CX3CL1 (fractalkine): A signpost for biliary inflammation in primary biliary cirrhosis


  • Potential conflict of interest: Nothing to report.


Improvements in the treatment of primary biliary cirrhosis (PBC) may depend upon dissection of mechanisms that determine recruitment of mononuclear cells to intralobular bile ducts, including the role of the chemokine-adhesion molecule CX3CL1 (fractalkine). We submit that there are unique interactions between intrahepatic biliary epithelial cells (BECs), endothelial cells (ECs), liver sinusoidal endothelial cells (LSECs), and liver-infiltrating mononuclear cells (LMCs), and that such interactions will in part dictate the biliary-specific inflammatory response. To address this, we studied fresh explanted livers from pretransplantation patients with PBC and with inflammatory liver disease due to viral infection (disease controls) and biopsy material from patients with a discrete liver tumor (normal controls). Using this clinical material, we isolated and stimulated BECs, ECs, LSECs, and LMCs with a panel of Toll-like receptor ligands. We also studied the interactions of these cell populations with LMCs with respect to adhesion capability and production of tumor necrosis factor α (TNF-α). Finally, we used fresh biopsy samples to evaluate mononuclear cells around intrahepatic biliary ductules using monoclonal antibodies specific to CD68 or CD154, markers for monocytes/macrophages, and activated T cells, respectively. Conclusion: There are common properties of ECs, LSECs, and BECs, whether derived from PBC or viral hepatitis, but there are also significant differences, particularly in the potential in PBC for LMCs to adhere to ECs and BECs and to produce TNF-α; such properties were associated with augmented CX3CL1 production by BEC from PBC liver. The processes defined herein suggest potential novel biotherapies for biliary specific inflammation. (HEPATOLOGY 2009.)

Previous work in our laboratory on inflammatory chemotaxis in primary biliary cirrhosis (PBC) has demonstrated participation of various chemokine ligands,1 and studies in other laboratories have implicated the chemokine CX3CL1 (fractalkine) in inflammatory liver disease.2, 3 We therefore set out to examine the contribution of CX3CL14–9 to the bile duct destruction of PBC. Our previous findings in PBC indicated that CX3CL1 is elevated in serum concurrent with increased expression of the CX3CR1 receptor in liver-infiltrating mononuclear cells (LMCs),10 leading us to posit that CX3CL1 could indeed be critical for the generation and persistence of the portal lymphocytic infiltration in PBC. We have herein taken advantage of our ability to isolate pure populations of multiple intrahepatic cell types, including endothelial cells (ECs), liver sinusoidal endothelial cells (LSECs), and biliary epithelial cells (BECs) to directly study the interaction of CX3CL1-producing cells with LMCs. We should note that several nonprofessional immunocompetent cells produce chemokines in response to ligands for Toll-like receptors (TLRs).11, 12 Here we have evaluated CX3CL1 production from ECs, LSECs, and BECs exposed to TLR ligands and report that ECs produced high amounts of CX3CL1 using one or another of several TLR ligands, whereas LSECs never produced CX3CL1 with any ligand; BECs produced CX3CL1 on exposure to autologous LMCs, tumor necrosis factor α (TNF-α), and a TLR3 ligand. In the process of simplifying the production system of CX3CL1 from BECs, we found that TLR3-stimulated BECs produced CX3CL1 after direct contact with TLR4-stimulated autologous monocytes. In conclusion, our data indicate that CX3CL1 and TNF-α, which are induced by TLR ligands, participate in processes that lead to disease-specific recruitment of lymphoid cells into the portal tracts of the liver and thereby to the characteristic chronic nonsuppurative destructive cholangitis of PBC. This new knowledge of the mechanisms of lymphocyte homing and recruitment induced by innate immunity and, potentially, the ability to inhibit abnormal chemoattractant homing is a fertile area for future therapeutic intervention in PBC.


BEC, biliary epithelial cell; EC, endothelial cell; IFN-γ, interferon-γ; LMC, liver-infiltrating mononuclear cell; LPS, lipopolysaccharide; LSEC, liver sinusoidal endothelial cell; LTA, lipoteichoic acid; mDC, myeloid dendritic cell; NK, natural killer; NKT, natural killer T; PBC, primary biliary cirrhosis; poly(I:C), polyinosinic:polycytidylic acid; TLR, Toll-like receptor; TNF-α, tumor necrosis factor α.

