Kinetic, Energetic, and Mechanical Differences between Dark-State Rhodopsin and Opsin

SUMMARY

Rhodopsin, the photoreceptor pigment of the retina, initiates vision upon photon capture by its covalently linked chromophore 11-cis-retinal. In the absence of light, the chromophore serves as an inverse agonist locking the receptor in the inactive dark state. In the absence of chromophore, the apoprotein opsin shows low-level constitutive activity. Toward revealing insight into receptor properties controlled by the chromophore, we applied dynamic single-molecule force spectroscopy to quantify the kinetic, energetic, and mechanical differences between dark-state rhodopsin and opsin in native membranes from the retina of mice. Both rhodopsin and opsin are stabilized by ten structural segments. Compared to dark-state rhodopsin, the structural segments stabilizing opsin showed higher interaction strengths and mechanical rigidities and lower conformational variabilities, lifetimes, and free energies. These changes outline a common mechanism toward activating G-protein-coupled receptors. Additionally, we detected that opsin was more pliable and frequently stabilized alternate structural intermediates.

INTRODUCTION

Rhodopsin is the light receptor in the rod outer segment (ROS) disc membranes of photoreceptor cells in the retina and a prototypical G-protein-coupled receptor (GPCR). Like other GPCRs, rhodopsin has seven transmembrane α helices (TMH) that are connected by intracellular and extracellular polypeptide loops and expose functionally important extracellular N termini and cytoplasmic C termini. Unlike other seven transmembrane α-helical GPCRs, rhodopsin covalently binds the light-sensitive chromophore 11-cis-retinal at lysine 296 (Lys296) in the seventh transmembrane α helix (TMH7). This covalently bound 11-cis-retinal is an inverse agonist that locks theGPCRin the inactive dark state. In the absence of the chromophore, the apoprotein opsin exhibits a lowlevel of constitutive activity (Melia et al., 1997), which causes retinal degeneration in Leber congenital amaurosis or vitamin A deficiency (Fain and Lisman, 1993; Woodruff et al., 2003).

A recent crystal structure of opsin reveals that the transmembrane region undergoes significant α-helical rearrangements compared to dark-state rhodopsin (Okada et al., 2004; Palczewski et al., 2000; Park et al., 2008). For instance, prominent structural changes are observed at the conserved D(E)RY and NPXXY motifs, resulting in the breakage of conformational constraints that lock the receptor in the inactive state. Outward tilting of the cytoplasmic portion of TMH6 and the elongation of TMH5 suggest formation of further interhelical interactions. These changes are similar to those occurring in a crystal structure of the active metarhodopsin II state of rhodopsin (Choe et al., 2011). However, for structure determination by X-ray crystallography, membrane proteins must first be extracted from their native membrane by solubilization in detergent and then packed into a crystal lattice. Structure determination requires that all (or most) of the crystallized membrane proteins adopt a single conformation, which may or may not represent the native conformation (Hite et al., 2007). Thus, it is unclear whether the crystal structure of opsin represents the structure the receptor adopts in the native membrane.

To better understand the mechanisms by which the absence of 11-cis-etinal causes constitutive activity in opsin under physiological conditions, we utilized single-molecule force spectroscopy (SMFS) and dynamic SMFS (DFS). These nano-technological methods quantify and localize molecular interactions of membrane proteins in their native membrane and physiologically relevant buffer (Bippes et al., 2009; Cisneros et al., 2005, 2008; Ge et al., 2011; Kedrov et al., 2005, 2007b, 2010; Park et al., 2007; Zocher et al., 2012b).We have previously applied SMFS and DFS to investigate the effects of ions, posttranslational modifications, and pathogenic mutations on rhodopsin in native ROS disc membranes (Kawamura et al., 2012; Park et al., 2007, 2009). SMFS and DFS have also shown that inter- and intramolecular interactions stabilize dark-state wild-type rhodopsin from bovine and mouse retina in a similar manner (Kawamura et al., 2010).

In the current study, opsin embedded in native ROS disc membranes was obtained from Rpe65 knockout (Rpe65−//−) mice that lack retinal pigment epithelium-specific 65 kDa protein (RPE65) (Redmond et al., 1998) and are an animal model for Leber congenital amaurosis (Caruso et al., 2010; Fan et al., 2008). The enzyme RPE65 is required to regenerate 11-cisfrom all-trans-retinal, and therefore the light receptor is present as opsin in photoreceptor cell membranes of Rpe65−//− mice. ROS disc membranes from this mouse model were investigated by SMFS and DFS to determine the differences in kinetic, energetic, and mechanical properties between dark-state rhodopsin and opsin. The differences observed outline acommon mechanism toward activating G-protein-coupled receptors.

RESULTS

ROS Disc Membranes from Rpe65–/– Mice Contain Properly Folded Opsin

RPE65 is required for the regeneration of 11-cis-retinal (Kiser et al., 2012). In the absence of RPE65, mice are unable to synthesize 11-cis-retinal (Redmond et al., 1998). By normal-phase HPLC, we confirmed the absence of retinal in retina extracts of Rpe65−//− mice, whereas both 11-cis-retinal and all-transretinal were present in retina extracts of wild-type mice (Figure 1A). Thus, only the apoprotein opsin is present in ROS disc membranes from the retina of Rpe65−//− mice.

Figure 1. Opsin Expressed in Rpe65–/– Mice Lacks 11-cis-Retinal.

(A) UV/Vis absorbance spectroscopy (at 325 nm) of retinoids extracted from Rpe65−//− (solid line) or wild-type (dashed line) mice and separated by normal-phase high-performance liquid chromatography (HPLC). Samples from wild-type mice displayed absorbance peaks representing anti and syn isomers of 11-cis-retinal oximes (1, 1′ ) and anti and syn isomers of all-trans-retinal oximes (2, 2′ ). No retinals were observed in extracts from Rpe65−//− mice. To avoid overlap, traces for extracts from wild-type mice and Rpe65−//− mice were offset on the y axis. Other predominant peaks include the buffer exchange point (*) and a nonretinoid peak (**) with a maximum absorbance between 290– 300 nm.

