|Home | About | Journals | Submit | Contact Us | Français|
We examined the role of GlyT1, the high-affinity glycine transporter, in the mouse retina with an emphasis on the role of glycine as a coagonist of N-methyl-D-aspartic acid (NMDA) receptors. We pursued this objective by studying heterozygote mice deficient in the GlyT1 transporter (GlyT1−/+) and compared those results with wild-type (WT) littermate controls (GlyT1+/+). Capillary electrophoresis was used to separate and quantitatively measure glycine release from isolated retina preparations; pharmacologically blocking GlyT1 with N-[3-([1,1-biphenyl]-4-yloxy)-3-(4-fluorophenyl)propyl]-N-methylglycine in the WT retina generated a significantly larger accumulation of glycine into the bathing environment when compared with the GlyT1−/+ retinas. The relative occupancy state of the NMDA receptor coagonist sites was tested using whole-cell recordings from ganglion cells while bath applying D-serine or D-serine + NMDA. The interpretation of these studies was simplified by blocking post-synaptic inhibition with picrotoxinin and strychnine. NMDA receptor coagonist sites were more saturated and less enhanced by D-serine in the GlyT1−/+ mice compared with the WT controls. Immunoblots of NMDA receptor subunits (NR1, NR2A and NR2B) in WT and GlyT1−+ animals showed that the NR1 subunits were identical. These observations are discussed in view of contemporary issues about NMDA receptor coagonist function in the vertebrate retina and the role of glycine vs. D-serine as the endogenous coagonist.
N-methyl-D-aspartic acid (NMDA) receptors (NMDARs) are distributed throughout the central nervous system and have diverse functions in mediating and modulating excitatory actions at glutamatergic synapses. These receptors are also found in the retina, where their predominant distribution is among ganglion and amacrine cells (Slaughter & Miller, 1983; Dixon & Copenhagen, 1992; Diamond & Copenhagen, 1993; Fletcher et al., 2000). NMDARs in the retina are of interest not only because of their physiological role in mediating light-evoked synaptic currents but also because of their excitotoxic properties, which may provide a risk factor for diseases of the retina such as glaucoma, ischemia or diabetic retinopathy (Lipton, 2001). Excitotoxic actions of glutamate on retinal ganglion cells have been demonstrated, with neuroprotection observed when NMDARs were blocked (Vorwerk et al., 1996).
Among the many unique features of NMDARs is their requirement for a second agonist other than glutamate, which must be bound to its recognition site on the NR1 subunit for channel gating by glutamate. When the need for this coagonist was first described (Johnson & Ascher, 1987), it was assigned to glycine. However, shortly after the discovery of this coagonist requirement, Kleckner & Dingledine (1988) showed that many amino acids could substitute for glycine, including enantiomers such as D-serine. At the time of this observation, the presence of D-amino acids was considered to be an aberration in vertebrate tissues, perhaps reflecting a dietary source or bacterial contamination. However, Hashimoto et al. (1992) succeeded in showing that D-serine was present at significant levels in the brain and that its distribution was similar to that of NMDARs (Hashimoto et al., 1993). Subsequently, Wolosker et al. (1999) succeeded in isolating and characterizing serine racemase, the enzyme that synthesizes D-serine from L-serine. Initially this enzyme was localized to astrocytes but more recent evidence suggests that it may also be present in neurons (Kartvelishvily et al., 2006).
D-serine and serine racemase have been identified in the vertebrate retina (Stevens et al., 2003; Williams et al., 2006) and initially localized to Müller cells and astrocytes. Previous studies have suggested that D-serine plays a prominent role as an NMDAR coagonist (Stevens et al., 2003; Gustafson et al., 2007). However, glycine is also present in the retina, but is tightly regulated through GlyT1, which has been localized to a group of amacrine cells in the mammalian and avian retinas (Pow & Hendrickson, 1999) and Müller cells (Lee et al., 2005) in the amphibian retina. What role then does GlyT1 have on the availability of glycine as a coagonist for NMDARs in the mammalian retina? In the current study, we approached this question by using a knockout mouse that has a deficiency in GlyT1 (GlyT1−/+) (Tsai et al., 2004). The homozygotes (GlyT1−/−) die at birth precluding their use for these studies. For this reason, we used the heterozygotes and their GlyT1+/+ littermates to evaluate the importance of high-affinity glycine transport and the role that this system plays in regulating the coagonist sites of NMDARs of retinal ganglion cells. A preliminary report of these findings was published previously (Reed et al., 2007).
