Human Liver Dendritic Cells Promote T Cell Hyporesponsiveness¹

The liver is charged with regulating differential immune responses to both harmless foreign Ags and invasive pathogens (1). Autoimmune hepatitis and clearance of certain hepatotropic pathogens like hepatitis A represent the liver’s capability to manifest a robust immune response. The overall balance in hepatic immunity, however, appears to favor the induction of immunologic hyporesponsiveness and sometimes tolerance. For example, the lack of systemic inflammation despite continuous exposure to ingested food Ags highlights the liver’s role in oral tolerance (2). The liver is a frequent site of chronic viral and parasitic infection and is the most common site of distant metastatic disease in cancer. Furthermore, whereas HLA-disparate heart, lung, and kidney transplants are readily rejected, the acceptance of allogeneic liver transplants in HLA-mismatched patients is typical. In addition, the occurrence of T cell-mediated rejection in liver transplant recipients is low, and in some cases patients can be weaned off immunosuppression altogether (3). Although these phenomena have led to wide-ranging hypotheses to explain the liver’s tolerogenic predisposition, none have been definitively established in humans.

The liver contains a distinct repertoire of immune cells. It is enriched in NKT cells that make up 30% of all hepatic CD3⁺ lymphocytes in human nonparenchymal cells (NPC)⁴ compared with less than 1% of PBMCs. Furthermore, the 1:3.5 ratio of CD4⁺ to CD8⁺ T cells in the liver is distinct from the 2:1 ratio in the blood (4). Hepatic NK cells also contain a larger fraction of CD16⁻ cells than does the blood (5). A variety of APCs, including Kupffer cells and dendritic cells (DCs), also reside in the liver.

DCs play a vital role in the initiation of innate and adaptive immune responses, and their role as immune mediators in cancer and infection has been studied extensively (6). More recently, several lines of evidence have emerged supporting the importance of DCs in the induction and maintenance of tolerance (7, 8, 9). The use of such tolerogenic DCs in animal models has revealed that their adoptive transfer can limit the incidence of transplant rejection and curb the progression of certain autoimmune diseases (10, 11). In humans, evidence of the tolerogenic capacity of DCs is limited to work using immature, cytokine-generated monocyte-derived DCs (MoDCs) that are prepared ex vivo. Injection of Ag-loaded immature MoDCs into patients decreased immunity and induced the generation of T regulatory cells (T_reg) (12, 13). The precise mechanisms governing development of tolerance in situ in humans, however, remain unknown.

Human DCs have been defined traditionally as HLA-DR⁺ cells that lack the hematopoietic cell lineage (lin) markers CD3, CD14, CD16, CD19, and CD56. DCs have been subdivided into conventional myeloid (CD11c⁺CD123^low) and plasmacytoid (CD11c⁻CD123^high) populations (14). Four additional surface Ags specific to human DCs have further distinguished subsets of DCs. BDCA-2 (CD303) and BDCA-4 (CD304) are expressed on all plasmacytoid DCs, whereas BDCA-1 (CD1c) and BDCA-3 (CD141) are each expressed by a subgroup of conventional myeloid DCs (15, 16). Meanwhile, a population of CD16⁺ DCs has been recognized to comprise the majority of freshly isolated blood DCs (17, 18, 19, 20, 21). CD16⁺ DCs are M-DC8⁺(6-sulfo LacNAc⁺)CD11c⁺BDCA-1⁻CD123⁻, which in addition to inducing T cell proliferation, mediate Ab-dependent cellular cytotoxicity, stimulate NK cells, and produce large amounts of TNF-α and IL-12p70 in response to LPS (18, 19, 21). Overall, at least four distinct DC subgroups are now recognized after fresh isolation from human peripheral blood: BDCA-1⁺, BDCA-3⁺, CD16⁺, and CD123^high.

The characterization of human liver DCs and the precise role they play in modulating the distinct immune environment within the liver have yet to be fully elucidated. To investigate whether DCs contribute to the tolerogenic environment of the liver, we analyzed freshly isolated DCs from human liver in 56 patients and compared them with freshly isolated autologous blood DCs.

