Spectroscopic and Photochemical Characterization of a Deep Ocean Proteorhodopsin*

A second group of proteorhodopsin-encoding genes (blue-absorbing proteorhodopsin, BPR) differing by 20–30% in predicted primary structure from the first-discovered green-absorbing (GPR) group has been detected in picoplankton from Hawaiian deep sea water. Here we compare BPR and GPR absorption spectra, photochemical reactions, and proton transport activity. The photochemical reaction cycle of Hawaiian deep ocean BPR in cells is 10-fold slower than that of GPR with very low accumulation of a deprotonated Schiff base intermediate in cells and exhibits mechanistic differences, some of which are due to its glutamine residue rather than leucine at position 105. In contrast to GPR and other characterized microbial rhodopsins, spectral titrations of BPR indicate that a second titratable group, in addition to the retinylidene Schiff base counterion Asp-97, modulates the absorption spectrum near neutral pH. Mutant analysis confirms that Asp-97 and Glu-108 are proton acceptor and proton donor, respectively, in retinylidene Schiff base proton transfer reactions during the BPR photocycle as previously shown for GPR, but BPR contains an alternative acceptor evident in its D97N mutant, possibly the same as the second titratable group modulating the absorption spectrum. BPR, similar to GPR, carries out outward light-driven proton transport in Escherichia coli vesicles but with a reduced translocation rate attributable to its slower photocycle. In energized E. coli cells at physiological pH, the net effect of BPR photocycling is to generate proton currents dominated by a triggered proton influx, rather than efflux as observed with GPR-containing cells. Reversal of the proton current with the K+-ionophore valinomycin supports that the influx is because of voltagegated channels in the E. coli cell membrane. These observations demonstrate diversity in photochemistry and mechanism among proteorhodopsins. Calculations of photon fluence rates at different ocean depths show that the difference in photocycle rates between GPR and BPR as well as their different absorption maxima may be explained as an adaptation to the different light intensities available in their respective marine environments. Finally, the results raise the possibility of regulatory (i.e. sensory) rather than energy harvesting functions of some members of the proteorhodopsin family.

A second group of proteorhodopsin-encoding genes (blue-absorbing proteorhodopsin, BPR) differing by 20 -30% in predicted primary structure from the first-discovered green-absorbing (GPR) group has been detected in picoplankton from Hawaiian deep sea water. Here we compare BPR and GPR absorption spectra, photochemical reactions, and proton transport activity. The photochemical reaction cycle of Hawaiian deep ocean BPR in cells is 10-fold slower than that of GPR with very low accumulation of a deprotonated Schiff base intermediate in cells and exhibits mechanistic differences, some of which are due to its glutamine residue rather than leucine at position 105. In contrast to GPR and other characterized microbial rhodopsins, spectral titrations of BPR indicate that a second titratable group, in addition to the retinylidene Schiff base counterion Asp-97, modulates the absorption spectrum near neutral pH. Mutant analysis confirms that Asp-97 and Glu-108 are proton acceptor and proton donor, respectively, in retinylidene Schiff base proton transfer reactions during the BPR photocycle as previously shown for GPR, but BPR contains an alternative acceptor evident in its D97N mutant, possibly the same as the second titratable group modulating the absorption spectrum. BPR, similar to GPR, carries out outward light-driven proton transport in Escherichia coli vesicles but with a reduced translocation rate attributable to its slower photocycle. In energized E. coli cells at physiological pH, the net effect of BPR photocycling is to generate proton currents dominated by a triggered proton influx, rather than efflux as observed with GPR-containing cells. Reversal of the proton current with the K ؉ -ionophore valinomycin supports that the influx is because of voltagegated channels in the E. coli cell membrane. These observations demonstrate diversity in photochemistry and mechanism among proteorhodopsins. Calculations of photon fluence rates at different ocean depths show that the difference in photocycle rates between GPR and BPR as well as their different absorption maxima may be explained as an adaptation to the different light intensities available in their respective marine environments. Finally, the results raise the possibility of regulatory (i.e. sensory) rather than energy harvesting functions of some members of the proteorhodopsin family.
