Experience alters hippocampal and cortical network communication via a KIBRA-dependent mechanism

SUMMARY

Synaptic plasticity is hypothesized to underlie “replay” of salient experience during hippocampal sharp-wave/ripple (SWR)-based ensemble activity and to facilitate systems-level memory consolidation coordinated by SWRs and cortical sleep spindles. It remains unclear how molecular changes at synapses contribute to experience-induced modification of network function. The synaptic protein KIBRA regulates plasticity and memory. To determine the impact of KIBRA-regulated plasticity on circuit dynamics, we recorded in vivo neural activity from wild-type (WT) mice and littermates lacking KIBRA and examined circuit function before, during, and after novel experience. In WT mice, experience altered population activity and oscillatory dynamics in a manner consistent with incorporation of new information content in replay and enhanced hippocampal-cortical communication. While baseline SWR features were normal in KIBRA conditional knockout (cKO) mice, experience-dependent alterations in SWRs were absent. Furthermore, intra-hippocampal and hippocampal-cortical communication during SWRs was disrupted following KIBRA deletion. These results indicate molecular mechanisms that underlie network-level adaptations to experience.

In brief

Hippocampal sharp wave/ripples (SWRs) and cortical spindles are coordinated network oscillations proposed to support memory consolidation. Quigley et al. show that KIBRA, a memory-associated regulator of AMPAR trafficking, is required for experience-induced modification and temporal coupling of SWRs and spindles. They identify plasticity mechanisms that enable networks to adapt to new experience.

Graphical Abstract

INTRODUCTION

Synaptic plasticity plays a critical role in adaptive cognition and memory.^1–6 In particular, trafficking of AMPA-type ionotropic glutamate receptors (GluAs or AMPARs) within the post-synapse has emerged as a key and conserved regulator of synaptic plasticity and learning.^2,7 The synapse-enriched protein KIBRA (kidney and brain protein, also called WWC1) is an excitatory post-synaptic scaffold that regulates AMPAR trafficking and expression.^8–10 Accordingly, KIBRA is required in the adult brain for learning and memory as well as synaptic plasticity.^8–11 Specifically, loss of KIBRA impairs hippocampal LTP (long-term potentiation) and prevents LTP-induced increases in AMPAR expression.^8,10 However, neither constitutive nor adult-onset KIBRA deletion affect basal synaptic transmission or basal synaptic AMPAR expression.^8,10 Gene variants of KIBRA are associated with normal variation in human episodic and working memory performance,^12,13 highlighting its importance in human cognition. In addition, KIBRA and the protein complexes it organizes are implicated in a wide range of neurological/psychiatric disorders known to have synaptic/circuit etiologies,^8,14 including autism,^15–17 schizophrenia,^18,19 bipolar disorder,²⁰ Tourette’s syndrome,²¹and Alzheimer’s disease.²² It is unclear how abnormal plasticity may contribute to pathological brain function or how plasticity at individual synapses contributes to larger-scale, circuit-level changes that facilitate memory. Thus, elucidating KIBRA’s role in experience-dependent changes in network function may reveal fundamental links between plasticity, memory, and neurological disorders.

Of particular relevance for the study of network-level memory processes are hippocampal sharp-wave/ripple (SWR) events, brief (50–200 ms), high-frequency (125–300 Hz) oscillations that occur during inactive behavioral states such as slow-wave sleep and quiet wakefulness.^23–25 SWRs coordinate temporally compressed “replay” of neural activity representing prior or possible future behaviors,^25–28 and considerable evidence implicates SWRs as a network-level mechanism for memory consolidation and/or retrieval in both rodents and humans.^{23,26,29–31} Disruption of SWRs impairs memory,^30,32–34 and artificial prolongation of SWRs can enhance memory,³¹ arguing for a causal role of SWRs in mnemonic processes. Further, SWRs are altered in human patients with schizophrenia³⁵ and several animal models of cognitive dysfunction and neuropsychiatric disorders,^36–38 highlighting their importance in normal brain function.

Of functional relevance to SWR generation by the hippocampal circuit, cortical networks produce brief (0.5–3 s) 10- to 15-Hz oscillations during slow-wave sleep termed spindles that are strongly associated with cortical plasticity and memory consolidation.^39–42 Recent studies indicate that memory traces are simultaneously encoded in the hippocampus and neocortex, with recent memory dependent on hippocampal activity and remote memory dependent on cortical activity.⁴³ The active systems consolidation hypothesis proposes a two-stage model of memory consolidation in which memories that initially rely on hippocampal activity are transferred to the cortex during sleep-dependent memory consolidation.⁴⁴ This process is believed to involve the interaction of SWR events in the hippocampus with sleep spindles in the cortex to support information transfer and memory consolidation.^45,46 Indeed, cortical spindles are temporally associated with hippocampal SWRs, and disrupting this relationship impairs sleep-based memory consolidation, arguing that these network-level features represent acute periods of enhanced inter-regional communication required for memory consolidation,^39,40,45,46 although the synaptic mechanisms that underlie and modulate this coordination are largely unknown.

To study how KIBRA contributes to circuit-level memory formation, we examined the role of KIBRA in experience-dependent modulation of hippocampal SWR events and cortical spindles. We evaluated properties of these emergent network features during a novel experience and within rest periods before and after the behavioral session in wild-type (WT) mice and littermates with forebrain-selective deletion of KIBRA from excitatory neurons. We focused our investigation on KIBRA function in CaMKIIα+ excitatory neurons to provide insight into cell-group selective function of KIBRA in circuit dynamics. Single-cell RNA sequencing data indicate that KIBRA is expressed in the majority of excitatory forebrain neurons (89% of excitatory cortical neuron subtypes and 94% of excitatory hippocampal neuron subtypes, including all CA1 and all CA3 subtypes)^47,48 (Figure S1). In contrast, only 30% of inhibitory neuron subclasses show detectible KIBRA gene expression, with somatostatin+ interneurons contributing most to KIBRA-expressing GABAergic neurons^47,48 (Figure S1). Additionally, the level of KIBRA gene expression is substantially higher in excitatory than inhibitory neurons^47,48 (Figure S1).

RESULTS

Basal hippocampal network dynamics are unaffected by loss of KIBRA

To assess how loss of KIBRA affects neural circuit mechanisms underlying memory and adaptive cognition, we performed in vivo electrophysiological recordings from freely behaving WT mice and littermates lacking KIBRA in excitatory neurons of the forebrain (KIBRA conditional knockout [cKO]: Kibra^{Floxed/Floxed 8} x αCaMKII-CreER^{T2 49}).¹⁰ We simultaneously monitored brain activity in hippocampal areas CA1 and CA3 as well as the anterior cingulate cortex (ACC) before, during, and after exploration of a novel environment with enriched visual and social cues (Figures 1A and S2). The environment provided a salient experience expected to produce long-lasting changes in synaptic and network function.^3,51–53 The total distance traveled, mean velocity, and time spent in active movement were similar between WT and KIBRA cKO mice (Figures 1B–1D), indicating that both groups had equivalent levels of experience.

Figure 1. KIBRA cKO mice show normal baseline hippocampal area CA1 circuit dynamics.

(A) Timeline for experiments.

(B–D) Total distance traveled (B: WT, 89.8 ± 7; cKO, 82.4 ± 4), total time spent moving ≥3 cm/s (C: WT, 20.7 ± 1 min; cKO, 19.5 ± 0.9 min), and mean velocity (D: WT, 3.8 ± 0.3; cKO, 3.4 ± 0.1) during behavior sessions.

(E and F) Baseline power spectral density from pre-experience home cage during active periods (E: velocity>3 cm/s for at least 2.5 s) and immobility periods (F: velocity <1 cm/s for 3–60 s).

(G–I) Mean theta band (6–12 Hz) oscillatory power (G: WT, 106.5 ± 7; cKO, 99.9 ± 8), mean theta frequency (H: WT, 8.71 ± 0.12 Hz; cKO, 8.75 ± 0.06 Hz), and peak theta frequency (I: WT, 9.16 ± 0.10 Hz; cKO, 9.01 ± 0.10 Hz) during active periods in the pre-experience home cage.

(J and K) Low-gamma-band (J: 25–50 Hz; WT, 68.5 ± 4; cKO, 65.8 ± 6) and high-gamma-band (K: 65–140 Hz; WT, 46.1 ± 2; cKO, 48.4 ± 4) oscillatory power for WT and KIBRA cKO mice during active periods in the pre-experience home cage session.

