|Home | About | Journals | Submit | Contact Us | Français|
The angiogenic capability of colorectal carcinomas (CRC), and their susceptibility to anti‐angiogenic therapy, is determined by expression of vascular endothelial growth factor (VEGF) isoforms. The intracellular protein T‐cell Intracellular Antigen (TIA‐1) alters post‐transcriptional RNA processing and binds VEGF‐A mRNA. We therefore tested the hypothesis that TIA‐1 could regulate VEGF‐A isoform expression in colorectal cancers. TIA‐1 and VEGF‐A isoform expression was measured in colorectal cancers and cell lines. We discovered that an endogenous splice variant of TIA‐1 encoding a truncated protein, short TIA‐1 (sTIA‐1) was expressed in CRC tissues and invasive K‐Ras mutant colon cancer cells and tissues but not in adenoma cell lines. sTIA‐1 was more highly expressed in CRC than in normal tissues and increased with tumour stage. Knockdown of sTIA‐1 or over‐expression of full length TIA‐1 (flTIA‐1) induced expression of the anti‐angiogenic VEGF isoform VEGF‐A165b. Whereas flTIA‐1 selectively bound VEGF‐A165 mRNA and increased translation of VEGF‐A165b, sTIA‐1 prevented this binding. In nude mice, xenografted colon cancer cells over‐expressing flTIA‐1 formed smaller, less vascular tumours than those expressing sTIA‐1, but flTIA‐1 expression inhibited the effect of anti‐VEGF antibodies. These results indicate that alternative splicing of an RNA binding protein can regulate isoform specific expression of VEGF providing an added layer of complexity to the angiogenic profile of colorectal cancer and their resistance to anti‐angiogenic therapy.
Alternative RNA splicing, export and translational regulation all contribute to cancer formation and progression (Venables, 2006). These are dependent on the complex secondary structure of the 3′ UTR of the mRNA recognised by binding proteins such as T‐cell Intracellular Antigen‐1 (TIA‐1) (Eberhardt et al., 2007). TIA‐1 contains three domains with RNA recognition motifs (RRM1, RRM2 and RRM3), the latter two of which have been shown to be important for RNA binding and selectivity (Bauer et al., 2012) (Cruz‐Gallardo et al., 2014). It has been shown to be alternatively spliced in exon 5 to form two isoforms (a and b, Supplementary Figure 1A) with b being the most active in its splicing activity, but TIA‐1 also acts as a stress induced translational inhibitor, localising to stress granules with polyA RNA (Izquierdo and Valcarcel, 2007). TIA‐1 alters both co‐transcriptional and post‐transcriptional RNA processing and binds VEGF‐A RNA (Suswam et al., 2005), the alternative splicing of which influences the angiogenic capability of colorectal carcinomas (CRC) (Ferrara et al., 1991).
VEGF‐A exists as multiple alternatively spliced isoforms. Exons 6 and 7 endow heparin binding (Houck et al., 1991), whereas exon 8 determines angiogenicity (Harper and Bates, 2008). Exon 8 has two splice acceptor sites, a proximal one, and a distal one, 66 bases downstream. The stop codon for the two isoforms lies the same number of bases downstream from the respective splice sites, resulting in two families of isoforms of the same amino acid number but with differing six amino acid sequence at the end, which alter the angiogenicity of the protein. Proximal splice site (PSS) selection in exon 8 results in the angiogenic family (VEGF‐Axxx where xxx is the number of amino acid e.g. VEGF‐A165), but distal splicing (DSS) results in VEGF‐Axxxb, anti‐angiogenic proteins of the same size (e.g. VEGF‐A165b) (Bates et al., 2002). A variable but specific upregulation of the angiogenic isoforms is seen in colorectal cancer (Varey et al., 2008), but although the mechanisms underlying this splice specificity are being elucidated in epithelial cells (Amin et al., 2011; Nowak et al., 2008) and lung cancer (Merdzhanova et al., 2010), isoform specific control of expression, particularly in colon cancer is still not known. We therefore tested the hypothesis that TIA‐1 could regulate VEGF splicing.
HEK293, colonic carcinoma cell lines HT29, HCT116, HAC7, SW480, Osteosarcoma SaOS‐2 and U2OS, the cervical cancer cell line, HeLa, prostate cancer cells, PC3 and LNCaP (ATCC; Rockville, MD) LS174t (ECACC, Salisbury, UK), adenoma cell lines AA/C1 and RG/C2, the in vitro transformed adenoma cell line AA/C1/SB/10, referred to as 10C cells) and breast cancer MDA‐MB231 (Caliper, PerkinElmer, USA), Human Haemangioma Endothelial Cells (ESC and stem cells (HSC, a kind gift from Joyce Bischoff at Harvard University), myeloma RPMI‐8226, and cardiac myocytes (HCM, from Promocell) were maintained as described previously (Williams et al., 1990) or according to manufacturer's instructions. Cancer cell lines were validated by STR profiling (Identicell, Aarhus, Denmark). Human tissues were collected under Local Ethics Committee Approval. An initial study was undertaken on six frozen paired colon samples (normal and cancerous) from partial colon resection. Additional RNA was extracted from 40 formalin fixed, paraffin embedded (FFPE) samples of normal and tumour tissues. These were taken from 22 male, 18 female patients, mean age 69 years, 28 with colon carcinoma, 18 with rectal carcinoma and 1 with carcinoma of the rectosigmoid junction.
