Specific exercise patterns generate an epigenetic molecular memory window that drives long-term memory formation and identifies ACVR1C as a bidirectional regulator of memory in mice

Abstract

Exercise has beneficial effects on cognition throughout the lifespan. Here, we demonstrate that specific exercise patterns transform insufficient, subthreshold training into long-term memory in mice. Our findings reveal a potential molecular memory window such that subthreshold training within this window enables long-term memory formation. We performed RNA-seq on dorsal hippocampus and identify genes whose expression correlate with conditions in which exercise enables long-term memory formation. Among these genes we found Acvr1c, a member of the TGF ß family. We find that exercise, in any amount, alleviates epigenetic repression at the Acvr1c promoter during consolidation. Additionally, we find that ACVR1C can bidirectionally regulate synaptic plasticity and long-term memory in mice. Furthermore, Acvr1c expression is impaired in the aging human and mouse brain, as well as in the 5xFAD mouse model, and over-expression of Acvr1c enables learning and facilitates plasticity in mice. These data suggest that promoting ACVR1C may protect against cognitive impairment.

Fig. 1. Exercise enhances hippocampus-dependent memory and synaptic plasticity and engages a molecular memory window leading to enhanced cognitive improvement and plasticity from subsequent 2-day reactivating exercise.

A Simplified schematic; see. Created with Biorender.com. B Discrimination Index (DI) scores during object location memory (OLM) test. 14 days of exercise transforms a normally subthreshold learning event into long-term memory. After a 7-day sedentary delay, only 2 days of reactivating exercise enables long-term memory formation. One-way ANOVA, Group: (F_(6,64) = 8.13, P < 0.0001). Tukey’s post hoc test: *P < 0.05, ***P < 0.001 compared to sedentary (0-0-0), ++P < 0.01 compared to 14-0-0, (0-0-0: n = 10, 0-0-2: n = 15, 14-0-0: n = 6, 14-7-0: n = 10, 14-7-2: n = 11, 14-14-0: n = 9, 14-14-2: n = 10). Data taken from our previous study Butler et al., 2019. Data are presented as mean ± SEM. C Exercise parameters and framework. D Left panel, time course of the mean ± SEM field excitatory postsynaptic potential (fEPSP) slope as a percentage of baseline recorded in slices. Right panel, independent two-sample t test (two-tailed): (t(10) = 8.261, P < 0.0001), (14-0-0: n = 6 mice, n = 12 slices, control 0-0-0: n = 6 mice, n = 12 slices). Inset; representative traces collected during baseline (black line) and 60 min post theta burst stimulation (TBS, arrow) (red line). Scale = 1 mV/5 ms. E Left panel, LTP time course. Right panel, quantification of mean ± SEM potentiation 50–60 min post-TBS, One-way ANOVA, Group: (F_(2,9) = 10.96, P = 0.0039). Tukey’s post hoc test, 50–60 min post TBS: **P < 0.01 for both 14-7-0 and 14-7-2 compared to 0-0-0. (14-7-0: n = 4 mice, n = 8 slices, 14-7-2: n = 4 mice, n = 8 slices, 0-0-0: n = 4 mice, n = 8 slices). Inset; representative traces collected during baseline and 60 min post-TBS. Scale = 1 mV/5 ms. F Left panel, LTP time course. Middle panel, representative traces. Scale = 1 mV/5 ms. Right panel, One-way ANOVA, Group: (F_(3,26) = 8.942, P = 0.0003). Tukey’s post hoc test, 50–60 min post TBS: _***P < 0.001 for 14-14-2 compared to 0-0-0, *P < 0.05 for 14-14-2 compared to 14-14-0, P = 0.137 for 0-0-2 compared to 0-0-0. (14-14-0: n = 5 mice, n = 10 slices, 14-14-2: n = 5 mice, n = 10 slices, 0-0-2: n = 5 mice, n = 10 slices, 0-0-0: n = 15 mice, n = 30 slices). Source data are provided as a Source Data file.

Fig. 2. Acvr1c expression in hippocampus is induced during object location memory (OLM) consolidation only under exercise conditions that drive memory encoding.