Microbial rhodopsins, seven-helix integral membrane proteins that use retinal as a chromophore, comprise a large family of photoactive proteins present in archaeal, bacterial, and eukaryotic microorganisms (1-3). Among the most abundant and widely distributed are the proteorhodopsins, the first of which was identified in the ubiquitous uncultured SAR86 group of marine proteobacteria in plankton from Monterey Bay coastal waters (4). Nearly identical genes detected in picoplankton from the Southern Ocean (1), Mediterranean, and Red Sea (5) define a phylogenetic group of Monterey Bay-type pigment genes (clade I). The Monterey Bay pigment, a green light-absorbing proteorhodopsin ( max 525 nm, hence referred to as "GPR" 1 (5)), has been demonstrated to function as a light-driven proton pump (4), and its transport, spectroscopic, and photochemical reactions have been characterized by a number of laboratories in Escherichia coli expressed forms (1, 4 -10). In particular, its rapid photocycle, characteristic of transport rhodopsins, in E. coli membranes (4) and in its native membranes from Monterey Bay plankton (1) provides strong evidence that GPR functions as a light-driven proton pump photoenergizing SAR86 cells in their natural environment.
A variant group (clade II) of proteorhodopsin genes differing by ϳ22% in predicted primary structure from the clade I group has been detected in the Antarctic and deep ocean plankton from Hawaiian waters (1). A member of clade II isolated from 75-meter deep plankton from the Hawaiian Ocean Time Station when expressed in E. coli exhibits a blue-shifted absorption spectrum ( max 490 nm, hence referred to as "BPR") with vibrational fine structure, unlike the unstructured spectrum of GPR (1). A second clade II gene (from Antarctica) matches exactly these absorption properties when expressed in E. coli. 2 Here we characterize the Hawaiian deep water BPR transport, spectroscopic, and photochemical reactions. We observe striking differences between the Hawaiian 75-meter BPR and the Monterey Bay GPR, a comparison with which provides insight into proteorhodopsin diversity in mechanism and possibly physiological function.

MATERIALS AND METHODS
Site-directed Mutagenesis-For simplicity, the Monterey Bay protein eBAC31A08 is designated GPR and Hawaiian Ocean Time Station variant HOT75m4 is designated BPR in this report. Mutants BPR_D97N and BPR_E108Q were constructed using two-step PCR (11) following the procedure described for BPR_Q105L and GPR_L105Q (5). All of the mutations were confirmed by DNA sequencing.
Expression of Proteorhodopsins in E. coli Cells-GPR, BPR, and their mutants were expressed in E. coli strain UT5600 using plasmid pBAD-Topo as described previously (4). Overnight cultures of transformants were diluted 1:50 and grown to 0.4 absorbance units at 600 nm at 30°C. Cells were induced with 0.2% (ϩ)-L-arabinose for 3 h at 30°C. To obtain protein for flash photolysis and proton transport measurements, alltrans-retinal was added to a final concentration of 5 M with the arabinose. For measurement of retinal-generated absorption spectra, the apoproteins were induced without the addition of retinal.
Preparation of Cell Membranes for Retinal-generated Spectra Measurements-Cells were disrupted by sonication, unlysed cells and large debris were removed by low speed centrifugation, and cell membranes were collected by ultracentrifugation at 180,000 ϫ g for 1 h. The membranes were resuspended with 100 mM sodium acetate, MES, potassium phosphate (KPi), and Tris-HCl buffers at different pH values between 4.5 and 9, and retinal-generated absorption spectra were measured by adding all-trans-retinal from an ethanolic solution as described previously (5,12).
