PMCCPMCCPMCC

Search tips
Search criteria 

Advanced

 
Logo of plosonePLoS OneView this ArticleSubmit to PLoSGet E-mail AlertsContact UsPublic Library of Science (PLoS)
 
PLoS One. 2017; 12(7): e0181902.
Published online 2017 July 21. doi:  10.1371/journal.pone.0181902
PMCID: PMC5521836

EINCR1 is an EGF inducible lincRNA overexpressed in lung adenocarcinomas

Karol Nowicki-Osuch, Conceptualization, Investigation, Writing – original draft, Writing – review & editing,1,¤ Yaoyong Li, Investigation, Writing – review & editing,1 Mairi Challinor, Investigation,1 David T. Gerrard, Investigation, Writing – review & editing,2 Neil A. Hanley, Supervision, Writing – review & editing,2,3 and Andrew D. Sharrocks, Conceptualization, Funding acquisition, Project administration, Supervision, Writing – original draft, Writing – review & editing1,*
David Raul Francisco Carter, Editor

Abstract

Long non-coding RNAs are being increasingly recognised as important molecules involved in regulating a diverse array of biological functions. For example, many long non-coding RNAs have been associated with tumourigenesis and in this context their molecular functions often involves impacting on chromatin and transcriptional control processes. One important cellular control system that is often deregulated in cancer cells is the ERK MAP kinase pathway. Here we have investigated whether ERK pathway signaling in response to EGF stimulation, leads to changes in the production of long non-coding RNAs. We identify several different classes of EGF pathway-regulated lncRNAs. We focus on one of the inducible lincRNAs, EGF inducible long intergenic non-coding RNA 1 (EINCR1). EINCR1 is predominantly nuclear and shows delayed activation kinetics compared to other immediate-early EGF-inducible genes. In humans it is expressed in a tissue-specific manner and is mainly confined to the heart but it exhibits little evolutionary conservation. Importantly, in several cancers EINCR1 shows elevated expression levels which correlate with poor survival in lung adenocarcinoma patients. In the context of lung adenocarcinomas, EINCR1 expression is anti-correlated with the expression of several protein coding EGF-regulated genes. A potential functional connection is demonstrated as EINCR1 overexpression is shown to reduce the expression of EGF-regulated protein coding genes including FOS and FOSB.

Introduction

Over recent years, long noncoding RNAs (lncRNAs) have emerged as a large and diverse group of genes. The discovery of these genes arose due to the advent of the next generation sequencing technologies which indicated that up to 80% of human genome is transcribed [1]. LncRNAs are defined as transcripts that are over 200 nucleotides long and do not contain protein coding open reading frames [2]. Functional studies have implicated lncRNAs in many cellular processes including the regulation of transcription, RNA processing, siRNA attenuation and RNA stabilisation [3]–[6]. Over recent years, lncRNAs have become prominent elements of regulatory networks associated with cell growth, differentiation and cancer progression. Many lncRNAs have been associated with controlling gene expression (reviewed in [7]) with arguably the best studied lncRNA being MALAT1 and its role in cancer. For example, MALAT1 was recently shown to be a critical factor in breast cancer metastasis [8]. These findings have triggered other studies into the role of lncRNAs in cancer. For example RNA-seq experiments in prostate cancer cells identified 121 novel lncRNA associated with that disease [9]. Further investigation of the most upregulated lncRNA, PCAT1, indicated that it is predominantly repressive in nature and it influences expression of genes associated with mitosis and cell cycle. Recently, it has been shown that transcription of PCAT1 is regulated by a specific distal regulatory element (via ONECUT2 transcription factor) and that PCAT1 itself is regulates expression of androgen receptor late response genes in LSD1-dependent manner [10]. The identification of additional lncRNAs whose expression changes in disease states or in response to cellular signalling would therefore likely identify new potential regulatory molecules that are important in signal-dependent processes.

The EGF signaling pathway and its effects on protein coding gene expression has been extensively studied and represents an excellent model system in which to identify and investigate the function of lncRNAs. For example previous studies using the human MCF10A breast epithelial cell line have identified dynamic changes in the expression of protein coding genes and miRNAs following EGF induction [11]. Protein coding genes which are upregulated are typically from the immediate-early class of genes including well studied genes like FOS and EGR1. Many of the genes activated by EGF signaling encode transcription factors that propagate downstream transcriptional events. However, a number of feedback loops are triggered that act to turn off or moderate the levels of the initial transcriptional wave, including a role for miRNAs [12]. The immediate-early genes are also activated by different growth factors and other mitogenic signals through activation of the ERK MAP kinase pathway [13]. Importantly, EGF receptor tyrosine kinases and the downstream ERK pathway are commonly upregulated in cancer [14], meaning that the identification of novel downstream effector molecules is potentially important for enhancing our understanding of the tumourigenic process. Here we used RNA-seq to identify EGF-regulated lncRNAs and identify several different classes of lncRNA. We focus on one of the lincRNAs, EGF inducible long intergenic non-coding RNA 1 (EINCR1), and characterise its transcriptional activation mechanism, its subcellular localisation, its expression profile and potential role in controlling gene expression.

Results

Identification EGF-inducible lncRNAs

To identify lncRNAs that are inducible by EGF pathway signalling, we treated non-tumourigenic human breast epithelial MCF10A cells with EGF for 30 mins and performed RNA-seq. We focussed on the nuclear RNA fraction as many lncRNAs are known to be preferentially localised to this subcellular compartment [1]. We identified a total of 482 transcripts (>200 bp) that are upregulated or downregulated by EGF (log2 fold change >0.58, P-value <0.05), the majority of which are upregulated (Fig 1A; S1 Table). Well characterised protein coding genes such as FOS and EGR1 are among the upregulated genes although lncRNAs are also upregulated including one which we have named EINCR1.

