Both CD4+ and CD8+ human memory but not naive T cells respond to allogeneic human dermal microvascular endothelial cells (HDMEC) in vitro by secreting cytokines and by proliferating. Several recently identified costimulators, namely, 4-1BB ligand, ICOS ligand, and OX40 ligand, are up-regulated on cultured HDMEC in response to TNF or coculture with allogeneic T cells. Blockade of these costimulators each partially reduces IFN-γ and IL-2 secretion and proliferation of previously resting memory T cells. The effects of these costimulators are overlapping but not identical. Memory but not naive T cells are the principal effectors of microvascular injury in human skin allografts following adoptive transfer into immunodeficient mice. Furthermore, blocking 4-1BB ligand, ICOS ligand, or OX40 ligand in this model reduces human skin allograft injury and T cell effector molecule expression. These data demonstrate that human memory T cells respond to microvascular endothelial cells and can injure allografts in vivo without priming. Furthermore, several recently described costimulators contribute to these processes.

Long-lived memory T cells may play an important role in early events following transplantation. Alloreactive memory T cell populations are not typically present in the young rodents commonly used to study transplantation. However, adoptive transfer experiments with memory T cells generated by Ag immunization have demonstrated that these cells can rapidly reject a cardiac graft in the absence of secondary lymphoid organs, whereas naive T cells cannot (1). In adult humans, about one-half of the T cells capable of responding to allogeneic cells in limiting dilution assays express CD45RO, a marker of memory cells (2, 3). However, it is unknown whether human memory T cells, like their murine counterparts, can mediate rejection.

Memory T cells have a special relationship to vascular endothelial cells (EC).3 Central memory T cells express chemokine receptors (e.g., CCR7) and adhesion molecules (e.g., CD62L) that preferentially interact with high endothelial venules (4). Effector memory cells express chemokine receptors like CXCR3 that respond to inflammatory chemokines and express high levels of adhesion molecules (e.g., LFA-1 or VLA-4) (5) whose ligands are preferentially expressed by cytokine-activated peripheral vascular EC (6). This phenotype allows effector memory cells to be directly recruited to sites of peripheral inflammatory reactions such as allograft rejection. Furthermore, resting memory T cells can be activated by alloantigens presented by EC but not by other stromal or parenchymal cells in vitro (7, 8).

The T cell response to Ag depends both on TCR signals (provided by peptide-MHC molecule complexes) and on Ag-independent costimulatory molecules. The best-described molecules on T cells for costimulation are CD28 (which binds B7-1 and B7-2, also known as CD80 and CD86, respectively) and CD2. CD2 binds CD48 in rodents but is preferentially activated in humans by LFA-3 (CD58), a molecule missing from the rodent genome (9). Human EC express LFA-3 (but generally not B7-1 or B7-2), and the LFA-3-CD2 pathway seems to be particularly important in human allogeneic responses to this cell type. However, Ab-blocking experiments have suggested that LFA-3 does not account for all of the costimulation that human EC provide to memory T cells (3, 10). Several newly discovered receptors for costimulation in both the B7 family (such as ICOS) and the TNF family (such as 4-1BB and OX40) are important for the generation of effector/memory T cells (11, 12, 13, 14), and ligands for these molecules can be expressed at high levels on cytokine-activated HUVEC (15, 16). It is not known whether any of these molecules also contribute to the reactivation of resting memory T cells by human EC.

Cumulatively, these observations suggest that circulating human alloreactive memory T cells may effectively recognize graft EC in vivo and may do so in a manner dependent on costimulators other than CD80 and CD86. We have developed a model in which we can test this hypothesis (17, 18). Specifically, we have engrafted C.B-17 SCID/beige mice with split thickness human skin containing the superficial vascular plexus of the papillary dermis. The healed graft is perfused through retained human EC-lined microvessels. At this point, usually 3–4 wk after skin transplantation, we adoptively transfer human allogeneic PBMC or T cell subpopulations by i.p. inoculation. Within 7 days, human T cells are present in the mouse circulation and by 12–14 days infiltrate the human graft. Between days 14 and 21, human graft EC are progressively destroyed and the graft frequently ulcerates, indicative of rejection. Previous studies have shown that this response depends on LFA-3 (19). In the present study, we used this model to show that the graft EC injury is mediated by human memory but not naive T cells. We also demonstrate that ICOS ligand (ICOSL), OX40 ligand (OX40L), and 4-1BB ligand (4-1BBL) are inducible on cultured human dermal microvascular EC (HDMEC) and that all three molecules contribute to EC-dependent memory T cell activation in vitro and graft rejection in vivo.

Recombinant human TNF (TNF-α) was purchased from R&D Systems, and IFN-γ was from BioSource International. Mouse anti-human 4-1BB and mouse anti-human OX-40 mAbs were purchased from Ancell. Mouse anti-human ICOS mAb was purchased from eBioscience. Mouse anti-human 4-1BBL mAb was a gift from X. G. Zhang (Soochow University, Suzhou, China), and mouse anti-human ICOSL mAb was a gift from H. W. Mages (Forschungs Institut for Molekulare Pharmacologie, Berlin, Germany). Mouse anti-human OX40L mAb and recombinant human 4-1BBFc were purchased from R&D Systems. Biotin-conjugated goat and donkey anti-mouse Abs were purchased from Jackson ImmunoResearch Laboratories.

All human cells and tissues were obtained under protocols approved by the Yale Human Investigations Committee. PBMCs were isolated by density gradient centrifugation of leukapharesis products by using lymphocyte separation medium (Invitrogen Life Technologies). Isolated cells were stored in 10% DMSO-90% FBS at −196°C and were thawed and washed before use (20). CD4/8+ T cells were isolated from PBMCs by positive selection by using Dynabeads (Dynal Biotech). The selected population obtained by this procedure was routinely >97% CD3+ by flow cytometry (data not shown). Activated T cells were removed by negative selection with an anti-HLA-DR Ab at a concentration of 5 μg/ml (LB3.1; gift of J. Strominger, Harvard University, Cambridge, MA) for 20 min, washing twice, and depleting by using magnetic beads conjugated to goat anti-mouse Ab (Dynal Biotech). Naive and memory subsets of T cells were isolated from the total T cell population by further negative selection by using anti-CD45RA or anti-CD45RO Abs at a concentration of 2 μg/ml (BioSource International).

HDMEC were isolated from discarded human skin, purified using anti-CD31 mini MACS beads (Miltenyi Biotec) and cultured in EGM2-MV growth medium (Cambrex) as previously described (21). When indicated, the cells were treated with 50 ng of IFN-γ (BioSource International) per milliliter for 3 days before cocultivation. After pretreatment with IFN-γ, the EC were uniformly HLA-DR positive (data not shown).

HDMEC (∼1.5 × 105 cells) were plated into the gelatin-coated well of 24-well culture plates (Falcon; BD Biosciences) and treated with IFN-γ where indicated. Purified T cell subsets were then added to each well (∼1 × 106 per well). All cultures were maintained in 5% CO2 at 37°C. The medium for coculture consisted of RPMI 1640 supplemented with 10% FBS serum (Invitrogen Life Technologies), 2 mM l-glutamine, 100 U/ml penicillin, and 100 μg/ml streptomycin.

RNA was isolated from T cells and EC using RNeasy mini kits (Qiagen) with on-column DNase treatment following the manufacturer’s supplied instructions. RNA was isolated from skin grafts using the RNeasy fibrous tissue mini kit (Qiagen) again with on-column DNase treatment and following the manufacturer’s supplied instructions. cDNA was synthesized using Taqman RT reagents (Applied Biosystems), following the manufacturer’s instructions. Quantitative real-time PCR (qRT-PCR) for 4-1BBL, ICOSL, OX40L, FasL, granzyme B, and perforin and CD3ε was performed exactly as described (22), using primers shown in Table I or as previously described (20). Samples were analyzed by generating a standard curve from plasmids containing the PCR fragment and expressed as copy number where indicated.

Table I.

PCR primers used in this studya

TargetPrimers (5′→3′)
4-1BBL ACTGCCCAGCTGGTACATTC 
 CACAGGTCCTTTGTCCACCT 
ICOSL AAACCTTGCAGGCAACAATC 
 CACTTGGCTTGGATCAGTCA 
OX40L GTGAATGGCGGAGAACTGAT 
 ATGCTGGTGCCTGGTTTTAG 
IL-2 AACTCACCAGGATGCTCACA 
 GCACTTCCTCCAGAGGTTTG 
IL-4 AGCTGATCCGATTCCTGAAAC 
 ACTCTGGTTGGCTTCCTTCA 
IFN-γ GTCCAACGCAAAGCAATACA 
 TGCTCTTCGACCTTGAAACA 
IL-10 TGAGAACCAAGACCCAGACA 
 CAGGGAAGAAATCGATGACA 
TargetPrimers (5′→3′)
4-1BBL ACTGCCCAGCTGGTACATTC 
 CACAGGTCCTTTGTCCACCT 
ICOSL AAACCTTGCAGGCAACAATC 
 CACTTGGCTTGGATCAGTCA 
OX40L GTGAATGGCGGAGAACTGAT 
 ATGCTGGTGCCTGGTTTTAG 
IL-2 AACTCACCAGGATGCTCACA 
 GCACTTCCTCCAGAGGTTTG 
IL-4 AGCTGATCCGATTCCTGAAAC 
 ACTCTGGTTGGCTTCCTTCA 
IFN-γ GTCCAACGCAAAGCAATACA 
 TGCTCTTCGACCTTGAAACA 
IL-10 TGAGAACCAAGACCCAGACA 
 CAGGGAAGAAATCGATGACA 
a

Each primer is listed 5′→3′. The forward primer is above and the reverse primer is below. In each case, the identity of the amplified fragment was confirmed by sequencing.

