| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Department of Neuroscience, University of Minnesota, Minneapolis, Minnesota
Submitted 24 January 2005; accepted in final form 31 January 2006
 |
ABSTRACT |
|---|
 TOP ABSTRACT INTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
|
INTRODUCTION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
; Werblin and Dowling 1969). These cells generate large transient light-evoked responses at the onset and offset of light stimulation associated with impulse activity. They receive excitatory input from both depolarizing and hyperpolarizing bipolars (Miller 1979; Miller and Dacheux 1976b), which provides their sensitivity to positive and negative contrast images (Burkhardt and Fahey 1999). In the amphibian, ON-OFF amacrine cells can be activated by rod and cone pathways (Yang and Wu 2004).
The impulse activity generated by ON-OFF amacrine cells consists of two types of TTX-sensitive events that have been attributed to a somatic and dendritic origin (Miller 1979; Miller and Dacheux 1976a; Werblin 1977), and there is evidence that impulses initiated in the soma propagate into the dendrites of these cells (Cook and Werblin 1994). Since their discovery in fish and salamanders, ON-OFF amacrine cells have been reported in the rabbit (Dacheux and Raviola 1995), including the polyaxonal cells (Famiglietti 1992a, b; Volgyi et al. 2001), cat (Freed et al. 1996), turtle (Marchiafava and Weiler 1982), and primate (Stafford and Dacey 1997); they are presumed to be widely represented in all vertebrate retinas. Further subdivision of ON-OFF amacrine cells includes a distinction based on the size of the dendritic arborization into small and large field cell types and differences as well in the sublamination branching pattern within the inner plexiform layer (Pang et al. 2002).
ON-OFF amacrine cells in the amphibian retina generate a strong feedforward inhibition onto ganglion cells; it appears that both GABA and glycinergic subtypes are present (Belgum et al. 1984; Deng et al. 2001; Miller et al. 1977, 1981). More recent studies have determined that impulse activity in these cells is required for secretion of inhibitory neurotransmitter for both feedforward (Cook and Werblin 1994; Taylor 1999) and feedback (Shields and Lukasiewicz 2003) inhibition onto bipolar cell terminals. Thus there is renewed interest in the role that impulses play in mediating the actions of ON-OFF amacrine cells with special emphasis on the structure-function properties of their dendritic trees (Miller et al. 2002).
In the present study, we have carried out whole cell recording (WCR) and cell-staining experiments in the intact retina-eyecup preparation of the mudpuppy and tiger salamander to gain additional insight into the relationship between dendritic morphology and physiological characteristics. Based on morphological features derived from cell-staining experiments, we developed computer simulations of amacrine cells with compartmental models derived from these realistic structures. Three-dimensional reconstructions of ON-OFF amacrine cells based on serial confocal images revealed that they contain spines and large numbers of varicosities. Modeling strategies were carried out to determine the influence of varicosities on impulse propagation through the dendritic trees. WCRs from amacrine cells in the intact retina strongly suggest that they have a multifocal impulse generating capability. The dendritic impulses are likely to be initiated at numerous sites, presumably confluent with regions of synaptic input. Computer simulation studies indicate that impulse activity in the dendrites can be local in function, but when the somatic impulse is activated, it rapidly invades the entire dendritic tree within a few milliseconds. A model of the ON-OFF amacrine cell has been generated and is contrasted with a similar model of the retinal ganglion cell reported in a previous publication (Fohlmeister and Miller 1997b).
|
METHODS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
; Fohlmeister and Miller 1997b; Miller and Dacheux 1976b). Animal-care facilities and methods followed approved guidelines at the University of Minnesota. The eyecup was superfused using a gravity-fed system that had a flow rate of 0.5–1.0 ml/min at a temperature of 22°C. The external Ringer solution consisted of (in mM) 111 NaCl, 2.5 KCl, 1.8 CaCl2, 1 MgCl2, 5 dextrose, and 10 HEPES and was titrated to pH 7.80 with NaOH. Visual stimuli were provided by a computer-assisted image synthesizer (Innisfree), displayed through a monitor (Tektronix) with an intensity of 0.1 µW/cm2. Software routines gave computer control over stimulus parameters such as stimulus size, position and intensity.
Physiological signals were recorded using a Dagan 3900 amplifier, band-pass limited to 10 kHz. The data were digitized and stored at 12-bit resolution (sampled every 0.200–1.25 ms) with an analog/digital board (Lab Master, Scientific Solutions) in a PC computer using the pCLAMP data-acquisition package (Molecular Devices, Sunnyvale, CA). Patch electrodes were made from 1.2 mm OD, 0.95 mm ID omega dot glass (Friedrich and Dimmock, Millville, NJ) and were fabricated with a Narashige P-88 puller, using a two stage pulling technique. Electrodes were filled with (in mM) 98.0 KCH3SO4, 3.5 NaCH3SO4, 3.0 MgSO4 3, 1.0 CaCl2, 11.0 ethylene glycol-bis (
-aminoethyl ether)-N,N,N',N'-tetraacetic acid (EGTA), 5.0 HEPES, 2.0 glucose, 1.0 glutathione, 1.0 ATP-Mg2+, and 0.5 GTP-3-Na+. The pH was adjusted to 7.4 with KOH. The filled electrodes measured 
10 M
in resistance. For WCRs from the intact eyecup preparation, an important additional component of electrode fabrication was that of dipping the electrode tip into hexamethyldisilizane (Sigma) and allowing it to air dry. The hydrophobic surface created by the hexamethyldisilizane presumably facilitated the formation of gigaseals and eliminated the need for fire polishing the tip of the pipette even when penetrating deep into the retina. A single pipette could be used several times in an attempt to obtain gigaseals from intact cells in the retina, but positive pressure had to be applied before attempting additional seal formation.
Phase plots
For some data analysis, we generated phase plots, using whole cell recording data. Phase plots provide an important way to exaggerate and thus separate fast and slow response components (Fohlmeister and Miller 1997a, b). These plots were generated in Origin 7.5 (OriginLabs, Northampton, MA). A single array of data is first plotted as amplitude versus time. This plot is then used to generate a derivatized (dv/dt) analysis column of data points in which a derivative of the voltage versus time is determined for each point using the following relationship
 |
All intracellular staining experiments were carried out with sharp electrodes. Electrodes were filled with Lucifer yellow (dilithium salt, Sigma, St. Louis, MO), Lucifer yellow and Neurobiotin (Vector Laboratories, Burlingame, CA), sulforhodamine 101B (Molecular Probes, Eugene, OR), or horse radish peroxidase (HRP; Worthington Biochemical, Lakewood, NJ). The use of HRP for staining retinal neurons has been previously described (Arkin and Miller 1988). Sharp electrodes for recording and staining were fabricated from capillaries (thin-walled triangular glass, Friedrich and Dimmock). Triangular glass proved superior to other shapes for dye ejection experiments and also provided reasonable recordings for studying the light-evoked responses. For some experiments, the electrodes were filled with 4% Neurobiotin dissolved in 50 mM Tris-HCI buffer with the pH adjusted to 7.6 and connected to a pressurized electrode holder mounted on a hydraulically controlled advancer. Electrodes had tip resistances of 100–350 M.
Physiological identification of ON-OFF amacrine cells
Some studies were carried out with whole cell recordings from ON-OFF amacrine cells that were not subsequently stained for cell identification. In these cases, cell identification was based on the response characteristics classically associated with ON-OFF amacrine cells, including large amplitude, light evoked ON and OFF excitatory postsynaptic potentials (EPSPs) and large- and small-amplitude spikes. These cells are typically associated with noisy baseline recordings, but it is the small- and large-amplitude spikes in the recording that discriminate these cells from ganglion cell recordings, which generally show a single amplitude impulse. We did not study amacrine cells physiologically, which did not show the characteristic somatic and dendritic spike activity (Miller and Dacheux 1976a).
Intracellular staining
After impaling a cell and characterizing its responses to light, cells were iontophoretically injected with the cell marker using a combination of current and pressure (+0.5 nA DC constant current with 20 lb pressure into the electrode for 2 min and –0.5 nA DC constant current with 20 lb pressure for 2 min). The location of the injection site was noted, and that quadrant of the retina was not used for the remainder of the experiment. After the experiment, we exposed the retina to sulforhodamine for 1 h to further facilitate dye uptake into amacrine cell varicosities, using 2 mM SR dissolved in Ringer (Miller et al. 2001). After this incubation period, the retina was removed and placed on filter paper (Millipore, 0.3-µm pore size) using a slight suction to ensure adhesion. The filter paper reduced deformities of the retina during the normal histochemical processing and limited the cell shrinkage during dehydration for the HRP-DAB product (Miller and Bloomfield 1983). The retina was then placed in a 4% paraformaldehyde, 0.1% glutaraldehyde, and 0.2% picric acid fixative dissolved in 0.15 M phosphate-buffer saline (PBS), pH 7.4 for 25 min.
