Vitamin A₁/A₂ chromophore exchange: its role in spectral tuning and visual plasticity

Abstract

Vertebrate rod and cone photoreceptors detect light via a specialized organelle called the outer segment. This structure is packed with light-sensitive molecules known as visual pigments that consist of a G-protein-coupled, seven-transmembrane protein known as opsin, and a chromophore prosthetic group, either 11-cis retinal (‘A₁’) or 11-cis 3,4-didehydroretinal (‘A₂’). The enzyme cyp27c1 converts A₁ into A₂ in the retinal pigment epithelium. Replacing A₁ with A₂ in a visual pigment red-shifts its spectral sensitivity and broadens its bandwidth of absorption at the expense of decreased photosensitivity and increased thermal noise. The use of vitamin A₂-based visual pigments is strongly associated with the occupation of turbid aquatic habitats in which the ambient light is red-shifted. By modulating the A₁/A₂ ratio in the retina, an organism can dynamically tune the spectral sensitivity of the visual system to match the predominant wavelengths of light in its environment. As many as a quarter of all vertebrate species utilize A₂, during part of their life cycle or under certain environmental conditions. A₂ utilization therefore represents an important and widespread mechanism of sensory plasticity. This review provides an up-to-date account of the A₁/A₂ chromophore exchange system.

Graphical Abstract

Introduction

Vertebrate rod and cone photoreceptors mediate vision in dim- and bright-light environments, respectively (Fig 1A,B). The light-sensitive molecule of a photoreceptor cell, known as the visual pigment, consists of two components: a G-protein-coupled, seven-transmembrane apo-protein known as opsin, and a chromophore prosthetic group bound via a Schiff base linkage to the side chain of a lysine within the opsin’s chromophore binding cleft (Fig. 1B). In vertebrates, two different chromophores are found: 11-cis retinal (derived from vitamin A₁ and henceforth referred to as ‘A₁’) and 11-cis 3,4-didehydroretinal (derived from vitamin A₂ and henceforth referred to as ‘A₂’)^11,12. The only difference between the two chromophores is the presence of an additional double bond within the β-ionone ring of A₂ (Fig. 1C). Replacing A₁ with A₂ in a visual pigment has four main effects: (1) it red-shifts the spectral absorption curve of the visual pigment; (2) it broadens the spectral bandwidth of absorption; (3) it decreases the pigment’s photosensitivity; and (4) it increases thermal noise (Fig. 1D–F)^12,14. The wavelength of maximal sensitivity of a visual pigment (referred to as λ_max) can be tuned toward shorter or longer wavelengths via two primary mechanisms: changes in the amino acids of the opsin; or exchange of one chromophore for the other. Opsin tuning via amino acid replacement has become an important model system for the study of molecular evolution in recent years^15–21. In contrast, nearly a half century has elapsed since publication of the most recent comprehensive review of the vitamin A₁/A₂ chromophore exchange system¹². The present review therefore aims to provide an up-to-date account of our understanding of the A₁/A₂ system and how it impacts spectral tuning and visual plasticity.

Figure 1. The vitamin A₁/A₂ chromophore exchange system.

(A) H&E-stained histologic section of adult zebrafish retina. The photoreceptor cell bodies reside in the outer nuclear layer (ONL). INL = inner nuclear layer; GCL = ganglion cell layer; RPE = retinal pigment epithelium. (B) Drawing of rod and cone photoreceptor subtypes of adult zebrafish: UV = ultraviolet cone; B = blue cone; G = green cone; and R = red cone. Also shown is the crystal structure of bovine rhodopsin (RH1) with the 11-cis retinal (A₁) chromophore in red. PDB code = 1F88¹. (C) Chemical structure of 11-cis retinal (A₁) and 11-cis 3,4-didehydroretinal (A₂). Note the position of the additional double bond (in red) within the terminal β-ionone ring of A₂. (D) Photosensitivity curves of a typical LWS visual pigment either with A₁ (λ_max = 565 nm) or A₂ (λ_max = 606 nm). Curves are based on templates in Govardovskii et al.⁶ Note that the λ_max of the A₂ pigment is red-shifted by 41 nm relative to the A₁ form in accordance with the formula in Fig. 3A. In addition, the photosensitivity of the A₂ pigment is only ~70% that of the A₁ form. (E) A₂-based visual pigments have a wider bandwidth than A₁ forms as shown here by superimposing the absorbance curves of A₁ and A₂ pigments with identical λ_max = 565 nm. Dotted line indicates half-maximal absorbance. (F) This figure shows the noise power spectral density from electrical recordings of the light-sensitive current of individual larval tiger salamander (Ambystoma tigrina) rods either in their native A₂-predominant form (red trace; A₁:A₂ ratio is ~0.26:0.74) or after regeneration with A₁ (black trace; A₁:A₂ ratio is ~0.91:0.09)¹⁰. Note that the A₂-based visual pigment is noisier than the A₁ form. Photoreceptor ‘dark noise’ has discrete (low frequency) and continuous (high frequency) components. Discrete noise results from thermal isomerization events which occur with greater frequency in A₂-based visual pigments. Continuous noise arises in components of the phototransduction cascade downstream of the visual pigment and occurs at similar rates in A₁- and A₂-based pigments. The inset shows recordings of a rod in the ‘A₂’ state (red trace) and the ‘A₁’ state (black trace). Note the large, lower-frequency deviations in the A₂ trace that are absent from the A₁ recording. Panel F is adapted, with permission from the author, from reference¹⁰.

