Faithful mRNA splicing depends on the Prp19 complex subunit faint sausage and is required for tracheal branching morphogenesis in Drosophila

mRNA splicing is required to process nearly all eukaryotic transcripts. As splicing can be rate-limiting for efficient gene expression (Guilgur et al., 2014; Hoyle and Ish-Horowicz, 2013), reduced splicing efficiency can perturb many cellular processes. Although defects in RNA processing have been associated with pleiotropic phenotypes during development (Golling et al., 2002), inactivation of splicing factors can cause surprisingly specific cellular defects (Chen et al., 1998; van der Lelij et al., 2014). Mutations in core components of the spliceosome are associated with human diseases, including retinitis pigmentosa and spinal muscular atrophy (Faustino and Cooper, 2003), and can contribute to carcinogenesis (Quesada et al., 2011). However, the basis for these specific phenotypes is not clear.

Morphogenesis requires dynamically regulated gene expression programs to coordinate cell movements and differentiation. Tracheal branching morphogenesis in Drosophila is guided by the dynamic expression of the fibroblast growth factor (FGF) branchless (Bnl; Sutherland et al., 1996) in cells surrounding the tracheal primordia. Bnl activates the FGF receptor (FGFR) breathless (Btl) on tracheal cells, which move as a cohort towards the Bnl source. Here, we show that mutations in the faint sausage (fas) gene (Nüsslein-Volhard et al., 1984) specifically affect branch outgrowth and cellular rearrangements during tracheal morphogenesis. We found that fas encodes a subunit of the spliceosome-activating Prp19 complex (Prp19C; also known as NineTeen complex, NTC; Chanarat and Sträßer, 2013), contrary to an earlier report that fas encodes a secreted immunoglobulin (Ig) domain protein (Lekven et al., 1998). Lack of zygotic fas/Prp19C function broadly impairs the efficiency of mRNA splicing and leads to extensive changes in global gene expression. Our findings suggest that tracheal branching morphogenesis is particularly sensitive to efficient spliceosome function, thus providing an entry point to investigate the requirements of splicing during organogenesis.

We isolated two allelic embryonic lethal mutants (P218 and H124) that were defective in primary tracheal branching (Fig. 1). While wild-type embryos have formed an interconnected tracheal network by 12 h after egg laying (AEL), tracheal branch outgrowth was impaired in P218 and H124 homozygous (Fig. 1A,B; Movie 1) and in P218/H124 trans-heterozygous embryos. Primary branching was severely reduced, but not entirely abolished as in btl mutants, which lack the FGF receptor (Fig. 1C-E). P218 and H124 tracheal primordia remained as elongated sacs with dorsal extensions that frequently detached from the remaining primordia (Fig. 1B; Movie 2). Adjacent metameres often formed a partially interconnected dorsal trunk (DT; Fig. 1B) that appeared to become stretched during tube elongation, especially at sites where relatively few cells were present (Fig. 1B; Movie 3). Persisting DT connections revealed deposition of luminal material (Fig. 1F,G). Although tracheal development was severely affected, earlier processes, including gastrulation and germband retraction (Fig. 1A,B), appeared normal. However, the mutant embryos showed defects during later embryogenesis, including abnormal dorsal closure and head involution (not shown; Liu et al., 1999).

We mapped the lethality of H124 and P218 to a small interval (50B4-B6) comprising six annotated genes (Fig. 2A; supplementary Materials and Methods). A lethal P-element insertion (P{PZ}05488) in this interval failed to complement H124 and P218 mutants. P{PZ}05488 is an allele of the fas (Nüsslein-Volhard et al., 1984) locus, which was reported to encode a secreted immunoglobulin domain protein (CG17761) expressed in the central nervous system (CNS) and in clusters of epidermal cells (Lekven et al., 1998). Furthermore, the ethyl methanesulfonate (EMS)-induced fas¹ mutation (Nüsslein-Volhard et al., 1984) failed to complement P218 and H124, indicating that the two mutations are allelic to fas. However, expression of a UAS-CG17761 transgene corresponding to the reported fas locus (Lekven et al., 1998) in the ectoderm of fas embryos did not lead to noticeable rescue of the tracheal defects (data not shown).

