On the Conflicting Estimations of Pigment Site Energies in Photosynthetic Complexes: A Case Study of the CP47 Complex.

We focus on problems with elucidation of site energies [Formula: see text] for photosynthetic complexes (PSCs) in order to raise some genuine concern regarding the conflicting estimations propagating in the literature. As an example, we provide a stern assessment of the site energies extracted from fits to optical spectra of the widely studied CP47 antenna complex of photosystem II from spinach, though many general comments apply to other PSCs as well. Correct values of [Formula: see text] for chlorophyll (Chl) a in CP47 are essential for understanding its excitonic structure, population dynamics, and excitation energy pathway(s). To demonstrate this, we present a case study where simultaneous fits of multiple spectra (absorption, emission, circular dichroism, and nonresonant hole-burned spectra) show that several sets of parameters can fit the spectra very well. Importantly, we show that variable emission maxima (690-695 nm) and sample-dependent bleaching in nonresonant hole-burning spectra reported in literature could be explained, assuming that many previously studied CP47 samples were a mixture of intact and destabilized proteins. It appears that the destabilized subpopulation of CP47 complexes could feature a weakened hydrogen bond between the 13(1)-keto group of Chl29 and the PsbH protein subunit, though other possibilities cannot be entirely excluded, as discussed in this work. Possible implications of our findings are briefly discussed.

Introduction

Photosynthetic complexes

Light-absorbing chromophores in various photosynthetic complexes (PSCs) initiate a highly complex series of processes.1 The latter drives chemical reactions that support nearly all life on the Earth.1 To understand the excitonic structure and dynamics of PSCs, it is essential to know: (i) pigment site energies , the transition energies in the absence of pigment–pigment interactions; (ii) intramolecular vibrational modes; (iii) interpigment couplings; and (iv) interactions between pigments and the protein environment, ie, electron–phonon (el–ph) coupling. Electronic coupling to phonon and intramolecular modes is characterized by the phonon and vibrational spectral densities, ie, Jph(ω) and Jvib(ω), respectively.2,3 While spectral densities can be measured experimentally4–8 and interpigment coupling matrix elements can be calculated (assuming that X-ray structures are available) by several approaches,9 the site energies are typically extracted from simultaneous fits of various experimental data.7,10,11 Quantum chemical approaches can also be used,9,12–15 but the calculated values have to be further optimized by fitting algorithms; however, the values obtained for structure-based calculations can aid in discarding unrealistic site energy sets obtained from the fitting algorithm. Conversely, the calculated site energies determined from fits of experimental results can test the accuracy of various quantum mechanical methods used in site energy calculations. The difficulties in finding real site energies are not surprising, as PSCs are very intricate biological systems.

Spectroscopic techniques often used in photosynthesis research

Widely used methods include high-resolution (laser-based) frequency-domain spectroscopies, ie, hole burning (HB) and fluorescence line narrowing (FLN),16,17 single photosynthetic complex spectroscopy,18,19 circular dichroism (CD) and linear dichroism (LD),20 and circularly polarized luminescence (CPL).21–24 Time-resolved techniques (eg, pump-probe25–27 and two-dimensional electronic spectroscopies28,29) are also used and provide more information about the dynamics of PSCs.

Challenges facing determination of chromophore site energies in PSCs

Although many individual components of photosynthetic machinery are quite well understood, some questions remain unanswered.30 The purpose of this work is not to criticize the published site energies for various PSCs but rather to raise a genuine concern regarding the underlying reasons for conflicting estimations and accuracy of pigment site energies reported in many publications for the CP47 antenna. Therefore, we comment on the uncertainties that may affect the fitted values of various chromophores. For example, sample purity (ie, the presence of contamination) and protein stability (ie, samples could consist of mixtures of intact and somewhat destabilized/damaged proteins). Furthermore, in low-temperature spectroscopy, high-fluence measurements may also modify optical spectra, changing absorption and emission spectra due to unaccounted HB. The latter problem is even more prominent in single photosynthetic complex spectroscopy studies where high-fluence light is routinely used to generate single-complex spectra (see discussions in Refs. 31 and 32). One also needs to be mindful of photochemistry, which can be either reversible or irreversible. For example, in bacterial reaction centers (RCs) (eg, Rhodobacter sphaeroides or Rhodopseudomonas viridis), the relatively weak probing light used to measure absorption spectra can not only significantly bleach the so-called P-band (due to very efficient charge separation) but also electrochromically shift the bands of accessory BChlL and BChlM. Thus, these absorption measurements can significantly modify site energies, and as a result, optical spectra.33

One of the main obstacles in providing a unified description of the structure–function relationship in PSCs, protein energy landscapes, and dynamics of intact, wild-type, and mutated LH antennas is that various techniques are often applied to samples that show significantly different shapes of basic optical spectra, ie, absorption and emission spectra. The latter is a real limitation that raises questions regarding the relevance of pigment site energies extracted from different spectra (vide infra). Another important issue, as discussed in Refs. 3 and 33, is using a proper shape of Jph(ω), which also affects the extracted site energies and population dynamics. For example, it is not clear: (i) whether an experimentally determined spectral density obtained via the delta FLN methodology, within the lowest energy state, is the same for all pigments; (ii) whether the el–ph coupling strength is sufficiently similar for all chromophores; (iii) to what extent inhomogeneous broadening (Γinh) varies from pigment to pigment; (iv) if the protein dielectric constant varies for different binding sites; and (v) how one should determine excitonic domains using a single coupling cutoff value. In fact, all of the above could depend on the protein-binding pocket. Thus, it is feasible to assume that a single solution based on modeling studies alone may not exist and a large number of experimental constraints (not available as of yet) are needed to narrow the possible choices.

