Oligomeric states of microbial rhodopsins determined by high-speed atomic force microscopy and circular dichroic spectroscopy

Oligomeric assembly is a common feature of membrane proteins and often relevant to their physiological functions. Determining the stoichiometry and the oligomeric state of membrane proteins in a lipid bilayer is generally challenging because of their large size, complexity, and structural alterations under experimental conditions. Here, we use high-speed atomic force microscopy (HS-AFM) to directly observe the oligomeric states in the lipid membrane of various microbial rhodopsins found within eubacteria to archaea. HS-AFM images show that eubacterial rhodopsins predominantly exist as pentamer forms, while archaeal rhodopsins are trimers in the lipid membrane. In addition, circular dichroism (CD) spectroscopy reveals that pentameric rhodopsins display inverted CD couplets compared to those of trimeric rhodopsins, indicating different types of exciton coupling of the retinal chromophore in each oligomer. The results clearly demonstrate that the stoichiometry of the fundamental oligomer of microbial rhodopsins strongly correlate with the phylogenetic tree, providing a new insight into the relationship between the oligomeric structure and function-structural evolution of microbial rhodopsins.

other microbial rhodopsins are also expected to be relevant for their molecular properties and functionalities. However, although numerous microbial rhodopsins have been recently identified in various organisms, their oligomeric states have remained unclear because of a lack of effective ways to determine them under near a physiological condition 7 .

Results and Discussion
Oligomeric structure of sodium-pumping Krokinobacter rhodopsin, KR2. First, we analyzed a light-driven sodium-pumping rhodopsin KR2. Crystal structures of KR2 at relatively high (pH 4.0 and 4.3) and low (4.9 and 5.6) acidic pHs were found to contain monomer and pentamer units, respectively 9,10 . HS-AFM images clearly reveal ring-shaped pentameric forms in lipids at physiological pH (pH 8) ( Fig. 2a and b). Interestingly, the HS-AFM images exhibit two distinct structures, which could correspond to the two different sides of KR2 in the lipid: C-(cytoplasmic) and N-terminal (extracellular) sides (Fig. 2b). The simulated AFM images of the N-and the C-terminal sides of KR2 constructed from the crystal structure reported as pentamers (PDB code 4XTN) 9 display two distinctive configurations: a round and a star-like ring shape (Fig. 2c). The simulated AFM image of the N-terminal side reproduces the real AFM image showing the round ring shape well (left panel in Fig. 2b). On the other hand, the simulated AFM image of the C-terminal side is similar to the star-like arrangement of the real AFM image, except for a large protrusion in the ring's center (right panel in Fig. 2b). This protrusion originates presumably from the additional 6 × His-tag on the C-terminal side of KR2. In fact, after deletion of KR2's His-tags (KR2 278 ), the AFM images show a central pore at the pentamer center similar to the simulated AFM image (see Supplementary Fig. S1a). It is, therefore, safe to conclude that the protrusion seen in Fig. 2b (right panel) corresponds to the additional His-tags of the C-terminal side of KR2. This indicates that the His-tags on KR2 monomers in the pentamer likely integrate at the center of the pentamer and are rigidly structured. Further, we confirmed that the pentamer form is retained in an acidic solution to compare it with the crystal structure of KR2 prepared at pH 4.0-4.3, which contains the monomeric form 9,10 . All of the KR2 278 observed at pH 4.3 are pentamers similar to those observed at pH 8.0 (see Supplementary Fig. S1b). This suggests that the monomeric form found in KR2 crystals represents an unphysiological state likely favored during crystallization but is not a native form in the lipid.
