Cutaneous leucocyte composition after human allogeneic bone marrow transplantation: relationship to marrow purging,...

Immunohistological and morphometric techniques were used to study the skin after marrow transplantation with particular reference to the relationship of marrow purging, the presence of a clinical rash and histological changes to leucocyte numbers and phenotype. Recipients of T-cell-depleted marrow showed significant reductions in CD2+, CD4+ and CD8+ T-lymphocytes in the first 22 d after transplantation but not after this time. T-cell numbers in recipients of unpurged marrow were similar to those of normal donors, indicating a rapid repopulation by cells from the graft. Langerhans cells (CD1+ dendritic cells) and macrophages, on the other hand, were present in similar numbers in both groups of patients within the first 22 d; the former in low and the latter in normal numbers. Biopsies exhibiting graft versus host disease showed increases in CD2+, CD4+ and CD8+ T-lymphocytes with significant lowering of the CD4:CD8 ratio. A proportion expressed markers of activation and HNK1+ cells and macrophages were also increased. Biopsies exhibiting epidermal basal abnormalities only (changes identical to graft versus host disease but without detectable leucocyte infiltration on conventional microscopy) showed a minor increase in macrophages and HNK1+ cells but no other leucocyte alterations to suggest a pathogenetic link with graft versus host disease. Langerhans cells were reduced in these biopsies, however, when taken more than 22 d post-transplant, suggesting that the epidermal changes are associated with Langerhans cell damage or repopulation. We were unable to identify any significant alteration in leucocytes in patients with strong clinical evidence of graft versus host disease but with histologically unremarkable biopsies. Although it is possible that perivascular increases in T-cells and expression of activation markers precede the characteristic histological picture of graft versus host disease the time scale is probably too short to allow diagnostic value.