Advances in basic retinal anatomy, genetics, biochemical pathways and neurochemistry have not only provided a better understanding of retinal function but have also allowed us to link basic science to retinal disease. The link with disease allowed measures to be developed that now provide an opportunity to intervene and slow down or even restore sight in previously ‘untreatable’ retinal diseases. One of the critical advances has been the understanding of the retinal amino acid neurotransmitters, related amino acids, their metabolites and functional receptors. This review provides an overview of amino acid localisation in the retina and examples of how retinal anatomy and amino acid neurochemistry directly links to understanding retinal disease. Also, the implications of retinal remodelling involving amino acid (glutamate) receptors are outlined in this review and insights are presented on how understanding of detrimental and beneficial retinal remodelling will provide better outcomes for patients using strategies for the preservation or restoration of vision. An internet-based database of retinal images of amino acid labelling patterns and other amino acid-related images in health and disease is located at http://www.aminoacidimmunoreactivity.com.
The aim of this review is to provide an overview of basic retinal neurochemistry, focusing on the amino acid neurotransmitters and metabolites (glutamate, γ-amino butyric acid [GABA], glycine, aspartate, glutamine) and osmoregulators (taurine), synthesising and degradation enzymes, (glutamic acid decarboxylase [GAD], glutamine synthetase [GS]), glutamate receptor subunits and functional activation. The extensive literature on this topic means that it is not possible to do justice to all published manuscripts. The focus of this review is to highlight the conservation of core amino acid neurochemistry across different vertebrate retinae and provide the foundations to understand retinal disease. This approach allows a holistic appreciation and an understanding of the vertebrate retina in health and disease, based on core knowledge of retinal neurochemistry, anatomical and functional data.
The internet-based archive located at http://www.aminoacidimmunoreactivity.com, provides an opportunity to undertake general or specific searches on a variety of retinal topics. Images can be identified, often with the corresponding figure legends, by undertaking a search using the following parameters:
Supplementary Figures 1 to 5 outline the phylogenetic classification for different species available in this review and the internet-based archive.
Basic neurochemistry and anatomy (a clinical perspective)
The vertebrate retina has a common structure that displays species-specific specialisations. Photoreceptors capture light and through a series of interactions within the retina, the fundamental encoding of the spatio-temporal and chromatic information is achieved and transmitted to other parts of the visual processing centres. The basic retinal architecture involves photoreceptors-bipolar cells-ganglion cells forming the ‘through’ retinal pathway associated with ‘lateral’ elements (horizontal cells, amacrine cells and interplexiform cells) interacting at the two retinal synaptic layers.[1-3]
The three spectral classes of cone photoreceptors in the primate retina synapse onto two distinct sets of bipolar cells: the depolarising bipolar cells (ON bipolar cells) and the hyperpolarising bipolar cells (OFF bipolar cells[4, 5]). The two bipolar cell types subsequently synapse with different types of ganglion cells to form two major streams of visual information flow: ON and OFF ganglion cells (Figure 1). Although the terminology of ‘ON’ and ‘OFF’ is commonly used to describe bipolar cell and amacrine cell responses, it was originally reserved for ganglion cell responses. Turning the light ‘ON’ or ‘OFF’ led to an increase in spike firing rate in the corresponding type of ganglion cells.
The encoding of visual information occurs through synaptic interactions at each of the two synaptic layers: the outer plexiform and inner plexiform layers (Figure 1). It is at these synaptic layers that the lateral elements (horizontal cells in the outer plexiform layer and amacrine cells in the inner plexiform layer), interact with the through pathway providing the feedback to encode spatial, chromatic and temporal information.[1, 7-10]
The mammalian scotopic system has only one type of bipolar cell: a depolarising or ON bipolar cell. Rod bipolar cells synapse on a unique type of amacrine cell (the AII amacrine cell) that subsequently conveys the scotopic information to the ON and OFF cone system (Figure 1). Thus, AII amacrine cells subserve a similar function to bipolar cells as part of the through pathway. There are no unique rod ganglion cells and thus, the mammalian rod circuitry ‘piggy backs’ onto cone circuitry to transmit scotopic information.
Clinical manifestations provide evidence of a single bipolar cell type in the scotopic system as well as its dependence upon cone circuitry to put out its signal. In the paraneoplastic syndrome, melanoma associated retinopathy (MAR), photoreceptors are intact yet nyctalopia results due to systemic antibodies destroying rod and cone ON bipolar cells by targeting the unique cation channel (transient receptor potential cation channel subfamily M member 1 or TRPM1) located on these cells.[12-14] Although the OFF cone pathway remains functional, the patients are afflicted with night-blindness as the rod photoreceptors lose their sole connection with the inner retina.
When ganglion cells are damaged in glaucoma or non-glaucomatous optic nerve damage, dark adaptation is affected, resulting in greater threshold elevation for the rod pathway, as well as elevated cone thresholds. In the clinical cone adaptation recovery test (photostress test), patients with glaucoma displayed a significantly longer recovery time (approximately 70 ms) compared to controls (approximately 42 ms). Thus, any disease where ganglion cell dysfunction occurs will also result in rod and cone visual dysfunction.
Non-mammalian retinae display similar cellular physiological responses to those of mammalian retinae, although one unique feature is that bipolar cells carry mixed rod/cone signals and synapse directly onto ganglion cells.[1, 3, 17] Subsequent to retinal processing, ganglion cells project to a multitude of locations within the central nervous system.
