The various LH1 and/or LH1-RC core, and peripheral LH2/LH3 complexes from sulfur (Ectothiorhodospira (E.) haloalkaliphila, Thermochromatium (Tch.) tepidum, Thiorhodovibrio strain 970 (Trv. 970)) and non-sulfur (Rba. sphaeroides, Rhodoblastus (Rbl.) acidophilus, Rhodospirillum (Rsp.) rubrum) purple bacteria studied here were kindly donated to the authors by Profs. R. Cogdell (Glasgow University), C. N. Hunter (Sheffield University), A. A. Moskalenko (IBBP, Pushchino), Z.-Y. Wang-Otomo (Ibaraki University), and N. Woodbury (Arizona State University). The purple sulfur and purple non-sulfur bacteria, respectively, use either sulfide and hydrogen or organic compounds as an electron donor (Hunter et al., 2008). Isolation and purification of the complexes were carried out following standard protocols, as described in refs. provided in bookkeeping Table 1. Specific samples investigated are enlisted in Table 2 and Table 3. Distinct aspects of the sample preparation and handling are commented at proper places of the manuscript.

The concentrated samples were stored at −78°C in deep freezer. Prior the use the samples were diluted with buffer (20 mM Tris-HCl or 20 mM HEPES, depending on samples) to meet the required for specific measurements optical density. The buffers for purified complexes additionally contained proper detergent (n-dodecyl β-D-maltopyranoside (DDM), dihexanoylphosphatidylcholine (DHPC), octyl β-glucoside (βOG) or lauryldimethylamine oxide (LDAO)) to prevent aggregation of the proteins. The Trv. 970 and Tch. tepidum solutions also contained 60 mM CaCl₂ to keep samples saturated with Ca²⁺. To produce the Ca²⁺-depleted samples, EDTA was added and the sample solution was incubated for several hours before the measurements. Glycerol with a 2:1 volume ratio was used to obtain transparent glassy samples at cryogenic temperatures.

xThe multipurpose spectroscopic setup applied in this work has been recently described (Rätsep et al., 2017; Timpmann et al., 2021). It comprises a model 3900S Ti: sapphire laser of 0.5 cm^–1 linewidth pumped by a Millennia Prime solid-state laser (both Spectra Physics), a 0.3-m focal length spectrograph Shamrock SR-303i, equipped with a thermo-electrically cooled CCD camera DV420A-OE (both Andor Technology) for the measurements of fluorescence excitation anisotropy and hole-burning spectra, and a high-stability broad-band tungsten light source BPS100 (BWTek) for absorption measurements. In the measurements of anisotropy spectra, the vertically polarized laser beam was scanned over a proper wavelength range. The fluorescence spectra were detected through high-contrast analyzing polarizer set parallel or perpendicular to the excitation laser beam. Another correcting polarizer was fixed at 45° to the polarization direction of the analyzing polarizer. The estimated experimental uncertainty of the baseline anisotropy value is ± 0.03. Photobleaching and scattering are potentially detrimental to the quality of anisotropy measurements. To minimize photobleaching, the excitation light flux density was kept as low as feasible for achieving reasonable signal-to-noise ratio. At the experimental power density of 1–10 mW/cm² some amount of photobleaching was only observed when excited in resonance with the main absorption band of the samples, which, however, didn’t concern determination of the anisotropy dip positions. Optimization of the sample concentration and blackening of the cuvettes’ holder were used to curtail the excitation laser-light scattering. The wavelength scale of the spectrometer was determined with a precision of ± 0.1 nm using a Ne−Ar calibration lamp. In low-temperature measurements the PMMA plastic cuvettes (Brand) filled with the sample solution were placed into a liquid helium bath cryostat (Utreks). The cryostat was equipped with a temperature controller (model 211 Lakeshore Cryotronics), stabilizing the temperature within ± 0.5 K.

The data were analyzed and fitted using the graphing and data analysis software Origin 9.0 SR1 (OriginLab).

