Charge separation in the photosystem II reaction center resolved by multispectral two-dimensional electronic spectroscopy

Abstract

The photosystem II reaction center (PSII RC) performs the primary energy conversion steps of oxygenic photosynthesis. While the PSII RC has been studied extensively, the similar time scales of energy transfer and charge separation and the severely overlapping pigment transitions in the Q_y region have led to multiple models of its charge separation mechanism and excitonic structure. Here, we combine two-dimensional electronic spectroscopy (2DES) with a continuum probe and two-dimensional electronic vibrational spectroscopy (2DEV) to study the cyt b559-D1D2 PSII RC at 77 K. This multispectral combination correlates the overlapping Q_y excitons with distinct anion and pigment-specific Q_x and mid-infrared transitions to resolve the charge separation mechanism and excitonic structure. Through extensive simultaneous analysis of the multispectral 2D data, we find that charge separation proceeds on multiple time scales from a delocalized excited state via a single pathway in which Pheo_D1 is the primary electron acceptor, while Chl_D1 and P_D1 act in concert as the primary electron donor.

Primary charge separation in the photosystem II reaction center is resolved by multispectral 2D spectroscopy.

INTRODUCTION

The primary charge separation steps of oxygenic photosynthesis occur within photosystem II (PSII), the only known natural enzyme capable of light-driven water splitting. The unique capabilities of PSII motivate a deep understanding of its structure-function relationship and highly efficient charge separation mechanism. In plants and green algae, PSII is composed of over 25 subunits; about 250 chlorophyll molecules act as light-harvesting antennas to transfer energy to the reaction center, which performs the initial charge separation. The PSII reaction center (PSII RC) contains six chlorophyll a (Chl a) molecules, two pheophytin (Phe), two β-carotenes (Car), and two quinones arranged along the D1 and D2 branches (depicted in Fig. 1A), with charge separation occurring along the D1 side with near-unity quantum efficiency (1). The mechanism of charge separation in the PSII RC has been a topic of debate for many years. Initially, a “multimer model” treated all pigment site energies as equal and suggested that “P680, the primary electron donor of PS II, should not be considered a dimer but a multimer of several weakly coupled pigments, including the pheophytin electron acceptor” (2). This model drew an analogy with the better understood purple bacterial reaction center in which the primary charge separation occurs between the strongly coupled “special pair” pigments (3) but concluded that the weaker relative coupling between the P_D1-P_D2 pair in the PSII RC would lead to a higher degree of delocalization. Later, exciton models assigned distinct site energies to the pigments (4–10) and proposed alternative models of charge separation.

Fig. 1. PSII RC structure and representative multispectral 2D spectra.

(A) Arrangement of the PSII RC pigments, based on the 1.9-Å structure (Protein Data Bank (PDB):3WU2) of Thermosynechococcus vulcanus (65). (B) Representative multispectral 2D spectroscopy used in this study, where we combine broadband 2DES and 2DEV spectra (shown here at t₂ = 200 fs). The 2D excitation axis spans the spectrally congested Q_y region, while broadband detection in the visible (2DES) and mid-IR (2DEV) accesses transitions specific to the RC pigments and charge-separated species. Positive signals (red) represent ground state bleach (GSB), while negative signals (blue) indicate excited state absorption (ESA).

Ultrafast spectroscopic measurements have played a key role in motivating and testing exciton and charge separation models of the PSII RC, with early work summarized in review papers (11, 12). Despite the extensive studies of the PSII RC, the similar time scales of energy transfer and charge separation in this system and the severely overlapping pigment transitions in the Q_y region have led to ongoing debate about its charge separation mechanism and excitonic structure. Two main charge separation mechanisms have been proposed for the PSII RC. On the basis of a photon echo study, Prokhorenko and Holzwarth (13) proposed that the accessory chlorophyll on the D1 branch (Chl_D1) is the primary electron donor in the charge separation process (the “Chl_D1 pathway”: RC* → Chl_D1⁺Pheo_D1⁻ → P_D1⁺Pheo_D1⁻). Alternatively, transient absorption studies by Shuvalov and coworkers (14) claim that charge separation proceeds in a manner similar to the purple bacterial reaction center, in which charge separation begins at the special pair and the accessory Chl_D1 acts as the primary electron acceptor (the “P_D1 pathway”: RC* → P_D2⁺P_D1⁻ → P_D1⁻Chl_D1⁺ → P_D1⁺Pheo_D1⁻). Additional spectroscopic measurements including transient absorption measurements with broadband visible and mid-infrared (mid-IR) probes as well as recent two-dimensional electronic spectroscopy (2DES) experiments either have supported the Chl_D1 hypothesis (15, 16) or have proposed that charge separation proceeds via both the Chl_D1 and P_D1 pathways (17, 18).

