Impacts of traumatic brain injury severity and sex on sleep architecture, duration, and fragmentation

Abstract

Traumatic brain injury (TBI) is frequently associated with acute and chronic disturbances in sleep architecture. However, the extent to which injury severity and biological sex influence post-traumatic sleep patterns remains underexplored in preclinical models. Here, we used a validated, noninvasive piezoelectric monitoring system to assess sleep in male and female mice following sham (n = 30), mild (n = 32), or moderate (n = 32) midline fluid percussion injury (mFPI). Physiological parameters were recorded non-invasively to determine sleep for 48 h post-injury. Hierarchical mixed-effects models were used to evaluate effects of injury severity and sex on sleep duration, architecture, and fragmentation. We found that sleep increased during the acute post-injury period regardless of TBI severity, but that sleep fragmentation was selectively elevated after moderate injury. Notably, female mice exhibited greater overall sleep disturbances compared to males, highlighting a sex-dependent vulnerability. These effects varied across the light-dark cycle. This study provides the first detailed characterization of sex- and severity-specific changes in sleep architecture and fragmentation following diffuse TBI using a high-throughput, noninvasive method. Importantly, it reveals that injury severity predicts the extent of sleep fragmentation highlighting a direct link between injury severity and disrupted sleep architecture. These findings contribute to the growing recognition of sleep fragmentation as a relevant biomarker in TBI and establish a framework for future mechanistic and interventional studies.

Highlights

• •Sleep fragmentation increases with TBI severity in a preclinical mouse model.
• •Female mice show greater post-TBI sleep disruption than male mice.
• •Righting reflex time predicts acute sleep-wake fragmentation after TBI.
• •Piezoelectric monitoring enables non-invasive, high-throughput sleep assessment.

1. Introduction

Each year, approximately 70 million people worldwide, and 1.5 million in the United States, suffer from a traumatic brain injury (TBI) (Dewan et al., 2018). Accordingly, TBI is a substantive issue in global public health, affecting children, athletes, military personnel, and many others via events such as falls, assaults, or motor vehicle collisions (Aoun et al., 2019). At the time of injury, TBI is often classified as mild, moderate, or severe based on diagnostic scoring criteria referenced to a standardized scale, such as the Glasgow Coma Scale (GCS). (Teasdale and Gentleman, 1982; Teasdale and Jennett, 1974). Despite the undeniable benefits of these classification schemes, TBI has considerable variation; consequently, relative injury severity scores alone are insufficient to predict outcomes like patient pathology or recovery rate and further investigation into severity-specific responses is warranted (McAllister, 2011; Nelson et al., 2021; van Reekum et al., 2000). This heterogeneity also leads to a broad spectrum of cognitive, emotional, and physical symptoms among TBI survivors. Notably, over 50 % of all TBI survivors experience sleep disturbances-a rate significantly higher than that observed in the general population. (Gardani et al., 2015; Mathias and Alvaro, 2012; Sandsmark et al., 2017; Verma et al., 2007). Thus, sleep-wake disturbances are amongst the most prevalent sequelae of TBI.

Following TBI, common sleep-wake disturbances include excessive daytime sleepiness, sleep fragmentation, hypersomnia, and insomnia (Sandsmark et al., 2017). While not yet fully understood, proposed hypotheses for the pathophysiology of sleep disorders following TBI include diffuse axonal injury affecting the sleep regulation systems, disruptions in sleep-regulating hormones, and insults to the hypothalamus, reticular activating system (RAS), and brainstem (Castriotta and Murthy, 2011; Grima et al., 2016). Acute increases in sleep are also observed in TBI survivors, largely independent of injury severity (Aoun et al., 2019; Viola-Saltzman and Watson, 2012). Specifically, clinical data have shown that diffuse TBI increases slow wave sleep (SWS) following injuries of differing severities (Aoun et al., 2019; Suzuki et al., 2017). Increased cumulative or total sleep following TBI is associated with acute increases in sleep-regulating interleukin (IL)-1β, suggesting that altered sleep could be a useful biomarker for TBI (Rowe et al., 2014; Saber et al., 2020). Similarly, patients with acute moderate-to-severe TBI tend to have considerably lower levels of hypocretin-1 in cerebrospinal fluid compared to controls, emphasizing the need for research on post-TBI sleep at acute timepoints (Baumann et al., 2005). Such disruptions to sleep delay recovery and can precipitate or exacerbate long-term neurological morbidities. Ultimately, TBI-induced sleep disturbances translate to a worse prognosis for TBI survivors, longer hospital stays, and more severe disabilities (Green et al., 2020; Thomas and Greenwald, 2018).

Despite a growing appreciation for both the prevalence and societal costs of sleep disturbances after TBI, few clinical studies have investigated sleep architecture in the immediate post-injury period. This is likely partly due to concerns about using polysomnography (PSG) and other elaborate monitoring approaches to assess sleep in patients who are confused or severely injured (Wiseman-Hakes et al., 2016). Moreover, most mild traumatic brain injuries go unreported, and when reported, are often not treated in the clinic until weeks or months following the primary injury (Cassidy et al., 2004). As a result, clinical studies of acute post-traumatic sleep changes are sparse and highly biased towards patients with moderate-to-severe TBI. Nevertheless, data from PSG and electroencephalography (EEG)/electromyography (EMG) recordings of human subjects suggest that TBI generally reduces sleep efficiency, shortens sleep duration, and increases sleep fragmentation and wakefulness (Grima et al., 2016; Khoury et al., 2013; Mantua et al., 2019; Rao et al., 2011; Rao and Rollings, 2002; Sandsmark et al., 2017). Although limited, evidence also exists that TBI survivors may exhibit less rapid eye movement (REM) sleep or altered REM onset latencies (Grima et al., 2016; Mantua et al., 2019; Williams et al., 2008). Furthermore, the role of injury severity in post-traumatic sleep outcomes remains unclear in the current clinical literature. Although other consequences of TBI, such as motor and memory impairments, generally worsen with increased injury severity, sleep outcomes in the acute post-injury period are more complex (Sandsmark et al., 2017). Multiple reviews suggest that mild and moderate TBI acutely increase non-rapid eye movement (NREM) sleep but present conflicting claims regarding how injury severity affects sleep duration and specific stages (i.e., N1, N2, and N3) (Aoun et al., 2019; Castriotta et al., 2007; Mantua et al., 2019; Viola-Saltzman and Watson, 2012). Since mild and moderate TBIs account for ∼90 % of all brain injuries in the United States, a better understanding of the relationship between changes in sleep architecture and injury severity is essential to improve diagnostics and treatments (Dewan et al., 2018).

