Fonsatti et al. in this issue have offered some criticisms of our article on the role of methylation in the turn-off of MHC class I antigens in the MSR3 melanoma cell line. In our report,1 we described the role of methylation only as one more mechanism among those reported to date that lead to HLA total loss (Phenotype no. I). We did not intend to claim that the behavior shown by cell line MSR-3 was the rule, and indeed it may be the exception, as Fonsatti et al. note. In fact, other mechanisms (β2m inactivation, TAP/LMP downregulation, altered DNA-binding factors and oncogene overexpression) are observed more frequently and also contribute to a similar phenotype.2, 3, 4, 5, 6 In particular, alterations caused by those mechanisms involving structural alterations to β2m lead to irreversible losses that will not respond to treatment with IFN-γ or demethylation. We proposed a role for methylation in MSR3-mel only after careful evaluation of the cells to rule out β2m mutation, genomic loss, nuclear transcription factors downregulation and loss of inducibility by cytokines. The conclusions Fonsatti et al. reach after their study of cell lines with total loss (Table I, p. XX) cannot be sustained unless structural mechanisms are ruled out because some of these mechanisms are likely to have occurred in at least some of their lines. Moreover, the special behavior of MSR3-mel after demethylation may result from the fact that HLA genes in this cell line are hypermethylated, as shown in Southern blot analyses. Do the authors have information regarding the methylation state of the HLA genes in the lines they studied? This information is fundamental to observe the effects of 5-AZA-CdR. In our study, we reconciled our findings with those of earlier studies on the role of methylation in regulating HLA class I gene expression by recalling that a minimum level of methylation may be needed to maintain HLA class I antigen expression.7 This is why our findings for level of expression were clear in mRNA analyses but less evident in assays to demonstrate surface expression. To better document HLA class I expression, we used INF-γ treatment in studies designed to observe the recognition of MAGE-specific cytotoxic T lymphocytes. We feel that the arguments presented by Fonsatti et al. would have been more solid, and more directly comparable to our findings, if they had reported mRNA levels rather than FACs analyses. The authors also claim that HLA-A2 expression is not recovered in lines in which this antigen has been lost. From this they deduce that hypermethylation does not represent a mechanism used to turn off HLA-A2 expression. Have the authors taken into account the possibility of a structural alteration in the HLA-A2 gene, as a result, for example, of loss of heterozygosity associated with chromosome 6p21? Their findings point in this direction, unless the pattern of HLA-A2 methylation is assumed to differ somehow from the methylation of other HLA antigens that were induced in the demethylation assays. This would account for the lack of response to 5-AZA-CdR. In this connection, our experience with HLA allelic losses suggests that most such instances are due to loss of heterozygosity (HLA haplotypic loss), mutational events or aberrant splicing.8, 9, 10, 11, 12, 13 Fonsatti et al. should determine the mechanism of HLA-A2 loss in the cell lines they tested before they undertake experiments to recover the expression of this molecule. In some experiments, we used 3 μM and up to 10 μM 5-AZA. Cell line MSR3-mel was resistant to prolonged culture in the presence of 5-AZA-CdR for periods longer than those reported in our article. Cultures were initiated and maintained at both participating laboratories, and no toxic effects were reported during the study. We note, however, that this behavior may have reflected hypermethylation not only of the HLA genes but also of many other genes. This would probably lead to the particular behavior of the cell line during demethylation, in which both cis- and trans-acting factors may modify not only HLA expression but also many other characteristics. In conclusion, we have emphasized that multiple mechanisms are responsible for HLA class I defects observed in tumor cells.14, 15 and that the characteristics of the tumor tissue, and in particular the degree of conservation of different components of the antigen presentation machinery, need to be determined in detail before immunotherapeutic decision-making processes can be placed on a firmer footing. Finally, we suggest that the treatment of demethylating agents should be attempted only after the intrinsic mechanisms involved in downregulation of MHC antigens have been identified and verified. Yours sincerely, Francisco Ruiz-Cabello*, Catia Traversari , Federico Garrido*, * Department of Analisis Clinicos, Hospital Universitario Virgen de las Nieves, Universidad de Granada, Granada, Spain, Cancer Immunotherapy and Gene Therapy Program, Scientific Institute H.S.Rafaele, Milano, Italy.