Human spliceosomal protein CWC22 plays a role in coupling splicing to exon junction complex deposition and nonsense-mediated decay

Abstract

The multiprotein exon junction complex (EJC) that is deposited upstream of spliced junctions orchestrates downstream events in the life of a metazoan mRNA, including its surveillance via the nonsense-mediated decay (NMD) pathway. However, the mechanism by which the spliceosome mediates EJC formation is not well understood. We show that human eIF4G-like spliceosomal protein (h)CWC22 directly interacts with the core EJC component eIF4AIII in vitro and in vivo; mutations at the predicted hCWC22/eIF4AIII interface disrupt association. In vivo depletion of hCWC22, as for yeast Cwc22p, causes a splicing defect, resulting in decreased levels of mature cellular mRNAs. Nonetheless, hCWC22 depletion yields increased levels of spliced RNA from the unusual nonsense codon-containing U22 host gene, which is a natural substrate of NMD. To test whether hCWC22 acts in NMD through coupling splicing to EJC deposition, we searched for mutations in hCWC22 that affect eIF4AIII deposition without affecting splicing. Addition of hCWC22(G168Y) with a mutation at the putative hCWC22/eIF4AIII interface exacerbates the defect in splicing-dependent deposition of eIF4AIII(T334V) with a mutation reported to be in direct contact with mRNA, linking hCWC22 to the process of EJC deposition in vitro. Importantly, the addition of hCWC22(G168Y) affects deposition of eIF4AIII(T334V) without inhibiting splicing or the efficiency of deposition of the endogenous eF4AIII(WT) in the same reaction, demonstrating hCWC22’s specific role in eIF4AIII deposition in addition to its role in splicing. The essential splicing factor CWC22 has, therefore, acquired functions in EJC assembly and NMD during evolution from single-celled to complex eukaryotes.

Messenger ribonucleoprotein (mRNP) complexes play important roles in all aspects of mRNA metabolism (1, 2). One example is the multiprotein exon junction complex (EJC) that is deposited ∼24 nt upstream of mRNA exon–exon junctions (3, 4). The EJC affects downstream events in the lifetime of a metazoan mRNA, including surveillance by the nonsense-mediated decay (NMD) machinery (5, 6), mRNA localization (7, 8), mRNA export from the nucleus (5), and the enhancement of translation attributable to splicing (9). Moreover, EJC components have been shown to be required for proper splicing of MAPK and other long-intron–containing transcripts (10, 11).

Four EJC components [eIF4AIII, human homolog of D. melanogaster mago nashi (Magoh), Y14, and Metastatic Lymph Node 51 (MLN51)] form a stable tetrameric EJC core on the mRNA in vitro. The DEAD-box helicase eIF4AIII plays a key role in clamping and locking the core onto RNA, as a result of inhibition of its ATPase activity by the Magoh/Y14 heterodimer (12–14).

Whereas deposition of the EJC strictly depends on splicing (3, 4), the mechanisms by which the spliceosome mediates EJC formation and defines the position of the EJC on the spliced mRNA are not well understood. The human intron-binding protein IBP160 (15) was shown to recruit EJC components to the spliceosome, even in the absence of the final EJC-binding site on the exon, and was found to be required for EJC deposition and NMD of at least two spliced cytoplasmic RNAs, the U22 host gene (UHG) and growth arrest specific transcript 5 (GAS5). However, it was not established whether the association of EJC components with IBP160 is direct or mediated by other spliceosomal factors.

We recently reported identification of an eIF4G-like binding partner of human eIF4AIII, nucleolar protein with MIF4G domain 1 (NOM1), and demonstrated the role of the eIF4AIII/NOM1 complex in 18S rRNA biogenesis (16). Here, we identify yet another directly interacting eIF4A/eIF4G-like protein pair formed by human eIF4AIII and an eIF4G-like protein (h)CWC22, a core spliceosomal protein (17–20). We show that hCWC22 is required for mRNA splicing and has a specific in vivo role in maintaining low cellular levels of the spliced UHG RNA, which is a natural substrate of NMD. We suggest that the role of hCWC22 in NMD is mediated by its function in EJC deposition and show that eIF4AIII and hCWC22 are linked by demonstrating a synthetic effect of mutations in the two proteins on the process of the EJC deposition in vitro.

