Characterization of the Regulation of CD46 Alternative Splicing

Here we present a detailed analysis of the alternative splicing regulation of human CD46 , which generates different isoforms with distinct functions. CD46 is a ubiquitous membrane protein which protects host cells from the complement and plays other roles in immunity, autophagy and cell adhesion. CD46 deficiency causes an autoimmune disorder, and this protein is also involved in pathogen infection and cancer. Before this study, the mechanisms of CD46 alternative splicing remained unexplored, even though dysregulation of this process has been associated with autoimmune diseases. We proved that the 5' splice sites of CD46 cassette exons 7 and 8 encoding extracellular domains are defined by noncanonical mechanisms of base-pairing to U1 small nuclear RNA. Next we characterized the regulation of CD46 cassette exon 13, whose inclusion or skipping generates different cytoplasmic tails with distinct functions. Using splicing minigenes, we identified multiple exonic or intronic splicing enhancers or silencers, which regulate exon 13 inclusion via trans -acting splicing factors like PTBP1 and TIAL1.

messenger RNA (25)(26)(27). Cassette exons 7, 8 and 9 encode the extracellular domain known as serine, threonine and proline rich (STP) region, which is the binding site for Neisseria gonorrhoeae and Neisseria meningitides. The distinct STP isoforms preferentially trigger different complement regulatory pathways and exhibit different bacterial adherence affinities (20,28,29). On the other hand, inclusion or skipping of cassette exon 13 respectively generate two mutually exclusive cytoplasmic tails, denoted as CYT1 and CYT2, which are encoded by either exon 13 or exon 14. These two cytoplasmic tails have different binding partners and regulatory functions in autophagy, Tcell activation as well as different posttranslational processing times (7,10,11,(30)(31)(32)(33). The proteolytic cleavage of CYT1 upon co-activation of CD46 induces interferon-γ (IFNγ) production in T helper 1 (T H 1) cells, and subsequent differentiation to T regulatory 1 (Tr1) cells, which typically secrete the anti-inflammatory cytokine IL-10. CYT2 is processed at a later time to terminate the anti-inflammatory Tr1 response and restore homeostasis (31,32). The expression of both cytoplasmic tails is essential for modulating the T H 1 response, because defective CD46induced Tr1 differentiation or Tr1-induced IL-10 production is associated with pathogenesis of several autoimmune diseases like asthma, rheumatoid arthritis and multiple sclerosis (15,17,18). This also highlights the importance of proper splicing regulation to generate both CD46-CYT1 and CD46-CYT2.
Splicing is catalyzed by the spliceosome, and regulated by cis-acting elements, trans-acting factors, transcription and chromatin structure (34)(35)(36)(37)(38). During alternative splicing, exons are defined by their 5' and 3' boundaries or splice sites, as well as by multiple auxiliary cis-acting splicing elements (39). Short RNA sequences that act as splicing enhancers or silencers provide binding sites for protein regulators to either promote or repress exon inclusion. The regulatory mechanisms of these trans-acting factors include direct interactions with the spliceosome, inhibition of exon or intron definition, or looping out an exon (34,(40)(41)(42)(43)(44). The regulation of alternative splicing is complex and context dependent and thus it requires intensive studies. Here we performed serial deletions in splicing minigenes to identify enhancers and silencers that regulate exon 13 inclusion. Additionally, we identified several proteins that likely act via these cis-acting elements and elucidated their functions in alternative splicing of exon 13 by loss-and gain-of function assays. This is the first report to describe the regulation of CD46 alternative splicing.

RESULTS
Alternative splicing of CD46 is tissue selective-A previous study reported that CD46 splice isoforms are differentially expressed across tissues (45). We verified these findings by analyzing the alternative splicing patterns of CD46 in 20 human tissues using semi-qRT-PCR. Both CD46 cytoplasmic tail isoforms were detected in all tissue RNAs with a predominance of CYT2 by exon 13 skipping, suggesting that inclusion of exon 13 is generally repressed (Fig. 1A, upper panel). The exon 13 inclusion level varied in different tissues, ranging from 6% to 47%.
At least two STP isoforms were observed in all tissues, and the two predominant isoforms are inclusion of a single exon and of two exons (Fig.  1A, lower panel). The identities of the included exons were determined by PCR with exon 7-, 8-or 9-specific primers. The exon that was included alone is exon 9 giving rise to the C isoform, while exons 8 and 9 were included together in mRNA giving rise to the BC isoform. Most tissues predominantly express either C or BC isoforms, while nearly equivalent expression of both C and BC isoform is seen in prostate, thyroid and trachea (Fig. 1A, lanes 13 and 19-20). Few tissues, such as bladder and brain, show a faint band corresponding to skipping of all three STP exons (Fig. 1A, lanes 2-3), and a low level of exon 7 containing isoforms (with all three STP exons) was detected in lung, ovary, placenta and testes (Fig. 1A, lanes 10-12 and 17). Furthermore, there was virtually no correlation for the inclusion of exons 8 and 13 across the 20 tissues (data not shown), suggesting that the regulation of these two cassette exons is independent. Largely consistent with previous findings (45), CD46 isoform expression changes across different tissues and the two most predominant isoforms are C-CYT2 and BC-CYT2.