(L) Theta phase/low-gamma-amplitude cross-frequency coupling (phase/amplitude coupling [PAC]). Left: theta phase vs. normalized gamma amplitude (20° bins). Right: PAC measure (modulation index [MI]⁵⁰: WT, 0.0106 ± 0.0005; cKO, 0.0114 ± 0.0006).

(M) Theta-phase/high-gamma-amplitude coupling (MI: WT, 0.0118 ± 0.0007; cKO, 0.0126 ± 0.0005).

(N) Example of raw and ripple-filtered traces depicting detected SWRs in WT (top) and cKO (bottom) mice. Scale bars: 100 ms, 200 μV; same scale for WT and cKO traces.

(O) Pre-experience SWR rate by detection threshold (standard deviations above the mean; WT, 2 SD 0.76 ± 0.02, 3 SD 0.57 ± 0.02, 4 SD 0.43 ± 0.01, 5 SD 0.31 ± 0.01, 6 SD 0.22 ± 0.01, 7 SD 0.14 ± 0.01; cKO, 2 SD 0.76 ± 0.05, 3 SD 0.53 ± 0.05, 4 SD 0.38 ± 0.05, 5 SD 0.26 ± −0.04, 6 SD 17 ± 0.03, 7 SD 0.1 ± 0.02).

(P) Ripple-band power spectrograms averaged across all detected SWRs for WT (top) or cKO (bottom) mice, scale bar indicates Z-scored power.

(Q–S) Baseline ripple properties during pre-experience home cage immobility using detection threshold of 5 SDs above mean SWR power; SWR rate (Q: WT, 0.24 ± 0.01; cKO, 0.22 ± 0.01 events/s immobile), mean duration of SWR events (R: WT, 0.13 ± 0.004; cKO, 0.13 ± 0.006 s), and peak SWR power (S: WT, 7.15 ± 0.19; cKO, 6.96 ± 0.16). Data values reported are mean ± SEM. Statistics: unpaired Mann-Whitney rank-sum test (B) or unpaired t test (C and D, G–M, Q–S). Two-group test for equal spectra (E and F). Multiple Mann-Whitney rank-sum test with Holm-Šidák correction (O). All comparisons were not significant (WT vs. cKO, p > 0.1). n = 8 (cKO)/10 (WT) mice.

We first quantified basal hippocampal network dynamics by analyzing local field potential (LFP) oscillatory power within area CA1, the primary output of the hippocampus, across several well-studied frequency ranges important for organizing hippocampal information processing and coordinating inter-regional communication. As hippocampal oscillations across several frequency bands are influenced by recording site position relative to the pyramidal layer, we determined the relative electrode location(s) for each mouse using the sharp-wave component of the SWR complex, which reverses across the pyramidal layer.⁵⁴ We found no difference in CA1 sharp-wave deflection between WT and KIBRA cKO mice (Figures S2D and S2E), indicating that recording locations relative to the CA1 pyramidal layer are similarly distributed across genotypes. Thus, any alterations in network properties observed between WT and KIBRA cKO mice are unlikely to trivially arise due to systematic differences in recording depth and instead likely reflect biological distinctions.

Analysis of hippocampal activity during the pre-experience home cage session revealed no differences between genotypes in theta power or mean or peak oscillatory frequency (Figures 1G–1I), no difference in low or high gamma oscillatory power (Figures 1 J and 1K), and no difference in theta-gamma phase-amplitude coupling (Figures 1L and 1M). Furthermore, in the home cage prior to novel experience, we observed no difference in SWR duration, peak ripple power, or SWR rate at any detection threshold (Figures 1N–1S). Thus, basic network mechanisms of regional synchronization and SWR generation are intact in KIBRA cKO mice, consistent with normal basal synaptic transmission in synapses lacking KIBRA.^8,10

KIBRA is required for experience-induced changes in SWRs that reflect information updating

Having established that basal circuit properties are largely normal in KIBRA cKO mice, we next explored whether the loss of KIBRA affects experience-dependent changes in circuit-level hippocampal function. The occurrence rate of SWRs has been shown to transiently increase following salient experience,^55–57 an effect that is blocked by systemic pharmacological inhibition of NMDA receptors (NMDARs).⁵⁸ Given KIBRA’s critical role in regulating NMDAR-dependent synaptic plasticity,^8–10 we hypothesized that experience-induced increases in SWR drive would be impaired in KIBRA cKO mice. However, both WT and KIBRA cKO mice displayed similar experience-dependent modulation of SWR rate (Figures 2A and 2B). Notably, this increase in pre- to post-experience SWR drive was independent of the threshold used to detect SWRs (Figure S3). These findings suggest that KIBRA expression in excitatory hippocampal neurons is unnecessary for experience-dependent changes in SWR drive and argue that the source for this plastic change resides elsewhere.

Figure 2. Experience-dependent modulation of CA1 SWR properties is impaired in KIBRA cKO mice, while modulation of ripple abundance is intact.

(A) Pre- and post-experience SWR abundance shown in 5-min segments for a representative WT (gray, top) and cKO (purple, bottom) mouse.

(B) SWR rate (events/s during immobility) in pre- and post-experience home cage sessions (main effect of experience p < 0.0001, pre vs. post WT p < 0.0001, pre vs. post cKO p < 0.0001; WT, pre 0.24 ± 0.01, post 0.37 ± 0.02; cKO, pre 0.22 ± 0.01, post 0.33 ± 0.02).

(C) Example trace showing experience-dependent increases in SWR power and duration in WT but not cKO mice. Scale bar: 100 ms, 200 mV.

(D) Peak SWR power during pre- and post-experience home cage sessions (main effect of experience p = 0.0284, pre vs. post WT p = 0.0168, pre vs. post cKO p = 0.8335; WT, pre 0.13 ± 0.004, post 0.14 ± 0.005; cKO, pre 0.13 ± 0.006, post 0.13 ± 0.004).

(E) SWR duration during pre- and post-experience home cage sessions (main effect of experience p = 0.0279, experience × genotype p = .0656, pre vs. post WT p = 0.0091, pre vs. post cKO p = 0.9476; WT, pre 7.15 ± 0.19, post 7.55 ± 0.19; cKO, pre 6.96 ± 0.16, post 7.04 ± 0.19). Data values reported are mean ± SEM. Statistics: two-way repeated measures (RM) ANOVA with Šidák’s multiple comparisons. ****p < 0.0001, ***p < 0.001, **p < 0.01, *p < 0.05. n = 8 (cKO)/10 (WT) mice per group. See also Figures S3–S5.

We next assessed whether KIBRA expression was necessary for experience-dependent changes in the information content of SWRs. Experience dramatically changes the representation of information within SWR-based replay.^59–61 Like the experience-dependent change in SWR drive, incorporation of novel information into SWRs can be blocked by systemic pharmacological inhibition of NMDAR-dependent synaptic plasticity.⁶¹ Neuronal synchrony, which contributes to oscillatory power, also increases during SWRs after salient experience.^24,55,62 In addition, replays representing longer spatial trajectories are encoded by SWRs of longer duration,⁶³ and long-duration SWRs are increased during memory tasks.³¹ Thus, we hypothesized that, if features of experience were incorporated into SWR-based replay via a KIBRA-dependent plasticity mechanism, we would observe an increase in the duration of SWRs and the power of the ripple component of SWRs in WT mice but not in mice lacking KIBRA expression. Supporting our prediction and consistent with prior data in rodents,^56,64,65 WT mice displayed a significant increase in both ripple power and SWR duration in the post-experience rest period (Figures 2C–2E). However, ripple power and SWR duration were unchanged following novel experience in KIBRA cKO mice (Figures 2C–2E), consistent with impairments in synaptic plasticity following KIBRA loss.^8–10

During SWRs, hippocampal networks are coordinated via the 25- to 50-Hz low gamma oscillation.^60,66 Low gamma power is transiently increased in hippocampal area CA1 during SWRs, and the representation of information during SWR-based replay is temporally segmented by hippocampal low gamma.^60,66 We therefore examined ripple-associated low gamma power (Figure 3). We observed a transient increase in low gamma power during SWRs from both WT and KIBRA cKO mice (Figures 3A–3C). During pre-experience home cage periods, there was no significant difference in SWR-associated low gamma power between genotypes (Figure 3D), in agreement with a lack of basal changes in synaptic function and SWR properties in KIBRA cKO mice^8,10 (Figures 1O–1S). However, the experience-induced increase in SWR-associated low gamma power was significantly impaired in KIBRA cKO mice (Figure 3D). These data provide further support for the hypothesis that plasticity deficits following KIBRA loss impair increased network synchrony and incorporation of salient information content in SWRs.

