Proc. Nati. Acad. Sci. USAVol. 83, pp. 9774-9778, December 1986Neurobiology

Blockade of electrical activity promotes the death of mammalianretinal ganglion cells in culture

(natural cell death/patch clamp/tetrodotoxin/low Ca/high Mg/conditioned medium)

STUART A. LIPTONDepartment of Neurology, Children's Hospital, and Program in Neuroscience, Harvard Medical School, Boston, MA 02115

Communicated by Torsten N. Wiesel, August 7, 1986

ABSTRACT During the first 2 postnatal weeks, up to 50%of the ganglion cells in the mammalian retina normally die.Natural cell death may result from several factors, andelectrical activity has been proposed as one critical element.Recent experiments in vivo using intraocular injection oftetrodotoxin (TTX) have suggested that competition for sur-vival between ganglion cells from the two eyes is mediated bytheir degree of neuronal activity. In addition, the level ofactivity of afferents to the ganglion cells has been postulated tobe an important variable in determining their survival. Toinvestigate the mechanism of cell death engendered by alteredactivity, I studied the effect of electrical blockade with TTX (toblock sodium channels and thus action potentials) or lowCa/high Mg (to block transmitter release and hence synapticactivity) on individual neurons in vitro. For this purpose,identified retinal ganglion cells (RGCs) from postnatal ratswere maintained in culture. Unlike the previous in vivoexperiments, this approach permitted the exact concentrationof each agent to be controlled and the electrical activity of theRGCs to be recorded. In cultures from animals of postnatal day2-10 (P2-10), 1 ,IM TTX or 0.2 mM Ca/20 mM Mg resultedin the death of about 50% of the RGCs, representing those cellsthat had displayed spontaneous electrical activity, but did notaffect RGCs that lacked activity. However, the death of RGCswith spontaneous activity from P11-13 animals was not influ-enced by these drugs. These findings suggest that during acritical period ofdevelopment neurons become dependent uponelectrical activity, and the cessation of this activity can result intheir death. In addition, conditioned medium, collected fromcultures lacking TTX, rescued from death a large proportionof TTX-treated RGCs. Thus, the critical element for survivalmay represent modulation of a trophic factor related to thelevel of activity rather than electrical activity itself. Since, invivo, natural cell death occurs in neurons of similar type andage, and in the same proportion as that induced by the artificialblockade of electrical activity in culture, these findings may begermane to the mechanism of natural cell death in the retina.

During periods of normal development of the vertebratecentral nervous system (CNS), a large proportion of neuronsdie (1). For example, in the early postnatal period, about halfof the retinal ganglion cells (RGCs) of rats, hamsters, andhigher mammals naturally die (2-6). In the pigmented rat,normal cell death proceeds for the first 1½ weeks (7, 8). It hasbeen suggested that this loss of neurons is due to competitionbetween RGCs from the two eyes for their postsynaptictargets (6, 9). The results of several experiments have furthersuggested that the competition, and consequent cell death,may be mediated by neuronal activity since intraocularinjection of tetrodotoxin (TTX) can affect the extent of RGCsurvival (9-11). In addition to the effect of efferents, a

number of investigators have argued that cell death may beinfluenced by the degree of functional afferent innervation(12, 13). The mechanism for activity-dependent neuronalsurvival is not understood; one suggestion is that competitionfor trophic substances during synaptic stabilization may be acritical factor in determining the outcome of survival or death(14).

I studied the effect of electrical blockade by TTX or lowCa/high Mg solutions on ganglion cells in cultures of disso-ciated retinas from neonatal rats. This approach affordedcomplete control of the extracellular environment and facil-itated physiological recordings from the RGCs in order tomonitor their electrical activity. Identified RGCs in vitro aresusceptible to the effects of electrical blockade if they arespontaneously active at the time of onset of treatment and,even then, only if the cells are cultured from retinas of theappropriate age. These findings suggest that RGCs maybecome dependent upon electrical activity during a criticalperiod of development; cessation of activity during thisperiod results in their death. Furthermore, conditioned me-dium (CM) can prevent the demise ofRGCs exposed to TTX.Thus, neuronal survival may be related to a chemical factorwhose presence is dependent on electrical activity; blockadeof activity might prevent the release of this trophic factor intothe culture medium and result in cell death. A preliminaryreport of this study has appeared (15).