Substantial clinical data suggest women have worse recovery from TBI than men, but sex differences in post-traumatic sleep have been largely unexplored in preclinical experiments (Farace and Alves, 2000; Ma et al., 2019). Most preclinical TBI studies have used only male animals (Rubin and Lipton, 2019); for instance, none of the preclinical studies mentioned in a comprehensive review of sleep after TBI included female rodents (Sandsmark et al., 2017). However, significant biological sex differences exist in both the baseline (i.e., pre-injury) sleep and in post-TBI outcomes of C57BL/6 J mice subjected to controlled cortical impact (CCI) and fluid percussion injury (FPI) injury models (Saber et al., 2020). Sex hormones also have a key role in pathogenic processes and recovery in experimental TBI and have therefore been implicated in post-TBI sex differences observed in the clinic (Duncan, 2020; Golz et al., 2019; Kovesdi et al., 2020). Consequently, much of the contemporary knowledge regarding sleep architecture after TBI has only been tested and demonstrated in models with male rodents and, therefore, may not accurately reflect what happens in females.

Although previous studies have reported alterations in sleep following TBI (Rowe et al., 2014), few have systematically examined how injury severity and biological sex interact to shape acute changes in sleep architecture and continuity. Moreover, many preclinical investigations rely on invasive methods or focus exclusively on sleep duration, limiting insight into fragmentation and sleep quality. The present study addresses these gaps by employing a non-invasive, high-throughput piezoelectric monitoring system to quantify changes in sleep duration, architecture, and fragmentation in male and female mice following sham, mild, or moderate midline fluid percussion injury (mFPI). By stratifying analyses across time-of-day, injury severity, and sex, this study provides a detailed assessment of early post-traumatic sleep disturbances and identifies specific features, such as sex-dependent fragmentation after moderate injury, that may serve as sensitive behavioral indicators of injury burden. These findings contribute to a growing recognition of sleep fragmentation as a biologically relevant and clinically translatable outcome in TBI research.

2. Materials and methods

2.1. Rigor

All animal studies were conducted in accordance with the guidelines established by the Institutional Animal Care and Use Committee (IACUC) at the University of Colorado Boulder (protocol 2819) and the National Institutes of Health (NIH) guidelines for the care and use of laboratory animals. Studies are reported following the Animal Research: Reporting In Vivo Experiments (ARRIVE) guidelines (Kilkenny et al., 2010). In consideration of sex as a relevant biological variable, both male and female mice were used. Determination of sleep-wake behavior based on physiological parameters was performed by investigators blinded to experimental conditions.

2.2. Animals

Adult (12-25 weeks-old) male and female C57BL/6 J mice from 7 total sex-specific cohorts were used (total n = 94). All mice were bred in-house from breeder pairs obtained from Jackson Laboratories (Bar Harbor, ME). Mice were group-housed until the initiation of the study protocol. Female mice were naturally cycling, and to limit handling and stress, estrous stage was not determined or controlled for as a variable. Once on protocol, all mice were singly housed and maintained on a 12-h light:dark cycle (200 lux, cool white, fluorescent light) at an ambient temperature of 24 °C ± 2 °C. All mice were acclimated to non-invasive piezoelectric sleep cages (Signal Solutions, Lexington, KY, USA) for a minimum of 5 days prior to initiation of data collection. Mice were fed a normal diet of standard rodent chow, and food and water were available ab libitum. Prior to study initiation, mice were randomly assigned to an experimental group (moderate TBI, mild TBI, sham surgery; Fig. 1). All brain injuries were administered between zeitgeber (ZT) 4 and 6. Measurement of post-injury sleep parameters began at ZT 6 for all cohorts. All surgeries were performed under anesthesia with isoflurane, and exclusion criteria were predetermined: mice that lost >20 % of their bodyweight or had visible signs of pain or distress were euthanized and excluded from the study (n = 0). Predetermined inclusion criteria included a righting reflex time of >90 s for mild TBI (232 ± 54.3 s), and >315 s for moderate TBI (414 ± 103.6 s). Righting reflex times have previously been used to indicate injury severity (Grin'kina et al., 2016; Hosseini and Lifshitz, 2009; Rowe et al., 2014). The group sizes used in statistical analyses were: mild TBI female n = 15, moderate TBI female n = 15, sham (control) female n = 14, mild TBI male n = 17, moderate TBI male n = 17, sham (control) male n = 16.

Fig. 1.

Study Design. Male and female mice were subjected to sham (n = 30), mild (n = 32), or moderate (n = 32) midline fluid percussion injury (mFPI) at zeitgeber (ZT) 6. Physiological parameters were recorded using non-invasive piezoelectric cages to determine sleep in the acute post-injury period. All data were disaggregated by sex and analyzed using sex-specific models.

2.3. Midline fluid percussion injury

Mice were anesthetized with 5 % isoflurane in 100 % oxygen for 5 min, after which they were secured in a stereotaxic frame with a continuous flow of 2.5 % isoflurane in 100 % oxygen via a nosecone. Eye ointment was applied, and the surgical site was cleaned with betadine and ethanol. A midline excision was made to expose the skull. A craniectomy (3 mm diameter) was trephined midway between bregma and lambda on the sagittal suture. An injury cap was prepared from a Luer-Loc needle hub and fixed over the craniectomy using cyanoacrylate gel and methyl-methacrylate (Hygenic Corp., Akron, OH). The hub was filled with saline and closed with a cap made from a modified syringe tip to prevent debris and air exposure.

