Activation of vascular adhesion protein-1 on liver endothelium results in an NF-κB–dependent increase in lymphocyte adhesion


  • Potential conflict of interest: Nothing to report.


Vascular adhesion protein-1 (VAP-1) is an adhesion molecule and amine oxidase that is expressed at high levels in the human liver. It promotes leukocyte adhesion to the liver in vivo and drives lymphocyte transmigration across hepatic sinusoidal endothelial cells in vitro. We report that in addition to supporting leukocyte adhesion, provision of specific substrate to VAP-1 results in hepatic endothelial cell activation, which can be abrogated by treatment with the enzyme inhibitor semicarbazide. VAP-1–mediated activation was rapid; dependent upon nuclear factor-κB, phosphatidylinositol-3 kinase, and mitogen-activated protein kinase pathways; and led to upregulation of the adhesion molecules E-selectin, intercellular adhesion molecule-1, and vascular cell adhesion molecule-1 and secretion of the chemokine CXCL8. This response resulted in enhanced lymphocyte adhesion, was restricted to hepatic endothelial cells that expressed VAP-1, and was not observed in human umbilical vein endothelial cells. Conclusion: We propose that as well as directly promoting adhesion via interactions with the as yet unknown ligand, binding of enzyme substrate to VAP-1 can indirectly promote inflammatory cell recruitment via upregulation of adhesion molecules and chemokines. This response is likely to be important for the recruitment of leukocytes to the liver and suggests that VAP-1 inhibitors have therapeutic potential for treating chronic inflammatory liver disease. (HEPATOLOGY 2007;45:465–474.)

Immune function depends on the continuous recirculation of lymphocytes between blood and tissue regulated by molecular interactions between the circulating lymphocytes and ligands on the surface of endothelial cells.1 One such receptor is vascular adhesion protein-1 (VAP-1), a 170- to 180-kd homodimer.2 The cloning of VAP-1 revealed no similarity to any known adhesion molecule, but significant identity to the copper-binding semicarbazide-sensitive amine oxidases (SSAO) resulted in VAP-1 being classified as amine oxidase, copper containing-3 (AOC3).3, 4

VAP-1 is predominantly expressed as a membrane-associated or soluble protein in vascular endothelial5 and smooth muscle cells6 and adipocytes.7 VAP-1 is detected on normal and inflamed hepatic endothelium, where it supports the recruitment of specific lymphocyte subsets.8 It is also released as a soluble protein from the hepatic vascular bed, which accounts for most of the monoamine oxidase in human serum. Circulating levels are elevated in inflammatory liver disease, but not in other immunomediated conditions.9, 10 The enzyme activity of VAP-1 may be involved in detoxifying xenobiotic amines, regulating glucose uptake, modulating cell adhesion,11 and angiogenesis.12

Although the leukocyte ligand for VAP-1 has not been identified, evidence from several sources suggests that endothelial transmembrane VAP-1 regulates leukocyte adhesion and migration into tissue.8, 11, 13 Studies using either specific inhibitors of AOC3 activity or cells transfected with mutant forms of VAP-1 that lack the enzymatic active site demonstrate that enzymatic function is critical for transendothelial migration both in vitro and in vivo. The mechanism by which the enzyme modulates leukocyte recruitment has not yet been defined.14 One model proposes that leukocyte adhesion to VAP-1 mediated by sialic acid–containing side chains on the protein precedes enzyme activation during leukocyte interactions with endothelium.15 This adhesive interaction allows the leukocyte to present substrate to the enzyme with the generation of aldehydes and hydrogen peroxide, which activate promigratory pathways.2 In the present study, we report how the enzymatic activity of VAP-1 modulates the activation of human hepatic endothelium to promote leukocyte transendothelial migration.


AOC3, amine oxidase, copper containing-3; EMSA, electrophoretic mobility shift assay; HSEC, hepatic sinusoidal endothelial cell; HUVEC, human umbilical vein endothelial cell; ICAM-1, intercellular adhesion molecule-1; mAb, monoclonal antibody; MAPK, mitogen-activated protein kinase; MEK, mitogen-activated protein kinase kinase; NF-κB, nuclear factor-κB; PI3, phosphatidylinositol-3; SSAO, semicarbazide-sensitive amine oxidase; TNF-α, tumor necrosis factor-α; VAP-1, vascular adhesion protein-1.

Materials and Methods

Isolation and Culture of Human Hepatic Sinusoidal and Umbilical Vein Endothelial Cells.

Tissue was collected at the Queen Elizabeth Hospital or The Birmingham Women's Hospital after informed consent and approval by the local research ethics committee. Human umbilical vein endothelial cells (HUVECs) were isolated via collagenase digestion of umbilical cords according to standard procedures16 and cultured to confluence in complete endothelial media (Invitrogen, UK) containing 10% human AB serum (HD Supplies, UK), 10 ng/ml epidermal growth factor (Peprotec, UK) and 10 μg/ml hydrocortisone (Sigma, UK).

