Antibody-Mediated Rejection: Emergence of Animal Models to Answer Clinical □Questions

Abstract

Decades of experiments in small animals had tipped the balance of opinion away from antibodies as a cause of transplant rejection, but clinical experience, especially with sensitized patients, has convinced basic immunologists of the need to develop models to investigate mechanisms underlying antibody-mediated rejection (AMR). This resurgent interest has resulted in several new rodent models to investigate antibody-mediated mechanisms of heart and renal allograft injury, but satisfactory models of chronic AMR remain more elusive. Nevertheless, these new studies have begun to reveal many insights into the molecular and pathological sequelae of antibody binding to the allograft endothelium. In addition, complement-independent and –dependent effects of antibodies on endothelial cells have been identified in vitro. As small animals models become better defined, it is anticipated that they will be more widely used to answer further questions concerning mechanisms of antibody-mediated tissue injury as well as to design therapeutic interventions.

Introduction

In the first decade of clinical transplantation the incidence of antibody-mediated hyperacute rejection was frequent. Williams reported that before 1969 about 50% of second renal transplants were rejected hyperacutely (1). Hyperacute rejection was virtually eliminated following publication of the first cross-match technique in 1969 (2). As tests to detect donor-reactive antibodies became more sensitive, low levels of antibodies were associated with increased frequency of reversible acute rejection. However, antibodies were generally thought to be an epiphenomenon because passive transfer of immune serum to rodent allograft recipients convinced basic scientists that antibodies were not sufficient and usually not necessary to mediate acute rejection of skin or organ transplants. Often antibody transfer enhanced rather than shortened graft survival in rats and mice (3, 4). In contrast, passively transferred T cells accelerated graft rejection and elimination of T cells prevented rejection. As a result, T cells became the focus of experimental transplantation and basic research on antibody-mediated rejection (AMR) was very limited. While research on donor-reactive T cells yielded immunosuppressive approaches that greatly decreased acute rejection, there was little effect on chronic rejection.

Clinical interest in AMR was reignited in 1990 when Halloran and colleagues (5, 6) described a small number of renal transplants with pathological features of “pure” AMR. With the identification of C4d as a pathological marker for AMR in clinical transplants (7, 8), the question evolved from whether to how much of acute and chronic rejection is caused by antibodies. This has prompted the need for better animal models to define molecular mechanisms of antibody-mediated graft injury.

Relevance of Clinical Experience to Development of Relevant Animal Models

Experimental models that were often too reductionist to be clinically relevant convinced many basic immunologists and clinicians that mice and rats were not appropriate animals for testing mechanisms underlying AMR. With the appropriate experimental design, however, mice offer unmatched advantages of genetic manipulability and extensive treatment options. Therefore, it is important to evaluate animal models in the context of clinical AMR.

The largest clinical experience with AMR has been in renal allografts. This is due to several factors, including the larger volume of transplants to both unsenstized and sensitized recipients, comparison between transplants from living and deceased donors, and qualitative aspects of renal biopsies. Studies on biopsies from sensitized patients with suspected rejection reported very high incidences of diffuse C4d deposits on peritubular capillaries that usually occurred in the first few months after transplantation. Based on this experience, stringent criteria were established for acute AMR in renal transplants (9, 10). These criteria are: 1) detection of circulating antibodies to donor MHC antigens; 2) diffuse deposition of the complement split product C4d in peritubular capillaries as an indicator of antibody activity; 3) morphologic indications of acute tissue injury; and, 4) evidence of graft dysfunction. Using these criteria, AMR was diagnosed in 1- 6% of protocol biopsies from unsensitized patients (11), and more frequently (reaching 50-70%) in biopsies from patients with suspected rejection (12). With increased acknowledgement of antibodies as a cause of graft injury, the perspective has changed to determining the full extent of antibody effector functions in transplants. The concept of subclinical AMR was introduced to test whether deposits of C4d and vascular inflammation in the absence of concurrent graft dysfunction progressed to subsequent acute or chronic rejection (13, 14). More recently, pathological, physiological or molecular evidence of endothelial disturbance in the absence of demonstrable C4d deposits has been correlated with chronic graft failure (15).

