Allogeneic Memory T Cells Derived from Host DC-Activated CD44^loCD8⁺ T Precursors Sustain Host Tissue Injury of Ongoing Graft-versus-Host Disease.

Graft-versus-host disease (GVHD) directed against minor histocompatibility antigen (miHAs) evolves over weeks to months, suggesting a requirement for persistent alloreactive donor T cells. In patients with allogeneic bone marrow transplantation (allo-BMT), persistency of GVHD is accompanied with elevated allogeneic CD4⁺ T cells with memory phenotype in peripheral blood. In contrast, several other studies have recently shown that T cells with memory phenotype (CD44^hiCD62L^hi/lo) from normal donor mice do not induce acute GVHD. While these T cells with memory phenotype may be induced by environmental antigenic stimulation or may represent cells undergoing homeostasis in vivo, we found that early activated donor CD44^hiCD8⁺ T cells with effector/memory phenotype upon ex vivo host DC stimulation are also functional defective in GVHD induction in vivo. However, whether alloreactive memory T cells might develop in vivo in recipient with ongoing GVHD, and if this is the case, whether these in vivo generated alloreactive memory T cells may be responsbile for persistency of GVHD, remain unknown. Using the C3H.SW anti-C57BL/6 (B6) and B6 anti-BALB.B mouse models of human GVHD directed against miHAs, we found that alloreactive CD8⁺ T cells secreting high levels of IFN-γ in recipient mice receiving C3H.SW CD44^loCD8⁺ T cells + T⁻BM peaked by day 14, declined by day 28, and increased again after 35 days of transplantation, corresponding to the kinetics of primary and memory T cell responses. Indeed, while donor C3H.SW CD8⁺ T cells recovered from these B6 mice receiving C3H.SW CD44^loCD8⁺ T cells + T⁻BM 10 days after allo-BMT, at the peak time of primary allogeneic immune response, upregulated the expression of effector marker CD25, donor CD8⁺ T cells recovered 42 days after allo-BMT from B6 mice with ongoing GVHD, at the time of memory T cell development, expressed high levels of CD44 and CD122 but down-regulated CD25. However, both d10-CD8⁺ and d42-CD8+ T cells expressed identical levels of cytotoxic molecules including granzyme B, perforin and FasL and were able to kill B6 mouse-derived EL-4 leukemic cells. Compared to naïve CD44^loCD8⁺ T cells that were lost after cultured in the presence of IL-2+IL-15 for 5 days, d42-donor CD8⁺ T cells recovered from B6 mice with ongoing GVHD survived over 5 days in the presence of IL-2+IL-15 alone and these surviving d42-CD8⁺ T cells were able to rapidly proliferate in responding to B6 DCs+IL-2+IL-15 in secondary culture. Flow cytometry analysis showed that d42-CD8⁺ T cells contained at least two distinct subsets: CD44^hiCD62L^lo (80% to 85%) and CD44^hiCD62L^hiCD8⁺(3% to 6%) T cells, resembling to the phenotype of effector memory and central memory CD8⁺ T cells, respectively. Administration of irradiated secondary B6 mice with either d42-CD44^hiCD62L^hi or d42-CD44^hiCD62L^lo CD8⁺ T cell subset recovered at day 42 from primary B6 mice receiving C3H.SW CD44^loCD8⁺ T cells + T⁻BM caused virulent GVHD. These results indicate that alloreactive memory T cells develop in vivo in recipient mice with acute GVHD where host mHAs persist and may be responsible for the persistence of GVHD. Accordingly, we suggest that in vivo blockade of both alloreactive effector and memory T cell responses will be necessary for GVHD prevention and treatment.