Cytotoxic T lymphocytes reactive with endothelium have

Transcription

1 Human Vascular Endothelial Cells Stimulate Memory But Not Naive CD8 T Cells to Differentiate into CTL Retaining an Early Activation Phenotype 1 Thomas J. Dengler and Jordan S. Pober 2 Endothelial cell (EC)-selective alloreactive CTL may mediate alloimmune vascular injury. In the present study, EC-selective CTL were generated in cocultures of purified human CD8 T cells with allogeneic EC and were compared with conventional CTL against corresponding B lymphoblastoid cells (BLC). EC caused activation and expansion of memory but not naive CD8 T cells, which differentiated into EC-selective CTL that retained high surface expression of CD69, CD25, and CD62L and displayed low intracellular perforin content. In contrast, BLC-stimulated CTL could be generated from naive or memory CD8 T cells and showed a more mature phenotype (low CD69, CD25, and CD62L with higher levels of perforin). The expansion of alloreactive T cells by EC stimulation was 5- to 20-fold less effective than in corresponding BLC-stimulated cultures, accounting for a reduction in the assayable cytotoxicity of individual microcultures. In these IL-2-supplemented cocultures, no effect on CTL generation or phenotype was observed by mab blocking of costimulation provided by LFA-3, ICAM-1, or CD40, by addition of comitogenic anti-cd28 mab, or by preactivation of EC with CD40 ligand. Cyclosporine inhibited CTL expansion and cytotoxicity similarly in both EC- and BLC-stimulated cultures but did not affect the phenotype of those CTL that did emerge. This study extends the characterization of endothelium as an immunoregulatory cell type distinct from conventional APC and may explain why graft rejection within the arterial intima, an anatomic compartment in which EC may be the primary type of APC, is separable from rejection in the graft parenchyma. The Journal of Immunology, 2000, 164: Cytotoxic T lymphocytes reactive with endothelium have been proposed to be the effectors of cell-mediated vascular rejection (endothelialitis), a potential precursor lesion of chronic graft rejection (transplant vasculopathy) (1, 2). Endothelial cell (EC) 3 -selective, alloreactive CTL have been isolated from endomyocardial biopsies of acutely rejecting heart transplants (3), and it is likely that such EC-selective CTL arise in response to alloantigen presented by graft EC. Previous studies from our laboratory have demonstrated that allogeneic human EC stimulators can generate EC-selective CTL with atypical characteristics, including weak cytotoxic activity and absent IFN- secretion (4). Further studies of EC-stimulated CTL clones confirmed these unusual features, especially EC selectivity and low production of IFN-, which correlated closely with constitutive surface expression of CD40 ligand (CD154) (5), a molecule that is not usually seen on CD8 T cells. Although our initial studies highlighted the capacity of EC to stimulate an unusual CTL response, these analyses were limited to Interdepartmental Program in Vascular Biology and Transplantation, Boyer Center for Molecular Medicine, Yale University School of Medicine, New Haven, CT Received for publication December 20, Accepted for publication March 3, 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 grants from the Deutsche Forschungsgemeinschaft (DE 591/2-1) and National Institutes of Health (HL-51014). 2 Address correspondence and reprint requests to Dr. Jordan S. Pober, Boyer Center for Molecular Medicine, Yale University School of Medicine, 295 Congress Avenue, New Haven, CT address: jordan.pober@yale.edu 3 Abbreviations used in this paper: EC, endothelial cell; BLC, B lymphoblastoid cell; CFSE, 5-(and -6)-carboxyfluorecein diacetate succinimidyl ester; CsA, cyclosporine; CD62L, CD62 ligand; CD40L, CD40 ligand. a relatively late time point, i.e., after CD8 T cell clonal expansion had progressed sufficiently to permit measurement of cytolysis or after cloning by limiting dilution with further propagation in culture. In the present study, we have used 5-(and -6)-carboxyfluorecein diacetate succinimidyl ester (CFSE) labeling to identify and selectively analyze proliferating lymphocytes (6), which allowed us to look at earlier events in the primary coculture system. We have used this approach to compare EC with professional APC, e.g., B lymphoblastoid cells (BLC), as stimulators of CTL by analyzing the kinetics of clonal expansion, the identity of the resting T cells that are able to respond to alloantigen by proliferation, and the phenotype of the proliferating T cell populations during their clonal expansion. We have also been able to analyze the effects of costimulation blockade, costimulatory augmentation, and cyclosporine (CsA) on both clonal expansion and CTL maturation. Our experiments show that EC stimulate a smaller number of alloreactive CD8 T cells drawn entirely from the memory T cell subset compared with BLC, which stimulate expansion of both naive and memory alloreactive T cells. Interestingly, EC-stimulated CTL appear to retain an early activation phenotype characterized by persistent expression of CD69, CD25, and L-selectin (CD62L). In the presence of exogenous IL-2, which is required for CTL generation in these cocultures with EC or BLC stimulators, we have not detected a role for costimulation of CD8 T cells through B7, LFA-3, ICAM-1, or CD40. CsA inhibits EC- and BLCstimulated cocultures primarily by reducing clonal expansion despite the presence of exogenous IL-2 and does not appear to significantly affect other CTL differentiation events. We conclude that EC cause a form of immune deviation resulting in activation of unusual CTL effector populations that could be important for allogeneic responses that develop within the graft vessel wall. Copyright 2000 by The American Association of Immunologists /00/$02.00

