Loss of Pseudouridine Synthases in the RluA Family Causes Hypersensitive Nociception in Drosophila

Nociceptive neurons of Drosophila melanogaster larvae are characterized by highly branched dendritic processes whose proper morphogenesis relies on a large number of RNA-binding proteins. Post-transcriptional regulation of RNA in these dendrites has been found to play an important role in their function. Here, we investigate the neuronal functions of two putative RNA modification genes, RluA-1 and RluA-2, which are predicted to encode pseudouridine synthases. RluA-1 is specifically expressed in larval sensory neurons while RluA-2 expression is ubiquitous. Nociceptor-specific RNAi knockdown of RluA-1 caused hypersensitive nociception phenotypes, which were recapitulated with genetic null alleles. These were rescued with genomic duplication and nociceptor-specific expression of UAS-RluA-1-cDNA. As with RluA-1, RluA-2 loss of function mutants also displayed hyperalgesia. Interestingly, nociceptor neuron dendrites showed a hyperbranched morphology in the RluA-1 mutants. The latter may be a cause or a consequence of heightened sensitivity in mutant nociception behaviors.

Pain serves an indispensable, protective role but when pain becomes pathological it can have a debilitating impact on human life. The total annual cost of pain to society in the United States was estimated by the Institute of Medicine to be up to $635 billion, which is greater than that of heart disease, cancer, and diabetes combined (Gaskin and Richard 2012). It is therefore of urgent importance to uncover the basic molecular and cellular mechanisms involved in pain in order to better treat it. Laboratory animal models of pain and nociception have played an essential role in identifying such mechanisms. Drosophila larvae respond to noxious thermal and mechanical stimuli through stereotyped rolling escape locomotion (in which the larva rotates around its long body axis) which is easily distinguishable from other forms of locomotion (Tracey et al. 2003). Combined with the unparalleled genetic tools available for Drosophila melanogaster, this behavioral readout provides an excellent system to study the genetics of nociception and pain (Tracey et al. 2003;Caldwell and Tracey 2010;Milinkeviciute et al. 2012;Im and Galko 2012;Khuong et al. 2019). Previous studies have demonstrated a specific subset of dendritic arborization (da) sensory neurons in the peripheral nervous system, Class IV multidendritic da (cIVda) neurons, are of critical importance for thermal, mechanical, and high intensity light nociception (Grueber et al. 2002;Hwang et al. 2007;Xiang et al. 2010). Further evidence suggests a lesser but significant contribution of Class II (cIIda) and Class III da (cIIIda) neurons in mechanical nociception (Hwang et al. 2007;Kim et al. 2012;Hu et al. 2017). In addition, great progress has been made in identifying the circuits in the larval abdominal ganglion that are involved in rolling escape locomotion (Ohyama et al. 2013;Ohyama et al. 2015;Chin and Tracey 2017;Burgos et al. 2018).
Genes that are specifically expressed in multidendritic (md) neurons have been found to play a role in nociception. For instance, painless (Tracey et al. 2003) is required for mechanical and thermal nociception and it is expressed in all four classes of md neurons. Similarly, straightjacket (Neely et al. 2010) is expressed in md neurons and required for avoidance of noxious heat. Mechanical nociception et al. 2014;Gorczyca et al. 2014;Guo et al. 2014) and the polymodal nociception gene dTRPA1-C/D (Zhong et al. 2012) each show very specific expression in the cIVda neurons. Forward genetic screens have also identified a set of genes with enriched expression in the cIVda neurons which either inhibit or activate nociceptive pathways (Honjo et al. 2016).
The historically first known genetic marker with specific expression pattern in the da neurons was the lacZ enhancer trap E7-2-36 (Brewster and Bodmer 1995). Later studies reported that this enhancer trap gene was inserted upstream of the RluA-1 gene and DNA sequences from upstream of RluA-1 caused expression of GAL4 in multidendritic neurons (Wang et al. 2011). The RluA-1 gene encodes an enzyme that is predicted to have the conserved pseudouridine synthase domain required to catalyze the isomerization of uridines to pseudouridines on RNA (Sivaraman et al. 2004;Wang et al. 2011;Waterhouse et al. 2018), but how this RNA modifying protein is involved in the function of nociceptive multidendritic neurons remains unknown.
