TGF-β Signaling Prevents MHC Class II–Expressing Lymphatic Endothelial Cells from Reactivating Human Allogenic Memory CD4⁺ T Cells

As the primary route of transport for fluid, Ags, exosomes, and immune cells from peripheral tissues to their draining lymph nodes (LNs), lymphatic vessels play critical transport functions in regulating immune responses. In addition to these transport functions, there is a growing appreciation for the direct immunomodulatory functions of lymphatic endothelial cells (LECs) that line lymphatic vessels, which can actively modulate the functions of various types of immune cell to shape the outcome of the immune response (1, 2). By expressing immunosuppressive factors such as NO, LECs can regulate the extent to which dendritic cell (DC)–activated T cells proliferate and protect inflamed LNs from excessive immune responses (3, 4). In peripheral tissue, LEC-expressed PD-L1 helps resolve the expansion of effector CD8⁺ T cells, preventing disproportionate tissue damage at the site of inflammation (5). LECs have also been shown to modulate the function of immune cells directly by acting as nonprofessional APCs (6); expression of peripheral tissue Ags, cross-presentation of exogenous Ags in the context of MHC class I, upregulated expression of PD-L1, and the lack of major costimulatory molecules such as CD40, CD80, and CD86 endows LECs with the potential to induce anergy in autoreactive naive CD8⁺ T cells (7–9).

In addition to modulating CD8⁺ T cells, LECs may also (both indirectly and directly) regulate the function of CD4⁺ T cells. Transfer of LEC-expressed peripheral tissue Ags to DCs for presentation in the context of MHC class II (MHC-II) leads to anergy in autoreactive naive CD4⁺ T cells under steady-state conditions (10). LECs can also acquire peptide-loaded MHC-II complexes from DCs and directly impair the activation of Ag-specific naive CD4⁺ T cells by DCs (11). Furthermore, IFN-γ stimulation induces upregulation of MHC-II expression by LECs, which may support the immunosuppressive activity of regulatory T cells (Tregs) (12, 13), but to date, this has not been shown to play a role in activating CD4⁺ T cells; rather, our laboratory and others have found that MHC-II–expressing LECs do not directly activate naive CD4⁺ T cells or even memory CD4⁺ T cells (T_Ms). In contrast, human MHC-II–expressing blood endothelial cells (BECs) have been shown to reactivate allogenic CD4⁺ T_Ms (14, 15), which can be driven either by TCR-dependent recognition of allopeptides presented in the context of the allogenic MHC-II (16, 17). In contrast with BECs, it has been reported that human MHC-II–expressing LECs are unable to reactivate allogeneic T_Ms (18).

TGF-β has been shown to regulate vital cellular processes, such as proliferation, migration, differentiation, and apoptosis, and therefore plays an important role in the maintenance of homeostasis (19). TGF-β is secreted in an inactive, latent form in which mature TGF-β is noncovalently bound to a latency-associated peptide (LAP) associated with a large latent TGF-β–binding protein or with the transmembrane protein GARP (glycoprotein A repetitions predominant). The release of active TGF-β from the latent complex is tightly controlled and molecules such as proteases or integrins have been shown to activate TGF-β (20). Once activated, TGF-β signals as a dimer through a receptor complex consisting of two types of serine/threonine kinase receptors: TGF-β receptor type II and TGF-β receptor type I, also known as activin receptor-like kinase (ALK) 5 (21). TGF-β is among the best-studied examples of how TGF-β superfamily members regulate innate and adaptive immune responses (22). TGF-β1 knockout mice suffer from severe immune dysregulation, providing evidence for the immunoregulatory role of TGF-β (23). T cells are among a wide range of immune cells whose biological activity is controlled by TGF-β. TGF-β has been shown to inhibit Th1 and Th2 differentiation (24, 25), impair IL-2 expression and subsequent proliferation of T cells (26), and suppress the cytotoxic activity of CD8⁺ T cells (27). TGF-β–mediated induction of FOXP3 expression in naive CD4⁺CD25⁻ T cells and subsequent conversion into Tregs represents another mechanism through which TGF-β regulates T cell–mediated immune responses and exerts immunosuppressive activity (28). In addition to naive T cells, TGF-β has also been shown to impact the biology of memory T_Ms by modulating their generation, maintenance, and reactivation (29–31).

