Postoperative Impairment of Cognitive Function in Rats: A Possible Role for Cytokine-mediated Inflammation in the Hippocampus

Ninety-eight Sprague-Dawley male rats (weighing 300–350 g, approximately 90 days old) were housed individually in a light-, temperature-, and humidity-controlled environment with free access to food and water. Experiments were conducted in accordance with the guidelines for care and use of laboratory animals from the Chinese Academy of Science, Shanghai, China, and from the Home Office, United Kingdom.

Rats were randomly assigned to three groups. Naive controls (group A; n = 14) received no intervention. In the anesthetic group (group B; n = 42), rats were anesthetized with a combination of 20 μg/kg fentanyl and 500 μg/kg droperidol (intraperitoneal). Pilot studies established that rats anesthetized with this anesthetic regimen exhibited surgical analgesia (no purposeful response to supramaximal noxious stimuli) with no evidence of cardiorespiratory dysfunction (as reflected by normal arterial blood gas measurements; data not shown). The anesthesia-plus-surgery group (group C; n = 42) underwent splenectomy during neuroleptic anesthesia (as specified for the anesthetic group). For splenectomy, a small incision was made in the left upper abdominal quadrant and the spleen was mobilized, isolated, and removed; this procedure was chosen because it represents a standardized organ or tissue removal intervention. The wound was infiltrated with 0.25% bupivacaine and then closed by suture. Postintervention animals were recovered and then returned to their cages and housed individually. Rats in groups B and C were killed at day 1, 3, or 7 after intervention (n = 14/time point).

Aversive memory, a hippocampal-dependent place learning,20was evaluated on days 1, 3, and 7 after intervention by an investigator who was blinded to intervention (surgical incision area was covered by tape in all groups) using the modified Y-maze apparatus. The apparatus consists of three opaque-white arms (30 × 5 × 20 cm) connected into a Y shape (fig. 1). The floor of the “stem” of the Y arm and one of the two branch arms was installed with electric wires through which an electric shock (40 ± 5 V for 10 s) can be applied. The branch arm lacking wires remained illuminated throughout the test (because rats are typically neophobic, they prefer to avoid an illuminated area). During the testing, rats were placed in the stem of the Y maze; when animals received an electric shock, they entered into one or the other of the two branch arms. Before the next trial, the rats were manually returned to the stem of the Y maze. The test was stopped when the animal entered into the nonshock (illuminated) arm in 9 out of 10 consecutive trials; at this trial number, the rat was considered to have developed aversive memory.

After behavioral testing, half of the animals in each group were killed by an overdose of pentobarbitone (100 mg/kg intraperitoneal). The hippocampus was quickly removed and immediately cooled in liquid nitrogen and kept at −70°C for messenger RNA (mRNA) and Western blot studies. The remaining animals in each group were killed by an overdose of pentobarbitone (100 mg/kg intraperitoneal) and perfused transcardially with heparinized 0.9% saline followed by 4% paraformaldehyde in 0.1 m phosphate buffer. Brains from these perfused animals were placed in 4% paraformaldehyde in 0.1 m phosphate buffer for in situ  immunohistochemistry.

Wax sections (5 μm) of the hippocampus were immunostained on slides. Briefly, after preincubation with 0.3% hydrogen peroxide in methanol for 30 min to block endogenous peroxidase, the sections were rinsed in phosphate-buffered saline. Glial fibrillary acidic protein (GFAP), S100β, Bax, Bcl-2, and caspase 3 were visualized immunohistochemically using rabbit anti-GFAP antibody (1:80; Sigma, St. Louis, MO), rabbit anti-S100β antibody (1:100; Dako, Carpenteria, CA), Bax, Bcl-2, and caspase 3 (1:100; Delta Biolab, Gilroy, CA), respectively. After incubation for 60 min with the primary antibodies, the sections were washed three times in phosphate-buffered saline at room temperature. The biotinylated secondary antibodies (Sigma) and the avidin-biotin-peroxidase complex (Vectastain Elite ABC-Kit; Vector Lab, Burlingame, CA) were applied (both diluted 1:100), and the sections were incubated for 30 min at room temperature. After the sections were washed with phosphate-buffered saline, the peroxidase reaction was developed by incubating the section in 0.02% 3,3′diaminobenzidine tetrahydrochloride (Sigma) solution containing 0.003% hydrogen peroxide. The sections were then counterstained with Mayer hematoxylin. Finally, the sections were dehydrated through a gradient of ethanol solutions (70–100%). Control sections were processed identically and in parallel; however, they were incubated with phosphate-buffered saline instead of the primary antibodies. No labeling was detected in these controls. For quantification, the number of immunopositive cells per a total of 1,000 in the hippocampal sections was counted at 400× magnification.