The discovery of vitamin A₁/A₂ chromophore exchange

Scientists first studied purified extracts of vertebrate visual pigments toward the end of the 19^th century. They found that solutions of visual pigments obtained from aerial and terrestrial species had a striking rose-pink color which rapidly bleached upon exposure to light. This visual pigment was named ‘rhodopsin’ based on the ancient Greek words, ῥóδoν, rose, and ὄψις, vision²². In 1880, Kühne and Sewall first noticed that visual pigment extracts from some species of fish were not rose-colored but purple²⁵. This new visual pigment was subsequently named ‘porphyropsin’, based on the Greek term for the gastropod mollusks used in antiquity to produce ‘royal purple’ dye (πορφύρα, purple-fish, and ὄψις, vision)^11,26,27. In 1896, Köttgen and Abelsdorff showed that the absorption spectra of visual pigments derived from reptiles, birds, and mammals peaked around 500 nm, whereas those of freshwater fish were red-shifted, peaking around 540 nm²⁸. George Wald later demonstrated that ‘rhodopsins’ contain an A₁ chromophore whereas ‘porphyropsins’ contain A₂¹¹. He also showed that replacement of A₁ with A₂ in the same opsin produces a red shift in the λ_max of the visual pigment¹¹. While the terms ‘rhodopsin’ and ‘porphyropsin’ were once used to refer to any visual pigment containing A₁ and A₂, respectively, the subsequent discovery of cone opsins made this terminological distinction obsolete. The term ‘rhodopsin’ is now only used to refer to ‘rod opsins’ (i.e., RH1 opsins), whereas ‘porphyropsin’ is rarely used.

Species distribution and uses of vitamin A₁/A₂ chromophore exchange

The use of vitamin A₂-based visual pigments is strongly associated with the occupation of turbid aquatic habitats, particularly ones with red-shifted or highly variable ambient light¹². While A₂-based visual pigments have never been identified in any species of bird, mammal, or fully terrestrial reptile (with the exception of two lizards, Anolis carolinensis and Podarcis siculus²⁹), these pigments are widely distributed among fishes, amphibians, aquatic reptiles, and lamprey¹². Indeed, data suggest that the vast majority of freshwater fishes utilize A₂-based visual pigments³⁰. While early studies showed A₂ to be common in diadromous species (i.e., salmon, trout, eels, and lamprey) during the freshwater phase of their life cycle^{12,26,31–34}, they found the use of A₂ to be rare among fully marine species. Subsequent studies, however, have demonstrated the use of A₂-based visual pigments in marine fishes from multiple families^{30,33,35–42}, especially nearshore species inhabiting spectrally variable environments³⁷. It is therefore unwise to assume a priori that a marine species does not utilize A₂-based visual pigments. In total, it is likely that more than a quarter of all vertebrate species utilize A₂-based visual pigments, at least during a part of their life cycle or under certain environmental conditions.

The vitamin A₁/A₂ chromophore exchange system is sometimes referred to as a ‘switch’, implying that organisms toggle between the use of one chromophore or the other in an all-or-none fashion. In fact, many species utilize both chromophores simultaneously, adjusting the A₁/A₂ ratio in response to physiological or environmental cues in a continuous manner¹². The presence of both A₁- and A₂-based visual pigments in a single outer segment endows a photoreceptor with a broad, unimodal spectral response curve with a λ_max intermediate between those of the pure A₁- and A₂-based pigments¹². Thus, by adjusting the A₁/A₂ ratio, the organism can continuously tune λ_max on a physiological time scale. It appears that some species (e.g., the ninespine stickleback, Pungitius pungitius) can independently tune the A₁/A₂ ratio in rods and cones and even within cone subtypes in the same retina⁴³. Interestingly, A₁/A₂ ratio can vary among individual fish of a single species caught at the same time and place^32,44,45. One author suggested that this phenomenon is particularly notable in schooling species and proposed that the broadening of spectral sensitivity of the school as a whole by individually variable A₁/A₂ ratios might confer a selective advantage in detecting predators^12,44.

Fishes inhabiting clear waters tend to utilize A₁-based visual pigments, while those found in more turbid environments tend to have a high proportion of A₂ in their eyes. However, classifying fish species as either ‘A₁’ or ‘A₂’ is overly simplistic, because most published reports describe fish collected at a single location at only one time of the year, and thus ignore potential temporal dynamics of chromophore usage. For example, most published studies (with one exception⁴⁶) indicate that zebrafish (Danio rerio) almost exclusively utilize A₁-based pigments under standard laboratory conditions^47,48. Yet, the zebrafish’s native streams and ponds in India and Bangladesh are likely subject to conditions of widely varying turbidity, particularly during the monsoon season^49–51. It is possible, and even likely, that zebrafish utilize A₂-based visual pigments in the wild when they encounter highly turbid environments. Indeed, laboratory studies have demonstrated that application of thyroid hormone (TH) to the water of zebrafish can induce a complete switch to A₂-based visual pigments^47,48, revealing a latent capacity to synthesize A₂. Clearly, one cannot rule out the use of A₂-based visual pigments by a given species based on examination of individuals caught at one time or in a single locale.

