In the present study, we report on the structural features of the bacteriorhodopsin triple mutant E9Q/E194Q/E204Q (3Glu) of bacteriorhodopsin by combining experimental and molecular dynamics (MD) approaches. In 3Glu mutant, Glu9, Glu194 and Glu204 residues located at the extracellular side of the protein were mutated altogether to glutamines. UV-visible and differential scanning calorimetry experiments served as diagnostic tools for monitoring the resistance against thermal stress of the active site and the tertiary structures of the 3Glu. The analyses of the UV-visible thermal difference spectra demonstrate that the spectral forms at room temperature and the thermal unfolding path differ in the wild-type bacteriorhodopsin and the 3Glu. Even with these spectral differences, the thermal unfolding of the active site occurs at rather similar melting temperatures in both proteins. A noteworthy consequence of the mutations is the altered two-dimensional packing revealed by the lack of the pre-transition peak in differential scanning calorimetry traces of 3Glu mutant, as previously detected in wild-type and the corresponding single mutants. The infrared spectroscopy data agree with the loss of paracrystalinity, illustrating a substantial conversion of αII to αI helical conformation in the 3Glu mutant. Molecular dynamics simulations show higher dynamics flexibility of most of the extracellular regions of 3Glu, which may account for the somewhat lower tertiary structural stability of the mutated protein. Finally, hydrogen bond analysis reveals that the mutated Glu194 and Glu204 residues create ~ 50% less hydrogen bonds with water molecules compared to wild-type bacteriorhodopsin. These results exemplify the role of the water hydrogen-bonding network for structural integrity and conformational flexibility of bacteriorhodopsin.
Bacteriorhodopsin (bR) is a seven-helix transmembrane protein found in the halophilic archaea Halobacterium salinarum. The protein acts as a light-driven proton pump, converting light energy into an electrochemical gradient of protons, which is further used by the archaebacterium to energize cellular processes [1, 2]. The biological activity of bR as an ion pump is set off by the absorption of a photon, which causes a very rapid isomerization of the covalently bound retinal chromophore. This event initiates bacteriorhodopsin's photocycle, consisting of several spectroscopically distinguishable short-lived photostates with absorption bands from blue (410 nm) to red (640 nm) optical regions . During the photocycle, which takes ~ 10 ms to complete at room temperature, bR undergoes significant conformational changes that drive the transfer of one proton across the cell membrane.
Extensive work carried out to explore the structure–function relationship of bR has provided a remarkably complete view of the protein, including atomistic details of an interplay between key residues and water molecules . Point mutation studies of bR, have been aimed essentially at the identification of the residues implicated in the maintenance of the functionally active protein structure [5-8]. Furthermore, the engineering of bR mutants able to stabilize a particular photointermediate or, in other words, to freeze the protein intermediate conformations has opened new possibilities for an application of bR in data storage devices [9-12]. Several single and multiple mutants of the four negatively-charged glutamate side chains (Glu9, Glu74, Glu194 and Glu204) located in the extracellular (EC) region of bR have been engineered and extensively studied by several groups in a search for the proton release group. Initially, Glu204 and/or Glu194 were proposed as proton release residues [13, 14]. However, a number of subsequent spectroscopic and quantum mechanics studies have led to the suggestion that the proton release group actually is a complex formed by a shared protonated water cluster  or a shared proton between the Glu194/Glu204 pair . Currently, both residues (Glu194 and Glu204) are recognized as essential elements of the proton release machinery [17-19]. The question remains as to whether they share similar tasks throughout the activation of the protein. We have demonstrated that the second pKa increase of the Asp85 during the photocycle is inhibited in E194Q but not in E204Q . The replacement of Glu194 with Gln affects not only the photocycle, but also the thermal isomerization of the unphotolysed bR by stabilizing the all-trans configuration and restricting all-trans to 13-cis isomerization . These findings clearly emphasize the different functional roles of Glu194 and Glu204 residues.
During recent decades, apart from being a paradigm for studying other transduction membrane proteins, bR has also attracted attention as a possible material for biotechnological applications . bR shows a high yield of expression, high thermal and chemical stability, and easy charge separation. The intrinsic stability of bR is unusually high compared to other proteins found in archaebacteria, apart from proteins isolated from other extremophiles. Essentially, the capability of bR to complete a high number of photo-conversions with no loss of its photonic properties is far beyond the capacity of any synthetic material. These intrinsic properties have made this membrane protein an excellent candidate in the search for new strategies where biomaterials with attractive physical functions can be used as components in nanoscale devices .
