NMDA receptors figure it out

Our ability to perceive, attend, and memorize depends on bidirectional information exchange between hierarchically organized brain areas. Areas lower in the hierarchy send information through feedforward connections to higher areas. Higher areas modulate the activity in lower areas by means of feedback connections, and thereby influence the local processing as well as the information they receive. Although this scheme is well accepted, the exact role of feedforward and feedback connections in cognition is still poorly understood, let alone the mechanisms by which they mediate their effects. A report by Self et al. (1) in PNAS sheds light on the mechanisms involved. They delineate the different receptor types that contribute to feedforward and to feedback processing. They recorded neuronal activity from macaque primary visual cortex (V1) while animals performed a figure-ground detection task. V1 neurons in awake animals show differential activity depending on whether they represent the figure (2, 3) or whether they represent the background, even if the image within their receptive field is identical under both conditions. This differential activity is known as figure-ground modulation, and it depends on feedback from higher areas (4). Based on modeling studies (5, 6), studies from anesthetized animals (7), and the known receptor properties (8), Self et al. (1) hypothesize that the main stimulus-driven response of the V1 neurons will be mediated by glutamatergic AMPA receptors, whereas the figure-ground modulation will be largely mediated by glutamatergic NMDA receptors. They tested this hypothesis by applying AMPA and NMDA receptor antagonists on some of their trials and studied which aspects of the neuronal response were affected by selective receptor blockade. Blockade of AMPA receptors reduced the neuronal activity overall, but it had very little effect on figure-ground modulation. This supports the idea that the main stimulus drive, which is carried by feedforward connections, is reliant on glutamate’s action on AMPA receptors (Fig. 1). However, it equally suggests that AMPA receptors have little role in figure-ground modulation. Intriguingly, blockade of NMDA receptors selectively reduced the figure-ground modulation, although it had relatively little effect on the stimulus-driven activity. This is an important study, one of only a few directly investigating specific receptor contributions to cognitive function (9–11), and it demonstrates a special role of NMDA receptors in feedback-mediated activity.

Fig. 1.

Cartoon of glutamatergic ionotropic receptor involvement in figure-ground coding. A figure-ground stimulus is displayed at the bottom. Self et al. (1) train monkeys to report whether a figure is present on any given trial and simultaneously record neuronal activity across different layers in V1. The example shows a trial in which a figure would have been present, and nonfigure trials would show an isooriented structured image in which the center stimulus would be unchanged and the surround stimulus would have an orientation identical to the center. V1 neurons, representing parts of the figure, showed higher activity on figure trials than on nonfigure trials. This elevated activity is called figure-ground modulation; it occurs in a somewhat delayed manner relative to response onset and is highlighted in blue in the central histogram. Self et al. (1) demonstrate that figure-ground modulation was almost entirely abolished when NMDA receptors (NMDAr) were blocked in V1, whereas the feedforward component of the response (red curve) was largely unaffected. Conversely, blocking AMPA receptors did not affect the figure-ground modulation, but it strongly reduced the feedforward response component (not shown).

The results are convincing, but it is extremely unlikely that feedback connections target synapses containing exclusively NMDA receptors or that feedforward connections target synapses containing exclusively AMPA receptors, a fact that the authors themselves discuss (1). We do know that this is not the case (12). Even so, the results suggest that feedback connections target synapses especially rich in NMDA receptors, whereas feedforward connections target synapses especially rich in AMPA receptors. It will be important, even if difficult, to verify this by combining tract-tracing, ultrastructural, and immune-histochemical methods in macaque monkeys.

Figure-ground segmentation is only one of many cognitive functions that rely on feedback connections. Attention is possibly the most studied cognitive function, wherein feedback connections influence neuronal processing in sensory cortex (13–16). A recent paper reported that muscarinic, rather than NMDA, receptors are vital for attentional rate modulation in primary visual cortex (9). This does not contradict the study by Self et al. (1), because a joint role for cholinergic and NMDA mechanisms in attention-mediated rate modulation is possible. Modeling suggests that acetylcholine simply adjusts the network properties, such that feedback, possibly acting through NMDA-rich synapses, can induce attentional modulation in V1 (17).

Attention affects a variety of neuronal signatures in sensory areas. It alters neuronal firing rate (18, 19), similar to the rate modulation that Self et al. (1) study, but it also alters the strength of oscillatory activity (20, 21), it reduces rate variability (22), and it reduces firing rate correlations in pools of neurons (23, 24). We currently do not know whether figure-ground segmentation has equivalent effects on these higher order statistics of neuronal activity, and whether these are equally dependent on NMDA receptor activation.

This short list of questions highlights how little we know regarding the neuropharmacology of cognition at the cellular level. Self et al. (1) make an important contribution toward filling the knowledge gaps.