Immunohistochemistry was performed as described previously [
10]. In summary, 5 μm sections were taken, and for each case matched biopsy and resection samples were placed on the same single slide (SuperFrost Plus; Menzel-Glaser, Braunschweig, Germany) to allow direct comparisons of staining between biopsy and resection. Slides were air-dried, and samples dewaxed with xylene and submerged in ethanol. Epitopes were retrieved by heat (900 W microwave, 10 min) in 10 mM citrate buffer (pH 6.0) and endogenous peroxidase activity was blocked using 0.3 % H
2O
2 (10 min). For anti-activated Notch1, non-specific primary antibody binding was blocked using casein (SP5020; Vector, Burlingame, USA) diluted 10-fold in antibody diluent reagent (Invitrogen, Carlsbad, USA) for 20 min. Slides were stained with the following primary antibodies in antibody diluent reagent for 1 h at room temperature; MRP1 (1:50, QCRL1 clone; sc18835, Santa Cruz Biotech, Santa Cruz, USA); activated Notch1 (Notch1
IC: 1:100; ab8925, Abcam, Cambridge, UK). Both antibodies have been used previously for immunohistochemistry on breast tissue [
25,
10], and for western blots with breast cell lines [
26,
27]. Immuno-staining was visualised using Envision reagents (Dako, Gostrup, Denmark). Sections were counter-stained with Mayer’s haematoxylin and mounted in DPX (Fluka, Gillingham, UK). MRP1 expression was quantified using computer-aided scoring as a histoscore from 0 (no staining) to 300 (strong staining throughout epithelial areas) as described previously [
10]. Notch1
IC nuclear staining was scored manually by two independent scorers (BK, BJW) using the Allred method [
28,
29], giving scores from 0 (no staining) to 8 (strong staining in >66 % of tumour cells). Weighted Kappa coefficients (
k) for the two independent scores were 0.78 for biopsies (good agreement), and 0.9 for resections (near perfect agreement); overall
k = 0.89 (
n = 58). Averages of the two scores were used as final expression scores.