In a recent study, we used the bR triple mutant E9Q/E194Q/E204Q (3Glu) to construct an excitonic solar cell and found that it transfers electrons from the redox electrolyte to the anode better than the wild-type does, and thus it can be successfully used as a photosensitizer . One of the specific requirements for using naturally occurring organic molecules in technological devices is their structural and functional stability over wide temperature and pH ranges . To improve the performance of photoactive nanostructured materials, a knowledge of molecular electrostatics and dynamics is essential  not only for the understanding of the transport mechanism, but also for the successful sensing and optimal binding of bR to nanofibres.
In the present study, we extend our study on the effect of the triple mutation on the stability of the active site and the tertiary structure of bR by using spectroscopic methods and differential scanning calorimetry (DSC). The ability of the 3Glu mutant to retain the protein in a proper active conformation was further tested by molecular dynamics (MD) simulations. We compared wild-type bR (WT) and the triple mutant, 3Glu, with the aim of understanding the driving forces for the structural integrity and the conformational dynamics of the protein.
Conformational variability and stability of the active site of the 3Glu mutant: thermal difference UV-visible spectroscopy
The retinal and the retinal binding pocket form the active site of bR. It is generally assumed that the absorbance maximum of the WT is defined by the protonation of the Schiff base (SB) and the interactions of retinal with the binding pocket. Therefore it reflects the conformation of retinal and the protein domain in the vicinity of the retinal moiety [22, 23]. When exposed to elevated temperatures, membrane proteins lose their functionality and denature. In the case of bR, the conformation of the active site is affected as well [22, 24]. To examine the effect of the substitution of the three Glu for Gln (Fig. 1) on the destabilization of the active site, we performed UV-visible absorption studies at various temperatures. Thermal difference spectra (TDS) of WT and 3Glu are presented in Fig. 2. TDS allow visualization of the formation and the disappearance (decay) of thermally-induced spectral forms and also help to separate partial spectral transitions, which overlap.
At pH 7, on an increase in temperature up to 60 °C, the room temperature (RT) absorption maximum of the WT at 559 nm exhibits gradual blue shifts to 530 nm, accompanied by an intensity decrease (Fig. S1, top left). Above 85 °C, the absorption maximum of WT is shifted to 370 nm as a result of hydrolysis of the SB linkage and the release of retinal from the protein. These changes are irreversible upon cooling of the sample [22, 25]. The corresponding TDS spectra of WT (Fig. 2, top left) display one negative band, with the maximum shifted from 571 nm (25 °C) to 562 nm (85 °C), and two positive bands at 460 and 370 nm, respectively. The amplitude of the absorption changes at 460 nm, caused mainly by the transformation of bRRT into species absorbing at shorter wavelengths, the so-called bR ‘ red’ form [26-28], increases with a raising of the temperature. Moreover, the isosbestic point at 510 nm for TDS in the range 25–80 °C suggests a two-state thermal transition from purple to ‘red’ form. However, at temperatures above 85 °C, the band at 370 nm dominates over the bandat 460 nm because the latter band begins gradually to decay.
At pH 7, the absorption maximum of 3Glu at 530 nm (Fig. S1, top right) is blue shifted compared to WT, implying some conformational alteration of the active site of the protein. An increase in temperature causes a further blue shift of the maximum, accompanied by a decrease of its intensity. TDS of the mutant (Fig. 2, top right) display one negative band and two positive bands, at 454 and 374 nm, with two cross-over points at 503 and 374 nm, respectively. Compared to WT, the amplitude of the thermally-induced ‘red’ spectral form at 454 nm is lower, most likely as a result of its presence at RT (for more details, see the blue shifted maximum of 3Glu and the text below). The thermal denaturation of 3Glu, followed by the band at 370 nm, starts after the decay of the ‘red’ form. This behavior is different in the WT, in which the spectral transitions from bRRT to the ‘red’ form and to retinal release essentially proceed simultaneously.
At pH 4, the absorption maximum of the WT at 560 nm decreases in amplitude as the temperatures increases, concomitantly with a shift to 590 nm. TDS of WT at pH 4 (Fig. 2, middle left) show that the amplitude of the positive band at 631 nm increases up to 55 °C because, at more elevated temperatures, it starts to decline, generating the positive band at 370 nm. The temperature-induced spectral transitions at pH 4 proceed through two isosbestic points as the cross-over at 582 nm defines the two-state transition from the ground state spectral form (bRRT at 560 nm) to the ‘blue’ spectral form . The second isosbestic point at 456 nm suggests that the thermal unfolding of the protein starts from the ‘blue’ spectral transient form.
