Abstract
Subconcussive head impacts are associated with the development of acute and chronic cognitive deficits. We recently reported that high-frequency head impact (HFHI) causes chronic cognitive deficits in mice through synaptic changes. To better understand the mechanisms underlying HFHI-induced memory decline, we used TRAP2/Ai32 transgenic mice to enable visualization and manipulation of memory engrams. We labeled the fear memory engram in male and female mice exposed to an aversive experience and subjected them to sham or HFHI. Upon subsequent exposure to natural memory recall cues, sham, but not HFHI, mice successfully retrieved fearful memories. In sham mice the hippocampal engram neurons exhibited synaptic plasticity, evident in amplified AMPA:NMDA ratio, enhanced AMPA-weighted tau, and increased dendritic spine volume compared with nonengram neurons. In contrast, although HFHI mice retained a comparable number of hippocampal engram neurons, these neurons did not undergo synaptic plasticity. This lack of plasticity coincided with impaired activation of the engram network, leading to retrograde amnesia in HFHI mice. We validated that the memory deficits induced by HFHI stem from synaptic plasticity impairments by artificially activating the engram using optogenetics and found that stimulated memory recall was identical in both sham and HFHI mice. Our work shows that chronic cognitive impairment after HFHI is a result of deficiencies in synaptic plasticity instead of a loss in neuronal infrastructure, and we can reinstate a forgotten memory in the amnestic brain by stimulating the memory engram. Targeting synaptic plasticity may have therapeutic potential for treating memory impairments caused by repeated head impacts.
Introduction
Repeated exposure to head impacts during an athletic career has been linked to memory impairments and increased risk of neurodegenerative diseases such as chronic traumatic encephalopathy. Even minor subconcussive impacts, which occur through activities such as heading a soccer ball, or the sustained low-amplitude cranial movement in professional sled athletes, have been shown to cause chronic cognitive symptoms (Colvin et al., 2009; Talavage et al., 2014; McCradden and Cusimano, 2018). However, the mechanisms by which these subconcussive impacts lead to lasting cognitive impairments and whether lost cognitive function can be regained are not well understood. While the presence of amyloid and tau pathology in the brain has been associated with dementia, a causal relationship between concussion, neurodegenerative pathology, and cognitive dysfunction has not been established (Stewart et al., 2019).
The presence of cognitive impairments can be replicated in animal models of repeat mild traumatic brain injury (rmTBI), even in the absence of significant pathological changes (Meehan et al., 2012; Mouzon et al., 2012; Petraglia et al., 2014; Xu et al., 2016). This suggests that nonpathological mechanisms may be involved in the development of cognitive impairment. We recently reported that mice exposed to a high frequency of head impacts (HFHI) show physiological changes in the brain leading to reduced hippocampal brain function and anterograde spatial and working memory deficits (Sloley et al., 2021). These HFHI-induced memory deficits are caused by chronic synaptic adaptations in hippocampal neurons occurring in response to glutamate released following head impacts, and pretreatment with an NMDA receptor antagonist prevents these synaptic adaptations and the resulting memory deficits from occurring (Sloley et al., 2021). This prophylactic treatment has the potential to prevent head impact-induced cognitive impairment; however, treatments that can reverse cognitive impairments in individuals already living with TBI have not been developed.
The memory engram refers to the collection of changes that happen in the brain that hold the substrate for a specific memory (Josselyn et al., 2015; Poo et al., 2016; Josselyn and Tonegawa, 2020). The physical manifestation of these changes for a specific memory involves changes in the structure and function of neurons and synapses (Josselyn et al., 2015; Poo et al., 2016; Josselyn and Tonegawa, 2020). Previous research using engram mice has shown that engram neurons display synaptic changes including typical of those associated with long-term potentiation including the stronger α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) evoked excitatory postsynaptic currents (EPSC), increased spontaneous EPSC and amplitude, and increased dendritic spine density (Ryan et al., 2015; Roy et al., 2016; Kim and Cho, 2017; Sun et al., 2020; Naik et al., 2021; Lee et al., 2023). The engram neurons are reactivated by retrieval cues (Ryan et al., 2015) and direct optogenetic activation of engram neurons for contextual memories associated with fear/threat is sufficient (Liu et al., 2012; Ramirez et al., 2013), as well as necessary (Denny et al., 2014; Tanaka et al., 2014; Trouche et al., 2016), to recall a specific memory even in amnesia states (Ryan et al., 2015; Roy et al., 2016; Power et al., 2022).
Here, we use TRAP2/Ai32 mice to examine if HFHI-induced retrograde amnesia is associated with changes in engram neuron physiology, and if head impact-induced amnesia can be reversed. We focused on the dentate gyrus as this hippocampal area is associated with memory encoding and retrieval (Bernier et al., 2017), and the electrophysiological profile of c-Fos-expressing dentate gyrus engram cells is established (Pignatelli et al., 2019). We report that HFHI does not alter the number of dentate gyrus neurons in the memory engram, but it does interfere with the ability of memory engram neurons to physiologically distinguish themselves from nonengram dentate gyrus neurons. We report the loss of engram-specific electrophysiological signatures and decreased spontaneous reactivation of the memory engram in HFHI mice. We also find that it is possible to reactivate the amnestic memory in HFHI mice and elicit a temporary reinstatement of the forgotten memory in the head impact brain.