Human hepatic sinusoidal endothelial cells (HSECs) were isolated from 30 g human liver tissue as described previously.10 Briefly, tissue was enzymatically digested using collagenase Type 1A (Sigma), filtered, and purified via density gradient centrifugation using 33/77% (w/v) Percoll (Amersham Biosciences, UK). HSECs were extracted from the mixed nonparenchymal population via magnetic selection with antibody HEA125 (Progen Biotechnik, Germany) to deplete cholangiocytes followed by selection of endothelial cells with anti-CD31. Endothelial cultures were maintained in collagen-coated flasks in complete medium with 10 ng/ml vascular endothelial growth factor and 10 ng/ml hepatocyte growth factor (R&D Systems, UK). All cells were used within 4 passages of isolation except for cells that were used as controls after more prolonged passage and loss of cell surface VAP-1.


HSECs were plated into collagen-coated CultureSlides (BD Falcon, UK) and maintained until confluent before being fixed in ethanol for 5 minutes, washed with phosphate-buffered saline, and incubated with mouse anti-human VAP-1 (1B2 10 μg/ml; gift of S. Jalkanen17) for 60 minutes. Purified mouse IgG (X0931; Dako, UK) was used as a negative control. Cells were incubated with secondary rabbit anti-mouse FITC (1:100 dilution; BD Pharmingen, UK) for 30 minutes before washing and mounting. For nuclear factor-κB (NF-κB) p65 staining, cells were stimulated with tumor necrosis factor-α (TNF-α) (10 ng/ml), benzylamine (100 μM), or a combination of both for 1 hour before fixation and permeabilization using 3.7% formaldehyde and 0.1% Triton-X 100. Cells were incubated with NF-κB p65 antibody (BD Pharmingen) or isotype control and washed with phosphate-buffered saline before incubation with Alexa 594-labeled secondary monoclonal antibody (mAb) (1:400 dilution; Molecular Probes, UK) for 30 minutes before washing and mounting.

Electrophoretic Mobility Shift Assay.

Activation of NF-κB was measured via electrophoretic mobility shift assay (EMSA). Cells were cultured in the presence of 1B2 (10 μg/ml) or benzylamine plus sodium orthovanadate (100 μM; Sigma) for the periods indicated. Ligand binding to VAP-1 was mimicked by cross-linking 1B2 with rabbit anti-mouse (50 μg/ml ZO259; Dako). Stimulation for 2 hours with TNF-α (10 ng/ml; Peprotech) was a positive control for NF-κB activation. VAP-1–mediated NF-κB activation was inhibited using caffeic acid phenethyl ester (CAPE, 25 μg/ml, 30 minutes pretreatment; Calbiochem). The contribution of phosphatidylinositol-3 (PI3) kinase, p38 mitogen-activated protein kinase (MAPK), and mitogen-activated protein kinase kinase (MEK) to endothelial NF-κB activation was assessed via pretreatment with the inhibitors LY294002 (50 μM, 30 minutes pretreatment; Biosource, UK), SB203580 (100 μM, Calbiochem), and PD98059 (50 μM, Calbiochem), respectively, for 30 minutes before incubation with VAP-1 substrate. The selectivity of activation responses was confirmed using the VAP-1 enzyme activity inhibitor semicarbazide hydrochloride (100 μM; Sigma). Following treatment, nuclear extracts were prepared from the endothelial cells using standard protocols,18 and protein content was determined by the bicinchoninic acid protein assay (Pierce Chemical Co., Rockford, IL). EMSAs were performed as described previously18 using 32P end-labeled NF-κB oligoduplex probes (Promega, UK) with binding specificity confirmed via supershift assays using specific antisera to NF-κB RelA (p65) (1:300 dilution of stock SC-8008X; Santa Cruz Biotechnology, Santa Cruz, CA).

PCR Analysis.

