Overview
Since the turn of this century, compelling evidence has emerged for a novel class of photoreceptors in the mammalian retina. These neurons are ganglion cells that express the photopigment melanopsin and respond autonomously to relatively bright light with a sustained depolarization and increase in spike frequency. Named intrinsically photosensitive retinal ganglion cells (ipRGCs), they differ radically in form and function from the classical rod and cone photoreceptors. In this chapter, we discuss the origins of this discovery, the idiosyncratic physiology of these cells, and the roles they play in retinal processing and visual behavior in health and disease.
Historical roots
The discovery of photoreceptors in the inner retina may have come as a shock initially to many retinal scientists, but the roots of the idea are actually traceable to the very beginnings of retinal science (reviewed in ). Beginning with Descartes and through the first half of the 19th century, it was assumed that the photoreceptive elements of the retina lined its inner surface, nearest the incoming light. For example, Treviranus, the first to conduct systematic microscopic studies of the retina, suggested that optic fibers passed through all retinal layers to end as photosensitive papillae at the vitreal surface. Bidder accurately described the outer retinal location of the rods, but hypothesized that they served a reflective function, like a tapetum, and continued to view optic fibers in the inner retina as the locus of phototransduction. The lack of photosensitivity of the fibers of the optic disc, as indicated by the blind spot, eventually undermined this view, and prompted Bowman and Helmholtz to propose ganglion cells as the true photoreceptors. It was not until the latter half of the 19th century that photoreception was convincingly localized to the outer retina. The key insights were made by Heinrich Müller, who inferred the retinal depth of the photoreceptors by a clever geometric analysis of the parallactic displacement of shadows of retinal vasculature (the “Purkinje tree”). He also detected “visual purple” or rhodopsin in the rods, which Franz Boll soon thereafter demonstrated was a photosensitive pigment.
Fifty years would pass before interest in the possibility of inner retinal photoreception would reemerge. The catalyst was the groundbreaking work of Clyde Keeler, who identified the first animal model of inherited retinal degeneration. This strain of mice, which Keeler termed “ rodless ,” is allelic with the rd1 strain (formerly rd ); both carry the Pde6b rd1 mutation and phenotypically resemble human autosomal recessive retinitis pigmentosa (OMIM database number 180072). Keeler eventually achieved wide acclaim for this seminal contribution to the genetics of retinal disease, but a key observation he made in these mice received little notice at the time. Although his rodless mice were apparently blind when tested behaviorally or electroretinographically, they unexpectedly retained robust pupillary responses to light. This led him to suggest that “in mammals the iris may function independent of vision” either through intrinsic photosensitivity of the smooth muscle itself or by “direct stimulation of the internal nuclear or ganglionic cells” by light.
Over the next 70 years, a series of studies in retinally degenerate mammals confirmed the persistence of the pupillary response but suggested that the functional blindness might not be as complete as Keeler had believed. Various photic effects on behavior or physiology were noted, despite the absence of an outer retina in histologic material or a detectable electroretinogram (reviewed in ). These included avoidance of a visual cliff, photic suppression of activity in open field test, visually guided avoidance of shock in a shuttlebox, chromatic discrimination, entrainment of circadian rhythms, suppression of pineal melatonin synthesis, and suppression of spontaneous firing in the superior colliculus. A number of these studies demonstrated that eye removal abolished these residual photoresponses and overtly echoed Keeler’s speculation about the possible existence of inner retinal photoreceptors, including ganglion cells. These suggestions gained only limited currency, probably because improved anatomical studies had begun to cast doubt on the completeness of the loss of outer retinal photoreceptors, especially in the peripheral retina. These studies showed that the mouse retina, once believed to possess only rod photoreceptors, in fact also had a modest population of cones, and that these degenerated far more slowly than rods in the rd1 model. Perhaps 20% survived beyond 80 days of age, albeit without intact outer segments, and a few lasted at least a year. This clouded the interpretation of the behavioral studies, virtually all of which had been conducted in mice young enough to have had many surviving cones. With the benefit of hindsight, however, it seems plausible that at least some of the residual light-driven effects reported in these early studies were indeed mediated by inner retinal photoreceptors. Even in rd1 animals as young as 6 to 10 weeks old, surviving rods or cones seem unable to support visual function: silencing ganglion-cell phototransduction in such animals abolishes the photic influences on circadian rhythms, the pupil, locomotor activity, and melatonin synthesis that these mice would otherwise exhibit (see “Central projections” further in the chapter).
In the 1990s, Russell Foster and colleagues took up the work on retinally degenerate mice, applying more rigorous methods to provide compelling evidence for inner retinal photoreception. Their innovations included appropriate controls for genetic background and the use of very old rd1 animals and several genetically modified mice strains in which they confirmed nearly complete loss of cones, as well as rods. They also showed that there was little loss of sensitivity of circadian photoreception as rod and cone degeneration progressed and that the residual visually evoked behaviors extended beyond circadian regulation to other outputs of what they called the nonimage-forming (NIF) visual centers of the brain. These included pupillary constriction, acute suppression of locomotor activity, and acute suppression of pineal melatonin release by light. The case for novel inner photoreceptors was bolstered by parallel observations in human patients with advanced outer retinal disease (see also ) and by assessment of the spectral tuning of residual photoresponses in retinally degenerate animals. Yoshimura and Ebihara (see also ) showed that the photic influence on circadian rhythms in retinally degenerate mice was most effective at 480 nm, clearly distinct from the optimal wavelengths for activating the known rod and cone photopigments. Remarkably similar spectral data were obtained for residual pupillary responses.
