To the Editor:

In a recent Nature Genetics letter entitled “Estrogen receptor alpha (ESR1) gene amplification is frequent in breast cancer,” Holst et al. report that more than 20% of breast cancers harbor genomic amplification of the ESR1 gene1. The authors also suggest that ESR1 amplification may help to identify a subgroup of estrogen-positive breast cancers likely to have a good response to anti-estrogen therapy. As the authors acknowledge, such an observation constitutes a rather unexpected finding in light of a number and variety of studies that have focused on the structure, expression and function of the estrogen receptor gene in breast cancer cells since its discovery more than 20 years ago.

Given the clinical importance of this finding, this report prompted us to investigate the status of the estrogen receptor gene in a series of 381 breast cancers studied by BAC-array CGH (aCGH). This series included 360 tumors (184 invasive ductal, 88 ductal in situ, 27 lobular, 24 micropapillary and 37 medullary carcinomas) and 21 breast cancer cell lines. The aCGH contained 3,342 sequence-validated BACs covering the human genome at the mean density of one BAC per megabase2,3,4. In particular, it included the RP11-450E24 BAC, which was used by Holst et al. to monitor ESR1 gene amplification, and BAC CTD-2019C10, which contains the ERBB2 gene. Cy5 (tumor DNA) to Cy3 (control DNA) ratios at each BAC locus were determined and analyzed using previously published spatial normalization, VAMP (visualization and analysis of CGH array, transcriptome and other molecular profiles) and GLAD (gain and loss analysis of DNA) analysis procedures5,6,7 (Supplementary Methods online).

We investigated these two loci with reference to the twofold copy number increase (2× threshold) used by Holst et al. to define amplification. On our aCGH platform, the 2× threshold was calculated on the basis of the log2 ratios for single-copy gain of chromosome X loci in normal female/male hybridizations (the median log2 ratio of 130 chromosome X clones is +0.49, approximated to 0.5 in Supplementary Fig. 1a,b online). The single-copy loss was determined by chromosome 1p log2 ratios in a series of 34 oligodendrogliomas with 1p/19q deletions3 (the median log2 ratio of 340 chromosome 1p deleted loci is −0.48). To estimate the relationship between FISH and aCGH data, we also took advantage of 49 cases that could be studied by both approaches at the ERBB2 locus. Results obtained with both techniques were strongly correlated (r = 0.83), with an aCGH log2 ratio of 0.5 corresponding to an absolute FISH copy number of 4.8, hence close to the twofold copy number increase defined above if taking into account the hyperdiploid status of most breast cancers.

It seemed that the distribution of ratios for the ERBB2 and ESR1 loci were markedly different. All tumors defined as HER2-positive by immunohistochemistry8 showed aCGH ratios higher than the 2× threshold (Supplementary Fig. 1a). In contrast, concerning the ESR1 locus, only three cases (3/341 interpretable cases; 0.9%) crossed the 2× threshold (Supplementary Fig. 1b), one of these being clearly estrogen receptor (ER)-negative by immunohistochemistry9. Moreover, we did not observe any statistically significant difference in ESR1 aCGH ratios between ER-positive and ER-negative tumors.

To rule out the possibility that low-level genomic amplification of ESR1 may have escaped detection on aCGH because of technical reasons, we carried out quantitative PCR (qPCR) and FISH for validation. To investigate copy number at the ESR1, ESR2 and TGFBR3 loci, we used qPCR on a subset of 168 cases, including 2 out of the 3 cases with aCGH ratios higher than the 2× threshold (Supplementary Table 1 online lists primer sequences). The aCGH and qPCR results were strongly correlated at corresponding loci (r = 0.6). Four cases, including the two cases previously detected by aCGH, had qPCR ratios higher than the 2× threshold (Supplementary Fig. 1c). In the other two cases, this increased qPCR ratio was rather due to a relative copy number loss at control loci, as suggested by decreased aCGH ratios at corresponding and flanking BACs. We did FISH analysis on two of the breast carcinomas with ESR1 aCGH ratios over the 2× threshold and for which material was available. As positive controls for amplification, sarcomas with 6q amplicons encompassing the ESR1 locus that were initially characterized on a dedicated aCGH10 and further analyzed on the genome-wide aCGH used herein were also investigated. As for ERBB2, we observed a very strong correlation between aCGH ratios and FISH copy numbers at the ESR1 locus (r = 0.9; Supplementary Fig. 1d,e), showing that the very few breast cancers with notably increased aCGH ratios indeed harbored an increased copy number by FISH. The aCGH ratios of 1.86 and 1.69 for the two breast cancers corresponded to FISH ratios (number of spots at the ESR1 locus as compared to the control locus) of 1.65 and 1.56, respectively. This correlation curve pointed out that our aCGH platform was even more sensitive to detect copy number increase at the ESR1 locus than at the ERBB2 locus.

The study of this large series of breast cancers with aCGH, qPCR and FISH indicated that less than 1% of breast cancers harbored a notable increase of ESR1 copy number and that, notwithstanding subtle variations between techniques, these increases never reached a level similar to that of the ERBB2 amplification, even in ER-positive cases. These results sharply contrast with those reported by Holst et al., although it has to be mentioned that cribriform and mucinous carcinomas, which showed the highest frequency of ESR1 amplification in the series studied by Holst et al., were not included in our analysis.

Note: Supplementary information is available on the Nature Genetics website.