For cell surface immunostaining, HDMEC were washed twice with HBSS and incubated with trypsin-EDTA for 1 min. Detached cells were collected by centrifugation at 1000 × g for 5 min and washed twice with ice-cold PBS containing 1% BSA and 0.1% sodium azide (PBS/BSA). Detached cells or T cells were then incubated with the Abs or isotype controls at 2 μg/ml in PBS/BSA for 2 h at 4°C. After two further washes, cells were incubated with 2 μg/ml biotin-conjugated secondary Abs in PBS/BSA for 1 h at 4°C. Following another two washes, cells were incubated with 1 μg/ml strepavidin-R PE (Molecular Probes) in PBS/BSA for 30 min at 4°C, washed a further two times, and analyzed on a FACSort using Cellquest software (BD Biosciences).

Supernatants were collected from cocultures of T cells and HDMEC after 24 h of coculture. The samples were then assessed using an ELISA kit for IFN-γ (BioSource International) or IL-2 (eBioscience). Both ELISAs were performed as described by the manufacturers.

To measure proliferation by [3H]thymidine incorporation, HDMEC were cultured in 96-well U-bottom plates (Falcon) as described previously (21). HDMEC were then pretreated with 50 ng of IFN-γ (BioSource International) before coculture with CD4+ T cells or untreated before coculture with CD8+ T cells. All HDMEC were then treated with mitomycin C (50 μg/ml in PBS, 30 min; Sigma-Aldrich) before coculture to prevent proliferation as described previously (3). Twenty-four hours before each indicated time point, 1 μCi [3H]thymidine (Amersham Biosciences) was added to each well. Plates were frozen, thawed, and then harvested on a 96-well harvester (Tomtec) and counted on a Microbeta scintillation counter (Wallac). The mean of the replicates (n = 16) was calculated, and the mean [3H]thymidine incorporation into EC alone was subtracted.

To measure proliferation by CFSE dilution, the cells were stained with 250 nM CFSE (Molecular Probes) for 15 min before coculture with EC. Cells were then collected and subjected to FACS analysis.

C.B-17 SCID/beige female mice (Taconic Farms) were used at 5–8 wk of age. All protocols involving animals were approved by the Yale Animal Care and Use Committee. The animals were housed individually in microisolator cages and were fed autoclaved food and water. Serum IgG levels were determined by sandwich ELISA using reagents from Cappel as previously described (18). SCID/beige animals were considered “leaky” at IgG levels of >1 μg/ml and excluded from experimental use.

Human skin was obtained from cadaveric donors through the Yale University Skin Bank under a protocol approved by the Yale Human Investigations Committee. Human skin was orthotopically transplanted to SCID/beige mice as previously described (18). In brief, 0.5-mm-thick sheets were divided into 1-cm2 pieces, kept at 4°C in RPMI 1640 medium (Invitrogen Life Technologies), and fixed onto similarly sized defects on the dorsum of C.B-17 SCID/beige recipients using staples (3M). The resultant surface area of healed grafts was kept constant between animals when possible. The skin reproducibly grafted with a >95% success rate and was allowed to heal for 4–5 wk before manipulating the graft. Rare animals that did not successfully engraft were excluded from the experimental groups before treatments.

CD45RA+ or CD45RO+ T cells were isolated as described above, and cells were >95% pure populations; both populations contained approximately a 2:3 ratio of CD8+ to CD4+ cells. SCID/beige mice were reconstituted with 3 × 108 human PBMC by i.p. inoculation 4 wk after skin engraftment. Animals demonstrated no signs of graft-vs-host disease. Rare animals that failed to reconstitute with human T cells were, by prior design, excluded from analysis. Alternatively, separated naive or memory T cells were then counted using a Coulter particle counter (Beckman Coulter), and 1.5 × 108 cells were i.p. inoculated into SCID/beige mice 4 wk after skin engraftment.

Circulating human T cells were evaluated by flow cytometry as previously described (20). In brief, heparinized retro-orbital venous samples were obtained 14 days after reconstitution, and the erythrocytes were lysed. The leukocytes were incubated with FITC-conjugated mouse anti-human CD3 (Immunotech) and PE-conjugated rat anti-mouse CD45 (Sigma-Aldrich) mAbs or PE-conjugated CD45RO or CD45RA Abs (Immunotech). Samples were then analyzed using a FACSort (BD Biosciences). None of the treatments used in this study influenced the frequency or extent of engraftment.

Human skin grafts, harvested at indicated times, were processed for paraffin-embedded or frozen sections as previously described (20). Immunostaining was performed using isotype-matched, nonbinding control Abs or the following Ab: mouse anti-human CD3 (UCHT1, IgG1) (DakoCytomation).

The degree of graft microvascular damage was evaluated from H&E-stained sections by a dermatopathologist (J. M. McNiff) blinded to treatment protocols as previously described (19). In brief, the percentages of dermal vessels showing injury, defined as EC loss or sloughing, and thrombosis were assessed from an average of three high-power (×200) fields using the following semiquantitative grading scale: grade 0, all vessels patent and uninvolved; grade 1, <25% of vessels show injury; grade 2, ∼50% of vessels show injury; and grade 3, >75% of vessels show injury. The number of human CD3+ T cell infiltrates were assessed by counting and averaging the number of T cells in 10 random, ×40 high-power fields using the ImageJ program (National Institutes of Health).

Statistical differences between groups with respect to T cell infiltrates were evaluated using a two-tailed t test. Statistical difference between the pathology scores were evaluated using a nonparametric analysis Mann-Whitney U test.

Cultured HUVEC have been shown to preferentially activate allogeneic memory T cells compared with naive T cells, and they do so in an MHC- and costimulation-dependent manner (3, 23). Cultured HUVEC are the most widely used model of human EC, but the behaviors of microvascular EC, the targets of graft injury, are not always reflected accurately by this cell type. We therefore began our study by examining the capacity of EC isolated from human skin microvessels (i.e., HDMEC) to activate memory and naive T cells in coculture. Cultured HDMEC, like HUVEC, express HLA-A/B but not HLA-DR and increase expression of both type of molecules in response to IFN-γ (24). In our experiments, HDMEC were treated with IFN-γ for 3 days (to induce HLA-DR) for optimal CD4+ T cell stimulation and then cocultured with allogeneic, CFSE-labeled T cell subsets. Medium was collected at 24 h for measurement of cytokines, and T cells were collected after 7 days for analysis of proliferation. These time points were judged optimal in pilot experiments. Significant production of both IFN-γ and IL-2 was observed in cocultures containing memory but not naive T cells (Fig. 1,A). By FACS analysis, purified memory (CD45RO+) T cells, but not naive (CD45RO) T cells, proliferated in coculture with HDMEC (Fig. 1,B). When further separated into CD4+ and CD8+ memory and naive cells, again memory cells, but not naive cells proliferated in response to HDEMC (Fig. 1 C). In control experiments CD4+ T cells did not respond to HDMEC that were not pretreated with IFN-γ (data not shown). Thus, HDMEC, like HUVEC, selectively activate allogeneic memory T cells and need to be pretreated with IFN-γ to stimulate CD4+ T cells.

FIGURE 1.

HDMEC activate memory T cells. HDMEC were treated with IFN-γ to up-regulate MHC class I/II and subsequently cocultured with either naive (CD45RA+) or memory (CD45RO+) T cells. A, Media were taken for ELISA from HDMEC cultured with either naive or memory T cells for 24 h. Memory T cells cultured with HDMEC showed significant production of both IFN-γ and IL-2 (striped bars) when compared with naive T cells (filled bars). B, After 1 wk of coculture, CFSE-labeled T cells were stained with anti-CD3 and subjected to FACS analysis. Memory T cells demonstrated significant proliferation (4.51%; right panel, top left, circled population), whereas naive T cells showed very little proliferation (0.49%; left panel, top left, circled population) as judged by CFSE dilution. C, Following 1 wk of coculture, purified and CFSE-labeled CD4+ or CD8+ naive and memory T cells were stained with CD4 or CD8, respectively, and analyzed using FACS. Again, purified memory T cells (left panels, top left, circled populations) or either subset showed significantly greater proliferation than their purified naive counterparts (right panels, top left, circled populations). Data represent one of four experiments with similar results.

FIGURE 1.

HDMEC activate memory T cells. HDMEC were treated with IFN-γ to up-regulate MHC class I/II and subsequently cocultured with either naive (CD45RA+) or memory (CD45RO+) T cells. A, Media were taken for ELISA from HDMEC cultured with either naive or memory T cells for 24 h. Memory T cells cultured with HDMEC showed significant production of both IFN-γ and IL-2 (striped bars) when compared with naive T cells (filled bars). B, After 1 wk of coculture, CFSE-labeled T cells were stained with anti-CD3 and subjected to FACS analysis. Memory T cells demonstrated significant proliferation (4.51%; right panel, top left, circled population), whereas naive T cells showed very little proliferation (0.49%; left panel, top left, circled population) as judged by CFSE dilution. C, Following 1 wk of coculture, purified and CFSE-labeled CD4+ or CD8+ naive and memory T cells were stained with CD4 or CD8, respectively, and analyzed using FACS. Again, purified memory T cells (left panels, top left, circled populations) or either subset showed significantly greater proliferation than their purified naive counterparts (right panels, top left, circled populations). Data represent one of four experiments with similar results.

