Depletion of Bone Marrow–derived Macrophages Perturbs the Innate Immune Response to Surgery and Reduces Postoperative Memory Dysfunction

• Animal models of postoperative cognitive dysfunction implicate an innate immune response, with increased circulating inflammatory mediators and migration of bone marrow–derived cells into the brain

• Inhibitors of inflammatory mediators reduce postoperative memory deficits in mice but also reduce wound healing

• In mice, treatment with a drug that depletes bone marrow–derived macrophages reduced circulating inflammatory mediators after surgical orthopedic injury, reduced migration of immune cells into the brain, and reduced postoperative memory deficits

ACUTE postsurgical memory deterioration leads to persistent cognitive decline1that can result in considerable morbidity and increased mortality.2The specter of memory dysfunction, including acute delirium, postoperative decline, and dementia, is a source of anxiety for patients and their families.3Knowledge about the molecular and cellular pathways involved in postoperative memory dysfunction may provide a launching pad for the development of biomarkers to identify the most vulnerable patients as well as preventive strategies.

Using a murine model of aseptic surgical trauma with a long-bone fracture, we previously demonstrated that postoperative cognitive decline requires the engagement of the innate immune response. This engagement includes increased systemic expression of alarmins and proinflammatory cytokines such as interleukin (IL)-6 in the blood4,5; increased ratio of CD11b⁺cells corresponding to macrophages/microglia cells,4and specifically the ratio of CCR2⁺bone marrow-derived macrophages6; and elaboration of proinflammatory cytokines that are capable of disrupting hippocampal long-term potentiation, a neurobiologic correlate of learning and memory.3,7–10 

Strategies designed to block the effect of proinflammatory cytokines with IL-1 receptor antagonist (anakinra) or tumor necrosis factor (TNF)-α antibody (etanercept) prevented murine postoperative memory dysfunction.4,5These interventions also prevented inflammation-dependent wound healing.11,12 

In this study, we tested the hypothesis that mediation of postoperative memory decline requires recruitment of systemic bone marrow–derived macrophages into the brain, using a specific pharmacologic strategy to acutely deplete systemic phagocytes before an aseptic surgical trauma with an experimental tibial fracture.

All experimental procedures involving animals were approved by the University of California, San Francisco Institutional Animal Care and Use Committee, and conformed to National Institutes of Health guidelines. Twelve 8- to 12-week-old CCR2  ^RFP/+ CX3CR1  ^GFP/+ male mice6,13(fig. 1A) were used to identify bone marrow–derived macrophages. CCR2 and CX3CR1 are acronyms for chemokine (C-C motif) receptor 2 (whose cognate ligand is monocyte chemoattractant protein [MCP]-1) that is highly expressed in bone marrow–derived macrophages, and CX3C chemokine receptor 1 (CX3CR1, fractalkine receptor) that is highly expressed in resident microglia. Eighty-nine wild-type male mice (C57BL/6J, 10–12 weeks old) were purchased from The Jackson Laboratory (Bar Harbor, ME): 29 for the cytokine expression (fig. 1B) and 70 for the behavior tests (fig. 1C). Mice did not experience unexpected lethality in the study and were euthanized according to our institutional animal care and use committee guidelines.

Clodrolip is a liposomal formulation of clodronate (dichloromethylene bisphosphonic acid), a nontoxic bisphosphonate. Liposomes are lipid vesicles consisting of concentric phospholipid bilayers surrounding aqueous compartments. In this case, liposomes are used as “Trojan horses” encapsulating clodronate, which are then ingested and digested by phagocytes, followed by an intracellular release and accumulation of clodronate. At a certain intracellular concentration, clodronate induces apoptosis of the phagocytes. Clodronate liposomes were obtained from clodronateliposomes.org**(Vrije Universiteit, Amsterdam, The Netherlands) at a concentration of 7 mg/ml and prepared as described previously.14,15Clodrolip (200 μl, approximately 100 mg/kg) was injected intraperitoneally 60 min before the bone fracture. Control animals received 200 μl of control liposomal solution (CT-lip). No intraperitoneal or extraperitoneal damage was observed after clodrolip intraperitoneal administration.

