Independent initiation of primary electron transfer in the two branches of the photosystem I reaction center

Abstract

Photosystem I (PSI) is a large pigment-protein complex that unites a reaction center (RC) at the core with ∼100 core antenna chlorophylls surrounding it. The RC is composed of two cofactor branches related by a pseudo-C2 symmetry axis. The ultimate electron donor, P₇₀₀ (a pair of chlorophylls), and the tertiary acceptor, F_X (a Fe₄S₄ cluster), are both located on this axis, while each of the two branches is made up of a pair of chlorophylls (ec2 and ec3) and a phylloquinone (PhQ). Based on the observed biphasic reduction of F_X, it has been suggested that both branches in PSI are competent for electron transfer (ET), but the nature and rate of the initial electron transfer steps have not been established. We report an ultrafast transient absorption study of Chlamydomonas reinhardtii mutants in which specific amino acids donating H-bonds to the 13¹-keto oxygen of either ec3_A (PsaA-Tyr696) or ec3_B (PsaB-Tyr676) are converted to Phe, thus breaking the H-bond to a specific ec3 cofactor. We find that the rate of primary charge separation (CS) is lowered in both mutants, providing direct evidence that the primary ET event can be initiated independently in each branch. Furthermore, the data provide further support for the previously published model in which the initial CS event occurs within an ec2/ec3 pair, generating a primary ec2⁺ec3^- radical pair, followed by rapid reduction by P₇₀₀ in the second ET step. A unique kinetic modeling approach allows estimation of the individual ET rates within the two cofactor branches.

In oxygenic photosynthesis, the primary reactions of light utilization are driven by two multisubunit, pigment-protein complexes—photosystem II (PSII) and photosystem I (PSI) (1–3). The structures of PSI from the cyanobacterium Thermosynechococcus elongatus (4) and from the plant Pisum sativum (5) have been resolved to 2.5 Å and 3.4 Å, respectively. The PSI core complex consists of an extensive antenna (ANT) system of ∼100 densely packed Chls and a relatively isolated group of redox (reduction-oxidation reaction) active cofactors at the center composing the reaction center (RC). As in all other RCs, the cofactors in the RC of PSI form two quasi-symmetric branches (Fig. 1), diverging from a Chl a^′/Chl a pair (ec1_A/ec1_B) traditionally called P₇₀₀ (4). In each branch is a pair of Chl a molecules (ec2_A/ec3_A or ec2_B/ec3_B) and a phylloquinone (PhQ_A or PhQ_B). Finally, the branches join again at the F_X iron-sulfur (FeS) cluster.

Fig. 1.

Organization of the ET cofactors in the RC of PSI, based on the X-ray crystal structure of cyanobacterial PSI [1JB0, (4)], and using the nomenclature suggested by Redding and van der Est (43). (Figure was generated using UCSF Chimera.)

The symmetry of the cofactor branches is key to discussion of the directionality of electron transfer (ET) in PSI. Originally, it had been assumed that, analogous to the type II RCs, ET in PSI proceeded along only one of the cofactor branches, and that the other branch was not used. This idea would have been supported by the structural asymmetry of P₇₀₀, if P₇₀₀ were the primary electron donor in PSI. However, evidence was provided that mutations near P₇₀₀ did not affect the primary charge separation (CS) rate, but rather the second ET step, thus excluding P₇₀₀ as the primary donor (6).