Close modal

LFA-3 is an important costimulator expressed by HUVEC, but blocking the LFA-3 interaction with its ligand, CD2, does not fully inhibit proliferation or cytokine production by allogeneic T cells (25). To characterize additional molecules involved in costimulation, we initially examined HDMEC with or without treatment with proinflammatory cytokines (at doses and times shown to be optimal for other HDMEC responses), for the expression of three molecules proposed to act preferentially on activated T cells, namely 4-1BBL, ICOSL, and OX40L (CD134L). Specifically, HDMEC were mock treated or treated with either TNF (10 ng/ml) or IFN-γ (50 ng/ml) for 24, 48, or 72 h and collected for qRT-PCR and FACS analysis. Mock-treated HDMEC expressed minimal transcript levels and surface expression for all three of these memory-selective costimulators. Transcript levels of all three molecules were increased manyfold at 24 h and remained elevated for 4-1BBL and OX40L, but not ICOSL, over 72 h (Fig. 2,A). There were some differences between surface expression and mRNA expression. 4-1BBL surface expression only increased modestly, whereas ICOSL surface expression increased substantially with TNF treatment (Fig. 2,B). TNF treatment shifted OX40L, which has high basal expression, from a heterogeneous population of low and high expressors to a mostly high expressing population leading to an overall increase in OX40L expression assessed as corrected mean fluorescence intensity (Fig. 2 B). IFN-γ did not affect costimulator mRNA or protein for any of the molecules (data not shown).

FIGURE 2.

HDMEC express inducbile memory T cell-selective costimulatory molecules. HDMEC were cultured with cytokines for the indicated times, and mRNA levels were calculated as a percent change from untreated cells. A, HDMEC cultured with TNF (10 ng/ml) showed a 50-fold increase in mRNA for 4-1BBL at 24 h that remained elevated for 72 h; 6-fold and 8-fold increases in mRNA were observed for ICOSL and OX40L, respectively, at 24 h, which remained elevated for 72 h for OX40L but slowly declined for ICOSL over the same time period (▪). HDMEC cultured with IFN-γ (50 ng/ml) showed no increase in mRNA for either 4-1BBL, ICOSL, or OX40L over 72 h (□). Data are pooled from four separate experiments. B, By FACS analysis, 4-1BBL, ICOSL, and OX40L all showed increased surface expression (black lines) with 72 h of TNF treatment but no increase over the same time period with IFN-γ treatment compared with control, untreated EC (gray lines). IgG controls, which do not change with cytokine treatment, are shown as filled histograms. Data represent one of four experiments with similar results.

FIGURE 2.

HDMEC express inducbile memory T cell-selective costimulatory molecules. HDMEC were cultured with cytokines for the indicated times, and mRNA levels were calculated as a percent change from untreated cells. A, HDMEC cultured with TNF (10 ng/ml) showed a 50-fold increase in mRNA for 4-1BBL at 24 h that remained elevated for 72 h; 6-fold and 8-fold increases in mRNA were observed for ICOSL and OX40L, respectively, at 24 h, which remained elevated for 72 h for OX40L but slowly declined for ICOSL over the same time period (▪). HDMEC cultured with IFN-γ (50 ng/ml) showed no increase in mRNA for either 4-1BBL, ICOSL, or OX40L over 72 h (□). Data are pooled from four separate experiments. B, By FACS analysis, 4-1BBL, ICOSL, and OX40L all showed increased surface expression (black lines) with 72 h of TNF treatment but no increase over the same time period with IFN-γ treatment compared with control, untreated EC (gray lines). IgG controls, which do not change with cytokine treatment, are shown as filled histograms. Data represent one of four experiments with similar results.

Close modal

We next evaluated whether these three costimulators molecules are induced when CD4+ T cells are cocultured with IFN-γ-pretreated allogeneic HDMEC. Using qRT-PCR analysis, we found that mRNA for 4-1BBL, ICOSL, and OX40L all increase in HDMEC after coculture with allogeneic T cells. The induction of 4-1BBL and OX40L was most pronounced at 48 h, with almost a 100-fold induction, whereas ICOSL showed a more modest 20-fold induction (Fig. 3,A). Surface expression was then assessed by FACS. 4-1BBL increased modestly at day 1 and steadily increased through day 7. ICOSL showed immediate increase by day 1 and also continued to increase throughout the coculture. OX40L, in contrast, increased at day 1, but then returned to baseline by day 3 (Fig. 3,B). T cells recovered from the same cocultures at days 1, 3, 5, and 7 expressed increasing levels of the corresponding receptors (4-1BB, ICOS, and OX-40) to these EC ligands on cells that had proliferated, although OX40 expression actually declined by day 7. Surprisingly, these molecules were also increased in expression on some T cells that had not proliferated, as assessed by dilution of CFSE fluorescence (Fig. 3 C). We do not know whether this is a bystander response to cytokines or a response to suboptimal TCR stimulation.

FIGURE 3.

HDMEC up-regulate memory costimulatory molecules in coculture with T cells and stimulate T cells to express the corresponding ligands. HDMEC were cocultured with allogeneic T cells, and cells were harvested at 3, 6, 12, 24, and 48 h for RNA and at day 7 for FACS analysis. A, HDMEC up-regulate for mRNA 4-1BBL (•), ICOSL (▦), and OX40L (♦) over the course of 48 h. B, 4-1BBL, ICOSL, and OX40L protein expression on HDMEC after 1, 3, 5, and 7 days of coculture with allogeneic T cells. cMFI, corrected mean fluorescent intensities, i.e., cMFI = Ab MFI − IgG control. C, Both proliferating T cells (circled population) and nonproliferating T cells identified by CFSE dye dilution show increasing expression of 4-1BB, ICOS, and OX40 at day 1, 3, 5, and 7 of coculture (numbers represent percentage of total T cell population). Data represent one experiment of four with similar results.

FIGURE 3.

HDMEC up-regulate memory costimulatory molecules in coculture with T cells and stimulate T cells to express the corresponding ligands. HDMEC were cocultured with allogeneic T cells, and cells were harvested at 3, 6, 12, 24, and 48 h for RNA and at day 7 for FACS analysis. A, HDMEC up-regulate for mRNA 4-1BBL (•), ICOSL (▦), and OX40L (♦) over the course of 48 h. B, 4-1BBL, ICOSL, and OX40L protein expression on HDMEC after 1, 3, 5, and 7 days of coculture with allogeneic T cells. cMFI, corrected mean fluorescent intensities, i.e., cMFI = Ab MFI − IgG control. C, Both proliferating T cells (circled population) and nonproliferating T cells identified by CFSE dye dilution show increasing expression of 4-1BB, ICOS, and OX40 at day 1, 3, 5, and 7 of coculture (numbers represent percentage of total T cell population). Data represent one experiment of four with similar results.

Close modal

The parallel induction of the receptors on T cells and their EC costimulators suggest that these interactions could contribute to memory T cell activation by allogeneic HDMEC. We therefore examined the effect of blocking these pathways using Abs and/or fusion proteins in T cell-EC cocultures. In our experiments, HDMEC were treated with IFN-γ for 3 days (to induce HLA-DR) for CD4+ T cell stimulation or mock treated for CD8+ T cell stimulation and then cocultured with allogeneic T cell subsets. Pilot titrations were done in T cell-EC cocultures to determine maximally inhibiting doses of each Ab and fusion protein typically between 4 and 10 μg/ml (data not shown). Cytokines were assessed by ELISA in the supernatants collected from the cocultures at 24 h. By ELISA, we found that blocking either ICOSL or OX40L with specific Abs led to a substantial decrease in IFN-γ production by CD4+ T cells, whereas blockade of 4-1BBL with a fusion protein had little effect (Fig. 4,A). IL-2 production by CD4+ T cells showed similar decreases (Fig. 4,A). In contrast, blockade of any of the three costimulators in CD8+ T cell-EC cocultures led to at least a 70% decrease in IFN-γ production (Fig. 4,B) when compared with control Ab. Overall, CD8+ T cells produce substantially less IL-2 and exhibited a lesser decrease in IL-2 production than CD4+ T cells with blockade of any of the three costimulators (Fig. 4,B). T cell proliferation was also reduced by blocking each of these costimulators. For CD4+ T cells, blocking 4-1BBL or OX40L decreased proliferation by almost 35% and blocking ICOSL led to a decrease of almost 45% at day 5 (Fig. 5,A). CD8+ T cells also demonstrated decreased proliferation, with 4-1BBL blockade reducing proliferation by 40% and ICOSL or OX40L blockade causing lesser, but significant, reductions of 26% and 20% (Fig. 5,B). Our results suggest that these costimulators are not interchangeable but that all three contribute to resting memory T cell activation by allogeneic HDMEC. In several experiments, combined blockade did not appear more effective than blockade of individual costimulators (data not shown). The effects of blocking these pathways were comparable in magnitude to those attained with blockade of LFA-3 (Fig. 5 and data not shown). Overall, our in vitro experiments demonstrate that allogeneic HDMEC, like HUVEC, selectively activate memory T cells and that several newly described costimulators, which are inducible on HDMEC, contribute to the alloresponse.