As described above, visible CD spectroscopy is commonly used to assess the oligomeric structures of microbial rhodopsins. For example, the CD spectra of BR and halorhodopsin from H. salinarum (HsHR) show bilobe-shaped spectra (CD couplet) with positive and negative peaks at the short and long wavelength sides of the absorption maximum (λ max ), respectively 28,29 . The couplet originates from exciton coupling between the retinal chromophores in the trimer 30 . When BR or HsHR is solubilized to the monomeric state, no exciton coupling occurs 29 . Thus, the CD spectrum does not show the bilobed shape and thus only a single positive peak (intrinsic CD) at λ max is observed, whose shape is identical to that of the visible absorption spectrum. In KR2, we observed a single peak around 544 nm, which is similar to those for monomeric BR and HsHR, but the peak wavelength is significantly longer than the λ max (524 nm) (Fig. 2d). This shift to the longer wavelength compared to the λ max can be explained as the sum of intrinsic CD and inverted CD couplets, which have negative and positive peaks at the short and long wavelength sides of λ max , respectively. The pattern of exciton coupling strongly depends on the relative orientation of retinal in the oligomer. While the retinals are aligned in a head (the β-ionone ring)-to-tail (the retinal Schiff base) manner in the BR trimer, the retinals in the KR2 pentamer have a tail-to-tail type orientation ( Fig. 3). Thus, the exciton coupling of the retinals in KR2 should differ from that of the retinals in BR, and hence the inverted CD couplet observed in KR2 is expected to be a characteristic of the pentameric oligomer. The CD spectrum of DDM-solubilized KR2 shows an almost identical peak at 544 nm (dotted line in Fig. 2d), indicating that the pentameric form of KR2 is preserved even under DDM-solubilized conditions.
We also attempted to visualize the light-induced conformational changes of KR2 as observed for BR 25 . In BR, our HS-AFM movies clearly showed that the C-terminus of BR monomer undergoes a large conformational change towards the outside of the trimer center 25,26 . Because the photocycle of wild-type KR2 is completed in 8.7 ms, we used an N112A variant of KR2 which has a longer photocycle (1200 ms decay of the M-intermediate) (see Supplementary Fig. S2a) 8,31 . Here, the N112A variant pumps protons, but not sodium ions 8 . The KR2 278 N112A was observed in the dark and after green light illumination (see Supplementary Fig. S2b and Video S1). Unlike BR, clear conformational changes of KR2 triggered by light were not detected, although we imaged both the C-and N-terminal sides. This result suggests that KR2 does not undergo large conformational changes which can be resolved by our HS-AFM during the photocycle. Translocation of sodium ions inside KR2 may occur by diffusion without large conformational changes.
Oligomeric structures of microbial rhodopsins from eubacteria. Next, we observed other microbial rhodopsins, recently identified in different species, to systematically determine their oligomer states. Figure 4a shows an AFM image of proton-pumping GR in the lipid, revealing that GR is in a pentameric form similar to KR2 (left and middle panels of Fig. 4a). In addition, the GR pentamers also exhibit both round ring and star-like arrangements similar to the KR2 pentamers (left panel of Fig. 4a). However, the protrusion at the center of the pentamer is more blurred compared to that of KR2, although the GR monomers also have a 6 × His-tag at the C-terminal side. This may represent structural differences in the C-terminal sides between KR2 and GR monomers.
The CD spectrum of GR in the lipid (POPE/POPG) shows a couplet with a negative peak at 503 nm and positive peak at 566 nm (right panel of Fig. 4a), which is clearer than that of KR2. This blue-side negative/red-side positive couplet suggests that GR also forms pentamers like KR2, which is consistent with the AFM image. The CD spectrum of solubilized GR in DDM shows a similar CD couplet to that in the lipid, indicating that the solubilization does not change the oligomer form of GR. A previous report showing the similar CD couplets both in detergent and an E. coli membrane concluded that the inverted CD couplet of GR represents the formation of trimers with different molecular arrangements from BR and NpHR 32 . However, the combination of HS-AFM and CD spectroscopy indicates that the inverted CD spectrum is a characteristic of pentamer form. In fact, the size-exclusion chromatography pattern in the previous report showed that the main peak of GR in DDM was significantly higher than that of NpHR 32 , which was considered to be in the trimeric form [33][34][35] . This also suggests that GR forms larger oligomers than trimers even in DDM. In the same way, we analyzed the oligomeric state of inward chloride-pumping FR. The AFM image of FR clearly indicates the pentameric form in the lipid (left and middle panels of Fig. 4b).
