Quantitative assessment of chlorophyll types in cryo-EM maps of photosystem I acclimated to far-red light

Highlights

• •A method of cryo-EM map analysis to determine chlorophyll substituents is developed.
• •An improved molecular structure of far-red light photosystem I is presented.
• •Cryo-EM maps of far-red light photosystem I provide direct evidence for chlorophyll f.
• •Chlorophyll f-binding sites are predicted in far-red light-acclimated photosystem II.

Abstract

Chlorophyll cofactors are vital for the metabolism of photosynthetic organisms. Cryo-electron microscopy (cryo-EM) has been used to elucidate molecular structures of pigment-protein complexes, but the minor structural differences between multiple types of chlorophylls make them difficult to distinguish in cryo-EM maps. This is exemplified by inconsistencies in the assignments of chlorophyll f molecules in structures of photosystem I acclimated to far-red light (FRL-PSI). A quantitative assessment of chlorophyll substituents in cryo-EM maps was used to identify chlorophyll f-binding sites in structures of FRL-PSI from two cyanobacteria. The two cryo-EM maps provide direct evidence for chlorophyll f-binding at two and three binding sites, respectively, and three more sites in each structure exhibit strong indirect evidence for chlorophyll f-binding. Common themes in chlorophyll f-binding are described that clarify the current understanding of the molecular basis for FRL photoacclimation in photosystems.

Introduction

An important goal of structural biology is to obtain molecular models that elucidate the structural and functional relationships of biomolecules in the context of the physiology of an organism. Most macromolecules contain cofactors that are integral to their specific function, and thus careful assignment of these cofactors is of utmost importance in gleaning functional attributes. This is exemplified in structures of photosystem I (PSI) and photosystem II (PSII), the cofactor-rich reaction center complexes central to oxygenic photosynthesis. These membrane proteins absorb light by photoexcitation of chlorophylls, ultimately leading to special pair photooxidation that derives reducing equivalents to power metabolic processes.

Approximately 25% of the molecular mass of PSI and PSII corresponds to chlorophyll (Chl) molecules. There are five known types of Chls – Chls a, b, c, d, and f—in oxygen-evolving phototrophs [1,2]. The differences in molecular structure among the Chls are relatively minor but result in important differences in electronic properties that confer niche-specific biological fitness. Chl a is found in all known oxygenic phototrophs and is the most common and well-studied of all the Chls in oxygenic photosynthesis. In most cases, the other Chl types serve to extend light absorption into wavelengths that are enriched in specific environments. Chls b and c preferentially absorb blue light [2]; the former is found in plants, algae and some cyanobacteria, and the latter is found in some marine diatoms and dinoflagellates. Chls d and f preferentially absorb far-red light (FRL) and are only found in some cyanobacteria [3].

Prior to the “resolution revolution” in cryo-EM [4], only photosystem structures containing solely Chl a, or a mixture of Chl a and b had been solved, and only relatively few examples of structures containing Chl b were known. Since then, however, many cryo-EM structures have been reported with multiple Chl types, such as giant PSI and PSII supercomplexes containing hundreds of Chls a and b, or Chls a and c, respectively, and cyanobacterial PSI structures containing both Chl a and Chl f [5]. Elucidating the number of various Chl types and their specific sites in the photosystem structures is critical for disentangling energy and electron transfer mechanisms, but their minor structural differences make them difficult to distinguish at commonly achieved resolutions [6]. Therefore, it is vital to develop tools that assist in the accurate assignment of cofactors in cryo-EM maps as new and unique Chl-containing molecular structures continue to be discovered.

Chl f can be found in some cyanobacteria that have evolved to absorb the far-red part of the solar spectrum, a mechanism called FRL photoacclimation (FaRLiP) [7]. FaRLiP is currently the subject of great research interest due to its potential application to modifying crops [8,9]; therefore, understanding the molecular basis of FaRLiP is highly desirable. During FaRLiP, alternative genes for six PSI subunits are expressed which allows the binding of Chl f in addition to Chl a [7]. FRL-PSI is thought to contain 6-8 Chl f molecules per monomer [10], [11], [12], possibly exhibiting some species-specific variation in the number of Chl f molecules bound per PSI [6,13]. The locations of Chl f in FRL-PSI are thought to be specific in order to avoid the creation of energy traps, which could ultimately result in energy dissipation rather than productive charge separation. Some Chl f sites in FRL-PSI had been suggested on the basis of spectroscopic observations [11]; however, they are yet to be confirmed by more direct evidence from molecular structures of FRL-PSI.