METHODS

Incubation of Cultured Retinal Cells with TTX. Retinas ofLong Evans rat pups were dissociated and cultured asdescribed (16). RGCs were identified with specific fluores-cent labels (16). A dosage of 1 gM TTX (Sigma) was chosensince this concentration has been shown to completely blockthe fast action potentials and Na current of RGCs (17). TTXwas added to most cultures at the time of plating. RGCs werescored for survival 24 hr later.

In another set of experiments TTX was added 10 hr afterplating; this protocol permitted patch-clamp recording fromindividual RGCs before treatment with TTX. The cells wereidentified by marking the bottom of the culture dish. Afterexposure to TTX for 14 hr, the survival ofthe identified RGCswas determined by observing each cell. Dead cells couldreliably be scored by the presence of swollen, vacuolatedmembranes producing an appearance as if they had "burst."Upon washing out the TTX, patch-clamp recording andfluorescein diacetate were used to confirm the state ofthe cellin multiple cases (n = 57); the result invariably agreed withthe prior assessment based on the physical appearance of thecell. In addition, cells were scored as dead if they werecompletely absent from their original location. Previous

Abbreviations: CM, conditioned medium, CNS, central nervoussystem; EPSP, excitatory postsynaptic potential; P, postnatal day;RGC, retinal ganglion cell; TTX, tetrodotoxin.

9774

The publication costs of this article were defrayed in part by page chargepayment. This article must therefore be hereby marked "advertisement"in accordance with 18 U.S.C. §1734 solely to indicate this fact.

Dow

nloa

ded

by g

uest

on

May

26,

202

1

Proc. Natl. Acad. Sci. USA 83 (1986) 9775

experiments using time-lapse video recording had shown thatthe cells did not migrate after a few hours in culture.As a control, sister cultures were maintained without TTX

for an equivalent period of time. Thus, death of RGCs fromcauses not related to the experimental paradigm could bemonitored. The cells were plated on several different sub-strates (16) [plain glass, collagen, poly(L-lysine), or Thy-1antibody, 2G12] to ensure that no one substrate artifactuallyaffected cell death in the presence of TTX.

Incubation of Cultures in Low Ca/High Mg. In otherexperiments retinal cultures were plated in low Ca (0.2 mMinstead of 1.8 mM) and high Mg (nominally 20 mM instead of2 mM). These concentrations were chosen since they areknown to block transmitter release and synaptic activity inthe retina (18). In three experiments the ratio of Mg:Ca waslowered from 100:1 to about 10:1 (2.41 mM Mg, 0.2 mM Ca)to maintain a constant concentration of divalent cations.

Incubation of RGCs in CM. CM was harvested from 24-hrretinal cultures obtained from postnatal day 7-8 (P7-8)animals. Then, freshly dissociated cells were plated in CMwith and without added TTX. As a control, CM was alsocollected from cultures initially grown in the presence ofTTX.

Electrical Recordings of RGCs. Recordings were madeusing standard intracellular pipettes as well as patch elec-trodes (17, 19). Intracellular pipettes were filled with 4 Mpotassium acetate, whereas patch electrodes for cell-at-tached and whole-cell recordings generally contained a po-tassium saline (in mM: KCl, 140; MgCl2, 2; CaCl2, 1; EGTA,1.5; Hepes/NaOH, 10, pH 7.2). Other patch recordings weremade in the cell-attached mode with a pipette solution similarto that in the bath (in mM: NaCl, 137; NaHCO3, 1; Na2HPO4,0.34; KCl, 5.4; KH2PO4, 0.44; CaCl2, 2.5; MgSO4, 0.5;MgCl2, 0.5; Hepes/NaOH, 5; glucose, 22.2, pH 7.2, with0.001% phenol red indicator).

RESULTSSolitary and Clustered RGCs. About 10-15% of the RGCs

were found on the second day in culture as solitary neurons(for pictures, see ref. 16). The remainder of the RGCs waslocated among small clusters of other retinal cells. A typicaldish contained =20 solitary RGCs and 150 "clustered"RGCs.