Proton Pumping Measurements-Right-side-out membrane vesicles were prepared as follows. 100 ml of induced cells were harvested by centrifugation at 3600 ϫ g for 10 min, resuspended into 10 ml of 30 mM Tris-HCl, pH 8.0, and 20% sucrose. 100 g (10 4 units) of lysozyme was added and stirred gently at room temperature. Spheroplasts were collected by centrifugation at 3600 ϫ g for 15 min at room temperature and resuspended in 1 ml of buffer (100 mM KPi, pH 7.0, 20 mM MgSO 4 , 20% sucrose, and 4 mg of DNase) and injected using a 1-ml syringe (18gauge needle) into 500 ml of a rapidly stirring solution of 50 mM KPi, pH 7.0, at 37°C. After 15 min of gently stirring at 37°C, Na-EDTA was added to a final concentration of 10 mM and stirred for another 15 min, after which MgSO 4 was added to a final concentration of 15 mM, followed by 15 min of stirring. Cell debris was removed at 3600 ϫ g for 10 min, and vesicles were collected at 16,000 ϫ g for 1 h and resuspended in 1 ml of 100 mM KPi, pH 7.0, and 10 mM MgSO 4 .
For the measurement of proton pumping of cells, the E. coli cells with expressed proteorhodopsins were pelleted at 3600 ϫ g for 10 min at room temperature and washed twice with unbuffered solution (10 mM NaCl, 10 mM MgSO 4 ⅐7H 2 O, 100 M CaCl 2 ) at room temperature. Samples were illuminated with 100 watts/m 2 at 500 ϩ 20 nm using a 200-watt Tungsten halogen lamp in combination with wide-band interference and heat-protecting (CuSO 4 and glass) filters. The pH was monitored with a computerized Beckman F72 pH meter.
Flash Photolysis-Flash-induced absorption changes were measured on a laboratory-constructed cross-beam flash photolysis system (similar to that used in Ref. 12). The actinic flash was from a Nd-YAG pulse laser (Continuum, Surelight I, 532 nm, 6 ns, 40 mJ). Signals from the photomultiplier were digitized by a DIGIDATA 1320A at 20 s/point and stored in a PC using the Clampex 8.0 program (both from Axon Instruments). 20 -100 signals obtained with a 10-s flash interval were averaged. To minimize excitation artifacts, equal amounts of signals with the measuring beam blocked (scattered light) were averaged and subtracted from the measuring beam-monitored signals. The Clampfit 8.0 program (Axon Instruments) was used for data analysis. For flash photolysis, membranes were prepared as described above and suspended in 50 mM Tris-HCl buffer at pH 9 and E. coli cells with expressed proteorhodopsins were pelleted at 3600 ϫ g for 10 min at room temperature and washed and suspended with 50 mM Tris-HCl at pH 9.
Absorption Spectroscopy and pK a of the Acid and Alkaline Form Transition-Retinal-generated absorption spectra of proteorhodopsin apoproteins in E. coli membrane were measured by difference spectroscopy as described previously (5,12). After adding the retinal solutions, the spectra were recorded with an SLM Aminco DW2000 spectrophotometer every 10 min for 1.5 h, after which no further absorption changes occurred. The spectra were used for calculating the pK a of the transition as follows. The max at different pH values were measured according to the reconstitution spectra of the membrane with proteorhodopsin apoproteins ("opsins"). The position of the absorption maxima of the mixtures of alkaline and acidic forms at different pH values does not depend linearly on the relative concentrations of the two forms because of different structures and half-bands of their spectra. For calculating the pK a of the transition, a calibration curve, which relates the relative concentration of alkaline and acidic forms with the position of the absorption maximum of their mixture, was calculated by mathematical summation of absorption spectra of pure acid and alkaline forms with corresponding coefficients. The corrected ratio of protonated and non-protonated forms at different pH values was determined using this calibration curve from the position of absorption maximum of their mixture and fitted with functions containing one or two pK a components (y ϭ A/(1 ϩ 10 pHϪpKa ).

Two Titratable Groups Control BPR Absorption Spectra-
One of the most notable distinguishing properties of retinal as a chromophore is the large variation of its absorption spectrum depending on interaction with the apoprotein ("spectral tuning") (13,14). In rhodopsins, retinal is covalently attached to the ⑀-amino group of a lysine residue, forming a protonated retinylidene Schiff base. In methanol, a protonated retinylidene Schiff base exhibits a max of 440 nm. Typically, the protein microenvironment shifts the max (the "opsin shift" (15)) to longer wavelengths, e.g. to 490 nm in BPR and 525 nm in GPR.
