TNF-α–secreting monocytes are recruited into the brain of cholestatic mice

Authors

  • Steven M. Kerfoot,

    1. Liver Unit, Immunology Research Group, Health Sciences Center, University of Calgary, Calgary, Alberta, Canada
    Search for more papers by this author
  • Charlotte D'Mello,

    1. Liver Unit, Gastroenterology Research Group, Health Sciences Center, University of Calgary, Calgary, Alberta, Canada
    Search for more papers by this author
  • Henry Nguyen,

    1. Liver Unit, Gastroenterology Research Group, Health Sciences Center, University of Calgary, Calgary, Alberta, Canada
    Search for more papers by this author
  • Maureen N. Ajuebor,

    1. Liver Unit, Gastroenterology Research Group, Health Sciences Center, University of Calgary, Calgary, Alberta, Canada
    Search for more papers by this author
  • Paul Kubes,

    1. Liver Unit, Immunology Research Group, Health Sciences Center, University of Calgary, Calgary, Alberta, Canada
    Search for more papers by this author
    • P. K. is a CIHR and an AHFMR Scientist.

  • Tai Le,

    1. Liver Unit, Gastroenterology Research Group, Health Sciences Center, University of Calgary, Calgary, Alberta, Canada
    Search for more papers by this author
  • Mark G. Swain

    Corresponding author
    1. Liver Unit, Gastroenterology Research Group, Health Sciences Center, University of Calgary, Calgary, Alberta, Canada
    • University of Calgary, Calgary, Alberta, Canada T2N 4N1
    Search for more papers by this author
    • fax: 403-270-0995

    • M.G. S. is a CIHR Investigator and an Alberta Heritage Foundation for Medical Research (AHFMR) Senior Scholar.


  • See Editorial on Page 20

  • Potential conflict of interest: Nothing to report.

Abstract

Signaling occurs between the liver and brain in cholestatic liver disease, giving rise to sickness behaviors such as fatigue. However, the signaling pathways involved are poorly defined. Circulating inflammatory mediator levels are increased in cholestasis, leading to speculation that they may be capable of activating circulating immune cells that subsequently could gain access to the brain. Indeed, we have identified that at day 10 after bile duct resection–induced cholestasis, there is activation of circulating monocytes that express tumor necrosis factor α (TNF-α) in conjunction with increased expression of adhesion molecules by cerebral endothelium. Moreover, using intravital microscopy, we have identified markedly enhanced leukocytes rolling along cerebral endothelial cells, mediated by P-selectin, in bile duct–resected (BDR) but not control mice. In addition, we have identified increased infiltration of monocytes (but not lymphocytes) into the brains of BDR mice and found that these infiltrating monocytes produce TNF-α. Furthermore, infiltration of TNF-α–secreting monocytes into the brains of cholestatic mice is associated with a broad activation of resident brain macrophages to produce TNF-α. In conclusion, cholestasis is associated with an activation of cerebral endothelium that recruits TNF-α–producing monocytes into the brain. We hypothesize that enhanced TNF-α release within the brain may contribute to the development of cholestasis-associated sickness behaviors, including fatigue. (HEPATOLOGY 2006;43:154–162.)

Nonspecific symptoms—collectively termed sickness behaviors—are commonly associated with several disease states, including liver disease.1–4 These symptoms typically include fatigue, malaise, listlessness, anorexia, decreased social interaction, and difficulty concentrating.1, 3, 4 In cholestatic liver disease, including primary biliary cirrhosis, sickness behaviors are commonly encountered. Specifically, fatigue occurs in up to 86% of patients with primary biliary cirrhosis, is often the presenting symptom, and can be incapacitating.5, 6 However, the causes of sickness behaviors associated with liver disease are poorly understood and have received limited scientific attention.

It has become increasingly clear from clinical and animal experimental observations that sickness behaviors in cholestasis originate from changes in neurotransmission within the central nervous system.7–9 However, one question that clearly stands out is: How does the cholestatic syndrome, with its complex combination of liver damage and retention of substances in blood that are normally secreted in bile, lead to changes within the central nervous system (CNS)? Moreover, the pathways that link the peripheral (i.e., outside the CNS) changes occurring in cholestasis with those that occur within the CNS have not been explored but are of significant potential importance.

