Abstract P4-04-07: Heterogeneous gene fusions detected by RNASeq show enrichment of insulin signaling pathway genes in breast cancer

Background: Recent discoveries of recurrent and targetable gene fusions in breast cancer suggest the need to characterize the functional significance of such genomic aberrations within larger cohorts. We quantify fusion transcript expression in patient samples using RNASeq and evaluate their functional significance using biological pathway enrichment analysis.

Methods: We sequenced transcriptomes of core biopsy RNA from 97 breast tumors obtained from brief-exposure preoperative clinical trials BrUOG 211A/211B. HER2- patients were treated with brief exposure to bevacizumab (B) or nab-paclitaxel (nP) followed by treatment with B/nP/carboplatin while HER2+ patients received brief exposure to trastuzumab (T) or nP followed by T/nP/carboplatin. Paired-end sequencing on 55 baseline biopsies and 42 post-exposure biopsies using amplified total RNA yielded 55 million reads on average per sample. We assigned RNASeq-based PAM50 subtypes for each of the samples using standard methodology. Fusion transcript abundance was evaluated using two independent pipelines, TopHat and deFuse, due to their complementary strategies in fusion detection. We eliminated fusions of genes with their respective pseudogenes as likely false positives arising due to alignment artifacts. TopHat fusion calls with total supporting reads ≥10 and deFuse calls with probability of fusion ≥0.7 were considered reliable.

Results: We identified high confidence gene fusions, detected by both TopHat and deFuse, in 30 of the 55 baseline biopsies (54.4%), with 3.3 fusions on average per sample and a maximum of 10. Fusions were predominantly associated with chromosomal aberrations (75%), with putative deletions responsible for 32% of fusions and translocations responsible for 43%. We find a high level of fusion transcript heterogeneity within breast cancers, detecting a total of 80 fusions across the 30 samples with only three fusions recurrent in two samples with high expression in each: MDN1-GAS5 in two basal breast cancers, KRAS-GRIP1 and ITPR2-CCDC91 in two LumB cancers. Several cancer-related genes were found to be fusion partners: AKT3-SMYD3, CREB1-PPP1R1C, FLOT2-TOP2A and FOXC1-ARID1B. Pathway analysis of the fusion genes at baseline revealed enrichment of proteasome (p = 0.000752), tight junction (p = 0.027), insulin signaling (p = 0.0284) and melanogenesis (p = 0.05) pathways after multiple testing correction (FDR≤0.25). We looked for modulation of gene fusions upon brief exposure to therapy in 18 patients and found a majority of the baseline fusion transcripts to be present post-brief exposure in 44% of the patients, irrespective of therapy regimen.

Conclusions: We find that gene fusions in breast cancer are highly heterogeneous but are enriched with cancer-related pathway genes. This is the first study to report a novel gene-lincRNA fusion transcript (MDN1-GAS5). We are currently validating the fusion calls using qRT-PCR. The heterogeneity of detected fusions suggests that multiple mechanisms could underlie the selective advantage of tumor cells expressing fusion transcripts. The brief-exposure preoperative paradigm provides a unique opportunity to evaluate modulation of fusion transcripts that can shed light on their functional importance.

Citation Information: Cancer Res 2013;73(24 Suppl): Abstract nr P4-04-07.

Abstract P4-04-02: RNA-seq reveals functional lncRNAs associated with estrogen-receptor status in breast cancer

Background: Long non-coding RNAs (lncRNAs) are pervasively transcribed in the genome yet their role in human disease is not well understood. LncRNAs can have regulatory effects on coding mRNAs through a number of mechanisms, including repressing their sense-strand protein-coding partners. There is also emerging evidence that estrogen signaling affects the expression of a wide variety of non-coding functional RNA transcripts in addition to protein-coding transcripts. We performed an RNA-seq study to characterize changes in lncRNA expression and evaluate their association to estrogen receptor (ER) status and estrogen signaling.

