Rhodopsin, the founding father of the type-2 family was first identified as the visual pigment of the rod photoreceptor cell. In the 19th century, groundbreaking research on vision by Müller, Boll and Kühne led to the visual perception, that light capture occurred in the distal part of the human retina (Figure 2), in particular the outer segments of the photoreceptor cells (Müller, 1855; Boll, 1877; Ewald and Kühne, 1878). The typical red color of this tissue disappeared upon illumination, which was termed “bleaching,” and could to some extent be regenerated upon subsequent dark adaptation of the isolated eyecup. As of the 1930s it became apparent that a membrane-bound protein in the rod photoreceptor cell was responsible for the red color (Tansley, 1931; Bliss, 1948; Wald, 1953). This protein was named rhodopsin, after the ancient Greek words ροδεοσ (rhodeos, rose-coloured) and οψισ (opsis, which appropriately can be translated as sight or eyes). It was found to owe its spectral properties to a covalently bound cofactor, eventually named retinal (Wald, 1935, 1953, 1968; Hubbard et al., 1971). Subsequently, it was discovered that the cone photoreceptors in the vertebrate retina harboured closely related visual pigments (Morton and Pitt, 1957; Dartnall, 1962c; Mustafi et al., 2009). Thereafter, it became known that the invertebrate retina applied structurally very similar, but photochemically slightly differently operating visual pigments (Hara et al., 1967; Suzuki et al., 1993; Gärtner, 2000). Similar “bi-stable” pigments in fact are also active in the vertebrate retina, like the well-known melanopsins (Provencio et al., 1998; Kumbalasiri and Provencio, 2005). Another highlight was the growing insight that the visual pigments form part of the superfamily of G protein-coupled receptors (Kühn, 1984; Hargrave and McDowell, 1992; Palczewski and Orban, 2013). As a matter of fact, rhodopsin is the cornerstone of the major subfamily in this widespread receptor family.

In the early 1970s, fascinated by the dark-purple colonies of the salt lake thriving archaeon Halobacterium salinarum, Oesterhelt reported the surprising discovery that an intrinsic membrane protein was dominating purple patches in the cellular membrane of this archaeon and also harboured retinal as the chromophoric cofactor (Oesterhelt and Stoeckenius, 1971; Oesterhelt and Hess, 1973). At that time archaea were considered a subfamily of bacteria, and Oesterhelt coined the name bacteriorhodopsin (BR). Surprisingly, it was discovered that bacteriorhodopsin functions as a light-driven outward-directed proton pump, creating a proton-motive force enabling the cellular ATP-synthase complex to supply the cell with metabolic energy in the form of ATP (Oesterhelt et al., 1991). While bacteriorhodopsin is the dominant photoreceptor in Halobacterium salinarum, this archaeon eventually turned out to harbour several related photosensitive proteins, both with ion transport and sensory functions (Oesterhelt, 1998). Since the 1990s this field exploded, with more strains, including eukaryotic organisms like algae and fungi, and other functionalities being revealed every year (Béjà et al., 2000; Spudich et al., 2000; Brown, 2004; Rozenberg et al., 2021; Broser, 2022; Nagata and Inoue, 2022). More recently even viral rhodopsins have been discovered (Philosof and Béjà, 2013; Bratanov et al., 2019; Zabelskii et al., 2020). The overall structure and photochemistry of these pigments are very similar, and they are now considered to be a primary factor in marine phototrophy and solar energy conversion (Kirchman and Hanson, 2013; Gómez-Consarnau et al., 2019).

Most type-2 rhodopsins, and in particular cone visual pigments and invertebrate pigments are very sensitive to at least partial denaturation upon solubilization in detergent solution (Bliss, 1948; Kropf, 1982; Okano et al., 1989). While commercial detergents like Triton X-100, CTAB, LDAO and Emulphogene BC-720 could dissolve the vertebrate rod pigment rhodopsin into mixed micelles with none or only very slow loss of spectral properties at room temperature, for most other pigments only the very mild agent digitonin could be applied (Tansley, 1931; Knudsen and Hubbell, 1978; Okano et al., 1989; Hofmann and Palczewski, 2015). This natural compound, a steroidal glycone extracted from Digitalis purpurea, however has the disadvantage that its commercial preparations were quite expensive and did vary in composition and aqueous solubility (Bridges, 1977). Major progress was attained in the 1970s upon development of the alkylsaccharide detergents 1-O-n-β-D-octylglucoside (octylglucoside, OG), nonylglucoside (NG) and dodecylmaltoside (DDM) (Stubbs et al., 1976; DeGrip and Bovee-Geurts, 1979). DDM in particular turned out to maintain thermal stability and spectral and photochemical properties of rhodopsin almost as well as digitonin (DeGrip, 1982; VanAken et al., 1986). Additionally, DDM is well accessible and affordable through organic synthesis, and has therefore become the most popular detergent in the membrane protein field. Also, in case a protein purified in DDM needs to be reconstituted in a lipid bilayer for certain applications (nanodisc or proteoliposome, Figure 4), DDM can be easily extracted via cyclodextrin inclusion (DeGrip et al., 1998). More recently, a large number of novel detergents based upon the structural principle of DDM have been developed, some of which provide better thermal stability or better crystallization conditions for selected membrane proteins than DDM, but all requiring more complex synthesis (Hussain et al., 2016; Nguyen et al., 2018; Ehsan et al., 2020; Urner et al., 2020).

