Secondary-blast injury in rodents produces cognitive sequelae and distinct motor recovery trajectories

1. Introduction

Secondary blast injuries (SBIs) are injuries induced by fragments and debris propelled by high-pressure waves caused by explosions(McAllister, 2011). Increased use of improvised explosive devices due to terrorist activities has raised the incidence of secondary blast injuries(Tahtabasi et al., 2020). Although multiple body parts may be affected, recent studies have shown that the height at which the explosive device is placed increases the likelihood of inflicting damage in specific body areas(Singh et al., 2014). For instance, the head, face, and neck are susceptible areas when civilians are seated near an explosive device(Tahtabasi et al., 2020). In head injuries, multiple brain regions are commonly affected by this type of penetrating damage, which leads to fundamental changes in cognitive, motor, and emotional functions. Indeed, blast-injured patients undergo multiple sequelae, including impaired memory, decreased motor function, and mood disorders(McAllister, 2011; Swanson et al., 2017). Although restoration of these functions is a critical goal for patient recovery, how different neural substrates recover function and contribute to the short- and long-term rehabilitation of blast injury patients is unclear. Interventions focused on motor dysfunction recovery aim at improving specific muscle groups, inhibiting excessive tone, or facilitating movement patterns that enhance aspects of gait(Seif-Naraghi and Herman, 1999). However, these targeted approaches are unable to promote systemic recovery(Scherer, 2007). Similar events occur during cognitive rehabilitation, in which compensatory strategies seek to overcome specific problems and improve performance through interventions relying on theoretical conceptions(Bogdanova and Verfaellie, 2012). With this in mind, we designed a secondary blast-injury rodent model in which we consistently target susceptible brain structures to study cognitive, motor and mood deficits. Since several reports in humans have outlined gender differences in post-trauma outcome(Spani et al., 2018), we analyzed the consequences of blast injury in male and female mice. Finally, because neuronal and axonal injury are common neuropathological consequences of secondary-blast injury in humans that can have long term consequences(Aldag et al., 2017), we also examined brain areas undergoing degeneration post-trauma. We hypothesized that animals exposed to secondary blast injury would show cognitive impairment and that blast injury produces consistent motor impairment in both males and females. Indeed, cognitive deficit was significant in male mice, but surprisingly, we observed both divergent and consistent motor outcomes across non-control cohorts. Both male and female cohorts contained individuals with motor trajectories that actually surpassed control animals. However, both cohorts also had mice that performed worse than or similar to the control group. Intriguingly, we found that groups with different recovery trajectories had distinct patterns of injury in brain areas. The consistency of our secondary-blast injury model will help to elucidate the neural basis of axonal remodeling, plasticity, and adaptation associated with individual differences in performance in male and female mice after trauma. The model also has the potential to help develop novel rehabilitation strategies for blast-injured patients.

2. Results

2.1. Blast model features

Skull penetration and intracranial injuries are common consequences of blast and propelled shrapnel upon exposure to improvised explosive devices(Singh et al., 2016; Tahtabasi et al., 2020). We therefore designed a device that delivers a controlled blast and mimics these lesions, but is confined to particular brain regions to increase reproducibility and minimize effects from blast overpressure and debris in other vital organs. This blast model leads to injuries, including skull fractures, cerebral contusion and penetrating injuries observed in individuals exposed to explosions. Since we fixed the animal’s head using a micro-switch sensor (Figure 1A) that also controls the distance traveled by a motor stepper in which the device rests (Figure 1D), we reliably targeted the parieto-temporal area. The device routinely caused local skull fracture from which blood leaked, minimizing the possibility of extensive subarachnoid hemorrhages and likelihood of brainstem herniation. Nevertheless, subarachnoid hemorrhages were evident within the region impacted by the shrapnel and blast wave (Figure 1 B&C). We adjusted the air blast pressures to 50 psi for males and 45 psi for females as to avoid frontal and parietal bone fractures together with extensive hematomas and rupture of jaw muscles as described in (Guley et al., 2016) (Kuehn et al., 2011). Lower pressures (<30 psi) did not show obvious injuries or pathological changes, setting a lower limit for the model.

Figure 1. Development of a secondary blast injury model and timeline of experimental design.

(A) Close up of head holder and lateral view of the impacted area (mesocircuit). The photograph shows a micro-switch sensor that limits the penetration of the piston (which represents shrapnel associated in SBI) in the region of interest. The animal is exposed to isoflurane via an anesthetic mask as depicted. (B) Gross pathology showing the injured area on a perfused brain after right SBI from a lateral(B) and dorsal view (C) at 4 hours and 21 days post-SBI. Pressure-dependent apparatus for producing SBI including the components used to align the head and the modified blast gun aligned with the target area using a motor stepper controlled by an Arduino (D) Experimental timeline including all behavioral assessments conducted in male and female mice. The brains were perfused following behavioral experiments. Scale bar: 3.5 mm

In order to examine the effects of secondary blast injury, we assessed recovery time based on physiological parameters in male and female mice. All animals exposed to secondary blast injury (Figure 1) experienced apnea for 2-4 s. However, animals exposed to sham trauma (controls) showed normal breathing patterns. Blast-injured mice recovered from anesthesia and re-righted themselves within an average time period of 8 ± 0.76 min. In contrast, control mice stood up within a significantly shorter time (2.5± 0.18 min; p=<0.001, two-tailed Welch’s T-test). We analyzed males and females together as no differences were found between them when we quantified this parameter. However, we found gender differences when monitoring the body weight before and after (day 4) post-injury. On average, male mice lost 3.2 ± 3 g, whereas blast injured females did not show significant fluctuations in their body weight (p=0.10, two-tailed paired T-test). The differences in control (male and female) body weights were statistically negligible (see Table 1). Overall, a localized blast injury of (50 psi pressure) moderately affects the health of males within the first-week post-trauma. A blast injury in females (45 psi) caused at most mild effects on the health of females.

Table 1.

Male and female body weights in grams(g) before and after secondary blast injury.

Subjects	No. of animals tested (n)	Average weight on Day 1 (g)	Average weight on Day 4 (g)
Male - Control	37	26.4 ± 0.4	25.4 ± 0.5
Male - SBI	60	25.2 ± 0.3***	22.0 ± 0.4***
Female - Control	15	20.6 ± 0.4	21.5 ± 0.3
Female - SBI	17	20.3 ± 0.3	18.6 ± 0.6

^***represents significant differences P = < 0.0001 between day 1 (immediately before SBI) and after (day 4) post trauma.

2.2. Secondary blast injury produces memory. and spatial learning deficits in male, but not female mice.

Given the susceptibility of the hippocampal and frontal cortex area to blast injury in humans(McAllister, 2011; Swanson et al., 2017) and rodents(Hernandez et al., 2018), and that impaired memory is a frequent manifestation in humans after secondary blast injury(McAllister, 2011; Swanson et al., 2017), we hypothesized that our SBI model exhibits spatial memory and learning deficits. We used the Morris Water Maze (MWM) to test differences between experimental groups during both the learning and extinguishing/re-learning phases. For the learning phase, we applied an innovative analytical approach for behavioral studies, survival analysis(Andersen et al., 2020). Specifically, we applied the accelerated failure time to event model (AFT). This approach was recently used in the analysis of MWM data, and defines latency to reaching the platform as a time-to-event problem. Our analysis therefore considered both time (the latency to reaching the platform) and event information (whether the animal reached the platform within the available time(90s)). Overall, our data show that control and SBI mice have reduced latency and increased success rate in reaching the platform (Figure 2A, S1A, Tables 2&3) over the learning days. Both male and female mice show an increased probability of finding the platform over time (Figure 2B, S1B). However, SBI male mice frequently showed a lower probability of reaching the platform than control mice (Figure 2B, and Table 4) or female mice (Figure S1, S2Table 5).

Figure 2. Blast injured male mice exhibit memory impairment during the learning phase of the Morris water maze task.

(A) Swim duration of each animal per trial and day during the learning phase (4 days). SBI data is shown in red, and control (CTRL) in black. If animals reached the platform before the established time (90s), we identified the trace with a closed circle. However, if the animal could not reach the platform within the 90s, we identified it with an open circle (censored time) (B) Probability of reaching the platform as a function of time (4 days) for SBI (red) and control (black) mice applying predictions from the accelerated failure time (AFT) time-to-event model. See table 4 for p-values. The interval (gray) represents the standard error. See methods for data analysis in section 2.2 & 4.7.3

Table 2.

Univariate Summaries of Swim Duration in male mice and Incidence of Reaching Platform by Group, Day, and Swim per Day SD = standard deviation; SBI = secondary blast injury.