In this study, we asked whether TGF-β plays a role in limiting the function of MHC-II on LECs, using human allogeneic CD4⁺ T_M reactivation as a model system. We found that human dermal LECs (hdLECs) possess the capacity to reactivate allogeneic human memory CD4⁺ T cells (hCD4⁺ T_Ms), but the extent of LEC-mediated reactivation is tightly controlled by TGF-β. MHC-II–expressing hdLECs reactivated allogeneic hCD4⁺ T_Ms significantly only when TGF-β signaling was inhibited. Moreover, although LECs are considered to lack costimulatory molecules such as CD40, CD80, and CD86, we found that in addition to CD58, the proinflammatory cytokines IFN-γ and TNF-α induce the expression of ICOSL and OX40L by LECs. Altogether, our results provide evidence that hdLECs hitherto seen as one of the maintainers of the peripheral tolerance and negative regulators of immune responses may also function as nonhematopoietic APCs for hCD4⁺ T_Ms.

Both primary and immortalized hdLECs were used for in vitro experiments. Primary neonatal human dermal-derived LECs (hLECs; HMVEC-DLyNeo, CC-2812; Lonza, Walkersville, MD) were cultured in EGM2 Bulletkit medium (Lonza), later referred to as the growing medium up to passage 8. Human telomerase-immortalized dermal LEC line (hiLECs; a gift from Prof. D. Kerjaschki, Medical University of Vienna, Austria) were isolated from a population of telomerase-immortalized human dermal microvascular endothelial cells derived from the foreskin. It has been shown that these cells maintain LEC phenotype and functional features for >30 passages (32). hiLECs were cultured in EGM2 Bulletkit medium up to passage 40. We verified the LEC characteristic of hLECs and hiLECs by flow cytometry analysis using podoplanin and CD31 as LEC markers. We confirmed that both cell lines express CD31 and podoplanin on their cell surface. Although hiLECs and hLECs are not significantly different in terms of CD31 and podoplanin expression, hiLECs are smaller than hLECs (Supplemental Fig. 1A, 1B). If not indicated otherwise, Accutase (Sigma-Aldrich, St. Louis, MO) was used to passage the cells. Both cell types were cultured at 37°C and 5% CO₂ conditions. PCR for mycoplasma 16S rDNA was routinely done to test for the presence of mycoplasma.

hdLECs were seeded on six-well plates (2 × 10⁵/well) and cultured overnight. Then the following cytokines were added to the growing medium: (1) human IFN-γ for 72 h (50 ng/ml; PeproTech, Caranbury, NJ), (2) human TNF-α for 24 h (10 ng/ml; PeproTech), or a combination of both cytokines. Non–cytokine-treated hdLECs served as a control. Cells were collected from the plate either with Accutase or, as in the case of membrane-bound LAP–TGF-β1 analysis, with 3 mM EDTA/PBS and subjected to RT-PCR or flow cytometry where their phenotype was analyzed.

The acid guanidinium thiocyanate-phenol-chloroform extraction method was used to isolate the total RNA from hdLECs (33). One microgram of RNA was reverse transcribed with moloney murine leukemia virus reverse transcriptase (Promega, Madison, WI). Next, obtained cDNA was subjected to quantitative reverse transcription PCR and amplified using FastStart Essential DNA Green Master (Roche, Indianapolis, IN) on the LightCycler 480 System (Roche). The following primers with the annealing temperature of 60°C were used: human GARP forward: 5′-GCTGCACAACACCAAGACAA-3′, reverse: 5′-GCTGATCTCATTGGTGCTCA-3′; human TBP forward: 5′-TAGAAGGCCTTGTGCTCACC-3′, reverse: 5′-GAGCCATTACGTCGTCTTCC-3′. Fold change in the expression of the gene of interest in the cytokine-treated samples relative to the unstimulated controls was quantified using the ΔΔCt relative quantitation method and TBP as reference gene.

hiLECs were plated on a 12-well plate (4 × 10³/well), grown overnight, and then transduced with lentiviral vectors (multiplicity of infection 5) encoding for GARP shRNA (TRCN0000419644) or scramble shRNA (Mission shRNA; Sigma-Aldrich), according to the manufacturer’s protocol. Puromycin (0.5 µg/ml; Thermo Fisher Scientific, Waltham, MA) was used to select stably transduced sublines. The level of GARP silencing was evaluated at mRNA and protein level by RT-PCR and flow cytometry, respectively.