An antibody to sTIA‐1 was generated by Abgent Europe using a KLH‐conjugated VSLKNGQ NCPG peptide. For access please contact the corresponding author.
Cell experiments used at least 3 independent biological replicates (independent experiments), with each western blot, ELISA and RT‐PCR carried out on independently replicated samples. For western blot, and RT‐PCR where gels are shown, each is indicative of a representative of 3 experiments, and quantitated as per the figures. ELISAs were undertaken on 3 separate sets of cells. Protein samples from homogenised tissue (100 μg) and whole cell extract (50 μg) were prepared using standard protocols. Protein lysates were resolved on SDS PAGE, and probed with a mouse monoclonal anti‐COX‐2 antibody (sc‐19999; Santa Cruz), a/b‐tubulin (Sigma) anti‐VEGF‐A165b IgG (A56/1; R&D Systems), rabbit polyclonal anti‐TIA‐1 antibody (sc‐28237; Santa Cruz, detects full length and sTIA‐1 isoforms), or sc‐166246 C10 mouse monoclonal N‐terminal; Santa Cruz and rabbit anti‐VEGF‐A IgG (A‐20, sc‐120; Santa Cruz), goat polyclonal laminin (Santa Cruz) and mouse monoclonal lamin (Cell Signalling) antibodies. Proteins were detected using chemiluminescence and analysed by NIH image, or by fluorescently labelled secondary antibody imaging and imaged using a LiCOR Odyssey, and quantified using LiCOR imaging software. ELISAs were carried out as described previously (Varey et al., 2008). The VEGF‐A165b ELISA uses an antibody that can detect all VEGF‐Axxxb isoforms but no VEGF‐Axxx isoforms, but its affinity to all of the different VEGF‐Axxxb isoforms is not known. We have assumed that it is the same as VEGF‐A165b and therefore calculated the levels making this assumption.
For immunoprecipitation, 50 μl of protein A or protein G (depending on capture antibody) magnetic beads (Millipore) was washed with 500 μl PBS containing 0.1% Tween 20. After removing the buffer, the beads were resuspended in 100 μl of PBS/0.1% Tween, capture antibody added to the beads and incubated at room temperature for 10 min before adding the protein samples. The protein samples and antibody were incubated at 2–8 °C overnight with continuous mixing, washed with PBS/Tween0.1% before adding sample buffer. For spectrometry, after running the samples on PAGE, the gel was fixed in methanol:acetic acid (50%:7%) for 30 min and stained with SYPRO Ruby (Molecular Probes) overnight. The gel was de‐stained in methanol:Acetic Acid (10%:7%), washed and imaged. The band was excised and subjected to in‐gel tryptic digestion using a ProGest automated digestion unit (Digilab UK). The resulting peptides were fractionated using a Dionex Ultimate 3000 nanoHPLC system in line with an LTQ‐Orbitrap Velos mass spectrometer (Thermo Scientific) as described in the supplementary material. Raw data files were processed and quantified using Proteome Discoverer software v1.2 (Thermo Scientific) and searched against the SwissProt Human database using the Mascot search engine (Matrix Science) with parameters as described in the supplementary material.
Total RNAs from tissue (200 mg) and cells (107) was extracted using the RNeasy kit (Qiagen) and quantified. RNA was extracted from 25 μm thick sections of the FFPE samples by RNeasy FFPE kit (Qiagen) according to the manufacturer's protocol. Standard RT‐PCR protocols were used. For cDNA synthesis form FFPE samples random primer or gene specific primers were used. Extracted DNA was cloned into the p‐GEM T‐Easy plasmid (Promega) and sequenced using standard synthetic primers complementary to the vector DNA at the junction of the cloning site (T7, SP6) and a pTarget sequencing primer 5′‐TTACGCCAAGTTATTTAGGTGACA‐3′ by the Dundee sequencing service (Dundee University). For each cell line 10 clones were sequenced.
