|Home | About | Journals | Submit | Contact Us | Français|
In the few days prior to eye-opening in mice, the excitatory drive underlying waves switches from cholinergic to glutamatergic. Here, we describe the unique synaptic and spatiotemporal properties of waves generated by the retina’s glutamatergic circuits. First, knockout mice lacking vesicular glutamate transporter type 1 do not have glutamatergic waves, but continue to exhibit cholinergic waves, demonstrating for the first time that the two wave generating circuits are linked. Second, simultaneous outside-out patch and whole-cell recordings reveal that retinal waves are accompanied by transient increases in extrasynaptic glutamate, directly demonstrating the existence of glutamate spillover during waves. Third, the initiation rate and propagation speed of retinal waves, as assayed by calcium imaging, are sensitive to pharmacological manipulations of spillover and inhibition, demonstrating a role for both signaling pathways in shaping the spatiotemporal properties of glutamatergic retinal waves.
Spontaneous activity in developing circuits plays a multitude of important roles. For example, in the developing visual system spontaneous activity in the retina is critical for neuronal survival (Goldberg et al., 2002), neuronal path finding (Nicol et al., 2007), and in the refinement and maintenance of visual maps (for reviews, see Huberman et al., 2008; Torborg and Feller, 2005; Wong, 1999). Indeed, similar functions have been assigned to spontaneous activity in the cochlea (Tritsch et al., 2007), spinal cord (Gonzalez-Islas and Wenner, 2006; Hanson et al., 2008), hippocampus and cortex (Leinekugel, 2003; Mohajerani and Cherubini, 2006), indicating that spontaneous activity may represent a general strategy used by the developing nervous system to assemble neural circuits (Feller, 1999).
We study this phenomenon in the developing retina, which exhibits propagating bursts of action potentials termed retinal waves. Retinal waves persist in a period of development during which the retina is rapidly maturing (Morgan and Wong, 2006; Sernagor et al., 2001). These changes in retinal circuitry are reflected by changes in the mechanisms that mediate waves, which have been divided into three stages (for review, see Torborg and Feller, 2005). Stage I retinal waves (embryonic day 16 – postnatal day 0 [P0] in mice) are blocked by gap junction blockers but are only weakly affected by antagonists to fast neurotransmitter receptors. Stage II waves (P0-P10) are blocked by nicotinic acetylcholine receptor (nAChR) antagonists. Stage III waves (P10-P14) are blocked by ionotropic glutamate receptor antagonists and not nAChR antagonists. The transition from Stage II to Stage III waves occurs when bipolar cell axons begin to make synapses within the inner plexiform layer (Fisher, 1979; Morgan et al., 2008) and light responses are first detected (Demas et al., 2003; Tian and Copenhagen, 2003).
The mechanisms underlying Stage II waves and their role in visual system development have been well characterized (Huberman et al., 2008; Torborg and Feller, 2005). However, there is little understanding of Stage III waves. Recently, it has been proposed that glutamate acting outside of the synaptic cleft provides a source of depolarization before mature synaptic structures are established (Demarque et al., 2002). There is also evidence that glutamate spillover, in which glutamate spills out of the synaptic cleft and activates extrasynaptic glutamate receptors, plays a role in circuit development. Spillover in hippocampus has been detected as early as P14, with a developmental reduction in glutamate spillover correlated with an increase in the expression of glutamate transporters (Diamond, 2005), and may be critical for activation of “silent synapses” (Balland et al., 2008). In developing neocortex (Demarque et al., 2004; Milh et al., 2007), hippocampus (Cattani et al., 2007) and brainstem (Sharifullina and Nistri, 2006) blockade of glutamate transporters profoundly alters spontaneous firing patterns, indicating that regulation of spillover has implications for network function. Furthermore, ambient levels of extracellular glutamate regulate spontaneous activity in perinatal neocortex (Allene et al., 2008).
Here, we describe several experiments that provide key insights into the mechanisms underlying glutamatergic Stage III waves. First, we use a knockout mouse lacking glutamate release from bipolar cells to demonstrate that the maturation of glutamate signaling is required for the cessation of cholinergic Stage II waves. Second, we use glutamate receptor-containing outside-out patches to establish that Stage III waves induce spontaneous increases in glutamate spillover. Third, we use calcium imaging to determine how manipulating glutamate spillover and inhibition alters the spatiotemporal properties of waves.