Leukocyte reduction blood filters were obtained from healthy blood donors (University of Chicago Medicine Blood Donation Center) and PBMCs were isolated by a density gradient centrifugation using a lymphocyte separation medium (Lonza) according to the manufacturer’s protocol. Freshly isolated PBMCs were suspended in PBS/2% FBS/1 mM EDTA at the density 5 × 10⁷/ml and subjected to T_M isolation or CD14⁺ monocyte isolation.

Allogeneic CD45RO⁺CD45RA⁻ T_Ms, later referred to as hCD4⁺ T_Ms, were isolated from freshly isolated PBMCs by negative selection using EasySep Human Memory CD4⁺ T cell enrichment kit (Stemcell Technologies, Vancouver, BC, Canada) according to the manufacturer’s protocol. The purity of isolated cells was routinely checked and ranged from 91 to 96%. Prior to coculture with hdLECs or mature DCs (mDCs), hCD4⁺ T_Ms were washed twice with PBS and labeled with CFSE (1 µM; Invitrogen, Thermo Fisher Scientific, Waltham, MA) according to the manufacturer’s protocol.

Human DCs were generated from CD14⁺ monocytes. CD14⁺ monocytes were isolated from the freshly isolated PBMCs by positive selection using The EasySep Human CD14 Positive Selection Kit II (Stemcell Technologies) according to the manufacturer’s protocol. Isolated CD14⁺ monocytes were cultured for 5 d in RPMI-1040 supplemented with 10% FBS, 2 mM l-glutamine, 1 mM sodium pyruvate, 0.0012% (v/v) 2-ME (all from Life Technologies, Waltham, MA), and 500 U IL-4 and 1000 U GM-CSF (each from PeproTech). To obtain mDCs, we stimulated cells with 10 ng/ml LPS (Sigma-Aldrich) for 1 additional day.

Prior to coculture with hCD4⁺ T_Ms, hdLECs were plated in 96-well plates at 8000 cells/well in triplicates in EGM2 medium. To induce MHC-II expression, we treated hdLECs with 50 ng/ml IFN-γ (PeproTech) for 72 h. In some wells, 10 ng/ml TNF-α (PeproTech) was added for the last 24 h of the culture. Just before the start of the coculture, hdLECs were washed twice with PBS, and 100 µl of the coculture medium was added [RPMI-1040 with 10% FBS, 2 mM l-glutamine, 1 mM sodium pyruvate, 0.0012% (v/v) 2-ME, and 1% penicillin/streptomycin; all from Life Technologies]. At the same time, LPS-matured DCs, which served as a control of hCD4⁺ T_M reactivation, were washed twice with PBS and then plated on a 96-well plate at 10,000 cells/well in triplicates in 100 µl coculture medium. Subsequently, CFSE-labeled hCD4⁺ T_Ms suspended in the 100 µl of the coculture medium were added to the hdLECs or mDCs at 1 × 10⁵ cells/well. ALK4, ALK5, and ALK7 inhibitor SB431542, later referred to as ALK4/5/7 inh, or ALK5 inhibitor LY364947, later referred to as ALK5 inh (Sigma-Aldrich), both at a final concentration of 3 µM, were used to block the TGF-β signaling pathway and were present throughout the coculture. Coculture not treated by TGF-β receptor inhibitors, later referred to as control, was enriched with the solvent for TGF-β receptor inhibitors, DMSO. The volume of added DMSO was equal to the volume of added inhibitors. For human ICOSL, OX40L, or MHC-II blocking experiments, hdLECs were preincubated for 30 min with the neutralizing Abs or their isotype controls before CFSE-labeled hCD4⁺ T_Ms were added to the coculture. The following neutralizing Abs or their respective isotype controls were used: (1) anti-human ICOSL (clone 136716; R&D Systems, Minneapolis, MN)/mouse IgG2b; (2) anti-human OX40L (clone 159403; R&D Systems)/mouse IgG1; and (3) anti-human HLA-DR, DP, DQ (Tü39; BioLegend, San Diego, CA) (34)/mouse IgG2aκ in 100 µl of coculture medium. Blocking Abs or their isotype controls were present throughout the coculture time at a final concentration of 10 µg/ml. hCD4⁺ T_Ms were cocultured with hdLECs or mDCs for 5 d. hCD4⁺ T_Ms cultured without hdLECs or mDCs were used as a control for each condition. Next, hCD4⁺ T_Ms were collected from the coculture plate, the supernatant was saved for the analysis of expression of different cytokines, and cells were washed with PBS and subjected to flow cytometry, where their phenotype was analyzed. Analysis of CFSE dilution was used to assess the proliferation of hCD4⁺ T_Ms. hCD4⁺ T_Ms with decreased CFSE intensity were considered as proliferating cells, and their percentage was analyzed by flow cytometry.

