Can Graft Endothelial Cells Initiate a Host Anti-Graft Immune Response?1

In this Overview, we discuss whether graft endothelial cells (ECs*) can initiate anti-graft immune responses. In other words, are human ECs sufficiently competent antigen-presenting cells (APCs) to activate resting T cells? In 1975, Thorsby and colleagues reported that cultured human umbilical vein ECs could stimulate allogeneic T cells to proliferate in vitro (i.e., initiate a positive mixed leukocyte-endothelial reaction), suggesting that these cells possess the capability to initiate allogeneic T cell responses in vivo (1). Several years later, Stastny and colleagues re-examined this question and attributed the observed T cell response to the presence of a contaminating MHC class II-bearing subpopulation in the EC cultures, presumably of leukocyte origin (2). This debate continues to some extent in 1996. The relevant properties of ECs as APCs can now be analyzed in a much more precise manner, and at least some of the apparent discrepancies among laboratories can be adequately explained. We begin by reviewing these new insights into APC function in transplantation. Two key concepts have emerged from recent studies of the molecular and cellular basis of the host anti-graft immune response, responsible for the rejection of vascularized organ grafts. The first of these concepts is that host T cell recognition of graft antigens consists of two components, allorestricted (sometimes called “direct”) recognition and self-restricted (sometimes called “indirect”) recognition(3). Allorestricted recognition of alloantigen is a molecular cross-reaction: host T cells that bear antigen receptors selected to recognize a complex formed by a particular peptide (derived from a non-self protein) bound to a self-allelic form of a class I or class II MHC molecule also recognize different complexes formed by other peptides (derived from either self or non-self proteins) associated with particular non-self allelic forms of MHC molecules that are expressed by the cells of a foreign (i.e., genetically disparate) graft. The unexpected ability of host T cells to react with self peptides associated with non-self MHC molecules arises for two reasons: (1) the mechanisms to remove self-reactive T cells during thymic selection or in the periphery actually only eliminate T cells that react with self peptides complexed to self-MHC molecules; and (2) allogeneic MHC molecules may bind different peptides derived from the same self protein than do self-MHC molecules. In this case, tolerance never developed because T cells were never presented with these peptides until foreign MHC molecules were introduced into the system. Many different foreign peptide-specific T cells can react with the same foreign MHC molecule, although much of this cross-reactivity may depend upon structurally distinct peptides that are bound. In other words, different allorestricted T cells reactive with a particular non-self MHC molecule may actually recognize different peptide-MHC complexes. The consequence of multiple cross-reactivity is that an unimmunized host may have as many as 1 in 500 T cells capable of reacting with a particular graft cell bearing foreign MHC molecules. In contrast, an unimmunized host may only possess 1 in 100,000 to 1 in 1,000,000 T cells specific for a particular foreign protein. Therefore, allorestricted recognition of alloantigens results in a very strong primary reaction and can be measured in an unimmunized host. Self-restricted presentation results when host APCs present graft proteins in the form of peptides associated with self-MHC molecules. In the unimmunized host, the number of alloreactive self-restricted cells will be small, comparable to the number that react to other foreign proteins. The primary response is likely to be so small that it cannot be measured in vitro. However, like an immune response to an infectious agent, alloreactive self-restricted T cells will expand with time, eventually leading to a reaction capable of initiating graft rejection and being measured in vitro. There are several other important differences between allorestricted and self-restricted recognition of alloantigen. For example, self-restricted recognition largely involves treating graft cells as sources of extracellular(to the APC) protein, and will overwhelmingly involve presentation of peptides complexed with class II MHC molecules to CD4+ T cells. In contrast, allorestricted recognition of alloantigen will also involve CD8+ T cell activation by foreign class I molecules. Cytolytic T lymphocyte (CTL) responses, which are largely mediated by CD8+ T cells, are therefore more prominent in allorestricted