(B) UV/V is absorbance spectroscopy of opsin purified from Rpe65−//− mice (black line), opsin purified from Rpe65−//− mice, and reconstituted with 9-cis-retinal (gray line) and rhodopsin purified from wild-type mice (dashed line). Each experiment was repeated at least three times.

The presence of opsin in ROS of Rpe65−//− mice was further tested by UV/Vis absorbance spectroscopy. Receptors were affinity-purified from the retina of wild-type or Rpe65−//− mice. Purified rhodopsin from wild-type mice displayed absorbance maxima at 280 and ≈500 nm (Figure 1B). The absorbance maximum at 500 nm originates from 11-cis-retinal bound via a Schiff base linkage to the apoprotein opsin (Wald, 1968). In contrast to purified wild-type rhodopsin, purified receptors from the retina of Rpe65−//− mice only displayed an absorbance maximum at 280 nm, indicating the presence of opsin and absence of 11-cis-retinal.

The ability of opsin in ROS disc membranes of Rpe65−//−mice to form isorhodopsin was tested by reconstituting ROS membranes with an isomer of 11-cis-retinal, 9-cis-retinal. Isorhodopsin (i.e., opsin conjugated with 9-cis-retinal) has functional properties similar to wild-type rhodopsin (Hubbard and Wald, 1952). Receptors from 9-cis-retinal reconstituted ROS membranes from Rpe65−//− mice were affinity-purified and their UV/Vis absorbance spectra determined. The 9-cis-retinal reconstituted and purified receptors exhibited an absorbance maximum at 487 nm (Figure 1B), which is characteristic of isorhodopsin (Hubbard and Wald, 1952). Thus, opsin expressed in Rpe65−//− mice is properly folded into a functional three-dimensional structure and can bind chromophore similarly to wild-type rhodopsin. These results demonstrate that ROS disc membranes from Rpe65−//−mice are a native source of the apoprotein opsin and suitable for investigation by SMFS.

Mechanical Unfolding of Individual Opsins

We have previously characterized the interactions stabilizing dark-state rhodopsin in isolated ROS disc membranes from wild-type mice using SMFS (Kawamura et al., 2012). We use this SMFS data as a reference for the inactive state of the receptor. We conducted SMFS on opsin of ROS disc membranes isolated from Rpe65−//− mice in Ringer’s buffer at ≈28°C, employing the same experimental conditions and AFM-based equipment used to characterize dark-state rhodopsin. SMFS was conducted as follows: the tip of the AFM cantilever was pushed to the extracellular surface of ROS discs to promote the nonspecific attachment of the N-terminal end of the receptor to the AFM tip (Figure S1 available online). The AFM cantilever was then withdrawn to stretch the N-terminal polypeptide. As the distance between tip and surface increased, a mechanical force was applied to the receptor. The response of the receptor to this force was recorded in a force-distance (F-D) curve (Figures 2A and 2B). With increasing distance the receptor unfolds in sequential steps, each occurring at a distinct force. Each force peak of an F-D curve represents the unfolding of a single structural segment that stabilizes the receptor. The pattern of force peaks in an F-D curve characterizes the sequential unfolding of structural segments, and the height of the force peak reflects the strength of the interactions stabilizing a structural segment. The force peaks of rhodopsin clustered into three distinct regions (Figure 2A). Opsin displayed the same cluster of force peaks found in rhodopsin, although additional force peaks at irregular distances were detected (Figure 2B). These irregular force peaks occurred less frequently than the clustered force peaks.

Figure 2. F-D Curves Recorded upon Mechanically Unfolding Dark-State Rhodopsin or Opsin.

(A and B) Individual F-D curves recorded during the mechanical unfolding of single rhodopsin (blue) and opsin (red) molecules. Force peaks represent the rupture of molecular interactions that stabilize unfolding intermediates of the receptor. F-D curves shown were recorded at a pulling velocity of 300 nm/s.

(C and D) Superimposed F-D curves of rhodopsin (n = 799) or opsin (n = 628) represented as density plots with the color spectrum (see color scale bar), indicating the frequency of force peaks. Rhodopsin data were taken from Kawamura et al. (2012).

See also Figures S1, S2, S4, and S5.

To determine which force peaks occurred reproducibly, F-D curves of rhodopsin and opsin were superimposed and displayed as density plots (Figures 2C and 2D). Superimposed F-D curves displayed high-density regions (blue areas, Figures 2C and 2D), which represent force peaks detected with high frequency. The patterns of these high-density regions of rhodopsin and opsin were similar and did not depend on the pulling velocities at which the receptors were unfolded (Figure S2). However, the opsin superimpositions displayed more data points from irregular force peaks (red area, Figure 2D). The same effect is observed for superimposed F-D curves recorded at different pulling velocities (Figure S2). These superimpositions show that opsin displays similar reoccurring force peaks as rhodopsin, but additionally frequently display irregular force peaks. Because each force peak in a F-D curve is the result of interactions established by the unfolding receptor (Kedrov et al., 2007a), the irregular force peaks show that, compared to dark-state rhodopsin, opsin more frequently establishes randomly distributed interactions.

Automated Statistical Analysis Using a Gaussian Mixture Model

To determine the mean contour lengths at which force peaks occurred, every force peak of every F-D curve was fitted using the worm-like chain (WLC) model (see the Experimental Procedures) (Oesterhelt et al., 2000; Rief et al., 1997). Histograms show the frequency of force peaks at all contour lengths (Cisneros et al., 2008; Puchner et al., 2008) of dark-state rhodopsin and opsin (Figure 3A). Recurring force peaks at particular contour lengths indicate force peak classes that reveal the unfolding of stable structural segments. The histogram of opsin displayed a pattern similar to that of rhodopsin; however, some force peak classes were less distinct in the opsin histogram.

