Epilepsy Phenotype and Its Reproducibility After Lateral Fluid-Percussion -Induced Traumatic Brain Injury In Rats – A Multicenter EpiBioS4Rx Study Project 1

Abstract

Objective:

To assess reproducibility of the epilepsy outcome and phenotype in lateral fluid-percussion model of post-traumatic epilepsy across three study sites.

Methods:

A total of 525 adult male Sprague-Dawley rats were randomized to lateral fluid-percussion -induced brain injury (FPI) or sham-operation. Of these, 264 were assigned to magnetic resonance imaging (MRI cohort, 43 sham, 221 TBI) and 261 for electrophysiological follow-up (EEG cohort, 41 sham, 220 TBI). A major effort was made to harmonize the rats, materials, equipment, procedures and monitoring systems. On the 7^th post-TBI month, rats were video-EEG monitored for for epilepsy diagnosis.

Results:

A total of 245 rats were video-EEG phenotyped for epilepsy on the 7^th post-injury month (121 in MRI cohort, 124 in EEG cohort). In the whole cohort (n=245), the prevalence of PTE in rats with TBI was 22%, being 27% in the MRI and 18% in the EEG cohort (p>0.05). Prevalence of PTE did not differ between the three study sites (p>0.05). The average seizure frequency was 0.317 ± 0.725 seizures/d in UEF (Finland), 0.085 ± 0.067 in Monash (Australia) and 0.299 ± 0.266 in UCLA (USA)(p<0.01 as compared to Monash). The average seizure duration did not differ between the UEF (104 ± 48 s), Monash (90 ± 33 s) and UCLA (105 ± 473 s)(p>0.05). Of the 219 seizures, 53% occurred as part of a seizure cluster (≥3 seizures/24 h)(p<0.05 between the study sites). Of the 209 seizures, 56% occurred during lights-on period and 44% during lights-off period (p>0.05 between the study sites).

Significance:

The PTE phenotype induced by lateral FPI is reproducible in a multi-center design. Our study supports the feasibility of performing pre-clinical multicenter trials in PTE to increase statistical power and experimental rigor to produce clinically translatable data to combat epileptogenesis after TBI.

1 |. INTRODUCTION

Traumatic brain injury (TBI) is defined as an alteration in brain function, or other evidence of brain pathology, caused by an external force¹. Annually, about 2.5 million people both in Europe and USA experience TBI (www.center-tbi.eu; www.cdc.gov/traumaticbraininjury). TBI is a major etiology for structural epilepsy in humans, with post-traumatic epilepsy (PTE) estimated to account for approximately 20% of structural epilepsies and 5% of all epilepsies^2–4. Despite of about 20 promising pre-clinical proof-of-concept antiepileptogenic or biomarker discovery studies in animal models of PTE, none of the interventions or candidate biomarkers has progressed to a clinical use⁵.

One of the major obstacles on the path from laboratory to clinic is the low reproducibility of preclinical studies⁶. This often relates to a low sample size, and consequently, an underpowered study⁷. A potential solution to this problem is multisite preclinical randomized controlled trials (pRCT), but for this to be successful there is a need for harmonization of data collection with the use of common data elements and standardization of procedures to achieve a greater methodological rigor^8,9. The first centrally coordinated, randomized and blinded pRCT tested the effect of anti-CD49d antibodies on infarct volume in two stroke models in six European centers¹⁰. Efforts to conduct preclinical multi-center biomarker and therapy studies in TBI were reported by Operation Brain Trauma Therapy network (OBTT)¹¹. More recently, methods optimization for preclinical stroke intervention studies have been reported by Italian Stroke Organization (ISO) Basic Science Network¹² and the Stroke Preclinical Assessment Network (SPAN) investigators¹³. These projects have demonstrated promising data, supporting the feasibility and benefits of conducting pre-clinical multicenter trials.

The Epilepsy Bioinformatics Study for Antiepileptogenic Therapy (EpiBioS4Rx) consortium, a NINDS funded Center-Without-Walls international study is the first to apply preclinical multi-center design in antiepileptogenesis biomarker identification and therapy development (https://epibios.loni.usc.edu/). Project 1 of the EpiBioS4Rx intends to facilitate the development of antiepileptogenic therapies after traumatic brain injury (TBI) through the discovery of preclinical blood, electrophysiologic and imaging biomarkers of epileptogenesis, using harmonized protocols and statistically powered study designs. The EpiBioS4Rx Project 1 was performed in three international sites located in Finland (University of Eastern Finland), Australia (Monash University, Melbourne University) and USA (David Geffen School of Medicine at UCLA).

