Bidirectional transendothelial migration of monocytes across hepatic sinusoidal endothelium shapes monocyte differentiation and regulates the balance between immunity and tolerance in liver

Isolation and Culture of Endothelial Cells

ECs were isolated from human explanted or resected liver and surplus donor tissue as described.15, 16 Briefly, parenchymal cells were isolated from collagenase-digested tissue over a 33%/77% Percoll gradient (Amersham Biosciences, GE Healthcare, Little Chalfont, UK), and magnetic selection with antibody HEA125 (Progen Biotechnik, Germany) was used to deplete cholangiocytes followed by anti-CD31 selection of HSECs (Invitrogen, Paisley, UK) and culture in human endothelial basal growth medium (Invitrogen), 10% (v/v) heat-inactivated AB human serum (HD Supplies, Glasgow, UK), 10 ng/mL vascular endothelial growth factor, and 10 ng/mL hepatocyte growth factor (PeproTech, Peterborough, UK) in collagen-coated culture flasks. All human tissue and blood was collected with local research ethics committee approval and patient consent.

Bidirectional Monocyte Transmigration Across HSECs

Bovine collagen I (Gibco, Life Technologies), 3 mg/mL in 10 × minimal essential medium (Life Technologies) pH 7.4, was polymerized in 24-well cell culture standing inserts (pore size 0.4 µm) at 37 °C (Merck Millipore, Watford, UK) equilibrated in supplemented HSEC medium for at least 3 days. To study phagocytosis, collagen matrix was supplemented with 1 × 10⁷ unopsonized fluorescence-labeled Zymosan A BioParticles (Molecular Probes, Invitrogen). Collagen gels were coated with 100 μL fibronectin at 37 °C for 30 minutes (Invitrogen/Gibco; 50 µg/mL in phosphate-buffered saline) and seeded with HSECs grown until confluence following stimulation with 10 ng/mL tumor necrosis factor-α (TNF-α) (PeproTech) and 10 ng/mL/interferon-γ (IFN-γ) (PeproTech) for 24 hours. CD14⁺ monocytes (2 × 10⁶) or presorted monocyte subsets in endothelial basal growth medium/0.1% bovine serum albumin were layered onto the fibronectin-coated collagen plugs. After 1.5 hours, nonadherent cells were washed off and the medium was replaced by endothelial basal growth medium containing 2.5% (v/v) heat-inactivated human serum. During the subsequent 48-hour incubation spontaneously reverse transmigrating (RT) monocytes were harvested by thoroughly washing off cells from above the endothelial layer with ice-cold phosphate-buffered saline/0.5 mM ethylene glycol tetraacetic acid/1% fetal calf serum. The collagen plugs containing retained subendothelial (SE) monocytes were gently digested in 20% (v/v) collagenase from Clostridium histolyticum (Sigma-Aldrich) and 20% (v/v) heat-inactivated fetal calf serum at 37 °C for 10 minutes. Digests were placed on ice, filtered, and resuspended in phosphate-buffered saline/2 mM ethylene diamine tetraacetic acid/1% fetal calf serum. We confirmed that surface markers were not lost during collagenase digestion (data not shown). Contaminating HSECs were depleted with biotinylated Ulex Europaeus Agglutinin I (Vector Laboratories, Burlingame, CA) and streptavidin-conjugated Dynabeads (Life Technologies, Carlsbad, CA) and magnetic depletion. In some experiments human umbilical vein endothelial cells (HUVECs) were used as endothelial cells. RT and SE monocytes were counted and subjected to further analysis or in vitro experiments. Trypan blue exclusion confirmed viability.

Statistical Analysis

The Student t test and GraphPad Prism software were used to compare numerical variables between two groups and one-way analysis of variance followed by Bonferroni's posttest, for comparisons between more than two groups. Results are expressed as mean ± standard error of the mean. P < 0.05 was considered statistically significant. *P < 0.05, **P < 0.01, ***P < 0.001.

For further information on materials and methods, please refer to Supporting Information.