2 The Journal of Immunology Materials and Methods Cell isolation PBMC were obtained from healthy volunteers by density-gradient centrifugation of leukapheresis products and were stored in liquid nitrogen as described previously (7). CD8 T cells were isolated from PBMC by positive selection (4) using Dynabeads (Dynal, Lake Success, NY). The selected population obtained by this procedure was routinely 98% CD8 / CD3 by flow cytometry and 99% viable, as shown by trypan blue exclusion. Use of positively selected CD8 T cells in this assay had previously been shown to yield identical results compared with negative selection (4). Naive and memory subsets of T cells were isolated from the CD8-selected population by further negative selection. CD8 T cells were incubated with anti-cd45ra (B-C15; 1 l/10 7 cells) or anti-cd45ro mab (5 l/10 7 cells; UCHL-1, both mab from Biosource, Camarillo, CA), respectively, for 30 min at 4 C. After three washes, cells were incubated with magnetic beads coated with goat anti-mouse IgG antiserum (Dynal) for 30 min at 4 C. Cells not attached to beads were recovered during application of a magnet; negatively selected populations obtained in this manner were routinely 95% double-positive for CD8 and either CD45RA or CD45RO, respectively, and were 99% viable. HUVEC were isolated from umbilical cords by enzymatic digestion and maintained in culture, as described (8). EC cultures were free from detectable CD45 contaminating leukocytes and uniformly expressed von Willebrand factor and CD31. BLC from the same donors as the HUVEC were generated by EBV immortalization of cord blood mononuclear cells (4). After 6 8 wk in culture, BLC lines were uniformly CD19-positive. HUVEC and BLC of the same donor were used as stimulator cells in cocultures and as target cells in cytotoxicity assays. When BLC were used as stimulator cells in cocultures, they were pretreated with mitomycin C (50 g/ml in PBS, 30 min; Sigma-Aldrich, St. Louis, MO) to prevent proliferation. This treatment did not affect the outcome of cultures stimulated with HUVEC and was routinely omitted. CTL generation The procedure for CTL differentiation has been described in detail elsewhere (4). In brief, purified CD8 T cells were incubated with EC or BLC stimulator cells in 96-well microculture plates (Falcon; Becton Dickinson, Bedford, MA) at responder-to-stimulator cell numbers of 200,000/20,000 for EC cocultures and 100,000/20,000 for BLC cocultures. A total of microculture replicates were initiated for each experimental group within a given responder-stimulator combination. All cultures were maintained in 5% CO 2 room air at 37 C. The medium for coculture consisted of RPMI 1640 supplemented with 10% human AB serum (Irvine Scientific, Santa Ana, CA), 2 mm L-glutamine, 100 U/ml penicillin, and 100 g/ml streptomycin. The medium was further supplemented with exogenous IL-2 (R&D Systems, Minneapolis, MN; or National Cancer Institute, Frederick, MD) on day 3 (final concentration, 5 ng/ml (12.5 U/ml, R&D conversion table)). On day 7, T cells were transferred to fresh stimulator cells in fresh medium containing IL-2, and cultures were fed with fresh medium containing IL-2 on day 10. On day 14, CTL were harvested from microculture wells, counted manually in a hemocytometer, and subjected to further analyses described below either from individual wells or after pooling cells from 3 12 microwells. CD8 T cells in the absence of any stimulator cells but receiving the same IL-2 treatment were cultured in replicate wells as unstimulated controls. In experiments involving the addition of blocking mab, mab were added at the initiation of cocultures at final concentrations indicated below; treatment with mab was repeated at times of medium change on days 3, 7, and 10. The mab used were TS2/9 (anti-cd58; 1:200 dilution of ascites; gift of T. Springer, Center for Blood Research, Boston, MA), M3 (anti- CD40; 10 g/ml; gift of W. C. Fanslow, Immunex, Seattle, WA), and D5 (anti-icam-1; 10 g/ml; gift of Dr. D. Altieri, Yale Medical School, New Haven, CT). The inhibitory potential of mabs used in these studies had been extensively demonstrated previously (9, 10); saturating concentrations of all mab were used. To exclude interference of blocking Abs with cytotoxicity assays, recovered cells were washed before use. For costimulation experiments, anti-cd28 mab (clone ; R&D Systems) was added at the initiation of cocultures at 10 g/ml and again at times of medium change on days 3, 7, and 10. For preactivation of EC with CD40 ligand, CD40 ligand trimer (gift from W. C. Fanslow) was added to EC at 100 ng/ml for 24 h before initiation of cocultures. For experiments involving CsA, Sandimmune Neoral (liquid formulation, 100 mg/ml; Novartis, East Hanover, NJ) was added at initiation of cocultures or 4 h before cytotoxicity assays at 500 ng/ml. Cells recovered from cocultures were washed before use in cytotoxicity assays. Cytotoxicity assay 5147 Cytotoxicity by CTL was measured by a calcein fluorescent dye release assay as described previously (4, 11). In brief, target cells were loaded with calcein-am (Molecular Probes, Eugene, OR) and incubated with effector CTL from individual microculture wells for 4hat37 C. In experiments involving CsA or mab, CTL were washed three times before addition to cytotoxicity assays. The supernatant was harvested and calcein release was quantitated with a fluorescence plate reader (Cytofluor 2; Perseptive Biosystems, Framingham, MA; excitation wavelength, 485 nm; emission wavelength, 530 nm). Killing was considered positive for individual microculture wells if released fluorescence in a sample well exceeded the mean 3 SD of the spontaneous release. Percent specific killing was calculated as follows: (release sample spontaneous release)/(maximal release spontaneous release) 100. Spontaneous release was obtained by medium alone, and maximal release was obtained by adding lysis buffer (50 mm sodium borate and 0.1% Triton X-100 (ph 9)). Corresponding BLC or EC were used as targets for BLC-stimulated or EC-stimulated cultures, respectively. In most experiments, the cytotoxicity of EC-stimulated cultures against BLC autologous to the EC was also measured to confirm EC selectivity. Immunophenotyping of CTL from microcultures For flow cytometric analysis, T cells were harvested from microculture wells, and cells were further processed either from individual wells or after pooling of cells from 6 18 wells, as indicated. Cells were washed twice in ice-cold PBS/1% BSA and incubated with saturating concentrations of directly FITC- or PE-conjugated mouse anti-human CD3, CD8, CD25, CD69, CD45RA/RO, CD62L, HLA-DR (all from Coulter Immunotech, Miami, FL), or CD154 (TRAP-1; PharMingen, San Diego, CA) for 1hat 4 C. Cells were washed twice in ice-cold PBS/1% BSA and fixed with 1% paraformaldehyde in PBS. Samples were analyzed using FACSort (Becton Dickinson, San Jose, CA) and CellQuest analysis software by gating on viable cells and collecting 5000 gated events per sample. For intracellular staining for flow cytometric analysis, cells were washed twice in ice-cold PBS and fixed in 4% paraformaldehyde in PBS for 15 min at 4 C. After