The first indications of bidirectional ET in PSI can be found in the work of Sétif and Brettel on PSI particles from spinach (7), where biexponential decay kinetics of PhQ^- were observed (with t_1/2 = 25 ns and 150 ns). Biphasic PhQ^- reoxidation was later seen in cyanobacterial PSI (8), yet the original explanation did not invoke bibranched CS. Later, Joliot and Joliot (9), however, observed two phases of PhQ^- oxidation with t_1/2 of ∼18 ns and ∼160 ns and similar amplitudes in Chlorella sorokiniana cells, and suggested that they be assigned to reoxidation of the two PhQs within PSI. This hypothesis was tested directly by examination of the kinetics of PhQ^- reoxidation in Chlamydomonas reinhardtii mutants in which the Trp residues in π-stacking contact with the PhQs had been converted to Phe either singly or together (10). These results allowed for the assignment of the two phases to particular cofactor branches: the fast phase (∼20 ns) to reoxidation of PhQ_B and the slow phase (∼200 ns) to reoxidation of PhQ_A. Similar results were obtained in optical studies performed with cyanobacterial mutants (11, 12).

Initially, there seemed to be little support from EPR experiments for participation of the two cofactor branches in CS (13–15). The time resolution of this method precludes direct detection of the fast phase of PhQ^- reoxidation observed by optical spectroscopy (11, 13). However, the fact that the spin density of is localized on the ec1_B Chl (16) opens the door for distinguishing the and radical pairs (RP) by transient (laser flash-induced) EPR, due to the different relative orientations of the and PhQ^- radicals in the two radical pairs. This, indeed, has been observed in PSI prereduced with dithionite (to reduce the FeS clusters), using PSI RCs purified from perdeuterated Synechococcus lividus (17), mutants of C. reinhardtii that affect the axial ligands of the ec3 Chls (18), or in a mutant of Synechocystis sp. PCC6803 targeting the PhQ_A site (19).

All of the above-mentioned works describe only the final stage of the ET along the cofactor branches: the ET step from PhQ to F_X. However, the decision about the branch along which the electron will travel is already taken in the primary CS step. In order to resolve the branching problem and understand the underlying mechanisms, ultrafast time-resolved studies are required. In the past, such studies on PSI were mainly concentrated on solving another major issue—the type of the energy trapping kinetics (cf. reviews in refs. 3, 20, and 21). The major question was, is charge separation in PSI limited by the photochemical trap or by transfer of excitation energy to the trap? Most of these studies either did not attempt to describe the ET reactions in detail or used a single-branched and, in most cases, irreversible CS model (22–26). Many of these studies also assumed a priori that excitation energy transfer (EET) from the ANT to the RC of PSI is slower than the primary CS reaction, and little attempt had been made to resolve the RC* population kinetics. Thus, the actual rate of primary CS could not be determined (24, 25). If EET to the RC indeed represented the bottleneck of the overall trapping reaction, it would be extremely difficult to unravel the following faster ET steps. Our recent studies have demonstrated, however, that the RC* population kinetics and the rate of primary CS can be resolved and that trapping is limited by the ET process rather than by the EET to the RC [(6, 27, 28), cf. refs. 3 and 27 for extensive discussion of these issues]. A key point in the decision for (ET) trap-limited vs. EET-limited kinetics is the fact that the trap-limited model is the only one that simultaneously satisfies both the transient absorption (TA) as well as the fluorescence kinetic data (29). We note, however, that the structure-based modeling simulations that concluded that trapping is either intermediate (neither trap- nor diffusion-limited) (23) or diffusion-limited (26) did not take charge recombination of the initial radical pair into account (27, 29). Ignoring the charge recombination reactions will have a pronounced influence on the calculated EET rates (27, 29). Furthermore, the fluorescence kinetics cannot be described consistently if the charge recombination is not included (29). Another important point is that the source of material matters—the core antenna of C. reinhardtii PSI, unlike that of cyanobacteria and higher plants, lacks “red” Chls (27, 29), which simplifies the trapping kinetics substantially.

