Allorecognition Unveiled: Integrating Recent Breakthroughs Into the Current Paradigm

Direct T cell allorecognition depends on the recognition by recipient T cells of intact allogeneic MHC molecules expressed on graft cells. Left uncontrolled, this phenomenon rapidly leads to TCMR (Figure 1), since 1%–10% of a given individual’s T cell repertoire is capable of recognising intact allogeneic MHC [5, 6].

There are two theories to explain this phenomenon [7–10], restricted to major alloantigens.

The “multiple binary complex” hypothesis (Figure 2) suggests that the principal factor determining the strength of T cell allorecognition is the wide variety of antigenic peptides presented by the donor’s MHC molecules. Because differences in the MHC molecules between donor and recipient are concentrated in the peptide-binding grooves, donor and recipient cells present different peptides derived from the exact same protein. Thus, one donor MHC molecule can potentially represent a myriad of different antigenic complexes depending on the particular peptides bound. The “multiple binary complex” hypothesis was first suggested by the observation that T cell clones that responded to allogeneic antigen presenting cells (APCs) presenting peptides derived from human albumin did not respond to the same APCs presenting peptides from bovine albumin [11]. According to this theory, the ability to recognise endogenous peptides, and not the ability to “directly” interact with allogeneic MHC, is the main factor determining alloreactivity [12–14]. Heterologous immunity is a special case, which nevertheless fits into this theory. During an antiviral response, memory T lymphocytes are generated that recognise a self MHC/viral peptide complex. Molecular mimicry between this complex and donor MHC/peptide can divert the antiviral memory response into an allogeneic response [15, 16]. The multiple binary complex theory also explains the occurrence of TCMR in recipients of HLA-identical transplants. In this case, alloreactivity is not driven by polymorphisms in the binding groove (it is identical in donor and recipient), but rather by the fact that intra-familial HLA-identical donors are not genetically identical like monozygotic twins. While they may share the same HLA genes, the rest of their genomes can differ significantly.

Finally, it shall be kept in mind that the expression of MHC-I-peptide complexes on the cell surface depends on the function of various intracellular assembly factors, such as the transporter associated with antigen presentation (TAP), tapasin, calreticulin, ERp57, TAP-binding protein related (TAPBPR), endoplasmic reticulum aminopeptidases (ERAPs), and proteasomes. It is therefore conceivable that polymorphism in these proteins also contribute to differences in the spectrum of HLA bound peptides (independently of differences in HLA binding groove and polymorphism of minor histocompatibility antigen) [17, 18].

According to the second theory – “the high determinant density theory” - the TCR directly recognises the MHC polymorphism, independently of the peptide antigen (Figure 2). Therefore, because potentially every peptide-MHC complex on the surface of the allogeneic APC is recognised as foreign, a strong overall response is triggered, even though the strength of the interaction between the TCR and individual alloMHC/peptide complexes may be very low [19].

These cognitive studies were mainly conducted in mouse models, or on a few specific HLA molecules, which do not encompass the complexity and extreme variability of HLA in humans. It is reasonable to think that for a particular donor/recipient pairing, the modality of interaction between the recipient TCR and the donor MHC lies within a range between the two theories described above.

In transplantation, it has long been theorised that donor’s leukocytes leave the graft rapidly after transplantation [20]. These passenger leukocytes, in particular the donor’s APCs, reach the recipient’s secondary lymphoid organs (SLOs), and present intact allogeneic MHC molecules on their surface to recipient’s T cells [21]. According to this theory, the incidence of TCMR is directly linked to the presence of donor passenger leukocytes, and it is therefore expected for this incidence to decrease over time as the pool of donor leukocytes fades progressively and cannot be replenished [22] (Figure 3). The fact that vascularised composite allografts (VCA), in which the stock of donor APC can self-renew (i.e; Langerhans cells in the skin), are often targeted by late TCMR, can also be seen as an indirect validation of this theory [23, 24].

The development of an effective cytotoxic response requires the provision of CD4⁺ T cell help to the CD8⁺ T cell [25]. In the context of direct allorecognition, this help can be provided by CD4⁺ T cells with direct specificity [26], presumably because within the recipient’s SLO, a single donor APC can establish a productive three-cell cluster with CD4⁺ and CD8⁺ T cells that directly recognise the MHC class II and class I alloantigens, respectively, on the cell surface (Figure 1).

