Surgical samples have long been used as important subjects for cancer research. In accordance with an increase of neoadjuvant therapy, biopsy samples have recently become imperative for cancer transcriptome. On the other hand, both biopsy and surgical samples are available for expression profiling for predicting clinical outcome by adjuvant therapy; however, it is still unclear whether surgical sample expression profiles are useful for prediction via biopsy samples, because little has been done about comparative gene expression profiling between the two kinds of samples.

A total of 166 samples (77 biopsy and 89 surgical) of normal and malignant lesions of the esophagus were analyzed by microarrays. Gene expression profiles were compared between biopsy and surgical samples. Artificially induced epithelial-mesenchymal transition (aiEMT) was found in the surgical samples, and also occurred in mouse esophageal epithelial cell layers under an ischemic condition. Identification of clinically significant subgroups was thought to be disrupted by the disorder of the expression profile through this aiEMT.

This study will evoke the fundamental misinterpretation including underestimation of the prognostic evaluation power of markers by overestimation of EMT in past cancer research, and will furnish some advice for the near future as follows: 1) Understanding how long the tissues were under an ischemic condition. 2) Prevalence of biopsy samples for in vivo expression profiling with low biases on basic and clinical research. 3) Checking cancer cell contents and normal- or necrotic-tissue contamination in biopsy samples for prevalence.