Actigraphic evidence of persistent sleep disruption following repetitive mild traumatic brain injury in a gyrencephalic model

Abstract

We studied the effect of multimodal traumatic brain injuries on daily sleep/activity patterns and related histology. Gyrencephalic ferrets wore actigraphs and received military-relevant brain injuries including shockwaves, strong rotational impact, and variable stress, which were evaluated up to 6 months post injury. Sham and Baseline animals exhibited activity patterns occurring in distinct clusters of high activity, interspersed with periods of low activity. In the Injury and Injury + Stress groups, activity clusters diminished and overall activity patterns became significantly more dispersed at 4 weeks post injury with significant sleep fragmentation. Additionally, the Injury + Stress group exhibited a significant decrease in daytime high activity up to 4 months post injury. At 4 weeks post injury, the reactive astrocyte (GFAP) immunoreactivity was significantly greater in both the injury groups compared to Sham, but did not differ at 6 months post injury. The intensity of immunoreactivity of the astrocytic endfeet that surround blood vessels (visualized with aquaporin 4; AQP4), however, differed significantly from Sham at 4 weeks post injury (in both injured groups) and at 6 months (Injury + Stress only). As the distribution of AQP4 plays a key role in the glymphatic system, we suggest that glymphatic disruption occurs in ferrets after the injuries described here.

Disruptive sleep is one of the most reported symptoms by military service members after a traumatic brain injury (TBI) and is more frequently described than by the general population after TBI (Barshikar and Bell 2017; Piantino et al. 2021). Sleep problems also correlate highly with stress; some researchers suggest that if a person presents with sleep impairment, we should look for stress in their life (Ohayon and Shapiro 2000; Good et al. 2020). Depending on the study, up to 84% of those experiencing TBI report sleep difficulties (Mathias and Alvaro 2012). Overall, however, around 50% of TBI patients show elements of sleep impairment (Mathias and Alvaro 2012). Sleep is also associated with many aspects of cognition. Adequate sleep may be necessary for memory formation and consolidation, including strengthening of visual tasks (Stickgold et al. 2000; Mander et al. 2011; Patterson et al. 2019). Disrupted sleep is prevalent after TBI and often characterized by insomnia, increased awakenings, increased sleep fragmentation, and daytime somnolence, therefore clearly represented in overall activity patterns. Sleep disruption frequently exacerbates other issues, such as depression, anxiety, stress, and pain. As adequate sleep is essential for optimal performance and function, it is important that we study this issue in a realistic fashion.

In order to clarify the precise effects of TBI and stress on sleep and subsequent functional abilities, animal studies are essential. As the smallest mammal with a gyrencephalic cerebral cortex, the ferret is an excellent animal to investigate this problem. The ferret brain has several features similar to those of the human brain, including many sulci and gyri, a large amount of white matter, and the hippocampus largely in a ventral position (Schwerin et al. 2017; Schwerin et al. 2018; Schwerin et al. 2021). A gyrencephalic brain with multiple small interfaces will likely incur greater damage than a lissencephalic brain, which may result in distinct sleep disruptions compared to those sustained in a smooth brain (Gupta and Przekwas 2013). To take advantage of these features we used actigraphy to study sleep and activity levels in the ferret after TBI. Although actigraphy does not directly measure sleep, several studies find that this method assesses sleep as well, or better, than other forms of sleep detection, such as polysomnography (Montserrat Sanchez-Ortuno and Edinger 2010; Blackwell et al. 2011a, 2017; Williams et al. 2020). Actigraphy studies used to assess overall activity and sleep have been carried out in animals including swine (Olson et al. 2016), rats (Suzuki et al. 2018), rhesus monkeys (Berro et al. 2018), and lambs (Rurak et al. 2008). This method has also been used to evaluate sleep for many years in people (Blackwell et al. 2011b; Ancoli-Israel et al. 2015).

The glymphatic system clears waste from the brain and consists of channels composed of astrocytes and their processes (Jessen et al. 2015). This system is paravascular and bounded by the endfeet of astrocytes that express aquaporin 4 (AQP4) and surround blood vessels (Jessen et al. 2015). The glymphatics also function most strongly during sleep, while being curbed when awake (Jessen et al. 2015; Kiviniemi et al. 2016). This interpretation of glymphatic function strongly suggests that a major purpose of sleep employs the glymphatic system by clearing toxic waste from the brain. The glymphatic waste clearance process is disrupted after TBI, most likely due to the injury resulting in scarring and inflammation of astrocytic processes and the AQP4 expressing end feet (Iliff et al. 2014; Kress et al. 2014; Thrane et al. 2015; Shively et al. 2016; Schwerin et al. 2021). The glymphatic pathway that relies on astrocytes is an essential component of normal sleep and operates in part through AQ4P+ astrocytic endfeet surrounding blood vessels (Jessen et al. 2015).

In order to study a situation closely related to the experiences of military personnel in theater, we exposed ferrets to a TBI that had two distinct mechanistic components (i.e. multimodal) as well as an element of stress. We evaluated the changes in activity/sleep patterns along with paravascular astrocytic changes in ferrets after experiencing a repetitive multimodal TBI. Since adequate sleep is crucial, it is important to study this issue using realistic scenarios by employing TBI mechanisms similar to those experienced by service members. Although blast-induced mild TBI has been referred to as the signature injury of recent military conflicts, it usually occurs in the presence of stress and often coincides with other types of brain injuries (e.g. head rotation and/or impact with an object or the ground) (Kwon et al. 2011; MacDonald et al. 2014). In addition, many military personnel report more than one brain injury with an increase in the number of TBIs correlated with sleep difficulties (Vanderploeg et al. 2012; Miller et al. 2013; Yee et al. 2017; Merritt et al. 2020). In order to align our experimental model with clinical presentations, we used three consecutive cohorts with an increasingly greater number of brain injuries using both the explosive blast injury as well as the Closed Head Impact Model of Engineered Rotational Acceleration (CHIMERA) injury (Namjoshi et al. 2014) with a subset of animals in each cohort also receiving stress prior to and throughout the brain injury time period.

