Marc: Functional Neuroanatomy of the Retina

Download 0.58 Mb.
Size0.58 Mb.
  1   2   3   4   5   6   7   8

Marc: Functional Neuroanatomy of the Retina

Functional Neuroanatomy of the Retina

Robert E. Marc

Dept. Ophthalmology, Moran Eye Center, University of Utah;

65 Mario Capecchi Dr., Salt Lake City 84132 UT

PHONE 801-585-6501

FAX 801-587-7724
Manuscript checklist:

  1. Complete MS in *.doc, *, *.pdf and *.rtf format at URI:

  1. Letter of Agreement: *.pdf at URI:

  1. Permissions: Only permissions from the previous version are needed. No new permissions.

  2. Figures: 32 figures: all 1200 dpi, CMYK, LZW compressed TIFFs at URI:

  1. Tables: 2 included in the manuscript

  2. References: 264


In the eight years since Dr. Paul Witkovsky reviewed retinal anatomy and function for this series, a sea change has occurred. We have progressively uncovered detailed molecular architectures for many structures and processes. We have concretely linked circuitry and retinal disease. There is now an overwhelming amount of literature to be screened. A PubMed search on the topic of “retina AND neuron” over 2001-2008 retrieves > 10,000 papers. If you started today and read all day, every day without fail, it would take two-to-three years to catch up to 2008, and then you would still be three years behind an exponentially growing field. After severe winnowing the reference list has grown to over 250 papers, yet it is certain that I have neglected key publications. I apologize to those authors in advance. What justifies this expansion? First, we have now assembled a nearly complete catalogue of cells in the retina (perhaps 90%) and have learned a tremendous amount about neuronal phenotypes and connections. We simultaneously know more and, paradoxically, understand less about primate color coding mechanisms. New retinal cell types have been discovered and refined models of synaptic signaling have emerged. Our understanding of the molecular mechanisms of synaptic function, of neurotransmitter receptor molecular biology, of modulatory mechanisms, and gap junctions has exploded. Further, we now have strong evidence of postnatal and disease-induced neuroplasticity in the mature retina. Some problems persist. We still do not know how horizontal cells (HCs) work; how red/green color coding happens; why we need so many kinds of bipolar cells (BCs) and amacrine cells (ACs); how the retina develops (though great advances have been made); what retinal efferents do; nor the exact details of any ganglion cell (GC) micronetwork. The study of retinal structure remains a dynamic, challenging enterprise.


The vertebrate retina encodes visual scenes in starlight and at the solar zenith, discriminates spectral reflectances, detects motion, outlines form, and even mediates non-imaging-forming light encoding in some species. Most vertebrate retinas are duplex, using rod photoreceptors for nocturnal scotopic vision and cones for diurnal photopic vision. Our focus will be the mammalian retina in general, and the primate retina in particular. While the mammalian retina is complex, it is a reduced and re-derived vertebrate retina with less neuronal diversity and sensory bandwidth than those of avians, reptiles, amphibians and fishes. The evolutionary mechanisms underlying this reduction are beyond the scope of this chapter, but every mammalian attribute ought to be viewed in the light of this evolutionary transformation.

