{"paper_id":"8c642336-dcce-4f0b-ab67-c33e9912bbf7","body_text":"1 \nELAV mediates circular RNA biogenesis in neurons  \nCarlos Alfonso-Gonzalez1,2, Sarah Holec1, Sakshi Gorey1,2, Mengjin Shi1,2, Michael Rauer1,4,  \nJudit Carrasco1,5, Stylianos Tsagkris1,6, and Valérie Hilgers1,3,7,* \n  \n1 Max Planck Institute of Immunobiology and Epigenetics, Freiburg, Germany. \n2 Faculty of Biology, University of Freiburg, Freiburg, Germany. \n3 Signalling Research Centre CIBSS, University of Freiburg, Schänzlestraße 18, 79104 Freiburg, Germany \n4 Present address: Institut für Medizinische Bioinformatik und Systemmedizin, University of Freiburg, Freiburg, \nGermany \n5 Present address: Discovery Sciences, BioPharmaceuticals R&D, AstraZeneca, CB2 0AA, Cambridge, United \nKingdom \n6 Present address: Epigenetics & Neurobiology Unit, European Molecular Biology Laboratory (EMBL) -Rome, \nAdriano Buzzati-Traverso Campus, Rome, Italy  \n7 Lead contact \n*Correspondence: hilgers@ie-freiburg.mpg.de. \nSummary \nCircular RNAs (circRNAs) arise from back-splicing of precursor RNAs and accumulate in the nervous system of \nanimals, where they are thought to regulate gene expression and synaptic function. Here, we show that neuronal \ncircRNA biosynthesis is mediated by the pan-neuronal RNA-binding protein ELAV. In Drosophila embryos, we \ncharacterize the circRNA landscape in normal and elav mutant neurons. We find that neu ronal circRNAs are \nglobally (>75%) depleted upon ELAV knockout, and induction of ELAV expression drives ectopic RNA \ncircularization. In brain tissue, ELAV binds to pre -mRNA introns flanking putative circRNAs and decreases \nefficiency of linear splicing in favor of intron pairing at reverse complementary matches, inducing circularization. \nTogether, our data demonstrate that ELAV directly modulates splicing decisions to generate the neuronal \ncircRNA landscape. \nKeywords \ncircular RNAs, nervous system, RNA -binding proteins, ELAV, Drosophila, splicing, back -splicing, reverse \ncomplementary matches (RCMs), RNA processing, RNA binding motif. \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted June 22, 2024. ; https://doi.org/10.1101/2024.06.22.600180doi: bioRxiv preprint \n\n 2 \nIntroduction \nCircular RNAs (circRNAs) are single-stranded RNA molecules covalently bound into a continuous loop, found \nubiquitously across all domains of life 1. In animals, circRNAs are expressed in tissue- and developmental stage-\nspecific patterns in animals, with a particularly high abundance in the nervous system 2-5. CircRNAs typically \narise from genes linked to synaptic fun ctions. Although their physiological functions remain much less well \nstudied compared to the linear version of the gene, circRNAs have been shown to play important roles in various \naspects of brain development and functionality, including neural stemness and neurodegeneration during aging, \ncognition, and plasticity 6-11. For instance, the loss of Cdr1as circRNA disrupts sensorimotor gating, a phenotype \nassociated with several human neuropsychiatric disorders 12,13. Moreover, circRNAs expression is altered i n \nmultiple neuropathological conditions such as addiction and neurodegeneration, making them potential markers \nfor these diseases 14,15. Mechanisms of action differ for distinct circRNAs 1 ; they may act as regulators of \nindividual genes 16, function globally as sponges for microRNAs and RNA -binding proteins (RBPs) 17,18, while \nothers are translated into a functional protein 19,20. \nCircRNA formation is a co -transcriptional process in which a 5'  splice site is ligated “back” to a 3' splice site \nlocated upstream, thereby forming a circular structure with a characteristic 3' -5' phosphodiester bond at the \nback-splicing junction site (BSJ) 21,22. Intron pairing is facilitated by co -transcriptional secondary structures of \nflanking introns and promotes circRNA formation by bringing splice sites of distal exons into proximity of each \nother 23,24. Both back -splicing and linear splicing use fundamental RNA processing components such as the \nspliceosome machinery and canonical splice sites 18,25, which suggests that the two processes occur in a \ncompetitive manner. Specific RBPs have been shown to influence circRNA expression in developmental \ntransitions and in various cellular contexts. During epithelial-mesenchymal transition, the splicing factor Quaking \nregulates the formation of over one-third of abundant circRNAs through intron binding 26; in motor neurons, the \nRBP FUS interacts with the pre-mRNA to control the biogenesis of specific circRNAs 27, and the RNA processing \nfactor NOVA2 promotes the biogenesis of numerous circRNAs in mouse cortical neurons 28. The mechanisms \ngoverning circRNA regulation in vivo are yet to be fully elucidated; in particular, the intriguing prevalence of \ncircRNAs in the nervous system indicates the involvement of a general mechanism or effecto r that shifts the \nbalance between linear and circularizing splicing.  \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted June 22, 2024. ; https://doi.org/10.1101/2024.06.22.600180doi: bioRxiv preprint \n\n 3 \nELAV/Hu proteins are a family of highly conserved RBPs, of which at least one member is expressed in a \nnervous-system-specific manner and used as marker of neuronal identity across anima ls 29-31. elav null \nmutations are embryonic lethal and ELAV activity is essential in neurogenesis for establishing and maintaining \nthe neuronal RNA transcriptome via alternative splicing and alternative polyadenylation 32-34. This regulation \noccurs co -transcriptionally, with ELAV proteins targeting sequence elements at the 3' ends or splice sites. \nELAV’s global role in RNA processing prompted the hypothesis that the RBP may be involved in co -\ntranscriptional RNA circularization in the nervous system. In this work, we unveil a previously unrecognized role \nof ELAV proteins as key mediators of circRNA expression in the nervous system, and provide mechanistic \ninsights into circRNA biogenesis. \nResults \nCharacterization of the circRNA landscape in Drosophila embryos \nIn order to identify neuron-enriched circRNAs in vivo, we compared transcriptomes in distinct cell populations. \nWe crossed heterozygous Δelav or ΔelavΔfne mutant parental flies and collected embryonic progeny at two \nstages of late embryogenesis, 14–16 and 18–20 hours after fertilization. The ELAV paralogue FNE can partially \ncompensate for ELAV functions in later embryonic stages 32,35, therefore we used ΔelavΔfne mutants for the 18–\n20h time point (Fig. 1A and S1A). Embryonic tissues were dissociated into i ndividual cells, fixed, fluorescently \nlabeled and flow-sorted into three distinct populations: wild -type neurons, Δelav mutant neurons, and pooled \ncells from all other embryonic tissues (“non-neurons”). RNA sequencing and gene expression profiling confirmed \nthe purity (Fig. 1B and S1B) and identity (Fig. 1C) of each population. \nTo annotate and quantify circRNAs, we counted RNA -seq reads that span back-splice junctions (BSJs) using \nCIRI2 36, including only circRNAs containing canonical splice signals. We treated bulk tissue samples from adult \nfly heads and 14–20 h embryos with RNaseR, a treatment that enriches for circRNAs, and compiled a circRNA \nreference transcriptome by pooling BSJ reads from all sample replicates (Fig. S1C). We subsampled BSJ reads \nfrom each dataset and assessed the number of identified circRNAs in each fraction for each cell population. \nNear-saturation was achieved for circRNAs detected with 5 or more BSJ reads (Fig. S1D); therefore, we set this \nvalue as a threshold for stringent circRNA  identification and subsequent analyses, in line with previous good \npractice large benchmarking studies 37.  \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted June 22, 2024. ; https://doi.org/10.1101/2024.06.22.600180doi: bioRxiv preprint \n\n 4 \n58% of the circRNAs identified (≥5 BSJs) in the sorted cell populations were also present in the RNaseR-treated \nsamples (Fig. S1E, F), validating the accuracy of our annotation and indicating that isolating cell populations \nhelps identify circRNAs that are expressed either lowly or in a tissue -specific manner. Comparing BSJ \nexpression between neuronal and non -neuronal populations with stringent cut -offs, we identified around 400 \ncircRNAs specifically enriched in neurons at either developmental time (hereafter referred to as “neuronal \ncircRNAs”, Fig. 1D and Fig.  S1G-J, Table S1), with only 55 circRNAs depleted compared to other embryonic \ncells (“non-neuronal”). The majority of circRNAs enriched in the neuronal population are transcribed from genes \nassociated with neuronal functions, with a notable enrichment for synaptic signaling and complex behavior (Fig. \nS1K). This aligns with research in mouse tissues demonstrating that host genes of brain-specific circRNAs are \nhighly enriched for synaptic proteins 6, and indicates that circRNAs have a conserved function in the regulation \nof synaptic processes.  \nBSJs are unique to circRNAs and may constitute binding sites for RBPs that regulate the circRNA in a manner \ndistinct from that of the linear transcript. When we compared back-splice junction regions of neuronal circRNAs \nwith those of broadly expressed circRNAs, we observed a significant enrichment for neuronal RBP motifs (Fig. \nS1L, M). Interestingly, when comparing only BSJ regions (BSJ ±25 nucleotides), we found specific neuronal \nRBPs enriched, including Mushroom-body expressed (MUB), Musashi (MSI) and Fragile X mental retardation \nprotein (FMR1). Some RBPs displayed exquisite positional enrichment at the center of the BSJ, i.e., the precise \nregion discriminating circRNAs from their linear cognate (Fig. 1E). Individual circRNAs have been shown to act \nindependently in cell-type-specific gene expression regulation in different tissues 16,38-41 and were proposed to \nact as tissue-specific RBP and microRNA “sponges” 7,13,17. Our results show that circRNAs represent molecular \nsignatures of developing neurons, and suggest they perform cellular functions distinct from those of linear \ntranscripts. \nELAV mediates neuronal circRNA expression  \nThe RBP ELAV is a key regulator of neuron -specific alternative RNA processing ac ross neuronal cell types, \nincluding the generation of neuron -specific (linear) splice isoforms 32-35,42. To test a role for ELAV in circRNA \nexpression, we compared the circular transcriptome of wild -type flow-sorted neurons to those of Δelav and \nΔelavΔfne neurons. We found a marked and global decrease in neuronal circRNA expression in absence of \nELAV proteins, with over 75% neuronal circRNAs significantly downregulated (Fig. 2A, B and Fig. S2A). circRNA \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted June 22, 2024. ; https://doi.org/10.1101/2024.06.22.600180doi: bioRxiv preprint \n\n 5 \ndepletion was more pronounced in ΔelavΔfne double mutants (Fig. 2C, D and Fig. S2B), showing that FNE \nrescues circRNA-related functions of ELAV like it does alternative linear RNA processing 32-34. \nClustering of circRNA expression in ΔelavΔfne neurons identified a distinct pattern: neuronal circRNAs were \nglobally downregulated in ΔelavΔfne mutants while non-neuronal or “other” circRNAs were upregulated (Fig. 2E, \nF), showing that ELAV drives the neuronal identity (“neuronality”) of the circRNA landscape. Additionally, we \nnoted the appearance of circRNAs in ΔelavΔfne neurons not detected in any other cell population, “ ectopic \ncircRNAs”, most of which were undetectable (BSJ<1) even in RNaseR-treated wild-type tissues (Fig. S2C and \nFig. S2D). \nNext, we expressed ELAV in Drosophila S2 cells, which are macrophage -like with very low natural ELAV \nexpression. Ectopic ELAV caused an upregulation of many circRNAs, with very few downregulated (Fig 2G), \nindicating that ELAV constitutes a general, positive regulator of circRNA expression. Thus, it is likely that the \npan-neuronal expression of ELAV protein is the main cause for the high circRNA diversity and abundance in \nneural tissues. \nELAV binds to nascent RNAs but not mature circles \nTo test whether ELAV directly binds neuronal circRNAs in vivo, we performed ELAV RNA immunoprecipitation \nwith UV-crosslinking followed by RNA sequencing (xRIP -seq) in extract from adult head tissue (Fig. 3A). We \nensured identification of high-confidence targets by using a polyclonal antibody directed against native ELAV 32, \nand in an independent experiment, an anti-Flag antibody in head extract of flies in which the endogenous elav \ngene was N-terminally Flag-tagged (elavFLAG flies). Control xRIP-seq samples were obtained from head extract \nof untagged flies treated with the anti -Flag antibody. As ex pected, xRIP-seq recovered almost all transcripts \npreviously identified as ELAV functional targets for AS or APA 32-35 (Fig. S3A). However, we found that ELAV \ndoes not bind circRNAs: BSJs were strongly underrepresented in xRIP samples compared to input (Fig. 3B, C, \nand S3B-D). In contrast, xRIP samples were highly enriched in reads originating from genes that host neuronal \ncircRNAs (Fig. 3B, D, and Table S2). In order to discriminate between pre -mRNA and mRNA binding, we \nperformed an analysis based on eisaR 43: we compared the enrichment of signals from exon-intron boundaries \n(pre-mRNA) against that from exon  sequences (pre -mRNA and mRNA) for each protein -coding gene. \nRemarkably, 90% of genes from which neuronal circRNAs originate were enriched at the pre -mRNA level \n(Fig. 3E and S3E, F). From these data, we conclude that ELAV binds to pre-mRNAs of select genes to promote \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted June 22, 2024. ; https://doi.org/10.1101/2024.06.22.600180doi: bioRxiv preprint \n\n 6 \nneuronal circRNA biogenesis. Absence of ELAV binding to closed circles could mean that ELAV binds to intronic \nsequences that will later b e spliced out of the pre -mRNA. Indeed, analysis of ELAV iCLIP data from adult fly \nheads 32 revealed a significant ELAV binding enrichment in introns that flank neuronal circRNAs (Fig. 3F). \nTogether, our data demonstrate that ELAV globally mediates neuronal circRNA expression through binding the \npre-mRNA of circRNA host genes. \nELAV binds to RCMs to promote back-splicing \nWe hypothesize that in neurons, ELAV binds to flanking introns of neuronal circRNAs in the nascent transcripts \nto promote intron pairing and back-splicing (Fig. 4A). To assess whether the role of ELAV in circRNA biogenesis \nis linked to its established function in regulating neuronal AS, we quantified (linear) splice junction counts in wild-\ntype and ΔelavΔfne neurons (Table S3), thereby identifying neural -specific splice events. We found that the \nmajority of (linear) neuronal splicing events are ELAV -dependent (Fig. 4B, Table S3), consistent with prior \nstudies demonstrating the impact of ELAV proteins in the establishment of neuronal mRNA signatures 32,34, and \nalso suggesting that ELAV’s role in AS is more extensive than previously recognized. Notably, most (60%) \nELAV-dependent circRNAs are produced from genes with no measurable linear AS, indicating that ELAV can \nindependently regulate linear and circular splicing (Fig 4C). circRNAs are often flanked by long introns 13; \ncomparing linear and circular splicing targets of ELAV, we noted a nearly ten -fold difference in median intron \nlength (Fig. 4C). We measured ELAV xRIP signal in introns of genes undergoing either neuronal AS or neuronal \nback-splicing, and found comparable and high enrichment of ELAV binding in both scenarios (Fig 4D). These \ndata show that the ELAV -dependent generation of neuronal circRNAs constitutes a regulated  process, which \nlikely evolved to benefit neuron development and functionality. In contrast, ectopic or ELAV -independent \ncircRNAs did not display such specific regulation (Fig. 4D, see also Figs. 1E, 2D-F and S2D), and may constitute \nnon-functional or even deleterious side-products of linear splicing, as has been proposed to be the case for \nmany mammalian circRNAs 44.  \nOne possible mechanism through which ELAV binding promotes back-splicing is by inhibiting linear splicing of \nlong introns that flank circRNAs. We evaluated co-transcriptional splicing efficiency using nascent RNA-seq data \nfrom Drosophila heads 45, comparing signal in unspliced introns to that in their respective downstream exons 46 \n(Fig. S4A). Consistent with findings in Drosophila S2 cells 47, splicing efficiencies generally decreased with intron \nlength (up until 10  kb; Fig. S4B). Accordingly, the —typically long — introns flanking circRNAs exhibited \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted June 22, 2024. ; https://doi.org/10.1101/2024.06.22.600180doi: bioRxiv preprint \n\n 7 \nsignificantly lower splicing efficiencies than non -circRNA-associated introns (Fig. S4C). Inter estingly, ELAV-\ndependent introns are spliced significantly less efficiently compared to other introns, whether in the context of \nAS or circRNA formation (Fig. 4E), even within the same gene (Fig.  S4C). These observations suggest that \nELAV binding reduces splicing efficiency, thereby increasing the window of opportunity for AS and back-splicing. \nWe also hypothesized that ELAV binding may facilitate the formation of secondary structures that promote intron \npairing at reverse complementary match sequences (RCM s) (Fig. 4A). RCMs constitute hallmarks of inter -\nintronic interaction and can predict circRNA formation 23,24,48. We identified multiple RCMs in flanking introns of \nmost circRNAs, with a particular prevalence in ELAV -dependent circRNAs (Fig. S4D). We found  that RCM \nregions were highly enriched in ELAV binding motifs (Fig. 4F); ELAV iCLIP from brain tissue showed significant \nlocal binding of ELAV at RCMs flanking neuronal circRNAs, with signal particularly strong in the upstream intron, \ni.e., the intron tran scribed first (Fig. 4G). Together, our data show that ELAV regulates back -splicing and \nneuronal circRNA formation by binding to RCMs in flanking introns.  \nDiscussion  \nIn this study, we demonstrate that the pan -neuronal protein ELAV is the global mediator o f neuronal circRNA \nsynthesis; ELAV’s tissue-specific expression underlies the extraordinary abundance of circRNAs in the nervous \nsystem. Mechanistically, ELAV binds to the pre -mRNA of genes that produce neuronal circRNAs, specifically \ntargeting RCM sequences in BSJ-flanking introns. Long introns are associated with circRNA formation in flies \n(this study; 13) and mammals 49 as well as with low splicing efficiencies 50,51, which supports the notion that many \ncircRNAs constitute functionally neutral or deleter ious side-products of splicing. Our data indicate that most \nneuronal circRNAs do not belong to that category: ELAV -dependent circRNA formation occurs in a regulated \nand highly specific fashion, with reduced splicing efficiencies in ELAV-bound introns. In our model, ELAV binding \nto the RCM in the upstream flanking intron represents the crucial step, ensuring retention of that intron and its \nRCM while the downstream intron is being transcribed, which can take several minutes in long neuronal genes. \nIt is stil l not clear how ELAV achieves the targeting specificity required to induce circularization at distinct \nneuronal BSJs. The ELAV binding motif is found broadly distributed across the genome and can by itself not \naccount for the observed positional enrichment  at RCMs. This is reminiscent of the high prevalence of Alu \nelements in the introns of primates, where Alu pairing regulates linear 52,53 and circular 54 splicing. Even within \nthe nervous system, circRNAs display cell -type-specific expression patterns 15,55,56; it is possible that the \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted June 22, 2024. ; https://doi.org/10.1101/2024.06.22.600180doi: bioRxiv preprint \n\n 8 \ncombination of transcription and RNA processing factor specificities selects and modulates transcript \ncircularization in distinct neurons, perhaps even in distinct synapses.  \nIn fly embryos, at least 75% of neuronal circRNAs depend on ELAV/FNE. The accumulation of circRNAs in the \nbrain is a molecular landmark across animals 1,12,13. Considering the similarity, in terms of sequence and \nexpression patterns, of neuronal ELAV -like (nELAV) proteins from flies to humans 29,57,58, con served \nmechanisms may be at play to regulate intron pairing and neuronal circRNA biogenesis. Moreover, the \npredominantly cytoplasmically localized human nELAV protein HuD binds to a quarter of all brain -expressed \ncircRNAs as well as many host transcripts 59; in context with our findings, this raises the hypothesis that in \nevolutionarily distant species, nuclear nELAV proteins (such as ELAV) mediate circRNA biogenesis, and that \nbinding of mature circRNAs by cytoplasmic nELAVs (such as HuD) regulates their local expression and function. \nThe exquisite co -regulation of neuronal circRNAs at the levels of biosynthesis (this study, 28), subcellular \nlocalization, and response to extrinsic signals 6,12,60, could indicate that neuron -specific, nELAV -dependent \ncircRNAs may function in concert, for example by binding a common set of RBPs. Our finding that BSJ regions \nof neuronal circRNAs are enriched for binding motifs for neuronal RBPs are consistent with this possibility. RBPs \noften localize in synapses, where they reg ulate the local translation and transport of neuronal transcripts 61,62. \nWith their long half-life, unique RBP binding sites, and versatile functionalities, circRNAs possess great potential \nfor modulating gene expression in a cell-specific and even synapse-specific manner. \nLimitations of the Study \nOur cell sorting approach aimed at enriching neuronal cell types, providing insights into nervous system specific \ncircRNA expression. It is important to note that circRNAs can exhibit highly cell-type specific expression patterns. \nThis specificity cannot be resolved in our sorting methodology. To address this limitation, future studies using \nsingle-cell RNA sequencing technologies are essential. Such approaches would allow for the high -resolution \ndetection of circRNAs within individual brain cells. Additionally, single -cell studies could uncover the dynamic \nregulation and functional roles of circRNAs in diverse neuronal populations, contributing to a deeper \nunderstanding of their involvement in brain development and neurological diseases. \n \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted June 22, 2024. ; https://doi.org/10.1101/2024.06.22.600180doi: bioRxiv preprint \n\n 9 \nAcknowledgements \nWe thank Fernando Mateos for technical help. We are grateful to Alejandro Gomez Auli and Gerhard Mittler at \nthe Proteomics Core, Ulrike Bönisch and the Deep Sequencing Core, and Thomas Manke and the Bioinformatics \nCore at MPI-IE. We thank Andreas Lingel for his help in designing elavRBD. We thank Hasan-Can Ozbulut and \nNikolaus Rajewsky for helpful discussions, and Anton Heß for critical reading of the manuscript. Stocks obtained \nfrom the Bloomington Drosophila Stock Center (NIH P40OD018537) were used in this study. This work was \nfunded by the Max Planck Society, the Deutsche Forschungsgemeinschaft (DFG, German Research \nFoundation) SFB 1381 (Project-ID 403222702) and under Germany’s Excellence Strategy (CIBSS-EXC-2189-\nProject-ID 390939984), and the European Research Council (ERC) under the European Union’s Horizon 2020 \nresearch and innovation program (grant agreement ERC-2018-STG-803258). \nAuthor contributions \nC.A.-G. and V.H. conceptualized the study. C.A. -G., S.H., S.G., M.S., J.C., and S.T. performed experiments. \nC.A.-G., S.H., S.G., M.S. and V.H. designed and analyzed experiments. C.A. -G. and V.H. designed \ncomputational data analysis. C.A -G. and M.R. performed computational data analysis. C.A. -G. and V.H. \nprepared the figures. C.A.-G. and V.H. wrote the manuscript with input from all authors. V.H. supervised the \nstudy and acquired funding. \nDeclaration of interests \nAuthors declare no competing interests. \n  \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted June 22, 2024. ; https://doi.org/10.1101/2024.06.22.600180doi: bioRxiv preprint \n\n 10 \nFigures and figure legends \n \nFigure 1. The circRNA landscape in Drosophila embryos. See also Figure S1 and Table S1. \n(A) Experimental overview: cells from embryonic progeny of Δelav,fneFLAG heterozygous flies (carrying an elav \nnull mutation recombined with a Flag-tagged allele of the neuronal marker FNE) were FACS-sorted into three \ndistinct populations: wild-type neurons (ELAV+, Flag+), Δelav mutant neurons (ELAV-, Flag+), and non-neuronal \ncells (ELAV-, Flag-). circRNA expression was quantified from total RNA -seq data, measuring reads spanning \nthe back-splice junction (BSJ reads) unique to circRNAs. \n(B) Total RNA-seq tracks at the elav locus, in the sorted cell populations. The loss of signal in the elav coding \nregion in Δelav neurons is highlighted. \n(C) Principal-component analysis plot of gene expression across the embryonic cell populations. The heatmap \non the right indicates the top differentially expressed genes between neuronal and non -neuronal populations. \nReplicates and identity of cell populations are indicated with colored dots. (D) Differential circRNA expression in \nneurons compared to non-neuronal populations represented as a function of BSJ counts per million. Significantly \nenriched (neuronal) circRNAs (p<0.05 and Log(CPM)> -5.4, Log2FC ≥1, blue) or depleted (non -neuronal) \ncircRNAs (Log2FC≤-1, brown) are highlighted. \n(E) Left, RBP binding motifs significantly enriched at the BSJ (±25 nt) of neuronal circRNAs, compared to the \nentire linearized circRNA sequence. Right, positional enrichment of binding motifs for three neuronal RBPs are \nindicated from the center of the back-splice junction (right). \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted June 22, 2024. ; https://doi.org/10.1101/2024.06.22.600180doi: bioRxiv preprint \n\n 11 \n \nFigure 2. ELAV regulates neuronal circRNA formation. See also Figure S2 and Table S1.  \n(A) Total RNA-seq signal tracks and representation of the BSJ count of a portion of the gene discs large 1 (dlg1), \nin the sorted cell populations. Arrows in the gene model indicate the back -splicing events that produce the \nneuronal (blue) or non-neuronal (brown) circRNAs, respectively.  \n(B) Differential circRNA expression in wild -type neurons compared  to ΔelavΔfne neurons, represented as a \nfunction of BSJ counts per million. Highlighted dots represent circRNAs classified as neuronal (Fig. 1D, 376 \nneuronal circRNAs). The dotted line indicates the abs(log2FC)≥0.3 cutoff. \n(C) Differential circRNA expressi on in Δelav and ΔelavΔfne mutant neurons compared to wild -type neurons. \n****p<0.0001 (two-tailed Welch’s t-test).  \n(D) Proportion of circRNAs with neuron -specific expression (neuronal, non-neuronal, other) affected in Δelav \nand in ΔelavΔfne neurons. circRNAs were considered significantly affected (up or down) in mutant compared to \nwild-type neurons if abs(Log2FC)≥0.3 (ΔelavΔfne) or abs(Log2FC)≥0.5 (Δelav), and Log(CPM)>-5.4.  \n(E) Confusion matrix displaying the fraction of circRNAs with neuron -specific expression that are affected (up, \ndown, unchanged) in ΔelavΔfne mutants. \n(F) Heatmap representing circRNA expression in neurons (differential expression in neurons compared to non-\nneurons) and in ΔelavΔfne (differential expression in mutant compared to wild -type neurons). circRNAs are \ngrouped according to their expression in cell populations (neuronal, non -neuronal, other). Only circRNAs \nsignificantly affected in ΔelavΔfne are represented.  \n(G) Differentially expressed circRNAs upon ELAV expression in S2 cells compared to an empty vector control. \nSignificantly (p<0.1) upregulated (Log2FC≥0.33, purple) and downregulated (Log2FC≤-0.33, brown) circRNAs \nare highlighted. Dots are jittered to reduce overlap. \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted June 22, 2024. ; https://doi.org/10.1101/2024.06.22.600180doi: bioRxiv preprint \n\n 12 \n \nFigure 3. ELAV binds to pre-mRNA of neuronal circRNAs host genes. See also Figure S3 and Table S2. \n(A) xRIP-seq workflow. UV -cross-linked heads from adult flies underwent nuclear fractionation followed by \nisolation of protein -RNA complexes and RNA sequencing. In the shown example, Flag -tagged ELAV was \ncaptured using anti-Flag beads in flies of the genotype elavFLAG or w1118 (control). \n(B) Total RNA-seq signal tracks and representation of the BSJ count of a portion of the gene nAChRalpha4 in \nFlag-ELAV xRIP compared to input. The arrow in the gene model indicates the back-splicing event that produces \nthe neuronal circRNA.  \n(C) ELAV binding to circRNAs, calculated as the ratio of circular to linear transcript expression in Flag -ELAV \nxRIP compared to input, in elavFLAG and w1118 (control) flies. ****p<0.0001 (one-tailed Welch’s t-test).  \n(D) Left, proportion of genes that displayed significant enrichment in Flag -ELAV xRIP-seq compared to input \n(Log2FC>1 and p<0.01), in host genes of neuronal circRNAs and in all other genes (expressed in input sample). \n****p<0.0001 (Pearson's Chi-squared test). Right, differential linear transcript expression in Flag -ELAV xRIP \ncompared to input, for the same gene groups. ****p<0.0001 (one-tailed Welch’s t-test).  \n(E) Exon-intron split analysis of Flag-ELAV xRIP compared to input. Each d ot represents one gene. For each \ngene, ELAV enrichment at exon-intron boundaries (pre-mRNA) is shown as a function of ELAV enrichment in \nexons (total transcript). Host genes of neuronal circRNAs are highlighted in blue. Total transcript enrichment and \npre-mRNA enrichment were calculated from exon reads and from reads overlapping exon -intron boundaries, \nrespectively.  \n(F) Enrichment of ELAV iCLIP signal in introns. Introns immediately upstream or downstream of neuronal \ncircRNA back-splice sites (flanking neuronal circRNAs) are compared for enrichment against other introns of the \nsame gene (non-flanking) and introns of genes that do not host a neuronal circRNA (all other).  \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted June 22, 2024. ; https://doi.org/10.1101/2024.06.22.600180doi: bioRxiv preprint \n\n 13 \n \nFigure 4. ELAV regulates circRNA production in inefficiently spliced introns.  See also Figure S4 and \nTable S3. \n(A) Model of ELAV -mediated circRNA biogenesis: ELAV binds to reverse complementary match sequences \n(RCM) in introns flanking the circRNA back -splice site (BSJ) and inhibits splicing efficiency, resulting in \nsecondary structures that promote neuronal circRNA expression. \n(B) Proportion of neuronal alternative splicing events (differential splicing events in neurons compared to non -\nneurons) that are ELAV-dependent (differential in ΔelavΔfne compared to wild-type neurons), for the indicated \ntypes of splicing. \n(C) Venn diagram showing the number of genes that undergo ELAV-dependent alternative splicing (linear) and \nof genes that host an ELAV-dependent circRNA (circular). Below, box plot representing the length of the flanking \nintrons for each type of splicing event. ****p<0.0001 (one-tailed Welch’s t-test).  \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted June 22, 2024. ; https://doi.org/10.1101/2024.06.22.600180doi: bioRxiv preprint \n\n 14 \n(D) Gene metaplot representing ELAV xRIP-seq enrichment in flanking introns of the indicated types of ELAV-\nregulated splicing events. Forward and reverse arrows in the gene model represent linear and back -splicing \nevents, respectively.   \n(E) Splicing efficiency of introns undergoing linear splicing or back-splicing, comparing ELAV-regulated AS and \ncircRNA formation to ELAV-independent (other) splicing events. ****p<0.0001 (one-tailed Welch’s t-test). \n(F) Heatmap showing the number and distribution of ELAV motifs in RCM regions of introns flanking neuronal \ncircRNAs. RCMs of introns that contain at least one ELAV motif (independently of the motif’s position within the \nintron) are displayed.  \n(G) Profile plot of ELAV iCLIP signal at RCMs and in 1kb surrounding, in introns flanking (upstream and \ndownstream) neuronal circRNAs. \n \nSupplemental Information \nDocument S1. Figures S1–S4.  \nSupplemental Tables. Tables S1-S4. Excel files containing additional data too large to fit in a PDF. \nTable S1. CircRNA identification and quantification in flow -sorted cell populations from Drosophila \nembryos. Related to Fig. 1 and Fig. 2. circRNAs identified as differentially expressed in each indicated dataset \n(neurons vs. non-neurons at 14–16h, neurons vs. non-neurons at 18–20h, Δelav neurons vs. wild-type neurons \nat 14–16h, ΔelavΔfne neurons vs. wild-type neurons at 18–20h), with quantification results.  \nTable S2. xRIP -seq identification of transcripts directly bound by ELAV. Related to Fig.  3. Transcripts \nidentified as enriched in each indicated dataset (anti -ELAV antibody xRIP vs. input in wild -type flies, and anti-\nFlag antibody xRIP vs. input in elavFLAG flies), with quantification results. \nTable S3. Identification and quantification of alternatively spliced exons in flow-sorted cell populations \nfrom Drosophila embryos. Related to Fig. 4.  Exons identified as differentially expres sed in each indicated \ndataset (neurons vs. non -neurons at 18 –20h, ΔelavΔfne neurons vs. wild -type neurons at 18 –20h), with \nquantification results.  \nTable S4. Recombinant DNA and RT -qPCR oligonucleotides used in this study. Related to STAR \nMethods. \n  \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted June 22, 2024. ; https://doi.org/10.1101/2024.06.22.600180doi: bioRxiv preprint \n\n 15 \nSTAR Methods \nContact for reagents and resource sharing \nFurther information and requests for resources and reagents should be directed to and will be fulfilled by Valérie \nHilgers (hilgers@ie-freiburg.mpg.de).  \nData and code availability \n● All sequencing data generated during this study can be accessed at NCBI Gene Expression Omnibus under \nthe accession number (GSE269179). \n● Code used in this study to generate all figures is deposited in https://github.com/hilgers-lab/circles2024 and \nhttps://github.com/hilgers-lab/SpliceFlow.  \n● Any additional information required to reanalyze the data reported in this paper is available from the lead \ncontact upon request. \nExperimental model and genome editing \nIn this study, all experiments used Drosophila melanogaster male embryos. The age of the embryos is specified \nin hours after egg laying (AEL) when maintained at 25°C. Control flies ( w1118) and GFP -marked balancer \nchromosomes were sourced from the Bloomington Stock Center (lines 5905, 4559, 6 662). Null alleles for fne, \ndenoted as Δfne, were acquired from M. Soller (Zaharieva et al., 2015) 63; Δelav and fneFLAG alleles are from \n(Carrasco et al., 2020)32. elavFLAG and elavRBD fly lines were generated through CRISPR/Cas9 genome editing, \nadhering to the methods outlined by Port and Bullock 64. Embryo injections were performed by Bestgene, Inc. \nelavFLAG expressing an endogenously, N-terminally Flag-HA-tagged ELAV protein, was generated using a guide \nRNA (GTCTACTCCGCCGCCAGCTC) targeting elav and a plasmid comprising 1406 nucleotides upstream of \nthe tag insertion site, the Flag-HA sequence, and 1509 nucleotides downstream the insertion site (sequence in \nTable S4). To create elavRBD, a total of ten single-nucleotide mutations in all six RNP motifs of the three ELAV \nRRMs, previously described to eliminate ELAV RNA -binding (Lisbin, M. et al., 2000), were introduced by \ninjecting the elavFLAG line with a plasmid containing two guide RNAs (AACCACAGCAGGCGCAGCCC, \nGTCTACTCCGCCGCCAGCTC) and a 1257-nucleotide gBlock (Integrated DNA Technologies) as a homology \ndonor (sequence in Table S4). The resulting amino acid mutations are the following: RRM1, RNP1: Y205A and \nF207D. RRM1, RNP2: I152A and N154A. RRM2, RNP1: F294A. RRM2, RNP2: Y251A. RRM3, RNP1: Y445A \nand F447D. RRM3, RNP2: F405A and Y407A. \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted June 22, 2024. ; https://doi.org/10.1101/2024.06.22.600180doi: bioRxiv preprint \n\n 16 \nGenetic strategy to sort wild-type and Δelav mutant neurons (14–16h dataset) \nWe used the progeny of Δelav, fneFLAG/FM7 and FM7/Y flies that contain the elav null allele Δelav 32 recombined \nwith an endogenously V5- flagged fne (fneFLAG) 65. Flag was used as a marker of neuronal cells and ELAV protein \nwas used as a marker of wild-type neurons, which allowed to distinguish Δelav mutant neurons (Flag+, ELAV-), \nwild-type neurons (Flag+, ELAV+), and non-neuronal cells (Flag-, ELAV-). \nGenetic strategy to sort wild-type and ΔelavΔfne mutant neurons (18–20h dataset) \nFlies described as ΔelavΔfne flies are progeny of elavRBD,Δfne/FM7 crossed with FM7/Y males. As a result, the \nRNA-binding dead ELAV protein can be discriminated from the wild-type ELAV by Flag, and never co-expresses \nwith FNE. In heterozygous flies, the elavRBD allele is robustly suppressed (through a mechanism we have not \ncharacterized), leading to the absence of detectable elavRBD protein (Figure S1A). This suppression allows for \nthe positive identification of neuronal populations that lack the ELAV RBD protein using Flag as a marker. Wild -\ntype neurons (Flag- and ELAV+) were distinguished from mutant neurons ( elavRBD,Δfne: Flag+, ELAV+), and \nnon-neuronal cells were identified as (Flag-, ELAV-).  \nRNaseR treatment  \nFor head samples, 3-day-old w1118 flies were collected, flash-frozen in liquid nitrogen, decapitated by shaking \nand heads were manually collected with a thin brush. For embryo samples, eggs from w1118 flies were collected \nfor two hours on agar plates and aged for 14h (14 –16h AEL embryos) at 25°C. In both cases, samples were \nhomogenized in QIAzol Lysis Reagent (QIAGEN 79306) for RNA extraction. 