Figure 3. Contour Length Histograms of Force Peaks Detected upon Unfolding of Dark-State Rhodopsin and Opsin.

(A) Histograms of contour lengths at which force peaks were detected in F-D curves recorded from dark-state rhodopsin (blue) and opsin (red). Bin size, 1 amino acid (aa). A total of 799 and 628 F-D curves were analyzed for rhodopsin and opsin, respectively. The frequency difference, Δ, between rhodopsin and opsin is indicated below the histogram. Arrows indicate the contour length of force peak classes as determined by the Gaussian mixture model (B–C).

(B and C) Gaussian mixture model fits to determine the contour length and width of force peak classes in each histogram. In the rhodopsin histogram, ten force peak classes were determined. Each force peak class is represented by a weighted Gaussian model component and is indicated by a unique color. Color segments on the x axis indicate the range of contour lengths determined for each force peak class by the Bayes classifier. Boundaries of each force peak class are listed in Table 1. The same ten Gaussian model components fit the opsin histogram. The sums of weighted contour lengths detected for all force peaks, that is, the full mixture models, are shown as black lines. Gray bars denote histograms shown in (A). Dashed lines indicate the uniform baseline noise determined by Equation 2.

See also Figures S2–S5.

In our previous SMFS analysis of dark-state rhodopsin, the mean contour length of each force peak class of a contour length histogram was determined by manually fitting a Gaussian distribution (Kawamura et al., 2010, 2012). The dense histograms make this conventional approach difficult because several classes of force peaks overlap. Therefore, these force peak classes may be overlooked and their overlapping distributions prevented proper fitting. To alleviate this issue, we implemented an automated statistical algorithm that uses a Gaussian mixture model (Equation S1) to analyze opsin and rhodopsin histograms (Supplemental Experimental Procedures). This algorithm estimates the location and spread of all significant classes of force peaks simultaneously (Figures 3B and 3C). We restricted the analysis to contour lengths ranging from 0 to 265 aa, which resulted in the removal of <1% outliers and facilitated reliable estimation of the background noise (Supplemental Experimental Procedures). Using the Bayesian information criterion as a statistical model selection procedure (Supplemental Experimental Procedures), we found ten classes of force peaks in both rhodopsin and opsin histograms, whereas only nine classes of force peaks were detected using the manual approach (Kawamura et al., 2012).

Assigning Structural Segments Stabilizing Dark-State Rhodopsin and Opsin

The Gaussian mixture model fit provided the mean contour lengths and standard deviations of all force peak classes (Table 1). The mean contour length of a given force peak class estimates the number of amino acid residues that have been unfolded and stretched by the AFM tip. This contour length localizes the amino acid position of the beginning of a stable structural segment. Mapping the stable structural segments to secondary structural models of rhodopsin and opsin revealed which structural regions of the receptors were stabilized (Figure 4). The positions of structural segments stabilizing rhodopsin and opsin agreed within one standard error of the mean (Table 1). Stable structural segments [N1], [N2], [E1], [H3-C2-H4-E2], [H5-C3-H6.1], [H6.2-E3-H7], [H8], and [CT] were identical to those detected previously for dark-state bovine and mouse rhodopsin (Kawamura et al., 2010). However, the newly introduced fitting algorithm separated the previously observed structural segment [H1-C1-H2] (Kawamura et al., 2010, 2012) into two structural segments [H1] and [C1-H2].

Table 1.

Mean Contour Lengths and SD of Force Peak Classes that Assign Structural Segments Stabilizing Dark-State Rhodopsin and Opsin

Stable Structural Segment	Rhodopsin, Mean ± SD (aa)	Opsin, Mean ± SD (aa)
[N1]	19.0 ± 4.5	18.6 ± 5.1
[N2]	27.4 ± 2.1	27.5 ± 1.7
[H1]	38.1 ± 4.9	37.5 ± 4.8
[C1-H2]	55.0 ± 6.7	55.0 ± 6.5
[E1]	97.4 ± 3.7	97.8 ± 3.1
[H3-C2-H4-E2]	110.0 ± 3.9	108.4 ± 3.7
[H5-C3-H6.1]	125.2 ± 8.6	121.9 ± 11.5
[H6.2-E3-H7]	176.4 ± 6.8	176.9 ± 7.4
[H8]	219.9 ± 4.6	223.7 ± 10.9
[CT]	239.6 ± 6.5	240.0 ± 5.7

As described in the Experimental Procedures, each force peak class is described by Gaussian distribution with mean length μ_s and variance σ_s². See also Figures S3 and S4.

Figure 4. Stable Structural Segments Detected in Dark-State Rhodopsin and Opsin.

Secondary structure of rhodopsin and opsin with amino acid residues colored to highlight the stable structural segments ([N1], red), ([N2], blue), ([H1], purple), ([C1-H2], dark pink), ([E1], cyan), ([H3-C2- H4-E2], orange), ([H5-C3-H6.1], dark green), ([H6.2-E3-H7], yellow), ([H8], light pink), and ([CT], light green). Black numbers denote the amino acid residue at which a stable structural segment starts. Numbers in parentheses are contour lengths obtained by fitting contour length histograms (Table 1). Differences between values are due to the disulfide bridge (S-S) or membrane compensation (Experimental Procedures). The chromophore 11-cis-retinal forms a Schiff-base linkage at Lys296 (circled black).