Methods: We sequenced transcriptomes of core biopsy RNA from 45 breast tumors obtained from neoadjuvant clinical trials BrUOG 211A/211B. RNA was derived from biopsy samples obtained before exposure to run-in monotherapy with either nab-paclitaxel, bevacizumab or trastuzumab. Paired-end sequencing was performed using amplified total RNA with 74bp read length, yielding genome-wide transcriptomic data. Transcriptomic abundance and differential expression were estimated assuming Poisson-distributed read-counts. Paired-end sequence data was aligned to a lncRNA database containing 14,572 unique lncRNAs. Changes in relative abundance of lncRNA transcripts were tested for association with estrogen receptor status using the Wilcoxon rank-sum test. Expression levels of the ER-associated lncRNAs were investigated in RNA-seq data from ER-positive MCF7 cells in response to treatment with E2 and tamoxifen (3hr, 12 hr and a non-treated control). ER-responsive binding sites on or near the ER-associated lncRNAs were investigated in a ChIP-seq study in MCF7 cells following estrogen treatment. MCF7 cell-line RNA-seq and ChIP-seq data were obtained from publicly available Short Read Archive (ERX030990 and SRX113365) at NCBI.

Results: On average, in each patient 5000 lncRNAs were detectable. Expression of 22 lncRNAs were associated ER status (p<0.05) in breast tumors. 17 of the 22 ER-associated lncRNAs were detectable in ER-positive MCF7 breast cancer cells and were all responsive to E2 (estradiol) treatment. Interestingly, all 17 lncRNAs showed significant downregulation (> = 2 fold) upon tamoxifen treatment. To further evaluate the regulatory relationship between ER and the 22 lncRNAs, we identified the binding sites of ER in estrogen-treated MCF7 cells using ChIP-seq. We found that >50% of the ER-associated lncRNAs had ER binding site either overlapping or neighboring the lncRNA.

Discussion: We have shown that lncRNA expression levels are associated with ER status in breast tumors. We have further established that they are estrogen-responsive with a majority being direct targets of ER using cell-line data. Further functional validation studies are ongoing. We are also exploring lncRNA-mRNA expression for coding partners of lncRNAs to identify coding/noncoding gene regulatory networks important in estrogen response. Understanding the regulatory effects of lncRNA expression opens up new opportunities for stratification and management of breast cancers.

Citation Information: Cancer Res 2013;73(24 Suppl): Abstract nr P4-04-02.

Abstract P4-01-02: Association of DCE-MRI texture features with molecular phenotypes and neoadjuvant therapy response in breast cancer

Purpose: MRI imaging phenotype features such as volume and morphology are used to characterize tumor heterogeneity and tumor response. Texture-based imaging features are important in lesion characterization but their relationship to molecular phenotypes and response is unclear. Molecular stratification of breast cancer into luminal, basal, ERBB2, and normal-like can be made based on gene expression profiles. We investigate how texture-based imaging features relate to tumor biology, genetic subtypes and neoadjuvant therapy response using MRI, histopathological and RNA-sequencing data.

Materials and Methods: Data from 74 Stage IIA to IIIB breast cancer patients enrolled in neo-adjuvant clinical trials NCT00617942 and NCT00723125 were retrospectively reviewed. We evaluated 37 gray-level co-occurrence matrix features (GLCM) on post-contrast T1 fat-suppressed images of 38 HER2− tumors and 35 HER2+ tumors. The texture features included angular second moment, contrast, correlation, first diagonal moment, entropy, regularity, roughness, line likeness and other statistical summaries. We also performed RNA-sequencing on 23 tumors and compared RNAseq-based PAM50 clustering with texture-based clustering. Patients with pCR and RCB class=I were determined to be responders and the rest were labeled non-responders. Wilcoxon signed rank test was used to compare luminal vs. basal, ER+ vs. ER− and PR+ vs. PR- tumors and determine the discriminative power of the texture features. We then performed hierarchical clustering on our patient data set based on the significant texture features and evaluated their association with subtypes, hormone receptor status and response. Statistical significance of clusters was determined by hypergeometric test.

Results: We found five MRI texture features to be significantly associated with tumor subtypes: first diagonal moment, contrast range, correlation range and entropy range (p < 0.05). These five features together differentiated basal and luminal PAM50 subtypes with p = 0.001. Our analysis also showed an association between texture features and tumor hormone status. ER− tumors clustered strongly (13 of 20 ER− cases clustered, p = 0.009) with the 23 significant ER-associated texture features. Similarly, the PR- tumors formed tight grouping (15 of 24 PR- cases clustered, p = 0.006) with 26 significant PR-associated texture features in HER2− patients. Interestingly, only two texture features, entropy range and regularity, distinguished between responders and non-responders (p = 0.04). These features will be further evaluated with DCE-MRI data capturing post brief exposure dynamics.