2µg RNA per sample was treated \nwith 10 U RNAseR (Applied Biological Materials, E049) in 20 µL for 20 min at 37°C. For non -treated control \nsamples, the enzyme was omitted. After addition of spike-in mouse embryonic stem cell RNA at 0.33 ng/µL final \nconcentration, RNA was extracted with Trizol LS (Ambion 10296028) and glycogen (Invitrogen AM9515) \naccording to the manufacturer's protocol. RNAs were reverse transcribed using Maxima RT (Thermo Scientific \nEP0741) with random hexamers (Jena Bioscience 94824), according to the manufacturer's protocol. \nTotal RNA sequencing library preparation   \nRNA integrity was analyzed using a 2100 Bioanalyzer (Agilent Technologies).  Libraries for total RNA-seq were \nprepared with 100 ng of total RNA using TruSeq Stranded total RNA Library Prep (Gold) (Illumina 20020599) \naccording to the manufacturer’s instructions. Paired -end sequencing was performed using the NovaSeq6000 \nplatform (Illumina) and 101-bp reads. \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted June 22, 2024. ; https://doi.org/10.1101/2024.06.22.600180doi: bioRxiv preprint \n\n 17 \nFluorescence-activated cell sorting in Drosophila embryos  \nDrosophila embryos were collected at the appropriate times after egg laying. Sorting of embryo populations was \ndone based on (McCorkindale et al., 2019)66, with modifications. All steps until FACS were carried on at 4°C, \nand all washes used cold, RNAse -free PBS (Alfa Aesar, J62851) with RNase inhibitor (1:250) (Ribolock, \nThermoFischer) and 1000g for 2 min to preserve RNAs from degradation. Embryos were dounced in RNAse-\nfree PBS with RNase inhibitor (1:250), and cell debris were filtered out by filtration through a 40µm strainer at \n400g. Cells were washed twice, and stained with Zombie Aqua (BioLegend, 423101) for live/dead separation \nfor 15 min in the dark, in PBS with 1:100 RNase inhibitor. Cells were washed twice and then fixed with 4% \nformaldehyde (ThermoFischer scientific, 28908) in PBS for 15  min in the dark. Fixation was stopped with \nincubation for 3 min with a quenching buffer (750 mM TrisHCl pH 7.5 in DEPC-PBS, 1:100 RNase inhibitor) and \ncells were washed again twice. Cells were incubated with the following primary antibodies: polyclonal anti-ELAV \n1:500 (rabbit, generated in Carrasco et al., 2020)32 and mouse-anti-Flag M2 (Sigma, F1804) at 1:500 for 45 min \nin the dark, with agitation, at 4 °C in staining buffer (1% BSA  (w/v)( Sigma, 6917), 0.1% saponin, 1:200 \nRiboLock). Cells were washed twice in 0.2% BSA, 0.1% saponin in PBS, 1:100 RiboLock and incubated with \nfluorescently conjugated antibodies at 1:500 (14–16h embryos) or 1:1000 (18–20h embryos) for 45 min in the \ndark, with agitation, at 4 °C in staining buffer. Sorting used 488 anti -mouse (Abcam) and 555 anti -rabbit \n(Invitrogen) in Δelav embryos; and 488 anti -rabbit (Life Technologies) and 555 anti -mouse (Abcam) in \nΔelavΔfne. Cells were then washed twice, resuspended in sorting buffer (0.5% BSA, 2mM EDTA, 1:500 \nRiboLock), filtered through a 40µm nylon mesh, sorted using a FACSymphony (Becton Dickinson), and \ncollected in cold sorting buffer. Cells were pelleted and RNA-protein complexes were reverse-cross-linked with \nproteinase K (Ambion AM2546) in 10mM Tris, pH8, 10mM NaCl, 1mM EDTA, 1:100 proteinase K, 1:100 \nRiboLock), for 20 min at 50° C. RNA was extracted in 3:1 Trizol LS (Ambion) according to the manufacturer’s \nprotocol. \nCell transfection  \nS2 cells were transfected in six-well plates using Effectene (Qiagen) with 300 ng of the 𝜆N expression plasmid: \n𝜆N empty plasmid or 𝜆N-ELAV from Hilgers et al., 2012 67. Transfections were conducted in three replicates.  \nCells were harvested 48 hours after transfection by centrifugation and lysed in QIAzol Lysis Reagent (QIAGEN \n79306) for RNA extraction. \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted June 22, 2024. ; https://doi.org/10.1101/2024.06.22.600180doi: bioRxiv preprint \n\n 18 \ncircRNA validation by qPCR \n300 ng of total RNA was used per replicate. RNA was reverse transcribed using Maxima RT (Thermo Scientific \nEP0741) with random hexamers (Jena Bioscience 94824) according to the manufacturer's protocol. RT-qPCR \nwas performed in a LightCycler 480 II instrument using FastStart SYBR Green Master (Roche). Divergent \nprimers were designed to amplify only circular RNAs from the target genes. RT -qPCR primer sequences are \nlisted in Table S4. \nCross-linking RNA-Immunopurification and sequencing (xRIP-seq) \nDrosophila heads were separated from bodies by shaking in liquid nitrogen, and ground into powder in liquid \nnitrogen. For one sample, 100 mg head powder was subjected to UV cross-linking (UV-C, 254 nm) through six \nrounds at 300 mJ/cm² in a BLX-312 crosslinker (Bio-Link). All subsequent steps were performed at 4°C. Nuclear \nextracts were obtained by homogenizing the UV-cross-linked head powder in 1 mL homogenization buffer (0.2% \nTriton X, 10% sucrose, 0.5 mM DTT, double the standard concentration of protein inhibitors (Roche, \n11873580001), and 1:500 RiboLock RNase inhibitor (Thermo Fisher Scientific, EO0384), using ten strokes with \na loose pestle. For each replicate, three samples (3mL of homogenate) were pooled, homogenate was filtered \nusing a 40µm strainer and centrifuged at 200 g for 3 min. Supernatants were transferred to 15-mL Protein LoBind \n(Eppendorf, 0030122216) tubes and centrifuged at 800 g for 10 min to pellet nuclei. Pellets were resuspended \nand centrifuged at 600 g for 10 min to pellet nuclei in 6 mL homogenization buffer excluding Triton X and \ncentrifuged at 800 g for 10 min to pellet nuclei. This pellet was then resuspended in Lysis buffer (50 mM Tris \n(pH 7), 500 mM LiCl, 10 mM EDTA, 5 mM DTT, 2% lithium dodecyl sulfate (LDS), 0.5% SDS, and 0.5% sodium \ndeoxycholate) and rotated for 10 min. The lysate was centrifuged at 15,000 g for 10 min, and the supernatant \ntransferred to a new tube. After another centr ifugation under the same conditions, ionic detergents were \nneutralized with 1% NP -40. For the Flag -ELAV experiment, the clarified lysate (750 µL; referred to as 'input') \nwas incubated with 40 µL anti-Flag M2 magnetic beads (Invitrogen, M8823) for 1.5 hours. The beads were rinsed \nwith lysis buffer (0.1% SDS and 0.1% Na -deoxycholate), washed three times for 5 min with lysis buffer (0.1% \nSDS, 0.1% Na-deoxycholate, and 1% Triton X-100), and washed twice with lithium chloride buffer (350 mM LiCl, \n50 mM Hepes -KOH (pH 7.5), 1 mM EDTA, 1% NP -40, and 0.7% Na -deoxycholate) for 5 min. Protein -RNA \ncomplexes were eluted with 120 µL of elution buffer (10 mM Tris -HCl (pH 7.4), 150 mM NaCl, 0.1% Triton X -\n100, without protease inhibitors) containing Flag peptide (0.2 mg/ml; Sigma-Aldrich, F3290) for 1 hour. The \neluates and corresponding inputs underwent treatment with proteinase K (Ambion, AM2546) for 30 min at 50°C \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted June 22, 2024. ; https://doi.org/10.1101/2024.06.22.600180doi: bioRxiv preprint \n\n 19 \nand 1100 rpm. For the endogenous ELAV IP experiment, input  was incubated with 40uL of the conjugated \nDynabeads protein A-antibody complex with anti-ELAV antibody (rabbit, generated in Carrasco et al., 2020) 32 \nfor 1 hour. Protein-RNA complexes were eluted in 120uL of 1x Proteinase K Buffer (Tris HCl pH 7.5, NaCl EDTA, \n6 uL 10% SDS, 1.5 uL Proteinase K (20 mg/mL) an d 1 uL Ribolock). RNA was then purified using TRIzol LS \nReagent (Ambion, 10296028) following the manufacturer's protocol. Equal volumes of total RNA from both input \nand IP samples were used to prepare libraries for total RNA-seq.  \nmRNA sequencing data processing. \nSequencing data were processed using the RNA-seq module from snakePipes v2.4.3 68, adding flags for --trim, \n-m “alignment -free,alignment”. Reads were mapped to the Drosophila melanogaster  reference genome \n(Ensembl assembly release dm6), and the transcriptome reference annotation release-96 using STAR 69. Quality \ncontrol of RNA -seq reads was done using FASTQC 70. To compare gene expression estimates across cell \npopulations, a variance stabilizing transformation (VST) was applied using the DESeq 71 function vst() on raw \ngene counts data from the different samples. The transformed data was used to compute a PCA using the \nDESeq2 function plotPCA() with standard parameters. Differential gene expression analysis was performed \nusing DESeq2 filtering for baseMean>=4, logFC>1 and padj<0.01 to determine neuronal enriched genes. \nGeneration of the circRNA reference database  \nThe circRNA reference database was constructed from a compilation of libraries from RNaseR -treated RNA \nsamples (from heads and embryos) an d from RNA samples obtained from flow -sorted embryonic cell \npopulations. For each group, replicates were pooled. Libraries were aligned using Burrows-Wheeler Alignment \ntool with the parameter  \"-T 19 \" 72,  and circRNAs identified using CIRI2 under default settings as recommended \nby the tool's developers 36, based on the dm6 genome and Ensembl 96 annotations. This process involved \nfiltering and consolidating the identified circRNA entities. Filtering criteria included a minimum of five back-splice \njunctions (BSJ) per circRNA per cell  population, with mutant -specific circRNAs identified by the presence of \nmore than five BSJs