Deep cerebellar stimulation enhances cognitive recovery after prefrontal traumatic brain injury in rodent

Abstract

Functional outcome following traumatic brain injury (TBI) varies greatly, with approximately half of those who survive suffering long-term motor and cognitive deficits despite contemporary rehabilitation efforts. We have previously shown that deep brain stimulation (DBS) of the lateral cerebellar nucleus (LCN) enhances rehabilitation of motor deficits that result from brain injury. The objective of the present study was to evaluate the efficacy of LCN DBS on recovery from rodent TBI that uniquely models the injury location, chronicity and resultant cognitive symptoms observed in most human TBI patients. We used controlled cortical impact (CCI) to produce an injury that targeted the medial prefrontal cortex (mPFC-CCI) bilaterally, resulting in cognitive deficits. Unilateral LCN DBS electrode implantation was performed six weeks post-injury. Electrical stimulation started at week eight post-injury and continued for an additional four weeks. Cognition was evaluated using baited Y-maze, novel object recognition task and Barnes maze. Post-mortem analyses, including Western Blot and immunohistochemistry, were conducted to elucidate the cellular and molecular mechanisms of recovery. We found that mPFC-CCI produced significant cognitive deficits compared to pre-injury and naïve animals. Moreover, LCN DBS treatment significantly enhanced the long-term memory process and executive functions of applying strategy. Analyses of post-mortem tissues showed significantly greater expression of CaMKIIα, BDNF and p75^NTR across perilesional cortex and higher expression of postsynaptic formations in LCN DBS-treated animals compared to untreated. Overall, these data suggest that LCN DBS is an effective treatment of cognitive deficits that result from TBI, possibly by activation of ascending, glutamatergic projections to thalamus and subsequent upregulation of thalamocortical activity that engages neuroplastic mechanisms for facilitation of functional reorganization. These results support a role for cerebellar output neuromodulation as a novel therapeutic approach to enhance rehabilitation for patients with chronic, post-TBI cognitive deficits that are unresponsive to traditional rehabilitative efforts.

Introduction

Traumatic brain injury (TBI) disrupts normal brain function as a result of mechanical damage to brain tissue. In the United States alone, it is estimated that there are between 3.2 and 5.3 million individuals suffering from chronic, post-TBI disability (Corrigan et al., 2014). It disproportionally affects younger individuals, and causes long-term sequelae such as altered consciousness as well as a wide range motor, cognitive and behavioral deficits (Chen and Loya, 2014; Menon et al., 2010; Tefertiller et al., 2011). Following emergent treatment, individuals initially undergo physical and occupational rehabilitation (Khan et al., 2003; Perlesz, 1999). However, outcomes greatly vary, which is due to the heterogeneity of the traumatic severity and further increasing severity, with an estimated 40–57% of patients left with persistent motor or cognitive disabilities (Corrigan et al., 2014; Marklund et al., 2019; Oberholzer and Müri, 2019). Such numbers underscore the need to develop novel therapeutic approaches to improve rehabilitation outcomes. Such efforts often involve pharmacologic-(Pavlovskaya et al., 2007; Whyte et al., 2002) or neurostimulation-based (Issar et al., 2013; Schiff et al., 2007; Shin et al., 2014) strategies to enhance functional reorganization, most often across peri- or contralesional cortical brain regions.

Our group has shown previously that electrical stimulation of the cerebellothalamocortical (CTC) pathway using a deep brain stimulation (DBS)-based approach enhances rehabilitation of chronic motor deficits in a rat model of stroke as well as in a rat fluid percussion injury model of sensorimotor TBI (Chan et al., 2018a; Chan et al., 2018b; Machado and Baker, 2012; Machado et al., 2009; Machado et al., 2013). Our targeting of the lateral cerebellar nucleus (LCN; the rodent homologue of the human dentate nucleus) was driven by its position as a final nodal point in the cerebellum’s ascending projections, with dense, glutamatergic terminals projecting contralaterally across both mediodorsal and ventrolateral thalamic nuclei. These connections place the LCN in an ideal position to modulate thalamocortical interactions across widespread prefrontal, frontal, and parietal circuits (Dum et al., 2002; Dum and Strick, 2003; Tracey et al., 1980). We have established previously that LCN DBS modulates contralateral cerebral cortical excitability in both naïve (Baker et al., 2010) and post-stroke animals (Park et al., 2015). The magnitude and persistence of that effect was found to be sensitive to pulse frequency, with stimulation in the range of 30 Hz associated with maximal, subacute enhancement of cortical excitability. Since then, we have shown repeatedly that 30 Hz LCN DBS enhances motor rehabilitation in chronic, post-stroke animals (Machado et al., 2009; Machado et al., 2013; Cooperrider et al. 2014) and, more recently, in a chronic rodent model of sensorimotor TBI (Chan et al. 2018b). At the cellular and molecular level, the behavioral benefits were accompanied by increased synaptogenesis and evidence of neurogenesis in perilesional cerebral cortex when compared to surgical controls (Chan et al., 2018a; Cooperrider et al., 2014; Machado et al., 2013). The opportunity to translate this intervention to patient care is significant given the well-established safety of DBS in surgical group is currently conducting a phase I clinical trial to enhance motor rehabilitation in chronic, post-stroke patients (NCT02835443) (Wathen et al., 2018).

In the current study, we sought to evaluate whether, given that its widespread connections include modulation of frontal and prefrontal brain regions, the behavioral benefits of LCN DBS could be extended to promoting cognitive rehabilitation in a model of medial prefrontal cortex TBI. Our TBI model emulates the injury location (prefrontal cortex) and resultant cognitive symptoms (long-term memory and executive functions) observed in most human TBI patients. The LCN DBS electrode was implanted after the acute recovery phase to model further the likely delayed application of an invasive treatment approach in human TBI treatment.

The experiments, which included naïve, untreated animals with mPFC-CCI and treated animals with mPFC-CCI, were carried out over a period of 13-weeks (Figure 1). Cognitive function was compared across LCN DBS-treated and untreated cohorts. To elucidate the therapeutic mechanism of LCN DBS stimulation enhancing recovery from cognitive deficits, we performed Western Blot and immunohistochemistry (IHC) of the perilesional tissues quantifying BDNF, p75^NTR, and CaMKIIα, which are recognized as factors involved in synaptogenesis and cognitive improvement (Gonzalez et al., 2019; Li et al., 2013; Sonoyama et al., 2020).

Figure 1: Timeline of the study, verification of electrode placement and injury size and elimination of subjects.

A) Experimental timeline. B) The flow chart of elimination of subjects in this study. C) The tip location for electrodes of the twenty-four implanted animals included in the final analyses is co-registered to the brain atlas (Paxinos, 2007).

Materials and Methods

Animals

Thirty-three male Long Evans rats (3–4 month-old and 200–224g at study onset; Envigo, IN) were used in this study. Animals were individually housed in a plexiglass cage and maintained on a 12:12 light/dark cycle. Food and water were provided ad libitum, with the exception of two days of restricted food access prior to baited Y-maze (BYM) testing as described below. All experiments were conducted under a protocol approved by the Institutional Animal Care and Use Committee of the Cleveland Clinic (protocol number: 2019–2133).