Materials and methods

Animals and housing

All experiments were designed to minimize the number of animals used and were approved by the Uniformed Services University Animal Care and Use Committee and conducted in accordance with the NIH Guide for Care and Use of Laboratory Animals. Seventy-one 7–20 month old male ferrets (Mustela putorius furo, Marshall Bioresources, NY, USA) weighing 1.08–2.03 kg were used in these experiments. Ferrets were group-housed (2–3 per cage unless single-housed as necessary due to aggression) under a 12-hour light/dark cycle (lights on: 7 a.m. –7 p.m.) at 61–72°F and 30–70% humidity, with food and water ad libitum. Cage enrichment consisted of a cloth hammock and a toy.

All experiments were carried out blinded; ferrets were randomly divided into three groups: Sham, Injury alone, or Injury + Stress. Animals were tested consecutively in three cohorts that varied in age and experienced slightly different injury/stress parameters (population age range 33–90 weeks: Cohort 1 (24 ferrets): 33–36 weeks, Cohort 2 (23 ferrets): 47–52 weeks, Cohort 3 (24 ferrets): 75–90 weeks). Cagemates were of the same Cohort but participated in various experimental groups (Sham, Injury, Injury + Stress) resulting from random assignment.

Animals were part of a larger behavioral study not included here, as such they were trained and tested for these behaviors prior to and after the stress/injury intervention discussed below. The ferrets were comfortable interacting with the investigator conducting the behavioral testing who was not involved in the stress application or the injuries. The behavioral tests and baseline data from the animals in this study can be found in this normative ferret behavior article (Obasa et al. 2022). Actigraphy was not evaluated during time periods where the animal was involved in any training, testing, or after any other procedure or activity expected to affect the actigraphic pattern.

Timeline

The cohorts differed by the number of injuries and duration of stress (Fig. 1A). Cohort 1 animals received one CHIMERA and three Blast injuries over 5 days. A subset of animals received stress, which occurred for a total of 14 days (prior to and during the injuries). Only one injury type occurred per day. Cohort 2 animals received two CHIMERA and five Blast injuries over 13 days, which included 1 day where the blast was immediately followed by the CHIMERA injury. The stress was administered to a subset of animals for 21 days prior to and during the injuries. Cohort 3 animals received two CHIMERA and seven blast injuries over 10 days, including three back-to-back injuries—two with the blast followed by another blast and one with the blast followed by a CHIMERA injury. A subset of animals received stress for 21 days prior to and during the injuries. Cohorts 1 and 2 animals were terminated at 4 weeks post the last day of the injuries. Cohort 3 animals were terminated at 6 months post the last day of the injuries. Curves of the shock waves delivered by the blast injury and the velocity and path of movement delivered by the CHIMERA can be seen in Fig. 1(B–D).

Chronic variable stress

We applied intermittent unpredictable stress to a subset of each cohort prior to and during the period of brain injury for a total of 14–21 days. Each animal received two to three sessions per daytime (7 a.m.–7 p.m.)—roughly morning, afternoon, or evening and an occasional session overnight (7 p.m.–7 a.m.). The stressors were either applied individually or combined within a session. The duration and start time of each session was varied. Using the 21-day stress group as an example, there were 63 possible stress sessions (3 × 21 days, not including overnight). In eight of those sessions, we did not apply a stressor. In 22 of the sessions, we applied a single stressor. Each stressor was applied 12–15 times in a pseudo-randomized fashion. We applied stress overnight five times spread out across the stress period. Stress was applied to individual animals in a room separate from the housing room, in four dedicated cages. Water was available ad libitum, but food was withheld during the stress session unless it was an overnight session.

The types of stress we used were designed to mimic real life events that appear as disruptors or hassles. This form of stress, whether combined with trauma or not, are often found to be predictive of poor outcomes after natural disasters or wartime conflicts; this is true whether civilians or servicemembers are involved (Galea et al. 2007; Dickstein et al. 2010; Heron et al. 2013). Our stressors were chosen to realistically interrupt the daily life of the animals involved. They made it difficult to walk around in the cage (wet floor, uneven floor) or difficult to sleep (strobe light, sound/vibration, moving floor, lights on overnight) or to freely move (restraint).

Stressors included:

Wet floor: water-soaked towels covered the floor of the cage. Duration varied between 45 minutes and 6 hours, or overnight.

Uneven floor: 36 hard plastic balls of three different sizes (4 large: 10 inches in diameter, 12 medium: 8 inches, 20 small: 4 inches) covered the floor. The duration varied between 45 minutes and 3 hours, and could occur twice in a session.

Strobe light(s): one strobe light (108 LEDs, stroboscopic speed: 6 flashes/second, 454 lux, Roxant, St.Peters, MO) was placed on the ceiling of the cage and one on the front of the cage. The duration varied between 45 minutes and 6 hours, or overnight.

Sound/vibration: an alarm clock (Sonic Bomb by Sonic Alert, Troy, MI) was placed on the cage ceiling and an attached vibrating unit was placed under the cage. Each alarm clock had two alarms set to varying durations of 1–5 minutes (sound level = 113 dB). Four cages were used at a time, so during the stress session, eight alarms went off, typically individually but, occasionally, two were on simultaneously. The vibrating unit jumps and rattles creating additional noise. The duration (time in which the eight alarms could go off) varied between 45 minutes and 6 hours, or overnight.

Moving floor: we placed a large rotating disk on the floor of the cage and surrounded it with metal mesh to enclose it. We also installed a floor to ceiling wall from the edge of the disk to the center. The wall had protrusions so that if the animal rode the rotating disk, it would encounter the protrusions, which required the ferret to stand and move away from the protrusion. Essentially, animals cannot fall asleep for the duration of this stressor. The duration was 45 minutes to 3 hours (or overnight) and could be conducted twice in a session.

Lights on overnight: animals were kept in the stress cage overnight with the lights on (single-housed, food and water ad libitum, sometimes combined with another stressor).

Restraint: we used a clear acrylic ferret restrainer (PLAS-LABS, Inc., Lansing, MI) for a maximum of 40 minutes and could occur twice in a session.

