Concussion Increases CA1 Activity During Prolonged Inactivity in a Familiar Environment

Abstract

Although hippocampal damage plays a key role in impairments after concussion, differences in hippocampal information processing during recovery are unknown. Micro-endoscopic calcium imaging after concussion showed that CA1 activity increased in awake, immobile animals in a familiar environment, but was unaltered when animals explored a novel environment. As awake immobility parallels cognitive rest, a common treatment for patients, the results imply that prolonged cognitive rest may unwittingly impede concussion recovery.

The architecture of neural circuits and the coordination of their activity are instrumental for cognitive function. Mechanical impact to the brain (concussion or mild Traumatic Brain Injury; mTBI) disrupts the coordination between neural circuits by damaging their architecture and leads to pronounced changes in cognition that can affect the quality of life for years (Meaney et al., 2014). The circuits of the hippocampal formation involved in learning and memory are among those affected by mTBI (Paterno et al., 2017). However, despite decades of research and numerous in situ studies on the electrophysiological changes occurring in the hippocampal formation (Atkins, 2011), it is unknown how hippocampal circuits respond in vivo after mTBI, especially as the animal explores and processes information about its environment. Here, through calcium imaging in area CA1 of animals behaving freely in their familiar home cages (HC) and a novel open field (OF), we report for the first time that mTBI modifies CA1 activity during prolonged awake immobility in the familiar HC but does not impede the ability to detect contextual novelty in an OF.

Area CA1 is integral for detecting and encoding contextual novelty (Lisman and Otmakhova, 2001). Previous work shows enhanced inhibition and reduced Schaffer collateral evoked field responses in CA1 after mTBI (Beamer et al., 2016; Chen et al., 2018; Johnson et al., 2014; Witgen et al., 2005). To appreciate how this increased inhibition affects CA1 circuits in vivo, we designed a simple experimental paradigm that observed CA1 activity in both familiar and novel environments on two consecutive days prior to mTBI (−2d, −1d) and on the fifth day (5d) after mTBI (see methods, Figures 1a, b, Supplementary Figure 1). Regions of interest (ROI) in the calcium images were first identified automatically(Zhou et al., 2018) and those corresponding with distinct cell bodies (neurons) were retained after manual inspection (Supplementary Figure 2). The total number of active neurons varied across days in each of the animals (Figure 1c). Although the total number of active cells appeared to increase after mTBI, the difference was not significant (p=0.652, Friedman’s one-way repeated measures ANOVA). Of all the neurons active on any given day, we found that about 13% were active in both the HC and OF environments (Figure 1d). Examining the event rates (Figure 1e) and amplitudes (Supplementary Figure 3b) of these cells, we found that mTBI did not alter the correlation in their activity in the two environments. The remaining neurons were uniquely active in either the HC (around 42%) or OF (around 45%). These fractions were similar across all the days and were not altered by injury. However, as images could not be satisfactorily registered across days, we were unable to judge if the active population in each environment was the same on all the days.

Figure 1.

a. Modified base plate design enabling better focus. Inset: Imaging setup. b. Experimental timeline. Closed arrows in row 3 indicate the start of imaging in the environment. Animals in their HCs were placed in the imaging enclosure twice for 10 minutes on the day of (0) and on 1d, after which cages were changed. c. Total number of active neurons in both environments on each day. All group sizes: n =10 animals (−2d & −1d) & n=4 animals each (sham & mTBI). Data from two casualties of blast was retained in the pre-injury groups. All bar graphs: grey open circles represent individual animals and data is presented as mean ± sem. d. Fraction of neurons active in each environment on the different days. Data is mean ± sem. e. Event rates of individual neurons active in both environments from all animals. Colors for each group (black: −2d, grey: −1d, blue: sham, green: mTBI) are preserved throughout. f. Average frequency histograms of event amplitudes (solid lines from Supplementary Figure 3e), and bar graphs showing mean amplitudes. g. Frequency histograms of calcium event rates. Left: Histograms in both environments on the imaging days. Light grey lines represent histograms of individual animals. Solid and dashed colored lines represent the average and standard deviation of the individual animal histograms in each group. Right: Average histograms (solid lines in left panels) in the HC (standard deviations of sham and mTBI are shown in the inset figure) and OF, and bar graphs showing distribution medians. * p=0.0286 for both, Mann-Whitney test. g, h, i Fraction of time animals were (g) quiet or active in the 2 environments, (h) in the center and periphery of the OF, and (i) explored each wall of the OF.

