Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2021 Apr 15.
Published in final edited form as: Mol Cell Endocrinol. 2020 Feb 5;506:110746. doi: 10.1016/j.mce.2020.110746

Characterization of Basal and Estrogen-Regulated Antisense Transcription in Breast Cancer Cells: Role in Regulating Sense Transcription

Tim Y Hou 1,2,4, Tulip Nandu 1,2,4, Rui Li 1,2,3,4, Minho Chae 1,2, Shino Murakami 1,2,3, W Lee Kraus 1,2,3,5
PMCID: PMC7089808  NIHMSID: NIHMS1561704  PMID: 32035111

Abstract

Estrogen-responsive breast cancer cells exhibit both basal and estrogen-regulated transcriptional programs, which lead to the transcription of many different transcription units (i.e., genes), including those that produce coding and non-coding sense (e.g., mRNA, lncRNA) and antisense (i.e., asRNA) transcripts. We have previously characterized the global basal and estrogen-regulated transcriptomes estrogen receptor alpha (ERα)-positive MCF-7 breast cancer cells. Herein, we have mined genomic data to define three classes of antisense transcription in MCF-7 cells based on where their antisense transcription termination sites reside relative to their cognate sense mRNA and lncRNA genes. These three classes differ in their response to estrogen treatment, the enrichment of a number of genomic features associated with active promoters (H3K4me3, RNA polymerase II, open chromatin architecture), and the biological functions of their cognate sense genes as analyzed by DAVID gene ontology. We further characterized two estrogen-regulated antisense transcripts arising from the MYC gene in MCF-7 cells, showing that these antisense transcripts are five-prime capped, three-prime polyadenylated, and localized to different compartments of the cell. Together, our analyses have revealed distinct classes of antisense transcription correlated to different biological processes and response to estrogen stimulation, uncovering another layer of hormone-regulated gene regulation.

Keywords: Antisense transcription, Antisense RNA, Estrogen receptor, Estrogen signaling, MYC gene, Transcriptional regulation

1. Introduction

The human genome is extensively transcribed, generating a large pool of non-coding RNAs, in addition to a relatively small fraction of coding mRNAs (Carninci et al., 2005; Djebali et al., 2012; Hah et al., 2011; Kapranov et al., 2007). One class of non-coding transcripts, antisense RNAs (asRNAs), are transcribed in the opposite direction of a sense mRNA or a long non-coding RNA (lncRNA) (Katayama et al., 2005; Pelechano et al., 2013). More than 30% of annotated transcripts in the human genome are paired with an antisense transcript (Ozsolak et al., 2010), whose expression can be driven by an independent promoter near the transcription termination sites of the sense gene or cryptic promoters within the body of the sense gene (Pelechano et al., 2013). These overlapping sense-antisense transcription units create opportunities for the regulation of sense gene transcription through a variety of mechanisms. Some potential mechanisms include (1) DNA methylation of the sense promoter (Tufarelli et al., 2003), (2) recruitment of the polycomb repressive complex 2 (PRC2) to modify the histone proteins residing in the promoters of the sense gene to silence transcription (Yap et al., 2010; Yu et al., 2008), and (3) head-to-head collision of RNA polymerase II as it transcribes in opposing directions simultaneously (Crampton et al., 2006; Hobson et al., 2012). Antisense transcripts have also been found in the cytoplasm, where they can hybridize to sense transcripts to increase stability of mRNA by preventing miRNA-mediated silencing (Faghihi et al., 2008; Faghihi et al., 2010).

The potential of antisense transcripts and transcription to regulate gene expression has put a greater focus on their genesis and biological functions (Wanowska et al., 2018; Werner, 2013). In this regard, the expression of antisense transcripts has been studied in a variety of biological systems, such as human cancers (Balbin et al., 2015; Wanowska et al., 2018). Like protein-coding mRNA genes, the asRNA genes are subject to regulation by various signaling pathways, such as the nuclear estrogen signaling pathway, which we have investigated in this study (Hah et al., 2011). Estrogenic hormones, acting through nuclear estrogen receptor (ER) proteins (ERα and ERβ), are potent regulators of gene expression programs (Hewitt et al., 2018). When estrogen-bound ER dimerizes and binds to thousands of sites across the genome to act as a regulator of transcription, resulting in estrogen-regulated alterations in the transcriptome (Franco et al., 2015; Franco et al., 2018; Hah et al., 2011; Hah et al., 2013; Murakami et al., 2017; Sun et al., 2015). We have previously documented the rapid, extensive, and transient transcriptional response to estrogen treatment in ER-positive breast cancer cells, profiling the transcription of not only protein-coding genes, but also non-coding RNAs transcribed by all three RNA polymerases (Hah et al., 2011; Hah et al., 2013; Sun et al., 2015). How antisense transcription responds to hormone stimulation and its relationship to sense gene transcription are largely unexplored.

In this study, we have characterized basal and estrogen-stimulated antisense transcription in ERα-positive MCF-7 human breast cancer cells. Using a computational pipeline that we described previously (Chae et al., 2015; Danko et al., 2014; Hah et al., 2011), we annotated over 6,000 antisense transcription units in MCF-7 cells. We classified these antisense transcription units based on the antisense transcription termination site (TTS) relative to the corresponding sense transcription start site (TSS). We found that different classes of antisense transcription are associated with different biological processes and responses to estrogen. Finally, molecular characterization of two antisense transcripts of the MYC gene confirm our identification strategy and suggest potential roles in estrogen responses and breast cancer biology.

2. Methods and Materials

2.1. GRO-seq, RNA-seq, ChIP-seq, and DNase-seq data sets

We used the following published GRO-seq, RNA-seq, and ChIP-seq data sets from MCF-7 cells (± treatment with E2) for the analyses described herein. All data sets are available from the NCBI’s Gene Expression Omnibus repository (GEO; http://www.ncbi.nlm.nih.gov/geo/) using the accession numbers listed below.

2.2. Defining antisense transcription units from GRO-seq data

We used an established data analysis pipeline, which we have described previously (Hah et al., 2011; Luo et al., 2014), to annotate antisense transcription units from GRO-seq data obtained from MCF-7 ERα-positive human breast cancer cells. Our pipeline can be summarized as follows: (1) Collect all published annotations for sense transcription units from mRNA and lncRNA genes; (2) Define all transcription units in MCF-7 cells from GRO-seq data using the groHMM analysis tool (Chae et al., 2015; Danko et al., 2014); (3) Align GRO-seq-called transcription units with the sense annotations to define antisense transcripts; and (4) Define steady-state RNA levels for sense and antisense transcripts using RNA-seq data. The details are as follows.

Collecting all annotations for sense transcription units for mRNA and lncRNA genes.

We collected the annotations for all protein coding and lncRNA genes from RefSeq (GRCh37/hg19), UCSC Genes (GRCh37/hg19) and GENCODE Comprehensive gene annotation (v.15). We also included a set of 726 previously unannotated lncRNAs identified by our lab (Sun et al., 2015), and collapsed them into a set of non-overlapping consensus annotations for all unique gene symbols. This collection of annotations was used as the reference set for sense transcription units.

