Extrasynaptic NMDA Receptors on Rod Pathway Amacrine Cells: Molecular Composition, Activation, and Signaling

Identification of AII and A17 amacrine cells and rod bipolar cells in retinal slices

The cellular elements and basic synaptic connectivity of the rod bipolar cell-AII-A17 microcircuit are illustrated in Figure 1. As reliable identification of the two types of amacrine cells, both with respect to each other and with respect to other types of amacrine cells, is important for the conclusions of this study, we emphasize the procedure followed for the visual targeting, recording, and identification of these cells in retinal slices. The primary targeting criterion was the shape and location of the cell bodies at the border between the inner nuclear layer and the inner plexiform layer (Fig. 1A,D). The AII amacrine cell is a bistratified, narrow-field amacrine cell with a cell body that spans the border between the inner nuclear and the inner plexiform layer, and a thick apical dendrite that tapers as it descends into the inner plexiform layer (Fig. 1B). In contrast, the A17 amacrine cell is a diffuse, wide-field amacrine cell with a dome-shaped cell body and long, thin processes that carry distinct varicosities and terminate near the ganglion cell layer of the retina (Fig. 1E). For dual recordings between rod bipolar cells and either AII or A17 amacrines, rod bipolar cells were targeted based on the location of a cell body at the distal end of the inner nuclear layer, close to the outer plexiform layer. For unequivocal identification, all recorded cells were visualized with fluorescence microscopy following the recording. For AII amacrine cells, this allowed us to verify their narrow-field bistratified morphology (Fig. 1B) and for A17 amacrine cells we could identify their long, thin processes that carry varicosities and terminate near the ganglion cell layer of the retina (Fig. 1E). Rod bipolar cells have a single long axon that terminates with a number of large axon terminals in the proximal part of the inner plexiform close to the ganglion cell layer (Fig. 1C).

• Download figure
• Open in new tab
• Download powerpoint

Figure 1.

Identification of microcircuit with presynaptic rod bipolar cells and postsynaptic AII and A17 amacrine cells in the rat retinal slice preparation. A, IR-DIC videomicrograph of a retinal slice with cell body (arrow) and apical dendrite of an AII amacrine cell. Scale bar: A, B, 10 μm. B, Wide-field fluorescence image of AII amacrine cell filled with Alexa-594 via patch pipette (MIP of image stack after deconvolution). C, Wide-field fluorescence image of rod bipolar cell filled with Alexa-594 via patch pipette (MIP of image stack after deconvolution) overlaid on IR-DIC videomicrograph of retinal slice. Scale bar, 10 μm. D, IR-DIC videomicrograph of a retinal slice with cell body (arrow) of an A17 amacrine cell. Scale bar: D, E, 10 μm. E, Wide-field fluorescence image of A17 amacrine cell filled with Alexa-594 via patch pipette (MIP of image stack after deconvolution). F, G, NMDA receptors do not contribute to synaptic input from rod bipolar cells to AII (F) or A17 (G) amacrine cells. F, Schematic figure of experimental configuration with simultaneous dual whole-cell voltage-clamp recording of presynaptic rod bipolar cell (blue) and postsynaptic AII amacrine cell (magenta) in rat retinal slice (top). Top, Voltage command applied to rod bipolar cell with depolarization (100 ms) from −60 to −10 mV. Three bottom traces represent current measured in AII amacrine cell in response to rod bipolar cell depolarization in the following conditions: during control (average of three traces), during application of the non-NMDA receptor antagonist CNQX (10 μm; average of five traces), and after washout of CNQX (average of four traces). Note the transient inward current in response to depolarization of rod bipolar cell in control condition, complete block by CNQX, and partial recovery after washout of CNQX. In this and subsequent figures, the nominal concentration of Mg²⁺ in the extracellular solution was zero (unless otherwise indicated), and the bath also contained TTX (300 nm) to block voltage-gated Na⁺ channels, bicuculline (10 μm) to block GABA_A receptors, strychnine (1 μm) to block glycine receptors, and the NMDA receptor coagonist d-serine (200 μm). G, Schematic figure of experimental configuration with simultaneous dual whole-cell voltage-clamp recording of presynaptic rod bipolar cell (blue) and postsynaptic A17 amacrine cell (green) in rat retinal slice (top). Top, Voltage command applied to rod bipolar cell with depolarization (100 ms) from −60 to −10 mV. Three bottom traces represent current measured in A17 amacrine cell in response to rod bipolar cell depolarization in the following conditions: during control (average of four traces), during application of the non-NMDA receptor antagonist CNQX (10 μm; average of four traces), and after washout of CNQX (average of two traces). Note the transient inward current in response to depolarization of rod bipolar cell in control condition, complete block by CNQX, and partial recovery after washout of CNQX.

NMDA receptors do not contribute to EPSCs in AII and A17 amacrine cells evoked by depolarization of rod bipolar cells

There is evidence for the expression of NMDA receptors by both AII and A17 amacrine cells (Hartveit and Veruki, 1997; Zhou and Dacheux, 2004; Kothmann et al., 2012; Zhou et al., 2016), but it is less clear whether the receptors contribute to mediating glutamatergic synaptic input from rod bipolar cells. To investigate the potential involvement of NMDA receptors, we performed simultaneous dual patch-clamp recordings between pairs of rod bipolar cells and AII amacrine cells and between pairs of rod bipolar cells and A17 amacrine cells (Fig. 1F,G). For both sets of recordings, the bath solution was changed to a Mg²⁺-free solution (including 200 μm of the NMDA receptor coagonist d-serine), at least 10 min before breaking into the cells, to eliminate the voltage-dependent Mg²⁺ block of NMDA receptors.

In dual recordings from pairs of rod bipolar cells and AII amacrine cells, with both cells in the whole-cell voltage-clamp configuration (V_hold = −60 mV for both cells), we first verified that the cells were synaptically connected by applying a depolarizing voltage step to the rod bipolar cell (to −10 mV). In the presence of a synaptic connection, the depolarization evoked a large transient inward current in the AII amacrine cell (Fig. 1F), corresponding to an EPSC (Singer and Diamond, 2003; Veruki et al., 2003; Trexler et al., 2005). Next, the bath solution was exchanged to include 10 μm CNQX to block non-NMDA receptors and potentially reveal a non-NMDA receptor-independent component of the EPSC. In all cases, however, application of CNQX completely blocked the rod bipolar cell-evoked EPSC in AII amacrine cells (Fig. 1F; n = 6 cell pairs). In the control condition, the average peak amplitude (relative to baseline) was 116 ± 20 pA and in the presence of CNQX the average was 3.2 ± 1.0 pA. The current in the presence of CNQX was indistinguishable from baseline membrane noise. For three rod bipolar-AII cell pairs, we observed not only a block of the EPSC in the presence of CNQX, but also a partial recovery of the EPSC after washout of CNQX (Fig. 1F), suggesting that rundown of transmitter release was not the reason for the lack of response.

We next performed the same type of experiment for synaptically connected pairs of rod bipolar cells and A17 amacrine cells. In contrast to the synapse between a rod bipolar cell and an AII amacrine cell, depolarization of a rod bipolar cell (to −10 mV) evoked a much smaller EPSC in an A17 amacrine cell (Fig. 1G). For seven synaptically connected cell pairs, the average peak amplitude of the depolarization-evoked EPSC in the control condition was 8.2 ± 1.2 pA (range 5.3–15 pA). In the presence of CNQX, the average peak response in the same A17 amacrines was 2.1 ± 0.5 pA (range 0.2–3.8 pA). Similar to AII amacrine cells, the current in the presence of CNQX was indistinguishable from the baseline noise. After washout of CNQX, recovery (partial) was observed in three of the paired recordings (Fig. 1G). Together, our results suggest that NMDA receptors do not contribute directly to synaptic transmission between rod bipolar cells and either A17 or AII amacrine cells. In the case of the AII, these results corroborate earlier studies by Singer and Diamond (2003) and Trexler et al. (2005).

Whereas A17 amacrines receive glutamatergic input only from rod bipolar cells, AII amacrines are likely to receive glutamatergic input from at least some types of OFF-cone bipolar cells (Veruki et al., 2003; Graydon et al., 2018). However, it is unlikely that NMDA receptors are expressed at the OFF-cone bipolar cell to AII synapse. First, EM immunolabeling for the GluN1 subunit suggests that only one of the two profiles postsynaptic to OFF-cone bipolar cell ribbons (presumably one ganglion cell and one amacrine cell) expresses NMDA receptors (Fletcher et al., 2000). This profile most likely belongs to a ganglion cell, as a synaptic NMDA component has consistently been identified in OFF-ganglion cells (e.g., Manookin et al., 2010; Buldyrev et al., 2012; Stafford et al., 2014). In addition, no NMDA component was revealed in the synaptic input to AII amacrines from Type 2 OFF-cone bipolar cells (that mediate the predominant input from OFF-cone bipolar cells to AIIs) (Graydon et al., 2018). For these reasons, we did not further test for synaptic NMDA receptors at OFF-cone bipolar cell to AII synapses in this study.

Ambient glutamate can activate NMDA receptors on both AII and A17 amacrine cells

That NMDA receptors are expressed by AII and A17 amacrines, but not activated directly after neurotransmitter release from presynaptic rod bipolar cells, suggests that the receptors are located outside the postsynaptic density of the dyad synapses, at an extrasynaptic location. In the present context, extrasynaptic could mean perisynaptic (i.e., outside but in the immediate vicinity of the synapse) or further away. In a previous study, our laboratory provided physiological evidence that few, if any, NMDA receptors are located at the cell bodies of AII and A17 amacrine cells, and therefore are unlikely to be responsible for the responses evoked in either cell type by application of NMDA (Zhou et al., 2016). Regardless of the exact location of the NMDA receptors, we can propose two different mechanisms for their potential activation under physiological conditions. One mechanism involves direct spillover of synaptically released glutamate, and the other mechanism involves activation by ambient glutamate. With respect to the first mechanism, there is evidence that glutamate spillover at the rod bipolar cell axon terminal can activate glutamate transporters on the same and neighboring axon terminals (Veruki et al., 2006; Wersinger et al., 2006). Depending on their exact location, this suggests that NMDA receptors on AII and A17 amacrine cells under certain conditions could be activated by spillover of glutamate from rod bipolar cells after synaptic release. With respect to the second mechanism, there is evidence that the concentration of ambient glutamate (of neuronal and/or glial origin) in the extracellular environment of the CNS is sufficiently high to activate different extrasynaptic receptors (Chen and Diamond, 2002; Cavelier et al., 2005; Tzingounis and Wadiche, 2007).