It is estimated that there are approximately 60 different cell types in the mammalian retina.[18, 19] The classical studies using Golgi impregnation by Ramón y Cajal provided the first retinal morphological and morphometric information, which has been expanded with the application of specific cell markers, where whole populations of neurons are labelled.[21-23] New analytical methods of cell imaging and visualisation using self-reporting methods (for example, green fluorescent protein), have further transformed our understanding of retinal structure/function relationships.[24, 25]
The early anatomical work has been combined with cell marking techniques and physiology to identify adult cellular architecture and retinal developmental characteristics.[21, 23, 26-30] Despite the diverse morphology of individual cells in the vertebrate retina, there is an overall conservation of core neurochemical architecture, particularly in the amino acid neurochemistry. This is not surprising, given that amino acid manufacturing and degradation pathways are preserved across different vertebrate species. The study of amino acids forms a core foundation in our knowledge base to complement that of retinal structure, metabolism and neurotransmission.[31-40]
The perplexing outcome—a successful operation but poor recovery of vision
In rhematogenous retinal detachment, the retina is deprived of one of its major blood supplies, the choriocapillaris, as well as disruption of other metabolic and cellular functions (Figures 2A, B and C). Various surgical techniques are available to treat retinal detachment with a high rate of success. Thelen and colleagues reported a surgical success rate of over 83 per cent for the combined group (macula on and macula off). Pastor and colleagues achieved an surgical reattachment success rate of approximately 95 per cent in a study of 546 patients. In this study, the majority of patients with reattached retinas at three months (n = 517) had a poor visual outcome: 21.5 per cent worse than 6/30, 36 per cent between 6/15 and 6/30 and only 42.7 per cent had visual acuities of 6/12 or better. The best predictor of final visual outcome appears to be presenting visual acuity before surgery. Thus, despite a high rate of surgical success, visual outcomes appear to be poor, even when surgery was performed very soon after detachment.
Although initial retinal hypoxia and death of some photoreceptors occur before surgery, the sequelae of anatomical and functional detrimental retinal remodelling are a possible reason for the poor visual outcome following surgical reattachment.[44-48] The concept of retinal remodelling explains not only the poor outcomes when some vision preservation/restoration therapies are attempted in patients with inherited retinal dystrophies but is also applicable to other forms of retinal degeneration.
The need to understand retinal remodelling
Retinal neuronal cell death and remodelling occur in a variety of ocular conditions, leading to irreversible visual loss in conditions such as age-related macular degeneration, optic nerve disease including glaucoma, and retinopathies including diabetic retinopathy.[47, 50-65] Some conditions result in long-term visual affliction (such as inherited retinal dystrophies, including retinitis pigmentosa and cone-rod dystrophies such as Stargardt's disease [Figures 2D and E]).[66-69]
Currently, there are some exciting developments in treatment options for inherited retinal dystrophies. Although associated with controversy, nutritional supplements may slow the progression of some forms of inherited retinal dystrophies (that is, vitamin A for retinitis pigmentosa). Through basic scientific research, we are also aware of the potential dangers of ingesting vitamin A in other retinal conditions, that is, in Stargardt's disease and related cone-rod dystrophies and in age-related maculopathy. The use of gene therapy,[72-78] stem cells[79-82] and nutritional supplements or pharmacological agents[74, 83-86] is restoring or maintaining sight in both humans and non-human models of retinal disease.
Retinal remodelling may restore visual function (beneficial remodelling). In animal models, beneficial anatomical remodelling occurs with photoreceptor migration, particularly when low-energy laser photocoagulation is used.[87, 88] The photoreceptor migration to fill in the damaged area appears to be dependent upon retinal pigment epithelial viability, whereby their migration into damaged areas allows subsequent photoreceptor filling in and presumably a reduced area of visual dysfunction. In human studies, the use of small, short (20 milliseconds) pulse laser lead to restoration of anatomical photoreceptor characteristics and overall sustained healing response over six months.
The future is promising for many retinal conditions that were thought to be ‘untreatable’ a decade or so ago; however, understanding detrimental retinal remodelling is crucial. This is highlighted in the disparity between high surgical success for retinal reattachment versus poor visual function post-surgery, most likely due to detrimental retinal remodelling.[41-43, 46-48]
Some intervention measures assume that neurons and neuronal circuitry ‘downstream’ of the disease locus are functionally normal. Yet, a wealth of evidence indicates that neurons and glia undergo anatomical and functional remodelling secondary to retinal disease that may impede intervention success.[47-49, 56, 57, 60, 90-97] The recent successful transplantation of photoreceptors in the adult retina occurred in an animal model of congenital stationary night blindness, where photoreceptors are intact but dysfunctional. In dystrophies where photoreceptors degenerate, the anatomical and glial remodelling, including the formation of a glial seal, would make it difficult for transplanted cells to make functional contacts. Conversely, it is also critical to understand beneficial remodelling[87-89] and use it to maximise visual outcomes for patients.
Changes associated with anatomical remodelling in the retina can be identified using a range of cell markers or using clinical imaging modalities. This knowledge is useful to characterise cell death or cell anatomical changes during disease. Functional (neurochemical) remodelling may alter neurotransmitter content, recycling and release, neurotransmitter receptor disposition or neurotransmitter receptor functionality. Functional remodelling may result in anatomically intact neural tissue displaying poor function, thereby creating a quandary in understanding structure/function relationships. If retinal circuitry is altered, any new signal from implanted cells/devices may be ‘jumbled’ and thus higher-order neurons are unable to decode, or worse, the neurons may not be able to send a signal, that is, the collision circuit theory of Marc and colleagues. Functional remodelling is a relatively new discovery and occurs in the early stages of the disease process. By identifying the time course, the specific neurotransmitter receptors affected and modifying the time over which the retina is pliable, it will be possible to improve the success of intervention measures. For example, the suitable time of intervention for procedures such as cell transplantation or retinal prosthetic implants will result in the maximum possibility of success.