Defining all transcription units in MCF-7 cells from GRO-seq data using the groHMM analysis tool.

GRO-seq reads were aligned to human reference genome (GRCh37/hg19), including autosomes, the X chromosome, and one complete copy of an rDNA repeat (GenBank ID: U13369.1) using the SOAP2 (v.2.21) software package (Li et al., 2009) with the following parameters: (1) all n mappings were removed (−r 0), (2) three mismatches were allowed in each mapped read (−v 3), (3) low-quality reads with more than 10 ambiguous bases were removed (−N 10), and (4) for reads failing to align over the entire length of the read, the first 32 bp was used (−l 32). To determine the transcription units for expressed genes in MCF-7 cells from GRO-seq data, we used groHMM (v.1.0.2), a computational tool based on a two-state hidden Markov model (Chae et al., 2015; Danko et al., 2014). groHMM takes as input the read counts across the genome and assigns them to two states, “transcribed” and “non-transcribed,” which are then used to call transcription units based on a number of tunable parameters. We used the shape setting parameter of 5 and -log transition probability of 200. All other parameters were set at the default.

Aligning GRO-seq-called transcription units with sense annotations to define antisense transcripts.

We compared the GRO-seq-called transcription units with the collection of sense annotations described above. All GRO-seq-called transcription units that met the following parameters were classified as antisense: (1) run antisense to a sense gene in the consensus annotation and (2) >20% of the GRO-seq-called transcription unit overlapped >20% of a well-annotated gene on the opposite strand.

2.3. Transcript maps

Individual antisense transcription units were visualized in transcript maps (e.g., Fig. 1B) that relate the genomic position of each transcription unit to an associated sense transcript using custom PERL and R scripts (available on request from W.L.K.). The sense genes were scaled to a standard length of 1 kb and the cognate antisense transcript partners were scaled accordingly to show the relative lengths and positions. Based on these analyses, we could define three classes of antisense transcription units: (1) those that run through the sense transcription start site (TSS) (Class I), (2) those that terminate at or less than 1 kb before the sense TSS (Class II), and (3) those that terminate more than 1 kb before the sense TSS (Class III). Sense transcripts with no cognate antisense transcripts were used as a background reference.

Fig. 1. Defining antisense transcription units in MCF-7 cells.

Fig. 1.

Open in a new tab

(A) Flowchart describing the computational analysis pipeline used to define and characterize antisense transcription units based on GRO-seq, RNA-seq, ChIP-seq, and DNase-seq data.

(B) Graphical representation of the orientation, position, and length of antisense transcription units in MCF-7 cells relative to their sense gene. All sense transcription units are scaled to 1 kb (grey box) for visualization. The red line indicates the transcription start site (TSS) of the sense gene with sense transcription going from left to right as indicated by the red arrow at the top, and each blue line, scaled accordingly, indicates the relative length and position of the corresponding antisense transcription unit.

(C) Stability of sense (S) and antisense (AS) transcripts in MCF-7 cells treated with either vehicle (Veh) or 17β-estradiol (E2). RNA stability was calculated based on log2(RNA-seq FPKM/GRO-seq RPKM). Bars marked with different letters are significantly different (Wilcoxon rank sum test, p < 2 × 10−16).

(D and E) Correlation between the stability of sense and antisense transcripts in MCF-7 cells treated with either (D) vehicle or (E) E2.

2.4. Preparation of strand-specific total RNA-seq libraries

Total RNA-seq libraries were generated as previously described (Zhong et al., 2011) with the following modifications. Total RNA was isolated from MCF-7 cells treated with 100 nM E2 for three hours using the RNeasy Plus kit (Qiagen; 74134). Ribosomal RNA was depleted from 2 μg total RNA using the Ribo-Zero rRNA Removal kit (Epicentre; MRZH116), and fragmented at 94 °C in 2 x SuperScript III first-strand buffer (LifeTechnologies; 18080) supplemented with 10 mM DTT. The fragmented RNA was subjected to first strand cDNA synthesis and subsequent procedures as described by Zhong et al. (Zhong et al., 2011). The libraries were barcoded with indices as described for the Illumina TruSeq™ RNA Prep Kit, multiplexed, and sequenced in Illumina HiSeq 2000.

2.5. Defining steady-state RNA levels from total RNA-seq data

Read Alignment.

We developed a computational pipeline to analyze total RNA-seq data, which includes the following steps: (1) Raw data were analyzed using FastQC tool for quality control; (2) Paired-end Total RNA-seq reads were aligned to the human reference genome (GRCh37/hg19) including autosomes, X chromosome, and one complete copy of an rDNA repeat (GenBank ID: U13369.1) using the spliced read aligner TopHat version 1.4.0 (Trapnell et al., 2009) (default parameters and ‘max-multihitis=1’,’mate-inner-dist=100’). The reads were aligned to the rDNA repeat to check rRNA contamination in the library; and (3) The output from TopHat was converted into BED files using SAMtools (v.0.1.18) (Li et al., 2009) and BEDTools (v.2.16.2) (Quinlan et al., 2010).

Determining steady-state RNA levels for sense and antisense transcripts.

All strand-specific RNA-seq reads falling within the sense transcript annotations or the GRO-seq-called antisense transcription units were collected and expressed as RPKM. We did the same for the GRO-seq reads so that we could compare the levels of transcription and steady-state RNA. The antisense transcripts were categorized as annotated or previously unannotated based on significant overlap (20% or more) with an annotation in the same direction from our consensus annotation.

2.6. Analysis of ChIP-seq and DNase-seq data sets

The raw reads were aligned to the human reference genome (GRCh37/hg19) using Bowtie (v1.0.0) (Langmead et al., 2009). Uniquely mappable reads were converted into (1) bigWig files using BEDTools (v.2.16.2) (Quinlan et al., 2010) for visualization in the UCSC genome browser and (2) R data files for subsequent analyses. The aligned data sets were then used to determine the enrichment of RNA polymerase II (RNA Pol) II and histone H3 lysine 4 trimethylation (H3K4me3), as well as chromatin accessibility, around the TSSs of the promoters driving expression of the sense and antisense transcripts.

2.7. Metagene and boxplot analyses

Metagenes were used to illustrate the distribution of GRO-seq, ChIP-seq, and DNase-seq reads in 8 kb windows around the TSSs (± 4 kb) of sense and antisense promoters. They were generated using the metagene function in groHMM (Chae et al., 2015; Danko et al., 2014), as we have described previously (Hah et al., 2011; Hah et al., 2013). We used boxplot representations in order to minimize the bias that can be caused by outliers in the metagene analyses and for efficient comparison across different groups. The read distributions in 8 kb windows around the TSSs of sense and antisense promoters were calculated and plotted using the boxplot function in R. All of the metagenes and boxplots were scaled to a library size of 15 million reads to minimize differences caused by variability in sequencing depth among samples.

