Address correspondence and reprint requests to Bang V Bui, Department of Optometry and Vision Sciences, The University of Melbourne, Vic. 3010, Australia. E-mail: firstname.lastname@example.org
Glutamate is a major neurotransmitter in the CNS but is also a key metabolite intimately coupled to amino acid production/degradation. We consider the effect of inhibition of two key glutamate metabolic enzymes: glutamine synthetase (GS) and aspartate aminotransferase on retinal function assessed using the electroretinogram to consider photoreceptoral (a-wave) and post-receptoral (b-wave) amplitudes. Quantitative immunocytochemistry was used to assess amino acid levels within photoreceptors, ganglion and Müller cells secondary to GS inhibition. Intravitreal injections of methionine sulfoximine reduced GS immunoreactivity in the rat retina. Additionally, glutamate and its precursor aspartate was reduced in photoreceptors and ganglion cells, but elevated in Müller cells. This reduction in neuronal glutamate was consistent with a deficit in neurotransmission (−75% b-wave reduction). Exogenous glutamine supply completely restored the b-wave, whereas other amino acid substrates (lactate, pyruvate, α-ketoglutarate, and succinate) only partially restored the b-wave (16–20%). Inhibition of the aminotranferases using aminooxyacetic acid had no effect on retinal function. However, aminooxyacetic acid application after methionine sulfoximine further reduced the b-wave (from −75% to −92%). The above data suggest that de novo glutamate synthesis involving aspartate aminotransferase can partially sustain neurotransmission when glutamate recycling is impaired. We also show that altered glutamate homeostasis results in a greater change in amino acid distribution in ganglion cells compared with photoreceptors.
Inhibition of GS using l-methionine sulfoximine (MSO) leads to a reduction of the electroretinogram (ERG) b-wave but not the upstream photoreceptoral response (Barnett et al. 2000). Similarly, inhibition of glutamate uptake using threo-hydroxyaspartic acid and d-aspartate (Winkler et al. 1999) as well as antisense knockdown of high affinity glutamate transporters (excitatory amino acid transporter-1 and glutamate-aspartate transporters) in rats (Barnett and Pow 2000) selectively attenuates the depolarizing-bipolar cell dominated b-wave (Stockton and Slaughter 1989). Thus, effective glutamate clearance and recycling are both necessary for neurotransmission. Not surprisingly, inhibiting GS resulted in substantial reduction of glutamate, glutamine and GABA within retinal neurons, but little difference in glycine and taurine labeling patterns (Pow and Robinson 1994). The reduction in GABA is expected as glutamate is its immediate precursor (Bender 1985). What is not known is the relative change in amino acid content, particularly amino acids that are known to participate in glutamate production via transamination reactions.
In addition to the neurotransmitter role, glutamate is also a key metabolic intermediate (Bender 1985). Under normal conditions a proportion of the glutamate taken up into glial cells is metabolized and the carbon skeleton enters the tri-carboxylic acid (TCA) cycle (Ereciñska et al. 1988; Yudkoff et al. 1989; Sonnewald et al. 1996) which reflects the intimate coupling between glutamatergic and metabolic pathways. Indeed, anaplerotic reactions involving amino acids such as glutamate, aspartate, and alanine are important to replenish carbon skeletons by producing the TCA cycle intermediates α-ketoglutarate, oxaloacetate, and pyruvate, via reactions catalyzed by the aminotransferase group of enzymes (Ereciñska and Silver 1990). These pathways appear to be important during metabolic stress, where amino acid carbon skeletons are seconded to maintain metabolism (Kalloniatis and Napper 1996; Zeevalk et al. 1998; Acosta and Kalloniatis 2005), at the expense of neurotransmission (Bui et al. 2004). To maintain efficient neurotransmission, de novo glutamate synthesis may be needed to compensate for the loss of glutamate to metabolic pathways (Lieth et al. 2001). However, it is unclear whether glutamate synthesis contributes to neurotransmission under normal conditions and when the glutamate/glutamine cycle becomes impaired.
We consider whether glutamate production via transamination reactions contribute to normal neurotransmission by assessing the effect of aminotransferase inhibition on retinal function. We assess the effect of GS inhibition on the distribution of a range of amino acids involved in the glutamate/glutamine cycle. In particular, we consider whether precursors for glutamate are reduced, which would indicate that TCA intermediates are being used for glutamate production.