Microbial rhodopsins in general exhibit two spectral forms in a proton-dependent equilibrium: a blue-shifted alkaline species and red-shifted acidic species. The two species differ in absorption maximum by ϳ20 -40 nm depending on the pigment and conditions of assay. In bacteriorhodopsin (16), sensory rhodopsins I (17) and II (18), and GPR (6), the spectral shift is attributable to the ionization state of the primary protonated retinylidene Schiff base counterion (Asp-85 in bacteriorhodopsin, Asp-97 in proteorhodopsins), because mutating this residue to the isosteric asparagine in each of these pigments eliminates the pH-induced shift and stabilizes the absorption spectrum in the wild-type acidic form.
We titrated BPR in E. coli membranes to determine its alkaline and acidic spectra and the pK a of the spectral transition. Acid induces a shift of 46 nm (Fig. 1), a larger shift than the 20 nm seen in GPR (5) and a loss of vibrational fine structure in the spectrum as observed for sensory rhodopsin II (19). The GPR spectral transition fits well to a single pK a of 7.2 (Fig 2A); however, a second component is evident in the BPR titration according to hyperbolic fits to the data that yields a diprotic fit with pK a values 6.2 and 7.8 (Fig. 2B). Furthermore, unlike other microbial rhodopsins tested, the mutation of the expected Schiff base counterion Asp-97 to Asn does not completely eliminate an acid-induced shift but rather a 10-nm red-shift (Fig. 1) with a pK a of 7.1 (Fig. 2C) occurs in the BPR_D97N mutant pigment. This unique behavior among microbial rhodopsins indicates that an additional titratable group modulates the environment of the chromophore in BPR. Note that the diprotic titration behavior of BPR differs from the two pK a spectral transitions reported for bacteriorhodopsin (20) and sensory rhodopsin I (21). The diprotic behavior in those cases are attributable to the dependence of the pK a of the retinylidene Schiff base counterion aspartyl carboxylate (Asp-85 in BR and Asp-76 in SRI, which correspond to Asp-97 in BPR) on the titration of another residue in the protein. In those cases, substitution of the aspartyl counterion residue with asparagine eliminates the spectral shift entirely and the aspartyl counterion-independent shift evident in BPR_D97N (Fig. 2C) does not occur.
BPR Photochemical Reactions Are an Order of Magnitude Slower than in GPR-Laser flash photolysis of the BPR pigment in E. coli cells reveals major differences in kinetics of the BPR photocycle from that of GPR (Fig. 3). In cells, laser flashinduced absorption difference spectra in the range of 350 -600  (Fig. 3A). The return of BPR490 absorption (Fig. 3B) governed by the decay of the O species (Fig. 3C) is Ͼ10-fold slower than that for GPR in cells under the same conditions.
In contrast to GPR, the blue-shifted M intermediate accumulates in BPR in very low concentrations with maximum amplitude of Յ7% of the depletion of the unphotolysed state. This low extent of accumulation and high scattering renders the signal/noise ratio very poor below 400 nm in whole cell preparations. Furthermore, the appearance of M is masked above ϳ380 nm by the depletion of the unphotolysed state. In membranes, we detected the rise and decay of BPR M at wavelengths below 380 nm, although its kinetics is difficult to determine because of perturbation by a depletion of absorption in the near-UV region (data not shown). Clear formation of M was observed in purified BPR in dodecyl maltoside micelles and in membranes of the BPR_E108Q mutant (Fig. 4).
Asp-97 and Glu-108 Function as a Proton Acceptor and Donor, Respectively, for Schiff Base Proton Transfer Reactions during the BPR Photocycle-The appearance of absorption changes near 370 nm is diagnostic of an unprotonated retinylidene Schiff base in this BPR intermediate. In GPR, Asp-97 is the proton acceptor from the Schiff base in the corresponding photocycle transition and M formation in GPR_D97N is essentially completely blocked (6). BPR_D97N shows a 7-fold reduced accumulation of M, confirming a similar role of Asp-97 as a proton acceptor in BPR (Fig. 4A). However, some M formation is still evident in BPR_D97N, indicating that an alternative proton acceptor residue is present in this mutant. Possibly, the titratable group that influences the absorption maximum of BPR_D97N is the alternative proton acceptor.