Throughout this study we use synaptic inputs onto RGCs to monitor retinal waves. Therefore it is critical to establish the correspondence between synaptic inputs and network activation. We performed simultaneous current and voltage clamp recordings from neighboring RGCs while monitoring spontaneous calcium increases in the region around these cells (Fig. 1A,B). Because bath-loading of calcium indicators is significantly less effective in mouse retinas older than P4 (Bansal et al., 2000) we used the multicell bolus loading method to label cells in the RGC layer of P10-P13 mouse retinas with Oregon Green BAPTA-1-AM (Ohki et al., 2005; Stosiek et al., 2003). Simultaneous calcium imaging, whole-cell voltage clamp (Vhold=−60 mV) and current clamp recordings revealed that correlated increases in calcium indicative of retinal waves were concurrent with RGCs firing bursts of action potentials and receiving excitatory input (Fig. 1B,C). Retinal waves were always accompanied by compound excitatory synaptic currents (cEPSCs) in RGCs (n=6 cells in 6 retinas). Hence, Stage III waves represent waves of depolarization in RGCs that are driven by synaptic input, and cEPSCs reliably indicate the presence of retinal waves.
What inputs do RGCs receive during retinal waves? Simultaneously monitoring excitatory currents in one RGC (Vhold = −60 mV) and inhibitory currents in another (Vhold = 0 mV) revealed that RGCs receive both excitatory and inhibitory synaptic inputs during retinal waves (Fig. 1D). Wave-associated cEPSCs were blocked by a combination of the AMPA/Kainate receptor antagonist DNQX (20–50 μM) and the NMDA receptor antagonist AP5 (50 μM, n=6 pairs; Fig. 1D), suggesting that during retinal waves RGCs receive direct excitatory inputs from bipolar cells, which are the primary glutamatergic input to RGCs (Bansal et al., 2000; Demas et al., 2006; Syed et al., 2004; Wong et al., 2000). Wave-associated inhibitory synaptic currents (cIPSCs) were also blocked by DNQX and AP5, indicating that RGCs receive feedforward inhibition from amacrine cells during retinal waves.
The GABAA antagonist gabazine (5 μM) and the glycine antagonist strychnine (4 μM) together (n=11), but neither alone (n=6 cells in gabazine, n=5 cells in strychnine), blocked all cIPSCs recorded from RGCs, establishing that RGCs receive both GABAergic and glycinergic inputs during waves (Fig. 1E).
In the recordings of the excitatory synaptic inputs RGCs receive during retinal waves, we observed that some cEPSCs and cIPSCs occurred in clusters (Fig. 1F), which is visible in a histogram of inter-cEPSC intervals (Fig. 1G).
Stage III waves are blocked by ionotropic glutamate receptor antagonists, but the source of glutamate is not known. In the adult retina, bipolar cells are the primary source of glutamatergic inputs to RGCs, but there is a class of amacrine cells which express vesicular glutamate transporter 3 (VGLUT3) on their synaptic vesicles and form synapses onto RGCs (Haverkamp and Wassle, 2004; Johnson et al., 2004). Bipolar cells express only VGLUT1 (Fremeau et al., 2004; Wassle et al., 2006). (Photoreceptors also express VGLUT1, however they are not thought to contribute to Stage III waves since our experiments are carried out in bright light, which hyperpolarizes photoreceptors.) To determine which cell supplies the glutamate required for Stage III waves, we examined mice lacking either VGLUT1 or VGLUT3.
We monitored retinal waves by recording wave-associated synaptic currents from pairs of RGCs in P10 – P12 VGLUT1 −/− mice and in age-matched littermate VGLUT1 +/− controls (Fig. 2A,C). In VGLUT1 +/− retinas, as in wild type retinas, wave-associated synaptic currents were blocked by a combination of either DNQX (20 μM) or NBQX (20 μM) and AP5 (50 μM, n=10 cells in 5 mice), whereas wave-associated synaptic currents were not blocked by DHβE, a nicotinic acetylcholine receptor antagonist (8 μM, n=11 cells in 5 mice; Fig. 2A,B). Conversely, wave-associated synaptic currents in VGLUT1 −/− retinas were not blocked by DNQX or NBQX and AP5 (n=13 cells in 4 mice), but were instead blocked by DHβE (n=12 cells in 4 mice; Fig. 2C,D), demonstrating that VGLUT1−/− mice do not have Stage III glutamatergic waves but rather have maintained Stage II cholinergic waves. These results indicate that 1) Stage III waves require glutamate release from bipolar cells or photoreceptors and that 2) glutamatergic signaling is necessary for the cessation of Stage II waves.
