Multiple intron motifs have redundant functions in the trans-splicing of the mod(mdg4) locus

Abstract

In contrast to canonical cis-splicing, trans-splicing combines exons from two distinct transcripts, yielding chimeric mRNAs. One striking example is the ubiquitously expressed modifier of the mdg4 (mod(mdg4)) locus in Drosophila, where all mRNAs spanning over 30 isoforms are generated exclusively by trans-splicing. This process combines common constitutive exons with one of many alternative 3′ variable exons transcribed from independent promoters. Here, we demonstrate that trans-splicing is determined by a 600-bp region located in the proximal part of the last common intron. Unlike other known examples of trans-splicing, complementary sequences between donor and acceptor pre-mRNAs do not trigger trans-splicing of the mod(mdg4) gene. Deletion analysis revealed that the 600-bp intronic region contains multiple redundant motifs for RNA-binding proteins (RBPs) that collectively determine the level of trans-splicing. We identified specific binding of a large group of spliceosome-associated RBPs by affinity purification of the in vitro transcribed proximal part of the intronic RNA. A multifold reduction in mRNA expression caused by the partial removal of intronic sequences is partially compensated for by an increase in the translation efficiency of Mod(mdg4) isoforms. The results were consistent with a model in which multiple RBP motifs stabilize pre-mRNA binding to the spliceosome, thereby inducing trans-splicing.

Graphical Abstract

Graphical Abstract.

Introduction

Most eukaryotic precursor messenger RNAs (pre-mRNAs) undergo splicing, a two-step process that requires the precise definition of intron–exon boundaries and comprises intron removal followed by exon ligation [1]. Splicing is a highly regulated process catalyzed by the spliceosome, a dynamic ribonucleoprotein complex comprising five noncoding U-rich small nuclear RNAs (snRNAs) and over 200 associated proteins [1–3]. The spliceosome plays a central role in identifying core regulatory sequences within pre-mRNAs and orchestrates a complex cascade of RNA–protein interactions. The U1 snRNP complex mediates initial recognition of the 5′ splice site of an intron through base-pairing interactions involving the 5′ end of U1 snRNA [4, 5]. This specific recognition mechanism does not rely solely on the primary sequences of the splice sites, as their consensus features are rather rudimentary. Instead, specific motifs within exons and introns adjacent to donor and acceptor sites interact with RNA-binding proteins (RBPs), thereby ensuring efficient spliceosome assembly and splicing [3].

More than 200 RBPs have been annotated in Drosophila [6, 7]. The predominant class of Drosophila RBPs includes one or several copies of a conserved RNA-binding domain (RBD) [8, 9] that typically recognizes short and degenerate motifs. In addition to RBDs, these proteins often contain auxiliary domains, such as serine/arginine-rich (SR) domains, K-homology (KH) domains, and zinc fingers [10–13]. While some RBPs are ubiquitously expressed and contribute to constitutive splicing, others exhibit highly tissue-specific expression patterns. These specialized RBPs play a key role in triggering alternative splicing and, hence, the synthesis of distinct protein isoforms [10, 12, 13]. RBPs bind to cis-regulatory RNA elements, termed splicing enhancers or silencers, that modulate the efficiency of donor- and acceptor-site utilization during alternative splicing [1, 14–16].

Alternative splicing is widely regarded as a major driver of proteome diversity, generating multiple mRNA variants from a single pre-mRNA through the differential selection of splice sites [14, 17]. Here, we focus on a highly specialized form of alternative splicing known as trans-splicing, in which exons from distinct pre-mRNAs that originate either from the same genomic locus (intragenic trans-splicing) or from completely different genes (intergenic trans-splicing) are joined [18–20]. Despite its remarkable capacity to generate transcript diversity, trans-splicing is normally repressed to prevent the formation of aberrant or deleterious chimeric mRNAs. Nevertheless, functional trans-splicing has been documented in numerous higher eukaryotes, including humans [21–23]. Current models suggest that long complementary sequences typically mediate the pairing of distinct transcripts to facilitate trans-splicing [20, 21, 24]. For example, complementary base pairing brings pre-mRNAs from different sDscamβ clusters of Tetranychus urticae into proximity, enabling the assembly of chimeric transcripts and expanding isoform diversity [25].

The mod(mdg4) locus of Drosophila represents a unique paradigm in which trans-splicing generates 31 diverse mRNAs encoding isoforms with distinct C-terminal domains [26, 27]. The mod(mdg4) locus (Fig. 1A) contains a common region encoding the constant core of the Mod(mdg4) proteins. This region harbors an N-terminal BTB domain and alternative 3′ exons organized into five clusters, each transcribed from an independent promoter. The well-characterized Mod(mdg4)-67.2 isoform contains a unique C-terminal domain that specifically interacts with the architectural protein Su(Hw). This complex is involved in organizing chromosomal architecture, insulation, and the direct silencing of promoters [28–31]. The mod(mdg4) locus is exceptional because all of its mature mRNAs are produced exclusively via trans-splicing [32, 33]. To date, no discernible complementary sequences have been identified between the introns involved in trans-splicing at this locus [33–35]. Consequently, the assembly of alternative mod(mdg4) mRNAs must proceed via a novel mechanism that couples the 5′ and 3′ splice sites of transcripts derived from independent promoters.

Figure 1.

Organization of the mod(mdg4) locus and an illustration of trans-splicing. (A) The structure of the mod(mdg4) locus in D. melanogaster. Arrows indicate annotated promoters. Constitutive exons shared by all isoforms are shown in green and are numbered from 1 to 4. Alternative exons located on the same DNA strand as the constitutive exons are shown in orange, and those located on the opposite strand are shown in purple. Black arcs indicate constitutive introns. Trans-splicing events connecting constitutive and alternative exons are shown by green curves. The gray boxes indicate genomic fragments used to generate the donor and acceptor transgenes employed in this study. The alternative exons corresponding to the T and K isoforms used in the acceptor construct are indicated. The scheme is based on the FlyBase annotation of the mod(mdg4) locus [36–38], and genomic coordinates are according to the D. melanogaster reference genome assembly dm6 [39]. Schematic representation adapted from Beginyazova et al. [40] under the CC BY 4.0 license. (B) Schematic representation of the fourth intron of mod(mdg4). The intron is subdivided into regions (a–i); genomic coordinates (dm6), coordinates relative to the beginning of the intron, and fragment lengths are indicated above the proportional intron map. Functionally important sequence elements are shown below, including the conserved 5′ region [41], the predicted RNA secondary structures located at the end of region b and the beginning of region c [42], the conserved c¹³ motif required for efficient trans-splicing [34], and the RNA secondary structure associated with noncanonical transcription termination [32]. (C) Schematic representation of mature mRNA formation by trans-splicing. The constitutive region of the locus produces the donor pre-mRNA (green), whereas the alternative regions generate multiple acceptor pre-mRNAs (orange and purple). Trans-splicing between donor and acceptor pre-mRNAs produces 31 mature mRNA isoforms, each containing four constitutive exons fused to one or two alternative exons.

Several cis-regulatory elements critical for trans-splicing have been mapped to the proximal region of the fourth (and last) common intron of the mod(mdg4) gene (Fig. 1B). A conserved 13-bp motif capable of robustly recruiting the U1 snRNP through strong base pairing [34] and a conserved sequence flanking the donor site [41] contribute to trans-splicing. However, their in vivo deletion only marginally impairs trans-splicing efficiency [41, 33]. In contrast, deletion of a 73-bp sequence encompassing this 13-bp core severely reduces trans-splicing in vivo, suggesting that the process is driven by multiple distinct motifs acting in concert. There is a conserved sequence located distally within the intron that forms a stable secondary RNA structure [34] and is involved in the termination of transcription without mRNA polyadenylation [32, 33]. Notably, RNA polymerase II (RNA pol II) pauses within the proximal part of this fourth intron, with no transcription having been detected in its distal region [33].

In this study, we mapped the specific sequences and motifs that govern trans-splicing at the mod(mdg4) locus using a previously established model system [33] in combination with in vivo genome editing via CRISPR–Cas9 and the φC31 integrase system [43–47]. We demonstrate that a 600-bp region of the fourth intron contains multiple, redundant RBP motifs that cooperate to produce efficient trans-splicing. RNA affinity purification analysis revealed three sub-complexes, within which five distinct RBPs possess specific binding sites inside this critical trans-splicing-determining region. Furthermore, we show that partial disruption of trans-splicing, which leads to a multifold reduction in mod(mdg4) isoform transcripts, does not compromise mutant viability, even though mod(mdg4) is an essential gene in Drosophila. Finally, we demonstrate that a severe reduction in transcripts encoding the Mod(mdg4)-67.2 isoform—a critical component of the Su(Hw) insulator complex—does not notably diminish total Mod(mdg4) protein levels.