Total cellular RNA was extracted from endothelial cells using TRI Reagent kit (Sigma) and RNA purified and precipitated using standard chloroform/isopropanol methodology and reverse-transcribed into complementary DNA using dNTP mix (Sigma), random hexamers (Amersham Biosciences), and Superscript II RnaseH-reverse transcriptase (Invitrogen). Sequence-specific oligonucleotide primers for CXCL8 and E-selectin were obtained from ALTA Bioscience, University of Birmingham, UK (CXCL8: forward 5′-CTTGGCAGCCTTCCTGATTT-3′, reverse 5′-CTCAGCCCTCTTCAAAAACT-3′; E-selectin: forward 5′-CTCTGACAGAAGAAGCCAAG-3′, reverse 5′-ACTTGAGTCCACTGAAGTCA-3′). Polymerase chain reaction was performed using a 25-μl solution containing 0.5 μl complementary DNA, 10× PCR buffer (Sigma), 200 μM deoxyribonucleoside triphosphate, 25 pmol primer mix, 2 mM MgCl2 (Sigma), and 2.5 U Taq polymerase (Invitrogen). Samples were overlaid with mineral oil and incubated for 3 minutes at 94°C, followed by 30 cycles of 1 minute at 94°C, 1 minute at 52°C for E-selectin, or 60°C for IL-8 and 1 minute at 72°C plus a final extension for 10 minutes at 72°C in a PTC-100 Programmable Thermal Controller (MJ Research Inc). Amplified complementary DNA was separated on a 2% agarose gel in a Tris–acetic acid EDTA buffer and visualized with Quantity One Software on Gel Doc 2000 (Bio-RAD).


Adhesion molecule expression was quantified using a cell-based ELISA as described.11 Endothelial cells were cultured in 96-well plates until confluent. Following incubation with VAP-1, substrate and inhibitor cells were methanol-fixed, washed with phosphate-buffered saline, and incubated with blocking buffer for 60 minutes before being incubated with mouse anti-human mAb (5 μg/ml; Dako). Staining was detected with peroxidase-conjugated goat anti-mouse and OPD-substrate read on a Synergy-HT plate reader. Isotype-matched mouse mAb was used as a control. Data are expressed as the mean absorbance of triplicate wells minus isotype-matched controls.

CXCL8 protein secreted into the endothelial culture media was determined using a commercial capture ELISA kit (R&D Systems) according to the manufacturer's instructions.

Flow-Based Adhesion Assay.

To determine the physiological consequences of VAP-1–mediated NF-κB activation on lymphocyte adhesion to HSECs under physiological blood flow, HSECs were grown to confluence in glass capillary tubes before being stimulated for 4 hours with TNF-α (10 ng/ml), benzylamine (200 μM), or both, and connected to the flow-based system as described previously.19 Peripheral blood lymphocytes were perfused through microslides at a physiologically relevant hepatic wall shear-stress of 0.05 Pa. Adherent cells were visualized microscopically under phase contrast and recorded for offline analysis. The number of adherent cells was converted to adherent cells per millimeter2 and corrected for the number of cells perfused. The pattern of adhesion was analyzed to determine the number of cells rolling, statically adherent, or transmigrated.

Statistical Analyses.

Data were analyzed using SPSS software (SPSS Inc.). Differences between groups were compared using a paired Student t test.


Low-passage HSECs expressed VAP-1 (Fig. 1A), whereas no VAP-1 was detected on the surface of cultured HUVECs. Staining with an isotype-matched control antibody was negative on both. VAP-1 has been shown to mediate SSAO-dependent leukocyte adhesion and transendothelial migration,20 leading us to hypothesize that activation of VAP-1 may trigger intracellular signaling cascades, resulting in endothelial cell activation. Because the identity of the physiological ligand for VAP-1 is unknown, we used 2 strategies to mimic VAP-1 activation on HSECs. VAP-1 on the cell surface was cross-linked using antibodies or the effect of VAP-1 enzyme activity was investigated by providing the specific SSAO substrates benzylamine or hydroxylamine. Substrates were given with vanadate to inhibit protein tyrosine phosphatases and prolong phosphorylation of signaling intermediates, allowing us to detect the downstream signaling pathways involved. Endothelial activation was determined by activation of NF-κB. Antibody cross-linking of VAP-1 on HSECs resulted in activation of NF-κB, whereas treatment of HUVECs that lacked detectable amounts of VAP-1 did not (Fig. 1B). VAP-1 substrate—either benzylamine or hydroxylamine—also led to activation of NF-κB in HSECs, which was prevented by pretreatment of the cells with the SSAO inhibitor semicarbazide (Fig. 1C,D). Again, the response was specific to HSECs, because no activation was detected when HUVECs were treated with benzylamine. Fluorescence microscopy confirmed that benzylamine resulted in increased nuclear localization of NF-κB p65 in HSECs (Fig. 1E) but not HUVECs. Treatment of HSECs with a combination of benzylamine and TNF-α had an additive effect.

Figure 1.