Group	Day	Swim	N	Mean Swim Duration (s)	Swim Duration SD	Median	Q1	Q3	Min	Max	Reached Platform (N)	Reached Platform (%)
CONTROL	1	1	28	39.04	32.10	27	14.5	62.25	3	90	21	75.00
CONTROL	1	2	28	27.50	23.40	19	13.75	26.75	3	90	27	96.43
CONTROL	1	3	27	26.96	24.77	17	12	32	6	90	25	92.59
CONTROL	1	4	28	29.04	25.05	22	13.75	36.5	1	90	26	92.86
CONTROL	2	1	28	24.86	24.56	20.5	8	29	1	90	26	92.86
CONTROL	2	2	27	22.19	20.18	18	8.5	26	5	90	26	96.30
CONTROL	2	3	27	26.89	21.80	22	11.5	31	4	90	26	96.30
CONTROL	2	4	27	16.78	13.70	10	6	24.5	4	43	27	100.00
CONTROL	3	1	28	14.93	13.55	12.5	4	18.75	1	50	27	96.43
CONTROL	3	2	28	18.46	12.44	17.5	10.5	21.25	4	59	28	100.00
CONTROL	3	3	28	28.54	18.67	23.5	16.75	32.25	3	90	27	96.43
CONTROL	3	4	28	24.29	22.29	17	10.75	24.75	4	90	27	96.43
CONTROL	4	1	28	11.32	18.76	5	3	11	1	90	27	96.43
CONTROL	4	2	28	24.50	20.84	19.5	9.75	28.75	4	90	27	96.43
CONTROL	4	3	28	24.86	24.37	16.5	10	26.5	3	90	26	92.86
CONTROL	4	4	28	11.43	7.90	9	5.75	16.25	2	33	28	100.00
SBI	1	1	53	41.85	34.34	34.5	11	78.75	1	90	41	77.36
SBI	1	2	53	45.28	33.44	34.5	15	90	4	90	39	73.58
SBI	1	3	53	41.52	31.99	30	14.25	82	5	90	42	79.25
SBI	1	4	51	39.31	32.87	23	11.25	78.25	2	90	45	88.24
SBI	2	1	53	25.35	30.17	12	5	25	1	90	47	88.68
SBI	2	2	52	29.47	28.20	17	10	34	3	90	47	90.38
SBI	2	3	53	32.67	27.39	25.5	15	37	5	90	46	86.79
SBI	2	4	51	28.93	28.23	17	12	33.25	3	90	46	90.20
SBI	3	1	53	23.50	28.93	12	3.25	28.75	1	90	49	92.45
SBI	3	2	53	30.74	28.83	18.5	11	34.5	3	90	46	86.79
SBI	3	3	53	26.56	25.95	14	9	33.75	4	90	49	92.45
SBI	3	4	51	31.06	31.00	14	8.25	47.25	2	90	47	92.16
SBI	4	1	53	21.83	27.21	9	5	25.75	1	90	50	94.34
SBI	4	2	53	28.87	27.68	16	10	36	5	90	46	86.79
SBI	4	3	53	31.09	30.40	17	9	40.5	4	90	45	84.91
SBI	4	4	51	26.87	24.88	15	8	36	4	90	50	98.04

Table 3.

Univariate Summaries of female mice Swim Duration and Incidence of Reaching Platform by Group, Day, and Swim per Day. SD = standard deviation; SBI = secondary blast injury.

the Group	Day	Swim	N	Mean Swim Time (s)	Swim Time SD	Median	Q1	Q3	Min	Max	Reached Platform (N)	Reached Platform (%)
Control	1	1	14	40.14	38.00	20	5.75	83.75	2	90	11	78.57
Control	1	2	14	48.57	33.40	46.5	18.5	86.25	5	90	10	71.43
Control	1	3	14	38.21	33.09	29	11.25	59.5	4	90	11	78.57
Control	1	4	14	23.86	29.14	10.5	8.25	25.5	6	90	12	85.71
Control	2	1	14	22.64	23.90	15	10.5	30	1	90	13	92.86
Control	2	2	14	17.64	26.73	7	5	11.75	3	90	13	92.86
Control	2	3	14	19.57	22.26	12.5	7.25	21.75	4	90	13	92.86
Control	2	4	14	21.71	29.43	11.5	5	18	4	90	12	85.71
Control	3	1	14	2.86	3.13	2	1.25	2	1	13	14	100.00
Control	3	2	14	16.64	19.97	12	5	17	4	82	14	100.00
Control	3	3	14	21.21	23.17	13	8	25.5	4	90	13	92.86
Control	3	4	14	21.50	21.93	16.5	9.5	22.75	3	90	13	92.86
Control	4	1	14	5.00	5.33	3	2	4	1	19	14	100.00
Control	4	2	14	20.86	29.01	9.5	6.25	15.75	4	90	13	92.86
Control	4	3	14	12.50	5.60	10.5	8.25	16	5	25	14	100.00
Control	4	4	14	12.36	7.87	12	7.5	14.75	3	36	14	100.00
SBI	1	1	16	34.31	33.82	17	8.75	61.25	2	90	14	87.50
SBI	1	2	16	21.31	18.98	19	7.5	25.25	4	72	16	100.00
SBI	1	3	16	28.63	20.32	24.5	15	36.75	5	76	16	100.00
SBI	1	4	16	19.69	15.92	14.5	7.25	29.75	4	52	16	100.00
SBI	2	1	16	16.88	23.49	9.5	3.5	18	1	90	15	93.75
SBI	2	2	16	23.81	28.78	14.5	4	23.75	2	90	14	87.50
SBI	2	3	16	26.50	29.01	10.5	6	41	4	90	14	87.50
SBI	2	4	16	21.44	24.17	12.5	4	33.5	3	90	15	93.75
SBI	3	1	16	11.44	12.37	8.5	2	15.25	1	49	16	100.00
SBI	3	2	16	11.81	8.35	7.5	5.75	17.25	4	27	16	100.00
SBI	3	3	16	14.56	10.61	12	8.5	14.25	3	42	16	100.00
SBI	3	4	16	16.50	20.99	9.5	5	19.5	4	90	15	93.75
SBI	4	1	16	14.69	22.54	6	2.75	19	1	90	15	93.75
SBI	4	2	16	19.31	18.25	12.5	4	29	2	62	16	100.00
SBI	4	3	16	20.88	27.75	10	6	20	4	90	14	87.50
SBI	4	4	16	27.94	32.82	9	8.25	31.75	4	90	13	81.25

Table 4.

Time-to-Event (Reaching Platform) Model-Adjusted Differences between Control and SBI male mice, by Day and Swim per Day Exp (estimate) provides the ratio of model-adjusted mean swim durations between Control and SBI groups. The Hommel-adjusted p-values compensate for multiple testing. Bold = p < 0.05. CI = confidence interval; SE = standard error; SBI = secondary blast injury.

Day	Swim	Estimate	SE	Exp (estimate)	CI 95 Min	CI 95 Max	p-value	Hommel p-value
1	1	0.01	0.33	1.01	0.53	1.91	0.9738	0.9738
1	2	0.57	0.24	1.77	1.10	2.83	0.0180	0.2342
1	3	0.52	0.23	1.69	1.08	2.63	0.0212	0.2549
1	4	0.25	0.27	1.29	0.76	2.19	0.3520	0.9738
2	1	−0.29	0.29	0.74	0.43	1.30	0.3022	0.9738
2	2	0.18	0.21	1.20	0.79	1.82	0.3824	0.9738
2	3	0.16	0.20	1.17	0.79	1.74	0.4415	0.9738
2	4	0.47	0.23	1.61	1.03	2.51	0.0373	0.4100
3	1	0.09	0.31	1.09	0.60	1.99	0.7775	0.9738
3	2	0.29	0.19	1.33	0.92	1.93	0.1243	0.9738
3	3	−0.35	0.18	0.71	0.50	1.01	0.0555	0.6108
3	4	0.02	0.24	1.02	0.64	1.62	0.9215	0.9738
4	1	0.72	0.29	2.06	1.18	3.60	0.0116	0.1505
4	2	0.06	0.22	1.06	0.69	1.64	0.7764	0.9738
4	3	0.14	0.24	1.15	0.72	1.84	0.5530	0.9738
4	4	0.67	0.21	1.96	1.31	2.94	0.0010	0.0166

Table 5.

Time-to-Event (Reaching Platform) Model-Adjusted Differences between Control and SBI female mice, by Day and Swim per Day Exp (estimate) provides the ratio of model-adjusted mean swim durations between Control and SBI groups. The Hommel-adjusted p-values compensate for multiple testing. Bold = p < 0.05. CI = confidence interval; SE = standard error; SBI = secondary blast injury.

Day	Swim	Estimate	SE	Exp (estimate)	CI 95 Min	CI 95 Max	p-value	Hommel p-value
1	1	−0.20	0.70	0.82	0.21	3.20	0.7702	0.9580
1	2	−1.02	0.42	0.36	0.16	0.83	0.0168	0.2527
1	3	−0.14	0.44	0.87	0.37	2.07	0.7547	0.9580
1	4	0.04	0.36	1.04	0.51	2.13	0.9045	0.9580
2	1	−0.60	0.48	0.55	0.22	1.39	0.2052	0.9580
2	2	0.43	0.40	1.53	0.70	3.35	0.2862	0.9580
2	3	0.08	0.42	1.08	0.47	2.46	0.8536	0.9580
2	4	0.05	0.42	1.05	0.46	2.40	0.9109	0.9580
3	1	1.23	0.38	3.41	1.61	7.22	0.0014	0.0229
3	2	−0.16	0.29	0.85	0.48	1.51	0.5846	0.9580
3	3	−0.14	0.29	0.87	0.49	1.53	0.6201	0.9580
3	4	−0.40	0.32	0.67	0.36	1.25	0.2114	0.9580
4	1	0.65	0.41	1.92	0.86	4.28	0.1103	0.9580
4	2	0.13	0.39	1.14	0.53	2.43	0.7410	0.9580
4	3	−0.01	0.25	0.99	0.61	1.61	0.9580	0.9580
4	4	0.26	0.35	1.30	0.66	2.57	0.4477	0.9580

According to the results obtained by applying the time to event AFT model, while individual male SBI mice (n=35) displayed a trend of increased latencies across days and swimming trials in comparison with the control group, there were initially no significant changes when considering the cohort (n=30; Figure 2B & Table 4). However, the fourth swim on the last day of learning showed significant differences between the male SBI cohort and controls (P = 0.01; Hommel-adjusted for multiple comparisons). Table 6 shows consistent and significant reductions in mean swim time of SBI animals when comparing day 1 to each of the days, indicating that SBI mice ultimately were able to learn. The control group showed similar results, but with drastic reductions in mean swim times between days 1 and 3&4, suggesting faster learning capabilities.

