Ketamine Alters Hippocampal Cell Proliferation and Improves Learning in Mice after Traumatic Brain Injury

Traumatic brain injury is a potentially devastating condition associated with significant long-term morbidity, including memory deficits, depression, and seizure disorders. Though traumatic brain injury involves neuronal loss and dysfunction, the postinjury response includes increased glial proliferation and activation, as well as an increase in hippocampal neurogenesis.¹  Although adult-born hippocampal neurons have beneficial roles in healthy brains,²  the contribution of posttraumatic neurogenesis to cognitive recovery after injury is less clear: it may be beneficial if new neurons compensate for circuits disrupted by injury, or it may be harmful, if neurons generated after traumatic brain injury have maladaptive properties, and contribute to the formation of aberrant circuits.³ 

Hippocampal radial glial-like stem cells reside in the subgranular zone of the dentate gyrus, and can differentiate into three types of cells: new radial glia-like cells, astrocytes, or neurons.⁴  The mitotic rate, fate specification, and long-term survival of cells generated from radial glia-like cells and their progeny are modulated by numerous endogenous and exogenous contingencies, including activity at ionotropic neurotransmitter receptors,⁵  which might provide a link between activity-related signaling and hippocampal structure. Of particular importance is the N-methyl-d-aspartate (NMDA)-type inotropic glutamate receptor (NMDAR), which is involved in synaptic transmission and plasticity. Activation of NMDARs stimulates radial glia-like cell mitosis⁶  and plays a vital role in the survival of adult-born neurons⁷  as well as in the formation of synapses.⁸ 

NMDARs are also known to potently modulate glial function and proliferation, either directly or indirectly via alterations in neuronal function and signaling.⁹  Glial function likely has important roles in recovery from traumatic brain injury, as two types of glia—microglia and astrocytes—are activated after injury¹⁰  and are known to modulate neuronal structure and function.¹¹  It is still not clear, however, whether various glial subtypes contribute positively or negatively to recovery after injury, as their activity likely has both beneficial and harmful consequences in terms of circuit function.¹¹ 

NMDARs are targets of several anesthetic drugs, including ketamine, a noncompetitive NMDAR antagonist, which is frequently employed by medical providers in both intensive care units and operating rooms. Although it has historically been contraindicated in the setting of head injury due to its perceived effects on intracranial pressure, it is increasingly recognized for its neuroprotective potential.¹²  Based on positive outcomes from its use on battlefields and after trauma, it is no longer contraindicated in the setting of head injuries.¹³  Thus, it is likely that increasing numbers of people will be exposed to ketamine early after head injury. In this study, we investigate the effects of ketamine administration after traumatic brain injury on posttraumatic hippocampal cell proliferation in mice and their subsequent performance in hippocampus-dependent learning tasks.

All mouse housing, handling, and procedures were performed in accordance with National Institutes of Health guidelines and were in compliance with Oregon Health and Science University Institutional Animal Care and Use Committee (Portland, Oregon) approved protocols. We utilized male and female adult (8 to 12 weeks old) C57BL/6J wild-type mice. Mice were euthanized at different time points after traumatic brain injury for histologic analysis, and included 11 male and 12 female mice for the 3-day time point, 28 male and 26 female mice at the 2-week time point, and 29 male and 18 female mice at the 6-week time point. All mice had access to food and water ad libitum and access to 24-h/7-day veterinary consultation if needed. To minimize the number of mice needed for this study, the 6-week histology data were obtained on the same mice that underwent behavioral testing at 4 to 5 weeks.

Traumatic brain injury was modeled using a controlled cortical impact injury in mice as previously described.¹⁴  Briefly, mice were anesthetized with isoflurane (1.8%) and placed on a stereotaxic frame. After 10% betadine sterilization and 2% lidocaine gel application, a scalp incision was made at midline, followed by a sterile 4-mm craniotomy to the right of the midline sagittal suture, between lambda and bregma, leaving dura intact. A 0.9-mm cortical deformation was applied using an electronic impactor (Leica Microsystems, Germany; speed 4.4 m/s; dwell time 800 ms) to the exposed area with a sterile 3.0-mm diameter stainless steel impact tip. After controlled cortical impact, the scalp was sutured and bioadhesive glue was used to secure the incision site. Sham (noninjured) mice underwent similar anesthesia, frame mounting, and the scalp incision/closure, but no craniotomy. Immediately after sham or controlled cortical impact, mice were implanted with an Alzet osmotic pump (Durect, USA) delivering 1 µl/h of one of the following: vehicle (sterile 0.9% normal saline) or ketamine (± ketamine hydrochloride; Sigma-Aldrich, USA; 25 mg/ml in saline, for a dose of 30 mg · kg^–1 · day^–1)¹⁵  into the subcutaneous space on the back (dorsum) of the mouse. After pump placement, mice were marked using ear punches, and allowed to emerge from anesthesia in a heated, padded chamber, with access to acetaminophen-soaked food for 48 h. Mice typically resumed normal exploratory behaviors within 15 min after emergence. All subsequent experiments were performed by experimenters blinded to group assignment.

All mice survived sham or controlled cortical impact injury, apart from one mouse in the 3-day group that had been assigned to receive controlled cortical impact injury/vehicle, which died within 1 h of controlled cortical impact and was excluded from analysis. In the 2-week and 6-week groups, osmotic pumps remained in place for 7 days, after which they were sterilely removed under anesthesia. In the 3-day group, pumps remained in place throughout the entire study period. Residual pump volumes were measured from all mice to confirm drug delivery, and in all cases, the volumes remaining were ±10% of that expected based on the calculated pump delivery rates.

Bromodeoxyuridine was used to label cells undergoing mitosis after sham or controlled cortical impact injury. For mice undergoing delayed analysis of cell survival and fate specification at 2 and 6 weeks after controlled cortical impact, bromodeoxyuridine (20 mg/ml in saline; Sigma-Aldrich) was injected intraperitoneally at a dose of 300 mg/kg twice a day, 4 h apart, on post–controlled cortical impact days 2 and 3. Mice examined at 3 days after injury received 50 mg/kg bromodeoxyuridine three times a day, 2 h apart, on post–controlled cortical impact day 2 to assess proliferation rates.