Experimental design

The experimental design is provided in Fig 1A. Baseline performance on BYM and novel object recognition (NOR) task was recorded the week prior to mPFC-CCI induction (week 0). Five animals received no surgery and assigned into Naïve group (n = 5). All remaining 28 animals underwent mPFC-CCI and were allowed to recover spontaneously for a period of six weeks (Weeks 1 to 6). At the start of week 7, all 27 surviving animals underwent DBS macroelectrode implantation in the left cerebellar LCN of all animals. We chose to implant the electrode unilaterally to allow for side-to-side comparisons with an internal control for better characterization of treatment effects. After electrode implantation the animals were then allowed to recover for one week post-surgery (Week 7). During week 8, the BYM and NOR tests were repeated to establish pre-treatment, baseline performance levels. Performance on the BYM was used to pseudo-randomize (severity matching) the 26 animals that survived mPFC CCI across two experimental groups: STIM (CCI with stimulation; n = 13); Ctrl (CCI without stimulation; n = 13). At the start of week 9, LCN DBS was activated for animals in the STIM group and continued for a total of four weeks (Week 9–12). Four days after the onset of the stimulation protocol, all experimental animals underwent Barnes maze testing for 10 consecutive days (Weeks 9 and 10; the first and second weeks of stimulation phase [SP] labeled as SP1 and SP2 respectively). The Barnes maze probe test was subsequently performed at the end of Weeks 11 and 12 (SP3 and SP4, respectively). Finally, during week 12 (SP4), the BYM and NOR tests were repeated. Animals were euthanized after Week 12 (SP4) and post-mortem tissue was collected for further analysis.

Induction of mPFC-CCI

The TBI lesion was induced by applying a CCI across the mPFC using coordinates previously described (Cope et al., 2012; Vonder Haar et al., 2014). Briefly, animals were anesthetized with ketamine (50mg/kg, i.p.) and dexmedetomidine (0.2mg/kg, i.p.) and placed in a stereotaxic frame. Following a midline incision, a 6.0 mm circular craniotomy centered over the mPFC (anterioposterial, AP: +3.00mm, ML: 0.0mm; relative to bregma) was created. The impactor (5.0 mm in diameter) of the CCI system (Impact One, Leica, IL) was placed on the surface of the mPFC without depressing brain tissue. The parameters for impact were pre-set as: impact speed: 2.25m/s, depth advanced 2.5mm and dwell time 0.5 sec. Afterward, the craniotomy was sealed with sterile cellulose paper fixed with VetBond followed by approximation and suturing the scalp wound margins. Anesthesia was reversed using atipemezole (1mg/kg, i.p.). Post-operative care included analgesia (buprenorphine: 0.05 mg/kg, s.c.; 2 days post-surgery) and antibiotic administration (15mg/kg cephazolin, s.c.) as well as regular monitoring for changes in weight or signs of pain and distress.

Electrode implantation

Six weeks after CCI induction, animals in the Ctrl and STIM groups underwent a second surgical procedure for placement of the LCN DBS macroelectrode (MS306, Plastics One, Inc., VA, USA). Briefly, animals were anesthetized as described for CCI induction and secured in the stereotaxic frame. Following a midline incision, a bur hole was created and the macroelectrode was inserted into the left LCN (AP: −2.00 mm; ML: +3.6 mm; and DV: +3.7 mm; relative to the interaural point). The interaural system was used as reference as bregma was typically damaged during the CCI craniotomy (De Vloo and Nuttin, 2019; Paxinos et al., 1985). Four screws were placed across the exposed calvaria to serve as anchors and dental acrylic was applied to secure the electrode. Anesthesia was reversed using atipemezole (1mg/kg, i.p.). Post-operative care included prophylactic analgesia (buprenorphine: 0.05 mg/kg, s.c.; 2 days post-surgery) and antibiotic administration (15mg/kg cephazolin, s.c.) as well as regular monitoring for changes in weight or signs of pain and distress.

Electrical stimulation in the LCN

At the end of Week 8 and immediately prior to the start of LCN DBS, the impedance of the implanted electrode was evaluated to ensure implant integrity. Animals with impedance values greater than 50kΩ were removed from the study. Impedance value across the two implanted experimental cohorts ranged from 16 to 42kΩ. Thereafter, the stimulation-induced side-effect threshold was established through visual observation by an experienced investigator as described previously (Chan et al., 2018b; Cooperrider et al., 2014). Briefly, 30Hz stimulation was delivered in the form of charge-balanced, square wave pulses (pulse width: 400 μsec) using a constant-current stimulus isolation unit (SIU; Model SIU-102, Warner Instruments, Hamden, CT) controlled by a stimulator (Grass S88, West Warwick, RI). Pulse amplitude was individually-set for each animal at 80% of the value that elicited a reproducible contralateral or axial motor response (threshold range: 72–144 μA). During the stimulation phase (weeks 9–12 [SP1–SP4]) animals in the STIM group received DBS for 12 h daily during the active phase of the dark/light cycle. Ctrl animals were similarly tethered to the commutator but no stimulation was delivered. The integrity of the electrodes and cables was confirmed regularly through periodic impedance checks. Animals in the naïve group were housed in identical cages, but not tethered.

Cognitive function assessments

The Barnes maze is used to assess long-term spatial memory (Maegele et al., 2015; McAteer et al., 2016), with subsequent probe tests administered to evaluate retention. The task involved 10 days of serial training/assessment followed by two additional probe tests administered one (PT1) and two (PT2) weeks after the end of training. Briefly, the animal was placed on a circular platform (1.2 meter in diameter) with 20 holes (10 cm in diameter) evenly-spaced around the perimeter, one of which passed through to an escape box secured underneath the platform. During an initial training session, the animal was placed on the platform and guided to the escape box where, upon entry, it was allowed to habituate for a period of two minutes. Thereafter, the test was repeated such that the animal was allowed up to five minutes to locate and enter the escape box independently. If it failed to do so by the end of five minutes, the operator guided the animal into it. The latency for successfully box entry is recorded. To evaluate memory retention, additional probe tests were performed seven (PT1: the end of SP3) and 14 (PT2: the end of SP4) days after the last day of the 10-day testing period. All testing was performed in a dedicated room under white ambient lighting. All sessions were video-recorded. The navigation pattern of each animals in each session was scored as “random” (defined as crossing the maze center area three times or more while checking different holes without a discernible pattern), “Serial” (defined as navigating along the maze perimeter in a clockwise or counter-clockwise direction until the escape box is located) andefined as locating and entering the escape box after checking three or fewer non-target holes) by an operator blinded to experimental cohort (Rosenfeld and Ferguson, 2014).

