Enhanced T cell transmigration across the murine liver sinusoidal endothelium is mediated by transcytosis and surface presentation of chemokines


  • Potential conflict of interest: Nothing to report.


Transmigration through the liver endothelium is a prerequisite for the homeostatic balance of intrahepatic T cells and a key regulator of inflammatory processes within the liver. Extravasation into the liver parenchyma is regulated by the distinct expression patterns of adhesion molecules and chemokines and their receptors on the lymphocyte and endothelial cell surface. In the present study, we investigated whether liver sinusoidal endothelial cells (LSEC) inhibit or support the chemokine-driven transmigration and differentially influence the transmigration of pro-inflammatory or anti-inflammatory CD4+ T cells, indicating a mechanism of hepatic immunoregulation. Finally, the results shed light on the molecular mechanisms by which LSEC modulate chemokine-dependent transmigration. LSEC significantly enhanced the chemotactic effect of CXC-motif chemokine ligand 12 (CXCL12) and CXCL9, but not of CXCL16 or CCL20, on naive and memory CD4+ T cells of a T helper 1, T helper 2, or interleukin-10–producing phenotype. In contrast, brain and lymphatic endothelioma cells and ex vivo isolated lung endothelia inhibited chemokine-driven transmigration. As for the molecular mechanisms, chemokine-induced activation of LSEC was excluded by blockage of Gi-protein–coupled signaling and the use of knockout mice. After preincubation of CXCL12 to the basal side, LSEC took up CXCL12 and enhanced transmigration as efficiently as in the presence of the soluble chemokine. Blockage of transcytosis in LSEC significantly inhibited this effect, and this suggested that chemokines taken up from the basolateral side and presented on the luminal side of endothelial cells trigger T cell transmigration. Conclusion: Our findings demonstrate a unique capacity of LSEC to present chemokines to circulating lymphocytes and highlight the importance of endothelial cells for the in vivo effects of chemokines. Chemokine presentation by LSEC could provide a future therapeutic target for inhibiting lymphocyte immigration and suppressing hepatic inflammation. (HEPATOLOGY 2008.)

Homeostatic extravasation into the liver parenchyma and transmigration of T cells into sites of inflammation under pathological conditions such as hepatitis and cholangitis are mediated by multiple steps involving activation and firm adhesion followed by egress. The process is closely regulated by the distinct expression patterns of adhesion molecules and chemokines and their receptors on the cell surface.1, 2 Within the liver, lymphocyte adhesion, a prerequisite for transmigration, occurs mainly within the sinusoidal circulation. Thus, T cell immigration into the liver is presumed to happen predominantly via the sinusoids.3 Liver sinusoidal endothelial cells (LSEC) lining the sinusoids form a morphologically and functionally distinct endothelium that lacks a basement membrane, fails to form tight junctions, and displays fenestrae grouped to sieve plates.4 In contrast to other tissues, selectins are not necessary for the initial adhesion of leukocytes to the liver microvasculature. The slow hepatic blood flow and the small luminal diameter of the sinusoids seem to be sufficient for the first interaction.5 Subsequently, firm adhesion can occur through constitutively expressed adhesion molecules, such as intracellular adhesion molecule-16 and vascular adhesion protein-1 in humans.3 Chemokines can activate integrins and support firm adhesion by binding lymphocyte integrins to the immunoglobulin superfamily adhesion molecules on the endothelial cell. Ultimately, cells leave the blood vessel and enter the extravascular tissue in response to chemokines.1, 2

Chemokines are classified by typical amino acid motifs. The largest groups are CC chemokines with conserved tandem cysteine residues and CXC chemokines with one amino acid separating adjacent cysteine residues.7 CXC-motif chemokine ligand 12 (CXCL12) is constitutively expressed in many tissues, such as bone marrow, lung, and liver tissues. Its receptor, CXC-motif chemokine receptor 4 (CXCR4), is found on the surface of naive and memory CD4+ T cells.8 Consequently, CXCL12 promotes chemotaxis of naive and antigen-experienced CD4+ T cells such as in vitro polarized9 or ex vivo isolated T helper 1 (Th1) and T helper 2 (Th2) cells.10 CXCL9, predominantly synthesized under inflammatory conditions, is also found in the healthy liver but not in the spleen, gut, lungs, or brain. The only known receptor for CXCL9, CXCR3, is restricted to effector/memory cells.11, 12 Although CXCR3 is preferentially associated with a Th1 phenotype,7 CXCL9 also promotes transmigration of Th2 cells.10

Against the background of the particular immunological capacity of the liver to support the induction of peripheral tolerance rather than to establish immunity,13, 14 we hypothesized whether LSEC might possess distinct gatekeeper functions for leukocytes. Here we used a transwell assay system and time-lapse microscopy with ex vivo isolated murine LSEC to investigate whether LSEC inhibit or support the chemokine-driven transmigration of CD4+ T cells. Second, we wished to clarify whether LSEC differentially influence the transmigration of distinct pro-inflammatory or anti-inflammatory CD4+ T cell subsets such as Th1 or Th2 cells. Finally, the molecular mechanisms by which LSEC might modulate chemokine-dependent transmigration are addressed.