Figure 3. Experience-dependent increase SWR-associated low gamma power is impaired in KIBRA cKO mice.

(A and B) Transient increase in low gamma power during SWR events recorded from area CA1 in pre- and post-experience home cage for (A) WT and (B) cKO mice. Low gamma power was averaged across 100-ms bins within 450 ms of SWR onset (data plotted at bin center and normalized to −450- to −350-ms bin for each SWR).

(C) Time-frequency spectrograms showing experience modulation of SWR-associated low gamma power in WT (top) and cKO (bottom) mice, scale bar to right indicates Z-scored power.

(D) Low gamma power at peak ripple power during SWR (main effect of experience p < 0.0001; experience × genotype interaction p = 0.0095; pre WT vs. cKO p = 0.9916; post WT vs. cKO p = 0.0469; WT, pre 1.00 ± 0.07, post 1.61 ± 0.06, pre vs. post p < 0.0001; cKO, pre 0.99 ± 0.09, post 1.29 ± 0.15, pre vs. post p = 0.0018). Data values reported are mean ± SEM. Statistics: twoway RM ANOVA with Šidák’s multiple comparisons (D). *p < 0.05. n = 8 (cKO)/10 (WT) mice per group.

Deficits in SWRs begin to emerge during experience and are not state specific

The above results indicate that novel experience generates plastic changes in the hippocampal network that are reflected in altered SWR features during post-experience rest and that KIBRA is necessary to either induce or maintain those changes. To dissociate these two possibilities, we examined SWRs that occurred during brief pauses within the novel experience itself (“on-task” SWRs). Because incorporation of task-relevant information into on-task replay occurs rapidly²⁵ and requires synaptic plasticity,⁶¹ we hypothesized that WT mice should display modification of SWRs during behavior, whereas such modifications should be impaired or absent in KIBRA cKO mice. Indeed, when we quantified features of SWRs arising during the novel experience, we observed that KIBRA cKO mice had significantly weaker ripple power than WT littermates (Figure 4A). These data are consistent with a role for KIBRA in rapid, plastic changes within the hippocampal network during experience, changes that normally persist into post-experience periods of memory consolidation. Unexpectedly, SWR duration and SWR modulation of low gamma power were similar across genotypes for on-task SWRs (Figures 4C–4E), with differences in these measures emerging in post-experience rest (Figures 2 and 3). On-task SWR rate was variable and not significantly different in KIBRA cKO mice during novel exploration (Figure 4B).

Figure 4. Loss of KIBRA impairs awake CA1 SWR properties during novel experience.

(A) Z-scored peak ripple power (p = 0.0377, WT, 7.17 ± 0.29; cKO, 6.33 ± 0.16).

(B and C) (B) SWR rate (p = 0.0769, WT, 0.15 ± 0.04; cKO, 0.07 ± 0.02) and (C) SWR duration (p = 0.6240, WT, 0.13 ± 0.005; cKO, 0.13 ± 0.006) for SWRs with 3-SD threshold that occurred in periods of immobility during a novel experience.

(D) SWR-associated increase in low gamma power.

(E) Low gamma power at peak ripple power (p = 0.9420, WT, 1.08 ± 0.15; cKO, 1.10 ± 0.15). Data values reported are mean ± SEM. Statistics: unpaired t test (A–C and E), *p < 0.05. cKO n = 7 mice (A and C–E), n = 8 mice (B, note that one mouse had no SWRs during novel experience, so the SWR rate of 0 was included in B, but excluded from all other analyses in this figure); WT n = 10 mice. See also Figure S3.

SWRs that occur during quiet wake are both functionally and mechanistically distinct from those that arise during slow-wave sleep.^33,67 Our analyses during home cage recording sessions combined all SWRs irrespective of behavioral state (quiet wakefulness vs. sleep). To gain insight into whether KIBRA deletion affects SWRs that occur during sleep, we defined epochs of putative nonrapid eye movement (NREM) sleep based on extended immobility duration and hippocampal oscillatory dynamics.⁶⁸ We observe no difference in overall putative sleep durations or sleep architecture across genotypes (Figure S4). WT mice display a similar experience-dependent increase in SWR duration, power, and rate for SWRs occurring in putative sleep (Figure S5). The duration and power of SWRs in KIBRA cKO mice do not change from pre to post experience in putative NREM sleep, whereas the experience-dependent increase in SWR rate is intact (Figure S5), similar to results with mixed-state analyses (Figure 2).

KIBRA is required for experience-induced recruitment of hippocampal place cell activity within SWRs

We next examined the activity of individual hippocampal neurons before, during, and after the novel experience. We quantified the spatial firing fields⁶⁹ for hippocampal place cells during exploration of the novel arena (Figure 5A). Place cells from WT and KIBRA cKO mice had similar overall firing rates (Figure 5B) and mean in-field firing rates (Figure 5C), although WT neurons had slightly but significantly higher peak firing rates (Figure 5D). While both genotypes had similar numbers of place fields per cell (Figure 5E), the place-field sizes for WT neurons were significantly more spatially restricted (Figure 5F), leading to an increased amount of information per spike (Figure 5G) compared with KIBRA cKO neurons. We next assessed recruitment of neural activity into SWRs. Both WT and KIBRA cKO mice showed similar overall per-cell firing rates in pre-experience SWRs (Figure 5H), consistent with our prior findings that basal network function is normal following loss of KIBRA. However, while WT mice significantly increased overall percell firing rate (Figure 5I) and cell participation (Figure 5J) from pre- to post-experience in SWRs, consistent with prior work,⁶¹ KIBRA cKO mice failed to show a similar experience-dependent recruitment of neurons into SWRs.

Figure 5. KIBRA is required for experience-induced increase in hippocampal pyramidal cell participation in SWRs.

(A) Example place fields for five cells each from WT and KIBRA cKO mice. Arena dimensions 40 × 60 cm. Heatmap scale bar, 0 to maximum firing rate, where the maximum firing rate (Hz) for each cell is indicated in the top right corner.

(B) Mean firing rate during exploration of a novel environment for each cell with an identified place field (WT, 0.98 ± 0.22 Hz; cKO, 1.04 ± 0.18 Hz; WT vs. cKO p = 0.5336).

(C) Mean in-field firing rate during active exploration (WT, 2.34 ± 0.30 Hz; cKO, 1.72 ± 0.23 Hz; WT vs. cKO p = 0.0595).

(D) Maximum in-field firing rate during active exploration (WT, 4.94 ± 0.66 Hz; cKO, 3.52 ± 0.42 Hz; WT vs. cKO p = 0.0410).

(E) Number of place fields per cell in novel environment shown in (A) (WT, 1.70 ± 0.22; cKO, 1.74 ± 0.19; WT vs. cKO p = 0.9091).

(F) Mean place-field size for each cell (WT, 420.2 ± 105.5 cm²; cKO, 689.2 ± 87.0 cm²; WT vs. cKO p = 0.0351).

(G) Information (bits) per spike for each WT and KIBRA cKO cell with identified place field(s) (WT, 2.05 ± 0.22; cKO, 1.46 ± 0.20; WT vs. cKO p = 0.0416).

(H) Average participating cell firing rate during pre-experience SWRs (WT, 12.14 ± 0.69 Hz; cKO, 12.44 ± 0.68 Hz; WT vs. cKO p = 0.9463).

(I) Experience modulation of firing rate in SWRs (for each cell, firing rate in post-experience SWRs/firing rate in pre-experience SWRs. WT, 2.54 ± 0.54; cKO, 1.30 ± 0.14; WT vs. cKO p = 0.0043. One-sample Wilcoxon test vs. 1; WT, p = 0.0002; cKO p = 0.0635).