BYM is used to assess the long-term spatial memory and was performed over a period of two days using the modifications described by Yang (Yang et al., 2014). As shown in Figure 1A, it was repeated three times across the experimental timeline: Pre-CCI, post-CCI (pre-treatment) and during week four of stimulation (SP4). Briefly, the animal is food restricted (11.5–12.5g rat diet per day was provided) for two days prior to testing. Thereafter, on day 1 of testing, it is placed in one of the three arms of the Y-maze (Start arm; dimension of all three arms: Length: 50cm; Width: 11cm; Height: 30cm) and allowed to freely navigate the Start arm (height from floor: 70cm). After 20 seconds, a door at the end of the Start arm is opened by the experimenter, allowing the animal free access to either of the two remaining arms. Once the animal enters another arm, the door for that arm is closed. This process is repeated a total of ten times per animal in order to establish a preferred arm choice for each. Thereafter, a sugar pellet reward (45mg) is placed in the arm opposite to the preferred arm. This design is intended to minimize the possibility of the animal entering the rewarded arm as a result of natural preference. During actual testing, upon entering one of the two arms, the door was closed by experimenter and the animal allowed to explore for the reward for 30 seconds. Afterward, the animal was guided back to the Start arm and stationed there for 20 seconds while both arms are cleaned with 10% ethanol. Tests were repeated in the same manner until the animal achieved an index success rate marked by five consecutive entries into the rewarded arm. On average, animals required approximately 30 trials to establish success. On day two, each animal was re-tested, with the pattern of arm entry recorded across 20 consecutive trial administrations. The successful entry rate was defined as:

$$\frac{[Rewardedarmentry]}{20}*100\%$$

The NOR task is a metric of recognition memory (Thanapreedawat et al., 2013) that was applied at the same intervals as described for the BYM (Figure 1A). It consists of two different sessions: habituation and testing, which are administered with a 30 min interval (Dong et al., 2018; Pezze et al., 2015). During the habituation session, animals are placed in a chamber (dimension: Length: 55cm; Width: 40cm; Height: 27cm) with two identical objects and allowed to explore the objects for four minutes. During the testing session, animals are placed in the same chamber with one of the familiar objects still present but the other replaced with a novel object. The animal is again allowed four minutes to explore the environment. Exploration is video-recorded and the time spent exploring each object is quantified off-line by an operator blind to experimental cohort. The discrimination index, indicating the preference to the novel object, is defined as:

$$\frac{Timefornovelobject}{Timefornovelobject+Timeforoldobject}*100\%\text{.}$$

Elimination of experimental animals and grouping

The study started with 33 animals, four of which were withdrawn over the course of the study (Fig 1B). Two animals died prior to randomization, with one lost during CCI induction and the other during LCN electrode implantation. As a result, thirty-one animals were randomized across three different experimental groups: Naïve (n=5), Ctrl (n=13) and STIM (n=13). Two additional animals were withdrawn post-randomization from the STIM group due to poor electrode placement, defined as the tip of electrode track located outside (>50μm) of the LCN, as determined through post-mortem examination of the histological samples (Fig. 1C). The final N for the STIM cohort was eleven. One animal from the Ctrl group did not experience a post-CCI deficit in NOR performance (less than 5% decrease relative to the pre-CCI baseline performance) and was removed from that analysis only. Sample size was determined based on pilot data from the Novel Object Recognition task, wherein it was determined that ten animals per group (STIM and Ctrl) were required to detect a post-stimulation group difference of at least 15 percentage points with 80% power.

Euthanasia and tissue processing

Upon reaching the experimental endpoint, animals were randomly assigned to two different protocols of perfusion for different post-mortem analyses, namely immunohistochemistry (IHC) and Western blot. All animals were first administered an overdose of sodium pentobarbital (100 mg/kg, i.p.), with those assigned to IHC (13 animals: 7 Ctrl and 6 STIM) undergoing transcardial perfusion using cold phosphate buffered saline (PBS) followed by 4% PFA (EMS, PA) and those assigned to Western blot (16 animals: 5 Naïve, 6 Ctrl and 5 STIM) perfused with only PBS. After harvesting the tissue sample for Western blot, the brain blocks of all animals with PBS-only perfusion, including the Naïve animals, were further fixed with 4% PFA. Tissue from all Naïve animals was analyzed in both IHC and Western blot.

For Western Blot, tissue samples from contra- and ipsilateral (relative to the implanted LCN electrode) perilesional frontal cortex were harvested, snap-frozen in liquid nitrogen, and stored in −80°C freezer.

PFA-perfused brain blocks were preserved in 4% PFA for an extra 24 hours followed by cryoprotection in 30% sucrose solution/PBS. 30μm free-floating sections (a postmortem cut was created at the left ventral part of the brain blocks to indicate the hemisphere ipsilateral to the implanted LCN electrode) in the region of the perilesional prefrontal cortex were collected and stored in PBS with sodium azide until the histological analyses were performed. Cerebellar sections (30μm) were cryostated and collected onto gelatin-coated slides for further histological and IHC analyses.

Nissl’s staining

Electrode placement in the cerebellar sections and lesion in the cortical sections were visualized using Nissl-stained sections. Briefly, sections were incubated in 0.4% cresyl violet (CX2065–1, EMD Chemicals, NJ) in 0.4M acetate buffer for 20 min followed by serial dehydration with ascending concentration of ethanol. Each section was then mounted with DPX (06522, Sigma, MO) and cover-slipped.

IHC

In order to investigate the cellular and molecular mechanism of the LCN DBS-mediated cognitive recovery, IHC was performed to identify those cells expressing CaMKIIα, BDNF and p75^NTR, which are with important role in the cognitive-associated synaptogenesis (Gonzalez et al., 2019; Li et al., 2013; Sonoyama et al., 2020). Briefly, to ensure antigens were not hindered due to over-fixation, antigens were retrieved by treating free-floating sections with hot (90–100 °C) citrate buffer (10mM sodium citrate and 0.1% Tween-20, pH 6.0) for 10 min. Afterward, endogenous peroxidase activities were quenched by treatment of 1% H₂O₂/PBS for 10 min. Following blockage of the non-specific binding sites by 5% normal goat serum (Millipore, CA, USA) in 0.1% Triton X-100 containing PBS (PBST), sections were incubated with rabbit anti-BDNF (1:250; Cat #: ab108319, Abcam, CA; RRID: AB_10862052), p75 neurotrophin receptor (p75^NTR, 1:500; Cat #: 07–476, Millipore, MA; RRID: AB_310649) and mouse anti-CaMKIIα (1:500; Cat #: 05–532, Millipore, MA; RRID: AB_309787) for 48 hours. Afterward, sections were incubated with biotinylated anti-rabbit (1:500; Cat #: BA-1000, Vector laboratories, CA; RRID: AB_2313606) or mouse IgG (1:500; Cat #: BP-9200, Vector laboratories, CA; RRID: AB_2827937) for two hours at room temperature. Immunosignal was then magnified by treatment of Avidin-biotin complex (ABC, 1:200; Vector Laboratories, CA) for two hours at room temperature. IHC signal was visualized by using DAB kit (Cat #: SK-4100, Vector Laboratories, CA). Stained sections were mounted on gelatin-coated slides followed by dehydration and cover-slipped for further image analysis. Images were captured under 10X and 20X magnifications for quantification and illustration, respectively.

Immunofluorescence

To determine the presence and location of molecular changes consistent with the cognitive improvement we observed, we performed immunofluorescent staining to visualize the co-localization of pre-synaptic (synapsin I) and post-synaptic (PSD-95) markers at the perilesional cortex (Brito et al., 2014) with an emphasis on the side contralateral to the LCN DBS electrode. For this double immunofluorescent staining, free-floating cortical sections were incubated in a mixture of mouse anti-PSD-95 (1:250, Cat #: ab2723, Abcam, CA; RRID: AB_303248) and rabbit anti-synapsin I (1:500, Cat #: AB1543, Abcam, CA; RRID: AB_90757) antibodies for 48 hours. Thereafter, Alexa FluorTM 488-conjugated anti-mouse (1:500, Cat #: A28175, Thermofisher, MA; RRID: AB_2536161) and biotinylated anti-rabbit IgG (1:500, Cat #: BA-1000, Vector laboratories, CA; RRID: AB_2313606) were incubated for 2 hours, followed by 2-hour incubation of 1:500 Dylight^® 633-conjugated streptavidin (21844, ThermoFisher, MA). Fluorescent stained sections were mounted with anti-fading agents (H-10000, Vector Laboratories, CA, USA) for confocal microscopic analysis (Leica SP5, IL, USA).