Spontaneous Action Potentials in Clustered Cells. Record-ings with intracellular and patch electrodes from solitary andclustered RGCs have shown that both types are electricallyexcitable. Action potentials, which were reversibly blockedby short-term additions of 1 AxM TTX, were elicited whendepolarizing current was injected (17). However, the pres-ence of spontaneous spikes was quite different between thepopulations of solitary and clustered RGCs. Solitary RGCsonly very rarely, if ever, had spontaneous activity. Incontrast, about 50% of the clustered RGCs displayed spon-taneous spikes. Fig. la shows a whole-cell voltage recordingobtained with a patch electrode from a clustered RGC thathad spontaneous action potentials and a smaller positivepotential, presumably an EPSP; this finding is consistent withthe presence of synaptic activity. In Fig. lb a solitary RGCwas recorded from for 20 min without encountering sponta-neous activity.Using patch electrodes to record from the neuronal surface

(20), the presence or absence of spontaneous action poten-tials could be ascertained with minimal disturbance from eachof the RGCs in a culture dish. However, because of reuse ofthe electrode in these experiments, relatively loose sealswere formed (on the order of 10-100 MW) and were thus oflower resolution than whole-cell recording. One loose patchrecording is illustrated in Fig. 2, which reveals spontaneousaction potentials from a clustered RGC. Fig. 3 summarizes

50 mV

10 s

a b

FIG. 1. Transmembrane voltage recorded with a whole-cell patchelectrode from a clustered and a solitary RGC. (a) The clusteredRGCdisplayed spontaneous action potentials and an excitatorypostsynaptic potential (EPSP, marked by arrow). (b) The solitaryRGC had no spontaneous spikes or postsynaptic potentials. Potas-sium aspartate substituted for KC1 in the pipette solution.

the findings of the loose patch recordings (n = 2250). About50% of the RGCs in clusters had spontaneous action poten-tials after 1 day in culture. On the other hand, virtually noneof the solitary cells had spontaneous activity. The addition of1 ,M TTX blocked action potentials in all of the clusteredRGCs.TTX and Cell Death. RGCs were plated in medium con-

taining 1 ,M TTX, whereas sister cultures had normalmedium without TTX. After 24 hr in culture, the RGCs werecounted. Fig. 4a shows that there were fewer clustered RGCsif the medium contained TTX. In fact, the ratio ofthe numberof RGCs in TTX versus control was 1:2. This finding isconsistent with the notion that 50% more clustered RGCs haddied in the dishes containing TTX than in the controls. Therighthand column of Fig. 4a indicates that after washing outthe TTX, the remaining clustered RGCs had no spontaneousaction potentials. The inference is that the RGCs in clustersthat had had spontaneous spikes were the cells that had diedwhen exposed to TTX for 24 hr. The surviving clusteredRGCs still had excitable sodium channels since sodiumcurrents could be demonstrated with depolarization duringwhole-cell recording after TTX washout (not illustrated).To directly demonstrate that RGCs from P7 animals with

spontaneous electrical activity were susceptible to dyingbecause ofTTX, six additional experiments were performed.In these experiments 10 hr after plating in normal medium,loose patch recordings were obtained on 433 clustered RGCs(Table 1). The location of each cell was marked so it could beidentified later. After recording, the medium was changed toinclude TTX. Within 14 hr, all 79 RGCs that had displayed

40 pA10 ms

-105.65 pAVH = 0 mV

FIG. 2. Cell-attached current recording obtained with a patchelectrode that was loosely sealed to a clustered RGC. No holdingpotential was applied to the pipette (VH = 0), and the amplitude ofthe first action current (denoted by the +) is listed above VH.Spontaneous activity in these records consisted of one or two burstsof action currents per s, often with clustering of three to five eventsover 100 ms.

Neurobiology: Lipton

Dow

nloa

ded

by g

uest

on

May

26,

202

1

Proc. Natl. Acad. Sci. USA 83 (1986)

0

n-

0

-2

C I

0

U. cell s n sol itary clIu stered cell sclusters cells ill TTX

FIG. 3. Histogram of the number of RGCs displaying spontane-ous action potentials (APs) after 24 hr in culture. On each of the threesubstrates, about half of the clustered RGCs had spontaneous activity(n = 1680). Virtually none of the solitary RGCs had spontaneousaction potentials in recordings lasting at least 5 min (second column,n = 570); the term "virtually" is used because on occasion as thepatch electrode first contacted the surface of the cell (n = 12), a fewaction potentials were noted, presumably because of transient injuryand consequent brief depolarization. Incubation or microperfusion in1 ,uM TTX blocked the spontaneous spikes of the clustered RGCs (n= 227). Ab, antibody; PLL, poly(L-lysine). Data were from 11experiments, P7-8 animals. Error bars in this and subsequent figuresrepresent 1 SD.

