Distinct cellular mediators drive the Janus Faces of Toll-like Receptor 4 regulation of network excitability which impacts working memory performance after brain Injury

Abstract

The mechanisms by which the neurophysiological and inflammatory responses to brain injury contribute to memory impairments are not fully understood. Recently, we reported that the innate immune receptor, toll-like receptor 4 (TLR4) enhances AMPA receptor (AMPAR) currents and excitability in the dentate gyrus after fluid percussion brain injury (FPI) while limiting excitability in controls. Here, we examine the cellular mediators underlying TLR4 regulation of dentate excitability and its impact on memory performance. In ex vivo slices, astrocytic and microglial metabolic inhibitors selectively abolished TLR4 antagonist modulation of excitability in controls, but not in rats after FPI, demonstrating that glial signaling contributes to TLR4 regulation of excitability in controls. In glia-depleted neuronal cultures from naïve mice, TLR4 ligands bidirectionally modulated AMPAR charge transfer consistent with neuronal TLR4 regulation of excitability, as observed after brain injury. In vivo TLR4 antagonism reduced early post-injury increases in mediators of MyD88-dependent and independent TLR4 signaling without altering expression in controls. Blocking TNFα, a downstream effector of TLR4, mimicked effects of TLR4 antagonist and occluded TLR4 agonist modulation of excitability in slices from both control and FPI rats. Functionally, transiently blocking TLR4 in vivo improved impairments in working memory observed one week and one month after FPI, while the same treatment impaired memory function in uninjured controls. Together these data identify that distinct cellular signaling mechanisms converge on TNFα to mediate TLR4 modulation of network excitability in the uninjured and injured brain and demonstrate a role for TLR4 in regulation of working memory function.

1.1. Introduction

A growing number of studies suggest that immune signaling can modulate nervous system function and plasticity (Pribiag and Stellwagen, 2014; van Vliet et al., 2018). Activation of glial inflammatory signals, including TNFα and purinergic receptors, have been shown to impact synaptic plasticity and hippocampal memory function (Beattie et al., 2002; Belarbi et al., 2012; Pascual et al., 2012; Pribiag and Stellwagen, 2014). Increases in neuronal excitability and immune activation are hallmarks of traumatic brain injury and present a condition in which neuroimmune interactions are potentially accentuated (Chiu et al., 2016; Neuberger et al., 2017a). While post-traumatic inflammatory responses are implicated in neurodegeneration, and immunosuppressants can improve neurological outcomes after brain injury (Saletti et al., 2019), mechanisms underlying immune regulation of neuronal function and behaviors are not fully understood. Among the immune pathways activated by brain injury, Toll-like receptor 4 (TLR4), an innate immune Pattern Recognition Receptor, has been shown to contribute to sterile inflammatory responses (Laird et al., 2014; Vezzani et al., 2011; Zhu et al., 2014). Although TLR4 is traditionally recruited in defense against microbial pathogens, it can be activated by endogenous ligands released from injured tissue and dying cells, making it a critical link between brain injury and the ensuing inflammatory response (Kielian, 2006). While TLR4 upregulation after brain injury is well documented (Ahmad et al., 2013; Laird et al., 2014; Ye et al., 2014), we recently demonstrated a preferential localization of TLR4 in hippocampal dentate granule cell layer and hilar neurons rather than in glia (Li et al., 2015). Our results revealed that TLR4 signaling reduced excitability in ex vivo slices from sham injured animals while enhancing excitability after brain injury (Li et al., 2015). Moreover, we recently reported that brief in vivo antagonism of TLR4 reduced posttraumatic increases in seizure susceptibility one to three months after injury, while augmenting risk for seizures in controls (Korgaonkar et al., 2020), indicating that brain injury and changes in TLR4 signaling have lasting effects on dentate network excitability. The potential contribution of glial signaling and behavioral consequences of this differential TLR4 regulation of dentate network excitability are currently unknown.

TLR4 is known to be expressed in both neurons and glia (Okun et al., 2011). However, a majority of the established functional effects of TLR4 in the CNS involve glial signaling and little is known about the role of neuronal TLR4. Consequently, signaling downstream of TLR4 has been elucidated primarily in glia. Activation of microglial and astrocytic TLR4 is known to engage two distinct signaling cascades, the pathway dependent on the Myeloid Differentiation factor 88 (MyD88) and the MyD88-independent pathway which activates the TIR-domain-containing adapter-inducing interferon-β (TRIF) (Kawai and Akira, 2007). Recruitment of the MyD88-dependent pathway in astrocytes and microglia leads to transcription of nuclear factor-kappa B (NF-κB) and production of the pro-inflammatory cytokine Tumor Necrosis Factor α (TNFα). Glia-derived TNFα has been shown to activate neuronal TNF receptor 1 (TNFR1) and mediate increases in NMDA receptor-dependent synaptic plasticity and AMPA receptor (AMPAR) surface expression (Stellwagen et al., 2005). The MyD88-independent pathway involves the recruitment of TRIF to activate Interferon regulatory factor 3 (IRF3), which in turn causes the production of interferon-β. Whether these or other signaling pathways are recruited by neuronal TLR4 remains unknown.

A second aspect of interest concerns the functional consequences of TLR4 regulation in hippocampal dentate excitability. Sparse granule cell activity is critical for dentate spatial working memory function (Dengler and Coulter, 2016). Conditions that enhance dentate excitability or impair inhibition compromise memory function (Kahn et al., 2019). Indeed, brain injury leads to both increased dentate excitability and impairments in memory performance (Gupta et al., 2012; Hamm et al., 1996; Santhakumar et al., 2001; Semple et al., 2018). Several lines of evidence suggest a role for TLR4 in hippocampal memory processing. Mice lacking TLR4 show enhanced memory processing and mutations that alter TLR4 signaling impact hippocampal long-term potentiation and spatial reference memory (Costello et al., 2011; Okun et al., 2012). Changes in hippocampal memory processing in mice lacking TLR4 have been associated with increases in cells expressing the immediate early gene Arc, which is consistent with enhanced excitability. These findings suggest that TLR4 modulation of dentate excitability could impact memory performance (Jeltsch et al., 2001; Xavier et al., 1999). The current study was conducted to determine the signaling mechanisms underlying TLR4 regulation of network excitability in the injured brain and its impact on behavioral outcomes.

2. Materials and Methods

All procedures were performed under protocols approved by the Institutional Animal Care and Use Committee of the Rutgers New Jersey Medical School, Newark, New Jersey and University of California, Riverside, California and are consistent with the ARRIVE guidelines.

2.1. Fluid percussion injury:

Juvenile male Wistar rats (25–27 days old) were subject to moderate (2–2.2 atm) lateral fluid percussion injury (FPI) or sham-injury using standard methods (Li et al., 2015; Neuberger et al., 2017b). Studies were restricted to males to avoid confounds due to cyclical changes in immune responses and behavioral effects of injury in females (Potter et al., 2019; Roof and Hall, 2000). Briefly, under ketamine (80mg/kg)-xylazine (10mg/kg) anesthesia (i.p.), rats underwent stereotaxic craniotomy (3mm dia, −3mm bregma, and 3.5mm lateral to sagittal suture) to expose the dura and bond a Luer-Lock syringe hub. The next day, randomly selected animals received a brief (20ms) impact on the intact dura using the FPI device (Virginia Commonwealth University, VA) under isoflurane anesthesia. Sham rats underwent all procedures except delivery of the pressure wave. Animals with implants dislodged during impact and those with <10 sec apnea following injury or injuries in which the pressure waveforms were jagged were excluded.

2.2. Field Electrophysiology:

One week after FPI or sham-injury rats were euthanized under isoflurane and horizontal brain slices (400μm for field recordings and 300 μm for patch recordings) were prepared in ice-cold sucrose-artificial cerebrospinal fluid (ACSF) containing (in mM) 85 NaCl, 75 sucrose, 24 NaHCO₃, 25 glucose, 4 MgCl₂, 2.5 KCl, 1.25 NaH₂PO₄, and 0.5 CaCl₂ (Li et al., 2015; Yu et al., 2016). Recordings were restricted to the side of injury (ipsilateral).

