Excitation transfer and charge separation kinetics in purple bacteria. (1) Picosecond fluorescence of chromatophores
from Rhodobacter capsulatus wild type
Marc G. Miiller a, Gerhart Drews b and Alfred R. Holzwarth a
a Max-Planck-lnstitut fiir Strahlenchemie, Miilheim a. d. Ruhr (Germany) and b Institut J~r Biologie H Mikrobiologie, Albert-Ludwigs-Universitiit, Freiburg (Germany)
(Received 3 September 1992)
Key words: Energy transfer kinetics; Bacterial photosynthesis; Purple bacterium; Picosecond fluorescence; Target (compartmental) analysis; Model calculation; Free energy; (Rb. capsulatus)
The fluorescence kinetics of chromatophores from Rhodobacter capsulatus wild type have been measured with picosecond time resolution over the wavelength range from 850 to 940 nm. The data have been analyzed both by global lifetime analysis yielding lifetimes and decay-associated spectra (DAS) and by global target analysis techniques yielding rate constants and species-associ- ated spectra (SAS). In the global lifetime analysis five lifetime components were necessary for a good fit. The lifetimes are: ~'~ = 9 ps showing a DAS with positive and negative amplitudes; ~'2 = 40 ps with a DAS maximum at --- 890 nm; ~3 = 95 ps with a DAS maximum at = 895 nm; z4 = 260 ps with a DAS similar to that of lifetime component ~'3; a fifth lifetime component of 940 ps has nearly negligible amplitude. In the global target analysis several kinetic models were tested. A homogeneous model with sequential energy transfer LHC II ,--, LHC I ,--, RC does formally fit the data only if the detailed charge separation and charge recombination processes at the reaction center (RC) are taken into account explicitly. However, despite the good fit, such a model must be excluded as a valid description both on the basis of simple thermodynamic considerations as well as due to the fact that it predicts unreasonable RC rate constants. Instead a heterogeneous model, assuming a mixture of chromatophores with open and closed RCs describes the situation both formally and physically quite well. For such a model the rate constants of the RC electron transfer processes and their free energy values as well as the energy transfer rate constants and SAS of the antenna pools were obtained. The extrapolated RC kinetics for open RCs agrees well with that known from isolated purple bacterial RCs. The rate constants for energy transfer processes among antenna pools and from antenna to RCs indicate that at room temperature the exciton kinetics of the entire antenna system is limited and determined by the RC charge separation, i.e., the exciton decay is trap limited. Our data are compared with the corresponding data for Rb. sphaeroides.
Introduction
Time- re so lved opt ica l spec t roscopy r ep resen t s the m e t h o d of choice for the s tudy of the ear ly events of energy t rans fe r and charge sepa ra t ion in photo- synthet ic systems. However , e luc ida t ion of the s truc- tural and funct ional o rgan iza t ion of the a n t e n n a sys-
Correspondence to: A.R. Holzwarth, Max-Planck-lnstitut fiir Strahlenchemie, W-4330 Miilheim a.d. Ruhr, Germany. Abbreviations: B800-850, light-harvesting complex II (LHC II); B875, light-harvesting complex I (LHC I); B896, minor long-wavelength antenna form of LHC I; SPT, single photon timing; Rb., Rhodobac- ter; DAS, decay-associated spectrum; SAS, species-associated spec- trum; RC, reaction center; PS II, Photosystem II; PSU, photo- synthetic unit.
t ems of pho tosyn the t i c bac te r i a by t ime-reso lved meth- ods in the p icosecond range has begun only very re- cent ly (for reviews see Refs. 1 and 2). Both f luores- cence kinet ics as well as t rans ien t absorp t ion tech- niques have been app l i ed dur ing the last few years to var ious pu rp l e as well as g reen bac te r ia l systems. The d e v e l o p m e n t in this f ield was largely d e t e r m i n e d by the technica l diff icult ies of measur ing u l t rafas t f luores- cence in the nea r - i n f r a r ed region on the one hand, and the d e v e l o p m e n t of low-intens i ty u l t ra fas t lasers to be used in t rans ien t absorp t ion m e a s u r e m e n t s on the o the r hand.
T h e aims of p icosecond s tudies on in tac t pho to - synthet ic systems are manifold : they consist , in ter alia, in the cha rac te r i za t ion of the energy t ransfe r pa thways be tween p igments , eventua l ly down to the s ingle- t rans-
50
fer step level, in the determination of the pigment arrangement and the role of spectral heterogeneity of antenna systems, and in the answer to questions relat- ing to the overall energy trapping kinetics and the efficiency of the absorbed light energy that is trans- ported to the reaction center(s) (RCs). Intimately re- lated to these questions are some fundamental prob- lems for the understanding of photosynthetic light- harvesting, e.g., the question whether special 'energetic funnel-type' pigment arrangements are necessary for efficient light-harvesting [2-5] and whether the overall energy trapping kinetics and yield are limited by diffu- sion of the excitations in the antenna (diffusion-limited kinetics) or by the rate of charge-separation at the RC (trap-limited kinetics) [6-10].