After fixation, the cells containing Neurobiotin were imaged as a horseradish peroxidase-diaminobenzidine tertrahydrochloride (HRP-DAB) reaction product viewed with light microscopy and suitable for computer tracing on the Eutectic Neuronal Tracing System (ENTS) system. Alternatively, tissue suitable for obtaining fluorescent images using the same staining procedure was accomplished using streptavidin conjugated to the fluorescent markers Cy-5 or Cy-3 (Jackson Laboratories, West Grove, PA); this mode of tissue preparation was better suited for three-dimensional reconstruction methods based on confocal imaging. Cells imaged for the HRP-DAB product were rinsed three times with PBS for 10 min each after fixation. To enhance penetration of the HRP reagent, the retina was treated with 1% Triton-X in PBS overnight at 4°C using an orbital tissue agitator (Thermolyne, Dubuque, IA). Subsequently, the tissue was incubated with VectaStain Elite ABC HRP reagent kit (Vector Laboratories) for 2 days at 4°C with tissue agitation. After another rinse in PBS for 10 min, the retina was reacted with a DAB solution to visualize the Neurobiotin-HRP product for 10 min or until a dark product appeared. Before mounting the tissue, it was rinsed with PBS for 
15 min then dehydrated using 50, 75, 80, 95, and 100% EtOH treatments for 3 min followed by two xylene (100%) rinses. Finally, the filter paper was dissolved using acetone (100%), and the retina was mounted in DePeX (Hopkin and Williams. Essex, UK).
When a fluorescent product was desired, the retina was placed on transparent filter paper before fixation (Millipore.0.3-µm pore size). The retina was rinsed twice for 10 min post fixation in PBS. To enhance penetration of the fluorescent reagent, the retina was treated with 1% Triton-X in PBS overnight at 4°C on an agitator. Next, the retina was incubated with streptavidin-Cy-5 or steptavidin-Cy-3 (diluted to 1:50) in 1% Triton-X in PBS for 2 days at 4°C on an agitator. The retina was rinsed for 5 h in PBS before mounting with Vectashield mounting medium (Vector Laboratories).
Cell tracing and confocal imaging
The retinas processed for the HRP-DAB reaction product were allowed to stabilize for 1 day before being viewed on an Olympus Vanox microscope equipped with S Plano x63 and x100 oil-immersion objectives (NA 1.4). Photographs were obtained in conventional brightfield microscopy or while using differential interference contrast (DIC) optics. The slides were transferred to a Nikon microscope (Optiphot) that was integrated with a Eutectic neuronal tracing system (ENTS). The tracing system allowed the cell's morphology to be entered into the computer as x, y, and z coordinates and included branch diameter, length, spines, and varicosities. In several instances, confocal images were used as the source rather than the bright field microscope and were traced using the same software. The computer-assisted trace rendition of the cell could be used for statistical analysis (Tocris et al.1995) and computer simulations (Velte and Miller 1996, 1997) or the morphology could be displayed and printed in any geometrical plane.
Retinas containing cells that were fluorescently labeled were imaged using a scanning laser confocal microscope to generate optical sections for three-dimensional reconstruction. The depth of any single z section image ranged from 0.275 to 4.0 µm, while the total depth of images varied from 15 to 70 µm. Most images were stored at 768 - 512 pixels with an 8-bit brightness pixel depth. All images were line or frame averaged over 8–12 samples; no other filtering methods were employed. All processing of the raw data were off-line using the volume rendering software program VoxelView (Vital Images, Fairfield, IA) running on a Silicon Graphics Crimson workstation or Imaris (Bitplane, Zurich) running on a PC.
Modeling ON-OFF amacrine cells
Models of ON-OFF amacrine cells, based on realistic morphologies, were derived from the digitized ENTS data, converted to compartmental models with ntscable (Velte and Miller 1995, 1996) and used for computer simulations with NEURON (Hines 1997). For each compartmental model, the spines and varicosities were included in the conversion from the ENTS representation and converted to compartmental models using ntscable. However, they were not included as separate structures but were instead integrated into the determination of a single compartment that was set as a length of dendrite equal to 
10 µm. Further division of each compartment was done in NEURON so that the maximum compartment size used for computer simulations was within a 1- to 3.3-µm-length window. Although the increase in compartment number added to the time required for simulations, the small size of each compartment ensured that active mechanisms could be faithfully simulated even in small processes. This was tested by studying the rate of impulse propagation along a dendritic branch while adjusting the number of compartments but keeping the other parameters of diameter, length and ion channel density constant. We found that the number of compartments we used for our simulations of amacrine cells was higher than that required to give the maximum impulse conduction time along the structure. This analysis suggested that our simulations were sufficiently optimized to provide good fidelity for the response properties under study.
We modeled impulse activity in ON-OFF amacrine cells using the five-channel model of impulse generation developed by Fohlmeister et al. (Fohlmeister and Miller 1997a, b; Fohlmeister et al. 1990) and included INa, IK, IK,A, IK,Ca, ICa, and the leakage channel (ILeak) to provide the background membrane resistance. These currents have been identified in amacrine cells of the tiger salamander retina (Eliasof et al. 1987).
Synaptic activation of ON-OFF amacrine cells was modeled with 
conductance changes, in which the time variant conductance (gt) varies as
 |
, which defines the time of the peak conductance change. The current injected into the compartment followed the conventional relationship
 |
conductance changes used in amacrine simulations used a value of 0 mV for the reversal potential of the Esyn. Using a single-compartment model with conductance changes, we determined that modifications of the peak delay were sufficient to provide an accurate representation of light-evoked EPSPs at the onset and offset of the stimulus, after impulses were blocked with TTX. |
RESULTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
, 1996) for computer simulations using the program NEURON. During the course of this study, we discovered that ON-OFF amacrine cells consisted of both small- and large-field subtypes. The small-field cells had dendritic spreads of about 300 µm, whereas the large-field cells were substantially larger in overall diameter. Although different in size, these two cell groups showed similarities in their overall structure and physiological properties. However, because the small-field cells formed a minority of the ON-OFF amacrine cell population in our hands, we have confined our analysis to the large field subtypes. ON-OFF amacrine cells are not coupled
We stained several cells with Neurobiotin to determine if ON-OFF amacrine cells were coupled to other neurons. Neurobiotin is a relatively small marker that effectively passes through gap junctions. Experiments in which the electrodes were filled with Lucifer Yellow (LY) and Neurobiotin allowed us to see the cell from which the recording was made (LY does not pass well through gap junctions) as well as the cells to which the impaled neuron was coupled through the transcellular spread of Neurobiotin. This method revealed dye coupling among horizontal cells, Müller cells, and bipolar cells, but we did not see coupling between ON-OFF amacrine cells or between them and other neurons. Thus we believe that the modeling analysis of this study, based on a single uncoupled cell, matches our experimental conditions.
Prominence of varicosities
Figure 1 illustrates a large-field ON-OFF amacrine cell the image of which was reconstructed from 40 signal averaged optical sections obtained with a confocal microscope. The large, nonspherical soma gives rise to a single prominent dendritic trunk that appears to give off all of the smaller branches as it progresses, divides, and diminishes in diameter. Varicosities are a prominent feature of the smaller dendritic branches but are largely absent from the main trunk. Varicosities on the smaller branches usually begin near their point of departure from the trunk. Each varicosity is connected by very fine dendritic branches that are often faintly stained, presumably because of their small diameter. Dendritic varicosities vary in size and sometimes exist as a very subtle expansion of the dendrite but are generally 1 µm in thickness and somewhat oblong to tear-shaped in their appearance.
|
A more subtle feature of the amacrine dendrites is the presence of spines, some of which have been highlighted by positive contrast circles that are expanded to the right (Fig. 1, 1–4). These structures contain very fine spine necks that typically end in a spine head or terminal varicosity. In a few cases, we have observed that these small structures give rise to two consecutive varicosities and thus appear to be more like a very small, fine terminal dendritic branch, but the majority have a single neck and a single spine head.
The extent of the dendritic dimensions of this large-field ON-OFF amacrine cell was not established because the confocal images were all obtained within a single microscopic field of view. Indeed, all large-field ON-OFF amacrine cells had their dendritic tree dimensions exceed the area covered by a single x40 objective field of view.
Figure 2 illustrates four different ON-OFF amacrine cells that were dye filled with intracellular injections, after which serial optical depth sections were obtained with the confocal microscope and served as the basis for three-dimensional (3D) reconstructions. Both volume (left) and surface (middle) reconstructions were carried out, and the third vertical panel (right) illustrates the attempts to isolate and highlight the population of varicosities in each cell by thresholding out the soma and the proximal dendritic branches together with some of the interconnecting processes. Differences between these cells are apparent. The cells in Fig. 2, A, C, and D, all showed a more dense branching pattern than the cell in B. Differences in the density of varicosities were also apparent, as the cells in Fig. 2, C and D, seemed to have a higher varicosity density. Some variance was also observed in the structure of the primary dendrites, which included a large primary dendrite (Fig. 2, A and B) or a less obvious primary dendritic structure (B and C). It was not possible to determine if all dendritic branches emerged from a common primary dendrite, as seemed to be the case in Fig. 2A, or whether more than one primary dendrite was present (B). Thus there appear to be some subtle morphological differences within the large-field, ON-OFF amacrine cell population.
|
Whole cell recordings from intact ON-OFF amacrine cells: is there only one type of dendritic impulse?