To extract the richest information content from the system with simultaneous high frequency and temporal resolution, we use Fourier transform 2D spectroscopy (19), employing a combination of probes in the visible and mid-IR region as depicted in Fig. 1B. Previous 2DES experiments of the PSII RC by our group (20, 21) and others (18, 22–25) have focused on the Q_y region where the severe overlap of the excitons complicates the interpretation of the data and hinders tests of the exciton model and charge separation mechanism. While exciting within the crowded Q_y region is key to unraveling the sequence of events that proceed from the population of specific Q_y excitonic states, it is critical to obtain pigment-specific spectroscopic signatures to distinguish between distinct charge separation pathways. Previous transient absorption spectroscopy using broadband probes have exploited the improved spectral separation of Chl and Pheo Q_x transitions and the formation of distinct anion absorption bands for this purpose (16, 17). In addition, the C═O keto and ester vibrational bands of Chl and Pheo have been shown to be sensitive to the protein environment and to charge separation and have been used by Groot et al. (15) in transient absorption studies of the PSII RC with a mid-IR probe. The approach of combining 2D electronic excitation with a mid-IR probe [2DEV spectroscopy (26)] was recently used by Yoneda et al. (27) to study the PSII RC. Here, we combine broadband 2DES (28) and 2DEV (26), exciting the Q_y band to interrogate its excitonic structure and initiate charge separation. We probe over the visible and mid-IR regions to access a unique combination of pigment-specific spectroscopic markers and use lifetime density analysis (LDA) and global target analysis with simultaneous fitting of the broadband 2DES and 2DEV data to test models of charge separation. We observe and characterize delocalized excitonic and charge transfer (CT) states and trap states (29) that are responsible for slower phases of charge separation. The combination of visible and mid-IR probes reveal a delocalized charge separation process in which Pheo_D1 acts as the primary electron acceptor, with Chl_D1 and P_D1 acting in concert as the primary electron donor. This mechanism unites the distinct Chl_D1 and P_D1 pathways that have been proposed based on lower-dimensional measurements that probed a subset of the transitions considered here.

A 2D spectroscopy experiment uses a sequence of three ultrafast laser pulses with carefully controlled interpulse timing to induce a third-order polarization in the interrogated sample (19). The first two laser pulses (the pump pulses) separated by a coherence time t₁ generate an excited state or ground state population, which decays over the “waiting time” t₂ until the arrival of the third pulse (the probe pulse). The signal radiated as a result of the third-order polarization generated in the sample can be detected in the frequency domain with a spectrometer, defining the detection frequency axis. Fourier transformation of the signal with respect to t₁ generates the excitation axis. 2D correlation plots of detection versus excitation frequency can be generated at a series of t₂ waiting time points to map out the excited-state evolution of the system.

RESULTS

Figure 2A shows the waiting time dependence of the broadband 2DES data. Upon Q_y excitation, the dominant features of the 2DES spectra are an initial positive ground state bleach (GSB) signature from the red edge of the Soret band and a broadband negative excited state absorption (ESA) signal spanning the Q_x and carotenoid regions (30, 31). Charge separation is evident from the increasingly negative signal at 455 nm at longer waiting times, corresponding to the formation of ESA of the Pheo⁻ anion (17, 32, 33). Also evident is the expected increase in Pheo Q_x 0-0 GSB at 545 nm that appears after a few picoseconds and continues to grow more positive at long waiting times, consistent with formation of a charge separated species involving Pheo.

Fig. 2. Broadband 2DES and 2DEV spectra at different t₂ delays.