The Visual Process and Neurotransmission
Localisation of glutamate, GABA and glycine
The mammalian retina is a complex array of glial and neuronal cells that convert and encode light energy into electrical impulses to begin the visual process. Glutamate is the major excitatory neurotransmitter in the retina (Figures 3A, 4 and 5).[37, 98-102] Figure 3A displays a schematic of mammalian retinal structure (for simplicity only the cone pathway is shown) and the concurrent glutamate labelling pattern (rabbit retina). The photoreceptor layer is typically glutamate immunoreactive as are bipolar and ganglion cells. Glutamate is a metabolite and precursor for GABA[103, 104] and therefore, is also localised in GABAergic and other amacrine cells. Müller cells are virtually devoid of glutamate (arrows in Figure 3A) and in the high cellular density of the central primate retina, the light band of immunoreactivity within the middle of the inner nuclear layer reflects the location of Müller cell somata (Figure 4D).
Examples of amino acid labelling from a variety of species (Figures 4, 5, 7-11) illustrate the conservation of amino acid labelling patterns but also reveal some interesting species differences. Figure 4 is the primate retina (monkey—Macaca fascicularis) and includes a Nissl stained central area (Figure 4A), and a sample away from the fovea (the low ganglion cell density but large nerve fibre layer indicates that it is close to the optic nerve head, Figure 4B), followed by immunostaining for a range of amino acids in the central retina (Figures 4C–H). Other species presented include the domestic cat (Felis catus), cougar (Puma concolour), Shingleback or stumpy tail lizard (Tiliqua rugosa), freshwater crocodile (Crocodylus johnstoni), South African clawed frog (Xenopus laevis), Port Jackson shark (Heterodontus portusjacksoni), collared peccary (Pecari tajacu) and eastern grey kangaroo (Macropus giganteus).
The cat visual system has excellent binocular vision and the physiology and neuroanatomy are well characterised.[105-108] Being a predator, the cougar has keen vision, high visual sensitivity in poor light, a rod-dominated retina, large pupil, tapetum, reduced visual acuity during day light, good depth perception and a broad field of view. The blue-tongue lizards have well developed colour vision, reaction to prey and predator and also contain oil droplets in their photoreceptors. Crocodiles are thought to possess good eyesight that senses motion, have a wide sight range, good colour vision above water but poor vision under water, as they use their nictitating membrane. The Xenopus visual system is well studied with colour discrimination possible, good motion detection and good vision at low light levels.[112, 113] Sharks have low packing density, large diameter rod photoreceptors and presumed low resolution.[114, 115] The collared peccary is thought to have poor vision in daylight and poor movement perception. The eastern grey kangaroo feeds during both daylight and nighttime, different from other kangaroos that have daytime eating habits. Although there is strong evidence that glutamate is used by photoreceptors,[35-37] considerable variations in glutamate immunoreactivity have been reported in primate tissue.[118-121] Post-mortem changes, alterations due to experimental manipulations or even light adaptation changes are some of the reasons for these differences within a single species. In monkey retina, Figure 4D shows strong glutamate immunoreactivity in most cells with the ganglion cells and their axons displaying the highest level of labelling. Photoreceptors are also strongly labelled, as are their axons (Henle's fibres, the thick band below the outer nuclear layer in Figure 4D). In contrast, Figure 5 illustrates considerable diversity in glutamate labelling of photoreceptors, from being strong in the cat, shingleback, crocodile, frog, shark and kangaroo (Figures 5A, C–F and H) to weak in the collared peccary and puma (Figures 5B and G), that likely reflects real differences in glutamate content.
Bipolar cells often display the highest level of immunoreactivity while ganglion cells may be extremely large and highly glutamate immunoreactive (for example, Figure 5G). When evident, the nerve fibre layer is also strongly glutamate immunoreactive (Figures 4D, 5A, B, D, F and H). Lateral elements most likely use glutamate as a precursor and display distinct morphology (large layered horizontal cells in the shark, Figure 5F) or smaller highly immunoreactive somata in the crocodile (Figure 5D). Müller cells display very low levels of glutamate immunoreactivity under physiological conditions.
Both ionotropic and metabotropic glutamate receptors are dispersed throughout the retina. In the outer retinal synaptic layer (outer plexiform layer; OPL), ON bipolar cells display functional metabotropic glutamate receptor 6 (mGluR6), while OFF bipolar cells and horizontal cells display functional kainate or 2-amino-3-(5-methyl-3-oxo-1,2- oxazol-4-yl) propanoic acid (AMPA) receptors. Within the inner retina, kainate/AMPA and n-methyl-D-aspartate (NMDA) receptors are located on amacrine and ganglion cells.[4, 22, 31, 122, 125-128] Distribution of glutamate receptors can be observed in Figure 6, where the small focal puncta identify the location of receptor subunits. Receptors are composed of numerous subunits that combine to impart the physiological characteristics for various neurotransmitter receptors.
The two major inhibitory neurotransmitters, GABA and glycine, are located in the inner retina (amacrine cells and interplexiform cells, although in some species GABA is also found in horizontal cells: Figure 3B). The volatility of GABA localisation is well recognised with examples of the primate retina showing very low GABA labelling (Figure 4C) in contrast to other strongly labelled examples.[121, 131] In the rabbit retina, GABA is predominantly localised in conventional and occasionally displaced amacrine cells and their processes within the inner plexiform layer (Figure 3B). GABA horizontal cell labelling is evident in five of the eight species illustrated in Figure 7, with the shingleback lizard devoid of horizontal cell GABA immunoreactivity. Amacrine cells are the dominant cell class displaying GABA immunoreactivity in all vertebrates.[32, 40, 132]
Stronger GABA labelling in central bipolar cells allowed the classification of a unique neurochemical class reflective of rod bipolar cells.[121, 133] Recent evidence suggests that these bipolar cells may tonically release GABA through reversal of a GABA transporter.[134, 135] The GABA immunoreactivity within ganglion cells and their axons, evident particularly in the crocodile and kangaroo, is most likely from leakage through gap junctions.[137-139] A solitary GABA immunoreactive ganglion cell is evident in the central ganglion cell layer of the primate (Figure 4C).