Results

eIF4AIII-Binding Protein NOM1 Is Homologous to hCWC22 (Nucampholin), a Spliceosomal Component.

We recently identified the human eIF4G-like protein NOM1 [suppressor of glycerol defect (Sgd1p) in yeast] as a direct binding partner of human eIF4AIII [four A like (Fal1p) in yeast] (16). We further demonstrated that the NOM1/eIF4AIII complex in human and the orthologous Sgd1p/Fal1p complex in Saccharomyces cerevisiae play essential and conserved roles in 18S rRNA biogenesis (16). Interestingly, the central regions of the NOM1 and human eIF4G proteins share domain organization [one middle portion of eIF4G (MIF4G) domain followed by one MA3 domain] with hCWC22 [nucampholin (ncm)], a core human spliceosomal protein (Fig. S1) (17–20). The crystal structure of the yeast eIF4A/eIF4G complex (21) and our identification of the 12-aa eIF4AIII-interacting motif in NOM1 (16) then allowed identification of a conserved sequence in hCWC22 (Fig. 1A). This striking similarity between the putative eIF4AIII-binding interfaces of human NOM1 and hCWC22 suggested the possibility that the hCWC22 and eIF4AIII proteins interact.

Fig. 1.

hCWC22 directly interacts with eIF4AIII in vitro and in vivo; mutations in the MIF4G domain of hCWC22 disrupt the interaction. (A) eIF4AIII-binding sites of CWC22 and NOM1 are more homologous to each other than to the eIF4A-binding sites of eIF4G or death-associated protein 5 (DAP5) as seen from the alignment of their major (12-aa motif) and minor (second motif) binding sites. Positions in eIF4G that directly contact eIF4A in the yeast crystal structure (21) are indicated with black dots. Positions in Sgd1p that resulted in (i) allele-specific suppression and (ii) synthetic interaction with fal1(T322V) (16) are indicated with a triangle and diamond, respectively. (B) Structure of the yeast eIF4G/eIF4A complex (21); binding surfaces are outlined with red ovals. Amino acids that make direct contacts and were used for alanine substitutions in this work are underlined. (C) GST pull-down shows direct physical interaction between human eIF4AIII and hCWC22 recombinant proteins. The black arrow indicates the Coomassie-stained hCWC22 band copurifying with GST-eIF4AIII. The white arrow shows the absence of detectable hCWC22 copurifying with GST-Trm82p, which provides a negative control. GST-eIF4AIII and GST-Trm82p were expressed and purified from yeast, whereas hCWC22 was expressed and purified from E. coli. I1 and I2 denote inputs (100%) for pull-downs shown in lanes 1 and 2 and lanes 3 and 4, respectively. (D) Human eIF4AIII and hCWC22 coimmunoprecipitate from cell extracts; mutations in the MIF4G domain of hCWC22 disrupt the interaction. The Western blot shows anti-FLAG immunoprecipitates from nuclear extracts of HEK293T cells transiently transfected with plasmids expressing CMV-driven tagged versions of eIF4III, hCWC22, and a negative control protein METTL1, as indicated. The IP reactions were performed in the presence of RNase A (Qiagen) as described in the SI Materials and Methods. The FLAG and c-Myc epitopes were detected using mouse monoclonal M2 and HRP-conjugated mouse monoclonal 9E10 antibodies, respectively.

hCWC22 Interacts with eIF4AIII in Vitro and in Vivo, with Mutations at the Putative Interface Disrupting the Interaction.

