Vascular endothelial cells (EC) are an exposed target tissue in the course of CTL-mediated alloimmune diseases such as graft-vs-host disease (GVHD) or solid organ transplant rejection. The outcome of an interaction between CTL and target cells is determined by the amount of Ag presented and the costimulatory signals delivered by the target cells. We compared human EC with leukocytes and epithelial cells as targets for peptide-specific, MHC class I-restricted CTL clones. EC were poor targets for immunodominant CTL. Both endogenously processed antigenic proteins and exogenously added antigenic peptides are presented at 50- to 5000-fold lower levels on EC compared with any other target cell analyzed. This quantitative difference fully explained the poor CTL-mediated killing of EC. There was no evidence that lack of costimulation would contribute significantly to this cell type-specific difference in CTL activation. An HLA-A2-specific CTL clone that killed a broad selection of HLA A2-positive target cells equally well, killed EC less efficiently. Our data suggest that EC present a different Ag repertoire compared with other cell types. By this mechanism, these cells may escape an attack by effector CTL, which have been educated by professional APCs and are specific for immunodominant antigenic peptides.
Vascular endothelial cells (EC)3 are among the first peripheral cells that present Ag to alloreactive T lymphocytes after allogeneic stem cell transplantation. CD8+ CTL are major effector cells of tissue injury in graft-vs-host disease (GVHD) (1, 2). Effector CTL, which are activated by an MHC-peptide complex, rapidly degranulate and deliver a lethal hit to target cells. In this context, it is intriguing that vascular EC, which express alloantigen, are not rapidly killed by transmigrating, Ag-specific effector CTL (3, 4). During acute GVHD, mild, immune-mediated endothelial injury was described in the skin (5) and gut (6). We found significant cutaneous microvessel loss only late in the course of persistent chronic GVHD (7). The question arises of whether EC present the same amount and the same spectrum of antigenic peptides as leukocytes (including professional APCs) or epithelial cells (e.g., colon epithelial cells or keratinocytes), which are important target cells during acute GVHD.
GVHD after allogeneic stem cell transplantation from HLA-identical donors typically involves allorecognition of minor histocompatibility Ags. Minor histocompatibility Ags are peptides derived from proteins that are encoded by polymorphic or sex-specific genes (8). They are presented by the recipient’s MHC class I molecules and can elicit vigorous donor T cell responses (9).
The amount of MHC class I-bound peptide presented to CTL critically determines the type of CTL activation; CTL-mediated target cell lysis requires the lowest, cytokine secretion an intermediate, and CTL proliferation the highest amount of antigenic peptide presented by target cells (10). Activating or inhibiting costimulatory signals provided by target cells may further enhance or prevent CTL activation (11, 12).
In this study we compared EC with leukocytes and epithelial cells as targets of Ag-specific MHC class I-restricted CTL. We measured target cell lysis, esterase release by CTL, and the secretion of IFN-γ. We chose two immunodominant, high affinity, HLA-A2-restricted peptides: the Y chromosome-encoded, male-specific minor histocompatibility Ag FIDSYICQV (SMCY) (13), and the influenza matrix protein-derived peptide GILGFVFTL (flu) (14). We used male, HLA-A2-positive cells to investigate processing and presentation of endogenous Ag and assessed the amount of Ag required to trigger CTL-mediated EC lysis in peptide titration experiments. We found that EC are poor targets for peptide-specific CTL. This observation was fully explained by the impaired capacity of EC to present immunodominant antigenic peptides.
Materials and Methods
All studies involving primary human cell lines and tissues were approved by the institutional ethical review board. HUVEC were isolated from umbilical cords by enzymatic digestion as described previously (15). EC were cultured in complete medium 199 containing 20% FCS, 2 mM l-glutamine, 100 U/ml penicillin, and 100 μg/ml streptomycin (all from Invitrogen Life Technologies), supplemented with fibroblast growth factors (20 ng/ml human acidic fibroblast growth factor and 20 ng/ml human basic fibroblast growth factor; both from PeproTech) and heparin (0.2 mg/ml; Sigma-Aldrich). Human dermal microvascular EC were isolated and cultured as described previously (16). EBV-immortalized B lymphoblastoid cells (BLC) were grown from cord blood mononuclear cells obtained from the same donors as the EC (17). BLC were cultured in complete RPMI 1640 (Invitrogen Life Technologies) containing 10% FCS, 2 mM l-glutamine, 100 U/ml penicillin, and 100 μg/ml streptomycin.