Fig 1
EGF-stimulation leads to transcriptional induction of lncRNAs.

Next, we plotted the average sequencing reads found in a 2 kb region spanning the putative transcript transcriptional start sites (TSS) of either protein coding genes or lncRNAs. As expected, for upregulated protein coding genes we observed a general increase of downstream sequencing reads on the sense strand but little indication of significant activity or changes in transcriptional levels either upstream from the TSS or on the anti-sense strand (Fig 1B; left panel). In contrast, for upregulated lncRNAs, in addition to the expected increase in downstream sense strand transcription, we also observe a high level of sequencing reads upstream from the TSS on the sense strand (Fig 1B, right panel). This indicates the presence of upstream sense strand transcriptional activity in a large proportion of the lncRNAs. To further investigate this phenomenon we plotted the average read densities around the TSS of individual genes and subjected these to cluster analysis (Fig 1C). There was generally little evidence of transcriptional activity on the sense strand upstream from the TSS of protein coding genes. However, a large number of lncRNA transcripts showed evidence for upstream transcriptional activity on the sense strand of the lncRNA encoding genes (Fig 1C, bottom panel). This indicates the presence of upstream transcriptional units originating from coding genes as exemplified by the DUSP5 locus (Fig 1D, top panel). Here, both the DUSP5 locus and the downstream lncRNA are EGF-inducible. Little anti-sense activity was detected around the transcriptional start sites for both protein coding and lncRNA genes (Fig 1C, right panels). However, there are examples of lncRNAs that are closely positioned in the anti-sense direction downstream from protein coding genes as observed at the EDN1 locus. This gene is EGF-inducible, and at this locus, the downstream anti-sense lncRNA transcript (termed EINCR2) is also EGF inducible (Fig 1D, bottom panel).

We also examined the RNA levels surrounding genes which are downregulated following EGF treatment. Very few protein coding genes are downregulated but a large number of lncRNAs are downregulated (see S1B Fig). In contrast to the EGF-induced lncRNAs, there is evidence for large amounts of anti-sense transcription in the region upstream from the TSS of many EGF-downregulated lncRNAs, suggesting the existence of divergent promoters (S1A and S1B Fig). Collectively, this data indicates that many EGF-regulated lncRNAs form a subclass that is located immediately up or downstream from other transcriptionally active genes. In the case of the EGF-inducible lncRNAs, the existence of upstream sense strand transcription units is suggestive of polymerase run-ons from the upstream gene, a phenomenon that has recently been reported by others [15].

We focussed on the EGF upregulated transcripts and subclassified the lncRNAs according to known genomic annotations or by belonging to the “polymerase run-on” or “anti-sense” classes identified above (Fig 1E).173 of these transcripts (55%) are non-coding RNAs, and of these 67 (21%) are known pseudogenes or are derived from “polymerase run-ons”. The rest of the lncRNAs can be subdivided into either antisense or intronic lncRNAs or long intergenic non-coding RNAs (lincRNA). We identified 13 known lincRNAs and 76 novel lincRNAs. To validate these findings, we performed RT-qPCR on a selection of EGF up- and down-regulated lincRNAs and observed the expected patterns of regulation (Fig 1F). The kinetics of activation and repression varied across a 6 hour time course depending on the lincRNA studied. For example, XLOC-036745 is rapidly induced and rapidly returns to basal levels within 1 hour of stimulation whereas XLOC-014669 (EINCR1) shows more sustained activation beyond 90 mins.

We have therefore identified a large number of lncRNAs that are either up or downregulated following EGF stimulation that can be subcategorised into different classes. Due to its rapid and sustained induction kinetics and also its genomic location which is clearly distinct from any neighbouring genes, we decided to focus on characterising one of the lincRNAs EINCR1.

EINCR1 is an EGF regulated lincRNA

EINCR1 is located >50 kb from the nearest annotated coding gene, indicating that it should be classified as a long intergenic non-coding RNA (lincRNA). To establish whether EINCR1 is a true non-coding RNA, we examined its coding potential using PhyloCSF [16]. EINCR1 scored as a non-coding RNA like XIST rather than a protein coding gene such as GAPDH (Fig 2A). Next we asked whether EINCR1 is associated with the ribosomal fraction, and hence its likelihood for undergoing translation. In the presence of cycloheximide, there was little difference between EINCR1 levels and GAPDH levels associated with the ribosomal fraction (Fig 2B, top). However, upon treatment with EDTA, GAPDH was released from the ribosome as expected for an RNA undergoing translation upon ribosomal subunit dissociation, whereas the relative levels of EINCR1 increased, suggesting that any association was likely due to contamination during the fractionation process, rather than undergoing active translation. These results therefore indicate that EINCR1 is a non-coding RNA but we cannot rule out that under some circumstances, short peptides might be generated from the EINCR1 transcript.

Fig 2
EINCR1 is an EGF-regulated, nuclear long intergenic non-coding RNA.

Having established that EINCR1 is likely a non-coding RNA, we next investigated its subcellular localisation. We isolated RNA from the nuclear and cytoplasmic compartments and verified their validity by demonstrating an enrichment of U4 snRNA in the nucleus and 18S rRNA in the cytoplasm (Fig 2C). As expected, the mRNA from protein coding gene is enriched in the cytoplasm. However, EINCR1 exhibits reciprocal enrichment and was found predominantly in the nuclear fraction (Fig 2C). This nuclear localisation was verified by analysing the expression of EINCR1 in published RNAseq data from the MCF7 breast cancer cell line (Fig 2D) and is consistent with a previous study that also showed that this lincRNA is localised to the nucleus in PTEC kidney cells [17]. Thus, in common with many lincRNAs, EINCR1 exhibits a predominantly nuclear localisation.

