Rhodopsin forms a dimer with cytoplasmic helix 8 contacts in native membranes

Abstract

G protein-coupled receptors (GPCRs) form dimers and higher-order oligomers in membranes, but the precise mode of receptor-receptor interaction remains unknown. To probe the intradimeric proximity of helix 8 (H8), we carried out chemical crosslinking of endogenous cysteines in rhodopsin in disk membranes. We identified a Cys316–Cys316 crosslink using partial proteolysis and liquid chromatography-mass spectrometry (LC-MS). These results show that a symmetric dimer interface mediated by H1 and H8 contacts is present in native membranes.

GPCRs are versatile signaling machines that constitute the largest group of membrane proteins in the human genome. GPCRs activate and modulate numerous intracellular signaling pathways, including catalysis of nucleotide exchange in the alpha subunit of heterotrimeric G proteins. Like many other membrane receptors, GPCRs are known to form dimers and higher-order oligomers in membranes. Atomic force microscopy (AFM) images of native rod outer segment (ROS) disk membranes showing rows of rhodopsin dimers provide a striking demonstration of their possible supramolecular organization (1). Like most class A (rhodopsin-like) GPCRs, the functional role of rhodopsin dimerization is unknown. In fact, it has been shown that isolated monomers can activate G proteins when segregated in membrane nanoparticles (2).

Recently, the GPCR field has witnessed a surge in structural information, including the first crystal structure of a receptor complexed with heterotrimeric G protein (3). Despite these advances, and ever mounting evidence for oligomerization, the precise interface(s) mediating receptor-receptor contacts remains controversial (4). Spatial constraints from the AFM images were used to predict that the primary rhodopsin dimer interface involved transmembrane (TM) helices H4 and H5 (5). However, two-dimensional (6) and three-dimensional (7) densities obtained from electron microscopy (EM), as well as X-ray data on packing of rhodopsin crystals (8) show that dimer contacts involve TM H1 and cytoplasmic H8. In addition, a recent crosslinking study suggests this interface exists for dopamine D2 receptors heterologously expressed at physiological densities in membranes (9). These results pointed to an oligomeric model in which both the H4/H5 and H1/H8 symmetrical interfaces are simultaneously present.

We set out to demonstrate the possibility of the H1/H8 dimer in native disk membranes by crosslinking the endogenous cysteines of rhodopsin and identifying the site(s) involved. Here we go beyond previous studies of rhodopsin in that we identify additional intermolecular crosslinking sites (10). Working with receptors in the native membrane environment eliminates artifacts associated with dimerization in detergent.

Rhodopsin contains two primary reactive cysteines (positions 140 and 316), and two additional cysteine residues in H8 (positions 322 and 323) may also be reactive due to incomplete palmitoylation. Structural evidence suggests that the Cys316–Cys316 distance in a H1/H8 interface would be 2–3 nm (7, 8). In contrast, these side chains are on opposite faces and thus farther from each other in a H4/H5 dimer arrangement. We hypothesized that a Cys316–Cys316 crosslink using a reagent with a spacer length close to the predicted distance would support the presence of the H1/H8-H1/H8 dimer.

Dark state ROS membranes were first crosslinked with two homobifunctional bis-maleimide reagents containing polyethylene glycol (PEG) spacers of different lengths (Figure 1A–B). The crosslinker-to-rhodopsin stoichiometry was controlled to optimize crosslinking. If too large excess is used, the reactive cysteines saturate with the reagent rather than crosslinking to a neighboring receptor. We found that a 5:1 excess of crosslinker resulted in appreciable formation of dimers and higher-order oligomers that could be observed on a gel. Addition of either BM(PEG)₃ (2.1 nm, the sulfur-to-sulfur distance in the extended conformation of the crosslink) or Bis-MAL-dPEG₃ (3.1 nm) reduced the amount of monomer present relative to a negative control.

FIGURE 1.

Chemical crosslinking of rhodopsin in ROS disc membranes followed by limited proteolysis and SDS-PAGE analysis with silver (A) and Coomassie (B) staining. Samples were analyzed before (lanes 1–3) and after (lanes 4–6) thermolysin treatment (proteolysis) in the absence of crosslinker (lanes 1 and 4), after crosslinking with BM(PEG)₃ (lanes 2 and 5), and after crosslinking with Bis-MAL-dPEG₃ (lanes 3 and 6). The crosslinkers are non-cleavable homobifunctional cysteine-reactive reagents with different spacer lengths. R’ results from cleavage of short C-terminal peptides, F1 is a ~28 kDa N-terminal peptide, and F2 is a ~12 kDa C-terminal peptide. The appearance of the crosslinker-dependent band (F2)₂ in panel B lanes 5 and 6 demonstrates the proximity of H8 in adjacent rhodopsin monomers in ROS, as described in the text. This result is consistent with an H1/H8–H1/H8 dimer model.

