Importance of Control Groups for Evaluating Long-Term Behavioral and Cognitive Outcomes of Controlled Cortical Impact in Immature Rats

Study and group design

The study was approved by the Animal Care and Use Committee of the Johns Hopkins University School of Medicine (Baltimore, MD) and in compliance with the National Institutes of Health (NIH) guidelines. Surgeries were conducted at the postnatal Days 10-11 (PND) in Sprague Dawley rats. Only male pups were used in this study. Rats were randomly assigned to four different treatment groups: Naïve controls (n = 14), Anesthesia only group (n = 12), Craniotomy only group (n = 17), and TBI (n = 17). The study design is shown in Figure 1. After weaning from dams, rats were housed in standard rat cages and kept in groups of three per cage for up to 2 months of age, and, later, in groups of two per cage. After the end of behavioral testing, rats were anesthetized, sacrificed and their brains harvested. Detailed methods and procedures can be found in the Supplementary Methods.

OPEN IN VIEWER

FIG. 1.

Timeline of experimental procedures. Traumatic brain injury (TBI) was performed on the left hemisphere (see details in the “Methods” section) at post-natal Day 10 (PND 10). To assess for possible TBI induced motor deficits, Foot Fault and Gait Analysis tests were performed at PND 25-26 and PND 27-P30, respectively. Novel object recognition was conducted at PND 40-50 to test recognition memory. Then, a number of water maze protocols were employed to test for spatial reference memory as well as episodic-like and working memory at PND 65-100. Finally, possible alterations in social interactions were tested at PND 120-140 using social recognition and social motivation tasks. Color image is available online.

Controlled cortical impact (CCI) injury

TBI was performed using a controlled cortical impact (CCI) device, as previously described. ^{10 -12} Injury was produced using a 3-mm metal impactor tip with a depth of penetration of 1.5 mm, at velocity of 5.5 ± 0.3 m/sec, and a duration of deformation of 50 msec. At the completion of surgery, isoflurane was discontinued. When the rats emerged from anesthesia, they were returned to their cages with their littermates and mother.

The rats assigned to the Craniotomy group underwent identical surgeries, with the exclusion of the CCI. The rats assigned to the Anesthesia group received isoflurane for the same length of time as the TBI and Craniotomy groups (30 min) without any surgical interventions. Representative brain images from each of the groups are presented in Supplementary Figure S1 demonstrating the extent of brain damage in the TBI group and preserved brain tissue in the Craniotomy group, as expected from our previous studies. ^{10 -12}

Behavioral testing

Behavioral testing started when rats reached 25 days of age in the order described below (Fig. 1). The battery of tasks included tests to access multiple domains affected in children with TBI: locomotor function and coordination, different types of learning and memory, and social interactions (social recognition and social motivation). ANY-maze software (Stoelting Co, Wood Dale, IL) was used to record data in all tasks unless noted otherwise.

Foot fault test

The test was conducted at PND 25-26, 2 weeks after surgery. Animals were placed on an elevated, horizontal metal grid with a mirror underneath for video recording of the paw placement. The number of foot slips for each limb over 5 min was analyzed during video playback.

Gait analysis

A runway test was conducted at PND 27-33 using a 1 m long enclosed glass runway using Gait Scan software (Clever Sys, Reston, VA). The following parameters were analyzed: foot pressure, foot angles, run speed and distance, step base, step length, and whole-body length.

Novel object recognition (NOR)

NOR test was conducted at PND 40-50. Rats were habituated to the box (60 × 60 cm) twice a day 10 min each for 2 consecutive days. In Session 1, two similar objects from LEGO^® were placed in the box and rats were allowed to investigate the objects for 5 min. On the next day, session 2 was conducted when one of the objects was replaced with a new one, and rats were allowed to investigate the objects for 5 min.

Morris water maze

Morris water maze tasks were performed at PND 65-100 as described before ¹³ with modifications (see the Supplementary Methods for detailed descriptions of the protocols).

Social interaction testing

Statistical analysis

The data were analyzed using the statistical package STATISTICA 13 (TIBCO Software Inc., Palo Alto, CA) and a minimal level of significance p < 0.05. The data were analyzed using mixed design analysis of variance (ANOVA) with Treatment (Naïve, Anesthesia, Craniotomy, TBI) as an independent factor and repeated measure(s) (RM; trials, blocks, arms, etc.). The least significant difference post hoc test was applied to significant Treatment or Treatment × RM interactions to evaluate differences between four sets of means: Naïve versus Anesthesia; Naïve versus Craniotomy; Naïve versus TBI; and Craniotomy versus TBI. Two-tailed t-test was used to analyze differences from chance levels. The details of statistical results including F, df, and post hoc tests are presented in Supplementary Tables S1-S6. Data in figures represent means ± standard error of the means unless otherwise noted.

Locomotor function in foot fault test and gait analysis

The assessment of locomotor function started in a foot fault test conducted 2 weeks after TBI when rats reached the age of 25-26 days (PND 25-26; Fig. 1). Rats from the Anesthesia group were not significantly different from Naïve controls in any of the measures. The Craniotomy group demonstrated more errors than Naïve controls for the rear paw on the side contralateral to craniotomy (right rear paw, p < 0.0001). TBI rats had more misplacements than Naïve rats for both rear paws (Fig. 2A; Supplementary Table S1). For the front paws, both Craniotomy and TBI groups showed a notable tendency making more missteps as compared with Naïve rats; however, these differences did not reach significance.

OPEN IN VIEWER

FIG. 2.

Traumatic brain injury (TBI) and Craniotomy rats show deficits in the Foot Fault test (A) but not in Gait analysis (B, C). (A) Number of foot placement errors in the foot fault test conducted 2 weeks after TBI. Single and double asterisks indicate significant differences from Naïve group at p < 0.01 and 0.0001 levels, respectively, as a result of least significant difference post hoc test applied to significant Group × Paw interaction [analysis of variance, F(9,171) = 2.68, p < 0.0061]. (B, C) Width of step base (B) and step length (C) assessed in Gait analysis conducted at the 3rd week after TBI. There were no significant differences between any of the groups. See Supplementary Table S1 for details of statistical analyses. Color image is available online.

