Contribution of extracellular Glu residues to the structure and function of bacteriorhodopsin. Presence of specific cation-binding sites.

Single and multiple mutants of extracellular Glu side chains of bacteriorhodopsin were analyzed by acid and calcium titration, differential scanning calorimetry, and thermal difference spectrophotometry. Acid titration spectra show that the second group protonating with Asp(85) is revealed in E204Q in the absence of Cl(-) but is not observed in the triple mutant E9Q/E194Q/E204Q or in the quadruple mutant E9Q/E74Q/E194Q/E204Q. The results point to Glu(9) as the second group protonating cooperatively with Asp(85). Comparison of the apparent pK(a) of Asp(85) protonation in water and in the deionized forms and results of calcium titration suggest that cation-binding sites are of low affinity in the multiple Glu mutants. Like for deionized wild type bacteriorhodopsin, differential scanning calorimetry reveals a lack of the pretransition in the multiple mutants, whereas in E9Q it appears at lower temperature and with lower cooperativity. Additionally, at neutral pH the band at 630 nm arising from cation release upon temperature increase is absent for the multiple mutants. Based on these results, we propose the presence of two cation-binding sites in the extracellular region of bacteriorhodopsin having as ligands Glu(9), Glu(194), Glu(204), and water molecules.

The elucidation of three-dimensional structures of bacteriorhodopsin (BR) 1 at high resolution and the increased use of mutants have improved greatly the knowledge of the proton transport mechanism, allowing the identification of the principal side chains undergoing protonations/deprotonations during this process (1)(2)(3)(4). Among key groups, Asp 85 and Asp 96 are the primary proton acceptor and proton donor of the Schiff base, respectively. An important achievement has been the description of several water molecules in the extracellular region, forming a hydrogen-bonded network with key residues (2,5,6). This network serves most likely to transmit protonation changes and to conduct the proton. The role of Asp 85 in the extracellular region is not limited to be the first acceptor of the Schiff base proton. According to a coupling model, protonation of Asp 85 induces the deprotonation of the proton release group through a relationship existing between their pK a values (7,8). Although this model is useful in understanding protonation/ deprotonation mutual influences between ionizable groups, it cannot give the answer about the exact mechanism of proton transfer steps. In fact, several important questions remain still elusive. One of these refers to the identity of the assemblies involved in proton release, the proton release group. Recent studies suggest that it is a complex including at least Glu 194 , Glu 204 , and some water molecules (9 -11). It should be noted, however, that these two Glu side chains do not behave equivalently, as demonstrated by the inhibition of the second increase of the Asp 85 pK a during the photocycle in the E194Q mutant but not in the E204Q mutant (12). This is a clear indication that Glu 194 forms part of XЈH, the group coupled to the Asp 85 pK a .
From the time when the presence of divalent cations in the purple membrane was described, numerous efforts have been devoted to identify their role and location (13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26). Some authors claim a nonspecific location of the cations in the double layer (18,21,22) or in the lipid phase (26). However, the majority of works argue for the existence of specific binding sites in the purple membrane, including a site near Asp 85 (24,27,28), or a more external location (29,30). Among the most outstanding results pointing toward specific cation locations in the protein moiety are: (i) the evidence for carboxyl participation in cation binding from studies using specific reagents (13,17); (ii) the presence of fewer cation-binding sites and with lower affinity in the bleached and pink membranes as compared with purple membranes (14,16,19), even that the electric surface potential of bleached membrane remains unchanged (31); (iii) extended x-ray absorption fine structure data describing a different environment for the bound Mn 2ϩ with respect to free Mn 2ϩ in water and ruling out the interaction of Mn 2ϩ with P or S atoms (32); (iv) Fourier transform infrared studies showing that binding of Mn 2ϩ to deionized membrane induces changes in the reverse turns, located in the loops (33); and (v) 13 C NMR studies detecting changes in the Ala 196 environment upon divalent cation binding (25).
Despite the substantial experimental background suggesting the existence of specific cation-binding sites, none of the recent bacteriorhodopsin models reflects their presence (2,5,34). One possible exception is the electron microscopy structural data of Mitsuoka et al. (35), which detected the presence of charges and polarized water molecules in some locations of their structure.
