Biophysical Journal 84:1161-1179 (2003)
© 2003 The Biophysical Society
Excitation Energy Transfer Dynamics and Excited-State Structure in Chlorosomes of Chlorobium phaeobacteroides
Jakub P
en
ík*,
,
Ying-Zhong Ma*,
,
Juan B. Arellano*,
,
Jan Hála and
Tomas Gillbro*
* Department of Chemistry, Biophysical Chemistry, Umeå University, S-901 87 Umeå, Sweden; Department of Chemical Physics and Optics, Faculty of Mathematics and Physics, Charles University, 121 16 Prague, Czech Republic; Department of Chemistry, University of California, Berkeley, California 94720-1460 USA; Instituto de Recursos Naturales y Agrobiologia (CSIC), 37008 Salamanca, Spain
Correspondence: Address reprint requests to Jakub Peník, Faculty of Mathematics and Physics, Dept. of Chemical Physics and Optics, Charles University, Ke Karlovu 3, 121 16 Prague 2, Czech Republic. Tel.: 420-2-2191 1627; Fax: 420-2-2191 1249; E-mail: jakub.psencik{at}mff.cuni.cz.
 |
ABSTRACT
|
|---|
The excited-state relaxation within bacteriochlorophyll (BChl) e and a in chlorosomes of Chlorobium phaeobacteroides has been studied by femtosecond transient absorption spectroscopy at room temperature. Singlet-singlet annihilation was observed to strongly influence both the isotropic and anisotropic decays. Pump intensities in the order of 1011 photons x pulse-1 x cm-2 were required to obtain annihilation-free conditions. The most important consequence of applied very low excitation doses is an observation of a subpicosecond process within the BChl e manifold (
200–500 fs), manifesting itself as a rise in the red part of the Qy absorption band of the BChl e aggregates. The subsequent decay of the kinetics measured in the BChl e region and the corresponding rise in the baseplate BChl a is not single-exponential, and at least two components are necessary to fit the data, corresponding to several BChl e
BChl a transfer steps. Under annihilation-free conditions, the anisotropic kinetics show a generally slow decay within the BChl e band (10–20 ps) whereas it decays more rapidly in the BChl a region (1 ps). Analysis of the experimental data gives a detailed picture of the overall time evolution of the energy relaxation and energy transfer processes within the chlorosome. The results are interpreted within an exciton model based on the proposed structure.
|
INTRODUCTION
|
|---|
Green photosynthetic bacteria involve two otherwise not closely related families, green filamentous (Chloroflexaceae) and green sulfur bacteria (Chlorobiaceae). The common property of both groups is the presence of similar extramembrane particles, called chlorosomes, attached to the inner side of cytoplasmic membrane (Blankenship et al., 1995
; Olson, 1998). Chlorosomes serve as the main light-harvesting antennae and contain a large amount of bacteriochlorophyll (BChl) c, d, or e (depending on species). In addition, a minor amount of BChl a (1% of total BChls) is present in the baseplate of the chlorosome envelope, which interfaces the entire chlorosome to the membrane. Chlorosomes also contain a substantial amount of carotenoids, whose exact location and function are still a matter of debate, but they are supposed to be in close contact with BChls (Psencik et al., 1994a; Frese et al., 1997; Arellano et al., 2000; Carbonera et al., 2001). A distinct feature of chlorosomes from other light-harvesting antennae is the formation of BChl c, d, and e aggregates, which are believed to form several rod-like elements, without considerable involvement of protein. Several models of the BChl organization in the rod have been proposed, among which the closely related structures reported by Holzwarth and Schaffner (1994), Steensgaard et al. (2000a), Prokhorenko et al. (2000), and van Rossum et al. (2001) are perhaps in best agreement with the available experimental data.
Chlorobium (Cb.) phaeobacteroides is a BChl e containing, strictly anaerobic green sulfur bacterium, which is able to survive at extremely low light conditions. It was even found in the Black Sea at a depth of 80–100 m, where the solar irradiance is lower than 0.003 µE x m-2 x s-1 (Overmann et al., 1992). The chlorosomes have to be thus very efficient in light harvesting. In the case of green sulfur bacteria, the photons absorbed by chlorosome pigments are transferred to BChl a in the baseplate only at low redox potentials, whereas in the presence of oxygen the excitation is quenched by quinones (Frigaard et al., 1997) and/or BChl radicals (van Noort et al., 1997) in a process that seems to protect reaction centers against reactive oxygen species. From the baseplate, the excitation is transferred to reaction centers in the cytoplasmic membrane, presumably via the so-called FMO protein complex.
So far there exist only few clear evidences about energy transfer within the BChl c, d, or e manifold. For an unique strain of Cb. limicola containing a mixture of BChl c and d, a 4-ps component corresponding to the energy transfer between these two types of pigments was resolved by single-photon-timing (SPT), however only at low temperature (Steensgaard et al., 2000b). Previously, a 5-ps transfer was observed by the same technique in the BChl a-free, BChl c containing chlorosomes of Chloroflexus (Cf.) aurantiacus, but it was not observed in untreated chlorosomes (Holzwarth et al., 1990). Both these reported transfer lifetimes are similar to the lifetime of the lowest exciton level of BChl c (
5–6 ps) as determined by spectral hole burning at 4 K (Psencik et al., 1994b, 1998). Using femtosecond transient absorption (TA) technique, the only direct evidence about energy transfer within the main chlorosome BChl was an observation of a 300-fs rise component for Cf. aurantiacus at low temperature (Savikhin et al., 1996a). Other evidences of exciton relaxation and energy transfer processes within the chlorosome aggregates are indirect, based on a wide variety of decay components (in the order of 0.1–100 ps) obtained for BChls c, d, and e by means of various time-resolved spectroscopy techniques (Blankenship et al., 1995 and references therein, Savikhin et al., 1995, 1996b, 1998; Mimuro et al., 1996; Ma et al., 1996; Psencik et al., 1998; van Walree et al., 1999; Prokhorenko et al., 2000).