FIGURE 4.

CD4+ and CD8+ T cell production of IFN-γ and IL-2 in allogeneic T cell-EC cocultures is decreased when memory costimulator molecules are blocked. CD4+ (A) or CD8+ (B) T cells were cocultured with allogeneic HDMEC pretreated with IFN-γ for 3 days (to up-regulate MHC class II). Cytokine production was then assessed by ELISA on supernatant collected at 24 h. A, CD4+ T cells exhibit >50% reduction of both IFN-γ and IL-2 production in the presence of blocking Abs to ICOSL or OX40L but not in the presence of the blocking fusion protein 4-1BB-Fc. Combinations of Abs do not show additive blocking effects. B, CD8+ T cells demonstrate a large decrease in IFN-γ production in the presence of all three blocking reagents but only modest a decrease in IL-2 production. Data represent one of four experiments with similar results.

FIGURE 4.

CD4+ and CD8+ T cell production of IFN-γ and IL-2 in allogeneic T cell-EC cocultures is decreased when memory costimulator molecules are blocked. CD4+ (A) or CD8+ (B) T cells were cocultured with allogeneic HDMEC pretreated with IFN-γ for 3 days (to up-regulate MHC class II). Cytokine production was then assessed by ELISA on supernatant collected at 24 h. A, CD4+ T cells exhibit >50% reduction of both IFN-γ and IL-2 production in the presence of blocking Abs to ICOSL or OX40L but not in the presence of the blocking fusion protein 4-1BB-Fc. Combinations of Abs do not show additive blocking effects. B, CD8+ T cells demonstrate a large decrease in IFN-γ production in the presence of all three blocking reagents but only modest a decrease in IL-2 production. Data represent one of four experiments with similar results.

Close modal
FIGURE 5.

CD4+ and CD8+ T cells show decreased proliferation in the presence of blocking reagents to the memory costimulatory molecules. Thymidine was added to the cocultures 24 h before the time points indicated, and the control is T cells alone. A, CD4+ T cells were cocultured with allogeneic HDMEC that had been pretreated with IFN-γ for 3 days (p < 0.02 for all points at day 5). B, CD8+ T cells were cocultured with allogeneic HDMEC (p < 0.02 for all points at day 5). Data represent one of five experiments with similar results.

FIGURE 5.

CD4+ and CD8+ T cells show decreased proliferation in the presence of blocking reagents to the memory costimulatory molecules. Thymidine was added to the cocultures 24 h before the time points indicated, and the control is T cells alone. A, CD4+ T cells were cocultured with allogeneic HDMEC that had been pretreated with IFN-γ for 3 days (p < 0.02 for all points at day 5). B, CD8+ T cells were cocultured with allogeneic HDMEC (p < 0.02 for all points at day 5). Data represent one of five experiments with similar results.

Close modal

We have previously described a model of human T cell responses to allogeneic skin microvascular EC in vivo involving human skin grafts and adoptive transfer of human PBMC in immunodeficient mice (18). To determine which T cell subsets contribute to graft EC injury, we adoptively transferred purified T cell subpopulations and compared the results to those observed following transfer of PBMC. CD4+ plus CD8+ T cells were isolated from whole PBMC and then further fractionated into memory and naive subsets based on negative selection of CD45 isoform expression (Fig. 6,A). Memory (CD45RO+) or naive (CD45RA+) T cells were then injected i.p. into SCID/beige mice engrafted with human skin. One week later, circulating CD3+ T cells were present in similar numbers following adoptive transfer of either CD45RO+ or CD45RA+ T cell subsets, as shown by FACS analysis of mouse blood (Fig. 6,B). Interestingly, mice reconstituted with CD45RA+ T cells developed some circulating CD45RO+ T cells (Fig. 6 B) in addition to circulating CD45RA+ T cells, consistent with observations that naive T cells acquire some memory markers during homeostatic proliferation (26). In contrast, mice receiving purified CD45RO+ cells displayed exclusively CD45RO+ cells in their circulation. Similar to effects following adoptive transfer of whole PBMC, neither human T cell subset appeared to infiltrate or injure any mouse tissue (data not shown).

FIGURE 6.

Injection of CD45RO+ (memory) T cells or CD45RA+ (naive) T cells results in circulating human T cells in SCID-beige mice. A, T cell subsets were isolated as described in the text and subjected to FACS analysis. CD45RA+ T cells (naive), CD45RO+ T cells (memory) (1.5 × 108 cells) or whole PBMC (3 × 108) were injected i.p. into SCID mice that had been grafted with skin allogeneic to the mononuclear cells. B, After 7 days, peripheral blood lymphocytes where then stained with anti-human CD3-FITC Ab and either anti-human CD45RA-PE (top panel) or anti-human CD45RO-PE and analyzed by flow cytometry. As seen in the top panel, only mice injected with naive T cells or PBMC have circulating naive cells, whereas mice receiving only memory cells do not. Bottom panel, mice receiving naive T cells, memory T cells, or whole PBMC all have circulating memory T cells. Data represent one of five experiments with similar results (n = 15 per group).

FIGURE 6.

Injection of CD45RO+ (memory) T cells or CD45RA+ (naive) T cells results in circulating human T cells in SCID-beige mice. A, T cell subsets were isolated as described in the text and subjected to FACS analysis. CD45RA+ T cells (naive), CD45RO+ T cells (memory) (1.5 × 108 cells) or whole PBMC (3 × 108) were injected i.p. into SCID mice that had been grafted with skin allogeneic to the mononuclear cells. B, After 7 days, peripheral blood lymphocytes where then stained with anti-human CD3-FITC Ab and either anti-human CD45RA-PE (top panel) or anti-human CD45RO-PE and analyzed by flow cytometry. As seen in the top panel, only mice injected with naive T cells or PBMC have circulating naive cells, whereas mice receiving only memory cells do not. Bottom panel, mice receiving naive T cells, memory T cells, or whole PBMC all have circulating memory T cells. Data represent one of five experiments with similar results (n = 15 per group).

Close modal

Human skin grafts were harvested from the mice at 15 days after adoptive transfer of T cells and analyzed using histology, immunohistochemistry, and qRT-PCR. H&E (Fig. 7,A) and anti-CD3 (Fig. 7,B) staining showed a significantly different pattern of infiltration and tissue destruction following transfer of naive or memory T cells. Transferred CD45RA+ T cells produced only limited perivascular infiltrates, whereas CD45RO+ T cells, like whole PBMC, infiltrated the dermis in a more diffuse pattern. Quantitatively, naive T cells showed less total infiltration of CD3+ T cells compared with memory T cells (Fig. 7,C). Naive T cells did not produce histological evidence of EC injury, whereas memory T cells, like PBMC, did do so (Fig. 7,C), even when mice given naive cells were followed out to 5 wk (data not shown). RNA transcripts encoding effector molecules (FasL, granzyme B, and perforin) were minimally expressed in the CD45RA+ inoculated mice compared with expression following either CD45RO+ T cells or unfractionated PBMC inoculation (Fig. 7 D). These data suggest that memory T cells largely account for the rejection response observed with whole PBMC because naive T cells, although they can enter the circulation and focally infiltrate human skin grafts, do not express effector molecules or cause graft injury. Thus, our model provides a means to study human memory T cell alloresponses in vivo.

FIGURE 7.

Reconstitution of mice with memory T cells, but not naive cells, exhibits a rejection response comparable to whole PBMC. Mice were reconstituted with CD45RA+ (naive) T cells, CD45RO+ (memory) T cells, or PBMC, and grafts were then harvested at 21 days for analysis. A, H&E staining demonstrates that mice reconstituted with memory cells exhibit an infiltration pattern (full dermis, ablation of the dermal papillae architecture) similar to whole PBMC, whereas naive cells exhibit a much more focal, perivascular infiltrate that does not damage the dermal architecture (n = 18). B, CD3 staining of the infiltrated dermis confirms that the infiltrate observed in the H&E is human T cells. C, CD3+ T cell counts showed a decreased number of naive T cells infiltrating the skin graft compared with memory T cells and only the mice reconstituted with memory T cells or whole PBMC show significant endothelial damage and vascular thrombosis. D, qRT-PCR of RNA isolated from grafts at day 10 indicate that memory cells express much higher levels of FasL, granzyme B, and perforin than naive cells, although less than whole PBMC (n = 9). Data were analyzed using standard curves generated from plasmids and normalized to CD3ς transcripts.

FIGURE 7.

Reconstitution of mice with memory T cells, but not naive cells, exhibits a rejection response comparable to whole PBMC. Mice were reconstituted with CD45RA+ (naive) T cells, CD45RO+ (memory) T cells, or PBMC, and grafts were then harvested at 21 days for analysis. A, H&E staining demonstrates that mice reconstituted with memory cells exhibit an infiltration pattern (full dermis, ablation of the dermal papillae architecture) similar to whole PBMC, whereas naive cells exhibit a much more focal, perivascular infiltrate that does not damage the dermal architecture (n = 18). B, CD3 staining of the infiltrated dermis confirms that the infiltrate observed in the H&E is human T cells. C, CD3+ T cell counts showed a decreased number of naive T cells infiltrating the skin graft compared with memory T cells and only the mice reconstituted with memory T cells or whole PBMC show significant endothelial damage and vascular thrombosis. D, qRT-PCR of RNA isolated from grafts at day 10 indicate that memory cells express much higher levels of FasL, granzyme B, and perforin than naive cells, although less than whole PBMC (n = 9). Data were analyzed using standard curves generated from plasmids and normalized to CD3ς transcripts.