(A) Broadband-2DES spectra plotted at various t₂ times. The dashed lines denote specific spectral features in the Q_x region: Chl Q_x 0-1 at 585 nm, Pheo Q_x 0-0 at 545 nm, Pheo Q_x 0-1 at 512 nm, Car_D2 0-0 at 506 nm, Car_D2 0-1 at 474 nm, and Car_D1 0-0 at 489 nm. The spectral locations of these peaks are taken from the linear absorption analysis done by Shipman et al. (66) and Rätsep et al. (67). (B) 2DEV spectra at corresponding t₂, with assignments derived from the literature (15, 34–36). Frequencies of notable transitions are denoted by horizontal dashed lines. The 2DEV spectra shows clearer excitation frequency dependence. Positive signals (red) represent GSB, while negative signals (blue) indicate ESA. Contours are drawn at 22 evenly spaced contours from minimum to maximum.

The waiting time dependence of the 2DEV data is shown in Fig. 2B. Compared to the broadband 2DES spectra, the 2DEV spectra show considerably more pronounced excitation frequency dependence. This is likely due to the lack of broad ESA that dominates the 2DES spectra, obscuring the excitation-dependent features that are more clearly observed upon approximate subtraction of the ESA signal (see fig. S1B). The excitation dependence in both the 2DES and 2DEV spectra can be roughly partitioned into two distinct regions centered at ~665 and ~680 nm with rich and distinct waiting time-dependent structure. We base our assignment of the mid-IR features on the work of Nabedryk et al. (34), Noguchi et al. (35, 36), and Groot et al. (15) as summarized in the Supplementary Materials (section S2). Upon excitation at 680 nm, we observe immediate GSB of Pheo via positive peaks at 1722 and 1677 cm⁻¹ (34, 36), as well as GSB of P via a positive peak at ~1704 cm⁻¹ (15, 36). A broadband negative signal is present across the mid-IR detection window, with the initial strongest contribution at ~1664 cm⁻¹, which, in combination with the positive feature at 1657 cm⁻¹, has been attributed to the amide C═O response to charge separation (15, 34, 36). Also notable is a negative feature at ~1711 cm⁻¹ that is present at t₂ = 200 fs and grows in amplitude that been attributed to the keto C═O of P⁺ (15, 36). Upon excitation of the higher-energy Q_y excitons (centered at 665 nm), the dominant initial features include a broad positive GSB feature spanning ~1670 to 1680 cm⁻¹, encompassing assigned Chl_D1 and Pheo GSB features at 1670 and 1677 cm⁻¹, respectively (15, 34). Also present, although with smaller amplitude than seen with 680-nm excitation, is the P GSB at 1704 cm⁻¹. We note that the negative peak we observe near 1635 cm⁻¹ at early times is present in Pheo⁻-minus-Pheo light-induced Fourier transform infrared (FTIR) difference spectra of dithionite-treated samples (34).

Data analysis

To obtain an overview of the kinetics in the broadband 2DES and 2DEV datasets, we applied LDA (16, 37). LDA fits the data using exponential functions with a continuous distribution of time constants. As no prior knowledge of the system dynamics is required for this analysis, this approach is well-suited for studying complicated systems such as the PSII RC. We further process the lifetime density maps (LDMs) obtained from the LDA [implemented via the OPTIMUS software package (37)] using a convolution method to reduce unphysical oscillatory components following reference (38) and personal communication with M. Berg at University of South Carolina. Details of this treatment are given in the Supplementary Materials (section S3).

Figure 3 displays the LDMs obtained from integration of the 2D spectra within two different excitation windows centered at 680 nm (top row) and 665 nm (bottom row) to capture the main excitation frequency–dependent kinetics present in the broadband 2DES (left column) and 2DEV (right column) data. In the LDMs, blue features indicate a rise of GSB or decay of ESA, while conversely, red indicates decay of GSB or rise of ESA. Direct comparison of the kinetics revealed by the LDMs of the broadband 2DES and 2DEV data shows similar overall waiting time–dependent evolution of the spectra, consistent with the expectation that the electronic and vibrational probes provide complementary views of the energy conversion process in the PSII RC. Within both excitation windows, spectral changes can be roughly characterized as taking place with distinct processes within the first ~1 ps, ~1 to 100 ps, and >100 ps. Additional LDMs derived from the 2D data in 3-nm excitation windows spanning the Q_y region are provided in the Supplementary Materials.