This allorecognition pathway offers the advantage of a treatment that seems readily available, namely, depletion of passenger leukocytes. Thus, several studies have been able to show, in heart, kidney [27] and liver transplantation [28], that depletion of donor APCs induces prolonged graft survival in the absence of immunosuppressants. However, there is a major bias in these studies: depletion of donor APCs not only eliminates the function they carry, but also the stock of antigens (in particular MHC-II) that they strongly express. Notably, Mandelbrot et al have reported that cardiac allografts from donors whose APCs were unable to deliver costimulatory signals to recipient T cells were still rejected, despite the apparent inability to trigger conventional direct-pathway T cell responses [29].

The limitations discussed above and the observation that TCMR can occur late, at a time when the pool of passenger donor leukocytes is expected to be exhausted, led to reconsideration as to how alloimmune responses with T cells with direct specificity are triggered.

A series of publications from the early 2000s have instead suggested that presentation of MHC alloantigen that has been acquired by the recipient APC and represented intact without processing may be key [30] (Figure 1). These recipient’s APCs could have taken up donor antigens by direct cell-cell contact [31] with either migrating donor leukocytes or graft parenchymal cells (recipient APCs rapidly infiltrate the grafts after transplantation [32]). Alternatively, alloantigen may be acquired from donor-derived exosomes or extracellular vesicles drained from the graft to recipient’s SLOs [33, 34]. It should be noted that donor-derived exosomes cannot activate T cells on their own but must be presented by donor APCs. This capture by recipient’s APCs and the subsequent presentation to T cells only occurs in an inflammatory context [35], possibly explaining the link sometimes observed between infection and rejection, the causality of which has been difficult to establish [36, 37].

The semi-direct pathway has become gradually recognised as an important mechanism leading to TCMR, since studies (almost exclusively on murine models) by independent groups suggest that it elicits much stronger CD8⁺ T cell immunity than interaction of the CD8⁺ T cell with a donor leukocyte, principally because alloantigen from a single donor APC can be presented intact by many-more fold recipient APCs. The semi-direct pathway can still explain the observed kinetics of TCMR after transplantation (Figure 3). Indeed, in animal models of chronic renal graft rejection, cross-dressing, although declining with time, persists for many weeks after transplantation [2]. The initial peak is probably of mixed cause, linked to the donor’s dendritic cells, which remain a major source of alloantigens in the early stages of the transplant, but also to ischaemia-reperfusion lesions, responsible for the immunogenic death of graft cells and the release of extracellular vesicles. Finally, although, as discussed above, the default pathway for provision of CD4⁺ T cell help to alloreactive CD8⁺ T cells is through direct pathway allorecognition of MHC class II alloantigen on the surface of donor APCs, Lee and Auchnicloss’ seminal paper has demonstrated that indirect-pathway CD4⁺ T cells can also provide effective help for generating effector cytotoxic CD8⁺ T cell alloimmunity [38]. Explaining this phenomenon through the immunological paradigms prevalent at publication was not straightforward, because it necessitated formation of a cumbersome four cell cluster, comprising a donor dendritic cell presenting intact MHC class I alloantigen to an alloreactive CD8⁺ T cell, and a recipient dendritic cell presenting processed alloantigen to an indirect-pathway helper CD4⁺ T cell, with moreover no apparent physical linkage between the former two and the latter two cell types. This risks uncontrolled and damaging bystander T cell activation. Semi-direct recognition provides an elegant solution to obviate these concerns, because, assuming a single recipient dendritic cell presents both intact MHC class I alloantigen and self-MHC class II restricted allopeptide for, respectively, direct-pathway CD8⁺ and indirect-pathway CD4⁺ T cell allorecognition, this recreates a 3-cell cluster model (Figure 1), with physical linkage possible between the three [39–41].

