH 8 and 0.015 M Tris acetate at pH 7.7 in 0.2 M sodium acetate. The procedure lasted less than 1 h at room temperature. Fig. 1 shows the chromatographic profile of an elution followed at 280 nm. We used the major peak near 35 min, which on rechromatography showed only one single peak in the elution profile.
Deoxygenation of hemoglobin was achieved by diluting the protein with nitrogen-bubbled buffer in a closed cuvette with no liquid-gas interface. Na+ dithionite was added with a syringe through a serum stopper, to a final concentration of 0.01 mg/ml. Spectrophotometric analyses indicated the complete deoxygenation of the samples.
Carbonmonoxyhemoglobin was obtained by flushing the air phase of a closed fluorimetric cuvette with CO and then gently inverting a few times. The complete saturation of the samples with carbon monoxide was monitored by spectrophotometry.
All measurements were conducted at a protein concentration near 0.2 mg/ml in 0.05 M phosphate buffer at pH 7.0 at 4 "C. Protein concentration was measured spectrophotometrically using E = 0.868 cm2 mg" for the carbonmonoxy derivative at 540 nm. Optical densities were measured using a Cary 14 spectrophotometer.
Steady state fluorescence measurements were performed with a SLM photon-counting spectrofluorometer. Lifetimes and correlation times were measured using a fluorometer operating between 4 and 2000 MHz (14). The modulated excitation was provided by the harmonic content of a laser pulse train with a repetition rate of 3.76 MHz and a pulse width of 5 picosecond, from a synchronously pumped and cavity-dumped rhodamine 6G dye laser. The dye laser was pumped with a mode-locked argon ion laser (Coherent, Innova 15). The dye laser output was frequency-doubled to 295 nm with an angletuned KDP crystal. The emitted signals were observed with a microchannel photomultiplier, and the cross-correlation detection was performed outside the PMT. The frequency-domain intensity data were fit to the time-resolved expression days.

I ( t ) = C CY, e-'"'
(1) where a, are the pre-exponential factors, 7 ; the decay times, and x CY; = 1.0. The frequency-domain anisotropy data were fit to where r; are the amplitudes of the rotational correlation times 0;. The parameters were recovered by nonlinear least squares using the theory and software described elsewhere (16,17). Data for lifetimes and correlation times were obtained from three different preparations of hemoglobin, with very similar results. The analyses here reported reflect the actual data of a single preparation of the protein.

RESULTS
The emission spectrum of Hb, excited at 295 nm, is shown in Fig. 2. This emission is similar to that of tryptophan, except for the sharp peak near 325 nm. This component is almost certainly due to Raman scatter off the water, which is expected to decrease the energy of the scattered light by about 3600 cm-l. Additionally, this component was found to be almost completely vertically polarized. It was necessary to eliminate this scattered light prior to the detector. Otherwise, the 0 picosecond time delay due to scattered light would corrupt the data and prevent reliable determination of the tryptophan decay times. Also shown in Fig. 2 are the emission spectra recorded through two emission filters, Corning 0-52 and 7-51. The 0-52 eliminates light below 340 nm but transmits longer wavelengths. Since some emission was seen near 450 nm (---), we also used the 7-51, which transmits a broad band centered a t 360 nm. In this way the Raman scattering was eliminated, along with residual emission above 400 nm. All subsequent measurements were performed using the two filters in the optical path and observing the emission a t the resultant maximum of 365 nm.
One possible source of non-Hb fluorescence is heme-free hemoglobin. Small percentages of apohemoglobin could make a major contribution to the total emission. Suppose a solution contains 1% apoHb and that the tryptophan emission is quenched 100-fold in Hb. Then, 50% of the total emission would be from the 1% contaminant. We attempted to detect the presence of apoHb by titrating the solutions with hemin chloride, while monitoring the emission a t 365 nm. If the emission was due to apoHb, then titration with hemin should result in ligation and quenching. Fig. 3 shows that the decreasing emission followed a simple exponential function, produced by the increasing optical density of the solutions. This argues against the presence of detectable amounts of apohemoglobin, at least in a form capable of binding hemin.