To determine whether a direct interaction between human eIF4AIII and hCWC22 can be reconstituted using purified proteins, we heterologously expressed and purified GST-tagged human eIF4AIII from S. cerevisiae (Fig. 1C, lane 7) and FLAG-tagged hCWC22 from Escherichia coli (Fig. 1C, lane 8). After mixing the two proteins (Fig. 1C, lane 5), we examined GST-bound eIF4AIII for its ability to select hCWC22. Whereas a majority of hCWC22 was efficiently bound by GST-eIF4AIII (Fig. 1C, lane 2, bound hCWC22 is indicated with a black arrow), a control GST-tagged protein tRNA methyltransferase 82 (Trm82p) (lane 9) similarly purified from S. cerevisiae and mixed with hCWC22 (lane 6) failed to select hCWC22 (compare lanes 2 and 4, black and white arrows). This result confirms a direct and specific interaction between hCWC22 and eIF4AIII in vitro.

Similarly, transiently expressed FLAG-eIF4AIII coimmunoprecipitated Myc-hCWC22 from human HEK293T cell nuclear extract (Fig. 1D, lane 1) but failed to coimmunoprecipitate a mutant variant of hCWC22 in which three key residues of a major eIF4AIII-binding interface (NGLINKVNISNI, where underlined amino acids make direct contacts) (Fig. 1B) were replaced with alanines (NKVN→AAVA) (Fig. 1D, lane 2). Coimmunoprecipitation (co-IP) of hCWC22 containing mutations in the minor eIF4AIII-binding interface (RVQY→AVQA) was diminished (Fig. 1D, compare lanes 1 and 3), whereas hCWC22(WT) was not detectably coimmunoprecipitated by a control protein FLAG-METTL1 (lane 4). These findings confirm the specificity of the hCWC22/eIF4AIII interaction in vivo.

Depletion of hCWC22 Results in a Splicing Defect in Vivo.

Consistent with the role of the yeast ortholog of hCWC22 as an essential splicing factor (19), we observe a significant decrease in the levels of cellular spliced mRNAs [glyceraldehyde-3-phosphate dehydrogenase (GAPDH), β-2-microglobulin (B2M), and hypoxanthine phosphoribosyltransferase (HPRT)1] (Fig. 2B) upon siRNA knockdown of hCWC22 (to less than 30%) in HeLa cells (Fig. 2A). Furthermore, hCWC22 knockdown results in accumulation of an intron-containing human β-globin reporter RNA, as shown by Northern blotting (Fig. S2, lane 2). In all cases, expression of a plasmid-borne siRNA-resistant hCWC22 (phCWC22; Fig. 2A) partially rescued the defect (Fig. 2B, gray bars, and Fig. S2, lane 3), confirming the specificity of the hCWC22 knockdown.

Fig. 2.

Knockdown of hCWC22 decreases levels of spliced cellular mRNAs but increases levels of spliced RNA of UHG, a natural NMD substrate. (A) The Western blot shows the level of hCWC22 in HeLa cells (lane 1) and the results of siRNA-mediated knockdown of hCWC22 without (lane 2) or with expression (lane 3) of its siRNA-resistant variant; α-tubulin provides a loading control. (B) RT-qPCR analysis of the levels of spliced GAPDH, B2M, and HPRT1 mRNAs in the cells used in A. (C) Schematic representation of the positions of starting methionines (Met) and PTCs (shown as stop signs) in the UHG(WT), an NMD-resistant mutant variant UHG(NMD^R), and the NMD-resistant variants into which a single premature stop codon (PTC1 or PTC3) was reintroduced [the 19 stop codons in the last (11th) exon are not expected to elicit NMD]. (D) Human UHG is an efficient NMD substrate. RT-qPCR analysis of spliced transcripts from the WT and mutant UHG constructs diagramed in F was performed with primers complementary to the CMV promoter-specific 5′-untranslated sequence (UHG_q1F) and exon 3 of UHG (UHG_q2R). The resulting qPCR products were gel-purified and sequenced to ensure their identity as spliced UHG RNA with introns 1 (399 nt) and 2 (207 nt) removed. Levels of cotransfected GFP (pmaxGFP) served as a control for transfection efficiency. (E) RT-qPCR analysis of the levels of spliced endogenous UHG RNA in the cells used in A reveals a marked increase upon hCWC22 knockdown. qPCR primers were complementary to exon 2 (UHG1F) and exon 3 (UHG1R) (set 1) or to the junction of exons 9 and 10 (UHG3F) and exon 11 (UHG3R) (set 2).