2.8. Gene ontology (GO) analyses

Gene ontology analyses were performed using the DAVID (Database for Annotation, Visualization, and Integrated Discovery) tool (Huang et al., 2007; Huang et al., 2009). As input, we used the three classes of antisense transcription units described above. Sense transcripts with no cognate antisense transcripts were used as a background reference. DAVID returns clusters of related ontological terms that are ranked according to an enrichment score. We listed the top four GO biological process term in each cluster (based on the p-value) from the top ten clusters (based on the enrichment score).

2.9. Cell culture and treatments

MCF-7 cells, kindly provided by Benita Katzenellenbogen (University of Illinois, Urbana-Champaign), were maintained in MEM medium with Hank’s salts (Sigma, M1018) supplemented with 5% HyClone calf serum (GE Healthcare, SH30072). The cells were tested and verified as mycoplasma-free every 6 months and screened regularly for expression of estrogen receptor alpha (ERα) by Western blotting to confirm cell identity. For experiments involving estrogen treatment, the cells were grown for at least 3 days in phenol red-free MEM Eagle medium with Earle’s salts (Sigma, M3024) supplemented with 5% charcoal-dextran-treated calf serum and then treated with ethanol (vehicle) or 17β-estradiol (E2; 100 nM) for the times specified in the figures and legends.

MCF-10A and MDA-MB-231 cells were purchased from the ATCC. MCF-10A cells were maintained in mammary epithelial cell growth medium (MEGM) containing the supplement pack (Lonza, CC-3151). MDA-MB-231 cells were maintained in phenol red-free Dulbecco’s modified Eagle medium/nutrient mixture F-12 Ham (DMEM/F-12; Sigma, D2906) supplemented with 10% charcoal-dextran-treated calf serum, 6 ng/mL of human recombinant insulin (Sigma, I6634), 3.75 ng/mL hydrocortisone (Sigma, H0888), 16 μg/mL glutathione (Sigma, G6013), 100 units/mL penicillin/streptomycin (Gibco, 15140122), and 25 μg/mL gentamicin (Gibco, 15710064). For experiments involving estrogen treatment, the cells were grown for at least 3 days in estrogen-free medium and then treated with ethanol (vehicle) or 17β-estradiol (E2; 100 nM) for the times specified in the figures and legends. The cells were tested and verified as mycoplasma-free every 6 months.

2.10. 5’ and 3’ RACE, and cloning of MYC-AS1 and MYC-AS2

To confirm the GRO-seq-based antisense transcript annotation approach, we mapped the 5’ and 3’ ends of, and subsequently cloned, two antisense transcripts of the MYC gene (MYCAS1 and MYCAS2) from MCF-7 cells. Total RNA from MCF-7 cells was isolated using TRIzol Reagent (Invitrogen, 15596) according to manufacturer’s protocol. 5’ and 3’ rapid amplification of cDNA ends (RACE) was performed using the FirstChoice RLM-RACE kit (Life Technologies, AM1700M). The 5’ RACE assays were performed with or without tobacco acid pyrophosphatase (TAP, Epicentre, T19050) to determine if the MYC antisense transcripts were 5’ capped. The 3’ RACE assays were performed with or without an in vitro RNA polyA tailing step using E. coli poly(A) polymerase (NEB, M0276S) to determine if the MYC antisense transcripts were polyadenylated at their 3’ ends. Based on the RACE analyses, cDNAs for MYCAS1 and MYCAS2 were PCR amplified using oligonucleotide primers listed below, cloned into pcDNA3 (Invitrogen) vector, followed by subcloning to pLPCX vector and sequenced.

  • 5’ RACE primers:
    • MYCAS1 5’ RACE: TCATCCAGGACTGTATGTGGAG
    • MYCAS2 5’ RACE: AGAGGTGTTAGGACGTGGTGTT
  • 3’ RACE primers:
    • MYCAS1 3’ RACE: TGAAACTCTGGTTCACCATGTC
    • MYCAS2 3’ RACE: CTACTCCAAGGAGCTCAGGATG
  • cDNA cloning primers:
    • MYCAS1 Cloning Forward: CCGGAATTCATGGAGCACCAGGGGCT
    • MYCAS1 Cloning Reverse: ATAAGAATGCGGCCGCGGCAAATTGTTTTCCTC
    • MYCAS2 Cloning Forward: CCCAAGCTTTTTCTCTCGCCGGCTGGA
    • MYCAS2 Cloning Reverse: CCGGAATTCCACTTTAATGCTGAGATGAGTC

2.11. RNA isolation and RT-qPCR

RNA isolation and RT-qPCR were performed as described previously (Luo et al., 2014). Briefly, MCF-7 cells were grown and treated as described above. After collecting the cells, total RNA was isolated using TRIzol Reagent (Invitrogen, 15596). Cytoplasmic and nuclear RNAs were isolated using the Ambion PARIS kit (Life Technologies, AM1921), according to manufacturer’s protocol. RNA was reverse transcribed with MMLV reverse transcriptase (Promega, M1701) using either oligo(dT) or gene-specific primers for AS1 and AS2, and then analyzed by quantitative real-time PCR (qPCR) using the gene-specific primers listed below. Briefly, cDNA, 1x SYBR Green PCR master mix, and forward and reverse primers (250 nM) were mixed and amplified with 45 cycles (95°C for 10 second, 60°C for 10 second, 72°C for 1 second) following an initial 5 minute incubation at 95°C using a Roche LightCycler 480 384-well detection system. Melting curve analyses were performed to ensure that only the targeted amplicon was amplified. All target gene expression was normalized to the expression of the gene encoding beta-actin. All experiments were performed on at least three separate biological replicates to ensure reproducibility. The sequences of the primers used for RT-qPCR are listed below.

  • RT-qPCR primers:
    • MYCAS1 RT primer: GCGAGCACAGAATTAATACGACTTAGGGGATAGCTCTGCAAGG
    • MYCAS1 forward: CGGGAGGCAGTCTTGAGTTA
    • MYCAS1 reverse: GCGAGCACAGAATTAATACGACT
    • MYCAS2 RT primer: GCGAGCACAGAATTAATACGACTGGCATTTAAATTTCGGCTCA
    • MYCAS2 forward: ACCCAACACCACGTCCTAAC
    • MYCAS2 reverse: GCGAGCACAGAATTAATACGACT
      The RT primers for gene specific reverse transcription contain a universal adaptor sequence followed by gene specific sequences. During the PCR step, we used gene specific forward PCR primers with Universal Adaptor Reverse primer to ensure strand-specific amplification.
    • MYC sense forward: TCGGATTCTCTGCTCTCCTC
    • MYC sense reverse: CCTGCCTCTTTTCCACAGAA
    • ACTB control forward: AGCTACGAGCTGCCTGAC
    • ACTB control reverse: AAGGTAGTTTCGTGGATGC
    • MALAT1 control forward:
    • MALAT1 control reverse:
    • MALAT1 Amplicon A forward qPCR GAATTGCGTCATTTAAAGCCTAGTT
    • MALAT1 Amplicon A reverse qPCR GTTTCATCCTACCACTCCCAATTAAT
    • MALAT1 Amplicon B forward qPCR GACGGAGGTTGAGATGAAGC
    • MALAT1 Amplicon B reverse qPCR ATTCGGGGCTCTGTAGTCCT

