Prevention of Graft Versus Host Disease by Inactivation of Host Antigen-Presenting Cells

See allHide authors and affiliations

Science  16 Jul 1999:
Vol. 285, Issue 5426, pp. 412-415
DOI: 10.1126/science.285.5426.412


Graft versus host disease, an alloimmune attack on host tissues mounted by donor T cells, is the most important toxicity of allogeneic bone marrow transplantation. The mechanism by which allogeneic T cells are initially stimulated is unknown. In a murine allogeneic bone marrow transplantation model it was found that, despite the presence of numerous donor antigen-presenting cells, only host-derived antigen-presenting cells initiated graft versus host disease. Thus, strategies for preventing graft versus host disease could be developed that are based on inactivating host antigen-presenting cells. Such strategies could expand the safety and application of allogeneic bone marrow transplantation in treatment of common genetic and neoplastic diseases.

Allogeneic bone marrow transplantation (alloBMT) has revolutionized the treatment of hematopoietic malignancies, inherited hematopoietic disorders, and aplastic anemia. Unfortunately, graft versus host disease (GVHD) remains a major toxicity that greatly limits the application and efficacy of alloBMT (1). Most patients who undergo alloBMT receive stem cells from major histocompatibility complex (MHC)–identical donors. In these patients, GVHD is initiated by donor T cells that recognize a subset of host peptides, called minor histocompatibility antigens (miHAs), which are derived from the expression of polymorphic genes that distinguish host from donor (2). Presently, therapy for GVHD is limited to immunosuppression directed largely against T cells (3).

T cell responses are initiated on antigen-presenting cells (APCs). In alloBMT the unusual situation arises in which both host- and donor-derived APCs are present. We therefore examined the roles of host- and donor-derived APCs in initiating GVHD with the goal of identifying another target for GVHD prevention (4–6). Because recipient hematopoiesis is ablated by cytotoxic therapies before the transplant, and more intensive radiation augments GVHD, donor APCs might be crucial (7). On the other hand, peptides presented to CD8+ T cells by MHC class I molecules are derived primarily from endogenously expressed genes (8), so host APCs might be essential. However, extracellular antigens can be “cross-presented” on class I, thereby “cross-priming” donor T cells (9,10).

Using a MHC-identical, multiple miHA mismatched murine model of alloBMT analogous to most human alloBMTs, we examined whether host mice with APCs unable to present class I–restricted peptides would develop GVHD. We generated bone marrow chimeric mice without class I on their APCs but with class I on target tissues (11). Wild-type C57BL/6 (B6; H-2b) hosts received 1000 cGy of ionizing radiation to eradicate host hematopoiesis, including APCs, followed by reconstitution with T cell–depleted bone marrow (TdeplBM) from B6 β2-microglobulin knock out mice (β2M−/−). β2M−/− cells do not express class I and therefore cannot present peptide antigens to CD8+T cells (12). After waiting 4 months for β2M−/− hematopoietic engraftment and APC repopulation, we used these chimeras (designated β2M−/−→B6) as recipients in a GVHD-inducing alloBMT (11). The chimeras were reirradiated and then injected with T cell–depleted C3H.SW (H-2b) bone marrow (C3H.SW TdeplBM) (13) with or without 106 or 2 × 106 purified C3H.SW CD8+ T cells (14). As controls, B6→B6 syngeneic chimeras were treated identically.

In each of three experiments, the β2M−/− →B6 recipients of C3H.SW TdeplBM plus CD8+ T cells were resistant to GVHD induction. In contrast, the B6→B6 recipients of C3H.SW TdeplBM plus CD8+ cells developed severe GVHD manifested by hunched posture, skin and ear erythema, alopecia, weight loss (Fig. 1A), and death. Only 1 of 30 tissue samples from β2M−/−→B6 CD8 recipients showed evidence of GVHD, in contrast to 25 of 30 tissue samples from B6→B6 CD8 recipients (Fig. 2 and Table 1). CD8+ but not CD4+ T cells were detected in skin lesions, confirming the pathogenic role of CD8+ T cells (Fig. 2, G and H).

Figure 1

(A) Percent weight loss. Average percent weight change versus time is plotted. Groups are (squares) B6→B6 recipients of C3H.SW TdeplBM; (circles) β2M−/−→B6 recipients of C3H.SW TdeplBM; (triangles) B6→B6 recipients of C3H.SW TdeplBM plus 106 CD8+ T cells; and (diamonds) β2M−/−→B6 recipients of C3H.SW TdeplBM plus 106 CD8+ cells. (B) Survival. β2M−/−→B6 and B6→B6 chimeras were reirradiated and received C3H.SW TdeplBM with or without purified C3H.SW CD8+ T cells. Groups are (closed squares) B6→B6 recipients of TdeplBM; (closed circles) B6→B6 recipients of TdeplBM plus 106 CD8+ cells; (closed triangles) B6→B6 recipients of TdeplBM plus 2 × 106 CD8+ cells; (open squares) β2M−/−→B6 recipients of TdeplBM; (open circles) β2M−/−→B6 recipients of TdeplBM plus 106 CD8+ cells; and (open triangles) β2M−/−→B6 recipients of TdeplBM plus 2 × 106 CD8+cells.

