Animals
Eight-week-old CD1 male mice (~ 25–30 g; Envigo, Italy) were maintained in SPF facilities and housed at constant room temperature (23 °C) and relative humidity (60 ± 5%) with free access to food and water and a fixed 12 h light/dark cycle. Mice were individually housed with environmental enrichment (toilet paper, straw, nesting material) [24].
Procedures involving animals and their care were conducted in conformity with institutional guidelines that are in compliance with national (D.L. n.26, G.U. March 4, 2014) and international guidelines and laws (EEC Council Directive 86/609, OJ L 358, 1, December 12, 1987, Guide for the Care and Use of Laboratory Animals, U.S. National Research Council, 1996; laws of the United States and regulations of the Department of Agriculture), and were reviewed and approved by the intramural ethics committee and by the Animal Care and Use Review Office (ACURO; mice in cohort 2, see below). All animal experiments were designed in accordance with Animal Research Reporting of In Vivo Experiments (ARRIVE) guidelines [25], with a commitment to refinement, reduction, and replacement, and using biostatistics to optimise the number of mice.
Experimental design
Figure 1 shows the timeline of the experiments. We prepared two independent cohorts of TBI and their respective control mice (cohorts 1 and 2). In cohort 1, we ran a longitudinal ECoG monitoring 3 and 5 months after TBI (24/7, 3 weeks each recording session) to measure PTE incidence, and the frequency and duration of spontaneous seizures. We selected these two time points on the basis of previous evidence that they illustrate the early and late (chronic) epilepsy phases in the model, respectively [19]. Sensorimotor deficits and white matter damage were evaluated in the same mice.
In a subsequent cohort 2, we ran a longitudinal study to characterise behavioral (sensorimotor deficits, as well as motor and cognitive tests) and brain structural changes by MRI, with the main aim of identifying biomarkers of PTE development. In this cohort, mice with or without spontaneous seizures were identified at 5 months post-TBI only (chronic phase of epilepsy) since a progression in the incidence of PTE was observed between 3 and 5 months in cohort 1.
Histopathology was evaluated at the end of the experiments in mice taken at random from cohorts 1 and 2.
Induction of traumatic brain injury
Mice (n = 45) were anesthetized with isoflurane inhalation (2% for induction, 1.5% for maintenance) in an N2O/O2 (70%/30%) mixture and placed in a stereotaxic frame, then subjected to craniectomy followed by induction of CCI brain injury as previously described [26,27,28,29]. Briefly, the injury was induced using a 3 mm diameter rigid impactor driven by a pneumatic piston rigidly mounted at an angle of 20° from the vertical plane and applied perpendicularly to the exposed dura mater, between bregma and lambda, over the left parieto-temporal cortex (antero-posteriority: − 2.5 mm, laterality: left 2.5 mm), at an impactor velocity of 5 m/s and deformation depth 2 mm, resulting in severe injury [19, 30]. The craniectomy was then covered with a cranioplasty and the scalp sutured.
Control sham mice received identical anesthesia, and surgery, but no brain trauma. During all surgical procedures, mice were maintained at a body temperature of 37 °C.
Behavioral tests
Sensorimotor function
Sensorimotor deficits were evaluated by neuroscore, beam walk and Simple Neuroassessment of Asymmetric Impairment (SNAP) tests at specific time points (Fig. 1), as previously described [26,27,28,29, 31, 32].
Neuroscore
Mice were scored from 4 (normal) to 0 (severely impaired) for each of the following: (1) forelimb function while walking on the grid and flexion response during suspension by the tail; (2) hindlimb function while walking on the grid and extension during suspension by the tail; (3) resistance to lateral right and left pushes (best score 12) [26,27,28,29, 31].
Beam walking test
Mice were analysed for motor coordination and balance by measuring the number of foot faults of a trained mouse walking twice on an elevated, narrow wooden beam (5-mm wide, 100 cm long). Before each test, mice were trained in three habituation trials. Performance was scored as the sum of the number of foot faults during the two tests (best score 0) [32].
