Sensory Rhodopsin I and Sensory Rhodopsin II Form Trimers of Dimers in Complex with their Cognate Transducers^†

Expression and purification of NpSRII:NpHtrII and its C-terminally truncated version NpSRII:NpHtrII_(1-114)

Sensory rhodopsin II and its cognate transducer NpHtrII were expressed and purified from E. coli as described before 28-31. Membranes were solubilized in 2% (w/v) n-dodecyl-β-D-maltopyranoside (DDM, Anatrace) in buffer A (300 mm sodium chloride, 50 mm NaPi, pH 8). Purifications of C-terminally hexa His-tagged proteins were performed by immobilized metal ion affinity chromatography 32 on Ni-NTA resin (Qiagen) using a linear gradient from 30 to 250 mm imidazole in buffer B (300 mm sodium chloride, 50 mm sodium phosphate, 0.05% (w/v) DDM, pH 8). It was followed by exclusion chromatography on a HiLoad 16/600 Superdex 200 pg column (GE Healthcare) with a flow rate of 0.5–1 mL min⁻¹ (300 mm NaCl, 50 mm NaPi, 0.05% (w/v) DDM; pH 8).

For the formation of the NpSRII:NpHtrII complex, the purified proteins were combined with two-fold molar excess of NpSRII and incubated for 12 h at 4°C in buffer B. Subsequently, the mixture was applied to preparative size exclusion chromatography on a HiLoad 16/600 Superdex 200 pg column at 4°C using the same buffer (Figure S2). Fractions containing solely NpSRII and NpHtrII were selected for EM.

Negative stain transmission electron microscopy of NpSRII:NpHtrII and NpSRII:NpHtrII_(1–114)

Purified fractions of NpSRII:NpHtrII and NpSRII:NpHtrII_(1–114) complexes were examined by negative stain electron microscopy. Colloidon/carbon-coated copper grids were treated in a glow discharger (Femto version A, Diener electronic). After incubation with the protein solution, grids were washed three times in double-distilled water and stained with an aqueous 0.7% (w/v) uranyl formate solution 33. Images were acquired using a JEOL JEM 1400 transmission electron microscope with an acceleration voltage of 120 kV. Processing of data acquired for the NpSRII:NpHtrII complex by negative stain electron microscopy was done using the EMAN2 34 and SPARX 35 program suites.

Expression and purification of HsSRI-HtrI

The strategy for cloning, expression and purification of the full-length fusion protein HsSRI-HtrI is essentially the same as for the already described version HsSRI-HtrI_1–52 31, 36, 37. A shortened sopI gene from H. salinarum, which lacks downstream sequences for the last 15 amino acids, was fused upstream of the full-length HsHtrI gene using the megaprimer PCR method 38. A coding sequence for six histidines was added to give a His-tag at the C-terminus of the fusion protein. Both chimera constructs were inserted into the expression vector pET11a (Novagen) under the control of the T7 promotor. Expression was done in E. coli BL21(DE3) RP cells at 37°C in double yeast tryptone medium supplemented with 200 mg L⁻¹ ampicillin. As soon as OD₆₀₀ of 0.6–0.8 was reached, the cells were fed with 1 mm isopropyl-β-D-thiogalactopyranoside and 12 μm of all-trans retinal (Sigma) for 4 h and then harvested by centrifugation (10 000 g).

The pelleted cells were resuspended in 500 mm NaCl, 50 mm 2-(N-morpholino)ethanesulfonic acid (MES) at pH 6.0, and 2 mm EDTA and disrupted in a French press. The insoluble membrane fraction was harvested by centrifugation (130 000 g, 1 h, 4°C) and solubilized overnight at 4°C in 4 m NaCl, 50 mm MES at pH 6.0 and 2% (w/v) DDM (Anatrace). Still insoluble material was removed by an additional centrifugation. The solubilized HsSRI-HtrI variants were further purified on a Ni-NTA agarose column (Qiagen), which was equilibrated with buffer containing 4 m NaCl, 50 mm MES at pH 6.0 and 0.05% (w/v) DDM. Three washing steps were performed: first with 10 bed volumes of the plain buffer, second with buffer plus 20 mm imidazole and finally again with the plain buffer. Elution was done in buffer at pH 4.5, and the purified proteins were immediately diluted with buffer at pH 6.8 to adjust the final pH in the range of 5.5–6.0. Shortly before usage, the HsSRI-HtrI proteins were separated on a gel filtration column (HiLoad 16/60 Superdex 200 pg, GE Healthcare) to get rid of aggregated protein.