Field recordings were obtained in an interface recording chamber (BSC2, Automate Scientific, Berkeley, CA) perfused with ACSF containing (in mM) 126 NaCl, 2.5 KCl, 2 CaCl₂, 2 MgCl₂, 1.25 NaH₂PO₄, 26 NaHCO₃ and 10 D-glucose at 32–33°C. Granule cell population responses were recorded using patch pipettes with ACSF, in response to stimuli delivered through bipolar tungsten stimulating electrodes in the perforant path. Population spike amplitude was measured as the amplitude of the first negative deflection overriding the field EPSP waveform, as described previously (Neuberger et al., 2014). In all drug incubation experiments, pre and post drug recordings were conducted in the same slices with a 45 min incubation between recordings. Slices were recorded in control ACSF and transferred to incubation chambers containing ACSF or one of the following drug cocktails (a) CLI-095, (b) CLI-095 + fluoroacetate + minocycline (c) fluoroacetate + minocycline (d) HMGB1 (e) anti-TNFα or (f) HMGB1 + anti-TNFα for 45 minutes before returning slices back to the recording chamber. The drug were used at the following concentrations: CLI-095, 10ng/ml; fluoroacetate, 1mM; minocycline, 50nM; HMGB1, 10 ng/ml and anti-TNFα, 1μg/ml. The drugs and their concentrations are listed in Supplemental Table 1. Electrode positions were maintained during pre and post incubation recordings. Control incubations in ACSF between recordings confirmed that perforant path-evoked dentate population spike amplitude was stable during experimental manipulations (Suppl. Fig. 1). All electrophysiology data were low pass filtered at 3 kHz, digitized using DigiData 1440A, acquired using pClamp10 at 10 kHz sampling frequency, and analyzed using ClampFit 10. In experiments in which the percent change in population spike was analyzed, data from recordings in which no population spikes were observed in ACSF were excluded from analysis.

For whole cell patch clamp recordings, slices were visualized using infrared differential interference contrast techniques (IR-DIC) with a Nikon Eclipse FN-1 microscope and a 40X water-immersion objective. A MultiClamp 700B amplifier (Molecular Devices) was used to perform voltage clamp recordings from granule cells at 32–33°C using borosilicate microelectrodes (4–6 MΩ) containing (in mM) 140 Cs-Methanesulfonate, 10 HEPES, 2 MgCl2, 0.2 EGTA, 2 Na-ATP, 0.5 Na-GTP, 10 phosphocreatine and 0.2% biocytin. Currents were evoked using a tungsten stimulating electrode located in the perforant path. Afferent evoked excitatory postsynaptic currents (eEPSCs) were recorded at −60 mV and data from slices incubated in ACSF or drugs were compared. Data were low pass filtered at 3 kHz, digitized using DigiData 1440A, and acquired using pClamp10 at 10 kHz sampling frequency. Recordings were rejected if the cell had more than a 20% change in series resistance over time. ClampFit 10 and Sigma Plot 12.3 were used to analyze the data.

2.3. Neuronal Cultures:

Primary hippocampal neurons were obtained from the hippocampi of E17–18 C57Bl/6J mouse embryos of both sexes. Hippocampi were transferred to 15ml tissue culture tubes and the volume was adjusted to 10ml with Hanks balanced salt solution (HBSS) and were then gently centrifuged at 1000 rpm for 1 min. HBSS was then removed from the tube and tissue was digested in 5ml of 0.25% trypsin and 75μl of DNase (200Units/mg) solution for 20 min at 37ºC. Trypsin/DNase mix was then removed from tissue and the tissue was washed with characterized fetal bovine serum. Tissue was transferred to Neurobasal plating media-A (NB-A) and mechanically dissociated by trituration with a Pasteur pipet. Cells were pelleted by centrifugation at 1500 rpm for 5 min. Cells were taken up in 20μl of NB-A plating medium and counted in a hemocytometer. Approximately 0.5–1.0 × 10⁶ cells per well were plated on Poly-D-Lysine (PDL) coated 15mm coverslips in NB-A medium containing B27 and L-glutamine and cultured at 37°C in 5% C0₂/95% air. Medium was changed every 3–4 days. Ara-C was added to media on day 3 for complete removal of glia. Cultured coverslips were stained for GFAP and IBA-1 for confirmation of complete absence of glia. 11-to 13-day old cultures were used for physiological recordings. Recordings were obtained from independent culture plates obtained from embryos from 3–4 pregnant dams.

Cultured neurons were visualized under IR-DIC using a Nikon Eclipse FN-1 microscope and a 40X, 0.8 NA water-immersion objective. Whole cell recordings were obtained using MultiClamp 700B (Molecular Devices) at 32–33°C. Voltage clamp recordings of AMPAR currents were performed using borosilicate microelectrodes (4–6 MΩ) containing (in mM) 140 Cs-gluconate, 10 HEPES, 2 MgCl2, 0.2 EGTA, 2 Na-ATP, 0.5 Na-GTP, 10 phosphocreatine and 0.2% biocytin in the presence of SR95531 (10μM) to block GABA_A receptors and D-APV (50μM) to block NMDA receptors. Recordings were rejected if the cell had more than a 20% change in series resistance over time.

2.4. Drug administration:

One day after injury, a randomly assigned cohort of FPI and sham rats received either vehicle (saline) or a synthetic TLR4 antagonist, CLI-095 (0.5mg/kg, s.c.) for 3 doses at 8–12 hour intervals starting 8–12. after injury. A second group underwent stereotaxic injection (Hamilton syringe-26 G) of 5μl of saline or the highly selective TLR4 antagonist LPS-RS Ultrapure (LPS-RSU, 2mg/ml) in the hippocampus on the side of injury. Injections were delivered through the implanted syringe hub (AP: 2.5–3.0mm, ML: 2.5–3.5 mm, DV: 3–3.5mm). Drugs were delivered one day after injury at a rate of 1μl/5 mins under isoflurane anesthesia (2% in 95% O₂, 5% CO₂) delivered through a nose cone.

2.5. Western blotting:

Western blots of protein from fresh hippocampal tissue were obtained from rats perfused with cold ACSF (4°C) 3 days and 1 week after FPI or sham-injury as described previously (Li et al., 2015) using antibodies listed in Table 1. Protein concentration of the sample lysates was measured using BCA assay (Santa Cruz). Equal amounts of protein samples were diluted at a ratio of 1:1 in Laemmli sample buffer (Sigma) and separated on pre-cast gel (4–12% Tris–glycineBio-Rad). Chemiluminescent detection was performed with ECL western blotting detection reagent (Westdura, Thermo Scientific) using FluoroChem 8800. Densitometric quantification was determined using Image-J software (NIH) and normalized to β-actin density.

Table 1:

Antibodies

Antibody	Molecular	Type Weight	Dilution	Company	Catalog Number	Figure #
MAP2		Chicken Polyclonal	1:1000	AbCam	ab5392	Figure 3
GFAP		Mouse monoclonal	1:1000	Millipore	MAB360	Data not shown
IBA-1		Mouse monoclonal	1:1000	Millipore	MABN92	Data not shown
β-actin	42kDa	Mouse Monoclonal	1:5000	Sigma-Aldrich	A1978	Figure 4
HMGB1	29kDa	Rabbit Polyclonal	1:1000	Abcam	AB18256	Figure 4
MyD88	35kDa	Rabbit Polyclonal	1:1000	Abcam	Ab2064	Figure 4
IKB	40kDa	Mouse Monoclonal	1:500	Novus	6A920 (NB100-56507)	Figure 4
NFkB	60kDa	Rabbit Polyclonal	1:1000	Abcam	Ab16502	Figure 4
TICAM2	32kDa	Rabbit Polyclonal	1:500	Millipore	ABF235	Figure 4
IRF3	47kDa	Rabbit Polyclonal	1:1000	Abcam	Ab25950	Figure 4
TNFa	25kDa	Rabbit Polyclonal	1:1000	Abcam	Ab9635	Figure 4

2.6. Spatial Working Memory test:

A delayed match to position working memory task was conducted using a Morris Water Maze (Pang et al., 2015). Rats were trained to locate a hidden escape platform (10 × 10cm) located below the water surface in a pool (1.5m diameter). The working memory procedure consisted of 6 trials in one day; each trial had a sample and a choice phase. During the sample phase, rats began from a predetermined location in a zone without the platform and had 60 sec to locate a randomly placed escape platform in the pool. Rats which failed to locate the platform were manually led to the platform. The choice phase commenced 60 seconds after the sample phase and was identical in procedure to the sample phase. Rats spent a minimum of 30 minutes in a holding cage between trials. In each session, the start and the escape platform locations were distributed to equally throughout the pool (3 zones X 2 trials each for a total of 6 trials). Swim paths were recorded for offline analysis of path efficiency (defined as ratio of the straight-line distance between the start and the escape platform location and total distance traveled by a rat; a value of 1 indicates the most efficient path) using ANYmaze software. Rats were tested for spatial working memory performance prior to injury, rats with matched pre-injury path efficiency scores were assigned to either sham or FPI groups such that the average pre-injury path efficiency scores were not different between rats receiving sham or FPI. Rats in each stratum were randomly assigned to an injury group (Sham or FPI) and treatment (saline, LPS-RSU, CLI-095) groups. Rats had one session of testing at 1 week (early phase) and 1 month and/or 3 months (late phase) after injury.