RNA Immunoprecipitation RIP was performed as previously described (Cok et al., 2003; Tenenbaum et al., 2002). In brief, 5 × 107 cells were harvested, cross linked with 1%formaldehyde at room temperature for 10 min and quenched with 0.25 M glycine. Protein was extracted, and 100 μg of pre‐cleared protein incubated with 3 μg TIA‐1 antibody immobilized using Seize Primary immunoprecipitation kit (Pierce Biotechnology) and immunoprecipitates isolated by centrifugation. Total RNA was extracted from immunoprecipitates and subjected to RT‐PCR as described.
The VEGF‐A sequence from 35 bp upstream of exon 8a to 35 bp downstream of exon 8b was inserted into the ADML‐MS2 plasmid as previously described (Nowak et al., 2008). Nuclear protein (75 μg) from HEK 293 or AA/C1/SB/10 (10C) cells were prepared and added to the MS2‐MBP fusion protein/VEGF‐RNA mix in 0.5 mM ATP, 6.4 mM MgCl2, 20 mM creatine phosphate for 1 h at 30 °C. Proteins bound to the MS2‐MBP/VEGF‐MS2 RNA complex were affinity selected on amylose beads and immunoblot analysis was performed on 30 μg of protein samples using TIA‐1 antibody and β‐tubulin antibodies as previously described (Nowak et al., 2008). Blots are representative of at least three experiments.
The cDNA for flTIA‐1was generated by RT‐PCR amplification using Pfu Turbo DNA polymerase (Stratagene) and primer TIA‐1‐forward and TIA‐1‐reverse from RNA from the RG/C2 adenoma cell line. The 1239 bp fragment was cloned in p‐GEM T‐easy vector and sequenced in both directions. The cDNA for both variants (13‐exon TIA‐1NCBI NM_022173.2 and 12‐exon TIA‐1NCBI NM_022037.2) were obtained using these primers. For in vitro expression, the cDNA were subcloned into mammalian expression vectors, pTarget (Promega) and pcDNA3 according to the manufacturer's instructions. siRNA to exon 6A was generated by Qiagene. Two custom siRNAs targeting sTIA‐1 (siRNA1: CAGAATTGCCCTGGCTAACTA and siRNA2: GCCCTGGCTAA CTACAAGCTA) and control siRNA (Lopez de Silanes et al., 2005) were assessed. Two colon cancer cell lines were transfected with mixture of the 2 siRNAs (50 nM) using HiPerfect transfection reagents (Qiagen). Knockdown was monitored after 24 (RNA) and 48 (protein) hours. TIA‐GFP plasmid is a kind gift from Paul Anderson (Harvard).
Transfection: LS174t and 10C cells were electroporated with 2 μg cDNA of pTarget‐flTIA‐1 cDNA in pTarget and pmaxGFP™ vectors pTarget vector or pmaxGFP™ with an AMAXA Nucleofector system using a basic epithelial cell transfection Kit (VPI‐1005, Amaxa GmbH). Transfectants were analysed 24 h post nucleofection by Trypan blue staining and fluorescence microscopy for GFP control plasmid. Transfected cells were selected using G418 (500 μg/ml) for 4 weeks.
The cells were 4% formaldehyde fixed (15 min) and incubated in 1%BSA/10% normal goat serum in TBS‐Tween (0.1%) for 1 h. The cells were then incubated with the antibody (ab40693 rabbit polyclonal C‐terminal, or sc‐166246 C10 mouse monoclonal N‐terminal, 5 μg/ml) overnight at +4 °C. The secondary antibody (red) was Alexa Fluor donkey anti‐rabbit and (green) Alexa Fluor 488, used at a 1/500 dilution for 1 h. Vector shield Hoescht was used to stain the cell nuclei (blue). For confocal imaging a Nikon D‐Eclipse C1 Laser Scanning Microscope was used. Cells were imaged using a 60× oil immersion objective with a numerical aperture (NA) of 1.4–5 cell clumps were selected at random and a z stack of 25 images taken for each clump of cells. The central slice of each of these stacks was selected and used for the analysis of the ratio of staining between the nucleus (Hoescht) and cytoplasmic regions of the cell. In NIH image, the TIA‐1 positive area and the Hoescht positive area were determined by tracing and nuclear area subtracted from the TIA‐1 positive area and expressed as a proportion of the TIA‐1 positive area.
The VEGF‐A full length 3′UTR was cloned downstream of the Thymidine Kinase promoter and the Firefly Luciferase ORF (construct LUC‐VEGF165), (from Jean‐Jacques Feige and Nadia Cherradi, CEA Grenoble). This was modified using the Quick‐Change II XL Site‐Directed Mutagenesis Kit (Stratagene) to delete the first 66 nucleotides (LUC‐VEGF‐A165b) of exon 8. The plasmid pRL‐TK (Promega) encoding Renilla Luciferase was used as an internal control. 10C cells were transfected with the both reporter constructs using the Amaxa system. Luciferase activities were measured with the DUAL‐Luciferase reporter assay system (Promega) on a VICTOR 3 1420 Multilabel Counter Luminometer (PerkinElmer).