The 5' splice sites (5'ss) of exons 7 and 8 are recognized by U1 using asymmetric loop 1 register-The exons 7 and 8 5'ss sequences substantially deviate from the consensus, and they can only establish 5 Watson-Crick base pairs with U1 snRNA in the canonical register, which is the main mechanism of 5'ss selection (Fig. 1B) (46,47). However, if two nucleotides at 5'ss positions +3 and +4 and one nucleotide at U1 snRNA position 6 form an asymmetric loop in the U1/5'ss helix, this allows formation of 4 additional base pairs thus stabilizing this interaction (Fig. 1B). We hypothesized that these 5'ss are recognized by this register that we named asymmetric loop 1 (+3/+4), which indicates the difference of the looped nucleotides and the looped 5'ss positions in parenthesis (48).
We performed mutational and suppressor U1 experiments to test if these two 5'ss employ the asymmetric loop register for U1 recognition (46,(48)(49)(50). Exon 7 was very poorly recognized and included in minigene mRNA (Fig. 1C, lane 1), yet its inclusion could be improved by mutating position +1 of the 3'ss of this exon (first exonic nucleotide) from uridine (non-consensus) to guanosine (consensus) (Fig. 1C, lane 2) (51). As expected, the 5'ss mutation at position -2 in both exons 7 and 8 disrupted one base pair in both registers, causing reduction in exon inclusion ( ). This observation suggests that these two 5'ss use the asymmetric loop register for U1 recognition. The effects brought by these mutations were only rescued by suppressor U1 snRNAs that restore base-pairing in the noncanonical register (Fig. 1C, lanes 7-9 and 15-17). Altogether, these data show that the 5'ss of exons 7 and 8 base-pair with U1 in the asymmetric loop 1 (+3/+4) register.
Exon 13 inclusion is regulated by multiple enhancers and silencers-Due to the biological significance of CD46 exon 13 in immunity and other cellular processes, from here on we focused on the splicing regulation of this exon. Serial deletions of exon 13 were introduced into the CD46 splicing minigene containing exons 12, 13 and 14 with shortened introns 12 and 13 to determine the importance of these sequences in splicing. The majority of splice products from the wild-type CD46 minigene include exon 13 ( Fig.  2A, lane 1) upon HEK293T transfection. All 8nucleotide exonic deletions in the A series affected the alternative splicing pattern of CD46 exon 13 to different extents ( Fig. 2A, top left panel). Deletion of A2, A3, A4 or A7 sequences resulted in nearly complete skipping of exon 13. This result strongly suggests that these sequences contain exonic splicing enhancer/s (ESE/s) which promote inclusion of exon 13. In contrast, A5, A6, A8, A9, A10 and A11 likely contain exonic splicing silencer/s (ESS/s), as exon 13 inclusion was largely increased when either of these sequences was removed. The staggered 8-nucleotide deletions of exon 13 or B series were then generated to further map the regions of ESEs and ESSs, and to limit the false positives from the first deletion set that could arise from the creation of new sequence junctions ( Fig. 2A, top right panel). These staggered deletions showed consistent splicing patterns compared to the first deletion set, strongly supporting the isolation of multiple ESEs and ESSs within exon 13.
Likewise, eight 10-nucleotide deletions in the C and D series were introduced into introns 12 and 13 near exon 13, respectively. These deletions do not perturb the 3'ss, the predicted branch point sequence (BPS) and the 5'ss, thus leaving intact the last 32 nucleotides of intron 12 and the first 8 nucleotides of intron 13 (Fig. 2B). The majority of deletions in either intron did not trigger much change in exon 13 inclusion ( Fig. 2A, lower panels). Nevertheless, a strong intronic splicing enhancer (ISE) was revealed in intron 13 immediately downstream of the 5'ss in exon 13, as its removal led to a strong reduction in exon inclusion ( Fig. 2A, bottom right panel, lane 13 vs. 14 and 15). Overall, these results suggest that the sequences of introns 12 and 13 near exon 13 play a minor role in regulating exon 13 inclusion. Consistent results were seen across HEK293T, HeLa and Jurkat cell lines (Fig. 2, C and D), suggesting that these cis-acting elements are similarly regulated in different cell types.
