JOURNALOF NEUROPHYSIOLOGY Vol. 68. No. 3. Scptcmber 1992. Yrinred

in C’.S.A.

Relationship Between Receptive and Dendritic Field Size of Amacrine Cells in the Rabbit Retina STEWART

A. BLOOMFIELD

Department

oj’Ophthalmology,

SUMMARY

AND

New York University IMedical Center, New York, New York 10016

CONCLUSIONS

2. Intracellular recordings were obtained from 40 amacrine cells in the isolated, superfused retina eyecup of the rabbit. Cells were subsequently labeled with horseradish peroxidase for morphological identification. Many of these cells displayed dendritic morphology consistent with that of amacrine cells described in prior anatomic studies, including starburst, A 17, AII, and DAPI-3 cells. 2. The center receptive field of amacrine cells was measured with a 50- or 95-pm-wide, 6.0-mm-long rectangular slit of light that was displaced along its minor axis (parallel to the visual streak) in increments as small as 3 pm. The extent of the receptive field was calculated as the total distance over which the displaced slit could evoke a center response. Area summation of amacrine cells was measured with concentric spots of light with increasing diameters centered over the cell. 3. For a single amacrine cell, the receptive field size was comparable to the extent of its dendritic arbor. For the total population of amacrine cells, there was a strong, linear relationship between receptive field and dendritic field size. The receptive fields were, on average, 27% larger than the corresponding dendritic arbors, but this discrepancy can be accounted for entirely by tissue shrinkage associated with histological processing and a small imprecision of the light stimuli. Area summation measurements were consistent with those of receptive fields and were also related linearly to the dendritic field size of cells. 4. These findings indicate that even when the slit of light was placed at the distal edges of the dendritic arbor, synaptic inputs activated there were propagated effectively to the soma and recorded by microelectrodes placed there. In addition, amacrine cells were capable of summating synaptic inputs distributed throughout the entire arbor. 5. These results are inconsistent with the findings of prior computational modeling studies of passive, dendritic current flow in Al 7 and starburst amacrine cells that synaptic inputs on distal dendritic branches are isolated electrically from the soma and that these branches form autonomous, functional subunits. 6. The majority of amacrine cells encountered displayed lightevoked and/or spontaneous action potentials. These action potentials often took the form of high-amplitude somatic and lowamplitude dendritic spikes. On average, spiking amacrine cells showed considerably larger dendritic fields than nonspiking amacrine cells. In f’act, all amacrine cells with arbors >436 pm, which formed 45% of the total population, displayed spike activity. These data suggest that spike generation is a feature common to most medium- and large-field amacrine cells, permitting active propagation of synaptic currents over distances often > 1 mm. 7. Thus the discrepancy between the present electrophysiological data and those of prior modeling studies may be explained by the fact that synaptic inputs to most amacrine cells are not propagated passively within the arbor but rather by an active, regenerative process.

INTRODUCTION

The amacrine cells are laterally oriented interneurons that subserve a number of complex synaptic interactions in the proximal retina. In the mammal, it has been estimated that ~40 morphological classes of amacrine cell exist, based on differences in dendritic architecture, retinal distribution, and neurotransmitters used (Kolb et al. 198 1; Massey and Redburn 1987; Vaney 1990). Inherent to this classification is the concept of structure-function associations, suggesting that a correspondingly large number of physiological or functional classes of amacrine cells exist as well. Unfortunately, due mainly to technical considerations, there have been relatively few studies of the physiology of amacrine cells in the mammalian retina. However, the available data suggest that mammalian amacrine cells do, indeed, display a rich variety of physiological response properties (Dacheux and Raviola 1986; Nelson 1982; Nelson and Kolb 1985; Raviola and Dacheux 1987), including complex trigger features such as orientation sensitivity (Bloomfield 1991). An alternative method employed to study the function of amacrine cells is computational modeling of synaptic current flow. Rall’s ( 1959, 1964) formulations first showed directly how the dendritic geometry of a cell affects the generation of its response properties. Assuming constant specific membrane and internal resistivity, Rall showed that the branching pattern of dendrites strongly influences the propagation and integration of synaptic currents that, in turn, shape a cell’s somatic and axonal responses. In the 1980s Rall’s models were used by Ellias and Stevens ( 1980) to examine passive current flow within the dendritic arbor of Al7 amacrine cells in the cat retina and by Miller and Bloomfield ( 1983) to study the passive cable properties of the cholinergic starburst amacrine cells in the rabbit. Interestingly, both of these studies concluded that portions of the dendritic trees of these cells are isolated electrically from the soma and one another. For Al7 amacrine cells, the electrical isolation of dendrites derives from the large varicosities found throughout the arbor acting to limit axial current flow, whereas the relative thinness of the primary dendrites of starburst amacrine cells serves to isolate distal processes. Thus these models suggest that the dendritic arbors of these cells consist of autonomous, functional subunits capable of performing numerous computations simultaneously. Further, for at least these two types of amacrine cells, responses recorded with soma-placed microelectrodes should not reflect accurately the events occurring within the dendritic tree. In fact, only proximally located synaptic 711

Downloaded from www.physiology.org/journal/jn by ${individualUser.givenNames} ${individualUser.surname} (132.174.255.223) on October 4, 2018. Copyright © 1992 American Physiological Society. All rights reserved.

S. A. BLOOMFIELD

712

inputs should be recorded at the soma, whereas more distally placed synaptic events should be hidden completely. Assuming that retinotopic space is mapped linearly onto these amacrine cells, then their receptive fields should be considerably smaller than the extent of their dendritic arbors. To test this hypothesis, I compared the dendritic and receptive field size of several different morphological types of amacrine cells in the rabbit retina, including A 17 and starburst amacrine cells. Inconsistent with the conclusions of the modeling studies, the present results show a remarkable correspondence between the size of the dendritic and receptive fields of single amacrine cells. In contrast to prior studies (Bloomfield and Miller 1986; Miller and Bloomfield 1983; Nelson and Kolb 1985; Raviola and Dacheux 1987), I found that A 17 and starburst amacrine cells generate both spontaneous and light-evoked action potentials. In fact, with the exception of some small-field cells, all amacrine cell types were capable of generating action potentials. Thus one explanation for the discrepancy between the present results and those of prior modeling studies is that synaptic currents within the dendritic trees of amacrine cells may not be propagated passively, but actively by a regenerative process. METHODS

FIG. 1. Schematic diagram illustrating the 2 principle types of light stimulation used. A: displaced slit of light used to measure receptive field size. Flatmount view of a camera lucida drawing of a starburst amacrine The generalmethodsusedduring this study havebeendescribed cell and slit stimuli. The rectangular slit of light was first placed over the previously (Bloomfield and Miller 1982). Briefly, adult Dutch- center of the cell by alignment with the tip of the microelectrode. Adjustbeltedrabbits( 1S-3.0 kg) wereanesthetizedwith an intraperito- ments were made to this center position as dictated by cell response. The neal injection of ethyl carbamate(2.0 g/kg) and a local injection slit was then displaced in either direction in increments as small as 3 pm and presented at new positions over the dendritic arbor. B: concentric of 2% lidocainehydrochloride to the eyelidsand surroundingtis- small spots of light used to measure area summation. Smallest spot (50 pm sue.The eye wasthen removed under dim, red illumination and diam) was aligned with slit and the tip of the microelectrode. Larger conhemisected.The vitreoushumor wasremovedwith an ophthalmic centric spots of light were then presented while constant light intensity was spongeand the resultantretina eyecup mountedin a superfusion maintained.