VEGF‐A mRNA decay was measured in 10C cells after transfection with TIA‐1 and treatment with 10 μmol/L actinomycin D. Cells were harvested every hour over a 4‐h period, and VEGF‐A isoform mRNA levels analysed by RT‐PCR.
Experiments were undertaken under home office license after review by the University of Bristol Ethical review group. 2 × 106 untransfected, flTIA‐1, VEGF‐A165, doubly or vector transfected cells in 0.2 ml PBS were injected subcutaneously in the dorsum of 6 Balb/C nude mice (from University of Bristol colony) for each group. Mice were treated bi weekly with mouse IgG (Sigma), or an anti‐mouse VEGF‐A neutralising antibody (Liang et al., 2006) intraperitoneally twice weekly at 2 mg/kg. Sample size was calculated a priori using G‐Power on the basis of being able to have 80% power to detect a mean ± SD 2 ± 0.5 fold change in size and vascular density. All experiments were carried out with the investigator blinded to the groups.
5 μm thick de‐waxed, re‐hydrated sections were microwaved in 0.01 M sodium citrate, pH 6.0 for 15 min at 800 W. Sections were incubated with VEGF‐A Receptor 2 (Rabbit Mab 2479 Cell Signalling) or CD31 antibodies at 1:100 overnight at 4 °C in a humidity chamber and after application of secondary antibodies and appropriate enzymatic process (Vector DAB kit), vessel number counted in 10 randomly selected areas for each tumour (n = 6).
Results were analysed using Prism software, given as Mean ± SEM unless otherwise stated and compared using one‐way ANOVA or t tests for comparison of means. Proportions were compared using Fisher's exact test. p < 0.05 was considered significant. Power analysis on non‐significant results were carried out to ensure power of at least 80% (using G‐Power).
In five carcinoma (LS174t, HCT116, SW480, HT29, HCA7) and an adenocarcinoma line (10C) TIA‐1 RNA, but not two other RNA binding proteins, TTP and HuR, was lower than in adenoma lines RG/C2, and AA/C1 (Figure S1A, B), amplified using primers to detect full length (flTIA‐1). Expression of the 42–43 kDa TIA‐1 protein was similar across cell lines, but in all carcinoma and the 10C cells, an additional band at ~16 kDa was seen (Figure 1A, Figure S1C) suggesting the presence of a truncated TIA‐1 protein. Amplification of TIA‐1 mRNA sequence from the 10C cells using primers that flank exons 6 and 7 revealed an additional product only in carcinoma and 10C cells (Figure 1B, Figure S1D), due to the insertion of a 66bp sequence (Figure 1C) from the intron between exons 6 and 7 of the tia‐1 gene (Figure 1E). This was also present in BRAF V600E + ve A375 melanoma cells (Figure S1E). The 66base sequence is flanked by a classical consensus splice sequence (AG…GU) and has a 73% CU sequence immediately upstream, suggesting a cassette exon undergoing alternative splicing. This sequence results in a unique 11 amino acid sequence and a premature stop codon, which should generate a 144 amino acid truncated TIA‐1 peptide of 15.8 kDa (Figure 1D), with only RNA recognition motif (RRM)‐1 and part of the RRM 2, consistent with the short band (~16 kDa) detected by the anti‐TIA‐1 antibody in Figure 1A. Sequencing of TIA‐1 genomic DNA from 10C cells showed no mutations in the region around either the 5′ or 3′ acceptor or donor sites. This mRNA is identical to a splicing event described as exon 6A when TIA‐1 itself was over‐expressed in a cancer cell line (Le Guiner et al., 2001), but its endogenous expression has not previously been demonstrated. A BLAST search of this sequence showed identity in 66 ESTs, all consistent with it being this TIA‐1 isoform. A rabbit antibody raised to the unique peptide sequence of sTIA‐1 detected the protein in 10C cells (Figure S1F). Immunoprecipitation of 10C protein (Figure S1G) and immunoblotting for TIA‐1 showed both short and full length TIA‐1. IP with a flTIA‐1 antibody and IB with sTIA‐1 demonstrated that the protein detected with this antibody was TIA‐1 in LS174t cells as well. This protein was not present in AAC1 cells. (Figure S1H). Both protein (Figure 1E top image) and RNA expression (Figure 1E bottom image) of sTIA‐1 was seen in sarcoma (SaOS), HeLa, prostate cancer (PC3 and LNCaP), and MDA‐MB‐231 breast cancer cells, but not osteosarcoma (U2OS), haemangioma endothelial (a differentiated phenotype), haemangioma stem (HSC), human embryonic kidney (HEK), or RPMI‐8226 leukaemia cells or human cardiac myocytes (HCM). Of the 18 cell lines in which sTIA‐1 was investigated, only 2 of the 8 that lacked sTIA‐1 have ras mutations or over‐expression, whereas 8 of the 10 that did express sTIA‐1 had ras over‐expression or activating mutations, and the other two are BRAF mutants, (HT29 and A375 cells). In both LS174t and 10C cells the Ras inhibitor Farnesyl Thiosalysilic Acid (FTA) significantly inhibited splicing to sTIA‐1 in a dose dependent manner (Figure 1F). Downstream signalling through MEK and PI3K was required for this splicing in both cell lines (Figure S2A). Ras is active in AAC1 and RGC2 cell lines (sTIA‐1 negative) indicating that it is necessary, but not sufficient to induce splicing of TIA‐1.