Traditionally, communication between the periphery and the CNS has been considered to involve neural (i.e., nerve projections, principally the vagus nerve) and/or humoral (i.e., substances within the circulation, principally cytokines) pathways.10 The liver and peritoneum are richly inervated by the vagus nerve, and stimulation of vagal nerve afferents can signal the brain to induce sickness behaviors.11–14 In addition, sepsis or the systemic administration of endotoxin or cytokines (e.g., interleukin 1, interleukin 6, tumor necrosis factor α [TNF-α]) can also induce sickness behaviors in humans and animals,15–17 possibly by stimulating cerebral endothelial cells to release secondary messengers (e.g., PGE2) into the brain.15, 16 It is likely that both of these pathways participate in the genesis of cholestasis-associated sickness behaviors.

However, we hypothesized that in cholestatic liver disease a third communication pathway between the peripheral circulation and the CNS may exist. Cholestasis is associated with the retention (or secretion) of a number of substances into the blood that are capable of activating immune cells as well as endothelium. These substances include endotoxins and cytokines.18–21 Therefore, we speculated that cholestatic liver damage may be associated with the activation of the endothelial cells that form the blood–brain barrier, thereby resulting in the recruitment of immune cells (e.g., monocytes) from the bloodstream into the CNS. Immune cells entering the CNS, if they were activated and secreted cytokines, could represent a potentially important communication pathway between the circulation and the brain in cholestatic liver disease. Therefore, we designed the following series of experiments to test this hypothesis.

Abbreviations

TNF-α, tumor necrosis factor α; BDR, bile duct–resected; CNS, central nervous system; VCAM-1, vascular cell adhesion molecule 1.

Materials and Methods

Model of Cholestasis.

Male C57BL/6 mice (Jackson Laboratories, Bar Harbor, ME) 8 weeks of age were housed in a light-controlled room maintained at 22°C with a 12-hour day/night cycle and were given free access to food and water. All animals were treated humanely under University of Calgary Animal Care Committee guidelines, and all experiments were performed in accordance with the guidelines of the Canadian Council on Animal Care.

The model of cholestasis used was the well-characterized model of obstructive cholestasis due to bile duct ligation and resection as previously described.22 Briefly, laparotomy was performed under halothane anesthesia and the bile duct was isolated, doubly ligated, and resected between the ligatures. Sham resection consisted of laparotomy and bile duct identification and manipulation without ligation or resection. Experiments were performed 10 days after surgery, at which time bile duct–resected (BDR) mice were overtly cholestatic.

Cerebral Endothelial Cell Isolation and Flow Cytometry Analysis.

Mice were anesthetized with halothane and their brains were perfused in ice-cold phosphate-buffered saline. Cerebral endothelial cells were then isolated as described by Tontsch and Bauer.23 Briefly, brains were removed, and cortices were dissected free and rinsed in ice-cold sucrose buffer (0.32 mol/L sucrose, 3 mmol/L HEPES [pH 7.4]), cleared from the pia with fine forceps, and minced into small pieces. Brain tissue was then homogenized using a hand-held tissue Dounce homogenizer in 3 volumes of ice-cold sucrose buffer. The homogenate was diluted fourfold with ice-cold sucrose buffer and centrifuged at 100g for 10 minutes at 4°C. The pellet was homogenized again in 3 volumes of ice-cold sucrose buffer and cleared through a series of centrifugation steps as described.23 The final pellet contained exclusively microvessels that were dissociated with 0.075% collagenase (Type 1; Sigma, Mississauga, Ontario, Canada) in Ca2+- and Mg2+-free phosphate-buffered saline by shaking for 10 minutes at room temperature. The resulting cells were centrifuged at 200g for 10 minutes and the cell pellet was used for FACS (FACSscan; Becton Dickinson, Mountain View, CA). Cerebral endothelial cells were identified via FACS by cell size characteristics and by surface expression of CD3424 (phycoerythrin-labeled anti-CD34 antibody; Pharmingen, Mississauga, Ontario, Canada). In addition, cerebral endothelial cell activation was assessed by expression of vascular cell adhesion molecule 1 (VCAM-1). VCAM-1 is expressed by endothelium activated by a variety of stimuli.25 VCAM-1 expressing cerebral endothelial cells were identified by FACS as those cells coexpressing CD34 and VCAM-1 (FITC-labeled anti-VCAM-1 antibody; Pharmingen).

