Glutamic acid 204 is the terminal proton release group at the extracellular surface of bacteriorhodopsin.

We have measured proton release into the medium after proton transfer from the retinal Schiff base to Asp85 in the photocycle and the C=O stretch bands of carboxylic acids in wild type bacteriorhodopsin and the E204Q and E204D mutants. In E204Q, but not in E204D, the normal proton release is absent. Consistent with this, a negative band in the Fourier transform infrared difference spectra at 1700 cm−1 in the wild type, which we now attribute to depletion of the protonated E204, is also absent in E204Q. In E204D, this band is shifted to 1714 cm−1, as expected from the higher frequency for a protonated aspartic than for a glutamic acid. Consistent with their origin from protonated carboxyls, the depletion bands in the wild type and E204D shift in D2O to 1690 and 1703 cm−1, respectively. In the protein structure, Glu204 seems to be connected to the Schiff base region by a chain of hydrogen-bonded water. As with other residues closer to the Schiff base, replacement of Glu204 with glutamine changes the O-H stretch frequency of the bound water molecule near Asp85 that undergoes hydrogen-bonding change in the photocycle. The results therefore identify Glu204 as XH, the earlier postulated residue that is the source of the released proton during the transport, and suggest that its deprotonation is triggered by the protonation of Asp85 through a network that contains water dipoles.

We have measured proton release into the medium after proton transfer from the retinal Schiff base to Asp 85 in the photocycle and the C‫؍‬O stretch bands of carboxylic acids in wild type bacteriorhodopsin and the E204Q and E204D mutants. In E204Q, but not in E204D, the normal proton release is absent. Consistent with this, a negative band in the Fourier transform infrared difference spectra at 1700 cm ؊1 in the wild type, which we now attribute to depletion of the protonated E204, is also absent in E204Q. In E204D, this band is shifted to 1714 cm ؊1 , as expected from the higher frequency for a protonated aspartic than for a glutamic acid. Consistent with their origin from protonated carboxyls, the depletion bands in the wild type and E204D shift in D 2 O to 1690 and 1703 cm ؊1 , respectively. In the protein structure, Glu 204 seems to be connected to the Schiff base region by a chain of hydrogen-bonded water. As with other residues closer to the Schiff base, replacement of Glu 204 with glutamine changes the O-H stretch frequency of the bound water molecule near Asp 85 that undergoes hydrogen-bonding change in the photocycle. The results therefore identify Glu 204 as XH, the earlier postulated residue that is the source of the released proton during the transport, and suggest that its deprotonation is triggered by the protonation of Asp 85 through a network that contains water dipoles.
Bacteriorhodopsin is the light-driven proton pump in the cell membranes of halobacteria. The transport cycle ("photocycle") is initiated by photoisomerization of retinal from all-trans to 13-cis and the ensuing transfer of a proton from the retinal Schiff base to Asp 85 , an anionic residue that lies between the Schiff base and the extracellular surface of the protein (Mathies et al., 1991;Tittor, 1991;Ebrey, 1993;Krebs and Khorana, 1993;Lanyi, 1993). After protonation of Asp 85 , a proton is released to the aqueous phase. A protein conformational shift observed while the Schiff base is unprotonated (Subramaniam et al., 1993) has been associated with the kinetically detected M 1 to M 2 reaction (Váró and Lanyi, 1995). It was suggested to change the access of the Schiff base away from Asp 85 and toward the initially protonated Asp 96 that lies near the cytoplasmic surface (Kataoka et al., 1994) and to lower the pK a of Asp 96 (Cao et al., 1993b;Brown et al., 1995a). Asp 96 thus becomes a proton donor and reprotonates the Schiff base. This is followed by reprotonation of Asp 96 from the cytoplasmic surface, reisomerization of the retinal to all-trans, and deprotonation of Asp 85 in the extracellular domain, thereby recovering the initial state. The intermediate states that arise and decay in this cycle have been named J, K, L, M, N, and O (Lozier et al., 1975).
The proton transfers inside the protein to and from the Schiff base trigger the proton transfers between groups near the surfaces and the aqueous phase. The pK a of the release group in the extracellular domain is lowered after protonation of Asp 85 (Zimá nyi et al., 1992), and the pK a of the proton uptake group in the cytoplasmic domain is raised after deprotonation of Asp 96 (Zimá nyi et al., 1993). In this way the extracellular and cytoplasmic peripheral domains facilitate the unidirectional passage of protons. The details of this coupling are not yet clear because both regions are structurally complex (Henderson et al., 1990) and contain several protonatable and hydrogen-bonding residues, as well as functionally important bound water (Maeda et al., 1992Brown et al., 1994;Fischer et al., 1994;Kandori et al., 1995;Yamazaki et al., 1995).