Purple bacteria, due to their generally well-resolved spectral properties of the antenna and their variability in the antenna composition are highly suited to study these problems. We report here on the fluorescence kinetics and kinetic modelling of the energy transfer in Rhodobacter capsulatus which belongs to a group of purple bacteria that contains two major types of an- tenna pigment pools called LHC 1 (B875) and LHC II (B800-850) with postulated minor antenna forms of B870 and B896 [11,12]. From that group Rb. sphaeroides has been studied most extensively so far (see Ref. 2 for a review) while time-resolved studies of Rb. capsulatus have not been reported to our knowledge. In view of the expected similarities in antenna structure between these organisms we shall thus compare our results with the reported kinetic data for Rb. sphaeroides [13-16].
The following kinetic scheme for energy transfer under room temperature conditions has been proposed for chromatophores with open RCs from Rb. sphaeroides based on picosecond transient absorption data [14]:
,; 2 ps 35-40 ps ~J ps [13800, ' B8501 ~ ' B 8 7 5 , ' R C
(B8%)
where the numbers indicate the equilibration a n d / o r trapping times between pigment pools and (B896) indi- cates the involvement of the minor long-wavelength form of LHC I in the energy transfer between B875 and RC. This scheme has been questioned both by Freiberg et al. [15] and Shimada et al. [16] at least with respect to the B850/B875 equilibration time. Shimada et al. [16] reported a 12-ps decay for the B850 excited state but rise terms of 20 ps and 35 ps at emission wavelengths of 875 nm and 909 nm, respectively, were reported as well. The latter authors suggested an even longer equilibration time of --- 50 ps between B896 and RC. Freiberg et al. [15] found two decay times of 10 and 50 ps for B850 fluorescence which they interpreted in terms of two pools of B850 which were coupled more (10 ps) and less (50 ps) efficiently to LHC 1.
Further discrepancies exist in the interpretations of these groups with respect to the trapping times on the RC.
Materials and Methods
Chromatophores have been isolated from Rb. capsu- latus wild type cells (strain B10) grown under low light conditions as described in Ref. 17. For measurement the chromatophores were suspended in 20 mM Tris buffer (pH 7.8) and the suspension was purged with nitrogen gas. The sample was pumped through a flow cell (1.5 × 1.5 mm) at a pump speed of --- 20 ml /min by a peristaltic pump. All measurements were carried out at ambient temperature (= 22°C, i.e., k BT= 25 meV or 205 cm-I) .
Fluorescence kinetics has been measured by the single photon timing (SPT) method as described [18] using a near-infrared sensitive microchannel plate photomultiplier as detector (R2809U-05, Hamamatsu) and near-infrared monochromator (DH 10.VIR, Jobin Yvon). The width of the apparatus response was --- 30-35 ps (fwhm) using excitation pulses of = 10 ps width (fwhm) at 798 nm [18]. The fluorescence kinetics was measured with a resolution of 5 ps/channel . Data were analyzed by two different methods. First, in the global lifetime analysis procedure, the fluorescence kinetics IF(t,Aem) are described as a sum of n expo- nentials
IF(t, Acre)= LAi(A¢~).exp(- t / ~i) i=1
with lifetimes 7 i. The plot of the amplitudes A i ( ,~em)
vs. emission wavelength h e m is called a decay-associ- ated spectrum (DAS). This type of analysis makes no assumptions about the kinetic scheme involved except that the same set of lifetimes is considered suitable for all emission wavelengths included in the analysis [19,20].
In the second type of analysis, the so-called global target analysis, specific kinetic models are directly tested on the set of experimental decays in order to check for their compatibility with the data [21,22]. Apart from avoiding many problems inherent in sums- of-exponentials fitting, this method also yields the real physical parameters of interest of the system, i.e., rate constants of energy transfer kij between pools (com- partments) of antenna pigments, charge separation rates etc., and the pure emission spectra of the kineti- cally resolved pigment pools, which are called the species-associated spectra (SAS) [21]. These proce- dures are described in further detail in Ref. 22. Quality of the fits was judged on the basis of residual plots as well as individual and global A, 2 values [18]. All spectra shown are corrected for the wavelength dependent sensitivity of the detector system.
Results
The chromatophores of Rb. capsulatus have been excited at 798 nm, a wavelength where the B800 pig- ments of the B800-850 (LHC II) complex absorb al- most exclusively along with a very minor (< 3%) ab- sorption by the monomeric BChls of the RC. The measurement conditions were chosen to keep the RCs in the active 'open' state (P860 reduced, quinone QA oxidized) as much as possible• The repetition rate of the laser was 4 MHz. All decays were recorded with a high S /N ratio of 40000 to 50000 counts in the peak channel.
Global lifetime analysis Fig. 1 compares residual plots for some selected
wavelengths from the global lifetime analysis by four and five exponential functions of the decay data over the emission wavelength range 850 nm to 940 nm. These plots clearly indicate that five lifetime compo- nents are necessary to describe the kinetics over this wavelength range, since in the four-component analysis large deviations in the residual plots around the maxi- mum of the decay are observed along with a drastic increase in g 2 values (Fig. 1A). For five exponentials the fits are excellent (Fig. 1B). In the four-exponential analysis lifetimes of = 10 ps, = 65 ps, --165 ps and 750 ps are obtained.