Blast shock wave exposure

Ferrets were exposed to a shock wave as previously described (Vu et al. 2018; Schwerin et al. 2021). Animals received isoflurane (2–5% in oxygen) anesthesia for 10 minutes in an induction box placed on a heating pad set to low. When insensitive to pain, each animal was removed from anesthesia and positioned prone in a mesh hammock in the center of the shock tube with the head facing forward and exposed. The shock wave was then delivered using a custom-built Advanced Blast Simulator (ABS, Stumptown Research & Development, Black Mountain, NC). In this study, a VALMEXⓇ FR 1400 Type IV 7270 membrane (Mehler Texnologies GmbH, Huckelhoven, Germany) and compressed air simulate the blast shock wave. The mean positive peak pressure of blast waves was 21.3 ± 0.8 psi (mean ± standard deviation) and coefficient of variance (COV) was 3.8%. The mean negative peak pressure was 4.25 ± 0.23 psi (COV: 5.4%). The duration of the positive phase was 7.3 ± 0.3 ms, the mean positive impulse 0.065 ± 0.002 psi × second, and the speed 510.3 ± 6.8 m/second (See Fig. 1B). Following the blast, we transferred the animal to a heating pad to recover. Once ambulatory, the animal was returned to their home cage. Sham animals received identical treatments but were placed prone on a table immediately next to the ABS during a shock wave. For the combined blast/blast injury, the ferret was returned to the induction box for another 10 minutes of anesthesia (2–5%) between blast exposures.

CHIMERA procedure

CHIMERA procedures were performed as previously described (mice: Namjoshi et al. 2014, ferret: Whyte et al. 2019). On the day of injury, the ferret was anesthetized in a clear induction box using inhaled isoflurane (5% induction, 1–3% maintenance), given a subcutaneous injection of 10-mL warm saline and maintained via nosecone, while positioned on the CHIMERA device. The animal was placed supine on a plastic bag filled with rice and adjusted so that the top of the head could be placed flat over the piston opening approximately at the anterior aspect of the ears along the midline. This position targets the posterior portion of the frontal lobes. The head plate was angled 150° relative to the body plate and positioned −5° from horizontal. The body was secured with hook and loop straps to the body plate under the front legs and across the pelvis, leaving the head, neck, and upper body free to rotate in the sagittal plane. The impactor is pneumatically driven by a compressor. The impactor weighs 200 g and its tip is 12.7 mm in diameter and 31.7 mm long including a 6.35-mm-thick rubber cap. The nose cone was removed for impact. Instrument pressure was set at 34.8–45.9 psi to obtain a piston speed of 12.406–15.614 m/second, which corresponds to an energy level of 15.39–24.38 J. The actual measured speed for all impacts had an average of 13.456 m/second, resulting in calculated energy of 18.28 J with a 20.2% COV. Following the impact, the ferret was given a subcutaneous injection of 0.01–0.03 mg/kg buprenorphine while recovering on a heating pad. Ferrets were monitored and returned to their home cages once ambulatory. Sham animals underwent all procedures, including anesthesia, analgesia, and positioning on a table next to the device while it was triggered, but no impact delivered.

Combined blast/CHIMERA injury

For combined injuries, the blast was conducted first and then the animal was immediately transferred to the CHIMERA device with anesthesia maintained via nosecone as the CHIMERA procedure continued as described above. The saline was injected prior to the blast.

Actigraphy

All animals wore a wGT3X-BT (ActiGraph, Pensacola, FL) activity monitor on a woven nylon wristband (ActiGraph, Pensacola, FL) around their neck. Animals acclimated to wearing the monitor for at least 7 days prior to recording.

Sample selection

Animals from three separate cohorts were represented in the analysis. For two successive 24-hour periods (i.e. Day 1, Day 2), we collected actigraphy data starting at 12:00 a.m. and ending at 11:59 p.m. Given potential variability related to handling, we restricted the sample data for analysis based on Day 2 only. In Cohort 1, Day 1 and Day 2 data were collected for some animals over Wednesday–Thursday and others over Saturday–Sunday. In Cohorts 2 and 3, data we collected only on Saturday–Sunday. as laboratory activity was greater during the weekdays, to minimize external factors influencing the animals’ resting and activity patterns, we restricted data analysis to Sunday only. For daytime analysis, data were analyzed starting at 7:00 a.m. and ending at 6:59 p.m. on Sunday. For nighttime analysis, we analyzed data from 7:00 p.m. to 6:59 a.m. for both Saturday night and Sunday night and then divided by 2 (due to limitations of analysis software described below). We scored specific time-points over the course of the experiment: Baseline, 4 weeks post injury for all cohorts and 4 months and 6 months post injury for Cohort 3.

We used ActiLife v6.13.4 software (ActiGraph, Pensacola, FL) to analyze the data collected from the actigraphs. ActiGraph uses an accelerometer and a proprietary filtering procedure to generate counts that vary based on the frequency and intensity of the raw acceleration (see https://actigraphcorp.my.site.com/support/s/article/What-are-counts). ActiGraph offers acceleration outputs (counts) in each of the three planes of movement as well as the combined three-dimensional vector magnitude |$\sqrt{{\left( Axis\ 1\right)}^2+{\left( Axis\ 2\right)}^2+{\left( Axis\ 3\right)}^2}$|⁠. Our activity results are based on the vector magnitude counts over a 60-second epoch. We defined sedentary activity (0–100 counts) and high activity (2001 and above counts with no ceiling), see Supplemental Material and below. Non-wear time was identified and excluded using Wear Time Validation using the Troiano technique (Troiano et al. 2008) with manual review. Counts are based on 60 second epochs. We defined sedentary activity (0–100 counts) and high activity (2001 and above counts, see Supplemental Material and below).

Activity level percentage

The percent of time spent highly active (>2,000 counts) was averaged across animals within experimental groups for daytime and nighttime periods.

Bout analysis to calculate sleep fragmentation index

Disrupted sleep is quantified in many ways across sleep studies (Shrivastava et al. 2014). One such measure is sleep fragmentation; however, this often has many mathematical definitions depending on the study. In general, sleep fragmentation typically refers to brief arousals that occur during a sleep period. When assessed with actigraphy, sleep fragmentation may refer to the amount of movement or restlessness in a sleep period. Here, we used a similar mathematical definition described by Fekedulegn et al. for the sleep fragmentation index, which refers to the ratio of the number of awakenings to the total sleep time in minutes (Fekedulegn et al. 2020). We did, however, modify the equation to work with the ferret’s sleeping habits. The ferret sleeps 60–75% of the 24-hour period, with frequent awakenings. To more accurately reflect the changes we observed, we further defined awakenings as sustained high activity requiring a minimum of 10 consecutive minutes where each minute had greater than 2,000 counts, with a permissible drop time of 2 minutes (meaning up to two individual minutes could have a count less than 2,000). We chose 10 minutes to specifically capture the sustained high activity observed in the healthy ferret. The total sleep time was further defined as sedentary time requiring a bout length minimum of 3 minutes and a maximum count level of 100 counts per minute. We chose a short time requirement for sedentary behavior because we observed that healthy ferrets exhibited frequent, short sleep periods, but we required three consecutive minutes to avoid overestimating sedentary behavior. The sleep fragmentation index was then calculated as the total number of sustained high activity bouts divided by the total time (in minutes) spent in sedentary bouts over a 24-hour time period (daytime + nighttime); the closer to zero, the more fragmented the sleep:

