CX3CR1 and vascular adhesion protein-1-dependent recruitment of CD16+ monocytes across human liver sinusoidal endothelium


  • Alexander I. Aspinall,

    1. The Liver Unit, University of Calgary Health Sciences Center, Calgary, Alberta, Canada
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Stuart M. Curbishley,

    1. Division of Immunity and Infection, Centre for Liver Research and National Institute for Health Research Biomedical Research Unit, College of Medicine and Dentistry, University of Birmingham, Edgbaston, Birmingham, UK
    Search for more papers by this author
    • These authors contributed equally to this work.

  • Patricia F. Lalor,

    1. Division of Immunity and Infection, Centre for Liver Research and National Institute for Health Research Biomedical Research Unit, College of Medicine and Dentistry, University of Birmingham, Edgbaston, Birmingham, UK
    Search for more papers by this author
  • Chris J. Weston,

    1. Division of Immunity and Infection, Centre for Liver Research and National Institute for Health Research Biomedical Research Unit, College of Medicine and Dentistry, University of Birmingham, Edgbaston, Birmingham, UK
    Search for more papers by this author
  • Miroslava Blahova,

    1. Division of Immunity and Infection, Centre for Liver Research and National Institute for Health Research Biomedical Research Unit, College of Medicine and Dentistry, University of Birmingham, Edgbaston, Birmingham, UK
    Search for more papers by this author
  • Evaggelia Liaskou,

    1. Division of Immunity and Infection, Centre for Liver Research and National Institute for Health Research Biomedical Research Unit, College of Medicine and Dentistry, University of Birmingham, Edgbaston, Birmingham, UK
    Search for more papers by this author
  • Rebecca M. Adams,

    1. Division of Immunity and Infection, Centre for Liver Research and National Institute for Health Research Biomedical Research Unit, College of Medicine and Dentistry, University of Birmingham, Edgbaston, Birmingham, UK
    Search for more papers by this author
  • Andrew P. Holt,

    1. Division of Immunity and Infection, Centre for Liver Research and National Institute for Health Research Biomedical Research Unit, College of Medicine and Dentistry, University of Birmingham, Edgbaston, Birmingham, UK
    2. The Liver Unit, Queen Elizabeth Hospital, Edgbaston, Birmingham, UK
    Search for more papers by this author
  • David H. Adams

    Corresponding author
    1. Division of Immunity and Infection, Centre for Liver Research and National Institute for Health Research Biomedical Research Unit, College of Medicine and Dentistry, University of Birmingham, Edgbaston, Birmingham, UK
    • Division of Immunity and Infection, Centre for Liver Research and National Institute for Health Research Biomedical Research Unit, College of Medicine and Dentistry, University of Birmingham, Edgbaston, Birmingham, United Kingdom B15 2TT
    Search for more papers by this author
    • fax: +44121 4158701

  • Potential conflict of interest: Nothing to report.


The liver contains macrophages and myeloid dendritic cells (mDCs) that are critical for the regulation of hepatic inflammation. Most hepatic macrophages and mDCs are derived from monocytes recruited from the blood through poorly understood interactions with hepatic sinusoidal endothelial cells (HSECs). Human CD16+ monocytes are thought to contain the precursor populations for tissue macrophages and mDCs. We report that CD16+ cells localize to areas of active inflammation and fibrosis in chronic inflammatory liver disease and that a unique combination of cell surface receptors promotes the transendothelial migration of CD16+ monocytes through human HSECs under physiological flow. CX3CR1 activation was the dominant pertussis-sensitive mechanism controlling transendothelial migration under flow, and expression of the CX3CR1 ligand CX3CL1 is increased on hepatic sinusoids in chronic inflammatory liver disease. Exposure of CD16+ monocytes to immobilized purified CX3CL1 triggered β1-integrin-mediated adhesion to vascular cell adhesion molecule-1 and induced the development of a migratory phenotype. Following transmigration or exposure to soluble CX3CL1, CD16+ monocytes rapidly but transiently lost expression of CX3CR1. Adhesion and transmigration across HSECs under flow was also dependent on vascular adhesion protein-1 (VAP-1) on the HSECs. Conclusion: Our data suggest that CD16+ monocytes are recruited by a combination of adhesive signals involving VAP-1 and CX3CR1 mediated integrin-activation. Thus a novel combination of surface molecules, including VAP-1 and CX3CL1 promotes the recruitment of CD16+ monocytes to the liver, allowing them to localize at sites of chronic inflammation and fibrosis. (Hepatology 2010)