The data presented in Fig. 3 addresses two major questions about dendritic and somatic spike activity: can dendritic spikes originate from more than one site and is the somatic spike always preceded by a dendritic impulse? In this figure, both physiological results (A–F) and modeling analysis (G and H) are presented for comparison. Figure 3A shows a whole cell recording from an ON-OFF amacrine cell in response to a focal light stimulus to the receptive field center. The light response consisted of an initial EPSP associated with large-amplitude impulses present at the onset and offset of the light stimulus. In this example, the ON response had a relatively sustained component, more so than one would see with broad field illumination. A closer examination of the responses shows small, fast, aperiodic events (marked by 
) that are interspersed between the large-amplitude impulses. We have referred to the large-amplitude spikes as somatic spikes, whereas the small fast events have been referred to as dendritic impulses (Miller 1979; Miller and Dacheux 1976a; Werblin 1977). Figure 3B illustrates an attempt to separate the small and large spikes by moving the focal light stimulus 100 µm away from the center position. In this case, the light response was smaller, and no impulse activity at light offset was evident. The frequency of the somatic impulses was reduced, and each somatic spike was preceded by a smaller spike with a considerable delay between the two events, suggesting a very weak but significant functional relationship between them. Figure 3C provides yet another example from a different cell with the focal spot displaced from the center. Note that before the first large-amplitude spike, several smaller fast events are apparent and the large impulses have smaller spike events before and in between them.
|
Light stimulation of ON-OFF amacrine cells always evoked a dendritic spike before the somatic spike could be observed (Miller 1979; Miller and Dacheux 1976a). In the early days of intracellular recording, it was thought that because each somatic spike was preceded by a dendritic spike, somatic spikes could only be generated by that sequence (Miller 1979; Miller and Dacheux 1976a) of events, implying a tight obligatory coupling between them. However, WCR data from intact amacrine cells suggests that tight coupling between the two spike events is not always evident and that it is more a product of the stimulation method with broadfield light stimuli more likely to evoke a seemingly tight relationship between dendritic and somatic impulses.
In a series of experiments, we used displaced focal light stimulation to determine if a consistent temporal relationship existed between the dendritic and somatic impulses. The displaced spot experiments also allowed us to determine if each dendritic spike had the same waveform when recorded in the soma; a uniform dendritic spike amplitude and waveform, independent of spot location, would imply that dendritic spikes either had a single site of generation or, alternatively, propagated toward the soma, with the active to passive transition taking place at the same electrotonic distance from the soma. In contrast, if each locally activated dendritic site generated an impulse near that site, one might expect to see differences in dendritic spike amplitude and waveform if active impulse propagation toward the soma failed at variable distances from the cell body. The thick traces of Fig. 3D show two recordings from the same cell evoked by a different spatial position of the small spot, whereas the thin traces illustrate the time derivative of each recording. The trace labeled 13, which reflects a stronger activation of the cell, shows a slowly rising, light-evoked EPSP, with a somatic spike that was not immediately preceded by a dendritic spike. The arrow on the rising phase points to a smooth transition to the somatic impulse with the time derivative indicative of a single event. In contrast, the second trace (2) reflects a weaker stimulus response than that of trace 13; in this case, the somatic spike was directly preceded by a dendritic spike (arrow) and the time derivative reflects two events. In other words, the recordings illustrated in Fig. 3D suggest that somatic spikes can be generated independently of dendritic spikes and are therefore not obligatorily dependent on them. Figure 3E illustrates another example from a different cell in which a displaced focal light stimulus evoked two somatic spikes that were not preceded by an obvious dendritic impulse as evident in both the voltage recording and its derivative (dotted line).
Dendritic spikes can sum like synaptic potentials
Another way of evoking dendritic and somatic spike activity is through intracellular current injection. This approach provides a different way of viewing the dendritic and somatic spike relationships. Figure 3F illustrates a recording from an ON-OFF amacrine cell in response to current injected into the soma (+20 pA). In this example, two dendritic spikes show a summation with one another, and the following somatic impulse is preceded by a dendritic impulse, which creates the notch on the rising phase of the action potential. We interpret the spike summation to indicate that the injected current gave rise to two independent dendritic spikes, presumably generated in two different dendrites. Spike generation was followed by active propagation of these independent impulses, both of which failed as they traveled toward the soma, such that their passively decayed currents summed in the soma recording. The impulses are too close to one another to be generated at a single site because the first spike would be in its refractory period and could not generate the second impulse. This type of recording event favors the concept that more than one dendritic site is involved in impulse generation and that two different dendritic action potentials can be generated at different sites with near simultaneous initiation.
Although single vesicular events are too small and slow to be confused with dendritic impulses, Yang et al. (2002) have described spontaneous and light-evoked responses they refer to as transient depolarizing potentials (TDP). These events were large in amplitude and comparatively fast, although they were readily distinguished from action potentials. They attributed these large, postsynaptic events to multivesicular release of glutamate induced by regenerative calcium currents observed in bipolar terminals (Burrone and Lagnado 1997; Protti et al. 2000; Zenisek and Matthews 1998). These TDPs were only observed in very dark-adapted preparations, using a retinal slice preparation. We used the whole retina-eyecup preparation in our study and avoided a dark-adapted environment. We did not detect TDP events in our study either as spontaneous or light-evoked responses; these responses will not be considered further.
Figure 3, G and H, illustrates modeling studies that are relevant for the physiological data presented in A–F. The traces of Fig. 3G illustrate a simulated synaptic event using the time course of vesicular events derived from modeling studies based on the release of glutamate and the activation of AMPA receptors (Sikora et al. 2005). In this example, the conductance change of the event was magnified by 20-fold over the conductance change derived for a single vesicular event. One of these events was generated in the soma compartment; the bottom trace illustrates how the identical event would look if it had been generated in a proximal dendrite but recorded in the soma. The peak delay for the synaptic event is 8.4 ms, whereas the peak delay for the first spike event in Fig. 3F is 2 ms. Of equal importance is the fact that the spike event has a clear sharp peak, whereas the synaptic event has a slow time course, which makes it difficult to determine the peak position. A second event was introduced into the soma at the arrow (sMini). This would be the size of a single vesicular event generated in the soma. Note that the small size and slow time course of the response make it difficult to resolve as a single event and that it is much smaller than the spike events we attribute to proximal dendritic spike activity.
Figure 3H illustrates a modeling study using a compartmental model of an ON-OFF amacrine cell with synaptic current injected as an function into the soma compartment. Synaptic events are readily recognized as such and should not be confused with dendritic impulses, provided that the dendritic spikes travel close enough to the soma to be recognized as spike-like events, such as those we have identified as dendritic spikes in Figs. 3 and 4. In this computer simulation, a current-clamp record is illustrated (v) together with the time course of the conductance change underlying the voltage response. Note that the peak conductance change occurs well before the peak of the voltage response and is effectively completed shortly after the peak of the voltage response. The repolarization that takes place after the peak voltage is largely determined by the time constant of the cell because it takes place after the conductance change is completed. In contrast to the synaptic event, the impulse has active mechanisms for both the depolarizing and hyperpolarizing components; indeed one can see the afterhyperpolarization present in the somatic impulses of Fig. 3, A–F. The active depolarization-repolarization sequence gives the spike a sharp peak and serves to discriminate impulse activity from that of synaptic events. However, if the dendritic impulse is generated at a very remote distance (see Fig. 9) and fails to propagate at a very distal site, it can be virtually impossible to distinguish these heavily attenuated impulses from synaptic events when both are recorded in the soma.
|
|
|
The slowing of the light-evoked EPSP, evident in several traces in Fig. 4 as a result of TTX application, was unexpected. The loss of an amacrine-mediated TTX-sensitive feedback onto bipolar terminals (Shields and Lukasiewicz 2003) should enhance release of glutamate from bipolars, not suppress it. One possible explanation of EPSP suppression as a result of TTX application can be appreciated from recent experiments by Ichinose et al. (2005), who described TTX-sensitive Na+ channels in a group of transient ON bipolars. In these bipolar cells, the Na+ channels serve as a booster to light-evoked synaptic inputs that feed into transient ganglion cells with smaller synaptic inputs into these cells when the channels are blocked with TTX. It is not known if these bipolars also feed into ON-OFF amacrine cells, but the presence of this boosting mechanism could account for the slower and smaller EPSP we observed for focal stimuli in the presence of TTX. Although this mechanism may account for the changes we observed in EPSP amplitude and time course, it cannot account for the loss of the fast events we have identified as dendritic spike activity.