In this work, we aim to differentiate the early ET steps in the two cofactor branches of the PSI RC, in order to get insight into the origin of the biexponential kinetics previously observed at the PhQ/F_X level. Because of the chemical identity of the cofactors on the two branches, differentiation of them requires mutations directed to specific cofactors. Savikhin and coworkers studied ET kinetics in mutant cyanobacterial PSI RCs where the methionines serving as axial ligand to the ec3 Chls had been modified in either of the branches (30) and concluded that ET in cyanobacteria is strongly asymmetric, using predominantly the A-branch. Similar mutations to these Met residues were made in C. reinhardtii, and ultrafast data from them were used to support bidirectional ET with a branching ratio near 1∶1 (31). A different approach was taken by Li et. al. (32) who studied C. reinhardtii PSI in which the Tyr residues involved in H-bond formation to the ec3 Chls had been converted to Phe, with the predicted effect of destabilizing the ec3^- Chl anion radical. Mutation of the A-side Tyr resulted in a decrease in the amplitude of the slower kinetic component (assigned to ET from PhQ_A to F_X), which normally represents ∼60% of the total absorbance change at 380 nm in the wild-type (WT), and an increase in the amplitude of the faster kinetic component (assigned to ET from PhQ_B to F_X). The opposite result was seen in the B-side Tyr mutant. These data not only provided strong support for bidirectionality of ET in PSI, but also they suggested that there was some sort of competition between the branches in terms of CS, and that the energetics of a state involving ec3^- could affect this competition.

There is now general agreement that both cofactor branches in PSI are active for CS, and that the A-branch is the dominant one, with the ratio of A-side/B-side ET varying in a species-dependent manner from ∼3∶2 (green algae) to ∼3–4∶1 (cyanobacteria). What is lacking is a mechanistic model of CS within PSI that would naturally explain the observed bidirectionality, tracing it back to its origins in the initial ET steps. In the current study, we used these same C. reinhardtii PSI mutants, PsaA-Y696F and PsaB-Y676F, to probe the effects of loss of the H-bond to specific ec3 cofactors on the earliest ET reactions in PSI. It has previously been shown that the primary CS reaction in PSI is not affected by mutations to P₇₀₀ (6), strongly suggesting that the initial CS event is between ec2 and ec3. This study again tests that hypothesis directly, as a clear prediction of it would be a significant effect on the CS rate in the branch in which the H-bond donor to ec3 had been removed.

Results

Figure 2 displays the TA difference spectra at selected delay times after preferential PSI RC excitation at 700 nm. (See Fig. S1 in SI Text for the complete hypersurfaces.) The early blue shift of the initial bleaching band (690–695 nm) indicates that RC/ANT EET is finished for the most part within ∼2 ps. Hence, the difference spectra at later times should reflect the overall trapping of excitation and the formation/decay of various RP states. The fast energy equilibration kinetics occurring on the time scale of 1–2 ps are remarkably similar in the WT and both mutants. Differences are observed only in the shape of the later spectra beyond ∼10 ps. The kinetics in the spectral range above 720 nm, where there is virtually no ground state absorption, is crucial for a qualitative and quantitative understanding of the reaction steps. This range provides a window to selectively monitor the formation of the early RPs (33). For example, the appearance of the primary RP(s) from the excited state can be directly observed from the decay of the stimulated emission (SE) and the concomitant rise of the absorption of the RPs at 730 nm (27) (Fig. S2). The loss of SE (negative signal) and rise of RP absorption (positive) are nearly a factor of two faster in the WT (∼10 ps) than in the mutants (∼20 ps), and are somewhat slower in PsaA-Y696F than in PsaB-Y676F. This effect of the mutation on the primary CS process in both mutants, which can be seen without any further analysis or modeling, not only demonstrates that both cofactor branches contribute to the primary CS reaction with similar yields, but also indicates that it is the first CS step that is slowed down by either mutation. This key observation logically excludes all single-branched ET models and demands a detailed examination of bibranched models.

Fig. 2.

Transient absorption spectra at different delay times after excitation at 700 nm for the WT (A), PsaA-Y696F (B) and PsaB-Y676F (C) PSI particles.

Lifetime Density Analysis.

