Cellular Neurobiology Laboratory
Information Processing in the Retina

Dr Masland | Recent Publications | Links

Summary: Richard Masland's laboratory studies the microcircuitry of the retina.

The retina is a complex sample of neural tissue, a microprocessor located in the eye. Specialized photoreceptor cells detect light and communicate synaptically with a network of subsequent neurons, ultimately leading to the transmission of nerve impulses down the optic nerve. Along the way, the visual input is shaped and modified for transmission to the brain's higher centers.

This recoding of the visual input is surprisingly sophisticated: certain features of the visual input are accentuated, and others are downplayed. We are still unraveling the intricacies of the retina's neural codes. The retinal hardware is concomitantly complex but unraveling its intricacies is worth the effort because the retina, which is an extension of the central nervous system, is an accurate prototype of many brain nuclei-their neuronal structure is no less heterogeneous than that of the retina. The lessons learned from retinal structure, and the methods developed for its analysis, thus are applicable to other brain regions.

The Cell Populations of the Retina

About 10 years ago we set out to make a comprehensive accounting of the types of neurons that participate in the retina's computations. As a strategy, this was a distant cousin to the human genome project: the program is first to list the players-the gears and wheels that make the machine turn-and later find out how they are assembled.

The initial phase of this effort is now complete. A typical mammalian retina contains rod and cone photoreceptors, 2 types of horizontal cell, 13 types of bipolar cell (more about these below), ~30 types of amacrine cell, and a dozen types of ganglion cell. Each cell is a distinct computational element: it carries out a distinct job in the retina's circuitry. This shows, among other things, that the florid neuronal complexity described by the great turn-of-the-century anatomists is real, reflecting the actual components of an intricate circuitry. The next task is to make sense of it.

The Fundamental Plan of the Retina

To understand a retina that contains 60 different computational elements seems at first a daunting problem, but the job becomes simpler once a central rule is recognized: the output of any individual cone photoreceptor is tapped by each of the dozen distinct types of cone bipolar cell. Why is this simple statement so important? Because it creates, in the basic backbone of retinal structure, a dozen separate informational channels, each transmitting a different aspect of the photoreceptor cell's output. For example, a single cone photoreceptor (rod photoreceptor cells work on a different and simpler plan) drives about a half-dozen different types of ON bipolar cell, which increase their output when the cone is illuminated. The same cone also innervates a half-dozen OFF bipolar cells, which decrease their responses during illumination. In this way two fundamentally different codings of the visual input are created, at the very first synapse of the visual system.

Each functional class (ON and OFF) of bipolar cell is further subdivided. For example, there are cells that respond best to sudden changes in illumination and other cells that respond best to steady levels of light. To decipher further the bipolar cells' encodings of the cone's output is an important task: these dozen or so informational channels represent "primitives" of vision, the fundamental set of signals used by the retina to represent the visual world. In a television set, the world's palette of colors is represented by admixtures of three signals: red, green, and blue. In the retina, mixtures of a dozen signals are used to represent the world's panoply of brightnesses, shapes, motion, and colors. These individual signals are collected in different combinations by retinal ganglion cells, which then transmit a combinatorial message to the brain.

Signaling from the Retina to the Brain

The distinct types of bipolar cell synapse upon distinct types of retinal ganglion cell, transmitting to each that bipolar cell's unique view of the world. But the matching is not always one-to-one, and the final output of a ganglion cell is also controlled by feedback and feedforward signals from amacrine cells, an enormously diverse group of locally acting interneurons. These can create highly specialized properties in ganglion cells. For example, an amacrine cell called the starburst cell enables some ganglion cells to signal the direction of stimulus motion. (We proposed that this was the case in 1984; after much seesawing of the evidence, it was finally proved correct during the past year.) Thus, ganglion cells mix and match among inputs from bipolar and amacrine cells to create the retina's final array of encodings of the visual world.

To which features of the visual input are the ganglion cells sensitive? Remarkably, the answer is only partly known. We recently surveyed the structural types of ganglion cell present in the rabbit, a fairly representative mammalian retina. We used four independent methods of identifying the cells, and each method revealed 12 morphologically distinct ganglion cell types. Only about half of these cell types have been well characterized with respect to their coding of visual stimuli.

What Does the Retina's Structural Complexity Mean for Vision?

Our new understanding of the retina's structure points to computational sophistication, and a spate of recent studies from our lab and others document unsuspected subtleties in the retina's processing of visual information. Some of the most dramatic effects are mediated by wide-field amacrine cells, which cause the retina's messages to the brain to depend not only on the immediate stimulus but also on the visual context in which it occurs.

Perhaps the most fundamental issue, though, is to learn how the codings represented by the dozen or so retinal ganglion cell types are matched to the structure of the natural visual world. We seek here a more precise and powerful "language of vision" with which to parse the world of visual inputs and match them to the properties of the different ganglion cell types. The circuitry of retinal ganglion cells did not evolve in a vacuum; it evolved in response to the set of objects present in the natural world. We hope that the ganglion cell structures, which define the discrete coding streams into which the visual input is parceled, will point us toward the fundamental visual code.

A Great Three-Dimensional Jigsaw Puzzle

Although the central organizing principle of the retina now seems clear, we have far to go to learn how specific microcircuits function. A classic example is the direction-selective ganglion cell, but it is also safe to say that the circuitry for even the simplest ganglion cell types remains obscure in detail. We have the list of players, but can only begin to see how they are wired.

The daunting task of deciphering that wiring can hardly be approached by classical techniques. Recently, however, genetic introduction of marker compounds and transneuronal tracing methods have made a connectivity map of the retina an attainable goal. In addition, new mouse strains reliably and reproducibly express fluorescent proteins in defined subsets of neurons. These allow certain types of dual-fluorescence experiments that would otherwise be impractical.

Neurosurgery Clinical Units MGH Neurosurgical Service Spine Tumor Center at MGH Brain Tumor Center at MGH Research at MGH Neurosurgery Neuroscience Care Units at MGH Residents at Neurosurgery