A bichromatic reporter template has been used based on previously published designs^14,17. This was engineered to mimic VEGF-A exon 8 splicing. The reporter (termed pRG-VEGF8ab) consists of the endogenous VEGF-A intron 7 and required parts of exon 8, followed by the coding sequences for two fluorescent proteins [see details in Methods section and Stevens et al.¹⁵]. Depending on which 3′ splice site is used for exon 8, a different fluorescent protein is expressed. When the canonical, proximal splice site is used, dsRED is formed followed by a stop codon, so EGFP is not produced. When the distal splice site is used, there is a shift in the reading frame (dsRED + 1) and a fusion protein with EGFP is produced (Fig. 1B). Therefore, dsRED denotes the pro-angiogenic VEGF-A_xxxa isoforms, and EGFP denotes the anti-angiogenic VEGF-A_xxxb isoforms.

While this reporter has been used and validated by our lab in vivo in a transgenic mouse model¹⁵ for the purpose of this study – to use it in a screen – we wanted to pursue more validation in vitro. This is to ensure it reproduces the endogenous VEGF-A splicing pattern, including response to several compounds known to affect VEGF-A terminal exon splicing through inhibition of SRPK1¹⁸.

The endogenous VEGF-A is present exclusively as the pro-angiogenic isoform in HEK293 cells¹⁹. Transient transfection of pRG-VEGF8ab into HEK293 cells resulted in both a dsRED and EGFP signal being produced (Fig. 2A, upper panel). This is not surprising, as in a transient setting there are many copies of the reporter inside the cells and this can swamp the splicing machinery and the regulation of a particular splice site²⁰. However, when stable transfected HEK293 cells are obtained, the reporter displays only dsRED fluorescence (Fig. 2A, lower panel), which is expected for the exclusive proximal splice site usage in this cell line. RT-PCR analysis confirmed this expected splicing pattern, with only a band observed for dsRED (Fig. 2B), and FACS analysis shows exclusively dsRED positive cells (Fig. 2C). On the other hand, when pRG-VEGF8ab is transfected into proliferating conditionally immortalised podocytes, reported before to have high levels of VEGF-A₁₆₅b²¹, EGFP can predominantly be seen (Fig. 2D, upper panel). However, when transfected in Denys-Drash podocytes – previously reported to express predominantly the pro-angiogenic VEGF-A₁₆₅a isoform¹⁹ – only dsRED is observed as expected.Fig. 2

A Transfection of the reporter in HEK293 cells – upper panel: transient transfections (24 h) show both dsRED and EGFP signal due to multiple copies of the reporter; lower panel – stable transfected cells display only dsRED fluorescence corresponding to the VEGF-A terminal exon proximal splice site usage in these cells. B RT-PCR for the reporter shows correct splicing in HEK293 cells. C FACS analysis of HEK293 stably transfected with the reporter shows dsRED fluorescence only (parentFluor area is the trace for naïve untransfected cells). D In proliferating conditionally immortalised human podocytes, selection of the reporter’s distal splice site leads to EGFP expression. No dsRED, only EGFP expression was observed in transfected podocytes indicating an anti-angiogenic splicing pattern (upper panel). In Denys Drash Syndrome (DDS) podocytes, where the VEGF-A_xxxb isoforms are not expressed, we see only dsRED expression. Transfection efficiency is estimated at 20%. E SRPK1 inhibition increases use of the pRG-VEGF8ab reporter distal splice site (dsRED/EGFP), as measured by the plate reader. n = 3, *P < 0.05. No significant difference in dsRED+EGFP measurements. F The reporter was further validated after stable transfection into PC3 cells, Both SRPK1 inhibitors resulted in an increase in the EGFP/dsRED ratio. n = 3, *P < 0.05 vs DMSO. G To determine whether the reporter splicing mimics that of endogenous VEGF-A, PC3 cells were treated with the same SRPK1 inhibitors. The protein VEGF-A₁₆₅b/panVEGF-A₁₆₅ ratio was increased after treatment with SRPIN340 and SPHINX, although only significant with the more potent SRPK1 inhibitor, SPHINX. n = 3, *P < 0.05 vs DMSO.