In contrast to VGLUT1 −/− RGCs, all compound PSCs were blocked by DNQX and AP5 in both VGLUT3 +/− (n=8 cells in 4 mice) and VGLUT3 −/− RGCs (n=6 cells in 4 mice; data not shown). Hence, we conclude that glutamate release from VGLUT3-positive amacrine cells is not required for Stage III retinal waves and that bipolar cells are the source of glutamatergic currents recorded in RGCs during Stage III waves (Fig. 1).
Spontaneous release of glutamate from bipolar cells is necessary for the generation of Stage III waves. However, at this age glutamatergic synaptic structures (Morgan et al., 2008; Sherry et al., 2003) and transmission (Tian and Copenhagen, 2001) are not yet mature. We detected very few spontaneous synaptic events in between retinal waves (Fig. 1D), indicating that there are fewer glutamate synapses than detected in later stages of development (Tian and Copenhagen, 2001). We found that cEPSCs associated with retinal waves were extremely slow (Fig. 1D), suggestive of spillover-mediated currents (Bergles et al., 1999).
Is there glutamate spillover during retinal waves? To directly detect extrasynaptic glutamate, we used glutamate receptor-containing outside-out patches (Allen, 1997; Copenhagen and Jahr, 1989; Maeda et al., 1995), a technique that has also been used to detect extrasynaptic ACh (Hume et al., 1983; Young and Poo, 1983) and GABA (Isaacson et al., 1993). Patches of membrane excised from RGC somas in the outside-out patch configuration respond to pressure-applied glutamate, indicating that they contain sufficient numbers of glutamate receptors to detect extrasynaptic glutamate (see Experimental Procedures). To test whether there is glutamate spillover during Stage III retinal waves, we monitored extracellular glutamate in the inner plexiform layer with outside-out patches while monitoring cEPSCs in RGCs (Fig. 3). Most wave-associated cEPSCs were accompanied by glutamate currents in outside-out patches (Fig. 3; currents were present in 6/6 patches; 65±26% of cEPSCs were accompanied by patch currents). Wave-associated patch currents reversed at 0 mV, consistent with activation of glutamate receptors (Fig. 3A). These data indicate that retinal waves are accompanied by episodic increases in the amount of glutamate spillover.
There is a significant amount of GABA and glycine release during retinal waves (Figure 1). To ensure that GABA and glycine currents (Isaacson et al., 1993) were not contaminating our recordings we repeated these experiments while blocking GABAA and glycine receptors using solutions containing 5μM gabazine and 4μM strychnine. In addition, to increase the current through NMDA receptors, we also performed these experiments in 0 Mg2+. There was no significant increase in the percent of wave-associated cEPSCs accompanied by measurable spillover currents under these pharmacological conditions (79±6%, n=3 patches, p=.42, unpaired t test vs. control; Fig. 3B), suggesting that GABA and glycine currents do not contribute significantly to wave-associated currents in outside-out patches.
We have demonstrated for the first time that Stage III waves are accompanied by robust glutamate spillover (Fig. 3). Our next goal was to assay the role of glutamate spillover in circuit function. To do so we performed a detailed characterization of Stage III wave properties by visualizing intracellular calcium concentration in many cells over several hundred microns simultaneously. This method has been used extensively to determine the spatiotemporal properties of Stage II waves but has been limited in its application to Stage III waves due to poor loading of calcium indicators at older ages (Bansal et al., 2000; Feller et al., 1997; Sernagor et al., 2000; Wong et al., 1995; Zhou and Zhao, 2000). In retinas loaded with the multicell bolus loading technique (Stosiek et al., 2003) (Fig. 1), Stage III waves induced a large change in fluorescence and therefore the time evolution and the spatial boundaries of waves were clearly distinguishable (Fig. 4A; Supplemental Movie 1).
First, we assayed wave initiation rates. Stage III waves were similar to Stage II waves in that they initiated at several locations across the imaged region, and over time the entire ganglion cell layer participated in waves. However, Stage III waves were different from Stage II waves in that they often occurred in episodes during which a cluster of 2–5 waves initiated rapidly in roughly the same location of the retina (Fig. 4B,C). These episodes were separated by much longer intervals, after which a distinct wave or cluster of waves initiated in a different region of the retina (Fig. 4B–D).
