Reduced function of the RNA export factor, Nxt1, in Drosophila causes muscle degeneration and lowers expression of genes with long introns and circular RNA

The RNA export pathway is essential for export-competent mRNAs to pass from the nucleus into the cytoplasm, and thus is essential for protein production and normal function of cells. Drosophila with partial loss of function of Nxt1, a core factor in the pathway, show reduced viability and male and female sterility. The male sterility has previously been shown to be caused by defects in testis-specific gene expression, particularly of genes without introns. Here we describe a specific defect in growth and maintenance of the larval muscles, leading to muscle degeneration in Nxt1 mutants. RNA-seq revealed reduced expression of many mRNAs, particularly from genes with long introns in Nxt1 mutant muscles. We further determined that circRNAs derived from these same genes are also reduced in the mutants. Despite this, the degeneration was rescued by increased expression in muscles of a single gene, the costamere component tn (abba). This is the first report of a specific role for the RNA export pathway gene Nxt1 in muscle integrity. Our data on Nxt1 links the mRNA export pathway to a global role in expression of mRNA and circRNA from common precursor genes, in vivo. Author summary In eukaryotic cells, the DNA encoding instructions for protein synthesis is located in the nucleus. It is transcribed into pre-mRNA, which is processed at both ends and spliced to remove internal spacer regions (introns) to generate mRNA. This mRNA is then transported to the cytoplasm by the mRNA export pathway via nuclear pores, and then used as a template for protein synthesis. We have previously shown that reduction in activity of a specific protein in the mRNA export pathway, Nxt1, has an additional role in testis-specific transcription. Here we describe a further role for this protein specifically in gene expression, particularly of genes with long introns and circular RNAs, and in muscle maintenance. Drosophila larvae with reduced Nxt1 activity have normal muscle pattern when they are small, but show muscular atrophy and degeneration as they grow, resulting in significant defects in their movement speed. We discovered that expression of many genes is reduced in the Nxt1 mutant larvae, but that restoring the expression of just one of these, abba, the Drosophila homologue of Trim32 (a human gene involved in muscular dystrophy) is capable of preventing the muscle degeneration. Our data provide a new insight into how defects in generally acting cellular factors can lead to specific defects similar to human diseases.


49
The formation, growth and maintenance of the musculature is critical for normal animal function, 50 and defects in these processes can lead to impaired mobility, shorter lifespan or early lethality. The 51 somatic muscular tissue of Drosophila melanogaster is generated via a highly regulated 52 developmental programme in embryogenesis, such that the first instar larva contains a stereotyped 53 pattern of muscles [1]. The abdominal A2-A7 segments each contain 30 muscles on each side of 54 the animal, each with defined size, shape, position and attachment sites. Each muscle is a single 55 multinucleated cell, derived from fusion of muscle founder cells with fusion competent myoblasts 56 (reviewed in [2]). Very severe defects in embryonic myogenesis result in a failure of the embryo to 57 hatch; less severe defects in the development of the muscle pattern or in muscle function can 58 result in animals with impaired mobility. During larval stages, no new cells are added but the 59 muscles grow extensively by addition of new myofibrils, particularly during the final larval instar 60 (reviewed in [3]). 61 Normal muscle function is essential during pupariation, and defects in larval muscles can lead to 62 lethality at this stage. About 12 hours before pupariation, a pulse of ecdysone triggers the larvae to 63 stop feeding and move out of the food. About 6-10 hours later they stop moving, contract 64 longitudinally, evert their spiracles and become a white pre-pupa. Three hours later the prepupa 65 contracts from the anterior partially withdrawing the anterior tracheal lining [4]. At this stage, an air 66 bubble forms in the abdominal cavity. The bubble gradually increases in size and the abdominal 67 tissue is forced against the body wall. One hour later, the prepupa becomes separated from the 68 puparium due to the secretion of the prepupal cuticle. The air bubble migrates to the anterior by 120 We have previously described a role for Nxt1 in regulation of transcription in Drosophila testes, 121 indicating that the level of this gene product is critical for processes