For a variety of reasons (simplicity of surgery and vigor of rejection), small animal models more frequently use heterotopic cardiac transplants than orthotopic renal transplants to investigate mechanisms of allograft injury. In this regard, it is pertinent that concepts about AMR have been more controversial for cardiac transplants than renal transplants (16).

Experimental Models of Acute AMR with Markers of Complement Activation

Reagents have now been developed to detect C4d and C3d in rats and mice. These reagents have demonstrated that rejection of cardiac transplants is frequently associated with diffuse, linear deposits of C4d and C3d on vascular endothelium and is accompanied by circulating alloantibodies as measured by clinically relevant flow cytometry assays (17-21). The deposition of C4d and C3d is related to alloantibody activation of complement because it is absent in cardiac allografts in immunoglobulin knockout mice (19). In some strain combinations, acute rejection of MHC incompatible allografts is also delayed in immunoglobulin knockout recipients. In these models, passive transfer of complement activating monoclonal antibodies to donor MHC antigens restores complement deposition and accelerates graft rejection (22). As in clinical episodes of AMR, macrophages, neutrophils and platelets are the predominant inflammatory constituents at sites of complement deposition. The convergence of complement and these leukocytes is logical because the complement split product C5a is an extremely potent chemoattractant and activator of neutrophils and macrophages. That C5a is an important chemoattractant in AMR is supported by experiments in which administration of monoclonal antibodies to C5 inhibits macrophage infiltration of cardiac and renal allografts in sensitized mice (23, 24). In a murine renal allograft model, a C5a receptor antagonist was found to block upregulation of chemokines and adhesion molecules on endothelial cells as well as macrophage infiltration (25). It should be emphasized that the strains of mice used in experiments to investigate complement need to be selected carefully because some strains of mice have complement deficiencies. For example, B10.D2 old, DBA/2, AKR, and A/J mice are C5 deficient (Table 1).

Table 1.

Examples of significant differences between mice and humans of relevance to AMR

	Mouse	Human
Complement	Complete or partial deficiencies in some strains1	Complement deficiencies are uncommon2
MHC expression on endothelium  Class I  Class II	Constitutive Not on resting endothelium, but upregulated by inflammation3	Constitutive Constitutive on most vascular endothelium4
Anatomy of coronaries	Short epicardial segment; Mostly intramyocardial	Mostly epicardial; Branches intramyocardial
Pre-existing injury	Usually none	Frequent5

¹For example, some strains have complete deficiencies of C5 (A/J, AKR, B10.D2 old and DBA/2), some have low levels of C3 (B10.D2 old and new, C57BL/10 and DBA/2) and others have low levels of C4 (CBA/J and C3H).

²Deficiencies of terminal complement components more common than early components.

³Upregulated on vascular endothelium in response to ischemia, infection and rejection .

⁴Expressed by endothelial cells in capillaries and veins in kidneys; capillaries, veins and arteries of hearts.

⁵For example, ischemic injury, pre-existing athromas, etc.

Induction of acute AMR in T and B cell immunodeficient mice (RAG^−/−) by passive transfer of antibodies to donor MHC antigens has been difficult to achieve. Nozaki et al. (20) concluded that antibody titer is critical to successful passive transfer of AMR to RAG-1^−/− recipients. These investigators found that wild-type recipients of MHC incompatible cardiac transplants do not produce sufficiently high titers of alloantibodies to cause AMR in passive transfer studies with RAG-1^−/− recipients. In contrast, CCR5^−/− mice produce titers of alloantibodies to MHC mismatched A/J cardiac allografts that are 15- to 25-fold higher than their wild-type counterparts. Transfer of undiluted sera containing these higher alloantibody titers from CCR5^−/− mice to RAG-1^−/− mice 2 days after transplantation with an A/J heart caused AMR within 12 days. As in clinical AMR, this process was accompanied by neutrophil and macrophage infiltration that was immunohistologically documented at the time of rejection. AMR did not result when the sera were diluted to wild-type titers. Understanding the importance of antibody titer may allow the design of passive transfer experiments that result in C4d or C3d deposition in the absence of acute rejection to provide useful models of subclinical or asymptomatic AMR.