washing twice, cells were incubated for 15 min in permeabilization solution consisting of 1% FCS, 0.1% saponin (Sigma-Aldrich), and 0.1% sodium azide. Cells were then incubated with primary unconjugated mouse anti-human perforin mab (T-Cell Sciences, Woburn, MA) in permeabilization solution for 1hat4 C. After washing twice in permeabilization solution, cells were incubated with PE-conjugated goat anti-mouse IgG antiserum for 1hat4 C. Cells were finally washed twice in permeabilization solution and once in PBS/1% BSA and stored in PBS/1% BSA until analysis by flow cytometry as above. For immunofluorescence microscopy, CTL were fixed, spun onto glass slides, permeabilized, and stained as described previously (4). Cells were examined with a Microphot FXA microscope (Nikon, Natick, MA). Labeling studies of CD8 T cells with CFSE Labeling and FACS analysis of CD8 T cells were performed as described elsewhere (6, 12). In brief, total CD8 T cells or separated CD45RA / RO subsets were washed twice in cold PBS and then incubated in 0.25 M CFSE (Molecular Probes) in PBS at 37 C for 15 min. Cells were again washed twice before addition to cocultures. At indicated times, CFSElabeled cells were collected, washed in PBS/1% BSA twice, and fixed in PBS/1% paraformaldehyde. For two-color fluorescence studies, CFSE-labeled cells were incubated with saturating concentrations of directly PEconjugated mouse anti-human mabs reactive with CD3, CD8, CD69, CD25, CD62L, CD45RA, CD45RO, HLA-DR, and CD154 for 30 min at 4 C, washed twice, and subjected to flow cytometric analysis. Unlabeled CD8 T cells and cells fixed immediately after CFSE labeling were used as negative and positive internal controls for staining, respectively. Results T cell expansion after stimulation with EC or BLC To analyze the extent of lymphocyte expansion after stimulation with allogeneic EC or BLC, CD8 T cells were labeled with CFSE before addition to microwell cocultures. Fig. 1 is a representative example of two-color FACS analysis of surface CD8 labeled with PE-conjugated mab and CFSE content after 14 days of coculture with EC or BLC. Unstimulated CFSE-labeled CD8 T cells and unlabeled CD8 T cells are included as positive and negative staining controls. In both types of cocultures, CD8 T cells fall into discrete CFSE high and CFSE low populations. CFSE high cells,

3 5148 EARLY EVENTS IN ENDOTHELIAL STIMULATION OF CTL FIGURE 1. CD8 T cell division in BLC- or EC-stimulated cocultures analyzed by CFSE labeling. CFSE-labeled CD8 T cells were harvested from BLC- or EC-stimulated microwell cocultures and analyzed by two-color fluorescence flow cytometry as described in Materials and Methods. CFSE low cell populations represent alloactivated cells that have undergone multiple cell divisions. CFSE-labeled unstimulated CD8 T cells and unlabeled CD8 T cells were analyzed as controls. Note that CFSE fluorescence intensity of alloactivated (expanded) cells in BLC- and EC-stimulated cocultures have declined to the level of background signal obtained using unlabeled cells. One of five similar experiments is shown. recovered from cocultures with either type of stimulator cell, express the same intensity of CFSE fluorescence as unstimulated CD8 T cells or CD8 T cells immediately after labeling (not shown), indicating that no cell divisions have occurred in these populations. CFSE low /CD8 T cells show a fluorescence intensity comparable to that of unlabeled CD8 T cells. These populations represent cells that have been activated by alloantigen and have undergone multiple cell divisions such that there has been a dilution of the CFSE signal below the limit of detection. Although the number of cell divisions that these alloactivated CD8 T cells have undergone cannot be exactly defined, a minimum of seven to eight rounds of cell divisions is required to dilute CFSE to background levels of fluorescence. Further enhancement of the CFSE signal to detect more rounds of division was not possible because higher concentrations of CFSE inhibited CD8 T cell proliferation (not shown). The absence of cells with intermediate CFSE labeling, which are readily detectable in PHA-activated cultures (our unpublished observation), is consistent with the interpretation that the alloreactive precursor populations of CD8 T cells (as low as 1:20,000 for EC stimulators) are too small to be detectable until multiple divisions have occurred (7). Absolute cell numbers for the alloactivated and nonactivated populations of CD8 T cells were determined by multiplying the relative percentage of such cells, determined by CFSE labeling analysis, with the total number of T cells per microwell determined by counting with a hemocytometer. To facilitate comparison between EC- and BLC-stimulated microcultures initiated with inputs of 200,000 and 100,000 CD8 T cells per well, respectively, absolute cell numbers are expressed as a percentage of the input cell number. Using this method, microwell cocultures of CD8 T cells stimulated with EC or corresponding BLC were followed over several time points during the 14-day coculture period (Fig. 2). On day 3 of the coculture period, no significant population of CFSE low cells was detectable either after EC or BLC stimulation (not shown). With BLC stimulators, significant numbers of alloactivated (CFSE low ) T cells became detectable on day 7, expanding until day 10 and then either leveling off in absolute number (Fig. 2, Exp. A) or expanding further (Fig. 2, Exp. B) until day 14. Total numbers of expanded CD8 T cells always reached at least 100% of input cell numbers with BLC stimulation. On day 14 of coculture, essentially all microculture wells stimulated with BLC con- FIGURE 2. Expansion of CD8 T cells in cocultures stimulated with BLC or EC. Cells were harvested from cocultures containing CFSE-labeled CD8 T cells at indicated time points, counted, and analyzed by two-color flow cytometry. Left panels, The relative proportion of alloactivated (expanded) cells in cocultures (percentage of CFSE low cells of total CD8 T cells) was combined with the total number of cells harvested per microwell (open bars) to yield absolute numbers for alloactivated, expanded CD8 T cells (filled bars), both of which are expressed as a percentage of the original input cell number per well. Right panels, Proportion of BLC- or ECstimulated coculture microwells with detectable significant cytotoxicity against corresponding target cells for respective cell expansion experiments. Two of five similar experiments are shown.