Close modal

To evaluate the function of the memory T cell-EC costimulatory interaction in vivo, we again used our model of human skin allograft rejection in immunodeficient mice. We first investigated whether skin grafts in our alloimmune response express these ligands. Although basal expression was low, we found a large increase in the mRNA expression of 4-1BBL and ICOSL following PBMC transfer; OX40L showed a more modest increase (Fig. 8 A). These increases may be explained by the presence of TNF in the circulation following adoptive transfer of human PBMC (27). Unfortunately, the reagents available for FACS analysis of cultured cells are unable to detect these molecules in tissues; therefore, we could not directly assess protein expression in the skin grafts.

FIGURE 8.

Rejecting skin grafts in the human-SCID chimera up-regulate mRNA for the memory costimulatory ligands, and blocking these pathways decreases rejection pathology. mRNA was harvested from skin grafts that had been on mice reconstituted for 10 days with PBMC. A, qRT-PCR on the mRNA harvested from these grafts showed an induction of mRNA for 4-1BBL, ICOSL, and OX40L (n = 12). Similar grafts were then harvested from mice that had been reconstituted for 10 days with PBMC that had also been receiving i.p. injections of blocking Ab (100 μg) every 3 days. B, 10 random high-powered fields were taken of slides cut from the grafts and stained for human CD3ε for assessment. Counts of the CD3ε+ T cells showed similar levels of T cell infiltration except for LFA-3, which showed a significantly decreased number of infiltrating T cells (p < 0.001). H&E slides were generated from these grafts and subsequently evaluated by a dermatopathologist blinded to the treatment. Blocking 4-1BBL, ICOSL, or OX40L reduced endothelial injury (p < 0.004, 0.05, and 0.0003, respectively) and thrombosis (p < 0.01, 0.1, and 0.002), despite having similar levels of infiltration (C). LFA-3, as described previously, showed decreases in both endothelial injury (p < 10−6) and thrombosis (p < 0.0004) corresponding with its lower level of infiltration (n = 9 per treatment group).

FIGURE 8.

Rejecting skin grafts in the human-SCID chimera up-regulate mRNA for the memory costimulatory ligands, and blocking these pathways decreases rejection pathology. mRNA was harvested from skin grafts that had been on mice reconstituted for 10 days with PBMC. A, qRT-PCR on the mRNA harvested from these grafts showed an induction of mRNA for 4-1BBL, ICOSL, and OX40L (n = 12). Similar grafts were then harvested from mice that had been reconstituted for 10 days with PBMC that had also been receiving i.p. injections of blocking Ab (100 μg) every 3 days. B, 10 random high-powered fields were taken of slides cut from the grafts and stained for human CD3ε for assessment. Counts of the CD3ε+ T cells showed similar levels of T cell infiltration except for LFA-3, which showed a significantly decreased number of infiltrating T cells (p < 0.001). H&E slides were generated from these grafts and subsequently evaluated by a dermatopathologist blinded to the treatment. Blocking 4-1BBL, ICOSL, or OX40L reduced endothelial injury (p < 0.004, 0.05, and 0.0003, respectively) and thrombosis (p < 0.01, 0.1, and 0.002), despite having similar levels of infiltration (C). LFA-3, as described previously, showed decreases in both endothelial injury (p < 10−6) and thrombosis (p < 0.0004) corresponding with its lower level of infiltration (n = 9 per treatment group).

Close modal

To investigate the role of memory T cell-selective costimulators in the progression of immune-mediated damage in human skin grafts, i.p. injections of blocking Abs or IgG control were administered three times per week to groups of animals with healed human skin grafts starting on the day before i.p. inoculation with PBMC. The dosing schedule used was one that gives circulating trough levels of Ig sufficient to match in vitro levels. Within each experiment, all skin grafts were obtained from a single donor and all PBMC came from a second donor allogeneic to the skin graft donor. Data from five independent experiments were pooled for analysis; in each experiment, additional control animals bearing skin grafts were injected with blocking Abs or IgG control and were not inoculated with PBMC to be certain that injury was not caused by the Abs. Injury to graft EC was evaluated as microvascular luminal occlusion or shedding of EC, initial steps in human vessel loss (18). Blockade of ICOSL, 4-1BBL, or OX40L all failed to reduce T cell infiltration, whereas blockade of LFA-3 did do so (Fig. 8,B). ICOSL blockade decreased endothelial injury and thrombosis, but the difference in thrombosis did not achieve statistical significance. Blockade of the 4-1BBL or OX40L pathway reduced both endothelial injury and thrombosis to a significant degree (Fig. 8 C). However, the decrease in endothelial damage was not as effective as that observed with blockade of LFA-3.

In a final series of experiments, we used qRT-PCR to examine RNA obtained from grafts on animals treated with blocking Abs. Blockade of 4-1BBL, ICOSL, or OX40L all significantly decreased FasL and perforin mRNA production in T cells, although blockade had little effect on granzyme B expression (Table II). Analysis of cytokine mRNA expression within skin grafts was less clear cut. Blockade of 4-1BBL caused a decrease in IFN-γ, IL-2, and IL-10 mRNA expression compared with control grafts but no change in IL-4 mRNA expression (Table II). In contrast, ICOSL blockade showed an increase in IFN-γ, a decrease in IL-4 mRNA and a modest decrease in IL-2 and IL-10 mRNA levels. OX40L blockade, when compared with IgG control, showed dramatic reductions in IFN-γ, IL-2, and IL-4 expression but a comparable level of IL-10 mRNA. Overall, there is a better correlation of reduction in cytotoxic effector molecule than on cytokine expression with protection of grafts from injury, consistent with the correlation of effector molecule transcripts with rejection of human allografts (28, 29).

Table II.

Effect of Ab blockade on expression of effector molecules and cytokinesa

TreatmentPercent Inhibition of Expression
FasLGranzyme BPerforinIFN-γIL-2IL-4IL-10
4-1BBL 85.6 ± 9.8 8.8 ± 8.0 32.3 ± 21.9 42.1 ± 1.7 59.6 ± 8.6 (−4.0 ± 5.1) 47.4 ± 2.4 
ICOS-L 98.3 ± 0.7 (−28.3 ± 14.9) 80.5 ± 4.9 (−51.8 ± 10.0) 30.4 ± 15.0 63.1 ± 7.7 59.7 ± 11.9 
OX-40L 81.8 ± 14.5 0.0 ± 8.2 70.8 ± 10.3 59.3 ± 5.0 92.8 ± 3.5 95.8 ± 4.2 6.3 ± 14.3 
LFA-3 98.7 ± 0.01 82.3 ± 5.9 74.8 ± 8.9     
TreatmentPercent Inhibition of Expression
FasLGranzyme BPerforinIFN-γIL-2IL-4IL-10
4-1BBL 85.6 ± 9.8 8.8 ± 8.0 32.3 ± 21.9 42.1 ± 1.7 59.6 ± 8.6 (−4.0 ± 5.1) 47.4 ± 2.4 
ICOS-L 98.3 ± 0.7 (−28.3 ± 14.9) 80.5 ± 4.9 (−51.8 ± 10.0) 30.4 ± 15.0 63.1 ± 7.7 59.7 ± 11.9 
OX-40L 81.8 ± 14.5 0.0 ± 8.2 70.8 ± 10.3 59.3 ± 5.0 92.8 ± 3.5 95.8 ± 4.2 6.3 ± 14.3 
LFA-3 98.7 ± 0.01 82.3 ± 5.9 74.8 ± 8.9     
a

mRNA from human skin grafts harvested from mice that had been reconstituted for 10 days with PBMC and receiving i.p. injections of blocking mAbs (100 μg) every 3 days was obtained for qRT-PCR. Data was analyzed by comparing threshold cycles to a standard curve generated using plasmid DNA and then expressed as a percentage of IgG control.

In this report, we present four new observations relevant for human transplantation. First, we show that human microvascular EC, in this instance HDMEC, preferentially activate allogeneic CD3+ T memory T cells, defined by expression of CD45RO, but not naive T cells. Second, we show that three recently described costimulatory molecules, namely, ICOSL, OX40L, and 4-1BBL, are inducible on HDMEC by TNF or by culture with allogeneic T cells and that these ligands contribute to activation of resting memory T cells by allogeneic HDMEC. Third, we show that, following adoptive transfer in immunodeficient mice, memory but not naive T cells can attack human microvessels in allogeneic (to the T cell) skin grafts. Finally, we show that the three costimulators we have determined as inducible on HDMEC in culture are also inducible in human skin grafts and contribute to human memory T cell-mediated injury during an alloreaction. Protection correlates with reduction of effector molecule transcripts within the infiltrated graft.