At pH 4, the absorption maximum of 3Glu is at 575 nm, and is red shifted in comparison with that of pH 7, implying the presence of the ‘blue’ species in the mutant at this pH. Moreover, in contrast to WT spectra at pH 4, the absorption maximum of 3Glu undergoes a slight blue shift and a decrease in the intensity on an increase in temperature. TDS of 3Glu at pH 4 (Fig. 2, middle right) display a negative band with maxima changing from 606 mm to 579 nm, a strong positive band at 379 nm and a small positive band at ~ 500 nm. The isosbestic point at 535 nm of TDS in the range 25–45 °C implies a two-state transition from ground state spectral form (3GluRT) to the ‘red’ (at ~ 480 nm). Above 55 °C, the decay of the positive band at 480 nm is concurrent with the intensity increase of the band at 370 nm. The second isosbestic point at 442 nm of TDS at temperatures above 55 °C most likely reflects the temperature-dependent equilibrium between the spectral form (579 nm) and retinal release (band at 372 nm). Moreover, in contrast to WT, the increase of temperature does not cause the formation of the ‘blue’ form in the 3Glu mutant (Fig. 2, middle right). This is expected because, as reported previously, in 150 mm KCl the pKa of Asp85 is 3.8 and 2.7 for 3Glu and WT, respectively . Thus, at pH 4, approximately half the amount of the 3Glu mutant should be in the ‘blue’ form, whereas, in the WT, this form is only marginally formed at this pH.
At alkaline pH, thermally-induced spectral transitions of WT and 3Glu are similar to those identified at pH 7 (for further details, see Fig. 2, top and bottom). However, at alkaline pH, both proteins are more prone to unspecific aggregation, especially at elevated temperatures, where a gradual loss of absorbance was observed. At pH 9.5 and at RT, the absorption maximum of WT is at 555 nm, blue shifted compared to that at pH 7 (Fig. S1). On an increase in temperature, it undergoes intensity losses and a further blue shift to 532 nm. Above 85 °C, the absorption maximum already is shifted to 370 nm, corresponding to free retinal. TDS of WT (Fig. 2, bottom left) show a negative band with a maximum ranging between 575 and 560 nm and two positive bands at 460 and 372 nm. The cross-over point at 511 nm of TDS in the range 30–75 °C reflects the transition from the ground state spectral form (bRRT) to the ‘red’ spectral form. Moreover, above 55 °C, the intensity of the band at 460 nm drops, whereas that at 372 nm rises.
Unexpectedly for the WT, at pH 9.5, the absorption maximum of 3GluRT is at 527 nm, which is very similar to that at pH 7 (Fig. S1, spectrum of 3Glu, pH 7). With an increase in temperature up to 70°C, it blue shifts to 499 nm but essentially keeps its magnitude. Moreover, up to this temperature, TDS (Fig. 2, bottom right) cross-over at 514 nm, indicating a two-state temperature transition from the ground (3GluRT) (negative band at 561 nm) to the ‘red’ spectral form (positive band at 460 nm). Finally, above 80 °C, the ‘red’ form initiates the release of retinal (i.e. at the band at 372 nm).
From data analysis, the overall spectral features of TDS of WT and 3Glu appear to be quite similar. Basically, they consist of one negative band, as a result of the disappearance of the ground state (bRRT) spectral form, and two positive bands, at ~ 372 and 460 nm, as a result of deprotonation of the SB and retinal release from the protein, and also the formation of the ‘red’ form, respectively. These spectral features illustrate the perturbations of the protein's active site as a result of temperature stress, leading to the formation of the transient thermally-induced spectral form and, consequently, in the thermal unfolding of the protein in both WT and 3Glu. However, the shape of the difference spectra and the position and amplitudes of the bands differ within both proteins and with pH. For example, at pH 4, the positive band at 460 nm is missing in the TDS spectra of WT but a positive band at 630 nm is resolved instead. In 3Glu, the thermally-induced ‘red’ form is present to a very low extent, generated from the ‘blue’ spectral form (Fig. 2, right middle; band at 606 nm). Moreover, at this pH, the two-state unfolding proceeds via different isosbestic points in both proteins. The isosbestic points at 456 and 442 nm for WT and 3Glu, respectively (Fig. 2, middle) reflect the same concentration of ‘red’ and protonated Schiff base  species during the complete unfolding of WT and 3Glu, respectively. These observations imply that the spectral forms comprising the unfolding path of the active site are different in both proteins.
