Conserved sequences in the Drosophila mod(mdg4) intron promote poly(A)-independent transcription termination and trans-splicing.

Alternative splicing (AS) is a regulatory mechanism of gene expression that greatly expands the coding capacities of genomes by allowing the generation of multiple mRNAs from a single gene. In Drosophila, the mod(mdg4) locus is an extreme example of AS that produces more than 30 different mRNAs via trans-splicing that joins together the common exons and the 3' variable exons generated from alternative promoters. To map the regions required for trans-splicing, we have developed an assay for measuring trans-splicing events and identified a 73-bp region in the last common intron that is critical for trans-splicing of three pre-mRNAs synthesized from different DNA strands. We have also found that conserved sequences in the distal part of the last common intron induce polyadenylation-independent transcription termination and are enriched by paused RNA polymerase II (RNAP II). These results suggest that all mod(mdg4) mRNAs are formed by joining in trans the 5' splice site in the last common exon with the 3' splice site in one of the alternative exons.

INTRODUCTION

Pre-mRNA splicing is a fundamental process in eukaryotic gene expression (1). The boundaries of exons and introns are marked by the splice donor site at the 5′-end of the intron and the splice acceptor site at its 3′-end. Splicing is catalyzed by the spliceosome, a complex that is comprised of five small nuclear ribonucleoproteins (snRNPs) and a large number of non-snRNP proteins. The most common type of splicing is known as ‘cis-splicing’, which involves the ligation of exons within one pre-mRNA. Cis-splicing ensures the correct assembly of mRNAs, whose processing is mechanistically coupled with transcription (2,3). Eukaryotic genes usually consist of many exons, and their inclusion in mature mRNA is dependent on alternative splicing (AS). Alternative splicing is a ubiquitous regulatory mechanism of gene expression that allows generation of more than one unique mRNA species from a single gene (4,5). It has been shown that AS greatly expands the coding capacities of genomes by allowing the generation of multiple mRNAs from a limited number of genes (6).

In Drosophila, there are several extreme examples of AS generating multiple mRNAs from a single locus (7,8). In two cases, different mRNAs are formed by trans-splicing of mRNAs produced by different promoters. The first is the lola locus that encodes a transcription factor involved in neuron and stem cell differentiation (9). The 5′-end of lola transcripts contain five exons that splice into 20 variants of 3′ exons to generate 20 protein isoforms (10). The second is the mod(mdg4) locus that produces at least 31 variants of mRNAs with the same four 5′-terminal exons and alternative 3′-terminal exons (11,12). Unexpectedly, the 3′ alternative exons in both cases are transcribed from different promoters and are sometimes located on opposite DNA strands (10,11,13–15). These facts suggest that the high diversity of mRNAs is formed by trans-splicing, with the splice donor site at the 5′-end of last constitutive exon joining with one of the 3′ splice sites of alternative 3′ exons transcribed from independent promoters. Such trans-splicing of mod(mdg4) is conserved in insects, including silkworms, mosquitoes, and cotton bollworms (16–19). Using paired-end deep sequencing of mRNA in Drosophila interspecies hybrids, McManus et al. (20) confirmed that mod(mdg4) and lola mRNAs can be formed from transcripts originating from different DNA strands. They also found that at least 80 other Drosophila genes encode alternative mRNAs that can be generated by trans-splicing. Thus, such trans-splicing is one of the AS mechanisms that ensure generation of highly diverse mRNAs from the same locus.

The mechanisms of trans-splicing in higher eukaryotes are not yet well understood. Current models favor the concept of long complementary sequences that bring two separate transcripts together to promote trans-splicing (19,21,22). However, no obvious base-paired regions have been found between the introns predicted to be involved in trans-splicing in the mod(mdg4) locus (11,12).

To make a step toward understanding a trans-splicing mechanism, we used the φC31-based integration system (23) to obtain an in vivo model for mapping DNA regions in the last common intron that was previously shown to be critical for trans-splicing (12,13). As a result, we identified a 73 bp sequence that proved to be critical for trans-splicing and the mod(mdg4) gene activity in vivo. At the same time, deletion of highly conserved 13-bp motif from the center of the 73 bp sequence had a slight effect on trans-splicing and mod(mdg4) functions. This motif was previously shown to bind U1 small nuclear RNP through strong base-pairing with U1 snRNA (12). We also found that RNAP II paused in the region of the last common intron that could form a conserved secondary structure in RNA. These sequences define a new polyadenylation-independent transcriptional terminator that prevents transcription into the cluster of 3′ exons located downstream. Thus, our results show that all alternative mod(mdg4) mRNAs are formed by the trans-splicing mechanism that joins the 5′ and 3′ ends of exons generated from different promoters.