The lifetimes and DAS obtained from five-exponen- tial global analysis are shown in Fig. 2. The fastest
51
23 0 " ~ ~ ~ - - ' - -
L d
E /~ 937
262 u m ~ / < --5 94
Stotionory Spectr < ) ~ i 1 0
I I I I | I I I I I
8 4 0 8 6 0 8 8 0 9 0 0 9 2 0 9 4 0
W a v e l e n g t h , n m
Fig. 2. DAS for fluorescence decay o f chromatophores from Rb. capsulatu.~" at room temperature ()rex c = 798 nm) as calculated by
global l i fetime fi t t ing. The stationary fluorescence spectrum meas-
ured under the same conditions is shown as a dashed line.
lifetime component ( T 1 = 9 ps) shows positive ampli- tude in the DAS below 875 nm and a negative ampli- tude (rise-term) above that wavelength. The posi- tive/negative shape directly identifies this lifetime component as indicating an energy transfer process with an equilibration time of 9 ps. From the shape of this DAS one can tentatively assign it to the energy equilibration kinetics between the B850 pigments of LHC II and LHC I (B875). We have not attempted to
18
,3.2, x,m - 050 '
" " I g T ~ l l l i ' ! ~ ' l / r C / ~ l P l ~ ' l , ' l " ~ I _ 3 . 2 / ~ 1 4. exponential~ X = 1.261 I
3.8
0.0
- ,3.8
4.3
I X - 895 nm
,~ tll.l,..~a IAli l l i~i~i, , .~li . i .J, . a , '1 p'~'rW"r'rr'r"' , r ' l . ' " , l l ~ l " ~
4. exponentlal~ X" " 1.34.9
b 0.0
-¢ . ,3
,3.7
;k__ - 910 nm
" " ' " " i " , 4. exponentl~e X I - 1.394
0.0
-,3.7
• . 1 ~ , . , . - 9 4 0 r ~
111 ' , ~ e ~ " r " " ~ ' ~ l ~ t ~ r ', r ' 4, exponentl(l~ X' - 1,324 i 0.5 1.0 1.5 2.0
T i m e , ns
3 0
00 tU~lha~,t ,~=1 I~i,a - 3 . 0 l" P 1 ~ =xporNmU~= ~ m 1.0.31 I
Ih IJ ~- o o ~ h ~ ' ' ~ t . ~ w ' - ' ~ = " " r l ' l , r~ l 'W"l l"r~Ir l~ r ? l . " ."1 W l
- ,3.5 | / / I S exponential= ~' i 1•047 I
,3.2 = 940 nm
• rlr'll~l~'llrpv~'r'l~V" I " ' " ' , .r!lll1lV~V'r! _,3.21 ' 1 / , ~, .~,.,..~u(p x - ~ , ~ I
0.5 1.0 1.5 2.0
T i m e , ns
Fig. 1. Residual plots and X 2 values for some wavelengths from the global lifetime analysis of one data set containing 7 decays in the emission wavelength range of 850-940 nm for Rb. capsulatus chromatophores at room temperature. (A) Four-component and (B) five-component analysis.
52
k21 k32 k43 k44 1 , " 2 , " 3 , - " 4 •
k12 k23 k34 [B800-850] ° [Be75]* [P860]* [P+H-Q]
Scheme I. Sequential model.
PSUs with open RCs
k21 k32 1 , " 2 , " 3
kt 2 k23 [B800-850]* [B875/P860]" [P+H-Q]
k33
resolve the ultrafast B800 ~ B850 transfer kinetics which was reported to occur in a time < 2 ps, i.e., faster than the time-resolution of our measurements [23,24].
The lifetime components ~'2 --- 40 ps, "/'3 ~ 95 pS and ~'4 = 260 ps all have very similar DAS with all positive amplitudes only and maxima at 895 + 5 nm. The z 3 (95 ps) component has slightly smaller amplitude in its DAS below 890 nm than the DAS of the 7" 2 (40 ps) component. The 7"4-component (260 ps) has a small relative amplitude at all wavelengths (A4 < 10%) while the fifth lifetime component (940 ps) exhibits nearly negligible amplitude (A5 < 0.5%) but it cannot be ig- nored in the analysis for a good fit. Its maximum in the DAS occurs around 865 nm and we suppose that it arises from a very small amount of detached B800-850 complex, not capable of transferring energy to the RC via LHC I. We will thus ignore this component in the further discussion. The very similar shape of the DAS
PSUs with closed RCs
k21 k42 k4 4
1 , " 2 , " 4 •
k12 k24 [B800-850]* [B875/P860]* [P+H-Q-]
Scheme II. Parallel model. Schemes I and II. The bold numbers indicate pigment pools or reaction center states while the ko are the rate constants of energy or electron transfer from pool (species) j to i. The symbols in parentheses give the most likely identification of the respective pools or states. In Scheme I a uniform redox state of the RC is assumed prior to excitation. In Scheme II two different redox states of the RC prior to the excitation pulse (PSUs with open and with closed RCs)
but a uniform antenna structure is assumed.
of components r2-z 4 indicates that they should all be related to the energy trapping and charge separation kinetics in the RC, representing mostly fluorescence
TABLE I
Rate constants k~j from global target analysis for kinetic Schemes I (sequential model) and H (parallel or heterogenous model) and corresponding lifetimes
The times from global lifetime analysis are also given. All rate constants in units n s - l and lifetimes in ps. In general the errors in lifetimes and rate constants are less than 10% except those which were indicated specifically. The lifetimes for the kinetic models have been recalculated from the fitted set of rate constants.
a Lifetimes from global lifetime analysis for comparison. ~' This lifetime was considered as an extra exponential component in the target analysis due to a very small amount of 'impurity' which is not
incorporated into the kinetic model but is required for fitting. This component has always very small ( < 0.5%) amplitude. c The rate constants k24 and k44 are not determined very precisely in the model II2; i.e., they have a large error (see also model I13 which
results in a similar good fit without these components). In model I rate constant kz3 could be as high as 1 ns-1 within the error limits.