The concentrations of Th cytokines in the supernatant from the hdLEC– or mDC–hCD4⁺ T_M cocultures were analyzed using Legendplex Th cytokine kit (BioLegend) according to the manufacture’s protocol. A total of 50 µl of the coculture medium was used for the analysis. Samples were acquired on the LSR Fortessa flow cytometer (BD Bioscience) and analyzed using Legendplex data analysis software (BioLegend).

Cells were washed 2× with PBS and incubated at 4°C for 15 min with a fixable live/dead marker (Life Technologies), and then cell-surface markers were stained for 20 min at 4°C with the fluorochrome-conjugated Abs in PBS with 5% FBS. After surface staining cells were fixed and permeabilized using FOXP3 Fix/Perm Buffer Set (BioLegend), the intracellular Ags were stained for 30 min at 4°C with the appropriate fluorochrome-conjugated Abs in permeabilization buffer enriched with 5% FBS. The following anti-human Abs were used, all from BioLegend: CD4-Brilliant Violet 785 (clone RM4-5), CD45RO-allophycocyanin-Cy7 (clone UCHL1), CD25-PE-Cy7(clone BC96), FOXP3-PE (clone 259D), CD80-FITC (clone 2D10), CD86-allophycocyanin (clone BU63), OX40L-PE (clone 11C3.1), GARP-PE (clone 7B11), LAP-allophycocyanin (clone TW4-2F8), HLA-DR-allophycocyanin-Cy7 (clone L243), CD106-PE (clone STA), CD54-allophycocyanin (clone HCD54), and HLA-DM (clone MaPDM1), as well as anti-ICOSL-allophycocyanin (clone MIH12; eBioscience). Flow cytometry was performed on a LSR Fortessa flow cytometer (BD Bioscience) and analyzed with FlowJo software v10.7.1 (Becton, Dickinson and Company, Ashland, OR).

Statistical analysis was done using Prism 6 software (GraphPad, San Diego, CA). Paired two-tailed Student t test was used for the comparison of two groups, whereas one-way ANOVA or two-way ANOVA with Bonferroni test for multiple comparisons correction were used for the comparison of more than two groups. Data are presented as mean ± SD unless otherwise noted; p < 0.05 was considered as statistically significant.

First, we confirmed that IFN-γ could induce the expression of both HLA-DR and HLA-DM in both primary (hLECs) and immortalized (hiLECs) hdLECs (Fig. 1A). Thus, not only mouse LECs (12, 35) but also hdLECs express HLA-DM, which is necessary for efficient Ag presentation in the context of MHC-II (36). Because TNF-α is needed for the upregulation of some costimulatory molecules, we added TNF-α in addition to IFN-γ to activate hdLECs; in those conditions, the TNF-α was added for the last 24 h because longer incubation times led to decreased levels of IFN-γ–driven MHC-II expression (Supplemental Fig. 1C), consistent with previous reports (37, 38).