The HLA-A2-positive, male EBV-transformed B cell line JY, the HLA-A2-positive, female, TAP-deficient T2 cell line (both gifts from A. Cerny, University of Bern, Bern, Switzerland) and the HLA-A2-positive, male and female colon carcinoma cell lines (CC; provided by G. Spagnoli, University of Basel, Basel, Switzerland) were cultured in complete RPMI 1640.
Peptides, cytokines, and Abs
The HLA-A2-restricted minor histocompatibility Ag derived from the protein SMCY (FIDSYICQV, SMCY) (13, 18) and the HLA-A2-restricted peptide derived from influenza virus matrix protein (GILGFVFTL, flu) (14, 19) were obtained from American Peptides. Peptides were dissolved in DMSO at 10 mg/ml and stored at 4°C. Human IFN-γ was reconstituted at 100 μg/ml, and human TNF (both from PeproTech) was reconstituted at 50 μg/ml; both cytokines were stored in aliquots at −70°C. To activate resting EC, the cells were incubated overnight with IFN-γ (final concentration, 100 ng/ml) and TNF (final concentration, 50 ng/ml). For FACS analysis, the following Abs were used: BB7.2 and K1616 (gifts from P. Cresswell, Yale University Medical School, New Haven, CT), FITC-conjugated goat anti-mouse IgG(H+L) (Southern Biotechnology Associates), monomeric Fab M1-D12 and G9-280-2F1 and FITC-conjugated goat anti-human IgG Fab (Jackson ImmunoResearch Laboratories).
Generation of Ag-specific CTL clones
Peptide-specific, HLA-A2-restricted CTL clones were generated from PBMC according to the method described by Cerny et al. (20). Buffy coats were obtained from healthy, HLA-A2-positive blood donors (Blutspendezentrum SRK), PBMC were isolated by Ficoll gradient (Lymphoprep; Axis-Shield) and stored in liquid nitrogen. Freshly thawed PBMC were incubated for 2 h at 37°C with the peptide of interest (10−5 M) in complete RPMI 1640-AB containing 10% heat-inactivated human AB serum (Blutspendezentrum SRK) instead of FCS. Cells were plated at 4 × 106 cells/ml in a 24-well plate in complete RPMI 1640-AB. After 3 days, medium was supplemented with rIL-2 (proleukin; final concentration, 50 U/ml; gift from Roche). On day 7 and weekly thereafter, cultures were restimulated with peptide-pulsed, irradiated (3000 rad) autologous feeder cells (106/ml). After 5–6 wk, the CTL lines were tested for peptide-specific lysis in a calcein-release cytotoxicity assay. Primary cultures displaying a peptide specific lysis of >40% were cloned by limiting dilution as described previously (21). Cells were suspended in cloning medium (complete RPMI 1640 supplemented with 5 mM nonessential amino acids, 5 mM sodium pyruvate (both from Invitrogen Life Technologies), and 200 U/ml IL-2) containing 500,000 irradiated (3000 rad) autologous feeder cells/ml and 20,000 mitomycin C-treated (Roche) JY cells/ml and plated in 96-well, round-bottom plates at 100, 10, and 1 cells/well. After 1 wk, the cultures were fed 100 μl/well cloning medium. Microcultures that grew at the frequency conforming to clonal growth were expanded and tested for peptide-specific lysis. Every 2–3 wk, the CTL clones were restimulated with cloning medium containing feeder cells. HLA-A2-specific CTL lines were generated by incubating HLA-A2-negative PBMC with the HLA-A2-positive cell line JY. Cultures were restimulated weekly with mitomycin C-treated JY cells and tested for HLA-A2-restricted killing using T2-, JY-, and HLA-A2-negative BLC as targets. Cell lines that showed >40% specific killing were cloned as described above.