" " " - " 0 - Ass ignmen t : 10 / x \ . . J " ",, x 1
,ll "" ', A 2 ," ", [] 3
• / , . / ~ ~.w, • Stati . . . . y
" ~//J
E ",
o - o o o o o o o . . . . .
I I I I I I I I I I
840 860 880 900 920 940 Wavelength, nm
Fig. 3. SAS for the fluorescence kinetics of chromatophores from Rb. capsulatus (same data as used in Fig. 2) at room temperature as calculated for Scheme I by global target analysis• The corresponding rate constants are collected in Table I. The numbering of the SAS
corresponds with the one used in Scheme I.
from the LHC I excited state, along with some emis- sion from the RC and the presumed long-wavelength pigments.
Global target analysis A detailed understanding of the kinetic components
in these complex kinetic data can be achieved only by kinetic modelling. We ~ested various kinetic schemes by global target analysis for their compatibility with the data. The important common feature of all these mod- els consists in the explicit inclusion of the charge separation as well as the charge recombination pro- cesses within the RC in the kinetic scheme. Without this inclusion clearly no reasonable fitting to the data whatsoever could be achieved. All deactivation rates to the respective ground states (sums of radiative and nonradiative rate constants) were assumed to be be- tween 0.4 ns-1 and 1 ns-1 in the analyses. The choice of this value is not critical.
Sequential model. The simplest kinetic model for- mally fitting the data well contains two antenna pig- ment pools with sequential energy transfer kinetics and a homogeneous RC redox state (Scheme I, 'sequential model'). This scheme involves formally four states or species, three of which we expect to be capable of fluorescing. We thus expect three SAS. Fig. 3 shows the SAS calculated for this model while the corre- sponding optimal rate constants are given in Table I. The SAS component 1 peaks at -= 865 + 5 nm with a shoulder at = 895 nm and significant contributions out to 940 nm. The most reasonable assignment is B850 (see Scheme I) but smaller contributions from some longer-wavelength antenna cannot be excluded. SAS- components 2 and 3 result in basically identical spectra
53
in an unrestricted fit. This would be more or less expected for this model. SAS-component 4, represent- ing the RC primary radical pair state, gets negligible fluorescence amplitude, thus characterizing it as a non-fluorescent species (state) as expected. Despite the formally good fit of this model (X z value = 1.10, the residuals are essentially identical to those shown for the global lifetime fit, Fig. 1B) we do not consider it as a physically reasonable interpretation of the data, as will be discussed below in detail.
Parallel or heterogeneous model. A second kinetic model (shown in Scheme II, 'parallel model'), assumes some heterogeneity in the RC kinetics. This hetero- geneity could have various origins. In principle it could be due to a heterogeneity in the redox state of the RCs or to a heterogeneity in the size of the B875 pool. However, for purple bacteria the B875 /RC ratio is a fairly stable parameter that is quite independent from growth conditions. Thus the most likely reason for a kinetic heterogeneity would be different RC redox states due to a contribution from closed RCs, either in the state P+ or in the state PQA- These models con- tain only two fluorescing pools tentatively attributed to B850 (pool 1) and B875 (pool 2). We have calculated the rates for three options: One involves connectivity of the photosynthetic units (model II1, connected PSUs). A further option assumes a separate package model (model II2, isolated PSUs) [10,25,26]. Addition- ally we tested a parallel model in which the excitation energy in the second subset of a n t e n n a / R C complexes is irreversibly quenched at the RC possibly by an oxidized P860 with a high rate constant (model II3, isolated PSUs with quencher)• The fits are also very good in these cases (global X 2 value is ---1.13 for
~ . - " " - " ~ I , Ass ignmen t : 10 / x ~ . - - ~ , x 1
," ~ [] 3
• / / / ~ ~ • Stationary
e" 6 x , , " / × \ "~,
•~_4 E .<
2
0 n o [] 0 - ~ - - - ~ o °
I I I I I I I I I I
840 860 880 900 920 940 Wavelength, nm
Fig. 4. SAS for the fluorescence kinetics of chromatophores from Rb. capsulatus (same data as used in Fig. 2 and 3) at room tempera- ture as calculated for Scheme I1 by global target analysis• The corresponding rate constants are collected in Table I. The number-
ing of the SAS corresponds with the one used in Scheme If.
