Design of RNA splicing analysis null models for post hoc filtering of Drosophila head RNA-Seq data with the splicing analysis kit (Spanki)

BACKGROUND

The production of multiple transcript isoforms from one gene is a major source of transcriptome complexity. RNA-Seq experiments, in which transcripts are converted to cDNA and sequenced, allow the resolution and quantification of alternative transcript isoforms. However, methods to analyze splicing are underdeveloped and errors resulting in incorrect splicing calls occur in every experiment.

RESULTS

We used RNA-Seq data to develop sequencing and aligner error models. By applying these error models to known input from simulations, we found that errors result from false alignment to minor splice motifs and antisense stands, shifted junction positions, paralog joining, and repeat induced gaps. By using a series of quantitative and qualitative filters, we eliminated diagnosed errors in the simulation, and applied this to RNA-Seq data from Drosophila melanogaster heads. We used high-confidence junction detections to specifically interrogate local splicing differences between transcripts. This method out-performed commonly used RNA-seq methods to identify known alternative splicing events in the Drosophila sex determination pathway. We describe a flexible software package to perform these tasks called Splicing Analysis Kit (Spanki), available at http://www.cbcb.umd.edu/software/spanki.

CONCLUSIONS

Splice-junction centric analysis of RNA-Seq data provides advantages in specificity for detection of alternative splicing. Our software provides tools to better understand error profiles in RNA-Seq data and improve inference from this new technology. The splice-junction centric approach that this software enables will provide more accurate estimates of differentially regulated splicing than current tools.

Rapid functional diversification in the structurally conserved ELAV family of neuronal RNA binding proteins

BACKGROUND

The Drosophila gene embryonic lethal abnormal visual system (elav) is the prototype of a gene family present in all metazoans. Its members encode structurally conserved neuronal proteins with three RNA Recognition Motifs (RRM) but they paradoxically act at diverse levels of post-transcriptional regulation. In an attempt to understand the history of this family, we searched for orthologs in eleven completely sequenced genomes, including those of humans, D. melanogaster and C. elegans, for which cDNAs are available.

RESULTS

We analyzed 23 orthologs/paralogs of elav, and found evidence of gain/loss of gene copy number. For one set of genes, including elav itself, the coding sequences are free of introns and their products most resemble ELAV. The remaining genes show remarkable conservation of their exon organization, and their products most resemble FNE and RBP9, proteins encoded by the two elav paralogs of Drosophila. Remarkably, three of the conserved exon junctions are both close to structural elements, involved respectively in protein-RNA interactions and in the regulation of sub-cellular localization, and in the vicinity of diverse sequence variations.

CONCLUSION

The data indicate that the essential elav gene of Drosophila is newly emerged, restricted to dipterans and of retrotransposed origin. We propose that the conserved exon junctions constitute potential sites for sequence/function modifications, and that RRM binding proteins, whose function relies upon plastic RNA-protein interactions, may have played an important role in brain evolution.

Shared RNA-binding sites for interacting members of the Drosophila ELAV family of neuronal proteins

The product of the Drosophila embryonic lethal abnormal visual system is a conserved protein (ELAV) necessary for normal neuronal differentiation and maintenance. It possesses three RNA-binding domains and is involved in the regulation of RNA metabolism. The long elav 3'-untranslated region (3'-UTR) is necessary for autoregulation. We used RNA-binding assays and in vitro selection to identify the ELAV best binding site in the elav 3'-UTR. This site resembles ELAV-binding sites identified previously in heterologous targets, both for its nucleotide sequence and its significant affinity for ELAV (K(d) 40 nM). This finding supports our model that elav autoregulation depends upon direct interaction between ELAV and elav RNA. We narrowed down the best binding site to a 20 nt long sequence A(U5)A(U3)G(U2)A(U6) in an alternative 3' exon. We propose and test a model in which the regulated use of this alternative 3' exon is involved in normal elav regulation. Found in NEurons (FNE), another neuronal RNA-binding protein paralogous to ELAV, also binds this site. These observations provide a molecular basis for the in vivo interactions reported previously between elav and fne.

found in neurons, a third member of the Drosophila elav gene family, encodes a neuronal protein and interacts with elav

elav, a gene necessary for neuronal differentiation and maintenance in Drosophila, encodes the prototype of a family of conserved proteins involved in post-transcriptional regulation. We identified found in neurons (fne), a gene encoding a new ELAV paralogue. We showed that FNE binds RNA in vitro. fne transcripts are present throughout development and contain long untranslated regions. Transcripts and proteins are restricted to neurons of the CNS and PNS during embryogenesis. These features are reminiscent of elav. However, fne expression is delayed compared to elav's, and FNE protein appears cytoplasmic, while ELAV is nuclear. GAL4-directed overexpression of fne in neurons leads to a reduction of stable transcripts produced from both the fne and elav endogenous loci, suggesting that fne autoregulates and also regulates elav.