The relevant buried protonatable residues to the extracellular side of the Schiff base are Asp 85 , Asp 212 , and Arg 82 and possibly Tyr 57 and Tyr 185 . The first three and coordinated water constitute a diffuse counter-ion to the Schiff base (De Groot et al., 1989Dér et al., 1991). Interaction of the anionic Asp 85 with Arg 82 is indicated by the fact that the pK a of Asp 85 is increased by about 5 pH units in the R82Q and R82A mutants (Subramaniam et al., 1990;Thorgeirsson et al., 1991;Balashov et al., 1992;Brown et al., 1993), and interaction of Asp 85 with the Schiff base is inferred from the fact that the pK a of the Schiff base is decreased by about 4 pH units in the D85N mutant (Otto et al., 1990;Marti et al., 1992;Brown et al., 1993;Tittor et al., 1994;Kataoka et al., 1994). The participation of Asp 212 in these interactions is indicated by the observations that in D212N the Schiff base does not deprotonate after photoisomerization of the retinal near neutral pH (Otto et al., 1990;Needleman et al., 1991;Cao et al., 1993b), and in the D212N/ R82Q mutant the pK a of Asp 85 is not raised as in R82Q, but the pK a of the Schiff base appears to be several pH units higher than in the wild type (Brown et al., 1995b). Understanding the first half of the photocycle means understanding the nature of dipole interactions and proton transfers within this complex. There are some suggestive clues. A newly formed hydrogen bond between water and Asp 85 appears to play an essential role in destabilizing the Schiff base proton in the L intermediate . After transfer of this proton to Asp 85 in the L to M reaction, a proton is released to the extracellular surface. The released proton originates not from Asp 85 , because this residue remains protonated until the final reaction of the photocycle, but from an unknown group termed XH. The pK a of XH must be high in the unphotolyzed protein Balashov et al., 1993) but will have decreased at this time to about 6 (Zimá nyi et al., 1992). Above pH 6 XH releases its proton, but at more acid pH it remains protonated throughout, and a proton is released at the extracellular surface only at the end of the photocycle, presumably directly from Asp 85 as its pK a returns to its initial low value (about 2.5). Arg 82 and Tyr 57 have been proposed as candidates for XH, but the former would have to be in a very unusual environment for its pK a to become as low as 6 (Brown et al., 1993), and as for the latter, neither UV Raman (Ames et al., 1990) nor NMR McDermott et al., 1991) indicate the formation of a tyrosinate in the photocycle. In view of these problems, until now a water liganded to Arg 82 (Braiman et al., 1988) seemed to be the most likely source of the released proton.
In this report we identify Glu 204 , a residue located nearly at the surface end of the extracellular proton channel and about 15 Å from the Schiff base (Henderson et al., 1990), as part of the proton release mechanism. The evidence indicates that Glu 204 is protonated in BR but unprotonated in the M state and strongly suggests therefore that Glu 204 is the proton release group XH. This role for Glu 204 was recently predicted on the basis of molecular dynamics calculations (Scharnagl et al., 1995).

MATERIALS AND METHODS
The site-specific residue replacements E204Q and E204D were introduced into the bop gene and expressed in Halobacterium salinarium with a vector based on a halobacterial plasmid to be described elsewhere. The mutant R82Q was described before (Brown et al., 1993). The purple membranes containing these proteins were purified by a standard method (Oesterhelt and Stoeckenius, 1974).
Time-resolved spectroscopy was as described previously (Cao et al., 1993a(Cao et al., , 1993b. The experiments that measured proton release were done in 2 M NaCl at the pH indicated, without buffer, and with or without 50 M pyranine. When it occurred several milliseconds after photoexcitation or later, the absorption change of pyranine was assumed to approximate the kinetics of the interaction of the protons with the protein, because protons at the surface and in the bulk equilibrate in about 1 ms (Heberle and Dencher, 1992). All spectroscopy in the visible was at a regulated 22°C.
