Pumilio differentially binds to mRNA 3’ UTR isoforms to regulate localization of synaptic proteins

In neuronal cells, the regulation of RNA is crucial for the spatiotemporal control of gene expression, but how the correct localization, levels, and function of synaptic proteins are achieved is not well understood. In this study, we globally investigate the role of alternative 3’ UTRs in regulating RNA localization in the synaptic regions of the Drosophila brain. We identify direct mRNA targets of the translational repressor Pumilio, finding that mRNAs bound by Pumilio encode proteins enriched in synaptosomes. Pumilio differentially binds to RNA isoforms of the same gene, favoring long, neuronal 3’ UTRs. These longer 3’ UTRs tend to remain in the neuronal soma, whereas shorter UTR isoforms localize to the synapse. In cultured pumilio mutant neurons, severe axon outgrowth defects were accompanied by mRNA isoform mislocalization, and proteins encoded by these Pumilio targets displayed excessive abundance at synaptic boutons. Our study identifies an important and widespread mechanism for the spatiotemporal regulation of protein function in neurons.


Introduction
In neurons, the localization of mRNA molecules within specific subcellular compartments, and their local translation, is essential for the spatial regulation of gene expression, and for a rapid and efficient stimulus response (Sun & Schuman, 2023).Neuron-specific RNA-binding proteins (RBPs) interact with sequence or structure elements, usually located in the mRNA 3' untranslated region (3' UTR), to regulate protein function at multiple levels, including transcript localization, translation, and degradation (Bourke et al, 2023;Mayr, 2017;Mitschka & Mayr, 2022).Most metazoan genes express mRNAs with several 3' UTR variants, generated by the alternative use of polyadenylation (poly(A)) sites (Gruber & Zavolan, 2019).
In animal neurons, the highly conserved RBP Embryonic Lethal Abnormal Vision (ELAV) promotes the synthesis of longer, neuron-specific 3' UTRs in hundreds of genes (Carrasco et al, 2020;Hilgers et al, 2012;Wei et al, 2020).By providing an additional platform for regulatory factors to bind, these sequences are thought to modify the physiological function of the encoded protein by altering the localization, stability, and plasticity of mRNAs isoforms in neuronal compartments (Bourke et al., 2023;Tushev et al, 2018).The differential usage of 3' UTR in neuronal projections vs. cell bodies is widespread (Arora et al, 2021;Bauer et al, 2019;Ciolli Mattioli et al, 2019;Goering et al, 2023;Mendonsa et al, 2023;Taliaferro et al, 2016), and functions for distinct 3' UTR variants have been reported for individual genes.For example, the dendritic targeting of the long 3' UTR isoform of brain-derived neurotrophic factor (BDNF) is essential for proper dendritic development and long-term potentiation (An et al, 2008), whereas in sympathetic neurons, shortening of the Inositol Monophosphatase 1 (IMPA1) 3' UTR by cleavage is required for maintaining axon integrity (Andreassi et al, 2021).
Binding motifs for the RBP Pumilio (Pum) were found enriched in the neuron-specific 3' UTR sequences of Drosophila mRNAs (Hilgers et al, 2011;Sanfilippo et al, 2017), raising the possibility that specific RBPs are involved in the global, isoform-dependent modulation of mRNA behavior in neurons.Pumilio proteins are highly conserved members of the PUF (pumilio and fem-3 mRNA-binding factor) family (Goldstrohm et al, 2018).The gene was discovered in Drosophila melanogaster, where, in addition to roles in embryonic patterning (Lehmann & Nüsslein-Volhard, 1987), Pum regulates stem cell selfrenewal (Gilboa & Lehmann, 2004;Wang & Lin, 2004), synaptogenesis, neuronal excitability, synaptic excitability and plasticity, and circadian rhythms (Menon et al, 2009;Muraro et al, 2008;Schweers et al, 2002).In mammals, Pum 1 and 2 are involved in a variety of important nervous-system related processes such as neurogenesis, axon guidance, membrane excitability and synaptic plasticity (Dong et al, 2018;Driscoll et al, 2013;Vessey et al, 2010).Molecularly, Pumilio proteins regulate gene expression at the post-transcriptional level by binding to specific sequences in the 3' UTR of target mRNAs (Jarmoskaite et al, 2019;Zhang et al, 2017).One well-described consequence of Pum binding is the translational repression of target mRNAs, often in conjunction with transport and localized protein synthesis (Goldstrohm et al., 2018;Nishanth & Simon, 2020).
Here, we show that 3' UTR diversity regulates the expression of protein-coding genes in synaptic regions of the Drosophila brain.We biochemically isolate synaptosomes -structures that contain presynaptic terminals, synaptic vesicles, and often a portion of the postsynaptic membrane, and find that hundreds of mRNAs express 3' UTR variants that are differentially localized in neuronal subcompartments.We identify direct mRNA targets of Pum in head tissue and find that the role of Pum family proteins in binding synaptic mRNAs is conserved, with many common targets in flies and mammals.Pum binds to long, neuron-specific 3' UTRs to promote mRNA localization and regulate the expression of the encoded protein in synaptic compartments.Our results reveal a mechanism of 3' UTR-dependent regulation, demonstrating how neurons can achieve high spatial protein complexity with a restricted set of genes.