2.7. Statistical analysis:

Statistical analyses were performed using GraphPad Prism 8. A total of 140 rats and 6 mice (pregnant dams) were used in the study. All independent samples were tested for normality and homogeneity of variance in GraphPad Prism 8 using descriptive statistics from Levene’s test. Post-hoc Tukey’s test was used to assess statistical significance of between-group differences. In behavioral studies (Fig. 7 and 8), data from sample and choice phases were analyzed separately using mixed-design ANOVA or two-way repeated measures ANOVA with time as repeated measure variable. Appropriate tests were selected depending upon the samples and their distribution for each experiment and on consultation with the Rutgers Biostatistics core. The significance level was set to p < 0.05. Data are shown as mean ± s.e.m. All data and statistical results are presented in Supplemental Tables 2 and 3 respectively.

Figure 7. Brain injury impairs working memory function.

(A). Schematic of experimental design shows timeline for working memory testing pre and post injury using Morris Water Maze. (B) Representative heatmaps of time spent at each location as the rat traced the path to the platform during sample and choice phase of a delayed match to position task in sham and FPI rats 1 week, after injury. Arrows indicate insertion point in pool. (C) Summary of path efficiency during sample and choice phases, *p<0.05 by mixed design ANOVA followed by pairwise Tukey’s test.

Figure 8. TLR4 antagonist treatment early after FPI improves working memory function.

(A) Schematic of experimental design shows timeline for working memory testing pre and post injury using Morris Water Maze. (B) Summary of path efficiency during sample and choice phases of a delayed match to position task in vehicle or CLI-095 treated (systemic) sham and preinjury, 1 week and 1 month after FPI. (C) Summary of path efficiency during sample and choice phases of a delayed match to position task in vehicle or LPS-RSU treated (systemic) sham and pre-injury, 1 week and 1 month after FPI. *indicates p<0.05 compared to sham and # indicates p<0.05 compared to corresponding vehicle treatment by TW-RM ANOVA followed by pairwise Tukey’s test, n= 9 rats each group.

3.0. Results

3.1. Differential contribution of glial signaling to TLR4 modulation of excitability in sham and injured rats

The sterile inflammatory response after brain injury is known to enhance TLR4 signaling in the hippocampus (Ahmad et al., 2013; Laird et al., 2014). Brain injury results in dentate hilar neuronal loss and increases in expression of microglia and reactive astrocytes (Gupta et al., 2012; Li et al., 2015; Neuberger et al., 2017a). Our earlier studies identified that TLR4 is expressed in neurons in the hippocampal dentate gyrus and has divergent effects on hippocampal dentate excitability; reducing excitability in controls while enhancing excitability early after brain injury (Korgaonkar et al., 2020; Li et al., 2015). However, prior studies have attributed TLR4 regulation of glutamatergic currents to recruitment of glial TLR4 signaling (Balosso et al., 2014). To specifically test the contribution of glial signaling to TLR4 modulation of dentate excitability in sham and FPI rats, we adopted a pharmacological cocktail of fluoroacetate (1mM) and minocycline (50nM) to acutely suppress astrocytes and microglia respectively. Fluoroacetate and its derivative flurocitrate are preferentially taken up by astrocytes and inhibit their metabolic function by blocking a critical Krebs cycle enzyme (Paulsen et al., 1987; Swanson and Graham, 1994). Minocycline, a tetracycline antibiotic and a potent inhibitor of microglial activation, has been shown to effectively suppress microglial TLR4 signaling (Li et al., 2014; Yrjanheikki et al., 1999; Zhou et al., 2006). If glial signaling contributes to the effects of TLR4 on dentate excitability a cocktail of fluoroacetate (1mM) and minocycline (50nM), referred hereafter as glial metabolic inhibitors (GMI), should occlude TLR4 modulation of dentate afferent evoked excitability. Alternatively, if neuronal TLR4 is solely responsible for modulating dentate excitability, glial metabolic inhibitors should not alter the TLR4 modulation. As in prior studies, we used two mechanistically distinct TLR4 antagonists CLI-095 (also known as TAK-242/Resatorvid, 10ng/ml) and LPS-RS Ultrapure (LPS-RSU, 50 μg/ml) which have similar effects on dentate excitability (Korgaonkar et al., 2020; Li et al., 2015). CLI-095 is a small molecule inhibitor of TLR4 signaling that acts on the intracellular domain to block both ligand dependent and independent TLR4 signaling (Matsunaga et al., 2011) and is approved for clinical trial in sepsis. LPS-RSU is a competitive TLR4 antagonist acting on the extracellular binding site. As reported previously, CLI-095 increased granule cell population spike amplitude in slices from rats one week after sham injury (Fig. 1A, Note: the CLI-095 data, previously reported in Korgaonkar et al., 2020, are shown for comparison). In contrast, CLI095 failed to alter granule cell population spike amplitude when the slices from sham rats were incubated for 45 minutes in GMI (Fig 1B, summarized in sham plots in Fig.1E, population spike amplitude in mV in response to a 4mA stimulation, sham in ACSF: 0.39 ± 0.08, n=8 slices, 3 rats, and sham incubated in CLI-095 and GMI: 0.28±0.06, n=8 slices, 3 rats, p>0.05 by two-way ANOVA followed by post-hoc Tukey’s test; sham-CLI095: 1.04±0.17, n=5 slices, 3 rats, reported in Korgaonkar et. al., 2020). Similarly, LPS-RSU did not increase granule cell population spike amplitude in the presence of GMI (population spike amplitude in mV in response to a 4mA stimulation, sham in ACSF: 0.33 ± 0.05; sham in LPS-RSU and GMI: 0.28±0.08, n=6 slices, 3 rats, p>0.05 by two-way ANOVA followed by post-hoc Tukey’s test). These data indicate a critical role for glia in CLI-095 and LPS-RSU mediated increases in population spike amplitude in sham controls (Fig. 1F, sham plots).

Figure 1. Differential role for glial signaling in TLR4 modulation of dentate excitability in sham and injured rats.

(A-D). Granule cell population responses evoked by a 4mA stimulus to the perforant path in slices from sham rats before and after CLI-095 (10ng/ml) in A and before and after incubation in CLI-095+Fluororoacetate (1mM)+Minocycline (50nM) in B. Responses in slices from FPI rats before and after CLI-095 (10ng/ml) in C and before and after incubation in CLI-095+Fluororoacetate (1mM)+Minocycline (50nM) in D. Arrows indicate stimulus artifact. (E-F) Summary plots of population spike amplitude (E) and % change in population spike amplitude compared to corresponding ACSF treatment condition (F). * indicates p<0.05 compared to corresponding recordings in sham, # indicates p<0.05 and n.s. indicates p>0.05 for pairwise comparison with corresponding control drug incubation by TW-ANOVA followed by pairwise Tukey’s test. Note: CLI-095 data in panel E was previously reported in Korgaonkar et al., 2020 and included for comparison with effect of GMI.