We next employed 5-d cocultures of CFSE-labeled allogenic hCD4⁺ T_Ms isolated from healthy blood donors with cytokine-pretreated LECs (or mDCs as controls). As expected, hCD4⁺ T_Ms robustly proliferated when cocultured with LPS-matured hDCs (Fig. 1B–D), where the expression of costimulatory molecules CD83 and CD86 was highly upregulated (Supplemental Fig. 1D). In contrast, a very low percentage of allogeneic hCD4⁺ T_Ms (1–9% depending on the donor) responded to MHC-II–expressing (IFN-γ–pretreated) hdLECs by dividing (Fig. 1B–D). Proliferation of hCD4⁺ T_Ms when cocultured with non–cytokine-pretreated LECs was similarly negligible (≤1%) as when left in monoculture (no APCs); because 1–3% of freshly isolated hCD4⁺ T_Ms expressed the activation marker CD25⁺ (data now shown), this suggested that any proliferation in coculture conditions with non–cytokine-pretreated LECs resulted from the presence of already activated hCD4⁺ T_Ms in the pool of freshly isolated cells.

LECs are known to express immunosuppressive molecules, including PD-L1, NO, and IDO (3, 8, 18). Given this, we hypothesized that MHC-II–expressing hdLECs might be able to provide sufficient activation signals to hCD4⁺ T_Ms, but that these could be suppressed by LEC-expressed inhibitory molecules. However, we found that even when blocking the biological activity of these inhibitory molecules, hCD4⁺ T_M proliferation did not appreciably increase after LEC coculture (data not shown). Thus, we turned our attention to TGF-β, which has been shown to impact the biology of memory T_Ms by modulating their generation, maintenance, and reactivation (29–31). To block TGF-β signaling, we used either SB431542, a broad inhibitor of type I receptors (ALK4, ALK5, and ALK7, later referred to as ALK4/5/7 inh) (39), or LY364947, an inhibitor of TGF-β receptor type I (ALK5, later referred to as ALK5 inh).

Interestingly, we found that coculture of hCD4⁺ T_Ms with cytokine-activated hdLECs in the presence of TGF-β signaling inhibitors led to proliferation of allogeneic hCD4⁺ T_Ms (Fig. 1B–D). Depending on the donor, the percentages of dividing (CFSE^low) hCD4⁺ T_Ms after coculture with stimulated, TGF-β–inhibited hiLECs ranged from 5 to 23% of the total cells, and based on CFSE levels to determine generation, this corresponded to 1–7% of the original hCD4⁺ T_Ms that responded by dividing. We note that primary LECs (hLEC) induced slightly lower proliferation responses than did hiLECs, which is not surprising considering that the two sources of hdLECs would not be expected to have identical expression patterns of the numerous factors that influence T cell proliferation. The LEC-induced hCD4⁺ T_M proliferation was dependent on MHC-II expression, because non–cytokine-pretreated LECs failed to induce hCD4⁺ T_M proliferation despite the presence of TGF-β signaling inhibitors. To further confirm this, we added MHC-II blocking Abs to the coculture and found significant inhibition of LEC-induced CD4⁺ T cell proliferation (Fig. 1E). This result ruled out the possibility of bystander hCD4⁺ T_M activation via LEC-expressed soluble factors.

To verify the activated state of hdLEC-experienced hCD4⁺ T_Ms, we analyzed the phenotype of CD4⁺ T cells after the hdLEC coculture and found that the late activation marker CD25 was expressed by the majority of proliferating FOXP3⁻ hCD4⁺ T_Ms (Fig. 1F). However, this expression was lower in hdLEC-reactivated hCD4⁺ T_Ms compared with those reactivated by mDCs. To further assess the functional state of proliferating CD4⁺ T cells, we assessed cytokine levels in the media collected at the end of coculture. We found increased levels of IFN-γ, IL-5, IL-10, IL-13, and IL-22 when hCD4⁺ T_Ms were activated by cytokine-pretreated hdLECs in the presence of TGF-β inhibitors, whereas in the absence of TGF-β inhibitors, cytokine levels were low or undetectable (Fig. 1G, 1H). This was consistent with the proliferation data (Fig. 1B–D), as was the lower levels of IFN-γ in hLEC versus hiLEC cultures. Furthermore, none of these cytokines could be detected in monocultures of hCD4⁺ T_Ms or in cocultures with non–cytokine-pretreated (MHC-II⁻) hdLECs.