The first objective of EpiBioS4Rx Project 1 was to demonstrate that the three study sites can perform lateral FPI-induced TBI, resulting in PTE with similar prevalence and epilepsy phenotype. At all study sites, a major effort was made to harmonize the materials, equipment and rats used as well as the procedures and monitoring systems, including blood sampling, video-electroencephalogram (video-EEG) and magnetic resonance imaging (MRI). Our data show that (a) post-traumatic epileptogenesis and epilepsy phenotype induced by lateral fluid-percussion injury (FPI) -induced TBI in adult male rats is highly reproducible, tolerating some site-specific procedural differences and (b) application of multi-center approach is feasible for preclinical studies on epileptogenesis after TBI.

2 |. MATERIALS AND METHODS

Figure 1 summarizes the study design that was applied by all three study sites [University of Eastern Finland (UEF), Kuopio, Finland; Monash University (Monash), Melbourne, Australia; UCLA, Los Angeles, USA]. Table 1 summarizes the information on animals, feeding, housing, surgery and EEG recording systems, data acquisition and data analysis at different study sites. Figure S1 shows the randomization based on preliminary data^14–16, study flow and causes of exclusions. Detailed methodologic descriptions, including ethics and statistics are given in Appendix S1.

Figure 1. Study design.

EpiBioS4Rx Project 1 consisted of two separate animal cohorts: the magnetic resonance imaging (MRI) cohort and video-electroencephalography (video-EEG) cohort. They were generated and monitored by using the same protocols at the three study sites (UEF, Monash, UCLA. In both cohorts, the rats were randomized either to the sham-operated experimental control or TBI groups. Epileptogenesis was triggered using lateral fluid-percussion-induced (FPI) traumatic brain injury (TBI). During the first follow-up month (M1), rats underwent an extensive physiologic and somatomotor monitoring to assess animal well-being and injury severity at different study sites. These included body weight and core temperature (UEF only) at baseline (BL) and on day (D)1, D2, D3, D4, D5, D6, D7, D8, D9, D14 and D30 after TBI (injury day D0) and composite neuroscore at BL and on D2, D7, D14, D21 and D28. For blood biomarker analysis, tail vein blood sampling was performed in both cohorts at BL and on D2 (48 h), D9, D30 and 5 months post-TBI. In the MRI cohort, rats were MR imaged on day D2 (48 h), D9, D30 and 5 months post-TBI using T2-weighted imaging (T2-w), multi-echo gradient echo (MGE) imaging, diffusion-weighted imaging (DWI) for diffusion tensor imaging (DTI) and tractography and magnetization transfer imaging (MT). Then, the rats were implanted with epidural and intracerebral electrodes for video-electroencephalography (video-EEG). In the EEG cohort, electrodes were implanted right after TBI and rats were recorded during the 1^st week, and then, monthly with a 1-wk-long high-density (HD) EEG for detecting high-frequency oscillations, starting right after the TBI or sham-operation till the 6^th month. During the 7^th month, both the MRI and EEG cohorts were continuously (24/7) video-EEG monitored for 30 d to detect spontaneous seizures to diagnose post-traumatic epilepsy (no video was recorded in UCLA). Finally at 7 months after TBI or sham-operation, rats were perfused for ex vivo MRI (MGE, DWI and MT modalities), after which the brains were processed for histologic analysis.

Table 1.

Rats, material and equipments used for model production and video-EEG analysis at the three study sites