Statistical Analysis (% Highly Active and Sleep Fragmentation Index)

All three experimental groups (Sham, Injury, Injury + Stress) and all time-points (Baseline, 4 weeks, 4 months, and 6 months post injury) were analyzed. Only Cohort 3 continued through the 4 and 6 months post injury time-points. Repeated measures ANOVA cannot handle missing values so we analyzed the data by fitting a mixed-effects model with repeated measures followed by post hoc comparisons within experimental group as implemented in GraphPad Prism 9.4.1. This mixed model uses a compound symmetry covariance matrix and fit using Restricted Maximum Likelihood (REML). We used the Geisser–Greenhouse correction, with repeated measures of time, followed by multiple comparisons test within experimental group (either the Tukey or Fisher’s LSD). A significance level of 0.05 was applied for each test.

LOESS analysis

We examined the Actigraphy data to assess activity levels and potential differences in their distribution among ferret groups. Non-parametric methods, using locally estimated scatterplot smoothing (LOESS), were applied to each animal’s activity count. Briefly, LOESS represents a smoothing function where a regression line is fit at each data point. At each of these points, adjacent data points in the vicinity of these points are included in the regression. A smoothing curve is generated that compiles the locally fit regression lines across the data. A span parameter controls the adjacent data points that contribute to each locally fit regression and determines the level of smoothness of the resultant curve. LOESS was applied with the current data using (i) activity counts > 0, based on the assumption that only the non-zero data varied across animals, and (ii) the same span for each individual ferret’s data. As part of LOESS, a 95% confidence interval (band) was calculated. Intervals at which the 95% lower confidence limit exceeded 2,000 counts over the 24-hr period for each ferret were identified and a threshold of 2,000 counts was defined as “high sustained activity” (see Supplemental Material). Two metrics were calculated based on these “sustained high activity” intervals: (i) a summation of the LOESS-derived fitted counts > 2,000 and (ii) a summation of minutes spent in these intervals over the 24-hour period that the actigraphy data were recorded for each ferret (fitted counts).

Statistical analysis (LOESS)

Measures of sustained high activity over a 24-hour period using LOESS were compared across Sham, Injury, and Injury + Stress groups at the time-points indicated above (Baseline, 4 weeks post injury, 4 months post injury, and 6 months post injury). A Baseline comparison between the three groups was conducted prior to stress application and/or brain injury. At 4 weeks post injury, comparisons were again conducted for the three experimental groups. With Cohort 3 data only, at 4 and 6 months post injury, given that only this cohort survived past 4 weeks post injury, comparisons were conducted for each experimental group. Boxplots depict the distribution of activity among the groups. Outlier values were plotted for viewing purposes but were not included in the analysis given they do not comprise the bulk of the distribution of each respective group’s data. Because the skewness of the data and lack of homogeneity of variance for the different groups, a Kruskal–Wallis test was applied to assess group differences at the different time-points. A significance level of 0.05 was applied for each test. We used an ANCOVA analysis to assess differences between cohorts at the 4-week time point.

Assessment of sustained high activity threshold

Based on animals’ activity counts, we sought to determine a threshold where counts exceeding that level would represent sustained high activity. In contrast to a short burst of activity, high sustained activity represents a cluster—or engagement in activity by an individual animal—over a period of time (see Supplementary Fig. 1). We examined data from animals in each cohort at baseline who represent a healthy, normal ferret group. Our analysis to establish a threshold occurred in two parts: (i) we applied the LOESS fits to assess animals’ activity patterns; ii. we evaluated different threshold levels in terms of activity counts (i.e. counts = 1,000, 1,500, 2,000, 2,500, 3,000). We then compared the fitted LOESS values for each animal’s data with binary measures reflective of each animal’s actual activity counts (e.g. activity counts > 2,000 vs ≤ 2,000). Predictive measures were obtained based on this comparison in 49 animals (i.e. sensitivity (SEN), specificity (SPE), positive predictive value (PPV), negative predictive value (NPV), accuracy (ACC)). Two additional measures were derived: sensitivity + specificity − 1 and sensitivity + 2 × specificity. (i). We evaluated these measures with respect to the five different thresholds and, based on an optimal set of these measures, selected a threshold of 2,000 (see Supplementary Fig. 2A–G). For example, in terms of evaluating SEN and SPE (Supplementary Fig. 2A and B, respectively), a count threshold of 1,000 was optimal for SEN, whereas a count threshold of 3,000 was optimal for SPE. In order to optimize SEN and SPE jointly, we selected 1,000 < count threshold < 3,000. Additionally, the research question of assessing individual ferret’s sustained high activity informed the threshold assessment. First, a lower threshold (e.g. 1,000) would record shorter durations (bursts) of activity, and this would be represented by greater SEN (e.g. Supplementary Fig. 1 activity recorded between 2:00 and 3:00 hours:minutes), but the activity recorded would not reflect sustained high activity. A lower threshold (see Supplementary Fig. 1, horizontal dashed line activity count = 1,000) would include activity counts near zero as high sustained activity, which would represent lower SPE. By contrast, selection of a higher threshold (e.g. 2,000), would preclude the presence of activity counts near zero, represent higher SPE, and would reflect greater consistency with sustained high activity (e.g. Supplementary Fig. 1 activity recorded between 7:00 and 9:00 hours:minutes). In addition to examining other metrics (PPV, NPV, ACC) (Supplementary Fig. 2C, D, and G), which indicated increased optimal levels given higher thresholds (e.g. > 1,000), we also assessed thresholds with respect to the two derived measures mentioned above. In the case of sensitivity + specificity − 1, the SEN and SPE are weighted equally, for which a count threshold = 1,500 appears optimal (See Supplementary Fig. 2E). However, as stated earlier, we wished to optimize SPE relative to SEN given our research question, therefore, we derived another measure (i.e. sensitivity + 2 × specificity −1) to weight SPE more relative to SEN. Based on this derived measure, a count threshold = 2,000 appears optimal for defining sustained high activity in the ferrets (see Supplementary Fig. 2F).