Recently we have described some spectral and functional properties of the quadruple mutant E9Q/E74Q/E194Q/E204Q (4Glu), including high hydroxylamine and Cl Ϫ accessibility to the Schiff base environment and severalfold decreased effect of Ca 2ϩ on the regeneration of the purple color of deionized 4Glu (36). These results suggest a more open structure for the extracellular region of the 4Glu mutant and reduced affinity of cations for the color-controlling binding site(s) as compared with wild type. Therefore, we proposed that the extracellular Glu side chains could be involved in cation binding and in the maintenance of the correct protein structure. To get further insight on the role of extracellular Glu side chains in these issues, as well as in proton transport, in this work we study BR mutants in which the extracellular Glu have been substituted by Gln. Even if these changes are very conservative, Gln cannot deprotonate on one hand and has a reduced capability of forming hydrogen bonds on the other (37). Although some characteristics of these mutants (especially E194Q and E204Q) have already been published (11, 26, 38 -40), we complete these studies by presenting new data in a thorough manner. Finally, we propose a model of cation binding in the extracellular region, involved in protein structure and function.

EXPERIMENTAL PROCEDURES
The construction of the mutants and their expression in Halobacterium salinarum was done as described previously (12,36). The wild type and mutagenized membranes were isolated as purple membrane sheets according to Oesterhelt and Stoeckenius (41). Deionized membranes were obtained by dialyzing the samples against a cation exchange resin (Dowex-50W). All manipulations of deionized samples were done using well rinsed plastic materials. Concentration of deionized samples was calculated by the method of Lowry et al. (42), modified for membrane proteins (43). Absorption spectra were recorded on a Varian Cary3-Bio spectrophotometer in the 250 -800-nm range. To avoid light-scattering artifacts, an integrating sphere was used.
The pH of purple and deionized membranes (at a concentration of 1.5-2.0 ϫ 10 Ϫ5 M) was adjusted by microadditions of NaOH, HCl, or H 2 SO 4 solutions. To avoid contamination, pH adjustment of deionized samples was done using duplicates. All experiments were performed in the dark using dark-adapted samples. Because of abnormal dark adaptation kinetics of the E194Q, 3Glu, and 4Glu mutants (36), the samples were kept in the dark for about 25 days.
Absorbance changes at 630 nm as a function of pH were used to monitor the purple-to-blue transition. Experimental points were normalized with respect to the largest value at 630 nm and fitted to the Boltzmann equation.
Calcium binding experiments were done in darkness with deionized samples adjusted to pH 4 -4.5. Absorbance changes induced by the addition of small quantities of calcium were monitored spectrophotometrically. BR concentration was in the 1.5-2 ϫ 10 Ϫ5 M range.
DSC experiments were performed using a Micro-Cal MC2 instrument. The samples were dialyzed previously in water adjusted at pH 6.5-7, giving a final concentration of 1.5-2 mg/ml. Experiments were done under 1.7 atm nitrogen pressure to avoid sample evaporation at high temperatures. Scanning speed was set at 1.5 K/min. Three thermograms were registered for each sample. The first informs about the heat released or taken by the protein upon temperature increase. After cooling down to room temperature, second and a third thermograms were run to check the reversibility of the transitions. Two corrections were applied to the first thermogram: (i) subtraction of the second thermogram, which acts as a blank, and (ii) subtraction of the chemical base line using the Takahashi and Sturtevant method (44). T m was defined as the point where the C p value is maximal.
Thermal denaturation experiments were carried out on darkadapted samples in H 2 O and pH 7.0 by following changes in the UVvisible absorption spectra upon temperature increase. Spectra were taken every 5°C in the range 250 -800 nm, starting at 20°C and allowing 8 min for stabilization at each temperature. BR concentration was 1 ϫ 10 Ϫ5 M.

Spectral Behavior of Extracellular Glu Mutants upon
Acidification: the Purple-to-Blue Transition As has been widely described, acidification of purple membrane samples causes the formation of the blue form, because of Asp 85 protonation (45,46). Fig. 1 (A and B) shows the absorbance and difference spectra of the dark-adapted E74Q mutant in 150 mM KCl upon acidification of the medium from pH 5.7 to 2.1. As in wild type, the difference spectra of E74Q reveal the presence of two isosbestic points at about 620 and 578 nm. The isosbestic point at about 620 nm reflects a transition that produces a small band broadening in the red edge of the spectrum and a small blue shift (47,48), yielding the so-called FIG. 1. Purple-to-blue transition of dark-adapted E74Q and E204Q mutants. A, curves 1-8, absorption spectra of the E74Q mutant from pH 5.7 to 2.1 in the presence of 150 mM KCl. B, curves 2-8, difference absorption spectra between the sample at pH i minus the sample at pH 5.7, where pH i is 4.3, 3.8, 3.5, 3.0, 2.8, 2.6, and 2.1, from the spectra in A. C, curves 1-10, absorption spectra of the E204Q mutant from pH 7.5 to 1.7 in H 2 O titrated with H 2 SO 4. D, curves 2-10, difference spectra between the sample at pH i minus the sample at pH 7.5, where pH i is 6.7, 6.1, 5.6, 4.7, 4.3, 3.9, 3.3, 2.8, and 1.7, from the spectra in C.