Fresh blood and liver samples were collected from 56 individual patients undergoing elective hepatic resection at Memorial Sloan-Kettering Cancer Center. Operations were performed for benign disease (n = 7), primary cancer (n = 9), and metastatic disease (n = 40). Informed consent was obtained according to an Institutional Review Board-approved protocol. Blood was drawn intraoperatively, and PBMC were isolated by density centrifugation over Ficoll-Paque^Plus (LPS free; GE Healthcare). Liver tissue was procured at least 5 cm away from any tumor and processed immediately following removal from the patient. To remove contaminating circulating blood cells, liver vessels were flushed with HBSS containing type IV collagenase (1 mg/ml; Sigma-Aldrich), DNase (50 ng/ml; Roche Diagnostics), and 2% endotoxin-free human AB serum (Invitrogen). Liver tissue was then morselized and digested in the collagenase solution at 37°C for 30 min. The digestion was quenched with cold HBSS containing DNase (25 ng/ml), and then passed through a 100-μm filter. The cell suspension was centrifuged twice at 250 × g for 10 min to remove fat and debris. The cell pellet was then reconstituted with HBSS and DNase, and the parenchymal hepatocytes were removed with a low speed spin (30 × g for 1 min). Mononuclear cells from the remaining NPC were isolated by Ficoll-Paque^Plus density centrifugation.

Spleen samples were processed directly following removal from patients undergoing elective pancreatic resections in which the spleen was also removed. There was no pathologic evidence of disease in any of the spleen specimens removed. Five gram samples were digested and processed by an identical process, as described above. After quenching and filtering, the cell suspension was centrifuged at 300 × g for 7 min. The supernatant was discarded, and the cell pellet was resuspended in PBS and layered over a Ficoll-Paque^Plus density gradient, after which the interface layer was harvested.

Liver NPC and PBMCs were stained with FITC, PE, PerCP-Cy5.5, allophycocyanin, allophycocyanin-Cy7, and AF700 conjugated to Abs against CD3 (UCHT1), CD4 (RPA-T4), CD14 (M5E2), CD19 (HIB19), CD56 (NCAM 16.2), HLA-DR (L243), CD11c (B-ly6), CD16 (3G8), CD34 (581), CD123 (9F5), CD40 (5C3), CD80 (L307.4), CD25 (M-A251), CD83 (HB15e), CD86 (2331) (all BD Pharmingen), BDCA-1 (AD5-8E7), and BDCA-3 (AD5-14H12) (Miltenyi Biotec). We defined a set of lin-specific markers using a combination of anti-CD3, anti-CD14, anti-CD19, and anti-CD56 Abs. Appropriate isotype controls were used. Flow cytometry was performed on a FACS Aria flow cytometer (BD Biosciences), and data were analyzed with FlowJo software (Tree Star). Intracellular FoxP3 expression was assessed using the FITC anti-human FoxP3 staining kit (eBiosciences), and rat IgG2a FITC (BD Biosciences) was used as an isotype control. Sorting was performed on a MoFlo (DakoCytomation) or BD FACSAria (BD Biosciences) cell sorter. Bulk DCs were defined as lin⁻HLA-DR⁺. Postsort purities were routinely >98%.

A total of 1 × 10⁶ blood PBMCs and liver NPC was incubated at 37°C for 0, 5, 15, or 30 min in a solution of PBS with 1 mg/ml 10,000 m.w. FITC dextran or 50 μg/ml DQ OVA (Molecular Probes). The cells were then quenched with ice-cold PBS, washed twice, stained with Abs to lin and DC surface markers, and analyzed on a FACSCalibur flow cytometer (BD Biosciences).

FACS-purified bulk liver and blood DCs were cultured in triplicate at varying concentrations with 1 × 10⁵ allogeneic naive CD4⁺CD45RA⁺ T cells purified by immunomagnetic selection using the naive CD4⁺ T cell isolation kit (Miltenyi Biotec). The T cells were isolated from donor buffy coat leukocytes from healthy adult volunteers (Greater NY Blood Center, American Red Cross). Cells were routinely greater than 93% pure (for CD4⁺ and CD45RA⁺) and displayed no FoxP3 staining. On day 5 of coculture, supernatants were harvested and cytokine levels were measured using the human Th1/Th2 cytometric bead array kit (BD Biosciences). Cells were then pulsed with [³H]thymidine (1 μCi/well; PerkinElmer), and radioactive uptake was measured 18 h later with a TopCount NXT microplate scintillation and luminescence counter (PerkinElmer). For secondary MLRs, liver and blood DC-primed T cells were harvested on day 6, recounted, and restimulated with irradiated (2500 rad) PBMCs that were enriched for APCs by depletion of CD3⁺ cells using immunomagnetic beads (Miltenyi Biotec). These CD3⁻ PBMC APCs for restimulations were from the same donor as the DCs used in the primary MLR. [³H]Thymidine (1 μCi/well) was added to wells at the indicated time points after restimulation, and proliferation was measured 18 h later.