across the combined single and double mutant libraries. This consolidated collection of \ncircRNAs formed the reference database that was used in subsequent analyses. \nSaturation analysis (sub-sampling) \nSaturation analysis was performed by pooling all replicates from one sample group, and randomly sampling \ndifferent fractions from 1% to 100% from the raw read files using seqtkV1.2 -r94 73. Then, the CIRI2 36 pipeline \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted June 22, 2024. ; https://doi.org/10.1101/2024.06.22.600180doi: bioRxiv preprint \n\n 20 \nwas applied to each of the sample fractions. Results were summarized as a fraction of recovered compared to \nthe full set. \nDifferential circRNA expression \nWe computed differential circRNA expression by executing the CIRIquant 74 workflow and using the circRNA \nreference database. CIRIquant was run with circRNA in format \" --ciri --library-type 2 \"and using script prepDE.py \nto handle replicates to compute the differential status for each circRNA. After the circRNA quantificat ion with \nCIRIquant, circRNAs were classified based on their expression in a given embryonic cell population compared \nto another cell population. We constructed four groups, two for the neuronal comparison (neuronal, non -\nneuronal) and two for the mutant com parison (ELAV-down, ELAV-upregulated). For classification, we require \nthat a circRNA comply with edgeR parameters of logCPM>-5, p-value<0.1 and |LogFC|>0. \nMotif enrichment at BSJs \nRBP enrichment in BSJs of neuronal circRNAs was performed by generating a se quence of 50 nt flanking the \nBSJ site (±25nt) using the BSgenome.Dmelanogaster.UCSC.dm6 reference genome package in R. The FASTA \nfiles were submitted to the MEME suite server and the AME 75 program was used to calculate enrichment over \nthe sequences. CentriMO 76 was used for assessing positional enrichment. For the comparisons, circRNAs that \nwere neither enriched nor depleted in neurons were used as control sequences.  \nIdentification of ELAV target genes by xRIP-seq \nTo identify ELAV target genes, we used th e DESeq2 pipeline DESeq 71 and assessed read counts in xRIP \ncompared to input. Prior to applying DESeq2, we filtered for genes with more than 10 read counts across all \nsamples and replicates. For xRIP -seq performed with the anti -ELAV antibody in wild-type fly heads, DESeq2 \nwas executed with default parameters using a simplified design: \" design = ~type\", where 'type' represents the \ncomparison between input and IP. For xRIP-seq performed with the anti-Flag antibody in elavFLAG flies, we used \na likelihood ratio test with the following parameters: \"design = ~type + condition + condition:type and DESeq(test \n= \"LRT\", reduced = ~type + condition)\" . In this context, 'condition' accounts for the confounding factor of non -\nspecific binding to the anti-Flag antibody, which we controlled for by performing Flag IP in untagged w1118 flies. \nELAV target genes for each xRIP experiment were selected with the criteria p.adj<0.05 and log2FoldChange>1.  \nExon-intron split analysis (EISA) on xRIP-seq data \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted June 22, 2024. ; https://doi.org/10.1101/2024.06.22.600180doi: bioRxiv preprint \n\n 21 \nWe counted reads specifically spanning exon-exon (EEJ) junctions and exon-intron boundaries (EIB). EIB reads \nwere defined as reads that do not span exon-exon junctions and fully match the reference genome, indicated by \nhaving only 'M' (match or mismatch) i n their CIGAR string from STAR mapping. These reads are assigned to \nregions that overlap both exons and introns. EEJ and EIB reads were counted using featureCounts 77 with the \nparameters: \"featureCounts -p -B -C -t ‘gene’ -g ‘gene_id’ -f -s 2 -J \". EISA analysis was performed using the R \npackage eisaR with parameters from Gaidatzis et al., 2015 43: \"runEISA(modelSamples = FALSE, geneSelection \n= 'filterByExpr', statFramework='QLF', effects='predFC', pscnt=2, sizeFactor = 'individual', \nrecalcNormFactAfterFilt = TRUE, recalcLibSizeAfterFilt = FALSE) \". In this analysis, read counts from EEJs were \nused for mRNA counts and reads of EIBs for intron counts (pre-mRNA). Genes with fewer than 10 EEJ counts \non average across replicates were discarded. \nELAV iCLIP data processing \nPreviously published ELAV iCLIP data was processed as described in Carrasco et a l. 2020. In brief, we used \niCount 78 to align iCLIP reads and call significant cross -link regions. Intronic cross-link sites were called using \niCount and the exonic signal of isoform-specific exons was removed before computing the mean coverage per \nregion across replicates. These signal tracks represent highly confident intronic cross-link sites. For the analysis \nof ELAV iCLIP signals in flanking introns, we first constructed an intron database to identify sets of introns \nflanking circRNAs and differentially expressed exons. This enabled us to compute signal enrichment in distinct \ngroups. To create the intron database, we selected protein -coding genes, excluded first/last exons from \nalternative transcription start sites (TSSs) and alternative transcription e nd sites (TESs), and discarded non \nexpressed exons. The remaining exons were then subtracted from gene coordinates, resulting in the intron \ndatabase. We excluded introns smaller than 10 nucleotides. To compute iCLIP enrichment, we selected the \nfollowing features: i) Upstream and downstream introns immediately flanking a circRNA, ii) Non-flanking introns, \niii) All introns of the gene set. To assign introns to circRNAs, we collapsed overlapping circRNAs whose \nrespective 5' and 3' ends lie within 10 nt of each other. Introns were assigned to a feature if they were within a \n10-nt window. For each set of introns, we assessed enrichment by computing the fraction of iCLIP signal in \nintrons normalized by intron length. The iCLIP profile represents the piled -up count of cross-link sites within \nnormalized intron sizes.  \nDetection of differential alternative splicing events.  \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted June 22, 2024. ; https://doi.org/10.1101/2024.06.22.600180doi: bioRxiv preprint \n\n 22 \nTo identify alternative splicing events, rMATS 79 was performed using the snakePipes mRNA -seq pipeline, \nadding the flag \"--rMATS\". Splicing events were classified as neuronal if, comparing neuronal and non-neuronal \ncell populations, the following applied: FDR<0.01 and abs(IncLevelDifference)>1. Parameters for the definition \nof ELAV-dependent splicing were FDR<0.01 and abs(IncLevelDifference)>0.1.  \nCalculation of splicing efficiencies from nascent RNA-seq data. \nTo calculate splicing efficiencies, we used previously published nascent RNA -seq data and a modified \nquantification method 46. First, we generated a reference annotation by selecting the last nucleotides of each \nintron, i.e., the region covering t he last (3') 30% of each intron’s length. For the downstream exon, the \nquantification used the nucleotides constituting the first (5') 30% of the exon’s length. Reads aligned to either \nthe intron or the exon were counted using the summarizeOverlaps function from the GenomicAlignments \npackage 80. These reads were then normalized using DEXseq 81. To quantify splicing efficiencies, we calculated \nthe ratio of reads in introns to reads in exons and subtracted this value from 1. Values closer to 1 indicate a \nhigher splicing efficiency (“faster splicing”), as they represent a lower amount of intron reads relative to exon \nreads. Values closer to 0 indicate lower splicing efficiencies (“slower splicing”). The methodology and code for \nthis analysis are available at: https://github.com/hilgers-lab/SpliceFlow. \nPrediction of RCMS within flanking intron sequences  \nTo identify RCMs, we used the method described in Ivanov et al., 2015 24. The procedure involves a BLAST \nalignment for each pair of introns flanking a circRNA to identify all the potential candidates. The parameters we \nused for the alignments were \"-strand minus -word_size 7 -outfmt 0\". RCMs of more than 20 nt in length were \nused for downstream analysis.  \n  \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted June 22, 2024. ; https://doi.org/10.1101/2024.06.22.600180doi: bioRxiv preprint \n\n 23 \nReferences \n \n1. Liu, C.-X., and Chen, L.-L. (2022). Circular RNAs: Characterization, cellular roles, and applications. \nCell 185, 2016-2034. 10.1016/j.cell.2022.04.021. \n2. Salzman, J., Gawad, C., Wang, P.L., Lacayo, N., and Brown, P.O. (2012). Circular RNAs are the \npredominant transcript isoform from hundreds of human genes in diverse cell types. PLoS One 7, \ne30733. 10.1371/journal.pone.0030733. \n3. Memczak, S., Jens, M., Elefsinioti, A., Torti, F., Krueger, J., Rybak, A., Maier, L., Mackowiak, S.D., \nGregersen, L.H., Munschauer, M., et al. (2013). Circular RNAs are a large class of animal RNAs \nwith regulatory potency. Nature 495, 333-338. 10.1038/nature11928. \n4. Wang, P.L., Bao, Y., Yee, M.C., Barrett, S.P., Hogan, G.J., Olsen, M.N., Dinneny, J.R., Brown, \nP.O., and Salzman, J. (2014). Circular RNA is expressed across the eukaryotic tree of life. PLoS \nOne 9, e90859. 10.1371/journal.pone.0090859. \n5. Patop, I.L., Wust, S., and Kadener, S. (2019). Past, present, and future of circRNAs. EMBO J 38, \ne100836. 