Functional neuroanatomy addresses not only the neuronal architecture of signal processing, but also the synaptic connectivity, network topology, and signaling biophysics of retinal networks. The retina has been the focus of intense investigation since the earliest days of neuroanatomical research. The great Spanish neuroanatomist Santiago Ramón y Cajal, who shared 1906 Nobel Prize for Medicine with Camillo Golgi (and whose methods he used) set forth reasons for the attractiveness of the retina as an experimental tissue in the introduction to his classic study (1): the basic flow of information from photoreceptors toward ganglion cells was largely understood; the retinal neurons were arrayed in well-defined cellular layers; their contacts with other neurons were separated into clear zones (the inner and outer plexiform layers); and the compact nature of a cell's dendritic and axonal arbors facilitated study of its nervous connections. Ramón y Cajal considered the retina to be a true nervous center, but one whose thinness and transparency made it ideal for histologic analysis.
We still find this framework essential, but there is much molecular and structural information to add. In addition to new imaging and electrophysiological tools, we add increasingly detailed descriptions of molecular networks that participate in retinal signaling, growth, normal and pathologic function, and cell death. One important and clinically potent finding is that the nervous system, including the retina, is highly plastic. In addition to the basic “hard-wiring” of the system, parallel arrays of neuromodulators (2, 3) such as peptides, amines, metabolites and even free gases modify the properties of circuits. Neuromodulation is predominantly linked through transmembrane signaling systems, usually G-protein coupled receptors (GPCRs) whose transduction networks converge on a variety of intracellular proteins (e.g., kinases), which exert subtle control over target receptors and channels. In most cases neuromodulators reach their targets by diffusion from distant sources (4). Of high clinical import for translational work is the discovery that the retina is plastic under pathologic conditions, displaying structural remodeling, physical rewiring and molecular reprogramming, requiring a new neuroscience-based understanding of these diseases. Neural circuitry is no longer simply the playground of the cognoscenti but a fundamental part of medical praxis.
Ramón y Cajal (1) and Stephen Polyak (5) catalogued the fundamental shapes and laminar locations of a given cell's dendritic and axonal arborizations. To this we now add the display of electrical and chemical synapses among neurons, neurotransmitter and receptor expression patterns, new mechanisms of signal integration, plasticity / adaptation and developmental history. This chapter reviews many of these topics and (i) considers persistent problems, (ii) addresses revisions in our thinking about specific networks, and (iii) summarizes the implications of plasticity for retinal disease.
General Organization of the Retina

The retina is a heterocellular (Fig. 1A,B) collection of interacting cellular systems and is assembled from three developmentally distinct neuron-like groups (Table 1). Superclass 1, the sensory neuron phenotype, is a superset of rod and cone photoreceptors and BCs, all characterized by polarized epithelial forms with apical ciliary-dendritic and basal axonal-exocytotic poles (6). These cells uniquely use high fusion-rate synaptic ribbons as their output elements. Superclass 2, the multipolar neuron phenotype (7), is a superset of ACs, axonal cells (AxCs) and GCs characterized by numerous branching neurites (often separable into dendrites and classical axons) and classical CNS Gray type I and II synapses. Superclass 3 contains the gliaform cell phenotype (7) and the superclass of HCs. Though they are multipolar and may display axons, they do not spike and also express many otherwise uniquely glial attributes. A complete vertebrate retina also requires two traditional classes of glial cells radial Müller’s cells (MCs) and astrocytes (AsCs). Mammalian retinas are also unique in expressing retinal vascularization (Fig. 1C). This requires the migration of vascular endothelial phenotypes and pericytes into a permissive retina and the molecular mechanisms of this process remain unknown. Finally, vertebrate retinas contain populations of surveillant microglia (8, 9).

[-------------------------------BEGIN FIGURE 1 INSTRUCTIONS--------------------------]
Figure 1: CMYK 1200 dpi 3.5”w LZW compression

Filename: Marc_Duanes_FNAR_Fig_01_1200dpi_CMYK_LZW.tif

Filesize: 24.6 Mb


Panel A: recover original Witkovsky figure DA2DB1DC1C20FF1 part A, scale to 1.2”w x 2.7”h, insert as shown

Panel B: recover original Witkovsky figure DA2DB1DC1C20FF1 part B, scale to 1.7”w x 2.2”h, insert as shown

Panel C: As is.
[-------------------------------END FIGURE 1 INSTRUCTIONS--------------------------]
Like other complex CNS assemblies, neuronal populations of the retina are segregated into distinct layers of cell somas (nuclear layers) with interposed layers of synaptic connections (plexiform layers). The outer nuclear layer contains the nuclei of rod and cone photoreceptors embedded in a meshwork of distal MC processes, and constitutes the nuclear zone of the image-forming sensory retina. The inner nuclear layer is the distal part of the true neural retina and contains four cell groups layered in distal-to-proximal order: HCs, BCs, MCs and ACs (Fig. 1C). The GC layer contains the somas of the CGs, the true projection neurons of the retina, which decode BC signals and and re-code them as spike trains. These packets of coded information are distributed by GC axonal projections to thalamus and midbrain. Interposed between the ONL and INL is a thin outer plexiform layer containing the synaptic output of photoreceptors and the dendrites of BCs and HCs. Interposed between the inner nuclear and ganglion cell laters is a thick inner plexiform layer containing the axonal outflow of BCs, the dendrites of GCs and the dendrites and synaptic output of diverse classes ACs and AxCs. Together, these elements form the basic set of components for vision (Fig. 2).
[-------------------------------BEGIN FIGURE 2 INSTRUCTIONS--------------------------]
Figure 2: CMYK 1200 dpi 7”w LZW compression