4 The Journal of Immunology 5149 FIGURE 3. Expression of surface markers and perforin in alloactivated CD8 T cells after 14 days of stimulation with BLC or EC. A, Expression of surface markers on alloactivated (CFSE low ; cf Fig. 1 and text) cells stimulated with BLC or EC; for comparison, expression on unstimulated cells (total population) is shown. B, Two-color FACS analysis for CFSE and perforin of cells from BLCor EC-stimulated cocultures. One representative experiment of three similar experiments is shown. sistently tested positive for cytotoxicity (Fig. 2, insets) with mean specific killing exceeding 30%. With EC stimulation, low but significant numbers of alloactivated, expanded CD8 T cells only became detectable on day 10. Such cells usually expanded moderately further until day 14. However, total cell numbers of alloactivated (CFSE low ) cells after EC stimulation never reached more than 50 60% of input cell numbers. After 14 days, 15 25% of microculture wells stimulated with EC were positive for cytotoxicity (Fig. 2, insets), and mean specific killing remained below 15%. Control experiments, demonstrating the absence of cytotoxicity of EC-stimulated CTL vs corresponding BLC, confirmed the EC specificity of killing in these cocultures as reported previously (Ref. 4 and not shown). Extension of EC-stimulated cocultures beyond day 14 by restimulation and refeeding failed to further enhance the absolute number of alloactivated cells or the extent of cytotoxicity. Indeed, in most EC-stimulated cultures the absolute and relative sizes of the expanded (CFSE low ) cell pool actually contracted between day 14 and day 21 (not shown). The number of surviving, nondividing (CFSE high ) CD8 T cells (relative to input cell number) was not significantly different between BLC or EC stimulation (Fig. 1). These data show that BLC stimulation of CTL consistently displayed significantly greater expansion of relative and absolute numbers of alloactivated cells than EC stimulation. This greater expansion of the alloactivated population correlates with and can explain the higher degree of measured cytotoxic activity of cells recovered from BLC-stimulated cocultures. Immunophenotype of CD8 T cells after stimulation with BLC or EC After 14 days of microwell coculture, recovered CD8 T cells were analyzed by two-color flow cytometry for CFSE fluorescence levels and for expression of several surface activation markers or intracellular perforin expression. CFSE labeling of CD8 T cells in these experiments allowed the separate analysis of surface Ags and perforin expression for unactivated (CFSE high ) and for alloactivated, expanded (CFSE low ) cells. At 14 days after stimulation with BLC, the alloactivated CD8 T cells were almost exclusively CD45RO (memory phenotype) and expressed high levels of HLA-DR and very low levels of CD25 and CD69. These last two markers, indicative of early activation, had fallen to levels comparable to those of unstimulated, IL-2-treated control cells (Fig. 3A). Expression of L-selectin (CD62L) was significantly reduced compared with that in control cells but was not completely absent. In day 14 cultures stimulated with EC, the alloactivated, expanded CD8 T cells (CFSE low ) were also exclusively of CD45RO phenotype and expressed high levels of HLA-DR. In striking contrast to BLC-stimulated CD8 T cells, alloactivated (CFSE low ) T cells in EC-stimulated cultures retained high expression levels of the early activation markers CD69 and CD25. In addition, almost all activated cells were positive for CD62L, an even higher percentage than that observed in unstimulated control cells (Fig. 3A). In cocultures extended beyond 14 days, these phenotypic differences were essentially maintained (not shown); in most EC-stimulated cocultures, the number of alloactivated T cells began to decline after more than 14 days (see above). Comparisons of 10-day cultures, i.e., 3 days after restimulation, showed some expression of CD25 and CD69 in BLC-stimulated cultures, confirmimg that the low levels seen in day 14 cultures did represent a decline from peak levels. However, even at day 10 the level and percentage of positive cells were still markedly lower than those in EC-stimulated cultures (not shown). In BLC- or EC-stimulated cocultures, CD154 was not detectable on alloactivated (CFSE low ) cells by flow cytometry (Fig. 3A). The expression of perforin by alloactivated (CFSE low ) CD8 T cells after stimulation with BLC or corresponding EC was qualitatively similar but differed significantly in quantitative analysis (Fig. 3B). After stimulation with either cell type, 90% of the alloactivated CD8 T cell population expressed detectable levels of perforin; however, the expression level per cell was markedly higher after stimulation with BLC. Nondividing CD8 T cells in coculture and unstimulated, IL-2-treated control cells did not show detectable perforin expression (not shown). Activation of CD45RA (naive) and/or CD45RO (memory) subsets Although all of the alloactivated CD8 T cells recovered at day 14 expressed CD45RO, these cells could have arisen from resting memory (CD45RO ) or resting naive (CD45RA ) subsets or both. To directly identify the subset(s) of CD8 T cells that were activated by stimulation with BLC or EC, cocultures were set up

5 5150 EARLY EVENTS IN ENDOTHELIAL STIMULATION OF CTL using CFSE-labeled, purified CD45RA or CD45RO CD8 resting T cells. As in the case of unseparated CD8 T cells, populations were analyzed after 14 days by two-color flow cytometry and cytotoxicity assays. After stimulation with BLC, the proportion and absolute numbers of alloactivated (CFSE low ) CD8 T cells were similar in both subsets, indicating that both naive and memory CD8 T cells were activated. The expression of surface activation markers and perforin was very similar for both subsets after BLC stimulation and was comparable to the findings from studies of unseparated CD8 T cells (Fig. 4A; perforin not shown), namely showing low expression of CD25 and CD69. A high proportion ( 90%) of cells in the CD45RA (naive) subset had become CD45RA/RO double-positive after activation. These results differ from those obtained after stimulation with EC, when alloactivated CD8 T cells (CFSE low ) were only detectable in the CD45RO (memory) subset (Fig. 4A). Expression of surface activation markers was similar to findings with unseparated CD8 T cells (Fig. 4A) showing high levels of CD25 and CD69 in alloactivated cells. However, all alloactivated cells expressed perforin again at a markedly lower level than after stimulation with BLC (not shown). Cytotoxicity assays with CTL recovered from the CD45 subset cocultures were concordant with perforin expression: 100% of microwell cocultures of CD45RA and CD45RO phenotype stimulated with BLC displayed comparable, significant cytotoxicity. In contrast, after stimulation with EC, cytotoxicity was only detectable in the CD45RO subset (Fig. 4B). Similarly to experiments with unseparated CD8 