Materials and Methods


Twenty-one explanted livers were studied, derived from nine patients with PBC—three with hepatitis B virus infection, seven with hepatitis C virus infection, and two with primary sclerosing cholangitis. All patients had end-stage liver cirrhosis without signs of other acute liver injury from an unrelated cause. The diagnosis of PBC was based on established criteria13 and all such patients had a positive test for serum antimitochondrial antibodies.13 In addition, and for purposes of comparison in further nested substudies, we studied livers from an additional five noncirrhotic PBC patients for analysis of LMCs, as well as normal control livers from four patients undergoing liver resection for isolated metastatic tumors, for concurrent analysis of intrahepatic BECs, ECs, and LSECs. For studies on immunohistochemistry, fresh samples from wedge biopsies were available from 12 patients with PBC, 15 patients with hepatitis C infection, and seven patients with a hepatic neoplasm but normal surrounding liver. All such samples were studied after informed consent of the donor, and all experimental protocols were approved by the Research Ethics Committee of Kyushu University and the University of California Davis.

Isolation of Intrahepatic BECs, ECs, LSECs, and LMCs.

The isolation of nearly pure cell subpopulations from livers was achieved using methods described previously.14 Liver specimens were first digested with 1 mg/mL of collagenase type I (Sigma-Aldrich, Tokyo, Japan). Cells from the digested tissue were gradient-separated to obtain LMCs15 that were cultured overnight, the adherent cell population was maintained in culture until there was full confluence, usually by day 14, and the nonadherent cell populations were stored in liquid nitrogen. Further, in a nested study, fresh LMCs from noncirrhotic PBC liver were obtained from liver biopsies (needle, n = 3; surgical, n = 2) that were cut into smaller fragments and digested with collagenase type I for 20 minutes. Dissociated cells were filtered through a 150-μm mesh and separated by way of Ficoll centrifugation, and were then immediately used for study of TNF-α production by LMCs, as described below.

BECs were separated from adherent cells using CD326 (EpCAM) MicroBeads specific for epithelial cells as described.14 The cell phenotype was verified by immunohistochemistry with antibodies against cytokeratins 7 and 19 (Dako, Glostrup, Denmark); a cell purity exceeding 90% was deemed acceptable. The viability of all cells for each of the experiments of greater than 95% was established by way of trypan blue exclusion.

ECs were separated from adherent cells using CD31 microbeads specific for ECs. Because LSECs do not express CD31,16 but are positive for CD105,17 they could be separated from both BECs and ECs after separation of adherent cells using CD105 microbeads. LSECs were isolated using a density gradient.16–18 We confirmed that LSECs thus isolated were CD31-negative and CD105-positive. ECs and LSECs were cultured with endothelial-specific medium (HuMedia-EG2).14 For the two cases of primary sclerosing cholangitis, the limited size of the liver specimens provided precluded isolation of ECs and LSECs and thus limited the data available. For each cell population, the yield of cells differed between samples; however, all tissues were handled identically, and the total number of cells used in each assay was standardized. Cells were studied in early cultures, at passages 4-6, to obviate the potential loss of phenotype after prolonged in vitro culture.

CX3CL1 Production.

To study CX3CL1 production, cells (1 × 105/400 μL in 24-well plates) were initially cultured for 48 hours in the presence of one or another of multiple TLR ligands; these included lipoteichoic acid (LTA), polyinosinic:polycytidylic acid [poly(I:C)], purified lipopolysaccharide (LPS), flagellin, CL-097, ODN2216, and ODN2006, all purchased from Invitrogen (San Diego, CA). The optimal concentrations were 2-10 μg/mL. Then BECs (1 × 105/200 μL in 24-well plates) and autologous LMCs (2 × 106/200 μL in 24-well plates) were cultured for 24 hours to study CX3CL1 production in the presence of one or another of various ligands for TLR, and either interferon-γ (IFN-γ) or TNF-α, at final concentrations of 200 U/mL or 0.1 μg/mL, respectively. The supernatants from the cultured media with different TLR stimuli were analyzed for CX3CL1 production by sandwich enzyme-linked immunosorbent assay kits (R&D Systems, Minneapolis, MN), using a combination of unlabeled and biotin- or enzyme-coupled monoclonal antibody to CX3CL1, for which the lower level of detectability was 200 pg/mL.