Filename: Marc_Duanes_FNAR_Fig_02_1200dpi_CMYK_LZW.tif

Filesize: 19.7 Mb


[-------------------------------END FIGURE 2 INSTRUCTIONS--------------------------]
Photoreceptors form the sensory retina. Their distal ciliary extensions form the light-harvesting outer segment layer in mammals (Fig 3), and their proximal somatic and axonal extensions form the neural interface between photoreceptors and the afferent neural chain to the CNS. There are three classes of photoreceptors: (i) rods , (ii) long-wave system (LWS) cones and (iii) short-wave system 1 (SWS1) cones. Each class displays a distinct morphology as well as visual pigment (Fig. 3), but the full array of genes that confer rod, LWS or SWS1 cone identity remains unknown. Visual pigments (opsins) are GPCRs that bind 11-cis retinaldehyde as their ligand in mammals. Rods alone express the RH1 visual pigment rhodopsin that absorbs maximally at 499 nm (VP 499). LWS cones express either red (R) VP 560 or green (G) VP 530, via a yet uncertain semi-stochastic switch (10). There are no known gene expression differences in LWSR and LWSG cones other than the visual pigment. Indeed, LWSR and LWSG cones are morphologically indistinguishable. Conversely, SWS1 cones differ from LWS cones in subtleties of shape (they are slightly longer and slimmer), connectivity (fewer ribbons) and other gene expression patterns (7). Human or primate SWS1 cones express VP 420 and are also referred to as blue (B) or short-wave (S) cones. Similarly, R cones are also termed L or long-wave cones and G cones termed M mid-wave cones. We will generally use the R, G, B notation as it conforms to the psychophysical color percepts the cones drive.

[-------------------------------BEGIN FIGURE 3 INSTRUCTIONS--------------------------]

