Systemic Inhibition of Soluble Tumor Necrosis Factor with XPro1595 Exacerbates a Post-Spinal Cord Injury Depressive Phenotype in Female Rats

Abstract

Spinal cord injury (SCI) is associated with a three-fold risk of major depressive disorder compared with the general population. Current antidepressant therapy is often not as effective in this patient population, suggesting the need for a more efficacious therapeutic target. The goal of this study was to elucidate the role of inflammatory cytokine tumor necrosis factor (TNF) in the dorsal raphe nucleus (DRN, the principle source of serotonin to the brain) in the development and possible treatment of depression after SCI. A depressive phenotype following moderate T9 contusion was identified in adult female rats using a battery of behavioral tests (forced swim test, sucrose preference test, novel object recognition test, open field locomotion, and social exploration). Data revealed two clusters of injured rats (58%) that exhibit increased immobility in the forced swim test, indicating depressive phenotype or a melancholic-depressive phenotype with concomitant decrease in sucrose preference. ElevatedTNF levels in the DRN of these two clusters correlated with increased immobility in the forced swim test.

We then tested the efficacy of soluble TNF inhibition with XPro1595 treatment to prevent the depressive phenotype after SCI. Subcutaneous (s.c.) delivery of XPro1595 caused an exacerbation of depressive phenotype, with all treated clusters exhibiting increased forced swim immobility compared with saline-treated non-depressed rats. Intracerebroventricular (i.c.v.) administration of the drug did not prevent or enhance the development of depression after injury. These results suggest a complex role for TNF-based neuroinflammation in SCI-induced depression that needs to be further explored, perhaps in conjunction with a broader targeting of additional post-SCI inflammatory cytokines.

Introduction

Major depression is the leading cause of disability worldwide according to the World Health Organization, with an annual incidence of 6.7% in the general population (World Health Organization, 2013; https://www.who.int). The frequently debilitating constellation of symptoms characterizing this disorder include anhedonia (a loss of interest in things once considered pleasurable); changes in sleep, weight, and appetite; memory loss; difficulty concentrating; social withdrawal; and feelings of guilt, worthlessness, and/or hopelessness (Diagnostic and Statistical Manual of Mental Disorders, fifth edition [DSM-V]). Major depressive disorder (MDD) classically has been attributed to an imbalance of the serotonin system, which includes neurons of the dorsal raphe nucleus (DRN) involved in modulation of many important affective features associated with depression such as attention, working memory, and emotional control (reviewed in Hensler¹).

In addition to motor, sensory, and autonomic dysfunction, patients with spinal cord injury (SCI) are at three times the risk for MDD compared with the general population.² Current antidepressant treatment with selective serotonin reuptake inhibitors (SSRIs) is largely ineffective in this population and, despite the increased prevalence, there are few clinical guidelines to treat post-SCI depression.² Comorbidity of depression and SCI has a profound impact on quality of life, even correlating with less functional improvement during rehabilitation.² Elucidating possible underlying causes of MDD after SCI would allow for a more targeted therapeutic approach in these patients.

SCI causes a robust and prolonged inflammatory response that extends well beyond the lesion epicenter.^3–5 Evidence of remote, chronic inflammation after SCI has been found in supraspinal brain regions^6,7 as well as in distal spinal cord more than 12 segments caudal to the lesion.⁸ Inflammation has also been implicated in MDD pathology, with frequently elevated levels of pro-inflammatory cytokines such as interleukin (IL)-6 and tumor necrosis factor (TNF) in the serum of patients with major depression.^9–11 TNF signaling elicits a broad and pleiotropic inflammatory response and its inhibition has been effective in treating depression refractory to treatment with SSRIs in a subset of patients demonstrating elevated baseline inflammation.^12–14 Direct adminstration of TNF into the lateral cerebral ventricle of rodents can elicit many behaviors that mimic clinical depression, supporting its potential role in the development of MDD and SCI-depression.¹⁵

TNF is present in two biologically active forms, soluble and transmembrane, each with a diverse signaling cascade (reviewed in McCoy and Tansey¹⁶). The soluble form of TNF has a higher affinity for TNF receptor subtype 1 (TNFR1) and is more highly associated with pro-inflammatory and apoptotic effects.^17–23 Transmembrane TNF has a higher affinity for TNFR2 and is associated with decreased inflammation and increased cell survival and axon remyelination.^{18,19,24–26} Additionally, TNFR2 is mostly located on immune cells, resulting in increased susceptibility to specific pathogens following global TNF inhibition.^18,19,27 XPro1595 is a dominant-negative inhibitor of TNF with affinity for only the soluble form, leaving the neuroprotective functions of transmembrane TNF intact. Systemic administration of XPro1595 has had therapeutic effects in multiple animal models of neuroinflammatory disease, including increased dopaminergic neuron sparing in 6-OHDA induced Parkinson's disease and increased axon preservation and myelination in experimental autoimmune encephalomyelitis (EAE).^28,29

Interestingly, Novrup and colleagues³⁰ demonstrated that only local administration of XPro1595 via intrathecal pump was sufficient to increase axonal sparing and locomotor function in a mouse model of SCI and intrathecal XPro1595 administration was shown recently to diminish maladaptive plasticity of the spinal sympathetic reflex circuit associated with post-SCI autonomic dysreflexia.³¹ It was hypothesized that the lack of efficacy with subcutaneous (s.c.) treatment in SCI may be due to the acute release of cytokines by immune cells in the central nervous system (CNS) within hours after injury and a systemic dosing regimen would not have provided an adequate acute level of drug in the spinal cord.³⁰ Given the insidious onset of post-SCI depression, however, we tested whether peripheral administration of Xpro1595 would be sufficient to prevent depressive phenotype in our model, circumventing the clinical difficulties associated with drug administration directly into the human CNS. One goal of this study was to assess whether an increase in brainstem levels of TNF correlated with post-SCI depression, followed by a study to determine if peripheral (s.c.) or central intracerebroventricular (i.c.v.) administration of XPro1595 would provide an efficacious treatment for the SCI-MDD experimental animal population.