(J) Experience modulation of cell participation in SWRs (fraction of SWRs in which a cell fired at least one action potential, post-/pre-experience. WT, 2.24 ± 0.45; cKO, 1.17 ± 0.10; WT vs. cKO p = 0.0010 One-sample Wilcoxon test vs. 1, WT p < 0.0001, cKO p = 0.1663). (B–G) WT, 20 cells from six mice; cKO, 31 cells from five mice; cells with peak firing rate <1 Hz were excluded from these analyses. (H–J) WT, 24 cells from six mice; cKO, 36 cells from five mice. Note that consistent results were observed when analyses in (I) and (J) were restricted to neurons with place fields (experience modulation of firing rate in SWRs: WT = 2.52 ± 0.63, cKO = 1.35 ± 0.16, WT vs. cKO p = 0.0153. Experience modulation of cell participation in SWRs: WT = 2.14 ± 0.47, cKO = 1.19 ± 0.11, WT vs. cKO p = 0.0029). Statistics: Mann-Whitney rank sum (B–J) and one-sample Wilcoxon (I and J).

Intra-hippocampal coordination during SWRs requires KIBRA

During SWR-based replay, hippocampal areas CA3 and CA1 are tightly synchronized to facilitate information transfer.⁶⁶ CA3 input to CA1 is coordinated by the low gamma oscillation⁷⁰ and is crucial for organized, sequential activation of CA1 place cells during SWRs.⁷¹ Given prior work demonstrating a role for KIBRA in plasticity at CA3→CA1 synapses^8–10 and our data demonstrating that KIBRA cKO mice show impaired experience-dependent modulation of SWR-associated low gamma power (Figure 3), we hypothesized that coordination between CA1 and CA3 would be disrupted after loss of KIBRA. In both WT and KIBRA cKO mice, ripple power and low gamma power were transiently elevated in area CA3 during CA1-identified SWRs in pre-experience rest (Figure S6), and no difference was observed between genotypes in the basal CA3 low gamma power outside of SWRs (Figure 6A). Similar to area CA1, area CA3 of WT mice displayed an experience-dependent increase in both ripple power and SWR-associated low gamma power, an effect that was abolished in KIBRA cKO mice (Figures 6B and 6C), indicating that KIBRA loss produced similar impairments in experience-dependent changes in both CA1 and CA3 networks.

Figure 6. Loss of KIBRA impairs CA3-CA1 communication and experience modulation of CA3 ripple and low gamma power during SWRS.

(A) Baseline CA3 low gamma power in pre-experience rest (p = 0.3185; WT, 60.0 ± 3; cKO, 47.6 ± 10).

(B) Experience modulation of peak CA3 low gamma power during identified CA1 SWRs, shown as ratio of post/pre experience (p = 0.0411; WT, 1.71 ± 0.17; cKO, 1.10 ± 0.18).

(C) Experience modulation of peak CA3 ripple power during CA1 SWRs shown as ratio of post/pre experience (p = 0.0428; WT, 1.22 ± 0.06; cKO, 0.91 ± 0.13).

(D) Probability normalizied polar histogram (10° bins) showing distribution of CA3-CA1 low-gamma-phase offsets in SWRs during pre- (left) and post-experience (right) home cage sessions for all WT (gray) and KO (purple) SWRs. To combine phase-offset distributions across animals, normalized data from each mouse were rotated to set the peak phase offset bin to 0 (Rayleigh test for uniformity: WT and KO, pre and post p < 0.0001. Kuiper’s two-sample test: WT vs. KO distribution, pre p = 0.004, post p = 0.002).

(E and F) Across both pre- and post-experience sessions, mean CA3-CA1 phase coherence (ISPC) across SWRs (E: WT = 0.50 ± 0.26, cKO = 0.23 ± 0.06, p = 0.0457) and within SWRs (F: WT = 0.55 ± 0.11, cKO = 0.47 ± 0.02, p = 0.0381). Statistics: unpaired t test (A–C and E), with Welch’s correction (A and E). Mann-Whitney rank-sum test (F). Rayleigh test for uniformity, Kuiper two-sample test with Bonferroni-Holm corrections for multiple comparisons (D). WT n = 6, cKO n = 4 mice (A–C, E, and F). (D) n = # ripples. WT: pre (n = 3,894), post (n = 7,432). cKO: pre (n = 2,262), post (n = 3,002). See also Figures S6 and S7.

We next examined the precise temporal synchronization between CA1 and CA3 during SWRs. As low gamma oscillations are known to facilitate communication between CA3 and CA1⁷⁰ and discretize information during SWR-based replay,⁶⁰ we examined phase coupling of the low gamma oscillation across CA3 and CA1 networks. KIBRA cKO mice display substantially disrupted synchronization in both pre- and post-experience SWRs as indicated by a less polarized distribution of low-gamma-phase offsets (Figure 6D). We determined low-gamma-phase coherence across all SWRs for each mouse by calculating the mean phase offset between CA1 and CA3 for each SWR and calculating the inter-site phase clustering (ISPC) value of this distribution for each mouse (Figure 6E). These analyses demonstrated significantly reduced low-gamma-phase coherence between CA1 and CA3 in KIBRA cKO mice (Figure 6E). Importantly, it is possible that individual SWRs are tightly synchronized between CA1 and CA3 in KIBRA cKO mice but that each SWR is synchronized with a distinct phase offset, resulting in a more uniform distribution (Figure 6D) and lower ISPC (Figure 6E) across SWRs. To determine phase coherence within individual SWRs, we quantified a single ISPC value between CA1 and CA3 for every SWR and averaged those values for each mouse. This analysis again revealed significant reduction in low gamma coherence following KIBRA loss (Figure 6F). These data together suggest that loss of KIBRA produces a consistent reduction of intra-hippocampal coordination within individual SWRs.

Experience synchronizes hippocampal SWRs and cortical spindles during subsequent NREM sleep in a KIBRA-dependent manner

To assess the role of KIBRA in systems-level memory consolidation, we examined whether the coordination of ACC sleep spindles with CA1 SWRs is altered by the loss of KIBRA. We find no significant differences between WT and cKO mice in the sigma band power used for spindle detection (Figures 7A and 7B) or basal spindle properties in pre-experience rest, including spindle rate, duration, or power (Figures 7C–7E). We find that spindle density and duration increase with experience in WT mice, consistent with prior findings^72,73 (Figures 7F and 7G). However, while KIBRA cKO mice exhibit an increase in spindle rate, they fail to show experience-dependent changes in cortical spindle duration (Figures 7F and 7G). Interestingly, while we did not observe experience-dependent increases in power across all ACC spindles in WT or cKO mice (Figure 7H), when we restricted our analysis to ACC spindles occurring during identified hippocampal SWRs, we observed a significant experience-induced increase in spindle power in WT mice that was not present in cKO mice (Figure 7I). Several studies have demonstrated coupling of hippocampal SWR and neocortical spindle events during post-learning sleep.^39,45,74 Given the impairments in cross-regional (CA3-CA1) synchronization observed in KIBRA cKO mice (Figure 6), we predicted that hippocampal-cortical coordination might also be altered after loss of KIBRA. Therefore, we examined time lags from SWRs to the nearest spindle and found a shift closer to zero lag from pre to post experience in WT mice (Figures 7J and 7L), indicating enhanced synchronization of these two brain regions. However, this improvement in inter-regional coordination is absent in KIBRA cKO mice (Figures 7K and 7L). As SWRs tend to be nested within the troughs of spindles,⁷⁴ we next examined the properties of SWRs that were nested inside spindles. The proportion and peak ripple power of SWRs nested within spindles increased after experience in WT but not cKO mice (Figures 7M and 7N). Finally, we found that the number of cortical spindles that overlapped with multiple consecutive hippocampal SWRs increased across experience in WT but not cKO mice (Figure 7O). Importantly, this is not a trivial result of increased SWR and spindle rate, as both WT and KIBRA cKO mice show elevated SWR (Figures 2A and 2B) and spindle (Figure 7F) rate. Together these findings indicate that KIBRA is necessary for experience-dependent enhancement of SWR-spindle coordination.

Figure 7. Coordination of ACC spindles and CA1 SWRs during sleep is disrupted by loss of KIBRA.

(A) Example of a spindle event detected in the ACC. Top: raw LFP during spindle event. Middle: sigma-filtered trace (10–15 Hz). Bottom: root-mean-square (RMS) transformed trace used for spindle detection (black) and thresholds used to define occurrence, onset, and offset (red). Scale bars: raw LFP 500 μV, sigma filtered 200 μV, transformed 1 × 10⁶ μV³.