For the Fourier transform infrared (FTIR) 1 measurements, 40-l aliquots of the samples in 2.5 mM phosphate buffer, pH 7, were dried on a BaF 2 window, humidified with H 2 O or D 2 O before mounting into an Oxford cryostat, and illuminated at 274°K for 2 min with Ͼ500 nm light for light adaptation. The M minus BR spectra were measured at 230 K after 1 min of illumination with Ͼ500 nm light in a Bio-Rad FTIR spectrometer FTS60A/896 with 2 cm Ϫ1 resolution. The spectra shown are differences of the averages of four to six recordings, each calculated from 128 interferograms before and after illumination. All the spectra are scaled by normalizing to the amplitude of the 1169 cm Ϫ1 band.

Photocycle and Protonation/Deprotonation Kinetics of E204Q Bacteriorhodopsin-
The absorption maximum of the chromophore did not shift noticeably upon replacement of Glu 204 with glutamine (not shown); neither was the photocycle fundamentally changed. Some differences from wild type were noted, however. Fig. 1 shows difference spectra of E204Q measured at various delay times after photoexcitation with a nanosecond laser flash. The decay of the initial red-shifted product  (Zimá nyi et al., 1993). In addition to these traces, the net absorption of the pH indicator dye pyranine is included (Fig. 2, dashed line) to show the release of protons (from the extracellular surface) and their uptake (at the cytoplasmic surface) that occurs roughly during the decay of N. Although pyranine detects the protons only once they are in the bulk, the pH-dependent kinetics had argued (Zimá nyi et al., 1992) that when the protons are released to the membrane surface during the formation of M, it occurs during the M 1 to M 2 reaction.
As described earlier (Otto et al., 1990;Thorgeirsson et al., 1991;Cao et al., 1993a;Balashov et al., 1993), replacement of Arg 82 with glutamine or alanine causes distinct changes in the chromophore and proton kinetics. accumulates. Proton transport occurs in spite of the late release of the proton to the extracellular surface, apparently because the M 1 to M 2 reaction can proceed without deprotonation of the XH group, as in the wild type at pH Ͻ 6 (Zimá nyi et al., 1992). The effects of the R82Q mutation on the deprotonation of the Schiff base and the proton release have suggested that Arg 82 has a role in proton conduction between the Schiff base and the aqueous phase. The effect of Arg 82 on the accumulation of O was suggested to reflect a requirement for its positive charge for the early reisomerization of the retinal from 13-cis to all-trans in the N to O reaction (Otto et al., 1990;Cao et al., 1993a;Balashov et al., 1993). Fig. 3A shows single wavelength kinetics for the chromophore and for protons in E204Q at pH 7.3. This mutant resembles R82Q in that (a) the formation of M is more rapid than in wild type (by about 3ϫ), but more importantly in that (b) proton release is delayed until after proton uptake. Thus, as in R82Q, the proton release mechanism that utilizes deprotonation of XH is made inactive upon replacing Glu 204 . However, unlike in R82Q, the O intermediate accumulates in amounts even larger than in the wild type. The decay of M contains a second phase, larger than in wild type, with a time constant that roughly corresponds to the rise of the O state. The proton uptake coincides with the rise of the O state.
Protonation/Deprotonation Kinetics of E204D Bacteriorhodopsin-The photocycle of E204D contains a larger amount of the O state than the wild type, and in this respect it resembles that of E204Q. However, the decay of M and the rise and decay of the O state are much more rapid, as in the wild type. The rise of M is 1.5ϫ faster than in the wild type. Importantly, Fig. 3B shows that, unlike in E204Q, proton release is unaffected when Glu 204 is replaced with an aspartate. Proton release occurs before uptake, as in the wild type ( Fig. 2A).
