LETTERS TO NATURE
released by high levels of emotional arousal can produce memory
impairment, Findings of animal experiments indicate that
memory storage is influenced by drugs affecting many neuro-
modulatory systems, including a-adrenergic systems!"
Studies to date have not investigated the selective involvement
of other systems in regulating emotionally influenced memory
ins human subject
Although clinical reports suggest that fblockers may induc,
memory loss in some patients", controlled laboratory investi-
‘gations have so far generally failed to find consistent effects of
Beblockers on memory". However, previous studies generally
investigated memory for relatively unemotional events and typ-
ically studied only short-term memory. Future research using
B-blockers with different affinities for §, and Bs receptors and
differential effectiveness in passing the blood-brain barrier
should clarify the role of the adrenergic system in regulating
long-term memory for emotional experienc
53. Mcoagy J. Lis Renato moon aa hey: Research
aes ee ara ne anne sia
Changes in reliability of
synaptic function as a
mechanism for plasticity
Charles F. Stevens & Yanyan Wang
Laboratory for Molecular Neurobiology,
Howard Hughes Medical Institute, Salk Institute,
10010 North Torrey Pines Road, L3 Jolla, Califoria 92210, USA
SyNarric transmission in the hippocampus is rather unreliable,
ations in synaptic strength seen in long-term potentiation (L,TP)*
and long-term depression” (L'TDY? We find that L-TP increases
i y, and LTD decreases it, both without a change
in the size of those postsynaptic currents that do occur. Thus LTD
is a functional inverse of LTP.
We have used “minimal stimulation’, a method for activating
only one or a few synapses: the stimulus intensity is reduced to
a level that produces a postsynaptic response of uniform latency
fand shape in a target neuron with a probability of less than
about 0.5. This method permits us to separate {40 components
‘of synaptic strength: reliability, the fraction of stimuli that pro-
704
duce & postsynaptic current, and potency. the average peak size
of the postsynaptic current when one does occur
Whole-cell recording” provides a signal-to-noise ratio that is
adequate for clearly distinguishing between a release failure and
4 success in almost all instances. ‘This is illustrated in Fig. |
‘hich reveals clear separation betwen transmitter release and
no release
Data for typical LTP experiments are illustrated in Figs ta
and 2a e. Separate averages were made (before and after the
induction of LTP) for all responses. all stimulation trials for
‘hich postsynaptic current was detected, and all records that
represented Failures in synaptic transmission (Fig. 2a). Further,
non-overlapping groups of about 25 sugcessve traces were wver
‘aged throughout the experiment, and failure probabilities were
estimated forthe same groups of traces: these group averages
reveal stable LTP following the tetanie stimulus (Fig. 2). and
2 sustained post-etanic decrease in the failure rate (Fig. 20)
‘The group average potency is roushly constant throughout the
experiment (Fig. 2), and the histograms of excitatory postsyn
aptic potential (eps..) peak amplitude for the control and LTP
periods (Fig. 2c) are not significantly different (Kolmogoroy
Smirnov test, P>0.1)
Tn this instance, stable LTFP is represented as an increase in
the average response. a decrease in the synaptic failure rate
and no appreciable change in the potency. As with the LTP
experiment, LTD is represented by'a decrease in the average
response size (Fig. a, am inerease in the failure rate (Fig. 30),
and no appreciable change in synaptic potency (Fig. 3a. d.e)
Although we find. as reported previously that LTP is difi-
cult to induce after about 20 min of recording in the whole-cell
mode. LTD seems not to "wash out’. That is, we could produce
LTDas late as 80 min after the statt of whole-cell recording, the
longest time tested. In six experiments, LTD reversed previously
established LTP (data not shown),
To compare data fram different experiments, we note thatthe
average peak eps. 1. is (by definition) the product of the
probability, thata release will cur and the average size «of
a response (when release does occur): 1/= i, the subscript j
is 0 before the induction of LTP or LTD and 1 after, The probe
ability ofa failure j,= (1 ~,) is determined directly from experi-
ment ass the average poteney 4). Synaptic plasticity Sis defined
asthe ratio Sr, ry Inplasticity were only a change in synaptic
reliability Qwith no change in poteney), then a de Would equal
I and S would be just the ratio of success probabilities
wii. Fay dy. A plot of S worsus the ratio inp would then
fall on a straight line with slope 1, and a plot ofthe ratio ays
would be constant (=I). Our experiments provide the data
reajuired for testing the extent to which plasticity is determined
by changes in reliability and changes in poteney as we estimate
di fy and yy diet
The unfilled circles in Fig, 4a and b represent data from 26
experiments hike the ones illustrated in Figs 1, 2 and 3, Cleary.
