Depolarization Induces NR1 E21 Repression in Differentiated P19 Cells
Previously, we showed that splicing of NMDAR1 E21 is repressed by CaMK IV after transient expression of a minigene reporter in HEK 293T cells. [23
]. It was not clear whether E21 in endogenous NR1 transcripts could be similarly repressed by depolarization and CaMK IV. To confirm E21 repression in endogenous NR1 transcripts, we examined its splicing in differentiated P19 embryonal carcinoma (EC) cells by RT-PCR. Undifferentiated cells were aggregated in retinoic acid (RA) and then dispersed and replated [25
]. Ten days after RA treatment, 60% of the cells exhibited neuronal morphology with extended fasciculated processes synapsing on other cells. The remainder of the culture consisted of flat adherent cells that resembled either astrocytes or undifferentiated EC cells [26
]. The alternative splicing of E21 to E22a or E22b at the 3′ end of NMDAR1 transcript results in four possible different NMDAR1 isoforms. We used E22a- or E22b-specific reverse primers to separately measure the NR1–1:NR1–2 and NR1–3:NR1–4 ratios (A). All four NR1 mRNA isoforms were detected in differentiated P19 cells, with 26% of E21 inclusion in the NR1–1/NR1–2 isoform pair (Exon 22a; B, lane 1) and 12% inclusion in the NR1–3/NR1–4 isoform pair (Exon 22b; Figure S1
A). To examine the effect of depolarization on splicing, these cells were exposed to 50 mM KCl. E21 inclusion was reduced after 6 h of this treatment, and continued to decrease to about 25% of its original level after 24 h (B, lanes 2–5, and Figure S1
A). The extent of repression was dependent on the KCl concentration; in 30 mM KCl, E21 splicing was reduced by 50% at 24 h, whereas in 50 mM KCl, E21 splicing was reduced by 75% (B, lanes 5 and 8). These E21 changes were observed in both exon 22a– and exon 22b–containing mRNA. Thus, E21 splicing is indeed repressed by depolarization, and this repression is apparently independent of the choice of alternative 3′ splice site downstream.
Depolarization Induces Splicing Repression of NMDAR1 Exon 21 in Differentiated P19 Cells
We next tested whether the CaMK pathway is involved in the depolarization-induced repression, as seen with the BK channel STREX exon. Treatment of the cells with the CaMK inhibitor KN93 prior to depolarization completely blocked the splicing repression (C, lane 3, and Figure S1
B). The less-active analog KN92 showed an intermediate effect (C, lane 4). Thus, depolarization-induced splicing repression of E21 is mediated through the CaMK pathway in differentiated P19 cells, similar to what was observed for BK channel splicing in the pituitary cell line GH3
The observation of a splicing change after a stimulus requires removal of the pre-existing mRNA. Thus, the shift in splicing is observed in a timeframe of hours in response to a continuous application of high KCl concentrations. Although we did not observe a decrease in cell viability from the treatment, this splicing alteration could derive from a permanent change in the cells after chronic depolarization. Alternatively, the E21 skipping could be a reversible adaptation to the depolarizing media. To test the reversibility of this splicing repression, KCl was washed out after 18 or 24 h of depolarization, and then the cells were grown for another 24 h. Removal of the high concentration of KCl from the media restored E21 splicing to its normal level (D, lanes 5 and 6, and Figure S1
C). Thus, the splicing repression is reversible and is a dynamic response to environmental cues.
The Exonic Sequence of E21 Is Sufficient for CaMK IV–Induced Splicing Repression
To study the sequence elements needed for E21 regulation, we cloned the human E21 exon with flanking intron sequence between two constitutive β-globin exons of the Dup4–1 plasmid (A). This reporter was transiently expressed in HEK 293T cells, and the mRNAs were assayed for E21 inclusion by primer extension. The repression of E21 by CaMK IV was examined by co-transfection of plasmids expressing a constitutively active form of CaMK IV (CaMK IV dCT) or a kinase-dead mutant control (CaMK IV dCT K75E). E21 was strongly included in the NME21D300 mRNA (91%) when co-expressed with the dead CaMK IV (dCT K75E). In contrast, E21 splicing was reduced by the active CaMK IV (67% inclusion; B, lanes 1 and 2).
