Evidence for long range allosteric interactions between the extracellular and cytoplasmic parts of bacteriorhodopsin from the mutant R82A and its second site revertant R82A/G231C.

Evidence is presented for long range interactions between the extracellular and cytoplasmic parts of the heptahelical membrane protein bacteriorhodopsin in the mutant R82A and its second site revertant R82A/G231C. (i) In the double mutants R82A/G72C and R82A/A160C, with the cysteine mutation on the extracellular or cytoplasmic surface, respectively, the photocycle is the same as in the single mutant R82A with an accelerated deprotonation of the Schiff base and a reversed order of proton release and uptake. Proton release and uptake kinetics were measured directly at either surface by using the unique cysteine residue as attachment site for the pH indicator fluorescein. Whereas in wild type proton uptake on the cytoplasmic surface occurs during the M-decay (tau approximately 8 ms), in R82A it occurs already during the first phase of the M-rise (tau < 1 microseconds). (ii) The introduction of a second mutation at the cytoplasmic surface in position 231 (helix G) restores wild type ground state absorption properties, kinetics of photocycle and of proton release, and uptake in the mutant R82A/G231C. In addition, kinetic H/D isotope effects provide evidence that the proton release mechanism in R82A/G231C and in wild type is similar. These results suggest the existence of long range interactions between the cytoplasmic and extracellular surface domains of bacteriorhodopsin mediated by salt bridges and hydrogen-bonded networks between helices C (Arg-82) and G (Asp-212 and Gly-231). Such long range interactions are expected to be of functional significance for activation and signal transduction in heptahelical G-protein-coupled receptors.

G-protein at the opposite cytoplasmic surface of the protein.
The membrane protein bacteriorhodopsin (bR) 1 shares the heptahelical bundle motif with G-protein-coupled receptors. In this report, we present evidence for long range interactions between the extracellular and cytoplasmic parts of bR based on evidence from the mutant R82A and its second site revertant R82A/G231C.
Bacteriorhodopsin acts as a light-driven proton pump in the plasma membrane of the archaebacterium Halobacterium salinarium. A retinylidene chromophore is bound via a protonated Schiff base linkage to Lys-216. Upon flash excitation bR undergoes a cyclic photoreaction with distinct spectroscopic intermediates and proton translocation from one side of the membrane to the other (for reviews see Refs. [3][4][5]. In the first half of this photocycle, the Schiff base becomes deprotonated during the transition to the M-intermediate, and in the same time range a proton is released at the extracellular surface (6). During the second half of the photocycle, the Schiff base is reprotonated from Asp-96 (7), which is located in the cytoplasmic part of the protein, and a proton is taken up from the cytoplasm. In the proton release pathway, Asp-85 has been identified as the primary acceptor of the proton from the Schiff base by sitedirected mutagenesis and Fourier transform infrared difference spectroscopy (8). Arg-82 is located one helix turn away from Asp-85, facing toward the proton channel (9,10). Steadystate and time-resolved UV/Vis absorption (11)(12)(13) and Fourier transform infrared (14) spectroscopy results indicate that Asp-85, Glu-204, , and Arg-82 interact via a direct or indirect coupling of their pK values. Proton release to the extracellular medium may be facilitated by a hydrogen-bonded network between these residues (10,13,15,16). In the three-dimensional structure of the unphotolyzed bR, a defined density for the arginine 82 side chain, has recently been resolved by x-ray crystallography at a resolution of 2.5 (17) and 2.3 Å (10). The proposed movement of this side chain toward the extracellular surface upon formation of the M-intermediate, based on computer calculations (18), is not yet established. However, recent time-resolved electrical measurements support this proposal (19).
To further understand the mechanism of proton transport, we have studied the mutant R82A. In R82A an accelerated deprotonation of the Schiff base and a reversed order of proton release and uptake, detected with pH indicator dyes in the aqueous solution, were already observed (20,21). Moreover, in R82A the pK a of Asp-85, which controls the purple to blue transition, is increased by about 5 pH units compared with wild type (wt) (22). These dramatic changes are because of the replacement of the positively charged arginine by the neutral alanine.