A total of 5 × 10⁴ freshly sorted bulk DCs or DC subsets from liver or blood was cultured in triplicate with medium alone or with LPS (1 μg/ml from Escherichia coli 055:B5; Sigma-Aldrich). Medium consisted of RPMI 1640, 2 mM l-glutamine, 1% nonessential amino acids, 100 U/ml penicillin, 100 μg/ml streptomycin, 0.1% 2-ME, and 10% heat-inactivated, endotoxin-free human AB⁺ serum (Media Preparation Core Facility, Memorial Sloan-Kettering Cancer Center). After 24 h, supernatant was harvested and cytokine levels were measured using the human inflammation cytometric bead array kit (BD Biosciences).

For isolation of T cells primed by liver DCs for 7 days, staining was performed using anti-CD4 (RPA-T4), and anti-CD25 (M-A251) conjugated to PE or allophycocyanin, respectively. The CD4⁺CD25^high (2–3% of highest CD25⁺ cells) and CD4⁺CD25⁻ fractions were purified via FACS and used as suppressors or negative controls, respectively. Naive CD4⁺CD45RA⁺ T cells were freshly isolated by immunomagnetic selection (Miltenyi Biotec) and labeled with CFSE (5 μM; Invitrogen). These naive T cell responders were cultured alone or in the presence of anti-CD3/CD28 mAb-coated Dynabeads (Invitrogen; 2:1 bead-total cell ratio) with varying concentrations of CD4⁺CD25^high or CD4⁺CD25⁻ cells. Proliferation of the CFSE-labeled cells was determined after 2 or 3 days of coculture.

A total of 5 × 10⁴ allogeneic peripheral blood naive CD4⁺CD45RA⁺ T cells isolated from buffy coats was cultured in triplicate with FACS-purified bulk blood or liver DCs (DC:T cell ratio of 1:10) for 7 days. In some cultures, IL-10-blocking Ab (1 μg/ml; BD Biosciences) was added. Rat IgG1 (OX86, 1 μg/ml; mAb Core Facility, Memorial Sloan-Kettering Cancer Center) was used as an isotype control. Cells were collected after 7 days of primary stimulation with DCs, recounted, and then restimulated with 5 μg/ml plate-bound anti-CD3 and 1 μg/ml soluble anti-CD28 (BD Biosciences) for 24 h. Culture supernatants were harvested 18 h after restimulation, and the levels of IL-4 and IFN-γ were measured using a human Th1/Th2 cytometric bead array (BD Biosciences). For intracellular cytokine analysis, the primed CD4⁺ T cells were restimulated, as described above, with the addition of brefeldin A (10 μg/ml; BD Biosciences) during the final 6 h of culture. The cells were then harvested and stained with a combination of fluorochrome-labeled Abs to CD3 and CD4, fixed, permeabilized, and stained with 0.5 μg of FITC-conjugated anti-IFN-γ, 0.03 μg of PE-conjugated anti-IL-4, or the appropriate isotype controls (BD Biosciences).

A total of 2.0 × 10⁵ to 4 × 10⁵ FACS-purified bulk liver and blood DCs was pulsed for 3 h with a pool of 23 peptides corresponding to HLA class II-restricted T cell epitopes from CMV, EBV, influenza virus, and tetanus toxin (selected MHC class II Peptide Pool^Plus, 40 μg/ml; Axxora). Autologous peripheral blood CD4⁺ T cells were obtained using the CD4⁺ T isolation kit (Mitenyi Biotec). Peptide-pulsed DCs were washed twice in PBS and cultured with CD4⁺ T cells at a ratio of 1:10 for 7 days. IL-2 (12.5 IU/ml; R&D Systems) was added on day 2 and every other day for 7 days. Blood and liver DC-primed T cells were restimulated on day 7 using irradiated (2500 rad), peptide-pulsed (10 μg/ml) autologous CD3⁻ PBMCs at a ratio of 1:1. ELISPOT assays were performed 7 days later after 24 h of restimulation with irradiated, peptide-pulsed CD3⁻ PBMCs at a ratio of 2 APC:1 CD4⁺ T cell in X-VIVO 15 serum-free medium (BioWhittaker). Induction of IFN-γ and TNF-α production was measured using the human IFN-γ and TNF-α ELISPOT kits (BD Biosciences), and cells were counted with an automated immunospot device (Cellular Technology). T cells cultured in PHA (5 μg/ml; Sigma-Aldrich) were used as positive controls.

