Phototransduction by Retinal Ganglion Cells That Set the Circadian Clock

See allHide authors and affiliations

Science  08 Feb 2002:
Vol. 295, Issue 5557, pp. 1070-1073
DOI: 10.1126/science.1067262


Light synchronizes mammalian circadian rhythms with environmental time by modulating retinal input to the circadian pacemaker—the suprachiasmatic nucleus (SCN) of the hypothalamus. Such photic entrainment requires neither rods nor cones, the only known retinal photoreceptors. Here, we show that retinal ganglion cells innervating the SCN are intrinsically photosensitive. Unlike other ganglion cells, they depolarized in response to light even when all synaptic input from rods and cones was blocked. The sensitivity, spectral tuning, and slow kinetics of this light response matched those of the photic entrainment mechanism, suggesting that these ganglion cells may be the primary photoreceptors for this system.

The SCN is the circadian pacemaker of the mammalian brain, driving daily cycles in activity, hormonal levels, and other physiological variables. Light can phase-shift the endogenous oscillator in the SCN, synchronizing it with the environmental day-night cycle. This process, the photic entrainment of circadian rhythms, originates in the eye and involves a direct axonal pathway from a small fraction of retinal ganglion cells to the SCN (1–3). A striking feature of this neural circuit is its apparent independence from conventional retinal phototransduction. In functionally blind transgenic mice lacking virtually all known photoreceptors (rods and cones), photic entrainment persists with undiminished sensitivity (4). Candidate photoreceptors for this system are nonrod, noncone retinal neurons, including some ganglion cells, that contain novel opsins or cryptochromes (5–8).

To determine whether retinal ganglion cells innervating the SCN are capable of phototransduction, we labeled them in the rat retina by retrograde transport of fluorescent microspheres injected into the hypothalamus (9). In isolated retinas, whole-cell recordings were made of the responses of labeled ganglion cells to light (10) (Fig. 1, A to E). In most of these cells (n = 150), light evoked large depolarizations with superimposed fast action potentials (Fig. 1, E to G) (11). The light response persisted during bath application of 2 mM cobalt chloride (Fig. 1F; n = 42), which blocks calcium-mediated synaptic release from rods, cones, and other retinal neurons (12). In contrast, other ganglion cells prepared and recorded under identical conditions but not selectively labeled from the SCN (control cells) lacked detectable response to light even without synaptic blockade (47/50 cells; Fig. 1, I and J) (13). This is presumably because rod and cone photopigments were extensively bleached (10). A few control cells (3/50) exhibited weak, evanescent responses to light, but these were abolished by bath-applied cobalt (n = 2).

Figure 1

Labeling and light responses of rat ganglion cells innervating the SCN. (A) Fluorescence photomicrograph showing deposit (white arrow) of mixed red and green fluorescent microspheres (appears as yellow) in the SCN. The red arrows mark boundaries of contralateral SCN. Acridine orange was used for green fluorescent Nissl counterstain. ox, optic chiasm; III, third ventricle. Scale bar, 500 μm. (B) Two ganglion cells back-filled from the SCN, photographed in whole mount with rhodamine filter set to show retrograde labeling. The cell at right was patched and recorded. Scale bar, 20 μm. (C) Same cells viewed under blue excitation, to show LY filling of the recorded cell. (D) Camera lucida drawing of cell filled in (C), as viewed in the whole mount after antibody to LY immunostaining (35). Scale bar, 100 μm. (E) Strong depolarization and fast action potentials evoked in this cell (D) by a light pulse indicated by step in horizontal line below. (F to H) Evidence for the intrinsic photosensitivity of ganglion cells selectively retrolabeled from the SCN. (F and G) The light response apparent in control Ames solution (black traces) persisted during bath application of 2 mM CoCl2 (red traces) either alone (F) or in combination with a drug mixture blocking ionotropic and metabotropic glutamate receptors as well as ionotropic GABA and glycine receptors (G) (14). The absence of evoked spikes during drug application probably reflects depolarization block (tonic sodium channel inactivation) because weaker stimuli evoked spikes (40). (H) Light response recorded from the isolated soma of a ganglion cell retrolabeled from the SCN. The cell body was bathed in an enzyme solution (papain, ∼20 units/ml) with a puffer pipette and then mechanically removed from the retina with an empty patch pipette under visual control, amputating its dendrites and axon. (I and J) Control recordings from a conventional ganglion cell (Fig. 4C) labeled nonselectively by a deposit of rhodamine beads in the optic chiasm and filled with LY. Light evoked no detectable response (I), although synaptic transmission was not blocked and responses to current injection were normal (J) (+50 pA). Retinal irradiance of stimuli (in photons s−1 cm−2): (E) 7 × 1012, (F) 2.6 × 1013, (G) 7.2 × 1012, (H) 1 × 1013, and (I) ≫9 × 1013. Stimuli in (E), (F), and (G) were 500 nm.