To further demonstrate that the emission is due to tryptophan, we examined the wavelength dependence of the intensity and the steady state anisotropy (Fig. 4). The weak signals prevented a full investigation of the excitation and polarization spectra of our samples. However, the emission intensity decreased as the excitation wavelength was increased from 290 and 300 nm, and the anisotropy increased (Fig. 4). These features are consistent with the spectral properties of tryptophan.
We also examined the emission in the presence of increasing concentration of acrylamide, which is an efficient quencher of tryptophan fluorescence (18,19). It is known that collisional quenching is proportional to the fluorescence lifetime. Hence, acrylamide should selectively quench the longlived emission. We reasoned that the measured intensity would then contain an increase proportion of the presumed intrinsic Hb fluorescence. If this emission was dominated by scattered light, (with a decay time of 0) then the anisotropy in the quenched samples may rise above 0.4 (20). However, the anisotropy rose gradually, and the values remained reasonable (Fig. 5). Additionally, acrylamide was not able to quench more than 50% of the emission, which is consistent   with about 30% of the emission being due to nanosecond timescale components (below). The downward curvature seen in the Stern-Volmer plot (Fig. 5) is consistent with the emission being the sum of components with a range of decay times.
The frequency-domain intensity decays of oxy-, deoxy-, and carbonmonoxy Hb are shown in Fig. 6. For comparison this figure also contains the calculated response for a single exponential 1-nanosecond decay time (-).
Clearly, the frequency response of Hb is complex and shows considerable dispersion as compared to a single exponential decay. In each case it was necessary to use three decay times to account for the data. The adequacy of the three decay time fits is seen from the good match between the data (0, 0, A) and the three-exponential simulations (lines in Fig. 6). The parameters describing the decays are summarized in Table I. The standard deviations of the parameters in Table I show the Deoxy Hb  In each state the Hb decays show components with decay times near 20 picosecond, 0.8 and 8 nanosecond. The picosecond components dominate the decays (a1 0.99) but contribute only 40-55% to the integrated intensities. As expected quenching with acrylamide did not affect the picosecond lifetime of oxyhemoglobin, but it did decrease the decay times and intensities of the longer components.

Frequency-domain Hb Fluorescence
It is of interest to estimate the uncertainties in the picosecond decay times. If the uncertainties are less than the difference between the various liganded states, then one can interpret the picosecond decay times with respect to the molecular properties of the ligation states. Estimation of uncertainties for the decay times is complex. The decay times are correlated with the other parameters describing the decay, while the usual assumption of nonlinear least squares fitting is the lack of correlation. To circumvent this problem, we examined the xR2 surface for each of the picosecond decay times (Fig. 7). These curves were constructed by fixing the picosecond decay times at the values on the x axis, while the other parameters were adjusted to minimize xR2. We believe this method accounts for all possible correlations between the parameters and hence provides an estimate of the maximum uncertainty of the decay times. A very conservative estimate of the uncertainties is given by the standard deviations around the minima of the curves, which are indicated by the horizontal lines in Fig. 7. By this criterion the COHb decay time is distinct from the other two decay times, and it is likely that the oxy and deoxy decay times are also distinct. The uncertainty in these components appears to cover a range of about 3 picosecond.
And finally, we examined simulated data to determine how variations in the short decay time affect the frequency response. The simulated curves were generated using the parameters (a; and 7 ; ) found for oxy Hb, except for the picosecond correlation time which was varied from 9 to 27 picosecond (Fig. 8). These values bracket the recovered value of 16 picosecond. The modulation was not sensitive to these short decay times (9-27 picosecond), at least to our current measurement limit of 2 GHz. However, the high frequency phase  Table 11. Other conditions are as in Fig. 6. Excitation at 300   angles (0.5-2 GHz) were clearly sensitive to T , , which illustrates our ability to recover these values with reasonable precision.