Fig. 3. LDMs of the multispectral 2D data.

LDMs of broadband 2DES (left) and 2DEV (right) at different excitation frequency regions: 677 to 683 nm (top) and 662 to 668 nm (bottom). Blue features indicate a rise of GSB or decay of ESA, while conversely, red indicates decay of GSB or rise of ESA.

“Red” excitation (677 to 683 nm)

Upon red excitation, the broadband 2DES reveals immediate signatures of charge separation as indicated by the 455 nm anion ESA and the broadband ESA features throughout the visible and mid-IR. The initial rise in broadband ESA, most apparent in the 2DEV LDM, is complete within ~0.3 ps. At ~0.3 ps, there is an overall decrease of the GSB of P at 1705 cm⁻¹, which, at later times, is difficult to distinguish from the rising ESA from P⁺ at 1711 cm⁻¹. At ~3 ps, the 2DES LDM exhibits a peak in the anion ESA coincident with a rise in Pheo GSB, consistent with Pheo signatures in the 2DEV LDM on this time scale. Both 2DES and 2DEV LDMs exhibit a later stage of charge separation on the time scale of >100 ps as evident from the strong growth in the anion ESA at 455 nm and the P⁺ ESA at 1711 cm⁻¹, accompanied by Pheo GSB growth and positive/negative features at 1657/1664 cm⁻¹ associated with the amide C═O response to charge separation (15, 34, 36).

“Blue” excitation (662 to 668 nm)

Unlike the red excitation case, there is little evidence of immediate charge separation upon blue excitation, as indicated by the relative lack of initial anion band absorption in the 2DES LDM or P⁺/Pheo⁻ ESA markers in the 2DEV LDM. Within the first 300 fs, Pheo GSB appears in both the 2DES LDM (Pheo Q_x 0-0 and 0-1) and the 2DEV LDM. A decay of P GSB is also evident in the 2DEV LDM on the same time scale. Between ~3 and 20 ps, the growth of Pheo GSB competes with ESA from P⁺Pheo⁻ and P⁺ indicating increasing charge separation. We note the concomitant bandshift feature present in the Car region on this time scale, consistent with previous reports from transient absorption studies (39). The strong broad feature between 1660 and 1680 cm⁻¹ at ~20 ps has contributions from the C═O response to charge separation (15, 34, 36). The GSB features in this region are difficult to interpret due to their close proximity and competing oppositely signed processes occurring on this time scale. On the time scale of >100 ps, both 2DES and 2DEV LDMs exhibit a later stage of charge separation, with spectral markers that are consistent with the red excitation data.

Global target analysis

To further explore the charge separation mechanism and establish a model that can capture the kinetic processes in the multispectral data, we turn to global target analysis (40). This approach fits the data using a trial kinetic model and produces a set of species-associated spectra (SAS) and rate constants. Here, we perform simultaneous fits of broadband 2DES and 2DEV data to multiple kinetic models, chosen based on the time scales and spectral features obtained in the LDMs as well as commonly used models from the literature. These models and the resulting fits are discussed in further detail in the Supplementary Materials. We acknowledge that we cannot rule out untested kinetic models. We extensively examined the two-pathway model of charge separation proposed by Romero et al. (17) and found that it did not produce SAS consistent with a distinct P_D1 pathway lacking participation from Pheo. While a simplification of the two-pathway model that omitted the slow energy transfer from Chlz pigments was found to adequately fit the broadband 2DES data, it failed to give reasonable results for the 2DEV fits. The best model that adequately describes the kinetics in both broadband 2DES and 2DEV datasets is given in Fig. 4A. It consists of five compartments with 2D SAS shown in Fig. 4B: two distinct exciton states (RC* and Trap*), two intermediate radical pair (RP1 and RP2), and the final charge separated state P_D1⁺Pheo_D1⁻ (RP3). The lifetime of the final charge separated state RP3 is much longer than the ~1-ns waiting time limit of our experiment. We therefore fix the RP3 lifetime to 13 ns following the work of Romero et al. (17). Within our model, a rapid charge separation channel exists: RC* → RP1 → RP2, followed by a slow ~150-ps relaxation process of the charge separated state RP2 → RP3. A slower charge separation route also proceeds from a distinct trap state. The existence of trap states at 77 K (17), 4 K (29, 41), and room temperature (16) has been previously proposed, as has a slow phase associated with the formation of the relaxed charge separated state (17). We find that Trap* is depopulated within ~13 ps, proceeding to form RP2 via the same pathway as RC*. The 2D-SAS of the visible and IR show consistent excitation wavelength dependence, lending credibility to our target model through which the final charge separated state is reached upon excitation of two distinct exciton states (RC* and Trap*).