Mice were deeply anesthetized with isoflurane before receiving a lethal anesthetic dose of 2,2,2-tribromoethanol (Sigma-Aldrich) in accordance with Institutional Animal Care and Use Committee approved protocols. They were then transcardially perfused with chilled phosphate-buffered saline, followed by 4% paraformaldehyde (in phosphate-buffered saline) solution, and postfixed overnight. A small, rostral-to-caudal nick was made to the ventral left hemisphere (contralateral to the controlled cortical impact) to denote laterality. Free-floating coronal brain sections were prepared using a vibratome. For immunohistochemistry, four hippocampal sections were selected: two dorsal sections approximately 200 μm apart, and two ventral sections approximately 300 μm apart, and the rostrocaudal regions chosen were kept consistent between animals. In the rare event that damage to the dentate was apparent, samples were excluded and neighboring slices without dentate damage were chosen. Slices were permeabilized in 0.4% Triton X-100 (Sigma Aldrich) in 0.5 M potassium-phosphate-buffered saline with Triton X-100 for 45 min, placed in 2 N HCl potassium-phosphate-buffered saline with Triton X-100 at 37°C for 30 min, followed by a 10-min wash in pH 8.5 potassium-phosphate-buffered saline with Triton X-100, and two 10-min washes in 0.4% Triton X-100 potassium-phosphate-buffered saline (pH 7.4). Samples were blocked in 10% horse serum in 1% phosphate-buffered saline with Triton X-100, and then incubated overnight at 4°C with primary antibodies and 1.5% horse serum in 0.4% potassium-phosphate-buffered saline with Triton X-100. Primary antibodies and concentrations used for this study were: goat anti-doublecortin (1:500; Santa Cruz Biotechnology, USA), mouse anti-neuronal nuclear protein (1:500; EMD Millipore, USA), rat anti-bromodeoxyuridine (1:500; Abcam, USA), mouse anti–glial fibrillary acidic protein (cy3 conjugated; 1:500; Sigma-Aldrich), rabbit anti–ionized calcium-binding adaptor molecule 1 (1:500; Wako Chemicals, USA), rabbit anti–caspase-3 (1:500; Cell Signaling Technology, USA), and mouse anti-cellular Finkel–Biskis–Jinkins murine osteogenic sarcoma protein (cfos) (1:500; EMD Millipore). After primary antibody incubation, samples were washed in 0.4% potassium-phosphate-buffered saline with Triton X-100 for two 10-min washes and then incubated for 4 h at room temperature with secondary antibody. Secondary antibodies used in this study were Alexa-dye conjugated secondary antibodies at 1:500 concentrations (Invitrogen, USA). Samples were then washed in potassium-phosphate-buffered saline with Triton X-100 for two 10-min washes, incubated in 4´,6-diamidino-2-phenylindole (1:20,000; Sigma-Aldrich) for 30 min, and subsequently mounted onto slides.

Microscope slides were coded before microscopy so that blinding was maintained throughout imaging and analysis, until final results were tabulated. Four antibody-labeled hippocampal coronal sections from the hemisphere ipsilateral to sham/controlled cortical impact for each mouse were imaged using a Zeiss LSM780 confocal microscope with a 5×/0.16 numerical aperture, 20×/0.8 numerical aperture, or a 63×/1.4 numerical aperture lens (Zeiss, Germany). Contralateral sections were not analyzed, as previous studies demonstrated intermediate levels of proliferation in the contralateral hemisphere with high variability.^14,16  A z-stack of ~1.5-µm-thick intervals was obtained for 20× images and 0.5- to 1.0-µm thick intervals for 63× images traversing the entire depth of the sample. Sections from each cohort were stained identically in parallel, and laser strength and detector gain settings were kept constant across all samples. Cell counting was performed using Imaris Image Analysis software (Bitplane, USA) as follows: three-dimensional images obtained from confocal microscopy were evaluated through the entire depth (z-stack) of the image. 4’,6-diamidino-2-phenylindole staining provided clear delineation of the granule cell layer and subgranular zone, and Imaris software provided a calculation of this volume as the region of interest through its “surfaces” function. Bromodeoxyuridine+ cells within this defined region were identified by automated detection of antibody-stained objects with a diameter of 7 ± 2 µm. For all samples, cell counts were obtained from the dorsal blade of the hippocampal dentate gyrus granule cell layer including the subgranular zone, through the entire depth of the three-dimensional confocal images. Colocalization was determined blindly on coded slides and images; doublecortin+/bromodeoxyuridine+, neuronal nuclear protein+/bromodeoxyuridine+, and ionized calcium-binding adaptor molecule 1+/bromodeoxyuridine+ colocalization was determined by presence of bromodeoxyuridine throughout the nuclear envelope of an antibody-outlined cell; glial fibrillary acidic protein+/bromodeoxyuridine+ colocalization was determined if glial fibrillary acidic protein surrounded the bromodeoxyuridine-labeled nucleus in three planes. For each image, cells that were selected for counting, and which demonstrated colocalization, were identified by the automated protocols (Imaris), and confirmed or rejected manually by an investigator blinded to group assignment before data tabulation. Cell densities were calculated by dividing cell count by granule cell layer volume in cubic micrometers, and the four samples per subject were averaged.

Injury cavities were grossly visible in whole brains before slicing, and demonstrated no obvious asymmetries (cavities were round, based on gross visual inspection). Thus, we quantified injury magnitude using the cross-sectional area of the cavity in a single section through its midpoint using images from full coronal brain slices from both controlled cortical impact groups (vehicle vs. ketamine), at the same rostrocaudal level. The area of the dorsal quadrant ipsilateral to controlled cortical impact was measured using ImageJ (National Institutes of Health, USA), and subtracted from the area of the contralateral quadrant to provide a measurement of the area lost to injury.

Four weeks after injury, mice underwent behavioral assessment by a blinded observer. Mice were singly housed each day for 1 h before behavior in a holding room containing a white noise generator. The water maze consisted of a circular pool (122 cm wide) filled with water (19 to 21°C) and located in the center of a room containing distal visual cues around the pool. A stationed platform was hidden from the mice by submerging it 1 cm below the water and by adding chalk to the water to make the water opaque. The drop location of the mice in the pool was pseudorandomized between four imaginary quadrants. Mice were first trained to locate the hidden platform using two sessions per day for 2 days (2-h intersession intervals). Each session included three trials (10-min intertrial intervals). Mice were given a maximum of 60 s to locate the platform and were guided to the platform if they did not locate it. Once on the platform, mice were allowed to remain on it for 3 s. A probe test, in which the hidden platform was removed in order to assess the preference for the platform location, was conducted 24 h after the last training session (day 3). To avoid potential effects of the probe trial on the memory for the learned location of the first platform, a reinforcement session was conducted immediately after the probe trial in which the hidden platform was again placed in the original location.

The following day, the location of the platform was moved to a different quadrant and mice were trained to locate the hidden platform (reversal water maze), again using two sessions of three trials each with similar intervals as the first hidden platform training. A probe trial was then conducted on day 5. Three days later, all mice received four trials in which the platform was made visible using a clearly marked beacon. The location of the visible platform was moved to a different quadrant each trial. The time spent in the target quadrant, latency (time) to reach the platform, and swim speed were analyzed for potential group differences. Video recording and tracking software (Ethovision; Noldus, The Netherlands) analyzed samples at a rate of five samples per second.

