Fragmentation of grid cell maps in a multicompartment environment

In this paper, they recorded neural activity in grid cells and place cells when rats ran through a hairpin maze.

Figure 1

• grid cells have a primarily been recorded in environments with no internal boundaries (like open fields)

• grid maps repeat across alleys of the hairpin maze.

• in hairpin maze, although two-dimensional periodicity of the grid was lost, the positions were highly correlated across arms, especially all even-numbered arms or all odd-numbered arms.

Figure 2

• population analysis for a single cell emsemble.
• they showed repeating submaps and high correlation btw neural activity in the same direction, but in the opposite direction the correlation values were not high.
• grid cell representation segmentation (fragmentation)

Figure 3

• population analysis for all trials and all rats
• a,b: population vector correlation matrix (in one rat)
• c,d: all rats
• grid representations don't encode simple spatial coordinate but direction-specific and visual-input dependent maps

Figure 4

• representation were reset near the turning points.
• correlation values are high between adjacent bins in the population vector, however, around the turning point in the hairpoint (the start and end of each corridor).
• in c, tested on shuffled data where correlation value was high indicating that this is not the statistical result related to noise

Figure 5

• short cut experiments (most interesting results!)
• they intentionally cut one specific arm - truncated arm.
• on truncated arms, when shift is small (around starting point) or high (around end point), the correlation values with reference arms were high, but around middle shift value, the correlation values were low. <-- they're confusing where they are.

Figure 7

• open field, hairpin, virtual hairpin, open field 2
• in OF, VH, OF2 setting, the grid field were maintained. (spatial phase), but in HP, realignment occurs (?) and the corrlelation value were low
• the fact that such discontinuities were absent in the virtual hairpin task suggests that the realignments were imposed by the physical structure of the task.
• in the transparent wall condition, grid realignment between compartments still occurred, indicating that segmentation was not solely due to visual occlusion.

Discussion

• fragmentation of grid cells. firing patterns of grid cells were repeatedly similar in each corridor. but the firing patterns were sharply disconnected at turning points where the direction changed --> suggesting that grid cells were fragmented into "submaps"
• similar to place cell.
• fragmentation was not seen in open spaces or virtual hairpin without physical walls even if the trajectory was the same in hairpin map.