Materials and methods

Construct design and molecular cloning

Genetic constructs were prepared following standard molecular biology protocols, utilizing reagents from Thermo Fisher Scientific (Waltham, MA, USA). Plasmid propagation occurred in Escherichia coli DH5α under appropriate selective conditions. Site-specific integration constructs were developed from a pBluescript-derived backbone, incorporating mini-white, a loxP site, a multiple cloning site (MCS), and an attB site for φC31 integrase-mediated insertion. Promoters were obtained from their respective sources: the mod(mdg4) promoter was amplified by polymerase chain reaction (PCR) from Drosophila melanogaster genomic DNA, the actin5C promoter was excised from pAc5.1/V5-His, and the UAS promoter was obtained from pUASTm [48]. Genomic DNA was used as the template for PCR amplification of coding sequences for mod(mdg4) and grip, and the reaper (rpr) internal ribosome entry site (IRES). Intact introns were amplified directly, and deletions or site-specific mutations were subsequently introduced via fusion PCR or by fragment excision between restriction sites. Promoters, outrons, and short 5′ exon fragments for the acceptor constructs were also amplified from genomic DNA. Firefly luciferase (Fluc) was excised from pGL3-Basic (Promega), while mCherry was PCR-amplified from a laboratory stock vector. The SV40 polyadenylation signal was obtained from pAc5.1/V5-His and inserted downstream of the marker genes. sgRNAs for targeted mutagenesis of mod(mdg4) intron 4, targeting the b and d regions, were selected using the Fly CRISPR Optimal Target Finder [44]. Two sgRNAs were cloned into a dual U6 promoter vector based on pCFD4-U6:1_U6:3 [43] (Addgene #49411) for simultaneous expression. The homology-directed repair donor template comprised PCR-amplified genomic fragments, the mCherry coding sequence, the 3xP3 promoter (both from laboratory plasmids), and the SV40 polyadenylation signal, which were from pAc5.1/V5-His. attP recombination sites were incorporated at the primer 5′ ends. Sequences for targeted integration into the fourth intron landing platform were PCR-amplified with attB-containing primers and then cloned into a laboratory vector carrying the 3xP3-driven eGFP reporter gene.

The oligonucleotides used are listed in Supplementary Table S2

Drosophila stocks, transgenesis, and genetic crosses

D. melanogaster stocks were maintained at 25°C under standard conditions. Transgenic constructs were integrated site-specifically into the 22A2 locus using φC31 integrase-mediated recombination. Plasmid DNA (200 ng/μl) was microinjected into preblastoderm embryos [49] from a fly line (Bloomington Stock Center #24 481: y¹ M{vas-int.Dm}ZH-2A w¹¹¹⁸; M{3xP3-RFP.attP’}ZH-22A) [50] that carries an attP docking site and expresses φC31 integrase under the germline-specific vasa promoter. F0 flies were crossed with y¹w¹¹¹⁸, and F1 progeny with the transgene were selected via mini-white eye pigmentation. Homozygous lines were established by crossing with y¹w¹¹¹⁸; Kr^If−1/SM5. Integration was confirmed via PCR across transgene–genome junctions and sequencing of selected products.

To assess the effects of different mod(mdg4) mutant alleles on the ct⁶ allele phenotype, we generated specific genotypes using a three-step crossing scheme. First, y¹ w¹¹¹⁸ ct⁶; MKRS/TM6B females were crossed with y¹ w¹¹¹⁸; mod(mdg4)*/mod(mdg4)* males, where mod(mdg4)* represents various mutant alleles. The resulting y¹ w¹¹¹⁸ ct⁶/Y; mod(mdg4)*/TM6B males were then crossed with y¹ w¹¹¹⁸ ct⁶; MKRS/TM6B females. For the final cross, y¹ w¹¹¹⁸ ct⁶; mod(mdg4)*/TM6B females from this generation were crossed with y¹ w¹¹¹⁸ ct⁶/Y; mod(mdg4)*/TM6B males. Wings from the desired y¹ w¹¹¹⁸ ct⁶/Y; mod(mdg4)*/mod(mdg4)* males were subsequently dissected and imaged using a Nikon SMZ18 stereomicroscope with a Nikon DS-Ri2 camera (Tokyo, Japan).

Genome editing within the mod(mdg4) locus in vivo

Microinjection of the plasmid mixture (500 ng/μl) was performed on D. melanogaster embryos that express Cas9 under the ubiquitous actin5C promoter (BDSC #58492). The resulting F0 flies were crossed with y¹w¹¹¹⁸ flies. F1 progeny demonstrating successful gene editing, as evidenced by mCherry expression, were then crossed with y¹w¹¹¹⁸; MKRS/TM6B to establish the y¹w¹¹¹⁸; mod(mdg4)^attPx2/TM6B line. Subsequent successive crosses facilitated the creation of y¹ w¹¹¹⁸ P(y[+t7.7]=nanos-phiC31\int.NLS)X; mod(mdg4)^attPx2, which expresses φC31 integrase driven by the maternal nanos promoter.

Plasmid DNA containing mod(mdg4) intron fragments, flanked by attB sites (200 ng/μl), was microinjected into embryos of the established platform line. The resulting F0 flies were crossed with y¹w¹¹¹⁸; MKRS/TM6B. Transgenic progeny were screened based on the loss of mCherry fluorescence (indicating successful replacement of the target locus) and the absence of eGFP (confirming specific integration of only the desired fragment, not the plasmid backbone). Carriers of the integration were then crossed to MKRS/TM6B to remove the X-linked φC31 integrase. Integration was subsequently confirmed by PCR amplification and sequencing of the genomic region adjacent to the insertion site.

Quantification of splicing efficiency and gene expression

Biological replicates, consisting of 15–30 adult males (2- to 3-day-old), were flash-frozen in liquid nitrogen. Adult males were used because expression measurements in adult females are strongly influenced by the high background signal from the ovaries and developing oocytes, making it difficult to distinguish expression in other tissues. This choice also ensured consistency with our previous analyses of mod(mdg4) expression [33]. Samples were homogenized in 200 µl of TRI Reagent^® (MRC, Cincinnati, OH, USA), brought to a final volume of 1 ml, and total RNA was extracted according to the manufacturer’s protocol. RNA was then treated with DNase I (Thermo Fisher Scientific, Waltham, MA, USA) to eliminate residual genomic DNA. Complementary DNA (cDNA) synthesisc utilized random hexamer primers and was carried out using RevertAid, Maxima Reverse Transcriptase (both Thermo Fisher Scientific), or LunaScript^® RT Master Mix (Primer-free; New England Biolabs, Ipswich, MA, USA), following the respective manufacturer’s guidelines. For quantitative PCR, Diamant TaqA polymerase (BelBioLab, Moscow, Russia) was used with either EvaGreen dye (Biotium, Inc., Fremont, CA, USA) or TaqMan probes (Lumiprobe, Moscow, Russia) (using ROX as a reference dye) on a QuantStudio^™ 6 Flex Real-Time PCR System (Thermo Fisher Scientific). Relative mRNA levels were quantified through standard curves derived from serial dilutions of genomic DNA (for housekeeping genes) or PCR products spanning splice junctions (for splicing efficiency). Student’s t-test was used to assess statistical significance. The MIQE guidelines [51] are provided in Supplementary Table S3.

Fly extract preparation

Twenty adult flies were cooled in liquid nitrogen. Homogenization was performed for 30 s with a pestle in 200 μl of extraction buffer (20 mM 4-(2-hydroxyethyl)-1-piperazineethanesulfonic acid (HEPES), pH 7.5, 100 mM KCl, 5% glycerol, 10 mM ethylenediaminetetraacetic acid (EDTA), 1% NP-40, 1% sodium deoxycholate, 0.1% sodium dodecyl sulfate (SDS), 1 mM dithiothreitol (DTT), 5 mM phenylmethylsulfonyl fluoride (PMSF), and 1:100 Calbiochem Complete Protease Inhibitor Cocktails VII and V). The resulting suspension was incubated on ice for 10 min. Subsequently, the sample underwent sonication in a Bioruptor (Diagenode, USA) for 3 min at setting H, utilizing 15-s ON and 45-s OFF cycles. Following this, 4× SDS–PAGE sample buffer was added to the homogenate. The extracts were heated to 100°C for 10 min, centrifuged at 16 000 × g for 5 min, and then loaded onto a 10% SDS–PAGE gel. Immunoblot analysis employed rabbit antibodies against the C-terminal domain of Mod(mdg4)-67.2 and mouse monoclonal anti-lamin Dm0 (clone ADL84.12, #ADL84.12, DSHB, USA).

Target enrichment and RNA sequencing

Total RNA was extracted from imago males of the control y¹w¹¹¹⁸ line and the y¹w¹¹¹⁸; mod(mdg4) ^c13/mod(mdg4)^c13line. An aliquot of 50 µl of hydrophilic streptavidin magnetic beads (NEB) were washed with 500 µl of wash/binding buffer (0.5 M NaCl, 20 mM Tris–HCl pH 7.5, 1 mM EDTA), magnetically pelleted, and subsequently blocked with 100 µl of wash/binding buffer containing 250 ng/µl of sonicated salmon sperm DNA. After an additional wash with 500 µl of wash/binding buffer, the beads were incubated for 15 min at room temperature, with rotation, in 25 µl of wash/binding buffer containing 200 pmol of a biotinylated oligonucleotide complementary to exon 4 (8 pmol/µl) RNA. This incubation was followed by two subsequent washes with 100 µl of wash/binding buffer.

Total RNA (100 µg) was denatured at 65°C, chilled on ice, and then added to the beads. This mixture was incubated at 55°C for 30 min with agitation. Five sequential washes followed. The first two washes used wash buffer (1 M NaCl, 20 mM Tris–HCl pH 7.5, 1 mM EDTA, and 0.05% Tween-20) at 55°C. The third wash employed the same buffer but without Tween-20. The fourth wash used low salt buffer [150 mM NaCl, 20 mM Tris–HCl (pH 7.5), and 1 mM EDTA], and the fifth was with ice-cold low salt buffer. RNA was eluted by adding 200 µl of TRI Reagent^® and then extracted following the manufacturer’s protocol with glycogen added prior to isopropanol precipitation.

Preparation of RNA-seq libraries followed the manufacturer’s instructions for the NEBNext^® Ultra RNA Library Prep kit (NEB, USA). Amplified libraries underwent quantification by fluorometry using a DS-11 (DeNovix, USA) and a Bioanalyzer 2100 (Agilent, USA). Diluted libraries were subsequently clustered on a pair-read flowcell and sequenced on an Illumina NovaSeq 6000 system (USA). Alignment of sequencing reads to the D. melanogaster dm6 genome [39] was performed using STAR v2.7 [52], which simultaneously assessed splice junction coverage.