T cells, only about 20% of EC-stimulated microwell cocultures showed detectable cytotoxicity (Fig. 4B), and the percentage of specific killing was markedly lower than that observed in BLC-stimulated cocultures (not shown). FIGURE 4. Stimulation of CD45RA (naive) or CD45RO (memory) CD8 T cells by EC or BLC. BLC- or EC-stimulated cocultures of CFSElabeled purified CD45RA or CD45RO subsets of CD8 T cells were analyzed by two-color FACS and in cytotoxicity assays. A, Surface expression of CD45 isoforms and activation markers on alloactivated, expanded cells after 14 days of coculture, identified by loss of CFSE signal (cf Fig. 1)., No CFSE low, alloactivated cells were detectable in EC-stimulated cocultures of CD45RA CD8 T cells. B, Proportion of coculture microwells with detectable significant cytotoxicity against corresponding target cells from indicated cocultures (day 14). One representative experiment of three similar experiments is shown. Effect of inhibition or augmentation of costimulation upon ECstimulated CTL differentiation Previous studies have shown that neither differences in production of IL-12 nor MHC class I expression levels could account for the differences in the capacities of EC vs BLC to stimulate and expand CTL (4). BLC also differ from EC regarding the costimulator molecules expressed by these cell types (13), and these differences are reflected in the pathways used to stimulate IL-2 production. The influence of CD8 T cell costimulation by EC upon CTL generation was examined in cocultures using blocking mabs. Because these cocultures were also fed routinely with exogenous IL-2, these blocking experiments examine features of T cell activation and differentiation independent of IL-2 synthesis. Inhibition of costimulation with mabs vs LFA-3, ICAM-1, or CD40 did not influence any of the parameters analyzed after 14 days of coculture, including T cell numbers, extent of cytotoxicity, perforin positivity and expression level, and expression of surface activation markers (Fig. 5). As the generation of CTL by stimulation with EC had consistently been less effective than by corresponding BLC, we examined whether enhanced costimulation during stimulation of CD8 T cells could improve the CTL response. Because EC do not express the CD28 ligands CD80 and CD86, additional costimulation was provided via a comitogenic anti-cd28 mab. The comitogenicity of the mab used in soluble form in these studies had been confirmed in pilot experiments with purified CD8 T cells stimulated with anti-cd3 mab or PHA (not shown). However, addition of comitogenic concentrations of an anti-cd28 mab did not increase the generation of CTL by stimulation with EC as assessed by cell numbers, extent of cytotoxicity, perforin positivity and expression level, and expression of surface activation markers (Fig. 6A). An important role of CD40 ligation for activation of APC to induce CTL has been reported (14, 15). Therefore, cocultures were performed using EC preactivated via CD40 ligation for 24 h. Activity of the recombinant trimeric CD40 ligand molecule had been demonstrated in EC previously (Ref. 16) and our unpublished observations). Preactivation of EC used as stimulator cells in coculture assays had no effect on the differentiation of CTL in that all examined parameters were unchanged between untreated cocultures and cocultures after pretreatment with CD40 ligand trimer (Fig. 6B). Effect of CsA on CTL differentiation To address the influence of the immunosuppressant CsA on the differentiation of CTL after stimulation with BLC or EC, cocultures were performed in the presence of 500 ng/ml of CsA added at various time points after coculture initiation. In the presence of CsA, the absolute number of alloactivated cells (i.e., CFSE low ) was markedly reduced with both types of stimulator cells (Fig. 7), suggesting general inhibition of the expansion/proliferation of alloactivated CD8 T cells even in the presence of exogenous IL-2. In case of stimulation with BLC, the absolute size of the alloactivated (CFSE low ) cell population was reduced by 75% after 7, 10, and 14 days compared with cultures without CsA (Fig. 7, left half of panel). In EC-stimulated cultures, alloactivated CD8 T cells in the presence of CsA could only be detected on day 14, and the absolute number of expanded (activated) cells was reduced by

6 The Journal of Immunology 5151 FIGURE 5. Effect of costimulatory blockade on CTL differentiation by EC stimulation. Pooled cells from 14-day cocultures of CD8 T cells stimulated with EC in the absence and presence of blocking Abs against LFA-3, ICAM-1, and CD40 were analyzed for total cell number, perforin positivity, and expression of surface activation markers (horizontal bar graph). At the same time, the proportion of coculture microwells with detectable significant cytotoxicity from indicated cocultures was determined in cytotoxicity assays (inset). One representative experiment of three similar experiments is shown. CTX, Cytotoxicity. 80% compared with control cultures without CsA (Fig. 7, right side of panel). The influence of CsA on cytotoxic activity of CTL from cocultures was examined, using T cells recovered at day 14. CsA suppressed cytotoxicity displayed by cocultures after stimulation with either BLC or EC (Fig. 8). Addition of CsA at the initiation of BLC-stimulated cocultures resulted in a reduction in microwells testing positive for cytotoxicity from 100% to 20%, with a corresponding decrease in mean percentage of specific killing. Similarly, in EC-stimulated cocultures, detectable cytotoxicity was completely lost after early addition of CsA. Treatment of cocultures for 4 h before the cytotoxicity assay still resulted in a moderate decrease in cytotoxicity (Fig. 8) without affecting cell numbers (not shown) in both types of cocultures, but the magnitude of this effect was marginal compared with that produced by persistent CsA exposure. However, persistent presence of CsA in cocultures did not influence the expression of surface activation markers and perforin expression after stimulation with either BLC or EC on those CD8 T cells that did proliferate (Fig. 9). These observations suggest that the primary action of CsA is to reduce clonal expansion, despite the presence of IL-2, to a level that falls below the sensitivity of the cytotoxicity assays. Those cells that do expand appear resistant to the effects of CsA, and a smaller number of CsA-resistant clones emerge in the EC-stimulated cultures than in the BLC-stimulated cultures, which is consistent with the lower overall frequency of EC-responsive CTL. Discussion Cell type-selective CTL have been described after various forms of solid organ transplantation, and they have been implicated in tissue-specific injury during allograft rejection (3, 17, 18) or graftversus-host disease (19). We have previously described the in vitro generation of EC-selective CTL (4), from which CTL clones could be derived, that exhibited atypical features including absence