A recent study from Luedtke and associates³² established a model of SCI-induced depression in male rats, with reliable depressive phenotype emerging at 4 weeks post-injury. Although men are statistically more likely to sustain an SCI, MDD is more common in women in the general population National Institute of Mental Health (NIMH; https://www.nimh.nih.gov). Our model of moderate thoracic contusion injury in female rats was designed to assess possible sex differences in expression of depressive phenotype using a behavioral battery of tests, because all previous reports of post-SCI depression have been based on male rodents.^7,32,33 We tested for increased TNF in various brain regions that might be correlated with affective disorders such as depression. Subsequently, we tested the efficacy of soluble TNF inhibition with XPro1595 via peripheral s.c. and i.c.v. administration in prevention of post-SCI depression. Surprisingly, peripheral inhibition of TNF was associated with an increase in incidence of depression after SCI, whereas central administration provided no increase or decrease in the incidence. These results suggest that additional effects of post-SCI neuroinflammation likely are involved in expression of depressive phenotype and TNF modulation in SCI patients must be carefully considered due to potential adverse effects.

Methods

Subjects

Female Sprague Dawley rats (n = 112, 225–250 g, Charles River Laboratories, Wilmington, MA) were housed with access to food and water ad libitum on a 12-h light-dark cycle. All experiments were compliant with federal guidelines and approved by the Drexel University Institutional Animal Care and Use Committee. All subjects were handled 10 min/day for a week prior to study initiation, after which they were singly housed. Baseline behavioral tests (sucrose preference, novel object recognition, and social exploration) were conducted 1 week prior to SCI. Behavioral tests were repeated at 28 days post-injury for all groups, followed by a forced swim test. A subset of rats (n = 53) received treatment with either XPro1595 (s.c. n = 22, i.c.v. n = 15) or sterile saline (s.c. n = 10, i.c.v. n = 6). S.c. administration began at the time of injury and was continued until sacrifice. I.c.v. administration by osmotic pump began 1 day prior to injury and was continued until sacrifice.

T9 contusion

Subjects were anesthetized by hindlimb intramuscular injection with an XAK mixture (6 mg/kg xylaxine [Anased, Lloyd Laboratories, Shenandoah, IA], 6 mg/kg acepromazine [Promace, Vedco Inc., St. Joseph, MO], and 60 mg/kg ketamine [Ketaset, Vedco Inc.]). Sustained-release buprenorphine (1.0 mg/kg s.c., Zoopharm, Laramie, WY) was given prior to skin incision. Rats received laminectomy and were stabilized at the T8 and T10 vertebrae prior to midline T9 contusive injury of 150 Kdyn with a 1 sec dwell time using the Infinite Horizons Impact device (Precision Systems and Instrumentation, Lexington, KY). Probe tip diameter was 2 mm. The site of probe impact was submerged in sterile saline during injury. Dorsal musculature was closed using 5-0 Vicryl coated suture. The skin incision was closed using Michel wound clips. After injury, bladders were expressed twice daily until reflexive micturition was recovered. Subjects received cefazolin (160mg/kg s.c. 2 × daily, Sandoz, Princeton, NJ) and lactated Ringer's solution (3 cc, s.c.) for 7 days post-injury.

XPro1595 administration

Peripheral administration

Beginning at the time of T9 contusion, 32 rats received s.c. injection of XPro1595 (n = 22, 10mg/kg [generously provided by David E. Szymkowski of Xencor]) or equivalent volume of sterile saline solution (n = 10) every 3 days throughout completion of behavioral testing beginning at 4 weeks post-injury.

Intracerebroventricular cannulation

One day prior to T9 contusion, a cohort of 21 rats was implanted with a unilateral i.c.v. cannula (Alzet Brain Infusion Kit 2, Model 2004 Alzet Osmotic Pump, Durect Corporation, Cupertino, CA) into the left lateral ventricle (AP: −1.0. ML: +2.0, DV: −4.0 → −3.5). Subjects were anesthetized with isofluorane (5% induction level) prior to surgical procedure. Once hindlimb pinch reflex and corneal reflex were no longer elicited, rats were maintained at a 2.5–3% isofluorane level for the duration of surgery. Each subject was suspended in a stereotaxic apparatus (David Kopf Instruments, Tujunga, CA) using ear bars, and the surgical field was sterilized using ethanol and betadine. A midline skin incision was made to expose the skull and bregma was identified to calculate lateral ventricle coordinates for each animal.

A Dremel drill with 0.9-mm bit was used to penetrate the skull over the projected site of cannula implantation and bilaterally at two coordinates approximately 8 mm rostral and caudal to the implantation site for bone screws (1.59 mm O.D., 3.2 mm length, Stoelting, Wood Dale, IL) to help secure the head cap with dental cement (Stoelting). The cannula (28 gauge, 5 mm length) was subsequently lowered until the surface of the brain was reached and DV coordinates were calculated from this stereotaxic value. Once dental cement secured the head cap, the cannula was connected to an Alzet Osmotic Pump (28 day, 0.25 μL/h) filled with either saline (n = 6) or XPro1595 (n = 15). An incision was made between the shoulder blades and the pump was inserted subdermally. The incision was closed with 4-0 silk suture. At least 48 h prior to implantation, pumps were filled and primed by immersion in sterile saline at 37°C.

Behavioral testing

Rats were assessed with a battery of behavioral tests 1 week before injury and again at 28 days post-injury. Naïve rats were tested 1 week prior to single-housing and again at 28 days.

Sucrose preference

Rats were habituated to 2% sucrose solution for 48 h, 1 week prior to testing. For habituation, two bottles were placed one on each side of the home cage with 300 mL of either water or 2% sucrose. The position of the bottles was switched after 24 h to reduce bias from side preference. One week after sucrose habituation, rats were water deprived for 12 h and the testing phase was conducted. At 5:00 a.m., rats were presented with two bottles (one of water, one of 2% sucrose), with the location of the sucrose bottle alternating sides between each animal. Bottles were left for 2 h, with position of bottles being switched after 1 h. Pre- and post-test bottle weights were recorded and % sucrose preference was calculated as the volume of sucrose consumed/total liquid volume consumed × 100. A significant decrease in the preference for sucrose solution suggests an anhedonic response. Sucrose preference values of less that 10% were not included in clustering due to the sensitivity of clustering to extreme values. These instances were rare, with only eight values removed throughout the study.