Conclusion: Our results indicate that certain texture features from DCE-MRI images do capture biological heterogeneity in tumors and can potentially complement standard clinical evaluations. Texture features have previously been assessed for diagnostic settings but to our knowledge this is the first study that shows association of texture features with breast cancer subtyping and neoadjuvant therapy response. We speculate that this could potentially impact clinical management decisions and therapy selection.

Citation Information: Cancer Res 2012;72(24 Suppl):Abstract nr P4-01-02.

Abstract PD05-05: RNA-seq identifies unique transcriptomic changes after brief exposure to preoperative nab-paclitaxel (N), bevacizumab (B) or trastuzumab (T) and reveals down-regulation of TGF-β signaling associated with response to bevacizumab

Background: Identification of differentially expressed transcripts upon brief exposure to preoperative therapy can help determine likely response markers. We quantify and compare differential transcript expression using RNA-seq on patient samples before and after one dose of T or B or N. We also evaluate correlation of brief-exposure transcriptomic changes with response to B.

Methods: We sequenced transcriptomes of core biopsy RNA from 50 pairs of breast tumors obtained from neoadjuvant clinical trials BrUOG 211A/211B. Patients were given a run-in dose of B or N or T, followed by combination biologic/chemotherapy (HER2− with B/carboplatin/N; HER2+ with T/carboplatin/N). We sequenced biopsy pairs obtained pre/post 10 day exposure to run-in monotherapy. Paired-end sequencing was done on Illumina GAII platform using amplified total RNA with 74bp read length, yielding expression data for 22,302 genes and 35,768 transcripts. We evaluated transcriptomic changes upon brief exposure to monotherapy assuming Poisson-distributed read-counts, followed by multiple testing correction and enrichment analysis of 185 KEGG pathways. We investigated association of transcriptomic changes upon brief exposure and pathologic complete response (pCR) in the B arm. Differential expression of previously published signatures of tumor vasculature, TGF-β, β-catenin, MYC, E2F3 and RAF-MEK pathway activities were evaluated to identify associations with pCR.

Results: PAM50-based clustering showed individual samples cluster together, demonstrating that tumor subtypes do not change over the 10-day treatment. We identified unique transcripts that were significantly differentially expressed in each therapy arm (p < 0.05;FDR<0.1). Significant down-regulation of tumor vasculature-related genes was seen in B samples (p = 0.05). We found 1024 genes whose significant differential expression correlated with pCR in the B arm (Mann-Whitney p-val<0.05; abs log2-fold change≥0.5). Only 4 KEGG pathways, TGF-β signaling, Cell Cycle, DNA Replication and Steroid Biosynthesis were found to be enriched (p ≤ 0.05) in the pCR-associated gene list, and displayed significant down-regulation of member genes within the pCR group. To further evaluate the enrichment results, we used several published pathway activity gene signatures. Interestingly, clustering of the B-treated samples using a TGF-β response signature strongly clustered pCR cases due to down-regulation of TGF-β activity in that group (p = 0.004). We found that the TGF-β signature was most informative of pCR when compared to E2F3, RAF-MEK, β-catenin and MYC signatures in the B arm.

Conclusions: This is the first study to compare differential gene expression upon brief exposure across therapies using RNA-seq technology. The association of TGF-β activity with pCR in B arm was identified using both a bottom-up statistical approach and with a previously published TGF-β activity signature. The unique aspects of transcriptional response to each treatment and the association of transcriptional changes with response underscore the value of the brief-exposure paradigm to identify markers of neo-adjuvant therapeutic response.

Citation Information: Cancer Res 2012;72(24 Suppl):Abstract nr PD05-05.

P3-06-04: Sno/miRNA Expression Via Next Generation Sequencing: Variation in Patients before and after Treatment

Background

Aberrant expression of small RNA molecules has been shown in breast cancer. It is yet unclear if variation exists in small RNA molecule expression in response to treatment. Since next generation sequencing offers more globally sensitive detection of sno and miRNAs, we performed an RNA-seq study to explore patients pre- and post- brief exposure to treatment.