All experiments were performed in accordance with animal protocols approved by the Institutional Animal Care and Use Committee at the University of Minnesota. Anaesthesia and methods were the same for all protocols. Mice were killed by an overdose of Nembutal (150 mg/kg i.p.; Ovation Pharmaceuticals, Deerfield, IL, USA), followed by pneumothorax, after which the eyes were quickly removed and the retina was surgically excised and mounted on filter paper. We obtained heterozygote GlyT1−/+ mice (GlyT1 knockout 129S6/SvEvTac) from the laboratory of J.T.C. and bred them in-house at the University of Minnesota.
Genotypes of all of the animals used in this study were confirmed by polymerase chain reaction as previously described (Tsai et al., 2004). Briefly, tail snips were digested overnight in Proteinase K and the DNA was purified using a Dneasy Kit protocol (Qiagen, Valencia, CA, USA). DNA fragments were amplified by polymerase chain reaction with three primers. Primer 1 (5′-GCCTTGGGAAAAGCGCCTCC-3′) corresponds to the PGK-neo cassette inserted into the mutant locus, primer 2 (5′-CCCCTACTTCATCATGCTGATC-3′) corresponds to the wild-type (WT) allele and primer 3 (5′-TCATACCATAGCCCACGCCT-3′) corresponds to both the mutant and WT alleles. The polymerase chain reaction primers were supplied by the University of Minnesota Biomedical Genomics Center. Hybridization between primers 2 and 3 (WT) yields a single 1.3 kb amplicon, whereas primers 1 and 3 (GlyT1−/+) yield an additional 1.0 kb amplicon on DNA gel electrophoresis.
Whole-cell current and voltage-clamp recordings were obtained using patch recording techniques in an isolated, superfused retina preparation. Bath-applied pharmacological agents were used to study the relative level of coagonist binding in identified retinal ganglion cells. We examined the effects of exogenous D-serine on the amplitude of currents evoked by light stimulation as well as the exogenous application of NMDA. Impulse generation was blocked with tetrodotoxin citrate (TTX) and, unless indicated otherwise, we also blocked GABA and glycine responses with picrotoxinin and strychnine, respectively. For each electrophysiology experiment, eyes were removed, hemisected and the retinas were gently teased free of the eyecup and pigment epithelium. The isolated retina was then placed ganglion cell side up on a nitrocellulose membrane (8 μm, 13 mm, SCWP; Millipore, Billerica, MA, USA) whose center was punched out leaving a 2 mm concentric hole. The retina was mounted in a modified chamber (Newman & Bartosch, 1999) and secured with a platinum wire that was part of a nylon harp structure to secure the retina and filter paper within the chamber. The second retina was removed and placed in Ames medium for later use. The isolated retina preparation was continuously superfused with Ames medium (Ames & Nesbett, 1981) (Sigma, St Louis, MO, USA) but without serum. Perfusate solutions were maintained in gas-bubbled reservoirs above the recording cage and gravity fed over the retina with a flow rate of approximately 3 mL/min. The solution was continuously bubbled with a 95/5% O2/CO2 mixture. Prior to dissection, the pH had been adjusted to 7.4 with NaOH. Osmolality was approximately 284 mOsmol/L for both the extracellular and intracellular electrode filling solutions. All experiments were conducted at room temperature (22°C). Generally, cells with large somas in the most proximal portion of the ganglion cell layer were targeted for recording.
Patch recording electrodes were pulled from borosilicate glass (Freidrich & Dimmock; 1.2 mm OD; 0.8 mm ID) and polished in a Narishige microforge. The filling solution for the electrodes was as follows (in mM): 128 KCH3SO4, 5 NaCH3SO4, 2 MgCl2, 5 EGTA, 5 HEPES, 1 glutathione, 2 ATP-Mg2+ and 0.2 GTP Na3, pH was adjusted to 7.4. The filling solution was filtered with a 0.2 μm syringe filter at the time of electrode filling. Filled electrodes had tip resistances in the range of 7–10 MΩ. Recordings were made with an Axon Multiclamp 700A computer-controlled amplifier connected to an Axon Instruments Digidata 1320 16 bit A-D converter (Molecular Devices, Sunnyvale, CA, USA) and a Windows PC. To reduce noise, offline filtering was performed at 0.5–2 kHz with an eight-pole Bessel software filter. Data were analyzed offline with pCLAMP/Clampfit (Molecular Devices) and Origin Lab (Northampton, MA, USA) commercial software.