Materials and methods
All experimental protocols conformed to the principles regarding the care and use of animals adopted by the National Health and Medical Research Council of Australia. Animals were maintained in a 12-h light/dark (< 50 lux, on at 8 am) environment with normal rat chow and water available ad libitum.
Tissue collection and preparation
Sprague–Dawley rats (postnatal day 26–28, n = 6) were deeply anesthetized by intramuscular injection of ketamine (60 mg/kg) and xylazine (5 mg/kg; Troy laboratories, Smithfield, NSW, Australia). After tissue collection, animals were killed using an overdose of anesthetic injected intracardially. Whole eyes were excised and several cuts were introduced in the cornea and part of the sclera to increase fixative penetration.
After fixation in a mixture of 0.75%p-formaldehyde and 0.01% glutaraldehyde in 0.1 M phosphate buffer for 24 h, whole eyes were cryoprotected and sectioned. Cryosections at 16-μm thickness were then mounted onto Superfrost™ Plus (Erie Scientific Company, Portsmouth, NH, USA) microscope slides and processed for immunocytochemistry using procedures as previously published (Hu et al. 2008). The primary antibody to label retinal Müller cells was a mouse anti-GS (clone 6, 1 : 6000; BD Transduction Laboratories, San Jose, CA, USA), visualized with a goat anti-mouse Alexa 594 secondary antibody (1 : 400; Molecular Probes, Invitrogen, Carlsbad, CA, USA). The tissue was embedded in anti-fading medium (Citifluor; Alltech Associates, Auckland, New Zealand) and coverslipped.
The procedure for post-embedding silver-intensified immunogold detection of a range of antibodies was identical to previously published studies (Acosta et al. 2007; Sun et al. 2007). Antibodies used in this study to target free amino acids (anti-glutamate, 1 : 5000; anti-glutamine, 1 : 1000; anti-aspartate, 1 : 10 000; anti-arginine, 1 : 400; anti-alanine 1 : 1000; and anti-taurine, 1 : 10 000) were polyclonal antibodies raised in rabbit (kind gifts by Dr RE Marc, University of Utah, Salt Lake City, UT, USA). Previous studies have established the specificity of these glutaraldehyde-conjugated antibodies (Marc et al. 1990, 1995; Kalloniatis and Fletcher 1993). Good selectivity and low cross-reactivity were demonstrated by dot-blot immunoassays (Marc et al. 1990, 1995) and competitive inhibition of binding (resources from Signature Immunologics Inc, Salt Lake City, UT, USA). The selectivity of these antibodies for their target antigenic epitope is > 3 log units in most cases (Marc et al. 1995).
Image capture and analysis
For immunofluorescence, antibody labeling was visualized using a Leica SP2 confocal laser scanning microscope (Leica Microsystems Ltd, Heerbrugg, Germany) fitted with an argon/krypton mixed ion laser and appropriate filters. For post-embedding immunocytochemistry silver deposits were viewed on a Leica DMR light microscope (Leica Microsystems Ltd), and photographed under constant light with a fixed camera gain and gamma.
Silver-intensified amino acid immunoreactivity was quantified by measuring the labeling intensity in pixel value (0–255), where each pixel value equals 9.407 × 10−3 log units. A fixed area size of 100 pixels was drawn over a region of interest and ∼10 measurements were taken for each retinal sample to give the average pixel value. The final mean pixel value was calculated from six different animals. Differences in pixel value between the control eye and MSO-treated eye was expressed as the mean (± SEM). Increased immunoreactivity is expressed as fold change (treated/control), whereas decreased immunoreactivity is expressed as a fold decrease (−1/fold change).
Intravitreal injection of pharmacological agents
All pharmacological agents were diluted in normal saline, with the pH adjusted to ∼7.4. All concentrations given in this study represent the final vitreal concentration calculated assuming complete dilution, no leakage and an average vitreous chamber volume of 40 μL for the rat (Dureau et al. 2001; Naarendorp et al. 2001).
l-Methionine sulfoximine (ICN Biomedicals Inc., Aurora, OH, USA) irreversibly inhibits GS by binding to the active site of the enzyme as methionine sulfoximine phosphate (Rowe et al. 1969). MSO is known to completely inhibit GS activity after 10 min in isolated rat retina (Winkler et al. 1999). Alanine aminotransferase (EC 188.8.131.52) (Subbalakshmi and Murthy 1983) and glutamine transferase (EC 184.108.40.206) may also be inhibited by MSO. However, Barnett et al. (2000) have shown that at the concentrations used here the in vivo effect of MSO on these enzymes is minimal. MSO concentrations of 1 mM (n = 5), 5 mM (n = 5) and 20 mM (n = 5) were employed in the functional studies described below. A MSO concentration of 13.5 mM was used to assess the effect of GS inhibition on amino acid distribution. MSO or control saline (0.9% NaCl, pH 7.4) was injected to left and right eyes (n = 6), respectively. The animal was kept for 24 h before tissue collection.