One week after brain injury, CLI-095 reduced perforant path-evoked dentate population spike amplitude (Fig. 1C) indicating that TLR4 signaling enhances network excitability after FPI (Korgaonkar et al., 2020; Li et al., 2015). Unlike the findings in sham controls, the reduction in population spike amplitude with CLI-095 treatment persisted in the presence of GMI (Fig 1D and Fig. 1E FPI data, population spike amplitude in mV in response to a 4mA stimulation, FPI in ACSF: 1.646 ± 0.12, n=8 slices, 3 rats, and FPI in CLI-095 and GMI: 0.40±0.13, n=8 slices, 3 rats, p>0.05 by two-way ANOVA followed by post-hoc Tukey’s test; FPI-CLI095: 0.14±0.11, n=6 slices, 3 rats, reported in Korgaonkar et. al., 2020). Once again, effects of LPS-RSU paralleled those of CLI-095, with LPS-RSU reducing dentate population spike amplitude even in the presence of GMI (population spike amplitude in mV in response to a 4mA stimulation, FPI in ACSF: 1.61 ± 0.30; FPI in LPS-RSU and GMI: 0.38±0.14, n=9 slices, 3 rats, p>0.05 by two-way ANOVA followed by post-hoc Tukey’s test). These data indicate glial signaling is unlikely to mediate TLR4 enhancement of dentate excitability after brain injury (Fig. 1F, FPI plots).

Next we examined whether GMI could directly impact network excitability. While the population spike amplitude in slices from FPI rats was greater than in sham rats, treatment with GMI alone did not directly alter network excitability in slices from either sham or FPI rats (Fig. 2A–C, sham in ACSF: 0.29±0.05; sham in GMI: 0.23± 0.06, −18.33±12.11 % change in n=5 slices, 3 rats each; FPI in ACSF: 1.67 ± 0.13, and FPI in GMI: 1.68±0.12, 1.30±3.46% change, n=5 slices, 3 rats each; p<0.0001 for effect of injury F_(1,8)=101.8).

Figure 2. Effect of glial metabolic inhibitors on dentate excitability.

(A). Granule cell population responses evoked by a 4mA stimulus to the perforant path in slices from the various experimental conditions. Arrows indicate stimulus artifact. (B-C) Summary plots of population spike amplitude (B) and pairwise comparison of population spike amplitude compared to corresponding ACSF treatment condition (C). * indicates p<0.05 by TW-ANOVA followed by pairwise Tukey’s test. (D, E) Representative eEPSC traces from granule cells recorded following incubation (45 minute) in ACSF (above) or GMI (below). Responses were elicited by a 2mA stimulus to the perforant path in slices prepared one week after sham-injury (D) and FPI (E). Corresponding summary plots are presented in the panels to the right.

We previously reported that TLR4 signaling enhances granule cell AMPA currents after FPI (Li et al., 2015). To further verify whether glial signaling contributes to posttraumatic increase in granule cell EPSCs, we examine whether incubation in GMI altered perforant path evoked excitatory postsynaptic currents (eEPSCs) after FPI. Consistent with our prior data (Li et al., 2015), granule cell eEPSC amplitude was significantly increased one week after FPI (Amplitude in pA, sham-ACSF: 749.1±139.5, n=7 cells and FPI-acsf 1296±146.2, n=6 cells, p<0.05 by t-test). We previously demonstrated that TLR4 antagonist selectively reduced post-injury increases in perforant path evoked granule cell eEPSCs one week after FPI without altering eEPSCs in sham injured controls (Li et al., 2015). Granule cell eEPSC amplitude in slices from sham rats showed a small increase following incubation in GMI which failed to reach statistical significance (Fig. 2d, Amplitude in pA, sham-ACSF: 749.1±139.5, n=7 cells and sham-GMI: 1044±116.1, n=6 cells, p>0.05 by t-test). Importantly, in slices from FPI rats, eEPSC amplitude was not altered by incubation in GMI (FPI-ACSF: 1296±146.2, n=6 cells and FPI-GMI 920±180.7, n=8 cells, p>0.05 by t-test) indicating the lack of direct effect of GMI on granule cell eEPSCs amplitude after injury. These data further suggest that glial signaling does not underlie TLR4 regulation of eEPSC amplitude after FPI.

3.2. Neuronal signaling underlies TLR4 modulation of AMPAR currents

To further examine the role for neuronal signaling in TLR4 modulation of excitability we established pure neuronal cultures from wild-type embryonic mouse hippocampi (Fig. 3A). Cultured cells were confirmed to be purely neuronal by the presence of the microtubule associate protein (MAP2) and absence astrocytic (GFAP) or microglial (Iba1) markers (Fig. 3B). In neuronal cultures, stimulation of neuropil evoked synaptic AMPAR currents, which were enhanced in cultures treated with the TLR4 agonist HMGB1 (10ng/ml) (Fig. 3C-D, AMPAR charge transfer in pA.sec: ACSF: 152.3 ± 6.80, 7 cells from 4 different repeats, HMGB-1: 254.4± 15.57, 7 cells from 4 replicates, p<0.05 by one-way ANOVA followed by a Tukey’s multiple comparison test). The ability of TLR4 agonist to increase excitability is the opposite of what is observed in slices from control rats and similar to observations in ex vivo slices from the injured brain (Li et al., 2015). Similarly, when neuronal cultures in vitro were treated with the TLR4 antagonist LPS-RSU (1μg/ml) the AMPAR current charge transfer was reduced (Fig. 3C-D, AMPAR charge transfer in pA.sec: cultures incubated in LPS-RSU: 93.4± 13.75, 6 cells from 4 replicates, p<0.05 versus ACSF by one-way ANOVA followed by a Tukey’s multiple comparison test) as reported in granule cells in ex vivo slices from brain injured rats (Li et al., 2015). Note that, since LPS-RSU and CLI-095 have similar effects on AMPA currents in hippocampal slices (Korgaonkar et al., 2020; Li et al., 2015), we restricted the in vitro analysis to a single TLR4 antagonist. Once again, these data show that the response of cultured neurons to TLR4 ligands is similar to that observed in granule cells from slices one week after FPI, suggesting that the trauma of dissociation may confer an injury-like phenotype to TLR4 effects in neuronal cultures. The ability of TLR4 ligands to bidirectionally modulate AMPAR currents in the absence of glia, and in a direction similar to that observed in slices after brain injury, is consistent with our finding (Fig. 1) that glial TLR4 signaling is not central to TLR4-dependent enhancement of excitability after brain injury.

Figure 3. TLR4 modulation of AMPAR currents in hippocampal neurons in vitro.

(A). Schematic of experimental design shows timeline for preparation of cultures followed by drug treatments and whole cell recordings. (B). Example maximum intensity projection of confocal image stacks from a hippocampal neuronal culture plated at embryonic day 17 and stained for MAP2 at 12 days in vitro (DIV) to reveal neurites. (C). Overlay of sample AMPAR current traces recorded in response to neuropil stimulation in neuronal cultures. Neurons were held at −60 mV to obtain recordings in ACSF, TLR4 agonist HMGB1 and TLR4 antagonist LPS-RSU. (D). Summary plot of AMPAR current charge transfer. * and # indicates p<0.05 compared to ACSF by One Way ANOVA followed by post-hoc Tukey’s test.

