Research Article

Bacterial Rhodopsin: Evidence for a New Type of Phototrophy in the Sea

See allHide authors and affiliations

Science  15 Sep 2000:
Vol. 289, Issue 5486, pp. 1902-1906
DOI: 10.1126/science.289.5486.1902


Extremely halophilic archaea contain retinal-binding integral membrane proteins called bacteriorhodopsins that function as light-driven proton pumps. So far, bacteriorhodopsins capable of generating a chemiosmotic membrane potential in response to light have been demonstrated only in halophilic archaea. We describe here a type of rhodopsin derived from bacteria that was discovered through genomic analyses of naturally occuring marine bacterioplankton. The bacterial rhodopsin was encoded in the genome of an uncultivated γ-proteobacterium and shared highest amino acid sequence similarity with archaeal rhodopsins. The protein was functionally expressed inEscherichia coli and bound retinal to form an active, light-driven proton pump. The new rhodopsin exhibited a photochemical reaction cycle with intermediates and kinetics characteristic of archaeal proton-pumping rhodopsins. Our results demonstrate that archaeal-like rhodopsins are broadly distributed among different taxa, including members of the domain Bacteria. Our data also indicate that a previously unsuspected mode of bacterially mediated light-driven energy generation may commonly occur in oceanic surface waters worldwide.

Retinal (vitamin A aldehyde) is a chromophore that binds integral membrane proteins (opsins) to form light-absorbing pigments called rhodopsins. Rhodopsins are currently known to belong to two distinct protein families. The visual rhodopsins, found in eyes throughout the animal kingdom (1), are photosensory pigments. Archaeal rhodopsins, found in extreme halophiles, function as light-driven proton pumps (bacteriorhodopsins), chloride ion pumps (halorhodopsins), or photosensory receptors (sensory rhodopsins) (2–5). The two protein families show no significant sequence similarity and may have different origins. They do, however, share identical topologies characterized by seven transmembrane α-helices that form a pocket in which retinal is covalently linked, as a protonated Schiff base, to a lysine in the seventh transmembrane helix (helix G). Recently, a protein with high sequence similarity to the archaeal rhodopsins has also been found in the eukaryote Neurospora crassa (6). The eucaryal rhodopsin formed a photochemically reactive pigment when bound to all-trans retinal and exhibited photocycle kinetics similar to those of archaeal sensory rhodopsins (7). To date, however, no rhodopsin-like sequences have been reported in members of the domain Bacteria.

Cloning of proteorhodopsin. Sequence analysis of a 130-kb genomic fragment that encoded the ribosomal RNA (rRNA) operon from an uncultivated member of the marine γ-Proteobacteria(that is, the “SAR86” group) (8, 9) (Fig. 1A) also revealed an open reading frame (ORF) encoding a putative rhodopsin (referred to here as proteorhodopsin) (10). The inferred amino acid sequence of the proteorhodopsin showed statistically significant similarity to archaeal rhodopsins (11). The majority of predicted proteins encoded by ORFs upstream and downstream of the proteorhodopsin gene, as well as the rRNA operon, showed highest similarity to proteobacterial homologs. Given the large amount of apparent lateral gene transfer observed in recent whole genome studies, it is not surprising that some predicted proteins (17 of 74) had significantly greater similarity to those from other bacterial groups, including Actinomycetesand Gram-positive bacteria (12, 13). No other ORFs encoding archaeal-like genes, however, were detected in the vicinity of the proteorhodopsin gene, verifying the bacterial origin of the 130-kb genome fragment.

Figure 1

(A) Phylogenetic tree of bacterial 16S rRNA gene sequences, including that encoded on the 130-kb bacterioplankton BAC clone (EBAC31A08) (16). (B) Phylogenetic analysis of proteorhodopsin with archaeal (BR, HR, and SR prefixes) and Neurospora crassa (NOP1 prefix) rhodopsins (16). Nomenclature: (HR, halorhodopsin; SR, sensory rhodopsin; BR, bacteriorhodopsin). Halsod, Halorubrum sodomense; Halhal, Halobacterium salinarum(halobium); Halval, Haloarcula vallismortis; Natpha, Natronomonas pharaonis; Halsp, Halobacteriumsp; Neucra, Neurospora crassa.