Glycine is predominantly localised in amacrine cells and ON bipolar cells (Figures 3B, 4F and 8). Localisation in bipolar cells and ganglion cells again is likely due to heterologous gap junctions.[137, 140] Also, a sub-population of interplexiform cells in the teleost fish, amphibian and avian retinae are glycine immunoreactive.[33, 34, 100, 141] The corresponding GABA or glycine receptors have been identified on bipolar cell terminals, amacrine cells and ganglion cell dendrites.[31, 126, 142, 143]
Amino Acids as Precursors and Metabolites (Glutamine and Aspartate)
Glutamate is the immediate precursor to GABA as well as a cellular metabolite[103, 104] and therefore, is localised throughout different cell classes in the retina. Although not neurotransmitters, glutamine and aspartate are amino acids critical for both neurotransmission and metabolism. Glutamine is part of the glutamate-glutamine cycle and aspartate is involved in transamination reactions with glutamate (Figures 4E, H, 9 and 10). In both cat and rat retinas, the carbon skeleton of glutamine has been localised as GABA, glutamine, glutamate and to a lesser extent, aspartate, highlighting the key role this amino acid plays in amino acid neurotransmitter homeostasis.
Glutamine and aspartate are localised throughout the retina with ganglion cells in particular displaying intense labelling (Figures 4E, H, 9 and 10). Glutamine is also localised within Müller cells and this is particularly evident in the primate, collared peccary and shark retinae. The Müller cell processes are visible in the middle of the inner nuclear layer and strong endfeet labelling (Figures 4E, 9G and F).
Anaplerotic biochemical reactions are required to replenish the carbon skeleton to the tri-carboxylic acid (TCA) cycle, when amino acid production consumes TCA cycle intermediaries.[103, 104, 146] Inhibition of monocarboxylates resulted in dramatic reduction in retinal function (assessed using the electroretinogram) and also dramatic changes in amino acid immunoreactivity. Virtually all retinal cell types showed reduced glutamate and glutamine and amacrine cells showed reduced GABA and glycine and a concurrent increase of aspartate. Aspartate levels are also influenced by physiological changes in retinal metabolic activity and can be dramatically altered in more extreme insult, when the retina is incubated under anoxic conditions.
Preventing glutamine production through enzymatic inhibition of glutamine synthetase resulted in markedly altered amino acid levels and retinal dysfunction.[101, 149, 150] The provision of glutamine completely restored function secondary to glutamine synthetase enzyme inhibition,[149, 150] whereas the provision of other substrates such as a-ketoglutarate, succinate, lactate or pyruvate only partially restored function (16 to 20 per cent). A key finding by Bui and colleagues was the dramatic change in precursor amino acids in other retinal neurons consistent with a role of provision of metabolic substrates by Müller cells. Also, several studies have shown that the retina uses amino acids or keto-acids to maintain retinal function.[31, 131, 151]
Transport (uptake) pathways for glutamate, glutamine, GABA and glycine
Glutamate and Glutamine Uptake
Neurotransmitter systems require mechanisms to rapidly deactivate active chemicals. Furthermore, compounds required for cellular function must be selectively transported into or out of a cell through a mechanism that allows the maintenance of cellular integrity. One of the ways these criteria are achieved is through the expression of unique transport proteins that undertake either low-affinity or high-affinity transport of compounds into or out of a cell.[31, 33, 34, 38, 152-155] Figures 11 (cat) and 12 (rabbit) illustrate the high-affinity uptake of various amino acids (autoradiographic distribution of tritiated amino acid uptake) and how the uptake can be modified using a variety of analogues, such as the GABA analogue (muscimol) with GABA transporter inhibitors, 4,5,6,7-tetrahydroisoxazolo [4,5-c] pyridin-3-ol (THPO) or nipecotic acid.
Both neurons and glia have uptake systems for glutamate and glutamine illustrating a diffuse labelling pattern in both cat and rabbit retinae (Figures 11A, B and 12A, B). Glutamate uptake is not restricted to neurons for use as a neurotransmitter but also involves glial cells and other neurons which use glutamate as a precursor or metabolite.[31, 156, 157] In the rat retina, Rauen and colleagues established that glial high-affinity uptake (Km = 2.1 ± 0.4 μmol/l) is the predominant deactivation pathway for glutamate. Glutamate uptake by photoreceptors through high-affinity uptake appears to be the predominant means of deactivating the neurotransmitter in the outer plexiform layer.[22, 159]
There is a group of systemic drugs that includes, N-acetylcysteine (NAC: produces cystine), d-cycloserine (binds to glycine binding site of the NMDA receptor), memantine (NMDA receptor antagonist) and topiramate (an AMPA receptor antagonist) that acts on the glutamatergic system, particularly useful in managing depressive disorders, drug addictions and potentially in the treatment of dementia.[160, 161] With the exception of d-cycloserine, the other neurochemicals affect multiple neurotransmitter systems in addition to the glutamate system and therefore, the direct link to clinical conditions is not established. Laboratory work supports the role of glutamate neurotransmission in drug reward, reinforcement and relapse. There was considerable interest in memantine as a neuroprotective agent in the treatment of glaucoma; however, there is little support that the use of neuroprotective agents, including memantine, preserves retinal ganglion cells and thus preserves vision in patients with open angle glaucoma.