AcLDL, acetylated low-density lipoprotein; CXCL, CXC-motif chemokine ligand; CXCR, CXC-motif chemokine receptor; IFNγ, interferon gamma; IL, interleukin; LSEC, liver sinusoidal endothelial cell(s); Mac-1, macrophage 1; RPMI 1640, Roswell Park Memorial Institute 1640 medium; SD, standard deviation; Th1, T helper 1; Th2, T helper 2.

Materials and Methods


Directly labeled antibodies against surface markers and cytokines were purchased from BD Biosciences (Heidelberg, Germany), if not indicated otherwise. Anti-CXCL12 (79018) was from R&D Systems (Minneapolis, MN). Goat anti-mouse Alexa 594 was obtained from Molecular Probes (Leiden, The Netherlands). Anti-CD4 (GK1.5) and anti-interferon gamma (anti-IFNγ; XMG6) were kindly provided by the Deutsches Rheuma-Forschungszentrum (Berlin, Germany). Hybridomas for the preparation of supernatants for panning [anti-CD8, TIB105; anti-macrophage 1 (anti–Mac-1), M1/70; anti-Fcγ receptor II/III, 2.4G2; and anti-CD62L, MEL-14] were obtained from the American Type Culture Collection (Manassas, VA). The monoclonal mouse endothelial cell–reactive antibodies ME-9F1 and MECA32 were raised as previously described.15 Rat immunoglobulin G and goat serum were obtained from Dianova-Jackson (Hamburg, Germany).


Brefeldin A, bovine serum albumin, collagenase IV, chlorpromazine, cytochalasin B, DNase I, filipin, Histopaque-1083, ionomycin, nystatin, phorbol myristate acetate, pertussis toxin, saponin, and Triton X-100 were purchased from Sigma-Aldrich (Steinheim, Germany). Paraformaldehyde, glutaraldehyde, and sodium cacodylate buffer were from Merck (Darmstadt, Germany). Bovine plasma fibronectin, Dulbecco's modified Eagle's medium, L-glutamine, 100-fold–concentrated nonessential amino acid solution, β-mercaptoethanol, and Roswell Park Memorial Institute 1640 medium (RPMI 1640) were obtained from Invitrogen (Karlsruhe, Germany). Fetal calf serum, sodium pyruvate, and penicillin/streptomycin were from Biochrom (Berlin, Germany). Magnetic cell sorting reagents were bought from Miltenyi Biotec (Bergisch Gladbach, Germany). The chemokines CXCL12, CXCL9, CCL21, CCL20, and CXCL16 were purchased from R&D Systems (Minneapolis, MN). Accutase was obtained from Innovative Cell Technologies (San Diego, CA). Acetylated low-density lipoprotein (AcLDL)–Bodipy and 6′-diamidino-2-phenylindol were from Molecular Probes. Nycodenz was bought from Axis Shields (Oslo, Norway). Fluoresbrite plain YG 20-μm microspheres were obtained from Polysciences (Warrington, PA).


Female C57BL/6 and CXCR3−/− mice were obtained from the Bundesamt für Risikobewertung (Berlin, Germany). CXCR3−/− mice16 were kindly supplied by Craig Gerard (Boston, MA). All animals received humane care according to the criteria published by the National Institutes of Health (Bethesda, MD).

Endothelioma Cell Lines.

Murine brain endothelioma cells (bEnd.5)17, 18 were cultured in complete Dulbecco's modified Eagle's medium (supplemented with 10% fetal calf serum, 1 mM sodium pyruvate, 100 U/mL penicillin, 100 μg/mL streptomycin, 2 mM L-glutamine, 1% nonessential amino acid solution, and 0.1% β-mercaptoethanol). Murine mesenteric lymph node–derived endothelioma cells (mlEnd),19, 20 kindly provided by Rupert Hallmann (Münster, Germany), were cultured in complete RPMI 1640 (supplemented with 10% fetal calf serum, 1 mM sodium pyruvate, 100 U/mL penicillin, 100 μg/mL streptomycin, and 0.1% β-mercaptoethanol).

Ex Vivo Isolation of LSEC and Lung Endothelial Cells.