Methods: We sequenced transcriptomes of frozen biopsy samples from 8 breast cancer patients enrolled in a clinical trial for neoadjuvant therapy using trastuzumab (HER2 positive) or bevacizumab (HER2 negative) with chemotherapy. Tumor core biopsies were taken before and after 10 days of either biologic or nab-paclitaxel treatment and stored in OCT compound. Total RNA was extracted and libraries were constructed for the 16 samples using TruSeq (Illumina). We performed 74bp paired-end sequencing with the Illumina GAII platform. Sequences were aligned to the sno/miRNA track (containing 928 miRNAs and 413 snoRNAs) in UCSC and read counts were determined using Bowtie. We performed differential miRNA and snoRNA expression analysis pair-wise in all pre- and post-therapy samples. Given that miRNA deregulation relies on their protein-coding gene targets, we analyzed the predicted targets of the significantly varying miRNAs for functional enrichment.

Results: Each sample had on average 46 million paired-end reads, of which on average 70% were mapped to the human genome. Overall, we detected 138 miRNAs in at least one sample, with each sample expressing 33 miRNAs on average. We detected a total of 11 miRNAs (7%) that showed significant differential expression with treatment. Interestingly, 6 of these miRNAs varied in all patients. The predicted targets of these miRNAs were enriched in DNA-dependent transcription, gene expression, cell proliferation and cell communication. Similarly, we detected 202 snoRNAs in at least one sample, with each sample expressing 87 snoRNAs on average. Of these, we found 21 snoRNAs (10%) to vary significantly upon treatment and 6 of these snoRNAs showed expression changes in all patients.

Conclusions: These results suggest that variation in sno/miRNA expression may play a role in response to treatment. The consistent variation of sno/miRNAs in response to treatment suggests shared small RNA-mediated mechanisms. If validated, these results suggest that next generation sequencing technologies will allow lead to improved methods of stratifying, subclassifying and managing breast cancer.

Citation Information: Cancer Res 2011;71(24 Suppl):Abstract nr P3-06-04.

P3-06-01: Next Generation RNA Sequencing Reveals Changes in Gene Expression and Alternative Splicing upon Brief Exposure to Therapy in Early Breast Cancer

Background

The use of next generation RNA sequencing (RNA-seq) allows for the characterization of the transcriptome at levels of detail unachievable by array-based technologies. RNA-seq analysis can quantify expression of novel transcripts and alternatively spliced isoforms in addition to known genes. Alternative splicing allows for flexibility in production of protein isoforms and is frequently dysregulated in cancer. As splice variants may play a role in response to therapy (Solier, et al, Cancer Res., 2010), we studied differential gene and isoform expression in breast cancers after one dose of treatment, prior to a course of preoperative therapy.

Methods: We sequenced transcriptomes of core biopsy samples from 8 breast cancer patients enrolled in a preoperative clinical trial using trastuzumab (HER2 positive) or bevacizumab (HER2 negative) with chemotherapy. Tumor core biopsies were taken before and after 10 days of either biologic or nab-paclitaxel treatment and stored in OCT compound. Total RNA was extracted, amplified and libraries were constructed for the 16 samples using TruSeq (Illumina). Paired-end sequencing was performed on the Illumina GAII platform with read length of 74bp. Sequence data was mapped using TopHat and transcript abundance in FPKM units (Fragments per kilo-base of mRNA per million reads) estimated for a total of 22,160 unique genes and 34,449 unique transcripts from RefSeq. Differential expression of transcripts between baseline and 10-day samples was estimated using t-statistics with read-counts modeled as a Poisson distribution. Differentially expressed transcripts were selected at a significance level of 0.05 after multiple testing correction.

Results: Each sample had on average 46 million paired-end reads, of which on average 70% were mappable to the human reference genome (UCSC, hg19). A median of 138 (range 68–948) transcripts varied with treatment. GO analysis showed enrichment of cell-adhesion, apoptosis, differentiation and cell proliferation pathways. Interestingly, the isoforms of several known cancer genes such as TP53 were seen in all treatment types. Certain isoforms were only seen to change upon brief exposure to chemotherapy such as BCL2 whereas TNF ligand and PCDH isoforms showed significant change only with biologic agents.