In what we designate as transwell experiments, BECs were grown in the bottoms of wells, and LMCs were added to the 24-well plate filter inserts with a pore size of 0.4 μmol/L (BD Biosciences, Bedford, MA). Blockage of cellular interactions was achieved with various antibodies, including anti-CD154 (Ancell, Bayport, MN); anti-HLA class I; and anti-HLA DP, DQ, and DR (BD Biosciences, Bedford, MA). Each antibody was used at a predetermined optimal concentration of 5-40 μg/mL. Briefly, anti-HLA class I and anti-HLA DP, DQ, and DR for BECs and anti-CD154 for LMCs were preincubated overnight, and cells were washed twice; BECs and LMCs were cocultured in the presence of poly(I:C) and TNF-α. In selected nested experiments, T cells, monocytes, natural killer T (NKT) cells, natural killer (NK) cells, or myeloid dendritic cells (mDCs) were separated as described and pretreated with LPS (final concentration 10 μg/mL) for 24 hours, and washed three times; poly(I:C)-pretreated BECs and 2 × 106/200 μL T cells, monocytes, NKT cells, NK cells, or mDCs were cocultured in each well of a 24-well plate. In other selected nested experiments, BECs were pretreated with poly(I:C) or poly(I:C) and LPS (final concentration 10 μg/mL) for 24 hours, washed well, and then monocytes were added to the BEC preparation. Anti–TNF-α or an irrelevant control antibody (final concentration 10 μg/mL) was used to confirm the role of TNF-α in the production of CX3CL1 from BECs.

Cell Adhesion Assay.

Assays were performed for adhesion between ECs, BECs or LSECs, and LMCs. Confluent monolayers of ECs, BECs, or LSECs were cultured in 24-well plates (1 × 105cells/well) and then overlaid with LMCs (2 × 106/well) in the presence of poly(I:C) (final concentration 10 μg/mL) and TNF-α (final concentration 0.1 μg/mL) for 24 hours as described above. Nonadherent cells were removed by gentle rinsing and wells were washed four times. Adherent LMCs were fixed and stained with Diff-Quick Stain, and counted as the number of adherent cells in 10 random high-power fields as described.19

Isolation of T Cells, Monocytes, NK Cells, mDCs, and NKT Cells from LMCs.

T cells among LMCs were separated using a Pan T cell isolation kit II. Non–T cells (B cells, NK cells, DCs, monocytes, granulocytes, and erythroid cells) were indirectly magnetically labeled using a cocktail of biotin-conjugated antibodies against CD14, CD16, CD19, CD36, CD56, CD123, glycophorin A, and anti-biotin microbeads. Isolation of purified T cells was achieved by depletion of magnetically labeled cells by separation over a MACS column, which was placed in the magnetic field of a MACS Separator; a purity of CD3+ T cells of >90% was confirmed by flow cytometry. Monocytes were separated with a monocyte isolation kit. Non-monocytes were indirectly magnetically labeled with a cocktail of biotin-conjugated monoclonal antibodies against CD3, CD7, CD16, CD19, CD56, CD123, and glycophorin A, and anti-biotin microbeads. Isolation of monocytes was achieved by depletion of magnetically labeled cells; a purity of CD14+ monocytes of >90% was confirmed by way of flow cytometry. NK cells were separated with an NK isolation kit. Non-NK cells were indirectly magnetically labeled with a cocktail of biotin-conjugated antibodies against lineage-specific antigens and anti-biotin microbeads. Isolation of NK cells was achieved through the depletion of magnetically labeled cells; a purity of CD56+ NK cells of >90% was confirmed by way of flow cytometry. mDCs (CD1c+) were separated with an mDC isolation kit performed by two magnetic separation steps. In the first step, CD1c-expressing B cells were magnetically labeled with CD19 microbeads and subsequently depleted magnetically. In the second step, CD1c+ mDCs in the B cell–depleted flow-through fraction were indirectly magnetically labeled with CD1c-biotin and anti-biotin microbeads. Upon separation, the labeled CD1c+ mDCs were retained within the column and eluted after removing the column from the magnetic field. A purity of CD1c+ CD19 mDCs of >80% was confirmed by way of flow cytometry. NKT cells were separated with an NKT isolation kit. The isolation of NKT cells was performed in two magnetic separation steps. In the first step, NK cells and monocytes were indirectly magnetically labeled using a cocktail of biotin-conjugated antibodies and anti-biotin microbeads. The labeled cells were subsequently depleted by separation over a MACS Column. In the second step, CD3+CD56+ NKT cells were directly labeled with CD56 microbeads and isolated by positive selection from the pre-enriched NKT cell fraction. Upon separation, the labeled CD56+ cells were retained within the column and eluted after removing the column from the magnetic field. A purity of CD3+ CD56+ NKT cells of >80% was confirmed by flow cytometry.