MATERIALS AND METHODS

RNA purification and quantitative analysis

Total RNA was isolated from 2- to 3-day-old adult males using the TRI reagent (MRC) according to the manufacturer's instructions. The nuclear and cytoplasmic RNA fractions were prepared from S2 cells grown in 100-mm dishes. The amounts of specific cDNA fragments were quantified by real-time PCR. See Supplementary Materials and Methods for details.

Cell transfection

S2 cells grown in SFX medium were cotransfected with 200 ng of pAc-Fluc reference plasmid and 800 ng of experimental plasmids using Cellfectin II (Life Technologies) as recommended by the manufacturer. After transfection, the cells were grown for 48 h and then collected.

Luciferase analysis

Fly lysate was prepared from from 2- to 3-day-old adult males and assayed using the Firefly Luciferase Assay Kit (Biotium) following the manufacturer's instructions. See Supplementary Materials and Methods for details.

RACEseq and nascent RNAseq

The nuclear modified RNA from S2 cells was prepared as described in Supplementary Materials. The amplified cDNA copies were sequenced. Libraries for MiSeq (Illumina) were prepared with Nextera XT DNA Library Prep Kit according to manufacturer's recommendations. The adapter sequence was trimmed, and reads were mapped to the Drosophila dm3 genome using Bowtie2. The mapped data were visualized with IGB Browser. Nascent-seq data (GSM815908, GSM815909 (24)) and 3′NT-Seq data (3′end of native transcripts sequencing) (GSM1193873 (25)) were obtained from NCBI GEO. Reads were mapped to the mod(mdg4) locus using Bowtie2 and visualized with IGB Browser.

ChIP analysis

Chromatin preparation and immunoprecipitation were performed as described (26), with modifications. See Supplementary Materials and Methods for details.

Transgenic lines

Drosophila strains were grown at 25°C under standard culture conditions. The transgenic Drosophila lines were generated using the φC31-mediated site-specific integration system (23) or Cas9-induced (Bloomington Stock Center: 58492) homologous recombination. See Supplementary Materials and Methods for details. Details of the cloning procedure and the sequences of primers used for construct preparation are available upon request.

RESULTS

The mod(mdg4) transcription is terminated in the last common intron 4

The mod(mdg4) locus is critical for fly viability and survival of Drosophila cell lines. The mod(mdg4) mRNAs were generated by joining four common exons with one of alternative 3′ exons, suggesting that critical sequences for trans-splicing should be located in the last common intron (named ‘intron 4’) located downstream of exon 4 (Figure 1A). To map the essential sequences, we divided the intron into nine sections of unequal length, designated a to i (Figure 1B).

Figure 1.

(A) Schematic organization of the mod(mdg4) gene locus in Drosophila melanogaster. The mod(mdg4) mRNAs are generated by joining four common exons (green boxes) with one of alternative 3′-exons placed on the same (orange boxes) or opposite (purple boxes) DNA strand. The sequence between exon 4 and the alternative exon N is defined as intron 4 (pink horizontal bar). Arrows show the TSS and directions of transcription. (B) Nascent RNAseq, nuclear RNA 3′-RACE amplicon sequencing, and NET-Seq profiles. Intron 4 was divided into nine regions named a to i. Region c was divided into 44-bp (c) and 73-bp (c) sequences. The conserved 13-bp sequence was named c. The scheme below shows conserved regions for closely related Sophophora and more common Drosophilidae groups (light green and brown boxes, respectively). The lines with numbers indicate fragments used in qPCR analysis. (C) RNA levels within intron 4 in adult males and in the nuclear and cytoplasmic fractions of S2 cells. Primer positions are shown in (B), at the bottom. (D) Enrichment of RNAP II subunits within intron 4. ChIP-qPCR analysis was carried out with the same primers as in (C) and at the following control points: exon 3–intron 3 junction (ei3), hsp70 promoter (h70), and tubulin-γ37C (tub). Relative enrichment was calculated as the ratio of immunoprecipitated DNA to the input DNA. Error bars show standard deviations (n = 3).