Materials and Methods
Animals and HFHI model
All procedures were performed in accordance with protocols approved by the Georgetown University Animal Care and Use Committee. Homozygous TRAP2 mice (#030323, Jackson Laboratories) were crossed with homozygous Ai32 mice (#024109, Jackson Laboratories) to produce hemizygous offspring. Closed head HFHI procedures were performed on hemizygous mice as previously described (Main et al., 2017; Sloley et al., 2021). Briefly, male and female, 2–3-month-old, mice were anesthetized for 3 min in 3% isoflurane in 1.5 L/min oxygen. Mice were placed in the injury device with their unrestrained head resting on a gel pad and isoflurane delivered via a nosecone for an additional 1 min. The 10-mm-diameter Teflon tip was positioned to impact directly on the midline dorsal surface of the head with the front of the impact tip positioned immediately rostral to the eye socket and equidistance from the mouse ears. The area impacted is equivalent to the rostrocaudal length of the parietal bone, with inclusion of rostral areas of the frontal bone. The pneumatically controlled impact was delivered at an impact speed of 2.35 m/s, dwell time of 32 ms, and an impact depth of 7.5 mm. Five impacts were delivered over a 10 s interval. One cohort of mice received a single day of impacts [five head impact mice (5HI)]. HFHI mice received five impacts delivered per day for 6 d (totaling 30 hits). Shams received identical handling and anesthesia protocols, but no head impacts. Mice were exposed to sham, 5HI, or HFHI protocol 15 min following memory acquisition and daily thereafter for 6 d total.
Fear conditioning and memory recall
Training
To reduce background interference with engram labeling caused by transportation, animal handling, and novel room exposure, TRAP2/Ai32 transgenic mice were handled by experimenters performing the behavior tests for at least 3 d prior to acquisition in the room where conditioning occurs. Mice were fear conditioned for 6 min in a fear conditioning chamber (Context A). The training protocol consisted of 3 min of baseline followed by a 2 s 0.75 mA footshock every minute for 3 min (total of three shocks).
Engram labeling
The engram neurons were labeled by two intraperitoneal injections of 50 mg/kg 4-OH-tamoxifen in chen oil administered 15 min and 2 h after CFC.
Natural recall
Mice were placed back in Context A to test natural memory recall (day of recall specified in each experiment). The recall protocol consisted of 3 min in the conditioned context, but with no shocks. Percent freezing, latency to freeze, and freezing episodes were automatically calculated using AnyMaze software.
Optogenetic recall
Ten days after the training protocol, dual ferrule cannulae were bilaterally implanted into the dentate gyrus of the hippocampus. Burr holes were made using a 0.5 mm drill bit and cannulae were stereotactically implanted into the dorsal dentate gyrus (dDG; −2.0 mm anteroposterior AP, ±1.3 mm mediolateral ML, −1.9 mm dorsoventral DV) and mice were allowed to recover for 2 weeks. After the recovery period, and prior to optogenetic experiments, mice were habituated to handling and optogenetic patch cords for 2 d. Habituation consisted of a 5 min session in a neutral context with the patch cords attached to the cannula. Over the course of the experiment, we experienced an attenuation in the number of experimental animals due to a loss of optogenetic cannulae caused by handling. For the experiment, mice were attached to the patch cords and placed in a novel context (Context B). Context B consisted of a novel arena that was visually distinct from Context A and was housed in a different outer case to Context A and contained a different scent. Mice experienced a 6 min protocol consisting of one 3 min epoch of habituation (no optogenetic stimulation) followed by another 3 min epoch with optogenetic stimulation. Twenty-four hours after the optogenetic experiment, mice were tested for natural fear recall by replacing them in Context A for 3 min. No-shock control groups were included as a reference for baseline freezing and to determine the effect of optogenetic stimulation on engram neurons not linked to a conditioned stimulus. These mice did not receive footshocks during the conditioning session, and not all of these no-shock mice were re-exposed to Context A at the end of the study.
Freeze behavior was manually quantified by blinded experimenters due to optogenetic light stimulation and patch cords interfering with automatic detection. Optogenetic stimulation consisted of a nested stimulation in which five light pulses occurred at 20 Hz every 0.125 s via a Doric 450 nm LED light source driven by TTL input with a delayed onset of 25 ms.
Immunohistochemistry
Forty-five minutes following natural recall, animals were transcardially perfused with PBS and 4% PFA. Whole brains were extracted and drop fixed in 4% PFA overnight and then saturated in 20% sucrose. Fixed brains were sectioned into 60 µm coronal slices on a microtome following sucrose saturation and were subsequently stained for EYFP, endogenous c-Fos, and DAPI as follows. Slices were washed in phosphate-buffered saline (PBS) and blocked in 10% NGS with 0.2% PBS-triton prior to antibody staining. Staining for engrams was performed using chicken anti-YFP (#A10262, Thermo Fisher Scientific) and rabbit anti-c-Fos (#2250, Cell Signaling Technology) at a 1:1,000 concentration overnight. Following another wash, goat anti-chicken 488 (#A-11039, Thermo Fisher Scientific) and goat anti-rabbit 568 (# A-11011, Thermo Fisher Scientific) at 1:1,000 was added overnight to stain for EYFP and endogenous c-Fos, respectively. DAPI counterstaining was performed prior to mounting and imaging on a LSM 880 Zeiss laser scanning confocal microscope (Zeiss). Z-stacked images of the dentate gyrus were taken at 20× in 3 µm Z-slices. Images were then processed and projected by a blinded experimenter in FIJI (NIH, Bethesda, MD). All processing steps equally affected all pixels in the frame. The same blinded experimenter manually quantified EYFP + engram neurons, c-Fos + cells, and % overlap between EYFP and c-Fos. All values were acquired from at least three slices per animal, normalized to the dorsal hippocampus of the standard mouse brain, and presented as the number of cells per dDG.
