Relocating the active-site lysine in rhodopsin and implications for evolution of retinylidene proteins

Abstract

Type I and type II rhodopsins share several structural features including a G protein-coupled receptor fold and a highly conserved active-site Lys residue in the seventh transmembrane segment of the protein. However, the two families lack significant sequence similarity that would indicate common ancestry. Consequently, the rhodopsin fold and conserved Lys are widely thought to have arisen from functional constraints during convergent evolution. To test for the existence of such a constraint, we asked whether it were possible to relocate the highly conserved Lys296 in the visual pigment bovine rhodopsin. We show here that the Lys can be moved to three other locations in the protein while maintaining the ability to form a pigment with 11-cis-retinal and activate the G protein transducin in a light-dependent manner. These results contradict the convergent hypothesis and support the homology of type I and type II rhodopsins by divergent evolution from a common ancestral protein.

The retinylidene proteins are integral membrane proteins that covalently bind a retinal chromophore. Amino acid sequence comparison divides these proteins into two families known as type I and type II rhodopsins (1). Type I rhodopsins, such as bacteriorhodopsin from the archaeon Halobacterium salinarum, function as light-driven ion transporters, channels, and phototaxis receptors. Type II rhodopsins, best known for the visual pigment of mammalian rod photoreceptor cells, function primarily as photosensitive receptor proteins in metazoan eyes and in certain extraocular tissues. Henceforth, we will use the term “rhodopsin” to refer to the visual pigment of bovine rod photoreceptor cells and “bacteriorhodopsin” to refer to the light-driven proton pump of H. salinarum.

Rhodopsin is a prototypical member of the large family of G protein-coupled receptors (GPCRs; specifically class A GPCRs) (2). It is composed of an apoprotein (called “opsin”) and an 11-cis-retinal chromophore, resulting in a pigment with λ_max = 500 nm. The GPCR fold comprises seven transmembrane α-helices oriented in a particular spatial arrangement with a specific connectivity (SCOP classification scop.b.g.c.A; ref. 3). In rhodopsin, the N terminus resides in the intradiscal (i.e., extracellular) space and the C terminus in the cytoplasm. The 11-cis-retinal chromophore is covalently attached to the protein by means of a protonated Schiff base to the ε-amino group of Lys296 in the seventh helix. The GPCR fold and active-site Lys are absolutely conserved among all visual pigments of higher eukaryotes.

Upon absorption of light, the 11-cis-retinal isomerizes to the all-trans form. The protein responds with a conformational change leading to an enzymatic cascade that begins with activation of the G protein transducin and ends with closure of cation channels in the plasma membrane and hyperpolarization of the rod cell (4). A key intermediate in the photoactivation of rhodopsin is the species metarhodopsin II (MII) (5). MII is the only intermediate capable of activating transducin and is characterized by an absorption maximum in the near UV (λ_max = 380 nm), resulting from deprotonation of the Schiff base.

Like rhodopsin, bacteriorhodopsin adopts the GPCR fold (1, 3, 6). Bacteriorhodopsin is oriented with the N terminus in the extracellular space and the C terminus in the cytoplasm. The retinal chromophore is attached to the protein covalently by means of a protonated Schiff base to the ε-amino group of a Lys residue, Lys216, in the seventh transmembrane α-helix. Upon absorption of light, a key intermediate, M, in the proton pumping cycle forms in which the Schiff base nitrogen is no longer protonated. With the exception of a few fungal proteins of unknown function, the GPCR fold and the active-site Lys are also conserved among all type I rhodopsin homologs.

Despite the striking structural and functional similarities of the type I and type II rhodopsins, there is no significant sequence identity between these two families that would suggest a common ancestral origin (1). It is widely believed that the common fold and active-site Lys are products of convergent evolution resulting from functional constraints on the proteins (1, 7–12). To test this hypothesis, we have focused on the visual pigment rhodopsin and asked whether it is possible to move the active-site Lys296 to a different location in the protein. We attempted to move the Lys to five different locations: two positions in transmembrane helix (TM) 2, one in TM3, one to a different location in TM7, and one in the β-hairpin loop connecting TM4 and TM5 that forms part of the retinal binding pocket. Surprisingly, four of the five mutants combine with 11-cis-retinal to form pigments with near wild-type spectral properties, and three of these four activate transducin in a light-dependent manner with specific activities approaching that of wild-type rhodopsin. These results demonstrate that an absolutely conserved, common structural feature—the Schiff base Lys in helix seven—is not required for rhodopsin’s photosensitive function, contradicting a key prediction of convergent evolution resulting from functional constraint.