The gait analyses were conducted in a runway test in the 3rd week after TBI when rats reached age of PND 27-33. Different variables including foot pressure, foot angles, run speed and distance, foot spacing, and whole-body length where analyzed, and no significant treatment-related changes were found. Results for the width of step base and step length are shown in Fig. 2B, 2C. As expected, a significant effect of front versus rear paws was observed for the variable of step base (Fig. 2B) with the rear paws having wider step than the front paws (Supplementary Table S1).

Object recognition memory

The object recognition memory was tested in PND 40-50 rats at Week 5 after surgeries. In Session 1, no significant differences in preferences investigating two identical objects were found in any of the treatment groups (Fig. 3B; Supplementary Table S2). After a 24-h delay, Session 2 was conducted when one of the objects was exchanged for a novel one (Fig. 3A, 3B). As expected, rats from the Naïve control group investigated the new object significantly longer than the previously presented object (Fig. 3B). Rats from the Anesthesia group demonstrated significant preferences to the new object, similar to the naïve control group. However, Craniotomy and TBI groups showed no preferences between the new and old objects (Fig. 3A, 3B; Supplementary Table S2), implicating deficits in the object recognition memory.

OPEN IN VIEWER

FIG. 3.

Traumatic brain injury (TBI) and Craniotomy rats show deficits in the Novel Object Recognition (NOR) test. (A) Representative examples of exploratory activity in rats from each treatment group (Naïve, Anesthesia, Craniotomy and TBI) during investigation of a new (upper right corner, indicated as 1) and familiar object (low left corner, indicated as 2). (B) Duration of objects' investigation in Session 1 (two similar familiar objects, thin lines) and Session 2 (one new object in position 1, thick lines). No preferences to the objects were observed when two similar objects were presented in Session 1. In Session 2, when one of the objects was replaced by a novel object, only Naïve and Anesthesia groups spent more time investigating the novel object. Asterisks indicate difference between the novel vs familiar object in Session 2 (p < 0.05). Pound signs indicate differences in investigation at a particular location between Sessions 1 and 2 (p < 0.015). (C) Distance travelled during Session 1 (similar familiar objects) and Session 2 (one of the objects is novel). Pound signs indicate differences between Sessions at p < 0.002. See Supplementary Table S2 for details of statistical analyses. Color image is available online.

Analysis of distances travelled during object recognition testing showed that Craniotomy and TBI groups travelled longer distances in Session 2 as compared with Session 1 (Fig. 3C). Thus, in addition to significant deficits in object recognition memory, Craniotomy and TBI groups demonstrated behavioral activation during presentation of the novel object.

Reference memory in a classic Morris water maze

Cognitive testing in a classic version of the Morris water maze started at Week 10 after surgeries when animals reached approximately 11 weeks of age (Fig. 4A-C). This task assays spatial reference memory for a hidden platform that stays in the same position across trials and days. ^14,16 As training progressed, latency to find the platform decreased in all treatment groups (ANOVA, effect of trial p < 1.0E-6; Supplementary Table S3). There were no significant between-group differences in swim speed (data not shown). Despite the training-related improvements in the escape latencies, rats from the TBI group required significantly longer time than Naïve controls to find the platform (ANOVA, effect of Treatment, p < 0.012; Fig. 4A). Craniotomy group was not different from Naïve controls in this measure if averaged across the training (Fig. 4A insert); however, in contrast to Naïve and Anesthesia groups, an overnight delay significantly increased latency to find the platform in Craniotomy rats (Fig. 4A). This data indicated that spatial learning abilities in Craniotomy group were relatively preserved but deficits occurred in long-term retention of platform location.

OPEN IN VIEWER

FIG. 4.

Traumatic brain injury (TBI) and Craniotomy rats demonstrated deficits in the Reference (A-C) and Episodic-like memories (D-F) in Morris water maze. (A) Latency to find a hidden platform in training trials over a 2-day training. Insert shows significant effect of treatment for average latency (p < 0.012). Asterisk and triangle indicate differences from Naïve controls and Craniotomy groups (p < 0.01 and 0.05, respectively). Pound signs denote difference between the trials separated by a 24 h-delay (p < 0.02). Numbers above the X-axis indicate the order of probe trials during the testing. (B) Time (%) spent in the area around the position of the platform in the course of testing in probe trials. Pound signs indicate significant difference from the chance level of performance (p < 0.005). (C) Spatial preference to the position of the platform shown as time spent in the areas located symmetrically in the maze. Asterisks indicate differences from Naïve controls (p < 0.001). (D) Latency to reach the daily new platform position averaged over 3 days of reversal training. Insert shows effect of treatment for average latency. Triangles indicate differences from Craniotomy groups (p < 0.013). (E) Time (%) spent in the area around the previous platform location tested in a probe trial after a 24-h delay. Pound signs indicate significant difference from the chance level of performance (p < 0.005). (F) Time (%) spent in the area around the new position of the platform during consecutive reversals. Spatial preferences were tested in a probe trial 30 min after the last training trial. Solid lines in (B, C) and (E, F) show chance level of preference (18%). Color image is available online.

To assay the development of spatial preferences, the probe trials were conducted, in which the platform was initially lowered and then made available after a variable interval (40-45 sec). The good performance in the probe trials required not only spatial memory for the location of the platform (as in the training trials), but also persistence in looking for the platform in the appropriate location. This complexity led to longer dynamics of learning in the probe trials. Specifically, in Naïve controls, learning the spatial location of the platform (as judged by the escape latency in the training trials) reached the asymptotic performance after the first 4-5 trials, whereas the appropriate spatial preference was observed only at the end of the 2nd day of testing (Probe 3; Fig. 4B). All treatment groups reached spatial preferences above chance level (Probe 3; Fig. 4B, 4C). Nevertheless, TBI and Craniotomy groups showed less efficient performance than Naïve controls (Fig. 4C; p < 0.005). In addition, testing of long-term reference memory after a 24-h delay demonstrated the loss of spatial preferences in the TBI and Craniotomy groups (Fig. 4E, Previous Platform for Reversal 1).