A Detailed schematic displaying the timing of procedures for individual groups. B Simplified Schematic. Created with Biorender.com. C Heat map of genes differentially expressed in each condition compared to sedentary control (0-0-0). Positive log fold change indicates up-regulated genes vs sedentary whereas negative log fold change refers to down-regulated genes. D Number of up-regulated differentially expressed genes (DEGs) in dorsal hippocampus compared to sedentary. E RT-qPCR data demonstrating that Acvr1c is up-regulated in dorsal hippocampus only in conditions where exercise facilitates learning (14-0-0, n = 9 and 14-7-2, n = 7). One-way ANOVA, Group (F_(6,54) = 3.62, P = 0.004). Tukey’s post hoc test: *P < 0.05, +P = 0.05 compared to sedentary (0-0-0, n = 9). Normalization to sedentary (HC and 14-7-0, n = 9, 0-0-2 n = 10 and 14-14-0 n = 8). Data are presented as mean ± SEM. F–H Overlapping volcano plots illustrating the significance (Y-axis) and magnitude (X-axis) of experience-induced changes in each group. Volcano plots show fold change in gene expression between two conditions, using regularized t test and p value corrected for multiple testing. Acvr1c and Bdnf were up-regulated during memory consolidation in both conditions where exercise facilitates learning (14-0-0) (F) and (14-7-2) (H). I Top predicted upstream regulators of 14d initial exercise (14-0-0) vs. sedentary DEGs, 7d sedentary delay (14-7-0) vs. sedentary ( J), and 2d reactivating exercise (14-7-2) vs. sedentary (K). L Venn diagram highlighting common upstream regulators in exercise conditions that facilitate learning (14-0-0 and 14-7-2). Source data are provided as a Source Data file.

Fig. 3. Exercise modulates epigenetic regulation of Acvr1c and Bdnf during consolidation and reveals a specific permissive signature.

A Schematic. Created with Biorender.com. B, C Epigenetic regulation of Acvr1c during consolidation is modified by exercise. B Exercise does not impact H3K9Me3/H3K9Ac occupancy at the Acvr1c promoter following sedentary delay periods: One-way ANOVA, Group (F_(5,47) = 4.41, P = 0.002), (0-0-0: n = 8, 0-0-2: n = 8, 14-0-0: n = 10, 14-7-0: n = 10, 14-7-2: n = 9, 14-14-0: n = 8). C Engagement in either minimal (2-day), (**P < 0.01) or extensive (14-day), (****P < 0.0001) exercise reduces repressive H3K27Me3 at the Acvr1c promoter. Notably, this reduction occurs throughout the sedentary delay periods assessed (7-day), (****P < 0.0001) and (14-day), (****P < 0.0001) and following reactivating exercise (***P < 0.001): (One-way ANOVA), Group (F_(5,49) = 9.377, P < 0.0001), (0-0-0: n = 10, 0-0-2: n = 10, 14-0-0: n = 10, 14-7-0: n = 10, 14-7-2: n = 7, 14-14-0: n = 8). D, E Epigenetic regulation of Bdnf IV during consolidation is modified by exercise. D Exercise does not impact H3K9Me3/H3K9Ac occupancy at the Bdnf IV promoter: One-way ANOVA, Group (F_(5,51) = 1.44, P = 0.22), (0-0-0: n = 10, 0-0-2: n = 10, 14-0-0: n = 10, 14-7-0: n = 10, 14-7-2: n = 9, 14-14-0: n = 8). E Remarkably, engagement in either minimal (2-day), (****P < 0.0001) or extensive (14-day) (****P < 0.0001) exercise also reduces repressive H3K27me3 at the Bdnf IV promoter. Notably, this reduction also continues at the sedentary delay periods assessed (7-day), (****P < 0.0001) and (14-day), (****P < 0.0001) sedentary periods and following reactivating exercise (****P < 0.0001): (One-way ANOVA), Group (F_(5,53) = 13.90, P < 0.0001), (0-0-0: n = 10, 0-0-2: n = 10, 14-0-0: n = 10, 14-7-0: n = 10, 14-7-2: n = 9, 14-14-0: n = 10). Results provide evidence for the ability of exercise to effectively remove repressive marks at two distinct promoters of genes critical for memory in a specific manner during consolidation. Tukey’s post hoc test: *P < 0.05, **P < 0.01, ***P < 0.001, ****P < 0.0001, + P = 0.05. Data are presented as mean ± SEM. Source data are provided as a Source Data file.

Fig. 4. ACVR1C is required for hippocampus-dependent long-term memory and synaptic plasticity.

