Retinal parallel processors: more than 100 independent microcircuits operate within a single interneuron - PubMed (original) (raw)
Comparative Study
Retinal parallel processors: more than 100 independent microcircuits operate within a single interneuron
William N Grimes et al. Neuron. 2010.
Abstract
Most neurons are highly polarized cells with branched dendrites that receive and integrate synaptic inputs and extensive axons that deliver action potential output to distant targets. By contrast, amacrine cells, a diverse class of inhibitory interneurons in the inner retina, collect input and distribute output within the same neuritic network. The extent to which most amacrine cells integrate synaptic information and distribute their output is poorly understood. Here, we show that single A17 amacrine cells provide reciprocal feedback inhibition to presynaptic bipolar cells via hundreds of independent microcircuits operating in parallel. The A17 uses specialized morphological features, biophysical properties, and synaptic mechanisms to isolate feedback microcircuits and maximize its capacity to handle many independent processes. This example of a neuron employing distributed parallel processing rather than spatial integration provides insights into how unconventional neuronal morphology and physiology can maximize network function while minimizing wiring cost.
(c) 2010 Elsevier Inc. All rights reserved.
Figures
Figure 1. Quantitative Evidence for Submicron Feedback Circuits in the Inner Retina
(A) Two-photon fluorescence image (3D projection) of an A17 amacrine cell superimposed upon a transmitted differential interference contrast (DIC) image of the retinal slice (scale bar, 50 mm). Black arrows indicate examples of synaptic feedback varicosities. (B) Serial EM reconstructions of RBC-A17/AII synapses: (left) AII and A17 amacrine cells receive the majority of RBC inputs (scale bar, 500 nm). (center; A17 varicosity transparent) A17 amacrine cell varicosities make en passant synapses onto RBC terminals; these reciprocal inhibitory synapses (yellow markers) are located less than a micron from RBC ribbon inputs (blue marker). (right; AII removed) Although an AII and A17 are typically paired postsynaptic to a RBC ribbon synapse, one AII receives multiple inputs from one RBC (blue markers; Singer et al., 2004; Strettoi et al., 1990), whereas each A17 varicosity (Vaney, 1986; Zhang et al., 2002) receives only one ribbon input. (C) An example of three consecutive serial sections further illustrates the typical synaptic configuration: a pair of reciprocal synapses in a single varicosity appose each ribbon input (scale bar, 500 nm). Ribbon and reciprocal synapses are indicated with blue and yellow arrows, respectively. (D) Summary histogram showing the number of excitatory and inhibitory synapses per varicosity identified via EM (n = 17). (E) Histogram showing the distribution of distances between neighboring varicosities from 3D reconstructions of neurites from 16 different A17s.
Figure 2. A17 Morphology Limits the Spread of Synaptically Evoked Depolarizations into Neighboring Neurites and Varicosities
(A) Application of DC current to an infinite neurite of 0.1, 0.13, 0.3, or 0.5 mm diameter was modeled and compared to the cable theory solution. Incorporation of varicosities at uniform intervals (20 mm) reduced the steady-state length constant. (B) Simulation design: a synaptic conductance, derived to match the time course of AMPAR EPSCs recorded experimentally, was imposed at individual varicosities along a single neurite while monitoring the consequent depolarization at all varicosities on the stimulated neurite and in a neighboring neurite on the same cell (for simplicity, nonstimulated neurites are not shown). (C) Postsynaptic depolarizations were large and had a nonlinear dependence on distance from the soma, suggesting that electrical compartmentalization occurs within neurites. (D and E) Normalization of responses (to that of the stimulated varicosity) reveals the relative electrical coupling between varicosities on the stimulated neurite (D) and those on neighboring, unstimulated neurites (E). See also Figure S1.
