Thirdly our data on Chtop

Thirdly, our data on Chtop participating in APA regulation (Figure 4) support the view that the co-transcriptional loading of RNA export factors can influence 3′ end processing in human cells. Because Cpsf6 is a master regulator of APA in the cell, we looked at the set of transcripts displaying APA changes in both siCPSF6 (Martin et al., 2012) and siCHTOP conditions. It was restricted to only 21 transcripts, revealing the complex interplay taking place on the last exons to establish the pattern of pA sites used in the cell. The broad distribution of Chtop on the last exons might assist with restricting the spatial deposition of the CPAC. Nxf1 was upregulated in response to Chtop RNAi (Figure S3B), and Nxf1 RNAi triggers hyperadenylation of reporter transcripts (Qu et al., 2009). These results reinforce the functional connection we have previously uncovered between Nxf1 and Chtop (Chang et al., 2013). Consistent with involvement of Chtop in 3′ end-processing regulation, it preferentially coIP Ser2-phosphorylated Pol II (Figure 2D), and its overexpression specifically led to transcriptional readthrough downstream of its own gene (Figure S3E). Perhaps this constitutes another mechanism used by Chtop to auto-regulate its expression (Figures S1A and S3E) (Izumikawa et al., 2016). Although the EJC wasn’t originally implicated in RNA export, those data were obtained either in vitro (Cheng et al., 2006) or before the discovery of the EJC’s anchor eIF4A3. Only the combined depletion of all EJC subunits known at the time triggered a partial block of export, and the co-depletion of Alyref and these EJC proteins is lethal in Drosophila Cefepime (Gatfield and Izaurralde, 2002). Our study confirms a key functional link between the EJC and Alyref. In agreement, Alyref binding and enrichment patterns mirrored those of the EJC, whereas Chtop and Nxf1 were more alike (Figures 5A and S4A, and Figures 4B, S3A, and S5E). In addition, depletion of either Alyref or eIF4A3 increases the levels of Chtop. Both cases could be a cellular response to the loss of Alyref on the RNA, because Alyref works with Chtop to mediate TREX function (Figure 6D) (Chang et al., 2013). Because endogenous Alyref interacts poorly with EJC components (Cheng et al., 2006, Chi et al., 2013) and isn’t strongly enriched in the EJC interactome (Singh et al., 2012), its proximity to the EJC is perhaps achieved through RNA-enhanced protein-protein interactions. Yet our binding assays suggest that an EJC-driven  may contribute to the transfer of Alyref from the CBC next to the EJC (Figures S6D and S6E). We can reconcile Alyref 5′ enrichment in Figure 1F with its enrichment at EJC sites as the 5′ enrichment composed of Alyref strong binding to the 3′ end of a transcript’s first exon, because this junction is spliced first and is closest to the CBC. Surprisingly, our work has revealed the presence of the EJC on intronless RNAs. Our results are consistent with the ability of EJC subunits to bind nascent transcripts independently of pre-mRNA splicing in Drosophila (Choudhury et al., 2016). Moreover, eIF4A3 enhances translation of intronless mRNAs by binding to their 5′ end (Choe et al., 2014) (Figure S5F). The recruitment of splicing factors to intronless mRNAs is not unprecedented and could result from a scanning mechanism of the pre-mRNA by spliceosome components (Änkö et al., 2012, Brody et al., 2011). The whole EJC is associated with the 5′ exon before exon ligation, and eIF4A3 is the only EJC core protein to directly interact with spliceosome subunits (Zhang et al., 2017). In addition, eIF4A3 doesn’t seem to recognize exons as well as Casc3 (Figures 5D and S5A) or their positions within the transcript (Figure S5D). Therefore, it could be involved in such a scanning process. Intronless RNAs are notoriously poorly exported and are subject to greater exosome-mediated degradation than regular mRNAs (Dias et al., 2010, Fan et al., 2017). Alyref and the EJC were both more enriched on multi-exons than on single-exon RNAs (Figures 5D and S5A), whereas Chtop and Nxf1 were not discriminative (Figure S5A). Given that the major mRNA export adaptor Alyref seems to help RNAs evade nuclear degradation by the exosome (Fan et al., 2017), it is plausible that Alyref’s loading might be a rate-limiting step for the efficient export of intronless RNAs. In contrast, the single-exon histone mRNAs, which bind TREX subunits (but not the EJC) just as strongly as multi-exon RNAs, may have evolved more efficient ways to recruit RNA export factors (Figures 5D and S5A).