Below, we present a case study of the aforementioned issues applied to the CP47 antenna by demonstrating nearly perfect simultaneous fits of multiple spectra with different sets of parameters, raising concerns about the parameters used to describe the excitonic structure of this important photosynthetic antenna complex. The key questions addressed are as follows: (i) why are there conflicting optical spectra reported in the literature for this antenna system? and (ii) is it possible to reconcile the different maxima and shapes of the reported fluorescence spectra? First, we note that this is a distinctly challenging protein to study, as CP47 is more difficult to separate from the photosystem II core complex (PSII-cc) than, for example, the accompanying CP43 antenna.34,35 This is probably why somewhat different optical spectra have been reported over the years, which has led to disagreement about which spectra represent intact CP47 complexes and should they be used in modeling studies.36–38 As a result, there is no agreement as to which chlorophyll (Chl) contribute to the lowest energy exciton state(s) and what their corresponding site energies are.10,11,36

Results and Discussion

Figure 1 shows the arrangement of CP47 Chls in relation to the RC chromophores. This antenna facilitates excitation energy transfer to the PSII RC.11,25,39 Even though CP47 has been extensively studied over the years, the shape of various optical spectra of intact complexes (eg, absorption and emission spectra), extracted site energies (vide infra), and the resulting excitonic structure are still under debate. Typically, low-temperature steady-state absorption, emission, nonresonant HB, and room temperature (RT) CD and LD spectra of CP47 are used in modeling studies utilizing simple excitonic calculations7 and Redfield simulations, taking lifetime broadening effects into account.10,11 In this study, we focus only on Redfield simulations for a straightforward comparison. Our modeling studies described below are based on the 1.9 Å resolution structure of the PSII core (PDB ID: 3WU2) from cyanobacteria.40 While the experimental data are obtained for CP47 from spinach, it is believed that CP47 complexes are similar in PSII from all organisms. Unfortunately, no X-ray data for spinach PSII are available, and structure-based calculations by necessity are based on cyanobacterial structures.

Figure 1

Arrangement of CP47 Chls in relation to the RC pigments within PSII-cc dimer based on the 1.9 Å resolution structure (PDB ID: 3WU2).40 Each CP47 complex contains 16 Chls and 3 carotenoids (in orange). Pigments likely contributing to the lowest exciton states are separately colored and labeled (see text).

Typical absorption, emission, and spectral densities used in modeling studies of CP47

Peak maxima varying from 690 to 695 nm in CP47 fluorescence spectra have been reported,10,11,26,36,37,41–43 where the 695 nm band corresponds to intact CP47 complex.36,37 Thus, the variable peak maxima suggest that complexes with blueshifted emission (near 690–693 nm) must be partly modified. This suggestion is consistent with the results obtained for intentionally damaged CP47 complexes (data not shown). To illustrate the discrepancies mentioned above, Figure 2A shows typical 5 K absorption and emission spectra for CP47 from the literature.36 The emission spectrum in frame A is peaked at ~691 nm and has a full-width at half-maximum (FWHM) of 270 cm−1 (12.9 nm),36 which is in general agreement with the previously published data (~260 cm−1).10,26 Notably, the emission maximum in frame B, for a carefully prepared and treated sample, peaks near 695 nm with the FWHM of just 195 cm−1. The 695 nm emission is in agreement with the data obtained for PSII-cc, which is believed to contain intact CP47 proteins. This suggests, as argued in our previous articles,36,37 that this sample represents the most intact CP47 complex, whereas typical CP47 spectra shown in frame A (and many 77 K spectra reported in Refs. 41–44) are representative of a mixture of intact and destabilized complexes.36,37

Figure 2

5 K fluorescence and absorption spectra of the CP47 complex reported in Refs. 36 and 37. Spectra in frames A and B were measured for partly destabilized and intact samples from spinach.36 Only frame B shows emission spectrum consistent with that observed in thylakoid membranes41,45,46 and PSII-cc.39 Reprinted with permission from Ref. 36. Copyright 2010 American Chemical Society.

It has been suggested that the absence of ~695 nm emission,38 or its decrease,47,48 might be related to the absence of the PsbH protein subunit in the isolated CP47 complexes. However, it is possible that all CP47 complexes possess the PsbH subunit, and rather a subpopulation of complexes could have a weakened (or broken) hydrogen bond (H-bond) between one of the CP47 Chls and the amino acid residues of PsbH. For example, Thr5 of PsbH forms an H-bond with Chl29 of CP47.38,49,50 If this H-bond is broken or weakened in a fraction of CP47 complexes, a blueshift of the Chl29 site energy should occur.51 As a result, variable blueshifted emission spectra should be observed, in agreement with the experimental data. A similar mechanism involving weakened hydrogen bonds for other Chls could explain the absorption and emission spectra shown in Figure 2 (vide infra).