Figure 3. Biophysical Characterization of Pharmacologically Isolated Nav and Kv Conductances in A17 Amacrine Cells
(A) Nav activation protocol: when preceded by a hyperpolarizing voltage step (−100 mV; 100 ms), a series of depolarizing voltage steps (−70 to +30 mV in 20 mV increments; 100 ms) elicited fast and transient TTX (1 μM) -sensitive inward currents. (B) Nav activation data was converted to conductance, normalized to values obtained for steps to +30 mV, pooled (n = 9), and fit using the Boltzman equation to derive the half-maximal activation potential and Zδ (gating charge times distance traveled relative to membrane thickness). (C) Nav inactivation protocol: a series of conditioning steps (−140 to −30 mV in 10 mV increments; 100 ms) preceding a 100 ms step to −10 mV revealed the voltage dependence of Nav channel inactivation (trace data not shown). (D) Kv activation protocol: a series of depolarizing voltage steps (−70 to +30 mV in 20 mV increments; 200 ms) elicited large sustained outward currents that were partially sensitive to an A-type channel antagonist (4-AP; 4 mM). (E and F) Current-voltage plots of the transient (E; 3–8 ms) or sustained (F; 150–170 ms) Kv components for control and 4-AP conditions indicate that, although a rapidly inactivating 4-AP-sensitive current exists, it is only activated at potentials ≥ −10 mV. Error bars in (B), (C), (E), and (F) indicate standard deviation. See also Figure S2.
Figure 4. Active Membrane Conductances Do Not Amplify EPSPs or Enhance Spatial Signaling
(A) A17 Nav (top) and Kv (bottom) conductances were uniformly incorporated into the membranes of the electrotonic A17 model to match experimentally observed amplitudes (dashed lines represent average experimental values observed at corresponding potentials). (B) EPSPs elicited in individual varicosities of the active membrane A17 model (black circles) were negligibly larger than those elicited in the purely passive model (dashed line; note ordinate scale). (C) Compared to the passive model (dashed line), A17 active membrane conductances decreased the spatial extent of electrical signaling in neurites by ~12%. For comparison, the active model was also tested with Nav conductances expressed only in the neurites (4.6 mS/cm2; red lines and red circles); these simulations produced nearly identical results to those using uniform distributions. Synaptic input elicited spikes when the uniform Nav density was increased 70-fold (blue trace), but not 35-fold (green trace). (Inset) Voltage responses in varicosity #5 to synaptic input at varicosity #10 with 13 (1.8 mS/cm2; black), 353 (green), and 703 (blue) Nav channel density. (D–F) Experiments confirm modeling results: A17 Nav and Kv conductances do not contribute to membrane excitability. (D) EPSPs were not amplified by Nav conductances (i.e., were insensitive to 1 μM TTX) but were completely blocked by the AMPAR antagonist NBQX (10 μM). (E) Scatter plot indicates Nav and AMPAR antagonist effects on EPSPs for a group of five cells (dashed line represents unity). (F) Current-clamp protocols (left), designed to maximize the availability of Nav channels by minimizing inactivation, were unable to elicit action potentials (right). (G) Slowly inactivating L-type Cav channels were also incorporated into the active A17 model at uniform densities to match previous experimental results (Grimes et al., 2009). (H and I) Regardless of synaptic strength, Cav channels had negligible effects on local or spatial electrical signaling within neurites. Monitoring intravaricosity Cav current revealed that the spatial extent of Cav channel activation is 75% less than the spatial extent of membrane potential signaling (I). See also Figure S3.
Figure 5. Experimental Evidence for Highly Compartmentalized Functional Microcircuits
(A) Coupling experiment on a varicosity pair separated by ~30 mm. Synaptic Ca2+ transients were elicited in single varicosities under voltage clamp (blue, i). Ca2+ fluorescence was then monitored in the nearest neighboring varicosity without altering the stimulus (red, ii), but evoked Ca2+ transients in these varicosities were rarely observed. Following the switch to current clamp, Ca2+ signals were recorded again in both varicosities (iii and iv). Left panels indicate the EPSCs (top) and EPSPs (bottom) evoked by synaptic stimulation. (B) The same experiment in a varicosity pair separated by ~10 mm (different A17) demonstrates that closely neighboring microcircuits (<20 mm) are partially coupled. Note that synaptically evoked Ca2+ responses were observed in the neighboring varicosity only in the current-clamp configuration, indicating voltage-dependent interactions between closely neighboring varicosities. (C) Ratios of the time-locked Ca2+ transients, quantified by averaging over a 80 ms period following stimulation (20 ms after the stimulus onset, gray regions in A and B), plotted versus intervaricosity separation/distance (in microns; IVD3-D) for both voltage-clamp (black squares) and current-clamp (open triangles) configurations. Fitting the current-clamp Ca2+ ratios with a single exponential (see Experimental Procedures) yielded an effective space constant for neuritic Ca2+ signaling. Vertical black line and gray box indicate the mean and standard deviation of intervaricosity spacing, respectively. (D) Additional analysis shows a significant correlation between the relative Ca2+ response and IVD3-D (left) and also between the distance of var1 from the soma and the corresponding EPSP recorded with the somatic patch electrode (right) but not between the relative Ca2+ response and EPSP (middle). Additionally, there was no significant correlation between DG/R and distance to the soma (r2 = 0.0068, p = 0.82; data not shown).