Taken together, these results indicate that MHC-II–expressing hdLECs have the ability to reactivate allogeneic hCD4⁺ T_Ms, but that TGF-β signaling largely prevents such reactivation.

LECs are known to express two isoforms of TGF-β: TGF-β1 and TGF-β2 (40). However, it is unknown whether the latent forms of these isoforms (LAP–TGF-βs) are secreted predominantly into the surrounding microenvironment in a complex with latent TGF-β–binding protein or whether LECs also display the latent form of TGF-β1 on their surface. Given that membrane protein GARP is involved in tethering LAP–TGF-β1 to the cell membrane and participates in TGF-β activation (41, 42), we first asked whether this protein is expressed on the surface of hdLECs. To address this, hdLECs (immortalized and primary) were either left untreated or stimulated with IFN-γ, TNF-α, or a combination of both cytokines. We found that hdLECs express GARP both on the mRNA and the protein level, and that this expression was upregulated by IFN-γ and TNF-α (Fig. 2A–C). Furthermore, LAP–TGF-β1 was also displayed on the surface of hdLECs, and its expression levels were increased by IFN-γ and TNF-α stimulation (Fig. 2D), consistent with the cytokine-induced increased GARP expression. We then silenced GARP expression in hiLECs and found that this significantly decreased membrane-bound LAP–TGF-β1 (Fig. 2E, 2F), even though the expression of TGF-β1 mRNA in GARP-silenced LECs was unchanged (data not shown). Together, these data verify that hdLECs tether LAP–TGF-β1 to the cell surface via GARP.

Given this, we next asked whether membrane-bound LAP–TGF-β1 could account for the inhibition of LEC-mediated activation of hCD4⁺ T_M, or whether the interaction between hdLECs and hCD4⁺ T_M was regulated by TGF-β present in the extracellular space. To address this, we cocultured hCD4⁺ T_M with GARP-silenced hiLECs that had been pretreated with IFN-γ. Interestingly, GARP-silenced hiLECs could significantly activate hCD4⁺ T_M proliferation, and the addition of ALK4/5/7 inhibitor further increased this effect (Fig. 2G). Altogether, these results demonstrate that LAP–TGF-β1 is bound to the membrane of hdLECs via GARP and inhibits LEC-mediated activation of hCD4⁺ T_Ms.

It has been shown that lymph node stromal cells (LNSCs) may provide promaintenance signals to FOXP3⁺ Treg cells (35), so we asked whether hdLECs might impact the survival of memory Tregs. To this end, we evaluated the percentages of Tregs (Foxp3⁺CD25⁺) within the nonproliferating (CFSE^high) populations compared with those in freshly isolated hCD4⁺ T_Ms. In control conditions (no APCs), Treg percentages declined among the nonproliferating hCD4⁺ T_Ms, whereas coculture with either hdLECs or DCs prevented that loss (Fig. 3A, 3B). This general phenomenon was not MHC-II or TGF-β driven, because all coculture conditions maintained Treg percentages within the nonproliferating populations at levels similar to those within freshly isolated hCD4⁺ T_Ms. Furthermore, although the Treg fractions were not statistically different between the DC- versus LEC-experienced nonproliferating hCD4⁺ T_Ms, we noticed that LEC-experienced Tregs had higher levels of FOXP3 expression (assessed as median fluorescence intensity) as compared with DC-experienced and freshly isolated Treg populations (Fig. 3C), particularly in hiLEC cocultures. As with Treg percentages, FOXP3 expression levels were similar regardless of cytokine pretreatment or TGF-β inhibition.