The liver contains bone marrow-derived myeloid dendritic cells (mDCs) and macrophages (Kupffer cells) that are recruited from blood via the hepatic sinusoids. They act as immune sentinels to detect and coordinate responses to invading pathogens and antigens entering the liver through the portal vein.1-3 Under basal conditions, these cells are replenished by recruitment of precursors from blood, which increases with inflammation. The exact nature of the precursor cells is unclear, but they likely reside within the circulating CD16+ monocyte population.4-7

mDCs arise from bone marrow-derived progenitors within the monocyte pool.8-10 Several populations of precursors have been proposed, including lineage-negative CD11c+ monocytes, CD34+ progenitors,11 and human CD16+ monocytes.12 Human monocytes display heterogeneity defined by expression of chemokine receptors, adhesion molecules, CD14, and CD16.13-15 The CD14+CD16++ subset expresses high levels of the chemokine receptor CX3CR1 and is believed to give rise to DCs with potent antigen-presenting capabilities16 and inflammatory tissue macrophages.15, 17 Furthermore, transendothelial migration of CD16+ monocytes in vitro induces differentiation into functional DCs, suggesting that recruitment itself may shape their subsequent differentiation.18

Integral to mDC function is the capacity to traffic from one anatomical compartment to another. In the liver, this involves a pathway that traverses the space of Disse and takes the cells along the hepatic sinusoids to the portal tract lymphatics.19-21 The recruitment of precursor mDCs from the blood into tissues across endothelium is poorly understood.22 In the mouse, precursor mDCs enter inflamed skin using intercellular adhesion molecule (ICAM)-2, P-selectin, and E-selectin and the chemokine receptors CCR1, CCR2, and CCR5,23 but little is known about hepatic recruitment through the sinusoidal vascular bed. Because recruitment of neutrophils and lymphocytes to the liver involves distinct adhesion pathways,24, 25 we hypothesized that unique combinations of molecules might regulate monocyte recruitment. We report that recruitment of human CD16+ monocytes to the inflamed liver involves unique combinations of adhesion molecules in which interactions mediated by vascular adhesion protein-1 (VAP-1) and the chemokine CX3CL1 are critically important.


GPC, G protein-coupled; HSEC, hepatic sinusoidal endothelial cell; ICAM, intercellular adhesion molecule; mAb, monoclonal antibody; mDC, myeloid dendritic cell; PBS, phosphate-buffered saline; PTX, pertussis toxin; TNF-α, tumor necrosis factor-α; VAP-1, vascular adhesion protein-1; VCAM, vascular cell adhesion molecule.

Materials and Methods

Tissue and Blood.

Liver tissue was obtained from livers removed at transplantation at the Queen Elizabeth Hospital from patients with alcoholic liver disease (n = 6), primary biliary cirrhosis (n = 6), primary sclerosing cholangitis (n = 6), and autoimmune hepatitis (n = 6). Peripheral blood was obtained from healthy volunteers and liver transplant recipients. Samples were collected after informed consent following local Ethics Committee approval.

Antibodies and Reagents.

Soluble CX3CL1 and all anti-chemokine receptor monoclonal antibodies (mAbs) except anti-CX3CR1 were obtained from R&D Systems Europe and used at the recommended concentrations (Table 1).

Table 1. Clones and Suppliers of Antibodies Used
Antibody TargetCloneSource
CD16 (FACS)3G8Serotec
CD16 (IHC)DJ130cDako
CD86FUN-1BD Biosciences
CX3CL151637R&D Systems
Ep-CAMHEA-125Progen Biotec
HLA-DRL243BD Biosciences
ICAM-111C81R&D Systems
VAP-1TK8-14Kind gift of David Smith, Biotie Therapeutics
VCAM-14B2R&D Systems


Six-micrometer cryostat sections were air-dried on poly-L-lysine treated slides, acetone-fixed (10 minutes), and stained. Sections were preincubated with 2.5% horse serum (Vector Laboratories, Peterborough, UK) in TBS prior to mouse anti-human mAb against CD16 or CX3CL1 in Tris-buffered saline/0.1% normal horse serum. Control sections were incubated with isotype-matched control mAb. Antibody binding was assessed using ImmPress peroxidise visualisation with Novared chromogen (Vector Laboratories). Sections were counterstained with hematoxylin.