(B) ACC sigma power in pre-experience home cage sessions during all immobility periods (p = 0.1416; WT, 47.4 ± 3; cKO, 50.1 ± 1).

(C–E) Pre-experience baseline properties of identified spindles: (C) Rate (spindle events per second; WT vs. cKO, p = 0.1571; WT, 0.11 ± 0.01; cKO, 0.09 ± 0.01); (D) mean spindle duration (WT vs. cKO, p = 0.6135; WT, 1.81 ± 0.06; cKO, 1.85 ± 0.05); (E) Peak sigma power during spindles (WT vs. cKO, p = 0.8865; WT, 4.47 ± 0.07; cKO, 4.58 ± 0.19).

(F–H) Experience modulation of ACC spindles: (F) spindle rate (main effect of experience p = 0.0002; WT pre vs. post p = 0.0009; cKO pre vs. post p = 0.0453; WT, pre 0.11 ± 0.01, post 0.13 ± 0.01; cKO, pre 0.09 ± 0.01, post 0.10 ± 0.01); (G) spindle duration (main effect of experience p = 0.0631; WT pre vs. post p = 0.0343, cKO pre vs. post p = 0.9676; WT, pre 1.81 ± 0.06, post 2.00 ± 0.08; cKO, 1.85 ± 0.05, post 1.87 ± 0.07); (H) peak sigma power across all spindles (WT pre vs. post p = 0.9729; cKO pre vs. post p = 0.6787; WT, pre 4.47 ± 0.07, post 4.45 ± 0.06; cKO, pre 4.58 ± 0.19, post 4.50 ± 0.09).

(I) Peak ACC sigma power for spindles co-occurring with CA1 SWRs (main effect of experience, p = 0.0002; experience × genotype, p = 0.0376; WT pre vs. post p = 0.0002, cKO pre vs. post p = 0.1250; WT, pre 1.23 ± 0.10, post 1.47 ± 0.09; cKO, pre 1.12 ± 0.04, post 1.22 ± 0.07).

(J and K) Probability distribution for absolute lag between CA1 SWRs and ACC spindle onset during putative NREM sleep for WT (J) and cKO (K): pre vs. post, WT p < 0.0001, cKO p = 0.1063.

(L) Percentage of SWR events occurring within 2 s of a spindle during post-experience rest (WT vs. KO, p = 0.0217; WT, 0.49 ± 0.03; cKO, 0.36 ± 0.04).

(M) Proportion of ACC spindles that co-occur with CA1 ripples (main effect of experience, p = 0.0007; WT pre vs. post p = 0.0013, cKO pre vs. post p = 0.1694; WT, pre 0.20 ± 0.02, post 0.26 ± 0.02; cKO, pre 0.16 ± 0.01, post 0.19 ± 0.03).

(N) Peak ripple power of SWRs nested within spindles (experience × genotype, p = 0.0308; WT pre vs. post p = 0.0408, cKO pre vs. post p = 0.6521; WT, pre 6.96 ± 0.11, post 7.32 ± 0.06; cKO, pre 7.07 ± 0.24, post 6.94 ± 0.17).

(O) Fraction of spindles containing more than one nested SWR (# spindles containing >1 SWR/# spindles containing ≥1 SWRs; main effect of experience, p = 0.0425; experience × genotype, p = 0.0684; main effect of genotype, p = 0.0964; WT pre vs. post p = 0.0168, cKO pre vs. post p = 0.9806; WT, pre 0.40 ± 0.04, post 0.48 ± 0.03; cKO, pre 0.35 ± 0.03, post 0.35 ± 0.04). Data values reported are mean ± SEM. Statistics: unpaired t test (C, D, and L). Mann-Whitney rank-sum test (B and E). Two-way RM ANOVA with Šidák’s multiple comparisons (F–H, I, and M–O), Wilcoxon rank-sum test on distributions with Bonferroni-Holm correction for multiple comparisons (J and K). n = 8 (WT)/n = 7 (KO) mice per group (B–I and L–O). n = # ripples in J and K;WT, pre (n = 3,969) post (n = 5,262); cKO, pre (n = 3,352) post (n = 4,555). ****p < 0.0001,***p < 0.001,**p < 0.01,*p < 0.05.

DISCUSSION

KIBRA plays a key role in regulating AMPAR trafficking, synaptic plasticity, learning, and memory.^8–11 Substantiating this crucial role are numerous human genetic studies that implicate KIBRA in memory and disorders of complex cognition.^{12,13,16,18,21,75} However, whether and how KIBRA-dependent cellular mechanisms affect in vivo network dynamics were open questions. Here, we examined SWRs before, during, and after salient novel experience in adult mice lacking the synapse-enriched protein KIBRA. Considerable evidence implicates SWR-based replay as a memory consolidation and retrieval mechanism in both rodents and humans. Thus, SWRs serve as a quantifiable measure of circuit-level memory formation and consolidation.^{23,24,26,29,30,33} We find that, under pre-experience baseline conditions, SWRs are normal following removal of KIBRA, consistent with our previous findings that basal synaptic transmission is unaffected in these mice.¹⁰ However, loss of KIBRA abolishes experience-dependent changes in several SWR properties, including duration, ripple-band power, and SWR-associated increases in gamma-band power. KIBRA cKO mice also lack experience-dependent recruitment of neural activity in SWRs. In addition, KIBRA cKO mice display impaired basal intra-hippocampal (CA3-CA1) coordination at the low gamma frequency during SWRs and fail to produce experience-dependent changes in hippocampal-cortical communication during SWRs. These experiments were performed in a conditional knockout mouse line in which KIBRA was selectively deleted from mature, predominantly excitatory forebrain neurons (i.e., CaMKIIα+ neurons),^10,49 arguing against altered neuronal development as an explanation for these observations.

Thus, these results collectively indicate that KIBRA is acutely required for experience-dependent SWR modulation and that KIBRA-dependent mechanisms facilitate synchronization across inter-regional networks. Furthermore, while the current study was not designed to provide behavioral assessment of long-term spatial memory, these data further suggest that KIBRA-dependent deficits in updating SWR information content and hippocampal-cortical communication during SWRs may contribute to memory impairments resulting from loss of KIBRA.^8–11,22

Novel experience has been shown to increase SWR drive in post-experience rest periods.^55,56,58,76 In vitro slice experiments demonstrate that stimuli that establish LTP can facilitate the spontaneous generation of SWRs.⁷⁷ Consistent with a role for synaptic plasticity in experience-dependent changes in SWR expression, the increase in SWR drive is abolished by pharmacological blockade of NMDARs during the experience.⁵⁸ The experience-dependent increase in SWR drive observed in both WT and KIBRA cKO mice in our study indicates that the site of NMDAR-dependent plasticity that underlies this increased drive is unlikely to reside in synapses onto excitatory neurons of the hippocampus. Rather, this plasticity may occur at CA3 inputs onto CA1 inhibitory neurons. Indeed, both experimental and computational work suggest that excitatory input from CA3 pyramidal neurons onto local parvalbumin-positive (PV⁺) basket cells may trigger SWR initiation via generation of high-frequency oscillatory firing of inhibitory populations.^78,79 Our findings therefore suggest that excitatory input onto PV⁺ inhibitory neurons is a likely site of experience-dependent increases in SWR drive. As KIBRA was not deleted in inhibitory neurons and given the substantially lower expression of KIBRA in inhibitory neurons, whether experience-induced changes in SWR drive are modulated by KIBRA remains to be determined.

In addition to changes in SWR drive, considerable evidence supports a role for synaptic plasticity across the hippocampal network in the incorporation of experience-relevant information in SWR-based replay. The activity patterns of hippocampal pyramidal cells with overlapping firing fields produce NMDAR-dependent LTP in ex vivo hippocampal slices⁵³ and increase the probability for those neurons to co-fire in subsequent offline SWRs.^24,62 In agreement with these findings, pharmacological inhibition of NMDAR function during experience prevents coherent representation of the current experience in subsequent SWR-based replay, indicating that NMDAR function is important for establishing synaptic plasticity during experience that is later revealed in the experience-specific content of SWR-based replay.⁶¹ Changes in the information content of SWRs following novel experience are reflected in increased overall firing rate and ripple power.^56,62 SWR duration is correlated with the length of virtual trajectories,⁶³ and long-duration SWRs are increased during memory tasks.³¹ Thus, experience-dependent changes in SWR duration and ripple power are likely both driven by synaptic plasticity. However, NMDAR blockade broadly affects multiple forms of synaptic plasticity across many cell types. We observe increased ripple power and SWR duration following salient experience in WT mice, whereas KIBRA cKO mice fail to show such experience modulation of SWRs, consistent with the hypothesis that incorporation of recent experience into hippocampal SWRs requires KIBRA-dependent synaptic plasticity mechanisms within the excitatory pyramidal neurons of the hippocampal network. We further demonstrate that experience-dependent increases in firing rate and cell participation in SWRs observed in WT mice⁶¹ are absent in KIBRA cKO mice. While we did not record from the large single-unit populations necessary to unambiguously quantify replay content,⁸⁰ we cautiously interpret these results to support the hypothesis that KIBRA plays an important role in allowing the hippocampal network to organize coherent, experience-relevant replay.