TIA‐1 splicing is modified in colon carcinoma cells. A. Immunoblot for TIA‐1 in carcinoma, adenocarcinoma and adenoma cell lines. A smaller protein product (sTIA‐1) was seen in carcinoma and adenocarcinoma (AC) cells. B. RT‐PCR ...
Ras inhibition (Figure 2A, Figure S2B–D) and MEK and PI3K inhibition (Figure S2E) also switched splice form expression from VEGF‐A165a to VEGF‐A165b in both 10C and LS174t cell types. VEGF‐A165b protein expression was confirmed by immunoprecipitation and peptide sequencing (Figure S3A and B). These results suggest that hyperactivated ras signalling also acts to repress VEGF‐Axxxb splicing, so we investigated a possible link between VEGF and TIA‐1. Isoform specific siRNA knockdown of sTIA‐1 (Figure S3C–G), reduced VEGF‐A165a mRNA in both 10C and LS174t and increased VEGF‐A165b mRNA in 10C cells but made no difference in AA/C1 cells (Figure 2B). VEGF‐A165b but not VEGF‐A165 protein was significantly increased by siRNA to sTIA‐1 in LS174t cells (p = 0.013), whereas only a significant reduction in the VEGF‐A165a isoform expression was seen in 10C cells (p = 0.03, Figure 2C). Thus there was a significant switch in the ratio of isoforms in both cell types towards a less angiogenic phenotype after knockdown of sTIA‐1 (Figure 2D, p < 0.01 ANOVA). Conversely, VEGF‐A165b expression was rescued by flTIA‐1 expression in 10C and LS174t cells measured by RT‐PCR (Figure 3E), immunoblot (Figure 2F) and ELISA (Figure 2G), in 10C cells to 90% VEGF‐A165b (p < 0.001 compared with ctrl), similar to AA/C1 cells and in LS174t from 5 ± 1.8% to 12 ± 1.5% of total VEGF‐A (p < 0.01). The reverse experiment (over‐expression of sTIA‐1) resulted in lethality of unknown cause. flTIA‐1 also reduced cyclooxygenase‐2 protein expression (Figure S3H), a known TIA‐1 target. RNA immunoprecipitation (RIP) of TIA‐1 showed that, whereas 10C, AA/C1 and RG/C1 all expressed both VEGF‐A165 and VEGF‐A165b, immunoprecipitation with an anti‐TIA‐1 antibody pulled down only VEGF‐A165 in AA/C1 and RG/C1 (cells expressing only flTIA‐1), but no VEGF‐A mRNA in sTIA‐1 expressing cell lines (10C, HCA7 and LS174t, Figure 2H). In 10C and LS174t cells, transfection with flTIA‐1 restored TIA‐1 binding to VEGF‐A165a mRNA, as evidenced by pull down of VEGF‐A165 mRNA, compared with untransfected cells (ctrl, Figure 2I). sTIA‐1 also affected binding of TIA‐1 to COX‐2 (Figure S3I). The binding site of TIA‐1 to VEGF‐A mRNA was demonstrated, in HEK cell nuclear extract using an MS2‐MBP binding assay (Nowak et al., 2008) to reside in the VEGF‐A exon 8a sequence and 35 bases of intron 7 (Figure 2J). Nuclear extract from 10C cells, while containing TIA‐1 (as seen in the flow through unbound eluate in the lower panel), was unable to bind to VEGF‐A mRNA sequences (Figure 2J).
TIA‐1 regulates VEGF splice form expression. A. PCR for VEGF‐A showing ras inhibition by FTA increased VEGF‐A165b expression. B. Colonic tumour cells were transfected with 2 siRNAs targeted to exon 6A of TIA‐1 (sTIA), ...