BR acid form. The second isosbestic point at 578 nm corresponds to the red shift of the spectrum because of protonation of Asp 85 , giving the principal transition (the purple-to-blue transition). Similar spectral changes upon pH decrease are obtained for E74Q in H 2 O and for dark-adapted E9Q and E194Q single mutants in 150 mM KCl or in H 2 O (not shown).
Acidification of the dark-adapted E204Q mutant reveals more complex spectral changes. Whereas in 150 mM KCl a similar behavior to wild type is obtained, in H 2 O or in 75 mM Na 2 SO 4 where Cl Ϫ ions are not present, the absorption maximum undergoes a small blue shift and band broadening in both sides ( Fig. 1, C and D). This gives rise to two positive bands at 640 and 460 nm in the difference spectra (Fig. 1D). Further decrease of the pH leads to the main transition, where all species are converted to the blue form. This produces a positive band at 628 nm and a negative band at 538 nm.
Titration of deionized wild type and deionized forms of the single mutants E9Q, E74Q, and E194Q by increasing the pH from 4 (the initial pH after deionization) to 8 gives only one isosbestic point at 577 nm in the difference spectra ( Fig. 2A for the spectra of deionized E9Q). However, titration of deionized E204Q shows not only that the red form persists but also that it is better distinguished than in the presence of Na 2 SO 4 ( Fig.  2B). Thus, E204Q is unique among wild type and the single Glu mutants, in that titration of the deionized sample displays changes in the protonation state of more than one group. Moreover, the max never reaches 603 nm in any of the condition examined (not shown).
Upon acid titration, the multiple extracellular Glu mutants undergo spectral changes different from those of the single mutants. The difference spectra of 3Glu or 4Glu mutants do not show the red band at about 460 nm in any of the conditions analyzed (salt, H 2 O, deionized; see Fig. 2C for difference spectra of 3Glu in 75 mM Na 2 SO 4 ). However, acidification of the E194Q/E204Q mutant in Na 2 SO 4 or in water gives rise to the red band (Fig. 2D), arising from a small blue shift and broadening of the absorption band, similar to the single E204Q mutant.

Determination of the Apparent Asp 85 pK a
As is known, the plot of the absorbance increase at 630 nm as a function of pH yields the apparent pK a of Asp 85 (45,46). Table  I shows the calculated values for pK a of Asp 85 for all mutants and wild type in three different conditions: 150 mM KCl, water, and deionized membranes. Values for the Asp 85 pK a similar to those found in the presence of 150 mM KCl were obtained upon titration in 75 mM Na 2 SO 4 (not shown). In low salt concentration (150 mM KCl) all single mutants gave similar values (pK a ϳ 2.7, although E194Q shows slightly lower pK a ), whereas the pK a values of the multiple mutants are elevated (around one unit above wild type). In water, the absence of ions in the medium induces more negative electrical surface potential and increased proton concentration, giving rise to increased an apparent pK a value as compared with the salt-containing samples. In water, all samples give higher pK a values as compared with the same samples in salt, except the E74Q mutant (pK a about 2.8, similar to that in salt). Multiple mutants E194Q/ E204Q and 4Glu have an identical and elevated pK a of 4.7, whereas 3Glu has a pK a of 5.2 ( Table I).
Analysis of deionized membrane samples permit the evaluation of the effect of mutations themselves over the Asp 85 pK a , regardless of the presence or absence of endogenous cations. Additionally, comparison of the sample properties in water with those of deionized form allows estimation of the effects of the endogenous cations. As Table I shows, all deionized mutants give slightly lower pK a values as compared with deionized wild type, except for E9Q, which presents an increased pK a of 6.0. On the other hand, the absence of endogenous cations in the deionized form as compared with the sample in water cause an increase of the pK a of more than 2 pH units for wild type (13,49) and similar values for the single mutants except for E204Q. It is noteworthy that the multiple mutants exhibit differences of less than 1 pH unit between pK a values of the deionized form and in water. This suggests that the nondeionized forms of these samples have low content of endogenous cations, thus affecting only slightly the apparent Asp 85 pK a in comparison with deionized sample.