Peripheral Blood Monocyte Isolation and Flow Cytometry Analysis.

Whole blood was removed from sham-resected and BDR mice, and peripheral blood mononuclear cells were isolated using Lympholyte M (Cedarlane, Hornby, Ontario, Canada). Monocytes were then characterized via FACS after staining with anti-F4/80 antibody (RPE-labeled antibody; Serotec, Raleigh, NC). TNF-α–producing monocytes were identified via intracellular staining and analyzed via FACS as those cells coexpressing F4/80 and TNF-α (FITC-labeled antibody; BD Pharmingen).

Intravital Microscopy.

Intravital microscopy of the brain microcirculation was performed as previously described in detail.26, 27 Briefly, pial vessels in the brain were exposed by removing a piece of the parietal bone and underlying dura mater. Mice were then injected with rhodamine 6G (0.3 mg/kg intravenously; Sigma-Aldrich, Mississauga, Ontario, Canada) to label circulating leukocytes.

Leukocyte/endothelial interactions were observed using a microscope (Axioskop; Carl Zeiss Canada, Don Mills, Canada [10 eyepiece and ×25 objective lens]) outfitted with a fluorescent light source (epi-illumination at 510-560 nm using a 590-nm emission filter). All experiments were recorded onto videotape for playback analysis. Rolling leukocytes were defined as those cells moving at a velocity less than that of erythrocytes within a given vessel. Leukocytes were considered adherent to the venular endothelium if they remained stationary for 30 seconds or longer. To determine potential adhesion molecules involved in cerebral endothelial cell–leukocyte interactions, BDR mice were treated with either anti-P selectin (RB40-34; 20 μg/mouse) or anti-μ4 integrin (R1-2; 70 μg/mouse) antibodies (both from Pharmingen).26

To determine the potential roles of TNF-α and TLR4 in the increased leukocyte rolling and adhesion observed in BDR mice, the described experiments were repeated in BDR mice using TNF-α- and TLR4-deficient mice (Jackson Laboratories, Bar Harbor, ME) and their respective wild-type controls.

Isolation of Inflammatory Cells Infiltrating the Brain and Flow Cytometry.

Monocytes and lymphocytes were isolated from brains of day 10 BDR and sham-resected mice using previously described protocols.26, 28 Briefly, mice were anesthetized with halothane, and the brains were perfused with ice-cold phosphate-buffered saline and then removed and dispersed by passage through a size 40 wire mesh (Sigma) followed by passage through a size 50 wire mesh (Sigma) to obtain a cell suspension. The suspension was then centrifuged at 200g for 10 minutes at 4°C. The supernatant was discarded and the pellet was resuspended in 4 mL of 90% isotonic Percoll (Sigma), which was overlayed by 37% and then 30% isotonic Percoll. The Percoll gradient was centrifuged at 500g for 20 minutes at 4°C. Mononuclear cells were collected from the 90%/37% interface and washed three times with wash buffer (1% fetal bovine serum, 0.1% sodium azide in phosphate-buffered saline) and centrifugation at 3,000 rpm for 3 minutes at 4°C in a microcentrifuge. Cells obtained were stained with appropriate antibodies for 25-30 minutes at 4°C as follows: monocytes with Mac1 antibody (FITC-labeled anti-Mac1 [CD11b]; Pharmingen) plus anti-CD45 antibody (PerCP labeled anti-CD45; Pharmingen). Lymphocytes with anti-CD3 antibody (FITC-labeled anti-CD3; Pharmingen). Monocytes within the brain can be differentiated from microglia using flow cytometry by monocyte coexpression of Mac1 and CD45high in contrast to microglia that coexpress Mac1 and CD45low.28, 29

Mononuclear cells isolated from brains were also triple-stained with anti-CD45 antibody, anti-Mac1 antibody, and anti–TNF-α antibody (as described) to determine TNF-α–producing CD45high Mac1-expressing cells via FACS (i.e., TNF-α–producing infiltrating monocytes). Moreover, TNF-α–producing, Mac1-expressing cells (i.e., total Mac1-expressing immune cells in the CNS; mainly microglia29) were also identified via FACS.

