Molecular diversity in ductal carcinoma in situ (DCIS) and early invasive breast cancer

Abstract

Ductal carcinoma in situ (DCIS) is a non-invasive form of breast cancer where cells restricted to the ducts exhibit an atypical phenotype. Some DCIS lesions are believed to rapidly transit to invasive ductal carcinomas (IDCs), while others remain unchanged. Existing classification systems for DCIS fail to identify those lesions that transit to IDC. We studied gene expression patterns of 31 pure DCIS, 36 pure invasive cancers and 42 cases of mixed diagnosis (invasive cancer with an in situ component) using Agilent Whole Human Genome Oligo Microarrays 44k. Six normal breast tissue samples were also included as controls. qRT-PCR was used for validation. All DCIS and invasive samples could be classified into the "intrinsic" molecular subtypes defined for invasive breast cancer. Hierarchical clustering establishes that samples group by intrinsic subtype, and not by diagnosis. We observed heterogeneity in the transcriptomes among DCIS of high histological grade and identified a distinct subgroup containing seven of the 31 DCIS samples with gene expression characteristics more similar to advanced tumours. A set of genes independent of grade, ER-status and HER2-status was identified by logistic regression that univariately classified a sample as belonging to this distinct DCIS subgroup. qRT-PCR of single markers clearly separated this DCIS subgroup from the other DCIS, and contains samples from several histopathological and intrinsic molecular subtypes. The genes that differentiate between these two types of DCIS suggest several processes related to the re-organisation of the microenvironment. This raises interesting possibilities for identification of DCIS lesions both with and without invasive characteristics, which potentially could be used in clinical assessment of a woman's risk of progression, and lead to improved management that would avoid the current over- and under-treatment of patients.

Figure 1

Hierarchical clustering of pure DCIS. A: Clustering of 605 genes showing variation of expression of at least 2 fold across all pure DCIS samples (see Supplemental Figure S2 for heatmap and Supplemental Table S1 for gene list). B: Clustering of the top 100 ER‐, HER2‐ and grade‐independent genes identified by logistic regression (see Supplemental Figure S3 for heatmap and Supplemental Table S3 for all 492 genes identified by logistic regression). DCIS Type indicates the DCIS subgroup: type I = black, type II = grey. Colour coding panel: ER: ER‐positive = grey, ER‐negative = black. PR: PR‐positive = grey, PR‐negative = black. HER2 IHC: 1+ = grey, 2+ = dark grey, 3+ = black. HER2 CGH: amplified = black, non‐amplified = grey, missing data = white. HER2 mRNA: gray‐scale according to expression value; black indicates high gene expression value. Ki‐67: Ki‐67‐positive = black, Ki‐67‐negative = grey. EORTC: grade A = grey, grade B = dark grey, grade C = black. PAM50 represents the different “intrinsic” molecular subtypes previously identified for invasive breast cancer. Dark blue = luminal A, light blue = luminal B, purple = ERBB2+, red = basal‐like and green = normal‐like. ER, PR and Ki‐67 status determined by IHC.

Figure 2

Validation of the DCIS type I subgroup in an independent dataset. Hierarchical clustering of the 39 DCIS samples in the validation dataset (Hannemann et al., 2006) by 44 of the top 100 genes independent of ER, HER2 and grade identified by logistic regression (only 44 genes were matched on the cDNA platform used in (Hannemann et al., 2006)). Colour coding panel: DCIS grade: black = poorly‐differentiated, grey = intermediate differentiated, light grey = well‐differentiated. HER2 and ER: grey = positive by IHC and black = negative by IHC.

Figure 3

Validation of microarray gene expression by real‐time PCR. A–D: Relative expression levels of genes identified by SAM; FOXA1, TRAF3IP3, HCST and CYP1B1, measured by TaqMan qRT‐PCR in 22 type II and 6 type I DCIS cases. E–G: Relative expression level of EMT‐related genes identified by SAM; SNAI1, S100A8 and CXCL1, measured by TaqMan qRT‐PCR in 22 DCIS type II and 6 DCIS type I cases. Black horizontal bars represent median value for each DCIS phenotype.