VAP-1 activation on human HSECs results in activation of NF-κB. (A) HSECs express VAP-1 in culture. HSECs grown to confluence in chamber slides were fixed and stained with mAb directed against VAP-1 (1B2, 10 μg/ml). The strong green staining indicates VAP-1 expressing HSECs. No staining was seen with HUVECs. (B-D) Analysis of nuclear extracts from HSECs or HUVECs via EMSA. (B) Anti–VAP-1 antibodies activate NF-κB in HSECs. A representative EMSA in which HSECs or HUVECs were treated with TNF-α (10 ng/ml, 2 hours), cross-linked VAP-1 antibody (1B2 10 μg/ml plus rabbit anti-mouse secondary 50 μg/ml, 1 h; VAP-1 cross-link), VAP-1 antibody alone (1st mAb, 1B2 10 μg/ml, 1 hour), or secondary antibody alone (2nd mAb, 50 μg/ml, 1 hour) is shown. TNF-α activates NF-κB in both cell types. Anti–VAP-1 antibody either alone or cross-linked activates NF-κB in HSECs but not HUVECs. (C) VAP-1 substrates activate NF-κB in HSECs. A representative EMSA is shown in which HSECs or HUVECs were treated with TNF-α (10 ng/ml, 2 hours), benzylamine plus sodium orthovanadate (Benz+Van, both 100 μM, 1 hour) or semicarbazide (100 μM, 30 minutes pretreatment and maintained concurrently with substrate treatment). Whereas TNF-α–activated NF-κB in both cell types, increased activation in response to substrates was only seen in HSECs. (D) Pooled densitometry data generated by analysis of five replicate experiments using different primary isolates of HSECs and HUVECs. Data are expressed as the mean ± SEM increase in density of NF-κB bands compared with signal generated from a control lane of nuclear extract with control-labeled probe alone (Control, normalized to 1). *Significant increase in relative density compared with control (paired Student t test; P ≤ 0.05). (E) Activation of AOC3 enzyme activity in cultured endothelial cells results in nuclear translocation of NF-κB p65. HSECs grown to confluence in chamber slides were fixed, permeabilized, and stained with mAb directed against NF-κB detected using an Alexa 594-conjugated secondary antibody. Unstimulated HSECs (control) show predominantly cytoplasmic localization of NF-κB p65, whereas cells treated with TNF-α (10 ng/ml), benzylamine (200 μM), or both for 1 hour demonstrate relocalization of p65 to the nucleus.

The specificity of EMSA was confirmed using a 100-fold molar excess of unlabeled oligonucleotide probe for NF-κB and supershifts using mAb against NF-κB RelA (p65). Figure 2A shows EMSA generated using extracts prepared from HSECs treated with benzylamine and vanadate. Binding of NF-κB consensus sequence to nuclear extract was eliminated in the presence of an excess of unlabeled probe, and the migration of the complex was retarded by addition of antibody raised against NF-κB p65. Pretreatment of HSECs with the NF-κB inhibitor CAPE inhibited substrate induced NF-κB activation. CAPE alone did not activate NF-κB, whereas treatment with VAP-1 substrate activated NF-κB within 2 hours and persisted for up to 8 hours. Pretreatment of HSECs with CAPE for 30 minutes was sufficient to prevent substrate-induced activation of NF-κB for up to 8 hours after treatment (Fig. 2B).

Figure 2.

Hybridization of labeled NF-κB oligonucleotide probe with HSEC nuclear extracts is specific and activation of NF-κB is dependent upon PI3 kinase, p38, and MEK activity. (A) Representative EMSA for treatment of HSECs with TNF-α (10 ng/ml, 2 hours) or benzylamine plus sodium orthovanadate (Benz+Van, both 100 μM, 1 hour). Specificity of the NF-κB band was confirmed by addition of specific mouse monoclonal antisera to NF-κB RelA (p65) (1:300 dilution of manufacturer's stock; NF-κB p65) or a 100-fold molar excess of unlabeled probe (NF-κB unlabeled) to nuclear extract from cells treated with VAP-1 substrate plus vanadate. (B) Representative EMSA in which HSECs were treated with TNF-α (10 ng/ml, 2 hours) or benzylamine plus sodium orthovanadate (Benz+Van, both 100 μM, 2-8 hours as indicated). Some samples were treated with the selective NF-κB inhibitor CAPE (25 μg/ml for 8 hours alone, or for 30 minutes before substrate treatment with subsequent analysis at 2-8 hours). Activation of NF-κB occurs within 2 hours and has diminished by 24 hours. CAPE completely prevents NF-κB activation in response to benzylamine and vanadate. (C) Representative EMSA and densitometry data for experiments in which inhibitors were used to block activation of NF-κB in response to treatment of HSECs with TNF-α (10 ng/ml, 2 hours) or benzylamine plus sodium orthovanadate (Benz+Van, both 100 μM, 1 hour). All inhibitors were added for 30 minutes before substrate administration; PI3 kinase inhibitor LY294002 (50 μM), p38 inhibitor SB203580 (100 μM), MEK inhibitor PD 98059 (50 μM); semicarbazide (100 μM). EMSA data represents results from HSECs isolated from a representative liver specimen. Data in the densitometry graph are expressed as the mean ± SEM increase in density of NF-κB bands from unstimulated cells compared with signal generated from a control lane of nuclear extract with control-labeled probe alone (Cont). *Significant difference between inhibitor-treated samples compared with cells treated with benzylamine plus vanadate alone (paired Student t test; P ≤ 0.02; n = 4 samples from different cell isolates).