spontaneous activity prior to TTX died (line 4, Table 1);meanwhile, for those without activity only 20% died (line 5,Table 1). Control experiments showed that about 20% of theclustered RGCs died in normal medium whether they pos-sessed spontaneous activity or not (lines 2 and 3, Table 1).Furthermore, the patch-clamp technique itself was not re-

a 1

o 0 .8 DX glas's

0 PLL0.6-

C0.4-0

~02 0

0.0total number APs after

TTX washoutFor Clustered Cells -P7, P8

C1.0-

1-~ ~~~-e#°- 0.8- ztiEL ~~~~~~~~~~~~~~~~~~~~~~~~.1.

U 0~~~~~~~~~~~~~~~~~~~~~~~~~~~~~.6--......F~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~.........:.....

c) 0.64 - ......-.-X_T ,

. ..X0.2- I ~~~~~. +l

0 .4 I ....g.+... , ....-.-........

.e:..'|....0.2

Table 1. TTX-induced death of clustered RGCs as a function ofspontaneous electrical activity

Fraction of RGCs % of RGCsExperiment surviving 14 hr surviving

Without patch recording, P7 72/91 78+ activity, P7 56/74 76- activity, P7 72/87 83+ activity, + TTX, P7 0/79 0*- activity, + TTX, P7 81/101 80+ activity, + TTX, P13 48/68 71- activity, + TTX, P13 44/60 73

Activity was assessed with a cell-attached patch pipette 10 hr afterplating the cultures, following which treatment with TTX was begun.Values represent the number of clustered RGCs pooled from sixexperiments with retinas dissociated from P7 rat pups and twoexperiments with P13 animals.*Significantly different (P < 0.001) from corresponding row withoutTTX by x2 test with the Bonferroni correction.

sponsible for this mortality because approximately the samepercentage of RGCs died in control medium if no electricalrecording was made (line 1, Table 1). These experimentsshow that RGCs with spontaneous activity were more sus-ceptible to TTX than those lacking activity. However, fromthese data it cannot be determined if the 20% mortality inculture under control conditions was related to normal celldeath or to differences in the in vivo and in vitro environ-ments.The effect of TTX on cell death was observed on RGCs

from P2-10 animals but not older. Fig. 4b illustrates theresults of three experiments from P13 rats. The ratio 6f

b

00

x

F-.0ci

total numberFor Clustered Cells - P13

do 1.0-

0 0.8-V)_ 0.6-

o 0.4-

n: 0.2-

1 1 i 3 total number °Owith length ofAge (days postnatal) process processesFor Clustered Cells For Solitary Cells - P7, P8

FIG. 4. Histograms of clustered (a-c) or solitary (d) RGCs in TTX medium for 24 hr as a fraction of those in control medium. (a) Total numberof living RGCs after TTX was 50% of control. None of the surviving RGCs after TTX exposure had action potentials (APs). The number ofRGCs scored on plain glass, poly(L-lysine) (PLL), and 2G12 [Thy-1 antibody (Ab)] was 3243, 1430, and 2954, respectively. (b) For older animalsthe ratio of surviving RGCs in TTX to control medium is nearly equal to unity, so there was no significant increase in the death of clusteredRGCs in TTX. For glass, poly(L-lysine), and 2G12 the number of cells scored was 592, 646, and 688, respectively. (c) Ratio of surviving RGCsin TTX/control versus age. Each bar represents three experiments using 2G12 substrate, but similar results were found on plain glass andpoly(L-lysine). P13 data were significantly greater than those of P2, P4, and P7 by an analysis of variance using an F test followed by Duncan'smultiple comparison of means (P < 0.01). (d) For solitary RGCs there was no difference between TTX and control in the total number of livingRGCs, the number with at least one process longer than the cell body diameter, and the mean length of processes per RGC. In control mediaon glass, poly(L-lysine), and 2G12, the number of surviving RGCs was 474, 228, and 384, respectively; the percentage growing processes was10, 44, and 81%; mean process length was 13, 42, and 87 ,um.