We emphasize again that only the redshifted emission near 695 nm (see Fig. 2B) is clearly observed in PSII membrane fragments41,45,46 and PSII-cc from Thermosynechococcus elongatus,39 which are considered to be intact preparations. Figure 3 shows 5 K PSII-cc emission spectra before and after HB (curves a and b, respectively), while the inset shows 77 K spectra of intact PSII-cc and CP47 (red curve). These data clearly indicate that absorption of the CP47 complex has experienced nonphotochemical HB at 5 K and that the 695 nm emission of CP47 is consistent with that observed for intact PSII-cc samples (for more details, see Ref. 39). Only data in Figure 2B are consistent with Figure 3.

Figure 3

PSII-cc emission spectra of T. elongatus before (curve a) and after (curve b) bleaching of the lowest energy state of the CP47 complex responsible for the 695 nm emission at 5 K. λB = 496.5 nm and both spectra were obtained with λex = 496.5 nm and I = 50 μW cm−2. In the inset is the 77 K emission spectra obtained for PSII-cc from T. elongatus (black curve) and CP47 with a maximum at 695 nm (red curve).37 Reprinted from Ref. 39 (Copyright Springer Science + Business Media Dordrecht 2015) with the permission of Springer.

Spectral densities

Additionally, we point out that different spectral densities, Jph(ω), were used over the years to fit optical spectra of various PSCs—making an otherwise convincing assignment of site energies less tenable. The left frame in Figure 4 shows the single-site fluorescence spectrum calculated using a lognormal J1(ω) compared with ΔFLN spectrum for the CP47 sample characterized by a 695 nm emission band at low temperatures. The right frame compares ω2J(ω) (corresponding to the antisymmetric component of the Fourier–Laplace transform of the energy gap correlation function) for J1(ω) (curve a), J2(ω) (curve b), and the B777 J(ω) (curve c) from Refs. 10 and 11. J2(ω) is a broader lognormal distribution than J1(ω) and is used in modeling studies for higher energy pigments, as J1(ω), measured for the lowest energy state, may not account for the expected fast exciton relaxation dynamics if used for all Chls. Note that ω2J(ω) determines lifetime broadening effects and transition energy shifts. Thus, it is obvious that its shape will affect the extracted site energies. The B777 J(ω) (see curve c in Fig. 4), obtained from FLN data of the B777 subunit of LH1,52 was used in several previous modeling studies of CP47.10,11

Figure 4

(A) Experimental ΔFLN spectrum fit using a lognormal distribution for J(ω). ωc = 40 cm−1 and σ = 0.8, where ωc and σ are the cutoff frequency and standard deviation, respectively, and the Huang–Rhys factor (S) is 1.0. (B) Comparison of ω2J(ω) for J1(ω) from left frame (curve a), J2(ω) (a lognormal distribution with ωc = 80 cm−1, σ = 0.7, and S = 0.65; curve b), and the B777 J(ω) with S = 0.6 (curve c).52

Modeling studies

Disorder is introduced into the diagonal matrix elements (ie, ) by a Monte Carlo approach with normal distributions centered at (n labeling various pigments, ie, n = 1–16) and with FWHM representing Γinh, which can be site dependent or independent. Eigen decomposition of the interaction matrix provides eigen coefficients and eigenvalues (ω). All 16 Chls are tested as the red-most pigment with of 14,500 cm−1, while other pigments are initially set at higher energies. , Γinh, and phonon/vibrational S factors are used as free or fixed parameters (depending on the model) and are optimized simultaneously against the experimental spectra. Since it is not obvious that Jph(ω) measured for pigments contributing to the lowest exciton state is the same for higher energy pigments, the effect of a broader Jph(ω) shape (see Fig. 4B) with a different S factor and/or variable Γinh is also tested for comparison. Calculated spectra are modeled using a newly developed algorithm that can simultaneously fit several experimental spectra at different temperatures, providing constraints on the pigment site energies of interest.

To address the excitonic structure of the CP47 protein, we begin below with the best fits of our experimental steady-state absorption/emission (from Fig. 2B),36 RT CD,53 and our nonresonant HB spectrum (5 K, λB = 496.5 nm).36 All samples were prepared as described previously.36,37 As argued before,36,37 these data, in our opinion, correspond to the most intact CP47 complexes of PSII from spinach studied in our laboratory over the years. In simulations, we used a non-Markovian reduced density matrix theory52 with a Nelder–Mead simplex algorithm for parameter optimization.54 Errors between optimized fits and experimental spectra are reported as the root-mean-square deviation. The Chls within CP47 were divided into the same domains of coupled Chls as in Ref. 11, with coupling matrix elements calculated by the TrEsp method developed by Madjet et al.55 A coupling cutoff (Vc) of 40 cm−1 is used to split the Hamiltonian into the five exciton domains of Ref. 11. For brevity, we discuss only the calculated spectra, exciton composition (focusing on the lowest energy exciton states), and the extracted site energies.

Several different models are considered below. In Model A, we use the experimentally determined Jph(ω) for all pigments and allow Γinh to vary independently for all Chls. Herein, we use Vmn values calculated with the TrEsp method.55 The best fits were obtained for three sets of parameters in Model A, which are referred to below as Models 24A, 26A, and 29A (where the number identifies the lowest energy Chl). The results for Models 24A, 26A, and 29A are shown in Figures 5–7, respectively. This is consistent with our previous extensive searches for realistic low-energy pigments.7 That is, previously published articles suggested that Chl267 or Chl2910,11 are the best candidates to contribute mostly to the lowest energy state, although different shapes of emission/absorption spectra, spectral densities, and levels of theory were used in modeling studies. Left panels in Figures 5–7 show calculated curves as solid lines, and experimental curves are filled. Right panels in Figures 5–7 show the respective first three excitonic states (top) and the major pigment contributions (right bottom panels) to the respective lowest energy (α = 1) state.