RNA Extraction and Quantitative Polymerase Chain Reaction.

Total RNA was extracted from 30 mg human liver using RNEasy (Qiagen, UK) after DNAse treatment with RNAse-free DNAse (Qiagen). Fifty micrograms extracted RNA was transcribed into complementary DNA using iScript cDNA (BioRad, Hercules, CA), and eluted RNA and complementary DNA were measured (NanoDrop, Thermo-Fisher Scientific). Expression of human CX3CL1 messenger RNA was quantified using Taqman Fluorogenic 5′-nuclease assays and gene-specific 5′-FAM-labeled probes on an ABI Prism 7900 detector. Threshold cycle (Ct) values of the target gene were normalized to GAPDH and differential expression levels calculated using 2−ΔΔCt.

Flow Cytometry.

Blood mononuclear cells isolated using Lympholyte (Cedarlane Laboratories, Burlington, Canada) were resuspended in labelling medium (phosphate-buffered saline [PBS]/0.5% fetal bovine serum/0.1 mM ethylene diamine tetraacetic acid). Fc receptors were blocked with 10 μg/mL mouse immunoglobulin for 10 minutes. Directly conjugated mAbs were added for 20 minutes and cells were resuspended in labeling medium and analyzed using a CyAn flow cytometer (Beckman Coulter, Bucks, UK). Cells were maintained on ice throughout except for the demonstration of CX3CR1 internalization, which was performed at 37°C.

Isolation of CD16+ Monocytes.

OptiPrep density gradient centrifugation26 produced 75%-85% pure monocytes and CD16+ monocytes were then isolated from the enriched population by high-speed flow cytometric sorting. Fc receptors on the enriched monocytes were blocked with normal mouse Ig. Directly conjugated mAbs against CD16 and CD56 were added on ice for 20 minutes before washing and resuspending in labeling medium and sorted using a MoFlo cell sorter (Beckman Coulter). Three gates were set during the sort: one to exclude doublet events, one to include monocytes and exclude other cells/debris, and one to include only CD16+CD56 cells. This produced a population of >98% pure CD16+ monocytes.

Isolation and Culture of Hepatic Sinusoidal Endothelial Cells.

Hepatic sinusoidal endothelial cells (HSECs) were isolated using published methods with the substitution of NycoDenz (Axis Shield) for the discontinued metrizamide for density-gradient centrifugation).27 Fresh liver was minced and incubated with collagenase IV (Sigma, Poole, UK) for 30 minutes at 37°C before filtering through 40 μm nylon mesh, diluted in PBS and layered on 25% Nycodenz/PBS and centrifuged at 700g for 20 minutes. Cells at the interface were collected, washed and cholangiocytes removed by negative immunomagnetic selection with anti-HEA-125. Endothelial cells were positively selected using anti-CD31 antibody and cultured in human endothelial basal media plus penicillin/streptomycin/L-glutamine/10% human serum, hepatocyte growth factor, and vascular endothelial growth factor (10 ng/mL, Peprotech, UK) and used within four passages. This protocol was developed to isolate sufficient cells from either normal or diseased human liver for use in functional assays. In rats, it has been suggested that CD31 should not be used to isolate sinusoidal cells because cell-surface CD31 is absent from quiescent sinusoidal endothelium and its use generates cells with low frequencies of fenestrae.28 However, we find that human sinusoidal endothelial cells express cell surface CD31, albeit at lower levels than vascular endothelium, a finding consistent with other published reports.29 To confirm that CD31-selected cells from human liver have a sinusoidal phenotype, we demonstrated expression of several receptors that are present on sinusoidal but not vascular endothelium, including the hyaluronan receptor LYVE-130 and the C-type lectins L-SIGN,31 L-SECtin, mannose receptor, and CLEVER-1.25, 32, 33 These cells thus have a unique sinusoidal phenotype. They also express ICAM-1 and VAP-1 and increase expression of vascular cell adhesion molecule (VCAM)-1 in response to cytokines, a phenotype that corresponds to activated sinusoidal endothelium in vivo. Consistent with their distinct phenotype, we have reported functional and signaling responses in these cells that differ from those seen with vascular endothelial cells.27, 34 The relevance of in vitro studies using HSECs to the situation in vivo is illustrated by the fact that many of our in vitro findings with HSECs have subsequently been confirmed by others using in vivo models. For example, the involvement of VAP-1 in lymphocyte recruitment through hepatic sinusoids was first proposed using HSECs in vitro27 and subsequently confirmed by others using rat and mouse models in vivo.35, 36 Thus, HSECs cultured for up to four passages maintain expression of cell surface receptors that (1) define them as sinusoidal in origin and (2) allow them to interact with leukocytes during flow-based assays. We thus believe that these cells, despite losing fenestrations during passage in culture, are appropriate to model leukocyte recruitment to the liver—particularly as this process usually occurs through damaged or inflamed sinusoids—and we refer to them as human liver sinusoidal endothelial cells (HSECs).