54
TABLE I1
Extrapolated rate constants k~:~ ' (ns - :) and lifetimes 1 / k'c~ t (ps) for RCs and AG calues (meV) for radical pairs in chromatophores AGRe and hypothetically Lsolated RCs AGian~ as calculated according to Eqns. 1 and 2 from the data o f chromatophores shown in Table I
For all parameters that depend on antenna size n the values for n = 25 and n = 30, respectively, are separated by a slash.
Scheme 113 (isolated units with quencher) RC open 25/30 432/518 2.3/1.9 - 89
- 1 8 3 / - 187 - 4 7 / - 5 1
- 1 7 5 / - 1 7 9
" Taken directly from Table I.
model 111 and 112 whereas it is slightly higher for model II3: A ,2 value = 1.14). The SAS for these mod- els arc given in Fig. 4 (the SAS are the same for all parallel models) and the corresponding rate constants in Table I (see also Scheme II). The SAS of pools 1 and 2 are very similar to those for Scheme 1 with maxima at = 865 nm and --- 895 nm, respectively. Pool 1 again shows a SAS with a shoulder at = 895 nm, while the SAS of pool 2 extends down to 850 nm. The two species 3 and 4 get negligible fluorescencc ampli- tude as expected for (non-fluorescent) radical pair states.
The fit quality for Schemes I and I1, as judged from the global X 2 values and the residual plots of the global target analysis, are quite similar and indicate good fits in both cases. A number of other schemes, inter alia not including details of charge separation and recombination in the RC processes, have been tested as well. But they are not described here in detail because they yield unsatisfactory fits with large X 2
values and strong deviations in the residuals. In Table I the optimal rate constants for Schemes I and II are collected, along with the lifetimes recalculated for these schemes.
Discussion
In view of the fact that at least two kinetic schemes, i.e., Scheme I and Scheme II, fit the data about equally well, it is clear that the DAS as calculated by the global lifetime analysis prove insufficient to arrive at a proper interpretation and assignment of the kinetic compo- nents. We thus discuss mainly the results from the global target fitting of the various kinetic models in detail. However, the close similarity of the lifetimes found in both types of global analysis indicates that the lifetime analysis resolved in fact all components that were resolved also in the target analysis. For this discussion the (extrapolated) rate constants for charge
separation int (k s ) and the free energy differences AF~"int ~"-" RP of the radical pairs in the hypothetically isolated RC are important. They have been calculated according to Ref. 27 on the basis of the following equations:
i . . . . k B T . i n ( k , ~ t / k c , ) JGR~, - (1)
with
k csint _- kcsY[Ni.exp( _ AE i / k a T ) (2) i
where k B is the Boltzmann constant, kc, the rate of charge recombination, kc~ the rate of charge separa- tion in the a n t e n n a / R C complex, N i the number of chromophores in pigment pool i and AE i the energy difference relative to the excited special pair. These equations have been derived under the assumption that the excitation energy is equilibrated between the antenna and the special pair before charge separation takes place. In this case the product of the 'internal' rate constant for charge separation k~ t and the nor- malized probability for the excitation energy in the antenna to be located on the special pair yields directly the observed rate constant for the charge separation k~ as determined by the target analysis. The influence of the energy differences A E i between the antenna excited states and the excited state of the special pair is taken into account for N >> 1 by the Boltzmann factor. The rate constants k~ t and the AG values are col- lected in Table II (see also below for a more detailed discussion of these parameters).
Sequential model. It has been pointed out above that the sequential model (Scheme l) assuming a uniform RC redox state fits the data very well. Nevertheless, this model leads to physically unreasonable results. The most likely assignment of the pigments for this model is shown in Scheme I. This leads, however, to the problem that the ratio k f , ,~ /k ,ack ~ in the B875
kf~ P860 equilibration is > 100. Such a value is k b ~ - w
entirely unreasonable, given the fact that there should exist about 25-30 B875 molecules per RC and taking into account their nearly isoenergetic nature as indi- cated also by the SAS of pools 2 and 3. In fact, one might rather expect a ratio keo~,/k backw of about l / 30 , i.e., three orders of magnitude smaller than that actu- ally suggested by Scheme I. Even taking into account a small number of B896 pigments, which have been suggested for Rb. sphaeroides [28], does not help re- solve the discrepancy but would simply increase the expected ratio to 2 /30 or 3 /30 at the most. Turning the assignment of B875 and P860 to pools 2 and 3, respectively, around, also does not solve the problem and in fact leads to a very bad fit. This is not unex- pected, since it is well-known that LHC I I is not located in the vicinity of the RCs and thus cannot transfer energy directly to it.
There exists, however, an additional and equally serious problem. The extrapolated rate constant of charge separation from the (hypothetically isolated) RC in Scheme I is only 10 ns-~, i.e., by a factor of 30 too small as compared to isolated RCs (for a review see Ref. 29). Again taking into account a small number of RC-connected B896 complexes over which the exci- tation might be delocalized would by far not solve that discrepancy, apart from creating the paradox that the existence of these B896 complexes would slow down the effective charge separation rate, rather than in- creasing it, in contrast t~ the originally suggested func- tion of B896 [28]. We do not believe that these prob- lems can be solved in a physically reasonable way within the framework of this model and suggest, that Scheme I must be rejected as a valid description of the excited state kinetics in these chromatophores under our measurement conditions. The situation may be different at low temperatures, however [14].