Total RNA was isolated from homogenization of the hippocampal tissue samples in Trizol reagent. Chloroform (200 μl) was added to each tissue homogenate. Samples were centrifuged (12,200g , 15 min, 4°C), and the aqueous phase was transferred to a clean tube. An equivalent volume of isopropanol was added to the sample. Samples were further centrifuged (12,200g , 10 min, 4°C), and supernatants were discarded. The pellet was washed with 75% ethanol and centrifuged (7,500g , 10 min, 4°C). Supernatants were discarded, and the pellet was allowed to air dry before reconstitution in sterile DNase and RNase-free H²O.

Total RNA concentration was determined by spectrophotometry at 260 and 280 nm. First-strand complementary DNA (cDNA) was synthesized by random priming using 2 μg total RNA, 50 ng DNA random hexanucleotides, and 200 U Superscript II RNase H-reverse transcriptase (Life Technologies, Gaithersburg, MD). Reverse transcription was performed at 42°C for 60 min and terminated by heating to 95°C for 3 min. Aliquots of cDNA (2 μl) were amplified using 2 U Taq  polymerase in a reaction volume of 50 μl that contained 5 μl 10× PCR buffer, 1.5 mm MgCl², dNTPs (0.2 mm each), and 0.4 mm gene-specific primers for IL-1β or TNF-α. The primers for IL-1β were 5′-GACCTGTTCTTTGAGGCTGAC-3′ and 5′-TCCATCTTCTTCTTTGGGTATTGTT-3′, which amplified a 578-bp product. The primers for TNF-α were 5′-GTAGCCCACGTCGTAGCAAA-3′ and 5′-CCCTTCTCCAGCTGGAAGAC-3′, which amplified a 347-bp product. cDNA was amplified for 33 cycles for TNF-α and 35 cycles for IL-1β. After amplification, a 5-μl aliquot of the polymerase chain reaction products was electrophoresed on 2% (wt/vol) agarose gels. Gels were run at 100 V for 1 h, and the bands were visualized with ethidium bromide and photographed under ultraviolet light using a Gel Doc 2000 System (BioRad Laboratories, Hercules, CA). Band densities were quantified by densitometric measurements of the polymerase chain reaction products Scion Image Beta 4.02 Win (Scion Image, Frederick, MD). The amount of mRNA was expressed as a ratio of densitometric measurements derived from the gene-specific mRNA and β-actin (a housekeeping gene) mRNA.

Hippocampi were homogenized in lysis buffer (5 mm Tris-HCl, 1 mm phenylmethylsulfonyl fluoride, 150 mm NaCl, 1% NP-40, 0.5% Na deoxycholate, 0.1% SDS, and protease and phosphatase inhibitors; pH 7.5). The homogenates were centrifuged for 15 min at 12,000g  at 4°C. The quantity of protein in the supernatants was determined using a BCA protein assay kit (Pierce Biotechnology, Rockford, IL). Protein extracts (60 μg per sample) were denatured in Laemmli sample loading buffer (10% SDS, 0.1 m Tris pH 8.0, 50 mm DTT, 3 mm EDTA, 0.001% bromphenol blue) at 100°C for 5 min, separated by 15% SDS-PAGE and transferred to a nitrocellulose membrane (Amersham Biosciences, Piscataway, NJ). After blocking with 5% skim milk in TBS-T (0.1% Tween 20, 20 mm Tris-HCl, 137 mm NaCl, pH 7.6) for 2 h, the membrane was incubated with rabbit anti-TNF-1α (1:800) or anti-IL-1β (1:1,000; Sigma, St. Louis, MO) in TBS-T overnight at 4°C. Subsequently, membranes were incubated for 2 h with goat anti-rabbit HRP-conjugated secondary antibody (Amersham Pharmacia Biotech) at room temperature. The membranes were treated with the ECL system (Amersham) and then were developed on film. The bands were analyzed by densitometric analysis (Kodak Digital Science and 1D Image Analysis Software; Rochester, NY). The membrane blots were stripped and reprobed with polyclonal anti-β-actin (20-33; Sigma). Relative expression levels of TNF-1α and IL-1β protein were normalized by the ratio of TNF-1α and IL-1β to β-actin, respectively.