To further validate that the reporter behaves as expected we have used a set of SRPK1 inhibitors. We have previously shown^18,19,22 that SRPK1 phosphorylates the splice factor SRSF1 and consequently promotes the proximal splice site in the VEGF-A terminal exon, defining the pro-angiogenic VEGF-A_xxxa isoform. If SRPK1 function is abrogated either genetically by knocked-down or using inhibitors (SRPIN340 or SPHINX), there is a switch in splicing towards the distal splice site and promotion of the anti-angiogenic VEGF-A_xxxb isoform. We would therefore expect that treatment of cells transfected with the pRG-VEGF8ab reporter with SRPK1 inhibitors to shift fluorescence of the reporter from dsRED towards EGFP. Indeed, as seen in Fig. 2E, there is a progressive and significant decrease in dsRED/EGFP ratio (indicating splicing changes) when HEK293 cells stably transfected with pRG-VEGF8ab are treated with inhibitors with increasing potency (SPHINX7 more potent than SPHINX which is more potent than SRPIN340). Remarkably, when treated with a control inhibitor that mimics SPHINX7 but it is chemically inert, there is no shift in the splicing ratio. Additionally, there is no movement with any inhibitor in the (dsRED + EGFP) values on the X-axis, indicating that these inhibitors do not affect the reporter expression.

We next stably transfected PC3 cells with pRG-VEGF8ab to further validate the reporter at the mRNA level using RT-PCR. In control treated PC3s, dsRED was predominantly expressed at the mRNA level, which equates to this cells line having a high expression of pro-angiogenic VEGF-A isoforms¹⁸. Treatment of the transfected PC3 cells with SPRIN340 and SPHINX (10 μM, 48 h) resulted in a significant increase in the EGFP/dsRED ratio at the mRNA level (*P < 0.05; Fig. 2F). We next used Western blotting to determine the protein expression of the endogenous VEGF-A isoforms in PC3s exposed to the same conditions. Both SRPIN340 and SPHINX increased the VEGF-A₁₆₅b/panVEGF-A₁₆₅ ratio, although this was only significant with the more potent SPHINX (*P < 0.05; Fig. 2G).

Of note, even in an in vivo setting the reporter responds accurately to SRPK1 inhibitors – when transgenic mice harbouring the pRG-VEGF8ab reporter are treated intraperitoneally with SPHINX injections, there is a reporter switch in the expected direction ascertained by both fluorescence and RT-PCR¹⁵.

Additionally, we have recently reported²³ that a natural blueberry extract (DIAVIT) upregulates production of VEGF-A₁₆₅b isoforms in podocytes. Indeed, when stably transfected HEK293 cells were treated with DIAVIT, the reporter dsRED/EGFP splicing ratio decreases as expected, in a dose-dependent manner (Supplementary Fig. 1).

The fact that the reporter responds in a similar way to the endogenous VEGF-A terminal exon splicing gave us confidence that it can be used to study the regulation of this splicing event.

In a quest to uncover novel anti-angiogenic compounds that are able to modify VEGF-A splicing we set-up a screen using the splicing reporter described. We have chosen a repositioning screen using the LOPAC (library of pharmacologically active compounds; Sigma, Inc). The LOPAC library is formed of 1280 compounds that are pharmacologically active, either marketed drugs or pharmaceutically relevant structures. The library is designed to cover most signalling pathways and major drug target classes. Due to one of the goals of this project – to find modulators of angiogenesis with one major target being tumour angiogenesis – we chose the prostate cancer PC3 cell line for screening.

The screen was done in three steps (see Fig. 3). Following the primary screen, the first positive list of compounds was screened for elimination of false positives that may directly affect fluorescence or RNA stability. This was done using cells stably transfected with control reporters that lack the intron, cannot be spliced and mimic the resulting dsRED or EGFP RNA of the two spliced isoforms (see step 2 in Fig. 3). Finally, due to the inherent errors in any methodology, the resulting compound list after the control screen were re-screened using a different method – FACS. A hit-list of 9 compounds (called ESSOs) resulted after this last step of screening – see Table 1.Fig. 3

PC3 cells transfected with the fluorescent VEGF splicing reporter were treated with the LOPAC compound library and reporter splicing measured using a fluorescent plate reader (Primary screen, step 1). Hit compounds were used to treat non-splicing control reporter constructs (Control screen, step 2). Flow cytometry was used to measure the effect of the hit compounds on reporter splicing (Secondary screen, step 3).