A widespread importance of RNA-binding proteins in cIVda neuron dendrite morphogenesis and function was found in a recent large-scale RNAi screen that identified 88 genes encoding RNAbinding proteins whose knockdown caused aberrant dendrite morphogenesis (Olesnicky et al. 2014). The elaborate dendrite arbors of cIVda neurons project long distances from the neuronal cell body and mRNA granules are trafficked to these distant sites where they may undergo local translation. Indeed, RNA granules containing Nanos (Nos), Pumilio (Pum), Oskar (Osk), Fragile X Mental Retardation (FMRP) and other proteins have been shown to regulate the formation of higher order dendrites in these cells (Ye et al. 2004;Pan et al. 2004;Brechbiel and Gavis 2008;Bianco et al. 2010;Xu et al. 2013). Mechanical nociception defects are also observed in animals with disruption in these pathways (Xu et al. 2013).
Pseudouridylation is the most common post-transcriptional RNA modification. Pseudouridine (Psi, C), the C5-glycoside isomer of uridine, was initially found in many positions in rRNA, tRNA and snRNA in all organisms that have been investigated (Ge and Yu 2013). RNA-seq based global pseudouridine profiling has shown the presence of C in many mRNAs and a large number of those sites were found to be dynamically regulated in yeast and human cells (Carlile et al. 2014;Lovejoy et al. 2014;Schwartz et al. 2014;Khoddami et al. 2019). Dysfunctional pseudouridylation has been linked to several human diseases (Knight et al. 1999;Fujiwara and Harigae 2013;de Brouwer et al. 2018). Since many sites of pseudouridylation in different organisms are evolutionarily conserved, Drosophila melanogaster provides an excellent and genetically tractable metazoan system to elucidate some of these functions (Giordano et al. 1999;Deryusheva and Gall 2013;de Brouwer et al. 2018).
The isomerization of uridine to pseudouridine is catalyzed by six families of pseudouridine synthases. They function either as guide RNA directed ribonucleoprotein complexes or as stand-alone proteins (Koonin 1996;Hamma and Ferré-D'Amaré 2006). In the Drosophila genome, 9 proteins have been identified with annotated pseudouridine synthase domains. Minifly (mfl), the RNA-dependent pseudouridine synthase homolog of human dyskerin (mouse NAP57 and yeast Cbf5), is required for somatic stem cell homeostasis and is essential for Drosophila viability and fertility (Phillips et al. 1998;Giordano et al. 1999;Vicidomini et al. 2017). Knockout of Drosophila Pus7, the human and yeast Pus7 homolog, results in increased aggressiveness in adult flies (de Brouwer et al. 2018). The function and specificity of other predicted pseudouridine synthases are largely unknown. Among the six families, the RluA family, which does not rely on guide RNAs, appears to be the most complex based on divergent substrate specificities in bacteria and yeast (Hoang et al. 2006). RluA family members in bacteria are involved in ribosomal assembly and growth (Raychaudhuri et al. 1999;Gutgsell et al. 2005) but their function in multicellular organisms has not been studied. Although pseudouridine synthases appear to function ubiquitously, as noted above, Drosophila RluA-1, a member in RluA family, has been reported to be specifically expressed in md neurons (Wang et al. 2011). Thus, we have investigated the role for RluA-1 and its paralog RluA-2, in nociception pathways that are known to depend on md neurons. Our results indicate an important function for RluA-1 and RluA-2 in the regulation of nociception.

RNA isolation, RT and Q-PCR analysis
To evaluate the RluA-1-RNAi efficiency, a stock of the md-Gal4 driver built with UAS-dicer2 (w; md-GAL4; UAS-dicer2) was crossed to UAS-RluA-1-RNAi (test) or its genetic background (w 1118 , control). A batch of four L3 larvae were rinsed in 1xPBS and quickly frozen in liquid nitrogen for RNA extraction. RNA isolation was performed using TRIzol (Life Technologies) following the manufacturer's protocol. 1 mg of total RNA samples were subjected to DNase treatment and reverse-transcription using the SuperScript IV reverse transcriptase enzyme and Oligo(dT) 12-18 primer (Life technologies). cDNA was amplified in real time using the qPCR Master mix plus for power SYBR Green I assay (Invitrogen) and analyzed with the QuantStudio Real-Time PCR System (Applied Biosystems). Intron-spanning primer pairs (RluA-1-cDNA-F: 59-GAGCAGCAGATTCGCAA-CAG, RluA-1-cDNA-R: 59-ACTTCAATGGGCTCCTTGCA) and (Act5C-F: 59-GGGGCAGAGCAAGCGTGGTA, Act5C-R: 59-GGGTG-CCACACGCAGCTCAT) were used for RluA-1 and Actin5C, respectively. The level of expression for RluA-1 in RNAi expressing line (md-Gal4.RluA-1-RNAi) was normalized based on the Actin5C and expressed as a percentage of the control.