Figure 5

Model 24A. Left: best simultaneous fits (lines) of 5 K absorption (blue), emission (red), and low-fluence nonresonant HB (green) spectra compared to experiment (filled curves). Gray curves are experimental and calculated RT CD spectra. Lowest energy Chl24 and Chl29 have of 14,485/152 and 14,710/224 cm−1, respectively. Right: contributions to 5 K absorption of the three lowest energy exciton states (top) and the three pigments that contribute most to the α = 1 exciton state (bottom). Fitting error = 1.15 × 10−3.

Figure 6

Model 26A. The same experimental data as in Figure 5. Lowest energy Chl26 and Chl29 have of 14,501/161 and 14,662/187 cm−1, respectively. Chl26, Chl24, and Chl29 contribute 91%, 2%, and 2% to the lowest energy (α = 1) state, respectively. Fitting error = 1.19 × 10−3.

Figure 7

Model 29A. The same experimental data as in Figures 5 and 6. Lowest energy Chl29 and Chl24 have of 14,484/146 and 14,759/250 cm−1, respectively. Chl29, Chl24, and Chl27 contribute 94%, 2%, and 1% to the lowest energy (α = 1) state, respectively. Fitting error = 1.34 × 10−3.

Figure 5 presents Model 24A, where the lowest energy pigment, Chl24, has . Herein, Chl24, Chl26, and Chl29 contribute to the lowest (α = 1) exciton states 92%, 2%, and 2%, respectively. A virtually perfect fit of the ~695 nm emission spectrum was obtained. Figures 6 and 7 show fits to the same experimental spectra starting from different site energies, assuming that either Chl26 or Chl29, respectively, mostly contribute to the lowest energy trap (as reported in Refs. 7, 10, and 11). Again very good fits can be obtained; in these two cases, the lowest energy chromophores are characterized by of 14,501/161 and 14,484/146 cm−1, respectively. All fit parameters (including and Γinh of all Chls) are summarized in Tables 1 and 2, while the previously reported values from Refs. 10 and 11 are summarized for comparison in Supplementary Table 1.

Table 1

Site energies for Models 24A/B, 26A/B, and 29A/B/C (see text).

CHL	24A	24B	Δ24 (A-B)	26A	26B	Δ26 (A-B)	29A	29B	Δ29 (A-B)	29C	Δ29 (A-C)
11	678.7 (3)	678.2 (3)	0.5	676.2 (4)	676.2 (4)	0.0	674.8 (4)	674.7	0.1	672.9	1.9
12	672.2	673.6	−1.4	675.7	676.3 (3)	−0.7	668.4	668.3	0.0	671.9	−3.5
13	675.9 (4)	676.0 (3)	−0.1	673.4	672.8	0.6	673.2	672.3	0.9	671.5	1.7
14	661.8	660.3	1.6	670.7	671.3	−0.6	669.8	669.1	0.7	667.2	2.6
15	664.8	664.6	0.2	676.4 (3)	676.1	0.3	674.7	673.8	0.9	672.5	2.2
16	674.0	675.5	−1.5	676.1	675.9	0.2	672.6	674.0	−1.4	674.1	−1.5
17	669.0	670.2	−1.2	663.6	663.8	−0.2	663.5	662.7	0.7	662.9	0.6
21	675.3	674.7	0.5	673.2	673.5	−0.4	674.6	675.2 (4)	−0.6	673.2	1.4
22	665.3	664.8	0.5	662.6	662.0	0.7	661.5	661.2	0.3	662.9	−1.4
23	665.2	665.2	−0.1	665.9	665.4	0.5	6 67.9	667.8	0.1	664.7	3.1
24	690.4 (1)	690.4 (1)	0.0	670.0	669.6	0.4	677.6 (2)	677.7 (2)	−0.1	677.9 (2)	−0.3
25	673.5	673.4	0.2	671.4	671.7	−0.3	671.9	671.8	0.0	670.2	1.7
26	6 68.1	668.8	−0.8	689.6 (1)	689.6 (1)	0.0	673.5	675.4 (3)	−1.9	675.6 (3)	−2.2
27	672.6	672.5	0.1	668.4	668.5	−0.2	675.1 (3)	674.8	0.3	674.3 (4)	0.8
28	666.7	665.9	0.8	662.7	662.6	0.1	663.0	662.8	0.2	661.4	1.6
29	679.8 (2)	678.9 (2)	1.0	677.4 (2)	677.3 (2)	0.2	690.4 (1)	690.4 (1)	0.0	690.3 (1)	0.1

Notes: Pigments are labeled according to nomenclature of Loll et al.49 Δ shows the difference between site energies of various models. The red numbers in parentheses indicate energy ordering of the four lowest energy Chls for easy comparison. All values are in units of nanometer.

Table 2

Γinh (FWHM) for Models 24A, 26A, and 29A.

