Background
Hypoxia is defined as the state in which the availability or delivery of oxygen is insufficient to meet tissue demand [
1]. It is a common feature of aggressive cancers, as angiogenesis cannot match the rate of tumour growth [
2]. Insufficient angiogenesis results in poor delivery of nutrients, and this coupled with severe hypoxic stress can trigger apoptosis [
3]. Remarkably, cancer cells are often able to survive the harsh conditions associated with hypoxia.
Hypoxic tumours are more difficult to treat. In radiotherapy, high-energy photons directed at the tumour site react with biological molecules to form DNA-damaging molecules, causing cancer cells to undergo apoptosis leading to shrinkage of the tumour [
4]. Therefore, in hypoxic regions where there is a low level of oxygen molecules, radiotherapy is often not an effective form of treatment. In addition, tumour hypoxia can lead to resistance to chemotherapeutic drugs. A review published by Brown outlines three reasons for this [
5]. Firstly, with diminished vasculature, drugs are unable to travel to hypoxic regions. Secondly, many chemotherapeutics target rapidly proliferating cells; however hypoxic cells tend to progress through the cell cycle much more slowly. Thirdly, hypoxia increases the expression of proteins involved in drug resistance, such as metallothionein-IIA (MT-IIA) involved in cisplatin resistance [
6] and periostin involved in arsenic trioxide resistance [
7]. A greater understanding of how tumour cells are able to adapt to hypoxia could help improve existing therapies or even suggest new treatment solutions.
Changes in alternative splicing of pre-mRNA may be one mechanism through which cells are able to adapt to hypoxic conditions. Pre-mRNA splicing involves the excision of introns and precise joining of exons to form the mature RNA [
8]. 95% of pre-mRNAs are alternatively spliced in humans [
9], giving rise to at least 100,000 distinct proteins [
10] despite there only being an estimated 20,000 protein-coding genes. Splice isoforms encode proteins with distinct cellular functions. Hypoxia has been shown to have a significant effect on the alternative splicing of genes in normal endothelial cells [
11,
12]; in cartilage endplate-derived stem cells [
13]; in Hep3B liver [
14] and in MCF7 breast cancer cell lines [
15]. Hypoxia reduces expression of MAX (MYC-associated factor X) in endothelial cells through increased unproductive splicing [
16]. The splicing of the
SMN2 gene is also modified by hypoxia in the context of Spinal Muscular Atrophy [
17]. While several studies utilised exon arrays to study changes in exonic splicing during hypoxia, a recent study conducted in MCF7 breast cancer cells utilised whole transcriptome RNA-seq identifying intron retention as the most commonly altered alternative splicing event [
15].
Given these notable changes in alternative splicing in hypoxia, the question is which regulators of splicing are involved? The aim of the present study is firstly to examine the effect of hypoxia on alternative splicing in a prostate cancer cell line model, and secondly to examine the expression of several splice factors and splice factor in response to hypoxia.
Methods
Cell lines, hypoxia treatments and CLK1 inhibition
The prostate cancer cell lines, PC3 (Sigma-Aldrich, 90112714), VCaP (Sigma-Aldrich, 06020201) and DU145 (ATCC, HTB-81), the normal prostate epithelial PNT2 cell line (ECACC, 95012613), the colorectal cancer cell line HT29 (Sigma-Aldrich, 91072201), and the breast cancer cell line MCF7 (Sigma-Aldrich, 86012803) were cultured in high glucose Dulbecco’s modified Eagle’s medium (DMEM), supplemented with 10% fetal bovine serum (FBS) and 1% L-glutamine-penicillin-streptomycin and grown at 37 °C with 5% CO2. Hypoxia was attained in a modular incubator chamber (MIC-101; Billups-Rothenberg, USA) using a gas mixture containing; 1% O2 / 5% CO2 / 94% N2 (BOC, UK). The chamber was flooded with the hypoxic gas mixture for 2 min and then sealed and stored in an incubator for 48 h at 37 °C in 5% CO2. The normoxic control was stored in the same incubator for the same amount of time. CLK1 was inhibited with the benzothiazole TG003 (Sigma-Aldrich) at 10–50 μM; the TG003 stock was dissolved in dimethylsulphoxide (DMSO).
RNA was extracted using Tri reagent (Sigma-Aldrich) as per manufacturer’s instructions with the following modifications; the first three centrifugations (homogenate, phase separation and RNA isolation) were conducted at 10,000 g, phase separation centrifugation was for 30 min; and 0.1 ml chloroform was used for phase separation. RNA pellets were washed with 0.5 ml 70% ethanol and centrifuged at 3500 g, and the wash step was repeated twice. RNA pellets were resuspended in 30 μl nuclease-free H2O (Promega). Genomic DNA was removed using the Precision DNase kit (Primerdesign, UK). Total RNA concentration was determined using a Nanodrop spectrophotometer (ThermoFisher Scientific, Delaware).
cDNA synthesis and standard PCR
cDNA was synthesised with the GoScript reverse transcription system (Promega, UK). Total RNA was reverse transcribed using a 1:1 mixture of random primers and oligo(dT). The GoTaq hot start polymerase kit (Promega, UK) was used to perform standard PCR. PCR conditions were: initial denaturation at 94 °C for 2 min; then 35–40 cycles: 94 °C for 30 s; 58 °C for 30 s; and 72 °C for 30 s followed by a final extension at 72 °C for 5 min. Primer sequences used in standard PCR are shown in Additional file
1: Table S1. PCR products were run on 2% (
w/
v) agarose gels for 1 h at 95 V and analysed using FluorChem Q software, Alpha Innotech MultiImage III. Optical density peak values were generated with ImageJ software. Excel was used to calculate the percentage splicing index (PSI/ψ) where ψ = exon inclusion/ exon inclusion + exon skipping band intensities.