Immunohistochemistry

A subset of animals at 4 weeks post injury and 6 months post injury were evaluated immunohistochemically. Animals were deeply anesthetized with Euthasol (0.22 mL/kg) and transcardially perfused with 1-L ice-cold 0.1-M phosphate-buffered saline (PBS) (pH 7.4) containing 53.1 mg of heparin (Sigma-Aldrich, St. Louis, MO) followed by 1 L of ice-cold 4% paraformaldehyde solution in PBS (Santa Cruz Biotechnology, Santa Cruz, CA). The brains were post-fixed in the same solution for 7 days and then transferred to a storage solution of PBS mixed with 0.04% sodium azide. Coronal sections (50 mm) were cut using a vibratome (Leica VT1000; Leica), and immunohistochemistry carried out as described previously (Schwerin et al. 2017; Schwerin et al. 2018). Sections selected from the medial frontal cortex (FCM, i.e. representing the anterior cingulate gyrus) were double labeled with chicken anti-glial fibrillary acidic protein (GFAP) (1:500; Abcam Cat# ab4674, RRID: AB_304558) and rabbit anti-aquaporin 4 (AQP4) antibody (1:1,000; Abcam Cat# ab128906, RRID: AB_11143780) using antigen retrieval (1x citrate buffer (Thermo Scientific, Waltham, MA) in an 80°C water bath for 20 min). Immunoreactivity was revealed by incubation with either AlexaFluor 555- or 488-conjugated secondary antibody (1:500) and nuclei counterstained with 40,6-diamidino-2-phenylindole (DAPI) (Invitrogen, Waltham, MA: D21490, 1:2,000). Sections were mounted with Mowiol 4-88 medium (Polysciences, Warrington, PA).

Imaging and analysis

We acquired optical image fluorescent stacks at 10× magnification using a Zeiss Axio Observer.Z1 microscope with an Apotome using Zen 2012 software (Blue edition) version 1.1.2.0 (Carl Zeiss Microscopy). To quantify aspects of the immunoreactivity, measurements from 5 to 10 coronal sections were averaged across three to six animals per group. We were especially interested in visualizing astrocytes and used GFAP and aquaporin 4 (AQP4) to label reactive astrocytes (GFAP) and their endfeet (AQP4). We assessed the fluorescence intensity and area of immunoreactivity by adapting a procedure in Fiji (http://imagej.net) to measure the total area of immunoreactivity in histologic samples (adapted from https://www.youtube.com/ watch?v=nLfVSWcxMKw&list=TLPQMDUwNjIwMjAhHiT0TuG0X A&index=4). We used the medial part of the frontal cortex, as a neocortical region known to be affected for similar types of injuries (e.g. Schwerin et al. 2021). We obtained images using the same exposure settings for each wavelength across all samples; red: 400 ms, green: 60 ms, blue: 15 ms, to avoid fully saturated pixels from skewing the analysis. Images were taken using the tile and stitching (fuse) filter of 9-tiles (Zen Software), saved as TIF images and exported to Fiji software. Within Fiji, we used pixels “scale set”—with pixels as a scaling measure, “color threshold adjustment”, “brightness calibrator”, and the “analyze program” to determine the “percent area labeled” and “fluorescence intensity.” These measurements were averaged and compared between groups via a one-way ANOVA and Tukey multiple comparisons.

Results

Ferrets are naturally crepuscular, being more active at dawn and dusk, but otherwise sleep on and off for the majority of a 24-hour time period. Ferrets are known to adapt to the environment, however, and in the laboratory, this results in more awake time during the day than during the night (see Fig. 2, examples A, B, C). When they are awake, ferrets typically exhibit high activity for sustained time periods (> 10 minutes).

Baseline activity levels

At baseline, we observed sleep/activity patterns where ferrets show distinct bouts of high levels of activity that persist for at least 10 minutes, interspersed with periods of low level activity. Figure 2(A–C) shows examples of the clustered patterns of increased activity in three different ferrets at Baseline (i.e. before any injuries). Plots of the same data with a fitted LOESS (95% CI) appear in Fig. 2, examples (D), (E), (F). The LOESS plots show the same distribution of individuals’ activity data as seen in the actigraph plots. In these plots, animals’ intervals of sustained high activity are represented by areas where the lower 95% confidence limit of the LOESS exceeds 2,000 activity counts.

Sleep/activity patterns are altered after repetitive, multimodal TBI/stress

Although as indicated above, actigraphs do not measure sleep per se, they have been used to measure sleep-like patterns in many species and show a strong correlation with sleep (Williams et al. 2020). After delivery of the combined injury (blast + CHIMERA), we assessed each for activity patterns at 4 weeks after the last day of the delivery of the Injury or Injury + Stress. As the three Cohorts received slightly different injuries, we assessed whether each cohort demonstrated the same overall level of activity at the 4-week timepoint. Supplementary Fig. 3 shows that each cohort (Sham, Injury and Injury + Stress) presented similar activity levels at the 4-week time point. An ANCOVA test revealed no significant differences across cohorts. Figure 3 shows that the distinct clustering of maintained high bouts of activity seen clearly in the Sham animals (Fig. 3A–C), begins to break down and become more dispersed in the Injury animals (e.g. Fig. 3D–F). Although the clustered pattern can still be observed in the Injury animals, the activity is less sustained and scattered increases in activity counts appear. For the animals subjected to Injury + Stress (Fig. 3G–I), the clustered activity is less prominent and the overall distribution appears substantially more fragmented.

The fitted LOESS plots in Fig. 4 represent the same individuals shown in Fig. 3. These plots indicate the similar patterns of clustered activity, more apparent in the Sham animals (when the 95% confidence interval exceeds 2,000 counts) that become more dispersed and fragmented with the Injury animals and even more so in the Injury + Stress animals.