Novel object recognition

Rats were acclimated to novel object chambers (40.5 cm × 40.5 cm) for 3 days, 10 min each day before testing. The familiar trial consisted of placing the rat in the novel object chamber with two identical objects for a 5-min exploration period, followed by a period of 1.5 h in the home cage before initiation of the novel object trial. The novel object testing paradigm consisted of 5-min exploration time with one object from the familiar trial and one completely distinct object of similar size. All objects and chambers were cleaned with 70% ethanol after each trial to prevent scent-cue bias. Both the familiar and novel trials were video-recorded from above and scored post hoc. Interaction was defined as having the nose directed at the object from a distance of 2 cm or closer. Climbing or sitting on the objects was not counted as exploration. Percent novel object exploration was calculated as the time spent exploring novel object (sec)/time exploring both objects (sec) × 100. A decrease in time spent exploring the novel object suggests a memory deficit.

Open field

Rats were placed in open field chambers for 5 min to assess locomotive behavior. MotorMonitor Host software (Kinder Scientific, Poway, CA) calculated movement based on beam breaks in the chamber. The chamber was digitally separated into peripheral and central segments. Less time spent in the center of the chamber is suggested to be an indicator of anxious phenotype.

Social exploration

Rats were acclimated to open field chambers 3 times, 10 min each before testing. Rats were placed in open field alone for 5 min of exploration time. After 5 min, a novel female intruder rat of similar size and age was placed in the open field with test rat for 5 min. Social interaction was scored as any sniffing, climbing, following, or touching initiated by test rat. Social interaction percentage was calculated as time spent interacting with intruder (sec)/600 sec × 100.

Modified forced swim test

The forced swim test was only performed in one session post-injury (as opposed to the classical 2-day testing paradigm) to avoid learned helplessness adaptation to multiple bouts of swimming. Rats were placed in a cylindrical tank of water (50.3 cm height × 27.7 cm diameter, 30 cm water level, 23–25°C) for a 10-min swimming period. Each test was video-recorded from a point directly above the tank. Post hoc scoring was done via modified 5-sec interval method to assess time spent immobile, versus swimming or climbing.^34,35 The modified scoring method has been suggested to have greater predictive validity in response to antidepressant treatment.^34,36 Immobility was defined as a lack of movements, except for the small movements necessary for keeping the head above water.³⁷ Percent immobility ([time spent immobile (sec)/total swim time (600 sec)] × 100) has been established as an indicator of hopelessness.

BBB

Ambulation was assessed weekly in all non-treated rats using the Basso, Beattie, Bresnahan (BBB) open field rating scale.³⁸ Once characteristic spontaneous motor recovery was established for our model, all treated cohorts (XPro1595 or saline) were tested at 4 weeks post-injury only. Rats were placed in an open field apparatus (95 cm diameter) and ambulation was scored for 4 min by two trained observers. The BBB scale runs from 0 (no observable hindlimb joint movement) to 21 (naïve rat ambulation). A score of 10 or greater indicates hindlimb stepping with various levels of coordination and paw placement. A score of 14 indicates consistent weight-supported stepping with consistent forelimb/hindlimb coordination.

Automated von Frey

Six animals per testing period were placed in a long Plexiglas chamber with five divider walls so that neighboring animals could not be seen. The chamber floor was a wire mesh to allow for microfilament probe access. After a 10-min acclimation period, the monofilament was centered below the plantar surface of the hindpaw. The automated plantar aesthesiometer (Ugo Basile, Gemonia, Italy) was raised at a 20-sec ramp speed up to a maximum of 50g of force. The force at which the animal withdrew the paw with evidence of supraspinal stimulus awareness (lifting, looking at, licking the hindpaw, vocalizing, moving away from stimulus) was recorded for each trial. Five trials/hindpaw were conducted for each animal. Complete absence of paw withdrawal was scored as 50g force threshold (maximum). The withdrawal force for all five trials was averaged for each hindpaw to give the withdrawal threshold. A post-injury reduction in the withdrawal threshold of greater than 50% from baseline testing was considered positive for tactile allodynia. Each rat was given an inter-trial interval of at least 1 min to reduce a fatigue effect.

Hargreaves thermal testing

Three animals per testing period were placed in a Plexiglas chamber with glass-pane bottom and two divider walls so that neighboring animals could not be seen. After a 10-min acclimation period, the infrared heat source (I.R. intensity 48, Ugo Basile, Gemonia, Italy) was centered below the plantar surface of the hindpaw. Latency to withdraw the hindpaw with evidence of supraspinal stimulus awareness (lifting, looking at, licking the hindpaw, vocalizing, moving away from stimulus) after initiation of thermal stimulus was recorded for each trial. Five trials were conducted for each hindpaw in a testing period. The order of paw testing was randomized and each animal was given an inter-trial period of at least 1 min to minimize fatigue effect. The withdrawal latency for all five trials was averaged for each hindpaw. A post-injury reduction in the withdrawal threshold greater than 50% from baseline testing was considered positive for thermal allodynia.

Protein quantification

Protein extraction

Rats were euthanized at 5 weeks post-injury by isofluorane anesthesia and intracardial injection of Euthasol (0.3 mL, 390 mg/kg sodium pentobarbital and 50 mg/kg phenytoin) before decapitation. Brain tissue was removed and regions of interest (DRN, hippocampus [HP], prefrontal cortex [PFC]) were blocked into 2-mm thick coronal slices using a brain matrix. Three-millimeter diameter punches (Disposable Biopsy Punch, Robbins Instruments, Chatham, NJ) of tissue from each target region were collected and flash frozen on dry ice. Tissue was weighed and ice-cold radioimmunoprecipitation assay (RIPA) lysis buffer was added at a volume 10 × weight. Tissue was sonicated on ice in brief bursts and centrifuged for 40 min at 14,000 rpm, 4°C. Supernatant was removed and aliquots were stored at −80°C.

Protein concentration (μg/mL) was quantified for each supernatant using the Pierce BCA Protein Assay Kit (Pierce, Rockford, IL). Five μL of each supernatant was diluted in 95 μL nuclease-free water before BCA. Ten μL of each diluted sample and included albumin standards were added in duplicate to a 96-well microplate. The Pierce BCA Kit working reagent was prepared at a ratio of 50:1 (A:B) and 250 μL was added to each well of the microplate. The plate was incubated at 37°C for 30 min and read immediately using a TECAN Infinite 200 pro spectrophotometer (TECAN Trading AG, Switzerland) at 450 nm. Protein values were calculated from the linear regression equation of the standards and multiplied by a dilution factor of 20 to give the final μg/mL concentration.