Our concept of neurotransmitter release and deactivation is changing with the discovery of a glutamate-cystine transporter (Xc-) in photoreceptor terminals and evidence that supports the role of this transporter in tonic release of glutamate. This discovery of a function for a glutamate-cystine transporter in glutamate neurotransmission, may explain the pharmacological mechanism for the use of NAC to manage diseases of the central nervous system. A transporter-mediated mechanism for GABA release has been proposed for primate rod bipolar cells and in horizontal cells.[134, 163]
GABA Uptake and Clinical Significance of Inhibition of GABA Metabolism
The uptake of GABA in lower vertebrates is restricted to neurons and can be modified by the application of the GABA uptake inhibitor, nipecotic acid.[32, 164] In the mammalian retina, several GABA transporters have been identified[165, 166] and display a high-affinity uptake for GABA (Figures 11C and 12C). This uptake is modified with the addition of GABA uptake inhibitors (THPO or nipecotic acid: Figures 11D, E and 12D, E). Differences in uptake systems within mammalian retinae are illustrated by the almost complete suppression of glial GABA uptake in the rabbit (Figure 12E) but not in the cat (Figure 11E), similar to other approaches. Muscimol is not transported by the glial GABA transporter and displays a pattern consistent with exclusive uptake by conventional and displaced GABAergic amacrine cells (Figures 11F and 12F).
Altered GABA uptake or metabolism is the method used to manage a wide range of conditions, including epilepsy. One of these drugs, vigabatrin, a selective GABA-degradation enzyme inhibitor causes permanent peripheral visual field deficits, retinal electrophysiological changes and other visual disturbances.[169-171] Although the exact mechanism is unknown, the inhibited enzyme, GABA transaminase, is essential for recycling the GABA carbon skeleton (succinate) which would be an energy source for the retina. Similarly, when monocarboxylate transport was inhibited using 4-CIN, retinal function was impaired with the greatest recovery following the provision of glutamine (62 per cent recovery). Furthermore, the application of transaminase inhibitors or phosphate-activated glutaminase inhibitor totally suppressed glutamine-induced recovery. The supply of carboxylates also resulted in recovery (approximately 24 to 27 per cent). Overall, this confirms an intricate interrelationship between the tricarboxylic acid cycle and retinal metabolism and function.
High-affinity uptake of glycine is the exclusive domain of glycinergic amacrine cells[33, 34, 173] (Figure 11G). The glycine transporter (Glyt-1), the product of the LLC6A9 gene, has been located primarily in glycine immunoreactive amacrine cells in a number of species and in some where the glycinergic interplexiform cell exists, complying with the tritiated glycine uptake studies.[174, 175] The concurrent glycine labelling shown in ON cone bipolar cells is due to heterogenous coupling.[137, 140] In a knock-out mouse model, where one of the gap junctions is missing (connexin45 knock out), scotopic function was impaired and glycine immunoreactivity was eliminated in ON cone bipolar cells. These results confirm the importance of coupling in retinal function and that cone ON bipolar cells receive their glycine through heterologous coupling of retinal neurons.
Alterations in amino acid uptake
Müller cell function is altered in retinal disease. Altered amino acid uptake or degradation has been reported in glia during early retinal degeneration, diabetes or secondary to metabolic insult.[179-182] Figure 13 illustrates two examples of insult that altered Müller cell function, ultimately leading to modification of retinal function. Glutamine synthetase is a useful marker of retinal Müller cells and is localised throughout the soma (Figure 13A). Secondary to severe ischaemia, the retinal architecture is severely disrupted (Figure 13B), glutamine synthetase immunoreactivity is reduced and glutamate content redistributed.[183, 184] In addition, once Müller cells become reactive in retinal dystrophy and thus presumably alter their function, they display increased immunoreactivity to glial fibrillary acid protein (GFAP, Figures 13C and D).
Taurine is a sulphated amino acid, the function of which in the retina is not fully understood. Taurine is involved in cell volume regulation, neuromodulation, antioxidant defense, protein stabilisation, stress responses and via formation of taurine-chloramine in immunomodulation. A lack of taurine leads to severe photoreceptor degeneration.[185, 186] Taurine is also involved in rod photoreceptor production during development through interactions with a glycine receptor and a GABA receptor. Taurine levels are altered in retinal detachment, retinal degeneration or secondary to prolonged anaesthesia in mammals.
Figures 4G and 14 show the distribution of taurine in the nine species. Almost all cells are labelled, particularly photoreceptors and retinal Müller cells. A striking species difference exists, where the South African clawed frog displays little taurine labelling within the inner retina and Müller cells (Figure 14E).
Although most current neuroanatomical approaches provide analytical tools, where the integrity and density characteristics of specific sub-populations of neurons can be investigated, the sheer volume and complexity of retinal cell architecture does not allow a uniform analysis to occur of all retinal cells. Pattern recognition applied to serial-section immunocytochemistry employs computational methods identical to those applied in the analysis of gene arrays.[188, 189] The volume and complexity of information generated by gene arrays necessitates the use of clustering algorithms such as k-means and isodata approaches followed by statistical testing such as transformed divergence. Similarly, the reason for applying this approach to the retina was due to the volume of data and complexity of cell marking patterns.
Pattern recognition using small molecules as cell markers is the only method that provides a complete structural map of the vertebrate retina or brain and can classify all cells.[58, 62, 94, 95, 121, 138, 190-193] Pattern recognition using serial-section silver-intensified immunocytochemistry aims to classify data sets using rigid statistical paradigms and separates cells into distinct populations sharing a common signature (that is, a common labelling pattern). Virtually all retinal cells can be classified into taurine-rich, glutamate-rich, GABA-rich or glycine-rich cell classes, groups of cells or ‘theme classes’.
