Pugh et al. reply.

Replying to J.-M. Mosquera et al., Nature 520, http://dx.doi.org/10.1038/nature14265 (2015) In the accompanying Comment1, Mosquera and colleagues analysed MAGI3–AKT3 fusions in 236 formalin-fixed paraffin embedded (FFPE) triple-negative breast cancer (TNBC) specimens using break-apart fluorescence in situ hybridization (FISH) and detected no cases with fusions. In contrast, our previous published report found MAGI3–AKT3 in 8 of 235 breast cancer samples and 5 of 72 TNBC cases by reverse trancriptase polymerase chain reaction (RT–PCR) using gene specific primers2. To address this discrepancy, we analysed MAGI3–AKT3 fusions using a hybrid capture array, ‘ExomePlus’, that covers known exons, conserved non-coding regions and intronic regions involved in gene fusions including the ~150 kilobase first intron of AKT3. FFPE tumour and normal tissue was available from 3 positive TNBC cases from our original screen, including the index case, BR-M-045, and frozen tissue for BR-M-045. We performed ExomePlus hybrid capture and Illumina sequencing, achieving an average median read coverage of 76× (range of medians 29–144) across intron 1 of AKT3 (chr1:243,859,018–244,006,427), on DNA from these samples. We found 4 fusion read-pairs within intron 1 of AKT3 confirming the existence of the MAGI3–AKT3 fusion in genomic DNA obtained from frozen tissue of the index case, BR-M-045 (Fig. 1)3. In contrast, we failed to detect the fusion event in any tumour or normal genomic DNA obtained from FFPE tissue, including the BR-M-045 case. Screening additional DNA from 370 breast tumours, including 280 frozen tumours and 90 FFPE samples, and 372 normals (366 paired samples), also failed to find evidence of the fusion in any of these 370 tumour DNA samples at a threshold of three read pairs. Comparison of the relative allelic fraction of the fusion event to the median allelic fraction of somatic mutations (Fig. 2)4, suggests that the MAGI3–AKT3 fusion event in BR-M-045 may represent a sub-clonal population of tumour cells5. Our initial positive observations might therefore be explained by intra-tumour heterogeneity, as well as by rare contamination with the fusion cDNA—that is, we observed the fusion in 4% of cases and in 0 controls, but we analysed only 12 negative controls. In retrospect, any such study, even by a straightforward method such as PCR, would be better powered by using a number of controls equal to the number of experimental samples. We conclude that, although the MAGI3–AKT3 fusion occurs in at least one breast cancer case, the overall prevalence is lower than our original estimate. Indeed, the data from Mosquera and colleagues and our validation data, suggest a prevalence of <1%. This rare alteration is oncogenically transforming and its activity is sensitive to Akt inhibition, highlighting the potential of rare genome alterations in breast cancer therapy. Furthermore, the AKT3 pathway may be important in breast cancer in light of the overexpression of AKT3 that is observed in basal-like breast cancers6, the expression subtype corresponding to TNBC, and it remains possible that the MAGI3–AKT3 translocation will be observed in other cancer types by genomic studies. This Reply has been written on behalf of the original author list2, T. J. Pugh was not a co-author of the original submitted manuscript but led the analysis of the ‘ExomePlus’ data. Download references