Discovery of melanopsin and ganglion-cell photoreceptors
The identities of the mysterious inner retinal photoreceptors and their photopigment were established in a flurry of studies in the period 2000 to 2005. These developments have been thoroughly reviewed elsewhere, so we provide only an abbreviated summary here. A pivotal contribution was the discovery of melanopsin. This novel opsin (coded by the Opn4 gene) derives its name from the dermal melanophores of frogs, the photosensitive cells of the skin in which the gene was first identified. It was shown subsequently that in mice and primates, including humans, the protein was expressed solely in a small minority of cells in the inner retina, mainly in the ganglion- cell layer (GCL). It was speculated that melanopsin could be the photopigment of the postulated inner retinal photoreceptors and that the neurons expressing it might be the cells of origin of the retinohypothalamic tract. These ideas were at odds with an alternative suggestion, introduced a few years earlier, that the relevant photopigment might not be an opsin but, rather, a cryptochrome, a blue light–absorbing flavoprotein (reviewed in ). However, overwhelming evidence for melanopsin’s central role soon emerged from studies in rodents and in heterologous expression systems.
Two key studies in this cohort focused on rat retinal ganglion cells shown by retrograde axonal tracing to innervate the suprachiasmatic nucleus (SCN) of the hypothalamus, the brain’s circadian pacemaker. The first showed that these ganglion cells expressed melanopsin (see also ), while the second demonstrated that they generated robust electrical responses to light even when pharmacologically or mechanically isolated from other retinal neurons (see also ). Because this capacity for autonomous phototransduction distinguishes these cells from all other RGCs, they are termed intrinsically photosensitive retinal ganglion cells (ipRGCs), or photoreceptive ganglion cells, or ganglion-cell photoreceptors. Soon, both mice and primates were shown also to possess ipRGCs with structural and functional properties much like those in the rat. See Box 26.1 for an example of the impact of this discovery on ophthalmic practice.
Ocular enucleation and evisceration are common surgical therapies for managing chronic debilitating ocular pain and improving cosmesis of blind eyes. Such surgery is typically indicated only in eyes lacking useful vision. However, the discovery of inner retinal photoreceptors should prompt a reassessment of best practices in such cases. When the blindness is attributable to outer retinal diseases such as retinitis pigmentosa or Leber’s congenital amaurosis, or to chronic retinal detachment, ganglion-cell photoreceptors may continue to provide functionally useful signals to the brain (see ). These signals may not be consciously perceived by the patient and may escape detection by most standard clinical tests of visual function. Such residual functions can nonetheless have significant impact on the quality of life for blind patients, since they may entrain their circadian rhythms. Loss of intrinsically photosensitive retinal ganglion cells (ipRGCs) through enucleation or evisceration would be expected to abolish such synchronization of circadian rhythms, with associated disruption of normal sleep patterns, daytime sleepiness, and other complications that plague many blind patients. It seems likely that photic effects on physiology, including regulation of neuroendocrine physiology, alertness, and mood, may also persist in such patients and be sacrificed when the eyes are removed. There is thus a pressing need to develop both a more inclusive concept of “useful vision” in ocularly blind patients and a more comprehensive panel of assessments of photosensory capacity when enucleation or evisceration is being weighed as a treatment option. Although tests of circadian or neuroendocrine effects would in principle allow residual inner retinal photoreceptor function to be evaluated, these would require highly specialized testing. Assessments of residual pupillary responses may be a promising avenue to explore.
The presence of melanopsin within physiologically identified ipRGCs was first demonstrated directly by Hattar et al . (see also ). The opsin was found not only in the cell body, but also in their dendrites which, like the soma, are directly photosensitive. Further support for the view that melanopsin was the sensory photopigment in these cells came from the opsin-like action spectrum of the light response in ipRGCs and from the observation that the intrinsic photosensitivity of these cells was abolished in melanopsin knockout mice. The capacity of melanopsin to function as a sensory photopigment was first demonstrated biochemically, and subsequently confirmed by electrophysiology and calcium imaging in heterologous expression systems. There is remarkably good concordance in the spectral sensitivity in this system, as assessed from absorbance of heterologously expressed or purified melanopsin, from the action spectrum of ipRGCs, and from behaviors mediated by inner retinal photoreceptors; all closely adhere to a retinaldehyde template function with a best wavelength at approximately 480 nm (but see ).
Distinctive functional properties of ipRGCs
Ganglion-cell photoreceptors have physiologic properties that are optimized for their roles in NIF vision and contrast markedly with those of the rod and cone photoreceptors that feed the cortical circuits mediating the perception of form, color, and motion. The photoresponse kinetics of ipRGCs also differs drastically from those of conventional, nonphotoreceptive ganglion cells.
Melanopsin chromophore and pigment bistability
In both vertebrate and invertebrate photopigments, the opsin apoprotein is covalently linked to a retinaldehyde molecule derived from vitamin A that serves as the light-absorbing moiety, or chromophore. In darkness, the retinaldehyde is in the 11- cis form. Absorption of a photon converts it into all- trans retinaldehyde. This triggers a conformational change in the opsin which in turn activates a G protein, initiating the transduction cascade. Before the photopigment can undergo another cycle of photoexcitation, the all- trans retinaldehyde must be reisomerized to 11- cis . For rod and cone photoreceptors, such reisomerization is carried out through multiple enzymatic steps after bleaching, that is, the dissociation of all- trans retinaldehyde from the opsin apoprotein. Several of these enzymatic steps occur in cells of the retinal pigment epithelium (RPE), which are essential for the visual cycle in rods and important, if not obligatory, for cones (see Chapter 13 for more information). The visual cycle in invertebrate photoreceptors appears to be radically different, in part because their photopigments are bistable: after photoexcitation, all- trans retinaldehyde remains covalently bound to the opsin and is reisomerized to 11- cis by subsequent absorption of light in a process known as photoreversal.