To validate the main hit compounds described above, RT-PCR and western blot analysis were performed on PC3 cells individually treated with ESSO compounds at 10 μM for 48 h. At the mRNA level, ESSO 1, 3, 8 and 9 resulted in a significant switch in splicing towards the VEGF-A₁₆₅b isoforms (Fig. 4A). Furthermore, both ESSO1 and 3 showed a significant increase in the VEGF-A₁₆₅b/panVEGF-A₁₆₅ ratio at the protein level in PC3 cells, whereas ESSO 5 and 6 resulted in a significant decrease in panVEGF-A₁₆₅ expression Fig. 4B).Fig. 4

A The effect of ESSO compounds on the alternative splicing of endogenous VEGF-A mRNA transcripts. RNA was extracted from PC3 cells following 48 h treatment with ESSOs (10 μM), reverse transcribed and amplified by PCR. Anti-angiogenic isoform transcripts were expected to be 66 bp smaller than pro-angiogenic transcripts. ESSO1, 3, 8 and 9 resulted in a significant shift in splicing to increase VEGF-A₁₆₅b at the mRNA level. n = 3, *P < 0.05 vs DMSO. B Changes in expression of anti-angiogenic VEGF-A protein isoforms upon treatment of PC3s with ESSO compounds (example western blot). Following treatment with each small molecule (10 μM, 48 h), protein was extracted and used for western blotting. The bands observed correspond to VEGF-A₁₆₅b and VEGF-A₁₆₅a/b isoform dimers (46 kDa). ESSO1 and 3 resulted in a significant increase in VEGF-A₁₆₅b relative to panVEGF-A₁₆₅, whereas ESSO5 and 6 resulted in a significant decrease in both VEGF-A₁₆₅ isoforms. n = 5, *P < 0.05 vs DMSO. C Changes in expression of anti-angiogenic VEGF-A protein isoforms upon treatment of podocytes with ESSO 1 and 3 compounds (example western blot). Following treatment with each small molecule (10 μM, 48 h), protein was extracted and used for western blotting. The bands observed correspond to VEGF-A₁₆₅b and VEGF-A₁₆₅a/b isoform dimers (46 kDa). ESSO1 and 3 resulted in a significant increase in VEGF-A₁₆₅b relative to panVEGF-A₁₆₅, as well as a significant decrease in panVEGF-A₁₆₅ expression. n = 8, *P < 0.05 vs DMSO.

The effect of the compounds of the protein VEGF-A₁₆₅b/panVEGF-A₁₆₅ ratio was also assessed in podocytes, a cell line with a higher endogenous level of VEGF-A₁₆₅b^19,21 We found that treatment of podocytes for 48 h with 10 μM ESSO1 and ESSO3 resulted in a significant increase in the VEGF-A₁₆₅b/panVEGF-A₁₆₅ ratio at the protein level, as well as a significant decrease in panVEGF-A₁₆₅ expression (Fig. 4C).

ESSOs are part of a repositioning library composed of either substances that are approved drugs or chemicals that were developed for drug purposing. While for most of them, the mechanism of action and signalling pathways involved are known (Fig. 5A), it is not clear whether or not the new activity they are tested for (angiogenesis inhibition) occurs through the same mechanism – e.g. ESSO8 is an RTK (receptor tyrosine kinase) inhibitor – does its activity on VEGF-A splicing and angiogenesis depend on RTK signalling?Fig. 5

A Examples of known signalling mechanisms for ESSOs. B Splicing gels. Separation of products after 2 h splicing reaction. Separated on 5% denaturing polyacrylamide gels. Positions of precursor, lariat intermediate, spliced product and lariat product indicated. 5′ exon intermediate would be at 130 nt. Labelled DNA markers in left lane – sizes indicated to left.

A number of small molecules are known to directly target core components of the splicing machinery²⁴. To test the possibility that one or more of the ESSOs might affect VEGF-A alternative splicing by directly targeting components of the splicing machinery, we used cell-free in vitro splicing assays. Radiolabelled pre-mRNA substrates containing VEGF-A exon 8a and 8b 3′ splice sites and 365 nt of the adjacent upstream intron 7, with a constitutive Tpm1 5′ exon and splice site were transcribed in vitro and then incubated in HeLa cell nuclear extract. The RNAs spliced exclusively using the VEGF-A 8a 3′ splice site in HeLa extract (Fig. 5B and Supplementary Fig 2A). When the predicted exon 8a branch sites were weakened by A to G point mutations at −30 and −31 nt upstream (Supplementary Fig 2A), the efficiency of 8a splicing was substantially reduced, with no concomitant upregulation of 8b splicing. This suggests that splicing of exon 8b requires specific activation. We then tested the effects of ESSO compounds at 1, 5 and 10 μM concentration. The ESSOs either had no effect or modestly increased exon 8a splicing (Fig. 5B and Supp Fig. 2B). However, none of the ESSO compounds detectably activated exon 8b splicing. These results therefore suggest that the ESSO compounds do not directly target the splicing machinery to affect VEGF-A 8a vs 8b splicing, but more likely act upstream to regulate splicing indirectly. The assay and our conclusions have limitations though, as we do not have a positive control for exon 8b activation in this assay.