Conclusions: These results suggest that recurrent changes in both canonical genes and splice variants occur over the course of treatment in early breast cancer. This underscores the value of RNA-seq to provide novel information that may be clinically useful. Brief exposure to monotherapy prior to combination treatment may provide important mechanistic insights and produce predictive biomarkers. Biologic treatments may produce unique changes that can only be discovered with novel next generation sequencing techniques.

Citation Information: Cancer Res 2011;71(24 Suppl):Abstract nr P3-06-01.

Evaluation of ER/PR and HER2 status by RNA sequencing in tissue core biopsies from preoperative clinical trial specimens

46

Background: Next-generation sequencing for measuring RNA (RNASeq) offer increased sensitivity, dynamic range and provide unbiased detection of all transcripts. To evaluate the clinical utility of such methods, we sequenced entire transcriptomes from fresh-frozen biopsies in a cohort of 120 patients enrolled on a preoperative therapy trial receiving carboplatin, nab paclitaxel and either bevacizumab (HER2-) or trastuzumab (HER2+).

Methods: Total RNA was extracted and amplified from frozen breast core biopsies and libraries constructed using TruSeq (Illumina). Sequencing was performed on the Illumina GAII platform. 75bp reads were mapped using Tophat and transcript abundance in FPKM units (Fragments per kilo-base of mRNA per million reads) calculated using Cufflinks. CLIA approved assays were performed for ER, PR, HER2 (IHC+/- FISH) on patient tumors. Four tumors from each subtype (ER +ve/HER2 -ve; HER2 +ve; ER/HER2 -ve) were analyzed for correlation with clinical status. PAM50 classification will be provided for verification of molecular subtypes.

Results: RNA-Seq library construction/sequencing were successful in 12/12 samples with 50-90% reads mapped. ER +ve tumors ranged in FPKM values from 1.76-22.67 and ER -ve tumors ranged from 0.00-0.79. i.e. ER RNASeq measurements can separate clinical ER status. HER2 +ve tumors ranged in FPKM values from 2.62-21.71 and HER2 -ve tumors from 0.21-1.79. Of note, 7/8 HER2 -ve tumors ranged from 0.21-0.87 with one ‘outlier’ at 1.79±0.3. This outlier was HER2 IHC 2+, FISH ratio 1.1 with 45% of tumor cells with polysomy chromosome 17. Correspondence of ER/PR and HER2 status with molecular subtyping by PAM50 analysis will be presented.

Conclusions: RNASeq has potential to provide in depth analysis of the breast cancer transcriptome and a single analysis test for standard markers. In addition, RNASeq may uncover unexpected expression patterns in conventionally-defined HER2 -ve tumors. If reproducible in larger datasets, this technology may provide both standard and novel information previously unavailable to oncologists and their patients.

Genome-Wide DNA Methylation Profiles of Breast Tumors Reveal Loci Associated with Relapse Risk