Table 6.

Time-to-Event (Reaching Platform) Model-Adjusted Differences between Days, by Treatment Group and Swim per Day in male mice. Exp (estimate) provides the ratio of model-adjusted mean swim durations between TBI and Sham. The Hommel-adjusted p-values compensate for multiple testing. Bold = p < 0.05. CI = confidence interval; SE = standard error; SBI = secondary blast injury.

Contrast (Days)	Group	Swim	Estimate	SE	Exp (estimate)	CI 95 Min	CI 95 Max	p-value	Hommel p-value
2 - 1	CONTROL	1	−0.55	0.27	1.01	0.30	1.09	0.0421	0.8872
3 - 1	CONTROL	1	−1.05	0.29	1.77	0.22	0.56	0.0003	0.0108
3 - 2	CONTROL	1	−0.50	0.28	1.69	0.39	0.95	0.0722	0.9410
4 - 1	CONTROL	1	−1.77	0.28	1.29	0.10	0.29	0.0000	0.0000
4 - 2	CONTROL	1	−1.22	0.24	0.74	0.17	0.52	0.0000	0.0000
4 - 3	CONTROL	1	−0.72	0.27	1.20	0.32	0.74	0.0089	0.3017
2 - 1	CONTROL	2	−0.24	0.20	1.17	0.53	1.18	0.2291	0.9823
3 - 1	CONTROL	2	−0.28	0.20	1.61	0.48	1.18	0.1542	0.9823
3 - 2	CONTROL	2	−0.05	0.19	1.09	0.52	1.74	0.8066	0.9823
4 - 1	CONTROL	2	−0.13	0.22	1.33	0.61	1.27	0.5625	0.9823
4 - 2	CONTROL	2	0.11	0.24	0.71	0.78	1.58	0.6560	0.9823
4 - 3	CONTROL	2	0.15	0.21	1.02	0.73	1.85	0.4725	0.9823
2 - 1	CONTROL	3	0.09	0.20	2.06	0.62	1.91	0.6597	0.9823
3 - 1	CONTROL	3	0.25	0.16	1.06	0.83	1.97	0.1327	0.9823
3 - 2	CONTROL	3	0.16	0.19	1.15	0.73	1.86	0.4015	0.9823
4 - 1	CONTROL	3	−0.11	0.21	1.96	0.60	1.34	0.5852	0.9823
4 - 2	CONTROL	3	−0.20	0.23	1.01	0.43	1.55	0.3910	0.9823
4 - 3	CONTROL	3	−0.36	0.19	1.77	0.44	1.12	0.0547	0.9141
2 - 1	CONTROL	4	−0.57	0.17	1.69	0.36	0.88	0.0006	0.0252
3 - 1	CONTROL	4	−0.18	0.20	1.29	0.49	1.42	0.3706	0.9823
3 - 2	CONTROL	4	0.39	0.21	0.74	0.84	2.58	0.0621	0.9307
4 - 1	CONTROL	4	−0.84	0.20	1.20	0.29	0.65	0.0000	0.0013
4 - 2	CONTROL	4	−0.27	0.20	1.17	0.51	1.14	0.1735	0.9823
4 - 3	CONTROL	4	−0.66	0.18	1.61	0.33	0.81	0.0004	0.0149
2 - 1	SBI	1	−0.86	0.26	1.09	0.23	0.77	0.0012	0.0456
3 - 1	SBI	1	−0.98	0.26	1.33	0.26	0.54	0.0002	0.0092
3 - 2	SBI	1	−0.12	0.18	0.71	0.62	1.27	0.5187	0.9823
4 - 1	SBI	1	−1.06	0.28	1.02	0.22	0.55	0.0001	0.0056
4 - 2	SBI	1	−0.20	0.21	2.06	0.47	1.43	0.3353	0.9823
4 - 3	SBI	1	−0.08	0.19	1.06	0.60	1.42	0.6625	0.9823
2 - 1	SBI	2	−0.62	0.23	1.15	0.34	0.06	0.0064	0.2249
3 - 1	SBI	2	−0.56	0.22	1.96	0.38	0.85	0.0116	0.3719
3 - 2	SBI	2	0.06	0.19	1.01	0.56	2.01	0.7683	0.9823
4 - 1	SBI	2	−0.64	0.21	1.77	0.33	0.85	0.0026	0.0989
4 - 2	SBI	2	−0.02	0.15	1.69	0.63	1.54	0.9190	0.9823
4 - 3	SBI	2	−0.07	0.16	1.29	0.55	1.58	0.6521	0.9823
2 - 1	SBI	3	−0.28	0.20	0.74	0.43	1.33	0.1745	0.9823
3 - 1	SBI	3	−0.62	0.18	1.20	0.35	0.81	0.0004	0.0172
3 - 2	SBI	3	−0.35	0.15	1.17	0.40	1.05	0.0207	0.5952
4 - 1	SBI	3	−0.49	0.21	1.61	0.39	0.95	0.0202	0.5864
4 - 2	SBI	3	−0.22	0.16	1.09	0.44	1.47	0.1847	0.9823
4 - 3	SBI	3	0.13	0.15	1.33	0.79	1.64	0.4017	0.9823
2 - 1	SBI	4	−0.35	0.23	0.71	0.49	1.00	0.1257	0.9823
3 - 1	SBI	4	−0.41	0.23	1.02	0.42	1.05	0.0719	0.9410
3 - 2	SBI	4	−0.06	0.21	2.06	0.54	1.65	0.7746	0.9823
4 - 1	SBI	4	−0.42	0.21	1.06	0.43	1.01	0.0478	0.9015
4 - 2	SBI	4	−0.07	0.18	1.15	0.59	1.49	0.7135	0.9823
4 - 3	SBI	4	0.00	0.21	1.96	0.67	1.49	0.9823	0.9823

To assess whether blast-injured mice extinguish their initial learning of the platform’s location in Quadrant 1, we moved the hidden platform to Quadrant 3 (Figure 3A) after the learning phase. Both control and SBI males spent similar time on day 1 of the probe phase attempting to find the new platform location (Figure 3B; control: 44.7 ± 3.2 s, blast-injured: 50.7 ± 2.5 s). However, control males learned the new location much faster than the blast-injured males by day 2 (control: 29.7 ± 2.8 s, blast-injured: 43.5 ± 2.5 s). Blast-injury had a significant effect on the probe day latencies (F (1,268)=26.03, p<0.001; two-way repeated measures ANOVA (2RMA)). In contrast, the interaction between the blast-injury and probe days did not show a significant effect on probe day latencies (p=0.21).

Figure 3. Blast injured males are impaired in extinguishing previously learned behavior.

(A) The schematic displays the location of the platform in Quadrant 1 during the learning days. After the first four days of learning, the platform was moved to Quadrant 3 for the probe days.

(B) SBI male(n=54) average latencies (mean ± S.E.M) to find the platform compared to controls(n=28) on probe days. A two-way repeated measures ANOVA analysis conducted on the probe phase latencies reveal a significant effect between groups (F(1,268)=26.03, p<0.001). SBI males exhibited significantly longer latencies than controls on days 2 and 4 of the probe phase (p<0.05, Bonferroni-adjusted T-test). (C) During the probe phase, SBI males spent more time on average (mean ± S.E.M) in Quadrant 1 (Q1) and less time in Quadrant 3 (Q3), while controls spent more time on average in Q3 and less time in Q1 as they learned the new platform location (p<0.05, Bonferroni-adjusted T-test). Our analysis of the percentage of time the groups spent in each quadrant revealed that the quadrant*group interaction was significant (F(3,4480)=7.47, p<0.001 ; three-way ANOVA). No changes in other quadrants were observed. (*p<0.05, **p<0.01, ***p<0.001).