PI3 kinase has been implicated in SSAO signaling in adipocytes.21 Thus, we pretreated HSECs with inhibitors of PI3 kinase (LY294002), p38 MAPK (SB203580), or MEK (PD98059) for 30 minutes before incubation with VAP-1 substrate. Treatment with LY294002 almost completely abolished activation of NF-κB, and treatment with the MEK and p38 inhibitors also reduced NF-κB activation, suggesting that these pathways are involved in VAP-1–mediated activation of NF-κB (Fig. 2C). Sodium orthovanadate was included in the substrate incubations to inhibit protein tyrosine phosphatases and stabilize phosphorylated proteins and kinase activation.22 However, because protein tyrosine phosphorylation may alter cell adhesion,23 it was important that we considered effects mediated by vanadate alone. Figure 3A shows that vanadate can activate NF-κB in HSECs. However, the time course of NF-κB activation was different from that seen in combination with benzylamine or when benzylamine was used alone (Fig. 3A), and the level of NF-κB activation was increased by the combination of benzylamine plus vanadate compared with either compound alone. Furthermore, the specific SSAO enzyme inhibitor semicarbazide blocked the activation of NF-κB seen in response to benzylamine alone or in combination with vanadate, but not the activation induced by vanadate alone (Fig. 3B).

Figure 3.

Benzylamine with or without vanadate induces rapid activation of NF-κB activation in HSECs that can be inhibited by semicarbazide. (A) Representative EMSA for treatment of HSECs with benzylamine (100 μM for times indicated), sodium orthovanadate alone (Vanadate, 100 μM for times indicated), or benzylamine plus sodium orthovanadate (Benz+Van, both 100 μM, for times indicated). Benzylamine and benzylamine plus vanadate induced early activation of NF-κB that was detectable at 15 minutes, whereas vanadate alone had a delayed effect that was evident at 60 minutes. (B) Representative EMSA for treatment of HSECs with benzylamine alone (Benzylamine, 100 μM, 1 hour), sodium orthovanadate alone (Vanadate, 100 μM, 1 hour), or benzylamine plus sodium orthovanadate (Benz+Van, both 100 μM, 1 hour) in the presence or absence of semicarbazide (100 μM, 30 minutes pretreatment and maintained concurrently with substrate). NF-κB activation by vanadate alone was not inhibited by semicarbazide, whereas NF-κB activation in response to benzylamine and vanadate was.

To assess whether NF-κB activation in response to VAP-1 substrate treatment has physiological consequences related to leukocyte adhesion, we looked at the activation of adhesion-related genes that are downstream targets of NF-κB. that transcription of both E-selectin and CXCL8 (IL-8) increased in HSECs within 2 hours of treatment with benzylamine plus vanadate and was maintained for up to 24 hours (Fig. 4A). Vanadate treatment alone did not increase expression (Fig. 4A), and substrate-induced expression was inhibited by treatment of cells with semicarbazide. Analysis of CXCL8 protein production via ELISA confirmed that endothelial cells secreted significant amounts of protein within 6 hours of treatment (Fig. 4B).

Figure 4.

VAP-1 substrate-dependent NF-κB activation in HSECs is associated with upregulation of adhesion molecules and chemokines. (A) Representative reverse-transcription polymerase chain reaction data for upregulation of E-selectin and CXCL8 messenger RNA after treatment of HSECs grown in T75 flasks with lipopolysaccharide (10 μg/ml, 4 hours) as positive control, benzylamine plus sodium orthovanadate (Benz+Van, both 100 μM, for times indicated), or sodium orthovanadate alone (Vanadate, 100 μM, 8 hours) in the presence or absence of semicarbazide (100 μM, 30 minutes pretreatment and maintained concurrently with substrate). Benzylamine plus vanadate induced dose-dependent induction of CXCL8 and E-selectin messenger RNA that was inhibited by semicarbazide, whereas vanadate alone had no effect. (B) VAP-1 substrates also induced secretion of CXCL8 protein into HSEC supernatant. CXCL8 after VAP-1 substrate treatment was determined by commercially available capture ELISA with specific monoclonal antibodies. Cells were left unstimulated (Control) or treated with benzylamine plus vanadate (B+V, 100 μM for time indicated) before collection of supernatant. Signal from triplicate wells was compared with a standard curve of known protein concentration. The data are expressed as the mean ± SEM of 4 replicate experiments using different isolates of cells. *Significantly elevated production of CXCL8 protein compared with unstimulated control cells (paired Student t test; P ≤ 0.01).

