Teshima et al. reply

The issues raised by Mielcarek et al. are important and help to clarify our study's conclusions1, which we believe are consistent with their examples. Our first conclusion, that host APCs are needed for both the activation and effector phases of acute GVHD, is supported by additional data published in the dog study2. After infusion of naive donor lymphocytes, none of the nine chimeras developed significant clinical GVHD, and none developed histopathologic lesions in the skin. Thus, alloantigen on target epithelium alone did not elicit acute GVHD.

Our second conclusion was that alloantigen expression on target epithelium is not always required for GVHD damage. This is especially true for GVHD mediated by donor CD4+ T cells. We also showed that when acute GVHD is mediated by donor CD8+ T cells (as is the case for most GVHD elicited by mHAg), alloantigen on target epithelium increases GVHD mortality but does not account for all of it. The dog study demonstrated that GVHD occurred after donor lymphocyte infusion only when donors had been previously sensitized to the host. Still, only two of twelve animals showed pathologic changes of GVHD in the skin, and only one showed severe changes2. We therefore agree that donor cells, especially when primed, can sometimes attack GVHD target epithelium directly, a conclusion recently confirmed by a mouse GVHD model using donor CD8+ T cells activated by mHAg6.

The example of the patient with chronic GVHD who rejected a third-party skin graft but not a donor skin graft3 is consistent with a donor response to alloantigen on APCs but not necessarily on the epithelium, where damage could be mediated by cytokines, cytotoxic T lymphocytes or both. The protection of the donor skin graft in this example emphasizes the importance of local APCs for the induction of GVHD in that target tissue, a feature not stressed in our study but shown for hepatic GVHD in a mouse model7. Such APC geography is probably a factor in the unusual target organ distribution of GVHD, although other variables such as chemokines and trafficking molecules may also play a role.

Our third conclusion was that inflammatory cytokines can mediate mortality and target destruction, because extended blockade of both IL-1 and TNF-α prevented both of these outcomes. We agree that inflammatory cytokine activity alone is not sufficient to induce acute GVHD because cytokine infusion does not recreate the full disease spectrum, even in the mouse. Two randomized trials of single cytokine antagonists did not show any benefit in preventing GVHD4,5. Pentoxyfylline was poorly tolerated, and no cytokine data were recorded4. IL-1 receptor antagonist was only administered until day 10 and its administration did not change TNF receptor levels5. Prevention of clinical GVHD may require blockade of both cytokines, a longer blockade or both. Although we do not agree with the authors' conclusions that both recipient APCs and epithelial cells must express alloantigen for the induction of acute GVHD, we concur that insights from all animals models (both inbred and outbred) must be extrapolated to human patients with caution and that the effectiveness of any agent or approach must be verified first in well designed, carefully controlled clinical trials.

See “Is alloantigen expression by host epithelium required for acute graft-versus-host disease?” by Mielcarek et al.