9776 Neurobiology: Lipton

Dow

nloa

ded

by g

uest

on

May

26,

202

1

Proc. Natl. Acad. Sci. USA 83 (1986) 9777

clustered RGCs surviving in TTX and control is not signifi-cantly different from 1, indicating the absence of an effect ofTTX under these conditions. Additional experiments thatfollowed the fate of individual RGCs after patch recordingproduced similar findings (bottom two lines, Table 1). Coin-cidentally, by P13 natural cell death in vivo has ceased in thepigmented retina (7, 8).

If there is indeed a period of susceptibility to blockade withTTX, then there should be some time dependence of theeffect. This premise was tested by varying the age of theanimals. Fig. 4c shows that the ratio of surviving clusteredRGCs in TTX versus control cultures increased in olderanimals (P11 and 13).On the other hand, solitary RGCs, the cells without

spontaneous spikes, were relatively unaffected by the addi-tion ofTTX to the cultures despite variation in the age of therat pups (P2-13). For example, in Fig. 4d a comparison ismade between solitary RGCs from P7-8 animals in TTXversus control medium. Since the ratio of TTX to controlresults is roughly unity, there was no difference in the numberof solitary RGCs surviving (first column), the proportion ofcells that grew processes (second column), or the total lengthof processes per cell (third column).Thus, when spontaneous activity in cultures from P2-10

animals was prevented by using TTX, a significant proportionof the RGCs in clusters died, but there was no untowardeffect on solitary RGCs. Clearly, the increased death asso-ciated with TTX was not an indiscriminate toxic effect; it wasspecific only for the cells that had spontaneous actionpotentials, which represented 50% of the RGCs in clusters.

Effect of Low Ca/High Mg. Blockade of spontaneoussynaptic activity and action potentials in clustered RGCs wasobserved during low Ca (0.2 mM)/high Mg (20 mM)microperfusion. Presumably, therefore, the difference be-tween clustered RGCs with and without spontaneous activityis synaptic connectivity. Incubation of retinal cultures fromP2-10 animals in low Ca/high Mg had an effect similar to TTXon increasing the death of clustered RGCs. Fig. 5a shows thatthe number of clustered RGCs was reduced about 50% in lowCa/high Mg medium compared to controls. Furthermore,when these cultures were then washed with the normalincubation medium, the remaining clustered RGCs did nothave spontaneous electrical activity. This observationstrongly suggests that the population of RGCs killed by lowCa/high Mg was the same as that killed by TTX. In addition,similar to the findings with TTX, the number of survivingclustered RGCs from the older P13 animals was unaffected bylow Ca/high Mg.

0

a01.0i nlas

Table 2. Surviving clustered RGCs after treatment with TTXand/or CM during the critical period of development

Treatment No. of RGCs % of controlControl 162 ± 23TTX (1 ,uM) 98 ± 17 60*CM 151 ± 15 93CM +TTX 145 ± 12 90

RGCs were counted after 1 day in culture with control medium orwith medium containing TTX and/or CM (100% by volume, notincluding the TTX). Values are the mean + SD; n = 4. The retinalcultures were obtained from P7 animals and plated on Thy-1antibody. Other cultures were plated on plain glass or poly(L-lysine)and yielded similar results.*Statistical comparisons were made with an analysis of varianceusing an F test followed by Duncan's multiple comparison of meansyielding: control = CM = CM + TTX > TTX; P < 0.01.

The first column in Fig. 5b shows that the low Ca/high Mgsolution did not adversely affect the number of solitary RGCsafter 1 day in culture-that is, the ratio of the number ofsurviving solitary cells in test and control medium was nQtsignificantly different from unity. This result is in agreementwith that obtained in TTX. Unlike the TTX effect, however,is the interesting finding that in 20 mM Mg/0.2 mM Ca agreater proportion of solitary RGCs grew processes; inaddition, the total length of processes per cell increased(columns 2 and 3, Fig. 5b). Certainly, high Mg/low Ca at aratio of 100:1 was not adversely affecting the solitary RGCs;if anything, they grew better. Compared to the 100:1 ratio,medium containing an -10:1 ratio of Mg/Ca did not com-pletely block synaptic activity and resulted in the death of anintermediate number of clustered RGCs. In contrast, thesolitary RGCs in the 10:1 solution displayed no difference inthe various parameters of process outgrowth compared tocontrol. Possibly, then, the reason for the increase in processgrowth by solitary cells in the 100:1 medium was the increasein total divalents, which affects surface charge and thusadhesion.CM Prevents Cell Death. When CM-treated cultures re-

ceiving TTX were compared to sister cultures receiving TTXin control medium, a striking difference was noted: CMappeared to protect the clustered RGCs from the increaseddeath observed with TTX alone (Table 2). In contrast,preliminary results suggested that CM harvested from dishescontaining TTX from the time of initial plating did not rescueclustered RGCs from death. This finding implies that elec-

b

For Clustered Cells - P7, P8

,0

I= prcs prcese

0 F

24-

C.)