responses. Moreover, CTL activated by an allorestricted pathway are capable of directly lysing graft cells. In contrast, CTL activated by self-restricted pathways will actually attack host-derived cells. Another implication is that many of the alloreactive T cells responding to allorestricted presentation may be memory T cells, even in a previously unimmunized host, because these cross-reactive T cells may have been activated previously by prior contact with their specific foreign peptide, e.g., during an infection before transplantation. In contrast, self-restricted recognition will initially involve a predominant or exclusive response of naive T cells. Finally, and most relevant to this discussion, allorestricted recognition of alloantigens involves the provision of competent APC by the graft, whereas self-restricted recognition occurs when these cells are provided by the host. Consideration of the role of competent APCs brings us to the second key concept to emerge from recent studies of allogeneic immunology, namely the recent molecular refinement of the “two-signal hypothesis” of T cell activation (4). In brief, it is now understood that signals provided from the binding of peptide-MHC complexes by specific T cell antigen receptors are too weak to initiate interleukin (IL)-2 synthesis by T cells. The critical role of competent APCs in the initiation of immune responses is to provide cell surface ligands that interact with T cell accessory molecules and amplify the weak specific signal provided through the antigen receptor into a sufficiently strong signal to cause IL-2 synthesis and secretion. Several such “costimulatory” signals have been identified, including: MHC molecules on the APC, which signal through T cell CD4 (for class II) or CD8 (for class I); B7.1 and B7.2 on the APC, which signal through T cell CD28; and leukocyte function-associated antigen (LFA)-3(CD58) and CD59 on the APC, which signal through T cell CD2. Many stromal and epithelial cells lack sufficient costimulators and are therefore unable to effectively activate resting T cells, despite displaying appropriate peptide-MHC molecule complexes. Moreover, the costimulator requirements needed for activation of T cells change with T cell maturation and differentiation. Of relevance to transplantation is the observation that the costimulator requirements of naive T cells are greater than those of memory T cells(5), and only highly specialized APCs, such as dendritic cells, effectively activate resting naive T cells. Memory T cells can receive adequate costimulation from less specialized APC, including macrophages and, as we shall argue here, vascular ECs. We should note that recently or repetitively activated T cells, such as T cell blasts or clones, respectively, have even less stringent costimulatory requirements than resting memory cells and can often be stimulated by tissue cells bearing appropriate peptide-MHC molecule complexes. Finally, there seem to be important species differences in the tissue distribution of costimulator molecules. Mice, for example, do not display LFA-3 or other CD2 ligands on ECs whereas human do(6). Similarly, pigs apparently display B7.2 on ECs whereas humans do not (7). These differences in accessory molecule display may result in differences in the identity of APCs that initiate allogeneic responses across species, and it is dangerous to extrapolate results from nonprimates to human systems. MOLECULAR CHARACTERISTICS OF CULTURED ENDOTHELIAL CELLS AS APCs Endothelial cells have been cultured from a variety of different vascular beds and from a number of different mammalian species. However, most work has been accomplished with human umbilical vein ECs, the first type of endothelium to be cultured successfully. In general, similar results have been noted with other cultured EC types and we will only emphasize other ECs when differences from umbilical vein ECs have been reported. Under standard culture conditions, optimized for cell growth and viability, human ECs express class I MHC molecules (HLA-A, B, and C) but not class II molecules (HLA-DR, DP, or DQ)(8). There are no reports of expression of nonconventional class I-like or class II-like molecules on ECs. Treatment of cultured ECs with interferon (IFN)-α, IFN-β, tumor necrosis factor (TNF), lymphotoxin(LT), or CD40 ligand increases the level of expression of class I molecules without inducing class II molecules (9, 10, and unpublished observations [J.S.P.]). (An exception is that TNF has been reported to induce class II MHC molecule expression in porcine aortic endothelial cells [11].) IFN-γ also increases the level of class I MHC molecule expression and, on human ECs, is uniquely able to induce expression