FTIR Spectra of Photointermediates of Wild Type, E204Q, and E204D Bacteriorhodopsins-The results above demon-strate that like Arg 82 , residue 204 is a participant in the release of a proton to the extracellular surface. There are three alternatives: (a) Glu 204 influences XH but does not have a proton transfer role, (b) Glu 204 conducts the proton from XH to the aqueous medium, and (c) Glu 204 is in fact XH, the source of the released proton. The features of the CϭO stretch region for protonated carboxyls should reveal whether Glu 204 does not deprotonate or deprotonates only transiently and without accumulation of the carboxylate anion (alternatives a or b) or if Glu 204 deprotonates in M and remains unprotonated (alternative c). However, the CϭO stretch frequency for a glutamic acid is in general lower than that for aspartic acid, such as Asp 85 (Braiman et al., 1988) or Asp 96 (Bousché et al., 1991), and the intensity of the band is weaker; therefore the expected negative depletion band would be buried in the complex amide band region between 1650 and 1700 cm Ϫ1 . Indeed, an early FTIR study (Eisenstein et al., 1987) found no evidence that any of the glutamic acids participate in proton transport. On the other hand, the two mutants studied would help to decide this. The putative band from the deprotonation of Glu 204 would be absent in E204Q, and its frequency would be shifted in E204D to above 1700 cm Ϫ1 , a range where the amide bands do not interfere. Fig. 4 shows M minus BR spectra in the CϭO stretch region for the wild type protein, E204Q, and E204D. The spectrum of the wild type (Fig. 4A) contains the prominent positive band of the protonated Asp 85 at 1761 cm Ϫ1 , characteristic of the M state (Braiman et al., 1988;Fahmy et al., 1992). As in many earlier reported spectra under these conditions, there is no obvious negative band that would indicate the deprotonation of Glu 204 . However, the appearance of a negative band at 1714 cm Ϫ1 in E204D (Fig. 4C) raises the strong possibility that when residue 204 is an aspartic acid, it becomes deprotonated. If this is indeed the depletion band of the protonated Asp 204 , its frequency will be downshifted in D 2 O, similarly to the positive band of the protonated Asp 85 . The appearance of a negative band at 1703 cm Ϫ1 in D 2 O (Fig. 4C, dotted line) confirms this. The downshift of 11 cm Ϫ1 in D 2 O is similar to the 12 cm Ϫ1 downshift for Asp 85 in the same spectra. Fig. 4 shows also the differences between the spectra measured in H 2 O and D 2 O for E204Q and E204D. Recordings of two independent experiments are shown to illustrate the reproducibility of these difference spectra. In the E204D the H 2 O/D 2 O difference spectrum has the positive and negative features expected from the spectra in Fig. 4C. They are absent in E204Q (Fig. 4B). Importantly, the results with the wild type (Fig. 4A) show that there is a similar H 2 O/D 2 O difference feature at lower frequencies, where the original bands are obscured by the amide bands. The negative band is at 1700 cm Ϫ1 , and the positive band is at 1690 cm Ϫ1 . We suggest that the 1700 cm Ϫ1 band is the depletion band of the protonated Glu 204 . The small bands in the 1720 -1740 cm Ϫ1 region are ascribed to perturbation of Asp 96 and Asp 115 . The broad bands in the 1680 -1705 cm Ϫ1 region in E204Q and E204D are likely due to perturbation of peptide CϭO and arginine C-N (Braiman et al., 1994). These perturbations could be different in the different mutants. In E204Q Gln 204 might contribute also to these bands.
The amplitude of the CϭO band ascribed to Glu 204 is smaller than that of Asp 85 . Increasing the pH from 7 to 10, which would result in more complete deprotonation of Glu 204 in the photocycle but in less observed amplitudes if this pH is higher than the pK a of Glu 204 in the unphotolyzed state, did not change the amplitudes (not shown). Tilting the samples, which would reveal any orientational effects of the CϭO group in question, did not increase the amplitude relative to that of the Asp 85 band (not shown). Measuring the M minus BR spectrum in photostationary states at room temperature, which might result in more efficient proton release, did not increase the amplitude either (not shown). Thus, the low intensities might originate from some degree of delocalization of the proton of residue 204 within the extracellular hydrogen-bonded network. The greater intensity of the band in Asp 204 than that in Glu 204 might originate, likewise, from better localization of the proton on the aspartate than on the glutamate. However, we cannot exclude the possibility that the proton is localized fully on Glu 204 , but the intensity of the CϭO stretch band is lower than expected for some other (unknown) reason.