Species utilize A₁/A₂ exchange to fine-tune spectral sensitivity in accordance with season, migration status, and developmental stage^12,52. Longitudinal studies have shown that A₂ levels tend to be highest during the winter months and lowest during the summer, even at a single location^32,53–55. Some have speculated that the increase in A₂ during the winter may be a response to red-shifting of the ambient light spectrum caused by decreased solar elevation^12,32. Alternatively, colder water temperatures may play a role since lower temperatures are known to increase A₂ levels in several species^34,56–63. Colder habitats might also favor A₂ usage since the lower temperature mitigates the increased thermal noise associated with A₂-based pigments (see below)⁶⁴. As mentioned above, migratory species (salmon, trout, eels, and lamprey) alternate between A₁ and A₂-based pigments, shifting to A₁ upon entering clear marine environments, and favoring A₂ upon entering more turbid, inland waterways^{12,26,31–34,65}. Some species switch between A₁ and A₂ according to developmental stage^42,66–68. For example, the northern leopard frog (Lithobates pipiens) utilizes A₂-based pigments during the aquatic tadpole stage and then switches to A₁ upon metamorphosis into a semiterrestrial adult⁶⁶. In contrast, amphibians that remain aquatic as adults, such as the African clawed toad (Xenopus laevis), appear to have A₂-predominant retinas throughout life¹². One study reported an increasing proportion of A₂ in the eyes of older individuals of the common rudd (Scardinius erythrophthalmus) under controlled lighting conditions⁶⁹, but the relationship between A₂ levels and age has not been examined in other species.

Several species use A₂-based pigments for specialized purposes. For example, unlike most anurans, the American bullfrog (Lithobates catesbeianus) retains A₂-based pigments in the dorsal third of its retina, even as an adult⁷⁰. Bullfrogs spend considerable time with their eyes positioned right above the surface of the water⁷¹. Thus, the A₂-rich dorsal retina enables downward vision into the turbid, red-shifted aquatic environment, while the A₁-rich ventral retina scans the aerial milieu. Similarly, the four-eyed fish (Anableps anableps) inhabits the surface of turbid waterways, viewing the aquatic and aerial environments with its dorsal and ventral retinas, respectively^72,73. Despite marked morphologic differences between the dorsal and ventral retina and the expression of red-sensitive LWS opsin exclusively in the dorsal retina⁷³, the distribution of A₂ in the four-eyed fish’s retina is currently unknown⁷⁴. One might speculate that the LWS opsin expressed in the dorsal retina is likely to be paired with an A₂ chromophore to enhance aquatic vision. In contrast, some other fishes have been reported to have a higher proportion of A₂ in the ventral retina, indicating that intraretinal distribution of A₂ is species-specific^75,76.

A specialized optical adaptation in certain shallow-water, nearshore fishes favors the use of A₂. More than 100 species of fish can reversibly pigment their corneas upon exposure to bright light⁷⁷. Corneal pigmentation is mediated by the movement of yellow and red carotenoid-containing organelles within specialized chromatophores whose processes extend across the pupil⁷⁷. While the adaptive function of corneal pigmentation is debated and might differ between species^78,79, in the masked greenling (Hexagrammos octogrammus), pigment density can be so high that the cornea acts as a long-pass spectral filter, precluding the transmission of light less than 520 nm⁷⁹. This species has adopted the use of A₂-based visual pigments to red-shift their medium- and long-wavelength-sensitive opsins and thereby enhance their ability to detect the longer wavelengths that are passed by the corneal filter, even in summertime when A₂ levels in most species are low^39,79.

Another interesting optical adaptation found in three genera of deep-sea dragonfish (Stomiidae) involves the use of A₂-based visual pigments. In addition to the blue-green (450-500 nm) bioluminescent signals typically emitted by deep-sea fishes⁸⁰, these dragonfishes emit a far-red (>700 nm) signal from periorbital photophores^81–83. Since most deep-sea fishes are blind to long-wavelength light, far-red emission endows dragonfishes with the potential for covert prey illumination or ‘private’ signaling between individuals⁸⁴. To detect this signal, dragonfishes have evolved rod opsins (RH1) with red-shifted λ_max⁸⁵. These fishes couple their red-shifted opsin with an A₂ chromophore in a sub-population of rods, thereby enhancing sensitivity to far-red signals^85–88. Remarkably, the dragonfish Malacosteus niger red-shifts sensitivity even further by using derivatives of bacteriochlorophyll as photosensitizers^84,89,90.

Environmental factors controlling the A₁/A₂ ratio

What accounts for the widespread utilization of A₁/A₂ chromophore exchange among aquatic organisms and its rarity among terrestrial species? The most likely answer is the variable and labile quality of light in aquatic, and especially freshwater, habitats (Fig. 2). Light transmission in water can be affected by both biotic and abiotic factors (e.g., chlorophyll-containing plankton, suspended inorganic particles, and dissolved compounds). These factors modify both the amount and spectral distribution of light available for vision (Fig. 2B). Opsin switching (i.e., changes in the expression of opsin genes) and A₁/A₂ chromophore exchange are the primary mechanisms whereby species modify their spectral sensitivity in response to changes in their photic environment¹⁵. It is therefore not surprising that evolution has favored the emergence of both A₁/A₂ chromophore exchange and highly diverse opsin gene repertoires in fishes^12,15,91. Despite the established role of water temperature in controlling A₁/A₂ ratio in some fishes^34,56–62, there is broad consensus that light intensity, duration (i.e., day length), and wavelength are the most important environmental variables determining A₁/A₂ ratio in the majority of species¹². Environments with less light of shorter duration and longer wavelengths tend to favor an increase in A₂ levels¹². Indeed, emerging evidence from the evolutionarily diverse cichlid fishes (Cichlidae) indicates that selection for higher A₂ levels has likely played a key role in adaptation to turbid, red-shifted environments (Fig. 2B,C)^{13,24,92–95}.