Data were analyzed for statistical significance with the nonparametric, unpaired, two-tailed Mann-Whitney U test or the two-tailed Wilcoxon matched pairs test, as indicated (Prism statistical software; GraphPad). Values of p <0.05 were considered statistically significant in all cases except Fig. 5C, in which the α level was adjusted via the Bonferroni method to avoid an inflated type I error rate. In Fig. 5C, in which two simultaneous comparisons were made, p values <0.025 were considered statistically significant.

The subset composition of freshly isolated human liver DCs is unknown. We therefore analyzed liver lin⁻HLA-DR⁺ cells for the DC subsets known to exist in the blood. All four major subsets of freshly isolated blood DCs were also present in the liver (Fig. 1,A). CD16⁺ DCs were more prevalent in the blood (median 64 vs 30%), whereas BDCA-1⁺ DCs were more common in the liver (median 36 vs 15%). The percentage of CD123^high DCs was similar between blood and liver (median 13 vs 15%), whereas BDCA-3⁺ DCs (median 3% in blood and 6% in liver) were infrequent (Fig. 1,B). Overall, the trend of liver DC composition and the differences between liver and blood DCs were similar among patients, except for one patient with hemochromatosis and hepatocellular carcinoma who had a high proportion of CD123^high DCs in both sites (Fig. 1,B). The average yields of hepatic NPC and PBMCs were 1.8 × 10⁶ cells/g liver and 1.2 × 10⁶ cells/ml blood processed, respectively. The average yield of bulk DCs after sorting was 4.2 × 10³ cells/g liver and 4.3 × 10³ cells/ml peripheral blood (Fig. 1 C).

Maturation is critical in determining the functional ability of DCs to orchestrate protective immunity to infection and tumors (22, 23). Despite the differences in subset composition between bulk liver and blood DCs, we found that freshly isolated populations expressed similarly high surface levels of HLA-DR, low levels of CD86, and essentially no CD40, CD80, or CD83 (Fig. 2, A and B). Specific liver and blood DC subsets also had similar maturation. Nevertheless, using FITC-conjugated dextran to investigate Ag uptake, we found that bulk, CD16⁺, and BDCA-1⁺ DCs from the liver were less efficient at Ag uptake (Fig. 2, C and D). Additional experiments using DQ OVA, which requires uptake and proteolytic cleavage to become fluorescent, revealed that bulk liver DCs as well as the CD16⁺ and BDCA-1⁺ DC subsets, were less efficient at Ag processing than bulk blood DCs and their respective subsets (Fig. 2, C and D). Thus, liver DCs were as phenotypically immature as blood DCs, but functionally less able to capture and process Ag.

One of the hallmarks of DC function is the ability to stimulate naive T cells. Freshly isolated bulk liver DCs were weaker than autologous bulk blood DCs at inducing the proliferation of allogeneic naive CD4⁺ T cells (Fig. 3,A). Furthermore, supernatants harvested from MLR cultures stimulated by bulk liver DCs contained significantly lower levels of IFN-γ and TNF-α and a greater amount of IL-10 (Fig. 3,B). To determine whether disparity in the isolation procedure for blood and liver DCs may account for the observed differences between the two DC populations, we subjected whole blood to the collagenase-based procedure we used to isolate liver DCs. Both subset composition and activation state of blood DCs were unchanged (Fig. 3,C). In addition, there was no difference in T cell allostimulation by blood DCs isolated by the collagenase method compared with autologous blood DCs isolated via the standard protocol described in Materials and Methods (Fig. 3 D).

To determine whether priming by liver or blood DCs had a differential effect on T cell proliferative capacity following restimulation, we performed secondary MLRs. Irradiated PBMCs enriched for APCs by depletion of CD3⁺ cells were used to restimulate liver or blood DC-primed T cells isolated from primary MLRs. Liver DC-primed naive CD4⁺ T cells were significantly less responsive to restimulation than blood DC-primed T cells at serial time points and at different APC concentrations (Fig. 4).