Second, we assayed wave propagation speed (see Supplemental Experimental Procedures and Supplemental Fig. 1), which is reflective of the coupling that underlies the sequential recruitment of bipolar cells into retinal waves. Stage III waves propagate at a mean velocity of 198 μm/s with a very large variability across retinas (standard deviation=68 μm/s, n=341 waves recorded in 34 P10-P13 retinas, Fig. 4E). This high variability indicates that the coupling mechanisms that underlie wave propagation are likely to be easily modulated by environmental factors. These velocities are comparable to those previously reported for mice (Muir-Robinson et al., 2002), rabbit (Syed et al., 2004) and ferret (Feller et al., 1997; Wong et al., 1993), and significantly lower than turtle (Sernagor and Grzywacz, 1999; Sernagor et al., 2003) and previous reports of Stage III waves in mice based on multielectrode array (MEA) recordings (Demas et al., 2003). The different measurements of Stage III wave velocities in mice obtained using calcium imaging and the multielectrode array are due to the methods used to compute velocity (see Supplemental Experimental Procedures).
Traditionally, glutamate spillover is studied at the level of intercellular communication between pairs of cells. Our goal was to assay the role of glutamate spillover in the function of an entire circuit. To do so we assayed the effects of different manipulations of glutamate spillover on the spatiotemporal properties of Stage III waves.
First, we monitored the effects of enhancing glutamate spillover by blocking excitatory amino acid transporters (EAATs), which remove glutamate from the extracellular space. To characterize the effects of the non-transportable blocker of EAATs TBOA (25 μM) on glutamatergic synaptic input to RGCS and glutamate spillover we used whole-cell voltage clamp and outside-out patch recordings. The presence of TBOA increased the amount of glutamatergic activity recorded in RGCs (Fig. 5A,B), and in TBOA 73±12% of wave-associated cEPSCs were accompanied by measurable glutamate spillover, (a significant increase over control patches; 55±6% in control, n=4 patches, p=.04).
The presence of TBOA significantly altered the spatiotemporal properties of retinal waves as assayed by calcium imaging. Waves initiated much more frequently with very few quiescent periods between waves (Fig. 5C,D and Supplemental Movie 2). The presence of TBOA dramatically shifted the distribution of inter-wave intervals such that the vast majority of intervals were less than 10 seconds (Fig. 5E).
A second dramatic effect we observed was that TBOA reduced the variability of wave propagation speed. In contrast to the wide range of propagation speeds recorded in control retinas (Fig. 4E), waves in TBOA propagated within a significantly narrower range of velocities (mean±SD: 145±14 μm/s in TBOA vs. 173±79 μm/s in control, n = 8 retinas; t test, p=0.33; Levene’s test of variance, p=0.001; Fig. 5F). Interestingly, the effect of TBOA depended on control wave propagation speed. For retinas with waves slower than 140 microns per second TBOA significantly increased wave propagation speed (n=3, p<0.02 for each retina), for retinas with waves faster than 200 microns/sec TBOA significantly decreased wave propagation speed (n=4, p<0.02 for each retina), and for one retina with waves propagating near 150 μm/s TBOA had no effect (n=1, p=.16). Indeed, the magnitude of the TBOA-induced change in wave propagation speed was strongly correlated with the control propagation speed (linear regression r2=0.985; Fig. 5G). Furthermore, in a subset of retinas that exhibited waves of widely varying propagation speeds, TBOA reduced the variability across waves within a retina (Supplemental Fig. 2). The dramatic reduction in variability of propagation speed caused by TBOA leads to the hypothesis that variations in ambient levels of glutamate may control the spatial and temporal structure of correlations induced by retinal waves.
Are the TBOA-induced changes in wave properties due to effects of spillover on the Stage III wave-generating circuitry, or does TBOA induce waves generated by a novel mechanisms as observed in developing neocortex, hippocampus and brainstem (Cattani et al., 2007; Demarque et al., 2004; Milh et al., 2007; Sharifullina and Nistri, 2006)? If the retina contains a circuit that supports waves independent of the Stage III wave-generating circuitry, TBOA may induce waves in retinas not capable of generating Stage III waves. To test this, we performed calcium imaging while applying TBOA to retinas during Stage II waves and to adult retinas. TBOA did not induce waves in P5-P6 retinas; on the contrary, it made Stage II waves less frequent (IWI, mean±SD = 107±36 s in control, 226±49 s in TBOA, n=5 retinas, p = .01; Supplemental Fig. 3), and did not induce waves in adult retinas (n=8 P22-P24 retinas). Hence, enhancing glutamate spillover enhanced wave activity only during Stage III. Furthermore, like Stage III waves in the absence of TBOA (Fig. 6; Wong et al., 2000) but unlike TBOA-induced activity in other brain areas, wave-associated cEPSCs in TBOA were not blocked by AP5 alone (n=6/7 cells in 4 retinas) but were blocked by a combination of AP5 and DNQX (n=5/5 cells in 3 retinas). Though these data do not directly implicate glutamate spillover in wave propagation, they are consistent with the model that glutamate signaling via spillover is a component of the Stage III wave generating circuit.