The absorbance band profiles of type-2 rhodopsins are quite similar (Figure 3), but the position of the α-band varies strongly for the visual pigments. The vertebrate rod photoreceptor pigment rhodopsin has quite a broad range in its absorbance maximum (Rh1 subset, 440–520 nm), with fresh-water animals slightly red-shifted and marine animals blue-shifted depending on the depth of their habitat (Locket, 1977; Luk et al., 2016; Musilova et al., 2019). Vertebrate cone pigments cover the entire visible spectrum, and can be divided into four subsets, the long-wavelength (LWS, absorbance maximum range 520–640 nm), green (Rh2, 460–530 nm), blue (SWS2, 400–470 nm), and UV (SWS1, 350–450 nm) sensitive pigments (Crescitelli, 1991; Yokoyama and Yokoyama, 2000; Imamoto and Shichida, 2014). This classification is not only based upon spectral sensitivity, but also upon sequence similarity (Nathans, 1987; Hunt and Collin, 2014; Jacobs, 2018; El Khatib and Atamian, 2019). Invertebrate visual pigments are more scattered over the visible region and can range from 340 nm up to 600 nm (Gärtner, 2000; Katz and Minke, 2009; Tsukamoto and Terakita, 2010). Non-visual animal rhodopsins are scattered over the 340–550 nm region (Leung and Montell, 2017; Pérez J. H. et al., 2019; Moraes et al., 2021).

The spectral properties of the type-2 rhodopsins depend on the 11-cis configuration of the retinylidene chromophore. Next to the standard retinal (retinal A1, Figure 1), several natural modifications occur (analogs). In fresh-water and coastal vertebrates 11-cis 3-dehydroretinal (retinal A2) has been observed (Figure 5) (Bridges, 1972; Yoshizawa, 1984; Imai et al., 1999). The longer conjugated chain red-shifts the absorbance maximum by 20–40 nm in rod pigments and up to 70 nm in cone pigments, as compared to retinal A1, to compensate for the lower blue light intensity in their habitat (Dartnall, 1962c; Hubbard et al., 1971). These “A2-rhodopsins” are also referred to as porphyropsins. In insects and some other invertebrates, 11-cis 3-hydroxy- and 4-hydroxyretinals have been detected (Figure 5) (Vogt and Kirschfeld, 1984; Matsui et al., 1988; Seki and Vogt, 1998). These modifications blue-shift the absorbance maximum by 20–40 nm, as compared to retinal A1 (Sekharan et al., 2011).

While such natural modifications are exploited to modulate the spectral position of a rhodopsin, the most effective approaches to shift the absorbance spectrum of the rhodopsin chromophore away from that of free retinal (380 nm) are protonation of the Schiff base and mutation of selected opsin residues lining the retinal binding pocket. For instance, only the Schiff base in UV absorbing rhodopsins, which absorb in the 350–380 nm region, is not protonated, while in all other classes it is protonated (Kusnetzow et al., 2004; Imamoto and Shichida, 2014). The large variation in the spectral properties in the latter classes is mainly due to the combined inductive effect of opsin binding pocket residues, in combination with H-bonding networks involving water molecules. On top of that, some vertebrate LWS visual pigments have developed a unique mutation (Glu197->His) creating a chloride binding site that effectuates a further 20–30 nm red-shift (Wang et al., 1993).

With respect to structural biology, bovine rod rhodopsin was a forerunner among all animal intrinsic membrane proteins, presenting the first detailed 3-D structure via X-ray crystallography in 2000, with many more to follow (Palczewski et al., 2000; Li et al., 2004; Okada et al., 2004). The seven transmembrane α-helical scaffold surrounding an accessible cofactor binding pocket proved to be the general motif for the entire G protein-coupled receptor family (Figure 6) (Sanchez-Reyes et al., 2017). This feat has stimulated advances in many other research fields, including drug design in the pharmaceutical sciences, study of protein structure-function correlations, and membrane protein-lipid interactions, both from experimental, theoretical and in-silico standpoints. Several natural factors concurred to enable this important step forward. First of all, rod rhodopsin is one of the few intrinsic membrane proteins that is available in relatively large quantities in domesticated animals, the most used being cattle (up to 1 mg of rhodopsin per eye), bullfrogs (up to 100 μg per eye) and chick (up to 100 μg LWS cone pigment per eye) (DeGrip et al., 1980; Toba and Hanawa, 1985; Yoshizawa and Kuwata, 1991). After enucleation and proper dark adaptation of the eyes, intact rod or cone outer segments (ROS or COS) can be easily isolated in a dark room under dim red light (>650 nm) that will not activate and bleach the pigment (Figure 3). Further, in dark-adapted ROS, rhodopsin makes up about 85% of the total protein content (DeGrip et al., 1980). Eventually, dark-adapted bovine retinae even became commercially available (Hormel Co., Austin, Minnesota, United States). Finally, bovine rod rhodopsin was found to be relatively resistant to destabilization by detergents as compared to most other visual pigments, allowing extensive purification. Likewise, it proved to be sufficiently stable in less mild but more crystal-production-favoring small detergents like OG and NG to facilitate crystallization trials (Palczewski et al., 2000; Park et al., 2008).