Fig. 4. Global target model and resulting SAS from simultaneous fitting of the multispectral 2D data.

(A) Five-compartment model with one trap state and three RP states. (B) 2D SAS of each compartment, with both visible (top) and mid-IR (bottom) components. The sign convention in the 2D SAS is the same as for the 2D spectra, where positive signals (red) represent GSB, while negative signals (blue) indicate ESA. The 2D SAS amplitudes are normalized with respect to the final charge separated state, and contours are drawn at 22 evenly spaced contours between −1 and 1.

The 2D SAS in the mid-IR exhibit some spurious structure in regions of low amplitude, motivating a simpler representation that captures the excitation frequency dependence and enables visualization of the spectroscopic features associated with each compartment of our target model. For this purpose, we make the approximation that the 2D SAS of the two initial excitons (RC* and Trap*) can be represented as a product of a 1D SAS and a Gaussian distribution that captures the excitation frequency dependence of the 2D data. The resulting Gaussian distributions and 1D SAS are shown in Figs. 5 and 6, respectively. The fitting procedure is described in further detail in the Supplementary Materials (section S6), where we show that this reduced representation reproduces the main 2D spectral features and the kinetic processes present in both datasets.

Fig. 5. Linear absorption of the D1D2 RC and distinct exciton states used in the global target model.

The experimental linear absorption spectrum of the D1D2 RC at 77 K (green solid curve) fit with two Gaussian distributions that represent the two distinct initial exciton states RC* [red solid curve, centered at 679 nm, full width at half maximum (FWHM) 4.2 nm] and Trap* (blue solid curve, centered at 669 nm, FWHM 5.4 nm). The black dashed line is the sum of the two Gaussian distributions.

Fig. 6. 1D SAS of the compartments of the global target model.

SAS from the fits to the global target model shown in Fig. 4A in the visible (left) and mid-IR (right). The amplitudes are normalized with respect to the final charge separated state P_D1⁺Pheo_D1⁻ (RP3).

DISCUSSION

The 1D SAS shown in Fig. 6 allow us to clearly visualize the spectroscopic features associated with each compartment of the global target model. First, we examine the rapid charge separation process that proceeds upon excitation of RC*, with excitation centered at 679 nm. We then discuss the slower charge separation proceeding from the trap states and the implications for our understanding of the mechanism of charge separation in the PSII RC.