Attempts to provoke AMR in RAG^−/− recipients with monoclonal antibodies to MHC antigens on the transplanted heart have generally produced little pathology even though they cause C4d deposition on capillaries (26, 27). The use of monoclonal antibodies has several potential pitfalls. First, each monoclonal antibody recognizes a single epitope and this results in lower densities of antibodies arrayed on the vasculature. Mixtures of different classes and subclasses of monoclonal antibodies more closely approximate polyclonal antisera, and such mixtures can activate complement synergistically (19). Second, experimental models using transferred monoclonal antibodies have focused on antibodies to MHC class I antigens. In clinical studies, however, antibodies with specificities for MHC class II alone or mixtures of antibodies to MHC class I and II have been associated with AMR (28). This relates to a critical antigenic difference between murine and human organs (Table 1). MHC class II antigens are expressed constitutively by endothelial cells in most vessels in humans (capillaries and veins in kidneys; capillaries, veins and arteries of hearts), but not in mice (29, 30). However, MHC class II antigens are rapidly upregulated on murine endothelial cells in response to local and systemic inflammation including ischemia (31, 32). Therefore, antibodies to MHC class II antigens may be effective in particular experimental models, such as early after transplantation or in mixtures with antibodies to MHC class I antigens or in combination with T cells, all of which have clinical relevance.

The complete absence of T cells in RAG^−/− recipients may undermine the pathogenic effects of antibodies. Obviously, upregulation of MHC class I and II antigens on endothelial cells by IFNγ and other T cell secreted cytokines increases targets for antibody binding and enhances the proinflammatory properties of antibodies. Platelets are another potential link between antibodies and T cells. Clinically, platelet aggregates are found in peritubular capillaries of renal transplants with acute AMR (33). Morrell et al (34) have demonstrated that antibodies to MHC antigens on endothelial cells of skin grafts cause the release of von Willebrand factor (vWf) and adhesion of platelets to blood vessels. Activated platelets express and secrete many factors that can localize and activate T cells in allografts (35, 36).

Effects of Antibody Interactions with Capillary Endothelium

Because endothelial cells are the primary target of both complement-dependent and independent effects of antibodies, Sis et al (15) have hypothesized that altered expression of genes associated with endothelial cells would be a sensitive indicator of AMR. They reported that increased expression of a panel of endothelial associated transcripts correlated with circulating donor specific antibodies, C4d deposition, and AMR. Importantly, only about half of the patients with elevated endothelial associated transcripts and capillary pathology (peritubular capillaritis, glomerulitis, transplant glomerulopathy, and capillary basement membrane multilayering) or graft loss had C4d deposits in their biopsies. These data suggest that C4d deposition is an insensitive marker for AMR. Several genes within this panel of transcripts had high individual correlations with AMR including vWf, vascular cell adhesion molecule-1 (VCAM-1) and platelet endothelial cell adhesion molecule (PECAM). The increased expression of these adhesion molecules is consistent with antibody-mediated endothelial cell activation. In vitro experiments have demonstrated that antibody cross-linking of HLA on endothelial cells stimulates exocytosis of vWf and expression of VCAM-1 (37, 38). Administration of alloantibodies causes release of vWf from capillaries and intravascular aggregation of platelets in murine skin and cardiac allografts (22, 34). The findings are not limited to antibodies to MHC. Rose and colleagues (39) have shown that administration of antibodies to the autoantigen vimentin causes expression of P-selectin, which is stored together with vWf in Weibel-Palade granules of endothelial cells and alpha granules of platelets.

One explanation for AMR in the absence of demonstrable C4d deposition is fluctuating amounts of antibody that allow clearance of C4d. C4b and its final split product, C4d, can bind covalently to free hydroxyl or amino groups on proteins or carbohydrates. This reaction permits a significant amount of the C4b to bind to antibody itself. Any C4d bound to antibody would be cleared with the same short half-life of the antibody. The C4d bound to endothelial cells in transplanted rat hearts is also cleared within about 5 days (17). The kinetics of antibody and complement turnover make it possible that subclinical or asymptomatic AMR can initiate inflammatory responses that become clinically evident after the C4d deposits are cleared. Colvin and co-workers have recently modeled this in mouse cardiac allografts (27).