Many previous studies examining the role of EC in T cell activation have focused on EC derived from large vessels (namely HUVEC) (3, 10, 30). However, in allograft rejection and many other inflammatory situations, it is not the macrovasculature but rather the microvasculature that may be responsible for the recruitment of inflammatory cells, initiation of the immune response, and eventually the target of the alloresponse. Microvascular EC in rejecting grafts express adhesion molecules important for T cell recruitment, namely, ICAM-1, VCAM-1, and E-selectin (31, 32) and may have prolonged expression compared with HUVEC (21). Furthermore, MHC class II was shown to be expressed basally on microvasculature in unstimulated, untransplanted hearts and was up-regulated following rejection episodes (33). The microvasculature has also been shown to activate CD4+ and CD8+ T cells; however, they appear to possess different requirements for activating T cells compared with macrovascular EC (10, 34, 35). This difference may be due to a difference in costimulator expression. For example, HUVEC express the costimulatory molecules LFA-3 and CD40 but lack both B7 molecules, CD80 and CD86 (36); however, unlike HUVEC, microvascular EC from cardiac tissue express CD80 after CD40 ligation (32). We have not detected CD80 or CD86 on the HDMEC used in our experiments (M. S. Kluger, personal communication). Finally, microvascular EC are the primary target of rejection in vascularized organ transplants (18). Because of the important role of the microvasculature in the alloimmune response outlined above, we focused our study on their behavior. We show in our study that HDMEC, like HUVEC, have the capacity to activate memory (CD45RO+) T cells, but not naive (CD45RO) to proliferate and produce cytokines.

We next looked at the expression of several newly described molecules 4-1BBL, ICOSL (also known as B7Rp1, B7h, GL-50, or LICOS), and OX40L (gp34) in response to cytokines and T cell coculture. Previous studies have demonstrated that activation of T cells by EC involves the interaction of LFA-3, ICOSL, and OX40L (15, 25, 37); however, no data exist regarding 4-1BBL on EC and only ICOSL has been studied on microvascular EC (15). In addition to EC, 4-1BBL, OX40L, and ICOSL are expressed on most classical APCs, including dendritic cells, B cells, and activated monocytes (38, 39, 40). ICOS is induced following initial activation of T cells, and ligation of ICOS leads to augmented proliferation (41) and secretion of many effector cytokines (including IL-4, IL-10, IFN-γ) from CD4+ T cells (42). Like ICOS, in humans, OX40 is expressed by T cells after ligation of the TCR and was initially identified as a marker of T cell activation (43) and appears to be important for proliferation and cytokine production by CD4+ T cells when stimulated with HUVEC (44) or allogeneic DC (37). 4-1BB, like ICOS and OX40, augments cytokine production and proliferation; however, it does so for both CD4+ and CD8+ T cells (14).

In this report, we demonstrate that all three of these molecules are inducible on HDMEC in response to TNF, but not IFN-γ. This is consistent with a previous report of ICOSL on HDMEC (15); the expression of OX40L and 4-1BBL has not been previously reported on this cell type. Furthermore, previous studies have not examined expression in cocultures of T cells with human EC. We show that coculture of HDMEC and allogeneic CD4+ T cells induces expression of both the costimulator on T cells and their corresponding ligands on HDMEC. The mRNA for the ligands is induced early on EC for all three ligands; however, their surface expression differs markedly, with 4-1BBL peaking and persisting at day 7, ICOSL peaking and persisting at day 3, and OX40L peaking at day 1 and then returning to baseline by day 3 (Fig. 2 B). This difference in expression may explain the reduction on 24-h cytokine production in CD4+ T cells when ICOSL and OX40L are blocked but the lack of effect when 4-1BB is blocked. Surface expression of OX40L did not increase to the degree anticipated by the mRNA. The reason for this difference is unclear and under current investigation. Blockade of any of the three costimulator molecules seems to have moderate effects on both CD8+ T cell cytokine production and proliferation. Preliminary data regarding combinations of blockade indicate that the effects are less than additive, consistent with overlapping activity (data not shown). CD4+ T cells in coculture with EC showed increasing expression of 4-1BB, ICOS, and OX40 in both the proliferated and unproliferated population as measured by CFSE dilution. The increased expression on the proliferating cells is consistent with earlier data that showed that these molecules are up-regulated on T cells on activation (45, 46); however, the increased expression on unproliferated cells has not been observed previously and may be a response to cytokines produced by the activated cells or may represent responses of T cells that are incompletely activated by TCR signals.

Memory T cells are thought to play an important role in allograft rejection. In particular, memory T cells may have important roles in rejection resistant to immunosuppressants because the activation requirements of memory T cells appear to be less stringent that those of naive T cells (47), and may be more difficult to inhibit and to tolerize (48, 49). Several previous studies have reported the presence of memory T cells (CD45RO+) in heart and kidney allograft biopsies and in the peripheral circulation correlated with the incidence and intensity of rejection (50, 51). Heeger et al. (52) also showed that pretransplant frequency of donor-specific memory T cells correlates with risk of rejection posttransplantation despite high levels of immunosuppression. All of these studies suggest that memory cells could play a large role in determining the survival of human allografts; however, direct experimental evaluation in humans systems has not been possible. We establish in this study using purified human T cell subsets that the human PBL-SCID mouse model of skin allograft injury is an example of memory T cell-dependent allograft rejection. At 3 wk following adoptive transfer of memory T cells, the grafts had intense mononuclear infiltrates and significant injury to the microvasculature; however, naive cells transferred to the same mice demonstrated no evidence of injury even when followed out to 5 wk after the adoptive transfer.

Using our model, we examined the role of 4-1BBL, ICOSL, and OX40L in human allograft injury. Several recent studies using rodent transplant models have described roles for 4-1BB, OX40, and ICOS in vivo. Genetic ablation or Ab blockade of the 4-1BB-4-1BBL pathway prolongs graft survival by reducing T cell proliferation and cytotoxicity (53, 54, 55). ICOS deficiency or blockade have also been shown in rodent models to improve transplant survival and, in conjunction with B7 blockade, even induce tolerance (56, 57), although not in all cases (58). OX40L blockade also leads to increased graft survival in presensitized mice in combination with B7 blockade (58, 59). When any of these pathways are blocked in our chimeric model of human memory, we show significantly decreased endothelial damage, vascular destruction, and effector molecule mRNA at 10 days, although none seems to be as effective as blockade of LFA-3. The likely reason for this is that LFA-3 had some effects on T cell infiltration into the grafts in addition to any of its potential costimulatory effects, although the extent of infiltration was reduced by <50%. Thus, our model is the first to demonstrate a role for 4-1BBL, ICOSL, and OX40L in human allograft rejection in vivo. To date, previous data suggested a correlation between these costimulators and graft rejection in human transplantation (60, 61) as well as in other human diseases (46), but none has explored the effects of targeting these pathways in humans.

Although 4-1BBL, ICOSL, and OX40L appear to have similar potential to reduce the allograft rejection, the effects on cytokine transcript expression in vivo reveals differences among the functions of these costimulators in vivo. Blockade of 4-1BBL decreased mRNA production of IFN-γ, IL-2, as well as IL-10, but had no effect on IL-4. This is consistent with observations that 4-1BBL activates both CD4+ and CD8+ human T cells to produce IL-2 and IFN-γ independently of B7 interactions (14). In comparison to 4-1BBL, blockade of ICOSL increased IFN-γ and only minimally reduced IL-2 production but showed a much larger decrease in IL-10 and IL-4 production. Our in vivo study correlates with in vitro studies that have indicated a limited role for ICOS in IFN-γ regulation, while playing a large part in the regulation of IL-10 (62) and IL-4 (63, 64). OX40L blockade had no effect on the level of IL-10 mRNA, while reducing all the other cytokines, suggesting a distinct mechanism of action different from either 4-1BBL or ICOSL blockade and may be related to the role that the OX40 interaction plays in the control of regulatory T cells (65, 66).

In conclusion, we have shown in vitro that HDMEC activate resting memory T cells and that this activation depends in part on the costimulatory molecules 4-1BBL, ICOSL, and OX40L. Furthermore, we demonstrate that memory T cells are the major effector population of human allograft rejection following adoptive transfer into chimeric immunodeficient mice and that memory T cell-selective costimulation plays a significant role in this model of memory T cell-dependent allograft rejection. Our studies extend the understanding of the mechanism by which human memory T cells may be activated, introduce a small animal model to study human memory T cell reactions to allografts, and identify targets for reducing the memory T cell participation in clinical allograft rejection.

We thank Bruce Fichandler for providing cadaveric skin and Louise Benson, Gwendoline Davis, and Lisa Gras for excellent assistance in cell culture and animal handling.

The authors have no financial conflict of interest.

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.

1

This work was supported by National Institutes of Health (NIH) Grant R01 HL51014. S.L.S. was supported by NIH Training Grant GM07205 during the initial phase of this study.

3

Abbreviations used in this paper: EC, endothelial cell; HDMEC, human dermal microvascular endothelial cell; ICOSL, ICOS ligand; OX40L, OX40 ligand; 4-1BBL, 4-1BB ligand; qRT-PCR, quantitative RT-PCR.