Depletion of hCWC22 Results in Increased Levels of a Spliced Endogenous NMD Target, the UHG Transcript.

Because NMD in vertebrates requires deposition of the EJC, we also wished to examine the effect of hCWC22 knockdown on a premature termination codon (PTC)-containing spliced RNA that is subject to the NMD surveillance pathway. We chose the UHG transcript, in which the excised introns generate stable small nucleolar (sno)RNA products (22). The spliced UHG RNA, on the other hand, contains ORFs that are interrupted by multiple premature termination codons, which led to the original proposal that the spliced UHG RNA is a naturally occurring endogenous NMD substrate (22).

To unequivocally establish UHG RNA as a natural NMD substrate, we assessed the effect of the PTCs within the wild-type (WT) UHG transcript on the cellular levels of spliced UHG RNA. We constructed an NMD-resistant mutant of UHG, named UHG(NMD^R) (Fig. 2C), by eliminating the first ORF by mutating an AUG in the third exon (AUG→GUG) and then mutating all five natural stop codons in exons 2–10 (PTC1, PTC2, PTC3, PTC4, and PTC5) in the second ORF, which originates from an alternative AUG in the third exon. Because the third potential ORF of UHG RNA has no AUGs until the very last (11th) exon, this frame is not expected to elicit NMD. Moreover, the 19 stop codons (in all three frames) in the last (11th) exon are not expected to elicit NMD because of their location 3′ to the last exon–exon junction. The WT intron–containing UHG gene and its mutant variants were then placed under control of a CMV promoter, and the levels of the resulting exogenous spliced UHG RNA product were assessed in HeLa cells by quantitative RT-PCR (RT-qPCR); primers complementary to neighboring exons were used, resulting in PCR products that include the corresponding exon–exon junctions. Consistent with the observed PTC dependence of UHG RNA levels, we found that the level of spliced “NMD-resistant” UHG(NMD^R) RNA is eightfold higher than that of UHG(WT) (Fig. 2D). Reintroduction of either PTC1 or PTC3 (Fig. 2C) reduced the level of UHG(NMD^R) three- and twofold, respectively (Fig. 2D). These results confirm that the UHG transcript is a genuine NMD substrate.

Interestingly, despite the fact that knockdown of hCWC22 causes a splicing defect that results in a decrease in the levels of the cellular spliced mRNAs tested in Fig. 2B, we observed a threefold increase in the level of the spliced product of the UHG gene upon hCWC22 knockdown (Fig. 2E). This observed increase was in the spliced UHG RNA and not in its unspliced precursor(s), as determined by the size and sequence of the RT-qPCR products. Because a defect in splicing per se cannot generate an increase in the spliced RNA product, these data argue that knockdown of hCWC22 increases the level of spliced UHG RNA by affecting its NMD.

In Addition to Its Critical Role in Splicing, hCWC22 Functions to Couple Splicing to EJC Deposition.