	UEF	Monash	UCLA
Animals and housing
Number of rats randomized	184	194	147
Strain and species	Spraque-Dawley rats	Spraque-Dawley rats	Spraque-Dawley rats
Vendor (Country)	Envigo Laboratories B. V. (The Netherlands)	In-house breeding (Australia)	Charles River (USA)
Sex	Male	Male	Male
Food pellets	2016S (Teklan Diet) (Envigo Laboratories B.V., The Netherlands)	102108 (Barastoc, Australia)	LabDiet 5001* (LabDiet, St. Loius, MO, USA)
Duration of quarantine	7 days	3-7 days	At least 3 days
Lights-on/lights-off cycle	7.00 am light-on/7.00 pm lights-off	7.00 am light-on/7.00 pm lights-off at Monash University and 6.00 am light-on/6.00 pm lights-off at the Melbourne University	6:00 am light-on/6:00 pm lights-off
Room temperature (C)	22 ± 1 °C	22 ± 1 °C	20-26 °C
Surgery
Weight at the time of injury	354 ± 18 g (range 315 – 408)	349 ± 39 g (range 250 – 440 g)	336 ± 41 g (range 260 – 497 g)
Anesthesia system	Somnosuite #SS6069B (Kent Scientific)	Somnosuite #SS6069B (Kent Scientific)	Matrix VIP 3000 Vaporizer #91305430 (Patterson Veterinary)
Anesthetic	Isoflurane	Isoflurane (5% induction & 2% maintenance)	Isoflurane, USP (VET one)
Medical Oxygen	Not applicable	Mediquip Medical Equipment & Supplies, Australia	Not applicable
Trephine	#18004-50 (hand-held) (Fine Science Tools GmbH, Germany)	Model 300 (motorized) (Dremmel, Australia)	#18004-50 (hand-held) (Fine Science Tools GmbH, Germany)
Tissue adhesive	3M Vetbond (3M Deutschland GmbH, Germany)	Octyl cyanoacrylate (N/A) (Bostik, Australia)	3M Vetbond (3M Deutschland GmbH, Germany)
Dental acrylate	Selectaplus #10009210 or #D10009102 (DeguDent, Germany)	AVSCV00500 (Vertex, The Netherlands)	SNAP Liquid (P16-02-65) and Powder (P16-02-60) (Pearson Dental) (Pearson Dental, USA)
Fluid-percussion device	Model FP 302 (AmScien Instruments, USA)	Model FP 302 (AmScien Instruments, USA)	Model FP 302 (AmScien Instruments, USA)
Analgesia	Buprenorphine (Orion Pharma, Finland)	Buprenorphine (Indivior Pty Ltd, Australia)	Flunixin meglumine (MERK, USA)
Other treatments	None	None	Enrofloxacin (Norbrook)
Additional feeding	Powdered pellets	Milk powder, mixed with powdered pellet and water provided ad libitum, until rats recovered their pre-injury weight	Trimethoprim sulfamethoxazole (TMS) medicated rodent chow
Housing after impact surgery	Single-housed	Single-housing	Single-housed
EEG recording system, materials and data acquisition
Head cap	12-channel pedestal (MS12P EM12/20/2.4/ SP) (PlasticsOne Inc.)	12-channel pedestal (MS12 P EM12/20/2.4/SP (PlasticsOne Inc.)	6-channel pedestal (MS363) (PlasticsOne Inc.)
Cable	Flexible shielded cable (M12C-363/2) (PlasticsOne Inc.)	Flexible shielded cable (M12C-363/2) (PlasticsOne Inc.)	Flexible shielded cable (363/2-363/2) (PlasticsOne Inc.)
Commutator	12-channel double-brush (SL12C, (PlasticsOne Inc.)	12-channel double-brush (SL12C, 12-pin swivel) (PlasticsOne Inc.)	12-channel double-brush (SL-12C, 12-pin swivel) (PlasticsOne Inc.)
Epidural electrodes	E363/20/2.4/Spc (Stainless-Steel) (PlasticsOne Inc)	363/120/2.4 ELEC W/000-120 X 2.4 SCREW or E363/20/SPC ELEC W/3.2MM SCREW SS (PlasticsOne Inc)	Stainless steel epidural screw (F000CE096, J.I. MORRIS CO)
Intracerebral electrodes	EM12/3-2TW/Spc (tungsten, 100 μm) E363T/5-2TW/Spc (tungsten, 50 μm)	E363T/5-2TW ELEC (tungsten, 50 μm)	Bipolar intracerebral tungsten .002” twisted electrode
Electrode impedance	Initially below 5KΩ. Maintained below 10KΩ	Below 10KΩ (measured at 1000 Hz)	Female Socket-Contact E363/0 (PlasticsOne Inc.)- Below 5KΩ
Amplifier model	Digital Lynx 16SX (Neuralynx, USA)	Neuvo	Intan RHD2000
Acquisition software	Cheetah ver. 6.3.2	Profusion EEG 5	RHD2000
Sampling rate	5 kHz	2 kHz/channel	2 kHz/channel
Filter settings	FIR high-pass 0.1Hz	0.01-2030 Hz	0.1-1000 Hz
EEG file format	.NCS converted to EDF+	.rda2 converted to EDF+	.RHD converted to EDF+
EEG file duration	24 h	24 h	2 h
EEG file size	~18 GB	~3.8 GB	~1.2 GB
Video system and materials
Camera	acA1300-75gc - Basler ace	IP Camera (Vivotek, Australia)	not applicable
Video acquisition software	inhouse software	Profusion 5	not applicable
Frames per second	30fps	30fps	not applicable
Resolution	1.3 MP (1280 px x 1024 px)	2 MP	not applicable
Analysis of video-EEG
Seizure detection in EEG	Home-made seizure detection algorithm 1(Andrade et al., 2018) followed by visual validation of positive hits.	Either manual seizure detection or Assyst algorithm for automated seizure detection followed by visual validation of positive hits2. (Casillas Espinosa et al., 2019).	Manual seizure detection using EDF browser software
Assessment of behavioral seizure severity	Racine scoring of video imaged seizures	Racine scoring of video imaged seizures	not applicable

^1.Andrade P, Paananen T, Ciszek R, Lapinlampi N, Pitkänen A. Algorithm for automatic detection of spontaneous seizures in rats with post-traumatic epilepsy. J Neurosci Methods [Internet]. 2018; 307:37–45. Available from: https://linkinghub.elsevier.com/retrieve/pii/S0165027018301948

^2.Casillas-Espinosa PM, Sargsyan A, Melkonian D, O’Brien TJ. A universal automated tool for reliable detection of seizures in rodent models of acquired and genetic epilepsy. Epilepsia [Internet]. 2019; 60(4):783–91. Available from: http://www.ncbi.nlm.nih.gov/pubmed/30866062

3 |. RESULTS

Altogether 245 animals were EEG-phenotyped and included in the final analysis cohort (Figure S1). That is, 121 rats in the MRI cohort and 124 in the EEG cohort. Of these, 100 rats were generated in UEF, 84 in Monash and 61 in UCLA. We will first describe the injury characteristics, and then the epilepsy characteristics.