The Exonic Sequence of NR1 E21 Is Sufficient for CaMK IV–Induced Splicing Repression in HEK 293T Cells
To roughly localize the RNA elements responsible for CaMK IV repression, we made several chimeric minigenes containing portions of E21 and portions of the constitutive Dup175 exon, and examined their response to CaMK IV. Deletion of the downstream intron (NME21) had little effect on the overall splicing, but made E21 more repressible by CaMK IV dCT (B, lanes 3 and 4). The deleted downstream intron sequence presumably contains positive elements that counteract the repression. As seen previously, and unlike the BK channel STREX exon or NR1 exon 5, the E21 3′ splice site does not confer CaMK IV responsiveness when transferred to the constitutive Dup175 exon (Dup175E21L; B, lanes 5 and 6). In contrast, the E21 exon alone, when fused to the Dup175 exon 3′ splice site (NME21E), was highly repressed by CaMK IV (B, lanes 7 and 8). These results indicate that the E21 exonic sequence is sufficient to induce CaMK IV–dependent splicing repression. This is in contrast to other identified exons in which the regulatory element is located in the 3′ splice site.
A CaRRE-Like Element and a Purine-Rich Sequence Are Both Needed for Efficient CaMK IV Repression of E21
To identify the sequence elements involved in E21 repression, we performed linker scanning mutagenesis across the exon. Blocks of 20 nucleotides within the E21 exon were sequentially substituted with a 20-nucleotide sequence from the IgM gene that is known to lack splicing activity [27
]. To maintain the 3′ and 5′ splice sites, the sequence substitutions started at exon nucleotide 3 and ended four nucleotides from the 3′ end of the exon (A).
Exon Scanning and Point Mutagenesis Reveal Sequences Needed for CaMK IV–Induced Splicing Repression
Two different substitution mutations had strong effects on the CaMK IV–induced repression. The response was completely eliminated when nucleotides 3–22 (LS1) or nucleotides 33–52 (LS3) were substituted with the heterologous sequence (B, lanes 3 and 4 [LS1], and 7 and 8 [LS3]). Thus, at least two RNA elements, one from each region, are needed together for the CaMK IV–induced repression. Having either one of these RNA elements alone is not sufficient for this response. As described earlier, several ESEs and exonic splicing silencers (ESSs) have been identified in the E21 exon, and we found that a substitution that eliminated these ESEs (LS5) did reduce the overall exon inclusion [20
] (B, lanes 11 and 12). However, the CaMK IV response was not affected by this mutation, because the exon was still more excluded in the presence of the activated CaMK IV.
We showed previously that the splicing repression of the BK channel STREX exon is mediated by a CaRRE (CACAUNRUUAU) located in the 3′ splice site. Sequence analysis of the LS3 region identified a CaRRE-like sequence CACAUUUA within it. Point mutations (M1 and M2) within this CaRRE-like sequence led to significant decreases in CaMK IV repression (C, lanes 7–10). A larger 12-nucleotide substitution of the entire CaRRE motif (R1) nearly eliminated the CaMK IV effect (C, lanes 11 and 12). Interestingly, the LS4 scanning mutation also changes some nucleotides in this region, but only gives a small reduction in splicing repression. Subsequent analysis of this element indicates that by chance, the LS4 mutations only have a small effect on activity (see below). These results identify a functional CaRRE motif within the LS3 region and indicate that this motif can function within an exon as well as the 3′ splice site.
In contrast to LS3, no known ESS was identified in the LS1 region. To map the important elements within this sequence, we mutated a purine-rich element (M3–2) and the adjacent sequence downstream (M3–3; A). Although the mutation M3–2 had no effect on CaMK IV repression, the mutation M3–3 totally abolished this response (C, lanes 5 and 6), identifying critical nucleotides in the LS1 region. We called this sequence a CaRRE type 2 motif and will refer to the previous CaRRE as a CaRRE type 1 motif.