Using the data available from NCBI GEO (24), we analyzed the genome-wide distribution of nascent RNA molecules in S2 cells and noticed a strong decrease in RNA accumulation downstream of exon 4 (Figure 1B). To confirm this observation and detect the potential ends of RNA molecules, we performed 3′-RACE combined with high-throughput sequencing in S2 and Kc167 cells and revealed a gradual decrease in the amount of RNA (almost to zero) in the first 500 bp of intron 4 (Figure 1B and Supplementary Figure S1A). Moreover, no poly(A) tails were detected at the 3′-ends of sequenced RNAs from intron 4, in contrast to annotated polyadenylation signals at the 3′-ends of specific isoforms (Supplementary Figure S1B, S1C).

We then used RT-qPCR to compare S2 cells and adult 2-day males (line yw) for RNA profile in intron 4 (Figure 1C). The level of RNA was measured with 16 pairs of primers covering a 2-kb region of the intron. In S2 cells, nuclear and cytoplasmic RNA fractions were isolated separately. As expected, no transcripts related to intron 4 were detected in the cytoplasmic fraction, suggesting that the donor unspliced pre-mRNA was retained in the nucleus and not exported to the cytoplasm. The nuclear S2 RNA fraction and total RNA from adult males displayed similar qPCR profiles in intron 4. We observed a gradual decrease of pre-mRNA in the ef region (between points 7 and 11), and no pre-mRNA was detected in the g region (between points 12 and 14). Taken together, these results are consistent with transcription termination in the ef region delimited by points 7 and 11. This region contains a 79-bp sequence highly conserved among insects (Supplementary Figure S2), which was identified previously (12).

It is also evident that transcription is initiated in the h region (Figure 1C). According to ModEncode data ((27); ID 5096, 5098), sequences in the h region are enriched with the promoter-associated H3K4Me3 histone mark (Supplementary Figure S3A). To test for promoter activity in the h region, we used luciferase-based assay in S2 cells (Supplementary Figure S3B). The results confirmed that the ghi region can drive the expression of the firefly luciferase reporter gene (Fluc). It appears that this promoter is responsible for the transcription of the 3′ exon encoding isoform N and, possibly, of 3′ exons located downstream that encode other isoforms (Figure 1A).

Alternative splicing and transcription termination are usually associated with pausing of the RNAP II complex (4,6). Therefore, we performed ChIP experiments with antibodies against RpII33 and Rpb8 subunits to quantify the enrichment of RNAP II at different points along the mod(mdg4) intron 4 in 2-day yw males (Figure 1D). The level of RNAP II enrichment was found to be low at the control point in the exon–intron region (ex3-int3) of the mod(mdg4) locus and moderate in the bcd region, while an unexpectedly strong enrichment (with both antibodies) was observed in the efg regions, suggesting strong RNAP II pausing. Similar results were obtained by analyzing data of 3′end of native transcripts sequencing (3′NT-Seq) (25) available from NCBI GEO (GSM1193873). 3′NT-Seq detects nascent RNA transcripts, actively transcribed by RNAP II, through the capture of 3′ RNA. A high level of signal was detected at the fg region, linked in this case with 3′ ends formation of RNAs bound by RNAP II, which could be explained by RNAP II pausing and by active generation of RNA 3′ ends (25). Both these processes are likely to take place in the efg region. A high level of RNAs detected at the end of the exon 4 is due to the formation of splicing intermediates after the 5′SS cleavage. ChIP showed a high enrichment of RNAP II in the h region, which contains the potential promoter for the transcription of alternative downstream 3′ exons. Thus, intron 4 contains several regions in which RNAP II appears to be paused.

Mapping regions required for trans-splicing in intron 4

An attempt to map regions required for trans-splicing was made previously by using RT-PCR to test the products of trans-splicing between exon 4 with different lengths of intron 4 (expressed from plasmid transfected into Drosophila S2 cells) and the endogenous 3′ exons of mod(mdg4) (12). Such a model system did not take into account the role of chromatin and distance between the transcription units whose products are involved in trans-splicing. Therefore, we developed an alternative model system based on parts of the mod(mdg4) locus stably integrated into the cytogenetic region 22A (Supplementary Figure S4A).

To avoid possible RT-PCR artifacts, we developed an assay for evaluating trans-splicing events at the protein level (Figure 2A). One construct (named Donor) contained the endogenous region of the mod(mdg4) locus (including the promoter), all common exons and introns, and 3′ exon for the N isoform. The last common exon 4 was modified by insertion of an IRES sequence from the reaper gene (28) that could initiate translation from internal parts of the mRNA by a cap-independent mechanism. The coding region for the N isoform was separated by the Renilla luciferase (Rluc) gene sequence in order to measure, at the protein level, the amount of mRNA generated by splicing of intron 4. The second construct (named Acceptor) contained the promoter-driven 3′ exon encoding the V isoform that was separated by the firefly luciferase (Fluc) gene sequence. Cap-dependent translation of the Fluc was prevented due to premature open reading frame between the promoter and 3′ splice site. If trans-splicing occurred, the IRES would be fused to the Fluc coding region to induce translation of the firefly luciferase gene. Using the φC31-based integration system (23), both constructs with the white reporter were inserted into the 22A chromosome region located at approximately 10 kb from the nearest coding gene. Unfortunately, we failed to identify the Rluc product in our experimental system due to the problem with the too long distance between IRES and the AUG initiation codon of the Rluc gene.