Figure 6. Microcircuit Interactions Are Minimal under Scotopic Conditions
(A–C) Synchronized activation of inputs to multiple varicosities along a neurite can boost local responses. (A) Neuritic EPSPs in response to three, five, or ten synchronized inputs indicate that integration of inputs can occur but that neighboring neurites are indifferent regardless of input number. (B) The average electrical coupling between microcircuits depended on the number of inputs and the average inter-input distance. (C) Local Cav currents also can summate upon synchronized activation of multiple microcircuits. (Top) Cav currents in response to two synchronized inputs spaced by 167 μm were not boosted locally, but three synchronized inputs (bottom) with an average inter-input spacing of 84 μm significantly enhanced the local Cav current response. (D and E) Simulations of scotopic light conditions indicate that coincident activation of multiple microcircuits along A17 neurites is rare and that integration between microcircuits is minimal. Monitoring the neuritic Cav current response to the events/inputs occurring at varicosity #5 (for ten trials, gray lines) revealed the extent to which other inputs influence the local response (peak of black line) and the average spatial response to input at varicosity #5 (black line) for conditions mimicking the adaptation threshold and higher light intensities at which the gain of the RBC dyad has been significantly reduced. (D) At the adaptation threshold, each varicosity along a neurite receives randomly timed inputs at a rate of ~1.25 events/s. Although the rate of input to a simulated neurite was 10-fold higher than that of a single varicosity (i.e., convergence of ten RBCs to one neurite), there was no local summation, and spatial signaling was highly restricted under these conditions. (E) Same approach as in (D) but using higher light intensities that effectively reduce the gain at the RBC dyad. Under these conditions, the receptive field of a single A17 neurite sees ~50 Rh*effective/s, and under these conditions local responses and spatial signaling are only slightly influenced.
Figure 7. Wiring Optimization/Cost Theory Provides Insight into Why Evolution Would Produce a Wide-Field Retinal Interneuron of A17's Morphology that Contains Hundreds of Parallel Microcircuits that Operate Independently
(A) Several morphological configurations could provide functionally independent reciprocal inhibition to individual RBC ribbon synapses. Typical neurons in the CNS do not have multiple independent input-output units and therefore provide a one cell per one independent microcircuit cost (dashed line). Expanding a neuron's capacity to incorporate multiple independent microcircuits within a single neuron greatly reduces the cost per computational unit. Thin neurites connected directly to the cell body provide electrical and biochemical independence to distally located feedback microcircuits (gray line). Incorporation of additional microcircuits on single neurites separated by long (but shorter than the initial segment), thin neuritic segments further increases the single-cell capacity for independent processing and reduces the average cost (black line). (B) A17 morphology and parallel processing capacity reduces the net tissue volume attributed to RBC reciprocal feedback circuits by more than one hundred times that of the one microcircuit per cell configuration and could represent an evolutionary reduction in the thickness of the retina circuitry and an improvement in its optical transparency.
References
- Bloomfield SA. Relationship between receptive and dendritic field size of amacrine cells in the rabbit retina. J. Neurophysiol. 1992;68:711–725. -PubMed
- Bloomfield SA. Effect of spike blockade on the receptive-field size of amacrine and ganglion cells in the rabbit retina. J. Neurophysiol. 1996;75:1878–1893. -PubMed
- Chávez AE, Singer JH, Diamond JS. Fast neurotransmitter release triggered by Ca influx through AMPA-type glutamate receptors. Nature. 2006;443:705–708. -PubMed
- Chklovskii DB. Synaptic connectivity and neuronal morphology: two sides of the same coin. Neuron. 2004;43:609–617. -PubMed
Publication types
MeSH terms
LinkOut - more resources
Full Text Sources