The bovine rhodopsin amino acid sequence was established thanks to heroic protein sequencing efforts (Abdulaev et al., 1982; Hargrave et al., 1983). Over time, sequence information became available more easily via genome mining and c-DNA-sequencing. Thus, it came out that most invertebrate visual pigments are similar in size to the vertebrate pigments (36–42 kD), but mollusc pigments are significantly larger (46–55 kD), because of the presence of a much longer C-terminal (Ovchinnikov et al., 1988b; Gärtner, 2000). This additional stretch is unique in having an insertion of up to eleven copies of a peculiar pentapeptide sequence (Pro-Pro-Gln-Gly-Tyr), which probably helps in immobilization of the protein in the microvillar membrane (Ryba et al., 1993; Gärtner, 2000). Longer C-terminal stretches are also found in non-visual rhodopsins. For instance, the VA-opsin and melanopsin family also show this feature, except that the pentapeptide insertion does not occur. Here the extra sequence probably has a function in complex regulation of signal processing and desensitization (Valdez-Lopez et al., 2020; Contreras et al., 2021). Some VA-opsins and melanopsins are even produced in two or more splicing isoforms, with longer and shorter C-terminals (Davies et al., 2010).

While squid provides fair quantities of visual pigment, the first complete 3-D crystal structures only became available since 2008, both because of the much lower stability of the pigments in detergent solution and since crystallization could only be achieved after proteolytic removal of most of the long C-terminal (Murakami and Kouyama, 2008; Shimamura et al., 2008). The overall fold of the seven-transmembrane α-helical scaffold is quite similar to bovine rhodopsin, but the structure of the long C-terminal could not be determined, of course. The position of the retinal chromophore is slightly different, since the Glu residue functioning as the direct counterion for the protonated Schiff base is displaced from the site in the vertebrate pigments (Terakita et al., 2004). The first crystal structure of an arthropod rhodopsin (jumping spider) was only recently published in 2019, and again shows the familiar seven α-helical fold with overall high similarity with the squid structure (Varma et al., 2019). So far, crystal structures of non-visual rhodopsins have not been reported.

The crystal unit cell of bovine rod rhodopsin contains a dimer, but its interaction pattern is very different from the natural one (Fotiadis et al., 2006; Palczewski, 2006). In fact, rhodopsin is equally active as a monomer, and the organization in the ROS disc membranes is still debated (monomer, dimer, longer stretches?) (Fotiadis et al., 2004; Chabre and LeMaire, 2005; Mishra et al., 2016; Zhang et al., 2016; Feldman et al., 2019; Zhao et al., 2019). Invertebrate visual rhodopsins are probably rigidly immobilized in their native membrane, which allows to discern the polarization plane of the incoming light (Gärtner, 2000; Stavenga et al., 2000).

The crystal structures are essential to resolve the protein fold of the rhodopsins and have confirmed several conjectures of the binding pocket. Biochemical, vibrational (resonance Raman and FTIR spectroscopy) and solid-state NMR studies already produced very strong evidence that it indeed harboured the 11-cis configuration of retinal (Groenendijk et al., 1980; Mathies et al., 1987; Lugtenburg et al., 1988; DeGrip and Rothschild, 2000; Mathies and Lugtenburg, 2000). Surely enough, this configuration best fitted the non-protein electronic density in the binding pocket. The same is true for the covalent binding of retinal to a lysine residue, for which the above-mentioned techniques also already provided a wealth of evidence (Bownds, 1967; DeGrip et al., 1973; Creemers et al., 1999; Mathies and Lugtenburg, 2000). However, to firmly establish protonation of the Schiff base the resolution of the crystal structures is not high enough. Instead, the evidence produced by vibrational and NMR spectroscopy is very convincing and in fact was later underpinned by quantum-chemical computation (Palings et al., 1987; Herzfeld and Lansing, 2002; Gascón et al., 2005; Tastan et al., 2014).

The sensitivity to detergent action also varies strongly between microbial rhodopsins. For instance, while bacteriorhodopsin (BR) is quite stable in OG, Triton X-100 and dodecylphosphocholine (DPC) even as a monomer, the rhodopsin proton pump from the cyanobacterium Gloeobacter violaceus (GR) strongly prefers DDM and is very unstable in DPC (Dencher and Heyn, 1978; Brouillette et al., 1989; Ganapathy et al., 2020). In general, OG and DDM are the preferred agents for solubilization of type-1 rhodopsins.

The spectral range of type-1 rhodopsins (360–690 nm) is comparable to that of type-2. There is less evidence for a clear relation to activity or habitat, an exception being the proton pump proteorhodopsin, which exhibits a blue-shift in deeper marine environments (Béjà et al., 2001; Bielawski et al., 2004). In a major distinction from type-2, microbial rhodopsins invariably exploit retinal A1 in the all-trans configuration as the basis for their light absorbance. Here, as well, a plethora of experimental evidence has demonstrated retinal binding to a lysine residue via a protonated Schiff base (Haupts et al., 1999).

Advanced angular electron diffraction studies on 2-D BR crystals in membrane patches already afforded a first glimpse into the organization of the helical transmembrane segments of type-1 rhodopsins (Henderson and Unwin, 1975; Mitra et al., 1993; Grigorieff et al., 1996; Heymann et al., 1997; Mitsuoka et al., 1999). The first 3-D crystal structures were reported for BR from 1997 onwards, and at a very high resolution slightly before that of bovine rhodopsin (Luecke et al., 1999; Pebay-Peyroula et al., 2000). This progress was aided by its high stability in detergent solutions and the relatively simple isolation from its native source. Bacteriorhodopsin is organized in large singular patches in the cellular membrane of Halobacterium salinarum, which can visually be observed and separated from other membrane fragments quite easily (Oesterhelt and Stoeckenius, 1971). In addition, type-1 rhodopsins complete a full photocycle (see below) and after photo-activation do not release the retinal, but thermally return to the ground state. This obviates the complexity of using dark rooms and shielding all experimental manipulations from room light exposure. Meanwhile, quite a number of crystal structures have been resolved for various classes of type-1 rhodopsins (Table 1). The most recent high resolution 3-D structures actually capitalized on the fantastic progress in cryo-EM (Hirschi et al., 2021; Kishi et al., 2022).