CHL	MODEL 24A	MODEL 26A	MODEL 29A
11	198	234	253
12	273	234	258
13	229	231	230
14	230	220	264
15	249	225	232
16	248	176	263
17	235	216	255
21	154	240	244
22	236	219	256
23	238	223	262
24	152	227	250
25	228	221	255
26	266	153	273
27	196	208	231
28	223	199	262
29	224	169	146

Note: All values are in units of per centimeter.

Recall that in fitting simultaneously all four spectra in Models 24A, 26A, and 29A, we assume that all Chls could have somewhat different Γinh. The fits of the same four experimental spectra as those shown in Figures 5–7, assuming the same Γinh for all pigments except the lowest energy one (referred to as Model B), also provide very good fits when Chl24 (Model 24B), Chl26 (Model 26B), or Chl29 (Model 29B) has the lowest site energy; that is, in Model B, we assume that Γinh is site independent for all but the lowest energy Chl, which is constrained by the width of the fluorescence and HB spectra. The best fits in Models 24B, 26B, and 29B were obtained with Γinh of 221, 218, and 241 cm−1, respectively, for all high-energy Chls. The resulting fits are shown in Supplementary Figures 1–3.

Although the fits for Model B are good, as expected, the site energy differences between Models A and B are on the order of 1 nm (see Table 1 and the Supplementary File). However, the lack of detailed information on Γinh complicates the analysis, as multiple sets of parameters provide good fits to experimental spectra. Thus, we do not assert that the site energies shown in Table 1 are perfect. However, a comparison between site energies obtained for variable and the same Γinh for all but one pigment (vide supra) introduced relatively small changes. For example, the average difference between the between Models 29A and 29B is only 0.5 nm, though the difference ranges from values −1.9 to 0.9 nm. Variations for other models are also listed at the bottom of Table 1.

Next, one could test whether J1(ω) (curve a in Fig. 4) is a proper choice for all Chls. To test this, Model 29C assumes two Jph(ω), curves a and b from Figure 4, ie, J1(ω) measured experimentally for the lowest state (ωc = 40 cm−1, σ = 0.8, and S = 1) and J2(ω) with ωc = 80 cm−1, σ = 0.7, and S = 0.65 for higher energy pigments; that is, J2(ω) is used for all pigments except Chl29. As expected, addition of the broader J2(ω) somewhat compensates for Γinh, on average, down from 241 to 216 cm−1 (for Model 29C). However, we emphasize again that all spectra can be simultaneously well fitted with different sets of site energies, as shown in Table 1. The fits for Model 29C are shown in Figure 8.

Figure 8

Model 29C. Lowest energy Chl29 and Chl24 have of 14,486/147 and 14,751/227 cm−1, respectively. Chl29, Chl24, and Chl26 contribute 96%, 2%, and 1% to the lowest energy (α = 1) state, respectively. The el–ph coupling of Chl29 is described by J1(ω), with all other Chls described by J2(ω). Fitting error = 1.28 × 10−3.

Finally, there are also various methods for calculating the interaction between pigment transition dipoles.9 Although the TrEsp method has been used in both this work and Ref. 10, the more complicated Poisson–TrEsp method, accounting for electrostatic potential of the protein environment, was used to calculate the coupling constants of Ref. 11. On average, Poisson–TrEsp produces values ~0.7 smaller than the standard TrEsp method (see Supplementary Tables 2 and 3). Note that very good fits can also be obtained for the above models (24A/B, 26A/B, and 29A/B) using the smaller coupling constants from Ref. 11, ie, calculated with the Poisson–TrEsp method and Vc = 30 cm−1. Thus, the differences between TrEsp and Poisson–TrEsp are not critical for providing good fits of optical spectra. For comparison, Figure 9 shows optimized fits of Model 29C using the Poisson–TrEsp values11 (Model 29CP), as given in Supplementary Table 3. For brevity, we only mention that similar quality fits can be obtained for models in which Chl24 and Chl26 are the main contributors to the lowest energy state.

Figure 9

Model 29CP. Lowest energy Chl29 and Chl24 have of 14,486/147 and 14,721/188 cm−1, respectively. Chl29, Chl24, and Chl26 contribute 97%, 2%, and 1% to the lowest energy (α = 1) state, respectively. Poisson–TrEsp coupling constants.11 The el–ph coupling of Chl29 is described by J1(ω), with all other Chls described by J2(ω). Fitting error = 1.35 × 10−3.

The data discussed above, as well as data shown in Supplementary Figures 1–3, show that multiple sets of Chl site energies can fit absorption, emission, CD, and nonresonant HB spectra. However, the values of elucidated site energies somewhat depend on assumptions regarding Γinh and shapes of Jph(ω). Interestingly, taking mutational data into account,47,48 one could suggest that Chl29 might indeed be the pigment of choice, as briefly discussed below, but one still cannot exclude that Chl24 and Chl26 are the lowest energy pigments. Thus, the quality of the fits (which appear to be very good for all models), even when multiple spectra are fitted simultaneously, does not guarantee a unique outcome, as a number of parameters discussed above are not well defined.