Negative stain transmission electron microscopy of HsSRI-HtrI

Gel filtration fractions containing detergent-solubilized HsSRI-HtrI variants were adsorbed for 10–60 s to parlodion carbon-coated copper grids rendered hydrophilic by glow discharge at low pressure in air. Grids were washed with three drops of double-distilled water and stained with two drops of 0.75% (w/v) uranyl formate. Electron micrographs were recorded at a magnification of 50 000× on Eastman Kodak Co. S0-163 sheet films with a Hitachi H-7000 electron microscope operated at 100 kV. Negatives were digitized with a Heidelberg Primescan D 7100 at 4 Å per pixel resolution.

Protein–protein docking

The crystal structure of the NpSRII:NpHtrII dimer in the ground state (PDB access code 1H2S) was taken as an initial structure for the lattice building. We employed the protein–protein docking program M-ZDOCK 39 to predict cyclically symmetric multimeric conformations of the trimers of NpSRII:NpHtrII dimers and, then, of the hexamers of the obtained trimers of dimers. The M-ZDOCK algorithm uses a grid-based fast Fourier transform approach for searching for the best conformation in a space of symmetric (Cn) multimers consisting of a given number n of protomers. The obtained conformations are ranked according to ZDOCK 2.3.2 scoring function 40, which includes atomic contact energy (ACE) potential, shape complementarity and electrostatics terms.

All-atom molecular dynamics simulations

All-atom molecular dynamics simulations were carried out in Gromacs 5.0.5 using the CHARMM36 force field with CMAP corrections and TIP3P water. The initial structure of the transmembrane part of the NpSRII:NpHtrII dimer (the full NpSRII and residues 23 to 82 of NpHtrII) was placed into a 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphoethanolamine: 1-hexadecanoyl-2-(9Z-octadecenoyl)-sn-glycero-3-phospho-(1’-rac-glycerol) (POPE:POPG =3:1) model membrane and solvated using the CHARMM-GUI web server 41. The appropriate numbers of Na⁺ and Cl⁻ ions were placed into the simulation box to neutralize the system and set the ionic strength to 0.15 m. Simulations were done in the NPT ensemble controlled by means of a Nose-Hoover thermostat (T_ref = 303 K, τ_T = 2 ps) and a Parrinello-Rahman barostat (isotropic pressure, p_ref = 1 atm, τ_p = 2 ps, compressibility = 4.5 · 10^–5 bar⁻¹); the time step equaled 2 fs; a Verlet cutoff scheme 42 and PME 43 were used for nonbonded interactions. The initial equilibration (5 ns with the heavy atoms of the lipids and the protein restrained by the harmonic potential with k = 1000 kJ · mol⁻¹ · nm⁻²) was followed by the lipids relaxation simulation (200 ns, the protein backbone restrained with the harmonic potential of the same strength) and, then, by the production simulation (1 μs).

Calculations of the global modes of motion

Principal components analysis (PCA) is a technique, which is commonly used to extract information about global modes of macromolecular mobility from sets of coordinates, for example, MD trajectories 44. PCA requires construction of the symmetric pairwise covariance matrix C between the all of coordinates (e.g. Cartesian coordinates of all Cα-atoms of the protein complex were used in this study resulting in 3N × 3N matrix, where N is the number of residues):

(1)

where x_i represents the ith coordinate and <…> – averaging over all sampled conformations from MD trajectory, which are preliminary superimposed onto the initial conformation. Diagonalization of C gives 3N eigenvectors with the corresponding eigenvalues. The former describe the collective motions, while the latter—their respective amplitudes. Principal components are the projections of the original MD trajectory onto the obtained eigenvectors, and a limited number of them represent a reduced subset of dimensions, which often explain most of the variation present in a system.

The principal components were compared with the collective modes predicted from normal mode analysis (NMA) 45 used in its simplified form, the anisotropic network model (ANM) 46. This technique is computationally similar to PCA, however, instead of the covariance matrix the Hessian matrix (H) is used, which is a matrix of the second-order partial derivatives of the potential energy U by coordinates:

(2)

For the sake of computational efficiency, the macromolecular systems are often chosen to be coarse-grained, usually, by replacing individual residues with the beads corresponding to the Cα-atom positions. The potential energy is simply defined as the sum of uniform harmonic potentials acting between all pairs of the beads separated by less than a given distance cutoff (15 Å in the present study) with the equilibrium distances taken from the initial structure.