Drosophila Pumilio directly binds mRNAs encoding synaptic proteins
We undertook a systematic characterization of mRNAs regulated by Pum binding in the adult Drosophila brain.We C-terminally tagged the endogenous pumilio locus with FLAG-HA (Fig. S1A, B).Head tissue of the thus-generated pum FLAG flies was used to perform RNA immunopurification with UV-crosslinking (xRIP) and recover RNAs directly bound by Pum in vivo (Fig. 1A).We also performed FLAG IP on flies that do not express an endogenous tag, in order to control for unspecific binding.We eluted the Pumbound RNAs and sequenced them by 3'-end sequencing (3'-seq), a method that allows for quantitative measurements of mRNAs with distinct 3'-end isoforms.3'-seq reads were mapped, filtered for aberrant 3'-ends, clustered and quantified at the 3'-end cluster level (Fig. 1A).
We identified 460 genes whose mRNAs were highly and specifically enriched in the Pum IP compared to the input sample and to the control IP (Fig. 1B, C, Fig. S1C, Table S1); we hereafter refer to these transcripts as "Pum targets".Consistent with a highly efficient and specific Pum xRIP, the reported RNAbinding motif for Drosophila Pum (Pumilio Response Element (PRE) (Ray et al, 2013)) constituted the top-enriched motif in 3' UTR sequences of Pum-bound mRNAs (Fig. S1D, Table S1), and we recovered the mRNA encoding the sodium channel paralytic (para), a validated functional and physical interactor of Pum in the nervous system (Muraro et al., 2008).Direct Pum targets include numerous genes with a reported role in synaptic transmission (e.g., Dopamine 1-like receptor 1 (Dop1R1), Synaptotagmin 1 (Syt1), Vesicular acetylcholine transporter (VAChT), nicotinic acetylcholine receptor subunits (nAChRα1, nAChRβ1 and nAChRα6), memory (e.g., klingon (klg), orb2, dunce (dnc), rutabaga (rut)), and genes involved in phenotypes of neuronal excitability (e.g., shaker cognate w (Shaw), para); gene ontology terms for Pum mRNA targets were highly enriched in synaptic processes -most prominently, chemical synaptic transmission (Fig. 1D, Table S1).Interestingly, a substantial fraction of Pum targets represent homologues of genes bound by mouse homologues Pum 1 and/or Pum 2 in neonatal mouse brains (Zhang et al., 2017), and many of those mRNAs were bound by both mouse Pumilios (Fig. 1E, Fig. S1E).This group, which also included Pum itself, contained genes involved in the positive regulation of transcription (Fig. 1F, Table S2).This is consistent with Pum's described role in modulating nuclear effectors including transcription factors (Bohn et al, 2018;Elguindy & Mendell, 2021;Goldstrohm et al., 2018;Wreden et al, 1997).In addition, Gene Ontology (GO) terms were enriched for specialized synaptic functions such as learning and memory, and chemical synaptic transmission (Fig. 1F, Table S2).In conclusion, our data reveal the molecular targets of Pum in the Drosophila brain, and strongly suggest a conserved role for Pum in the translational regulation and/or localization of mRNAs important for synaptic function.

Proteins of the Drosophila synaptosome are encoded by Pumilio targets
Next, we aimed to assess whether Pum was involved in the synaptic localization of its target mRNAs and/or the proteins they encode.Several studies have separated neuronal compartments and assessed differential localization of mRNAs in neurites or synaptic regions of mammalian brains (Cajigas et al, 2012;Tushev et al., 2018) or cultured neurons (Gumy et al, 2011;Mendonsa et al., 2023;Taliaferro et al., 2016;Zappulo et al, 2017); however such experiments have not been performed in Drosophila.To determine the subcellular distribution of neuronal mRNAs and proteins, we prepared synaptosomes from wild-type adult fly heads.We modified the protocol for biochemical fractionation of neuronal cell compartments from frozen fly head tissue (Depner et al, 2014).Crucially, to preserve RNA/protein complexes, and to exclude content from nuclei -which, in our hands, burst upon homogenization of thawed material-, we employed freshly collected, never-frozen, fly heads.We used homogenization and differential centrifugation with the aim of isolating a fraction enriched in presynaptic and postsynaptic components (Fig. 2A).By transmission electron microscopy, we verified the isolation of intact synaptosomes composed of pre-synaptic membranes enclosing synaptic vesicles and mitochondria, and adjacent to postsynaptic densities (Fig. 2B).Western Blot analysis confirmed the enrichment of synaptic markers and the depletion of cytoplasmic and nuclear markers (Fig. 2C).Moreover, intronic regions of mRNAs, hallmarks of nuclear pre-mRNAs, were severely depleted in synaptosome fractions (Fig. S2A).We performed a proteomics analysis of Drosophila synaptosomes as well as crude head homogenate (input) using shotgun mass spectrometry (Table S3).Gene Ontology analysis of 989 proteins enriched in synaptosomes showed that our fraction contained a substantial number of synaptic cellular components, with a strong enrichment of highly specific biological processes related to synaptic function, including chemical synaptic transmission and neurotransmitter secretion (Fig. 2D, Table S3).
Together, these results show that our neuronal fractionation protocol yields synaptosomes highly enriched in soluble and membrane-bound synaptic proteins.
Strikingly, almost half of all Pum target mRNAs whose encoded protein was detected by mass spectrometry, encode a synaptosome-enriched protein (Fig. 2E, Fig. S2B).Unlike the bulk of Pum targets, many of which encode proteins involved in transcriptional regulation, this "synaptic" subset of Pum targets was exclusively enriched in functions related to synaptic transmission, neurite guidance and complex behaviors (Fig. 2F, compared to Fig. 1D, Table S3).Those mRNAs stood out by their particularly long 5' UTRs and 3' UTRs (over twice as long as non-target synaptosome proteins, Fig. S2C).Longer 3' UTRs confer added potential for post-transcriptional regulation; for example, Pum targets encoding synaptic proteins displayed an accumulation of predicted microRNA binding sites (Table S3).Our results suggest that Pum plays a global and wide-reaching role in the synaptic localization and/or local translation of mRNAs encoding synaptic proteins.

Synaptically localized mRNAs are not Pum targets
By 3'-seq, we identified mRNAs of 211 and 1195 genes significantly enriched and depleted in synaptosome fractions, respectively, compared to head homogenate (padj<0.05;Table S3).The subset of synaptosome-enriched transcripts, although not yielding any significant GO terms (likely due to low detection), contains multiple mRNAs encoding proteins well-known for specialized neuronal and/or synaptic functions, such as Ankyrin 2 (Ank2), nAChRβ3, Longitudinals lacking (Lola) and Neurotrophin 1 (NT1) (Fig. S2D).On the other hand, consistent with mRNAs localizing in the subcellular region in which they are translated, synapse-depleted mRNAs encoded proteins involved in broad cellular functions in cell somata and nuclei, predominantly transcription and chromatin organization (Fig. S2E).
Surprisingly, only seven Pum target mRNAs figured among synaptosome-enriched transcripts, constituting an actual depletion of Pum targets at the synapse; in contrast, Pum targets were enriched among synaptosome-depleted transcripts (Fig. 2G and S2F, G, Table S3).Hence, mRNAs bound by Pum tend to localize in the somatic compartment of neurons, indicating that Pum binding may cause mRNAs to be retained in the soma rather than being transported to the synapse.Although unexpected considering the enrichment of mRNAs encoding synaptic proteins among Pum targets, these results are consistent with findings from cultured mammalian neurons: PREs are overrepresented in the transcriptome of cell bodies compared to axons in mouse, and Pum2 knockdown in rat neurons increased axonal localization of the Pum2 target L1cam and PRE-containing mRNAs (Martínez et al, 2019;Minis et al, 2014).Further considering our finding that among genes encoding synaptosome proteins, Pum targets exhibit a much longer 3' UTR, we hypothesized a conserved role for Pum proteins in the isoform-specific regulation of protein expression: Pum binding to the longer mRNA isoform of synaptic genes would inhibit their translation and retain them in the soma, whereas the short-UTR isoform would be selectively transported to the synaptic compartment where it can be expressed locally (Fig. 2H).

