Abstract
Perioperative neurocognitive disorders (PNDs) are characterized by confusion, difficulty with executive function, and episodic memory impairment in the hours to months following a surgical procedure. Postoperative cognitive dysfunction (POCD) represents such impairments that last beyond 30 d postsurgery and is associated with increased risk of comorbidities, progression to dementia, and higher mortality. While it is clear that neuroinflammation plays a key role in PND development, what factors underlie shorter self-resolving versus persistent PNDs remains unclear. We have previously shown that postoperative morphine treatment extends POCD from 4 d (without morphine) to at least 8 weeks (with morphine) in aged male rats, and that this effect is likely dependent on the proinflammatory capabilities of morphine via activation of toll-like receptor 4 (TLR4). Here, we extend these findings to show that TLR4 blockade, using the selective TLR4 antagonist lipopolysaccharide from the bacterium Rhodobacter sphaeroides (LPS-RS Ultrapure), ameliorates morphine-induced POCD in aged male rats. Using either a single central preoperative treatment or a 1 week postoperative central treatment regimen, we demonstrate that TLR4 antagonism (1) prevents and reverses the long-term memory impairment associated with surgery and morphine treatment, (2) ameliorates morphine-induced dysregulation of the postsynaptic proteins postsynaptic density 95 and synaptopodin, (3) mitigates reductions in mature BDNF, and (4) prevents decreased activation of the BDNF receptor TrkB (tropomyosin-related kinase B), all at 4 weeks postsurgery. We also reveal that LPS-RS Ultrapure likely exerts its beneficial effects by preventing endogenous danger signal HMGB1 (high-mobility group box 1) from activating TLR4, rather than by blocking continuous activation by morphine or its metabolites. These findings suggest TLR4 as a promising therapeutic target to prevent or treat PNDs.
SIGNIFICANCE STATEMENT With humans living longer than ever, it is crucial that we identify mechanisms that contribute to aging-related vulnerability to cognitive impairment. Here, we show that the innate immune receptor toll-like receptor 4 (TLR4) is a key mediator of cognitive dysfunction in aged rodents following surgery and postoperative morphine treatment. Inhibition of TLR4 both prevented and reversed surgery plus morphine-associated memory impairment, dysregulation of synaptic elements, and reduced BDNF signaling. Together, these findings implicate TLR4 in the development of postoperative cognitive dysfunction, providing mechanistic insight and novel therapeutic targets for the treatment of cognitive impairments following immune challenges such as surgery in older individuals.
Introduction
Normal aging is associated with gradual declines in cognition. However, these declines can turn precipitous following insults such as surgery, infection, or injury that evoke an immune response (Moller et al., 1998; Barrientos et al., 2006; Jurgens et al., 2012; Sparkman et al., 2019; Muscat et al., 2021; Muscat and Barrientos, 2021). Perioperative neurocognitive disorders (PNDs) primarily affect older individuals and are characterized by confusion, difficulty with executive tasks, and episodic memory impairment in the hours to months, or even years, after a surgical procedure (Moller et al., 1998; Terrando et al., 2011; Evered et al., 2018; Mahanna-Gabrielli et al., 2019). PNDs are further classified based on symptom duration, with postoperative delirium representing shorter symptoms that resolve within 30 d of surgery, and postoperative cognitive dysfunction (POCD) being characterized by symptoms lasting beyond 30 d postsurgery (Evered et al., 2018). Notably, a longer-lasting PND is associated with increased risk of progression to dementia and mortality (Bickel et al., 2008; Mahanna-Gabrielli et al., 2019).
Preclinical models of PNDs, including our own, have established neuroinflammation as a critical mediator in the development of postsurgical cognitive impairments (Rosczyk et al., 2008; Barrientos et al., 2012; Wang et al., 2020a; Muscat et al., 2021). Importantly though, many of these preclinical models have largely failed to recapitulate the more persistent nature of POCD, with cognitive deficits lasting only ∼1 week. Clinically, >90% of surgical patients receive morphine or other opioid prescriptions for postsurgical pain management (Hill et al., 2017; Marcusa et al., 2017), and we recently demonstrated that postoperative morphine treatment extends postsurgical memory dysfunction from 4 d (without morphine) to at least 8 weeks (with morphine) in aged rodents (Muscat et al., 2021). The effects of morphine were likely mediated by its proinflammatory action via activation of toll-like receptor 4 (TLR4), as surgery plus morphine caused a robust neuroinflammatory response lasting at least 2 weeks in aged rats. However, the mechanisms between the proinflammatory effects of morphine and its ability to induce persistent postoperative cognitive dysfunction remain unclear.
TLR4 is a member of the toll-like receptor family, a group of innate immune pattern recognition receptors. Classically, TLR4 recognizes pathogen-associated molecular patterns, most notably lipopolysaccharides from Gram-negative bacteria, although a number of endogenous molecules has also been found to interact with it (Lu et al., 2008; Roy et al., 2016; Yang et al., 2020b). TLR4 can be activated directly or following ligation of its coreceptor MD-2, and stimulation results in a conformational change in and recruitment of intracellular toll-interleukin-1 receptor domains, which ultimately induce an inflammatory signaling cascade and the production of proinflammatory cytokines (for review, see Lu et al., 2008).
Given the role of neuroinflammation in morphine-induced persistent postsurgical memory deficits in aged rats, we speculated that morphine, surgery, and aging synergize via joint activation of a similar neuroimmune mechanism to induce exaggerated neuroinflammation sufficient to cause long-lasting cognitive impairment. Indeed, normal aging is associated with neuroinflammatory priming, or sensitization, including increased expression of TLR4 (Letiembre et al., 2007; Fonken et al., 2016; Wilhelm et al., 2017). Moreover, surgery has been shown to rapidly induce TLR4 expression in both spleen and hippocampus (Lu et al., 2015). Additionally, a growing body of literature has characterized the ability of morphine to induce TLR4 activation via its binding to MD-2 (Hutchinson et al., 2010; Zhang et al., 2020b; Gabr et al., 2021). Therefore, we chose to focus on a potential mediating role for TLR4 in the development of morphine-induced persistent POCD. In the present study, we used the TLR4-specific antagonist lipopolysaccharide from the bacterium Rhodobacter sphaeroides [LPS-RS Ultrapure (hereafter referred to simply as LPS-RS)] to block central TLR4 activation preoperatively and postoperatively in aged rats. A combination of behavioral, neurobiological, and neurochemical assays revealed a significant role for TLR4 in the development of persistent POCD, as TLR4 antagonism both prevented and reversed the dysregulation of synaptic plasticity and long-term memory impairment associated with the combination of aging, surgery, and morphine.
Materials and Methods
Experimental design.
This study was composed of five separate experiments that, for ease of reading, will be briefly summarized here. Specific experimental details will be described in the sections below. In experiment 1, rats received a surgical procedure (laparotomy) followed by 7 d of morphine treatment. Beginning the day after the last morphine injection, rats began 7 d of treatment with the specific TLR4 antagonist LPS-RS. For this experiment, three different doses of LPS-RS were used to assess for optimal dosage. We chose to begin with a postoperative treatment regimen because this is a clinically relevant time point. Many cases of perioperative cognitive dysfunction resolve within several days of surgery. Thus, patients or clinicians may not consider a need for treatment until cognitive symptoms persist beyond this time period. Importantly, we have previously shown that TLR4 and its endogenous ligand high-mobility group box 1 (HMGB1) remain elevated at least 2 weeks after surgery and morphine treatment, suggesting that inflammatory signaling continues even after morphine treatment ends (Muscat et al., 2021). Thus, the aim of this study was to determine whether TLR4 antagonism after symptom onset, but before neuroinflammation resolves, could reverse or improve persistent POCD-associated memory deficits. Four weeks after surgery (3 weeks after morphine treatment and 2 weeks after LPS-RS treatment) rats underwent contextual pre-exposure facilitation fear conditioning (CPF-FC) to evaluate long-term hippocampal-dependent memory function.
In experiment 2, rats received a single preoperative injection of LPS-RS (at the dose found to be effective in experiment 1) to determine the ability of TLR4 antagonism to prevent persistent POCD. Immediately following this injection, rats underwent laparotomy followed by 1 week of morphine treatment. As in experiment 1, long-term hippocampal-dependent memory function was assessed via CPF-FC 4 weeks postsurgery (3 weeks after the last morphine treatment).