Drosophila arginase is produced from a nonvital gene that contains the elav locus within its third intron

A Drosophila gene encoding a 351-amino acid-long predicted arginase (40% identity with vertebrate arginases) is reported. Interestingly, the third intron of the arginase gene includes the elav locus, whose coding sequence is on the complementary DNA strand to that of the arginase. Terrestrial vertebrates produce two arginases from duplicated genes. One form, essentially present in the liver, is a key enzyme of the urea cycle and eliminates excess ammonia through the excretion of urea. The function of the extrahepatic arginase, more ubiquitous, is not well understood. In macrophages, arginase competes with nitric-oxide synthase, which converts arginine into nitric oxide. Most organisms, including insects, produce only one type of arginase, whose function is not centered on ammonia detoxification. A Drosophila cDNA encoding a predicted arginase was isolated. It produces a 1.3-kilobase transcript present with highest levels toward the end of embryogenesis and thereafter. During embryogenesis, the arginase transcripts localize to the fat body. The first mutant allele of the Drosophila arginase gene was identified. It is predicted to produce a 199-amino acid-long C-terminally truncated protein, likely to be inactive. Preliminary characterization of the mutation shows that this recessive allele causes a developmental delay but does not affect viability.

Evidence for 3' untranslated region-dependent autoregulation of the Drosophila gene encoding the neuronal nuclear RNA-binding protein ELAV

The Drosophila locus embryonic lethal abnormal visual system (elav) encodes a nuclear RNA-binding protein essential for normal neuronal differentiation and maintenance of neurons. ELAV is thought to play its role by binding to RNAs produced by other genes necessary for neuronal differentiation and consequently to affect their metabolism by an as yet unknown mechanism. ELAV structural homologues have been identified in a wide range of organisms, including humans, indicating an important conserved role for the protein. Analysis of elav germline transformants presented here shows that one copy of elav minigenes lacking a complete 3' untranslated region (3' UTR) rescues null mutations at elav, but that two copies are lethal. Additional in vivo experiments demonstrate that elav expression is regulated through the 3' UTR of the gene and indicate that this level of regulation is dependent upon ELAV itself. Because ELAV is an RNA-binding protein, the simplest model to account for these findings is that ELAV binds to the 3' UTR of its own RNA to autoregulate its expression. I discuss the implications of these results for normal elav function.

Structural evolution of the Drosophila 5S ribosomal genes

We compare the 5S gene structure from nine Drosophila species. New sequence data (5S genes of D. melanogaster, D. mauritiana, D. sechellia, D. yakuba, D. erecta, D. orena, and D. takahashii) and already-published data (5S genes of D. melanogaster, D. simulans, and D. teissieri) are used in these comparisons. We show that four regions within the Drosophila 5S genes display distinct rates of evolution: the coding region (120 bp), the 5'-flanking region (54-55 bp), the 3'-flanking region (21-22 bp), and the internal spacer (149-206 bp). Intra- and interspecific heterogeneity is due mainly to insertions and deletions of 6-17-bp oligomers. These small rearrangements could be generated by fork slippages during replication and could produce rapid sequence divergence in a limited number of steps.

Two distinct temperature-sensitive alleles at the elav locus of Drosophila are suppressed nonsense mutations of the same tryptophan codon

The Drosophila gene elav encodes a 483-amino-acid-long nuclear RNA binding protein required for normal neuronal differentiation and maintenance. We molecularly analyzed the three known viable alleles of the gene, namely elavts1, elavFliJ1, and elavFliJ2, which manifest temperature-sensitive phenotypes. The modification of the elavFliJ1 allele corresponds to the change of glycine426 (GGA) into a glutamic acid (GAA). Surprisingly, elavts1 and elavFliJ2 were both found to have tryptophan419 (TGG) changed into two different stop codons, TAG and TGA, respectively. Unexpectedly, protein analysis from elavts1 and elavFliJ2 reveals not only the predicted 45-kD truncated ELAV protein due to translational truncation, but also a predominant full-size 50-kD ELAV protein, both at permissive and nonpermissive temperatures. The full-length protein present in elavts1 and elavFliJ2 can a priori be explained by one of several mechanisms leading to functional suppression of the nonsense mutation or by detection of a previously unrecognized ELAV isoform of similar size resulting from alternative splicing and unaffected by the stop codon. Experiments described in this article support the functional suppression of the nonsense mutation as the mechanism responsible for the full-length protein.