Figure 2

β2M−/−→B6 bone marrow chimeras are resistant to GVHD induction. Histology from β2M−/−→B6 (B, D, and F) and B6→B6 (A, C,E, G, and H) recipients of C3H.SW TdeplBM and CD8+ T cells are shown: (A and B) liver, (C and D) small intestine, and (E and F) skin. Periportal mononuclear infiltrates are present in (A); apoptotic cells are found in small bowel crypts in (C) (arrow); and mononuclear cell infiltrate, fibrosis, epidermal maturation disarray, and necrotic keratinocytes (arrow) are visible in (E). These changes were absent in β2M−/−→B6 recipients. (G) Horseradish peroxidase immunohistochemistry for CD8+ cells. Note CD8+ cells invading follicles (lower left) and epidermis (arrows). (H) Immunohistochemical staining for CD4+ cells from the same mouse as in (G). Note the absence of CD4+cells in epidermis. Diffuse red staining in the dermis is background. (For all panels, 1 mm = 6.5 μm.)

Table 1

Histologic scoring of GVHD. Formalin-fixed, paraffin-embedded sections were stained with hematoxylin and eosin, randomized, and read blindly by experienced pathologists. Findings were scored and given an overall interpretation of positive (+), indefinite (+/−), or negative (−) for GVHD. N, number of mice analyzed.

View this table:

In a second experiment, β2M−/−→B6 and B6→B6 chimeras received C3H.SW CD8+ cells and were monitored for survival (Fig. 1B). Six of eight B6→B6 recipients of 106 C3H.SW CD8+ cells and 8 of 8 recipients of 2 × 106 CD8+ cells died with clinical GVHD, whereas only two deaths occurred in the 24 β2M−/− →B6 chimeric recipients of C3H.SW CD8+ cells (P = 0.0024, Fisher's exact test; comparison between all B6→B6 and β2M−/−→B6 CD8+ T cell recipients). In a third experiment in which all mice received 2 × 106 CD8+ T cells, GVHD was again inhibited in the β2M−/− →B6 C3H.SW CD8 recipients. However, in this case, delayed (40% longer mean time to onset) and less severe (46% less mean weight loss) GVHD was observed among 3 of 8 mice compared with 7 of 8 B6→B6 C3H.SW CD8 recipients.

The finding of milder and delayed GVHD among 4 of 38 β2M−/−→B6 CD8+ T cell recipients over three experiments suggested that either replacement of host APCs with β2M−/− APCs was variably incomplete, or that in a minority of mice, donor-derived APCs cross-primed donor T cells. To quantitate host APC replacement, we analyzed spleen and lymph node cells from β2M−/−→B6 chimeras by flow cytometry before the second transplant (Fig. 3). The range of residual host dendritic cells (0.2 to 17.8%) suggests that “breakthrough” GVHD was due to variable APC turnover and that complete depletion of class I+ APCs may not be required for protection from GVHD. The greater degree of host macrophage persistence (3 to 30%) suggests that they may not be important APCs in these experiments.

Figure 3

Class I expression on dendritic cells, macrophages, and B cells. (A) Class I expression by dendritic cells. Dendritic cells were isolated by digesting spleens and lymph nodes with collagenase, followed by centrifugation through 30% bovine serum albumin. Dendritic cells were identified by four-color flow cytometry. Cells staining with a multilineage mixture of phycoerythrin-conjugated antibodies to Thy1.2 (T cells), Gr-1 (granulocytes), TERR 119 (erythroid), and CD45R (B220; B cells) were excluded. Class I expression on CD11c+ cells is shown. FITC, fluorescein isothiocyanate; PerCP, peridinin chlorophyll protein. (B) Class I expression by dendritic cells (DC), macrophages, and B cells in lymph nodes and spleens of β2M−/−→B6 chimeras. Circles, individual mouse; dashes, median. Twelve mice were analyzed for splenic and lymph node dendritic cell chimerism; seven mice were analyzed for macrophage and B cell chimerism.

The failure of the β2M−/−→B6 recipients of C3H.SW CD8+ T cells to develop GVHD was not due to rejection of donor CD8+ T cells or bone marrow. Using antibodies specific for an allelic form of the T cell marker CD5 expressed on C3H.SW but not on B6 T cells, we observed C3H.SW CD8+ T cells in the β2M−/−→B6 chimeras. Nearly all of the CD11b+ and CD11c+ cells in the β2M−/−→B6 chimeras that underwent the second transplant were class I+, confirming donor C3H.SW APC engraftment (15).