TNF ELISA

Sandwich ELISA (enzyme-linked immunosorbent assay) Kits (ab100785, Abcam, Cambridge, MA) were used to quantify TNF levels for each brain region (DRN, HP, PFC). One hundred μL of each sample (30 mg total protein diluted in Abcam 1 × sample diluent buffer) supernatant was loaded in duplicate into a 96-well plate pre-coated with anti-rat TNF antibody and incubated at 4°C overnight. Plate was rinsed with 300 μL wash buffer 4 times and incubated in biotinylated anti-TNF solution with shaking (500–600/min) for 1 h at room temperature. Wash procedure was repeated and plate was incubated in streptavidin horseradish peroxidase (HRP) solution with shaking for 45 min at room temperature. Wash procedure was repeated and plate was incubated in 3,3′,5,5′-tetramethylbenzidine (TMB) one-step solution in the dark for 30 min. Fifty μL of stop solution was added to each well and the plate was analyzed immediately using a TECAN Infinite 200 pro spectrophotometer (TECAN Trading) at 450 nm. Cytokine concentration (pg/mL) was determined from the linear regression of provided standards and multiplied by the appropriate dilution factor for each sample based on volume of supernatant required to reach 30 μg of protein.

WES capillary electrophoresis of pP65

Levels of TNFR1 signaling were assessed through quantification of phosphorylated P65 (pP65) protein levels in brain supernatant, normalized to actin levels. The Wes ProteinSimple System (Santa Clara, CA) utilizes automated capillary-based electrophoresis and immunoprobing of sample supernatant to measure protein with greater sensitivity and reduced sample size compared with traditional Western blotting. Compass for Simple Western software automatically provides quantitative analysis of assay data (area under the curve minus baseline) and creates a display of detected proteins as bands in a lane view (Fig. 1) similar to what one would see with a traditional Western blot. A primary antibody to pP65 (Novus Biologicals, NB100-82088) was used to detect levels of NF-κβ P65 only when it is phosphorylated at serine 536. Specificity of this antibody for the phosphorylated form was confirmed with whole cell lysate of serine/threonine phosphatase inhibited HeLa cells (Abcam, clone AC-74, A5316). A primary antibody to β-actin (Sigma, #A5316) was used as a loading control for normalization of pP65 values.

FIG. 1.

Graphs display mean + SEM area under the curve pP65 intensity normalized to beta actin for each treatment group. (A) WES results of rabbit anti-pP65 (65 kDa) and mouse anti-beta actin (48 kDa). (B) One-way ANOVA of pP65 signal in clusters of each treatment group did not reveal any significant differences. (C) One-way ANOVA of combined subjects from each treatment group revealed a significantly higher pP65 level in animals receiving s.c. XPro compared with i.c.v. XPro (*p < 0.05). ANOVA, analysis of variance; i.c.v., intracerebroventricular; pP65, phosphorylated P65; s.c., subcutaneous; SEM, standard error of the mean.

Lesion epicenter analysis

The thoracic spinal cord was extracted from each animal immediately after euthanasia and fixed overnight in 4% paraformaldehyde at 4°C. Each spinal cord then was cryoprotected in 30% sucrose solution. The lesion site (T9 spinal cord epicenter +2 mm rostral and caudal) was blocked and embedded in optimal cutting temperature (OCT) compound (Fisher Scientific, Pittsburgh, PA) for cryosectioning at −20°C (Leica Microsystems, Wetzlar, Germany). Twenty-five micrometer sections, 250 μm apart and spanning the entire tissue block were mounted directly onto Superfrost Plus slides and stored at 4°C for processing. Sections were stained with cresyl violet (Sigma-Aldrich, St. Louis, MO) to identify Nissl substance and euriochrome cyanine (Sigma-Aldrich) to identify myelin. Slides were cover-slipped immediately after staining with DPX (dibutylphthalate polystyrene xylene) mounting medium (Fisher Scientific). The area (mm²) of spared white matter at the lesion epicenter was measured using the Cavaleri estimator method (Stereo Investigator, MBF Bioscience, Burlington, VT). Epicenter spared white matter for all animals in each cluster was averaged and compared using one-way analysis of variance (ANOVA). Animals with area of sparing greater than 3 standard deviations above or below the mean were considered outliers and removed from further behavioral and biochemical analysis.

Statistical analysis

Principal components analysis

All post-injury behavioral data were analyzed with principal components analysis based on orthogonal varimax rotation. Correlation matrix generated from principal component analysis revealed that social exploration, novel object recognition, and total distance traveled in open field did not have a correlation coefficient of r ≥ 0.3 with any other behavioral tests. Therefore, these tests were excluded from principal components analysis. Data distribution passed Bartletts's test of sphericity (p = 0.048). Only one component was extracted, explaining 64% of cumulative variance. Forced swim test and sucrose preference loaded on component 1 and were used to establish behavioral clusters.

Hierarchical cluster analysis

Hierarchical cluster analysis was used to identify the number of clusters that arose within injured animals based on sucrose preference and forced swim tests. Data were standardized by transformation to z score. Ward's method of hierarchical clustering was applied with a squared Euclidian distance measure.³² The resultant dendrogram and agglomeration schedules were used to determine cluster number based on the presence of a large increase in agglomeration coefficient. Once the appropriate number of clusters was determined the data were analyzed with k-means cluster analysis. Although hierarchical cluster is good for determining the number of clusters that naturally occur in a data set, it is susceptible to change with addition of subjects. K-means cluster analysis is based on a predetermined number of clusters with more stable cluster centers.

K-means cluster analysis

K-mean cluster analysis was used to separate subjects into three clusters according to statistical rigor and fidelity of cluster centers. Data from forced swim and sucrose preference were analyzed with a maximum of 20 iterations.

One-way ANOVA

After k-means analysis, one-way ANOVA was used to determine significant differences between the resultant clusters in each behavioral test. The behavioral phenotype of each cluster was used to identify non-depressed, depressed, and melancholic-depressed phenotypes. One-way repeated measures analysis of covariance (ANCOVA) of BBB scores was used to confirm that ambulation did not differ between clusters, with BBB scores at 3 days post-injury used as a covariate. After BBB progression in our model was established in untreated rats, XPro1595-treated and saline-treated rats were only assessed for BBB ambulation differences at 4 weeks post-injury with one-way ANOVA. One-way ANOVA was used to assess differences in TNF and pP65 levels, as well as white matter sparing at the lesion epicenter between clusters. Tukey's multiple comparison test was used for post hoc analysis. If data did not adhere to a normal distribution, the Kruskal-Wallis test was used to assess statistical differences between clusters. Data points in each measure that were 3 standard deviations from the mean were considered outliers and removed.