As indicated in the Methods (Fig. 1), Cohorts 1 and 2 survived until 4 weeks after the injury. Cohort 3 survived until 6 months post injury. We examined the activity/sleep patterns of the animals in Cohort 3 at 4 weeks but also at 4 and 6 months post injury. At 4 months post injury (Fig. 5), the actigraph patterns (Fig. 5A–B) in the Sham animals were similar to those seen at Baseline and at 4 weeks post Injury show clusters of sustained high activity that persisted for at least 10 min and were interspersed with lower levels of activity below 2,000 counts. For the animals that received an Injury (Fig. 5C–D), the clustered distribution was not obvious and scattered activity patterns appeared with less extensive prolonged visible activity. The Injury + Stress animals displayed even less structured activity patterns compared with the Shams. In the LOESS plots of the same individuals at 4 months post injury (Fig. 5A’–F’), the clustering or scattered pattern of activity is similarly captured for the Injury and Injury + Stress groups. Additionally, based on these fitted plots, differences in the activity patterns (i.e. sustained high activity) among the individual ferrets can be examined quantitatively dependent on the lower 95% confidence limit of the LOESS, as indicated previously. Figure 6 reveals a similar depiction of animals at 6 months post injury, in terms of actigraphy patterns (Fig. 6A–F) and corresponding LOESS plots (Fig. 6A’–F’).

Quantification of patterns

Locally estimated scatterplot smoothing (LOESS)

Figure 7 shows a comparison of separate measures of sustained high activity (Minutes of High Activity—Fig. 7A–D and Fitted Counts of Activity—Fig. 7E–H), for the injury groups at different timepoints. The measures were aggregated across individual animals from the different groups to make the comparisons. Statistical tests indicated no significant differences appeared between the groups (Sham, Injury, Injury + Stress) at Baseline, but at 4 weeks post injury, a Kruskall Wallace H test showed significant differences across the groups of animals. Despite the distinctions in the individual patterns seen in Figs. 5 and 6, at 4 and 6 months post injury, this analysis does not indicate significant differences at these later time points based on the aggregated (group) data.

We used different analyses and also assessed if changes occurred in the amount of time that the ferrets spent active during the day after the Injury or Injury + Stress. Figure 8 shows that a mixed effects analysis finds a significant difference between the three groups of animals (Sham, Injury, and Injury + Stress) over the time points we measured (Baseline, 4 weeks post injury, 4 months post injury, and 6 months post injury) for both group and time, suggesting that both the Injury and Injury + Stress groups show significant differences from the Sham. Pairwise comparisons revealed that significant differences occurred in the Injury + Stress groups at 4 weeks and 4 months post injury with Baseline values. This strongly suggests that the groups of animals receiving added stress showed stronger changes in the amount of activity up to 4 months post injury, which were not seen in the Injury only ferrets.

Fragmented sleep

We further determined if the animals receiving an Injury or Injury + Stress showed changes in fragmented sleep by assessing the Sleep Fragmentation Index (SFX) as described in the Methods. The analysis conceptually represents the effect of brain injury to fragment sustained activity and sleep. We determined if the number of high activity bouts divided by the total sedentary bout length differed among groups to obtain a measure of fragmented sleep. Again, a mixed-effects analysis revealed that this effect across all groups was significant (P < 0.0049). Comparison using a post hoc pairwise test revealed that both the Injury and Injury + Stress groups at 4 weeks post injury showed a more fractured pattern of sleep (Fig. 9).

TBI/stress increases astrocyte and aquaporin immunoreactivity

TBI often results in increased astrocytic reactivity, frequently revealed with increases in GFAP. The higher levels of reactive astrocytes often locate around blood vessels and the astrocytic endfeet are visualized using AQP4 immunoreactivity. The arrangement of GFAP reactivity combined with the AQP4 endfeet form a paravascular structure that helps to clear the brain of waste products. This glymphatic system is an important method of transporting toxins from brain tissue and also seems to operate primarily during sleep. As our model shows impaired activity associated with sleep, we assessed the distributions of astrocytic reactivity using GFAP and AQP4 in the same animals studied here with actigraphy.

Figure 10 shows the immunoreactivity for the ferret medial prefrontal cortex (including the anterior cingulate gyrus) in the same sections immunoreacted for GFAP and AQP4 and the merged image of the two markers. Shown are examples for Sham animals as well as the other groups involved in our study (4 weeks post injury, 4 weeks post injury and stress, 6 months post injury, 6 months post injury and stress). The sections were quantified as described in the Methods for both intensity and the percent area labeled. Figure 10 presents the same images at a higher power, to better visualize the differences in fluorescent intensity and area labeled. Both the intensity and area of label were consistently higher relative to the Sham values, but not always significantly (Fig. 11). At 4 weeks post injury, the intensity and percent labeled areas were significantly greater than the levels seen in the Sham animals. At 6 months, we only observed significant increases in the intensity of AQP4, not the percent area labeled, in relation to the Sham (Fig. 11). The percent area labeled, however, at 6 months post injury for the Injury + Stress group had a P-value of 0.0636, close to significance. These immunoreactivity values correspond with the other comparisons reported here for the actigraphy/sleep values, where we most often saw significant differences at the 4-week post injury time point compared to Shams, but also consistently observed differences at longer time points.

Discussion

Ferrets received a combination of blast and CHIMERA injuries (Fig. 1), with or without stress, and survived for 4 weeks or 6 months. During this time, they wore actigraphs and we reported on their activity patterns, which correlate highly with sleep, at 4 weeks, 4 months, and 6 months post injury. The animals in each group (Sham, Injury, and Injury + Stress) showed no differences at baseline, but revealed significant disruption in activity patterns at 4 weeks, as measured with a LOESS analysis. The patterns at 4 and 6 months showed no differences from the Sham, using LOESS. Assessing levels of daytime activity, however, indicated an overall significant difference between the groups at all survival time points (using a Mixed Effects Analysis) and specific pairwise significant decreases between the Sham and the Injury + Stress group at 4 weeks and 4 months post injury. Evaluation of the sleep fragmentation index also revealed an overall significant difference between the groups at all timepoints using a Mixed Effects Analysis, and pairwise differences between Baseline and 4 weeks post injury for both the Injury and Injury + Stress. We also assessed immunoreactivity of GFAP and AQP4 in the Sham, Injury, and Injury + Stress groups at 4 weeks and 6 months post injury, showing increases in both these markers at 4 weeks post injury, but only in the Injury + Stress group at 6 months post injury. Our study reveals that activity patterns (interpreted as closely related to sleep) in the gyrencephalic ferret are disrupted after a combination of blast and CHIMERA injuries. The disruption persists in different forms up to 6 months post injury and is overall slightly greater in the animals receiving stress. The use of this gyrencephalic animal, and the findings of our study, represent a potentially highly useful translational model for investigating the critical clinical consequence of military and civilian TBI.