We quantified the time mice spent in each quadrant (Figure 3C). SBI mice struggled to associate Q3 with the platform location and spent more time in Q1 than controls (p=0.001, Bonferroni-adjusted T-test). Our ANOVA analysis showed that the trauma is also associated with significant differences between groups on the percentage of time spent in Q1 and Q3 (F(1,268)=11.65, p<0.001 and F(1,268)=5.52, p=0.02 respectively; 2RMA). Overall, blast injury had a significant effect on durations spent in quadrants (F(3,4480)=7.47, p<0.001; three-way ANOVA). These results suggest that blast-injured males are impaired in extinguishing previously learned behavior. Importantly, there were no statistical differences between the swim velocities of SBI and controls in the learning phase (Figure S3A; p=0.71, 2RMA), or probe phase (Figure S3B). Together, our results suggest that the negative impact on MWM performance was a result of cognitive impairment. In contrast, female mice show no significant differences in escape latencies or the probability of reaching the platform when compared to the control group (Figure S1, S2 and Tables 3,5,7). Additionally, there were no group differences in the percentage of time spent in Q1 and Q3 on probe days (Figure S2C;p=0.84 and p=0.92 respectively, 2RMA). Similarly, there were no differences in the swim velocities of the blast-injured females in the learning or probe phase (Figure S3C p=0.27 and Figure S3D p=0.38, respectively; 2RMA). The blast injury did not affect the MWM performance of female mice. These results indicate that blasted-injured males, but not females, show deficits in spatial learning, memory and reversal learning.

Table 7.

Time-to-Event (Reaching Platform) Model-Adjusted Differences between Days, by Treatment Group and Swim per Day in female mice. Exp (estimate) provides the ratio of model-adjusted mean swim durations between TBI and Sham. The Hommel-adjusted p-values compensate for multiple testing. Bold = p < 0.05. CI = confidence interval; SE = standard error; SBI = secondary blast injury.

Contrast	Group	Swim	Estimate	SE	Exp (estimate)	CI 95 Min	CI 95 Max	p-value	Hommel p-value
2 - 1	Control	1	−0.47	0.59	0.82	0.16	2.45	0.4109	0.9871
3 - 1	Control	1	−2.47	0.55	0.36	0.04	0.19	0.0000	0.0004
3 - 2	Control	1	−2.00	0.27	0.07	0.06	0.32	0.0000	0.0000
4 - 1	Control	1	−2.00	0.55	1.04	0.07	0.27	0.0003	0.0123
4 - 2	Control	1	−1.53	0.34	0.55	0.09	0.55	0.0000	0.0004
4 - 3	Control	1	0.47	0.32	1.53	0.73	3.49	0.1449	0.9871
2 - 1	Control	2	−1.63	0.37	1.00	0.09	0.45	0.0000	0.0006
3 - 1	Control	2	−1.35	0.37	1.05	0.11	0.59	0.0003	0.0124
3 - 2	Control	2	0.28	0.25	3.41	0.63	2.81	0.2647	0.9871
4 - 1	Control	2	−1.37	0.32	0.85	0.14	0.45	0.0000	0.0013
4 - 2	Control	2	0.26	0.24	0.87	0.74	2.29	0.2822	0.9871
4 - 3	Control	2	−0.02	0.30	0.67	0.52	1.83	0.9416	0.9871
2 - 1	Control	3	−0.70	0.46	1.92	0.22	1.10	0.1267	0.9871
3 - 1	Control	3	−0.65	0.47	1.14	0.24	1.12	0.1646	0.9871
3 - 2	Control	3	0.05	0.30	0.99	0.65	1.72	0.8556	0.9871
4 - 1	Control	3	−0.83	0.42	1.30	0.22	0.86	0.0496	0.9507
4 - 2	Control	3	−0.13	0.24	0.82	0.22	3.47	0.6084	0.9871
4 - 3	Control	3	−0.18	0.25	0.36	0.36	1.92	0.4676	0.9871
2 - 1	Control	4	−0.17	0.43	0.87	0.35	1.99	0.6859	0.9871
3 - 1	Control	4	0.14	0.31	1.04	0.56	2.33	0.6575	0.9871
3 - 2	Control	4	0.31	0.27	0.55	0.54	3.46	0.2476	0.9871
4 - 1	Control	4	−0.23	0.30	1.53	0.36	1.75	0.4579	0.9871
4 - 2	Control	4	−0.05	0.26	1.08	0.42	2.16	0.8418	0.9871
4 - 3	Control	4	−0.36	0.19	1.05	0.30	1.59	0.0635	0.9779
2 - 1	SBI	1	−0.87	0.46	3.41	0.20	0.88	0.0583	0.9697
3 - 1	SBI	1	−1.04	0.53	0.85	0.20	0.63	0.0496	0.9507
3 - 2	SBI	1	−0.17	0.58	0.87	0.48	1.49	0.7731	0.9871
4 - 1	SBI	1	−1.15	0.62	0.67	0.17	0.59	0.0655	0.9819
4 - 2	SBI	1	−0.27	0.49	1.92	0.34	1.69	0.5785	0.9871
4 - 3	SBI	1	−0.11	0.42	1.14	0.42	1.92	0.8023	0.9871
2 - 1	SBI	2	−0.19	0.40	0.99	0.51	1.35	0.6355	0.9871
3 - 1	SBI	2	−0.49	0.34	1.30	0.31	1.21	0.1430	0.9871
3 - 2	SBI	2	−0.30	0.34	0.82	0.19	2.90	0.3701	0.9871
4 - 1	SBI	2	—0.23	0.43	0.36	0.35	1.83	0.5981	0.9871
4 - 2	SBI	2	−0.04	0.21	0.87	0.41	2.28	0.8535	0.9871
4 - 3	SBI	2	0.27	0.32	1.04	0.64	2.65	0.4066	0.9871
2 - 1	SBI	3	−0.49	0.44	0.55	0.24	1.56	0.2629	0.9871
3 - 1	SBI	3	−0.66	0.24	1.53	0.24	1.14	0.0062	0.2465
3 - 2	SBI	3	−0.17	0.32	1.08	0.37	1.93	0.6068	0.9871
4 - 1	SBI	3	−0.70	0.25	1.05	0.22	1.13	0.0047	0.1898
4 - 2	SBI	3	−0.22	0.34	3.41	0.38	1.71	0.5228	0.9871
4 - 3	SBI	3	−0.05	0.21	0.85	0.54	1.69	0.8187	0.9871
2 - 1	SBI	4	−0.17	0.38	0.87	0.48	1.49	0.6549	0.9871
3 - 1	SBI	4	−0.31	0.31	0.67	0.39	1.38	0.3262	0.9871
3 - 2	SBI	4	−0.14	0.30	1.92	0.39	1.94	0.6474	0.9871
4 - 1	SBI	4	−0.01	0.36	1.14	0.46	2.13	0.9871	0.9871
4 - 2	SBI	4	0.16	0.37	0.99	0.72	1.91	0.6617	0.9871
4 - 3	SBI	4	0.30	0.26	1.30	0.69	2.67	0.2453	0.9871

2.3. Secondary blast injury results in distinct trajectories of motor recovery in males and females.

Patients with brain injury secondary to blasts experience difficulty walking, poor coordination, decreased motor function and control(Scherer, 2007). We used the rotarod test (RR) because this skill learning task has been modeled to establish the trajectories followed by wildtype animals. This test, therefore, allows quantitative benchmarking of general motor deficits in animal disease models relative to control mice(Buitrago et al., 2004; Hamm et al., 1994). Although others have evaluated motor performance immediately after trauma(Guley et al., 2016; Zhou et al., 2018), we assessed motor features 7 days after the animals recovered from blast injury (Figure 1E) completing 5 trials of the RR each day. We also sought to establish the rate of motor recovery(Hamm et al., 1994) of males and females by performing daily evaluations for 10 consecutive days. Motor performance was evaluated by the length of consecutive running time on the RR, which accelerated at a rate of 0.1 cm/s.

Wildtype rodents exposed to this paradigm show courses modeled by an exponential function in which performance improvement occurs within five days and subsequently plateaus. (Buitrago et al., 2004). Since our paradigm followed that of Buitrago et al. and others(Buitrago et al., 2004; Scholz et al., 2015), we expected a similar trajectory in our sham group and a possible decay or consistent course if animals restore performance after seven days post-SBI.

We evaluated trajectories in the control and SBI cohorts by applying a principal component analysis (PCA) to uncover components underlying maximal variance in performance in control and SBI mice. We included all animals in this analysis (n=75). The first two components explained 86% of the variance. The first principal component strongly correlated with the latencies to fall from the RR during the first 4 days (day 0: PC1, 0.77; day1: PC1 0.83; day 2: PC1, 0.92 and day 3: PC1, 0.90). This suggests that (1) these latencies vary together and (2) that these correlations are consistent with the exponential model described by others in which the first four days delineate the steepest part of the learning curve(Buitrago et al., 2004; Scholz et al., 2015).

Since we noticed wide dispersal of data points in the SBI group (Figure 4C) compared to the confined behavior of the control group along the first PC1 (Figure 4A), we classified the data points. To classify these data (including control and SBI groups), we first calculated the centroid of the control group which is the mean of its principal components (PCs; Figure 4A). Note that we removed two subjects within the control group (n=28) with z-scores greater than 2 (Figure 4A). We calculated the Euclidian distance between controls and the centroid. The integer that enclosed the majority (>90%) of controls was considered as the radius. If the distance was less than the radius of control PCs in our SBI group (Figure 4C), we identified the cluster as “standard”; we labeled the cluster as “low” when the 1^st PC was more negative and greater than the radius and “high” when 1^st PC was positive and greater than the radius. The corresponding trajectories per animal between days 0-3 are shown in Figure 4B&D. Since males and females showed similar classification, we combined to perform the analysis.

Figure 4. SBI and control rotarod (RR) performances under two different motor learning paradigms.