To identify the desired HDR mutants, G 0 flies were crossed to w 1118 and single F 1 founders were identified with DsRed fluorescence in the eyes (from 3XP3-DsRed reporter) and mated with a second chromosome balancer strain to establish independent lines. An initial molecular screening for the desired events was performed by PCR on gDNA extracted from candidates placed over a deficiency (Df(2L) Exelixis7048) with genomic primers located outside of the homology arms for RluA-1 and RluA-2, respectively (RluA-1-front-F: 59-GAG-TAATTGTGGGTGTGCCAGAG-39 and RluA-1-end-R: 59-CTG-GACTTTTGTTACCCCTT-39 for RluA-1, RluA-2-front-F: 59-CGGATTGGAAATGTGCCATC -39 and RluA-2-end-R: 59-TT-CCAGTTGAATATCGCCGTG -39 for RluA-2, data not shown). For positive candidates, subsequent rounds of PCRs were performed to demonstrate the desired HDR event comparing the homozygous (deletion allele over a deficiency), heterozygous (deletion allele over CyO) and wild type (the corresponding injection line) for RluA-1 del-HDR , RluA-2 del-HDR and RluA-1 del-HDR RluA-2 del-HDR (primer pairs marked in Figures S4A and S6A, PCR amplification in Figures S4B  and S6B). PCR products sequenced across the attp-loxP-3XP3-DsRed-SV40-loxP fragment ( Figure S4C and Figure S6C) confirmed accurate targeting of the loci. For behavioral analysis, the original deletion mutants were backcrossed to CS, w 1118 or isow 1118 for six times. For each generation, five heterozygous females were selected for six successive backcrosses in vials and about 10 heterozygous females were used to cross to a second chromosome balancer to produce balanced mutant males in bottles and finally heterozygous mutant virgins and balanced males were crossed en masse in bottles to establish the balanced and homozygous mutant lines. In all crosses the DsRed fluorescence marker was used to follow the presence of the mutant.
Generation of deletion line in RluA-1 and RluA-2 using FRT-mediated deficiency FRT bearing insertions {WH+} f02750 and {XP-} d2586 and in RluA-1 (locations marked in Figure S4A) were used for generating a deficiency allele RluA-1 del-FRT . Insertions of {WH-}RluA-2 F07702 and {WH-}Grip75 f05483 (locations marked in Figure S6A) were used for generating the deficiency allele RluA-2 del-FRT . Crossing and heat-shock schemes followed (Parks et al. 2004). Hybrid PCR with corresponding primers (WH59 plus / XP59 minus left and right primers) was used to screen for candidate lines with w-deletion in RluA-1 del-FRT and two-sided PCR with left and right primers for WH39 minus/ WH59 minus was used for screening for candidate lines with w+ deletion in RluA-2 del-FRT (Parks et al. 2004). Molecular testing for the deletion was performed by PCR on gDNA extracted from positive candidates placed over a deficiency (Df(2L)Exelixis7048) covering the RluA-1 and RluA-2 region.
Generation of GFSTF insertions in RluA-1 and RluA-2 by recombination mediated cassette exchange (RMCE) Crossing, heat shocking and screening for EGFP tagged MiMIC lines in RluA-1 and RluA-2 were essentially carried out as described (Nagarkar-Jaiswal et al. 2015). Males carrying a MiMIC insertion in a coding intron of RluA-1 (RluA-1 MI06897 ) or RluA-2 (RluA-2 MI12981 ) were crossed to females carrying the hs-FLP and vasa-phiC31 integrase on the X chromosome and a frame-specific ("phase 0") FRT flanked multiple tag (GFSTF) cassette on chromosome III. Candidate males with mosaic w-eyes and y-bodies were individually crossed to w; Sco/ CyO; Sb/TM3 Ser balancers to establish stocks. The presence and direction of the insertion were tested by PCR assays described (Venken et al. 2011). Since the original MiMIC insertion in RluA-1 or RluA-2 and the respective gene are in the same orientation, positive PCR reaction 1 (with primers MiLF and TagR) and 4 (with primers MilR and TagF) as described (Venken et al. 2011) indicated a successful RMCE event and resulted in expression.