of IFN- production and CD40L (CD154) expression (5). In the present study, EC-selective CTL were analyzed and compared with conventional CTL directly from primary cocultures with stimulating cells, thus eliminating potential selection bias and long-term culture changes that may occur during cloning procedures. Two principal findings of our study are that CD8 T cells expanded less well after stimulation with EC than with BLC and that stimulation of CD8 T cells by EC exclusively activated memory (CD45RO ) cells to differentiate into EC-selective CTL, whereas BLC stimulators were able to recruit both naive and memory CD8 T cells to differentiate into CTL. Specifically, there was a 5- to 20-fold reduction in the final absolute number of expanded (i.e., alloactivated) CD8 T cells after EC stimulation compared with that after BLC stimulation. Therefore, on the basis of these data, the reduced cytotoxicity seen with EC-selective CTL appears to be largely due to much lower E:T ratios in these cytotoxicity assays. However, EC-stimulated CTL also expressed significantly lower amounts of perforin in individual cells (Fig. 2, Exp. B), which could also contribute to the difference in cytotoxicity. The smaller number of expanded, alloactivated CD8 T cells in ECstimulated cocultures can in principle be explained either by a lower frequency of precursor cells or by fewer generations of cell division during cocultures. To account for the observed difference in CTL at the end of cocultures, precursor frequencies for BLCstimulated CTL would have to be 5 20 times higher than for EC-responsive CD8 T cells. Alternatively, the difference in absolute numbers of CTL at the end of cocultures could be caused by two to four fewer rounds of cell division in EC-stimulated cultures due to generally slower kinetics of CD8 T cell divisions or to a delayed start of T cell expansion in the presence of EC stimulation. As technical limitations preclude the exact determination of generational cycles in our cocultures, neither hypothesis could be tested experimentally. Considering that EC stimulation is only capable of activating memory CD8 T cells, the precursor frequency in EC-stimulated cultures at most can represent half of that in BLC-stimulated cultures. Moreover, previous studies determined that the precursor frequency in CD8 T cells comparing EC to monocyte stimulators was 5 times lower when measured by limiting dilution analysis of IL-2 production (7). Moreover, if ECselective CTL arise from precursors that only recognize EC-derived peptides, this precursor pool of pre-ctl may be even smaller than that determined by limiting dilution analysis. On the other hand, proliferation after stimulation with EC does appear to be

7 5152 EARLY EVENTS IN ENDOTHELIAL STIMULATION OF CTL FIGURE 6. Exogenous enhancement of costimulation does not improve CTL differentiation by stimulation with EC. Pooled cells from 14-day cocultures of CD8 T cells stimulated with EC in the absence and presence of enhanced costimulation were analyzed for total cell number, perforin positivity, and expression of surface activation markers (horizontal bar graph). At the same time, the proportion of coculture microwells with detectable significant cytotoxicity from indicated cocultures was determined in cytotoxicity assays (inset). A, Cocultures were performed in the absence and presence of comitogenic anti-cd28 mab. B, Cocultures were performed in the absence and presence of 24-h pretreatment with recombinant trimeric CD40L. One representative experiment of three similar experiments is shown each. CTX, Cytotoxicity. delayed compared with BLC stimulation (7), and EC reduce T cell proliferation stimulated by BLC in mixed cocultures (4). Thus, both factors may contribute to the lesser capacity of EC to stimulate alloreactive T cell populations. The inability of EC to activate naive T cells has been proposed to be due to limitations in Agindependent adhesion compared with other, professional APC (9) or to reduced capacity of cytokine secretion, especially IL-12 (9, 20). Previous observations that supplementation of EC-stimulated cocultures with IL-12 did not improve the generation of CTL (4) render the latter alternative less likely. CTL from EC-stimulated cultures continued to express surface markers CD69, CD25, and L-selectin (CD62L) at the end of the coculture period, representing an early, immature activation phenotype. CD69 and CD25 are generally considered early activation markers (21 23), and L-selectin is usually lost rapidly during the course of T cell activation (24 26). Also, the expression level of perforin per cell in EC-selective CTL was markedly lower than in corresponding conventional CTL. In contrast, on BLC-stimulated CTL, the expression of CD69 and CD25 had subsided, and CD62L expression, highly expressed in CD8 T cells at the beginning of coculture, had largely been lost. The findings in BLC-stimulated CD8 T cells are in accordance with numerous in vitro and in vivo reports showing the loss of early activation markers 2 3 days after activation of T cells by BLC or other professional APC (21 23). A direct correlation between the number of cell divisions and loss of CD25 and CD69 also has recently been reported (27). Time course of the activation phenotype on T cells after (allo)activation by EC has not been reported previously, and the basis for the retention of this early activation phenotype of EC-specific CTL is currently unknown. Several (not mutually exclusive) explanations appear possible. First, the phenotype seen on EC-stimulated CTL might represent an immature state of activation caused by slower or delayed activation of CD8 T cells by EC, such that the EC-selective CTL lag behind their conventional counterparts in the down-regulation of early activation markers and in the synthesis of perforin. Such defective activation in EC-stimulated cocultures may be caused by differences in EC-mediated adhesion, costimulation, or cytokine production. However, the persistence of the phenotypic differences between EC- and BLC-stimulated CTL beyond 14 days of coculture seems to argue against a simple temporal lag in the EC cocultures. It also appears unlikely that the differential phenotype FIGURE 7. Effect of CsA on expansion of CD8 T cells after stimulation with BLC or EC. Cells were harvested from cocultures containing CFSE-labeled CD8 T cells in the absence and presence of 500 ng/ml CsA at indicated time points, counted, and analyzed by two-color flow cytometry. The relative proportion of alloactivated (expanded) cells in cocultures (percentage of CF- SE low cells of total CD8 T cells) was combined with the total number of cells harvested per microwell (open bars) to yield absolute numbers for alloactivated, expanded CD8 T cells (filled bars), both of which are expressed as a percentage of the original input cell number per well. CsA, Coculture in presence of cyclosporine A. One representative experiment of three similar experiments is shown each.