Chl a and Chl f differ only by the C2 substituent on their tetrapyrrole ring which may be challenging to differentiate in cryo-EM maps: Chl a has a methyl moiety and Chl f has a formyl moiety (Supplementary Fig. 1). Two structures of FRL-PSI determined by cryo-EM have recently been reported [10,14]. Gisriel et al. determined the 3.19-Å global resolution cryo-EM structure of FRL-PSI from Fischerella thermalis PCC 7521 (hereafter F. thermalis) [10]. This resolution is insufficient to differentiate a formyl moiety from a methyl group, and thus Chl f from Chl a; therefore, the authors tentatively assigned Chl f sites based on possible H-bond donors to the C2 substituent of prospective Chl f sites that are specific to subunits expressed in FRL. Kato et al. determined the cryo-EM structure of FRL-PSI from Halomicronema hongdechloris at a higher global resolution of 2.41 Å [14]. The authors assigned seven Chl f sites, but most were called into question upon subsequent analysis due to unlikely chemical environments for the formyl substituents and questionable assignment strategies [6]. Due to the challenges that accompany direct observation of formyl substituents of Chl f, especially in cryo-EM maps, it was suggested that quantitative methods should be developed to distinguish Chl a from Chl f accurately [6].

Here, a cone scan method is introduced for the assessment of Chl substituents that is verified by density functional theory (DFT) calculations and practically applied to FRL-PSI structures from H. hongdechloris [14] and F. thermalis to identify Chl f-binding sites, the latter of which was refined to higher resolution than previously reported [10]. Common themes of Chl f-binding allow one to predict Chl f-binding sites in FRL-PSII for which there is not yet a molecular structure. These findings exemplify the importance of quantitative cofactor assignment in cryo-EM maps and provide a roadmap for elucidating the molecular basis of FaRLiP in cyanobacteria.

Results

Development of a quantitative method to assess electrostatic potential corresponding to chlorophyll substituents

Cryo-EM experiments yield 3-dimensional maps of the ensemble electrostatic potential (ESP) of molecules. Calculations of DFT-derived ESP maps were combined with DFT energy-minimized Chl models to establish a method for the quantitative analysis of experimental cryo-EM maps to identify Chl types. DFT energy-minimized structures of Chl f and Chl a, and a DFT-derived ESP map of Chl f (see Methods and Supplementary Fig. 2) were generated. The ESP map was smoothed at B-factors corresponding to 2.41-Å and 2.96-Å resolution (Fig. 1a and 1b). For a given substituent, scan parameters (Supplementary Table 1) were used to define the error-free ESP profile associated with its position in the model (Fig. 1d). For example, the ethyl substituent at C8 was characterized by scanning the calculated ESP in a cone about the C8-C8¹ bond axis of the model at an angle of 109.5° at a distance of 1.57 Å. A peak in the ESP cone scan is exhibited that corresponds to the orientation of the substituent. We show such cone scans for the C8 ethyl, C7 methyl, C3 vinyl, and C2 formyl substituents in Fig. 1d. These theoretical ESP cone scan profiles provide a basis for interpreting cone scans performed on experimental ESP maps.

Fig. 1.

Calculated ESP and free energy of Chl f. Panels a and b show calculated DFT-derived ESP maps for Chl f at 2.41 Å and 2.96 Å, respectively, which correspond to the resolutions of the two experimental maps analyzed below. Notable substituents of interest are labeled. The calculated ESP on van der Waals surfaces for Chl a and Chl f are shown in Supplementary Fig. 2. Panel c uses the formyl substituent of a Chl f to define angle assignments used herein. Panel d shows the ESP cone scans corresponding to the cones shown in a and b with black and red lines, respectively, for a Chl f model within the calculated ESP map for each substituent of interest using their respective parameters (Supplementary Table 1). Notable peaks are labelled.

Cone scans of Chl substituents

The formyl C2 substituent of Chl f is expected to have increased ESP compared to that of the methyl C2 substituent of Chl a (Supplementary Fig. 2). To identify which Chl sites exhibit direct evidence for binding Chl f in the cryo-EM maps, we compared cone scans of the C2 position, which could be methyl (Chl a) or formyl (Chl f), to a cone scan distribution derived from the C7 position, which is always methyl (Fig. 2 and Supplementary Fig. 3). These cone scans were generated from all Chl sites in the 2.41-Å resolution structure of FRL-PSI from H. hongdechloris (Protein Data Bank [PDB] 6KMX), and the new structure of FRL-PSI from F. thermalis that we solved to an improved global resolution of 2.96 Å (PDB 7LX0). Data statistics (Supplementary Table 2), data processing details (Supplementary Fig. 4), and the local resolution map (Supplementary Fig. 5) for the F. thermalis map and structure can be found in the supplemental material. We also validated the cone scan method (Supplementary Text 1) by calculating free energy profiles for Chl a and Chl f substituents as a function of torsion angle (Supplementary Fig. 6), and comparing them to cone scans of the C8, C7, and C3 substituents (Supplementary Fig. 7).