The information content in SWR-based replay is temporally organized by the low gamma oscillation,⁶⁰ the power of which transiently increases during SWRs.^38,60,66 Gamma activity during SWRs is strongly impaired in an age-dependent fashion in a mouse model of Alzheimer’s disease,³⁸ suggesting a role for SWR-embedded low gamma in mnemonic processes. While stronger gamma coherence across the CA3-CA1 networks during SWRs is predictive of the presence of coherently structured replay,⁶⁶ whether experience directly affected the low gamma-SWR relationship was unclear. Here, we show that the transient increase in gamma activity during SWRs is significantly elevated following a novel experience. Hippocampal neurons are phase-locked to low gamma during SWRs^60,66 and experience increases overall firing rates during SWRs⁶²; thus, the enhancement in gamma power we observe during SWRs in WT mice is consistent with these prior findings. Although KIBRA cKO mice display a normal relationship between low gamma power and SWRs during pre-experience rest periods, the absence of KIBRA impairs experience-dependent changes in this relationship.

While synaptic plasticity during experience can influence subsequent SWRs, the firing patterns within SWRs can themselves produce LTP,³ which is believed to facilitate information transfer from hippocampal to cortical networks and underlie long-term, systems-level memory consolidation.⁴⁴ Supporting this model, blocking SWRs during post-experience rest impairs long-term memory.^30,32,81,82 During SWRs, the hippocampal and cortical networks become more tightly synchronized.^45,83 In particular, cortical sleep spindles are temporally coupled to hippocampal SWRs.^39,40,45,46 Disrupting this relationship impairs sleep-based memory consolidation, while enhancing SWR-spindle coupling improves memory.^39,40,45,46 We observe that the duration of cortical spindles and the strength of SWR-associated cortical spindles increases across experience in WT mice but not in KIBRA cKO mice. Furthermore, we observe an experience-dependent enhancement of the temporal coupling between hippocampal SWRs and cortical spindles in WT but not cKO mice. While the present study was not designed to address prolonged, multi-day plastic changes across the hippocampal or cortical networks, we hypothesize that the lack of experience-dependent changes in SWR-spindle coordination likely results in impaired memory consolidation and may account for the long-term memory impairments observed in mice lacking KIBRA expression.^8,10,11

In contrast to SWR properties and SWR-spindle coupling, which were not different between WT and KIBRA cKO mice in pre-experience conditions, CA3-CA1 communication during SWRs was abnormal in cKO mice even during this baseline period. Consistent with prior work in rats,⁶⁶ we observe strong coherence across hippocampal areas CA3 and CA1 on a low gamma timescale during SWRs in WT mice. This coherence was significantly weaker in cKO mice, demonstrating impaired communication between these regions. Given the lack of basal changes in synaptic function observed in KIBRA cKO,^8,10 it is unclear what accounts for this impaired communication. Both areas CA1 and CA3 display similar overall gamma power, and a normal SWR-based increase in gamma power is observed prior to novel experience in KIBRA cKO mice; thus, the lack of low gamma coherence is unlikely to be a result of impairments in the mechanisms of gamma generation. Rather, the lack of coherence may result from a shift in the balance of CA1 drive from CA3 inputs relative to layer 3 entorhinal cortex (EC3) inputs.^68,70,84

It remains unclear how experience modifies the hippocampal network to allow for specific sequences to be re-expressed in SWR-based replay or how specific plasticity mechanisms regulate the expression of these sequences during different stages of memory acquisition and consolidation. Our data are consistent with a model in which novel or salient experience drives rapid KIBRA-dependent synaptic plasticity across the hippocampal and cortical networks and that these plastic changes fundamentally alter the properties and content of SWR-based hippocampal replay and facilitate enhanced hippocampal-cortical communication during subsequent sleep. As KIBRA is known to regulate AMPAR trafficking and expression at excitatory synapses,^8–10 the lack of experience-dependent changes in SWR function and inter-regional coordination are likely a result of impaired synaptic plasticity at these sites; however, our data cannot rule out contributions from other neuronal functions of KIBRA. Our findings are consistent with the hypothesis that synaptic plasticity at excitatory CA3 synapses onto PV⁺ basket cells may facilitate experience-dependent increases in SWR drive, while plasticity at excitatory synapses on CA1 pyramidal neurons is likely responsible for changes in SWR content. As KIBRA, plasticity, and SWRs are commonly linked to cognitive and neuropsychiatric disorders,^{14,18,21,36–38,75} additional work investigating how these mechanisms converge to regulate information processing will provide critical mechanistic insights into both normal and pathological brain function.

Limitations of the study

Our findings identify differences in several features of network activity between WT and KIBRA cKO mice using LFP measurements. Differences in electrode recording location along the radial axis of the hippocampus can affect many of the properties that we have measured,⁸⁵ and thus systematic differences in recording locations between WT and cKO mice could produce differences that do not reflect the underlying biology. We use the deflection of the sharp-wave component of the SWR complex as a measurement of radial recording depth in CA1, as the sharp wave reverses across the pyramidal layer.^54,85 We observe no difference in average sharp-wave amplitude across genotypes, arguing against a systematic bias in recording locations between WT and KIBRA cKO mice. In addition, radial depth has a substantial impact on the power of the theta oscillation,⁸⁶ and the lack of difference in this value between WT and cKO mice under pre-experience conditions further argues that recording site location is similar across genotypes. Finally, the majority of our measurements compare pre with post experience in the same mouse, an experimental strategy that minimizes the potential impact of recording site variability between mice.

The change in LFP oscillations across the radial axis of the hippocampus arises due to the spatial distribution of sinks and sources,^85,86 and it is possible that the differences we observe between WT and KIBRA cKO mice may arise due to altered topographical organization of CA3 projections. We consider this unlikely for several reasons. First, in our mouse model, KIBRA was deleted after mice reached adulthood, allowing normal circuit development. Additionally, no macroscale anatomical alterations were identified in the constitutive KIBRA KO mouse.⁸

Only male mice were used in our study, as explained in the section “exclusion criteria” in the STAR Methods. Because of this experimental limitation, our study is not able to determine the role of KIBRA-dependent network plasticity in female mice.

Our data support the hypothesis that KIBRA-dependent plasticity mechanisms in excitatory neurons are essential for experience-induced modification of hippocampal SWR information content and hippocampal SWR-cortical spindle coupling. However, while the primary neuronal function identified for KIBRA is regulation of synaptic plasticity through facilitating activity-induced AMPAR trafficking and expression,^8–10 our study cannot rule out contributions from other cellular functions of KIBRA that do not relate to synaptic plasticity. Additionally, because KIBRA was deleted from CaMKIIα+ neurons throughout the hippocampus and cortex, our study is not able to determine the specific region(s) undergoing experience-induced, KIBRA-dependent plasticity within hippocampal or cortical circuits.

Finally, we observe a non-significant trend toward decreased SWR rate during novel exploration (on-task SWRs). The lack of required behavioral pauses in our task (e.g., pausing at a reward location to consume food or at a forced delay during maze navigation^26,55) likely contributed to large variability in SWR rate. Thus, additional data from a task with designed behavioral pauses will be necessary in future studies to conclusively determine whether KIBRA shows state-dependent differences in experience modulation of SWR rate.

STAR★METHODS

RESOURCE AVAILABILITY

Lead contact

Further information and requests for resources and reagents should be directed to and will be fulfilled by the lead contact, Lenora Volk (Lenora.Volk@utsouthwestern.edu).

Materials availability

This study did not generate new unique reagents.