The energy transfer process from the aggregated BChls to the baseplate is much better characterized. In earlier SPT and TA studies, it has been found that this transfer can be described by a single time constant, with a value depending on the species (Holzwarth et al., 1990; Miller et al., 1991; Causgrove et al., 1992), growth conditions of the bacteria (Ma et al., 1996; Fetisova et al., 1996) and redox states (Causgrove et al., 1990; Wang et al., 1990; van Noort et al., 1997). However, using a two-color femtosecond pump-probe technique with the chlorosomes of Cf. aurantiacus, Savikhin et al. (1996b) observed a biexponential rise with lifetimes of 2–3 and 8–11 ps in the kinetics probed in the baseplate BChl a, after the excitation of the aggregated BChl c. This biphasic energy transfer feature was explained as due to the both intra-BChl c and BChl cBChl a energy transfer steps based on the kinetic simulation (Savikhin et al., 1996b). Two discrete time constants were also found recently to be necessary in the description of the energy transfer in the isolated BChl c and e containing chlorosomes from Cb. tepidum and Cb. phaeobacteroides, respectively, using picosecond SPT technique (van Walree et al., 1999, Steensgaard et al., 2000b). This biphasic energy transfer was also found for Cb. phaeobacteroides in a comparative TA and SPT study (Psencik et al., 2002), and the nature of the two processes will be further addressed here. In particular, it is interesting to know whether these resolved transfer times are correlated with the presence of multiple, spectrally different pigment pools in the aggregated BChls and/or the baseplate, as proposed for some green bacteria (Blankenship et al., 1995 and references therein; Psencik et al., 1994b, 1998; Savikhin et al., 1995; Mimuro et al., 1996; Somsen et al., 1996; Steensgaard et al., 2002b; Melo et al., 2000; Carbonera et al., 2001) or only with the architecture of the chlorosome.
In this study, we applied two-color femtosecond TA technique to investigate systematically the excitation energy transfer within chlorosome BChls. Special care was dedicated to avoid the onset of exciton annihilation, which enabled us to reveal otherwise hidden energy transfer processes. The experimental data thus provide information about the excitation migration within and between the aggregates, and finally to BChl a in the baseplate. For the interpretation of the data we used a model based on the structure of Holzwarth and Schaffner (1994). We show that with minor changes, this model allows to better explain main features of the steady-state optical spectra and provide a good basis for interpretation of time-resolved data.
|
MATERIALS AND METHODS
|
|---|
Sample preparation
Growth of cells of Cb. phaeobacteroides strain CL1401 and isolation of chlorosomes were carried out as described previously (Arellano et al., 2000), and samples were kept at -20°C until use. All TA experiments were performed on isolated chlorosomes diluted with 50 mM Tris-HCl buffer (pH 8.0) containing 2 M NaSCN to an absorbance ranging from 0.25 to 3.5 per mm at the respective BChl e Qy maxima, depending on the pump-probe wavelengths. 1-mm glass cuvettes were used and the chlorosomes were incubated with 20–25 mM sodium dithionite for 2 h before the measurements to achieve anaerobic conditions. Absorption spectra were measured before and after the experiments to ensure that no degradation occurred during the data acquisition.
Femtosecond TA
Two-color femtosecond TA kinetics were measured using a setup based on a mode-locked Ti:sapphire femtosecond oscillator (Tsunami, Spectra Physics, Mountain View, CA) together with a Ti:sapphire regenerative amplifier (Spitfire, Positive Light, Los Gatos, CA) producing 100-fs pulses centered around 800 nm at a repetition rate of 5 kHz. The output of the amplifier was split and the major part was used for pump pulse generation by frequency conversion in an optical parametric amplifier (OPA 800, Spectra Physics) to desired excitation wavelengths (505, 685, 715, or 745 nm). 505-nm pulses were further compressed by two SF14 prisms. Pulses of 
685 nm were used without compression. A white-light continuum generated in a sapphire plate by the minor part of the amplifier output was used as a probe pulse. The relative polarization of the pump beam with respect to the probe beam was set to the desired angle (54.7°, 0°, or 90°) by a Berek compensator (New Focus, Santa Clara, CA). The probe wavelength was selected after passing a monochromator with a bandwidth of 2.6–5.7 nm. Absorption changes were detected by silicon photodiodes. Excitation intensity was between 2 x 1011 and 4 x 1014 photons x cm-2 x pulse-1, and the spot size of the pump beam at the sample place was 0.5–0.7 mm. Beam intensity profile was measured by scanning a sharp blade mounted on a micrometer screw through the sample position and detecting the laser intensity behind the blade with a silicon photodiode. Spot size was determined at the point where the intensity equals to 1/e2 times the peak intensity. The cross correlation between the pump and probe pulses was 150 fs and 200 fs (full-width at half maximum (FWHM)) for the pump wavelengths at 505 and 685 nm, respectively. Deconvolution fitting of the isotropic and anisotropic kinetics, as well as the global lifetime analysis, were calculated using the Spectra program (S. Savikhin Software, Ames, IA).
|
EXPERIMENTAL RESULTS
|
|---|
Absorption and emission spectra of the Cb. phaeobacteroides chlorosomes at room temperature are shown in Fig. 1. The main absorption bands in the near infrared are the Qy bands of BChl e and BChl a, at 714 and 790 nm, respectively. The fluorescence emission spectra show bands at 742 and 808 nm, due to BChl e and BChl a, respectively. TA kinetics were measured upon excitation at 505 nm, 685 nm, 715 nm, and 745 nm, among which the kinetics excited at 685 nm were studied in a greater detail. Probe wavelengths (det exc) were chosen to cover the major part of the BChl e and BChl a absorption and emission spectra (685–825 nm, Fig. 1). The kinetics are dominated by excited state absorption (ESA) at the high energy side of the BChl e Qy absorption band, whereas photobleaching and/or stimulated emission (PB/SE) prevail on longer wavelengths. The zero-crossing wavelength (isosbestic point), which separates these two regions, is located at 700 nm.