Figure 3: CMYK 1200 dpi 3.5”w LZW compression

Filename: Marc_Duanes_FNAR_Fig_03_1200dpi_CMYK_LZW.tif

Filesize: 25.5 Mb


[-------------------------------END FIGURE 3 INSTRUCTIONS--------------------------]
The light-driven switch that activates signaling is the absorption of photons by 11-cis retinaldehyde-opsin to form all-trans retinaldehyde-opsin, which activates binding of the ATP-activated form of the G-protein transducin to opsin and triggers a transduction sequence that leads to closure of cation permeant channels on the photoreceptor outer segment. This closure represents a conductance decrease to a cation with a positive reversal potential and is manifest as photoreceptor hyperpolarization. This voltage change directly gates the rate of synaptic glutamate release by photoreceptor terminals. When depolarized in the dark, photoreceptors release maximal amounts of glutamate; when hyperpolarized in light, release is attenuated.
BCs decode photoreceptor glutamate signals and re-code them as their own voltage-driven glutamatergic synaptic outputs (reviewed in reference 3). Generic synaptic transfers (→) in retinal networks are classified as sign-conserving (>) or sign-inverting (>i). Sign-conserving synapses nominally copy the voltage patterns of the presynaptic cell to the postsynaptic cell. Sign-inverting synapses nominally invert those patterns. Such synapses also possess differences in kinetics and amplification based on their molecular targets. All vertical channel (rod, cone, BC) synapses use glutamate as their transmitter, provide high amplification (greater than 1), and most are sign-conserving. Thus the cone > HC synapse is a quintessential sign-conserving synapse in which HCs mirror the behavior of cones through the readout of glutamate fluctuations by ionotropic glutamate receptors (iGluRs) of the AMPA type: When cones hyperpolarize to light, so do HCs. There are two kinds of BCs: OFF and ON. Cones drive OFF BCs via high gain sign-conserving AMPA or kainate (KA) IGluRs. One of the most unique synapses in the CNS is the high-gain, glutamatergic, sign-inverting synapses between photoreceptors and ON BCs, mediated by a metabotropic glutamate-binding GPCR known as mGluR6. This special high-gain, inverting transition will be symbolized >m .
The fundamental glutamatergic signal flow of photoreceptors → BCs → GCs → LGN neurons → cortex in mammals (a prototypical CNS projection chain) is also shaped by sign-inverting, low-amplification (gain less than 1) feedback and feedforward micronetworks at every synaptic transfer. Most of these feedback / feedforward events in the inner plexiform layer are mediated by classical inhibitory transmitters: 4-aminobutyrate (GABA, γ) and glycine (gly). In the outer plexiform layer, the molecular mechanisms of sign-inverting feedback is unknown, but powerful cone ⇄ HC feedback networks predominate.
Non-neural cells also play a major role in control retinal signaling, though these mechanisms are largely beyond the scope of this chapter. First, as schematized in Fig. 1, the entire retina is sealed distally from the choroidal vascular compartment by the tight-junctions of the retinal pigmented epithelium (RPE) cell layer and internally mostly by the intermediate junctions of MC processes ensheathing the vascular endothelium (11). In most mammals, three capillary arcades emerge from the vitreal vessels of the ophthalmic artery: the first (c1) branches at the interface between the GC and inner plexiform layers; the second (c2) at the interface between the AC and inner plexiform layers; and the third (c3) exactly at the level of the HCs cells in the outer plexiform layer. There are some important exceptions. Like non-mammalians, retinas of lagomorph mammals (rabbits and hares) are avascular. Order Sciuridae (squirrels) lack the c3 arcade. Importantly, the blood-retinal barrier in arcades c1 and c2 are provided almost exclusively by MCs, whereas HCs and MCs cells form roughly equal contact zones in the OPL at arcade c3 in tree shrews (12). Whether this is generic for primates has not yet been established, but suggests that HCs have glial-like endothelial recognition systems. This is intriguing for two reasons. First, only mammals display retinal vascularization and recent discoveries of secreted VEGF inhibitors that maintain corneal clarity (13) suggests that similar molecular mechanisms may have played a role in evolutionary regulation of retinal vascularization. Second, the tolerance of the mammalian retina for vascular cells also invites pathological over-invasion, such as in neovascular macular degeneration. A recently characterized form of neovascularization apparently arises from arcade c3 and is termed retinal angiomatous proliferation (RAP) (14). It is plausible that RAP involves loss or modulation of HC. AsCs reside in the layer of optic nerve fibers and sometimes around retinal capillaries in arcade c1.
Retinal Patterning

Retinal photoreceptors convert a photon-based image into a mixed array of photoreceptors generating primary photocurrents transformed into synaptic drive (15). We can think of this drive as a synaptic image. The cone mosaic of humans and primates is patterned (Fig. 4) in a manner roughly analogous to the Bayer pattern of digital color CCD arrays (16, 17). While not tiled like non-mammalian cone arrays (18) and nearly random in organization, the cones of primates are differentially expressed (17, 19) so that B cones comprise a minority: about 7% of the human mosaic (20). The remaining cones are a mix of R and G, the ratio varying widely across individuals, ranging from a high of 16.5:1 to about 1:1 in males, and 0.37:1 in some females (21). In primates, the photoreceptor layer is dominated by rods except in the central 2 degrees of visual angle, where cone density rapidly rises to a sharp peak at 160,000 cones/ mm2 (Fig. 5) and the rod density drops to zero in the foveola (22), a region about 0.25 degrees in diameter. In the periphery (beyond 20 degrees), the cone density drops to 5000 cones/mm2. Importantly, the cone inner segments (the light capture compartment) become proportionally larger in the periphery, increase diameter from about 2 μm in the foveola to 7-8 μm beyond 20 degrees. Thus the coverage of image space by cones smoothly decreases from 100% to no less than 30%. This is quite different from rodent retinas where the cone density may be higher (> 10,000 cones/mm2) than peripheral primate retina but image coverage never exceeds 3% (23). Cone density is less important than the coverage.