3.3. TLR4 antagonist treatment in vivo reduces post-injury increases in inflammatory signaling downstream of TLR4

Earlier studies have shown that anti-inflammatory agents can limit cell death after brain trauma by suppressing the inflammatory responses mediated downstream of TLR4 signaling (Dong et al., 2011). Our data show that TLR4 signaling reduces excitability in shams and increases excitability after brain injury. Moreover, we recently demonstrated that blocking TLR4 signaling in vivo reduced posttraumatic increases in hippocampal cellular inflammation observed three days after FPI without altering the inflammatory milieu in controls (Korgaonkar et al., 2020). However, whether blocking TLR4 signaling causes opposite changes in signaling pathways downstream of TLR4 in the uninjured and injured brain, which could contribute to the differing effects TLR4 signaling on excitability in sham and FPI rats, is not known. To test this possibility, we examined western blots from hippocampal tissue ipsilateral to injury for the endogenous TLR4 ligand, HMGB1, and downstream signaling elements of the MyD88-dependent and independent pathways (See schematic in Supplementary Fig. 2). Fresh hippocampal tissue for western blot analysis was obtained from rats that had undergone sham or FPI followed by treatment with saline or CLI-095 treatment (0.5mg/kg, 3 doses at 8–12 hour intervals starting 8–12 hrs after injury). Since TLR4 expression peaks at 24 hours, this dosing strategy was based on reported biological availability of CLI-095 in rat (Jinno et al., 2011) and was designed to antagonize the brief increase in TLR4 which peaks 24 hours after injury. In a cohort of rats examined one week after FPI or sham injury, hippocampal expression of TLR4 was not different between groups (Sham saline: 1.00±0, Sham-CLI095: 1.04±0.10, FPI-Saline: 1.13±0.12 and FPI-CLI095: 0.946±0.083, n=4 each, no group differences by TW-ANOVA) which is consistent with the recovery of TLR4 levels back to those observed in controls one week after injury (Li et al., 2015). Therefore we decided to evaluate TLR4 signaling in hippocampal tissue 3 days after FPI, a time point at which we have reported a significant post traumatic increase in TLR4 expression which is suppressed by CLI-095 treatment in vivo (Korgaonkar et al., 2020; Li et al., 2015). In saline treated rats, hippocampal expression of HMGB1 was significantly enhanced three days after FPI compared to sham rats (Fig. 4A-B, p=0.02 by TW ANOVA test, followed by post-hoc pairwise Tukey’s test). While CLI-095 treatment failed to enhance HMGB1 levels in sham controls (p=0.99 for sham-saline vs. sham-CLI), HMGB1 levels in FPI rats treated with CLI-095 trended towards a decrease when compared to saline-treated FPI rats (p=0.065 for FPI-saline vs. FPI-CLI by pairwise Tukey’s test) and was not different from saline-treated sham rats. Next we examined the pathways downstream of MyD88-dependent TLR4 signaling: MyD88, IkBα, and NFκB (Schematic in Supplementary Fig. 2). Levels of the three proteins downstream of the MyD88-dependent TLR4 pathway were enhanced in saline-treated FPI rats compared to saline-treated sham rats. (Fig. 4C-F). Compared to saline, CLI-095 treatment reduced the level of all three proteins in FPI rats without altering their levels in sham (Fig. 4C-F, Supplementary Table 2–3). Similarly, proteins downstream of the MyD88-independent pathway, TICAM2 and IRF3 were also enhanced in saline-treated FPI rats than in sham controls (Fig. 4G-I; Supplementary Table 2–3). CLI-095 treatment returned TICAM2 and IRF3 levels in FPI rats back to levels comparable to sham controls without altering their levels in sham rats (Fig. 4G-I; Supplementary Table 2–3). Moreover, TNFα, which is downstream of both MyD88-dependent and independent pathways was also enhanced in saline-treated FPI rats and decreased to sham levels after CLI-095 treatment (Fig 4G, J; Supplementary Table 2–3). However, CLI-095 did not alter TNFα levels in sham rats (Fig 4G, J; Supplementary Table 2–3). Thus, in vivo CLI-095 treatment significantly reduced hippocampal expression of TLR4 effectors examined in the injured brain while the treatment failed to alter expression of any of the TLR4 effectors tested in sham rats. These findings confirm the ability of in vivo treatment with CLI-095 to reduce early injury-induced increases in TLR4 signaling in the hippocampus, possibly by reducing the activation glial and cellular inflammatory responses to brain injury (Korgaonkar et al., 2020), without altering inflammatory signaling in controls.

Figure 4. TLR4 antagonism in vivo suppresses increases in TLR4 signaling after brain injury.

Representative western blots of HMGB1 (A) MyD88, IkBα, and NFκB, (C) and, TICAM2, IRF3 and TNFα (G) in hippocampal samples from the injured side obtained 3 days after vehicle/CLI-095 treatment. Treatments were started 24 hours after injury. Corresponding β-actin bands are illustrated. (A, C and G) Summary histograms of expression of HMGB1 (B), MyD88 (D), IkBα (E), NFκB (F), TICAM2 (H), IRF3 (I) and TNFα (J), normalized to the expression levels in sham-vehicle treated controls. * indicates p<0.05 compared to sham and # indicates p<0.05 compared to corresponding ACSF by TW ANOVA followed by post-hoc Tukey’s test.

3.4. Role of TNF-α in TLR4 effects on network excitability

Since TNFα is a direct downstream effector of the MyD88-dependent pathway and is also recruited by MyD88-independent pathways (Akira and Takeda, 2004), we examined whether blocking basal TNFα signaling impacts network excitability in the absence of the TLR4 ligand. In ex vivo slices from sham rats, incubation in the TNFα antibody (anti-TNFα, 1μg/ml, 1hr) increased dentate population spike amplitude (Fig. 5A-B, population spike amplitude evoked by a 4mA stimulus, in mV, sham: before anti-TNFα 0.02±0.01; after anti-TNFα: 0.33±0.22, n=8 slices from 4 rats, p=0.028 by paired t-test). This increase in population spike amplitude in anti-TNFα is similar to what is observed in slices from sham injured rats treated with TLR4 antagonist (Fig. 1A and Li et al., 2015). In contrast to findings from sham rats, incubation in anti-TNFα significantly reduced perforant path-evoked population spike amplitude in slices from FPI rats (Fig. 5C-D, population spike amplitude in mV at 4mA stimulation, FPI before anti-TNFα: 1.43 ± 0.22, FPI after anti-TNFα incubation: 0.422 ± 0.19, n=14 slices from 5 rats, p<0.0001 by paired t-test). Blocking TNFα which is elevated after brain injury (Fig. 4J and Atkins et al., 2007; Sinha et al., 2017), reduced network excitability, consistent with the effect of TLR4 antagonist application in slices from rats after FPI.

Figure 5. Effect TNFα signaling on network excitability in sham and injured rats.

(A) Representative dentate population responses evoked by a 4mA stimulus to the perforant path in a slice from a sham rat before (above) and after (below) incubation in anti-TNFα (1μg/ml). Arrows indicate stimulus artifact. (B) Summary data of effect of anti-TNFα (1μg/ml) on perforant path-evoked granule cell population spike amplitude in slices from sham rats. (C) Example dentate population responses perforant path stimulation at 4 mA in slices from FPI rats before (above) and after (below) incubation in anti-TNFα (1μg/ml). Arrows indicate stimulus artifact. (D) Summary data of effect of anti-TNFα (1μg/ml) on perforant path-evoked granule cell population spike amplitude in slices from sham rats. * indicates p<0.05 by paired Student’s t-test.

Since the effects of anti-TNFα on excitability paralleled that of TLR4 antagonists, we examined the potential ability of anti-TNFα to occlude the effects of the TLR4 agonist HMGB1. We previously reported that HMGB1 (10 ng/ml) reduced dentate population spike amplitude in slices from sham rats (Fig 6A-B, Li et al., 2015, HMGB1 data from prior study were used to determine % change data in Fig. 6A-B). The ability of HMGB1 to reduce dentate population spike amplitude in slices from sham rats was decreased in the presence of anti-TNFα (1μg/ml) (Fig. 6A-B, % change in population spike amplitude compared to corresponding ACSF, sham-HMGB1: −55.01±9.45 and sham-HMGB1+anti-TNFα: − 27.81±2.59, n=8 slices from 3 rats, p=0.045 by Students t-test). In contrast to the increase in dentate population spike amplitude following HMGB1 treatment reported in slices from FPI rats (Fig 6C-D, Li et al., 2015, HMGB1 data from prior study were used to determine % change data in Fig. 6C-D), co-treatment with anti-TNFα and HMGB1 decreased dentate population spike amplitude in slices from FPI rats (Fig. 6c, % change in population spike amplitude compared to ACSF, FPI-HMGB1: 77.32±21.59 and FPI-HMGB1+anti-TNFα: − 47.51±3.89, n=11 slices from 4 rats, p=0.001 by Students t-test). These results suggests that, in spite of the opposite functional effects and differential neuro-glial involvement in TLR4 signaling in sham and FPI rats, TNFα signaling contributes to TLR4 regulation of excitability in both the uninjured brain and after brain injury.

Figure 6. Contribution of TNFα signaling toTLR4 effects on network excitability in sham and injured rats.