We then investigated the expression of cell surface adhesion molecules on endothelial cells following VAP-1 activation by substrate. Because vanadate has been reported to alter levels of adhesion molecules on HUVECs,23 we used benzylamine alone to stimulate the endothelial cells. Benzylamine treatment for 4 hours increased expression of both intercellular adhesion molecule-1 (ICAM-1) and E-selectin by HSECs in a concentration-dependent manner. There was also a trend for increased expression of ICAM-1 (Fig. 5A,B). When HSECs were treated with semicarbazide plus benzylamine, we observed significant, dose-dependent inhibition of induction of all three adhesion molecules. The functional relevance of these observations was tested in a flow-based adhesion assay. Treatment of low-passage HSECs that express demonstrable membrane VAP-1 (Supplementary Fig. 1) increased the total adhesion of lymphocytes compared with untreated control cells. The effect is less pronounced than the stimulation of adhesion observed following treatment of HSECs with TNF-α (Fig. 6A), and combined treatment with TNF-α and benzylamine results in a level of adhesion similar to that seen in the presence of TNF-α alone. When higher-passage cells lacking membranous VAP-1 were used, the adhesion-promoting effect of benzylamine was abolished, whereas the cells responded appropriately to TNF-α (Fig. 6B). Benzylamine treatment increased the number of adherent lymphocytes that underwent sustained rolling adhesion (Fig. 6C) or endothelial transmigration (Fig. 6D) to a similar degree seen with TNF-α.

Figure 5.

Benzylamine treatment of HSECs is associated with upregulation of vascular cell adhesion molecule-1, ICAM-1, and E-selectin protein, and the effect can be inhibited by semicarbazide treatment. (A) Cell-based ELISA data for treatment of HSECs with benzylamine at the indicated doses (50-200 μM) for 4 hours. Data represent absorbance signals from triplicate wells minus signals from isotype-matched controls and are expressed as the mean ± SEM of 4 experiments with different isolates of cells. *Significantly elevated protein expression compared with unstimulated control cells (0 μM; paired Student t test; P ≤ 0.05). Benzylamine alone induced significant dose-dependent increases in expression of adhesion molecules which was inhibited by semicarbazide. (B) Cell-based ELISA data for treatment of HSECs with benzylamine at 100 μM for 4 hours in the absence or presence of semicarbazide at the indicated doses. Data represent absorbance signals from triplicate wells minus signals from isotype-matched controls and are expressed as the mean ± SEM of 4 experiments with different isolates of cells. *Significantly elevated protein expression compared with unstimulated controls; #significantly decreased protein expression compared with cells treated with substrate alone (paired Student t test; P ≤ 0.05 for all).

Figure 6.

Benzylamine treatment of VAP-1 expressing HSECs is associated with increased ability to recruit lymphocytes under conditions of blood flow. Flow-based adhesion assay data for peripheral blood lymphocytes perfused over HSECs that had been pretreated with benzylamine (200 μM), TNF-α (10 ng/ml), or a combination of both for 4 hours. Peripheral blood lymphocytes were perfused over HSECs at a wall shear stress of 0.05 Pa, and adhesion events were counted and classified into static adhesion, (C) percent rolling, or (D) percent of cells that transmigrated through the endothelial monolayer. Total adhesion was normalized to the number of cells/mm2/million perfused. (A,B) Total adhesion of peripheral blood lymphocytes to HSEC monolayers that tested positive or negative for membranous expression of VAP-1, respectively (Supplementary Fig. 1). Data are expressed as the mean ± SEM adhesion from 4 or 3 replicate experiments, respectively, using different cell isolates on each occasion. (C,D) Percentage of adherent cells on VAP-1–positive HSECs that rolled or transmigrated. Data are expressed as the mean ± SEM adhesion from four experiments. Paired Student t tests were used to compare adhesion events on treated compared with unstimulated HSECs. *P = 0.05. **P ≤ 0.05. ***P ≤ 0.01.