TIA‐1 localisation regulates VEGF‐A mRNA translation but not degradation. A. 10C cells, or 10C cells expressing flTIA‐1 were treated with the transcriptional inhibitor Actinomycin D (ActD) followed by RT‐PCR for VEGF. ...
flTIA‐1 has been shown to function both as a translational repressor (Piecyk et al., 2000) and an alternative splicing factor (Barron and Lou, 2012; Del Gatto‐Konczak et al., 2000). Expression of flTIA‐1 did not affect the half‐lives of VEGF‐A165b and VEGF‐A165 mRNA in 10C cells, determined by investigating mRNA decay rates in cells treated with the transcriptional inhibitor Actinomycin D (ActD) (Figure 3A and B). Although it did increase the amount of VEGF‐A165b at the start of the experiment, Figure 3C shows that there was no difference in the half‐life. There was no significant difference in expression of a luciferase reporter containing the 3′UTRs of VEGF‐A165a or VEGF‐A165b (Figure 3D) in either 10C or LS174t cells (Figure 3E), but transfection with flTIA‐1 resulted in significantly increased luciferase activity with the VEGF‐A165b 3′UTR compared with the VEGF‐A165a 3′UTR (Figure 3F) indicating preferential translation of VEGF‐A165b by TIA‐1 as well as inhibiting splicing of VEGF‐A165. The flTIA‐1 isoform contains three RNA recognition motifs (RRMs). RRM 2 and 3 are required for 3′ translational repression (Bauer et al., 2012), and this is lacking in sTIA‐1, which only contains the full length RRM1, which cannot by itself bind to RNA (Dember et al., 1996) but is required for flTIA‐1 binding to specific sequences (e.g. U rich sequences) (Bauer et al., 2012), thus conferring specificity. Thus, while sTIA‐1 prevents flTIA‐1 mediated repression of splicing of the pro‐angiogenic isoforms of VEGF‐A, the increased level of VEGF‐A165b mRNA and protein induced by over‐expression of flTIA‐1 in colon cells suggest that it also acts to selectively represses translation of the exon 8a containing isoforms. This may explain why the effects of sTIA‐1 siRNA knockdown on RNA and protein were subtly different in the two cell types (in 10C it increased VEGF‐A165b RNA but not protein, and vice versa in LS174t cells).
Nuclear/cytoplasmic separation of proteins (Figure S4A) showed that in AAC1 cells, TIA‐1 is found as the full length form but only in the nucleus, whereas in LS174t and 10C cells a substantial proportion of all TIA‐1 isoforms and products were expressed in the cytoplasm, including sTIA‐1 (Figure 3G). This was confirmed by transfection of cells with full length TIA‐1 fused to GFP (Figure 3H), and by immunofluorescent staining with antibodies to the N terminus of TIA‐1 (detects all isoforms, Figure S4B). sTIA‐1 knockdown, surprisingly, resulted in enhanced cytoplasmic staining for TIA‐1 (Figure 3I). This was confirmed in flTIA‐1‐GFP over‐expressing 10C cells, with sTIA‐1 knockdown (Figure S4C). N‐terminal staining (which would detect sTIA‐1) was reduced, consistent with a reduction in only the sTIA‐1 component (Figure S4C). Nuclear localisation of GFP‐flTIA‐1 was reduced by sTIA‐1 knockdown (Figure 3J). sTIA‐1 therefore alters localisation of the flTIA‐1 within the cell, thereby preventing it from binding VEGF‐A mRNA. The increased not decreased cytoplasmic TIA‐1 resulting from reduced of sTIA‐1 in K‐Ras mutant cells suggests that the GFP‐flTIA‐1 fusion protein localisation is sTIA‐1 dependent. Thus sTIA‐1 expression in K‐Ras mutant cells may act by preventing flTIA‐1 nuclear trafficking (to and from the nucleus), preventing it from interacting with other potential regulators of splicing and/or translation. This in turn alters the expression of VEGF‐A isoforms and results in a more angiogenic phenotype of the tumours. Co‐immunoprecipitation indicated that it is possible that the two isoforms could complex together (Figure S1G), but how this happens is not clear. The localisation experiments indicate that in carcinoma cells flTIA‐1 excess either by over‐expression with GFP, by knockdown of sTIA‐1 or both results in cytoplasmic flTIA‐1 and/or VEGF‐A165b expression, whereas in adenoma cells, where K‐Ras is not over‐expressed, flTIA‐1 is nuclear. Thus when sTIA‐1 dominates in the nucleus, VEGF‐A165 is generated.