We further wanted to test whether the ESSO compounds are able to inhibit angiogenesis. We first used them in ex vivo Matrigel assays. There was a significant decrease in the Matrigel tubule length with all ESSO compounds compared to DMSO control (Fig. 6A), suggestive of anti-angiogenic properties. This was not due to a toxic effect on the HUVECs as there was no difference in cell viability with the various treatments, as assessed by Trypan blue staining (Fig. 6B). To show that the effect is, at least in part, due to a switch in VEGF-A splicing when using ESSO1 and ESSO3, we performed a rescue experiment in which antibodies specific to VEGF-A₁₆₅b (56-1; 10 µg/ml) were also added while treating with ESSO1 and 3, as examples. As expected, this partially rescued the effect of the ESSOs on Matrigel tubule length (Fig. 6C).Fig. 6

A Using a Matrigel angiogenesis assay ESSOs 1-9 resulted in a statistically significant decrease in HUVEC tubule length, thus angiogenesis, compared to DMSO control. n = 5, *P < 0.05. B HUVEC cell viability is not altered by treatment with ESSO compounds compared to DMSO control. n = 3, P = ns. C Anti-angiogenic VEGF-A isoforms mediate ex vivo inhibition of endothelial tube formation. The presence of anti-VEGF-A₁₆₅b (56-1; 10 μg/ml) significantly reversed the effect of ESSO 1 and 3 on tubule length. Three images of each well were captured and quantified. n = 3, *P < 0.05 vs. DMSO control. ^#P < 0.05 vs. ESSO control. D Schematic of the fibroblast (NHDF) and endothelial cell (HUVEC) co-culture experiment to assess angiogenic potential in the presence of ESSO-treated PC3 conditioned media. E Endothelial cell tubule formation was significantly reduced by ESSOs 1 and 7 during endothelial fibroblast co-culture. Ten microscopic fields of each co-culture were quantified for the total length of endothelial tubules formed normalised to the total cell area. n = 3, **P < 0.01 vs DMSO.

To further test ESSOs anti-angiogenic properties, another assay was employed; an endothelial cell - fibroblast co-culture assay (Fig. 6D)²⁵. Normal human dermal fibroblasts (NHDF) were plated and after settling for 24 h, HUVECs were added on top. Conditioned media from PC3 cells treated with ESSOs (as above) was added to the co-culture. Following incubation, cells were fixed and stained with CD31 (an endothelial cell marker) and DAPI, to visualise all cells in a particular well. While in control conditions (DMSO), endothelial cells start proliferating and migrating to form contacts (Fig. 6E, left upper panel), this is inhibited in the presence of most ESSOs, of which ESSO1 and ESSO7 were significant (see Fig. 6E, right upper panel and quantification in the lower panel).

Furthermore, we found that addition to HUVEC cells of conditioned media from PC3 cells treated with either ESSO 1 or ESSO 3 resulted in a significant decrease in the phosphorylation of VEGFR2 in comparison to a DMSO control, further indicating that the increase in the VEGF-A₁₆₅b/VEGF-A₁₆₅a ratio in response to ESSO treatment has anti-angiogenic properties (Supplementary Fig. 3).

These results are consistent with our hypothesis that ESSOs have anti-angiogenic activity ex vivo, some through an effect on VEGF-A splicing.

To further investigate the effects of certain ESSOs on angiogenesis, we employed an in vivo assay using Matrigel plugs. We chose first to test ESSO 9 because out of the four ESSOs that showed significant switch in VEGF-A splicing, ESSO9 had the lowest efficiency (Fig. 4A). Cells were incubated with either DMSO or ESSO9, mixed with Matrigel and implanted subcutaneously in the back of nude mice. Normally, small vessels from the mouse vasculature grow into the plug (see schematic Fig. 7A). In the presence of anti-angiogenic activity, there is less vasculature in the plug. Indeed, this is the case of ESSO9 treatment, as there was less red colour in the ESSO9-treated plugs (Fig. 7B, C). Furthermore, we confirmed a VEGF-A splicing switch to increase the VEGF-A_xxxb isoform at the mRNA level in some of the plugs treated with ESSO9 (Fig. 7D). Finally, we repeated this experiment with ESSO1 and obtained similar results (Supplementary Fig. 4). This shows that the two ESSOs, even if with different efficiencies in switching VEGF-A splicing, work well in vivo to inhibit angiogenesis.Fig. 7

A ESSO9-treated PC3 cells (48 h) were mixed with Matrigel. The Matrigel and control cells were injected subcutaneously into nude mice. Four days post-injection, the plugs were removed and RNA was extracted. B, C Photographs were taken of each plug and colour intensity quantified using ImageJ software. There was a significant reduction in ESSO9 plug intensity. n = 6, P = 0.0002. D RT-PCR for VEGF-A splice variants showed a switch in splicing to increase VEGF-A_xxxb isoforms in some ESSO9-treated plugs compared to DMSO controls (n = 3).