To ensure blockade of conventional synaptic influences from rods and cones, we supplemented cobalt with a mixture of drugs that independently disrupted both the glutamatergic synapses crucial to vertical signal transfer through the retina and the ionotropic receptors responsible for most inhibitory influences on ganglion cells (14). Robust light responses persisted in SCN-projecting ganglion cells under these conditions (Fig. 1G;n = 7). Furthermore, the somata of these ganglion cells exhibited photosensitivity even when completely detached from the retina by microdissection (Fig. 1H; n = 3). These light responses were not an artifact of photic excitation of either of the intracellular fluorophores we used, as the action spectrum of the light response (Fig. 2C) differed from the absorption spectra of both the retrograde tracer and Lucifer Yellow (LY) used for intracellular staining. Also, light-evoked increases in spike frequency were detectable in extracellular recordings, before patch rupture and LY dye filling (n = 5). Whole-cell recordings revealed normal light responses when LY was omitted from the internal solution (n = 8). In contrast, control cells lacked cobalt-resistant light responses even when labeled with both fluorescent beads and LY (n = 12; Fig. 1I). These data indicate that retinal ganglion cells innervating the SCN are intrinsically photosensitive.

Figure 2

Spectral tuning of light response in photosensitive ganglion cells. (A) Voltage responses of a single cell to a 500-nm narrow-band stimulus at indicated intensities (in log10 photons s−1 cm−2). Baseline ∼ −60 mV for each trace. (B) Plots of peak depolarization as a function of log retinal irradiance for each of several narrow-band spectral lights (400- to 600-nm wavelength, as indicated; 10-nm width at half height); same cell as in (A). Peak was obtained from a 1-s boxcar average of raw voltages. (C) Spectral sensitivity functions derived for photosensitive ganglion cells from relative displacements of intensity-response functions along the abscissa in (B). Red curve: same cell as in (A) and (B). Green curve: group data for all cells (n = 34; number of cells tested per wavelength as follows: 400 nm, 5; 420 nm, 3; 440 nm, 2; 460 nm, 4; 480 nm, 4; 500 nm, 34; 520 nm, 7; 540 nm, 5; 570 nm, 2; and 600 nm, 2). Black curve: nomogram for retinal1-based photopigment with λmax of 484 nm (41), fit by least squares method to the group data.

To determine if these cells could serve as the primary photoreceptors for circadian entrainment, we assessed congruence between their photic properties and those of the entrainment mechanism. The responses of a single cell to narrow-band stimuli of various intensities showed that at each wavelength, peak depolarization increased with stimulus energy (Fig. 2, A and B). Intensity-response curves exhibited a consistent slope when plotted in semilogarithmic coordinates (Fig. 2B), as expected for responses mediated by a single photopigment (principle of univariance). The horizontal displacements of the curves from one another reflect the spectral dependence of the pigment's quantum efficiency and yield the spectral sensitivity function shown in Fig. 2C (red curve). Other cells exhibited similar action spectra (Fig. 2C, green curve) (15). These action spectra closely matched that predicted for a retinal1-based pigment with peak sensitivity at 484 nm (Fig. 2C, black). They also resemble action spectra derived behaviorally for circadian entrainment in rodents (16,17), as expected if these ganglion cells function as primary circadian photoreceptors (18). Judging from available spectral evidence, the photopigment in these ganglion cells is more likely to be a retinaldehyde-based opsin such as melanopsin (5, 19, 20) than a flavin-based cryptochrome (21).

The threshold and dynamic range of the light response in these ganglion cells were also similar to those of the entrainment mechanism. Threshold retinal irradiance for a full-field stimulus was about 5 × 1011 photons s−1 cm−2 at 500 nm (∼λmax; n = 3) . This corresponds to an in vivo corneal irradiance of ∼2 × 1013 photons s−1 cm−2, comparable to thresholds for circadian phase shifts in rodents (∼1010 to 1013 photons s−1 cm−2 at 500 nm) (22–24) and to ocular illumination by the dawn sky. Response saturation in photosensitive ganglion cells occurred at irradiances ∼3 logarithmic units above threshold (Figs. 2B and3C), matching the dynamic range of entrainment behavior (16, 23,24) and many SCN neurons [(25), but see (26)].