One way of demonstrating the TTX effect on impulse activity in amacrine cells is to use phase plots in which the time derivative of each trace is plotted against the membrane potential. Phase plots have proven highly useful for interpreting impulse activity and separating it from other kinds of physiological events. To generate the phase plots (see METHODS for a more complete description), we first plotted the time variant voltage response for each trace (Fig. 4), after which the derivative of each plot was obtained using the derivative math routine in the plotting program (Origin 7.5). The differentiated trace (dv/dt) was then plotted as the ordinate against the membrane potential (abscissa) to illustrate how rapidly the membrane potential changes as a function of the membrane potential itself. The phase plots generated in this way produce a stationary plot reflecting the dynamic processes associated with impulse generation. Figure 5A shows the phase plot relationships before (black circles) and during (red circles) the application of TTX. These data were generated from 4 spot displacement responses of the data in Fig. 4 (2,4,5, and 7). In this phase plot, the somatic impulse is better understood as a clockwise event in which the impulse rises above the background noise between –45 and –40 mV and has maximum amplitude during the rising phase of the spike and reaches its peak (dv/dt = 0) at about –10 mV (the peak value of the impulse was attenuated by the bandwidth of the recording device), followed by impulse recovery with a maximum negative rate of change between –10 and –20 mV.
|
). Thus amacrine cell light responses may consist of four components, including an early slow-rising EPSP, distally evoked dendritic spike activity, more proximal, attenuated dendritic spike activity, and the conventional large-amplitude impulses we attribute to the soma. In summary, the action of dendritic spikes serves to accelerate the rise time of the light response and provides a boost that facilitates the generation of the somatic spike. In the absence of dendritic spikes, the remaining EPSP has a slower rise time and lacks the small, fast, irregular events that we attribute to dendritic spike activity, which decays passively into the soma. It is important to emphasize, however, that the events we have attributed to dendritic spike activity in our recordings have been identified as such because of their sensitivity to TTX. In the absence of TTX sensitivity, dendritic spikes that propagate close to the soma are readily identified as attenuated impulses, whereas dendritic spikes that begin and end more distally can experience an erosion in amplitude and distortion in waveform such that they may be difficult to distinguish from synaptic events generated in the same cell.
Passive properties of ON-OFF amacrine cells
Figure 6 schematically illustrates the morphology of an ON-OFF amacrine cell of the mudpuppy retina as a Sholl plot (Sholl 1953) modified to illustrate the electrotonic (
) length of the processes based on variations of the specific membrane resistance (Rm), which varied from 5,000 to 100,000 cm2. The electrotonic distances from the soma are indicated (- - -). The four representations of the same cell show how it becomes more electrotonically compact as Rm increases while the morphological parameters and internal resistance (Ri = 110 cm) remained constant. At 100,000 cm2, the electrotonic lengths of the entire dendritic tree are each <0.4 . Whole cell recordings obtained from intact ON-OFF amacrine cells in the mudpuppy and tiger salamander retina indicate that the input resistance is relatively high and the time constant appropriately large (Coleman and Miller 1988; Eliasof et al. 1987), so that Rm values of amacrine cells in these species are probably closer to the high end of this comparative scale. For the simulations presented in this study, we have used an Rm value of 70,000 cm2 and an internal resistance of Ri = 110 cm as we have previously used for models of ganglion cells in the amphibian retina (Fohlmeister and Miller 1997a, b; Sikora et al. 2005; Velte and Miller 1995, 1996). The amacrine cell illustrated in Fig. 6 was converted to a compartmental representation and used for all of the computer simulations in this study.
|
Do impulses travel faster than synaptic potentials in the dendrites of ON-OFF amacrine cells? Figure 7 summarizes comparative simulations of an ON-OFF amacrine cell in which the speed of passive versus active propagation was examined. The morphology of the cell is illustrated in Fig. 7A, in which a single dendrite—exaggerated with a thick dashed line—was selected with six displaced recording sites to register propagation velocity. Varicosities were not included but are separately considered below. When the soma is the only impulse generating compartment (Fig. 7B) with passive dendrites, an impulse initiated in the soma decays rapidly toward the distal dendritic branches due to the membrane capacitance, which quickly dissipates the fast rising current generated by the impulse. Thus passive propagation of impulses generated at a single site is a poor means of communicating information within ON-OFF amacrine cell dendrites. A more effective means of communication within these cells can be achieved by synaptic potentials, the rate of rise of which is much slower than that of the nerve impulse and therefore less degraded by the membrane capacitance. Using a single example, an conductance change with a peak delay of 50 ms was injected into the soma compartment of the model with passive dendrites and soma (Fig. 7C). In this case, the EPSP simulation was more effectively transmitted to the distal branches than the passively conducted impulse and decayed to about half of the soma peak value, with a peak-to-peak delay of 30 ms between the soma and distal dendritic compartment.
|
10 times faster than the propagation speed of the synaptic current: an animation illustrating impulse traffic from the soma into the dendrites can be visualized at [http://www2.neuroscience.umn.edu/RFM/On-OffAmCurrent_%20in_Soma_JNP2006.avi]. Clearly if speed of information is important for functional expression of these cell types, then impulse generation and propagation is by far the most effective and rapid means of conveying polarization changes throughout the cell.
|
In view of our observations that the dendrites of ON-OFF amacrine cells are richly endowed with varicosities, we considered the possibility that impulse propagation in the dendritic trees of these cells is affected by the size and distribution of this prominent component of their dendritic morphology. We evaluated this in two ways. Figure 8A diagrams the models used for evaluation of varicosities. First, we contrasted the influence of varicosity size (1–10 µm) versus dendritic diameter size (0.1–1 µm), and then we contrasted whether the varicosities and/or the interconnecting dendrites were active or passive. Ion channel density was selected that fell within the range that gave physiological amplitudes of spikes in amacrine cell recordings (0.75X and 1.0X). All models were compartmental models with each dendritic process divided into compartments of 3.33 µm, and each varicosity was divided into three compartments. The first evaluation used a long 1 process with a single, variable-diameter varicosity in the middle and stimulation at one end was used to trigger the propagated action potential. For the 1.0- and 0.5- µm processes, active impulse propagation was evident through the varicosity, whether it was active or passive for the full range of varicosity diameters. For the 0.1-µm process, active impulse propagation through the varicosity was blocked when it was 7.5 µm in diameter for an active varicosity and was blocked when the diameter was 5 µm for a passive varicosity. A varicosity diameter of 1 µm and a connecting process of 0.1 µm (probably the lower limit of process diameter) provided active propagation if the varicosity was active or passive.
|
simulations described previously. We concluded from the simulations based on models summarized in Fig. 8 that varicosities neither enhance nor significantly retard the propagation of the nerve impulse through the complex of fine dendritic branches interconnecting the varicosities of ON-OFF amacrine cells, provided that the ion channel density supporting impulse generation is appropriate.
Is saltatory conduction through varicosities possible?
We evaluated one other question about varicosity function: is it possible for active varicosities to form the basis of a saltatory transmission of impulse propagation to enhance the speed of conduction from one active varicosity to another through the array of varicosities found within each cell? These simulations represented a subset of the second simulation set (Fig. 8B) in which all of the varicosities were active, while the interconnecting structures were passive. For these simulations, we quickly determined that a saltatory varicosity to varicosity propagation model was unlikely because the membrane capacitance of the interconnecting passive processes substantially slowed conduction. Even reducing the membrane capacitance to 0.1 µF/cm2 (from the standard value of 1.0 µF/cm2) did not significantly improve the situation over that of having actively conducting dendritic processes, although it substantially enhanced the rate of rise of the impulse. Thus our simulation studies suggest that impulse propagation in the dendrites of ON-OFF amacrine cells is most effective and rapid when the varicosities and the interconnecting processes actively participate in impulse generation and propagation. As a corollary, we suggest that a varicosity to varicosity, saltatory conduction mechanism of the dendritic nerve impulse is unlikely based on the assumptions and parameters of our models.
Models of spiking ON-OFF amacrine cells
Models of spiking ON-OFF amacrine cells were established to simulate the kinds of impulse activity and recordings observed in WCRs. Our objective with this approach was to establish dendritic and somatic impulse generation sites that could initiate impulses independently of one another and to establish a range of dendritic ion channel density so that multiple dendritic sites could be independently activated and give rise to impulses that propagated sufficiently close to the soma to be recognized as attenuated impulses of similar size and shape to those observed in physiological recordings. We refer to this arrangement as a multifocal model of impulse generation in which dendritic sites can initiate impulses in localized regions of the dendrites and potentially play a role in focal activity of these cells.
What is the density of ion channels in dendritic processes?
We first established a range of dendritic ion channel density sufficient to generate dendritic impulse activity conforming to our physiologically based expectations. We independently manipulated the channel density of the dendrites and soma. Table 1 summarizes the ion channel subtypes and their density for each of the different models that we evaluated. The multichannel model consisted of time- and voltage-dependent ionic currents, including INa, IK, IK,Ca, ICa, IK,A, and the leakage channel (ILeak). Separate models were generated in which the dendrites had 0.25, 0.5, 0.75, 1.0, 2.0, and 4.0 times the soma values.
Figure 9 illustrates recordings from the cell body of a large-field ON-OFF amacrine cell model while sequentially stimulating the distal ends of each dendrite with a brief current pulse. Vertical displacement of the traces was done for clarity. When the dendritic channel density was 25% of the soma (0.25X), impulses could be generated in each of the terminal dendrites with sufficiently strong current, but they decayed into passively conducted impulses that failed too far away from the soma to be recognized as attenuated impulses in soma recordings. But these events did resemble miniature EPSPs. In contrast, with this channel density, an impulse generated in the soma could actively propagate into most (but not all) terminal dendritic branches. Impulse traffic from the soma out into the dendrites takes advantage of the more favorable impedance matching arrangements, such that active propagation for impulses is more likely for centrifugal movement than impulse movement in the opposite direction. As the dendritic ion channel density increased, the speed of impulse propagation was enhanced, and the active dendritic spike could propagate closer to the soma before failing. At the 0.5X channel density, an impulse generated in the soma actively propagated into all dendrites, including the distal ends of each terminal branch.