Anesthesia was induced and maintained with isoflurane by inhalation. We used a dedicated chamber for induction with 5% isoflurane for 3 min, and the operation was performed under 2% isoflurane for 10–12 min. Under aseptic surgical conditions, an open tibial fracture of the right hind limb with intramedullary fixation was performed as described previously.4Body temperature was maintained at 37° ± 0.5°C using a thermal blanket throughout the surgical procedure, and analgesia was provided by injection of buprenorphine (0.3 mg in 100 μl of saline). Sham mice for bone fracture (sham group) received the same anesthesia and analgesia as the bone fracture mice.

Mouse blood was collected using cardiac puncture under general anesthesia (isoflurane, 3%) in separate cohorts 12 and 24 h after the bone fracture procedure (fig. 1B). Blood samples were centrifuged at 1300 rpm for 10 min at room temperature, and the serum was collected and frozen at −80°C. IL-6 is secreted by bone marrow–derived macrophages in response to alarmins,16and the IL-6 level in the serum is increased within the first 24 h after the tibia fracture.4,5The IL-6 level in the serum is also associated with the postoperative memory dysfunction phenotype5and is affected by clodrolip in response to lipopolysaccharide infusion.17For these reasons, we decided to quantify IL-6 levels in the serum of mice exposed to clodrolip or CT-lip using the IL-6 enzyme-linked immunosorbent assay kit (KMC0062; Invitrogen, Grand Island, NY). Results are expressed as fold increase compared with that measured in five control mice that did not receive any treatment or surgery.

The hippocampi of the mice were collected rapidly under a dissecting microscope, 12 and 24 h after the tibia fracture (fig. 1B), and placed in RNAlater solution (Qiagen, Valencia, CA). To avoid blood contamination, mice were perfused with saline for 5 minutes before sample collection. Total RNA was extracted using the RNeasy Lipid tissue Kit (Qiagen) treated with recombinant DNase I using a RNase-Free Dnase set (Qiagen), and reverse-transcribed to complementary DNA with a High Capacity RNA to Complementary DNA Kit (Applied Biosystems, Bedford, MA). TaqMan Fast Advanced Master Mix (Applied Biosystems) and gene-specific primers and probes used for quantitative polymerase chain reaction are as follows: β-actin (NM_007393.1), IL-6 (Mm00446190_m1), TNF-α (Mm00443258_m1), IL-1β (Mm01336189_m1), and MCP-1 (Mm00441242_m1). Quantitative polymerase chain reaction was performed using StepOnePlus (Applied Biosystems). Each RNA sample was run in triplicate, and relative gene expression was calculated using the comparative threshold cycle (δCT) method and normalized to β-actin. Results are expressed as fold increase compared with that observed in five control mice that did not receive any treatment or surgery.

Twenty-four hours after the tibia fracture surgery, the brain and spleen of the CCR2  ^RFP/+ CX3CR1  ^GFP/+ mice were collected after intracardiac perfusion with paraformaldehyde 4% (fig. 1A). Spleen and brain (bregma, −1.0 to −1.4 mm, corresponding to interaural 2.7 to 2.3 mm in coronal orientation) were sectioned into 20-μm-thick slices and mounted with Vectashield DAPI (Vector Laboratories, Burlingame, CA). The expression of CCR2-RFP and CX3CR1-GFP cells was assessed using confocal images, performed with a Spectral Confocal microscope (Nikon Instruments, Melville, NY) using three laser lines (405, 488, and 561 nm). Z-stacks were rendered into a three-dimensional image using the NIS-Elements AR 3.0 software (Nikon), and the expression of CCR2-RFP and CX3CR1-GFP cells was quantified using ImageJ (National Institutes of Health, Bethesda, MD), with three different photographs per mouse taken with a 20× objective. Data are expressed as relative cell percentages normalized to the average value of the CT-lip group.

Fear conditioning is used to assess memory in rodents, which are trained to associate a conditional stimulus, such as a conditioning chamber, with an aversive, unconditional stimulus, such as a foot shock. Freezing behavior is an indicator of aversive memory that is measured when subjects are reexposed to the conditional stimulus. With this model, lesions of the hippocampus disrupt recall of fear responses to the presentation of the context, resulting in a diminution in freezing.18,19 

For this study, we used a previously published paradigm.4–6,20Briefly, the behavioral study was conducted using a conditioning chamber (Med Associates, Inc., St. Albans, VT) and an unconditional stimulus (two periods of foot shock of 0.75 mA during 2 s). An infrared video camera, mounted in front of the chamber, captured motion speed (Video Freeze; Med Associates).

