Reversal and Stabilization of Synaptic Modifications in a Developing Visual System

See allHide authors and affiliations

Science  20 Jun 2003:
Vol. 300, Issue 5627, pp. 1953-1957
DOI: 10.1126/science.1082212


Persistent synaptic modifications are essential for experience-dependent refinement of developing circuits. However, in the developing Xenopus retinotectal system, activity-induced synaptic modifications were quickly reversed either by subsequent spontaneous activity in the tectum or by exposure to random visual inputs. This reversal depended on the burst spiking and activation of the N-methyl-D-aspartate subtype of glutamate receptors. Stabilization of synaptic modifications can be achieved by an appropriately spaced pattern of induction stimuli. These findings underscore the vulnerable nature of activity-induced synaptic modifications in vivo and suggest a temporal constraint on the pattern of visual inputs for effective induction of stable synaptic modifications.

Synaptic modifications induced by patterned neuronal activity are essential for experience-dependent refinement of developing circuits (15) as well as learning and memory (69) in the adult brain. However, how these modifications become stabilized in the constant presence of neuronal activity in vivo remains largely unclear. We addressed this question using a specific in vivo preparation (10, 11). Synaptic connections in the developing tectum of Xenopus are highly susceptible to modification by activity in the visual pathway. Long-term potentiation (LTP) and long-term depression (LTD) can be readily induced at retinotectal synapses by correlated spiking of retinal ganglion cells (RGCs) and tectal neurons (12) and by repetitive visual stimuli (13). In these studies, the postsynaptic tectal neurons were usually voltage-clamped at a constant membrane potential, a condition that prevents neuronal spiking. Under natural conditions, however, postsynaptic neurons are likely to spike spontaneously, as a result of random sensory inputs due to ambient illumination (14) or inputs from other brain regions (15, 16).

We first examined whether spontaneous activity of tectal neurons affects the persistence of LTP and LTD at retinotectal synapses induced by correlated pre- and postsynaptic activity. Repetitive suprathreshold stimulation of an RGC input (at 1 Hz for 100 s) resulted in an immediate increase in the amplitude of excitatory postsynaptic currents (EPSCs), without affecting the unstimulated control input that converged onto the same tectal neuron (Fig. 1A). This potentiation was long-lasting when the tectal neuron was kept in voltage clamp (v.c.) after the correlated spiking (Fig. 1B). However, when the recording was switched to current clamp (c.c.) for a period of 10 min, the stimulation-induced synaptic potentiation was largely abolished (Fig. 1, A1 and C) (17). Despite reversing recently induced LTP, spontaneous activity did not affect the control RGC input on the same tectal neuron (Fig. 1, A2 and D). This reversal is not limited to LTP induced by the low-frequency correlated spiking. LTP induced by conventional theta burst stimulation (TBS) (18), which causes high-frequency correlated activity, was similarly reversed by spontaneous activity 10 min after LTP induction (Fig. 1E). Correlated pre- and postsynaptic spiking can also induce LTD at these retinotectal synapses (12). Repetitive stimulation of subthreshold RGC inputs at 1 Hz (for 50 s), when preceded within 5 to 10 ms by postsynaptic spiking (evoked by injection of a depolarizing current), induced a persistent reduction of the EPSC amplitude when monitored in v.c. (Fig. 1B). Switching the recording to c.c. for 10 min after correlated activation largely abolished the depression (Fig. 1C).

Fig. 1.

Rapid reversal of LTP and LTD by spontaneous activity in vivo. (A1) Monosynaptic EPSCs in the tectal neuron that were elicited by test electrical stimuli applied to RGCs (at 0.05 Hz). Synaptic potentiation was induced by correlated pre- and postsynaptic spiking (arrow). The recording was switched to c.c. for 10 min to allow spontaneous (spont.) activity (bar). EPSC amplitude is shown in pA. Inset traces: Sample EPSCs (an average of 5, arrowheads). Inset scales, 50 pA and 10 ms. (A2) Synaptic responses at a control RGC input converging onto the tectal neuron in (A1). (B) When neurons were monitored in v.c. throughout the experiment, there was no decrement of the established LTP (solid circles, n = 9 cells) or LTD (open circles, n = 11 cells). (C) Ten minutes of spontaneous activity reversed the established LTP (solid circles, n = 7 cells) or LTD (open circles, n = 10 cells). Arrows indicate the induction of LTP [(B) and (C)] or LTD (C). (D) Spontaneous activity did not affect control convergent inputs (n = 10 cells). (E) TBS-induced LTP was also reversed by spontaneous activity (n = 8 cells).