In the range of 0.5–1.0X, the soma recordings revealed small fast responses that would be interpreted as action potentials of dendritic origin, and they showed summation properties because they were generated in separate dendrites: an animation of the 1X model can be visualized at: [http://www2.neuroscience.umn.edu/RFM/On-OffAmSequentialFig_9_JNP2006.avi]. The colored lines of the selected dendrites illustrate how far toward the soma the impulse actively propagated before failing to a passive mode using the 1X model and sequential distal dendritic impulse activation. If the active impulses headed toward the soma and continued through the branch point as an active impulse, then an impulse was always generated in that branch point and moved distally to actively invade all daughter branches including their terminal endpoints. Note that the dendritic pathway outlined in green for terminal dendrite 13 propagated actively into the soma and evoked a somatic action potential that, in turn, activated the entire dendritic tree (except for the dendritic process connected to terminal branch 13). Further increases in the channel density to 2X and 4X resulted in the impulse activity generated by a single terminal dendrite evoking a somatic impulse and activating the entire dendritic tree in several examples. In each case, when the single dendrite initiated the impulse in the soma, all of the dendritic tree was activated with the exception of the dendrite that gave rise to the somatic spike because that dendritic membrane was in its refractory period recovering from its own prior spike event.
In summary, dendritic ion channel density ranging from 0.5 to 1.0X resulted in passive dendritic spike responses in soma recordings that fell within the range observed from physiological recordings in both amplitude (1–10 mV) and waveform. The models that seemed to correspond to many of the amacrine cell properties revealed by WCR in the intact retina were the 0.75X and 1X models as defined in Table 1. In these models, dendritic spikes generated in the distal dendrites did not actively propagate into the soma (except for one case: dendrite 13) but became passive sufficiently close to the soma to be recognized as attenuated impulses. However, because each dendrite experienced a block at different distances from the soma, dendritic spikes of several different amplitudes and shapes were evident in the soma recording. Also in this model, an impulse generated in the soma actively propagated throughout the entire dendritic tree and all terminal branches. Thus the 0.75X and 1X models have both local and global capabilities depending on whether the soma is sufficiently activated to generate independent impulse activity.
Are ON-OFF amacrine cells directionally biased?
Unexpectedly, the modeling studies presented in Fig. 9 revealed a directional bias for dendritic impulse traffic in the cell that could not be appreciated without computer simulations. Setting aside the issue of synaptic activation and considering only the 4X model, activation of dendrites 1–27 consistently generated an impulse in the soma that then spread to the entire dendritic tree, whereas activation of dendrites 18–28 generated impulses that decayed passively into the soma. One could imagine that a visual target moving from lower right to upward left (arrow) could generate activity that would be confined to the dendritic tree structure and would be less likely to activate a more global response of the cell. In contrast, a target moving in the opposite direction would have a more favorable opportunity to activate the entire cell. This directional bias is entirely accounted for by the diameter of the dendritic processes involved. The largest dendritic branch is on the upper left (activated by terminal branch 13), and this structure was sufficiently large near the soma that the current was able to trigger the somatic spike and result in global activation of the cell. Thus an amacrine cell such as that illustrated in Fig. 9 is likely to have an intrinsic directional bias for impulse traffic that originates in the dendrites, imposed by the impedance matching properties of the parent dendrite as it approaches the soma. Imposing a threshold for release of transmitter could create a synaptic basis for generating directional selectivity in an appropriately wired ganglion cell (Fried et al. 2002). Although directionally selective ganglion cells have been described in the mudpuppy, only 1% of recorded cells revealed this type of organization (Karwoski and Burkhardt 1976). Furthermore, there is evidence that cholinergic amacrine cells are present in the amphibian (Zhang and Wu 2001) and could subserve mechanisms of directional selectivity as has been postulated in the rabbit retina (Fried et al. 2002). We will not further consider the role of ON-OFF amacrine cells as substrates for directionally selective mechanisms in the amphibian retina.
Is the somatic spike essential for amacrine cell function?
Simulations were carried out to determine the role of the soma for the spiking behavior of ON-OFF amacrine cells. Using a 0.75X or 1X model, we made the soma active or passive by turning on or off all of the voltage-gated ion channels of the five-channel model. In these two models, making the soma passive did not alter the general properties of cell behavior because the proximal dendrites had sufficient ion channel density (and surface area) to serve as the soma-like pacemaker. The only difference in soma recordings between the two was that for a passive soma, the soma impulse was slightly smaller in amplitude compared with that observed when the soma was active. Thus although it may seem counterintuitive, we found that the proximal dendrites could serve as the impulse initiating site for global operations of these interneurons. Indeed, because the exact site of the large-amplitude spike cannot be established with intracellular recordings alone, it is probably better to think of the two spike events as "proximal" and "distal" or "dendritic" rather than assuming that the large impulse is necessarily of somatic origin.
Synaptic activation of ON-OFF amacrine cells
Figure 10 illustrates the results of simulating synaptic activation of an ON-OFF amacrine cell (A) using the 1X model. Two different simulations were carried out, and the results are illustrated in Fig. 10, B (broad center stimulation indicated by 
) and C (focal stimulation confined to 6 inputs indicated by 
). For each simulation, functions were used (peak delay: 50 ms) at 250 and 750 ms to emulate the ON and OFF light responses of the cell. At each input site, the conductance was tailored, through trial and error, to provide stimulation that activated the cell in an appropriate manner. These conductances ranged from 100 to 1,000 pS; the same set of inputswere activated for the OFF component but with conductances reduced to half of those for the ON component. It is worth emphasizing that in recent modeling studies in which synaptic activation between bipolars and ganglion cells was based on salamander data, Sikora et al. (2005) concluded that the size of the postsynaptic AMPA receptor bed could be modeled as a group of 128 receptors. Assuming a peak conductance of the AMPA receptors of 20 pS, the co-activation of this receptor bed could generate a peak conductance change of 2,560 pS. This conductance change is adequate to reach dendritic spike threshold within the intermediate and distal branches (where the input resistance is higher) but not in the more proximal branches or in the soma. However, this calculation does not include the N-methyl-D-aspartate (NMDA) receptor contributions. Nevertheless, the single-site conductance changes that we have applied to discrete regions of the dendritic tree are below those that could be generated if the entire hypothesized bed of AMPA receptors was co-activated. For both the diffuse and focal simulation analyses, the conductance changes were repeated with the Na+ channel conductances turned off, illustrated in the - - - and labeled as TTX. A faint box labeled 1 and 2 shows the region of expanded time scale presented on the right as Fig. 10, B, 1 and 2, and C, 1 and 2. For the diffuse stimulation in Fig. 10B, a train of impulses follows the time course of the transient EPSP, but the early events leading to somatic impulse activity—expanded to the right (Fig. 10B1)—when spiking was present shows that the rise time of the EPSP without impulse activity (TTX) was accelerated by dendritic spike current that reached the soma and presumably served to enhance the speed at which the impulse was generated. The OFF activity of Fig. 10B shows two somatic spikes that were preceded by dendritic spikes; a third event, which did not evoke a somatic spike, shows two dendritic spikes on an expanded scale (Fig. 10B2, 
) that sum in a manner virtually identical to the summation seen physiologically in Fig. 3F [a 3rd smaller dendritic spike is evident between the 2 similar size attenuated impulses (
)]. Each of these dendritic spikes was generated in separate branches, failed to actively propagate near the soma, and summed at that level. These two impulses could not be generated by the same dendrite due to limitations imposed by their refractory periods: an animation of this simulation can be visualized at [http://www2.neuroscience.umn.edu/RFM/On-OffAmacrineFullFieldStimFig10_A_B_JNP2006.avi].
The trace in Fig. 10C shows results from the focal simulation (6 input sites: in Fig. 10A) in which a reduced number of somatic spikes is evident. But in the absence of a large number of somatic spikes, one can see different amplitude dendritic spikes in the soma recording. The expanded view on the right (Fig. 10C1) illustrates that the initial rate of rise of the EPSP (in TTX) is substantially accelerated by the arrival of three dendritic spike events () that were not sufficient to bring the soma to spike threshold, but additional dendritic spike activity, combined with the EPSP, initiated the soma impulse. Notice in Fig. 10C1 that the · · · (TTX) would probably not have reached the action potential threshold without the superimposed dendritic spike activity: an animation of this simulation can be viewed at http://www2.neuroscience.umn.edu/RFM/On-OffAmacrineFocalStim_Fig10_A_C_JNP2006.avi.
|
DISCUSSION |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
; Vallerga 1981; Wong-Riley 1974; Yang et al. 1991), this is the first report based on three-dimensional reconstruction methods from flatmount preparations and confocal images. This approach has served to clarify some of the microscopic morphological details of amacrine dendrites, including the detection of spines and a more quantitative assessment about the number and dendritic location of varicosities that form a prominent component of these dendritic structures. A second important constituent to this study was derived from our ability to obtain whole cell recordings from single cells in the intact retina eyecup, thereby avoiding the potential distortion of physiological functions that can be imposed when using slice preparations or dissociated cells. Thus the combination of these two methods, when coupled with the compartmental simulation studies based on realistic morphologies, presents the first and most complete attempt to interrelate the form and function of these inhibitory interneurons. Spines in amacrine cells
Although spines appear to be a prominent part of the structure of ON-OFF amacrine cells, we know little about their synaptic connections. Spines have been observed in amacrine cells of many other species (Casini et al. 1995; Cuenca and Kolb 1998; Famiglietti 1992a, b; Teranishi and Negishi 1991; Wong and Collin 1989). In the rabbit retina, a developmental study suggested that amacrine cell spines were more common in young animals and diminished but did not disappear in the adult (Wong and Collin 1989). In the primate, Cuenca and Kolb (1998) reported that spines of substance-P-containing amacrine cells (that also contain GABA) received input from bipolar cells.