Next, we asked whether the reactivation of Tregs by MHC-II–expressing hdLECs is regulated by TGF-β, and we turned our attention to the proliferating populations (CFSE^low) of hCD4⁺ T_Ms. Interestingly, TGF-β inhibition did not affect Treg fractions in LEC-activated hCD4⁺ T_Ms, but it did significantly decrease Treg fractions in DC-activated cells (Fig. 3D). Furthermore, when comparing the ratio of Treg fractions in proliferating versus unproliferating populations, this ratio was significantly decreased by TGF-β inhibitors only in DC-activated hCD4⁺ T_Ms (Fig. 3E). This indicates that Tregs are activated to a similar degree as non-Tregs by hdLECs when TGF-β signaling is blocked.

These results suggest that in addition to acting as nonprofessional APCs for hCD4⁺ T_Ms, including Tregs, hdLECs may also support the maintenance of Tregs and increase the expression of FOXP3 within the population of nonproliferating Tregs. However, the functional relevance of this phenomenon and the underlying mechanisms remain to be elucidated.

To better understand the mechanisms that drive LEC-mediated reactivation of hCD4⁺ T_Ms, we asked whether hdLECs support hCD4⁺ T_M by other costimulatory molecules, such as ICOSL and OX40L, because these have been shown to support the activation of hCD4⁺ T_Ms by MHC-II–expressing BECs (43, 44). Using flow cytometry, we analyzed these together with CD80 and CD86 on both hdLEC populations. We found low expression levels of these costimulatory molecules under resting conditions (Fig. 4A, 4B). IFN-γ had no effect on any of their expression levels except for OX40L, which showed a small but statistically insignificant increase (Fig. 4B). In contrast, TNF-α caused potent upregulation of ICOSL on hdLECs, consistent with its known activation of ICOSL on APCs (45). TNF-α also caused a moderate increase in OX40L expression but did not induce CD80 or CD86 (Fig. 4A, 4B). It is worth noting that the TNF-α–upregulated levels of ICOSL and OX40L in hdLECs was similar in HUVECs (data not shown). Furthermore, the expression levels of ICOSL and OX40L on hdLECs were unaffected by ALK4/5/7 inhibition (Fig. 4B), suggesting that TGF-β signaling does not affect ICOSL and OX40L expression on LECs. Finally, because TGF-β may regulate the expression of MHC-II (46) along with two adhesion molecules, ICAM-1 (47) and VCAM-1 (48), we analyzed hdLEC expression of these by flow cytometry. We found that TGF-β inhibition did not alter their levels (Fig. 4C), neither in steady-state conditions nor after their upregulation by cytokine stimulation (Fig. 4C).

Finally, we assessed whether blocking ICOSL or OX40L on hdLECs affected their capability to reactivate memory T cells by supplementing hiLEC-hCD4⁺ T_M cocultures with blocking Abs against ICOSL or OX40L. We found that neither ICOSL nor OX40L blockade exerted any inhibitory effect on the LEC-induced proliferation of hCD4⁺ T_Ms (Fig. 4D, left). Moreover, we did not notice any alteration in the production of IFN-γ (Fig. 4D, right) or in hCD4⁺ T_M viability (data not shown).

These results demonstrate that even though hdLECs lack the major costimulatory molecules CD80 and CD86, they can upregulate other costimulatory molecules in response to proinflammatory cytokines, including ICOSL and OX40L, as do other nonprofessional APCs, including BECs. However, ICOSL and OX40L do not seem to be responsible for LEC-mediated hCD4⁺ T_M activation when TGF-β is blocked.