To determine whether monocyte recruitment into the brains of BDR mice required both P-selectin and α4 integrin, BDR mice were treated intraperitoneally with immunoglobulin (IgG) control (Pharmingen) or with anti-α4 integrin antibody alone (PS/2; 200 μg/mouse on days 6 and 8 after surgery; Chemicon, Temecula, CA), or anti-α4 integrin antibody plus anti-P selectin (RB40-34 20 μg/mouse on days 6 and 8 after surgery; Pharmingen) antibody and brain infiltrating monocytes (CD45highMac1+ cells) isolated (as described) on day 10 after surgery.

Immunohistochemical Analysis of Brain VCAM-1 and F4/80 Expression.

Brain VCAM-1 and F4/80 (monocyte/macrophage marker) were determined using an indirect immunoperoxidase technique on cryostat frozen brain sections using methods as previously described30 using primary antibody (rat anti-mouse VCAM-1 monoclonal antibody; BD Pharmingen; or rat anti-mouse F4/80 monoclonal antibody; Serotec, NC). P-selectin could not be identified within brains of either sham-resected or BDR mice via immunohistochemistry, which is consistent with our previous reports.30

Statistical Analysis.

Data are expressed as the mean ± SEM. For comparisons between two means, the Student t test was used; for comparisons between more than two groups, ANOVA followed by the Student-Neuman-Keuls post hoc test was used. A P value of less than .05 was considered significant.

Results

Characterization of Animal Model of Cholestasis.

BDR mice showed clinical evidence of cholestasis with dark urine and icteric plasma. Cholestasis was confirmed biochemically through the demonstration of significant elevations of plasma total bilirubin (0.6 ± 0.1 mg/dL [sham] vs. 25.2 ± 1.9 mg/dL [BDR], n = 4/group; P ≤ .0001) and alanine aminotransferase (34.9 ± 0.5 IU/L [sham] vs. BDR: 96.4 ± 8.6 IU/L [BDR], n = 4/group; P ≤ .004) levels.

Cerebral Endothelial Cells Are Activated in Cholestasis.

VCAM-1 expression was used as a marker of cerebral endothelial cell activation, given that endothelial cells activated by several mediators demonstrate increased surface VCAM-1 expression.31 More cerebral endothelial cells isolated from BDR mice expressed surface VCAM-1 (percentage of VCAM-1–expressing cerebral endothelial cells: 38.3% ± 2.7 [BDR] vs. 25.2% ± 4.3 [sham], n = 5/group; P ≤ .03). In addition, the magnitude of expression of VCAM-1 on cerebral endothelial cells as determined via mean fluorescence intensity was significantly higher in BDR versus sham-resected mice (mean fluorescence intensity: 270.1 arbitrary units ± 14.9 [BDR] vs. 178.6 arbitrary units ± 6.1 [sham], n = 5/group; P ≤ .004). Therefore, cerebral endothelial cells are activated in BDR compared with sham-resected mice (as confirmed by intravital microscopy studies outlined in next section).

Similarily, VCAM-1 expression was documented via immunohistochemistry in the brains of BDR but not sham-resected mice and was restricted to endothelium lining blood vessels (Fig. 1).

Figure 1.

Immunohistochemical staining of VCAM-1 in brain sections obtained from (A) sham-resected and (B) BDR mice 10 days after surgery. The arrow indicates positive VCAM-1 staining. BV, blood vessel.

Circulating Monocytes Demonstrate Augmented Production of TNF-α in Cholestatic Mice.