To test if the Gaussian mixture model provides consistent results, we reanalyzed SMFS data from a constitutively active G90D rhodopsin mutant (Kawamura et al., 2012). The same ten force peak classes detected for wild-type rhodopsin and opsin were also detected for G90D rhodopsin (Figure S3). However, a difference was observed at the mean contour length that assigns an additional stable structural segment [C1-H2], which consequently harbors the G90D mutation. This comparison demonstrates the applicability of the Gaussian mixture model and the detection of masked differences. Importantly, the Gaussian mixture model is a faster and more objective approach to detect the position of force peak classes in SMFS data, thereby avoiding subjective human judgments and possible errors.

Spurious Interactions within Rhodopsin and Opsin

Our SMFS experiments show that opsin stabilizes the same structural segments as rhodopsin. Although the F-D curves recorded for opsin showed overall consistency with F-D curves of rhodopsin, irregularly distributed force peaks appeared as dispersed red regions in superimposed F-D curves of opsin (Figures 2C and 2D). These irregular force peaks, which added noise to the superimposed F-D curves and the contour length histograms of force peaks, distributed over the entire contour length of the opsin polypeptide. The dissimilarities between rhodopsin and opsin can also be seen in the contour length histograms of force peaks (Figure 3A) and in the average force and standard deviation of all force peaks detected (Figures S4 and S5). In addition to the assignment of force peak classes, the Gaussian mixture model (Equation 1) includes a uniformly distributed component accounting for background noise, π₀g (Supplemental Experimental Procedures). The level of noise is denoted in the histograms by a horizontal dashed line (Figures 3B and 3C). F-D curves recorded from opsin (π_0,opsin = 0.38) exhibited ≈30% more noise than F-D curves recorded from dark-state rhodopsin (π_0,rhodopsin = 0.29; p < 1.7 × 10⁻³). This increased noise quantifies that opsin F-D curves show more irregularly distributed force peaks, which reflect irregularly positioned interactions within the receptor.

Kinetic, Energetic, and Mechanical Properties of Stable Structural Segments

The unfolding force of a protein depends on the pulling velocity of the experiment (Evans and Ritchie, 1997). Thus, the apparent strengths of interactions (e.g., force) that stabilize a structural segment are loading rate dependent. This relationship between unfolding force and loading rate provides information about the underlying unfolding free-energy barrier. The unfolding free-energy barrier was used to describe the energetic, kinetic, and mechanical properties of each structural segment stabilizing rhodopsin or opsin in the absence of an externally applied force (Figure S6). To quantify these properties, we conducted DFS of opsin embedded in native ROS disc membranes and collected F-D curves at six different loading rates (i.e., pulling velocities of 300, 700, 1,500, 3,000, 4,500, and 6,000 nm/s) (Figure S2). This opsin DFS data set was analyzed along with the DFS data previously obtained for dark-state rhodopsin (Figure 5).

Figure 5. DFS Plots of Dark-State Rhodopsin and Opsin.

Each DFS plot shows the dynamic behavior of a stable structural segment from dark-state rhodopsin (blue) and opsin (red). Plotted is the mean unfolding force of each stable structural segment versus loading rate. Slanted ellipses indicate one standard error of each data point. The Bell-Evans model was fitted (solid lines) to obtain the unfolding energy barrier parameters (Experimental Procedures; Equations S14 and S15). Dark- and light-colored regions indicate fitting confidence intervals of one (68%) and two (95%) standard deviations, respectively.

See also Figures S2, S6, and S7.

DFS plots were generated by plotting the mean unfolding force of each stable structural segment against the logarithmic loading rate (Figure 5). DFS plots of every stable structural segment showed log-linear relationships between force and logarithmic loading rate, indicating a two-state unfolding process by which a folded structural segment overcomes a single-energy barrier to unfold (Figure S6) (Bell, 1978; Evans and Ritchie, 1997). Linear regression was used for fitting the DFS plots, and the error propagation of measurement uncertainties was computed using Monte Carlo simulations (Supplemental Experimental Procedures). This approach has the advantage of correctly accounting for nonlinearities and correlations among measurement errors. Fitting the DFS data to the Bell-Evans model (Equation 3) approximates the equilibrium unfolding rate k₀ in the absence of applied force and the distance separating the folded state and transition state x_u. The reciprocal of the unfolding rate k₀ describes the lifetime of a structural segment, whereas x_u approximates the width of the energy valley that hosts the folded state. The number of conformational substates (i.e., conformational variability) that can be hosted by an energy valley depends on this width. Hence, a structural segment characterized by a small x_u shows lower conformational variability than one having a larger x_u. These parameters were determined for every stable structural segment of dark-state rhodopsin and opsin (Table 2; Figure 6).

Table 2.

Unfolding Energy Barrier Parameters and Mechanical Rigidity Characterizing the Properties of Structural Segments Stabilizing Rhodopsin and Opsin