Murine livers were perfused in situ with 2 mL of a prewarmed digestion medium (RPMI 1640 supplemented with 5% fetal calf serum, 2 mg/mL collagenase IV, and 0.2 mg/mL DNase I) injected into the portal vein. Organs were removed, cut into small pieces, and digested in the mixture (25 minutes, 37°C, and constant shaking at 250 rpm). Single cell suspensions passed through a sieve were subjected to a one-step density gradient centrifugation with 26% Nycodenz. Endothelial cells were further purified by magnetic cell sorting with the anti-mouse endothelium antibody ME-9F1, and this yielded 1 to 5 × 106 cells per liver in the positive fraction.21 After overnight culturing in complete RPMI 1640, nonadherent cells and debris were removed, and cells were used for experiments. To control the purity of LSEC, small aliquots were incubated overnight with 1 μg/mL AcLDL-Bodipy.22 According to AcLDL uptake, 95% to 98% of all cells were LSEC. Lung tissue was disintegrated in the digestion medium as outlined previously. Endothelial cells were further magnetically enriched from the resulting cell suspensions with the endothelial-specific monoclonal antibodies MECA32 and ME-9F1.

Isolation of Lymphocytes from the Spleen for the Transmigration Assay.

Splenic T cells were prepared from single spleen cell suspensions by Histopaque-1083 density gradient centrifugation. After incubation with anti–Mac-1, anti-Fcγ receptor II/III, and anti-CD8, CD4+ T cells were enriched by panning on rabbit anti-rat immunoglobulin G (cross-reactive to mouse) coated on plastic dishes.

To study transmigration of cytokine-expressing effector/memory T cells, naive CD62Lhigh cells were depleted by the addition of 1.5 μg/mL anti-CD62L to the panning mixture. In these experiments, 58% to 61% of all CD4+ T cells were CD45RBlow effector/memory cells.

Transmigration Assay.

One or two days prior to transmigration assays, cells from endothelioma cell lines and freshly isolated LSEC were cultured on fibronectin-coated, 24-well Transwell tissue culture inserts with a 5-μm pore size (Corning, New York, NY). Ex vivo isolated lung endothelial cells were initially expanded before being eventually transferred to the transwell membranes 1 or 2 days prior to the assay. Endothelial cells formed comparable layers. In phase contrast microscopy, approximately 80% of pores were covered. Immediately before the addition of the T cells, nonadherent and dead cells were removed from the cell monolayers via rinsing with phosphate-buffered saline. Fibronectin-coated endothelial cell–free transwells served as controls. Transmigration assays were performed as previously described.18, 23 Briefly, 5 × 105 T cells suspended in 100 μL of an assay medium (RPMI 1640 supplemented with 0.5% bovine serum albumin) were added to the upper chamber of the transwell. Soluble chemokines (CXCL9, 100 nM; CXCL12, 10 nM; CXCL16, 100 nM; CCL20, 100 nM; and CCL21, 10 nM) were applied to the lower chamber in 600 μL of the assay medium. After 90 minutes at 37°C, 500 μL of the cell suspension from the lower chamber or of the input cells was mixed with 10 μL of Fluoresbrite beads. Cells stained for CD4 and CD45RB were analyzed without washing by four-color flow cytometry with a FACSCanto II flow cytometer and FlowJo software (BD Biosciences). Absolute cell numbers were determined by gating on CD4+, CD4+CD45RBlow, or CD4+CD45RBhigh cells in relation to defined numbers of Fluoresbrite beads. Checkerboard analysis with 10 nM CXCL12 and 100 nM CXCL9 in the lower chamber, the upper chamber, or both chambers showed that transmigration in empty transwells is predominantly due to chemotaxis and less due to chemokinesis (Supplementary Fig. 1).

Gi-protein–coupled receptor signaling by the chemokine was blocked by preincubation of either T cells or LSEC for 2 hours at 37°C with 1 to 100 ng/mL pertussis toxin.24

To investigate chemokine uptake and immobilization on the endothelial cell surface, LSEC monolayers were preincubated for 2 hours at 37°C with 50 nM CXCL12 within the lower or upper chamber. Prior to the transmigration assay without additional soluble chemokine, LSEC were washed three times with phosphate-buffered saline, and the transwell membranes were placed into new cavities.

Transcytosis was blocked by the addition of 50 μM chlorpromazine25 during the transmigration assay. To exclude an effect of the inhibitor on the T cells, LSEC were alternatively preincubated for 2 hours at 37°C with 50 μM chlorpromazine or 50 μM monensin together with 50 nM CXCL12 from the lower chamber. In other experiments, 30 μg/mL nystatin,26 30 μg/mL filipin,27 10 μg/mL brefeldin A,28 and 10 μg/mL cytochalasin29 were used as inhibitors.