The datasets from both time ranges were submitted to a combined lifetime density (LFD) analysis. The LFD maps (Fig. 3) reveal all of the relevant lifetime components. The correlated bright yellow-green and dark blue nodes (660–700 nm) represent the ∼800-fs EET process. The most important feature, describing excited state trapping and the rise of the primary RP(s) can be observed above 720 nm (blue features in LFD maps and in Fig. S2). This takes place with lifetimes of 7–20 ps in the mutants and 5–9 ps in the WT (cf. Fig. 3), indicating that primary CS in both mutants is slowed down substantially compared to the WT (Fig. S2). The broad dark blue feature (675–700 nm) with lifetimes of 15–50 ps contains a complex mixture of RP kinetics that can be disentangled only by detailed kinetic modeling. This signal includes the loss of excited state absorption below 670 nm (lifetimes from ∼20–40 ps—bright yellow-green node). At the upper end of the LFD maps there is a long-lived nondecaying (ND) component that directly reflects the final difference spectrum.

Fig. 3.

Lifetime density maps for WT, PsaA-Y696F, and PsaB-Y676F mutants. The ordinate is a logarithmic lifetime axis and ranges from 100 fs to above 1 ns. On the abscissa are plotted the spectra calculated for each corresponding lifetime, using a color code for amplitude at each wavelength: orange corresponds to zero level, while negative amplitudes are blue to black, and positive amplitudes go from green to yellow to white. Thus, decay of a bleaching or rise of a new absorbance (e.g. at 8–20 ps above 710 nm) will appear as a dark feature, while a component with positive amplitude, indicating a loss in ESA or a rise in a bleaching during an EET will appear as a bright feature (e.g. the node at 0.8–1 ps between 660–680 nm due to uphill EET).

Kinetic Modeling.

The next step in the data analysis was to build a physically meaningful kinetic model describing the underlying processes in detail (34), based on the above-mentioned qualitative conclusions demanding a bibranched model. Again, one excited ANT compartment (ANT*) and one excited RC compartment (RC*) were used (Fig. 4). Using extensive analysis of simulated datasets, it turned out that by applying a minimal set of relatively unrestrictive boundary conditions (see SI Text for details), it was possible to solve the fitting problem for the bibranched model and to determine all of the rate constants (Fig. 4), the resulting species-associated absorption difference spectra (SADS), and transient populations (Fig. 5 and Table S1).

Fig. 4.

General kinetic scheme of the bibranched kinetic model. The rate constants are given in units of ns^-1 for the indicated ET steps in WT (black), PsaA-Y696F (red), and PsaB-Y676F (blue) PSI. (See SI Text for the lifetimes and the weighted eigenvector matrices.)

Fig. 5.

SADS (A, B, C) and transient populations (D, E, F) of the intermediates resulting from the bibranched model shown in Fig. 4. The WT data represent a reanalysis of our previously published TA data (27). Note that the WT and the mutants have different antenna sizes, thus explaining the different signal strengths between antenna and RC (see footnote in the Discussion).

Discussion

The Initial CS Event.

The fact that the PsaA-Y696F and PsaB-Y676F mutations preferentially affect primary CS indicates that the ec3 Chls are part of the RP formed during this process. We had earlier found that three mutations near P₇₀₀ (including an axial ligand and H-bond donor) had an effect upon the second ET step (RP1 → RP2), but not on primary CS (6). This indicated that P₇₀₀ was not part of RP1. Taking together both the negative evidence (lack of effect of P₇₀₀ mutations upon CS) and the positive evidence (observed effect of ec3 mutations upon CS), we can now unequivocally assign RP1 as ec2⁺ec3^- and state that primary CS consists of RC^∗ → ec2⁺ec3^-. The secondary step would then consist of ET from P₇₀₀ to rereduce ec2⁺ (). Furthermore, the fact that both Tyr mutations had quantitatively similar effects upon the primary CS rate demonstrates that RP1 can be formed on either branch, generating either or . Moreover, the effects of these mutations upon primary CS are mirrored by their effects upon directionality, as manifested by the amplitudes of the nanosecond components assigned to ET from PhQ_A and PhQ_B. It had earlier been shown that loss of the H-bond donor to P₇₀₀ had no effect upon directionality (35) and was later shown to have no effect upon primary CS (6). The ec3 mutations do affect directionality (32) and now have been shown to inhibit primary CS. Thus, the directionality phenomenon observed in the nanosecond time scale is merely a manifestation of the primary CS mechanism.