TNF-α Production.

Cell populations (2 × 104/200 μL in 96-well plates) were cultured for 48 hours in the presence of the TLR ligands described above at 10 μg/mL. Supernatants were analyzed for TNF-α production by sandwich enzyme-linked immunosorbent assay kits (R&D Systems), using a combination of unlabeled and biotin- or enzyme-coupled monoclonal antibody to TNF-α. In all instances, known positive and known negative controls were used throughout, and all assays were performed in triplicate.

Immunohistochemical Stainings of Liver Specimens.

Fresh liver specimens from 12 patients with PBC, 15 patients with hepatitis C infection, and seven patients with discrete intrahepatic tumors (unaffected non–tumor-bearing liver) were fixed in 10% neutral-buffer formalin and snap-frozen in OCT compound (Miles, Inc., Elkhart, IN). Deparaffinized and rehydrated sections, and frozen sections, were used for immunostaining for cell surface markers, CD68 (expressed particularly on monocytes/macrophages) and CD154 (expressed particularly on activated T cells). Endogenous peroxidase was blocked in normal goat serum diluted 1:10 (Vector Laboratories, Burlingame, CA) for 20 minutes; CD68 and CD154 were diluted 1:100 (Dako), and immunostaining was performed on coded sections and interpreted by a blinded qualified liver pathologist.

Statistical Analysis.

All experiments were performed in triplicate, and data points shown are means of results of these triplicates. Comparisons between the points for data items are expressed as the mean ± standard deviation, and the significance of differences was determined using the Student t test. All analyses were two-tailed, and P < 0.05 was considered significant. Statistical analyses were performed using Intercooled Stata 8.0 (Stata Corp, College Station, TX).


Production of CX3CL1 by Populations of Liver Cells.

We assessed the production of CX3CL1 by isolated populations of liver cells in PBC and control patients after stimulation by different TLR ligands. With ECs, production was induced by LTA, poly(I:C), LPS, and flagellin, but not by CL-097, ODN2216, or ODN2006. Levels of CX3CL1 in PBC versus non-PBC disease controls were as follows: LTA, 1.7 ± 0.9 versus 1.6 ± 0.9 ng/mL (P value not significant); poly(I:C), 7.8 ± 1.0 versus 7.9 ± 1.7 ng/mL (P value not significant); LPS, 4.9 ± 0.9 versus 5.1 ± 1.0 ng/mL (P value not significant); and flagellin, 0.5 ± 0.2 versus 0.6 ± 0.2 ng/mL (Fig. 1A). Levels of CX3CL1 in normal liver controls were as follows: LTA, 1.8 ± 0.6 ng/mL; poly(I:C), 8.0 ± 1.5 ng/mL; LPS, 4.9 ± 1.8 ng/mL; and flagellin, 0.6 ± 0.4 ng/mL (Fig. 1A); these differences were not significant. Although activated LSECs mediate CX3CL1 shedding and release of chemotactic peptides,20 neither LSECs nor BECs produced CX3CL1 after stimulation with any of the TLR ligands used (data not shown) in PBC, non-PBC disease controls, and normal liver controls.

Figure 1.

(A) CX3CL1 is produced by ECs from PBC, normal livers (n = 4), and disease control livers (three hepatitis B virus, seven hepatitis C virus) exposed to TLR ligands. Differences between PBC and controls were not significant. LSEC did not produce CX3CL1 with any TLR ligand (data not shown). (B) CX3CL1 is produced by BECs exposed to TLR ligands in the presence of autologous LMCs and TNF-α. BEC production of CX3CL1 required the presence of LMCs and TNF-α, and was elicited only with poly(I:C). Differences between PBC and disease control livers were not significant. BECs did not produce CX3CL1 when exposed only to LMCs or LMCs with IFN-γ (data not shown). Open circle: normal. Black diamond: PBC. Dashed diamond: hepatitis B virus. Open diamond: hepatitis C virus. Hatched diamond: primary sclerosing cholangitis.