Except for the following and where indicated elsewhere in the text, all of the chemicals that we used were obtained from Sigma Chemical (St Louis, MO, USA). We obtained the following chemicals from Tocris (Ellisville, MO, USA): TTX, DL-2-amino-7-phosphonoheptanoic acid (D,L-AP7), NMDA, 2,3-dioxo-6-nitro-1,2,3,4-tetrahydrobenzo[f]quinoxaline-7-sulfonamide and N-[3-([1,1-biphenyl]-4-yloxy)-3-(4-fluorophenyl)propyl]-N-methylglycine (NFPS). Glycine oxidase was kindly provided by Herman Wolosker (Technion-Israel Institute, Haifa, Israel).
Light stimuli were generated with customized software (VisionEgg; Straw, 2008) running on a Windows PC. A beam of light from a 100 W halogen lamp was collimated, projected through a computer-controlled LCD panel (Burkhardt et al., 1998) and then deflected into the Faraday cage with mirrors that directed the light into an optical port of an Olympus BX51WI microscope: a 40× water immersion lens was used to focus the light stimulus onto the retina. In this way, the light stimulus and retinal topology could be viewed simultaneously. A CCD camera (134B, Watec, Orangeburg, NY, USA) was mounted on an additional optical port for capturing images of dye-filled cells. For this purpose, the output of the camera was connected to a ConvertX (PX-M402U, Plextor, Fremont, CA, USA) image converter box, which fed into a computer through a USB terminal. Screen captures of dye-filled cells were taken from the screen image; for each cell, several images (four to nine) were obtained at different focal depths, manually adjusted to capture discrete image planes of the cell. These images were then convolved into a single image using Imaris 3D software (Imaris Bitplane, St Paul, MN, USA). In this way, the axon, cell body and portions of the proximal dendrites were included in the final image (Fig. 1).
We used capillary electrophoresis to separate and measure the glycine accumulation in the bathing medium of an isolated GlyT1−/+ or WT littermate mouse retina. Capillary electrophoresis separations were performed on a commercial capillary electrophoresis instrument (Bechman-Coulter MDQ, Fullerton, CA, USA) with laser-induced fluorescence detection. Retinas were isolated and incubated in separate 100 μL wells of a teflon chamber filled with bicarbonate Ringer and oxygenated with a continuous flow of 95% O2 and 5% CO2 gas, while gently shaking the entire chamber (Rotomix, Thermolyne, Dubuque, IA, USA). Retinas were incubated for 50 min.
Both retinas of a single animal were removed and transferred to separate wells of the multiwell chamber; one retina served as the control and the other was incubated with 10 μm of the GlyT1 inhibitor, NFPS. Following incubation, the medium from each retina was extracted and partitioned into three equal volumes of 25 μL for capillary electrophoresis analysis. One of the three fractions received 2 μL of 1 mM glycine to aid in determining and identifying the glycine peak; the second fraction was incubated for 30 min at 37°C with 2.5 μL of 8 mg/mL glycine oxidase to degrade any glycine in the sample and confirm the glycine peak identity; and the third fraction received no additional additives and was used for the quantitative measurement of the glycine level in the bathing environment (see Fig. 2). For all samples, we added 5 μM of α-amino adipic acid to serve as an internal standard for normalization of the data. Amino acids were fluorescently derivatized at 60°C for 15 min with 4-fluoro-7-nitrobenz-2oxa-1,3-diazole (Molecular Probes, Eugene, OR, USA).
Separations were performed at 15 kV (70 μA). A 4 mW argon laser (488 nm) was directed at the migrating derivatized amino acids, which induced a fluorescent signal at 520 nm that was detected by a photomultiplier tube and digitally plotted as the magnitude of the fluorescence signal vs. time (electropherogram). Quantification o6f the data was achieved by integrating the glycine peak, divided by the integrated peak of the amino adipic acid internal standard. The conversion to concentration was performed by comparing the results with a series of glycine standards. The software used for display and peak integration was 32Karat provided by Beckman-Coulter.