Mice were re-anesthetized 24 h after surgery with 5 % isoflurane in 100 % oxygen for 3 min. The cap was removed from the hub and the dura inspected to ensure it was intact. The hub was then filled with saline and attached to an extension tube connected to the fluid percussion injury (FPI) device (Custom Design and Fabrication, Virginia Commonwealth University, Richmond, VA). When a toe pinch response was detected, the pendulum was released, and a fluid impulse was delivered to the intact dura, as we routinely publish (Rowe et al., 2018; Rowe et al., 2018; Rowe et al., 2014; Saber et al., 2021). For mild TBI, the pendulum was released from a lower height. The righting reflex time was recorded and used as an indicator of injury severity. Sham (control) animals received a craniectomy and were attached to the FPI device, but the pendulum was not released. Once the mouse had returned to a supine position, it was briefly re-anesthetized, and the injury site was cleaned and closed with sutures. Antibiotic ointment (bacitracin) was applied, and the animal was returned to a heated holding cage.

2.4. Piezoelectric sleep recording parameters

To determine sleep-wake behavior, physiological parameters were recorded using a non-invasive piezoelectric cage system (Signal Solutions, Lexington, KY, USA), as previously described (Harrison et al., 2015; Rowe et al., 2018; Saber et al., 2019). Briefly, each cage has an open bottom that allows the mouse to be placed directly on a Polyvinylidine Difluoride sensor on the cage floor. These sensors are coupled to an input differential amplifier to generate pressure signals. These pressure signals are then converted to voltages, the amplitude of which is proportional to the pressure signal. Regular breathing movements characterize sleep [3 Hz, regular amplitude signals (Donohue et al., 2008)], whereas signals from awake mice are of higher amplitude with irregular spiking associated with volitional movements. In this study, the piezoelectric signals were analyzed over 10-s (epochs at a 2-s interval. Data collected from the sleep cages were binned at each hour using a rolling average of the percentage of recording time spent in sleep (Ouellette and Donohue, 2022). Data were also binned by length of individual bout to calculate the hourly mean bout length (duration in seconds). To be considered a bout, a minimum of 10 consecutive seconds had to be scored as sleep (or wake). Total minutes slept within each experimental period (e.g., day 1) were also calculated.

Using SleepStats Data Explorer Version 4, we also applied a classifier that discriminated between REM-like sleep and NREM sleep. A random forest decision tree classifier was developed to classify NREM and REM-like sleep based on piezo signal features related to irregularities in breathing patterns that are characteristic of changes in autonomic control during REM sleep. Specifically, cage-floor pressure signals from the animal's respiratory thorax motion become more irregular during REM-like sleep and exhibit speed changes and low-level amplitude variations over short time intervals, which are not typically observed during NREM sleep. The non-invasive piezoelectric data acquisition system, PiezoSleep with SleepStats software programs have been used to accurately determine REM and NREM sleep in both the mouse and the rat (Topchiy et al., 2022; Yaghouby et al., 2016). The 3-state algorithm to determine WAKE, NREM, and REM-like sleep from piezo system recordings has also been validated against data collected from mice that were simultaneously recorded using EEG/EMG human-scored vigilance state classification on 4-s intervals (Yaghouby et al., 2016). The classification accuracy and utility of the system have been discussed extensively in a previous article (Mannino et al., 2024).

To determine sleep-wake transitions, we used the sleep-wake decision statistic and decision threshold features of SleepStats Data Explorer Version 4. We plotted a time series of decision statistics, which represents the likelihood of each sleep state based on features derived every 2 s from the pressure signals. For each animal, a decision threshold is automatically calculated for each cage over the 48-h sleep period, based on maximizing the decision statistic separation between the sleep and wake clusters. Throughout the 48-h post-injury period, each 2-s interval is classified as sleep if its decision statistic exceeds the threshold; otherwise, it is classified as wakefulness. To identify transitions from wake to sleep, we counted instances where the decision statistic rose above the threshold after being below it during the preceding 2-s interval. Similarly, we counted transitions from sleep to wake when the opposite instances occurred. Summing these values provided the total number of sleep-wake transitions, which we report for both hourly and daily intervals.

2.5. Statistical analyses

To investigate sex-specific sleep differences between injury severities for each vigilance state (NREM, REM-like and WAKE), we fit hierarchical generalized linear mixed models (Faraway, 2016; Stroup, 2012) using the package glmmTMB in the R statistical computing environment (Brooks et al., 2017; Team, 2023). Because the outcome measures of interest were percentages (percent sleep) and overdispersed counts (bout lengths, total minutes slept, and number of transitions between sleep - wake states), we specified Beta and negative-binomial distributions, respectively, in the statistical models (Ferrari and Cribari-Neto, 2004; Hilbe, 2014).

Percentage of time spent in a vigilance state and sleep/wake bout lengths were sequential time-series data with cyclical light-dark period trends within which observations close to each other in time are expected to be more similar than observations further apart in time (i.e., temporal dependency among observations (Kedem, 2002; R. K. Rowe et al., 2022a, Rowe et al., 2022b; Woolridge, 2016);). Additionally, the rodent sleep cycle is expected to have temporally varying nonlinear interactions with predictor variables and temporally varying nonlinear effects on outcome measures (R. K. Rowe et al., 2022a, Rowe et al., 2022b; Saber et al., 2020; Saber et al., 2021). Therefore, we included three-knot and five-knot basis splines for time (ZT) in all percentage sleep and sleep bout length models, respectively, to accommodate both the expected similarity among observations within a period and the nonlinear effects of time (de Boor, 2001; Perperoglou et al., 2019; R. K. Rowe et al., 2022a, Rowe et al., 2022b). To determine the appropriate number of spline knots for each percentage sleep and sleep bout length model, we fit models with two to seven spline knots, conducted model selection using Akaike's Information Criterion corrected for small sample size (AIC_c), and produced estimates from the top-ranked model that best described the data for each outcome (Burnham, 2002; 2011).