beyond its known role in 122 mRNA nuclear export [19]. Here we describe a specific role for Nxt1 in the maintenance of larval 123 muscles. Specifically, Nxt1 partial loss of function animals showed muscle degeneration during the 124 extensive growth associated with the final larval instar stage. We demonstrate that this is probably 125 due to its role in the RNA export pathway as knock down of other RNA export factors caused a 126 similar muscle degeneration phenotype. We found that normal expression of many genes, 127 particularly those with long introns that are sources of circRNAs, requires Nxt1. We discovered that 128 both the mRNA and circRNA products of these Nxt1-responsive genes are reduced in mutants, 129 although the nascent transcripts of these genes are produced at normal levels. Despite the large 130 number of altered transcripts we were able to rescue the muscle degeneration of Nxt1 mutants, but 131 Nxt1 pupae have a distinctive curved shape, uneverted spiracles and fail head eversion 141 We have previously described that Nxt1 z2-0488 / Nxt1 DG05102 transheterozygotes are male and female 142 sterile, but also have significantly reduced viability [19]. Nxt1 DG05102 is a null allele caused by a P-143 element insertion into the coding sequence of the gene, while Nxt1 z2-0488 is a hypomorphic allele 144 that disrupts a hydrogen bonding network in the core of the protein thus reducing protein stability 145 [19]. Many Nxt1 z2-0488 / Nxt1 DG05102 transheterozygote pupae had a curved shape and uneverted 146 spiracles ( Figure 1A). To quantify this morphological defect we measured the axial ratios from the 147 pupa by measuring the length (excluding the posterior and anterior spiracles) and width. This 148 confirmed that the mutant pupae were significantly longer and thinner than wild type (t-test, p= 1e-149 08). Scanning electron microscopy (SEM) revealed that the surface structure of the pupal case 150 was similar between mutant and wild type ( Figure 1C-E) while in Nxt1 trans-heterozygotes, 50% 151 larvae (N=36) had uneverted anterior spiracles ( Figure 1E''). 152 We have also previously reported that the Nxt1 z2-0488 / Nxt1 DG05102 transheterozygote pupae show 153 failure in head eversion [19]. Air bubble formation and migration is implicated in head eversion and 154 therefore the viability of the pupa. To analyse this process we filmed w 1118 and Nxt1 z2-0488 / Nxt1 155 Crosses of Nxt1 DG05102 /CyO, Act-GFP to Nxt1 Z2-0488 / CyO, Act-GFP did not reveal the expected 167 Mendelian ratios of 1:2 (Nxt1 trans-heterozygotes: CyO, Act-GFP; note CyO, Act-GFP 168 homozygotes are lethal). Instead, the observed ratio was 1:10. We assessed the progression of 169 pupal development at four time points (24h, 48h, 72h and 96h pupa), and the number of viable 170 adults. Only about 20% (N=182) of the Nxt1 trans-heterozygote pupae survived through to 171 adulthood, compared to 90% (N= 287) viability for control w 1118 pupae (Supplementary Figure 2). 172 The Nxt1 DG05102 /+ showed a slight reduction to pupa viability (60% (N= 211); Supplementary Figure  173 2). Head eversion occurs roughly 12 hours into the pupa stage. Pupa metamorphosis continued 174 (such as body, bristles and wings) in Nxt1 trans-heterozygotes where most pupa did not have head 175 eversion. The consequence of no head eversion was seen between 48-96 hours into 176 metamorphosis as development stopped abruptly, and the pupae blackened. 177

179
The air bubble phenotype suggests that there are defects in the function of the larval/pupal 180 abdominal muscles, or in the developmental regulation of their function at this stage. Ecdysone 181 coordinates tissue-specific morphogenetic changes by several pulses throughout the Drosophila 182 development (reviewed in [20]). A high concentration pulse is observed immediately preceding the 183 larva-pupa transition, and this pulse is essential to trigger the muscular contractions on pupariation 184 and in the pre-pupa. To understand if a transcriptional defect downstream of ecdysone signalling is 185 responsible for the phenotype, in a manner analogous to the role for Nxt1 in regulation of testis-186 specific transcripts dependent on the transcriptional activation complex tMAC [19], we performed 187 RNA sequencing of pooled whole larvae before (wandering larvae), during (stationary larvae), and 188 after (white prepupae) the pulse of ecdysone. 