Slice preparation
Acute sagittal hippocampal slices were prepared from experimental animals (preference for slices containing the dorsal hippocampi) on day 10 following fear memory acquisition. Brain slices were prepared in NMDG and HEPES-buffered artificial cerebrospinal fluid (aCSF), as previously described (Ting et al., 2014). Briefly, mice were anesthetized in open isoflurane prior to transcardial perfusion, brain dissection, and brain slicing in 4°C NMDG solution (in mM, 92 NMDG, 2.5 KCl, 1.25 NaH2PO4•2H2O, 30 NaHCO3, 20 HEPES, 25 glucose, 10 sucrose, 5 ascorbic acid, 2 thiourea, 3 sodium pyruvate, 5 N-acetyl-L-cysteine, 10 MgSO4•7H2O, 0.5 CaCl2•2H2O, pH ∼7.4, osmolarity ∼300–310 mOsm). We prepared 275-μm-thick sagittal slices using a Vibratome Series 3000. Slices were immediately placed in 32°C NMDG solution for 25 min during which time the sodium chloride concentration was brought up to 90 mM with sodium spike-ins every 5 min. Slices were then transferred to an incubation chamber containing room temperature carboxygenated HEPES solution (in mM, 92 NaCl, 2.5 KCl, 1.25 NaH2PO4•2H2O, 30 NaHCO3, 20 HEPES, 25 glucose, 5 ascorbic acid, 2 thiourea, 3 sodium pyruvate, 5 N-acetyl-L-cysteine, 2 MgSO4•7H2O, 2 CaCl2•2H2O, pH ∼7.4, osmolarity ∼300–310 mOsm) and were allowed to recover until the recording.
Electrophysiology
At the time of recording, slices were transferred to a Siskiyou PC-H perfusion chamber, anchored to the bottom of the recording chamber and submerged in circulating carboxygenated aCSF (in mM, 124 NaCl, 3.5 KCl, 1.2 NaH2PO4•2H2O, 26 NaHCO3, 10 glucose, 1 MgCl2•6H2O, 2 CaCl2•2H2O, pH ∼7.4, osmolarity ∼300–310 mOsm) at 5 ml/min. Recordings were performed with a MultiClamp 700B amplifier (Molecular Devices), digitized to 20 kHz, and low-pass filtered at 2 kHz with a computer running Clampex 11 and DigiData 1440 (Molecular Devices). One recording channel for the whole-cell electrophysiology was recorded with 4–5 MΩ borosilicate pipettes pulled the day of recordings and filled with cesium methanesulfonate internal (in mM, 120 Cs-MeSO3, 5 NaCl, 10 TEA•Cl, 10 HEPES, 1.1 EGTA, 4 lidocaine N-ethyl bromide, 4 ATP•Na, and 0.3 GTP•Na) with 0.5% biocytin added on the day of the experiment. A monopolar stimulating electrode was placed in the inner 1/3 of the dentate gyrus molecular layer for evoked potentials. At least one engram and nonengram were recorded from different slices for each animal. Following a successful whole-cell patch clamp, the following recordings were performed: voltage-clamp photocurrent (confirm engram or nonengram), 20 stimulations at −70 mV holding potential (AMPA current), 20 stimulations at 40 mV holding potential (NMDA current), and, in current clamp, 40 current injections at currents ranging from −120to 40 pA for biocytin filling. Following completion of the patch-clamp session, an outside out patch was pulled, and the slice was allowed to sit for at least 1 h prior to fixation in 4% paraformaldehyde and 4% sucrose overnight for post hoc immunohistochemistry. Access resistance was constantly monitored throughout the recording and only cells with access resistance <20 MΩ were included in the analysis. Data were analyzed in Clampfit 11.3. IPSCs were detected using a template search for positive going currents, and the output was analyzed using a custom MATLAB script (https://github.com/dpchapma/PSC-Analysis). We quantified the weighted decay tau of the AMPA current (AMPA Tw) using a double exponential as previously described (Rumbaugh and Vicini, 1999). Briefly, a double exponential curve was fit to the AMPA current, and a weighted average of the decay times was taken.
Spine analysis/imaging
Following cell filling and overnight fixation, slices were washed in PBS, permeabilized in 0.5% triton for 2 h, and blocked in 10% NGS with 0.2% PBS-triton prior to staining. Staining for engrams was performed using chicken anti-YFP (Thermo Fisher Scientific) at a 1:1,000 concentration overnight. Following another wash, goat anti-chicken 488 (Thermo Fisher Scientific) at 1:1,000 and streptavidin CY3 (VectorLabs) at 1:500 were added overnight to stain for EYFP and biocytin respectively. DAPI counterstaining was performed prior to mounting and imaging on a Zeiss laser scanning confocal microscope (LSM 880). Z-stacked images were taken at 40× with optimized step levels for magnification, pinhole size, and zoom level (0.4–0.6 µm). Imaris software was then used to semiautomatically reconstruct whole neuron morphology including dendrites and spines. Output from Imaris was analyzed using custom MATLAB code (https://github.com/dpchapma/SpineAnalysis). Only spines between 100 and 200 µm from the soma, <0.8 µm3 volume, and <5 µm in length were included in the analysis and cells with <15 spines that met these criteria were excluded (n = 1 HFHI F engram).
