Total RNA was isolated using RNeasy Mini kit (Qiagen) following the manufacturer's protocol. DNase I in column digestion was included to ensure RNA quality. Two biological repeats were performed. For circRNA-seq, total RNA after DNase I treatment was digested with RNase R (Epicentre) at 37 °C for 1 h before RNA library preparation. RNA library preparation for both regular RNA-seq and circRNA-seq was performed by using NEBNext® Ultra™ Directional RNA Library Prep Kit for Illumina (E7420L). Paired-end sequencing was performed with Illumina HiSeq 3000 at RiboBio Co., Ltd. or Amogene Biotech Co., Ltd.

For regular RNA-seq, sequencing reads were aligned to hg19 reference genome by using TopHat ( 79. To determine circPGR-regulated gene program, Cuffdiff 80 was used to first quantify the expression of RefSeq annotated genes with the option -M (reads aligned to repetitive regions were masked) and -u (multiple aligned read are corrected using 'rescue method'). Coding genes with FPKM (fragments per kilobase per million mapped reads) larger than or equal to 0.5 in any one of the experimental conditions were included in our analysis. CircPGR-regulated gene program was determined by fold change (FC) of gene FPKM in si-CTL and si-circPGR-transfected samples (FC ≥ 1.5). FPKM of a gene was calculated as mapped reads on exons divided by exonic length and the total number of mapped reads.

For circRNA-seq analysis, sequencing reads were mapped to hg19 reference genome by using BWA MEM (-T 19), and circRNAs were predicted by using 81. To predict full length circRNAs, CIRI-full was used at default settings 81, 82. To visualize circRNAs, CIRI-vis was used at default settings 83.

Box plots were generated by R software and significance was determined using Student's t-test. Heat maps were visualized using Java TreeView or R software.

Gene ontology analysis (GO) was performed by using GSEA 71 or DAVID 84, and Kaplan Meier survival analysis was performed by using GOBO 85. Specifically, genes positively-regulated by circPGR or cell cycle genes positively-regulated by circPGR were submitted to GOBO, and “ER-positive” or “ER-negative” were selected for “Tumor selection”, “2 groups”, “10 years”, “overall survival” and “ER-status” were then selected. Results were exported in PDF format.