EINCR1 transcription is regulated via the ERK MAPK pathway

To examine EINCR1 expression in more detail we initially delineated the transcript structure. First we compared the extent of the transcription unit we had identified with the annotated transcripts in this genomic region. Two putative non-coding RNAs, RP11-7F17.7 and RP11-7F17.1 (also known as LINC01629 or linc-KIAA1737-2) had previously been documented in this region (S2A Fig). Our RNAseq data provide supporting data for the existence of two of the annotated exons, indicating that transcript splicing occurs and the EINCR1 transcription unit appears to span these two loci (S2A Fig). We confirmed the existence of the splice junction by RT-PCR and Sanger sequencing. Next we established the transcriptional start site (TSS) of the EINCR1 lincRNA by 5’ RACE. The start site we identified is located two bp upstream from the one identified for RP11-7F17.7 and corresponds exactly with the most common start site identified in CAGE data from cell lines examined by the FANTOM5 consortium (S2B Fig; [18]). The region surrounding this TSS is highly conserved amongst other primates and is also well conserved with rodents (Fig 3A). However, while the promoter region is conserved, the region further downstream from the TSS shows much lower conservation in rodents, suggesting a lack of functional conservation for the lincRNA (Fig 3A, middle). There is a putative TATA-like element which is specific to humans and is potentially associated with promoter functionality in this species. We also examined the histone marks which are associated with the locus in MCF7 cells and found evidence for H3K4me3 and H3K27ac which demarcate active chromatin regions but little sign of the repressive marks H3K9me3 or H3K27me3 (Fig 3A, bottom). However, the H3K4me3 peak is not located close to the TSS as commonly found in active genes and the H3K27ac mark is spread over a wide area rather than delineating a single promoter or enhancer unit. We also examined RNA polymerase II ChIP-seq data from EGF stimulated HeLa cells and found regions of enrichment around the TSS and H3K4me3 peak which increased in magnitude following EGF induction and spread across the entire locus (Fig 3A, bottom), consistent with the EGF-inducible transcription of EINCR1 that we observe in MCF10A cells. Finally we examined EINCR1 expression in two breast cancer cell lines and showed that, similar to FOS, it is inducible by EGF stimulation in MCF7 cells but not in MDA-MB-231 cells (S2C Fig).

Fig 3
EINCR1 transcription is regulated via the ERK MAPK pathway.

Next we asked whether EGF-mediated EINCR1 expression is mediated through the ERK MAP kinase signalling pathway. MCF10A cells were stimulated with EGF in the presence or absence of the MEK inhibitor U0126. EINCR1 was efficiently induced following EGF treatment for 15 mins but this response was abolished upon MEK inhibition (Fig 3B), indicating that EINCR1 is induced through ERK pathway signalling. Other EGF-inducible genes like FOS belong to the immediate-early class of genes and can be activated in the absence of new protein synthesis [19, 20]. We therefore tested whether EINCR1 expression is dependent on new protein synthesis by treating cells with cycloheximide prior to EGF stimulation. As expected, initial FOS induction proceeded as normal but expression was prolonged in the presence of cycloheximide due to the lack of feedback inhibition on its expression (Fig 3C, top). In contrast, although initially EINCR1 induction kinetics were unaffected, sustained EINCR1 message production is lost in the presence of cycloheximide, indicating that new protein synthesis is essential for the second phase of EINCR1 induction (Fig 3C, bottom). EGF stimulation leads to the production of new AP1 complex components such as FOS (reviewed in [13]) which might contribute to this later phase of EINCR1 induction. Indeed there are two evolutionarily conserved AP1 binding sites within the 250 bp region upstream from its TSS (S2D Fig). To establish whether AP1 might be playing a role in controlling EINCR1 expression, we transduced MCF10A cells with a lentivirus containing a dominant-negative version of FOS (A-FOS; [21]). Compared to control treated cells, the levels of EINCR1 were reduced following EGF treatment in the presence of A-FOS (Fig 3D).

Collectively, these data demonstrate that EINCR1 expression is activated by EGF through the ERK pathway and that the transcription factor AP1 plays a role in its activation.

EINCR1 expression patterns in tissues and cancer

To begin to understand the biological function of EINCR1, we first established its expression pattern in human tissues. In the adult, EINCR1 was generally expressed at a low level across the majority of tissues. However, high level expression is observed in the heart and to a moderate level in the bladder (Fig 4A). We also examined EINCR1 expression in the developing human embryo [22] and its expression is exclusively detected in the heart ventricle, alongside other heart-specific genes such as NKX2.5, NKX2.6 and TBX20 (Fig 4B). Therefore, under normal physiological conditions, EINCR1 likely functions in the context of the heart.

Fig 4
EINCR1 is specifically expressed in normal human heart tissues and is up-regulated in cancer.

To establish a potential disease link, we next looked for aberrant EINCR1 expression in cancer and interrogated 10,406 samples generated by the TCGA Research Network [23]. Little difference was seen for many cancers as illustrated by bladder cancer (Fig 4C). However, several other cancers showed enhanced expression of EINCR1 compared to normal tissue, as exemplified by the “bronchus and lung” category (Fig 4C). Expression of the lincRNA MALAT1 is shown for comparison, and although upregulation is seen in some cancers, there is little difference in the “bronchus and lung” group. Further sub-partitioning of the “bronchus and lung” category showed that EINCR1 is particularly overexpressed in several cancer subgroups, including squamous cell carcinomas and adenocarcinomas (S3A Fig). We further investigated this phenomenon by studying the expression of EINCR1 and other EGF-inducible lncRNAs in an independent dataset from lung adenocarcinomas [24]. The majority of EGF inducible protein coding genes show a general reduction in expression in these cancer samples (as exemplified by FOS) (S3B Fig), and the same was observed for EGF-inducible lncRNAs (S3C Fig). However, a subset of lncRNAs, including EINCR1, shows upregulation in lung adenocarcinomas. Given this strong association of EINCR1 expression with lung cancer, we investigated whether EINCR1 expression correlated with disease prognosis in lung adenocarcinoma [25]. High level EINCR1 expression gave a significantly worse disease prognosis with patients exhibiting high level expression showing lower long term survival (Fig 4D). As we initially identified EINCR1 in breast epithelial cells, we tested whether we could observe inducible expression in A549 lung adenocarcinoma cells. EGF treatment caused the expected transient induction of FOS in these cells and also a more sustained activation of EINCR1 (Fig 4E) in these lung adenocarcinoma cells.