The CD spectrum of FR shows a single positive peak at 544 nm whose the wavelength is longer than the λ max of FR (518 nm) 36 and is similar to that of KR2 (right panel of Fig. 4b). Furthermore, we evaluated the oligomeric states of KrActR and QsActR belonging to the recently reported actinorhodopsin sub-family of proton pumps found in freshwater (Fig. 1) 37 . Both the HS-AFM images of KrActR and QsActR show the ring-like pentamer forms (left and middle panels of Fig. 4c and d), same as KR2, GR, and FR. The CD spectra of KrActR and QsActR show positive peaks at 569 and 572 nm, which are longer than the λ max (KrActR: 532 nm; QsActR: 551 nm), respectively (right panels in Fig. 4c and d). These results suggest that KrActR and QsActR also have negative contribution of the couplet at the blue side of their absorption peaks. They are observed as dips at 520 and 519 nm, respectively, and do not decrease below zero because of the overlapping positive intrinsic CD. Thus, the orientation of retinal chromophores in ActR pentamers is identical to those of KR2, GR, and FR. Interestingly, the AFM image of ActR does not show protrusions at the center of the pentameric ring, in contrast to the KR2, GR, and FR. This is likely because the ring diameters of ActR pentamers (4.5 ± 0.21 nm for KrActR and 4.9 ± 0.050 nm for QsActR) are larger than those of the others (4.3 ± 0.070 nm for KR2, 4.3 ± 0.12 nm for GR and 4.1 ± 0.26 nm for FR), and therefore entanglement of the His-tags would hardly occur. This also implies a looser interaction between monomers in the ActR compared to in KR2, GR, and FR. Furthermore, both KrActR and QsActR have longer C-terminal tails than KR2 and FR after the end of 7th transmembrane helix (helix G) (see Supplementary Table 1). The longer tails might have large fluctuations and inhibit the formation of the rigid protrusion structure. Similarly, GR has a longer C-terminal tail than KR2. This can also facilitate large fluctuations and a less-structured protrusion.
A previous AFM study showed that proton-pumping GPR predominantly assembles into a hexameric form with a smaller fraction assembling into a pentameric form in the densely packed condition in a lipid membrane 7 . Our AFM images of GPR in a lipid membrane also reveal the coexistence of pentamers and hexamers (Fig. 4e), although pentamers are the dominant structure in our experimental conditions (pentamers: hexamers = 76% : 24% for 486 oligomers in total). The CD spectrum of GPR shows a couplet with a negative peak at 495 nm and positive peak at 550 nm in the lipid (rightmost panel in Fig. 4e). This negative/positive couplet is similar to those of other eubacterial rhodopsins (KR2, FR, KrActR and QsActR), and consistent with the pentamer dominant oligomeric state of GPR observed by HS-AFM. The inhomogeneous oligomeric mixture of GPR suggests that inter-molecular interaction is weaker than that of other eubacterial rhodopsins which form a strictly regulated pentameric structure. Since this indicates a decreased robustness of GPR oligomers, the CD spectrum was also measured in DDM. For the solubilized GPR, we observed a large change in the CD spectrum with a single positive peak at 541 nm. This is still significantly longer than the λ max of GPR (519 nm), indicating that GPR forms pentamers even in the detergent. The less significant contribution of the couplet, however, compared with that in the lipid implies that a part of the oligomeric mixture of pentamers and hexamers is disrupted and/or the exciton coupling is weakened by the change in the oligomeric structure in the solubilized condition. Taken together, the native oligomeric states of most eubacterial rhodopsins appear to be mainly pentamers. Interestingly, however, these microbial rhodopsins transport different ions in opposite directions, suggesting that oligomer formation is not crucial for the fundamental functions of microbial rhodopsins.

Oligomeric structures of XeR and microbial rhodopsins from archaea.
It is well-known that the fundamental oligomeric state of BR from archaea is a trimer in the native purple membrane 4 . Further, trimeric oligomers of some microbial rhodopsins from archaea have been determined using various methods 34 , but there has been no direct evidence under lipid-reconstituted conditions. The AFM image of chloride pump NpHR clearly shows the trimer forms in the lipid, which is consistent with a previous report (Fig. 5a) 34 . Additionally, the CD spectrum of NpHR shows a positive peak and negative peak on the blue (541 nm) and red sides (608 nm) of λ max = 576 nm, respectively (right panel in Fig. 5a) 38 . This is similar to the CD spectra of trimeric BR and HsHR 28,29 , and the CD couplet is inverted compared to those observed in the pentameric rhodopsins (KR2, FR, GR, KrActR, QsActR and GPR). Thus, NpHR forms trimer as observed with AFM in the lipid membrane, which is also suggested by crystallography 35 . The blue-side-positive/red-side-negative type of CD couplet would be a common signature of trimeric rhodopsins.
