Extended Data Fig. 7: Connectivity between functional cell types.

a, Estimated connection probability (in %, n = 16 animals, 1 session per animal) for projections originating from each of four functional classes (internal direction (‘ID’), conjunctive grid (‘conj’), bursty pure grid (‘bursty’) and non-bursty pure grid (‘non-bursty’). Dots show connection rates in individual animals (only plotted for animals and pairings with >1500 total cell pairs: n = 14, 6, 11, 8, 6, 4, 6, 2, 11, 6, 9, 5, 8, 2, 5 and 4 animals for each category, left to right). Bars and whiskers show means and standard deviations of estimated binomial distributions (data pooled across 16 animals). Note connections from internal direction cells to conjunctive grid cells and from conjunctive grid cells to pure grid cells, in addition to recurrent connections within each class. Also note that internal direction cells and conjunctive grid cells have putative connections to bursty pure grid cells, but not non-bursty pure grid cells (connection rates for direction-tuned to bursty: 237/102,786 or 0.23%; ID to non-bursty: 3/55,900 or 0.003%; conjunctive to bursty: 183/28,387 or 0.645%; conjunctive to non-bursty: 2/18,085 or 0.011%). Although estimated connection rates varied from animal to animal, the general pattern of connections was preserved. b, Examples of putative recurrent connections between internal direction (ID) cells. Each of the three rows shows firing rate cross-correlogram (left column) and directional tuning curves (right column) for an example pair of putatively connected ID → ID cells. Preferred firing direction (PFD) of each cell is indicated above directional tuning curves. Note similar directional tuning between connected cells. c, Three example cell pairs illustrating putative connections between ID-tuned cells (green) and conjunctive grid cells (pink). Plotted as in Fig. 3b. d, Three example cell pairs showing putative connections between conjunctive grid cells (pink) and pure grid cells (blue). Plotted as in Fig. 3e. e, Alignment of directional tuning for other putatively connected cell pairs. Scatter plot shows preferred directions of pairs of pre- and postsynaptic cells, plotted as in Fig. 3c. Note that all combinations of connected direction-tuned cells have similar directional tuning (angle between preferred directions: 12.1 ± 58.0 deg, circular correlation: r = 0.49, p < 2.2e-16 for 707 putative ID → ID pairs; angle: 4.6 ± 58.2 deg, circular correlation: r = 0.48 p = 5.7e-13 for 322 putative conjunctive grid → ID pairs; and angle: 11.1 ± 58.0 deg, circular correlation: r = 0.52 p = 5.5e-5 for 130 putative conjunctive grid → conjunctive grid pairs). f, Alignment of directional tuning is present across all recording sessions. Left: preferred directions for all pairs of connected ID cells to conjunctive grid cells (angle between preferred directions: 8.4 ± 51.6 deg, circular correlation: r = 0.56 p = 2.2e-16, n = 341 pairs from 40 sessions, 16 animals). Right: preferred direction of pre-synaptic grid cells and direction of grid phase offset between pre- and postsynaptic cell for all pairs of putatively connected conjunctive grid cells and ‘pure’ grid cells (angle: −0.7 ± 64.1 deg, circular correlation: r = 0.35 p = 1.4e-6, n = 213 pairs from 40 sessions). Each dot corresponds to one pair of cells. g, Left: tuning directions for randomly selected pairs of non-connected ID cells and conjunctive grid cells (1,600 pairs, 100 pairs per animal). Absolute angles between tuning directions were significantly smaller for connected ID → conjunctive grid pairs than for randomly selected cell pairs (mean absolute offset: 42.0 deg vs. 84.6 deg (chance 90 deg), p = 7.4e-17, two-tailed Mann-Whitney U test). Right: Tuning directions for randomly selected pairs of non-connected conjunctive grid cells and pure grid cells (1,600 pairs, 100 pairs per animal). Absolute angles between preferred directions were significantly smaller for connected conjunctive grid → pure grid cell pairs than for randomly selected cell pairs (mean absolute offset: 34.1 deg vs. 91.5 deg, p = 1.5e-11, Mann-Whitney U test). h, Connection rates between functional cell types estimated with different significance thresholds for identification of monosynaptic connections. Thresholds are specified in terms of standard deviations from the baseline firing-rate cross-correlograms. While connection probabilities are heavily dependent on detection thresholds (average connection probability ranging from 0.04% to 0.4%), the connection probability ratios are fairly stable. i, Alignment of tuning relationships of connected cell pairs is stable across significance thresholds (z-score of convolved cross-correlogram peak) for detecting monosynaptic connections. Left: directional tuning in ID → conjunctive grid cell pairs (mean angles: 12.7–16.3 deg, correlation coefficients 0.35–0.68); right: directional-tuning vs. grid-phase offset angle in conjunctive grid → pure grid cell pairs (mean angles between preferred directions: −2.0 to −0.4 deg, correlation coefficients 0.34–0.79).

Source data