Although cross-priming occurs in a variety of experimental situations (9, 10), in the β2M−/− →B6 chimeras described here, cross-presentation of host antigens by donor APCs was insufficient to generate GVHD. Therefore, either host APCs are more efficient than donor APCs in cross-presenting exogenous antigens, cross-presentation does not induce GVHD, or both. To test whether host-derived APCs could cross-present exogenous antigens, we next used C3H.SW→B6 bone marrow chimeras as recipients for an alloBMT with C3H.SW TdeplBM and CD8+ T cells. For these mice to develop GVHD, resident C3H.SW APCs would need to cross-present B6 antigens. However, such C3H.SW→B6 chimeras were completely resistant to GVHD (Fig. 4). Thus, when APCs are limited to the cross-presentation of exogenous antigens, they cannot induce clinical GVHD even when resident in a chimeric host for a long period. Although dendritic cells can efficiently cross-present antigens from apoptotic cells (10), which are in abundance after irradiation, this was insufficient to induce GVHD. Therefore, either cross-presentation is inefficient or apoptotic bodies are rapidly cleared by other mechanisms, thereby limiting their exposure to dendritic cells. These data also suggest that initial target antigens for CD8+ T cells in GVHD are restricted to proteins expressed by host APCs.

Figure 4

C3H.SW→B6 bone marrow chimeras are resistant to GVHD induction. B6 mice received two 500-cGy fractions then were reconstituted with 7 × 106 C3H.SW or B6 TdeplBM. Four months later, the chimeras were reirradiated with two 375-cGy fractions and received C3H.SW TdeplBM with or without 2 × 106 C3H.SW CD8+ T cells. Groups are (triangles) B6→B6 recipients of C3H.SW TdeplBM; (circles) B6→B6 recipients of C3H.SW TdeplBM plus CD8+ T cells; (closed squares) C3H.SW→B6 recipients of C3H.SW TdeplBM; and (open squares) C3H.SW→B6 recipients of C3H.SW TdeplBM plus C3H.SW CD8+ T cells.

Suppressor cells have been proposed as mediators of resistance to GVHD (6, 16), but appear unlikely to explain the present data. The simplest explanation for the resistance of both the β2M−/−→B6 and the C3H.SW→B6 chimeras is the absence of APCs capable of presenting host antigens via the endogenous pathway. If suppressor cells were responsible for the absence of GVHD, they would have to have developed only in the β2M−/−→B6 and C3H.SW→B6 chimeras, and not in the B6→B6 controls. In any event, if suppressor cells were involved, host APCs would be required for their development.

These results contrast with those of Sprent and colleagues (4). They found that in a CD8+T cell–dependent MHC-compatible but miHA-incompatible GVHD model donor→host bone marrow chimeras that were reirradiated and injected with donor bone marrow and lymph node cells developed GVHD. They suggested that either the chimeras were not devoid of host APCs because of insufficient radiation, or that host antigens were processed by donor marrow cells. Our data support the former explanation. Sprent and colleagues also found that heavily irradiated allogeneic parent→F1 bone marrow chimeras developed GVHD in response to very high doses of parental T cells (up to 8 × 107 T cells per recipient) and concluded that nonhematopoietic cells functioned as APCs. Given the increased precursor frequency of T cells recognizing allogeneic MHC molecules in comparison to miHA on self MHC and the large dose of T cells used, they may have unmasked the presence of small numbers of residual host APCs. Alternatively, nonhematopoietic cells when challenged with very large numbers of allogeneic T cells may cause T cell activation sufficient to induce GVHD.

Our results suggest that depleting host APCs before the conditioning regimen should abrogate GVHD without the need for prolonged T cell–targeted immunosuppression. Such an approach, perhaps using toxin-conjugated or radiolabeled antibodies, could expand the range of diseases treated with alloBMT. To test the feasibility of in vivo antibody-mediated depletion of host dendritic cells, we injected mice with N418, a hamster monoclonal antibody to the β integrin CD11c expressed on murine dendritic cells (17). The injected anti-CD11c bound to all CD11c-expressing dendritic cells in both lymph node and spleen (15), supporting the feasibility of antibody-mediated APC depletion.

A subset of alloBMT recipients have self-limited GVHD, which was presumed to reflect acquired T cell tolerance. Our data suggest another explanation: replacement of host with donor APCs abrogates T cell activation. Infusions of T cells from original bone marrow donors given to relapsed leukemia patients months to years after the initial alloBMT (18) cause less GVHD than has been observed when T cells are given at the time of transplantation (19). Although there may be other explanations (20), we suggest that the replacement of host with donor APCs reduces the chance of a donor CD8+ T cell interacting with a GVHD-inducing host APC. In addition to suggesting explanations for these clinical observations, our data provide the foundation for a different strategy for reducing GVHD-host APC depletion. This approach may avoid the problems associated with T cell depletion of marrow allografts: failure of engraftment, poor immune reconstitution, and lack of immunoreactivity against the tumor.

  • * These authors contributed equally to this work.

  • To whom correspondence should be addressed. E-mail: emersons{at}


View Abstract

Navigate This Article