Rapid charge separation

The visible SAS for RC* exhibits GSB of the Pheo Q_x peaks as well as ESA from Chl ESA and the Pheo anion band. In the mid-IR SAS, we observe Pheo, Chl_D1, and P GSB peaks. Thus, RC* is consistent with a delocalized exciton state spanning the RC pigments: (Chl_D1Pheo_D1P)*. This state also appears to have some degree of CT character, consistent with Stark spectroscopy experiments that propose direct excitation of mixed exciton-CT states (Chl_D1^δ+Phe_D1^δ−)^* at 681 nm and (P_D2⁺P_D1⁻)^δ* at 684 nm (42). This may explain the strong similarities between RC* and the RP1 state, particularly in the visible SAS. From RC* to RP1, the visible SAS displays a slight rise in the Pheo anion ESA as well as a small rise and red shift of the Pheo Q_x 0-0 GSB. These signatures are consistent with charge separation involving Pheo⁻. In the mid-IR SAS, the GSB peak associated with P decreases substantially from RC* to RP1, consistent with energy transfer from P. The decrease in P GSB is accompanied by a relative increase in ESA from P⁺. The GSB peak at 1670 cm⁻¹, assigned to Chl_D1 (15, 35, 36), is present in both the 1D SAS of RC* and RP1. These observations are consistent with assignment of RP1 as (Chl_D1P_D1)⁺Pheo_D1⁻. Comparing RP2 to RP1, the visible SAS reveal a substantial increase in the Pheo anion band and a slight increase in the Pheo Q_x 0-0 GSB with an additional small red shift. The GSB of Pheo Q_x 0-1 can also be resolved in RP2, which also shows a small increase in the GSB of Chl Q_x 0-1 and Chl ESA in comparison with RP1. In the mid-IR, a comparison of the SAS of RP1 and RP2 shows an increase in Pheo GSB, as well as P⁺ and Pheo⁻ ESA, and a vanishing of the Chl_D1 GSB. The reduced P GSB peak in the mid-IR is likely a result of the competing nearby P⁺ ESA. RP2 also exhibits the expected strong derivative-like amide response that has been associated with charge separation. The visible and mid-IR SAS support the assignment of RP2 as the final charge separated state P_D1⁺Pheo_D1⁻. We note a slight red shift in the P⁺ ESA peak from RC* to RP1 to RP2, which Yoneda et al. (27) assigned to the hole migration, consistent with our assignments. The mid-IR SAS for RP2 and RP3 are very similar and are in good agreement with the final radical pair spectrum reported by Groot et al. (15) in their pump mid-IR probe study of the PSII RC and with assignments from light-induced difference FTIR studies (34–36, 43). On the basis of the combination of visible and mid-IR SAS, we find that a consistent picture in which the charge separation in the PSII RC proceeds via the following sequential pathway

$$\begin{array}{rl}& ({\mathrm{Chl}}_{\mathrm{D}1}{\mathrm{P}\mathrm{heo}}_{\mathrm{D}1}\mathrm{P})\ast \stackrel{0.45\mathrm{ps}}{\to }{({\mathrm{Chl}}_{\mathrm{D}1}{\mathrm{P}}_{\mathrm{D}1})}^{+}{{\mathrm{P}\mathrm{heo}}_{\mathrm{D}1}}^{-}\stackrel{2.8\mathrm{ps}}{\to }{{\mathrm{P}}_{\mathrm{D}1}}^{+}{{\mathrm{P}\mathrm{heo}}_{\mathrm{D}1}}^{-}\stackrel{150\mathrm{ps}}{\to }\\ & {({{\mathrm{P}}_{\mathrm{D}1}}^{+}{{\mathrm{P}\mathrm{heo}}_{\mathrm{D}1}}^{-})}^{\mathrm{relaxed}}\end{array}$$ (1)

The time scales are in reasonable agreement with previous works which suggest that initial charge separation at 77 K occurs within 600 to 800 fs (15) or ~1 to 3 ps (17), followed by a slower relaxation process that has also been reported in both room temperature and 77 K experiments (15–17). This slow ~150-ps process may be caused by dielectric relaxation as the protein environment adjusts to the newly formed charge separated state. Stark shift measurements have revealed dielectric asymmetry in the protein environments of the active and inactive branches of the purple bacterial reaction center (44). Studies of wild-type and mutant purple bacterial reaction centers have compared the kinetics of transient absorption with Stark shift measurements, finding differences on longer time scales that have been attributed to slow dielectric relaxation induced by protein dynamics (45).

Charge separation from the trap state

Consistent with previous reports (16, 17, 32), we also observed slower phases of charge separation that we interpret as originating from trap states. Trap*, excited at ~669 nm, exhibits visible SAS features that include Chl ESA, Chl Q_x bleach, and a derivative-like feature in the Car region. Notably, no Pheo Q_x bleach or Pheo anion band signatures are present. The mid-IR features of Trap* are highly distinct from the other SAS, with a broad prominent GSB centered at ~1673 cm⁻¹ and an ESA at ~1635 cm⁻¹. The broad GSB encompasses several identified mid-IR bands, complicating its assignment. On the basis of the visible SAS and the lifetime of ~20 ps (16, 17), we propose that Trap* represents the Chlz pigments which have been reported to transfer energy to the center pigments on this time scale, accompanied by a Stark shift of the Car peaks (39). We note that FTIR studies of spinach membranes report absorption from neutral Chlz between ~1660 and 1687 cm⁻¹ (35, 46), which is reasonably consistent with our Trap* SAS. We note that a modified global target model that produces a comparable fit to the data and very similar SAS has Trap* decay directly to RP2 on a similar time scale (~24 ps) as described in the Supplementary Materials. We also considered an additional model with two trap states, similar to that proposed by Romero et al. (17), to account for the slowest charge separation processes evident in the LDMs. This model is presented in the Supplementary Materials and includes a Trap2* exciton at ~670 nm. While this model provided a reasonable fit to the data, it appeared unphysical: The visible and mid-IR SAS were inconsistent, and the mid-IR SAS closely resembled RP3, suggesting that RP3 is photoexcited directly, with little subsequent evolution.