Stable Structural Segment	xu (nm)		k0 (s−1)
	Rhodopsin	Opsin	Rhodopsin	Opsin
[N1]	1.24 ± 0.28	0.37 ± 0.13	0.00 ± 0.00	3.34 ± 4.94
[N2]	0.97 ± 0.35	0.37 ± 0.08	0.00 ± 0.00	0.36 ± 0.65
[H1]	0.60 ± 0.16	0.30 ± 0.03	0.06 ± 0.15	0.92 ± 0.81
[C1-H2]	0.58 ± 0.20	0.50 ± 0.66	0.85 ± 1.34	1.01 ± 1.91
[E1]	0.64 ± 0.39	0.41 ± 0.09	0.03 ± 0.08	0.12 ± 0.22
[H3-C2-H4-E2]	0.43 ± 0.11	0.31 ± 0.03	0.09 ± 0.17	0.11 ± 0.11
[H5-C3-H6.1]	0.46 ± 0.06	0.31 ± 0.07	0.04 ± 0.06	0.25 ± 0.41
[H6.2-E3-H7]	0.48 ± 0.13	0.34 ± 0.06	1.12 ± 1.29	1.05 ± 0.98
[H8]	0.38 ± 0.03	0.26 ± 0.02	0.34 ± 0.21	1.22 ± 0.42
[CT]	0.47 ± 0.09	0.33 ± 0.05	0.11 ± 0.15	0.37 ± 0.34
	ΔG (kBT)		κ (nm s−1)
[N1]	38.0 ± 5.0	20.8 ± 2.3	0.22 ± 0.06	1.41 ± 0.47
[N2]	34.8 ± 5.8	23.3 ± 2.5	0.35 ± 0.12	1.52 ± 0.43
[H1]	25.9 ± 3.4	21.2 ± 1.0	0.65 ± 0.20	1.94 ± 0.33
[C1-H2]	22.4 ± 2.7	23.3 ± 4.3	0.62 ± 0.23	1.03 ± 0.47
[E1]	28.1 ± 4.8	24.5 ± 2.6	0.67 ± 0.26	1.31 ± 0.39
[H3-C2-H4-E2]	25.1 ± 3.0	23.3 ± 1.0	1.26 ± 0.40	2.00 ± 0.31
[H5-C3-H6.1]	24.8 ± 1.6	23.5 ± 2.4	1.02 ± 0.21	2.22 ± 0.70
[H6.2-E3-H7]	21.5 ± 1.9	21.2 ± 1.2	0.88 ± 0.32	1.62 ± 0.45
[H8]	22.0 ± 0.7	20.6 ± 0.4	1.30 ± 0.18	2.51 ± 0.26
[CT]	24.0 ± 1.8	22.2 ± 1.1	0.96 ± 0.26	1.78 ± 0.39

Energy barrier parameters: x_u, k₀, and ΔG.

Mechanical rigidity: κ. Mean values of estimates are listed with their SD. See also Tables S1 and S2.

Figure 6. Conformational, Kinetic, Energetic, and Mechanical Properties of Dark-State Rhodopsin and Opsin.

(A and B) Structural models of (A) rhodopsin (Protein Data Bank [PDB] ID code 1U19) and (B) opsin (PDB ID code 3CAP) mapped with stable structural segments shown in Figure 4. The following structural models have been colored to map the distance separating the folded from the transition state, x_u (e.g., conformational variability), the transition rate, k₀ (reciprocal of lifetime), the unfolding free energy, ΔG, and the mechanical spring constant, κ. Color bars specify the range of values taken from Table 2. Maps were created using PyMOL Molecular Graphics System (v.1.2r3pre, Schrödinger).

See also Figure S7.

Opsin generally exhibited higher k₀ and lower x_u values than rhodopsin (Table 2). For stable structural segments, the distance from the folded to the transition state x_u ranged from 0.38 nm ([H8]) to 1.24 nm ([N1]) for rhodopsin and from 0.26 nm ([H8]) to 0.50 nm ([C1-H2]) for opsin (Table 2). In the absence of an externally applied force, the unfolding rates k₀ ranged from 0.00 s⁻¹ ([N1] and [N2]) to 1.12 s⁻¹ ([H6.2-E3-H7]) for structural segments stabilizing rhodopsin and from 0.11 s⁻¹ ([H3-C2-H4-E2]) to 3.34 s⁻¹ ([N1]) for opsin (Table 2).

Using k₀ and x_u, the height of the unfolding energy barrier ΔG and the mechanical spring constant κwere calculated for every structural segment (Table 2; Equations S18 and S19). ΔG denotes the height of the unfolding free-energy barrier stabilizing a structural segment, whereas κ describes its mechanical rigidity. In rhodopsin, structural segments exhibited unfolding energy barrier heights ranging from 21.5 k_BT ([H6.2- E3-H7]) to 38.0 k_BT ([N1]) and mechanical spring constants ranging from a minimum flexibility of 0.22 N/m ([N1]) to a maximum rigidity of 1.30 N/m ([H8]). In opsin, the structural segments had unfolding energy barrier heights ranging from 20.6 k_BT ([H8]) to 24.5 k_BT ([E1]) and mechanical spring constants ranging from 1.03 N/m ([C1-H2]) to 2.51 N/m ([H8]).

DISCUSSION

In the current study, we compared the molecular interactions stabilizing the apoprotein opsin and dark-state rhodopsin to determine the impact of the absence of the chromophore 11-cis-retinal and the mechanisms of constitutive activity in opsin. Therefore, we applied single-molecule experiments to localize and quantify energetic, kinetic, and mechanical properties of opsin embedded in native ROS discs membranes from Rpe65−//− mice. Significant improvements, introduced for the statistical analyses of SMFS and DFS data, revealed more reliable and more detailed structural insights into the molecular interactions stabilizing the receptor. These improved statistical analyses were applied to data collected newly for opsin and to data collected previously for dark-state wild-type rhodopsin and for dark-state mutant G90D rhodopsin. We first discuss the statistical approaches developed to improve SMFS and DFS data analysis and then discuss the biological significance of our experimental results.

Improved Statistical Analysis of Contour Length Histograms of Force Peaks

In most SMFS experiments, the force peaks detected when unfolding membrane proteins are manually grouped into classes (Bippes et al., 2009; Ge et al., 2011; Kawamura et al., 2010, 2012; Kedrov et al., 2007a; Park et al., 2007; Sapra et al., 2006, 2008). To determine the mean contour lengths of all force peak classes, a histogram of contour lengths of all force peaks is created (Figure 3), and every force peak class is fitted by a Gaussian distribution. This approach can introduce a bias because each force peak class is considered individually, and poorly resolved or overlapping force peak classes can be overlooked. To overcome these limitations, we have applied a single Gaussian mixture model to fit all force peak classes in contour length histograms simultaneously. The Bayesian information criterion indicated the presence of ten reoccurring force peak classes in both opsin and rhodopsin data sets (Figure 2). To test the validity of this approach, we fitted contour length histograms of force peaks previously recorded from ROS disc membranes containing G90D mutant rhodopsin (Figure S3). The same ten force peak classes as observed for opsin and wild-type rhodopsin were revealed. Using the Bayes classifier, the Gaussian mixture model can separate overlapping force peak classes. With this approach, we identified two additional populations of force peak classes and assigned their contour lengths to introduce two structural segments, namely, [H1] and [C1-H2], which replaced the former segment [H1-C1-H2] (Figure 4) (Kawamura et al., 2012). We conclude that applying the Gaussian mixture model to statistically analyze all force peaks in contour length histograms is sensitive and efficient. In addition, the Gaussian mixture model also estimates the fluctuation of the force peaks. For opsin, this analysis quantified a noise increase of ≈30% that was likely caused by irregular interactions within the receptor.

