To the editor—In the September issue of Nature Medicine, Sanchez-Prieto et al.1 reported that the adenoviral E1A protein induces the EWS–FLI1 rearrangement specific to Ewing tumor2 in a variety of human cells. These results led to the proposal that EWS–FLI1 expression may be important for the transforming properties of E1A, a protein already known to interfere with cellular pathways relevant to oncogenesis, such as those controlled by p53 and Rb. They also raised the revolutionary hypothesis of the Ewing tumor being of viral origin.

We have attempted to reproduce these results using HEK and HEK293 cells from various sources. Surprisingly, although HEK293 cells, in contrast to parental HEK cells, express high levels of both 289R and 243R isoforms of E1A, the EWS–FLI1 protein could not be detected in any of these cell lines. Moreover, the highly sensitive RT–PCR method did not detect an EWS–FLI1 fusion transcript.

We also note discrepancies in the figures presented by the authors. In Fig. 1 (ref. 1), the established HEK293 cell line expresses a type 1 EWS–FLI1 fusion transcript. Conversely, Fig. 2c and d indicate that this same cell line expresses a protein and a fusion transcript of the same molecular weight as that of the RDES cell line known to have an EWS–FLI1 type 2 RNA and protein3. The authors claim that the different sizes of the EWS–FLI1 proteins or cDNAs found throughout the manuscript are linked to different retroviral infections and correlated with the variety of fusion transcripts found in Ewing tumors4. Although this explanation may be relevant for the results obtained with retrovirally infected IMR90 cells, it does not account for the established HEK293 cell line.

Being familiar with fluorescence in situ hybridization of Ewing cells and having provided the authors with the cos1D1 and cosG9 cosmids5 (as mentioned in their acknowledgments), we believe that the results of Fig. 3 are inconsistent with a t(11;22) translocation. In Fig. 3a, only one derivative chromosome is seen, whereas two would be expected. More strikingly, Fig. 3b shows a chromosome hybridizing with both a centromeric chromosome 11 probe and the cosG9 chromosome 22 cosmid; this is therefore suspected to be a der(11) chromosome. However, as the cosG9 cosmid lies proximal to the translocation breakpoint region on chromosome 22, it is always localized on the der(22) and never on the der(11) chromosome in the case of a Ewing t(11;22) translocation.

We cannot rule out the possibility that some E1A-expressing IMR90 or HeLa cells may occasionally present an EWS–FLI1 fusion, as we have not studied these cells. However, we would like to emphasize that the presence of an EWS–FLI1 fusion is not a characteristic of the type 5 adenovirus-transformed HEK293 cell line. We can therefore conclude that the generation of the t(11;22) translocation does not necessarily follow transformation with E1A.