Results

The K296G and K296A mutant rhodopsins, in which the active-site Lys has been removed, cannot covalently bind 11-cis-retinal (13). However, both mutants can bind a Schiff base complex of 11-cis-retinal and an external alkyl amine, thereby producing a pigment with near wild-type properties (absorption maximum approximately 490 nm and restored ability to light-activate transducin) (13). The apoprotein forms of K296G and K296A activate transducin in the dark without retinal, an unusual ability known as constitutive activity (14). This behavior results from disruption of an internal salt bridge between the Lys296 ε-amino group and the Glu113 carboxylate, the so-called Schiff base “counterion” (15–17).

Because K296G and K296A cannot bind 11-cis-retinal covalently, they have lost the ability to activate transducin in a light-dependent manner. We wondered, therefore, whether we could introduce a Lys residue at a different site to rescue 11-cis-retinal binding and restore light-dependent activation of transducin. We selected five different sites for the newly introduced Lys residues (Fig. 1 and Fig. S1): Gly90, Thr94, Ala117, Ser186, and Phe293. All five positions are in the active site, located within 9 Å (range of 5–9 Å) of the retinal C15. The five sites are distributed among several different secondary structure elements, including TM2 (Gly90 and Thr94), TM3 (Ala117), the two-stranded β-sheet connecting TM4 and 5 (Ser186), and finally TM7 (Phe293), the same transmembrane segment as the original Lys296 of the WT protein. In each case, the target amino acid was changed to a Lys residue within the context of the N2C/D282C thermally stable mutant rhodopsin that also contains a mutation of the active-site Lys296 (to either a Gly or an Ala). In total, we made 10 mutant proteins. We evaluated pigment function with three criteria: (i) ability to bind the 11-cis-retinal chromophore covalently via a Schiff base to the introduced Lys; (ii) ability to form a pigment with a long-wavelength absorption maximum; and (iii) ability to activate the G protein transducin in a light-dependent manner.

Fig. 1.

The retinal-binding pocket of bovine rhodopsin (PDB ID code 1U19). 11-cis-retinal, orange; Glu113 counterion, red; Lys296, blue; positions at which a Lys was introduced (Gly90, Thr94, Ala117, Ser186, and Phe293), green.

With the exception of K296G/A117K and K296A/A117K, all of the mutant proteins bind retinal and form a pigment with a long-wavelength absorption maximum (Fig. 2 and Table 1). The mutants K296G/F293K, K296A/F293K, and K296A/G90K have a low yield of long-wavelength pigment and a significant peak at 380 nm, indicating a fraction of unprotonated Schiff base or free retinal. However, the remaining mutants were isolated in good yield and exhibit near wild-type absorption maxima (λ_max = 476–484 nm). We chose one mutant for each new Lys position to test for ability to activate transducin: K296G/G90K, K296A/T94K, K296A/S186K, and K296G/F293K.

Fig. 2.

Ability of rhodopsin mutants to form pigments with 11-cis-retinal. Normalized UV-visible absorption spectra for dark-adapted pigments in 0.02% (wt/vol) DDM at pH 7.5 at room temperature. Insets are an expanded view of the long-wavelength λ_max peak. Scale bars in upper left corner of each panel (upper right of Insets) represent 0.03 absolute absorbance.

Table 1.

UV-visible absorption maxima of rhodopsin mutants

Construct	Dark λmax, nm	Light λmax, nm	Acid trapped after light λmax, nm
WT	500	384	428
K296G	—	—	—
K296G/G90K	483	381	438
K296G/T94K	475	384	396
K296G/A117K	—	—	—
K296G/S186K	469	384	402
K296G/F293K	485	368	393
K296A	—	—	—
K296A/G90K	478	381	436
K296A/T94K	483	382	402
K296A/A117K	—	—	—
K296A/S186K	478	385	401
K296A/F293K	493	370	423

Three mutants activate transducin in a light-dependent manner (Fig. 3). The K296A/S186K and K296G/G90K mutants approach the activity observed with the wild-type protein, whereas the K296A/T94K mutant exhibits significantly less activity, in extent of reaction and initial rate. The initial rates for K296A/S186K and K296G/G90K are similar. However, the extent of the reaction is less for K296A/S186K, probably reflecting the greater stability of the MII intermediate in K296G/G90K as indicated by the yield of 440-nm pigment following acid trap of light-exposed samples (Fig. S2). In contrast, the K296G/F293K mutant displays decidedly nonwild-type behavior (Fig. 3). This mutant rhodopsin activates transducin, but in a light-independent manner. The K296G/F293K mutant is active in the dark, and exposure to light does not change the rate of transducin activation (although the spectrum is clearly converted to a peak with a λ_max of 380 nm; Fig. S2).