All of the animals underwent the same training session, regardless of the specific intervention, and received their training 30–40 min after the liposomal intraperitoneal injections (whether clodrolipid or CT-lip) that occurred 30 min before surgery (fig. 1C). Three days after conditioning, mice were returned to the same chamber where training had occurred for a context test. During the context test, mice were exposed just to the context and no tones or foot shocks were delivered. Freezing was recognized by the software as a total lack of movement, excluding breathing and movement of vibrissae (linear detection with a minimal freeze duration of 20 frames corresponding to 0.7 s and a motion threshold of 20 arbitrary units).4–6,20Decrease in the percentage of time spent freezing indicated impairment of memory.

The body weight of the animals was measured 3 days after surgery, following assessment of freezing behavior. An infrared video camera (Video Freeze) captured and quantified motion speed during the context test, and the maximal motion speed was recorded for each mouse.

Data are presented as mean ± 95% CI. Normality was tested with the d’Agostino–Pearson omnibus normality test. Equality of variances was tested with the F test. For two-sample comparisons, Student t  tests were used (using the Welch correction if necessary); Mann–Whitney U  tests were used if data were not normally distributed. For comparisons of more than two groups, means were compared using one-way ANOVA followed by Student t  tests with a Bonferroni-corrected alpha level.

We used the two-way ANOVA procedure to determine whether or not time and treatment were significant factors in predicting IL-6 concentration in the serum, and IL-6, IL-1β, TNF-α, and MCP-1 messenger ribonucleic acid (mRNA) expression in the hippocampi. Given the highly skewed nature of the mRNA expression, we checked the distribution of the residuals. We applied a log transformation (ln[X]) to the response of the mRNA expression before performing analysis to better adhere to the ANOVA model’s assumptions of normally distributed residuals and homoscedasticity of residuals.

For the behavior tests, animals were tagged and allocated randomly to each group before any treatment, and researchers were blinded to the group assignment that was revealed only after the analysis phase. A repeated measures ANOVA was performed to determine whether treatment (CT-lip and clodrolip) and the three time periods (baseline, first shock, and second shock) were significant predictors of percentage freezing time during the training session.

For this study, our primary outcome was percentage of freezing time during the context session. Based on previous freezing time data,4we estimated that a sample of 18 C57BL/6J surgical mice per group was necessary to demonstrate a 20% increase in percentage freezing time, with 80% power at the 0.017 alpha level (after adjusting for three comparisons) to find a significant difference between clodrolip and CT-lip.

A two-tailed value of P < 0.05 was considered statistically significant for two-group comparisons, and the significance threshold was adjusted for multiple comparisons with a Bonferroni correction. Prism 5 (GraphPad Software, Inc., La Jolla, CA) was used to conduct the statistical analyses.

Using CCR2  ^RFP/+ CX3CR1  ^GFP/+ mice (fig. 1A), in which RFP+ bone marrow–derived macrophages and GFP+ resident microglia can be tracked,6,13we found that clodrolip depleted splenic macrophages and surgery-induced bone marrow–derived macrophage infiltration into the hippocampus. The CCR2⁺cells, which are mainly present in the splenic red pulp (fig. 2A), decreased by 96% in the clodrolip-exposed mice (fig. 2B) (95% CI, 95–97%, P < 0.001). As shown in figure 3, the number of CCR2⁺cells was also significantly reduced in the hippocampi of clodrolip-treated mice compared with CT-lip–treated mice 24 h after surgery (decrease of 76% for the dentate gyrus and 87% in the cornu ammonis 3). However, clodrolip treatment did not change the number of CX3CR1⁺cells in the dentate gyrus and cornu ammonis 3 hippocampal regions (fig. 3).

We previously showed that proinflammatory cytokines in the blood and hippocampus increased within the first day after surgery.4To test whether clodrolip treatment would reduce the proinflammatory cytokines, we studied serum and hippocampal expression 12 and 24 h after surgery (fig. 1B). Twelve hours after surgery, the rise in IL-6 in the serum was significantly attenuated in mice exposed to clodrolip (two-way ANOVA, P = 0.004 for the treatment, P = 0.003 for the time effect, and P = 0.19 for interaction) (fig. 4).