View larger version (24K):
[in this window]
[in a new window]
|
FIGURE 1 Steady-state absorption and fluorescence spectra of Cb. phaeobacteroides chlorosomes at room temperature. All pump and probe wavelengths used in this work are denoted in the bottom of the figure. Fluorescence spectrum was measured after incubation with 20 mM sodium dithionite for 2 h.
|
|
Annihilation in BChl e manifold
Both the isotropic and anisotropic decays are strongly affected by exciton annihilation, and pump intensities in the order of 1011 photons x pulse-1 x cm-2 were required to obtain kinetics with negligible annihilation effect. An exact intensity is dependent on the excitation wavelength. For excitation at 715 nm, a pump intensity of 3 x 1011 photons x pulse-1 x cm-2 was required to get annihilation-free kinetics, although it was sufficient to use 7 x 1011 photons x pulse-1 x cm-2 at 685 and 745 nm, regardless of sample optical density (OD). The effect of excitation intensity on isotropic decays of BChl e is illustrated in Fig. 2 a. Note that the decays were normalized to their maximal intensities, which makes the difference between decays measured at the two lowest pump intensities less pronounced due to their low signal-to-noise ratio. Further attenuation of the pump intensity by a factor of 2 led to a profile that is undistinguishable from the curve measured at 7 x 1011 photons x pulse-1 x cm-2 (not shown). Under the intensity of negligible annihilation, the amplitude of the TA signal was found to be typically less than 0.001 OD for a sample of 1 OD at the absorption maximum. Due to the extremely low pump intensities used to obtain negligible annihilation, the experimental profiles exhibit a relatively low signal-to-noise ratio. Therefore, we cannot exclude the possibility that some residual contribution of annihilation is hidden within the relatively large noise, however, the effect is expected to be rather small. In addition, although the overall shape of the experimental profiles measured at given pump-probe wavelength combination were always well reproducible, the fitting parameters obtained for independent measurements were determined to be of standard deviations ±2–15%. To minimize the uncertainty in the estimation of the fitting parameters, 3–8 kinetic traces for each probe wavelength were measured independently upon excitation at 685 nm, and the parameters obtained from the best fits to the most representative profiles are presented.
The amplitude of the TA signal is linearly proportional to the pump intensity as it is 
1 x 1013 photons x pulse-1 x cm-2. For a sample with 1 OD per mm at 714 nm, the corresponding magnitudes of the TA amplitude are 
OD 0.01, i.e., their upper limit is one order of magnitude larger than that associated with the annihilation-free signal. At even higher pump intensities, the annihilation led not only to acceleration of the isotropic decay but also to a nonmonotonic course of decay, similar to that observed for B800-850 antenna complexes of purple bacteria (Ma et al., 1997). Under the highest pump intensities used (4 x 1014 photons x pulse-1 x cm-2), the TA signal probed at certain wavelengths even exhibits positive OD values during first 20 ps (not shown) after the initial fast decay, and the amplitude of the transient signal decreases with the pump intensity instead.
Rise in BChl e
Besides the apparent deceleration of the decay, the most striking effect observed for isotropic decays within the BChl e spectral region with the decrease of pump intensity was the appearance of a subpicosecond rise component for many combinations of sufficiently separated pump-probe wavelengths. Such a feature has never been observed previously for chlorosomes at room temperature, presumably due to the presence of annihilation to certain extent or insufficient separation between the pump and probe wavelengths. Previously, a rise was observed only for the chlorosomes from Cf. aurantiacus at 19–100 K, but it disappeared at higher temperatures (Savikhin et al., 1996a). From our experiments performed at 77 K on the chlorosomes from Cb. phaeobacteroides, we know that the rise in the kinetics measured at low temperatures is more pronounced and can be observed at higher pump intensities than those needed at room temperature (manuscript in preparation). This rise indicates an energy transfer within BChl e and can be resolved as long as the excitation intensity is equal to or below 1012 - 1013 photons x pulse-1 x cm-2 (depending on pump-probe wavelengths), but the time constant associated with this rise is intensity dependent (inset of Fig. 2 a). This dependence is negligible only at the pump intensity of 1011 photons x pulse-1 x cm-2. In addition, the actual lifetime and amplitude of the rise depends on the pump-probe wavelength separation. Upon excitation at 685 nm, we could resolve the rise for det 725 nm, and with the red shift of the probe wavelength the lifetime increases from 200 to 500 fs (Table 1; Fig. 3). Excitation at 715 nm leads to relatively shorter rise lifetimes at respective probe wavelengths (Table 2; Fig. 4), and even more pronounced shortening was observed upon excitation at 745 nm. On the other hand, excitation in the Soret band leads to similar lifetimes as obtained for excitation at 685 nm and no rise could be resolved at det 715 nm.
View this table:
[in this window]
[in a new window]
|
TABLE 1 Fitting parameters for the two-color isotropic kinetics measured upon excitation at 685 nm under a pump intensity of 7 x 1011 photons x pulse-1 x cm-2
|
|
View larger version (37K):
[in this window]
[in a new window]
|
FIGURE 4 The first 4 ps of the isotropic kinetics (symbols) probed at 745 nm upon excitation at 505 nm (a), 685 nm (b), 715 nm (c), and 745 nm (d) together with their fits (lines). Insets show the same data on extended timescale. Pump intensity was 3 x 1011 photons x cm-2 x pulse-1 (a), 7 x 1011 photons x cm-2 x pulse-1 (b, d), and 3.5 x 1011 photons x cm-2 x pulse-1 (c). Sample OD was 1 at the BChl e Qy maximum.
|
|
Decay of BChl e
After the subpicosecond rise, the kinetics of BChl e probed at det 700 nm exhibit wavelength-dependent decays, characterized by gradual slowdown of the decay with increasing probe wavelength (Table 1; Fig. 3). In contrast to the initial rise, the change of pump wavelength has only moderate effect on the decay: the decay times resolved from the kinetics probed at a given wavelength appear often slower with shorter pump wavelength (Table 2; Fig. 4). Table 1 summarizes the fitting parameters of the kinetics measured upon excitation at 685 nm. In general, most of the kinetics can be fitted by a model function consisting of one rise and two decay components. Distinguishable improvement of the fits to the first 10 ps of the kinetics measured at det 725 nm, i.e., those close to the isosbestic point (700 nm) and exhibiting more complicated shape, could be achieved by adding a fourth component, but it only gives a minor improvement of 
2(<5%). For kinetics probed at det 735 nm, the addition of the fourth component leads to an indistinguishable 2(<1%). In addition, the difference between lifetimes obtained from the fitting using three and four components becomes smaller with the increase of the probe wavelength.