Mice from each litter were randomly allocated to treatment groups in balanced distribution. Data are presented as means and SDs in text and figures. Statistical analysis was conducted using RStudio version 1.0.136 (RStudio, USA) with R version 3.3.2 (R Core Team 2016, available at: https://cran.r-project.org/bin/windows/base/old/3.3.2/; accessed March 28, 2018). All statistical tests were conducted using two-tailed hypothesis testing. Normality of data was assessed using the Shapiro–Wilk and Kolmogorov–Smirnov tests, and parametric tests were used in our analysis when data were normally distributed. Due to the variance observed in our data, unequal variance was assumed in all parametric comparisons. Immunohistochemistry results were analyzed using two-way ANOVA (drug × injury) followed by Tukey honest significant difference post hoc testing; results include weighted average main treatment effect comparisons of drug administration (all vehicle-treated mice vs. all ketamine-treated mice) and injury state (all sham-treated mice vs. all controlled cortical impact–treated mice), and post hoc pairwise comparisons (sham + vehicle, controlled cortical impact + vehicle, sham + ketamine, and controlled cortical impact + ketamine). For injury-size comparison between controlled cortical impact + vehicle and controlled cortical impact + ketamine groups, an independent t test was used. Animal behavioral testing utilized paired t tests, Mann–Whitney U tests, and repeated measures two-way ANOVA (drug × injury), with Newman–Keuls or Dunnett multiple comparison post hoc testing where stated. For all comparisons, P < 0.05 was considered statistically significant. Three mice from the 2-week time point (1 controlled cortical impact + vehicle, 2 controlled cortical impact + ketamine), two mice from the 6-week time point (1 controlled cortical impact + vehicle, 1 controlled cortical impact + ketamine), and one mouse from the 3-day time point (controlled cortical impact + vehicle) were excluded from histologic analysis due to predetermined criteria of extensive injury damage into the dentate itself. Power analysis of preliminary data indicated N = 11 per group would be sufficient to detect significance with a power of 0.80 and α = 0.05, based on pilot data, which demonstrated a 55% reduction in new neuron density, and variability of ±35% in each injury group. Potential sex differences were evaluated and not found to be statistically significant across any experiments, so male and female results were grouped together for analysis. Figures were prepared using Prism Software version 7.0 (GraphPad, USA).

To investigate the effect of ketamine on post–traumatic brain injury hippocampal cell proliferation, we used a controlled cortical impact model of traumatic brain injury in mice (fig. 1, A and B), and administered a continuous subanesthetic dose of ketamine for the first postinjury week via osmotic pump. This dose (30 mg · kg^–1 · day^–1) demonstrated preclinical efficacy in a variety of other assays,^15,17  and was chosen to avoid sedation/immobility/anorexia that might be present at higher doses,¹⁵  which could lead to potentially confounding effects. Ketamine did not cause gross differences in behavior or motor activity levels in either controlled cortical impact or sham-treated animals. Because weight loss or gain is another indicator of injury severity and can confound other results, we measured mouse weights at 1 week, 2 weeks, and 6 weeks postintervention. Mice in all groups gained weight throughout the evaluation period, as would be expected for healthy mice during this developmental window. There were no significant differences in weight gain induced by either controlled cortical impact or ketamine administration (mean weight ± SD at 6 weeks post–controlled cortical impact, in grams: sham + vehicle, 1.8 ± 1.2, n = 12; controlled cortical impact + vehicle, 2.0 ± 1.4, n = 10; sham + ketamine, 1.6 ± 0.8, n = 12; controlled cortical impact + ketamine, 2.5 ± 1.5, n = 10; two-way ANOVA: main effect (injury vs. sham): F_1,42 = 1.68, P = 0.203; main effect (ketamine vs. vehicle): F_1,42 = 0.10, P = 0.756).

Hippocampal cell proliferation was assessed using bromodeoxyuridine-mediated labeling of mitotic cells, as bromodeoxyuridine is incorporated into the DNA of dividing cells and can be detected long after cell maturation. Bromodeoxyuridine was administered to mice 2 to 3 days after sham or controlled cortical impact injury, to coincide with the peak window of postinjury cell proliferation.¹⁸  Mice were euthanized at various time points after controlled cortical impact or sham injury for histologic analysis of the hippocampal dentate gyrus using antibromodeoxyuridine immunohistochemistry.

First, to evaluate cellular proliferation early after controlled cortical impact, a nonsaturating dose of bromodeoxyuridine (50 mg/kg × 3 injections, 2 h apart) was administered to mice 2 days after injury, and they were euthanized 1 day later for analysis. Consistent with previous observations,^14,18,19  controlled cortical impact treatment increased hippocampal cell proliferation when compared with sham-treated animals; however, the subhippocampal distribution of these cells differed between drug treatment groups. In the dentate molecular layer, controlled cortical impact significantly increased proliferation in both injury groups compared to their respective shams (fig. 1C; cells/mm³ means ± SD: sham + vehicle [400 ± 400, n = 6] vs. controlled cortical impact + vehicle [11,800 ± 5,600, n = 4], two-way ANOVA, main effect [controlled cortical impact vs. sham]: F_1,18 = 33.20, P < 0.001, Tukey honest significant difference post hoc test, P = 0.002; sham + ketamine [600 ± 600, n = 6] vs. controlled cortical impact + ketamine [9,500 ± 6,200, n = 6], Tukey honest significant difference post hoc test, P = 0.006). However, when we focused exclusively on the dentate granule cell layer including the subgranular zone, controlled cortical impact only induced an early proliferative response in this subregion in the presence of ketamine (fig. 1 C; two-way ANOVA, main effect [controlled cortical impact vs. sham]: F_1,18 = 15.74, P = 0.001; main effect [ketamine vs. vehicle]: F_1,18 = 6.02, P = 0.025; interaction effect [controlled cortical impact:ketamine]: F_1,18 = 2.21, P = 0.154; controlled cortical impact + ketamine vs. sham + ketamine, Tukey honest significant difference post hoc test, P = 0.005). In contrast, vehicle-treated mice demonstrated a smaller, and not statistically significant, increase in granule cell layer cell proliferation after controlled cortical impact (fig. 1, C and D; controlled cortical impact + vehicle vs. sham + vehicle, Tukey honest significant difference post hoc test, P = 0.438). As the original goal of this study was to examine postinjury neurogenesis, from this point forwards we focus entirely on cell densities within the dentate granule cell layer, as this area contains the radial glial stem cells with the capacity to differentiate into neurons.

Two weeks after controlled cortical impact or sham surgery, a separate cohort of bromodeoxyuridine-labeled mice was examined to determine the effects of ketamine on continued proliferation and early survival of the cells born after injury. A saturating dose of bromodeoxyuridine²⁰  was administered on the second and third postinjury day to label proliferating cells and their progeny (fig. 1A). At this time point, bromodeoxyuridine+ cell labeling was significantly increased in the granule cell layer in vehicle-treated, controlled cortical impact–exposed mice when compared with their vehicle-treated sham littermates (two-way ANOVA, main effect [controlled cortical impact vs. sham]: F_1,48 = 17.35, P < 0.001; controlled cortical impact + vehicle vs. sham + vehicle, Tukey honest significant difference post hoc test, P = 0.001). In contrast, however, although ketamine-treated mice demonstrated increased bromodeoxyuridine+ labeling in injured mice, this was not statistically significant (two-way ANOVA, main effect [ketamine vs. vehicle]: F_1,48 = 3.88, P = 0.055). Although the differences between vehicle-treated mice at 3 days post–controlled cortical impact and 2 weeks post–controlled cortical impact could have resulted from differences in the exact timeframe of bromodeoxyuridine administration between the two time points examined (second postinjury day vs. second/third postinjury days), the increase in cell labeling in the injured mice could also have resulted from ongoing proliferation of bromodeoxyuridine-labeled cells (whose daughters would also be labeled) or differences in cell survival between groups.