Experiments 3 and 4 followed the same experimental setup as experiments 1 and 2, respectively. That is, experiment 3 consisted of laparotomy and 1 week of morphine treatment followed by 1 week of treatment with LPS-RS and experiment 4 involved a single preoperative administration of LPS-RS followed by laparotomy and 1 week of morphine treatment. Both experiments used LPS-RS at the dose found to be effective in experiment 1. In both experiments 3 and 4, rats underwent a learning experience 4 weeks after surgery. The purpose of the learning experience was to activate processes related to memory consolidation. Two hours after the learning experience (when memory consolidation is actively occurring), rats were killed and hippocampi and amygdala were dissected. Tissues were then processed for Western blot analysis of proteins involved in synaptic plasticity and memory consolidation. It should be noted that amygdala was used as an anatomic control, based on previous work indicating no effect of laparotomy on amygdala-dependent memory in our model (Barrientos et al., 2012).
To determine a potential mechanism by which TLR4 antagonism is able to elicit ameliorative and preventative effects in the present model of POCD, a separate experiment was performed. In experiment 5, rats underwent laparotomy followed by 1 week of morphine treatment, as before. A subset of subjects received either preoperative or postoperative LPS-RS treatment. Seventy-two hours after the last morphine injection, rats were killed and hippocampi were dissected. The rationale for this time point was twofold: first, it aligns with the midway point of LPS-RS administration in experiments 1 and 3; and second, up to 90% of the administered morphine dose is excreted by 72 h (Iwamoto and Klaassen, 1977; Saboory et al., 2012). Positive control samples were taken 30 min after final morphine injection, corresponding to peak plasma morphine levels (Kilpatrick and Smith, 2005; Trescot et al., 2008). One hemisphere of the hippocampus was used for mass spectrometric evaluation of morphine and its metabolites morphine-3-glucuronide (M3G) and morphine-6-glucuronide (M6G). The other hemisphere was used for protein analysis of the endogenous danger signal HMGB1, an inflammatory mediator and natural ligand of TLR4 that we have previously shown to be increased in aged rats 2 weeks after surgery plus morphine treatment (Muscat et al., 2021).
Subjects.
Subjects were aged (22–24 months old), male F344xBN F1 rats obtained from the National Institute on Aging (NIA) Rodent Colony managed by Charles River. Unfortunately, female rats of this strain and age, exclusively available through the NIA colony, were not available at the time these studies were conducted. F344xBN F1 rats are useful in the study of aging and aging-associated conditions because they remain relatively healthy in old age. Importantly, the aged rats are not senescent, and previous work from our group has shown that unchallenged rats at this age do not exhibit impairment on contextual memory tasks compared with young adult (3 months old) controls (Barrientos et al., 2006, 2012, 2015; Frank et al., 2010; Spencer et al., 2017; Butler et al., 2021; Muscat et al., 2021). Young adult rats were not used in this study based on our previous work that established they are not impaired in this model (Muscat et al., 2021). Condition-matched rats were housed 2 to a cage (length, 52 cm; width, 20 cm; height, 21 cm). Subjects did not receive environmental enrichment, based on literature showing it to be antineuroinflammatory (for review, see Muscat and Barrientos, 2020), to avoid potential confounds. The animal colony was maintained at 22 ± 2°C on a 12 h light/dark cycle (lights on at 7:00 A.M.). Rats were allowed ad libitum access to food and water. Animals were given at least 1 week to acclimate to colony conditions before experimentation. All experiments were conducted in accordance with protocols approved by The Ohio State University Animal Care and Use Committee. Every effort was made to minimize the number of animals used and their suffering.
Surgery.
Laparotomy (exploratory abdominal surgery) was performed using aseptic procedures under isoflurane anesthesia, as described (Martin et al., 2005; Barrientos et al., 2012; Muscat et al., 2021). Briefly, the abdominal region was shaved and cleaned with 70% ethanol and surgical scrub. Approximately 0.5 cm below the lower right rib, a 3 cm incision was made, revealing the peritoneal cavity. While wearing sterile gloves, the viscera and musculature were vigorously manipulated by the surgeon. Then, ∼10 cm of the small intestines was exteriorized and vigorously rubbed between the surgeon's thumb and index finger for 30 s, after which they were returned into the peritoneal cavity. Sterile chromic gut sutures (3–0) were used to suture the peritoneal lining and abdominal muscle in two layers; the skin was closed with surgical staples. The wound was then dressed with Polysporin to prevent infection. Each procedure lasted ∼20 min. Based on previous work demonstrating that sham-operated rats are unimpaired at any and all time points, sham surgery controls were not included in this study (Muscat et al., 2021). Because the present study assessed the impact of an analgesic (morphine), control animals did not receive any analgesic treatment to avoid potential confounds. All subjects were assessed twice daily for the first week after surgery, and then twice weekly thereafter, to monitor surgical recovery.
Drugs and administration procedures.
Morphine was gifted by the National Institute on Drug Abuse drug repository and administered intraperitoneally at a dose of 2 mg/kg/ml twice daily (at ∼9:00 A.M. and ∼5:00 P.M.) for 7 d, based on prior studies (Morgan et al., 2006; Hutchinson et al., 2009, 2010; Grace et al., 2019; Muscat et al., 2021). The human morphine dose equivalency to the rat dose used is 45 mg/d (based on calculations suggested in the study by Reagan-Shaw et al. (2008) and was chosen based on the recommended dose for opioid-naive patients (MD Anderson Cancer Center, 2019). Morphine is reported as a free base concentration and was diluted in sterile saline (0.9%). An equivolume of sterile saline was administered to control animals.
LPS-RS is a competitive antagonist of TLR4 via its binding to the MD-2 coreceptor binding site (Coats et al., 2005; Visintin et al., 2005). LPS-RS (InvivoGen) is a selective TLR4 inhibitor with no activity at TLR2 (Hirschfeld et al., 2000; Vallance et al., 2019). It was administered with intracisterna magna injection at a dose of 500 ng/5 μl, 5 μg/5 μl, or 50 μg/5 μl, based on previous work (Chen et al., 2015; Sun et al., 2015; Zhang et al., 2016; Liu et al., 2018a). In experiments 1 and 3, rats received a single injection once daily for 7 d, beginning the day after the last morphine injection. In experiments 2 and 4, rats received a single injection immediately before laparotomy. In experiment 5, rats received either a single intracisterna magna injection of LPS-RS immediately before surgery or once-daily intracisterna magna injections for 3 d, beginning the day after the last morphine treatment. Intracisterna magna administration was used to directly target the CNS in a minimally invasive manner to reduce additional inflammation that may be caused by stereotactic surgery. Under isoflurane anesthesia, the dorsal aspect of the skull was shaved and then cleaned with 70% ethanol. A 27 gauge needle attached via PE50 tubing to a 25 μl Hamilton syringe was inserted into the cisterna magna. Entry into the cistern was verified by drawing up ∼2 μl of clear CSF. The cerebral spinal fluid was then gently pushed back in, and a 5 μl total volume of LPS-RS was subsequently administered over 30 s. The entire procedure is completed in ∼3 min. Vehicle control animals received an equal volume of sterile water.
Behavior.
Memory function was assessed using the CPF-FC paradigm. This assay was chosen because it is widely accepted to be highly and specifically dependent on the hippocampus (Fanselow, 1990; Rudy et al., 2002) and because we have validated its use in assessing memory impairment in both young adult and aging rats following an array of insults (Barrientos et al., 2006; Sobesky et al., 2014; Muscat et al., 2021). Rats were conditioned using the Coulbourn Instruments Habitest Modular System. Each conditioning chamber (width, 12 inches; depth, 10 inches; height, 12 inches) had two solid metal walls and two walls made of Plexiglas. An audio speaker and a house light were mounted on the ceiling. A footshock could be delivered through a removable grid shock floor. The rods were wired to a shock generator for eight unique shock outputs. These conditioning chambers were placed inside an isolation cubicle (width, 23 inches; depth, 20 inches; height, 24 inches). Before each animal was conditioned or tested, the chambers were cleaned with water and 70% ethanol.