Together these results demonstrate that during normal development, EINCR1 is expressed in a tissue-specific manner and is largely confined to the heart. However, EINCR1 is aberrantly expressed in several cancers, and in the case of lung adenocarcinoma, high level expression correlates with poor prognosis.

High level EINCR1 expression reduces EGF-regulated protein coding gene expression

We noticed that in lung adenocarcinoma samples, many EGF-inducible genes, including FOS and IER2 are downregulated whereas EINCR1 expression is generally upregulated (S3B and S3C Fig). This suggests a potentially reciprocal relationship between EINCR1 and other EGF-regulated genes. To further explore this phenomenon, we analysed the expression of EINCR1 and two EGF-inducible genes, FOS and FOSB across the lung adenocarcinoma samples in the TCGA-LUAD dataset. While EINCR1 is upregulated in the cancer samples, both FOS and FOSB show reduced expression in cancer samples (Fig 5A). When analysed on a sample by sample basis, we observed that when EINCR1 expression is high, then the expression of the coding gene is almost always low and vice versa (Fig 5B and 5C). There is therefore a reciprocal relationship between EINCR1 and the expression of EGF regulated protein coding genes and suggests that the high levels of EINCR1 might be responsible for dampening down their expression. To test this we used a CRISPR-derived system (Fig 5D; [26]) to drive constitutive overexpression of EINCR1 in MCF10A cells. Efficient EINCR1 overexpression was achieved (Fig 5E) and in the presence of enhanced EINCR1 levels, the expression of the EGF regulated genes, FOS, FOSB, IER2 and IER3 was dampened (Fig 5F–5I).

Fig 5
High levels of EINCR1 transcription leads to changes in the expression profiles of EGF-regulated genes.

Collectively these data demonstrate overexpression of EINCR1 leads to defects in the induction of EGF-activated protein coding genes and likely explains the reciprocal relationship between EINCR1 expression and the expression of genes like FOS and FOSB in cancer cells.

Discussion

It is now appreciated that lncRNAs are important regulatory molecules that play a role in controlling a variety of cellular activities [3]–[6]. Here we have identified a large number of lncRNAs whose expression is altered in response to signalling pathway activation following EGF treatment. These lncRNAs belong to different categories but the major class is lincRNAs, of which the majority are novel transcripts. One interesting additional class we identified corresponds to the previously identified “polymerase run-ons” which occurs at the 3’ ends of genes where the transcript extends beyond the usual termination site [15]. In the case of DUSP5, this could be explained by inefficient transcriptional termination in the face of high levels of EGF-induced transcription. However, there are several instances where the expression of the upstream coding region is unaffected but the “run-on” transcription is enhanced following EGF induction. It is not clear what mechanism is involved in lncRNA induction at these loci, nor is it clear what the functional significance of these events is.

We focussed on one of the EGF inducible lincRNAs, EINCR1. This lincRNA is rapidly induced following EGF induction but unlike typical EGF-inducible protein coding genes such as FOS, it exhibits extended activation kinetics. FOS is classified as an immediate-early gene due to the lack of requirement for protein synthesis for its induction and subsequent super induction following treatment with the protein synthesis inhibitor cycloheximide. However, while EINCR1 is still induced in the presence of cycloheximide, its prolonged activation is lost, indicating that it is not susceptible to the same negative feedback loops as FOS and requires new protein synthesis for its sustained activation. One such factor appears to be AP1 family transcription factors, several of which are inducible by EGF (including FOS, FOSB, JUN and JUNB). This therefore suggests an interesting regulatory network in which EGF-inducible protein coding genes like FOS cause increases in intracellular AP1 levels that enhance the expression of EINCR1 which in turn could act to dampen down any further activation of EGF-coding genes.

Our data indicate that EINCR1 is predominantly nuclear although ideally future studies should utilise additional approaches such as FISH to confirm this and detect potential subnuclear localisation. However its nuclear localisation suggests that EINCR1 might have a role in controlling gene expression either directly or indirectly through alterations to the underlying chromatin state. Our initial studies point to such a possibility as overexpression of EINCR1 leads to decreased expression of a cohort of EGF-induced protein coding genes such as FOS and FOSB. This observation is consistent with the observation that in several human cancer subtypes, EINCR1 is overexpressed, and in the case of lung adenocarcinomas its expression is anti-correlated with the expression of these genes. Thus, one function of EINCR1 might be to dampen down the levels of gene induction in response to EGF pathway activation. In humans, EINCR1 is predominantly expressed in the heart, and future studies are needed to study whether EINCR1 plays a role in modulating the output from receptor tyrosine kinase pathway signalling in this context. Although not highly expressed in the majority of tissues, our observation that EINCR1 expression is EGF inducible might reflect low basal level expression and yet the potential for activation under defined signalling conditions in a wider cellular context. Moreover, induction by other signalling events might also broaden its expression repertoire and a lincRNA transcribed from the same locus (KIAA1737-2), has been shown to be induced by cytokine treatment of human PTEC kidney cells [17]. EINCR1/ KIAA1737-2 might therefore by a commonly upregulated gene that responds to many different stimuli.