For both percentage of time spent in a vigilance state and sleep/wake bout lengths, we further subdivided the data by vigilance state (NREM, REM-like, and WAKE) and day post-injury (1 vs. 2 days post-injury), and fit sex- and vigilance state-specific models with a three-way interaction among injury severity, light-dark period, and ZT with a basis spline (R. K. Rowe et al., 2022a, Rowe et al., 2022b; Saber et al., 2020; Saber et al., 2021). For total sleep and vigilance state transitions, we also subdivided the data by vigilance state and fit type-specific models, but we included a two-way interaction between injury severity and light-dark period, disregarding ZT, because total minutes slept were derived by summing across all ZT within each day. Because the data were comprised of repeat observations from the same animals, we included random intercepts in all models for individual mice; this hierarchical structure accounted for the potential dependency or clustering of observations from the same mouse across time (Faraway, 2016; Stroup, 2012). We based statistical inferences on a combination of coefficient estimates (β), differences between estimated conditional means (Δ) and associated 95 % confidence intervals, and p-values following Tukey's adjustments for multiple comparisons (Dunn, 1961), all of which we obtained using the package emmeans in R (R.K. Rowe et al., 2022a, Rowe et al., 2022b). Additionally, we used Kolmogorov-Smirnov (KS) tests to compare the predicted distributions of percentage sleep and sleep bout lengths among injury severities, using sham as the reference distribution; a statistically significant (p < 0.05) KS value indicated that the distributional patterns between sham and mild TBI and sham and moderate TBI were dissimilar.

To account for previously published large sex differences in baseline sleep (Mannino et al., 2024), we used geometric mean ratios to compare sleep outcomes between sexes. This approach allows for a standardized scaling of sleep parameters, ensuring that differences are expressed relative to baseline values (sham values) rather than absolute differences, which may be influenced by inherent sex-based variations in sleep architecture. Using geometric mean ratios reduces the impact of baseline variability and facilitates a more interpretable comparison of relative changes in sleep outcomes between male and female mice. Comparisons of geometric mean ratios for relative sex-specific differences in total NREM sleep (min), total REM sleep, and total WAKE between sham and mild TBI and sham and moderate TBI, during the light and dark periods across the first two days post-injury were determined. Geometric mean ratios were produced from vigilance state-specific generalized linear mixed models with negative-binomial response distributions, three-way interactions among sex, injury, and period, and random intercepts for individual animals nested within their respective cohorts. Ratios are centered on 1.00, which indicates no difference between sham and injury groups, whereas ratios >1.00 indicate the injury group had less total minutes and ratios <1.00 indicate the injury group had more total minutes in a vigilance state compared to respective shams. Comparing geometric mean ratios between sexes within a period within a day is a direct comparison of the relative sizes of differences, an approach often used in meta-analyses (Friedrich et al., 2012; Higgins et al., 2008).

To determine whether relationships existed between the number of sleep-wake transitions and righting reflex times, we fit simple linear regression models with Gaussian error distribution to the post-injury righting reflex times (seconds), with sex and number of sleep-wake transitions as the fixed effects, for each sex within each post-injury day.

3. Results

3.1. TBI altered NREM sleep and WAKE, but not REM-like sleep

TBI altered the distribution (Fig. 2) and total amount (Fig. 3) of NREM sleep and WAKE, but not REM-like sleep (Fig. 2), in an injury severity-, sex-, and time-dependent manner. The distributions of NREM sleep differed substantially between days for all groups within each sex (KS_{Female × Day} = 0.39, p < 0.0001; KS_{Male × Day} = 0.37, p < 0.0001; Fig. 2A–B). During the first day post-injury, both female TBI groups exhibited significantly more NREM sleep than sham control mice during the first light period (p < 0.0001) and the subsequent dark period (p = 0.0002), but not thereafter (p = 0.78). Relative to sham control mice, male TBI groups exhibited significantly more NREM sleep throughout the entire first day post-injury (p < 0.0001; p = 0.02; and p = 0.004, first light period, dark period, and second light period, respectively). By 2 days post-injury, only the moderate TBI group during the dark period was significantly different from shams for males and females (p = 0.001–0.004). Standardized effect sizes for supported differences in NREM sleep among injury groups were small (d ≤ 0.10) except for the first light period at 1 day post-injury for both females (d = 0.21) and males (d = 0.28).

Fig. 2.

Mild and moderate TBI increase NREM sleep and decrease WAKE in both female and male mice. Both mild and moderate TBI increased the percentage of time spent in NREM sleep and decreased time spent awake compared to sham controls, with consistent effects across sexes and time points. No group differences were observed for REM-like sleep. (A–B) Percent of time spent in NREM sleep was significantly higher in both mild and moderate TBI groups compared to shams on post-injury days 1 and 2. (C–D) Percent of time spent in REM-like sleep did not differ between injury groups. (E–F) Percent of time spent awake was significantly lower in TBI groups compared to sham on both post-injury days. Results shown are predicted conditional effects point estimates and 95 % confidence intervals estimated from hierarchical models with Beta error distributions. Individual background points represent the raw observed data. Shaded gray background areas represent the dark (active) period. Asterisks (∗) indicate significant difference between moderate TBI and sham; pound signs (#) denote significant difference between mild TBI and sham.

Fig. 3.