Western Blot

Western Blot was used to evaluate the effect of LCN DBS on the expression of certain proteins, namely BDNF, p75^NTR, CaMKIIα, PSD-95 and synapsin I in an effort to elucidate the cellular and molecular mechanism(s) of LCN DBS-mediated neuroplasticity along the CTC pathway. Briefly, protein was extracted using RIPA lysis buffer (Cat #: 89900, Thermofisher, MA) with protease inhibitors (Cat #: A32955, Thermoscientifics, IL) and the concentration of total protein was measured using BCA assay (Cat #: 23227, Thermoscientifics, IL). 40μg of protein from each sample was separated in an SDS-protein acrylamide gel electrophoresis followed by electro-transference onto a nitrocellulose membrane. After blocking the non-specific binding sites using 5% non-fat milk in TBST (Cat #: sc-24953, Santa Cruz, CS), rabbit anti-BDNF (1:2000, Cat #: ab108319, Abcam, CA; RRID: AB_10862052), p75^NTR, (1:1000, Cat #: 07–476, Millipore, MA; RRID: AB_310649), synapsin I (1:500, Cat #: AB1543, Millipore, MA; RRID: AB_90757) and GAPDH (1:2000, Cat #: ab9485, Abcam, CA; RRID: AB_307275), and mouse anti-CaMKIIα (1:100, Cat #: 05–532, Millipore, MA; RRID: AB_309787) and PSD-95 (1:500, Cat #: ab2723, Abcam, CA; RRID: AB_303248) antibodies were applied and incubated overnight at 4°C. The next day, horseradish peroxidase-conjugated secondary anti-rabbit (for BDNF, p75^NTR, synapsin I and GAPDH; 1:2000, Cat #: sc-2357, Santa Cruz Biotech, CA; RRID: AB_628497) and anti-mouse (for CaMKIIα and PSD-95; 1:2000, sc-516102, Santa Cruz Biotech, CA; RRID: AB_2687626) IgGs, were applied and incubated for two hours. The immunosignal was then developed by reacting with ECL (1705061, Bio-Rad, CA) and visualized with a GelDoc machine. Analyses were performed by comparing the intensity ratio of each marker relative to the GAPDH signal.

Image Analysis

To avoid quantifying any nonspecifically-stained content inside the CCI core, we visually defined the perilesional area as that between the border of lesion core and 0.05mm from the border. Quantification was performed over a fixed square (0.09×0.09mm²) in this defined perilesional area. The quantification of CaMKIIα, BDNF and p75^NTR positive cells in ipsilateral and contralateral perilesional area were sampled every ten coronal sections (~300μm) from cerebral cortical tissue between +5.64 and −0.72 mm anterior/posterior relative to bregma. This yielded an average of twenty sections for each animal. Cell counts were converted into density according to the area analyzed.

The double fluorescent IHC of contralateral perilesional cortex was examined using confocal microscopy and analyzed with a modified method previously reported (Brito et al., 2014; Castejón et al., 2004). In order to confirm the co-localization of PSD-95 and synapsin I, images at high magnification (63x) were utilized. The focal plane for all magnifications was approximately 0.2μm in order to reduce confounds from adjacent cells. Quantification in the contralateral perilesional cortex was performed on every tenth section (~300μm), yielding an average of twenty sections for each animal. The area of signal in different channels, namely green for PSD-95, red for synapsin I and yellow for the overlay of both channels, was measured using Image J. Experimenters were blinded to sample group assignment in order to prevent bias in the quantifying process.

Data analyses and statistics

Cognitive analyses:

Independent samples t-tests were used to confirm that the two CCI groups did not differ significantly on cognitive tasks prior to treatment onset. Note that some figures include Naïve group performance as a reference point, however statistical analyses were applied only on data from the two CCI groups.

Box and whisker plots stratified by group and time point were created to visualize performance on the BYM and NOR task. BYM and NOR task performance was evaluated using a random intercepts mixed effects model for each. A mixed effects model is a flexible extension of linear regression, which allows for repeated measurements from the same animal and does not require equality of variances between groups. Multivariate normality for each outcome was confirmed visually using Chi-squared quantile by squared Mahalanobis Distance plots. Each mixed effects model evaluates change in performance on the respective cognitive task from post-CCI to SP4 of stimulation using an interaction term between stimulation group and time. A significant result for the coefficient of the interaction indicates that the two treatment groups differed in the degree of performance change over the course of treatment. Each model assumed an unstructured covariance matrix and included a clustering term to account for the two cohorts of animals.

To visualize Barnes Maze performance over time, data were stratified by group and time point and box and whisker plots were created. Barnes Maze results were evaluated multi-dimensionally. First, simple change in latency from Post-CCI to treatment end was compared between CCI groups using a mixed-effects model as described for BYM and NOR tasks, with time treated as categorical. Next, because the human post-TBI cognitive profile often presents as inconsistent performance, rather than simply poor performance (Adams et al., 1985; Hetherington et al., 1996), week-to-week variability within and between groups was explored descriptively with reference to the Naïve group’s performance using Wilcoxon rank-sum tests. All behavioral analyses were two-sided and carried out using SAS Studio v 3.7 with alpha set to 0.05.

Pearsons’s correlation analysis was performed to determine the relationship between the cellular markers changes and the Barnes maze performance as in changes in the latency, while Spearman’s correlation analysis was performed to determine the relationship between the confocal signal of synapse formation and the Barnes maze performance.

Other statistical analyses:

For Western Blot, the signal of each band was normalized according to the loading control, presented as relative intensity (%) ±S.E.M.. For IHC, the data were expressed as the cell density in the analyzed area (/μm²)±S.E.M. For dual channel immunofluorescence, the data were expressed as percentage of area covered by either channel. The statistical comparison was performed in each of the two areas (contralateral and ipsilateral). The Western Blot, IHC and dual channel immunofluorescence data were analyzed using one-way ANOVA with Bonferroni post-test was performed to examine differences between the contralateral and ipsilateral sites within the STIM group. The Kolmogorov-Smirnov test and Bartlett’s test were used to verify normality and homoscedasticity, respectively. Tests were two-sided with significance thresholds at p<0.05, p<0.01 or p<0.0001 as indicated in each figure. All cellular and molecular analyses were carried out using GraphPad Prism (GraphPad, CA, RRID: SCR_002798).

Results

LCN DBS improved mPFC-CCI-induced long-term memory deficits on Barnes maze.

At the beginning of the treatment phase, both Ctrl and STIM had median escape latencies of 300 seconds (Day 1: STIM interquartile range, IQR, 104–300; Ctrl IQR 194–300; p =0.654; Fig. 2A). Over the treatment course (Trials 1 − 10 [T1-T10]) STIM animals improved to a median escape latency of 39 seconds (IQR 21–73) while Ctrl animals maintained a median latency of 300 by T10 (IQR 191–300; p = 0.022, mixed effects model interaction between group and time).