EINCR1 is upregulated in several different cancer subtypes, including lung adenocarcinomas. This is not a general phenomenon shared by all lincRNAs as MALAT1 is often associated with cancer progression (reviewed in [27]) but is only upregulated in a subset of the cancers where EINCR1 is overexpressed. It is not clear what role EINCR1 might have in the tumourigenic process but as EINCR1 is a target of the EGF pathway and EGFR-regulated pathways are often activated in cancer cells it appears likely that it might be an important mediator of dysregulated signalling. The reciprocal expression of EINCR1 and other EGF-regulated protein coding genes suggests a functional relationship which is reinforced by our over-expression studies. Additional work will be needed though to investigate the relevance of this relationship in the context of cancer.

In summary, we have identified a large number of EGF inducible lincRNAs and shown that one of these, EINCR1, is potentially important in the context of human cancer. Our findings provide a useful resource for further investigation of both EINCR1 and the dozens of other lncRNAs we have identified in modulating the output of EGF pathway signalling.

Materials and methods

Cell culture

MCF10A cells were grown in DMEM/F12 (Gibco, 11320–033) containing 5% horse serum (Biosera, DH291), 20 ng/ml EGF (Sigma, E1257), 10 μg/ml insulin (Sigma, I0516), 100 ng/ml cholera toxin (Sigma, C9903) and 0.5 μg/ml hydrocortisone (Sigma, H0396) (complete medium).

A549, MCF7, MDA-MD-231 and HEK293T cells were grown in either RPMI 1640 with L-glutamine, 25mM HEPES (for A549 cells)(Life Technologies, 52400041) or DMEM (Gibco, 22320–22) supplemented with 10% fetal bovine serum (Life Technologies, 10270–098). All of the cells were grown up to 90% confluence and were passaged every 2–3 days by brief washing of the cells with Dulbecco's phosphate-buffered saline (DPBS, Life Technologies, 14190–094), detached with Trypsin-EDTA (0.05%) (Life Technologies 25300–054) and replated at 1 in 5 dilution in full media. Where indicated, the MEK inhibitor U0126 was used at a final concentration of 10 μM in 0.1% DMSO and the protein synthesis inhibitor cycloheximide at 25 μg/ml. All cell lines were obtained from ATCC except A549 cells which were a gift from Caroline Ridley. We verified the identity of the genotypes of all the cell lines we used.

Nuclear RNA-seq

1.5 x 106 MCF10A cells were seeded onto a 60 mm dish in complete media without EGF and with 0.5% horse serum instead of 5% (see section 2.1). After 48 hrs incubation, EGF was added to a final concentration of 20 ng/ml. After 30 minutes, nuclear fractions were isolated according to the protocol described previously [1]. At the same time, nuclear fractions were isolated from cells which were not stimulated with EGF. Briefly, cells were washed with 1xPBS and then incubated for 10 min with 1 ml of RLN buffer (50 mM TrisHCl pH = 8.0, 140 mM NaCl, 1.5 mM MgCl2, 0.5% (v/v) IGEPAL CA-630 and 2 U/ml SUPERaseIn—RNase inhibitor, Invitrogen AM2696). Subsequently, the lysate was centrifuged at 1,000 g for 5 min at 4°C, the pellet was collected, resuspended in 1 ml of RLN buffer, incubated for a further 5 min on ice and centrifuged at 1 000 g for 5 min at 4°C. At this stage nuclear fractions from different experimental days were stored at -80°C for subsequent RNA extraction. The RNA was extracted from the nuclear fraction using an RNeasy plus kit (Qiagen, 74134) with DNase treatment according to the manufacturer’s protocol. After isolation, the purity and integrity of RNA were investigated using a Bioanalyser. Only non-degraded samples with a ratio of 260/280 nm absorption above 2 were used for sequencing library preparation.

The cDNA libraries were prepared TruSeq Stranded mRNA Sample Prep Kit (Illumina, RS-122-2101) for polyadenylated-tailed RNA sequencing. The preparations of libraries together with subsequent sequencing reactions were performed at the Genomic Technologies Facility, University of Manchester according to the manufacturer’s recommendations. The sequencing reaction was performed on the Illumina HiSeq 2500. Data are deposited in ArrayExpress (E-MTAB-5370).

RNA extraction and RT-qPCR

RNA was extracted using the RNeasy plus kit (Qiagen, 74134) following the manufacturer’s protocol. Subsequently, RNA samples were quantified using a Nanodrop 2000 (Thermo Scientific) and concentrations were normalized to 20 ng/μl. 40 ng of each sample was used per reverse transcription-quantitative polymerase chain reaction (RT-qPCR) reaction using the QuantiTect SYBR® Green RT-PCR Kit (Qiagen, 204243) on Rotor-Gene Q (Qiagen) real-time PCR machine.

When indicated, the nanolitre volume RT-qPCR was performed using the Fluidigm Biomark HD system using EvaGreen chemistry. The reactions were performed by Claire Morrisroe at The Genomic Facility, University of Manchester following manufacturer’s protocol. The output data were processed following the default quality protocol. Data points with more than one peak in the melt analysis were discarded.

The final results are normalised to the house keeping genes using the delta delta Ct method [28]. S2 Table contains the list of primers used.

