Sensing light without sight: The visual system’s ‘third eye’

ipRGCs provide non-image vision, responding to light independently of rods and cones
Intrinsically photosensitive retinal ganglion cells, rich in melanopsin, respond to light independently of rods and cones. (Courtesy Elliott Milner, PhD)

Michael Tri H. Do, PhD, is an investigator in the F.M. Kirby Neurobiology Center at Boston Children’s Hospital and an assistant professor of neurology at Harvard Medical School.

Light affects us even without impinging on our awareness. In 1995, Charles Czeisler and colleagues at Harvard Medical School described people who lacked visual perception due to retinal degeneration, but nevertheless responded to light subconsciously — despite being blind, their melatonin level was suppressed, and they appeared to synchronize their circadian clock with the solar day. Across the pond at Oxford, Russell Foster and colleagues were finding the same in mice, and learned that these responses began in the eye.

These discoveries spurred an intense research effort that continues to this day. What system confers subconscious sight, and how does it differ from the system that generates visual experience?

An outdated view of how vision begins

For well over a century, mammalian vision was understood to start with the rods and cones of the retina. These photoreceptors contain visual pigments, which absorb light and trigger a cascade of reactions that results in an electrical signal. That signal is processed by the retinal circuitry and conveyed to the brain by retinal ganglion cells (RGCs), which extend axons to form the optic nerve.

understanding of retina, before non-image vision
We now know that among the retinal ganglion cells (i and j at bottom) are cells that intrinsically respond to light. Rods and cones are shown at top. (Ramon y Cajal, 1900)

When rods and cones degenerate in diseases like retinitis pigmentosa or Leber congenital amaurosis, blindness results. The remaining cells in the retina do not have visual pigments, so light passes through them without consequence — or so textbooks held for generations.

The idea that light could affect a retina that lacked rods and cones was heretical. A third photoreceptor would have to exist in the eye. Yet entire lineages of superb anatomists and physiologists had scrutinized the retina without seeing any hint of one.

A third photoreceptor?

Robert Lucas, while a postdoctoral fellow with Foster, strengthened the case for a third photoreceptor. He studied a visual function that persisted in rodless/coneless mice, and whose simplicity invited precise analysis: pupillary constriction. In 2001, Lucas reported that the pupil’s sensitivity to light could not be explained by the rod and cone pigments. It seemed likely that the mysterious photoreceptor used an undiscovered pigment, one that is most sensitive to a particular hue of blue light.

There was no easy way forward from here. Sifting randomly through the retina for a new photoreceptor was impractical.The retina contains more than 100 types of cells, many of which remain poorly understood: they cannot be identified simply by microscopic examination, do not have obvious molecular signatures and lack apparent counterparts across species. Progress, it turns out, came from an unlikely quarter.

The discovery of ipRGCs

Certain cells in the skin of frogs, the melanophores, darken when illuminated. Ignacio Provencio and colleagues, at the Uniformed Services University of Health Sciences in Bethesda, sought the molecular trigger for this process. In 1998, they described a messenger RNA (mRNA) that acts as the blueprint for a new pigment they called melanopsin. Soon thereafter, they described a similar pigment within a small number of RGCs in mammals, including humans.

the eye is also the center of non-image vision
Male face with human eye antomy

Further north, a venturesome graduate student named Joshua Gooley was beginning a rotation with Clifford Saper at Harvard Medical School. Gooley decided to ask if cells that had melanopsin mRNA were wired up to the brain’s master circadian clock, a tiny region called the suprachiasmatic nucleus (SCN). They were. If these cells were photosensitive, they could explain the ability of rodless/coneless individuals to synchronize their circadian clocks with the solar day.

Are they photosensitive? That question was the subject of heroic experiments by David Berson and his lab at Brown University. Step one was precision brain surgery to inject a fluorescent dye into the SCN of rats. This dye eventually traveled up the optic nerve to mark the retinal cells of interest. The investigators then removed the retina, kept it alive, placed an electrode onto a marked cell and gave it light.

The RGCs were activated — remarkably, even when signals originating with the rods and cones were blocked. Moreover, the cells were best activated by the same blue light as the pupillary constriction of rodless/coneless mice. Berson and colleagues called these cells intrinsically photosensitive RGCs (ipRGCs).

A wide photoreceptive net

Concurrently, antibodies were being made that allowed the melanopsin protein to be visualized. Using them, King-Wai Yau’s group at the Johns Hopkins University School of Medicine found melanopsin in about 800 of the 50,000 total RGCs in the mouse retina. The rarity of these melanopsin-bearing cells explains why they were so long overlooked. Although sparse, the cells sent forth long, tortuous, melanopsin-rich dendrites that overlapped extensively and spanned the entire retina. Provencio, making a similar observation in the rat retina, called it a photoreceptive net. Working together, the Berson and Yau groups verified that ipRGCs and melanopsin RGCs were one and the same.

