Unconventional Transmission at Conventional Synapses of an Inhibitory Interneuron

Abstract

The AII amacrine cell (AII) is a key information hub in the retina, allowing rod-driven signals to piggyback onto cone-dominated circuitry. In this issue, Balakrishnan et al. (2015) use capacitance recordings to unravel mechanisms of inhibitory synaptic transmission by AIIs.

The AII amacrine cell (AII) is one of the best studied of more than 30 types of inhibitory interneurons in the vertebrate retina and is a key hub for information flow, contacting ≥28 types of retinal neurons (Marc et al., 2014) and linking rod-and cone-driven retinal circuits (Demb & Singer 2012). While AIIs also provide tonic crossover inhibition to some retinal ganglion cells (RGCs) during daylight conditions (Manookin et al., 2008), they are principally known for transferring rod-driven signals into the cone network (Figure 1) as a key part of the “primary rod pathway” (Bloomfield & Dacheux 2001). In dim light, photon absorption by rods depolarizes rod bipolar cells (RBCs). Cone bipolar cells (CBCs) make direct excitatory synapses onto RGCs, but RBCs do not and instead piggyback onto cone circuits by making excitatory glutamatergic synapses onto AII distal dendrites. AIIs then send the rod-driven signals to ON CBCs (excited by light) through sign-conserving electrical synapses and to OFF CBCs (excited by dark) through inhibitory glycinergic synapses. Despite the key roles played by AIIs in carrying signals through the retinal network, little is known about mechanisms regulating their synaptic output.

Figure 1. AII Amacrine Cell Synaptic Circuitry.

In the primary rod pathway, rod responses are transmitted to RBCs and piggyback on cone circuits via AIIs. This is accomplished with sign-conserving electrical synapses onto ON CBCs and sign-inverting glycinergic (gly) synapses onto OFF CBCs. CBCs make glutamatergic (glu) synapses onto RGCs.

In this issue of Neuron, Balakrishnan et al. (2015) tackle mechanisms underlying glycine release at this key retinal hub by measuring exocytotic capacitance signals in mouse AIIs. They characterize the Ca²⁺ dependence of exocytosis, postnatal synaptic maturation, and mechanisms supporting sustained synaptic transmission. Membrane capacitance (Cm) recordings allow direct measurement of exocytosis without requiring simultaneous recording from a postsynaptic neuron, permitting study of presynaptic mechanisms undistorted by postsynaptic receptor desensitization, saturation, or trafficking. Exocytotic Cm recordings have generally been limited to neurosecretory cells and large glutamatergic synapses; the present study is the first to employ Cm recordings for studying release from an inhibitory interneuron.

Because Cm measurements require extracting small phase changes in membrane current during a high-frequency sinusoidal voltage-clamp command, vesicle fusion sites must be located within the same electrotonic compartment as the recording site lest neuronal cable properties filter the sine wave and distort measurements of distant exocytotic events. With this constraint in mind, the authors go to great lengths to ensure that AII morphology is conducive to Cm recordings and that Cm signals recorded in AII somata truly reflect synaptic vesicle exocytosis. They show that the AII has two distinguishable electrotonic compartments, one of which likely corresponds to distal dendrites and the other likely corresponds to the soma, primary dendrite, and large lobular dendrites. Lobular dendrites are sites of glycine release from AIIs onto OFF CBCs and principal sites of Ca²⁺ influx through L-type channels (Habermann et al., 2003; Strettoi et al., 1992). Balakrishnan et al. show that depolarization of AIIs triggers Cm increases that are blocked by L-type Ca²⁺ channel blockers and increased intracellular Ca²⁺ buffering. Additionally, Cm increases were enhanced by increasing Ca²⁺ influx, reducing intracellular Ca²⁺ buffering, or raising the bathing solution temperature. These hallmarks of synaptic transmission suggest that the authors' Cm measurements represent true synaptic vesicle exocytosis.

A key task in the quest for understanding synaptic processing is to measure and characterize the functional subpopulations, or pools, of synaptic vesicles. Vesicle pools typically differ in their release kinetics, which can have important functional consequences for synaptic processing. To characterize synaptic vesicle pools in AIIs, Balakrishnan et al. measure Cm responses to steps of varying duration, finding two kinetically distinct pools (τ = 10 and 280 ms) of similar size (∼750 and ∼650 vesicles). Strikingly, with relatively long depolarizations, release begins to climb linearly with stimulus duration, suggesting that release can be sustained due to replenishment even when the stimulus is maintained for long periods of time.

Cm recordings measure the response across all active zones in a single electrotonic compartment. There is a wide range of estimates for the number of active zones in individual AIIs, which may reflect cell-to-cell variability and likely contributes to the response amplitude variability observed by Balakrishnan et al. Dividing the average total measured pool size of 1400 vesicles by the number of active zones suggests a releasable pool of 15–70 vesicles per active zone. Electron microscopy studies of AII lobular dendrites show that they are densely populated with vesicles (∼1500/μm³) with especially dense clusters at the presynaptic active zone surrounded by a slightly sparser cloud (Strettoi et al., 1992). Balakrishnan et al. suggest that the fast vesicle pool corresponds to vesicles docked at the membrane while the slower pool may correspond to vesicles in the surrounding cytoplasm.