Thus, to examine longer-term survival of cells proliferating early after controlled cortical impact, mice were allowed to recover for 6 weeks before histologic analysis. At 6 weeks after controlled cortical impact, across both groups, controlled cortical impact increased the number of bromodeoxyuridine+ cells compared to their respective Sham-treated mice (two-way ANOVA, main effect [controlled cortical impact vs. sham]: F_1,40 = 10.33, P = 0.003; main effect [ketamine vs. vehicle]: F_1,40 = 0.76, P = 0.390). However, when analyzed via pairwise comparisons, bromodeoxyuridine labeling was significantly increased only in the controlled cortical impact + ketamine group compared to its respective sham (fig. 1 D; Tukey honest significant difference post hoc test, controlled cortical impact + vehicle vs. sham + vehicle: P = 0.288; controlled cortical impact + ketamine vs. sham + ketamine: P = 0.043), indicating that ketamine exposure early after controlled cortical impact increases the survival of cells that were between 2 and 6 weeks of age.

Overall, the patterns of post–traumatic brain injury cell proliferation and longer-term cell survival indicate that ketamine has both an acute effect on cell proliferation during its administration, as well as longer-term effects on cell survival that outlast the window of ketamine exposure (fig. 1E). These effects could result from early induction of long-term changes in gene expression that drive longer-term cell survival, or alternatively, by changes in cell fate specification, as different cell types would be expected to have differential proliferation and survival trajectories. Thus, we examined the phenotypes of bromodeoxyuridine-labeled cells after controlled cortical impact using cell-type specific immunohistochemistry.

We analyzed the early phase of posttraumatic neurogenesis 2 weeks after traumatic brain injury using tissue from sham- and controlled cortical impact–treated mice that had received either ketamine or vehicle during the first postinjury week. Cells that had assumed a neuronal fate were identified by staining for the immature neuronal marker doublecortin, and bromodeoxyuridine colabeling was determined using confocal microscopy. As doublecortin is expressed in neurons approximately 3 to 21 days after mitosis,²¹  doublecortin+/bromodeoxyuridine+ colabeling can help define the fate of proliferating cells early after the injury, during the peak window for posttraumatic neurogenesis.²²  Additionally, the overall density of doublecortin+ cells can serve to identify and quantify all of the recently born neurons generated over a wider (~2.5 weeks) timeframe.

We first examined whether ketamine administration altered the density of doublecortin staining in injured brains. Ketamine administration significantly decreased neurogenesis in controlled cortical impact–treated mice at this 2-week time point, manifest as a decreased density of doublecortin+ cells in ketamine-treated mice after controlled cortical impact when compared with vehicle-treated, controlled cortical impact–exposed mice (fig. 2, A and B; two-way ANOVA, main effect [ketamine vs. vehicle]: F_1,48 = 4.93, P = 0.031; Tukey honest significant difference post hoc test: controlled cortical impact + ketamine vs. controlled cortical impact + vehicle, P = 0.048). Controlled cortical impact exposure itself did not significantly increase neurogenesis when analyzed at the 2-week time point using doublecortin immunohistochemistry (two-way ANOVA, main effect [controlled cortical impact vs. sham]: F_1,48 = 0.80, P = 0.376).

To more precisely determine the fate of cells born early after traumatic brain injury, we examined bromodeoxyuridine+/doublecortin+ colabeled cells in the dentate gyrus 2 weeks after controlled cortical impact. Surprisingly, despite the fact that ketamine treatment alone did not affect overall levels of hippocampal cell proliferation (fig. 1), ketamine treatment decreased the number of neurons born 2 to 3 days after controlled cortical impact across both sham and controlled cortical impact conditions (fig. 2, A and C; two-way ANOVA: F_1,48 = 4.05; main effect [ketamine vs. vehicle], P = 0.0499). However, controlled cortical impact itself did not increase neurogenesis (two-way ANOVA, main effect [controlled cortical impact vs. sham]: F_1,48 = 0.97, P = 0.329) at this time point.

Altered cell proliferation in mice after controlled cortical impact might have long-term effects on the ability of the hippocampus to sustain constitutive levels of adult neurogenesis, as has previously been observed in studies of postseizure neurogenesis.²³  Thus, we examined the density of immature neurons 6 weeks after injury by staining for the immature neuronal marker doublecortin, to focus on cells born several weeks after controlled cortical impact.

Six weeks after injury, the density of immature neurons was significantly reduced when comparing all animals that received controlled cortical impact to those that received sham injury (fig. 3, A and B; two-way ANOVA, main effect [controlled cortical impact vs. sham]: F_1,40 = 4.25, P = 0.046), suggesting that in contrast to the burst of cell proliferation noted after controlled cortical impact, the longer-term impact of controlled cortical impact is to reduce neurogenesis at more remote time points. This effect was not significantly modified by ketamine (fig. 3B; two-way ANOVA, main effect [ketamine vs. vehicle]: F_1.40 = 0.11, P = 0.744), indicating that ketamine treatment did not affect neurogenesis at remote time points after injury in either sham- or controlled cortical impact–treated mice.

At 3 to 4 weeks after mitosis, granule cells begin expressing neuronal nuclear protein, a marker of mature neurons. During constitutive adult neurogenesis, most immature adult-born neurons do not survive to maturity²⁴ ; similarly, during postischemic neurogenesis in the subventricular zone, many immature neurons born after stroke also do not survive to maturity.²⁵  As NMDAR activation is required for the long-term survival of constitutively generated adult-born neurons,⁷  we hypothesized that survival of neurons born after traumatic brain injury would be modulated by ketamine.

To evaluate ketamine’s impact on the long-term survival of newly born neurons, slices were evaluated 6 weeks after injury via antibody staining with the mature neuronal marker neuronal nuclear protein. Although neurogenesis (neuronal nuclear protein+/bromodeoxyuridine+ colabeled cell density) increased in the controlled cortical impact–treated groups versus the sham-treated groups, this did not reach statistical significance (cells/mm³, mean ± SD: sham + vehicle = 3,400 ± 3,100; controlled cortical impact + vehicle = 4,900 ± 5,100; sham + ketamine = 3,400 ± 1,600; controlled cortical impact + ketamine = 5,000 ± 3,500; two-way ANOVA, main effect [controlled cortical impact vs. sham]: F_1,40 = 2.44, P = 0.126). Importantly, there was no difference in the density of neurons born after controlled cortical impact that survived until maturity when ketamine was administered for the week after controlled cortical impact (main effect [ketamine vs. vehicle]: F_1,40 = 0.00, P = 0.959; Tukey honest significant difference post hoc test: controlled cortical impact + ketamine vs. controlled cortical impact + sham, P = 0.999), indicating that ketamine treatment did not negatively affect the ultimate survival of neurons born early after injury.

Because radial glia-like cells can differentiate into other cell types and only a minority of the bromodeoxyuridine+ cells observed colabeled with neuronal markers, we stained tissues at various time points for glial markers to determine whether ketamine exposure altered cell fate and differentially affected the proliferation of glia. Astrocytes were identified via antibody staining against the astrocyte-selective cytoskeletal protein glial fibrillary acidic protein, which also serves as a marker for astrocyte activation. Thus, increased glial fibrillary acidic protein staining could either result from increased astroglial reactivity, an increase in the number of astrocytes (or radial glia-like cells, which are also glial fibrillary acidic protein+), or both. Astrogliogenesis was quantified by costaining tissue with glial fibrillary acidic protein and bromodeoxyuridine antibodies, to identify both overall glial fibrillary acidic protein staining, as well as to specifically identify astrocytes born after controlled cortical impact or sham injury.