Recently, exonic UAGG and intronic G track motifs were shown to be important in silencing E21 splicing [20
]. The possible roles of these elements in the CaMK IV–dependent repression were examined with mutations M4 and M5. Consistent with the previous observations, mutation of the UAGG motif promoted E21 splicing in our system (C, lanes 13 and 14). The M4 mutant also exhibited a significantly weaker CaMK IV response. Thus, the reduced exon repression observed in LS6 and LS7 mutants is presumably due to the loss of this element. In contrast, disruption of the G track had no effect on CaMK IV response. These results indicate that CaMK IV–mediated exon repression requires multiple elements, and that there are two critical elements separate from the UAGG and G track motifs.
Either CaRRE Is Sufficient for CaMK IV Repression in a Heterologous Exon
Although the two exonic elements in E21 are both needed for efficient CaMK IV repression, we wanted to test if they could function independently in other exons. For this, we transferred the CaRRE 1 motif to the constitutive Dup175 exon and characterized its response (A). Introduction of CaRRE 1 into the 5′ region of Dup175 exon (Dup175CA10) caused slight exon skipping, indicating that CaRRE 1 is a weak splicing silencer. However, co-expression of this clone with CaMK IV strongly repressed exon inclusion (A, lanes 3 and 4). This effect was due to the introduced CaRRE 1, because a small point mutation in the element reduced its effect, and replacement with a neutral IgM sequence at the same position in Dup175 exon was not able to confer the repression (A, lanes 5–8). Thus, the CaRRE 1 motif can be sufficient to silence splicing within an exonic context. To investigate the effect of position on the activity of CaRRE 1, the element was moved to the middle of the Dup175 exon (Dup175CA13). This centrally located CaRRE 1 behaved similarly to an element upstream, but gave a somewhat reduced response to CaMK IV (A, lanes 9 and 10). When two CaRRE 1 motifs were introduced in the Dup175 exon (Dup175CA10+13), splicing was completely repressed even in the absence of CaMK IV activity (A, lanes 11 and 12). Thus, CaRRE 1 is a splicing silencer whose activity is dependent on both the sequence context and the number of elements in the exon.
Either Exonic CaRRE Is Sufficient for CaMK IV Repression in a Heterologous Exon
Similarly, the CaRRE 2 motif was also tested for its ability to function independently. CaRRE 2 was inserted at the +3 position of the Dup175 exon, the same position relative to the 3′ splice site as in E21 (B). Similar to CaRRE 1, CaRRE 2 had weak splicing silencer activity on its own, slightly reducing exon inclusion of the new exon. When co-expressed with CaMK IV dCT, exon skipping was sharply increased, indicating that CaRRE 2 is also sufficient for CaMK IV–dependent repression (B, lanes 1 and 2). To confirm that this effect was due to the same element identified in E21, the same mutations were tested. As before, M3–2 had a weak effect on the inducible repression (B, lanes 3 and 4). M3–3, the combined mutation (M), and a replacement sequence (R) all eliminated the repression activity, showing that M3–3 mutation covered the minimal CaRRE 2 sequence (B, lanes 5–10).
The interaction between the CaRRE 2 and CaRRE 1 motifs was examined by introducing CaRRE 1 to positions downstream of CaRRE 2 (C), mimicking the E21 situation. Similar to two copies of CaRRE 1, one copy of each silencer element led to the complete repression of the exon (C, lanes 1–4). To weaken the activity of CaRRE 1, we changed the inserted CaRRE 1 sequence to that of the CaRRE found in the 3′ splice site of NR1 exon 5 (Dup175PR1CA3). This partially restored splicing to 31% inclusion. Overexpression of CaMK IV reduced the splicing to 10% (C, lanes 5 and 6). Similarly, for these exons with both CaRRE 1 and CaRRE 2, mutations in CaRRE 2 restored splicing in the absence of CaMK (C, lanes 7 and 9). These exons were still repressed by the CaMK IV, and behaved like the exons with a single CaRRE motif (C, lanes 8 and 10). These results indicate that the level of constitutive and inducible repression from the CaRRE's can be finely tuned by the number, position, and exact sequence of the elements present in the exon.