A Experimental design. Created with Biorender.com. B Discrimination Index (DI) scores during object location memory (OLM) training reveal no differences between groups independent two-sample t-test (two-tailed): (t(17) = 0.006, P = 0.995), (empty vector (EV): n = 9, ACVR1C-kinase dead (ACVR1C-KD): n = 10). C Total amount of time in seconds exploring objects during training. Mice from both groups display similar levels of total object exploration independent two-sample t-test (two-tailed): (t(17) = 0.545, P = 0.592). D Disrupting ACVR1C function (ACVR1C-KD) leads to impaired OLM independent two-sample t test (two-tailed): (t(17) = 4.65, P = 0.0002), (EV: n = 9, ACVR1C-KD: n = 10). E Total amount of time in seconds exploring objects during test. Total object exploration does not differ between groups during test independent two-sample t test (two-tailed): (t(17) = 1.53, P = 0.14), (EV: n = 9, ACVR1C-KD: n = 10). F Percent freezing during 3-minute test session. Disrupting ACVR1C function (ACVR1C-KD) led to reduced freezing compared with EV control independent two-sample t-test (two-tailed): (t(18) = 4.790, P = 0.0001), (EV: n = 10, ACVR1C-KD: n = 10). G LTP as mean ± SEM excitatory postsynaptic potential (fEPSP) slope as percentage of baseline overtime (ACVR1C-KD; n = 7 mice, n = 14 slices, EV; n = 5 mice, n = 10 slices). Theta burst stimulation (TBS) applied at arrow. Inset; representative traces collected during baseline (black line) and 60 min post TBS (red line) from ACVR1C-KD and EV group. Scale = 1 mV/5 ms. H Mean ± SEM potentiation 50–60 min post TBS showing impaired LTP in slices from KD-infused mice vs EV-control. Disrupting ACVR1C function leads to impaired LTP independent two-sample t test (two-tailed): (t(10) = 3.795, P = 0.0035) relative to EV control (EV: n = 5, ACVR1C-KD: n = 5). ####P < 0.0001 compared with training day (within group). *P < 0.05, ***P < 0.001. Data are presented as mean ± SEM. Source data are provided as a Source Data file.

Fig. 5. Acvr1c overexpression transforms inadequate, subthreshold learning into long-term memory and facilitates LTP.

A Experimental design. Created with Biorender.com. B Discrimination Index (DI) scores during object location memory (OLM) training reveal no difference between groups independent two-sample t test (two-tailed): (t(18) = 0.564, P = 0.579), (empty vector (EV): n = 11, ACVR1C-wild type (ACVR1C-WT): n = 9). C Total amount of time in seconds exploring objects during training. Mice from both groups display similar levels of total object exploration independent two-sample t-test (two-tailed): (t(18) = 0.550, P = 0.589), (EV: n = 11, ACVR1C-WT: n = 9). D Acvr1c overexpression enhances OLM independent two-sample t test (two-tailed): (t(18) = 3.303, P = 0.004), (EV: n = 11, ACVR1C-WT: n = 9). E Exploration does not differ between groups on test independent two-sample t-test (two-tailed): (t(18) = 0.636, P = 0.532), (EV: n = 11, ACVR1C-WT: n = 9). F Percent freezing during 3-minute test session. Acvr1c overexpression has no effect on freezing compared with EV control independent two-sample t-test (two-tailed): (t(19) = 1.696, P = 0.106), (EV: n = 11, ACVR1C-WT: n = 10). G Measurement of LTP as mean ± SEM excitatory postsynaptic potential (fEPSP) slope as percentage of baseline overtime (ACVR1C-WT; n = 5 mice, n = 10 slices, EV; n = 3 mice, n = 6 slices). Theta burst stimulation (TBS) applied at arrow. Inset; representative traces collected during baseline (black line) and 60 min post-TBS (red line) from ACVR1C-WT and EV group. Scale = 1 mV/5 ms. H Mean ± SEM level of potentiation 50–60 min post TBS showing enhanced level of LTP in slices from mice overexpressing Acvr1c relative to slices from EV-infused mice independent two-sample t test (two-tailed): (t(6) = 3.51, P = 0.0127), (EV: n = 3, ACVR1C-WT: n = 5). ###P < 0.001 compared with training day (within group). *P < 0.05, **P < 0.01. Data are presented as mean ± SEM. Source data are provided as a Source Data file.

Fig. 6. Age and AD-related impairments in hippocampus-dependent memory and synaptic plasticity are associated with reduced Acvr1c expression which are ameliorated by overexpressing Acvr1c.