Three days after the intervention, glial fibrillary acidic protein expression was significantly increased in the controlled cortical impact–treated groups compared to their respective sham groups (two-way ANOVA, main effect [controlled cortical impact vs. sham]: F_1,18 = 18.46, P < 0.001; main effect [ketamine vs. vehicle]: F_1,18 = 0.27, P = 0.609). This increase was similar between both drug treatment groups; however, when analyzed via pairwise comparison, glial fibrillary acidic protein expression was significantly increased only in the controlled cortical impact + ketamine group compared to its respective sham (glial fibrillary acidic protein+ cells/mm³ ± SD, Tukey honest significant difference post hoc test: sham + vehicle [56,200 ± 22,200, n = 6] vs. controlled cortical impact + vehicle [91,100 ± 23,000, n = 4], P = 0.095; sham + ketamine [52,800 ± 14,800, n = 6] vs. controlled cortical impact + ketamine [97,200 ± 26,000, n = 6], P = 0.012). Examining astrocytes born specifically in the immediate postinjury period via glial fibrillary acidic protein+/bromodeoxyuridine+ colocalization, controlled cortical impact–treated groups demonstrated an increase in astrocytosis when compared to sham groups (two-way ANOVA, main effect [controlled cortical impact vs. sham]: F_1,18 = 6.16, P = 0.023; main effect [ketamine vs. vehicle]: F_1,18 = 1.30, P = 0.269); however, no significant pairwise differences between any groups were observed (glial fibrillary acidic protein+/bromodeoxyuridine+ cells/mm³ ± SD, Tukey honest significant difference post hoc test: sham + vehicle [1,400 ± 1,300, n = 6] vs. controlled cortical impact + vehicle [4,500 ± 2,000, n = 4], P = 0.210; sham + ketamine [2,800 ± 1,400, n = 6] vs. controlled cortical impact + ketamine [4,900 ± 3,700, n = 6], P = 0.456). Taken together, these findings indicate that ketamine either facilitated the activation of glial cells or increased the density of controlled cortical impact–induced astrocytes through proliferation before the second post–controlled cortical impact day.

To assess continued proliferation and survival of the astrocytes born after injury, we evaluated glial fibrillary acidic protein staining at 2 and 6 weeks postintervention. Bromodeoxyuridine was administered 2 and 3 days after sham or controlled cortical impact treatment, to identify astrocytes generated during the early postinjury period. Despite no difference in astrocyte generation at 3 days after injury, at the 2-week time point, we noted a dramatic increase in astrogliogenesis (glial fibrillary acidic protein+/bromodeoxyuridine+ colabeled cell density) after controlled cortical impact in vehicle-treated mice (fig. 4, A and B; two-way ANOVA, main effect (controlled cortical impact vs. sham): F_1,48 = 6.65, P = 0.013; main effect [ketamine vs. vehicle]: F_1,48 = 1.52, P = 0.224; interaction effect [controlled cortical impact:ketamine]: F_1,48 = 3.75, P = 0.059; Tukey honest significant difference post hoc test: sham + vehicle vs. controlled cortical impact + vehicle, P = 0.013), which was almost completely abrogated by ketamine exposure (fig. 4, A and B; Tukey honest significant difference post hoc test: sham + ketamine vs. controlled cortical impact + ketamine, P = 0.955). By 6 weeks after intervention, the populations of glial fibrillary acidic protein+/bromodeoxyuridine+ cells equalized between all groups (data not shown).

Microglia are the resident macrophages of the central nervous system and have a separate lineage than hippocampal radial glial-like stem cells and their progeny. In healthy tissue, microglia play important roles in sculpting neuronal circuits via synapse pruning and immune surveillance.²⁶  After traumatic brain injury, microglia become activated and mount a dramatic response, including the release of pro- and antiinflammatory cytokines and increased phagocytosis.²⁷  Moreover, activated microglia have complex effects on neurogenesis.²⁸  We evaluated microglial proliferation in the granule cell layer after injury, using antibodies against the macrophage/microglia-specific marker ionized calcium-binding adaptor molecule 1 and the marker for activated microglia, Galectin-3.

Three days after the controlled cortical impact, ketamine treatment dramatically increased the density of newly generated (bromodeoxyuridine+) ionized calcium-binding adaptor molecule 1–labeled microglia within the granule cell layer (fig. 5, A and B; two-way ANOVA, main effect [controlled cortical impact vs. sham]: F_1,18 = 16.45, P = 0.001; main effect [ketamine vs. vehicle]: F_1,18 = 6.70, P = 0.019; Tukey honest significant difference post hoc tests: sham + ketamine vs. controlled cortical impact + ketamine, P = 0.001; controlled cortical impact + ketamine vs. controlled cortical impact + vehicle, P = 0.012; controlled cortical impact + ketamine vs. sham + vehicle, P = 0.001). This increase was statistically significant compared with all other groups and included a significant interaction effect with injury, indicating a vastly different effect of ketamine on microglial proliferation after controlled cortical impact that was not present in sham-treated mice (fig. 5B; interaction effect [controlled cortical impact:ketamine]: F_1,18 = 7.46, P = 0.014). Surprisingly, there was no controlled cortical impact–associated increase in ionized calcium-binding adaptor molecule 1+/bromodeoxyuridine+ microglia in the granule cell layer of the vehicle-treated mice group at this early time point (fig. 5B; Tukey honest significant difference post hoc test: controlled cortical impact + vehicle vs. sham + vehicle, P = 0.907).

Galectin-3 staining of these samples demonstrated no significant differences between any groups, indicating the microglia present in the granule cell layer were not more activated with exposure to controlled cortical impact or ketamine at this time point (Galectin-3+ cells/mm³, mean ± SD: sham + vehicle, 55,700 ± 11,100; controlled cortical impact + vehicle, 61,400 ± 7,500; sham + ketamine, 42,500 ± 11,200; controlled cortical impact + ketamine, 44,800 ± 14,200; two-way ANOVA, main effect (controlled cortical impact vs. sham): F_1,18 = 0.10, P = 0.753; main effect [ketamine vs. vehicle]: F_1,18 = 1.42, P = 0.248).

Two weeks after injury, there was an overall increase in the density of ionized calcium-binding adaptor molecule 1+/bromodeoxyuridine+ microglia in all controlled cortical impact-treated mice relative to sham mice, but no significant difference between groups induced by ketamine treatment (two-way ANOVA: F_1,48 = 6.24; main effect [controlled cortical impact vs. sham], P = 0.016; main effect [ketamine vs. vehicle]: F_1,48 = 0.73, P = 0.398). This result indicates that either delayed microglial maturation or ongoing microglial proliferation in the vehicle-treated group resulted in similar densities of new microglia at this more delayed time point. By 6 weeks postinjury, the colabeled (bromodeoxyuridine+/ ionized calcium-binding adaptor molecule 1+) population of microglia in the controlled cortical impact–injured groups remained significantly increased versus their sham counterparts, suggesting increased persistence of injury-induced microglia in the granule cell layer, with no significant difference between vehicle- and ketamine-treated groups (fig. 5C; two-way ANOVA, main effect [controlled cortical impact vs. sham]: F_1,40 = 28.69, P < 0.001; main effect [ketamine vs. vehicle]: F_1,40 = 1.87, P = 0.179; Tukey honest significant difference post hoc test: sham + vehicle vs. controlled cortical impact + vehicle, P = 0.045; sham + ketamine vs. controlled cortical impact + ketamine, P < 0.001).