Mild and moderate TBI increase minutes of NREM sleep and decrease minutes of WAKE. Mice with mild or moderate TBI spent more time in NREM sleep and less time awake compared to sham controls, particularly during the dark (active) period. No group differences were observed in REM-like sleep across injury groups or time periods. (A–B) Mice with mild and moderate TBI showed significantly more minutes of NREM sleep during both the light and dark periods at 1 day post-injury, and during the dark period at 2 days post-injury. (C–D) There were no differences in minutes spent in REM sleep in the light or dark period following mild or moderate TBI compared to shams. (E–F) TBI groups exhibited significantly fewer minutes of wake during both the light and dark periods at 1 day post-injury, and during the dark period at 2 days post-injury. Results shown are predicted conditional effects point estimates and 95 % confidence intervals estimated using hierarchical models with negative-binomial error distributions. Individual background points represent the raw observed data. Asterisks (∗) indicate significant difference between moderate TBI and sham; pound signs (#) denote significant difference between mild TBI and sham.

The distributions of REM-like sleep differed between days for all groups within each sex (KS_{Female × Day} = 0.26, p = 0.02; KS_{Male × Day} = 0.24, p = 0.04; Fig. 2C–D). However, at both 1- and 2 days post-injury, REM-like sleep was similar among groups within each sex (p > 0.05). All standardized effect sizes for REM-like sleep comparisons among injury groups were nominal (d < 0.05).

The distributions of WAKE differed substantially between days for all groups within each sex (KS_{Female × Day} = 0.39, p < 0.0001; KS_{Male × Day} = 0.40, p < 0.0001; Fig. 2E–F). At both 1- and 2 days post-injury, both female and male TBI groups exhibited significantly less WAKE than shams during the first light period, the dark period, and second light period (p < 0.0001). Standardized effect sizes for supported differences in WAKE among injury groups were small (d ≤ 0.10).

Total NREM sleep was significantly greater for mild and moderate TBI groups compared to shams in males and females during the light period (first 6-h and second 6-h light period combined) and the dark period at 1 day post-injury, but only the dark period at 2 days post-injury (p < 0.0001; Fig. 3A–B). Total REM-like sleep did not notably differ between TBI groups and shams in males or females during the light or dark periods at 1- or 2 days post-injury (p > 0.20; Fig. 3C–D). Total WAKE was significantly lower for mild and moderate TBI groups compared to shams in males and females during the light and dark periods at 1 day post-injury (p < 0.0001), but only the dark period at 2 days post-injury (p = 0.005; Fig. 3E–F). All standardized effect sizes for supported differences in total REM-like and NREM sleep and total WAKE between sham and TBI groups were nominal (d < 0.01).

3.2. WAKE bout lengths substantially differed between sham and TBI during the light and dark periods

Although the distributions of NREM sleep bout lengths differed between days for all groups within each sex (KS_{Female × Day} = 0.24, p = 0.04; KS_{Male × Day} = 0.33, p = 0.0006; Fig. 4A–B), NREM sleep bout lengths were similar among groups within each day for both sexes (p > 0.05). Standardized effect sizes for NREM bout length comparisons among groups were nominal (d ≤ 0.01). The distributions of REM-like sleep bout lengths did not differ between days for any groups within either sex (KS_{Female × Day} = 0.05, p = 0.56; KS_{Male × Day} = 0.08, p = 0.53; Fig. 4C–D), and REM-like sleep bout lengths were similar among groups within each day for both sexes (p > 0.05).

Fig. 4.

TBI reduces the ability to sustain wakefulness without affecting NREM or REM bout lengths. Mice with mild or moderate TBI exhibited significantly shorter wake bout lengths compared to sham controls, indicating increased sleep fragmentation. No group differences were observed in the duration of NREM or REM-like sleep bouts. (A–B) There were no differences in NREM bout lengths, or (C–D) or REM bout lengths following mild or moderate TBI compared to shams. (E–F) WAKE bout lengths were significantly reduced in both mild and moderate TBI groups compared to sham controls on both post-injury days, particularly during the dark (active) period. Results shown are predicted conditional effects point estimates and 95 % confidence intervals estimated from hierarchical models with Beta error distributions. Individual background points represent the raw observed data. Shaded gray background areas represent the dark (active) period. Asterisks (∗) indicate significant difference between moderate TBI and sham; pound signs (#) denote significant difference between mild TBI and sham.

The distributions of WAKE bout lengths substantially differed between days for all groups within each sex (KS_{Female × Day} = 0.49, p < 0.0001; KS_{Male × Day} = 0.38, p < 0.0001; Fig. 4A–B). At 1 day post-injury, female mild and moderate TBI groups exhibited significantly shorter WAKE bout lengths than female shams during first light (p < 0.0001) and dark (p = 0.002) periods, but not the second light period (p = 0.21). Male mild and moderate TBI groups exhibited significantly shorter WAKE bout lengths at 1 day post-injury than male shams during both the first (p < 0.0001) and second (p = 0.002) light periods but not the dark period (p = 0.34). At 2 days post-injury, only the moderate TBI group during the dark period was significantly different from shams for males and females (p = 0.008–0.01). Standardized effect sizes for supported differences in WAKE bout lengths among injury groups were small (d ≤ 0.10).

3.3. TBI fragments sleep

At 1 day post-injury, moderately injured females and males exhibited significantly more transitions between sleep and wake than both shams and mice subjected to mild TBI (Fig. 5A). During the light period (first 6-h and second 6-h light period combined), females subjected to moderate TBI had 523 and 460 more transitions than female shams (p < 0.0001) and females subjected to mild TBI (p < 0.0001), respectively, whereas shams and mild TBI females had a similar number of transitions (p = 0.70). Similarly, during the light period, moderate TBI males had 490 and 229 more transitions than male shams (p < 0.0001) and mild TBI males (p = 0.008), respectively; however, mild TBI males had 260 more transitions than male shams (p = 0.0001). During the dark period, moderate TBI females had 601 and 497 more transitions than female shams (p < 0.0001) and mild TBI females (p < 0.0001), respectively, whereas shams and mild TBI females had a similar number of transitions (p = 0.32). Similarly, during the dark period, moderate TBI males had 400 and 240 more transitions than male shams (p < 0.0001) and mild TBI males (p = 0.003), respectively; however, mild TBI males had 159 more transitions than male shams (p = 0.0001).

Fig. 5.