3. Results

3.1. Characterization of estrogen-regulated antisense transcription in MCF-7 breast cancer cells

We developed a computational pipeline to characterize antisense transcription units regulated by estrogen in MCF-7 breast cancer cells using existing transcriptomic data (Hah et al., 2011) (Fig. 1A). We first defined all expressed transcription units in MCF-7 using our previously developed groHMM computational tool that defines transcription units from GRO-seq data generated from MCF-7 cells treated with either vehicle or 100 nM 17β-estradiol (E2) for 40 minutes (Chae et al., 2015; Danko et al., 2014). Next, we aligned the GRO-seq-called transcription units with annotated sense genes from RefSeq, UCSC, Ensembl, and GENCODE. We also included lncRNAs identified by our lab in MCF-7 cells (Sun et al., 2015). Collectively, these transcripts comprise the set of antisense transcription units in MCF-7 cells. Next, we overlapped the antisense transcription units from MCF-7 cells under basal and estrogen-stimulated conditions with the universe of sense transcription units from the public repositories, and characterized these antisense transcription units by examining steady-state RNA levels (RNA-seq), RNA polymerase II (RNA Pol) II and histone H3 lysine 4 trimethylation (H3K4me3) enrichment (ChIP-seq), and chromatin accessibility (DNase-seq) using available data sets (Fig. 1A). Our analyses are restricted to those antisense transcription units arising from independent and cryptic promoters, and do not include antisense transcription arising from bidirectional promoters shared with the sense gene (i.e., promoter upstream transcripts, PROMPTs). Overall, we found 6,173 antisense transcription units associated with sense genes in MCF-7 cells (Fig. 1B). To visualize the relationship between the sense and antisense transcription, we scaled all the sense transcription units to 1 kb (grey box in Fig. 1B) and plotted the antisense transcription units relative to the them (Fig. 1B).

We have previously devised a metric to determine transcript stability by computing the ratio of RNA-seq FPKM to GRO-seq RPKM, determining the ratio of their steady-state RNA levels to their transcription levels (Sun et al., 2015). Using this approach, we found that antisense transcription units were less stable than their cognate sense transcription units in MCF-7 cells treated with vehicle or E2 (Wilcoxon rank sum test, p < 2 × 10−16) (Fig. 1C). We also found no correlation between stability of sense transcripts and their cognate antisense transcripts in the vehicle (r = 0.07) (Fig. 1D) or E2-treated conditions (r = 0.02) (Fig. 1E). These results indicate that (1) antisense transcription units are less stable than their cognate sense transcription units and (2) the stability of these transcripts are not correlated with each other.

After calling these antisense transcription units in MCF-7 cells using groHMM, we further examined the enrichment of RNA Pol II, H3K4me3, and chromatin accessibility at the promoters of these antisense transcription units (Fig. 2). We first identified the sense and antisense promoters from transcription units called by groHMM. We then aligned and reoriented the promoters driving expression of these antisense transcription units so that they all ran in the same direction, then calculated the average read counts ±4 kb around the transcription start sites (TSSs) of the promoters (Fig. 2A). Both the sense and antisense promoters displayed the classical signatures of active promoters, including increased transcription levels (GRO-seq), enrichment of RNA Pol II and H3K4me3 (ChIP-seq), and an open chromatin architecture (DNaseI) (Fig. 2B). These analyses suggest that the promoters of the antisense transcription units in MCF-7 cells display classical markers of promoter, similar to that of their sense gene promoter counterparts.

Fig. 2. The TSSs of sense and antisense promoters in MCF-7 cells are enriched for active histone marks.

Fig. 2.

Open in a new tab

(A) Schematic diagram showing the identification and analysis of antisense transcription units.

(B) Metagene representations of average read counts for GRO-seq, Pol II, and H3K4me3 ChIP-seq, and DNase-seq around (± 4 kb) the TSSs of sense (top) and antisense (bottom) transcription units.

3.2. Antisense transcription can be classified based on the transcription termination sites relative to the TSS.

After defining antisense transcription units in MCF-7 cells using our computational pipeline, we sought to classify the antisense transcription units based on their transcription termination sites (TTSs) relative to the TSSs of their cognate sense genes. We surmised that the architecture of the antisense TSSs and transcribed gene bodies (i.e., transcription units) in relation to the sense TSSs might have some regulatory consequences. We clustered the antisense transcription units into three classes: (1) Class 1 - antisense transcription running through the sense gene TSS (antisense TTS upstream of the sense TSS); (2) Class 2 - antisense transcription terminating before the sense gene TSS (antisense TTS within 1 kb downstream of the sense TSS); and (3) Class 3 - antisense transcription terminating more than 1 kb downstream of the sense gene TSS (TTS >1 kb downstream of the sense TSS) (Fig. 3A). The average size of the antisense transcription units was 44.3 kb for Class I, 44.4 kb for Class 2, and 47.1 kb for Class 3. We observed that approximately 56% of the antisense transcription units pass through the sense gene TSS (Class 1), while the remaining antisense transcription units belonged to Class 2 (20%) or Class 3 (24%).

Fig. 3. Classification of antisense transcripts in MCF-7 cells.

Fig. 3.

Open in a new tab

(A) Classification of antisense transcription units into three categories based on their transcription termination sites (TTSs) relative to the TSSs of the cognate sense transcription units: (1) transcription units running through the sense TSS, (2) transcription units terminating within a 1 kb window downstream of the sense TSS, and (3) transcription units terminating more than 1 kb downstream of the sense TSS.

(B) Correlation analysis of antisense transcription versus sense transcription using GRO-seq for the defined three categories.

(C) Gene ontology (GO) terms enriched for the sense genes associated with the three categories of antisense transcription units. The analysis was done using the DAVID tool (Huang et al., 2007; Huang et al., 2009).

We then compared the sense and antisense transcription output using GRO-seq data to determine whether these three classes of antisense transcription units correlated with the levels of sense gene transcription (Fig. 3B). In most cases, antisense transcription levels were significantly lower than sense transcription. Although antisense transcription was positively correlated with sense transcription in general, antisense transcription in Class 1 (antisense transcription running through the sense TSS) was correlated most strongly with sense transcription (r = 0.50), whereas antisense transcription in Class 3 (antisense transcription terminating >1 kb upstream of the sense TSS) was correlated weakly to sense transcription (r = 0.29) (Fig. 3B). To discern whether these three classes of antisense transcription units are associated with particular biological functions, we performed gene ontology (GO) analyses using the DAVID tool (Huang et al., 2007; Huang et al., 2009) for the sense genes associated with each of the three classes of antisense transcription units (Fig. 3C). Interestingly, we discovered that the three classes of antisense transcription units are associated with sense genes that define distinct biological processes. For example, the sense genes paired with Class 1 antisense transcription units are associated with transcription regulation, Class 2 are associated with protein regulation, and Class 3 are associated with cellular morphology. This result suggests that sense/antisense gene pairs with different gene architectures control distinct and specific biological processes.