Figure 2: Recovery effect of LCN DBS against mPFC-CCI induced elevated variability of long-term cognitive performance in Barnes maze.

A) Box and whisker plots summarizing the latency to locating the escape hole/box is shown across each of the 10 sequential test days as a function of experimental group. In contrast to naïve animals, mPFC-CCI animals showed a much longer latency and slower learning curve in task performance with serial exposure. Notably, although still abnormal in relation to the naïve cohort, treated animals showed a faster learning curve with less variability in task performance beginning on day five of testing (random intercepts mixed effects model). B) Latency to locating the escape box by treatment group during each of the two subsequent probe tests (PT1: probe test day 1, and PT2: probe test day 2). During probe test, animals needed to learn the new location of escape box at PT1 and relocated it at PT2. STIM animals experienced a significantly shorter latency to locate the new spot than the Ctrl animals at PT1. However, the difference observed between the two groups at PT2 was not significant (Wilcoxon rank-sum tests).

Additionally, STIM animals demonstrated markedly reduced performance variability beginning as early as day 5 of testing, whereas variability in Ctrl remained large. STIM scores decreased from an IQR of 18–300 on Day 4 to 45–125 on Day 5 and further to 26–67 on Day 7, then remained stable. In contrast, IQR of Ctrl were between 45–300 and 65–300 from Day 4 to Day 9, only decreasing marginally on Day 10 (Fig. 2A).

At probe test 1 (PT1), STIM demonstrated a significantly shorter median latency of 54 seconds (IQR 33–220) relative to the Ctrl of 263 (IQR 60–300) (p = 0.003; Fig. 2B). During probe test 2 (PT2), group performances were not significantly different because of spontaneous improvement in the Ctrl group. STIM achieved a median latency of 32.5 (IQR 16–55) and Ctrl displayed a median latency of 67 (IQR 20–93) (p = 0.337; Fig. 2B). The positive effect on the performance in the initial probe test suggested that LCN DBS accelerated the mPFC-mediated learning process recovery relative to controls.

LCN DBS-treated animals demonstrated improved strategy use following CCI.

Early in learning, naïve animals used a highly efficient ‘Serial’ strategy of checking sequential outlets of the Barnes Maze until locating the escape box and, once learned, they converted to a direct strategy (running directly to the escape location). In contrast, CCI induced a strategy application deficit. Sixty-nine percent of Ctrl animals and 45% of STIM animals initially engaged in a “Random” strategy early in learning. At day 1, 31% of animals in Ctrl and 55% of animals in STIM applied “Serial” search strategy (p = 0.401; Fig. 3). By day 10, 73% of STIM animals applied the “Direct” strategy, with the remaining 27% using the “Serial” strategy. In contrast, only 16% of Ctrl animals used the “Direct” approach, 8% used “Serials” and 76% continued to “Random” (p < .001; Fig. 3). This supports a potential role for LCN DBS in enhancing strategy use, which may have important translational consequences given analogous strategy application deficits in human frontal lobe damage following TBI.

Figure 3: Effect of LCN DBS on improvement of the strategy use in Barnes maze following mPFC-CCI.

The strategy of Barnes maze exploration was visually classified for each animal and classified as either Random, Serial or Direct. Relative to naïve animals, there was an initially higher percentage of animals whose exploration pattern was Random in nature (i.e., failure to apply consistent strategy). The serial pattern initially observed in the Naïve animals was quickly replaced with a Direct approach within the first several days of testing. A similar strategy replacement pattern was observed in STIM animals, though not as wholly as that observed for Naïve animals, while animals in the Ctrl group showed a limited shift towards more strategic approaches to locating the escape hole across the 10-day period.

LCN DBS improved mPFC-CCI-induced long-term memory deficits on BYM and NOR.

All three groups demonstrated a successful entry rate of between 80% and 85% on BYM during the pre-CCI baseline assessment, with the Naïve group maintaining this performance level over the 13-week experiment. Following mPFC-CCI, both Ctrl and STIM showed reduced performance (Ctrl: 46.9% and STIM: 45.9%). At SP4, performance was significantly higher in the STIM (54.1%) relative to the Ctrl (44.6%) (p = 0.048, interaction between group and time, post-CCI to post-stimulation, Fig 4A). To further assess the spatial memory in BYM, the alternation rate at the beginning (First-5) and the end (Last-5) of the tasks across the experimental groups are compared. Ctrl and STIM performed similarly pre- and post-CCI. Pre-CCI, the difference in alternation rates between First-5 and Last-5 trials was 0.82 for the STIM and 0.54 for Ctrl animals (p = 0.263; Fig. 4B). Post-CCI, the difference in alternation rates between the First-5 and Last-5 trials was 0.09 for the STIM and 0.15 for the Ctrl group (p = 0.491; Fig. 4B), which is lower than that in pre-CCI. Such deficit indicates a compromised spatial memory in both Ctrl and STIM groups. Following treatment, however, the difference in alternation rates for the First-5 and Last-5 trials was 1.45 for the STIM group and only 0.08 for the Ctrl group (p < 0.001; Fig. 4B). Such findings suggest that the LCN DBS improves the performance in this BYM by improving the spatial memory during the navigation in the maze.

Figure 4: Effect of LCN DBS on mPFC-CCI-induced cognitive deficits in BYM and NOR task.

A) Animals were trained to enter the arm containing a food reward. mPFC-CCI induced a reduction of performance which was modestly but significantly reversed by the LCN DBS at SP4 (p<0.05; SP4, stimulation phase 4, is the last week of the four-week DBS stimulation phase). B) The alternation of arm entry in BYM at the beginning (First-5 trials) versus the end (Last-5 trials) of task in all three experimental groups were compared. In Naïve group, the number of alternation at the Last-5 was reduced when compared to the First-5. Following mPFC-CCI, the alternation at the Last-5 did not reduce when compared to the First-5 in both Ctrl and STIM. However, in STIM, the alternation at SP4 was reduced following the LCN DBS treatment (p < 0.001). C) The percentage of time spent exploring the novel object decreased post-CCI (pre-treatment) for animals in both the Ctrl and STIM cohort suggesting that memory formation of the familiar object was impaired. After four weeks of treatment, the STIM animals showed marked improvement as signified by greater time spent exploring the novel object (p<0.001; random intercepts mixed effects model).

On the NOR, Naïve showed consistent performance across the study (73.1%, 70.0%, and 71.7% of relative time spent with the novel object at week 0, 8 and 12, respectively; Figure 4C). Both Ctrl and STIM showed a comparable decrease performance following mPFC-CCI at 40.5% and 39.4%, respectively. At SP4, however, a significant difference was observed between Ctrl (42.2%) and STIM (61.2%) groups (p < 0.001, interaction between time and group from post-CCI to post-stimulation, Fig 4C).

LCN DBS-treated animals showed increased expression of CaMKIIα, BDNF and p75^NTR asymmetrically across perilesional cortex following mPFC-CCI.