Actigraphy

Although using actigraphs is not identical to measuring sleep, a number of researchers agree that actigraphy provides highly valuable information when assessing sleep, such as the ability to capture extended periods of “sleep” in a home environment and the ability to evaluate this behavior over several days (Ancoli-Israel et al. 2015). Actigraphy also measures other patterns related to sleep such as daytime sleepiness and sleep duration (Ancoli-Israel et al. 2015). Several studies validate that actigraphy can be used to assess sleep/wake problems and insomnia (Sadeh et al. 1994; Ancoli-Israel et al. 2003; Morgenthaler et al. 2007; Montserrat Sanchez-Ortuno and Edinger 2010; Ancoli-Israel et al. 2015). Actigraphy also corresponds well with other measures of sleep, such as polysomnography (Blackwell et al. 2011b; Williams et al. 2020). All in all, we are using actigraphs to measure the overall activity levels of ferrets in several conditions involving injury, with or without stress, and find that the patterns and levels of activity change after TBI.

Sleep/activity after TBI is disrupted

Sleep disruption after TBI has been described for many years and is an unfortunate frequent occurence for those involved in battle (Mathias and Alvaro 2012; Piantino et al. 2019; Piantino et al. 2021). Post TBI studies indicate that sleep disruption shows a number of different features including insomnia and interrupted sleep patterns (Bell et al. 2018). Veterans also have serious and persistent sleep problems, even with relatively mild TBIs, which contribute to increasing functional problems and poor recovery (Gilbert et al. 2015). Other studies show that sleep disorders evolve after TBI and tend to diminish with time (Paredes et al. 2021). Excessive daytime drowsiness is often a problem along with insomnia (Paredes et al. 2021). Many of these features correspond with our observations in ferrets. We show that high activity levels decrease after injury, both generally and in the daytime (Figs. 7–9), suggesting they may struggle to function normally. In addition, people with TBI and sleep disruption tend to improve after a period of 4 weeks post injury, as do the ferrets, using some measures of evaluation (Paredes et al. 2021).

Distractibility is another issue after TBI. Several reports indicate that after a TBI, individuals are easily diverted and disorganized in attention to tasks requiring focus (Chen and Loya 2014; Neyens et al. 2015). The distribution of sustained high activity levels after injury in ferrets becomes less organized and less clustered as seen in Figs. 3–6, which may suggest loss of the ability to focus after Injury or Injury plus Stress compared with Baseline or Shams. The loss of cohesive activity stretches appear even greater in the stressed animals. Not all of the analyses we carried out, however, showed a higher level of quantitative difference between the Injury and Injury + Stress groups. The data seen in Fig. 8, however, show a significant decrease for the daytime hours spent highly active in the animals that received stress, again implying that the ability to focus, or to be easily diverted from a task, is an outcome in ferrets after the types of injuries described here, and especially after Injury + Stress.

Astrocytes and the glymphatic system play a role in sleep

Interestingly, astrocytes have been recognized for a number of years to be important in sleep regulation; our findings of increased and persistent astrogliosis may play a role in the disruption of activity patterns shown here (Halassa et al. 2009; Clasadonte et al. 2017; Vanderheyden et al. 2018). Astrocytes use gliotransmission to release a number of neurotransmitters that play a role in sleep regulation. Halassa et al. (2009) found that manipulating gliotransmission in astrocytes disrupted sleep homeostasis, emphasizing their importance. The astrocytic release of neurotransmitters facilitates or inhibits neuronal oscillations prevalent in sleep (Fellin et al. 2009; Poskanzer and Yuste 2016). Vaidyanathan et al. specifically reported that astrocytes control non REM sleep by regulating two types of G protein coupled receptors (Vaidyanathan et al. 2021). A recent article suggests that a link between astrocytes and neurons become disrupted in Alzheimer‘s Disease, which may share some of the pathology seen after TBI, when an uncoupling occurs between astrocytes and the metabolic demands of neurons (Vanderheyden et al. 2018).

The glymphatic system comprises another important feature involving astrocytes and sleep. The glymphatics consist of a paravascular system that participates in eliminating waste from the parenchyma of the brain. Astrocytes surround blood vessels and use the paravascular glymphatic system, via astrocytic endfeet, to clear the brain of waste substances. This process most likely occurs during sleep (Iliff et al. 2012). The astrocytic endfeet are immunoreactive for AQP4, which as a water channel positioned in a unique interface between astrocytes and blood vessels, provides an excellent path for waste removal (Salman et al. 2022). Our findings show an increase of both astrocytic immunoreactivity (via GFAP) and their endfeet (AQP4 immunoreactivity) after an injury or injury plus stress (Figs. 10 and 11). The increase in astrocytic reactivity is expected as a typical response to brain injury (e.g. Schwerin et al. 2021 and previous reports also show increases in AQP4 after TBI; Dadgostar et al. 2022). AQP4 increases correlate with edema and poor function as measured in behavioral tests in mice (Lopez-Rodriguez et al. 2015; Dadgostar et al. 2022); AQP4 decreases also link with negative effects (Dadgostar et al. 2022). In our study, the AQP4 augmentations seem to persist for a longer time than the GFAP increases, especially in the animals receiving stress (Fig. 11). Together, the GFAP and AQP4 changes are a likely to be a key cause underlying poor sleep.