Next, we observed archaeal phototaxis receptor sensory rhodopsin II from N. pharaonis (NpSRII) incorporated into the lipid. NpSRII was found to form trimers, by AFM (Fig. 5b). The CD spectrum of NpSRII shows only a positive peak at 498 nm, which is nearly identical to the λ max (500 nm) 39 (right panel in Fig. 5b). This indicates that the CD couplet by exciton coupling between NpSRII molecules is much smaller than the intrinsic CD, although NpSRII molecules are in the trimeric form. Consequently, the oligomers of microbial rhodopsins from archaea typically predominantly exist as trimers in the lipid.
The xenorhodopsin (XeR) group is found in eubacteria, but is classified as an archaeal type rhodopsin in the phylogenetic tree of microbial rhodopsins (Fig. 1). We recently reported that inward proton-pumping xenorhodopsin from Parvularcula oceani (PoXeR) forms trimers in a lipid 40 . We reproduced the previous result that PoXeR trimers pack into a 2D crystal similar to BR (left panel of Fig. 5c). The CD couplet of PoXeR shows positive and negative peaks at 542 and 609 nm, which are shorter and longer than the λ max (569 nm), respectively 40 , confirming that PoXeR exists in the trimeric form in the lipid (right panel of Fig. 5c). Furthermore, we observed ASR which belongs to the same group as PoXeR (Fig. 1). X-ray crystal analysis of ASR suggested a dimeric oligomer 12,13 , while CD spectroscopy, SSNMR spectroscopy, and EPR spectroscopy suggested trimers in lipids 15 . In addition, a double electron-electron resonance EPR study using site-directed spin labeling of two ASR cysteine mutants confirmed the trimeric nature of ASR in lipids 16 . Also, both AFM images and CD spectrum indicate trimeric form for ASR (right panel of Fig. 5d), which is in good agreement with other experimental results. The dimeric form resolved by crystallography could be caused by crystallization processes of the solubilized molecules. Taken together, XeR likely exists as a trimer, similarly to BR, NpHR and NpSRII. Scientific  In Fig. 5a,b and c, the trimers in the two-dimensional packing show two different heights. It has been reported that the cytoplasmic side of BR is more protruded from the membrane surface than the extracellular side 41 . Thus, the height (image contrast) differences of trimers are probably caused by the height differences between the cytoplasmic (C-terminal) and the extracellular (N-terminal) sides of molecules protruded from the membrane surface.
Among four types of trimeric rhodopsins (NpHR, NpSRII, PoXeR, and ASR), only the CD couplet of NpSRII is very small. This may indicate different packing of NpSRII in the trimer compared to the others. The outer diameter of NpSRII trimers (2.8 ± 0.11 nm) is smaller than those of the others (3.5 ± 0.27 nm for NpHR, 4.0 ± 0.33 nm for PoXeR and 4.8 ± 0.36 nm for ASR). This suggests that NpSRII forms trimeric structures that are tighter packed compared to BR or ASR which have stronger CD couplets. Therefore, the relative orientation of retinals in NpSRII is different from those of BR and ASR, resulting in a loss of the CD couplet in the spectrum.
Correlation between stoichiometry of oligomers and the phylogenetic tree of microbial rhodopsins. We systematically observed various microbial rhodopsins reconstituted into a lipid membrane by HS-AFM and CD spectroscopy, and determined their native oligomeric states. Based on the HS-AFM and CD spectroscopic analyses, the fundamental oligomeric states of microbial rhodopsins can be classified into two groups; one is a trimeric form and the other is a pentameric form. Taking the phylogenetic tree of microbial rhodopsins into consideration, the branching point of the trimer and pentamer likely emerged at an early stage of evolution (Fig. 6). Whereas all pentameric rhodopsins (KR2, GR, FR, KrActR, QsActR, and GPR) originate from eubacteria, archaeal rhodopsins (BR, NpHR, and NpSRII) form trimers. Thus, pentameric and trimeric rhodopsins likely evolved after branching of archaea from eubacteria. The only exception from this rule is XeRs (ASR and PoXeR). ASR and PoXeR form trimers, but are present in the eubacteria, Anabaena sp. PCC7120 and P. oceani, respectively. Although further studies are needed to determine the evolutionary relationship between XeRs and other archaeal rhodopsins, many eubacteria contain genes derived from archaeal species by lateral gene transfer 42 . Thus, the XeR group likely evolved from some eubacteria after lateral transfer of genes of trimeric rhodopsins from archaeal species.