Demonstrated in Figure 1 are the Q_y absorption spectra of LH1-RC (Figure 1A) and LH2 (Figure 1B) complexes from selected WT bacteria in the proper light-harvesting exciton spectral range, see below. Broadband version of the absorption spectra spanning from 250 nm to 1,100 nm can be found in, Supplementary Figure 1. For better comparison, reciprocal (linear in energy) wavelength scale is applied whenever feasible. Also, for the convenience sake, a reference terrestrial solar irradiation spectrum is provided on top of the experimental absorption spectra. It is well-known that the absorption spectra of LH1-RC complexes are in the near-infrared part of the solar spectrum dominated by a sole Q_y exciton band of LH1, also known as B875. The RC cofactors contribute visibly into this spectrum only around 750 and 850 nm, in the high-energy tail of the absorbance of LH1 excitons. Absorption spectra of LH2 complexes encompass two peaks, due to weakly-coupled and strongly-coupled pigments in the B800 and B850 circular arrangements of BChls, respectively.

The peak-normalized spectra in Figure 1 demonstrate an exciting spectral tunability of LH1 excitons in different native complexes that almost reaches 100 nm. This is much more than the extent observed in the case of B850 (12 nm) or B800 (5 nm). For the sake of comparison, it is interesting to note that the variations of the Q_y energy of BChl in polar and nonpolar solvents are about 17 nm and 22 nm, respectively (Limantara et al., 1997). As follows, we will concentrate on the properties of the most responsive strongly coupled B875 and B850 excitons.

Presented in Figure 2 are the fluorescence excitation anisotropy spectra of the Q_y excitons in, respectively, LH1/LH1-RC (Figure 2A) and LH2 (Figure 2B) complexes from the same WT bacterial strains as exposed in Figure 1. Shown also with continuous line is absorption spectra of the samples.

The shape of the anisotropy spectrum generally depends on the recording wavelength, see Figure 3. Being an evidence of a spectrally disordered ensemble of light-harvesting complexes, this effect is the strongest at selective recording in the blue side of the emission spectrum, but almost vanishes when recording at red part, past the maximum of the spectrum, as demonstrated in the inset of Figure 3. Therefore, in order to handle comparable results, selective recording of fluorescence at red side of the spectra with 10-nm bandwidth was commonly applied in the anisotropy measurements.

In agreement with the previous work (Timpmann et al., 2004; Timpmann et al., 2005), the anisotropy spectra of different cyclic light-harvesting complexes look rather similar. They feature high anisotropy values at the long wavelength edge of the respective absorption spectra that strive towards a theoretical maximum value of 0.4 and two dips towards shorter wavelengths, where anisotropy values close to 0 are generally observed. According to numerical modeling (Trinkunas and Freiberg, 2006; Pajusalu et al., 2011) the higher-energy (or blue) anisotropy dip consistently showing up in a relatively narrow spectral range of 760 ± 10 nm rather precisely reproduces the high-energy edge of the Q_y exciton band, whereas the red dip appears around the strong (k = ±1, ±2) Q_y exciton transitions.

Hence, the energy difference ∆E between the blue and red dips, although underestimating the actual exciton bandwidth, can be considered as an effective measure of the exciton bandwidth in cyclic light-harvesting complexes (Timpmann et al., 2004). As shown in (Timpmann et al., 2005), a more precise estimation of the exciton bandwidth is obtained as the energy difference ∆E₀ between the blue dip of the anisotropy spectrum and λ₀. The λ₀ represents the peak wavelength of the maximum of the inhomogeneous distribution function (IDF) of the lowest-energy k = 0 exciton states, which is measured as the peak position of the hole burning action spectrum (Reddy and Small, 1991). From the ∆E and ∆E₀ data collected in Table 2 and Table 3, one can immediately grasp that exciton bandwidths in peripheral complexes are systematically narrower compared with those in core complexes, the ratio changing from about 2/3 in Rba. sphaeroides to almost 3/4 in Rbl. acidophilus.

Generally speaking, molecular excitons are sensitive to all physical factors that disturb transition energies of the sites and couplings between them. As follows, we will demonstrate using parallel measurements of fluorescence anisotropy and absorption spectra the responses of the light-harvesting exciton band structure upon altering the environment and internal structure of the chromoproteins. The problem is that in intricate systems such as the chromoproteins studied here, these effects can but rarely be uniquely unraveled. The answer to the question whether one is dealing with environmental or structural property depends on the degree of the effects observed as well as on subjective viewpoint of the investigators. Nevertheless, most researches agree that purification of membrane proteins out of its native membrane by mild detergents, a standard biochemical procedure for studying membrane proteins in isolation, provides a more-or-less pure case of an environmental change, see Figure 4.