3.1 |. Impact pressure, post-injury apnea and duration of righting reflex

Below we will focus the site differences. Impact severity and its effects on apnea time and righting reflex in the EEG-phenotyped MRI and EEG cohorts as well as in the animals that did (TBI+) or did not (TBI-) develop epilepsy at different study sites are summarized in Appendix S1 and Table S1.

3.1.1 |. Impact pressure

Data was available from 95% (178/187) of the rats with TBI (Table S1). The average impact pressure used was 2.65 ± 0.37 atm (n=178). The impact pressure varied between the sites (p<0.001). In UEF, the impact pressure used (2.88 ± 0.16 atm, n=75) was higher than that in Monash (2.67 ± 0.37 atm, range, n=59)(adj. p<0.01) or in UCLA (2.20 ± 0.20 atm, n=44)(adj. p<0.001). In Monash, the impact pressure was higher than that in UCLA (adj. p<0.001).

3.1.2 |. Post-impact apnea

Data was available from 100% (187/187) of the rats with TBI (Table S1). The average duration of post-impact apnea was 41 ± 38 sec (n=187). The mean post-impact apnea varied between the sites (p<0.001). In UEF, the apnea duration (28 ± 14 s, n=75) was shorter than that in Monash (67 ± 51 s, n=68)(adj. p<0.001). Also, in UCLA, the apnea duration (22 ± 11 s, n=44) was shorter than that in Monash (adj. p<0.001), and also shorter than that in UEF (adj. p<0.05).

3.1.3 |. Righting reflex

Data was available from 60% (35/58) of the sham-operated experimental controls and 99% (186/187) of the rats with TBI (Table S1). No data was available from sham-operated animals in UCLA.

The average duration of righting reflex in the TBI group (1 057 ± 512 s, n=186) was substantially longer than that in the sham-operated group (187 ± 100 s, n=35)(p<0.001). In the sham-group, the average duration of righting reflex did not differ between the UEF and Monash (178 ± 102 s vs. 200 ± 98 s; p>0.05). In the TBI group, the average duration of righting reflex varied between the sites (p<0.001). In UEF, the righting time (879 ± 338 s, n=74) was shorter than that in Monash (1 347 ± 581 s, n=68) (adj. p<0.001). Also in UCLA, the righting time (910 ± 447 s, n=44) was shorter than that in Monash (adj. p<0.001). In UEF and UCLA, the righting times did not differ (adj. p>0.05).

3.2 |. Duration of video-EEG recording on the 7^th post-injury month

Duration of the index month video-EEG recording at different study sites is summarized in Figure 2C and Table S2. Our objective was to record each animal for 30 d, which was anticipated to give a 99.6% probability to detect a seizure, if the animal had epilepsy with the expected seizure frequency of 0.2 sz/d¹⁷.

Figure 2. Epilepsy phenotyping.

(A) Electrode placement. Location of the craniotomy (light pink circle), 4 epidural recording electrodes (blue; C3, C4, O1, O2), 3 bipolar intracerebral electrodes [anterior cortical A1 (deeper tip) and A3 (superficial tip); hippocampal (H1 and H3), posterior cortical (Po1 and Po3), reference (purple; Ref), and ground (green; G) electrodes in the skull. Grid dimensions 1 mm x 1 mm. (B) Coronal plates were adapted from the rat brain atlas (Paxinos and Watson, 2007), showing the aimed locations of the intracerebral electrode tips (AP, antero-posterior; ML, medio-lateral; DV, dorso-ventral). (C) Box plots show the duration of video-EEG monitoring used to diagnose post-traumatic epilepsy in the EEG and MRI cohorts at different sites. Note that no video-monitoring was performed in UCLA. The green dashed line shows the aimed duration of the monitoring, that is, 30 d. In 73% (180/245) of the animals, the monitoring lasted at least for 30 d. Details of video-EEG monitoring of different cohorts at different sites is summarized in Table S2. Note that some rats presented status epilepticus as the 1^st seizure and died, resulting in a short monitoring period (UEF MRI cohort).