To identify the region of the crosslink, we employed a partial proteolysis procedure with thermolysin. Thermolysin cleaves rhodopsin at several C-terminal sites in addition to a primary cut site in the third intracellular loop. The C-terminal cuts result in a slightly shorter peptide, R’. Further proteolysis results in two predominant fragments: a ~28 kDa N-terminal peptide (F1) and a ~12 kDa C-terminal peptide (F2). The F1 peptide contains Cys140, and the F2 peptide contains Cys316, Cys322, and Cys323. Because these fragments are easily resolved on a gel, the region of the cysteine crosslink can be determined by observing which oligomerizes upon addition of the bis-maleimide reagent. As shown in Figure 1A–B lane 4, R’, F1, and several forms of F2 (depending on the extent of C-terminal proteolysis) are the only lower molecular weight bands observed in a control sample. The F1 band stains relatively poorly with Coomassie – likely because it is glycosylated – so it is clearer in the silver stained gel. On the other hand, the modified F2 bands are almost invisible with silver staining, but prominent with Coomassie. Upon addition of either crosslinker (Figure 1B lanes 5–6), the intensities of the F2 bands decrease and new bands appear near 20 kDa. These bands can only correspond to an F2 dimer, so we conclude that the H8 regions crosslink to each other and are thus in close proximity in neighboring receptors in ROS.

Next, we demonstrated the chemical specificity of the Cys crosslinking with two methanethiosulfonate (MTS) reagents, MTS-O4-MTS (2.2 nm) and MTS-O5-MTS (2.6 nm). The MTS groups are extremely reactive, and crosslinking is observed after just 5 min (Supporting Figure S1). These reagents have the advantage that the covalent linkage formed can be cleaved with a reducing agent. Treatment of crosslinked samples with dithiothreitol (DTT) before running a gel collapsed the higher molecular weight bands back down to primarily monomer, as observed in negative controls. This demonstrates that the crosslinker-dependent bands are the result of thiol reactions and not nonspecific oligomerization.

We used LC-MS to demonstrate definitively the presence of a Cys316–Cys316 dimer crosslink. ROS samples were solubilized, alkylated, and precipitated with trichloroacetic acid. The precipitate was then digested with cyanogen bromide (CNBr), which cleaves after methionines at positions 309 and 317 to yield a small peptide that can be detected by LC-MS. The C-terminal methionine is modified to a homoserine lactone as a result of the chemical cleavage, and iodoacetamide treatment adds a carbamidomethyl group to Cys316 if it has not been substituted with crosslinker.

In crosslink samples, the carbamidomethyl-modified 310–317 peptide was identified first (Supporting Table T1). It appears as several isotopes and its mass is accurate to several ppm with respect to theoretical monoisotopic masses. For Bis-MAL-dPEG₃ samples, the crosslinked product is simply two 310–317 peptides plus the mass of the reagent. This product appeared as a z=4 ion (Figure 2A). For MTS peaks, we identified the 310–317 peptide with carbamidomethyl and N-ethylmaleimide (NEM) (resulting from reaction quenching), and then found the peptide substituted with crosslinker. Monosubstituted MTS reagents have their second reactive group hydrolyzed to a thiol, which can then be modified with NEM, and this peak was also present. Finally, the crosslinked product was found, again with roughly ppm accuracy (Figure 2B). Crucially, in samples treated with DTT prior to precipitation and digestion, the peaks corresponding to peptide substituted with MTS crosslinker were not observed. This additional control confirms the peak assignments. The mass spectrometry data conclusively demonstrate the existence of Cys316–Cys316 crosslinks. It is notable that this crosslink was only observed with the longer reagents (≥2.6 nm) but not with the shorter reagents (≤2.2 nm), suggesting the distance between side chains is less than 2.6 nm. However, the thermal mobility and other factors limit the accuracy of such estimates.

FIGURE 2.