If activation of NMDA receptors contributes to membrane noise under baseline conditions, it should be possible to block their contribution by blocking the receptors. To test this hypothesis, we recorded membrane noise for ∼10 min before adding the NMDA receptor antagonist CPP (20 μm) to the bath solution. The bath solution was Mg²⁺-free and contained CNQX, bicuculline, strychnine, and TTX (to block non-NMDA, GABA_A, and glycine receptors, and voltage-gated Na⁺ channels, respectively), as well as d-serine. We analyzed the membrane noise by calculating the variance of the membrane current recorded during a 2 min period in the control condition and a 2 min period in the presence of CPP. For the AII amacrine cell illustrated in Figure 2A, the addition of CPP markedly decreased the membrane noise, from an average variance of 31 pA² in the control condition to 8.0 pA² in the presence of CPP. This suggested that gating of NMDA receptor channels contributed substantially to membrane noise in our baseline condition. That the membrane noise was suppressed by the competitive antagonist CPP also suggests that the channel gating was driven by the presence of an endogenous agonist of the NMDA receptor, which is likely to be glutamate. For AII amacrine cells tested in this way, the average variance in the control condition was 20.0 ± 3.9 pA²; and in the presence of CPP, it was 7.3 ± 1.2 pA² (p = 0.0395, paired t test; n = 4 cells), corresponding to an average reduction of 60 ± 8% (Fig. 2A).

• Download figure
• Open in new tab
• Download powerpoint

Figure 2.

Activation of NMDA receptors by ambient glutamate generates membrane current noise in AII and A17 amacrine cells. A, Continuous current recorded from an AII amacrine cell in whole-cell voltage clamp in control condition (top) and during application of CPP (20 μm) in the extracellular bath to block NMDA receptors (bottom). Note the reduction of membrane current noise in the presence of CPP, suggesting activation of NMDA receptors by ambient glutamate in the control condition. Here and later, the identity of the cell from which the recording was performed is indicated by the schematic (left), with color and shape as in Figure 1. Bar graphs (right) represent membrane current noise measured as variance (here and later, bars represent mean ± SEM) in AII amacrine cells (n = 4 cells) in control (without CPP; left bar) and in the presence of CPP (right bar). For each cell (here and below), current variance was averaged over 2 min. Here and later, data points for the same cell are connected by lines. Statistical comparisons between averages: n.s., No significant difference (p > 0.05). *p < 0.05. B, Continuous current recorded from an A17 amacrine cell in whole-cell voltage clamp in control condition (top) and during application of CPP (20 μm) in the extracellular solution to block NMDA receptors (bottom). Note the reduction of membrane current noise in the presence of CPP, suggesting activation of NMDA receptors by ambient glutamate in the control condition. Bar graphs (right) represent membrane current noise measured as variance in A17 amacrine cells (n = 6 cells) in control and in the presence of CPP. C, Continuous current recorded from an AII amacrine cell in whole-cell voltage clamp at 32°C in control condition (top) and during application of CPP (20 μm) in the extracellular solution to block NMDA receptors (bottom). Note the reduction of membrane current noise in the presence of CPP, suggesting activation of NMDA receptors by ambient glutamate in the control condition at 32°C. Bar graphs (right) represent membrane current noise measured as variance in AII amacrine cells (n = 5 cells) at 32°C in control and in the presence of CPP. D, E, No effect of CPP on membrane current noise when the NMDA receptors are blocked by the open-channel blocker MK-801 (2 mm) in the intracellular solution. D, Continuous current recorded from an AII amacrine cell in whole-cell voltage clamp in control condition (top) and during application of CPP (20 μm) in the extracellular solution to block NMDA receptors (bottom). Note that including MK-801 intracellularly blocks the effect of CPP on membrane current noise, suggesting that the noise is generated by activation by ambient glutamate of NMDA receptors in the AII amacrine cell itself. Bar graphs (right) represent membrane current noise measured as variance in AII amacrine cells (n = 5 cells; MK-801 in the intracellular solution) in control and in the presence of CPP. E, Continuous current recorded from an A17 amacrine cell in whole-cell voltage clamp in control condition (top) and during application of CPP (20 μm) in the extracellular solution to block NMDA receptors (bottom). Note that including MK-801 intracellularly blocks the effect of CPP on membrane current noise, suggesting that the noise is generated by activation by ambient glutamate of NMDA receptors in the A17 amacrine cell itself. Bar graphs (right) represent membrane current noise measured as variance in A17 amacrine cells (n = 4 cells; MK-801 in the intracellular solution) in control and in the presence of CPP. F, G, Effect of blocking glutamate uptake on membrane noise. F, Continuous current recorded from an AII amacrine cell in whole-cell voltage clamp in control condition (top), during application of the glutamate transporter blocker TBOA (50 μm; middle), and during combined application of TBOA and CPP (20 μm; bottom). Note that membrane current noise is reduced by TBOA and further reduced by CPP. Bar graphs (right) represent membrane current noise measured as variance in AII amacrine cells (n = 4 cells) in control (without TBOA and CPP; left bar), in the presence of TBOA, and in the presence of TBOA+CPP. G, Continuous current recorded from an A17 amacrine cell in whole-cell voltage clamp in control condition (top), during application of the glutamate transporter blocker TBOA (50 μm; middle), and during combined application of TBOA and CPP (20 μm; bottom). Note that membrane current noise is increased by TBOA and reduced by CPP. Bar graphs (right) represent membrane current noise measured as variance in A17 amacrine cells (n = 3 cells) in control, in the presence of TBOA, and in the presence of TBOA+CPP.

Similar results were seen for A17 amacrine cells tested the same way. For the cell illustrated in Figure 2B, the membrane noise was reduced from 4.4 pA² in control to 3.1 pA² in the presence of CPP. For all A17 amacrines tested, the membrane noise was reduced from 6.3 ± 0.9 pA² in control to 3.2 ± 0.2 pA² in the presence of CPP (p = 0.0169, paired t test; n = 6 cells), corresponding to an average reduction of 46 ± 6% (Fig. 2B). These results suggest that, for both AII and A17 amacrine cells, ambient levels of glutamate are sufficiently high to gate NMDA receptor channels.

We next examined whether the increased activity of glutamate transporters at a more physiological temperature would reduce the level of ambient glutamate and/or prevent spillover of synaptic glutamate (Wadiche et al., 1995; Asztely et al., 1997; Rauen et al., 1998) and thus reduce or eliminate the observed NMDA receptor-mediated membrane noise. When we repeated these experiments for AII amacrine cells at 32°C, the average variance in the control condition was 57.7 ± 15.4 pA²; and in the presence of CPP, it was 39.3 ± 11.7 pA² (p = 0.0223, paired t test; n = 5 cells; Fig. 2C), corresponding to an average reduction of 33 ± 9%. This suggested that, even at more physiologically relevant temperatures, with enhanced glutamate transporter activity, the level of ambient glutamate is sufficiently high to activate NMDA receptors.

The reduction of membrane noise in AII and A17 amacrines by CPP does not demonstrate conclusively that the NMDA receptor channels are located on these amacrine cells, as the reduction of noise could be due to transsynaptic effects. To explore this possibility, we performed experiments identical to those above but included the NMDA receptor antagonist MK-801 in the intracellular pipette solution. MK-801 acts as an open-channel blocker and can also block NMDA receptors from the intracellular side (Wong et al., 1986; Huettner and Bean, 1988). If intracellularly applied MK-801 reduces the effect of extracellularly applied CPP on the membrane noise, this would verify that the membrane noise must include gating of NMDA receptor channels on the AII and A17 amacrine cells themselves, and cannot be explained by gating of NMDA receptors exclusively located elsewhere in the network of neurons. For AII amacrine cells with 2 mm MK-801 in the recording pipette, the average variance was 13.4 ± 2.4 pA² in the control condition, not significantly different from 12.9 ± 1.9 pA² in the presence of CPP (p = 0.41, paired t test; n = 5 cells; Fig. 2D). For A17 amacrine cells (with 2 mm MK-801 in the recording pipette), the average control variance was 6.6 ± 1.8 pA², not significantly different from 7.1 ± 1.8 pA² in the presence of CPP (p = 0.385, paired t test; n = 4 cells; Fig. 2E). These results suggested that the NMDA receptor-mediated membrane noise observed in both AII and A17 amacrine cells, activated by ambient glutamate, is generated predominantly (even exclusively) by receptors located in the membrane of the recorded cells themselves and does not depend on potential transsynaptic effects.

Increasing glutamate spillover can increase activation of NMDA receptors on AII and A17 amacrine cells

In the retina (Rauen et al., 1998), like elsewhere in the CNS (for review, see Danbolt, 2001), the concentration of glutamate in different extracellular environments is a result of the balance between release, synaptic and nonsynaptic from neurons and potentially from glial cells, and uptake, primarily into glial cells, but also into neurons. We therefore hypothesized that pharmacologically blocking uptake mediated by glutamate transporters in glial cells and neurons could increase glutamate spillover and ambient glutamate and subsequently the level of activation of NMDA receptors on AII and A17 amacrine cells. To investigate this, we performed experiments where we used the specific, but nonselective, nontransported glutamate transporter blocker TBOA (Shimamoto et al., 1998) to increase extracellular glutamate.

In the first set of experiments, we examined whether application of TBOA (50 μm) increased membrane noise, measured as the variance of the membrane current recorded in whole-cell voltage clamp (see above), and whether a TBOA-evoked increase was mediated by activation of NMDA receptors. For the AII amacrine cell illustrated in Figure 2F, however, the effect of TBOA was opposite of that hypothesized. The total current variance decreased from 23 pA² in the baseline condition to 17 pA² in the presence of TBOA. As in the experiments presented above, non-NMDA receptors, GABA_A receptors, glycine receptors, and voltage-gated Na⁺ channels were blocked pharmacologically and d-serine was added to the bath solution. When we added CPP (20 μm) to block NMDA receptors, in the maintained presence of TBOA, the membrane noise was further reduced to a level of 8.7 pA². For all AII amacrine cells tested in this way, adding TBOA decreased the membrane noise from 22.9 ± 2.2 pA² to 13.5 ± 2.5 pA² (p = 0.03, adjusted p value, one-way ANOVA, followed by Holm–Sidak's multiple-comparisons test) and adding CPP further reduced the noise to 6.1 ± 1.2 pA² (p = 0.043, adjusted p value, one-way ANOVA, followed by Holm–Sidak's multiple-comparisons test; n = 4 cells; Fig. 2F). The reduction of current variance by CPP suggested that NMDA receptor channel gating contributed to the membrane noise both in the baseline condition and in the presence of TBOA, which is further investigated below.

When A17 amacrine cells were tested in the same way with TBOA (50 μm), the effect on membrane noise was very different from AII amacrines and consistent with our original hypothesis. For the A17 illustrated in Figure 2G, the membrane noise increased from 9.8 pA² in the baseline condition to 26 pA² in the presence of TBOA. When we added CPP to block NMDA receptors (in the maintained presence of TBOA), the membrane noise was reduced to 2.0 pA². For A17 amacrine cells tested in this way, adding TBOA increased the membrane noise from 12.0 ± 3.4 pA² to 28.7 ± 3.0 pA² (p = 0.006, adjusted p value, one-way ANOVA, followed by Holm–Sidak's multiple-comparisons test) and adding CPP reduced the noise to 1.89 ± 0.06 pA² (p = 0.002, adjusted p value, one-way ANOVA, followed by Holm–Sidak's multiple-comparisons test; n = 3 cells; Fig. 2G). These results suggested that, when glutamate transport was blocked with TBOA, increased extracellular glutamate enhanced the activation and gating of NMDA receptors on A17 amacrines, leading to an increase of membrane noise.