Addition of Calcium to Deionized Bacteriorhodopsin
When calcium or some other cations are added to blue deionized samples, the purple form of bacteriorhodopsin is recovered, because of the deprotonation of Asp 85 (13,17,49). Fig. 3A shows typical difference spectra obtained upon calcium addition to the wild type blue form. The presence of only one isosbestic point at 578 nm strongly indicates that only the deprotonation of Asp 85 is involved in the process, leading to a sigmoidal curve in the plot of absorbance change as a function of pCa (Fig. 3B).
The pCa values for 50% of the blue-to-purple transition are shown in Table II. As can be seen, the most drastic changes in Ca 2ϩ binding, as compared with wild type are obtained for the multiple mutants where the quadruple mutant 4Glu needs more than 10 Ca 2ϩ /BR molecule, 3Glu needs about 8 Ca 2ϩ /BR, and E194Q/E204Q needs 3.5 Ca 2ϩ /BR. Among the single mutants, only E204Q and E9Q need higher Ca 2ϩ amounts compared with wild type.

Thermal Denaturation Experiments
Differential Scanning Calorimetry-It has been demonstrated previously that BR samples with decreased cation content have lower thermal stability (14,50). To study the involvement of the mutated Glu side chains on thermal stability, DSC experiments were carried out. Fig. 4 shows thermograms of extracellular mutants in H 2 O, obtained after correction for instrumental and chemical base lines (see "Experimental Procedures"). A known feature of the thermal scan of native purple membranes is the presence of two transitions (51). The main transition at about 98°C accounts for the destruction of the helical interactions within the protein and retinal release. The small and reversible pretransition at about 80°C has been interpreted as resulting from disorganization of the hexagonal para-crystalline arrangement (51,52).
As is seen in Fig. 4, the main transition of all mutants appears in the range 90 -100°C, decreasing in the order E74Q Ͼ E194Q Ͼ E9Q ϭ E204Q Ͼ E194Q/E204Q Ͼ 4Glu. Besides that, the transitions are less cooperative except for E74Q. In comparison with the main transition, the temperature of the pretransition of the mutants is more variable. It appears at lower temperatures and with low cooperativity for E194Q, E204Q, and E9Q (at about 68°C for this latter sample), and it is absent for all multiple mutants.
Visible Spectroscopy-Heating the purple membrane suspension induces first the appearance of the blue form, because of cation release (14). This is followed by the appearance of the red form ( max at about 460 nm) and finally by the release of retinal above 80 -85°C. Fig. 5 shows the difference spectra obtained for E9Q and 4Glu on temperature increase, in H 2 O at pH 7.0. Similar to wild type, an increase of temperature from 25 to 55°C of E9Q causes a progressive red shift of the absorption spectrum, which gives rise to the appearance of the band at 630 nm in the difference spectrum. Moreover, the thermally induced blue form is a reversible process because lowering the temperature from 55°C (temperature for which maximum accumulation of blue form is observed) back to 25°C recovers the initial purple form. Importantly, whereas the rest of the single mutants (E74Q, E194Q, and E204Q) also show the temperature-induced band at 630 nm, the multiple mutants lack formation of the blue form, as represented in Fig. 5 for 4Glu. DISCUSSION In the first part of this work, we performed acid titration experiments to acquire new information about the influence of    (48) corroborated this observation and proposed a model postulating that more than one proton is bound cooperatively during acid titration. They identified a new species, the BR acid form, that results from BR by proton binding and appears before the blue form, BR blue . Although the additional group titrated in this process was not identified at that time, neither it is yet known whether it is worthy of mention that the authors suggested Glu 9 and Glu 204 residues as a possibility. However, in the majority of the following work dealing with the purple-to-blue transition, these early observations have been ignored, and only the protonation of Asp 85 has been considered.