Samer Hattar, then a postdoctoral fellow with Yau, showed that ipRGCs send axons through the optic nerve to connect with far-flung brain regions. Regions controlling the circadian clock and pupillary constriction were major targets. Other targets were involved in a diversity of functions, including acute sleep/wake regulation, emotion processing and pain modulation. IpRGC axons are now thought to reach dozens of brain targets, forming an extensive system of light detection whose impact remains largely unexplored.

non-image vision

Making use of melanopsin

An explosion of surprising findings has followed. For example, Richard Lang and colleagues at the Cincinnati Children’s Hospital showed that embryonic mice use melanopsin to sense light in the womb and that this process is required for proper retinal development.

Another surprise is that melanopsin, when put into practically any cell, renders it light sensitive (see here, here and here). Harnessing this phenomenon, melanopsin has been used for the artificial control of hormone secretion and heart rate. Richard Masland (Harvard Medical School), Satchidananda Panda (Salk Institute) and their teams delivered the melanopsin gene into the retinas of rodless/coneless mice, endowing RGCs with artificial photosensitivity and rudimentary visual function — enabling the mice to navigate toward a visual target. In this way, a protein found by studying the blind has become a candidate for restoring sight.

How ipRGCs sense light for non-image vision

Subconscious sight is referred to formally as “non-image-forming” vision, stressing the idea that ipRGCs sense the overall level of environmental illumination, or irradiance. Irradiance is what matters most for control of the pupils, circadian clock, wakefulness and related functions by light. On the other side, “image-forming” or “pattern” vision resolves details in the world by detecting differences in parameters like brightness and color across space and time.

IpRGCs are highly suited to non-image-forming vision. An obvious reason is their large size. The photoreceptive dendrites of a human ipRGC can span a diameter of up to about 1 millimeter, pooling light from a roughly 300-fold larger portion of the image than the cones that support our highest visual acuity (see here and here). Chunking the scene into big pixels reduces detail — consider the coarse graphics of a vintage computer — and is one step toward encoding irradiance.

Initial recordings from ipRGCs also indicated that their responses were slow. In my own postdoctoral work with Yau, we found that the response of an ipRGC to a single photon of light has an effective duration of 8 seconds, far greater than that of rods (0.25 seconds) and cones (0.08 seconds). Imagine a camera exposure that lasts this long: as the image moves, its features blur together. IpRGCs are set to reduce detail in favor of capturing overall brightness.

In my lab at Boston Children’s Hospital, Alan Emanuel found that the responses of ipRGCs can last far longer. If enough melanopsin molecules are activated, they appear to outpace the machinery that shuts them off — like a band of toddlers with too few adults to settle them down. Consequently, ipRGCs remain activated for several minutes even after illumination has ceased and may continue to influence physiology. We also found that melanopsin responds to an unusually broad interval of the visible spectrum. Thus, ipRGCs are less sensitive to the exact timing and color of light than they are to light intensity.

Taking shifts around the clock

How do ipRGCs signal different intensities of light? As the earth turns, environmental irradiance ranges over many orders of magnitude. The original assumption was that the higher the irradiance, the stronger an ipRGC’s signal to the brain. But Elliott Milner, completing his doctoral research in my group, revealed a different strategy: ipRGCs are tuned to different stretches of irradiance, dividing labor across the population of cells to work from moonlight to full daylight.

non-image vision is conveyed around the clock

Why divide labor? One idea is that doing so saves energy; as some ipRGCs turn on, others turn off. Another is that it provides flexibility. Brain regions that mediate responses at certain light levels may collect information from particular ipRGCs. We are currently exploring the implications of using a functionally diverse population of ipRGCs for non-image-forming vision.

Some of the many open questions

This discussion has concerned ipRGCs of a particular kind, the first to be identified. But others are now recognized. What is their nature? Pioneering experiments by David Berson, Samer Hattar, Tiffany Schmidt, Kwoon Wong and others indicate that there is much to discover in this regard. The types differ in their shapes, functional properties and connections to the brain.

Furthermore, most studies of ipRGCs have used nocturnal rodents like mice and rats. How different are the ipRGCs of species like humans that operate in daylight?

And what happens when non-image-forming vision malfunctions? Circadian dysregulation has been linked to cardiovascular disease, metabolic disorder, psychiatric illness and cancer. Melanopsin variants have also been associated with cases of seasonal affective disorder. As the field moves forward, ipRGCs may prove to have other links to our health.

More research from the F.M. Kirby Neurobiology Center