Fig. 2.

Chl f candidates identified by the cone scan method. a shows the three local resolution bins for H. hongdechloris, and b shows the two local resolution bins for F. thermalis. The global resolution, local resolution bins, and the number of Chl sites per bin are labeled. For each panel, the gray area corresponds to signals within the $\mu \pm 3\sigma$ boundaries derived from the methyl distribution (Supplementary Fig. 3). High-probability angles, calculated based on DFT free energy profiles (Supplementary Fig. 6), are highlighted in green. Gray lines are C2 cone scans that are similar to the methyl distribution. Red lines are C2 cone scans that are significantly different from the methyl distribution with extra ESP within high-probability angles. Orange lines are C2 cone scans that are significantly different from the methyl distribution with extra ESP but not within high-probability angles.

We describe the methyl cone scan distribution using a normal distribution, $\mathcal{N}(\mu ,{\sigma }^2)$, where mean $\mu$ and standard deviation $\sigma$ are both derived from cone scans of the C7 position. We refer to the $\mathcal{N}(\mu ,{\sigma }^2)$ distribution hereafter as “methyl distribution”. Then, we conducted a significance test on each C2 scan with a null hypothesis that the C2 scan also belongs to the $\mathcal{N}(\mu ,{\sigma }^2)$ distribution and exhibits a similar magnitude of ESP signal. If a C2 scan is much larger than the mean $\mu$, we reject the null hypothesis and accept the alternative hypothesis, which is that the C2 scan shows significantly larger ESP signal. This may indicate the presence of a formyl substituent or density from another molecule in proximity to C2, as addressed in the Discussion. We consider a C2 signal to be significantly different from the methyl distribution when it exceeds $\mu +3\sigma$, corresponding to a p-value of < 0.002, meaning that the probability of these candidates having the same $\mathcal{N}(\mu ,{\sigma }^2)$ distribution is less than 0.2%. Because the $\sigma$ changes as a function of resolution, as discussed in Supplementary Text 1 (see also Supplementary Fig. 7), we maintained the same local resolution bins as those used for the C8, C7, and C3 cone scans.

Because Chl sites are numbered differently among different PSI structures due to PDB processing, we normalize this discussion to the nomenclature that was first introduced by Jordan et al. [15] and provide a table for easy conversion (Supplementary Table 3). For the C2 cone scans from H. hongdechloris, seven have energy profiles that are significantly different from the corresponding methyl distributions. Because the orientations of Chl substituents tend to favor those identified as being low energy in our DFT free energy profiles (Supplementary Text 1 and Supplementary Fig. 6), the most convincing of these candidates are those whose additional ESP is within high-probability angles (highlighted in Fig. 2a) based on the Maxwell-Boltzmann statistics calculated using the DFT energy profiles (Supplementary Fig. 6). These Chl sites are B7, B37, A8 and A23 (red in Fig. 2a). At 2.27- and 2.25-Å local resolution, B37 and B7 are both found in the highest local resolution bin. At 2.42-Å local resolution, A23 has one of the highest local resolutions in the middle local resolution bin. A8, however, is found in the lowest resolution bin at 2.72-Å. The remaining three Chl sites whose cone scans exhibit peaks at unlikely angles are A17, B9, and A10 found at local resolutions of 2.44-, 2.64-, and 2.67-Å, respectively.

For the C2 cone scans from F. thermalis, four have energy profiles that are significantly different from the corresponding methyl distribution. Two have additional ESP present at high-confidence angles; sites B7 and B30 (Fig. 2b). At local resolutions of 2.98- and 3.02-Å, respectively, these sites are nearly identical in their local resolution but because we have set the cutoff at 3.0 Å for binning, B7 is found in the higher resolution bin, and B30 is found in the lower resolution bin. The cone scans of A10 and A38 are different than the methyl distribution but with additional ESP at improbable angles for a formyl oxygen. A10 and A38 exhibit local resolutions of 3.06- and 3.01-Å, respectively, and are therefore found in the lower resolution bin. Though our cone scans suggest that all sites exhibiting increased ESP compared to the methyl distribution are good candidate sites for being highly occupied by Chl f, a careful analysis of the ESP map and chemical environment must also agree with their assignment [6].