Generation of UAS-RluA-1 Drosophila RluA-1 full length cDNA clone for RluA-1 transcript A (FI04540) was obtained from the Drosophila Genome Resource Center (DGRC). The UAS-RluA-1 expression constructs were generated with the ENTR/gateway system following the instructions of the manufacturer (Invitrogen). The RluA-1-cDNA-F (59-CACCATG-CAGAATTCTCCGGCT-39, and RluA-1-cDNA + STOP-R (59-TCATG-CCGAGTCTAAGTG-39 primers were used to amplify the open reading frame and cloned into pENTR-D-TOPO vector. The sequence was verified and then cloned to the P-element based destination vector pTW. Model System Injections performed injections into w 1118 embryos. F 0 flies were crossed to w 1118 and single F 1 founders were identified based on the w + marker. Individual lines were mapped and balanced to establish stable stocks. A UAS-RluA-1 line with relatively weak expression evaluated by Q-PCR in RluA-1-cDNA lines driven by md-GAL4 was used for the cIV-neuron specific RluA-1-cDNA rescue experiment.

Larval behavioral analyses
Wandering 3 rd instar larvae were washed out from vials and acclimated for 5 min in petri dishes before testing. Larval thermal nociception assays were conducted essentially as previously described (Caldwell and Tracey 2010;Mauthner et al. 2014;Walcott et al. 2018), except that the probe is gently held against the lateral surface of abdominal segments 4, 5, or 6 until the animal completes a 360°roll along the dorsal-ventral axis. All animals tested eventually performed rolling and the response latency from all the animals was graphed for a given genotype. Larval mechanical nociception response assays were conducted as previously described (Mauthner et al. 2014). Behavioral recording and scoring were performed with the observer blinded to the genotype.
For gentle touch assays, early L3 larvae were scooped out from the top layer of the fly food in the vials and 5-10 larvae were briefly rinsed with PBS and allowed to acclimate on 1% agarose in a plate for 5 min before testing with an eyelash fixed to the end of a paintbrush. Each larva was brushed with the eyelash on segments T1-A3 for 4 times and the responses were recorded and summarized using a gentle touch scale (Tsubouchi et al. 2012).

CIV dendrite imaging, tracing and analysis
To image class IV dendrites, the DsRed marker in RluA-1 del-HDR was first removed with CRE recombinase (w 1118 ; RluA-1 del-HDRDDsRed ) and ppk1.9-CD4::tdTom was introduced to the mutant background and a stable homozygous stock was established. For CIV dendrite analysis, six virgins and three males were crossed in each vial for the mutant (w 1118 ; RluA-1 del-HDRDDsRed ; ppk1.9-CD4::tdTom) and control (ppk1.9-CD4::-td-Tom). Wandering larvae were anesthetized with diethyl ether in a sealed glass chamber for 15min before being arranged on a slide and covered with 50 mm glass coverslip. Neurons expressing the fluorescently tagged markers were visualized on a Zeiss LSM 5 Live confocal microscope with a 40X oil objective (Plan Apo M27, NA 0.8). Images were collected as 5x3 tile scans of z-stacks with 512x512 resolution. A MatLab build was used for initial automatic tracing of the ddaC neuron dendrites from the confocal z-stack series TIFF images (Gulyanon et al. 2016). The generated SWC files were overlaid onto the maximum intensity projected image of the neuron in neuTube (Feng et al. 2015) and manually curated to eliminate tracing errors made by MatLab. The corrected images were then analyzed with MatLab to extract neuron features of interest including number of branches, average branch length, and neuron size (the estimated size of the neuron, defined as the area of the minimum bounding circle) (Gulyanon et al. 2016). Isoneuronal crossover events were quantified manually from the traced dendrites for each genotype (n = 6 neurons).