Statistics
All statistics were performed in Prism 9 (GraphPad). For behavior experiments, one-way ANOVA with Tukey's post hoc test or unpaired t tests were used to test the difference between sham and HFHI. For the optogenetic experiment, a repeated-measure three-way ANOVA with main effects of light off–on was performed, followed by Šídák's multiple-comparisons test. For morphology experiments, a two-way ANOVA was used to compare morphology features across cell types and groups. For spine volume, cumulative distributions were created for each cell and averaged across group. A repeated measures two-way ANOVA was performed on the resulting averaged cumulative distributions. For electrophysiology experiments, we compared the engram cells against the nonengram cells from mice within the same group using an unpaired t test. In all datasets, outliers were identified using Grubb's outlier test with the alpha value set at 0.01. The following outliers were removed from the final analyzed data: AMPA:NMDA ratio (one sham pair) and spine volume (one sham nonengram).
Results
HFHI causes retrograde amnesia that is reversible through engram stimulation
We have previously shown that HFHI mice have anterograde memory deficits (Sloley et al., 2021). To determine if HFHI mice also have retrograde memory deficits, we fear conditioned male C57Bl/6 mice, and 15 min after training we randomly assigned mice to sham or HFHI groups for 6 d (Fig. 1A). We also included a second cohort of mice that only received head impacts on the first day (5HI) alongside our established HFHI protocol. At 24 h after the final impact, we found that HFHI mice, but not 5HI mice, had a significant decrease in the time spent freezing in a 180 s context test (F(3,32) = 21.02; p < 0.0001). Post hoc analysis revealed that while 5HI mice did not display reduced freezing time, HFHI mice spent significantly less time frozen than either their sham control or 5HI mice (p < 0.0001; Fig. 1B).
HFHI mice display chronic retrograde amnesia following contextual fear conditioning. A, Schematic of the experimental design. C57Bl/6 male mice were fear conditioned, exposed to sham or HFHI protocols, and then tested for natural recall on day 6 (24 h after the final impact) and day 28. B, Multiple days of head impact are required to induce retrograde amnesia. A single day of five head impact (5HI) does not cause memory impairments; however, HFHI exposure over 6 d causes a significant decrease in memory recall (ANOVA with Tukey's post hoc test; ****p < 0.001). C, Sham and HFHI mice display similar learning curves, demonstrating that both groups of mice successfully acquired the fear memory prior to the HFHI protocols. D, At day 6, HFHI showed significantly decreased time freezing and increased latency to freeze compared with sham mice. E, At day 28, HFHI showed significantly decreased time freezing and increased latency to freeze. Unpaired t test, **p < 0.01, ***p < 0.001, ****p < 0.0001.
HFHI mice did not have any difference in their pre-head impact memory acquisition when compared with the sham mice (Fig. 1C). One day after HFHI (6 d after memory acquisition), we found an 82% decrease in freezing time (Fig. 1C; unpaired t test; NSham = 10; NHFHI = 6; p < 0.0001; t = 6.237; df = 14; effect size = 51.61%; 95% CI = 69.35:33.86) and increased latency to freeze in HFHI mice compared with sham (Fig. 1D; unpaired t test; NSham = 10; NHFHI = 6; p = 0.0007; t = 4.301; df = 14; effect size = 73.80%; 95% CI = 37.0:110.5). This retrograde amnesia remained at 1 month post-HFHI with a 96% decrease in memory recall (Fig. 1E; unpaired t test; NSham = 10; NHFHI = 5; p = 0.0061; t = 3.268; df = 13; effect size = 30.43%; 95% CI = 50.54:10.31) and increased latency to freeze (Fig. 1F; unpaired t test; NSham = 10; NHFHI = 5; p < 0.0001; t = 6.267; df = 13; effect size = 116.8%; 95% CI = 76.54:157.1) These data show that HFHI causes retrograde amnesia, but a single day of 5HI does not.
HFHI does not alter the number of engram neurons in the dentate gyrus memory engram
To study the effects of HFHI on engram neurons, we repeated our behavior experiments using male and female TRAP2/Ai32 mice. As these experimental mice were being used for electrophysiological and immunohistochemical experiments, they were injured in small batches of two mice per week (one sham and one HFHI), and their behavioral data was grouped at the conclusion of the experiments. We trained mice using the fear-conditioning protocol, injected mice with 4-OHT, and assigned them to sham or HFHI groups. Memory recall testing for these experiments occurred at 10 d post-training (5 d after final head impact), a timeline that was required to allow sufficient expression of ChR2-EYFP to occur to visualize engram neurons for electrophysiology (Fig. 2A). We found that HFHI mice displayed a 28% decrease in freezing time compared with sham mice (Fig. 2B; unpaired t test; NSham = 23; NHFHI = 21; p = 0.0477; t = 2.040; df = 42; effect size = 12.4%; 95% CI = −24.66:−0.1324). This decrease in freezing time was accompanied by a 101% increase in latency to the first freezing episode (Fig. 2C; unpaired t test; NSham = 23; NHFHI = 21; p = 0.0007; t = 3.675; df = 42; effect size = 18.11%; 95% CI = 8.165:28.05) and 29% decrease in the number of freezing episodes in HFHI mice compared with sham mice (Fig. 2D; unpaired t test; NSham = 23; NHFHI = 21; p = 0.0016; t = 3.375; df = 42; effect size = 6.145%; 95% CI = 9.819:2.471).
HFHI prevents reactivation of the engram without affecting engram size. A, Schematic of the experimental design for TRAP2/Ai32 mice used for the electrophysiology and immunohistochemistry experiments. TRAP2/Ai32 mice were fear conditioned, injected with 4-OHT, and randomly selected for HFHI or sham protocols. At day 10 a natural recall test was performed, and mice were prepared for ex vivo electrophysiology or perfused and fixed for IHC. B–D, HFHI mice display a retrograde amnesia phenotype compared with sham mice, as measured by percent freezing, latency to freeze, and the number of freezing episodes. Unpaired t test, *p < 0.05, **p < 0.01, ***p < 0.001. E, Confocal images of the dentate gyrus engram from sham (left) and HFHI (right) mice. green, ChR2-EYFP + neurons that were active at memory acquisition; magenta, c-Fos + neurons that were active at memory recall; blue, DAPI. F, The number of ChR2-EYFP + neurons in the dDG is similar between sham and HFHI mice. G, The number of c-Fos + neurons in the dDG is similar between sham and HFHI mice. H, The percent of overlapping c-Fos/ChR2-EYFP neurons is significantly reduced in HFHI mice compared with sham controls. Unpaired t test, *p < 0.05. Closed circles represent male animals and open circles represent female animals.