The reduced membrane noise of AII amacrine cells in the presence of TBOA was surprising. Closer inspection of the raw data, however, suggested that the effect of TBOA was more subtle than indicated by the overall reduction of current variance. As can be seen by the current traces illustrated in Figure 2F, TBOA did strongly reduce the low-frequency noise, but this was accompanied by an increase in higher-frequency noise. This higher-frequency noise was virtually eliminated following addition of CPP, suggesting that it corresponded to NMDA receptor channel gating. We explored this by analyzing the spectral density of the current noise in the same four AII amacrine cells (Fig. 2F), both in control and in the presence of TBOA. We found that TBOA reduced the spectral density of frequencies ∼<10 Hz but increased the spectral density of frequencies ∼>10 Hz (Fig. 3A). To quantify this result, we measured the spectral density at 3.9 Hz and 39 Hz for each cell in both control and TBOA. Both the observed decrease (at 3.9 Hz) and the observed increase (at 39 Hz) of the spectral density was statistically significant (p = 0.020 and p = 0.021, respectively, ratio t test, two-tailed; Fig. 3B,C). At this time, we cannot explain the cellular or molecular mechanisms that result in the reduction of variance of membrane noise in AII amacrines exposed to TBOA, but they may involve transsynaptic effects and transmission through electrical synapses (Veruki and Hartveit, 2002a,b). Regardless of any accompanying changes in membrane noise, we decided to investigate directly whether blocking glutamate uptake could evoke an increase in NMDA receptor-mediated current in AII amacrine cells.

• Download figure
• Open in new tab
• Download powerpoint

Figure 3.

Frequency-dependent effect of TBOA on membrane noise in AII amacrine cells. A, Power density spectra obtained from the current noise in an AII in the control condition (black line), during application of TBOA (50 μm; red line), and during combined application of TBOA (50 μm) + CPP (20 μm; gray line). Note the crossing of the control and TBOA spectra at ∼10 Hz. B, Spectral density measured at 3.9 Hz for AII amacrine cells (n = 4) in control (without TBOA and CPP; left bar), in the presence of TBOA and in the presence of TBOA+CPP. Here and in C, the data are from the same cells as illustrated in Figure 2F. C, Spectral density measured at 39 Hz in control (without TBOA and CPP; left bar), in the presence of TBOA and in the presence of TBOA+CPP. Statistical comparisons between averages: *p < 0.05.

The following experiments were performed with normal Mg²⁺ in the extracellular solution. To alleviate the Mg²⁺-dependent voltage block of NMDA receptor channels, we applied voltage ramps (from 50 to −90 mV at 200 mV/s) and measured the average current in a window between 35 and 45 mV of the I–V curve obtained after subtracting a linear leak component calculated by fitting a straight line to the current measured between −90 and −60 mV (see Materials and Methods). To block depolarization-evoked activation of voltage-gated K⁺ channels that would interfere with the measurements, the pipette solution contained Cs⁺ and TEA (see Materials and Methods). In addition, the extracellular solution contained TEA (20 mm) and 3,4-DAP (0.1 mm) to block voltage-gated K⁺ channels faster and more effectively.

If application of TBOA evokes an increased activation of NMDA receptors, it should be revealed as an increase of the current measured in the averaging window between 35 and 45 mV. For the AII amacrine cell illustrated in Figure 4A, we applied voltage ramps approximately every 4 s. During the first period of ∼5 min, the current response (measured in the averaging window) decreased markedly, corresponding to a gradually increasing block of voltage-gated K⁺ currents mediated by diffusion into the cell of Cs⁺ and TEA from the pipette solution. After this period, we switched to the extracellular solution with TEA and 3,4-DAP and observed that the current response continued to decrease, reflecting increased block of voltage-gated K⁺ channels, before stabilizing at a relatively constant plateau (Fig. 4A). When TBOA (50 μm) was added to the bath solution, the current response increased by ∼60 pA (from 30 to 90 pA; Fig. 4A). When TBOA was washed out, the current decreased approximately to the level before TBOA was applied. For quantitative analysis, we used this more stable period as the control condition. The I–V curves illustrated in Figure 4B correspond to leak-subtracted currents evoked by voltage ramps applied at two time points during the recording in Figure 4A: during application of TBOA (1; “TBOA”) and after washout of TBOA (2; “Control”). The difference between the two I–V curves corresponds to the I–V curve of the net current evoked by TBOA (Fig. 4C) and displays a voltage dependence similar to that expected for a current mediated by NMDA receptors. For AII amacrine cells tested in this way, TBOA increased the current from a control level of 18.3 ± 3.8 pA to 54.7 ± 8.8 pA (p = 0.0033, paired t test; n = 6 cells; Fig. 4D).

• Download figure
• Open in new tab
• Download powerpoint

Figure 4.

Blocking glutamate uptake evokes a conductance with J-shaped I–V relationship in AII amacrine cells. A, Current activated in an AII increases when glutamate uptake is blocked by bath application of TBOA (50 μm; indicated by continuous horizontal line above data points). Here, and in E, G, and I, each data point corresponds to current activated by a voltage-ramp stimulus (50 mV to −90 mV; 200 mV/s; repeated at 1 Hz) and measured by averaging the current between 35 and 45 mV in the resulting I–V relationship. Linear leak current was subtracted by first fitting a straight line to the I–V relationship between −90 and 60 mV and then subtracting the resulting function from the ramp current over the whole range of the voltage ramp. Here, and in E, G, and I, the extracellular solution contained 1 mm Mg²⁺. During the early period of recording, the response is dominated by depolarization-evoked K⁺ currents that are gradually suppressed by Cs⁺ and TEA in the pipette solution (0–7 min). At ∼7 min recording time, TEA (20 mm) and 3,4-DAP (0.1 mm) are added to the extracellular bath (broken horizontal line), leading to additional suppression of depolarization-evoked K⁺ currents and a stable response plateau of ∼30 pA (20–40 min). Note the increase of depolarization-evoked current during application of TBOA and reduction after washout of TBOA. Numbers (1, red; 2, black) correspond to time points of I–V traces displayed in B. B, Individual I–V relationships (after linear leak subtraction) obtained by voltage-ramp stimuli applied during application of TBOA (1, red) and after washout of TBOA (2, black). C, I–V relationship of net TBOA-evoked current obtained by subtracting I–V relationships in left panel (“TBOA”-“Control”). Note the J-shaped I–V relationship. D, Membrane current measured (as in A) in AII amacrine cells (n = 6 cells) in control and in the presence of TBOA. Here and below, for each cell, current responses were averaged over 5 min in each condition. E, Current activated in an AII amacrine cell increases when glutamate uptake is blocked by bath application of TBOA (50 μm; indicated by thick, broken, horizontal line above data points), but not when TBOA is applied in the presence of CPP (20 μm; indicated by continuous horizontal line above data points). Depolarization-evoked K⁺ currents blocked by Cs⁺ and TEA in the pipette solution and by TEA (20 mm) and 3,4-DAP (0.1 mm) in the extracellular solution (from the start of the recording; thin, broken, horizontal line above data points). Note the increase of depolarization-evoked current during application of TBOA alone, but not when TBOA was applied with CPP. F, Membrane current measured (as in E) in AII amacrine cells (n = 6 cells) in control, in the presence of TBOA and CPP, and in the presence of TBOA alone. G, Current activated in an AII amacrine cell increases when glutamate uptake is blocked by bath application of TBOA (50 μm; indicated by broken, horizontal line above data points), but not when TBOA is applied in the presence of the NMDA receptor antagonist CPP (20 μm; indicated by continuous, horizontal line above data points). Depolarization-evoked K⁺ currents blocked by Cs⁺ and TEA in the pipette solution, but the extracellular solution did not contain K⁺ channel blockers. Note the increase of depolarization-evoked current during application of TBOA alone, but not when TBOA was applied with CPP. H, Membrane current measured (as in G) in AII amacrine cells (n = 6 cells) in control, in the presence of TBOA+CPP, and in the presence of TBOA alone. I, Current activated in an AII amacrine cell decreases when NMDA receptors are blocked by bath application of CPP (20 μm; indicated by horizontal line above data points). Depolarization-evoked K⁺ currents blocked by Cs⁺ and TEA in the pipette solution, but the extracellular solution did not contain K⁺ channel blockers. Note the decrease of depolarization-evoked current during application of CPP and reversal when CPP is washed out. J, Membrane current measured (as in I) in AII amacrine cells (n = 14 cells) in control and in the presence of CPP. Statistical comparisons between averages: n.s., No significant difference (p > 0.05). *p < 0.05.

If this TBOA-evoked current was mediated by activation of NMDA receptors, it should be blocked by CPP. For the cell illustrated in Figure 4E, the response stabilized to a relatively constant level after ∼10 min. When we applied TBOA (50 μm) and CPP (20 μm) together, there was no obvious change in the current response (Fig. 4E). Following washout of CPP, TBOA applied alone evoked an increase in response of ∼45 pA. This was followed by a slow reversal when TBOA was washed out at the end of the recording period (Fig. 4E). For AII amacrine cells tested in this way, the response increased from an average of 41.1 ± 6.6 pA in the presence of TBOA and CPP to an average of 80 ± 20 pA in the presence of TBOA alone after washout of CPP (p = 0.0263, adjusted p value, one-way ANOVA, followed by Holm–Sidak's multiple-comparisons test; n = 6 cells; Fig. 4F). These results suggest that CPP can block the effect of TBOA on the voltage ramp-evoked current, supporting the hypothesis that the increase of current evoked by TBOA is mediated by activation of NMDA receptors.

Whereas the addition of TEA and 3,4-DAP to the extracellular solution was useful in the sense that it led to a faster and more effective block of voltage-gated K⁺ currents (compared with Cs⁺ and TEA only in the pipette solution), it could overestimate the effect of TBOA because block of voltage-gated K⁺ channels in bipolar cells could lead to depolarization, increased excitability, and increased release of glutamate. Indeed, when we recorded from AII amacrine cells and switched to a bath solution with TEA and 3,4-DAP, we observed a strong increase in bursting synaptic input occurring at a frequency of ∼0.5 Hz (data not shown). Although increased release of glutamate does not invalidate the observed effect of TBOA as such, we performed control experiments where we examined the effect of TBOA using our standard bath solution without TEA and 3,4-DAP.