Episodic-like memory in Morris water maze

During the next 3-day long training, the platform position was changed daily, while the trial schedule as well as spatial cues around the water maze were kept the same. TBI rats showed longer latencies finding the platform as compared with Naïve controls (ANOVA, effect of treatment p < 0.013; Fig. 4D). Probe trials conducted at the end of each reversal indicated that all treatment groups were able to show spatial preferences to the new position of platform (Fig. 4F). These data indicated that deficits in spatial reversal learning observed during training trials in TBI rats did not prevent forming new spatial memories. This was documented in the probe trials with a short 30-min-long delay (Fig. 4F). However, these memories were not retained, as TBI rats demonstrated no spatial preferences when tested in the probe trials after a long over-night delay (Fig. 4E). Importantly, in contrast to TBI, Naïve controls showed significant preferences to the platform location of the previous day of testing (Fig. 4E). This data implicated that performance of Naïve controls in this task required not only learning the new platform but inhibiting visits to the old platform location as well. In the course of repeated reversals, the preferences of Naïve controls to the previous platform decreased to the chance level (Fig. 4E). In the last reversal, when learning the new platform position in Naïve controls was not complicated by interferences with memory of the previous platform location, their performances in the final probe trial were highly efficient (Fig. 4F) and similar to those in the reference memory task (Fig. 4B, 4C). This stage of repeated reversals revealed significant deficits in TBI and Craniotomy groups, as both groups demonstrated less accurate spatial preferences than in Naïve controls (Fig. 4F). Anesthesia group showed performances statistically indistinguishable from Naïve controls.

Episodic-like memory tested in the Radial Water Maze

Similar to reversal training in MWM, episodic-like memory in the Radial Water Maze (RWM) was tested by changing the location of the platform daily. During the 1st day of training (Fig. 5A), Naïve and Anesthesia groups demonstrated a one-trial learning; namely, the level of their performances in the latency and number of errors reached asymptotic levels already at the 2nd training trial (Fig. 5B, 5E; Supplementary Table S4). Craniotomy group showed a mildly delayed learning as they reached their asymptotic performance in the latency to find the platform at the 4th trial. TBI rats were not able to demonstrate the one-trial learning and improved incrementally up to the 4th trial.

OPEN IN VIEWER

FIG. 5.

The Radial Water Maze task revealed deficits in spatial learning and memory in traumatic brain injury (TBI) rats. (A) Scheme of the Radial Water Maze apparatus. The hidden platform was placed at the end of one of the eight arms and its position changed daily. Each of the six daily trials started randomly from different arms with exception of the arms baited on the current and previous testing day. Set of multiple spatial cues placed on the walls of the room was kept the same as for the previous water maze tasks. (B) Latency to find the hidden platform in the 1st testing day. Naïve and Anesthesia groups (but not Craniotomy and TBI) demonstrated a one-trial learning. Asterisk and triangle indicate a significant impairment in TBI group in Trial 2 as compared with Naïve control (p < 0.0001) and Craniotomy (p < 0.018) groups, respectively. Insert shows group averages for latencies in Trials 2-6 of the training. Asterisk and triangle indicate significant deficit of TBI group as compared with Naïve controls (p < 0.002) and Craniotomy (p < 0.017) groups, respectively, as a result of planned comparisons applied to a significant effect of Treatment [analysis of variance (ANOVA), F(3,55) = 8.07, p < 0.0002].

Starting from the 2nd day of testing, rats were required to learn a new position of the platform. In the orientation trial (Trial 1), rats from all groups visited the previous platform location demonstrating normal long-term spatial memory (Fig. 5G, 5H, Platform 1). Independent of the treatment condition, about 30% of errors in the orientation trial were made due to repeated visits to the previous platform location (Fig. 5H, Trial 1). Interestingly, the orientation trial in TBI rats was significantly shorter than in Naïve controls (Fig. 5C, Trial 1) while the number of errors was not different (Fig. 5F, Trial 1). Indeed, the rate of arm choices (the number of arm choices per sec; Fig. 5D) was significantly increased in TBI rats as compared with control. Craniotomy rats showed significantly higher rates of arm choices in the orientation trial as well (Fig. 5D, Trial 1).

After the orientation trial, TBI rats showed significantly longer latency and more errors to find the platform than the Naïve controls. Anesthesia and Craniotomy groups demonstrated performance comparable to controls. When memory for the platform was assessed after a 24 h-long delay (in Trial 1 of the next testing day), TBI rats made more errors before reaching the previous day's platform location than controls (Fig. 5G, Platforms 2-4). Despite this impairment, TBI rats visited the previous platform location more often than expected from a chance level across all 6 trials of testing (Fig. 5H). These data indicate that the deficit seen in TBI rats after the 24-h delay is not due to an impairment in long-term memory, but likely due to a deficit in its immediate recall during Trial 1. When given more time/trials, TBI rats demonstrated their spatial preferences to the previous platform. In fact, the inability of TBI rats to inhibit such preferences after the orientation trial might be a reason for their less efficient performance than in control groups.

Spatial and goal-directed strategies in the Radial Water Maze

All groups have learned the RWM task creating conditions for a more challenging test. At the next step of testing, a new type of trials was introduced in which the position of platform was prompted by a highly visible cue at the end of the arm (Fig. 6). Positions of the cued platform were changed between trials. Eight daily trials switched between two types of trials: from a trial with the cued platform to a trial with the hidden platforms. Thus, this stage of testing required acquisition of spatial memory for a new hidden platform while, if the cue is present, responding by a goal-directed behavior toward the cued platform.

OPEN IN VIEWER

FIG. 6.