The available type-1 3-D structures show high similarity in protein fold and retinal pocket location. The basic seven α-helical transmembrane organization is comparable to that of type-2 (Figure 6), but for type-1 the helical packing is somewhat different and more compact. The loop segments connecting the helices are generally shorter and the retinal pocket is positioned differently to accommodate the longer all-trans chromophore instead of the curved 11-cis one (Figures 1, 6). Aspects of the binding pocket (retinal isomer and binding to a lysine residue via a Schiff base) again were in line with a wealth of evidence generated by biochemical and spectroscopic techniques (Lanyi, 2004). A recent XFEL study of the bacteriorhodopsin photocycle achieved a very high structural (ca 1.5 Å) and temporal (femtosecond) resolution and produced evidence for protonation of the Schiff base (Nogly et al., 2018). Also in the type-1 case the evidence generated by biophysical techniques like vibrational, EPR and NMR spectroscopy and by quantum-chemical computation is most convincing (Ernst et al., 2014; Brown and Ernst, 2017; Ryazantsev et al., 2019; Nagata and Inoue, 2022).

A conspicuous feature of most type-1 rhodopsins is that they organize in homo-oligomers, whether observed in the native membrane or in host cells. The most common arrangement for bacterial and archaeal rhodopsins are trimers or pentamers, though occasionally hexamers do occur as well (Hussain et al., 2015; Shibata et al., 2018; Kao et al., 2019). Circular dichroism spectroscopy provides evidence for exciton coupling between the chromophores (Cassim, 1992; Fujimoto and Inoue, 2020; Fujimoto, 2021). For eukaryotic type-1 rhodopsins, homo-dimeric as well as hetero-dimeric complexes are observed (Mukherjee et al., 2019; Govorunova et al., 2021; Broser, 2022). Isolated type-1 monomers are also functionally active, indicating that the oligomeric assembly probably affords optimal packing and mutual stabilization, and/or the opportunity to modulate monomer activity by inter-subunit interplay (Iizuka et al., 2019).

A novel feature was discovered in the enzyme-rhodopsins i.e. an additional transmembrane segment at the N-terminal (TM8), which functions as a connector with the cognate soluble enzyme domain and seems to be essential for modulating its activity (Ikuta et al., 2020; Tsunoda et al., 2021).

Interestingly, several thermostable microbial rhodopsins have been discovered. The crystal structure of the highly thermophilic rhodopsin (TR) from Thermus thermophilus was resolved to be very similar to that of the much less thermally stable xanthorhodopsin (XR) from Salinibacter ruber, including the binding crevice for the carotenoid antenna (Tsukamoto et al., 2016). Likewise, the crystal structure of the thermostable rhodopsin proton-pump from Rubrobacter xylanophilus (RxR) is very similar to that of bacteriorhodopsin (Hayashi et al., 2020). An unusually widely stable proton pump (pH, detergent, temperature), named Tara76 rhodopsin, was isolated from uncultured bacteria (Shim et al., 2021). Such data shed new light on the design options to increase thermal and environmental stability without a significant sacrifice in dynamics and activity (Hayashi et al., 2020).

Additional selected references relevant for this section have been compiled in Table 1.

The animal rhodopsins have meanwhile been classified in at least nine gene families (Opn1–Opn9) and two separate sets with some members still awaiting further assignment (Table 2). Physiological function and tissue distribution show incredible diversity (Janssen et al., 2003; Leung and Montell, 2017; Liebert et al., 2021; Moraes et al., 2021; Calligaro et al., 2022). The classical visual pigments come within Opn1 (cone pigments) and Opn2 (rod pigments). Pigments discovered later in the vertebrate retina, such as melanopsin, VA-opsin or neuropsin, peropsin (RRH) and RGR fall under Opn4, unclassified and Opn5, respectively (Table 2). The common thread still is primary signal transduction via at least one of the available G-protein species (Gt, Go, Gi, Gq, and Gs), with cross-activation, modulation or desensitization via a variety of other mediators. However, RGR and its mollusc counterpart retinochrome are exceptional in this context, since they act as photo-isomerases, binding all-trans retinal in the dark state, and releasing 11-cis retinal after photo-activation as a supply for regeneration of visual opsins (Hara et al., 1967; Pepe and Cugnoli, 1992; Zhang et al., 2019; Choi et al., 2021; Vöcking et al., 2021). Another remarkable subset are Opn5L, peropsin and Opn7 members, which also bind all-trans retinal in the dark state, but that seems to be the active state binding the G protein. Upon illumination they generate the 11-cis chromophore, which represents the resting state that in the case of Opn5L members may even thermally revert to the active state (Nagata et al., 2018; Yamashita, 2020; Karapinar et al., 2021; Sakai et al., 2022). An even more surprising observation is that some type-2 pigments may be involved in recognizing temperature differences or mechanical changes, or function as chemosensors or tumorigenic elements, possibly even without requiring their retinal cofactor (Shen et al., 2011; Park et al., 2013; Baker et al., 2015; Pérez-Cerezales et al., 2015; Leung et al., 2020; Xu et al., 2020; Córdova et al., 2021; Moraes et al., 2021).