For the AII amacrine cell illustrated in Figure 4G, the voltage ramp-evoked response decreased slowly in the control condition and became relatively stable after ∼15 min. When we added a combination of TBOA (50 μm) and CPP (20 μm) to the bath solution, there was no change in the response (Fig. 4G). At ∼25 min, CPP was washed out and TBOA was applied alone. This led to a clear increase of the response that slowly decreased back to the baseline when TBOA was washed out (Fig. 4G). We tested a total of four AII amacrine cells in this condition (coapplication of TBOA and CPP followed by application of TBOA alone) and two cells with the opposite order of application (i.e., first TBOA alone, followed by coapplication of TBOA and CPP). The average current during coapplication of TBOA and CPP was 130.7 ± 7.8 pA, compared with an average current of 182 ± 22 pA when TBOA was applied alone (p = 0.0215, adjusted p value, one-way ANOVA, followed by Holm–Sidak's multiple-comparisons test; n = 6 cells; Fig. 4H).

The above results suggested that blocking glutamate transporters with TBOA can increase activation of NMDA receptors, both during baseline conditions of release and when release presumably is enhanced by increasing the general excitability after blocking voltage-gated K⁺ channels with TEA and 3,4-DAP in the bath solution. For some AII amacrine cells, we could detect activation of NMDA receptors under our baseline conditions not only by observing CPP-sensitive membrane noise (see above; Fig. 2A,C), but also by observing a tonic current that was revealed by depolarization. For the AII amacrine cell illustrated in Figure 4I, the current evoked by voltage ramps was suppressed when CPP (20 μm) was added to the bath solution. When CPP was washed out, the voltage ramp-evoked current reversed to the level measured before adding CPP (Fig. 4I). For AII amacrines tested in this way, the average current in the control condition was 102.1 ± 8.7 pA; and in the presence of CPP, it decreased to 71.4 ± 5.9 pA (p = 0.000015, paired t test; n = 14 cells; Fig. 4J). Together, these experiments directly demonstrate that increased ambient glutamate, evoked by blocking transporter activity, leads to an increase in NMDA receptor-mediated currents in AII amacrine cells.

Single-channel properties determined by stationary noise analysis of NMDA-evoked currents

We previously determined that the somas of AII and A17 amacrine cells lack detectable NMDA receptor responses (Zhou et al., 2016), limiting the possibilities for a more detailed investigation of the biophysical properties of these receptors by using outside-out or nucleated patches. Here, instead, we used stationary noise analysis of NMDA-evoked responses in whole-cell recordings of AII amacrine cells to elucidate single-channel properties of the NMDA receptor channels. Even for the relatively small AII amacrine cells, however, it is expected that channel gating noise generated in distal dendrites will be distorted by electrotonic filtering (Zandt et al., 2018); thus, no attempt was made to perform a similar analysis for A17 amacrine cells, which have much longer processes (Fig. 1E).

For these whole-cell recordings, we used a Cs⁺- and TEA-based intracellular solution to block potential activation of K⁺ channels activated by Ca²⁺ influx through the NMDA receptors, similar to what has been observed for olfactory bulb granule cells (Isaacson and Murphy, 2001). We first established a control condition by switching to a bath solution containing TTX, strychnine, bicuculline, and CNQX. Next, we changed to a Mg²⁺-free extracellular solution (containing the above blockers); and ∼5–10 min later, we added 50 μm NMDA and 200 μm d-serine to the bath solution. For the majority of AII amacrine cells tested (8 of 9 cells), bath application of NMDA evoked a slowly developing inward current. The inward current was accompanied by a marked increase of current noise (Fig. 5A). The noise increased in parallel with the increase of current amplitude as the concentration of NMDA in the recording chamber increased over time. For each epoch of continuous recording under semistationary conditions (see Materials and Methods), we calculated net NMDA-evoked noise spectra and fitted each averaged spectrum (typically between 1.95 and 500 Hz) with one (n = 6 cells) or two (n = 2 cells) Lorentzian components. Figure 5B shows the net spectrum and resulting curve fit for one such current epoch acquired after ∼9 min of NMDA application. The half-power frequency (f_c = 18 Hz) is indicated by a vertical arrow. The corresponding time constant (τ) was 8.9 ms. For the six AII amacrine cells that responded to NMDA and were well fit by a single Lorentzian component, the mean time constant for the later (“saturating”) responses was 8.03 ± 0.72 ms (range 5.5–9.6 ms). For the two cells where an adequate fit required two Lorentzian components, the mean time constants were as follows: τ₁ = 0.5 ms, τ₂ = 6.8 ms (cell 1) and τ₁ = 2.8 ms, τ₂ = 12.5 ms (cell 2).

• Download figure
• Open in new tab
• Download powerpoint

Figure 5.

Stationary noise analysis of NMDA-evoked current in an AII amacrine cell. A, Control recording and responses evoked by bath application of NMDA (50 μm). Traces 1–3 recorded at successive times after application of NMDA. Note the increasing membrane current noise during application of NMDA. B, Net power density spectrum (red markers) obtained from the current noise evoked by 50 μm NMDA during a 20 s period including trace 3 in A. The power spectrum for the (net) NMDA-evoked noise has been fitted (continuous curve) with a Lorentzian function and the half-power cutoff (corner) frequency (f_c) of the function is indicated by the vertical arrow. C, Net current variance plotted against mean current for a series of different periods (as in A) of NMDA-evoked responses in the same AII amacrine cell. Each data point corresponds to a specific time period after NMDA was added to the bath. Data points 1–3 correspond to the periods in A. The data points were obtained by power spectral density analysis (as in B) and were fitted with a parabolic function (Eq. 4). The apparent single-channel current (i) was 1.4 pA (corresponding to an apparent single-channel conductance of 22.6 pS) and the number of available receptor channels (N) was 60.

We estimated the net NMDA-evoked increase in current variance for each epoch as the integral under the sum of the Lorentzian function(s) fitted to each spectral density graph and plotted the resulting variance versus the mean current. When we obtained a series of epochs with slowly increasing mean current (as the concentration of NMDA in the recording chamber increased), the variance initially increased approximately linearly with mean current, consistent with a low open probability of the NMDA receptor channels (Fig. 5C). With increasing mean current, the slope of the variance versus mean curve eventually declined; and for the majority of cells, the variance reached a peak followed by a slight decline (Fig. 5C). The apparent unitary current (i) was estimated by fitting the data points with Equation 4 and the apparent unitary chord conductance (γ) was calculated as γ = i/(V_m − E_rev), where V_m is the voltage-clamp potential and E_rev is the reversal potential (assumed to be 0 mV) (compare Zhou et al., 2016). For the cell illustrated in Figure 5C, γ was 22.6 pS. For the AII amacrine cells where the variance versus mean data points spanned a range large enough to be well fitted with Equation 4 (n = 8 cells), γ was 22.6 ± 1.3 pS (range 17.1–27.1 pS). The curve fitting also estimated N (the number of available channels in each cell) to be 57.9 ± 7.0 (range 38–94), corresponding to a maximum P_open of 0.57 ± 0.04 (range 0.41–0.74). These numbers must be cautiously interpreted, however, as N is unlikely to equate the total number of NMDA receptors on an AII amacrine cell because many receptors will have desensitized. In addition, it is likely that the responses were influenced by electrotonic low-pass filtering and attenuation in the dendritic tree.

Pharmacological evidence for GluN2B-containing NMDA receptors in AII amacrine cells

The subunit composition of NMDA receptors confers specific and distinct functional properties, such as Mg²⁺ sensitivity, Ca²⁺ permeability, affinity for glutamate, single-channel conductance, kinetics, and pharmacology (Paoletti et al., 2013). We took advantage of the differences in pharmacological properties between NMDA receptors with different subunit composition (Ogden and Traynelis, 2011). Because the majority of NMDA receptors in the CNS are composed of either GluN2A or GluN2B subunits (in combination with GluN1 subunits), we started our investigation with pharmacological agents that can differentiate between receptors with these subunits.

We first examined the effect of ifenprodil (10 μm), an NMDA receptor antagonist with high selectivity for GluN2B-containing receptors (Williams, 1993; Perin-Dureau et al., 2002). For the AII cell illustrated in Figure 6A, ifenprodil reduced the amplitude of responses evoked by puffer-pipette application of NMDA (1 mm, 1 s) from ∼100 pA to ∼20 pA, a reduction of 80% relative to control. When ifenprodil was washed out, the response slowly recovered, but by the time the cell was lost after 20 min recording, the recovery had only reached an amplitude that was ∼50% of the initial control response (Fig. 6A). Slow and incomplete recovery from block by ifenprodil has been observed previously for other preparations (Williams, 1993; Kew et al., 1996). For AII amacrine cells tested in this way, the average NMDA-evoked response in the presence of ifenprodil (17.9 ± 5.7 pA) was signficantly lower than in the control condition (111 ± 33 pA; p = 0.029, paired t test), corresponding to a suppression of 84 ± 4% (n = 5 cells; Fig. 6A). Partial recovery after washout of ifenprodil was observed for 3 of 5 cells, with the response amplitude reaching an average of 63 ± 19% of the baseline response.

• Download figure
• Open in new tab
• Download powerpoint

Figure 6.

NMDA-evoked responses in AII, but not A17, amacrines are blocked by the GluN2B-selective antagonists ifenprodil and Ro 25–6981. A, Left, Responses evoked by NMDA application in an AII amacrine cell by pressure application of NMDA (1 mm, 1 s) from a puffer pipette in the control condition (1) and in the presence of 10 μm ifenprodil in the bath (2). In this and subsequent figures, the duration of drug application is indicated by the horizontal bar above the current traces. Middle, Peak amplitude of currents activated in the same AII amacrine cell by application of NMDA (as to the left) at 60 s intervals. Note the suppression of NMDA-evoked responses during addition of ifenprodil. Here and in Figures 7 and 8, numbered data points (1 and 2) correspond to responses of data traces displayed at left. Right, Peak amplitude of NMDA-evoked responses in AII amacrine cells (n = 5 cells) in control and in the presence of ifenprodil. B, Left, Responses evoked by NMDA application in an A17 amacrine cell by pressure application of NMDA (1 mm, 1 s) from a puffer pipette in the control condition (1) and in the presence of 10 μm ifenprodil in the bath (2). Middle, Peak amplitude of currents activated in the same A17 amacrine cell by application of NMDA (as to the left) at 60 s intervals. Note the lack of effect on NMDA-evoked responses during addition of ifenprodil. Right, Peak amplitude of NMDA-evoked responses in A17 amacrine cells (n = 8 cells) in control and in the presence of ifenprodil. C, Left, Responses evoked by NMDA application in an AII amacrine cell by pressure application of NMDA (1 mm, 1 s) from a puffer pipette in the control condition (1) and in the presence of 1 μm Ro 25–6981 (2) in the bath. Middle, Peak amplitude of currents activated in the same AII amacrine cell by application of NMDA (as to the left) at 60 s intervals. Note the suppression of NMDA-evoked responses during addition of Ro 25–6981. Right, Peak amplitude of NMDA-evoked responses in AII amacrine cells (n = 6 cells) in control and in the presence of Ro 25–6981. D, Left, Responses evoked by NMDA application in an A17 amacrine cell by pressure application of NMDA (1 mm, 1 s) from a puffer pipette in the control condition (1) and in the presence of 1 μm Ro 25–6981 (2). Middle, Peak amplitude of currents activated in the same A17 amacrine cell by application of NMDA (as to the left) at 60 s intervals. Note the lack of effect on NMDA-evoked responses during addition of Ro 25–6981. Right, Peak amplitude of NMDA-evoked responses in A17 amacrine cells (n = 6 cells) in control and in the presence of Ro 25–6981. Statistical comparisons between averages: n.s., No significant difference (p > 0.05). *p < 0.05.

