Pietropaolo et al. and Pugliese et al. are correct in arguing that our study does not entirely rule out the possibility that other thymic cells can express a diverse range of tissue-specific antigens. It was neither our intention, nor do we consider it appropriate, to address the issue of gene expression in a categorical yes-or-no fashion. It has long been known that a certain degree of so-called illegitimate transcription allows for the presence of one or few transcripts of any given gene in any given cell1,2. Consequently, the detection of a specific transcript reflects, to a large extent, the sensitivity of the method applied, and the designation of a positive signal has to be defined in the particular context. We compared expression of tissue-specific genes in five different thymic cell types by carefully calibrating the amount of input cDNA according to the expression levels of housekeeping genes. The PCR data thus allow for an estimate of the relative abundance of transcripts in the respective cell types. The central observation of our study is that mTECs are clearly distinct in the strength as well as the diversity of genes that are promiscuously expressed. We have, in the meantime, extended this observation to mTECs isolated from the human thymus (B. K. et al. unpublished data).

The failure of Pugliese et al.3 to detect the expression of endocrine pancreas gene products (for example, insulin and GAD) in keratin-positive cells from the human thymus and yet their success in detecting these gene products in DCs and macrophages from the thymus and spleen is puzzling.

Pugliese et al. incorrectly state that we used an identical approach to Throsby et al.4. Throsby et al. analyzed sorted DCs and macrophages, but did not relate gene expression in these cell types to equally pure preparations of cTECs or mTECs. Instead, TEC-like cell lines and small numbers of cells of deoxyguanosine-treated thymic lobes were compared and found to be negative. We have confirmed the finding that established thymic epithelial cell lines do not display promiscuous gene expression (L. K., unpublished data). There may be numerous reasons for differences in gene expression between ex vivo isolated cells and long-term cell lines, including signals from the in vivo microenvironment. We analyzed freshly isolated cells without an intervening culture period5.

Pietropaolo et al. allude to the ill-defined PAE cells as being “an additional cell type” (distinct from mTECs) that expresses peripheral antigens. We think it likely that PAE cells are, in fact, mTECs or a subset of mTECs. This is because of the observation that the tolerogenic competence of the PAE cells, as defined by Hanahan's group, could be transferred with the thymus into nude mice and the tolerance remained robust several months after reconstitution of the T cell compartment6.

Thus, our study does not exclude expression of certain peripheral antigens in thymic cells of hematopoietic origin, indeed we have reported tolerance induction to proteolipid protein associated with bone marrow–derived cells7. In the case of genes of the endocrine pancreas, however, expression in thymic and peripheral DCs appears at least two orders of magnitude lower than in mTECs. At present we cannot reconcile our data with those reported in studies that failed to detect insulin- and GAD-specific mRNA or protein in mouse4 and human mTEC3.