0

*0 total number % with length of

M ~~~process processes

For Solitary Cells P7, P8

FIG. 5. Histograms of clustered (a) or solitary (b) RGCs in 0.2 mM Ca/20 mM Mg medium for 24 hr as a fraction of those in control medium.(a) Total number of surviving clustered RGCs after low Ca/high Mg was 50% of control. (b) Solitary RGCs in low Ca/high Mg did as well orbetter than control with respect to total number of survivors, percentage with processes longer than the diameter of the cell body, and meanlength of processes per RGC. Data were collected in the same experiments as in a. In control medium on glass, poly(L-lysine) (PLL), and 2G12[Thy-1 antibody (Ab)], the number of surviving solitary RGCs was 216, 352, and 148, respectively; the percentage of RGCs growing processeswas 23, 50, and 71%; the mean length of processes per RGC was 16, 43, and 111 ,m.

Neurobiology: Lipton

Dow

nloa

ded

by g

uest

on

May

26,

202

1

Proc. Natl. Acad. Sci. USA 83 (1986)

trical activity is necessary for the production of effective CM.Solitary RGCs were unaffected by CM.

DISCUSSIONThe experiments of the present study concern the role ofelectrical activity in the death of mammalian RGCs. Asubpopulation of cultured RGCs died within 24 hr of expo-sure to TTX or low Ca/high Mg medium if the cells were ata particular stage of development in which they eitherputatively possessed or were acquiring spontaneous electri-cal activity. The subset ofRGCs that was susceptible to theseagents formed clusters with other retinal neurons. Theseneurons presumably contacted the RGCs, based upon thepresence of postsynaptic potentials and spikes that wereblocked by transient microperfusion with low Ca/high Mg. Inthe experiments assessing RGC death, TTX or low Ca/highMg was added at the time of plating of these cultures; this factmay imply that the susceptible clustered RGCs die only ifthey do not develop electrical activity by way of their newlyformed synaptic connections. In other experiments TTX wasadded 10 hr after plating retinal cells from P7 animals. All ofthe spontaneously active RGCs that were identified, andsubsequently followed, died in TTX, whereas only about 20%of the inactive cells died. This finding directly demonstratesif electrical activity is not maintained, RGCs will die'. Takentogether these results suggest that during development RGCsbecome dependent upon electrical activity, and failure toacquire or maintain this activity is associated with mortality.This effect of electrical activity could be linked to somerelated event such as uptake or release of a chemical factor.Along these lines, the addition of CM to culture dishes

containing TTX resulted in increased longevity of a signifi-cant number of clustered RGCs. These experiments supportthe notion that a trophic factor may be present in retinalcultures without' TTX, and it may be this factor that isnormally secreted in toe presence of electrical activity that isimportant for cell survival during the critical period. If onlythe RGCs have TTX-sensitive action potentials in the mam-mali'an retina, it may be that RGC activity is specificallyneeded for the factor to be released. If the spikes of amacrinecells also have a TTX-sensitive component, then amacrinecell activity may be necessary for the presence of trophicfactor. Nevertheless, one could envision mechanisms where-by either neurons or glialcells produce the factor, dependingon the connectivity of the amacrine and RGCs in culture. Thetrophic factor in these experiments, however, cannot origi-nate in the more central target neurons of RGCs since thesenormal targets (superior colliculus and lateral geniculatenucleus) were not present in the culture dishes.