When A17 amacrine cells were tested in the same way, ifenprodil had no effect on the NMDA-evoked responses, as illustrated for the cell in Figure 6B where the response amplitude was ∼80 pA in the control condition and ∼82 pA in the presence of ifenprodil. For A17 amacrine cells, the average NMDA-evoked response was 82 ± 10 pA in the control condition and 80 ± 11 in the presence of ifenprodil (p = 0.2725, paired t test), corresponding to a change of only 3 ± 2% (n = 8 cells; Fig. 6B). These results suggested that AII amacrine cells express NMDA receptors that contain the GluN2B subunit, whereas A17 amacrines do not.

To substantiate this conclusion, we also examined the effect of Ro 25–6981 (1 μm), a more potent derivative of ifenprodil that is highly specific for NMDA receptors containing the GluN2B subunit (Fischer et al., 1997). For the AII amacrine cell illustrated in Figure 6C, application of Ro 25–6981 reduced the NMDA-evoked response from ∼110 pA to ∼3 pA, corresponding to an amplitude reduction of 90%. The cell was lost before Ro 25–6981 could be washed out, but partial recovery of NMDA-evoked responses was observed for AII cells in which recordings could be maintained for at least 15 min after we started to wash out Ro 25–6981. For AII amacrine cells tested with Ro 25–6981, the average NMDA-evoked response in the presence of Ro 25–6981 (4 ± 1 pA) was significantly lower than in the control condition (68 ± 14 pA) corresponding to a suppression of 93 ± 2% (n = 6 cells; p = 0.0053, paired t test; Fig. 6C). For A17 amacrine cells, application of Ro 25–6981 did not change the NMDA-evoked currents. For the cell illustrated in Figure 6D, the response was ∼103 pA in the control condition and ∼105 pA in the presence of Ro 25–6981. For A17 amacrines, the average control response was 102 ± 15 pA and the average response in Ro 25–6981 was 110 ± 17 pA corresponding to a change of only 6 ± 6% (n = 6 cells; p = 0.25, paired t test; Fig. 6D). Together, these results provided strong evidence for the presence of GluN2B subunits in the NMDA receptors of AII amacrine cells and suggested that GluN2B is absent from the NMDA receptors of A17 amacrine cells.

Pharmacological evidence for GluN2A-containing NMDA receptors in A17 amacrine cells

To test for NMDA receptors containing the GluN2A subunit, we examined the effect of Zn²⁺, which acts as a specific, voltage-independent, noncompetitive, high-affinity (IC₅₀ = 20 nm) antagonist of GluN2A-containing NMDA receptors. The maximum degree of inhibition by Zn²⁺ that can be expected is ∼60%-80% (Paoletti et al., 2009). However, under normal experimental conditions, nanomolar concentrations of Zn²⁺ can be found as a contamination in different chemical compounds and are sufficient to block GluN2A-containing NMDA receptors, potentially obscuring their presence (Paoletti et al., 2009). In addition, Zn²⁺ can be colocalized with glutamate in synaptic vesicles and is coreleased with glutamate during exocytosis (Smart et al., 1994). For simplicity, we collectively refer to these sources as contaminating Zn²⁺. To work around this complication, we used tricine to chelate contaminating Zn²⁺ and buffer Zn²⁺ at a controlled level (Paoletti et al., 2009). To investigate the potential relevance of contaminating levels of Zn²⁺ under our conditions, we recorded from both AII and A17 amacrine cells and acquired baseline NMDA-evoked responses (as above). After a stable baseline had been established, we added tricine at a concentration of 10 mm (Paoletti et al., 2009) to chelate any contaminating Zn²⁺, which may have suppressed the baseline NMDA-evoked responses.

For AII amacrine cells, the NMDA responses evoked in the presence of tricine (10 mm), had an average peak amplitude of 74 ± 13 pA, not significantly different from the responses obtained in the control condition without tricine (79 ± 14 pA; n = 9 cells; p = 0.1177, paired t test; Fig. 7A). However, for A17 amacrine cells, the average peak amplitude of NMDA responses evoked in the presence of tricine was 92 ± 11 pA, significantly larger than the responses obtained in the control condition without tricine (66.5 ± 7.2 pA; n = 13 cells; p = 0.0004, paired t test; Fig. 7B), and corresponded to an increase of 39 ± 6%. These effects of tricine suggested that the NMDA receptors on A17, but not AII, amacrine cells were suppressed by endogenous or contaminating Zn²⁺ under our recording conditions.

• Download figure
• Open in new tab
• Download powerpoint

Figure 7.

NMDA-evoked responses in A17, but not AII amacrines, are increased by chelating potentially contaminating trace amounts of Zn²⁺ with tricine in the extracellular solution. NMDA-evoked responses in A17, but not AII amacrines, are blocked by the GluN2A-selective antagonist Zn²⁺. A, Peak amplitude of currents activated in AII amacrine cells (n = 9 cells) by repeated pressure application of NMDA (1 mm, 1 s) from a puffer pipette at 60 s intervals, first in control and then in the presence of tricine (10 mm). Note the lack of effect on NMDA responses after adding tricine. B, Peak amplitude of currents activated in A17 amacrine cells (n = 13 cells) by repeated pressure application of NMDA (1 mm, 1 s) from a puffer pipette at 60 s intervals, first in control and then in the presence of tricine (10 mm). Note the increase of responses after adding tricine compared with control. C, Left, Responses evoked by NMDA application in an AII amacrine cell by pressure application of NMDA (1 mm, 1 s) from a puffer pipette in the control condition (after adding 10 mm tricine; 1) and in the presence of tricine and ZnCl₂ (to achieve a free concentration of Zn²⁺ ([Zn²⁺]_free) of 100 nm; 2). Middle, Peak amplitude of currents activated in the same AII amacrine cell by application of NMDA (as to the left) at 60 s intervals. After a brief period, the extracellular solution was supplemented with tricine to chelate potentially contaminating trace amounts of Zn²⁺ before adding ZnCl₂. Note the lack of effect on NMDA-evoked responses by Zn²⁺ ([Zn²⁺]_free = 100 nm). Right, Peak amplitude of NMDA-evoked responses in AII amacrine cells (n = 4 cells) in tricine and with tricine and [Zn²⁺]_free = 100 nm (“Zinc”). D, Left, Responses evoked by NMDA application in an A17 amacrine cell by pressure application of NMDA (1 mm, 1 s) from a puffer pipette in the control condition (after adding tricine; 1) and in the presence of tricine and ZnCl₂ ([Zn²⁺]_free = 100 nm; 2). Middle, Peak amplitude of currents activated in the same A17 amacrine cell by application of NMDA (as to the left) at 60 s intervals. After a brief period, the extracellular solution was supplemented with tricine (10 mm) to chelate potentially contaminating trace amounts of Zn²⁺ before adding ZnCl₂. Note the enhancement of responses after adding tricine and suppression of responses after addition of ZnCl₂. Right, Peak amplitude of NMDA-evoked responses in A17 amacrine cells (n = 6 cells) in tricine (10 mm) and with tricine (10 mm) and [Zn²⁺]_free = 100 nm (“Zinc”). E, Left, Responses evoked by NMDA application in an AII amacrine cell by pressure application of NMDA (1 mm, 1 s) from a puffer pipette in the control condition (in the presence of 10 mm tricine; 1) and in the presence of tricine and 25 nm PEAQX (2). Middle, Peak amplitude of currents activated in the same AII amacrine cell by application of NMDA (as to the left) at 60 s intervals. After a brief period, the extracellular solution was supplemented with tricine (10 mm) to chelate potentially contaminating trace amounts of Zn²⁺ to alleviate any block of GluN2A-containing NMDA receptors. Note the lack of effect on NMDA-evoked responses by PEAQX. Right, Peak amplitude of NMDA-evoked responses in AII amacrine cells (n = 5 cells) in tricine and in the presence of tricine and PEAQX (“PEAQX”). F, Left, Responses evoked by NMDA application in an A17 amacrine cell by pressure application of NMDA (1 mm, 1 s) from a puffer pipette in the control condition (in the presence of 10 mm tricine; 1) and in the presence of tricine and PEAQX (2). Middle, Peak amplitude of currents activated in the same A17 amacrine cell by application of NMDA (as to the left) at 60 s intervals. After a brief period, the extracellular solution was supplemented with tricine (10 mm) to chelate potentially contaminating trace amounts of Zn²⁺ to alleviate any block of GluN2A-containing NMDA receptors. Note the suppression of NMDA-evoked responses by PEAQX. Right, Peak amplitude of NMDA-evoked responses in A17 amacrine cells (n = 5 cells) in tricine and in the presence of tricine and PEAQX (“PEAQX”). Statistical comparisons between averages: n.s., No significant difference (p > 0.05). *p < 0.05.

To test potential effects of Zn²⁺ directly, we recorded from both cell types with tricine present in the control condition and then changed to an extracellular solution with a free Zn²⁺ concentration ([Zn²⁺]_free) of 100 nm (see Materials and Methods). For AII amacrine cells, the average NMDA-evoked response in the presence of 100 nm free Zn²⁺ was 45.3 ± 9.4 pA, not significantly different from that in the control with tricine (44.8 ± 9.4 pA, n = 4 cells; p = 0.8763, paired t test; Fig. 7C). These results suggested that the NMDA receptors expressed by AII amacrine cells are not inhibited by nanomolar concentrations of Zn²⁺, and thus are unlikely to contain the GluN2A subunit.

When A17 amacrine cells were tested in the same way, we found that 100 nm Zn²⁺ (in the extracellular solution with 10 mm tricine) markedly suppressed the NMDA-evoked responses of these cells. For the cell illustrated in Figure 7D, Zn²⁺ reduced the response to NMDA to ∼30 pA, corresponding to a decrease of 50% relative to the control with tricine. For A17 amacrine cells tested in this way, the average NMDA-evoked response in the presence of 100 nm Zn²⁺ was 36.1 ± 7.1 pA, significantly lower than in the control condition with tricine (71.9 ± 9.1 pA, n = 6 cells; p = 0.0002, paired t test; Fig. 7D), and corresponded to a reduction of 51 ± 5%. These results suggested that the NMDA receptors expressed by A17 amacrine cells are inhibited by nanomolar concentrations of Zn²⁺, and thus are likely to contain the GluN2A subunit.