The liver microvasculature is one of the few extralymphoid vascular beds where VAP-1 is constitutively expressed,24 and in vitro and in vivo studies demonstrate an important role for VAP-1 in lymphocyte recruitment via the hepatic sinusoids.7, 10, 25 We previously reported that VAP-1 supports adhesion and transendothelial migration across human sinusoidal endothelium under flow and that this process is dependent on AOC3 enzyme activity.11 These findings are supported by recent in vivo studies. Bonder et al.8 reported a role for VAP-1 in the recruitment of Th2 CD4+ T cells via liver sinusoids and postsinusoidal vessels during concanavalin A–induced liver inflammation in the mouse, and Martelius et al.26 demonstrated that antibody blockade of VAP-1 reduces lymphocyte recruitment in a rat model of liver allograft rejection. Thus, VAP-1 appears to have a particular role in lymphocyte recruitment in the liver. Our own studies suggest this function is dependent, at least in part, on the AOC3 enzyme activity of VAP-1, because specific enzyme inhibitors are as effective as antibody in blocking lymphocyte transendothelial migration.11

Salmi et al.5 hypothesized that the VAP-1 ligand on leukocytes is an amino sugar or free NH2 group in an amino acid constituent of a surface protein that interacts with the enzyme after initial sialic acid–dependent adhesion. In the absence of an identified ligand, we adopted 2 strategies to investigate the downstream events in endothelial cells after binding of VAP-1. First, we cross-linked cell surface VAP-1 using antibody, an approach that has been shown to activate other endothelial adhesion molecules.27 We demonstrated activation of NF-κB in HSECs following both binding of VAP-1 primary antibody alone and when the mAb was cross-linked with a secondary antibody. We compared the response in HSECs to HUVECs. Although HUVECs do not express cell surface VAP-1,28 some isolates contain cytoplasmic VAP-1 devoid of the sialic acid residues essential for its function as an adhesion molecule.28 We were unable to detect surface VAP-1 in HUVECs by immunocytochemistry and flow cytometry but were able to detect a pool of cytoplasmic protein (Supplementary Fig. 1). Consistent with this lack of membranous VAP-1 neither antibody cross-linking nor substrate treatment resulted in NF-κB activation in HUVECs. Thus, we only observed activation of NF-κB in HSECs that expressed membranous VAP-1.

The observation that primary antibody alone was sufficient to induce activation of VAP-1 in the absence of cross-linking secondary antibody was surprising and suggests that antibody binding may mimic ligand occupancy, resulting in the formation of VAP-1 multimers and the generation of intracellular signals. However, the antibody we used in our study only recognizes monomeric VAP-19 and thus is unlikely to cause multimerization without cross-linking. Furthermore, the cytoplasmic tail of VAP-1 is short and contains only four amino acids, suggesting that cross-linking of the receptor would be unlikely to directly recruit intracellular signaling molecules. Alternatively, ligand/antibody binding may act to stabilize a complex involving other transmembrane proteins or make VAP-1 accessible to cell surface–associated substrate. Most function blocking antibodies raised against VAP-1 do not block the enzymatic capacity of the molecule. Indeed, there is a suggestion that the structures required for enzymatic activity are so well conserved between species that they are not immunogenic.15 Thus, our demonstration of NF-κB activation following antibody ligation, which does not alter enzyme function, implies that such activation can proceed independent of enzyme activity. However, we cannot completely discount the possibility that the antibody effect may relate to signaling via Fc receptors that are expressed constitutively on HSECs; thus, we used the model substrate benzylamine to activate endothelial cells in subsequent studies.

Treatment of HSECs with benzylamine plus vanadate induced rapid activation of NF-κB that was not observed in HUVECs. The effect was inhibited by pretreatment of the HSECs with the NF-κB inhibitor CAPE or the enzyme inhibitor semicarbazide. We then used inhibitors to investigate the involvement of specific kinases and found that the PI3 kinase inhibitor LY294002 completely inhibited NF-κB activation in response to VAP-1. PI3 kinase regulates many signaling pathways that lead to activation of NF-κB,29 including activation of MAPKs, a family of serine/threonine protein kinases with well-established roles in the regulation of inflammatory gene expression. One member of the family, p38, is activated by MEKs such as MKK3 and MKK4 and is responsible for the phosphorylation of transcription factors such as NF-κB by way of complex mechanisms that may involve modulation of other transcription factors.30 Our finding of reduced NF-κB activation following p38 and MEK inhibition suggests that AOC3 enzyme activity leads to activation of PI3 kinase in hepatic endothelial cells and subsequent MAPK-dependent activation of p38, resulting in nuclear translocation of NF-κB and proinflammatory gene activation. Alternatively, PI3 kinase and the MAP kinase signaling pathways may be activated in parallel, but the induction of NF-κB is dependent on both signals. In adipocytes, AOC3-mediated NF-κB activation has been shown to be dependent on PI3 kinase.21, 31

