We show here that EJIPT is an effective assay for measuring in vitro splicing, based on the specific assembly of EJC components on spliced mRNA. EJIPT has significant advantages over the conventional splicing assay using gel electrophoresis, including speed and ease of use and the ability to provide quantitative readout. EJIPT's favorable attributes of low-percent CV and high signal-to-noise ratio make it an excellent HTS tool. Combining this with the secondary assays we implemented provides a highly effective platform to identify modifiers of splicing and EJC formation or stability.
Our HTS identified 1,4-naphthoquinones and 1,4-heterocyclic quinones as splicing inhibitors. In addition to inhibiting splicing, NSC95397 also decreased snRNP assembly in a dose-dependent manner (see Fig. S3 in the supplemental material), consistent with results from a screen identifying inhibitors of Sm core assembly (62
). This activity is attributed to the compound's hydroxyethylthio side chains which generate reactive oxygen species (13
), causing intermolecular cross-linking of SMN (survival of motor neurons) and thus inactivation of the SMN complex activity in snRNP assembly (62
). It is likely that these moieties also contribute to its splicing inhibition activity, since 1,4 naphthoquinones without it () have higher IC50
s. The most potent compounds we found through SAR studies were the heterocyclic quinones, NSC663284 and BN82685, which inhibited in vitro
splicing with low IC50
s while having only a negligible effect on snRNP assembly, indicating selectivity for splicing (E and and ; see also Fig. S3). NSC95397 and several of the structurally related compounds have been described as CDC25 inhibitors (13
). However, the 50-fold discrepancy between in vitro
s for CDC25 phosphatase inhibition (~200 nM) versus splicing inhibition (~7 to 20 μM) and the lack of an effect by structurally unrelated CDC25 inhibitors (e.g., PM-20) ( and ) suggested that CDC25 is not the target of these drugs vis-a-vis splicing inhibition in vitro
. Therefore, albeit at higher concentrations, these anticancer drugs also target the spliceosome.
Of the relatively few small-molecule splicing inhibitors that have been recently described, all appear to target early stages of splicing. A natural product bioflavonoid, isoginkgetin, and histone acetyltransferase inhibitors (e.g., garcinol) have been shown to arrest splicing at prespliceosomal A complex stage, while histone deacetylase inhibitors (e.g., suberoylanilide hydroxamide [SAHA] and splitomicin) inhibit splicing at complex B (31
). The FR901464 derivatives, pladienolide-B and spliceostatin A, are antitumor agents which target U2 snRNP-associated SF3b components SAP130 and SAP155, respectively, and interfere with the A-to-B complex transition (21
). Meayamycin, another FR901464 analog, blocks complex A formation in vitro
with increased potency (2
). SR-dependent splicing inhibitors like TG003 (42
), which targets the Clk1/Sty kinase, or NB-506 (46
), which targets topoisomerase I, both block the phosphorylation of ASF/SF2 and prevent complex A formation. In contrast, the 1,4-naphthoquinone and 1,4-heterocyclic quinone scaffolds we identified here inhibit splicing at a later stage than these inhibitors, predominantly accumulating lariat-exon 2 and exon 1 intermediates and thus arresting splicing at a point between the first and second steps of catalysis. The persistence of complexes A and B at late times (A) and the accumulation of unspliced pre-mRNA ( and C), more prevalent at higher drug concentrations, suggest that these drugs may also target an earlier step. Although we did not detect convincing evidence of an effect on ASF/SF2 phosphorylation, it is possible that the benzothiazole scaffold of BN82685, which is also found in TG003, may additionally inhibit the Clk kinases and contribute to this early block to splicing (42
). Nevertheless, when these compounds are used at their IC50
s, a combination of denaturing gels, native gels, and size exclusion column fractionation demonstrated that spliceosome assembly and the first catalytic step typically occur with normal kinetics ( to ). Splicing reaction mixtures incubated longer (e.g., 3 h) continue to be blocked at this stage and do not produce more mRNA (data not shown). Thus, it is the efficiency and the rate at which the second step proceeds that are dramatically reduced, rather than an overall delay in splicing, causing lariat-exon 2 and exon 1 intermediates to accumulate (B).