Moderate TBI increases the number of sleep-wake transitions. Mice with moderate TBI exhibited significantly more sleep-wake transitions than both mild TBI and sham groups, particularly during the dark period. This effect was consistent across sexes and persisted for at least two days post-injury. (A) At 1 day post-injury, moderate TBI increased sleep-wake transitions in both the light and dark periods in female and male mice compared to respective mild TBI and sham groups, with mild TBI in males also increasing transitions compared to male shams. (B) At days post-injury, moderate TBI continued to increase sleep-wake transitions in both the light and dark periods in female and male mice compared to respective mild TBI and sham groups. Results shown are predicted conditional effects point estimates and 95 % confidence intervals estimated using hierarchical models with negative-binomial error distributions. Individual background points represent the raw observed data. Asterisks (∗) indicate significant difference between moderate TBI and sham; pound signs (#) denote significant difference between mild TBI and sham.

At 2 days post-injury, moderate TBI females exhibited significantly more transitions than both shams and mice subjected to mild TBI, whereas moderate TBI males exhibited significantly more transitions than shams only (Fig. 5B). During the light period, moderate TBI females had 179 and 205 more transitions than female shams (p = 0.03) and mild TBI females (p = 0.02), respectively, whereas shams and mild TBI females had a similar number of transitions (p = 0.93). During the light period, moderate TBI males had 219 more transitions than male shams (p = 0.001), but transitions were similar to mild TBI males (p = 0.13). During the dark period, moderate TBI females had 290 and 212 more transitions than female shams (p < 0.0001) and mild TBI females (p = 0.006), respectively, whereas shams and mild TBI females had a similar number of transitions (p = 0.46). During the dark period, moderate TBI males had 201 more transitions than male shams (p = 0.002), but transitions were similar to mild TBI males (p = 0.07).

3.4. Female mice exhibit greater sleep disruption than male mice relative to respective shams

Geometric mean ratios were used to directly compare the magnitude of change in sleep outcomes while accounting for baseline physiological sex differences in sleep (Table 1). Overall, TBI caused greater relative disturbances in sleep architecture in female mice compared to their respective sham controls. However, on day 1 post-injury, male mice with mild TBI exhibited larger relative differences in all three vigilance states during the light period compared to male shams, while those with moderate TBI had a larger relative difference in WAKE. Additionally, males with mild TBI showed greater relative changes in NREM sleep during the dark period on day 2 post-injury. Across all other time points and injury severities, female mice consistently exhibited larger relative deviations from sham controls than males.

Table 1.

Comparisons of geometric mean ratios for relative sex-specific differences in total NREM sleep (min), total REM sleep, and total WAKE between sham and mild TBI and sham and moderate TBI, during the light and dark periods across the first two days post-injury (DPI). Geometric mean ratios were produced from vigilance state-specific generalized linear mixed models with negative-binomial response distributions, three-way interactions among sex, injury, and period, and random intercepts for individual animals nested within their respective cohorts. Ratios are centered on 1.00, which indicates no difference between sham and injury groups, whereas ratios >1.00 indicate the sham group had more total minutes and ratios <1.00 indicate the injury group had more total minutes.

Comparison	Sex	DPI	Period	REM	NREM	WAKE
Sham vs. Mild TBI	Female	1	Light	1.03	0.80	1.26
	Male	1	Light	1.20	0.78	1.32
	Female	1	Dark	0.88	0.76	1.26
	Male	1	Dark	1.03	0.83	1.18
	Female	2	Light	1.02	0.96	1.05
	Male	2	Light	0.99	1.02	0.98
	Female	2	Dark	0.84	0.86	1.10
	Male	2	Dark	1.04	0.84	1.10
Sham vs. Moderate TBI	Female	1	Light	1.12	0.73	1.38
	Male	1	Light	0.94	0.75	1.49
	Female	1	Dark	0.86	0.78	1.26
	Male	1	Dark	0.91	0.83	1.20
	Female	2	Light	0.88	1.00	1.05
	Male	2	Light	0.94	0.99	1.02
	Female	2	Dark	0.73	0.77	1.21
	Male	2	Dark	1.03	0.78	1.16

3.5. Injury severity predicts the extent of sleep fragmentation

Righting reflex times were used in this study as an indicator of injury severity (Grin'kina et al., 2016). In both female and male mice, there was a positive relationship between righting reflex times and the number of sleep-wake transitions (Fig. 6). Longer righting reflex times were predictive of more sleep-wake transitions in female mice subjected to TBI at 1 day post-injury (p = 0.0007) and 2 days post-injury (p = 0.0228). Similarly, longer righting reflex times were predictive of more sleep-wake transitions in male mice subjected to TBI at 1 day post-injury (p = 0.0002) and 2 days post-injury (p < 0.0001).

Fig. 6.

Longer righting reflex times predict greater sleep fragmentation following TBI. Righting reflex time, a marker of injury severity, was positively associated with the number of sleep-wake transitions in both female and male mice across the first two days post-injury. These findings suggest that more severe injuries result in greater sleep fragmentation. (A) In female and (B) male mice, righting reflex time was significantly correlated with sleep-wake transitions on post-injury days 1 and 2. Each plot shows individual mouse data points with regression lines and 95 % confidence intervals. Results are from simple linear regressions and include the beta coefficient (β), 95 % confidence intervals (CI), and associated p values.

4. Discussion

Our findings indicate that under the conditions of this study, diffuse TBI acutely disrupts sleep in male and female C57BL/6 J mice, with female mice exhibiting larger relative differences in vigilance states compared to respective shams. In both male and female mice, irrespective of injury severity, sleep increases at the expense of WAKE. In contrast, injury severity, as quantified by righting reflex time immediately after injury, predicts the extent to which sleep is fragmented during the acute post-injury period. This latter finding suggests a dichotomy between TBI-induced effects on sleep duration and sleep fragmentation.