Given our previous finding of reduced excitability in CA1 after blast mTBI (Beamer et al., 2016), we first asked if there was a decrease in amplitudes of the calcium transients (Figure 1f, Supplementary Figure 3d). Mean amplitudes in the injured group were slightly smaller than those in sham but not significantly different (p>0.200 for both HC and OF, Mann-Whitney test). Next, we focused on the temporal dynamics of CA1 neurons in the two environments. Parsing the fraction of cells active in each recorded minute in both environments (Supplementary Figure 3e), we observed that a larger number were active initially in the first minute or two of recording after which the fraction stabilized, and that the trend persisted after mTBI. Event rates in the familiar HC and novel OF were distributed similarly across both pre-injury days and on 5d in the sham group, emphasizing the stability of the responses (Figure 1g). Median pre-injury and 5d sham event rates in the HC were slightly but not significantly smaller than those in naïve mice in a familiar OF on equivalent days (Supplementary Figure 4), suggesting that activity in the HC in a familiar context on 5d may be a measure of basal hippocampal activity. Unexpectedly, mTBI did not alter activity in the novel OF (Supplementary Figure 3c). More surprisingly and most strikingly, we found that mTBI significantly increased the median activity of neurons in the HC compared with those in the OF and sham HC (p=0.028 for both, Mann-Whitney test; p=0.037, Friedman’s one-way repeated measures ANOVA).

We then asked if the increased neuronal activity in the HC was due to an increase in the animals’ activity. Behavior was automatically classified into active and quiet intervals (see methods, Supplementary Figure 1). Mice were mostly quiet in their HC on all days (Figure 1h). Interestingly, we found that this behavior was more pronounced in the injured mice; 3 out 4 of which were quiet for the entire duration of the HC recording. In stark contrast, we found that mice were very active in the novel OF on all the days (Figure 1h), constantly moving throughout the field (Figure 1i) and examining all four novel walls by leaning and sniffing (Figure 1j). Together, these results indicate that while the ability of CA1 to detect contextual novelty is unimpaired, its basal state is altered after mTBI. Finally, we asked if the observed changes in activity after mTBI altered the dimensionality of the population i.e., the number of network states (Carrillo-Reid et al., 2017). There was no change in the dimensionality in either the HC or OF (Supplementary Figure 3g).

In summary, our results show that concussion increases CA1 activity during prolonged awake immobility in the familiar HC. We speculate that this finding is the result of increased excitation of the CA3 recurrent system due either to decreased subcortical modulation or failure of the local hippocampal mechanisms normally suppressing recurrent excitation after injury (Buzsáki, 2015). The explanation, if accurate, predicts an increase in the frequency of sharp wave ripples (SWRs) in CA1 during prolonged awake immobility in the HC, providing a locus for commonly reported memory impairments after concussion (Buzsáki, 2015). More broadly, our findings have major implications for post-concussion treatment in the military and civilian population. Cognitive rest, part of the clinical guidelines for treatment, is prescribed to reduce the metabolic load associated with neural processing and thereby accelerate recovery (Barkhoudarian et al., 2016). Intriguingly, some clinical studies show that cognitive rest does not impact recovery time but on the contrary prolonged cognitive rest exacerbates symptoms and delays their resolution (McLeod et al., 2017). As awake, immobile animals share many features of cognitive rest treatment in humans, our results imply that prolonged rest for the recovering brain may increase the metabolic burden on the brain, and raise the need for extensive work examining post-concussion metabolism during different brain states. As concussion can affect the fundamental processes of memory, mood and sleep simultaneously, it is hoped that these results will enable effective treatment for mTBI.

Methods

All experiments were performed on male C57BL6 mice (Charles River Laboratories) in accordance with the guidelines of the Institutional Animal Care and Use Committee (IACUC) of the University of Pennsylvania. For all survival procedures requiring anesthesia, anesthesia was induced with 3% isoflurane mixed with oxygen (100%, Air Gas, PA) and maintained with 1–2% of the same (the percentage of isoflurane was lowered to 2% immediately after the animal was under and subsequently lowered in decrements of 0.5% as the duration under increased). Depth of anesthesia was frequently gauged by the paw pinch test and the percentage of isoflurane was increased to a maximum of 2% if a reflex was observed. Animals were placed on a regulated heating pad throughout the surgical procedures and their eyes were lubricated to prevent corneal damage under anesthesia. Their weight and general health were monitored once a week throughout the study.