To investigate why liver DCs accomplished less alloproliferation, we examined their ability to produce cytokines that may modulate their interaction with T cells. Consistent with the previous findings, freshly sorted blood DCs activated by LPS produced high levels of inflammatory cytokines, but nearly undetectable amounts of IL-10 (24). In contrast, liver DCs made 10 times more IL-10 than blood DCs upon culture alone and over 1 ng/ml IL-10 after culture with LPS (Fig. 5,A). In addition, liver DCs made much less IL-1β, IL-6, and TNF-α than blood DCs after activation. To determine whether the differences in cytokine production were unique to liver DCs, we performed similar experiments with human spleen and autologous blood DCs. Although bulk spleen DCs made more IL-10 than their blood counterparts, they made comparable amounts of inflammatory cytokines upon LPS stimulation (Fig. 5,B) and did not induce T cell hyporesponsiveness (data not shown). Further analysis of liver DCs revealed that the liver BDCA-1⁺/BDCA-3⁺ DC subsets were largely responsible for IL-10 production, whereas liver CD16⁺ DCs accounted for IL-1β, IL-6, and TNF-α secretion (Fig. 5 C). As expected, CD123^high DCs did not produce cytokines in response to LPS because they do not express TLR4.

IL-10 is a potent immunosuppressive cytokine and is known to shape CD4⁺ Th cell responses (25). We therefore tested whether IL-10 from liver DCs conditions naive CD4⁺ T cells. After polyclonal restimulation of naive CD4⁺ T cells, which had been cultured for 7 days with allogeneic DCs freshly isolated from either liver or blood, we found that liver DCs generated a substantially lower percentage of IFN-γ⁺ T cells (Fig. 6,A). This effect depended on IL-10, because the presence of an IL-10-blocking Ab restored the percentage of IFN-γ⁺ T cells to that generated by blood DCs (Fig. 5,A). Pooled data from multiple patients revealed that the average percentage of IFN-γ⁺ T cells in liver DC cocultures was one-third of that in blood DC-T cell cultures (Fig. 6,B). We substantiated our intracellular cytokine data by measuring the amount of supernatant IFN-γ protein from T cells that were primed for 7 days with either freshly isolated blood or liver DCs and then polyclonally restimulated for 24 h. Liver DC-T cell cultures had significantly lower levels of IFN-γ unless IL-10 was blocked (Fig. 6,C). These findings were consistent with the lower supernatant IFN-γ levels in the liver DC MLR cultures (Fig. 3,B). In addition to the decreased Th1 cytokine profile, liver DC-T cell cultures yielded significantly higher percentages of IL-4⁺ T cells (Fig. 6, D and E). Similarly, supernatant amounts of IL-4 following polyclonal restimulation were higher in the T cell cultures primed with liver DCs (Fig. 6,F). Consistent with the increase in the Th1 response by liver DCs in the presence of an anti-IL-10 Ab, IL-10 blockade also abrogated the Th2 response induced by liver DCs. Simultaneous analysis of IFN-γ- and IL-4-producing effectors following polyclonal restimulation revealed that the cytokines were produced by distinct CD4⁺ T cells (Fig. 6 G).

To investigate additional immunomodulatory effects of liver DCs, we analyzed MLR cultures on day 7 for the presence of CD4⁺FoxP3⁺ T cells. We found that blood DC-T cell cultures contained relatively low percentages of CD4⁺FoxP3⁺ T cells, whereas liver DC cultures generated up to 6% CD4⁺FoxP3⁺ T cells (Fig. 7,A). Further phenotypic characterization of the CD4⁺FoxP3⁺ T cells that had been primed by fresh liver DCs revealed that 98 ± 2% stained positively for CD25 (data not shown). The generation of CD4⁺FoxP3⁺ T cells in liver DC-T cell cultures depended on IL-10, as demonstrated with a blocking IL-10 Ab (Fig. 7,A). Data pooled from multiple patients revealed that the percentage of CD4⁺FoxP3⁺ T cells generated by liver DCs was 3 times more than blood DCs (Fig. 7,B). Similarly, the absolute number of CD4⁺FoxP3⁺ T cells in the liver DC-T cell cultures also increased on day 7 in an IL-10-dependent manner (Fig. 7,C). To confirm the regulatory function of the CD4⁺FoxP3⁺ T cells, we sorted CD4⁺CD25^high and CD4⁺CD25⁻ T cells from the liver DC-T cell cultures and incubated them with third-party allogeneic naive CD4⁺ T cells and anti-CD3/CD28 beads. CD4⁺CD25^high T cells were able to inhibit T cell proliferation in a dose-dependent manner as measured by CFSE dilution (Fig. 7, D and E).