Detection of Cytokine-Producing Effector Cells by Intracytoplasmic Cytokine Staining.

For the detection of cytokine-expressing cells, cells pooled from 5 to 20 wells were restimulated with phorbol myristate acetate (10 ng/mL) and ionomycin (500 ng/mL) for 4 hours with the addition of brefeldin A (10 μg/mL) after 45 minutes. Cells were surface-stained with anti-CD4 and subsequently fixed with 2% paraformaldehyde in phosphate-buffered saline. After permeabilization using a buffer consisting of 0.5% saponin and 0.2% bovine serum albumin in phosphate-buffered saline, they were stained with anti-IFNγ, anti–interleukin 4 (IL-4), and anti–IL-10. Nonspecific binding was blocked with rat immunoglobulin and anti-Fcγ receptor II/III. Cells were analyzed by four-color flow cytometry and quantified with Fluoresbrite beads as outlined previously. Percentages of transmigrated cells of the respective cytokine phenotype were determined by gating on IFNγ+, IL-4+, or IL-10+ CD4+ T cells.23

Scanning Electron Microscopy and Confocal Laser Scanning Microscopy.

LSEC prepared as outlined previously were cultured on transwell membranes. For scanning electron microscopy, cell layers were fixed with 2.5% glutaraldehyde in a cacodylate buffer followed by 2% OsO4. Cells were dehydrated in alcohol and subsequently by a critical point drying technique using CO2 (Polaron E 3000, Gabler Instruments, Bad Schwalbach, Germany). Finally, samples were covered with a gold/palladium layer (MED 020, BAL-TEC, Liechtenstein) and analyzed with a DSM 982 Gemini scanning electron microscope (Zeiss, Oberkochen, Germany).

For confocal laser scanning microscopy, LSEC were fixed with 2% paraformaldehyde, permeabilized with 0.5% Triton X-100, and incubated with 6% goat serum and 1% bovine serum albumin to block unspecific binding. Subsequently, cells were stained for 60 minutes at room temperature with anti-CXCL12 (final concentration = 5 μg/mL) followed by anti-mouse Alexa 594. All incubation and washing steps were carried out in phosphate-buffered saline supplemented with 6% goat serum and 1% bovine serum albumin. Nuclei were stained with 1 μg/mL 6′-diamidino-2-phenylindol. Images were taken with an LSM 510 Meta confocal laser scanning microscope (Zeiss). The identity of the fluorescent label Alexa 594 within the specimen was verified with spectral signature scans.

Data Analysis.

Data are presented as mean values ± the standard deviation (SD). Mann-Whitney tests were performed with GraphPad Prism software (GraphPad Software, San Diego, CA).

Online Supplementary Material: Time-Lapse Video Microscopy.

LSEC, isolated ex vivo on the previous day (Supplementary Movie 1), or cells from the endothelioma cell line bEnd.5 (Supplementary Movie 2) were cultured overnight on fibronectin-coated μ-Slides VI (ibidi Integrated BioDiagnostics, Munich, Germany). Cells from murine spleens were sorted by magnetic cell sorting with anti-CD4, and this resulted in more than 99% CD4+ T cells. Endothelial cell monolayers were preincubated for 2 hours with 50 nM CXCL12, washed three times in complete RPMI 1640, and overlaid with 1 × 106 CD4+ T cells. After 20 minutes at 37°C, layers were flushed, and cells were monitored at 80-fold magnification by phase contrast microscopy with a Leica DM IRE2 microscope and Leica Confocal software (Leica Microsystems, Wetzlar, Germany). Transendothelial migration of CD4+ T cells was analyzed with time-lapse videos (one picture per 5 s) taken from three different areas (the original picture size was 117 × 117 μm2).


The Chemotactic Effect of CXCL12 and CXCL9 on Various CD4+ T Cell Subsets Is Significantly Enhanced by LSEC but Not by Other Endothelial Cells.

The CXCL12-triggered chemotactic response of CD4+ T cells was investigated in a transwell assay system. At low CXCL12 concentrations, no migration greater than spontaneous baseline mobility was observed. Chemotaxis continuously increased between 10 and 100 nM and eventually declined at 300 nM. In contrast, in the presence of LSEC, CXCL12 concentrations as low as 1 and 3 nM supported a significant transmigration of CD4+ T cells. At 10 and 30 nM CXCL12, LSEC further enhanced transmigration in comparison with endothelial cell–free transwell membranes (Fig. 1). In contrast to LSEC, the endothelioma cell lines bEnd.5 and mlEnd and ex vivo isolated lung endothelial cells strongly inhibited the chemokine-triggered transmigration (Fig. 2A). Phase contrast microscopy was used to visualize transendothelial migration of individual CD4+ T cells (Fig. 2B). Confirming the results from the transwell assay system, it showed that T cell transmigration was enhanced by LSEC (see the supplementary videos).