The average duration of “index month” (i.e., the 7^th month) recording to detect unprovoked (spontaneous) diagnostic seizures for epilepsy was 39 ± 26 d (median 34 d), ranging from 2 to 176 d. In UEF-MRI cohort one rat was recorded for 2 d and another for 14 d. The rats had a seizure cluster or SE as the “first” seizure, respectively, leading to death. In the UEF-EEG cohort, the video-EEG recording times were up to 3 months instead of 30 d, due to need of additional recordings related to pogo-pin malfunctions. Overall, in 73% (180/245) animals the index-month EEG recording lasted for at least 30 d [92% (92/100) in UEF, 94% (79/84) in Monash, 15% (9/61) in UCLA]. In the whole animal group as well as at each study site, the duration of index month EEG recordings did not differ between the TBI+ and TBI− groups.

3.3 |. Prevalence of epilepsy

An animal was considered to have PTE, if it had at least one unprovoked seizure detected during the 7^th month video-EEG recording, or at least one handling-related seizure observed by an experienced investigator¹⁸. Representative examples of electrographic seizures at each study site are shown in Figure S2.

In the whole TBI animal cohort, the prevalence of epilepsy was 22% (41/187), being 27% (24/90) in the MRI and 18% (17/97) in the EEG cohort (p>0.05)(Figure 3). There was no difference in the prevalence of epilepsy between the three study sites (p>0.05). When each site was assessed separately, the prevalence of epilepsy did not differ between the MRI and EEG cohorts (p>0.05). In UEF, 1 rat with video-EEG recorded seizures also had a handling-related seizure. In Monash and UCLA, all seizures detected were electrographic.

Figure 3. Prevalence of epilepsy.

Pie-charts summarizing the prevalence of epilepsy. (A) In the final analysis cohort (all study sites combined) 22% (41/187) of the rats had epilepsy. In the MRI cohort the prevalence was 27% (24/90) and in the EEG cohort 18% (17/97). (B) Prevalence of epilepsy at different study sites (UEF. Monash, UCLA). The pie-charts at the bottom show the prevalence of epilepsy in the MRI and EEG cohorts at each study site. Prevalence of epilepsy did not differ between the study sites and between the MRI and EEG cohorts. Abbreviations: EEG, electroencephalogram; MRI, magnetic resonance imaging.

3.4 |. Epilepsy phenotype

A seizure calendar showing the occurrence of all 219 unprovoked electrographic seizures recorded in UEF, Monash or UCLA is presented in Figure 4. Seizures occurred as isolated events or as part of a seizure cluster or SE.

Figure 4. Seizure calendar.

(A) Number of electrographic seizures in the MRI and EEG cohorts and in both cohorts combined at different study sites. Animal numbers are shown in parenthesis. Altogether, 219 seizures were found in 41 rats. Of these 112 were in rats in the MRI and 107 in rats in the EEG cohort. (B) Occurrence of seizures during the long-term video-EEG monitoring that was aimed to start on the 7^th post-TBI month (“index month”, blue shading). Animal IDs are shown on the y-axis. The time from injury is shown on the x-axis. A dashed line separates the MRI and EEG cohorts. Animals from different study sites are shown with different colors (UEF blue; Monash orange; UCLA green). The duration of monitoring varied between 2 to 176 d (Table S1). Purple arrows indicate rats in the continuously monitored EEG cohort that had seizures already during the 2^nd and 5^th post-TBI month. Green arrows indicate animals with seizure clusters (≥ 3 seizures/24 h).

Seizure frequency.

The average seizure frequency (seizures/day) in the whole TBI+ cohort was 0.228 ± 0.476 sz/d (41 rats, median 0.106)(Figure 5A–B). In UEF the average seizure frequency was 0.317 ± 0.725 sz/d (16 rats, median 0.105), in Monash 0.085 ± 0.067 sz/d (15 rats, median 0.065, n=15) and in UCLA 0.299 ± 0.266 sz/d (10 rats, median 0.188). Kruskal-Wallis indicated a difference between the sites (p<0.05), the seizure frequency being lower in Monash than in UCLA (adj. p <0.01).

Figure 5. Seizure characteristics.

(A) Violin plots showing the average seizure frequency (seizures/24 h) in each animal in the EEG and MRI cohorts at different study sites. Each dot refers to one animal. No differences were found between the EEG and MRI cohorts, when all animals from different study sites were combined or different sites were analyzed separately (all p>0.05). Number of animals is in parenthesis. (B) Distribution of average seizure frequencies (seizures/monitoring days) in 41 rats with PTE. The median seizure frequency (dashed line) was 0.106 seizures/d. (C) Duration of individual seizures in the EEG and MRI cohorts at different study sites. No differences were found between the EEG and MRI cohorts, when all animals from different study sites were combined or different sites were analyzed separately (all p>0.05). Number of seizures is in parenthesis. (D) Distribution of the duration of 208 electrographic seizures. The median seizure duration was 96 s. (E) Average seizure duration in each animal in the EEG and MRI cohorts at different study sites. No differences were found between the EEG and MRI cohorts, when all animals from different study sites were combined or different sites were analyzed separately (all p>0.05). (F) Distribution of average seizure duration in 40 rats with PTE. (G) Average Racine score was greater in the EEG (60 seizures) than MRI (68 seizures) cohort (4.2 ± 1.6 vs. 2.6 ± 1.9; Mann-Whitney U). When both sites were analyzed separately, no differences between the cohorts were found (p>0.05). Number of scored seizures is in parenthesis. (H) Distribution of average Racine score in 30 rats with PTE. The median Racine score was 3.2.