To conclude, exon 13 contains two strong ESEs and two strong ESSs which are denoted as ESE1, 2 and ESS1, 2, respectively. These ESEs and ESSs are alternatively arranged in the exon. A strong ISE is located downstream of exon 13 5'ss in addition to multiple weak ISEs and ISSs within the proximal intron 12 and 13 sequences (Fig. 2B). The strength of each element is not absolute but rather relative to other elements (see below).
Validation of cis-acting elements-Sets of nucleotide substitutions in ESE1 and ESS1 were generated to further confirm these cis-acting elements and to be used as a negative control for RNA pull-down (Fig. 3A). The point mutations were designed such that they disrupt the predicted protein binding without creating a novel cis-acting element by using the prediction tool Human Splicing Finder (52). The point mutations in ESE1 abolished exon inclusion while the point mutations in ESS1 caused complete exon inclusion, suggesting the disruption of an enhancer or silencer (Fig. 3B, lane 1 vs. 4-6 and 7-11). This result further validates ESE1 and ESS1 found in the deletion assays.
The pSXN splicing minigene plasmid containing a weak and small alternative exon was used to examine the enhancing effects of ESEs identified in the previous serial deletions (Fig. 3D) (53). ESE1 was shown to be a very strong enhancer as insertion of this sequence into the weak exon caused almost complete exon inclusion, which in turn was abolished by mut 5 (Fig. 3C, lanes 7-8). However, ESE2 appeared to include a weak enhancer as only a very faint inclusion band was detected after inserting this sequence into the alternative exon (Fig. 3C, lanes 9-10). The inclusion of this ESE2-containing alternative exon was improved by changing the 5'ss sequence for better base-pairing to U1 snRNA (Fig. 3C, lane 11). The mut 1 in ESE2 in the context of such improved 5'ss also strongly reduced exon inclusion (Fig. 3C, lane 12). The insertion of ESS1 and ESS2 into pSXN did not promote exon inclusion (Fig. 3C, lanes 3 and 5). This result suggests that the inclusion of ESE1/2 containing exons was not due to the increased exon size but rather to the presence of enhancers, and also that the ESS sequences probably possess silencing ability. This heterologous minigene experiment also validates the enhancers and silencers found in previous experiments.
Identification of splicing regulators through RNA-pulldown-Several trans-acting factors bound to the ESS1 and ESE1 were identified by RNA pull-down using the wild-type and mutated or ASO-annealed ESS1 and ESE1 RNA sequences and HeLa nuclear extract. Binding of SRSF1, hnRNP M and PTBP1 to ESS1 was detected by western blot, and the binding efficiency of the former two proteins was diminished in ESS1 mut1 ( Fig. 3E, left panel). Knockdown and overexpression of hnRNP M did not change exon 13 inclusion levels (data not shown). SRSF1 also interacted with ESE1 and the binding was reduced by ASO-mediated blocking of the enhancer (Fig.  3E, right panel). The roles of SRSF1 and PTBP1 in CD46 exon 13 inclusion were next examined.
SRSF1 and PTBP1 negatively regulate exon 13 inclusion through ESS1-SRSF1 was predicted to bind to exon 13 ESE1 region while two PTBP1 consensus binding sequences were predicted within ESS1 and ESS2 using Human Splicing Finder, SF map and Splicing Rainbow (52,54,55). The interaction between SRSF1 and ESS1, and between PTBP1 and ESS1 were observed in the previous pull-down assays (Fig. 3E). Therefore, loss-and gain-of-function assays were performed to test the predicted role of SRSF1 as an activator and PTBP1 as a repressor of exon 13 inclusion. Unexpectedly, endogenous exon 13 was included at higher levels in SRSF1-depleted cells by siRNAs (Fig. 4A, lane 1 vs. 2), which is contradictory to the prediction. The increase of exon 13 inclusion by SRSF1 depletion was partially reversed by exogenous SRSF1 (Fig. 4A, lanes 3-4). Supporting our experiment, RON exon 11 inclusion also increased upon SRSF1 knockdown as previously documented (data not shown) (56,57). These observations suggest that SRSF1, instead of activating, is repressing CD46 exon 13 inclusion. Furthermore, PTBP1 knockdown cells showed a small increase in endogenous exon 13 inclusion (Fig. 4B, lane 1 vs. 2), which was reversed by adding back exogenous PTBP1 (Fig. 4B, lanes 3-4). PTBP1 knockdown also increased inclusion of PLOD2 exon 14 and PTBP2 exon 10, as published (data not shown) (58,59). However, CD46 exon 13 inclusion did not change in either SRSF1 or PTBP1 overexpressing cells, consistent with some SRSF1-regulated splicing events only responding to knockdown (60) (Fig. 4, A and B, lanes 5-6). These data suggest that both SRSF1 and PTBP1 act as repressors of exon 13 inclusion.