Strategy competition in the Radial Water Maze (RWM) revealed cognitive impairments in traumatic brain injury (TBI) and Craniotomy rats. (A) Comparison of learning in two versions of the RWM. The spatial version used trials with a hidden platform (Fig. 5.). The mixed version of the RWM used two types of trials: with cued platform or with hidden platform. Number of Errors to find a hidden platform is shown for both versions of the RWM task. Pound signs for Naïve (p < 0.036) and Anesthesia group (p < 0.002) indicate significant differences between the task versions in the trials with hidden platform (see Supplementary Table S5 for details of statistical analyses). (B) Number of errors to find cued or hidden platform averaged over 2 days of training. Insert shows the order of two types of trials for each training day. Testing started from a trial with cued platform followed by a trial with hidden platform. The location of hidden platform was stable across testing day, while location of cued platform was changed every trial. Asterisks (p < 0.05) and triangles (p < 0.005) indicate significant differences from Naïve controls or Craniotomy group, respectively (least significant difference [LSD] post hoc tests). (C) Latency to the cued or hidden platform averaged over 2 days of training. Asterisks (p < 0.05) and triangles (p < 0.05) indicate significant differences from Naïve controls or Craniotomy group, respectively (LSD post hoc tests). (D) Decision speed calculated as the rate of arm choices (number of errors + one correct choice per sec). Asterisks (p < 0.05) and triangles (p < 0.05) indicate significant differences from Naïve controls or Craniotomy group, respectively (LSD post hoc tests). (E) Number of errors in consecutive trials according to different strategies: to find the hidden platform, to find the previous cued platform, or to find a platform from the previous trial. Trials are shown starting from Trial 2 as there was no previous cued platform for Trial 1. H—a trial with hidden platform, C—a trial with cued platform. (F) Average number of errors according to different strategies. The errors were averaged across trials 2-4 for each type of trials. The orientation trials (Trial 1) for both cued and hidden platform were excluded for consistency. Significant differences between strategies (pound signs, p < 0.02, LSD post hoc tests) were observed in all treatment groups. (G) Number of errors to find the previous trial's platform in consecutive trials. Insert shows number of errors averages across all trials. Asterisk (p < 0.002) and triangle (p < 0.01) indicate significant differences from Naïve controls or Craniotomy group, respectively (LSD post hoc tests). Data presented in panels (E-G) are from Day 1 of the training. Horizontal solid lines in panels (A, B) indicate the chance level of performance. Number of cases in each of the treatment groups is shown in the legend. See Supplementary Table S5 for further details of statistical analyses. Color image is available online.

The new challenging protocol showed that introducing the cued trials impaired learning the hidden platform. In particular, performances of Naïve and Anesthesia groups in the spatial trials were significantly worse when compared with their performances in the previous standard protocol (Fig. 6A). These impairments were most obvious in the 2nd trial (Fig. 6A). The standard Radial Water Maze protocol demonstrated that the 2nd trial was critical to dissociate differences in the speed of learning between treatment groups. With the new protocol, the between-group differences in the second trial were readily apparent as well. However, in this more challenged condition, the differences between treatment groups were represented by a dissociation between best performances of Craniotomy group and the worst performances of the Anesthesia group (Fig. 6B, 6C, Hidden Trial 2). Interestingly, Craniotomy group had dramatically fewer number of errors in spatial trials, but significantly more errors in the cued trials than Anesthesia group (Fig. 6B, Cued Trial). These data implied that better performance of Craniotomy rats in spatial trials came at the expense of performance in the cued trials. In the Anesthesia group, the cued trial performance was similar to the Naïve controls, however their attempts to adequately respond to cued platform might have deteriorated their ability to learn the location of new hidden platform.

Analysis of decision rates (number of arm choices per second) yielded the clear dissociation between high rates in TBI and Craniotomy groups and low rates in Naïve and Anesthesia controls (Fig. 6D). This dissociation was reminiscent of that observed in the orientation trials in the standard Radial Water Maze (Fig. 5D, Trial 1). In the more complex protocol, the between-group differences were observed in the orientation trials for both cued and hidden platforms. In addition, Naïve controls continued to show low decision rate in most of the trials with hidden platform. It is noteworthy that the dissociation of Naïve and Anesthesia groups from Craniotomy and TBI groups in the decision rate was independent of the actual success of those decisions. For example, the low rate of decision in Anesthesia rats was observed despite of their problematic performance in the spatial trials. Craniotomy rats demonstrated the high rate of arm choices in the trials with successful decisions (the spatial trials with hidden platform) and in the trials with less successful choices (the cued trials). These data indicated that the high decision rates observed in Craniotomy and TBI groups might be reflective of their inability inhibiting the action before making a decision.

The analysis of the order of choices within each trial revealed a highly organized behavior (Fig. 6E). Namely, during the cued trials all rats visited the hidden platform before going to the cued platform (Fig. 6F). During the trials with hidden platform, all rats visited the arm with the previous cued platform before going to the correct arm (Fig. 6F). Entering the correct arm and contacting a platform did not necessarily result in climbing on the platform, as animals continued to engage in orientation behavior and visited other arms before their final choice. These observations can be reconciled if in the beginning of every trial the choice was directed to the location of the platform in the previous trial. Indeed, the analyses of such choices revealed that visiting the previous trial's platform location was the 1st or 2nd choice for most of the animals (Fig. 6E, 6G). TBI group had followed this strategy as well, however, made significantly more errors finding the previous platform location than Naïve controls (Fig. 6G). This deficit in following spatial strategy coincided in TBI rats with the impairment in learning goal-directed behavior as they demonstrated long latencies and substantial number of errors in the final cued trials (Fig. 6B, 6C).

Social interactions

Social recognition task (Fig. 7A-D)

All treatment groups demonstrated robust habituation to the new cage environment, and no differences were observed in novelty-induced exploratory activity (Fig. 7B; Supplementary Table S6). Then, social recognition memory was tested using a habituation-dishabituation paradigm. Analyses of the latency to initiate the social contact revealed that TBI rats took significantly longer for their first social approach as compared with Naïve and Anesthesia groups (Fig. 7C). In accordance with this observation, analyses of duration of social investigation in the first 30 sec of each of the tests showed that TBI rats spent less time interacting with novel rats. This was true for both the 1st trial (when the first novel rat was presented) and the 4th trial (when the second novel rat was presented; Fig. 7D). In terms of social recognition at the dishabituation phase, TBI rats were the only treatment group that failed to significantly increase social engagement when the novel rat was presented (Fig. 7D). These observations were supported by significant effects of Treatment, Trial and Treatment × Trial interaction (ANOVA; Supplementary Table S6). In addition, analyses of total time of social investigation in this task revealed that Craniotomy group accumulated longer duration of social interactions than Anesthesia and TBI group. The size-effect of this effect was small (Fig. 7D).