As a second manipulation of glutamate spillover, we took advantage of glutamate transporters’ high sensitivity to temperature (Wadiche and Kavanaugh, 1998) by altering the temperature of the ACSF. We found a significant decrease in inter-wave intervals and increase in wave propagation speed for elevated temperatures (Supplemental Fig. 4). However, there was a significant increase in propagation speed and a shift in the IWI distribution as a function of temperature in the presence of TBOA, similar to the temperature effects observed in control solutions. Hence we cannot distinguish whether the effects of elevated temperatures on the spatiotemporal properties of waves were due to glutamate spillover or other temperature-dependent physiological processes.
Unfortunately, there is no unambiguous way to block glutamate spillover without also inhibiting conventional synaptic transmission. However, in many parts of the brain, glutamate spillover exerts its effects by preferentially activating NMDA receptors, which have a high affinity for glutamate and are therefore more likely to be activated by the lower concentrations of glutamate outside of the synaptic cleft (Chen and Diamond, 2002; Higgs and Lukasiewicz, 1999). In RGCs, over 90% of NMDA receptors are localized extrasynaptically while AMPA receptors, conversely, are expressed almost exclusively in synapses (Zhang and Diamond, 2006). Therefore, one way to test the role of glutamate spillover in the retinal wave-generating circuit is to block AMPA/Kainate receptors.
Bath application of NBQX, a selective AMPA/kainate receptor antagonist, did not block Stage III waves as assayed with calcium imaging (n=5, Fig. 6A,B). This demonstrates that activation of NMDA receptors provides sufficient depolarization to support retinal waves. This is consistent with previous results from calcium imaging in ferrets (Wong et al., 2000). However, these results differ from a report previous from our lab indicating that CNQX or DNQX are sufficient to block waves (Bansal et al., 2000). This difference could be due to poorer sensitivity in previous calcium imaging experiments or potential non-specificity of CNQX and DNQX.
Though waves persist in AMPA/kainate receptor antagonists, we observed a reduction in their frequency (Fig. 6C). AMPA/kainate receptor antagonists had no effect on wave propagation speed (mean±SD, 229±37μm/s in NBQX vs. 198±17 μm/sec in control, n=5, p=0.22; Fig. 6D), indicating that glutamate signaling via AMPA/kainate receptors determines the initiation rate but not the propagation speed of Stage III waves.
Next, we assayed the role of NMDA receptor activation in determining the spatiotemporal properties of Stage III waves. Similar to blockade of AMPA/kainate receptors and consistent with previous results we found that blockade of NMDA receptors by bath application of AP5 (50 μM) did not block waves or change the speed of wave propagation (205±16μm/s in AP5 vs. 193±35 in control, n=5, p=0.47, Fig. 6E,F,H), although the initiation rate was slightly shifted to higher inter-wave intervals (Fig. 6G).
These results suggest that both synaptic and extrasynaptic glutamate receptors contribute to the depolarizations during Stage III waves. However, they do not distinguish which aspects of wave propagation are mediated by synaptic vs. extrasynaptic glutamate receptor activation. First, the subcellular distribution of AMPA and NMDA receptors in the retina at this age is not known, and it may differ from adult as there is robust reorganization of glutamatergic synapses during development (Morgan et al., 2008). Second, in cerebellum glutamate spillover has been shown to activate AMPA receptors in neighboring synapses (DiGregorio et al., 2002), so an alternative interpretation is that glutamate spillover-mediated activation of both NMDA and AMPA receptors is required for wave propagation. Furthermore, during circuit development the distinction between synaptic and extrasynaptic signaling may represent a false dichotomy given the dramatic rearrangements of glutamate receptors that occur at this age (Morgan et al., 2008).
Elevating glutamate spillover increased the initiation rate of retinal waves. Previously, it was established that blocking GABAA and glycine receptors also increased the frequency of Stage III waves (Fischer et al., 1998; Kerschensteiner and Wong, 2008; Syed et al., 2004; Zhou, 2001). Application of gabazine (5 μM) and strychnine (4μM) increased the frequency of cEPSCs recorded in RGCs (Fig. 1E), indicating that GABA and glycine normally suppress glutamate release from bipolar cells. These observations lead to the hypothesis that the initiation rate of retinal waves is determined by the balance of excitation and inhibition at the level of bipolar cells. To test this hypothesis, we compared the effects of TBOA and the blockade of inhibition on the spatiotemporal properties of waves.