Hydrogen bonding and mutational studies in CP47

Recent experiments, where spectroscopic studies were performed on CP47 assembly modules containing or lacking the PsbH protein,38 shed more light on the spectral differences reported in the literature for CP47 complexes.26,36,37,41,42 An essential role of PsbH in the origin of the PSII red emission has been recently demonstrated.38 The authors argued that based on the crystal structure,40 PsbH directly interacts with Chl29 of CP47 through an H-bond between Thr5 and the Chl 131-keto group, which could be responsible for the red fluorescence state of CP47. These ideas were also previously discussed in Refs. 7 and 50. Now, if Chl29 is indeed the lowest energy pigment, as also proposed in Refs. 10, 11, 47, and 48, the stronger or weaker H-bond (due to protein destabilization) between Thr5 of the PsbH subunit and Chl29 (at least in a subpopulation of CP47 complexes) could explain the variable emission spectra reported in the literature. We test below whether or not it is feasible to assume that low-temperature spectra (5–77 K) reported and modeled in the past10,11,56 could, as suggested by us in Refs. 36 and 37, correspond to mixtures of intact and destabilized complexes.

Do the typically reported and modeled fluorescence spectra with FWHM of ~260–270 cm−1 correspond to a mixture of intact and destabilized CP47 complexes?

To test this suggestion, we use the parameters obtained from the fit of four optical spectra shown in Figure 8 with one exception; the site energy of Chl29 is shifted to higher energy, assuming that in samples with a contribution from destabilized proteins the H-bond with PsbH could be broken or significantly weakened. The latter is very feasible, though other minor changes could also occur. For example, de Weerd et al42 observed the 1633 cm−1 mode in FLN spectra, indicative of an extremely strong H-bond between the 131-carbonyl group and the protein, which (at least in a subpopulation of complexes) could produce the redshift of one of the pigments contributing to the lowest energy state. In our 2010 article,7 we questioned this interpretation, as in 2004 the above assignment was reversed by the same group,44 as they did not find any strong evidence for a 1633 cm−1 mode. However, it is very likely that both sets of experimental data were correct, but the experiments were done on more intact and less intact CP47 samples. This, in turn, would suggest that CP47 samples can be a mixture of subpopulations in which the H-bond from the PsbH subunit to Chl29 (or other low-energy pigment) is strong or significantly weakened, though other minor changes to site energies of other pigments cannot be entirely excluded.

The data for a mixture of intact/destabilized CP47 complexes (with Chl29 being the lowest energy pigment in both subpopulations) are shown in Figure 10. The top spectra are identical to those shown in Figure 8 (for intact CP47, Model 29C) for easy comparison of shifts. The lower two panels are for the mixture of intact/destabilized sample (Model 29CM, where M indicates a mixture model). The CD spectrum is not shown in Figure 10, as the calculated spectrum is nearly identical to that shown in Figure 8 for the intact CP47 complex. Figure 10, which shows reasonable fits of the often reported spectra without any optimization, reemphasizes our previous conclusion that the broad (FWHM ~260–270 cm−1) and blueshifted fluorescence spectra may indeed correspond to a mixture of intact and destabilized complexes; that is, all what was required to obtain good fits in Figure 10 was a 113 cm−1 shift of the site energy of Chl29 for the destabilized subpopulation and assuming at the same time a 40%/60% mixture of intact/destabilized contributions. Note that applying the optimization algorithm for site energies of higher energy Chls would improve the fits of absorption and HB spectra (data not shown). Of course, a different composition of intact and destabilized proteins will change the resulting emission spectrum in agreement with other experimental data, where often the maxima (for partly destabilized CP47 complexes) were reported to lie between 690 and 693 nm.36,37,39 That is, samples with emission near 690 nm would contain mostly destabilized complexes (data not shown for brevity). Interestingly, fits similar to those shown in Figure 10 can also be obtained for Models 29AM and 29BM, as illustrated in Supplementary Figures 4 and 5.

Figure 10

Filled spectra in the top and bottom frames are the experimental absorption, emission, and HB spectra for intact CP47 (Fig. 2B) and intact/destabilized CP47 complexes (Fig. 2A), respectively. All calculated spectra for Models 29C (top) and 29CM (bottom) are given by solid lines. The three spectra in the bottom panel can be well described, assuming a mixture of intact/destabilized (40%/60%) proteins. The only difference between intact and destabilized subpopulations is the site energy of Chl29 (see text).

Furthermore, at this point, it cannot be excluded that Chl26 could be the lowest energy Chl. In this case, one would have to assume that a putative H-bond between His9 of CP47 and the ester group of Chl26 exists and could be significantly weakened due to protein conformational changes. A similar scenario could also be feasible assuming that Chl24 contributes mostly to the lowest energy state, with a putative H-bond of the Chl24 ester group with either Trp468 or His9 of CP47 (in the latter case, based on Ref. 50, this could be an H-bond from the Nδ -atom). Data for the abovementioned scenarios involving Chl24 and Chl26 as the lowest energy pigments, though not shown for brevity, could also fit all the spectra in question. Although the mixture models with Chl26 or Chl24 as the lowest energy pigments could also be consistent with experimental data, they should be interpreted cautiously, as no strong H-bonds were found for these pigments in the structures in the studies by Loll et al49 and Guskov et al.57 Thus, there are no authoritative experimental or simulation data to provide just one solution for the lowest energy pigment(s), although based on mutational studies47,48 it is likely that the best candidate for the lowest energy pigment could be Chl29. However, the lowest energy state of the modified subpopulations maybe localized on different pigments than the intact subpopulation, which could provide the abovementioned weakly positive CD signal at ~690 nm. Importantly, our simulation results clearly support our earlier suggestion that many previously reported optical spectra and their simulation data were obtained for mixtures of intact and destabilized proteins, questioning the validity of parameters reported in the literature.7,10,11

