Integrative analyses that summarize and link molecular data to treatment sensitivity are crucial to capture the biological complexity which is essential to further precision medicine. We introduce Weighted Orthogonal Nonnegative parallel factor analysis (WON-PARAFAC), a data integration method that identifies sparse and interpretable factors. WON-PARAFAC summarizes the GDSC1000 cell line compendium in 130 factors. We interpret the factors based on their association with recurrent molecular alterations, pathway enrichment, cancer type, and drug-response. Crucially, the cell line derived factors capture the majority of the relevant biological variation in Patient-Derived Xenograft (PDX) models, strongly suggesting our factors capture invariant and generalizable aspects of cancer biology. Furthermore, drug response in cell lines is better and more consistently translated to PDXs using factor-based predictors as compared to raw feature-based predictors. WON-PARAFAC efficiently summarizes and integrates multiway high-dimensional genomic data and enhances translatability of drug response prediction from cell lines to patient-derived xenografts.

Chromatin immunoprecipitation (ChIP)-seq analyses of transcription factors in clinical specimens are challenging due to the technical limitations and low quantities of starting material, often resulting in low enrichments and poor signal-to-noise ratio. Here, we present an optimized protocol for transcription factor ChIP-seq analyses in human tissue, yielding an ~100% success rate for all transcription factors analyzed. As proof of concept and to illustrate general applicability of the approach, human tissue from the breast, prostate, and endometrial cancers were analyzed. In addition to standard formaldehyde fixation, disuccinimidyl glutarate was included in the procedure, greatly increasing data quality. To illustrate the sensitivity of the optimized protocol, we provide high-quality ChIP-seq data for three independent factors (AR, FOXA1, and H3K27ac) from a single core needle prostate cancer biopsy specimen. In summary, double-cross-linking strongly improved transcription factor ChIP-seq quality on human tumor samples, further facilitating and enhancing translational research on limited amounts of tissue.

The Androgen Receptor (AR) is the key-driving transcription factor in prostate cancer, tightly controlled by epigenetic regulation. To date, most epigenetic profiling has been performed in cell lines or limited tissue samples. Here, to comprehensively study the epigenetic landscape, we perform RNA-seq with ChIP-seq for AR and histone modification marks (H3K27ac, H3K4me3, H3K27me3) in 100 primary prostate carcinomas. Integrative molecular subtyping of the five data streams revealed three major subtypes of which two were clearly TMPRSS2-ERG dictated. Importantly, we identify a third subtype with low chromatin binding and activity of AR, but with high activity of FGF and WNT signaling. While positive for neuroendocrine-hallmark genes, these tumors were copy number-neutral with low mutational burden, significantly depleted for genes characteristic of poor-outcome associated luminal B-subtype. We present a unique resource on transcriptional and epigenetic control in prostate cancer, revealing tight control of gene regulation differentially dictated by AR over three subtypes.

Male breast cancer (MBC) is rare and largely hormonally driven. Here, the authors examine the action of steroid hormone receptors in male and female breast cancers and find gender selective hormone receptor action that associates with the survival of MBC patients.

The DNA-binding sites of estrogen receptor α (ERα) show great plasticity under the control of hormones and endocrine therapy. Tamoxifen is a widely applied therapy in breast cancer that affects ERα interactions with coregulators and shifts the DNA-binding signature of ERα upon prolonged exposure in breast cancer. Although tamoxifen inhibits the progression of breast cancer, it increases the risk of endometrial cancer in postmenopausal women. We therefore asked whether the DNA-binding signature of ERα differs between endometrial tumors that arise in the presence or absence of tamoxifen, indicating divergent enhancer activity for tumors that develop in different endocrine milieus. Using ChIP sequencing (ChIP-seq), we compared the ERα profiles of 10 endometrial tumors from tamoxifen users with those of six endometrial tumors from nonusers and integrated these results with the transcriptomic data of 47 endometrial tumors from tamoxifen users and 64 endometrial tumors from nonusers. The ERα-binding sites in tamoxifen-associated endometrial tumors differed from those in the tumors from nonusers and had distinct underlying DNA sequences and divergent enhancer activity as marked by histone 3 containing the acetylated lysine 27 (H3K27ac). Because tamoxifen acts as an agonist in the postmenopausal endometrium, similar to estrogen in the breast, we compared ERα sites in tamoxifen-associated endometrial cancers with publicly available ERα ChIP-seq data in breast tumors and found a striking resemblance in the ERα patterns of the two tissue types. Our study highlights the divergence between endometrial tumors that arise in different hormonal conditions and shows that ERα enhancer use in human cancer differs in the presence of nonphysiological endocrine stimuli.