Activated immune cells have an increased propensity to interact with activated endothelium, leading to enhanced adherent interactions between these two cell types.32 TNF-α production is a widely used marker of monocyte activation.33 Therefore, we determined whether cholestasis is associated with increased TNF-α production in circulating monocytes via FACS analysis. Cholestatic liver injury was associated with a 1.8-fold increase in the number of circulating peripheral blood mononuclear cells (1.70 ± 0.29 cells/mm3 [BDR] vs. 0.93 ± 0.09 cells/mm3 [sham], n = 5 and 4 mice/group, respectively; P ≤ .03). In addition, cholestasis was associated with a striking increase in the percentage of circulating monocytes producing TNF-α (25.8% ± 4.6 [BDR] vs. 9.3% ± 1.6 [sham], n = 5 and 4 mice/group, respectively; P ≤ .018). These results are consistent with cholestasis being associated with activation of circulating monocytes and their enhanced production of TNF-α.

Intravital Microscopy.

To determine whether cholestasis is associated with cerebral endothelial cell activation and associated leukocyte recruitment, intravital microscopy was employed. BDR mice demonstrated a striking increase in the number of rolling leukocytes observed within the cerebral vasculature compared with sham-resected controls (P ≤ .04)) (Fig. 2A-C). Increased leukocyte rolling was paralleled by a modest but significant increase in leukocyte adhesion to cerebral endothelium in BDR mice compared with sham-resected controls (Fig. 2D). The increase in leukocyte rolling along cerebral endothelial cells in BDR mice was completely abolished by treatment of BDR mice with an anti–P-selectin antibody (P ≤ .01) (Fig. 2E). Treatment of BDR mice with an α4 integrin antibody did not alter leukocyte rolling (data not shown).

Figure 2.

Representative photographs of cerebral blood vessels in (A) sham-resected and (B) BDR mice 10 days after surgery. Arrows indicate rolling leukocytes. Number of (C) rolling leukocytes (cells/min) and (D) adherent leukocytes (cells/100 μm) in sham-resected and BDR cerebral blood vessels. Bars represent the mean ± SEM of data from 5 sham-resected control mice and 8 BDR mice. *P < .012 and **P < .009 vs. sham-resected control group. (E) Leukocyte rolling in cerebral blood vessels of BDR mice is completely abrogated within 5 minutes after treatment with an anti–P-selectin antibody. Data are expressed as the mean ± SEM of results from 3 mice/group. *P < .01 versus time 0.

Intravital microscopy was performed on the brains of TLR4- and TNF-α–deficient BDR mice (Fig. 3A-B). Interestingly, leukocyte rolling was significantly greater in BDR TLR4-deficient mice compared with wild-type and TNF-α–deficient BDR mice (Fig. 3A). However, leukocyte adhesion was similar in BDR wild-type, TLR4-deficient, and TNF-α–deficient mice (Fig. 3B).

Figure 3.

Genetic deficiency of either TLR4 or TNF-α does not eliminate leukocyte–endothelial interactions in the brain vasculature of BDR mice. BDR was performed in TLR4-deficient or TNF-α knockout (TNF-α−/−) mice and intravital microscopy performed 10 days later to quantify leukocyte (A) rolling and (B) adhesion. Results are expressed as the mean ± SEM. Bars represent data from 4 TNF-α knockout and 3 TLR4-deficient mice/group. Dotted lines represent the mean of data presented in Fig. 2C and 2D for wild-type BDR mice. *P < .05 versus wild-type BDR mice and TNF-α knockout BDR mice. TLR4-def., TLR4-deficient; TNF-α−/−, TNF-α knockout.

Immune Cell Recruitment into the Brains of Cholestatic Mice.

Mononuclear cells were isolated from the brains of BDR and sham-resected mice to determine whether the increase in leukocyte rolling demonstrated in BDR mice was associated with a significant increase in the infiltration of brain tissue with mononuclear cells. BDR mice exhibited a significant increase in brain infiltration of monocytes (as represented by their coexpression of CD45high and Mac1+28,29) compared with sham-resected controls (Fig. 4A-B). In contrast, no significant increase in brain infiltration of T cells (i.e., CD3+ cells) was documented in BDR mice compared with sham-resected mice (2,796 ± 748 cells/brain [sham] vs. 4,192 ± 653 cells/brain [BDR], n = 4 and 6 mice/group, respectively; P value not significant).

Figure 4.