CPF-FC was performed as described in the study by Muscat et al. (2021). Briefly, CPF-FC is composed of the following three components: a pre-exposure phase, an immediate shock phase, and a testing phase. During the pre-exposure phase, rats were transported in a black bucket from their home cage to the conditioning context, where they were allowed to freely explore. This was repeated six times (rats remained in the conditioning context for 5 min on the first exposure and for 40 s on each of the five subsequent exposures, with an ∼40 s interval in their home cage between each subsequent exposure) to establish an association between the black bucket and the conjunctive representation of the conditioning context (for further details, see Rudy et al., 2002). During the first 5 min exposure, general locomotion was assessed to account for any confounding motor disturbances or generalized fear. Three days later, the immediate shock phase was conducted. Rats were transported from their home cage to the conditioned context in the same black bucket as before. Immediately upon being placed in the conditioned context, rats received one 2 s, 1.5 mA footshock. They were then removed from the conditioned context and transported back to their home cage. During the immediate shock phase, the time rats spent in the conditioned context never exceeded 10 s. Twenty-four hours after the immediate shock phase (4 d after pre-exposure), rats were tested for memory of the conditioned context. During this testing phase, rats were again transported from their home cage to the conditioned context in the black bucket. They were then observed and scored for freezing behavior; freezing is the dominant fear response of a rat and is characterized by a complete suppression of behavior, including immobility, shallow breathing, and autonomic changes such as increased heart rate and piloerection (Fanselow and Lester, 1988). In this study, freezing was defined as the absence of all visible movement, except for respiration. Rats were scored every 10 s while in the chamber for 6 min. Two hours later, locomotion and generalized fear were again assessed by placing the rats in a new, novel/neutral context (a round plastic chamber with wire walls and a solid floor with corn cob bedding) for another 6 min and observing them for freezing behavior. Scoring was conducted manually in real time by three observers blind to treatment conditions; scores were averaged, and inter-rater reliability exceeded 97% for all studies.
A separate cohort of rats underwent a learning experience to evaluate the effects of surgery, morphine administration, and LPS-RS on memory consolidation processes. In these experiments, the learning experience consisted of the pre-exposure phase of the CPF-FC paradigm. Rats were transported in a black bucket from their home cage to the conditioning chamber six times, as described above. After the last exposure, rats were returned to their home cages and were killed 2 h later, as described below.
Tissue dissection.
Rats received a lethal dose of sodium pentobarbital (Fatal-Plus) and were then transcardially perfused with ice-cold 0.9% saline for 3 min to remove any circulating immune leukocytes from the CNS vasculature. Then, brains were quickly extracted and placed on an ice-cold glass plate, and hippocampi and amygdala were dissected. Brain tissues were then frozen in liquid nitrogen before being stored at −80°C until the time of processing.
Western blots.
One hemisphere of the hippocampus or amygdala was manually sonicated in 0.3 or 0.2 ml buffer, respectively, for 10 s using an ultrasonic cell disrupter (Thermo Fisher Scientific). Sonication buffer contained a 50 mm Tris base and an enzyme inhibitor cocktail that included 100 mm amino-n-caproic acid, 1 mm EDTA, 5 mm benzamidine HCl, 0.2 mm phenylmethyl sulfonyl fluoride, and a cOmplete Mini Protease Inhibitor Cocktail tablet (Millipore Sigma). Following sonication, samples were centrifuged at 10,000 rpm at 4°C for 10 min, and supernatants were transferred into a clean tube. Bradford protein assays were performed on all samples (before the first freeze) to determine total protein concentrations. Samples were divided into 10 μl aliquots and frozen at −80°C until Western blots were performed.
An equal amount of total protein (60 μg for hippocampus or 40 μg for amygdala) from each sample was loaded into each lane. The NuPAGE Bis-Tris (10 wells, 4–12%, 1.5 mm) gel electrophoresis system was used under reducing conditions (Thermo Fisher Scientific). The iBlot dry-blotting system (Thermo Fisher Scientific) was used to electrophoretically transfer gels to nitrocellulose membranes. Odyssey TBS blocking solution (LI-COR) was used to prevent nonspecific protein binding [1 h at room temperature (RT)]. The primary antibodies used, in Odyssey blocking solution containing 0.2% Tween20 (overnight at 4°C), were as follows: HMGB1 (1:1000; catalog #ab18256, Abcam); TLR4 (1:1000; catalog #sc-293072, Santa Cruz Biotechnology); synaptophysin (SYP; 1:1000; catalog #sc-17 750, Santa Cruz Biotechnology); PSD-95 (1:1000; catalog #2507S, Cell Signaling Technology); synaptopodin (SYNPO; 1:1000; catalog #ab259976, Abcam); BDNF (1:500; catalog #ab108319, Abcam); total tropomyosin-related kinase B (TrkB; 1:1000; catalog #ab187041, Abcam); phosphor-TrkB (1:500; catalog #SAB4503785, Millipore); phosphor-phospholipase Cγ (PLCγ; 1:500; catalog #SAB4300082, Millipore); total PLCγ (1:500; catalog #ab76155, Abcam); total ERK (tERK; 1:1000; catalog #9102S, Cell Signaling Technology); and phosphor-ERK (pERK; 1:1000; catalog #9101, Cell Signaling Technology). Anti-GAPDH (1:50,000; catalog #8245, Abcam) was used as an internal loading control. Blots were washed 4× for 5 min with TBS plus 0.1% Tween 20 and then probed with the appropriate fluorescent secondary antibody (1:20,000; Thermo Fisher Scientific) for 1 h at RT. The Odyssey Infrared Imaging System (LI-COR) was used to image membranes, and Empiria Studio version 1.3 software was used for the quantification of bands. Each protein band was normalized to its respective loading control and then analyzed as a percentage of control (intraperitoneally injected saline plus intracisterna magna injected saline).
Mass spectrometry.
Both nominal and isotopically labeled standards were purchased from Cambridge Isotope Laboratories. Morphine, M3G, and M6G standards were prepared at 10 ppm in 100% water. A serial dilution was then performed to produce a calibration range from 5 ppm to 1 part per billion (ppb). Isotopically labeled internal standards were then added to a concentration of 100 ppb. Standards were prepared by the same methods as tissue and serum samples, respectively. One hundred microliters of the serum or standard solutions were mixed with 400 μl of ice-cold methanol to perform protein precipitation. Samples were then centrifuged at 13,500 × g for 15 min. Supernatant was then collected, and the solvent was evaporated to produce a metabolite pellet. Pellets were then dissolved in 85%/15% H2O/MeOH and placed on the instrument without further purification. Tissue samples were homogenized. Samples were first weighed and then placed inside the homogenizing tube with 10 μl/mg 80%/20% MeOH/H2O. Homogenized mixture was then spun down 13,500 × g for 15 min. Solid phase exchange was then performed to extract desired metabolites. The procedure was as follows: columns were first washed with 2 ml of MeOH, then followed by 2 ml of H2O, and, finally, 2 ml of 10 mm ammonium bicarbonate, pH 8.8. Samples were then loaded onto the column and washed with an additional 2 ml of ammonium bicarbonate. Samples were extracted with an additional 2 ml of MeOH. Solvent was then evaporated and resolubilized in 85%/15% H2O/MeOH.
Liquid chromatography separation was performed using a Dionex UltiMate 3000 RSLC HPLC system (Thermo Fisher Scientific). The column was a reverse phase InfinityLab Poroshell 120 (Agilent) column with a partial size of 2.7 μm, an inner diameter of 2.1 mm, and a column length of 100 mm. Water with 0.1% formic acid was used as mobile phase A, while 100% MeOH with 0.1% formic acid was used as mobile phase B. The initial flow gradient was set to 15% mobile phase B for the first 30 s. The gradient was increased to 70% B over the next 1 min and 30 s. The gradient was then reduced to 15% mobile phase B at 2.1 min. The flow rate was set to 250 μl/min with a column temperature of 40°C. Injection volume was set to 5 μl for each run.