Gene elav of Drosophila melanogaster: a prototype for neuronal-specific RNA binding protein gene family that is conserved in flies and humans

Regulated gene activity is crucial to the formation and function of the nervous system. It is well known that gene regulation can occur at the transcriptional, post-transcriptional, translational, and post-translational levels. In this review our focus has been on the post-transcriptional regulation in neurons and on neural-specific RNA binding proteins that may be involved in post-transcriptional modulation of gene activity. We have taken advantage of this opportunity to review our work on the elav gene of Drosophila melanogaster which encodes a neural-specific RNA binding protein and relate it to other members of this elav-like gene family. We report new data that suggests that elav is post-transcriptionally regulated and we demonstrate that below-threshold levels of ELAV protein severely affects neuronal differentiation.

Gene activation and DNA binding by Drosophila Ubx and abd-A proteins

The Ubx and abd-A gene products are required for proper development of thoracic and abdominal structures in Drosophila. We expressed LexA-Ubx and LexA-abdA fusion proteins in yeast. These proteins activated expression of target genes that carried either upstream LexA operators or upstream Ubx binding sites. Both proteins contain homeodomains. Experiments with mutant fusion proteins show that the homeodomain is not required for the proteins to form dimers or enter the nucleus, and that, when DNA binding is provided by the LexA moiety, the homeodomain is not required for gene activation. Our results suggest that the homeodomain is necessary for these proteins to bind Ubx sites, but that the homeodomain does not contact DNA exactly like bacterial helix-turn-helix proteins. Finally, our data suggest that gene activation by these proteins is a simple consequence of their binding to DNA, while negative gene regulation requires that these proteins act together with other Drosophila gene products.

An approach to study the evolution of the Drosophila 5S ribosomal genes using P-element transformation

The P-element-mediated gene transfer system was used to introduce Drosophila teissieri 5S genes into the Drosophila melanogaster genome. Eight transformed D. melanogaster strains that carry D. teissieri 5S mini-clusters consisting of 9-21 adjacent 5S units were characterized. No genetic exchanges between D. melanogaster and D. teissieri 5S clusters were detected over a 2-year survey of the eight strains. The occurrence of small rearrangements within the D. melanogaster 5S cluster was demonstrated in one of the transformed strains.

Bipartite structure of the 5S ribosomal gene family in a Drosophila melanogaster strain, and its evolutionary implications

Knowledge of multigenic family organization should provide insight into their mode of evolution. Accordingly, we characterized the 5S ribosomal gene family in the Drosophila melanogaster strain ry506. The 5S genes in this strain display a striking HindIII restriction difference compared to the "standard" D. melanogaster 5S genes. The sequence of three ry506 5S genes was determined. We show that the HindIII restriction site heterogeneity within the ry506 5S family most probably results from the same point mutation, suggesting that a single 5S variant was propagated into the 5S cluster of this strain. Furthermore, we demonstrate that the structural organization of the 5S genes in ry506 is a bipartite structure, i.e., that about 40% of the 5S genes constitute a HindIII+/HindIII- mixed cluster, while those remaining constitute an homogeneous HindIII- cluster. The events which might lead to such an heterogeneous pattern are discussed from an evolutionary point of view.

The 5S ribosomal genes in the Drosophila melanogaster species subgroup. Nucleotide sequence of a 5S unit from Drosophila simulans and Drosophila teissieri

The 5S genes of the eight species of the D. melanogaster subgroup have been mapped. The spacers, in contrast with coding regions, differ markedly between most species. One 5S gene unit has been sequenced for both D. simulans and D. teissieri. The mature 5S RNA region in these two species is identical to the corresponding region of D. melanogaster. Only 5 nucleotide variations occur between the D. melanogaster and D. simulans 5S gene spacers. The spacer in D. teissieri is very different. Only two segments, located one at each side of the coding region, are clearly homologous to corresponding sequences of D. melanogaster and D. simulans.