To extend our findings that allogeneic T cells conditioned by liver DCs were hyporesponsive after restimulation (Fig. 4), we tested how Ag-specific stimulation of CD4⁺ T cells by liver DCs affected subsequent Ag rechallenge. Sort-purified liver and blood DCs were pulsed with a pool of HLA class II-restricted peptides and then cultured with bulk CD4⁺ T cells for 7 days. T cells were then incubated with autologous, peptide-loaded, CD3⁻ PBMCs for another 7 days. An ELISPOT assay was then performed after 24 h of restimulation with Ag-loaded CD3⁻ PBMCs. Presentation of such class II-restricted peptides by APCs can induce the secretion of IFN-γ and TNF-α by Ag-specific CD4⁺ T cells. We consistently found that liver DCs induced significantly fewer IFN-γ- and TNF-α-producing cells when compared with blood DCs (Fig. 8). The differential effect on cytokine production was abrogated when CD4⁺ T cells stimulated by liver and blood DCs were restimulated nonspecifically with PHA.

Our understanding of human DCs is based almost exclusively on cultured peripheral blood DCs and various DC types generated with cytokines in vitro. There are scant data on the function of DCs from human organs because of limited access to tissue specimens and difficulty in isolating sufficient numbers of cells for analysis. It is becoming increasingly apparent, however, that organ-specific immune responses differ and adapt to their local environment. For example, the lung and gut, which are laden with foreign Ags and symbionts, display unique regulatory mechanisms in response to TLR activation (26, 27). Similarly, to protect against potentially harmful immune responses, the liver must adapt to the portal circulation, which is rich in endotoxin as well as harmless foreign Ags. Our results highlight the importance of studying organ-specific immunity and demonstrate that freshly isolated liver DCs most likely contribute to hepatic tolerance based on their distinct subset composition and function.

Although we showed that the majority of freshly isolated blood DCs were CD16⁺, as previously reported (20), BDCA-1⁺ DCs were the most prevalent subset within the liver. Similarly, a prior report of two patients also found that lin⁻CD11c⁺CD16⁻CD123^low hepatic DCs (which include BDCA-1⁺ and BDCA-3⁺ DCs) constituted the majority of liver DCs (28). Not only did liver and blood DC composition differ, but liver DC composition was also distinct from spleen (data not shown). It is unclear whether differences in subset composition within the liver are a consequence of selective recruitment of DC subsets or the result of differential intrahepatic development from circulating DC precursors. Importantly, the differential results between liver and blood DCs did not depend on the method of isolation.

The distinct cytokine profile of liver DCs compared with both blood and spleen DCs supports the hypothesis that DCs undergo local conditioning within the liver. Previous reports of cytokine production by human liver DCs have been confounded by over 50% contamination with CD14⁺ cells, which contain Kupffer cells (29, 30). We found that activated liver DCs produced IL-10, but comparatively low levels of proinflammatory cytokines. The predominance of liver BDCA-1⁺ DCs and their production of IL-10 may be relevant to chronic viral infection within the liver. Natterman et al. (31) demonstrated an elevation in the number of intrahepatic BDCA-1⁺ cells in patients with chronic hepatitis B or C when compared with normal subjects. In addition, in mouse models of chronic lymphocytic choriomeningitis, reduction of DC IL-10 production by a blocking Ab was sufficient to restore T cell function and eliminate viral infection (32). In contrast, the potential hepatoprotective role of IL-10 has been revealed using murine models of Con A-induced hepatitis, because IL-10 blockade resulted in increased hepatic inflammation and higher systemic levels of proinflammatory cytokines (33).