Virus Injection

AAV9.CamKII.GCaMP6f.WPRE.SV40 and AAV9.Syn.GCaMP6f.WPRE.SV40 were obtained from the Penn Vector Core. Mice were unilaterally injected in either their right (n=10) or left (n=9) dorsal CA1 (RC: 1.8–2.0 mm from bregma, ML: 1.4–1.6 mm from midline and DV: 1.3–1.4mm from the surface of the brain) with 1μl of either hSyn.GCaMP6f (4.75e¹² GC ml⁻¹, n=3; 2.37e¹² GC ml⁻¹, n=1 was used only for behavior control) or CamKII.GCaMP6f (1.14e¹³ GC ml⁻¹, n=8; 1.14e¹² GC ml⁻¹, n=7). After induction of anesthesia, animals were placed on a stereotaxic frame (Stoelting Co., IL) and their heads fixed. The skin on their heads was cleaned with povidone-iodine solution (10%, Betadine® Solution) and isopropyl alcohol (70%) after which an incision was made in the middle. Bupivacaine hydrochloride (2 mg kg⁻¹) was injected into the periosteum and the latter cleared with a cotton swab. Hydrogen peroxide (3%) was lightly dabbed on the sutures to better expose them. A small hole was carefully drilled with a burr (0.5mm tip diameter, Fine Science Tools Inc., CA) and the needle (26s/2”/2, Hamilton Company, NV) was positioned at the aforementioned coordinates. The tissue was allowed to settle for 2 minutes before starting infusion (0.125μl min⁻¹, Stoelting Co.). Two minutes after completion of infusion, the needle was slowly retracted over the course of 5–10 minutes and the skin sutured (5–0 dissolvable sutures, EMSCO Scientific Enterprises Inc., PA). Meloxicam (5 mg kg⁻¹) was administered subcutaneously at the nape prior to the start of the procedure and one day after injection.

Preparation of Hippocampal Window

One (n=17) or two (n=2) weeks after virus injection a stainless steel (1.3mm inner diameter, HTX-15R, Component Supply, TN) cannula sealed with a glass cover slip (0.13–0.17mm, Fisher Scientific, MA) at one end was inserted above the hippocampus as follows. Animals were first head fixed, their skulls exposed and cleared of overlying tissue as above. The skull was then thoroughly cleaned with hydrogen peroxide (30%, Sigma Aldrich) and the skin glued (Vetbond ^TM/MC, No. 1469 SB) along the circumference. A circular aperture centered at the injection site was slowly drilled into the skull with a trephine (1.8mm tip diameter, Fine Science Tools Inc.) and the brain exposed. The dura was removed, and the neo-cortex and external capsule were continuously flushed with sterile saline and gently aspirated as described(Dombeck et al., 2010) to expose the underlying hippocampus. Bleeding was stemmed with Gelfoam® (Midwest Veterinary Supply, MN), subsequent to which the cannula was gently inserted into the saline filled aspirated cavity. Once positioned in place, Kwik-Sil^™ (WPI, FL) was applied around it to hold it in place. Dental cement was then applied uniformly on the entire exposed skull to secure the cannula and provide a scaffold for the baseplate of a miniature microscope (iNSCOPIX, CA). Meloxicam (5 mg kg⁻¹) was administered subcutaneously at the nape prior to the start and for two days after the procedure.

Baseplates were cemented at least one, but typically four, days prior to the commencement of the study. The original camera base plate was customized to move both vertically and horizontally to ensure that the cells were always in focus (Figure 1a). Briefly, animals were placed on a sterotax under anasesthesia and their cannulas cleaned with distilled water. A GRIN lens, (1.0 mm diameter, iNSCOPIX) that serves as the relay lens between the glass cover slip of the cannula and the miniature microscope, was then inserted into the cannula. The lens was secured to the cannula to prevent displacement during imaging with silicone grease applied very lightly around its circumference halfway through its length. The baseplate was centered so that the complete field of view of the GRIN lens was visible through the camera and the assembly was then cemented to the cement scaffold with additional dental cement. Upon conclusion of the study, animals were deeply anaesthetized with pentobarbital (100 mg kg⁻¹) and perfused with 4% paraformaldehyde. The brains were removed, and the location of the imaging window was confirmed.