In the region of the O-H stretch of water, the negative band of the wild type at 3643 cm Ϫ1 is replaced in E204Q with a band shifted to a lower frequency, at 3638 cm Ϫ1 (Fig. 5). The water that changes its hydrogen bonding is therefore more strongly bound in this mutant, as in D212N . Replacing Glu 204 with aspartic acid preserved the higher frequency, although a second weaker depletion band of unknown origin at a lower frequency appeared also (Fig. 5C). DISCUSSION We report here that proton release at the extracellular surface of bacteriorhodopsin does not occur at its normal time in the photocycle when Glu 204 is replaced with a glutamine. The influence of this mutation is through removing the carboxyl group, because proton release is normal in E204D. Glu 204 is located near the extracellular surface (Henderson et al., 1990), and in a structural model from molecular dynamics calculations (Humphrey et al., 1994), it is connected by a chain of three hydrogen-bonded water molecules to Arg 82 in the cluster of residues that directly participate in the counter-ion to the Schiff base. In considering the possible role of Glu 204 in the proton release, it is instructive therefore to recall what is known about the hydrogen bonding of water in this region during the photocycle.
The properties of the water near Asp 85 , detected in the L and M states by FTIR, can be summarized as follows. In the wild type unphotolyzed protein it is weakly hydrogen-bonded (O-H stretch at 3643 cm Ϫ1 ) but becomes somewhat more strongly bonded (shift to 3636 -3638 cm Ϫ1 ) when either Asp 212  or Glu 204 (Fig. 5) is replaced. It becomes much more strongly hydrogen-bonded in the L intermediate of the wild type protein (shift to a diffuse region between 3460 and 3560 cm Ϫ1 ). These bands of the O-H stretch are eliminated when Asp 85 is replaced , suggesting that either Asp 85 participates in the increased hydrogen bonding or possibly this water is absent altogether in the D85N mutant. The extent of the shift is greatly reduced when Asp 212 is replaced with asparagine , suggesting that Asp 212 is an important participant in the stronger hydrogen bonding. Therefore, it seems likely that water is held in a hydrogenbonded network by Arg 82 , Asp 212 , Glu 204 , and Asp 85 , and this network in the unphotolyzed state is characterized by a balance of opposing hydrogen bonds. After photoexcitation the water becomes more strongly bound to Asp 212 and creates the conditions in the L state for the role of Asp 85 as proton acceptor to the Schiff base in the L to M reaction and for the subsequent release of a proton to the surface.
Arg 82 and Glu 204 have different functions from Asp 85 and Asp 212 in this network. When Asp 85 or Asp 212 is replaced, the complex in L becomes nonfunctional for proton transfer and the Schiff base remains protonated. Replacing Arg 82 or Glu 204 redistributes the hydrogen bonds of bound water to favor the remaining residues and accelerates rather than abolishes the proton transfer from Schiff base to Asp 85 . However, the release of a proton to the surface cannot occur at this time in the photocycle when Arg 82 or Glu 204 is replaced. This does not necessarily identify either of these residues as the group XH, the source of the proton, however. There has never been any direct evidence that the origin of the released proton is Arg 82 itself, and we had estimated (Brown et al., 1993) its pK a in the photocycle as much higher than that of XH. A normal proton release in the photocycle of D212N/R82Q (Brown et al., 1995b) argues that Arg 82 is not XH. On the other hand, we find not only that proton release depends on Glu 204 but that Glu 204 is protonated in BR and deprotonated in the M state. It appears therefore that Glu 204 is XH, and the proton to be released in the M state is either localized entirely on this carboxyl group or to some degree delocalized in a hydrogen-bonded network that contains also water.
Although much closer to the aqueous interface than the residues near the Schiff base (Henderson et al., 1990), Glu 204 can probably fulfill the requirements we had set earlier for XH (Zimá nyi et al., 1992). Because it is the source of the released proton, its pK a should be high in the unphotolyzed protein but decrease to about 6 in the photocycle. Molecular dynamics and electrostatic calculations have estimated the pK a of Glu 204 as 13 before photoexcitation but 6 -8 in the M state (Scharnagl et al., 1995). We therefore suggest the following sequence of events after photoisomerization of the retinal. First, the Schiff base proton is transferred to the anionic D85. Second, the hydrogen bond between Arg 82 and the now neutral Asp 85 is broken, and the water dipoles in the network redistribute so as to transfer some of the positive charge of Arg 82 toward Glu 204 . Although the location of Arg 82 in the model from electron diffraction is ambiguous (Henderson et al., 1990), in several structural models based on molecular dynamics (Humphrey et al., 1994;Scharnagl et al., 1995), Arg 82 does assume the position between Asp 85 and Glu 204 required for this. Alternatively, the side chain of Arg 82 could move from Asp 85 toward Glu 204 . Third, Glu 204 loses its proton to the surface and the aqueous phase. This sequence would account for the so far poorly understood events that result in proton release in the photocycle.