Isolation of mononuclear cells from brains of sham-resected and BDR mice 10 days after surgery. (A) Percentage of total mononuclear cells isolated from sham-resected and BDR mouse brains expressing CD45highMac1+ (i.e., infiltrating monocytes). Bars represent the mean ± SEM of data from 3 and 4 mice/group for sham-resected and BDR mice, respectively. *P < .005 versus sham-resected mice. (B) Total number of CD45highMac1+ cells isolated from brains of sham-resected and BDR mice. Bars represent the mean ± SEM of data from 3 and 4 mice/group for sham-resected and BDR mice, respectively. *P < .005 versus sham-resected mice. (C) Percentage of CD45highMac1+ cells isolated from sham-resected and BDR mice brains producing TNF-α. Bars represent the mean ± SEM of 3 and 4 mice/group for sham-resected and BDR mice, respectively. *P < .014 versus sham-resected mice. (D) Total number of CD45highMac1+ cells isolated from sham-resected and BDR mice brains producing TNF-α. Bars represent the mean ± SEM of data from 3 and 4 mice/group for sham-resected and BDR mice, respectively. *P < .005 versus sham-resected mice. BDR, bile duct–resected; TNF-α, tumor necrosis factor α.

More importantly, monocytes isolated from the brains of sham-resected mice did not produce TNF-α, whereas a significant proportion of monocytes isolated from the brains of BDR mice produced TNF-α (Fig. 4C-D). In addition, Mac1+ immune cells isolated from the brains of cholestatic but not sham-resected mice (these would be expected to be mainly microglia29) demonstrated significant TNF-α production, suggesting that cholestasis is also associated with the activation of immune cells that reside in the brain (i.e., microglia) to also produce TNF-α (brain Mac1+ cells producing TNF-α as (1) percentage of cells [1.04% + 0.12 (sham) vs. 11.9% + 4.6 (BDR); P < .0006 vs. sham, n = 7 mice/group] and (2) brain Mac1+ cells producing TNF-α [total number of cells/brain × 103; 1.37 + 0.36 (sham) vs. 18.2 + 7.74 (BDR); P < .0006 vs. sham, n = 7 mice/group]).

F4/80 staining of brains from BDR and sham-resected mice demonstrated that monocytes within the brains of cholestatic mice appear to cluster mainly around blood vessels to form perivascular cuffs, but also appear to penetrate into the brain parenchyma (Fig. 5).

Figure 5.

Immunohistochemical staining for F4/80 positive cells (monocyte and macrophage type cells) in brain sections obtained from (A) sham-resected and (B) BDR mice 10 days after surgery. Positive F4/80 staining cells are demonstrated (dark brown color) by arrows. BV, blood vessel.

Adhesion molecule blocking experiments revealed that the recruitment of monocytes into the brains of cholestatic mice could not be prevented by an α4 integrin–blocking antibody alone (Fig. 6) but required the simultaneous blocking of both α4 integrin and P-selectin (Fig. 6).

Figure 6.

Recruitment of CD45highMac1+ (monocytes) cells into the brains of BDR mice is significantly inhibited only by the combined blocking of P-selectin and α4 integrin, not by blocking α4 integrin alone. Mice (n = 3-4/group) were treated with α4 integrin blocking monoclonal antibody alone (α4 on days 6 and 8 after surgery) or α4 integrin plus P-selectin blocking monoclonal antibodies combined (α4 + P-selectin on days 6 and 8 after surgery) and monocyte recruitment into the brain determined at day 10 after bile duct resection. Data are expressed as a percentage of control (control BDR mice received isotype control antibody alone on days 6 and 8 after surgery). Bars represent the mean ± SEM of n = 3-4 mice/group. *P < .01 versus control isotype antibody–treated BDR mice and BDR mice treated with the α4 integrin antibody alone). P-sel, P-selectin; BDR, bile duct resection.