Results

Behavioral assessment

At 4 weeks post-injury, hierarchical cluster analysis revealed three behaviorally distinct clusters from the 53 injured rats not receiving treatment with XPro1595 or saline. Data from the tests that loaded on principle component analysis (PCA) component 1 (sucrose preference, forced swim test) was then analyzed with k-means analysis to determine cluster membership. Based on overall behavioral performance these clusters were phenotypically labeled as non-depressed, depressed, or melancholic-depressed (patients with persistent anhedonia as a subclass of MDD). Greater than 50% of injured rats reliably expressed the depressive phenotype after T9 contusion of four experimental groups prepared over an 8-month period.

Novel object recognition was used to explore retention memory in post-SCI depression. One-way ANOVA revealed no significant differences between clusters in novel object recognition (F(3,53) = 0.84, p = 0.48), suggesting that short-term memory deficit is not a reliable indicator of SCI-induced depressive phenotype in female rats (Fig. 2A). There was no significant difference in the total time each cluster spent exploring the objects (F(3,56) = 1.883, p = 0.144). Baseline novel object recognition did not differ significantly between clusters (one-way ANOVA F(2,42) = 0.59, p = 0.56).

FIG. 2.

Bar graphs display mean + SEM for behavioral tests assessing depressive phenotype at 4 weeks post-injury. (A) No significant difference between clusters was found for novel object recognition. (B) Depressed and melancholic-depressed clusters demonstrated significantly higher immobility in the forced swim test when compared to non-depressed (δ). (C) Melancholic-depressed cluster demonstrated decreased preference for sucrose. (D) No significant difference in social exploration was seen between clusters. (E) Open field ambulatory ability after SCI did not differ between clusters (****p < 0.0001, δ significantly different from non-depressed). SEM, standard error of the mean.

One-way ANOVA revealed a significant difference in forced swim immobility between the clusters (F(3,45) = 33.02, p < 0.0001). Tukey post hoc analysis revealed significantly greater mobility in one of the injured clusters compared with the other two. This cluster was therefore classified as non-depressed (n = 22). Post hoc analysis revealed no significant differences in forced swim immobility between the non-depressed cluster and the singly housed naïve rats (n = 6), indicating that injury alone was not sufficient to elicit this behavioral phenotype. Two of the injured clusters demonstrated significantly greater immobility than both the non-depressed and singly housed naïve rats (p < 0.0001), signifying a depressive phenotype (Fig. 2B, depressed n = 22, melancholic-depressed n = 9). Forced swim test was not assessed pre-injury to avoid a learned helplessness effect due to multiple bouts of swimming.

A main effect of injury on sucrose preference was demonstrated by one-way ANOVA (F(3,53) = 91.34). Tukey post hoc analysis revealed a diminished preference for sucrose in only one cluster (p < 0.0001), further denoted as melancholic-depressed. Injury alone was not associated with decreased sucrose consumption, as singly housed naïve rats did not differ significantly from non-depressed or depressed clusters (Fig. 2C). Baseline sucrose preference was not significantly different between the clusters (F(3,53) = 1.05, p = 0.38).

No significant effect was revealed in one-way ANOVA analysis of social exploration between clusters (F(3,42) = 2.02, p = 0.13), indicating that social exploration in female rats is not reliably altered after SCI (Fig. 2D).

Despite differences in performance in depressive behavioral tests, no significant difference was found in BBB open field score between the clusters at the time of 4 weeks post-injury behavioral testing with means adjusted for 3-day post-injury BBB scores (one-way repeated measures ANCOVA F(2,49) = 0.551, p = 0.58, partial η² = 0.02). Analysis of earlier time-points confirmed that BBB scores did not differ between the clusters (one-way repeated measures ANCOVA: 6 DPI F(2,49) = 2.194, p = 0.122, partial η² = 0.082; 14 DPI F(2,49) = 0.729, p = 0.488, partial η² = 0.029; 21 DPI F(2,49) = 1.711, p = 0.191, partial η² = 0.065). BBB ambulatory score at 4 weeks post-injury was not significantly correlated with performance in behavioral tasks (Pearson correlations NOR: r = 0.17, p = 0.45, FST: r = 0.17, p = 0.33, SP: r = −0.07, p = 0.75). Average adjusted BBB scores at 4 weeks post-injury were 11.3 (non-depressed), 12.0 (depressed), and 11.7 (melancholic-depressed) (Fig. 2E). Most rats were able to weight-bear by 4 weeks post-injury. To assess motor retardation, total distance traveled in the open field box was analyzed with one-way ANOVA. No significant difference was found between clusters (F(2,29) = 0.651, p = 0.529). The tendency of rats to avoid the center of the activity chamber was calculated to assess passive anxiety. The total time each cluster spent in the center of the chamber was not significantly different between the clusters (F(2,29) = 1.512, p = 0.237). Injury severity did not differ between clusters, with no main effect of cluster membership on spared white matter (F(2,10) = 1.59, p = 0.25) in one-way ANOVA.

Von Frey testing was used to assess tactile allodynia after injury (Fig. 3A,B). One-way ANOVA revealed no main effect of clustering on the change of paw withdrawal threshold from baseline (LHL: F(2,50) = 0.79, p = 0.46, RHL: F(2,50) = 1.84, p = 0.17). No significant difference in thermal allodynia was detected by Hargreaves testing (LHL: (F(2,49) = 1.43, p = 0.25, RHL: F(2,50) = 1.07, p = 0.35, Fig. 3C,D).

FIG. 3.

(A,B) Plots demonstrate the change in probe force required to elicit paw withdrawal compared with baseline for each of the clusters. (C,D) Change in infrared intensity required to elicit paw withdrawal compared with baseline for each cluster is demonstrated. Dashed lines indicate threshold for identification of neuropathic pain. No significant difference in paw withdrawal thresholds was seen between clusters in either test. Locomotion in an open field chamber was recorded for each cluster. No significant difference was found in the total distance traveled (E), or the total time spent in the center (F) or periphery (G) of the chamber.