These modifications of retinotectal synapses were induced by electrical stimulation of RGCs. Can synaptic modifications induced by visual experience be similarly reversed by spontaneous activity? We measured changes in the synaptic strength of a number of retinotectal connections before and after exposure of the retina to 60 sweeps of a unidirectional moving bar (11), a condition known to induce a persistent directional sensitivity in the tectal response when the tectal neuron is continuously monitored under v.c. (19). When the EPSCs were monitored in the tectal neuron under v.c. (Fig. 2, A and B), some connections became potentiated (20 out of 40), whereas others became depressed (11 out of 40) or remained unchanged (9 out of 40) (20). Such disparate changes are consistent with selective modifications of retinotectal connections that underlie the emergence of directional sensitivity. When the tectal neuron was allowed to spike for 10 min under c.c. within a few minutes after training with the moving bar, synaptic potentiation or depression was greatly diminished (Fig. 2, B1 and B2). This is consistent with the findings on the reversal of LTP and LTD induced by correlated activation of RGCs and tectal neurons. When we recorded the spiking activity from tectal neurons in cell-attached configurations, the enhanced spiking activity of the tectal neuron in response to movingbar stimuli after training with the same stimuli (19) also showed gradual reversal (Fig. 2D1). This reversal did not occur when spontaneous activity (including spiking) was eliminated by treatment with the glutamate receptor blocker kynurenic acid (KyA) (Fig. 2D2).

Fig. 2.

Spontaneous activity and random visual inputs reversed synaptic modifications induced by training with moving bars. (A) Retinotectal connections that exhibited (A1) LTP and (A2) LTD after exposure to repetitive, unidirectional, moving light bars across the retina. (B) Summary of all experiments in which (B1) stable potentiation (open circles, n = 6 cells), (B2) depression (open circles, n = 4 cells), or (B3) no change (open circles, n = 3 cells) of retinotectal connections was observed (in v.c) after moving-bar training as in (A). Spontaneous activity at 5 to 15 min after bar training reversed the potentiation and depression [solid circles in (B1), n = 8 cells, and (B2), n = 4 cells], with no effect on the connections not altered by moving-bar stimulation [solid circles in (B3), n = 3 cells]. (C) Random light flashes for 5 min reversed the (C1) potentiation (n = 6 cells) and (C2) depression (n = 3 cells) of retinotectal connections induced by training with a unidirectional moving bar, but had no effect on (C3) connections that were not modified by training (n = 3 cells). (D) Spontaneous activity reversed the enhancement of spiking activity of the tectal neuron induced by moving-bar training. (D1) The enhanced spiking activity in response to the moving bar in the trained direction fully decayed to the baseline level. (D2) Spontaneous activity, including spiking of tectal neurons, was blocked by KyA (10 mM); enhanced spiking to the trained stimuli was revealed after washout of KyA.

Spontaneous activity of the tectal neuron usually consists of random single spikes and bursts of spikes (Fig. 3, A1 and C). The extent of LTP reversal correlated well with the occurrence of bursts, but not with the mean spiking frequency (Fig. 3B). Light stimulation, such as random flashes of white light squares onto the retina (11), readily induced burst firing in the tectal neurons (Fig. 3A2), which was also reflected in the enriched high-frequency components in the distribution of interspike intervals (Fig. 3C). To further test the idea that burst spiking is important for the reversal of LTP and LTD, we applied random light flashes to the retina after correlated pre- and postsynaptic activation. A 5-min application of 60 episodes of random flashes (with tectal neurons in c.c.) completely abolished both potentiation and depression (Fig. 3D). Furthermore, the extent of reversal of LTP induced by random flashes correlated well with the mean number of bursts per minute (Fig. 3B1), but not with the mean frequency of spikes that occur in the tectal neuron (Fig. 3B2). Potentiation induced by TBS was also completely abolished by subsequent random flashes (n = 5 cells) (21). In addition, moving bar–induced potentiation and depression of retinotectal connections was similarly reversed by random flashes (Fig. 2C). Thus, random sensory inputs are effective in reversing synaptic modifications induced by patterned synaptic activation, and bursts occurring in the tectal neuron are likely to be the primary cause of this reversal. This result demonstrates that visual inputs can both induce and reverse synaptic modifications, depending on their specific spatial and temporal patterns.

Fig. 3.