Although complement activation can greatly augment chemotaxis of leukocytes (principally through C5a), activation of leukocytes and endothelial cells (through C3a and C5a receptors), as well as injury to endothelial cells (through the membrane attack complex), antibodies can initiate some of these processes through interactions with Fc receptors on leukocytes and by cross-linking HLA on endothelial cells. Extensive studies by Reed and co-workers have demonstrated that even in the absence of leukocytes and complement, binding of antibodies to MHC class I antigens on endothelial cells or smooth muscle cells stimulates release of growth factors, upregulation of receptors and cell proliferation (40). As noted previously, Lowenstein and his colleagues (38) have demonstrated that antibody-mediated cross-linking of MHC class I antigens on endothelial cells can also induce exocytosis of vWf and P-selectin from Weibel-Palade storage granules. Although most of these experiments were performed on endothelial cells in vitro without complement, purified components of the terminal membrane attack complex of complement can induce endothelial cell proliferation (41) and exocytosis of vWf and P-selectin (42). While non-complement activating antibodies stimulate the exocytosis of vWf and P-selectin followed by adhesion of platelets and neutrophils to capillaries in skin grafts, complement-activating antibodies have a more prolonged effect (22, 34). Similarly, antibodies alone can cause endothelial cells to produce cytokines such as IL-6 and MCP-1, but increased production of these cytokines occurs when antibodies activate macrophages through Fc receptors (43). These cytokine responses are likely to be further enhanced by complement in transplanted organs because macrophages express an array of receptors for complement split products.

The interaction of antibodies with complement and Fc receptors varies among different classes and subclasses of antibodies. In addition, Ravetch and colleagues have demonstrated that variations in the carbohydrate side chains within subclasses of antibodies modify the binding affinity of complement and Fc receptors (44). Variations in the carbohydrate structure of IgG antibodies have been noted in autoantibodies for many years. Autoantibodies that lack the terminal sialic acid and galactose (referred to as “G0” antibodies) expose terminal N-acetylglucosamine (GlcNAc) residues. Mannose binding lectin (MBL) binds avidly to GlcNAc. As a result, G0 autoantibodies activate complement through the lectin pathway in both humans and mice (44, 45). In mice, G0 antibodies also bind activating (FcγRIII and FcγRIV) Fc receptors more strongly and inhibitory (FcγRIIB) receptors more weakly (44). The dynamics of glycosylation of antibodies requires further investigation, but it has been reported to be influenced by pregnancy and treatment with infliximab and possibly mycophenolate mofetil (46, 47).

In renal biopsies, chronic manifestations of AMR are characterized by duplication of the basement membrane in glomerular (transplant glomerulopathy) and peritubular capillaries (48, 49). In cardiac transplants, C4d deposition on capillaries in endomyocardial biopsies has been correlated with increased chronic arteriopathy in the coronary arteries (50), but changes in capillary basement membrane structure in endomyocardial biopsies have not been identified as a diagnostic feature of chronic rejection in hearts. Therefore, renal transplant models may be more suitable for examining these changes. Orthotopic renal transplants have generally been less susceptible to acute rejection in mice. However, acute AMR of renal allografts does occur in CCR5^−/−, but not in μMT^−/−/CCR5^−/−, recipients with each of the four criteria for clinical diagnostics of AMR observed (21). The acute models of AMR have not yet been modified to produce chronic rejection in mice, but studies in non-human primates with mixed chimerism found that almost half of the animals with long-term renal allografts made donor specific alloantibodies, almost a third had diffuse C4d deposits in peritubular capillaries, and the majority of these biopsies had transplant glomerulopathy (51).

Manifestations of Antibody Interactions with Arterial Endothelium

In spite of the availability of protocol endomyocardial biopsies from most cardiac transplant recipients, the utility of these biopsies for studies of antibodies in chronic rejection is problematic because the pathological changes in this process are initially confined to the epicardial coronary arteries. This does not exclude the possibility that physiological or molecular changes might be found in the microvasculature of the endomyocardial biopsies that reflect processes in the coronary arteries and this is certainly open to investigation in animal models.