Based on these assessments, we propose plausible unfolding path schemes upon an increase in temperature:
Furthermore, to follow the resistance against thermal structural destabilization and unfolding of the active site of WT and 3Glu, we mapped the absorption changes at two wavelengths (560 and 370 nm) representative of the disappearance of the ground state spectral form and the hydrolysis of the SB linkage and retinal release, respectively. The unfolding temperatures or, more accurately, the temperatures of the mid-transition (Tm) for these spectral transitions, as calculated by plotting the absorption changes at these two wavelengths as a function of the temperature (Fig. S2), are presented in Table 1.
Table 1. Thermal denaturation temperatures (Tm) of WT and 3Glu at three pH values, measured at different wavelengths.
Tm (°C) at 370 nm
Tm (°C) 560 nm
Tm (°C) 460 nm
Tm (°C) at 370 nm
Tm (°C) at 560 nm
Tm (°C) at 460 nm
Tm values correspond to the midpoint of the transition between the native and denatured states.
The formation of the ‘red’ spectral form in WT and 3Glu
At room temperature and pH 7, the most noticeable spectral difference between WT and 3Glu relates to the blue shifted maximum of the latter, implying the presence of ‘red’ species in the ground state of the mutant (Fig. S1). Indeed, the stability of the WT at 560 nm over a wide range of pH and temperature as monitored by its absorption maximum is a well known phenomenon. However, bR can change its purple color (λmax = 560 nm) to blue (λmax = 603 nm) or red (λmax = 460 nm) upon the formation of spectral forms known as blue and red, respectively. The blue spectral form is a result of the protonation of Asp85 in unphotolysed bR and can be obtained at acidic pH . The molecular nature of the red form remains under debate; however, it has been confirmed experimentally that its formation in WT can be induced by different factors, such as alkaline pH, heating, mutations or incubation with different organic substances [8, 30, 32, 33].
At alkaline pH, the absorption maximum of WT is blue shifted, implying the formation of some ‘red’ species upon alkalization of the medium, whereas the absorption maximum of 3Glu is approximately the same as at pH 7 (Fig. S1). These findings suggest that the ‘red’ species are thermally induced in WT, whereas, in 3Glu, they coexist with the purple form at RT (Fig. 2). Plots of the absorption changes at 460 nm versus temperature improve visualization of the formation and decay of this transient spectral form in the two proteins (Fig. 3). A comparison of the plots of 3Glu at neutral and alkaline pH shows a comparable subtle increase of ‘red’ component on an increase in temperature but it decays with a different Tm (Table 1). These data further validate the presence of ‘red’ species in 3Glu mutant at RT (ground state). Moreover, the WT plots clearly show the increase of the ‘red’ form with temperature, although with two-fold magnitude at neutral pH compared to alkaline pH (Fig. 3 and Table 1). These data are in close agreement with a purple to red transition upon an increase in temperature and alkaline pH . Furthermore, the assessment of the absorption changes at 460 nm shows that, compared to WT, the amount of ‘red’ form increases with temperature only by 20% in the mutant (Fig. 3). Thus, considering that the absorption maximum of WT (560 nm) at pH 7 and RT represents the protein basically in purple form, it can be estimated that ~ 80% of the species constituting the absorption maximum of 3Glu (at 527 nm) at pH 7 and RT are ‘red’ species.
The formation and decay of the thermally-induced red spectral form can also be followed by plotting the absorbance at 560 nm as a function of temperature (Fig. S2). Two transitions for WT unfolding, giving two Tm values, highlight another difference between both proteins (Table 1). Furthermore, the first (Tm1) associates closely with the Tm of ‘red’ formation, whereas the second one, (Tm2) associates with the Tm of ‘red’ decay and the Tm of retinal release from the protein (Table 1). Except for the distinct thermal stability of the transient spectral forms, followed by the disappearance of ground state spectral form, the transition from the native state to the unfolded state monitored by means of the retinal release occurs at similar Tm for WT and 3Glu (Tables 1 and 2).
Table 2. Comparison of Tm values of WT and 3Glu obtained using UV-visible spectroscopy or DSC.
WT Tm (°C)
3Glu mutant Tm (°C)
Tm of unfolding of the active site is calculated from the temperature absorption changes monitored at 370 nm.