(A) Evaluation of trajectories of control group (n=28) by applying principal component analysis (PCA). We calculated the Euclidian distance between controls and the centroid. The integer that enclosed the majority (>90%) of controls was considered as the radius of the gray circle displayed in the figure. (B) Corresponding trajectories per animal of the control group between day 0-3 (C) Data distribution of SBI trajectories along the principal component 1 (PC1) and classification. If the distance was less than the radius of control PCs in our SBI group, we identified the cluster as standard (SBI-S;n=21). We defined low-performers (SBI-L;n=5) if the 1^st PC was negative and high-performers (SBI-H;n=19) if the 1^st PC was positive.(D) Corresponding classified trajectories per animal of the SBI group between day 0-3 (E) One week after the injury, SBI-mice that were high performers (SBI-H mice) ran longer average durations (mean ± S.E.M) on the RR than low-performers (SBI-L),standard and control mice. A two-way repeated measures ANOVA conducted on the run times revealed a significant between-group effect (F(3, 322)=87.20, p<0.001) and significant within-group effects in the group*day interaction (F(23.35, 2505.76)=1.66, p<0.05; Huynh-Feldt corrected). SBI-H mice had significantly longer run times compared to the SBI-L, (F(1,97)=73.40, p<0.001), SBI-S (F(1,175)=95.96, p<0.001), and control groups (n=28; F(1,228)=140.87, p<0.001). Additionally, SBI-L mice had significantly lower run times than control (F(1,147)=68.69, p<0.001), SBI-S (F(1,94)=88.83, p<0.001), and SBI-H mice (F(1,97)=73.40, p<0.001). There were no significant differences found between control and SBI-S mice (F(1,225)=0.56, p=0.45). (F) No significant differences were observed during the pretraining days (Days 6-8) between groups. SBI-H mice (n=9) significantly improved their RR performance after the injury (Day 17). In contrast, RR performance by SBI-L mice (n=5) worsened following injury. No change in motor learning behavior was observed among control mice (n=20) and SBI-S group(n=10). A two-way repeated measures ANOVA conducted on run times revealed a significant between-group effect (F(3,126)=12.68, p<0.001) and significant within-group effects of day and the cluster per day interaction (F(11.56,1456.43)=11.55, p<0.001 and F(F(34.67,1456.43)=7.86, p<0.001 respectively; Greenhouse-Geisser corrected). (*p<0.05, **p<0.01, ***p<0.001).

Our analysis showed (Figure 4E) that the high-performing SBI cluster (SBI-H) (n=19) had longer run times on the RR task compared to the low performing (SBI-L), (n=5; F(1,97)=73.40, p<0.001; 2RMA) standard(SBI-S), (n=21; F(1,175)=95.96, p<0.001; 2RMA) and control groups (n=28; ; F(1,228)=140.87, p<0.001; 2RMA).

This difference was observed from the first day of testing (SBI-H: 37.6 ± 1.2 s, SBI-L: 14.8 ± 3.8 s, SBI-S: 30.2 ± 1.2 s and control: 27.5 ± 1.0 s) and persisted over the entire length of the task. There was no significant difference in performance between SBI-S and controls (p=0.45; 2RMA). We did, however, find a significant effect between groups on the latency to fall, and a significant interaction between groups and the testing day in the ANOVA results (F(3, 322)=87.20, p<0.001 and F(23.35, 2505.76)=1.66, p<0.05 respectively; 2RMA).

Since several studies have shown that training per se may elicit plastic changes in neural circuits such as the cerebellar-thalamic cortical area(Holschneider et al., 2007; Wang et al., 2013). We therefore tested whether high-performance observed in mice was a consequence of learning the motor task or a plastic event independent of learning. To do so, we focused on trajectories that assess mice on the accelerating RR for eight consecutive days before the injury, and for nine additional days after a 7-day recovery period (Figure 1E).

Performance is retained after eight days of paused testing in wildtype animals (Buitrago et al., 2004). We therefore expected flat trajectories when comparing 3 days before the pause (plateau) and 3 days after the pause in our control group or in those animals that restored motor activity during recovery time. A deviation of this expected trajectory may indicate either worsening post-SBI, or a plastic event that is independent of initial motor learning. To examine this particular condition, we classified data trajectories by applying a template matching approach. This classification included male and female mice (Figure 4F). Before the injury, performance of mice on the RR was not significantly different between the groups determined by post-hoc cluster analysis.

This implies that the RR findings from mice without any pre-training (Figure 4E) likely did not exhibit significant differences in performance prior to injury, and that the post-blast injury outcome was not caused by pre-existing differences between the groups in learning the motor task. After the blast injury, SBI-H mice (n=9) showed significantly longer run times when compared to performance prior to injury (Day 6: 35.47 ± 1.5 s, Day 17: 52.2 ± 2.6 s). However, control mice (n=20) and SBI-S (n=10) did not show any changes in run time between the pre-SBI and post-SBI phases (Day 6: 39.3 ± 1.2 s, Day 17: 37.0 ± 1.0 s) and (Day 6: 33.1 ± 1.1 s, Day 17: 37.2 ± 1.7 s) respectively. Additionally, SBI-L mice (n=5) showed significantly worse performance compared to the pre-injury phase (Day 6: 45.9 ± 2.2 s, Day 17: 34.2 ± 2.6 s). The blast injury showed a significant effect on run times (F(3,126)=12.68, p<0.001; 2RMA), and a significant interaction between groups and days of RR testing (F(34.67,1456.43)=7.86, p<0.001; 2RMA). These results suggest that the blast-injury had distinct outcomes on the motor performance of individual blast-injured mice, but this can be clustered into consistently distinct outcomes. SBI-L mice showed expected deficits, and the SBI-H mice showed an enhancement of motor skills. These results were consistent over several replicates of mice. Thus, the enhanced motor performance robustly observed in SBI-H-males and females is attributed to plastic changes as a consequence of the blast injury, rather than differences in motor learning per se.

2.4. Anxiety-related behavior is not found in male and female mice exposed to secondary blast injury.

Anxiety is a central feature of post-traumatic stress disorder (PTSD), a disorder that often coexists with blast injury. It is currently unclear whether blast-injury directly induces PTSD(Perez-Garcia et al., 2019). Since anxiety is observed in subjects struck by a blast, we applied the open field test (OFT) to our SBI model to evaluate levels of anxiety triggered by exposing mice to a large arena relative to the animal’s natural environment(Seibenhener and Wooten, 2015). We assessed mice for 10 minutes and quantified distance traveled, speed, and time spent in the center versus outer zones. When compared to blast-injured mice and controls of the male or female cohorts, no significant differences were found in traveled distance (Figure 5A, males: p=0.97; Figure 5D, females: p=0.62; one-way ANOVA[1AN]) and speed (Figure 5B, males: p=0.96;Figure 5E, females: 0.79; 1AN) suggesting regular locomotor activity. Similar results were obtained when we examined exploratory behaviors (Figure 5C, males: p=0.46; Figure 5F, females, p=0.51; two-way ANOVA) indicating the absence of anxiety-like behaviors.

Figure 5. Absence of anxiety-like behaviors in SBI mice.

(A) SBI males (n=14) performed similarly to controls (n=5) in the Open Field Test (OFT) when assessing average distance traveled, speed (p=0.97 and p=0.96 respectively, one-way ANOVA), and time spent in center and peripheral zones within the chamber (p=0.46, two-way ANOVA). The circles represent averaged trial scores for each subject. (B) SBI females (n=8) performed similarly to controls (n=8) in distance traveled, speed (p=0.62 and p=0.79 respectively, one-way ANOVA), and time spent in each zone (p=0.51, two-way ANOVA). NS. Non-significant statistical differences.

2.5. Brain areas linked to spatial learning and navigation undergo degeneration post-SBI

To examine neurodegeneration, we perfused mice after completing behavioral tasks (3 weeks post-injury) and brains were processed and stained with Fluoro-Jade. We analyzed approximately 130 sections per brain including control and SBI cohorts. Since the distribution of cell counts was non-parametric, we applied a one-tailed Wilcoxon test when comparing brain regions between control and SBI mice (Figure 6). We found 87 regions with significant differences. We observed that areas such as the primary somatosensory, auditory, entorhinal, perirhinal, posterior parietal and temporal association areas were significantly affected by the lateral impact of the SBI (Figure 1B). Other affected subcortical areas include the basolateral and posterior amygdala, the lateral, preoptic and periventricular zone of the hypothalamus, subthalamic nucleus, and zona incerta. Interestingly, we found significant damage in particular regions distant from the impact such as the pons and the dorsal raphe nucleus in the brainstem. Moreover, we also found that visual pathways relays, including the superior colliculus, preoptic nucleus, the lateral geniculate complex, and the pretectal regions were injured, as well as those from the auditory pathway such as the lateral lemniscus and cochlear nucleus. Lastly, we noticed significant degeneration in hippocampus, in the CA2, CA3 fields, subiculum and the retrosplenial area, critical to spatial learning and navigation tasks.

Figure 6. Quantification of neuron degeneration in SBI mice.