In order to make a causal link between AMR and chronic arteriopathy of cardiac transplants, mechanisms of antibody-mediated injury to arteries need to be established. This could be mediated through direct interactions of antibodies with arterial endothelium, but it could also be through effects on the vasa vasorum that supply the coronary arteries or through indirect effects on the resistance arterioles. Human coronary arteries are nourished by a network of vasa vasorum within the adventitia, which are known to increase in natural atherosclerosis. A few studies have reported changes in vasa vasorum in chronic graft arteriopathy, but these have not been related to AMR (52). It is possible that models of segmental aortic grafts are more prone to immune injury because their nourishing vasa vasorum are disrupted (53).

High affinity antibodies have the capacity to bind in the high flow, high shear stress conditions of arteries. In fact, early reports that linked antibodies to vascular pathology emphasized deposition of antibodies and complement on arteries (54, 55). In contrast, most recent studies have found that C4d deposition is frequently confined to capillaries (9). This may reflect differences in intensity of rejection due to advances in cross-matching and immunosuppression. In animal models of AMR, C4d staining is usually stronger on arterial than capillary endothelium. Moreover, Galvani et al (56) have demonstrated that monoclonal antibodies to HLA class I antigens can cause neointimal formation in segments of human mesenteric arteries transplanted interpositionally into the aorta of SCID/beige mice. These observations indicate that antibodies and complement can have direct effects on arterial endothelium. In clinical specimens, however, complement deposition in arteries can also be related to other pathological processes such as arteriosclerosis.

The use of heterotopic cardiac allografts in mice to determine whether alloantibodies can contribute to chronic arteriopathy has had limited success. Again, the use of polyclonal antibodies apparently is more effective than monoclonal antibodies. Russell and colleagues reported that passive transfer of polyclonal antibodies to SCID mice recipients of heart allografts resulted in extensive graft arteriopathy (57). However, the same group reported that passive transfer of monoclonal antibodies to donor MHC antigens only causes lesions localized to the origin of the coronary arteries of cardiac allografts in RAG^−/− recipients (27).

It is important to appreciate that significant anatomical differences exist between mouse and human coronary arteries that may contribute to the apparent resistance to development of arteriopathy in cardiac transplants in mice. First, the main coronary arteries in humans are located on the surface of the heart surrounded by epicardial fat. In contrast, only the first few millimeters of the coronary arteries adjacent to the aortic root are on the surface of the heart in rodents. At that point, the coronary arteries penetrate the myocardium. This is relevant to the development of arteriopathy because epicardial fat produces proinflammatory cytokines including IL-6, IL-8, and MCP-1 (58). The second difference is that the larger human coronary arteries contain vasa vasorum, which provide an “outside-in” route for the migration of inflammatory cells and cytokines. In addition, adult human hearts often have superimposed abnormalities before transplantation such as subclinical amounts of atherosclerosis (59). Preexisting atheroma is an indicator of more global underlying aging alterations. Moreover, these organs are often transplanted into patients who have had previous surgery (e.g., repairs of congenital abnormalities, replacement of valves, application of coronary bypass grafts, and insertion of ventricular assist devices) with ensuing inflammation, angiogenesis, and scarring. In many patients, this local inflammatory environment is compounded by systemic sensitization as the result of transfusions received for the surgical procedure. In contrast to humans, the animals used as donors and recipients in experimental transplantation are usually young and healthy.

Enhancement and Accommodation

Alloantibodies are not always proinflammatory. The beneficial effects of alloantibodies have been studied in rodent models of transplantation with various underlying hypotheses. Over 50 years ago, alloantibodies were found to enhance the growth of allogeneic tumors in mice (60). This phenomenon of enhancement was translated to experiments in tissue transplantation in mice and even stimulated discussions of clinical trials (61). A decade later, the beneficial effects of alloantibodies were re-explored as a possible mechanism underlying the clinical observation that patients who had received prior blood transfusions – even from the prospective organ donor – had better graft outcomes (62). These efforts were abandoned when improved immunosuppressive protocols eclipsed the effects of “donor-specific transfusions”.