Ribosome analysis—Sucrose cushion

MCF7 cells were plated onto 150 mm petri dishes at the density required for ~70% confluency on the assay day (around 5 x 106 cells). On the day of experiment, cells were stimulated with EGF for the indicated times. Cells were incubated for 5 minutes before lysate collection with 100 μg/ml of cycloheximide (CHX, Sigma, C7698, protein synthesis inhibitor that ‘freezes’ the actively translated mRNA with ribosomes and nascent protein). Subsequently, dishes were transferred onto ice/water bath, media was removed and cells were washed with 1x PBS containing 100 μg/ml of CHX. Cells were gently scraped into 1 ml of 1x PBS and centrifuged at 200 x g for 5 minute at 4°C. Subsequently, cells were resuspended in 200 μl of CSB buffer (300 mM sorbitol, 20 mM HEPES, pH 7.5, 1 mM EGTA, 5 mM MgCl2, 10 mM KCl, 10% Glycerol, 100 μg/ml CHX and protease inhibitor cocktail) and the cell membrane was disrupted with glass beads for 45 s. The lysate was cleared by centrifugation at 9,300 g for 10 min. The Optical density (OD) at 260 nm was measured and lysates were normalized for OD. For EDTA treatment, EDTA was added to a final concentration of 50 mM and lysates were overlaid over 60% sucrose solution in CSB buffer (without Sorbitol and protease inhibitors). Samples were centrifuged for 150 min at 4°C at 55,000 g. RNA was collected from the pellet and supernatant fractions (supplemented with luciferase RNA (Promega, L4561)–normalization spike-in) using trizol LS reagent (Life technologies, 10296). RNA levels were quantified with RT-qPCR and normalized to the input.

Lentiviral transductions

Second generation lentiviral particles were produced as described previously [29] using psPAX2 and pMD2.G packaging plasmids (Addgene 12260 and 12259). The viral particles were concentrated using PEG-it solution (System Biosciences, LV825A-1) and quantified using a qPCR method [30].

Transductions of the human cell lines with concentrated viral particles were performed in 6-well plates. Firstly, cells were plated at ~40% confluency in the media appropriate for the cell line. After 8 hrs, growth media was changed to the fresh growth media supplemented with 10 μg/ml of polybrene (Millipore, TR-1003-G) and the appropriate amount of the viral particles. The following day, the media was changed to the assay media required for the subsequent analysis.

CRISPR methods

Guide target sequences were designed following the protocol described previously [31]. For the overexpression studies, we used the dCas9-VP64 and MS2-p65 system [26]. Briefly, DNA duplexes corresponding to the guide RNAs were cloned into the BsmBI site of the lenti sgRNA(MS2)_zeo plasmid (Addgene 61427), to give the following plasmid; pAS4516 (containing guide sequence 5-CACCGTTTATCCCAGCATGAGGCG-3’ targeting region 218 nucleotides upstream of the EINCR1 TSS). Guide RNAs were designed using E-CRISPR (e-crispr.org; [31]) using the strict setting. sgRNA(MS2)_zeo plasmid encoding scrambled guide was use as a control (pAS4519, 5’-CACCGTGGTTTACATGTCGACTAA-3’). iral particles were produced for the sgRNA(MS2) plasmids and for lenti MS2-P65-HSF1_Hygro (Addgene 61426) and lenti dCAS-VP64_Blast (Addgene 61425) and the target cell lines were sequentially transduced with these plasmids. Gene expression analysis was performed using RT-qPCR.

Computational data analysis

For the nuclear RNA-seq data, raw sequencing files (fastq format) were investigated for the quality of reads, duplication level and GC content using fastqc packages [32]. Subsequently, raw reads were processed with the Trimmomatic tool which removed low quality reads and contamination with sequencing adapters [33]. The trimmed sequencing reads were aligned to the Ensembl transcription (release 72, 2013-03-06) and human genome version 19 (hg19) using RNA-Star aligner (version 2.3.0e) [34]. The aligner was run with default settings with option:—outFilterMultimapNmax 100, which allows for up to 100 genomic locations per read. Subsequently, reads that were mapped to multiple loci and redundant reads were removed using Pickard tools (http://broadinstitute.github.io/picard/). De novo transcriptome assembly was performed with Cufflinks package version 2.2.1 and it was run with default setting. Transcripts were identified using Cuffmerge packages [35, 36] with default settings and merged with the GENCODE (v. 19) transcriptome. Differential analysis was performed with Cuffdiff package where the transcriptome assembled by the Cuffmerge package was used as a reference. Subsequently, data quality was analysed using CummeRbund package version 2.0.0 [36]. All of the above calculations were performed using the computational shared facility at the University of Manchester. Subsequent data analysis was performed using custom written Perl scripts and R programming language and data were also processed in the MS Excel software. TiBCO Spotfire software was used for data visualisation.

For the RNA-seq data downloaded from gastric adenocarcinomas [24], the same protocol was applied. The expression of EINCR1 in normal human tissues was analysed in data downloaded from http://www.gtexportal.org/ (version 6p; [37]). The annotated lncRNA encompassing part of EINCR1 (RP11-7F17.7) was used in the search. To study EINCR1 expression during human organogenesis, embryonic RNA-seq data [22] were mapped to hg38 using STAR [34] to generate gene level (GENCODE 23) read counts. After quantile normalisation of all genes {R package preprocessCore}, selected genes were row-normalised to the maximum observed expression.

In order to obtain the EINCR1 expression in the samples in RNA-seq data generated by the TCGA Research Network (http://cancergenome.nih.gov/), we first downloaded the bam files of the RNA-seq data from the data portal of the NCI Genomic Data Commons (GDC) [23]. Then from each bam file we counted the number of reads overlapping with the genomic region occupied by EINCR1 and used the number of reads per 1kb exonic genomic region per 10M reads (RPK10M) as the EINCR1 expression in the corresponding sample. The expression of other genes such as FOS, FOSB and MALAT1 were also obtained in the same way.

The RNA level in all of the RNA-seq based experiments is presented as fragments per kilobase per million reads (FPKM) or per 10 million reads (FPK10M).

Supporting information

S1 Fig

EGF-stimulation leads to transcriptional downregulation of lncRNAs.

(PDF)

S2 Fig

Gene structure of EINCR1.

(PDF)

S3 Fig

EINCR1 is up-regulated in lung cancer samples.