Representative brain section from a control(A) and SBI (C) mice displaying Fluoro-jade staining in the midbrain area (Bregma: −3.95 mm). Scale bar= 500 μm. Magnification of the region of interest (green rectangle) depicting the superior colliculus in control (B) and SBI mice(D). Scale bar=100 μm. (E) shows automated segmentation of cell counts represented as mean ± S.E.M where cell densities for SBI mice(n=6) were significantly higher than controls (n=3; one-tailed Wilcoxon test) *p<0.05, **p<0.01. PAG periaqueductal grey, PPN, pedunculopontine tegmentum, CA1 and CA3 area of the hippocampus, SUB, subiculum, SC, superior colliculus, and PG, pontine gray.

Given the differences in motor performance between SBI-H, SBI-S, and control mice, we further analyzed affected brain areas in these clusters, applying a Kruskal Wallis analysis (Figure S4A). Similar to the histological differences between the control and SBI group, we observed significant differences between the SBI-H and SBI-S in the visual and auditory pathways, parietal, hypothalamic, and entorhinal areas. Moreover, we noticed that within the brainstem, in addition to the dorsal raphe nucleus, we found significant differences in the ventral tegmental area(χ²(2)6.17 p=0.04), nucleus of the solitary tract(χ²(2)9.09 p=0.02), Barrington’s(χ²(2)6.17 p=0.04), trigeminal (χ²(2)6.36 p=0.04) and parabrachial nucleus(χ²(2)7.87 p=0.01). Interestingly, the SBI-H group, which exhibited the highest motor performance, displayed the highest degenerate-cell counts across the analyzed groups (Figure S4A). Although degenerating cells were found in multiple brain areas in the SBI-S group, these cell counts were not sufficient to show significant differences when compared to control group (p=0.0501; post-hoc Dunn’s test). This histological finding is consistent with our motor results.

All together, these injured areas reflect the consequences of a controlled lateral impact and record secondary blast brain injury in our rodent model.

3. Discussion:

We developed a secondary blast injury rodent model, which results in prominent cognitive impairment prevailing in male mice without visible signs of anxiety-like behavior. In addition, our model reveals distinct trajectories of motor recovery in our mice after trauma. High-performing SBI mice (SBI-H males and females) surpassed motor execution by control and SBI-S mice which recovered to previous baseline state (SBI-S group) on the accelerated rotarod paradigm. Low-performing SBI mice showed weaker performance. To our knowledge, this is the first time an animal model of blast injury has been shown to display a level of motor-plasticity where mice exceed the motor performance achieved prior to the blast-injury. Importantly, our rodent model is the first to emulate head injuries caused by secondary blasts. The incidence of this condition has increased from 16-21% in the last century to 43% in recent years, and models that help to understand and develop recovery therapies can greatly increase the quality of life and health for those exposed(Keller et al., 2015).

Overall, our SBI model mimics anatomical and functional damage previously reported by others when using a single localized blast (50-60 Psi) with physical and temporal dynamic forces associated with explosions experienced by humans(Elder et al., 2010). These forces have proven sufficient to produce brain injury due to rapid brain tissue deformation resulting in significant cognitive and motor deficit, visual damage, and widespread axonal injury(Guley et al., 2016). Thus, these lesions contribute to the development of chronic traumatic encephalopathy (Aldag et al., 2017). Results across studies of blast-induced TBI suffer from significant variability due to the blast wave (Aldag et al., 2017). However, since we used a micro-switch sensor to fix the animal’s head against the wall of the tube and always reach the same distance with the help of the motor stepper, we guaranteed our ability to consistently impact the target area in male and female mice. Our approach reduced variability among replicates, obtaining more reproducible results in male and female mice that allowed us to uncover and pinpoint diverse behavioral outcomes.

Mild, moderate, and severe brain injury caused by blast forces is associated with impaired concentration and memory in humans(Kaplan et al., 2018). Likewise, there are several reports demonstrating cognitive deficits during acute, subacute, and chronic injuries, mainly in rodent models. Previous reports assessed cognition by applying a variety of tests such as the Y maze, active avoidance response, and MWM(Cernak et al., 2011; Zhou et al., 2018). Similar to these results, our SBI model exhibited spatial memory deficits in the MWM, even when compared to a model with repetitive traumatic brain injury (Petraglia et al., 2014).This deficit was particularly accentuated in males due to impaired learning. In two related studies, although rodents showed significant delay reaching the platform in the MWM, they still exhibit linear learning despite prominent damage(Andersen et al., 2020; Shi et al., 2020). Our model found similar results, given the reduced latency and increased success rate of reaching the platform over time. In a recent study in which the time-to-event AFT model was applied to MWM(Andersen et al., 2020), the authors found significant differences on all swims on the last day of learning between the trauma and sham group. Although we found similar findings, our significance was limited to the last swim of the last day. Indeed, this study analyzed 107 rodents, a considerable number compared to our study, which provides increased power for this type of analysis.

The finding that our female mice have no evident spatial memory impairment is consistent with previous reports in which female mice performed better in cognitive tests than males after trauma(Spani et al., 2018). While we examined a small group of female mice, including control(n=3) and SBI(n=5), we noticed that areas such as the visual, temporal, entorhinal, and auditory cortical areas together with the hippocampus, hypothalamus and the primary somatosensory area showed increased counts of degenerating cells similar to SBI males. However, the differences did not reach significance when compared to controls (p=0.07) suggesting that injury was modest in the female SBI group.

A particular caveat of our study is that we used isoflurane as an anesthetic during the blast. O’Connor et al. found that isoflurane exerts a protective effect on cognitive outcome in females when performing a cognitive task after an impact-accelerated trauma(O’Connor et al., 2003). Future studies will address the protective effects of isoflurane and other anesthetics during blast injury using our animal model.

Consistent with our finding that spatial memory is impaired, fluoro-jade staining in our SBI model shows significant degeneration of the subiculum, hippocampal and retrosplenial area. Moreover, areas including brainstem, visual and auditory pathways were also affected as reported in studies in which rodents experienced blasts ranging between 9 to 60 PSI (Gilmore et al., 2016; Guley et al., 2016; Koliatsos et al., 2011; Wang et al., 2020). These overlap with injured regions in humans that are frequently observed as a result of explosive devices (Gondusky and Reiter, 2005).

In addition to these brain regions, increased degeneration is seen in areas such as the dorsal raphe nucleus and the ventral posterior nucleus of the thalamus, especially in our SBI-H mice. These are structures involved in chronic pain that are also associated with anxiety and depression symptoms. Although we expected traits of anxiety in our SBI model, the absence of these features suggests that lesions in the above mentioned areas improve depression/anxiety like behaviors as others have shown(Li et al., 2017).

The cerebellum is particularly vulnerable to single or repetitive blast injuries (Koliatsos et al., 2011; Mac Donald et al., 2013). Indeed, cerebellar Purkinje cells, ventral cerebellar lobes, and cerebellar white matter tracts are particularly affected by blast injury(Meabon et al., 2016). Based on this evidence, and understanding that the rotarod evaluates balance and coordination, we expected motor deficits in the rotarod test in all animals exposed to SBI. We observed a cohort (SBI-L mice) that shares these characteristics and mimics the findings of a study in which the rotarod’s performance worsened a week after blast injury (Koliatsos et al., 2011). In a single SBI-L subject, in which we performed a histological analysis, we found significant damage of the Crus 1 area (p=0.03). This structure is a cerebellar area involved in cognitive and visuomotor functions(Sugihara, 2018). Additionally, we distinguished a group (SBI-S) in which the rotarod’s execution was restored to baseline conditions. This is a condition that other trauma models have reported (Cernak et al., 2011). Importantly, we noticed male and female cohorts with exceptional motor performance (SBI-H), a finding that contrasts with the studies. above.

Several studies have shown that repetitive training enhances motor performance after brain injury in TBI rodent models. Bilateral movement training has been seen to promote axonal remodeling of the corticospinal tract and recovery of motor function following TBI(Nakagawa et al., 2013). Though regeneration was seen in the cortico-spinal tract, rotarod training only led to limited recovery in rodents. Hence, the regeneration of a single tract seems insufficient to promote the plasticity observed in our model. We postulate that the rewiring of multiple tracts is necessary to reach the level of motor performance seen in male and female mice in our model. Moreover, areas such as the ventral tegmental area, subthalamic nucleus and the zona incerta are significantly damaged in our SBI-model, likely altering overall basal ganglia function. We speculate that the cerebellum may be exerting a compensatory role after trauma. Future studies will address this compelling hypothesis. Certainly, our model may provide insight toward optimizing the injured brain to maximize the efficacy of rehabilitation(Kochanek et al., 2019).

Although the majority of blast injuries in humans result in motor impairment, there are a few cases in which savant motor skills have emerged after brain injury. It has been speculated that enhanced local connectivity in particular areas is responsible for savant skills (Hughes, 2010). Our model may serve to elucidate mechanisms inducing acquired savant syndrome.

Limitations:

Although we are aware that humans suffer severe injuries in the head and body after an explosion, we chose to inflict localized secondary blast injury to the brain to increase the likelihood of survival and assess the functional neurological deficits that result from secondary blast injury. We acknowledge that by confining the blast to the brain, we are partially mimicking the effects of SBI.

Along with the study, the low-performance SBI group showed the highest mortality numbers, and as a consequence, we gathered lower numbers in this group. Similarly, tissue damage in these animals precluded histological analysis of the whole brain in this group.