Because previous reports demonstrated that BECs produce chemokines in coculture with autologous LMCs,1 and because TNF-α and IFN-γ enhance CX3CL1 production from mucosal ECs,21 we used an LMC and BEC coculture system with or without the addition of TNF-α or IFN-γ. No production of CX3CL1 by BECs with LMCs was induced with any TLR ligands (data not shown). However, BECs in the presence of LMCs, with TNF-α but not with IFN-γ, together with poly(I:C) produced CX3CL1 (3.7 ± 0.1 versus 3.5 ± 1.2 ng/mL); again, the difference between PBC and controls was not significant (Fig. 1B). Thus the presence of TNF-α is critical for CX3CL1 production by BECs. The possibility that lymphocytes produced CX3CL122 was excluded by irradiation of LMCs, which did not significantly alter the results (data not shown). Also, LMCs without BECs never produced CX3CL1 with any TLR ligands, even after addition of IFN-γ or TNF-α. In the case of nondiseased controls, we were unable to study CX3CL1 production from BECs with LMCs and TNF-α, because sufficient LMCs were not available.

Production of CX3CL1 by BECs Requires Direct Contact with LMCs Through CD40–CD154 Interaction.

BECs did not produce CX3CL1 on coculture with poly(I:C)-pretreated LMCs in the presence of TNF-α, illustrated by representative data for one PBC liver (Fig. 2A), and indicating that BECs but not LMCs require poly(I:C) stimulation for production of CX3CL1. Such production decreased markedly when the BEC and LMC populations were separated by a filter in a transwell system (Fig. 2B). We assessed the functional effects of CD40, HLA class I, and HLA class II molecules on BECs by testing the capacity of blocking antibodies to CD154 and HLA molecules to suppress production of CX3CL1 by BECs. Production of CX3CL1 by BECs was significantly decreased when CD40 on BECs was blocked from interacting with CD154 on LMCs (Fig. 2C). Having shown that LMCs and TNF-α are critically required for production of CX3CL1 by BECs, we next examined in detail the role of LMCs and TNF-α production.

Figure 2.

CX3CL1 is produced by BECs in the presence of TNF-α and poly(I:C) with autologous LMCs. Representative data are shown from the liver of one PBC patient. (A) Poly I:C pretreatment of BECs plus autologous LMCs in the presence of TNF-α enabled production by BECs of CX3CL1 with LMCs, whereas poly(I:C) treatment of LMCs did not. Thus, BECs but not LMCs are required for poly(I:C). (B) CX3CL1 production by BECs in the presence of TNF-α and poly(I:C) required direct cell contact with autologous LMCs, because CX3CL1 production under transwell conditions with autologous LMCs was minimal. (C) CX3CL1 production by BECs in the presence of TNF-α and poly(I:C) with autologous LMCs was significantly inhibited by a monoclonal antibody specific to CD154 (a-154) but not by monoclonal antibodies specific to major histocompatibility complex class I or II (a-MHC I, II). Thus, interaction between CD40 from BECs and CD154 from LMCs facilitates CX3CL1 production. *P < 0.05.

Assay for Adherence of LMCs to ECs or LSECs.

LMCs in the presence of poly(I:C) and TNF-α adhered to ECs and BECs and, notably, the number of such adherent LMCs from PBC livers exceeded that for control cases (394 ± 94 versus 116 ± 45 cells [P < 0.01] for ECs; 180 ± 63 versus 65 ± 40 cells [P < 0.01] for BECs). However, only very few LMCs adhered to LSECs, whether from PBC livers (21 ± 14) or controls (20 ± 15) (P > 0.05) (Fig. 3).

Figure 3.

Autologous LMC adhesion assay using ECs, BECs, and LSECs. Adherent LMCs were stained and counted in 10 random high-power fields. LMCs from PBC livers adhered in greater numbers than did LMCs from disease control livers, using either ECs or BECs, whereas LMCs adhered only minimally to LSEC in all instances. **P < 0.01. Dashed diamond: hepatitis B virus. Open diamond: hepatitis C virus.

Production of TNF-α in the Presence of LPS by LMCs (T Cells, Monocytes, NKT Cells, NK Cells, and mDCs).