It is noteworthy that the three-component fits lead to similar lifetimes (17–20 and 70–100 ps) at all wavelength through the BChl e region, with an exception for the kinetics measured at 705 nm, which is close to the isosbestic point. The lifetimes associated with the main decay components determined using the four-component fits appear slower with the increase of the probe wavelength. As shown in Table 1, the isotropic kinetics detected in the range of 725 nm det 755 nm are characterized by somewhat similar time constants (for a comparison of decays measured at the blue and red edges of the BChl e band, see Fig. 5 a), in contrast to those probed around 765 and 775 nm. In the latter case, both BChl e and BChl a contribute and those decays depend dramatically on the probe wavelength.
Rise in BChl a
As observed for BChl e, the isotropic kinetics measured within BChl a spectral region are also strongly pump-intensity dependent (Fig. 2 b); however, the effect becomes apparent at somewhat higher pump intensity (1.5 x 1012 photons x pulse-1 x cm-2). A dominant positive spike was observed at higher intensities, which is significantly attenuated at lower pump intensities. Under the annihilation-free conditions and at wavelengths below 800 nm, this spike almost disappears and instead a minor rise component with a lifetime 1 ps appears (inset of Fig. 2 b). A similar lifetime can also fit the remaining spike seen at det > 800 nm (Table 1; Fig. 5 b). Except the initial part and the overall amplitude, the kinetics measured at all probe wavelengths within BChl a region (det > 775 nm) are very similar, showing negligible dependence on the detection wavelength (Fig. 5 b). The change of the pump wavelength has also only moderate effect on the kinetics, and this time shows an opposite trend in comparison to the decay BChl e: the resolved rise times at a given probe wavelength are faster with a shorter pump wavelength. The profiles are dominated by a slow rise, which requires two exponential components for satisfactory fits. Attempt to fit the kinetics with a single rise component leads to a lifetime of 70–80 ps, but the fits exhibit a remarkable deviation from experimental data. The faster rise component with an amplitude of 10–14% has a very similar lifetime (17–19 ps) as the faster component of the BChl e decay, obtained from the three-component fits. The slower and also the dominant rise component is characterized by a lifetime of 120–130 ps, which is considerably slower than the 70–100 ps decay component of BChl e. This slower rise is in fact rather close to the slowest lifetime resolved from the kinetics detected around the BChl e absorption maximum using four exponential components. The subsequent BChl a decay was characterized by a lifetime of 200–240 ps for kinetics excited at exc 685 nm.
Global lifetime analysis
The similarity between the lifetimes resolved using the three-component fitting within BChl e region, together with the similar lifetimes resolved from the kinetics probed within BChl a, justify the use of global lifetime analysis. This analysis enables the determination of the fitting parameters with a higher accuracy than from the single decay fit.
The global analysis was applied to the kinetics measured upon excitation at 685 nm. The first model function consists of three and four exponential components for the kinetics probed in the BChl e and a regions, respectively, in which one component (
1) and its amplitude were held as local free parameters. This enabled reasonably fitting of the initial part of the kinetics at all wavelengths, including the one at 705 nm, where 1 was obtained to be 0.74 ps. At other wavelengths, the fitting of the initial part gives a lifetime and an amplitude that are always in good agreement with the parameters obtained from the corresponding single decay fits (Table 1). The best fits and the smallest weighted 2 were obtained when different 3 lifetimes (see Table 1) are assumed for the kinetics measured in BChl e and BChl a spectral region. In fact, this difference is expected on closer inspection of the parameters determined from the corresponding single decay fits. The lifetimes obtained from the global analysis are 18.4 ps (2), 79.3 ps (3) for the kinetics probed within BChl e, and 18.4 ps (2), 123 ps (3) and 208 ps (4) for those measured in BChl a region. The corresponding amplitudes are plotted as decay associated spectra (DAS) in Fig. 6.
View larger version (29K):
[in this window]
[in a new window]
|
FIGURE 6 Decay associated spectra obtained from global lifetime analysis of the isotropic kinetics upon excitation at 685 nm. The steady-state absorption and fluorescence spectra are also shown.
|
|
For comparison, the global analysis was also performed with four components. Although the lifetime and amplitude associated with the fastest component were kept again as local parameters, we released the restrictions applied in the previous case and all three remaining components were considered as truly global ones within the whole spectral range. As a consequence, the time constants of the decay and rise terms in BChl e and BChl a, respectively, were forced to be identical. This analysis gives lifetimes of 20.7 ps (2), 102.3 ps (3), and 211 ps (4) for the global parameters. The corresponding DAS (not shown) was very similar to that obtained in the previous case, except that the DAS of 79.3 and 123 ps were replaced by the one with lifetime of the 102.3 ps, and the relative amplitudes of 20.7 and 102.3 ps components were almost the same (50%) in BChl e spectral region. In addition, the 212 ps component has a negligible amplitude for det < 765 nm. The weighted 2 (1.1) for this analysis was worse than that obtained in the previous case (1.05)
The DAS provide a straightforward visualization of the excitation relaxation processes within the chlorosome. The rise observed within first 1 ps in the BChl e region is connected with a relaxation from some high-energy excited states to the states from which eventual emission occurs. On closer inspection of Fig. 6, one will find that the DAS of the 0.2–1.1 ps component is nearly a perfect mirror image of the BChl e fluorescence spectrum, both having maxima around 745 nm. Energy transfer from BChl e to BChl a occurs in two main steps: first 15% of the energy is transferred in 18–21 ps, and the main portion of energy arrives with a substantial delay. The results obtained from the single decay fitting and from the first global analysis (Fig. 6) suggest that the time constants associated with the second step are different for the decay of BChl e and the rise of BChl a. If these two time constants are forced to be identical, a lifetime of 100 ps is obtained. The subsequent decay of BChl a occurred with a 200 ps lifetime.