We also examined the frequency-domain anisotropy decays (Fig. 9). The lines show the best fits to the data obtained using two correlation times, and the parameters are summarized in Table 11. Fig. 10 is self-explanatory. The tryptophan residues appear to be remarkably mobile. Most of the anisotropy decayed by a 50-60-picosecond correlation time. This correlation time and the amplitude are characteristic of freely rotating indole in water at 20 "C (21). In the case of oxy Hb the sum of the anisotropy amplitudes is 0.38, which is larger than the value found for tryptophan or indole in the absence of rotational motion (22).
The high value of the initial anisotropy was not anticipated. The anisotropy analyses in Table  I1 assume that a single emitting species was the origin of the intensity and of the anisotropy decay. However, an impurity would be a distinct emitting species, and its intensity decay would be associated with its own anisotropy decay. The formalism for analyzing associated and nonassociated systems is distinct (23)(24)(25). As shown by Ludescher et al. (23), when associated behavior is analyzed with the equations for dissociated systems this is likely to produce high values of initial anisotropies.

DISCUSSION
The results presented above strongly suggest that the lifetimes and correlation times detectable in the hemoglobin system have their origin in tryptophan fluorescence. One may question whether the emission was due to the hemoglobin molecules or to impurities present in the samples.
The strong overlap between the tryptophan emission and Soret absorption of hemoglobin results in very efficient energy transfer from the tryptophan residues to the heme. Using the Forster equation (51, with the assumption that the transition moments of the donor and the acceptor are randomly oriented, one can calculate that transfer occurs with an efficiency of 50% a t a distance of 70 A. Hemoglobin contains three different types of tryptophan residues, one in the (Y and two in the p subunits. The distances between heme and the tryptophans vary from 13 and 17 A. The intersubunits distances are near 35 A. In view of the 6th power relationship between distance and transfer efficiency, the quenching of tryptophan is expected to be between 100-and 10,000-fold. The lifetime of unquenched tryptophan is near 2 nanosecond (25). Hence, the 10-20-picosecond components we detected imply a modest 100-200-fold quenching. Probably, these quantities represent an average including even shorter lifetimes.
The picosecond decay times were affected by changing the ligation state of the protein. The lifetimes varied from 9 picosecond in deoxyhemoglobin to 16 picosecond in oxyhemoglobin and to 26 picosecond in carbonmonoxyhernoglobin. The sharp curvature functions shown in Fig. 7 indicate very little, if at all, overlap between these values. It should be stressed that these values are very consistent with the values reported by Hochstrasser and Negus (11) for ferric and CO myoglobin, whose intensity decays displayed components of 14 and 26 picosecond, respectively (11). These values imply a different quenching of the tryptophans produced by the oxy-, carbonmonoxy-, or deoxyheme.
The origin of the difference can be either a modification of the overlap integral in the 300-450 nm region, a change of the distance, or a modification of the angular relationships between tryptophan and heme. With regard to the overlap integral, it should be noted that in the 300-400 nm region the absorption spectra of oxy-and carbonmonoxyhemoglobin are practically superimposable. The absorption of deoxyhemoglobin is slightly shifted so that below 350 nm is lower than that of the liganded derivatives, while above 350 nm is higher, therefore compensating the areas in the overlap integral. In the 400-450-nm region, carbonmonoxyhernoglobin has the highest values of extinction coefficient, while its picosecond lifetime is three times longer than that of deoxyhemoglobin. Thus, the origin of the different efficiency of transfer must be either in the distance or in the angular relationships between tryptophans and heme or both.
The distance is probably a minor factor. At high levels of quenching, the transfer efficiency is not sensitive to modifications of 1 A or less (26) and would not explain the 3-fold change of the picosecond lifetime in carbonmonoxy-and deoxyhemoglobin. A major role is most likely played by the angular relationships. As described by Eaton and Hofrichter (27), for reasons of symmetry, the transition moments associated with the absorption of the light in the heme are polarized either perpendicular or parallel to the plane of the tetrapyr role ring. The former are z-polarized in the direction of the 4-fold symmetry axis passing through the center of the iron atom perpendicular to the plane of the heme. The latter transitions are x,y-polarized on the plane of the heme and show equal absorption for all directions of the electric vector parallel to the heme plane. For such transitions the heme behaves like a planar, circular absorber.