Figure 2. Light in aquatic habitats is highly variable.

(A) Aquatic species experience widely varying and labile photic environments in which the transmission of light is modulated by both suspended and dissolved matter in the water column. Light variability was likely a major impetus for the evolution of the A₁/A₂ chromophore system. D = downwelling light; S = sidewelling light (i.e., the horizontal visual field); U = upwelling light. Photo credits^2–5. (B, C) Spectral irradiance in two Central American cichlid habitats with marked differences in water clarity. Turbidity decreases the amount of light available for vision and preferentially absorbs shorter wavelengths, effectively red-shifting the spectral distribution. A recent study showed that expression of cyp27c1 (the enzymes that converts A₁ into A₂) in the eyes of cichlids correlates with the spectral distribution of light in these habitats¹³. (D, E) The spectral absorbance curves of the A₁ (D) and A₂ (E) forms of the four cone visual pigments of the goldfish, a typical tetrachromatic teleost. The curves are based on templates in Govardovskii et al.⁶, using the following values for λ_max from reference²³: A₁ forms (370.1, 447.2, 515.9, 565.9 nm); A₂ forms (381.9, 454.1, 534.9, 617.5 nm). The irradiance curves from panels B and C (gray) are superimposed on the spectral absorbance curves in D and E, respectively. Note how the λ_max of the A₁ and A₂ forms of the red cone (LWS) pigment are well-positioned to capture the predominant wavelengths of light in clear and turbid habitats, respectively. This figure demonstrates how switching between A₁ and A₂ allows an organism to tune its spectral sensitivity to match the predominant wavelengths in its environment. The spectral irradiance curves in panels B and C are adapted, with permission of the author, from reference²⁴.

Given the outsized importance of light quality and quantity in determining opsin and A₁/A₂ usage, it is surprising that so few studies of fish vision have included ambient light measurements. In studies that do incorporate such measures, the data are often limited to quantification of up- or downwelling light at varying depths^{13,36,43,96,97}. While certainly laudable, such studies do not typically measure the light that actually reaches the animal’s eye. Yet, in surfperches (Embiotocidae) photoreceptor spectral tuning most strongly correlates with sidewelling irradiance (i.e., the horizontal visual field), not up-or downwelling light (Fig. 2A)³⁷. Since changes in spectral tuning via opsin or chromophore exchange occur over days to weeks^12,32, these changes likely reflect cumulative light exposure at the retinal surface. Ideally, studies of spectral sensitivity should include a ‘fish eye’ view of light⁹⁸, summated over time via a miniature head-mounted camera or spectrophotometer, perhaps with special weighting of visual features critical to organismal fitness (e.g., the reflectance spectra of predators or potential mates⁹⁹). Such an approach would likely reveal much stronger correlations between spectral tuning and light exposure than have heretofore been observed using more indirect measures of ambient light.

The enzymatic mechanism and transcriptional control of vitamin A₁-to-A₂ conversion

The existence of an enzyme that converts vitamin A₁ into A₂ was proposed more than a half century ago¹². Early studies showed a strong correlation between A₂ levels in retina and in retinal pigment epithelium (RPE)^63,70,100, suggesting that A₂ might be synthesized in the RPE and then passed to the retina during the visual cycle¹⁰¹. To identify the enzyme mediating A₂ synthesis, my lab used RNA-seq to compare the transcriptomes of RPE from TH-treated zebrafish vs. untreated controls as well as the transcriptomes of dorsal vs. ventral bullfrog RPE⁴⁷. We identified a single gene that was both upregulated in TH-treated zebrafish RPE and enriched in the dorsal bullfrog RPE, the cytochrome P450 family member, cyp27c1⁴⁷. P450 enzymes are involved in the metabolism of a wide range of xenobiotic compounds and endogenous small molecules, including retinoids¹⁰². Thus, cyp27c1 was an excellent candidate for the long-hypothesized ‘vitamin A₁ 3,4-dehydrogenase’. Subsequent analysis demonstrated that this enzyme is localized to the RPE in zebrafish and American bullfrog, and that its expression correlates with the presence of A₂⁴⁷. We also showed that cyp27c1 is sufficient to convert vitamin A₁ and its congeners into their corresponding A₂ forms^47,103. Lastly, we engineered zebrafish with mutations in cyp27c1 and showed that the gene is required for endogenous synthesis of A₂⁴⁷. Knock-out of cyp27c1 also eliminates the zebrafish’s ability to red-shift its photoreceptor spectral sensitivity in response to TH treatment and reduces its ability to see and respond to near-infrared light of 770 nm⁴⁷. Some fishes display differences in the A₁/A₂ ratio between photoreceptor classes, suggesting that cyp27c1 might be differentially expressed in individual photoreceptor subtypes in these species, rather than exclusively in the RPE^43,86. Differences in the A₁/A₂ ratio between rods and cones might also be accounted for by expression of cyp27c1 in Müller glia which support cone, but not rod, pigment regeneration¹⁰⁴. Interestingly, there does not appear to be an enzyme that converts A₂ into A₁. Instead, a switch from A₂ to A₁ likely occurs via progressive turnover of the retinoid pool in the RPE.