Parallel models. For Scheme I1 we have assumed that there exists some kind of heterogeneity in the rates of trapping a n d / o r charge recombination in the RC. Such a heterogeneity could be easily brought about by closing some RCs in the measuring beam despite our attempt to keep RCs open. This could involve either the oxidized RC state (P860 +) or the state with QA reduced, which are both believed to be less efficient excited state quenchers than the reduced and open P860Q A [2]. In fact both states might accu- mulate to some extent under our measuring conditions. The redox state of the RC is not expected to modify the energy transfer rates between antenna pigments. For this reason we chose the energy transfer rate constants between pools 1 and 2 to be identical in both RC states. We have carried out the analysis for con- nected PSUs (model II1) and isolated PSUs (II2 and II3). Any of these models result in an equilibration time of = 9 ps between the B850 and the [B875/P860]
55
pools. The forward/backward ratio of rate constants k21/kl2 implies an energy difference of 379 cm -1 (model II1), 295 cm-1 (model II2) and 244 cm- i (model I13) between the B850 and the B875 pigments, assum- ing the same number of pigments in both pools, which is quite reasonable. Except for model II3 the energy differences are in good agreement with the energy difference of 336 cm- l obtained from the difference in the absorption maxima of the two antenna pools. The remaining small differences can be explained by a slightly different number of pigments in the B875 pool as compared to the B850 pool. Thus, pool 2 in Scheme II might represent photosynthetic units with open RC surrounded by the LHC I pigments (B875) while state 3 represents the corresponding primary radical pair state P + H - Q A. Rate constant k33 represents the rate of electron transfer to the quinone QA. This rate constant is in good agreement with the corresponding values measured in isolated RCs of purple bacteria [29]. The target analysis shows that = 60% of the RCs were in the open state as determined by the fitted relative absorptivities for each branch in the parallel models II2 and 113.
From the ratio of rate c o n s t a n t s k32/k23 we can calculate the free energy difference AGRp = - k a T . ln(k32/k23) between the excited state and the primary radical pair P +H- which results in values of - 9 7 meV (model II1), - 8 3 meV (model II2) and - 8 9 meV (model II3). The rate constant k32 (Table I) is the effective rate constant of charge separation from the excited state of the co re -an tenna /RC complex. Using the relationships given above (Eqn. 1 and 2) we can calculate both the intrinsic rate constant k~ t for charge separation from the isolated RC as well as the corre- sponding AG value for the radical pair of the isolated RC int (AGRp). In order to do so we need the energy differences between antenna (B850 and B875) and RC excited states and the number of antenna molecules comprizing the core. The latter is generally believed to be = 25-30 for the LHC l-core of most purple bacte- ria [2]. For the excited state energies of antenna and RC we take the respective mirror-image points of absorption and fluorescence spectra. The difference in absorption maxima themselves in this case would not provide a good measure of the difference in excited state energy due to the larger Stokes shift of the RC fluorescence as compared to the antenna fluorescence. These mirror image points can be estimated in a good approximation as the center between the correspond- ing absorption and emission maxima (Aemm ~x = 865 nm
m a x _ for the B850 complex and h e m - - 885 nm for the B875 complex and Ac~ ~x = 900 nm for the isolated RC as shown for Rb. sphaeroides in Ref. 30). In this way we obtain a mirror-image wavelength of ~-857.5 nm for the B850 antenna and = 880 nm for the B875 antenna as well as for the RC, i.e., there is effectively zero
56
energy difference between the excited states of the B875 antenna and the RC. We thus get a AG~, value between - 168 and - 187 meV and 1/ki~ t between 1.7 and 2.3 ps (see Table II). Notably, the AG values are essentially identical with the corresponding values for PS II [27] and isolated Chloroflexus RCs [31] while the extrapolated charge separation time corresponds well with that found for isolated purple bacteria [18,32]. The formulae used for the calculation for kic~ t and A / ~ i n t "-'aP apply only if the thermodynamic equilibrium between the excited states of the core antenna pig- ments and the RC is established. The validity of this approach has been also theoretically demonstrated re- cently [5,33]. The fact that both values (rate constants of charge separation and AG value) are in agreement with the values known for isolated purple bacterial RCs [32,34] can be taken as direct proof that this equilibrium is indeed established before charge separa- tion takes place.
The agreement between extrapolated charge separa- tion time in the RC (1.7-2.3 ps) with that measured in isolated RCs also allows us to calculate an upper limit for the energy transfer time from the closest antenna pigment(s) to the RC. If that transfer time were slower than about 3 ps, energy transfer to the RC would become the rate limiting step for the overall decay of the B875 excited state. Since this is not the case, as can be judged from the correctly extrapolated charge sepa- ration time of 1.7-2.3 ps, we can conclude that the equilibration time for the excited states of B875 an- tenna and RC must be __< 3 ps. This conclusion differs from that of Sundstr6m et al. [14] that the energy transfer B880 to B896 takes up to 50 ps and that the latter pigment transfers its energy to the RC in Rho- dospirillum rubrum. This comparison is valid, because the results found here for the LHC I decay correspond quite well with those reported in Ref. 14 for Rhodospir- ilium rubrum. We conclude that the charge-separation process is the rate-limiting step for the overall antenna deactivation and not energy migration in any part of the antenna, i.e., the kinetics is clearly 'trap-limited' and not 'diffusion-limited' at room temperature with respect to the core a n t e n n a / R C complex. Since the equilibration of the excitations in the LHC I1 complex with those in the core complex takes place with a time constant of = 9 ps, i.e., much faster than charge sepa- ration, it follows, that at room temperature in fact the whole antenna system is equilibrated both within itself as well as with the RC on a time scale several times faster than charge separation occurs. Under these con- ditions emission from LHC I, the long-wavelength pig- ments and the RC can not be distinguished, neither spectrally nor kinetically in our measurements.