Preparation

chamber.The chamberwasmountedwithin a light-tight Faraday cageand superfusedwith a modified Amesmedium(Sigma) with 25 mM sodiumbicarbonateand 10 mM glucoseat a flow rate of 30 cc/min. The superfusatewasmaintainedat 35°C with oxygenation and pH 7.4 provided by bubblingwith a gaseous mixture of 95Yooz-5% cop

Light stimulation Two 100-Wquartz-iodide lampsprovided light for a dual beam optical bench. Light intensity could be reduced (7.00 log units with calibrated neutral density filters placed in the light path of both beams.The maximum n-radiance(0.0 log) of both beams wasequalizedat 0.980 mW cmm2.The beamswerecombinedby a collectingprismand focusedonto the vitreal surfaceof the retina eyecupby meansof a final focusinglens.The bottom beamprovided smallconcentric spotsof light 50 pm-6.0 mm in diameter anda 50-or 95-pm-wide,6.0-mm-longrectangularslit of light that could be rotated and displacedalong its minor axis in stepsas smallas 3 pm. The top beamalsoprovided annular stimuli. All stimuli were alignedvisually with the microelectrodetip usinga dissectingmicroscopemountedwithin the Faradaycage.This was performedbefore retinal impalementto ensurethat stimuli were alignedcloselywith the center of cells’receptive fields. To assess the receptive field size of a cell, the slit of light was moved along its minor axis (parallel to the visual streak) in discrete stepsin both directionsfrom the central position ( Fig. 1A ). The location at which the slit of light failed to evoke a center l

responsewasviewed asthe edgeof a cell’s receptive field. Often, cellsdisplayedan antagonisticsurround responseas the slit was moved beyond this edge.The spatial extent and strength of the antagonisticsurround will influence the a cell’scenter receptive field size.However, becauseit is impossibleto quantify this influence, it was not factored into the receptive field measurements. Near the edgeof the center receptive field, the slit wasmoved in relatively smallsteps(aslow as3 pm) to provide a moreaccurate measure.The total extent of a cell’sreceptivefield wasthen taken asthe edge-to-edgedistancethe slit traversed in both directions from the central location: this value wasthen rounded off to the nearestmultiple of five. Often, it wasnecessaryto adjustthe center positionof the slit to produce the largestcenter response.This wasprobably due to a misalignmentof the light stimuli with the tip of the microelectrode. The concentric spotswere then realignedwith this new center position. For measurementof a cell’sareasummation,the concentric small spotsof light were presentedsequentiallywith light intensity maintained at 1.5-2.0 log units above threshold (Fig. 1B). The amplitude of the center responsewasthen plotted asa function of spot size(seeFig. 3B). The saturatingresponse amplitude wasfirst determinedfrom the graph, and the nonsaturating data points were then fit with a line usingthe least-squares method. The intersectionof this line with the saturatingresponse amplitudewasconsideredthe areaover which the cell wascapable of summatingthe center response;this value wasrounded off to the nearestmultiple of five.

Downloaded from www.physiology.org/journal/jn by ${individualUser.givenNames} ${individualUser.surname} (132.174.255.223) on October 4, 2018. Copyright © 1992 American Physiological Society. All rights reserved.

AMACRINE

CELL

RECEPTIVE

Of course. the quality of the light stimuli was an important factor in the accuracy of the receptive field measures. Thus accurate measurement of the dimensions of each stimulus (spot, annulus, or slit) was carried out. Stimuli were projected onto a dark sheet through 1.5 mm of water, mimicking the experimental situation in which stimuli were presented onto the retinal surface lying within the superfusate. The dimensions of the stimuli were then measured with a micrometer observed under a light microscope with a total magnification of x400. Light scatter was found to be insignificant as assessed visually under the microscope. Even the smallest spot or slit stimuli (50 ,um across) presented at moderate intensities showed mimmal light scatter, estimated at r5 pm.

Electricul rcwrdings Intracellular recordings were obtained from amacrine cells with glass microelectrodes fashioned from triangular glass tubing with wall lengths of 1.2 mm and wall thicknesses of 0.22 or 0.34 mm. Electrodes were filled at their tips with 10% horseradish peroxidase (HRP) ( Boehringer-Mannheim, Grade I) dissolved in 0.5 M potassium acetate-O. 1 M tris( hydroxymethyl)aminomethane (Tris) buffer ( pH 7.3) and then backfilled with 4 M potassium chloride. Final DC resistances of electrodes ranged from 150 to 500 MQ. After physiological characterization of a cell, HRP was iontophoresed into the neuron using a combination of sinusoidal (3 Hz, 0.3 nA peak-to-peak) and DC current (0.2 nA) applied simultaneously; this method facilitated the passage of HRP through the electrode tip without polarization. Recordings were displayed on an oscilloscope, recorded on magnetic tape, and digitized off-line for computer analyses. Only four neurons were labeled per retina to ensure that each labeled cell was matched correctly with its recorded responses.

One hour after labeling the final cell in an experiment, the retina was fixed immediately in a cold (4°C) solution of 1.5% gluteraldehyde- 1.5% paraformaldehyde in 0.1 M phosphate buffer (pH 7.3) for 12 min. The retina was then detached, trimmed, and fixed onto a gelatinized glass coverslip. After overnight incubation in phosphate buffer at 4OC, the retina was processed histologically using 3-3’diaminobenzidine with cobalt enhancement (Adams 1977) or benzidinc dihydrochloride ( Mesulum 1976) as chromagens. Labeled neurons were photographed in flatmount and drawn using a camera lucida microscope attachment. Drawings of cells were then digitized by scanner and entered into a computer for morphometric analysis and reconstruction. Measurements of dendritic field size were made along the axis parallel to the visual streak. This corresponds to the axis along which receptive field measurements were carried out. Tissue shrinkage associated with histological processing was not factored into the anatomic measures (see DISCUSSION).

All statistical analyses were made with Mann-Whitney c’test (Krauth 1983).

the nonparametric

RESULTS

CP// idm t$cu tion

This study includes intracellular recordings obtained from 40 amacrine cells. The dark resting potentials of these cells ranged from -38 to -76 mV. Recordings were limited to the area in and around the visual streak because the high density of amacrine cells there increased the number of cells

AND

DENDRITIC

FIELDS

713

encountered and impaled successfully; the most distal cell was located 4.8 mm from the optic disk. All cells were labeled with HRP, and their identification was based solely on morphological criteria. These criteria included the position of the soma and dendritic arbor within the proximal retina, clear lack of an axon, and, when possible, similarity to amacrine cells identified by prior morphological and/or histochemical studies. Of course, one problem in the identification was the possibility that a ganglion cell with an unlabeled axon could be identified mistakenly as an amacrine cell. However, because ganglion cell axons are emitted from the soma or proximal dendrite, it is my experience that axons are visible even in poorly labeled ganglion cells in which even the proximal dendrites are barely discernib le Further, except for starbu rst-b amacrine cells, the somas of all cells studied here were found within the proximal region of the inner nuclear layer (INL). Displaced ganglion cells are relatively rare within the rabbit retina and may not exist within the visual streak (Hughes 1985). Taken together, these findings all but exclude the possibility that some of the TABLE

1.

of‘amacrine

Cell Type

Physiological c&s Dendritic Field Size, Pm

AI1 AI1 AI1 AI1 AI1 DAPI-3 DAPI-3 DAPI-3

35 53 44 39 40 133 112 97

Small-field Small-field Small-field Small-field Small-field Starburst-a Starburst-a Starburst-a Starburst-b Starburst-b Medium-field Medium-field Medium-field Medium-field Medium-field Medium-field

81 112

Al7 Al7 Al7 Al7

Large-field Large-field Large-field Large-held Large-field Large-field Large-field Large-field Large-field Large-field Large-field Large-field

70 69 191 285 413 376 470 363 309 457 296 381 436 401 948 588 907 811 561 710 652 1,035 882 1,207 1,060 969 1,446 I ,03 1 661 561

and morphological

Receptive Field Size, Pm 55 75 65 45 60 150 145 95 110 235 90 100 275 350 510 410 590 410 635 315 555 440 505 1,100 680 1,235

1,020 700 925 730 870 1,030 1,620 1,510 1.030 1,265 845 700

properties

Area Summation, Pm

120

300 555 535 420 615 620 325 1,205 910 1,310 830 975 890 1,400 1.335 1,640 1,050 675

Spiking? No No No No No No No No No No No No Yes Yes Yes No Yes Yes Yes No Yes Yes No Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes Yes

Downloaded from www.physiology.org/journal/jn by ${individualUser.givenNames} ${individualUser.surname} (132.174.255.223) on October 4, 2018. Copyright © 1992 American Physiological Society. All rights reserved.

714

S. A. BLOOMFIELD

amacrine cells studied here were displaced ganglion cells with unlabeled axons. A large number of cells showed dendritic architectures that matched those of amacrine cells identified in prior morphological and/or histochemical studies and were classified accordingly. Other cells were arbitrarily separated into three classes: I ) small-field amacrine cells with dendritic arbors of O-249 pm, -3) medium-field cells with dendritic arbors of 259-499 pm. and 3) large-field cells with dendritic arbors 2500 pm. Physiological and morphological data for these cells are provided in Table 1.