Retinas from GlyT−/+ mice and age-matched controls were homogenized by 10 s sonication in fresh ice-cold radio-immuno precipitation assay buffer (50 mM Tris buffer, 150 mM NaCl, 1% Triton X-100, 1% sodium deoxycholate, 0.1% sodium dodecyl sulfate, pH 7.4) containing Complete protease inhibitor cocktail (Roche, Indianapolis, IN, USA). After holding on ice for 5 min, extracts were centrifuged at 13 000 g. in a microcentrifuge for 12 min to pellet insoluble material. Supernatants were adjusted to 1 mg/mL protein with radio-immuno precipitation assay, aliquoted and frozen at −80°C until use. Proteins (20 μg/well) were separated with sodium dodecyl sulfate-polyacrylamide gel electrophoresis (Novex, Invitrogen Corp., Carlsbad, CA, USA) and transferred to polyvinylidene fluoride for immunoblotting. Membranes were cut into strips, each containing one lane of WT and one lane of GlyT1−/+ proteins, and incubated for 1 h in blocking buffer [50 mM Tris, 150 mM NaCl, 0.1% Tween-20, 3% bovine serum albumin (Sigma)]. Strips were incubated overnight at 4°C in one of the following primary antibodies: anti-NMDAR1 (1 : 2000, Cell Signaling #4204; Millipore, Billerica, MA, USA), anti-NMDAR2A (1 : 2000, Millipore 9354) and anti-NMDAR2B (1 : 4000, Millipore AB1557P). Blots were washed in Tris buffered saline-TWEEN and incubated together in peroxidase-coupled donkey anti-rabbit secondary antibody (Jackson Immunoresearch, West Grove PA, USA) at 1 : 4000 for 1 h, washed and visualized on film with a Lumi-Light ECL kit (Roche). To visualize loading controls, blots were stripped for 2 h (Restore Plus, Pierce, Rockford, IL, USA), washed and incubated with anti-glyceralaldehyde-3-phosphate-dehydrogenase (GAPDH) (Chemicon MAB377, Millipore) at 1 : 50 000 for 20 min. GAPDH bands (38 kDa) were visualized with peroxidase-conjugated donkey anti-mouse secondary antibody (Jackson) and Lumi-Light enhanced chemiluminescence (Roche).
This study was based on whole-cell recordings from 66 ganglion cells in WT (30) and GlyT1−/+ (36) mice. During the course of these experiments, we dye-filled a number of cells in the ganglion cell layer and verified that each cell had an identifiable axon, which emanated from the cell body towards the optic nerve layer of the retina (Fig. 1). Twenty-seven cells in this group were studied with focal vs. diffuse light stimulation to evaluate receptive field properties; in this population we identified transient On (1), transient Off (3), sustained On (13), sustained Off (3) and On-Off (7) subtypes. Thus, a range of physiological polarities and response types were included in our ganglion cell population. However, for the purposes of this study, all subtypes behaved similarly in terms of NMDAR coagonist binding state when we compared the WT vs. the GlyT1−/+ mice and, for that reason, the data from all cells have been aggregated into a single pool.
Figure 2A shows a partial electropherogram from a single separation study, demonstrating the glycine peak sandwiched between the larger peaks of taurine and glutamine, with the internal standard, α-amino adipic acid, at the end of the trace. The migration time of glycine was characterized by adding a known quantity of glycine to the experimental sample and also by demonstrating that the peak in the experimental sample was eliminated by the glycine-degrading enzyme glycine oxidase (Fig. 2B). Comparing the baseline glycine levels between WT and GlyT−/+ revealed an apparent trend towards greater glycine levels in GlyT−/+, although this trend fell short of statistical significance (t8 = 1.35, P = 0.21). However, blocking GlyT1 with NFPS significantly elevated glycine levels above untreated control retinas in the WT (t4 = 2.63, P = 0.058) but not in GlyT−/+ (t4 = 1.17, P = 0.31; Fig. 2C and D) mice. The final glycine concentration following NFPS treatment was virtually identical between genotypes (t8 = 0.25, P = 0.81), suggesting that the reduced number of retinal GlyT1s in the GlyT1−/+ animals accounts for the difference observed in extracellular glycine levels. Comparing the percent increase in glycine following NFPS treatment between the genotypes revealed that WT retinal glycine levels were augmented significantly more by NFPS than they were in GlyT−/+ (t8 = 2.37, P = 0.046) (Fig. 2D).