(PDF)

S1 Table

RNAseq data from EGF stimulated MCF10A cells.

(XLSX)

S2 Table

List of primers used.

(XLSX)

Acknowledgments

We thank Karren Palmer for excellent technical assistance; Shen-Hsi Yang and members of our laboratory for comments on the manuscript and stimulating discussions. We thank Caroline Ridley for providing the A549 cells, Chris Kershaw and Mark Ashe for advice with the ribosome profiling experiment, Mike Jackson from the Flow Cytometry facility, Stacey Holden, Claire Morrisroe and Leo Zeef from the core Genomics Technology and Bioinformatics facilities. We would like to acknowledge the assistance given by Simon Hood in IT Services and for the use of the Computational Shared Facility at the University of Manchester We also thank Charles Vinson for reagents (DN-FOS).

Funding Statement

This work was supported by grants from the Wellcome Trust to ADS and to NAH (WT088566MA), including the Infrastructural Strategic Support Fund (WT097820). The funder had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Data Availability

Data Availability

RNAseq data are deposited in ArrayExpress (E-MTAB-5370).

References

1. Djebali S, Davis CA, Merkel A, Dobin A, Lassmann T, Mortazavi A, et al. (2012). Landscape of transcription in human cells. Nature 489: 101–108. doi: 10.1038/nature11233 [PMC free article] [PubMed]
2. Morris KV, Mattick JS. (2014) The rise of regulatory RNA. Nat Rev Genet. 15: 423–437. doi: 10.1038/nrg3722 [PMC free article] [PubMed]
3. Cech TR, Steitz JA. (2014). The Noncoding RNA Revolution—Trashing Old Rules to Forge New Ones. Cell 157: 77–94. doi: 10.1016/j.cell.2014.03.008 [PubMed]
4. Fatica A, Bozzoni I. (2013). Long non-coding RNAs: new players in cell differentiation and development. Nat. Rev. Genet. 15: 7–21. doi: 10.1038/nrg3606 [PubMed]
5. Holoch D, Moazed D. (2015). RNA-mediated epigenetic regulation of gene expression. Nat. Rev. Genet. 16: 71–84. doi: 10.1038/nrg3863 [PMC free article] [PubMed]
6. Yang L, Froberg JE, Lee JT. (2014) Long noncoding RNAs: fresh perspectives into the RNA world. Trends Biochem. Sci. 39: 35–43. doi: 10.1016/j.tibs.2013.10.002 [PMC free article] [PubMed]
7. Geisler S, Coller J. (2013) RNA in unexpected places: long non-coding RNA functions in diverse cellular contexts. Nat Rev Mol Cell Biol. 14: 699–712. doi: 10.1038/nrm3679 [PMC free article] [PubMed]
8. Arun G, Diermeier S, Akerman M, Chang KC, Wilkinson JE, Hearn S, et al. (2016) Differentiation of mammary tumors and reduction in metastasis upon Malat1 lncRNA loss. Genes Dev. 30: 34–51. doi: 10.1101/gad.270959.115 [PubMed]
9. Prensner JR, Iyer MK, Balbin OA, Dhanasekaran SM, Cao Q, Brenner JC, et al. (2011). Transcriptome sequencing across a prostate cancer cohort identifies PCAT-1, an unannotated lincRNA implicated in disease progression. Nat. Biotechnol. 29: 742–749. doi: 10.1038/nbt.1914 [PMC free article] [PubMed]
10. Guo H, Ahmed M, Zhang F, Yao CQ, Li S, Liang Y, et al. (2016) Modulation of long noncoding RNAs by risk SNPs underlying genetic predispositions to prostate cancer. Nature Genetics. 48: 1142–1150. doi: 10.1038/ng.3637 [PubMed]
11. Avraham R, Sas-Chen A, Manor O, Steinfeld I, Shalgi R, Tarcic G, et al. (2010). EGF decreases the abundance of microRNAs that restrain oncogenic transcription factors. Sci. Signal. 3: ra43 doi: 10.1126/scisignal.2000876 [PubMed]
12. Aitken S, Magi S, Alhendi AM, Itoh M, Kawaji H, Lassmann T, et al. (2015) Transcriptional dynamics reveal critical roles for non-coding RNAs in the immediate-early response. PLoS Comput Biol. 11: e1004217 doi: 10.1371/journal.pcbi.1004217 [PMC free article] [PubMed]
13. O'Donnell A, Odrowaz Z, Sharrocks AD. (2012) Immediate-early gene activation by the MAPK pathways: what do and don't we know? Biochem Soc Trans. 40: 58–66. doi: 10.1042/BST20110636 [PubMed]
14. Yarden Y, Pines G. (2012) The ERBB network: at last, cancer therapy meets systems biology. Nat Rev Cancer. 12: 553–563. doi: 10.1038/nrc3309 [PubMed]
15. Vilborg A, Passarelli MC, Yario TA, Tycowski KT, Steitz JA. (2015). Widespread Inducible Transcription Downstream of Human Genes. Mol. Cell 59: 449–461. doi: 10.1016/j.molcel.2015.06.016 [PMC free article] [PubMed]
16. Lin MF, Jungreis I, Kellis M. (2011) PhyloCSF: a comparative genomics method to distinguish protein coding and non-coding regions. Bioinformatics. 27: i275–282. doi: 10.1093/bioinformatics/btr209 [PMC free article] [PubMed]
17. Lin J, Zhang X, Xue C, Zhang H, Shashaty MG, Gosai SJ, et al. (2015) The long noncoding RNA landscape in hypoxic and inflammatory renal epithelial injury. Am J Physiol Renal Physiol. 309: F901–13. doi: 10.1152/ajprenal.00290.2015 [PubMed]