Cytokine levels are elevated in a regionally specific manner 5 weeks after SCI

Tissue from a subset of each cluster (singly housed naïve n = 6, non-depressed n = 14, depressed n = 14, melancholic-depressed n = 6) was used for analysis of TNF levels across multiple brain regions. TNF was significantly higher in the DRN of depressed and melancholic-depressed subjects when compared with both non-depressed and singly housed naïve subjects at 5 weeks post-injury (Fig. 4A, F(3,35) = 10.41, p < 0.0001). All injured rats displayed a higher level of TNF in the HP compared with singly housed controls, with no difference between the injured clusters (F(3,27) = 4.68, p = 0.009, Fig. 4B). TNF levels did not significantly differ between groups in the PFC (F(3,30) = 1.58, p = 0.21, Fig. 4C). Average TNF levels normalized to total protein appeared highest in the DRN and decreased in a caudal to rostral direction, with overall levels being the lowest in the PFC.

FIG. 4.

Graphs display mean + SEM supraspinal TNF levels. Depressed and melancholic-depressed clusters displayed increased TNF in the DRN compared with non-depressed (δ, A). All injured clusters demonstrated higher TNF in the hippocampus compared with naïve controls (B). No significant difference in PFC TNF levels was seen between clusters (C). TNF levels in the DRN of injured rats were significantly correlated with immobility in the forced swim test (D), but not sucrose preference (E) (*p < 0.05, **p < 0.01, ***p < 0.001). DRN, dorsal raphe nucleus; PFC, prefrontal cortex; SEM, standard error of the mean; TNF, tumor necrosis factor.

There was a significant positive correlation (r = 0.44, p = 0.01) between TNF levels in the DRN and immobility in the forced swim test (Fig. 4D); however, increased TNF levels were not significantly correlated with sucrose preference (r = 0.23, p = 0.187, Fig. 4E).

Peripheral administration of XPro1595 exacerbates depressive phenotype after SCI

Given the correlation between DRN TNF and immobility in the forced swim test, we assessed the efficacy of soluble TNF inhibition in prevention of SCI-depression. Rats received s.c. injections of XPro1595 (10mg/kg) or saline every 3 days, beginning at the time of injury and continuing to the day of euthanasia. K-means cluster analysis separated the subjects in each treatment group into three clusters: non-depressed, depressed, or melancholic-depressed (Fig. 5A,B). Saline-treated clusters exhibited a similar phenotypic distribution to non-treated rats, with greater than 50% of subjects falling into a depressed or melancholic-depressed cluster based on forced swim and sucrose preference testing (Fig. 5A,B). Surprisingly, one-way ANOVA revealed a main effect of treatment in the forced swim test, with all XPro1595 clusters exhibiting increased immobility compared with the non-depressed saline cluster (F(8,40) = 19.57, p < 0.0001, Fig. 5A). Based on the classification established in the non-treated subjects, all XPro1595 clusters would thus be considered in the depressed category, with the melancholic-depressed cluster designated as s.c. XPro melancholic-depressed and the two distinct clusters of depression designated as s.c. XPro depressed 1 and s.c. XPro depressed 2.

FIG. 5.

Graphs display mean + SEM behavior for each treatment group. Peripheral (s.c.) Xpro1595 administration resulted in increased forced swim immobility in all treated clusters when compared with saline non-depressed animals (δ). The two s.c. XPro clusters that were identified as depressed clustered separately from each other based on extent of immobility (s.c. XPro depressed 1 vs. s.c. XPro depressed 2). (A) Each treatment group included a melancholic cluster with significantly lower sucrose preference than at least one other cluster (B). Subjects treated with i.c.v. XPro1595 demonstrated a higher incidence of anhedonia than the other treatment groups (C) (*p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001, δ significantly different from saline non-depressed). i.c.v., intracerebroventricular; s.c., subcutaneous; SEM, standard error of the mean.

Tukey post hoc analysis revealed significantly higher immobility in the s.c. XPro depressed 1 cluster compared with the s.c. depressed 2, with no other differences among XPro1595 clusters. Additionally, significantly decreased sucrose preference was demonstrated in 27% of rats in the XPro1595 group, compared with 19% in the saline-treated subjects (F(8,36) = 26.68, p < 0.0001, Fig. 5B,C). Open field assessment of locomotor behavior revealed no significant difference in the total distance traveled in the activity chamber between treatment groups (F(8,40) = 0.565, p = 0.800), suggesting the increase in forced immobility did not reflect a general decrease in motor activity. The extent to which rats avoid the center of the activity chamber has been used to assess anxiety. One-way ANOVA revealed no significant difference in time spent in the center of the chamber between treatment group, suggesting no difference in anxious behavior (F(8,40) = 0.921, p = 0.509).

Intracerebroventricular administration of XPro1595 does not alter incidence of depressive phenotype after SCI

To determine whether ineffectiveness of s.c. XPro1595 administration was due to inadequate crossing of XPro1595 across the blood–brain barrier with our peripheral treatment paradigm, rats received either drug or sterile saline through intraventricular cannulation into the lateral ventricle. K-means analysis was used to separate the subjects in each treatment group into three clusters based on forced swim test and sucrose preference. One-way ANOVA revealed a significantly higher immobility in two i.c.v. XPro1595 clusters (depressed and melancholic-depressed) compared with the third (non-depressed), as well as the non-depressed saline control (F(8,40) = 19.57, p < 0.0001, Fig. 5A). The number of saline-treated rats in the i.c.v. group was relatively small, so i.c.v. saline rats were combined with s.c. saline-treated rats to increase statistical power. Before subjects were combined, non-depressed rats from each saline-treated paradigm were compared in forced swim immobility and sucrose preference to ensure there was no difference in behavioral phenotype. Student's t test confirmed that non-depressed rats treated with saline s.c. did not differ significantly in forced swim immobility (t(6) = 0.127, p = 0.90) or sucrose preference (t(6) = 0.87, p = 0.42) when compared with i.c.v. saline, confirming that route of administration did not affect behavioral phenotype.

Forty-six percent of rats treated with i.c.v. XPro1595 were classified as depressed, which is similar to the phenotype seen in saline controls (50%). Although i.c.v. treatment with XPro1595 did not greatly decrease the prevalence of depressive phenotype compared with saline treatment, importantly it did not exacerbate the phenotype as was seen with peripheral treatment. The effect of i.c.v. XPro1595 on anhedonia was distinct from that of hopelessness. One-way ANOVA revealed a significant difference between clusters in sucrose preference (F(8,36) = 26.68, p < 0.0001, Fig. 5B). Tukey's post hoc analysis revealed a melancholic-depressed cluster with significantly lower sucrose preference than the i.c.v. non-depressed cluster (p = 0.009) and trending toward significance compared with the i.c.v. depressed cluster (p = 0.075) (Fig. 5B). The prevalence of anhedonia was markedly higher with i.c.v. XPro1595, at 27%, than saline treated rats (Fig. 5C). Baseline sucrose preference was not significantly different between groups (Kruskal-Wallis test H = 8.82, p = 0.36). As with the untreated rats, novel object recognition did not differ between any of the clusters (one-way ANOVA F(8,29) = 0.30, p = 0.96). Anxiety phenotype did not seem to be affected by i.c.v. XPro1595, with no significant difference in time spent in the center of the open field activity chamber compared with saline-treated rats (F(8,40) = 0.921, p = 0.509).