Pum binds to specific 3' UTR isoforms of localized mRNAs
To test this hypothesis, we analyzed the expression of 3'-end variants of the same gene and compared their respective 3' UTR lengths in synaptosomes and input (Fig. 3A).For each gene with at least two expressed 3'-end isoforms (alternatively polyadenylated: APA genes), we assessed the expression of each isoform and calculated an "average 3' UTR length", which we represent as a percentage of the length of the longest 3' UTR.We interpret differences in these expression-weighted 3' UTR lengths between synaptosome and input as differential localization of mRNA isoforms of a given gene (Fig. 3B).
We found 542 genes with significantly localized 3' UTR isoforms (Fig. 3B, C, Table S3).Notably, Pum targets were strongly enriched among the 315 genes with shorter 3' UTRs at the synapse (Fig. 3B, C, D, Fig. S3A): 43 Pum targets displayed a significantly lower 3' UTR length-percentage in the synaptosome fraction, with only 10 showing the opposite trend (Fig. 3E, Table S4).Using Pum xRIP-3'seq data, we assessed binding of Pum to each differentially localized 3' UTR isoform.We found abundant Pum binding at the distal (longer) 3' UTR isoform for most genes, for both classes of localized Pum targets (Fig. 3F, Fig. S3B).Taken together, the integration of Pum xRIP-3'-seq and synaptosome RNA and protein analysis shows that Pum globally binds to long 3' UTR isoforms of synaptic genes; these long mRNAs are specifically de-localized from synaptic regions.Our results provide further evidence for the model that Pum inhibits long 3' UTR isoforms of synaptic genes in the soma, thereby promoting the expression of their short counterpart at the synapse.
Pum-mediated regulation of neuron-specific 3' UTRs mRNA isoforms with longer 3' UTRs, through their additional propensity for binding to effectors -such as Pum-may confer a context-or tissue-specific function to the encoded protein (Mitschka & Mayr, 2022).In light of our findings, we wondered whether Pum may confer additional functions to proteins in a neuron-specific manner.In animals from flies to humans, the inhibition of proximal poly(A) sites specifically in neurons, in favor of more distal ones, generates extended, neuronal 3' UTRs (nUTRs) in hundreds of genes, (Hilgers et al., 2011;Miura et al, 2013;Smibert et al, 2012;Ulitsky et al, 2012).In Drosophila, nUTRs are mediated by the pan-neuronal protein ELAV (Hilgers et al., 2012).nUTRcontaining genes tend to perform specialized neuronal functions, most notably synaptic transmission and complex behaviors (Fig. S4A, Table S4); interestingly, nUTRs are enriched in PREs (Hilgers et al., 2011;Smibert et al., 2012) compared to their short UTR (sUTR) counterpart (Fig. 4A).
Comparing our xRIP-3'-seq data from fly brains, we found that Pum extensively and specifically binds to this particular set of mRNAs: 71 out of 271 (26%) validated nUTR-containing genes (Carrasco et al., 2020) figure among direct targets of Pum (Fig. 4B, Fig. S4B, Table S4).Gene ontology analysis on these 71 genes compared to other neuron-expressed genes revealed chemical synaptic transmission as the only significant term (Fig. 4C, Table S4), indicating that this subset of Pum targets functions in neuronspecific pathways, including synaptic signaling, compared to the global list of targets (compare Fig. 1D and Fig. 4C).Consistent with this idea, nUTR-containing genes were highly enriched among genes with localized 3' UTR isoforms (48 out of 542 mRNAs with differential 3' UTR length in synaptosomes, Fig. S4C), and Pum binds more to the long (nUTR-containing) isoform; specifically, they constitute a substantial subset (33%) of Pum targets for which the short 3' UTR isoform is synaptically localized.
Similarly to other localized mRNA 3' UTR isoforms (Fig. 3), and to an even greater extent, nUTR (long) isoforms are depleted from synaptosomes; Pum specifically binds the somatically-localized isoforms of this subset of genes (Fig. 4D, E, Fig. S4C, D).We propose that by providing a binding platform for Pum, nUTRs promote the localized expression of encoded synaptic proteins.

Impaired neurite outgrowth, mRNA delocalization, and synaptic protein overexpression in Δpum neurons
To functionally test the involvement of Pum in the localized expression of neuronal mRNAs in vivo, we measured molecular and physiological outcomes of mutating pum in flies.We used a combination of two lack-of-function mutations that each abolish Pum RNA-binding activity, pum ET7 and pum ET9 (Forbes & Lehmann, 1998).The resulting Δpum animals are subvital due to severe developmental defects and typically die before eclosion (Forbes & Lehmann, 1998;Menon et al, 2004); therefore, we were unable to perform synaptosome purifications on these mutants.As an alternative, we prepared primary cultures of developing neurons from dissected larval brains.On a coated substrate, the cells elaborate complex neurite arbors over the course of several days in vitro.We grew the cells on coverslips for analysis of neuron morphology; we also employed coated microporous membrane inserts for neurites to extend towards the bottom side of the membrane, which allows for their physical separation from cell bodies (Fig. 5A, Fig. S5A, B).
By five days in vitro, differences between wild-type and Δpum neurons were evident by fluorescent microscopy (Fig. 5B).By day 7, the proportion of ELAV-expressing cells was decreased, indicating that loss of Pum specifically affects neuron survival (Fig. S5C).Moreover, neurite outgrowth was severely reduced, both in terms of neurite length and branching complexity (Fig. 5B, C, Fig. S5D, E).We investigated the subcellular localization of neuronal Pum mRNA targets using qPCR on soma and neurite fractions.In Δpum neurons, all RNAs we measured were shifted towards cell bodies, including two Pum-independent, localized RNAs, Arc1 and mimi (Fig. S5F), which is a likely consequence of the observed loss of neurite structures -and their content-upon Pum loss.In addition, the scarcity of material collectible from Δpum neurites rendered the detection of long 3' UTR isoforms unreliable; these circumstances precluded a direct comparison of 3' UTR isoform localization in neurites vs. soma.
Notably, we did find, for most genes with the largest 3' UTR difference between soma and synaptosome (Fig. 3E), that the relative abundance of long 3' UTR isoforms was disrupted in separated cell bodies of Δpum mutant neurons (Fig. 5D).Finally, we assessed whether the localized, nUTR-dependent Pum binding to target mRNAs affected local expression of the encoded proteins.Antibodies were only available for a subset of them, most of which did not produce reliable localized signal in synaptic boutons of larval neuromuscular junction (NMJs).We did successfully visualize two synaptic proteins, Bruchpilot (Brp) and Syt1 (Synaptotagmin 1).Syt1 is a conserved, nUTR-regulated Pum target localized to cell bodies, with the encoded protein enriched in synaptosomes, while brp, a Pum target encoding a synaptic marker, is present but not enriched in synaptosomes (Tables S1-S4).At NMJs of Δpum larvae, Brp and Syt1 proteins were significantly overexpressed, with Syt1 exhibiting a particularly drastic increase in abundance (Fig. 5E), suggesting that localization and translation of Pum mRNA targets are disrupted in the absence of Pum.Our results show that Pumilio regulates the localized expression of synaptic proteins in an isoform-specific, nUTR-dependent manner, a mechanism essential to preserve neurite integrity and neuron functionality.