Prediction of intermolecular RNA–RNA interactions

To identify potential intermolecular trans-interactions within the transcribed RNA, we employed a sliding-window approach combined with thermodynamic folding algorithms. The full-length transcripts of the mod(mdg4) locus was systematically partitioned into overlapping 300-nucleotide windows with a step size of 50 nt.

Pairwise RNA–RNA interactions between all generated windows were predicted using RNAcofold from the ViennaRNA Package [53]. This algorithm models the co-folding of two distinct RNA molecules and calculates the hybridization free energy while explicitly accounting for both intramolecular secondary structures and the accessibility of potential binding sites. The folding temperature was set to 25°C, and base-pairing probability estimates were enabled.

The resulting predictions were subjected to a stringent multistep filtering procedure. Interactions were retained for downstream analysis only if they satisfied all of the following criteria: (i) a predicted binding free energy ΔG ≤ −10 kcal/mol, (ii) an individual base-pairing probability > 0.7, (iii) >20 intermolecular base pairs, and (iv) detection in more than half of all relevant sliding-window predictions, defined as the interaction being predicted in >50% of the window pairs in which the corresponding interaction could be identified.

The filtered intermolecular interaction networks were visualized using Circos software [54].

Cell fractionation and nuclear extract preparation

D. melanogaster S2 cells were seeded in SFX medium (HyClone, Logan, UT, USA) and cultured at 25°C. For each pulldown, 10⁷ cells were collected, washed with ice-cold PBS, and homogenized with a Dounce homogenizer (tight pestle, 20 strokes) in lysis buffer (20 mM Tris–HCl 7.5, 10 mM KCl, 10 mM MgCl₂, 2 mM EDTA, 10% glycerol, 1% Triton X-100, 1 mM DTT, 1 mM PMSF, and 1× Halt Protease Inhibitor Cocktail). The homogenate was incubated on ice for 10 min and then centrifuged at 3000 × g to separate the cytoplasmic fraction from the nuclei. The nuclear pellet was resuspended in the same lysis buffer and sonicated in a Bioruptor (Diagenode, USA) for 3 min using the “L” setting (30-s ON/30-s OFF). NaCl was added to a final concentration of 500 mM, and the suspension was incubated on ice for an additional 60 min with periodic mixing. To reduce viscosity, sonication was repeated as above. Cell debris was pelleted by centrifugation at 10 000 × g for 30 min at 4°C. The supernatant (nuclear extract) was then diluted to 250 mM NaCl with incubation buffer-0 (20 mM Tris–HCl, pH 7.5, 10 mM MgCl₂, 10% glycerol, 0.1% NP-40, 0.5 mM PMSF, 2 mM Ribonucleoside Vanadyl Complex, and 1× Halt Protease Inhibitor Cocktail) and kept for subsequent incubation with RNA–Streptavidin magnetic beads.

RNA biotinylation

Two different RNAs (abc and cba) were transcribed in vitro from a T7 promoter using the Megascript kit (Ambion), according to the manufacturer’s instructions. For biotinylation, 1 μg of each RNA was mixed with 50 mM Tris–HCl pH 7.8, 10 mM MgCl₂, 1 mM DTT, 1 mM ATP, 10% DMSO, 15% PEG, 1 μM pCp biotin, 10 U T4 RNA ligase, and 50 U RNase inhibitor. The reaction was incubated overnight at 16°C. Biotinylated RNA was purified by chloroform:isoamyl alcohol (24:1) extraction and verified by dot blot analysis.

Immobilization of biotinylated RNA on streptavidin beads

Dynabeads MyOne Streptavidin T1 magnetic beads (Thermo Fisher Scientific) (20 µl per sample) were washed twice with buffer A (100 mM NaOH and 50 mM NaCl), once with buffer B (100 mM NaCl), and twice with buffer C (20 mM Tris–HCl pH 7.5). The beads were then resuspended in RNA-binding buffer (20 mM Tris–HCl pH 7.5, 1 M NaCl, and 1 mM EDTA). Biotinylated RNA was heated at 90°C for 2 min and immediately placed on ice, then added to the bead suspension and incubated at room temperature for 30 min. RNA-coated beads were washed twice with buffer C and once with RNA–protein buffer (20 mM Tris–HCl pH 7.5, 50 mM NaCl, 2 mM MgCl2, 0.1% Tween-20).

RNA pull down

For the binding reaction, 300 μl of nuclear extract (diluted to 250 mM NaCl) was mixed with 150 μl of 50% glycerol and 50 μl of 10× RNA–protein buffer (200 mM Tris–HCl pH 7.5, 500 mM NaCl, 2 mM MgCl₂, and 1% Tween-20). This mixture was added to the RNA-coated beads and incubated overnight at 4°C with gentle rotation. After three washes with wash buffer (10 mM Tris–HCl pH 8.0, 150 mM NaCl, and 2 M urea), bound proteins were eluted with elution buffer (10 mM Tris–HCl pH 8.0, and 8 M urea).

In-gel solution protein digestion

Protein concentrations in biological samples after RNA affinity purification were determined Quick Start Bradford protein assay (Bio-Rad), according to the manufacturer’s standard protocol (bovine serum albumin was used as the standard). Disulfide bonds were reduced with DTT in 5 mM final concentration for 30 min at room temperature (RT). Afterward, iodoacetamide was added to a final concentration of 10 mM. The samples were incubated at RT for 20 min in the dark, and the reaction was stopped by adding DTT to a final concentration of 5 mM for 20 min. Then, samples was diluted with ammonium bicarbonate solution to reduce urea concentration to 4 M. LysC (Promega) was added at the ratio of 1:100 w/w and the samples were incubated for 4 h at 37°C. Then, samples were diluted with ammonium bicarbonate solution to reduce urea concentration to 2 M, trypsin (Promega) was added at the ratio of 1:100 w/w, and the samples were incubated for 14 h at 37°C. After 14 h, the reaction was stopped by the addition of trifluoroacetic acid (TFA) to the final concentration of 0.5%. The tryptic peptides were desalted using SDB–RPS membrane (Sigma–Aldrich) and vacuum-dried. Prior to LC–MS/MS analysis, samples were redissolved in 5% acetonitrile (ACN) with 0.1% TFA solution and sonicated.

LC–MS/MS analysis

Proteomic analysis was performed using a Q Exactive HF mass spectrometer. Samples were loaded onto 50-cm columns packed in-house with C18 3 μM Acclaim PepMap 100 (Thermo Fisher), with an Ultimate 3000 Nano LC System (Thermo Fisher) coupled to the MS (Q Exactive HF, Thermo Fisher). Peptides were loaded onto the column thermostatically controlled at 40°C in buffer A (0.2% formic acid) and eluted with a linear (120 min) gradient of 4%–55% buffer B (0.1% formic acid and 80% acetonitrile) in A at a flow rate of 350 nl/min. Mass spectrometric data were stored during automatic switching between MS1 scans and up to 15 MS/MS scans (topN method). The target value for MS1 scanning was set to 3 × 10⁶ in the range 300–1200 m/z with a maximum ion injection time of 60 ms and a resolution of 60 000. The precursor ions were isolated at a window width of 1.4 m/z and a fixed first mass of 1000 m/z. Precursor ions were fragmented by high-energy dissociation in a C-trap with a normalized collision energy of 28 eV. MS/MS scans were saved with a resolution of 15 000 at 400 m/z and at a value of 1 × 10⁵ for target ions in the range of 200–2000 m/z with a maximum ion injection time of 30 ms.

Protein identification and spectral counting

Raw data files acquired on the Q Exactive HF-X were converted to Mascot generic format (mgf) peak lists using MSConvert (ProteoWizard Software Foundation) with the following parameters: “–mgf –filter peakPicking true [1,2].” The resulting peak lists were searched against the UniProt Knowledgebase (taxon 7227 D. melanogaster; downloaded 13/04/2026 with 44 174 protein sequences) concatenated with a reversed decoy set using MASCOT (version 2.5.1, Matrix Science Ltd., UK) and X! Tandem (ALANINE, 2017.02.01, The Global Proteome Machine Organization). For both search engines, precursor and fragment mass tolerances were set to 20 and 50 ppm, respectively. Database-searching parameters included the following: tryptic digestion with one possible missed cleavage, static modification for carbamidomethyl (C), and dynamic/flexible modifications for oxidation (M). For X! Tandem we also selected parameters that allowed a quick check for protein N-terminal residue acetylation, peptide N-terminal glutamine ammonia loss, or peptide N-terminal glutamic acid water loss. Result files were submitted to Scaffold 5 software (version 5.1.0) for validation and meta-analysis. We used the local false discovery rate scoring algorithm with standard experiment-wide protein grouping. For the evaluation of peptide and protein hits, a false discovery rate of 5% was selected for both. False positive identifications were based on reverse database analysis. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE [55] partner repository with the dataset identifier PXD079014 and 10.6019/PXD079014.

Results

The role of the proximal region of the fourth intron in trans-splicing

Previous studies have demonstrated that the fourth intron of the mod(mdg4) gene plays a key role in trans-splicing [34, 33, 40]. To determine whether the fourth intron alone is sufficient to induce trans-splicing, we used a previously established model system consisting of donor and acceptor transgenes integrated into the 22A genomic locus (Fig. 2A and Supplementary Fig. S1). The reference donor construct contained the mod(mdg4) promoter and the gene body comprising four exons and three introns (M:mod_i4). An IRES sequence from the reaper gene was inserted downstream of exon 4 to enable specific RT-PCR detection. In the M:mod_i4 donor, the IRES was followed by a 40 bp fragment of exon 4 and the fourth intron (i4) of mod(mdg4). The acceptor transgene contained the bidirectional promoter that, in the endogenous locus, drives transcription of the T and K isoforms, together with their corresponding outrons, while the coding regions were replaced with Fluc and mCherry reporters (F-mC) (Fig. 2A).

Figure 2.