In the current model, CCI was induced at midline, resulting in a bilateral mPFC lesion while LCN DBS was delivered unilaterally. At the end of the treatment, Ctrl demonstrated a reduction in the density of CaMKIIα+ (1-way ANOVA, df = 2; Contra: F = 46.28, p < 0.0001; Ipsil: F = 96.5, p < 0.0001;), BDNF+ (1-way ANOVA, df = 2; Contra: F = 27.86, p < 0.0001; Ipsil: F = 54.99, p < 0.0001) and p75^NTR+ (1-way ANOVA, df = 2; Contra: F = 36.25, p < 0.0001; Ipsil: F = 30.54, p < 0.0001) cells in both contralateral and ipsilateral (relative to the implanted LCN electrode) perilesional cortex as compared to Naïve animals, suggesting that mPFC-CCI suppressed calcium/BDNF pathway-mediated neuroplasticity at the mPFC (Fig 5A–C). LCN DBS was associated a reversal in this pattern, with STIM displaying levels of all three markers (Bonferroni post hoc; CaMKIIα+: Contra: p < 0.0001, Ipsil: p = 0.0127; BDNF+: Contra: p < 0.0001, Ipsil: p = 0.0236; p75^NTR: Contra: p = 0.0171; Ipsil: p = 0.0206) significantly higher than those in Ctrl (Fig 5A–C).

Figure 5:

Effect of LCN DBS against the mPFC-CCI induced reduction of density of CaMKIIα, BDNF and p75^NTR positive cells in the perilesional cortices. There was a significant reduction of A) CaMKII+, B) BDNF+ and C) p75^NTR+ cells following mPFC-CCI at the ipsilateral (Ipsil) and contralateral (Contra) perilesional cortices, which was reversed by the LCN DBS. Furthermore, the treatment-related effect on the density of cells expressing target proteins in Contra of STIM is significantly greater than that in Ipsil. Data are expressed as density of cells (/μm²)±S.E.M. and were analyzed using a one-way ANOVA followed by Bonferroni post hoc tests. * p < 0.05, ** p < 0.01 and *** p < 0.001 when compared to Ctrl and δ p < 0.05 when compared to Contra of STIM. Scale bars represent 200μm.

In comparing Ctrl to Naïve controls, there was a consistent reduction of CaMKIIα+, BDNF+ and p75^NTR+ cell density across both hemispheres. However, the unilateral LCN DBS significantly induced greater expression of these three markers (t test; CaMKIIα+ p = 0.0133, BDNF+ p = 0.0261 and p75^NTR+ p = 0.0387; Fig 5A–C) in contralateral relative to ipsilateral cortex.

Consistent with the cellular results, expression levels of CaMKII, BDNF and p75^NTR at both contralateral (1-way ANOVA, df = 2; CaMKIIα: F = 38.02, p < 0.0001; BDNF: F = 14.99, p < 0.0001; p75^NTR : F = 65.80, p < 0.0001; Fig 6A) and ipsilateral (1-way ANOVA, df = 2; CaMKIIα: F = 58.85, p < 0.0001; BDNF: F = 18.29, p = 0.0382; p75^NTR: F = 28.18, p = 0.0837; Fig 6B) perilesional cortex were reduced following mPFC-CCI, with Ctrl showing reduced expression of all markers, except p75^NTR in ipsilateral side, relative to Naïve (Fig 6A–B). In contrast, STIM showed higher levels of expression for each markers at the contralateral perilesional cortex evaluated (Bonferroni post hoc; all markers p < 0.0001; Fig 6A). STIM also showed higher expression level of CaMKIIα (Bonferroni post hoc; p = 0.0211), but not BDNF (Bonferroni post hoc; p = 0.1042) and p75^NTR (Bonferroni post hoc; p = 0.0981), in the ipsilateral perilesional cortex (Fig 6B).

Figure 6:

Effect of LCN DBS on mPFC-induced reduction of protein expression in both perilesional cortices. As determined through Western Blot, there was a significant reduction of CaMKIIα, BDNF, and p75^NTR following mPFC-CCI in A) contralateral (Contra) and B) Ipsilateral (Ipsil) perilesional cortices following mPFC-CCI, which was significantly reversed by LCN DBS. A similar pattern was observed in the expression of synapsin I and PSD-95 in A) and B). Data was expressed as the relative intensity to the loading control (GADPH) ±S.E.M.. Statistics was analyzed by 1-way ANOVA followed by Bonferroni post hoc tests for each protein target. * p < 0.05, ** p < 0.01 and *** p < 0.001 when compared to Ctrl.

LCN DBS promotes synaptic formations.

The mPFC-CCI reduced the expression of PSD-95 and synapsin I across contralateral (1-way ANOVA, df = 2; PSD-95: F = 16.61, p < 0.0001 and synapsin I: F = 25.97, p = 0.00784, Fig. 6A) and ipsilateral (1-way ANOVA, df = 2; PSD-95: F = 12.54, p < 0.0001 and synapsin I: F = 15.84, p < 0.0001, Fig. 6B) perilesional cortex. This reduction was reversed only the contralateral perilesional cortex in STIM (Bonferroni post hoc; Contralateral PSD-95 p < 0.0001 and synapsin I p = 0.0233; Ipsilateral, PSD-95 p = 0.0944; synapsin I p = 0.0817; Fig 6A–B). The increase of the expression of these pre- and post-synaptic markers suggests the possibility of increased synaptic formation. This was further examined with dual fluorescent IHC in order to evaluate for the superimposition of both PSD-95 and synapsin I signal in the contralateral perilesional area. Superimposition of the two markers was found to be reduced following the mPFC-CCI (1-way ANOVA, df = 2; F = 18.75, p < 0.0001), with STIM showing significantly higher superimposition than Ctrl (Bonferroni post hoc; p = 0.0383; Fig 7).

Figure 7:

Effect of LCN DBS on synaptogenesis following mPFC-CCI. There was significant reduction in the degree of overlay of synapsin I and PSD-95 in contralateral perilesional cortex following mPFC-CCI, which was reversed by the LCN DBS, indicating a possible role for synaptogenesis induced by the LCN DBS in mediating therapeutic efficacy. Data are expressed as % area covered by signal of PSD-95 (green), synapsin I (red) and overlay (yellow) ±S.E.M.. The data were analyzed using a one-way ANOVA followed by Bonferroni post hoc tests for each channel. * p < 0.05, ** p < 0.01 and *** p < 0.001 when compared to Ctrl. Scale bars represent 20μm.

Correlation between performance in Barnes Maze and neuroplasticity markers

In the analysis of the correlation between the cognitive performance and the protein markers analyzed in IHC, the lower magnitude of improvement in Barnes maze performance (as the changes of latency of escaping the maze across the training phase) in Ctrl animals was strongly and significantly correlated with the lower density of cells expressing CaMKIIα (contralateral: r = −0.74, p = 0.004; ipsilateral: r = −0.83, p < 0.001; Fig 8A–B), BDNF (contralateral: r = −0.75, p = 0.003; ipsilateral: r = −0.88, p < 0.001, Fig 8C–D), and p75^NTR (contralateral: r = −0.74, p = 0.004; ipsilateral: r = −0.79, p = 0.001 Fig 8E–F) in both contralateral and ipsilateral perilesional cortex. However, such correlation was not detected in STIM animals (CaMKIIα contralateral: r = 0.15, p = 0.663; ipsilateral: r = −0.28, p = 0.403; BDNF: contralateral: r = −0.08, p = 0.816; ipsilateral: r = −0.31, p = 0.35; p75^NTR contralateral: r = −0.05, p = 0.885; ipsilateral: r = −0.37, p = 0.256; Fig 8). Similarly, the lower rate of improvement in Barnes maze performance in Ctrl animals is strongly and significantly correlated to the lower numbers of synapse formation (Ctrl, Spearman’s Rho = −0.8, p = 0.001; STIM −0.19, p = 0.574, Fig 8G). Such findings suggest that the cognitive deficit in Barnes maze task following mPFC-CCI is related to the compromised CaMKIIα/BDNF pathways and the downstream of synapse formation, which were reversed by the LCN DBS although there is no correlation in STIM group. Correlation coefficients of all comparison were summarized in the Table 1.