Examples of the internal structure variations presented in this work include changes in spatial organization of pigment chromophores illustrated by Figure 4B and Figure 5, in the content of metal ions (Figure 6), in the tertiary structure H-bonding network (Figure 7), and in the type and presence of carotenoids (Supplementary Figure 2).

Native membrane-embedded versus detergent-isolated chromoproteins. Shown in Figure 4A are effects of detergent-purification on the absorption and fluorescence anisotropy spectra of LH2 complexes of Rba. sphaeroides. As seen, the purification result in subtle but characteristic spectral modifications, which foremost include a couple of nanometers blue-shift and broadening of the Q_y absorption band. The Q_y band broadening observed may be caused by 1) the less ordered detergent environment or by 2) the enhanced exciton coupling. The concurrently measured anisotropy spectra that demonstrate narrowing of the exciton band exclude option 2). The weakening of exciton couplings upon the membrane protein purification is also consistent with the common-sense loosening of the protein structure in the surroundings of detergent micelle rather than with tightening. At the same time, as seen in Table 3, choice 1) appears to contradict with the much narrower IDF observed in the case of detergent-isolated sample. An obvious solution to this conundrum is that Γ_IDF in native membrane-embedded LH2s is dominated by external disorder, while that in detergent isolated LH2s, by internal disorder. See continuation of this discussion in Section 3.5.

The Q_y band of LH1 complexes similarly broadens in response to detergent-purification, while the exciton band is narrowing (Figure 4B). Yet in this case, internal disorder appears to play leading role because of greater Γ_IDF recorded for purified LH1 complexes, see Table 2. Notable also is the much-shifted position of the blue anisotropy dip and somewhat distorted structure of the anisotropy spectrum of purified LH1 complexes compared with those of membrane complexes. Especially the deviating blue anisotropy dip, which we will repeatedly meet below (see Figures 5–7), is a sure indicator of structural rearrangements accompanying the various sample handling processes (Freiberg et al., 2010; Kunz et al., 2013). Spectral responses following the assembly of LH1 with RC to mimic core LH1-RC complexes in the native membrane environment appear rather small and uncharacteristic, as are the differences between the spectra of native dimeric (LH1-RC-pufX) and monomeric mutant (LH1-RC) core complexes, see Figure 4B and Table 2.

Although subtle, important effects of embodiment of bacterial light-harvesting complexes into lipid bilayer have been thus established combining absorption, fluorescence anisotropy and hole-burning techniques. These data add to the mounting previous evidence that some photosynthetic processes may proceed differently in the native membrane, as compared to artificial detergent environment (Bowyer et al., 1985; Hunter et al., 1985; Beekman et al., 1997; Pugh et al., 1998; Urboniene et al., 2007; Pflock et al., 2008; Freiberg et al., 2012b; Freiberg et al., 2016). According to data in Table 2 and Table 3, the relative exciton band narrowing observed in LH2 and LH1 complexes of different organisms is rather small: 2%–3% and 3%–9%, respectively. In water-soluble chlorophyll proteins, for example, much stronger environmental effects have been detected (Fresch et al., 2020).

LH2 complex with depleted B800 pigment ring. Figure 5 offers a comparison of the fluorescence excitation anisotropy spectra of Q_y excitons for the WT and mutant B800-less complexes of LH2 from Rba. sphaeroides. Removal of the BChls from the B800 binding sites implicates large perturbation on both the protein spatial structure and the B850 exciton structure, as compared the same characteristics in WT complex. In conformity with this, a 5–6 nm red-shift of the B850 absorption band was observed, associated with almost 200 cm^–1 broadening of ∆E (see Table 3).