Equivalently to the PCA approach, H is diagonalized and the obtained eigenvectors correspond to the collective low-frequency molecular modes (i.e. normal modes), while the eigenvalues characterize the energy cost of corresponding displacements.

The normal modes obtained from ANM are independent on force field selection and tiny structural details. They derive from the overall topology of macromolecules and describe global, low-frequency modes of motions potentially accessible for a system.

The similarity of collective modes acquired by PCA and ANM can be estimated by calculating their pairwise overlap, which is given by the correlation cosine of corresponding eigenvectors 44.

Here, PCA and ANM were performed using the ProDy toolkit 47.

Electron microscopy of trimers of dimers

For the electron microscopy experiments, different samples of HsSRI-HsHtrI and NpSRII:NpHtrII complexes were purified and analyzed. The list of samples included NpSRII complexed to full-length transducer and to a shortened transducer (NpHtrII_1–114). Further constructs encompassed HsSRI fused genetically to its cognate transducer. Also in this case, a full-length and a shortened HsHtrI_1–52 transducer was used (the constructs are shown in Figure S1). From the elution profile of NpSRII:NpHtrII complexes, a peak was observed were one would expect trimers of dimers to elute (Figure S2).

Figures 1 and 2 display selected examples of particles and class averages obtained by EM for HsSRI-HtrI and NpSRII:NpHtrII. The electron micrograph predominantly shows rounded hexagonal structures with diameters of roughly 10 nm. Generally, the samples appear to be heterogeneous, showing different shapes and sizes of objects. It is obvious that this general shape is observed for all samples. As also the shortened transducer forms similar structures (see Fig. 1, lower panels for HsSRI-HtrI_1–52), it seems that solely the membrane domain is responsible for forming the observed particles.

The full-length transducer NpHtrII in complex with the receptor NpSRII gave better results allowing for a preliminary classification (sectors of electron micrographs of NpSRII:NpHtrII and NpSRII:NpHtrII_1–114 are shown in Figure S3). In Fig. 2, eight representative classes for the complex NpSRII:NpHtrII are shown. Panels G and H show a top view of the complexes. Here, the electron density represents a cluster of six spheres arranged in a ring with high extra electron density in the center. The diameter of the arrangement is about 7–8 nm. One could interpret this result by an arrangement of three dimers forming a hexagonal ring. The central electron densities presumably stem from the transducer cytoplasmic domain. Similar observations can be made for HsSRI-HtrI (full length) in comparison with the truncated transducer (1–52, see Fig. 1). Other views of a possible cytoplasmic domain are depicted in panels C and E where an extra density protrudes out of the ring-like arrangement. Overall, the data are not accurate enough to warrant further processing. However, a major conclusion can be drawn: the complex of NpSRII with its transducer forms trimers of dimers in a hexagonal topology (see in particular Fig. 2H) even in detergents. This latter observation indicates quite strong binding, which cannot be disrupted by a mild detergent like DDM.

Molecular modeling of the transmembrane arrays

Reconstruction of the transmembrane arrays

Taking the EM data into account, we started with the prediction of the trimer-of-dimers conformations using rigid-body protein–protein docking, using the available structural data for the NpSRII:NpHtrII complex, as described in the Methods section. The two highest ranked conformations were termed “Y”-shaped and “O”-shaped trimers (Z-DOCK score = 44.27 and 39.48, respectively, with higher score value meaning better match) because their general topologies resemble these letters. In the “Y”-shaped conformation, only one monomer per NpSRII:NpHtrII dimer interacts with the other two dimers within the trimer. The interface consists of helices A and B of one monomer and helix D of the corresponding partner (Figure S4). Contrarily, in the “O”-shaped trimer, both monomers of each dimer contact to the adjacent dimers, leading to a ring-like topology with helices A and B of one monomer interacting with helix E of another one (Fig. 3).

These alternative conformations of trimer-of-dimers were subjected to another round of symmetrically restrained protein–protein docking calculations with C₆ symmetry. The condition of hexameric symmetry was imposed to obtain lattice conformations, which are congruent with the hexagonal arrangement observed in the EM micrographs and that of the cytoplasmic domains of the chemoreceptor clusters 26. The resulting lattice model for the “O”-shaped trimer-of-dimers is shown in Fig. 3 along with the determined intrahexagonal interface. The latter is represented by the contacts between helix A of one monomer with helices D and E of its partner. In the alternative lattice model based on the “Y”-shaped trimer, the intertrimer contacts are formed by helices A and B and helix D as shown in Figure S4.