Single Nucleotide Scanning Reveals the Degenerate Nature of CaRREs
Most splicing regulatory elements exhibit considerable degeneracy in the sequence that can bind to a particular regulatory protein and affect splicing. The intronic and the exonic CaRRE 1 have the same CACAY core sequence, but are different in the 3′ half and have different activities when placed at the same position in the Dup175PR1 exon (C, Dup175PR1CA13 and Dup175PR1CA3). Moreover, several CaRRE 1 mutants also have moderate activity, suggesting that CaRRE 1 might have other functional variants, but there is little information on which nucleotides are most important to their function. Similarly, very little is known about the sequence requirements for CaRRE 2 activity. The finding that exonic CaRRE motifs can also affect splicing indicates that there are likely to be many more exons regulated by CaRRE 1 or CaRRE 2. Thus, we needed to characterize the CaRRE in more detail before searching for them in other exons.
To characterize the functional sequence of these two CaRREs, we carried out single nucleotide scanning mutagenesis of CaRRE 1 and CaRRE 2. We mutated each CaRRE nucleotide to the other three possible nucleotides, one at a time. To facilitate cloning, we first changed the ApaI site in the Dup construct upstream intron to a XhoI site. Interestingly, this change reduced the CaMK IV responses of both the 175CA10-X and 175PR1-X exons compared to their original sequences. The original ApaI site contains a GGG motif that may weaken the exon and increase the silencing activity of the CaRREs.
In the absence of CaMK IV, most mutations did not have much effect on the splicing of the exons. When co-transfected with CaMK IV dCT, the different CaRRE mutants exhibited differential responses to the CaMK IV (A and B). To evaluate these differences, we defined the wild-type silencing activity as 100% and calculated the relative activity of each mutant ().
Degenerate Sequences Can Act as CaRRE 1 and CaRRE 2
As expected, many mutations reduced the CaMK IV repression to less than 75% of the wild-type activity, and there were four CaRRE 1 mutations and seven CaRRE 2 mutations that only moderately affected silencing activity (between 150% and 75% of their corresponding wild-type activity). Interestingly, there were six CaRRE 1 mutants and six CaRRE 2 mutants that significantly improved silencing activity to greater than 150% of wild type. Among these, a U to G or U to C mutation at position 5 in CaRRE 1 generated a 3-fold increase in silencing activity. In CaRRE 2, a U to G mutation at either position 2 or position 5 also gave a 3-fold increase in the silencing activity. These data indicate that both CaRRE 1 and CaRRE 2 can function through degenerate sequence elements and are likely more abundant than previously expected.
The Consensus CaRREs Identify a Group of Depolarization-Responsive Exons within the Mouse Genome
To identify exons that might be regulated by depolarization, we searched for CaRRE 1 and CaRRE 2 alone and in combination in a dataset of 2,594 mouse alternative exons. Exons are generally between 50 and 250 nucleotides in length, and only exons of length shorter than 253 nucleotides were examined (2,461 exons). Each exon and 50-nucleotide regions of upstream and downstream intron were separately searched for a list of putative CaRRE motifs. The list of CaRRE sequences included all those with an activity of at least 40% of the wild-type elements. The search identified 151 CaRRE 1 motifs and 104 CaRRE 2 motifs within 248 alternative exons or their flanking introns (A). Since NR1 E21 did not pass our automatic expressed sequence tag (EST)-genome mapping and filtering for cassette exons (see Materials and Methods
), it was not included in the searchable alternative exon set. There were 13 alternative exons that contained two CaRRE motifs. Of these, three alternative exons contain one exonic CaRRE 1 and one exonic CaRRE 2, as seen in NR1 E21.
Genome-Wide Identification of Alternative Exons with CaRRE 1 and CaRRE 2 Sequences
About 10% of the alternative exons were found to contain potential CaRRE 1 or CaRRE 2 motifs, suggesting that these elements contribute to the regulation of many alternative exons. To evaluate the significance of these CaRRE motifs in alternative exons, we compared the CaRRE frequencies in alternative exons and constitutive exons. We compiled a database of 10,000 constitutive exons, from which 9,401 exons with lengths shorter than 253 nucleotides were used in the calculation. The observed CaRRE motif frequencies in these two exon groups were calculated by dividing the number of CaRREs by the number of octamers in the searched region and then normalizing to the frequency of random octamers. The normalized CaRRE motif frequency distributions in alternative exons and constitutive exons are plotted separately for CaRRE1 and CaRRE 2 in B. From these calculations, CaRRE 1 motif frequencies were found to be significantly higher in alternative exons than in constitutive exons, and also significantly higher in the introns flanking alternative exons. In contrast, we did not observe a significant enrichment of CaRRE 2 motifs in these regions encompassing alternative exons relative to constitutive exons. Since this element can clearly affect splicing, this lack of enrichment could be due to the CaRRE 2 sequence not being sufficiently defined.