The addition of stress results in slightly increased activity disruptions

Many animal and human studies show that stress causes multiple changes in behavior and neural structures (Bay et al. 2004; Bay and Donders 2008; van Veldhoven et al. 2011; Portillo et al. 2023). Stress can set off a cascade of effects that include the parasympathetic and sympathetic nervous systems, oxidative stress, and changes in cytokines leading to disruption in neural function, the immune system, metabolism and the cardiovascular system (McEwen 2006). Experiences of stress highly correlate with sleep disruption (Armando et al. 2003; Saavedra and Benicky 2007). Specific neural centers show changes after stress, such as a reduction in newborn hippocampal neurons and a reduction in the dendrites of neurons in the hippocampus (Sousa et al. 2000; McEwen 2006). Structural changes are also seen in the amygdala and prefrontal cortex, which correlate with behavioral changes (McEwen and Gianaros 2011). In a study using mild repetitive TBI (similar to the injuries reported here) combined with stress, Portillo et al. report that sleep disruption occurs only in male mice receiving stress, either alone or combined with TBI, suggesting that stress may play a more important role than injury (Portillo et al. 2023). In our study, visual inspection of the actigraph plots show distinct alterations in many animals of the Injury alone and Injury + Stress groups compared with the Sham, at all post injury dates, with ferret activity less organized into clusters and more sporadic (Figs. 3–6). Our analysis also shows that TBI alone (of the type delivered here) results in altered activity patterns, up to at least at 4 weeks post injury as quantifying the LOESS analysis shows significant differences at the 4 week post injury time point. Assessing our data using other measures, including the amount of time highly active, shows a significant decrease in this activity in the Injury + Stress group at 4 weeks and 4 months post injury (Fig. 8). Our evaluation of sleep fragmentation shows a significant difference using a mixed effects analysis in both Injury alone and Injury + Stress groups at 4 weeks post injury. The immunoreactivity of AQP4 also displays an increase in intensity at 6 months post injury not seen in the Injury only group. Both Injury and Injury + Stress groups show significant increases in intensity and percent area labeled for GFAP and AQP4 at 4 weeks post injury. The Injury + Stress groups also show nearly significant increases at 6 months post injury; this is also indicated in Fig. 11. In addition, we see activity/sleep alterations in the Injury group alone. Although we observe several indications that the stressed animals show increased pathology, it was not consistent through all of our analyses and injury alone significantly effects sleep/activity disruption. This could be due to several reasons: (i) different levels or increases of stress may lead to significant differences between groups; (ii) we may have too low a number of animals to find significance; and (iii) we reduced the amount of stress after the injury period, and we may need to keep the stress delivery at the same level for the entire survival period.

Other animal studies show sleep alterations after TBI

Many other animal studies show sleep changes after TBI. Most publications use mice and study open skull impact type injuries and report changes in sleep wake behavior with loss of wakefulness or seizures, therefore reporting on a different type of injury than that used here (Skopin et al. 2015; Konduru et al. 2021). Many studies evaluate shorter time periods (several days), but several continue for many months (Bradshaw Jr et al. 2021). Olson et al (Olson et al. 2016) used actigraphy in pigs and also found disruptions in activity levels after both open skull impact and rotational injuries 4 days after the injury. In rats, Mountney et al (Mountney et al. 2021) report that a penetrating injury caused reduced time in REM sleep and also substantially altered gene expression, especially in immune and sleep related genes. Although our study in ferrets does not directly measure sleep, the use of actigraphs allows us to describe highly detailed periods of activity and rest not seen in many previous studies. We therefore present disrupted patterns after blast and CHIMERA injury that persist and also vary with the time of day.

Limitations

This is the first study to use actigraphy to study sleep/wake patterns in ferrets and builds upon our previous work establishing the ferret as an excellent model to study TBI (Schwerin et al. 2018; Schwerin et al. 2021). One of the limitations of this set of studies is that we used only male subjects. Sexually intact female ferrets are not often used in research due to serious estrus-related health problems if they do not become pregnant (Sherrill and Gorham 1985; Ball 2006). Interpreting sex differences after TBI studies is difficult, and more work needs to be done (Haynes and Goodwin 2021). Therefore, finding a way to incorporate female subjects in future ferret TBI studies is important, which we plan to do in future studies. We also have a relatively low number of animals, and several of our findings may have increased their significance level if a greater number of animals were included. Although ferrets are a very important gyrencephalic animal to use in TBI studies, their larger size requires greater housing space and overall expense, reducing the total number of animals that can be used in a given study. Related to this point, we do not have as many control groups as optimal. For example, we do not include a “stress only group”, which we would also like to include when possible. The absence of this group, however, does not diminish from our findings that mild repetitive TBI alone causes sleep/wake disruption, which is exacerbated by inclusion of stress. It is entirely possible (and even likely) that stress alone also causes similar disruptions.

Overall summary

Ferrets received a multimodal TBI (with and without stress) and wore actigraphs for up to 6 months after the injury. At Baseline before injury, or Sham animals, showed patterns of high activity that occurred in active bouts interspersed with periods of rest or sleep. Those with injuries displayed a breakdown of the clustered activity patterns that was significantly different from the Baseline or Sham at 4 weeks post injury using LOESS analysis or evaluation of sleep fragmentation. The amount of time spent highly active remained significantly different from Baseline up to 4 months post injury in the animals receiving Injury + Stress. Despite the lack of statistical significance in the measures we used, the activity patterns appear fractured in the Injury and Injury + Stress groups up to 6 months post injury. Analysis of reactive astrocytes using GFAP and AQP4 showed that these markers were significantly increased from Baseline for up to 6 months post injury in anterior cingulate cortex.

Acknowledgments

This research was supported by the Center for Neuroscience and Regenerative Medicine—CNRM-70-8956, USU- PAT-74-3439 and the CDMRP W81XWH-13-2-0018. We also thank the USUHS Department of Laboratory of Animal Resources and the Preclinical Studies Core staff- specifically Amanda Fu. We would also like to thank Michael Ray, Michael Strayhorn, Haley Spencer, Savannah Kounelis-Wuillaume, for their assistance conducting the experiments. We also thank Dr David Brody (Uniformed Services University) for his suggestion of the concepts presented here during his directorship of the Center for Neuroscience and Regenerative Medicine (CNRM).

CRediT author statement

Susan Schwerin (Conceptualization, Data curation, Formal analysis, Investigation, Methodology, Writing—original draft, Writing—review and editing), Nicholas Breehl (Data curation, Formal analysis, Methodology, Writing—review and editing), Adedunsola Obasa (Data curation, Investigation), Daniel Perl (Funding acquisition, Validation, Writing—review and editing), Yeonho Kim (Data curation, Formal analysis, Software), Joseph McCabe (Investigation, Methodology, Writing—review and editing), Thaddeus Haight (Data curation, Formal analysis, Methodology, Software, Writing—original draft, Writing—review and editing), Sharon Juliano (Conceptualization, Data curation, Formal analysis, Funding acquisition, Investigation, Methodology, Project administration, Writing—original draft, Writing—review and editing).

Conflict of interest statement: The authors have no conflict of interest to disclose.

Disclaimer

The views, information or content, and conclusions presented do not necessarily represent the official position or policy of, nor should any official endorsement be inferred on the part of, the Uniformed Services University, the Department of Defense, the US Government, or the Henry M. Jackson Foundation.

References