We have used time-resolved absorbance spectroscopy in combination with pH-sensitive dyes (pyranine in the aqueous bulk phase and fluorescein derivatives attached at specific sites on the protein surface) to detect the kinetics of the photocycle and of proton release and uptake. To detect proton concentration changes separately on each surface of the protein, the attachment site (single cysteine residue) for the pH indicator dye at the surface of the mutant protein R82A was varied. We have used the single cysteine residue in position 72 at the extracellular surface and in the positions 160 and 231 at the cytoplasmic surface as attachment sites. Fig. 1 shows a tertiary structural model of bR based on x-ray crystallographic data (23). The positions of the residues Gly-72, Arg-82, Ala-160, and Gly-231 are indicated. Unexpectedly, the cysteine mutation in position 231 at the end of helix G on the cytoplasmic surface leads to an almost complete restoration of the photocycle, pK a of Asp-85, and proton pumping in the double mutant R82A/G231C. The cysteine double mutant R82A/A160C, also with a cysteine mutation on the cytoplasmic surface (EF loop), has on the other hand still all the altered properties of the single mutant R82A. The reversion effect observed only at position Gly-231 suggests an interaction between helices C (Arg-82) and G (Gly-231). The second site revertant produced by the replacement of glycine by cysteine at the cytoplasmic surface of the protein far away from the location of the first mutation (R82A) can shed light on the molecular mechanisms that promote long range interactions in 7-helix transmembrane proteins. Such allosteric interactions are expected to be important for ligand binding and signal transduction in G-protein-coupled receptors (24). EXPERIMENTAL PROCEDURES 1,2-Dimyristoyl-sn-glycero-3-phosphatidylcholine (DMPC) and Tris-(hydroxy-methyl)-aminomethane (Tris) were obtained from Sigma. Ethylenediaminetetraacetic acid and 1,4-dithio-DL-threitol were from Fluka. 8-Hydroxy-1,3,6-pyrene-trisulfonic acid trisodium salt (pyranine) was from Serva. Sephadex G-25 (fine) was from Amersham Pharmacia Biotech, 5-(iodoacetamido)fluorescein and 5-(bromomethyl)fluo-rescein (BMF) were from Molecular Probes.
The preparation and expression of the bR mutants A160C and G231C in H. salinarium, in which cysteine replaces Ala-160 and Gly-231, respectively, and the preparation of R82A, in which alanine is substituted for arginine in position 82, has been reported (25,26). The bR double mutants R82A/A160C and R82A/G231C were prepared by following the described procedures (26) except that the synthetic restriction fragments BspHI-Psp14061 and SphI-NotI were used, respectively, to construct the mutant genes.
Solubilization of bR membrane fragments and regeneration of bR in DMPC/CHAPS micelles were performed as described (27). The regeneration procedure was modified according to Ref. 28. After regeneration the retinal band is shifted back completely to 550 nm, resulting in 96 -100% regeneration, using an estimated extinction coefficent of ⑀ 550 Ϸ 56,000 M Ϫ1 cm Ϫ1 . Labeling with 5-(iodoacetamido)-and 5-(bromomethyl)-fluorescein and the determination of the labeling stoichiometry were performed as described (25,28).
Flash spectroscopy and data analysis with a sum of exponentials were performed as described elsewhere (7). The excitation was with 10-ns pulses of 3-6 mJ of energy at 590 or 500 nm. Under these conditions about 15% of bR are cycling. Typically 30 -50 time traces were averaged for the kinetics of the M-intermediate and 70 -100 for the dye kinetics.
Proton release and uptake were detected in the aqueous bulk medium of the purple membrane suspension, containing 4 -15 M bR in 150 mM KCl, by calculating the difference of the measured flash-induced absorbance changes at 450 nm between samples with and without 45 M pyranine at pH 7.3 and 22°C (25,29). The light-induced proton concentration changes, detected in samples with fluorescein attached to cysteine residues in bR, were determined by calculating the measured flash-induced absorbance difference at 495 nm between samples with and without 10 mM Tris or MOPS buffer, pH 7.3, in 150 mM KCl at 22°C (25,29).
Titration experiments and analysis were performed as described in (30). To obtain the apparent pK a , the measured absorbance changes were fitted with the Henderson-Hasselbalch equation.
⌬A max is the maximal absorbance difference, n is the number of protons involved in the transitions and pK a is the midpoint of the titration. The data plotted in Fig. 4 were fitted with the following equation described by Balashov et al. (21), with ␣ ϭ 1 ϩ 10 (pH Ϫ pK3) , ␤ ϭ 1 ϩ 10 (pH Ϫ pK2) , and ␥ ϭ 10 (pH Ϫ pK1) .