Generalizing on the assumption that spine heads in amacrine cells receive glutamatergic, synaptic input from bipolar cells, a brief discussion of spine function in neurons found in other parts of the nervous system, such as the hippocampus, seems timely. In general it is currently believed that spines serve as an important subcellular site, underlying the generation of long-term potentiation (LTP). According to this view, LTP begins with synaptic activation of NMDA receptors in spine heads, which in turn mobilizes AMPA receptors into the spine head to enhance synaptic transmission (Nimchinsky et al. 2002; Shi et al. 1999). It is worth noting that ON-OFF amacrine cells have light-evoked responses dependent on NMDA receptors, whereas the sustained amacrine cells of the salamander lack NMDA receptor-mediated inputs (Dixon and Copenhagen 1992; Velte et al. 1997). It will be intriguing to learn more about the structure of spines in amacrine cells and whether these structures subserve modulation of synaptic efficiency through NMDA receptor activation. Because spine heads are electrically and chemically isolated from the dendrite, their high-input resistance serves to exaggerate voltages generated by synaptic currents (Miller et al. 1985; Segev and Rall 1988), and the small volume of the head allows calcium to accumulate to levels not possible if the same current was applied to a larger dendritic structure. In the case of ON-OFF amacrine cells, the additional presence of voltage-gated Na+ channels in spine heads raises the possibility that action potentials could be generated in these structures by the activation of a single synaptic input.
Varicosities in amacrine cells
All of the reconstructed ON-OFF amacrine cells in this study had numerous varicosities. A single large-field cell had >500 varicosities with spacing between neighboring varicosities as small as a few micrometers. Although they vary in size and shape, varicosities are commonly 1 µm in diameter and stand in sharp contrast to the fine dendritic segments with which they interconnect. These latter structures may be as small as a few tenths of 1 µm. To optimally visualize the varicosities, we found that three-dimensional reconstructions of single cells from serial confocal images were invaluable for appreciating their numbers, size and position along the dendritic tree. Filling cells with HRP or Neurobiotin and examining them with brightfield microscopy did not reveal the varicosities as effectively as reconstructions based on confocal images. Enhancement of varicosity staining was aided by applying an activity-dependent dye to the eyecup after recording the cell, presumably because the varicosities of amacrine cells acquire these dyes through synaptically related endocytosis (Miller et al. 2001).
Modeling the ON-OFF amacrine cell
Our modeling efforts of ON-OFF amacrine cells were based on the idea that the dendrites of these cells have a sufficient Na+ channel density to initiate impulse activity at or near regions of synaptic activation. All of the light-evoked responses we observed physiologically revealed that the initial response began with an EPSP, followed by one or more dendritic spikes, followed in turn by one or more somatic impulses if the stimulus was sufficiently large. Light stimulation did not obligatorily lead to any spiking events, but once the dendritic spike was initiated, it propagated to the distal dendritic branches but typically failed to actively propagate at some point as the impulse moved toward the soma (unless the dendritic Na+ channel density was high as in the 2x or 4x models).
Studies with focal light stimulation support the idea that each dendritic input site can generate impulse activity. The models in which ion channel density is uniform throughout the dendritic trees conform to this expectation. Focal light stimulation, coupled with the application of TTX, revealed that the light-evoked synaptic responses were slower and peaked well after impulse activity was initiated under control conditions. Because this kind of response was observed at virtually every stimulus location when focal stimulation was used, we concluded that dendritic spikes can be initiated within each dendrite, presumably close to the site of excitatory synaptic excitation. Thus our results favor a multifocal model of dendritic impulse initiation, consistent with the physiological observations of this report.
Bloomfield (1996) studied the effects of TTX on the receptive fields of ganglion cells and large-field amacrine cells in the rabbit retina. Whereas TTX did not change the receptive field size of ganglion cells, those of large field amacrine cells were reduced by 40% with TTX or intracellular application of QX-314. He concluded that Na+-dependent action potentials were required for normal physiological organization and receptive field characteristics of large-field amacrine cells. His study eliminated the possibility that the Na+-dependent action potential influence on the receptive field size could reflect input from other spike-generating neurons because the intracellular blocking agent QX-314 had similar effects. The analysis of the present study offers support for the idea that attenuated action potentials generated in the distal dendrites fail at regions sufficiently remote from the soma and appear in the soma by adding to the "synaptic current." This model can account for the presence of a TTX-sensitive component that appears to add to the synaptic currents recorded in the soma. In our simulations of ON-OFF amacrine cells, we observed impulse traffic in several distal dendritic branches that failed at the first branch point (14, 15 and 17 in Fig. 9) far removed from the soma in which case they contributed currents in the soma that would be difficult to clearly identify as impulse events.
Cook et al. (1998) have provided evidence that impulses are essential for the feedforward inhibitory actions of glycinergic ON-OFF amacrine cells onto ganglion cells. The results of this study provide an outline of the cellular events that may underlie this action. A surround (annulus) stimulus will provide synaptic input into many different input sites of a number of ON-OFF amacrine cells. For each amacrine that is activated, synaptic current will give rise to dendritic spike activity that will rapidly activate one or more somatic spikes, which in turn will reach all of the dendrites of that cell within a few milliseconds. Because the time constant of an amphibian ganglion cell is typically 70 ms (Coleman and Miller 1988), each potential site of transmitter release will be effectively depolarized, providing a near simultaneous release of transmitter over a time course that is an order of magnitude less than the time constant of the ganglion cell. Presumably the kind of surround stimulus commonly used in experiments will activate a number of ON-OFF amacrine cells, but if sufficiently activated, each amacrine cell can be expected to respond similarly. Thus the traveling impulse that begins at or near the soma is a rapid and effective means of ensuring that the dendrites that release transmitter will be encouraged to do so in rapid succession.
The experiments of Cook et al. (1998) showed that in the presence of TTX, inhibitory transmitter release was still evident for small spot stimulation to the center. In our recordings of ON-OFF amacrine cells, we observed that stimuli directed to the center in the presence of TTX evoked large-amplitude EPSPs that triggered a slower spike-like event that we have attributed to calcium channels. Thus the large center-evoked EPSP which appeared sufficient to activate calcium channels could explain how center spots retain their ability to evoke transmitter release from ON-OFF amacrine cells in the presence of TTX.
Although the somatic spike observed in ON-OFF amacrine cells appears to be a powerful means of communication with the entire dendritic tree, stimuli that are more localized, or below threshold for somatic spike activation, can activate dendritic spikes which propagate centrifugally and confine their actions to that dendrite and its daughter branches. That is, dendritic spike activity can provide a local function restricted to a subregion of the cell. It is not clear if this operational mode is important for the function of ON-OFF amacrine cells. However, because many of these cells have dendrites that exist in both the ON and OFF sublamina, the possibility of ON-OFF amacrine cells functioning as either ON or OFF small-field inhibitory units must be considered and could be subserved by the restricted distribution of dendritic spike activity and the synaptic connections mediated by a single dendrite and its daughter branches. In other words, the results of this study are consistent with the idea that ON-OFF amacrine cells can function as both local and global inhibitory interneurons as suggested in early studies of these cells (Miller 1979).
One cautionary note about the possibility of local dendritic actions mediated by dendritic impulses is appropriate. Our compartmental models were all derived from computer assisted tracings of amacrine cells stained with visible dyes (HRP/Neurobiotin) and viewed under flatmount conditions. Under these conditions, the dense staining of the cell body did not allow us to determine if the primary dendrites emerged from the soma or from a common central stalk. The resolution of this issue could determine whether local dendritic impulse traffic is practicable in these cells. If all of the dendrites emerged from a single common stalk, then the impedance matching for each dendritic ending near the soma could be optimized in such a way that centripetal impulse traffic from any peripheral dendrite could effectively give rise to somatic spikes. Based on our three-dimensional reconstructions it appears that some dendrites are arranged in this way, but more detailed anatomical reconstructions will be needed to resolve this issue. We should also emphasize that clustering of Na+ channels to form trigger regions or zones could change the site or sites of action potential initiation and impulse traffic.