Unibranched Versus Bibranched Kinetic Models.

Even with our previous analysis of the data from WT PSI, there were indications that the unibranched models were inadequate. The fitting of such a model to the mutant data is shown in Fig. S3. The average effective primary CS rates calculated from a unibranched model are 350 ns^-1 (WT), 210 ns^-1 (PsaA-Y696F), and 290 ns^-1 (PsaB-Y676F). However, we had pointed out earlier that the ET kinetics in the early steps in WT PSI appeared to be more complex than predicted from the unibranched model (6, 27), showing either distinct splitting of lifetime components or an unusual broadening in the LFD maps. The present mutant data go much further to corroborate the previous findings. For example, the broad distribution of lifetimes in the range of 10–40 ps for the mutant data (Fig. 3) cannot be modeled properly with three lifetimes only. The present qualitative and quantitative results demonstrate that loss of the H-bond to either of the ec3 Chls selectively slows down substantially the first ET step without substantially affecting the other rates (cf. Fig. S3), definitively excluding a unibranched ET model. Moreover, the fact that the two mutations have a quantitatively similar effect argues strongly that both ec3 Chls are involved in the CS event. It would be logically incoherent to claim that one mutation had a direct effect upon the cofactor in the active branch, while the other mutation (to a cofactor in the inactive branch) somehow had an indirect, yet quantitatively similar, effect upon the active branch. Thus, based on these mutant data, an extension of the ET chain model to a two-branched model was mandated even before analyzing the details of the kinetics.

The modeling must answer the following questions: (i) to what extent do the two branches differ in their ET rates; (ii) are there significant differences in the spectral properties of the two cofactor branches; and (iii) which factors finally determine the branching ratio in terms of yields of the two ? While the ET lifetimes from either PhQ to F_X is well beyond our experimental time scale, and thus cannot be resolved, the spectrum and yield of each RP are contained in the long-lived ND component (Figs. 4 and 5, cf. Table S1 for lifetimes and weighted eigenvectors).

Following excitation, the energy equilibrates between the RC and the ANT compartment with τ₁ ∼ 0.8–1 ps (Table S1). As expected, EET rates are virtually unaffected by the mutations. In WT PSI, the primary CS reaction proceeds with an effective rate constant of 220 ns^-1 in the A-branch and 140 ns^-1 in the B-branch.^† The charge recombination is about twice slower in the A-branch (∼8 ns^-1) than in the B-branch (∼16 ns^-1).

Comparison of the primary CS between the WT and the mutants reveals that the PsaA-Y696F and PsaB-Y676F mutations affect specifically the CS rate related to the branch in which they are located. The PsaA-Y696F mutation slows down the CS rate in the A-branch by a factor of about three to ∼80 ns^-1 (Fig. 4). However, no significant impact induced by this mutation is observed on B-side ET processes. Examination of the results from the PsaB-Y696F mutant indicates a substantial drop (by a factor of ∼2) of the primary CS rate in the B-branch. Again, the other rates are unaffected, within the error limits. These numbers also indicate that the effect of the mutations on the redox potential of the respective Chls is relatively mild, as expected, since neither mutation completely inhibits primary CS within the affected pair. The rates for the secondary ET step are in the range of 60–70 ns^-1 for the A-branch and 80–90 ns^-1 for the B-branch. Interestingly, although unaffected by mutation, the tertiary ET rates towards the PhQs differ significantly between the two branches, with ∼32 ns^-1 for the A-branch and ∼14–19 ns^-1 for the B-branch.