If ketamine has neuroprotective properties, ketamine-induced modification of injury severity could indirectly alter the nature of post–traumatic brain injury cell proliferation between groups. For example, if ketamine treatment modified the timing or degree of cell death, it could alter the proliferative response in a quantitative or qualitative manner. Moreover, as ketamine could also alter neuronal activity after traumatic brain injury, including potential modulation of subclinical seizure frequency/occurrence,²⁹  this could also alter neurogenesis and gliogenesis via activity-dependent mechanisms. We, therefore, attempted to identify possible indirect consequences of ketamine treatment that could lead to secondary changes in cell proliferation.

To grossly assess the impact of ketamine on neuronal injury, we first measured the degree of cortical cavitation using low-power images of coronal brain sections. We compared the amount of tissue lost after injury at both 2 and 6 weeks post–controlled cortical impact. The cross-sectional area of cortical cavity size was not affected by postinjury ketamine treatment at either time point (fig. 6, A and B; independent t tests: controlled cortical impact + vehicle vs. controlled cortical impact + ketamine, 2 weeks: P = 0.105; 6 weeks: P = 0.517). Although all cavities appeared round on gross inspection before sectioning, it remained possible that our coronal cross-sectional analysis may have missed subtle differences in three-dimensional cavity shape or profile. Thus, we performed additional analyses of injury severity.

The amount of cortical loss would likely not be sensitive to differences in hippocampal cell apoptosis early after controlled cortical impact, which might be more directly related to cell proliferation in the hippocampus. To evaluate hippocampal cell death, we performed immunohistochemical staining for activated caspase-3, a marker for the induction of apoptosis, at the same time point at which mitosis was measured (3 days postinjury or -sham), and quantified caspase-3 labeling in the dentate gyrus of the ipsilateral hippocampus. There were no significant differences in caspase-3 expression between any of the groups at 3 days postintervention (caspase-3+ cells/mm³, mean ± SD: sham + vehicle [1,100 ± 1,100, n = 6]; controlled cortical impact + vehicle [500 ± 400, n = 4]; sham + ketamine [1,000 ± 700, n = 6]; controlled cortical impact + ketamine [800 ± 600, n = 6]; two-way ANOVA, main effect [controlled cortical impact vs. sham]: F_1,18 = 1.50, P = 0.236; main effect [ketamine vs. vehicle]: F_1,18 = 0.03, P = 0.875). This lack of caspase activation suggests a lack of ongoing apoptosis at this time point, but might also have resulted from a rapid clearance of any apoptotic cells such that ongoing caspase-3 expression was not detected.

Subclinical seizures or changes in neuronal activity can influence postinjury histopathology and recovery.³⁰  To evaluate this, we performed antibody staining for cfos, a metabolic marker upregulated by neurons after recent activity. Though paradoxical, it has previously been shown that traumatic brain injury attenuates neuronal activity for up to a month after injury.³¹  Consistent with this previous finding, we observed a significant decrease in neuronal activity in all controlled cortical impact–treated mice versus all sham-treated mice 3 days after injury, with no difference in cfos activity as a result of ketamine administration (fig. 6, C and D; two-way ANOVA, main effect [controlled cortical impact vs. sham]: F_1,19 = 4.68, P = 0.044; main effect [ketamine vs. vehicle]: F_1,19 = 0.12, P = 0.735).

To determine how ketamine administration after traumatic brain injury might affect cognitive recovery, mice underwent behavioral testing 4 weeks after injury using a modified Morris water maze task. Although the standard reference memory water maze paradigm is sensitive to changes in spatial learning, there is considerable discrepancy in its ability to capture performance differences when hippocampal neurogenesis is altered.³²  Because the function of new neurons is believed to involve context separation, we focused our water maze test on a Morris water maze reversal paradigm, which requires distinction between very similar contexts (i.e., new vs. old target location).³³  Details of the water maze test are described in the methods section, but in general, once all groups of mice were sufficiently trained such that they showed equal levels of preference for the first target location (platform), we changed the target location and assessed whether mice could distinguish between the two learned target locations. Performance in this Morris water maze reversal task requires intact and properly functioning hippocampal neurogenesis, which is thought to allow mice to distinguish between different contextual location cues between the two different platform locations that are learned in the task.³³ 

In the first phase of the water maze test, mice attempted to locate and then recall the location of a submerged platform. Across the average of the training sessions, controlled cortical impact–treated mice required more time to locate the hidden platform (mean time in seconds ± SD: sham = 28.7 ± 6.2, n = 24; controlled cortical impact = 36.2 ± 7.3, n = 23; two-way ANOVA, main effect [controlled cortical impact vs. sham]: F_1,43 = 13.75, P = 0.001). Despite this difference, all groups of mice improved their ability to locate the hidden platform during repeated sessions (repeated measures ANOVA, effect of session: F_4,172 = 34.82, P = 0.001) and had no group differences in their learning rates (fig. 7A; repeated measures ANOVA, controlled cortical impact × session: F_4,172 = 1.45; P = 0.220; drug × session: F_4.172 = 0.52; P = 0.725). Furthermore, there were no group differences between sham- and controlled cortical impact–treated mice on session 4, the last session before the first probe trial (fig. 7A; two-way ANOVA, main effect [controlled cortical impact vs. sham]: F_1,43 = 2.55, P = 0.117). In the first probe trial, all groups of mice showed a preference for the learned location of the hidden platform as all groups of mice spent significantly more time in the target quadrant versus any of the other three nontarget quadrants (fig. 7, A and B; Dunnett multiple comparison post hoc test: P < 0.050 for each quadrant vs. the target quadrant, for all groups of mice).

After all mice had learned the first platform location, mice were subjected to a reversal Morris water maze test, in which the platform location was changed to another quadrant of the pool, a task more specific for the assessment of neurogenesis-dependent hippocampal function.³³  An effect of controlled cortical impact was again observed across the average of the hidden platform sessions, as controlled cortical impact-treated mice took longer to locate the hidden platform (mean time in seconds ± SD: sham = 29.9 ± 10.0, n = 24; controlled cortical impact = 37.3 ± 13.4, n = 20; two-way ANOVA, main effect [controlled cortical impact vs. sham]: F_1,43 = 4.43, P = 0.041). Similar to the first training, all groups of mice improved in their ability to locate the hidden platform during training (repeated measures ANOVA, effect of session: F_1,43 = 22.77, P = 0.001) without any group differences in the learning rates (repeated measures ANOVA, controlled cortical impact × session: F_1,43 = 0.02; P = 0.895; drug × session: F_1,43 = 1.58; P = 0.216). Although across the average of the sessions controlled cortical impact–treated mice took longer to locate the hidden platform, there were no significant group differences in their acquisition of the new target location during the last session before the second probe trial (two-way ANOVA, main effect [controlled cortical impact vs. sham]: F_1,43 = 2.92; P = 0.095).