Waves initiated much more frequently in GABAA and glycine receptor antagonists (5 μM gabazine and 4 μM strychnine; Fig. 7A–C). As in TBOA, this rapid initiation of waves eliminated the bimodal distribution of inter-wave intervals observed in control conditions, and replaced it with a smooth distribution in which the inter-wave interval for the vast majority of waves is less than 10 seconds.
Blockade of GABAA and glycine receptors slightly reduced wavefront propagation speed (mean±SD 172±18 μm/s in GZ+STR, 199±32 μm/s in control, n=8 retinas, p=.05; Fig. 7D). Furthermore, although the change in propagation speed caused by gabazine and strychnine on a per-retina basis was correlated with the initial propagation speed (linear regression r2 = .7; Fig. 7E), gabazine and strychnine did not reduce the variability in mean wave propagation speed among retinas (Levene test, p=0.1). However, for these experiments variability in propagation speed in control retinas was lower than for the TBOA experiments and therefore direct comparisons cannot be made.
This supports the hypothesis that the depolarization state of bipolar cells, set by spillover and inhibition, is a critical determinant of initiation rate. A schematic of these functional circuits consistent with the pharmacology presented here is provided in Figure 8 and discussed below.
We have explored the role of synaptic and extrasynaptic signaling in the generation of Stage III retinal waves. First, knockout mice lacking VGLUT1 did not have glutamatergic retinal waves but continued to exhibit cholinergic waves, demonstrating that glutamatergic signaling causes the disappearance of Stage II waves. Second, we provided direct evidence that glutamate spillover is a robust feature of Stage III retinal waves, demonstrating for the first time spontaneous episodic elevations in extrasynaptic glutamate in the developing retina. Third, elevating levels of glutamate spillover significantly reduced the variance in wave propagation speed, indicating that fluctuating endogenous levels of glutamate influence the propagation properties of retinal waves. In addition, wave initiation rates are reduced by blockade of either NMDA or AMPA receptors, indicating that initiation rates are influenced bidirectionally by levels of glutamate signaling. Last, blocking inhibition had similar effects on the spatiotemporal properties of waves as elevating glutamate spillover, supporting a model in which spontaneous initiation of Stage III waves is dictated by the balance of excitation and inhibition onto bipolar cells.
What role do glutamate waves play in development? Our observation that VGLUT1 −/− retinas do not have normal Stage III waves demonstrates that synaptic transmission from photoreceptors or bipolar cells is necessary for Stage III waves. Interestingly, while VGLUT1 +/−littermates have Stage III waves, VGLUT1 −/− mice have Stage II-like waves. This suggests that in wild type retinas the onset of glutamatergic signaling may turn off the Stage II wave-generating circuitry.
A similar phenomenon has been seen in a mouse lacking choline acetyl transferase (ChAT), the enzyme that synthesizes acetylcholine, in a portion of the retina (Stacy et al., 2005). During the period of development normally occupied by the cholinergic Stage II waves, ChAT-knockout retina exhibits spontaneous activity that is blocked by gap junction blockers but not by nAChR or ionotropic glutamate receptor antagonists (Stacy et al., 2005), indicating a persistence of mechanisms that mediate Stage I waves (for review, see Torborg and Feller, 2005). These findings suggest that the retina turns on a different form of spontaneous activity when one form is blocked, revealing powerful homeostatic mechanisms that ensure spontaneous activity is always present.
Glutamate spillover has been described in the adult retina (Chen and Diamond, 2002; DeVries et al., 2006; Veruki et al., 2006). A major goal of this study was to determine whether glutamate spillover exists at a stage in development when functional glutamatergic synapses are first forming. We found that retinal waves are accompanied by robust glutamate spillover by using glutamate receptor-containing outside-out patches (Fig. 3). It is possible that the presence of a recording pipette in the inner plexiform layer locally disrupts glutamate clearance, contributing to the spillover currents we see. However, glutamate spillover between pairs of bipolar cells has been measured in the adult retina, consistent with our conclusion that spillover occurs in the IPL (Veruki et al., 2006). This finding may have implications for spontaneous activity in other developing circuits that exhibit glutamate receptor-dependent spontaneous activity while glutamatergic synapses are still being formed, as in cortex (Dupont et al., 2006; Minlebaev et al., 2008) and hippocampus (Ben-Ari et al., 2004; Kasiyanov et al., 2008).