Improved Statistical Analysis of DFS Data

The determination of energy landscape parameters is sensitive to the analytical method applied (Friedsam et al., 2003). Thus, when comparing DFS experiments, they should be analyzed using the same methods. Here, we have modified our approach to analyze the DFS plot of every stable structural segment and introduced an automated procedure that, unlike previous approaches, includes a Monte Carlo simulation to propagate correlated errors. This method provides meaningful confidence intervals to compare the differences between DFS slopes obtained from different samples (Figure 5). For example, in the case of structural segments [N1], [N2], and [H5-C3-H6.1], the wider confidence intervals in opsin indicates a larger variety of interaction strengths. Accordingly, the x_u, k₀, ΔG, and κ values determined for stable structural segments of rhodopsin differed between manually and automated fitting methods (Tables 2, S2, and S3) (Kawamura et al., 2010, 2012). One reason for this discrepancy is that instead of propagating fitting uncertainties from k₀ and x_u to ΔG by means of a linear approximation, here we determined the errors of ΔG, as well as ofκ, x_u, and k₀ directly from DFS fits accounting for nonlinearities and correlated errors (Equations S16, S17, and S18).

Interactions in Rhodopsin and Opsin Stabilize Similar Structural Segments

The mean contour lengths of the ten force peak classes for rhodopsin and opsin differ by less than one standard deviation (Table 1; Figure 3). Thus, the absence of 11-cis-retinal does not alter the number or position of stable structural segments in opsin. This result is consistent with observations made on the human β₂ adrenergic receptor (β₂AR), which belongs to the same GPCR subfamily as rhodopsin (Zocher et al., 2012a). It was found that the β₂AR stabilizes the same structural segments independent of the presence of an agonist, antagonist, or inverse agonist. However, the kinetic, energetic, and mechanical properties of the stable structural segments were modulated upon ligand binding to the β₂AR. Similarly, we detected that although dark-state rhodopsin and opsin stabilized the same structural segments, the segments of opsin showed significantly different kinetic, energetic, and mechanical properties (Figure 6).

Opsin Establishes Significantly Stronger Interactions

Although the locations of stable structural segments are the same for dark-state rhodopsin and opsin, the unfolding forces detected for opsin were consistently higher (Figure 2). Higher unfolding forces of opsin were detected at all pulling velocities (Figures S4 and 5). Thus, interactions stabilizing the structural segments of opsin are of higher mechanical strengths than those stabilizing rhodopsin.

Stable Structural Segments in Opsin Reduce Conformational Variability

Compared to rhodopsin, all stable structural segments of opsin had a smaller distance to the transition state x_u (Table 2; Figure 6). Because a decrease in x_u describes the narrowing of the energy valley in which the stable structural segment resides, this finding implies that the conformational variability of structural segments stabilizing opsin is less than in rhodopsin.

Stable Structural Segments in Opsin Reduce Kinetic Stability

Compared to rhodopsin, all stable structural segments of opsin, except for segment [H6.2-E3-H7], exhibit higher transition rates, k₀, and therefore decreased lifetimes (Table 2; Figure 6). This finding implies that in the absence of the inverse agonist 11-cis-retinal, the stable structural segments of opsin are kinetically less stable. Similarly, it has been shown that stable structural segments of membrane proteins frequently change lifetime with ligand binding and unbinding (Bippes et al., 2009; Ge et al., 2011; Kedrov et al., 2006, 2010; Zocher et al., 2012a). Such a view is consistent with a recent observation that the affinity and type of the ligand (e.g., agonist, neutral antagonist, or inverse agonist) modulates the kinetic properties of the human β₂AR considerably (Zocher et al., 2012a). Thus, switching from a kinetically stable ligand-bound conformation to a kinetically unstable unliganded conformation may represent a common scheme among GPCRs.

Stable Structural Segments in Opsin Reduce Unfolding Free Energy

Generally, the unfolding free energy ΔG stabilizing a structural segment of opsin was reduced compared to rhodopsin (Table 2; Figure 6). On average, the unfolding free energy of a stable structural segment lowered by 3.6 k_BT. The thermal stability of rhodopsin and opsin has previously been studied by differential scanning calorimetry. The denaturation temperature was found to be 15°C lower for opsin (Albert et al., 1996; Khan et al., 1991; Miljanich et al., 1985; Shnyrov and Berman, 1988). This thermodynamic observation is complementary to our finding that opsin is energetically less stable.

Stable Structural Segments in Opsin Increase Mechanical Rigidity

The mechanical spring constant characterizes the rigidity of a structural segment. Compared to dark-state rhodopsin, all structural segments of opsin display significantly higher spring constants, thereby indicating a higher mechanical rigidity (Table 2; Figure 6).