the ratio w/in Falls along the diagonal in Fig. 4, and the ratio
4 y's roughly constant in Fig. 4b. For these experiments, then
wwe conclude that the mechanism of LTP and LTD is a change
in synaptic reliability with no appreciable change in poteney
The experimental situation here isa little different from previ
ous investigations of LTP that activated a population of sye
fipses so that all or most failures are obscured. Because we have
activated fewer synapses than was usual for earlier studies, we
felt that we must demonstrate thatthe plasticity examined here
exhibits the same properties as standard LTP.
LTP is sometimes elicited by the “pairing” stimulation model
instead of by tetanic stimulation. In these “pairing” experiments
low-frequency stimulation is paired witha postsynaptic depokar
zation imposed on the cell". Data from fixe "palring’ exper
tents, plotted in Fig. 4e and d as unfilled triangles, conform to
the behaviour of synapses potentiated with tetanie stimulation
LTP is blocked by the presence of the N-methyl-p-aspartate
(NMDA) receptor antagonist AP-S", and by hyperpolartzation
NATURE - VOL 371 - 20 OCTOBER 1994FIG. 1 a, Amplitude of synaptic currents as a function of time during
the experiment to estimate the uncertainty in classifying traces as
release’ or ‘no release’. Each point was calculated by averaging the
curtent over a fixed window 2:ms long centred atthe time that core:
Sponded to the peak curtent. Diamonds are from traces on which
release wes apparent. and dots from traces without detectable release
(currents everaged over the same 2s window). Tetanic stimulation
was gen at the time indicated by the arrow, and produced LTP of
41708. Three typical traces of eurent asa function of time, ane without
release and two with release, are superimposed In the inset. The cali
bration bars in b inset apply to these traces. b, Amplitude histogram
for the data presented in 3. The filed histogram was derived from all
ofthe races, represented by diamonds in which the experimenter could
Setect a synaptic current, and the open bars from all of the rolease
failures (Gots in a). The smooth histogram was derived from traces
measured in the same way (and from the same experiment) except no
stimulus was present to evoke synaptic current. The ‘ailures histogram
(uniled bars) and the ‘no stimulus’ histogram (smooth tine) are not
Signiicantly diferent (Kolmogoroy-Smimov test, P>0.1), The inset
presents average current as function of time from (1) all data in the
files histogram, (2) all data in the untied bar histogram, and (3) all
ata fom the line histogram. We estimate from the overlaps of the
releases’ and “alures'nistograms, and from uncertainties in making
eosiors with visual classification and inthis and other experiments,
that less than about 5% of the races would have been misclassified
In-most experiments, the classification of traces as “elease’ or 'no
release’ was done visually by the experimenter so that every trace was
Sorutnized. Most classfation was done bling in the sense that the
experimenter didnot know if the records came from an LTP or an LTD
experiment or ifthe traces being classified were obtained betore or after
the condoning event (tetanus, pairing or low-frequency stimulation. In
about 20% of the instances a second experimenter classified the traces
Independently, clasifcations by two independent experiments agree
to within a few per cent. The calibration bars apply to the three inset
‘waces in a and b, Vertical bar, 3 pA: Rorizontal bar, 5 ms.
METHODS. Slices were prepared as in refs 16 and 27. Transverse hip-
pocampal slices (~400 ym thick) were obtained from rats (P1d-P26,
most P15-P24, Harlan Sprague-Dawley). The tempersture in the
recording chamber was 31.5 0.5 °C. The extracellular solution is com.
posed of (in mM): NaCl, 120; KCl, 38; NaH,PO., 1.25; NaHCO, 26;
MeCls, 1.3: CaCl,, 2-2.5, Pieratoxin (50 4M) was added to the bath
LETTERS TO NATURE
&
2300
Relative frequency
e.p.s.c amplitude (pA)
%
Solution in ~60%6 ofthe experiments. Electrodes for whole-cell recorcing
were filed with in mM) caesium gluconate, £30; CsCl, 5; EGTA, 0:5;
MgCl, 2: Na-ATP, 2; GTP, 0.2: NaCl, 5; HEPES’ 10; pH 7.25. When
using the perforated patch technique, the pipette solution also contains
4180 pg mi * nystatin. E1ps.cs were evoked bypassing current through
fa bipolar tungsten electrode at 0.25 Hz while holding the neuron's
membrane potential at ~70 mv. Tetanic stimulation consisted of three
bursts, each at 100 He witha 200 ms duration and an iterburst interval
(F105, that were applied while the target neuron’s membrane potential
‘was maintained at 20 mV. LTP was alsa induced by @ paring protocol
Uwith 25 stimu applied at 0.2 He while maintaining the neuron's mem
brane potential at 0 mv.