189 We extracted the expression data for known ecdysone-responsive genes [21]. Out of 87 ecdysone-190 responsive genes, only four were mildly mis-regulated (Supplementary Table 1). Mutants of these 191 four genes do not show an air bubble phenotype [22,23], therefore, it is unlikely that failure of 192 expression of ecdysone-responsive genes is responsible for the air bubble phenotype in  Muscle degeneration occurs in Nxt1 mutant larvae 196 To determine whether the defects seen in pupal and pre-pupal stages in Nxt1 mutants are due to 197 muscular defects, we characterised the structure and function of muscles in the larval stages. A 198 larval movement assay was used to track 1 st , 2 nd , and 3 rd instar larvae to compare their speed 199 between stages and between w 1118 and Nxt1 trans-heterozygotes. Larvae were tracked on an agar 200 plate with a control odour on each side (Figure 2A and B). There was no significant difference in 201 the average speed of mutant larvae compared to wild type at either 1 st or 2 nd instar stage. 202 Normally, wild type 3 rd instar larvae travel up to 5x faster than 2 nd instars ( Figure 2F). In contrast, 203 Nxt1 trans-heterozygous 3 rd instar larvae were significantly slower than control animals ( Figure  204 2C), indeed, they were no faster than 2 nd instar larvae ( Figure 2G), despite being much larger. 205 To investigate the muscle structure in the mutant animals we stained of 3 rd instar larval body wall 206 muscles with phalloidin which labels F-actin. In Drosophila larvae, each of the abdominal 207 hemisegments A2-A7 has a stereotypical pattern of 30 different muscles [1]. This pattern was 208 clearly observed in the wild type ( Figure 3A and C), however, Nxt1 z2-0488 / Nxt1 DG05102 trans-209 heterozygotes showed clear signs of muscle degeneration ( Figure 3B, D and E). Defects seen 210 were variable, and included thinner muscles, fibre splits and torn muscles ( Figure 3F-K). We 211 counted the number and nature of defective muscles in 8 hemisegments of 6 control and 6 mutant 212 third instar larvae and categorized the nature of the defects. Defects seen in the mutants were 213 variable and included thinner muscles (15%), loss of sarcomeric structure (22%), degenerating 214 muscles (77%), torn muscles (8%); all control animals had no defects. 215 In addition to the gross morphological defects, Nxt1 z2-0488 / Nxt1 DG05102 trans-heterozygote mutant 216 animals also sometimes had defects in internal muscle structure. Normal muscle sarcomere 217 structure consists of thick and thin filaments, and phalloidin staining of actin reveals this structure 218 by labelling the thin filaments in the sarcomere ( Figure 4A). For Nxt1 trans-heterozygotes, the 219 sarcomere structure was compromised in 22% of the muscles examined. These muscles showed 220 more uniform phalloidin staining indicating weak or absent differentiation of thin vs thick filaments 221 ( Figure 4C). This muscle degeneration and sarcomere structure defect phenotype was not fully 222 penetrant, and some Nxt1 trans-heterozygotes had more normal sarcomere structure and less 223 obvious muscle degeneration. This is consistent with the finding that about 20% of the mutant third 224 instar larvae are able to develop to adulthood ( Figure 4B). We analysed muscles from first (N=8) 225 and second (N=10) instars with phalloidin staining. This revealed that earlier stage larvae have a 226 normal musculature, and thus that the defects are due to degeneration rather than a 227 developmental defect. 228 To confirm that the muscle phenotype observed was due to defects in Nxt1 rather than being a 229 non-specific effect, we used Mef2-Gal4 to drive UAS-RNAi expression specifically in muscles. We 230 combined this with UAS-dicer and high induction temperature to achieve a strong knockdown of 231 the transcript. Both RNAi lines 103146 (chromosome 2) and 52631 (chromosome 3) had 232 previously been shown to effectively knock down Nxt1 and phenocopy the mutant phenotype in 233 spermatocytes [19]. Phalloidin staining of these 2 nd instar larvae showed extensive muscle 234 degeneration ( Figure 5). When the temperature was 25 o C for embryonic development, with a shift 235 to 29 o C after hatching, line 52631 still induced early 2 nd instar lethality, while 103146 gave a 236 weaker phenotype of third instar larval lethality, very similar to the Nxt1 z2-0488 / Nxt1 DG05102 trans-237 heterozygote hypomorphic condition. This indicates that knock down of Nxt1 specifically in 238 muscles is able to phenocopy the Nxt1 mutant situation, and shows that the degeneration is due to 239 reduction in Nxt1 activity in muscles. 