A 175pm

750 pm

1000 pm

10mV

The Al7 amacrine cells in the rabbit retina produced a characteristic on-center light-evoked response. This consisted of a fast transient depolarization at light onset with multiple action potentials riding at the top and a sustained depolarizing phase for the duration of the light stimulus. A rather large hyperpolarization was produced at light offset (Fig. 2 ). Occasional spontaneous action potentials with large and small amplitudes were generated in the dark as

1250 pm

R Y

16 s .g

l5 14-

2

13-

12[rfz ll5 a

5 5

3 2

lo-

r\

-2lOpm

1 450 pm

#+p+kjv$

570pm

r

IlOmV

sec

-450 pm

-570pm I

1

I

I

Intracellular recording from an A 17 amacrine cell in the rabbit retina. At the to11 is the response to full-field illumination; light intensity = --4X) log units. lint below record is the light trace indicating the onset and offset of the light stimulus. At the bottom a 95pm-wide, 6.0-mm-long slit of light was presented at various positions across the arbor of this cell. The number to the lc;li of each trace indicates the distance the slit was displaced from the center position ( at the soma). Light traces at the bottom of the figure indicate the onset and offset of the light stimuli. This cell responded to the slit of light displaced over an extremely long distance of - 1.100 pm (550 pm in either direction ). Intensity d of slit stimuli = -4.0 log units. FIG. 2.

0

e

I

250

I

1

I

500

750

1000

Stimulus

Diameter

(Ctm)

I

1250

1500

FIG. 3. Measurement of area summation of the same Al7 amacrine cell whose responses are illustrated in Fig. 2. A: responses to concentric small spots of light centered over the soma. Light intensity = -4.0 log units. Values to the left of each trace indicate diameter of spot stimulus. B: plot of the data illustrated in A. The maximum response was taken as the peak response at light onset. The maximum or saturating response was 15.4 mV and is indicated by the dotted line. A line (- - -) was fit to the nonsaturating data points using the least-squares method. The intersection of the 2 lines indicates the area over which the cell could summate its response. Area summation = 1,205 pm. The area summation of all amacrine cells was measured by this method.

-3Opm

210pm

l a

98-

7!

90 pm

0

well; the generation of action potentials is a response feature not seen previously in responsesof A 17 cells [cf. Nelson and Kolb 1985; Raviola and Dacheux 1987 (see DISCUSSION)]. Figure 2 shows the responsesof an A 17 amacrine cell to the presentation of a 95-irn-wide, 6.0-mm-long slit of light at various positions across the retinal surface. The value to the lc-‘ffof each trace indicates the distance the slit was displaced from the central position. A large-amplitude center response was evoked by the slit, even when it was moved 450 pm from the center position. In fact, it was not until the slit was -570 pm off center that the cell became unresponsive. The total extent of the center receptive field of this cell was extremely large, covering a distance of 1,100 pm. The hyperpolarization seen at light offset was not surround mediated. Note that the off-hyperpolarization maintained a fairly constant amplitude asthe slit was repositioned within the first 2 10 pm but was reduced when the slit was moved farther off center. In addition, whereas the hyperpolarization was evoked by the small spots of light that stimulated only the center receptive field, spots of light with larger diameters were incapable of eliciting the response [ Figs. 2 (top) and 3A]. No antagonistic surround

Downloaded from www.physiology.org/journal/jn by ${individualUser.givenNames} ${individualUser.surname} (132.174.255.223) on October 4, 2018. Copyright © 1992 American Physiological Society. All rights reserved.

AMACRINE

CELL RECEPTIVE

responses were apparent in recordings from this or three other Al7 amacrine cells encountered during this study. Concentric small spots of light of increasing size were used to calculate the area summation of this cell (Fig. 3A ). A saturating response of 15.4 mV was produced by spots with diameters of > 1,250 pm. In Fig. 3B, these data are graphed, and the saturating response is represented by the dotted line. A line was fit to the nonsaturating data points using the least-squares method (---), and the intersection of the two lines reveals the area over which the cell could summate its response. The intersection point corresponds to an area summation of - 1,205 pm, consistent with the receptive field size of this cell measured with the displaced slit of light. This neuron was stained with HRP, and a flatmount view camera lucida drawing is presented in Fig. 4. The morphology of this cell is characteristic of A 17 or S 1/S2 amacrine cells described in both rabbit and cat retinas (Nelson and Kolb, 1985; Raviola and Dacheux, 1987; Sandell and Masland, 1986; Vaney, 1986). This cell has a small cell body - 11 pm diam from which extremely thin [approaching 0.1 pm (Fig. 4, inset)], wavy dendrites emerge and branch radially. The dendrites display numerous swellings, l-4 pm diam, throughout their extent. In fact, the arbor appeared under the microscope as numerous beads “floating” in space because the thin, interconnecting dendritic branches were barely visible under low power. The dendrites descend through the inner plexiform layer (IPL) and stratify within the most proximal strata adjacent to the ganglion cells. The arbor is fairly symmetrical and extends -950 pm along the axis parallel to the visual streak (horizontal across the

AND DENDRITIC

715

30 pm

-30 pm

60 ,um

-60 pm

90 pm

-90 pm

120pm

-120jdm 10mV 1 set

150pm

-150pm

180pm

-180pm

210pm

-210 pm I

20pm

FIELDS

I

I

I

FIG. 5. Intracellular recording from a starburst-b amacrine cell in the rabbit retina. Conventions are the same as in Fig. 2. Top: response of the cell to full-field illumination. Light intensity = -4.0 log units. Bottom: responses to a 50-pm-wide slit of light presented at various positions across the cell’s dendritic arbor. The on-center response of this cell was evoked by the slit displaced 1205 pm in either direction; receptive field = 4 10 pm. When the slit was moved more distally, a surround response was evoked as evidenced by the burst of spikes at light offset in the responses shown in the two bottom panels. Light intensity = -4.0 log units.

page). The shaded bar at the hottowz of the figure represents the size of the receptive field as measured with the slit of light displaced along this same axis (Fig. 2). Clearly, the dendritic and receptive fields of this cell correspond closely in size, indicating that a center response was evoked even when the slit was placed over the distal margins of the dendritic arbor.

Receptive field = 1100 pm ~~~--_-._.---~_l.ll--.lll-l~l-

._...I. _-......-.- ~-

FIG. 4. Flatmount view camera lucida drawing of A 17 amacrine cell whose responses are illustrated in Figs. 2 and 3. The full dendritic arbor of this large-field cell can be seen in this drawing, extending 950 pm along the horizontal axis. Imc~t : higher-magnification drawing of portion of the arbor showing the frequent varicosities and thinness of the interconnecting dendritic segments. The shaded bar at the bottom shows the extent of cell’s receptive size, 1,100 pm, so that a comparison with the cell’s dendritic field can be made easily. In this and all subsequent drawings, the visual streak runs horizontallv across the page.

Starburst amacrine cells Two types of starburst amacrine cell responses, on- and off-center, were encountered during this study. The characteristic response of on-center starburst cells consisted of a transient depolarization at light onset with a burst of spikes riding at the top and a large-amplitude, sustained depolarization for the stimulus duration [Fig. 5 (top)]. Unlike A 17 cells, there was no hyperpolarization evoked by the offset of the light stimulus but rather a small oscillation in the declining depolarization. Spontaneous spiking was also evident in the dark at a frequency of 2-3 Hz. Like Al7 amacrine cells, spiking has not been seen previously in re-

Downloaded from www.physiology.org/journal/jn by ${individualUser.givenNames} ${individualUser.surname} (132.174.255.223) on October 4, 2018. Copyright © 1992 American Physiological Society. All rights reserved.