Flow-Based Adhesion Assays.

Flow-based adhesion assays were conducted as described.37 HSECs were grown to confluence in glass capillaries for 24 hours, stimulated for 24 hours with tumor necrosis factor-α (TNF-α) (10 ng/mL, R&D Systems) and connected to the flow-based system. In some experiments, soluble proteins were coated onto the slides, including recombinant CX3CL1 (500 ng/mL, R&D Systems), recombinant VAP-1 (1 μg/mL, a gift from David Smith, Biotie Therapies), recombinant VCAM-1 (5 μg/mL, R&D Systems), or human albumin (1%, Sigma, Poole, UK ). CD16+ monocytes were perfused through the capillaries at wall shear stresses that mimic flow through hepatic sinusoids (0.05 Pa). Adherent cells were visualized under phase contrast microscopy and rolling, adherent (phase bright), and transmigrated cells (phase dark) quantified by offline digital analysis. The number of adherent cells was calculated as mm2/106 cells perfused. Adhesion was blocked using anti-VCAM-1 (BBIG-V1), anti-ICAM-1 (BBIG-I1, R&D Systems), anti-CX3CR1 (2A9-1, MBL, MA), anti-VAP-1 (TK8-14, a gift from David Smith, Biotie Therapies), and pertussis toxin (PTX) (Sigma, UK). Blocking reagents were incubated with HSECs or monocytes for 30 minutes and washed out before assay.

Transwell Assay.

Falcon transwells (24 wells/plate, 5 μm pore, Becton-Dickinson) coated on the upper membrane surface with 20% rat-tail collagen, were seeded with 5 × 104 HSECs, grown to confluence for 24 hours, and stimulated with TNF-α (10 ng/mL) for a further 24 hours. CD16+ monocytes (5 × 105) were added to the top chamber for 1.5 hours before rinsing the top well to remove nonadherent cells. After 24 hours, transmigrated cells were harvested from the bottom well, washed in PBS/BSA, and analyzed by way of flow cytometry.

Isolation and Characterization of Liver mDCs.

Hepatic mDCs were isolated from freshly explanted liver using mechanical and density gradient separation as described.7 Hepatic mDCs were identified by way of flow cytometry as cells within the monocyte cloud that expressed CD86 and high levels of HLA-DR.

Activation of β1-Integrin by Occupancy of CX3CR1.

CX3CR1 was engaged using purified recombinant CX3CL1. VLA-4 activation by 2 mM MnCl2 was used as a positive control. A 12G10 antibody that recognizes a conformation-dependent CD29 epitope was used to detect activated VLA-4.38 Incubation with primary mAb or isotype-matched control was performed for 45 minutes at room temperature during CX3CR1 engagement. Fluorescein isothiocyanate-conjugated secondary antibody (goat anti-mouse or goat anti-rat) was used to detect 12G10 binding by way of flow cytometry.

Statistical Analysis.

Data were analyzed using two-way analysis of variance with Tukey's posttest using GraphPad InStat (GraphPad Software, San Diego, CA).


Phenotype of CD16+ Monocytes.

We detected three subsets of monocytes from human blood CD16+CD14; CD16+CD14+ and CD14+CD16 (Fig. 1). CD16+ monocytes from human blood expressed low levels of two molecules associated with lymph node entry: CD62L and CCR7. The chemokine receptors CCR1, CCR2, CCR4, CCR5, CCR6, CXCR1, CXCR3, and CXCR5 were expressed on more CD14+ cells than CD16+ cells. The CD16+/CD14 subset had the most limited chemokine receptor repertoire, with CD14+ cells having a more inflammatory phenotype. CCR8, CXCR4, and CX3CR1 were expressed at similar levels on all three subsets.