Discussion

Traditionally, communication between the periphery and the CNS during systemic immune activation has been postulated to occur through one of two pathways: neural or humoral.10 Neural signals travel mainly in vagal afferent nerve projections from peripheral sites (e.g., liver) and synapse within the brainstem, from which they are subsequently relayed to higher centers within the CNS.10–14 Humoral signaling involves circulating soluble factors (e.g., cytokines) that signal the brain either by activating cerebral endothelial cells with the subsequent production of secondary messengers (e.g., PGE2) that are released into the brain parenchyma or by entering the brain through areas devoid of an intact blood–brain barrier to directly stimulate brain parenchymal structures.15, 16 Specifically, TNF-α has been implicated in both of these phenomena.15, 16, 34 Cholestatic liver damage is associated with increased systemic immune cell activation mediated at least in part by increased circulating levels of endotoxin and proinflammatory cytokines.18–22 It is well known that endotoxins and cytokines within the peripheral circulation can activate cerebral endothelial cells in addition to producing sickness behaviors such as lethargy, malaise, and fatigue.1, 3, 35, 36 In the current study, we documented peripheral immune activation in cholestasis as reflected by a striking increase in TNF-α production by circulating monocytes in BDR compared with sham-resected mice. Interestingly, the increase in monocyte TNF-α expression in BDR mice was not paralleled by an increase in plasma TNF-α levels as determined via a sensitive mouse-specific TNF-α ELISA (data not shown). Therefore, cholestatic liver injury is associated with an increase in circulating, activated, TNF-α–producing monocytes.

This observation raised the possibility that cholestasis might also be associated with increased infiltration of immune cells from the circulation into the brain. These brain-infiltrating immune cells might then produce cytokines (e.g., TNF-α) within the CNS. Importantly, this pathway would allow humoral messengers to access areas of the brain with an intact blood–brain barrier rather than being restricted to the few areas of the brain that do not have a blood–brain barrier. However, for this to be possible, cerebral endothelial cells would have to be activated in the setting of cholestatic liver injury, because under normal conditions very few leukocytes gain access to the CNS.37, 38

Activation of endothelium occurs after exposure to numerous stimuli. More importantly, activated endothelium expresses adhesion molecules that are critical for cellular recruitment into tissues, including P-selectin and VCAM-1.26, 30, 31 Therefore, we initially isolated cerebral endothelial cells from cholestatic and control mice and determined VCAM-1 expression; a commonly used marker of endothelial cell activation.26, 30, 31 We observed an increase in VCAM-1 expression in endothelial cells isolated from BDR compared with sham-resected mice, a finding consistent with cholestasis-associated activation of cerebral endothelial cells. This observation was supported by our documentation of increased VCAM-1 expression on blood vessel endothelium in brain sections obtained from cholestatic but not control mice (P-selectin expression could not be detected via immunohistochemistry, which is consistent with our previous observations30). This was further supported by our intravital microscopy studies. We identified a marked increase in leukocyte rolling along cerebral endothelium in cholestatic compared with noncholestatic mice. Sham-resected mice demonstrated very little leukocyte rolling in cerebral blood vessels, whereas leukocyte rolling in BDR mice was strikingly increased. Moreover, this increase in leukocyte rolling in BDR mice was completely abrogated by an anti-P selectin antibody, but not by an anti-α4 integrin antibody. These findings are consistent with activation of cerebral endothelium as being a consequence of cholestasis and suggest that leukocyte rolling in the cerebral vasculature in cholestasis is mediated by P-selectin and not VCAM-1 (which binds to α4 integrin). Despite the fact that VCAM-1 expression was increased in BDR mice, an anti-α4 antibody did not block rolling. These findings are consistent with our previous observations in the CNS vasculature that demonstrated that the role of VCAM-1 is largely restricted to adhesion.26 This observation is further supported by the fact that we and others26, 39 have identified constitutive expression of VCAM-1 at low levels on cerebral endothelial cells of control mice, but no leukocyte rolling was detected in these mice.26 Moreover, in a different model of chronic inflammation of the CNS (i.e., experimental allergic encephalomyelitis), leukocyte–endothelial interactions were dependent on the induction of P-selectin in cerebral endothelial cells.26 Similarily, we clearly show through a functional assay that leukocyte rolling along cerebral endothelial cells in BDR mice is also dependent on P-selectin expression, because this rolling is completely blocked by a P-selectin antibody.