The present findings extend prior work by demonstrating that sleep fragmentation, rather than total sleep duration alone, may serve as a more sensitive and sex-dependent indicator of TBI severity during the acute post-injury period. Although prior clinical and preclinical studies have identified broad associations between TBI and altered sleep, few have simultaneously examined the interactive effects of injury severity, biological sex, and sleep fragmentation using a standardized, non-invasive platform. By leveraging high-resolution piezoelectric monitoring in a rigorously stratified murine model, this study provides new insight into how moderate TBI disproportionately disrupts NREM sleep continuity in female mice, suggesting differential vulnerability across biological sex. These findings also highlight the importance of diurnal periods in shaping post-traumatic sleep disruption.

Although our classifier distinguishes between WAKE, NREM, and REM-like states, it is important to acknowledge limitations in detecting REM sleep with non-invasive systems such as piezoelectric monitoring. REM sleep has lower amplitude respiratory and postural signals compared to NREM and WAKE, making it more difficult to classify without EEG validation. While prior work supports the classifier's utility and general accuracy (Mannino et al., 2024; Ouellette and Donohue, 2022), REM classification remains more prone to error than NREM or WAKE detection. Accordingly, the null findings in REM-like sleep in our dataset should be interpreted cautiously, and future studies employing EEG/EMG validation are warranted to confirm these results.

Consistent with our current findings, previous fluid percussion injury studies in male rodents also demonstrate acute increases in sleep (Lim et al., 2013; Rowe et al., 2014; Saber et al., 2019). To the best of our knowledge, the present study is the first to examine post-traumatic sleep architecture using a non-invasive approach capable of determining all three rodent vigilance states (i.e., NREM, REM-like and WAKE) using a validated piezoelectric monitoring system. This classifier, previously validated against EEG/EMG recordings, has demonstrated strong agreement with gold-standard methods for detecting NREM and WAKE states, and it enables a scalable, longitudinal, and minimally invasive sleep assessment (Yaghouby et al., 2016). Additionally, recent work in our lab (Mannino et al., 2024) has confirmed the stability and reliability of this classifier across sex and light-dark cycle, establishing its utility for detecting robust sex differences in C57BL/6 J mice. Our data are also the first to date to detail acute increases in NREM sleep following midline fluid percussion injury. Notably, past rodent studies that examine sleep post-TBI at acute timepoints (i.e., 24–48 h post-injury), show no change to the amount of time spent in NREM sleep (Buchele et al., 2016; Willie et al., 2012). However, these disparities are likely due to differences in experimental injury models and the underlying model-specific biomechanics that differentially affect inflammation and neurovascular hemodynamics (Green et al., 2024). Our data are clinically relevant, as acute increases in slow wave sleep (SWS) are also observed in human TBI survivors; changes which are largely independent of injury severity (Aoun et al., 2019; Viola-Saltzman and Watson, 2012).

In this study, WAKE bouts in mice of both sexes subjected to mild or moderate TBI are shorter, particularly during the dark (active) period. The effects of TBI on WAKE bout duration persist for at least two days post-injury, the duration of our study protocol. Decreases in the length of WAKE bouts have been previously observed after fluid percussion injury (Rowe et al., 2014; Rowe et al., 2014) and following exposure to other TBI models (Foda and Marmarou, 1994; Willie et al., 2012). Such findings indicate that brain-injured mice exhibit a decreased ability to maintain a state of prolonged wakefulness. Similarly, individuals suffering from brain injury commonly report excessive daytime sleepiness during the acute phase post-injury (Baumann, 2012; Castriotta et al., 2007; Ouellet and Morin, 2006). Notably, our data replicate these clinical observations, which reinforces the relevance of the fluid percussion injury model for resembling some aspects of diffuse clinical TBI (Lifshitz et al., 2016; Rowe et al., 2016).

Sleep continuity, or the stability of sleep across a night, is characterized by minimal interruptions or awakenings and is a critical factor for determining the quality of sleep. Disruptions in sleep continuity, such as sleep fragmentation, are commonly associated with insomnia (Rosenberg, 2006), which is one of the most frequently reported sleep-wake disturbances following a TBI (Sandsmark et al., 2017; Viola-Saltzman and Watson, 2012). Furthermore, mice subjected to sleep fragmentation after TBI have delayed cognitive recovery and chronic alterations in neuroimmune function (Houle et al., 2024), highlighting the negative implications of fragmented sleep in the post-injury period. Here, we show that brain injury severity, as determined immediately after injury, predicts the number of sleep-wake transitions, a key marker of sleep fragmentation. Sleep of male and female mice with moderate TBI is more fragmented than that of mice subjected to mild TBI or sham surgery. However, the total amount of sleep during the acute post-injury period is not injury severity-dependent, indicating that TBI in this model impacts sleep quality and not sleep quantity. These findings highlight the need to evaluate sleep metrics beyond quantity in experimental studies and demonstrate that injury severity plays a critical role in sleep fragmentation. Frequent transitions between sleep and wake states, indicative of fragmented sleep, may serve as a measurable biomarker for TBI severity, offering insight into the disruption of sleep architecture caused by brain injury.

Dysfunction in the hypocretin (orexin) system likely underlies some of the observed sleep-wake disturbances after TBI. A potential role for hypocretin in post-TBI sleep-wake disturbances is supported by its upstream position relative to other wake-promoting systems, by its well-documented role in stabilizing wakefulness, and by several key clinical observations (Lim et al., 2013; Saper et al., 2005; Thomasy et al., 2017). For example, the number of hypocretin neurons is significantly reduced in post-mortem brains following fatal TBI, and abnormally low or undetectable levels of hypocretin are measured in the cerebrospinal fluid of patients within the first four days post-injury (Baumann et al., 2005, 2009). Similarly, TBI survivors experiencing excessive daytime sleepiness exhibit persistently low cerebrospinal fluid hypocretin levels up to six months post-injury. (Baumann et al., 2007). Studies on mice with genetic ablation of hypocretin highlight its role in mediating TBI-induced decreases in wakefulness, while genetically intact mice demonstrate a reduction in hypothalamic hypocretin-producing neurons at 7- and 15 days post-injury (Thomasy et al., 2017; Thomasy and Opp, 2019). In vivo microdialysis in mice reveals decreased extracellular hypocretin three days post injury, and fluid percussion injury impairs hypocretin neuronal activation (Lim et al., 2013; Willie et al., 2012). Our results align with evidence that reduced hypocretin activity after TBI may impair the normal regulation of sleep states, resulting in an increased number of transitions between sleep and wakefulness. Together, these findings highlight the critical role of hypocretin in the pathophysiology of post-TBI sleep-wake disruptions. Future studies incorporating molecular, inflammatory, or neural circuit-level markers, particularly those linked to arousal regulation and homeostatic sleep drive, are warranted to elucidate the underlying mechanisms of these effects and evaluate potential therapeutic targets.