To test whether SRSF1-mediated exon 13 repression occurs via this exon and its proximal flanking introns, this region was cloned into a heterologous minigene containing MCAD constitutive exons 8 and 10 and flanking intronic sequences (Fig. 4C). In this chimeric context, the depletion or enrichment of SRSF1 affected the alternative splicing of exon 13 in the same direction as that of endogenous CD46 (Fig. 4C, lanes 1-4), while the inclusion of MCAD exon 9 in the wild-type minigene did not respond to the changes in SRSF1 (Fig. 4C, lanes [5][6][7][8]. This result strongly suggests that SRSF1 regulates the splicing through exon 13 and/or its nearby flanking introns. Further HEK293T cotransfection with overexpression plasmids and CD46 exon 13 deletion minigenes suggested that SRSF1 acts via A6 and B9, and PTBP1 partially via A6 and A10 regions (data available upon request).
TIA1 and TIAL1 strongly activate exon 13 inclusion-TIA1 and TIAL1 induce inclusion of exons with weak 5'ss by binding to the U-rich sequences downstream of these 5'ss, thereby recruiting U1 small nuclear ribonucleoprotein (snRNP) via direct interaction with the U1-snRNP specific polypeptide U1-C (40,61,62). CD46 exon 13 has a relatively weak 5'ss followed by a stretch of U-rich sequence, and removal of this U-rich sequence strongly suppressed exon inclusion ( Fig.  2A, D series, lanes 14-15), making it a potential binding site for TIA1/TIAL1. Overexpression of TIA1 and TIAL1 both increased endogenous exon 13 inclusion, yet TIAL1 seemed to have a stronger activity (Fig. 4D, lane 1 vs. 2 and 3). Consistently, TIAL1 knockdown substantially decreased exon 13 inclusion, confirming the role of TIAL1 in exon 13 inclusion (Fig. 4D, lane 1 vs. 7). As published, TIAL1 knockdown also decreased inclusion of PLOD2 exon 14 and ANLN exon 9 (data not shown) (63). In addition, the repression caused by TIAL1 knockdown was partially reversed by TIA1 overexpression and TIAL1 reconstitution (Fig. 4D, lanes 8 and 9). In contrast, exon 13 inclusion level in TIA1-depleted cells was similar to the control (Fig. 4D, lane 1 vs. 4). The double knockdown of TIA1 and TIAL1 had more impact on exon 13 inclusion than each single knockdown, and their effects could be just weakly rescued by individual reconstitution of TIA1 or TIAL1 (Fig. 4D, lanes 10-12). These data indicate that both TIA1 and TIAL1 induce exon 13 inclusion, with TIAL1 likely as the major activator. HEK293T co-transfection with overexpression plasmids and mutant minigenes suggested that TIA1 and TIAL1 act via the strong U-rich ISE immediately downstream of the 5'ss in exon 13 (data available upon request).

Manipulation of exon 13 inclusion by
ASOs-Two ASOs with 2'O-methoxyethyl and phosphorothioate backbones were designed for each ESE1 and ESS1 to modulate alternative splicing of exon 13, and the annealing positions of the two ASOs differed by 4 or 5 nucleotides ( Fig.  2B and 5A). As expected, both ESE1.1 and ESE1.2 ASOs efficiently blocked endogenous exon 13 recognition, causing nearly complete exon skipping in Jurkat cells (Fig. 5B, lane 1 vs. 2 and  3). In turn, only ESS1.2 ASO weakly improved exon inclusion as intended, while the ESS1.1 ASO unexpectedly repressed exon inclusion (Fig. 5B, lane 1 vs. 4 and 5). The repressive effects of ESS1.1 could be explained by interference with the upstream enhancer ESE2, as this ASO partially covers the ESE2-adjacent B8 region whose deletion showed mildly enhancing effects in HEK293T and Jurkat.
The enhancing or repressing effects of these ASOs on exon 13 inclusion are dose-dependent in Jurkat cells (Fig. 5C). Treatment with 25 pmol of ESE1.2 ASO resulted in only 1% of exon inclusion, which was completely eliminated upon increase to 100 pmol (Fig. 5C, lanes 4-6). ESS1.2 ASO weakly promoted exon inclusion (by 5%) at 25 pmol, and the enhancing effects reached a maximum of 25% inclusion at 50 pmol (Fig. 5C, lanes 7-9). Overall, three out of four ASOs showed a CD46 splicing switch in the expected direction, albeit with different efficiencies. This experiment illustrates the feasibility of CD46 splicing manipulation through ASOs targeting exon 13. Similar results were observed in HEK293T and HeLa (Fig. 5B, lanes 10-15).
