Aromatase inhibitors--gene discovery

Microarray analysis of tumour RNA is an extremely powerful tool which allows global gene expression to be measured. When used in combination with neoadjuvant treatment protocols in which therapy is given with the primary tumour within the breast, sequential biopsies may be analysed and results correlated with clinical and pathological response. In the present study, a neoadjuvant protocol has been used, administering the third generation inhibitor, letrozole, for 3 months and subjecting RNA extracted from biopsies taken before and after 10–14 days of treatment to microarray analysis. The objectives were to discover: (i) genes that change with estrogen deprivation (the only known biological effect of letrozole is to inhibit aromatase activity and reduce endogenous estrogens in postmenopausal women) and (ii) genes whose basal, on treatment or change in expression differ between tumours which are either responsive or resistant to treatment (so that predictive indices of response/resistance may be developed).

Early changes in gene expression were identified by comparing paired tumour core biopsies taken before and after 14 days treatment in 58 patients using three different approaches based on frequency of changes, magnitude of changes and SAM analysis. All three approaches showed a greater number of genes were down-regulated than up-regulated. Merging of the data produced a total of 143 genes which were subject to gene ontology and cluster analysis. The ontology of the 91 down-regulated genes showed that they were functionally associated with cell cycle progression, particularly mitosis. In contrast, up-regulated genes were associated with organ development and extra-cellular matrix turnover and regulation.

Clinical response was assessable in 52 patients; 37 (71%) tumours were classified as clinical responders (>50% reduction in volume at 3 months). Microarray analysis of pre- and 14-day biopsies identified 291 covariates (84 baselines, 72 14-day and 135 changes) highly predictive of response status. A similarity matrix using the covariates showed responding tumours have a similar genetic profile which was dissimilar to non-responding cancers whereas non-responsive cases were distinctive from each other. Changed genes predicting for response showed no concordance with those changed significantly by treatment in the overall group.