Discussion

Data considerations and limitations

Of the ~90 Chl sites in FRL-PSI, 6-8 are expected to be Chl f [10], [11], [12], which exhibit a formyl moiety at the C2 position; the remaining 82-84 molecules should be Chl a, which have a methyl moiety at the C2 position (Supplementary Fig. 1). A lack of evidence for a formyl substituent in our cone scans does not exclude the possibility of a site binding Chl f. Our analysis suggests that the resolution cutoff to observe any given Chl substituent is map-dependent, even when the substituent is fully occupied without orientation variation. Such map-dependent resolution cutoffs can be estimated in our cone scan method by using the methyl distribution and resolution bins, but Chl f sites have the added complication that the total occupancy of Chl f versus Chl a may differ among and between sites. Sites that strongly bind a single Chl type likely evolved as a result of energy transfer constraints. For example, placement of Chl f in a site of too low energy may result in futile energy dissipation, and placement of Chl a in some sites might create an energy barrier too high for energy transfer from Chl f, though efficient uphill energy transfer from Chl f to Chl a has been observed [12,13,16]. It is possible that some sites are more promiscuous, and thus exhibit lower occupancy of Chl f, as has been demonstrated by Chl f-binding to the PSI complexes of Synechococcus sp. PCC 7002, which normally only bind Chl a [12,17]. This may explain why only a few Chl f sites are identified by our cone scans in the two structures – the resolution requirement for resolving a site with a low occupancy of Chl f would be much higher than a site with high occupancy. An additional consideration is that if a site binds Chl f with high occupancy but the formyl substituent exhibits alternate conformations (e.g., 0° and 180°), as is exemplified by Chl site B30 in the FRL-PSI structure from F. thermalis discussed below, it would also increase the resolution requirement to resolve the formyl moiety. Thus, the resolution requirement for obtaining direct evidence for Chl f in the cryo-EM maps is not only map-dependent but site-dependent as well.

It should also be noted that the methyl distribution is not ideally robust. At the C7 position, all Chl sites exhibit a methyl moiety, but there are still some C7 cone scans that peak outside of $\mu +3\sigma$ derived from the methyl distribution (Supplementary Fig. 3). It is likely, then, that a methyl distribution comprised of more methyl substituents in an environment similar to the C2 position of Chl a would increase the robustness of the distributions used for C2 comparison. We reiterate that the standard deviation for a group of cone scans is resolution-dependent; therefore, highly robust data would be achieved by collecting more Chl site data with more resolution bins. This may be achieved in the future upon the elucidation of higher resolution benchmark structures such as photosystem supercomplexes that contain hundreds of Chl sites, especially those in which the tetrapyrrole rings can be placed with higher confidence.

Despite these considerations, the most confident Chl f site assignments should be those whose C2 cone scans are significantly different from the methyl distribution with extra ESP at probable angles and whose formyl substituents exhibit obvious H-bond donors. We proposed previously that the formyl substituent of Chl f is likely to exhibit H-bonding to its formyl substituent [6], similar to Chl b in light-harvesting complex structures. We can expand upon this chemical observation by noting that, whereas ~70% of Chl a sites are coordinated by His sidechains in PSI and PSII structures that contain only Chl a, Chl b is never coordinated by a His sidechain in any known molecular structure (Supplementary Table 4), consistent with previous observations [18]. This is presumably because the imidazole sidechain of His is not a sufficiently strong Lewis base to displace axially coordinated waters [19], and this may also be the case for Chl f. With the addition of this chemical observation, the improved map for F. thermalis FRL-PSI, and the cone scan data presented herein, we have increased the lines of evidence for assigning Chl f-binding sites.

Probable Chl f-binding sites

B7. Chl site B7 in both structures exhibits a C2 cone scan with extra ESP at ~325° compared to the methyl distribution and is in the highest resolution bin, confidently suggesting direct evidence for Chl f-binding in both structures. This is supported by the chemical environment also being suitable for a formyl substituent at C2 in both structures [6]. In the H. hongdechloris structure where the resolution allows for placement of water molecules, the formyl substituent is seen as accepting a H-bond from a water that is H-bonded to a conserved, FRL-specific Tyr residue from PsaI2 [14]. Like Chl b sites in light-harvesting complex structures, Chl B7 is not coordinated by a His sidechain, which may be further evidence that this site binds Chl f at high occupancy. Though the ESP peak maxima are just outside the expected high probability orientation based on our calculated free energy profile for the formyl moiety of Chl f, much of the extra ESP is present in high probability orientations; therefore, it is clear that the protein environment stabilizes the ~325° orientation preferentially.