Whole microbial rhodopsins have a seven-transmembrane structure, but the oligomeric states are distinctly different. The factors affecting these different oligomeric structures remain elusive. To address this issue, we compared the amino acid sequences of microbial rhodopsins used in this study (see Supplementary Table 1). However, there are no completely conserved amino acid residues at the protein interfaces, nor specificity among trimers and pentamers. These results strongly suggest that a single specific amino acid does not play a crucial role in oligomer formation, and integration of several non-polar interactions and shapes at the protein interface is likely essential for determining the oligomeric state.
Eukaryotic microbial rhodopsins such as channel rhodopsins exist in a dimeric form, stabilized by a disulfide bond between molecules 43,44 . This indicates that the stability and function of eukaryotic rhodopsin are optimized by dimer formation and that the molecular interactions differ from those in eubacterial and archaeal proteins. The functional units of many channels are tetramers and form pores to transport ions. Additionally, some receptors, such as glutamate receptors (NMDAR and AMPAR) form tetramers, while visual rhodopsin, G protein-coupled receptors, and receptor tyrosine kinases form dimers in lipids. Further experimental studies are required to address these aspects; if an ancestor of microbial rhodopsin is identified, the biological meaning of trimer and pentamer formation in the evolutionary process can be determined.

Materials and Methods
Sample preparation. We used synthesized and E. coli codon-optimized genes for protein expression of KR2, FR, KrActR, QsActR, and PoXeR. All genes were inserted into the multi-cloning site of the pET21a vector (Novagen, Madison, WI, USA) between the NdeI and XhoI restriction sites 40 . The genetic constructs used to express GR, GPR, NpHR, NpSRII and ASR have been described previously 38,39,[45][46][47] . Microbial rhodopsins were expressed in E. coli cells in the following strains: NpHR and NpSRII: BL21(DE3); ASR, KR2 (wt, KR2 278 and N112A), FR and GR: C41(DE3); GPR, KrActR and QsActR: C43(DE3). All proteins were solubilized in 1.5% n-dodecyl-β-D-maltoside (DDM) (Anatrace, OH, USA) and then purified by Co-affinity column (TALON, Qiagen, Hilden, Germany) chromatography 40 . The final protein concentration was adjusted to 2.5 mg/mL for the reconstitution into the lipid membrane. KR2 278 was constructed in pET21a vector with 6 residues deleted at the C-terminus of the KR2 gene, and contained a thrombin cleavage site and 6 × His tag at the 3′ end. The thrombin cleavage site and 6 × His tag was attached at the C-terminus of KR2-derived genes by polymerase chain reaction using PrimeSTAR Max DNA Polymerase (Takara Bio, Shiga, Japan). NdeI and XhoI restriction sites were introduced at the 5′ and 3′ ends, respectively. The gene was then inserted into the pET21a vector. The expression vectors were transformed into the E. coli C43 (DE3) strain. After protein purification, the 6 × His tag was removed by overnight treatment with 10 U/mg protein of thrombin at 4 °C. To remove uncleaved His-tag molecules, thrombin-treated protein was purified using Micro-spin columns (Thermo Scientific, Waltham, MA, USA) containing Co-NTA resin. The flow-through fraction was used in each experiment.