sTIA‐1 and VEGF‐A165b RNA were detected in freshly isolated human colorectal cancer samples (Figure 4A) confirmed by sequencing. Quantification showed both a significant increase in sTIA‐1 expression in cancer (Figure 4B) and a significantly lower VEGF‐A165b:VEGF‐A165 ratio in the tumours than the controls (Figure 4C). In histological samples, flTIA‐1 RNA (Figure 4D) was found only in 13 of 31 tumours (42%) but 22 of 30 normal tissues (73%, p = 0.02, Fisher's exact test) samples (Figure 4E). The proportion of tumours that were K‐Ras mutant was significantly higher (p < 0.01) in those that had no flTIA‐1, and K‐Ras mutant tumours rarely (2/13 cases) had flTIA‐1 detected (Figure 4F). Quantification showed that sTIA‐1 exceeded flTIA‐1 in 20 of 31 tumours (65%) but in only 9 of 30 normal tissues (30%, p = 0.01, Figure 4G). Moreover, sTIA‐1 was present in 14 of the 15 Dukes' C (93%), but only 9 of the 16 Dukes' B (56%, p < 0.01, Figure 4H). There was a significantly higher microvascular density (Figure 4I) in samples in which sTIA‐1 but not flTIA‐1 was found (Figure 4J).
sTIA‐1 splice variants are present in human colon cancer. A. Upper panel, RT‐PCR of mRNA extracted from paired tumour (T) and normal (N) tissues from 6 patients undergoing bowel resection for colon cancer. In 4 of 6 patients, the splice ...
Over‐expression of flTIA‐1 inhibited tumour growth in nude mice (Figure 5A), resulting in significantly smaller tumours at 18 and 21 days after injection (Figure 5B). This reduction was rescued by VEGF‐A165 cDNA expression i.e. VEGF‐A that was not under the control of alternative splicing (Figure 5B). Tumours from flTIA‐1 transfected cells had reduced tumour vessel density (Figure 5C and D, p < 0.01). Mice with LS174t tumours expressing flTIA‐1 treated with an anti‐VEGF‐A antibody (Liang et al., 2006) grew significantly faster compared with IgG (Figure 5E), at the same rate as wild type tumours treated with anti‐VEGF‐A treatment (Figure 5F), but still more slowly than untreated wild type tumours (Figure 5G).
flTIA‐1 over‐expression reduces growth of sTIA‐1 expressing tumours in vivo in a VEGF dependent manner A. 2 × 106 control transfected (vector), flTIA‐1 transfected cells or double VEGF‐A165a ...
Here we identify a novel K‐ras dependent endogenous splice variant of TIA‐1, short TIA‐1 (sTIA‐1) in CRC patients and cancer cells, which controls VEGF‐A isoform expression. This sTIA‐1 had been identified as being capable of being induced by over‐expression of recombinant mouse TIA‐1 in human HEK cells (Le Guiner et al., 2001). This is the first demonstration that this splice variant may be present in human cancers, and that it may regulate VEGF splice form expression. sTIA‐1, containing only one RNA recognition motif (RRM), prevented binding of full length TIA‐1 to VEGF‐A165 mRNA, blocking both its angiogenic VEGF splice site repression and its promotion of translation of anti‐angiogenic VEGF‐A165b. It has previously been shown that RRM2 and 3 are required for binding of TIA‐1 to RNA (Bauer et al., 2012), and interestingly, RRM3 is required for binding to C rich stretches (Cruz‐Gallardo et al., 2014). Moreover, inclusion of exon 5 in TIA‐1 results in a higher splice activity, indicating that the activity of the protein can be altered by alternative splicing (Izquierdo and Valcarcel, 2007). As both of these domains are disrupted in sTIA, it is likely that sTIA‐1 does not act by directly binding to the VEGF RNA and preventing flTIA‐1 from exerting its effects, and the direct mechanism of repression of distal splicing is not known. As sTIA‐1 did co‐immunoprecipitate with flTIA‐1 it is possible that sTIA‐1 acts by heterodimerising with flTIA‐1 and preventing it from repressing splicing and/or translation of the pro‐angiogenic isoform, VEGF‐A165a. Cancer cells over‐expressing flTIA‐1 formed smaller, less vascular tumours than those expressing sTIA‐1, and were more resistant to anti‐VEGF antibodies such as bevacizumab. These results indicate that K‐ras mediated splicing of an RNA binding protein may regulate isoform specific expression of VEGF, providing an added layer of complexity to the angiogenic profile of colorectal cancer and their resistance to anti‐angiogenic therapy. This opens up the question as to what is the downstream mechanism through which K‐Ras regulates TIA‐1 splicing. There is no clear evidence as to how K‐Ras activation can induce splicing switches. Splicing is regulated by the complex interaction of splicing factor proteins that can be phosphorylated, localise to the nucleus and take part in a complicated interaction with other RNA binding proteins and auxillary factors that control splice site location. The mechanisms through which TIA‐1 is spliced is not known, and none of the known splicing factors have yet been identified as controlling Exon 6A splicing. Interestingly, over‐expression of human flTIA‐1 did not induce sTIA‐1 splicing in cancer cells, in contrast to that shown by mouse flYIA‐1 in human HEK cells (Le Guiner et al., 2001).