Although, as discussed, indirect pathway of allorecognition can provide help for generating cellular cytototoxic alloimmunity, its role in the provision of help to allospecific B cells is also important for determining graft outcomes. Indirect allorecognition is akin to conventional CD4⁺ T cell recognition of model protein antigen, whereby alloantigen is presented as processed peptide held in the binding groove of the recipient’s MHC class II antigen. Essential help for thymo (T) dependent antibody responses is provided by CD4⁺ T cells that similarly recognise complexes of MHC class II and bound peptide antigen, after that antigen has been internalised and processed via the B cell receptor (BCR). Thus, recipient indirect-pathway CD4⁺ T cells provide help for generating alloantibody [48–50] (Figure 1). The scientific literature on T-dependent humoral immunity has expanded prolifically over the last decade, with several recent publications considering how these advances shape our understanding of the B cell alloresponse.

Unlike cellular alloimmunity, whose functional relevance is confined to the response against donor major histocompatibility alloantigens [51], the humoral alloimmune response can be initiated by all alloantigens, both major and minor [52]. Alloreactive B cell responses are targeted generally against intact, conformational antigens [53, 54], which means that alloantigens must be transported from the graft to the SLOs for presentation to B cells. This may be achieved by exodus of donor DCs from the graft [20]. The graft may also release extra-cellular vesicles covered with donor MHC that are then captured by recipient DCs [54] or by subcapsular sinus macrophages in lymph nodes (or their equivalents in the spleen) [55] for presentation to allospecific B cells. In parallel, recipient DCs will acquire alloantigen either within SLOs or the graft itself and present it to indirect-pathway CD4⁺ T cells [34], which then migrate to the border of B cell zone and provide cognate B cell help for triggering initial production of class-switched alloantibody via short-lived extrafollicular foci. It is, however, the subsequent relocation to the B cell follicle and the formation of the Germinal Centre (GC) reaction, with essential help provided by further differentiation of indirect-pathway CD4⁺ T follicular helper cells [49], that likely determines transplant outcome. GC responses are uniquely capable of producing somatically-mutated, high-affinity alloantibody and animal models have confirmed the essential role of the GC response in mediating chronic antibody mediated rejection [56, 57]. The GC reaction also produces robust humoral memory, composed of memory B cells and bone-marrow resident, long-lived-plasma cells, with the latter potentially capable of producing alloantibody for the life of the individual, long after the GC response has dissipated. It is this aspect of the GC reaction that makes treatment of AMR and the desensitisation of patients awaiting transplantation so challenging [58, 59].

Clinicians’ experience and some studies have reported unusually strong de novo DSA responses developing early (within the first week) following lung and intestinal transplantation [60]. This early onset of de novo DSA suggests that in addition to the canonical indirect pathways, other mechanisms may be responsible for an alloimmune humoral response. In line with this hypothesis, recent work, published independently by Pettigrew’s and our own team [61, 62], has highlighted a major role for passenger donor CD4⁺ T cells that are contained in vascularised grafts and transferred to the recipient after the transplantation. This passenger CD4⁺ T cell population would be expected to contain a relatively large proportion (1%–10%) of cells with direct allospecificity for the recipient’s MHC class II antigens [what is true for recipient cells is also true for donor-derived cells [5, 6]], and are therefore capable of binding with these antigens on the surface of recipient haematopoietic cells. Thus, during inverted direct allorecognition (Figure 1), initial activation of recipient alloreactive B cells is triggered by binding B cell receptor (BCR) to its target alloantigen (as occurs with encounter with any classical antigen). Internalising and processing of the bound alloantigen results in upregulated expression of surface MHC class II presenting bound peptide alloantigen, but rather than cognate help being provided by indirect-pathway CD4⁺ T cells, these MHC class II complexes are recognised in a peptide-degenerate manner by activated passenger donor CD4⁺ T cells, which in turn deliver a second (costimulatory) signal to the B cell, enabling it to differentiate into a DSA-producing plasma cell [61, 62].

To date, a definitive analysis of how the kinetics of DSA production via the inverted direct pathway compares to conventional indirect-pathway CD4⁺ T cell help has not been performed, and it remains unclear whether the very early onset of alloantibody production (as soon as day 7) is mediated exclusively via the inverted direct pathway (Figure 3). A study of the precursor frequencies of the relative T cell pools mediating each pathway may provide some insight. In the indirect pathway, the frequency of the T cells involved is very low (clonal frequency of around 1/10,000). As a result, time-consuming clonal expansion is necessary to achieve an effective response. In contrast, the recently described pathway involves a very large pool of T cells (1%–10% of the T cells contained in the graft), capable of immediately providing effective assistance to B cells, without amplification. Moreover, the majority of transferred donor CD4⁺ T lymphocytes with direct allospecificity for the recipient would likely exhibit a memory phenotype, because of cross-reactive heterologous immunity [15, 16], one would expect that this would provide robust help for even more rapid production of alloantibody [50, 63]. This also explains the seemingly paradoxical finding that heart grafts from donors sensitised to recipients were rejected more rapidly than grafts from naïve donors [61].