Figure 2.

Mapping regions required for trans-splicing in intron 4. (A) Scheme of the model system used to map regions involved in trans-splicing. The Donor construct (on the left) contains the mod(mdg4) promoter (arrow) with the mod(mdg4) coding region, including four common exons (green boxes), IRES, intron 4 (pink horizontal bar), and alternative exon N (orange box) with the inserted Rluc gene (blue box). The Acceptor construct (on the right) contains the promoter (arrow) and alternative exon V with the Fluc reporter gene (yellow box). (B) Testing for trans-splicing efficiency in Donor/Acceptor heterozygotes using the Firefly luciferase assay. The broken blue line shows the background autoluminescence level. Each combination of trans-heterozygotes was analyzed in at least five replicates. (C) Testing for trans-splicing efficiency in Donor/Acceptor heterozygotes using RT-qPCR. All RT-qPCR assays were quantified against a standard sample. Levels of exon 4–IRES junction were used as reference. The histogram shows percent proportions of trans-spliced transcripts relative to the total amount of transcripts from the Donor construct. Error bars show standard deviations (n = 3). Asterisks indicate significance levels: *P < 0.05, **P < 0.01, ***P < 0.001. (D) Electrophoregram of RT-PCR products from trans-spliced transcripts (amplified with IRES and Fluc primer pair) and from normalization target (amplified with exon 4 and IRES primer pair).

We constructed transgenic lines expressing the intact intron and derivatives with deletions of either single or several regions indicated in Figure 1B. Trans-splicing was tested in 2- to 3-day males heterozygous for the Donor and Acceptor constructs (Supplementary Figure S4B). A close correlation was observed between the results obtained by measuring Fluc luminescence (Figure 2B) and by RT-qPCR with hydrolysis probes (Figure 2C).

A single deletion of region c was the only one that strongly affected trans-splicing. Single deletions of regions a, e and f showed a moderate effect, with deletions of other regions (b, d, h or g) having no significant influence on trans-splicing (Figure 2B, C; Supplementary Table S1). Deletion of a larger fragment (abc) resulted in complete inactivation of trans-splicing, suggesting that regions a and c have partially redundant functions. Deletion of ef had the same effect as deletion of either e or f, which could be explained by the involvement of these two regions in the same functional module. Indeed, the most conserved structure overlaps with sequences from both regions (Supplementary Figure S2).

Region c contains the previously identified highly conserved 13-bp core motif (GGURAGUAGAGUG) that resembles a pseudo-5′SS sequence with the 9-base region matching to U1 snRNA (12). To test for the functional role of this motif, we made smaller deletions in the 117-bp region c: Δ1–44 bp (Δc), Δ45–117 bp (Δc), and Δ73–85 bp (Δc) (Supplementary Figure S2). Unexpectedly, deletion of c and c had no influence on the efficiency of trans-splicing, while deletion of c had an even stronger effect than deletion of the entire region c, resulting in almost complete inactivation of trans-splicing (Figure 2B–D; Supplementary Table S1).

To confirm these results, we made an Acceptor construct that contained two promoters driving divergent transcription of the 3′ exons of the T and K isoforms, with the coding regions of the T and K isoforms being substituted by Fluc and mCherry reporters (Figure 1A, Supplementary Figure S5A). The construct was integrated into the 22A chromosome region and tested for trans-splicing by RT-PCR in combination with the Donor constructs. The results were generally similar to those described above, but deletion of region c and complex deletion (bcd) resulted in complete inactivation of trans-splicing, while deletion of region a had only a slight effect (Supplementary Figure S5B, S5C; Table S1). Thus, the role of region c in trans-splicing is critical but also depends on the nature of the promoter region regulating transcription of the 3′ alternative exon.