Characterization of the Unphotolyzed Single and Double
Mutants (Purple to Blue Transition)-The position of the maximum of the visible absorbance spectrum of bacteriorhodopsin reflects the charge distribution in the vicinity of the retinylidene chromophore. Fig. 2 shows the absorption spectra of wt bR in the unphotolyzed state and that of the mutants R82A, G231C, and R82A/G231C in 150 mM KCl at pH 6. The chromophore absorbance maxima ( max ) of wt, G231C, and R82A/ G231C are nearly identical (568, 568, and 567 nm, respectively, in the light-adapted form). In the single mutant R82A max is red shifted about 40 nm compared with wt to 602 nm. In the mutant R82A/G231C, the additional mutation G231C at the cytoplasmic surface (C terminus) restores the wt absorbance spectrum and, therefore, presumably the charge distribution in the vicinity of the chromophore.
The absorption spectrum of unphotolyzed bR is characterized by a purple form with a max of 568 nm around neutral pH, as shown above, which changes to a blue form at acidic pH with a max of about 602 nm. In wt the pK a of this so called purple to blue transition is about 2.6 (in 150 mM KCl). It has been shown that this transition reflects the protonation state of the primary proton acceptor Asp-85 (22,31). The purple to blue transition is directly affected by the proton concentration at the protein surface, which depends on the surface charge (30,32). Therefore, the apparent pK a of this transition is strongly dependent on ionic strength. We have compared the pK a values of the purple to blue transition in the various mutants at a given salt concentration. The pH titration curves of wt and the different mutants in 150 mM KCl are shown in Fig. 3. The absorbance changes at 628 nm are plotted as a function of pH. The data were fitted to Equation 1. In wild type the Hill coefficient of n Ϸ 1.6 indicates a cooperative effect in the protonation/ deprotonation reaction of the Asp-85 carboxyl group (33). In the bR mutant R82A, the pK a of the purple to blue transition is shifted from 2.6 to 7.3 (n ϭ 0.8). In the double mutants R82A/ G72C and R82A/A160C, the pK a values of the purple to blue transition are 7.2 and 7.3, respectively (data not shown), i.e. similar to that of the single mutant R82A. In the double mutant R82A/G231C, however, in which the second mutation is located at the cytoplasmic surface at the end of helix G, the pK a of the purple to blue transition is shifted completely back to the wild type value, even to a somewhat lower number (pK a ϭ 2.1, n ϭ 0.9). The single mutant G231C has a pK a of 2.7 (n ϭ 1.1), similar to that of wild type. Note that for all mutants the n value is close to 1 (0.8 -1.1) and clearly smaller than for wild type (1.6). The pK a and n values are summarized in Table I.
For wt and the mutant R82A/G231C the pH titration was also performed in the regenerated form in CHAPS/DMPC micelles. Doing the titration with bR micelles has the advantage of having a higher fraction of the blue state at alkaline pH values than is the case with membranes. pH-dependent absorption spectra of wt and the mutant were recorded between pH 1 and 9.5. In Fig. 4 the fraction of the blue membrane of the double mutant is plotted versus the pH. The titration data were analyzed according to the model proposed by Balashov et al. (21) using Equation 2. The scheme in the inset of Fig. 4 describes the coupling between the pK values of residues Asp-85 and another group denoted X'H (probably Glu-204) and their dependence on the ionization state of each of the residues. The pK a values of wt and the double mutant R82A/G231C are presented in the scheme. The comparison of the pK a values clearly shows that in the double mutant R82A/G231C the pK a of Asp-85 in the presence of the protonated group X'H is shifted back from 6.9 in R82A (pK in micelles (30)) to 3.0, i.e. lower than the wild type value (pK a ϭ 4.7). This should be compared with the Asp-85 pK a value of 2.6 and 2.1 in membranes for wt and R82A/G231C, respectively (Table I). These results indicate that the electrostatic environments for these residues (Asp-85 and X'H) differ in the double mutant and in wt. Also the pK a of Asp-85 in the presence of the deprotonated group X' Ϫ (like in the M-state), pK a ϭ 5.8, differs slightly from the wild type value (pK a ϭ 6.3). Similar shifts are observed for the pK a of X'H in the double mutant R82A/G231C, i.e. the pK a of X'H in the presence of the deprotonated Asp-85 is 7.5 for wild type and 6.7 for the double mutant. The corresponding numbers in the membrane fragments are 9.6 and 9.0, respectively. The kinetics of light-dark adaptation of the chromophore in the unphotolyzed state of bR is also controlled by the protonation state of Asp-85 (12). In the light-adapted state the retinylidene chromophore is 100% in the all trans conformation. In the dark adapted state, the chromophore is in an equilibrium between the all trans (34%) and 13-cis (66%) configurations (34). In wt-bR the pH dependence of the dark adaptation rate and the fraction of molecules in the blue state were shown to be the same (12). We have measured the dark adaptation rate for wt, R82A, G231C, and R82A/G231C at selected pH values (data not shown). wt and the single mutant G231C have similar values of k DA Ϸ 2.3-2.5 ϫ 10 Ϫ4 s Ϫ1 at pH 6. In the double mutant R82A/G231C a slightly higher value of 3.3 ϫ 10 Ϫ4 s Ϫ1 (pH 6) was observed. The fastest rate was observed for R82A with k DA Ϸ 120 ϫ 10 Ϫ4 s Ϫ1 at pH 7.3. In wt the fastest rates are  Table I. measured at acidic pH, where also the highest fraction of blue membrane was observed. Therefore, the faster rate for R82A, k DA Ϸ 120 ϫ 10 Ϫ4 s Ϫ1 , compared with wt, k DA Ϸ 2.3 ϫ 10 Ϫ4 s Ϫ1 , and the close values for wt, G231C, and R82A/G231C are in qualitative agreement with the results obtained from the blue to purple chromophore titration.