OPEN IN VIEWER

FIG. 7.

Deficits in social interactions were revealed in traumatic brain injury (TBI) rats in reciprocal social recognition (A-D) but not in social motivation (E-G). (A) Scheme of Social Recognition task. An adult rat from one of the treatment groups was initially familiarized with the environment where the social interactions will take place (Orientation trial). Then, adult rats had 3 trials in which it had an opportunity to socially engage with the same young stimulus rat (Habituation stage). At the final 4th trial, the adult rat was introduced to a novel stimulus rat with an expectation of renewed interest in social interactions dur to dis-habituation. (B) Distance travelled during a 5-min-long orientation trial. No significant treatment-related differences were observed. (C) Latency to the 1st social encounter initiated by an adult rat. Asterisks indicate significant differences from Naïve controls (analysis of variance [ANOVA] least significant difference [LSD] post hoc tests, p < 0.016; see Supplementary Table S6). Insert shows group means for the latency averaged across all trials. Asterisk (p < 0.0043) indicate differences from Naïve controls (significant effect of treatment, ANOVA, p < 0.011). (D) Duration of social interactions in consecutive trials. Asterisks indicate significant differences from Naïve controls (ANOVA LSD post hoc tests, p < 0.023). Insert depicts total duration of social investigations accumulated across all trials. Triangles (p < 0.0078) indicate differences from Craniotomy group (significant effect of treatment, ANOVA, p < 0.0172). (E) Scheme of Social Motivation task in a three-chamber apparatus. An orientation trial served for familiarization of an adult rat with the environment and empty enclosures. During the next two trials, an adult rat had an opportunity to investigate one randomly-assigned enclosure with young stimulus rat and another, non-animated, enclosure (Habituation stage). In the 3rd trial, the adult rat was introduced to the same environment, but an additional novel stimulus rat was present in the previously empty enclosure. (F) Investigation time of two enclosures in consecutive trials. Pound signs indicate significant differences between animated and non-animated enclosures in Trials 1-2 (p < 0.021, ANOVA LSD post hoc tests). Dollar signs indicate significant differences between the enclosure baited with familiar stimulus rat and the enclosure containing novel stimulus rat in Trial 3 (p < 0.029, ANOVA LSD post hoc tests). (G) Preferences in enclosure investigation in consecutive trials. No significant treatment-related differences were observed. All treatment groups switched their preferences to new social stimulus in Trial 3 (p < 0.0001, effect of trial, ANOVA). Horizontal solid line indicates the chance level of preference. Number of cases in each of the treatment groups is shown in the legend. See Supplementary Table S6 for details of statistical analyses. Color image is available online.

Data integration from different tasks

Multiple tasks are necessary to systematically characterize the effects of brain injury on behavioral and cognitive domains. However, as outcomes from different tasks are represented on different scales, such studies bring an additional complexity in reconciling the multitude of behavioral measures and delineation of relative sensitivity of different tasks/measures. To address this issue, we calculated effects sizes for a set of variables drawn from all tasks that yielded significant main effect of group (ANOVA; Supplementary Tables S2-S6). The effect sizes were calculated using Hedges q statistic as described (Fig. 8). ²² Comparisons between Naïve and TBI groups yielded variables with medium to large effect sizes in all tasks (Fig. 8A, 8C). Maximum effect sizes were observed for the time spent around the platform during the last reversal in the Morris water maze (q = 1.50) and for the number of errors in the Radial Water Maze (q = 1.57). The same two variables were among the most sensitive to differentiate Naïve versus Craniotomy (q = 1.16) and Craniotomy versus TBI (q = 1.54) groups (Fig. 8C). In addition, dissociations of Naïve versus Craniotomy groups were particularly strong for the variables of foot faults (q = 1.17), motor activation in Novel Object Recognition task (q = 1.20), and errors to visible platform in the Radial Water Maze (q = 1.07).

OPEN IN VIEWER

FIG. 8.

Effect sizes of traumatic brain injury (TBI) depend on the choice of control group (Naive or Craniotomy). (A) Hedges q statistic for the effects of TBI as compared with Naïve controls (blue line) and the effects of Craniotomy as compared with Naïve controls (orange line). Dark gray, light gray, and yellow areas indicate small, medium, and big effect sizes, respectively. (B) Hedges q statistic for the effects of TBI as compared with Craniotomy group. (C) Description of variables shown in (A, B) and respective effect sizes for each comparison. Values highlighted in green satisfied the Benjamini-Hochberg p value for false discovery rate = 0.05. Color image is available online.

When controlling the false discovery rate (FDR = 0.05), we noticed that Naïve versus TBI comparisons, which satisfied the Benjamini-Hochberg p value, had effect sizes q > 0.80 (Fig. 8A, 8C). For Naïve versus Craniotomy and Craniotomy versus TBI comparisons, variables that satisfied the FDR had bigger effect sizes, q > 0.90 and q > 1.50, respectively (Fig. 8C).

Foot fault test but not gait analysis is sensitive to the effects of TBI and Craniotomy

TBI was modeled using one of the most widespread protocols, controlled cortical impact (CCI) with the impact directed at the sensorimotor cortex, so it was important to assess the extent of functional motor deficits. Analyses of locomotor functions at the 3rd week after trauma revealed significant deficits in both Craniotomy and TBI groups (Fig. 2; Fig. 8C). The impairments were easily detectable in the foot fault test but not in gait analyses. In contrast to the latter, the foot fault test requires integration of visual and proprioceptive information to construct prediction and then timely execute appropriate foot placement. Although we did not analyze the extent of functional or histological tissue damage in this study, the behavioral data suggest that the requirement of complex sensorimotor integration might make the foot fault test more sensitive to deteriorating effects of TBI and craniotomy than sensorimotor reflexes supporting normal gait. In all rats, the number of missteps were higher for hind paws than for front paws, implicating that correct prediction of placement is more difficult for hind paws. Not surprisingly, the effects of TBI and craniotomy were most significant for the number of hind paws missteps. It is important to note that despite significant differences observed in the Foot Fault test, the lack of gross motor abnormalities in gait analysis, motor activity and swim speed indicated that the deficits observed in cognitive tasks are not likely to be explained by sensorimotor impairments.