Lesion severity among treatment groups

Lesion analysis (not shown) indicated no main effect of treatment on white matter sparing and one-way ANOVA of BBB scores at 4 weeks post-injury were not significantly different between the treatment groups (F(8,41) = 0.939, p = 0.496), indicating behavioral results are not a direct result of ambulation ability (F(8,36) = 1.486, p = 0.197).

NF-κb signaling with XPro1595 administration

WES analysis of phosphorylated NF-κb subunit P65 levels in samples from the DRN did not reveal a significant difference between clusters of each treatment group (F(9,35) = 1.447, p = 0.207, Fig. 1B). Despite lack of statistical significance, the average pP65 levels appeared higher in animals treated with s.c. XPro than the other groups. The pP65 data from all subjects in a treatment group were combined to analyze the overall difference in NF-κb signaling between each treatment paradigm. One-way ANOVA demonstrated a main effect of treatment (F(3,41) = 3.024, p = 0.040, Fig. 1C). Tukey's post hoc analysis revealed a significantly higher level of pP65 in animals treated with s.c. XPro than those receiving i.c.v. XPro, suggesting a paradoxical increase in NF-κb signaling with peripheral XPro administration that is not seen with i.c.v. administration (p = 0.041). Average levels of pP65 in s.c. XPro subjects were also higher than saline and naïve animals, although these differences did not reach significance (saline p = 0.13, naïve p = 0.19). These results suggest that peripheral TNF inhibition at this dosage was not sufficient to stop cytokine signaling in the dorsal raphe.

Discussion

The goal of this study was to assess the role of TNF in the development of depression after moderate SCI in female rats. We determined that increased levels of the cytokine in the dorsal raphe nucleus, the primary source of serotonin to the brain, correlated with a hopelessness phenotype 4 weeks after a mid-thoracic contusion injury. To determine the therapeutic efficacy of soluble TNF inhibition in this population, we administered dominant negative inhibitor XPro1595 through s.c. and i.c.v. delivery starting at the time of injury. Surprisingly, s.c. XPro1595 exacerbated the depressive phenotype, whereas i.c.v. administration did not increase or decrease the incidence of depression. Our results indicate a complicated role for TNF and likely other pro-inflammatory cytokines in depression after SCI that needs to be elucidated before therapeutically targeting these molecules for treatment of patients.

Behavioral phenotype of SCI-induced depression in female rats

This study indicates that after a moderate spinal cord injury, over half of female rats develop a depressive phenotype suggestive of common MDD symptoms in patients. Previous models of SCI using adult male rats reported a depression incidence of 35–57%.^7,32 The prevalence of post-SCI depression in the patient population is estimated to be about 26%.³⁹ The apparent discrepancy in incidence between animal model and patients could be due to under-diagnosis of MDD in SCI patients, which also is reflected in the variability of prevalence reported in clinical literature.^2,40

The symptom profile in animal models of post-SCI depression appears to differ between sexes and supports the need for a variety of behavioral tests when assessing affective disorders. In female rats, hopelessness was the primary symptom defining the depressive phenotype with a subset also demonstrating anhedonia (based on sucrose preference), which could correspond to the melancholic form of MDD in patients.⁴¹ Male rats with a similar injury presented with anhedonia, social withdrawal, and motor retardation.⁷ Notably, there was no effect of injury on social exploration or locomotor behavior in our model, which may reflect a sexually dimorphic presentation of post-SCI depression. This symptom profile has been demonstrated in other rodent models of anxiety/depression, where females tend to display higher immobility in forced swim test and less motor retardation in open field locomotor assessment when compared with males.^42,43 Anxious phenotype as indicated by time spent in the center of the activity chamber did not differ between groups; however, it is possible that a more sensitive test of anxiety would be necessary to identify this behavior.

Given the elevation of TNF in the HP of all injured animals, the absence of significant novel object recognition deficit in this model may seem surprising, as this assay classically has been associated with impaired HP function.^44,45 Our novel object paradigm tested short-term memory due to disruption of this process in established depression models.³³ It is possible, however, that a longer interval between familiar and novel object trials would reveal a deficit in injured rats. Alternative forms of memory deficit, such as working memory, could also be more relevant to our model of depression. In a mouse model of SCI, Wu and co-workers³³ demonstrated a significant deficit in special memory in the Morris water maze and Y-maze compared with naïve controls. Given the heterogeneous presentation of MDD in the patient population, however, memory deficit is not necessary to model this disorder.

Behavioral battery performance was not correlated with BBB score or spared white matter at the lesion site, suggesting that a decreased ability to ambulate was not a confounding factor in cluster generation. Incidence of tactile and thermal allodynia also did not differ between the clusters, precluding depression as a product of chronic pain.

Changes in DRN circuitry are indicated by increase in TNF levels

Increased levels of the inflammatory cytokine TNF in the DRN of animals exhibiting a depressive phenotype lends support to potential serotonergic/gamma-aminobutyric acid (GABA)ergic dysfunction in SCI-induced depression and is reinforced by a significant correlation of TNF levels with immobility in the forced swim test. TNF alters trafficking of α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) and GABA receptors to the neuronal membrane.^46–49 An alteration in excitatory/inhibitory signaling could result in disruption in the activity of tightly regulated serotonergic circuitry. Many of the excitatory projections from the PFC to the DRN synapse on local GABAergic interneurons.^50,51 An increase in AMPA receptor trafficking in this cell type could cause an overall decrease in serotonergic activity that cannot be compensated for by inhibition of serotonin transport alone.

A pathological decrease in serotonergic activity has been demonstrated in other affective disorders, such as a rodent model of social defeat.^51,52 Lack of significant correlation between TNF and sucrose preference could be a result of the low number of melancholic-depressed subjects. It is also possible that the two measures of depression are assessing separate neurotransmitter systems. Anhedonia and reward drive have long been associated with the VTA-dopaminergic system (reviewed by Belujon and Grace⁵³). It is possible that melancholic depression represents a subset of animals with more severe dopaminergic disruption than the depressed cluster, which we believe to be serotonin-disruption based.