Regulation of CD46 transcript variants by other processes-In addition to alternative splicing, we also report that CD46 transcript identity and abundance are determined by transcription speed and non-sense mediated mRNA decay (NMD). Stalling of RNA polymerase II at the initiation stage induced by DRB, a P-TEFb (transcription elongation factor) inhibitor, as well as inhibition of elongation by CPT, a DNA topoisomerase I inhibitor (37), slightly reduced endogenous CD46 exon 13 inclusion (Fig. 6A). The positive control mRNA, hnRNP DL (64), showed increased exon inclusion, strongly suggesting that the two compounds had the expected effects. The reduction of exon 13 inclusion by stalling or slowing down transcription by guest on March 24, 2020 http://www.jbc.org/ Downloaded from suggests the co-transcriptional recruitment of negative regulator/s. Another interpretation of the DRB effects is that, in the absence of ongoing transcription elongation, the CD46 transcripts with exon 13 are preferentially degraded by NMD, on top of their NMD-mediated downregulation in normal conditions (see below).
Exon 13 contains an in-frame stop codon located 41 nucleotides upstream from the last exon-exon junction. NMD, a quality control mechanism that degrades transcripts with premature termination codons, is usually activated by in-frame stop codons that are >50 nucleotides upstream of the last exon/exon junction (65,66). As the boundary of 50 nucleotides could be flexible, we sought to test whether the endogenous CD46 transcripts with exon 13 are downregulated by NMD. We treated HEK293T and HeLa with CHX, which represses both translation and NMD, and a 10% increase in exon 13 inclusion was observed, together with a strong increase in the reported NMD-sensitive SRSF1 isoforms (67) (Fig.  6B). Consistently, dsiRNA-mediated depletion of the essential NMD factor UPF1 resulted in slightly higher (4% increase) exon 13-containing mRNA (Fig. 6C). For the two experiments, the increase of exon 13 inclusion mirrors that in NMD-sensitive SRSF1 isoforms. These observations strongly suggest that CD46 mRNA with exon 13 inclusion is weakly sensitive to NMD.

DISCUSSION
Here we report an initial characterization of the cis-acting elements and trans-acting factors that regulate CD46 alternative splicing in human cells. We first illustrate that the CD46 alternative splicing patterns exhibit a high degree of tissueselectivity. These patterns are likely established by combinations of tissue-specific trans-acting factors or ubiquitous trans-acting factors with different expression levels across tissues (68,69), whose activities might be modulated by posttranslational modifications affecting their subcellular location or RNA-binding affinity (70). The three STP exons have different inclusion efficiencies, with exon 9 as the most efficient followed by exons 8 and 7 (Fig. 1). The imperfect inclusion of exons 7 and 8 is in part due to their 5'ss deviating from the consensus 5'ss sequence. We proved that these two 5'ss are recognized by U1 snRNA using noncanonical registers, which were previously found to be associated with alternative exons (46). Importantly, the disruption of base pairs that only occur in the noncanonical register (almost) completely abolishes exon inclusion, indicating that formation of these basepairing registers is essential for inclusion of these exons. Furthermore, the strong differences between the inclusion levels of exons 7 and 8 are not due to their 5' splice sites but rather to other cis-acting elements. The STP domain is the binding site for Neiseria bacteria and the BC isoform has a higher bacterial adherence than the C isoform (20,29), suggesting that inefficient inclusion of exons 7 and 8 could be favored to prevent bacterial infection.
We derived a detailed map of cis-acting elements that regulate the alternative splicing pattern of CD46 exon 13, and these enhancers and silencers play very similar roles across three cell lines. However, the mRNAs derived from the wild-type minigene mostly included exon 13 ( Fig.  2A), while endogenous mRNAs predominantly exhibited exon 13 skipping (Fig. 1A). The inverted inclusion ratios between endogenous and minigene exon 13 do not invalidate the regulatory maps, but imply that the derived strength of each enhancer or silencer is not absolute but rather relative to other elements in the same minigene. The weak NMD sensitivity of the endogenous (but not minigene) exon 13-containing mRNA partly accounts for this discrepancy (Fig. 6, B and C). In addition, we found that slow pol II transcription increases exon 13 skipping, consistent with the co-transcriptional recruitment of a repressor (37) (Fig. 6A). The longer time to transcribe the endogenous compared to the minigene's intron 13 (3,173 vs. 600 bp) could also facilitate recruitment of this repressor. Furthermore, chromatin structures, histone modifications and different promoters could also contribute to differences in exon 13 inclusion between endogenous and plasmidderived CD46 transcripts (38,71). Thus, the CD46 exon 13 minigene does not completely reconstitute the endogenous regulation, yet it allowed the identification of many general cis-acting elements. The importance of these elements is further supported by our finding that three out of four ASOs targeting ESE1 or ESS1 show a splicing switch in the predicted direction.