During the reversal task probe trial the following day, all sham-treated mice (both vehicle- and ketamine-exposed) demonstrated a clear bias toward the new location of the platform (target quadrant vs. any other nontarget quadrant: Dunnett multiple comparison post hoc test, P < 0.050 for each quadrant vs. the target quadrant). This correct preference was deficient in controlled cortical impact–injured mice treated with vehicle (target quadrant vs. other quadrants: Dunnett multiple comparison post hoc test, P = 0.637), indicating this group had a deficit in water maze reversal memory (figure 7, C and D). Surprisingly, ketamine treatment after controlled cortical impact rescued the ability of injured mice to retain the new platform location during the probe trial, as these mice demonstrated preservation of spatial memory retention during reversal water maze testing (fig. 7, C and D; Dunnett multiple comparison post hoc test, P < 0.050 for each quadrant vs. the target quadrant). These differences were not attributable to altered motor coordination, vision, or motivation to perform the task, as there was no effect of controlled cortical impact, drug treatment, or interaction on swim speed or latency of mice to reach the “visible” platform when it was later clearly marked with a beacon and there were no group differences across the average of these visible platform trials (two-way ANOVA, main effect [controlled cortical impact vs. sham]: F_1,42 = 2.48; P = 0.123; main effect [ketamine vs. vehicle]: F_1,42 = 2.36; P = 0.132).

Although vehicle-treated controlled cortical impact mice did not spend statistically more time in the target quadrant during the reversal probe trial, we found that the general pattern between groups appeared grossly similar (all spent the most time in the target quadrant) and thus we performed additional analyses of our data. First, we performed a different version of the quadrant preference analysis, in which all nontarget quadrants are averaged together in an attempt to minimize the effect of a single spurious quadrant, and then compared this average with the time spent in the target quadrant. This analysis again demonstrated significant target preference in the two sham groups and the controlled cortical impact + ketamine group (sham + vehicle, P = 0.0174, n = 12; sham + ketamine, P < 0.001, n = 12; controlled cortical impact + ketamine, P = 0.008, n = 12; paired t tests) but not the vehicle-treated controlled cortical impact group (P = 0.088, n = 11, paired t test). Additionally, we reanalyzed the trajectory data during the probe trial to determine the amount of time spent above the target platform location during the reversal probe trial, and also found that these data had the same pattern. Compared to sham + vehicle mice, controlled cortical impact + vehicle treated mice had significantly fewer crossings (sham + vehicle = 4.0 ± 2.9; controlled cortical impact + vehicle = 2.0 ± 1.3, P = 0.046, Mann–Whitney test), whereas the number of crossings between sham + ketamine and controlled cortical impact + ketamine treated mice were not different from each other (sham + ketamine = 2.9 ± 2.1; controlled cortical impact + ketamine = 2.5 ± 1.6, P = 0.610, Mann–Whitney test). As all groups of mice successfully learned the location of the new target location during training (see fig. 7A), but only the controlled cortical impact + vehicle mice were selectively deficient in recalling during the probe trial 24 h later, our data suggest that their deficit was more likely related to recall rather than the ability to acquire the new location per se.

Traumatic brain injury induces robust cellular proliferation in the hippocampus, which includes the generation of immature neurons and astrocytes from radial glial-like cells.^18,19  In this study, when ketamine was administered after controlled cortical impact, hippocampal cell proliferation in the dentate granule cell layer was markedly accelerated, as was evident by an increase in bromodeoxyuridine labeling of mitotic cells at the 3-day time point, which was mostly a product of early increases in microgliogenesis—a surprising finding given ketamine’s antiinflammatory properties.³⁴  Interestingly, these changes in proliferation were accompanied by improvement in behavioral outcomes, suggesting there may be a benefit to ketamine administration in the setting of head injury.

Neurologic insults such strokes, seizures, and traumatic brain injuries are associated with increased hippocampal cell proliferation, including neurogenesis.⁴  These pathologies often involve pathologic neuronal excitation, leading to the release of glutamate and other neurotransmitters, which can increase metabolic demands and cause excitotoxic neuronal death. Injury-associated glutamate release can also activate NMDARs on neuronal stem cells and immature neurons. As NMDA-type inotropic glutamate receptor activation modulates both the proliferation^5,6  and the survival of adult-born neurons,⁷  these receptors could mediate injury-induced changes in adult neurogenesis. Additionally, glutamatergic signaling via NMDARs can also influence glial cell signaling both directly and indirectly,⁹  which could mediate additional quantitative and qualitative changes in both gliogenesis and neurogenesis.¹¹ 

Our data demonstrating that ketamine alters cell proliferation after traumatic brain injury are consistent with a role for NMDARs in the control of cell proliferation after injury. Several unexpected findings, however, suggest potentially novel mechanisms at play in mediating these responses. Ketamine administration notably reduced neurogenesis in both injured and noninjured mice 2 weeks postinjury. Ketamine has proven antidepressant efficacy in human subjects,³⁵  yet classic antidepressants have been shown to increase neurogenesis³⁶ ; this finding suggests a potentially different mechanism of action for ketamine’s antidepressant action. Previous studies of short-term ketamine administration did not demonstrate changes in adult neural cell proliferation in healthy rodents.^37,38  One study did report that intermittent ketamine administration could increase neurogenesis in some portions of the hippocampus, but not others,³⁹  and a separate study demonstrated that repeated ketamine administration could decrease survival of adult-born neurons.⁴⁰  However, no studies have previously examined a longer-term continuous administration of ketamine after traumatic brain injury, which might be a better model for continuous human exposure via infusion, particularly in light of the extremely short serum half-life of ketamine in mice (13 min).

Although 1 week of ketamine administration reduced postinjury neurogenesis, its most dramatic effects were on glial proliferation (both microglia and astrocytes), suggesting that ketamine might not be working exclusively via the control of neuronal function. In this regard, microglial activation directly inhibits neurogenesis²⁶  and astrocytes release factors that have positive influences on neurogenesis,¹¹  raising the possibility that either the early increase in microglial proliferation or ketamine-associated decrease in astrocytosis could have contributed to the reduction in doublecortin+ cell counts at 2 weeks after injury in the ketamine-treated group. Moreover, a lack of a ketamine-induced change in cfos activity after injury argue against alterations in neuronal circuit activity as the primary mediator of ketamine’s effects. Future studies will hopefully delineate the specific receptor mechanisms mediating ketamine’s actions on postinjury neurogenesis.