of class II MHC molecules (8). HLA-DR is the most strongly expressed class II locus, followed by HLA-DP; very little HLA-DQ is detectable on human ECs. The IFNs are more potent than TNF or LT as signals to up-regulate class I MHC molecule expression(9, 10). When used in combinations, IFNs (α,β, or γ) seem to act interchangeably with each other and multiplicatively (often described as “synergistically”) with TNF or LT to increase class I (10). IFN-α and IFN-β strongly inhibit the ability of IFN-γ to induce class II molecule expression; TNF or LT also inhibit, but less strongly than IFN-α or IFN-β (10). Endothelial MHC molecule expression is largely unaffected by other cytokines, including IL-1, IL-2, IL-3, IL-4, IL-6, IL-8, and granulocyte-macrophage colony-stimulating factor. Natural killer cells are able to increase EC expression of MHC molecules through a contact-dependent signaling pathway that is independent of IFN-γ (12). In all cases, effects on MHC molecule expression are gradual (onset in hours, peak effects in days to weeks) and expression of induced molecules persists at elevated levels for days to weeks after withdrawal of cytokines. In general, the MHC patterns of expression exhibited by cultured EC are recapitulated in vivo. Human ECs are uniformly positive for class I MHC molecules (13, 14). Class II MHC molecules are expressed constitutively on some ECs, including most postcapillary venules, veins, and some arteries (e.g., coronary arteries)(13, 14). Consequently, foreign class I and class II MHC molecules are introduced into a host every time a vascularized allograft is transplanted. Expression is increased at sites of inflammation or at sites of IFN-γ injection (15). In dogs, constitutive class II expression is inhibited by cyclosporine (CsA), presumably due to reduction in basal levels of local IFN-γ(16). However, IFN-γ “knockout” mice have normal levels of constitutive class I and class II MHC molecules compared with wild-type littermates, but do show reduced levels of inducible MHC molecule expression in response to systemic inflammation(17). Rats do not show basal endothelial class II MHC molecule expression in vivo, although induction has been observed in transplant rejection reactions (18). As we noted in the Introduction, the actual structure recognized by T cells is a complex of the MHC molecule and a bound peptide. Class I MHC molecules acquire their peptides predominantly from proteins degraded in the cytoplasm by the proteosome; cytoplasmic peptides are translocated into the endoplasmic reticulum and loaded on to nascent class I molecules by the action of transporter in antigen processing (TAP) proteins (19). Although peptide loading has not been examined directly in ECs, it has been shown that ECs express TAP proteins (20) and that allorestricted CD8+ CTL lines can specifically lyse EC, implying that the TAP system is functional (21, 22). Class II MHC molecules associate with a non-MHC-encoded protein, commonly called the invariant chain, until they reach an intracellular compartment that fuses with late endosomes/lysosomes (14). In this compartment, the invariant chain is degraded and peptides derived from endocytosed proteins are loaded. The loading of peptides into class II molecules depends on an MHC encoded molecule called DM (23, 24). Endothelial cells have been shown to express invariant chain after IFN-γ treatment(8), but there has been no report of the presence or absence of DM in this cell However, human EC are able to proteins to a II MHC molecule complex recognized by CD4+ T These observations that ECs be able to form specific peptide complexes with class II MHC molecules. human ECs express a variety of costimulator molecules. As noted MHC molecules can provide costimulatory signals through CD4 or In human ECs express costimulators that act through namely LFA-3 and CD59 In contrast, human ECs do not express costimulators that act through EC molecules may also to For example, in some T cells can be through endothelial such as molecule and vascular cell molecule Human ECs constitutively express and each of which can bind to T cell expression is constitutively but it can be and in response to IL-1, CD40 or IFN-γ IFN-γ effects multiplicatively with these other IL-1, CD40 and, to a can induce ECs to express which to T cell and IFN-γ are only weak of multiplicatively with IL-1, or LT to increase As reported by and by many and are on ECs of human also induce ECs to express and as as ligands for but it has not been shown that these molecules can act as costimulators for T cell However, reactive with can T cell and ligands may have similar In general, the effects of and in costimulation are much than those of