Discussion
A more complete understanding of neuron-specific RNA regulation is necessary to fully comprehend the complexity of synaptic function.In our study, we reveal one strategy that neurons use to target synaptic protein function to specific cellular compartments: the 3' UTR-dependent localization of distinct mRNA isoforms of multi-UTR genes, by the RBP Pumilio.We show that in Drosophila brains, the translational repressor binds to distal, often neuron-specific, 3' UTR regions of mRNAs encoding synaptic proteins, and that these long mRNA isoforms tend to localize to cell bodies.The interaction with Pum likely leads to the retention and translational repression of nUTR isoforms of synaptic genes in somata, while short-UTR variants tend to localize to distal neuronal compartments.Although we did not directly demonstrate local translation of short isoforms at the synapse, the scenario is highly consistent with the observed enrichment of Pum target genes among synaptosome-enriched proteins.
The specific localization of mRNAs to distal neuronal compartments is mediated through 3' UTR sequence elements that are recognized and bound by neuronal RBPs including ZBP1 (Ross et al, 1997), FMRP (Goering et al, 2019), Staufen 2 (Bauer et al., 2019), and HBS1L (Mendonsa et al., 2023).Since longer 3' UTRs contain more RBP binding sites and may allow increased interactions with such regulatory factors, this has led to the general view that the additional UTR sequences mediate specialized mRNA transport to neurites and synaptic regions (Mitschka & Mayr, 2022).Individual long 3' UTR isoforms have indeed been reported to be targeted towards neuronal projections (An et al., 2008;Andreassi et al, 2010;Perry et al, 2012).While we find that some Pum targets display a similar regulation (Fig. 3), i.e. localization of the nUTR isoform to synaptosomes, our study suggests that the converse mechanism is more widely used: a specific repression of long isoforms to favor synaptic localization and functionality of their short counterpart.In rat neurons, Pum2 retains specific mRNAs in cell somata, a process important for axonogenesis (Martínez et al., 2019); moreover, mammalian nUTRs are enriched in Pum motifs, and interactions between nUTR genes and Pum proteins are conserved in mouse (Fig. 1), raising the possibility that the nUTR-dependent regulation of mRNA localization by Pum proteins is a conserved feature of neuronal development and function.
Complementing the notion that long 3' UTRs specifically regulate localization and protein output of the mRNA that carries them, our study supports the existence of a "passive" mode of mRNA localization in neurons, in which nUTRs act as a regulatory entity to allow for the proper fate of the short 3' UTR isoform, with Pum sites representing "anti-zipcodes".This goes in line with reports that short 3' UTRs of tandem-UTR genes tend to be neurite-localized in mouse cortical and mESC-derived neurons (Ciolli Mattioli et al., 2019;Taliaferro et al., 2016), and does not contradict findings that 3′ UTR isoforms of localized (to either compartment) transcripts are significantly longer compared to those of non-localized transcripts, although neuropil regions of the rodent brain were reported to contain the longest 3' UTR isoforms (Tushev et al., 2018).mRNAs with longer 3' UTRs are generally less stable, presumably because they tend to harbor more destabilizing motifs such as AU-rich elements (AREs) (Mitschka & Mayr, 2022;Siegel et al, 2022).Indeed, the microRNA let-7 and ARE-binding proteins preferentially destabilize target mRNAs in the soma of neurons, thereby causing an enrichment of mRNAs harboring let-7 motifs in neurites (Mendonsa et al., 2023).Overall, stable mRNAs and mRNA isoforms are enriched in neuronal projections and synaptic regions, where they are efficiently translated (Glock et al, 2021;Loedige et al, 2023;Tushev et al., 2018).Together, this body of evidence suggests that the nUTRmediated destabilization and translational inhibition of long mRNA isoforms by factors such as let-7 and Pum in cell bodies, coupled to the escape of short 3′ UTR isoforms from this repression, accounts for the differential localization of many mRNA isoforms, leading to the compartment-specific function of the encoded protein.
Binding of the long isoform in neuronal cell bodies by Pum may serve cellular functions beyond mRNA regulation.The relative enrichment of Pum binding motifs in cell bodies and distal compartments could help maintain precise levels of localized Pum protein.In flies and mammals, Pum plays important roles in apparently distinct molecular contexts, such as transcription and genome stability -in the nucleusand synaptic function; accordingly, Pum localizes both in cell bodies as well as at distal compartments including neurites and neuromuscular junctions, often within discrete particles (Elguindy & Mendell, 2021;Menon et al., 2004;Vessey et al, 2006).It will be interesting to uncover the direct biological implications of these RNA/protein interactions.

Drosophila melanogaster husbandry and mutagenesis
Experiments in this study used larvae or adult Drosophila melanogaster flies, males and females (in equal amounts).Flies were raised at 25ºC.w 1118 flies were obtained from the Bloomington stock center (5905).The pumilio loss-of-function alleles pum ET7 and pum ET9 were obtained from Ruth Lehmann (Forbes & Lehmann, 1998) and kept in heterozygosis over GFP-marked balancer chromosomes.To generate pum FLAG flies, which express an endogenously, C-terminally FLAG-HA-tagged Pum protein, a guide RNA (acggcaacgttgtgctgtaa) targeted the pum locus at chromosome 3. Genome editing (Port & Bullock, 2016) used a 2602 bp homology donor fragment (sequence listed in Table S5).Embryo injection was performed by Bestgene, Inc.
Cells were dissociated in 200 µL culture medium by pipetting CNSs up-and-down against the bottom of the tube 100 times.The cells were plated on 13 mm plastic coverslips (Sarstedt no.83.1840.002)or microporous membrane inserts (cellQART 931 10 12, pore size 1µm) coated with Concanavalin A (Sigma-Aldrich C-2010) and Laminin (VWR 47743-734).Coverslips were coated on top, cell culture inserts were coated only on the bottom.Cells were allowed to adhere for 2h at 25°C, 80% humidity.
Culture medium containing cell debris and non-adherent cells was removed and cells attached to the cover slip were flooded with 500 µL culture medium.Cells were allowed to grow in 25°C, 80% humidity for up to seven days.