All mass spectrometry experiments were performed on a Q Exactive Plus Hybrid Quadrupole-Orbitrap Mass Spectrometer (Thermo Fisher Scientific). The following settings were used for all experiments using a HESI source. Spray voltage was set to 4000 V. Sheath gas pressure was set to 30. Auxiliary gas pressure was set to 20. Capillary temperature was set to 365°C. A parallel reaction monitoring setup was used as the primary method for detection of morphine, M3G, and M6G. Briefly, parent ion masses were set to 286.14 Da for morphine and 462.18 Da for M3G and M6G. Analytes were detected by their product ions, which was 153.14 Da for morphine (found with 65 V of collisional energy) and 286.14 for M6G and M3G (found with 65 V of collisional energy). All chromatograms were processed using Xcalibur (Thermo Fisher Scientific).
Statistical analysis.
For all experiments, n = 6–8 rats/group were used, based on statistical power established by our previous work (Barrientos et al., 2012; Spencer et al., 2017; Muscat et al., 2021). Statistical analyses were performed using Prism version 9 software. Data met the requirements for parametric analysis, based on Shapiro–Wilk tests for normality and Q–Q plots. One-way and two-way ANOVAs or Student's t tests were used where appropriate, based on the experimental design. Following significant main effects or interactions, Tukey's post hoc tests were conducted to assess for pairwise differences between groups. Statistical significance for all tests was set at α = 0.05.
Results
TLR4 antagonism reverses hippocampal memory deficits associated with surgery and morphine
To determine the extent to which TLR4 mediates the development of morphine-induced persistent POCD, we first sought to determine whether central TLR4 antagonism after surgery and morphine treatment would ameliorate or reverse hippocampal memory dysfunction 4 weeks after surgery (Fig. 1A). We chose to focus on hippocampus-dependent memory because our previous work has established it to be highly and specifically affected in this POCD model, with no impairments observed in amygdala-dependent memory (Barrientos et al., 2012; Muscat et al., 2021). Importantly, the hippocampus-specific mechanisms that mediate contextual fear conditioning (conjunctive representations and pattern completion) are the same ones that support episodic declarative memory in humans (Rudy and O'Reilly, 1999; Rudy et al., 2002). Furthermore, episodic declarative memory is robustly impaired in patients with POCD and Alzheimer's disease (Nestor et al., 2006; Jungwirth et al., 2009; Schwarz et al., 2013); thus, this is an appropriate behavioral paradigm for our model. A two-way ANOVA with intraperitoneal treatment (saline or morphine) and intracisternal magna treatment (vehicle, 500 ng/5 μl, 5 μg/5 μl, or 50 μg/5 μl LPS-RS; n = 6–7/group) revealed a main effect of intraperitoneal treatment (F(1,43) = 33.66, p < 0.0001) and a significant interaction of intraperitoneal treatment × intracisterna magna treatment (F(3,43) = 3.439, p = 0.0250). Post hoc analyses revealed that morphine-treated rats who received 500 ng/5 μl, 5 μg/5 μl, or vehicle were significantly impaired on the memory test compared with saline-matched controls (p = 0.0065; p = 0.0212; p = 0.0060; Fig. 1B). In contrast, animals that received 50 μg/5 μl following morphine treatment did not differ from their saline-matched controls (p > 0.99; Fig. 1B). No saline-treated groups significantly differed from one another (Fig. 1B).
TLR4 antagonism ameliorates hippocampal memory deficits associated with surgery and morphine. A, Schematic depiction of experimental timeline. B, Treatment with 50 μg/5 μl LPS-RS ameliorated the persistent hippocampal memory deficit 4 weeks after surgery and morphine treatment. C, No differences were observed in the novel context test, and all groups froze <20% of the time, indicating little-to-no generalized fear. n = 6–7/group. Data are presented as the mean ± SEM. *p < 0.05, **p < 0.01.
Freezing to a novel context was also assessed to rule out the possibility of generalized fear or anxiety. No differences between groups were observed (n = 6–7/group; p > 0.05), and all groups froze <20% of the time, indicating little-to-no fear (Fig. 1C). It should also be noted that pain has the potential to confound the memory task. Pain was not assessed in the present study, as we have previously established that the effects of postsurgical morphine administration on memory are independent of its analgesic function, as the memory impairment is observed even in the absence of opioid receptor activation (Muscat et al., 2021). Additionally, pain is classically associated with increased freezing behavior (Vuralli et al., 2019), whereas memory impairments are associated with reduced freezing; thus, if pain were contributing to the behavioral output in this task, it would likely mask the memory impairment, which is not what we observed herein.
TLR4 antagonism prevents persistent hippocampal memory deficits associated with surgery and morphine
We next sought to determine whether central TLR4 antagonism immediately before surgery and morphine treatment would prevent hippocampal-dependent memory impairments 4 weeks after surgery (Fig. 2A). Here, only the effective dose from experiment 1 (50 μg/5 μl) was administered to minimize the number of rats used. A two-way ANOVA with intraperitoneal treatment (saline or morphine) and intracisterna magna treatment (vehicle or 50 μg/5 μl LPS-RS; n = 6–10/group) was performed, revealing significant main effects of both intraperitoneal treatment (F(1,27) = 4.338, p = 0.0469) and intracisterna magna treatment (F(1,27) = 5.702, p = 0.0242). An intraperitoneal treatment × intracisterna magna treatment interaction was also observed (F(1,27) = 4.745, p = 0.0383), and post hoc analysis again revealed that morphine-treated rats that were pretreated with the vehicle intracisterna magna injection were significantly impaired on the memory test compared with saline-matched control rats (p = 0.0442; Fig. 2B). Post hoc analyses also revealed that morphine-treated rats that received LPS-RS before surgery were not impaired relative to controls (p > 0.99), and that their memory function was significantly improved relative to morphine-treated vehicle controls (p = 0.0191). Freezing to a novel context again indicated no significant interaction (p > 0.05) ruling out the potential confounds of generalized fear or anxiety, and all groups froze <20% of the time, indicating little-to-no fear, consistent with our previous findings in this model (Muscat et al., 2021; Fig. 2C).
Preoperative TLR4 antagonism prevents hippocampal memory deficits associated with surgery and morphine. A, Schematic depiction of experimental timeline. B, Treatment with 50 μg/5 μl LPS-RS prevented the persistent hippocampal memory deficit 4 weeks after surgery and morphine treatment. C, No pairwise differences were observed in the novel context, and all groups froze <20% of the time, indicating little-to-no generalized fear. n = 6–10/group. Data are presented as the mean ± SEM. *p < 0.05, **p < 0.01.
TLR4 antagonism ameliorates postsynaptic protein dysregulation associated with surgery and morphine
Previous studies using this model have shown a dysregulation of synaptic element transcripts in the hippocampus of aged rats who received surgery and morphine treatment (Muscat et al., 2021). Therefore, we aimed to determine whether this dysregulation exists at the protein level and, if so, whether it is ameliorated by TLR4 antagonism. In the hippocampus, a two-way ANOVA, with intraperitoneal treatment (saline or morphine) and intracisterna magna treatment (vehicle or 50 μg/5 μl LPS-RS; Fig. 3A) for the presynaptic marker SYP showed no between-group differences in animals treated postoperatively with LPS-RS or vehicle (n = 6–7/group, p > 0.05; Fig. 3C). However, analysis of the postsynaptic element postsynaptic density-95 (PSD-95) revealed a main effect of intracisterna magna treatment (n = 6–7/group, F(1,21) = 10.63, p = 0.0037), as well as an intraperitoneal treatment × intracisterna magna treatment interaction (F(1,21) = 23.51, p < 0.0001; Fig. 3D). Post hoc analysis showed a significant reduction of PSD-95 in morphine-treated animals compared with saline-treated animals (p = 0.0071), and this reduction was ameliorated by LPS-RS treatment (p < 0.0001; Fig. 3D). Given the apparent specificity of the effects of morphine to the postsynaptic region, we further probed this relationship by assessing the postsynaptic factor SYNPO, an actin-binding protein critical for dendritic spine formation and synaptic plasticity (Vlachos et al., 2009; Segal et al., 2010). In rats treated postoperatively with LPS-RS or vehicle, Western blot assessment by two-way ANOVA, with intracisterna magna treatment (vehicle or 50 μg/5 μl LPS-RS) and intraperitoneal treatment (saline or morphine; n = 6/group), revealed a significant main effect of intraperitoneal treatment (F(1,20) = 8.151, p = 0.0098), where morphine caused a reduction in SYNPO expression regardless of intracisterna magna condition (Fig. 3E). No between-group differences were observed for the internal loading control GAPDH (data not shown).