Preparation of lipid-reconstituted microbial rhodopsins. In general, reconstituted proteoliposomes have been used to incorporate membrane proteins into lipids. For AFM imaging, the proteoliposome is ruptured on a solid substrate and then the supported lipid membrane containing proteins is imaged. However, HS-AFM observations of the ruptured proteoliposomes showed large bending of the lipid under dense conditions (see Supplementary Fig. S3a). At dilute protein concentrations, images of proteins on the flat membrane area were unclear, likely because of the rapid diffusion of molecules. Instead, we observed aggregates at the edge of the membrane (see Supplementary Fig. S3b). Membrane bending and aggregation prevented high-resolution imaging by HS-AFM. To overcome these problems, we applied membrane scaffolding proteins (MSP), which were developed for nanodisc technology 48,49 . Because of the limited size of the membranes used with MSP, diffusion of membrane proteins in the supported membrane can be suppressed. Additionally, the membrane can be maintained as a flat sheet by MSP. Therefore, microbial rhodopsins reconstituted into the size-limited membrane can be used to obtain high-quality AFM images, which are crucial for determining the oligomeric states of microbial rhodopsins.
We followed the manufacturer's protocol for the nanodisc (Sigma-Aldrich, St. Louis, MO, USA) with minor modifications. Briefly, for reconstituted lipids, we used a mixture of phospholipids, asolectin from soybean (Sigma-Aldrich, No. 11145). Asolectin (120 μg) was dissolved in chloroform and then evaporated under N 2 gas to completely remove the solvent. Next, the lipids were suspended in 50 μL buffer A (20 mM HEPES-KOH (pH 7.4), 100 mM NaCl 2 , and 4% DDM) and sonicated for ~1 min with a tip sonicator. Next, dissolved membrane proteins (1 nmol) and MSP (50 μL, 1 mg/mL) (MSP1E3D1, Sigma-Aldrich, No. M7074) were added to the lipid suspension and mixed for ~1 h while rotating in the dark at 4 °C. Finally, we added 60 mg Bio-beads SM-2 (Bio-Rad, Hercules, CA, USA, No. 1523920) and dialyzed the samples in detergent overnight at 4 °C. According to the manufacturer's protocol, nanodisc samples should be fractionated on a column to purify the nanodiscs based on size (~10 nm in diameter). Here, we did not purify the reconstituted samples, but obtained flat membranes with limited sizes <30 nm in diameter.

High-speed AFM measurements.
A laboratory-built HS-AFM operated in tapping mode was used 19,26 .
Briefly, cantilever deflection was detected with an optical beam deflection detector using an infrared laser (0.7 mW, 780 nm). The infrared laser beam was focused onto the back side of a cantilever (Olympus, Tokyo, Japan: BL-AC7DS-KU4) covered with a gold film through a ×60 objective lens (Nikon, Tokyo, Japan: CFI S Plan Fluor ELWD 60×) and the reflected infrared laser was detected by a two-segmented PIN photodiode. The cantilever was 6-7 μm long, 2 μm wide, and 90 nm thick. The spring constant, resonant frequency, and quality factor were 0.1-0.2 N/ m, ∼1 MHz, and ∼2 in liquid, respectively. The free oscillation amplitude was ∼1 nm and set-point amplitude was approximately 90% of the free amplitude for the feedback control of HS-AFM observation. An amorphous carbon tip, grown by electron beam deposition, was used as an AFM probe. The length of the electron beam deposition tip was ∼500 nm, and the end radius of the tip was ∼4 nm. For the photo-illumination experiment of KR2, a green laser with a wavelength of 532 nm was periodically irradiated onto the sample through the objective lens during the imaging. As a substrate, a mica surface treated with 0.01% (3-aminopropyl)triethoxysilane (Shin-Etsu Silicone, Tokyo, Japan) was used. HS-AFM experiments were carried out in a buffer solution containing 20 mM Tris-HCl (pH 8.0) and 50 mM NaCl, or 20 mM citrate-NaOH (pH 4.3), 50 mM NaCl at room temperature.

Construction of simulated AFM image.
To compare the HS-AFM images of KR2 with its previously reported crystal structure 9 , we simulated AFM images using a simple hard sphere model. The AFM probe was modeled as a cone shape with a half angle of 5° and an end radius of 0.5 nm. Simulation was carried out for either the N-or C terminus side of the crystal structure of KR2 (PDB code, 4XTN). The simulated images were processed with a low-pass filter with a cut-off wavelength of 2 nm because the spatial resolution of the AFM image is typically limited to approximately 2 nm. Data availability. The raw AFM image files and the raw data of CD spectra are available from the authors on request.