B37. In both structures, the C2 substituent of B37 is located near a FRL-specific loop of PsaB2 that may donate a H-bond from one of two possible backbone amides to the formyl substituent if Chl f is bound (Supplementary Fig. 8) and the axial ligand to this site is a water molecule, both of which are evidence that this site may bind Chl f. In H. hongdechloris, the C2 cone scan of B37 has a single peak above $\mu +3\sigma$ derived from the methyl distribution centered at ~180°, an orientation consistent with the calculated low free energy orientations, providing direct evidence for binding Chl f. Conversely, in F. thermalis, the C2 cone scan of B37 does not exhibit ESP above $\mu +3\sigma$ derived from the methyl distribution; however, it does exhibit two peaks of approximately equal height at ~0° and 180°. Unlike in H. hongdechloris, the C3 vinyl substituent is modeled toward C2 rather than away (Supplementary Fig. 8) which, based on our free energy calculations (Supplementary Text 1 and Supplementary Fig. 6), would be expected to impact the energy of the C2 moiety if it is formyl. It is possible that the C3 vinyl substituent being oriented toward C2 results in two alternate conformations of the C2 formyl substituent, which would increase the resolution requirement for its direct observation in the ESP map. This hypothesis is supported by the strong indirect evidence for Chl f-binding at B37: the non-His axial ligation and FRL-specific H-bond donors. Thus, although the C2 cone scan of B37 from the F. thermalis structure does not exhibit ESP signal greater than $\mu +3\sigma$ derived from the methyl distribution, it still exhibits compelling evidence for binding Chl f.

A23. Chl f occupancy in the A23 site from H. hongdechloris was suggested to be species-specific compared to F. thermalis because there is a possible H-bond donor nearby the C2 position in for former but not the latter [6]. The A23 site is axially coordinated by a water molecule, also suggesting it could bind Chl f. The C2 cone scan of A23 in H. hongdechloris exhibits a peak above $\mu +3\sigma$ derived from the methyl distribution at ~0°, consistent with its assignment as Chl f, but no significant peak is observed in the C2 cone scan of A23 in F. thermalis, supporting its identity as a species-specific Chl f-binding site. An important observation for the C2 cone scan of A23 in H. hongdechloris is that, of the peaks confidently identified as arising from formyl moieties in the H. hongdechloris cone scans, it has the lowest local resolution at 2.42 Å. For the H. hongdechloris ESP map, this may set a lower limit on the resolution required for direct evidence for Chl f-binding.

B30. B30 in F. thermalis was previously suggested to bind Chl f based upon a possible H-bond to its C2 formyl moiety from a FRL-specific Tyr residue in PsaJ2 [10]. The F. thermalis cone scans of B30 presented here suggest that the C3 vinyl substituent is oriented toward the C2 position and the C2 cone scan exhibits a two-peak profile with maxima at ~5° and 195° (Supplementary Fig. 9), similar to that observed at site B37 in F. thermalis. Unlike site B37, both peaks in the C2 cone scan for B30 are above $\mu +3\sigma$ derived from the methyl distribution. This suggests that B30 in F. thermalis exhibits direct evidence for binding Chl f and that alternate conformations of the formyl moiety are present due to the orientation of the C3 vinyl substituent toward C2. In H. hongdechloris, the H-bonding Tyr residue is conserved in sequence [6], but the associated FRL-PSI structure lacks both the PsaJ2 and PsaF2 subunits. The map quality for B30 is poor as might be expected for a region that suffered nearby subunit losses, rendering the cone scan results unreliable. Though direct evidence for Chl f-binding at B30 is not present in the H. hondechloris map, we think it highly likely that it, too, binds Chl f because the FRL-specific Tyr of PsaJ2 is conserved between the two species. Of the highly confident Chl f assignments discussed here, this site is the only one having a His residue near an axial ligand position. However, in F. thermalis the His-Mg distance for B30 is ~3.5 Å, which is too long for a direct coordination bond. Therefore, an unresolved water ligand may be present between the His sidechain and central Mg. In H. hongdechloris, the density for B30 is poor but the His-Mg distance is ~2.2 Å, suggesting direct His coordination. More complete and higher resolution structural analyses will be required to assign the axial ligand to B30 confidently.

A21. In the structure of FRL-PSI from F. thermalis, A21 was suggested to bind Chl f because the C2 formyl substituent is nearby a FRL-specific loop of PsaA2 that may donate a H-bond from a backbone amide [10]. The C2 cone scans for this site in H. hongdechloris and F. thermalis do not exhibit peaks above $\mu +3\sigma$ derived from the methyl distribution. However, this site is on the periphery of the FRL-PSI monomer where it may be more difficult to resolve substituents on the tetrapyrrole ring. In white light-adapted PSI, a His sidechain provides the axial ligand to A21, but in FRL the axial ligand is instead a water molecule, suggesting that it binds Chl f in the latter. Though no direct evidence is observed for Chl f-binding at site A21 in either of the FRL-PSI structures, we think that the FRL-specific axial coordination and the likely H-bond donor to the C2 substituent are strong indirect evidence that this site binds Chl f. It may be that the binding affinities for Chl a versus Chl f at this site are similar, thereby increasing the resolution required for direct observation of the contribution of a partially occupied formyl substituent at this site.