Consistent with the essential role of the yeast ortholog of hCWC22 in splicing (19), knockdown of hCWC22 in HeLa cells results in a splicing defect (Fig. 2B). Because deposition of the EJC on a nascent mRNA is splicing-dependent (3, 4), knockdown of hCWC22 is, therefore, expected to decrease the level of EJC deposition simply because of the decreased availability of splicing-generated exon–exon junctions. However, direct interaction of the core EJC component eIF4AIII with the human spliceosomal protein hCWC22 in vitro and in vivo, as well as the increased levels of spliced endogenous NMD substrate UHG RNA upon hCWC22 knockdown, raises the possibility that in addition to its role as a splicing factor, hCWC22 may be the missing link between the processes of splicing and EJC deposition.

To demonstrate such a role for hCWC22, its functions in (i) splicing and (ii) EJC deposition must be separated by identifying mutations in hCWC22 that (i) do not affect splicing but (ii) do affect EJC deposition. Because multiple simultaneous substitutions at the major and minor eIF4AIII-binding surfaces of hCWC22 (the 12-aa motif and RVQY domains, respectively; Fig. 1B) resulted in a pronounced growth defect in HeLa cells (see SI Materials and Methods), they are not helpful in uncoupling splicing from EJC deposition. Instead, we searched for point mutations at the major eIF4AIII-binding surface of hCWC22 that (i) do not affect splicing but (ii) affect deposition of an eIF4AIII mutant that is itself partially defective in assembling EJCs because of a T304V substitution at its mRNA-binding surface (Fig. 1B). The eIF4AIII(T304V) mutant was chosen for three reasons, all based on the predicted structural similarity of hCWC22 and yeast Sgd1p to eIF4G, as well as that of human eIF4AIII and yeast Fal1p to eIF4A. First, the hydroxyl group of threonine corresponding to eIF4AIII T304 was shown previously to interact directly with the RNA phosphate backbone in the initial crystal structure of the Drosophila melanogaster helicase VASA (23) and, subsequently, in both crystal structures of the EJC core bound to RNA (12, 14). Second, the T322V substitution in Fal1p, the yeast eIF4A-like helicase, which is homologous to the T304V substitution in eIF4AIII, was only partially (as opposed to entirely) disruptive for function; the resulting growth and 18S rRNA biogenesis defects of fal1(T322V) were more pronounced at low temperature (16 °C) (16). Third, the growth and 18S rRNA biogenesis defects of the T322V substitution in yeast Fal1p were either rescued or exacerbated by point mutations in the eIF4G-like yeast protein Sgd1p (16); the rescuing/exacerbating mutations remarkably mapped to the major Fal1p/Sgd1p-interacting surface (equivalent to eIF4A-interacting 12-aa motif in eIF4G) just 2 aa apart from each other. Because eIF4AIII itself was shown previously not to be required for splicing of the adenovirus major late (AdML) pre-mRNA substrate in vitro (24), we reasoned that we might observe a compensatory or synthetic effect between the hCWC22(G168Y) mutation at the eIF4AIII-binding interface, which corresponds to allele-specific suppressor mutation in Sgd1p (Fig. 1A) (16), and the eIF4AIII(T304V) mutation. Such an effect, in the clear absence of an effect of the hCWC22 mutation on splicing, would then be suggestive of a specific role of hCWC22 in EJC deposition.

We performed in vitro splicing reactions with biotinylated AdML RNA substrate using HeLa nuclear extract supplemented with total extracts of HEK293T cells overexpressing hCWC22, hCWC22(G168Y) (∼2.5-fold; Fig. S3A), or FLAG-tagged versions of eIF4AIII and eIF4AIII(T304V). We used a protocol derived from the splicing-dependent EJC IP (EJIPT) assay (25) in which interaction of endogenous eIF4AIII with spliced mRNA provides a quantification of both EJC formation and splicing (25). We then isolated the RNA from the in vitro splicing reactions (with associated mRNP complexes) using streptavidin beads and analyzed the protein composition using Western blots (Fig. 3B). In this experiment, the extent of deposition of the endogenous WT eIF4AIII serves as a sensitive indicator of any effect of the hCWC22 mutation on splicing in the same reaction. Based on the temperature dependence of the fal1(T322V) substitution in yeast, the splicing reactions were performed at 20 °C to enhance the putative deposition defect of the eIF4AIII(T304V).