P{PZ}05488 is inserted upstream of the fandango (fand, CG6197) gene, which encodes a subunit of the NineTeen Complex/Prp19 complex (NTC/Prp19C; Fig. 2A; Guilgur et al., 2014) that is essential for spliceosome activation (Chanarat and Sträßer, 2013). Surprisingly, two fand alleles, fand¹ and fand² (Guilgur et al., 2014), failed to complement P218 and H124 mutants. Consistent with these findings, sequence analysis revealed premature stop codons in the CG6197/fand gene in H124 and P218 mutants (Fig. 2A). Moreover, we found that the fas¹ allele (Nüsslein-Volhard et al., 1984) carries a premature stop codon in CG6197 within the same codon as P218 (Fig. 2A). Conversely, we did not find any missense or nonsense mutations in the CG17716 coding sequence in H124, P218 and fas¹ mutants. Finally, the tracheal and CNS defects of fas^H124/fas¹ embryos were completely rescued by a genomic construct containing a wild-type copy of CG6197 (Fig. 2C-F; Guilgur et al., 2014). Together, these findings indicate that the fas phenotype is caused by mutations in the NTC/Prp19C subunit CG6197 also known as Fand (Guilgur et al., 2014), and not, as previously reported, in the immunoglobulin domain protein CG17716 (Lekven et al., 1998). We therefore refer to the CG6197 gene as fas from here onwards.

To investigate whether fas function is required in tracheal cells for their normal migration, we expressed a Myc-tagged Fas (Fas-Myc) construct in tracheal cells of fas mutants. However, tracheal-specific expression of Fas-Myc was not sufficient to rescue tracheal morphogenesis (Fig. 2G,G′). Conversely, tracheal development was rescued when Fas-Myc was expressed in all cells within an entire body segment (Fig. 2H,H′). Furthermore, expression of Fas-Myc throughout the embryo except for tracheal cells (using ubiquitous da-Gal combined with btl-Gal80 to inhibit Gal4 specifically in tracheal cells) led to nearly complete rescue of tracheal development in fas embryos (Fig. 2I-J′). Although we cannot exclude that btl-Gal80-mediated inhibition of da-Gal4-driven Fas-Myc expression was incomplete, these findings suggest that tracheal cell migration depends on fas function in surrounding tissues.

Maternally provided Fas protein is essential for efficient mRNA splicing in the early embryo (Guilgur et al., 2014; Martinho et al., 2015), but zygotic functions of fas have not been addressed thus far. We found that Fas protein levels were strongly reduced in fas^H124 and fas^P218 late-stage embryos (15-18 h AEL), although residual amounts of Fas protein, presumably representing maternal Fas protein, were still detectable by immunoblot (Fig. 2B).

To systematically identify changes in mRNAs, which may account for the tracheal defects in fas embryos, we performed high-throughput transcriptome sequencing (RNAseq). Total RNA was extracted from fas^P218 and control embryos at 12-13 h AEL, when tracheal development showed first signs of perturbation in fas mutants. Most strikingly, we observed an accumulation of transcripts with retained intronic sequences in fas mutants (Fig. 3A). Out of 47,590 introns present in 11,059 genes, the relative inclusion rate of 6453 introns (13.6%) was increased significantly (FDR<0.01) and at least 2.83-fold in fas^P218 embryos (Fig. 3B; log2 FC≥1.5; Table S1). These retained introns were present in 3629 genes, suggesting that lack of Fas does not uniformly affect splicing of all introns, but that a subset of transcripts or introns is more susceptible to escape splicing in the absence of Fas. Similarly, among the set of retained introns, some were more frequently retained than others. This was observed also for different introns within the same transcript (e.g. CG17716; Fig. 3B,C). For example, of the 6453 introns with at least 2.83-fold increased retention (FDR<0.01, log2 FC≥1.5), 2968 were increased fourfold or more (log2 FC≥2) and 896 were increased at least 16-fold (log2 FC≥4), suggesting that specific introns may be especially prone to splicing defects when Prp19C function becomes limiting. Although introns with high retention rates show some association with intron length, GC content, number of introns per gene and transcript expression level, these weak associations could be partially explained by higher statistical power when counts are higher (Fig. S1).