We next tested whether the presence of a CaRRE motif in the exon is predictive of depolarization-induced repression. Sixty exons from different categories were chosen for analysis, and of these, expression of 45 exons was detected by RT-PCR in differentiated P19 cells (unpublished data). Most of these exons showed partial inclusion, confirming their categorization as alternative exons. From this set, 27 exons were selected for further analysis. Because every exon has different splicing efficiency in these cells, we used one of two criteria to categorize the splicing as changed after depolarization. If an exon shows an intermediate level of inclusion between 30% and 70%, it is easier to detect change in splicing. For these exons, a 15% change in inclusion was considered significant. In contrast, when an exon is spliced in less than 30% or greater than 70% of transcripts, it is more difficult to detect a large percent change in splicing. For this group, responsive exons were defined as those showing more than a 2-fold difference in exon inclusion or skipping by depolarization. By these criteria, 13 exons were repressed after depolarization (A and B), and can be classified into two groups by their responses to the depolarization. Group 1 contained six exons whose splicing remained repressed through 24 h, as was seen for NR1 E21. Group 2 contained seven exons whose splicing was repressed at 12 h, but then began to recover after 24 h under depolarizing conditions. These exons apparently adapt to the condition of chronic depolarization and are initially repressed by the treatment, but may not remain so. Interestingly, three exons showed the opposite response of splicing activation after depolarization (group 3). These results indicate a range of responses and regulatory mechanisms for exons affected by depolarization. In all, about 60% of the CaRRE-containing exons that were tested showed regulation by depolarization (C and ). As a control, we selected 17 alternative exons that do not contain any of the degenerate CaRRE motifs and tested their response to depolarization. As expected, the splicing of these exons was unaffected by depolarization. The one exception was Adcyap1r1 exon 14, which is discussed below (A and B, group 2). Overall, we find that the presence of CaRRE motifs is highly predictive of depolarization-induced splicing regulation.
RT-PCR Examination of Depolarization-Induced Splicing Changes of CaRRE-Containing Exons in Differentiated P19 Cells
Summary of RT-PCR Analysis on CaRRE-Containing Exons
To confirm that the predicted CaRREs were functioning in these regulated exons, we cloned 11 of these exons into the Dup4–1 splicing reporter and examined their splicing in HEK 293T cells. Six of these exons showed either complete inclusion or exclusion in the HEK 293T cells after transient expression (unpublished data ). Five exons showed partial inclusion. Of these, three exons showed CaMK IV–dependent repression in HEK 293T cells (). For these exons, we tested the effect of CaRRE mutation on the CaMK IV–induced repression of these exons. Nf2 exon 16 contains a CaRRE 1 in the 3′ splice site whose deletion greatly reduces the CaMK IV effect (). Rnf14 exon 4 contains an exonic CaRRE 1 whose replacement with the neutral IgM sequence again significantly reduces the CaMK IV repression. For Adcyap1r1 exon 14, no CaRRE was found in our motif search, but the exon was seen to be repressed by depolarization in differentiated P19 cells. Interestingly, a CACAYNNA sequence very similar to CaRRE 1 is seen in the 3′ splice site of this exon. Deletion of the CACA nucleotides of this element again significantly weakened the CaMK IV response. Thus a CaRRE 1–like element presumably also mediates the repression of this exon. These results confirm that the type 1 CaRREs in these exons are at least partially mediating their regulation. Unfortunately, none of the CaRRE 2–containing exons tested was spliced in the Dup construct in HEK 293T cells, and we were not able to confirm the role of this element for the computationally identified exons. Because the CaRRE 2 sequences were clearly functional in the CaMK IV repression assay (B), behaving very similarly to CaRRE 1, CaRRE 2 is also likely to play a role in this response. However, its correlation with the group of depolarization-dependent exons is not as clear as CaRRE 1.
Mutagenesis Analysis Confirms the Activity of CaRRE 1 in the Regulated Exons