Mechanisms underlying the reversal of synaptic modifications. (A) Sample traces of (A1) spontaneous spiking activity (bursts marked by asterisks) and (A2) activity induced by light flashes. (B) Correlation between the extent of LTP reversal with (B1) the mean burst rate or (B2) the mean spiking frequency of spontaneous activity (open circles) or light flashes (solid circles). The extent of reversal is shown as the percent reduction in the magnitude of potentiation after spontaneous activity (10 min) or random flashes (5 min), relative to the initial level of potentiation at 5 to 10 min after LTP induction. Bursts are defined as clustered spiking with an interspike interval ≤200 ms. (C) The histogram distribution of interspike intervals of spontaneous activity (white bars, n = 10 cells) or during exposure to random flashes (black bars, n = 8 cells). (D) Random flashes completely reversed the potentiation (n = 7 cells, solid circles) and depression (n = 4 cells, open circles) induced by correlated pre- and postsynaptic spiking. (E) Reversal of synaptic modifications by random flashes was blocked by hyperpolarization (hyper.) of the tectal neuron (solid circles, n = 5 cells), or by d-APV infusion (APV) of the tectum (open circles, n = 5 cells). (F) Burst spiking (60 pulses at 0.2 Hz, each pulse containing 5 spikes at 20 Hz) evoked by current injections into the tectal neuron while synaptic transmission was blocked by KyA did not reverse LTP (n = 7 cells, open circles). In contrast, LTP was reversed when the same pattern of burst spiking was driven by EPSCs (n = 5 cells, solid circles). Sti., stimulation of tectal neurons (open circles) or retinal inputs (solid circles). Arrows in (D), (E), and (F) indicate the induction of LTP or LTD. (G) Time window for the effective reversal of LTP by random light flashes. Tectal neurons were held in v.c. for different periods between the induction protocol and the exposure to 5-min random flashes. Numbers in parentheses are the number of cells examined.

An immediate effect of random light flashes is the spiking activity of tectal neurons. We first examined whether such spiking activity is necessary for the reversal of LTP. Postsynaptic spiking caused by random light flashes was prevented by hyperpolarization of the tectal neuron to –80 mV (in c.c.). Under this condition, exposure to the same 5-min random flashes at 10 to 15 min after LTP induction did not reverse the potentiation (Fig. 3E). However, burst spiking in the tectal neurons alone is not sufficient for LTP reversal, because burst spikes elicited by current injection in the tectal neurons, while synaptic transmission was blocked by KyA, did not reverse LTP (Fig. 3F). In contrast, the same spiking activity driven by RGC input completely abolished potentiation (Fig. 3F). Furthermore, random light flashes were ineffective in reversing LTP when the tectum was infused with D-aminophosphovalerate (D-APV) (50 μM), a specific N-methyl-D-aspartate (NMDA) receptor antagonist (Fig. 3E) (22). Thus, this reversal of LTP requires RGC-driven burst spiking in the tectal neurons and the activation of NMDA receptors. Burst spiking in the postsynaptic cell can lead to reversal of LTP through a homosynaptic (Fig. 3F) as well as a heterosynaptic (11) effect. Because the bursting spikes triggered by random flashes are largely uncorrelated with the activity of the specific retinotectal connection under examination, the reversal of synaptic modification appeared to result largely from uncorrelated postsynaptic spiking induced by other converging inputs on the tectal neuron (11). This heterosynaptic nature of the reversal is distinct from the depotentiation (23, 24) or de-depression induced by subsequent activity at potentiated or depressed synapses (25, 26), but is similar to the reversal of tetanus-induced hippocampal LTP that is associated with rats who enter a new environment (27). One potential factor mediating the reversal of LTP is protein phosphatase 1/2A (PP-1/2A) (28). Incubation of the tectum with okadaic acid, a selective PP-1/2A blocker, eliminated the reversal of LTP (11). Furthermore, we found that random flashes were effective in reversing LTP only when they were presented during the first 20-min period after the induction of LTP (Fig. 3G). This temporal constraint is analogous to that of the homosynaptic depotentiation of tetanus-induced LTP at hippocampal synapses by low-frequency stimulation of the potentiated synapses (2830).