The splicing pathway requires extensive rearrangement of proteins and RNA-RNA interactions of spliceosomal complexes, chief of which involve the remodeling of the U5/U4-U6 tri-snRNP (for a review, see reference 60
). The presence of a complex comprised only of the U2/U5/U6 snRNAs in equivalent amounts at late reaction times (A, 90 min) and the absence of U1 and U4 indicated that BN82685-stalled splicing reactions proceed beyond activation of complex B. A decrease in PRP2, the helicase which drives spliceosomes from B* to catalysis, suggests that even the activation step may also be impaired. Biochemical characterization and mass spectrometry of streptavidin-pulled down splicing complexes (B and C) were oriented by an inventory of spliceosomal proteins found in each complex as systematically determined by the Lührmann group (for a review, see references 60
). Importantly, BN82685 nonuniformly disrupted association of individual protein components, pointing to interference with specific rearrangements, rather than a block in the recruitment of the entire tri-snRNP. Indeed, levels of components in the stable core of U5 snRNP (PRP8, Brr2, and U5-116K) (1
) were largely unaffected, as were the Sm proteins and the core PRP19/CDC5L complex, which stably interacts with U5 during both catalytic steps (12
). Despite the egress of U4 snRNA concomitant with activation, U4/U6-specific proteins (e.g., PRP3 and PRP4) are still variably detected in complex C (5
). Reduced levels of PRP6 are also expected at late stages since this protein also dissociates from the U5 snRNP upon activation (60
). Levels of these proteins were all dramatically decreased in BN82685 reactions compared to reactions with DMSO. Interestingly, peptidyl-prolyl cis/trans
isomerases of the tri-snRNP (PPIH) and those associated with the PRP19/CDC5L complex (PPIE and PPIL1) were also consistently decreased. These enzymes are members of the cyclophilin family which associate with spliceosomes at discrete stages and are believed to play a role in both catalytic steps, possibly by determining the rate of protein folding (19
). Finally, given the block to step II and the failure to resolve C complexes, it is perhaps not surprising that the U2-associated DEXH/D-box helicase, PRP43, required for spliceosome disassembly, is also deficient.
Currently, it is unclear what the actual target(s) of BN82685 is and whether these changes we observe represent the cause or result of the inhibition. Considering the role currently described for PRP6, which is to bridge U5 to the U4-U6 di-snRNP in the precatalytic spliceosome, reactions would not be expected to proceed beyond complex B in its absence (17
). Indeed, mutations in the conserved phosphorylation sites at the N terminus of PRP6 prevent spliceosome activation and cause an accumulation of complex B containing all five snRNPs (8
). In contrast, BN82685 reactions clearly produced step I intermediates and formed complex C, with purified splicing complexes from this reaction containing only U2, U5, and U6 snRNAs. Therefore, a deficiency in PRP6, or inhibition of PRPF4 kinase, for example, which phosphorylates PRP6 and PRP31 as they are incorporated into complex B, is not likely the cause of splicing inhibition by BN82685 (49
). Upon catalytic activation, while PRP8, U5-116K, and U5-200K remain tightly bound to U5 snRNA, there is an exodus of several proteins from the spliceosome, involving PRP6, PRP31, PRP3, and PRP4, all of whose associations are decreased by BN82685. Thus, whereas DMSO-treated profiles represent a mixture of completed and ongoing reactions, reactions with BN82685 are “frozen” after catalytic step I, and proteins that exit upon activation immediately prior to this are decreased by comparison (e.g., PRP6). Furthermore, as a result of not completing step II, BN82685-treated spliceosomes fail to disassemble properly (A) and recycle the tri-snRNP, which in turn would affect the next round of splicing. This could explain why the predominant effect of BN82685 is in the accumulation of step I intermediates (~250 to 350%) and why the accumulation of pre-mRNA (~120 to 150%) appears secondary. As a final note, we considered PP1/PP2A phosphatases as candidate targets. These are required to dephosphorylate SAP155 and U5-116K prior to the second step of splicing; however, our compounds had little to no effect on the phosphorylation status of these substrates, nor was the addition of purified phosphatase able to rescue drug-treated reaction mixtures (data not shown) (50
Rearrangements during the early stages of splicing are fairly well understood, but the later steps are not as well characterized. For example, what the U2/U6 base pairing interactions immediately preceding catalysis are, whether the U4/U6 associated proteins exit catalytically activated spliceosomes en masse
or in distinct steps, and a detailed description of which events occur during spliceosome dissociation are just some of the open issues remaining (60
). BN82685 and the related compounds we identified can provide important tools for dissecting these steps, identifying additional rearrangements, intermediates, or players involved late in the splicing process. With these additional anticancer drugs now also shown to be splicing inhibitors, it has becoming increasingly appreciated that pre-mRNA splicing is another cellular process that can be targeted for cancer treatment (18
). The small but growing number of splicing inhibitors known so far and the ability of EJIPT and other HTS assays to discover additional ones should help dissect this important cellular pathway and have important biomedical applications.