Our recent work demonstrates that significant biological sex differences exist in the sleep of C57BL/6 J mice; male mice exhibit significantly more NREM sleep under normal physiological conditions than female mice (Mannino et al., 2024). The post-injury sleep-wake phenotypes shown herein reflect changes from normal sex-specific sleep patterns in this mouse strain. Accordingly, we compared sex-specific geometric mean ratios of the differences in total time spent in each vigilance state following mild and moderate TBI to that of the sham control animals. Results from that analysis demonstrate that TBI generally alters the time spent in each vigilance state proportionally more in females than in males compared to their respective shams (Table 1). These findings further our understanding of sex-specific responses to experimental TBI and demonstrate the importance of including and accounting for both sexes in preclinical TBI research. Continued research is warranted to further elucidate the ways in which biological sex may impact post-TBI sleep architecture.

As with any study, there are some limitations to this one. Our protocol focused on acute timepoints post-TBI, which have previously been utilized as a key window for identifying therapeutic targets and thereby improving long-term functional outcomes (Apostol et al., 2022; Lim et al., 2013; Rowe et al., 2018). However, men and women exhibit distinct recovery trajectories and sleep phenotypes over the course of acute and chronic periods following TBI (Howell and Griesbach, 2024; Ledger et al., 2020). The present study did not determine chronic changes to post-TBI sleep architecture, and thus we do not know if the acute sex differences in sleep persist, worsen, or resolve over time. Future studies incorporating longitudinal sleep monitoring over weeks to months post-injury are needed to map the trajectory of sleep disturbances, determine whether sleep fragmentation resolves or becomes persistent, and identify opportunities for long-term intervention.

Another important consideration is that to minimize handling and non-specific stress responses, we did not track the estrous cycles of the female mice. As such, it is possible that some of the sex differences we observed may be influenced by fluctuating levels of estrogen and progesterone, which are known to modulate both sleep architecture and neuroinflammatory responses (Collins et al., 2022; Copinschi and Caufriez, 2022; Deurveilher et al., 2011). While this omission was intentional to reduce potential confounds related to stress and disturbed sleep, it limits our ability to separate hormonally driven variability from more stable sex-dependent effects. Clinical data suggest that biological sex can predict multiple components of TBI prognosis such as sensitivity to sensory stimuli, affective symptoms, and functional outcomes (Mikolic et al., 2021; Starkey et al., 2022). These marked sex differences in TBI recovery have led to the longstanding hypothesis that sex hormone signaling may play a key role in driving these differences (Ma et al., 2019). Without estrous cycle tracking, we cannot determine the extent to which post-TBI sleep architecture is influenced by hormonal fluctuations. Future studies are needed to explore the role of sex hormones in shaping post-traumatic sleep architecture following diffuse TBI.

In conclusion, this study shows that TBI increases NREM sleep duration irrespective of injury severity. We further identify injury severity as a predictor of sleep fragmentation, directly linking greater injury severity to disrupted sleep continuity. Fragmented sleep, characterized by frequent transitions between sleep and wake states, may serve as a biomarker for TBI severity, providing a measurable indicator of brain injury. Given prior evidence that post-TBI sleep fragmentation is associated with impaired cognitive recovery and altered neuroimmune signaling, these findings suggest that sleep continuity may also reflect underlying mechanisms that shape long-term functional outcomes. These results emphasize the importance of considering injury severity when managing post-traumatic sleep disturbances and suggest that tailored approaches could enhance recovery outcomes for individuals with different levels of brain injury. Future studies that include behavioral, cognitive, or inflammatory endpoints alongside sleep outcome measures will be essential for clarifying how sleep quality influences recovery trajectories and for identifying actionable targets for intervention.

CRediT authorship contribution statement

Grant S. Mannino: Writing – original draft, Visualization, Methodology, Investigation, Formal analysis, Data curation, Conceptualization. Tabitha R.F. Green: Writing – review & editing, Methodology, Data curation. Sean M. Murphy: Writing – review & editing, Formal analysis. Michael R. Sierks: Writing – review & editing, Supervision, Funding acquisition. Mark R. Opp: Writing – review & editing, Supervision, Resources, Project administration. Rachel K. Rowe: Writing – review & editing, Supervision, Resources, Methodology, Funding acquisition, Formal analysis, Data curation, Conceptualization.

Ethics approval statement

All animal procedures were conducted in accordance with the guidelines of the University of Colorado Boulder Institutional Animal Care and Use Committee (IACUC) and were approved under protocol #2819. Experimental data are reported in compliance with the ARRIVE guidelines to ensure rigorous and transparent reporting of animal research.

Financial disclosures

None.

Non-financial disclosures

None.

Funding

This work was supported in part by grants from the Department of Defense awards W81XWH-22-1-0383 (RR), W81XWH-14-1-0467 (MS and RR), and W81XWH-22-1-0384 (MO). RR and GM are supported, in part, by funds from the Department of Integrative Physiology at the University of Colorado Boulder. -

Declaration of competing interest

None.

Data availability

The data supporting the conclusions of this manuscript are publicly available in the Open Data Commons for Traumatic Brain Injury (ODCTBI): http://doi.org/10.34945/F5K30W.