Although ketamine reduced the density of doublecortin-stained immature neurons 2 weeks after injury, we did not find that it substantially altered the long-term survival of neurons born after traumatic brain injury. One limitation of our study resulted from an unexpectedly high variability in posttraumatic neurogenesis within our controlled cortical impact–treated mice, such that our data displayed enhanced injury-induced neurogenesis, which was not statistically significant. It has been well established in the literature that controlled cortical impact induces neurogenesis,^14,18  however, experimental variability has led to negative results in some studies.⁴¹  This higher-than-expected variability may have left our study underpowered to determine statistically significant increases in neurogenesis, and thus, left us unable to find group differences between drug conditions. Our sample sizes were based on power analyses using data obtained in our own lab using the same model and, thus, we did not feel it appropriate to continue to add additional mice in the quest for statistical significance in regards to neurogenesis, particularly in light of the multiple findings in regards to glial proliferation. However, it does have implications for sample sizes necessary for future studies.

One of our most surprising findings involved the ketamine-associated changes in postinjury gliogenesis. To our knowledge, no previous studies have reported an effect of ketamine on the generation of new glial cells after traumatic brain injury. There are, however, previous studies evaluating the effects of ketamine on glial function, which suggest possible antiinflammatory effects. Ketamine suppressed astrocyte activation in a mouse model of neuropathic pain,⁴²  and can directly alter astrocyte function in vitro.⁴³  Ketamine also reduces cytokine production by both astrocytes and microglia in vitro,⁴⁴  and proinflammatory cytokines have been shown to favor astrocytogenesis,⁴⁵  suggesting that ketamine could be inhibiting astrocytosis after injury by decreasing injury-induced cytokine signaling.

Furthermore, although ketamine is an NMDAR antagonist, it interacts with numerous other receptors,^17,46  and its metabolites have therapeutic effects that are not mediated by actions at NMDARs.⁴⁷  Thus, the effects of ketamine noted in this study may have been due to non–NMDA-type inotropic glutamate receptor–mediated mechanisms. Interestingly, activation of the proinflammatory fibroblast growth factor receptor promotes astrocytosis, and as ketamine downregulates fibroblast growth factor signaling,⁴⁸  this could provide one potential mechanism through which ketamine suppresses astrocytosis after traumatic brain injury.

Fascinatingly, despite no differences in the long-term survival of neurons generated during early posttraumatic neurogenesis, mice treated with ketamine after injury had improved learning in the water maze reversal task, suggesting a benefit to receiving ketamine after injury. The Morris water maze reversal test relies on the presumed neurogenesis-dependent ability to distinguish between similar environments and to supplant previously learned preferences given new, subtle environmental cues. Although performance in this task has been interpreted as related to rodents’ ability to successfully engage in pattern separation between the environmental contexts between the two different platform locations,³³  this task also involves other cognitive processes, such as behavioral adaptability, and we expect that future work will help elucidate the specific features that might be improved by ketamine treatment.

Increased neurogenesis after traumatic brain injury has previously been associated with improved Morris water maze performance, and Blaiss et al. demonstrated that ablation of neurogenesis after traumatic brain injury specifically impairs reversal Morris water maze.⁴⁹  Conversely, Gradari et al. detailed a very specific association between decreased immature neuron populations and improved performance in the reversal Morris water maze,⁵⁰  which may best explain the improvement seen in this study, as ketamine exposure reduced neurogenesis at the 2-week time point. It is not immediately clear, however, how an early decrease in neurogenesis after injury might explain such an association, given that we did not find any ketamine-induced change in neurogenesis at the 6-week time point. One potential explanation would be that the improved performance after ketamine exposure related to qualitative, and not quantitative, differences in postinjury neurogenesis.

Another explanation for the improved behavioral outcome after ketamine exposure might relate to the changes in glial function after injury. Intense astrocyte activation and scar formation has been associated with behavioral deficits, raising the possibility that muting this reactivity (as was noted 2 weeks after controlled cortical impact in the ketamine-treated group) was beneficial. Similarly, the accelerated proliferation and sustained survival of microglia generated after traumatic brain injury might also altered the functional properties of hippocampal neurons, as it has recently become clear that microglia play important roles in sculpting neuronal circuit function.²⁶  As the behavioral improvements and histologic findings in our study are purely associative, and not causal, further evaluation of either of these possibilities through the specific targeting of glial function using genetic or pharmacologic means would hopefully shed light on these potential mechanisms.

Alternatively, hippocampal neurons born after various forms of neuronal injury have altered structural and functional properties,¹⁴  and ketamine could serve to improve the functional capacity of neurons born after traumatic brain injury. Ketamine rapidly induces synaptogenesis,³⁶  although it is not clear how long these changes would persist or translate to cells which had not yet received excitatory inputs during the period of ketamine administration. Although we attempted to morphologically characterize immature doublecortin+ neurons in this study, neither dendritic morphology nor dendritic spine density (a surrogate for excitatory synaptic innervation) could be accurately discerned using this method (data not shown), thus the potential effect of ketamine on structural maturation was not assessed. Further studies using genetic or retrovirus-based cell labeling techniques, in conjunction with electrophysiologic analyses, could help define the morphologic and functional properties of neurons born after controlled cortical impact in the presence of ketamine to identify potential differences. Additionally, a broader array of behavioral assessments will allow us to more precisely delineate the scope of ketamine’s effects on outcomes after traumatic brain injury, as changes in hippocampal cell proliferation after injury could also have deleterious consequences that were not detected in our current study.

Ketamine administration after traumatic brain injury alters hippocampal cell proliferation. These changes are associated with an improvement in learning weeks after ketamine cessation, as determined by the Morris water maze reversal task. Additional studies focusing on the cellular mechanisms underlying these responses to ketamine after traumatic brain injury are needed, so that we may help refine strategies to improve neurologic outcomes after brain injury, and determine whether ketamine might also improve outcomes in humans after traumatic brain injury.

The authors would like to thank the following: Gary Westbrook, M.D., senior scientist at the Vollum Institute, Portland, Oregon, for project guidance; members of the Westbrook and Schnell labs, Portland, Oregon, for helpful discussions; David Yanez, Ph.D., associate professor of biostatistics and codirector of the Biostatistics and Design Program, Oregon Health and Science University, Portland, Oregon, for statistical guidance and interpretation; and Sarah Mader, B.S., research assistant, Department of Anesthesiology, Oregon Health and Science University, for outstanding technical support. The contents of this manuscript do not represent the views of the U.S. Department of Veterans Affairs or the United States government.

Primary funding for this project was provided by a Foundation for Anesthesia Education and Research (Schaumburg, Illinois) Research Fellowship Grant (to Dr. Peters). Additional funding was provided by a BIRCWH K12 award made possible through the Eunice Kennedy Shriver National Institute of Child Health and Human Development (Bethesda, Maryland) and the Office of Research on Women’s Health (Bethesda, Maryland; grant No. K12 HD 043488; to Dr. Villasana), a Department of Veterans Affairs, Veterans Health Administration, Office of Research and Development (Washington, D.C.), Biomedical Laboratory Research and Development (Washington, D.C.) CDA-2 Award 005-10S (to Dr. Schnell), a Department of Veterans Affairs Merit Review Award I01-BX002949 (to Dr. Schnell), and Oregon Health and Science University (Portland, Oregon) Anesthesiology and Perioperative Medicine departmental funds. Supplementary equipment and developmental funds were provided by the Oregon Clinical and Translational Research Institute (Portland, Oregon) grant No. TL1TR000129 and National Institute for Neurological Disorders and Stroke (Bethesda, Maryland) grant No. P30 NS061800.

The authors declare no competing interests.