Type-1 rhodopsins have been identified in archaea and eubacteria, including cyanobacteria, as well as in unicellular eukaryotes (algae, fungi, yeast) and more recently also in choanoflagellates and viruses (Lamarche et al., 2017; Bratanov et al., 2019; Zabelskii et al., 2020; Rozenberg et al., 2021; Govorunova et al., 2022b; Nagata and Inoue, 2022). Most of these pigments function as light-driven ion transporters or ion channels (Figure 7). The newly discovered xenorhodopsins and schizorhodopsins are exceptional as they perform inward-directed proton transport (Inoue et al., 2018; Inoue et al., 2020; Weissbecker et al., 2021; Brown, 2022). However, some type-1 rhodopsins display a photosensory function (sensory rhodopsins) and signal via a cognate transducer protein, which is totally different functionally and structurally from the animal G-proteins (Bogomolni and Spudich, 1991; Krah et al., 1994; Deininger et al., 1995). Overall, type-1 pigments are the dominant contributors to marine phototrophy (Casey et al., 2017; Larkum et al., 2018; Gómez-Consarnau et al., 2019). In addition, eukaryotic type-1 rhodopsins have been discovered which are intracellularly fused to an enzymatic domain and mediate light-driven enzyme activation (guanylyl cyclase, phosphodiesterase) or inhibition (guanylyl cyclase, based upon histidine kinase activity) (Avelar et al., 2014; Lamarche et al., 2017; Luck et al., 2019; Mukherjee et al., 2019; Tsunoda et al., 2021; Broser, 2022; Tian et al., 2022). These pigments have been termed as enzyme-rhodopsins.

The overall structure and photochemistry of all type-1 rhodopsins present a very similar pattern, though the sequence identity can be as low as 12%, and the kinetics of the photocycle can vary up to at least thousand-fold. The most recent addition, the heliorhodopsins, are not very different in their protein fold from e.g. BR in spite of a very low sequence identity (<10%) (Shihoya et al., 2019). Considering their inverted insertion into the membrane, very long photocycle and so far unknown functionality, they probably are better classified separately as type-3 rhodopsins (Tanaka et al., 2020; Chazan et al., 2022).

Type-2 rhodopsins can undergo a variety of posttranslational modifications (disulfide-bridge formation, N- and O-glycosylation, methylation, acetylation, myristylation, palmitoylation, phosphorylation), most of which are not properly executed by the bacterial or archaeal biosynthetic machinery (Table 3). Expression of bovine rhodopsin in bacteria and even yeast did not yield promising results (Mollaaghababa et al., 1996; Abdulaev and Ridge, 2000). Hence, for optimal heterologous expression a eukaryotic cell type had to be selected as a host. Attempts have been made to express type-2 pigments and related receptors in the eye of whole organisms (mouse, Xenopus) and in Caenorhabditis elegans using viral vectors or transgenic animals, but this gave relatively low yields or even led to retinal degeneration (Zhang et al., 2005; Salom et al., 2008; Cao et al., 2012; Salom et al., 2012). Eventually, the best results with sufficient posttranslational modification and targeting to the plasma membrane were obtained in some mammalian cell lines using plasmid transfection (COS, HEK, Neuroblastoma cell lines), in insect cell lines using baculoviral infection (Spodoptera Sf9 and Sf12 and Trichoplusia “High-Five”) and in Xenopus oocytes (Oprian et al., 1987; Janssen et al., 1988; Khorana et al., 1988; Karnik et al., 1993; Kazmi et al., 1996). The highest expression levels of functional pigments, with addition of 11-cis retinal during culture or after isolation of the cells, were obtained in suspension culture of insect cells or specially adapted HEK293 cells, with yields up to 130 nmol/L, equivalent to ca 5 mg bovine rhodopsin per liter (Klaassen and DeGrip, 2000; Reeves et al., 2002). Even then the pigment accounts for maximally 5 percent of the total cellular membrane protein, and further purification is inevitable. Eventually, gene manipulation lent a helping hand and it has now become common practice to add a small sequence tag to the pigment c-DNA, encoding a short peptide sequence to easily identify and purify the expressed pigment. Two approaches have become the most popular in the type-2 rhodopsin field. One exploited the availability of a monoclonal antibody against the C-terminal octapeptide of bovine rhodopsin (Molday, 1989). This allows for highly selective immuno-affinity purification using a suitable detergent like DDM for solubilization (Ridge et al., 1995). By adding to or replacing the native C-terminal with this octapeptide, the resulting tagged protein can be comfortably isolated. The second approach involved extending the C-terminal with six to ten histidine residues (His6-tag to His10-tag), which upon solubilization with a suitable detergent allows metal affinity purification over a matrix containing immobilized Ni²⁺ or Co²⁺ complexes (Janknecht et al., 1991; Janssen et al., 1995). Both approaches are very effective with hardly any perturbation of expression level and functionality of the pigment (Reeves et al., 1999; Bosman et al., 2003). Nevertheless, if necessary, a short target peptide sequence for a selective proteolytic enzyme can be introduced in front of the purification tag to remove it after purification (Sarramegna et al., 2006). Most Opn1 and Opn2 pigments can be satisfactorily purified by either procedure (Vissers and DeGrip, 1996; Shirzad-Wasei and DeGrip, 2016; Katayama et al., 2017; Katayama et al., 2019). Some pigments from the other subsets have been difficult to solubilize or are too unstable in detergent solution to survive purification. The alternative option then is to transfer the protein into the stabilizing lipid environment of nanodiscs (Figure 4), which requires hardly any detergent (amphipol or SMA-type) or very brief exposure to a suitable mild detergent (MSP-type). Exploiting the sequence tag on the incorporated protein, the protein-nanodisc unit is then easily purified again by affinity chromatography (Shirzad-Wasei et al., 2015; Cai et al., 2017; Ganapathy et al., 2020).