Photocycle Kinetics-The formation and decay of the M-intermediate represent the deprotonation and reprotonation, respectively, of the Schiff base during the photocycle. We believe that a low pK a of the proton acceptor Asp-85 and a wild typelike dark isomerization rate are prerequisites for wild type like photocycling and proton pumping (for further details see "Discussion"). Because, as shown in the previous section, these two prerequisites are fulfilled in the double mutant R82A/G231C, one may expect a restoration of wild type kinetics of the photocycle and of proton release and uptake. In show that in comparison to wild type the rise of M in R82A is accelerated by approximately a factor of 10, whereas the decay is faster by less than a factor of 2. In the double mutant R82A/G231C on the other hand, the kinetics of the rise of M are virtually identical to that of wild type, and the decay is delayed by about a factor of 2. The major effect in R82A, the greatly altered rise of M, is thus restored in the double mutant. Fig. 6 shows the photocycle kinetics at three selected wavelengths (410, 570, and 650 nm) of wild type and the mutants G231C and R82A/G231C. These three wavelengths were chosen because they are diagnostic for the kinetics of the M intermediate (410 nm) of the L intermediate and ground state depletion (570 nm) and of the K and O intermediates (650 nm). The data of Fig. 6 show that the photocycles of these three samples are closely similar until about 1 ms. Small differences are apparent in the second half of the cycle with a slow down of the later intermediates N and O in particular in G231C. A multiexponential global fit of the photocycle, measured at 17 different wavelengths between 370 and 690 nm with a sum of 7 exponentials, was applied for each mutant and wild type. The fit results in the following seven time constants for wild type, 1.2/40/138 s and 0.71/2.11/6.18/15.2 ms; for G231C, 1.0/22/98 s and 1.6/8.9/41/204 ms; and for R82A/G231C, 1/30/129 s and 0.8/2.1/7.9/50 ms. These numbers confirm that in both mutants, G231C and R82A/G231C, the kinetics of the second half of the photocycle are somewhat slower than for wt (Fig. 6). The difference is smallest for the double mutant. The changed kinetics compared with wt may represent a different quasi-stationary equilibrium between M 7 N/O 3 bR. The equilibrium between M, N, and O is pH-dependent (7,35). Therefore, the kinetics of the M-intermediate of wild type and R82A/G231C were measured in the pH range from 7 to 11. In

TABLE II Kinetics of M formation and proton release
The M-formation is measured as the absorbance increase at 410 nm in 150 mM KCl, pH 7.3, at 22°C. The relative contributions of the three rise components are given in percentages in parentheses following the rise times. When two components occur in the H ϩ release, the times are separated by a slash. A total gives the total amplitude as a percentage of the wild type amplitude of the M-intermediate. The proton release is detected with fluorescein bound to the cysteine at the indicated position at the protein surface and with pyranine in the aqueous bulk phase. A gives the amplitude of the proton signal as percent of the wild type amplitude of the proton signal.  ent, 89 mOD for wild type and 70 mOD for R82A/G231C (amplitudes from a multiexponential fit; the apparent amplitudes in Fig. 7 are approximately 74 and 64 mOD, respectively). The overall conclusion, however, deduced from the multiexponential fits of the time traces shown in Fig. 7 (Fig. 8, inset). This results in the overall smaller amplitude of the M-intermediate. The pK a values are 9.6 for wt and 9.1 for R82A/G231C. It was shown previously, that this pK a value represents the pK a of the terminal proton release group (X'H) in the unphotolyzed state (13). The lower pK a of the terminal proton release group in the double mutant compared with wt is in agreement with the results for the group X'H from the pH titration of the unphotolyzed protein (pK a ϭ 9.6 and 9.0 for wt and R82A/G231C, respectively, in bR membrane fragments) described in the previous section. Kinetic Isotope Effects in the Photocycle of Wild Type and R82A/G231C-Kinetic isotope effects because of H/D