The conserved c region is critical for trans-splicing in the endogenous mod(mdg4) locus

Our next task was to test whether the results obtained in our model system could be confirmed at the endogenous mod(mdg4) locus. To make deletions of the c and c regions we used the CRISPR/Cas9 system (29,30) (Figure 3, Supplementary Figure S6A). The sgRNAs were designed for regions b and d that, according to our test, are not essential for trans-splicing. The DNA template for substitution contained the region of intron 4 between the sites for the sgRNAs. The proximal site for sgRNA in the DNA template was replaced by an ApaI restriction site, and the mCherry reporter flanked by loxP sites was inserted into the distal site. The region of substitution was flanked by the 500-bp homology arms. We prepared DNA templates for deletions of the 13 bp core motif (Δc), the c and c regions (Δc and Δc) and for control substitution without deletion of the sequences (intact).

Figure 3.

Functional role of the c regions from intron 4 in trans-splicing of the mod(mdg4) locus in vivo. (A) Scheme of CRISPR/Cas9 technique used to delete designed sequences from intron 4 of the mod(mdg4) locus. The Cas9 cleavage site in region b was altered into ApaI site. The mCherry reporter flanked by loxP sites was added into d cut site. (B) Scheme of the second variant of intron 4 editing. The template for substitution contained the unaltered region of intron 4 and the mCherry reporter between loxP sites and followed by the same region with one of the tested deletions (c*). (C, D) Comparing trans-splicing efficiency in mod(mdg4) (WT), mod(mdg4)Δc13 lines (C) and mod(mdg4) (WT), mod(mdg4)Δc44, mod(mdg4)Δc73, mod(mdg4)Δc13 lines (D) based on the results of RT-qPCR for five mod(mdg4) isoforms: N, P, T, V, and W (see Figure 1A) and for a point from intron 4. Lines with altered genome were obtained in accordance with the schemes shown in (A) and (B). Analysis of RNA isolated from 2-day-old males was performed in three independent replicates. The histogram shows proportions (%) of particular transcripts relative to the total amount of mod(mdg4) transcripts. Error bars show standard deviations (n = 3). Asterisks indicate significance levels: *P < 0.05, **P < 0.01.

Seven independent events were obtained with the intact template. All of them had insertion of the mCherry reporter (mod(mdg4)) and the ApaI site, suggesting that both sgRNAs functioned effectively (Supplementary Figure S6B). Homozygous mod(mdg4) flies and mod(mdg4) flies generated by loxP-mediated deletion of the mCherry reporter were fully viable. Since inactivation of the mod(mdg4) gene leads to embryonic lethality (31), we concluded that insertion of the reporter gene in region d did not interfere with the mod(mdg4) function.

Three independent events with the simultaneous insertion of the mCherry reporter and ApaI were obtained with the Δc template. In two cases, the 13-bp deletion was also detected and confirmed by sequencing (Supplementary Figure S6B). Flies homozygous for the mod(mdg4) allele were fully viable. Finally, eight and seven independent events were obtained for the deletion of the regions c and c, respectively. Unexpectedly, integration of the mCherry reporter in seven cases for Δc and five cases for Δc was not accompanied by that of the ApaI site. Moreover, none of integration events had deletion of either of the regions, including the three remaining events that contained the ApaI site. The absence of deletions in the integration events could be interpreted as evidence for a dominant lethal effect of the deletions or, alternatively, for a high level of recombination events during integration of the constructs.

Taking into account that deletions might induce a dominant lethal effect, we used an alternative approach with the CRISPR/Cas9 (Figure 3B, Supplementary Figure S7A). The same sgRNAs and the 500-bp homology arms were used, but the DNA template for substitution contained the wt region of intron 4 between the sites for the sgRNAs (c) and the mCherry reporter flanked by loxP and followed by the same c region with one of the tested deletions (c*). It was expected that the presence of the wt copy of the c region would support trans-splicing at the mod(mdg4) locus after successful substitutions. The resulting events were tested by PCR and sequencing to select lines with correct substitutions (Supplementary Figure S7B). Next, the wt c region and the mCherry reporter were deleted by loxP-mediated recombination. The resulting transgenic lines contained loxP in the region b and one of the c variants. Using this approach, we obtained two intact, one Δc, one Δc, and one Δc fly lines. All new mod(mdg4) alleles were viable and could be maintained in the homozygous state. Only line homozygous for the mod(mdg4) allele showed reduced viability and fertility.