Anisotropy decays
Additional information about the excitation relaxation processes within the chlorosome was obtained from the anisotropy decays. Similarly as for the isotropic decays, the anisotropic decays are also strongly pump-intensity dependent (Table 3; Fig. 7). Besides obviously faster decay, the increase of pump intensity further leads to a lower value of residual anisotropy. Under the annihilation-free condition, the anisotropy decays measured within the BChl e region are rather slow, and this decay can be fairly well fitted by a single exponential component with a lifetime between 10–20 ps (Fig. 8). The parameters characterizing the anisotropy decay are quite well reproducible considering the low pump intensities used. For instance, independent measurements for 685745 nm pump-probe combination give lifetimes between 13.5–17.5 ps, initial anisotropy from 0.26 to 0.30, and the residual anisotropy always 0.16. Addition of the second decay component does not lead to any improvement of the fit. The lifetime also depends on pump-probe wavelength separation in a similar way like the isotropic decays. However, the difference of r(0) - r(
) is similar for different pump-probe wavelength combinations at low pump intensities, and values between 0.12–0.15 were found (Fig. 8). For the decays measured in the red part of the BChl e absorption band (Figs. 7 and 8), the initial anisotropy values were found to be significantly less than 0.4, indicating nonzero angle between transition dipole moments at the pump and probe wavelengths. Only in these cases do the residual values differ considerably from steady-state fluorescence anisotropy, 0.24, determined previously (Arellano et al., 2000).
View this table:
[in this window]
[in a new window]
|
TABLE 3 Comparison of pump intensity dependence for the fitting parameters determined in the isotropic and anisotropy decays excited at 715 nm and probed at 745 nm
|
|
View larger version (33K):
[in this window]
[in a new window]
|
FIGURE 7 Pump-intensity dependence of the anisotropy decays probed at 745 nm upon excitation at 715 nm together with the corresponding single exponential fits. Insets show the first 5 ps of the same data. Intensities are in photons x cm-2 x pulse-1, and the sample OD 1 at the BChl e Qy maximum. The decay time, initial, and residual anisotropies are listed in Table 3.
|
|
View larger version (43K):
[in this window]
[in a new window]
|
FIGURE 8 Anisotropy decays obtained for different pump-probe wavelength combinations as indicated, together with the corresponding fits. Insets show the first 5 ps of the anisotropy decay. The decay time, initial, and residual anisotropies are also shown. Pump intensity was 1.4 x 1012 photons x cm-2 x pulse-1 (a, d), 7 x 1011 photons x cm-2 x pulse-1 (b), and 3.5 x 1011 photons x cm-2 x pulse-1 (c). Sample OD was 0.25 (a, c) and 1 (b, d).
|
|
For a given pump-probe wavelength combination, the lifetime of the anisotropy decay probed in the BChl e region is fairly close to the 10–20 ps decay component of the corresponding isotropic kinetics. This correspondence holds true even at higher pump intensities as long as anisotropy decays are fitted with a single exponential, although use of a second faster component gives somewhat better fit to the calculated anisotropy decays in this case (data not shown). Under even higher intensities, the anisotropy decay appears no longer monotonic, as observed also for the isotropic decays. For instance, under a pump intensity of 7 x 1013 photons x pulse-1 x cm-2, the anisotropy obtained at 745 nm upon excitation at 715 nm reaches a minimum of -0.6 in the first 10 ps and then rises again up to 0.05 (not shown).
Anisotropy decays in the BChl a region were measured at 795 nm upon excitation at 685 nm. Both the parallel and perpendicular polarized kinetics are of similar shape, and the amplitude of the perpendicular one is slightly larger than the parallel one (not shown). As a result, the calculated anisotropy decays are rather noisy, so their lifetimes and especially the initial anisotropies are determined with a relatively large uncertainty. Nevertheless, the decay is obviously much faster than the decay obtained for BChl e. Best fits were obtained with a lifetime of 0.9–1.4 ps and initial values of 0.2 ± 0.07, which is in the same range as the residual values found for BChl e, 0.15–0.25. The residual anisotropy of BChl a was (-0.09)–(-0.1), in good agreement with the steady-state values (Arellano et al., 2000).
|
MODELING OF OPTICAL SPECTRA
|
|---|
Our interpretation of the experimental data is based on the structural model for the bacteriochlorophyllide d aggregate, proposed according to the quantum and molecular modeling calculations (Holzwarth and Schaffner, 1994). Slightly different structural parameters were reported for BChl c (Prokhorenko et al., 2000). These models accommodate the bonds that are responsible for vibrations observed in Raman and infrared spectroscopy (Hildebrandt et al., 1991; Chiefari et al., 1995) and the intermolecular cross correlations observed in magic-angle spinning NMR spectroscopy (Balaban et al., 1995, Boender et al., 1995). The latter method was recently used to refine the model (van Rossum et al., 2001). Assembly of BChl molecules in these structures is also in agreement with optically detected magnetic resonance data (Psencik et al., 1997). Even if the model still cannot be regarded as definitive, it represents a solid base for testing the optical properties of aggregates. The aggregates are characterized by strong pigment-pigment interactions between the BChls, and, in addition, there is very little amount of protein in chlorosomes, which means a minimal disturbance from pigment-protein interaction. These features make the chlorosomes ideally suited for the application of exciton theory (Lin et al., 1991; Alden et al., 1992; Buck and Struve, 1996; Somsen et al., 1996; Arellano et al., 2000; Prokhorenko et al., 2000).
Optical spectra were calculated in the point-dipole approximation as described in the Appendix. If the macrocycles of the pigments are in or close to van der Waals contact, such as the BChls in the chlorosomes, the point-dipole approximation is no longer valid and more realistic description of charge density is necessary. However, it often turns out in practice that point dipoles work well even for closer approaches (Pearlstein, 1991). Although the use of extended transition-charge distributions certainly can affect the calculated values of exciton interaction energies, they have no effect on the forms of the expressions for spectra calculations (Pearlstein, 1991). In addition to chlorosomes (see above), point-dipole approximation was successfully used for numerous coupled pigment complexes of various photosynthetic organisms (for a recent review, see van Amerongen et al., 2000).