We tested the status of the NMDAR coagonist binding sites in single ganglion cells using D-serine added to the bathing medium to evaluate the comparative degree of NMDAR coagonist site occupancy (Gustafson et al., 2007). We simplified the retinal network by blocking post-synaptic inhibitory inputs as well as impulse activity. For these experiments, the superfusate (Ames medium) contained 1 μM TTX, 50 μM picrotoxinin and 10 μM strychnine. Unless otherwise indicated, all responses illustrated are the result of averaging 20 consecutive light-evoked responses. Figure 3A shows a currentclamp recording from a sustained On ganglion cell; the light-evoked control response showed transient and sustained components. In the presence of D-serine, the peak response increased, whereas the steadystate component remained relatively constant. When D,L-AP7 was added after returning to the control, the response was reduced to a small sustained component, indicating that NMDARs contributed substantially to the light response. To avoid the influence of shifting membrane potentials on the driving force for synaptic currents, we confined our experimental observations almost entirely to voltage-clamp studies. Figure 3B shows a voltage-clamp whole-cell recording (WCR) from a sustained On ganglion cell of a WT retina. The control response showed transient and sustained components, both of which increased during the application of 100 μM D-serine. When D,L-AP7 was added, the response decreased below that of the control, indicating a contribution from NMDARs. In several studies, we added 100 μM NBQX(2,3-dioxo-6-nitro-1,2,3,4-tetrahydrobenzo [f]quinoxaline-7-sulfonamide) (100 μM) to the D,L-AP7 bathing solution to determine the pharmacological identity of the response components that were insensitive to D,L-AP7. The addition of NBQX (100 μM) completely eliminated the light response, indicating that the D,L-AP7-insensitive light response can be attributed to AMPA/kainic acid receptors (not illustrated). Figure 3C shows a WCR from a GlyT1−/+ ganglion cell, in which the control response had a significant peak and plateau component, with a small overshoot at light off. When D-serine was added, the response was minimally increased, whereas D,L-AP7 significantly reduced the peak response, with a slight increase in the plateau component. Figure 3D presents a graphic summary of the relative change in the light-evoked responses of ganglion cells in the presence of D-serine. For the GlyT1−/+ mice, only 7/14 cells showed an increase in response to D-serine, whereas 9/9 cells in the WT retinas were enhanced through D-serine application; this latter population included seven On ganglion cells. Very little difference was evident in the outcome when comparing the peak amplitude and the response measurement obtained by integrating the current to give total charge.
A second method used for evaluating the state of NMDAR coagonist occupancy included a comparison of bath-applied NMDA with and without concurrent application of D-serine. These experiments involved WCRs from ganglion cells in the voltage-clamp mode in the absence of light stimulation. Bath application of NMDA can be expected to reach both synaptic and non-synaptic pools of NMDARs and, for that reason, might conceivably reveal different results, particularly if a large non-synaptic pool of NMDARs exists, which have a different level of coagonist site occupancy. We carried out this study by applying the sequence of (i) 30 s of NMDA (500 μM); (ii) several minutes of control; (iii) 30 s of NMDA + D-serine (100 μM); (iv) several minutes of control and (v) 30 s of NMDA. For these experiments, each cell recording was obtained from a virgin retina that had no prior exposure to pharmacological agents. The magnitude of the response to exogenous NMDA varied considerably and did not reveal any consistent difference between the WT and GlyT1−/+ mice. We were unable to determine whether this reflected differences in the size or type of the ganglion cell. Figure 4A illustrates a WCR obtained from a GlyT1−/+ ganglion cell bathed in TTX, in which a 30 s application of NMDA (upper trace) resulted in a slow inward current of 142 pA; after recovery and washout with control Ringer, the middle trace shows the results of NMDA + D-serine, with some enhancement of the peak response (219 pA at peak), although the response itself had different phasic components. The bottom trace shows recovery of the response to 500 μM NMDA after a washout period of several minutes (peak amplitude 124 pA). The duration bar in the upper trace of this and the other panels indicates a 30 s application of the agents. Figure 4B shows the same test sequence in a WT mouse in TTX (control response 27 pA) with D-serine enhancement (40 pA) and a return to NMDA after washout (31 pA). The D-serine increase was observed in 4/5 of the GlyT1−/+ mice, with one animal showing no change in the response to NMDA + D-serine. In contrast, 5/5 of the WT ganglion cells showed a measurable increase induced by