The classification uses signal correlations from n-dimensional space (‘n’ depends upon the number of small molecule labels that are used, for example, the number of amino acids). Theme classes reflect the result of the classification approach identifying pixels that have the same labelling pattern (for example, metabolic signature due to similar amino acid labelling patterns). Theme classes arise exclusively from their labelling intensity signals and do not depend on morphological identities. The anatomical location allows the cell identity to be determined through correlation of cellular position and morphometrics (the ‘truth point’ principle outlined by Sun, Vingrys and Kalloniatis). A ‘truth point’ is defined from satellite imaging analysis defined as the observation, measurement, and collection of information about the actual conditions on the ground to determine the relationship between remote sensing data and the observed objects.
Pattern recognition only classifies pixels based upon their unique signal correlations with no spatial information used in the classification. Once the classification is made, the addition of spatial information using known retinal anatomy (applying the anatomical truth point principle), allows for retinal cell types to be identified at the individual neuron level, ‘horizontal cells’ or groups of cells, ‘amacrine cells’. For example, the spatial patterns and neurochemical labelling of sub-populations of amacrine cells[195, 196] allow for the identification of these cells during functional mapping.[95, 197] Marc identified the high sensitivity of the cholinergic amacrine cells to kainate activation due to their known inner plexiform stratification pattern and somata location. Sun, Vingrys and Kalloniatis also used anatomical truth points to link theme maps and extract out the amino acid signature patterns for the displaced cholinergic amacrine cell population.
Once signal correlations are determined and theme maps completed, statistical analysis is required. The statistical analysis tests for separability of the different clusters and is not based upon differences in the ‘mean’ labelling intensity but rather in the overlap of labelled distributions. The approach is analogous to d' and associated values derived from ‘signal detection theory’. Once the clusters are identified into their different classes, the degree of overlap in their probability distributions (that is, separation index) is tested using transformed divergence DT. The criterion routinely employed is DT = 1.9 that translates to less then one per cent error in errors of classification. Thus, the probability of correct classification is 99 per cent. This criterion in ‘signal detection theory’ is analogous to having an overlap of less than one per cent between the ‘noise’ distribution to the ‘signal plus noise’ distribution in ‘signal detection theory’. As such, one can predict that out of 100 cells that have the same classification, 99 will be correctly classified, with only one cell being misclassified.
The number of theme classes that have been separable has varied among species and the number of reporting molecules that have been used. Marc, Murry and Basinger reported a total of nine theme classes using a six amino acid set in the goldfish and 14 cell classes using a seven amino acid set in the cat retina. Kalloniatis, Marc and Murry found 16 separable classes in the monkey retina with a six amino acid set.
The theme class in the adult mouse retina derived from the analysis of amino acid labelling is shown in Figure 15. Photoreceptors are separated into various theme classes reflective of outer segment, inner segment and nuclear layer. The different pseudocolours indicate that the amino acid labelling pattern is distinctly different as determined by the separability index (transform divergence). The multitude of colours within the inner nuclear and ganglion cell layers illustrates the unique amino acid labelling pattern of inner retinal neurons allowing for discrete signature patterns to be identified. Müller cell endfeet form a distinct signature (black) within the inner-most retinal layers (Müller cell somata within the inner nuclear layer were not classified). Using this approach, new insights have been provided in retinal disease.[62, 94, 95, 97, 138, 193]
In both ischaemia-reperfusion[93-95] and retinopathy of prematurity, the results of pattern recognition allowed the investigation of specific cell classes (using cell markers) or retinal function (using electroretinography). Downie and colleagues were able to show a specific loss of AII amacrine cells, indicating the potential loss of rod function in this disease. Sun and colleagues[93-95] were able to identify a dramatic loss of cone bipolar cells and dysfunctional neurotransmission between rod photoreceptors and rod bipolar cells. Further, it was only through the unequivocal identification of the amino acid signature patterns that the different stages in anatomical remodelling revealed by theme maps during retinal degeneration were characterised.[97, 193] The concept of a glial seal has evolved because of these studies,[97, 193] allowing us to understand why in moderate-to-advanced stages of retinal dystrophy some vision restoration strategies will encounter major obstacles.
Theme maps are important in that they not only provide a statistically rigid classification for the unequivocal identification of cell types but can be used to extract the amino acid distribution patterns of individual classes of neurons. The extraction of univariate, bivariate or trivariate amino acid distributions allows for the individual amino acid distribution within classes of cells to be determined but also establishes differences due to the disease process.[62, 94, 95, 121, 190, 193]
Functional mapping using an organic cation (agmatine—AGB)
The activity of glutamate receptors can be probed by using an organic cation, agmatine (AGB). Agmatine is a divalent polyamine (de-carboxylated arginine) that penetrates acetylcholine-activated ion channels in sympathetic neurons and frog endplate, as well as indicating physiological activity of invertebrate visual and vestibular neurons.[199-201] Agmatine or related molecules (guanidinium compounds) enter cells via ionic channels in a Nernstian fashion[200, 201] with AGB tracking allowing for excitatory mapping of neuronal tissue.[93-95, 197, 202-209] The amount of AGB entering a cell is proportional to the channel permeability.[197, 202, 203, 207] Agmatine concentration is a key driving force with AGB reporting current flow over time.
Agmatine entry into different sub-populations of neurons will be distinctly different due to two factors.[128, 197, 202, 203] First, basal AGB entry is dependent upon the overall ion channel activity in the ‘resting’ state and may be further modified if L-2-amino-4-phosphono butyric acid (L-APB or L-AP4), a specific metabotropic glutamate receptor agonist, is used to modify AGB entry into ON bipolar cells.[93, 128, 197, 202] The second reason relates to the diverse disposition of glutamate receptors within the vertebrate retina. The activation of these receptors using AMPA, kainate and NMDA produces distinctly different AGB labelling patterns that can be combined with amino acid labelling or cell markers to either classify the retina or provide cell identification of activated neurons (Figures 16 and 17[47, 48, 197, 203, 208, 209]).