@
FG. 2 9, average records of caret 8 2 futon of Movs 5
time with current and time given by the calibration bars uw
forai races te lot toes were taken rom 4a
txperment sing heme pocecing te tetanus anc Foie
the righthand traces from @ final 10-min period 30 min
after the tetanus, The top traces are averages of synap-
tic currents (all events, N= 110 left, N= 450 right), the
middle traces aro averages of those records for which
no release was detected (failure, N=72 left, N=67
Fight), and the bottom traces are’ averages of records
{or which transmitter release was detected (release, N=
‘36 left, N~ 83 right. b, Synaptic strength as @ function
Of time (min) during the experiment demonstrating LTP
for a sample neuron. Average peak synaptic current,
including falures, was found by averaging across non-
‘overlapping grouss of successive Incvidual peak syn
‘apie currents and Is potted as a rato to the average
fof the five baseline points. The same time base is also
Used for c and d, Tetanic stimulation was presented,
Using the some stimulus svength, atthe time indicated
by the arrow (also in ¢ and a). The horizontal line ater
the tetanic. stimulation represents LTP of 2.18. c,
Release failure rate, estimated over the same groups
of successive individual aces as used forthe average
‘synaptic stength in b, 282 function of time. The failure
fate decreased from 0.6 to 0.2. 0, Average amplitude
(of the peak synaptic current for these tnals on which
‘lease was detected, as @ function of time: averages
‘ver the same groups of successive traces as used in
and c.e, Histograms presenting the reatve frequency
of peak elpssc. (when release occurred) before (bar,
N=18%) and after stable LTP has been established
(smooth line, N=339). The two histograms are not Si
nificantly different (Koimogorov-Smirnov test, P> 0.1).
NATURE - VOL 371 20 OCTOBER 1904
ol... |
°
os Fe Ty
$08 ere
bos
Boal tee yewd
20
imeimn)
| pee
ay E
| Eo
trices Poe coped g
:
°
10 20 30 40 50 ° 5 0
Time (min) eps. amplitude (pA)
705LETTERS TO NATURE
FIG, 3 a, Averages fr a typical experiment as for Fig a
2a All vents, N=150 lek, N= 180 rit; failure, N=
658 lof, N~ 108 right; release, N~82 lef, N72 right.
», Characteristics of responses for an LTO experiment
pitied in the same way as the LTP data in Fig. 2. Low
foqueney stimulation le indieated by bars in be, and
{TD of 0.51 incleated by the ine afer te low-frequency
stimulation. ¢, Failure rate inereases from 0.6 t0 0.77.
€, Histogram of peak synaptic currents before low-tre-
‘quency stimulation (open bars, N'=80) and after (sold
line, N=100). The amplitude histograms are not Sig:
nificantly different (Kolmogorov-Smirnov test, P> 0.1).
METHODS. LTO was inducee by apalyng stimuli at 1 He
for 10-12 min while maintaining the membrane poten-
Lal at ~TOmW. The magnitude of LTP or LTD was typ
‘cally determines between 15 to 35min afer the
conditioning event, Access resistance (25 to 30 M2)
and input resistance (100-300 MO) were monitored
‘rroughout experiments and only recordings with stable
‘access and input resstance were used for date analysis
10
os
18 FB
poo
Failure rate
e2oo,
38
‘release success or fallure was usually determined by
eye, rather than by setting an arbitrary threshoid, to
ensure that any changes in the smallest response size
were detected. A failure was determined by te following
o
10
10 20 30 40 50
Time (min)
20.30 40 80
Criteria: (2) the recorded Wace ts not eiscernbly diferent
from background noise level, (2) the time course oF
shape of recorded ‘ace Is not consistent with. the
establisned time course or shape for the epp.s.cs from
‘that neuron In some cells, a smal fraction of recorded
‘races (5%) could not be fairly judged as ether a fai:
Ute ora success and were rected. The distribution and
average of ep... amplitudes in our study are compat-
e.p.s.c. amplitude (ratio of bs!
g
Relative frequency
lle wth those of spontaneous miniature €.p... O°
‘single fore e:ps.c. in CAL reporteo previously". :AP-
5 (GRE), picrotonin (Aldrich) and nystain (Sigma).