Conclusion:

Our novel SBI rodent model displays a variety of cognitive and motor disabilities comparable to humans exposed to blast injuries caused by improvised explosive devices. Given the chronic nature of injuries, as well as the reproducibility of diverse outcomes in male and female mice, this model will be useful for discovering the neural basis of recovery from SBI in both groups. Further studies using this model will also deepen our understanding of structural and functional differences in our low and high-performing SBI-exposed mice. Understanding such variations in recovery from SBI may serve to promote systemic restoration of motor performance during rehabilitation of patients.

4. Experimental Procedure:

4.1. Animals

This study was conducted in the Belfer Research Animal Facility using 8-10-week-old C-57BL6/J mice (males and females) from Jackson Labs. All use of laboratory animals was consistent with the Guide for Care and Use of Laboratory Animals and approved by the Weill Cornell Medical College IACUC (Protocol No. 2016-0054). Our animals were maintained in a gated animal facility in which the inhabited cages were kept in cubicles with a reverse light cycle; the lights within the cubicle turned off at 9:00h and turned on at 21:00h. The animals were given unrestricted access to food and water, and were fed according to Weill Cornell Medical College IACUC’s standard diet. The temperature of the cubicles was maintained at 23°C. Animals were initially housed in groups within their home cages. After trauma, we singled house mice (including SBI and sham group) to closely evaluate behavior of animlas as described below in the recovery section (4.2.1). All animals were provided with the enrichment material, EnviroPak. Each cage received a 7-ounce EnviroPak every month. EnviroPaks consist of Envirodri crinkled paper encased in a paper bag for animals to build their nests. All procedural and behavioral rooms within the facility were soundproof and devoid of external odors.

4.2. Secondary blast injury device

We designed a device that transmits blast waves and mimics the impact of airborne debris as a result of the blast wave. The air blast was delivered using a 6-gallon portable electric air compressor (porter-cable) attached to a 16-gauge nail gun (nails were removed). The shrapnel energized by the blast wave was represented by a piston propelled by the air blast triggered by the nail gun. The piston penetrated into an area of 2 x 2 mm located at the mesocircuit (Schiff, 2010) which corresponds to the frontal/prefrontal cortical-pallidal thalamocortical loop systems (Figure 1B). Half of the individual’s body (Figure 1) was placed in a 50 ml Falcon tube perpendicular to the gun and surrounded with molded foam to hold the head steady during the head impact. A window of approximately 7.5 mm in diameter was made in the tube to reach the parietal region of the right side of the animal’s head (Figure 1A). The tube opening was aligned with the gun’s piston using a X-Y-Z control stage in which the body of the animal laid over (Figure 1D). A micro-switch sensor (Figure 1A) was placed near the targeted brain area to fix the head against the wall of the tube and control the distance traveled by a motor stepper in which the gun rested. We used an Arduino to accurately control the rotational speed of the stepper motor and assure that the piston reached a consistent brain-piston distance of 1 mm. Prior to the impact, mice were induced with isoflurane (3% vol.), weighted, and injected with flunixin 5 mg/Kg. Opthalmic ointment was applied to the animal’s eyes. Animals remained anesthetized using isoflurane (1.5% vol.) via an anesthesia mask located at the end of the tube (Figure 1A). Animal temperature was monitored and maintained using a temperature controller (CWE incorporated) set to at 36 ± 0.5 °C. The air blast pressures were adjusted to 50 psi for males and 45 psi for females. Controls were subject to the same induction procedure as the SBI mice and placed in the blast apparatus. The blast was administered several centimeters away from the head of the mouse, without direct impact to the control animal. After the procedure, mice were immediately transferred to a clean cage to rest under a heat lamp for a half hour.

4.2.1. Recovery and monitoring of animals

After the experimental procedure, animals were monitored at least three times each day for a week, weighed to monitor their condition, and fed manually with an oral gavage syringe when necessary. Additionally, moistened food pellets were placed in a container at the bottom of the cage to facilitate eating. Flunixin meglumine (1.5 mg/Kg) and saline solution (0.9%) were administered subcutaneously each day (max. saline volume 1-2 mL per 24 hr). Urination and defecation were also recorded for each animal to monitor their status. Body temperature was also tracked, and mice with low body temperatures placed under heat lamps to maximize recovery. Signs of pain, respiratory distress and dehydration were closely monitored. When necessary, humane treatments were provided according to IACUC’s guidelines.

4.3. Morris water maze

The Morris Water Maze (MWM) task assessed the learning and memory capabilities of male and female mice 7 days post-SBI. All testing occurred between the hours of 9:00 and 13:00. Training and testing phases were performed at the same time of day. The MWM pool had a diameter of 120 cm and a height of 56 cm. The pool was conceptually divided into 4 quadrants of equal size and shape, labeled Quadrant 1, 2, 3, and 4 (Figure 3A). Proximal and distal visual cues were placed and maintained in the same locations for the remainder of the task. The moveable platform had a diameter of 10 cm and was placed 30 cm away from the wall of the pool. The water temperature was maintained at 20°C throughout the task. Acclimation took place on the first day of experiments, followed by 4 days of the learning phase and 4 days of the probe phase. During the acclimation and learning phases, a platform was placed in Quadrant 1 (Figure 3A). On the acclimation day, the water in the pool was left clear and a flag was placed on the platform to guide the animals. EthoVisionXT software was used to track the animal’s movement and speed across the different quadrants of the maze, and to collect data on the latency of the animals to find the platform. The trials were set to last for 90 seconds. If the animal did not find the platform within the allotted 90 s, it was placed on the platform for 10 s. If the mouse found the platform within the 90 s, the animal was left in place for 5 s before transferring it back to its recovery cage, which was supplied with heat and paper towels. Each animal completed 4 trials, starting each trial in a new quadrant and placed facing the wall of the pool. The learning phase began the following day and lasted for 4 days. For the learning phase, non-toxic temper paint was added to the pool and mixed so that the water became opaque and hid the platform. The flag was removed from the platform for the remainder of the experiments and sufficient water was added to cover the platform. Each mouse ran 4 trials each day with these configurations, for the next 4 days. A resting intertrial interval of 4 minutes was provided to each animal.

To examine cognitive flexibility, the location of the platform was moved to Quadrant 3 (opposite of Quadrant 1; Figure 3A) on the sixth day of experiments, otherwise known as the probe phase. Probe experiments were conducted just like the learning phase, with each mouse running four 90 s trials each day for the next 4 days.

The task was conducted in groups (four mice). After each group completed all four rounds of trials, the cages were cleaned and wiped with Clidox-S Dilution 1:18:1 (chlorine dioxide liquid) before the next group of mice was brought into the room. Between individual trials, the pool was cleaned of visible mouse droppings.

Mice were excluded from the MWM analysis if they showed external damage to the eye or if they exhibited anxiety-induced thigmotaxis consistently throughout the experiments. A total of 2 (2 female) mice were excluded from the study based on these criteria.

4.4. Rotarod

The Rotarod task (RR) was used to assess motor performance in mice post-SBI. The Rotarod (Rotamex-5, Columbus Instruments) was configured to increase in speed from 0 - 40 cm/s over the course of 400 seconds, accelerating at 0.1 cm/s throughout the trial. RR testing began in conjunction with the MWM task, 7 days post-SBI (Figure 1E-Experimental timeline). Experiments started 2 hours after completing the MWM. For 10 consecutive days, mice performed 5 daily trials of the RR. A resting interval of 2 minutes was provided between trials, consistent with (Fremont et al., 2014). The enclosures were cleaned using Clidox-S Dilution and dried between trials. Run time, the length of time until the animals lost control of coordinated limb movements, was recorded using the Rotamex laser system and used to measure RR performance. Mice were excluded from the RR analysis if entire days of data were missing due to errors in data collection caused by the equipment.

4.4.1. Pretraining group

Baseline performances on the RR task were assessed in male and female mice before SBI induction. Animals were trained until they plateaued during the motor learning task (Figure 1E). The experiments were conducted each day using the protocol mentioned in Section 2.4. Animals were tested again 7 days post-SBI for 9 consecutive days.

4.5. Open field

Exploratory and anxiety-like behaviors were assessed using the open field test (OFT) in males and females. OFT trials were conducted after MWM experiments were completed (Figure 1E). OFT trials were conducted from 13:00 to 16:00. To preserve reverse cycle conditions, experiments were conducted in a room illuminated by red light. Animals were habituated to the Photobeam Activity System (PAS-OF, 16”x16”x15” plexiglass enclosure, San Diego Instruments, San Diego, CA) for 30 minutes the day before starting the experiments. Each day for 3 days, animals were placed in the center of the enclosure and their activity was recorded by the PAS-OF software for 10 min. Measures included the time spent in the center versus the periphery of the enclosure, as well as speed and distance traveled. The enclosures were cleaned using Clidox-S Dilution.

4.6. Histology

Following the completion of the behavioral tests, mice were anesthetized with isoflurane and intracardially perfused using 0.1 M phosphate buffered saline (PBS) and 4% paraformaldehyde (PFA) in PBS. Brains were extracted and submersed in 4% PFA overnight at 4°C and transferred to a 30% sucrose solution until equilibrated. The tissue was coronally sectioned with a thickness of 25 μm using a freezing sliding microtome (Leica Biosystems RM2245). Sections were stored in PBS at 4°C until mounted. The sections were mounted onto slides.