HEK293, PC3, Podocyte, HUVEC and NHDF cells were sub-cultured from existing cultures within the lab. HEK293, PC3 and NHDF cells were cultured in DMEM (D6429, Sigma - St Louis, Missouri, USA) supplemented with 10% foetal bovine serum (10270, GIBCO – Waltham, Massachusetts, USA) and 1% penicillin streptomycin (GIBCO). Podocytes and Human umbilical vein endothelial cells (HUVECs) were cultured in EBM-2 media (Lonza – Basel, Switzerland)) supplemented with EGM-2 Bulletkit (Lonza). All cell lines were cultured at 37 °C and in a humidified incubator with 5% CO₂.

Details about reporter design are given in Stevens et al.¹⁵ where we used it to construct a transgenic mouse.

A library containing 1280 FDA-approved chemicals, LOPAC, Library of Pharmaceutically Active Compounds (LO4200, Sigma-Aldrich) was used. Chemicals in the library were dissolved DMSO at a stock concentration of 10 mM. PC3 pRG8ab cells were trypsinised and diluted to 8 × 10⁴ cell/ml using DMEM. In all, 100 μl of the cell solution was seeded into each well of a black 96-well plate. The cells in each 96-well plate were treated with DMSO only and 16 different compounds from the LOPAC in triplicate at 10 μM. Wells of the outside edges of 96-well plates were not used for treatments or measured for fluorescence. Plates were incubated for 48 h at 37 °C. Changes in reporter alternative splicing caused by treatment with LOPAC compounds were measured using a VICTOR X multi-label plate reader (Perkin Elmer – Waltham, Massachusetts, USA). Each plate was measured using VICTOR plate reader three times. DMSO treated wells were used as control measurements and statistically compared to treated wells from the same 96-well plate using a one-way ANOVA.

These were performed by standard methods/conditions – for details please see Supplementary Information

This was performed as previously described;²³ for details please see Supplementary Information

These were performed as previously described;¹⁹ for details please see Supplementary Information

Cells were washed and trypsinised and described previously and diluted to an approximate concentration of 2 × 10⁵ cells per ml. In all, 0.5 ml of cell suspension was transferred to a screw cap tube and mixed thoroughly with 100 μl of 0.4% Trypan Blue stain. Cells and the stain were incubated for 5 mins at room temperature. A haemocytometer was filled and used to count cells under a microscope. Stained cells were non-viable and unstained cells were viable.

PC3 cells were cultured in media containing 10 μM ESSO or DMSO for 48 h. Cells were detached from culture plates and diluted in DMEM to a concentration of 2 × 10⁷ per ml and placed on ice. Matrigel basement membrane protein (734–1100, VWR) was thawed on ice. In total, 100 μl of cell suspension was mixed with 400 μl of Matrigel and 10 μM of the test compound. The mix of cells and matrix was subcutaneously injected into each upper flank of Crl:CD1-Foxn1^nu nude mice (males, 2 months old; Charles River). Four days post-injection, mice were culled by cervical dislocation (Schedule 1) and the Matrigel plugs were extracted. Images of each plug were taken, and their colour quantified using Photoshop as an indication of blood vessel infiltration into the matrix. Plugs were flash frozen in liquid nitrogen. In total, 1 ml of Trizol was added to each plug and samples were homogenised for RNA extraction as described.

were carried out in HeLa cell nuclear extract as described previously⁴³. Transcripts contained 365 or 165 nt of VEGF-A intron 7 immediately upstream of the exon 8a 3′ splice site and 220 nt of exon 8, including the 8b 3′ splice site followed by a BamHI restriction site. The 5′ exon, 5′ splice site and first 150 nt of the intron were from rat Tpm1 exon 1. Run-off transcripts were generated with T7 RNA polymerase from BamHI linearised template in the presence of α-[³²P]UTP. Spliced RNA products were separated on 5 or 6% denaturing polyacrylamide gels and analysed by autoradiography or phosphorimager for quantitation.

Comparisons of two datasets were performed using Students’ t test or a Mann–Whitney U test, depending on whether the data met the normal distribution. A comparison of three or more groups was performed using one-way analysis of variance (ANOVA) with Dunnett’s post-test. P < 0.05 was considered to indicate a statistically significant difference.