The existence of the critical window for the reversal of LTP and the observation that the reversal appears to be gradual (Fig. 2D) suggest that stable synaptic modifications may be achieved by repetitive induction of LTP or LTD that counteracts the reversal action of spontaneous activity. We hypothesized that, whereas one episode of induction stimuli fails to induce persistent synaptic modifications, repetition of induction episodes in a “spaced” manner might result in persistent synaptic modifications, and the extent of stable modification might depend on the spacing of the induction episodes (Fig. 4A). We tested this hypothesis on TBS-induced LTP at retinotectal synapses. The magnitude of LTP is saturated when the total number of bursts in the TBS exceeded 20. However, synaptic potentiation induced by TBS (60 bursts) was completely abolished by 20 min of spontaneous activity (in c.c.), interrupted briefly three times for monitoring of EPSCs, after the induction of LTP (Fig. 4B). This reversal of synaptic potentiation was not due to deleterious neuronal or synaptic conditions, because subsequent TBS (20 bursts) was fully effective in inducing LTP to a similar extent (Fig. 4B). In contrast to the “massed” TBS (60 bursts) in one induction episode, spaced application of three episodes of TBS (20 bursts each) within the 20-min period counteracted the effect of spontaneous activity, resulting in persistent synaptic potentiation. Residual potentiation resulting from each TBS episode accumulated when the subsequent TBS episode was applied before the reversal was completed (Fig. 4C) (11). Spaced TBS was effective in producing stable LTP, to a level that occluded further potentiation by subsequent TBS application (Fig. 4D). Furthermore, when the interval between TBS episodes increased beyond an optimal interval of 5 min, the extent of stabilization became progressively diminished (Fig. 4E). Finally, we tested whether movingbar stimulation with a spaced rather than a massed pattern could lead to the long-lasting appearance of directional sensitivity in tectal neurons (11). Preference toward the trained direction was found in 12 out of 25 tectal neurons that were trained with the spaced pattern, whereas no preference was observed in 15 out of 15 cells that were trained with the massed pattern of stimuli (Fig. 4F).

Fig. 4.

Stabilization of synaptic modifications under a spaced pattern of induction stimuli. (A) A simple model depicting (left) massed and spaced induction paradigms and (right) the expected stabilization of synaptic modification as a function of the intervals of induction episodes. The shaded area marks the critical window for stabilization. The model predicts the existence of an optimal interval for maximum stabilization. (B) Massed induction with 60-burst TBS did not result in persistent potentiation in the presence of spontaneous activity (n = 9 cells). Potentiation was induced subsequently with one 20-burst TBS episode and was persistent when tectal neurons were in v.c. (C) Residual potentiation cumulated with spaced TBS (n = 8 cells) with three 5-min episodes of spontaneous activity, each after one 20-burst TBS episode (arrows). (D) A spaced induction protocol (n = 6 cells) with three 20-burst TBS episodes (spaced at 5-min intervals) led to a stabilized potentiation similar to that in (C), which was resistant to spontaneous activity [at the same level as in (B)]. (E) The level of stabilized potentiation depended on the inter-TBS intervals. Experiments were performed as in (D), with three TBS episodes. We measured mean potentiation levels immediately after the first TBS episode and after 20 min of in vivo spontaneous activity after the last TBS episode. The ratio of the latter to the former yielded the values for the stabilized potentiation. Numbers in parentheses are the number of cells examined. (F) Long-lasting directional sensitivity of tectal responses induced by spaced training with moving-bar stimuli. Average spike numbers in a tectal neuron evoked by the moving bar in a given direction were normalized to the average spike numbers evoked by the bar moving in all four directions. U, up; D, down; R, right; L(T), left (trained direction). For the cells trained with the massed pattern, there was no statistical difference for all four directions [analysis of variance (ANOVA), P = 0.265]. For cells trained with the spaced pattern, the difference was significant (*) in the trained direction [ANOVA test, F(25,96) = 14.30, P = 0.0001; L/U, L/D, L/R, P < 0.01; U/R or D, R/D, P > 0.05; Tukey's honestly significantly different (HSD) test].

Our results demonstrate the disruptive influence of spontaneous activity on experience-induced synaptic modifications in vivo and suggest the existence of a temporal constraint on the pattern of visual inputs necessary for induction of stable synaptic modifications. Because the developing visual system is highly modifiable by light exposure (13, 14, 31), the susceptibility of synaptic modification to reversal by spontaneous activity may serve as a protective mechanism against long-lasting synaptic changes triggered by incidental episodes of correlated activity. In the face of this susceptibility, a spaced pattern of visual inputs becomes essential for stable synaptic modifications.

Supporting Online Material

Materials and Methods

Fig. S1

References and Notes

View Abstract

Stay Connected to Science

Navigate This Article