Figure 1.

CD16+ and CD16 monocytes were analyzed for the expression of cell adhesion molecules and chemokine receptors. Some surface molecules—such as CX3CR1, CD49d, and CD18—were expressed at the same level on both populations (A,B,E). When compared with CD16 monocytes, molecules such as CD62L and CCR7 were expressed at a lower level or not at all on CD16+ monocytes (C,F). CD11c was noted on both populations, but at a slightly higher level on CD16+ monocytes (D), consistent with their putative role as mDC precursors. The complete characterization of chemokine receptor expression on three populations of monocytes (H) defined by expression of CD14 and CD16 as CD14+/CD16+, CD14/CD16+ and CD14+/CD16 is shown in (G). The plots in (A-F) are representative profiles. The mean ± standard error of the mean of five experiments is shown in (G).

Localization of CX3CL1 and CD16+ Cells in Human Liver.

CX3CL1 in normal human liver was largely limited to bile ducts, whereas in diseased liver it was also detected on sinusoids (Fig. 2). Increased expression in inflammatory disease was confirmed by way of real-time quantitative polymerase chain reaction (Fig. 2C). In normal liver, CD16+ cells were detected throughout the parenchyma, consistent with Kupffer cells and on mononuclear cells within portal tracts. In diseased liver, CD16+ cells were increased at areas of inflammation, including fibrotic septa and expanded portal tracts, where they were seen in close association with bile ducts. In cirrhotic liver, there was a relative loss of CD16+ cells within regenerative nodules associated with increased numbers at sites of inflammation/fibrosis (Fig. 3).

Figure 2.

Normal and diseased human liver tissue was analyzed for the expression of CX3CL1. (A) Total RNA was analyzed for the expression of CX3CL1 messenger RNA by way of quantitative polymerase chain reaction, demonstrating a 2-fold increase in primary biliary cirrhosis and a 3- to 3.5-fold increase in alcoholic liver disease (ALD) and autoimmune hepatitis (AIH) compared with noninflamed liver. A similar increase was detected in primary sclerosing cholangitis (PSC), though this did not reach significance. All values were relative to normal liver (NL) (n = 6 ± standard error of the mean). *P ≤ 0.01. **P ≤ 0.001. (B,C) Immunohistochemical analysis of normal liver (B) revealed expression of CX3CL1 was limited to biliary structures (arrows), whereas in diseased liver expression persisted on bile ducts and increased on sinusoids (C).

Figure 3.

The localization of CD16+ cells was assessed in normal and diseased human liver tissue by way of immunohistochemical analysis. In normal liver (A), CD16+ cells were located throughout the parenchyma and portal tracts (PT), but in diseased liver (B) they were associated with the inflammatory infiltrate and fibrotic septa and were largely absent from regenerative nodules (RN).

Adhesion Under Conditions That Mimic Physiological Flow Through Liver Sinusoids.

CD16+ monocytes purified from peripheral blood as described above were perfused through microslides containing confluent HSECs stimulated with TNF-α for 24 hours. The number of CD16+ monocytes binding HSECs was determined (Fig. 4A), and adhesion was subclassified into cells that became activated, changed shape, and migrated across the endothelial monolayer (phase dark, Fig. 4B). Several inhibitors had no effect on adhesion or migration on HSECs, including antibodies against P-selectin and E-selectin (data not shown), confirming the lack of involvement of selectins in this vascular bed.39 Heterotrimeric Gαi proteins are involved in chemokine receptor signaling and can be inhibited using PTX. Preincubation of CD16+ monocytes with PTX caused a decrease in total adherent cells and virtually abolished transmigration as demonstrated by the lack of phase dark, monocytes beneath the endothelium in Fig. 4B. Because CD16+ cells express high levels of CX3CR1, we looked to inhibit this receptor using antibodies and by desensitization by prior exposure to ligand CX3CL1. Both approaches reduced transmigration (normalized to number of adherent cells) to a similar level seen with PTX treatment (Fig. 4B), suggesting that CX3CR1 is the dominant G protein-coupled (GPC) receptor involved. Total adhesion was more efficiently inhibited by anti-CX3CR1 antibody than by PTX, suggesting that some adhesion is GPC-independent; this finding is consistent with previous studies showing that transmembrane CX3CL1 can support leukocyte adhesion directly (Fig. 4A). Antibodies against VCAM-1 and ICAM-1 in combination or VAP-1 decreased total adherent cells (Fig. 4A), whereas anti-ICAM-1 or anti-VCAM-1 alone had no effect (data not shown). Inhibition of HSECs with anti-VAP-1 antibodies immediately before and during the flow-based adhesion assay reduced the proportion of cells undergoing transendothelial migration (Fig. 4B).