We demonstrated a role for glutamate spillover in shaping the spatiotemporal properties of retinal waves in two ways. First, altering the level of glutamate spillover altered the rate of wave initiations, i.e. the rate at which retinas entered the active state. Reducing signaling via ionotropic glutamate receptors by blocking either AMPA/kainate or NMDA receptors reduced the frequency of wave initiations (Fig. 6), while increasing glutamate spillover by blocking glutamate transporters dramatically increased the frequency of wave initiations (Fig. 5). A similar increase in wave initiations was observed when inhibition was blocked (Fig. 7 and Fischer et al., 1998; Kerschensteiner and Wong, 2008; Syed et al., 2004; Zhou, 2001).
Second, altering the level of glutamate spillover affected wavefront propagation speed. Wavefront propagation is a measure of the rate at which bipolar cells are recruited into waves. In control conditions, Stage III waves propagate with a highly variable wavefront propagation speed. Though wavefront propagation varied across waves within a retina (Supplemental Fig. 2), it varied most significantly across preparations, with a few retinas supporting very rapid waves with propagation speeds greater than 250 μm/s (Fig. 4D, Supplemental Fig. 1). Surprisingly, TBOA reduced the variability in propagation speed by causing all waves to propagate at speeds near 150 μm/s (Fig. 5, Supplemental Fig. 2).
Based on this observation, we propose that glutamate activates two distinct processes: one that positively regulates wave speed, such as activation of ionotropic glutamate receptors, and one that negatively regulates it, such as activation of presynaptic mGluRs. In this scenario variation in propagation speed from retina to retina may have to do with varying ambient levels of glutamate, perhaps due to variations in the effectiveness of transporters. In TBOA, there is maximum activation of both positive and negative regulators of wave speed and therefore there is less variation across retinas.
We propose a model that is consistent with these experimental observations (Figure 8). Stage III waves are initiated by spontaneous depolarizations in bipolar cells, and coupled to surrounding bipolar cells via spillover. In this way, bipolar cells function in a manner similar to starburst amacrine cells during the cholinergic stage II waves. The strong depolarization provided by a wave also excites inhibitory circuits, which contribute to the delay between waves. Hence, the balance of excitation and inhibition onto bipolar cells sets the initiation rate. The question remains – how does spillover couple bipolar cells when its function in the adult retina is to inhibit release from neighboring bipolar cells? More experiments are required to determine whether glutamate release from one bipolar cell directly depolarizes its neighbor during development via conductances that are not described in adult or whether there is an intermediate excitatory cell.
In summary, we have demonstrated that an absence of glutamate signaling prevents the transition from Stage II to Stage III waves. In addition, there is robust inhibitory signaling and glutamate spillover during Stage III waves, both of which shape the spatiotemporal properties of waves. These findings represent key insights into the features of transient circuits that mediate spontaneous activity during development.
C57Bl/6 mice obtained from Harlan were used for all WT recordings. VGLUT1 knockout mice (Fremeau et al., 2004) and VLGUT3 knockout mice (Seal et al., 2008) were generated as reported previously. Heterozygous littermates were used as controls for experiments using VGLUT1- and VGLUT3-null mice. All animal procedures were approved by the University of California, Berkeley; University of California, San Diego; or University of California, San Francisco Institutional Animal Care and Use Committees and conformed to the National Institutes of Health Guide for the Care and Use of Laboratory Animals, the Public Health Service Policy, and the Society for Neuroscience Policy on the Use of Animals in Neuroscience Research.
P10 – P24 mice were anesthetized with halothane or isoflurane and decapitated. Retinas were isolated in artificial cerebrospinal fluid (ACSF) (in mM: 119.0 NaCl, 26.2 NaHCO3, 11 glucose, 2.5 KCl, 1.0 K2HPO4, 2.5 CaCl2, 1.3 MgCl2) and mounted RGC side up on filter paper. Retinas were incubated at room temperature in bubbled (95% O2/5% CO2) ACSF until transfer to the recording chamber, where they were constantly superfused with bubbled ACSF (32–34°C unless otherwise specified).