Soma/neurite separation in Drosophila primary neuron cultures
Three Drosophila 3rd instar larval brains were dissected for each sample.Primary neurons were incubated for seven days in vitro.After aspiration of the medium, the cell culture inserts (cellQART 931 10 12) were transferred to a new 12-well plate containing 1 mL of ice-cold PBS.The cell culture insert was inverted and neurites were removed using a cell scraper (FisherScientific 08100241) pre-wetted in 250 µL ice-cold PBS.The neurites were transferred by washing the cell scraper in 250 µL ice-cold PBS.
After the neurite removal, the cell culture insert was inverted again and the somata were removed by adding 250 µL ice-cold PBS to the insert and pipetting up/down 25 times.750 µL TRIzol LS Reagent (Ambion 10296028) were added immediately to each 250 µL-sample and RNA was extracted according to the manufacturer's instructions.

Immunofluorescence staining and analysis of cultured neurons
Six Drosophila 3rd instar larval brains were dissected per genotype, resulting in 1.5 larval brains used per coverslip and timepoint.Cells were fixed at RT for 10 min in 4% paraformaldehyde (PFA) in PBS, washed twice for 5 min in PBS and stored in PBS at 4°C.Cells were blocked for 1 h at RT with 3% BSA (Sigma #A6947 (LOT #SLBZ6761)) in 0.2% PBS-Triton X-100 (Sigma-Aldrich #X100).Primary antibodies were incubated overnight at 4°C in 0.2% PBS-Triton X-100, rinsed once with PBS and washed thrice 5 min with PBS.The secondary and conjugated antibodies were incubated for 1 h at RT in the dark, rinsed once with PBS and washed thrice 5 min with PBS.The penultimate wash included a counterstaining with 4′,6-diamidino-2-phenylindole (DAPI) (Abcam #AB228549) for 10 min at RT.The coverslips were mounted using VECTASHIELD antifade mounting medium (Vector Laboratories #H-1000) and imaged using a Zeiss LSM 900 confocal microscope in a sequential scanning mode.Neurites were traced in a semi-automated manner using the SNT framework for neuronal morphometry (Arshadi et al, 2021) on the Fiji (ImageJ) platform.To trace a neurite, the start point was set at the edge of the cell soma; a subsequent point on the neurite was selected next and the neurite was traced automatically between the two points.
Following three washes for 15 min at RT with 1xTBS, secondary antibodies were incubated for 4 hours at 35°C in 5% BSA in 0.1% TBS-Triton X100 in the dark.Fillets were washed three times for 15 min at RT with 1xTBS, mounted using VECTASHIELD antifade mounting medium (Vector Laboratories #H-1000) and imaged using a Zeiss LSM 900 confocal microscope in Z-project scanning mode.

Quantification of synaptic SYT1 and BRP expression
Images were analyzed using the Fiji (ImageJ) platform.Z-stacks were imported, the recorded channels split and a maximum intensity projection created.Using the ROI Manager, synapses were marked and the selected areas were copied to the channel of interest (Syt1 or Brp).To not introduce any bias, the images were randomly numbered and the synapses were marked in the HRP channel.Mean intensities and sizes of the marked areas were exported and quantified.

Isolation of Drosophila synaptosomes
~25g of anesthetized 3-days old w 1118 flies were collected in a kitchen blender (Philips ProBlend 6) in 400 mL ice-cold PBS and blended with 5 short pulses.The homogenate was poured through a sieve system (grid size top-to-bottom: 710 µm -425 µm -355 µm -125 µm).Each sieve was washed with pressurized water to separate different fly parts.Heads were collected in 355 µm sieves and transferred into a tube containing sucrose buffer (0.32 M sucrose, 1mM EDTA, 5mM Tris pH 7.4, 0.25 M DTT, supplemented with 1x Protease inhibitor cocktail Roche, 11873580001 and 1x RiboLock RNase inhibitor Thermo Fisher Scientific, EO0384) with a spatula.The suspension containing heads was poured through a 100 µm cell strainer to separate tissue from the collection sucrose buffer and separated into 4 equally aliquots that were each separately homogenized with 7 gentle strokes in a KONTES Tissue Grinder (VWR) (loose pestle) in 10 mL sucrose buffer.Homogenates were pooled and poured through a 40 µm strainer.The input sample for RNA sequencing and proteomics was collected at this step.The homogenate was centrifuged twice at 1000 g for 20 min at 4 °C, then at 2000 g for 20 min at 4 °C, each time carefully transferring the supernatant to a new tube, and finally at 15,000 g for 15 min at 4 °C.The resulting pellet was resuspended in 2 mL sucrose buffer, loaded onto a precooled Percoll gradient (from top to bottom 3, 10, 15, 23% Percoll diluted in sucrose buffer (0.32M sucrose, 1mM EDTA, 5mM Tris pH 7.4, 0.25mM DTT, 1x Protease inhibitor cocktail and 1x RiboLock RNase inhibitor) and centrifuged at 31,000 g at 4 °C for 5 min (acceleration 9, deceleration 7).Resulting layers contained membranes (fraction 1 and 2), membranes and synaptosomes (fraction 3), synaptosomes (fraction 4) and mitochondria (fraction 5) (see also Fig. 2).Synaptosome fractions (fractions 3 and 4) were diluted in 30 mL sucrose buffer and centrifuged at 20,000 g for 35 min 4 °C (acceleration 9, deceleration 6).Most of the supernatant was discarded, leaving ~65 ul of volume atop to not disturb the pellet containing synaptosomes.Each synaptosome sample was split in two: one half was diluted in TRIzol LS Reagent (Ambion, 10296028) for RNA purification according to the manufacturer's instructions.The other half was diluted in RIPA buffer (50 mM tris-HCl pH 8, 150 mM NaCl, 1% NP-40, 0.5% Na-deoxycholate, 0.1% SDS, 5 mM TCEP), homogenized for 1 min with a mechanical Pellet Pestle Motor (KONTES), incubated for 15 min on ice, centrifuged 5 min at 16,000 g at 4°C to clear the lysate, and submitted for proteomics analysis.