SNAP test
Neurological parameters were assessed in eight tests including vision, proprioception, motor strength and posture, as previously described [32]. The result obtained in each of the eight tests was summed to obtain the overall SNAP score. A neurologically intact animal is expected to have a SNAP score of 0.
Open field
Mice were placed in the center of a square arena with walls (40 × 40 × 30 cm) with the floor divided into 25 squares (8 × 8 cm). The nine central squares (24 × 24 cm) represent the “central area” and the surrounding border zone the “outer squares”. Mice were tested under dim illumination. Their behaviour was then video-recorded for 5 min (Ethovision XT 5.0, Noldus). The total distance moved (a measure of locomotor activity), and the time spent in the central area (a measure of thigmotaxis and indicative of anxiety-related behavior) were recorded [33].
Y maze
The Y maze is a two-trial spatial recognition memory test where performance does not involve learning a rule, but is based on the innate propensity of rodents to explore new environments (i.e. not encountered before). The Y maze apparatus consisted of three arms joined in the middle to form a “Y”. The insides of the arms are identical, providing no intra-maze cues. Visual cues were placed around the perimeter of the maze, and kept constant during the behavioral test. During the acquisition phase (trial 1), one arm of the Y maze was closed with a guillotine door. The position of the closed arm was chosen randomly among the three arms. Each mouse was placed in one of the other two open arms (“starting arm”), and allowed to visit the two accessible arms of the maze for 5 min. At the end of the trial, mice were placed in their home cage. After an inter-trial interval of 1 h, mice were placed in the same “starting arm” as in trial 1, with free access to all three arms for 5 min (trial 2, retrieval phase) [34, 35]. During each trial, the number and the duration of visits to each arm were recorded (Ethovision XT 5.0, Noldus).
Magnetic resonance imaging analysis
Animals were anesthetized with isoflurane in a mixture of O2 (30%) and N2O (70%). Body temperature was maintained at approximately 37 °C by a warm water-circulated heating cradle, and the respiratory rate was continuously monitored. Experiments were done on a 7 T Bruker Biospec 70/30 Avance III system, equipped with a 12 cm diameter gradient coil (400 mT/m maximum amplitude) (Bruker Biospin, Ettlingen, Germany). Due to a hardware upgrade between experimental sessions, two different coils setup had to be employed. A transmit cylindrical radiofrequency (rf) coil (7.2 cm inner diameter) for transmission and a quadrature receive surface rf (2 × 2 cm) coil array positioned over the animal’s head were used for cohort 1, while a quadrature cryogenic surface coil was used as transmitter and receiver for cohort 2.
In the same imaging session, two subsequent protocols were applied to monitor the structural alterations (2D T2-weighted sequence) and white matter damage (diffusion tensor images, DTI). A 2D T2-weighted SE RARE sequence of the whole brain was done. The images were obtained with a voxel size of 100 × 100 µm (matrix size 150 × 150 and field of view 1.5 × 1.5 cm), slice thickness 0.3 mm, repetition time [TR] = 7500 ms, effective echo time [TE] = 66 ms. MRI images were analysed to measure the volume of selected brain regions using the automatic multiatlas-based segmentation approach embedded in the ANTs software library [36, 37]. The brain regions of interest in the hemisphere contralateral to injury were co-registered on the mouse brain histology atlas of Franklin and Paxinos [38]. Every MRI section was visually inspected, and the brain regions manually corrected, if needed, to take account of the cavity in the TBI group. Measurements were taken using freely available ITK-SNAP software [39].