Figure 4.

CD16+ monocytes adhere to and transmigrate across HSECs under flow. (A) Pretreatment of CD16+ monocytes with PTX or HSECs with blocking antibodies to VAP-1, ICAM-1, and VCAM-1 or CX3CL1 resulted in a significant decrease in the total number of cells adhering to the endothelial monolayer from flow. (B) We quantified the number of adherent cells that subsequently went on to transmigrate through the HSEC monolayer. Transmigrated cells that have migrated beneath the endothelial monolayer were phase dark and easily distinguished from the phase bright cells adherent on surface of the monolayer. The numbers of adherent cells transmigrating was significantly reduced in the presence of a blocking anti-VAP-1 antibody and virtually abolished by the treatment of HSECs with anti-CX3CL1 or of CD16+ monocytes by pertussis toxin. In contrast blocking antibodies to VCAM-1 and ICAM-1 prevented adhesion, but not the transmigration of CD16+ monocytes. Blocking antibodies to E-selectin and P-selectin on HSECs did not affect adhesion or transmigration of CD16+ monocytes (not shown), consistent with previous findings that selectins have no role in adhesion in low flow state of liver sinusoids (n = 5). *P ≤ 0.01. **P ≤ 0.001. HSECs were confirmed to be CD31, L-SIGN, LYVE-1, and VAP-1+ as reported.27, 30, 31

Adhesion Under Flow to Soluble CX3CL1 and VCAM-1.

To further investigate the roles of CX3CL1 and VCAM-1, adhesion and migration under flow were studied with combinations of purified proteins. Microslides were coated with soluble CX3CL1 and VCAM-1. VCAM-1 but not CX3CL1 alone (data not shown), was able to support CD16+ monocyte adhesion; of the adherent cells, ≈40% changed shape and developed a migratory phenotype. When VCAM-1 was coimmobilized with CX3CL1, the total number of adherent cells increased, and the proportion undergoing shape-change increased to 70% (Fig. 5A). No change was seen in the level of adhesion or shape-change on VCAM-1 when an irrelevant chemokine was coimmobilized with VCAM-1. This adhesion and shape-change was associated with activation of the VLA-4 integrin (Fig. 5B) as demonstrated by increased binding of mAb 12G10, which recognizes the conformation-dependent active site on VLA-4,40 following exposure of CD16+ monocytes to soluble CX3CL1. Thus, the engagement of CX3CR1 by immobilized CX3CL1 induces downstream activation of integrins.

Figure 5.

CX3CL1 treatment induces adhesion and shape change in CD16+ monocytes on VCAM-1. (A) CD16+ monocytes bound to immobilized VCAM-1 in microslides. Less than 50% of the cells undergo shape change. The combination of immobilized VCAM-1 and CX3CL1 supported a higher number of adherent cells, the majority of which showed morphological changes consistent with a migratory phenotype. (B) Furthermore, analysis of the activation of the α4β1-integrin on CD16+ monocytes demonstrated a near doubling of median channel fluorescence following exposure to CX3CL1. Flow data represent the mean ± standard error of the mean of three experiments; cytometry data are representative of three repeats. **P ≤ 0.001.

CX3CR1 Is Down-regulated After Transmigration Across HSECs.

The expression of CX3CR1 on CD16+ monocytes following transmigration was studied in transwells in which HSECs were cultured on membrane inserts and CD16+ monocytes were applied to the top chamber. Cells that migrated were removed from the bottom chamber, and levels of CX3CR1 were determined. Following transmigration through HSECs, the expression of CX3CR1 decreased on CD16+ monocytes (Fig. 6), and preincubation of CD16+ monocytes with soluble CX3CL1 reduced surface CX3CR1, which was re-expressed 1 hour after removal of soluble CX3CL1. This was not due to receptor masking, because expression remained detectable when the experiment was repeated at 0°C (Fig. 6B).

Figure 6.