The presence of foot fault deficits without gait alterations is consistent with the limited data in developmental TBI models. To our knowledge, only one other study employed foot fault testing in a pediatric TBI model. Ajao and colleagues showed increased foot faults at post-injury Days 7 and 60 after CCI in PND 17 rats. ²³ The pediatric TBI studies using gait analysis have been performed primarily in piglets, demonstrating sustained deficits out to 12 weeks after CCI in 4-week-old piglets, ^24,25 and correlating with injury severity. ²⁶ One group evaluated a panel of sensorimotor tests for one month after CCI in PND 17 rats, and concluded that gait analysis was not suitable for measuring long-term recovery following CCI in immature rats, due to lack of motivation in performing the task during development, and interference in the signal by male rat testes after puberty. ⁶ This study also raised concerns with the use of grid walking for foot fault analysis, due to the lack of scalability with growth of the animals.

NOR task is highly sensitive to the effects of TBI and Craniotomy

Testing of object recognition memory revealed that TBI and craniotomy resulted in significant deficits observed as a lack of exploratory preferences for a novel object (Fig. 3; Fig. 8C). The impairments in novel object recognition coincided with significant motor activation in these groups, implying that detection of novelty was preserved; however, discrimination of the source of novelty was impaired. The object recognition task is based on a natural motivation of rodents to explore novelty and has been shown to require normal functioning of the perirhinal cortex and hippocampus. ^27,28 Significant deficits in the TBI group indicate that an early-life damage of the sensorimotor cortex may result in long-term secondary changes in the integrity of other cortical areas such as perirhinal cortex and hippocampus. These effects are consistent with a secondary (delayed) cell loss in this model, particularly in the hippocampus. ²⁹ The presence of object recognition deficits in the Craniotomy group suggests that deteriorating functional outcomes can be observed even if the injury is limited to dura matter.

Similar deficits in novel object recognition have been demonstrated by our group and others, using the CCI model in immature rats injured at PND 10-21. ^11,30,31 Importantly, post-injury impairments in object recognition were prevented by several multi-mechanistic, neuroprotective treatments, including erythropoietin, ³² aceytyl-L-carnitine, ³¹ progesterone, ³³ docosahexaenoic acid, ³⁰ MK-801, ³⁴ and the 20-HETE inhibitor, HET0016. ¹¹ In addition, a decrease in novel object recognition in pediatric TBI models appears to relate to injury severity in closed head injury models, ³⁵ with repeat TBI events leading to worsening memory impairments. ³⁶

Multi-stage Morris water maze is sensitive to the effects of TBI and Craniotomy

To further analyze TBI-related cognitive deficits, we employed a battery of tasks based on a motivation to escape from water. These tasks were based on similar procedural approaches (such as a hidden escape platform, similar inter-trial intervals, different start location for every trial, etc.) facilitating between-task comparisons. In contrast to the novel object recognition task, in the water maze tasks information about objects (spatial cues and escape platform[s]) had to be integrated into a spatial map of the environment with critical roles of hippocampus-entorhinal cortex and hippocampus—prefrontal cortex networks. ^{37 -39} Importantly, the task complexity increased as training progressed starting from testing spatial reference memory to testing features of episodic-like memory that require use of “where” and “when” categories in an integrated and flexible manner. ⁴⁰ The latter stages have increased sensitivity to hippocampal dysfunctions. ^14,41

In the Morris water maze, the TBI rats demonstrated significant deficits in measures of both, reference and episodic-like memories (Fig. 4; Fig. 8C). In addition, they showed delayed learning and spent longer time to reach the hidden platform(s) than the other groups. Despite training-related improvements in performance, the deficits of the TBI rats in finding the platform persisted up to the end of the training. As start locations varied every trial, these observations suggest that TBI rats could have difficulties re-orienting a spatial map as quickly as controls. The performance of the TBI rats in the probe trials is consistent with this interpretation as they showed significant deficits in spatial preferences as compared with Naïve controls; nevertheless, their preferences were significantly higher than a chance level.

In addition to the TBI, the Craniotomy group had deficits in the Morris water maze as well. However, the nature of craniotomy-induced deficits seems different than in the TBI group. Craniotomy rats were not impaired in latency to reach the platform demonstrating normal spatial learning and the ability to re-orient the spatial map to varied start positions. Nevertheless, the Craniotomy rats demonstrated deficits in the probe trials similar to the TBI rats. Thus, in the Craniotomy group, the deficits in the probe trials coexisted with the normal performance in the training trials and could reflect a poor acquisition of a waiting response for the platform rather than an impairment in spatial memory.

In addition, in the reference memory task, the Craniotomy rats performed poorly after an overnight delay, indicating an impairment in memory retrieval. This deficit was temporal and disappeared after additional training as the Craniotomy group demonstrated normal retrieval visiting previous platform locations during Repeated Reversals and Radial Water maze testing. The deficits were not debilitating, as performance of the TBI rats was above the chance level. In addition, the low accuracy of spatial exploration in the probe trials observed after craniotomy was not consistent with primary deficits in spatial memory. As poor performance in the probe trials can reflect not only impaired spatial memory but a response strategy that is not compliant with persistent presence in the vicinity of the platform, utilization of an additional spatial task that does not require acquisition of such response (like RWM) might be helpful in delineating deficits specific to spatial memory.

The Morris water maze deficits we found are comparable to those seen in other pediatric CCI studies in rodents. Adelson and colleagues found a significant Morris water maze deficit in both PND 7 and PND 17 rats, compared with craniotomy controls, with a graded response based on severity of injury. ⁴² Interestingly, one study evaluated long-term water maze deficits at 3-6 months following PND 17 CCI in rats. They did not find differences in cued or spatial learning between craniotomy controls and CCI rats. However, they did see that injured rats had fewer entries into the target area, and the cumulative distance traveled in spatial learning was negatively correlated with the number of target area entries in the probe trials. ²⁹ Several studies have utilized water maze testing to demonstrate improved behavior with neuroprotective treatments after CCI, such as buspirone, ⁴³ COX-2, minocycline, ⁴⁴ ketogenic diet, ⁴⁵ and flumazenil. ⁴⁶ In contrast, some studies did not show improvements in Morris water maze deficits with well-studied treatments such as progesterone ⁴⁷ and COX-2 inhibition. ⁴⁸ Further, some developmental TBI studies did not detect any water maze deficits in their injured animals, when compared with craniotomy controls. These findings speak to the importance of the use of appropriate age-matched control groups, with inclusion of a naïve control group.