The chromophore of melanopsin in situ is 11- cis retinaldehyde. Sequence homology indicates that melanopsin resembles invertebrate opsins more than it does vertebrate ones, implying that it might be bistable. There is growing evidence in support of this view (but see ). If melanopsin is indeed bistable, it would be expected to be less reliant than rods or cones on the enzymatic machinery of the RPE for the regeneration of its chromophore, and might be utterly independent. Indeed, ipRGC photoresponses appear highly resistant to lighting conditions that fully bleach rod and cone photopigments, suggesting that all- trans retinaldehyde remains covalently bound to melanopsin following photoexcitation. Furthermore, neither acute pharmacologic disruption of the retinoid cycle nor genetic deletion of the essential isomerohydrolase of the RPE (RPE65) abolishes the melanopsin-based photoresponse of ipRGCs or their ability to photoentrain circadian rhythms (but see ). The evidence implies that to a remarkable degree, the melanopsin in ipRGCs can be loaded with its chromophore and can recover from photoactivation independent of retinoid processing by RPE. Melanopsin might acquire its retinoid mainly in the form of all- trans retinal which it then photoisomerizes to a cis isomer to form a photoexcitable pigment. This autonomy is important for the intrinsic photosensitivity of ipRGCs because these cells are located far (>100 µm) from the RPE (see Box 26.2 ). Nevertheless, during prolonged exposure to bright light, melanopsin regeneration in ipRGCs depends partly on 11- cis retinaldehyde supplied by the RPE, analogous to the partial dependence of Drosophila rhodopsin regeneration on the enzymatic visual cycle in RPE-like retinal pigment cells. During prolonged stimulation by bright light, 11- cis retinaldehyde appears to be transported from the RPE to ipRGCs by cellular retinaldehyde-binding protein (CRALBP) in Müller glial cells, which span the thickness of the retina.
There are at least two clinical conditions in which disruption of normal interactions between retinal pigment epithelium (RPE) and neural retina compromise rod and cone function but may permit continued phototransduction by intrinsically photosensitive retinal ganglion cells (ipRGCs). Retinal detachment disrupts the close contact between rod/cone outer segments and the RPE. This compromises the bidirectional exchange of retinoids across the subretinal space and thus the photosensitivity of rods and cones. Leber’s congenital amaurosis (LCA), an autosomal recessive, early onset form of retinitis pigmentosa, is caused in some cases by mutations in the Rpe65 gene. This disrupts the function of RPE65, a retinoid isomerase essential for the regeneration of 11- cis chromophore for rod and cone photopigments. Both retinal detachment and LCA have devastating consequences for photosensitivity of rods and cones. Evidence from animal studies (see main text) predicts that photosensitivity of ipRGCs should be substantially preserved under these conditions and should thus provide some useful photic information to the brain.
Spectral tuning
As noted previously, melanopsin has peak sensitivity in the blue region at around 480 nm, different from the spectral sensitivities of the rod and cone visual pigments ( Fig. 26.1 ). This wavelength roughly coincides with the peak of solar emission, perhaps the result of evolutionary pressures to optimize the sensitivity of this system to daylight. Although melanopsin is often described as a blue light–sensitive pigment, it is important to recognize that, like other opsin-based photopigments, its spectral tuning is rather broad. Throughout the spectral range from 340 to 580 nm, sensitivity is within 2 log units of that seen with the optimal wavelength ( Fig. 26.1 ). Although there is evidence that ipRGCs project to the lateral geniculate nucleus and could thus contribute to cortical function (see “Central projections” further in the chapter), we are unaware of any evidence that the unique spectral tuning of melanopsin is exploited in the conscious appreciation of color, which is well accounted for by the three cone photopigments and trichromatic theory (but see ). Melanopsin’s spectral tuning, however, should be considered in the design of intraocular lenses (see Box 26.3 ).
Designers of intraocular lens implants must grapple with the fact that there are both harmful and salutary effects of short-wavelength visible light. The goal must be to strike the appropriate balance between photoprotection and photoreception. High-pass filters that block blue, as well as ultraviolet, wavelengths offer better retinal photoprotection than those that block only ultraviolet light, while leaving rod and cone photoresponses (which peak near 500 nm and 555 nm respectively) relatively unaffected. However, because melanopsin is most sensitive to blue light with peak sensitivity at 480 nm, lenses that block blue wavelengths could compromise nonimage-forming (NIF) photoreception. On the other hand, reducing NIF photoreception in this manner might have the benefit of reducing the harmful effects of light exposure at night on the NIF visual system, as discussed in the main text. These factors need to be considered in the design and selection of lens implants. It has been shown that for the typical person, long-pass filtering with a sharp filter cutoff near 445 nm may provide the best compromise between photoprotection and light reception.
The spectral behavior of the ipRGC output signal is much more complex than predicted simply from the action spectrum of the intrinsic light response as measured in dark-adapted ipRGCs. This is in part because melanopsin appears to be a bistable photopigment (see the previous section). The absorption spectrum of the activated form of the pigment, and thus of photoreversal, is shifted to longer wavelengths with respect to that of melanopsin in the dark state (e.g., ), and this suggests the possibility of a form of spectral opponency under some conditions. A second source of complexity is the functional input to the ipRGCs from rod and cone photoreceptors, discussed in detail in “Synaptic input” further in the chapter. Because the intensity threshold for melanopsin activation is above that for rods and cones, the spectral tuning of ipRGCs can be expected to be intensity dependent. Furthermore, there is evidence in primates that the cone input to ipRGCs is itself spectrally opponent, with activation of short-wavelength cones driving OFF responses in ipRGCs, and activation of mid- and long-wavelength cones driving ON responses.