Statistical analysis
Statistics were performed using GraphPad Prism 4. Thermal and gentle touch behavioral data were compared with an unpaired non-parametric Mann-Whitney test when comparing two groups and Kruskal-Wallis test when comparing three or more groups. Mechanical nociception behavioral data were compared with Fisher's exact test. Dendrite morphology data were compared with the Student's t-test. Error bars represent standard deviation (S.D.) in all the figures unless otherwise specified. Q-RT-PCR data were analyzed and plotted with Estimation Stats (Ho et al. 2019).

Data availability
Strains and plasmids are available upon request. All supplementary figures have been uploaded to figshare. The authors affirm that all data necessary for confirming the conclusions of the article are present within the article and its figures. Supplemental material available at figshare: https://doi.org/10.25387/g3.12755873.

RESULTS
Structural Conservation of RluA-1 and RluA-2 With Pseudouridine Synthases Both RluA-1 and RluA-2 are annotated as putative pseudouridine synthases but their Psi synthase activities have not yet been proven biochemically. Thus, we explored features of their predicted amino acid sequences to further scrutinize their hypothesized enzymatic activity. First, we aligned the predicted amino acid sequences of RluA-1 and RluA-2 with their closest homologous pseudouridine synthases across the kingdoms of life. This sequence alignment shows that RluA-1 and RluA-2 are highly conserved throughout the entire predicted pseudouridine synthase domain ( Figure 1A). Importantly, highly conserved residues of pseudouridine synthase domain are distributed across the primary amino acid structure in four sequence motifs (motif I, motif IIa, motif IIb and motif III) as defined by Koonin (Koonin 1996). These motifs harbor key residues that are close to 100% conserved across superfamily members and species. Each of the fully conserved non-catalytic residues (i.e., motif I, Figure  1A) and critical residues thought to be important for catalysis (asterisks, Figure 1A) are conserved in both RluA-1 and RluA-2. To further evaluate the level of structural conservation within the pseudouridine synthase domains of RluA-1 and RluA-2, we generated structural homology models based on solved structures of pseudouridine synthases. Interestingly, we found that the best model for RluA-1 was based on E. coli RluA (PDB ID 2i82), whereas the best RluA-2 model was generated from the template structure of E. coli RluC (PDB ID 1vk6) ( Figure 1B, 1C, Figure S1A, S1B). In our model, key residues and their side chains (DYI/LR), within motifs II, IIa and III of RluA-1 and RluA-2 models, are clearly superimposed with the position of the template residues in 3D space ( Figure 1D). Together, the primary sequence and homology model analyses provide strong evidence of conservation of function of pseudouridine synthase domains in RluA-1 and RluA-2, supporting the hypothesis that RluA-1 and RluA-2 act as pseudouridine synthases, leaving to future investigation the nature of their substrates and whether they interact with RNA similarly as other synthases ( Figure S1C, S1D). However, we cannot exclude the possibility that these enzymes may possess additional activities. For instance, both RluA-1 and RluA-2 possess amino acid similarity to an RNA-binding S4 domain (located N-terminal to the Psi synthase domain (data not shown)).
RluA-1 is expressed in the multidendritic neurons of the peripheral nervous system and in the cells of brain Previously described reporter genes for RluA-1 showed specific expression in larval multidendritic neurons (Wang et al. 2011). We replicated this finding by generating an RluA-1 GAL4 driver at the endogenous gene locus through recombination mediated cassette exchange (RMCE) of RluA-1 MI06897 (an intronic MiMIC) and a Trojan exon cassette (Diao et al. 2015). An mCD8GFP reporter driven by RluA-1 GAL4 was expressed in peripheral sensory neurons in each segment of the larval body wall (Figure 2A). In the dorsal cluster, GFP-positive signals were clearly detected in all four classes of md-da sensory neurons, dorsal multiple dendrite neuron (dmd1), external sensory (ES) and dorsal bipolar dendritic (dbd) neurons ( Figure 2B). In the larval ventral nerve cord, the GFP-positive signals were seen in the axonal projections of the sensory neurons ( Figure 2C). GFP signal was also observed in the unidentified clusters of neurons in the larval brain ( Figure 2C). In the adult brain, significant signals were detected in the optic lobes and other small cell clusters of the central brain ( Figure  S2A).