To examine the similarities and differences between the secondary structures of ‘red’ spectral forms of 3Glu and WT, we compared their IR spectra. Given that the ‘red’ form in WT is the most accumulated at 80 °C and at alkaline pH (Fig. S2), the IR spectra of WT were recorded under these conditions (Fig. S3). In accordance with previous observations [35-37], the increase of the temperature up to 80 °C, just above DSC pre-melting temperature, causes a blue shift and a widening of the amide I band in the IR spectra of WT [38, 39]. On the other hand, the IR spectrum of 3Glu at room temperature looks almost the same as the WT spectrum at 80 °C (Fig. S3, top). The most notable features of the deconvoluted spectra of WT is the downshift of the main band at 1666 cm−1, in a favor of a band at 1658 cm−1 at 80 °C (red form conformation of WT). Similarly to WT at 80 °C, the deconvoluted spectrum of 3Glu at RT shows the main amide band at 1660 cm−1. (Fig. S3, middle and bottom). Previously, the band at 1665 cm−1 in the IR of WT was assigned to the αII helical conformation, whereas the band at 1660 cm−1 was assigned to the αI helical conformation . It is assumed that both helical conformations share the same parameters, although they differ in the dihedral angles and hydrogen-bonding scheme . These observations imply that the major difference between the secondary structures of WT and 3Glu corresponds to a change in α-helical conformation.
To probe the thermal stability of tertiary structure of WT and the 3Glu mutant, we carried out DSC experiments at three different pH, similar to the UV-visible experiments (Table 2). It is well established that the thermal unfolding of the WT goes through two temperature-dependent transitions. The first one is reversible and occurs at ~ 78 °C, also known as a pre-melting transition or pre-transition, followed by the main and irreversible transition at ~ 97 °C . As monitored by DSC traces (Fig. S4), the most noteworthy structural consequence of the substitution of the three EC Glu residues for Gln is the lack of the pre-transition peak in the mutant, which is seen in WT and the single mutants E9Q, E204Q and E194Q . The pre-melting transition has been associated with gel-to-liquid transition of the hexagonal BR lattice and is assigned to reversible disorganization of the two-dimensional paracrystalline array of bR molecules [42, 43]. In accordance with previous studies, the main irreversible transition temperature (Tm) of the WT is pH-dependent . Furthermore, Tm values for WT at acid and alkaline pH (Table 2) are in good agreement with previously reported data [36, 45]. In comparison with WT, the main transitions of unfolding of 3Glu are not downshifted by very much, between 6 and 12 °C, depending on pH , implying destabilization of the tertiary structure of the mutated protein to some extent (Table 2). Moreover, in both WT and 3Glu, the transitions from native to unfolded state of the active site occur at lower temperatures than those of tertiary structure (Table 2).
Structural and vibrational analysis of monomeric bR
To obtain information about the structural differences between WT and 3Glu monomers and their fluctuations, we applied molecular mechanical calculations, using the CHARMM force field. After minimization, the differences between WT and 3Glu are only 0.2 Å over all nonhydrogen atoms and 0.1 Å over backbone atoms, suggesting that there are no large-scale structural changes in this simulated monomeric form (Fig. 4). The magnitudes of atoms fluctuations along the second lowest frequency normal modes of WT and 3Glu range from 0.1 to 0.5 Å. Moreover, the differences in the fluctuations along these modes demonstrate that some residues, such as Val199, Pro200, Gln204 and Phe208, show ~ 50% greater conformational flexibility in the mutant than in WT. The whole loop between helix F and helix G (holding Glu194), the nearby residue Glu204 and the Glu9 environment appear to be affected most (Fig. 4).
MD simulations of oligomeric bR in the native membrane
The system used for MD consisted of nine bR molecules arranged in a hexagonal lattice of bR trimers. The calculated violin plots (Fig. 5) represent the symmetrical density distribution of rmsd and illustrate the most notable differences between WT and 3Glu. Based on rmsd values, WT appears to have slightly more stable structure than 3Glu (0.10 nm versus 0.12 nm), although there is a smaller population of less stable WT structures (a second maximum at 0.15 nm), which is less stable than most of 3Glu.
To examine the mobility of the individual residues, the root mean square fluctuations (rmsf) of 3Glu and WT proteins as well as the 3Glu/WT rmsf ratio were calculated (Fig. 6). Unlike variance differences, the variance ratio distributions (F-distributions) are well characterized and can be transformed to an approximate normal distribution by taking the natural logarithm. This procedure allows for a precise presentation of differences in thermal movement. Based on those data, the structure of 3Glu is less stable than WT in area spanning from helix B to the center of helix E, including the β-sheet loop. The second area of lesser stability of 3Glu is from end of helix F to the center of helix G. The logarithm values of rmsf ratio of 3Glu/WT were mapped onto the bR structure (Fig. 7). There are two clearly separated regions of different stabilities: almost all of the extracellular part of 3Glu, including the β-sheet loop, is less stable than WT, whereas the intracellular parts of helices A, E, F and G are more stable. Only one helix, helix B, is less stable in 3Glu along its whole length.