°
10 20 30
alll sen
e.p.s.c. amplitude (pA)
40 50 °
Time (min)
FIG. 4 a, Ratio of success
Minimal stimalus
probablites (afte/before
‘conditioning » stimulation
consisting of tetanus, oF
low-frequency stimula
tion} as @ function of syn
apie strength rato fatter?
before) calculated fom
averages such a8. illus
‘rated in the top traces
(of Figs 2a and 3a. Un-
files cncies are trom 12
LIP synaptic strengin
ratio>1) and 14 UID
{synaptic Strength ratio
<1) experiments, Filled
ccioles. are from ATP
‘experiments, of the sort
ihustated in e ang f that
Release probability ratio
Response size ratio
Interleaved strong. and
minimal stimuli at te
‘some stimulation site in
‘the slice. b, Ratio fatter
before conditioning stimulation) of average peak e.p:.c. for tals on
whieh release vecured, obtained from averaged traces like those ius
trated at the bottom of Figs 2a and 3a. asa function of the synaptic
‘strength ratio (rom average traces like thase atthe top af Figs 29 and
3a). Symbols nave the same meanings as those in a.c, LTP experiments,
‘as 3. Diamonds are from 5 LTP experiments done in the presence of
AP-5 (93) or with the membrane potential maintained at ~80 mv
‘uring the tetanic stimulation (1~ 2). Unfilled tangles are from 5
‘experiments in which ‘pairing’ ater than tetanic stimulation was the
Conditioning event. Two ofthese experiments used weaky trong condi
tioning stmt fs in @and f), Uniled squares are from six experiments
that used perforated patch recording. d. Same type of plat as b with
‘symbo's having the same meaning as c.e, Peak synantic current as a
Tunetion of time during an LTP experiment with minimal stimulation.
‘Standard tetanic stimulation with nigh intensity just before the 20 min
time point Insets are averages (over 10 min) of traces from before and
30 min after the tetanic stimulation, witn curent and time calibrations,
ranges trom tals on which release was detected and dots for
°
o 1 2 3 0
706
1
Relative synapte strengths (S)
eps campitude (pA)
2 3
Time (min)
nowelease tra0es. f, AS for ¢, but €.9.5.cs evoked by strong stimulus
‘applied to the same stimulating electrode. No release failures were
Seen with the stiong stimulus. Note thatthe ordinate in ds 10 times
greater than tne ordinate inc. In these experiments, minimal ana strong
Stimuli were applied alternatively to the stimulation site; the stimulus
‘uration for minimal stimulation was reduced to 50 us and the stimulus
‘uration for strong stimulation was 100 ys with the same current a5
‘minimal stimulation.
NATURE - VOL 372 - 20 OCTOBER 1994of the postsynaptic cell during the conditioningevent"™"*", Data
from five experiments that used hyperpolarization or AP-5 10
block LTP, plotted as untilled diamonds in Fig. 4e and d. all
fall near S= 1. wy /wo= and ay /do= |. Thus, LTP is produced
tunder the conditions of our experiments by “pairing” as well as
tetanic stimulation, and LTP is blocked by hyperpolarization
and AP-S)
Although we are aware of no firm evidence that genuine LTP
requires the activation of synuptic population, one can imagine
some intersynaptic cooperative effect for which plasticity
‘mechanisms would be different from those that are invoked when
only one or two synapses are involved, as is the case for our
preceding experiments. To test this, we have done the Following
experiment on six cells, Ata single stimulation site, microscopic
‘and macroscopic stimuli are interleaved with a pattern of five
‘minimal stimuli (failure rate >0.5) followed by one strong stimu:
lus. After obtaining a baseline response size with this repeating
5/1 pattern, LTP was produced with the strong stimulus and
then the 5 weak/I strong stimulation pattern was resumed. Typ
ical LTP was produced in the population of synapses activated
by the strong stimulus, and LTP as we have measured it above
\was produced in the subset of synapses activated by the minimal
stimulus (Fig. 4e,.)