Returning to Figure 10, we note that shifting the site energy of Chl29 in Model 29CM (see above), or Models 29AM and 29BM (see Supplementary Figs. 4 and 5), leads to very small changes in the absorption and CD spectra, while significant changes occur in fluorescence and HB spectra, in very good agreement with experimental data.36 Thus, our data provide more insight into possible composition of previously published fluorescence spectra (with maxima between 690 and 693 nm),36,47,58 which revealed variable contributions from ~695 nm emission,36,37 as well as of HB spectra. The mixture model is consistent with the measured fluorescence spectra and the observed (sample-dependent) shift in the nonresonant HB spectra and supports our previous conclusions.36 This is why the width (FWHM) of various steady-state 77 K fluorescence spectra reported in the literature also varied from 16 to 24 nm, while the FWHM of fluorescence spectrum of intact CP47 complex at 77 K is only 12 nm.36,37 In general, the FWHM of various ~5 K emission spectra are by a factor of ~1.3–1.7 larger than our emission spectrum obtained for the intact sample, which peaks at 694.8 nm and has a width of ~9.5 nm.36,37 We hasten to add that the temperature dependence of various emission spectra can also be explained by our mixture model, but the results will be reported elsewhere.

Finally, we comment on parameters elucidated in Refs. 10 and 11 from the fits of 77 K spectra. Figure 11 shows calculated spectra using all parameters of Ref. 11, with Chl29 as the lowest energy Chl (that is, Vnm, Jph(ω), S, and Γinh) compared with experimental data of Figure 2, as well as the corresponding HB spectra36 and the RT CD spectrum.53 Experimental data are shown in black, while calculated curves are shown in red. Note that the parameter set was optimized not to our experimental data shown in Ref. 36, but to various spectra taken from the literature.26,42,53 It is obvious that parameters reported in Refs. 10 and 11 do not fit spectra shown in Figure 2, although the fit to CD is reasonable as this spectrum was also included in fits reported in Ref. 11. We admit that comparison of calculated parameters obtained from fits of different experimental spectra does not provide any insight, though it reveals again that samples of variable intact/destabilized composition were studied by various laboratories. This is also supported by the fact that parameters set from Refs. 10 and 11 do not fit emission spectra reported in Ref. 42 (not shown for brevity). Clearly, the results of this work have shown that until the shapes of standard spectra for the intact CP47 complex are established, spectral densities, inhomogeneous disorder, etc. are further constrained by experimental data, and ambiguous results can be obtained, in agreement with the main conclusions of this work.

Figure 11

Experimental (black) and simulated (red) 5 K absorption (a/a′), fluorescence (b/b′), nonresonant HB (c/c′), and RT CD (d/d′) spectra. (A) and (B) correspond to the experimental spectra of Figure 2A and B, respectively (for details, see Ref. 36). The CD spectrum is the same for both frames.53 The red curves are calculated using the parameters of Ref. 11 (see Supplementary Table 3 and text in the Supplementary File) and are the same in both frames.

Possible implications of our findings on the structure–function relationship of Chls in the CP47 complex

In light of the data presented above, we think that the analysis of the structure–function relationship in CP47 must await further work and confirmation of our mixture model(s). We still believe that even if Chl29 is the lowest energy pigment (with a localized excited state) it may not be important for excitation energy transfer to the RC. That is, Chl29 (in proximity to CarD2 and ChlZD2) could dissipate excess energy under conditions of high light intensity.11 In the case where Chl24 or Chl26, located in the vicinity of the RC, contribute mostly to the lowest energy state, they could efficiently transfer energy to the RC. Of course, the assignment of low-energy pigments and the connection of these low-temperature assignments to function in vivo will depend on the shape of the low-temperature spectra being fit.

Concluding Remarks

In light of the difficulties and issues surrounding the extraction of pigment site energies discussed in this work, we believe that more insight into possible sets of site energies for the most intact and intact/destabilized CP47 samples was provided via computerized optimization cycles while simultaneously fitting absorption, emission, CD, and HB spectra. The assignment of site energies differs significantly from those reported recently by Shibata et al11 for CP47 obtained by modeling 77 K spectra,26,42 highlighting the complications involved in comparing the results from various modeling studies, especially, when different types of spectra are simulated using different Jph(ω). Although, as shown above, assigning the lowest site energy to Chl24, Chl26, or Chl29 can fit the data very well, including both intact and mixtures of intact/destabilized complexes. Although it is likely (considering all mutagenesis data)47,48 that Chl29, with an H-bond to the PsbH subunit, might be the best candidate as the major contributor to the lowest energy trap in intact CP47 complex, other assignments (eg, with Chl24 or Chl26 as the lowest energy pigments) cannot be entirely excluded. Very recent modeling of fluorescence kinetics of PSII in Ref. 56 also suggested that assigning Chl26 as the reddest absorbing Chl cannot be entirely excluded.

We conclude that spectra simulated in Refs. 10 and 11, as well as many spectra reported in the literature, likely correspond to a mixture of intact and destabilized complexes, where a H-bond between one of the Chls and the protein environment could be altered, though other reasons for the changes in Chl site energies cannot be excluded at this time, such as Chl conformations and the presence of nearby polarizable molecules. Thus, if Chl29 is indeed the lowest energy pigment, then the H-bond between its 131-keto group and Thr5 of PsbH could be broken or weakened, as indicated by very good fits of spectra corresponding to a mixture of intact/destabilized complexes (Fig. 10 and Supplementary Figs. 5 and 6).