Constitutively Active G90D Mutant Rhodopsin Shows Similar Trends as Opsin

Structural segments stabilizing opsin showed increased interaction strengths, reduced conformational variability, decreased lifetime, decreased unfolding free energy, and increased mechanical rigidity (Figures 6 and 7). A similar trend was observed previously for G90D mutant rhodopsin, which is constitutively active and causes congenital night blindness (Kawamura et al., 2012). Because we modified and improved the SMFS and DFS data analysis in the current study, we reanalyzed the DFS data recorded previously for G90D rhodopsin to ensure that the trends are independent of the analysis method used (Figure S7; Table S1) (Kawamura et al., 2012). Again, G90D rhodopsin exhibited narrower energy valleys, increased transition rates, lower unfolding free-energy barriers, and higher mechanical rigidities compared to rhodopsin (Tables 2 and S1). Because both opsin and G90D rhodopsin exhibit constitutive activity, the trends observed in DFS parameters may characterize the nature of interactions stabilizing the receptor in the constitutively active state.

Figure 7. Overall Conformational, Kinetic, Energetic, and Mechanical Differences between Dark-State Rhodopsin and Opsin.

Compared to dark-state rhodopsin, the valleys of the energy barriers stabilizing structural segments of opsin are narrower (smaller x_u). In addition, structural segments of opsin generally display a reduced lifetime (reciprocal of k₀), are stabilized by a lower activation free energy (ΔG), and show increased mechanical rigidity (κ). Upon binding 11-cis-retinal, structural segments stabilizing rhodopsin increase structural flexibility, increase kinetic stability, and decrease mechanical rigidity.

Similar Effects of Inverse Agonists on Opsin and β₂AR

The general trends observed in the differences of DFS parameters between free receptor (opsin) and inverse agonist bound receptor (rhodopsin) are also observed for the β₂AR (Zocher et al., 2012a). Half of the structural segments stabilizing the β₂AR exhibit decreased conformational variability and mechanical rigidity when the receptor was unliganded compared to the receptor bound to an inverse agonist. Also, a majority of the stable structural segments of the unliganded β₂AR showed decreased lifetimes and lower energetic stability compared to β₂AR in the presence of an inverse agonist. Thus, the general trends observed in the current study for opsin may be conserved across other GPCR members as well.

Opsin Adopts More Conformational Substates

The increased noise covering the common force peaks (Figure S4) indicates that the opsin molecule shows an increased number of conformational substates. In addition to this fluctuation of force peaks characterizing the reproducibly detected stable structural segments, F-D curves recorded for opsin are peppered with irregularly distributed force peaks that add noise along the entire contour length of the superimposed F-D curves (Figures 2C, 2D, and S2). This noise shows that opsin establishes irregular interactions. These irregular interactions, which are different between those reproducibly detected in opsin and rhodopsin (Figure 3), suggest that individual opsin molecules can adopt additional stable structural segments that occur at much lower frequency compared to the reproducibly detected structural segments. These irregularly detected structural segments indicate alternate structural intermediates.

Conclusions

We applied SMFS and DFS to quantify kinetic, energetic, and mechanical differences between dark-state rhodopsin and opsin. In general, the strength and the variability of the molecular interactions stabilizing structural segments in opsin were considerably larger than in rhodopsin. Although the number and location of structural segments stabilizing rhodopsin and opsin did not differ significantly, their properties differed between chromophore-bound (i.e., ligand-bound) rhodopsin and unliganded opsin. The structural segments stabilizing opsin narrowed their energy valleys, reduced their kinetic stability, decreased their free-energy barriers, and increased their mechanical rigidity (Figure 7). These changes are in agreement with those observed for β₂AR upon inverse agonist binding (Zocher et al., 2012a) and therefore may represent a common feature among GPCRs. Moreover, because the structural segments stabilizing opsin have properties similar to the constitutively active G90D rhodopsin mutant (Kawamura et al., 2012), it may be speculated that these changes facilitate constitutive activity of opsin in a similar manner. The higher conformational variability of dark-state rhodopsin indicates that the structural segments stabilizing the receptor adopt more conformational substates in the inactive than in the chromophore-free state. However, compared to dark-state rhodopsin, opsin establishes ≈30% more irregular interactions. These interactions likely represent alternative conformations of the receptor that are less frequently observed in dark-state rhodopsin.

EXPERIMENTAL PROCEDURES

Mice and ROS Disc Preparation

Opsin in ROS disc membranes was prepared from 4- to 6-weeks-old Rpe65−//− mice (Redmond et al., 1998). ROS purification and ROS disc membrane preparation were conducted as described for wild-type mice (Kawamura et al., 2010). ROS disc preparations were aliquoted and stored at − 80°C.

Retinoid Analysis

Retinoid extraction, derivatization, and separation by high-performance liquid chromatography (HPLC) were performed on samples extracted from eyes of dark-adapted mice as previously described (Van Hooser et al., 2002). Briefly, eyes were homogenized in 1 ml retinoid analysis buffer (50 mM MOPS, 10 mM NH₂OH, and 50% ethanol in H₂O [pH 7.0]). Retinoids were separated twice by centrifugation after adding 4 ml hexane. Extracted retinoids in hexane were dried by speed-vac. Dried retinoids were resuspended in 300 µl hexane and then separated by normal-phase HPLC (Ultrasphere-Si, 4.6 × 250 mm; Beckman Coulter, Brea, CA, USA) using 10% ethyl acetate and 90% hexane at a flow rate of 1.4 ml/min.

UV/Vis Absorbance Spectroscopy

UV/Vis spectroscopy was performed on purified dark-state rhodopsin and opsin as previously described (Kawamura et al., 2012). Dark-state rhodopsin from C57BL/6J (Jackson Laboratories, Bar Harbor, ME, USA) mice and opsin from Rpe65−//− mice were purified as previously described (Kawamura et al., 2010). Reconstitution of opsin with 9-cis-retinal was performed as follows: ROS purified from Rpe65−//− mice were resuspended in Ringer’s buffer (130 mM NaCl, 3.6 mM KCl, 2.4 mM MgCl₂, 1.2 mM CaCl₂, 10 mM HEPES, and 0.02 mM EDTA [pH 7.4]) and incubated with 1 µM 9-cis-retinal (Sigma-Aldrich, St. Louis, MO, USA) for 90 min in the dark at room temperature. The membrane suspension was centrifuged at 16,100 g for 5 min at 4°C, and the supernatant was discarded. The pellet was washed twice in Ringer’s buffer and subsequently solubilized in 10 mM Bis-Tris propane (pH 7.5), 500 mM NaCl, and 20 mM n-dodecylβ-d-maltoside and was purified as previously described (Kawamura et al., 2010).