The opportunity to modify, bio-generate and purify type-2 pigments in sufficient quantities has given a tremendous boost to all mechanistic and functional studies. Analysis of the native proteins or binding pocket mutants, often in combination with ²H-, ¹³C- or F-labeling and/or chemical modification of retinal and/or with ¹⁵N- and/or ¹³C-labeling of protein residues or inserting modified amino acids, has provided a wealth of data, underpinning, extending and refining the information obtained from crystal structures (see next sections). Groundbreaking details of dark state structures have been excavated by biochemical (e.g. limited proteolysis, selective chemical modification, selective deuteration, atomic force microscopy, cryo-EM) and biophysical studies (e.g. FTIR and resonance Raman spectroscopy, solid-state NMR spectroscopy, EPR spectroscopy) (Table 3). This also fueled a large body of theoretical and in-silico efforts (molecular dynamics, quantum-chemical calculation and modeling) (Ryazantsev et al., 2019; Pedraza-González et al., 2020; Fujimoto, 2021; Mroginski et al., 2021; Church et al., 2022b). As a result of all these exertions, it has already been possible to construct a highly detailed picture of the dark state of bovine rhodopsin.

For the archaeal and bacterial type-1 rhodopsins, a heterologous expression host was more easily identified. Escherichia coli strains had already been developed for uncomplicated suspension culture, high productivity, low proteolytic activity and easy transformation. Plasmids with inducible promoters became available, and were further engineered with specific features, like producing the necessary enzymatic machinery to generate all-trans retinal from its precursor β-carotene (Kim et al., 2008). Nevertheless, in most cases just supplementing the cell culture with all-trans retinal together with inducing opsin expression or even after membrane isolation was sufficient to produce the full equivalent of the corresponding rhodopsin (Spudich et al., 2000; Ganapathy et al., 2015). In this way yields up to 20 mg/L have been reported (Ganapathy et al., 2015; Song et al., 2020). For some archaeal pigments, this straightforward approach only gave low yields and had to be adapted e.g. for bacteriorhodopsin itself (Bratanov et al., 2015; Tu et al., 2018). On the other hand, heterologous expression was more problematic for the eukaryotic type-1 rhodopsins, again because of their more complex posttranslational modification profile. Channelrhodopsins are commonly produced in yeast (Pichia pistoris), but successful production of eukaryotic type-1 pigments in insect and mammalian cell lines, Caenorhabditis elegans and Xenopus oocytes is also reported (Nagel et al., 2003; Bruun et al., 2015; Govorunova et al., 2017). An interesting new approach is using the trypanosome Leishmania tarentolae for over-expression (Volkov et al., 2017). For optogenetic applications (see below), functional production and targeting in a mammalian context is imperative, and often requires insertion of trafficking or targeting signals and/or sequence optimization to mammalian genetic code preferences.

The C-terminal His-tag has become the most popular option for purification of archaeal and eubacterial rhodopsins. For eukaryotic type-1 rhodopsins, several tags are used, including the His-tag, although the latter may sometimes interfere with particular electrophysiological or enzymatic analyses (Govorunova et al., 2021; Rozenberg et al., 2021; Tsunoda et al., 2021; Govorunova et al., 2022b).

Thanks to the powerful combination of the recombinant DNA toolbox with heterologous expression and purification making sufficient protein material available, an astounding repertoire of structural and functional data has also become available for the type-1 rhodopsins (Table 3). As a result, bacteriorhodopsin has become the best studied and fathomed membrane protein, with unprecedented insight into its structure and function (Ernst et al., 2014; Larkum et al., 2018; Nogly et al., 2018; Weinert et al., 2019). Next to that, the type-1 community has delivered prospects for a wealth of biotechnological and biomimical applications, far beyond any prognosis (see below).

For the type-2 rhodopsins ultrarapid spectroscopy data are limited, and mainly available for Opn1, Opn2, and R-type pigments (Shichida et al., 1978; Shichida, 1990; Schoenlein et al., 1991; Vought et al., 2000; Imamoto and Shichida, 2014; Schnedermann et al., 2018). Generally speaking, two schemes have been identified: Monostable pigments eventually release all-trans retinal (all Opn1 and Opn2 rhodopsins, Figure 8A) following which the opsins require supplementation with retinal re-isomerized elsewhere to regenerate the original “dark” state. Bistable pigments (most other type-2 pigments investigated, Table 2) progress until a stable M-intermediate is reached (all-trans chromophore), that requires photo-isomerization to return to the original “dark” state (11-cis chromophore) (Figure 8B) (Hillman et al., 1983; Gärtner, 2000; Stavenga et al., 2000).

The photochemical profile of the monostable bovine rod rhodopsin has been explored in great detail. The native pigment and a variety of isotopically labeled and/or mutant pigments have been investigated by femtosecond optical spectroscopy and vibrational and NMR spectroscopy. These studies have revealed intimate details on the kinetics, conformational changes in the chromophore and surrounding H-bonded networks with constrained water molecules, protein-chromophore interplay and Schiff base (de)protonation (Table 4). Overall protein conformational changes have been elucidated by fluorescence, ESR and NMR spectroscopy and TR-WAXS (DeGrip et al., 1999; Kusnetzow et al., 2006; Alexiev and Farrens, 2014; Malmerberg et al., 2015; Van Eps et al., 2017; Smith, 2021). Crystal structures have been resolved for all photo-intermediates and present a broad structural basis (Table 4). The power of theoretical and quantum-chemical calculations has grown immensely, laying a strong foundation for electronic and energetic elements of the process, in particular (Schapiro et al., 2011; Gozem et al., 2017; Schnedermann et al., 2018; Agathangelou et al., 2021; Nikolaev et al., 2021).