Target cell lysis was measured by a calcein release assay as described previously (17). Adherent target cells (EC and CC) grown to confluence in 96-well, flat-bottom plates were loaded for 30 min at 37°C with 20 μM calcein-AM (Molecular Probes) in medium 199. Then, complete medium 199 containing peptides at the indicated concentrations was added for another 4–8 h. Cells were washed twice with assay medium (medium 199, 2% FCS, 5 mM HEPES, 2 mM l-glutamine, 100 U/ml penicillin, and 100 μg/ml streptomycin). Target cells grown in suspension were incubated for 30 min at 37°C with 10 μM calcein-AM. Cells were then washed and incubated for the same time as the adherent target cells in complete RPMI 1640 containing peptides at the indicated concentrations. After washing twice, target cells were counted and transferred to 96-well, round-bottom plates at 104 cells/well. CTL were washed, counted, and added to the calcein-loaded target cells (final volume, 200 μl/well). Spontaneous release (SR) was determined by adding assay medium; maximum release (MR) was determined by the addition of lysis buffer (50 mM sodium borate in 0.1% Triton X-100 (both from Sigma-Aldrich), pH 9.0). After 2 h, 75 μl of supernatant was carefully removed and transferred to a 96-well, flat-bottom plate. Released calcein was measured in a fluorescence multiwell plate reader (SpectraMax GeminiXS; Molecular Devices; excitation wavelength, 485 nm; emission wavelength, 538 nm). The percent specific lysis was calculated as (sample release − SR)/(MR − SR) × 100%. SR in all experiments was <20%.
IFN-γ secretion by CTL was measured in the supernatant of the cytotoxicity assay. After harvesting supernatants for the cytotoxicity assay, the medium was replaced, and the cells incubated at 37°C overnight. At the highest E:T cell ratio, 150 μl of supernatant/well was collected, pooled, and stored frozen at −70°C. IFN-γ was measured in a sandwich ELISA using a commercially available Ab pair to detect human IFN-γ (monoclonal mouse anti-human IFN-γ (MAB285) and biotinylated, polyclonal goat anti-human IFN-γ Ab (BAF285; both from R&D Systems) according to the manufacturer’s instructions.
Esterase release assay
Esterase release by CTL was measured in a CTL assay essentially as described above, but using serum-free assay medium throughout the procedure. After 2 h of incubation of CTL with target cells at 37°C, granzyme A activity was measured using N-α-benzyloxycarbonyl-l-lysine-thiobenzylester (Calbiochem) as a substrate (22). Twenty microliters of cell-free supernatant was harvested and mixed with 180 μl of substrate solution (0.2 mM N-α-benzyloxycarbonyl-l-lysine-thiobenzylester and 0.22 mM 5,5′-dithio-bis(2-nitrobenzoic acid) (Sigma-Aldrich) in PBS) in 96-well, flat-bottom plates. After 1 h of incubation at 37°C, OD at 405 nm (OD405) was measured using an ELISA plate reader (SpectraMax 190; Molecular Devices). SR was determined by adding assay medium without target cells; MR was determined by adding lysis buffer to the CTL. The percent specific esterase release was calculated as (OD405 of sample − OD405 of SR)/(OD405 MR − OD405 SR) × 100%.
Cell lines were tested for HLA-A2 expression using the allele-specific mAb BB7.2 or the nonbinding isotype control, K1616. All staining procedures were conducted on ice. Cells (5 × 104/sample) were washed once with PBS/1% BSA and incubated with the first Ab for 30 min at 4°C. Cells were washed with PBS/BSA and incubated with FITC-conjugated goat-anti-mouse IgG(H+L) for 30 min. Cells were washed again, fixed in 2% paraformaldehyde in PBS, and analyzed (FACScan; BD Biosciences).
Binding of GILGFVFTL to HLA-A2 was analyzed using the flu/HLA-A2-specific monoclonal, monomeric Fab M1-D12 (23). The melanoma differentiation Ag gp100-specific Fab G9-280-2F1 was used as a control. Cells (105 /sample) were treated using the staining protocol outlined above, but with FITC-conjugated goat anti-human IgG Fab as a secondary reagent.
Total RNA was isolated from 2 × 106 target cells using Tri-Reagent (Molecular Research Center) according to the manufacturer’s protocol. After RT (Superscript; Invitrogen Life Technologies) cDNA coding for SMCY was amplified using the following primers: 5′-CTGTTACGGTGAAGGATGAG-3′ and 5′-CTCTGCAAACTGTACTCCTG-3′ (Microsynth). GAPDH cDNA was amplified using the following primers: 5′-ATGGGGAAGGTGAAGGTCGG-3′ and 5′-AGGGATGATGTTCTGGAGAG-3′ (Microsynth). PCR was performed with an initial denaturation step at 96°C for 5 min, then 35 cycles with 30-s denaturation at 96°C, 30-s annealing at 54°C, 1-min elongation at 72°C, followed by the final extension for 10 min at 72°C.