Analysis of differences in the number of hydrogen bonds
A summary of the hydrogen bond analysis is given in Fig. 8, which shows the box plots for the distributions of the number of hydrogen bonds between residues Glu194 and Glu204 to the protein (other residues of BR), β-sheets (residues 67–78) and the solvent, respectively. For WT, the median number of hydrogen bonds in nonamer between residues 194 and 204 and protein or solvent are 32 and 29, respectively, whereas, for 3Glu, they are 27 and 15, respectively (Fig. 8). These data imply that the mutation of the three Glu residues to Gln results in a significant decrease in the number of hydrogen bonds with the solvent (~ 50%) but, at the same time, the number of hydrogen bonds within the protein is only affected slightly. The hydrogen bonds to the β-sheet-containing fragment are absent for WT but present in the 3Glu mutant. In most cases, the amide nitrogen atom is a donor, whereas oxygen atoms from residues 76–78 (the end of the β-sheet) serve as acceptors.
In the present study, we combined calorimetric and spectroscopic methodologies with MD simulations and modeling approaches to obtain insight into the conformational dynamics and the thermal stability of genetically engineered 3Glu mutant of bR. UV-visible TDS and DSC served as diagnostic tools for monitoring the resistance of the active site and the tertiary structure of the 3Glu against thermal stress.
The TDS, which follows the perturbation of the protein's active site by an increase in temperature, illustrate similar consecutive steps of the unfolding path in both WT and 3Glu proteins: first, a loss of the ground state spectral form, followed by an accumulation of transient thermally-induced spectral form and/or a breakage of the SB linkage with the retinal and, finally, a release of retinal from the protein, in accordance with previous studies [35-37]. On the other hand, the TDS analyses clearly demonstrate that the spectral forms, comprising the basic RT state and unfolding path, differ in both proteins. For example, ‘red’ species comprise ~ 80% of the 3GluRT, whereas the WT ‘red form’ is basically thermally induced (Fig. 3). Even with these spectral differences, the thermal unfolding of the active site occurs at rather similar melting temperatures (Tm) for both proteins (Tables 1 and 2). Structurally, the most notable consequence of the mutation of the three EC Glu residues for Gln is the lack of the pre-transition peak in DSC traces of 3Glu, which is seen in WT and corresponding single mutants [29, 30], implying disorganization of the two-dimensional paracrystalline arrangement of 3Glu trimers. Considering that the three Glu residues are on helices A, F and G near to the EC surface, their mutation may affect the intermolecular lipid–protein and protein–protein interactions responsible for the two-dimensional paracrystalline arrangement (Fig. 1) . This is in line with MD simulations showing less stability for the EC parts of these helices in the mutant. High-resolution structures of WT have revealed that bR trimers are stabilized by interactions between helices B and D and, to a lesser extent, between helices A and F of adjacent bR molecules forming the trimer . Therefore, the participation of helices B and E in higher flexible EC regions of 3Glu, as revealed by MD simulations, holds for the destabilization of the trimers in 3Glu. Further support for a disorganization of the paracrystalline arrays of the 3Glu trimers comes from IR data showing that the amide I band shifted at 1660 cm−1 (Fig. S3). The amide I band of the WT found at higher frequency (1665 cm−1) has been assigned to the αII helical conformation, and it is considered to be indicative for the organization of bR monomers into trimers and the paracrystalline arrangement. In accordance with previous studies, the αII to αI transition in WT occurs at a temperature just above the reversible DSC peak (Fig. S3) , whereas the deconvoluted IR spectrum of the 3Glu mutant presents as a dominant αI conformation at room temperature (Fig. S3). The conversion of the band at 1665 cm−1 into the band at 1655 cm−1 or 1653 cm−1 was found for bR, being in monomer form after its solubilization with SDS . It was suggested that the band at 1665 cm−1 is not present in IR spectra if bR is a monomer but is present if the bR monomers are organized in trimers, although not obligatorily being hexagonally packed. Furthermore, it was reported that the trimer to monomer transition leads to a downshift of 10 cm−1 [49, 50] and it was concluded that the interactions between monomers concomitant with the formation of the trimers lead to substantial conversion of αI to αII helices. Therefore, the presence of αII helical band in the deconvoluted IR spectrum of the purple form of WT (at room temperature) is an indication that bR molecules are arranged in highly interacting trimers, which is apparently not the case for the red samples of 3Glu at room temperature or for WT at 80 ºC (Fig. S3). On the other hand, a smaller downshift of the amide I band of 3Glu (of ~ 5 cm−1) compared to the downshift (of ~ 10 cm−1) for WT upon trimer to monomer transition  strongly implies that the mutations most likely result in disorganization of the trimer array, rather than the formation of monomers. As monitored by DSC, the thermal unfolding of the 3Glu occurs, depending on the pH, at temperatures that are lower only by 6–12 °C compared to WT (Table 2).