Seizure duration.

The seizure duration was available from 95% (208/219) of the seizures. The average seizure duration was 100 ± 53 s (208 seizures, median 96 s, range 13-390 s)(Figure 5C–D). The average seizure duration did not differ between the UEF (102 seizures, 102 ± 43 s, median 96 s, range 13-242 s), Monash (44 seizures, 87 ± 41 s, median 82 s, range 23-180 s) and UCLA (62 seizures, 108 ± 73 s, median 101 s, range 15-390)(p>0.05).

The average seizure duration per rat was 100 ± 50 s (40 rats, median 95 s, range 21-254 s)(Figure 5E–F). The average seizure duration did not differ between the UEF (15 rats, 104 ± 48 s, median 97 s, range 21-180 s), Monash (15 rats, 76 ± 37 s, median 73 s, range 25-135 s) and UCLA (10 rats, 129 ± 58 s, median 132 s, range 57-254)(p>0.05).

Behavioral severity of electrographic seizures.

Racine score was available from 58% (128/219) of the seizures. No videos were available in UCLA for Racine scoring. The average Racine score was 3.34 ± 1.94 (median 5.00, 128 seizures). Of the 128 seizures, 9% (11/128) had a score 0, 22% (28/128) score 1, 7% (9/128) score 2, 3% (4/128) score 3, 9% (11/128) score 4 and 51% (65/128) score 5. The average Racine score of all seizures was greater in UEF (3.69 ± 1.81, median 5.00, 89 seizures) as compared to that in Monash (2.54 ± 2.00, median 1.00, 39 seizures)(p<0.001).

The average Racine score per rat was 2.72 ± 2.04 (median 3.20, 30 rats)(Figure 5G–H). The average Racine score per rat did not differ between UEF (3.15 ± 1.99, median 3.85, 16 rats) and Monash (2.24 ± 2.06, median 1.00, 14 rats)(p>0.05).

Seizure clusters.

Of the 219 seizures, 53% (117/219) occurred as part of a seizure cluster (≥3 seizures/24 h) and 9% (19/219) as part of SE (Figure S3). The remaining 38% (83/219) of the seizures occurred more isolated. As shown in Figure 4B, 42% (10/24) of the rats with epilepsy in the MRI cohort and 29% (5/17) in the EEG cohort had one or more seizure clusters (p>0.05). There were no site differences in the occurrence of seizure clusters (p>0.05).

Seizure occurrence during lights-on/lights-off periods.

Of the 209 seizures with data available, 56% (116/209) occurred during lights-on period and 44% (93/209) during lights-off period (p>0.05)(Figure S4). There were no differences between the EEG and MRI cohorts (p>0.05). Also, there were no site differences when data from the EEG or the MRI cohorts were analyzed separately (p>0.05).

Seizure onset electrode.

We were able to locate the seizure onset electrode in 39% (86/219) of the seizures (Figure S5). Of the 86 seizures, 97% (83/86) originated ipsilaterally. In 86% (74/86) of seizures, the onset was ipsilateral neocortical and in 10% (9/86) ipsilateral hippocampal. Of the neocortical seizures with ipsilateral onset, 54% (40/74) originated caudal to craniotomy in the epidural O1 or intracortical Po1/Po3 electrodes.

Seizure propagation.

Seizures were considered fast-propagating, if they spread from the seizure onset electrode (if it could be determined) to another (either ipsilateral or contralateral) within 3 s. Three propagation patterns were categorized: generalized (no focal onset could be defined), fast propagation and slow propagation. We were able to define seizure propagation in 79% (174/219) of the seizures (Figure S6). Of the 174 seizures, 66% (114/174) were fast propagating, 17% (29/174) slow propagating and 18% (31/174) generalized.

3.5 |. MRI

Figure S7 shows the coronal in vivo T2-weighted MRIs of each rat with epilepsy in the MRI and coronal ex vivo T2*-weighted MGE MRIs in the EEG cohort. In each case the impact-related lesion was visible and its epicenter was located in the auditory association cortex.

4 |. DISCUSSION

The major objective of EpiBioS4Rx Project 1 is to identify translational diagnostic and prognostic plasma, EEG and MRI biomarkers for post-traumatic epileptogenesis in a rat model of PTE induced with lateral FPI. To achieve adequate statistical power to discover a biomarker with AUC 0.700, the project was performed in a multi-center design involving three centers – UEF in Finland, Monash in Australia and UCLA in the USA. Our first task was to assess whether the lateral FPI -induced model of PTE was similarly produced at all three study sites. Specifically, the following questions were addressed: is there a difference in the (a) prevalence of epilepsy or (b) epilepsy phenotype between the centers and (c) do the procedural differences at different study sites affect epilepsy outcome? As the data show, the lateral FPI rat model of PTE is highly reproducible. The prevalence and epilepsy phenotype did not differ between centers despite of some procedural differences.