TLR4 antagonism ameliorates dysregulation of hippocampal synaptic markers following surgery and morphine treatment. A, Schematic depiction of the experimental timeline associated with B–E. B, Representative Western blot images for C–E. C, Western blot quantification of the presynaptic marker SYP. D, Quantification of the postsynaptic element PSD-95. E, Quantification of the postsynaptic element SYNPO. F, Schematic depiction of experimental timeline associated with G–J. G, Representative Western blot images for H–J. H, Quantification of SYP. I, Western blot quantification of PSD-95. J, Quantification of SYNPO. n = 6–7/group. Data are presented as the mean ± SEM. *p < 0.05, **p < 0.01, ****p < 0.0001.
A similar trend was observed in animals treated preoperatively with the TLR4 antagonist (Fig. 3F). A two-way ANOVA showed no between-group differences for SYP (n = 6–7/group, p > 0.05; Fig. 3H). For PSD-95, main effects of intraperitoneal treatment (n = 6–7/group, F(1,22) = 7.418, p = 0.0124) and intracisterna magna treatment (F(1,22) = 5.249, p = 0.0319) were observed (Fig. 3I). A significant intraperitoneal treatment × intracisterna magna treatment interaction was also revealed (F(1.22) = 4.873, p = 0.0380; Fig. 3I). As in the postoperatively treated rats, post hoc analysis showed a significant reduction of PSD-95 in morphine-treated rats compared with saline-treated controls (p = 0.0104), which was ameliorated by LPS-RS treatment (p = 0.0209; Fig. 3I). A two-way ANOVA of SYNPO in rats treated preoperatively with LPS-RS or vehicle (n = 6–7/group) revealed a slightly different effect (Fig. 3J). There was a main effect of intracisterna magna condition (F(1,21) = 6.547, p = 0.0183), indicating increased expression in LPS-RS-treated groups. Also, an intraperitoneal treatment × intracisterna magna treatment interaction was observed (F(1,21) = 4.522, p = 0.0455); morphine treatment caused a significant reduction in SYNPO compared with saline-treated controls (p = 0.0450), and this reduction was significantly prevented by LPS-RS (p = 0.0184; Fig. 3J). No between-group differences were observed for the internal loading control GAPDH in preoperatively treated subjects (data not shown).
BDNF expression is dysregulated by surgery and morphine and ameliorated by TLR4 antagonism
To further investigate the mechanisms by which morphine exerts its deleterious effects on cognition in the postoperative setting, we next evaluated expression of the trophic factor BDNF, as it is critical to proper long-term memory function and is known to be vulnerable to dysregulation by neuroinflammation (Barrientos et al., 2004; Cortese et al., 2011; Patterson, 2015; Tanaka et al., 2018; Muscat and Barrientos, 2021). Protein expression of a precursor of BDNF (proBDNF) and mature BDNF (mBDNF) were assessed via Western blot in tissues taken 2 h after a learning experience, when memory consolidation is actively occurring. BDNF is synthesized as proBDNF, and mBDNF is then generated following cleavage of the prodomain (Hempstead, 2015; De Vincenti et al., 2019; Wang et al., 2021); mBDNF is critical for neuronal survival, differentiation, and synaptic plasticity, while proBDNF has been implicated in apoptosis and long-term depression (Lee et al., 2001; Lu et al., 2014; Leal et al., 2017; Li et al., 2017; Kowiański et al., 2018; von Bohlen und Halbach and von Bohlen und Halbach, 2018; Liu et al., 2018b; De Vincenti et al., 2019; Wang et al., 2020b). In rats treated postoperatively with LPS-RS or vehicle (Fig. 4A), a two-way ANOVA with intraperitoneal treatment (saline or morphine) and intracisterna magna treatment (vehicle or 50 μg/5 μl LPS-RS; n = 6–8/group) showed no between-group differences in levels of proBDNF in the hippocampus (p > 0.05; Fig. 4C). However, for mBDNF, a significant intraperitoneal treatment × intracisterna magna treatment interaction effect was observed (F(1,20) = 5.261, p = 0.0328; n = 6–7/group), with morphine causing a marked reduction in mBDNF that was ameliorated by LPS-RS (p = 0.0339; Fig. 4D). The ratio of hippocampal mBDNF to proBDNF was also quantified, and statistical analysis revealed a similar trend to that of mature BDNF. A two-way ANOVA again revealed a intraperitoneal treatment × intracisterna magna treatment interaction effect (F(1,21) = 8.665, p = 0.0078; n = 6–7/group) as well as a main effect of intraperitoneal treatment (F(1,21) = 10.07, p = 0.0046), where the mBDNF/proBDNF ratio was significantly reduced in morphine-treated animals relative to control animals (p = 0.0012); this blunting was diminished by LPS-RS treatment, as morphine plus LPS-RS-treated subjects did not differ from vehicle-treated controls (p = 3.122; Fig. 4E).
TLR4 antagonism ameliorates hippocampal BDNF dysregulation. A, Schematic depiction of experimental timeline associated with B–E. B, Representative Western blot images for C–E. C, Western blot quantification of proBDNF. D, Quantification of mBDNF. E, Quantification of the ratio of mBDNF to proBDNF. F, Schematic depiction of experimental timeline associated with G–J. G, Representative Western blot images for H–J. H, Quantification of proBDNF. I, Quantification of mBDNF. J, Quantification of the ratio of mBDNF to proBDNF. n = 6–7/group. Data are presented as the mean ± SEM. *p < 0.05, **p < 0.01.
A similar phenomenon was observed in rats treated preoperatively with LPS-RS (Fig. 4F). A two-way ANOVA with intracisterna magna treatment (vehicle or 50 μg/5 μl LPS-RS) and intraperitoneal treatment (saline or morphine; n = 6–7/group) showed no between-group differences for proBDNF (p > 0.05; Fig. 4H) but did reveal a significant intracisterna magna treatment × intraperitoneal treatment interaction effect (F(1,21) = 9.228, p = 0.0063) for mBDNF (Fig. 4I). Post hoc analysis showed that mBDNF was significantly reduced in morphine-treated rats relative to saline-treated controls (p = 0.0279); this effect was prevented by LPS-RS treatment, as mBDNF expression in LPS-RS plus morphine-treated rats did not differ from that in saline-treated controls (p > 0.05; Fig. 4I). The ratio of mBDNF to proBDNF again mirrored the expression levels of mBDNF; a two-way ANOVA revealed an intracisterna magna treatment × intraperitoneal treatment interaction (F(1,19) = 5.204, p = 0.0342), as well as a main effect of intracisterna magna treatment (F(1,19) = 4.539, p = 0.0464; Fig. 4J). Post hoc analysis showed that the mBDNF/proBDNF ratio was significantly reduced in vehicle plus morphine-treated rats compared with all other groups (p < 0.05; Fig. 4J).
TLR4 antagonism mitigates reduction of BDNF receptor activation
Given the impact of surgery plus morphine treatment on mBDNF expression, we sought to further investigate the influences of these factors on specific BDNF signaling pathways. BDNF elicits a variety of biological effects through activation of multiple receptors, the most prominent of which is TrkB (Hempstead, 2015; De Vincenti et al., 2019). Protein expression of total and phosphorylated (activated) TrkB, as well as the activation of ERK and PLCγ—two primary signaling pathways induced by BDNF (Greene and Kaplan, 1995)—were measured via Western blot.