Poor killing of male EC by SMCY-specific CTL
We compared male vascular EC with BLC from the same donor as targets for SMCY-specific, HLA-A2-restricted CTL. SMCY-specific CTL killed male BLC, but failed to lyse male EC (Fig. 1,A). Female or HLA-A2-negative target cells were not killed by this CTL clone (Fig. 1,A). To rule out that EC do not express the antigenic protein, we first confirmed that male EC express the SMCY gene. All male cells, but no female target cell line, transcribed the gene coding for the protein that contains the H-Y minor histocompatibility antigenic peptide FIDSYICQV used in this study (Fig. 1, B and C). Resting EC express low levels of HLA class I molecules (24). Therefore, low levels of the Ag-presenting molecule HLA-A2 could explain the poor killing of male EC by CTL. To up-regulate HLA-A2 expression, EC were incubated overnight with IFN-γ (100 ng/ml) and TNF (50 ng/ml). This treatment led to a substantial increase in surface HLA-A2 expression (Fig. 1,D). Cytokine-treated male EC became susceptible for lysis by SMCY-specific CTL (Fig. 1,A). To compare the CTL-mediated killing of male target cells, specific lysis was normalized to the maximal specific lysis obtained by pulsing target cells with a high SMCY peptide concentration. At a peptide concentration of 10−4 M, male EC were killed to 32%, male colon cancer cells to 19%, male BLC to 38%, and male PBMC to 29%. For each male target cell line, this maximal killing level was set at 100% for the representation of normalized specific lysis. This maximal lysis was determined in parallel using the same effector cells as those used for assessment of specific lysis of unpulsed male targets. However, male EC were killed less efficiently than male BLC. Using the same CTL clone at the same E:T cell ratio, male EC were the least susceptible target cell line compared with BLC, CC, and fresh PBMC (Fig. 1 E).
Poor killing of peptide-pulsed EC by peptide-specific CTL
To assess Ag-specific EC lysis quantitatively, we used peptide-pulsed cells as targets. CTL activation was examined by measuring target cell lysis, esterase release, and IFN-γ secretion (Fig. 2). EC were compared with epithelial (CC) and a variety of different leukocyte-derived target cells, such as the HLA-A2-positive cell line T2, B lymphoblastoid cell lines from the same donor as the EC (BLC), the HLA-A2-positive B lymphoblastoid cell line JY, and primary PBMC (Fig. 3). T2 cells are known to express empty HLA-A2 molecules on the surface, which might be loaded more easily by exogenous peptides. Therefore, we also used more physiological leukocyte cell lines as targets for comparison with EC, such as primary PBMC and BLC obtained from the same donor as the EC. HLA-A2-restricted CTL clones specific for SMCY (FIDSYICQV) and flu peptide (GILGFVFTL) were used as effector cells. Compared with the T2 cell line, vascular EC required significantly higher peptide concentrations to trigger half-maximal CTL-mediated lysis (Fig. 2,A), granule exocytosis by CTL (Fig. 2,B), and IFN-γ secretion by CTL (Fig. 2,C). When all experiments were considered together, 50- to 5000-fold higher peptide concentrations were necessary to trigger CTL responses using EC as targets compared with T2 cells. On the average, the peptide concentration required for half-maximal lysis by SMCY-specific CTL was 10−6.4 M for EC, 10−9.2 M for T2, 10−8.7 M for JY, 10−8.2 M for BLC, 10−8.9 M for the CC, and 10−8.6 M for PBMC (Fig. 3,A). Human dermal microvascular EC required a peptide concentration of 10−6.9 M for half-maximal lysis. The poor capacity of peptide-pulsed EC to activate CTL was qualitatively and quantitatively similar for both antigenic peptides used (Figs. 2 and 3).