We attempt to bypass GS inhibition by supplying exogenous carbon substrates including glutamine (2 μL, 10 mM), lactate, pyruvate, α-ketoglutarate, and succinate (2 μL, 10 mM; Sigma) at 90 min following MSO application (20 mM). These concentrations were chosen to be in excess of normal retinal levels of these compounds (Adler and Southwick 1992; Winkler et al. 1997). If carbon substrates can support neurotransmission then all substrates should display similar improvement in function to that previously reported for glutamine (Barnett et al. 2000).
Aspartate aminotransferase assay
Aspartate aminotransferase (AAT; EC 220.127.116.11) activity was assayed in Sprague–Dawley rats (9- to 10-week-old, n = 6). Individual retinas were homogenized in a 0.9% NaCl solution and centrifuged at 5000 g for 7 min (5415R; Eppendorf, Hamburg, Germany). A known amount of protein in the supernatant was added to the AAT reaction mixture (Trace, Noble Park, Vic., Australia) containing 2-oxoglutarate (13.2 mM), l-aspartate (220 mM), malate dehydrogenase (600 U/L; EC 18.104.22.168), lactate dehydrogenase (1000 U/L; EC 22.214.171.124), NADH (0.18 mM), Tris buffer (88 mM), and EDTA (5.5 mM). AAT activity was measured spectrophotometrically at 37°C by monitoring reduction in NADH absorbance at 340 nm, within 30–60 min of dissection in an enzymatically coupled reaction. The results were calculated a International Units per mg of protein (IU/mg protein) and plotted as relative activity of individual drug-treated tissue over saline injected tissue, where 1 indicates no change. AAT activity was assessed in the presence of 13.5 or 5 mM final concentration of MSO or 13.5, 5, 0.5, 0.1, or 0.05 mM final concentration of AOAA. Protein concentration was detected in a colorimetric reaction using a bicinchoninic acid protein assay reagent (Pierce, Rockford, IL, USA). AAT activity was also measure in tissue harvested 3 h, 24 h, and 7 days following intravitreal injection of 5 mM final vitreal concentration of AOAA.
Procedures for ERG recording are identical to those previously described (Bui et al. 2003). In brief, ERGs were recorded from anesthetized (60 mg/kg ketamine and 5 mg/kg xylazine; Troy laboratories), dark adapted (overnight) Long–Evans rats (9 weeks, n = 5 or 6 per group). Following mydriasis and corneal anesthesia recordings were made using silver–silver chloride (Ag–AgCl) electrodes. ERGs were elicited using a calibrated photographic flash (285 V; Vivitar Photographics, Newbury Park, CA, USA) delivered via a Ganzfeld sphere. Responses (−3 dB at 0.1–3000 Hz) to a single intensity (1.5 log cd s/m2) were collected prior to and at 5-min intervals following intravitreal injection of pharmacological agents for 180 min. Exogenous substrates were injected 90 min following MSO application.
As previously described, the photoreceptoral a-wave was fit with the phototransduction model of Hood and Birch (1990). This model describes the rod photocurrent (P3) as a function of stimulus energy (i, cd s/m2) and time (t, s) by its saturated amplitude (RmP3, μV), sensitivity (S, m2/cd/s3), and a fixed delay of 3.3 ms (td, s). This model was fit to the raw data up to the first minimum of each a-wave by minimizing the sum-of-square error term using an Excel™ spreadsheet (Microsoft Corporation, Redmond, WA, USA).
Post-receptoral b-wave was isolated by subtracting the P3 model from the raw data. Peak ampltitude and time were then measured. Post-treatment ERG parameters were normalized to their respective baselines, with group data shown as a mean (± SEM) as a function of time following drug application.