3.3. Genomic characterization of antisense transcripts reveals distinct promoter features for the three classes of antisense transcription units

To better characterize the features of the sense promoters associated with the three different classes of antisense transcription units under basal condition in MCF-7 cells, we mined existing genomic datasets, focusing on (1) the levels of sense and antisense transcription (i.e., GRO-seq; Fig. 4A), (2) recruitment of RNA Pol II to the sense promoters (i.e., Pol II ChIP-seq; Fig. 4B), (3) enrichment of H3K4me3 at the sense promoters (i.e., H3K4me3 ChIP-seq; Fig. 4C), and extent of chromatin accessibility at the sense promoters (i.e., DNase-seq; Fig. 4D). The overall level of transcription for both the sense and antisense genes was significantly greater for genes in Class 1 than Classes 2 and 3 (Wilcoxon rank sum test, p < 2 × 10−14) (Fig. 4A). In accordance with this result, we observed significantly higher levels of RNA Pol II recruitment, H3K4me3 enrichment, and chromatin accessibility for genes in Class 1 than Classes 2 and 3 (Wilcoxon rank sum test, p < 2 × 10−14) (Fig. 4, B–D). These analyses illustrate the different features of the different classes of antisense transcription units and suggest a link between antisense transcription through sense promoters and genomic features associated with active promoters.

Fig. 4. Genomic characterization of antisense transcript promoters in MCF-7 cells.

Fig. 4.

Open in a new tab

Metagene (three left panels) and boxplot (right panel) representations showing genomic features (transcription, Pol II loading, H3K4me3 levels, and chromatin accessibility) of the three antisense transcript unit categories in MCF-7 cells. The plots show the average read counts around (± 4 kb) TSSs. Bars marked with different letters are significantly different (Wilcoxon rank sum test, p < 2 × 10−14).

(A) Results for GRO-seq data showing transcription.

(B) Results Pol II ChIP-seq data showing Pol II loading.

(C) Results H3K4me3 ChIP-seq data showing H3K4me3 levels.

(D) Results DNase-seq chromatin accessibility.

Next, we characterized the antisense transcription units in MCF-7 cells upon estrogen stimulation (i.e., treatment with 17β-estradiol; E2). Antisense transcription was positively correlated with sense transcription in Class 1 (r = 0.49) , similar to what we observed under basal conditions, but was not correlated with sense transcription in Class 2 (r = 0.01) or Class 3 (r = 0.19) (Fig. 5A). Likewise, the overall level of transcription for both the E2-stimulated sense and antisense genes was significantly greater for genes in Class 1 than Classes 2 and 3 (Wilcoxon rank sum test, p < 2 × 10−14) (Fig. 5B). Again, in accordance with this result, we observed significantly higher levels of RNA Pol II recruitment, H3K4me3 enrichment, and chromatin accessibility for genes in Class 1 than Classes 2 and 3 (Wilcoxon rank sum test, p < 2 × 10−14) (Fig. 5, B–E). Furthermore, Class 1 antisense transcription units are associated with sense genes that are highly upregulated by estrogen (compare Figs. 4A and 5B), pointing to a similar mechanism by which E2-induced transcription of sense genes and antisense genes are linked.

Fig. 5. Genomic characterization of antisense transcript promoters in E2-treated MCF-7 cells.

Fig. 5.

Open in a new tab

(A) Correlation analysis of antisense transcription versus sense transcription using GRO-seq for the defined three categories in E2-treated MCF-7 cells.

(B-E) Metagene (three left panels) and boxplot (right panel) representations showing genomic features (transcription, Pol II loading, H3K4me3 levels, and chromatin accessibility) of the three antisense transcript unit categories in MCF-7 cells. The plots show the average read counts around (± 4 kb) TSSs. Bars marked with different letters are significantly different (Wilcoxon rank sum test, p < 2 × 10−14). (A) GRO-seq, (B) Pol II ChIP-seq, (C) H3K4me3 ChIP-seq, and (D) DNase-seq.

3.4. Estrogen-regulated antisense transcripts are 5’-capped, 3’-polyadenylated, and localized to both nuclear and cytosolic fractions

Although antisense transcription occurs in MCF-7 cells in response to estrogen treatment, the act of transcription may not result in the synthesis of stable transcripts. Thus, we generated total RNA-seq libraries from MCF-7 cells treated with either vehicle or E2 for 3 hours (Fig. 6A). Using the MYC gene locus as an example, we observed not only E2-induced sense gene transcription, but also two antisense transcripts, MYC-AS1 and MYC-AS2 (Fig. 6A). To determine whether these antisense transcripts are five-prime capped, we treated the PCR products from 5’ rapid amplification of cDNA ends (5’ RACE) reactions with tobacco acid pyrophosphatase, which cleaves the pyrophosphate bond of the 5′-terminal methylated guanine nucleotide cap of mRNAs. The resulting 5′-monophosphorylated terminus can be ligated to a 3′-hydroxylated terminus. The presence of detectable PCR products for MYC-AS1 or MYC-AS2 only after TAP treatment to expose a 5′-monophosphorylated terminus indicates that these two antisense transcripts are 5’-capped (Fig. 6B). Interestingly, we detected three products for MYC-AS1 in the 5’ RACE reactions, indicating at least three TSSs for MYC-AS1. Next, to determine whether MYC-AS2 and MYC-AS2 are polyadenylated, we performed 3’ RACE with or without in vitro polyadenylation using E. coli poly(A) polymerase. In this case, in vitro polyadenylation serves as a positive control. The presence of PCR products in the absence in vitro polyadenylation indicates that these two antisense transcripts are polyadenylated (Fig. 6C). Again, similar to the multiple TSSs observed for MYC-AS1, we observed two TTSs for MYC- AS1. Together, these results demonstrate that the antisense transcripts MYC-AS1 and MYC-AS2 are 5’-capped and 3’-polyadenylated.

Fig. 6. Molecular characterization of MYC antisense transcripts.

Fig. 6.

Open in a new tab

(A) Genome-browser view of the MYC gene locus (top) showing the annotation of two antisense transcripts (black arrows), mapped by RACE. These annotations correspond to the antisense (blue signals) from both GRO-seq (middle) and total RNA-seq (bottom) data in control and E2-treated MCF-7 cells at the indicated time points. A size scale in kb is shown.

(B) PCR products from 5’ RACE assays for MYC-AS1 and MYC-AS2 with or without tobacco acid pyrophosphatase (TAP) treatment as indicated. The asterisks denote sequenced PCR products that mapped to the negative strand of the MYC locus.

(C) PCR products from 3’ RACE assays for MYC-AS1 and MYC-AS2 with or without in vitro polyA tailing of the RNA as indicated. The asterisks denote sequenced PCR products mapped to the negative strand of the MYC locus.