Primary Blast Injury

To enable the collection of pre and post injury data in the same animal, mTBI was caused by blast overpressure to the head as described previously(Beamer et al., 2016). Briefly, animals were first deeply anaesthetized with isoflurane (3% for induction for 5’ and 2% for maintenance for 3’) and then placed on a holder outside the exit end of a shock-tube with their snouts facing the tube. Their torsos and extremities were protected from the shockwave by sorbothane-lined aluminum casing and their heads constrained from moving with a thin metal rod encircling the snout and a cervical collar positioned between the occiput and shoulders. A subset of animals (n=8; mTBI group) were then exposed to a mild blast peak incident overpressure of 215 ±13 kPa, duration of 0.65 ± 0.04ms, and an impulse of 46 ± 5 kPa*msec. A subset of animals (n=7; sham group) were not exposed to the blast and served as procedural controls. After blast or sham exposure, the righting time was used to assess severity of injury. Mice took 143.84 ± 120.41 (mean ± std) seconds to right after injury and 85.22 ± 92.40 seconds after sham procedures. Blast did not displace or damage the imaging preparation unless the cement scaffold securing the cannula was insufficiently adhered to the skull, in which case it dislodged the intact preparation from the brain. In the latter event, animals were euthanized immediately.

Imaging

Activity of CA1 neurons was imaged over the course of two weeks starting at least 7 weeks after virus injection when the animals were between 13–20 weeks old. Animals were handled once a week for 1–2 minutes in the interim and on the day before the first imaging session. On each imaging day, animals were briefly (6.70 ± 3.81, mean ± std) anaesthetized and the miniscope was mounted onto its baseplate. Upon completion of imaging, the miniscope was usually removed by constraining the mouse briefly with one hand. Anesthesia was only induced if the animal was difficult to constrain making the maneuver difficult.

Animals were transferred to their own separate cages (referred to as home cage in this manuscript) at least one day prior to the start of the imaging study but typically on the day the base plate was attached and were housed individually for the two-week duration. Cages were changed once a week and cage changes were avoided in the middle of the block of consecutive imaging days (first change was performed on the first day after injury). Images, on all the days of study, were acquired during the day in a small enclosure near a corner of a dark room (Supplementary Figure 1a) under red light. The time of the day at which the activity of any given animal was imaged was kept approximately the same (within 25 minutes of the start time on the first day of imaging) for the animal throughout the study. After mounting the camera and adjusting the focus, animals were allowed to recover from anesthesia in their home cage (HC) for at least an hour (proportionally more if the duration under anesthesia was a little longer than usual) on a table perpendicular to the imaging enclosure (see Supplementary Figure 1a). Activity was imaged in two environments for ten minutes each. First in the animal’s HC and then in a ‘novel’ open field (45.1 × 45.1 × 17.6 cm polycarbonate arena) with ten minutes of rest in between each, during which it was replaced in its HC on the recovery table. To maintain animals’ familiarity with the imaging enclosure and procedures on 5d, animals were placed in their HCs in the enclosure twice on the day of (day 0; prior to and after injury) and 1d after injury (separated by 30 minutes), for the same duration as on the imaged days. Calcium transients were captured at a rate of 20 frames s⁻¹. The LED illumination power and gain for a given animal were kept constant on all the days it was imaged. The LED illumination power for each animal was determined on the first day as the minimum power that clearly showed individual cells and the gain was set at one for all animals. Behavior was simultaneously captured at a rate of ~8 frames s⁻¹ with a webcam controlled by user-defined MATLAB® scripts. The rate of acquisition of behavior data (~ 8 frames s⁻¹) was constrained such that the least number of frames were dropped by the iNSCOPIX acquisition software. The acquired data was subsequently processed in MATLAB®.