Figure 3

Tonic encoding of light intensity by photosensitive ganglion cells. (A) Maintained depolarization evoked by 20 min of constant, diffuse illumination (retinal irradiance = 3 × 1013 photons s−1cm−2; 500 nm); spikes obscured by low sampling rate. (B) Voltage responses evoked in another cell by 4-min stimuli varying in intensity (log10 irradiance in photons s−1 cm−2 indicated at left; 500 nm). Note tonic depolarization, maintained spiking, and, for near-threshold stimulus (bottom trace), extremely long response latency. (C) Magnitude of maintained depolarization as a function of stimulus energy for the cell in (B) (open circles) and for two other cells. Bathing medium: Ames.

The circadian entrainment mechanism integrates light energy over very long time scales, exhibits little adaptation, and responds poorly to brief stimuli (24, 27). Similar features were evident in the behavior of photosensitive ganglion cells. Constant illumination depolarized cells tonically and elevated spike frequency, and the amount of depolarization was monotonically related to stimulus energy (Fig. 3). Response kinetics were much slower than typical for ganglion cells. Latencies to response onset (V m > 3 standard deviations above baseline) were typically several seconds and ranged from several hundred milliseconds for saturating stimuli (Figs. 1F and 2A) to ∼1 min near threshold (Fig. 3B, bottom trace). Latencies from stimulus onset to peak depolarization were typically 10 to 20 s (range: ∼2 s to 2 min) and inversely related to stimulus energy. Repolarization after intense stimuli required several minutes and was sometimes punctuated by spontaneous depolarizations and spike bursts lasting up to a minute each (Fig. 3B).

Photosensitive ganglion cells shared a common morphology (Fig. 4, A and B), as revealed by intracellular staining with LY (28). Somata were intermediate in diameter among neurons of the ganglion cell layer (14.7 ± 1.2 μm, mean ± SD; n = 18). Many cells sent an axon into the optic fiber layer; those lacking one had presumably lost it during mechanical exposure of the soma before recording. The sparsely branching, tortuous dendrites of these cells arborized primarily in the outer part (OFF sublayer) of the inner plexiform layer (IPL; Fig. 4B). Although some dendrites coursed within the inner IPL (ON sublayer) for 100 to 200 μm, nearly all terminated in the OFF sublayer. Such stratification is highly unusual for ganglion cells depolarized by light [but see (29, 30)]. Dendritic fields were large (diameter 497 ± 115 μm; mean ± SD;n = 21). Stimuli illuminating the dendrites but not the soma consistently evoked light responses (31). Control cells, which lacked cobalt-resistant light responses, had markedly different dendritic morphology (e.g., Fig. 4C).

Figure 4

Morphology of photosensitive ganglion cells as revealed by intracellular staining. (A) Camera lucida drawings of cells stained with LY and viewed in retinal flat mounts. Arrowheads indicate axons. (B) Schematic summary of dendritic stratification of these cells, predominantly in the OFF sublayer of the inner plexiform layer. (C) Morphology of a control ganglion cell that lacked any light response under these conditions (Fig. 1I). Scale bar applies to all panels.

These data identify a distinct ganglion cell type in the mammalian retina with characteristic dendritic profile and stratification pattern, extraordinarily sluggish and tonic light responses encoding ambient light level, and axonal projections to the SCN. The most striking feature of this cell type, however, is its apparent capacity for intrinsic phototransduction. The correspondences between the photic properties of these cells and those of the entrainment mechanism suggest that these unconventional ganglion cells may represent the primary photoreceptors for synchronizing the circadian clock to environmental time.

Melanopsin (5, 7) is probably the photopigment responsible for the intrinsic sensitivity of these cells to light, as it is selectively expressed in the small subset of ganglion cells that are intrinsically photosensitive and innervate the SCN (32, 33). In amphibians, certain nonretinal cells contain melanopsin, and these cells, too, are photosensitive, with action spectra resembling those of photosensitive rat ganglion cells (Fig. 2C) (5, 19, 20). Melanopsin exhibits marked sequence similarity to invertebrate opsins (5, 7), which, unlike vertebrate opsins, retain their photoisomerized retinaldehyde chromophore and typically trigger depolarizing light responses when activated. These properties may help to explain why photosensitive ganglion cells differ from conventional retinal photoreceptors in their response polarity and lack of dependence on the pigment epithelium. Cryptochromes, blue-light–absorbing, flavin-based pigments, have been proposed as circadian photopigments (8), but spectral evidence (Fig. 2C) (21) weighs against their mediating the light response in intrinsically photosensitive ganglion cells.

Note added in proof: Further evidence for the presence of melanopsin in ganglion cells innervating the SCN has emerged (34).

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

  • These authors contributed equally to this work.


Stay Connected to Science

Navigate This Article