(D) Relative expression of MYC-AS1 (left), MYC-AS2 (middle), and MYC mRNA (right) transcripts determined by strand-specific RT-qPCR assays in breast cancer cell lines treated with E2 for the indicated time. MCF-10A: normal mammary epithelial cells; MCF-7: ERα-positive breast cancer cells; and MDA-MB-231: ERα-negative breast cancer cells. Each bar represents the mean ± SEM. The asterisks indicate significant differences from the corresponding control (E2, 0h) (Student’s t-test, * p-value < 0.05; ** p-value < 0.005).

To examine the kinetics of MYC-AS1 and MYC-AS2 transcription by E2 in MCF-7 cells, we analyzed the steady-state transcript levels of MYC-AS1, MYC-AS2, and MYC mRNA using RT-qPCR over a time course of E2 treatment (Fig. 6D). As controls, we included similar analyses using RNA isolated from the human MCF-10A normal-like breast epithelial cell line and the ER-negative MDA-MB-231 breast cancer cell line, both of which should not exhibit E2 induced expression of the MYC-AS1, MYC-AS2, or MYC RNAs. MYC-AS1 and MYC-AS2 transcript levels peaked one hour after E2 treatment and returned to basal levels three hours after treatment, mirroring the induction of MYC mRNA (Fig. 6D). As expected, we did not observe the induction of MYC-AS1, MYC-AS2, or MYC mRNA in either MCF-10A or MDA-MB-231 cells (Fig. 6D). Based on the stability metric described above, MYC-AS1 and MYC-AS2 are both less stable than MYC mRNA (~8-fold less stable). Collectively, the results demonstrate that the estrogen-regulated MYC antisense transcripts are transcribed with similar kinetics as the MYC mRNA transcript.

To further determine the localization of MYC-AS1 and MYC-AS2, we mined our previous data from subcellular RNA-seq in MCF-7 cells treated with or without E2 for 3 hours, corresponding to fractions of RNA from chromatin, the nucleus, and the cytoplasm (Sun et al., 2015). Browser track representations of the MYC locus show the induction of the MYC-AS1, MYC-AS2, and MYC RNAs by estrogen in all three fractions, with the greatest abundance in the nuclear fraction (Fig. 7A). Next, we examined the kinetics of MYC-AS1 and MYC-AS2 induction in response to E2 treatment in the nuclear and cytosolic fractions. As controls, we used the lncRNA MALAT1 as a nuclear fraction marker and ACTB mRNA as a cytosolic fraction marker. Similar to the results from our RT-qPCR analysis using total RNA (Fig. 6D), MYC-AS1 and MYC-AS2 transcript levels both peaked at 1 hour of E2 treatment in the cytosolic fraction, but not in the nuclear fraction, although they were detectable in the latter (Fig. 7B). These results indicate that the MYC-AS1 and MYC-AS2 antisense transcripts are stable and localize to the cytoplasm in MCF-7 cells.

Fig. 7. Subcellular localization of MYC antisense transcripts.

Fig. 7.

Open in a new tab

(A) Genome-browser view of the MYC gene locus showing fractionated polyA+ RNA-seq data from MCF-7 cells treated with E2 for 3 hrs. RNA was isolated from cytoplasmic, nucleoplasmic, and chromatin fractions. Black arrows represent MYC antisense transcript annotations.

(B) Relative expression of the indicated transcripts in subcellular RNA fractions determined by RT-qPCR. MCF-7 cells were treated with E2 for the indicated time, and separated into cytosolic and nuclear fractions. RNA was isolated and the specific transcripts indicated were quantified by RT-qPCR. Each bar represents the mean ± SEM. The asterisks indicate significant differences from the corresponding control (Student’s t-test, * p-value < 0.05).

4. Discussion

While great strides have been made in understanding the biology of noncoding RNAs, the genomic organization of antisense genes and the production antisense RNAs remain poorly understood. In this study, we identified 6,173 antisense transcription units in MCF-7 human breast cancer cells. Genomic analyses revealed distinct classes of antisense transcription units relative to the sense gene TSSs, with particular genomic features associated with each class (i.e., recruitment of RNA Pol II, enrichment of H3K4me3, DNaseI hypersensitivity). Surprisingly, the sense genes from the different classes are associated with different biological processes and correlate with different levels of estrogen responsiveness; sense genes with antisense transcription units running through their TSS are associated with increased transcription and are highly induced by estrogen. Finally, molecular characterization of two antisense transcripts (MYC-AS1 and MYC-AS2) indicates that they are 5’-capped, 3’-polyadenylated, and can localize to both the nucleus and cytoplasm.

4.1. The functional architecture and biology of sense/antisense gene pairs

Grouping antisense transcription units based on their TTSs relative to their cognate sense gene TSSs revealed interesting and unexpected relationships. First, antisense transcription units running through a sense gene TSS (Class 1) exhibit the strongest correlation between sense and antisense transcription under basal conditions and after E2 treatment (Figs. 3B and 5A) compared to the other classes. Furthermore, sense genes associated with antisense transcription units in Class 1 are associated exhibit greater induction in response to E2 treatment than sense genes associated with the antisense transcription units in Classes 2 and 3 (Figs. 4A and 5B). Indeed, genomic features such as RNA Pol II enrichment, H3K4me3 enrichment, and DNase I hypersensitivity are all significantly higher at the promoters of sense genes associated with antisense transcription units in Class 1 (Fig. 4, B–D) and are more responsive to estrogen treatment (Fig. 5, C–E).

Gene ontology analyses revealed that the sense genes associated with each of the three different classes of antisense transcription units that we identified control distinct biological (Fig. 3C). Sense genes associated with Class 1 antisense transcription units, which are highly expressed and strongly up-regulated by estrogen, code for proteins involved in transcriptional regulation. Sense genes associated with Class 1 and Class 2 antisense transcription units, which are more modestly expressed and up-regulated by estrogen, code for proteins involved cytoskeleton function and organization (Fig. 3C). These results suggest that nature has devised different regulatory mechanisms controlling the expression of different groups of genes specifying distinct biological endpoints. In this case, the expression of sense genes coding for proteins involved in transcriptional regulation is enhanced by antisense transcription running through their promoters, which produces a more open chromatin architecture.

Previous studies have shown that antisense transcription can exhibit concordant (i.e., stimulatory) or discordant (i.e., inhibitory) regulation with respect to the cognate sense gene (Faghihi et al., 2009). This regulation may occur through a variety of mechanisms, some involving the act of transcription, with others involving a role for a stable antisense RNA (Faghihi et al., 2009; Lee, 2012; Pelechano et al., 2013). In our studies, antisense transcription running through a sense promoter is associated with enhanced transcription. In contrast, the transcription of a gene locus in both directions may promote the formation of RNA hybrids that can lead to transcriptional interference or induce gene silencing (Faghihi et al., 2009; Lee, 2012; Pelechano et al., 2013). Future analyses of how antisense transcription controls sense gene transcription will benefit from the addition of a kinetic component to better understand the complex regulatory circuitry of sense/antisense gene pairs, as well as its relationship to biological outcomes.