Data from these experiments (filled circles for tetanic stimul-
tion in Fig. 4a, , and two unfilled triangles for “pairing” in Fig
4e and d) show that when LTP is elicited in a population of
synapses, the modification of synaptic strength in the subset of
synapses activated by the weak stimulus occurs by a change in
synaptic reliability not potency
All of the experiments presented above used standard whoie-
call recording which can gradually change the concentration of
small intracellular molecules: for example, LTP is known to
‘wash out’ afier about 20 min, Could we have washed out some-
LETTERS TO NATURE
hat is required for increasing potency? To test this, we
did six LTP experiments using the perforated patch technique!
and found that, as for the other experiments. LTP involves @
change in reliability without modification of potency (squares
in Fig. de and a),
Although many different mechanisms are used to modify syn:
aaptic strength, we conclude that, under the conditions of our
experiments, LTP and LTD operate by’ modifying synaptic rei
ability. not poteney. Whatever the precise mechanism by which
it is achieved, the fact that symapses ai le and that
synaptic plasticity reflects modifications in reliability rather than
potency has implications for information processing. The brain
rust use a design that is fault tolerant and produces reliable
results when constructed with unreliable communications links
that modily the strength of connections by chan,
and not po!
eR ig ie seas ata
Cloning of an avermectin-
sensitive glutamate-gated
chloride channel from
Caenorhabditis elegans
Doris F. Cully, Demetrios K. Vassilatis**,
Ken K. Liu, Philip S. Paress,
Lex H. T. Van der Ploeg’, James M. Schaeffe
& Joseph P. Arena
Department of Cellular Biochemistry and Physiology and
* Department of Genetics and Molecular Biology,
Merck Research Laboratories, Rahway, New Jersey 07065-0300, USA
Th avermectins are a family of macrocyclic lactones used inthe
conteol of nematode and arthropod parasites. Ivermectin (2223-
er blindness
in humans", Abamectin (avermectin By.) isa miticide and insec-
ticide used in erop protection’. Avermectins interact with verte-
bbrate and invertebrate GABA receptors™’ and invertebrate
sglutamate-gated chloride channels". The soil nematode Cacnor-
habiitis elegans has served as a useful model to study the mechan-
ism of action of avermectins''"*. AC. elegans messenger RNA
expressed in Xenopus oocytes encodes an avermectin-sensitive
glutamate-gated chloride channel", To elucidate the structure
and properties of this channel, we used Xenopus ooeytes for expres-
NATURE - VOL 371 - 20 OCTOBER 1994
sion cloning of two functional complementary DNAs encoding an
avermectin-sensitive glutamate-gated chloride channel. We find
that the electrophy siological and structural properties ofthese pro-
Jns indicate that they are new members of the ligand-gated ion
‘channel superfari
Xenopus oocytes injected with C. elegans poly(A)” RNA
exhibit a rapidly activating reversible ghutamate- and irreversible
ivermectin 4"-O-phosphate (IVMPO,)-sensitive — current
(Fig, 1a)", A dizectional cDNA library was constructed from
size-fractionated RNA and sereened by expression in Xenopus
oocytes. Glutamate: and TVMPO,-sensitive currents were
observed after injection of i vitye RNA from a pool of 5.000
eDNAs (Fig. 1a). Subjractionation yielded two cDNA clones
(PGluCha and pGluCl-f) that expressed glutamate- and
IVMPO,-sensitive currents (Fig. 1a)
‘Oocytes simultaneously expressing GluCl a and proteins.
exhibited rapidly activating reversible glutamatc- and irrevers-
ible IVMPO,-sensitive currents comparable with those found
in oocytes injected with poly(A)” RNA (Fig. 1a). The time
for maximal activation of IVMPO,sensitive current was
42425 for GluCl wand f compared to 36+35 for poly(A)
RNA. The desensitization of the glitamate-sensitive current
observed in oocytes injected with poly(A)” RNA was also
observed in oocytes injected with GluCl @ and Bat glutamate
concentrations greater than TmM., The individual: subunits,
GluCla or GluCl-p, expressed functions homomerie channels
which were selectively responsive to IVMPO, or glutamate,
respectively (Fig. 1a). The time course for IVMPO, activation
of homomeric GluClar channels was 181s, which is faster
than that observed for GluCl @ and and poly(A)” RNA
(P0001),
‘The dose-response curves indicate that coexpression of GluCl
‘@ and i result in changes in the effector concentration for half
maximum response (EC) for glutamate and in the Hill
107