Human transcriptome array 2

200–400 ng total RNA aliquots of MIR618 cells (clones MIR618WT1-3, MIR618SNP1-3) and MIR204 cells (clones MIR204WT4-5, MIR204SNP4-6) were submitted for analysis with the Human Transcriptome Array 2.0 (Affymetrix, CA, USA) and data normalization using RMA to AROS Applied Biotechnology (Denmark). Prefiltering was performed on the normalized data to exclude transcript clusters lacking a gene symbol or RNA accession number. Differential expression of genes between wild type and mutant cell lines was assessed by SAM v1.0 as implemented in MeV software v4.9.0 [52 (link),53 (link)]. A two class unpaired analysis was used (group 1: WT clones, group 2: SNP clones). GSEA was performed using GSEA v2.2.0 software [54 (link),55 (link)] using MSigDB v5.0 C5 Gene Ontology molecular function gene sets (c5.mf.v5.0.symbols.gmt) [56 (link)]. For all comparisons (group 1: WT clones, group 2: SNP clones), data was collapsed to gene symbols, with max_probe to collapse multiple probes. Gene sets of > 500 and < 15 genes were excluded. 1000 permutations based on gene sets were performed. Gene sets were called significant when FDR < 0.01 and nominal p-value < 0.001.

For gene expression profiling, total RNA was extracted and hybridized to Affymetrix Human Transcriptome Array 2.0. Background correction and normalization were done using RMA (Robust Multichip Average) algorithm [45 (link)] using Affymetrix Power Tools. After quality assessment, a filtering process was performed to eliminate low-expression probe sets. Applying the criterion of an expression value > 16 in two samples for each experimental condition, 41,697 probe sets were selected for statistical analysis. R and Bioconductor were used for preprocessing and statistical analysis. LIMMA (Linear Models for Microarray Data) [46 (link)] was used to find out the probe sets that showed significant differential expression between experimental conditions. Genes were selected as significant using a P value > 0.01. The biological knowledge extraction was complemented through the use of Ingenuity Pathway Analysis (QIAGEN Inc., https://www.qiagenbioinformatics.com/products/ingenuity-pathway-analysis).

Three gene expression datasets of CCA and normal tissues were retrieved from the GEO database (https://www.ncbi.nlm.nih.gov/geo/) including GSE26566, GSE32225 and GSE76297. GSE26566 based on GPL6104 Illumina Humanref-8 v2.0 expression bead chip platforms, GSE32225 based on GPL8432 Illumina Humanref-8 WG-DASL v3.0 platforms and GSE76297 based on GPL17586 [HTA-2_0] Affymetrix Human Transcriptome Array 2.0 [transcript (gene) version].

The Human Transcriptome Array 2.0 (HTA 2.0; manufactured by Affymetrix Inc., Santa Clara, CA, USA) was employed in this study. HTA 2.0 covers global profiling of full-length transcripts, containing more than 40,000 non-coding and 245,000 coding transcripts in human genome; each transcript is accurately identified by specific exon or exon-exon splice junction probes. All of the transcripts were collected from multiple public sources such as NCBI RefSeq, Ensembl, UCSC (known genes and lincRNA transcripts), Vertebrate Genome Annotation (Vega) database, Mammalian Gene Collection (MGC) (v10), www.noncode.org, lncRNA db, Broad Institute, Human Body Map lincRNAs and TUCP catalog. In the cases of small samples, random variance model (RVM) t-test was applied to filter the aberrantly expressed transcripts for experimental and control group. After significant analysis and false discovery rate (FDR) analysis, the differentially expressed transcripts were selected according to the P < 0.05.[10 (link)–12 (link)]. The cDNA labeling, microarray hybridization and bioinformatics analysis were performed by Genminix Informatics, Shanghai, China.

To analyze gene enrichment in parous women who breastfed < 6 months vs. ≥ 6 months, we utilized gene expression data obtained from women who were undergoing reduction mammoplasty and were enrolled in the tissue collection study (OSU-2011C0094). Detailed reproductive and other demographic data were also collected at the time of enrollment (Additional file 1: Table S2). Because of the limited number of samples available, we considered the total number of month breastfed for all pregnancies (cumulative no. of months) and did not match for number of pregnancies. Briefly, reduction mammoplasty samples were collected from healthy women and snap frozen in liquid nitrogen. Total RNA isolated from the flash frozen breast tissue was then subjected to gene expression analysis using Human Transcriptome Array 2.0 (Affymetrix Inc, Santa Clara, CA). The affymetrix gene expression data was deposited in NCBI GEO database (GSE102088) [24 (link)]. Gene set enrichment analysis (GSEA; http://software.broadinstitute.org/gsea/msigdb/index.jsp) was performed using this published data set [24 (link)], querying the C2 curated, hallmark 34 gene sets, and Lim_Luminal_Mammary_Progenitor gene sets (Additional file 1: Table S3) [21 (link)] within the Molecular Signatures Database (MSigDB).

We downloaded the microarray datasets GSE95504 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE95504) and GSE85511 (https://www.ncbi.nlm.nih.gov/geo/query/acc.cgi?acc=GSE85511) from GEO. GSE95504 data was based on GPL17586 ([HTA-2_0] Affymetrix Human Transcriptome Array 2.0 [transcript (gene) version]). GSE85511 data was based on GPL21572 ([miRNA-4] Affymetrix Multispecies miRNA-4 Array [ProbeSet ID version]). These two datasets both contain three groups treated with DMSO, 40 μM baicalein, and 80 μM baicalein, respectively. Each group contains three replicates.