18. Forrest AR, Kawaji H, Rehli M, Baillie JK, de Hoon MJ, Haberle V, et al. (2014) A promoter-level mammalian expression atlas. Nature. 507: 462–470. doi: 10.1038/nature13182 [PMC free article] [PubMed]
19. Lanahan A, Williams JB, Sanders LK, Nathans D. (1992). Growth factor-induced delayed early response genes. Mol. Cell. Biol. 12: 3919–3929. [PMC free article] [PubMed]
20. Tullai JW, Schaffer ME, Mullenbrock S, Sholder G, Kasif S, Cooper GM. (2007). Immediate-early and delayed primary response genes are distinct in function and genomic architecture. J. Biol. Chem. 282: 23981–23995. doi: 10.1074/jbc.M702044200 [PMC free article] [PubMed]
21. Olive M, Krylov D, Echlin DR, Gardner K, Taparowsky E, Vinson C. (1997). A dominant negative to activation protein-1 (AP1) that abolishes DNA binding and inhibits oncogenesis. J. Biol. Chem. 272: 18586–18594. [PubMed]
22. Gerrard DT, Berry AA, Jennings RE, Piper Hanley K, Bobola N, Hanley NA. (2016) An integrative transcriptomic atlas of organogenesis in human embryos. Elife. 5: pii: e15657. [PMC free article] [PubMed]
23. Grossman RL, Heath AP, Ferretti V, Varmus HE, Lowy DR, Kibbe WA, et al. (2016) Toward a shared vision for cancer genomic data. New England Journal of Medicine 375: 1109–1112. doi: 10.1056/NEJMp1607591 [PubMed]
24. Seo JS, Ju YS, Lee WC, Shin JY, Lee JK, Bleazard T, et al. (2012) The transcriptional landscape and mutational profile of lung adenocarcinoma. Genome Res. 22: 2109–2119. doi: 10.1101/gr.145144.112 [PubMed]
25. The Cancer Genome Atlas Research Network (2014) Comprehensive molecular profiling of lung adenocarcinoma. Nature 511: 543–550. doi: 10.1038/nature13385 [PMC free article] [PubMed]
26. Konermann S, Brigham MD, Trevino AE, Joung J, Abudayyeh OO, Barcena C, et al. (2014). Genome-scale transcriptional activation by an engineered CRISPR-Cas9 complex. Nature 517: 583–588. doi: 10.1038/nature14136 [PMC free article] [PubMed]
27. Yoshimoto R, Mayeda A, Yoshida M, Nakagawa S. (2016) MALAT1 long non-coding RNA in cancer. Biochim Biophys Acta. 1859: 192–199. doi: 10.1016/j.bbagrm.2015.09.012 [PubMed]
28. Livak KJ, Schmittgen TD. (2001). Analysis of Relative Gene Expression Data Using Real-Time Quantitative PCR and the 2−ΔΔCT Method. Methods 25: 402–408. doi: 10.1006/meth.2001.1262 [PubMed]
29. Tiscornia G, Singer O, Verma IM. (2006). Production and purification of lentiviral vectors. Nat. Protoc. 1: 241–245. doi: 10.1038/nprot.2006.37 [PubMed]
30. Scherr M, Battmer K, Blömer U, Ganser A, Grez M. (2001). Quantitative determination of lentiviral vector particle numbers by real-time PCR. Biotechniques 31: 520–526. [PubMed]
31. Heigwer F, Kerr G, Boutros M. (2014). E-CRISP: fast CRISPR target site identification. Nat. Methods 11: 122–123. doi: 10.1038/nmeth.2812 [PubMed]
33. Bolger AM, Lohse M, Usadel B. (2014). Trimmomatic: A flexible trimmer for Illumina sequence data. Bioinformatics 30: 2114–2120. doi: 10.1093/bioinformatics/btu170 [PMC free article] [PubMed]
34. Dobin A, Davis CA, Schlesinger F, Drenkow J, Zaleski C, Jha S, et al. (2013). STAR: ultrafast universal RNA-seq aligner. Bioinformatics 29: 15–21. doi: 10.1093/bioinformatics/bts635 [PMC free article] [PubMed]
35. Kim D, Pertea G, Trapnell C, Pimentel H, Kelley R, Salzberg SL. (2013). TopHat2: accurate alignment of transcriptomes in the presence of insertions, deletions and gene fusions. Genome Biol. 14: R36 doi: 10.1186/gb-2013-14-4-r36 [PMC free article] [PubMed]
36. Trapnell C, Roberts A, Goff L, Pertea G, Kim D, Kelley DR, et al. (2012). Differential gene and transcript expression analysis of RNA-seq experiments with TopHat and Cufflinks. Nat. Protoc. 7: 562–578. doi: 10.1038/nprot.2012.016 [PMC free article] [PubMed]
37. GTEx Consortium. (2013) The Genotype-Tissue Expression (GTEx) project. Nat Genet. 45: 580–585 doi: 10.1038/ng.2653 [PMC free article] [PubMed]
38. ENCODE Project Consortium. (2012) An integrated encyclopedia of DNA elements in the human genome. Nature. 489: 57–74. doi: 10.1038/nature11247 [PMC free article] [PubMed]
39. Gardini A, Baillat D, Cesaroni M, Hu D, Marinis JM, Wagner EJ, et al. (2014) Integrator regulates transcriptional initiation and pause release following activation. Mol Cell. 56: 128–139. doi: 10.1016/j.molcel.2014.08.004 [PMC free article] [PubMed]
40. Blanchette M, Kent WJ, Riemer C, Elnitski L, Smit AF, Roskin KM, et al. (2004) Aligning multiple genomic sequences with the threaded blockset aligner. Genome Res. 14: 708–715. doi: 10.1101/gr.1933104 [PubMed]

Articles from PLoS ONE are provided here courtesy of Public Library of Science