Functional analysis of the proximal region of intron 4 in trans-splicing of mod(mdg4). (A) Schematic representation of the model trans-splicing system. The donor transgene (top) contains the mod(mdg4) promoter (arrows and semicircles), constitutive exons (rectangles), introns (lines), an IRES sequence, and intron 4 with the 5′ splice donor site (the red semicircle). The acceptor transgene (bottom) contains the bidirectional promoter associated with the T and K isoforms, together with their corresponding outrons (gray lines). Coding regions were replaced with the Fluc (yellow) and mCherry (red) reporter genes. Transcription produces one donor pre-mRNA and two alternative acceptor pre-mRNAs. Trans-splicing generates two mature mRNA isoforms consisting of the common 5′ region fused to either the T- or K-specific 3′ region. Red and black arrows indicate primer positions used to quantify trans-splicing products and donor transcript levels, respectively. (B) Trans-splicing supported by a donor construct containing only intron 4 and a short fragment of exon 4 from mod(mdg4). The chimeric donor transgene consists of the UAS promoter and a grip-derived gene body retaining only intron 4 from mod(mdg4). Relative donor transcription levels and trans-splicing efficiencies for the T and K isoforms are shown on the right. (C) The effects of deletions and inversions within regions a and b on trans-splicing efficiency. The schematics of the donor constructs are shown on the left. The donor segment is driven by either the endogenous mod(mdg4) promoter (green), the Act5C promoter (orange), or the UAS promoter (blue), and contains gene-body sequences derived from either mod(mdg4) (green) or grip (pink). The splice donor site is indicated by the red semicircle. Deletions are shown as dashed lines and inversions as black arrows. The dark gray segment corresponds to a control intronic fragment derived from ATPsynC. The construct labeled “i3^+ab; Δab” contains the a–b region inserted into intron 3 of the grip gene derivative. Construct names are centrally located. The relative trans-splicing efficiencies for the T and K isoforms are shown on the right. (D) The effects of inversion of region c or the entire abc fragment on trans-splicing efficiency. Construct organization and graphical annotations are as described in panel (C). Each trans-heterozygous combination was analyzed in at least three biological replicates. Error bars represent standard deviations (n = 3–13). Statistical significance was assessed relative to the corresponding control: *P < 0.05, **P < 0.01, ***P < 0.001.

We generated a chimeric donor construct composed predominantly of heterologous sequences to determine whether the fourth intron alone was sufficient to induce trans-splicing. Specifically, we used a chimeric UAS promoter consisting of five binding sites for the yeast transcriptional activator GAL4 and the minimal hsp70 promoter core that drives weak ubiquitous transcription in the absence of GAL4 activation. We inserted a fragment of the grip gene downstream of the promoter that featured an exon-intron organization comparable to that of mod(mdg4) [40] (UAS:grip_i4; Fig. 2B and Supplementary Fig. S2). In this construct, only the fourth intron of mod(mdg4) and a short fragment of exon 4 were retained.

The trans-splicing efficiency observed in heterozygous M:mod_i4/F-mC males was defined as the reference value of 1. In UAS:grip_i4/F-mC males, trans-splicing efficiency was reduced to approximately 10% of the control level (Fig. 2B), consistent with the substantially lower transcriptional activity of the UAS promoter relative to the endogenous mod(mdg4) promoter. To increase transcription from the UAS promoter, we introduced an Act5C-GAL4 transgene (Bloomington Drosophila Stock Center #4414) into the UAS:grip_i4/F-mC background by genetic crossing (Fig. 2B). Activation of GAL4 increased transcription from the UAS promoter by approximately 30-fold, resulting in a corresponding 4–5-fold increase in trans-splicing efficiency. Because the UAS:grip_i4 donor retains only the fourth intron of mod(mdg4), these results indicate that sequences within this intron are sufficient to promote trans-splicing.

The cis-elements required for trans-splicing are expected to reside within the proximal 600 bp of the fourth intron, as transcription is interrupted downstream at a conserved termination site (Fig. 1B) [32]. Previous studies have focused on a 300–450 bp interval corresponding to TSA/region c that contains the conserved 13 bp motif (c¹³). Although deletion of this region substantially reduced trans-splicing efficiency, it did not completely abolish it, suggesting the existence of additional regulatory elements within the intron. In a previous study, we subdivided the proximal 600 bp region of intron 4 into four segments: region a (11–108 bp), region b (109–295 bp), region c (296–412 bp), and region d (413–600 bp) (Fig. 1B) [33]. Deletion of either region a or b individually only slightly reduced trans-splicing efficiency. In this study, we repeated the analysis of these deletion variants and generated an additional construct that lacks both regions. Consistent with previous observations [33], deletion of region a had a minimal impact on trans-splicing, and deletion of region b had no effect (Fig. 2C and Supplementary Fig. S3A). In contrast, simultaneous deletion of regions a and b completely abolished trans-splicing. To preclude the possibility that this resulted from altered spacing between the splice donor site and region c, we replaced the deleted a–b interval with a similarly sized intronic fragment derived from the ATPsynC gene. This substitution failed to restore trans-splicing, indicating that regions a and b contained partially redundant cis-elements required for efficient trans-splicing.

To determine whether the function of the a–b region depends on its localization within intron 4, we transferred this combined fragment into intron 3 of the grip gene within the Act:grip_i4 construct driven by the Act5C promoter (Fig. 2D and Supplementary Fig. S3B). The process completely abolished trans-splicing, indicating that the a–b region must be positioned within intron 4 to support trans-splicing.

To further investigate the function of the a–b region, we generated a derivative of the M:mod_i4 donor in which the a–b region was inverted. This inversion was designed to disrupt cis-elements recognized at the RNA level while preserving the underlying DNA sequence and potential DNA-binding motifs. Inversion of the a–b region completely suppressed trans-splicing (Fig. 2C), indicating that this region likely contained sequence-specific RNA elements required for trans-splicing.

Because region c is also essential for efficient trans-splicing, we next tested whether its activity was preserved upon inversion (Fig. 2D and Supplementary Fig. S3B). Using a previously characterized donor construct lacking region c as a reference, we observed a similarly strong reduction in trans-splicing for both variants. The results indicated that inversion inactivates the function of region c. To exclude the possibility that independent inversion of regions ab and c disrupts their relative arrangement rather than the cis-elements themselves, we generated a donor construct in which the entire abc was inverted. This modification also completely abolished trans-splicing, indicating that the proximal region of intron 4 contained orientation-dependent RNA cis-elements required for trans-splicing.

Testing the role of complementary sequences in trans-splicing of the mod(mdg4) gene

According to the widely accepted model of trans-splicing, pairing between complementary sequences in distinct pre-mRNAs promotes splice-site proximity and facilitates trans-splicing (Fig. 3A) [20, 21, 24]. Previous analyses of potential RNA–RNA interactions within the mod(mdg4) locus were performed using BLAST-based searches [34] and local pairwise sequence alignment [33]. However, these approaches would not identify significant complementary interactions. Importantly, neither analysis considered the contribution of intramolecular RNA secondary structures, which can substantially alter nucleotide accessibility and the spatial organization of potential interaction sites.

Figure 3.

Analysis of potential complementary RNA interactions in mod(mdg4) trans-splicing. (A) Schematic representation of a potential intermolecular RNA duplex formed between donor and acceptor pre-mRNAs. Base pairing between complementary sequences may promote spatial proximity of the transcripts and facilitate trans-splicing. (B) CIRCOS diagram showing predicted intermolecular RNA duplexes within the mod(mdg4) locus. Only interactions with duplex lengths ≥ 25 nt and predicted free energy ΔG < −10 kcal/mol are shown. The thickness of lines is proportional to duplex length. For clarity, lines denoting interactions are connected to the centers of locus features and do not indicate the precise positions of predicted duplexes. (C) CIRCOS diagram showing predicted RNA duplexes between the proximal region of intron 4 and other transcribed regions of the mod(mdg4) locus. Ribbons represent interactions. For visualization purposes, intron 4 is shown enlarged, and ribbon positions correspond to the predicted locations of intermolecular RNA duplexes. (D) Functional analysis of the roles of outron sequences and complementary RNA pairing in trans-splicing. Schematics of the acceptor constructs are shown on the left. Relative transcript levels and trans-splicing efficiencies are shown in the center and on the right, respectively. The c-region was inserted in either forward or reverse orientation as indicated by the arrows within the pink boxes. Error bars represent standard deviations from three biological replicates. Statistical significance was assessed relative to the corresponding control: *P < 0.05, **P < 0.01, ***P < 0.001.

To address this limitation, we reanalyzed potential trans-interactions between RNAs transcribed from the mod(mdg4) locus using RNA secondary structure-aware predictions (Fig. 3B). The entire transcribed region of the mod(mdg4) locus was subdivided into 300 nt fragments using a sliding window with a 50 nt step size. Pairwise interactions among all fragments were then predicted using RNAcofold [53], which models intermolecular duplex formation while accounting for intramolecular secondary structures and nucleotide accessibility. Candidate interactions were selected based on low predicted ΔG values and reproducibility across different window positions.

The analysis identified multiple regions with the potential to mediate intermolecular RNA pairing. However, the majority of predicted interactions were localized within the clusters of alternative 3′ exons (Fig. 3B). Only a small fraction of the predicted complementary sequences were mapped within the proximal 600 bp of the fourth intron, and most of these were confined to region d, which previously showed minimal contribution to trans-splicing efficiency [33]. Notably, only a single interaction was predicted between the donor and acceptor transgenes of our model system, corresponding to pairing between region d of the donor intron and the K outron of the acceptor construct. However, deletion of region d did not impair trans-splicing [33]. In addition, the T outron–Fluc acceptor construct, for which no stable interactions with intron 4 were predicted, remained competent for trans-splicing. Taken together, these observations argue against a major role for complementary RNA pairing in promoting trans-splicing at the mod(mdg4) locus.