Out of 326 genes with annotated functions during tracheal development (GO:0007424; FlyBase), 214 genes were considered as expressed (see supplementary Materials and Methods). Ninety percent (193 genes) of these expressed tracheal genes revealed abnormal transcript processing (gene-level FDR<0.01; Fig. 3D). Most of the remaining normally processed transcripts were derived either from intron-less genes or from genes with only a few short introns (Fig. 3D). The tracheal cell migration defects in fas embryos could in principle be explained by abnormal splicing of transcripts encoding components of FGF and EGF signalling (Fig. 3D). However, transcripts with highest intronic fold-changes in fas mutants include several receptor tyrosine kinase (RTK) signalling components acting either upstream (Bnl, Sfl) or downstream (Stumps, Sos) of the RTKs (Fig. 3E), suggesting that the tracheal defects in fas embryos are unlikely to be attributable to changes in the expression of a specific gene.

In addition to the splicing defects, many genes showed dramatic changes in transcript levels (Table S2). Of 9606 expressed genes, 11.5% were downregulated at least 0.35-fold (log2 FC≤−1.5; FDR <0.01) and 6.7% were upregulated at least 2.83-fold (log2 FC≥1.5; FDR<0.01). We conclude that lack of zygotic Fas protein severely affects the efficiency of RNA splicing in the developing embryo, resulting in qualitative (intron retention) as well as quantitative (transcript abundance) changes in gene expression.

Analysis of all mis-spliced tracheal transcripts in fas mutants revealed abnormal transcript processing and changes in expression levels of several components of EGF and FGF signalling (Fig. S2A,B). To validate the functional consequences of the splicing defects, we analysed EGF and FGF signalling activity during tracheal morphogenesis. The branch identity gene spalt (sal; Kuhnlein and Schuh, 1996) is expressed in tracheal placodes upon activation of Wnt and EGF signalling (Chihara and Hayashi, 2000). Sal signals in tracheal cells of fas embryos were strongly reduced (Fig. 4A,B). Similarly, FGF-dependent ERK activation (dpERK) at tracheal branch tips was severely diminished (Fig. 4C,D), and FGF-induced specification of DSRF-positive tracheal terminal cells was completely abolished (Fig. 4E,F). However, the residual tracheal branching in fas mutants (Fig. 1E) suggests that FGF and EGF signalling pathways are partially active, but that their output is quantitatively reduced. We therefore asked whether the output of FGF signalling can be restored in fas mutants through expression of a constitutively active FGF receptor (λBtl; Lee et al., 1996). Indeed, we observed high ERK activation and DSRF-positive terminal cells upon constitutive activation of Btl FGFR signalling in tracheal cells of fas embryos (Fig. 4G-I).

The requirement of fas outside tracheal cells suggested that events upstream of FGFR, possibly including the production of Bnl FGF, may be compromised in the absence of fas function. We therefore tested whether misexpression of Bnl FGF was able to restore tracheal branching in fas mutants. Misexpression of a bnl cDNA in the epidermis caused ectopic tracheal branching in wild-type controls, but not in fas embryos (Fig. 4J,K). Together, these findings suggest that while zygotic Fas function is not strictly required for signalling downstream of FGFR, it is essential for the activation of signalling by Bnl FGF, although the tracheal branching defects in fas mutants are not solely attributable to mis-splicing of bnl transcripts.