4.6.1. Fluoro-Jade staining

To assess neuronal degeneration, we initially dried the slides in an oven at 50°C for 30 min. Slides were then dehydrated in solutions of 0.2% sodium hydroxide in 80% ethanol for 5 min, 70% ethanol for 2 min, and rinsed in distilled water (dH2O) for 2 min. To reduce background noise and preserve the quality of the fluorescence signal, slides were treated in 0.06% potassium permanganate for 10 min and rinsed in dH2O for 2 min. Sections were submersed for 15 min in the Fluoro-Jade B (FJB) staining solution, containing 0.0004% FJB (Histo-Chem, Inc.) and 0.0002% 4′,6-Diamidino-2-Phenylindole, Dilactate (DAPI; Invitrogen) in 0.1% acetic acid vehicle. Slides were rinsed in dH2O for 1 min, followed by two more dH2O rinses for 2 min each, then dried in the oven at 50°C for at least 10 min. When the slides were visibly dry, they were cleared in xylene for 1 min. Slides were coverslipped using Krystalon mounting medium (Harleco/EMD Millipore).

The slides were digitized using the Pannoramic MIDI digital slide scanner (3D HISTECH Ltd., Budapest, Hungary) with a x20/NA0.8 Zeiss Plan Apochromat objective (Carl Zeiss Microimaging, Inc, Thornwood, NY) and a sCMOS camera (2048 x 2048 pixels, with a pixel size of 0.325 μm; pco.edge 4.2, PCO GmbH, Kehlheim, Germany). Fluorophores were detected at the following wavelengths: 405 nm (DAPI) and 525 nm (FJB) using corresponding filters (DAPI, 39 ms exposure; FITC, 120 ms exposure).

Individual FITC composite sections for each brain were extracted from the microscopy slide files in CaseViewer and converted to TIFF format. The sections were then loaded into RStudio to be converted to grayscale. Using WholeBrain software(Furth et al., 2018), bregma coordinates were assigned to brain sections. We then manually set parameters for segmentation to ensure detection of true signal in single sections, and automated the segmentation and registration for the remaining sections. We calculated the signal-to-noise ratio (SNR) for each detected cell per section. We found the maximum and minimum pixel intensity for each brain, and normalized pixel intensity for all the detected cells using the formula N = i - x/m, where N = normalized value, i = pixel intensity of a detected cell, x = minimum pixel intensity, and m = interval between the minimum and maximum pixel intensity. Histograms of the SNR and pixel intensity were plotted for each brain, with any detected cell with a threshold of SNR < 1.3 or a normalized pixel intensity < 0.2 deemed to be noise. Any detected cell with SNR or pixel intensity values below these values were filtered from the master dataset. We then extracted a list of 859 regions of interest from our master data frame that were detected in our quantification. We used the “roi.cell.count” command to assess total cell counts for each animal (n=17) and within each region. We then used the “data.summary” command to calculate the average cell count for each region in each mouse group. We determined that the data was non-parametric using the Shapiro-Wilkes test for normality and performed a Kruskal-Wallis test to determine which regions had significantly different cell counts among groups. When necessary, we also applied a pairwise post-hoc Dunn’s test to determine significant differences between specific pairs of groups; p-values were corrected using Benjanini Hochberg correction.

4.7. Statistical analysis

A total of 132 male C57s were either subjected to the blast injury (SBI, n=91) or to the sham/control procedure (control, n=41). Of the SBI males, 16.5% died immediately after the blast (n=15) and 17.6% died during the recovery period (n=16). Four control males died from unknown complications. These experiments were conducted in 8 different replicas between June 2017 and July 2019.

A total of 42 female C57s were subjected to SBI (n=27) or to the sham procedure (n=15). Of the SBI females, 16.1% died immediately after the blast (n=5) and 16.1% died during the recovery period (n=5). No control females died during our experiments. 3 different replicates of these experiments were conducted between June 2017 and November 2019.

Individual research that administered the blast participated in the recovery and monitoring phase. However, they did not work on behavioral testing. Mice were randomly assigned to SBI and control groups. After SBI, we identified each subject with a number, so that individuals who performed the behavior tests remained blind to the condition. Note that on occasion, SBI animals were evident due to poor overall performance or because animals drastically increased execution. We have different sequences of entry points in MWM. Similarly, we did not follow a particular order when conducting rotarod or open field tests. Notably, the researchers that conducted PCA, template clustering, and cell-count quantification were blind to the experiment being assessed and the treatment condition.

4.7.1. Clustering of SBI and control mice.

The rotarod test evaluated the impact of trauma on motor coordination. We measured the length of time a subject spent running on a rotarod before falling off. We carried out five trials per day per animal. We denoted $x_i=[x_i^0,x_i^1,{\text{x}}_{\text{i}}^2,{\text{x}}_{\text{i}}^3,]$ as a 4-dimension vector for animal i that contained the time length on rod (averaged over 5 trials) from day 0 to day 3. We first pooled x_i for both groups and applied principal component analysis (PCA) to visualize the datapoints using the sklearn package in Python. In our case, the first two principal components (PCs) explained 86% of the variance. When analyzing the control group, we calculated the Euclidean distance between control-data points and the centroid (which equals the mean value) and plotted a histogram. The integer that enclosed the majority (>90%) of controls was considered as the radius of the circle that enclosed controls. We removed two subjects with z-scores greater than two. Using the obtained circle as a reference, SBI (PCs) were classified in three groups. If the distance was less than the radius of control PCs in our SBI group, we identified the cluster as “standard”; otherwise we labeled it “low” when the 1^st PC was more negative and greater than the radius and “high” when 1^st PC was positive and greater than the radius.

For the pretraining experiments, we assigned the motor performance of each subject to templates for clustering. We chose $x_i=[x_i^{-3},x_i^{-2},x_i^{-1},x_i^1,x_i^2,x_i^3]$ a 6-dimension vector that encompassed 3 days before and 3 days after trauma, x_i was normalized by removing the $\mathrm{offset}=x_i^{-3}$ so that x_i started from zero. For control animals, x_i was roughly a plateau. We then clustered SBIs through a template matching approach. We first defined x₁ as template 1. Then from x_j(j > 1) we did the following: we examined x₁,x₂,…,x_j−1, if there existed a subject highly similar to x_j such that each pairwise element has an absolute difference <10, we assigned x_j the same template as that animal. Otherwise, we defined x_j as a new template. We obtained three combined units: low performer (SBI-L), high performer(SBI-H) and the standard performer(SBI-S).

4.7.2. Data exclusion for statistical analysis

Certain trials had to be excluded due to errors in data collection by the Ethovision software. For cases in which the Ethovision data was determined to be unreliable, the manually recorded escape latencies were used, while the quadrant analysis and speed data were excluded.

4.7.3. Behavioral studies analysis

Results from the quantitative post-trauma observations and behavioral experiments were analyzed using Origin(Pro) 2020 (OriginLab, Northampton, MA). For all analyses, statistical significance was established at p<0.05. A two-tailed Welch’s t-test was used to compare the righting times between control and blast-injury groups. A two-tailed paired t-test was used to compare weight changes within groups. For the Morris Water Maze, we examined the learning phase by applying the accelerated failure time to event model (AFT). This approach defines latency to reaching the platform as a time-to-event problem. We used the log-logistic distribution model as suggested by(Andersen et al., 2020). The differences between control and SBI groups by swim per day were estimated by Hommel-adjusted contrasts. We ran all parameters (percentage of time in quadrant, and velocity) through a two-way repeated measures ANOVA, with group (control, SBI) as one factor and day as the repeated measure. To compare the time spent in each quadrant, we used a three-way ANOVA with quadrant (1,2,3,4), group, and day as the factors. For the Rotarod, we analyzed run times using a two-way repeated measures ANOVA with group (control, SBI-L, SBI-S, SBI-H) as one factor and day as the repeated measure. The Rotarod data violated the assumption of sphericity, thus the two-way repeated measures ANOVAs were corrected using either the Huynh-Feldt (RR) or the Greenhouse-Geiser (pretrained RR) correction.

The adjusted F- and P- values were reported in the text accordingly. Distance traveled and speed measures from the Open Field Test were analyzed using one-way ANOVAs with group (control, SBI) as the factor. The amount of time spent in each zone was analyzed using a two-way ANOVA with zone (center, peripheral) and group as the factors. All reported ANOVAs were corrected using the Bonferroni-adjustment.

4.7.4. Cell counts of cell degeneration

We calculated the average cell count for each region per mouse group (Control vs. SBI) and determined that data was non-parametric for the majority of regions using the Shapiro-Wilkes test for normality. We performed a Kruskal-Wallis test to determine which regions had significantly different cell counts between the Control and SBI groups. When necessary, we also applied a pairwise post-hoc Dunn’s test to determine significant differences between specific pairs of groups; p-values were corrected using Benjanini Hochberg correction. The full list of significant brain structures is available at Figshare. 10.6084/m9.figshare.12900017

Highlights.

1. Our novel secondary blast injury model exhibits cognitive and motor impairment comparable to human subjects.

2. Impaired cognition prevails in male mice after secondary blast injury (SBI)

3. SBI model displays distinct trajectories of motor recovery