The core principle of AGB labelling is illustrated in Figure 16. Basal AGB entry is shown in Figure 16A, where AGB labelling is diffusely located throughout the retina indicative of basal AGB entry. Cone photoreceptors are evident as more darkly stained nuclei in the inner segment region. In animal models of retinal dystrophy or secondary to light damage, AGB labelling of photoreceptors appears to precede photoreceptor loss.[210-212] Functional mapping of the retina is reflected by altered AGB labelling when glutamate analogues are added in the incubation medium. The inclusion of NMDA in the incubation medium resulted in labelling of processes within the inner plexiform layer and somata in the amacrine cell and ganglion cell layer (Figure 16B).
The inclusion of AGB labelling assists in creating greater diversity in the results provided by image analysis. Pattern recognition analysis of overlapping amino acid and functional profile using the cation channel probe (AGB) has provided 28 separable ‘theme classes’ within the rat retina. The inclusion of AGB labelling (functional profiling) almost doubled the number of separable theme classes. The 14 retinal ganglion cell classes reported by Marc and Jones were only possible when AMPA-activated AGB signal was added to the metabolic profiling.
The introduction of techniques to combine cell markers with functional tracking has allowed the identification of selective neurochemically identified neurons that display functional glutamate receptors in normal retinal development and in animal models of retinal disease.[47, 48, 93-96, 208, 209, 213-215] Using a glutamate agonist (kainate), it is possible to combine labelling with a cell marker for the cholinergic amacrine cells to confirm that this population of neurons contain kainate receptors (Figures 17A–C). Similarly, the synthesising enzyme for GABA, glutamic acid decarboxylase (GAD), can be co-localised with activated neurons to display localisation patterns (Figures 17D–F). Using this approach, the development of functional glutamate receptors in the retina has been characterised and this neurochemical development has been associated with changes in retinal cell spatial localisation during development.
Insights Into Retinal Disease Revealed by Basic Anatomical, Neurochemical and Functional Studies
Remodelling triggered after metabolic insult
Metabolic insult triggers a range of cellular damaging mechanisms including excessive glutamate release (excitotoxicity).[54, 216, 217] Altered vascularisation of the vertebrate retina not only results in anomalous vessel formation and disposition but dramatic changes in amino acid distribution patterns.[62, 64, 198] In more severe metabolic insult, ischaemia-reperfusion deprives the vascular tissues of oxygen, metabolites and removal of waste products. There is evidence that anatomical changes may begin within hours of ischaemic insult, particularly ischaemia due to elevated intraocular pressure (IOP) that reaches 50 mmHg or more.[218-220] Ganglion cell survival after transient intervals of retinal ischaemia proceeds for different lengths of time and the severity and duration is related to the length of the ischaemic interval. Amacrine cell death is also well established although possible further death as a function of time after reperfusion is not fully understood.[222, 223] Overall, there is little change in overall retinal amino acid levels from two to 14 days after ischaemic insult but a dramatic alteration in ganglion cell numbers with continuous ganglion cell death for months after 90 minutes of ocular ischaemia. The factors causing this prolonged ganglion cell death are unknown but this model may provide a unique opportunity to study cell death post-insult. Such studies would provide useful insights as to cell death and function in common clinical conditions such as retinal detachment and thus provide a better understanding as to the reason(s) for poor recovery of vision secondary to reattachment surgery.
Functional profiling of the ischaemic-reperfused rat retina was also dramatically altered.[94, 95] Altered glutamate release by photoreceptors was observed through alterations in AGB permeation in the ischaemic-reperfused retina. In addition, a sub-population of amacrine cells loses its exquisite sensitivity to activation by the glutamate analogue kainite. Following ischaemia-reperfusion, the remaining rod bipolar cells are dysfunctional due to a lack of glutamate release by photoreceptors as well as altered metabolic profiles.
Remodelling in retinal dystrophy
Anatomical reshaping of inner retinal neurons occurs during retinal degeneration, including loss of bipolar cell dendrites, neurite sprouting and soma displacement.[90-92, 224] Inner retinal neurons also display neurochemical and functional changes.[56, 102, 225-228] At late stages of retinal degeneration (after complete loss of photoreceptors), rod bipolar cells lose their responsiveness to glutamate.
An emerging concept from models of retinal degeneration is alterations in glutamate receptor function. For example, in the retinal degeneration animal model, the rd1 mouse, ON cone bipolar cells show an aberrant expression of functional AMPA/kainate receptors, with the apparent continual preservation of receptors on OFF cone bipolar cells during the early stages of retinal degeneration.[56, 96] ON cone bipolar cells with the normal metabotropic receptors and aberrant functional ionotropic glutamate receptors would be unable to respond to glutamate activation, reflecting a detrimental effect of remodelling.
During later stages of inherited forms of retinal degeneration where there is photoreceptor loss, ON and OFF cone bipolar cells lose both mGluR6 and ionotropic glutamate receptor expression, respectively.[56, 91] mGluR6 function (as assessed via AGB gating) was not present in the rd1 mouse retina.
The response to NMDA of amacrine cells is also gradually altered during degeneration in the rd1 mouse retina. There is a gradual loss of functional NMDA receptor response in immunocytochemically identified amacrine cells that increased with age with under 10 per cent of cells being activated in the adult rd1 retina.