We next tested whether acceptor-derived sequences were required for trans-splicing. To this end, we generated an F-mC(rpl) acceptor transgene, in which the outron regions extending from the transcription start sites to the acceptor splice sites, together with short fragments of the T and K exons, were replaced with sequences of an equivalent length derived from the 3′ regions of the rpl10 and rpl23 introns, respectively (Fig. 3D). These heterologous intronic fragments retained the canonical 3′ splicing signals. These substitutions did not reduce transcription from the bidirectional promoter; instead, transcript levels increased. Nevertheless, despite the presence of the native bidirectional promoter and intact 3′ splice acceptor signals, trans-splicing was completely abolished for both acceptor variants. The results suggested that sequences located within the native outron regions upstream of the 3′ exon clusters provide essential determinants required for trans-splicing.

We next asked whether sequences from region c of intron 4 could restore trans-splicing in the F-mC(rpl) construct. To test this hypothesis, we generated two derivative acceptor transgenes in which region c was inserted into the rpl10-derived sequence either in the forward orientation (F-mC(cf)), preserving the native arrangement of the cis-elements, or in the reverse orientation (F-mC(cr)), enabling potential base pairing between complementary region c sequences in the donor and acceptor pre-mRNAs (Fig. 3D). However, none of these constructs restored trans-splicing activity. Together, the results do not support a model in which complementary pairing between donor and acceptor pre-mRNAs is the primary determinant of trans-splicing of the mod(mdg4) locus. Instead, trans-splicing efficiency appears to depend on specific orientation-dependent cis-elements within the donor intron and, in addition, sequence determinants located within the native acceptor outron regions.

Trans-splicing is stimulated by several redundant sequences in the 117-bp region c

Previous studies [33] established that a 117-bp region (designated as the “c region,” located at positions 296–412 bp relative to the 5′ end of intron 4; see Fig. 4A) is critical for trans-splicing. Within this region lies a highly conserved 13-bp motif (c¹³) that is complementary to U1 snRNA [34]. However, deletion of c¹³ had minimal impact on trans-splicing in a model system and at the endogenous locus [33], implying that other crucial motifs must reside within the c region.

Figure 4.

Identification of potentially functional motifs within the c region and their role in trans-splicing. (A) Sequence organization of the 117 bp c region, subdivided into the c⁴⁴ and c⁷³ fragments. Partially repeated motifs are highlighted in green, red, and blue. The conserved c¹³ motif among Drosophilidae is shown in yellow. A duplicated motif identified in D. virilis is indicated in magenta. (B) Alignment of repeated sequence motifs identified within region c. Consensus sequences are shown below. (C) The sequence of the region homologous to c from Drosophila virilis. Conserved motifs shared with D. melanogaster are highlighted using the same color scheme as in panel (A). (D–F) Functional analysis of mutations within region c. Schematics of donor constructs are shown on the left, and the corresponding trans-splicing efficiencies for the T and K isoforms are shown on the right. All constructs were derived from the donor transgene containing the mod(mdg4) promoter and constitutive gene region (green). The colored boxes correspond to the motifs shown in panel (A–C). Dashed lines indicate deletions, and nucleotide substitutions are indicated by diagonal hatching. Each trans-heterozygous combination was analyzed in at least three biological replicates. Error bars represent standard deviations (n = 3–13). Statistical significance was assessed relative to the corresponding control: *P < 0.05, **P < 0.01, ***P < 0.001.

To identify these elements, we conducted a detailed sequence analysis of the c region utilizing the MEME Suite software [56]. The analysis revealed three potential repetitive motifs (Fig. 4A and B). Although the E-values for these motifs exceeded 0.05, we considered their potential biological significance based on conservation and context. The motifs were subsequently color-coded and designated as G (green), R (red), and B (blue). Given our prior finding that the D. virilis intron supports trans-splicing in our model system [57], we conducted a sequence analysis of this orthologous intron, and identified both a conserved c¹³ motif and a G motif (Fig. 4C). Additionally, we identified a novel duplicated CG-rich motif that we designated M (magenta). Notably, a similar M motif was identified in D. melanogaster near c¹³, suggesting potential functional conservation across Drosophila species.

To evaluate the importance of these motifs within the c region, we utilized a genetic model system based on crossing flies carrying donor transgenes with those carrying acceptor transgenes (Fig. 2A). In accordance with our previous findings, deletion of the entire c region almost completely abolished trans-splicing in this model system (Fig. 4D).

In our first series of experiments, we investigated the roles in trans-splicing of the proximal c⁴⁴ (G1R1B1R2) and distal G2R3 sequences from the c region (296–412 bp) (Fig. 4E, and Supplementary Figs S4 and S5). Deletion of the proximal c⁴⁴ region (Δc⁴⁴), which encompasses motifs G1, R1, B1, and R2, had no discernible effect on trans-splicing efficiency. Secondary structure modeling indicated that sequences within the c⁴⁴ region could form a stable RNA hairpin (Fig. 1B), which, together with individual motifs, might contribute to trans-splicing. Conversely, deletion of the region spanning 382–572 bp (which includes motifs G2 and R3) led to an approximately 50% reduction in trans-splicing efficiency for isoform T, without significantly affecting isoform K.

While our model system confirmed that the deletion of the c¹³ motif alone had only a minimal effect on trans-splicing (Fig. 4E), we aimed to investigate the potential additive effects arising from its simultaneous deletion with other segments of the 117-bp c region. The simultaneous deletion of c¹³ and c⁴⁴, or c¹³ and G2R3, resulted in a severe reduction in trans-splicing (Fig. 4F). Thus, c⁴⁴, c¹³, and G2R3 appear to function cooperatively to enhance the efficiency of trans-splicing.

Our subsequent investigation focused on the roles of individual segments within the 340–367-bp (M-B2) region (Fig. 4F and Supplementary Fig. S4). Neither individual deletion nor mutation of these components significantly affected trans-splicing. However, the most substantial decrease was observed following the deletion of a segment containing the M motif (a motif conserved between D. virilis and D. melanogaster) combined with its adjacent sequences (Δ340–350). These findings collectively supported a model wherein multiple motifs within the 340–367-bp (M + B) region function cooperatively. Furthermore, mutation of the M motif in combination with the deletion of c¹³ led to an almost complete inactivation of trans-splicing.

Taken together, these findings demonstrate an unexpectedly high degree of redundancy and cooperation among the diverse motifs in this region that collectively act to promote efficient trans-splicing.

Confirmation of interchangeability and redundancy of motifs in trans-splicing at the endogenous mod(mdg4) locus

The model system based on donor and acceptor transgenes integrated into the 22A genomic locus indicated that a cluster of motifs within the c region plays an important role in mediating trans-splicing.

To substantiate these findings at the endogenous mod(mdg4) locus, we engineered a landing platform for targeted integration of various segments of intron 4 (mod(mdg4)^attPx2). This platform involved replacing the native sequence spanning 226–553 bp of the fourth intron with the same mod(mdg4) sequences along with an mCherry reporter gene; both were flanked by two attP recombination sites (Fig. 5A). The viability of homozygous mod(mdg4)^attPx2 flies confirmed that this genomic modification did not disrupt essential biological functions. Using genetic crosses, we first established a homozygous mod(mdg4)^attPx2 line that also carried the transgene expressing the φC31 integrase under the control of the nanos promoter (y¹w¹¹¹⁸ P(y[+t7.7]=nanos-phiC31\int.NLS)X; mod(mdg4)^attPx2). The integrated sequences were flanked by attB sites. The eGFP reporter gene was inserted outside the attB sites, marking the plasmid backbone and enabling the selection of successful transformants that did not express eGFP or mCherry. Plasmids containing the test sequences, along with a control plasmid carrying the intact DNA fragment spanning the 226–553-bp positions, were then injected into mod(mdg4)^attPx2 embryos.

Figure 5.

Functional dissection of cis-elements required for trans-splicing at the endogenous mod(mdg4) locus. (A) The strategy employed in genome engineering of the fourth intron. Cas9 cleavage sites within regions b and d are indicated by the scissors. The donor repair plasmid contains homologous arms (dashed lines) flanking the intact c region together with two attP integration sites. Successful integration events were identified using the 3xP3-mCherry marker cassette. Subsequently, φC31-mediated recombination was used to introduce modified intronic fragments flanked by attB sites, thereby replacing the marker cassette. The 3xP3-eGFP marker was used to confirm integration of the plasmid backbone and was not incorporated into the tested alleles. (B) Functional analysis of distinct c-region fragments. Schematics of tested intronic variants are shown on the left. Putative motifs are represented by colored rectangles, and deleted regions are indicated by dashed lines. Relative trans-splicing efficiencies for the T and K isoforms are shown on the right. (C) Functional analysis of various combinations of conserved motifs. The percentage of viable homozygous adults is indicated in the center. Relative trans-splicing efficiencies for the T and K isoforms are shown on the right. (D) The effect of spacer-sequence mutations between conserved motifs on trans-splicing efficiency. Putative motifs are shown as colored rectangles, whereas mutated regions are indicated by diagonal hatching. Trans-splicing efficiencies for the T and K isoforms measured in pupae are shown on the right. Each line was analyzed in at least three biological replicates. Error bars represent standard deviations (n = 3–18). Statistical significance was assessed relative to the corresponding control: *P < 0.05, **P < 0.01, ***P < 0.001.

We initiated the experiment by substituting the 226–553 bp region with a linker sequence comprising only a MCS, thus generating the mod(mdg4)^Δbcd. This modification led to a pronounced decrease in fly viability. Approximately 90% of homozygous flies died during the pupal stage, with the remaining few survivors being sterile and dying within days. Considering that null mod(mdg4) mutants typically display lethality during metamorphosis [58, 59], the mod(mdg4)^Δbcd mutant likely exhibits considerably diminished expression of its isoforms. This outcome highlights the indispensable role of this genomic region for the proper functioning of the mod(mdg4) locus.