Mass spectrometry on Drosophila synaptosomes
Synaptosome and input fractions derived from three biological replicates were compared.
Corresponding samples (10 µg total protein in RIPA buffer containing 5 mM TCEP) were adjusted to 1xLDS sample buffer (Thermo Fisher; 5 mM TCEP final concentration) and heated for 5 min at 95°C.Proteins were separated on a 4-12% Novex BOLT gel (Thermo Fisher) and stained with colloidal Coomassie (InstantBlue, Expedeon).For each replicate, entire gel lanes were cut into five evenly distant slices, which were further processed by standard trypsin (Promega) in-gel digestion procedure.

Mass spectrometry analysis on Q Exactive Plus
Survey full scan MS spectra (m/z 300-1650) were acquired in the Orbitrap with 70,000 resolution (m/z 200) after accumulation of ions at a normalized AGC target of 3x10 6 at an expected LC peak width of 20 s and a default charge state of 2. Dynamic exclusion was set to 20 s (mass tolerance of ± 10 ppm).
The 12 most intense multiply charged ions (z ≥ 2 ≤ 6) were sequentially isolated (2.0 amu window) and fragmented in the octopole by higher-energy collisional dissociation (HCD) with a fixed maximum injection time of 60 ms at a normalized AGC target value of 1x10 5 and 17,500 resolution.General mass spectrometric conditions were as follows.Spray voltage, 2.1 kV; heated capillary temperature, 275°C; normalized HCD collision energy 28%, RF lens value 60% and MS/MS ion minimal AGC target value set to 1.2×10 4 counts.

Mass spectrometry data analysis
MaxQuant (v.1.6.14.0) employing standard parameters (enabling match between runs in between groups, match window 0.5 min) was used to identify peptides and final protein identification role-up (both at 1%FDR).MS raw data were searched simultaneously with the target-decoy standard settings against the Uniprot Drosophila melanogaster database (Uniprot_reviewed+Trembl including canonical isoforms; downloaded on August 2020) concatenated with an in-house curated FASTA file containing commonly observed contaminant proteins.Cysteine carbamidomethylation (+57.021464Da) was set as a fixed modification and N-acetylation of protein (N-term +42.010565Da), NQ deamidation (+0.984016Da) and methionine oxidation (+15.994915Da) as variable modifications.Maximum (allowed) variable PTMs per peptide were set to four.The MaxLFQ algorithm was utilized for relative label free quantification (min.ratio count ≥1) together with the iBAQ information.The MaxQuant output (proteingroups.txt;log2 transformed LFQ values) was further analyzed by standard R packages (removal of contaminant and reverse database hit IDs, vsn normalization, missing value imputation, limma package moderated T test and multiple testing correction).The fold change of protein abundance in synaptosomes vs. input was determined by calculating the difference between the Log2 transformed LFQ intensity values in the respective fractions.Proteins with Log2FC>0.6(after rounding up to the 2nd decimal place) and p-value<0.05(moderated limma p-value; not adjusted) were classified as enriched in synaptosomes.

Transmission electron microscopy on purified synaptosomes
Synaptosomes were fixed in a solution containing 4% paraformaldehyde (PFA, w/v) and 2% glutaraldehyde (w/v) in phosphate buffer (PB, 0.1 M, pH 7.4) for 30 minutes at RT.After fixation, the synaptosomes were washed with PB and incubated with 1% osmium tetroxide for 30 minutes.After every step the synaptosomes were centrifuged at 7,000 g for 3 minutes.Before dehydration, the pellet was coated with Agar.Graded ethanol (up to 50%) washes were performed for 5 minutes each, followed by incubation in 1% uranyl acetate for 30 minutes.The synaptosomes were then dehydrated with graded ethanol solutions (80%, 90%, 98% once for 5 minutes each, 100% twice for 10 minutes) and incubated in 100% propylene oxide (twice for 10 minutes) and for 1 hour in Propylenoxid/Durcupan 1:1.Subsequently, the synaptosomes were embedded in Durcupan resin (Sigma-Aldrich).Ultrathin sectioning (55 nm) was done with a Leica UC6 Ultracut.Sections were mounted onto copper grids (Plano) and contrasted using Pb-citrate for 3 minutes.Electron microscopy was performed using a Philips CM100 microscope equipped with a Gatan Kamera Orius SC600 at magnifications from 700x to 2950x.Image analysis was carried out with ImageJ/Fiji.

RT-qPCR
For synaptosome RT-qPCR, 500 ng/equal volume total RNA was used.Reverse transcription used iScript gDNA Clear cDNA Synthesis Kit (Bio-Rad).RT-qPCR was performed in a LightCycler 480 II instrument using FastStart SYBR Green Master (Roche).RT-qPCR primer sequences are listed in Table S5.

3'-end sequencing (3'-seq)
10 ng of total RNA from input or Pum xRIP, or 15 ng of total RNA from each purified fraction (input and synaptosome) was used for library preparation using Lexogen QuantSeq 3' mRNA-Seq reverse library prep kit (Lexogen) according to manufacturer's instructions.Paired-end sequencing was performed using the NovaSeq6000 platform (Illumina) and 2x150-bp reads.