Assignment of the Radical Pairs and Their Spectra.

As stated above, the first RP state on either branch is ec2⁺ec3^-. Hence, on the A-branch this RP is , while it is on the B-branch. The SADS of both RP3 states (Fig. 5) resemble the difference spectrum found in transient spectroscopy studies in the nanosecond time scale, allowing us to assign it with confidence to . Finally, the intermediate species (RP2) must be attributed to , in agreement with the conclusions drawn earlier based on single-branch ET models (6).

The SADS of ANT* and RC* are typical of Chl excited singlet states, showing a single bleaching band at about 683 nm and 690 nm, respectively, and SE up to 750 nm. In contrast, the RP SADS of both the A- and the B-branch feature a typical double band structure with a main band possessing a pronounced shoulder. Together with the positive absorption above 720 nm (27, 33), these intermediates are clearly characterized as Chl RPs. These features play an important role in deciding whether a kinetic model is consistent with the experimental data in a physically meaningful way, beyond a mere perfect mathematical fit to the data.

Despite the pronounced similarity between the SADS of the analogous RPs in the two branches, those involving the ec3_B Chl (RP1_B and RP2_B) display a substantially more pronounced shoulder/peak on the blue side than those involving ec3_A (RP1_A and RP2_A, see Fig. 5). This finding is also supported by the original transient absorption spectra (Fig. 2) at delay times longer than 10 ps. These spectra in the WT have a well-pronounced double band shape, which is indicative of bleaching of two Chls related to the formation of the RPs (see above). The mutations studied here introduce a noticeable modification of this double band structure. In the case of the PsaA-Y696F mutant, where the formation of the RP is reduced, a pronounced increase of the short-wavelength part of the double band is seen relative to the WT (see Fig. 2, transient absorption spectra in the range 10–30 ps). In contrast, in the PsaB-Y676F mutant, where the yield of RP is reduced, the short-wavelength shoulder is less distinct than in the WT. It follows that the TA spectra after ∼10 ps in the PsaA-Y696F mutant resemble the SADS of the RPs involving ec3_B Chl (Fig. 5) and vice versa for the PsaB-Y676F mutant. These effects are in fact expected and fully consistent with the mutations’ effects on the rate constants, since (partial) suppression of ET in one of the branches leads to a stronger contribution of the spectra from the RPs of the other pathway in the overall decay (Fig. 2).

Branching Ratio.

Based on these ultrafast measurements, the A-branch is the faster one in primary CS and in tertiary ET. However, it is known from the nanosecond measurements (10, 32) that the subsequent ET step to the F_X cluster is about 10-fold slower in the A-branch (τ ∼ 200 ns vs. ∼20 ns). These facts are likely explained by the different ways in which the two branches divide up the available free energy in going from the Ant/RC* state to the F_X state, if one assumes that the differences in rates are primarily due to differences in driving force. However, we cannot exclude differences in reorganization energy either.

The branching ratio (B/A) can be best determined after ∼300 ps after excitation, using the ratio of the population size of RP3_B and RP3_A. We obtain a B/A ratio of 0.61 in WT, which increases to 1.44 in PsaA-Y696F and drops to 0.35 in PsaB-Y676F (Table S2). Similar B/A branching ratios (WT—0.67, PsaA-Y696F—1.77, and PsaB-Y676F—0.21) were estimated from the amplitudes of the kinetic components representing reduction of F_X by PhQ_A or PhQ_B (32).

Some Thoughts Concerning the Intrinsic CS Rate, the Evolutionary Conservation of Bidirectionality in Type I RCS, and Future Approaches in Artificial Photosynthesis.

