Neuromodulation of Attention

Abstract

Attention is critical to high-level cognition and attention deficits are a hallmark of neurologic and neuropsychiatric disorders. Although years of research indicates that distinct neuromodulators influence attentional control, a mechanistic account that traverses levels of analysis (cells, circuits, behavior) is missing. However, such an account is critical to guide the development of next-generation pharmacotherapies aimed at forestalling or remediating the global burden associated with disorders of attention. Here, we summarize current neuroscientific understanding of how attention affects single neurons and networks of neurons. We then review key results that have informed our understanding of how neuromodulation shapes these neuron and network properties and thereby enables the appropriate allocation of attention to relevant external or internal events. Finally, we highlight areas where we believe hypotheses can be formulated and tackled experimentally in the near future, thereby critically increasing our mechanistic understanding of how attention is implemented at the cellular and network levels.

Keywords: attention; attractor networks; neuromodulators; pharmacology; population coding; top-down.

Figure 1

Schematic that Exemplifies Some of the Effects Attention Has on Cellular Activity and on Population Activity Feedback and neuromodulator influences alter the drive in excitatory (blue) and in inhibitory (red) cells, which thereby exert increased influence on one another and the rest of the network, leading to overall increased, but balanced excitation, and inhibition. Exemplified are three scenarios, which have been described in the literature when attention is deployed to a spatial location and/or to specific stimulus features. These can also occur when neuromodulators are applied to the local network. (A) A result thereof is a gain change, i.e., a change of the output a neuron produces given a specific input. (B) Another effect of increased balanced excitation and inhibition is an increase in gamma oscillations of the local field potential, indicative of higher neuronal coherence, proposed to improve information exchange between selected neuronal populations. (C) Attention and neuromodulators also change the relationship between tuning similarity and noise correlations of neuron pairs, such that population coding properties are improved. (D) These changes are reminiscent of altered network attractor dynamics, which stabilize network states, reduce fluctuations, and increase the ability to stay task focused. a.u., arbitrary units; Hz, frequency in Hertz.

Figure 2

Influence of Tonic and Phasic Modes of Cholinergic Signaling on Attention, and Spatial Specificity of Cholinergic Input-Output Relations (A) Tonic levels of ACh in rodent prefrontal cortex are interspersed by brief phasic increases in ACh, which occur after behaviorally relevant cues, but only if they occur after “non-cue” trials. These are preceded by glutamate increases, which occur on all “cue detect” trials (after Sarter et al., 2014). (B) Potential source of spatial and temporal specificity of ACh signals in the cortex. ACh release in rodent prefrontal cortex is partly dependent on local glutamate activating presynaptic NMDA receptors. Local glutamate control of ACh in rat prefrontal cortex is released from mediodorsal thalamic inputs. Whether other glutamate sources (e.g., feedback from higher cortical areas) equally control ACh release locally is unknown. (C) Input to and output from cholinergic basal forebrain neuron is segregated into specific sub-circuits. Cholinergic BF neurons that project to the prefrontal cortex receive input mostly from neurons in the lateral septum, and from small, but segregated populations of the central amygdala. Cholinergic BF neurons that project to the motor cortex receive inputs from neurons in the somatosensory cortex, and segregated populations of neurons in the central amygdala and the caudate nucleus. Cholinergic BF neurons that project to the basolateral amygdala receive input from segregated populations in the central amygdala and the caudate nucleus (after Gielow and Zaborszky, 2017).

Figure 3

Neuromodulation of WM Fields, Remote Feature Tuning, and Specificity of Dopaminergic Output Signals (A) Spatial tuning of WM fields is enhanced when small amounts of D1 agonists are applied in the vicinity of the neurons, by selectively reducing activity for non-preferred locations (memory fields). This is equivalent to a noise reduction. Spatial tuning of WM fields is equally enhanced when small amounts of NA α2A agonists are applied in the vicinity of the neurons, by selectively increasing activity for preferred locations (memory fields). This is equivalent to signal enhancement. Both changes increase the SNR. (B) Application of D1 antagonists to area FEF enhances the tuning of area V4 neurons that have overlapping receptive fields with the affected FEF locations. (C) Hypothetical interactions of dopaminergic subpopulations carrying specific; based on projections found in rat. dlPFC, dorsolateral prefrontal cortex; FEF, frontal eye field; VTA, ventral tegmental area; SN, substantia nigra; a.u., arbitrary units.