716

S. A. BLOOMFIELD

sponses of starburst amacrine cells (Bloomfield and Miller 1986; Miller and Bloomfield 1983). A 50-pm-wide displaced slit of light of used to calculate the size of the receptive field of this starburst cell [Fig. 5 (bottom)]. The slit evoked an on-center response when displaced up to -200 pm from center in either direction, corresponding to a center receptive field of 410 pm. When displaced by 2 10 pm, the slit evoked a surround response composed of a train of spikes evoked at light offset. This cell was labeled with HRP, and a flatmount view drawing is provided in Fig. 6. The dendritic morphology of this cell is typical of starburst-b amacrine cells (Famiglietti, 1983a; Tauchi and Masland, 1984). The soma is -9 pm diam and is displaced within the ganglion cell layer. The symmetrical dendritic arbor consists of fine, wavy segments that branch dichotomously with little overlap and show numerous swellings in the distal regions. The arbor stratifies within a narrow band in sublamina-b of the IPL. The shaded bar at the bottom of Fig. 6 represents the extent of the cell’s receptive field, which closely approximates the size of its dendritic arbor. Figure 7 shows the response of an off-center starburst amacrine cell. The response to full-field illumination (top) consists of a depolarization and train of spikes at light offset with no concomitant change in baseline potential during presentation of the stimulus. (The response of off-center starburst amacrine cells was somewhat variable with other cells displaying sustained light-evoked hyperpolarizations.) There was some spontaneous spiking seen in the recording, but it was very sporadic, with a frequency ~0.5 Hz. A 50pm-wide slit of light was used to measure the size of the center receptive field. As the slit was moved distally (i.e., t 150 pm), the spiking disappeared, leaving a depolarization with duration of - 1 s. When the slit was displaced further, to 180 pm off center, the center response was replaced by a surround response evidenced by a burst of spikes at light onset ( bottom left). This cell was labeled with HRP, and a camera lucida drawing is provided in Fig. 8. This cell shows the classic

60 pm

-60 pm

90 pm

-90 pm

150pm

-15Opm

180pm

-180 pm

I

L

I

I

15mV 1 set

7. Intracellular recording from a starburst-a amacrine cell in the rabbit retina. Conventions are the same as in Fig. 2. Top: response of the cell to full-field illumination, Light intensity = -3.5 log units. Bottom: a %-pm-wide slit of light was presented at various distances from the soma. The off-center response of this cell was evoked when the slit was moved over a distance of 350 pm. When the slit was positioned more distally, a surround response was evoked, as evidenced by the burst of spikes at light onset (bottom left). Light intensity = -3.5 log units. FIG.

dendritic morphology of a starburst-a amacrine cell. In contrast to the starburst-b cell described above, the soma of this cell lies within the proximal edge of the INL and the dendrites stratify within a restricted band in sublamina-a of the IPL. Once again, the receptive field of this cell, represented

1

50 pm

50 pm

Receptive field = 410 pm Recbptive field = 350 pm

6. Flatmount view camera lucida drawing of horseradish peroxidase ( HRP)-labeled starburst-b amacrine cell whose responses are illustrated in Fig. 5. The cell body of this cell was displaced within the ganglion cell layer. The dendritic arbor of this cell is 363 pm across its longest axis. The shaded bar at the bottom represents the size of the cell’s center receptive field, which is comparable in size to the dendritic arbor. FIG.

FIG. 8. Flatmount view drawing of HRP-labeled starburst-a amacrine cell whose responses are illustrated in Fig. 7. The dendritic arbor of this cell extends 285 pm along the horizontal axis. The receptive field of this cell is 350 pm, as measured by the slit of light displaced along the same axis (shaded bar).

Downloaded from www.physiology.org/journal/jn by ${individualUser.givenNames} ${individualUser.surname} (132.174.255.223) on October 4, 2018. Copyright © 1992 American Physiological Society. All rights reserved.

AMACRINE

CELL RECEPTIVE

by the shaded bar, matched closely the size of its dendritic arbor. Thus the data from starburst amacrine cells, as shown for A 17 cells, indicate that even when light stimuli were presented over the distal margins of their dendritic arbors, the synaptic inputs activated there were propagated effectively to and recorded by soma-placed microelectrodes. The receptive and dendritic fields of many other types of amacrine cells in the rabbit retina were also examined. Data from some well-known morphological types of amacrine cells are provided below.

AND DENDRITIC

FIELDS

distal

proximal

AI1 amacvine ceh The AI1 amacrine cells displayed one of the most unique and characteristic responses, making for relatively easy identification of these cells even before visualization with HRP. At the top of Fig. 9 is the response of an AI1 amacrine cell to a centered, 50-pm-wide slit of light. The response consists of a small, oscillating transient depolarization followed by a rising plateau phase and a large amplitude hyperpolarization at light offset. The hyperpolarization was often larger in amplitude than the light-evoked depolarization. As the slit was displaced off center [Fig. 9 (bottoms)], the entire response including the hyperpolarization was reduced in amplitude, suggesting that the hyperpolarization was not manufactured by a surround mechanism. In fact,

10pm

1

20pm fl

-2Opmp

25pm $umvr\

-25 pm

1

10mV OSeC

9. Intracellular recording from an AI1 amacrine cell in the rabbit retina. Conventions are the same as in Fig. 2. Tup: response to a 50-pmwide slit of light centered over the cell. Bottom:the slit was repositioned by the distance indicated by the value to the Ze@of each trace. The center receptive field of this cell was extremely small, covering only 55 pm. When the slit was positioned at -t45 pm off center, a surround-mediated hyperpolarization was evoked. FIG.

1Opm Receptive field = 55 pm FIG. 10. Flatmount view camera lucida drawing of horseradish peroxidase (HRP)-labeled AI1 amacrine cell whose responses are provided in Fig. 9. The dendritic arbor of this cell was bistratified, and detailed drawings of the arbor within 2 planes in the distal and proximal inner plexiform layer (IPL) are provided. This is a small-field cell with a dendritic field size of only 35 pm. The extent of the cell’s receptive field is indicated by the shaded bar at bottom.

the receptive field was only 55 pm across, and a clear, sustained hyperpolarizing surround response was evoked when the slit was moved more distally. Five AI1 amacrine cells were encountered during this study, and none showed evidence of light-evoked and/ or spontaneous spiking. The responses seen during this study are consistent with those described previously for AI1 amacrines in cat and rabbit retinas (Dacheux and Raviola 1986; Nelson 1982). This cell was labeled with HRP and displayed morphology characteristic of the small-field AI1 amacrine cell found in both cat and rabbit retinas [ Dacheux and Raviola 1986; Famiglietti and Kolb 1975; Kolb and Famiglietti 1974; Mills and Massey 199 la; Vaney et al. 199 1 (Fig. lo)]. This cell has a relatively large cell body, - 10 pm across, and a bistratified dendritic field extending 35 pm along its major axis. The portion arborizing in the proximal IPL consists of relatively fine branches that stratify adjacent to the ganglion cells. The distal arborization is formed by thicker branches emitting lobular appendages that ring the soma. In addition, some of these lobular appendages were apparently emitted from the proximal arborization. The shaded bar shows the receptive field size relative to the cell’s dendritic arbor. Although both are relatively small in size, the receptive field is -60% larger than the cell’s arbor. Although this discrepancy may reflect physiological factors such as electrical coupling, it is probably due to technical artifacts such as tissue shrinkage associated with histological processing and an imprecision of the light stimulus. The small absolute size of the receptive field of AI1 sizes make the measurements more vulnerable to such artifacts (see below and DISCUSSION). DAPI-3 amacrine cells Recently, Vaney ( 1990) and colleagues described a second type of small-field, bistratified amacrine cell the rabbit

Downloaded from www.physiology.org/journal/jn by ${individualUser.givenNames} ${individualUser.surname} (132.174.255.223) on October 4, 2018. Copyright © 1992 American Physiological Society. All rights reserved.