A comparison of CX3CR1 expression on CD16+ monocytes prior to and after transmigration across HSECs in vitro. (A) Purified CD16+ monocytes were applied to the top chamber of a transwell coated with HSECs that had been stimulated with TNF-α. Transmigrated cells were collected and compared with pre-emigrant cells for CX3CR1 expression. The transmigrated cells exhibited a near-complete loss of CX3CR1 expression. (B) Additionally, this loss of expression could be recapitulated by incubating CD16+ monocytes with CX3CL1 prior to antibody labeling. Preincubation for 1 hour completely diminished receptor expression, which recovered following an additional 1 hour of rest following removal of exogenous CX3CL1. We confirmed this was not due to receptor masking by repeating the experiment on ice. This receptor loss following engagement of CX3CR1 may contribute to retaining precursor DCs in the liver promoting their maturation into liver-specific mDCs. Representative flow cytometric profiles from three replicate experiments.

Expression of CX3CR1 on Intrahepatic CD16+ Cells.

Matched blood and liver tissue from patients undergoing liver transplantation was used to compare expression of CX3CR1 on mDCs freshly isolated from liver tissue with CD16+ monocytes from the same patient's blood. Figure 7 demonstrates the intermediate level of CX3CR1 on CD16+ monocytes in blood. Hepatic mDCs defined by expression of CD86 and MHC class II show heterogeneous CX3CR1 expression, with some cells expressing high levels, whereas others have levels comparable to those seen in blood CD16+ monocytes.

Figure 7.

A comparison of the expression of CX3CR1 on CD16+ monocytes and mDCs from human liver. Peripheral blood CD16+ monocytes taken from a patient about to undergo liver transplantation for end-stage liver disease exhibited an intermediate level of expression of CX3CR1, similar to the level of expression noted in healthy donors (Fig. 1A). The mDCs, representing mature forms of DC precursors that transmigrated across HSECs, were isolated from the liver of the same patient a few hours later and identified by way of flow cytometry as cells within the monocyte cloud that expressed CD86 and high levels of MHC class II. The mDCs exhibited either high or low expression of CX3CR1 and when compared with the intermediate level of expression on CD16+ monocytes, suggesting that the microenvironment of the liver may regulate the expression of CX3CR1. Representative flow cytometric profiles from three replicate experiments.


The liver contains mDCs and tissue macrophages that regulate local immune responses. Under basal conditions, these cells are continually replenished by the recruitment of precursors from blood, and this increases with inflammation. We show that CD16+ cells redistribute in the chronically inflamed liver and can be detected at sites of inflammation and fibrosis (Fig. 3). This is consistent with studies suggesting that CD16+ cells are the precursors of inflammatory tissue macrophages and inflammatory DCs.13 Human CD16+ cells undergo differentiation into DCs after migration through endothelium,10, 16, 18 and CD16+ DCs are present in the human liver,41 leading us to investigate the molecules involved in CD16+ monocyte recruitment from blood into the liver through hepatic sinusoids. We used flow-based adhesion assays incorporating primary human hepatic sinusoidal endothelium to model the liver sinusoid.27 We showed that human CD16+ monocytes express a cell adhesion molecules and chemokine receptors that could promote recruitment from blood, including CCR4, CXCR4, CX3CR1, CD18, and CD49d. We then demonstrated that they use a combination of CX3CL1, VCAM-1, and VAP-1 to arrest on HSECs from flow and a combination of CX3CR1 and VAP-1 to undergo transendothelial migration. Thus, we show for the first time that VAP-1 is involved in DC precursor recruitment to peripheral tissue and define the role played by CX3CR1 in transmigration through human endothelium under flow.

CX3CR1, which is expressed at uniformly high levels on CD16+ monocytes (Fig. 1), was the dominant chemokine receptor involved (Figs. 4 and 5) and antibodies against CX3CR1 inhibited both adhesion and transmigration across HSECs. CX3CL1 is a transmembrane chemokine that can support adhesion independently of signaling through its GPC receptor, consistent with our finding that adhesion of CD16+ cells on HSECs is more efficiently inhibited by anti-CX3CL1 antibodies than PTX. However, transmigration was clearly dependent on signaling by way of CX3CR1, because it was almost completely prevented by PTX. To investigate the role of CX3CL1 further, we studied adhesion under flow to recombinant proteins. In this system, CX3CL1 was unable to support adhesion on its own but facilitated adhesion when coimmobilized with VCAM-1 by activating the α4β1-integrin. The recombinant CX3CL1 used in these experiments is not fully glycosylated, and because GPC-independent adhesion is mediated by the mucin domains that decorate transmembrane CX3CL1, this may explain its inability to directly support adhesion.