D-serine. Because some amacrine cells are likely to be activated by these procedures and secondarily provide a synaptic input into ganglion cells, we also performed some experiments in which the influence of GABAA and glycine inputs was reduced through the application of picrotoxinin (50 μM) and strychnine (10 μM), in addition to TTX. Figure 4C illustrates the actions of NMDA in a GlyT1−/+ mouse ganglion cell in which NMDA was added to the bathing environment, which contained picrotoxinin and strychnine at the indicated levels. The control response had a peak value of −81 pA, whereas the NMDA + D-serine response was larger (−121 pA) with a washout and repeat NMDA response of −63 pA. Figure 4D shows the actions of NMDA on a WT mouse retina. The control response was −81 pA, whereas NMDA evoked a response of −244 pA and returning to a control NMDA application gave a peak response of −63 pA. Figure 4D summarizes the differences in the GlyT1−/+ and WT cells when D-serine was added to the NMDA bathing medium in TTX. The peak amplitude responses are expressed as a percent of the normalized control values. In TTX alone, the WT showed a slight increase over the GlyT1−/+ mice that was significant (t7 = 3.40, P = 0.014).
Figure 4E demonstrates that, when the experiments were performed in the presence of picrotoxinin/strychnine, the enhancement was greater for both the WT and GlyT1−/+ mice than those exposed to NMDA and D-serine in the presence of TTX. However, in this case, the increase in the presence of D-serine also remained greater for the WT ganglion cells. It is important to emphasize that D-serine potentiated the NMDA-evoked currents of ganglion cells in only some of the GlyT1−/+ (2/6) cells but it potentiated the NMDA-evoked currents in all of the cells from WT retinas (5/5); the difference between the two groups was significant (t8 = 2.31, P = 0.049).
We used Western blotting to test whether changes in NMDAR subunit expression could account for the functional saturation of NMDAR that we observed in GlyT+/− retinas. For example, a significant loss of expression of the NR1 subunit, which contains the coagonist binding site, could lead to apparent saturation of NMDARs even with unchanged glycine concentrations. Figure 5 shows that expression of the three NMDAR subunits tested (NR1, NR2A and NR2B) does not change substantially between WT and GlyT+/− mice. From densitometry measurements, the expression levels of NMDAR subunits in GlyT+/− as a percentage of the corresponding WT expression were: NR1, 130.9%; NR2A, 76.50% and NR2B, 88.3%. It should be noted that these are measurements of total subunit expression in the retina and thus do not distinguish among membrane-bound, cytoplasmic, synaptic or extrasynaptic proteins.
Two different high-affinity transport systems for glycine have been identified in the brain, including GlyT1 and GlyT2, each of which requires external Na+ to drive the uptake. Although it was originally thought that the expression of GlyT1s was confined to glial cells and GlyT2s were generally localized to neurons (Guastella et al., 1992; Smith et al., 1992), more recent evidence suggests that GlyT1s play an important role in glutamatergic pathways and are expressed in both pre- and post-synaptic neurons (Cubelos et al., 2005). Amacrine cells play an important role in controlling extracellular glycine levels in the retina (Pow, 1998). However, as opposed to the brain, the uptake of glycine in the retina is mediated exclusively by GlyT1, which in the mammalian and avian retinas is distributed in amacrine cells (Pow & Hendrickson, 1999), whereas GlyT1 is expressed in Müller cells in the amphibian retina (Lee et al., 2005). In the present study we used a mouse deficient in GlyT1 (Tsai et al., 2004) and found that the glycine concentration in the chamber containing an isolated retina showed very little change when GlyT1 was blocked with a selective antagonist (NFPS) (Bergeron et al., 1998; Aubrey & Vandenberg, 2001). In contrast, the same experiment carried out in a WT retina showed a low level of glycine released into the bathing medium and a comparatively larger increase in glycine when NFPS was applied. The measured levels of glycine that accumulated in the bathing chamber (low μM concentrations) were sufficient to activate the GlyT1 (Matsui et al., 1995; Supplisson & Bergman, 1997). Thus, GlyT1 contributed to the observed levels of glycine in the chamber and accounted for the large difference in glycine release when NFPS was applied. However, if the concentration of glycine in the chamber bathing environment was uniformly distributed throughout the extracellular spaces of the retina, all of the NMDAR coagonist sites would be saturated (Supplisson & Bergman, 1997) and this was not what we observed in ganglion cell recordings from WT retinas. In contrast, we observed NMDAR coagonist site saturation in many ganglion cells from the GlyT1−/+ retinas, as if the glycine levels that we saw in the chamber volume were relevant for those in the region of the NMDARs.