Invertebrate-like phototransduction cascade
In a photoreceptor cell, the light-absorbing pigment signals through an intracellular biochemical pathway to transduce light into electrical activity across the cell membrane. Despite its discovery in vertebrate tissues, melanopsin shares greater sequence homology to that of the rhabdomeric opsins (R-opsins) of invertebrates rather than the ciliary opsins (C-opsins) typically found among vertebrate species. This similarity to R-opsins suggested that melanopsin might activate a phototransduction cascade similar to that found in Drosophila photoreceptors rather than that in vertebrate rods and cones (see Chapter 18 ). In rhabdomeric phototransduction, R-opsins couple to Gq proteins, which activate phospholipase Cβ4 (PLCβ4), an enzyme that hydrolyzes the membrane phospholipid phosphatidylinositol (4,5)-bisphosphate (PIP2) into two second messengers: membrane-delimited diacylglycerol, and cytosolic inositol 1,4,5 trisphosphate (IP3). Canonical transient receptor potential (TRPC) channels are the terminal effectors of this cascade, mediating an influx of cations that results in depolarization across the plasma membrane.
Early melanopsin phototransduction experiments were conducted in heterologous expression systems, and melanopsin expression was found to render nonphotosensitive cells light-responsive, a necessary characteristic of a putative photopigment. With melanopsin introduced into cultured HEK-293 cells and Xenopus oocytes, illumination could activate Gq proteins and the downstream target, PLCβ4. Additionally, these HEK-293 cells, which were engineered to express TRPC channels, exhibited membrane depolarization in response to light, a quantitative experimental end point. This response, however, could be blocked by pharmacologic inhibition of Gq proteins or PLC. Similarly, in an oocyte expression system, light-induced inward currents were diminished by blockers of Gq and PLC. In cultured dermal melanophores, a native expression system in which melanopsin was originally identified, light increased IP3 levels and inhibition of PLC blocked light-induced dispersion of melanosomes, again implicating a rhabdomeric phototransduction pathway.
Electrophysiological studies in rodent ipRGCs have similarly shown the involvement of Gq and PLCβ4. The light-gated channel mediating membrane depolarization has remained elusive although a heterotetramer of TRPC6 and TRPC7 seems likely. These TRPC subunits have been identified in ipRGCs, and mice null for TRPC6 and TRPC7 show no photoresponses in M1-type ipRGCs ( Fig. 26.2 top ).
The subsequent discovery and characterization of multiple types of physiologically diverse ipRGCs, named M1–M6 (see “Morphological types” further in the chapter), raised the possibility that melanopsin-initiated signaling might also be more diverse than previous believed. Indeed, two isoforms of melanopsin have been identified. Moreover, while both M1 and M2 cells activate a typical rhabdomeric phototransduction cascade, as described previously, that terminates in the opening of heteromeric TRPC6/7 channels, M2 cells simultaneously activate a yet-to-be-characterized pathway that opens hyperpolarization-activated cyclic nucleotide–gated (HCN) cation channels ( Fig. 26.2 bottom ). The fast and slow components of photocurrents observed in cells have been attributed to the TRPC and HCN channels, respectively. As mentioned, the intermediate G proteins, effectors, and second messengers leading to the gating of HCN channels remain to be identified.
The melanopsin signaling pathways in M4 cells are less clear. Two hypotheses have been proposed. The first posits that the rhabdomeric cascade observed in M1 and M2 cells is conserved in M4 cells with the additional feature that a class of potassium leak channel closes in response to light. The second hypothesis suggests that the M2 cell pathway leading to HCN channel opening is also at play in M4 cells. These competing hypotheses are difficult to reconcile and require further investigation. Further study will also be necessary to characterize the phototransduction cascades of M3, M5, and M6 cells, which are yet to be elucidated.
Depolarizing photoresponse with action potentials
As mentioned in the previous section, an end point of the melanopsin phototransduction cascade is the opening of cation-selective channels in the plasma membrane. As for all other RGCs, the interior of an ipRGC is negatively charged relative to the extracellular space; thus, the opening of light-gated cation channels results in net influx of cations and membrane depolarization. In this, ipRGCs once again resemble invertebrate rhabdomeric receptors, which likewise depolarize when illuminated, but differ from the vertebrate rods and cones, which are hyperpolarized by light (see Chapter 19 ). If the depolarizing ipRGC light response is large enough, it brings the membrane potential above the threshold for activating voltage-gated sodium channels, triggering action potentials ( Fig. 26.3 ). This represents yet another divergence from the rod/cone photoreceptors, which do not spike. Action potentials are necessary for ipRGCs, as for other RGCs, to propagate electrical information faithfully along the optic nerve and tract to relatively distant central visual centers. By contrast, rod and cone axons are very short, and passive spread of the electrical signals generated in the outer segment is sufficient for appropriate voltage modulation of the axon terminal.
Sensitivity
A salient feature of ipRGCs is the relative insensitivity of their melanopsin-mediated responses to light. The threshold intensity for the ipRGC intrinsic photoresponse is 1 to 2 orders of magnitude (i.e., 10–100 times) higher than that for cones, and up to 5 log units higher than that for rods. Such insensitivity is apparently due mainly to the low abundance of melanopsin molecules in ipRGCs, and the resulting low probability of photon absorption. This in turn is related to the fact that ipRGCs lack any structural specialization for increasing the packing of pigment-laden membrane in the cell, such as the discs in the outer segments of rods and cones or the rhabdomeric microvilli of invertebrate photoreceptors. The insensitivity of intrinsic, melanopsin-based responses in ipRGCs means that such responses are triggered mainly by relatively bright light. Because the dynamic range of these responses complements those of the rod and cone photoreceptors, and because all three of these signals converge on ipRGCs (see “Synaptic input” further in the chapter), these retinal output neurons and at least some of the NIF visual responses they drive are able to function over the entire intensity range of naturally encountered light stimuli.