Our subsequent experiment aimed to determine whether the intact intron fragment (226–553 bp), the c¹¹⁷ segment (296–412 bp), and its subregions c⁴⁴ (296–339 bp) and c⁷³ (340–412 bp) could restore trans-splicing and thereby restore fly viability (Fig. 5B and Supplementary Figs S6, S9A, and S11, and Supplementary Table S1). In contrast to the lethal effects observed in homozygous mod(mdg4)^Δbcd flies, the integration of each of these tested DNA fragments restored the viability of the mod(mdg4) mutants (Supplementary Fig. S11 and Supplementary Table S1).

Trans-splicing efficiency was quantified by extracting total RNA from 2- to 3-day-old males, followed by RT-qPCR. The assay utilized primers designed to span the trans-spliced junctions, with the forward primer annealed to the fourth exon of the donor and the reverse primer targeting one of the alternative exons (T, K, or Z). The probes were specifically designed to cover the junction between the donor and acceptor exons. All results were normalized against the expression of the housekeeping genes vha100-1 and CG9067. For comparative analysis, trans-splicing efficiency for isoforms T, K, and Z was set relative to the control line y¹w¹¹¹⁸, which was defined as 1.0.

Restoration of the intact mod(mdg4) intronic sequence (a 226–553 bp fragment) resulted in trans-splicing levels comparable to those observed in the control y¹w¹¹¹⁸ line (Fig. 5B and Supplementary Fig. S9A). This result demonstrated that neither the genomic editing of the mod(mdg4) intron nor the retention of attL/R sites within the b and d regions compromised trans-splicing efficiency. Conversely, introduction of only the c¹¹⁷ sequence resulted in a slight reduction in trans-splicing relative to the control, likely due to the partial deletions of the adjacent b and d regions in this construct. Insertion of the b (226–296 bp) or b + d (226–296 + 413–553 bp) DNA fragments into mod(mdg4)Δbcd flies rescued their viability and partially restored trans-splicing efficiency (Fig. 5B).

When evaluated separately, the c¹¹⁷ subregions–the c⁴⁴ (the proximal portion) and c⁷³ (the distal portion)–yielded distinct results. The c⁴⁴ fragment exhibited weak induction of trans-splicing, whereas c⁷³ sustained trans-splicing at a level only slightly below that of c¹¹⁷. Notably, c⁷³ contains the c¹³ and M motifs, both of which were implicated in promoting trans-splicing in our model system. We then generated several truncated derivatives of the c⁷³ segment to identify the critical determinants of trans-splicing within this region (Fig. 5B). In the 347–412 bp and 368–412 bp DNA fragments, sequential deletion of the M and B2 motifs from the proximal side of the c⁷³ region led to a progressive reduction in trans-splicing, consistent with a functional role for both motifs. Similarly, in the 340–394 bp and 340–380 bp DNA fragments, the G2–R3 and spacer sequences were sequentially removed from the distal end of the c⁷³ region, and this also correlated with a stepwise decrease in trans-splicing efficiency. While the c¹³ motif alone induced relatively low trans-splicing activity, the c¹³ motif flanked by the B2 motif and spacer sequences (the 347–394 bp fragment) resulted in a relatively high level of trans-splicing. Taken together, these results suggest that the intervening spacer sequences also contribute to trans-splicing efficiency. The role of these spacer elements is further supported by a twofold increase in trans-splicing induced by the native 340–380 bp (Fig. 5B) fragment compared to an artificial analog containing only the M, B2, and c¹³ motifs (M + B2 + c¹³) (Fig. 5C). It is likely that the functional motifs are wider than predicted, that strict requirements for precise inter-motif spacing exist, or that additional essential motifs reside within these spacers. Therefore, the results collectively suggest that both the exact length of the trans-splicing-determining motifs and their relative genomic positions are critical to their functioning.

We also tested whether duplication of the upstream a–b region could compensate for the 226–553 bp deletion and restore trans-splicing by inserting the a + b (11–295 bp) DNA fragment into the mod(mdg4)attPx2 platform (Fig. 5B). Duplication of the a + b region only weakly restored trans-splicing relative to the b (226–295 bp) control. This result suggests that efficient trans-splicing requires a diverse combination of motifs for different RBPs, rather than the simple duplication of a specific subset of these elements.

Motifs and spacer sequences are essential for trans-splicing

Since the c¹³ motif alone yielded very low trans-splicing activity, we asked whether the addition of another motif, such as M, R3, B2, or G2, could enhance trans-splicing efficiency (Fig. 5C and Supplementary Fig. S7). We utilized the mod(mdg4)^attPx2 platform to integrate c¹³ in combination with each individual motif. Despite the demonstrated role of these elements in promoting trans-splicing, their pairwise combination with c¹³ did not lead to a marked increase in trans-splicing efficiency. This observation further supported the hypothesis that the intervening sequences between motifs are essential for their functional activity.

Although the isolated c¹³ motif only marginally increased trans-splicing when integrated into the mod(mdg4)^attPx2 platform, it completely restored mutant viability compared to the Δbcd line (Fig. 5C, and Supplementary S11 and Supplementary Table S1). Thus, even minor increases in trans-splicing efficiency are critical for the sufficient expression of Mod(mdg4) isoforms necessary for survival. We then tested whether other individual motifs integrated into the mod(mdg4)^attPx2 platform could rescue viability similarly to c¹³. We found that all tested motifs (M, B2, G2, and R3) only weakly restored trans-splicing efficiency, mirroring the effect of c¹³ alone (Fig. 5C). Notably, M, G2, and R3 partially restored mutant viability, albeit less effectively than c¹³, whereas integration of the B2 motif resulted in only a slight increase in survival. Interestingly, the combined integration of M and B2 almost completely rescued the viability of the resulting mutants.

These results demonstrate that the G2, R3, M, and B2 motifs are actively involved in driving trans-splicing, corroborating the conclusions derived from the transgenic model system (see Fig. 4). Collectively, a distinct group of motifs within the fourth intron contributes to trans-splicing efficiency, with varying degrees of impact: the G and R motifs have minor roles, whereas the M and c¹³ motifs provide the most substantial contributions. Remarkably, the absence of any single motif has only a minimal impact on trans-splicing efficiency, further underscoring their functional redundancy.

To further of the functions of the sequences between motifs, derivatives of the c¹¹⁷ fragment were integrated into the mod(mdg4)^attPx2 platform. These included a variant in which all nucleotides located between the motifs underwent replacement, effectively retaining only the motifs themselves (designated as mod(mdg4)^om, only motifs), and another variant that combined substitutions of these inter-motif nucleotides with specific mutations within the c¹³ and M motifs (mod(mdg4)^mutM&c13) (Fig. 5D, and Supplementary Fig. S8 and S9C). Substitution of the nucleotides between motifs within the mod(mdg4)^om line resulted in a considerable reduction in trans-splicing efficiency relative to the y¹w¹¹¹⁸ control line. Surprisingly, homozygous mod(mdg4)^mutM&c13 flies died during the pupal stage. Thus, insertion of the mutant DNA fragment at the attP site results in lethality of pupae compared to the pattern observed in the mod(mdg4) null mutants [58, 59]. Accordingly, trans-splicing was detected only at a rudimentary level in mod(mdg4)^mutM&c13 flies.

Taken together, the results indicate that the spacers between the identified motifs play an active role in trans-splicing, likely contributing to the core motifs recognized by RBPs.

Identification of proteins interacting with the proximal region of the last common intron

We performed in vitro RNA affinity purification followed by mass spectrometry to identify RBPs associated with the proximal 412 bp region of the fourth intron. We conducted in vitro transcription of an RNA fragment termed abc that contained the last 20 nucleotides of exon 4 together with intronic regions a, b, and c. As a control, we generated a second RNA fragment, termed cba, in which the sequence spanning nucleotides 11–412 of the intron was inverted. This inversion corresponds to the abc^inv transgene previously analyzed in vivo (Fig. 2D), where trans-splicing was completely abolished. The RNAs were biotinylated at the 3′ end and used for pull-down of RNA-associated protein complexes on streptavidin-coated magnetic beads.

RNA-coated magnetic beads were incubated with a nuclear extract from S2 cells. RNA-associated proteins were identified by label-free quantitative mass spectrometry (MS). After filtering (see Materials and methods), 188 proteins (Supplementary Table S4) were retained for further analysis. A protein interaction analysis using the STRING database [60] identified three groups of proteins with annotated functional interactions, suggesting the presence of distinct RNA-associated protein complexes (Fig. 6).

Figure 6.

Identification of the protein complexes associated with the proximal region of intron 4. (A, E, and F) Proteins enriched following in vitro RNA affinity purification coupled with mass spectrometry using the abc RNA fragment or the inverted cba control RNA. The abc RNA contained the proximal intronic regions abc in the native orientation, whereas the control cba RNA carried the same sequence in the inverted orientation corresponding to the abc^inv construct. The tables show normalized count values obtained from three independent experiments, the average fold enrichment, the annotated RNA-binding activity, the availability of experimentally determined RNA-binding motifs, and the presence of predicted binding sites within the c¹¹⁷ region. Statistical significance of protein enrichment relative to the control sample: *P < 0.05, **P < 0.01, ***P < 0.001. (B, G, and H) Interaction networks of proteins enriched on the abc RNA fragment based on annotated interactions from the STRING [60] and FlyBase [36] databases. Nodes represent proteins identified by affinity purification, and edges indicate experimentally supported or annotated functional interactions. (C) The distribution of predicted binding motifs for proteins from the first RNA-binding subcomplex across the c¹¹⁷ region of D. melanogaster. Binding sites were identified using experimentally determined RNA-binding motifs available for D. melanogaster proteins [61]. Conserved and repeated sequence motifs shown in Fig. 4 are indicated using the same color scheme. (D) The distribution of predicted binding motifs within the homologous c¹¹⁷ region of D. virilis. Binding sites were predicted using RNA-binding motif models available for the corresponding D. virilis orthologs [61]. Conserved sequence motifs are shown using the same color coding as in (C).