In rats treated postoperatively with LPS-RS or vehicle, a two-way ANOVA with intraperitoneal treatment (saline or morphine) and intracisterna magna treatment (vehicle or 50 μg/5 μl LPS-RS; n = 6–7/group; Fig. 5A) revealed no effect of treatment condition on total TrkB (tTrkB; p > 0.05; Fig. 5C). However, analysis of activated TrkB [phosphor-TrkB (pTrkB)] revealed a main effect of intraperitoneal treatment (F(1,21) = 5.255, p = 0.0323) and an intraperitoneal treatment × intracisterna magna treatment interaction (F(1,21) = 11.00, p = 0.0033; Fig. 5D). Post hoc analysis showed that morphine treatment caused a significant reduction in pTrkB relative to saline-treated controls (p = 0.0031) and that TLR4 antagonism reversed this effect (p = 0.0302; Fig. 5D). Assessment of the ratio of pTrkB to tTrkB by two-way ANOVA revealed a similar result to pTrkB; a main effect of intraperitoneal treatment (F(1,20) = 5.595, p = 0.0282) and an intraperitoneal treatment × intracisterna magna treatment interaction (F(1,20) = 8.087, p = 0.0100; Fig. 5E). Tukey's post hoc analysis showed that morphine treatment reduced pTrkB/tTrkB (p = 0.0051) relative to saline-treated controls, and that LPS-RS mitigated this effect (p = 0.0421). For ERK, a two-way ANOVA revealed no effect of intraperitoneal or postoperative intracisterna magna treatment on either tERK (p > 0.05) or pERK (p > 0.05; Fig. 5F,G). Consistent with this, no effect of intraperitoneal or intracisterna magna treatment was observed for the ratio of pERK to tERK (p > 0.05; Fig. 5H). A two-way ANOVA of PLCγ showed no impact of treatment condition on total PLCγ (tPLCγ; p > 0.05; Fig. 5I), but did reveal a significant main effect of both intraperitoneal treatment (F(1,20) = 6.296, p = 0.0208) and intracisterna magna treatment (F(1,20) = 4.739, p = 0.0416) for activated PLCγ [phosphor-PLCγ (pPLCγ); n = 6/group; Fig. 5J]. Post hoc analyses showed that morphine treatment caused significant reduction in pPLCγ compared with both vehicle-treated and LPS-RS-treated saline controls (p = 0.0236 and p = 0.0168, respectively) and that this reduction was reversed by LPS-RS pretreatment (p = 0.0388; Fig. 5J). Consistent with this, the ratio of pPLCγ to tPLCγ exhibited a main effect of intraperitoneal treatment (F(1,20) = 6.239, p = 0.0213) and intracisterna magna treatment (F(1,20) = 5.227, p = 0.0333); post hoc analyses revealed that morphine caused a reduction in PLCγ activation relative to vehicle-treated controls (p = 0.0393), and that this effect was not present in the LPS-RS-treated groups (p = 0.9791; Fig. 5K).
Activation of BDNF signaling pathways in the hippocampus is reduced by morphine and restored by postoperative TLR4 antagonism. A, Schematic depiction of experimental timeline associated with B–K. B, Representative Western blot images for C–K. C–H, Western blot quantification of tTrkB (C), pTrkB (D), and the ratio of pTrkB to tTrkB (E). F–K, Quantification of tERK (F), pERK (G), and the ratio of pERK to tERK (H). I–K, Quantification of tPLCγ (I), pPLCγ (J), and the ratio of pPLCγ to tPLCγ (K). n = 6–7/group. Data are presented as the mean ± SEM. *p < 0.05, **p < 0.01.
For rats treated preoperatively with LPS-RS or vehicle, a two-way ANOVA with intraperitoneal treatment (saline or morphine) and intracisterna magna treatment (vehicle or 50 μg/5 μl LPS-RS; n = 6–7/group; Fig. 6A) showed no effect of treatment condition on tTrkB (p > 0.05; Fig. 6C). For pTrkB, a main effect of intracisterna magna treatment was revealed (F(1,21) = 12.30, p = 0.0021), as well as an interaction effect between intraperitoneal treatment and intracisterna magna treatment (F(1,21) = 11.69, p = 0.0026; Fig. 6D). Post hoc analysis revealed that in animals treated with morphine, LPS-RS significantly increased pTrkB relative to vehicle controls (p = 0.0005; Fig. 6D). In line with this, two-way ANOVA of pTrkB/tTrkB revealed a main effect of intracisterna magna treatment (F(1,20) = 5.580, p = 0.0284) and an intraperitoneal treatment × intracisterna magna treatment interaction (F(1,20) = 5.061, p = 0.0359; Fig. 6E). Post hoc analysis showed that in animals treated with morphine, TLR4 antagonism significantly increased the ratio of pTrkB to tTrkB relative to vehicle controls (p = 0.0232; Fig. 6E). For ERK, two-way ANOVAs revealed no effect of intraperitoneal or preoperative intracisterna magna treatment on either total ERK (p > 0.05) or pERK (p > 0.05; Fig. 6F,G). Consistent with this, no effect of intraperitoneal or intracisterna magna treatment was observed for the ratio of pERK to tERK (p > 0.05; Fig. 6H). Two-way ANOVA assessment of PLCγ revealed no effect of treatment condition on total PLCγ (p > 0.05; Fig. 6I), but did determine a significant intraperitoneal treatment × intracisterna magna interaction effect (F(1,20) = 6.596, p = 0.0183) for activated PLCγ (Fig. 6J). Post hoc analyses showed that morphine treatment caused significant reduction in pPLCγ compared with saline controls (p = 0.0421) and that this was significantly prevented by LPS-RS pretreatment (p = 0.0312; Fig. 6J). The ratio of pPLCγ to tPLCγ followed a similar trend—a two-way ANOVA revealed an intraperitoneal treatment × intracisterna magna treatment interaction (F(1,21) = 5.312, p = 0.0315), and post hoc assessment indicated that morphine caused a reduction in PLCγ activation relative to saline-treated controls (p = 0.0477) and that TLR4 antagonism ameliorated this reduction (p = 0.0363; Fig. 6K).
Preoperative TLR4 antagonism prevents morphine-induced dysregulation of hippocampal BDNF signaling. A, Schematic depiction of experimental timeline associated with B–K. B, Representative Western blot images for C–K. C–E, Western blot quantification of tTrkB (C), pTrkB (D), and the ratio of pTrkB to tTrkB (E). F–H, Quantification of tERK (F), pERK (G), and the ratio of pERK to tERK (H). I–K, Quantification of tPLCγ (I), pPLCγ (J), and the ratio of pPLCγ to tPLCγ (K). n = 6–7/group. Data are presented as the mean ± SEM. *p < 0.05, ***p < 0.001.
Surgery and morphine does not impact synaptic elements or BDNF in the amygdala
Thus far, synaptic and memory-related proteins were assessed only in the hippocampus, based on the results of the hippocampus-dependent memory task. To determine the specificity of the dysregulation of memory-associated proteins to the hippocampus, we next assessed the impact of surgery plus morphine on these proteins in the amygdala. The amygdala was chosen based on previous work showing that our model of laparotomy-associated memory deficit is specific to the hippocampus, with no effect on amygdala-dependent memory (Barrientos et al., 2012).
Western blot assessment of the synaptic elements SYN, PSD-95, and SYNPO was performed on amygdala of saline-treated and morphine-treated rats (n = 6/group; Fig. 7A). A t test revealed no effect of morphine treatment on the presynaptic protein SYN (t(10) = 0.02,769, p = 0.9785; Fig. 7C). Similarly, levels of the postsynaptic elements PSD-95 and SYNPO were not affected by morphine treatment (t(10) = 0.01,314, p = 0.9898 and t(10) = 0.4343, p = 0.6733, respectively; Fig. 7D,E). Levels of BDNF were also assessed; both proBDNF and mBDNF were unaffected by morphine in the hippocampus, based on statistical assessment by t test (t(10) = 0.9480, p = 0.3655 and t(10) = 0.4057, p = 0.6935, respectively; Fig. 7F,G). In line with these findings, the ratio of mBDNF to proBDNF in the amygdala did not differ between saline-treated and morphine-treated rats (t(10) = 0.7092, p = 0.4944; Fig. 7H). Because amygdala BDNF was not affected by morphine treatment, levels of downstream BDNF-related signaling proteins were not evaluated. Similarly, given that morphine did not cause dysregulation of synaptic-associated and memory-associated molecules in the amygdala, we did not assess the impact of LPS-RS treatment in this brain region.
Synaptic-associated and memory-associated proteins are unaffected by morphine plus surgery in the amygdala. A, Schematic timeline associated with B–H. B, Representative Western blot images for C–H. C–H, Western blot quantification of SYP (C), PSD-95 (D), SYNP (E), proBDNF (F), mBDNF (G), and mBDNF/proBDNF ratio (H). n = 6/group. Data are presented as the mean ± SEM.