718

S. A. BLOOMFIELD

retina. This cell was labeled by the fluorescent dye DAPI and showed positive for glycinergic immunocytochemistry. Figure 11 shows the response of a so-called DAPI-3 amacrine cell to presentation of a 50-pm-wide slit of light. When the slit is centered over the cell (top), the response consists of a small, transient depolarization at light onset followed by a sustained, depolarized phase for the stimulus duration. At light offset, there is a slow repolarization to the dark membrane potential with a small oscillation similar to that seen in on-center starburst amacrine cell responses. Although the response of this cell shows a great deal of noise, no clear evidence of spiking was evident in this recording or in those of two other DAPI-3 amacrine cells encountered. As the slit was moved distally, the response was reduced until a sustained hyperpolarizing surround response was evoked when the slit was 90 pm off center (Fig. 11, bottom). The center receptive field of this cell extended over a distance of only 160 pm. A camera lucida drawing of this HRP-labeled cell is provided in Fig. 12. The soma of this cell is only 8 pm diam and lies within the proximal border of the INL. The dendritic arbor is bistratified and extends for - 135 pm along an axis parallel to the visual streak. The proximal arbor consists of thin, curving branches that show numerous small swellings throughout as well as larger bouton-like processes 54 pm diam. The distal arbor shows far fewer small swellings and

. proximal

25 pm Receptive field = 160 pm FIG. 12. Camera lucida drawing of DAPI-3 amacrine cell whose responses are illustrated in Fig. 1 1. Detailed drawings of the bistratified dendritic arbor in 2 planes in the IPL are provided. Shaded bar at the bottom: extent of the receptive field as measured from the data illustrated in Fig. 11.

only occasional larger swellings limited to distal branches. As shown in Fig. 12, the receptive field of this cell is comparable to the extent of its dendritic arbor. Receptive .field size versus dendritic .field size

IOpm

30 pm

-30 pm

50 pm

10mV

-50 pm

-70 pm

70 pm 90 pm

-gOpm I

1

v I

1

Intracellular recording from a DAPI-3 amacrine cell in the rabbit retina. Conventions are the same as in Fig. 2. Top: response of the cell to a 50-pm-wide slit of light aligned with the microelectrode tip and centered over the cell. Bottom: the slit was moved in 20-pm increments away from the center position. The center response of this cell was evoked when the slit was moved 180 pm in either direction. More distally, the slit evoked a surround-mediated hyperpolarization. Light intensity = -4.0 log units. FIG.

1 1.

In total, receptive and dendritic field sizes were compared for 38 amacrine cells. It should be emphasized that these measurements were both made along the axis running parallel to the visual streak. Figure 13A summarizes these data showing the remarkable correspondence between the size of these neurons and the extent of their receptive fields (r = 0.98). The ratio of receptive field to dendritic field size for the entire population was 1.27 t 0.2 1 (SD). This discrepancy can be explained by experimental errors introduced by tissue shrinkage as well as the thickness of the light stimuli (see DIscussIoN). A similar linear correspondence was seen between cells’ area summation values and dendritic field sizes [r = 0.9 1, n = 20 (Fig. 13 B)]. The slight reduction in linearity seen for this second relationship with arbor size may be explained by the fact that the spots of light were limited to only 13 preset sizes and that each step increase in spot diameter corresponded to a 66- 100% enlargement. In contrast, the displaced slit of light was moved in steps as small as 3 pm, thus providing a far more accurate measure. Nevertheless, for the entire population, there was a strong correspondence between the area summation measured with concentric small spots and the receptive field size measured with the displaced slit of light [r = 0.89 (Fig. 14)]. For single cells, the ratio of receptive field size to area summation was 1.13 t 0.27 (n = 16), indicating a strong correspondence between the two measures. Interestingly, this result contradicts findings that A 17 amacrine cells in the cat retina have large receptive fields but extremely small area summations [ Nelson and Kolb 1985 ( see DISCUSSION)] .

Downloaded from www.physiology.org/journal/jn by ${individualUser.givenNames} ${individualUser.surname} (132.174.255.223) on October 4, 2018. Copyright © 1992 American Physiological Society. All rights reserved.

AMACRINE

z g 8 v) E *f !j lg $ $

CELL

RECEPTIVE

1.7 1.6 1.5 1 l4 1.3 1.2 1.1 1 0.9 0.8 0.7 0.6 0.5 0.4

AND

DENDRITIC

719

FIELDS

1.7 1.6 1.5

T

0

1.4 l

0

1.3 1 z

1.2

g 5 .% E E

l*l

l

0

0.9 I 0.7 -

z

0.6 -

l

l 0

0.8

$ a

l

1

l

l

0

l e

0.5 0.4 0.3 -

0

l

0.2 0.1 -

l

0

III 0

1.6

II 0.6

11

11

0.8

1

11

11

1.2

1.4

1.6

Receptive Field Size (mm)

l

B

1.5

III 0.4

0.2

FIG. 14. Scatter plot showing relationship between receptive field size and area summations of amacrine cells.

z la29 E w 5

l

l.l-

l

l-

0

ls 0.9E

0.8 -

0= g $

0.7 0.6 0.5-

le l

l

0 l

0

.o

0.4 -

l

0.3 -

0

I

10 mV

.

1 set

I

0.2 O.l-

.

B

0 0

0.2

0.4

0.6

Dendritic

0.8

1

1.2

1.4

Field Size (mm)

FIG. 13. A: scatter plot of the relationship between the receptive field size of amacrine cells measured with a displaced slit of light and their dendritic field size. Sizes of receptive and dendritic fields are taken from measurements made along the axis parallel to the visual streak. B: scatter plot of relationship between area summations of amacrine cells measured with concentric spots of light and dendritic field size.

Spiking in amacrine cells As shown above, the majority of amacrine cells studied here displayed light-evoked and / or spontaneous action potentials (see Table 1). It was relatively easy to differentiate action potentials from slow synaptic potentials. Lightevoked spikes showed durations r5 ms and often rode at the top of the slow depolarizations (Fig. 15 ). Spontaneous spikes were usually > 10 mV and considerably larger in amplitude than spontaneous slow potentials. Further, these spikes also showed durations r5 ms, whereas the time constant of the slow potentials was 21 order of magnitude greater.

Ikid

C

1lOmV

I’\

I

1 i

If

/ I

2 msec I \

FIG. 15. Intracellular recordings from a large-field amacrine cell in the rabbit retina. A: responses to full-field illumination. Cell shows an oncenter response consisting of a rapid depolarization with numerous spikes generated at its peak that declines during presentation of the light stimulus. B: response at the peak of the depolarization at light onset shown in A, but displayed at a faster time base; this trace is taken from a different response than shown in A, but was evoked with the same stimulus. The cell shows large-amplitude spikes between 20 and 40 mV and small-amplitude spikes (-+ ) of +3 mV. C: response of the cell at the peak of the depolarization at light onset at an even faster time base. Five separate responses are superimposed. Durations of large-amplitude and small-amplitude (--t ) spikes are similar, with values of -2-4 ms. The small-amplitude spikes are similar to the dendritic spikes described in amacrine cells in amphibian retina.

Downloaded from www.physiology.org/journal/jn by ${individualUser.givenNames} ${individualUser.surname} (132.174.255.223) on October 4, 2018. Copyright © 1992 American Physiological Society. All rights reserved.

S. A. BLOOMFIELD

720

Figure 15 shows the response of a large-field, on-center amacrine cell to full-field illumination. The response is similar to that of several amacrine cells described above, consisting of a rapid depolarization at light onset with a burst of spikes riding at the top. When this burst of spikes is viewed at a faster time base (Fig. 15, B and C), the individual action potentials can be observed. Large-amplitude action potentials, presumably somatic in origin, are present with amplitudes of 20-40 mV and durations of 2-3 msec. However, potentials are also generated with smaller amplitudes of - 10 mV, but with the same durations of 2-3 msec (+ ). The extremely short durations of these potentials rules out the possibility that they are passively generated excitatory postsynaptic potentials ( EPSPs). In fact, these low-amplitude spikes are similar to the dendritic spikes found in amacrine cell responses from amphibian retina (Miller and Dacheux, 1976). Both the large- and small-amplitude spikes are abolished by tetrodotoxin (TTX), indicating that they are Na+-dependent and confirming their identification as regenerative potentials (Bloomfield, 1992). Although most amacrine cells showed a burst of spikes at light onset or offset, this burst was often short-lived, and any spontaneous spiking in the dark showed extremely low periodicity. However, a number of amacrine cells showed high-frequency spiking. Figure 16 shows an example of a cell that generates light-evoked and spontaneous high-frequency spiking. Light stimuli consist of concentric spots of light used for area summation measurements. Initially, the

175pm

350 pm

Recedve

field = 925 urn

FIG. 17. Camera lucida drawing of horseradish peroxidase (HRP)-labeled amacrine cell whose responses are shown in Fig. 16. The arbor ofthis cell consists of a few long, spiny processes showing only Znd-order branching. The dendritic field of this cell measures 7 10 pm along the axis parallel to the visual streak (horizontal across the page). The receptive field of this cell is 925 pm as measured with a 95rm-wide displaced slit of light. The shaded bar at the bottom represents the extent of the receptive field.

cell showed only low-frequency spiking (response to 175 pm spot), but after the membrane seal around the microelectrode improved, the cell hyperpolarized 25 to -68 mV and the spike frequency increased to - 12 Hz in the dark and to >60 Hz during presentation of light. Such high-frequency spiking has previously been thought to be exclusive to ganglion cells. However, this cell was injected with HRP and displayed a somatic and dendritic morphology identifying it clearly as a large-field amacrine cell (Figs. 17 and 18). The cell body lies within the proximal INL, and the dendritic arbor is very large, with a diameter of >700 pm. The arbor consists of only a few dendritic branches that show only first-order branching (Fig. 17). A prominent feature of the cell is the numerous spines, some > 1 mm in length, found throughout the dendritic arbor (Fig. 18, +).

