We read the correspondence regarding the recent Review (The different roles of ER subtypes in cancer biology and therapy. Nature Rev. Cancer 11, 597–608 (2011))1 between Ke-Da Yu and Zhi-Ming Shao and the authors of the Review with great interest1,2,3. The ongoing debate concerning the prevalence and the clinical relevance of oestrogen receptor 1 (ESR1) amplification reflects the highly visible evidence for this genetic aberration despite the challenging data.

There are now three studies from three different groups, which all confirm ESR1 amplification in 20.6%, 22.6% and 21.6% of analysed breast cancers by fluorescence in situ hybridization (FISH)4,5,6, including one study by Tsiambas et al.6 that is only rarely cited. Likewise, it has hardly been recognized that Moelans et al.7 detected 16% and 20%7,8ESR1 (mostly low level) gains and amplifications in two independent multiplex ligation-dependent probe amplification (MLPA) studies, which is substantially higher than the 8%9 cited by Thomas and Gustafsson3 and which may be related to the different patient sets in their three publications. Using a cut-off of 2 (for the grade of copy number imbalance) instead of >2, Nessling et al.10 would have found around 19% instead of 10% amplification by array comparative genomic hybridization (aCGH)10, illustrating that minor changes in the threshold criteria may have major effects on the outcome of ESR1 studies. There are five studies by four different groups using DNA-specific methods that are independent of observer subjectivity, such as Southern blot, aCGH or MLPA PCR, which have found rates of 10% and greater ESR1 amplification7,8,10,11 or increased copy number7,8,12 in breast cancer. Including the 20% increased ESR1 copy numbers in oestrogen receptor (ER)-positive tumours reported by Dunbier et al.12, these studies typically reported low-level copy number gains. It is easily imaginable that there could be copy number gains both just above and just below the detection level and within the background of measurement.

When non-morphological methods are used, contamination with non-tumour cells (which could be resolved by laser microdissection) and intermingled mosaic-patterned intratumoural heterogeneity of copy number status — as also shown in case of ERBB2 (Refs 13, 14) — could obscure the presence of low-level amplification, although this may be readily picked up by fluorescence in situ hybridization (FISH).

Evidence of correlation between elevated ESR1 copy numbers and high ER expression was found in five studies from five independent groups4,5,6,9,12, raising the question of whether ESR1 copy numbers could be a better clinical marker for response to hormone therapy than protein expression. In their recent ESR1 study, Ejlertsen et al.15 did not reproduce previously claimed therapeutic resistance of tumours with amplified ESR1, and there was no difference in distribution in the percentage of ER-expressing tumour cells between tumours with amplified ESR1 (13.6% ER+) and those of normal copy number (82.2% ER+). However, they found significantly lower ER expression in ESR1-deleted tumours15,16.

In summary, there is growing evidence that ESR1 gain or amplification is a fairly frequent event in breast cancer, although it is difficult to detect with current analysis and threshold techniques, which were initially established for ERBB2 measurement. The clinical relevance of ESR1 gains and amplifications in the sense of prognostic value and prediction of response to hormonal treatment requires further studies, similar to earlier debates for ERBB2 amplification.