Excitonic structure and charge separation in the heliobacterial reaction center probed by multispectral multidimensional spectroscopy

Photochemical reaction centers are the engines that drive photosynthesis. The reaction center from heliobacteria (HbRC) has been proposed to most closely resemble the common ancestor of photosynthetic reaction centers, motivating a detailed understanding of its structure-function relationship. The recent elucidation of the HbRC crystal structure motivates advanced spectroscopic studies of its excitonic structure and charge separation mechanism. We perform multispectral two-dimensional electronic spectroscopy of the HbRC and corresponding numerical simulations, resolving the electronic structure and testing and refining recent excitonic models. Through extensive examination of the kinetic data by lifetime density analysis and global target analysis, we reveal that charge separation proceeds via a single pathway in which the distinct A0 chlorophyll a pigment is the primary electron acceptor. In addition, we find strong delocalization of the charge separation intermediate. Our findings have general implications for the understanding of photosynthetic charge separation mechanisms, and how they might be tuned to achieve different functional goals.

and 807 nm, providing firm evidence for the formation of the CSS (P800 + A0 -). A slow CS component (t > 100 ps) has been observed previously, although only at cryogenic temperatures 1 . We also observe the slow component following antenna excitation at 690 nm ( Supplementary Fig. 4 panels b1-b3), suggesting a role for antenna excitons in this process. Fig. 3 | Lifetime density analysis shows energy transfer and charge separation dynamics after excitation of the RC. Panels a1-a3) Lifetime density maps under excitation at 666 nm. Dashed lines indicate the various peaks of key participants during charge separation. From left to right, these dashed lines are located at 573 nm, 578 nm, 666 nm, 690 nm, 786 nm, 794 nm, 807 nm and 814 nm to enable comparison of peak locations. The arrows are used as guidance for the lifetime peaks corresponding to the crucial photoexcited dynamics as discussed in the main text. The lifetime constants are annotated beside the arrows. The arrows are color-coded for assignment of features associated with vibrational relaxation or energy transfer (red), charge separation (black), and a combination of both (green). The contour levels are -1:0.1:1 of the maximum amplitude. Panels b1-b3) Slice spectra at λex = 666 nm and τ = 1.74 ps and 228 ps corresponding to the main charge separation processes. The derivative features in the BChl g Qx band with minima and maxima at 573 and 578 nm reveal the formation of the charge separated states. The peaks at 666 nm indicate involvement of A0 in these processes.

Supplementary
While the spectrum at 1.7 ps shows Stark lineshapes of the BChl g Qx peak (Supplementary The signal in the spectral range from 550 to 600 nm is multiplied by a factor of 4 to better visualize the Stark line shape. We used a sequential model to fit the transient data produced by excitation at 666 nm ( Supplementary Fig.   5). Five compartments are required to obtain a satisfactory fit. The first evolution-associated spectra (EAS1) is assigned to the primary product of photoexcitation and exhibits an A0 peak at 668 nm, a weak BChl g Qx peak at 573 nm and a BChl g Qy peak at 780 nm. This observation suggests that the primary photoexcitation is delocalized over A0 and Acc, which is consistent with the excitonic assignments proposed in the manuscript. EAS1 evolves to EAS2 with a time constant of 1.1 ps via vibrational relaxation, excitonic relaxation or energy transfer. Thus, EAS2 could be a mixture of both the relaxed RC exciton and the antenna exciton. The EAS3 exhibits a Stark lineshape at 576 nm, an A0 peak at 666 nm and a BChl g Qy at 814 nm.
The rapid formation of EAS3 suggests that the initial charge separation occurs at about 1.1 ps. The simultaneous appearance of the Stark lineshapes and the 814-nm feature also implies that the latter stems from a charge separated state (CSS). The spectral profile of EAS4 resembles EAS5 except that it exhibits a strong peak at 816 nm. The 816 nm peak can be attributed to the lowest energy exciton, generated either To test alternative charge separation models and determine whether charge separation produces a common intermediate CSS from either the antenna excitons or the high-energy RC exciton, we extracted 'pumpprobe' spectra upon excitation at 690 nm from 2DES and then fit the transient data using different kinetic models. Since only antenna excitons are photoexcited in this case, if the Acc acts as the primary acceptor, these fits promise to capture the predicted intermediate state, P800 + Acc -.
As discussed above, LDMs (Supplementary Fig. S4) show charge separation signatures at 1.1 ps, 8.6, and 180 ps. Thus, we first fit the data using a kinetic model with three parallel charge separation pathways (Supplementary Fig. 6a). In this model, we also take into account excitonic relaxation and vibrational relaxation, which take place from a few hundreds of femtosecond to several picoseconds, as reported in previous studies 1- 4 . We find that the SAS X3 exhibits a Stark lineshape of the BChl g Qx peak and an A0 peak. The resemblance of X3 and (P800Acc) + A0suggests that A0 is the primary acceptor in the initial CS step. No clear spectral signatures of a CSS are seen in X2. One feature that cannot be well explained in X4 is the significant negative-going Chl peak. We conclude that this model cannot give a satisfactory fit with physically meaningful SAS.
As shown in Supplementary Fig. 6c, we also considered a kinetic model involving two single step  5) and the fits of the 2D dataset in the manuscript (i.e., CSI in the Figure 3 of the manuscript). This observation provides further strong evidence that a common CSS is involved in the charge-separation pathways and that A0 is the primary electron acceptor, for both the antenna exciton and the high energy RC exciton.