Between 12 and 24 h after surgery, the increase in mRNA hippocampal expression of IL-6, TNF-α, and IL-1 induced by surgery returned to almost baseline values at 24 h (fig. 5). Clodrolip exposure significantly inhibited the surgery-induced increased expression of mRNA IL-6 (two-way ANOVA, P < 0.001 for the treatment, P = 0.002 for the time effect, and P = 0.51 for interaction), and interacted with the time-dependent decrease for TNF-α (two-way ANOVA, P = 0.03 for interaction). Clodrolip treatment did not change IL-1β mRNA expression (two-way ANOVA, P = 0.42 for the treatment, P < 0.001 for the time effect, and P = 0.66 for interaction) (fig. 5).

Systemic macrophages are recruited into tissues by the chemoattractant MCP-1 that binds to CCR2, which is expressed on the surface of bone marrow–derived macrophages.21,22Following surgery, MCP-1 mRNA expression increases and is unaffected by prior exposure to clodrolip (two-way ANOVA, P = 0.64 for the treatment, P < 0.001 for the time effect, and P = 0.67 for interaction) (fig. 5D).

During the preoperative training period, learning was similar in the clodrolip-exposed and the control (nonexposed) groups, with the percentage of freezing being highly associated with time (fig. 6). During the context session, surgery significantly decreased percentage of freezing time in comparison with the sham group (52% [95% CI, 41–63%] vs . 29% [95% CI, 21 to 37%], P = 0.0012); preoperative exposure to clodrolip resulted in significantly greater freezing time than in the nonexposed surgical cohort (29% [95% CI, 21–38%] vs . 48% [95% CI: 38–58%], P = 0.004), reaching a level similar to that observed in the sham-operated clodrolip-exposed mice (49% [95% CI, 36–63%], P = 0.86) (fig. 7, Aand B).

Clodrolip did not affect the body weight of the mice 3 days after the injection (fig. 7C). As for maximal motion speed, the clodrolip-treated groups were no different from the CT-lip groups, even though the maximal motion speed of the surgical groups was significantly slower than the sham groups (fig. 7D).

In this study, we report for the first time that bone marrow–derived macrophages are required in the pathogenesis of the neuroinflammatory and memory dysfunction induced by surgery. Also, we report that a possible hippocampal signal through MCP-1 is involved in the recruitment of bone marrow–derived macrophages to this brain region. Data from rodent surgical models have provided insight into the neuroinflammatory basis for postoperative cognitive decline. This usually transient process appears to be part of a motivational system that reorganizes the organism’s priorities to facilitate recovery. To date, we have established a pivotal early role for the proinflammatory cytokine TNF-α,6and our study demonstrated that hippocampal infiltration of bone marrow–derived macrophages also plays a role in the initiation of neuroinflammation.

Monocyte infiltration into the brain is mainly described in acute brain injuries such as stroke23and traumatic brain injuries,24as well as chronic inflammatory brain injuries such as multiple sclerosis.13Using long-bone fracture as a surrogate for a peripheral orthopedic surgical insult, we previously reported that CCR2⁺cells were present in the hippocampus.6Because microglia can also express CCR2 under certain conditions,25,26we could not ascertain whether these CCR2-expressing cells arose from the resident macrophage population (microglia) or through an infiltration from outside of the central nervous system. Using clodrolip to specifically deplete the systemic pool of phagocytes, including bone marrow–derived macrophages, we were able to demonstrate that the CCR2⁺cells in the hippocampus are a result of the recruitment of bone marrow–derived macrophages into the brain.

For passage into the brain, monocytes are required to overcome the blood–brain and/or blood–cerebral spinal fluid barrier27,28; these barriers can be disrupted by direct acute brain injury.29,30Interestingly, after peripheral surgery, the blood–brain barrier is disrupted, although there is no discernible brain lesion.6Now we show that after surgery, the hippocampus expresses MCP-1, which is capable of attracting CCR2⁺-expressing cells migrating through the disrupted blood–brain barrier. This increased expression of MCP-1 is unaffected by clodrolip treatment, indicating that bone marrow–derived macrophages are not a self-perpetuating source of this chemoattractant for its own recruitment. Future understanding of the source and the triggers for hippocampal MCP-1 following peripheral surgery may result in interventional strategies designed to prevent recruitment of bone marrow–derived macrophages into the brain.