Despite the apparent absence of A₂-based visual pigments in some groups (i.e., birds and mammals), orthologs of cyp27c1 are found in all vertebrate classes. Expression of cyp27c1 strongly correlates with the presence of A₂ in the retina of the sea lamprey (Petromyzon marinus)¹⁰⁵, an agnathan that diverged from jawed vertebrates during the Cambrian period ~500 million years ago¹⁰⁶. The sea lamprey switches between A₁- and A₂-predominance at different stages of its migratory life cycle³¹, suggesting that the capacity for A₂ production may have facilitated the initial invasion of turbid inland waterways by early vertebrates^105,107. As expected, cyp27c1 orthologs are nearly ubiquitous among the sequenced genomes of fishes, amphibians, and reptiles. More surprisingly, nearly all sequenced bird genomes also retain an intact copy of cyp27c1. The role of this enzyme in birds is currently unknown, but studies suggest potential functions outside of the eye. For example, 3,4-didehydroretinoic acid (a derivative of vitamin A₂), is the predominant form of ‘retinoic acid’ found in the developing chicken (Gallus gallus) embryo^108,109.

Orthologs of cyp27c1 are present in most mammalian species including humans, but the gene appears to have been lost in three groups. BLAST searches using human CYP27C1 as a query revealed loss of cyp27c1 in bats (Chiroptera), rodents (Rodentia) with the exception of squirrel-related clades (Sciuromorpha), and Afrotheria (with the possible exception of manatees). The retention of cyp27c1 orthologs among squirrel-like clades suggests early evolutionary branching of Sciuromorpha prior to cyp27c1 loss, a finding consistent with recent phylogenetic studies^110,111. Interestingly, cyp27c1 orthologs appear to be absent from all sequenced Afrotheria genomes, with the exception of that of the West Indian manatee (Trichechus manatus) which retains a gene encoding a protein with ~68% amino acid identity to human CYP27C1 and with shared synteny (BIN1 — CYP27C1 — ERCC3). The retention of a possible ortholog of cyp27c1 in West Indian manatee is intriguing, because this species inhabits turbid coastal waters, estuaries, and rivers and, along with the three other species of sea cow (Sirenia), represents the only fully aquatic sub-clade within Afrotheria¹¹². This finding raises the possibility that sea cows (and perhaps other aquatic mammals inhabiting turbid water such as river dolphins) might use A₂-based visual pigments. The presence of cyp27c1 orthologs in terrestrial mammals is more puzzling, but the enrichment of both CYP27C1 transcripts and 3,4-didhydroretinoids in human skin^113–116, and the ability of human CYP27C1 to convert vitamin A₁ into A₂^103,113, suggest a role for this enzyme in the integument.

Organisms dynamically modulate A₁/A₂ ratio in response to both physiological changes and environmental variables. In some species, such as salmon preparing to migrate, the initiation of the A₁-to-A₂ switch precedes the fish’s entry into the new photic environment^12,32, suggesting that chromophore exchange is part of a suite of anticipatory physiologic changes and is therefore likely under systemic hormonal control. In other species, the A₁/A₂ ratio can be modulated locally within the eye. Bridges and Yoshikami showed that when held in constant darkness the common rudd converts nearly all chromophore to A₂^69,100. Upon re-exposure to light the fish then reverts to A₁. This reversion can be prevented by placing an opaque plastic cap over one eye, while the uncapped eye reverts normally^69,100. Thus, in the rudd, changes in the photic environment can be sensed locally within a single eye and transduced into changes in the A₁/A₂ ratio independent of the other eye.

How do animals sense changes in their internal state or external milieu and transduce this signal into changes in cyp27c1 expression? The answer to this question is currently unknown, but TH signaling appears to play a role, at least in some species. Studies in salmon, trout, zebrafish, goldfish (Carassius auratus), and shiners (Richardsonius balteatus, Notemigonus crysoleucas, and Laxilus cornutus) indicate that application of TH increases the percentage of A₂ in the eye^34,117–122, while in sunfish (Lepomis sp.) and American bullfrog it has the opposite effect^67,123–125. To identify the transcription factors that mediate induction of cyp27c1 in response to TH, my lab assayed zebrafish with mutations in the three known TH nuclear receptors (thraa, thrab, and thrb). We found that no single TH nuclear receptor is required for TH-mediated induction of cyp27c1 but that deletion of all three completely eliminates cyp27c1 expression and the resulting conversion of A₁ to A₂¹²⁶. Despite this knowledge, we still do not understand the mechanism whereby some species flip the polarity of the response, reducing A₂ levels upon exposure to TH^67,123–125. We also do not know how changes in the light environment are sensed and transduced into changes in TH signaling in the RPE. These are important problems for future work.