In addition, it must be confirmed that the extremely weak positive CD spectrum near 692 nm observed in PSII core from spinach (at low-T)59 is also observed in isolated, intact CP47 complexes. The latter is critically important as none of the site energies reported so far in the literature,10,11 as well as parameters reported in this article (though excellent fits were obtained while simultaneously fitting multiple spectra), could reveal a positive CD peak near 692 nm within the 5–300 K temperature range. If the lowest exciton state in CP47 is delocalized, the CD signal should be strong. Thus, more reliable low-T CD spectra are needed to address the nature of the low-energy excited state(s) in the CP47 complex. Herein, CPL, successfully obtained recently for the CP43 protein complex,24 could also provide some insight into the composition of the lowest exciton state in the CP47 antenna. Thus, more experimental constraints are needed to reduce the variability in the reported Chl site energies and/or reveal new sets of parameters if necessary. In particular, the uncertainties for higher energy pigments are much larger, which in turn could affect the energy transfer rates/pathways. A final set of parameters must be able to describe all frequency- and time-domain data generated for the same well-characterized, intact protein complexes, and to the best of our knowledge, this has not been done as of yet. Our preliminary simulation data indicate that a different composition of the lowest exciton state might be necessary to describe positive CD bands near 692 nm. However, the latter must be demonstrated for intact isolated complexes, whose spectra are shown in Figure 2B.

Finally, more efforts are needed to theoretically explore spectral densities for pigments residing in different protein environments.60–62 Furthermore, efforts should be made to determine if Jph(ω) shapes could be elucidated from measurements of the average lifetimes of high-energy exciton states. The latter, if possible, would definitely improve the theoretical description of various linear and nonlinear optical spectra of many PSCs. This is very important, as small differences in the shape of Jph(ω) (in particular, the high-frequency spectral region) can lead to large changes in exciton relaxation. Therefore, Jph(ω) should be measured for each pigment–protein complex of interest. Finally, we emphasize that Jph(ω), via ΔFLN or ΔHB spectroscopies, reveal spectral density and el–ph coupling parameters only for pigments contributing to the lowest energy exciton state.

In summary, our findings for the spectra reported in this article can be stated as follows: (i) the 695 nm emission band (with FWHM ~195 cm−1) is the likely emission of intact CP47 at both 5 and 77 K, as such emission is also observable at both temperatures in PSII-cc39,41 and thylakoid membranes;41,45,46 (ii) blueshifted emission often observed near 690–693 nm must contain subpopulations of intact and destabilized complexes, and thus, an emission band with FWHM ~260–270 cm−1 (at ~5 K) peaking at 690–693 nm likely indicates a mixture of intact and destabilized CP47 proteins; and (iii) although simultaneous fits of 5 K absorption, emission, nonresonant holes, and RT CD spectra provide reasonable fits when Chl24, Chl26, or Chl29 contributes the most to the lowest energy exciton state, it cannot be excluded that the best candidate might be Chl29, as also suggested in Refs. 10, 11, 38, 47, and 48. The latter suggestion is based on fits of the optical spectra simulated in this article. If new CD and CPL spectra are obtained for destabilized and, in particular, intact, isolated CP47, new simultaneous fits should be completed, as the positive feature near 692–693 nm (if present, vide supra) could change the identity of low-energy pigments as mentioned above. For example, Chl11 and Chl14, based on simple Monte Carlo simulations reported in Ref. 7 (see Fig. 3), could also contribute to the lowest energy state.

We suggest that in order to obtain a final set of parameters, which can describe both the excitonic structure and dynamics of CP47 complexes, all laboratories should use stringent analytical tools to characterize the quality of extracted complexes. Standards for low-temperature spectra have to be established, and such spectra should be routinely shown in publications or supporting/supplementary information for easy comparison. In addition, various experimental techniques should be applied to samples from the same batch. For example, it is not advisable to simultaneously fit spectra from various publications, as sometimes done in the literature. If the complexes measured in different laboratories at 77 and 5 K constituted different mixtures, one set of parameters will never simultaneously describe both sets of data (see Supplementary Fig. 6 and the corresponding discussion above). Regarding time-domain data, it is often not clear to what extent the samples were composed from a mixture of intact and destabilized complexes, as low-temperature absorption and emission spectra are not reported. Finally, in light of the data and simulation studies for CP47 complexes discussed in this work, it appears that the modeling of data obtained for the entire PSII core11 may be even less reliable, as the site energies of CP43 and RC pigments are also not well established. Nevertheless, modeling and optical spectra for PSII core (and its components) should continue as proper assignment of pigment site energies is critical to their function.

Supplementary figure 1. Spectral fits for Model 24B.

Supplementary figure 2. Spectral fits for Model 26B.

Supplementary figure 3. Spectral fits for Model 29B.

Supplementary figure 4. Simulated spectra for Model 29AM.

Supplementary figure 5. Simulated spectra for Model 29BM.

Supplementary figure 6. Simulated spectra for Model 29CPM.

Supplementary table 1. Comparison of CP47 site energies of Renger and coworkers.

Supplementary table 2. TrEsp electronic coupling constants used in this work.

Supplementary table 3. Poisson-TrEsp electronic coupling constants used by Shibata et al.