Table S1: Additional Tissue Cohorts for Validation

Table S1: Additional tissue cohorts for validation. Samples Sample Format Number IDC-DCIS Cohort (Table 1) 17 FFPE Tissue Samples used for WG-DASL and qRT-PCR Normal (RM-NS)a 3 Total 20 RNA IDC-DCIS Cohort (Table 1) 16 FFPE Tissue Samples used for qRT-PCR Normal (RM-NS) 20 Total 36 RNA TMA 1: IDC NST with/without DCIS 45 - Grade 1 6 - Grade 2 19 - Grade 3 20 TMA 1: DCIS 11 FFPE Tissue Samples used for SOX10 IHCb TMA 2: ILC 49 - Grade 1 2 - Grade 2 41 - Grade 3 6 Total 105 TMA

- Normal Breast only 4 Fresh-Frozen Tissue Samples used for SFRP1 IHC - DCIS with (2) /without (3) IDC 5 - Invasive Carcinoma 14 Total 23 Whole section Normal (RM) 5 IDC and DCIS (+/- normal) 5 Fresh-Frozen Tissue Samples used for COL11A1 IF - IDC NST 5 - DCIS with/without IDC 3 Total 10 Whole section a Reduction Mammoplasties were carefully assessed to exclude samples showing inflammation and fibrocystic change. b Tissue Microarrays (TMAs) were constructed using a manual Tissue Arrayer (MTA-1, Beecher Instruments and Estigen Tissue Science) and 0.6 mm needles (Beecher Instruments, Cat No MP06). Scoring was performed on duplicate tissue cores and the final score was obtained from the average of the replicates or alternatively, from a single core when the replicate was not available. Any positive staining localised to the nucleus of the tumour epithelial cells of IDC and DCIS was regarded as positive expression for SOX10. Table S2: Immunohistochemistry data for SFRP1 in normal epithelial cells, DCIS and invasive carcinomas.

Case ID Normal TDLU DCIS Invasive Cancer % Intensity Grade % Intensity Type Grade % Intensity Q4 50 2-3+ Q19 20 2+ Q78 100 3+ Q161 100 2-3+ Q16 60 2+ High 30% 1+ Q52 High 10% 1+ Q106 100 2-3+ Intermediate 40% 1+ Q272 Intermediate 20% 1+ IDC NST 2 20 1+ Q25 30 2-3+ Intermediate 40% 1+ Tubular 1 30% 1+ Q27 100 3+ IDC NST 3 10% 2+ Q46 100 3+ IDC NST 3 50% 1+ Q48 Metaplastic 3 60% 1 / 2+ Q83 60 2+ IDC NST 3 0 Q84 100 3+ ILC 2 40 1+ Q90 50 3+ IDC NST 3 20 1+ Q93 100 3+ IDC NST 2 40 1+ Q118 100 3+ IDC NST 3 10 1+ Q137 IDC NST 2 0 Q154 100 2-3+ IDC NST 2 30 1+ Q248 Mixed Ductolobular 2 20 1+ Q267 Mixed Ductolobular 3 50 2+ %, percentage of cells staining; IDC NST, invasive ductal carcinoma no special type; ILC, invasive lobular carcinoma.

Cells featuring a strikethrough represent cases for which that specific sample type was not available. Statistical analysis (performed in GraphPad Prism 5.0c), comparing median percentage expression between paired compartments and using Fisher’s Exact Test: TDLU vs. DCIS = P<0.001; TDLU vs. invasive carcinoma = P<0.001; DCIS vs. invasive carcinoma = not significant; invasive carcinomas: IDC NST vs. special types (tubular, ILC, mixed): P= 0.003. Table S3: List of GO terms attributed to the genes differentially expressed between DCIS and IDC (n = 58 genes; this study) and between DCIS and IDC (n=546; ), using a significance value cut-off of P<0.05.