Background:Breast cancer prognosis is used in determining the course of adjuvant therapy for patients. Clinical prognostic indices like the Nottingham Prognostic Index have poor specificity, overestimate the risk of disease recurrence and necessitate more specific and robust prognostic markers. Prognostic gene expression markers are already in clinical use and show improved decision support. Methylation of CpG islands, an important regulator of gene expression, is reported to be disregulated in tumors, thus making methylation markers an important alternative to gene expression markers. We present the results of a genome-wide study that explored loci whose methylation status was significantly associated with recurrence risk.Methods:We used 108 frozen primary breast cancer specimens with ten year follow-up and extensive clinical data including histopathological measurements to identify potential epigenetic markers associated with recurrence risk. Using a previously validated array based method (Kamalakaran et. al., Nucleic Acids Research, 2009) we performed genome-wide measurements of differential CpG island methylation covering over 150,000 loci. We evaluated each locus for its ability to stratify patients into good or poor prognosis groups depending on its methylation status. Statistical significance was established using permutation analysis with appropriate multiple testing corrections. Prognostic markers independent of histopathological factors (ER, PR, HER2, tumor size, grade, node status, age) were identified using multivariate Cox regression analysis.Results:The methylation status of several loci proximal to genes significantly stratified samples independent of other clinical variables. Demethylation of several loci were associated with poor prognosis including ADAMTS4 (Hazard Ratio = 17.5, p-value<<0.001), a metalloproteinase previously implicated in the progression of glioma; DNA Topoisomerase I (HR = 3.81; 95% CI = 1.953-7.462; p<<0.001), implicated in chemotherapy resistance; and JMJD2C (HR = 3.7; 95% CI = 1.828-7.519; p<<0.001), which was found to be frequently amplified in esophageal cancers. The methylation of several loci also had significant association with poor prognosis, such as several members in the forkhead box family (FOXF1, FOXG1B, FOXJ1, FOXL2) and FHL1 (HR = 4.78; 95% CI = 2.38-9.62; p<<0.001). We selected several loci to form an ensemble classifier with statistically significant performance on our dataset. We show that this classifier achieved much higher specificity when compared to the Nottingham Prognostic Index, while maintaining high sensitivity.Discussion:Our retrospective study of genome-wide DNA methylation in breast cancer has identified several novel markers for prognosis. We found methylation deregulation of CpG islands proximal to genes implicated in metastasis and chemotherapy resistance is associated with poor prognosis. Furthermore, the potential for clinical benefit of these markers is their ability to jointly identify significantly larger number of low-risk patients compared to the Nottingham Prognostic Index.

Citation Information: Cancer Res 2009;69(24 Suppl):Abstract nr 4046.

Subtype Dependent Alterations of the DNA Methylation Landscape in Breast Cancer

Background: The diversity of breast cancers at the clinical, histopathological and molecular level reflects variation in underlying biology and affects the clinical implications for patients. Gene expression studies have identified five breast cancer subtypes with distinct expression profiles – Luminal A, Luminal B, basal, ErbB2 enriched and Normal-Like. DNA methylation is an important regulator of gene expression that is also known to be deregulated in tumors. We set out to determine the relationship between DNA methylation and breast cancer subtypes in 108 breast cancer samples with previously determined expression subtypes.Methods: We performed high-throughput genome-wide scans of CpG methylation in 108 tumors and 11 normal tissues using our previously validated Methylation Oligonucleotide Microarray Analysis (MOMA) method [Kamalakaran, S et al. Nucleic Acids Research, 2009)]. We identified loci that were most varied across all tumors or had the most significant alterations and performed unsupervised hierarchical clustering on those loci. We then used a genetic algorithm based feature selection method to identify a subset of those loci that could cluster the sample set by expression subtype. We then characterized the loci contributing to subsetting and where possible, the relationship between methylation and gene expression.Results: Unsupervised hierarchical clustering using the 500 most differentially methylated loci across all tumors and 100 most significant altered loci between tumors and Normal tissues clustered the tumors into 3 major clusters – 82% of Cluster I belonged to Luminal Subtypes (22 Luminal A and 4 Luminal B), and 86% of Cluster II samples were of Basal or ErbB2+ subtypes. Cluster III did not show any expression subtype specific enrichment, but contained samples whose expression subtype was inconclusive with weak correlations to multiple expression subtypes. Interestingly, methylation loci that contributed to this clustering were not localized to CpG islands immediately upstream of genes, with 354 loci far from gene transcription start sites. These non-geneic loci did not show any significant regulatory potential based on cross-species conservation measures and no clear function could be assigned to these regions. The remaining 146 loci could be mapped to known genes. Gene expression microarray measurements were available for 79 of these geneic loci and 36 showed significant correlation of methylation to expression levels (p<0.05), implying possible functional effects of the methylation on gene expression. Additionally, distinct subtype specific patterns of methylation could also be detected in known cancer associated genes. CpG islands in the HOXA gene cluster and many other homeobox genes were significantly more methylated in Luminal A tumors.Conclusions: Our results suggest that there are subtype dependant genome-wide alterations in the methylation landscape in breast cancers, especially near homeobox genes. Many more CpG islands with no apparent functional significance get methylated according to subtype in addition to those CpG islands associated with genes with known cancer related functions.

Citation Information: Cancer Res 2009;69(24 Suppl):Abstract nr 1144.