Similar to the adult CCI literature, many studies to date in immature rats have used only sham craniotomy controls. Schober and colleagues did include both craniotomy and naïve controls in water maze testing after CCI in PND 17 rats, with the two control groups performing similarly on hidden platform testing. ⁴⁹ Other important factors that may affect variability of findings in water maze testing are the age at injury and testing ^42,50 as well as specifics of water maze testing paradigms utilized. ^29,51

Radial Water Maze is sensitive and differentiates to the effects of TBI and Craniotomy

Although the Radial Water Maze has been used in adult rodents after TBI ^52,53 and after hypoxia-ischemia, ^54,55 to our knowledge, there are no published studies of Radial Water Maze in immature rats after CCI, indicating that utilization of this task in the future research may increase sensitivity of study design to effects of TBI and efficacy of treatments. In our study, the rats were trained in the Radial Water Maze, in which most procedural aspects of the task were the same as in the preceding repeated reversals in the Morris water maze. For example, the location of hidden platform was changed daily, and the start position was varied between trials. However, the swimming area was structured by a radial maze enclosure, which allowed for the outcome measures that did not require a waiting response to the platform as in the previous tasks. In this setting, the TBI rats demonstrated delayed learning but, at the end of the 1st training session, they were as efficient finding the platform as control rats. Importantly, when memory for the platform was tested after an overnight delay, the performance of the TBI rats was similar to controls as well. These data documented that in the Radial Water Maze, the TBI rats were able to learn and remember the 1st platform location. Their performance, however, deteriorated when the platform location started to be switched daily.

Similar to the repeated reversal in the Morris water maze, the TBI rats were deficient in finding a new platform location up to the end of the daily sessions. After an overnight delay, the TBI rats were not able to locate the previous day's platform as efficiently as Naïve controls; however, when given more time/trials, the TBI rats demonstrated their spatial preferences to the previous platform. Importantly, the inability of the TBI rats to inhibit such preferences after the orientation trial (Trial 1) might be a reason for their underperformance. Similar deficits inhibiting previous platform visits were detected in the Craniotomy rats.

The Radial Water Maze task showed that an early-life injury to the somatosensory cortex resulted in impairments of episodic-like memory, observed as difficulties with an immediate recall and deficits inhibiting responses based on previous memory traces (Fig. 5; Fig. 8C). The latter deficit also was observed after pediatric craniotomy. It is difficult to conclude whether preserved reference memory of the TBI rats in the Radial Water Maze is due to task-related differences or additional training received during previous tasks. Irrespective of these possibilities, demonstration of specific impairments in episodic-like but not reference memory in the TBI rats suggests that the former is more sensitive to deteriorating effects of TBI.

New RWM protocol with multiple alternative choices uncovers disinhibition in TBI and Craniotomy

Because rats are able to show more sophisticated spatial learning than mice, ^56,57 we explored the additional effects of TBI in more challenging conditions. To this end, trials with cued platform were introduced between hidden platform trials allowing for the analyses of competition between spatial (hippocampus-dependent) and goal-directed (caudate-dependent) strategies. As switching between these strategies requires medial prefrontal cortex (mPFC) activity, ^58,59 this task was particularly suited to challenge mPFC interactions with hippocampus and nucleus Caudatus. The analysis of strategies employed in this task revealed that rats from all treatment groups memorized the platform location from the previous trial.

For example, in a hidden platform trial, the animals visited first the location of cued platform from the previous trial despite the fact that the cue was no longer present. This behavior indicated that the rats continued to use previously acquired spatial strategy independently of the type of trials (with hidden or cued platform). This behavior was consistent with a one-trial learning demonstrated by Naïve control, Anesthesia and, partially, Craniotomy rats at the preceding stage of the Radial Water Maze. In the new task, the rats revealed an amazing fit of their cognitive abilities, in that they were able to learn one-trial associations seven times in row. As expected from the previous tasks, the TBI rats were significantly impaired in their abilities to form one-trial spatial associations (Fig. 6; Fig. 8C). It is important to note that such spatial strategy was no longer adaptive for the new task; however, it was not easily abandoned. In the uncertain situation of multiple platform locations, the performances of Craniotomy and Anesthesia groups were clearly dissociated. Both groups had difficulties balancing spatial and goal-directed strategies. In the Anesthesia rats, the attempts to follow the cue coincided with delayed learning of the hidden platform location, whereas in the Craniotomy rats the spatial strategy remained dominant resulting in the best performance in the hidden platform trials at the expense of success in the cued platform trials. Low efficacy in the cued platform trials was one of the most sensitive measures dissociating Craniotomy rats from Naïve controls. The TBI rats were also impaired adopting goal-directed strategy leading to significant deficits in the cued trial learning.

Testing in a situation of multiple alternative choices revealed that TBI as well as Craniotomy rats had significantly higher rate of arm choices than Naïve and Anesthesia groups. Similar observations were made for the orientation trials at the previous stage of the Radial Water Maze. The higher rates of choices in the TBI and craniotomy groups suggest possible impairment inhibiting immediate motor responses. Comparison of deficits observed across multiple tasks revealed that TBI- and craniotomy-induced changes in rate of choices had big effect sizes supporting (Fig. 8C) biological significance of this behavioral phenotype.