A Left: Acvr1c mRNA is reduced in aging mouse dorsal hippocampus independent two-sample t-test (two-tailed): (t(26) = 2.72, P = 0.01). Sex differences were not observed (Two-way ANOVA Sex): (F_(1,24) = 0.43, P = 0.50, interaction P > 0.05), indicating that Acvr1c is reduced with age in both sexes, (young: n = 12, aging: n = 16). Right: Acvr1c levels (transcripts per million (TPM)) are reduced in the aging human hippocampus independent two-sample t-test (two-tailed): (t(91) = 6.64, P = 0.0001). Ages displayed map to mouse ages for young (3 mo.) and aging (20 mo.), (n = 5 20–29-year-olds, n = 88 60–69-year-olds), accessed from the GTEx project expression database. B Discrimination Index (DI) scores during object location memory (OLM) training reveal no differences between groups independent two-sample t test (two-tailed): (t(12) = 1.988, P = 0.070), (n = 7/group). C Mice from both groups have similar levels of total object exploration independent two-sample t test (two-tailed): (t(12) = 0.259, P = 0.799), (n = 7/group). D Acvr1c overexpression (WT) ameliorates age-related impairments in OLM independent two-sample t test (two-tailed): (t(12) = 2.350, P = 0.036), (n = 7/group). E Exploration did not differ between groups on test independent two-sample t test (two-tailed): (t(12) = 0.817, P = 0.429), (n = 7/group). F Measurement of LTP as mean ± SEM fEPSP slope overtime (empty vector (EV); n = 5 mice, n = 10 slices; ACVR1C-WT; n = 3 mice, n = 6 slices). Theta burst stimulation (TBS) applied at arrow. Inset; representative traces collected during baseline (black line) and 60 min post-TBS (red line) from ACVR1C-WT and EV group. Scale = 1 mV/5 ms. G Mean ± SEM level of potentiation 50–60 min post TBS showing that Acvr1c overexpression enhances LTP in slices from 18 mo. C57BL6/J mice relative to EV control independent two-sample t test (two-tailed): (t(6) = 3.540, P = 0.0122), (EV; n = 5 mice, n = 10 slices; ACVR1C-WT; n = 3 mice, n = 6 slices). H RNA-Seq data displaying transcripts per million (TPM) shows decreases in Acvr1c with age (Three-way ANOVA), Age: (F_(2,48) = 54.95, P < 0.0001), that become exacerbated in 5xFAD mice (Genotype: F_(1,48) = 6.39, P = 0.01). Sex differences were not observed, Sex: (F_(1,48) = 0.40), P = 0.53, Age x Sex: P = 0.233, Age x Genotype: P = 0.387, Genotype x Sex: P = 0.037, Age x Genotype x Sex: P = 0.244. Tukey’s post hoc test (compared between C57BL/6 J groups): **P < 0.01, ***P < 0.001, ****P < 0.0001, (compared between 5xFAD groups): ++P < 0.01, ++++P < 0.0001, (C57: 4 mo.: n = 10, 8 mo.: n = 11, 12 mo.: n = 10; 5xFAD: 4 mo.: n = 10, 8 mo.: n = 9, 12 mo.: n = 10). I Discrimination Index (DI) scores during training reveal no differences between groups independent two-sample t test (two-tailed): (t(21) = 0.556, P = 0.583), (EV: n = 10, ACVR1C-WT: n = 13). J Mice from both groups have similar levels of total object exploration on training independent two-sample t test (two-tailed): (t(21) = 1.459, P = 0.159), (EV: n = 10, ACVR1C-WT: n = 13). K Acvr1c overexpression (WT) ameliorates AD-related impairments in OLM independent two-sample t-test (two-tailed): (t(21) = 2.287, P = 0.032), (EV: n = 10, ACVR1C-WT: n = 13). L Exploration did not differ between groups on test independent two-sample t-test (two-tailed): (t(21) = 0.149, P = 0.883), (EV: n = 10, ACVR1C-WT: n = 13). M Measurement of LTP as mean ± SEM fEPSP slope overtime (EV; n = 4 mice, n = 8 slices; ACVR1C-WT; n = 5 mice n = 10 slices). TBS applied at arrow. Inset; representative traces collected during baseline (black line) and 60 min post-TBS (red line) from ACVR1C-WT and EV group. Scale = 1 mV/5 ms. N Mean ± SEM level of potentiation 50–60 min post TBS showing that Acvr1c overexpression enhances LTP in slices from 12 mo. 5xFAD mice relative to EV control independent two-sample t test (two-tailed): (t(7) = 3.852, P = 0.0063), (EV; n = 4 mice, n = 8 slices; ACVR1C-WT; n = 5 mice n = 10 slices). ###P < 0.001 compared with training day (within group). *P < 0.05, **P < 0.01, ****P < 0.0001. Data are presented as mean ± SEM.