As indicated under "Results," single extracellular Glu mutants show similar pH titrations as wild type, with the minor transition caused by protonation of a second group remaining masked but already suggested by the presence of two isosbestic points. The only exception to this behavior corresponds to the E204Q mutant in the absence of Cl Ϫ ions, which shows clearly a band at 460 nm (red form). However, in the presence of Cl Ϫ ions, the red band disappears. One reasonable explanation is that Cl Ϫ binds near Gln 204 , restoring the negative charge and the water network. Therefore, unlike the rest of single mutants, E204Q senses the presence of Cl Ϫ ions in the medium and exhibits different spectral changes upon acid titration. Although at present we are not able to identify the second protein residue protonating cooperatively with Asp 85 , comparison of titration results for single and multiple mutants suggest that it can be one of the acidic side chains located in the extracellular region. Moreover, the fact that the red form is observed for E194Q/E204Q but not for 3Glu or 4Glu points to Glu 9 as being responsible for the second transition.
Our data provide new evidence for cation binding to the extracellular side of the purple membrane. In low salt concentrations, the apparent pK a values of Asp 85 of the single mutants are similar to that of wild type, being independent of the type of salt in the medium (sodium sulfate or chloride). However, titrations in water or in deionized membranes exhibit perturbed pK a values of Asp 85 as compared with wild type. The increased pK a obtained for a particular sample in water, as compared with the same sample in salt can be explained essentially by the increased proton concentration over the negatively charged membrane surface because of the absence of counter ions. Especially in the multiple mutants, the pK a increase is considerably higher than in wild type. This effect can be attributed to a loss of endogenous cations and a subsequent increase of proton concentration on the surface. Similar pK a values found for E74Q in water and in salt can be explained by loss of a negative charge in the membrane surface upon mutation of this external residue.
Comparison of the pK a of the purple-to-blue transition in water with that of the deionized sample gives an indication of the effects of bound cations on this transition and thus reveal their affinities, analogous to the comparison of pK a values between salt and water. Reasonably, large differences in pK a between a sample in water and in the deionized state reflect mainly strong affinity of cations, whereas small differences reflect low affinity of cations. Therefore, the difference of about 2.3 pH units found for wild type indicates the presence of cations bound with high affinity. This difference is strongly reduced to less than 1 pH unit for all multiple mutants, indicating clearly that cations have a low affinity in these samples (Table I). On the other hand, Ca 2ϩ binding experiments to deionized samples give further support to the suggested role of extracellular Glu side chains in cation binding. As described under "Results," the single mutant E9Q and the multiple mutants require higher cation concentrations for reaching the 50% of the blue-to-purple transition as compared with wild type. Thus, both acid titration and Ca 2ϩ binding experiments point to Glu 9 , Glu 194 , and Glu 204 as being involved in cation binding.
Thermal denaturation experiments can provide further information about the role of extracellular side chains in cation binding. Therefore, we analyzed thermal behavior of extracellular mutants by DSC experiments and UV-visible spectrophotometry. In some mutants, the DSC main transition appears at slightly lower temperature as compared with wild type, indicating somewhat decreased stability of the secondary and/or tertiary structures. However, the most interesting effect was obtained for the pretransition, resulting from disorganization of the hexagonal para-crystalline arrangement (51,52). As presented under "Results," there is a concurrence between the lack of the pretransition and the absence of cations. Particularly, the deionized wild type membrane lacks the pretransition but, upon cation addition of at least 2 Mn 2ϩ /BR, it is partly recovered (50). On the other hand, the recently reported decrease in the content of the BR-specific ␣ II helical structure upon temperature increase (54) is likely to be due also to cation release. First, there is a coincidence of the temperature of ␣ II decrease with the temperature of the pretransition (54); second, removal of cations by deionization causes a decrease in the ␣ II content as has been reported previously by Duñ ach et al. (33). Therefore, much experimental evidence indicates that the DSC pretransition is caused by cation release as temperature increases, which in turn induces the cooperative disorganization of the para-crystalline arrangement.
Appearance of the pretransition at lower temperatures and with lower cooperativity for the single mutants E194Q, E204Q, and especially E9Q point out that these mutations induce a loosening of the hexagonal disposition of trimers. Furthermore, the mutants E194Q/E204Q, 3Glu, and 4Glu do not show the pretransition, indicating either that the hexagonal arrangement is strongly perturbed in these mutants or that the disorganization occurs continuously as temperature increases, thus lacking any cooperativity and becoming unobservable by DSC. Most likely, these effects do not come directly from the loss of the Glu negative charges themselves but from the loss of cation-binding sites. There is a complete correlation between DSC and calcium titration data; as more calcium is needed to recover the purple form in the different mutants, the DSC pretransition is less evident and appears at lower temperatures.
Importantly, for the multiple mutants, UV-visible spectrophotometry establishes the absence of the blue form ensuing from cation release upon temperature increase. These results, which are a sign of the absence of cations in these multiple mutants, correlate closely with the observed lack of pretransition.