(A) Representative dentate population responses evoked by a 4mA stimulus to the perforant path in slices from sham rats before and after HMGB1 (10ng/ml) in upper panel and before and after incubation in HMGB1+ anti-TNFα (1μg/ml) in lower panel. Arrows indicate stimulus artifact. (B) Summary of % change in population spike amplitude in the drug(s) compared to corresponding ACSF treatment condition in slices from sham rats (C) Dentate population spike traces evoked by a 4mA stimulus to the perforant path in slices from FPI rats before and after HMGB1 in upper panel and before and after incubation in HMGB1+ anti-TNFα) in lower panel. (D) Summary of % change in population spike amplitude in the drug(s) compared to corresponding ACSF treatment condition in slices from sham rats. Error bars indicate s.e.m. * indicates p<0.05 by unpaired Student’s t-test. (E) Representative eEPSC traces from granule cells from FPI rats recorded following incubation (45 minute) in HMGB1 and GMI with and without anti-TNFα. Responses were elicited by a 2mA stimulus to the performant path in slices prepared one week. (F) Summary plot of granule cell eEPSC peak amplitude in rats one week after FPI under various recording conditions.

To further examine the involvement of glial signaling in TNFα modulation of granule cell eEPSCs after FPI, we incubated hippocampal slices from rats one week after FPI in GMI and HMGB1 (10 ng/ml, 45 min) and examined the effect of anti-TNFα on eEPSC amplitude. Even in the presence of GMI, anti-TNFα (1μg/ml) significantly reduced the peak amplitude of granule cell eEPSCs recorded in the presence of HMGB1(Fig. 6E-F). These findings support a role for neuronal TNFα in mediating TLR4 induced increases in eEPSC amplitude.

3.5. TLR4 modulation of network excitability alters working memory function

Recent studies have identified that optimal activity in the dentate is crucial for its role in memory processing and that both increases and decreases in excitability can impair memory function (Kahn et al., 2019). Brain injury is associated with early increase in dentate excitability (Gupta et al., 2012; Neuberger et al., 2017b) and is known to affect dentate and hippocampus-dependent memory function (D’Ambrosio et al., 1998; Hamm et al., 1996; Pang et al., 2015; Smith et al., 2015). Moreover, there is persistent enhancement in seizure susceptibility and development of epilepsy several months after FPI which is consistent with a chronic increase in network excitability (Kharatishvili et al., 2006; Korgaonkar et al., 2020; Neuberger et al., 2017b; Santhakumar et al., 2001). Interestingly, TLR4 antagonist treatment leads to opposing changes in early dentate excitability in injured and uninjured rats which influences long-term seizure susceptibility (Korgaonkar et al., 2020). Moreover, in vivo treatment with TLR4 antagonist induces hilar neuronal loss in sham injured rats while reducing loss of hilar neurons after FPI (Korgaonkar et al., 2020). Therefore, we reasoned that blocking TLR4 signaling would differentially impact memory processing in the uninjured and injured brain. We focused on a short-term spatial working memory task dependent on dentate function to examine whether FPI, and the associated network alterations, impairs memory processing. Rats were assessed for spatial working memory on a modified Morris Water Maze (MWM) task before injury, at 1 week (Early phase), and at 1 and 3 months (Late phase) after injury. Path efficiency (Fig. 7A, see methods), a measure independent of distances between start and escape platform locations and swimming speed of rats, was used to quantify performance (Pang et al., 2015). Rats with matched pre-injury path efficiency scores were assigned to either sham or FPI groups such that the average pre-injury path efficiency scores were not different between rats receiving sham or FPI. Path efficiency on the sample phase increased as training proceeded, but did not show any significant differences between FPI and sham rats (Fig. 7B-C, path efficiency, sham: pre:0.16±0.01; 1week: 0.166±0.02; 1month: 0.21±0.03; 3month: 0.25±0.03; FPI: pre: 0.17±0.01; 1 week: 0.151±0.02; 1 month: 0.22±0.03; 3 month: 0.21±0.02, n=8 rats each, F_(1,14)=0.026 and p=0.87 for effect of Injury; F_(3,14)=3.15 and p=0.057 for effect time and F_(3,14)=0.215 and p=0.8 for interaction by Mixed design ANOVA). Path efficiency during the choice phase after a 60 sec retention interval was not different between sham and FPI rats prior to injury (Fig. 7C, pre-injury path efficiency, sham: 0.310±0.04, FPI: 0.312±0.03, n=8 rats, p=0.96 by Mixed design ANOVA followed by post hoc Tukey’s test). However, injured rats demonstrated a significant reduction in path efficiency 1 week and 1 month (Fig. 7B-C, path efficiency, sham-1wk: 0.41±0.03; FPI-1wk: 0.183±0.028; sham-1mo: 0.40±0.04; FPI-1mo: 0.21±0.01, n=8 rats/group, p<0.001 for FPI vs. sham at same time point by Mixed design ANOVA followed by post hoc Tukey’s test) which recovered by 3 months (Fig. 7C. Sham-3mo; 0.364±0.05, n=4 rats; FPI-3mo 0.34±0.04, n=4 rats, p>0.99 by Mixed design ANOVA followed by post hoc Tukey’s test). These data identify a transient deficit in memory function 1 week and 1 month after FPI.

Next, we examined whether treatment with a TLR4 antagonist could improve deficits in working memory performance in brain injured rats. Based on the initial findings (Fig 7), we chose to focus on one week and one month after injury to test the effects of pharmacologic treatment (Schematic in Fig. 8A). As before, rats were matched and stratified within the four experimental groups based on pre-injury path efficiency scores. Path efficiency on the sample phase did not show any significant effect of injury/drug or time (Fig. 8B, n=9 rats per group, F_(3,32)=0.36 and p=0.78 for effect of Injury/Drug; F_(2,32)=0.3 and p=0.74 for effect time and F_(6,32)=0.23 and p=0.95 for interaction by TW RM ANOVA). In pre-injury trials, path efficiency during choice phase following a 60 sec retention interval was not different between groups (Fig. 8B, n=9 rats per group, p>0.99 by TW RM ANOVA followed by post-hoc pairwise Tukey’s test). As expected, saline-treated brain injured rats showed an impairment in path efficiency during choice phase compared to sham-saline rats one week and one month after FPI. Consistent with our prediction based on changes in dentate excitability, CLI-095-treatment improved path efficiency in FPI rats compared to saline treated counterparts both one week and one month after injury (Fig. 8B, FPI-saline: 1 week after injury 0.152±0.008 and FPI-CLI-095: 1 week after injury 0.41±0.03 n=9 rats, p<0.001 and FPI-saline: 1 month after injury 0.24±0.020, n=9 rats and FPI-CLI-095 1month after injury 0.49±0.014, n=9 rats, p<0.005 by TW-RM ANOVA followed by pairwise Tukey’s test). In contrast, sham rats treated with systemic CLI-095 showed a reduction in path efficiency compared to sham-saline rats one week and one month after sham-injury/treatment (Fig. 8B sham-saline: 1 week after injury 0.44±0.04 and Sham-CLI-095: 1 week after injury 0.332±0.007 n=9 rats, p=0.04 by post hoc Tukey’s test and sham-saline: 1 month after injury 0.60±0.029, n=9 rats and Sham-CLI-095 1 month after injury 0.349±0.03, n=9 rats, p<0.005 by TW-RM ANOVA followed by pairwise Tukey’s test). While the impairment in memory performance following TLR4 antagonist treatment in uninjured rats is surprising, it is consistent with the ability of CLI-095 to enhance network excitability which would impair dentate memory function.