750 pm

1250pm

1750 .um 1OmV c

2500 pm

6200 pm

FIG. 16. Intracellular recording from a large-field amacrine cell in the rabbit retina displaying high-frequency light-evoked and spontaneous spiking. Responses to concentric spots of light; diameter of spot given to the /e/r of each trace. Light intensity = -3.5 log units.

FIG. 18. Flatmount view photomicrograph of portion ofarbor of horseradish peroxidase (HRP)-labeled amacrine cell illustrated in Fig. 17. Dendrites show numerous spines (+ ) that project almost perpendicular to the main branches. Calibration bar = 15 pm.

Downloaded from www.physiology.org/journal/jn by ${individualUser.givenNames} ${individualUser.surname} (132.174.255.223) on October 4, 2018. Copyright © 1992 American Physiological Society. All rights reserved.

AMACRINE

CELL

RECEPTIVE

14 12

Spiking

10

Cells

Z = 686 + 331 pm

8 6

L2

4

$2 5 til

0

f2 3 =4 6 8 10 12

0

0.2

0.4

0.6

0.8

1.0

1.2

1.4

1.6

Dendritic Field Size (mm) FIG. 19. Histogram showing the dendritic size distribution of all amacrine cells relative to their ability to generate action potentials. Spiking amacrine cells displayed dendritic field sizes distributed across the entire range, whereas nonspiking amacrine cells were all 5436 pm across. The difference between the average dendritic field size of spiking and nonspiking amacrine cells was statistically significant (P < 0.00 1).

This type of high-frequency spiking was found for a few other types of large-field amacrine cells, including the recently reported orientation-selective amacrine cells (Bloomfield, 199 1). In fact, it appeared throughout the study that the larger amacrine cells showed a preponderance of spike activity. Figure 19 shows a histogram comparing the dendritic field size of amacrine cells with their ability to produce action potentials. Although amacrine cells in all size bins displayed the ability to generate action potentials, every cell with a dendritic arbor >467 ,urn generated action potentials (Table 1). In contrast, nonspiking cells were distributed more toward the low end of the range of dendritic field sizes encountered. In fact, the difference in the distribution of dendritic field sizes of spiking and nonspiking cells was statistically significant (P < 0.00 1). DISCUSSION

Technical considerations The major finding of this study is that the dendritic and center receptive fields of amacrine cells in the rabbit retina are comparable in size. A strong linear relationship between receptive and dendritic field size held for all amacrine cells studied, which included > 10 different morphological types with dendritic arbors ranging in size from 19 1 to 1,446 pm. Although the receptive fields were, on average, 27% larger than the corresponding dendritic fields, this discrepancy can be accounted for almost entirely by shrinkage of cells associated with histological processing. Following the exact procedures used here, Bloomfield and Miller ( 1982) estimated the linear shrinkage of retinal tissue at 25%. Clearly, adjusting the size of dendritic fields upward by 25% would

AND

DENDRITIC

FIELDS

721

produce values that match almost exactly those for the corresponding receptive fields. An additional error that could cause an overestimate of receptive field size derives from the thickness of the slit of light along its minor axis (axis of displacement). In the ideal case, the slit stimulus would be infinitely thin and thus would activate currents only orthogonal to the axis of movement. Although the slits used were relatively narrow (50 or 95 pm), their finite dimension nevertheless introduced an error into the receptive field measurements. To illustrate this error, consider the extreme case in which the receptive field equals the thickness of the slit, i.e., 50 pm. As the slit is displaced along its minor axis to measure the receptive field, it must move 50 pm in either direction to pass completely over each hemifield. Assuming that a response is elicited when any portion of the slit covers the dendritic arbor, then the receptive field will be measured mistakenly as 100 pm. As a result, the error in the measure will be equal to the thickness of the slit or 50 pm, an error of 100%. In contrast, the error will be (5% using the same stimulus to measure a receptive field covering 1,000 pm. Clearly, the error introduced by the thickness of the slit will be more significant for measurements of the receptive field size of small-field cells. Consistent with this notion, the ratio of receptive field to dendritic field size for small-field amacrine cells was 1.40 t 0.26, significantly greater than the ratio for medium- and large-field cells (P < 0.05 ). Nevertheless, the strong linear relationship found between receptive field and dendritic field size for all cells suggests that the error in the measurements of small-field cells, which composed about one third of the total, was not large enough to distort the overall findings. Electrical coupling bet ween amacrine cells An alternative explanation for the relatively large discrepancy between the receptive and dendritic field size of smallfield amacrine cells relates to current spread through electrical junctions. Ultrastructural studies of cat and rabbit retinas have reported gap junctions between the proximal dendrites of AI1 amacrine cells as well as with cone bipolar cells (Dacheux and Raviola 1986; Famiglietti and Kolb 1975; Kolb 1979; Strettoi et al. 1990). Although the junction-permeant dye Lucifer yellow does not appear to cross these gap junctions (Vaney 1985 ), tracer coupling between AI1 amacrines has been demonstrated recently with the tracers biocytin and Neurobiotin (Massey 199 1; Vaney 199 1). These tracers have been used recently to demonstrate coupling between the small-field DAPI-3 amacrine cells in the rabbit as well (Vaney 199 1). The spread of current between coupled cells would presumably extend the receptive fields of these cells beyond the physical dimensions of their dendritic arbors, as demonstrated for the electrically coupled A- and B-type horizontal cells (Bloomfield 1992; Bloomfield and Miller 1982; Dacheux and Raviola 1982; Nelson et al. 1976). Although the receptive fields of the small-field amacrine cells were 40% larger than their dendritic fields, this discrepancy corresponds in absolute numbers to only 20-30 pm. Thus any electrical coupling between small-field amacrine cells results in an insignificant spread of current. Although the

Downloaded from www.physiology.org/journal/jn by ${individualUser.givenNames} ${individualUser.surname} (132.174.255.223) on October 4, 2018. Copyright © 1992 American Physiological Society. All rights reserved.