Kinetics
The temporal characteristics of the melanopsin-based ipRGC photoresponse are different from those of the outer retinal photoreceptors, with remarkably slower onset and termination than both rod and cone photoresponses, and greater stability in the face of sustained illumination than for the cone response. Rod and cone photoresponses begin within several milliseconds of light onset and peak within tens of milliseconds. Even the fastest ipRGC intrinsic responses, evoked by very high light intensities (>10 14 photons/cm 2 per second at 480 nm; for reference, direct sunlight is approximately 10 17 photons/cm 2 per second), are much slower, with the first action potential appearing no earlier than a few hundred milliseconds after light onset and peak firing rate achieved only after several seconds. Latencies rise dramatically for less intense flashes, and for near-threshold light stimuli (~10 11 photons/cm 2 per second at 480 nm) can be as long as tens of seconds to the first spike and several minutes before the peak discharge is reached. The basis for such sluggish onset arises in part from the phosphorylation of melanopsin: when melanopsin was engineered to contain reduced numbers of phosphorylable sites, photoresponse latency was reduced by as much as approximately threefold. However, this relatively modest effect of phosphorylation cannot account for the very long latencies near threshold, which seem more likely to result from the temporal integration of many low-amplitude, long-lasting, single-photon events.
In response to a prolonged light step, the melanopsin-driven response gradually decays from an early peak to a lower steady-state level. This reflects light adaptation, which prevents saturation of the phototransduction cascade, and it is partly calcium dependent as in the outer retinal photoreceptors (see also Chapter 20 ). After light adaptation is complete (which typically requires several minutes), membrane potential and thus spike frequency remain elevated above baseline levels for as long as the light stimulus persists. Electrophysiological recordings from ipRGCs have shown melanopsin-based photoresponses lasting at least 10 hours, while analysis of activity-dependent induction of the immediate early gene cFos in putative ipRGCs implies that these photoreceptors can respond continuously to light for at least 19 hours. Thus, the ipRGC intrinsic photoresponse is far more sustained ( Fig. 26.3 ) than the cone response, which adapts much faster and more completely, making the response very transient. The ability of ipRGCs to continuously signal the presence of bright light for many hours is a key characteristic of the NIF visual system, which features tonic responses to steady diffuse illumination and the ability to integrate photon flux over periods of more than an hour, as explained in more detail in “Central projections” further in the chapter.
The termination of the melanopsin light response is extraordinarily slow. After the cessation of light stimulation, cone and rod responses terminate within hundreds of milliseconds and several seconds, respectively. But termination of the intrinsic photoresponse of ipRGCs takes up to several minutes, especially for very high light intensities. The reason for this slow recovery is not established, but the similarity of melanopsin to invertebrate rhabdomeric opsins suggests that it results in part from the thermal stability of the light-activated (metarhodopsin) form of the photopigment. If so, the slowly decaying poststimulus voltage response is directly analogous to the persistent depolarizing afterpotential of invertebrate photoreceptors. This is supported by the observation that the poststimulus depolarization can be suppressed by exposure to long-wavelength light, presumably by triggering photoreversal of the pigment. However, light is not required for response termination, because this can occur in complete darkness. As expected for all G protein–coupled receptors, the light-independent mechanisms for response termination involve phosphorylation of melanopsin followed by arrestin binding.
The very sustained nature of ipRGC photoresponses is unique among ganglion cells. In response to a light step, nearly all other ganglion cells respond to the stimulus in a far more transient manner, with spike rate changes that last for no more than several seconds, indicating that whereas ipRGCs stably encode absolute light intensity, most conventional ganglion cells detect mainly changes in light intensity. This sustained versus transient dichotomy reflects differences in synaptic inputs and in biophysical properties intrinsic to the ganglion cells.
Morphologic types
In all mammals studied to date, melanopsin-expressing ganglion cells represent a small minority of all RGCs. These are now recognized as comprising multiple cell types, which have been studied most extensively in the mouse, in which six types of melanopsin-expressing RGCs have been identified. They can be distinguished by their differing morphologies, melanopsin expression levels, and central targets. As form tracks function, the distinct configurations and projections of the ipRGC types also suggest different physiologic properties and behavioral roles, as is explained in subsequent sections.
Schematics of the six mouse ipRGC types, named M1–M6, are shown in Fig. 26.4 . M1 and M2 were the first two types to be identified due to their robust melanopsin immunoreactivity. Mouse M1 cells have relatively small soma diameters of around 15 µm on average, and sparsely branched, broad dendritic arbors (~350 μm average field diameter) that stratify in the OFF sublayer of the inner plexiform layer (IPL), abutting the inner nuclear layer (INL). As explained in “Synaptic input” further on, despite such stratification, M1 cells and the other ipRGC types that contain OFF-stratifying dendrites do not exhibit OFF responses under normal conditions. M1d is a displaced subset of the M1 type, differing only in having somas within the innermost aspect of the INL. Additionally, two subsets of M1 cells differentially express the POU-domain transcription factor Brn3b, which correlates to the central targets and potential functions of those cells (see “Central projections” further in this chapter), suggesting that the M1 type consists of two subtypes. On the whole, M1 cells exhibit the highest intrinsic photosensitivity among the known types of mouse ipRGCs, although substantial heterogeneity in intensity dynamic range has been noted among M1 cells, which has been proposed to allow the M1 population to collectively encode a wide range of absolute light intensities.