4.1 |. Prevalence of PTE did not differ between the sites or between the MRI and EEG cohorts

In the present study, 22% of the rats with lateral FPI -induced severe TBI developed epilepsy, that is, had at least one spontaneous seizure during the 7^th month video-EEG recording, in accordance with the ILAE definition of PTE¹⁹. Previous studies by the UEF team demonstrated a 11% (7/65) epilepsy rate after 3–4 months, a 22% to 29% epilepsy rate after 6–7 months and a 46% epilepsy rate after 12 months in adult male Sprague-Dawley rats who had experienced a lateral FPI^14,20,21. Accordingly, the Monash team had previously reported a 30% (7/23) PTE rate after 6 months post-lateral FPI in adult male Wistar rats¹⁵. The UCLA team had previously reported a 25% (4/12) epilepsy rate in adult male Sprague-Dawley rats EEG-monitored for up to 4 months²². Even though the cohort sizes and video-EEG monitoring periods in previous studies have been variable, affecting the sensitivity of detecting seizures, occurring in a low frequency and in a low proportion of animals, the epilepsy prevalence has been remarkably similar between the studies at 6 months post-TBI timepoint. Importantly, 68% (28/41) of the rats with two seizures also had the 3rd seizure within the month, justifying epilepsy diagnosis after the 1st unprovoked seizure¹⁸.

Did isoflurane anesthesia affect post-traumatic epileptogenesis? In the MRI cohort, rats underwent five long procedures under anesthesia, including a surgery-related, four MRI-related and an electrode implantation-related anesthesia. By contrast, in the EEG cohort, rats were exposed to one long anesthesia session that involved the surgery for injury followed by electrode implantation. In addition, both cohorts were exposed to 4 shorter sedations during post-TBI tail vein blood sampling. Interestingly, the epilepsy rate in the whole MRI cohort was 27% and in the EEG cohort 18%, suggesting no anti-epileptogenic effect of repeated isoflurane anesthesia. This is in correlation with our previous study, showing no effect of MRI-related isoflurane anesthesia duration on the prevalence of PTE²³. Also, our data show that the presence of chronically implanted intracerebral and epidural electrodes did not increase the prevalence of epilepsy.

The prevalence of epilepsy tended to be higher in the MRI than the EEG cohort both in UEF (28% vs. 16%) and Monash (29% vs. 16%), whereas in UCLA the epilepsy rate did not differ (22% vs. 24%) between the cohorts. This can reflect the “normal” variability in epilepsy prevalence after lateral FPI, when analysis is done by comparing smaller sub-cohorts. However, the tendency to a lower epilepsy prevalence in the UCLA MRI cohort can also relate to a shorter 7^th month EEG monitoring time as compared to that in UEF or Monash.

Epileptogenesis is a “moving target” with an increase in prevalence over time. Consequently, its detection rate is time dependent. Our data emphasize the need of standardization of the inter-site and inter-cohort video-EEG follow-up timing and duration to optimize the detection of PTE at a given follow-up point without recording bias. As we will summarize in paragraph 4.4., the unforeseen procedural obstacles can challenge the standardization efforts. Availability of early prognostic biomarkers for PTE will likely reduce the need of a long-term monitoring, and overall, reduce the proportion of false negative cases in the study cohort at a given analysis time period.

4.2 |. Epilepsy phenotype did not differ between the sites or between the MRI and EEG cohorts

At all study sites, a large majority of the seizures had a perilesional cortical onset. In UEF, 24% of the seizures had an ipsilateral hippocampal onset, which was not detected in Monash or UCLA. One explanation is that in UEF all TBI animals were monitored with a 12-channel montage, including a hippocampal electrode. According to histology, 93% (70/75) of the electrode tips in UEF rats with TBI were in the hippocampus proper or the dentate gyrus²⁴. However, in Monash, only 76% (52/68) of the TBI animals were recorded with hippocampal electrodes, six of which required re-implantation with epidural electrodes only. Consequently, only in 44% (30/68) of the TBI cases, ex vivo MRI could confirm the location of the electrode tip in the hippocampus or the dentate gyrus. In UCLA, all animals were recorded with a 12-channel montage, but due to a high re-implantation rate (16/44) and missing ex vivo MRIs (6/44), a correct positioning of the hippocampal electrodes could be confirmed in only 18% (8/44) of the TBI cases in ex vivo MRI. It should be noted that a correct hippocampal electrode location found in ex vivo MRI could be validated in histology in 87% of the cases²⁴. Taken together, the lack of hippocampal seizure onset in Monash and UCLA could relate to a lower number of animals recorded with hippocampal electrodes.