High-throughput PCR
High-throughput PCR was performed at the RNOmics platform at the University of Sherbrooke, Canada. RNA integrity was determined using an Agilent 2100 Bioanalyzer (Agilent Technologies). 2.2 μg total RNA were reverse transcribed with Transcriptor reverse transcriptase, random hexamers, dNTPs (Roche Diagnostics), and 10 units of RNAse OUT (Invitrogen) in 20 μl. All primers were resuspended in 20–100 μM stock solutions in Tris-EDTA buffer (IDT) and diluted as primer pairs to 1.2 μM in RNase and DNase-free water (IDT). RT-PCR reactions were performed using 10 ng cDNA as template in a 10 μL final volume containing 0.2 mmol/L dNTP, 1.5 mmol/L MgCl
2, 0.6 μmol/L each primer, and 0.2 units of Platinum Taq DNA polymerase (Invitrogen). After an initial denaturation at 95 °C for 2 min, 35 PCR cycles were performed; 94 °C 30 s, 55 °C 30 s, and 72 °C 60 s, followed by a final extension step at 72 °C for 2 min. PCR reactions were carried out on the GeneAmp PCR System 9700 (ABI), and the amplicons were examined by automated microcapillary electrophoresis using Caliper LC-90 instruments (Caliper LifeSciences). The percent spliced index (PSI, ψ) was calculated as the ratio of the band intensity of the exon inclusion amplicon divided by the sum of both inclusion and exclusion amplicon band intensities, measured using ImageJ software. The selection of cassette exons was based on genes from the NCBI RefSeq database associated with cancer and apoptosis pathways, all of which suitable for PSI analysis [
18,
19].
qRT-PCR
qRT-PCR was performed in 96-well plates as per the protocol by Primerdesign (UK), using SensiFAST Sybr Hi-Rox kit (Bioline, UK) mastermix. Three technical repeats were conducted per experimental sample, and nuclease free water was used as a negative control. qRT-PCR plates were run on an ABI 7300 qRT-PCR thermal cycler (Applied Biosciences). PCR conditions were as follows; initially 95 °C for 10 min; then 95 °C for 15 s; 60 °C for 1 min for 40 cycles. Melting curve conditions were 95 °C for 15 s; 60 °C for 1 min; 95 °C for 15 s. Primer sequences used in qRT-PCR are shown in Additional file
2: Table S2. CT values were normalised to the
UBC and
RPL13A housekeeping genes. All results were calculated using Excel software (Microsoft, USA).
Protein extraction and quantification
For the extraction of protein, cells were washed with phosphate buffered saline (PBS) before the addition RIPA lysis buffer (Sigma, UK) and phosphatase inhibitor tablets (Complete Mini EDTA-free PI tablets, Roche Diagnostics, UK). Cell lysates were collected in 1.5 ml Eppendorf tubes and kept on ice. The Pierce BCA assay (Thermo Fisher Scientific, USA) was used to measure protein concentration.
Western blotting
Protein samples were mixed 1:1 with 2× Laemmli buffer (Sigma, UK) and incubated at 100 °C for 5 min before being chilled on ice. Protein samples of equal amounts were subjected to SDS-PAGE and transferred to PDVF membranes. Membranes were incubated at 4 °C overnight with one of the following antibodies; anti-CA IX (M75 monoclonal antibody, Novus Biologicals); anti-SRSF1 (sc-33652, SantaCruz Biotechnology, UK); anti-CLK1 (R1471–1 s, Abiocode, USA); anti-phospho SR(1H4) (sc-13509, SantaCruz Biotechnology, UK); anti-β-actin (ab8226, Abcam, UK). Horse anti-mouse IgG HRP-linked antibody (7076S, Cell Signalling, UK) was used to detect all antibodies apart from anti-CLK1, which was detected by goat anti-rabbit IgG HRP-linked antibody (7074, Cell Signalling, UK). Laminata Forta Western HRP substrate (Millipore, UK) was used to image protein bands.
The promoter sequences of
SRSF1, SRSF2, SRSF3, SAM68, HuR, hnRNP A1, CLK1 and
SRPK1 splice factors and splice factor kinases were examined for the presence of HIF response element (HRE) sequences (5’-RCGTG-3′) using the Eukaryotic Promoter Database (
http://epd.vital-it.ch/). Parameters were set from − 1000 to + 100 of the transcription start site (TSS), as HIF-1α mostly binds close to the TSS [
20].
