Neuronal vector coding in spatial cognition

Abstract

Several types of neurons involved in spatial navigation and memory encode the distance and direction (that is, the vector) between an agent and items in its environment. Such vectorial information provides a powerful basis for spatial cognition by representing the geometric relationships between the self and the external world. Here, we review the explicit encoding of vectorial information by neurons in and around the hippocampal formation, far from the sensory periphery. The parahippocampal, retrosplenial and parietal cortices, as well as the hippocampal formation and striatum, provide a plethora of examples of vector coding at the single neuron level. We provide a functional taxonomy of cells with vectorial receptive fields as reported in experiments and proposed in theoretical work. The responses of these neurons may provide the fundamental neural basis for the (bottom-up) representation of environmental layout and (top-down) memoryguided generation of visuo-spatial imagery and navigational planning. [H1] Introduction Place cells fire whenever an animal traverses a specific location in an environment (the spatial receptive field [G] of that neuron, also known as its 'place field', FIG. 1a). Since the discovery of place cells in the rat hippocampus by O'Keefe and Dostrovsky 1 , researchers have uncovered a multitude of additional spatially-selective cell types in rodents: that is, neurons whose receptive fields reference some aspect of an organism's location, state of motion, pose or its relationship to environmental features (such as boundaries, landmarks and other objects). The spatial receptive fields of some of these cells correspond to vectors, indicating the distance and direction in space (relative to the animal's current location) within which the presence of an environmental feature will drive the neuron to fire. Such 'vectorial codes' for space have received comparatively less attention than the coding performed by place cells or other well-known spatially-selective cell types, such as grid cells [G] 2 (FIG. 1a) and head direction cells [G] 3,4. Vectorial codes for space are expressed by boundary vector cells 5,6 , border cells 7,8 , landmark vector cells 9 and object vector cells 10 in allocentric (that is, world-centered) coordinates. Just like place fields, the receptive fields of these vector-coding cells do not reflect the orientation of the animal but do rotate together with the prominent environmental features that control head direction cell firing 11. Egocentric counterparts of some of these cells-in which the direction of receptive fields are referenced relative to the facing direction of the agent-have also been found 12-15 , as well as cells that exhibit head direction-modulated boundary responses 16,17 .

Thalamus and claustrum control parallel layer 1 circuits in retrosplenial cortex

“…Consistent with this hypothesis, we observed that the magnitude of the advantage of STF over STD is proportional to the conductance of the Kv1 current (early spiking window: R 2 = 0.9, p = 2.88e-08; late-spiking window: R 2 = 0.82, p = 1.81e-06; Figure 4H&I). Overall, the synaptic dynamics of the inputs to LR neurons are thus better suited for temporal latency coding of thalamic directional inputs but a rate code for dorsal subicular distance-related inputs, such as those likely to be provided by bursty boundary vector cells in the dorsal subiculum (Bicanski and Burgess, 2020; Lever et al, 2009; Simonnet and Brecht, 2019; Stewart et al, 2014).…”

“…Further work is needed to determine whether any particular subtype of layer 5 pyramidal neurons may exhibit substantial TC-evoked responses, but our results, reporting significantly larger responses to TC input by LR neurons, suggest LR cells are the predominant spatial information-encoding subtype within RSG. The dorsal subiculum, which serves to transmit allocentric spatial information, including axis cell and boundary vector signals (Lever et al, 2009; Derdikman, 2009; Olson et al, 2017; Simonnet and Brecht, 2019; Bicanski & Burgess, 2020), also precisely targets LR neurons via projections localized to layer 3 of RSG (Figure S3; Nitzan et al, 2020; Yamawaki et al, 2019a). These inputs overlap with LR cell bodies and basal dendrites and have been shown to evoke larger excitatory postsynaptic currents in superficial compared to layer 5 pyramidal neurons (Figure S3; Nitzan et al, 2020; Yamawaki et al, 2019a).…”

A compact spatial map in V2 visual cortex

“…Proprioceptive feedback and sensorimotor integration may play a vital role in reshaping spatial selectivity of hippocampal and cortical neurons (Acharya et al, 2016; Aghajan et al, 2015; Ravassard et al, 2013). It may not come as a surprise that the spatio-visual map in visual cortex encoded both egocentric and allocentric spatial cues in parallel (Bicanski and Burgess, 2020; LaChance et al, 2019; Wang et al, 2020). How and whether hippocampal-entorhinal spatial signals interact with non-visual navigational signals in the visual cortex awaits for further investigation.…”

Striatal and hippocampal contributions to flexible navigation in rats and humans

“…Recent studies have begun to explore how different types of spatial information may be tracked by specific brain regions during navigation. Two important metrics for flexible navigation are vector-to-goal and path-to-goal (Bicanski and Burgess, 2020; Chadwick et al, 2015; Spiers and Barry, 2015). Using in situ learning experience and film simulation of Soho in London (UK), Howard et al (2014) identified neural correlates of path distance to goal in the right posterior hippocampus.…”