The proteorhodopsin gene encoded a polypeptide of 249 amino acids, with a molecular weight of 27 kD. Hydropathy plots indicated seven transmembrane domains, a typical feature of the rhodopsin protein family, that aligned well with the corresponding helices of the archaeal rhodopsins. The amino acid residues that form a retinal binding pocket in archaeal rhodopsins are also highly conserved in proteorhodopsin (Fig. 2). In particular, the critical lysine residue in helix G, which forms the Schiff base linkage with retinal in archaeal rhodopsins, is present in proteorhodopsin. Analysis of a structural model of proteorhodopsin (14), in conjunction with multiple sequence alignments, indicates that the majority of active site residues are well conserved between proteorhodopsin and archaeal bacteriorhodopsins (15).

Figure 2

Secondary structure of proteorhodopsin. Single-letter amino acid codes are used (33), and the numbering is as in bacteriorhodopsin. Predicted retinal binding pocket residues are marked in red.

A phylogenetic comparison with archaeal rhodopsins placed proteorhodopsin on an independent long branch, with moderate statistical support for an affiliation with sensory rhodopsins (16) (Fig. 1B). The finding of archaeal-like rhodopsins in organisms as diverse as marine proteobacteria and eukarya (6) suggests a potential role for lateral gene transfer in their dissemination. Available genome sequence data are insufficient to identify the evolutionary origins of the proteorhodopsin genes. The environments from which the archaeal and bacterial rhodopsins originate are, however, strikingly different. Proteorhodopsin is of marine origin, whereas the archaeal rhodopsins of extreme halophiles experience salinity 4 to 10 times greater than that in the sea (14).

Functional analysis. To determine whether proteorhodopsin binds retinal, we expressed the protein in Escherichia coli(17). After 3 hours of induction in the presence of retinal, cells expressing the protein acquired a reddish pigmentation (Fig. 3A). When retinal was added to the membranes of cells expressing the proteorhodopsin apoprotein, an absorbance peak at 520 nm was observed after 10 min of incubation (Fig. 3B). On further incubation, the peak at 520 nm increased and had a ∼100-nm half-bandwidth. The 520-nm pigment was generated only in membranes containing proteorhodopsin apoprotein, and only in the presence of retinal, and its ∼100-nm half-bandwidth is typical of retinylidene protein absorption spectra found in other rhodopsins. The red-shifted λmax of retinal (λmax = 370 nm in the free state) is indicative of a protonated Schiff base linkage of the retinal, presumably to the lysine residue in helix G (18).

Figure 3

(A) Proteorhodopsin-expressing E. coli cell suspension (+) compared to control cells (−), both with all-trans retinal. (B) Absorption spectra of retinal-reconstituted proteorhodopsin in E. coli membranes (17). A time series of spectra is shown for reconstituted proteorhodopsin membranes (red) and a negative control (black). Time points for spectra after retinal addition, progressing from low to high absorbance values, are 10, 20, 30, and 40 min.

Light-mediated proton translocation was determined by measuring pH changes in a cell suspension exposed to light. Net outward transport of protons was observed solely in proteorhodopsin-containing E. coli cells and only in the presence of retinal and light (Fig. 4A). Light-induced acidification of the medium was completely abolished by the presence of a 10 μM concentration of the protonophore carbonyl cyanide m-chlorophenylhydrazone (19). Illumination generated a membrane electrical potential in proteorhodopsin-containing right-side-out membrane vesicles, in the presence of retinal, reaching –90 mV 2 min after light onset (20) (Fig. 4B). These data indicate that proteorhodopsin translocates protons and is capable of generating membrane potential in a physiologically relevant range. Because these activities were observed in E. coli membranes containing overexpressed protein, the levels of proteorhodopsin activity in its native state remain to be determined. The ability of proteorhodopsin to generate a physiologically significant membrane potential, however, even when heterologously expressed in nonnative membranes, is consistent with a postulated proton-pumping function for proteorhodopsin.

Figure 4

(A) Light-driven transport of protons by a proteorhodopsin-expressing E. coli cell suspension. The beginning and cessation of illumination (with yellow light >485 nm) is indicated by arrows labeled ON and OFF, respectively. The cells were suspended in 10 mM NaCl, 10 mM MgSO4·7H2O, and 100 μM CaCl2. (B) Transport of 3H+-labeled tetraphenylphosphonium ([3H+]TPP) in E. coli right-side-out vesicles containing expressed proteorhodopsin, reconstituted with (squares) or without (circles) 10 μM retinal in the presence of light (open symbols) or in the dark (solid symbols) (20).