In addition to controlling cyp27c1 expression, TH signaling is required for red-sensitive LWS opsin expression in many vertebrates^126–129. In zebrafish, mutations in thrb cause LWS cone precursors to be transfated into UV cones¹²⁶. TH signaling also appears to play a role in controlling expression of paralogous opsin genes, possibly in response to changes in the photic environment. Temple and colleagues showed that TH treatment of coho salmon (Oncorhynchus kisutch) can induce increased expression of a RH2 paralog with red-shifted λ_max¹³⁰. Another study demonstrated that TH treatment of zebrafish induces a shift in expression toward red-shifted RH2 and LWS paralogs¹³¹. Taken together, these findings suggest that TH signaling coordinates a multi-level response to changes in long-wavelength light in the environment. One might speculate that in early vertebrate evolution both A₁/A₂ exchange and the expression of red-shifted opsins came under the control of TH signaling as a mechanism of coordinating physiologic changes, perhaps in a jawless ancestor undergoing metamorphosis or in one preparing to migrate into fresh water.

The extent of red shift upon switching from A₁ to A₂ is correlated with λ_max

One of the most remarkable features of the A₁/A₂ system is that the longer the λ_max-A1, the greater the red shift upon switching to A₂ (Fig. 3A,B)^{12,66,132,133}. For example, in an early study of the northern leopard frog using microspectrophotometry (MSP)⁶⁶, the authors found that the LWS pigment (λ_max-A1 = 575 nm) underwent a red-shift of 45 nm upon switching from A₁ to A₂, whereas the RH1 pigment (λ_max-A1 = 502 nm) underwent a red shift of 25 nm and the SWS2 pigment (λ_max-A1 = 432 nm) a red shift of only 6 nm. The authors noted that these data points fall exactly on a straight line when graphed against λ_max-A1. Multiple subsequent studies have confirmed this linear relationship^23,43,48,134, at least for λ_max-A1 > 407 nm (Fig. 3A,B). The functional consequence of this relationship is that switching from A₁ to A₂ results in a large extension of visual sensitivity into the far-red region without much of a corresponding loss of sensitivity at the short-wavelength end of the spectrum (Fig. 2D,E).

Figure 3. The relationship between λ_max-A1 and λ_max-A2.

(A) This graph shows the relationship between λ_max-A1 and λ_max-A2 for 57 different rod (RH1) and cone (SWS1, SWS2, RH2, LWS) visual pigment pairs from numerous species (primary data and references in Supplemental Table S1). The relationship is well described by a straight line (R² = 0.985; equation for the fitted line is shown). The dotted lines indicate the point at which λ_max-A1 and λ_max-A2 are equal (~407 nm). (B) The same data as in panel A but with λ_max-A1 plotted against λ_max-A2 - λ_max-A1 to highlight how the red shift increases with increasing λ_max-A1. The marked deviation of the one SWS1 pigment (purple) is discussed in the main text. Values for LWS pigments (red) also appear to deviate somewhat from the fitted line. It is not currently known whether this deviation is real or attributable to measurement errors. (C) The retinal chromophore consists of three planes (A, B, C) which can rotate relative to each other. Rotation about the C6-C7 bond (dihedral angle α) alters the degree of co-planarity between planes A and B, thereby modifying the extent of electron delocalization into the β-ionone ring. Blatz and Liebman have proposed that the relationship between λ_max-A1 and λ_max-A2 might be explained by differences in α across visual pigments^4,5,7. (D) A theoretical modeling analysis suggested that λ_max can be tuned over a range of >100 nm by changing α, the dihedral angle between planes A and B^8,9. Panel D is adapted from reference⁸.

A notable implication of the linear relationship between λ_max-A1 and λ_max-A2 shown in Fig. 3A is that the red shift caused by an A₁-to-A₂ switch is predicted to be zero when the λ_max-A1 equals ~407 nm. Two published datasets are consistent with this prediction. First, an MSP study of zebrafish (Danio rerio) found no statistically significant red shift of the λ_max of the SWS2 pigment (λ_max-A1 = 411 nm) upon switching to A₂ (λ_max-A2 = 412 nm)⁴⁸. The authors did, however, observe that the half-bandwidth of the absorption curve of the A₂-based SWS2 pigment was broader than that of the A₁-based, indicating that chromophore exchange had indeed occurred (Fig. 1E). In a second study, Makino and colleagues measured the λ_max of the SWS2 visual pigment of the tiger salamander (Ambystoma tigrinum) reconstituted with three different 9-cis retinals: 9-cis 5,6-dihydroretinal, which lacks double bonds in the β-ionone ring and is referred to by the authors as A₀, 9-cis retinal (referred to as A₁), and 9-cis 3,4-didehydroretinal (referred to as A₂)¹³⁵. As expected from the equation in Fig. 1A, the authors observed only very small red shifts of the SWS2 pigment upon addition of one or two double bonds to the β-ionone ring of the chromophore (λ_max-A0 = 415 nm; λ_max-A1 = 418 nm; λ_max-A2 = 422 nm). Although more data are needed to precisely define the λ_max-A1 value at which the red shift equals zero, the published data strongly suggest that such a point exists near 407 nm.

A counterintuitive prediction of the equation in Fig. 3A is that at λ_max-A1 < 407 nm, A₂-based pigments should absorb at shorter wavelengths than their A₁ counterparts! For example, an SWS1 pigment with λ_max-A1 = 370.1 nm would be predicted to have λ_max-A2 = 360.5 nm. Instead, measurements of the goldfish SWS1 pigment (λ_max-A1 = 370.1 nm) indicate a red shift of nearly 12 nm upon switching to A₂ (λ_max-A2 = 381.9 nm)²³. This apparent deviation from linearity has led some authors to suggest that A₁/A₂ data would be better fit by a non-linear (e.g., parabolic) function^23,136. Yet, careful measurement of multiple A₁/A₂ pairs within individual labs have almost always found a nearly perfect linear relationship between λ_max-A1 and λ_max-A2 (at least for λ_max > 407 nm)^23,48,66, suggesting that deviations from linearity are more likely due to differences in technique between laboratories or experimental error. How can we reconcile these conflicting viewpoints?