exchange allow insight into the nature of proton transfer steps inside the protein (36). We have performed photocycle measurements of wt, R82A, and R82A/G231C in H 2 O and D 2 O at selected wavelengths. Fig. 9 shows the time traces at 410 nm for the two mutants (A and B) and wt (C) in 150 mM KCl and pH/pD 8. The measuring conditions were chosen such that the mutant R82A is more than 50% in the purple state. The isotope effects on the kinetics of the main M-rise components are smaller for R82A (A) than for R82A/G231C (B) and wt (C). For a better comparison of the multiexponential M-rise kinetics (see Table II)  Proton Release and Uptake Kinetics-Proton concentration changes at the protein surface and in the aqueous bulk phase were detected with the surface bound dye fluorescein and the bulk pH indicator pyranine, respectively. Acetamido-or methylfluorescein was covalently bound to single cysteine residues introduced by site-directed mutagenesis at position Gly-72 (BC loop) on the extracellular side and at position Ala-160 (EF loop) and Gly-231 (C-terminal tail) on the cytoplasmic side of bacteriorhodopsin ( Fig. 1) both in the single cysteine mutants and in the double mutants R82A/G72C, R82A/A160C, and R82A/ G231C. Using the method described under "Experimental Procedures," on average 0.65-0.95 mol of BMF or 0.5-0.7 mol of 5-iodoacetamidofluorescein were normally incorporated/mol of mutant bR. In wild type bR, which lacks cysteine, less than 5 mol% fluorescein was bound under these conditions. A major change in the reactivity of the dye was observed for the cysteine residue in position 231. In the single cysteine mutant G231C, a labeling stoichiometry of only 0.15-0.2 mol 5-iodoacetamidofluorescein/mol bR was obtained, whereas in the double mutant R82A/G231C an approximately 3-fold higher stoichiometry (0.55-0.7) was observed under identical conditions. In Fig. 10 the kinetics of the rise and decay of the M-intermediate (upper panel) are compared with the kinetics of proton release and uptake, detected in the aqueous bulk phase (lower panel) of R82A (Fig. 10A) and at the protein surface (Fig. 10,  B-D, lower panel) of the three double mutants R82A/G72C, R82A/A160C, and R82A/G231C, respectively. All measurements were performed with membrane fragments under the same conditions (150 mM KCl, pH 7.3, and 22°C). All proton release and uptake times as detected by fluorescein and pyranine for wt and the mutants are collected together with the time constants and amplitudes for the kinetics of M in Tables II  and III. Note from Fig. 10 and Tables II and III that in contrast  to the double mutant R82A/G231C, the photocycles and the  proton signals, measured (Tables II and III). With R82A/ G231C on the other hand, proton release occurred first (0.58 ms) followed by proton uptake (11.8 ms), and these values are virtually identical to those obtained with wild type (Tables II  and III).
With the surface-bound dye fluorescein at position A160C at the cytoplasmic surface, two proton uptake components were detected for R82A/A160C with time constants Ͻ1 s (Ϸ40%) and 150 Ϯ 15 s (Ϸ60%) (Fig. 10C, lower panel). The proton uptake with the two time constants was reproducible as verified in three independent experiments. The proton uptake measured with the dye bound at position G72C on the extracellular surface was fitted with two time constants of 1 ϭ 84 Ϯ 10 s (Ϸ50%) and 2 ϭ 185 Ϯ 12 s (Ϸ50%) (Fig. 10B, lower  panel). For both double mutants, the proton uptake time constants detected at the protein surface are in the s time range and thus at least 50 times faster than those detected in the bulk phase (milliseconds). Moreover, they clearly precede the normal H ϩ uptake time in wild type, detected at the protein surface (ms time range), also by more than one order of magnitude. The main proton release component at the surface of R82A is about 1 ms ( ϭ 0.73 ms for R82A/G72C and 1 ϭ 1.4 ms for R82A/A160C) or slower ( 2 ϭ 14 ms for R82A/A160C). Our experimental data thus clearly show that in the mutant R82A the release is slowed down, and the uptake is accelerated compared with wt (proton release ϳ71 s and proton uptake ϳ8ms, as measured with the wt-like bR-mutant protein G72C-AF (25)).