Compared with M1 cells, mouse M2 cells have slightly larger somas averaging around 18 µm in diameter and broad, sparse arbors (~380 μm field diameter) that ramify in the ON sublayer of the IPL. They are also less melanopsin immunoreactive than M1 cells, suggesting a lower level of melanopsin expression. Accordingly, in terms of autonomous phototransduction, M2 ipRGCs are an order of magnitude less sensitive to light.
M3 somas can be visualized using antimelanopsin antisera but the neurites are difficult to resolve with standard immunohistochemical techniques, suggesting relatively low melanopsin expression, and indeed M3 cells generate weaker intrinsic photoresponses than both M1 and M2. M3 somas are intermediate in size to M1 and M2 somas, and their sparse dendrites bistratify in both ON and OFF sublaminae of the IPL, with an average dendritic field diameter of ~460 μm, the largest among mouse ipRGCs.
It is now understood that M4 cells correspond to what have been previously characterized as ON alpha RGCs. ON alpha RGCs were originally not believed to be intrinsically photoreceptive, but now are known to express a very low level of melanopsin and possess autonomous photosensitivity. The M4 type has the largest soma (~24 µm) of any class of mouse RGC, with dendritic field diameter ranging from approximately 270 μm to approximately 370 μm, depending on retinal location. The entire extent of the cells can be labeled using antibodies against SMI-32, which stain both ON alpha and OFF alpha RGCs. Owing to the low melanopsin content of M4 cells, their intrinsic photoresponse has an even higher threshold than that of M2 cells.
Because M5 and M6 cells express the lowest melanopsin levels among mouse ipRGC types, they have been identified using Cre recombinase–based melanopsin-specific reporter mice, or tyramide-enhanced melanopsin immunoreactivity. M5 cells have small somas of around 14 µm in diameter and compact, bushy dendritic arbors (~220 μm diameter) that terminate in the ON sublamina of the IPL. M6 cell morphology resembles that of M5 cells with small somas (~12 μm diameter) and small dendritic arbors (~220 μm diameter), except that M6 cells show slightly more exuberant branching than M5 cells and are bistratified, with additional dendrites terminating in the OFF sublamina of the IPL. As expected from the very low melanopsin expression in M5 cells, the peak amplitude of their intrinsic photoresponse appears even smaller than that for M3 cells. The intrinsic response of M6 cells is likewise relatively small, although it has not been systematically compared with those of the other ipRGCs types.
In addition to having diverse intrinsic photoresponse sensitivities reflecting different melanopsin expression levels (see previously) and differential expression of melanopsin isoforms, the various mouse ipRGC types have been shown to differ in terms of additional physiologic measures such as resting membrane potential, membrane resistance, spontaneous spike rate, spike waveform, and voltage-gated currents. These ipRGC types also exhibit diverse extrinsic photoresponses reflecting differences in synaptic input from rod/cone-driven circuits, as explained in “Synaptic input” further in the chapter.
The morphology and distribution of human ipRGCs have been described in detail. Of the approximately 1.5 million RGCs in a human, only about 7300, or 0.5%, express melanopsin. They comprise five morphologic types, four of which appear homologous to mouse ipRGCs: a strongly melanopsin-immunoreactive M1-like type with OFF-stratifying dendrites; an M2-like ON-stratifying type with somewhat weaker melanopsin immunoreactivity; a relatively rare M3-like ON/OFF–bistratifying type; and an M4-like ON-stratifying type with very weak melanopsin immunoreactivity. The fifth type of human ipRGC, named “gigantic M1” or “GM1,” has OFF-stratifying dendrites and very large somas averaging about 33 µm. For both M1-like and GM1 types, a subset of cells have somas displaced to the INL. The dendritic fields of human ipRGCs are the largest of any ganglion cell type, ranging from 300 µm in the central retina to 1200 µm in the periphery ( Fig. 26.5 ). The dendrites of neighboring cells overlap extensively so that the entire retina, except for the fovea, is covered by a network of melanopsin-containing processes ( Fig. 26.6 ). Because both the cell bodies and the dendrites of these cells are intrinsically photosensitive, each ipRGC integrates photons over a region comparable to its dendritic field size. These large receptive fields and low spatial densities make ipRGCs ill-suited for fine spatial discriminations, but ideal for the pronounced spatial integration that characterizes NIF functions. They contrast sharply with the very small receptive fields of rods and cones (<10 µm in diameter), the foundation for the fine-grained retinal representations at the cortical level that permit high acuity spatial vision (see Chapter 33 ). The melanopsin-mediated photoresponses of human ipRGCs have been investigated using extracellular spike recording and were found to comprise three physiologic varieties with different intensity thresholds and response kinetics.
Multiple ipRGC types have also been identified in many other species. The rat retina contains M1–M5, the morphologic properties and physiologic diversity of which are remarkably similar to their mouse counterparts. M1-like OFF-stratifying ipRGCs and M2-like ON-stratifying ipRGCs are present in nonhuman primates, including macaque and marmoset. Both rabbit and tree shrew possess three morphologic types of melanopsin-immunopositive RGCs that somewhat resemble M1–M3, whereas only M1-like and M3-like ipRGCs have been detected in the Mongolian gerbil retina. Extracellular spike recording has revealed three physiologic types of ipRGCs in the Sudanian grass rat, a diurnal rodent.