We can make similar considerations for the PSUs represented by the second part of Scheme II, i.e., those PSUs representing 'closed RCs'. Assuming that these
PSUs contain PQA we get the rate constants for charge separation and recombination as shown in Scheme II which results in A/'7"int values of about - 5 0 meV "~"-" RP
(model II1) and - 195 meV (model II2). For model II1 this implies that charge separation in closed centers would be less exergonic than in open ones, quite in agreement with PS II RCs [27]. The drop in the rate of charge separation (k32 vs. k4z) as predicted for closed RCs is much larger than expected if compared with data from isolated reduced RCs [35,36]. The reason is probably that the redox state of 'closed RCs' is not entirely PQA but contains also P+, which is not unex- pected. We alternatively tried to fit a model where all closed RCs were assumed in the state P +, which would mean direct antenna quenching with a rate = 40 ns- [2] and k24 = 0, i.e., no process equivalent to charge recombination to the excited state (model II3). How- ever, this model should be rejected because it is based on a quenching process which is much faster than quenching of the antenna excitation energy by open RCs. An increase in the rate constant of antenna deactivation upon oxidizing the RCs is not very likely [37]. On the other hand model II2 creates the para- doxon that the radical pair energy for closed RCs (PQA) occurs to be about 30 meV lower than for open RCs (Table II). A mixed model with two types of closed RCs, which might describe the situation better, is at this time too complex for the analysis. Despite the fact that for 'closed RCs' we do not so far get a completely satisfactory set of rate constants we con- sider Scheme Il l or a slightly extended version of it as the most reasonable interpretation of the observed kinetics.
Finally, we should like to discuss the shape of the SAS obtained for Scheme II (and in fact quite similar for Scheme I). Pool 1 should be identical to the initially excited pool, i.e., in our case the B800-850 complex, while pool 2 should correspond essentially to the core complex comprizing LHC I and the RC. As pointed out already, the SAS of pool 1 shows a shoulder at 895 nm, i.e., the emission maximum of B875, and extends well up to 940 nm. This probably indicates that a fraction (--- 10-20%) of B850 complexes transfer en- ergy to B875 complexes in a time even much faster than = 9 ps, which is the experimentally resolved equi- libration component. This is not very surprising. Our analysis considers the two main antenna complexes LHC I and LHC II as uniform 'compartments ' or 'pools' and describes the energy transfer with one rate constant each, for forward and backward transfer, and ignores all further details of their internal organization. This is clearly an oversimplification which at the pre- sent level of accuracy and time-resolution of the exper- iment is still useful, but must eventually be given up. Similar arguments can be used to explain the fact that the SAS of pool 2 has significant contribution down to
57
850 nm, a wavelength where we would expect only very little fluorescence from B875 a n d / o r the RC complex.
Comparison with Rb. sphaeroides kinetics. In view of the lack of relevant kinetic data for Rb. capsulatus we compare our results with those measured by several other groups for Rb. sphaeroides chromatophores a n d / o r cells [13-16]. In fact, it can be assumed that Rb. sphaeroides is very similar in all essential features of the antenna organization and excitation kinetics to Rb. capsulatus studied here. For example, Freiberg et al. [15] reported basically identical lifetimes for Rb. sphaeroides as we obtained in our (unsatisfactory) four-exponential fit, i.e., = 10 ps, 50 ps and 150 ps. Due to their lower S / N ratio they did not resolve the ---40- and = 95-ps components that were in fact re- solved in the work by Sundstr6m et al. [14]. All this suggests both very similar kinetics for Rb. sphaeroides and Rb. capsulatus on the one hand and perhaps very similar RC redox states in the measurements of Freiberg [15] and ours. A major difference is the signif- icantly higher LHC I I / L H C I ratio which in the Rb. sphaeroides samples was above 2 : 1 while it was about 1 : 1 in Rb. capsulatus measured here. However, inter- estingly this influences apparently the kinetics only slightly. This fact allows interesting conclusions with regard to the relative arrangement and attachment of the B850 and B875 complexes. Why, in spite of the very similar experimental kinetics, do we arrive at some different conclusions as compared to the other groups? The main reason for this discrepancy consists in the fact that Freiberg et al. [15], Sundstr6m et al. [14] and Shimada et al. [16] ignored the complexities of the RC processes, in particular the charge recombination, which reports back to the antenna up to B850 and influences its kinetics. Any model that ignores this feature does not even fit the data formally well in our case. Sundstr6m et al. determined a LHC II to LHC I transfer time of about 40 ps in