4.2. MYC-AS1 and MYC-AS2 are stable RNAs

Our genomic analyses and the identification of the Class 1 antisense transcription units suggest that the act of antisense transcription may have important regulatory effects on the expression of sense genes. But, as noted above, the production of stable antisense transcripts may also play a role in regulating the expression of sense genes and downstream biological processes (Faghihi et al., 2009). In this regard, we characterized two stable antisense transcripts originating from the MYC locus and showed that these transcripts are 5’-capped and 3’-polyadenylated (Fig. 6). Furthermore, a significant fraction of these transcripts localizes to the cytoplasm (Fig. 7), pointing to their potential to regulate not only transcription and nuclear processes, but also other cellular processes occurring in the cytoplasm. Indeed, many long non-coding RNAs have been shown to play a role in regulating translation and mRNA stability in the cytoplasm (Carrieri et al., 2012; Faghihi et al., 2008; Su et al., 2012). The specific molecular functions of MYC-AS1 and MYC-AS2 remain to be determined in future functional assays.

5. Conclusions

Using genomic data, we have defined three classes of antisense transcription in MCF-7 cells based on where their antisense transcription termination sites reside relative to their cognate sense mRNA and lncRNA genes. These three classes differ in their response to estrogen treatment, in the enrichment of a number of genomic features associated with active promoters (H3K4me3, RNA polymerase II, open chromatin architecture), and in the biological processes of sense genes. Molecular analyses of two estrogen-regulated antisense transcripts arising from the MYC gene in MCF-7 cells demonstrated that these transcripts are five-prime capped, three-prime polyadenylated, and localized to different compartments of the cell. Together, our analyses have identified distinct classes of antisense transcription correlated to different biological processes and response to estrogen stimulation, uncovering another layer of hormone-regulated gene regulation. Overall, the results from our characterization of estrogen-regulated antisense transcription in human breast cancer cells supports for a role of antisense transcription in estrogen-regulated gene transcription and breast cancer biology (Balbin et al., 2015). Since GRO-seq detects transcription for all three RNA polymerases, it is possible that RNAPI and RNAPIII could mediate antisense transcription. Thus, in future studies, it will be interesting to categorize the different classes of antisense transcription/transcripts associated with the different RNA polymerases. In addition, it will be also be interesting to elucidate the biochemical and molecular mechanisms by which sense and antisense transcription/transcripts are regulated by estrogen to drive breast cancer biology.

Supplementary Material

1

Table S1. Genomic coordinates of antisense transcription units called by groHMM in MCF-7 breast cancer cells. GRO-seq data obtained from MCF-7 breast cancer cells were analyzed using the groHMM analysis tool. Antisense transcription units were grouped based on antisense transcription termination sites relative to sense gene transcription start sites. The “Contents” provides details about all of the other worksheets contained within. The “Key” provides annotation information describing the antisense transcription units and sense genes.

  • Defined three classes of antisense transcription in ER+ MCF-7 cells.

  • Found that antisense transcription is enriched with active promoter marks.

  • Identified distinct biological processes for three classes of antisense transcription.

  • Characterized two estrogen-regulated antisense transcripts from the MYC gene.

ACKNOWLEDGEMENTS

The authors would like to thank members of the Kraus lab for helpful feedback during the course of this study and Venkat Malladi for assistance with the figures. This work was supported by a grant from the NIH/NIDDK (DK058110), grants from the Cancer Prevention and Research Institute of Texas (CPRIT; RP160318 and RP190235), and funds from the Cecil H. and Ida Green Center for Reproductive Biology Sciences Endowment to W.L.K.