Resistance to pathologic states
Compared to other types of ganglion cell, ipRGCs appear more resistant to a variety of acute insults including N -methyl- D -aspartic acid–induced excitotoxicity, optic nerve transection, and optic nerve crush. Two weeks following optic nerve crush, M1 ipRGCs are reduced by 30%, whereas “few if any” M2 cells survive. In experimental models of glaucoma, ipRGCs also exhibit some resilience relative to other (i.e., conventional) RGCs. After 6 to 8 months of experimentally elevated intraocular pressure (IOP) by laser coagulation, the general population of RGCs is reduced by 25% to 33%, while ipRGCs are reduced by only about 17%. Like optic nerve crush, ipRGC types are differentially affected by IOP elevation. At least 6 months after elevated IOP, M1 ipRGCs show no significant decline in number. Yet, M4 cells are reduced by 25%, in line with the reduction observed in the general RGC population. Additionally, chronically elevated IOP has no impact on circadian entrainment, a primarily M1-mediated NIF visual function. Contrast sensitivity, however, is diminished in these animals, consistent with the loss of image-forming RGCs including M4 ipRGCs. The participation of several ipRGC types in image-forming vision will be described in “Central projections” further in the chapter.
The molecular basis for these kinds of injury resistance is largely unknown, although the PI3K/Akt signaling pathway appears to be involved. Moreover, melanopsin expression levels seem to correlate with resilience to insult. Overexpression of melanopsin in RGCs promotes axonal regeneration by activating the mammalian target of rapamycin (mTOR) pathway. Further investigation into the molecular basis of ipRGC survival after retinal injuries may lead to novel strategies for preventing and/or treating various retinal diseases (see also Box 26.4 ).
Transient transfection with the melanopsin gene is sufficient to induce photosensitivity in a variety of cell types in culture ( Fig. 26.7 top ). Introduction of melanopsin into inner retinal cells thus holds promise as a candidate gene therapy for restoring sight in retinitis pigmentosa and other forms of outer retinal blindness. The practicality of this approach is enhanced by the fact that intrinsically photosensitive retinal ganglion cells (ipRGCs) do not absolutely require the retinal pigment epithelium for chromophore regeneration. Another advantage is that all of the critical elements of melanopsin-driven phototransduction appear to be associated with the plasma membrane, thus reducing the requirements for specialized cytosolic machinery. Indeed, three studies using a mouse model of retinitis pigmentosa report that transfecting conventional RGCs with the melanopsin gene renders most of them photosensitive and can rescue some light-driven behavioral responses ( Fig. 26.7 bottom ).
However, several disadvantages of this approach need to be considered. First, in both ipRGCs and the various heterologous expression systems, the melanopsin-based photoresponse is far more sluggish than those of the rods and cones. Thus, even if image-forming vision is restored, the patient’s perception of stimulus motion and of the timing of transient visual events would presumably be highly distorted. Second, ipRGCs are even less sensitive to light than cones, so visual threshold would be even higher than for people suffering from night blindness. Still, melanopsin is more light sensitive than the most promising alternative candidate for gene therapeutic induction of inner retinal photosensitivity, namely the microbial light-gated ionophore channelrhodopsin-2. Third, if melanopsin is targeted to RGCs, all of which have dendritic fields vastly larger than the profile of rod and cone outer segments, then spatial resolution would be quite poor. The average receptive field diameter of melanopsin-transfected mouse RGCs in one study was about 100 µm. To improve spatial resolution, it might be better to selectively target bipolar cells for melanopsin transfection (see for example ). It is unknown if these cells express critical components of the melanopsin phototransduction cascade, and so Van Wyk and colleagues developed a chimeric protein containing the light-sensitive domains of melanopsin and the intracellular domains of the ON bipolar cell–specific metabotropic glutamate receptor type 6, to ensure that photostimulation of this protein can activate the G protein transduction cascade in ON bipolar cells. Introducing this chimeric protein into the ON bipolar cells of rd1 mice improved visual function.
Synaptic input
Unlike conventional RGCs, ipRGCs do not require synaptic input to respond to light because they are directly photosensitive. Nonetheless, both primate and rodent ipRGCs receive synaptic inputs from bipolar and amacrine cells and thereby generate rod/cone-driven, as well as melanopsin-based, light responses. Receiving these inputs has consequences for the timing, spectral behavior, and sensitivity of the light-evoked discharges of these cells. A schematic diagram summarizing the intraretinal synaptic inputs to M1 and M2 ipRGCs appears in Fig. 26.8 .
Bipolar cell input
There is strong structural and physiologic evidence for synaptic contacts from bipolar cell axons onto ipRGC dendrites. The earliest anatomical evidence came from electron-microscopic immunohistochemical data in mice showing ribbon synaptic (bipolar) contacts onto melanopsin-expressing dendrites and this has been supported by subsequent structural observations. Bipolar cell input to ipRGCs has also been confirmed electrophysiologically. Rodent ipRGCs are excited by applied glutamate, the transmitter released by bipolar cell terminals. Spontaneous glutamate-mediated excitatory postsynaptic currents are detectable in ipRGCs, and these almost certainly derive from bipolar cell synapses. Under appropriate recording conditions, broad-spectrum (white) light elicits two excitatory ON response components in ipRGCs. One has a high threshold, is sluggish in its onset and termination, and is insensitive to synaptic blockers; these properties testify to its origin in intrinsic melanopsin-based phototransduction. The other component occurs much more rapidly after light onset, can be abolished by pharmacologic blockade of synaptic transmission, and is attributable to excitatory synaptic input from ON bipolar cells. Both rods and cones contribute to these extrinsic light responses, making them as many as five orders of magnitude more sensitive to light than the melanopsin photoresponse ( Fig. 26.9 ).