Supplementary Note 4.3 Other kinetic models to fit the 2D data
Supplementary Fig. 7 | Global-target analysis using an alternative kinetic model that includes two independent charge separation pathways for the antenna exciton and the high energy RC exciton (AccA0). a) Kinetic model with time constants obtained from the fits. b) Species-associated spectra. The contour levels are -1:0.1:1 of the maximum amplitude.
We also considered other kinetic models with independent charge separation pathways for excitation of the antenna and RC domains. These kinetic models promise to capture the intermediate state of P800 + Accif Acc acts as the primary acceptor in the charge separation pathway for the antenna exciton.
Previous studies 1,2,5,6 and the LDA in the current work have shown that the (AccA0) exciton transition to the CSS on the time scale of several picoseconds while for the antenna excitons, charge separation takes place in three time windows ranging from several picoseconds to hundreds of picoseconds. Supplementary   Fig. 7 shows target analysis using a kinetic model with two charge separation pathway: 1) one for the (AccA0) exciton and 2) one for the antenna excitons. Since charge separation takes place in three time windows, we included a back energy transfer pathway from the lowest energy exciton to the antenna exciton, which takes into account of the slowest charge separation on the time scale of hundreds of picosecond. However, two features in SASs indicates that this model does not work well. The peaks in the SASs, X2 and X3 are centered at the same wavelengths but their amplitudes have opposite signs. This is an artifact which is often observed in the exponential fit. In the charge-separation pathway for antenna excitons, the SASs, X4 and X5 exhibits weak Chl a peaks at 666 nm, suggesting that A0 may be involved in the initial step(s) of charge separation.
The analysis above also shows that for the antenna-exciton pathway (X3®X5®X6), the final charge separation step takes place with a time constant of 22 ps. Previous studies 2,7 have proposed that the rate-limiting step for this pathway may be the energy-transfer step from the antenna to the RC and charge separation may take place at about 1 ps. Therefore, we fit the 2D data using the same kinetic model but put a constraint of the charge separation rate (X5®X6) to be in the range of (5 ps) -1 to (0.5 ps) -1 . The fitting results are shown in Supplementary Fig. 8. We find that SAS X5 exhibits a Chl a peak at 666 nm and a Stark shift feature at the Qx band of BChl g. This observation suggests the involvement of the A0 Chl in the initial charge-separated step. Supplementary Fig. 8 | Global-target analysis using an alternative kinetic model that includes two independent charge separation pathways for the antenna exciton and the high energy RC exciton (AccA0). a) Kinetic model with time constants obtained from the fits. b) Species-associated spectra. The contour levels are -1:0.1:1 of the maximum amplitude. Supplementary Fig. 9 | Global-target analysis using an alternative kinetic model that includes two independent charge separation pathways for the antenna exciton and the high energy RC exciton (AccA0). a) Kinetic model with time constants obtained from the fits. b) Species-associated spectra. The contour levels are -1:0.1:1 of the maximum amplitude. Supplementary Fig. 9 displays the global fits of the 2DES data using another kinetic model with two charge separation pathways: 1) one for the (AccA0) exciton and 2) one for the antenna excitons. Different from the previous kinetic models, both forward and backward energy transfer between all antenna excitons (X4DX3DX5) are considered, and charge separation takes place via a one-step mechanism.
However, this fit does not produce a satisfactory result. First, SAS of an antenna exciton X3 exhibits an A0 peak while the RC exciton X2 does not. Second, the peaks at around 575 nm and 800 nm in the SASs, X2 and X3 are centered at the same wavelengths but their amplitudes have opposite signs, which is an artifact in the exponential fit. Supplementary Fig. 10. Global-target analysis using an alternative kinetic model that includes three independent charge separation pathways for the antenna exciton and the high energy RC exciton (AccA0). a) Kinetic model with time constants obtained from the fits. b) Species-associated spectra. The contour levels are -1:0.1:1 of the maximum amplitude. Supplementary Fig. 10 displays the global fits of the 2DES data using a kinetic model with three charge separation pathways: 1) one for the (AccA0) exciton and 2) two for the antenna excitons. In the fit, we find that the antenna-exciton pathway exhibits an SAS (X4) with a Stark feature in the BChl g Qx band, an A0 peak at 666 nm and two BChl g Qy peaks at 790 nm and 814 nm. This feature resembles the intermediate pathways has also been tested. However, no convergent answer was obtained for the fits using four charge separation pathways.