Eliasof et al. (1987) and Barnes and Werblin (1987) studied action potentials in ON-OFF amacrine cells of the tiger salamander. Their work was primarily directed to the somatic action potential, which showed rapid accommodation. They concluded that this rapidly accommodating somatic spike was the result of an I Na that was partially inactivated at the resting membrane potential. We did not attempt to study this phenomenon in the current modeling approach for two reasons. First, we had the impression that rapidly accommodating somatic spikes were more common with intracellular recordings when compared with WCRs. In our experience, whole cell recordings from intact cells tended to reveal less rapid accommodation of the somatic spike. Second, some stimulus conditions, such as intermittent light stimulation, could generate high-frequency repetitive firing that included somatic spike activity. Another issue relates to the origin of the somatic spike. Our modeling studies raised the possibility that the somatic spike could be generated by the proximal dendrites and that a completely passive soma still revealed a large amplitude spike from the soma compartment. This possibility raises questions about the validity of voltage-clamping the soma and intuiting properties of the large spike, which likely has a strong dendritic component to it. It is well established that the ion channel composition of the dendritic tree has an influence on the pattern of impulse activity generated in the soma (Fohlmeister and Miller 1997b; Mainen and Sejnowski 1996).
The large-amplitude EPSPs observed in ON-OFF amacrine cells seem designed to activate impulses and the thin dendritic processes may serve to enhance the background input resistance so that small synaptic currents can more effectively reach spike threshold. Rarely did we detect spontaneous impulse activity in ON-OFF amacrine cells, but in the few cases in which spontaneous impulse activity was observed, it was clear that this unusual event was not similar to the spontaneous pacemaker discharge reported for the dopaminergic amacrine cells studied in the mouse retina (Feigenspan et al. 1998) in which spontaneous impulse activity is generated in these cells due to an I Na that can be activated in the interspike voltage ranges. In our experience, dendritic spikes were not typically active spontaneously and had thresholds that required a depolarization to initiate them. The somatic impulses were similar in that they too had a threshold voltage, although the exact threshold value was not determined.
Advantages of dendritic impulses in amacrine cells
When a diffuse light stimulus activates a ganglion cell, the response begins with a slow rising EPSP. But in the amacrine cell, the EPSP is quickly converted to dendritic impulse activity that provides a powerful, nonlinear boost to the voltage and can rapidly initiate the events that release neurotransmitter at synaptic sites. Furthermore, these dendritic impulses serve to reduce the latency for somatic spike initiation, which then rapidly activates the remaining dendrites to provide a powerful enhancement for transmitter release and its inhibitory consequences. In principle, the ganglion cell can be inhibited before it can generate impulse activity. Thus co-equal activation of the amacrine and ganglion cell favors the amacrine cell inhibiting the ganglion cell either before it can fire impulses or shortly after impulse activity begins. This simple fact may help explain why ganglion cells in the amphibian retina respond poorly to diffuse light stimulation, which strongly activates ON-OFF amacrine cells. This powerful physiological advantage means that to activate the ganglion cell, the light stimulus must be appropriate for the excitatory receptive field of the ganglion cell and less appropriate for strongly activating the amacrine cell. Furthermore because the inhibition from the amacrine cell is largely on the dendrites of the ganglion cell, local computations can be carried out so that asymmetry of inhibition versus excitation can be achieved and can perhaps contribute to more specialized features of the receptive field, such as directional selectivity or other specialized trigger features.
Amacrine versus ganglion cell model
The multifocal model of the ON-OFF amacrine cell developed in this study contrasts sharply with the ganglion cell model developed to account for impulse generation in the axon-bearing ganglion cells (Fohlmeister and Miller 1997b). At one level, it seems that the amacrine and ganglion cell models are not very different. In the ganglion cell model, Na+ channels were also required in the dendrites and evidence of dendritic action potentials has been reported (Velte and Masland 1999). Voltage-gated Na+ channels in the ganglion cell model were sufficient in density to generate a back-propagated impulse (Fohlmeister and Miller 1997b) but not a forward-propagated impulse as appears to be present in the dendrites of ON-OFF amacrine cells. To meet this requirement, the ganglion cell model had dendritic Na+ channel density of 25 mS/cm2, whereas in the amacrine cell models of this study, a range of 50–100 mS/cm2 was sufficient to create feedforward local dendritic impulse activity. Thus this difference in ion channel density between the amacrine and ganglion cell models was sufficient to account for the feedforward generation of spikes in amacrine cell dendrites versus the exclusive back-propagated impulses suggested for ganglion cell dendrites (Fohlmeister and Miller 1997b).
|
GRANTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
|
ACKNOWLEDGMENTS |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
Present address of N. P. Staff: Dept. of Neurology, Mayo Clinic, 200 First St. SW, Rochester, MN 55905.
|
FOOTNOTES |
|---|
Address for reprint requests and other correspondence: R. F. Miller, Dept. of Neuroscience, University of Minnesota, 6-145 Jackson Hall, 321 Church St. SE, Minneapolis, MN 55455 (E-mail rfm{at}umn.edu)
|
REFERENCES |
|---|
|
TOPABSTRACTINTRODUCTIONMETHODSRESULTSDISCUSSIONGRANTSACKNOWLEDGMENTSREFERENCES |
|---|
Barnes S and Werblin F. Direct excitatory and lateral inhibitory synaptic inputs to amacrine cells in the tiger salamander retina. Brain Res 406: 233–237, 1987.[CrossRef][Web of Science][Medline]
Belgum JH, Dvorak DR, and McReynolds JS. Strychnine blocks transient but not sustained inhibition in mudpuppy retinal ganglion cells. J Physiol 354: 273–286, 1984.
Bloomfield SA. Effect of spike blockade on the receptive-field size of amacrine and ganglion cells in the rabbit retina. J Neurophysiol 75: 1878–1893, 1996.
Burkhardt DA, and Fahey PK. Contrast rectification and distributed encoding by ON-OFF amacrine cells in the retina. J Neurophysiol 82: 1676–1688, 1999.
Burrone J and Lagnado L. Electrical resonance and Ca2+ influx in the synaptic terminal of depolarizing bipolar cells from the goldfish retina. J Physiol 505: 571–584, 1997.
Casini G, Rickman DW, and Brecha NC. AII amacrine cell population in the rabbit retina identification by parvalbumin immunoreactivity. J Comp Neurol 356: 132–142, 1995.[CrossRef][Web of Science][Medline]
Coleman P and Miller RF. Measurement of passive membrane parameters with whole cell recording of neurons in the intact amphibian retina. J Neurophysiol 61: 218–230, 1988.[Web of Science]
Cook PB, Lukasiewicz PD, and McReynolds JS. Action potentials are required for the lateral transmission of glycinergic transient inhibition in the amphibian retina. J Neurosci 18: 2301–2308, 1998.
Cook PB and Werblin FS. Spike initiation and propagation in wide field transient amacrine cells of the salamander retina. J Neurosci 14: 3852–3861, 1994.[Abstract]
Cuenca N and Kolb H. Circuitry and role of substance P-immunoreactive neurons in the primate retina. J Comp Neurol 393: 439–456, 1998.[CrossRef][Web of Science][Medline]
Dacheux RF and Raviola E. Light responses from one type of ON-OFF amacrine cells in the rabbit retina. J Neurophysiol 74: 2460–2468, 1995.
Deng P, Cuenca N, Doerr T, Pow DV, Miller R, and Kolb H. Localization of neurotransmitters and calcium binding proteins to neurons of salamander and mudpuppy retinas. Vision Res 41: 1771–1783, 2001.[CrossRef][Web of Science][Medline]
Dixon DB and Copenhagen DR. Two types of glutamate receptors differentially excite amacrine cells in the tiger salamander retina. J Physiol 449: 589–606, 1992.
Eliasof S, Barnes S, and Werblin F. The interaction of ionic currents mediating single spike activity in retinal amacrine cells of the tiger salamander. J Neurosci 7: 3512–3524, 1987.[Abstract]
Famiglietti EV. Polyaxonal amacrine cells of rabbit retina: morphology and stratification of PA1 cells. J Comp Neurol 316: 391–405, 1992a.[CrossRef][Web of Science][Medline]
Famiglietti EV. Polyaxonal amacrine cells of rabbit retina: PA2, PA3, and PA4 cells. Light and electron microscopic studies with a functional interpretation. J Comp Neurol 316: 422–446, 1992b.[CrossRef][Web of Science][Medline]
Feigenspan A, Gustincich S, Bean BP, and Raviola E. Spontaneous activity of solitary dopaminergic cells of the retina. J Neurosci 18: 6776–6789, 1998.
Fohlmeister JF, Coleman PA, and Miller RF. Modeling the repetitive firing of retinal ganglion cells. Brain Res 510: 343–345, 1990.[CrossRef][Web of Science][Medline]
Fohlmeister JF and Miller RF. Impulse encoding mechanisms of ganglion cells in the tiger salamander retina. J Neurophysiol 78: 1935–1947, 1997a.
Fohlmeister JF and Miller RF. Mechanisms by which cell geometry controls repetitive impulse firing in retinal ganglion cells. J Neurophysiol 78: 1948–1964, 1997b.
Freed MA, Pflug R, Kolb H, and Nelson R. ON-OFF amacrine cells in cat retina. J Comp Neurol 364: 556–566, 1996.[CrossRef][Web of Science][Medline]
Fried SI, Munch TA, and Werblin FS. Mechanisms and circuitry underlying directional selectivity in the retina. Nature 420: 411–414, 2002.[CrossRef][Medline]
Hines M. NEURON: a program for simulation of nerve equations. In: Neural Systems: Analysis and Modeling, edited by Eeckman FH. Norwell, MA: Kluwer Academic, 1997, p. 127–136.