Fig. 3.

Synthetic effect of mutations hCWC22(G168Y) and eIF4AIII(T304V) results in decreased EJC deposition on AdML RNA without affecting splicing in vitro. (A) Schematic of the experiment [based on the EJIPT assay (25)] that simultaneously detects (i) the efficiency of deposition of FLAG-tagged eIF4AIII variants on spliced biotinylated AdML RNA and (ii) the efficiency of splicing in the same reaction. Whereas the deposition efficiency of FLAG-tagged eIF4AIII variants on biotinylated RNA differs, the equivalent deposition of endogenous eIF4AIII provides a sensitive measure of splicing. (B) Western blots show FLAG-tagged and endogenous eIF4AIII coprecipitated with biotinylated AdML RNA after in vitro splicing at 20 °C for 12 h. HeLa nuclear extract (60 μL) (33) was supplemented with 25 μL each of two HEK293T whole-cell extracts (34) expressing one FLAG-eIF4AIII, one hCWC22 variant, or a control green fluorescent protein (−), as indicated. Intronless biotinylated AdML mRNA substrate was used in place of the AdML pre-mRNA in lane 10 (which was otherwise identical to lane 8) to determine the background level of splicing-independent coprecipitation. Total eIF4AIII and the FLAG-tagged eIF4AIII were detected as described in Materials and Methods. (C) The bar graph summarizes quantification of three blots as shown in B; quantification of individual blots (Fig. S5) showed the synthetic effect in each experiment.

Consistent with partial functional defects observed for yeast fal1(T322V), the efficiency of deposition of the FLAG-eIF4AIII(T304V) on biotinylated AdML mRNA was lower than the efficiency of deposition of WT FLAG-eIF4AIII(WT) [Fig. 3B, compare the IP lanes 7 and 8; quantified in Fig. 3C]. Moreover, addition of the hCWC22(G168Y) mutant protein further decreased the amount of the FLAG-eIF4AIII(T304V) deposited on the spliced AdML RNA (Fig. 3B, compare the IP lanes 6 and 7; quantified in Fig. 3C). These results show a synthetic effect between the eIF4AIII(T304V) and hCWC22(G168Y) mutations in EJC deposition: whereas the hCWC22(G168Y) mutation alone is not sufficient to affect deposition of eIF4AIII, the combination of two mutations [hCWC22(G168Y) with FLAG-eIF4AIII(T304V)] disturbs the process of EJC deposition enough to yield a negative effect. Importantly, deposition of the endogenous WT eIF4AIII protein in these reactions [eIF4AIII(endog), marked by an arrow in Fig. 3B] was unchanged by the addition of mutant hCWC22(G168Y) (Fig. 3B, lanes 6, 7, and 8), demonstrating that the splicing efficiency was not decreased (splicing of [³²P]U body–labeled AdML RNA in these extracts is shown in Fig. S4). The levels of splicing-independent IP of the endogenous eIF4AIII and the exogenous FLAG-eIF4AIII were negligible, as shown in Fig. 3B, lane 10, where an intronless AdML mRNA substrate was used in place of the AdML pre-mRNA in the splicing reaction [as in the EJIPT assay (25)]. We conclude that the hCWC22(G168Y) and eIF4AIII(T304V) mutations exhibit synthetic effects, arguing that action of the two proteins is linked during EJC deposition.

Discussion

Here, we have demonstrated that the human spliceosomal protein hCWC22: (i) is an eIF4AIII-binding protein in vitro and in vivo; (ii) is essential for splicing, as is its yeast counterpart, Cwc22p; (iii) plays a role in the in vivo degradation of an endogenous spliced NMD target, the UHG transcript; and (iv) can be functionally linked through the synthetic effect of mutations to the process of EJC deposition in vitro. While this manuscript was in preparation, two other reports appeared showing that hCWC22 interacts with eIF4AIII (26, 27) and is necessary for splicing (27), but neither paper provides evidence for the role of hCWC22 in NMD or for its collaboration with eIF4AIII during EJC deposition.