Intrahepatic Accumulation of Monocytes/Macrophages Is Driven by Activated Endothelial Cells

In order to study the fate of monocytes after transmigration to the subendothelial compartment, we established a model of monocyte transmigration and reverse transmigration involving primary human HSECs, adapted from Randolph et al.17 Monocytes were allowed to migrate across activated HSECs and a thin layer of fibronectin into a three-dimensional collagen scaffold. After depletion of nonmigrating cells and 2 days of continued coculture, resident monocytes from the subendothelial compartment (SE monocytes) and monocytes that underwent spontaneous reverse transmigration (RT monocytes) from the subendothelial compartment back through the endothelial layer were harvested. To verify that RT monocytes had sampled the subendothelial compartment and had not simply detached from the endothelial monolayer, we supplemented the collagen scaffold with fluorescein isothiocyanate–labeled zymosan particles and tracked the localization of phagocytic monocytes/macrophages. Migration into zymosan-supplemented collagen increased the number of retained subendothelial cells, most of which were CD16^low. The RT monocytes contained zymosan, albeit at lower levels than the SE monocytes, and were predominantly CD16^high (Fig. 1B). Longtime imaging could visualize forward and reverse transmigration of monocytes across HSECs under cell culture conditions (Supporting Video). In order to confirm that the TEM was an active process, we reduced migration in both directions by inhibiting Janus kinase–signal transducer and activator of transcription signaling, which is required for macrophage TEM (Supporting Fig. S1).18 Interestingly, direct comparison of cell compartmentalization after monocyte TEM across HSECs or HUVECs revealed that HSECs clearly favor subendothelial retention (90% SE monocytes), whereas HUVECs permitted approximately 50% of monocytes to reverse transmigrate (Fig. 1C). To mimic the inflammatory environment present in the subendothelial compartment during liver disease, the collagen matrix was supplemented with conditioned medium of activated primary liver myofibroblasts. TNF-α/IFN-γ stimulation of activated primary liver myofibroblasts led to a profound increase in the total number of transmigrating cells, though the ratio of RT/SE was not significantly altered (Supporting Fig. S2). These data demonstrate that after active recruitment across HSECs monocytes differentiate into either sessile macrophage-like cells with high phagocytic capacity or mobile DC-like cells. Recruitment is augmented by cell-derived chemotactic stimuli.

RT Monocytes Express CD16 and Can Be Derived From All Monocyte Subsets

The majority of SE (mean 78.9 ± 9.8%) monocytes were classical CD14⁺⁺CD16^– monocytes, whereas 69.4% (±12.6%) of RT monocytes were intermediate CD14⁺⁺CD16⁺ and few were classical monocytes. Very few cells in either compartment were nonclassical CD14⁺CD16⁺⁺ cells, suggesting that this subset does not readily undergo TEM (Fig. 2A). Monocytes are highly plastic cells, and different subsets represent various states of maturity and differentiation, prompting us to determine how the different subsets in blood contributed to either SE or RT monocytes. When classical monocytes were used as the starting cell type, >80% were retained in the SE compartment (80.7 ± 12.6%) and fewer cells underwent reverse TEM compared with intermediate and nonclassical subsets (Fig. 2B,C), suggesting that CD16 expression is associated with the ability to undergo RT. Most RT monocytes were CD16⁺, indicating that these cells gain CD16 expression either during TEM or in the subendothelial space and that this confers on some cells the ability to undergo reverse TEM (Fig. 2C).

Reverse Transmigration of Monocytes Across HSECs Imparts a Phenotype Consistent With Immature DCs