Functional data for MSO-treated eyes were compared with MSO + substrates treated eyes (from 90 to 180 min) using two-way (treatment and time) repeated measures (RM) anova (prism, v4.0; GraphPad Software, Inc., San Diego, CA, USA). The outcomes of biochemical assays and amino acid quantification were compared using a t-test. For all experiments, an alpha value of 0.01 was employed to protect against type-1 errors.
Effect of MSO on glutamine synthetase immunoreactivity
To determine whether MSO has a uniform or regional effect on GS, we analyzed the pattern of GS labeling in control and MSO-treated retinae. GS immunoreactivity is exclusively found throughout the cytosol of Müller cells in the rat retina (Riepe and Norenburg 1977; Riepe and Norenberg 1978; Derouiche and Rauen 1995). Figure 1a shows that the somata of Müller cells were apparent in the middle of the inner nuclear layer and because of the labeling in the Müller cells processes, the outer and inner limiting membrane can be visualized in the distal outer nuclear layer and ganglion cell layer, respectively.
Twelve hours after MSO injection, GS expression was reduced throughout the retina (Fig. 1b). The distinctive somal labeling disappeared, though patchy GS expression was still evident in the processes of Müller cells. Because of the reduction in GS immunoreactivity, the outer and inner limiting membrane regions were poorly defined. A reduction in GS immunoreactivity has also been demonstrated in the rat brain (Tanigami et al. 2005). This is because of MSO interfering with antibody binding to GS as total GS-protein levels are unaffected by MSO (Bidmon et al. 2008).
Effect of MSO on amino acid distribution
To facilitate comparison of amino acid distribution, control and MSO-treated retinae were simultaneously processed and immunolabeled. Our results show the same amino acid distribution in the rat retina as previously described (Fletcher and Kalloniatis 1996; Sun et al. 2007). Briefly, glutamate was abundant in the photoreceptor, bipolar, and ganglion cells (Fig. 2a). Horizontal and amacrine cells also contained high levels of glutamate. Müller cells located in the middle of the inner nuclear layer, show minimal glutamate labeling (Fig. 2a, arrowheads). Following GS inhibition, the somata and processes of Müller cells were intensely labeled (Fig. 2b). The proximal Müller cell endfeet displayed increased glutamate immunoreactivity, which was 57 times greater than control (57.0 ± 3.4, p <0.0001, t-test; Fig. 4). In contrast, most neurons showed reduced immunoreactivity. The immunoreactivity in the ganglion cells was reduced by 16-fold (16.1 ± 1.1, p <0.0001, t-test; Fig. 4). Glutamate immunoreactivity was undetectable in photoreceptor and bipolar cells.
Glutamine, a precursor for glutamate was seen in virtually all neurons as well as the Müller cells (Fig. 2c). Horizontal cells and Müller cell endfeet displayed the highest glutamine immunoreactivity, while bipolar and amacrine cells were less strongly labeled. Application of MSO resulted in significantly reduced glutamine in Müller cells (5.9 ± 0.4-fold, p <0.0001, t-test; Fig. 2d). Reduced labeling was also evident throughout the retina.
Aspartate and alanine are also precursors of glutamate via transamination reactions (Ereciñska and Silver 1990; Kalloniatis and Tomisich 1999). Diffuse aspartate labeling was evident throughout the retina (Fig. 2e). Ganglion cells contained the highest aspartate level, whereas the photoreceptor inner segments and subpopulation of bipolar and amacrine cells were moderately labeled. The change in aspartate immunoreactivity following GS inhibition in Müller cells mirrored that of glutamate (Fig. 2f), increasing by 10-fold (9.9 ± 0.5, p <0.0001, t-test; Fig. 4), whereas most neurons showed reduced immunoreactivity. The greatest decrease in aspartate immunoreactivity was found in ganglion cells, with 95 times less compared with control retinae (p <0.0001, t-test; Fig. 4).
Alanine, was localized in horizontal, amacrine, and ganglion cells (Fig. 3a). Strong immunoreactivity was detected in the somata of horizontal cells, while amacrine cells displayed variation in alanine labeling. Alanine immunoreactivity was reduced in horizontal cells (Fig. 3b). The inner plexiform layer also showed lower alanine immunoreactivity, but the rest of the retina appeared similar to control.
Arginine was predominantly localized in the Müller cell endfeet (Fig. 3c, arrowheads). Following GS inhibition, higher arginine levels were found in Müller cells (Fig. 3d), with a 38-fold increase (38.1 ± 0.4, p <0.0001, t-test; Fig. 4).