LC-MS analysis of Cys316–Cys316 crosslinked peptides. (A) Two 310–317 peptides crosslinked with Bis-MAL-dPEG₃. The monoisotopic peaks are z=4 ions. Liquid chromatograms integrated across the two MS peaks show that the species elute at similar times. (B) Two 310–317 peptides crosslinked with MTS-O5-MTS. The monoisotopic peaks are z=2 ions, and disappear in samples treated with DTT prior to precipitation. Liquid chromatograms integrated across the MS peaks show that they elute simultaneously.

The partial proteolysis and LC-MS data are complementary, highlighting the importance of integrating diverse experimental data to identify crosslinking sites. The gels in Figure 1 and Supporting Figure S1 show extensive oligomerization, so more than one dimerization interface must be relevant. The presence of the (F2)₂ band demonstrates that one of them involves the proximity of H8 regions on adjacent receptors. The LC-MS experiments enable precise identification of the Cys316–Cys316 crosslinking site. The results are especially convincing because the assigned MTS peaks disappear upon addition of DTT, which cleaves the crosslink disulfide bond. Further, because the MTS reagent hydrolyzes so quickly, it is likely that the crosslinks result from prearranged dimers and not from random collisions over extended periods of time. Interestingly, the partial proteolysis data show H8 crosslinks for all four reagents employed, but the Cys316–Cys316 peaks were only identified with the two longer (Bis-MAL-dPEG₃ and MTS-O5-MTS) crosslinkers. It is possible that either Cys322 or Cys323 crosslinks are responsible for the gel shift with the shorter reagents. As with Cys140-containing fragments, we were not able to identify these sites with our CNBr cleavage strategy.

We also probed dimerization interfaces in coarse-grained molecular dynamics (CGMD) studies of spontaneous rhodopsin assembly (Periole, et al., in preparation). The second most frequently populated dimer cluster involved an H1/H8 interface similar to structures observed in X-ray crystallography and EM (Figure 3). The Cys316–Cys316 side chain bead distance of this cluster (Figure 3C) was 2.3 nm (2.6 nm when fitting residues 310–322 of PDB 1U19; see Supporting Information for discussion of side chain distances). Notably, a similar but less populated cluster (Figure 3D) showed a Cys316–Cys316 distance closer to 1.9 nm (2.1 nm from local fit of 1U19). It should be noted that the analysis of the CGMD self-assembly simulations on a microsecond time scale does not predict thermodynamic stability, and consequently, cluster populations do not reflect equilibrium distributions. The two structures are spatiotemporally mutually exclusive, but it is difficult to determine which of these arrangements is responsible for the crosslinks presented here. If the monomer–dimer exchange is slow on the time scale of the crosslinking experiment, the more stable interface may predominate and thereby exclude the other possible orientation. On the other hand, if the dimers are more transient then the structure that brings Cys316 residues closer and thus increases the likelihood of crosslinking could account for the data presented.

FIGURE 3.

Structural and CGMD analysis of H1/H1 dimer orientations. (A) Two symmetry-related promoters from the crystal structure of photoactivated rhodopsin (PDB ID: 2I35). (B) Crystal structure of rhodopsin (PDB ID: 1GZM) fit to electron density metarhodopsin I dimer observed in EM image (7). (C–D) Crystal structure of rhodopsin (PDB ID: 1U19) fit to two distinct H1/H8–H1/H8 dimer clusters observed in CGMD analysis.

In summary, we report biochemical crosslinking experiments that conclusively identify a Cys316–Cys316 crosslinked dimer and demonstrate the proximity of these residues between rhodopsins in native disc membranes. These data suggest a rhodopsin dimer interface mediated by H1/H8 contacts exists, as suggested by previous crosslinking, EM, and crystallography data. Because crosslinking caused the formation of oligomers, a second interface, perhaps involving H4/H5 contacts, must also be present. Despite studies that have demonstrated a single receptor is sufficient for full G protein activation, the extremely high density of rhodopsin in disc membranes hint that a dimer may be the primary structural unit that interacts with the heterotrimer. The high degree of homology in class A GPCRs suggests that the results reported here might be relevant for other receptors that are known to oligomerize.

ABBREVIATIONS

CGMD

coarse-grained molecular dynamics

CNBr

cyanogens bromide

DTT

dithiothreitol

EM

electron microscopy

GPCR

G protein-coupled receptor

H

helix

LC-MS

liquid chromatography-mass spectrometry

MTS

methanethiosulfonate

NEM

N-ethylmaleimide

PEG

polyethyelene glycol

ROS

rod outer segment

TM

transmembrane