For comparison, absorption and circular dichroism (CD) spectra were calculated using the parameters given by Prokhorenko et al. (2000), and essentially identical results were obtained as those presented therein (not shown). The only difference is the opposite sign of the CD signal, indicating opposite chirality of the aggregate (mirror image). One common result of these comparative calculations is worth mentioning here: the two exciton states possessing almost all the oscillator strength are located within 100 cm-1 from the lowest exciton state for the two structures (without inclusion of the disorder) presented by Prokhorenko et al. (2000).
As there is no structural model available for BChl e aggregates, we use those parameters of the original model with small changes. Our major modification to the model is an alteration of the angle between the direction of the Qy transition dipole moment of BChl molecule and the z axis of the rod, denoted here by 
. From fluorescence anisotropy data measured for Cb. phaeobacteroides, it was calculated that = 19–23° depending on the extent of the initial delocalization (Arellano et al., 2000). These values are in overall good agreement with previous results of linear dichroism (LD) in gels and in electric fields, polarized fluorescence on random samples, and on ordered samples and ps pump-probe experiments, which give values between 15 and 25° for Cf. aurantiacus (van Amerongen et al., 1988; Griebenow et al., 1991; van Amerongen et al., 1991; Ma et al., 1996; Frese et al., 1997). Values out of this range reported include the measurements performed by van Dorrsen et al. (1986) and by Fetisova et al. (1986) on the chlorosomes from Cf. aurantiacus and Cb. limicola, respectively, giving an angle of 37° and 0°. The former is identical to the angle of 36.7° determined for the original model (Prokhorenko et al., 2000). However, it was shown by van Amerongen et al. (1988) that this angle is too large, probably due to an overestimation of the extent of ordering of the chlorosomes. It should be noted that the values obtained from LD reflect the (weighted) average of the transition dipole moment orientation of the states absorbing at the given wavelength with respect to the rod axis. In contrast, the values determined by means of fluorescence anisotropy are given by the average angle between transition dipole moments of the absorbing and emitting states, which for excitation near the absorption maximum and detection in the red part of the absorption band should be close to (see Discussion).
In this work, we intend to show that the optical spectra calculated using the value of 20° (calculated from the steady-state anisotropy data) show better agreement with some experimental data. The remaining structural parameters were recalculated in such a way that the distances given by hydrogen bond network are kept the same as those given in the original model (Holzwarth and Schaffner, 1994, Prokhorenko et al., 2000). Furthermore, the main features of the model were conserved, namely an organization of the BChl molecules into linear stacks parallel to the symmetry axis of the aggregate and additional helical hydrogen bond network (Fig. 9). The recalculation led to a smaller distance between the centers of the stacks (0.46 nm instead of 0.8 nm), a larger distance between the molecules within the stack (0.83 nm instead of 0.65 nm, measured in term of Mg-Mg distance), as well as a larger angle of hydrogen-to-keto oxygen hydrogen bond (170° instead of 139–153°) with respect to the original model (Holzwarth and Schaffner, 1994, Prokhorenko et al., 2000). As a result, the number of stacks per 360° increases from 36 to 62 for the rod with the same diameter, and the overall number of pigments per unit length of the rod increases 1.25 times compared to the original model. It should be noted that similar results to those presented here could be obtained with an alternative model with the same density of pigments as in the original model, providing the distance between the planes was increased and dielectric constant of 1.00 was used. However, neither quantum calculations and molecular modeling (which have not been done for BChl e so far) nor experimental data supports such a model.
View larger version (24K):
[in this window]
[in a new window]
|
FIGURE 9 Model of a point-dipole distribution used for the spectra calculations. Panorama view of the 62 x 15 aggregate (a) and a detailed view of the 5 x 5 aggregate (b). Position of the dipoles are denoted by dots and their orientation by a line ended with symbol +. The length of the line roughly corresponds to the diameter of the porphyrin ring. Some planes of BChl molecules are schematically drawn, assuming for simplicity that the Qy transition dipole moment is parallel to the y axis of the porphyrin ring. All distances are in nanometers, structural parameters in Table 4.
|
|
The structural parameters used in this work are slightly different from those used by Arellano et al. (2000), where exciton calculation was applied to simulate the effect of the angle change on absorption and CD spectra. The angle of hydrogen-to-keto oxygen hydrogen bond was treated as an additional variable, whereas it was kept 180° previously. In addition, we selected differently the 15 x 15 aggregate for our calculation. We believe that these changes provide better correspondence to the original model. However, because these changes give rise to only very moderate changes in calculated spectra, they do not affect the conclusions presented in our previous work (Arellano et al., 2000).
Fig. 10 a shows the stick spectra calculated for 15 x 15 and 62 x 20 nondisordered aggregates. The latter represents a close tubular structure. A comparison between the calculation results shows that for the 15 x 15 aggregate the basic spectral features are already well established. Fig. 10 a also shows the envelope of the stick spectrum convoluted with a Gaussian function, and for illustration only, one CD spectrum is shown. The calculated absorption maximum is 380–400 cm-1 blue shifted from the respective lowest exciton level located at 735–740 nm for both 15 x 15 and 62 x 20 aggregates. This shift is more than four times larger compared to that calculated using the parameters of the original model (60–90 cm-1), either for the 18 x 8 or the 36 x 8 aggregate (Prokhorenko et al., 2000). However, as the lower levels possess almost no oscillator strength, the calculated fluorescence spectrum exhibits only a slightly enhanced Stokes shift compared to the original model (not shown). In addition, the higher-lying exciton states located above 650 nm also carry negligible oscillator strengths.
