In a new JGP study (Thoreson et al. https://doi.org/10.1085/jgp.202413746), anatomically realistic simulations reveal how the complex architecture of rod synapses influences glutamate dynamics and postsynaptic responses.

In darkness, rod photoreceptors release a stream of synaptic vesicles containing the neurotransmitter glutamate, which activates receptors on postsynaptic bipolar and horizontal cells. Absorption of even a single photon momentarily disrupts this stream of glutamate release and triggers bipolar cell depolarization. But how do bipolar cells distinguish this signal from the noise of random fluctuations in glutamate release rates? In this issue of JGP, Thoreson et al. perform new, anatomically realistic simulations of rod synapses and suggest that glutamate exits the synapses much more slowly than previously predicted, enabling slower vesicle release rates in the dark but increasing quantal variability in ways that could potentially complicate the ability of bipolar cells to distinguish genuine light responses from background noise (1).

Wallace Thoreson, Thomas Bartol, and Jeffrey Diamond.

The synaptic terminals of rod photoreceptors are spherical structures with a deep invagination that is typically filled by two bipolar cell dendrites flanked by two horizontal cell dendrites. These dendrites terminate near the presynaptic active zone, which contains a plate-like structure called the presynaptic ribbon that is associated with a pool of readily releasable synaptic vesicles.

Thoreson et al. performed anatomically realistic simulations of rod photoreceptor synapses based on reconstructions from serial electron micrographs. Glutamate released from a single synaptic vesicle diffuses much more slowly out of these geometrically complex synapses than it does out of simplified, spherical models with the same extracellular volume.

Previous studies modeled this complex architecture as a simple, saline-filled sphere and suggested that glutamate would be cleared from the synaptic cleft within a few milliseconds of its release from a synaptic vesicle (2). Such rapid clearance would mean that rods have to release ∼100 vesicles/s to prevent a random fluctuation in release from being mistaken for a genuine light signal, but measurements of the readily releasable vesicle pool indicate that the maximum release rate is only ∼36 vesicles/s (3).

“We decided to model the kinetics of single vesicle events in a more anatomically realistic setting,” explains Wallace Thoreson, of the University of Nebraska Medical Center. Thoreson and colleagues, including Thomas Bartol at the Salk Institute and Jeffrey Diamond at the National Institute of Neurological Diseases and Stroke, obtained serial block-face scanning electron micrographs of mouse retina and reconstructed four different rod synapses (1). These realistic structures were then incorporated into Monte Carlo simulations of glutamate diffusion and receptor activation.

“We were immediately surprised by how much longer glutamate persisted in the synapse,” Thoreson says. Indeed, according to the simulations, glutamate exits anatomically realistic synapses 10 times more slowly than it leaves simple, saline-filled spheres of identical volumes, largely because the extra geometric complexity constrains glutamate diffusion.

After exiting the synaptic cleft, glutamate is taken up by Müller glial cells that surround the rod synapse, but a 2006 study reported that a large amount of glutamate is retrieved directly into rods by the glutamate transporter EAAT5 (4). Thoreson et al.’s simulations suggest that there are ∼3,000 EAAT5 transporters in the rod cell membrane, and that they do, indeed, take up a functionally significant amount of glutamate. Nevertheless, most of the glutamate is retrieved by Müller cells outside of the cleft, which likely helps maintain a steep concentration gradient so that the neurotransmitter can slowly diffuse out of the synapse.

What does the persistence of glutamate in anatomically realistic synapses mean for the activity of postsynaptic glutamate receptors? Thoreson et al. found that incorporating a realistic geometry into their simulations made little difference to the predicted activity of horizontal cell AMPA receptors. But the activity of mGluR6 receptors on bipolar cells may be significantly prolonged, with a simple, binding-based model of mGluR6 suggesting that these receptors may remain active for >100 ms after each vesicle release.

“That allows these receptors to integrate individual release events over time, which would support slower vesicle release rates in resting rod cells,” Thoreson says. The required vesicle release rates could be lowered even further if, as Thoreson and colleagues have previously suggested, resting rod cells secrete multiple vesicles in semi-regular bursts instead of single vesicles at random intervals (5).

However, while geometric complexity may benefit rod synapses by enabling slower vesicle release rates, it could also increase variability in postsynaptic responses. The slower diffusion of glutamate means that it will take longer for the neurotransmitter to equilibrate throughout the cleft so that the concentration of glutamate that a postsynaptic receptor is exposed to could vary greatly depending on the precise site of vesicle release. Thoreson et al. simulated the release of vesicles from different positions along the presynaptic ribbon and found that the activity of mGluR6 and, especially, AMPA receptors was highly impacted by release site.

Thoreson and colleagues are now using their simulations to explore how this additional source of variability complicates the ability of downstream neurons to discriminate genuine, light-induced changes in glutamate release, as well as to investigate the contribution of regular, multivesicular bursts to this crucial step in low-light vision.

1.
Thoreson
,
W.B.
, et al
.
2025
.
J. Gen. Physiol.
2.
Rao-Mirotznik
,
R.
, et al
.
1998
.
J. Neurophysiol.
3.
Grabner
,
C.P.
, et al
.
2023
.
Nat. Commun.
5.
Hays
,
C.L.
, et al
.
2020
.
J. Gen. Physiol.
This article is distributed under the terms as described at https://rupress.org/pages/terms102024/.