We previously showed that HFHI does not cause neuron cell death or diffusion injuries in major memory centers (Sloley et al., 2021); however, this does not preclude the possibility of alterations in c-Fos and other IEG expression profiles. To determine if HFHI alters the size of the memory engram, we quantified engram neurons in the dentate gyrus. Forty-five minutes after exposure to the 10 d recall paradigm, engram tagged TRAP2/Ai32 mice were perfused with PBS and fixed in 4% paraformaldehyde. The 45 min window is critical to allow for c-Fos protein expression to occur following memory recall. The number of ChR2-EYFP+ engram neurons (green, neurons active at the time of acquisition) and the number of endogenous c-Fos ensemble (magenta, active at the time of recall) were quantified in the dentate gyrus (Fig. 2E).
HFHI did not alter the total number of ChR2-EYFP+ neurons in the dentate gyrus compared with sham mice, demonstrating that HFHI procedure did not affect either the expression of c-Fos following memory acquisition and did not interfere with the labeling of the engram by 4-OHT (Fig. 2F; unpaired t test; NSham = 5; NHFHI = 6; p = 0.8546; t = 0.1885; df = 9; effect size = 0.5767%; 95% CI = −7.496:6.342). In addition, HFHI did not alter the size of the recall neuronal ensemble, as detected by c-Fos protein. We found similar numbers of c-Fos + neurons in sham and HFHI mice at 45 min following memory recall (Fig. 2G; unpaired t test; NSham = 5; NHFHI = 6; p = 0.6002; t = 0.5432; df = 9; effect size = 3.468%; 95% CI = −17.91:10.97).
Recent evidence shows that upon natural recall, there is reactivation of a subset of the engram ensemble (Reijmers et al., 2007; Ryan et al., 2015). We can detect this reactivation of the memory engram in our experiment by quantifying the overlap of c-Fos and ChR2-EYFP. Following exposure to recall cues, HFHI animals had a 50% reduction in the overlap of ChR2-EYFP+ and c-Fos+ neurons in the dentate gyrus (Fig. 2H; unpaired t test; NSham = 5; NHFHI = 6; p = 0.0128; t = 3.098; df = 9; effect size = 4.669%; 95% CI = −8.078:1.260). These data show that HFHI is not affecting the formation of a stable engram but does impair the ability of the engram to reactivate upon presentation of natural recall cues.
HFHI prevents engram-specific increases in the AMPA/NMDA ratio and AMPA receptor kinetics
The inability of HFHI mice to recall the memory in response to natural recall, despite similar memory engram size, suggests the presence of a physiological mechanism underlying HFHI-induced amnesia. Hippocampal engram neurons have a specific physiological profile that differentiates them from surrounding nonengram neurons following recall of the target memory (Ryan et al., 2015; Lee et al., 2023), including a greater AMPA/NMDA ratio in engram neurons. We therefore characterized the excitatory synaptic properties in engram neurons and assessed how these changes were affected by HFHI. Evoked responses were captured using a cesium internal solution, in the presence of gabazine, at −70 and +40 mV. The AMPA/NMDA receptor current ratio (Fig. 3A) and the AMPA weighted decay time (Fig. 3B) were recorded from engram and nonengram dentate gyrus neurons from ex vivo slices from the same mice. Sham engram neurons displayed 47.3% greater AMPA/NMDA ratio compared with sham nonengram neurons (Fig. 3C; unpaired t test; N = 13–15; p = 0.0462; t = 2.094; df = 26; effect size = 0.9975%; 95% CI = 0.01790:1.937); however, we did not detect a change in the AMPA/NMDA ratio in HFHI engram neurons compared with HFHI nonengram neurons (Fig. 3C; unpaired t test; N = 15–17; p = 0.3802; t = 0.8907; df = 30; effect size = 0.5536%; 95% CI = −0.71258:1.823).
Engram-specific excitatory synaptic changes are abolished after HFHI. A, Example traces from sham (left) and HFHI (right) engram (top) and nonengram (bottom) cell AMPA/NMDA ratios. B, Normalized trace of AMPA current shows slowed decay in sham engram neurons compared with sham nonengram neurons. Sham (top) and HFHI (bottom). C, Quantification of the AMPA/NMDA ratio show that sham engram neurons have an increased ratio compared with nonengram neurons. There is no change in the AMPA/NMDA ratio in HFHI engram neurons. D, Quantification of AMPA Tw. Sham engram neurons display an increase in AMPA Tw compared with sham nonengram neurons. HFHI mice display no engram specific changes to AMPA Tw. E, No NMDA decay time changes occurred in neurons in either sham or HFHI mice. F, No AMPA rise time changes occurred in engram neurons in either sham or HFHI mice. Unpaired t test, *p < 0.05, **p < 0.01. Closed circles represent male animals and open circles represent female animals. Black circles represent mean ± SEM.