10.15252/embj.2018100836. \n6. You, X., Vlatkovic, I., Babic, A., Will, T., Epstein, I., Tushev, G., Akbalik, G., Wang, M., Glock, C., \nQuedenau, C., et al. (2015). Neural circular RNAs are derived from synaptic genes and regulated \nby development and plasticity. Nature neuroscience 18, 603-610. 10.1038/nn.3975. \n7. Piwecka, M., Glažar, P., Hernandez-Miranda, L.R., Memczak, S., Wolf, S.A., Rybak-Wolf, A., \nFilipchyk, A., Klironomos, F., Cerda Jara, C.A., Fenske, P., et al. (2017). Loss of a mammalian \ncircular RNA locus causes miRNA deregulation and affects brain function. Science 357. \n10.1126/science.aam8526. \n8. Suenkel, C., Cavalli, D., Massalini, S., Calegari, F., and Rajewsky, N. (2020). A Highly Conserved \nCircular RNA Is Required to Keep Neural Cells in a Progenitor State in the Mammalian Brain. Cell \nRep 30, 2170-2179 e2175. 10.1016/j.celrep.2020.01.083. \n9. Gruner, H., Cortés-López, M., Cooper, D.A., Bauer, M., and Miura, P. (2016). CircRNA \naccumulation in the aging mouse brain. Sci Rep 6, 38907. 10.1038/srep38907. \n10. Morandell, J., Monziani, A., Lazioli, M., Donzel, D., Döring, J., Oss Pegorar, C., D’Anzi, A., \nPellegrini, M., Mattiello, A., Bortolotti, D., et al. (2024). CircHTT(2,3,4,5,6) — co-evolving with the \nHTT CAG-repeat tract — modulates Huntington's disease phenotypes. Molecular Therapy - Nucleic \nAcids 35. 10.1016/j.omtn.2024.102234. \n11. Zimmerman, A.J., Hafez, A.K., Amoah, S.K., Rodriguez, B.A., Dell'Orco, M., Lozano, E., Hartley, \nB.J., Alural, B., Lalonde, J., Chander, P., et al. (2020). A psychiatric disease-related circular RNA \ncontrols synaptic gene expression and cognition. Mol Psychiatry 25, 2712-2727. 10.1038/s41380-\n020-0653-4. \n12. Rybak-Wolf, A., Stottmeister, C., Glažar, P., Jens, M., Pino, N., Giusti, S., Hanan, M., Behm, M., \nBartok, O., Ashwal-Fluss, R., et al. (2015). Circular RNAs in the Mammalian Brain Are Highly \nAbundant, Conserved, and Dynamically Expressed. Mol Cell 58, 870-885. \n10.1016/j.molcel.2015.03.027. \n13. Westholm, J.O., Miura, P., Olson, S., Shenker, S., Joseph, B., Sanfilippo, P., Celniker, S.E., \nGraveley, B.R., and Lai, E.C. (2014). Genome-wide analysis of drosophila circular RNAs reveals \ntheir structural and sequence properties and age-dependent neural accumulation. Cell reports 9, \n1966-1980. 10.1016/j.celrep.2014.10.062. \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted June 22, 2024. ; https://doi.org/10.1101/2024.06.22.600180doi: bioRxiv preprint \n\n 24 \n14. Dong, X., Bai, Y., Liao, Z., Gritsch, D., Liu, X., Wang, T., Borges-Monroy, R., Ehrlich, A., Serrano, \nG.E., Feany, M.B., et al. (2023). Circular RNAs in the human brain are tailored to neuron identity \nand neuropsychiatric disease. Nature communications 14, 5327. 10.1038/s41467-023-40348-0. \n15. Dube, U., Del-Aguila, J.L., Li, Z., Budde, J.P., Jiang, S., Hsu, S., Ibanez, L., Fernandez, M.V., \nFarias, F., Norton, J., et al. (2019). An atlas of cortical circular RNA expression in Alzheimer \ndisease brains demonstrates clinical and pathological associations. Nat Neurosci 22, 1903-1912. \n10.1038/s41593-019-0501-5. \n16. Pamudurti, N.R., Patop, I.L., Krishnamoorthy, A., Bartok, O., Maya, R., Lerner, N., Ashwall-Fluss, \nR., Konakondla, J.V.V., Beatus, T., and Kadener, S. (2022). circMbl functions in cis and in trans to \nregulate gene expression and physiology in a tissue-specific fashion. Cell Rep 39, 110740. \n10.1016/j.celrep.2022.110740. \n17. Hansen, T.B., Jensen, T.I., Clausen, B.H., Bramsen, J.B., Finsen, B., Damgaard, C.K., and Kjems, \nJ. (2013). Natural RNA circles function as efficient microRNA sponges. Nature 495, 384-388. \n10.1038/nature11993. \n18. Ashwal-Fluss, R., Meyer, M., Pamudurti, N.R., Ivanov, A., Bartok, O., Hanan, M., Evantal, N., \nMemczak, S., Rajewsky, N., and Kadener, S. (2014). circRNA biogenesis competes with pre-\nmRNA splicing. Mol Cell 56, 55-66. 10.1016/j.molcel.2014.08.019. \n19. Weigelt, C.M., Sehgal, R., Tain, L.S., Cheng, J., Eßer, J., Pahl, A., Dieterich, C., Grönke, S., and \nPartridge, L. (2020). An Insulin-Sensitive Circular RNA that Regulates Lifespan in Drosophila. \nMolecular cell 79, 268-279.e265. 10.1016/j.molcel.2020.06.011. \n20. van Heesch, S., Witte, F., Schneider-Lunitz, V., Schulz, J.F., Adami, E., Faber, A.B., Kirchner, M., \nMaatz, H., Blachut, S., Sandmann, C.L., et al. (2019). The Translational Landscape of the Human \nHeart. Cell 178, 242-260 e229. 10.1016/j.cell.2019.05.010. \n21. Chen, L.L., and Yang, L. (2015). Regulation of circRNA biogenesis. RNA Biol 12, 381-388. \n10.1080/15476286.2015.1020271. \n22. Kristensen, L.S., Andersen, M.S., Stagsted, L.V.W., Ebbesen, K.K., Hansen, T.B., and Kjems, J. \n(2019). The biogenesis, biology and characterization of circular RNAs. Nature Reviews Genetics \n20, 675-691. 10.1038/s41576-019-0158-7. \n23. Zhang, X.O., Wang, H.B., Zhang, Y., Lu, X., Chen, L.L., and Yang, L. (2014). Complementary \nsequence-mediated exon circularization. Cell 159, 134-147. 10.1016/j.cell.2014.09.001. \n24. Ivanov, A., Memczak, S., Wyler, E., Torti, F., Porath, Hagit T., Orejuela, Marta R., Piechotta, M., \nLevanon, Erez Y., Landthaler, M., Dieterich, C., and Rajewsky, N. (2015). Analysis of Intron \nSequences Reveals Hallmarks of Circular RNA Biogenesis in Animals. Cell Reports 10, 170-177. \nhttps://doi.org/10.1016/j.celrep.2014.12.019. \n25. Li, X., Liu, S., Zhang, L., Issaian, A., Hill, R.C., Espinosa, S., Shi, S., Cui, Y., Kappel, K., Das, R., et \nal. (2019). A unified mechanism for intron and exon definition and back-splicing. Nature 573, 375-\n380. 10.1038/s41586-019-1523-6. \n26. Conn, Simon J., Pillman, Katherine A., Toubia, J., Conn, Vanessa M., Salmanidis, M., Phillips, \nCaroline A., Roslan, S., Schreiber, Andreas W., Gregory, Philip A., and Goodall, Gregory J. (2015). \nThe RNA Binding Protein Quaking Regulates Formation of circRNAs. Cell 160, 1125-1134. \nhttps://doi.org/10.1016/j.cell.2015.02.014. \n27. Errichelli, L., Dini Modigliani, S., Laneve, P., Colantoni, A., Legnini, I., Capauto, D., Rosa, A., De \nSantis, R., Scarfò, R., Peruzzi, G., et al. (2017). FUS affects circular RNA expression in murine \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted June 22, 2024. ; https://doi.org/10.1101/2024.06.22.600180doi: bioRxiv preprint \n\n 25 \nembryonic stem cell-derived motor neurons. Nature communications 8, 14741. \n10.1038/ncomms14741. \n28. Knupp, D., Cooper, D.A., Saito, Y., Darnell, R.B., and Miura, P. (2021). NOVA2 regulates neural \ncircRNA biogenesis. Nucleic Acids Res 49, 6849-6862. 10.1093/nar/gkab523. \n29. Hilgers, V. (2022). Regulation of neuronal RNA signatures by ELAV/Hu proteins. Wiley Interdiscip \nRev RNA, e1733. 10.1002/wrna.1733. \n30. Yao, K.M., Samson, M.L., Reeves, R., and White, K. (1993). Gene elav of Drosophila \nmelanogaster: a prototype for neuronal-specific RNA binding protein gene family that is conserved \nin flies and humans. J Neurobiol 24, 723-739. 10.1002/neu.480240604. \n31. Pascale, A., Amadio, M., and Quattrone, A. (2008). Defining a neuron: neuronal ELAV proteins. \nCell Mol Life Sci 65, 128-140. 10.1007/s00018-007-7017-y. \n32. Carrasco, J., Rauer, M., Hummel, B., Grzejda, D., Alfonso-Gonzalez, C., Lee, Y., Wang, Q., \nPuchalska, M., Mittler, G., and Hilgers, V. (2020). ELAV and FNE Determine Neuronal Transcript \nSignatures through EXon-Activated Rescue. Mol Cell 80, 156-163.e156. \n10.1016/j.molcel.2020.09.011. \n33. Wei, L., Lee, S., Majumdar, S., Zhang, B., Sanfilippo, P., Joseph, B., Miura, P., Soller, M., and Lai, \nE.C. (2020). Overlapping Activities of ELAV/Hu Family RNA Binding Proteins Specify the Extended \nNeuronal 3' UTR Landscape in Drosophila. Mol Cell 80, 140-155.e146. \n10.1016/j.molcel.2020.09.007. \n34. Lee, S., Wei, L., Zhang, B., Goering, R., Majumdar, S., Wen, J., Taliaferro, J.M., and Lai, E.C. \n(2021). ELAV/Hu RNA binding proteins determine multiple programs of neural alternative splicing. \nPLoS Genet 17, e1009439. 10.1371/journal.pgen.1009439. \n35. Carrasco, J., Mateos, F., and Hilgers, V. (2022). A critical developmental window for ELAV/Hu-\ndependent mRNA signatures at the onset of neuronal differentiation. Cell Reports 41. \n10.1016/j.celrep.2022.111542. \n36. Gao, Y., Zhang, J., and Zhao, F. (2018). Circular RNA identification based on multiple seed \nmatching. Brief Bioinform 19, 803-810. 10.1093/bib/bbx014. \n37. Vromman, M., Anckaert, J., Bortoluzzi, S., Buratin, A., Chen, C.Y., Chu, Q., Chuang, T.J., \nDehghannasiri, R., Dieterich, C., Dong, X., et al. (2023). Large-scale benchmarking of circRNA \ndetection tools reveals large differences in sensitivity but not in precision. Nat Methods 20, 1159-\n1169. 10.1038/s41592-023-01944-6. \n38. Khan, M.A., Reckman, Y.J., Aufiero, S., van den Hoogenhof, M.M., van der Made, I., Beqqali, A., \nKoolbergen, D.R., Rasmussen, T.B., van der Velden, J., Creemers, E.E., and Pinto, Y.M. (2016). \nRBM20 Regulates Circular RNA Production From the Titin Gene. Circ Res 119, 996-1003. \n10.1161/circresaha.116.309568. \n39. Legnini, I., Di Timoteo, G., Rossi, F., Morlando, M., Briganti, F., Sthandier, O., Fatica, A., Santini, \nT., Andronache, A., Wade, M., et al. (2017). Circ-ZNF609 Is a Circular RNA that Can Be Translated \nand Functions in Myogenesis. Mol Cell 66, 22-37 e29. 10.1016/j.molcel.2017.02.017. \n40. Cherubini, A., Barilani, M., Rossi, R.L., Jalal, M.M.K., Rusconi, F., Buono, G., Ragni, E., Cantarella, \nG., Simpson, H., Peault, B., and Lazzari, L. (2019). FOXP1 circular RNA sustains mesenchymal \nstem cell identity via microRNA inhibition. Nucleic Acids Res 47, 5325-5340. 