It is surprising that the ON channel provides the dominant bipolar input to all ipRGCs. For decades, ON bipolar cell axons were thought to contact ganglion cell dendrites only in the inner half of the IPL, that is, in the ON sublamina (“IPL: On ” in Fig. 26.8 ). Whereas this arrangement is congruent with the ON input to M2–M6 ipRGCs, which possess ON-stratifying dendrites, it seems at odds with the observation that the ON channel also dominates the synaptic inputs to the OFF-stratifying M1 and M1-like ipRGCs, which arborize largely or exclusively within the OFF sublamina. Because most of the OFF-stratifying cells have somas in the GCL, their dendrites must traverse the ON sublayer en route to the OFF sublayer, so some of the ON channel input could conceivably occur in the ON sublayer on proximal dendrites ( Fig. 26.4 top ). Indeed, Belenky et al. reported that virtually all bipolar cell contacts onto melanopsin dendrites were found in the inner half of the IPL, although they did not indicate whether these dendrites derived from the outer- or inner-stratifying types of melanopsin RGCs. Two studies have also suggested that rod bipolar cells (a type of ON bipolar cell) make direct contact onto the GCL somas and proximal dendrites of melanopsin-expressing RGCs, including M1 cells (but see ). However, contacts solely on the soma or proximal dendrites would severely restrict the size of the synaptically driven receptive fields of these ipRGCs, when in fact they are large and coextensive with the dendritic arbor. This suggests that there must be additional ON bipolar input to the distal dendrites of M1 cells in the OFF sublayer. This inference is further supported by the observation that displaced M1 ipRGCs, which have somas in the INL and dendrites restricted to the OFF sublamina, nonetheless exhibit synaptically mediated ON responses. There is compelling evidence that these paradoxical ON bipolar inputs are made by ectopic ON bipolar cell axon terminals and synaptic release sites in the OFF sublamina, belonging primarily to type-6 ON cone bipolar cells ( Fig. 26.8 top ).
Anatomical studies have not detected any direct synaptic contact between OFF bipolar cells and ipRGCs. However, pharmacologic blockade of ON bipolar cells and amacrine cells reveals in some rat M1 ipRGCs a very small depolarizing ipRGC response at light offset, consistent with a weak input from OFF bipolar cells. It is plausible that these M1 cells receive OFF bipolar input polysynaptically, via gap junctional electrical synapses with OFF bipolar–driven amacrine cells, as has been documented for certain conventional ON ganglion cells.
Amacrine cell input and neuromodulation
Ganglion-cell photoreceptors also receive substantial synaptic input from amacrine cells. Initial evidence came from an electron-microscopic study in mouse documenting amacrine cell synaptic release sites in close apposition to melanopsin-containing dendrites in the IPL. The dominant effect of these inputs is almost certainly inhibitory because the vast majority of amacrine cells contain one of two major inhibitory transmitters: γ-aminobutyric acid (GABA) or glycine. Studies in rodent retina have demonstrated that exogenously applied or endogenously released GABA and/or glycine trigger inhibitory chloride currents in ipRGCs and that such currents can be activated by light, primarily at light onset. In primate ipRGCs, amacrine inputs have not been investigated electrophysiologically but there is extensive anatomical evidence for them in marmoset and macaque, where they contact both outer- and inner-stratifying melanopsin-expressing dendrites.
There are at least 50 types of amacrine cells in mammalian retinas, differing in dendritic stratification and field size, as well as in their neurochemical signature. Several amacrine cell types have been found to form chemical synapses with ipRGCs, and the best understood is the dopaminergic amacrine cell, which has processes stratifying almost exclusively in the most distal sublayer of the IPL, the same stratum in which dendrites of outer-stratifying ipRGCs arborize. Dopamine is a key retinal neuromodulator, and its primary role is to help retinal cells and circuits adapt to different background lighting conditions. Thus, ipRGCs can potentially adapt to light not only through intrinsic mechanisms (“photoreceptor adaptation” ), but also through synaptic inputs (“network adaptation”). Supporting this possibility, dopamine has been shown to regulate the transcription of melanopsin in rat over a time course of several hours, and electrophysiological studies have demonstrated additional, more acute effects of dopamine on the intrinsic photoresponse of ipRGCs. There is emerging evidence that ipRGC electrophysiology can also be influenced by several additional neuromodulators. Some of these substances exert inhibitory effects on ipRGCs, such as opioids, somatostatin, and adenosine, whereas orexin-A has been found to potentiate spontaneous spiking and light-evoked responses in M2 ipRGCs.
In addition to being regulated by amacrine cell neuromodulators, ipRGCs express receptors for melatonin, a hormone secreted at night by outer retinal photoreceptors, and melatonin has been shown to alter synaptic inputs to M4 ipRGCs. Moreover, various aspects of ipRGC electrophysiology are under circadian control, causing these neurons to exhibit different photoresponses, resting membrane potentials, and spontaneous spike rates at different times of day. Such time-dependent variations could contribute to a circadian rhythm in the ipRGC-mediated pupillary light reflex (PLR).
Extrinsic versus intrinsic photoresponses
In primate M1- and M2-like ipRGCs, the intrinsic melanopsin-based responses are invariably depolarizing, but the polarity of the extrinsic, synaptically mediated light response is wavelength-dependent : blue light elicits hyperpolarizing extrinsic responses, whereas longer wavelengths evoke depolarizing ones ( Fig. 26.10 ). Whereas the depolarizing responses are driven by ON bipolar cells, the hyperpolarizing response to short-wavelength light is mediated by an amacrine cell type that receives input exclusively from blue cone–selective ON bipolar cells. In mice, M5 ipRGCs also exhibit color opponency, with green cone input in the surround region of the receptive field driving amacrine-mediated inhibitory responses. The overall spectral behavior of these color-opponent ipRGCs is unusually complex, because it is shaped not only by their synaptic inputs but also by the two spectrally distinct states of their melanopsin photopigment, which exert opposing photic effects on the phototransduction cascade (see “Spectral tuning” previously in the chapter).