Figure 1.

LSEC enhance the chemotactic effect of CXCL12. Ex vivo isolated LSEC were cultured on transwell membranes. CD4+ T cells were placed in the upper chamber of the transwell assay system, and increasing concentrations of CXCL12 were added to the lower chamber. Cells were harvested after 90 minutes from the lower chamber, and aliquots of input cells were surface-stained for CD4. Fluoresbrite beads were added to determine the absolute cell numbers, and cells were analyzed by flow cytometry. Mean values from four independent experiments with triplicate determinations ± SD are shown. **P < 0.01. The dotted line shows spontaneous baseline transmigration in the absence of chemokine.

Figure 2.

Endothelioma cells and ex vivo isolated lung endothelial cells inhibit CXCL12-driven transmigration. (A) Cells from endothelioma lines (mlEnd and bEnd.5), ex vivo isolated lung endothelial cells, and LSEC were cultured on transwell membranes. Transmigration assays were performed as outlined for Fig. 1. Mean values from up to 32 independent experiments with triplicate determinations ± SD are shown. *P < 0.05, **P < 0.01. (B) CD4+ T cells were placed onto capillary slides covered with ex vivo isolated LSEC preincubated with CXCL12. After the slides were flushed, time-lapse videos were recorded. Representative pictures from three or four independent experiments with videos from three different areas are shown.

To investigate whether the chemotactic responses of distinct CD4+ T cell subsets are differentially modulated and whether the biological effects of other chemokines are also enhanced by LSEC, naive and memory CD4+ T cells, as identified by high or low CD45RB expression, respectively (Fig. 3), were studied. In line with the ubiquitous CXCL12 receptor expression, both cell types transmigrated through LSEC layers at higher rates in comparison with endothelial cell–free membranes. Selective expression of CXCR3 on antigen-experienced cells was reflected by higher rates of transmigrating memory, but not naive, CD4+ T cells in the presence of CXCL9 (Fig. 3).

Figure 3.

LSEC equally enhance transmigration of naive and memory CD4+ T cells in response to CXCL12 and CXCL9. Transwell assays were performed after the addition of CXCL9 or CXCL12, and T cells were analyzed as outlined for Fig. 1. Naive and memory T cells were distinguished by flow cytometry according to their CD45RB expression. Mean values ± SD from eight independent experiments with triplicate determinations are shown. *P < 0.05, **P < 0.01.

To address the effect of LSEC on the transmigration of Th1, Th2, and IL-10–producing cells, memory CD4+ T cells were enriched ex vivo, and transmigrating IFNγ+, IL-4+, or IL-10+ cells were identified by intracytoplasmic cytokine staining. Regardless of their effector cytokine phenotype, all effector cell subsets responded to CXCL12 and CXCL9 with significantly enhanced transmigration through LSEC monolayers in comparison with endothelial cell–free membranes, with IFNγ+ cells displaying a slightly higher response to CXCL9 than IL-4+ and IL-10+ cells (Fig. 4). In summary, LSEC equally amplified the biological effects of CXCL12 and CXCL9 on the transmigration of all CD4+ T cell populations studied. Transmigration of CD4+ T cells in the presence of CXCL16, CCL20, or CCL21 was not increased by LSEC (Fig. 5).

Figure 4.

CXCL12-dependent and CXCL9-dependent transmigration of Th1, Th2, and IL-10–producing cells is supported by LSEC. Transwell assays were performed, and absolute cell counts or frequencies of CD4+ T cells were determined as outlined for Fig. 1. The cytokine phenotype of input and transmigrated CD4+ T cells was determined by flow cytometry after intracytoplasmic cytokine staining. Mean values ± SD from three to five independent experiments with triplicate determinations are shown. *P < 0.05, **P < 0.01.

Figure 5.

Transmigration of CD4+ T cells in the presence of CXCL16, CCL20, and CCL21 is not enhanced by LSEC. CD4+ T cells were placed in the upper chamber of the transwell assay system, and 100 nM CXCL16, 100 nM CCL20, or 10 nM CCL21 was added to the lower chamber. Transmigration assays were carried out as described for Fig. 1. Mean values ± SD from two to three independent experiments with triplicate determinations are shown; the dotted line shows spontaneous baseline transmigration in the absence of chemokines.

Inhibition of Gi-Protein–Coupled Signaling and Chemokine Receptor Deficiency in LSEC Does Not Reduce Chemokine-Triggered CD4+ T Cell Transmigration.