Data and code availability

• All data reported in this paper will be shared by the lead contact upon request.

• Original code used for data analysis has been deposited at Zenodo and is publicly available as of the date of publication. DOIs are listed in the key resources table.

• Any additional information required to reanalyze the data reported in this paper is available from the lead contact upon request.

KEY RESOURCES TABLE.

REAGENT or RESOURCE	SOURCE	IDENTIFIER
Chemicals, peptides, and recombinant proteins
Tamoxifen	Sigma-Aldrich Selleck Chemicals	Cat# T5648 Cat# S1238
Deposited data
MATLAB code for data analysis	This paper	Zenodo: https://doi.org/10.5281/zenodo.7939187
Experimental models: Organisms/strains
KIBRA cKO	Makuch et al.8; Mendoza et al.10	N/A
CaMKIIα-CreERT2	Erdmann et al.49	N/A
Software and algorithms
MATLAB	MathWorks	https://matlab.mathworks.com/
DeepLabCut	Mathis et al.87	http://www.mackenziemathislab.org/deeplabcut-home
Chronux	Bokil et al.88	http://chronux.org/
Circular Statistics Toolbox	Berens89	https://www.mathworks.com/matlabcentral/fileexchange/10676-circular-statistics-toolbox-directional-statistics
Prism	GraphPad	https://www.graphpad.com/

EXPERIMENTAL MODEL AND STUDY PARTICIPANT DETAILS

The University of Texas Southwestern Institutional Animal Care and Use Committees approved all animal protocols in this study. Male mice ages 2.5–5 months were used for all experiments. A tamoxifen-inducible deletion strategy was used to selectively remove KIBRA from mature, excitatory forebrain neurons. KIBRA cKO (CaMKIIα-CreERT2⁺:KIBRA^flox/flox) and WT littermate controls (CaMKIIα-CreERT2⁻:KIBRA^flox/flox) mice of C57Bl/6N genetic backgrounds were generated by crossing floxed KIBRA mice (KIBRA^flox/flox) with hemizygous CaMKIIα-CreERT2 transgenic mice, which express CreERT2 (Cre recombinase fused to a mutated estrogen receptor) under the CaMKIIα promoter.⁴⁹ KIBRA cKO and WT mice between 2.5 and 4 months of age were injected with tamoxifen (IP, 100 mg/kg, 2x/day, 5 days), resulting in ~70% reduction of KIBRA protein in the hippocampus of KIBRA cKO mice.¹⁰ Mice were allowed to recover for at least two weeks after the last tamoxifen injection before microdrive implantation.

METHOD DETAILS

In vivo electrophysiology

Mice were surgically implanted with custom-made tetrode microdrives (4–4.5g fully assembled) using previously reported surgical procedures.⁶⁰ Microdrive components were designed using Solidworks software, printed using a 3D printer (Formlabs), and assembled using methods adapted from(26). Each microdrive array contained 8 independently adjustable tetrodes (twisted bundles of four 17.8 μm 90% platinum/10% iridium wires, California Fine Wire) bilaterally targeting (in mm) ACC (ML: ±.4-.5 AP: +.5 DV: 1.25), CA1 (ML: ±1.5 AP: −2.28 DV: 1.5), CA3 (ML: ±2.56 AP: −2.08 DV: 2.25) and a reference in overlying cortex. The tetrode tips were electroplated with gold solution (Neuralynx) to reduce the impedance to 150–200 kΩ before surgery. Small jewelry screws (Component Supply) and dental cement were used to anchor microdrives to the skull of the animal. One anchor screw behind lambda, overlying the cerebellum, served as a ground. Subjects were implanted under regulated isoflurane-oxygen anesthesia and fixed in a flat-skull position in a stereotaxic frame. For a majority of recordings, hippocampal tetrodes were individually lowered to their final recording locations over 1–2 weeks until ripples and hippocampal spike activity were identified. Hippocampal tetrodes for a subset of mice (3 WT, 2 cKO) were lowered to the target recording depths during surgery. No differences were observed between these groups in either recording quality or experimental results. ACC tetrodes were lowered to the target recording depths during surgery. Subcutaneous buprenorphine (0.006 mg/mL at 0.01 mL/g body weight) was given as a postoperative analgesic. Post-surgery, all implanted animals were individually housed on a 12 hr light/dark cycle with ad libitum access to chow and water in their home cages.

At the end of experiments, mice were given a lethal dose of Euthasol, and electrolytic lesions were made to confirm recording locations. Mice were then transcardially perfused with ice-cold sterile saline followed by 4% paraformaldehyde. Postmortem electrode locations were verified by examining 75mm brain coronal sections made with a Precisionary Instruments Compresstome (VF-300).

Data acquisition

Data were collected using a Neuralynx (Bozeman, MT) data acquisition system. Recording electrodes were connected to a Digital Lynx recording system via an HS-36 headstage and tether. During electrophysiological recordings the animal’s behavior was monitored by overhead video recorded continuously at 30 FPS and synchronized to electrophysiological data acquisition (Cheetah software). The animal’s position was determined using LEDs or infrared tracking stickers positioned at the front and back of the microdrive array (pointing towards nose or tail, respectively). Analog neural signals were digitized at 32,556 Hz. Continuous local field potential (sampled at 3,255.6 Hz) were recorded using Cheetah data acquisition software (Neuralynx).

Behavior

Mice were handled for at least 5 min on each of 5 consecutive days prior to experiments. Recordings were conducted at least 7 days post-implantation during the animal’s light cycle to ensure adequate sleep in the pre- and post-experience rest periods. Continuous electrophysiological recordings were conducted throughout the following: a 1-h pre-experience home cage session in which subjects were allowed to freely behave in the home cage, mice were then transferred to the center chamber of a novel 40 cm × 60 cm plexiglass box (22 cm tall) with four interior baffles (22 cm tall, angled 10–14.5 cm long) that protruded from long-axis walls and provided visual and tactile cues as mice explored the space. High contrast visual cues were also present on the walls in the room where the recording was performed. Following a 20-min exploration of the novel arena, a novel con-specific male mouse and a novel object were placed into the chamber (10 min) after which the novel object was replaced by an additional novel con-specific mouse (10 min). The unimplanted mice providing social salience were kept in a small wire mesh cage (10 cm tall, 9 cm diameter) which prevented them from causing damage to the recording apparatus on the implanted subject, but allowed for close interaction. The novel behavioral experience was followed immediately by a 1-h post-experience home cage session. Animal positions were quantified from behavior videos via deeplabcut.⁸⁷ Velocity was used to define behavior states, including active movement (velocity >3 cm/s), immobility (<1 cm/s), and sleep (<1 cm/s for at least 60s).

LFP analysis

Analyses were performed using custom MATLAB scripts following accepted analysis methodology. Reference signals were digitally subtracted from hippocampal electrodes prior to analysis. For each tetrode, one representative electrode from each region was selected for local field potential analyses. Theta (6–12 Hz), low gamma (25–50 Hz), high gamma (65–140 Hz), and ripple (125–300 Hz) bands were isolated from the raw LFP by bandpass filtering in the respective frequency bands. Power was defined as the absolute value of the smoothed (Gaussian kernel, sigma = theta 300ms; low gamma 85ms; high gamma 30ms; ripple 12.5ms) Hilbert transform of this filtered signal. 0°/360° was defined as the trough of the filtered oscillation. Removal of 60 Hz line noise was done using spectrum interpolation. Power spectral densities were generated via Chronux using the multitaper method.

Theta frequency

Theta frequency was calculated by identifying peaks in theta filtered LFP and calculating the inverse of the inter-peak interval identified as the time difference between adjacent theta peaks. Peak theta frequency was calculated as the frequency at peak power during periods of activity.

Nested theta-gamma

Theta-gamma phase-amplitude coupling was quantified by extracting the theta phase and gamma amplitude to generate a phase-amplitude distribution as described in Tort et al. 2008.⁵⁰ The strength of coupling or modulation index reflects the deviation of the Theta Phase-Gamma amplitude distribution from a uniform distribution.

SWR-triggered spectrograms

SWR-triggered spectrograms were computed using the short-time Fourier transform with 200ms temporal bins, 190ms overlap, and 2hz frequency resolution. A Z score was computed for each frequency band using the mean and standard deviation of the LFP power during the pre-experience home cage session from CA1. Spectral power was averaged for all SWRs for each animal.