Whereas the participation of eIF4A/eIF4G in translation initiation is well established (28), it was not known whether the core EJC component eIF4AIII, a helicase with more than 60% identity to eIF4A (29), has an eIF4G-like partner(s). We reported recently that eIF4AIII interacts with an eIF4G-like partner, NOM1, and that this eIF4A/eIF4G-like protein pair plays a crucial role in 18S rRNA biogenesis in human cells (16). Finding NOM1 prompted us to search for possible interactions of other eIF4G-like proteins with eIF4AIII. In particular, NOM1 and a little-studied human spliceosomal protein hCWC22 (17–20) exhibited not only similar domain topology (30) (Fig. S1) but also homology between their putative eIF4A-binding surfaces. In fact, conservation at the eIF4A-binding interface is more pronounced between NOM1 and hCWC22 (Fig. 1A) than between either of these proteins and eIF4G, suggesting interaction with a common partner (eIF4AIII). Our findings of a direct physical interaction between eIF4AIII and hCWC22, together with the observations that mutations in the MIF4G domains of hCWC22 alter its association with eIF4AIII (Fig. 1) and that synthetic mutations at the eIF4III-binding interface of hCWC22 affect EJC deposition (Fig. 3), provide strong evidence that eIF4AIII/hCWC22 represents a functional eI4FA/eIF4G-like protein pair. Thus, the roles of eIF4A/eIF4G-like pairs include EJC deposition and mRNA surveillance, in addition to translation initiation and ribosomal RNA biogenesis.

The dual role of hCWC22 in splicing and EJC deposition raises the question of how mutually dependent these two processes are. Whereas EJC deposition depends on splicing (3, 4), completion of splicing (or a subsequent step such as mRNA release), at least in theory, could conversely depend on completion of EJC deposition. This is not true of in vitro splicing because eIF4AIII is dispensable for the generation of spliced products (24). However, the possibility of a splicing/EJC deposition checkpoint in vivo remains. Intriguingly, the recently reported role of eIF4AIII in the splicing of mRNAs with long introns (MAPK) (10, 11) may be explained by a direct role of EJC components in the splicing of certain introns. Alternatively, these observations may reflect the action of a checkpoint that allows completion of splicing of certain introns only after correct EJC deposition in vivo. Mutations in hCWC22 that uncouple splicing and EJC deposition should allow distinguishing between these two possibilities.

The extent of hCWC22 involvement in degradation of other natural NMD substrates (31, 32) is, likewise, unknown at this time. Our data show that for at least one substrate (UHG RNA), the functions of hCWC22 in splicing and EJC deposition may not be strictly interdependent. Splicing of the UHG transcript can proceed after hCWC22 depletion, but NMD and, presumably, EJC deposition are affected, perhaps because of a differential requirement of hCWC22 for splicing of different pre-mRNAs. Alternatively, knockdown of hCWC22 may lead to competition between the many resulting unspliced mRNAs for limiting NMD factors; IBP160, an intron binding protein that was previously implicated in recruiting EJC components to the intron before deposition on the upstream exon, could be one such protein. Future studies will be needed to address the degree of hCWC22-dependent interplay between splicing and EJC deposition and to illuminate the involvement of hCWC22 in degradation of numerous PTC-containing mRNAs associated with human genetic diseases.

Materials and Methods

Protein Purification and IP.

For information regarding protein purification and IP, see SI Materials and Methods.