We analyzed the different subsets for features of DC differentiation to see if reverse transmigration selects monocytes with the capability to become DCs and to reenter the vasculature before being recruited to lymph nodes through high endothelial venules.17 Both RT and SE monocytes expressed high levels of major histocompatibility class 2 and CD40 (Fig. 3). CD86 (B7-2) was expressed at higher levels on SE monocytes with little CD80 (B7-1) or CD83 detected on either subset. RT monocytes expressed higher levels of two scavenger receptors, the mannose receptor (CD206) and CD163, both of which are associated with alternatively activated macrophages, while CD206 is also present on immature DCs. Neither RT nor SE cells expressed high levels of DC-SIGN (CD209). The classical macrophage marker CD68 was expressed at similar levels on both SE and RT monocytes. CD11b (Mac-1) and CD11c are integrin alpha chains that form heterodimers with CD18 and bind complement and intercellular cell adhesion molecule 1. CD11c expression is characteristic of DCs and was detected at high levels on RT monocytes. Chemokine (C-C motif) receptor 7 (CCR7) and CCR8 regulate homing of DCs to the afferent lymph node.19 CCR8 was expressed at higher levels on RT monocytes, suggesting that it might be involved in reverse transmigration. There were no detectable differences in expression of CCR7 or CCR5 on RT versus SE cells, and neither subset expressed CCR2 (data not shown). RT monocytes tended to express higher levels of CX3CR1 and macrophage colony-stimulating factor receptor (CD115), both of which are characteristic of DCs.20 Thus SE and RT cells are similar in terms of costimulatory molecule expression, although RT monocytes have characteristics of promigratory, DC-like cells.

RT, but not SE, Monocytes Induce Robust T-Cell Proliferation

We compared RT monocytes with SE cells for their ability to induce CD4⁺ T-cell proliferation, a defining feature of DCs. While SE monocytes had little effect on T-cell proliferation, RT monocytes induced potent proliferative responses after pulsing with pp65 (Fig. 4A). We saw a similar although weaker effect in an allogeneic T-cell response (Fig. 4B). Thus, DC-like function resides in the RT population. In the presence of SE monocytes T-cell proliferation in response to either OKT3 or CD3/CD28 stimulation was significantly suppressed at T-cell ratio/monocyte ratios of 1:1 and below, whereas RT monocytes amplified baseline proliferation of CD4⁺ T cells (Fig. 4C). The inhibitory effect of SE monocytes was not due to the induction of apoptosis in CD4⁺ T cells (Fig. 4D).

RT Monocytes Are Capable of Inducing CD4⁺ T-Cell Activation, Whereas SE Macrophage-Like Monocytes Dampen Activation

Having observed that RT, but not SE, monocytes can induce antigen-specific CD4⁺ T-cell proliferation, we next compared the ability of SE and RT monocytes to drive T-cell activation using CD14⁺ monocytes from cytomegalovirus (CMV)–seropositive donors (Fig. 5A). RT monocytes activated CD4⁺ T cells as demonstrated by up-regulation of the late-activation markers CD25 and HLA-DR, whereas SE monocytes had no effect on activation and suppressed expression of the mid-early T-cell activation marker CD71 (Fig. 5B,C). We observed similar effects in an allogeneic system (data not shown). To obtain insights into the consequences of T-cell activation by RT monocytes, we quantified changes in genes involved in T-cell differentiation. SE monocytes increased expression of NFATC2 (NFAT1), which is linked to CD4⁺ T-cell anergy and suppression along with Fas ligand (FASLG) that triggers activation-induced T-cell apoptosis.21-23 In contradistinction CD4⁺ T cells activated by RT monocytes expressed T-box transcription factor EOMES (Eomesodermin) and TNFRSF9 (CD137) (five-fold and 4.76-fold increases, respectively, compared with cells stimulated with SE monocyte) (Fig. 5D). These genes are associated with T-cell memory and effector function and promote clonal expansion and T-cell survival.21, 22 Cells coincubated with RT monocytes also increased expression of several genes that drive Th2 commitment, including IL13, IRF4, GFI1, IL4R, and IL13RA1 (Fig. 5D), whereas genes that counteract Th2 polarization (POU2F2 and RUNX1) were more highly expressed in T cells cocultivated with SE monocytes.