Supplementary Note 5 Spectroscopic Measurements
2DES spectra were measured by using a pump-probe geometry 2DES setup as described previously 8 .
Briefly, two regenerative amplifiers (i.e., Spitfire Pro and Solstice from Spectra Physics) seeded by a Ti:Sapphire oscillator (MaiTai SP from Spectra Physics) are used as the laser sources. The output from the Spitfire Pro (40-fs pulses, 4-mJ, 800 nm, 500 Hz) feeds a home-built two-stage non-collinear optical parametric amplifiers (NOPAs) 9 used as the pump beam. The pump beam is sent through a precompensating combination of two gratings and two prisms and then into an acousto-optic pulse shaper (Dazzler, Fastlite), where a compressed pulse pair with a programmable time delay (t1) is generated. As shown in Supplementary Fig. 11, the pump NOPA are compressed to 12 fs using the SPEAR method 10 .
The 1-kHz output from the Solstice, chopped at 500 Hz, pumps a commercial optical parametric amplifier (TOPAS Twins from Light Conversion) to generate a 1300-nm seed beam. A portion of the 1300-nm seed beam is then focused into a 1-mm sapphire plate to generate the white light continuum (i.e., the probe beam). The rest of the 1300-nm beam, combined with the rest of the Solstice output (i.e., 1 mJ), are used to feed a home-built degenerate optical parametric amplifier (DOPA) 11 to generate the near-IR pump beam.
The pump and probe pulses are focused at the sample position to generate the third-order 2DES signal, which is detected by a CCD camera (Princeton Instruments PIXIS 100B). During the experiments, t1 is scanned using the Dazzler from 0 to 150 fs with time steps of 2 fs. The 2DES data have been collected in a rotating frame. The pump-probe delay (T) is controlled by an optical delay line (DDS220, Thorlabs Inc.) and scanned from -5 to 1000 ps for both the pump-probe and 2DES. A two-phase cycling scheme is used as described previously to remove scattering and background signals 12 . A shutter added in the probe arm removed residual scattering from the pump. In the experiments, the pulse energy of pump pulses was ~40 nJ and the beam waists (1/e 2 ) for both pump and probe were ~200 μm. 2D experiments were performed under the magic-angle condition and at least three times to ensure reproducibility. The data is analyzed using home-written MatLab scripts. The chirp of the probe pulse has been corrected using a third-order polynomial function as described previously 13 . The lifetime density analysis is performed using the OPTIMUS software 14 . The global-target analysis is performed using CarpetView3D (Light Conversion).

Supplementary Note 7 Simulations of the Stark Shift of A0
To verify that the blue shift of the A0 GSB at 946 ps is caused by the Stark effect, we simulated the Stark spectrum of A0 in the passive branch induced by P800 + A0 -. As described in a previous study 4 , the Stark shift (Du) can be calculated using the following formula: where the changes in permanent dipole moment (Dµ) and polarizability (Δα) of Chl a are set to 0.5 D/f and 1.5 Å 3 /f 2 , respectively, where f is the local field correction factor and is set to 1. The angle θ between (Dµ) and the electric field is calculated using the HbRC crystallographic data 18 . The orientation of dipole moment is deviated from transition dipole moment by 15˚ according to the previous study 19 . The Stark spectrum of A0 in the presence of P800 + A0is shown in Supplementary Fig. 13. Furthermore, we also simulated the overall transient absorption spectrum in the A0 Qy band by summing the A0 GSB spectrum and the Stark spectrum of A0 in the passive branch. In this calculation, we estimated that about 6% of A0 is photoexcited according to our transient absorption measurement. As shown in Supplementary Fig. 13, our simulation is consistent with the experimental observation, supporting our assignment that the blue shift of the A0 GSB peak is due to the Stark effect.
reference, the reduction potentials of Chl a and BChl a determined electrochemically are indicated on the right side; the bar is the mean of the reported values, whose range is indicated by the brackets 22 .
(Although the potential of BChl g has not been determined experimentally, due to its tendency to irreversibly oxidize, its potential was calculated to be only 10 mV more reducing than that of BChl a 23 .) Arrows indicate various processes occurring within the RC: excitation (Exc, blue dotted), charge separation (CS, green solid), secondary electron transfer (2° ET, orange solid), and charge recombination (CR, red dashed).