Depressive phenotype increases with peripheral but not central block of TNF activity

A surprising finding in this study was the exacerbation of depressive phenotype after chronic s.c. XPro1595 administration. Pharmacological inhibition of both soluble and transmembrane TNF previously has resulted in a paradoxical increase in demyelinating lesions in a subset of patients with multiple sclerosis, leading to further scientific exploration of TNFR2 neuroprotective and remyelinating functions.^54–56 Based on this knowledge, it was hypothesized that maintaining TNFR2 signaling in disease states was favorable. Selective inhibition of soluble TNF with XPro1595 has been a successful treatment in several animal models of inflammatory disease, such as EAE.²⁸

Given our results, however, it is possible that TNFR2 signaling is not always beneficial and may need to be studied further before it is therapeutically targeted. Recent work demonstrated a difference in the effect of TNFR2 on peripheral macrophages versus central microglia.⁵⁷ Centrally, TNFR2 signaling seemed to be neuroprotective in an EAE model.⁵⁷ Peripherally, however, TNFR2 signaling in macrophages contributes to Th17 lymphocyte generation.⁵⁷ This particular lymphocyte class has been implicated in CNS damage from autoimmunity, as well as MDD in the general population.^58–60 Bethea and colleagues⁶¹ demonstrated that SCI induces peripheral macrophage production of TNF after injury, which possibly could initiate disruption of the blood–brain barrier to facilitate immune cell entry into the CNS. It is possible that selectively inhibiting soluble TNF shifted the peripheral balance toward TNFR2 signaling in macrophages, resulting in an increase in Th17 lymphocyte infiltration and the development of a depressive phenotype. Mironets and associates³¹ recently demonstrated an elevated level of regulatory T cells with intrathecal administration of XPro1595, along with maintained astrocyte count in the injured spinal cord of rats with thoracic transection. These results further support the conclusion that XPro1595 alters the immune profile.

Central administration of XPro1595, although not an efficacious antidepressant at our dosage, did not cause the same exacerbation of hopelessness phenotype, suggesting that peripheral delivery is unique in its mechanism of action. Another interesting finding is the increased expression of anhedonia with i.c.v. delivery of XPro1595. A study by Simen and co-workers⁶² demonstrated that TNFR1 and TNFR2 subtypes may have a pleiotropic effect on depressive symptoms, with TNFR2 highly correlating with an anhedonic response. If inhibition of soluble TNF pushed signaling toward TNFR2, an increase in anhedonia would support this conclusion.

NF-κb signaling with XPro1595 administration

Simple WES analysis of pP65, a downstream effector of the NF-κb pathway, suggested continued cytokine signaling with XPro1595 treatment. These results support the behavioral data from the s.c. XPro animals, demonstrating a much higher level of depressive behavior than the other clusters. Although TNFR1 signaling is a major upstream regulator of P65 phosphorylation, a variety of other inflammatory cytokines and immune receptors can activate this pathway (reviewed by Liu and colleagues⁶³). Therefore, remaining pP65 levels with soluble TNF inhibition could have several interpretations.

It is possible that the dosage of XPro1595 was not high enough to completely eliminate TNFR1 signaling. This is unlikely in the peripheral administration paradigm given the results from Brambilla and associates²⁸ demonstrating efficacy of the drug at the dosage administered in an EAE model. A preliminary study of five rats receiving s.c. XPro1595 every day to assess potential dose response did not differ greatly from the rats receiving a dose every third day. These results would indicate that s.c. dosage was not the source of depressive exacerbation or the lack of a therapeutic response. Given the small sample size, however, additional experiments should be conducted to make conclusive claims on s.c. dose response. An interesting finding was the significantly higher levels of pP65 in the dorsal raphe of animals treated with s.c. XPro compared with i.c.v. XPro. These results suggest that the exacerbation of depressive phenotype with s.c. treatment may have been due to increased cytokine signaling that is not present after i.c.v. administration of the drug. The data also suggest that i.c.v. XPro delivery had an overall anti-inflammatory effect that may not have been sufficient to cause a behavioral change. The XPro1595 dosage administered via i.c.v. was lower than that administered by Novrup and colleagues³⁰ to avoid lesion effects, which could complicate behavioral interpretation. It is possible that a higher i.c.v. dosage of XPro would more effectively prevent TNF signaling and the subsequent development of depression after SCI.

Our pP65 data could indicate continued signaling of cytokines other than TNF. IL-17, a cytokine that is released by Th17 cells, can signal through the NF-κb pathway (reviewed by Ouyang and colleagues⁶⁴). This cytokine also works synergistically with TNF to elicit inflammatory responses.^64–67 It is possible that signaling from cytokines such as IL-1β and IL-6 resulted in sustained pP65 levels even with decreased TNFR1 signaling. An alternative approach to our TNF targeted anti-inflammatory therapy could be use of a broader pharmacological inhibitor. Minocycline, a tetracycline compound that has been shown to decrease microglial activation and the production of cytokines such as IL-1β, had neuroprotective effects when used in a rat model of spinal cord transection.^68,69 Stirling and co-workers⁶⁸ demonstrated a significantly lower number of macrophages and microglia in the spinal cord, even distant to the lesion site. Thus, it is possible that the broad cellular and anti-inflammatory effects of minocycline could be more effective at limiting the development of depression after SCI.⁶⁹

Summary

Utilizing a moderate midline T9 contusion injury, we established a reliable depressive phenotype in greater than 50% of female rats at 4 weeks post-injury. There was a positive correlation with chronic increase in TNF levels in the DRN after thoracic SCI, yet targeting this increase was not effective in reducing the depressive phenotype. Rather, the incidence of depression increased significantly with peripheral but not i.c.v. administration of XPro1595.

Future experiments will utilize whole cell patch clamp electrophysiological recording to directly measure differences in activity of serotonergic and GABAergic DRN neurons in SCI-induced depression. Identifying neuronal changes after injury could help us elucidate both the mechanism of SCI-induced depression, but also potential therapeutic targets. This model and the associated neuroinflammation could be useful for future development of genetic, pharmacological, and/or exercise-based therapeutic strategies tailored to this specific patient population.

Author Disclosure Statement

No competing financial interests exist