Overall, the “O”-conformation is in good agreement with the ring-like topology obtained by EM for trimers of dimers. In addition, the putative inter- and intratrimer interfaces determined for this conformation are consistent with the distribution of lipid molecules, which are bound to the protein in the available crystal structures of NpSRII (Fig. 4). We assumed that the regions of the protein surface, which are important for protein–protein interactions, should have lower affinity to lipids and vice versa. In accordance with this assumption, in the crystal structure of NpSRII obtained without the cognate transducer (PDB access code 3QAP, 48) there are no resolved lipids in the vicinity of helices F and G, which form the interface with the NpHtrII transducer. Another region lacking bound lipids is an area comprising helices D and E. These helices are involved in formation of both, the interdimer and the intertrimer contacts in the “O” model.

The transmembrane array is concave

The distance between two neighboring trimers of dimers within a hexameric cell of the tightly packed transmembrane lattice equals 9.5 nm for the “O”-shaped model (and 9.2 nm for the alternative “Y” lattice). This value is slightly larger compared to similar distances measured between the neighboring cytoplasmic tips of receptors in the honeycomb lattices formed by chemoreceptors with the CheA/CheW baseplate (varying between 6.9 nm 49 and 8.0 nm 26), which is believed to be highly conserved across multiple bacterial and archaeal species and thus unlikely to exceed these limits 26. We hypothesized that this difference can represent an intrinsic property of the receptors, which assemble in concave transmembrane lattices. The concavity of the receptor arrays facilitates their localization preferably at the cell poles 50, 51. A rough estimation of the curvature of the membrane-inserted arrays is given in Fig. 5. The calculated inverse membrane curvature equals 150–300 nm, which is in agreement with the experimentally determined curvature values (~400 nm) in the polar regions of the cylinder-shaped bacterial cells 52.

Global motions of the NpSRII:NpHtrII dimer suggest a model of receptor cooperativity

Above, we have shown that the organization of the cytoplasmic tips of the transducers in hexagonal arrays, which form a platform for binding the histidine-kinase CheA and the adapter protein CheW, necessitates formation of similarly highly ordered lattices of the NpSRII:NpHtrII membrane domain. Such densely packed two-dimensional clusters consisting of identical subunits potentially can function in a highly cooperative manner, which might be the basis for signal amplification 53-55. If the activation of a single receptor causes significant changes of its overall structure and/or dynamics, these changes can allosterically spread throughout the lattice, leading to a cooperative behavior 56,which might account for the cooperativity in the cytoplasmic baseplate 57. However, the existence of equally dense lattices of the transmembrane domains suggests that they can act in a similar way, thus providing an additional layer for signal amplification.

In order to study plausible cooperativity at the transmembrane level, we applied principal component analysis (PCA) to a 1-μs-long unbiased MD trajectory of a model of the transmembrane part of the NpSRII:NpHtrII_1-82 dimer (the length of transducer in the 1H2S X-ray structure) in a lipid bilayer to reveal possible global motional modes of the NpSII:NpHtrII dimer. The collective motions detected by PCA (principal components, PCs) were compared with the global modes predicted by normal mode analysis (NMA) (shown in Fig. 6).

Both methods point to the presence of similar global motions as confirmed by the high overlap values between the PCA and ANM modes (Table S1).

The first two normal modes correspond to global twisting and wagging of the SRII molecules. The third normal mode largely overlaps with the PC1 of PCA analysis. This motional mode is of special interest as it closely resembles the transition between the “U”- and “V”-shaped conformations of the NpSRII:NpHtrII complex, which was previously argued to represent a functionally relevant global conformational change accompanying the activation process 58, 59. Indeed, the histogram of the trajectory projection on the PC1 as well as the histograms of the intermonomer SRII-SRII angle (between helices F) and distance (between the COMs of the cytoplasmic segment of helix F in both monomers) distributions are clearly bimodal indicating the presence of two conformations. While one of them remains structurally close to the “V”-shaped conformation, another one can be attributed to the “U”-shaped conformer (see Figure S5).

Interestingly, the membrane insertion energies, calculated for the “U”- and “V”-shaped dimers with the help of the PPM web server 60, differ by only ~5 kT (−503.3 and −516.3 kJ mol⁻¹, respectively), suggesting that the two conformations can easily interconvert (e.g. upon absorption of a photon, which energy is about 20 times higher). However, the mode of interaction of the transmembrane region with the HAMP domain seems to be different in the “U”- and “V”-shaped conformations 61. The latter implies feasible coupling between conformation of the HAMP and orientation of the SRII, which may facilitate a global cooperative response of the receptor:transducer clusters upon their activation.