Taurine was found in virtually all retinal neurons (Fig. 3e), with the highest level of immunoreactivity in the photoreceptors, bipolar cells, and Müller cells (Kuriyama et al. 1990; Lake 1994; Fletcher and Kalloniatis 1996). Several populations of amacrine cells were immunonegative for taurine. Taurine immunoreactivity was reduced throughout the retina following MSO treatment (Figs 3f and 4).
Exogenous substrate application and retinal function
Given the significant changes in amino acid distribution following MSO application we next consider the effect on neurotransmission and the potential benefit of providing exogenous amino acid carbon substrates. Although aspartate would be ideal for this purpose, this amino acid is neuroactive and is known to impair b-wave activity (Kleinschmidt and Dowling 1975).
Figure 5a shows the conventional ERG response, which is characterized by a photoreceptoral negative component (a-wave), described here using the phototransduction model of Hood and Birch (1990) as shown by the solid line. The a-wave is followed by a large corneal positive b-wave, which reflects responses from depolarizing bipolar cells (Fig. 5a). Consistent with a previous report, in vivo MSO application selectively attenuates the b-wave over the course of 90 min (Barnett et al. 2000).
Figure 5b and c shows that MSO treatment (filled circles) did not affect phototransduction amplitude (treated/control) at the drug concentrations used here (Fig. 5b, RM-anova, RmP3: 1 mM, p = 0.35; 5 mM, p = 0.36; 20 mM, p = 0.29). Figure 5d confirms a large b-wave reduction following MSO application (p < 0.001 for all), which was similar for the various MSO concentrations (Fig. 5e, 1 mM, −75 ± 4%; 5 mM, −77 ± 9%; 20 mM, −77 ± 5%).
Figures 6a and b show that the b-wave loss induced by MSO application (20 mM) could be completely reversed with the introduction of exogenous glutamine (10 mM). Exogenous glutamine injection at 90 min following MSO application (filled circles) gradually ameliorated the deficit in post-receptoral b-wave function (Fig. 6d, p < 0.001), without changing the photoreceptoral response (RmP3, Fig. 6c, p = 0.50). The b-wave had completely recovered by 40 min after glutamine injection (130 min total).
Figure 6d shows that exogenous lactate supply significantly improved b-wave amplitudes by 19 ± 7% (filled circles, p < 0.01). Similarly, pyruvate (16 ± 3%, p < 0.01), α-ketoglutarate (20 ± 4%, p < 0.01) and succinate (16 ± 6%, p < 0.01) were all able to provide significant but incomplete b-wave amplitude recovery (Fig. 6d). Although the improvement in photoreceptoral amplitude after exogenous substrate administration was not significant, the magnitude of change could explain the enhancement in the post-receptoral b-wave. To consider this possibility the b-wave change because of exogenous substrates was correlated with change in a-wave amplitude (Fig. 6c vs. d combined lactate, succinate, pyruvate, and α-ketoglurtarate), for each animal across all exogenous substrates. There was no correlation between the a-wave and photoreceptoral change following exogenous substrate application (r = 0.18, p = 0.40), which suggests that the b-wave improvement occurs independently of any photoreceptoral effect.
Activity of aminotransferases
Given the observed changes in amino acid distribution and retinal function following MSO treatment we wanted to rule out the possibility that this arose because of a secondary effect of MSO on the aminotransferases. Our average AAT activity (Acosta and Kalloniatis 2005) was similar to that previously reported in rat retina (0.78 IU/mg of protein reported (Endo et al. 1999) and brain (1.17 IU/mg of protein (Benuck et al. 1972). These values are lower than that reported by Ross and Godfrey (1985). However, that study assessed activity in various retinal layers and not the entire tissue. Figure 7 shows that incubation of retinal tissue with MSO (unfilled bars, 5 and 13.5 mM) did not result in significant changes in AAT activity that could account for the retinal dysfunction. However, tissue incubated in AOAA showed reduced AAT activity in a concentration dependent manner (Fig. 7 filled bars, p < 0.01 for 0.05 to 13.5 mM, t-test). An intraocular injection of 5 mM AOAA was necessary to achieve inhibition of the AAT enzyme (Fig. 7 gray bars, p < 0.01, t-test) which was effective at 3 h (−63 ± 8%). A smaller effect was noted 24 h (−34 ± 6%) after treatment and AAT activity appeared to have returned to normal after 7 days (Fig. 7, −9 ± 4%).