In addition to differences in AMPA/NMDA ratio, we observed substantially slower AMPA decay kinetics of sham engram neurons compared with sham nonengram neurons (Fig. 3B). We quantified the weighted decay tau of the AMPA current and observed 46.8% greater AMPA Tw in sham engrams compared with their nonengram counterparts (Fig. 3D; unpaired t test; N = 14–15; p = 0.0075; t = 2,889; df = 27; effect size = 6.192%; 95% CI = 1.794:10.59). This difference in AMPA Tw did not occur in HFHI engram neurons (Fig. 3D; unpaired t test; N = 13–16; p = 0.8819; t = 0.1500; df = 27; effect size = −0.4345%; 95% CI = −6.378:5.509). This finding was specific to the AMPA current as neither NMDA Tw (Fig. 3E) nor AMPA rise time (Fig. 3F) displayed differences in either sham or HFHI engram neurons.
HFHI prevents engram-specific increases in dendritic spine volume
There is a direct relationship between the number and type of surface excitatory receptors and the morphology of the dendritic spine, with spine volume increasing as AMPA receptors are inserted into the synaptic membrane (Chater and Goda, 2014). Given that engram neurons have an increased AMPA/NMDA ratio, we examined the spine morphology of engram neurons compared with nonengram neurons. At the conclusion of our patch-clamp experiments, we injected biocytin into the patched neuron, imaged the neurons using confocal microscopy, and reconstructed the engram and nonengram neurons including the dendritic arborizations and dendritic spines (Fig. 4A).
HFHI abolishes engram-specific increases in spine volume. A, After electrophysiological recordings, nonengram (top) and engram (bottom) neurons were filled with biocytin for post hoc immunohistochemistry. B, Dendritic spines from sham engram neurons had significantly higher spine volume compared with nonengram neurons, but this differentiation did not occur in HFHI mouse brains. Two-way ANOVA with Šídák's multiple-comparisons test, *p < 0.05. C, D, Cumulative distribution plots of dendritic volume reveal the differentiation of engram and nonengram dendritic spines in sham, but not HFHI mice. Mean ± SEM, two-way ANOVA, *p < 0.05. There were no differences in (E) spine density, (F) total dendritic length, or (G) dendritic complexity between engram and nonengram neurons in either sham or HFHI. Closed circles represent male animals and open circles represent female animals.
In sham mice, we observed a 49.7% greater engram spine volume compared with nonengram spines (Fig. 4B; two-way ANOVA; NShamNonEng = 10; NShamEng = 10; NHFHINonEng = 10; NHFHIEng = 9; cell type, p = 0.0303; injury, p = 0.9710; interaction, p = 0.1124; Šídák's multiple-comparisons test; sham E vs NE, p = 0.0170; HFHI E vs NE, p = 0.8864) that was not observed in HFHI engram neurons. The cumulative distribution functions highlight this increased engram synapse volume in sham synapses (Fig. 4C; two-way ANOVA; NShamNonEng = 10; NShamEng = 10; bin × cell type, p < 0.0001; cell type, p = 0.0283), but not HFHI synapses (Fig. 4D; two-way ANOVA; NHFHINonEng = 10; NHFHIEng = 9; bin × cell type, p = 0.9999; cell type, p = 0.5710). We did not observe differences in spine density in engram versus nonengram neurons in either sham or HFHI neurons (Fig. 4E). We also did not observe any changes to the total dendritic length (Fig. 4F) or the dendritic complexity (Fig. 4G) of engram versus nonengram neurons in either sham or HFHI mice. Taken together, these data show morphological changes in sham engram dendritic spines that are associated with the excitatory synaptic changes (Fig. 3) that allow engram neurons to be differentiated from nonengram neurons in the same mice. HFHI interferes with these processes such that engram neurons and spines are not able to be differentiated from the nonengram neurons and spines.
HFHI-induced retrograde amnesia is reversible through engram stimulation
The mild nature of the head impacts delivered in our HFHI model and the subsequent lack of brain pathology makes it an excellent candidate for memory recovery through engram neuron stimulation. TRAP2 mice express iCre under the endogenous c-fos promoter, allowing tagging of neurons activated in a specific temporal window through the activation of iCre with tamoxifen and subsequent recombination of target constructs (Allen et al., 2017). When crossed with Ai32 mouse line which expresses channelrhodopsin (ChR2-EYFP), only in neurons where iCre-based recombination happens, ensembles of cells that were active at the time of a targeted contextual experience can be labeled and visualized (Madisen et al., 2010, 2012). In order to test this hypothesis, we used the TRAP2/Ai32 transgenic mouse line to label engram cells through expression of c-Fos (Fig. 5A). We fear conditioned TRAP2/Ai32 mice in Context A, labeled engrams with 4-OHT, and subjected mice to HFHI or sham conditions (Fig. 5B). Optogenetic cannulae were surgically implanted bilaterally into the dentate gyrus 10 d following the fear conditioning training, and mice were allowed to recover for 2 weeks before any testing occurred. To test for natural recall, the mice were placed in Context A and percent time spent frozen was determined. We found a significant difference between groups (one-way ANOVA; NSham = 7; NHFHI = 9; NNo Shock = 12; F(2,25) = 6.55; p = 0.0051), and post hoc analysis revealed a 64.1% decrease in freezing between sham mice and HFHI (p = 0.0148) and a 67.7% decrease in freeze time between sham and no-shock controls (p = 0.0063). There was no difference in freeze time between HFHI mice and no-shock controls. To test whether optogenetic stimulation of the engram can induce recall, we placed mice in a novel Context B for a 6 min session consisting of 3 min of habituation and 3 min with optogenetic activation of the dentate gyrus engram using a theta/gamma nested stimulation paradigm. Three-way ANOVA (NSham = 7; NHFHI = 9; NNo-Shock Sham = 10; NNo-Shock HFHI = 12) revealed a significant effect of light (F(1,34) = 39; p < 0.001), shock (F(1,34) = 28; p < 0.001), and a significant light × shock interaction (F(1,34) = 25; p < 0.001). There was no significant effect of injury or any interaction of injury with either light or shock or any three-way interactions. Post hoc analysis revealed that light stimulation in sham mice caused a 112% increase in freezing compared with the baseline (p = 0.0001). Similarly, in HFHI mice light stimulation caused a 121% increase in freezing compared with that in baseline (p = 0.0004). In contrast, sham and HFHI mice that did not receive any shocks during the acquisition phase (no-shock controls) but did receive 4-OHT to activate engram neurons after no-shock training did not have any change in freezing during the light stimulation epoch (Fig. 5D). These data show that similar to sham mice, the fear memory can be elicited from HFHI mice, demonstrating that it is possible to override retrograde amnesia caused by HFHI in mice.