To determine the localization of the RluA-1 proteins, we generated a GFSTF exon trap (Nagarkar-Jaiswal et al. 2015) that expresses an in frame GFP fusion with RluA-1 at the endogenous genomic locus (with RMCE of the RluA-1 MI06897 MiMIC element) (Nagarkar-Jaiswal et al. 2015). Although live imaging did not detect the EGFP tagged RluA-1 protein, immunostaining with anti-GFP labeled nuclei of larval multidendritic neurons, ES and dbd neurons also expressing a membrane-localized RFP (Gal4109(2)80 . UAS-CD8-RFP, Figure  2D). The tagged RluA-1 protein was also detected in the cell bodies of neurons in optic lobes and other yet-to-be identified cells in the adult brain ( Figure S2B).
Reducing or removing RluA-1 results in a hypersensitive thermal nociception phenotype Given our confirmation of the expression of the RluA1 gene and protein in the multidendritic neurons we tested for potential roles of RluA-1 in the regulation of nociception. To do so, we first performed tissue-specific knockdown using GAL4/UAS based RNA interference (RNAi) (with a UAS line that showed a trend in reducing the expression of RluA-1 transcripts in multidendritic neurons ( Figure S3)). To investigate the behavioral consequence, a cIVda specific driver (ppk1.9-GAL4; UAS-dicer2) (Ainsley et al. 2003) was employed to drive the RluA-1-RNAi in the larval cIVda nociceptors. We assessed potential insensitive phenotypes (with a probe temperature of 46°) and potential hypersensitive phenotypes (with a temperature of 42°) as previously described (Honjo et al. 2016). When stimulated with the higher temperature 46°p robe the RluA-1-RNAi knockdown larvae responded significantly faster than the ppk-GAL4 driver alone controls. This genotype also showed a trend toward responding faster than the no driver UAS-RNAi/+ controls but this difference was not statistically significant ( Figure 3A). The results suggested that reducing RluA-1 in classIV neurons may have made the larvae more sensitive to noxious heat. Indeed, this was clearly observed for the RluA1-RNAi animals when testing with 42°probe that allows for easier detection of hypersensitivity ( Figure 3A). The average latency to roll in the RluA-1 knock-down animals was significantly faster than the driver alone animals or the UAS-RNAi control animals. These data combined suggest that reducing the activity of RluA-1 in the noxious heat-responsive clVda cells caused thermal hyperalgesia.
For the remainder of our behavioral studies, we focused on the isogenized w 1118 background. This had the advantage of greater genetic uniformity relative to w 1118 and CS, as well as showing the strongest hypersensitive mutant phenotype for RluA-1.
Genetic rescue of RluA-1 mutant restores thermal nociception response To test that the mutation in RluA-1 was the underlying cause of the hypersensitive nociception phenotype, we introduced a genomic rescue construct into the RluA-1 del-HDR mutant background (a Bacterial Artificial Chromosome (BAC) (P6-D7)) (Venken et al. 2009) covering the RluA-1 gene region). With the 42°stimulus, larvae with two copies of the duplication (Dp) in the background of RluA-1 del-HDR showed rescue of the hypersensitivity ( Figure 4A). This rescue with the genomic duplication was dosage dependent. Larvae with only one copy of the Dp in the RluA-1 del-HDR background (RluA-1 2/2 ; Dp +/2 ) responded more slowly than the mutant but this difference was not statistically significant ( Figure 4A). Combined, the results of genomic rescue experiments support the hypothesis that mutation of RluA-1 is indeed the cause of the hypersensitive thermal nociception phenotype.
A caveat remained in that the genomic rescue construct included other genes in addition to RluA-1. Thus, genomic rescue did not rule out the possibility that a mutation tightly linked to RluA-1, but not in RluA-1 itself, was responsible for the mutant phenotype. Thus, as a final test, we generated transgenic lines to express an RluA-1 cDNA under the control of the GAL4/UAS system (UAS-RluA-1). Using the UAS-RluA-1 we specifically restored RluA-1 to cIVda neurons in the RluA-1 null mutant background. When stimulated with the 42°probe the animals with both the ppkGal4 driver and the UAS-RluA-1-cDNA transgene in the RluA-1 del-HDR showed a complete rescue from the hypersensitivity seen in the null mutant ( Figure 4B). Neither the ppkGal4 driver alone nor the UAS-RluA-1 had an effect on the hypersensitive thermal nociception phenotype in the RluA-1 null mutant background, excluding the possibility of non-specific effects of these transgenes ( Figure 4B). In addition, overexpression of UAS-RluA1 with a md neuron driver (MD-Gal4) had no effect on nociception behavior at 42°stimulus ruling out the possibility that the increased latency seen in the rescue effect was a non-specific consequence of over-expression ( Figure S5). Combined, these nociceptor-specific rescue experiments provide genetic confirmation that loss-of-function mutation in RluA-1 causes hypersensitive thermal nociception and localizes the site of action for RluA-1 in this process to the nociceptors.