Lobular dendrites do not possess synaptic ribbons but instead appear similar to other conventional synapses (Strettoi et al., 1992). Nevertheless, glycine release from lobular dendrites behaves in many respects like glutamate release from ribbon-bearing cells. First is the remarkable ability to sustain synaptic release for long periods of time. While rapid sustained release can be observed at conventional synapses (Hallermann and Silver, 2013), it is often thought to be the raison d'être for ribbon synapses. Second is the finding that release appears to be mediated by activation of L-type channels (CaV1) rather than CaV2 channels that more typically control release at conventional synapses. L-type Ca²⁺ channels show less inactivation than CaV2 channels and this property may contribute to the ability of AIIs to sustain release. Third, they find that varying the strength of the step depolarization triggers capacitance responses that appear to vary linearly with Ca²⁺ current-voltage relationship. Such a relationship could be explained by a linear release sensor (Ca²⁺ cooperativity ∼1) or by exocytosis being triggered by Ca²⁺ nanodomains in which single vesicles sense the Ca²⁺ cloud within ∼100 nm of single channels. However, the authors also find that release is diminished with higher EGTA levels in the patch pipette, which would not be expected for nanodomain control. In addition, release appears to decline linearly with the decline in Ca²⁺ driving force at positive potentials, consistent with a linear release sensor (Goutman and Glowatzki, 2007). Most synapses studied to date employ a Ca²⁺ sensor with a Ca²⁺ cooperativity of 3–4. Rod and cone ribbon synapses are an exception to this, possessing an as-yet-unidentified exocytotic Ca²⁺ sensor with a cooperativity of 1–2 (Thoreson et al., 2004).

Short-term synaptic plasticity, the enhancement or suppression of synaptic transmission occurring on the scale of milliseconds to minutes, is critical for shaping information processing by neural circuits (Abbott & Regehr, 2004). Measuring exocytotic Cm responses to pairs of stimuli, the authors show that AIIs exhibit a combination of facilitation and depression, depressing with short inter-stimulus intervals (0.5–1.5 s) and facilitating with longer, 2 s intervals. The slow facilitation onset is somewhat unusual, as facilitation typically occurs over shorter intervals (tens to hundreds of milliseconds) while depression tends to dominate over longer intervals. A plausible explanation for this pattern is that the depression due to synaptic vesicle pool depletion recovers after 2 s but Ca²⁺ levels recover more slowly and this residual Ca²⁺ facilitates further release. As short-term plasticity is important in synaptic filtering and information processing, it will be important to more fully characterize these properties in AIIs and determine how they influence the kinetically distinct rod- and cone-driven visual signals.

As is the case throughout the rest of the CNS, synapses in the retina continue to mature for several weeks after birth, a process that is mostly complete by eye opening, around postnatal day 14 (P14) in mouse retina. Glycinergic inhibition in the retina begins to appear around P10 (Schubert et al., 2008). Consistent with this, the authors find that exocytotic Cm responses also begin to emerge around P10. At P8, there was an almost total lack of voltage-gated Ca²⁺ influx. Just a day later at P9, they can evoke Ca²⁺ currents, albeit of variable amplitude, but little exocytosis. Yet when they reduce the Ca²⁺ buffering in the pipette solution, they are able to trigger exocytosis at P9. Because lower buffering allows Ca²⁺ to spread further from the channels, this suggests that the Ca²⁺ channels and synaptic vesicles are not tightly coupled at P9. These findings allow Balakrishnan et al. to infer that Ca²⁺ channels undergo developmental changes in localization. To test this, they make use of two-photon imaging to monitor the localization of Ca²⁺ signals throughout maturation. As with electrophysiology, the authors are not able to record fluorescent Ca²⁺ signals in AIIs at P8. By P10, however, they can detect Ca²⁺ signals in lobular dendrites, somata, and the primary dendrites. By P25, these Ca²⁺ signals are considerably larger and almost entirely restricted to the lobular dendrites, suggesting a migration of Ca²⁺ channels throughout maturation and pointing toward Ca²⁺ channel migration to active zones as underlying the changes in release efficiency. This might be a common mechanism underlying changes in exocytotic efficiency during synaptic maturation (Fedchyshyn & Wang, 2005; Wong et al., 2014).

Having revealed some of the exocytotic properties of AIIs, the authors turn to a series of experiments to test how light-evoked excitatory signals driven through the retinal network stimulate AII depolarization and trigger glycine release. To accomplish this, the authors record baseline Cm under voltage clamp and then switch to current clamp to deliver a light flash that depolarizes the AII. After resuming voltage-clamp Cm recording, they find intensity-dependent increases in exocytosis that do not saturate with increases in the AII voltage integral. Together, these data suggest that AII glycinergic output can respond indefinitely to sustained input, allowing the synapse to signal graded inhibition continuously to OFF CBCs.

Balakrishnan et al. describe properties of glycinergic transmission by AII lobular dendrites. AIIs are an important hub for information processing in the retina, contacting ≥28 different retinal cell types including each type of retinal bipolar cell (Marc et al., 2014), thereby shaping cone-driven responses at high light levels and sharing information between rod and cone pathways at lower light levels. The authors find that in many ways, glycine release from lobular dendrites of AIIs behaves surprisingly like transmission by ribbon-bearing synapses, exhibiting a remarkable capability for sustained release from a large pool of vesicles regulated by Ca²⁺ influx through L-type channels. The present results are consistent with other data showing that the ability for rapid sustained release is not limited to ribbon synapses (Hallermann and Silver, 2013). These findings fit with an emerging picture of highly specialized mechanisms operating at different synapses to serve diverse signaling demands (O'Rourke et al., 2012) and open the door to future study of how AII exocytotic properties are specifically suited for carrying retinal signals under different signaling regimes.