Novelty in the Open Field

Novelty in the open field (OF) arena on each of the imaging days was provided by completely covering its walls with materials of different textures and with different pictures (see Supplementary Figure 1b). Textures ranged from that of the bare polycarbonate of the arena to reusable coarse domestic shelf liners and file enclosures containing pictures to disposable brown paper of grocery bags, crinkly tissue paper and the smooth and highly reflective printer paper. Pictures varied from simple shapes such as a circle or triangle to a more complex pattern of squares or tiles to natural scenes such as grass. The reusable textures viz., shelf liners and file holders were doused with ethanol and cleaned between animals on the same day and at the end of the day they were washed with soap water. The disposable paper textures were replaced with new ones for each animal. The arena was thoroughly wiped with water and ethanol, and the floor was covered with a fresh piece of laboratory shelf liner before placing an animal. The floor on each of the days was randomly arranged with chocolate shavings. Animals were given their first taste of a small shaving of chocolate on the day that they were transferred to their own cage. Animals (n=5) prepared for but excluded for imaging due to the absence of visible cells were used to control for behavior in the absence of chocolate shavings (Supplementary Figure 1f).

Analysis of Behavior

The animal’s image in each frame of the behavior sequence was automatically segmented and its centroid determined as described previously (Patel et al., 2014). Due to movement of the miniscope wire and the dim red lights, centroids were frequently off the animal’s image and were corrected manually in those frames. The image centroids in each frame were used to mark the animal’s position in the environment in all subsequent analysis. Behavior in both environments was automatically divided into two intervals of ‘active’ and ‘quiet’ based on empirical displacement of the centroid (Supplementary Figure 1c). These intervals were subsequently manually checked by an investigator blind to the treatment group. In the manual classification, active intervals included the larger displacements of ambulation, leaning, rearing and smaller displacements of turns and limb extensions. Quiet intervals excluded all movements except for small head movements and centroid jitter. The automatic classifier overestimated quiet times by approximately 5% in the HC and 16% in the OF (Supplementary Figure 1d). Results from the automatic classifier are reported here. For the standard OF measures of time spent in center and periphery, the periphery was set at 11% of the total area i.e., approximately 5cm from each wall.

Analysis of Calcium Transients or Events

The acquired time-series in the two environments on each day were processed individually. Images in a series were registered to correct for horizontal or in-plane motion and normalized by the mean intensity image after subtraction of the same (background subtraction, $\frac{dF}{F}=\frac{fij-Fi}{Fi}$, f_ij is the intensity in pixel i in frame j and F_i is the mean intensity in pixel i over time). Regions of Interest (ROI) and the transients therein were first automatically extracted from the raw motion corrected time-series using constrained non-matrix factorization (CNMF-E) (Zhou et al., 2018). To ascertain that the ROI corresponded with a neuron and that the calcium transients were true calcium events in the neuron and not the result of bleed through fluorescence from adjacent overlapping cells or local neuropil fluorescence, each ROI and transient associated with it was then manually inspected in the background subtracted time-series (Supplementary Figure 2) as follows. First, maximum intensity projections and waveforms of the ROI from the background subtracted time-series in its entirety were examined and all ROI that did not obviously correspond with cells were discarded. Second, local maximum intensity images of all transients (i.e., maximum intensity within the frames comprising FWHM of a transient) with amplitudes greater than 10 in the CNMF-E waveform of a retained ROI were generated from the background subtracted time-series and all transients that were contained within the ROI with amplitudes larger than 3 standard deviations of the dF/F waveform were retained. The dF/F waveforms were then cleaned by setting all points except those of the retained transients to zero. These waveforms were used for all subsequent analysis.

ROI common to both time-series on each day were identified as follows. ROI from both time-series with centroids within 3×3μm square neighborhood and distance less than 4.5μm of each other were first identified. The boundaries of these were then overlaid on the maximum intensity projection images of the two series and their correspondence manually verified. All pairs that did not enclose cells in both images were discarded.

Event rate was calculated as the number of above threshold transients (events) in the entire duration of recording. Transients that started prior to the start of the recording and decayed after the recording was stopped were excluded. Event rates were found to be below 0.035Hz and were similarly distributed between neurons transfected with GCamp6f under the CamKII and Synapsin promoters (Supplementary Figure 3a). Population rate vectors with a bin size of one-minute were constructed from z-scored number of transients that started within the one-minute intervals. The eigenvalues of their covariance matrices were determined, and the population vectors were projected onto the first two eigenvectors or principal components to observe the number of network states or dimensionality of the population (Carrillo-Reid et al., 2017).

Statistics

Data was plotted in MATLAB® or GraphPad (San Diego, CA) and the final figures assembled in Adobe Illustrator®. Non-parametric statistics were used for desired comparisons (GraphPad). Comparisons were made with each animal treated as an independent sample. A p-value of less than 0.05 was considered significant.