Liver DCs promoted the induction of functional T_reg from naive CD4⁺ precursors. Their immature state alone does not explain the capacity of fresh liver DCs to generate T_reg, because freshly isolated blood DCs were equally immature (Fig. 2, A and B). The generation of T_reg de novo by liver DCs is consistent with previous reports of CD4⁺CD25⁺ T_reg generated from naive (CD4⁺CD25⁻CD45RA⁺) T cell precursors using MoDCs (34). Liver DC-derived T_reg do not appear to fall into any of the currently known T_reg subsets. Their CD4⁺CD25⁺FoxP3⁺ phenotype is consistent with the naturally occurring subset of T_reg induced in the periphery, but their IL-10 dependency suggests that they may also share some similarities to the type 1 T regulatory subset of T_reg (35). Immature MoDCs possess the capacity to induce the generation of type 1 T_reg from allogeneic naive CD4⁺ precursors, but only following repetitive rounds of in vitro restimulation (7, 36). Furthermore, MoDCs rendered tolerogenic via treatment with exogenous IL-10 have also been shown to induce the differentiation of T_reg in vitro (37). The fact that freshly isolated liver DCs reproducibly generated higher numbers of T_reg in the absence of any exogenous treatment suggests that they may possess an inherent tolerogenic capacity. Additional studies will determine whether T_reg induced by liver DCs comprise a known or entirely novel subset of CD4⁺ T_reg.

In combination with an increase in T_reg, liver DCs simultaneously promoted a Th2 response. Although there is an extensive understanding of the specific pathways governing the differentiation of Th1 cells, the events controlling Th2-driven responses remain unclear. Although there are several models for the selective induction of a Th2 response by DCs, none have been definitively established (38). The current paradigm for DC-driven Th cell development is based on a cascade of events that occur after maturation of resting DCs (39). Notably, we found that bulk liver and blood DCs had similar increases in maturation during culture (data not shown). The three seminal events thought to mediate naive CD4⁺ T cell polarization include the following: stable Ag-MHC-II presentation, adequate costimulation, and a third signal, which in the case of Th1 differentiation is secretion of IL-12p70. Although alternate models exist that propose different sets of signaling cascades for both Th1 and Th2 development (40, 41), none have conclusively determined the third signal controlling Th2 development. Our data demonstrate that the Th2-polarizing capacity of liver DCs depends on IL-10. Evidence supporting such a role for IL-10 in murine models has been well established, because previous reports have revealed that IL-10 production by DCs in hepatic NPC and Peyer’s patches promoted Th2 responses (42, 43). Similarly, work on human MoDCs showed that IL-10 released after TLR2 stimulation was responsible for blocking the production of a number of Th1 cytokines that would otherwise be produced by stimulating MoDCs with TLR3 or 4 (44).

Although we provide direct functional assessment of human liver DCs supporting their tolerogenic role, this study does not rule out the possibility that other organ-resident DCs may exhibit similar properties. Our data are consistent with reports by Lu et al. (45), who provided early insight into the inherent tolerogenicity of murine liver DCs. Furthermore, in congruence with important observations on the regulatory role of liver DCs and the significance of donor-derived DCs in animal models of liver transplantation, our results lend support to the notion that liver DCs are critical regulators of alloimmunity and may contribute to liver allograft acceptance in humans (46, 47, 48). We demonstrate that freshly isolated human liver DCs promote T cell hyporesponsiveness after allogeneic and Ag-specific restimulation. The relative proportions of CD16⁺ and BDCA-1⁺ DC subsets are reversed in the liver compared with freshly isolated blood DCs. Paradoxically, liver DCs secrete more IL-10 and less proinflammatory cytokines such as IL-1β and TNF-α, which is unexpected from a population containing a large proportion of BDCA-1⁺ conventional DCs. Finally, freshly isolated liver DCs simultaneously promote a Th2 response and generate suppressive allogeneic CD4⁺FoxP3⁺CD25⁺ T_reg by an IL-10-dependent mechanism.

Although these properties of liver DCs may limit unnecessary immune responses to Ags in the portal circulation or to hepatic allografts, they may hinder immune responses to pathogens or tumors. Our results, therefore, have important implications to understanding hepatic immunity and should facilitate the development of therapeutic strategies for the treatment of liver diseases.

We are grateful to J. Hendrikx, M. Kweens, and P. Anderson (all from Flow Cytometry Core Facility, Memorial Sloan-Kettering Cancer Center) for their assistance with cell sorting.

The authors have no financial conflict of interest.