RluA-1 requirement for mechanosensory thresholds
The cIVda neurons are not only required for detection of noxious heat, they also contribute to sensing harsh mechanical stimulation (Hwang et al. 2007;Zhong et al. 2010;Mauthner et al. 2014). Thus, we investigated the RluA-1 mutant responses to noxious mechanical stimuli. When stimulated with a 30mN/720kPa Von Frey fiber, significantly more RluA-del-HDR null mutant larvae performed the typical nociceptive rolling behavior compared to the isow 1118 controls (Figure 5A). Even when the probe was reduced to 15mN/360kPa, the majority of RluA-1 2/2 larvae still rolled while less than half of control larvae rolled ( Figure 5A), indicating the defect in RluA-1 also caused hypersensitive mechanical nociception. Loss of RluA-1 did not have any impact on behavioral responses to gentle touch ( Figure 5B) suggesting a more specific involvement in nociception than for sensory processing in general.

Expression pattern and nociception functions for RluA-2
The RluA-2 locus is adjacent to RluA-1 on the second chromosome of Drosophila melanogaster. RluA-2 has significant sequence similarity with RluA-1 within the evolutionarily conserved pseudouridine synthase domain (Figure 1, Figure S1), suggesting possible functional overlap for the encoded proteins. To investigate the expression of RluA-2 we generated a GFSTF line with RMCE of the RluA-2 MI12981 mimic element ( Figure S6A) (Nagarkar-Jaiswal et al. 2015). Immunostaining with anti-GFP labeled nuclei in all of the cell types that we observed in third instar larvae ( Figure 6A), including md neurons ( Figure 6B).
To test the function of RluA-2 we generated a deletion (RluA-2 del-HDR ) to remove its pseudouridine synthase domain via CRISPR/Cas9 HDR ( Figure S6A-C). Note that RluA-2 del-HDR is not an RNA null allele and some residual function from the RNA binding S4 domain may remain. However, the RluA-2 pseudouridine synthase domain is completely removed. Since RluA-1 and RluA-2 are both expressed in the md neurons ( Figure 2D and Figure 6B), we generated a double mutant (RluA-1 del-HDR RluA-2 del-HDR ), by injecting the RluA-2 del-HDR constructs in the RluA-1 del-HDR null mutant background ( Figure S6). As RluA-1 and RluA-2 are the only annotated RluA family members in D. melanogaster, the double mutant completely removes RluA family pseudouridine synthase activity from the flies. Prior to functional assessment, the single mutant and the double mutant were backcrossed six times to the genetic background of isow 1118 .
We tested each single mutant (RluA-1 del-HDR and RluA-2 del-HDR ) together with the double mutant RluA-1 del-HDR RluA-2 del-HDR side by side in thermal nociception assays with the 42°thermal stimulus. RluA-1 del-HDR larvae (RluA-1 2/2 ) again responded significantly faster than the genetic background control (Figure 7). The RluA-2 del-HDR single mutant larvae also displayed a faster response to the stimulus ( Figure 7) and similar hypersensitivity was also seen in an independent allele for RluA-2 that we generated by FLP/FRT mediated recombination (RluA-2 del-FRT ) allele ( Figure S7). Finally, the double mutant RluA-1 del-HDR RluA-2 del-HDR larvae showed a faster response than the control larvae to the same extent as the single mutant of RluA-2 del-HDR (Figure 7). These results indicated that RluA-2, like RluA-1, negatively regulates nociception. The finding that the double mutant did not show a more severe phenotype than either single mutant suggests that RluA-2 and RluA-1 have non-redundant functional roles, and that they may function in the same molecular pathway. When this pathway is disrupted, hypersensitive nociception results.