In contrast to the double mutants R82A/G72C and R82A/ A160C, which have photocycles similar to that of the single mutant R82A, the photocycle of R82A/G231C is similar to wt as described in the preceding two sections. It is thus not very surprising that the kinetics of the proton signals detected at the protein surface are also similar to wild type (Fig. 10D, lower  panel, and Tables II and III). In particular the order is such that proton release (75 s) precedes uptake (6 and 10.8 ms).

Kinetics of Proton Release and Uptake
Detected at the Surface of the bR Mutant R82A-The reversed order of proton release and uptake in R82A and R82Q, as detected with the pH indicator dye pyranine in the aqueous bulk phase, was first reported by Otto et al. (20) using DMPC/CHAPS micelles. The effect of the arginine 82 to alanine mutation on dark adaptation, proton release, and photochemical cycle in membrane fragments was analyzed by Balashov et al. (21). These authors confirmed the reversed order of proton release and uptake observed in Ref. 20 using the same pH indicator dye, obtaining times of 8 ms (uptake) and 30 ms (release). Our measurements of the proton transfer kinetics in the aqueous bulk phase in the single mutant R82A also give proton uptake kinetics in the same time range of Ϸ 7 ms (Table III and Fig. 10A). The proton release kinetics is somewhat faster (10 ms compared with Ϸ30 ms) as described in Ref. 21, and this difference may be because of the different salt concentrations used in the experiments (150 mM and 2 M salt, respectively). The analysis of the kinetics of proton concentration change measured with pyranine in the aqueous bulk phase is complicated by the fact that it represents the kinetics of the H ϩ transfer between the protein surface and the bulk phase (6,25,28,37,38). Surface attached pH indicators dyes at both the extracellular and cytoplasmic protein surface overcome this limitation (25,37,39). We therefore constructed double mutants with the R82A substitution as well as another residue at either surface mutated to cysteine as an attachment site for the pH indicator dye. We chose G72C at the extracellular surface and A160C at the cytoplasmic surface. Both single cysteine mutants exhibit photocycle kinetics like wild type, with minor changes in the second half of the photocycle for A160C (25). The photocycles of R82A, R82A/G72C, and R82A/A160C are very similar. The M-rise was fitted with two time constants, 1 ϳ300 -700 ns and  2 ϳ5-10 s (Fig. 10). Two H ϩ uptake components were observed both at the cytoplasmic and extracellular surface, which precede the proton release. The very fast proton uptake time constant at the cytoplasmic surface of Ͻ1 s seems to be correlated with the ns M-rise component. On the extracellular side (R82A/G72C) the fastest uptake component is about 85 s. The proton uptake time of Ͻ1 s detected on the cytoplasmic surface at position 160, faster than the one detected on the extracellular surface, shows clearly that uptake occurs at this side as in wild type. On the other hand, when the uptake occurs on the same side as in wild type (i.e. cytoplasmic side) with a time constant faster ( Ͻ 1 s as detected in position 160) than the equilibration time of about 70 -80 s for a proton around the bR membrane fragment (25)  Furthermore, net proton transfer is observed in the electrical measurements of the purple form of R82A. 2 Together with our results that proton uptake in R82A clearly takes place at the cytoplasmic surface, we conclude that the direction of proton pumping is the same as in wild type, with proton release at the extracellular and uptake at the cytoplasmic surface.
From studies of the double mutant R82Q/D96N (40) it was concluded that proton release and uptake are independent processes. This interpretation was based on the fact that the proton uptake time in R82Q, monitored in the aqueous bulk phase, was virtually the same as in wild type. In the double mutant R82Q/D96N proton uptake was delayed as in the single mutant D96N. Our results, on the other hand, provide evidence that the mutation R82A, the substitution of a residue in the proton release channel close to the extracellular surface, affects not only the proton release, but in particular the proton uptake time at the opposite surface (Ͻ1 s compared with ϳ8 ms in the wt-like mutant A160C), as monitored directly at the cytoplasmatic protein surface. This suggests long range interactions between the extracellular and cytoplasmic domains of the protein.