Theoretical studies of the PSII RC are highly challenging due to the number of pigments and the necessity of accounting for the effect of the protein environment. Recent high-level QM/MM calculations find that Chl_D1 has the lowest site energy and support its assignment as the primary electron donor (7, 47). The calculation of CT states is particularly difficult, and it is our hope that our spectroscopic measurements will provide feedback for testing and refining theoretical models of the electronic excited states and charge separation process in the PSII RC. Ideally, such models would provide quantitative agreement with 2D spectroscopic data, obviating the need for global target analysis and motivating new experimental measurements to further validate the model. Our findings are in reasonable agreement with the empirical exciton models of Novoderezhkin (9), Gelzinis (10), and Renger (6, 48). The Novoderezhkin model has high participation from Chl_D1, Pheo_D1, and P_D1 pigments in the lowest-energy Q_y exciton states, consistent with our RC* exciton assignment of (Chl_D1Pheo_D1P)* that is excited on the red edge (~679 nm). In both the Gelzinis and Renger models, the lowest-energy exciton state is dominated by Chl_D1 but also contains nonzero contributions from Pheo_D1 and P. Our study suggests that Chl_D1, Pheo_D1, and P_D1 pigments all participate in the RC* exciton but does not determine their relative participation ratios as would be needed for quantitative comparison with the exciton models. We note that the exciton delocalization in our 77 K experiments is expected to be greater than at room temperature where disorder and coupling to vibrational degrees of freedom has been shown to decrease exciton delocalization (49). Our identification of Trap* as the Chlz pigments is consistent with the Novoderezhkin and Gelzinis models, in which they participate heavily in the highest energy Q_y excitons (9, 10). It is also in agreement with the work of Raszewski et al. (48), who predict slow transfer from Chlz to the core pigments upon excitation at ~665 nm.

To date, there has been little consensus on the charge separation mechanism of the PSII RC. Transient absorption measurements exciting the Q_y region and probing in the visible or mid-IR have favored either the Chl_D1 (15, 16) or P_D1 pathway (14). Still, others have suggested that both pathways occur, with protein conformation dictating which pathway will be taken (17). A 2DES study in the Q_y region used global analysis to find support for the Chl_D1 pathway (18). On the basis of calculations they proposed that, while the Chl_D1 pathway was dominant, the P_D1 pathway was viable but not well resolved in the experiments due to its minor relative contribution. Here, we have used 2D excitation spanning the Q_y region combined with a multispectral probe to access a broad range of pigment-specific and anion/ion spectral markers spanning the visible and mid-IR. Without relying on a global target model, the LDMs demonstrate that sub-picosecond spectral signatures associated with Pheo⁻ are present independent of Q_y excitation frequency (see the Supplementary Materials for additional LDMs beyond those shown in Fig. 3). This rules out the possibility that a distinct P_D1 pathway can be accessed via selective excitation of specific Q_y excitonic states. While such a pathway might be accessed on slower time scales following energy transfer from Chlz, our global target analysis failed to produce SAS with consistent visible and mid-IR signatures to support the existence of such a pathway. We note that the spectral resolution of our study with respect to the Q_y excitation frequency axis compares favorably with previous work (15, 17, 27), ruling out the possibility that we had insufficient resolution to isolate processes that have been observed in other measurements. Further comparison of our work with other measurements is detailed in the Supplementary Materials.