The next question was, whether the deletions in the intron 4 could affect trans-splicing of the mod(mdg4) isoforms. Using RT-qPCR, we measured splicing efficiency for five 3′ exons located on either the same or opposite DNA strand for lines carrying the following mod(mdg4) alleles: wt, intact, Δc, Δc, and Δc (Figure 3C, D). Trans-splicing in the mod(mdg4) and mod(mdg4)Δ alleles was at the wild-type level, being slightly changed in two independently obtained mod(mdg4) alleles. A significant reduction of trans-splicing was observed only in the case of isoform T. Thus, the core 13-bp conserved sequence proved to have only a limited impact on trans-splicing. At the same time, RT-qPCR showed that formation of all tested isoforms was obviously decreased in mod(mdg4)Δflies (Figure 3D). Accumulation of unspliced RNA carrying the intron 4 sequences was also observed only in mod(mdg4)Δ line (Figure 3D, right panel). These results are in complete agreement with the results obtained in the model transgenic system, demonstrating that the c region contains sequences that are critical for trans-splicing.

The conserved sequence in regions e, f and g are required for transcription termination in intron 4

According to our results (Figure 1), transcription is likely to be terminated in the ef region of intron 4. This region contains two cores that are invariant in Drosophila species and was predicted to form a stable secondary structure containing four stems and two loops (12). To directly map the regions that are essential for transcription termination, we compared RNA levels at different sites of intron 4 in males homozygous for the Donor constructs by means of RT-qPCR with either random primers or a specific primer located in the Rluc coding region. In both cases, deletion of e, f or g alone resulted in a severalfold increase in the amount of transcripts in downstream regions, while that of any other region had no significant effect (Figure 4A, Supplementary Figures S8, S9). Deletion of larger fragments (ef or efg) had the same effect as deletion of each individual region, suggesting that regions e, f, and g contain sequences that are important for the same activity.

Figure 4.

Transcription termination in the intron 4 of the mod(mdg4) locus. (A) Mapping the regions involved in transcription termination in intron 4 using the transgenic line homozygous for the Donor construct (see Figure 2). The histogram shows the accumulated and throughout-read transcripts within intron 4 in intact and deletion variants of the Donor construct. Dotted line shows the endogenous level of RNA corresponding to the test region. (B) Mapping the regions involved in transcription termination in intron 4 using the transgenic line homozygous for the construct carrying the abcdefg region of intron 4 inserted between the Rluc and Fluc genes under control of the Act5C promoter. All qRT-PCR assays were quantified against a standard sample. Levels of ex4–IRES junction and tubulin-γ37C were taken as reference. Random primers or specific primers (blue arrows) were used for RT. (C) Testing the ability of the efg region to terminate transcription in the heterologous yellow intron. The 2.7-kb intron with 200-bp flanking exon fragments from the yellow gene was inserted between the Rluc and Fluc genes driven by the Act5C promoter. Transgenic flies with the y background were used in order to eliminate endogenous yellow gene influence. The histogram shows RNA levels for control points of Rluc, exon1 (ex1), and unspliced (u.s.; d.s.) and spliced (spl) products from constructs with the intact yellow gene intron and this intron with efg insertion. (D) Model system for testing the ability of the efg region to terminate transcription in the transiently transfected S2 cells. The efg region, linker with similar length and polyadenylation signal from SV40 were inserted between the Rluc and eGFP genes driven by the Act5C promoter. pAct5C-Fluc reporter was cotransfected as reference. (E) The histogram shows Rluc-to-Fluc luminescence ratio. (F) The histograms show eGFP/Rluc and Rluc/Fluc ratios of RNA levels. Nuclear and cytoplasmic fractions of RNA were analyzed separately. Error bars show standard deviations (n = 3). Asterisks indicate significance levels: *P < 0.05, **P < 0.01, ***P < 0.001.

To corroborate these results, we examined transcription in the abcdefg region using a heterologous transgenic model system in which this region was inserted between the Rluc gene driven by the Act5C promoter and the Fluc gene followed by the poly(A) signal (Figure 4B, Supplementary Figures S10, S11). The construct was inserted into the 38D chromosome region. The level of RNA in different regions of the construct was tested by RT-qPCR. As previously, deletion of e, f or g alone proved to strongly increase transcription in downstream regions.

We then tested the ability of the efg region to affect transcription in the heterologous intron from the yellow gene using the construct where this region was inserted in the yellow intron placed between Act5C-Rluc and Fluc-poly(A) (Figure 4C). The construct was integrated into the 38D chromosome region. As a result, we observed that insertion of the efg region induced a strong decrease in the steady-state RNA level, with the amount of Rluc being only half that in the transgenic line carrying the construct without this region. The amount of RNA corresponding to the yellow intron and spliced RNA was decreased almost tenfold. Thus, the efg region inserted into the intron of unrelated gene reduces RNA level, which may be explained by induction of transcription termination and decrease in RNA stability.