All data were analyzed by an observer who was blinded to the experimental protocol. The results are expressed as mean ± SD. The data were analyzed with one-way analysis of variance, followed by Fisher post hoc  test where necessary. A P  value of less than 0.05 was considered to be statistically significant.

To elucidate the effect of surgery on learning and memory, we conducted the Y-maze test, which is used widely to evaluate spatial learning and memory in rodents.18In naive animals, it took 28 ± 7 trials for them to remember the impending risk of being shocked. The anesthetic regimen used in this study did not affect their learning ability, with the number of trials (28 ± 7, 29 ± 6, and 29 ± 7 at days 1, 3, and 7 after experiments, respectively) being no different than that seen in the control group. However, on days 1 and 3 after surgery, the number of trials of the test was significantly increased to 67 ± 26 and 60 ± 18 (P < 0.01 vs.  control) in the surgical animals. Although the number of trials was still higher in the splenectomized animals (41 ± 11) on day 7 after surgery, this did not differ statistically from that of naive control (P > 0.05 vs.  control; fig. 1). These results suggest that surgery induces a temporary impairment in spatial learning and memory.

Glial fibrillary acidic protein immunoreactivity was increased in the splenectomized rats when compared with the naive and anesthetized rats that did not undergo splenectomy. The GFAP-positive cells increased from 99 ± 24 in controls to 228 ± 39 (P < 0.01) and 169 ± 18 (P < 0.05) at days 1 and 3 after surgery, respectively (fig. 2). A similar pattern change was seen with S100β; S100β-positive cells were significantly increased from 106 ± 10 in controls to 203 ± 18 (P < 0.01) and 193 ± 60 (P < 0.05) at days 1 and 3 after surgery, respectively (fig. 3).

The anesthetics did not increase IL-1β mRNA expression at any time point when compared with the naive controls (P > 0.05). However, on days 1 and 3 after splenectomy, the level of IL-1β mRNA was increased by 67% (P < 0.01) and 46% (P < 0.05), respectively, when compared with the naive controls. Protein expression of IL-1β was significantly increased by 40% on the first postsurgical day (P < 0.05) but not at other time points (fig. 4).

The general anesthetic did not induce an increase of TNF-α mRNA expression at any time point when compared with the naive controls (P > 0.05). However, on postsurgical day 1, the level of TNF-α mRNA was increased by 38% (P < 0.05), but no difference in transcription was noted at days 3 and 7. There was no change in the expression of TNF-α protein at any time after surgery (fig. 5).

There was almost no Bcl-2 and Bax staining in hippocampal sections from controls and animals treated with anesthesia without surgery; a strong cytoplasmic signal for both Bax and Bcl-2 was detected in neurons in the CA1 area of the hippocampus in postsplenectomy rats (fig. 6). On days 1, 3, and 7, the numbers of Bax-positive cells were 378 ± 72, 272 ± 48, and 73 ± 1, respectively. The numbers of Bcl-2–positive cells on postoperative days 1, 3, and 7 were 214 ± 25, 216 ± 31, and 62 ± 9, respectively. When considered as a ratio of Bcl-2:Bax, the values on postoperative days 1, 3, and 7 were 0.57 ± 0.12, 0.79 ± 0.13, and 0.85 ± 0.09, respectively. Caspase 3 was not detected in any section examined.

We investigated the association of an inflammatory response in the hippocampus with the development of POCD and found that after surgery (splenectomy) under neuroleptic general anesthesia (fentanyl-droperidol), rats exhibited impaired memory that was temporally associated with glial activation and proinflammatory cytokine expression in the hippocampus; neither cognitive dysfunction nor neuroinflammatory changes were noted in rats that were subjected only to neuroleptic general anesthesia.

Postoperative cognitive dysfunction affects postoperative surgical patients irrespective of the anesthetic technique used.21–23POCD is particularly prevalent in cardiac surgical patients but also occurs in noncardiac surgical patients in whom putative mechanisms such as embolism, hypotension, and hypoxemia do not obtain.5Although especially frequent in elderly patients,24the incidence depends to a large degree on the sensitivity of the tests that are used to detect the cognitive deficit. In fact, the unequivocal establishment of POCD as a distinct condition has been questioned because its occurrence in the postoperative period has not been prospectively compared with an appropriate control group, e.g. , patients with a surgical condition who are not subjected to surgery and anesthesia.25Intercurrent illness (both medical and surgical) may be associated with “sickness behavior,” one component of which is cognitive dysfunction.26A series of studies have now demonstrated the pivotal role that neuroinflammation exerts to impair the hippocampal-dependent memory that is part of sickness behavior.6,26,27 