To substantiate these results, we examined the effect of PEAQX, an antagonist that is selective for GluN2A-containing NMDA receptors (Frizelle et al., 2006). We tested the effect of PEAQX (25 nm) in the presence of tricine to eliminate the possibility that potential GluN2A-containing NMDA receptors would already be blocked by contaminating levels of Zn²⁺. In this condition, PEAQX had no effect on NMDA-evoked responses in AII amacrine cells (Fig. 7E). The average response in the control with tricine was 98 ± 16 pA, and the average response in the presence of PEAQX was 94 ± 15 pA (n = 5 cells; p = 0.0879, paired t test; Fig. 7E).

When A17 amacrine cells were tested with PEAQX (in the presence of tricine), the results were very different from those obtained for AII amacrine cells. For the cell illustrated in Figure 7F, PEAQX suppressed the NMDA-evoked response from ∼60 pA in the control with tricine to ∼36 pA after 10 min in PEAQX, corresponding to a reduction of 40%. In the presence of PEAQX, the average NMDA-evoked response in A17 amacrines was 55 ± 12 pA, significantly lower than in the tricine condition (84 ± 14 pA; n = 5 cells; p = 0.00127, paired t test; Fig. 7F), corresponding to an average suppression of 36 ± 4%. Together, these results provided strong evidence for the presence of the GluN2A subunit in NMDA receptors of A17 amacrine cells and suggested that GluN2A is absent from the NMDA receptors of AII amacrine cells.

No evidence for expression of GluN2C and GluN2D receptor subunits in AII and A17 amacrine cells

The functional properties of NMDA receptors containing GluN2C and/or GluN2D subunits differ substantially from those containing GluN2A and/or GluN2B subunits (Paoletti et al., 2013). The receptor antagonists QNZ46 (Hansen and Traynelis, 2011) and DQP-1105 (Acker et al., 2011) are sufficiently selective for GluN2C- and GluN2D-containing NMDA receptors that they can be used to differentiate them from GluN2A- and GluN2B-containing receptors. Unfortunately, there are currently no NMDA receptor antagonists available that can adequately distinguish between GluN2C- and GluN2D-containing receptors. However, there is evidence from in situ hybridization studies in adult rat retina that GluN2C, but not GluN2D, is expressed by subsets of amacrine cells (in addition to GluN2A and GluN2B) (Brandstätter et al., 1994).

There was no effect of QNZ46 (10 μm) on NMDA-evoked responses in either AII or A17 amacrine cells (Fig. 8A,B). For AII amacrines, the average response amplitude was 74 ± 11 pA in control and 70 ± 11 pA in QNZ46 (p = 0.1295, paired t test; n = 8 cells; Fig. 8A). For A17 amacrines, the average was 79 ± 12 pA in control and 74 ± 11 pA in QNZ46 (p = 0.0742, paired t test; n = 5 cells; Fig. 8B). These results suggested that neither AII nor A17 amacrine cells express GluN2C or GluN2D subunits. The results with QNZ46 were supported by those obtained with DQP-1105 (10 μm), which also had no effect on the NMDA-evoked responses in either AII or A17 amacrine cells. For AII amacrines, the average was 87 ± 17 pA in control and 83 ± 16 pA in DQP-1105 (p = 0.4561, paired t test; n = 8 cells; data not shown). For A17 amacrines, the average was 77 ± 17 pA in control and 75 ± 17 in DQP-1105 (p = 0.6218, paired t test; n = 7 cells; data not shown). Together, the results with GluN2C- and GluN2D-selective antagonists strongly suggested the absence of these subunits from the NMDA receptors expressed by AII and A17 amacrine cells.

• Download figure
• Open in new tab
• Download powerpoint

Figure 8.

No evidence for expression of GluN2C/GluN2D subunits or triheteromeric GluN1/GluN2A/GluN2B NMDA receptors in AII or A17 amacrine cells. A, Left, Responses evoked by NMDA application in an AII amacrine cell by pressure application of NMDA (1 mm, 1 s) from a puffer pipette in the control condition and in the presence of 10 μm QNZ46 (2). Middle, Peak amplitude of currents activated in the same AII amacrine cell by application of NMDA (as to the left) at 60 s intervals. Note the lack of effect on NMDA-evoked responses during addition of QNZ46 (10 μm) to the bath. Right, Peak amplitude of NMDA-evoked responses in AII amacrine cells (n = 8 cells) in control and in the presence of QNZ46. B, Left, Responses evoked by NMDA application in an A17 amacrine cell by pressure application of NMDA (1 mm, 1 s) from a puffer pipette in the control condition and in the presence of 10 μm QNZ46 (2). Middle, Peak amplitude of currents activated in the same A17 amacrine cell by application of NMDA (as to the left) at 60 s intervals. Note the lack of effect on NMDA-evoked responses during addition of QNZ46 (10 μm) to the bath. Right, Peak amplitude of NMDA-evoked responses in A17 amacrine cells (n = 5 cells) in control and in the presence of QNZ46. C, Left, Peak amplitude of currents activated in an AII amacrine cell by pressure application of NMDA (1 mm, 1 s, 60 s intervals). During the first recording period, tricine (10 mm) was added to the bath to chelate potentially contaminating trace amounts of Zn²⁺ to avoid baseline suppression of GluN2A-containing receptors. Next, a relatively low concentration of ifenprodil (3 μm) was added to block diheteromeric GluN1/GluN2B receptors. Finally, the free Zn²⁺ concentration was set to 100 nm (by adding ZnCl₂) to test for the presence of potential triheteromeric GluN1/GluN2A/GluN2B receptors that would presumably be blocked by the combination of ifenprodil and Zn²⁺. There was no additional suppression by adding Zn²⁺. Right, Peak amplitude of NMDA-evoked responses in AII amacrine cells (n = 3 cells) in the presence of tricine and ifenprodil (3 μm; “Ifenprodil”), and in the presence of tricine, ifenprodil, and Zn²⁺ (100 nm; “Ifenprodil + Zinc”). D, Left, Peak amplitude of currents activated in an A17 amacrine cell by pressure application of NMDA (1 mm, 1 s, 60 s intervals). During the first recording period, tricine (10 mm) was added to the bath to chelate potentially contaminating trace amounts of Zn²⁺ to avoid baseline suppression of GluN2A-containing receptors. Next, the free Zn²⁺ concentration was set to 100 nm (by adding ZnCl₂) to block diheteromeric GluN1/GluN2A receptors. Finally, ifenprodil was added (3 μm) to test for block of triheteromeric GluN1/GluN2A/GluN2B receptors by the combination of Zn²⁺ and ifenprodil. There was no additional suppression by adding ifenprodil. Right, Peak amplitude of NMDA-evoked responses in A17 amacrine cells (n = 5 cells) in the presence of tricine and Zn²⁺ (100 nm; “Zinc”), and in the presence of tricine, Zn²⁺, and ifenprodil (3 μm; “Zinc + Ifenprodil”). Statistical comparisons between averages: n.s., No significant difference (p > 0.05).

No evidence for expression of triheteromeric NMDA receptors in AII and A17 amacrines

The majority of NMDA receptors in the CNS contain either two GluN2A or two GluN2B subunits in combination with two obligatory GluN1 subunits (Hansen et al., 2014). However, there is also the potential for NMDA receptors to form triheteromeric receptors with one GluN2A and one GluN2B subunit in addition to two GluN1 subunits, as demonstrated in the hippocampus and forebrain (for review, see Hansen et al., 2014). The functional and pharmacological properties of triheteromeric NMDA receptors are a mix between the two diheteromeric receptor subtypes (Hatton and Paoletti, 2005; Tovar et al., 2013). Specifically, Zn²⁺ has been demonstrated to be equally effective as an antagonist of diheteromeric GluN2A-containing receptors and triheteromeric GluN2A/GluN2B-containing receptors (Hatton and Paoletti, 2005). Furthermore, in recombinant expression systems, ifenprodil (at the low concentration of 1–3 μm) blocks triheteromeric GluN2A/GluN2B-containing receptors, but only by ∼20% (Hatton and Paoletti, 2005; Hansen et al., 2014). On this background, our results described above suggest that it is unlikely that AII amacrine cells have a significant population of functional triheteromeric receptors, as neither tricine nor nanomolar concentrations of Zn²⁺ had any effect on the NMDA-evoked currents. Similarly, it is unlikely that A17 amacrine cells have a significant population of functional triheteromeric receptors, as ifenprodil had no effect on the NMDA-evoked currents in these cells.

However, if there is only a small population of triheteromeric NMDA receptors on either cell type, responses mediated by these receptors might be masked by the pharmacological properties of the predominant diheteromeric receptors. Importantly, the inhibition of triheteromeric receptors by ifenprodil and Zn²⁺ is superadditive (Hatton and Paoletti, 2005), meaning that the inhibition observed when one antagonist is applied in a background of the other, is larger than if the same antagonist is applied alone. Thus, we decided to use a combination of the antagonists ifenprodil and Zn²⁺ to directly test for the presence of triheteromeric receptors among the AII and A17 NMDA receptors. For AII amacrines, we tested the effect of Zn²⁺ in a background of ifenprodil; and for A17 amacrines, we tested the effect of ifenprodil in a background of Zn²⁺.

For both AII and A17 amacrine cells, we applied NMDA (1 mm, 1 s) in an extracellular solution that contained tricine (10 mm) to ensure that any potential triheteromeric GluN2A/GluN2B receptors were not already blocked by contaminating Zn²⁺ (Fig. 8C,D). For the AII amacrines, we then switched to an extracellular solution with a low concentration of ifenprodil (3 μm) to block diheteromeric GluN2B-containing receptors without completely blocking potential triheteromeric GluN2A/GluN2B-containing receptors (Hatton and Paoletti, 2005; Hansen et al., 2014). Similar to the higher concentration of ifenprodil (10 μm; see above), 3 μm ifenprodil had a strong inhibitory effect on the NMDA-evoked responses in AII amacrines (Fig. 8C). After obtaining a stable baseline, we switched to an extracellular solution with a free Zn²⁺ concentration of 100 nm, but the addition of Zn²⁺ had no additional inhibitory effect (Fig. 8C). For AII amacrines tested in this way, the average NMDA-evoked current in the presence of ifenprodil was 20.1 ± 3.5 pA. With addition of 100 nm free Zn²⁺ in a background of ifenprodil, the NMDA response did not change, with an average of 22.2 ± 4.8 pA (p = 0.5493, paired t test; n = 3 cells; Fig. 8C). Thus, it is unlikely that AII amacrine cells express a significant population of functional triheteromeric GluN2A/GluN2B-containing receptors.