In conclusion, we have used multispectral 2DES, exciting the spectrally congested Q_y region and combining the rich spectroscopic signatures in the visible and mid-IR to gain insight into the Q_y excitonic structure and reveal the charge separation mechanism of the PSII RC. Our measurements resolve excitonic states with CT character within the Q_y region, in reasonable agreement with Stark spectroscopy measurements (42). Our extensive kinetic analysis, combining lifetime density and global target analysis of the multispectral data, reveals that charge separation in the PSII RC proceeds via a sequential mechanism that borrows elements of the distinct Chl_D1 and P_D1 pathways. We find that excitation of a delocalized RC exciton (Chl_D1Pheo_D1P)* leads to rapid charge separation. A highly delocalized primary charge separated state was proposed in the original multimer model (2), although the model’s assumption of equal site energies for the RC pigments has been shown to be inconsistent with experimental and theoretical work (6, 9–11, 48, 50–52). Our observations are consistent with our recent study of the heliobacterial RC (53), which has been proposed to be the RC most similar to the common ancestor of all photosynthetic RCs (54, 55), in which we observed charge separation from a delocalized state via a similar mechanism to that proposed here. Like the Chl_D1 pathway, and consistent with many previous studies, we find that Pheo_D1 acts as the primary electron acceptor. However, our combination of visible and mid-IR spectral signatures indicate that the primary electron donor is not strictly Chl_D1 but involves Chl_D1 and P_D1 acting in concert. Such delocalization of the hole in the primary charge separated state would draw cation density away from Pheo⁻, helping prevent wasteful charge recombination, as we have proposed for the heliobacterial RC (53).

MATERIALS AND METHODS

The D1-D2-Cytb559 reaction center (PSII RC) complexes were extracted from spinach following a method based on the work of Berthold et al. (56) and van Leeuwen et al. (57). We verified purity of the sample and removal of CP47 and CP43 via a ratio of 1.2 for the linear absorption at 417:435 nm as discussed by Eijckelhoff et al. (58). The RC samples were further diluted with glycerol with a ratio of 1:1 (volume:volume). In the broadband 2DES experiment, the RC sample was concentrated to obtain an optical density (OD) of ~1.5 measured at 675 nm at room temperature for a path length of 380 μm. In the 2DEV experiment, the OD of the sample measured at 675 nm was ~0.6 at room temperature for a path length of 100 μm. Both 2DES and 2DEV measurements were collected at 77 K using an Oxford Instruments cryostat (MicrostatN).

The broadband 2DES and 2DEV experiments were performed at the Laboratory for Ultrafast Multidimensional Optical Spectroscopy (LUMOS) (59) at the University of Michigan, using pulse shaping–based 2D spectroscopy in the pump-probe geometry (60). The laser repetition rate was set to 500 Hz to avoid a buildup of long-lived triplet states in the PSII RC (48). The pump was generated by using a home-built noncollinear optical parametric amplifier (NOPA) (61) with 50 and 100 nJ of pulse energy at sample position in the broadband 2DES and 2DEV experiments, respectively. The pump was compressed to 12 fs of pulse duration using the Spectral Phase of Electric field by Analytic Reconstruction (SPEAR) method (62). The 1/e² beam waist of the pump at sample position was ~120 μm, giving a beaching of ~5 to 6% in both experiments. In broadband 2DES, the probe was a white-light continuum generated by focusing 800-nm light from a regenerative amplifier into a CaF₂ crystal. The chirp of the white-light was corrected after data collection (63). In the 2DEV experiment, the mid-IR probe, generated by difference frequency generation (DFG), was centered at ~5988 nm with a full width at half maximum (FWHM) of >300 nm. The mid-IR probe pulse energy was kept below 100 nJ before the sample to avoid both nonlinear processes induced by the probe and saturation of the detector. A two-step phase cycling scheme was used to remove scattering and background signals (64). In both experiments, t₁ was scanned to a maximum delay of 210 fs. Waiting time t₂ was scanned from −1 to 5 ps in linear steps and then from 5 to 1 ns in logarithmically spaced steps. A total of 21,000 and 693,000 laser shots were used to construct each averaged 2D spectrum in broadband 2DES and 2DEV experiments, respectively. Experiments were repeated at least three times on different samples to confirm reproducibility.

Supplementary Materials

This PDF file includes:

Figs. S1 to S14

Text S1 to S9

Table S1

References

View/request a protocol for this paper from Bio-protocol.