Statistical analysis
All data obtained for the standard and qRT-PCR experiments was subjected to Shapiro-Wilks and Bartlett’s test in order to deduce the distribution and variance of the data, respectively. For the standard PCR, data that was found to be normally distributed and homoschedastic was then subjected to a student’s T-test in order to examine the statistical significance of the data. Results found not to be normally distributed and/or heteroschedastic were subjected to a Mann Whitney U statistical test of significance. A Mann Whitney U non-parametric test was used for test for statistical significance for all the real-time PCR data. For all results obtained from the standard PCR and real-time PCR experiments, the means were calculated from the results and plotted on graphs, and error bars were added using the 95% confidence interval.
Discussion
Recently there has been interest in examining hypoxia-driven changes in alternative splicing. A genome-wide study was conducted in hypoxic endothelial cells using the Affymetrix human exon 1.0 array [
12]. Alterations were found in 294 genes including nine alternative splicing events that were not previously related to hypoxia. The changes involved to a large extent genes involved in angiogenesis, glucose metabolism, cell cycle, DNA repair, and cytoskeletal remodelling. In the DNA damage response, hypoxia drives the alternative splicing of genes towards noncoding isoforms through increased intron retention [
29]. In a similar vein, in endothelial cells hypoxia promotes the expression of a splice isoform of MAX (Myc associated factor X) that is degraded by nonsense-mediated decay, and another splice isoform that encodes an unstable protein; the net effect is less active MAX protein is produced [
16]. Investigations using the Hep3B hepatocarcinoma cell line revealed the hypoxia altered alternative splicing patterns of HIF and non-HIF genes including
CA IX, RAP1GDS1 and
MBP [
14];
MBP was also identified in our high throughput PCR screen (Fig.
1b). In MCF7 breast cancer cells hypoxia was found to affect the alternative splicing of genes that are involved in a number of cancer-driving processes including angiogenesis, cell growth and apoptosis, suggesting that hypoxia plays a role in cancer progression [
15]. Furthermore, the latter study also confirmed intron retention as the most common type of splicing event to be altered in hypoxia. Typically, intron retention can result in the use of a premature stop codon, which in turn can trigger nonsense mediated decay (NMD) or produce a truncated protein. Thus there are a number of clearly documented cases of cells adapting to hypoxia by changing alternative splicing of key genes [
11‐
17]. In the present study we can confirm that changes in alternative splicing of cancer genes also occur in response to hypoxia in the widely used prostate cancer cell lines PC3 and VCaP, and in the normal epithelial cell line PNT2.
Hypoxic tumour cells can become resistant to apoptosis by expressing apoptosis inhibitors such as IAP-2 [
30]. There are several components of the apoptosis machinery and genes express pro- or anti-apoptotic splice isoforms (
BCL2L1, MCL1, CASP9 and
APAF1 are notable examples). A recent study shows that the pharmacological inhibition of APAF1 allows cells in which apoptosis is induced by hypoxia to recover [
31‐
33]. We examined the alternative splicing of
CASP9, and noted that in hypoxia expression of the pro-apoptotic full length caspase 9a isoform increased significantly in PC3 and PNT2 cells (Fig.
4c). We suggest that changes in alternative splicing in hypoxia result in a different balance of splice isoforms that contribute to the induction of apoptosis.
We examined the promoters of a representative set of splice factors and splice factor kinases, and found putative HIF consensus binding sites in 7/8 (the exception being SRPK1). The expression of all eight splice factors and kinases appeared to increase in hypoxia, suggesting a general activation of the machinery of alternative splicing in conditions of severe hypoxic stress. We turned our attention to the CDC-2 like (CLK) family of splice factor kinases, because recent literature has identified an increase in expression of CLK1 in response to hypoxia [
23,
34]. We confirmed the induction of CLK1, not just in PC3 cells, but also in another prostate cancer cell line, DU145, and in HT29 colorectal and MCF7 breast cancer cell lines (Fig.
6a and
b). We observed the parallel induction of CLK1 and CLK3 in the four cell lines. CLK3 is best known for its expression in normal testis [
25]; however we suggest that along with CLK1, CLK3 might contribute to adaptation to hypoxia in cancer cells.
To address the functional significance of CLK1 induction in the context of alternative splicing in hypoxia, we treated PC3 cells with the CLK1 inhibitor, the benzothiazole TG003 [
28]. Inhibition of CLK1 with up to 50 μM TG003 increased the expression of caspase 9b, the anti-apoptotic isoform in which exons 3–6 are skipped (Fig.
6c). There is interest in the potential of targeting CLK1 in the context of the influenza [
35] and HIV-1 viruses [
36] and in cancer cells [
37,
38]. The fact that CLK kinases appear to help cancer cells adapt to hypoxia suggests that they could be useful therapeutic targets.
Acknowledgments
We are very grateful to Professor Silvia Pastorekova for hosting EB in her laboratory and providing valuable expertise in setting up hypoxia experiments. We are very grateful to Professor Masatoshi Hagiwara for advice on CLK1 inhibition. We would also like to thank David Corry for valuable technical assistance. ML is a member of the European Organisation for Research and Treatment of Cancer (EORTC) Pathobiology group.