Robust recall of T_Ms requires Ag presentation by DCs in the context of MHC-II concurrent with CD28 activation by its ligands CD80 and CD86 (49). Nonhematopoietic cells, such as fibroblasts, epithelial cells, endothelial cells, and hepatocytes, can also induce expression of MHC-II upon IFN-γ stimulation (50–52), and these cells have been shown to present exogenous Ags in the context of the MHC-II (12, 50, 53). Furthermore, although they are unable to reactivate allogeneic naive CD4⁺ T cells because of the lack of CD80 and CD86 expression, BECs can reactivate allogeneic hCD4⁺ T_Ms (14). LECs, of interest in this study, also upregulate MHC-II upon IFN-γ stimulation and furthermore can acquire peptide-loaded MHC-II complexes from DCs (11). To date, however, most evidence has suggested that MHC-II expression on LECs can drive only quiescence, anergy, or tolerance of CD4⁺ T cells (6–8, 10–13). In contrast, our results in this article demonstrate that MHC-II–expressing LECs can in fact drive CD4⁺ T_M proliferation, but only in a substantial way when LEC-expressed TGF-β is inhibited, at least in the context of allogenic hCD4⁺ T_Ms.

To compensate for the lack of CD80/CD86 expression, BECs and LECs express other costimulatory molecules, including CD58 (54), which can play a significant role in the activation of CD28⁻CD8⁺ T cells (55); despite this, IFN-γ–pretreated hdLECs initiated only minimal proliferation of allogeneic hCD4⁺ T_Ms. Given the suppressive role of TGF-β in T cell activity (56, 57), we hypothesized that TGF-β signaling by LECs could be inhibiting their ability to activate allogenic hCD4⁺ T_M proliferation. Indeed, we found that either inhibition of TGF-β signaling or silencing of GARP, which tethers TGF-β to the cell membrane, significantly and substantially increased hCD4⁺ T_M proliferation by LECs. Unsurprisingly, this degree of proliferation was less than that induced by mDCs, which express high levels of CD80 and CD86, but it was still far greater than in the absence of TGF-β inhibition.

TGF-β is known to help maintain CD4⁺ T cells in a resting state and can suppress their activation, especially when presented by APCs with lower levels of Ag presentation or that lack strong costimulatory signals (58). Moreover, it has been shown that the proliferation of Ag-specific hCD4⁺ T_Ms may be efficiently inhibited by TGF-β (59), particularly in cells with a Th1 phenotype. Strong activation signals such as those triggered by potent TCR stimulation may release some of the memory Th1 cells from TGF-β–mediated inhibition (60). We speculate that the lower level of costimulation provided by MHC-II–expressing hdLECs because of the lack of CD80/CD86 expression, despite their expression of ICOSL and OX40L, may not overcome the inhibitory signals triggered by TGF-β in hdLEC-experienced allogeneic hCD4⁺ T_Ms enough to induce substantial proliferation.

With regard to Tregs in particular, LECs have been shown to modulate their biology in subtle ways. For example, it has been reported that mouse LNSCs can support the maintenance of Tregs and may promote their conversion from naive cells in an MHC-II– and IL-2–dependent manner (35, 61). In vivo, reduced expression of MHC-II in LNSCs, which include LECs, resulted in reduced Tregs and increased incidence of autoimmune reactions (12). In tumor-bearing mice, LEC-specific depletion of MHC-II led to decreases in intratumoral Tregs, both in numbers and suppressive functions (13). Our results in this article add further insight by suggesting that hdLECs may support the maintenance of CD25⁺ Tregs among hCD4⁺ T_Ms, yet apparently in an MHC-II– and TGF-β–independent manner. Thus, although TGF-β in general can induce FOXP3 expression in naive T cells and help maintain FOXP3 levels in peripheral Tregs (28, 62), the mechanisms by which LECs promote Treg maintenance may involve other pathways that are independent of TGF-β signaling.

In conclusion, these results implicate TGF-β as an important factor that shapes the outcome of LEC-CD4⁺ T_M interactions by restraining CD4⁺ T_M activation. Currently, different TGF-β pathway inhibitors are being tested in preclinical oncology models and clinical trials (63). Our findings suggest that blockade of TGF-β signaling could potentially, among other biological effects, also unleash LEC-mediated activation of CD4⁺ T_Ms and allow LECs to trigger activation signals within a broader population of CD4⁺ T_Ms.

The authors have no financial conflicts of interest.

We thank Marcin Kwissa for providing helpful technical guidance and valuable discussions and suggestions, Yue Wang and Suzana Gomes for technical assistance, and the University of Chicago Medicine Blood Donation Center and blood donors.