The CYT1 protein domain, derived from inclusion of exon 13 in mRNA, expands the CD46 cellular functions like maintenance of epithelial cell integrity, autophagy and Tr1 cell differentiation (10,11,31). According to the UCSC and Ensembl genome browser annotations, this exonic sequence exists in the genome of primates and other mammalian species like dog, elephant and dolphin, but is missing in mice and other metazoans like birds. Remarkably, this sequence is not recognized as an exon but rather it is part of an intron or UTR in most mammals except human, gorilla, gibbon, macaque, marmoset, baboon and green monkey. These observations suggest that exon 13 is newly evolved through exonization, by which a sequence gains novel and functional splicing signals by mutations (39), and as such its inclusion efficiency may not be optimal. CYT1 is required to induce Tr1 cell differentiation in order to suppress T-cell response at a post-infection stage (72). It is possible that low CYT1 expression is sufficient for its function but its high expression may cause immunodeficiency. Furthermore, pathogens like measles virus (MV) use CD46 as entry receptor and highjack its function to evade elimination (4,21). Upon binding to CD46, pathogens induce CD46-CYT1 mediated autophagy to get into the cytoplasm where they escape from the autophagosome and start the infection (10). Therefore, the low level of exon 13 inclusion could also be an adaptation to reduce immunodeficiency or MV invasion.
We also identified four trans-acting splicing factors for exon 13 inclusion, whose regulation appears to be direct. Trans-acting splicing factors are often cross-regulated by each other (73,74), making the interpretation of splicing regulatory networks challenging. TIAL1 probably is a stronger activator of exon 13 inclusion than TIA1, or it has less impact on regulating other transacting factors. Exon 13 inclusion did not change much upon TIA1 depletion, so TIAL1 alone might be sufficient to support exon 13 inclusion. In addition, neither single knockdown of one repressor (SRSF1) nor of one activator (TIAL1) completely included or excluded exon 13 from mRNA (Fig. 4), indicating that the splicing regulation of this exon is a combination of different trans-acting factors.
SR proteins and hnRNPs regulate splicing in a position-dependent manner (75). SR proteins typically activate exon inclusion when bound to the exon, but they repress inclusion when recruited downstream of the 5'ss. However, SRSF1 appears to repress exon 13 inclusion via ESS1 by RNA pulldown (Fig. 3E) and cotransfection experiments (data not shown), even though this exon was never reported as a target of SRSF1 regulation. Further experiments like in vitro splicing are required to confirm the direct regulation of exon 13 by SRSF1, and the region where it binds. A very recent paper reported that binding of SRSF1 to exonic regions near the 5'ss mostly promotes exon inclusion while binding to the region near the 3'ss either activates or represses exon inclusion (76). Instead, our results show that SRSF1 likely inhibits inclusion via the 3' half of exon 13. We hypothesize that SRSF1 might likely interfere with the exon definition or U1 recruitment. In addition, an SRSF1-mediated hyperstabilization of U1 snRNP binding to the 5'ss may inhibit subsequent spliceosome rearrangements (44). Besides, SRSF1 may also interact with other proteins to inhibit splicing (76). SRSF1 binding motifs have been studied using SELEX, CLIP and RNA-seq upon overexpression (76)(77)(78). These three methods generated three different consensus SRSF1 motifs with some similarity. SRSF1 motifs (CGCACGA) identified by SELEX are GC rich with similar distribution frequency along exons. Exonic SRSF1 motifs derived from CLIP are purine rich (UGAUGAA) and mainly locate at the 5' end or middle of exons, while the intronic motifs are Crich. The motif recently derived from overexpression and RNA-seq is UCAGAGGA. The difference in motif sequence also suggests the broad and degenerate binding specificity of SRSF1. However, the potential SRSF1-mediated sites at CD46 exon 13 are pyrimidine rich rather than purine rich. As our RNA pulldowns show that SRSF1 can bind to ESS1 despite of this sequence's lack of match to any previous consensus, we hypothesize that SRSF1 might be recruited to ESS1 by other exon 13-binding proteins. Our study illustrates that, on a case-by-case basis, it is inaccurate to predict the role of a trans-acting splicing factor just from its binding position and protein family, so detailed analyses of individual cases are needed to fully understand all splicing regulatory mechanisms.