Amnesia after HFHI is recovered by activating the DG engram. A, Diagram for transgenic engram labeling strategy using TRAP2/Ai32 mice. B, Schematic of the experimental design. TRAP2/Ai32 mice were fear conditioned and engram neurons labeled with 4-OHT. Following sham or HFHI procedures, bilateral hippocampal cannulae were surgically implanted and mice were allowed to recover for 2 weeks before being tested for optogenetic stimulation in a novel context and natural recall. C, For optogenetic stimulation of the memory engram, mice were placed in novel Context B and the engram stimulated with nested theta/gamma stimulation of the dentate gyrus. Percent freeze was quantified for a 3 min light off epoch and a 3 min light on epoch. Optogenetic stimulation elicited a similar increase in the freezing in both sham and HFHI mice. This response is absent in no-shock (NS) controls. Three-way repeated measures ANOVA with Šídák's multiple-comparisons test, ***p < 0.001. Closed circles represent male animals and open circles represent female animals. D, Mice were tested for natural recall at day 30. Quantification of the percent time freezing demonstrate that HFHI animals have retrograde amnesia compared with sham mice. Dotted line represents average no-shock freezing time in Context A. Unpaired t test, *p < 0.05.
Discussion
In this study, we explored why repetitive head impact causes retrograde amnesia, focusing on memory engrams in the dentate gyrus of the hippocampus. We stimulated memory recall in HFHI mice using optogenetic stimulation of the memory engram, demonstrating that retrograde amnesia in HFHI mice is due to aberrant engram physiology and not irreversible neuronal damage. We report that the memory engram neurons are still physically present in HFHI mice, but in the head impact brain the engram is not able to physiologically differentiate itself from the surrounding nonengram neurons. In sham mice, we first validated recent research showing that engram neurons have a specific neuronal signature that differentiates them from the nonengram network (Ryan et al., 2015; Roy et al., 2016; Kim and Cho, 2017; Pignatelli et al., 2019; Sun et al., 2020; Naik et al., 2021; Lee et al., 2023). Next, we revealed new aspects of engram-specific plasticity that have not been previously reported including changes in AMPA Tw in c-Fos labeled engram neurons. Finally, we add to the literature on amnesia to identify engram-specific changes that cause head impact-induced memory loss and highlight engram-specific changes that are maintained despite the manifestation of amnesia. Our results provide new data on how memories are stored and retrieved, uncover new insights on the mechanisms of amnesia, and reveal new targets for treating head impact-induced memory loss.
The hippocampal memory engram is a network of neurons that are active at the time of learning and contribute to the encoding of a new memory. Following memory recall, a unique neuronal signature is present in engram neurons that allow them to be differentiated from their nonengram counterparts in the same brain region. Previous reports show that engram-specific changes occur to the neuron and at the excitatory synapse, with an overall increase in synaptic excitability in the engram neurons compared with that in nonengram neurons (Ryan et al., 2015; Roy et al., 2016; Kim and Cho, 2017; Pignatelli et al., 2019; Sun et al., 2020; Naik et al., 2021; Lee et al., 2023). We confirmed that sham engram neurons have an increased AMPA/NMDA ratio and found a novel engram neuron phenotype in increased AMPA Tw, indicating an increased decay time for engram AMPA current. Both these phenotypes would increase the excitability of the engram neuron. Additionally, there is a direct relationship between the number and type of surface excitatory receptors and the morphology of the dendritic spine, with spine volume increasing as AMPA receptors are inserted into the synaptic membrane (Chater and Goda, 2014), and there is evidence of structural plasticity in the engram (J. H. Choi et al., 2018; D. I. Choi et al., 2021; Hwang et al., 2022). Similar to previous reports on a change to dendritic spines on engram neurons, we found an increased spine volume on engram neurons in sham mice.