S. A. BLOOMFIELD

722

spatial component may be minor, electrical coupling between amacrines may play an important, yet subtle, functional role that is not revealed by gross measures of receptive field size. Area summation The measurements of area summation made with concentric spots of light were consistent with those of receptive field size made with the displaced slits of light and correlated well with dendritic field size. Thus, as viewed from the cell body, synaptic inputs activated throughout the dendritic arbor of amacrine cells can be detected and have the ability to summate algebraically. Although this finding may appear unremarkable, it contrasts with previous data from A 17 cells. Nelson and Kolb ( 1985 ) described A 17 amacrine cells in cat retina with large receptive fields of nearly 1 mm, consistent with the present findings, but with area summations of only 100-200 pm. Other than asserting species differences, it is difficult to account for the discrepancies between the data from cat and rabbit. However, the recording provided by Nelson and Kolb was relatively small in amplitude, possibly due to a shunt associated with electrode impalement. In my experience, recordings from injuried cells where large shunts are present tend to saturate easily because the shunt “clamps” the voltage to a low level. This may explain why recordings from A 17 cells in the cat show an inability to summate the synaptic inputs activated with larger stimuli. However, it is of interest to note that I have found a class of wide-field ganglion cells (dendritic diameters of 500-800 pm) in the rabbit with correspondingly large receptive fields but narrow area summations of only 100 pm. Dendritic current flow in amacrine cells The strong relationship found between the dendritic and receptive field size of amacrine cells indicates that even when the slit of light was placed over the distal margins of their arbors, synapses activated there were visible to somaplaced microelectrodes. That is, synaptic inputs to the distal arbor were propagated efficiently to the soma. These data are totally inconsistent with the results of prior modeling studies of amacrine cells, which suggested the idea of electrically isolated dendritic branches impeding the flow of distally generated synaptic currents to the soma (Ellias and Stevens 1980; Miller and Bloomfield 1983). Ellias and Stevens ( 1980) proposed that the numerous varicosities found throughout the arbor of Al7 amacrine cells served to isolate regions of the dendritic arbor. Using average resistivity values, they concluded that synaptic potentials would be attenuated completely within the distance covered by only six or seven varicosities. In fact, as seen in Fig. 4, six to seven varicosities cover a distance of only lOO150 pm within the arbor of A 17 amacrine cells in the rabbit. Thus, on the basis of the computational model, synaptic inputs generated beyond this distance should be isolated from the soma, resulting in a correspondingly small center receptive field. In contrast, the present results show that for A 17 cells, signals generated as distally as 600 pm could be seen at the soma. In fact, 50% attenuation did not occur

until the slit was displaced off center by 85% of the total dendritic field. Using Rall’s ( 1959) formulations, Miller and Bloomfield ( 1983) computed the passive dendritic current flow within the starburst amacrine cells in the rabbit. Their findings, similar to those for Al7 cells, suggest that dendritic branches are electrically isolated from one another and form autonomous input-output circuits capable of independent computations. Their calculations indicate that a synaptic potential generated as proximally as the first branch point is attenuated >25-fold as recorded by a soma-placed microelectrode. The present electrophysiological data dispute these findings, suggesting that even the most distally placed synaptic inputs do not show this level of attenuation at the soma. Although the present data do not eliminate the possibility of electrically isolated dendritic subunits, they do suggest that the conclusions drawn from the theoretical modeling studies were incorrect. There are, however, several possible explanations for the discrepancies between the results of these studies. As detailed below, these explanations focus mainly on assumptions of the computational models as well as on those made during the present study. 1) Measurements of the receptive field size with the displaced slit of light assumed that there is a topographical alignment of presynaptic inputs to an amacrine cell such that retinotopic space is mapped linearly across the dendritic arbor. That is, when the slit of light is placed over a particular portion of the cell’s arbor, it is activating receptors “looking” at a corresponding portion of visual space. One can imagine contrasting situations in which there is lateral displacement of presynaptic cells, tremendous convergence or divergence of inputs to an amacrine cell, and/ or synaptic inputs limited to one portion of the dendritic arbor. In these cases, the receptive field size of an amacrine cell, as measured here, says little about the propagation of synaptic currents within its arbor, but rather reflects the spatial properties of presynaptic neurons. However, anatomic data indicate that topographic maps are, indeed, aligned radially in the rabbit, thus validating the assumption (Hughes 197 1). Changes in the densities and sizes of receptors, horizontal cells, bipolar cells, amacrine cells, and ganglion cells occur uniformly over the retina, including the visual streak (Honrubia and Elliot 1969; Hughes 197 1; Vaney 1990), indicating no changes in convergence or divergence of visual information as it flows radially through the retina. In addition, available ultrastructural data indicate that, at least for starburst (Brandon 1987; Famiglietti 1983b) and A 17 (Sandell et al. 1989) amacrine cells, presynaptic inputs from bipolar cells (thought to provide the excitatory center receptive field) are distributed throughout the arbor. Finally, the present finding of a correspondence between center receptive field and dendritic field size provides additional strong evidence that visual space is mapped linearly over amacrine cell dendritic arbors. 2) The resistivity values used for the modeling studies may have been underestimates, resulting in mistakenly low values for dendritic space constants and receptive field sizes. For example, the studies assumed average values for specific membrane resistivity of 2,000 Q/cm2 for A 17 cells and 5,000 Q/cm2 for starburst amacrines (Ellias and Ste-

Downloaded from www.physiology.org/journal/jn by ${individualUser.givenNames} ${individualUser.surname} (132.174.255.223) on October 4, 2018. Copyright © 1992 American Physiological Society. All rights reserved.

AMACRINE

CELL

RECEPTIVE

AND

DENDRITIC

FIELDS

723

vens 1980; Miller and Bloomfield 1983). Recent data using whole-cell patch recordings suggest that neurons can have resistances 2 1 order of magnitude greater than assumed previously using conventional intracellular recording techniques (Coleman and Miller 1989 ) . Certainly, higher resistivity values would increase cells’ space constants and the distances covered by synaptic currents within the arbor, thereby increasing their receptive field sizes. 3) The measure of synaptic current flow made in the modeling studies was based on the application of a single voltage to the dendritic arbor. That is, the calculations assumed activation of only one synapse whose current propagates unaffected by other excitatory or inhibitory inputs. However, as pointed out by Miller and Bloomfield ( 1983 ), the situation would change dramatically if a large portion of the arbor were activated simultaneously. In this situation, the current flow from active to inactive dendritic branches would be significantly greater as signals summated and would no longer be divided by a large number of inactive branches. The question, then, is whether the slit stimulus used during this study served to activate a considerable portion of the dendritic arbor of an amacrine cell, so as to present a situation significantly different from that described by the modeling studies. Clearly, concentric spots of light activated many of the synaptic inputs to an amacrine cell, with the largest spot illuminating and activating synapses across and well beyond the entire receptive field. However, the slit of light illuminated a much more limited area. By comparing the area covered by the dendritic field and the 50-pm-wide, 6-mm-long slit of light moving over the field, I calculated that the slit covered from 14 (placed centrally over the soma) to 3% (placed at a peripheral edge) of the dendritic field of starburst amacrine cells. For A 17 amacrine cells, the coverage of a 95-pm-wide slit ranged from 10 to 0.3% as it was stepped across the cell’s dendritic field. Certainly more than one synapse was activated in these experiments. However, the calculations suggest that the slit activated only a small percentage of the total synaptic inputs to a given cell and thus did not depart significantly from the situation assumed in the modeling studies. In addition, the fact that the results using the slit were consistent with those using spots of light indicates that area coverage, and thus percentage of synaptic inputs activated, was not an important factor in the measurements.

cells. This would be necessary to ensure current flow over distances 5 1.5 mm. The finding that most small-field amacrine cells do not show spike activity is appropriate considering that passive propagation of synaptic currents would be sufficient to spread signals throughout the small arbor and to the soma (but see Wassle and Boos, 1992). If the discrepancy between the present physiological data and the results of the computational modeling studies is due to the presence of active channels, then blocking these channels should cause a concomitant reduction in receptive field size. Amacrine cell spiking has been shown to be blocked by TTX, suggesting that it is Na+-mediated (Barnes and Werblin 1986; Eliasof et al. 1984; Huba and Hofmann 1990; Miller and Dacheux 1976). I have recently confirmed that both somatic and dendritic spikes in rabbit amacrine cells are blocked by TTX (Bloomfield, 1992). Moreover, my preliminary data (to be published separately) show that application of TTX produces an -50% reduction in the receptive cell size of large-field amacrine cells, including A 17 cells, but has no effect on those of medium-field amacrine cells (Bloomfield, 1992). These data suggest, then, that Na+-mediated regenerative spiking plays a crucial role in shaping the receptive fields of some, but not all, amacrine cells. Clearly, additional work on this problem is called for. Spiking in amacrine cells has been reported in fish (Djamgoz 1986; Kaneko 1970, 1973; Kaneko and Hashimoto 1969; Murakami and Shimoda 1977; Sakuranaga and Naka 1985a,b; Teranishi et al. 1987) and amphibian (Ammermuller and Weiler 1988; Barnes and Werblin 1986; Eliasof et al. 1984; Marchiafava 1976; Marchiafava and Torre 1978; Marchiafava and Weiler 1982; Miller and Dacheux 1976; Werblin and Dowling 1969) retinas. However, much less is known about the physiology of amacrine cells in the mammalian retina. Although spiking amacrine cells have been reported in rabbit (Bloomfield and Miller 1986; Dacheux and Miller 198 1) and chicken (Huba and Hofmann 1990), the amacrine cells in cat (Kolb and Nelson 198 1, 1985) and the majority of amacrine cells in rabbit (Bloomfield and Miller 1986; Dacheux and Raviola 1986; Raviola and Dacheux 1987) were found not to generate spikes. In fact, prior recordings from starburst amacrine cells in rabbit (Bloomfield and Miller, 1986; Miller and Bloomfield 1983), as well as A 17 amacrine cells in both rabbit (Raviola and Dacheux 1987) and cat (Nelson and Kolb 1985 ), showed no evidence of spike activity. This discrepancy between the present results and those of prior studProbably the most obvious explanation for the discrep- ies can be explained most easily by the lability of spiking in ancy between the modeling results and the present electro- these cells. Amacrine cells have very small somas, usually physiological data is that most amacrine cells showed spike - 10 pm in diameter, and thus can be damaged severely by In fact, within a single class activity. Thus, in contrast to a basic assumption of the mod- microelectrode impalement. els, synaptic current is not propagated passively within the (e.g., starburst amacrine cells), some cells showed the abildendritic arbor of amacrine cells but rather by an active, ity to generate spikes and some cells did not, whereas others regenerative process. This would provide a mechanism by showed spiking only initially, which was then abolished preIn my experiwhich synaptic signals arising at distal dendrites are propa- sumably because of injury depolarization. ence, however, most amacrine cells showed poor or no spikgated effectively to and recorded at the soma. The discovery of dendritic spikes reinforces the concept of active propagaing during initial penetration and only after a few minutes tion of synaptic inputs within the arbors of these cells. In did high-amplitude spiking appear concomitant with hyperaddition, the correspondence between arbor size and the polarization of the membrane, presumably because of a ability to generate spikes suggests that active propagation is better seal around the microelectrode tip. The inability of prior studies to detect spiking in amacrine cells, then, can a feature common to medium- and large-field amacrine