3'-end site definition and filtering
For Pum xRIP-3'-seq, signal originating from internal priming was mitigated by blacklisting bases with downstream genomic stretches of As ≥6 bases long, or with 7 out of 10 downstream bases being As.
Regions within 25 bases of Ensembl-annotated 3'-ends (Ensembl assembly release dm6) were not subject to blackout.Only 3'-seq signal within 500 bp downstream of an Ensembl-annotated gene was considered.In Synaptosome 3'-seq, internal priming signal was mitigated by blacklisting bases with downstream genomic stretches of As ≥4 bases long, or with 6 of 8 downstream bases being As.Regions within 25 bases of Ensembl 3'-ends were not subject to blackout.Only 3'-seq signal within 2000 bases downstream of an Ensembl annotated gene was considered.3'-seq peaks were clustered outward up to 15 bases from peaks with a read depth ≥ 6 in a single sample.Intersecting clusters were merged into single clusters.Reads were then re-counted for all samples across these clusters.Both xRIP-3'-seq and Synaptosome 3'-seq clusters were further subjected to validation against a set of high-confidence 3'ends as described in Alfonso-Gonzalez et al. (Alfonso-Gonzalez et al, 2023).Briefly, these 3'-ends were either identified through direct-RNA long-read sequencing, or represent Ensembl (release 93) annotated 3'-ends that either i) are the most distal 3'-end of a gene with no observed long-read derived 3'-end, or ii) are a most-distal Ensembl 3'-end that is more distal than any observed long-read 3'-end.Only clusters intersecting 50-nt windows around these validated 3'-ends were considered in downstream length enrichment analyses and in differential gene/cluster expression analysis of the synaptosome 3'-seq dataset.Relative xRIP-3'-seq expression across proximal and distal (nUTR) 3' UTR regions of 271 validated nUTR-containing genes, as described in Carrasco et al. (Carrasco et al., 2020), was assessed as follows.xRIP-3'-seq signal within 100 bases of proximal and distal 3'-ends for each gene was whitelisted for respective regions.Additional signal within proximal and distal regions upstream of 100base 3'-end windows was included as valid signal only if they intersected 50-nt windows around highconfidence 3'-ends identified and described in Alfonso-Gonzalez et al. (Alfonso-Gonzalez et al., 2023).
Expression of proximal and distal regions were then assessed in DESeq2 (Love et al, 2014).

3' UTR length analysis in synaptosome 3'-seq
Normalized read counts of validated 3'-ends were exported from DESeq for each replicate and averaged for both conditions (Synaptosome and Input).3' UTR lengths of each 3' UTR isoform were assigned by subtracting the most proximal 3' UTR start site in the Ensembl dm6 (release 93) gene annotation from the 3'-end coordinate of each isoform.For genes showing more than one expressed 3'-end isoform (APA genes), the relative 3'-end isoform abundance was calculated for each of these isoforms as the percentage of total expression.To determine the expression-weighted 3' UTR length of each APA gene in input and synaptosomes, the distal-most 3' UTR length of each gene was normalized to 1, and the average 3' UTR length was calculated by summing [% of distal length * % of total gene expression] of all isoforms of a gene.Alternatively, ranked length metric was enforced whereby the distance between 3'-ends of 3' UTR isoforms was normalized to a ranked fraction of 1, with 1 representing the length of the most distal isoform (e.g., 0.33, 0.67, and 1 for a three-isoform gene).Average percent of ranked distal 3' UTR length was calculated for each APA gene for both synaptosome and input.Example: a two-isoform gene with each isoform expressed equally (50% each), ranked 3' UTR length=0.5*50%proximal expressed + 1.0 * 50% distal Isoform expressed.Both log2-fold difference and subtracted difference between the average percent of 3' UTR length in synaptosome vs. input were assessed using both the real length and rank-length metrics.5% Z-score thresholds (p<0.05,1-tailed) were calculated on all four distributions.Any gene with p<0.05 in either of these four distributions qualified that gene as displaying a significant 3' UTR length change between input and synaptosome.

Motif analysis
Motifs of known Drosophila melanogaster RBPs as defined in Ray et al. (Ray et al., 2013) were scanned for enrichment using the AME tool within the MEME motif analysis suite (v 5.5.0)(McLeay & Bailey, 2010).Using default AME parameters, the most distal expressed 3' UTRs of 460 identified Pum targets, and a subset of 71 nUTR-containing Pum targets, were assessed against backgrounds of the most distal expressed 3' UTRs of all 8,931 expressed genes, and 3,757 expressed APA genes, respectively.The -max enrichment scoring was used in addition to default -avg, when scoring enrichment of motifs in 3' UTRs of mRNAs encoding 989 synaptosome-enriched proteins vs. 3' UTRs of a background of 4,843 total genes detected by proteomics.HOMER motif analysis (Heinz et al, 2010) was used (find options: -offset 1 -strand +) to scan for presence and location of Pum motifs within proximal and nUTR regions of the set of nUTR-containing Pum targets, and the 200 nUTR-containing non-targets.microRNA enrichment analysis microRNA target sites as defined in the TargetScanFly repository (v7.2) (Agarwal et al, 2018) were scanned in most distal expressed 3' UTRs of mRNAs encoding 989 proteins enriched in the synaptosome fraction, and assessed for enrichment against a 3' UTR background of 4,843 proteins found in synaptosome or input.If a gene's 3' UTR contained a target site, the gene was qualified as a target gene.Enrichment of 91 conserved microRNAs was assessed.Log2-fold enrichment in the number of genes containing target sites was compared both with and without adjusting for the greater average 3' UTR length of mRNAs encoding proteins enriched in synaptosome relative to the all-protein background.A parallel scan and enrichment assessment were carried out on a set of 211 genes with RNA isoforms enriched in synaptosomes against an all expressed (based on mRNA 3'-seq) gene background of 9,034 genes.

Gene Ontology analysis
GO enrichment analysis was performed using the DAVID online server (version 2021).Pum mRNA targets (either all targets or targets with neuronal 3' UTRs) or all genes with neuronal 3' UTRs were queried against the background of RNAs expressed in Pum xRIP-3'-seq with base mean >10.Proteins enriched in the synaptosome fraction were queried against the background of all proteins detected in the synaptosome experiment.GO terms with Benjamini-Hochberg adjusted p-values <0.01 (Fig. S2E) or <0.05 (all other figures) were classified as significant.S1 for all significant terms (p<0.05;one-sided EASE score adjusted using the Benjamini-Hochberg method).E. Overlap between mouse (Zhang et al., 2017)(Zhang et al., 2017) and Drosophila (this study) Pum targets.118 genes represent Drosophila homologues of mouse Pum targets that also figure among the 460 Drosophila Pum targets.