Supplementary Material

Supp 1

Supplementary Figure 1. Exploratory behavior of injured animals in a novel open field (OF) on the different days is unaltered by chocolate shavings. a. Schematic of the experimental area. Mice were placed in their home cages (HCs) on the recovery stand to recover from anesthesia used to mount the camera and in between imaging sessions. Imaging was performed in the imaging enclosure, a portion of the table isolated on all sides with white paper board. The lower half of the opening was closed with a paper board during recording sessions. b. Examples of wallpapers used to make the OF novel on each day. Mice were exposed to each texture and picture only once. Chocolate shavings were placed in random grids on the floor mat (black dots in bottom left image). c. Automatic classification of behavior into active and quiet intervals. Animals’ image centroids were represented by a position vector from the origin of the image coordinates (top left-hand corner) as shown in the images. The magnitude of this vector (arbitrary units) was then plotted as a function of time (plots). Critical points were roughly estimated as the point where the gradient reverses and the interval between these points were classified as quiet if the difference in magnitude of the position vector was less than 0.7 cm. The automatically classified intervals (represented by black dots on the magnitude plots) were then compared with manually classified ones (green crosses). d. Difference in the automatically and manually estimated total time animals were quiet on each of the days in the two environments expressed as a fraction of total time. e. Behavior in novel OF in the absence of chocolate shavings. Animals that were prepared for calcium imaging but in which no fluorescence was detected (n=5) were used to determine the effect of chocolate. The procedure on each imaging day was similar to that used for calcium imaging. The iNSCOPIX dummy camera was mounted under brief anesthesia and animals were placed in their HC on the recovery stand during recovery. For consistency with the calcium imaging cohort, behavior was first imaged in the HC and then in the OF with a break in between. n=2, 3 for mTBI and sham respectively.

Supp 2

Supplementary Figure 2. Manual verification of ROI and transients extracted using CNMF-E. a. An example of CNMF-E output: spatial footprint (left) and its associated temporal waveform (right). b. Obtaining ROI and dF/F waveforms from the CNMF-E output. The cores of the SF (enclosed by dotted lines on left figure) were used as ROI. Middle: ROI boundary on the dF/F maximum intensity image in the acquired time-series. The temporal waveforms used for analysis were extracted from the dF/F time-series. c. Examples of ROI determined by CNMF-E. All the ROI generated by CNMF-E did not correspond with cell bodies or the waveforms with physiological signals. The ROI were ascertained to enclose cell bodies by overlaying their boundaries on the local (restricted to the pixels comprising the spatial footprint) maximum intensity image in the acquired time-series. ROI that corresponded with cells and had physiological signals were retained (left column) and the rest discarded (right column). In this and all subsequent figures red traces represent the CNMF-E waveform and black traces the extracted dF/F waveform. d. Examples of retained and discarded transients from different ROI. Transients (numbered on the waveform) with amplitudes greater than 10 arbitrary units (threshold indicated by dotted lines on the waveforms) were extracted from the CNMF-E waveforms of the retained ROI and a local maximum intensity image for each i.e., maximum intensity in the frames containing its FWHM was generated (numbered in white). These were then individually inspected to ascertain that the signal was contained within the ROI. Transients that were completely enclosed (numbered in green) and whose amplitude in the dF/F waveform exceeded 3 standard deviations of its total signal (threshold indicated by solid lines on the waveforms) were retained and the rest of the waveform set to zero. e. Retaining transients from overlapping ROI. When selecting the ROI initially (step illustrated in c), all fluorescent ROI with physiological signals were retained. However, many of these overlapped with each other. To identify distinct neuron images, ROI were clustered together when their centroids were within 10um of each other and their boundaries were then overlain on the maximum intensity image of the entire time-series (top left, numbers indicate the number of the ROI, green the boundaries of the retained ROI and magenta the boundaries of the discarded ROI). The transients in the waveforms corresponding with each of the clustered ROI (numbers on top right-hand corner above the waveforms) were max projected as described in d and with the aid of the cluster overlay image the ROI that best contained the transients and represented distinct neurons were retained (bottom right). f. Illustration of neurons common to both environments (cyan: boundaries of ROI extracted from the HC time-series, white: boundaries of ROI extracted from the OF time-series). ROI from the two time-series with centroids within a 3μm square grid of each other and less than 4.5μm apart were identified and their boundaries superimposed on the maximum intensity projection images of the two series. Pairs that corresponded with cells in both images were considered as common to both.