We describe the requirement of the NTC/Prp19C subunit Fas for tracheal branching morphogenesis. Although mRNA splicing is generally required for transcript maturation, we found that embryos lacking zygotic fas function display surprisingly specific organogenesis defects. First, we show that the NTC/Prp19C subunit CG6197 is required for tracheal branching. We demonstrate that the fas locus (Nüsslein-Volhard et al., 1984), previously reported to encode the secreted immunoglobulin domain protein CG17716 (Lekven et al., 1998), in fact encodes the NTC/Prp19C subunit CG6197 also known as fand (Guilgur et al., 2014). Second, we show that loss of zygotic fas function results in widespread perturbation of splicing, consistent with previous work demonstrating an essential requirement of maternal CG6197/fand function for efficient splicing during early embryogenesis (Guilgur et al., 2014; Martinho et al., 2015). Abnormal transcript processing in fas mutants manifests predominantly in intron retention, accompanied by changes in transcript abundance. Third, we show that compromised FGF and EGF signalling may contribute to the tracheal branching defects in fas mutants. The requirement of Fas in non-tracheal cells (Fig. 2G-J) and the results of our epistatic analysis (Fig. 4G-K) suggest that fas function is essential for the activation of the Btl FGFR.

The late onset and the specific nature of tracheal and CNS (Lekven et al., 1998; Liu et al., 1999) defects in zygotic fas embryos was surprising, given that pre-mRNA processing is generally perturbed in the mutants. However, maternally provided gene products, including Fas protein itself, are likely to allow for largely normal development during early embryogenesis. As maternal Fas protein decays over time, Fas levels appear to become limiting at the onset of tracheal morphogenesis during mid-embryogenesis. Perturbed pre-mRNA processing in fas mutants is accompanied by substantial changes in the levels of many transcripts. Nonsense-mediated mRNA decay (Wilusz et al., 2001) is expected to degrade a large fraction of mis-spliced transcripts. In addition, indirect effects on transcriptional regulation are likely to influence gene expression in fas mutants. Of note, the abnormally processed transcripts in fas embryos include CG17716 mRNA (Fig. 3C), consistent with the finding of Lekven et al. (1998) that CG17716 protein was undetectable in fas embryos.

We report an example of strikingly specific developmental defects associated with a lack of efficient splicing. How can a general perturbation of splicing lead to specific phenotypes? First, tissue-specific expression of some splicing factors, including Prp19 (Urano et al., 2006), might account for tissue-dependent differences in splicing efficiency. Second, pre-mRNAs contain diverse auxiliary cis-acting regulatory elements that are recognized by a multitude of splicing factors (Zhang et al., 2008). Consequently, different introns may show distinct sensitivities towards the lack of a given splicing factor. In addition, intron number, transcript abundance and transcript stability may render some RNAs more prone to accumulating splicing errors than others. Finally, dynamic cellular processes, such as cellularization (Guilgur et al., 2014) and tracheal branching, which involve rapid regulation of gene expression, may depend more acutely on efficient mRNA processing. Consistent with this idea, highly expressed and rapidly regulated genes tend to have only few and short introns (Castillo-Davis et al., 2002; Jeffares et al., 2008).

Our findings allow us to define the most sensitive gene expression events required for proper organogenesis. Characterizing NTC/Prp19C function and regulation in different organs could therefore contribute to a better understanding of how differential gene expression is regulated during organogenesis.

Unless noted otherwise, Drosophila stocks are described in FlyBase and were obtained from the Bloomington Stock center: btl⁷²⁴ (Ghabrial and Krasnow, 2006); fas¹ (Nüsslein-Volhard et al., 1984); CG6197⁺ (fas⁺, genomic fas transgene); UAS-CG6197-myc (UAS-Fas-myc), fand¹, fand² (Guilgur et al., 2014); UAS-bnl (Sutherland et al., 1996); UAS-λBtl (Lee et al., 1996); UAS-palm-mNeonGreen (this study); UAS-CG17716 (this study); btl-Gal4 (Shiga et al., 1996); Abd-B-Gal4 (de Navas et al., 2006); btl-Gal80 (Nikolova and Metzstein, 2015); da-Gal4; and 69B-Gal4. fas^H124 and fas^P218 were isolated in an EMS mutagenesis screen (Förster et al., 2010; see supplementary Materials and Methods).