Unfortunately, because of high affinity of B800 binding sites to native as well as to non-native pigments (Swainsbury et al., 2019), the ambition to prepare the complexes completely free from the B800 site inclusions is untenable. As an example, all the multiple samples in our disposal show different degree of contamination of the B800 sites with the BChl pigments, as shown in Supplementary Figure 3. To our mind, this provides a reasonable explanation for the long-lasting confusion about the variability of the relevant literature data (Leupold et al., 1999; Timpmann et al., 2004; Gall et al., 2010). In particularly, this removes any credibility behind the identification of an upper exciton component in the LH2 complex with depleted B800 of Rba. sphaeroides using CD (Koolhaas et al., 1998).

Effects of structurally relevant metal ion inclusions. Metal ions are frequent embellishments of photosynthetic proteins. However, to the best of our knowledge, only sulfur purple bacteria utilize Ca²⁺ ions to gain supremacy by extending LH1 spectra maximally towards red. Hence, the Q_y absorption of LH1 in WT Ca²⁺ containing Tch. tepidum exhibits at ambient temperature a peak at 914 nm (Fathir et al., 1998), while that in Trv. 970 at record long wavelength of 960 nm (Permentier et al., 2001). The depletion of 16 Ca²⁺ ions in the C-terminal domain of LH1 complexes results in significant blue-shifting of their absorbance, see Figure 6A and Table 2. The new structure is less ordered, confirmed by substantial broadening of both the Q_y absorption band and IDF. The shift toward higher energies is reversible, so that the original spectra almost precisely recover upon the reconstitution of Ca²⁺ back into the respective protein structures (Imanishi et al., 2019). In contrast, the effect of metal ions (Ca²⁺ or Ba²⁺) on the spectra of LH2 complexes is fairly weak, see Table 3.

We have previously established an enhancement of exciton couplings as the main mechanism of the Ca-facilitated spectral red-shift (Timpmann et al., 2021). These data are reproduced in Figure 6A. Shown in Figure 6B is a fresh result of exchange of Ca²⁺ to Ba²⁺, another group IIA alkaline earth metal with slightly greater ion radius. As seen from spectral responses, the effect of ions exchange is much less dramatic than depletion of the ions. This is reasonable, because the ion replacement is expected only slightly distort the native Ca²⁺ binding sites, while the removal of ions is prone to collapse the sites that may initiate a large-scale restructuring of the protein (Timpmann et al., 2023). The widely separate positions of the blue anisotropy dips observed in case of Trv. 970 complexes and the nearly overlapping dips in case of Tch. tepidum complexes support these initial conclusions.

Tertiary-structure hydrogen-bonding effects. Hydrogen bonding applied at different stages of protein folding is vital for stabilization of the protein. A few well-defined tertiary structure H-bonds have been identified, which are instrumental in color tuning of Q_y bands in LH1 and LH2 complexes (Fowler et al., 1994; Sturgis et al., 1997; Sturgis and Robert, 1997; Kangur et al., 2008; Uyeda et al., 2010; Freiberg et al., 2012a). These first and foremost concern the bonds formed between the conserved tryptophan (Trp) and/or tyrosine (Tyr) protein residues and the C3 acetyl carbonyl group of BChls. According to recent theoretical evaluations, the band shifts created are due to joint effect of the dihedral angle of the acetyl moiety and the shuttling of the proton in the H-bond (De Vico et al., 2018).

Any adjustment of the H-bonding pattern necessarily involves transformation of the ground and excited electronic state structures of the excitonically coupled aggregates, hence their exciton properties. A straightforward way to demonstrate this is by direct comparison of the exciton state bandwidths, as revealed by fluorescence excitation anisotropy (data in Tables 2, 3).

Figure 7 provides examples of anisotropy spectra in case of WT and various H-bond mutant LH1-RC and LH2 complexes from Rba. sphaeroides. In the LH1-RC complex, one of these mutations (αTrp₊₁₁Phe, termed α-mutant in Figure 7A), replaces the tryptophan that H-bonds to the C3 acetyl carbonyl group of one of the BChls in a αβ-BChl₂ apoprotein unit, disrupting the H-bond, while another (βTrp₊₉Phe, β-mutant), disrupts the H-bond to the C3 acetyl carbonyl of the pairing BChl. Double mutant core complexes with replacements in both sites are structurally compromised and are unstable in detergent. For LH2 only membrane-embedded complexes are available carrying the H-bond breaking single (αTyr44Phe, α-mutant) or double (αTyr44Phe-αTyr45Leu, αβ-mutant) residue replacements.