8 The Journal of Immunology FIGURE 8. Effect of CsA on cytotoxicity displayed by CTL cocultures stimulated with BLC or EC. After 14 days of coculture of CD8 T cells with indicated stimulator cells, cells were harvested from individual microwells and tested for cytotoxicity against respective target cells. CsA was added to cocultures at indicated time points. Bars represent the proportion of microwells displaying significant cytotoxicity., Numbers indicate mean percentage of specific killing of all tested microwells. One representative experiment of three similar experiments is shown. of CTL after stimulation with EC or BLC should be exclusively due to fewer cell divisions after EC stimulation, especially because similar differences in the expression of early activation markers could already be detected after 10 days of coculture with EC. If the differences in expansion and phenotype between the cocultures were entirely due to the number of cell divisions, the number of cell divisions at day 14 after stimulation with EC would be expected to be more equivalent to the cell divisions at day 10 after stimulation with BLC, and cell phenotypes would likely correspond. Also, the up-regulation of CD62L was only seen in ECstimulated cultures and represents a unique feature of the EC-selective CTL phenotype. EC-derived effects could also account for the increased expression of CD69, even on nondividing (CFSE high ) CD8 T cells, in EC-stimulated cocultures compared with IL-2- treated control cultures (not shown). EC have been shown to secrete IL-7 and IL-15 and could thus up-regulate the expression of CD69 and CD25 (28 31). Finally, continuing activation of CD8 T cells by alloantigen in cocultures with EC but not with BLC 5153 could account for the described phenotype. Persistent expression of early activation markers has been demonstrated on T cells in disease states associated with continuous long-term immune activation, especially HIV and CMV infection or rheumatoid arthritis (32 34). In the current study, persistent T cell activation might theoretically result from a greater persistence of EC than BLC in the cocultures such that EC would continue to activate CD8 T cells. EC stimulators (in contrast to BLC stimulators) were not treated with mitomycin for growth arrest, but pilot experiments failed to show any difference in CTL generation irrespective of mitomycin treatment (not shown). In addition, no viable EC were detectable by microscopy in coculture wells after 5 days. Previous studies from our group and others have emphasized differences in the costimulators used by human EC and professional APC. In IL-2-supplemented EC-stimulated cultures, clonal expansion, surface phenotype, perforin positivity, and cytotoxicity toward EC were unchanged in the presence of blocking Abs to the three most relevant accessory receptor/ligand pairs involved in the interaction of EC and T cells: LFA-1/ICAM-1, CD2/CD58, and CD40/CD154. HUVEC do not express the ligands for either CD28 (B7.1 and B7.2; Ref. 10) or CD137, the human equivalent for mouse 4-1BB (ILA; T. J. Dengler, unpublished observation), and the T cell costimulatory potential of VCAM-1 and E-selectin are thought to be of minor importance (20, 35). However, exogenous costimulatory enhancement with a comitogenic anti-cd28 mab was also without effect on CTL generation. The central role of the CD40/CD154 interaction for the priming of professional APC as stimulators of CTL has been demonstrated extensively (14, 15, 36), but we did not observe a similar effect on EC in our experiments using recombinant CD40 ligand trimer. These results of our costimulation studies are in accordance with the classic concept that T cell costimulation by accessory cells primarily induces increased IL-2 production (37), an effect that was bypassed in our experiments by exogenous IL-2. Finally, as EC-selective CTL might play an important part in the pathogenesis of forms of rejection (e.g., endothelialitis) that are less amenable to treatment, we investigated the sensitivity of ECselective CTL to CsA. Our results demonstrate a similar inhibition of alloactivated CD8 T cell expansion by CsA in EC- or BLCstimulated cocultures. In BLC-stimulated cultures, the proportion of CD8 T cells expanding despite the presence of CsA increased FIGURE 9. Effect of CsA on the expression of surface markers and perforin of alloactivated CD8 T cells after stimulation with BLC or EC. Cells were analyzed by two-color flow cytometry after 14 days of microwell coculture of CD8 T cells with respective stimulator cells in the absence or presence of 500 ng/ml CsA. Expression of surface markers and perforin in alloactivated (expanded; CFSE low ; cf Fig. 1) cells stimulated with BLC or EC; for comparison, expression on unstimulated cells (total population) is shown. A, EC-stimulated coculture; B, BLC-stimulated coculture. One representative experiment of three similar experiments is shown in each panel.

9 5154 EARLY EVENTS IN ENDOTHELIAL STIMULATION OF CTL over time and was markedly higher than in EC-stimulated cocultures, which is again compatible with a delayed activation process in the presence of EC stimulators. Interestingly, the suppression of proliferation occurs in the presence of significant amounts of exogenous IL-2. Similar suppression of T cell growth by CsA has been reported previously (38, 39) and may be explained by more recent data indicating that CsA can inhibit T cell activation by additional mechanisms such as TGF- production, up-regulation of the cdk inhibitor p21, and inhibition of IL-2 receptor expression (40 42). Unspecific toxic or apoptotic cell death induced by CsA can be excluded as a cause of the reduced cell numbers in our experiments on the basis of the low concentration of CsA (500 ng/ml) used and the absence of apoptotic cells. CsA also reduced cytotoxicity, primarily via the reduction in CTL numbers, and consequently reduced E:T ratios. Differentiation or maturation of CTL at the level of the individual effector cell was not affected by CsA, as indicated by normal expression of activation markers and perforin. Late treatment of cocultures with CsA seemed to have a marginal inhibitory effect on cytotoxicity, which is in line with previous reports suggesting inhibition of perforin degranulation by CsA (43, 44). Taken together, EC stimulation of CD8 T cells for CTL generation does not appear to impart any CsA resistance. The effects of EC upon CTL activation and differentiation described here and in our previous studies have significant implications for transplantation and vascular biology. As shown earlier (4), the generation of highly cytotoxic, conventional CTL is suppressed by EC, a finding with particular pertinence for microenvironments in which EC are closely apposed to infiltrating immune cells, such as the vascular intima. On the other hand, stimulation of CD8 T cells by EC appears to favor the differentiation to a unique population of CTL displaying a variety of characteristic features, including an immature early activation phenotype, low perforin content, reduced capacity for clonal expansion, and