B38. Recent spectroscopic data [13] suggested that a Chl f dimer is located near the electron transfer chain in FRL-PSI [11]. Tros et al. suggested that B37 and B38 comprise this Chl f dimer [13], the former of which we assign herein as a highly occupied Chl f site [6]. In H. hongdechloris, B38 is found in a region of very high local resolution, 2.26-Å, and the C2 cone scan of B38 does not exhibit peaks above $\mu \pm 3\sigma$ derived from the methyl distribution, suggesting it does not bind Chl f at high occupancy; however, Tros et al. noted that a Trp sidechain found nearby C2 of B37 in the white light isoform of PsaB is instead a smaller Phe in the FRL isoform, and that spurious peaks are observed in the ESP map of H. hongdechloris FRL-PSI that may correspond to low occupancy water molecules, one of which could H-bond to the formyl moiety if B38 is Chl f (Supplementary Fig. 10) [13]. Based on the spectroscopic evidence and FRL-specific changes, it is likely that B38 binds Chl f, but the lack of cone scan evidence and unclear map features imply heterogeneity, suggesting that Chl f may bind at low occupancy. In the C2 cone scan of B38 from F. thermalis, peaks are observed at both ~340° and ~190°, though neither are above $\mu +3\sigma$ derived from the methyl distribution, and the C3 vinyl is oriented toward C2, possibly resulting in two conformations of the C2 formyl orientation as similarly observed for B30 and B37. In both structures, the axial ligands to both B37 and B38 are water molecules, consistent with the possibility of these sites binding Chl f.

Consensus of Chl f-binding sites in FRL-PSI structures

Some of the C2 cone scans exhibit extra ESP relative to the methyl distribution (Fig. 2) due to close proximity of adjacent tetrapyrrole rings rather than the presence of a formyl moiety (Supplementary Text 2 and Supplementary Figs. 11 and 12), including Chl site A8 in H. hongdechloris that exhibits extra ESP at a high probability orientation (Fig. 2). Sites that exhibit direct evidence or strong indirect evidence for binding Chl f are shown in Fig. 3. Other previously proposed Chl f sites are addressed in Supplementary Text 3. Biochemical analyses of FRL-PSI complexes from both F. thermalis and H. hongdechloris suggest that ~7 Chl f molecules are associated with each FRL-PSI monomer [10,20]. Recent spectroscopic data suggested at least 5 Chl f per FRL-PSI monomer in F. thermalis based on the deconvolution of low-temperature absorbance spectra [13]. Our observations here suggest confident assignments for five Chl f-binding sites in FRL-PSI from F. thermalis; therefore, it is possible that 1-3 more sites bind Chl f that have yet to be discovered, or that some Chl sites bind Chl f less specifically. Species-specific Chl f-binding sites in FRL-PSI appear to be possible but uncommon because only one such site was identified here. Structures of FRL-PSI from other cyanobacterial species will help to better understand conserved features of Chl f-binding sites in FRL-PSI.

Fig. 3.

Summary of Chl f sites in the two FRL-PSI structures. Left and right panels show one monomer of the FRL-PSI structures from H. hongdechloris and F. thermalis, respectively. The top panels show a stromal view and the bottom panels show a membrane plane view from the center of the trimer. Only Chl tetrapyrrole rings are shown for clarity. Pink glows are shown for Chl sites that exhibit strong direct and indirect evidence for being highly occupied Chl f, and yellow glows are shown for Chl sites that exhibit strong indirect evidence only for binding Chl f at high occupancy.

Common themes of Chl f sites

Of the six and five probable Chl f-binding sites between the two structures, all exhibit H-bond donors to their C2 formyl substituent, and none are axially ligated by His sidechains except possibly B30, whose axial ligand is unclear. Previously, it was shown that His sidechains readily axially ligate Chl a and d, but not Chl b or c [19]. Our observations regarding the Chl f sites suggest that His sidechains also are unlikely to coordinate Chl f axially. When considering all photosystem structures, ~30% of Chls are not axially coordinated by His sidechains. It may be that in FRL-PSI, Chl f binds to such sites at low occupancy unless there is a stable H-bond donor to the C2 formyl moiety. This could explain why pigment analyses consistently predict more Chl f per FRL-PSI monomer than is observed in the cryo-EM structures. This suggestion is supported by the observation that non-FaRLiP cyanobacterial PSI can bind small amounts of Chl f promiscuously [12,17,21].