Social recognition but not motivation is affected by TBI

Finally, we tested the long-term effects of early-age TBI on social interactions using two tasks (Fig. 7; Fig. 8C)—social recognition ^19,60 and three-chamber social motivation ²¹ —that assess overlapping as well as differing aspects of social behaviors. Both of these tasks require preserved levels of social motivation to ensure appropriate performance. The protocols utilized for both of these tasks included stages of habituation in which animals were presented the opportunity to interact repeatedly with the same social stimulus. The later stages, however, differed between the tasks. In the social recognition task, a dishabituation stage required the animal to recognize a novel social stimulus using memory representations of the previous social stimulus. In contrast, the social motivation task included a stage of discrimination, in which a novel and already familial social stimuli were presented simultaneously reducing requirement for stable social memory traces.

Another important between-task difference was that only the social recognition task was based on reciprocal interactions of free roaming animals. The social recognition task proved to be more sensitive to the effects of TBI. In particular, TBI rats took longer to initiate the 1st social encounter and spent less time investigating novel social stimulus. As TBI rats did not show any changes in the level of social investigation in the social motivation task, the TBI-induced social deficit seemed to be limited to a setting with higher-risk reciprocal interactions. An additional TBI-related deficit was revealed at the stage of dishabituation when TBI rats failed to reactivate social behaviors with a novel conspecific. This finding could be interpreted as an impairment in social recognition memory, which would be consistent with TBI-induced impairments in memory documented in the previous tasks. However, considering the deficit in social interactions in the 1st trial, the alternative explanation could suggest the primary TBI effect on social motivation.

Recent developmental TBI studies have demonstrated deficits of social recognition and social motivation after injury. Using CCI in PND 21 mice, Semple and colleagues found that injury-induced social deficits emerged during development. When tested in adulthood, injured mice had a lack of preference for sociability in the three-chamber task, and a lack of preference for social novelty that were not seen when tested in adolescence. ⁵ In follow-up studies, this group also demonstrated important age-at-injury and sex-dependent effect of these post-injury sociability changes, with younger age and male sex having greater deficits. ^61,62 Another study showed impairments in both social novelty and sociability in adolescence, following CCI in immature (PND 14) rats. ⁶³ Similar to our results, but using a closed-head injury model of moderate severity in PND 11 rats, Runyan and colleagues showed social novelty recognition impairment, but did not find sociability deficits using 3-chamber testing at 4 and 8 weeks after CCI in both male and female rats. ⁶⁰ Further, treatment with intranasal oxytocin before behavioral testing reversed the injury-related social recognition deficits. These studies, along with our results, would suggest that social interaction testing is an important component in defining post-injury behavioral profiles following TBI in pediatric models. Further, the translational relevance of social interaction testing after TBI, and its sensitivity to treatment therapies, support inclusion of this testing in future pediatric TBI studies.

Limitations of the study

There are a number of limitations that were not addressed in the current study. All experiments were carried out using only male rats, leaving possible sex-related differences untested. We were also unable to pretest the rats prior to injury, as they are developmentally unable to perform most of the tasks prior to CCI on postnatal Day 10. In addition, behavioral testing was performed sequentially rather than in parallel, which could limit direct comparison between tested outcomes, as they were tested at different times during maturation and in adulthood. Possible effects of enrichment must also be considered in this longitudinal study. The same animals were tested in a lot of tasks and handled from a very young age. The combination of these procedures can be viewed as cognitive and behavioral enrichment. Enrichment is known to improve inhibitory control, behavioral performance and decrease the likelihood of detecting cognitive deficits. ^64,65 From this point of view, the enrichment could underestimate the effects of TBI and other treatments in this study. However, it has also been shown that cognitive benefits of enrichment can be higher in normal controls than in experimentally affected groups. ⁶⁶ The latter consideration would have opposite expectations for the effect of enrichment, which can be specifically tested using combination of longitudinal and cross-sectional design. These questions will require further investigation.

Summary

Using detailed systematic analyses of learning and memory in multiple tasks/protocols, we demonstrated long-term cognitive and behavioral effects of pediatric TBI. Despite remarkable neuronal loss known to occur in the juvenile CCI model long-term, ²⁹ the deficits in most cognitive tasks were not debilitating, a testament to the high adaptive capacities of the developing brain. The most sensitive cognitive processes affected by TBI included those that require fast one-trial learning and flexibility of acquired memory traces or response strategies. These processes are known to require multiple cortical and sub-cortical regions (such as prefrontal cortex, perirhinal cortex, dorsal striatum) ^14,41,58,59 that are not affected by mechanical force of the CCI, however, are very likely to be involved in the long-term sequelae of pediatric TBI. The TBI effects observed in this study were compared not only with Craniotomy group, but also with Naïve and anesthesia controls. This experimental design allowed delineating additional deficits due to early-life craniotomy and anesthesia.

Very limited deficits observed in the Anesthesia group is consistent with a recent report on relatively safe profile of isoflurane used in this study ⁶⁷ ; however, the effects of anesthesia might be more pronounced with prolonged exposure and different anesthetics. Considering that craniotomy did not involve any experimental damage to neural tissue, the long-term cognitive impairments were impressive as some of the effect sizes were comparable to those in the TBI rats. These remarkable effects of pediatric craniotomy suggest that in studies that use the craniotomy group as the only control group, multiple effects of TBI might be missed. Craniotomy-induced cognitive impairments were more readily detectable in more complex tasks indicating that these effects can also be missed if study design incudes only simple tasks with low cognitive demands. Our study demonstrates that inclusion of multiple control groups allows for better characterization of long-term effects of TBI on cognition. Noteworthy, the lack of Naïve controls in study design, a practice that is unfortunately widely accepted in the field, prevents addressing mechanisms of craniotomy-related deficits, which may be more responsive to treatments.

In conclusion, we identified short- and long-term deficits between post-natal Days (PND) 25 and PND 140 and found that TBI rats, when compared with naïve controls, had impairments across most tests studied. When TBI group was compared with Craniotomy as a control, the deficits were detected in a limited number of cognitive outcomes related to learning speed and cognitive flexibility. Notably, effects of craniotomy, when compared with Naïve controls, spanned across multiple tasks, and sometimes reached the effect sizes comparable to that of TBI. Peri-operative anesthesia exposure alone did not produce deficits. These results highlight the importance of selection of appropriate control groups in pediatric CCI models. In addition, the results demonstrate that performing a wide array of comprehensive cognitive testing improves the ability to detect deficits following developmental TBI.