Figure 8:

Correlation between Barnes maze performance and expression of molecular markers. Analyses of correlation between the Barnes maze performance (in term of changes in latency) and density of A and B) CaMKIIα, C and D) BDNF, E and F) p75^NTR positive cells and G) the confocal signal of synapse formation were performed. The data distribution of all samples within the same group was highlighted by the 95% prediction ellipses. The less improvement in Barnes maze was highly correlated with the low levels of all markers in both contralateral (Contra) and ipsilateral (Ipsil) in Ctrl group. Such correlation was not detected in STIM group.

Table 1:

The correlation coefficients (r) of comparisons between the Barnes maze performance and the neuroplasticity markers.

		Ctrl		STIM
		r	p value	r	p value
CaMKIIα	Ipsil	−0.83	< 0.001**	−0.28	0.403
	Contra	−0.74	0.004*	0.15	0.663
BDNF	Ipsil	−0.88	< 0.001**	−0.31	0.35
	Contra	−0.75	0.003*	−0.08	0.816
p75NTR	Ipsil	−0.79	0.001*	−0.37	0.256
	Contra	−0.74	0.004*	−0.05	0.885
Synapse formation#	Contra	−0.8	0.001*	−0.19	0.574

Correlation related to CaMKIIα, BDNF and p75^NTR were analyzed by using Pearsons’ s correlation analysis while that to the synapse formation signal was analyzed by using Spearman’ s correlation analysis.

^#the coefficient is Spearman’s rho.

^*p<0.01 and

^**p<0.001.

Lesion at the cerebral cortex and hippocampi following mPFC-CCI.

There was no difference in average lesion volume between the Ctrl and STIM cohorts (t test; p = 0.6326; Fig 9A), suggesting that chronic LCN DBS, initiated eight weeks post-injury, did not impact post-mPFC-CCI degeneration. Furthermore, there was no structural damage detected in the anterior part of the hippocampi, anatomically closest to the lesioned prefrontal area, suggesting that the mPFC-CCI or the subsequent diffuse effect of lesion did not affect the hippocampal structure (Fig 9B). Such finding suggested that the observed cognitive deficits were not induced by the dysfunction of hippocampus following CCI.

Figure 9:

Lesion at the cerebral cortex and hippocampi. A) Representative examples of mPFC-CCI lesion in rat brain and histological demonstration of Nissl’s stained brain section for volume quantification. The comparison of the lesion volume of Ctrl and STIM groups, by t test, was not significant. Data was presented as lesion volume (mm³) ±S.E.M.. B) Absence of lesion in the Naïve, Ctrl and STIM groups.

Discussion

This study is the first to demonstrate that electrical stimulation of the DTC pathway promotes cognitive recovery in a rodent model of chronic, bilateral mPFC TBI. Treatment-related behavioral improvements were associated with cellular and molecular changes that support stimulation-enhanced functional connectivity across both perilesional cortical and thalamic brain regions as potential therapeutic mechanisms. Notably, cognitive enhancement in this bi-hemispheric cortical TBI model was observed in response to unilateral LCN stimulation and evidence of significant thalamocortical changes were observed not only contralateral, but also ipsilateral, to the targeted LCN. Overall, these findings support DBS of the ascending cerebellar pathways as a therapeutic approach to enhancing neurorehabilitation and, more specifically, provide novel evidence of its impact on cognitive recovery.

Cognitive improvement by LCN DBS

In humans, the frontotemporal cortices are highly susceptible to traumatic injury (Abdel-Dayem et al., 1998; List et al., 2015; Marek et al., 2018; Seamans et al., 2008), with involvement of prefrontal cortex leading to impairment of executive function, memory retrieval, strategic thinking, decision making and etc (Janowsky et al., 1989; Levin, 1990; Mani et al., 2017; McInnes et al., 2017; Rabinowitz and Levin, 2014; Salmond and Sahakian, 2005). Moreover, humans TBI produces a highly variable in cognitive performance such that a patient may perform a task completely normally on one occasion and then show significant impairment at the same task on another occasion (Adams et al., 1985; Hetherington et al., 1996), even years post-injury for moderate-to-severe victims (Sengupta, 2013). In the current study, we used a frontal, midline CCI model in an effort to induce bilateral damage across the rodent mPFC (functionally analogous to the dorsolateral PFC in humans) (Brito et al., 1982; Seamans et al., 2008; Yang et al., 2014).

Even after eight weeks of spontaneous recovery, the LCN DBS treatment was associated with significant improvement in performance on both Barnes maze and BYM. Treated animals demonstrated greater day-to-day reductions in escape latency with repeated task exposure on the Barnes maze. In probe tests (PT1), at which animals were required to locate the new escape hole, the greater performance in treated animals suggested that the learning process was improved by the LCN DBS. On BYM, treated animals showed a higher entry rate to the rewarded arm relative to untreated animals. Despite the hippocampal dependency of these task variables, the absence of gross hippocampal lesion in the current study and the difference between treated and untreated animals suggests that deficits may be the result of the compromised prefrontal cortex failing to provide strategic control over the learning and memory process (Dobbins et al., 2002; Miller and Cohen, 2001). Data from the Barnes maze supports this concept. Treated animals demonstrated a stepwise reduction in the intra-cohort variability in day-to-day performance in Barnes maze, suggesting that the mPFC-CCI mediated disorganized behavior was improved (Adams et al., 1985; Hetherington et al., 1996; Saxena, 2008). Furthermore, the application of an efficient search strategy, a primary function attributed to the mPFC, facilitates the spatial reference memory formation on the Barnes maze (Negrón-Oyarzo et al., 2018), while disruption of the prefrontal-hippocampal projections in rodent may result in a failure of applying strategy on the same task (Binder et al., 2019). In our data, both Ctrl and STIM animals initially showed poor strategy application, searching for escape in a Random fashion. Only the LCN DBS-treated animals adopted a more efficient search strategy, resulting in better learning.

Our findings support our hypothesis that mPFC-CCI mediated cognitive deficits are due to the disruption of the prefrontal hippocampal interactions. Deficits on both Barnes maze and BYM were restored by LCN DBS. Overall, the cognitive improvement in treated versus untreated animals supports LCN DBS as a potential translational emergent therapy to promote rehabilitation for individuals with chronic, long-term cognitive disabilities post-TBI.

Unlike those reports demonstrating that focal administration of cytotoxic compounds into prelimbic cortex induced no deficits in the object-based novelty (Hannesson et al., 2004a; Hannesson et al., 2004b), we observed a robust deficit in distinguishing novel objects following mPFC-CCI. Such finding is consistent to certain lines of evidence of the deficit in NOR detected in prefrontal TBI models (Darkazalli et al., 2016; Zhang et al., 2014). Such a discrepancy may be due to the CCI lesion involving multiple frontal areas that participate in functional regulation during object recognition processes. The recovery effect of the LCN DBS over the deficit in NOR in the current study is possibly due to the restoration of the overall mPFC functionalities.