Comparison of Wild Type and the Second Site Revertant R82A/G231C, Implications for the Revertancy Mechanism-Double mutants that restore the wild type phenotype in a single mutant with altered function are of particular interest because they allow the comparison of two proteins with different primary structures but identical phenotypes. For instance, the removal of a charged residue can often be compensated by a second mutation, which removes the opposite charge at a nearby position and therefore restores the electrostatic environment in this particular part of the protein. An example of such a short range electrostatic interaction in bR is the double mutant R82Q/D212N (41,42). In this double mutant the complex counterion of the protonated Schiff base, which involves Asp-85, Arg-82, and Asp-212, is simplified by removing both the positive charge at position 82 and the nearby negative charge at position 212, resulting in a dipole instead of the original quadrupole. Even though the photochemical properties are only partially restored, a proton signal with the normal order of release and uptake and similar time constants as in wild type are observed in the aqueous bulk phase (41). Long range electrostatic interactions have been observed in second site suppressor mutants of the photosynthetic reaction center of Rhodobacter capsulatus (43). These second site mutations restore proton transfer, which is interrupted in the initial single-site mutant protein. It was suggested that the compensatory mutation propagates its effect over large distances by mutation-induced realignments of salt bridges within a network of ionizable residues.
The second site revertants of mutants containing single replacements at position 82 in bR (like R82A and R82Q) are of particular interest, because Arg-82 has been proposed to be involved in multiple types of interactions, as part of the counterion of the Schiff base proton (electrostatic interaction), in interactions with the primary and terminal proton release groups (electrostatic and structural-conformational interactions), and during folding (structural-electrostatic interaction). Though the second site revertant R82A/G231C displays in many respects wild type properties (pK a of Asp-85, rate of light-dark adaptation, kinetics of photocycle and of proton release and uptake, and kinetic isotope effects), small but distinct differences exist: (i) the pK a of Asp-85 is down shifted by 0.5 pH units compared with wt, (ii) the Hill coefficient is 0.9 in R82A/ G231C but 1.6 in wt, (iii) the pK a of the terminal release group (X'H) was determined both in the unphotolyzed protein and from the pH dependence of the M-rise to be about 0.5 pH units lower than in wt, and (iv) the amplitude of the M-rise and of the proton signal are about 80% of the wt amplitude.
These differences may help to explain why the mutational change in position 231 at the cytoplasmic surface far away from position 82 (see Fig. 1) is able to revert many of the altered properties in R82A back to wild type. Based on our results and structural evidence from electron microscope and x-ray crystallographic data (9, 10), we present a hypothesis for the mechanism of the rescue in R82A/G231C. One possible explanation of the revertancy observed for R82A/G231C is a rearrangement of the native H-bonded network connected with Asp-212, leading to an equivalent hydrogen-bonded network as in wild type and facilitating proton release to the extracellular surface but without the participation of Asp-212 and Arg-82. The following arguments support this hypothesis.
In wild type, Asp-212 (helix G) presumably is directly hydrogen-bonded to Tyr-185 (helix F), Tyr-57 (helix B) and via water molecules (44) to Arg-82 and Asp-85 (helix C) forming a threedimensional hydrogen-bonding network (9,10,23). Asp-85 and Asp-212, together with Arg-82 form the complex counterion to the protonated Schiff base in the retinal binding pocket (45). Note, that the neutral alanine residue in the mutant R82A is not capable of participating in a hydrogen-bonded network. We found that the most striking similarity between the mutants R82A, G231C, and R82A/G231C is the loss of cooperativity in the purple to blue transition (n ϭ 1.6 in wild type to n ϭ 0.8 -1.1 in the mutants indicated). In addition, the pK a values of both Asp-85 and the terminal release group X'H in the unphotolyzed protein R82A/G231C are down shifted by about 0.5 pH units compared with wt. These results indicate a changed electrostatic environment of Asp-85 in the complex counterion and an altered interaction of ionizable residues in the proton release channel. This different electrostatic environment in the retinal binding pocket of R82A/G231C compensates for the removed positive charge of the arginine residue. A compensation of this positive charge might be realized by a removal of the negative charge of Asp-212 from the vicinity of Asp-85. This could occur by a structural change or by protonation. Conformational changes on the cytoplasmic surface close to helix G are suggested from the altered labeling stoichiometry with 5-iodoacetamidofluorescein observed between G231C and R82A/G231C. Asp-212 resides in the same helix (helix G) as Gly-231. A conformational change in helix G introduced by the mutation G231C can lead to a different interaction of the Asp-212 side chain with the neighboring residues in R82A/G231C in such a way that the side chain of Asp-212 no longer participates in the complex counterion. Recent time-resolved electrical measurements (19) provide evidence that the movement of the positively charged arginine 82 side chain makes a substantial contribution to the amplitude of the electrical signal in wt. This amplitude is clearly reduced in the mutants G231C and R82A/ G231C. 2 This result can be interpreted with the absence of the movement of Arg-82 already in the single mutant G231C. Note that the loss of cooperativity in the purple to blue transition was also observed in the single mutant G231C. Therefore, it might be that in the mutant protein G231C the arginine 82 side chain is already pointing outward to the extracellular surface and does not contribute to the proton release process. A replacement of Arg-82 with alanine is then not expected to alter substantially the properties of G231C. Furthermore, the results of the H/D isotope effect on the kinetics of the M-rise, on the other hand, indicate a reestablishing of a wt-like hydrogenbonded network in the proton release channel in the mutant R82A/G231C.