LFA-3 and the molecules are not by cytokines, the of human ECs with on costimulator is ECs, which lack LFA-3 may display a of function due to of or Although are not to be important sources of T cell it is that or either or in with other cytokines, can cause culture human ECs to a number of that can act on T cells, including IL-1, IL-6, and in Human ECs have not been to IL-2, IL-3, or is not whether ECs can be induced to express other on T cells such as or CHARACTERISTICS OF CULTURED ECs AS APCs As noted in the T cells at different of and activation have requirements for stimulated T cell proliferate in response to cultured human ECs peptides associated with self MHC in some cases, allogeneic MHC molecules CD4+ and CD8+ CTL can ECs in a upon MHC molecule recognition (21, However, these cannot be attributed to specialized properties of ECs, many stromal and epithelial cell types also can present antigen in with MHC molecules to T cell (for example, some cultured EC from as as T EC have been noted to be at CD4+ T cell that and but not IL-2, and the same EC are less able to stimulate other CD4+ T cell that express the cytokines. In other words, immune reactions or by graft EC involve CD4+ T cell it reported recently that of the ECs with can with the to activate as The of this and the extent to which it can be to other EC are These properties of ECs are of to transplant because are to the of T cell that whereas are to the of T cell that are more of but types of T cell responses may to graft human EC have been to from most other stromal or epithelial cell types in that they can activate of allogeneic T cells from The time of this T cell response to allogeneic ECs is comparable to in response to allogeneic cells mixed leukocyte reaction), although the of the mixed leukocyte-endothelial reaction is IL-2 can be by or in the of such T when IL-2 is by of to the of the IL-2 receptor The ECs in such are induced to express class II MHC molecules the first presumably in response to which also may be in the should be noted that IL-2 and IFN-γ are derived from the T cells, not the ECs. has been shown that when T cell are and cultured with allogeneic ECs, the induction of class II molecules on the EC is to CD8+ T cells CD8+ and CD4+ T cells but CD4+ T cells also an as that class II induction CD8+ T cells with allogeneic EC some IL-2 and but CD4+ T cells more IL-2 and proliferate even However, the CD4+ response depends upon of the ECs with IFN-γ with killer or CD8+ T to cause class II MHC molecule whereas constitutive levels of class I expression are adequate to initiate CD8+ T cell responses. When CD8+ and CD4+ T cell than when either is analyzed CD8+ but not CD4+ T cell can be inhibited by that bind to class I MHC molecules or to CD8 in contrast, CD4+ but not CD8+ T cell can be inhibited by that bind to class II MHC molecules or to of is inhibited by that IL-2 can be used to the of CD8+ and CD4+ T cells in that to allogeneic human EC The among and among laboratories the but the is that or umbilical 1 in to CD4+ T cells and 1 in to CD8+ T cells that to allogeneic The CD4+ response can be only when the ECs have been with whereas the CD8+ response is not by this should be noted that these are than the of T cells that can be stimulated by allogeneic cells, a of CD8+ T and CD4+ T cells, ECs only stimulate alloreactive memory cells, whereas cells stimulate memory and naive T cells. However, even among memory cells, ECs are not as potent as cells, a of CD4+ and CD8+ T cells. that the ECs activate some CD4+ T cells by presentation of class II as APCs, and different CD4+ T cells by allorestricted presentation of class II to allogeneic EC can be inhibited by that costimulatory between T cells and of CD2 with LFA-3 or to a molecule in are also by to and are without and effects of reactive with or are that T cell with ligands and are without in allogeneic responses to these same are very in responses of T cells to allogeneic cells The costimulator of EC can be analyzed more directly by cell stimulated by such as or to interact with In this the presence of ECs T cells to increase the of and of IL-2 and cells have similar but cells and stromal cells do Although many were ECs that were allogeneic to the T cells, recent work ECs to the T cells have these that costimulation is independent of allogeneic should be noted that as there are no to which costimulator molecules expressed by ECs or cells are on or vascular cells. some molecules that are expressed on and cells have been by to be important in costimulation provided by EC or cells. the same that costimulation in this more effectively costimulatory signals in allogeneic In CD2 and ligands seem to be of critical and no role