3.6. Focal and transient hippocampal TLR4 antagonism alters working memory function

Systemic administration of CLI-095 can potentially alter central and peripheral immune responses, which raises the possibility that the suppression of global immune responses may underlie the efficacy of TLR4 modulation on neurobehavioral outcomes after FPI. To limit the potential contribution of systemic effects of TLR4 antagonism, we examined whether local, unilateral hippocampal injection of LPS-RSU (5μl of a 2mg/ml solution, intrahippocampal injection 24 hrs after FPI/sham) on the injured side would reduce the working memory impairment following FPI, as observed after systemic treatment. Similar to what was observed after systemic CLI-095 treatment, path efficiency on the sample phase and during pre-injury trials did not show any significant effect of injury/drug or time (Fig. 8C). Focal TLR4 antagonist treatment reduced path efficiency in uninjured controls at one month (Fig. 8C, sham-saline: 1 month after injury 0.57±0.034, n=9 rats and Sham-LPS-RSU 1 month after injury 0.338±0.04, n=9 rats, p=0.028 by TW-RM ANOVA followed by pairwise Tukey’s test). Additionally, there was a trend for the treatment to decrease path efficiency one week after sham injury which did not reach statistical significance (Fig. 8C sham-saline: 1 week after injury 0.36±0.033 and Sham-LPS-RSU: 1 week after injury 0.29±0.020 n=9 rats, p=0.074 by TW-RM ANOVA followed by pairwise Tukey’s test). In contrast, LPS-RSU-treatment improved path efficiency in brain injured FPI rats compared to saline treated counterparts both one week and one month after injury (Fig. 8C, FPI-saline: 1week after injury 0.162±0.008 and FPI-LPS-RSU: 1week after injury 0.426±0.04 n=9 rats, p<0.001 and FPI-saline: 1 month after injury 0.173±0.011, n=9 rats and FPI-LPS-RSU 1month after injury 0.482±0.015, n=9 rats, p<0.005 by TW-RM ANOVA followed by pairwise Tukey’s test). Thus, as with systemic CLI-095, focal LPS-RSU improved memory performance one week and one month after FPI while impairing path efficiency at the one-month time point in sham controls (Fig. 8C). Together, these data support our hypothesis that early changes in dentate excitability and TLR4 modulation of dentate excitability contribute to neurological deficits including impaired working memory performance after brain injury.

4.0. Discussion

4.1. Janus Faced effects of TLR4 signaling on excitability and memory function

This study identifies that distinct cellular mediators contribute to the divergent effects of TLR4 on dentate excitability in the normal and injured brain which have a lasting impact on behavioral outcomes. The data show that the ‘Janus faced’ effects of TLR4 on excitability are likely determined by differential recruitment of glial versus neuronal TLR4 signaling.

It has long been recognized that sparse firing is an essential functional feature of the dentate gyrus and its ability to ‘gate’ activity is critical for memory processing and for limiting pathology (Acsady and Kali, 2007; Dengler and Coulter, 2016; Kahn et al., 2019; Lothman et al., 1992). TLR4 activation enhances dentate network excitability after brain injury (Korgaonkar et al., 2020; Li et al., 2015) which would compromise the dentate gate. Consequently, TLR4 antagonists reduce dentate excitability after brain injury and improve working memory function in vivo. Simultaneously, TLR4 antagonist treatment in vivo reduces injury-induced increases in molecular signals associated with both the MyD88-dependent and MyD88-independant pathways effectively suppressing TLR4-dependent immune response early after brain injury. Following brain injury, TLR4 antagonists reduced dentate excitability in ex vivo slices even in the presence of glial metabolic inhibitors. Moreover, AMPAR currents in neuronal cultures depleted of glia were enhanced by HMGB1, a TLR4 agonist, and reduced by LPS-RSU, a selective TLR4 antagonist. These results using both pharmacological and neuronal culture systems strongly support a role for neuronal TLR4 signaling in augmenting excitability in the injured brain. Interestingly, blocking TNFα, a proinflammatory cytokine downstream of MyD88-dependent TLR4 signaling, reduced dentate excitability, reduced granule cell eEPSC amplitude even in the presence of GMI and occluded the ability of HMGB1 to increase excitability indicating that TLR4 signaling acts through TNFα to enhance excitability after brain injury.

In contrast, TLR4 signaling appears to reinforce the dentate gate in the uninjured brain, where TLR4 signaling limits dentate network excitability during afferent stimulation (Li et al., 2015). Consistently, blocking basal TLR4 signaling enhances dentate excitability in response to input activation in ex vivo slices and compromises working memory function in vivo, demonstrating that basal TLR4 signaling regulates memory processing. Distinct from the effects on excitability, in vivo treatment with TLR4 antagonist failed to alter effectors downstream of the MyD88-dependent and MyD88-indepandant pathways in controls, indicating that antagonists did not have paradoxical effects on inflammatory signaling. Unlike after injury, glial metabolic inhibitors blocked TLR4 antagonist enhancement of dentate excitability in slices from sham rats indicating involvement of glial signaling in TLR4 modulation of excitability in controls. Blocking TNFα tended to enhance excitability suggesting a role for basal TNFα signaling in limiting network excitability. Anti-TNFα reduced the ability of a TLR4 agonist to decrease excitability indicating that TNFα is downstream of TLR4 signaling in controls. These findings coupled with the ability of GMI to block effects of TLR4 modulation in sham rats suggest that TLR4 recruitment of signaling pathways in glia and glia-derived TNFα contribute to suppression of basal excitability in the uninjured dentate gyrus. Thus, the Janus faces of TLR4 signaling in the brain are distinguished by recruitment of glial metabolic processes and TNFα to limit excitability which supports working memory function in the uninjured brain. On the other hand, our data support a role for neuronal TLR4 and TNFα in increasing network excitability and contributing to deficits in working memory function after brain injury.

4.2. Immune signaling and memory function

Several lines of evidence point to multiple roles for molecules, once thought to be confined to the peripheral immune system, in regulating normal function in healthy brains and in neurological disorders (Pickering and O’Connor, 2007; Williamson and Bilbo, 2013). Basal levels of neuroimmune signaling through specific cytokines such as TNFα have been shown to contribute to synaptic plasticity and memory processing (Avital et al., 2003; Balschun et al., 2004; Ben Menachem-Zidon et al., 2011; Goshen et al., 2007). Studies in transgenic mice lacking specific immune receptors, including TLR4, have identified memory and cognitive changes underscoring the importance of neuroimmune interactions in shaping brain function (Okun et al., 2012; Potter et al., 2019). Since neuro-immune interplay is critical for the developmental establishment of circuits (Szepesi et al., 2018), use of germline knockouts confounds the ability to distinguish between immune regulation of circuit development and basal neuroimmune signaling in behaviors. Here we show that even a brief and transient block of basal TLR4 signaling, which increases network excitability, leads to prolonged impairments in working memory function for up to a month after treatment. Moreover, two mechanistically distinct TLR4 antagonists administered either focally in the hippocampus (LPS-RSU) or systemically (CLI-095) had similar effects in impairing working memory function, reducing the potential for off target effects. Curiously, an earlier study in mice reported impaired memory performance one week after a single treatment with the exogenous TLR4 agonist LPS and showed that TLR4 antagonist treatment abolished LPS induced memory impairments and suppressed the associated inflammatory response (Zhang et al., 2018). These findings contrast with what we would expect based on results demonstrating that TLR4 antagonist impairs memory function in uninjured rats (Fig. 8). However, it should be noted that LPS treatment in Zhang et al. (2018) enhanced inflammatory cytokines and TLR4 mediated signaling and may better reflect the neuroimmune environment present in the injured brain in our study.

We find that TLR4 antagonist treatment reduced network excitability, neuronal loss and seizure susceptibility after FPI (Korgaonkar et al., 2020) and improved working memory after brain injury. The behavioral outcomes paralleled the opposing effect of TLR4 ligands on excitability, rather than the effects on inflammatory signaling which were unchanged in controls. Moreover, TLR4 expression returned to control levels and was unchanged by antagonist treatment one week (current study) after FPI. Similarly both GFAP and TLR4 expression were not different between vehicle and drug treated sham and FPI groups one month after FPI (Korgaonkar et al., 2020). Thus, it is reasonable to propose that, while TLR4 antagonism after FPI reduces glial activation and may thereby reduce inflammatory signaling, the changes in excitability underlie the effects on memory function. Indeed, in controls, TLR4 antagonists enhanced excitability (Korgaonkar et al., 2020) which would be expected to compromise sparse neuronal activity and impair working memory performance (Kahn et al., 2019; Madar et al., 2019; Scharfman and Bernstein, 2015; Wixted et al., 2014). In contrast, when neuronal excitability and the glutamate receptor subtype GluA1 are increased following brain injury, TLR4 antagonists reduced excitability (Korgaonkar et al., 2020), potentially aiding in maintaining sparse neuronal activity, which improved working memory. These data suggest that there may be an optimal range for TLR4 signaling to maintain physiological levels of excitability; both blocking basal levels, as with TLR4 antagonism in controls, as well as enhanced TLR4 signaling, as occurs after injury, can be pathological. Our findings are in line with the ability of innate immune signals to regulate the immediate early gene Arc and reports that Arc expression both above and below the optimal range can impair synaptic plasticity and cognitive function (Pickering and O’Connor, 2007; Rosi, 2011). Curiously, anti-TNFα mimics the effects TLR4 antagonists in slices from both the control and injured brains. Moreover, TLR4 mediated decrease in excitability in controls and increase in excitability after FPI appear to involve TNFα signaling suggesting that there may be a physiological range for TNFα that regulates excitability and synaptic plasticity at optimal levels. Our findings are consistent with the ability of exogenous TNFα to modulate synaptic excitability and plasticity (Belarbi et al., 2012; Maggio and Vlachos, 2018; Stellwagen et al., 2005). However, HMGB1 effects on dentate excitability in controls were reduced but not eliminated in anti-TNFα. It is possible HMGB1 may recruit additional receptor and or downstream pathways to impact network excitability.