Memory loss is one of the most common symptoms of head injury (Frost et al., 2013; Harmon et al., 2013; Gardner and Yaffe, 2015), and almost all animal models of mTBI and severe TBI present with anterograde amnesia, which can be caused by a combination of both memory encoding and retrieval deficits (Washington et al., 2012; Xiong et al., 2013). Retrograde amnesia is studied much less frequently in TBI models, but it has been reported in severe models such as CCI and lateral fluid percussion (Dash et al., 2002; Witgen et al., 2005; Chen et al., 2009; Quigley et al., 2009; Kaufman et al., 2010). This amnesia can be partially rescued by preventing cell death and salvaging hippocampal neurons, which speaks to the role of neuronal networks in memory retrieval (Dash et al., 2002; Witgen et al., 2005; Chen et al., 2009; Quigley et al., 2009; Kaufman et al., 2010). The HFHI model is much less severe than other TBI models; however, HFHI mice also have impaired learning and anterograde memory deficits (Sloley et al., 2021), and here we demonstrate that HFHI also induces retrograde amnesia. HFHI does not cause retrograde amnesia by interfering with memory consolidation, as we found that five head impacts delivered 15 min after training (which is the same day 1 protocol for HFHI) does not cause a significant reduction in memory recall at 24 h post-head impact. HFHI amnesia is also not caused by a reduction of engram neurons associated with the memory, as the number of engram neurons in the dentate gyrus of HFHI mice was similar to sham mice. Rather, there was impaired reactivation of the engram in HFHI mice, as revealed by a reduction in the number of engram cells that were positive for c-Fos protein following presentation of natural recall cues. In addition, we found that HFHI engram neurons did not develop the characteristic changes in the AMPA/NMDA ratio or AMPA Tw. In addition, spine volume was not increased in HFHI engram neurons. It is this inability of the HFHI engram neurons to differentiate themselves from the surrounding nonengram neurons that impairs reactivation when stimulated with natural cues. Very few studies have used electrophysiology to study the engram phenotype, and how it is impacted by amnesia (Ryan et al., 2015; Roy et al., 2016; Naik et al., 2021), and our work strengthens the concept that AMPA insertion and synaptic plasticity are key components of a functional memory engram, and it is the inability to replicate these key components that causes memory failure. Thus, we hypothesize that it is the inability of the engram neurons to differentiate themselves from nonengram neurons that underlies the inability of the HFHI mice to fully recall the memory—essentially the memory exists, but with a high background signal-to-noise ratio. The consequence of this noise is that the memory can either be impaired when presented with aversive stimuli or can be inappropriately activated by neutral stimuli. A recently published study on the engram in a severe TBI model shows that memory traces are altered in the brain injury mice such that fear generalization occurs where mice cannot distinguish between neutral and aversive stimuli (McGowan et al., 2024). In an Alzheimer's disease mouse model, unrelated memory engrams compete for activity and impair normal memory recall (Poll et al., 2020). These data suggest that when distinguishing synaptic features of the engram are absent, it cannot outcompete other ensembles for activity when recall is required.
Given that the engram neurons carrying the specific memory are not lost after HFHI, we tested the hypothesis that the forgotten memory could be fully retrieved many weeks after the head impacts had ceased. Such stimulated recall has been demonstrated for retrograde amnesia induced by Alzheimer's disease, pharmacologically disrupted memory consolidation using protein synthesis inhibitors, and sleep deprivation (Ryan et al., 2015; Roy et al., 2016; Perusini et al., 2017; Poll et al., 2020; Power et al., 2022; Bolsius et al., 2023). These seminal papers demonstrated that the trace of a forgotten memory still exists in the engram and that activation of this circuit is sufficient to drive memory recall, and here we add head impact-induced amnesia as a potentially reversible event. We found that optogenetic activation of the dentate gyrus engram could completely recover fear memory in HFHI mice, demonstrating that the memory trace remains intact in the HFHI brain. Our study tells us several things about head impact-induced amnesia. Contextual memory is reliant on hippocampal circuitry, and here we report that the dentate gyrus of the hippocampus is able to form a unique memory engram in HFHI mice but is unable to initiate the required synaptic strengthening required for this to be a functional engram. This follows our recent characterization of the HFHI model where we described a decrease in excitability in CA1 neurons of the hippocampus and a decrease in the AMPA/NMDA ratio. Together, these data show that hippocampal excitatory synapses are unable to normally respond to endogenous signals, resulting in both anterograde and retrograde amnesia. It is possible that therapies that can stimulate or modulate synaptic plasticity, such as transmagnetic stimulation (TMS) or transcranial direct current stimulation (tDCS), will have the ability to alleviate the cognitive symptoms of repetitive head impact (Villamar et al., 2012). In contrast to a progressive neurodegenerative disease such as Alzheimer's disease, the head impacts causing trauma-induced memory loss are time-limited events that can be prevented from continuing (e.g., retiring from the sport that is causing the head impacts). This allows us to deliberately intervene in cases of cognitive dysfunction to halt the head impacts and design individual therapies to stimulate synaptic plasticity and restore function.
In conclusion, memory loss is one of the most common symptoms of head injury, and here we use memory engram mice to probe the mechanisms that cause these cognitive deficits. We find that poor reactivation of memory ensembles in HFHI mice causes retrograde amnesia, and this poor reactivation is due to an inability of the engram network to develop the neural correlates necessary to distinguish itself from the surrounding neurons. Our data adds to the literature showing that the synaptic changes occurring in engram neurons serve as a gain–control mechanism to improve the fear memory engram's signal-to-noise ratio against other ensembles in the hippocampus. Interventions targeting synaptic plasticity may have therapeutic potential for treating memory impairments caused by repeated head impact exposure.
Footnotes
This research was supported by the Mouse Behavior Core in the Georgetown University Neuroscience Department. Experimental timeline graphics created under license in biorender.com. This work was supported by the National Institutes of Health (NIH)/National Institute of Neurological Disorders and Stroke (NINDS; R01NS107370 and RF1NS121316 to M.P.B.), NINDS also supported F30NS122281 to D.P.C. and the Neural Injury and Plasticity Training Grant housed in the Center for Neural Injury and Recovery (CNIR) at Georgetown University (T32NS041218 to D.P.C.). Seed funding from the CTE Research Fund at Georgetown University.
↵*S.V., T.J.R., and M.P.B. contributed equally to this work.
The authors declare no competing financial interests.
- Correspondence should be addressed to Mark P. Burns at mpb37{at}georgetown.edu.