For the A17 amacrines, switching to an extracellular solution with 100 nm free Zn²⁺ strongly inhibited the NMDA-evoked responses (Fig. 8D). After obtaining a stable baseline, we switched to an extracellular solution with a low concentration of ifenprodil (3 μm), but the addition of ifenprodil had no additional inhibitory effect (Fig. 8D). For A17 amacrines tested in this way, the average NMDA-evoked current in the presence of 100 nm free Zn²⁺ (and tricine) was 38.5 ± 8.6 pA. With the addition of ifenprodil in a background of 100 nm free Zn²⁺, the NMDA response did not change, with an average of 39.4 ± 9.7 pA (p = 0.6855, paired t test; n = 5 cells; Fig. 8D). Thus, as for AII amacrines, it is unlikely that A17 amacrines express a significant population of functional triheteromeric GluN2A/GluN2B-containing receptors.

Morphological evidence for expression of GluN2B in AII amacrine cells and GluN2A in A17 amacrine cells

Our pharmacological experiments strongly suggested a remarkable complementary expression of NMDA receptors in the postsynaptic targets of rod bipolar cells, with GluN2B-containing receptors in AII amacrine cells and GluN2A-containing receptors in A17 amacrine cells. To examine the relationship between the dendritic processes of these amacrine cells and NMDA receptor subunit expression, we injected AII and A17 amacrine cells in retinal slices with Alexa-488, immunolabeled the slices for either GluN2B or GluN2A, and acquired confocal image stacks.

We observed clear colocalization of GluN2B with AII, but not A17 amacrine cells (Fig. 9). Figure 9A shows a maximum intensity projection (MIP) of a confocal stack from a slice with an injected AII amacrine cell and immunolabeled for GluN2B. The labeling for GluN2B appears punctate across the inner plexiform layer, with higher density of punctae in the proximal part, corresponding approximately to sublamina b. It is commonly assumed that when synaptic (in contrast to extrasynaptic) receptors are immunolabeled with fluorescent antibodies, the labeling will appear punctate, with each punctum corresponding to a cluster of receptors in a postsynaptic density. In the case of NMDA receptors, however, this cannot be assumed to be correct, as there is evidence that extrasynaptic NMDA receptors can be organized in a clustered fashion (Petralia et al., 2010; Papouin and Oliet, 2014). Visual inspection of single slices (in the XY plane) of the confocal stack, with the two channels (AII and GluN2B) overlaid, identified multiple examples of GluN2B punctae that overlapped with processes of the AII amacrine cell (Fig. 9B,C). In most cases, colocalization between AII processes and GluN2B punctae was detected at the arboreal dendrites (Fig. 9C), but we could also find a few examples of GluN2B colocalized with lobular dendrites (Fig. 9B, left). For the majority of lobular processes, however, no colocalization could be observed despite the frequent presence of immunolabeled punctae in close proximity to these processes (Fig. 9B, right). Similar results were seen for a total of five AII amacrine cells in slices immunolabeled for GluN2B.

• Download figure
• Open in new tab
• Download powerpoint

Figure 9.

Colocalization between AII (but not A17) amacrine cells and immunolabeled punctae of the NMDA receptor subunit GluN2B. A, MIP of confocal image stack of retinal slice with an AII amacrine cell injected with Alexa-488 (green) and immunolabeled for GluN2B (magenta). Scale bar, 10 μm. A–C, Images from same AII amacrine. B, Left, Example of colocalization (arrowhead) of GluN2B punctae (magenta) and AII lobular dendrites (green) overlaid in a single XY confocal slice. Right, For the large majority of AII lobular dendrites, no colocalization with GluN2B punctae could be observed, despite close proximity. Here and later, each image was taken from a single XY confocal slice. Scale bar, 1 μm. C, Examples of colocalization (arrowheads) of GluN2B punctae (magenta) and AII arboreal dendrites (green) overlaid in single XY confocal slices. Scale bars, 1 μm. D, MIP of confocal image stack of retinal slice with an A17 amacrine cell injected with Alexa-488 (green) and immunolabeled for GluN2B (magenta). Scale bar, 10 μm. D, E, Images from same A17 amacrine. E, No colocalization of GluN2B punctae (magenta) and A17 lobular dendrites (green) overlaid in single XY confocal slices. Scale bar, 1 μm. Here and later, all images and graphs were generated from digital data after deconvolution (for details, see Materials and Methods).

For the larger dendritic trees of A17 amacrine cells, we were only able to acquire confocal image stacks, at the required resolution, of smaller portions of the total extent of these cells. Figure 9D shows an MIP of two adjacent confocal stacks with an A17 in a retinal slice immunolabeled for GluN2B. The confocal stacks were stitched together as a montage with the region of overlap running horizontally along the inner plexiform layer. In contrast to AII amacrine cells, we were not able to find any unequivocal examples of GluN2B punctae that overlapped with processes of the A17 amacrine cell when we inspected XY slices of the confocal stack (Fig. 9E). Similar results were seen for a total of three A17 amacrine cells in slices immunolabeled for GluN2B.

In parallel experiments, slices with injected AII amacrine cells were immunolabeled for the GluN2A subunit. Figure 10A shows an MIP of a confocal stack with an AII amacrine cell and immunolabeling for GluN2A. Similar to the labeling for GluN2B, the labeling for GluN2A is also punctate and is distributed across the inner plexiform layer (Hartveit et al., 1994). In contrast to the labeling for GluN2B, we were not able to find any unequivocal examples of GluN2A punctae that overlapped with processes of the AII amacrine cell when we inspected single slices (XY) of the confocal stack, both for lobular dendrites (Fig. 10B) and arboreal dendrites (Fig. 10C). While we sometimes identified apparent overlap in the XY plane, inspection of the associated XZ and YZ planes indicated that the CM for the GluN2A punctum and the filled AII process were further apart and that the appearance of overlap was due to the lower resolution along the z axis. Similar results were seen for a total of four AII amacrine cells in slices immunolabeled for GluN2A.

• Download figure
• Open in new tab
• Download powerpoint

Figure 10.

Colocalization between A17 (but not AII) amacrine cells and immunolabeled punctae of the NMDA receptor subunit GluN2A. A, MIP of confocal image stack of retinal slice with an AII amacrine cell injected with Alexa-488 (green) and immunolabeled for GluN2A (magenta). Scale bar, 10 μm. A–C, Images from same AII amacrine. B, No colocalization of GluN2A punctae (magenta) and AII lobular dendrites (green) overlaid in single XY confocal slices. Scale bars, 1 μm. C, No colocalization of GluN2A punctae (magenta) and AII arboreal dendrites (green) overlaid in single XY confocal slices. Scale bars, 1 μm. D, MIP of confocal image stack of retinal slice with a section of processes from an A17 amacrine cell injected with Alexa-488 (green) and immunolabeled for GluN2A (magenta). Scale bar, 3 μm. E, Examples of colocalization (arrowheads) of GluN2A punctae (magenta) and A17 dendrites (green) overlaid in single XY confocal slices. Scale bars, 1 μm. D, E, Images from same A17 amacrine, except for E (bottom right).

For A17 amacrine cells, however, we observed clear colocalization with GluN2A punctae. Figure 10D shows an MIP of a confocal stack from a portion of an injected A17 amacrine cell in a retinal slice immunolabeled for GluN2A. Visual inspection of single slices (XY) of the confocal stack, with the two channels (A17 and GluN2A) overlaid, identified several examples of GluN2A punctae that overlapped with processes of the A17 amacrine cell (Fig. 10E). There were cases of GluN2A punctae that overlapped at putative varicosities of the A17 (Fig. 10E, left panels), as well as GluN2A punctae that overlapped with intervaricosity segments (Fig. 10E, right panels). Similar results were seen for a total of three A17 amacrine cells in slices immunolabeled for GluN2A.

GluN2B is colocalized with Cx36 on AII amacrine cells

Gap junction coupling between AII amacrine cells is mediated by Cx36 (Feigenspan et al., 2001; Mills et al., 2001; Deans et al., 2002). There is evidence that NMDA receptors are involved in modulatory regulation of tracer coupling between AII amacrine cells, potentially via phosphorylation of Cx36 and a consequent increase in the junctional conductance (Kothmann et al., 2012). This predicts a close spatial association between Cx36-containing gap junction plaques and NMDA receptors with GluN2B. To investigate this, we performed double immunolabeling with antibodies against GluN2B and Cx36. Figure 11A shows an MIP of a confocal stack with an injected AII amacrine cell in a retinal slice immunolabeled for both GluN2B and Cx36. Consistent with previous reports (Mills et al., 2001), immunolabeled Cx36 punctae are distributed across the inner plexiform layer, with a higher density in sublamina b than in sublamina a (Fig. 11A). Visual inspection of single confocal slices (XY) identified several examples of complete or partial overlap between punctae corresponding to labeling for GluN2B and Cx36 that also overlap with processes of the AII amacrine (Fig. 11B–E).

• Download figure
• Open in new tab
• Download powerpoint

Figure 11.

Colocalization between AII amacrine cells, immunolabeled punctae of the NMDA receptor subunit GluN2B, and immunolabeled punctae of Cx36. A, MIP of confocal image stack of retinal slice with an AII amacrine cell injected with Alexa-488 (green) and immunolabeled for GluN2B (magenta) and Cx36 (turqoise). Scale bar, 5 μm. A–E, From the same cell. B–E, Left, Examples of colocalization (arrowheads) of GluN2B punctae, Cx36 punctae, and AII amacrine arboreal processes. Middle, GluN2B punctae displayed alone. Right, Cx36 punctae displayed alone. Scale bar: B–E, 1 μm.

For a quantitative analysis, we first applied simple threshold segmentation to identify structures in the AII, GluN2B, and Cx36 channels. Following segmentation, we isolated all segmented GluN2B objects that overlapped (≥50%) with structures belonging to the segmented AII amacrine cell. For the AII amacrine cell illustrated in Figure 11A, a total of 16 GluN2B objects overlapping with AII processes were detected (14 at the arboreals and two at the lobulars). We used the 3D (i.e., XYZ) CM coordinates of the GluN2B punctae overlapping with the AII amacrine cell to estimate the average signal in the Cx36 channel. The presence of a GluN2B punctum, overlapping with the AII amacrine, resulted in a localized signal increase above background in the Cx36 channel (Fig. 12A,C). This strongly suggested that the presence of a GluN2B cluster associated with an AII amacrine process predicts the nearby presence of a Cx36 cluster corresponding to a gap junction. We repeated the analysis for Cx36 punctae overlapping the AII amacrine cell. We used the 3D (i.e., XYZ) CM coordinates of the Cx36 punctae overlapping (≥50%) with the AII amacrine cell to estimate the average signal in the GluN2B channel. The presence of a Cx36 punctum, overlapping with the AII amacrine, resulted in a localized signal increase above background in the GluN2B channel (Fig. 12B,C). Similar results were obtained for three other AII amacrine cells in slices double immunolabeled for GluN2B and Cx36. These results suggest that the GluN2B-containing NMDA receptors expressed by AII amacrine cells could be involved in modulating the Cx36-mediated coupling of these cells.