Supplisson & Bergman (1997) have provided convincing evidence from studies in oocytes, which coexpressed NMDARs and GlyT1s, that the unstirred layer effect at the surface of the membrane allowed the GlyT1 to effectively reduce the local concentration of glycine to levels below the NMDAR coagonist site threshold. Studies in the central nervous system have demonstrated that GlyT1s are found near NMDARs (Smith et al., 1992; Cubelos et al., 2005), suggesting a functional linkage between these two mechanisms. Thus, small regions of the extracellular space surrounding the glycine transporters can result in a local environment that is relatively depleted in glycine compared with adjacent regions where glycine transporters are absent. Of course, glycine from more distant regions surrounding the glycine transporter, where the background concentration is higher, will continually diffuse into the region of the transporter, establishing an equilibrium condition between uptake, nearby release and diffusion. This concept explains why the mean extracellular levels that we found in the bathing chamber do not reveal how local high-affinity transport mechanisms can change the microenvironment concentration of glycine.
The findings of this study clearly establish that the NMDAR coagonist sites have higher occupancy levels when GlyT1s are deficient in their expression. This deficiency could come about in two different ways, including an abnormal spatial distribution of the transporters or a deficiency in density, although they are not mutually exclusive. Evidence suggests that the brains of GlyT1−/+ mice have a reduced expression of the GlyT1 with a reduction to about half of their WT littermates (Tsai et al., 2004; Martina et al., 2005). If we translate our interpretation of glycine accumulation in the chamber, with the demonstrated importance of the GlyT1 activity for keeping external glycine levels low, it is inescapable that glycine levels surrounding the NMDARs of retinal ganglion cells are tonically exposed to a higher level of glycine when compared with their WT counterparts. For that reason, ganglion cell recordings from the GlyT1−/+ retinas are less enhanced by exogenous D-serine compared with the augmenting action of D-serine on ganglion cells from WT retinas.
Western blots of whole retina demonstrated the presence of NR1, NR2A and NR2B NMDAR subunits in our WT and Glyt1−/+ animals, consistent with other studies that have demonstrated a major representation of these subunits in the inner retina (Fletcher et al., 2000). The apparent saturation of GlyT1−/+ NMDAR coagonist sites in our results could be explained by a substantial loss of NR1 subunits, which contain the ‘glycine’ binding site. Alternatively, differences in glycine/D-serine sensitivity have been reported for NMDAR containing NR2A vs. NR2B subunits (Mishina et al., 1993; Matsui et al., 1995), so changes in the expression of these two subunits could also affect NMDAR coagonist sensitivity. Our Western blot data indicate that NR1, NR2A and NR2B subunit expression differs little between WT and GlyT1 transporter-deficient retinas, suggesting that there is little developmental compensation in NMDAR subunit composition in response to the knockout condition. It should be noted that the Western blot technique measures total levels of subunits and does not distinguish between membrane-bound subunits and subunits found in other cellular compartments.
Previous work has demonstrated that NMDARs must be functionally present for D-serine to enhance the light-evoked responses of retinal ganglion cells; the same applies to D-serine enhancement to the proximal negative response, a light-evoked field potential of the inner retina (Burkhardt, 1970; Gustafson et al., 2007). Blocking NMDARs with antagonists essentially eliminates the enhancing actions of exogenous D-serine. However, as D-serine has no affinity for the GlyT1 (Thomson, 1990; Grimwood et al., 1992; Broer et al., 1999; Ribeiro et al., 2002), we can assume that the more saturated condition of NMDARs in the GlyT1−/+ retinas reflects an elevated level of glycine in the extracellular space adjacent to NMDARs. Because the expression of GlyT1 is reduced in the GlyT1−/+ mice (Tsai et al., 2004; Martina et al., 2005), it seems likely that an insufficient number of transporters are available to keep all regions surrounding the NMDARs at reduced levels of glycine. This could explain why some ganglion cells in the GlyT1−/+ retinas showed near saturation of the NMDAR coagonist sites, whereas others showed complete saturation, based on the exogenous application of D-serine.