The consistency of the indirect observations with the direct evidence from our cone scans offers predictive power regarding the structure of FRL-PSII, where ~4-5 Chl f and 1-2 Chl d are expected [11,[22], [23], [24]]. Chl d has the same structure as Chl a, except that the C3 vinyl substituent is instead formyl. To predict possible Chl f-binding sites in FRL-PSII from F. thermalis, we created its homology model (see Methods) and analyzed sites for possible Chl d and f binding. Eight Chl sites are not coordinated by His sidechains in the FRL-PSII homology model and these are thus candidates to be Chl f. These are Chl_D1 and Chl_D2 in the electron transfer chain, PsbC2 (CP43)-associated Chls 504, 511, and 507, and PsbB2 (CP47)-associated Chls 602, 608, and 611. Spectroscopic evidence suggested that in FRL-PSII from Chroococcidiopsis thermalis, Chl_D1, the second Chl site in the electron transfer chain, may bind Chl f [11]. The authors noted that homology modelling showed possible FRL-specific H-bond donors nearby C2. Because Chl d is likely to be coordinated by a His sidechain [19], and because Chl f sites are commonly axially ligated by water molecules and exhibit H-bonding to their C2 formyl moiety, we suggest that Chl_D1 in FRL-PSII is Chl f, consistent with the previously discussed energetic considerations [11,24,25]. Though Chl_D2 is also axially coordinated by a water molecule, there is no evidence in our homology model to suggest that FRL-specific H-bonding exists near the C2 position.

Though they are not coordinated by His sidechains, none of the Chls PsbC2-504, PsbC2-507, PsbB2-608, or PsbB2-602 exhibit obvious FRL-specific sequence differences near the C2 substituents in our homology model. However, PsbC2-507 and PsbB2-611 do. These two Chls are symmetry-related (Fig. 4) and probably coordinated by water molecules. PsbC2-507 exhibits a FRL-specific Asn sidechain near its C2 position, and PsbB2-611 exhibits a FRL-specific Thr residue nearby its C2 position, both of which may H-bond to the C2 formyl substituent if those sites bind Chl f (Supplementary Fig. 13). The FRL-specific sidechain we suggest provides a H-bond to PsbC2-507, PsbC2-Gln279, is conserved in various FRL isoforms of PsbC, but not those expressed in white light or non-FaRLiP sequences. For PsbB2-611, however, the FRL-specific sidechain we suggest provides a H-bond to the C2 formyl, PsbB2-Thr244, is conserved in some FRL-specific sequences, but not others. This may suggest that PsbB2-611 is species-specific for binding Chl f. For F. thermalis, if the three proposed Chl f sites are correct, which are shown in Fig. 4, they account for all but one Chl f and one or two Chl d sites in FRL-PSII. Confirmation of these sites in the future may be achievable by site-directed mutagenesis coupled with spectroscopic analyses and/or by structures of FRL-PSII.

Fig. 4.

Predicted Chl f sites in FRL-PSII. On the top, a stromal view of the FRL-PSII homology model from F. thermalis is shown, and on the bottom, a membrane plane view is shown. For clarity, only the tetrapyrrole rings of Chls and pheophytins are shown, and the isoprenoid tails of plastoquinones are truncated. The oxygen evolving complex is shown for orientation. Proposed Chl f sites are highlighted in yellow.

Concluding remarks

We have developed a quantitative cone-scan method for assessing Chl types in cryo-EM maps. This method was used to identify those Chl sites in two FRL-PSI cryo-EM maps from H. hongdechloris and F. thermalis that bind Chl f. C2 formyl substituents of Chl f are typically found nearly in plane with the tetrapyrrole ring. When the C3 vinyl substituent is oriented away from the C2 position, a single orientation of the formyl C2 substituent is observed. When the C3 vinyl substituent is oriented toward the C2 position, the formyl moiety may adopt multiple conformations. Such variation in the position of the formyl substituent, and possible partial occupancy of Chl f in some sites, increase the resolution requirement for their direct observation in cryo-EM maps. We identified three and two Chl sites that exhibit direct evidence for binding Chl f in the FRL-PSI structures from H. hongdechloris and F. thermalis, respectively. Three more likely Chl f-binding sites are suggested per structure, and common themes in Chl f-binding coupled with homology modeling allowed us to suggest three Chl f-binding sites in FRL-PSII from F. thermalis. As cryo-EM continues to grow as the preferred method for obtaining molecular structures of biological macromolecules, and as higher resolution cryo-EM structures become available, authors should be rigorous in determining what extent of molecular detail can be obtained from the data. Quantitative analyses like those performed here may be used to assist structural biologists in careful modeling of cofactors that are essential to enzymatic or functional activity.

Methods

Energy scans

Chls a and f were constructed based on PDB ligand entries CLA and F6C, respectively. To simplify optimizations, the hydrocarbon tail was truncated at C20. Both Chls were optimized by the Gaussian 16 software package [26] and calculated using DFT/B3LYP [27], with the 6-31G(d,p) [28], [29], [30] basis set applied to all atoms. Tetrapyrrole rings were optimized to a planar conformation with the Mg²⁺ cation placed in-plane with the ring (N-Mg-N angle of at least 177°). Torsion angle scans were then performed on the optimized structures, rotating the selected angle a full 360° over seventy-two 5° steps. The Chls were reoptimized at each step to achieve the minimized structure for each new torsion angle. The four atoms making up each torsion angle scanned are given in Supplementary Table 5.