A very effective combination of selectively labeled chromophore with femtosecond spectroscopy and advanced quantum chemical computation resolved many remaining issues in the photoisomerization process of bovine rhodopsin (Schnedermann et al., 2018). The global picture has arisen that after photo-excitation of the chromophore into the Franck-Condon state it rapidly relaxes along a barrierless trajectory on the potential surface to a minimal energy conical intersection (Figure 8D). Here, productive resonance of the electronic wave packet at the excited state potential surface with torsional and HOOP vibrational modes in the twisted C10-C13 segment of the 11-cis chromophore, can prime very effective cross-over to a ground state energy surface, generating a hot all-transoid state (photorhodopsin) within tens of fs (Johnson et al., 2015). This relaxes thermally in about 200 fs into the photoproduct bathorhodopsin, which contains a still highly twisted all-trans chromophore, but is stable below 130 K (Yoshizawa and Wald, 1963). In free retinal, photoexcitation results in formation of several isomers (predominantly all-trans, 13-cis, 9-cis, and 11-cis), but in rhodopsins this conversion is remarkably selective from 11-cis to all-trans. This is clearly facilitated by the constraints of the binding site and the twist in the C10-C13 segment of the chromophore (Bismuth et al., 2007; Weingart, 2007; Schnedermann et al., 2018).

At room temperature, the ca 35 kcal of excitation energy stored in bathorhodopsin (Cooper, 1979) drives further relaxation via several intermediates until the metarhodopsin IIa-IIb equilibrium is reached within ms. This relaxation process subtly rearranges chromophore, protein residues and H-bonded networks up to the metarhodopsin stage, where the Schiff base transfers its proton, the counterion and another Glu at the intracellular side of the protein become protonated and an interhelical activity switch reshuffles helical segments to open up binding residues for the G-protein (Hofmann, 2000; Vogel et al., 2007; Vogel et al., 2008; Pope et al., 2020). The chromophore is subsequently slowly released via hydrolysis of the Schiff base to generate the nearly inactive apoprotein opsin (Wald, 1953; Rothschild et al., 1987; Jastrzebska et al., 2011). In vivo the active state is rapidly inactivated through phosphorylation and arrestin binding, however, which blocks activation of the G protein (Ranganathan and Stevens, 1995).

The photochemical profile of other monostable pigments (human rod rhodopsin, several cone pigments) has been investigated to much less depth, but is quite comparable to the bovine rod pigment (Barry and Mathies, 1987; Kusnetzow et al., 2001; Hofmann and Palczewski, 2015; Kazmin et al., 2015). However, the kinetics differ somewhat. For instance, the investigated cone pigments show more rapid kinetics in most steps (Imai et al., 1997; Vissers et al., 1998; Chen et al., 2012; Sato et al., 2012). Ultra-violet absorbing cone pigments may be more complex, as photoisomerization is accompanied by protonation of the Schiff base (Kusnetzow et al., 2004; Mooney et al., 2012).

The photochemical profile of bistable pigments, investigated thus far (squid, octopus and some insect pigments), follow a scheme similar to the monostable pigments up through formation of the M-intermediate and with comparable kinetics (Figure 8B) (Gärtner, 2000; Stavenga et al., 2000; Vought et al., 2000; Murakami and Kouyama, 2015). It is reported that in cephalopods the M-intermediate in fact forms a pH-dependent equilibrium between a protonated (acid M) and a deprotonated state (alkaline M). This involves the Schiff base of the chromophore, and the alkaline M is strongly blue-shifted (Liang et al., 1994; Vought et al., 2000). Photo-reisomerization of the M state to the original “dark” state is again quite efficient with a quantum yield around 0.4 (Stavenga et al., 2000).

The “dark” state of type-1 pigments contains a chromophore with the all-trans, 15-syn configuration (Figure 1). Rapid spectroscopy has been performed on quite a number of type-1 pigments, and the global scheme is quite similar to that of bacteriorhodopsin (Figure 8C). However, the kinetics of the slower steps (M and subsequent ones) and thereby the overall cycle time can vary considerably from ms up to minutes (Rozenberg et al., 2021; Tsunoda et al., 2021; Broser, 2022; Nagata and Inoue, 2022).

Out of all rhodopsins the photochemistry of BR is understood in most detail (Wickstrand et al., 2015; Nango et al., 2016). Femtosecond XFEL crystallography has even revealed very early responses to photoexcitation of the chromophore (Nogly et al., 2018). The adjacent protein residues and water molecules already react to the charge delocalization in the excited chromophore before the isomerization is initiated (Tahara et al., 2019a). During the isomerization process more of the protein environment becomes involved while the chromophore rapidly relaxes along a 2-state trajectory on the excited state potential surface to a conical intersection, where it effectively crosses in ca 500 fs over to a ground state energy surface into a “hot” transient hybrid state (J) and then relaxes thermally in about 3 ps into the photoproduct K, which contains a still significantly twisted 13-cis, 15-anti chromophore, but is stable below 150 K (Lanyi, 2004). Here, a major driving force is the elongation of the C13-C14 bond in the excited state in combination with electrostatic re-arrangement and weakening of the hydrated H-bonded network in the Schiff base region. At room temperature, the ca 15 kcal of excitation energy stored in K (this can be higher in sensory rhodopsins) (Birge et al., 1991; Govorunova et al., 2017; Rozenberg et al., 2021) drives further relaxation via the spectrally distinguishable L intermediate until the M states are reached in ca 50 µs. This relaxation process again subtly re-arranges chromophore, protein helices and H-bonded networks up to the M states, where the Schiff base transfers its proton via a water molecule to the counterion and the hydrated H-bonded network opens up a proton gateway to the extracellular membrane surface. The M-states thermally decay via several intermediates in tens of ms to the BR ground state, during which the Schiff base is reprotonated via proton transfer from residue Asp96, a proton is taken up from the intracellular surface and the chromophore is re-isomerized to the all-trans, 15-syn configuration. In fact, all-trans is the most stable configuration for free retinal (Ganapathy and Liu, 1992). Nevertheless, in some archaeal rhodopsins including BR the chromophore slowly enters an all-trans, 15-anti ↔ 13-cis, 15-syn equilibrium when stored in the dark (dark adaptation). The latter chromophore is photo-excited in the light and via a separate non-productive photocycle rerouted to the ground state BR (Smith et al., 1989; Oesterhelt et al., 1991). In channelrhodopsins the opposite phenomenon is observed, where prolonged illumination reduces the activity, since an equilibrium between pigments with an all-trans, 15-anti and a 13-cis, 15-syn chromophore configuration is generated (light-adaptation with partial desensitization) (Bruun et al., 2015; Kuhne et al., 2019; Rozenberg et al., 2021; Govorunova et al., 2022b).