Thermal denaturation of proteins occurs by incrementally breaking the noncovalent hydrogen bonds and salt bridges in order of their relative strengths. Thus, the lower Tm of 3Glu can account for the different structural stability of the ground state spectral forms constituting both proteins: the purple spectral form for WT versus blue and the red form at acid and at neutral and alkaline pH for 3Glu, respectively. Furthermore, both proteins appear to be more resistant to unfolding at alkaline rather than acidic pH (Table 2). These results can account for the large structural conformational changes induced upon the purple to blue transition and no significant conformational changes upon alkalization of the media, as reported for the crystal structures of WT at acidic and alkaline pH .
The DSC data also suggest a less compact arrangement compared to WT with a decreased activation barrier (ΔH = 395 kcal·mol−1 for WT versus 205 kcal·mol−1 for 3Glu). MD simulations of the 3Glu nonamer indicate that almost the entire extracellular part of 3Glu is less stable than WT (Fig. 7). The greater flexibility of the extracellular part, especially in areas containing the mutated residues, including the β-sheet loop, accounts for a lower temperature of unfolding for the tertiary structure of 3Glu. These observations are in agreement with previous NMR data reporting the disappearance of some interhelical interactions and the appearance of extra space in 3Glu  Previously, altered interactions between 3Glu trimers, most likely comprising helix F and adjacent loops, have been revealed by single-molecule dynamic force spectroscopy .
The formation of new spectral forms, comprising the ground state of the 3Glu mutant at different pH, indicates that the mutation strongly affects the conformation of the protein segment responsible for color changes. Because the three mutated residues are far away from the active site, they are expected to exert a small (or no) effect on the conformation of the protein in the vicinity of retinal. Indeed, as we have shown previously, the single mutants of the residues comprising the triple mutant, show properties similar to the WT BR . However, simultaneous mutation of the three EC residues results in a significantly blue shifted absorption maximum as a result of conversion of 80% of the purple form to the red spectral form with an altered conformation of the retinal binding site. The most reasonable scenario for the long-range effect of the triple mutation is that it is transmitted through the hydrogen-bonded water network, expanding through the whole protein, and is affected by the mutation. This view is supported by hydrogen bond analysis revealing that the mutated Glu194 and Glu204 residues create a lower number of hydrogen bonds with water molecules, thus significantly disorganizing the water hydrogen-bonding network in EC side of 3Glu mutant (Fig. 7). This disorganization drastically affects the extracellular region, leading to late proton release and a slower photocycle of the mutant . Apparently these changes also propagate within the whole proton transfer pathway inside the mutated protein, affecting Asp85  and the conformation of the active site, responsible for the color. Such a long-range effect is not unexpected and similar effects have been observed and reported before upon the mutation of extracellular residues of bR [19, 54]. In an attempt to understand the molecular nature of the ‘red’ form, the question arises as to whether the treatments inducing the red form (e.g. anesthetics and diethyl ether) and the heat share some common properties. Previous studies on the ‘red’ form [55, 56] indicate that the common property shared by the compounds causing the appearance of the red form is that they are better hydrogen bond acceptors than water. Therefore, beyond a certain concentration, they break the water-bonded network. We consider that the hydrogen-bonding breaking activity of the heat of some mutations or of the indicated substances is the most likely reason for the formation of this spectral form. This scenario is strongly supported by the MD simulations of the 3Glu, which show disorganization of the water hydrogen-bonding network. Therefore, the results reported in the present study highlight the role of the water hydrogen-bonding network for both conformational flexibility and dynamics of the bR. In particular, they focus attention on a dominant contribution of the network comprising water molecules and functional residues not only for proper functioning of bR as a proton pump, but also for its structural integrity.
As a final point, it is worth noting that the 3Glu mutant shows an interesting property with respect to the search for new biomaterials with technological applications: it has the ability to slow down the photocycle kinetics and to stabilize some intermediates [18, 20, 30]. Because 3Glu retains a high degree of thermal stability of the active site and tertiary structure despite the introduced mutations, it may be a good candidate for technological applications.