Figures and figure legends
F. Gene ontology analysis of 118 Pum targets shared between fly and mouse.Significant terms related to nervous system function are shown.See Table S2 for all significant terms (p<0.05;one-sided EASE score adjusted using the Benjamini-Hochberg method).S3 for all significant terms (p<0.05;one-sided EASE score adjusted using the Benjamini-Hochberg method).E. Proportion of mRNAs that encode synaptic proteins in each category; "other RNAs" refers to mRNAs that are not Pum binding targets and that encode proteins detected by proteomics.***p<0.001(twotailed Fisher's exact test).Only RNAs that encode proteins detected by proteomics in the synaptosome isolation experiment were considered (see Fig. S2B).
F. Gene ontology analysis of 116 Pum targets that encode a protein found enriched in the synaptosome fraction.The top eight terms are shown.See Table S3 for all significant terms (p<0.05;one-sided EASE score adjusted using the Benjamini-Hochberg method).  A. Workflow of differential 3'-end isoform analysis.After mapping and clustering, 3'-seq read clusters were filtered against 3'-ends validated by long-read direct-RNA sequencing (see methods).For each gene, 3'-seq signal was scored for individual clusters in synaptosome fractions and compared to input to identify synaptosome-enriched 3' UTR isoforms of each gene.B. Differential 3' UTR lengths in synaptosomes compared to input.For each gene, the length of each expressed 3' UTR isoform was represented as a % of the length of the longest 3' UTR, weighed by expression of each isoform, and combined into one % 3' UTR length value for each gene.Genes with significant 3' UTR length differences between synaptosome and input fractions are depicted in orange (longer 3' UTR in synaptosomes) and blue (shorter 3' UTR in synaptosomes) (p<0.05,1-tailed Z-score).
Pum targets are marked in the darker shade of color.C. Gene ontology analysis of 71 Pum targets that possess a neuronal 3' UTR.The top three terms are shown but only one is significant (p=0.0085;one-sided EASE score adjusted using the Benjamini-Hochberg method).ns, non-significant.See Table S4 for

Figure 1 .
Figure 1.Pumilio directly binds mRNAs encoding synaptic proteins.A. xRIP-3'-seq experimental and data analysis workflow.Flag-tagged Pum protein was immuno-purified together with UV-crosslinked target RNAs from adult fly heads.RNA was subjected to 3'-end sequencing.To determine the exact 3'-end position, sequencing reads were mapped and unreliable clusters were filtered out (see methods).Pum target RNAs were identified by enrichment of 3'-seq signal in the xRIP sample over input.B. Visualization of 3'-seq signal for an exemplary Pum target (GluRIA) and the neighboring two nontarget genes.Control xRIP represents samples from flies not expressing FLAG-tagged Pum (w 1118 ).C. Differential RNA expression in the Pum xRIP sample compared to input (head homogenate), represented as a function of mean expression levels.Dark blue represents: |log2 fold change (Pum xRIP/input)| >1 with p-value<0.05and base mean >10, and |log2 fold change (Pum xRIP/control xRIP)| >1.

Figure 2 .
Figure 2. Pumilio targets encode proteins enriched in synaptosome fractions.A. Workflow of Drosophila synaptosome isolation.Fresh head tissue from adult Drosophila was dissociated and shear force was used to separate pre-and postsynaptic membranes from nuclei,

G
. Proportion of Pum targets in each category of mRNA subcellular localization, represented as enrichment over non-localized genes.*p<0.1 **p<0.05(two-tailed Fisher's exact test).H. Model of 3' UTR-dependent localization of Pum target mRNAs.In neuronal cell bodies, Pum binds to mRNAs encoding synaptic proteins, with a preference for long 3' UTR isoforms.This leads to the enriched localization of translationally competent short isoforms in synaptic compartments.

C
. Representative examples (3'-seq tracks and 3'-end isoform models) of genes with differentially localized 3' UTR isoforms.The upper two traces in each panel show enrichment of the shorter Beadex (Bx) 3' UTR isoform (blue), of the longer Octopamine receptor in mushroom bodies (Oamb) 3' UTR isoform (orange), or no change for Semaphorin 2b (Sema2b) in synaptosomes compared to input.The lower three traces show Pum preferential binding to the long (Bx, OamB) or short (Sema2b) 3' UTR isoforms of its target mRNAs, respectively.D. Proportion of Pum targets in each category of 3' UTR isoform subcellular localization, represented as enrichment over all expressed genes.ns, non-significant, ***p<0.001(two-tailed Fisher's exact test).E. 3' UTR length percentages in synaptosome (orange) and input (blue) fractions of Pum target mRNAs from Fig. 3B, that display differentially localized 3' UTR isoforms.10 and 43 genes display longer and shorter 3' UTR isoforms in synaptosomes, respectively.F. Heatmaps showing Pum binding and the number of Pum binding motifs in distal 3' UTR regions of Pum target mRNAs that display shorter 3' UTR isoforms in synaptosomes (43 genes), ranked by Pum xRIP-3'-seq signal compared to input.

Figure 4 .
Figure 4. Pumilio binds to soma-localized neuronal 3' UTRs of synaptic genes.A. Pum binding motif found enriched in neuronal 3' UTRs.B. Proportion of nUTR-containing genes in the indicated gene categories, represented as enrichment over all expressed genes.***p<0.001(two-tailed Fisher's exact test).
all terms.D. Number and proportion of nUTR-containing genes among mRNAs with a shorter 3' UTR in synaptosomes, in each gene category.**p<0.05(two-tailed Fisher's exact test).Only mRNAs expressed in both the synaptosome and input samples were considered.E. Heatmaps showing Pum binding and the number of Pum binding motifs in the nUTR of Pum targets whose nUTR-containing 3' UTR isoform is depleted in synaptosome fractions (14 genes), ranked by Pum xRIP-3'-seq signal compared to input.

Figure S2 .
Figure S2.Pum target mRNAs encode proteins enriched in synaptosome fractions.A. RT-qPCR quantification of the indicated transcript regions in synaptosome fractions relative to input fraction.For each gene, intron levels were normalized to coding exon mRNA levels, which were set to the value 1. Ratios represent the average of two biological replicates.

Figure S3 .
Figure S3.Synaptic localization of short 3' UTR isoforms of Pumilio target mRNAs.A. Number and proportion of Pum targets in each category of 3' UTR isoform subcellular localization.ns, non-significant, ***p<0.001(two-tailed Fisher's exact test).Only Pum target mRNAs expressed in both synaptosome and input (306 genes) were considered.B. Heatmaps showing Pum binding and the number of Pum binding motifs in distal 3' UTR regions of Pum target mRNAs that display longer 3' UTR isoforms in synaptosomes (10 genes), ranked by Pum xRIP-3'-seq signal compared to input.

Figure S4 .
Figure S4.Pumilio binds to soma-localized long neuronal 3' UTRs of synaptic genes.A. Gene ontology analysis of 271 genes that possess a neuronal 3' UTR.All significant terms are shown.See also TableS4.

Figure S5 .
Figure S5.Impaired neurite outgrowth, mRNA delocalization, and synaptic protein overexpression in neurons of Δpum flies.A. Light micrographs of Drosophila primary neuronal cells cultured on a microporous membrane for separation of neurite and soma compartments.Before removal, neurites are visible as dark protrusions (white arrowheads) below the cell bodies.After soma removal, only cell debris and micropores are visible on the membrane.The three images represent three distinct cultures at 7 days in vitro.Scale bars: 200 μm.