We used sodium orthovanadate with benzylamine in some experiments to inhibit protein tyrosine phosphatases and increase the half-life of phosphorylated signaling intermediaries allowing their detection.22 However, 100 μM vanadate (the concentration used in this study) can enhance benzylamine-stimulated, VAP-1–dependent glucose uptake, PI3 kinase activity, and receptor tyrosine phosphorylation in adipocytes, despite the fact that vanadate alone has no effect.21, 31 This led us to investigate the effect of benzylamine alone and in combination with vanadate. In contrast to the previous studies with adipocytes, vanadate alone caused NF-κB activation in HSECs, although the time course was different when compared with benzylamine alone or combinations of benzylamine and vanadate. Benzylamine caused early activation of NF-κB that lasted for under 2 hours, whereas vanadate took longer to have an effect but ultimately persisted. It has been proposed that vanadate can cause activation of both NF-κB and c-Jun N-terminal kinase in macrophage cell lines.32 We confirmed that the effect of benzylamine was specific to AOC3 activity, because semicarbazide inhibited NF-κB activation in response to either benzylamine alone or benzylamine in combination with vanadate but had no effect on cells treated with vanadate alone.

Semicarbazide has inhibitory effects on another amine oxidase, lysyl oxidase,33 which is expressed in vascular endothelial cells and for which benzylamine is a substrate.34 In the liver, lysyl oxidase is expressed by hepatocytes and myofibroblasts and is responsible for promoting the cross-linking of matrix constituents such as collagen during hepatic fibrosis. However, it is unlikely that lysyl oxidase is contributing to the effect seen in our studies, because it has not been reported in sinusoidal endothelial cells; in addition, HUVECs, which have demonstrable lysyl oxidase,35 did not show NF-κB activation in response to benzylamine treatment. Thus, NF-κB activation in HSECs in the presence of benzylamine can be attributed to VAP-1 activity.

H2O2 is produced during VAP-1–mediated amine oxidation,33 and in vitro studies with cultured endothelial cells have demonstrated NF-κB activation and upregulation of adhesion molecules and chemokines in response to treatment with H2O2.20, 36–38 In this study, we demonstrated secretion of the chemokine CXCL8 and increased cell surface expression of vascular cell adhesion molecule-1, ICAM-1 and E-selectin after treatment of HSECs with benzylamine. This effect was significantly inhibited if the endothelial cells were treated with semicarbazide, confirming the contribution of AOC3 to the process. It has been suggested that VAP-1 promotes stable rolling adhesion of neutrophils within mesenteric vessels in peritonitis39 and, because this model is dependent on the expression of ICAM-1 and selectins, part of the effect of VAP-1 could be mediated by altered expression of these molecules. H2O2 produced by VAP-1–dependent oxidation of primary amines increases the binding of amino sugars to VAP-1,40 which might stabilize binding between VAP-1 and a glycated ligand. Our demonstration that AOC3 enzyme activity results in rapid activation of NF-κB within 15 minutes of substrate interaction with VAP-1, combined with upregulation of CXCL8 and E-selectin within 2 hours, suggests that the adhesive functions of VAP-1 may be linked to its capacity to rapidly regulate the expression and function of other pathways. We confirmed the physiological relevance of AOC3 activation on HSECs using flow-based adhesion assays. Activation of membranous VAP-1 on HSECs with benzylamine resulted in a significant increase in total lymphocyte adhesion that was not seen on HSECs that failed to express membranous VAP-1. The increased adhesion resulting from AOC3 enzyme activity was less than that observed with optimal doses of TNF-α, suggesting that, in the presence of TNF-α, other pathways also contribute. However, signals resulting from activation of constitutively expressed VAP-1 in the liver are likely to be physiologically relevant, because AOC3 activation by benzylamine increased both the rolling and transmigration steps of lymphocyte recruitment. In addition, the circulating form of AOC3, which retains enzyme activity and is secreted by hepatic endothelium and adipose tissue, may have functional effects in organs remote from the site of origin.

In conclusion, we have demonstrated that a rapid activation of NF-κB occurs after binding of the enzyme substrate benzylamine to VAP-1 expressed in hepatic endothelial cells. This results in expression of proinflammatory proteins such as CXCL8, E-selectin, and ICAM-1 and the upregulation of leukocyte adhesion, adding a new dimension to the prevailing multistep model of leukocyte adhesion. Although our study was focused upon the significance of VAP-1 in the liver microenvironment, similar mechanisms may apply to other vascular beds where VAP-1 is expressed at the endothelial cell surface. Furthermore, the evidence that soluble VAP-1 can influence leukocyte recruitment in vitro9 suggests that the increased concentrations of enzymatically active, soluble VAP-1 demonstrated in inflammatory disease may affect inflammatory responses in vivo.9 The ability to disrupt VAP-1/AOC3 function with small molecule inhibitors suggests a promising avenue for the development of novel anti-inflammatory therapies in chronic liver disease.


We are grateful to our colleagues at the Queen Elizabeth Hospital, Birmingham, UK, for help with sample collection.