These results also demonstrate that inner retinal glutamatergic drive is altered from an early age in the rd1 mouse retina. The morphology and dendritic arbour of ganglion cells appear to be normal in both the rd1 and rd10 models.[229, 230] Stasheff has shown a high sustained spontaneous hyperactivity in OFF responses from ganglion cells in the rd1 mouse early in the degenerative process (P14) with rhythmic (10 Hz) firing present when photoreceptors have degenerated. Although the intrinsic physiology of ganglion cells in the rd1 mouse is normal, the neurotransmitter receptors on ganglion cells may be abnormal.
The altered physiology of ganglion cells confirms that the input is abnormal and any attempts to target ganglion cells as output cells in intervention studies must take into consideration the altered input. Possible causes for the altered ganglion cell physiology could include changes in retinal circuitry,[56, 229, 231] including alterations in NMDA receptors in the inner retina.
Afferent input to the NMDA receptor as well as complex protein interactions are responsible for receptor clustering at the synapse. De-afferentation could stimulate a down regulation and/or a change in the subunit composition of glutamate receptors.[223, 233]
Functional changes at the neurotransmitter receptor level
Glutamate receptors are altered in rodent models of inherited retinal dystrophy. Two major changes have been reported: first, the down regulation of the metabotropic (mGluR6) receptors on the depolarising bipolar cells; and second, a concurrent expression of aberrant functional ionotropic receptors on depolarising bipolar cells.[56, 90-92, 96, 234] Puthussery and colleagues propose that the down regulation of first the mGluR6 receptor on rod bipolar cells, followed by the cone bipolar cells while largely retaining ionotropic receptor function on OFF cone bipolar cells, signifies a need for the mGluR6 pathway to receive glutamate activation. Altered inner retinal function (presumed altered NMDA receptor), following known cone activity in the rd1 mouse retina, also suggests an activity-dependent alteration in function.
Further evidence of ‘plasticity’ in the retina includes glutamate receptor changes secondary to altered gene expression (CPG15) activated by a range of stimulants, including light. CPG15 gene expression leads to an increased growth rate of retinal axons and promoted synaptic maturation by recruitment of synaptic AMPA receptors. Trafficking of AMPA receptors also occurs within the retina through an activity-dependent plasticity that may be regulated in the course of a normal light/dark cycle. Guenther and colleagues showed light and development influence expression of NMDA receptors. Kamphuis, Dijk and O'Brien found the gene expression of an aberrant glutamate receptor in rod bipolar cells under different physiological light levels. Whether these receptor subunits combined to form functional ion channels is unknown. Rapid aberrant expression of AMPA receptor subunit (GluR2) occurs secondary to light damage.
Advances in the basic sciences over the past 60 years have lead to the discovery and characterisation of the biochemical pathways involved in the visual cycle and phototransduction followed by gene expression and associated mutations that lead to ocular disease. The development of genetic techniques that allow the insertion of the correct genetic code into cells, has led to viable treatment options. For example, gene therapy is now possible with the restoration of sight in patients affected with the RPE65 mutation causing Leber's congenital amaurosis.[72-78] Similarly, exciting developments with nutritional supplements (for example, using the retinoid QLT091001) has the potential to restore sight in mutations associated with lecithin:retinol acyltransferase (LRAT) or RPE65 that causes Leber's congenital amaurosis or retinitis pigmentosa.
Similar advances are required in understanding the poor visual outcome secondary to, for example, retinal detachment surgery. Recent studies providing an explanation for this poor outcome have come about through application of basic scientific techniques, such as amino acid localisation in the vertebrate retina, followed by understanding and localisation of the amino acid receptors, and more recently, development of techniques to functionally map responses from retinal neurons.[47, 56, 96, 97, 193] These and other methods have revealed detrimental remodelling involving specific subclasses of glutamate receptors that occur over a precise time frame. Further advances in fully characterising the time course, the receptors involved and the physiology of the aberrant expression of glutamate receptors will provide the foundations to target therapies for the restoration of vision before detrimental remodelling occurs.
Finally, other parts of the eye are being explored to better understand both normal and abnormal amino acid neurochemistry. Recently, work has characterised amino acid labelling and amino acid transporters in the anterior eye.[154, 239-242] Changes to amino acid and glutamate metabolic enzyme are not restricted to the retina secondary to metabolic insult: altered enzymatic expression in the ciliary processes and dramatic alterations in amino acid labelling patterns were found in the ciliary processes and lens.[154, 242] Despite these dramatic amino acid changes in the ciliary processes and in particular the lens, aqueous amino acid levels remained relatively unchanged. We have proposed that the lens is a reservoir for amino acids and may provide a mechanism of amino acid homeostasis for the eye. As we further explore this possibility, such interesting findings highlight the need to better understand the intricate interrelationships between different ocular structures, thereby expanding our understanding and provide further pathways for therapeutic intervention leading to better management of ocular disease.
This work was supported, in part, by grants from the National Health and Medical Research Council of Australia (1009342 and 1021042) and Optometrists Association Australia.
We are grateful to Dr Robert Marc (University of Utah) for the kind gift of the AGB and amino acid antibodies and for his guidance.
Ethics approval was obtained for the collection of eyes from university laboratories and from other institutions. Samples of different animal species not previously published were obtained either as part of experiments within the Retinal Networks Laboratory at the University of Melbourne, the University of New South Wales or University of Auckland or post-mortem from animals at the Melbourne Zoo, Queenscliff Marine Station, Healesville bird sanctuary or from Northern Territory crocodile farms.
We are grateful for access to the eyes provided to us by:
The Melbourne Zoo, Parkville Victoria 3052, Australia
Queenscliff Marine Station, Queenscliff Victoria 3225, Australia
Healesville Sanctuary, Badger Creek Road, Healesville Victoria 3777, Australia
Northern Territory Crocodile Farms (Crocodylus Park), Karama NT 0812, Australia (Grahame Webb)