Low expression of Ag-presenting MHC class I molecules on resting EC could explain the poor peptide presentation by EC. Overnight pretreatment with IFN-γ and TNF increased the expression of HLA-A2 molecules on the surface of EC (Fig. 4,A). However, this up-regulation of HLA-A2 expression did not change the peptide concentration required for efficient target cell lysis (Fig. 4,B). Furthermore, T2 cells and CC expressed equally low levels of surface HLA-A2 molecules (Fig. 4,A), but still required much lower peptide concentrations for half-maximal lysis. Rapid internalization of peptide-HLA-A2 complexes could also explain the poor killing of EC loaded with exogenous peptide. To rule out this mechanism, peptide-pulsed target cells were washed and preincubated at 37°C for the indicated time before the CTL assay was performed. Compared with T2 and BLC, the peptide-HLA-A2 complexes detected by Ag-specific CTL disappeared with a similar rate from the surface of EC (Fig. 4 C).
EC have impaired capacity to present immunodominant antigenic peptides
We next quantified the amount of peptide bound to HLA-A2 on T2 and EC using the flu peptide/HLA-A2-specific Fab M1-D12 (23). This recombinant humanized Fab monomer is specific for flu peptide bound to HLA-A2 and therefore can be considered a TCR-like protein (23). Compared with CTL-mediated target cell lysis, Fab binding to the peptide-HLA-A2 complex was less sensitive to detect antigenic peptides on the surface of target cells. Despite this restriction, the monomeric recombinant Fab was able to detect flu peptide bound to the target cell T2 (Fig. 5,A). Fab binding to the peptide-HLA-A2 complex was gradually reduced at decreasing peptide loading concentrations. Using this Fab, peptide binding to EC was not detectable even at the highest concentration tested, suggesting that exogenous peptides have impaired capacity to bind to HLA-A2 molecules on the surface of vascular EC (Fig. 5 B).
EC are not resistant and have enough costimulatory signals to be fully susceptible for CTL-mediated target cell lysis
The poor killing of EC by peptide-specific CTL could be the result of at least partial resistance to CTL-mediated lysis. To test this hypothesis, target cell lysis by CTL was assessed in the presence of PHA (5 μg/ml), a lectin that activates TCR pharmacologically. In the presence of PHA, EC were killed equally well compared with T2 as target cells (Fig. 6). This rules out the idea that EC are resistant to CTL-mediated lysis. The treatment of T lymphocytes with PHA activates the TCR very selectively and essentially does not involve activation of costimulatory molecules (12). The equipotent killing of T2 cells and EC by PHA-activated CTL provides strong evidence that endothelial costimulation allows efficient target cell lysis if the TCR is sufficiently activated. In all these experiments, target cell lysis was mediated by the granzyme/perforin pathway, because pretreatment of CTL with concanamycin A (25) abolished PHA-induced CTL-mediated target cell lysis completely (data not shown).
EC may present a different antigenic surface profile
To date we have demonstrated that EC have impaired capacity to present endogenously processed SMCY peptide to CTL. Using two completely different peptides we have shown that exogenous peptide loading of EC is very inefficient. To comprehensively assess whether EC present a different antigenic profile on their HLA-A2 molecules, we used an HLA-A2-specific CTL clone in the cytotoxicity assay. This clone was shown to kill a broad selection of HLA-A2-positive target cells equally well (Fig. 7, data not shown). Lysis of these target cells was maximal and was not further enhanced by PHA. Peptides bound to the HLA-A2 molecules may influence conformation of the molecular site of HLA-A2, which binds to the TCR of this CTL clone. Therefore, a different antigenic peptide repertoire should indirectly affect killing of target cells by this clone. Indeed, EC were the only cells analyzed that were killed less efficiently in this experiment. Pretreatment of EC with IFN-γ and TNF did not significantly improve lysis (Fig. 7) despite a substantial up-regulation of HLA-A2 molecules on these cells (Fig. 1 D). These findings provide indirect evidence that EC present a different antigenic profile to CTL.