The most prominent subgroup comprised 11 proteins, 10 of which possessed annotated RNA-binding activity (Fig. 6A and B). RNA-binding motifs have been experimentally characterized for six proteins in this group [61, 62]. Notably, the motifs for Hrb98DE and TBPH, which directly interact, were located adjacent to one another within the c¹³ element (Fig. 6C). The motif for Syp is positioned near c¹³ within a spacer region that also contributes to trans-splicing efficiency (Fig. 5B). The SC35 motif matched the R2 sequence, while a similar motif recognized by the related splicing factor Sf2 was likewise found exclusively within R2. No binding motif for Rsf1 was identified within region c. Unfortunately, the binding motifs of the remaining four RBPs in this subgroup are unknown.

We also compared the distribution of predicted RNA-binding motifs between D. virilis and D. melanogaster. As expected, orthologs of Hrb98DE and TBPH were predicted to bind the conserved c¹³ region. The predicted binding sites for the D. virilis orthologs of Sf2 and SC35 were shifted downstream of c¹³ relative to their positions in D. melanogaster. Notably, SC35 may potentially bind two adjacent sites in the D. virilis sequence. Although no Syp-binding site was detected in D. virilis, a predicted binding site for Yps, detected among the enriched proteins, was identified adjacent to the Sf2 and SC35 motifs.

A second subgroup corresponded to components of the U2 snRNP-associated complex, although only four proteins in this group were annotated as RNA-binding (Fig. 6E and G). Because binding motifs for these proteins have not yet been determined, their direct interaction with the intronic RNA cannot currently be assessed.

Finally, we identified a third subgroup consisting of proteins associated with RNA interference effector complexes [63], including Ago2, Vig, and dFXR/Fmr1 (Fig. 6F and H). A known binding motif for Fmr1 was identified within region b of the mod(mdg4) intron. However, the binding motifs of the other RBPs in this group remain unknown.

Taken together, these analyses identified three interacting protein subnetworks enriched in the proximal region of intron 4. Notably, one protein complex showed a strong association with the intronic region functionally required for trans-splicing.

Impaired trans-splicing reduces mod(mdg4) mRNA isoforms but spares Mod(mdg4)-67.2 protein abundance and function

The c¹³ flies were fully viable and displayed a normal phenotype despite a more-than-10-fold reduction in trans-splicing efficiency for the T, K, and Z isoforms. The lack of an observable functional consequence despite such a substantial decrease in mRNA encoding these protein isoforms remained unclear. One plausible explanation would be that the observed reduction in trans-splicing impacts only a subset of the 3′ alternative exons. Consequently, to explore this hypothesis, we conducted RNA sequencing on mod(mdg4) gene-enriched RNA samples derived from the mod(mdg4)^c13 and control y¹w¹¹¹⁸ lines. To quantify trans-splicing levels, we compared the number of reads spanning alternative trans-splice junctions in both lines, normalized against reads spanning the constitutive exon 3–exon 4 splice junction (Fig. 7A). Our analysis revealed that all isoforms exhibited considerably reduced trans-splicing efficiency in the mod(mdg4)^c13 line. Simultaneously, there was an approximately 2-fold increase in the splicing efficiency of 3′ exons encoding the N and O isoforms. This latter phenomenon was most likely a consequence of cis-splicing becoming active, a pathway typically suppressed at the native mod(mdg4) locus [33].

Figure 7.

Reduced trans-splicing efficiency does not impair Mod(mdg4)-67.2 protein function. (A) RNA-seq analysis of the trans-splicing efficiency of mod(mdg4) isoforms in 2- and 3-day-old adult males from the c¹³ mutant and the y¹w¹¹¹⁸ control line. For each isoform, the number of reads spanning the trans-splice junction was normalized to the number of reads spanning the exon 3–4 splice junction within the constitutive region of mod(mdg4). The bars represent mean values, and dots indicate individual biological replicates (n = 2 for y¹w¹¹¹⁸ and n = 3 for c¹³). (B) Western blot analysis of Mod(mdg4)-67.2 protein levels in flies carrying mutations within intron 4. Protein extracts from 2- and 3-day-old adult males were probed with antibodies specific to the Mod(mdg4)-67.2 isoform. Anti-lamin antibodies were used as a loading control. The mod(mdg4)^u1 mutant, which lacks expression of Mod(mdg4)-67.2, was used as a negative control. (C) Schematic representation of the functional assay for Mod(mdg4)-67.2 activity based on the ct⁶ allele. The gypsy retrotransposon inserted between the wing-margin enhancer and the cut promoter contains Su(Hw)-binding sites that recruit the Su(Hw)/Mod(mdg4)-67.2 insulator complex, thereby blocking enhancer–promoter communication and producing the characteristic cut-wing phenotype. Loss of Mod(mdg4)-67.2 function disrupts insulator activity and restores the wild-type wing phenotype. (D) Representative wing phenotypes of ct⁶ males homozygous for the indicated mod(mdg4) alleles. Mutations affecting trans-splicing efficiency retain the characteristic cut-wing phenotype, whereas the mod(mdg4)^u1 mutation restores the wild-type wing morphology.

An alternative hypothesis to account for the viability of mod(mdg4)^c13 flies posits that normal isoform expression may occur during early developmental stages, with reduced levels appearing exclusively in adult flies. To examine this hypothesis, we quantified the expression levels of the T, K, and Z isoforms throughout the larval and pupal stages of Drosophila development (Supplementary Fig. S10). Our analysis revealed a significant decrease in trans-splicing at both developmental stages, demonstrating that the reduction in trans-splicing is not restricted to adult flies.

Our investigation into the correlation between synthesized mRNA levels and their corresponding protein products focused on the Mod(mdg4)^T isoform, also known as Mod(mdg4)-67.2. This specific isoform interacts with the architectural C2H2 zinc-finger protein Su(Hw), forming an insulator complex that blocks enhancer–promoter communication [30]. The mod(mdg4)^u1 mutation, in which the Mod(mdg4)-67.2 isoform is not expressed, was used as a negative control. We conducted a western blot analysis to quantify Mod(mdg4)-67.2 protein levels in the imago from the control lines y¹w¹¹¹⁸ and mod(mdg4)^intact, generated by restoring the deleted intronic sequences in the mod(mdg4)^attPx2 background and from the mutant lines mod(mdg4)^c44, mod(mdg4)^c73, mod(mdg4)^340–380, mod(mdg4)^368–412, mod(mdg4)^c13, and mod(mdg4)^u1.

Protein extracts prepared from 2- to 3-day-old adult males were probed with antibodies specific to Mod(mdg4)-67.2 (Fig. 7B). As expected, the Mod(mdg4)-67.2 protein was undetectable in mod(mdg4)^u1 extracts, whereas there was a clear signal in the control line, confirming antibody specificity. Despite the observed considerable reduction in mod(mdg4)^T mRNA trans-splicing efficiency, Mod(mdg4)-67.2 protein levels were comparable among the tested controls and all mutant variants. This observation indicates that the marked reduction in trans-spliced mRNA is not reflected in a corresponding decrease in protein abundance.

We employed a sensitive functional assay based on the ct⁶ mutation to gain further insight into Mod(mdg4)-67.2. The ct⁶ allele arises from the insertion of the gypsy mobile element into the 85-kb region separating the wing margin enhancer from the cut promoter [28, 64]. The Su(Hw)/Mod(mdg4)-67.2 complex binds to this gypsy insulator, consequently inhibiting wing margin enhancer activity and producing the characteristic serrated wing phenotype (Fig. 7C). When ct⁶ was combined with homozygous mod(mdg4)^u1 mutants, which are deficient in Mod(mdg4)-67.2, the wild-type wing blade phenotype was almost completely restored due to loss of enhancer blocking.

In stark contrast, ct⁶ flies carrying all evaluated mod(mdg4) mutations consistently displayed the characteristic cut wing phenotype, confirming robust enhancer blocking mediated by the gypsy insulator (Fig. 7D). Thus, Mod(mdg4)-67.2 protein abundance and function are maintained despite an approximately 10-fold reduction in functional mod(mdg4)^T mRNA.

Discussion

The mod(mdg4) locus represents a unique model for investigating the mechanisms of trans-splicing, as all mod(mdg4) mRNAs are generated through trans-splicing between a pre-mRNA encoding the common N-terminal region and one of 31 alternative exons specifying distinct C-terminal domains. In Drosophila, as in other metazoans, trans-splicing is generally suppressed due to its potential to generate aberrant chimeric mRNAs encoding deleterious proteins. Canonical cis-splicing is therefore tightly coupled to transcription and occurs predominantly co-transcriptionally, thereby ensuring accurate and efficient mRNA maturation [1, 65, 66]. Although some cis-splicing events occur post-transcriptionally and contribute to developmental and stress-responsive regulation of gene expression [67], polyadenylated transcripts that retain introns typically remain associated with chromatin until splicing is completed. In this context, the mod(mdg4) locus represents an exceptional example of extensive reprogramming of the canonical pre-mRNA processing pathway.

Our results indicate that the proximal 600 bp region of intron 4 is sufficient to redirect pre-mRNA processing toward trans-splicing independent of the promoter and upstream constitutive sequences. We found no convincing evidence supporting a model in which trans-splicing is driven by stable RNA duplex formation between the regulatory region of intron 4 and acceptor pre-mRNAs. Computational analyses identified only limited, functionally unsupported complementary interactions, while the introduction of an artificial complementary interaction failed to restore trans-splicing. These findings argue against a major role for intermolecular RNA pairing in the mod(mdg4) locus and instead support a distinct mechanism that joins 5′ and 3′ exons transcribed from separate promoters.