We propose a model for the alternative splicing regulation of CD46 exon 13 based on the alternative arrangement of ESEs and ESSs in exon 13 (Fig. 7). We suggest that the competitive by guest on March 24, 2020 http://www.jbc.org/ Downloaded from binding of activators and repressors to exon 13 regulates each other by steric blocking (79), although future experiments should prove the direct links between the elements and factors identified here. ESE1 or associated factors may directly interact with U2AF35 and hence stabilize the binding of U2AF to the 3'ss (79)(80)(81). ESS1 most likely prevents the recruitment of U1 snRNP to the 5'ss by steric blocking or inhibits the subsequent spliceosome assembly (44). ESS2 and ESE2 located in the middle of exon 13 may regulate splicing by either promoting or inhibiting exon definition (40). Our RNA pulldowns suggest that hnRNP A1 specifically binds ESE1 (Fig. 3E), so further work should elucidate the role of this protein in regulating exon 13 inclusion. The ISE downstream of the 5'ss is likely bound by TIA1/TIAL1 thus probably interacting with U1C to help stabilize the U1-5'ss interaction (62). Further studies are needed to clarify the regulatory mechanisms of exon 13 inclusion by the cis-acting elements and trans-acting factors identified here. Unexpectedly, real-time RT-PCR analyses showed that the RNA levels of the four identified transacting factors do not correlate with exon 13 inclusion in the 20 human tissues (data not shown). A possible explanation is that these factors are regulated at translational or posttranslational steps (phoshorylation, turnover, localization, etc), or that there are compensatory mechanisms. Nevertheless, this experiment also suggests that the master regulator/s of CD46 alternative splicing remain to be identified. CD46 further exemplifies that the splicing outcome of each transcript is dependent on the balance of positive and negative regulators. Manipulation of CD46 exon 13 inclusion based on the understanding of its splicing regulation could be useful to investigate the role of CD46 in the pathogenesis of autoimmune diseases and perhaps for therapeutics.

Minigene
and protein expression plasmids-CD46 exons 12, 13 and 14 with intervening sequences, and CD46 exons 6, 7, 8 and 10 with intervening sequences were amplified from Human Genomic DNA (Promega) and subcloned into pcDNA3.1+ plasmid. Internal sequences of introns 12 and 13 were removed to leave 300 nucleotides at each end, while only the first 425 nucleotides of exon 14 were amplified. Exon 9 was deleted, and introns 8 and 10 were cropped to leave the first and last 350 nucleotides of each, respectively. Serial deletions and point mutations were introduced into minigene plasmids by PCR mutagenesis using KAPA HiFi polymerase (KAPA Biosystems). Double stranded DNA oligonucleotides with SalI and BamHI overhangs were ligated into the corresponding sites in pSXN plasmid (53). pCGT7-SRSF1 plasmid (82) was kindly given by Prof Javier F Cáceres from MRC Human Genetics Unit at Edinburgh, UK. TIA1 and TIAL1 cDNAs were amplified by PCR using PrimeSTAR Max DNA polymerase (TAKARA Bio) and cloned into pCGT7 plasmid. PTBP1 cDNA was subcloned from pEM830#48+PTBP plasmid (gift by Prof Eugene Makeyev from SBS NTU, Singapore) into pCGT7 plasmid. All plasmids were verified by sequencing.
RNA extraction and semi-quantitative RT-PCR-RNA was extracted using Purelink RNA mini kit (Ambion) and treated with RQ1-RNase free DNase I (Promega) to remove residual DNA, followed by ethanol precipitation. 1 µg RNA was then used to generate complementary DNA (cDNA) using M-MuLV reverse transcriptase (New England Biolabs) and oligo-dT. Subsequently, cDNAs derived from minigene plasmids were amplified with a radiolabeled vector-specific primer pair using GoTaq polymerase (Promega) for 24-35 cycles. Endogenous cDNAs were amplified for 30-35 cycles using radiolabeled primers mapping to the flanking constitutive exons. Primers were 5'-end radiolabeled using T4 polynucleotide kinase (New England Biolabs) and γ-32 P-ATP (Perkin Elmer) and then purified by Microspin G-25 columns (GE Healthcare). PCR products were separated by 4.5%-8% native polyacrylamide gel electrophoresis (PAGE), followed by gel drying and scanning of exposed phosphor-storage screens using Typhoon imager (GE Healthcare). Band intensities were quantified using ImageQuant TL software (GE Healthcare). PCR products were identified by gel purification and sequencing. Each experiment was done in at least triplicate to generate average percentage of inclusion and standard deviation.
CD46 alternative splicing patterns in human tissues-Total RNA from 20 human tissues (Ambion) was reverse transcribed, and the cDNAs were amplified with radiolabeled CD46 exon 6 and exon 10, or CD46 exon 12 and exon 14 primer sets to analyze the splice products of STP and cytoplasmic regions. As there is only one sample per tissue, biological replicates were not obtained, yet technical replicates confirmed the accuracy of the exon inclusion measurements.