In GPR, Glu-108 is the proton donor to the Schiff base for the FIG. 4. Laser flash-induced absorption changes in BPR and its D97N and E108Q mutants. A, absorption changes at wavelengths characteristic for M states in His 6 -tagged BPR and BPR_D97N at equal concentrations purified by Ni 2ϩ -affinity chromatography in 0.025% dodecyl maltoside. B, absorption changes in E. coli membranes containing BPR or BPR_E108Q. Signals were normalized to give equal amplitudes at 1 ms. proton transfer occurring during the transition of the M intermediate to later red-shifted states as evidenced by a greatly retarded M decay in GPR_E108Q. The corresponding photocycle transition in BPR is also greatly slowed in GPR_E108Q, indicating a similar role as a Schiff base proton donor for this residue (Fig. 4B). As noted above, in the wild type BPR photocycle. most accumulation is in the red-shifted O state without marked accumulation of M (Figs. 3A and 4B). In BPR_E108Q, the reversed situation was observed, i.e. the dominant intermediate is M with negligible accumulation of the O state (Fig.  4B). This accumulation allows determination of the absorption maximum of the M state of BPR_E108Q, which is ϳ390 nm (data not shown), slightly shorter than that of the M intermediate in GPR.
The Spectral Tuning Switch Residue at Position 105 Modulates the Photocycle-By site-specific mutagenesis of residues in the predicted retinal-binding pockets of both pigments, we have found that a single residue substitution (L105Q in GPR versus Q105L in BPR) nearly completely interconverts the absorption spectra of the two pigments. GPR_L105Q is blueshifted and gains vibrational fine structure, whereas BPR_Q105L is shifted to green absorption and loses fine structure in its spectrum (5).
In addition to functioning as a spectral tuning switch, the interconversion of residue 105 between GPR and BPR has pronounced effects on the photocycles of the pigments. As noted above, the photocycle of GPR is ϳ10 times faster than that of BPR with evident M accumulation unlike BPR when monitored at 400 nm (Fig. 5). The mutation L105Q in GPR slows the photocycle to a value typical of BPR with no M signal at 400 nm. The reciprocal mutation Q105L in BPR does not accelerate the photocycle but does cause clear accumulation of M in the 400-nm trace. Therefore, the residue at position 105 acts as a switch for both spectral and photochemical properties of GPR and BPR, but differences in addition to that at position 105 must contribute to the relatively rapid photocycle of GPR.

BPR-mediated Proton Fluxes in E. coli Vesicles and Cells-
Previous work has shown that GPR exhibits light-induced electrogenic ejection of protons from E. coli cells (4), and subsequent studies provide compelling evidence that GPR functions as a light-driven proton pump (1, 4 -6, 8, 9) in its natural marine environment (1). Here we examine the transport activity of GPR and BPR both in E. coli right-side-out vesicles and cells.
In E. coli vesicles, both GRP and BPR exhibit outward lightinduced proton transport (Fig. 6A). At similar pigment concentrations of GPR and BPR, protons are ejected at a slower rate in BPR, as expected from its slower photocycle rate. In agreement with a study of GPR proton transport (7), we do not observe the inverse direction of pumping by either rhodopsin under acidic pH in vesicles or in intact cells, contrary to what has been reported for GPR in a model system (8). At pH 6, cells with GPR and BPR both carry out light-driven proton ejection (data not shown); however, at pH 7, there is a net proton influx for the cells with BPR upon illumination (Fig. 6B). GPR photoexcitation induces proton efflux, but an initial transient net proton influx is observed following illumination. E. coli cells without proteorhodopsin do not exhibit light-induced proton fluxes under our conditions.