1
Chalasani, G., Z. Dai, B. T. Konieczny, F. K. Baddoura, F. G. Lakkis.
2002
. Recall and propagation of allospecific memory T cells independent of secondary lymphoid organs.
Proc. Natl. Acad. Sci. USA
99
:
6175
.-6180.
2
Hernandez-Fuentes, M. P., S. Stevenson, O. Barroso-Herrera, R. I. Lechler.
2002
. Establishing the optimal method to estimate the frequencies of alloreactive CD4+ T cells.
Transplant Proc.
34
:
2855
.-2857.
3
Epperson, D. E., J. S. Pober.
1994
. Ag-presenting function of human endothelial cells. Direct activation of resting CD8 T cells.
J. Immunol.
153
:
5402
.-5412.
4
Sallusto, F., D. Lenig, R. Forster, M. Lipp, A. Lanzavecchia.
1999
. Two subsets of memory T lymphocytes with distinct homing potentials and effector functions.
Nature
401
:
708
.-712.
5
Sallusto, F., J. Geginat, A. Lanzavecchia.
2004
. Central memory and effector memory T cell subsets: function, generation, and maintenance.
Annu. Rev. Immunol.
22
:
745
.-763.
6
Choi, J., D. R. Enis, K. P. Koh, S. L. Shiao, J. S. Pober.
2004
. T lymphocyte-endothelial cell interactions.
Annu. Rev. Immunol.
22
:
683
.-709.
7
Murray, A. G., P. Libby, J. S. Pober.
1995
. Human vascular smooth muscle cells poorly co-stimulate and actively inhibit allogeneic CD4+ T cell proliferation in vitro.
J. Immunol.
154
:
151
.-161.
8
Pober, J. S., T. Collins, M. A. Gimbrone, Jr, R. S. Cotran, J. D. Gitlin, W. Fiers, C. Clayberger, A. M. Krensky, S. J. Burakoff, C. S. Reiss.
1983
. Lymphocytes recognize human vascular endothelial and dermal fibroblast Ia antigens induced by recombinant immune interferon.
Nature
305
:
726
.-734.
9
Davis, S. J., P. A. van der Merwe.
1996
. The structure and ligand interactions of CD2: implications for T-cell function.
Immunol. Today
17
:
177
.-187.
10
Savage, C. O., C. C. Hughes, B. W. McIntyre, J. K. Picard, J. S. Pober.
1993
. Human CD4+ T cells proliferate to HLA-DR+ allogeneic vascular endothelium. Identification of accessory interactions.
Transplantation
56
:
128
.-134.
11
Rogers, P. R., J. Song, I. Gramaglia, N. Killeen, M. Croft.
2001
. OX40 promotes Bcl-xL and Bcl-2 expression and is essential for long-term survival of CD4 T cells.
Immunity
15
:
445
.-455.
12
Cooper, D., P. Bansal-Pakala, M. Croft.
2002
. 4-1BB (CD137) controls the clonal expansion and survival of CD8 T cells in vivo but does not contribute to the development of cytotoxicity.
Eur. J. Immunol.
32
:
521
.-529.
13
Yoshinaga, S. K., J. S. Whoriskey, S. D. Khare, U. Sarmiento, J. Guo, T. Horan, G. Shih, M. Zhang, M. A. Coccia, T. Kohno, et al
1999
. T-cell co-stimulation through B7RP-1 and ICOS.
Nature
402
:
827
.-832.
14
Wen, T., J. Bukczynski, T. H. Watts.
2002
. 4-1BB ligand-mediated costimulation of human T cells induces CD4 and CD8 T cell expansion, cytokine production, and the development of cytolytic effector function.
J. Immunol.
168
:
4897
.-4906.
15
Khayyamian, S., A. Hutloff, K. Buchner, M. Grafe, V. Henn, R. A. Kroczek, H. W. Mages.
2002
. ICOS-ligand, expressed on human endothelial cells, costimulates Th1 and Th2 cytokine secretion by memory CD4+ T cells.
Proc. Natl. Acad. Sci. USA
99
:
6198
.-6203.
16
Imura, A., T. Hori, K. Imada, T. Ishikawa, Y. Tanaka, M. Maeda, S. Imamura, T. Uchiyama.
1996
. The human OX40/gp34 system directly mediates adhesion of activated T cells to vascular endothelial cells.
J. Exp. Med.
183
:
2185
.-2195.
17
Pober, J. S., A. L. Bothwell, M. I. Lorber, J. M. McNiff, J. S. Schechner, G. Tellides.
2003
. Immunopathology of human T cell responses to skin, artery and endothelial cell grafts in the human peripheral blood lymphocyte/severe combined immunodeficient mouse.
Springer Semin. Immunopathol.
25
:
167
.-180.
18
Murray, A. G., P. Petzelbauer, C. C. Hughes, J. Costa, P. Askenase, J. S. Pober.
1994
. Human T-cell-mediated destruction of allogeneic dermal microvessels in a severe combined immunodeficient mouse.
Proc. Natl. Acad. Sci. USA
91
:
9146
.-9150.
19
Sultan, P., J. S. Schechner, J. M. McNiff, P. S. Hochman, C. C. Hughes, M. I. Lorber, P. W. Askenase, J. S. Pober.
1997
. Blockade of CD2-LFA-3 interactions protects human skin allografts in immunodeficient mouse/human chimeras.
Nat. Biotechnol.
15
:
759
.-762.
20
Kirkiles-Smith, N. C., K. Mahboubi, J. Plescia, J. M. McNiff, J. Karras, J. S. Schechner, D. C. Altieri, J. S. Pober.
2004
. IL-11 protects human microvascular endothelium from alloinjury in vivo by induction of survivin expression.
J. Immunol.
172
:
1391
.-1396.
21
Kluger, M. S., D. R. Johnson, J. S. Pober.
1997
. Mechanism of sustained E-selectin expression in cultured human dermal microvascular endothelial cells.
J. Immunol.
158
:
887
.-896.
22
Torpey, N., S. E. Maher, A. L. Bothwell, J. S. Pober.
2004
. Interferon α but not interleukin 12 activates STAT4 signaling in human vascular endothelial cells.
J. Biol. Chem.
279
:
26789
.-26796.
23
Ma, W., J. S. Pober.
1998
. Human endothelial cells effectively costimulate cytokine production by, but not differentiation of, naive CD4+ T cells.
J. Immunol.
161
:
2158
.-2167.
24
Sana, T. R., M. J. Janatpour, M. Sathe, L. M. McEvoy, T. K. McClanahan.
2005
. Microarray analysis of primary endothelial cells challenged with different inflammatory and immune cytokines.
Cytokine
29
:
256
.-269.
25
Hughes, C. C., C. O. Savage, J. S. Pober.
1990
. Endothelial cells augment T cell interleukin 2 production by a contact-dependent mechanism involving CD2/LFA-3 interaction.
J. Exp. Med.
171
:
1453
.-1467.
26
Goldrath, A. W., L. Y. Bogatzki, M. J. Bevan.
2000
. Naive T cells transiently acquire a memory-like phenotype during homeostasis-driven proliferation.
J. Exp. Med.
192
:
557
.-564.
27
Koh, K. P., Y. Wang, T. Yi, S. L. Shiao, M. I. Lorber, W. C. Sessa, G. Tellides, J. S. Pober.
2004
. T cell-mediated vascular dysfunction of human allografts results from IFN-γ dysregulation of NO synthase.
J. Clin. Invest.
114
:
846
.-856.
28
Strehlau, J., M. Pavlakis, M. Lipman, M. Shapiro, L. Vasconcellos, W. Harmon, T. B. Strom.
1997
. Quantitative detection of immune activation transcripts as a diagnostic tool in kidney transplantation.
Proc. Natl. Acad. Sci. USA
94
:
695
.-700.
29
Pascoe, M. D., S. E. Marshall, K. I. Welsh, L. M. Fulton, D. A. Hughes.
2000
. Increased accuracy of renal allograft rejection diagnosis using combined perforin, granzyme B, and Fas ligand fine-needle aspiration immunocytology.
Transplantation
69
:
2547
.-2553.
30
Marelli-Berg, F. M., R. E. Hargreaves, P. Carmichael, A. Dorling, G. Lombardi, R. I. Lechler.
1996
. Major histocompatibility complex class II-expressing endothelial cells induce allospecific nonresponsiveness in naive T cells.
J. Exp. Med.
183
:
1603
.-1612.
31
Lemstrom, K., P. Koskinen, P. Hayry.
1995
. Induction of adhesion molecules on the endothelia of rejecting cardiac allografts.
J. Heart Lung Transplant.
14
:
205
.-213.
32
Jollow, K. C., J. C. Zimring, J. B. Sundstrom, A. A. Ansari.
1999
. CD40 ligation induced phenotypic and functional expression of CD80 by human cardiac microvascular endothelial cells.
Transplantation
68
:
430
.-439.
33
Rose, M. L., M. I. Coles, R. J. Griffin, A. Pomerance, M. H. Yacoub.
1986
. Expression of class I and class II major histocompatibility antigens in normal and transplanted human heart.
Transplantation
41
:
776
.-780.
34
Biedermann, B. C., J. S. Pober.
1998
. Human endothelial cells induce and regulate cytolytic T cell differentiation.