Interestingly, when PSI trapping kinetics are modeled with a unibranched model (Fig. S3), the effective primary CS rate is about 350 ns^-1, which makes it about two times faster than in PSII (36). The same applies also for the intrinsic primary CS rate from any singular electron donor pigment (SI Text and ref. 27), which is obtained by multiplying the effective CS rate by the number of the pigments amongst which the RC* excitation is equilibrated, e.g. 3–6, depending on the exciton model (see ref. 27 for definitions of these terms). Hence, in PSI the intrinsic CS rate is ∼2.1 ps^-1 (using the unibranched model and assuming a 6-fold degeneracy of the RC), while it is ∼1 ps^-1 in PSII (6). Interestingly, not much attention has been paid to this significant difference between PSI and PSII. Based on the bibranched model, however, both the effective and the intrinsic primary CS rates in each ET branch of PSI are much closer to the ones in PSII. The intrinsic primary CS rate is ∼1.5 ps^-1 in the A-branch and ∼0.8 ps^-1 in the B-branch. This may imply that ET rates from a specific donor cannot much exceed ∼1 ps^-1. This upper limit may be imposed by protein relaxation dynamics (37, 38), which has to provide the ultimate free energy difference that drives the energy trapping in the forward direction. It was discussed earlier (37, 38) that without a dynamic energy relaxation of the early RPs, caused by protein relaxation, efficient photosynthetic CS from a unit with a large antenna system would not be possible.

This may help solve the mystery of why bidirectional CS was maintained in PSI. Although PSI is now heterodimeric, it surely evolved from a homodimeric RC, and in fact the three other members of the type I RC family (heliobacterial, chlorobial, and chloroacidobacterial) are all homodimeric (39–41). PSI is the one member of the type I RCs that became heterodimeric, presumably after a gene duplication event giving rise to the psaA and psaB genes. The subsequent diversification of the PsaA and PsaB polypeptides could have led to inactivation of one of the branches, as it did in all of the type II RCs. However, it did not—both branches remain active, although there are quantitative differences between them. Why this should be so has not been obvious, but we can make a speculative suggestion. The core antenna of PSI contains ∼100 Chls, making it by far the largest of all known antenna/RC units. This increase in antenna size would tend to decrease the total rate and efficiency of CS and trapping. By keeping both branches active, effectively doubling the overall CS rate, that undesirable effect has been minimized in PSI.

Our findings may have implications for the design of artificial photosynthetic units. Using a RC construct with two (or more) active branches instead of one, and accumulating the charge on a single electron acceptor, may significantly reduce the required driving force. New artificial photosynthetic systems could benefit from this concept using multibranch RCs, allowing for an efficient energy trapping even from very large antenna systems using moderate ET rates in each branch.

Materials and Methods

Preparation of Mutant PSI Samples.

Construction of the PsaA-Y696F and PsaB-Y676F mutants in C. reinhardtii was described in ref. 32. PSI particles were purified from P71 FuD7 transformants, which lack PSII and have a low light-harvesting complex (LHC) content (reduced antenna size). The protocol used was based upon standard protocols but modified in order to minimize contamination by LHC (see SI Text for details). Neither of the mutations caused any significant influence on the stationary absorption spectra (cf. Fig. S4).

Ultrafast Spectroscopy.

Femtosecond transient absorption data were recorded with a home-built camera detection system (42) at room temperature using an excitation wavelength of 700 nm with a narrow bandwidth (ca. 9–10 nm) to preferentially excite the RC and thus to enhance RC signals. The polarization of pump and probe light was set at the magic angle. The excitation intensity (pulses with a FWHM of about 70 fs) was kept sufficiently low (∼6 × 10¹² photon cm^-2 per pulse) to avoid annihilation (see SI Text for details).

Data Analysis.

Global lifetime analysis was performed employing lifetime density analysis to obtain the LFD maps as described previously (42). Detailed kinetic modeling was performed on the LFD maps with procedures described in detail in SI Text.

Simulations.

Extensive data analysis using simulated data from bibranched models with realistic noise added was performed using the same analysis procedures to check the integrity and reliability of the approach (see SI Text for details).