Supp 4

Supplementary Figure 4. Median event rates in the HC are slightly but not significantly smaller than those in a familiar OF despite the same extent of behavioral inactivity. In a separate set of experiments, a subset (n=4) of naïve mice injected at 7 weeks (n=3) and 10 weeks (n=1) were imaged in the same OF (50 × 50 × 24 cm white poster board arena) on equivalent days to observe the familiarization response in an initially novel environment and compare it to that in the HC, where the familiarization was to the procedure and enclosure. Mice were group housed and were separated prior to imaging on each day, when they were placed individually in a cage on the recovery stand after mounting the camera. Imaging was performed under white LED and room lights during the afternoon. Animals were imaged around the same time (in a 30-minute window) on all the days. Calcium transients were captured at a rate of 30 frames s⁻¹ and behavior at 10–15 frames s⁻¹. Animals were only placed in this environment on the days of data collection. Excepting these all other procedures are identical to the main experiments.

a. Experimental timeline. Closed arrows in row 3 indicate the start of imaging in the environment. Imaging days 2, 3 and 9 are equivalent to −2d, −1d and 5d in the injury study. b. The OF in which animals were placed on all the imaging days. c. Total duration for which animals were quiet or active on the different days of imaging as a fraction of total recording time. The time-course shows familiarization occurs within 3 days. All bar graphs: open circles represent individual animals and data is presented as mean ± sem. All group sizes: n =3 animals (1d & 2d) & n=4 animals (3d & 9d). d. Fraction of time animals spent in the center and periphery of the OF (left) and near each wall (right). In contrast to in the novel OF on equivalent days, mice spent most of the time sitting in the periphery near a wall. e. Total number of active neurons on each day. f, g Frequency histograms of amplitudes (f) and event rates (g). Light grey solid lines represent histograms of individual animals. Solid and dashed colored lines represent the average and standard deviation of the individual animals in the group. Colors for each group are preserved throughout the figure. Extreme right plots in each: Average histograms on the multiple days (colored solid lines in first 4 plots) and bar graphs showing a measure of the distribution used for statistical comparison. h. Fraction of neurons active in each environment in successive minutes of recording on the 4 days of imaging. Data is presented as mean ± sem. i, j. Comparison of behavior (i) and calcium event rates (j) of the uninjured groups (−2d, −1d and sham) in the familiar HC with that of naïve mice in the familiar OF. Note that for our injury model the sham procedure only involves a very brief exposure to isoflurane and head constraint so that animals are essentially naïve.

Supp 3

Supplementary Figure 3. Large fraction of the recorded cells in both environments are active in the first minute of imaging and the dimensionality of the population network is unchanged after mTBI. a. Frequency histograms of event rates of all cells (from both environments) in mice injected with constructs under the synapsin (n=3, 4.75e¹² GC ml⁻¹) and CamKII promoters (n=7; 1.14e¹³ GC ml⁻¹, n=4; 1.14e¹² GC ml⁻¹, n=3) on the two days prior to injury. These results confirmed our histological observation that this titer of the synapsin construct labelled pyramidal cells. Solid lines: average distribution across animals. Dashed lines: standard deviation from the average. b. Amplitudes of individual transients in neurons common to both environments from all mice on −2d (black), −1d (grey), 5d (sham: blue; injured: green). c. Average (solid lines) and standard deviations (dotted lines) of the frequency histograms of event rates in the novel OF in sham (blue) and mTBI (green) animals on 5d after injury. d. Frequency histograms of event amplitudes on the imaging days. Light grey solid lines represent histograms of individual animals. Solid and dashed colored lines represent the average and standard deviation of the individual animals in the group (black: −2d, grey: −1d, blue: sham, green: mTBI). e. Fraction of neurons in each environment that are active in successive minutes of recording on the 3 days of imaging. Data is presented as mean ± sem. f. Normalized eigenvalues (10 largest normalized with the sum of all the eigenvalues) of the covariance matrix of population rate vectors and g. projections of the population rate vectors onto the first two eigenvectors or principal components of sham (blue) and mTBI (green) mice in HC and OF on 5d, showing that the dimensionality of the population was unchanged after mTBI.