An echo-planar imaging (EPI) sequence of the whole brain was acquired with an in-plane image. The images were obtained with a voxel size of 125 × 125 µm (matrix size 120 × 120 and field of view 1.5 × 1.5 cm), slice thickness: 0.3 mm, repetition time [TR] = 7000 ms, effective echo time [TE] = 27.5 ms, t and 2 repetitions. Each repetition was co-registered and averaged to increase the signal-to-noise ratio (SNR) after correcting EPI distortions between repetitions. Diffusion, encoding b factors of 800 mm2/s is applied along 19 isotropic directions and two B0 unweighted images for each repetition. The diffusion tensor was computed using FSL software. A group mean full tensor template was first created using a population-based DTI atlas construction algorithm that adopts a tensor-based registration procedure embedded in the DTI-TK software library [40]. The average template was resampled to an in-plane resolution of 100 × 100 μm2 and a slice thickness of 0.2 mm, and is thinned to obtain a mean skeleton. DTI metrics (fractional anisotropy, FA; axial diffusivity, AD; radial diffusivity, RD; mean diffusivity, MD) were normalized to a mean template with a diffeomorphic transformation, then projected onto the mean skeleton for ROI-based analysis. The SNR within the regions of interest (ROIs) did not differ across experimental groups (sham: 26.03 ± 2.5; PTE: 27.5 ± 3.2; No-SRS: 27.2 ± 2.3), and it was higher than 20, thus allowing to perform reliable quantification of DTI metrics [41]. Reproducibility of the DTI measurements was also checked, and the coefficient of variation was below the 5% critical threshold (range: 2.6–3.8%) [42].
Detection and quantification of spontaneous seizures
At 3 months (cohort 1) or 5 months post-TBI (cohort 2) (Fig. 1), mice were surgically implanted under general gas anesthesia (1–3% isoflurane in O2) and stereotaxic guidance [43]. Four epidural screw electrodes were implanted in the skull, two ipsilaterally and two contralaterally to the injured hemisphere (Fig. 2a).
The electrodes were positioned from bregma, as follows (Fig. 2a): perilesional cortex (mm: nose bar 0; anteroposterior − 1, lateral left 2.8; black dot); frontal cortex (mm: nose bar 0; anteroposterior + 1.6, lateral left 1.8; yellow dot); parieto-occipital cortex (mm: nose bar 0; anteroposterior − 1 and − 3, lateral right 2.8; green and red dots, respectively). Two additional screw electrodes were positioned over the nasal sinus and the cerebellum, and used as ground and reference electrodes, respectively. Electrodes were connected to a multipin socket and secured to the skull with acrylic cement.
We reckoned the number and duration of spontaneous recurrent seizures (SRS) in electrode-implanted mice by continuous (24/7) video-ECoG monitoring for 3 weeks in each recording period. Seizure frequency was estimated by dividing the total number of seizures by the number of recording days.
Spontaneous seizures consist of ECoG paroxysmal events characterised by high frequency (> 5 Hz, usually 7–10 Hz) and/or multispike complexes and/or high amplitude (700 μV–1.0 mV vs 100–300 μV at baseline) spikes. ECoG activity was recorded using the Twin EEG Recording System (version 4.5.3.23) connected with a Comet AS-40 32/8 Amplifier (sampling rate 400 Hz, high-pass filter 0.3 Hz, low-pass filter 70 Hz, sensitivity 2000 mV/cm; Grass-Telefactor, West Warwick, R.I., USA). ECoG seizures were accompanied by generalized motor convulsions recorded using WFL-II/LED15W infrared video-cameras (Videor Technical, GmbH, Germany) synchronized with the ECoG recording system. Digitized video-ECoG data were processed using the Twin record and review software. ECoG analysis of seizures was done by two independent investigators blinded to the treatment, who reviewed all the ECoG tracings in the electronic files of each mouse. Deviation of ≤ 5% from concordance was considered acceptable; otherwise the ECoG tracing was analyzed by a third person to seek consensus. Mice were defined as epileptic (PTE) or non-epileptic (No-SRS) based on the presence or absence of spontaneous seizures at either 3 or 5 months, or both (for cohort 1) or at 5 months (for cohort 2).
Cortical screw electrodes were similarly implanted in 5 naïve mice and in 5 sham mice (with craniectomy only) to monitor the effect of electrode implantation and surgery on the ECoG signal. These mice were used as control for the histological analysis.