To further evaluate the functional role of the efg region in transcription termination, we prepared a set of constructs containing the Rluc and eGFP genes under control of the Act5C promoter. Different sequences such as the 717-bp negative control DNA (from mCherry), 243-bp SV40 poly(A) signal, and 765-bp efg region were inserted between the Rluc and eGFP genes. Expression of Rluc and eGFP in S2 cells was compared relative to that of Fluc expressed from the reference Act5C-Fluc construct (Figure 4D). Since insertion of the 717-bp linker between the Rluc and eGFP genes did not significantly change the expression patterns of the reporter genes, we regarded both these variants as negative controls (Figure 4E, F). The expression of reporter genes was examined separately in the nuclear and cytoplasmic RNA fractions. In the negative controls, the eGFP and Rluc genes were transcribed as a single RNA, and the ratio of eGFP to Rluc was therefore close to 1 (Figure 4F, left panels). This ratio in the nuclear RNA fraction was slightly lower because of the presence of incompletely transcribed nascent RNAs (Figure 4F, top panels). Insertion of either efg or SV40 resulted in a tenfold reduction of the eGFP/Rluc pre-mRNA ratio in the nuclear fraction, which in both cases could be explained by transcription termination in the region between the reporter genes. The Rluc/Fluc ratio in the nuclear RNA fraction (Figure 4F, right panel) was almost equal for all constructs, suggesting similar levels of expression. In contrast, different RNA patterns for the efg and SV40 constructs were observed in the cytoplasmic fraction (Figure 4F, bottom panels). As expected, the Rluc/Fluc ratio in the presence of SV40 was four times higher than in the negative controls (Figure 4F, bottom panel), and this correlated with more than tenfold difference in the Rluc/Fluc luminescence ratio, which reflected the amount of Rluc protein (Figure 4E). These results are in agreement with effective transcription termination by SV40 poly(A) signal and formation of the mature Rluc mRNA. The opposite effect was observed in the case of efg: compared to the negative controls, the eGFP/Rluc ratio was two times lower, while the Rluc/Fluc ratio was reduced threefold (Figure 4F, bottom panel). Thus, the cytoplasmic fraction was enriched with Rluc-efg-eGFP transcripts, while the nuclear fraction was enriched with Rluc transcripts. These results correlate with data on the Rluc/Fluc luminescence ratio for efg, which was decreased by factors of 6 and 60 relative to the negative controls and SV40 poly(A), respectively (Figure 4E). Taken together, these results confirm that the efg region accounts for the polyadenylation-independent mechanism of RNAP II transcription termination and that the RNA terminated at the efg region cannot be exported from the nucleus to the cytoplasm.

DISCUSSION

The Drosophila mod(mdg4) locus provides an extreme example of AS generating multiple mRNAs (at least 31 variants) from a single locus with the same four 5′-terminal exons and alternative 3′-terminal exons transcribed from different promoters and sometimes located on the opposite DNA strands. Thus, trans-splicing should account for equally efficient generation of mRNA variants with different 3′ exons. We developed an in vivo model for mapping DNA regions in the last common intron of mod(mdg4) that was previously shown to be critical for trans-splicing (12,13). As a result, we identified a 73-bp sequence that proved to be critical for trans-splicing and the mod(mdg4) gene activity in vivo. At the same time, deletion of highly conserved 13-bp motif from the center of the 73-bp sequence had no significant effect on trans-splicing and mod(mdg4) functions. This 13-bp motif was previously shown to bind U1 small nuclear RNP through strong base-pairing with U1 snRNA (12).

The results presented above are partly different from those obtained by Gao et al. (12), which may be explained by differences in model systems used for identification of sequences that are essential for trans-splicing (Figure 5A). Gao et al. transfected S2 cells with plasmid-borne constructs containing the mod(mdg4) exon 4 with different lengths of intron 4 under control of the metallothionein promoter and tested for trans-splicing products between the exogenous exon 4 and the endogenous 3′ exons of mod(mdg4) (Figure 5B). The documented trans-splicing events suggest that the exogenous pre-mRNAs containing only the exon 4 and part of intron 4 form stable intermediates that efficiently search for the pre-mRNAs with endogenous 3′ exons of mod(mdg4) located in the distant nuclear compartment. Moreover, these exogenous pre-mRNAs can efficiently compete for the endogenous 3′-exon pre-mRNAs with the endogenous 5′-exon pre-mRNAs expressed by the mod(mdg4) locus. In our model system, the native mod(mdg4) promoters generated pre-mRNAs for common 5′ exons or 3′ alternative exons (Figure 5C), and tests were performed for trans-splicing products between pre-mRNAs transcribed from genes located at the same site on homologous chromosomes, which recapitulated the situation in the endogenous mod(mdg4) locus.

