Emergence of preconfigured and plastic time-compressed sequences in early postnatal development

Abstract

When and how hippocampal neuronal ensembles first organize to support encoding and consolidation of memory episodes, a critical cognitive function of the brain, are unknown. We recorded electrophysiological activity from large ensembles of hippocampal neurons starting on the first day after eye opening as naïve rats navigated linear environments and slept. We found a gradual age-dependent, navigational experience-independent assembly of preconfigured trajectory-like sequences from persistent, location-depicting ensembles during postnatal week 3. Adult-like compressed binding of adjacent locations into trajectories during navigation and their navigational experience-dependent replay during sleep emerged in concert from spontaneous preconfigured sequences only during early postnatal week 4. Our findings reveal ethologically relevant distinct phases in the development of hippocampal preconfigured and experience-dependent sequential patterns thought to be important for episodic memory formation.

Fig. 1.. Early-life development of hippocampal neuronal ensemble spatial representation during first-time navigation.

(A) Schematic of experimental protocol. (B) Illustration of bilateral silicon probe implants in the dorsal hippocampal area CA1. (C) Sagittal brain section depicting tracks of four silicon probe shanks (white arrowhead) recording in CA1 (blue arrowhead) at P22. (D) Color-coded illustration of clusters of eight single cells recorded from one shank in a P15 animal during sleep-run-sleep sessions. (E) Simultaneous recording of 67 CA1 neurons across eight shanks at P15. (F) Similar velocity during first-time (de novo) run across development [P = 0.209; analysis of variance (ANOVA)]. (G) Example of first-time run trajectory at P15. (H) Examples of simultaneously recorded place cell sequences during first-time run across development. (I) Increased accuracy in neuronal ensemble representation of spatial locations during first-time navigation across ages throughout development, based on a memoryless Bayesian decoding algorithm (heatmap). Yellow line: actual animal trajectory. The color map shows the probability of a decoded location. (J) Significance (from P15–P16 onward, run vs. shuffle; P < 10⁻¹⁰, rank sum tests) and age-dependent improvement (P < 10⁻¹⁰, Kruskal-Wallis ANOVA) of neuronal ensemble–decoded locations during first-time navigation across development. Representation becomes adult-like at P23–P24. For post hoc comparisons with adult data, Dunn’s test was used. ***P < 0.001; ns, not significant.

Fig. 2.. Delayed development of compressed theta sequences during first-time navigation on linear tracks.

(A) Age-dependent increase in CA1 theta oscillation frequency across development (P < 10⁻¹⁰, ANOVA). The middle and bottom panels show wide band signals and the normalized power spectra for example P15 and P24 animals during the run. (B) Examples of developmental timeline (P15 to P24, panels 2 to 11) for compressed theta sequences binding past, current, and future locations (cartoon at top left). White curves: session-averaged theta oscillations. Note the absence of compressed theta rhythmic binding of past, current, and future locations at younger ages. (C) Quantification method for compressed theta sequence strength (quadrant ratio) in binding past, current, and future locations. (D) Developmental timeline (P15 to P24) of theta sequence strength. Note the emergence of consistent adult-like theta sequences (>shuffle) at P23–P24 (two directions per animal; n = 5, 3, 4, 4, 3, and 3 animals per age group, from P15–P16 to adult). (Inset) The quadrant ratio index significantly increases with age and becomes adult-like at P23–24. (E) Proportion of animals within age groups with theta sequences above the 95th percentile of their shuffles. (F) Effects of navigational experience on theta sequence strength (quadrant ratio index). Note that within-day experience (green bar; P = 0.75, paired t test, two directions per animal, three animals) does not reveal significant increases in theta sequence strength at P21–P22, while the increase in age, from P21–P22 to P23–P24, reveals increased strength during first-time navigation (red bar; P = 0.04, t test, two directions per animal, n = 4 and 3 animals). The effect of age is significantly higher than that of within-day experience (P = 0.03, t test). (G) Theta sequence slope/spatial extent (top) and compression ratio (bottom) increase with age (P < 10⁻¹⁰, ANOVAs). Data are means ± SEM. ***P < 10⁻¹⁰; *P < 0.05; ns, not significant.

Fig. 3.. Developmental timeline for trajectory-depicting sequences during on-track awake rest epochs.

(A) Neuronal ensemble activity during rest within first-time navigation (left panel) depicts individual locations at P15–P16 (stationary ensembles in frames, middle panel) and entire trajectories from P19–P20 (right panel). (B) Examples of stationary location-depicting frames and partial and entire trajectory–depicting frames across age groups during on-track rest epochs for first-time navigation (Day1Run, left 30 panels), within-day runs on same track (Day1Run2, middle 5 panels), and Day2Run on the same track (right 5 panels). Bayesian decoding analysis was used. (C) Age-dependent increases in proportion of significant trajectory-depicting frames (left; P < 10⁻⁴, ANOVA), correlation strength (middle; P < 10⁻¹⁰), and extent of depicted trajectory (right; P < 10⁻⁶). (D) Within-day (P = 0.92, paired t test) and previous-day (P = 0.67, t test) experience reveal no changes in the proportion of significant trajectory-depicting frames at P15–P16 and P16 (two directions per animal, n = 3 and 4 animals). Data are means ± SEM. ***P < 0.005; **P < 0.01; *P < 0.05; ns, not significant.

Fig. 4.. A three-stage developmental timeline for experience-dependent trajectory replay during sleep.

(A) Examples of stationary location–depicting (P15–P16), partial trajectory–depicting (P17–P18), and entire trajectory–depicting (from P19–P20) frames across age groups during sleep preceding first-time navigation (Pre-Run, preplay) and postnavigation sleep (Post-Run, replay) using Bayesian decoding. (B) (Top) Age-dependent increase in proportion of trajectory-depicting frames during Pre-Run (i.e., preplay) and Post-Run (i.e., replay) sleep (P < 10⁻¹⁰, ANOVAs) and age- and experience-dependent increased replay vs. preplay at P23–P24 (P < 0.009, t test). Color backgrounds demarcate the three stages. (Bottom) Preplay, replay, and plasticity at the individual animal level across development. (C) Age-dependent and navigational experience–independent decreases in stationary location–depicting frames in Pre-Run and Post-Run sleep (P < 10⁻⁷, ANOVAs). (D) Distribution of decoded track locations from ensemble activity within stationary frames at P15–P16. (E) Developmental timeline for time compression of decoded trajectory (P < 10⁻¹⁰, ANOVAs). (F) Proportions of stationary frames out of all significant frames (>shuffles, stationary and trajectory-depicting; left) and distributions of extents of decoded track (right) at P15–P16 and P23–P24. (Inset) Age-dependent increase in extent of decoded trajectory (P < 10⁻⁸, ANOVA). (G) Age-dependent increase in ripple power associated with location- and trajectory-depicting sleep frames during Pre-Run sleep (P < 0.004, ANOVAs). (H) Cross-development, two-parameter analysis of experience-dependent plastic changes in decoded trajectory from Pre-Run to Post-Run sleep (replay > preplay starting at P23–P34, Z-test for two proportions). (I) Prior navigational experience does not accelerate expression of Pre-Run to Post-Run sleep plasticity at P21–P22 (day 2, replay ≈ preplay). Data are means ± SEM. For (B) to (I), ***P < 0.005; **P < 0.01; *P < 0.05; ns, not significant. In (B) and (C), colored asterisks indicate P values from comparisons with chance.