In agreement with the previous knowledge (Fowler et al., 1994), the disruption of particular H-bonds results in blue-shifting of the absorption spectra in comparison with the spectrum of the WT complex. With one notable exception, these spectral shifts are according to expectation accompanied by exciton band narrowing. Except the β-mutant of LH1-RC-pufX, which shows broadening, not narrowing of the exciton bandwidth. Although quantitative evaluation of this effect awaits to be done, the shift of the blue side anisotropy dip allows to conclude that the replacement of the residue at the β₊₉ position causes greater amount of structural rearrangements in LH1 than the replacement at the α₊₁₁ position, confirming a prior notion obtained from analyses of Raman spectra (Sturgis et al., 1997).

Carotenoid depletion or exchange. The significance of carotenoids in the assembly and reinforcement of light-harvesting protein structures is widely recognized (Loach and Parkes-Loach, 2008). Yet the role carotenoids play in modulation of the light-harvesting exciton properties is less well documented. Here, we have been assessing the consequences of carotenoid exchange or depletion in a number of bacterial light-harvesting complexes. The results obtained are collected in Table 2 and Table 3, being also partially illustrated by Supplementary Figure 2.

Supplementary Figure 2 shows that the depletion of carotenoids from either LH1 or LH2 complexes has by far greater impact on light-harvesting excitons than their exchange. The exchange such as demonstrated in Supplementary Figure 2A generally well keeps the structure of the exciton band. However, collapse of the voids subsequent to removal of carotenoid and the accompanying relaxation of the B850/B875 pigment assembles may (as shown in Supplementary Figure 2B) or may not (Supplementary Figure 2C) save the intact exciton structure.

As it was presented in Introduction, attempts have been made to reveal the electronic structure of BChl excitons in LH2 and LH1 complexes by other experimental techniques, apart from polarized fluorescence spectroscopy. Especially CD method has been popular, because the structure-based model calculations promise clear CD signatures related to both red and blue exciton band edges of light-harvesting complexes that contain cyclic B850/B875 aggregates of BChls. These attempts, however, have all failed in intact peripheral and core light-harvesting units. Not so much because the expected weakness of the CD signal corresponding to the top of the exciton band, but more typically because of unfortunate spectral overlaps with other complements of these units such as RC in LH1-RC or B800 in LH2, which both overwhelmingly contribute into the CD spectra in the spectral range of interest.

Since the efforts to fix the LH2 complexes sufficiently free of B800 pigments proved unsuccessful, we were paying attention to purified LH1 complexes free of RC. In these complexes, a weak negative CD band was previously observed around 754 nm at 77 K and around 763 nm at room temperature (Georgakopoulou et al., 2006b). According to exciton modeling, these features were ascribed to the high-energy exciton component of the Q_y band.

Figure 8A confronts the CD spectrum of the LH1 complex from Rba. sphaeroides measured in this work at 77 K with its fluorescence anisotropy spectrum recorded at 4.5 K. Let us note that the CD spectrum of our sample precisely reproduces the already published spectra (Georgakopoulou et al., 2006a; Georgakopoulou et al., 2006b). The close match observed between the location of the CD and anisotropy dips reconfirms that the latter can be considered as an adequate measure of the top exciton band position.

The measurements of fluorescence excitation spectra provide yet another independent validation of the fluorescence anisotropy technique. Demonstrated in Figure 8B is the experimental determination of the top exciton band edge in the intact LH1-RC core complex from Trv. 970, in which case the CD method fails totally. The inset represents the blow-up version of the figure in the spectral region of the LH1 exciton band top, where the 1–T spectrum is dominated by the RC absorbance.

As seen, the photons absorbed at cryogenic temperatures by the RC get trapped and do not practically contribute into the LH1 exciton fluorescence. This exposes a rather weak (∼235-fold smaller than the main Q_y exciton absorption band at 987 nm) absorption band peaking at 772 nm. Exact overlap with the high-energy anisotropy dip allows this band to be assigned to the exiton states at the very top of the exciton band.