BDNF-dependent synaptogenesis as an underlying mechanism of LCN DBS-mediated cognitive improvement.

The potential for LCN-targeted neuromodulation to influence cerebral cortical activity is well-supported by prior electrophysiologic (Baker et al., 2010; Cooperrider et al., 2014; Rogers et al., 2011), imaging (Sultan et al., 2012), and anatomical tracing (Dum and Strick, 2003) studies. Importantly, this neuromodulation could have widespread impact on cortical activity, including frontal and prefrontal cortical regions (Middleton and Strick, 2001; Rogers et al., 2011; Sultan et al., 2012). We hypothesized that LCN DBS-related improvements would arise through activation of its ascending, glutamatergic projections to thalamus, with a subsequent, upregulating effect on corticothalamic activity and interactions that included mPFC and surrounding prefrontal regions. Overall, our comparison of microstructural and molecular changes between treated and untreated animals supports this hypothesis, with post-mortem analyses showing proliferation of BDNF, p75^NTR and CaMKIIα expressing cells and protein marker expression in perilesional cortex that surpass levels observed in control animals. Collectively, these changes suggest that LCN DBS-mediated improvements may depend, in part, on activation of BDNF and related signaling pathways (Lisman et al., 2012; Righi et al., 2000; Su et al., 2017; Takei et al., 2004; Wurzelmann et al., 2017) and support a potential role for enhanced long-term-potentiation-like plasticity and synaptogenesis (Aguado et al., 2003; Sonoyama et al., 2020) in mediating therapeutic gains. The critical role of BDNF in synaptogenesis and memory reconsolidation has been well recognized (Gonzalez et al., 2019). Abnormality of BDNF pathway results in reduction of neurite growth and synaptogenesis, and impairment of short-term memory (Sonoyama et al., 2020), which is consistent to our findings of the correlation between the reducing expression of BDNF related markers/synapse formation and cognitive deficit. Further, the antidepressant-like effect of the activation of the BDNF signaling and related synaptogenesis in the prefrontal cortex (Sales et al., 2019) suggests that the LCN DBS-mediated cognitive improvement may also be associated with the activation of BDNF pathway and synaptogenesis at the perilesional area of prefrontal cortex following TBI. Although there is no correlation between the cognitive improvement and any of the BDNF, p75^NTR and CaMKIIα expression, it is possibly due to the ceiling effect as the expression levels of those three markers in the treated animals were reversed to naïve levels.

The asymmetric effect of LCN DBS on marker expression was largely anticipated, as the decussate nature of the ascending CTC pathways would bias its impact towards cortical and thalamic targets contralateral to the implanted LCN electrode. Notably, however, expression was also found to be increased in cortical and thalamic targets ipsilateral to the electrode (See supplementary data). Still, those ipsilateral changes were consistently smaller than those observed in contralateral targets. Such changes may reflect the effects on the minority of ascending LCN projections that do not decussate (Bostan et al., 2013; Dum and Strick, 2003; Iwata and Ugawa, 2005), activity mediated by cortico-cortical connections between the two hemispheres (López-Gil et al., 2012; Richieri et al., 2017), or some combination of the two.

Limitations.

Although this study utilized a more translational and clinically relevant TBI model with frontal damage, the tasks used in this study had a large hippocampal dependent component of memory. Other than uncover the role of hippocampal-prefrontal connection in the LCN DBS-mediated cognitive improvement, future studies using this TBI model will also investigate LCN-DBS treatment effects on behavioral paradigms that focus on executive function and decision making, which are commonly impaired in moderate-to-severe TBI patients. Similarly, future studies will also evaluate the anxiety to any contribution to outcomes as prior work has suggested that mPFC TBI increases anxiety as indexed, for example, using a plus-style maze (Cope et al., 2011) although this is somewhat controversial (Johnson et al., 2013). Overall, however, the effects against the psychiatric symptoms remains unclear. Furthermore, the heterogeneity of the severity in TBI survivors and their symptoms involving a series of sensory/motor, cognitive, and psychiatric deficits makes the standardization and optimization of this neuromodulation therapy difficult. Further researches of the therapeutic significance of LCN DBS in different types TBI models, severity and therapeutic paradigms are essential for a future clinical investigation.

As noted, there was an asymmetric effect of LCN DBS on the microstructural and cellular changes identified in the current work. Our choice for unilateral stimulation enabled us to better evaluate the effects of LCN, having other side as a control. While beneficial from a mechanistic examination standpoint, it may have limited the overall potential therapeutic benefits. It is unknown, albeit seemingly likely, that further therapeutic LCN DBS gains could be achieved using a bilateral targeting approach that seeks to upregulate activity across both ascending cerebellar output pathways. As such, future efforts should be geared towards such an approach.

Future work should also include electrophysiological studies to characterize more directly the modulatory effects of LCN DBS on prefrontal cortex as well as thalamocortical interactions for the purpose of improving our understanding of its potential therapeutic mechanisms as well as to facilitate further refinement of the approach using closed-loop control. Such studies may also provide insight to help guide non- or more minimally-invasive approaches to activating ascending the circuits modulated by ascending cerebellar output pathways. It is important to note, however, that the effects observed in the current study were derived from the continuous delivery of stimulation during the animals’ active period of light-cycle, which is difficult to achieve through non-invasive treatment approaches at this time. Certainly, the relative benefits of an invasive, DBS-based approach to enhancing rehabilitation will have to be weighed in relation to its overall risks, both surgical and in relation to any potential side-effects. Although not observed in the animal model, previous studies of deep cerebellar DBS side-effects have reported stimulation-induced dizziness, dysphagia and dysarthria in human (Horisawa et al., 2021). Overall, however, the side-effects were generally mild and could be eliminated through reprogramming of the DBS output.

Conclusion

This translational study directly supports LCN DBS as a promising approach to enhance chronic, post-TBI cognitive rehabilitation and both highlights and extends the potential for cerebellar neuromodulation as part of a comprehensive approach that includes conventional physical and cognitive rehabilitation for the treatment of deficits secondary to acquired brain injury (Chan et al., 2018b).

Abbreviations:

ANOVA

Analysis of variance

AP

anterioposterial

BCA

bicinchoninic acid assay

BDNF

Brain Derived Neurotrophic Factor

BYM

baited Y-maze

CCI

controlled cortical impact

CTC

cerebellothalamocortical

DAB

3,3’-diaminobenzidine

DBS

deep brain stimulation

DV

dorsoventral

ECL

Enhanced chemiluminescence

GAPDH

Glyceraldehyde-3-Phosphate Dehydrogenase

IHC

immunohistochemistry

IQR

interquartile range

LCN

lateral cerebellar nucleus

MD

mediodorsal

ML

Mediolateral

mPFC

medial prefrontal cortex

NOR

novel object recognition

p75^NTR

p75 neurotrophin receptor

PBS

phosphate-buffered saline

PFA

paraformaldehyde

PSD-95

Postsynaptic density protein-95

PT

probe test

RIPA

Radioimmunoprecipitation assay

SDS

Sodium Dodecyl Sulfate

SP

stimulation phase

TBI

traumatic brain injury

TBS

Tris-buffered saline

VL

ventrolateral