To consider whether the aminotransferase reactions are involved in the normal maintenance of retinal neurotransmission, we assessed the ERG following application of the aminotransferase inhibitor AOAA. Figure 8 shows that introduction of 5 mM AOAA had no significant effect on the photoreceptoral (RM-anova, p = 0.09) or b-wave amplitude (RM-anova, p = 0.45). However, when 5 mM AOAA was injected at 90 min after MSO (20 mM) the b-wave was further reduced from −72 ± 6% to −95 ± 2% (Fig. 8c, p < 0.05, t-test).
A myriad of pathways couple glutamate neurotransmission to metabolism (Bender 1985). We provide further evidence that the interaction between these pathways are dynamic, thereby allowing the retina to react to alterations in metabolic and signaling needs. Specifically, we show that when glutamate recycling is impaired secondary to GS inhibition, glutamate is routed into metabolic pathways in Müller cells via transaminase reactions. On the other hand glutamate required for neurotransmission is being produced from metabolic TCA cycle intermediates via the same reactions in photoreceptors. We show that important reactions in this process are the aminotranferases. However, the retina is able to maintain function if transaminase reactions are impaired secondary to AOAA application. Therefore, the retina displays considerable resilience to maintain suitable levels of glutamate for neurotransmission and metabolic function.
Glutamine synthetase inhibition and neurotransmission
Pow and Robinson (1994) showed that neurons predominantly derive glutamate from glutamine. Our findings and those of Barnett et al. (2000) show that in vivo GS inhibition leads to an attenuation of the post-receptoral ERG b-wave. The rate of b-wave loss observed in our study (∼90 min, Fig. 5e) closely follows the time course of MSO induced glutamate depletion in the retina (Pow and Robinson 1994). In contrast, Winkler et al. (1999) found that MSO had no effect on the b-wave in isolated rat retina. This outcome may have been expected as glutamine and glutamate available in the perfusion media effectively negates any MSO effect. Winkler et al. (1999) also showed, that glutamate transport inhibition with threo-hydroxyaspartic acid results in b-wave loss even in the presence of glutamine, which suggests that glutamate reuptake into photoreceptors is more important than glutamate recycling for neurotransmission in isolated retina. However, their data are also suggestive of a greater b-wave amplitude loss and a faster decline when glutamine was excluded from the media, which is consistent with a contribution from glutamate recycling. These discrepancies may arise from differences between in vivo and isolated retina in terms of extracellular glutamate and glutamine concentrations (Winkler et al. 1999).
Our finding and that of Barnett et al. (2000) show that the b-wave loss could be completely ameliorated with the introduction of exogenous glutamine, which provides evidence that the b-wave deficiency induced by in vivo MSO application largely reflects impaired neurotransmission. This is consistent with the finding in isolated rat retinae that MSO induced b-wave loss could be ameliorated with the addition of glutamine or glutamate (Winkler et al. 1999).
Pharmacological inhibition (Stockton and Slaughter 1989) or targeted knockout (Masu et al. 1995) of metabotropic glutamate receptor 6, the metabotropic receptor found on depolarizing bipolar cells, leads to complete b-wave loss. That the b-wave was not completely abolished by MSO (Figs 5 and 8) suggests that glutamate recycling via GS is not entirely responsible for supplying all the glutamate needed for neurotransmission. In addition to glutamate recycling via GS glutamate reuptake and de novo glutamate synthesis can also contribute to replenishing neurotransmitter pools.
The change in amino acid immunoreactivity provides a clue as to the potential pathways that contribute to glutamate production during in vivo GS inhibition. Aspartate and alanine are precursors for glutamate production via AAT and alanine aminotransferase, respectively (Ereciñska and Silver 1990). Both aspartate and alanine are increased in Müller cells following GS inhibition (Fig. 4), which suggest that glutamate is shunted into metabolic pathways in glia. On the other hand, aspartate is reduced in retinal neurons, which suggests a shift in the equilibrium of AAT toward glutamate synthesis, with the concomitant production of the four carbon keto-acid oxaloacetate (Fig. 9). As AAT activity, was unaffected by MSO (Fig. 7) the overall reduction in neuronal glutamate with MSO application is likely to reflect the absence of the glutamate recycling. Alanine, despite being a precursor for glutamate shows a paradoxical elevation in neurons (Fig. 4). This is in line with the notion that alanine is predominantly involved in energy production (Tsacopoulos et al. 1994).