Fig. 3.

[³⁵S]-GTPγS binding to transducin following light activation of select mutant pigments. Each reaction contained 5 nM rhodopsin, 1 μM transducin, 3 μM GTPγS in 10 mM Tris buffer at pH 7.5, 100 mM NaCl, 5 mM MgCl₂, 0.1 mM EDTA, and 0.01% (wt/vol) DDM. Blue circles, WT; green squares, K296G/G90K; orange triangles, K296A/S186K; maroon diamonds, K296G/F293K; black inverted triangles, K296A/T94K; shading, reactions run in the dark. Error bars represent the SD (n = 4).

The K296G and K296A mutants are both constitutively active, whereas wild-type rhodopsin is not. Hence, it was also of interest to determine how the introduced Lys residues affect the constitutive activity of the K296G and K296A parents. Only the S186K mutant displays residual constitutive activity, which is significantly reduced relative to the parents (Fig. 4). Thus, all of the Lys mutations (even at position 293) successfully rescue wild-type behavior of the apoprotein forms.

Fig. 4.

Constitutive activation of transducin by select mutant opsins, monitored by [³⁵S]-GTPγS binding to activated transducin. Each reaction contained 5 nM opsin, 1 μM transducin, 3 μM GTPγS in 10 mM Tris buffer at pH 7.5, 100 mM NaCl, 5 mM MgCl₂, 0.1 mM EDTA, and 0.01% (wt/vol) DDM. Purple open triangles, K296A; light blue open squares, K296G; orange triangles, K296A/S186K; light blue +, K296A/A117K; green squares, K296G/G90K; black inverted triangles, K296A/T94K; maroon diamonds, K296G/F293K; light purple X, K296G/A117K; blue circles, WT. Error bars represent the SD (n = 4).

Discussion

Recently, the rational design of retinal binding proteins has emphasized the importance of active site geometry for forming a protonated Schiff base (18, 19). Successful reengineering of the Cellular Retinoic Acid Binding Protein II required proper orientation of the Lys ε-amino group relative to the carbonyl center of 11-cis-retinal for nucleophilic attack. However, we describe four different Lys positions that are capable of binding retinal (G90K, T94K, S186K, and F293K). Each of these alternative positions forces the Lys to react with retinal from drastically different angles, indicating that ideal Bürgi–Dunitz and Flippin–Lodge angles are not required for Schiff base formation in rhodopsin. This result is perhaps not entirely surprising, because the Bürgi–Dunitz and Flippin–Lodge angles were originally described for two molecules that collide in solution, not held in a specific orientation by a protein scaffold.

The Lys mutations may also provide insight into the mechanism of rhodopsin activation. In current models, movement of TM5 and TM6 is related to a corresponding displacement of the β-sheet connecting TM4 and 5 (EL2) (5, 20). In the functional K296A/S186K mutant, the Schiff base is located on EL2. If activation involves the movement of EL2 away from retinal in the binding pocket, there must be considerable plasticity in the active site, such that minor structural rearrangements are easily accommodated by the mechanism leading to the global conformational change for active rhodopsin.

Surprisingly, several second-site Lys mutations suppress the constitutive activity of the K296G and K296A parents (Fig. 4). With the exception of K296A/S186K, the constitutive activity of all of the Lys mutants is reduced to within experimental error of the N2C/D282C host. Second-site mutations that reverse the constitutive activity of a functional rhodopsin are rare. These results are consistent with a model in which a key determinant of constitutive activity is neutralization of charge on the Schiff base counterion residue, Glu113.