The enhanced biological effects of low CXCL12 concentrations suggested active mechanisms mediated by, for example, chemokine receptors. First, Gi-protein–coupled signaling was inhibited by pertussis toxin. As expected, preincubation of T cells with increasing toxin concentrations reduced transmigration in response to CXCL12 (Fig. 6A). In contrast, treatment of LSEC even with high pertussis toxin concentrations (100 ng/mL) did not affect the rates of transmigrating CD4+ T cells (Fig. 6B).

Figure 6.

Pertussis toxin treatment of LSEC does not reduce transmigration. (A) CD4+ T cells were pretreated with increasing concentrations of pertussis toxin and washed before being used in the transmigration assay with endothelial cell–free membranes. (B) Transwell inserts coated with LSEC monolayers were pretreated with 100 ng/mL pertussis toxin and were thoroughly washed immediately before the transmigration assay was carried out as described for Fig. 1. Mean values ± SD from two to four independent experiments with double and triplicate determinations are shown. **P < 0.01.

In addition, cells from CXCR3−/− mice were compared to wild-type cells. Although the transmigration of memory T cells from wild-type mice was increased in the presence of wild-type LSEC in response to CXCL9, the migratory response of CXCR3−/− memory CD4+ T cells was reduced to baseline levels. However, the outcome was independent of the CXCR3 expression on LSEC (Table 1), demonstrating that CXCR3 is essential on T cells but not on LSEC.

Table 1. CXCR3 Expression on LSEC Is Not Essential for the Transmigration of Memory CD4+ T Cells
Responding CD4+ T CellsNo Chemokine (% of Input)CXCL9 (% of input)
  1. LSEC from wild-type and CXCR3−/− mice, cultured on transwell membranes, were used in transmigration assays performed as outlined for Fig. 1. During the assay, soluble CXCL9 (100 nM) was added to the lower chamber. Mean values ± SD are presented from three to five independent experiments with triplicate determinations.

Wild-type8.8 ± 5.43.7 ± 1.026.3 ± 8.831.3 ± 3.6
CXCR3−/−3.2 ± 0.44.1 ± 0.73.1 ± 0.83.9 ± 1.2

CD4+ T Cell Transmigration Increases After Preincubation of LSEC Layers with CXCL12 and CXCL9 from the Abluminal Side and Can Be Inhibited by Blocking Transcytosis.

We investigated the possibility that LSEC layers actively take up and present chemokines clustered on their surface, thereby enhancing their biological activity. When LSEC-covered membranes were preincubated with CXCL12 in the lower chamber of the transwell, the numbers of transmigrating CD4+ T cells were comparable to those in the presence of LSEC and soluble CXCL12. In contrast, when endothelial cell–free membranes were preincubated with CXCL12 in the lower chamber, CD4+ T cell migration remained at the level of LSEC without any additional chemokine, excluding a significant effect of plate-bound chemokine. Interestingly, transmigration also did not differ from the baseline level when LSEC monolayers were preincubated with CXCL12 in the upper chamber (Fig. 7A). These findings suggest that LSEC might actively sample CXCL12 from the lower side. LSEC form a tight layer on transwell membranes and retain fenestrae grouped to sieve plates, as visualized by scanning electron microscopy (Fig. 7B). Indeed, preincubation of LSEC with CXCL12 in the lower chamber led to an accumulation of CXCL12 in cytosolic vesicles (Fig. 7C), and this implied transcytosis for chemokine presentation on the endothelial cell surface. To address the contribution of uptake and intracellular transport of CXCL12, transcytosis was blocked by chlorpromazine, which effectively inhibited the uptake of fluorochrome-labeled AcLDL by LSEC (Fig. 8A). Added during the assay, chlorpromazine abolished CD4+ T cell transmigration. To exclude direct effects of the inhibitor on the T cells, LSEC were preincubated with chlorpromazine or monensin together with CXCL12. After thorough washing of the LSEC layers, the biological activity of CXCL12 was significantly reduced (Fig. 8B). Spontaneous transmigration was not altered, and this argued against an effect of chlorpromazine and monensin on, for example, junctional retraction (Supplementary Fig. 2). In contrast, preincubation with nystatin, filipin, brefeldin A, or cytochalasin B did not influence the CXCL12-mediated transmigration (Table 2). Preincubation of LSEC with CXCL9 also resulted in increased transmigration in comparison with untreated LSEC. As observed for CXCL12, LSEC might also take up and present CXCL9 because transmigration was inhibited by the addition of monensin (data not shown) and chlorpromazine (Fig. 8C) in the preincubation phase.

Figure 7.