cell type selectivity. These results strengthen a function of vascular EC as semiprofessional APC of intermediate stimulatory capacity as proposed earlier (7, 9) and help to define an evolving concept of the vascular endothelium as an immunomodulatory cell type. In the intimal space, where EC are likely to represent the predominant APC, that may create a distinct immunologic compartment endowed with specific regulatory characteristics, similar to other organ-specific immune functions. For example, the immaturity and comparatively ineffective clonal expansion of EC-stimulated CTL and the suppression of conventional CTL suggest that the predominant modulatory role of the endothelium might be inhibitory for immune-mediated injury. Based on our in vitro findings and recent clinical data (3), such EC-mediated immunoregulation can be overcome and can result in activation and differentiation of CTL selectively directed against EC. By analogy to other organ-specific CTL (17, 18), the generation of EC-selective CTL could be an independent, tissue-specific alloimmune process occurring in the absence of generalized alloactivation. Vascular rejection and intimal arteritis could be interpreted as forms of organ -specific endothelial rejection mediated by such EC-selective CTL, developing independently of wide-spread parenchymal rejection. Further characterization of EC-selective CTL and their activation requirements could lead to a better understanding of the pathogenesis and to novel therapeutic approaches for immune-mediated vascular pathology. Acknowledgments We thank Drs. Barbara Biedermann and David R. Johnson for helpful experimental suggestions and constructive discussion. We are also grateful to Louise Benson, Gwendolyn Davis, and Lisa Gras for excellent technical assistance with cell culture. References 1. Colvin, R. B., A. H. Cohen, C. Saiontz, S. Bonsib, M. Buick, B. Burke, S. Carter, T. Cavallo, M. Haas, A. Lindblad, et al Evaluation of pathologic criteria for acute renal allograft rejection: reproducibility, sensitivity, and clinical correlation. J. Am. Soc. Nephrol. 8: Dong, C., D. Redenbach, S. Wood, B. Battistini, J. E. Wilson, and B. M. McManus The pathogenesis of cardiac allograft vasculopathy. Curr. Opin. Cardiol. 11: Jutte, N. H., C. J. Knoop, P. Heijse, A. H. Balk, B. Mochtar, F. H. Claas, and W. Weimar Human heart endothelial-cell-restricted allorecognition. Transplantation 62: Biedermann, B. C., and J. S. Pober Human endothelial cells induce and regulate cytolytic T cell differentiation. J. Immunol. 161: Biedermann, B. C., and J. S. Pober Human vascular endothelial cells favor clonal expansion of unusual alloreactive CTL. J. Immunol. 162: Weston, S. A., and C. R. Parish New fluorescent dyes for lymphocyte migration studies: analysis by flow cytometry and fluorescence microscopy. J. Immunol. Methods 133: Epperson, D. E., and J. S. Pober Antigen-presenting function of human endothelial cells: direct activation of resting CD8 T cells. J. Immunol. 153: Pober, J. S., M. P. Bevilacqua, D. L. Mendrick, L. A. Lapierre, W. Fiers, and M. A. Gimbrone, Jr Two distinct monokines, interleukin 1 and tumor necrosis factor, each independently induce biosynthesis and transient expression of the same antigen on the surface of cultured human vascular endothelial cells. J. Immunol. 136: Ma, W., and J. S. Pober Human endothelial cells effectively costimulate cytokine production by, but not differentiation of, naive CD4 T cells. J. Immunol. 161: Karmann, K., J. S. Pober, and C. C. Hughes Endothelial cell-induced resistance to cyclosporin A in human peripheral blood T cells requires contactdependent interactions involving CD2 but not CD28. J. Immunol. 153: Lichtenfels, R., W. E. Biddison, H. Schulz, A. B. Vogt, and R. Martin CARE-LASS (calcein-release-assay), an improved fluorescence-based test system to measure cytotoxic T lymphocyte activity. J. Immunol. Methods 172: Song, H. K., H. Noorchashm, Y. K. Lieu, S. Rostami, S. A. Greeley, C. F. Barker, and A. Naji Tracking alloreactive cell division in vivo. Transplantation 68: Hughes, C. C., and J. S. Pober Transcriptional regulation of the interleukin-2 gene in normal human peripheral blood T cells. Convergence of costimulatory signals and differences from transformed T cells. J. Biol. Chem. 271: Schoenberger, S. P., R. E. Toes, E. I. van der Voort, R. Offringa, and C. J. Melief T-cell help for cytotoxic T lymphocytes is mediated by CD40-CD40L interactions. Nature 393: Ridge, J. P., F. Di Rosa, and P. Matzinger A conditioned dendritic cell can be a temporal bridge between a CD4 T-helper and a T-killer cell. Nature 393: Karmann, K., C. C. Hughes, J. Schechner, W. C. Fanslow, and J. S. Pober CD40 on human endothelial cells: inducibility by cytokines and functional regulation of adhesion molecule expression. Proc. Natl. Acad. Sci. USA 92: Hadley, G. A., E. A. Rostapshova, and S. T. Bartlett Dominance of tissuerestricted cytotoxic T lymphocytes in the response to allogeneic renal epithelial cell lines. Transplantation 62: Heath, W. R., and L. A. Sherman Cell-type-specific recognition of allogeneic cells by alloreactive cytotoxic T cells: a consequence of peptide-dependent allorecognition. Eur. J. Immunol. 21: Warren, E. H., P. D. Greenberg, and S. R. Riddell Cytotoxic T-lymphocyte-defined human minor histocompatibility antigens with a restricted tissue distribution. Blood 91: Briscoe, D. M., L. E. Henault, C. Geehan, S. I. Alexander, and A. H. Lichtman Human endothelial cell costimulation of T cell IFN- production. J. Immunol. 159: Santis, A. G., M. Lopez-Cabrera, F. Sanchez-Madrid, and N. Proudfoot Expression of the early lymphocyte activation antigen CD69, a C-type lectin, is regulated by mrna degradation associated with AU-rich sequence motifs. Eur. J. Immunol. 25: Marzio, R., J. Mauel, and S. Betz-Corradin CD69 and regulation of the immune function. Immunopharmacol. Immunotoxicol. 21: Ziegler, S. F., F. Ramsdell, and M. R. Alderson The activation antigen CD69. Stem Cells (Dayton) 12: Picker, L. J., J. R. Treer, B. Ferguson-Darnell, P. A. Collins, D. Buck, and L. W. Terstappen Control of lymphocyte recirculation in man. I. Differential regulation of the peripheral lymph node homing receptor L-selection on T cells during the virgin to memory cell transition. J. Immunol. 150: Giorgi, J. V., M. A. Hausner, and L. E. Hultin Detailed immunophenotype of CD8 memory cytotoxic T-lymphocytes (CTL) against HIV-1 with respect to expression of CD45RA/RO, CD62L and CD28 antigens. Immunol. Lett. 66: Razvi, E. S., R. M. Welsh, and H. I. McFarland In vivo state of antiviral CTL precursors: characterization of a cycling cell population containing CTL precursors in immune mice. J. Immunol. 154: Oehen, S., and K. Brduscha-Riem Differentiation of naive CTL to effector and memory CTL: correlation of effector function with phenotype and cell division. J. Immunol. 161: Kanegane, H., and G. Tosato Activation of naive and memory T cells by interleukin-15. Blood 88: Oppenheimer-Marks, N., R. I. Brezinschek, M. Mohamadzadeh, R. Vita, and P. E. Lipsky Interleukin 15 is produced by endothelial cells and increases