Rb. sphaeroides, which is much slower than the one found by us. It is interest- ing that in recent low-temperature measurements (77 K) [37] they also found this transfer time to be 10 ps, i.e., identical to the time determined by us at room temperature. Although interpreted differently by the two groups, both Freiberg et al. [15] and Sundstr6m et al. [14] assigned the = 40-ps component to energy transfer from the B850 to the B875 complexes. While Sundstr6m et al. assigned the = 40-ps component to the B850--* B875 transfer directly, Freiberg et al. as- sumed a heterogeneous B850 pool with energy transfer times of -- 10 and --- 50 ps. Both of these interpreta- tions seem to be inconsistent with our data in Rb. capsulatus. It is important to note that a) we do not observe any energy transfer components longer than 10 ps over the entire spectrum from 850 nm to 940 nm while both Freiberg et ai. [15] and Shimada et al. [16] reported components with longer rise terms b) the
= 10-ps equilibration between B850 and B875 is en- tirely independent from the kinetic model chosen and c) the overall exciton kinetics is not diffusion-limited. The only agreement between our interpretation and that reported for Rb. sphaeroides consists in the main B850 ~ B875 transfer time of = 10 ps reported by Shimada et al. [16] and Freiberg et al. [15] for the fast B850 pool. Beyond that there exist fundamental differ- enccs. The interpretations of all other groups come to the conclusion that the kinetics is diffusion-limited [15,16] or at least that there exist some rate-limiting energy transfer steps in the antenna or from antenna to RC [14]. This is in contrast to our findings in Rb. capsulatus. We do not observe equilibration of several tens of ps between thc B850/B875/B896 and RC pigment pools. In particular, wc can also exclude a slow, i.e., much longer than about 3 ps energy transfer from B875 to hypothetical long-wavelength B896 pig- ments. Despite somc difficulties in the exact kinetic modelling that we have pointed out above, we are confident that at room temperaturc all lifetimes longer than = 15 ps are reflecting exclusively charge transfer processes in the RC and not energy transfer processes in the antenna nor antenna to RC transfer processes. The fact that all kinetic components are seen at all wavelengths (except for the B800 emission range) di- rectly proves the fast antenna equilibration processes. We also do not find any necessity to assume two different B850 pools as assumed by Freiberg et al. [15] for Rb. sphaeroides. We have tested such models and found that these assumptions seem unsuitable to rem- edy the still existing problems in the interpretation of the kinetics for closed RCs in Scheme II, let alone the physically unreasonable results that are obtained in Scheme I. Further studies of similar kind will have to show whether indeed there exist substantial differences in the kinetics and thus functional organisation of Rb. sphaeroides and Rb. capsulatus or not.
The fact that at least two entirely different kinetic schemes can formally describe the data very well, stresses the necessity of kinetic modelling for the un- derstanding of such complex kinetics. Only then the various possible models and interpretations can be tested for their physical relevance. It is clear, further- more, that the still simple scheme proposed here for the excitation kinetics will have to be extended eventu- ally. However, despite some unresolved difficulties we believe that most essential features of the system are described properly already at this level of approxima- tion.
Acknowledgements
Partial financial support for this work to A.R.H. and to G.D. from the Deutsche Forschungsgemeinschaft is acknowledged. We thank U. Pieper (MPI Miilheim) for able technical assistance.
58
References
1 Holzwarth, A.R. (1987) in The Light Reactions (Barber, J., ed.), Topics in Photosynthesis, Vol. 8, pp. 95-157, Elsevier, Amster- dam.
2 SundstrSm, V. and Van Grondelle, R. (1991) in The Chlorophylls (Scheer, H., ed.), pp. 1097-1124, CRC Press, Boca Raton.
3 Bergstr6m, tt., Van Grondelle, R. and SundstrSm, V. (1989) FEBS Lett. 250, 503-508.
4 Visscher, K.J., BergstrSm, H., Sundstr(im, V., Hunter, C.N. and Van Grondelle, R. (1989) Photosynth. Res. 22, 211-217.
5 Beauregard, M., Martin, I. and Holzwarth, A.R. (1991) Biochim. Biophys. Acta 1060, 271-283.
6 Robinson, G.W. (1967) Brookhaven Symp. Biol. 19, 16-48. 7 Holzwarth, A.R. (1987) in Progress in Photosynthesis Research
(Biggins, J., ed.), Vol 1., pp. 53-60, Nijhoff, Dordrecht. 8 Knox, R.S. (1986) in Encyclopedia of Plant Physiology: Photo-
synthesis 11I (Staehelin, L.A. and Arntzen, C.J., eds.), Vol. 19, pp. 286-298, Springer, Berlin.
9 Holzwarth, A.R. (1991) in The Chlorophylls (Scheer, H., ed.), pp. 1125-1151, CRC Press, Boca Raton.
10 Van Grondelle, R. (1985) Biochim. Biophys. Acta 811, 147-195. 11 Tadros, M.H., Frank, R., Dorge, B., Gad'on, N., Takemoto, J.Y.
and Drews, G. (1987) Biochemistry 26, 7680-7687. 12 Drews, G. (1988) in Photosynthetic Light-Harvesting Systems
(Scheer, H. and Schneider, S., eds.), pp. 233-246, de Gruyter, Berlin.