I propose that deviations from linearity at very short λ_max might be due to fundamental differences in the mechanisms of spectral tuning used by visual pigments with an unprotonated Schiff base and those with a protonated Schiff base^136–138. It has long been known that all SWS2, RH1, RH2, and LWS visual pigments contain a protonated Schiff base linkage^8,139,140. Furthermore, the presence of this positive charge on the chromophore and the distance of the counterion within the binding cleft play a major role in tuning the λ_max of the visual pigment^8,141–144. In contrast, SWS1-based visual pigments fall into two distinct spectral classes: those with λ_max-A1 < 400 nm (i.e., ultraviolet-sensitive) and those with λ_max-A1 > 400 nm (violet-sensitive)^145–148. Visual pigments in the former class contain an unprotonated Schiff base, whereas those in the latter class have a protonated Schiff base^17,136. Given the fundamental role played by protonation in defining the electronic state of the chromophore and consequently its spectral tuning, it is reasonable to conclude that a visual pigment with an unprotonated Schiff base might be tuned differently¹³⁷.

I therefore suggest that the relationship between λ_max-A1 and λ_max-A2 is linear for all visual pigments with a protonated Schiff base, and that a different, and currently unknown equation describes the relationship between λ_max-A1 and λ_max-A2 for visual pigments with an unprotonated Schiff base. At the present time, there are very few high-confidence measurements of A₁/A₂ pairs of visual pigments with an unprotonated Schiff base. In fact, the above-mentioned study of the goldfish SWS1 pigment is the only high-quality data point I have found in this range²³. Another study purporting to analyze A₁/A₂ pairs of ultraviolet pigments relied on paired A₁ and A₂ values derived from different species of fish (under the unproven assumption that the opsins were identical) or from the absorption spectra of pure A₁ and A₂ chromophores dissolved in ethanol¹³⁶. Clearly, to define the relationship between λ_max-A1 and λ_max-A2 in the ultraviolet region, more high-quality measurements of A₁/A₂ pairs are needed, especially in the range λ_max-A1 = 350-385 nm.

Why is the extent of red shift upon switching from A₁ to A₂ linearly correlated with λ_max in visual pigments with a protonated Schiff base? A definitive answer to this question is not yet available and will depend on a deeper understanding of the physical mechanisms of visual pigment spectral tuning. Spectral tuning largely depends on the presence/absence of a protonated Schiff base and the extent of π-electron delocalization along the polyene chain and into the β-ionone ring of the chromophore^{141–144,149}. A greater number of conjugated double bonds (as in A₂-based visual pigments) and a greater extent of π-electron delocalization result in greater red shifts in λ_max^7,8,138,150. Charge delocalization is modulated by both electrostatic interactions between the chromophore and the amino acid side chains within the opsin, as well as by steric interactions that distort the geometry of the chromophore, in turn, affecting its electronic state^{8,138,141–144}.

Nearly fifty years ago, Blatz and Liebman proposed a simple mechanism to account for the relationship between λ_max-A1 and λ_max-A2⁷. They suggested that the extent of π-electron delocalization into the β-ionone ring is modulated by the degree of co-planarity between the plane of the β-ionone ring (plane A in Fig. 3C) and that of the polyene chain (plane B in Fig. 3C). When the two planes are entirely co-planar (i.e, when α, the dihedral angle between planes A and B, is equal to 0°), maximal π-electron delocalization into the β-ionone ring occurs, and a maximal red shift is achieved^{8,9,137,142–144}. In contrast, when planes A and B are at a right angle to each other (α = 90° in Fig. 3C,D), π-electron delocalization cannot extend into the β-ionone ring, and a maximal blue shift results. Values of α between 0° and 90° would produce intermediate λ_max values.

This tuning mechanism could explain the observed relationship between λ_max-A1 and λ_max-A2 because the full effect of the additional double bond of the A₂ chromophore would be accessible for conjugation at α = 0°, while neither of the ring double bonds would be accessible for conjugation at α = 90°. Several modeling studies are consistent with a role for Blatz and Liebman’s mechanism in spectral tuning^8,137,142, and the crystal structure of the medium-wavelength-sensitive bovine rhodopsin (an RH1 opsin) demonstrates an ‘intermediate’ value of α as would be predicted by this model¹⁵¹. Nonetheless, a number of experiments using ‘locked’ chromophores¹³⁴ (in which a chemical bridge prevents changes in α) or 5,6-dihydroretinals¹³⁵ (which lack ring double bonds) indicate that changes in α alone cannot account for the full-range of spectral tuning observed in naturally occurring visual pigments. In conclusion, the Blatz and Liebman mechanism may play a role in spectral tuning, particularly in the ‘violet-blue’ region of the spectrum, but further modeling and experiments are required to evaluate this idea.