View larger version (42K):
[in this window]
[in a new window]
|
FIGURE 10 (a) Calculated absorption stick spectra for perfect aggregates consisting of 15 x 15 and 62 x 20 molecules, together with the sum of their Gaussian envelopes. Each of the envelopes has a FWHM of 350 cm-1 (Prokhorenko et al., 2000). The inset shows the calculated CD spectrum for the 15 x 15 aggregate. (b) Gaussian envelope of absorption and fluorescence spectra for the 15 x 15 aggregate obtained by averaging 5000 different realizations of the diagonal disorder, which is assumed to have normal distribution with standard deviation of 415 cm-1. For comparison, the absorption and fluorescence stick spectra obtained for one particular realization of the disorder are also shown. The fluorescence components were calculated for the temperature of 300 K and were normalized to the maximal absorption component. Absorbance of monomer corresponds to 23 units (BChl e dipole strength in Debye2) in the arbitrary scale used and is indicated by an arrow. The dielectric constant used both in (a) and (b) is 1.05, and all other parameters are given in the Appendix. The dotted lines are experimental curves measured at room temperature.
|
|
Due to expected random fluctuations of the molecular energies, the aggregate cannot be regarded as perfect and an effect of energetic disorder has to be included to get realistic spectra. To fit the experimental absorption and fluorescence spectra, a normal distribution of site energies (i in Eq. 3) with standard deviation of 400 cm-1 was used. Inclusion of the diagonal disorder leads to broadening of the spectrum and, more importantly, to redistribution of the oscillator strength from the highly allowed transitions mainly to those in the red part of the absorption spectrum (Fig. 10 b). Nevertheless, the lowest states still posses small oscillator strength. Fig. 10 b also shows the Gaussian envelope of absorption spectrum averaged over 5000 different realizations of the disorder (standard deviation of 415 cm-1) for the 15 x 15 aggregate. Among the overall 225 exciton states for each realization of the disorder, there are typically 100 states having negligible dipole strength (less than 0.1 of the monomer value), another 80 states have their dipole strength between 0.1 and 1 of the monomer value, 40 states fit between 1–10 of the monomer value, and around five states posses a dipole strength that is more than 10 times higher than that of the monomer. The distribution of the 5000 lowest exciton levels is characterized by a distribution function with a maximum at 765 nm and FWHM of 220 cm-1 (not shown). Individual lowest exciton levels are characterized by a transition dipole strength ranging from <1% to 300% (mean value 90%) of the monomer BChl e dipole strength. The width of the distribution function of the lowest exciton levels fits well with that determined by hole burning in fluorescence spectra of Cb. tepidum (Psencik et al., 1998), where it was shown that this distribution function corresponds to the overall emission spectrum of aggregated BChl at 4.2 K (FWHM of 250 cm-1).
The calculated spectrum retains well the main features of the experimental spectrum: 1), a large red shift of the absorption maximum (715 nm) compared to that of the monomer (660 nm); 2), asymmetry of the band with an enhanced absorption on the blue side together with the shoulder at 675 nm; and 3), transitions to the higher exciton states possess the dominant oscillator strength as predicted by hole-burning data (Fetisova and Mauring, 1992, Psencik et al., 1998 and references therein). Moreover, the same parameters used to obtain the absorption spectrum that is in the best agreement with the experimental data also give rise to the best results for calculated room temperature fluorescence spectrum in terms of its maximum (745 nm), Stokes shift, and the width of the band (Fig. 10 b). This calculation shows that use of an angle = 20°, determined experimentally, is capable of reproducing the absorption and CD spectra and, in contrast to the original model, the unusually large red shift of the fluorescence emission spectra. Further, it is unnecessary to take into account other interactions, such as the coupling between the aggregates (rods), to get realistic spectra. Therefore this calculation may also be applied to simulate the spectra of (individual) aggregates formed spontaneously in solutions.
With the parameters used by Prokhorenko et al. (2000), the strongest coupling is between the monomer 1 and 2 within the same stack (Table 4; Fig. 9 b). In our modified model, the situation is more complicated, as the modification of changes the mutual orientation of the interacting dipoles, which in turn affects the exciton couplings. The strongest interactions are between dipoles 1 and 3 as well as 1 and 4; both of them have similar energy but opposite signs. Nevertheless, the negative coupling energies, which arise from the head-to-tail interaction between the dipoles (typical for J-aggregates) play a dominant role in the aggregate, similarly as in the original model. This is clearly demonstrated by the pronounced red shift (more than 1000 cm-1) of the calculated absorption maximum with respect to that of monomeric BChl e, in agreement with experimental data. Moreover, the interaction in parallel fashion (typical for H-aggregates) is enhanced in our model. That results in a substantial energy difference between transitions with the largest oscillator strength and the lowest-lying exciton levels. This difference leads to a >500 cm-1 Stokes shift, which is also observed experimentally, owing to Boltzmann population distribution between these states.
Recently a refined model including a bilayer structure for the aggregates of Cb. tepidum was published (van Rossum et al., 2001); however, exciton calculation based on this model has not been reported yet.
As noticed, the choice of smaller (20°), compared to 36.7° used in the original model, is based merely on results of various experimental techniques utilizing polarized light. However, we believe that the good agreement between the model calculations and experimental data justifies the use of this modified model as a working hypothesis, which should be either proved or rejected by further work. The question remaining to be justified is if the proposed changes are consistent with the NMR data (Rossum et al., 2001 and the references therein). Also we believe that it is important to consider the possible distortion of the structure by other molecules, such as carotenoids (Ma et al., 1996; Arellano et al., 2000), which has not been well defined yet. It should be emphasized, however, that the main conclusions of this work are also valid within the original model, unless mentioned specifically in the text.
|
DISCUSSION
|
|---|
The densely packed pigments in chlorosomes and the rapid energy transfer between them lead to onset of exciton annihilation at remarkably low light level. However, this process is unlikely to operate at physiological conditions because of the low light available. Considering the excitation intensities used in this work to obtain annihilation-free kinetics (2–7 x 1011 photons x pulse-1 x cm-2), one will find that they are strikingly different from the light-harvesting complexes of purple bacteria, where annihilation is usually observable at an intensity that is 2 orders of magnitude higher (Ma et al., 1997). Actually, the applied pump intensities correspond to only a few photons per chlorosome, each consisting presumably of 105 molecules. To reveal the actual energy transfer dynamics in the chlorosomes, it is crucial to perform all the measurements under very low pump intensities, and to extract kinetic information from annihilation-free kinetics. Due to this consideration, we will not discuss the kinetics dominated by ESA (det 700 nm) in the rest of the paper, for which it was difficult to eliminate the annihilation effect.