Materials and methods
Construction of the E9Q/E194Q/E204Q (3Glu) mutant
Construction of the 3Glu mutant was accomplished by cloning of the single mutants together, taking advantage of unique restriction sites. After screening of the mutant by DNA sequencing, it was transformed and expressed in the Halobacterium salinarum L33 strain . The purple membranes were grown and purified in accordance with standard procedures, as described previously .
UV-visible absorption spectroscopy
UV-visible absorption spectra were recorded on a Varian Cary3 Bio spectrophotometer, (Varian Inc., Palo Alto, CA, USA) supplied with integrating sphere to avoid light scattering artifacts. Thermal studies were carried out using a temperature-controlled cuvette holder cell, as the samples were allowed to equilibrate for 10 min at each temperature, before recording the spectrum. The spectra were taken at increments of 5 °C over a temperature range from 25 to 100 °C, in the visible and the UV spectral range. The experiments were performed at three different pH values: pH 4.0 for the purple to blue transition, pH 9.5 for the purple to red transition and at neutral pH . The proteins (at ~ 15 μm) were incubated in three different buffers at pH 4.0, 7.0 and 9.5 and in presence of 150 mm KCl. Buffer signals were subtracted from the spectra of the proteins.
DSC experiments were performed using a MicroCal MC2 instrument (MicroCal, Inc., Northampton, MA, USA). Before the DSC scans, the samples were dialyzed against 3 mm appropriate buffer for acid and alkaline pH, giving a final protein concentration of 1.5 mg·mL−1. Experiments were performed at a scanning rate of 1.5 K·min−1 and under a nitrogen pressure of 1.7 atm to avoid sample evaporation at higher temperatures. For each sample, three consecutive thermograms were recorded. The first of these provides the heat released or taken up by the protein upon an increase in temperature. The second and third thermograms were recorded after cooling the sample back to the room temperature. Two corrections were applied to the first thermogram: (a) subtraction of the second thermogram, which was used as a blank, and (b) subtraction of the chemical baseline using the method of Takahashi and Sturtevant .
Building molecular models of bR
3The initial structural coordinates of bR were taken from the Protein Data Bank (PDB) file 1C3W , corresponding to the X ray crystal structure of the WT protein in the ground state, at 1.55 Å resolution. The missing coordinates for the residues 157–161 were built by a superposition with another crystal structure of bR (PDB code: 1QM8), which was resolved at 100 K to 2.5 Å . Nine water molecules were considered to be buried into protein and were included in the calculations. 3Glu was constructed by introducing the three mutated residues: Gln9, Gln194 and Gln204. All lysines, Asp96, Asp115 and Glu194 (in the case of WT), were protonated. Protonation of Glu194 was carried out to mimic the presence of a proton near the proton release group comprising of Glu194 and Glu204.
The software pdb2pqr  was used for assigning the atomic radii and the partial charges, at pH 7.0, consistent with the CHARMM force field  for the WT and the 3Glu monomers. The parameters for the Lys216 residue were built by combining the CHARMM parameters of lysine with the parameters for the retinal derived previously [63, 64].
MD of bR in the purple membrane
All the simulations of bR nonamer and most of the analyses were performed using gromacs, version 4.5.6 MD package . Two simulations were performed, each lasting 150 ns, and data for plots were collected from the last 100 ns. For visualization of trajectories, vmd  was used. The structure of bR with mapped rmsf values was prepared in pymol . The rmsd and rmsf were calculated using appropriate tools from gromacs . The rmsd was determined against the average structure obtained from the simulations of the wild-type protein, fitting on the helical regions. A detailed description of the models and parameters used is included in the Supporting information (Doc. S1).
Modeling of 3Glu bR was partially carried out at the Barcelona Supercomputer Centre. VR acknowledges support received from NIH, NSF, Wallace H Coulter Foundation, Rothschild Foundation and Harvard Medical School. VR would like to thank the National University of Singapore President and Professor Shih Choon Fong's office special program fund. Research in the EP laboratory was supported by the Ministerio de Ciencia e Innovación and FEDER (Fondo Europeo de Desarrollo Regional) grant BFU2012-40137-C02-01. The authors would like to thank Neus Ontiveros and Elodia Serrano for their skillful technical help, Dr Chandra S. Verma (Bioinformatics Instititute, A Star, Singapore) for electrostatic potential calculations and Dr Peter Fojan (Aalborg University, Aalborg, Denmark) for CD experiments. VR initiated the CD studies when at Aalborg University as a visiting Professor in 2006. VR dedicates the present paper to his father, Varun.