[-------------------------------BEGIN FIGURE 4 INSTRUCTIONS--------------------------]
Figure 4: CMYK 1200 dpi 3.5”w LZW compression

Filename: Marc_Duanes_FNAR_Fig_04_1200dpi_CMYK_LZW.tif

Filesize: 5.3 Mb


[-------------------------------END FIGURE 4 INSTRUCTIONS--------------------------]
Neurons are also patterned on global and local scales (24, 25). Local patterns are termed mosaics (Fig. 5) and each cell class possesses its own own portion of a segmented synaptic image. The functional attributes of these mosaics in forming synaptic images are determined by several features of individual cells: their abundance, regularity of spacing, dendritic overlap and synaptic density. Dendritic overlap and abundance are often combined into a dimensionless coverage factor (CF): the product of the projected planar area subtended by the dendritic arbor of an average cell (mm2/cell) in the class and its spatial density (cells/mm2). Three distinct pattern classes (15) or mosaics are found in the vertebrate retina and are characterized by their overlap features. Packings are mosaics with no overlaps allowed, though there may be gaps. Photoreceptors form packings with CF < 1. In other words, each point in visual space is sampled by only one photoreceptor, which necessary excludes all other photoreceptor classes. Coverings are mosaics that allow no gaps, but permit overlap. ACs overlap their dendritic arbors to form coverings with CF >>1. Perfect tilings are idealizations that admit neither gaps nor overlaps and CF = 1. The classes of the GC cohort roughly approximate tilings. Thus the initial synaptic image is fractionated into rod, LWSR, LWSG and SWS1B packings and these collections partially smoothed by tiles of each GC class, with further smoothing by high-coverage factor ACs and HCs. Paradoxically, the dendritic arbors of HC have small overlap compared to ACs and more resemble CNS astrocytes. However, a powerful mechanism generates a high physiological coverage factor: connexin-based coupling. This will be addressed in detail later.

[-------------------------------BEGIN FIGURE 5 INSTRUCTIONS--------------------------]

Figure 5: CMYK 1200 dpi 3.5”w LZW compression

Filename: Marc_Duanes_FNAR_Fig_05_1200dpi_CMYK_LZW.tif

Filesize: 2.6 Mb


[-------------------------------END FIGURE 5 INSTRUCTIONS--------------------------]
Local patterning is a measure of the spatial precision of a class. Spacing in a class can be statistically orderly, ranging from nearly perfect crystalline patterns such as cone (26) and BC patterns in fish eyes (27), to random as shown by AxC classes such as interplexiform cells (28). For the latter and for neurons that have high coverage factors, the spacing is measured as the distance between somas, which is a rough gauge of the center of mass for a mosaic element. The precision of spacing is gauged several ways, the most common being the non-dimensional conformity ratio (CR)(25, 29), also known as the regularity index (24): the ratio of the mean minimum distance between cells to its standard deviation. Thus highly precise patterns with little spacing variability have large CR values and random ones have CR of about 1. Normally, CRs of mammalian retinal neurons rarely exceed 4, while non-mammalian fish photoreceptor CRs can approach 30. Such high precision in patterning is likely closely associated with velocity detection in more complex visual environments than experienced by mammals.
Global patterning rules impact retinal function as cell abundance can vary with eccentricity, forming central zones with class mixture specialized for acuity and chromatic processing (7, 25). In the humans, the retina possesses a rod-free declivity where neuronal cell bodies (intense Rayliegh scatterers) are displaced in an aster pattern around the foveola and cone diameters decrease to about 2.5 um, achieving densities of 160,000 cones/mm2 (22). The peak rod density is found at about 20 degrees parafoveally in nasal and temporal retinas (Fig. 5). Thus, cone-selective BCs and GCs are at their highest densities in central retina, while rod-specific neurons are concentrated in an annulus around the fovea. Other highly visual mammals such as cat and rabbit lack optically optimized foveas but still display elevated concentrations of cones in central regions in which cones are found at a higher density, though lower than in primates. A correlate of this density gradient is that, like cones, most neurons are smaller in the central retina (packed more densely) and larger in the periphery (25), including HCs, rod BCs (30), GCs of the primate retina (31), and ACs (32). Figure 6 illustrates the variation in starburst AC size with increasing eccentricity. Correspondingly, larger dendritic arbors provide larger physiological receptive field sizes (33)and decreases in the resolving power in peripheral retina.
[-------------------------------BEGIN FIGURE 6 INSTRUCTIONS--------------------------]
Figure 6: recover original Witkovsky figure DA2DB1DC1C20FF2
[-------------------------------END FIGURE 6 INSTRUCTIONS--------------------------]

Share with your friends:
  1   2   3   4   5   6   7   8

The database is protected by copyright © 2019
send message

    Main page