RluA-1 regulates neuronal dendrite morphology of nociceptors A nociceptor-specific RNAi screen with thermal nociception assay discovered dozens of genes whose reduction caused either insensitive or hypersensitive thermal nociception (Honjo et al. 2016). Interestingly, some of those genes targeted with RNAi showed a reduced or increased branching of Class IV neuron dendrites. Reduced dendrite branching was often seen with nociceptive insensitivity while increased branching was found in some hypersensitive genotypes. Thus, regulation of Class IV neuron dendrite morphology is a commonly affected developmental pathway that is related to nociception phenotypes. Given this, we investigated the dendrite morphology of the cIVda neuron dendrites in the RluA-1 del-HDR mutant. In mutant ddaC neurons visualized with ppk-CD4-tdTomato, we observed a modest but significant increase in the number of dendrite branches (normalized by neuron size) and shorter average branch length in comparison to control animals ( Figures 8A and 8B). We also found that dendritic branches in the RluA-1 del-HDR ddaC neurons had higher frequency of isoneuronal cross-over events compared to the control ( Figures 8C and 8D). This latter phenotype is suggestive of an isoneuronal tiling defect. Increased isoneuronal crossovers are also seen in mutants that affect dendrite attachment to the basal lamina (Han et al. 2012;Kim et al. 2012;Meltzer et al. 2016;Tenenbaum et al. 2017). Whether or not these dendrite abnormalities play a causal role in the hypersensitive nociception phenotypes of the RluA-1 mutant will be an interesting subject for future investigation.
Although loss of RluA-2 also caused hyperalgesia, its expression pattern was ubiquitous and included multidendritic neurons. Like the RluA-1 GFP exon trap, the RluA-2 GFP exon trap labeled nuclei. The nuclear localization for both RluA-1 and RluA-2 may indicate that these proteins act on RNA targets prior to export of the nucleus, or that they predominantly act upon nuclear localized RNAs. Structure based homology modeling and sequence alignments indicate that both RluA-1 and RluA-2 have 88% sequence similarity throughout their pseudouridine synthase domains. This evidence suggests that both of proteins are very likely to possess pseudouridine synthase activity but we have yet to formally prove this. In order to do so, we are working to identify potential target RNAs that could be used as substrates in experiments with purified RluA-1 and/or RluA-2 proteins.
We also observed that RluA-1 mutants showed an increase in the number of dendrite branches relative to control genotypes as well as an increase in isoneuronal crossovers. Transcription factors, cytoskeletal regulators, motor proteins, secretory pathways and cell adhesion molecules all function in concert to develop and maintain optimum dendrite morphology (Corty et al. 2009). RluA-1 may modify RNAs for those genes that regulate dendritic morphology, or changes in the dendrite morphology could be an indirect consequence of neuronal sensitivity that is regulated by RluA-1. It is noteworthy that prior studies have noted a potential link between the degree of dendrite branching and the sensitivity of nociception behaviors (Honjo et al. 2016). In other cases, axonal factors such as the ion channel SK have been found to be important in regulating the cIVda neuron excitability (Onodera et al. 2017;Walcott et al. 2018). Whether the dendrite branching phenotype that we observe in RluA-1 mutants is a cause or a consequence of hypersensitivity will be an interesting question for future studies.
shown to suppress translation termination both in vitro and in vivo (Karijolich and Yu 2011). Thus, another possible function for pseudouridylation in nociceptive neurons could be to favor read-through of pseudouridylated stop codons to generate novel sequences at protein carboxy termini.
In summary, the data presented in this study showed the RNA pseudouridine synthases RluA-1 and RluA-2 are involved in nociception in D. melanogaster. The precise underlying mechanism can only be elucidated by identifying the RNA targets of these enzymes. Several groups have developed methods using next generation sequencing methods to identify the pseudouridine sites in transcriptomes (Carlile et al. 2014;Schwartz et al. 2014;Lovejoy et al. 2014;Li et al. 2015;Khoddami et al. 2019). We anticipate that future investigations applying these methods to wild type and RluA mutants in Drosophila will help us to identify the RluA targets.

ACKNOWLEDGMENTS
The authors thank Drosophila Genomic Resource Center (supported by NIH grant 2P40OD010949) for clones. Stocks obtained from the Bloomington Drosophila Stock Center (NIH P40OD018537) were used in this study. We also thank the Stock Center in Kyoto Institute of Technology and Harvard Medical School for providing fly stocks used in this study. Members of Tracey lab provided helpful feedback on the manuscript and Jayce Brown Culbertson assisted with dendrite tracing.