Our analyses revealed a high density of partially redundant RBP motifs distributed throughout the c¹¹⁷ region (296–412 bp). At the same time, efficient trans-splicing depends on currently unidentified elements located within regions a (11–108 bp) and b (109–295 bp). Region d contributes more modestly, but its functional importance becomes apparent upon disruption of motifs within region c. This extensive redundancy likely explains why deleting individual motifs or small groups of motifs has only limited effects on trans-splicing efficiency, whereas progressively removing multiple motifs produces cumulative defects. Importantly, attempts to isolate individual core motifs were insufficient to sustain efficient trans-splicing. Taken together, these observations suggest that trans-splicing depends not on a single determinant sequence but on the cooperative action of multiple cis-elements distributed across the proximal intronic region.

Using in vitro RNA affinity purification, we identified several protein groups that were specifically enriched on the native abc RNA fragment compared with the inverted control sequence. The most compelling candidates are proteins forming interaction-connected subnetworks that likely assemble on the proximal 412 bp intronic region and directly participate in trans-splicing induction. Particularly notable is the first RNA-binding complex, in which five proteins possess predicted binding sites within the c¹¹⁷ region. The Hrb98DE/TBPH heterodimer binds adjacent motifs within the conserved c¹³ element, which is critical for efficient trans-splicing. Two additional proteins in this complex, SC35 and Sf2, recognize overlapping motifs within the R2 region but not within the related R1 or R3 repeats, suggesting functional specialization among partially homologous sequence elements. In addition, Syp recognizes a motif immediately downstream of c¹³ that also contributes to trans-splicing. The c¹¹⁷ region likely contains additional recognition sites for cooperatively interacting RBPs. Some proteins enriched by RNA affinity purification but lacking characterized RNA-binding specificity may recognize motifs located within regions a, b, and c¹¹⁷.

Previous studies have demonstrated that closely spaced RBP motifs can cooperatively regulate splicing through multivalent interactions among RBPs bound to pre-mRNAs [68–71]. Biophysical studies further showed that multivalent RBP–RNA interactions can drive the formation of dynamic condensates through liquid–liquid phase separation (LLPS) [72–74]. Such condensates can increase local molecular concentrations by orders of magnitude relative to the surrounding nucleoplasm, thereby stabilizing assembly on RNAs containing multiple clustered binding motifs. We propose that a similar mechanism operates at the mod(mdg4) locus, where multiple interacting RBPs recognize densely packed intronic motifs and assemble into a specialized RNP compartment that alters canonical cis-splicing and transcriptional progression. Collective recruitment of these factors may induce currently uncharacterized modifications of both the spliceosome and the RNA pol II complex.

The identified protein complexes establish extensive connections with one another, as well as with components of the transcriptional machinery and the spliceosome (Fig. 8A). Within the first RNA-binding complex, Glo may provide a link to the transcriptional machinery, as its human homolog, HNRNPF, physically associates with the RNA pol II holoenzyme complex [75]. Glo is also connected to Fmr1, a component of the third complex [76], potentially linking transcriptional and RNA-associated pathways. The human homolog of Yps, YBX1, interacts with homologs of proteins from both the first and third complexes [77], suggesting a possible connection between these protein assemblies. The first complex is further linked to U1 snRNP through interactions involving U1-70K, SC35, Sf2, and Rsf1. Additional components may interact with the U2AF50/U2AF38 heterodimer, which is involved in 3′ splice-site recognition. Another spliceosomal component, U2A, is incorporated into the second extended complex. Together, the proteins associated with the intron and their putative connections to spliceosomal and transcriptional components suggest that a large ribonucleoprotein assembly may form on the proximal region of the nascent intron 4. The composition and spatial organization of this assembly could alter the canonical transcriptional and splicing program, thereby promoting trans-splicing.

Figure 8.

Several proposed models of protein complex assembly and trans-splicing at the mod(mdg4) locus. (A) Schematic representation of interactions between RNA pol II, spliceosomal components, and protein complexes identified on the proximal region of intron 4. Proteins belonging to the first RNA-binding subcomplex are shown in green; components of the U2-associated complex are shown in purple, and proteins associated with RNA interference-related complexes are shown in orange. Core spliceosomal factors interacting with the fourth intron, including U1 snRNP-70K and U2AF38/50, are highlighted in cyan and yellow, respectively. (B) Proposed models of trans-splicing following the assembly of RNA–protein complexes on intron 4. In the co-transcriptional model, the donor transcript remains associated with its transcription site until it interacts with acceptor pre-mRNAs. In the delayed co-transcriptional model, the donor transcript, together with associated protein complexes, relocates to the transcription site of acceptor transcripts before trans-splicing occurs. In the post-transcriptional model, donor and acceptor pre-mRNAs interact after the transcript is released from the chromatin to generate mature trans-spliced mRNAs.

At least three mechanistic scenarios can be envisioned as occurring following initiation of the trans-splicing program (Fig. 8B). In the co-transcriptional trans-splicing model, the donor transcript, together with RNA pol II and associated RNP complexes, remains near the donor transcription site after transcriptional pausing or termination until it encounters an acceptor pre-mRNA. In this scenario, specificity may be determined by protein factors associated with acceptor outrons and by spatial proximity between donor and acceptor transcription sites. Thus, donor splicing remains coupled to transcription, whereas acceptor processing occurs after the transcript is released from the chromatin. In the delayed co-transcriptional model, RNA pol II, together with the donor pre-mRNA and associated RNP complexes, relocates to promoters, driving transcription of the 3′ exon clusters. Alternatively, donor-associated RNP complexes are recruited to assembling acceptor transcription complexes. Trans-splicing would then occur during transcription of the alternative exons and would therefore resemble conventional co-transcriptional alternative splicing at the acceptor locus. Finally, in the post-transcriptional model, trans-splicing between donor and acceptor pre-mRNA occurs only after the transcript is released into the interchromosomal space of the nucleus. Further research is required to determine which of these models operates in vivo.

The extensive redundancy of RNA motifs within the proximal region of intron 4 likely contributes to the efficiency and robustness of trans-splicing. Notably, deletions reducing trans-splicing efficiency by approximately one order of magnitude did not substantially alter Mod(mdg4)-67.2 protein levels or its biological activity. These findings therefore indicate that a compensatory mechanism maintains Mod(mdg4)-67.2 protein abundance and function despite markedly reduced trans-splicing efficiency. The molecular basis of this compensation remains unknown. One possibility is that elevated translational efficiency originally evolved to compensate for inefficient ancestral trans-splicing, with subsequent evolutionary optimization of trans-splicing reducing the selective pressure for stringent translational control. Consistent with this idea, several proteins identified within the Ago2-associated complex, including Vig [78], Fmr1 [76], and GW182 [79], function as translational repressors. One hypothetical scenario is that the Ago2-containing complex simultaneously promotes trans-splicing while suppressing translation of the resulting mRNAs. If so, disruption of intronic motifs could reduce recruitment of this complex and partially compensate for reduced trans-splicing by increasing translation of mod(mdg4) mRNAs.

Our results suggest that trans-splicing may have evolved through the gradual accumulation of intronic motifs recognized by RBPs that normally participate in canonical cis-splicing. Although the mod(mdg4) locus generates all isoforms exclusively through trans-splicing, many other loci combine cis- and trans-splicing to produce alternative mRNAs. The lola locus represents a prominent example in which numerous isoforms are generated through mixed cis/trans-splicing mechanisms [80, 81]. Moreover, analysis of D. melanogaster/D. sechellia hybrids identified trans-splicing events at approximately 80 genes [81], suggesting that this mechanism is more widespread than previously thought.

In other metazoans, trans-splicing likely provides an efficient strategy for generating extensive repertoires of alternative mRNAs and protein isoforms from complex loci [25, 82]. The silkworm mod(mdg4) locus illustrates this concept, encoding >200 distinct protein isoforms [57, 83]. Because trans-splicing appears to depend primarily on combinatorial arrays of intronic RNA motifs rather than on specialized dedicated machinery, similar mechanisms may operate broadly across the animal kingdom. However, rapid evolutionary turnover of RBP-binding motifs likely generates substantial species-specific variation in trans-splicing-inducing sequences. Consistent with this idea, intronic sequences from D. virilis remain capable of inducing trans-splicing in D. melanogaster, whereas the corresponding silkworm sequences do not [57].

Trans-splicing substantially expands the coding and regulatory capacity of complex loci by enabling combinatorial assembly of alternative mRNAs from independently transcribed modules. Because this mechanism can coexist with canonical cis-splicing and appears to rely primarily on distributed intronic regulatory motifs, trans-splicing may be considerably more prevalent than currently believed. Defining its full biological significance, however, will require elucidation of the molecular architecture and dynamics of the RNP assemblies that initiate and coordinate trans-splicing.

A limitation of this study is the lack of direct experimental validation of the role of the identified RBPs in trans-splicing. Given the high redundancy among motifs recognized by different RBPs, inactivating a single RBP is unlikely to disrupt trans-splicing. Therefore, dissecting of the individual contributions of these proteins will probably require either combined inactivation of multiple RBPs or the use of a sensitized genetic background, both of which represent a promising avenue for future research.

Supplementary data

Supplementary data is available at NAR online.

Conflict of interest

None declared.

Funding

This work was supported by grant 23-44-00038 (to P.G.) from the Russian Science Foundation and NSFC 32261133522 (to Y.-Z.X.) from the National Natural Science Foundation of China.

Data availability

NGS data have been deposited at the NCBI Gene Expression Omnibus (GEO) under accession number GSE302152. The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE partner repository with the dataset identifier PXD079014 and 10.6019/PXD079014.