10.1093/nar/gkz199. \n41. Zheng, H., Huang, S., Wei, G., Sun, Y., Li, C., Si, X., Chen, Y., Tang, Z., Li, X., Chen, Y., et al. \n(2022). CircRNA Samd4 induces cardiac repair after myocardial infarction by blocking \nmitochondria-derived ROS output. Mol Ther 30, 3477-3498. 10.1016/j.ymthe.2022.06.016. \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted June 22, 2024. ; https://doi.org/10.1101/2024.06.22.600180doi: bioRxiv preprint \n\n 26 \n42. Torres-Méndez, A., Pop, S., Bonnal, S., Almudi, I., Avola, A., Roberts, R.J.V., Paolantoni, C., \nAlcaina-Caro, A., Martín-Anduaga, A., Haussmann, I.U., et al. (2022). Parallel evolution of a \nsplicing program controlling neuronal excitability in flies and mammals. Science Advances 8, \neabk0445. 10.1126/sciadv.abk0445. \n43. Gaidatzis, D., Burger, L., Florescu, M., and Stadler, M.B. (2015). Analysis of intronic and exonic \nreads in RNA-seq data characterizes transcriptional and post-transcriptional regulation. Nat \nBiotechnol 33, 722-729. 10.1038/nbt.3269. \n44. Xu, C., and Zhang, J. (2021). Mammalian circular RNAs result largely from splicing errors. Cell \nReports 36, 109439. https://doi.org/10.1016/j.celrep.2021.109439. \n45. Menet, J.S., Rodriguez, J., Abruzzi, K.C., and Rosbash, M. (2012). Nascent-Seq reveals novel \nfeatures of mouse circadian transcriptional regulation. Elife 1, e00011. 10.7554/eLife.00011. \n46. Khodor, Y.L., Menet, J.S., Tolan, M., and Rosbash, M. (2012). Cotranscriptional splicing efficiency \ndiffers dramatically between Drosophila and mouse. RNA 18, 2174-2186. 10.1261/rna.034090.112. \n47. Pai, A.A., Henriques, T., McCue, K., Burkholder, A., Adelman, K., and Burge, C.B. (2017). The \nkinetics of pre-mRNA splicing in the Drosophila genome and the influence of gene architecture. \nElife 6. 10.7554/eLife.32537. \n48. Liang, D., and Wilusz, J.E. (2014). Short intronic repeat sequences facilitate circular RNA \nproduction. Genes Dev 28, 2233-2247. 10.1101/gad.251926.114. \n49. Salzman, J., Chen, R.E., Olsen, M.N., Wang, P.L., and Brown, P.O. (2013). Cell-type specific \nfeatures of circular RNA expression. PLoS genetics 9, e1003777-e1003777. \n10.1371/journal.pgen.1003777. \n50. Wilusz, J.E. (2017). Circular RNAs: Unexpected outputs of many protein-coding genes. RNA Biol \n14, 1007-1017. 10.1080/15476286.2016.1227905. \n51. Zeng, Y., Fair, B.J., Zeng, H., Krishnamohan, A., Hou, Y., Hall, J.M., Ruthenburg, A.J., Li, Y.I., and \nStaley, J.P. (2022). Profiling lariat intermediates reveals genetic determinants of early and late co-\ntranscriptional splicing. Mol Cell 82, 4681-4699 e4688. 10.1016/j.molcel.2022.11.004. \n52. Sorek, R., Ast, G., and Graur, D. (2002). Alu-containing exons are alternatively spliced. Genome \nRes 12, 1060-1067. 10.1101/gr.229302. \n53. Liang, L., Cao, C., Ji, L., Cai, Z., Wang, D., Ye, R., Chen, J., Yu, X., Zhou, J., Bai, Z., et al. (2023). \nComplementary Alu sequences mediate enhancer–promoter selectivity. Nature. 10.1038/s41586-\n023-06323-x. \n54. Aktaş, T., Avşar Ilık, İ., Maticzka, D., Bhardwaj, V., Pessoa Rodrigues, C., Mittler, G., Manke, T., \nBackofen, R., and Akhtar, A. (2017). DHX9 suppresses RNA processing defects originating from \nthe Alu invasion of the human genome. Nature 544, 115-119. 10.1038/nature21715 \nhttp://www.nature.com/nature/journal/v544/n7648/abs/nature21715.html#supplementary-information. \n55. Ji, P., Wu, W., Chen, S., Zheng, Y., Zhou, L., Zhang, J., Cheng, H., Yan, J., Zhang, S., Yang, P., \nand Zhao, F. (2019). Expanded Expression Landscape and Prioritization of Circular RNAs in \nMammals. Cell Rep 26, 3444-3460 e3445. 10.1016/j.celrep.2019.02.078. \n56. Wu, W., Ji, P., and Zhao, F. (2020). CircAtlas: an integrated resource of one million highly accurate \ncircular RNAs from 1070 vertebrate transcriptomes. Genome Biology 21. 10.1186/s13059-020-\n02018-y. \n57. Wei, L., and Lai, E.C. (2022). Regulation of the Alternative Neural Transcriptome by ELAV/Hu RNA \nBinding Proteins. Front Genet 13, 848626. 10.3389/fgene.2022.848626. \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted June 22, 2024. ; https://doi.org/10.1101/2024.06.22.600180doi: bioRxiv preprint \n\n 27 \n58. Mulligan, M.R., and Bicknell, L.S. (2023). The molecular genetics of nELAVL in brain development \nand disease. European Journal of Human Genetics 31, 1209-1217. 10.1038/s41431-023-01456-z. \n59. Dell'Orco, M., Oliver, R.J., and Perrone-Bizzozero, N. (2020). HuD Binds to and Regulates Circular \nRNAs Derived From Neuronal Development- and Synaptic Plasticity-Associated Genes. Front \nGenet 11, 790. 10.3389/fgene.2020.00790. \n60. Cao, S.M., Wu, H., Yuan, G.H., Pan, Y.H., Zhang, J., Liu, Y.X., Li, S., Xu, Y.F., Wei, M.Y., Yang, L., \nand Chen, L.L. (2024). Altered nucleocytoplasmic export of adenosine-rich circRNAs by PABPC1 \ncontributes to neuronal function. Mol Cell. 10.1016/j.molcel.2024.05.011. \n61. Prashad, S., and Gopal, P.P. (2021). RNA-binding proteins in neurological development and \ndisease. RNA Biol 18, 972-987. 10.1080/15476286.2020.1809186. \n62. Sun, C., and Schuman, E. (2023). A multi-omics view of neuronal subcellular protein synthesis. \nCurr Opin Neurobiol 80, 102705. 10.1016/j.conb.2023.102705. \n63. Zaharieva, E., Haussmann, I.U., Brauer, U., and Soller, M. (2015). Concentration and Localization \nof Coexpressed ELAV/Hu Proteins Control Specificity of mRNA Processing. Mol Cell Biol 35, 3104-\n3115. 10.1128/MCB.00473-15. \n64. Port, F., and Bullock, S.L. (2016). Augmenting CRISPR applications in Drosophila with tRNA-\nflanked sgRNAs. Nat Methods 13, 852-854. 10.1038/nmeth.3972. \n65. Grzejda, D., Mach, J., Schweizer, J.A., Hummel, B., Rezansoff, A.M., Eggenhofer, F., Panhale, A., \nLalioti, M.E., Cabezas Wallscheid, N., Backofen, R., et al. (2022). The long noncoding RNA mimi \nscaffolds neuronal granules to maintain nervous system maturity. Science Advances 8, eabo5578. \n10.1126/sciadv.abo5578. \n66. McCorkindale, A.L., Wahle, P., Werner, S., Jungreis, I., Menzel, P., Shukla, C.J., Abreu, R.L.P., \nIrizarry, R.A., Meyer, I.M., Kellis, M., and Zinzen, R.P. (2019). A gene expression atlas of \nembryonic neurogenesis in Drosophila reveals complex spatiotemporal regulation of lncRNAs. \nDevelopment 146. 10.1242/dev.175265. \n67. Hilgers, V., Lemke, S.B., and Levine, M. (2012). ELAV mediates 3 ' UTR extension in the \nDrosophila nervous system. Genes & Development 26, 2259-2264. 10.1101/gad.199653.112. \n68. Bhardwaj, V., Heyne, S., Sikora, K., Rabbani, L., Rauer, M., Kilpert, F., Richter, A.S., Ryan, D.P., \nand Manke, T. (2019). snakePipes: facilitating flexible, scalable and integrative epigenomic \nanalysis. Bioinformatics 35, 4757-4759. 10.1093/bioinformatics/btz436. \n69. Dobin, A., Davis, C.A., Schlesinger, F., Drenkow, J., Zaleski, C., Jha, S., Batut, P., Chaisson, M., \nand Gingeras, T.R. (2013). STAR: ultrafast universal RNA-seq aligner. Bioinformatics 29, 15-21. \n10.1093/bioinformatics/bts635. \n70. Andrews, S. (2010). FastQC: A quality control tool for high throughput  sequence data. Available: \nhttp://www.bioinformatics.babraham.ac.uk. \n71. Love, M.I., Huber, W., and Anders, S. (2014). Moderated estimation of fold change and dispersion \nfor RNA-seq data with DESeq2. Genome Biol 15, 550. 10.1186/s13059-014-0550-8. \n72. Li, H., and Durbin, R. (2009). Fast and accurate short read alignment with Burrows-Wheeler \ntransform. Bioinformatics 25, 1754-1760. 10.1093/bioinformatics/btp324. \n73. Shen, W., Le, S., Li, Y., and Hu, F. (2016). SeqKit: A Cross-Platform and Ultrafast Toolkit for \nFASTA/Q File Manipulation. PLoS One 11, e0163962. 10.1371/journal.pone.0163962. \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted June 22, 2024. ; https://doi.org/10.1101/2024.06.22.600180doi: bioRxiv preprint \n\n 28 \n74. Zhang, J., Chen, S., Yang, J., and Zhao, F. (2020). Accurate quantification of circular RNAs \nidentifies extensive circular isoform switching events. Nature communications 11, 90. \n10.1038/s41467-019-13840-9. \n75. McLeay, R.C., and Bailey, T.L. (2010). Motif Enrichment Analysis: a unified framework and an \nevaluation on ChIP data. BMC Bioinformatics 11, 165. 10.1186/1471-2105-11-165. \n76. Bailey, T.L., and Machanick, P. (2012). Inferring direct DNA binding from ChIP-seq. Nucleic Acids \nRes 40, e128. 10.1093/nar/gks433. \n77. Liao, Y., Smyth, G.K., and Shi, W. (2014). featureCounts: an efficient general purpose program for \nassigning sequence reads to genomic features. Bioinformatics 30, 923-930. \n10.1093/bioinformatics/btt656. \n78. Curk, T. (2019). iCount: protein-RNA interaction iCLIP data analysis (in preparation). \n79. Shen, S., Park, J.W., Lu, Z.X., Lin, L., Henry, M.D., Wu, Y.N., Zhou, Q., and Xing, Y. (2014). \nrMATS: robust and flexible detection of differential alternative splicing from replicate RNA-Seq \ndata. Proc Natl Acad Sci U S A 111, E5593-5601. 10.1073/pnas.1419161111. \n80. Lawrence, M., Huber, W., Pages, H., Aboyoun, P., Carlson, M., Gentleman, R., Morgan, M.T., and \nCarey, V.J. (2013). Software for computing and annotating genomic ranges. PLoS computational \nbiology 9, e1003118. 10.1371/journal.pcbi.1003118. \n81. Anders, S., Reyes, A., and Huber, W. (2012). Detecting differential usage of exons from RNA-seq \ndata. Genome Res 22, 2008-2017. 10.1101/gr.133744.111. \n \n.CC-BY 4.0 International licenseavailable under a \n(which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made \nThe copyright holder for this preprintthis version posted June 22, 2024. ; https://doi.org/10.1101/2024.06.22.600180doi: bioRxiv preprint","source_license":"CC-BY-4.0","license_restricted":false}