The disadvantages of vitamin A₂-based visual pigments

The advantages of A₂-based pigments (red-shifted and broadened spectral sensitivity and the potential for continuous tuning of λ_max) are counterbalanced by two notable disadvantages: they have lower thermal stability and lower intrinsic photosensitivity than A₁-based pigments. Cis-to-trans isomerization of a visual pigment and consequent activation of the phototransduction cascade can be caused either by absorption of a photon (light) or by random thermal fluctuations (heat)^10,14,152. Thermal isomerization is sometimes referred to as ‘dark noise’ or ‘dark light’ because the resultant activation of the phototransduction cascade is indistinguishable from that caused by light-induced isomerization¹⁴. Dark noise sets a fundamental limit to an organism’s ability to detect photons at very low light levels, because it is impossible to distinguish between photoisomerization and thermal isomerization of the visual pigment^153,154. Thus, visual detection in dim light is a signal-to-noise discrimination task: a switch from A₁ to A₂ might increase signal by more precisely matching λ_max to the predominant wavelengths of transmitted light, but this increase is offset by an increase in thermal noise¹⁰. Ala-Laurila and colleagues have estimated that replacing A₁ with A₂ in tiger salamander RH1 results in a red shift of 26 nm and a 36-fold increase in dark noise¹⁰. These antagonistic effects pose a conundrum for organisms in turbid environments in which the predominant wavelengths of light are red-shifted (favoring the use of A₂-based pigments) while the amount of transmitted light is simultaneously reduced (favoring less noisy A₁-based pigments).

The potential advantages of a switch to A₂ are further offset by a second factor: the lower photosensitivity of A₂-based visual pigments compared to those with A₁^10,155. Photosensitivity (αγ) is a measure of the efficiency with which absorption of light by a visual pigment (or other molecule) induces a specific change in that pigment^155–157. It is the product of two terms: α, the absorption coefficient, which is a measure of the efficiency of light absorption, and γ, the quantum efficiency, which is a measure of the efficiency with which the absorbed light causes isomerization^155,157. Dartnall found that the average photosensitivity of A₁-based RH1 visual pigments was 10.5 (cm² × 10⁻¹⁷ per chromophore), while the average photosensitivity of A₂-based RH1 pigments was 7.4¹⁵⁷. Thus, the photosensitivity of A₂-based visual pigments is only ~70% that of A₁-based pigments (Fig. 1D). Both increased noise and decreased photosensitivity counterbalance the advantages of A₂-based visual pigments in low light, but these disadvantages are likely of little consequence in bright light. It is therefore possible that the main selective advantage of A₂ is the large red shift (>40 nm) it confers on long-wavelength-sensitive cone opsins (e.g., Fig. 2E). However, Donner has noted that some A₂-utilizing species appear to have evolved RH1 opsins with greater thermal stability than species that do not use A₂^14,158. The implication is that thermally stabilizing RH1 mutations act to counter the increased noise of the A₂ chromophore. Furthermore, the presence of such mutations implies that the species in question must use A₂ under low-light conditions where thermal noise would be selectively relevant. Overall, the nearly ubiquitous utilization of A₂ in turbid aquatic environments suggests that its advantages outweigh the disadvantages, irrespective of whether natural selection is acting primarily on the photopic or scotopic visual system.

Unsolved problems related to the A₁/A₂ chromophore system

In this section I recap what I consider to be the most interesting outstanding questions related to the A₁/A₂ system, listed in the order in which they arise in the main text. (1) Why are A₂-based visual pigments so rare among fully terrestrial vertebrates, and what is their role in the two species (Anolis carolinensis and Podarcis siculus) known to possess them? (2) What factors determine the distribution of A₂ usage among fully marine fishes? (3) What are the mechanisms that control differential A₂ utilization in different parts of the retina (e.g., in the American bullfrog, the four-eyed fish, etc.)? (4) What features of the environment (temperature; salinity; turbidity; light intensity, duration, and wavelength etc.) play the biggest role in determining A₁/A₂ ratio, and what is the best way to measure them? (5) What are the molecular mechanisms whereby environmental signals are sensed, transduced, and integrated into changes in A₁/A₂ ratio in the eye? (6) Are A₂-based visual pigments used by any birds or mammals? (7) What is the function of cyp27c1 in species that do not use A₂ in the eye? (8) What are the mechanisms that permit some species (e.g., deep-sea dragonfishes and ninespine sticklebacks) to differentially tune the A₁/A₂ ratio in individual photoreceptor subtypes? (9) Are there enzymes other than cyp27c1 that mediate production of A₂ in vertebrates? (10) Is TH signaling always involved in the control of cyp27c1 expression or do TH-independent mechanisms exist? (11) Is the coordinate regulation of A₁/A₂ exchange and the expression of red-shifted opsins by TH signaling fortuitous or does it have a deeper physiologic or evolutionary significance? (12) What is the equation (or equations) that describes the relationship between λ_max-A1 and λ_max-A2, especially at wavelengths <400 nm, and what are the mechanisms of spectral tuning that underlie this relationship? (13) To what extent is visual pigment spectral tuning mediated by the degree of co-planarity between the β-ionone ring and the polyene chain of the chromophore? (14) To what degree does the lower photosensitivity and increased noise of A₂-based visual pigments limit their utility? (15) Is the selective advantage of A₂ utilization mainly attributable to its effects on rod or cone vision? (16) What is the magnitude of the selective advantage that A₂ utilization confers? (17) Can we identify mutations that independently control thermal stability and spectral tuning of opsins? (18) Are thermally stabilizing opsin mutations a necessary accompaniment of A₂ utilization?