The K296G/F293K mutant forms a pigment with 11-cis-retinal and activates transducin, but the activity does not depend on light. The apoprotein form, in contrast, does not activate transducin. Thus, the K296G/F293K mutation suppresses the constitutive activity of the K296G parent and allows the protein to form a protonated Schiff base with retinal, but once retinal is bound, the protein is locked in an active conformation. This behavior is significantly different from that of the previously reported dark-active rhodopsin mutant E113Q/M257Y (21). Individually, the E113Q and M257Y mutations constitutively activate rhodopsin. The double mutant is also constitutively active but, in addition, displays dark activity not seen with either single mutation alone. The dark activity of this mutant likely results from conformational uncoupling of the transducin activation domain and retinal-binding site. In contrast, the K296G/F293K apoprotein is inactive but becomes trapped in an active conformation after binding retinal. In this way, K296G/F293K is reminiscent of the “steric doorstop” mutant T118W described by Kono and coworkers (22). Rhodopsin activates transiently in the process of binding retinal, as if the protein adopts an active conformation while opening to allow the ligand to enter the binding pocket (23–25). In the T118W mutant, changing Thr118 to a bulkier side chain results in a steric clash with the 9-methyl group of the retinal, preventing closure of the binding pocket and trapping the protein in an active conformation. Perhaps the dark activity of the K296G/F293K mutant similarly results from a steric clash due to a slight repositioning of the chromophore.

The K296G/A117K and K296A/A117K mutants deserve comment. Previous mutagenesis studies indicated that an acidic residue (either Asp or Glu) at the 117 position can substitute for the Schiff base counterion (26, 27). Thus, even before the crystal structure of rhodopsin confirmed that Ala117 is one helical turn from Glu113, indirect methods had shown that Ala117 is close to the Schiff base nitrogen. In addition, the E113A/A117E mutant forms a protonated MII intermediate upon exposure to light (28), demonstrating that the 117 side chain is close to the Schiff base nitrogen in both the excited state and the ground (or dark) state of the protein. In the present study, we attempted to move the Schiff base Lys from position 296 in TM7 to position 117 in TM3, but the mutant is incapable of forming a pigment with added 11-cis-retinal (Fig. 2). The inability to form a pigment could simply indicate that the mutant protein does not fold properly. However, in the A117K mutant, the Glu113 counterion may be so close to the Lys ε-amino group that the nitrogen remains protonated, unable to provide a lone pair of electrons for nucleophilic attack on the carbonyl carbon of the retinal. This interpretation is supported by the high yield of protein from the immunoaffinity column, because denatured rhodopsin is usually retained on the column under our purification conditions (29, 30). Close proximity of the Schiff base and Glu113 is also consistent with the loss of constitutive activity in this mutant (Fig. 4) (14).

Is the highly conserved active site Lys residue in TM7 of type I and type II rhodopsins a consequence of convergent evolution due to a shared functional constraint on two unrelated protein families? If so, it should not be feasible to move the active-site Lys to another location in the protein and retain the ability to form a functional pigment with retinal. In fact, we were able to construct multiple functional type II pigments in which the active site Lys is moved to three different positions in different secondary structure elements. Therefore, photosensitive rhodopsin function does not require a Lys-retinal Schiff base linkage in TM7. These results support the homology of type I and type II rhodopsins by demonstrating that the shared Schiff base position is not a product of convergence due to functional constraint.

Evolution has optimized modern type II rhodopsins over millions of years, regardless of whether type I and type II rhodopsins are convergent or divergent. Unlike with the natural rhodopsins, we have made little effort to optimize the function of our mutant constructs, other than the N2C/D282C background mutations that confer some stability to the protein. It is plausible that we could find other compensatory mutations that would improve the function of our mutants. For example, other mutations could suppress the low constitutive activity of K296A/S186K or could red-shift the absorbance maximum of the mutants by 10 nm. It is likewise possible that, say, the K296A/S186K mutant would have no constitutive activity in a different protein background from another species; we have tried alternate Lys positions in only one protein from one species. In any case, convergent evolution can access all these potential variants (different protein backgrounds and compensatory mutations). The fact that our crude mutagenesis found functional alternative Lys locations so easily—and that evidently nature has not—further underscores the implausibility of evolutionary convergence to the same Lys location in helix seven.

In summary, we have tested the hypothesis that the function of rhodopsin (in terms of binding retinal, formation of a long-wavelength pigment, and activation of transducin) constrains the active-site Lys to a location in TM7. The mutants clearly demonstrate that rhodopsin can retain function when the Lys is moved to a different location. These results support a divergent evolutionary scenario in which a common ancestor gave rise to the modern retinylidene proteins. One might wonder why the active-site Lys has not migrated to other locations during divergent evolution. Relocating the Lys would likely require an intermediate either with no active-site Lys or with Lys residues at both locations. Perhaps the Lys has been retained in TM7 because neither of these intermediates is capable of forming a viable pigment.