Brain tissue preparation for histological analysis
At the end of the experiment, TBI (n = 29) and control mice (n = 5 naïve and n = 5 sham) were deeply anesthetized (10% ketamine + 10% medetomidine + 80% saline; 10 ml/kg, i.p.) and perfused intracardially with ice-cold phosphate buffered saline (PBS, pH 7.4) followed by 4% paraformaldehyde (PAF) in PBS. Brains were removed from the skull and post-fixed for 90 min in 4% PAF in PBS at 4 °C, transferred to 20% sucrose in PBS for 24 h at 4 °C, then frozen in n-pentane for 3 min at − 50 °C, and stored at − 80 °C until assay. Coronal brain sections (30 μm) were cut on a cryostat from 0.98 mm to − 3.88 mm from bregma [43].
Nissl staining
Three antero-posterior levels (0.38 mm, − 1.70 mm and − 3.08 mm from bregma) were examined. Quantitative analysis was done in the striatum, thalamus (dorsal and ventral nuclei), septal pole of the hippocampus, and perilesional and entorhinal cortices. The ipsilateral cortex was analysed across two different antero-posterior levels (− 1.70 mm and − 3.08 mm from bregma), over a region 600 µm thick from the edge of the contusion. Striatum was quantified bilaterally, and thalamus, hippocampus and the entorhinal cortex were analysed only contralaterally, because of the brain cavity in the ipsilateral hemisphere of TBI mice. Two Nissl-stained slices per brain area 360 µm apart were matched for antero-posterior location in the various experimental group, and used for quantitative analysis.
Cell loss was measured as previously described [44, 45]. Briefly, images of the whole-brain coronal sections were captured at 20 × magnification using a Virtual Slide scanning microscopy system (Olympus, Germany) and digitized. In the hippocampus, neuronal cell loss was quantified by manually counting the number of Nissl-stained neurons in CA1 and CA3/CA4 pyramidal cell layers, and the hilar interneurons. For the remaining brain regions, images were processed using Fiji software [46]: an algorithm was created to segment and analyse every Nissl-stained cell over the entire manually selected ROI, as previously described [10, 28, 45]. Briefly, images were scaled into microns, and the background subtracted. Then an optimised threshold selected in a pilot study was applied across all the experimental groups to identify the Nissl-stained positive area, and the images were binarized. In order to unequivocally distinguish Nissl-stained neurons from glial cells and count only neuronal cells, we excluded from the counting those cells with an area below the cut-off of 25 µm2, using Fiji software [10, 28, 45]. Once segmented, all cells positive for Nissl were automatically quantified.
Data for each of the 2 slices/brain area/mouse were averaged, giving a single value for each brain area/mouse, and this value was used for statistical analysis.
Randomization, blinding procedures and statistical analysis of data
A simple random allocation was applied to assign a subject to an experimental group (TBI or control). All evaluations were done blind to the sample identity.
Statistical analysis was done with GraphPad Prism 6 (GraphPad Software, USA) for Windows, using absolute values. The choice between parametric or non-parametric tests was based on passing the Shapiro–Wilk normality test, and data distribution was inspected by QQ plot. For each experiment, the figure legend reports the statistical analysis of data. We calculated the effect size (d) for all statistical comparisons, and we report only the exact d value associated to a large effect size (d ≥ 0.8) [47]. Data for TBI mice are presented as bargrams depicting the mean ± SEM and the single values (n = number of individual samples).
Differences between groups were reported as statistically significant for p < 0.05.
We assessed the performance of the measure(s) to distinguish PTE from No-SRS mice using non-parametric Receiver Operating Characteristics (ROC) curves: the area under the curve (AUC) was calculated and compared with chance (AUC = 0.5). The performance of the measure(s) was considered excellent for AUC values close to 1.
Mice omitted from data analysis are described in the figure legends, where appropriate.