Our interpretation is that the pumping by BPR triggers the opening of voltage-gated proton channels in the E. coli membrane, which results in proton influx driven by the generated and preexisting electrochemical potential of protons across the membranes of the energized E. coli cells. A similar interpretation was advanced for light-induced bacteriorhodopsin-mediated inward proton fluxes in Halobacterium membranes attributed to voltage gating of an H ϩ -ATPase and Na ϩ /H ϩ antiporter (22). We examined proton fluxes in E. coli cells treated with valinomycin and K ϩ , which eliminates the electrical component of the H ϩ electrochemical gradient generated by the outwardly directed H ϩ transport by BPR. Confirming our interpretation, the direction of proton fluxes induced by BPR at pH 7 is reversed (i.e. converted to proton efflux) by K ϩ /valinomycin (Fig. 6C). Elimination of all of the BPR-mediated proton fluxes by the proton-ionophore carbonyl cyanide p-chlorophenylhydrazone (10 M) confirms that the proton fluxes are not passive currents responding to electrogenic transport of another ion by BPR (data not shown).
We explain the different proton fluxes from BPR and GPR photocycling as the result of the slower rate of proton ejection of BPR. In BPR-containing cells, the proton currents are dominated by the triggered proton influx through voltage-gated channels, whereas in GPR-containing cells, the triggered influx is overcome by the greater proton ejection rate of the pigment. Confirming this interpretation, low light intensity illumination of GPR, which reduces the rate of proton ejection to that of BPR at high intensity, results in a proton influx response similar to that of BPR in Fig. 6B.
Biological Significance of GPR and BPR Variation-The difference in photocycle rates between GPR and BPR and their different absorption maxima may be explained as an adaptation to the different light intensities in their respective marine environments based on measured spectral distributions of intensities of solar illumination at the ocean surface and at various depths (23). At solar radiation intensity of 1.5 watts⅐m Ϫ2 ⅐nm Ϫ1 at 525 nm (the absorption maximum of GPR) on the ocean surface under brightest conditions using a 95-nm half-bandwidth of GPR and 2 A 2 cross-sectional area, we calculate that 1 photon/proteorhodopsin molecule would be available for absorption for each ϳ150 ms. During this time, a population of GPR molecules reaches 98% recovery (Fig. 3B) and thus the GPR photocycle rate is well matched to the incident fluence rate. At 75 m, the depth at which the BPR gene was identified, the solar intensity at this wavelength is ϳ50 times less and the quanta available for absorption through the entire absorption spectrum are even smaller due to the very sharp attenuation in light at wavelengths above 500 nm. Because of the shift of solar radiation to the blue-width depth, the spectral shift to 488 nm in BPR compensates this decrease in available light energy Ͼ5-fold, but at most 1 quantum/1.5 s/BPR molecule would be available for absorption. Matching this photon fluence rate would require a 10-fold slower photocycle in BPR than in GPR as we observe (Fig. 3). Hence, there is no selective pressure for a photocycle faster than that of BPR at that depth.
BPR may function to energize cells by light-driven electro- genic proton pumping as does GPR. However, we do not expect a significant contribution of solar energy capture from BPR, which is limited by the low light intensities in deep waters. In E. coli, at neutral pH evidently the initial inward movement of protons, observed in both GPR and BPR, could not be overcome by BPR-driven proton extrusion from the cell even at saturating light intensities because of its slow cycling. This observation may indicate a regulatory rather than energy-harvesting function of BPR, which may be based on light-induced opening of proton channels in response to the initial increase in membrane electrical potential and dissipation of preexisting gradients. A possible function might be modulation of cell physiology depending on light intensity variations due to daily or seasonal changes or depth. Because little is known regarding the cells producing BPR, we cannot exclude that BPR may be coupled by protein-protein interaction as are haloarchaeal sensory rhodopsins with a transducer component not present in the E. coli membrane. In this case, BPR may not exhibit even its relatively slow light-driven proton transport in nature if its transport activity is inhibited by transducer interaction as is the case with the haloarchaeal sensory rhodopsins (24 -26).