J. Immunol.
161
:
4679
.-4687.
35
Pardi, R., J. R. Bender.
1991
. Signal requirements for the generation of CD4+ and CD8+ T-cell responses to human allogeneic microvascular endothelium.
Circ. Res.
69
:
1269
.-1279.
36
Karmann, K., C. C. Hughes, J. Schechner, W. C. Fanslow, J. S. Pober.
1995
. CD40 on human endothelial cells: inducibility by cytokines and functional regulation of adhesion molecule expression.
Proc. Natl. Acad. Sci. USA
92
:
4342
.-4346.
37
Kunitomi, A., T. Hori, A. Imura, T. Uchiyama.
2000
. Vascular endothelial cells provide T cells with costimulatory signals via the OX40/gp34 system.
J. Leukocyte Biol.
68
:
111
.-118.
38
Alderson, M. R., C. A. Smith, T. W. Tough, T. Davis-Smith, R. J. Armitage, B. Falk, E. Roux, E. Baker, G. R. Sutherland, W. S. Din.
1994
. Molecular and biological characterization of human 4-1BB and its ligand.
Eur. J. Immunol.
24
:
2219
.-2227.
39
Ling, V., P. W. Wu, H. F. Finnerty, K. M. Bean, V. Spaulding, L. A. Fouser, J. P. Leonard, S. E. Hunter, R. Zollner, J. L. Thomas, et al
2000
. Cutting edge: identification of GL50, a novel B7-like protein that functionally binds to ICOS receptor.
J. Immunol.
164
:
1653
.-1657.
40
Ohshima, Y., Y. Tanaka, H. Tozawa, Y. Takahashi, C. Maliszewski, G. Delespesse.
1997
. Expression and function of OX40 ligand on human dendritic cells.
J. Immunol.
159
:
3838
.-3848.
41
Aicher, A., M. Hayden-Ledbetter, W. A. Brady, A. Pezzutto, G. Richter, D. Magaletti, S. Buckwalter, J. A. Ledbetter, E. A. Clark.
2000
. Characterization of human inducible costimulator ligand expression and function.
J. Immunol.
164
:
4689
.-4696.
42
Beier, K. C., A. Hutloff, A. M. Dittrich, C. Heuck, A. Rauch, K. Buchner, B. Ludewig, H. D. Ochs, H. W. Mages, R. A. Kroczek.
2000
. Induction, binding specificity and function of human ICOS.
Eur. J. Immunol.
30
:
3707
.-3717.
43
Godfrey, W. R., F. F. Fagnoni, M. A. Harara, D. Buck, E. G. Engleman.
1994
. Identification of a human OX-40 ligand, a costimulator of CD4+ T cells with homology to tumor necrosis factor.
J. Exp. Med.
180
:
757
.-762.
44
Ukyo, N., T. Hori, S. Yanagita, T. Ishikawa, T. Uchiyama.
2003
. Costimulation through OX40 is crucial for induction of an alloreactive human T-cell response.
Immunology
109
:
226
.-231.
45
Greenwald, R. J., G. J. Freeman, A. H. Sharpe.
2005
. The B7 family revisited.
Annu. Rev. Immunol.
23
:
515
.-548.
46
Watts, T. H..
2005
. Tnf/Tnfr family members in costimulation of T cell responses.
Annu. Rev. Immunol.
23
:
23
.-68.
47
Lakkis, F. G., M. H. Sayegh.
2003
. Memory T cells: a hurdle to immunologic tolerance.
J. Am. Soc. Nephrol.
14
:
2402
.-2410.
48
Adams, A. B., M. A. Williams, T. R. Jones, N. Shirasugi, M. M. Durham, S. M. Kaech, E. J. Wherry, T. Onami, J. G. Lanier, K. E. Kokko, et al
2003
. Heterologous immunity provides a potent barrier to transplantation tolerance.
J. Clin. Invest.
111
:
1887
.-1895.
49
Chen, Y., P. S. Heeger, A. Valujskikh.
2004
. In vivo helper functions of alloreactive memory CD4+ T cells remain intact despite donor-specific transfusion and anti-CD40 ligand therapy.
J. Immunol.
172
:
5456
.-5466.
50
Salom, R. N., J. A. Maguire, D. Esmore, W. W. Hancock.
1994
. Analysis of proliferating cell nuclear antigen expression aids histological diagnosis and is predictive of progression of human cardiac allograft rejection.
Am. J. Pathol.
145
:
876
.-882.
51
Ibrahim, S., D. V. Dawson, F. Sanfilippo.
1995
. Predominant infiltration of rejecting human renal allografts with T cells expressing CD8 and CD45RO.
Transplantation
59
:
724
.-728.
52
Heeger, P. S., N. S. Greenspan, S. Kuhlenschmidt, C. Dejelo, D. E. Hricik, J. A. Schulak, M. Tary-Lehmann.
1999
. Pretransplant frequency of donor-specific, IFN-γ-producing lymphocytes is a manifestation of immunologic memory and correlates with the risk of posttransplant rejection episodes.
J. Immunol.
163
:
2267
.-2275.
53
Blazar, B. R., B. S. Kwon, A. Panoskaltsis-Mortari, K. B. Kwak, J. J. Peschon, P. A. Taylor.
2001
. Ligation of 4–1BB (CDw137) regulates graft-versus-host disease, graft-versus-leukemia, and graft rejection in allogeneic bone marrow transplant recipients.
J. Immunol.
166
:
3174
.-3183.
54
DeBenedette, M. A., T. Wen, M. F. Bachmann, P. S. Ohashi, B. H. Barber, K. L. Stocking, J. J. Peschon, T. H. Watts.
1999
. Analysis of 4-1BB ligand (4-1BBL)-deficient mice and of mice lacking both 4-1BBL and CD28 reveals a role for 4-1BBL in skin allograft rejection and in the cytotoxic T cell response to influenza virus.
J. Immunol.
163
:
4833
.-4841.
55
Cho, H. R., B. Kwon, H. Yagita, S. La, E. A. Lee, J. E. Kim, H. Akiba, J. Kim, J. H. Suh, D. S. Vinay, et al
2004
. Blockade of 4-1BB (CD137)/4-1BB ligand interactions increases allograft survival.
Transpl. Int.
17
:
351
.-361.
56
Ozkaynak, E., W. Gao, N. Shemmeri, C. Wang, J. C. Gutierrez-Ramos, J. Amaral, S. Qin, J. B. Rottman, A. J. Coyle, W. W. Hancock.
2001
. Importance of ICOS-B7RP-1 costimulation in acute and chronic allograft rejection.
Nat. Immunol.
2
:
591
.-596.
57
Harada, H., A. D. Salama, M. Sho, A. Izawa, S. E. Sandner, T. Ito, H. Akiba, H. Yagita, A. H. Sharpe, G. J. Freeman, M. H. Sayegh.
2003
. The role of the ICOS-B7h T cell costimulatory pathway in transplantation immunity.
J. Clin. Invest.
112
:
234
.-243.
58
Demirci, G., F. Amanullah, R. Kewalaramani, H. Yagita, T. B. Strom, M. H. Sayegh, X. C. Li.
2004
. Critical role of OX40 in CD28 and CD154-independent rejection.
J. Immunol.
172
:
1691
.-1698.
59
Yuan, X., A. D. Salama, V. Dong, I. Schmitt, N. Najafian, A. Chandraker, H. Akiba, H. Yagita, M. H. Sayegh.
2003
. The role of the CD134-CD134 ligand costimulatory pathway in alloimmune responses in vivo.
J. Immunol.
170
:
2949
.-2955.
60
Akalin, E., S. Dikman, B. Murphy, J. S. Bromberg, W. W. Hancock.
2003
. Glomerular infiltration by CXCR3+ ICOS+ activated T cells in chronic allograft nephropathy with transplant glomerulopathy.
Am. J. Transplant.
3
:
1116
.-1120.
61
Streeter, P. R., X. Zhang, T. V. Tittle, C. N. Schon, A. D. Weinberg, R. T. Maziarz.
2004
. CD25 expression distinguishes functionally distinct alloreactive CD4 CD134 (OX40) T-cell subsets in acute graft-versus-host disease.
Biol. Blood Marrow Transplant.
10
:
298
.-309.
62
Witsch, E. J., M. Peiser, A. Hutloff, K. Buchner, B. G. Dorner, H. Jonuleit, H. W. Mages, R. A. Kroczek.
2002
. ICOS and CD28 reversely regulate IL-10 on re-activation of human effector T cells with mature dendritic cells.
Eur. J. Immunol.
32
:
2680
.-2686.
63
Nurieva, R. I., X. M. Mai, K. Forbush, M. J. Bevan, C. Dong.
2003
. B7h is required for T cell activation, differentiation, and effector function.
Proc. Natl. Acad. Sci. USA
100
:
14163
.-14168.
64
Tafuri, A., A. Shahinian, F. Bladt, S. K. Yoshinaga, M. Jordana, A. Wakeham, L. M. Boucher, D. Bouchard, V. S. Chan, G. Duncan, et al
2001
. ICOS is essential for effective T-helper-cell responses.
Nature
409
:
105
.-109.
65
Takeda, I., S. Ine, N. Killeen, L. C. Ndhlovu, K. Murata, S. Satomi, K. Sugamura, N. Ishii.
2004
. Distinct roles for the OX40-OX40 ligand interaction in regulatory and nonregulatory T cells.
J. Immunol.
172
:
3580
.-3589.
66
Valzasina, B., C. Guiducci, H. Dislich, N. Killeen, A. D. Weinberg, M. P. Colombo.
2004
. Triggering of OX40 (CD134) on CD4+CD25+ T cells blocks their inhibitory activity: a novel regulatory role for OX40 and its comparison with GITR.
Blood
105
:
2845
.-2851.