Early energy transfer processes within BChl e
Under the low excitation intensities, we observed for the first time a subpicosecond relaxation process within the chlorosome BChl aggregates at room temperature, which is manifested by a rise in the kinetics probed in the red part of the Qy absorption band of the BChl e. The rise component has a decay counterpart in the kinetics probed at wavelengths shorter than 725 nm. However, at least part of this decay might be an artifact inasmuch as the probe wavelength is close to the isosbestic point at 700 nm. The lifetime (0.2–0.5 ps) and amplitude associated with the rise were found to increase with the spectral separation between pump and probe wavelengths within the BChl e Qy band. When excited at 505 nm, two additional processes contribute to the rise observed in the BChl e Qy band apart from the relaxation within the Qy manifold itself: a <100 fs energy transfer from the carotenoid S2 state (Psencik et al., 2002) and the internal conversion from the Soret-to-Qy band. The latter is also expected to occur with a <100 fs lifetime (Psencik et al., 2002). However, the rise observed upon 505 nm excitation is characterized by a smaller amplitude, and the rise time at 745 nm is only by 80 fs slower compared to those resolved upon excitation at 685 nm. These results can be explained by a nonselective population of all the states within the Qy manifold through the internal conversion from the Soret-to-Qy band. Subsequent relaxation within the Qy band can be described as a weighted average of relaxation from all the states located between the blue edge of the band and 745 nm. The idea is supported by the fact that excitation at 715 nm, where the density of states is the highest, leads to a significantly faster rise with a smaller amplitude at 745 nm compared to the excitation at 685 nm (Table 2).
By comparing the DAS of the rise component with steady-state absorption and emission of BChl e (Fig. 6), it can be clearly seen that the rise is due to relaxation and/or energy transfer from the higher energy excited states to the lower energy states, which give rise to fluorescence. The process is rather fast: for instance, the relaxation from the states absorbing at 685 nm to those at 745 nm with an energy difference of 1200 cm-1 occurs with a lifetime less than 0.5 ps. It suggests that the rise may be due to an exciton relaxation rather than stepwise Förster hopping of localized excitations, which would involve many transfer steps owing to the huge number of molecules in the chlorosome. This attribution is supported by following considerations: 1), Strong exciton coupling between the BChl e molecules leads to energy level splitting (Fig. 10 for our model), and the relaxation between the exciton levels is induced by electron-phonon coupling. In the presence of disorder, the relaxation is connected with a spatial redistribution of the energy as the excitation is not perfectly delocalized. 2), The rise was observed for certain pump–probe wavelength combinations, such as 715 nm745 nm but not for 685 nm715 nm, although their energy difference is similar. However, only in the former case the low-energy and presumably longer-lived exciton states are probed. 3), Low temperature hole burning showed that zero phonon holes can be burned only in the red edge of the aggregated BChl Qy absorption band. This is explained by a longer lifetime of the lower exciton levels, where the energy is accumulated after fast exciton relaxation and/or energy transfer (Fetisova and Mauring, 1992; Psencik et al., 1998 and references therein). Also it is unlikely that the process is due to a vibrational relaxation, inasmuch as Qy(0,0) band was most probably both excited and probed in the experiments with excitation at 685 nm.
Another important observation at low pump intensity is the minor depolarization within the first 1 ps in the BChl e region and no need for a subpicosecond component to fit the data. Instead, the anisotropy decay is remarkably slow (10–20 ps) compared to previously reported data (300 fs–1.5 ps, Savikhin et al., 1994; 1.7–3.7 ps, Savikhin et al.; 1995; 4–7 ps, Lin et al., 1991). This is in contrast to the situation in the exciton coupled B850 ring of purple bacteria, where the depolarization is completed within few hundreds of femtoseconds (e.g., Bradforth et al., 1995; Ma et al., 1997, 1998). In the B850, both the transition dipoles of individual molecules and the main exciton transitions are oriented nearly parallel to the ring plane. Calculations based on our model aggregate show that for both perfect and disordered aggregates, most of the exciton transitions carrying significant oscillator strength orient in close to parallel with respect to each other and to the symmetry axis z, provided that a fragment representing closed tubular structure is considered. Relaxation between these states, which are responsible for the major part of the TA signal in an isotropic decay, would not induce a significant anisotropy change. This explains well the presence of the fast decay or rise only in the isotropic kinetics, but not in the corresponding anisotropic profiles.
From the above discussion, it follows that within our model it is impossible to attribute the anisotropy decay to the exciton relaxation between the main absorbing exciton levels belonging to closed tubular structure. To explain the experimentally observed anisotropy decays, we need to consider the states with different orientations. Our calculations show that in the presence of disorder there are many states possessing small but not negligible oscillator strengths (comparable to that of the monomer) and having a localized character. In addition, the transition moments associated with these states orient often in such a way that are close to the transition moments of the individual molecules, at which position the amplitude of the corresponding exciton state wave function is dominant. The portion of the states with small oscillator strengths increases toward the edges of the absorption band. Relaxation and/or energy transfer between these states will lead to the observed anisotropy decay, even at 715 nm, where the main transitions are presumably parallel to the chlorosome axis. Such a behavior seems to be in agreement with the results of calculations for the exciton coupled B850 ring, where the presence of energetic disorder brings the most localized character to those exciton states at the edges of the absorption band, whereas the extent of delocalization is maximal in the middle of the band (Pullerits, 2000).
Experimental support for the prevailing localization of the long-wavelength absorbing BChl e states is given by the initial anisotropies. The initial anisotropy was found to be significantly lower than 0.4, when probe wavelength is tuned into the SE band, e.g., to 745 nm (Fig. 7 and Fig. 8 b). For the 715745 nm pump-probe combination, the initial anisotropy was found to be 0.32. The pump wavelength corresponds to the absorption maximum in this case, where all the main absorbing transition dipoles were calculated to be nearly parallel to the rod axis even in a presence of a diagonal disor
TOP
ABSTRACT
INTRODUCTION
) and 755 nm (
, divided by 3.57 for the sake of comparison) and in the BChl a region (b) at 775 nm (
), and 810 nm (