Arginine, an amino acid important in the urea cycle and a precursor of nitric oxide, showed a small reduction in neurons (Fig. 4). In contrast, arginine levels were dramatically increased in Müller cells following GS inhibition. Given that the brain and retina do not have a urea cycle and remove excess nitrogen via GS amidation of glutamate (Koshiyama et al. 2000; Wiesinger 2001), the likely explanation is a ‘flow back’ effect secondary to altered glutamate catabolism. Arginine is metabolized via two pathways in the retina: one using nitric oxide synthatase (EC 126.96.36.199) to produce nitric oxide and the second via arginase (EC 188.8.131.52) to produce ornithine (Koshiyama et al. 2000). The putative change in the equilibrium of glutamate transaminase reactions that produced the dramatic increase in aspartate in Müller cells, is also likely to affect ornithine aminotransferase (EC 184.108.40.206) that colocalizes with arginase in the retina (Koshiyama et al. 2000). Consequently, arginine metabolism would be disrupted leading to a marked increase in labeling within Müller cells. Although arginine metabolic pathways are found in other parts of the retina, particularly, photoreceptors (Koshiyama et al. 2000), the arginine elevation within Müller cells following MSO application, demonstrates another example of neural–glial metabolic compartmentalization.
Given the above patterns of amino acid changes following GS inhibition, it appears that glutamate transaminase reactions (e.g. AAT) play a role in glutamate production from amino acid carbon substrates. Using TCA intermediates to support neurotransmission would impact on energy production, thus it is not surprising that an equivalent molar concentration of amino acid carbon substrates could only partially restore the b-wave (Fig. 6, ∼16% improvement), when compared with glutamine which returned the b-wave back to normal. Of the exogenous substrates that were injected into the vitreous only glutamine can directly contribute to glutamate production without first entering metabolic pathways. Further evidence for this contention comes from our finding that the addition of AOAA following MSO resulted in an added reduction of the b-wave. Indeed, at its maximal effect (35 min post-injection) the b-wave (40 ± 8 μV) was only slightly larger than noise (∼20 μV). This outcome suggests that under the current conditions endogenous glutamate production plays a larger role than glutamate reuptake in neurotransmission.
Amino acid changes in photoreceptors and ganglion cells
Rapid uptake of extracellular glutamate limits the activation of neurons (Matsui et al. 1999; Chen and Diamond 2002; Fyk-Kolodziej et al. 2004) and prevents excitotoxicity. Retinal ganglion cells are particularly susceptible to this type of insult (Luo et al. 2001). Given that both photoreceptors and ganglion cells use glutamate and localize AAT (Acosta and Kalloniatis 2005) it is not surprising that we observed similar changes in amino acid distribution following GS inhibition in the two cell classes (Fig. 4). Theses similarities suggest that photoreceptors and ganglion cells use similar glutamate metabolic pathways. However, ganglion cells showed a more dramatic reduction in aspartate (Fig. 4), an outcome that suggests a much stronger shift in the equilibrium of AAT toward increased glutamate production compared with photoreceptors. Why might these pathways be producing glutamate in ganglion cells when GS inhibition leads to excess glutamate? One explanation is that glutamate taken up into ganglion cells can fuel the TCA cycle as α-ketoglutarate via glutamate dehydrogenase (Fig. 9). The excess five carbon α-ketoglutarate can then be transaminated back to glutamate via AAT and also produce the four carbon oxaloacetate, which would enter the TCA cycle. Together these pathways might act to maintain TCA cycle activity in ganglion cells with a net reduction in glutamate.
We confirm that glutamate recycling is important for photoreceptor-to-bipolar cell neurotransmission in vivo. We provide evidence that de novo glutamate synthesis involving AAT can partially sustain neurotransmission under stress, this contribution can be slightly increased given the availability of amino acid carbon substrates. In response to stress ganglion cells show a greater change in amino acid distribution compared with photoreceptors, which might suggest that ganglion cells are less able to buffer against changes in glutamate homeostasis.
This project has been supported, in part, by the Health Research Council of New Zealand 05/247 (MK/MLA), the New Zealand Optometric Vision Research Foundation (MK/MLA), the Auckland Medical Research Foundation (MK/MLA), the Robert G. Leitl estate (MK), and Retina Australia (to MK/AJV). BVB is supported by a NHMRC Grant (400127).