To investigate whether this increase in leukocyte rolling in the brain vasculature in cholestatic mice might be related to an effect of endotoxin or TNF-α on cerebral endothelium, we performed brain intravital microscopy in mice with gene deletions, making them unresponsive to endotoxin (TLR4 knockout) or incapable of making TNF-α (TNF-α knockout). Interestingly, we found an increase in leukocyte rolling in cerebral vessels in BDR TLR4-deficient mice compared with both BDR wild-type and BDR TNF-α knockout mice. However, leukocyte adhesion was similar in all three of these groups. The finding of increased leukocyte rolling in BDR TLR4-deficient mice is of interest and suggests that in the setting of presumed persistent low-grade endotoxemia (as has been reported in bile duct resected rodents21), TLR4 may act to decrease leukocyte rolling. Therefore, other inflammatory mediators in cholestasis must be responsible for augmented P-selectin expression on cerebral endothelial cells in cholestasis; this warrants further investigation.

Our brain intravital microscopy data suggested that cholestasis is associated with an increase in leukocyte adhesion (I.e., the final step in the leukocyte recruitment cascade before leukocyte emigration into a tissue occurs31). Therefore, to determine if cholestasis results in immune cell recruitment into the brains of BDR mice, we isolated mononuclear cells from the brains of BDR and sham-resected mice. Microglia are the main resident macrophages of the CNS and can be difficult to differentiate from monocytes infiltrating the CNS by using cell surface marker expression.28, 29 However, several recent publications have documented monocyte infiltration into the brain in experimental cerebral inflammatory disorders by using the differential expression of CD45.28, 29, 40, 41 Specifically, monocytes are CD45high-expressing cells and microglia are CD45low-expressing cells, thus allowing for the differentiation of these two cell types via FACS analysis. Using this methodology, we have identified an increase in monocyte infiltration in the brains of cholestatic mice compared with noncholestatic controls, demonstrating that leukocyte recruitment into the CNS does indeed occur in BDR mice. This finding appears to be specific for monocytes given our observation that cholestatic mice did not recruit more lymphocytes into their brains than noncholestatic controls. In addition, it is unlikely that this increase in brain monocyte recruitment in BDR mice is due to a nonspecific increase in blood–brain barrier permeability in cholestatic mice, because we have previously demonstrated that cholestasis is associated with decreased permeability of the blood–brain barrier.42 In addition, blood–brain barrier permeability is not a limiting factor in leukocyte entry into the CNS, whereas adhesion molecule expression is.27 Moreover, we could inhibit monocyte infiltration into the brains of BDR mice by blocking both P-selectin and α4 integrin, whereas blocking α4 integrin alone had no effect. These observations are consistent with our previous work that demonstrated that blocking one adhesion molecule pathway in a complex disease state is not enough to prevent the recruitment of cells into a tissue.43

Monocytes that enter the CNS may or may not be activated to release cytokines. Our current observations clearly show that a high proportion of the monocytes recruited into the CNS of BDR mice express TNF-α, whereas this was not the case with sham-resected mice. Immunohistochemistry clearly demonstrated that monocytes that have been recruited into the brains of cholestatic mice mainly cluster around blood vessels to form perivascular cuffs, but also appear to penetrate into the brain parenchyma. In addition, it is apparent that cholestasis is associated with a broad activation of other immune cells (presumably mainly microglia) within the CNS to also produce TNF-α. Although the infiltration of activated TNF-α–secreting monocytes into the brains of cholestatic mice may contribute to the subsequent activation of resident cells within the brain, we believe that this activation of resident cells in the brain likely also involves other signals that might include those from soluble factors and possibly from neural signals and warrants further investigation. Given the significant behavioral effects of TNF-α within the CNS (I.e., sickness behaviors44–46), the production of TNF-α within the brains of cholestatic mice is likely to be important in the alterations in behavior, as well as in the changes in the neurotransmitter systems that subserve these behaviors,47–49 within the brains of cholestatic mice and may have direct implications for these systems in cholestatic patients.

In conclusion, we have documented that cholestatic mice display activation of monocytes within the peripheral circulation as well as activation of cerebral endothelial cells. Moreover, activated P-selectin expressing cerebral endothelium in cholestasis recruits leukocytes and ultimately results in an influx of TNF-α–producing monocytes into the CNS. Our current observations may have direct relevance to liver–brain communication and the possible generation of sickness behaviors in cholestasis.

Ancillary