SWR associated gamma power and coherence

SWR associated low gamma power was quantified using 100ms time bins. The change in low gamma power during SWRs was quantified by subtracting the baseline (−450 to −350 preceding each SWR) power from each time point. Gamma synchrony was computed during SWRs by determining the phase offset between CA1 and CA3 tetrodes for all time points during an SWR and using the inter-site phase clustering method (ISPC).⁹⁰ ISPC is defined as the length of the average of phase angle difference vectors from two electrodes. An ISPC value was calculated for each SWR and averaged per animal. For polar histograms, the mean phase offset was calculated for each SWR and values were combined to obtain a distribution of gamma phase offsets for all SWRs. For Figure 6D, to facilitate analysis of CA3-CA1 coherence (consistency of distribution of phase offsets) without confounds from differences in absolute phase offset between animals, the phase offset data for each animal was ‘normalized’ (reoriented) by calculating the peak phase for each CA1-CA3 low gamma phase offset distribution, determining the deviation in peak phase from zero, and shifting the data by that deviation such that the peak phase for each distribution was oriented to zero.

Ripple detection and analysis

Sharp-wave/ripples were detected on CA1-targeted tetrodes as peaks above the mean power (5 SD above the mean unless otherwise noted), in the smoothed (Gaussian, sigma 12.5ms) ripple band (125–300 Hz) power, during only periods of immobility (velocity <1 cm/s). The start and end of each ripple were defined as the point when the smoothed ripple power crossed the mean+1SD. For all SWR events, thresholds were set using the pre-experience home cage session to calculate the mean and SD. SWR rate was calculated as the number of detected events divided by the number of seconds spent immobile. Power and duration were calculated for each SWR event and then values were averaged for all events per subject.

Spindle detection and analysis

Spindle events were detected in the ACC during putative NREM sleep periods as described⁹¹ using increases in the processed, cubed root-mean-square (RMS) transformed sigma power (10–15Hz). Detection thresholds for each mouse were set using the RMS transformed sigma power during quiescent periods in the pre-experience home cage session. Briefly, an upper threshold was used to identify peaks in the cubed RMS signal and a lower threshold was used to define the start and ends. Spindle events shorter than 0.5 s or longer than 10 s were eliminated from analysis. Events with an inter-event interval less than 0.05 s were combined. Spindle rate was calculated as the number of detected events divided by the time spent in NREM sleep. Power and duration were calculated for each spindle event, and then values were averaged for all events per subject. Temporal coordination of ACC spindles with CA1 SWRs was estimated by finding the offset between each NREM SWR and the nearest spindle start time. Probability distributions were generated for each genotype using the absolute value of each ripple-spindle time lag.

NREM and quiet wake definition

Putative sleep was defined^68,92–94 as times of prolonged immobility (>60 s) with NREM sleep classified by a ratio of theta-to-delta power below a threshold set by a blinded observer based on visual inspection of the CA1 spectrograms (0–20Hz, 2 s window with 0.2 s advancement, see example sleep spectrograms in Figure S4). To calculate the theta-to-delta ratio, a moving average of the envelope power was taken for theta (5s window) and delta (12.5s window). After calculating the ratio, the data was z-scored. For analyses of both pre and post experience sleep, only mice with detected sleep for both sessions were used for analysis.

Single unit analysis

Individual units from hippocampal electrodes were manually clustered (xclust2, Matt Wilson) based on spike waveform peak amplitudes. Only well-isolated units were included in the analysis. Modified L_ratio values⁹⁵ were calculated as previously described.²⁶ L_ratio was not different between WT and KIBRA cKO units (WT: 0.0057 ± 0.0017, cKO: 0.0077 ± 0.0023, WT vs. cKO Mann-Whitney p = 0.3331). Position was binned (1 cm) and position tuning curves (place fields) were calculated as a the smoothed (Gaussian kernel, st. dev. of 3 cm) histogram of firing activity normalized by occupancy using only periods when velocity was greater than 5 cm/s. Units were considered to have a place field if the peak of the smoothed tuning curve was greater than 1 Hz. Units with mean firing rate greater than 10 Hz were considered putative inhibitory neurons and were excluded from all analyses. Place fields were defined as contiguous bins with firing at least 30% of the maximal firing rate, with a minimum of 20 bins. Spatial information was calculated as previously described.⁹⁶ Cell participation was defined as the fraction of SWRs in which a given cell participated.

Sample size calculations

A priori sample size ranges were based on 1) power analysis using effect sizes estimated from our related previous experimental results and relevant published studies (G*Power, paired and unpaired two-tailed t-tests and ANOVAs, target power 0.8, alpha 0.05) and, 2) considerable prior experience and publications in the field of in vivo electrophysiology.

Exclusion criteria

Our initial study design included both male and female mice. However, while there is no difference in body weight between genotypes, the body weight of adult female mice mice was 83.5% that of male littermates in this mouse line (KO: 11F/12M, WT: 9F/14M, two-way ANOVA, significant effect of gender p > 0.0001, no significant effect of genotype p = 0.291 or gender × genotype interaction p = 0.210). Implanted female mice had a higher mortailty rate and had more difficulty carrying the drive weight (~4.2g). Due to concerns about how these experimental differences would influence data quality, we proceeded to collect and analyze data only in male mice with this iteration of our drive design.

For analyses during pre and post experience sleep, only mice with detected sleep for both sessions were used for analysis.

Randomization and blinding

The random allocation of mice to experimental groups was driven by Mendelian Inheritance. Initial experiments were conducted blinded as data acquisition and tamoxifen/vehicle injections were conducted by different experimenters. A subset of recordings were acquired by an experimenter not blinded to genotype for the entire experiment due to the necessity of one experimenter doing both Tamoxifen injections and acquiring data, although alternate ID numbers decoupled from genotype information were assigned prior to drive implantation.

Theta/delta ratio threshold used for segregating REM from NREM putative sleep was conducted by an experimenter blind to genotype/treatment. All other analysis parameters were optimized using data from the WT control group, then the standardized analysis parameters were run on all subjects in an automated manner.

QUANTIFICATION AND STATISTICAL ANALYSIS

Unless otherwise noted, data from each mouse were averaged such that every mouse contributed one data point for each comparison. All data and test residuals were assessed for normality using the Shapiro-wilk and D’Agostino-Pearson tests, or Lilliefors test, in addition to Q-Q plots. Equality of variance between groups was assessed using either an F test or Spearman’s test, and examination of residual homoscedasticity plots. Statistical tests were systematically chosen and uniformly applied based on hypotheses, and evaluation of statistical test assumptions. For circular data (polar histograms) uniformity of distributions was assessed with a Rayleigh Test for uniformity, and Kuiper’s Two-Sample Test was used to compare distributions between genotypes. For non-circular data, significance was evaluated using one or two-way ANOVAs (repeated measures applied as appropriate, e.g. comparison of pre vs. post experience in the same group) with Šidák correction for post hoc multiple comparisons, or t-tests. Welch’s correction was applied to data evaluated with t-tests that did not meet criteria for equal variance (Figures 1L, 4B, 6A, and 6E). For data which did not meet criteria for normally distributed test residuals, nonparametric statistics were used; Mann-Whitney rank sum in lieu of unpaired t test, Multiple Mann-Whitney rank sum with Holm-Šidák correction for multiple comparisons in lieu of two-way ANOVA with post hoc multiple comparisons, Multiple Wilcoxon matched-pairs with Holm-Šidák correction for multiple comparisons in lieu of repeated measures two-way ANOVA w/post hoc multiple comparisons, Freidman test with Dunn’s correction for multiple comparisons in lieu of one-way ANOVA, one-sample Wilcoxon in lieu of one-sample t test. All data are shown as mean ± SEM unless otherwise indicated. Statistical analyses were carried out with GraphPad Prism version 9.4.1 or custom analysis code in MATLAB. For Power Spectral Densities, a two-group test for equal population spectra was conducted via Chronux.⁸⁸ The Circular Statistics toolbox⁸⁹ was used to analyze phase angle data.

Highlights.

• KIBRA is necessary for experience modulation of neural activity during SWRs

• SWR-associated low gamma increases after experience and is disrupted by KIBRA loss

• Experience-induced coupling of hippocampal SWRs and cortical spindles requires KIBRA

• KIBRA is not required to maintain basal hippocampal or cortical oscillatory dynamics