Traumatic brain injury is known to impair memory and cognitive function (Azouvi et al., 2017; Hamm et al., 1996; Lyeth et al., 1990; Titus et al., 2016; Vallat-Azouvi et al., 2007) and lead to persistent deficits in hippocampal long-term potentiation (Sanders et al., 2000). In earlier studies, mild lateral FPI (1.2 atm) was shown to contribute to transient working memory deficits which recovered by 3 weeks (Pang et al., 2015). At the moderate injury strength (2–2.2 atm) used in the current study working memory deficits persist 4 weeks after injury and recover by 3 months, which is consistent with increasing neuropathology with injury severity. There is considerable evidence for the ability of non-specific immune modulators to limit post-traumatic inflammatory responses including TLR4 expression and improve neurological outcomes (Mao et al., 2012; Wang et al., 2015; Yu et al., 2012). Our study using both local and systemic administration of specific TLR4 antagonists clearly demonstrates that suppression of TLR4 is beneficial following experimental brain injury in rats. The paradoxical detrimental effects of TLR4 antagonist treatment in the uninjured brain, interestingly, are coupled to both early and persistent increase in network excitability (Korgaonkar et al., 2020; Li et al., 2015) by mechanisms that are currently unknown. However, our data indicate that TLR4 does not modulate AMPA and NMDA receptor currents in slices from uninjured rats (Li et al., 2015), suggesting that modulation of GABA currents or intrinsic neuronal excitability need to be examined in future studies. Regardless of the underlying mechanisms, the effects of TLR4 antagonists on excitability and memory function in uninjured brain warrants caution while proposing to use TLR4 antagonists for therapeutics.

4.3. Post-injury switch in cellular mediators of TLR4 effects on excitability

We previously identified that post-traumatic increase in TLR4 expression is predominantly neuronal (Li et al., 2015). However, activation of glial purinergic signaling by TLR4 is known to acutely increase excitatory synaptic drive to CA1 neurons by modulating metabotropic glutamate receptors (Pascual et al., 2012). Additionally, LPS induced neurotoxicity in cortical cultures was found to be mediated through glia (Lehnardt et al., 2003). Moreover, glial TLR4 signaling through endogenous TLR4 ligands such as HMGB1, can enhance proinflammatory cytokines such as IL-1β and TNFα in immune cells including microglia and promote excitability (Maroso et al., 2010). While we acknowledge that fluoroacetate and minocycline may have nonspecific effects, our studies using complementary strategies of pharmacological glial metabolic inhibitors and isolated neuronal cultures, indicate that TLR4 modulation of AMPAR currents is independent of glia. It is possible that the lack of TLR4 in cortical neurons (Lehnardt et al., 2003) or use of LPS, which activates a broader immune response, contributes to the glial dependence of TLR4 signaling in prior studies. Future studies using cell specific deletion of TLR4 could further validate the role for neuronal TLR4 in modulating AMPA currents after brain injury. It is notable that our data in control slices demonstrate that glial signaling is necessary for maintaining basal levels of excitability and TLR4 modulation of excitability in the uninjured brain.

Our findings suggest two distinct cellular and signaling pathways for TLR4 function depending on whether there is injury to the brain. Classically, TLR4 signaling in glia is mediated by MyD88-TNFα dependent pathway and MyD88-independent pathway. TLR4 activation induces a MyD88-dependent increase in TNFα (Gao et al., 2009) and is proposed to enhance NMDA currents (Balosso et al., 2014; Maroso et al., 2010) and recruit glial purinergic receptors (Pascual et al., 2012). However, the post-traumatic increase in dentate excitability is mediated by enhancement of polysynaptic AMPAR currents with no change in NMDAR currents (Santhakumar et al., 2000). Consistent with this data, TLR4 enhances granule cell AMPAR, specifically, calcium permeable AMPARs (CP-AMPARs), and not NMDAR currents after FPI (Korgaonkar et al., 2020; Li et al., 2015). Interestingly, glial TNFα has been shown to recruit CP-AMPARs to the synapse (Pickering et al., 2005; Pribiag and Stellwagen, 2014; Stellwagen and Malenka, 2006). Since our results with glial metabolic inhibitors and neuronal cultures indicate that TLR4-mediated increases in excitability after injury does not involve glial signaling, it is possible that TLR4 activation after brain injury recruits “neuronal” TNF-α signaling. This would constitute a novel function for neuronal TNFα. TNFα is a likely candidate for the neuropathological effects of TLR4 after brain injury since it is known to contribute to excitotoxicity (Pickering et al., 2005; Stellwagen, 2011). Indeed, an inhibitor of TNFα synthesis has been shown to reverse cognitive deficits induced by chronic neuroinflammation (Belarbi et al., 2012). Unlike after injury, our data suggest that classical glial-TNFα signaling may reduce excitability in the uninjured hippocampus by mechanisms of which are currently under investigation.

4.4. Basal TLR4 modulation of network function and behaviors

A particularly interesting finding is the ability of transient TLR4 antagonism to increase excitability and impair working memory performance in control, uninjured animals. These data suggest a critical neurophysiological function for the low levels of TLR4 signaling observed in the uninjured brain. Moreover, the lack of change in immune signaling alongside changes in excitability observed after TLR4 antagonism in the controls suggests that non-immune responses underlie this effect. Yet, blocking glial signaling eliminated the ability of CLI-095 to increase excitability in controls suggesting that glia are necessary intermediaries for this process. We demonstrate changes in memory function following acute and transient modulation of TLR4. These results extend prior works demonstrating a developmental role for TLR4 in learning and memory function in mice with constitutive TLR4 knockout (Okun et al., 2012). It remains to be seen whether TLR4 can modulate glial glutamate uptake and glutamate-glutamine cycling in the brain as has been shown in spinal circuits during LPS challenge (Yan et al., 2015). Regardless, the basal role for TLR4 in modulating dentate excitability and its impact on memory function need to be considered while proposing TLR4 modulators for use in therapies (Gnjatic et al., 2010; Hanke and Kielian, 2011).

5.0. Conclusions

Our data indicate that TLR4 signaling in neurons contributes to pathological increase in excitability after brain injury by enhancing AMPAR currents through a mechanism that involves TNFα. Blocking TLR4 signaling early after brain injury reduces injury-induced decreases in hippocampal working memory function. However, we find that blocking basal TLR4 signaling in the uninjured brain augments dentate excitability through recruitment of glial signaling and TLR4 demonstrating a role for basal TLR4 signaling in maintaining sparse dentate activity. Thus, blocking the basal TLR4 signaling impairs working memory function. Taken together, the potential therapeutic effect of TLR4 antagonists may be better harnessed by isolating and selectively targeting the distinct pathological signaling downstream of neuronal TLR4.

Highlights.

• TLR4 suppresses dentate excitability in controls through signaling involving glia

• Neuronal TLR4 signaling underlies enhanced dentate excitability after brain injury

• TNFα contributes to TLR4 regulation of excitability in the injured brain

• Altering TLR4 signaling impacts working memory performance

• TLR4 signaling is a potential target to improve working memory after brain trauma