DCIS vs. IDC (n = 58; this study) DCIS vs. IDC (n=546; Schuetz et al [1]) GO Term P value GO Term P value Phosphate transport 0.0002 Proteinaceous extracellular matrix 2.10E-28 Tissue development 0.0029 Extracellular matrix 3.77E-28 Inorganic anion transport 0.0030 Extracellular region 6.28E-25 Organ development 0.0031 Extracellular region part 1.45E-22 Multicellular organismal process 0.0037 Extracellular matrix part 1.61E-21 System development 0.0037 Extracellular matrix structural constituent 3.38E-15 Ectoderm development 0.0037 Multicellular organismal process 1.63E-11 Collagen 0.0040 Response to stimulus 1.35E-10 Anion transport 0.0040 Immune system process 1.35E-10 Anatomical structure development 0.0048 Immune response 9.50E-10 Proteinaceous extracellular matrix 0.0050 Cell adhesion 1.22E-09 Extracellular matrix 0.0056 Biological adhesion 1.22E-09 Epidermis development 0.0237 Phosphate transport 1.22E-09 Multicellular organismal development 0.0267 Structural molecule activity 2.07E-06 Extracellular region part 0.0482 Response to wounding 6.53E-06 Inorganic anion transport 9.49E-06 Response to external stimulus 2.05E-05 Anion transport 2.92E-05 Multicellular organismal development 1.46E-04 Developmental process 0.0025 Plasma membrane 0.003 Intrinsic to plasma membrane 0.0038 Extracellular space 0.0041 Response to stress 0.0041 Integral to plasma membrane 0.0058 Calcium ion binding 0.0058 Plasma membrane part 0.0205 Cellular component 0.027 Online IHC and IF Methods

IHC for SFRP1 (1:200; 1 hour incubation at room temperature; clone ab4193; Abnova, Taipei City, Taiwan) and SOX10 (1:200; 1 hour at room temperature; clone N-20; Santa Cruz Biotechnology Inc., Santa Cruz, CA, USA) was performed on FF and FFPE sections, respectively. Tissue sections (4 µm) from fresh-frozen samples were cut and kept at -80C until required. Antigen retrieval using citrate buffer was only required for SOX10. The MACH 1 Universal HRP-Polymer kit (Biocare Medical, Concord, CA, USA) was used for detection; DAB chromogenic staining was prepared by mixing 30 l of DAB Chromogen in 1 ml of DAB Substrate Buffer and sections incubated for 10 min. Sections were then rinsed with deionized water and counterstained with Haematoxylin. Normal breast tissue was used as positive control for both antibodies. Positive expression for SOX10 is observed in the nuclei of the myoepithelial cells [2], whereas immunostaining for SFRP1 is found in the cytoplasm of the normal epithelial cells [3,4]. Negative controls, where the primary antibody was omitted and only the secondary antibodies were included, were performed for each run.

Dual IF for COL11A1 (1:50, overnight incubation at 4°C; clone H-179, Santa Cruz Biotechnology Inc) and CK8/18 (1:100, 1 hour incubation; clone 5D-3, Novocastra, Leica Microsystems Pty Ltd, North Ryde, NSW, Australia) was performed on methanol fixed frozen tissue sections. Secondary antibodies (goat anti rabbit AlexaFluor488 IgG (1:300 dilution) and goat anti-mouse IgG1-γ1 AlexaFluor594 (1:400); Invitrogen, Mulgrave, Vic, Australia) were incubated for 30 minutes and DAPI was used for nuclear counterstaining. Slides were imaged using a confocal microscope (Carl Zeiss Pty Ltd, North Ryde, NSW, Australia) and analysed using Volocity 5.1.1 (PerkinElmer, Glen Waverley, VIC, Australia). Normal breast tissue was used as the positive control. Negative controls, where the primary antibody was omitted and only the secondary antibodies were included, were performed for each run.

Staining intensity was compared between epithelial and adjacent stromal compartments using two methods. A pathologist (ACV) manually separated the two compartments into i) Regions of Interest (ROI) or ii) defined 50 individual points within each compartment. The software calculated the signal intensity of each ‘ROI’ or ‘point’ and a mean ROI or point calculated per compartment within each sample. Mann-Whitney U test was used for statistical analysis. Figure S1: Dual Immunofluorescence (IF) for COL11A1 and CK8/18 Regions of Interest (ROI) separating the epithelial (A) and stromal (B) breast compartments; point by point imaging from epithelium (C) and stroma (D). Scale is 100 m. Figure S2: ROI vs. Point by Point Analysis of COL11A1 expression in Cancer and RM samples.