In fact, these results suggest rather convincingly that RGCdeath can be regulated locally (i.e., within the retina) andwithout target-dependent binocular interactions. It has beenreported that the dendrites of certain RGCs behave as if theymust normally compete with one another for inputs intrinsicto the retina in the inner plexiform layer (12, 13). Heretofore,there has been little direct evidence to link this dendriticcompetition to cell death or activity. The experiments re-ported in the present paper, however, could be interpreted tosupport the notion that presynaptic activity to the RGCs is animportant variable in determining cell death.An influence of electrical activity on cell death has also

been reported by Brenneman and colleagues on cultures ofspinal neurons. They found that the cytotoxic effect of TTXcould be antagonized by vasoactive intestinal peptide, ap-

parently working by way ofcAMP (21). My experiments alsosuggest that the effect of electrical activity may be mediatedby a soluble factor in the culture fluid; one interpretation isthat the level of this substance is related to the degree ofactivity and, possibly, in the critical period once a neuron iscontacted to form a synapse it becomes dependent on thisfactor. Further work will be necessary to uncover the factorresponsible for this protective effect in retinal cultures. Inany event, the observations of the present study in combi-nation with those of Brenneman et al. strongly suggest thatblockade of ongoing or developing electrical activity duringa critical period of development can lead to the progressiveloss of neurons.

In vivo, natural cell death occurs in a large number (about50%) of RGCs that are the same age as the cultured RGCsstudied here (7, 8). Whether the findings in culture, however,are a true reflection on the normal process of cell death in theCNS remains to be settled. Nevertheless, since the period ofsusceptibility to impulse blockade in vitro corresponds to theinterval of normal neuronal death in vivo and since, theproportion of RGCs dying normally is similar to that inducedby TTX or low Ca/high Mg, it is tempting to speculate thatcessation of spontaneous electrical activity is involved in thephysiological mechanism of natural cell death.

I thank Ursula Drager and Matthew Frosch for helpful discussionsand Paul Harcourt, Mary Hunt, and Micheal Phillips for technicalassistance. This work was supported in part by National Institutes ofHealth Grants EY05477 and NS00879 and by a Mary and AlexanderP. Hirsch Award from Fight For Sight, Inc., New York City.

1. Oppenheim, R. W. (1981) in Studies in DevelopmentalNeurobiology: Essays in Honor of Viktor Hamburger, ed.Cowan, W. M. (Oxford Univ. Press, New York), pp. 74-133.

2. Land, P. W. & Lund, R. D. (1979) Science 205, 698-700.3. Thompson, I. D. (1979) Nature (London) 279, 63-66.4. Finlay, B. L., Wilson, K. G. & Schneider, G. E. (1979) J.

Comp. Neurol. 183, 721-740.5. Insausti, R., Blakemore, C. & Cowan, W. M. (1984) Nature

(London) 308, 362-365.6. Rakic, P. & Riley, K. P. (1984) Nature (London) 305, 135-137.7. Cunningham, T. J., Mohler, I. M. & Giordano, D. L. (1982)

Dev. Brain Res. 2, 203-215.8. Perry, V. H., Henderson, Z. & Linden, R. (1983) J. Comp.

Neurol. 219, 356-368.9. Fawcett, J. W., O'Leary, D. D. M. & Cowan, W. M. (1984)

Proc. Natl. Acad. Sci. USA 81, 5589-5593.10. O'Leary, D. D. M., Fawcett, J. W. & Cowan, W. M. (1983)

Soc. Neurosci. Abstr. 9, 856.11. Cowan, W. M., Fawcett, J. W., O'Leary, D. D. M. &

Stanfield, B. B. (1984) Science 225, 1258-1265.12. Perry, V. H. & Linden, R. (1982) Nature (London) 297,

683-685.13. Cunningham, T. J. (1982) Int. Rev. Cytol. 74, 163-186.14. Hamburger, V. & Levi-Montalcini, R. (1949) J. Exp. Zool. 111,

457-502.15. Lipton, S. A. & Harcourt, P. (1984) Soc. Neurosci. Abstr. 10,

1081.16. Leifer, D., Lipton, S. A., Barnstable, C. J. & Masland, R. H.

(1984) Science 224, 303-306.17. Lipton, S. A. & Tauck, D. L. (1987) J. Physiol. (London), in

press.18. Masland, R. H. & Ames, A., III (1976) J. Neurophysiol. 39,

1220-1235.19. Hamill, 0. P., Marty, A., Neher, E., Sakmann, B. & Sig-

worth, F. J. (1981) Pflagers Arch. 39, 85-100.20. Forda, S. R., Jessell, T. M., Kelly, J. S. & Rand, R. P. (1982)

Brain Res. 249, 371-378.21. Brenneman, D. E. & Eiden, L. E. (1986) Proc. Natl. Acad.

Sci. USA 83, 1159-1162.

9778 Neurobiology: Lipton

Dow

nloa

ded

by g

uest

on

May

26,

202

1