Previously, it had been speculated that proinflammatory cytokines could play a role in the development of cognitive decline that may acutely follow surgery.28The peripheral IL-6 response to surgical injury exemplifies the vast literature on the proinflammatory cytokine response to surgery. Within 2 h of skin incision, IL-6 levels increase, peak between 4 and 12 h, and remain elevated for up to 3 days postoperatively.29–31Extent of surgical trauma is an important factor; there was a greater degree of IL-6 elevation after abdominal aortic and colorectal procedures (mean peak concentrations, 400–550 U/ml, ranging as high as 1,220 U/ml) than after hip replacement (208 U/ml) despite similar surgical procedure times.30Further, four separate studies illustrate that the increment in IL-6 is less after laparoscopic versus  open cholecystectomy, despite similar surgical times.32–35The IL-6 response is not modified by anesthetic regimen36,37but is decreased by use of antiinflammatory analgesic agents.38 

Cytokines originating from the periphery can exert effects within the central nervous system (CNS) through both direct and indirect means. Both IL-139and TNF-α40have been shown to gain entry into the CNS through the relatively permeable blood-brain barrier in the periventricular regions. Further, IL-1β can also directly bind to its cognate receptors on the endothelial cells within the brain microvasculature where these can elaborate a central inflammatory response.13Indirectly, cytokines can induce changes within the CNS through vagal afferent nerves.41In an elegant series of studies, either lipopolysaccharide42or IL-1β,43when administered peripherally, induced brainstem and hippocampal IL-1β production and CNS expression of IL-1, IL-6, and TNF-α, all of which could be abolished by vagotomy.

Within the CNS, the glia can be activated by above-mentioned events that are initiated in the periphery.44Glial activation and the consequent release of proinflammatory cytokines within the hippocampus interfere with cognitive function as evidenced by abnormal memory and learning in the intact organism and/or inability to develop long-term potentiation in hippocampal slice preparation.45Interestingly, cytokines seem to have a biphasic effect on long-term potentiation, the synaptic plasticity marker that is the neurobiologic correlate of memory; low concentrations seem to enhance long-term potentiation, whereas higher concentrations impair memory. Furthermore, the memory-enhancing and memory-impairing effects of the proinflammatory cytokine TNF-α seem to be mediated by different receptor subtypes.46 

Within the intrinsic apoptotic pathway, interactions between proapoptotic members of Bcl family seem to play a critical role in determining the ultimate fate of injured cells.47Neurons in the hippocampus of splenectomized rats increased expression of both Bcl-2 and Bax. However, the ratio of Bcl-2:Bax was decreased after surgery, suggesting that neurons in the hippocampus are inclined toward apoptosis48; however, caspase-3 activation, a marker of apoptotic cell death, was not detected in splenectomized rats at the time points examined. We speculate that the cognitive dysfunction that we observed is not necessarily associated with the presence of actual cell death in hippocampus; rather, there may be abnormal interaction between neurons and glial cells in the hippocampus.49,50Such hippocampal dysfunction in the absence of morphologic injury has been recently noted after induction of a proinflammatory cytokine response.51 

There are several caveats that must be considered. Splenectomy itself can alter the inflammatory response of an organism; however, this results in a decrease in inflammation rather than a proinflammatory state.52The abnormality in cognitive function did not persist up to 7 days postoperatively; therefore, it is possible that our observed phenomenon is more reflective of the cognitive impairment associated with delirium, a postoperative state that does not necessarily progress to POCD.53Possibly, an additional provocation (e.g. , infection, advanced age, or underlying subclinical neuropathology) is required to convert this phenomenon into a more permanent cognitive impairment. At this juncture, it is not possible to establish that the cognitive impairment is causally related to the inflammatory response and whether the proinflammatory cytokine expression and/or transcription initiated, or followed, glial activation.

In summary, we have demonstrated that surgery triggers a transient neurocognitive decline in a rat model that is temporally associated with glial activation and increase in proinflammatory cytokines in the hippocampus. Although further data are needed to establish the robustness of these observations (including other anesthetic regimens and surgical procedure), we interpret these data to mean that glial activation and consequent release of proinflammatory cytokines in the CNS (noted with a variety of “stressors”) may also occur after the “stress” of a surgical intervention. Strategies designed to mitigate the central neuroinflammatory component of surgery may represent a means of preventing postoperative cognitive dysfunction.