Exogenous application of NMDA generated a significantly larger current in the presence of inhibitory antagonists (picrotoxinin/strychnine/TTX) than it did when applied in TTX alone. As the chloride reversal potential was established at approximately −88 mV, the holding potential of −65 mV in our experiments meant that chloridemediated inhibition would generate a current of opposite polarity to that evoked by the direction actions of NMDA on the ganglion cell. As NMDARs are also found on amacrine cells (Dowling & Boycott, 1965; Witkovsky & Dowling, 1969; Slaughter & Miller, 1983; Muller & Marc, 1990; Dixon & Copenhagen, 1992), their activation could enhance feedforward inhibition onto ganglion cells. Thus, blocking post-synaptic inhibition with picrotoxinin/strychnine will eliminate the algebraic interaction and enhance the light responses. However, as feedforward inhibition from amacrine cells onto ganglion cells is carried out through dendrodendritic connections, inhibitory synaptic inputs into the dendrites can shunt the flow of excitatory current generated at more distal dendritic locations (Miller, 1979); this inhibitory action will also reduce the flow of excitation through the dendrites into the soma. Thus, the larger currents evoked by NMDA in the presence of toxins is consistent with the elimination of shunting inhibition in the dendrites as well as the algebraic interaction of the currents in the soma.
Although the present study strongly indicates that a deficiency of GlyT1s leads to an elevated extracellular level of glycine and a more saturated coagonist state of NMDARs, this does not mean that glycine is the normal tissue coagonist of NMDARs. In fact, studies performed in the amphibian and rat retina have raised the possibility that D-serine is the primary, functional NMDAR coagonist (Stevens et al., 2003). Additional studies in the retina have demonstrated that, when D-serine levels are reduced using enzymatic degradation applied to the whole retina, the light-evoked NMDAR contribution is similar to that observed when the NMDARs are blocked with an antagonist (Gustafson et al., 2007). Thus, one possible view of GlyT1 found in amacrine and glial cells is that its high sensitivity and proximity to NMDARs of ganglion cells are sufficient to reduce glycine in the microenvironment of these receptor sites, such that D-serine is the major player as the NMDAR coagonist. It remains unclear why GlyT1 expression seems to be restricted to amacrine cells in the mammalian and avian retina (Pow & Hendrickson, 1999) or to Müller cells in the amphibian retina (Lee et al., 2005). It is equally unclear at the moment whether glycine plays a role in more dynamic modulation of NMDARs, as glycine plays a role as an inhibitory neurotransmitter (Miller et al., 1977; Marc, 1989) and presumably goes through lightand dark-driven fluctuations in release and local concentration changes. However, because D-serine does not appear to be taken up by a high-affinity transport system (O’Brien et al., 2005) but through a lower affinity amino solute carrier type 2-type amino acid transport pathway (Dun et al., 2007), the extracellular levels of D-serine may more readily reach the coagonist threshold in strategically placed release sites near NMDARs. Additional studies of the D-serine release sites and the release mechanism will undoubtedly be needed to further clarify the role of D-serine vs. glycine as dynamic and static contributors to NMDAR function in ganglion cells of the retina.
Recent studies using rat and mouse retinal slices (Kalbaugh et al., 2009) have suggested that D-serine and glycine both play a role as NMDAR coagonists, with glycine contributing to a lightevoked modulation and D-serine serving as a background, steadystate modulator of NMDAR coagonist occupancy. Unfortunately, the evaluation of the D-serine contribution in the study of Kalbaugh et al. (2009) was performed using the enzyme D-amino acid oxidase, which is much slower and less efficacious than the faster, more effective enzyme D-serine deaminase. The use of the latter enzyme in studies of the salamander retina resulted in a suppression of light-evoked NMDAR currents similar to that observed when NMDARs were blocked with antagonists (Gustafson et al., 2007). Nevertheless, the concept of both dynamic and static regulation of NMDAR function through the influence of coagonists brings a new and exciting challenge to our understanding of NMDAR function in the retina as well as the other partitions of the central nervous system.
We appreciate many helpful discussions with Eric Gustafson and excellent editing and figure illustration assistance from Derek Miller.