More recently, interest in beneficial or at least harmless antibodies has re-emerged in the context of accommodation. The term accommodation was introduced to describe the survival of major blood group incompatible renal transplants in patients with circulating antibodies to the blood group (63). These antibodies apparently bind to the target antigen on the vascular endothelium of the graft as evidenced by the deposition of C4d. When circulating antibodies and C4d deposition are not associated with margination of leukocytes in vessels, and there is no evidence of graft dysfunction, then the graft is described as accommodated. Accommodation has been reported in patients with donor specific antibodies to HLA, but this is much less frequent and less stable than with antibodies to blood groups A or B (64). Several mechanisms for accommodation have been explored, including the increased expression of complement regulatory pathways and anti-apoptotic genes.

Upregulation of complement regulators has been reported to correlate with less severe AMR in both cardiac and renal transplants (65, 66). Interestingly, in these clinical studies decay accelerating factor (CD55) was demonstrated on capillaries of biopsies. CD55 inhibits the complement cascade after cleavage of C4 and therefore, would not diminish C4d deposits. Moreover, the early complement components have important anti-inflammatory functions. In fact, deficiencies of C1 or C4 are associated with autoimmune diseases. C1q in particular has been demonstrated to be critical in clearing apoptotic bodies by macrophages (67). It has been hypothesized that clearance of apoptotic cells by non-inflammatory mechanisms accounts for the prevention of autoimmunity by C1. In line with this hypothesis, Bishop and colleagues have recently reported that C1-deficient mice reject cardiac allografts more rapidly than normal mice (68). In summary, the mechanisms underling accommodation require further investigation as well as the possibility that these mechanisms could be applied to prevent AMR.

Therapeutic Approaches

Knowledge of the mechanisms of antibody-mediated tissue injury has provided insights for the generation of potential therapeutic interventions. For example, the finding that MHC cross-linking by antibody involves the mTOR pathway has obvious implications for the use of rapamycin (40). Understanding feedback regulation through FcγRIIB on B cells may lead to improvements in formulations of intravenous immunoglobulin (IVIg) to inhibit alloantibody production. The studies by Ravetch and colleagues demonstrating that the affinity of Ig binding to FcγRIIB is dependent on the carbohydrate side chain on the IgG (44) provide a rationale for designing modifications of the carbohydrate side chain to increase the inhibitory effect of IVIg on B cells. The complement cascade is also a therapeutic target. Studies in mouse models indicate that blocking C5a or C5a receptor inhibits macrophage infiltration and endothelial cell activation in allografts (23-25). Clinically, monoclonal antibodies to C5 have been used to inhibit AMR (69).

Finally, inhibiting the source of antibody production can be achieved by depleting B cells. Transgenic mice that express human CD20 have been developed to test the effects of Rituximab and other antibodies to CD20 (70). Tedder (71) has proposed that antibodies to CD19 may be more effective at abrogating antibody production because this target is expressed on more cells in the B cell lineage including some plasma cells.

Summary

New diagnostic tools and heightened clinical awareness have increased the recognition of AMR. However, the pathogenesis of AMR is incompletely understood. It is evident that not all of the potential mediators involved in the pathology of AMR have been defined. After decades of neglect, clinical urgency has rekindled interest in animal models to increase our understanding of AMR and to test potential therapeutic interventions. However, it will be important to view these models from the clinical perspective in order to identify relevant mechanisms and to develop applicable strategies to inhibit antibody-mediated graft injury.

Figure 1.

Diagram of major antibody initiated processes, including antigen cross-linking, Fc receptor interaction and complement activation. Antigen cross-linking causes exocytosis of vWf, expression of adhesion molecules (such as P-selectin), release of growth factors, upregulation of receptors and cell proliferation. Leukocytes with Fc receptors (eg, neutrophils and macrophages) can be stimulated by antibodies bound to antigen. The strength of Fc receptor interactions with IgG is dependent on the carbohydrate side chains (yellow hexagons) on the antibodies. The carbohydrate side chains on IgG also interact with the first components of the classical and lectin complement pathways. The many split products of complement expand the effects of antibodies in part through complement receptors (CR) on leukocytes.