Footnotes

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

References

  1. Balbin OA, Malik R, Dhanasekaran SM, Prensner JR, Cao X, Wu YM, Robinson D, Wang R, Chen G, Beer DG et al. , 2015. The landscape of antisense gene expression in human cancers, Genome Res. 25, 1068–1079. [DOI] [PMC free article] [PubMed] [Google Scholar]
  2. Carninci P, Kasukawa T, Katayama S, Gough J, Frith MC, Maeda N, Oyama R, Ravasi T, Lenhard B, Wells C et al. , 2005. The transcriptional landscape of the mammalian genome, Science. 309, 1559–1563. [DOI] [PubMed] [Google Scholar]
  3. Carrieri C, Cimatti L, Biagioli M, Beugnet A, Zucchelli S, Fedele S, Pesce E, Ferrer I, Collavin L, Santoro C et al. , 2012. Long non-coding antisense RNA controls Uchl1 translation through an embedded SINEB2 repeat, Nature. 491, 454–457. [DOI] [PubMed] [Google Scholar]
  4. Chae M, Danko CG and Kraus WL, 2015. groHMM: a computational tool for identifying unannotated and cell type-specific transcription units from global run-on sequencing data, BMC Bioinformatics. 16, 222. [DOI] [PMC free article] [PubMed] [Google Scholar]
  5. Crampton N, Bonass WA, Kirkham J, Rivetti C and Thomson NH, 2006. Collision events between RNA polymerases in convergent transcription studied by atomic force microscopy, Nucleic Acids Res. 34, 5416–5425. [DOI] [PMC free article] [PubMed] [Google Scholar]
  6. Danko CG, Chae M, Martins A and Kraus WL, 2014. groHMM: GRO-seq Analysis Pipeline, Bioconductor. Bioconductor, http://bioconductor.org/packages/release/bioc/html/groHMM.html. [Google Scholar]
  7. Djebali S, Davis CA, Merkel A, Dobin A, Lassmann T, Mortazavi A, Tanzer A, Lagarde J, Lin W, Schlesinger F et al. , 2012. Landscape of transcription in human cells, Nature. 489, 101–108. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Faghihi MA, Modarresi F, Khalil AM, Wood DE, Sahagan BG, Morgan TE, Finch CE, St Laurent G 3rd, Kenny PJ and Wahlestedt C, 2008. Expression of a noncoding RNA is elevated in Alzheimer’s disease and drives rapid feed-forward regulation of beta-secretase, Nat Med. 14, 723–730. [DOI] [PMC free article] [PubMed] [Google Scholar]
  9. Faghihi MA and Wahlestedt C, 2009. Regulatory roles of natural antisense transcripts, Nat Rev Mol Cell Biol. 10, 637–643. [DOI] [PMC free article] [PubMed] [Google Scholar]
  10. Faghihi MA, Zhang M, Huang J, Modarresi F, Van der Brug MP, Nalls MA, Cookson MR, St-Laurent G 3rd and Wahlestedt C, 2010. Evidence for natural antisense transcript-mediated inhibition of microRNA function, Genome Biol. 11, R56. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Franco HL, Nagari A and Kraus WL, 2015. TNFalpha signaling exposes latent estrogen receptor binding sites to alter the breast cancer cell transcriptome, Mol Cell. 58, 21–34. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Franco HL, Nagari A, Malladi VS, Li W, Xi Y, Richardson D, Allton KL, Tanaka K, Li J, Murakami S et al. , 2018. Enhancer transcription reveals subtype-specific gene expression programs controlling breast cancer pathogenesis, Genome Res. 28, 159–170. [DOI] [PMC free article] [PubMed] [Google Scholar]
  13. Hah N, Danko CG, Core L, Waterfall JJ, Siepel A, Lis JT and Kraus WL, 2011. A rapid, extensive, and transient transcriptional response to estrogen signaling in breast cancer cells, Cell. 145, 622–634. [DOI] [PMC free article] [PubMed] [Google Scholar]
  14. Hah N, Murakami S, Nagari A, Danko CG and Kraus WL, 2013. Enhancer transcripts mark active estrogen receptor binding sites, Genome Res. 23, 1210–1223. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Hewitt SC and Korach KS, 2018. Estrogen Receptors: New Directions in the New Millennium, Endocr Rev. 39, 664–675. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Hobson DJ, Wei W, Steinmetz LM and Svejstrup JQ, 2012. RNA polymerase II collision interrupts convergent transcription, Mol Cell. 48, 365–374. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Huang DW, Sherman BT, Tan Q, Collins JR, Alvord WG, Roayaei J, Stephens R, Baseler MW, Lane HC and Lempicki RA, 2007. The DAVID Gene Functional Classification Tool: a novel biological module-centric algorithm to functionally analyze large gene lists, Genome Biol. 8, R183. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Huang DW, Sherman BT and Lempicki RA, 2009. Systematic and integrative analysis of large gene lists using DAVID bioinformatics resources, Nat Protoc. 4, 44–57. [DOI] [PubMed] [Google Scholar]
  19. Kapranov P, Cheng J, Dike S, Nix DA, Duttagupta R, Willingham AT, Stadler PF, Hertel J, Hackermuller J, Hofacker IL et al. , 2007. RNA maps reveal new RNA classes and a possible function for pervasive transcription, Science. 316, 1484–1488. [DOI] [PubMed] [Google Scholar]
  20. Katayama S, Tomaru Y, Kasukawa T, Waki K, Nakanishi M, Nakamura M, Nishida H, Yap CC, Suzuki M, Kawai J et al. , 2005. Antisense transcription in the mammalian transcriptome, Science. 309, 1564–1566. [DOI] [PubMed] [Google Scholar]
  21. Langmead B, Trapnell C, Pop M and Salzberg SL, 2009. Ultrafast and memory-efficient alignment of short DNA sequences to the human genome, Genome Biol. 10, R25. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Lee JT, 2012. Epigenetic regulation by long noncoding RNAs, Science. 338, 1435–1439. [DOI] [PubMed] [Google Scholar]
  23. Li H, Handsaker B, Wysoker A, Fennell T, Ruan J, Homer N, Marth G, Abecasis G, Durbin R and Subgroup GPDP, 2009. The Sequence Alignment/Map format and SAMtools, Bioinformatics. 25, 2078–2079. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Li R, Yu C, Li Y, Lam TW, Yiu SM, Kristiansen K and Wang J, 2009. SOAP2: an improved ultrafast tool for short read alignment, Bioinformatics. 25, 1966–1967. [DOI] [PubMed] [Google Scholar]
  25. Luo X, Chae M, Krishnakumar R, Danko CG and Kraus WL, 2014. Dynamic reorganization of the AC16 cardiomyocyte transcriptome in response to TNFalpha signaling revealed by integrated genomic analyses, BMC Genomics. 15, 155. [DOI] [PMC free article] [PubMed] [Google Scholar]
  26. Murakami S, Nagari A and Kraus WL, 2017. Dynamic assembly and activation of estrogen receptor alpha enhancers through coregulator switching, Genes Dev. 31, 1535–1548. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Ozsolak F, Kapranov P, Foissac S, Kim SW, Fishilevich E, Monaghan AP, John B and Milos PM, 2010. Comprehensive polyadenylation site maps in yeast and human reveal pervasive alternative polyadenylation, Cell. 143, 1018–1029. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Pelechano V and Steinmetz LM, 2013. Gene regulation by antisense transcription, Nat Rev Genet. 14, 880–893. [DOI] [PubMed] [Google Scholar]
  29. Quinlan AR and Hall IM, 2010. BEDTools: a flexible suite of utilities for comparing genomic features, Bioinformatics. 26, 841–842. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Su WY, Li JT, Cui Y, Hong J, Du W, Wang YC, Lin YW, Xiong H, Wang JL, Kong X et al. , 2012. Bidirectional regulation between WDR83 and its natural antisense transcript DHPS in gastric cancer, Cell Res. 22, 1374–1389. [DOI] [PMC free article] [PubMed] [Google Scholar]
  31. Sun M, Gadad SS, Kim DS and Kraus WL, 2015. Discovery, annotation, and functional analysis of long noncoding RNAs controlling cell-cycle gene expression and proliferation in breast cancer cells, Mol Cell. 59, 698–711. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Trapnell C, Pachter L and Salzberg SL, 2009. TopHat: discovering splice junctions with RNA-Seq, Bioinformatics. 25, 1105–1111. [DOI] [PMC free article] [PubMed] [Google Scholar]
  33. Tufarelli C, Stanley JA, Garrick D, Sharpe JA, Ayyub H, Wood WG and Higgs DR, 2003. Transcription of antisense RNA leading to gene silencing and methylation as a novel cause of human genetic disease, Nat Genet. 34, 157–165. [DOI] [PubMed] [Google Scholar]
  34. Wanowska E, Kubiak MR, Rosikiewicz W, Makalowska I and Szczesniak MW, 2018. Natural antisense transcripts in diseases: From modes of action to targeted therapies, Wiley Interdiscip Rev RNA. 9. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Werner A, 2013. Biological functions of natural antisense transcripts, BMC Biol. 11, 31. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. Yap KL, Li S, Munoz-Cabello AM, Raguz S, Zeng L, Mujtaba S, Gil J, Walsh MJ and Zhou MM, 2010. Molecular interplay of the noncoding RNA ANRIL and methylated histone H3 lysine 27 by polycomb CBX7 in transcriptional silencing of INK4a, Mol Cell. 38, 662–674. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Yu W, Gius D, Onyango P, Muldoon-Jacobs K, Karp J, Feinberg AP and Cui H, 2008. Epigenetic silencing of tumour suppressor gene p15 by its antisense RNA, Nature. 451, 202–206. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Zhong S, Joung JG, Zheng Y, Chen YR, Liu B, Shao Y, Xiang JZ, Fei Z and Giovannoni JJ, 2011. High-throughput illumina strand-specific RNA sequencing library preparation, Cold Spring Harb Protoc. 2011, 940–949. [DOI] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

1

Table S1. Genomic coordinates of antisense transcription units called by groHMM in MCF-7 breast cancer cells. GRO-seq data obtained from MCF-7 breast cancer cells were analyzed using the groHMM analysis tool. Antisense transcription units were grouped based on antisense transcription termination sites relative to sense gene transcription start sites. The “Contents” provides details about all of the other worksheets contained within. The “Key” provides annotation information describing the antisense transcription units and sense genes.

NIHMS1561704-supplement-1.xlsx (506.3KB, xlsx)

RESOURCES