The necessity of TNF-α for production by BECs of CX3CL1 led us to assess the source of available liver TNF-α. As shown in Fig. 4, LMCs produced TNF-α following stimulation with most TLR ligands, and values for PBC exceeded those for disease controls. The data were as follows: LTA, 751 ± 163 versus 547 ± 138 pg/mL (P < 0.05); LPS, 1,699 ± 253 versus 1,303 ± 244 pg/mL (P < 0.01); and CL-097, 956 ± 188 versus 726 ± 154 pg/mL (P < 0.05) (Fig. 4). In the case of early noncirrhotic PBC, only a limited quantity of LMCs was available so that TNF-α production was measured only with or without LPS stimulation; here, TNF levels were 1,825 ± 334 pg/mL, which did not differ significantly from cirrhotic PBC (P > 0.05). There were, however, differences between noncirrhotic PBC and cirrhotic disease controls (P < 0.05) (Fig. 4). We then determined which subpopulations of LPS-stimulated LMCs produced TNF-α and, as shown in Fig. 5, the data for PBC livers versus disease control livers were as follows: monocytes, 476 ± 131 versus 336 ± 65 pg/mL (P < 0.05); NK cells, 179 ± 51 versus 107 ± 36 pg/mL (P < 0.01); and mDCs, 264 ± 60 versus 199 ± 38 pg/mL (P < 0.05). Thus, the major but not exclusive source of TNF-α was monocytes, and production of TNF-α was relatively greater in PBC.

Figure 4.

TNF-α production by LMCs from PBC and control livers with different TLR ligand stimulation. Greater TNF-α responses were elicited by LTA, LPS, and CL-097, and the TNF-α response for PBC and disease control livers was significantly greater for each of these (LTA [*P < 0.05], LPS [**P < 0.01], CL-097 [*P < 0.05]). LMCs from cirrhotic and noncirrhotic PBC were compared, and no differences were seen. Here, the open circle represents noncirrhotic PBC.

Figure 5.

TNF-α production from different cell types under LPS stimulation using isolated cells from PBC and disease controls. From a starting population of LMCs, separation was made of T cells, monocytes (Mo), NKT cells, NK cells, and mDCs. Monocytes elicited the highest TNF-α production among these populations, NK cells and mDCs produced measurable TNF-α responses, and T cell and NKT cell responses were minimal. TNF-α responses differed for PBC and disease control livers for monocytes (*P < 0.05), NK cells (*P < 0.01), and mDCs (*P < 0.05). Dashed diamond: hepatitis B virus; Open diamond: hepatitis C virus; Hatched diamond: primary sclerosing cholangitis.

BECs Pretreated with Poly I:C Produce CX3CL1 in the Presence of Monocytes Pretreated with LPS.

We have shown that direct contact of LMCs and TNF-α were necessary for production of CX3CL1 by BECs. In addition, it is known that TLR4 ligands stimulate LMCs to produce TNF-α. Accordingly, we sought to ascertain which cell population among LMCs is critical for CX3CL1 production by BECs. Our procedures included measurement of production of CX3CL1 by poly(I:C) pretreated BECs, with LPS pretreated mononuclear cells of either T cells, monocytes, NKT cells, NK cells, or mDCs (Fig. 6). Even though NK cells and mDCs did produce small amounts of TNF-α with LPS, production of CX3CL1 was rarely detectable when poly(I:C)-pretreated BECs were cocultured with LPS-pretreated T cells, NKT cells, or NK cells, or mDCs; Fig. 6A shows representative data for one PBC liver. On the other hand, CX3CL1 production was prolific when poly(I:C)-pretreated BECs were cultured with LPS-pretreated monocytes. Such production was not observed in the absence of LPS-pretreated monocytes, and the production was markedly inhibited after addition of anti–TNF-α (Fig. 6B), indicating that LPS-pretreated monocytes provided the necessary direct contact, and TNF-α, for subsequent CX3CL1 production by BECs. Comparison of PBC and disease control livers showed that poly(I:C)-pretreated BECs from PBC livers produced relatively large amounts of CX3CL1 when cultured with LPS-pretreated monocytes (2.1 ± 0.5 ng/mL versus 1.3 ± 0.4 ng/mL [P < 0.01]) (Fig. 6C). Of note, in these experiments, only small amounts of CX3CL1 were produced from the two primary sclerosing cholangitis livers.

Figure 6.

