Bidirectional synaptic mechanisms of ocular dominance plasticity in visual cortex

Abstract

As in other mammals with binocular vision, monocular lid suture in mice induces bidirectional plasticity: rapid weakening of responses evoked through the deprived eye followed by delayed strengthening of responses through the open eye. It has been proposed that these bidirectional changes occur through three distinct processes: first, deprived-eye responses rapidly weaken through homosynaptic long-term depression (LTD); second, as the period of deprivation progresses, the modification threshold determining the boundary between synaptic depression and synaptic potentiation becomes lower, favouring potentiation; and third, facilitated by the decreased modification threshold, open-eye responses are strengthened via homosynaptic long-term potentiation (LTP). Of these processes, deprived-eye depression has received the greatest attention, and although several alternative hypotheses are also supported by current research, evidence suggests that alpha-amino-3- hydroxy-5-methyl-4-isoxazolepropionate (AMPA) receptor endocytosis through LTD is a key mechanism. The change in modification threshold appears to occur partly through changes in N-methyl-D-aspartate (NMDA) receptor subunit composition, with decreases in the ratio of NR2A to NR2B facilitating potentiation. Although limited research has directly addressed the question of open-eye potentiation, several studies suggest that LTP could account for observed changes in vivo. This review will discuss evidence supporting this three-stage model, along with outstanding issues in the field.

Figure 1

Mechanisms of deprived-eye depression vary across cortical layers. (a) Thalamocortical axons project to both layers 4 and 3 of the mouse visual cortex. Axons were labelled with biotin-conjugated dextran (3000 MW) injected into binocular dLGN of a P28 mouse. Imaging took place 4 days later from fixed coronal sections containing visual cortex. Laminar borders were determined based on Nissl staining. Image courtesy of J. Coleman. (b) Schematic showing CB1 receptor density variations across cortical layers, with high levels of expression in supragranular layers and limited expression in layer 4. Drawing based on Deshmukh et al. (2007). (c) Three days of MD produces depression of deprived-eye responses in both layers 3 and 4 through distinct mechanisms. In layer 3, both LTD and deprived-eye depression require CB1 activation, whereas LTD is independent of postsynaptic AMPAR internalization. The absence of high CB1 expression in layer 4 correlates with the lack of a requirement for CB1 activation in both LTD and deprived-eye depression. By contrast, LTD in layer 4 requires AMPAR endocytosis, suggesting that deprived-eye depression at this synapse may also occur through AMPAR internalization.

Figure 2

Layer 4 VEPs display binocular competition but not synaptic scaling. (a) MD fails to elicit changes in responses recorded in (i) the monocular zone ((ii) record C VEPs), and (b) 4 days of complete darkness fail to modify responses in the binocular zone, showing that deprived-eye depression is impaired in neurons that do not experience binocular competition (record C and I VEPs). (c) Prolonged binocular deprivation fails to produce scaling up of VEPs recorded in layer 4 of the binocular cortex. Responses to stimulation of the contralateral (C) and ipsilateral (I) eyes were recorded prior to and following 7 days of binocular lid suture. No change in the response to either eye was observed, contrary to the prediction that synaptic scaling would lead to increased responses following decreased input activity. Experimental treatment: white, no manipulation; grey, eyelid suture; black, dark exposure. Data are replotted from (a) Frenkel & Bear (2008), (b) Blais et al. (2008) and (c) Frenkel & Bear (2004).

Figure 3

Alternative mechanisms for achieving open-eye potentiation and homeostasis following MD. Under both the synaptic scaling and BCM hypotheses, the initial response to MD is the depression of deprived-eye responses, which results from the decorrelation of deprived-eye inputs following lid suture. This decorrelation of input activity drives homosynaptic LTD in the binocular visual cortex. A consequence of the degraded input coupled with the weakening of synapses via LTD is the reduced spiking activity of neurons in the binocular visual cortex, which, after several days, leads to compensatory changes. The nature of these changes differs between the two models of open-eye potentiation. According to the synaptic scaling model, neurons respond to reduced spiking activity by globally scaling up synaptic weights, thereby increasing incoming drive and returning spiking to baseline levels. By contrast, under the BCM model, neurons respond to decreased spiking activity by lowering the modification threshold (θ_m), thus promoting LTP across a larger range of inputs. The inset shows that prior to MD, the boundary between potentiating and depressing inputs is equal to the open-eye input level, thereby maintaining stable open-eye responses if visual experience is not manipulated. Following 3 days of MD, the boundary shifts to the left, and open-eye inputs to the cortex (the strength of which has not changed as visual experience through the open eye has remained constant) now induce homosynaptic potentiation. This increase in drive from the open eye in turn elevates postsynaptic spiking to baseline levels. Therefore, both the synaptic scaling and BCM models can account for both the potentiation of open-eye responses and the output homeostasis of visual cortical neurons. The difference lies in the behaviour of the deprived-eye inputs during the later stages of MD. Synaptic scaling is heterosynaptic; therefore, both open- and deprived-eye inputs are predicted to increase equally. By clear contrast, BCM-mediated homeostasis occurs via homosynaptic mechanisms, thus only open-eye responses potentiate. Current data suggest that the deprived eye does not potentiate proportionally to the open eye, suggesting that global scaling of responses is not occurring in response to MD.