Fig 2: MMP-8 expression alters myoepithelial cell adhesion and migration to ECM proteins. a Upper panel: immunoprecipitation of matrix metalloproteinase-8 (MMP-8) from 20× concentrated conditioned media of ß6-1089 transiently transfected with wild-type MMP-8 (WT), inactive point mutant of MMP-8 (EA) or empty vector (EMPTY). Conditioned media (CM) was incubated overnight with MMP-8 antibody (R& D Systems, mAb 908) or IgG (Sigma-Aldrich) and bound proteins were precipitated with Protein A beads (B) and unbound supernatant was also collected (UB). Samples were separated by 10% SDS-PAGE and probed with anti-V5 antibody directed to the V5 tag on the transfected MMP-8. Lower panel: Collagen-I zymography. CM was collected from the ß6-1089 cells as described above and concentrated 20× using 3 K spin columns. Equal volumes were run by SDS-PAGE containing 3 mg of Collagen-I. The gel was incubated overnight in substrate solution and then stained with Coomassie. Only the lane containing CM from ß6-1089 transfected with WT MMP-8 demonstrated a clear band indicating Collagen degradation at the same molecular weight as MMP-8. b (i) Adhesion to ECM proteins. ß6-1089 cells transiently transfected with empty vector (EMPTY), wild-type MMP-8 (WT) or inactive mutant MMP-8 (EA) were plated on Fibronectin (Fib), Collagen-I (Col-I), Collagen-IV (Col-IV), Laminin-I (Lam-I), Tenascin-C (Ten-C), latency-associated peptide (LAP) or BSA (Control) for 1 hour. Then cells were treated with Calcein AM for 15 minutes before reading the fluorescence. The adhesion is calculated relative to the EMPTY. Expression of WT MMP-8 increased cell adhesion to Fib, Col-I, Lam-I and Ten-C, and decreased adhesion to LAP. The EA transfected cells did not show alterations in adhesion.(ii) ß6-1089 cells treated in the same way as above were plated onto transwells coated on the underside with the same ECM proteins and cells were allowed to migrate for 8 hours before quantifying the number of cells that were on the top and bottom of the transwell. Migration was calculated relative to the EMPTY. Expression of WT MMP-8 decreased migration towards Fib, Col-I, Lam-I, Ten-C and LAP, while expression of the EA had no effect. c (i) N-1089 cells were transfected with siRNA targeting MMP-8 (siMMP-8) or Luciferase (siLuc) as a control. Ninety-six hours after transfection protein and RNA were collected and analysed by western blot and RT Q-PCR respectively. Both the western blot (upper panel) and RT Q-PCR (graph) indicate a reduction in MMP-8 at the protein and mRNA level in response to siMMP-8 treatment as compared to the siLUC control. (ii) N-1089 cells treated with siMMP-8 as described were plated onto the same selection of ECM proteins as previously described in (b). A significant reduction in adhesion to Fib, Col-I, Col-IV, Lam-I and Ten-C was observed in N-1089 transfected with siMMP-8 as compared to N-1089 transfected with siLUC. (iii) N-1089 cells treated with siMMP-8 were also plated on transwells coated with the same selection of ECM proteins as described in (b). The N-1089 transfected with siMMP-8 demonstrated a significant increase in migration towards Col-IV, Lam-I and Ten-C as compared to N-1089 + siLUC. ** p = 0.01, *** p = 0.001 (Student’s t test). Error bars = SEM, n = 3 independent experiments

Fig 3: MMP-8 expression alters a6ß4 subcellular distribution and myoepithelial cell morphology. a (i) Confocal images of ß6-1089 cells transfected with empty vector (EMPTY) wild-type matrix metalloproteinase-8 (MMP-8 WT) or inactive mutant MMP-8 (EA) and grown on Fibronectin for 24 hours then fixed and co-stained for a6ß4 (red) and plectin (green). Insert: cartoon illustrating hemidesmosome (HD) components, indicating the integrin a6ß4 associates with plectin to link the complex to the keratin cytoskeleton. (ii) Co-localisation of a6ß4 and plectin was determined by quantifying the number of yellow pixels per field of view in each condition then expressed as a ratio as compared to EMPTY. ß6-1089 transfected with WT demonstrated a significant increase in a6ß4 and plectin co-localisation, while the ß6-1089 transfected with EA exhibited no change as compared to EMPTY. b (i) Confocal images of ß6-1089 cells transfected with empty vector (EMPTY) wild-type MMP-8 (WT) or inactive mutant MMP-8 (EA) and grown on Fibronectin for 24 hours then fixed and co-stained for a6ß4 (green) and phallodin (red). (ii) Morphological features of the transfected ß6-1089 were analysed and length of trailing fibres was noted to be altered. Analysis of the length of these fibres indicated that ß6-1089 transfected with MMP-8 WT significantly reduced the fibre length as compared to ß6-1089 transfected with empty vector or MMP-8 EA. (iii) Analysis of the number of these fibres per cell also demonstrated a significant reduction in ß6-1089 transfected with MMP-8 WT as compared to ß6-1089 transfected with empty vector or MMP-8 EA. (iv) The final morphological alteration analysed was the physical space the cells occupied, the relative area occupied by ß6-1089 transfected with MMP-8 WT was significantly larger than that occupied by ß6-1089 transfected with empty vector or MMP-8 EA. c (i) Confocal images of N-1089 cells transfected with siRNA targeting MMP-8 (siMMP-8) or Luciferase (siLuc) as a control grown on Fibronectin for 24 hours then fixed and co-stained for a6ß4 (green) and phallodin (red). (ii) Trailing fibre length was significantly longer in N-1089 transfected with siMMP8 as compared to siLUC. (iii) There were significantly more trailing fibres per cell in the N-1089 + siMMP8 group as compared to N-1089 transfected with siLUC. * p = 0.05 ** p = 0.01, *** p = 0.001 (Student’s t test). Error bars = SEM, n = 3 independent experiments