Preincubation of LSEC with CXCL12 from the abluminal side results in T cell transmigration and uptake of CXCL12 by LSEC. (A) Transmigration assays were performed in the transwell assay system as outlined for Fig. 1. LSEC without chemokine served as a negative control, and the addition of soluble CXCL12 to the lower chamber served as a positive control. LSEC monolayers were preincubated with CXCL12 within the lower or upper chamber. After preincubation, LSEC were washed, and the transwell membrane was placed into new cavities before the addition of T cells. Mean values ± SD from four to seven independent experiments with triplicate determinations are shown. **P < 0.01. The dotted line shows spontaneous baseline transmigration in the absence of chemokine (negative control). (B) Images of LSEC cultured on transwell membranes were taken by scanning electron microscopy. The arrow indicates fenestrae grouped to sieve plates. (C) Confocal laser scanning microscopy: untreated LSEC or LSEC preincubated with CXCL12 in the lower chamber were stained for CXCL12 (red). Nuclei labeled with 6′-diamidino-2-phenylindol appear blue; ME-9F1 labeled with fluorescein isothiocyanate appears green. The insert shows the cellular distribution of CXCL12. Representative images are shown from up to four independent experiments.

Figure 8.

Blockage of transcytosis by chlorpromazine and monensin inhibits the uptake of AcLDL and increased transmigration. (A) LSEC were incubated with 5 μg/mL AcLDL alone (bold line) or together with 50 μM chlorpromazine (gray plot). Unstained cells served as a negative control (thin line). Cells were harvested after 4 hours and analyzed by flow cytometry. A representative plot from two independent experiments with triplicate determinations is shown. (B) Transmigration assays were performed in the transwell assay system as outlined for Fig. 1. Soluble CXCL12 was added to the lower chamber alone or together with 50 μM chlorpromazine. Alternatively, LSEC were preincubated with CXCL12 alone or in the presence of 50 μM chlorpromazine or 50 μM monensin. Mean values ± SD from four to seven independent experiments with triplicate determinations are shown. *P < 0.05, **P < 0.01. The dotted line shows spontaneous baseline transmigration in the absence of chemokine (negative control). (C) LSEC were preincubated with CXCL9 alone or in the presence of 50 μM chlorpromazine. Mean values ± SD from four independent experiments with triplicate determinations are shown. *P < 0.05, **P < 0.01.

Table 2. Preincubation of LSEC with Inhibitors for Caveolae, the Golgi Complex, or Contractile Microfilaments Does Not Affect CXCL12-Mediated Transmigration of CD4+ T Cells
Treatment% of CXCL12-Mediated Transmigration
  1. LSEC, cultured on transwell membranes preincubated with CXCL12 (50 nM) and the indicated inhibitors, were used in transmigration assays performed as outlined for Fig. 1. Data are expressed as the percentage of transmigration with respect to LSEC preincubated with CXCL12 without an inhibitor. Mean values ± SD are presented from three to four independent experiments with triplicate determinations.

Nystatin94.3 ± 12.8
Filipin107.2 ± 19.9
Brefeldin A99.0 ± 25.7
Cytochalasin B93.3 ± 26.9


Actively supporting T cell transmigration seems to be a unique feature of the liver sinusoidal endothelium. We here report that the biological activity of CXCL12 and CXCL9 can be amplified by LSEC monolayers, which result in significantly increased transendothelial migration of CD4+ T cells in comparison with empty transwells even at very low concentrations of the chemokine. Under comparable experimental conditions, endothelioma cell lines derived from the brain (bEnd.5) and mesenteric lymph nodes (mlEnd), which retained typical features of primary endothelial cells,18 inhibited the chemokine-driven transendothelial migration. Ex vivo isolated endothelial cells from the kidneys (data not shown) and lungs also decreased the CXCL12-dependent transmigration of CD4+ T cells in our system. These findings are in line with previous observations that human umbilical vein cells reduce the CXCL12-triggered transendothelial migration of naive lymphocytes.30 In summary, our findings suggest that various endothelial cell populations might differentially modulate the biological effects of chemokines. Endothelial cells from different tissues are highly diverse in their expression of adhesion molecules, intercellular junctions, or surface glycosaminoglycans. Because control endothelial cells also inhibited the spontaneous transmigration of CD4+ T cells (Supplementary Fig. 3), mechanisms independent of the chemokine might be involved.31–33

LSEC enhanced the CXCL12-driven transmigration of all CD4+ T cell subsets investigated and the CXCL9-dependent transmigration of cytokine-producing memory cells, whereas the chemotactic effects of CXCL16, CCL20, and CCL21 on CD4+ T cells were not significantly altered under the given experimental conditions. Thus, LSEC seem to exert differential effects on certain chemokines. Furthermore, our findings do not allow the conclusion that LSEC-enhanced chemotaxis generally establishes an additional filter for certain pro-inflammatory or anti-inflammatory T cell subsets of a distinct cytokine phenotype. The transmigrating subset itself was still determined by the chemokine with an additional quantitative effect of the endothelial cell.