• Download figure
• Open in new tab
• Download powerpoint

Figure 12.

Colocalization analysis of GluN2B- and Cx36-immunolabeled punctae overlapping with an AII amacrine cell. A, Average signal in Cx36 channel around 3D coordinates (XYZ; CM) of all thresholded GluN2B punctae overlapping ≥50% with AII amacrine cell, displayed as 2D image (in XY plane) through center of extracted volume of average signal. A, B, Data are from the same cell as in Figure 11. 2D images represent the same intensity range. B, Average signal in GluN2B channel around 3D coordinates (XYZ; CM) of all thresholded Cx36 punctae overlapping ≥50% with AII amacrine cell, displayed as 2D image (in XY plane) through center of extracted volume of average signal. C, Overlaid linear intensity profiles (measured in arbitrary units, a.u.) across center of 2D image in A for average Cx36 signal around GluN2B coordinates (black line) and across center of 2D image in B for average GluN2B signal around Cx36 coordinates (red line).

Activation of NMDA receptors evokes an increase of intracellular Ca²⁺ in both AII and A17 amacrine cells

The large relative Ca²⁺ permeability is a hallmark of the NMDA receptor (MacDermott et al., 1986; Mayer and Westbrook, 1987; Mayer et al., 1987). To directly investigate the putative influx of Ca²⁺ mediated by activation of NMDA receptors on AII and A17 amacrine cells, we loaded cells with the Ca²⁺ indicator OGB-1 (200 μm), in addition to the Ca²⁺-insensitive dye Alexa-594 (40 μm) for structural imaging, via diffusion from the pipette solution during whole-cell voltage-clamp recording. A total of six AII amacrine cells that were examined with Ca²⁺ imaging displayed stable values for ΔG/R during the baseline period. All cells displayed an electrophysiological response to application of NMDA and a clear NMDA-evoked increase in Ca²⁺. An example of an AII amacrine cell with a strong increase of Ca²⁺ is shown in Figure 13A–C. Figure 13A shows an MIP from a complete stack of the cell acquired by structural imaging. For the focal plane used for Ca²⁺ imaging, a number of ROIs were selected manually and covered locations along both arboreal and lobular dendrites (including arboreal varicosities and lobular appendages), as well as the apical dendrite (Fig. 13B). For each ROI, we calculated ΔG/R as a function of time. Figure 13C shows the average ΔG/R time series for ROIs at arboreal dendrites (magenta; compare Fig. 13B) and at lobular/apical dendrites (green; compare Fig. 13B). For this cell, the peak value of the average ΔG/R time series was markedly larger for arboreal ROIs (105 ± 13%; n = 13 ROIs) than for lobular/apical ROIs (28.3 ± 5.6%; n = 7 ROIs; unpaired t test, p = 7.2 × 10⁻⁵) (Fig. 13C). For all AII amacrine cells tested in this way, the corresponding values were 63 ± 14% (range 11%–105%) for arboreal processes (n = 6 cells) and 26.3 ± 4.8% (range 12.5%–34.7%) for lobular/apical processes (n = 4 cells; the orientation in the slice compromised simultaneous imaging of arboreal and lobular dendrites for two of the six cells). For two of the four AII cells where we obtained simultaneous measurements from ROIs along both arboreal and lobular dendrites, the NMDA-evoked Ca²⁺ increase (peak value of ΔG/R) was significantly higher for the arboreal than for the lobular ROIs (p = 9.7 × 10⁻⁴ for the cell not illustrated in Fig. 13). For the other two cells, there was no significant difference between the Ca²⁺ responses in the lobular and arboreal dendrites.

• Download figure
• Open in new tab
• Download powerpoint

Figure 13.

NMDA receptor activation increases intracellular Ca²⁺ in AII and A17 amacrine cells. A, AII amacrine cell filled with Alexa-594 via patch pipette. MIP (along z axis) generated from MPE fluorescence image stack. Scale bar, 5 μm. B, Same cell as in A, image from single focal plane in a smaller subregion of A, generated by averaging all images of Alexa-594 fluorescence acquired during time period displayed in C. Magenta and green circles indicate ROIs in arboreal (ROI1-ROI13) and lobular (ROI14-ROI20) dendrites, respectively, used for calculating changes in intracellular Ca²⁺. Scale bar, 5 μm. C, Ca²⁺ signals in arboreal and lobular dendrites (averaged over corresponding ROIs in B) evoked by bath application of NMDA (100 μm; marked by the continuous horizontal line above data traces). Here and in F, Ca²⁺ signal calculated as ΔG/R, i.e., the change in OGB-1 (green; G) fluorescence normalized by the change in Alexa-594 (red; R) fluorescence. Each trace represents the average for all corresponding arboreal (magenta) and lobular (green) ROIs in B. Shaded area represents ±95% CI. Here and in F, time 0 corresponds to breaking into the cell and establishing the whole-cell recording configuration. D, A17 amacrine cell filled with Alexa-594 via patch pipette. MIP (along z axis) generated from MPE fluorescence image stack. Scale bar, 10 μm. Dashed rectangle represents the region used for Ca²⁺ imaging (in E). E, Same cell as in D, image from single focal plane, generated by averaging all images of Alexa-594 fluorescence acquired during time period displayed in F. Circles represent ROIs in dendritic tree, used for calculating changes in intracellular Ca²⁺. Scale bar, 5 μm. F, Ca²⁺ signal in A17 dendrites (averaged over corresponding ROIs in E) evoked by bath application of NMDA (100 μm; marked by the continuous horizontal line above data trace). Trace represents the average for all ROIs in E. Shaded area represents ±95% CI.

Similar results with NMDA-evoked increases of intracellular Ca²⁺ were obtained for three A17 amacrine cells tested in the same way. For the cell illustrated in Figure 13D-F, the peak increase of Ca²⁺ (measured as the peak ΔG/R value of the average time series for 13 ROIs) was 94%. For the other two A17 cells, the corresponding values were 37% and 120%. Because of the relative sparsity of processes in the single focal planes used for repeated Ca²⁺ imaging, it was difficult to obtain simultaneous measurements for a large number of ROIs, but we did not observe any obvious differences between the response magnitudes of ROIs located at varicosities and intervaricosity segments.

NMDA receptor activation increases GABA release from A17 amacrine cells

The increase in intracellular Ca²⁺ following addition of NMDA suggest that the extrasynaptic NMDA receptors could serve multiple functions on both AII and A17 amacrine cells. Whereas there is evidence that both AII and A17 amacrines are homologously coupled via gap junctions (AII: Strettoi et al., 1992; Veruki and Hartveit, 2002a; A17: Xin and Bloomfield, 1997; Grimes et al., 2014), it is currently unknown whether the coupling strength between A17 amacrines is physiologically regulated, potentially via a mechanism involving NMDA receptors, as for AII amacrines (Kothmann et al., 2012). Instead, for A17 amacrine cells, we focused on a potential role of NMDA receptors in the control of GABA release in the feedback circuit from A17 amacrines to rod bipolar cells. We recorded from rod bipolar cells in whole-cell voltage clamp with strychnine (to block glycine receptors) and TTX (to block release from GABAergic amacrine cells with nonreciprocal input to rod bipolar cells) (Chávez et al., 2010) added to the bath solution. Any responses evoked in rod bipolar cells following application of NMDA cannot be a direct effect, as we have previously demonstrated that rod bipolar cells do not express NMDA receptors (Zhou et al., 2016).

The rod bipolar cell illustrated in Figure 14A displayed a low frequency of spontaneous IPSCs in the control condition (∼0.1 Hz). The amplitude of the average IPSC in the control condition was 5.6 pA (n = 9 events recorded during 90 s; Fig. 14B). When we added NMDA (50 μm) to the bath solution, the frequency of IPSCs markedly increased to ∼2.8 Hz (Fig. 14A). The amplitude of the average IPSC in the presence of NMDA was 5.4 pA (n = 252 events recorded during 90 s; Fig. 14B), very similar to that in the control condition. For a total of five rod bipolar cells, the average IPSC frequency increased from 0.19 ± 0.08 Hz in the control condition to 2.14 ± 0.51 Hz during application of NMDA (p = 0.015, paired t test; Fig. 14C). There was no statistically significant difference in the amplitude of the average IPSC between the control condition (5.0 ± 0.3 pA) and during application of NMDA (5.4 ± 0.5 pA) (p = 0.10, paired t test; n = 4 cells; Fig. 14D). One cell did not display IPSCs in the control condition and was excluded from the amplitude analysis. When we repeated the application of NMDA in the presence of bicuculline (10 μm) and TPMPA (50 μm) to block GABA_A and GABA_C receptors, respectively (in addition to strychnine and TTX), there were no IPSCs in the control condition and no IPSCs were evoked after adding NMDA (n = 3 cells; Fig. 14E). These results suggested that the IPSCs were mediated by ionotropic GABA_A and/or GABA_C receptors and that NMDA receptors on A17 amacrine cells can contribute to release of GABA from these cells.

• Download figure
• Open in new tab
• Download powerpoint

Figure 14.

Activation of NMDA receptors on presumed A17 amacrine cells increases GABA-ergic input to rod bipolar cells. A, Continuous current recorded from a rod bipolar cell (V_hold = −60 mV) in control condition (top) and during bath application of NMDA (50 μm; bottom). The equilibrium potential for chloride (E_Cl) was ∼0 mV, such that responses mediated by activation of ionotropic GABA receptors appear as inward currents when V_hold = −60 mV. In both conditions (here and in E), the bath solution contained strychnine (300 nm; to block glycine receptors) and TTX (300 nm; to block voltage-gated Na⁺ channels) and the pipette solution contained TBOA (50 μm; to block glutamate transporter activity). NMDA was added with 200 μm d-serine, to saturate the coagonist binding site, and 0 Mg²⁺, to alleviate the voltage-dependent block. Note the increased frequency of IPSCs in the presence of NMDA. B, Averaged waveforms of IPSCs in the control condition (black trace; n = 9 events detected during 90 s recording) and in the presence of NMDA (red trace; n = 252 events detected during 90 s recording). C, Frequency of IPSCs in rod bipolar cells (n = 5 cells) in the control condition (left bar) and in the presence of NMDA (50 μm; right bar). D, Amplitude of average IPSCs in rod bipolar cells (same cells as in C) in the control condition (left bar) and in the presence of NMDA (50 μm; right bar). E, Continuous current recorded from a rod bipolar cell in control condition (top; as in A) and during bath application of NMDA (50 μm; bottom; as in A). In both conditions, the bath solution also contained bicuculline (10 μm; to block GABA_A receptors) and TPMPA (50 μm; to block GABA_C receptors). Note the absence of IPSCs both in control and during application of NMDA. Statistical comparisons between averages: n.s., No significant difference (p > 0.05). *p < 0.05.