Whole-cell current and voltage clamp recordings were made from whole-mount retinas. Retinas were visualized with differential interference contrast optics on a Zeiss Axioskop 2 FS Plus microscope with an Achroplan 40x water immersion objective. The inner limiting membrane of the retina was removed with a glass recording pipette and RGCs were targeted under control of a micromanipulator (MP-225; Sutter Instruments). RGCs were identified by their large somas and the presence of a voltage-gated sodium conductance. Recording pipettes (Garner Glass or Sutter Instruments) were pulled (PP-830; Narishige) with a tip resistance of 3–6 MΩ and filled with either a cesium gluconate (for all voltage-clamp recordings; in mM: 100 CsOH, 100 gluconic acid, 1.7 CsCl, 40 HEPES, 10 EGTA, 5 MgCl2, 1 QX-314, 2 Na2ATP, and 0.3 Na-GTP; pH adjusted to 7.25 with CsOH; ECl = −60 mV) or potassium gluconate internal solution (for current-clamp recordings; in mM: 98.3 K-gluconate, 40 HEPES, 1.7 KCl, 0.6 EGTA, 5 MgCl2, 2 Na2ATP, and 0.3 Na-GTP; pH adjusted to 7.25 with KOH). Data were acquired using pCLAMP 9 recording software and a Multiclamp 700A amplifier (Molecular Devices), sampled at 5 kHz and low-pass filtered at 1 kHz. 6,7-Dinitroquinoxaline-2,3-dione disodium salt (DNQX), D-(−)-2-Amino-5-phosphonopentanoic acid (AP5), 2,3-Dioxo-6-nitro-1,2,3,4-tetrahydrobenzo [f]quinoxaline-7-sulfonamide disodium salt (NBQX), SR-95531 hydrobromide (gabazine), strychnine, and Dihydro-β-erythroidine hydrobromide (DHβE) were added to ACSF as stock solutions prepared at ≥1000x concentration in water. DL-threo-b-Benzyloxyaspartic acid (TBOA) stock solution was prepared at 1000x in dimethyl sulfoxide. All antagonists were acquired from Tocris; all other chemicals were acquired from Sigma-Aldrich. Data were analyzed in pCLAMP, IGOR Pro (Wavemetrics Inc.) or MATLAB (Mathworks).
To calculate inter-cEPSC intervals in whole-cell voltage clamp recordings, cEPSCs were identified by eye. For Figure 2, a cluster was defined as a series of cEPSCs separated by not more than 10 s.
Outside-out patches were excised from RGC somas > 500 μm away from the simultaneously recorded RGC. Cesium gluconate was used for outside-out patch recordings. To confirm that patches contained glutamate receptors, L-glutamate was pressure-applied (5 psi) to patches at a concentration of 1mM in a solution containing (in mM): 150 NaCl, 2.5 KCl, 10 Hepes, pH 7.4. Patches were inserted to a depth of 10 μm in the inner plexiform layer, within 10 μm of the recorded RGC’s soma. All patch traces included were low-pass filtered (Bessel, 8-pole) at 150 Hz for clarity.
Retinas were bulk loaded with the calcium indicator Oregon Green 488 BAPTA-1 AM (OGB-1 AM; Invitrogen) using the multicell bolous loading technique (Stosiek et al., 2003). OGB-1 AM was prepared at a concentration of 10mM in a solution of 2% pluronic in DMSO, which was then diluted 1/10 in a solution containing (in mM): 150 NaCl, 2.5 KCl, 10 Hepes, pH 7.4 (Stosiek et al., 2003). The OGB-1 AM solution was pressure ejected from a borosilicate glass micropipette (Garner Glass or Sutter Instruments) using a PV-820 Pneumatic PicoPump (World Precision Instruments) at a pressure of 10 – 20 psi. The pipette was positioned just under the inner limiting membrane and dye was injected for 700 ms in 3–5 locations per retina. Epifluorescent calcium imaging was performed on a Zeiss Axioskop 2 FS Plus microscope using a 10x or 40x water immersion objective (Zeiss Achroplan), with illumination provided by a Sutter Lambda LS and controlled by a Uniblitz shutter. Images were acquired at 2–4 Hz in MetaMorph (Universal Imaging Corporation).
Paired t-tests were used for all paired comparisons. For multiple comparisons, a repeated measures ANOVA with a Newman-Keuls post hoc test was used. Levene’s test was used to compare variances. Alpha was always 0.05.
Support contributed by NIH Grants RO1EY013528 (MF), F31NS058167 (AB), EY01869 & NS016033 (DC), NSF Grant IOS-0818983 (MF), Research to Prevent Blindness (DC), That Man May See (DC), NIH (JJ), and Knights Templar Eye Foundation (JJ). We thank Dr. J. Isaacson for suggestions on the design of the experiments and for commenting on previous versions of the manuscript. We also thank Will Barkis for adapting the multicell bolus loading technique to developing retinal tissue.
Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final citable form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.