What molecular mechanisms might be responsible for the enhanced CXCL12-driven and CXCL9-driven CD4+ T cell transmigration through the liver endothelium? CXC chemokine receptors are expressed on the surface of certain endothelial cell types.34 Therefore, we first asked whether chemokine-driven transmigration is enhanced by chemokine receptor–mediated activation of LSEC. As previously shown, in the brain endothelium and to a lesser extent in the aortic or high vascular endothelium, transendothelial migration is completely dependent on Gi-protein–coupled signaling in endothelial cells.24 Pretreatment of LSEC with pertussis toxin had no effect on the CXCL12-driven transmigration of CD4+ T cells. With cells from CXCR3−/− mice, direct signaling via CXCR3 on LSEC was shown not to be crucial for CXCL9-mediated enhanced transmigration. Both observations argue against a contribution of receptor-mediated mechanisms that require Gi-protein signaling on the side of endothelial cells derived from the liver sinusoids, and this is in line with previous findings in which pretreatment of activated human umbilical vein endothelial cells with anti-CXCR4 or pertussis toxin did not inhibit lymphocyte transmigration in response to apical CXCL12.35

With the chemokine applied only to the upper side of the LSEC layers, T cell transmigration was not enhanced, and this is compatible with the proposal that a soluble chemokine gradient is mandatory underneath the endothelial cell layer to effectively promote transendothelial migration (chemotaxis).30 IL-8 taken up from the basal side of endothelial cells, transcytosed, and eventually presented on the apical side displayed strongly enhanced biological activity.36 Under shear forces, apical endothelial cell–bound chemokines were found to be crucial for transendothelial migration across an inflamed endothelium.35, 37 These findings support an alternative haptotaxis model in which leukocyte migration is facilitated by surface-presented chemokines in close vicinity to adhesion molecules.38, 39 Our observations of LSEC pretreated with CXCL9 and CXCL12 only from the lower side of the cell layer support this explanation. LSEC retain fenestrae, which allow the diffusion of soluble CXCL12 to the luminal side. However, our data instead point to transcytosis of the chemokine: LSEC take up and store abluminal CXCL12, and preincubation from the luminal side did not lead to enhanced transmigration. Therefore, blockage of transcellular pathways should inhibit chemokine-driven transmigration. Indeed, the use of chlorpromazine and monensin, which cause misassembly of clathrin lattices on endosomes and prevent coated pit formation at the cell surface,25 significantly reduced the chemotactic effect of CXCL9 and CXCL12 preincubated with LSEC. Transcytosis of these two chemokines seems to be mediated by clathrin-coated pits because transmigration could be blocked by monensin and chlorpromazine but not by nystatin and filipin acting on caveolae,26, 27 brefeldin A interfering with the Golgi complex,28 or cytochalasin A inhibiting contractile microfilaments.29

Taken together, our findings underline the importance of LSEC forming a specialized endothelium, serving as gatekeepers for leukocyte access to the tissue by regulating the chemokine-driven transendothelial migration. They are in line with a previous investigation demonstrating transendothelial migration of lymphocytes promoted by CXC3 ligands bound on the surface of activated human hepatic endothelium.40 In our model system, activated LSEC also enhance chemokine-driven transmigration, albeit not at higher rates than resting LSEC (data not shown). Preliminary observations suggest that LSEC activation might be crucial for antigen-dependent adhesion and transmigration, presumably because of increased major histocompatibility complex II expression (K.N., unpublished data, 2008). Translating to the situation in vivo in which within the normal liver hepatic stellate cells41 and hepatocytes,42 for example, release chemokines into the space of Dissé, we hypothesize that these chemokines can be taken up by LSEC from the basal side and are then presented on the luminal side, supporting CD4+ T cell trafficking. Transmigrating T cells might rapidly remigrate, receive suppressive signals upon transmigration,43, 44 or subsequently interact with tolerogenic Ito cells.45 In this respect, endothelial cells from the liver sinusoids seem to possess a special feature in facilitating T cell balance and immune surveillance under homeostatic conditions and presumably even more strongly under inflammatory conditions. It is tempting to speculate that inhibition of chemokine presentation by LSEC might provide a future therapeutic approach to suppressing lymphocyte immigration to treat hepatic inflammation.


We thank Craig Gerard (Boston, MA) for providing the CXCR3−/− mice and Gudrun Debes (Stanford, CA) for helping to establish the transmigration assay. We are grateful to Claudia May for performing the CXCL12 staining and to Petra Schrade and the Electron Microscopy Core Facility of the Charité (Prof. Dr. S. Bachmann) for the scanning electron microscopy.