It is important to note that the seizure onset could be defined only in 40% of the seizures recorded. The analysis was challenged by a limited number of recording electrodes as well as by bad electrode connections and poor EEG quality at the early phase of the study. Also, many unprovoked seizures appeared generalized or showed a very fast propagation. Importantly, the seizure frequency, seizure duration and behavioral seizure severity did not differ between the three study sites. Of the seizures, 53% were part of the seizure cluster, which were found at all sites. As reported previously, seizures occurred quite equally during light-on and lights-off periods¹⁴.

There were no phenotypic differences between the MRI and EEG cohorts despite of rather different study designs. In Monash, however, none of the 12 seizures recorded in the EEG cohort occurred within a seizure cluster.

Analysis of the seizure-related behavioral symptoms in UEF and Monash recordings indicated that 9% of electrographic seizures did not have a behavioral correlate and had been unrecognized without EEG. Moreover, 32% of the seizures with behavioral symptom were classified as focal with Racine score 1–2. Overall, these data suggest that without EEG recording there is a risk of underdiagnosing PTE after lateral FPI.

Taken together, like the prevalence of epilepsy also the epilepsy phenotype was reproducible between the sites. The occurrence of seizure clusters together with an overall low seizure frequency needs to be taken into consideration when planning the treatment trials and defining, for example, the 50% responder rate.

4.3 |. Some variability in animal and TBI-induction related factors did not affect epileptogenesis

Genetic background can affect the risk of structural epileptogenesis after TBI as shown in mice, the CD1 or APP/PS1 mice being more sensitive to post-traumatic epileptogenesis as compared to B6 mice after controlled cortical injury^25–30. Here, all study sites used adult male Sprague-Dawley rats, even though from different vendors.

Also, there were minor differences in pellets used for feeding, use of antibiotics, use of a drill vs. hand-held trephine for craniectomy and/or surgery-related analgesics. As the prevalence and epilepsy phenotype did not differ between the study sites, we conclude that the model reproducibility can tolerate such procedural variability.

4.4 |. Lessons learned

The present study is the first major preclinical attempt to harmonize the production of an animal model across multiple centres for a powered multicenter study on epileptogenesis. All study sites had used the lateral FPI model before initiation of the project. For the needs of EpiBioS4Rx, we made a major attempt to harmonize somewhat variable procedures. We used the same strain, sex and age of the rats as well as aimed at comparable housing, anesthesia, craniectomy site and size and impactor. We exchanged information of surgical and other procedural details and practiced the procedures. All procedures and their timing were reported using common data elements on the excel sheet. Finally, we conducted an interim analysis to assess the harmonization at the early stage of the study to identify the action points to improve the performance^31–37. Considering that the project was undertaken over a period of about three years, we realized the need of continuing training and monitoring to optimize the project success.

We encountered some obstacles that were not be foreseen in project planning. The first one related to malfunction of the pogo-pin electrode connectors after some time of their implantation. This resulted in exclusion of over 40 animals after the 7^th month video-EEG monitoring. Moreover, even though seizures could still be reliable detected in some animals, the seizure origin could not be analyzed. This experience emphasized the need of a preliminary study to test the materials which are planned to be used in the final study.

Another challenge related to presence of brain lesions detected in MRI at later stages of the study, including surgery-related brain lesions and abscesses. In the MRI cohort, the surgery-related lesions could be detected early in in vivo MRIs and the animals were excluded. In the EEG cohort, however, the ex vivo MRI became available in the end of the study, that is, after a 7-months follow-up, including the labor intensive video-EEG monitoring. These observations emphasize the benefits of the use of structural MRI, if available, for early screening of the animal cohort to stratify only the animals with TBI-related brain lesions into the follow-up. Moreover, either MRI or histology should be performed to exclude non-TBI related structural epilepsy in TBI animals. Structural analysis is also important to verify the depth electrode positioning for analysis of signal origin.

5 |. CONCLUSIONS

The lateral FPI-induced epileptogenesis and epilepsy phenotype are highly reproducible across sites if harmonization of methodology and data collection is undertaken. Consequently, the model is suitable for a multi-center design, which may be needed to produce larger animal numbers for a statistically powered biomarker study and/or a treatment trial. However, a preliminary study, assessing the harmonization of the procedures and testing the materials to be used is highly recommended. Also, the constant training and monitoring of the progress at different study sites should be anticipated and included in the budgeting for the study. Further studies are needed to explore the model reproducibility in females and different age groups.

Key points.

• The first large scale preclinical multi-center study in epilepsy

• Epilepsy phenotype after lateral fluid-percussion induced traumatic brain injury is highly reproducible across different study sites

• Preclinical multicenter studies are feasible

DATA AVAILABILITY STATEMENT

The data that support the findings of this study are available from the senior authors (A.P., T.O.B., R.S) upon reasonable request.