The finding that over‐expression of VEGF‐A165 can reverse the effect of flTIA‐1 expression on tumour growth inhibition suggests that the effect was VEGF splicing dependent, through anti‐angiogenic activity, or through direct effects on the cells. These findings may have significant therapeutic implications, but are not directly translatable. Low VEGF‐A165b levels predict bevacizumab response in CRC (Bates et al., 2012), but, bevacizumab provides significant clinical benefit in patients irrespective of K‐ras mutations (Hurwitz et al., 2009). Although patients with K‐ras mutations did do better on bevacizumab compared with controls (9.3 compared with 5.5 months, an increase of 3.8 months), they did not perform as well as patients with wild type K‐ras (13.5 versus 7.4, an increase of 6.1 months). It would therefore be interesting to know how patients with K‐ras mutations, in which sTIA‐1 is expressed, responded. It is not yet clear whether sTIA‐1 predicts survival, or can drive angiogenesis by reducing the anti‐angiogenic VEGF‐A165b splice variants, or by increasing the VEGF‐A165 variant. These VEGF‐A165b variants have been shown to be down regulated so far in melanoma (Pritchard‐Jones et al., 2007), colon (Varey et al., 2008), prostate (Woolard et al., 2004), and kidney (Bates et al., 2002) cancers, but the mechanism of regulation of these variants in cancer is still poorly described (Amin et al., 2011, 2010, 2008). The loose relationship between sTIA‐1 and VEGF‐A165b down regulation in patients suggests a complex interaction, just as for alternative splicing of p73 and VEGF‐A165b isoforms, also demonstrated in colon cancer (Diaz et al., 2008). This is supported by the findings that Ras inhibition in 10C cells was more effective at lower doses than in LS174t cells, and that the RNA and protein levels of VEGF‐A165b in the presence or absence of shTIA‐1 are not directly correlated. The effect of TIA‐1 on RNA stability, as shown in Figure 3 indicates that TIA‐1 is likely to act in a pleiotropic manner on the amount of VEGF isoforms present – acting by altering translation specifically for one set of isoforms, by regulating splicing at the proximal splice site, or, most likely, by changing both splicing and translation. Binding of TIA‐1 to the VEGF RNA was shown, but it is also possible that other vascular growth factors are regulated by TIA‐1, although these have not yet been published.
In summary, we show that alternative splicing of TIA‐1 in human colon cancer may regulate the balance of the pro‐ and anti‐angiogenic VEGF‐A isoforms, and control tumour growth and angiogenesis in animal models. This suggests that TIA‐1 can act as a splicing and translational regulator of VEGF‐A and alternative splicing of TIA‐1 may lead to enhanced angiogenesis and tumour growth in colorectal cancer, and resistance to anti‐VEGF‐A therapy.
MHZ, EMA, CHA designed and carried out experimental procedures, ED, KS, XY, KH, AS, OD'S, KBB carried out experimental procedures, ACW, DJK, AHJS, RSM, MRL, and SJH contributed to the design and secured funding for the experimental work, AHRV contributed to design and interpretation of the experiments, DOB designed the experiments, secured funding, and wrote the paper. All authors contributed to the writing of the manuscript and revised the draft paper.
DOB and SJH are holders of a patent on VEGF‐A165b.
The following is the supplementary data related to this article:
This work was supported by the Association for International Cancer Research (MHZ 07‐0605), Cancer Research UK KS, CF11392/A8451, CHA, C11392/A10484 and CP/MHZ CITRINA, the Medical Research Council (MR/K020366/1), the Oxford Comprehensive Biomedical Research Centre (ED), the British Heart Foundation (DOB, BS/06/005), the North Bristol Specific Cancer Project Fund (96464), University of West of England PhD Studentship to EA and MRL, the Richard Bright VEGF‐A Research Trust (SJH).
Supplementary data related to this article can be found at doi:10.1016/j.molonc.2014.07.017.
Hamdollah Zadeh Maryam A., Amin Elianna M., Hoareau-Aveilla Coralie, Domingo Enric, Symonds Kirsty E., Ye Xi, Heesom Katherine J., Salmon Andrew, D'Silva Olivia, Betteridge Kai B., Williams Ann C., Kerr David J., Salmon Andrew H.J., Oltean Sebastian, Midgley Rachel S., Ladomery Michael R., Harper Steven J., Varey Alexander H.R., Bates David O., (2015), Alternative splicing of TIA-1 in human colon cancer regulates VEGF isoform expression, angiogenesis, tumour growth and bevacizumab resistance, Molecular Oncology, 9, doi: 10.1016/j.molonc.2014.07.017.