Calculated ESP maps

ESP maps based on the Merz-Singh-Kollman scheme [31,32] were calculated for the optimized structures of Chl a and f, described above. As in the energy scans, the DFT/B3LYP [27] and 6-31G(d,p) [28], [29], [30] basis sets were used for calculations. Supplementary Movies 1 and 2 were created with VMD [33] (Humphrey 96) by mapping the ESP charges onto the electron density.

Cone scans

Ideal cone coordinates were constructed for a reference tailless Chl that was placed in the center of a cubic box with edge length of 25 Å according to given cone angles (Table S1). On a given cone surface, 72 cone axes were generated, one every 5°. Every Chl molecule from refined structures was least-squares aligned onto the reference Chl inside the cubic box using the first atom of the substituent plus three co-planar adjacent atoms where the substitute was located alongside the corresponding experimental map using the program suites CCP4 [34] and Rave [35]. The map was inverted using Phenix [36]. Rescaled experimental ESP values on each cone axis were extracted with an increment of 0.01 Å using direct Fourier summation of corresponding structure factors. On each axis, the experimental values were extracted for plotting at an expected bond length (Table S1).

Cryo-EM

Cryo-EM sample preparation and data collection details were provided previously [10]. The data were re-processed using RELION 3.1 [37] rather than RELION 3.0 used previously. Data processing workflow, Fourier shell correlations, and statistics are shown in Supplementary Fig. 3 and Supplementary Table 1. Movie stacks were gain-corrected, aligned, and dose-weighted within RELION using its implementation of MotionCor2 [38]. The implementation of CTFFIND4 [39] in RELION was used for estimation of defocus. Manual picking of ~2,000 particles allowed for initial 2D classes that were used for automated picking of 326,978 particles. This particle selection was edited manually, removing incorrectly chosen particles and adding particles not picked automatically. 2D classification led to a data set containing 232,000 particles. 3D classification led to a data set containing 201,104 particles. The Initial Model function in RELION was used to create a low-resolution ab initio model of FRL-PSI from these particle images, and 3D reconstruction was performed, leading to a structure at 4.65-Å resolution. One round of CTF refinement, two rounds of Bayesian Polishing, a final round of CTF refinement, and 3D refinement led to a final masked map at 2.96-Å resolution as determined by the gold-standard Fourier shell correlation. The starting model in model building was that previously published [10]. The model was fit into the density using University of California, San Francisco (UCSF) Chimera [40] and manually edited in Coot [41]. For automated refinement, real_space_refine in Phenix software suite was used [36,42].

Homology modeling

To create a homology model of FRL-PSII from F. thermalis, SwissModel [43] was used, providing the sequence of FRL-specific subunits and the structures of individual subunits of PSII from T. vulcanus (PDB 3WU2) with cofactors removed. Subunits generated in this way were superimposed with the corresponding subunits from T. vulcanus PSII, and merged into a single PDB file along with the cofactors from T. vulcanus PSII.

Data availability

The cryo-EM map of FRL-PSI from F. thermalis has been deposited into the Electron Microscopy Data Bank under accession number EMD-23563. The corresponding molecular structure has been deposited in the Protein Data Bank with accession number 7PX0 replacing the previous 6PNJ. Source code for the cone scan method can be found in Supplementary Code 1.

Funding

This work was supported by Department of Energy, Office of Basic Energy Sciences, Division of Chemical Sciences grant DE-FG02-05ER15646 to G.W.B. and Grant DE-SC0001423 to V.S.B, National Science Foundation grant MCB-1613022 to D.A.B., and the Forest B.H. and Elizabeth D.W. Brown Postdoctoral Fellowship for Plant Science to C.J.G.

Author contributions

C.J.G. and J.W. conceptualized the project. Data were curated by C.J.G., H-L.H., K.M.R., and J.W. Formal analysis was performed by H-L.H., D.A.F., and J.W. Funding was acquired by C.J.G., V.S.B., D.A.B., and G.W.B. Investigation was performed by C.J.G., H-L.H., D.A.F., and J.W. with input from all other authors. Methodology was developed by H-L.H., K.M.R., V.S.B., and J.W. Project administration was performed by C.J.G. Resources were provided by C.J.G., V.S.B., D.A.B., and G.W.B. Software was written by J.W. The project was supervised by C.J.G., V.S.B., D.A.B., G.W.B., and J.W. Results were validated by C.J.G. and J.W. Visualizations were created by C.J.G., H-L.H., K.M.R., and J.W. The original draft was written by C.J.G. and H-L.H. The final manuscript was edited and reviewed by all authors.

Declaration of Competing interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to influence the work reported in this paper.