Immunoblotting-Protein lysates were prepared with lysis buffer [50 mM Tris-HCl, pH8; 150 mM NaCl; 1% Triton X-100; 10% glycerol; 1 mM EDTA; 1x cOmplete EDTA-free protease inhibitor cocktail (Roche)] and quantified using Bradford assay. 20 μg of protein lysates were separated by 12.5% SDS-PAGE followed by incubation with primary antibodies purchased from Santa Cruz Biotechnology unless stated otherwise RNA pull-down assay RNA-bound proteins were isolated with the RNA-Protein Pull-down kit (Pierce). For each pull-down, 50 pmol RNA and 120 μg HeLa nuclear extract were used. For ASO annealing, a mixture of biotin labeled RNA and ASOs at a 1:3 ratio was first heated at 90 °C for 2 min and then gradually cooled down to room temperature. Pull-down eluates and flow through were separated by 12.5% SDS-PAGE and analyzed by western blot. Protein binding levels are defined by the eluate band intensity over the corresponding flow-through band intensity as measured by GS-800 Calibrated Densitometer (Bio-Rad). Each pull-down was performed in triplicate to derive averages and standard deviations.  Tissue-selective alternative splicing of CD46 with two cassette exons employing noncanonical U1 recognition. A, the predominant endogenous CD46 isoforms in 20 human tissues are BC/C-CYT2. Exon 13 is predominantly skipped while exons 8 and 9 are mostly included in mRNA. Solid arrows depict primers. Asterisk, faint high-molecular-weight STP bands which likely correspond to ABC isoform and ABC plus intron 7 (collectively quantified as +Ex 7 below the gel), as determined by exon 7 specific RT-PCR (data not shown). B, schematic of canonical and asymmetric loop base-pairing registers for exon 7 and 8 5'ss. Exon 7 or 8 have a G or A at position +9, respectively. C, mutational and suppressor U1 experiments in HEK293T cells transfected with CD46-exon 6-7-10 (left) and CD46-exon 6-7-8-10 (right) minigenes. Mutations disrupted the U1 snRNA binding (lanes 3,4,7,11,12 and 15). The effects of the +7C mutant were only rescued by suppressor U1 with the G3 mutation that restores basepairing in the asymmetric loop register, but not U1 with G2 which restored the canonical base-pairing (lanes 7-9 and 15-17). A cryptic 5'ss (AAU|gucaguuu) which is 212 nucleotides downstream of exon 7 was activated and resulted in partial intron retention (Band #1). Besides, a 123-nucleotide long cryptic exon within intron 9 was activated (Bands #2 and #3). This cryptic exon is 89 nucleotides away from exon 7 and uses the same cryptic 5'ss and a novel cryptic 3'ss (uaauuag|AAU) for splicing. The activation of these cryptic splice products could be due to the artificially shortened introns in the minigenes, but it does not change the conclusions of this experiment. In the minigenes, mean exon inclusion values and standard deviations (SD) were derived from at least three experimental replicas (samples from different transfections).   . SRSF1 and PTBP1 repress CD46 exon 13 inclusion via ESSs while TIA1 and TIAL1 promote exon 13 inclusion via a poly-U rich sequence downstream of the 5'ss. All mean exon inclusion values and standard deviations (SD) were derived from at least three experimental replicas (samples from different transfections). Knockdown and overexpression of SRSF1 (A) and PTBP1 (B) revealed their repressive effects on endogenous exon 13 inclusion. Expression levels of each protein were confirmed by Western Blot (bottom panels, also in C,D). C, CD46 exon 13 inclusion is susceptible to SRSF1 level in the context of a chimeric minigene with MCAD exons 8 and 10 and flanking intronic sequences. Inclusion of CD46 exon 13 increased upon SRSF1 depletion and decreased upon SRSF1 overexpression (lanes 1-4). On the other hand, inclusion of MCAD exon 9 was not affected by changes in SRSF1 (lanes 5-8). D, knockdown and overexpression of TIA1 and TIAL1 revealed that they activate endogenous exon 13 inclusion.  HEK293T cells were treated with DRB or CPT to stall or reduce transcription rate. CPT induced exon skipping after 5 h incubation (lane 5), while DRB did so after 24 h (lane 4). hnRNP DL exon 8 was used as a positive control because its inclusion increases upon slow transcription. B, Exon 13-containing CD46 mRNA is weakly sensitive to NMD. CHX-treated cells showed slightly increased exon 13 inclusion, suggesting that exon 13-containing mRNA is mildly predisposed to degradation. As a positive control, NMD-sensitive SRSF1 splice isoforms V and VI greatly increased upon CHX treatment (67). C, knockdown of UPF1, which is a core NMD factor, mildly increased CD46 exon 13 inclusion as well as SRSF1 NMD-sensitive isoforms.