Downloaded from www.physiology.org/journal/jn by ${individualUser.givenNames} ${individualUser.surname} (132.174.255.223) on October 4, 2018. Copyright © 1992 American Physiological Society. All rights reserved.

Downloaded from www.physiology.org/journal/jn by ${individualUser.givenNames} ${individualUser.surname} (132.174.255.223) on October 4, 2018. Copyright © 1992 American Physiological Society. All rights reserved.

AMACRINE

CELL RECEPTIVE

identified amacrine cells in the turtle retina. Proc. R. Sot. Land. Ser. B Biol. Sci. 2 14: 403-4 15, 1982. MASLAND, R. H., MILLS, J. W., AND CASSIDY, C. The functions of acetylcholine in the rabbit retina. Proc. R. Sot. Land. Ser. B Biol. Sci. 223: 121-139, 1984. MASSEY, S. C. AND REDBURN, D. A. Transmitter circuits in the vertebrate retina. Prog. Neurobiol. 28: 55-96, 1987. MESULUM, M.-M. The blue reaction product in horseradish neurohistochemistry: incubation parameters and visibility. J. Histochem. Cytothem. 24: 1273- 1280, 1976. MILLER, R. F. The neuronal basis of ganglion cell receptive field organization and the physiology of amacrine cells. In: The Neurosciences, Fourth Study Program, edited by F. 0. Schmitt and F. G. Worden. Cambridge, MA: MIT Press, 1979, p. 227-245. MILLER, R. F. AND BLOOMFIELD, S. A. Electroanatomy of a unique amacrine cell in the rabbit retina. Proc. Natl. Acad. Sci. USA 80: 3069-3073, 1983. MILLER, R. F. AND DACHEUX, R. F. Dendritic and somatic spikes in mudpuppy amacrine cells: identification and TTX sensitivity. Brain Res. 104: 157-162, 1976. MILLS, S. L. AND MASSEY, S. C. Labeling and distribution of AI1 amacrine cells in the rabbit retina. J. Comp. Neural. 304: 49 l-50 1, 199 1a. MILLS, S. L. AND MASSEY, S. C. Intracellular injection of Neurobiotin reveals dye coupling in the rabbit retina ( Abstract). Invest. Ophthal. Vis. Sci. 32, Suppl: 1130, 199 1b. MURAKAMI, M. AND SHIMODA, Y. Identification of amacrine and ganglion cells in the carp retina. J. Physiol. Lond. 246: 80 l-8 18, 1977. NAKA, K.-I. AND CHRISTENSEN, B. N. Direct electrical connections between transient amacrine cells in the catfish retina. Science Wash. DC 214: 462-464, 1981. NELSON, R. AI1 amacrine cells quicken time course of rod signals in the cat retina. J. Neurophysiol. 47: 928-947, 1982. NELSON, R., AND KOLB, H. A17: a broad-field amacrine cell in the rod system of the cat retina. J. Neurophysiol. 54: 592-6 14, 1985. NELSON, R., KOLB, H., FAMIGLIETTI, E. V., AND GOURAS, P. Neural responses in the rod and cone systems of the cat retina: intracellular records and Procion stains. Invest. Ophthal. Vis. Sci. 15: 946-953, 1976. RALL, W. Branching dendritic trees and motoneuron membrane resistivity. Exp. Neural. 1: 49 l-527, 1959. RALL, W. Theoretical significance of dendritic trees for neuronal input-

AND DENDRITIC

FIELDS

725

output relations. In: Neural Theory and Modeling, edited by R. Reis. Stanford, CA: Stanford Univ. Press, 1964, p. 73-79. RAVIOLA, E. AND DACHEUX, R. F. Excitatory dyad synapse in rabbit retina. Proc. Natl. Acad. Sci. USA 84: 7324-7328, 1987. SAKURANAGA, M. AND NAKA, K.-I. Signal transmission in the catfish retina. II. Transmission to type-N cells. J. Neurophysiol. 53: 390-4 10, 1985a. SAKURANAGA, M. AND NAKA, K.-I. Signal transmission in the catfish retina. III. Transmission to type-C cells. J. Neurophysiol. 53: 4 1l-428, 1985b. SANDELL, J. H. AND MASLAND, R. H. A system of indoleamine-accumulating neurons in the rabbit retina. J. Neurosci. 6: 333 l-3347, 1986. SANDELL, J. H., MASLAND, R. H., RAVIOLA, E., AND DACHEUX, R. F. Connections of indoleamine-accumulating cells in the rabbit retina. J. Comp. Neurol. 283: 303-3 13, 1989. TAUCHI, M. AND MASLAND, R. H. The shape and arrangement of the cholinergic neurons in the rabbit retina. Proc. R. Sot. Lond. Ser. B Biol. Sci. 223: 101-119, 1984. TERANISHI, T., NEGISHI, K., AND KATO, S. Functional and morphological correlates of amacrine cells in carp retina. Neuroscience 20: 935-950, 1987. VANEY, D. I. The morphology and topographic distribution of AI1 amacrine cells in the cat retina. Proc. R. Sot. Land. Ser. B 224: 474-488, 1985. VANEY, D. I. Morphological identification of serotonin-accumulating neurons in the living retina. Science Wash. DC 233: 444-446, 1986. VANEY, D. I. The mosaic of amacrine cells in the mammalian retina. Prog. Ret. Res. 9: 49- 100, 1990. VANEY, D. I. Many diverse types of retinal neurons show tracer coupling when injected with biocytin or Neurobiotin. Neurosci. Lett. 125: 187190, 1991. VANEY, D. I., GYNTHER, I. C., AND YOUNG, H. M. Rod-signal interneurons in the rabbit retina. II. AI1 amacrine cells. J. Comp. Neural. 3 10: 154-169, 1991. W~SSLE, H. AND Boos, R. Voltage- and transmitter-gated currents of AI1 amacrine cells in the slice preparation of the rat retina (Abstract). Invest. Ophthal. Vis. Sci. 33, Suppl. : 1172, 1992. WERBLIN, F. S. AND DOWLING, J. E. Organization of the retina of the mudpuppy Necturus maculosus. II. Intracellular recording. J. Neurophysiol. 32: 339-355, 1969.

Downloaded from www.physiology.org/journal/jn by ${individualUser.givenNames} ${individualUser.surname} (132.174.255.223) on October 4, 2018. Copyright © 1992 American Physiological Society. All rights reserved.