In clinical terms, the incidence of early DSAs is immediately correlated with the richness of the graft in lymphoid cells: renal grafts produce very few early DSAs, whereas lung and intestinal grafts (which contain a mucosa-associated lymphoid tissue) induce a massive wave of early de novo DSAs, in 25%–80% of patients [62]. It is also worth noting that B cells activated via inverted direct allorecognition should also be capable of ‘soliciting’ their own help from recipient CD4⁺ T cells [64]. Our recent work has suggested that whereas passenger donor CD4⁺ T cells trigger a rapid humoral response in the recipient, maintenance of this response as a GC reaction was dependent upon secondary differentiation of allopeptide-specific T follicular helper cells of recipient origin [57, 65].

One further possible consequence of inverted direct recognition is that, because the help provided by the donor CD4⁺ T cells is likely to be promiscuous and provided to all B cells in a peptide-degenerate fashion, the limiting factor in antibody production is availability of target antigen to bind the BCR. Consequently, in some murine models, inverted direct recognition was associated with production of class-switched anti-nuclear autoantibody responses [65]. Thus, this phenomenon may be responsible for the various autoantibody responses that have been detailed in human transplant recipients [66].

NK cells are prototypical of the paradigm shift from adaptive to innate rejection. Long implicated as secondary effectors of the humoral arm through their ability to mediate antibody-dependent cellular cytotoxicity (ADCC), they may have additional and independent role in allorecognition and innate allograft damage.

For the sake of clarity, we have separated in this review the mechanisms of allorecognition dependent on innate immunity from those dependent on adaptive immunity, but this last part brings us back to the very beginning of this review: the priming of alloreactive T cells.

The initiation of a T cell response requires the activation and full differentiation of APCs. According to the danger theory, this activation is triggered by danger-associated molecular patterns (DAMPs), molecules of all kinds that can be released by certain types of cell death (necroptosis for example). However, in the context of transplantation, the danger is probably not sufficient to initiate an alloimmune response. Indeed, recent work in mice has unequivocally demonstrated that i) the mouse immune system (particularly monocytes) can distinguish between self and non-self [77] and that ii) this recognition of non-self by monocytes is necessary for the development of an alloimmune T response [78]. Indeed, in syngeneic transplantation, monocytes that differentiate into DCs are incapable of sustaining a T response despite the DAMPs provided by ischaemia-reperfusion inherent in transplantation. In contrast, in an allogeneic context, monocyte-derived DCs induced by allogeneic cells are perfectly mature, express IL-12 and effectively stimulate T cell proliferation. This mechanism of innate myeloid recognition is independent and acts on its own account, since it does not depend on DAMPs receptors [78]. Indeed, the same authors demonstrated that it depends (at least in part) on the recognition, by the recipient’s monocytic CD47, of signal-regulatory protein (SIRP) α polymorphisms in the donor [79] (Figure 1).

Myeloid allorecognition can also act on its own behalf. Monocytes carry paired Ig-like receptors (PIR) which can recognise allogeneic MHC-I molecules on the surface of donor cells. Monocytes activated in this way can contribute to allograft rejection, and even establish an innate memory response [80]. The establishment of memory via PIRs requires the SIRPα signal described above at the time of monocyte priming [79, 80].

These recent findings have yet to be formally confirmed in humans. Recent studies have characterised an infiltrate of non-classical monocytes expressing both CD47 and leukocyte immunoglobulin-like receptors (LILRs, orthologs of murine PIRs) in rejection biopsies of kidney grafts [81, 82], and suggest that these monocytes may support the CD8⁺ T cytotoxic response [82]. Further studies are needed to confirm the role of innate myeloid allorecognition in human transplantation.