In agreement with the location of cations in the extracellular region, previous DSC results established the absence of the pretransition when the extracellular BC loop is cleaved but its presence when the cytoplasmatic loop EF is cleaved (55). On the other hand, normal pretransition is obtained for the cytoplasmatic mutants D36N/D38N and D102N/D102N, 2 indicating that neutralization of Asp side chains in the cytoplasmatic side do not affect the pretransition, thus leaving intact the para-crystalline arrangement.
On the basis of our findings, we propose a model of the extracellular region containing two cation-binding sites: one linked to Glu 194 /Glu 204 and the other linked to Glu 9 (Fig. 6). In this model, Glu 194 is postulated to control the Asp 85 pK a in the resting state as in the photocycle, participating in the diffuse water moiety that shares the proton that is released to the bulk (represented in the model by a protonated water molecule). Mainly from DSC and calcium titration results, it can also be suggested that Glu 204 is interacting more strongly with the cation and is less linked with the water network. Thus, it is not coupled with Asp 85 (or only weakly coupled). The different role envisaged for Glu 194 and Glu 204 is in keeping with our recent data demonstrating that, in the presence of Cl Ϫ ions, the Asp 85 pK a during the photocycle depends on Glu 194 but not on Glu 204 (12).
The two cations placed in the extracellular side may have not only structural significance by helping to maintain the correct structure in the extracellular region (36) and the para-crystalline arrangement, but most probably they are also involved in the functioning of the protein. Especially, the cation linked to Glu 194 /Glu 204 may be implicated in the maintenance of the optimal spatial relationship between these two residues and the water molecules that form the hydrogen-bonded network. This cation may constitute an essential element of the proton release machinery, by providing the electrical potential necessary to release the proton. In other words, the cation may help to regulate the pK a of the proton release group for it to be decreased sufficiently in the proton release step. What is more, the cation may have a gate-like function, allowing the proton release group to be reprotonated from the interior of the protein in the last steps of the photocycle but keeping it isolated from the exterior. In this context, it is known that the blue form does not have a normal photocycle, because of the presence of the already protonated Asp 85 (13,47). On the other hand, the purple-deionized form has a normal-like photocycle (56) with a yield of M intermediate around 50% (57) (5). It is suitable for pH Ͼ pK a of cation release. For simplicity, only the side chain ligands to Ca 2ϩ are drawn; the other Ca 2ϩ ligands are supposed to be water molecules. Location of cations are indicated schematically to illustrate that one divalent cation is linked to both Glu 194 and Glu 204 , whereas the second divalent cation has only one side chain, Glu 9 , as ligand. According to the results, the cation linked to Glu 194 / Glu 204 is assumed to be of higher affinity than that linked to Glu9 (see the text). The protonated water molecule interacting with Glu 194 represents the proton release group, which can consist of several water molecules. We use Ca 2ϩ to symbolize the divalent cations, but the physiologically bound Mg 2ϩ may occupy these sites instead. port capacity has not been measured yet. The possibility of a normal photocycle lacking proton transport has been raised before based on electro-optical studies of purple deionized BR (58).
An intriguing question is why the endogenous cations are not identified in the high resolution crystallographic structures published so far. One reason may be that the methods applied for crystal preparation cause loss of bound cations, which are then substituted by monovalent cations. This is likely to occur, because purple membrane is first delipidated by detergent solubilization and finally placed in a highly concentrated Na/ K-P i (59). It is well established that to recover purple membrane from blue, about 50 times more monovalent than divalent cations are needed (17). Moreover, it is known that monovalently regenerated BR turns blue by dilution (49). These facts indicate that monovalent cations have low affinity for BR and thus have high mobility around the binding sites, making them unlikely to be observed by diffraction techniques.
Finally, it should be indicated that other workers have also proposed the occurrence of cation-binding sites in the extracellular region. A site near Glu 194 was proposed based on 13 C NMR studies (25). Cation-binding site(s) near the membrane surface was proposed by Fu et al. (29), and a site involving Glu 204 was anticipated because of theoretical considerations (30). While this article was in the revision process, a new work appeared claiming for a cation site located at less than 9.8 Å from Glu 74 (60). Thus, although no cations have been seen in the crystallographic structures, accumulated evidence suggests the presence of cation-binding sites in BR. Our results place two of them in the extracellular region, linked to Glu 9 , Glu 194 , Glu 204 , and several water molecules.