CX3CL1 has been shown to support monocyte adhesion to human umbilical vein endothelial cells under flow,17 and our observations extend its role to a liver-specific vascular bed and implicate it in transendothelial migration. CX3CL1 is expressed in human liver, and in the present study we confirm its expression on bile ducts and show induced expression in chronic inflammatory liver disease on sinusoids, where it could act to promote recruitment of CX3CR1+ cells from blood (Fig. 2).42, 43 CD16+ monocytes in blood showed homogenous intermediate levels of CX3CR1, whereas mDCs isolated from the liver included CX3CR1high cells and a population that expressed levels comparable to blood CD16+ monocytes. CD16+ monocytes lost CX3CR1 surface expression during the process of transmigration and as a consequence of receptor engagement by CX3CL1. Exposure to soluble CX3CL1 in vitro resulted in a profound but transient loss of cell surface CX3CR1 and could explain why prior exposure to soluble CX3CL1 prevents transendothelial migration; a similar effect has been reported for other chemokine receptors.44 Thus, CX3CL1 must be appropriately retained and presented on endothelium to function efficiently. Re-expression of CX3CR1 after cells have been recruited to the liver parenchyma could be important for their onward migration to areas of portal or lobular inflammation, where CX3CL1 is also strongly expressed42 (Fig. 2).

In addition to CX3CL1 and VCAM-1, VAP-1 was involved in CD16+ monocyte transendothelial migration under flow. VAP-1 belongs to an expanding family of ectoenzymes involved in cellular trafficking.45 VAP-1 is present on liver sinusoidal endothelium, where it has been implicated in lymphocyte recruitment in humans and rodents.27, 35, 36 Soluble VAP-1 is detected at high levels in the serum of patients with chronic liver disease, but not other inflammatory conditions such as rheumatoid arthritis.46, 47 VAP-1 can mediate sialic acid-dependent tethering and transendothelial migration of lymphocytes on sinusoidal endothelium.27, 48 This is the first time that VAP-1 has been implicated in monocyte transendothelial migration, although reduced monocyte recruitment to inflammatory sites has been reported in mice after VAP-1 blockade.49 We found that VAP-1 was involved in both adhesion and transendothelial migration. The combination of immobilized CX3CL1 and VAP-1 proteins on their own failed to support significant levels of adhesion, suggesting that VAP-1 operates in conjunction with other receptors to mediate transendothelial migration, consistent with data showing that enzymatic activity of VAP-1 modulates the expression of other adhesion molecules34 (Fig. 5). These findings add to an evolving body of literature that implicates VAP-1 as an important molecule in leukocyte transmigration across hepatic sinusoidal endothelium in vitro and in vivo and provide further evidence that the sinusoidal bed uses distinct combinations of molecules to recruit leukocytes to the liver parenchyma.25, 34-36, 50 The role of VAP-1 in transendothelial migration is particularly interesting. Monocyte transendothelial migration across other endothelium involves CD31, CD99, CD226, and the junctional adhesion molecules (JAMs), which are present at tight junctions22, 51-53; the lack of this migration in hepatic sinusoids could explain why VAP-1 has a dominant role in this vascular bed.

The recruitment of monocytes to the liver is a major factor in determining the outcome of hepatic inflammation. Inflammatory monocytes drive liver injury and fibrosis,54 myeloid DCs play critical roles in regulating immune responses to injury and infection, and M2 macrophages are central to the resolution of hepatic inflammation and scarring.1, 4, 55 The demonstration that VAP-1 and CX3CL1 are implicated in the recruitment of CD16+ monocytes across inflamed hepatic endothelium and that CD16+ cells localize at areas of inflammation and fibrosis has important implications for the pathogenesis of hepatic inflammation and the design of therapies to modulate monocyte recruitment in liver disease. The recent appreciation that VAP-1 can be inhibited by small molecule enzyme inhibitors opens up exciting therapeutic approaches to target this receptor. It is thus critical to understand which leukocytes rely on VAP-1 for entry into tissue and the outcome of inhibiting this receptor in vivo.27, 56, 57


We thank our clinical colleagues and patient donors for provision of blood and tissue samples.