Fig 4: MMP-8 expression in; primary normal and DCIS myoepithelial cells and a cell line model. a Normal and ductal carcinoma in situ (DCIS) tissue immunohistochemically stained for matrix metalloproteinase-8 (MMP-8) (Atlas, HPA02122, 1:1000). Clear MMP-8 staining can be observed in the normal myoepithelial cells (MECs), which is absent in the DCIS section. Magnification × 400. b Normal MECs and luminal epithelial cells (LECs) were isolated from reduction mammoplasty tissue and DCIS MEC and DCIS LEC were isolated from a patient with high-grade DCIS. Tissues were enzymatically digested and magnetic bead separation was used to isolated specific populations (MEC - ITGB4, LEC - EpCAM). RNA was extracted from the isolated cells and G361 cells (a positive control for MMP-8), converted to cDNA by reverse transcription and then subjected to nested PCR for MMP-8 or PCR of housekeeping gene 18S. The MMP-8 PCR produces a product of 150 bp in normal MEC and G361 lanes, and other lanes only give a weak band. 18S gives a size of 52 bp in all sample lanes. The water (H2O) control lane shows no band. c (i) RNA was extracted from cell lines; G361, N-1089 and ß6-1089 and converted to cDNA as above. The cDNA was subjected to nested PCR for MMP-8 and bands were identified in G361 and N-1089 but only a weak band was observed in the ß6-1089 lane. All samples demonstrated a band for 18S. (ii) RIPA protein lysates were produced from N-1089 and ß6-1089 cells and 20 ug of protein was run on 8% PAGE then transferred to nitrocellulose and probed for MMP-8 (R&D Systems, mAb 908) avß6 (Santa Cruz Biotechnology, SC-6632) or HSC70 as a loading control. The gel indicates that N-1089 expresses higher levels of MMP-8 and much lower levels of avß6. While ß6-1089 exhibit much higher levels of avß6 and lower levels of MMP-8. (iii) Densitometry of western blots indicates that MMP-8 expression in N-1089 is 2.5-fold higher than N-1089 and (iv) avß6 expression in ß6-1089 is 10-fold higher than N-1089

Fig 5: MMP-8 expression by myoepithelial cells has a paracrine effect on tumour cell invasion. a ß6-1089 cells transfected with empty vector (EMPTY) wild-type matrix metalloproteinase-8 (MMP-8 WT) or inactive mutant MMP-8 (EA) were plated into 24-well plates and grown in serum-free media for 24 hours. Transwells coated with Matrigel were placed into the wells and MDA-MB-231 or SUM159 cells were placed on top of the Matrigel. After 24 hours the transwells were removed and the number of invading cancer cells counted. Invasion was calculated relative to the ß6-1089 transfected with empty vector. MDA-MB-231 and SUM159 cells exhibited a significant reduction in invasion when in the presence of ß6-1089 transfected with WT cells, this was not observed with the ß6-1089 transfected with EA cells. b MTS assay for viability on MDA-MB-231 and SUM159 cells grown for 24 hours in the presence of conditioned media (CM) from ß6-1089 cells transfected with empty vector (EMPTY) wild-type MMP-8 (WT) or inactive mutant MMP-8 (EA). No difference was observed between any of the groups relative to ß6-1089 transfected with empty vector. c Transwells were set up as described above with N-1089 cells transfected with siRNA targeting MMP-8 (siMMP-8) or Luciferase (siLuc) were plated into the bottom wells. MCF-7, MDA-MB-231 and SUM159 cells were plated into Matrigel-coated transwells above. After 24 hours (or 48 hours for MCF-7) the transwells were removed and the number of invading cancer cells counted. Invasion was calculated relative to the N-1089 transfected with siLUC. MDA-MB-231 and SUM159 cells exhibited a significant increase in invasion when in the presence of N-1089 transfected with siMMP-8. d MTS assay for viability on MCF-7, MDA-MB-231 and SUM159 cells grown for 24 hours in the presence of conditioned media (CM) from N-1089 cells transfected with siRNA targeting MMP-8 (siMMP-8) or Luciferase (siLuc). No difference was observed between the two groups. ** p = 0.01, (Student’s t test). Error bars = SEM, n = 3 independent experiments