Figure 5.

(A) Comparison of the mod(mdg4) intron 4 regions essential for trans-splicing as identified in this study (on the top) and by Gao et al. (12). The most critical regions are shown in red; the regions whose deletion has a moderate or no apparent effect are shown in orange and green, respectively. (B, C) Model systems used for mapping the regions required for trans-splicing (B) in Gao et al. (12) and (C) in this study. (D) A model of trans-splicing.

On the other hand, our results are consistent with (12) in showing that the 300–400 bp region (TSA/region c) is most critical for trans-splicing (Figure 5A, Supplementary Figure S12); while deletion of the 600–800 bp region (TSB/region ef) has only a slight effect on this process (Supplementary Figure S13). This was confirmed in experiments with deletions generated in the mod(mdg4) locus using the CRISPR/Cas9 system (Figure 5A, Supplementary Figure S12). The main contradiction with (12) concerns the role of the 13-bp core TSA region that is bound by U1 snRNA. Using the S2 model system, Gao et al. (12) found the core TSA to be essential for trans-splicing, but its deletion in our model system and in the endogenous mod(mdg4) locus had no significant effect on trans-splicing. It is likely that these contradictory results are explained by high stringency of the S2 model system (Figure 5B): the exogenous pre-mRNA should search in the nucleus for the endogenous 3′-exon pre-mRNAs and compete for trans-splicing with the endogenous pre-mRNA encoded by the common part of the mod(mdg4) locus. In this case, the binding of U1 snRNA to the TSA core may improve the stability of pre-mRNA and its ability to compete with the endogenous mod(mdg4) pre-mRNA for trans-splicing.

Alternative splicing and transcription termination are usually associated with pausing of the RNAP II complex. We also found that the intron 4 contains several regions in which RNAP II appears to be paused and that the ef region of the last common intron contains the most conserved sequence of intron 4 with predictable secondary structure. It is possible that such an RNA structure, in combination with unknown proteins bound to region g, induces RNAPII pausing and conformational changes that induce transcription termination without polyadenylation and prevents transcription into the cluster of 3′ exons located downstream. Similar mechanisms have been described for transcriptional termination of long noncoding microRNA genes (32,33). Such a mechanism of RNAP II termination is likely to be essential in view of previous data that U1 snRNP protects pre-mRNAs from cleavage and polyadenylation at conventional AATAAA signals in gene introns (33,34).

Transcription termination in intron 4 suggests that all mod(mdg4) mRNAs are formed by joining in trans of the 5′ splice site after the last common exon with the 3′ splice site before one of the alternative exons. As shown previously, different mod(mdg4) isoforms are expressed in a manner that depends on the tissue and the stage of development [(11); FlyAtlas (www.flyatlas.org)]. The mod(mdg4) pre-mRNAs start at multiple promoters that transcribe sequences from both DNA strands within the locus (11). Thus, the efficiency of splicing for 3′ exons should be the same, regardless of distance and orientation relative to the last common exon 4.

According to Gao et al. (12), all sequences required for trans-splicing are located in the intron 4 (Figure 5B). The c/TSA region contains the U1 snRNA binding site and additional key sites for unknown proteins involved in stimulation of trans-splicing. It is important to note that the deletion of c did not completely suppress trans-splicing in the mod(mdg4) locus, suggesting that additional, partially redundant sites for unknown proteins in the intron 4 are involved in this process. One of trans-splicing models (21) suggests that the spliceosome can mediate trans-splicing between the unsaturated donor and acceptor splice sites from different pre-mRNAs. In accordance with this model, regulatory proteins specifically bind to the intron 4 and, by unknown mechanism, form pre-mRNA with the 5′ unsaturated donor splice site, while transcription of alternative 3′ exons produces pre-mRNAs with unsaturated acceptor splice sites. Binding of the U1 snRNA can stabilize the unsaturated 5′ pre-mRNA complex that retains its trans-splicing ability while searching for a 3′ pre-mRNA with unsaturated acceptor site (Figure 5D).

The results of this study and the proposed model will hopefully contribute to our understanding of trans-splicing. However, a number of questions on the details of this phenomenon are still open, and further research is required to gain a deeper insight into the mechanism that generates, with equal efficiency, multiple alternative mRNAs from the same locus.

Click here for additional data file.