RT Monocytes Induce Effector T Cells With Th1-Like Features

To further characterize the CD4⁺ T-cell antigen–specific response driven by RT monocytes, we assessed the release of cytokines by activated T cells upon coculture with CMV pp65 exposed autologous RT/SE monocytes. RT monocytes induced a significantly higher proportion of IFN-γ producing T cells than SE monocytes, suggesting Th1 polarization (Fig. 6A), although responding CD4⁺ T cells also increased CTLA-4 and FOXP3 expression, genes that are linked to regulatory T cells but also up-regulated during early T-cell activation. In line with the proliferation assays described above, we suggest that RT monocytes induce activated effector CD4⁺ T cells rather than a suppressive phenotype (Fig. 6B,C). Despite a genetic profile compatible with Th2 commitment, we failed to detect interleukin-4 (IL-4) in CD4⁺ T cells stimulated with RT monocytes (Fig. 6C). Consistent with local IFN-γ secretion, the RT monocytes in the cocultures showed significantly increased levels of programmed death ligand-1 (PD-L1, which is strongly induced by IFNs) compared with SE monocytes (Fig. 6D), whereas its cognate ligand PD-1 was not regulated on cocultivated CD4⁺ T cells (Fig. 6E). Comparing HSEC-shaped and HUVEC-shaped monocytes, we observed that SE monocytes derived from HUVEC interaction were able to elicit IFN-γ secretion by autologous CD4⁺ T cells, indicating that failure to induce a trans-Th1 response is a unique feature of liver-specific endothelium (Fig. 6F). We conclude that RT monocytes effectively activate CD4⁺ T cells and potentially prime T-helper cell function, whereas SE monocytes suppress T-cell activation and blunt their differentiation.

Resident SE Monocytes Fail to Mount a Proinflammatory Response Upon Toll-Like Receptor Engagement

Having shown differences in the ability of SE and RT monocytes to induce T-cell activation, we compared their ability to respond to toll-like receptor (TLR) engagement. Under baseline conditions after 24 hours in vitro culture, RT and SE monocytes secreted low levels of macrophage migration inhibitory factor, IL-8, and the anti-inflammatory IL-1 receptor antagonist. SE monocytes also secreted the chemokine (C-X-C motif) ligand 1 (CXCL1), chemokine (C-C motif) ligand 2 (CCL2), and CCL5. Neither RT nor SE monocytes spontaneously secreted the proinflammatory cytokines IL-1β, TNF-α, or IL-6 (Fig. 7A,B). Upon lipopolysaccharide (LPS) stimulation, both subsets secreted IL-6, showing the potential to respond to TLR-4 engagement; but while RT monocytes showed a proinflammatory response with increased IL-1β, TNF-α, IL-6, granulocyte colony-stimulating factor, and granulocyte-macrophage colony-stimulating factor secretion, SE monocytes were relatively refractory to LPS stimulation and failed to produce IL-1β, IL-10, IL-13, or TNF-α (Fig. 7A-C). In contrast, RT monocytes responded to LPS with increased secretion of CCL3, CCL4, CCL2, and CCL5 as well as a marked increase in IL-1β and TNF-α secretion. Thus, SE monocytes display phagocytosis, poor T-cell stimulation, and resistance to LPS, whereas RT monocytes are proinflammatory and able to activate T cells (Fig. 7A-C).

Functional Disparities Between RT and SE Monocytes Are Reflected by Extensive Transcriptional Differences

To get a more profound insight into the effects of TEM on monocytes, we analyzed transcriptional differences between RT and SE monocytes belonging to the different subsets. Intermediate CD14⁺⁺CD16⁺ contained the most distinct gene signature with 432 genes more than 10-fold up-regulated and 1179 genes more than 10-fold down-regulated when comparing SE and RT monocytes, whereas around 200 genes were differentially expressed (>10-fold) between SE and RT monocytes in both CD14⁺⁺CD16^– and CD14⁺CD16⁺⁺ monocytes. RT monocytes originating from all subsets displayed increased CCL3, CCL4, CXCL1, CXCL2, and IL-8, and proinflammatory genes including TNF family members and IL-1 compared with SE monocytes. CD14, which is typically maintained in macrophages that do not follow the DC pathway, was consistently up-regulated on SE monocytes of all subsets. In line with the immunosuppressive phenotype of SE monocytes, they expressed 15-fold higher levels of SIGLEC-7, a marker of anti-inflammatory macrophages (Supporting Tables S1-S3).