Having determined the boundaries of the Q_y exciton state manifolds for all the 33 light-harvesting complexes studied, as represented in Table 2 and Table 3, it would be instrumental to plot the exciton bandwidths as a function of the Q_y absorption peak energy, E (Figure 9). The graphs are presented separately for the ∆E (Figure 9B) and ∆E₀ (Figure 9C) bandwidth definitions. They draw (almost) parallel lines, establishing linear negative correlations between the exciton bandwidths and the E. This means that the complexes with broader bandwidths (with stronger coupled excitons) are adapted to absorb lower-energy (redder) light and vice versa. A spectral division of the complexes into two major groups (LH1/LH1-RC core complexes and LH2 peripheral complexes) can also be clearly seen in Figures 9B, C. In energy terms, the color-tuning range of core complexes (1,481 cm^–1) is almost twice greater than that of LH2 complexes (752 cm^–1).

The ∆E₀ exciton bandwidths as a function of E well follow a linear function: ∆E₀ = 11,700–0.850 × E. A great practical value of this insight is that exciton bandwidths (thus the exciton couplings) and their changes can be readily evaluated with reasonable accuracy using simple measurements of absorption spectra. We have verified that the data that appear deviating from the above linear law correspond to structurally modified complexes, the LH3 complex from Rbl. acidophilus being an extreme case. Note, however, that this deceivingly simple relationship determined at 4.5 K is not expected to hold at elevated temperatures, because of thermally-induced variations of site energies, exciton coupling energies as well as the structure of the chromoprotein complexes in certain cases (Rätsep et al., 2018). Additional investigations are, therefore, required to find the bandwidth dependences on temperature (Pajusalu et al., 2011; Freiberg et al., 2013).

The exciton bandwidths recoded in the LH1/LH1-RC core complexes vary between 3,005 and 1,857 cm^–1, while those in LH2 peripheral complexes, between 2,013 and 1,347 cm^–1. The exciton couplings energies can be deduced from a “rule of thumb” equation, ∆E₀ ≈ 2 (V₁+V₂), of a model dimerized BChl chain in nearest-neighbor approximation, where V₁ and V₂ are the intra- and inter-dimer coupling energies. Assuming V₁ ≈ V₂, the low-temperature coupling energies may thus reach about 750 cm^–1 in LH1/LH1-RC and about 500 cm^–1 in LH2.

The formation of excitons is usually accompanied by sharpening of monomer spectra, a celebrated example being J-aggregates (Hestand and Spano, 2018). The degree of exciton line narrowing is a function of the ratio Γ_m/V, where Γ_m is the width of the inhomogeneously broadened monomer spectrum and V = (V₁+V₂)/2. Therefore, it is expected that the complexes with stronger coupling expose narrower IDF, a results of the so-called exciton motional narrowing effect (Knapp, 1984). This qualitative expectation is indeed roughly followed by our data, see Supplementary Figure 4. It also appears reasonable that the relatively weaker-coupled LH2 complexes are more widely spread with respect to the average value of Γ_IDF. Yet in several cases such as LH1-RC (Ca−) of Trv. 970, LH1-RC (Ba+) of Tch. tepidum, LH2 of E. haloalkaliphila, and all membrane-embedded LH2 complexes of Rba. sphaeroides, anomalously broad IDFs (Γ_IDF ≥ 200 cm^–1) are observed, see Table 2 and Table 3.

In the literature, the width Γ_IDF has been widely considered as a proper measure of the exciton disorder (Reddy et al., 1992; Purchase and Völker, 2009). This is not quite correct, because in the chromoprotein systems two qualitatively different types of spectral disorders have been identified, one caused by intra-protein (or internal) and second, due to inter-protein (external) variations (Freiberg et al., 1999), whereas only the former component is subject to motional narrowing. The systems with anomalously broad IDFs listed above thus apparently belong into the realm governed by external disorder.

The phototropic bacteria that in addition to core complexes contain peripheral light-harvesting complexes have to carefully tune their respective spectra to achieve optimal transport of excitons towards RCs. As can be seen from Figure 10 and Supplementary Figure 5, the exciton state manifolds in core and peripheral complexes perfectly match each other, whereby the high-energy edges of the exciton state manifolds in LH1 and LH2 nearly coincide. These facts were noticed years ago (Freiberg et al., 2013), but their comprehensiveness wasn’t immediately evident.