Our recent data suggest that transient hippocampal inflammation is the key element in postoperative memory dysfunction because (1) hippocampal areas are known to be involved in memory tasks7; (2) hippocampal neuroinflammation profile correlates with the level of memory dysfunction4,5; and (3) hippocampal neuroinflammation leads to long-term potentiation disruption.31,32Now we report here that the absence of bone marrow–derived macrophage infiltration, produced by systemic depletion by clodrolip, decreases surgery-induced hippocampal inflammation and memory dysfunction. Therefore, postoperative bone marrow–derived macrophage recruitment into the hippocampus plays a key role in the initiation of postoperative memory dysfunction.

In the context of postoperative cognitive decline, determining which cells are involved in the initiation of the inflammation response is important because, when exaggerated, this could overwhelm resolving responses and produce persistent postoperative cognitive decline. Earlier, we described that surgical trauma induces systemic release of alarmins (i.e ., high-mobility group protein 1) and proinflammatory cytokines (i.e. , TNF-α and IL-6).4,5Improving our knowledge of cellular and molecular initiation mechanisms will allow insight into an ex vivo  bioassay to prospectively determine whether patients are at risk.

We used experimental tibial fracture to generate animal postoperative memory acute dysfunction. With this model, we traumatized the bone marrow directly, which could play a key role. However, other models that did not damage bone marrow with a splenectomy33also showed that surgery generated postoperative cognitive dysfunction.

Fibrin is deposited in the hippocampus after tibia fracture,6suggesting that the blood–brain barrier becomes disrupted and may allow the passage of clodrolip to act directly on the microglia population.34However, we found that the systemic administration of clodrolip acts only on the number of CCR2⁺cells without significantly affecting the number of CX3CR1⁺cells (fig. 3); if clodrolip has an effect on microglia, it may be to functionally modify them. For this reason, we cannot exclude the possibility that clodrolip does not affect the function of microglia, and that microglia do not play a key role in postoperative cognitive dysfunction.

For this study, we used a pharmacologic strategy to quickly deplete the pool of systemic macrophages. However, because clodrolip is highly toxic for monocytes and macrophages,35it can increase the risk of postsurgical infections, generating a phenotype of its own. With a single dose, we did not observe loss of weight or other signs of sickness within the 3 days. We performed a very short-term study, focusing on the acute exaggeration phase of neuroinflammation, and did not perform any long-term study with clodrolip. Clodrolip should be considered as a tool for mechanistic studies but cannot be proposed for clinical therapy.

To distinguish whether these recruited cells were resident or recruited systemic macrophages, we used CCR2  ^RFP/+ CX3CR1  ^GFP/+ mice. CCR2 is receptor for MCP-1 and is mainly expressed in bone marrow–derived monocytes–macrophages. We previously showed that CD11b⁺macrophages–microglia cells were recruited in the hippocampus after tibial fracture.4In this study, however, we did not determine that CCR2⁺cells were only bone marrow–derived macrophages. Further study on the role of other systemic phagocytes, including neutrophils, in our phenotype will be of most interest.

Hippocampal cytokine expression was performed with mRNA and not with protein. This is a limitation, but we considered that the potential extravasation from blood may affect protein levels. Indeed, in this model we found blood–brain barrier leakage after the tibia fracture,6and the increase of circulating IL-6 protein in the serum could contaminate the hippocampal samples with passive movement in the parenchyma (this phenomenon could be amplified by the perfusion itself). By analyzing mRNA expression in the brain collected after purging blood from the vessels, we ensured that hippocampal cells were the source of proinflammatory cytokines.

In conclusion, we showed in this study that bone marrow–derived macrophage activation after experimental tibial fracture is directly involved in the tibia fracture–induced hippocampal bone marrow–derived macrophage infiltration and animal memory dysfunction. Understanding the cellular and biologic pathways involved in postoperative cognitive decline is a key element in designing interventions to prevent this disease. Reducing activation and/or migration of innate immune cells, such as systemic macrophages, into the brain represents a viable preemptive strategy.

The authors thank members of the University of California, San Francisco Center for Cerebrovascular Research and the Maze Laboratory for their support; Niccolo Terrando Ph.D., Assistant Professor, Karolinska Institute, Stockholm, Sweden, for assistance with setting up memory tests; and Israel F. Charo, M.D., Associate Director, University of California, San Francisco, Gladstone Institute, San Francisco, California, for providing the CCR2  ^RFP/+ CX3CR1  ^GFP/+ mice.