In deoxy-and carbonmonoxyhernoglobin the transition moments associated with the absorption in the 300-400 nm region are x,y-polarized, therefore, a tilting of the plane of the heme can justify a different efficiency of energy transfer from tryptophans. Crystallographic studies have given clear indications that a tilting of the heme occurs upon ligand binding (26), thereby explaining the different picosecond lifetimes in these two derivatives.
Crystallographic studies suggest that oxyhemoglobin and carbonmonoxyhemoglobin are isomorphous, therefore, the heme has the same "tilt" in both derivatives. However, the transition moments associated with the absorption in the 300-400-nm region in oxyhemoglobin are partly z-polarized (27) perpendicular to the plane of the heme. The different orientation of the electric vectors in oxy-and carbonmonoxyhemoglobin may contribute to the different picosecond lifetimes in the two derivatives.
If our speculations are correct, then measurements of the quenched lifetimes may become a sensitive probe for exploring conformational changes in hemoproteins.
The nanosecond decay times found for the Hb samples were not sensitive to ligation. This suggests that they were produced by non-hemoglobin impurities. Consistent with these observations and with the high mobility of the tryptophan residues in hemoglobin, the simulations described by Henry and Hochstrasser (28), using metmyoglobin coordinates, failed to show tryptophan positions stable enough to prevent energy transfer to the heme.
The anisotropy decays of all three hemoglobins showed the presence of two correlation times, one in the picosecond and one in the nanosecond range. The one in the picosecond range was endowed with unusual characteristics. Correlation times of about 50 picosecond are somewhat shorter than the correlation time generally found for tryptophans in proteins and are more characteristic of free tryptophan in water or tryptophan residues in random coil peptides. This seems to indicate a lack of conformational constraints limiting the motion of the residues that is consistent with the large amplitude of the picosecond motion. However, this motional freedom exceeds that found for even small peptides. The apparent rapid depolarization may be a result of the short tryptophan decay times, which bias the observations toward short correlation times. In fact these correlation times are probably averages of complex librational motions, and the detectable distribution of these motions probably depends both on the structure of the surrounding protein and on the duration of the lifetimes which allow the observation. Therefore, our data may still be consistent with tryptophyl residues having motional freedom on the picosecond timescale but not necessarily faster or with greater amplitude than those present in other proteins (21, [29][30][31]. It should be stressed that for the first time protein anisotropy decays were time-resolved in such highly quenched samples. The measurements are probing new territories which may need additional considerations. Another anomaly of the system is the high initial anisotropy that is distinctly higher than the anisotropy of immobilized tryptophan (20). The simulation reports by Ludescher et al. (23) suggest that they are produced by the presence of associated correlation times in our samples. This is consistent with the hypothesis that the nanosecond lifetimes are produced by non-hemoglobin impurities. It should be stressed that associated behavior may also arise from intrinsic heterogeneity of emission of a single macromolecular specie. This situation may be present in hemoglobin, where three different tryptophans are present.
Hochstrasser and Negus (11) and Jane et al. (12) report somewhat low values for the initial anisotropy of sperm whale, horse, and Aplysia myoglobins. For Aplysia myoglobin, they explain the phenomenon by energy transfer between 2 tryptophyl residues very near to each other e ( 12). In hemoglobin the three tryptophans are a t least 25 A distant from each other. Therefore, energy transfer among them can be excluded as a significant factor.
The anisotropy analyses also indicated the presence of longer correlation times ranging from 4.5 to 19.5 ns. These values are uncertain because the amplitudes are small (Table  11) and because they were determined primarily by the nanosecond timescale emission. This emission was probably contaminated with the emission from non-Hb impurities. Nonetheless, these values are comparable to those obtained in this laboratory for extrinsically labeled Hb and its isolated subunits (32,33).
In summary, our data provide reliable measurements of the intrinsic tryptophan decay times in Hb. These values are sensitive to the ligation state and may provide a sensitive probe for the structure and dynamics of Hb.