In this study we have compared EC from two different vascular beds with other cell types as targets for peptide-specific, MHC class I-restricted CTL clones. EC were poor targets for these CTL. This escape from CTL-mediated lysis was fully explained by the inability of EC to present equal amounts of antigenic peptides compared with other target cells. Neither resistance of EC to CTL-mediated target cell lysis nor lack of costimulation by EC explains this difference. On epithelial cells and leukocytes, but not on EC, peptide Ag that was presented by endogenous protein processing or exogenous peptide loading could efficiently activate CTL. The poor binding of the TCR-like Fab monomer, which specifically detects flu peptide bound to HLA-A2, to peptide-loaded EC demonstrates that EC express lower amounts of flu peptide on their HLA-A2 molecules. At concentrations at which T2 cells show titrated peptide loading, EC have no equivalent numbers of flu/HLA-A2 complexes detectable by this method. Similarly, to pulse EC with peptide to reach equipotent killing compared with T2 cells, two to three orders of magnitude higher peptide concentrations were required. The fact that an HLA-A2-specific CTL clone that killed a variety of target cells to 90–100%, but EC to only 50%, of the expected level lends additional support to the conclusion that EC may present a different peptide repertoire to circulating T cells compared with leukocytes or epithelial cells.
Cell type-specific differences in Ag presentation have been reported previously. Dendritic cells, for example, present different viral peptides compared with nondendritic cells (26, 27), and the implication for peptide-based vaccine strategies has been discussed (27). Cell type-specific expression of minor histocompatibility Ags such as HA-1 can separate the beneficial graft-vs-leukemia from the harmful graft-vs-host effect (2). We report for the first time that by this mechanism, vascular EC can escape CTL-mediated lysis. The low amount of an immunodominant minor histocompatibility Ag presented by vascular EC explains why these cells are ignored by peptide-specific CTL. Up-regulation of HLA-A2 expression by cytokines to high levels does partially improve the presentation of endogenously processed, male-specific Ag SMCY. There is still >10-fold less Ag present on the surface of cytokine-treated EC, as judged by the slope of the killing curve obtained by peptide pulsing. Endothelial cells are larger than leukocyte-derived target cells and therefore might present a lower number of HLA-A2 molecules per surface area. We measured the average surface area of the two target cell types, BLC and HUVEC, and found the latter to have a 3.7-fold greater surface area. Taking into account that HUVEC are adherent cells and present <50% of their surface to the CTL in suspension, this difference is even smaller. Therefore, pure geometric aspects cannot account for the substantial differences seen in the susceptibility for killing by peptide-specific CTL. In contrast, exogenous peptide pulsing of EC remains completely unaffected by cytokine treatment. This implies that the mechanism that causes the poor peptide presentation by EC is not exactly the same for endogenous Ag processing and exogenous peptide loading.
Our findings have implications for the pathogenesis of alloimmune diseases such as GVHD. They may explain why mature, fully differentiated effector CTL that were activated by dendritic cells in secondary lymphoid organs tend to ignore vascular EC on their way to the epithelial compartment of the affected organ, e.g., skin, gut, or liver (7). Unless EC express a sufficiently high amount of the Ag against which specific CTL are directed, CTL are not triggered to release cytolytic granules and therefore they do not kill these cells. When human EC were cocultured with allogeneic CD8+ T cells, we reproducibly generated CTL that are specific for EC, but fail to kill leukocyte-derived target cells from the same donor (21). These EC-selective CTL, when they arise in vivo, may recognize antigenic epitopes that are preferentially expressed at high levels on EC. Such cells may well be involved in the rather selective CTL-mediated microvessel loss observed in the course of severe chronic GVHD (7).
Cross-presentation of Ag by vascular EC is involved in the recruitment of Ag-specific CD8+ T cells to peripheral tissues (28). The amount of peptide presented by EC that is necessary to trigger Ag-specific transmigration of CTL may be lower than the amount required to trigger cytolytic granule release. This could explain the fact that in vivo transmigration of Ag-specific CTL can be observed without any sign of endothelial injury (3, 4).
In summary, the quite selective escape of vascular EC from cell death caused by immunodominant effector CTL may well explain why blood vessels are spared for a remarkably long time in the course of GVHD. Our data demonstrate that EC are killed if they express a sufficient amount of Ag to be recognized by CTL, and vascular injury or microvessel loss may occur only under these conditions.
We thank Denise Wittwer for excellent technical assistance.
The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked advertisement in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
This work was supported by the Swiss National Science Foundation (Grant 3200-664121), the Krebsliga beider Basel, and The Erich and Gertrud Roggenbuck Stiftung.
Abbreviations used in this paper: EC, vascular endothelial cell; BLC, EBV-immortalized B lymphoblastoid cell; CC, colon cancer cell; GVHD, graft-vs-host disease; MR, maximum release; SR, spontaneous release.