SMFS

Freshly thawed aliquots of ROS disc membranes (10 µl) were diluted in 200 µl of Ringer’s buffer. 30 µl of ROS disc membranes in Ringer’s buffer were adsorbed onto freshly cleaved mica for 30 min (Müller and Engel, 2007). The buffer was then exchanged several times to remove debris, while the sample was kept hydrated. SMFS was conducted at ≈28°C with an 850 nm laser equipped AFM (NanoWizardII, JPK Instruments, Berlin, Germany). SMFS was performed in a darkroom. Si₃N₄ cantilevers (NPS, Bruker, Billerica, MA, USA) with nominal spring constants of 0.06–0.08 N/m were calibrated using a thermal noise method in Ringer’s buffer (Butt and Jaschke, 1995). For DFS, SMFS was performed at pulling velocities of 300, 700, 1,500, 3,000, 4,500, and 6,000 nm/s. For high-frequency data acquisition at pulling velocities of >1,500 nm/s, a 16-bit data acquisition card (NI PCI-6221; National Instruments, Austin, TX, USA) was used. The extracellular side of ROS discs was identified by AFM imaging and probed by SMFS (Fotiadis et al., 2003; Liang et al., 2003; Sapra et al., 2006). SMFS recorded the deflection of the cantilever (force) and the distance between the tip and sample surface. For analysis, we selected force-distance (F-D) curves that showed force peak patterns extending to the length of fully unfolded and stretched opsin (≥65–75 nm) with an intact Cys110-Cys187 disulfide bond (Sapra et al., 2006).

Statistical Analysis of SMFS Data

Histograms showing how frequently a force peak was detected at a given contour length were generated. The frequency was calculated dividing all counts of force peaks through the total number of F-D curves analyzed. Because each F-D curve had many force peaks the total frequency of all force peaks of a histogram summed up to > >100%. However, the probability of each individual force peak class was ≤100%. These histograms were fitted with a Gaussian mixture model (Figure 3). In this model, the i-th observed contour length l_i, stems from one force peak s = 1, …, M with probability π_s, or from background noise with probability π₀. Force peaks are therefore modeled as a mixture of M different force peak classes, each at a distinct contour length. The contour length for a given force peak class s is described by a Gaussian distribution with mean length μ_s and variance σ _s². As it is not known in which force peak class an observed contour length originates, the probability density f of l_i is a mixture of Gaussians with weights π_s and background noise with weight π₀,

$$f(l_i)=\sum _{s=1}^M{\pi }_s\varphi (l_i,{\mu }_s,{\sigma }_s^2)+{\pi }_{0}g(l_i),$$ (Equation 1)

where ϕ(l_i, μ_s, σ_s²) denotes the probability density of the Gaussian distribution, and the uniform distribution g(l_i) describes the background noise. The model parameters π_s, μ_s, and s_s² were found by the expectation maximization algorithm (Dempster et al., 1977), and the optimal number of force peak classes M was found using the Bayesian information criterion (Schwarz, 1978). We assigned the most probable force peak class s_i to any given contour length l_i with the Bayes classifier by setting

$$S_i={\text{argmax}}_s({\pi }_s\varphi (l_i,{\mu }_s,{\sigma }_s^2),{\pi }_{0}g(l_i)).$$ (Equation 2)

After assigning all observations to force peak classes s, the mean unfolding force F_sv and loading rate r_sv were determined for each pulling speed v. For a detailed description of the statistical analysis, see the Supplemental Experimental Procedures.

Estimating Energy Landscape Parameters and Mechanical Rigidity from DFS Data

For DFS analysis, the Bell-Evans model (Bell, 1978; Evans, 1998; Evans and Ritchie, 1997) was employed. The model states that the most probable force F_sv, at which a protein transitions from the folded to the unfolded state, depends on the logarithm of the loading rate r_sv,

$$F_{sv}=\frac{k_{B}T}{X_u}\text{ln}\left(\frac{X_u{r}_{sv}}{k_{B}T{k}_0}\right),$$ (Equation 3)

where k_B is the Boltzmann constant, T the temperature in Kelvin, k₀ the rate of the folded state transiting to the unfolded state in absence of an externally applied force, and x_u the distance separating the folded from the transition state (Figure S6). The Bell-Evans equation (Equation 3) was fitted by the linear regression F_sv = a ln (r_sv) + b to the DFS plot of every stable structural segment s (Figure 5). To account for uncertainty in both F_sv and r_sv, the reciprocals of the determinant of the covariance matrices of the pulling speed served as weights in the DFS fit. From the fitted slope a and intercept b, one can derive x_u and k₀ (Table 2; Equations S16, S17, S18, and S19). The height difference of the free-energy barrier ΔG was calculated using

$$\mathrm{\Delta }G=-k_{B}T\text{ln}\left({\tau }_{\mathrm{A}}{k}_0\right)),$$ (Equation 4)

where τ_A is the Arrhenius frequency (10⁻⁹ s⁻¹) (Gräter et al., 2005).

In absence of any further information, the transition of a structural segment toward unfolding was assumed to be a simple harmonic motion. In this case, the spring constant κ of the structural segment can be estimated by

$$\kappa =\frac{2\mathrm{\Delta }G}{x_u^2}.$$ (Equation 5)