Mean Signal Intensity for COL11A1 (Alexa Fluor 488) displayed as ROI (A) and Point by point analysis (C). Mean signal intensity of ROIs for CK8/18 (Alexa Fluor 594, B). Note that the ROI and Points analysis of COL11A1 replicate well ((A) and (C)). Similarly high levels were observed in the normal epithelium (NE) from both breast cancer (BC-NE) and healthy patients (RM-NE; mean signal intensity, point analysis: 74.7 and 63.9, respectively, data not shown). COL11A1 expression was higher in IDC compared to DCIS (mean signal intensity, point analysis; 73.9 vs 51.5, respectively) and expression levels are lower still in all types of stroma (IDC-S, DCIS-S, NB-S; mean signal intensity: 32.4, 21.3 and 28.5 respectively). The ROI analysis of CK8/18 is also an internal control to validate the approach, as expected, CK8/18 is exclusively detected in the epithelial samples. NB-E: Normal breast epithelium from both BC-NE and RM-NE; NB-S: Normal breast epithelium from both BC-NS and RM-NS. Figure S3: Pairwise sample correlations of five technical replicates from WG-DASL.

Sample pairs are: 4801489029_C – 4804882009_F; 4804882017_E – 4804882008_G; 4804882011_D – 4801489028_E; 4801489029_A – 4804882018_D; 4804882009_A – 4804882008_F. Note correlations are between 0.89-0.97. Comparisons performed post normalisation. Figure S4: Supervised K-means clustering of epithelial samples (A) k=2 and (B) k=3 based K-means clustering analysis of the epithelial sample cohort using the 64 probes (58 genes) that were differentially expressed between DCIS and IDC (IDC grade 3, purple; IDC grade 2, pale blue; DCIS high-grade, yellow; DCIS intermediate (Int) grade, pink). Data were quantile normalised and the clustering was performed in Genespring GX 12.0 using Euclidean distance and 50 iterations. Expression level is as detailed in the colour range legend. Figure S5: Meta-analysis of SFRP1 expression in three publically available gene expression profiling datasets (Jonsson et al., ), NKI ) and The Cancer Genome Atlas (TGCA, http://tcga-data.nci.nih.gov/tcga/tcgaHome2.jsp; level 2 gene expression data (log2 lowess normalized)). Datasets were stratified into molecular subtype [7], and an analysis of variance (ANOVA) was employed to determine any differential gene expression of SFRP1 according to the breast molecular subtypes. A Tukey (multiple comparison) post hoc test was used to detect the difference between each pair of subtypes and showed that SFRP1 was significantly overexpressed in Basal breast cancer (the molecular subtype encompassing most Triple Negative cancers) compared to other molecular subtypes (Lum A, luminal A; Lum B, luminal B and HER2) in all datasets (P ≤5e-05). This analysis was performed in R version 2.10.1 (http://www.r-project.org/). Figure S6: IF co-staining for COL11A1 in stromal fibroblasts and normal epithelium (RM).

Fibroblasts displayed positive staining with COL11A1 (arrowhead) but not with CK8/18 (A), which confirmed the mesenchymal nature of these positive cells. In contrast, normal epithelium (B) co-stained with both COL11A1 (arrow) and CK8/19. A stromal fibroblast (arrowhead) is displayed adjacent to the epithelium. DAPI counterstain in blue; COL11A1 is green (Alexa Fluor 488) and CK8/18 is red (Alexa Fluor 594). Merged image is displayed in the right panel. References

1. Schuetz, C.S., et al., Progression-specific genes identified by expression profiling of matched ductal carcinomas in situ and invasive breast tumors, combining laser capture microdissection and oligonucleotide microarray analysis. Cancer research, 2006. 66(10): p. 5278- 86.

2. Nonaka D, Chiriboga L, Rubin BP (2008) Sox10: a pan-schwannian and melanocytic marker. Am J Surg Pathol 32(9): p. 1291-8.

3. Dahl, et al. Frequent loss of SFRP1 expression in multiple human solid tumours: association with aberrant promoter methylation in renal cell carcinoma. Oncogene, 2007; 26: 5680-5691.

4. Halsted KC, Bowen KB, Bond L, Luman SE, Jorcyk CL, Fyffe WE, Kronz JD, Oxford JT (2008) Collagen alpha1(XI) in normal and malignant breast tissue. Mod Pathol 21(10): p. 1246-54 5. Jonsson, G., et al., Genomic subtypes of breast cancer identified by array-comparative genomic hybridization display distinct molecular and clinical characteristics. Breast cancer research : BCR, 2010. 12(3): p. R42. 6. van de Vijver M.J., et al., A gene-expression signature as a predictor of survival in breast cancer. The New England Journal of Medicine, 2002. 347(25): p. 1999-2009. 7. Hu, Z., et al., The molecular portraits of breast tumors are conserved across microarray platforms. BMC Genomics, 2006. 7: p. 96.