Transfections and siRNA-Mediated Knockdown.

siRNA-mediated knockdown of hCWC22 (with or without expression of its siRNA-resistant variant) was achieved by electroporation of HeLa cells (twice, 48 h apart) resuspended in RPMI medium using Gene Pulser II (Bio-Rad) at 210 V and 975 μF, using a 0.4-cm gap; cells were immediately transferred into DMEM media supplemented with 10% (vol/vol) FBS and 2 mM L-glutamine. siRNA and plasmid concentrations were 5 nM and 0.5 μg/mL, respectively. hCWC22-specific Silencer Select siRNA s33631 (5′-GGGCGAUUUUGCAUGCUAAtt-3′ and 5-UUAGCAUGCAAAAUCGCCCag-3) and Negative Control #1 siRNA were from Ambion. Knockdown of hCWC22 was assessed by Western blotting using purified MaxPab mouse polyclonal antibody KIAA1604 (B01P) H00057703-B01P from Abnova.

RT-qPCR and Northern Blot.

HeLa cells were lysed in TRIzol reagent (Invitrogen), and RNA was isolated according to the manufacturer’s protocol, treated with 2 units of RNase-free RQ1 DNase (Promega), PCA-extracted, and ethanol-precipitated; cDNA was made with random primers using SuperScript III Reverse Transcriptase (Life Technologies) according to the manufacturer’s protocol. qPCR was performed using FastStart Universal SYBR Green Master mix (Roche) and StepOnePlus real-time PCR system (Applied Biosystems) and quantified using StepOne software. For Northern blots, RNA was separated on formaldehyde agarose, transferred to a Zeta-Probe membrane (Bio-Rad), cross-linked using a UV254 (Stratalinker 2400), and visualized with [³²P]U body–labeled RNA complementary to spliced β-globin or synthetic 5′-end ³²P-labeled DNA oligonucleotide probes.

Extract Preparation, in Vitro Splicing, RNA Isolation, and eIF4AIII Detection.

Nuclear extracts were prepared from HeLa cells as described (33). Whole-cell extracts were prepared from HEK293T cells as described (34). Biotinylated mRNA substrates were prepared as described (25). In vitro splicing reactions with biotinylated pre-mRNA were performed as follows: in vitro splicing reactions consisting of 13.6 μL of 10× SP buffer [5 mM adenosine 5′-triphosphate (ATP), 200 mM creatine phosphate, 24 mM MgCl₂], 5 nM biotinylated RNA (AdML pre-mRNA or intronless AdML mRNA), 0.7 units per μL SUPERaseIN, 60 μL of HeLa nuclear extract, and 25 μL each of two HEK293T whole-cell extracts expressing one FLAG-eIF4AIII (containing 3× FLAG on the N terminus), one hCWC22 variant, or a control green fluorescent protein were assembled on ice and incubated at 20 °C for 12 h. Reaction mixtures were then diluted with 564 μL of HNT buffer [20 mM Hepes-KOH, 150 mM NaCl, 0.5% (wt/vol) Triton X-100] and incubated with M-280 streptavidin-coated magnetic beads at 25 °C for 1 h. Beads were washed 10 times with HNT buffer at 25 °C; RNA-bound proteins were eluted from beads by the addition of 3× SDS loading buffer and boiling for 5 min. Eluted proteins were resolved by 4–20% (wt/vol) SDS/PAGE and transferred to a nitrocellulose membrane. Total eIF4AIII and FLAG-tagged eIF4AIII were detected using mouse monoclonal anti-eIF4AIII antibody [provided by Adrian Krainer (Cold Spring Harbor Laboratory, Cold Spring Harbor, NY)] (24) and mouse monoclonal anti-FLAG M2 antibody, respectively. Western Lightning Plus-ECL Enhanced Chemiluminescence Substrate from PerkinElmer and SuperSignal West Femto Maximum Sensitivity Substrates from Pierce were used for detection. Chemiluminescence was captured using the G:BOX (Syngene) system and quantified using GeneSys software. In vitro splicing with [³²P]U body–labeled pre-RNA AdML substrate was similarly performed in HeLa nuclear extract at 20 °C.