TLR4 antagonism blocks endogenous signaling
The fact that TLR4 blockade occurring after the end of morphine treatment was effective in ameliorating later cognitive deficits suggests that morphine plus surgery leads to an enduring signal that is capable of activating TLR4, which LPS-RS administered after morphine administration would then be blocking. We considered the following two major possibilities: (1) morphine or its TLR4-active metabolite M3G is still elevated in brain parenchyma during post-morphine LPS-RS treatment; or (2) an endogenous ligand capable of activating TLR4 is persistently elevated. To answer the first possibility, mass spectrometric analysis of morphine and its major metabolites M3G and M6G was performed. For the second, Western blot was used to quantify HMGB1, an endogenous danger signal and ligand of TLR4. HMGB1 was chosen for assessment based on our previous work indicating that elevated endogenous protein levels of HMGB1 in aged rats mediate contextual fear memory deficits (Fonken, et al., 2016), and that Hmgb1 mRNA is potently elevated for at least 2 weeks following surgery and morphine treatment in aged, but not in young, rats (Muscat et al., 2021). For both analyses, hippocampal tissue was taken 72 h after the last morphine injection (corresponding to both the midway point of the postoperative LPS-RS treatment regimen and the time by which ∼90% of administered morphine is excreted; Iwamoto and Klaassen, 1977; Saboory et al., 2012); positive controls were taken 30 min after the last morphine treatment, when plasma morphine levels are at their peak following administration (Kilpatrick and Smith, 2005).
Mass spectrometry revealed that, in the hippocampus, morphine treatment caused an increase in morphine, M3G, and M6G concentrations in the brain region 30 min after the last injection, confirming that morphine indeed enters the CNS following intraperitoneal administration (Table 1, Fig. 8A,D,G). While morphine was present at concentrations high enough to reliably quantify, the metabolites M3G and M6G, while detectable, were present at levels too low to accurately quantify (Table 1, Fig. 8A,D,G). This likely relates to the pharmacodynamics of morphine and its metabolites; both morphine and M6G can diffuse across the blood–brain barrier, albeit slowly, whereas M3G cannot (De Gregori et al., 2012). Thus, the presence of M6G and especially M3G in brain parenchyma largely depends on morphine first entering the brain tissue and then being metabolized. As a result, the presence of M3G and M6G in the brain lags behind that of morphine (De Gregori et al., 2012; Klimas and Mikus, 2014), so it may be that 30 min was simply too soon to detect M3G and M6G in the hippocampus with our methods. By 72 h after the last injection, morphine, M3G, and M6G levels are undetectable, indistinguishable from saline-treated controls (Table 1, Fig. 8B,C,E,F,H,I). A similar trend was observed in serum samples, where levels of morphine, M3G, and M6G were elevated 30 min after morphine treatment, but greatly reduced by 72 h (Table 1, Fig. 9A–I). These findings indicate that morphine and its TLR4-active metabolite M3G do not stay elevated in the CNS long enough to elicit prolonged activation of TLR4, suggesting that postoperatively administered LPS-RS does not reverse the morphine-induced memory impairments through the blockade of long-lasting morphine-evoked or M3G-evoked TLR4 activation.
Mass spectrometry of morphine and its metabolites in hippocampus and serum
Mass spectrometry of morphine and its metabolites in hippocampus after last morphine injection. A–C, Representative full retention time scan for 30 min after last morphine injection (A), 72 h after last morphine injection (B), and 72 h after last saline injection (C). D–F, Representative signal for morphine with standard depicted in inset panel for 30 min after last morphine injection (D), 72 h after last morphine injection (E), and 72 h after last saline injection (F). G–I, Representative signal for metabolites of morphine, M3G and M6G, with the standard depicted in the inset panel for 30 min after last morphine injection (G), 72 h after last morphine injection (H), and 72 h after last saline injection (I).
Mass spectrometry of morphine and its metabolites in serum after the last morphine injection. A–C, Representative full retention time scan for 30 min after last morphine injection (A), 72 h after last morphine injection (B), and 72 h after last saline injection (C). D–F, Representative signal for morphine with standard depicted in the inset panel for 30 min after last morphine injection (D), 72 h after last morphine injection (E), and 72 h after last saline injection (F). G–I, Representative signal for metabolites morphine, M3G and M6G, with the standard depicted in the inset panel for 30 min after last morphine injection (G), 72 h after last morphine injection (H), and 72 h after last saline injection (I).
In support of the second possibility (antagonism of an endogenous TLR4 ligand), Western blot analysis and a one-way ANOVA revealed a significant effect of treatment (vehicle only, morphine only, morphine plus preoperative LPS-RS, or morphine plus postoperative LPS-RS; F(3,21) = 4.639, p = 0.0122; Fig. 10A). An increase of HMGB1 was observed in morphine-treated rats 72 h after the last morphine injection, relative to vehicle-treated controls (p = 0.0451; Fig. 10C). Evaluation of HMGB1 after surgery plus morphine administration and either preoperative or postoperative LPS-RS treatment revealed that TLR4 antagonism mitigated the morphine-induced increase in HMGB1; levels of HMGB1 in preoperatively and postoperatively treated groups were statistically reduced compared with morphine-treated groups (p = 0.0437 and p = 0.0200, respectively) and did not differ from those of saline-treated controls (p > 0.9999 and p = 0.9677, respectively; Fig. 10C). TLR4 levels were also assessed to determine whether they mirror the increase in receptor ligand; a one-way ANOVA showed no significant difference in TLR4 expression among vehicle-treated, morphine-treated, or morphine plus LPS-RS-treated rats (F(3,23) = 0.3351, p = 0.8001; Fig. 10D). These findings suggest that the combination of aging, surgery, and morphine treatment induces long-lasting upregulation of the endogenous TLR4 ligand HMGB1, causing prolonged neuroinflammation and synaptic dysregulation, which LPS-RS blocks. Of course, these data do not rule out the possibility that additional endogenous TLR4 ligands may also be contributing to the observed memory deficits.
HMGB1 is increased in hippocampus 72 h after morphine treatment and ameliorated by TLR4 antagonism. A, Schematic depiction of experimental timeline. B, Representative Western blot images. C, Quantification of the danger signal HMGB1. D, Quantification of TLR4. n = 6–8/group. Data are presented as the mean ± SEM. *p < 0.05.
Discussion
We found that central TLR4 antagonism both prevented and reversed the persistent long-term hippocampal-dependent memory impairment associated with surgery and morphine in aged male rats. We have previously shown that this memory deficit lasts at least 8 weeks postlaparotomy (Muscat et al., 2021), recapitulating the persistent nature of POCD experienced by some patients (Evered et al., 2018). Postoperative administration of the TLR4-specific antagonist LPS-RS, beginning 1 week postsurgery (the day after the last morphine treatment) dose-dependently ameliorated the episodic memory impairment. This finding is particularly striking and clinically relevant, as the memory impairment associated with surgery and morphine has already developed by this time, suggesting that TLR4 antagonism may be a promising target to treat perioperative neurocognitive disorders, even after symptoms have developed. Additionally, we found that a single preoperative injection of LPS-RS prevented the long-term memory deficit altogether.