CXC3L1 production by BECs in the presence of monocytes: representative data from one PBC liver is shown in (A) and (B). (A) T cells, monocytes, NKT cells, NK cells, or mDCs were each preincubated with LPS for 24 hours and cocultured with BECs that had been preincubated with poly(I:C) for 24 hours. Only monocytes showed augmented production of CX3CL1. (B) BECs preincubated with poly(I:C) or poly(I:C) and LPS were cocultured with monocytes or LPS-preincubated monocytes, as shown in (A); anti–TNF-α abrogated the increased production by monocytes of CX3CL1 (**P < 0.01). (C) LPS pretreatment of monocytes (Mo) induced a greater degree of CX3CL1 production by poly(I:C) pretreated BECs from PBC livers than disease control livers (**P < 0.01). Each of the TLR ligands was used at 10 μg/mL. Pretreatment was performed for 24 hours, and cells were washed twice before coculture.

Monocytes Around BECs in the Liver.

Finally, we investigated the presence of monocytes around bile ducts in the liver by way of immunohistochemical analysis. Comparing livers of patients with PBC and those with hepatitis C (disease controls), CD68+ monocytes/macrophages were enriched in PBC, predominantly in the portal area (Fig. 7A), as were CD154+-activated T cells around biliary ductules (Fig. 7B); this is indicative of greater invasion of CD68 and CD154+ cells into portal areas of liver in patients with PBC compared with hepatitis C patients. Actual cell counts are shown in Table 1.

Figure 7.

Immunohistochemical detection of CD68+ cells (mainly monocytes) and CD154+ cells (mainly activated T cells) in PBC liver, hepatitis C liver, and normal liver. (A) Expression of CD68 on mononuclear cells around bile ducts. Mononuclear cells expressing CD68 (brown staining) are seen in the biliary epithelial layer and periductal tissue. In PBC, CD68+ cells are seen within the biliary epithelium (arrowheads) and are plentiful around the bile ducts (arrows), whereas in hepatitis C, CD68+ cells are scattered, and in normal are rarely seen around bile ducts. (B) CD154 expression on mononuclear cells (arrows) around bile ducts is high in PBC liver, but weak on mononuclear cells around bile ducts in hepatitis C liver.

Table 1. Immunohistochemical Staining and Counts of CD68+ and CD154 Cells Around Bile Ducts in Liver Sections from Patients with PBC, Chronic Hepatitis C, and Discrete Liver Tumor but Otherwise Normal Liver
 PBC (n = 12)Hepatitis C (n = 15)Normal Liver (n = 7)
  • *

    In PBC, counts of CD68+ monocytes exceeded counts in hepatitis C and in normal liver (P 0.01, for both comparisons), as did counts of CD154+-activated T cells (P 0.01 for both comparisons).

CD68147.3 + 62.3*69.5 + 35.810.4 + 6.3
CD1543.4 + 0.8*1.0 + 0.90.5 + 0.4


To facilitate understanding of the data herein, we have developed a schema to reflect the chain of events among the liver subpopulations studied (Fig. 8). We also note that the hypothesis that aberrant homing of T cell subsets are involved in the pathogenesis of PBC is based on earlier data in primary sclerosing cholangitis.23 Samples from the study herein were primarily derived from end-stage disease, thus raising the issue of whether pathogenetic mechanisms that induce disease are overwhelmed by secondary immunological processes, including the contributions of fibrosis and extensive cholestasis. However, by reason of tissue access, this was a necessity. We should also note that prolonged culturing of ECs, LSECs, and BECs introduces the possibility of loss of cell differentiation, which we attempted to obviate by studies of early cultures (passages 5-6). However, such reservations notwithstanding, we emphasize that the LMCs from PBC strongly induce production of CX3CL1 from BECs. In future studies, these reservations could potentially be addressed by use of laser-captured microdissection and in real-time analysis for study of site-specific expression of messenger RNA from the relevant hepatic subpopulations. Indeed, a weakness of the data herein is the absence of completely normal nondiseased liver; such tissue is not readily available.

Figure 8.