Protein extracts from embryos (15-18 h AEL) were analysed on immunoblots. Antibodies were rabbit anti-Fas (1:1000; Guilgur et al., 2014), mouse anti-α-tubulin Dm1A (1:10,000; Sigma), goat anti-rabbit Superclonal HRP conjugate (1:5000; Thermo Fischer) and goat anti-mouse Superclonal HRP conjugate (1:5000; Thermo Fischer). Three independent replicates (embryo collections) were analysed.

A DNA fragment corresponding to CG17716-PA cDNA (FlyBase) was synthesized (GenScript), cloned into pUAST-attB (EcoRI/HindIII) and integrated into the attP2 (68A4) landing site using PhiC31 integrase (Bischof et al., 2007). The coding sequence of monomeric yellow-green fluorescent protein (mNeonGreen; Shaner et al., 2013) fused to an N-terminal palmitoylation signal was synthesized (GenScript) using the D. melanogaster codon distribution, cloned into pUASt-attB (EcoRI/XbaI) and integrated into the attP2 landing site. See supplementary Materials and Methods and Table S3 for sequencing of fas alleles.

Embryos were fixed in 4% formaldehyde for 20 min and devitellinized in methanol/heptane. Primary antibodies were chicken anti-GFP (1:500; Abcam #13970), mouse anti-DSRF (1:300; Samakovlis et al., 1996), rabbit anti-Sal (1:40; Kuhnlein and Schuh, 1996), mouse anti-Tango (1:100; DSHB), mouse anti-myc (9E10; 1:300; DSHB) and rabbit anti-dpERK (1:100; Cell Signaling Technology #4370). Goat secondary antibodies were conjugated with DyLight 488 (1:500; Abcam), Alexa Fluor 568 (1:300; Molecular Probes) or Cy5 (1:500; Jackson ImmunoResearch). Chitin was detected as previously described (Caviglia and Luschnig, 2013). Neurons were labelled using anti-HRP (1:1000; Dianova) conjugated with Alexa Fluor 647.

Imaging was performed on an Olympus FV1000 confocal microscope with 20×/0.75 NA, 40×/1.3 NA and 60×/1.35 NA objectives or on a Zeiss LSM710 with a 20×/1.0 NA objective. For live imaging, staged embryos were dechorionated, glued on a coverslip and immersed in Voltalef 10S oil. Images were processed using ImageJ (v2.0.0), Imaris (v8.3.1; Bitplane) and Adobe Photoshop. At least three embryos per genotype were analysed. See supplementary Materials and Methods for statistics and reproducibility.

Total RNA was isolated from 12-13 h AEL fas^P218 btl-Gal4 UAS-GFP UAS-Verm-mRFP embryos and from control embryos (carrying the parental chromosome) using TRIzol (Thermo Fischer). RNA was precipitated in isopropanol with 0.3 M sodium acetate and treated with DNase I (Ambion) for 25 min at 37°C. cDNA libraries were generated using the Illumina RNASeq protocol and were sequenced with an Illumina HiSeq instrument. RNASeq experiments in three biological replicates (independent embryo collections) of control and fas^P218 embryos yielded 27 to 69 million paired-end reads per sample. Details on RNA-seq data analyses are provided in the supplementary Materials and Methods.

We thank Rui Gonçalo Martinho, the Developmental Studies Hybridoma Bank and the Bloomington Drosophila Stock Center for providing fly stocks and antibodies. The Functional Genomics Center Zurich (FGCZ) provided support with RNA sequencing experiments. We thank Dirk Beuchle for help with genetic mapping, Simone Mumbauer for generating UAS-CG17716 flies, Wilko Backer for technical support and Michaela Clever for comments on the manuscript. We are indebted to Christian Lehner for continuous support and discussions. We are grateful to Sofia Araujo and Rui Gonçalo Martinho for communication of unpublished work.