There is a well established link between postsurgical neurocognitive impairment and neuroinflammation, evidenced by both preclinical and clinical studies (Buvanendran et al., 2006; Rosczyk et al., 2008; Barrientos et al., 2012; Hirsch et al., 2016; Berger et al., 2017; Wang et al., 2020a; Muscat et al., 2021), and proinflammatory mediators are capable of modulating synaptic plasticity processes, including expression of synaptic proteins, neurotrophic factors, and long-term potentiation (for review, see Patterson, 2015; Muscat and Barrientos, 2021). Additionally, we previously showed that gene expression of synaptic elements is dysregulated in hippocampus of aged male rats 2 weeks after surgery and morphine treatment (Muscat et al., 2021). Therefore, we investigated the impact of surgery and morphine, as well as TLR4 antagonism, on a variety of presynaptic and postsynaptic proteins. We found that in hippocampus 4 weeks postlaparotomy, the presynaptic marker SYP was not affected by the combination of surgery and morphine in aged male rats. Conversely, the postsynaptic factors PSD-95 and SYNPO were both reduced by laparotomy and morphine treatment. These findings suggest that, in the context of aging and surgery, morphine specifically impairs protein expression of postsynaptic, but not presynaptic, markers. It should be noted that we have previously shown upregulation of both Syp and Psd95 mRNA in aged rats in this model (Muscat et al., 2021), although a few notable differences could contribute to these differential findings, as follows: (1) our previous study assessed mRNA, whereas this study measured protein expression; (2) our previous study assessed this expression at 2 weeks postsurgery, whereas here we evaluated synaptic markers 4 weeks postsurgery—it is possible that the upregulation of mRNA 2 weeks postsurgery reflects an early compensatory response that dampens by 4 weeks postsurgery; and (3) whereas we previously measured gene expression in the absence of a learning experience, here we measured synaptic protein expression during active memory consolidation, which is notable given that SYP and PSD-95 are regulated in an activity-dependent manner (Hinz et al., 2001; Kim et al., 2019; Wiesner et al., 2020). Importantly, we showed that preoperative and postoperative TLR4 antagonism ameliorated the morphine-induced reduction of PSD-95. Interestingly, differential effects of preoperative and postoperative LPS-RS were observed for SYNPO. Like PSD-95, preoperative TLR4 antagonism prevented any changes in SYNPO levels relative to those of controls. However, no effect of TLR4 antagonism was observed in postoperatively treated animal. These results suggest that, while postoperative TLR4 antagonism can improve cognitive function and ameliorate some pathology, other pathologic characteristics will remain unmitigated, suggesting that, in the case of POCD, precautionary prevention may be better than treatment.
BDNF is expressed throughout the CNS but is particularly prominent in the hippocampus (Yan et al., 1997; Wang et al., 2021) and is highly responsive to stress, including physiological challenges (Bilbo et al., 2008; Carlos et al., 2017; Notaras and van den Buuse, 2020). Aged rodents are especially susceptible to BDNF dysregulation and, consequently, to memory impairment following immunologic insults (Barrientos et al., 2011; Cortese et al., 2011; Chapman et al., 2012; Tanaka et al., 2018). Consistent with this, we observed a significant reduction in hippocampal mature BDNF protein following surgery and morphine treatment. This reduction was mitigated by both preoperative and postoperative TLR4 antagonism. Notably, there was no difference in proBDNF expression. BDNF is initially synthesized as proBDNF, and cleavage of the prodomain is required for generation of mBDNF (Hempstead, 2015; De Vincenti et al., 2019; Wang et al., 2021). Growing evidence indicates that proBDNF and mBDNF have opposing functions; whereas mBDNF is critical for neuronal survival, differentiation, and synaptic plasticity, proBDNF can induce apoptosis, inhibit cellular proliferation, and contribute to long-term depression (Lee et al., 2001; Lu et al., 2014; Leal et al., 2017; Li et al., 2017; Kowiański et al., 2018; Liu et al., 2018b; von Bohlen und Halbach and von Bohlen und Halbach, 2018; De Vincenti et al., 2019; Wang et al., 2020b). Importantly, increased cleavage of proBDNF is associated with long-term potentiation and memory function in the hippocampus, whereas reduced cleavage is linked to memory impairment (Pang et al., 2004; Barnes and Thomas, 2008; Carlino et al., 2013; Gibon and Barker, 2017). Therefore, our observation that mBDNF, but not proBDNF, is reduced by surgery plus morphine suggests a deficit in the conversion of proBDNF to mBDNF as underlying the observed memory deficit. LPS-RS administered both preoperatively and postoperatively prevented these reductions in mBDNF. Several studies have indeed associated TLR4 inhibition with enhanced BDNF expression (Lv et al., 2016; Aboul-Fotouh et al., 2018; Goel et al., 2018; Zhang et al., 2020a), although, to our knowledge, the impact of TLR4 modulation specifically on the conversion of proBDNF to mBDNF has not been examined. Further studies are needed to clarify the specific cause of mBDNF reduction in POCD and the mechanisms by which TLR4 antagonism mitigates these effects.
Given the dysregulation of mBDNF, we further characterized this deficit by evaluating activation of the critical mBDNF receptor and signaling pathways within the hippocampus. TrkB is the primary receptor mediating the prosurvival and differentiative effects of mBDNF, as well as its role in synaptic plasticity (Hempstead, 2015; De Vincenti et al., 2019; Johnstone and Mobley, 2020). Activation of TrkB by mBDNF results in receptor dimerization and autophosphorylation, which leads to activation of the major BDNF-regulated pathways, including the PLCγ and ERK pathways (Greene and Kaplan, 1995). We found that morphine treatment significantly reduced phosphorylation/activation of TrkB (consistent with reduced availability of mBDNF) and that both preoperative and postoperative TLR4 antagonism ameliorated this effect. Interestingly, no differences were observed in the activation of ERK, which plays an important role in cellular proliferation, differentiation, and neuroprotection (Bonni et al., 1999; Lavoie et al., 2020). Activation of PLCγ, however, was significantly reduced by morphine treatment, and this effect was ameliorated by TLR4 antagonism. PLCγ is critical for the release of intracellular calcium and thus synaptic plasticity (Gärtner et al., 2006; Yoshii and Constantine-Paton, 2010). Together, these findings suggest that postsurgical morphine exerts its detrimental effects on cognition via dysregulation of synaptic plasticity-related mechanisms, rather than by inhibiting neuroprotective processes.
Finally, we sought to understand the mechanism by which surgery plus morphine induces prolonged TLR4 activation, such that LPS-RS treatment administered after morphine cessation could ameliorate the associated memory impairment. Mass spectrometric analysis revealed that by 72 h after the last morphine injection, the levels of neither morphine nor its metabolites M3G and M6G were elevated in the hippocampus compared with vehicle controls. These findings indicate that morphine, or its TLR4-active metabolite M3G, is unlikely to be responsible for prolonged activation of TLR4, and that LPS-RS may be effective by blocking an endogenous TLR4 ligand instead. Indeed, Western blot analysis confirmed that HMGB1—an endogenous danger signal and natural ligand of TLR4—was significantly elevated 72 h post-morphine treatment relative to controls, and that this was ameliorated by either preoperative or postoperative LPS-RS treatment. We chose to assess HMGB1 based on previous work demonstrating that Hmgb1 mRNA is increased in aged rats 2 weeks following surgery plus morphine administration. While HMGB1 is a prominent TLR4 ligand, it is not the only endogenous factor with TLR4 activity, and thus it is possible that other physiological signaling molecules also contribute to the observed factors.
While previous studies have indicated a role for TLR4 in perioperative neurocognitive disorders (Wang et al., 2013; Lu et al., 2015; Shi et al., 2020; Yang et al., 2020a; Zhou et al., 2020), to our knowledge, this is the first study to show a mediating role for TLR4 specifically in long-lasting morphine-induced POCD, as well as the efficacy of TLR4 antagonism to reverse POCD-associated cognitive impairment. Moreover, this is the first time that TLR4 has been identified as a key mediator of neuroinflammation-induced dysregulation of synaptic elements in POCD, as inhibition of TLR4 both prevented and reversed dysregulated expression of postsynaptic densities and plasticity-related signaling molecules associated with POCD. Our findings reveal TLR4 as a potential therapeutic target to not only prevent but also to treat POCD, even after the development of cognitive symptomology. Given that our studies administered LPS-RS centrally, future studies will need to examine TLR4 antagonists that may be administered peripherally and still achieve efficacy, to increase translational validity.
Footnotes
This work was supported in part by National Institute on Aging Grant RF1-AG-028271 to R.M.B. and S.F.M., and Grant AG-067061 to R.M.B.; National Institute of Neurological Disorders and Stroke Grant T32-NS-105864 to S.M.M.; and National Center for Advancing Translational Sciences Grant UL1-TR-002733. The content is solely the responsibility of the authors and does not necessarily represent the official view of the National Center for Advancing Translational Sciences or the National Institutes of Health (NIH). This study made use of the Campus Chemical Instrument Center NMR facility at The Ohio State University, supported by NIH Grant P30-CA-016058.
The authors declare no competing financial interests.
- Correspondence should be addressed to Ruth M. Barrientos at ruth.barrientos{at}osumc.edu