Using serial synchrotron crystallography, the slower conformational changes from 5 to ca 40 ms were recorded in the BR photocycle and involve small α-helical rearrangements, chromophore re-isomerization and proton uptake, ending in formation of the ground state (Weinert et al., 2019). A very recent study using advanced high-resolution atomic force spectroscopy at the single-molecule level investigated the BR photocycle after M formation (Perrino et al., 2021). It was concluded that a cytoplasmic gate for proton uptake opens up at about 3 ms after photo-excitation lasting for about 14 ms. Surprisingly, this same study observes a “black-out period” of tens of ms before a recycled ground state can be photo-reactivated. This uncovers a very interesting new phenomenon reminiscent of comparable nonresponsive states in animal voltage-regulated channels (Armstrong, 1992). Meanwhile, XFEL studies have also been performed on other ion pumps and channels. A femtosecond XFEL study of the sodium-pumping rhodopsin from Krokinobacter eikastus (KR2) again observed photo-isomerization of the chromophore to start in the femtosecond range and completed within 2 ps (Skopintsev et al., 2020). Changes in the local structure of the binding site and early conformational changes in the protein backbone are observed in the early nanosecond range. Further subtle rearrangements result in Schiff base deprotonation in µs and in the early ms range a gate opens up and transient binding of a Na ⁺ ion in the vicinity of the Schiff base is observed with release within 20 ms. A femtosecond XFEL study of the chloride pump from the flavobacterium Nonlabens marinus follows the conformational adaptations between 1 and 100 ps after photo-excitation (Yun et al., 2021). It shows the final rearrangements of the chromophore to the 13-cis configuration within 50 ps, together with the dynamics of the hydrated H-bonded network and deformations in the local α-helical elements. Following chromophore isomerization the chloride ion first dissociates from the protonated Schiff base and then starts to diffuse away. Additional molecular details of the interactions and trajectory of the chloride ion are provided by recent ps up to ms studies using time-resolved serial crystallography in combination with spectral and theoretical analysis (Hosaka et al., 2022; Mous et al., 2022). An XFEL study of the channelrhodopsin chimera C1C2, that photochemically behaves like ChR1, investigated the photo-induced conformational changes from 1 µs to 4 ms (Oda et al., 2021). Photo-isomerization induces a kink in the chromophore structure, triggering shifts in the retinal binding lysine residue and TM7, starting at around 1 µs and increasing during formation of the M-state up to 4 ms. This induces small lateral shifts of the chromophore and in TM7 and TM3 at around 50 µs. It is postulated that these rearrangements forebode the subsequent opening of the gates in the cation channel pore, although these were not observed in the crystal. The XFEL and serial crystallography studies beautifully illustrate the powerful but subtle design and the broad potential of the photo-driven nanomachinery. Less detailed studies basically show a similar pattern (Table 4). Subtle differences in early kinetics and conformational adaptation in chromophore and adjacent protein elements following photo-excitation are observed in the ultrarapid studies. A cautious interpretation could be that the structure of the hydrated H-bonding network in the complex counterion is an important roadmap for the light-triggered protein activity, which also depends on the pKa of the direct counterion (Hontani et al., 2017b; Oda et al., 2021; Chang et al., 2022).

In this context it should be realized that crystal structures have their limitations (García-Nafría and Tate, 2020; Guo, 2020). Detergent exposure may affect elements of the protein structure, and the crystal will certainly constrain larger conformational alterations in the protein, which may occur in the slower phase of the photocycle (Weinert et al., 2019; Oda et al., 2021; Govorunova et al., 2022b). Hence, it would be preferable to study the slower photocycle phases with experimental approaches that can handle membrane-bound systems as shown in Figure 4, like time-resolved AFM, cryo-EM and vibrational spectroscopy.

The general scheme for the photocycle of BR (Figure 8C) also holds for other type-1 pigments, though the kinetics after M formation can vary significantly (Wand et al., 2013; Tahara et al., 2015; Han et al., 2020; Smitienko et al., 2021). The decay is much slower for sensory rhodopsins, enzyme-rhodopsins and heliorhodopsins, possibly since longer interaction with their cognate partner is required for regulated signal transduction. In fact, some sensory rhodopsins and enzyme-rhodopsins exhibit a bistable photocycle (Kawanabe et al., 2007; Broser et al., 2020) and proton transfer to the counterion may not occur (Bergo et al., 2006).