Quantitation and statistical analysis
For each experiment, bands corresponding to the variant messenger 			 RNA (ie ERC4, ERD3 or ERD5) and to WT-ER were excised from the gel and counted 			 in a scintillation counter. For each set of primers (ie ERC4, ERD3 and ERD5 			 primer set) and for each sample, four independent PCR assays were performed. 			 The ratios between ERC4, ERD3 or ERD5 signals and corresponding WT-ER signal 			 were calculated. For each experiment, in order to correct for overall 			 interassay variations (due to different batches of radiolabelled [α 			 -32P] dCTP or of Taq DNA polymerase), the ratio observed in the same 			 particular tumor (case number 12) was arbitrarily given the value of one and 			 all other ratios expressed relatively. Under our experimental conditions, some 			 samples did not have measurable levels (ie signal lower than twice the 			 background value) of ERD3 or ERD5 variant messenger RNAs (see Figs 			 2a and 3a) in any of the four 			 repetitions performed. Only cases that had detectable levels in at least three 			 of the replicates in both their normal and tumor compartments were included in 			 the statistical analysis. The significance of the differences in the relative 			 levels of expression of ERC4, ERD3 and ERD5 messenger RNAs between matched 			 normal and tumor components was determined using the Wilcoxon signed-rank 			 test.