Materials and Methods

Materials.

Synthesis and purification of 11-cis-retinal was as described (30). Dodecyl β-d-maltoside (DDM) was from Calbiochem. DMEM was purchased from GIBCO. Bovine growth serum and Dulbecco’s PBS were from HyClone Laboratories. GTP was from Amersham Biosciences, GTPγS was from Sigma-Aldrich, and [³⁵S]-GTPγS (1,250 Ci/mmol) was from Perkin-Elmer. Oligos for mutagenesis were purchased from Integrated DNA Technologies.

The antirhodopsin monoclonal antibody 1D4 was from the National Cell Culture Center (Minneapolis, MN). The 1D4-Sepharose 4B immunoaffinity matrix used to purify rhodopsin was prepared as described (30). The 1D4-peptide, corresponding to the carboxyl-terminal 8 amino acids of rhodopsin, was used to elute the protein from the immunoaffinity matrix.

Transducin was purified from frozen bovine retinae (Schenk Packing Co.), as reported (30, 31). Transducin concentration was determined spectrophotometrically by using an absorption coefficient of 93,570 M⁻¹⋅cm⁻¹ at 280 nm and by active-site titration using [³⁵S]-GTPγS of known specific radioactivity (31). Transducin was monitored for contamination by rhodopsin using either a transducin activation assay or Western blot analysis with the 1D4 antibody.

Mutagenesis, Expression, and Purification of the Proteins.

All mutations were made in the context of the thermally stable N2C/D282C mutant of bovine rhodopsin, which was considered to be WT in this study. The N2C/D282C mutant contains an engineered disulfide bond between two introduced cysteine residues, N2C and D282C (30). The engineered disulfide confers enhanced thermal stability to the opsin form of the protein (30, 32). Crystal structures of the N2C/D282C mutant, in both the dark state (33) and active state (34, 35), show the protein to be identical to that of native rhodopsin (36–39), except for the missing oligosaccharyl chain at position 2 and the presence of electron density corresponding to a disulfide bond connecting the two side-chain sulfur atoms at positions 2 and 282. In addition, functional studies performed to date show that this mutant behaves as does WT in all ways except with respect to stability of the opsin form in detergent solution (30, 32).

Mutations were introduced into the cDNA for rhodopsin by QuikChange mutagenesis (Stratagene). All rhodopsin mutants were expressed transiently in HEK293S-GnT1⁻ cells by using calcium phosphate precipitation for transfection (40). Proteins were purified and then reconstituted with 11-cis-retinal essentially as described (30) except that mutants were eluted from the 1D4-Sepharose matrix following a 1-h incubation with 0.02% (wt/vol) DDM in 5 mM Hepes buffer at pH 7.5, 0.1 mM MgCl₂, and 3 mM NaN₃ containing 80 µM 1D4-peptide at room temperature. Molar absorption coefficients were calculated relative to WT (ε₅₀₀ = 40,600 M⁻¹⋅cm⁻¹; K296G/G90K, ε₄₈₃ = 42,200 M⁻¹⋅cm⁻¹; K296A/T94K, ε₄₈₃ = 36,300 M⁻¹⋅cm⁻¹; K296A/S186K, ε₄₇₈ = 36,100 M⁻¹⋅cm⁻¹; K296G/F293K, ε₄₈₅ = 24,200 M⁻¹⋅cm⁻¹) by acid-trapping the chromophore (ε₄₄₀ = 31,000 M⁻¹⋅cm⁻¹) in the dark with 0.5% (wt/vol) SDS in 50 mM sodium phosphate buffer at pH 3.5 (final concentrations) as has been described (41).

Absorption Spectroscopy.

UV-visible absorption spectra were recorded with a Hitachi model U-3210 that was specifically modified by the manufacturer for use in a darkroom. Data were collected with a microcomputer by using GraphPad Prism from GraphPad Software. All spectra were recorded with samples at 25 °C and a path length of 1.0 cm. Pigments were bleached by exposure to light from a 300-W tungsten bulb filtered through a 435-nm cut-on filter for 30 s.

Transducin Activation Assays.

A filter-binding assay, described (30, 31, 41), was used to monitor the ability of mutant pigments or opsins (ε₂₈₀ = 65,000 M⁻¹⋅cm⁻¹ used for all mutants) to catalyze the exchange of GDP for [³⁵S]-GTPγS in transducin.