Archaeal bacteriorhodopsin, and to a lesser extent sensory rhodopsins (21), can both mediate light-driven proton-pumping activity. However, sensory rhodopsins are generally cotranscribed with genes encoding their own transducer of light stimuli [for example, Htr (22, 23)]. Although sequence analysis of proteorhodopsin shows moderate statistical support for a specific relationship with sensory rhodopsins, there is no gene for an Htr-like regulator adjacent to the proteorhodopsin gene. The absence of an Htr-like gene in close proximity to the proteorhodopsin gene suggests that proteorhodopsin may function primarily as a light-driven proton pump. It is possible, however, that such a regulator might be encoded elsewhere in the proteobacterial genome.

To further verify a proton-pumping function for proteorhodopsin, we characterized the kinetics of its photochemical reaction cycle. The transport rhodopsins (bacteriorhodopsins and halorhodopsins) are characterized by cyclic photochemical reaction sequences (photocycles) that are typically <20 ms, whereas sensory rhodopsins are slow-cycling pigments with photocycle half-times >300 ms (3). This large kinetic difference is functionally important, because a rapid photocycling rate is advantageous for efficient ion pumping, whereas a slower cycle provides more efficient light detection because signaling states persist for longer times. To assess the photochemical reactivity of proteorhodopsin and its kinetics, we subjected membranes containing the pigment to a 532-nm laser flash and analyzed flash-induced absorption changes in the 50-μs to 10-s time window. We observed transient flash-induced absorption changes in the early times in this range (Fig. 5). Transient depletion occurred near the absorption maximum of the pigment (500-nm trace, Fig. 5, top panel), and transient absorption increase was detected at 400 nm and 590 nm, indicating a functional photocyclic reaction pathway. The absorption difference spectrum shows that within 0.5 ms, an intermediate with maximal absorption near 400 nm is produced (Fig. 5, bottom panel), which is typical of unprotonated Schiff base forms (M intermediates) of retinylidene pigments. The 5-ms minus 0.5-ms difference spectrum shows that after M decay, an intermediate species that is red-shifted from the unphotolyzed 520-nm state appears, which is analogous to the final intermediate (O) in bacteriorhodopsin. The decay of proteorhodopsin O is the rate-limiting step in the photocycle and is fit well by a single exponential process of 15 ms, with an upward baseline shift of 13% of the initial amplitude. A possible explanation is heterogeneity in the proteorhodopsin population, with 87% of the molecules exhibiting a 15-ms photocycle and 13% exhibiting a slower recovery. An alternative explanation is that photocycle complexity such as branching produces a biphasic O decay. Consistent with this alternative, the O recovery is fit equally well as a two-exponential process with a fast component, with a 9-ms half-time (61% of the total amplitude) and a slow component with a 45-ms half-time (39% of the total amplitude). In either case, the rapid photocycle rate, which is a distinguishing characteristic of ion pumps, provides additional strong evidence that proteorhodopsin functions as a transporter rather than as a sensory rhodopsin.

Figure 5

Laser flash-induced absorbance changes in suspensions of E. coli membranes containing proteorhodopsin. A 532-nm pulse (6 ns duration, 40 mJ) was delivered at time 0, and absorption changes were monitored at various wavelengths in the visible range in a lab-constructed flash photolysis system as described (34). Sixty-four transients were collected for each wavelength. (A) Transients at the three wavelengths exhibiting maximal amplitudes. (B) Absorption difference spectra calculated from amplitudes at 0.5 ms (blue) and between 0.5 ms and 5.0 ms (red).

Implications. The γ-proteobacteria that harbor the proteorhodopsin are widely distributed in the marine environment. These bacteria have been frequently detected in culture-independent surveys (24) in coastal and oceanic regions of the Atlantic and Pacific Oceans, as well as in the Mediterranean Sea (8,25–29). In addition to its widespread distribution, preliminary data also suggest that this γ-proteobacterial group is abundant (30, 31) and specifically localized in marine surface waters. Preliminary data (30) also indicate that the abundance of SAR86-like bacteria positively correlates with proteorhodopsin mRNA expression. The absorbance λmax of proteorhodopsin at 520 nm matches well with the photosynthetically available irradiance in the ocean's upper water column. Furthermore, some phylogenetic relatives of the proteorhodopsin-containing bacteria are chemolithoautotrophs that use CO2 as a sole carbon source (32). Proteorhodopsin could support a photoheterotrophic lifestyle, or it might in fact support a previously unrecognized type of photoautotrophy in the sea. Either of these alternatives suggests the possibility of a previously unrecognized phototrophic pathway that may influence the flux of carbon and energy in the ocean's photic zone worldwide.

  • * To whom correspondence should be addressed. E-mail: delong{at}


Stay Connected to Science

Navigate This Article