Among the double mutants with R82A or R82Q previously described in the literature (30, 40 -42), only one, R82Q/D212N (41,42), with the second site mutation nearby at position 212 in helix G, is able to partially rescue the normal functional features, like normal kinetics of proton release and uptake and pK a of Asp-85 (41). Also the down shift of the pK a of Asp-85 compared with wt, similar to R82A/G231C, was reported in the double mutant R82Q/D212N. Interestingly, the compensatory mutation found in our study, G231C, is like Asp-212 also located in helix G. Whereas Arg-82 and Asp-212 are close together, Gly-231 is located at the end of helix G far away from Arg-82 (see Fig. 1). Therefore, in our case the effect of G231C on R82A in the double mutant R82A/G231C is clearly not a short range electrostatic compensation as in R82Q/D212N where a neutral charge pair was removed, but rather long range, and suggests a coupling of the two residues in the different parts of the protein. This coupling seems to mediated by an interaction of the helices C and G.
A final remaining question is concerned with the origin of the putative conformational changes in helix G leading to the altered electrostatic interaction in the counterion environment. It is known that arginines are stabilizing elements in proteins (46). In peptides, Lys 3 Arg substitutions increase the helix content of designed helical peptides. It was suggested that a single strategically placed arginine can exert long range control on helix structure (46). Two arginines, Arg-225 and Arg-227, are located at the cytoplasmic end of helix G, one helix turn below Gly-231. Arg-227 mutations affect the kinetics of Mdecay (47) and proton uptake (48), indicating the interaction of the positively charged guanidinium group with a hydrogenbonded network, which is proposed to be part of a proton uptake cluster (48). Exchange of the small unpolar and highly flexible glycine with the polar and ionizable cysteine in position 231 may lead to an alteration in the hydrogen-bonded network and/or water-filled cavities and to changes in the peptide backbone. On the other hand, the removal of all C-terminal amino acids in the mutant protein R82A by enzymatic digestion with papain, which cuts between residues 231 and 232 and leaves Gly-231 as the terminal residue, does not lead to the rescue as observed in the double mutant R82A/G231C but preserves the properties of R82A. The absorption maximum of the chromophore in the C terminus-less R82A mutant protein is 600 nm at pH 6 and the pK of the purple to blue transition is 7.2. These values are similar to the ones observed in the full-length R82A. This finding implies that all residues following Gly-231 in the C terminus (232-248) are not involved in the rescue mechanism, at least not by their absence. Further, it supports the idea that the residues involved in the revertancy mechanism are located in the helical region (helix G). Among the three double mutants involving R82A studied in this report, the five double or triple mutants involving R82A described in the literature (30, 40 -42) and the double mutant R82A/T178V, 3 only R82A/G231C reverts the properties of R82A(Q) back to wt. Except for the double mutants with the second mutation in the BC-(R82A/G72C) and EF-loop (R82A/A160C), the second or third mutation is located either in helix C, F, or G. Residues in these three helices are supposed to participate in the hydrogenbonded network facilitating proton pumping (e.g. Asp-96, Asp-212, Asp-85, Tyr-185, and Glu-204). However, whether double mutants other than R82A/G231C are able to rescue R82A and which residues are involved in the revertancy mechanism is subject to further extensive mutagenesis studies.
In conclusion, we have shown that changes at the cytoplasmic surface can alter structural and functional interactions, which occur in the extracellular part of the protein and vice versa, probably because of a rearrangement of H-bonded networks mediated by conformational changes. We demonstrated with experiments on the mutants R82A, G231C, and R82A/ G231C that allosteric interactions occur between the extracellular and cytoplasmic surfaces of bacteriorhodopsin. Such long range propagation of electrostatic and conformational changes is of potential importance for signal transduction in heptahelical G-protein-coupled receptors, where binding of ligands occurs mostly on the extracellular side resulting in the active form of the receptor and where binding and activation of the G-protein takes place at the cytoplasmic surface.