has been observed for and T cell contact with EC is to and EC may for cells it has been that EC can even provide costimulation to T cells that are responding to allogeneic cell types that are unable to cause a response, such as with IFN-γ to express class II MHC In other words, EC costimulators seem to be even when provided by cells different from those the primary antigen with EC costimulators can provide which EC presentation of antigen tolerance to of note is that EC costimulation to the ability of and but not to IL-2 This also cell contact between the EC and T cells, but IL-2 the EC will not to they have been to to is not until after has been but IL-2 synthesis is as as in the absence of cells lymphocyte cells can this is whether EC costimulation is responsible for the of T cells that have been from there are only a of the antigen-presenting of human ECs from distinct and there are no tissue with ECs from of to transplant are of antigen presentation by ECs of origin have aortic ECs are unable to express class II MHC molecules and do not allogeneic responses, but can act as competent accessory cells to responses ECs are APCs whereas ECs seem competent to initiate allogeneic responses, but not protein antigen responses, endothelial cells seem to activate CD8+ T cells, but are less at initiating CD4+ T cell responses ECs, when in vitro with seem capable of initiating in vivo rejection responses of a CD4+ T response In general, these results that species and tissue may in the of ECs to act as should be that tissue culture and of ECs involves and some of these differences may also from differences in the selection used to these EC OF ENDOTHELIAL AS APCs the that several independent including those of the of this Overview, have observed and by T cells with allogeneic human ECs, there some as to whether an cell to provide is The needed to this question of resting T cells as of ECs to of a cell However, as noted T cell have less stringent requirements for costimulation and cannot provide the activation of resting cells. it is to normal ECs from these cells display a depend on than of the T cell and EC In general, highly ECs can stimulate the activation of highly allogeneic T cell Therefore, costimulator cell in these highly cell be at very The only APC that is at is the dendritic Several observations that dendritic cells, which constitutively express levels of class I and class II MHC molecules as as not be present in the EC to activate highly allogeneic T cell These (1) that EC from are as despite their reduced for by other cell (2) that resting EC cultures, which are of or MHC class II-bearing cells, are able to stimulate T cells some EC including the and the T human EC and unpublished can stimulate allogeneic T cells to that EC can activate CD4+ T cells only they are with IFN-γ to cause class II whereas dendritic cells do not such that EC do not costimulator molecules such as B7.1 and which are used by dendritic cells and that the identity of the T cells that are activated by EC only a of those activated by dendritic that the T cell do not to dendritic cell in to to ECs observations (1) that the same T cell that do to EC are unable to to or cells derived from the same as the ECs and the same level of MHC and (2) that the not to the distribution a level of by cell type in the lymphocyte to cause is most to that the positive results in all laboratories from cell that are independent of ECs can directly stimulate of allogeneic T cells, have some laboratories been unable to between highly T cells and is that the activation of T cells by EC is than that provided by allogeneic the EC may be the system is The EC to be it is optimized and from the lack of response to or cells. relevance the T cell response to ECs have to or to transplant it is that local ECs can present foreign protein antigens to a of memory T cells, this could the by which responses are at For example, the response to a second to an infectious could be before memory T cells even the a not be to an immune response (i.e., T cells could a by inflammation that often but immune reactions could more host the in vivo for endothelial presentation of antigen is the of of inflammation is a of injection of specific protein, a time that is as as the response to injection of cytokines. the of ECs to activate alloreactive memory cells that graft ECs may initiate rejection This is the implication for transplantation to human of APCs (i.e., should not be to have on graft is not a that the for the leukocyte from transplant MHC molecule and LFA-3 expression by ECs are reduced or compared with Finally, ECs are important of graft ECs for in to the of an organ graft. This is ECs can be used to induce T cell