A diagrammatic representation of invasion of liver and destruction of bile ducts by CX3CR1+ cells. Shown are the pathways of activation by TLR3 and TLR4 ligands; sources of CX3CR1 include vascular endothelial cells of the portal vein, the likely site of leukocyte entry. Up-regulation of CX3CL1 on BECs is promoted by TLR3 ligand, by activated monocytes through direct contact and TNF-α secretion by monocytes. A chemotactically driven leukocyte–CX3CL1 interaction involving BECs promotes the destructive cholangitis of PBC. Events in detail are as follows. (A) Various TLR ligands, particularly TLR3 and TLR4, stimulate vascular endothelial cells of the portal vein to produce CX3CL1. (B) CX3CR1+ leukocytes invade the liver from the portal vein. (C) TLR3 ligand stimulates BECs to up-regulate CX3CL1. (D) TLR4 ligand activates resident and CX3CR1+ invading monocytes. (E) Monocytes stimulate BECs by direct contact to up-regulate CX3CL1. (F) Monocytes stimulate BECs by secretion of TNF-α to up-regulate CX3CL1. (G) CX3CR1+ leukocytes invade BECs by engaging CX3CL1, thereby causing cholangitis.

CX3CL1 is potentially involved in multiple other inflammatory processes. This has already been described not only in noncholestatic disease, but also in lung inflammation with associated smoke injury.8, 24 Hence, the data herein is not necessarily specific for PBC. Our data should also be contrasted with studies in gut. Intestinal microvascular ECs do not spontaneously produce CX3CL1, but can do so after stimulation with TNF-α or IFN-γ or direct leukocyte contact, and this effect is significantly stronger using ECs from patients with inflammatory bowel disease versus control ECs.21 Interestingly, liver ECs that are epithelial cell marker–negative and CD31+-adherent mononuclear cells produced CX3CL1 upon TLR stimulation, but production did not differ between livers from PBC patients and those from chronic viral hepatitis. Notably with LSECs (epithelial cell marker–negative, CD31 and CD105+), adherent mononuclear cells failed to produce CX3CL1 under any form of stimulation, perhaps a reflection of the capacity of LSECs to induce antigen-specific tolerance within the liver.25 The CXCR3 ligands CXCL9 to CXCL11 are dominant on LSECs, whereas the CCR5 ligands are dominant on the portal vascular endothelium.26 Thus, our findings suggest that CX3CR1+ cells invade the liver by way of the portal vascular endothelium. As noted in the data herein, we have demonstrated that ECs, LSECs and BECs from disease controls behave similar to cell populations isolated from cirrhotic PBC and the other control populations studied in our CX3CL1 production assays, indicating that these liver cell subpopulations respond equally well against danger signals (i.e., TLR ligands) irrespective of changes related to fibrosis or in vitro culture artifacts.

In order for CX3CL1 to be produced by BECs, direct contact with autologous LMCs is clearly required, because this production was inhibited when the CD40–CD154 interaction was blocked, in line with a previous report that there was reduced production of chemokines from BECs exposed to activated liver macrophages after the CD40–CD154 interaction had been blocked.27 These data take on added importance in light of the known capacity of biliary ductular cells in PBC to express CD40.28 Furthermore, TLR3 stimulation induces high levels of CD40 on BECs,1 and pretreatment of BECs with the TLR ligand poly(I:C) may increase the CD40–CD154-dependent interaction between BECs and LMCs.

We particularly examined differences between LMCs that were derived from patients with PBC versus those with an inflammatory disease of another causation, chronic viral hepatitis. Because CX3CL1 also functions as an adhesion molecule, we note that ECs produced high levels of CX3CL1 compared with BECs, and that LMCs from PBC patients attached to ECs at higher frequencies than to BECs, whereas LMCs from viral hepatitis patients showed only minimal attachment to ECs or BECs. There were also significant differences in adhesion to either ECs or BECs when LMCs were compared from patients with PBC versus comparison cases. LMCs after TLR4 stimulation produced TNF-α, and in this particular case there were significantly higher levels of TNF-α produced from LMCs from PBC compared with control cases.

There are clearly multiple interactions that occur in PBC and other inflammatory liver diseases with respect to cytokines, chemokines, and their cognate receptors; such is the case for murine models of PBC as well.29, 30 Within this context, as well as the schema presented above, the immunobiology of CX3CL1 has been recently demonstrated to interact with multiple other receptors and molecules. Indeed, as examples, ADAM10, ADAM17, and MMP2, produced by activated hepatic stellate cells, may also lead to shedding of CX3CL1.20 Thus, we report that the atypical chemokine-adhesion molecule CX3CL1 (fractalkine) is an important participant in PBC that leads to periductular accumulation of lymphoid cells. This conclusion should be tempered with our availability of clinical samples, primarily end-stage patients that may not mirror early events.