Ichinose T, Shields CR, and Lukasiewicz PD. Sodium channels in transient retinal bipolar cells enhance visual responses in ganglion cells. J Neurosci 25: 1856–1865, 2005.
Kaneko A. Physiological and morphological identification of horizontal, bipolar and amacrine cells in goldfish retina. J Physiol 207: 623–633, 1970.
Karwoski CJ and Burkhardt DA. Ganglion cell responses of the mudpuppy retina to flashing and moving stimuli. Vision Res 16: 1483–1495, 1976.[CrossRef][Web of Science][Medline]
Mainen ZF and Sejnowski TJ. Influence of dendritic structure on firing pattern in model neocortical neurons. Nature 382: 363–366, 1996.[CrossRef][Medline]
Marchiafava PL and Weiler R. The photoresponses of structurally identified amacrine cells in the turtle retina. Proc R Soc Lond [Biol] 214: 403–415, 1982.[Medline]
Miller JP, Rall W, and Rinzel J.. Synaptic amplification by active membrane in dendritic spines. Brain Res 325: 325–330, 1985.[CrossRef][Web of Science][Medline]
Miller RF. The neuronal basis of ganglion cell receptive field organization and the physiology of amacrine cells. In: The Neurosciences, edited by Schmitt FO and Worden FG. Cambridge, MA: MIT Press, 1979, p. 227–245.
Miller RF and Bloomfield SA. Electroanatomy of a unique amacrine cell in the rabbit retina. Proc Natl Acad Sci USA 80: 3069–3073, 1983.
Miller RF and Dacheux R. Dendritic and somatic spikes in mudpuppy amacrine cells: indentification and TTX sensitivity. Brain Res 104: 157–162, 1976a.[CrossRef][Web of Science][Medline]
Miller RF and Dacheux RF. Synaptic organization and ionic basis of on and off channels in mudpuppy retina. I. Intracellular analysis of chloride-sensitive electrogenic properties of receptors, horizontal cells, bipolar cells, and amacrine cells. J Gen Physiol 67: 639–659, 1976b.
Miller RF and Dacheux RF. Synaptic organization and ionic basis of on and off channels in mudpuppy retina. II. Chloride-dependent ganglion cell mechanisms. J Gen Physiol 67: 661–678, 1976c.
Miller RF, Dacheux RF, and Frumkes TE. Amacrine cells in Necturus retina: evidence for independent gamma-aminobutyric acid- and glycine-releasing neurons. Science 198: 748–750, 1977.
Miller RF, Fagerson MH, Staff NP, Wolfe R, Doerr T, Gottesman J, Sikora MA, and Schuneman R. Structure and functional connections of presynaptic terminals in the vertebrate retina revealed by activity-dependent dyes and confocal microscopy. J Comp Neurol 437: 129–155, 2001.[CrossRef][Web of Science][Medline]
Miller RF, Frumkes TE, Slaughter M, and Dacheux RF. Physiological and pharmacological basis of GABA and glycine action on neurons of mudpuppy retina. II. Amacrine and ganglion cells. J Neurophysiol 45: 764–782, 1981.
Miller RF, Stenback K, Henderson D, and Sikora M. How voltage-gated ion channels alter the functional properties of ganglion and amacrine cell dendrites. Arch Ital Biol 140: 347–359, 2002.[Web of Science][Medline]
Nimchinsky EA, Sabatini BL, and Svoboda K. Structure and function of dendritic spines. Annu Rev Physiol 64: 313–353, 2002.[CrossRef][Web of Science][Medline]
Pang JJ, Gao F, and Wu SM. Segregation and integration of visual channels: layer-by-layer computation of ON-OFF signals by amacrine cell dendrites. J Neurosci 22: 4693–4701, 2002.
Protti DA, Flores-Herr N and von Gersdorff H. Light evokes Ca2+ spikes in the axon terminal of a retinal bipolar cell. Neuron 25: 215–227, 2000.[CrossRef][Web of Science][Medline]
Segev I and Rall W. Computational study of an excitable dendritic spine. J Neurophysiol 60: 499–523, 1988.
Shi SH, Hayashi Y, Petralia RS, Zaman SH, Wenthold RJ, Svoboda K, and Malinow R. Rapid spine delivery and redistribution of AMPA receptors after synaptic NMDA receptor activation. Science 284: 1811–1816, 1999.
Shields CR and Lukasiewicz PD. Spike-dependent GABA inputs to bipolar cell axon terminals contribute to lateral inhibition of retinal ganglion cells. J Neurophysiol 89: 2449–2458, 2003.
Sholl DA. Dendritic organization in the neurons of the visual and motor cortices of the cat. J Anat 87: 387–406, 1953.[Web of Science][Medline]
Sikora MA, Gottesman J, and Miller RF. A computational model of the ribbon synapse. J Neurosci Methods 145: 47–61, 2005.[CrossRef][Web of Science][Medline]
Stafford DK and Dacey DM. Physiology of the A1 amacrine: a spiking, axon-bearing interneuron of the macaque monkey retina. Vis Neurosci 14: 507–522, 1997.[Web of Science][Medline]
Taylor WR. TTX attenuates surround inhibition in rabbit retinal ganglion cells. Vis Neurosci 16: 285–290, 1999.[CrossRef][Web of Science][Medline]
Teranishi T and Negishi K. Dendritic morphology of a class of interstitial and normally placed amacrine cells revealed by intracellular Lucifer Yellow injection in carp retina. Vision Res 31: 463–475, 1991.[CrossRef][Web of Science][Medline]
Toris CB, Eiesland JL, and Miller RF. Morphology of ganglion cells in the neotenous tiger salamander retina. J Comp Neurol 352: 535–559, 1995.[CrossRef][Web of Science][Medline]
Vallerga S. Physiological and morphological identification of amacrine cells in the retina of the larval tiger salamander. Vision Res 21: 1307–1317, 1981.[CrossRef][Web of Science][Medline]
Velte TJ and Masland RH. Action potentials in the dendrites of retinal ganglion cells. J Neurophysiol 81: 1412–1417, 1999.
Velte TJ and Miller RF. Dendritic integration in ganglion cells of the mudpuppy retina. Visual Neurosci 12: 165–175, 1995.[Web of Science][Medline]
Velte TJ and Miller RF. Computer simulations of voltage clamping retinal ganglion cells through whole-cell electrodes in the soma. J Neurophysiol 75: 2129–2143, 1996.
Velte TJ and Miller RF. Spiking and nonspiking models of starburst amacrine cells in the rabbit retina. Vis Neurosci 14: 1073–1088, 1997.[Web of Science][Medline]
Velte TJ, Yu W, and Miller RF. Estimating the contributions of NMDA and non-NMDA currents to EPSPs in retinal ganglion cells. Vis Neurosci 14: 999–1014, 1997.[Web of Science][Medline]
Volgyi B, Xin D, Amarillo Y, and Bloomfield SA. Morphology and physiology of the polyaxonal amacrine cells in the rabbit retina. J Comp Neurol 440: 109–125, 2001.[CrossRef][Web of Science][Medline]
Werblin FS. Regenerative amacrine cell depolarization and formation of on-off ganglion cell response. J Physiol 264: 767–785, 1977.
Werblin FS and Dowling JE. Organization of the retina of the mudpuppy, Necturus maculosus. II. Intracellular recording. J Neurophysiol 32: 339–355, 1969.
Wong RO and Collin SP. Dendritic maturation of displaced putative cholinergic amacrine cells in the rabbit retina. J Comp Neurol 287: 164–178, 1989.[CrossRef][Web of Science][Medline]
Wong-Riley MTT. Synaptic organization of the inner plexiform layer in the retina of the tiger salamander. J Neurocytol 3: 1–33, 1974.[Medline]
Yang C-Y, Lukasiewicz P, Maguire G, Werblin FS, and Yazulla S. Amacrine cells in the tiger salamander retina: morphology, physiology, and neurotransmitter identification. J Comp Neurol 312: 19–32, 1991.[CrossRef][Web of Science][Medline]
Yang XL, Gao F, and Wu SM. Non-linear, high-gain and sustained-to-transient signal transmission from rods to amacrine cells in dark-adapted retina of Ambystoma. J Physiol 539: 239–251, 2002.
Yang XL and Wu SM. Signal transmission from cones to amacrine cells in dark- and light-adapted tiger salamander retina. Brain Res 1029: 155–161, 2004.[CrossRef][Web of Science][Medline]
Zenisek D and Matthews G. Calcium action potentials in retinal bipolar neurons. Vis Neurosci 15: 69–75, 1998.[CrossRef][Web of Science][Medline]
Zhang J and Wu SM. Immunocytochemical analysis of cholinergic amacrine cells in the tiger salamander retina. Neuroreport 12: 1371–1375, 2001.[CrossRef][Web of Science][Medline]
This article has been cited by other articles:
|
|
 |
|
  R. F. Miller Cell Communication Mechanisms in the Vertebrate Retina The Proctor Lecture Invest. Ophthalmol. Vis. Sci., December 1, 2008; 49(12): 5184 - 5198. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| Visit Other APS Journals Online |