240 We expressed GFP-Nxt1 via the UAS-gal4 system both exclusively in muscles at a high level (with 241 mef2-gal4>UAS-GFP-Nxt1) and ubiquitously at a lower level (with arm-gal4>UAS-GFP-Nxt1) in 242 Nxt1 trans-heterozygotes. High level, muscle-specific, GFP-Nxt1 expression was able to partially 243 rescue muscle integrity in Nxt1 trans-heterozygotes (~13-26% with two 47% outliers (N= 10) 244 Figure 6C) and partially rescue the pupa lethality (~40% (N= 186); Figure 6A). On average, the 245 axial ratios were similar to wild type, albeit more variable ( Figure 6B). Finally, mobility was 246 increased compared to Nxt1 trans-heterozygotes, but was still significantly reduced compared to 247 wild type (Student's t-test * p<0.05; Figure 6D). Lower level ubiquitous expression of GFP-Nxt1 in 248 Nxt1 trans-heterozygotes resulted in improved, but not fully rescued muscle integrity (~3-33% 249 damage with one 73% outlier (N= 9), increased the pupa viability (~75% (N= 163)) and increased 250 larval mobility. The axial ratios were similar to wild type ( Figure 6B). 251 252 Muscle degeneration is associated with larval growth 253 During the last instar phase, larvae will grow substantially compared to earlier instars. Larvae 254 removed from the food 70 hours after egg laying do not grow, but still crawl until the normal time 255 for pupation (~112 hours AEL) and may survive through to adulthood, generating very small flies 256 [24]. Larvae at 70 hours AEL are late 2 nd , or early 3 rd instars. 257 To test whether growth or use (movement) is implicated in the muscle degeneration in Nxt1 258 mutants, we fed ~60 larvae for 70-73 hours then removed them from the food. 14-25% of the 259 larvae (removed from food at 70-73 hr) were able to pupate although less than 5% of those pupae 260 were able to emerge as adults. We simultaneously examined ~60 of the Nxt1 trans-heterozygote 261 larvae that were starved from 70-hour AEL and found, that only a few larvae pupated and none 262 emerged as adults (similar to w 1118 ). The pupae that formed had everted spiracles ( Figure 7A) and 263 resembled the wild type controls. Interestingly, the larvae that failed to pupate were able to survive 264 for several additional days as larvae before dying. Nxt1 trans-heterozygote larvae were dissected 265 four days after they were removed from the food, and were stained with phalloidin. For all larvae 266 (n=10), no muscle degeneration was observed ( Figure 7B). These larvae had been moving 267 normally and thus using their body wall muscles for the four days of starvation. The lack of 268 abnormalities in these animals indicates that the muscle growth, or high levels of force generation 269 associated with third instar larval movement, rather than use per se, is critical for degeneration in 270  Nxt1's well characterised function is in the RNA export pathway. To determine whether the muscle 275 degeneration phenotype is caused by defects in this pathway, we assessed the phenotype of 276 knock down of other RNA export pathway genes. We used arm-gal4, to drive RNAi hairpin 277 constructs targeted against thoc5, sbr (Nxf1), Ref1 and Hel25E (UAP56). This is designed to 278 reduce, but not complete eliminate the RNAi target gene expression throughout the larva. As a 279 positive control we used RNAi against Nxt1, and as a negative control we used both w 1118 and arm-280 gal4 alone. RNAi against Hel25E caused early larval lethality, even when tested at lower 281 temperature, so we were not able to assess the role of this gene in muscle maintenance. For all 282 the other genotypes, we assayed third instar muscle integrity with phalloidin staining. Knock down 283 of these RNA export pathway components caused a similar phenotype to knock down of Nxt1, with 284 muscle thinning and tearing apparent ( Figure 8B-E'). Control larvae (arm-gal4 alone and w 1118 ) had 285 normal musculature ( Figure 8A and A' and data not shown). This indicates that the role of Nxt1 in muscle maintenance is most likely to be attributable to its role in the RNA export pathway rather 287 than an un-related function. 288 289 Genes with long introns that also produce circRNAs are sensitive to the loss of Nxt1 290 The known role of Nxt1 in RNA export and transcriptional control suggested that the muscle 291 phenotype could be due to transcriptional or post-transcriptional defects in gene expression. To 292 identify somatically expressed differentially expressed genes, we performed RNA sequencing from 293 stationary third instar larval carcasses (comprising primarily muscle and cuticle). Each genotype 294 (w 1118 and Nxt1 trans-heterozygotes) was analysed in triplicate. 572 genes were 2x or more up-295 regulated in mutant compared to wild type, while 1340 genes were similarly down-regulated. The 296 numbers of genes up-and down-regulated at different fold change cut offs are shown in Table 1. 297 To investigate properties of differentially expressed genes we focussed on aspects of RNA 298 processing in the nucleus, and particularly intron length. Nxt1 is important for mRNA export, is 299 recruited indirectly by EJC to spliced transcripts [25], and mutations in the EJC can cause reduced 300 accumulation of transcripts with long introns [26]. In contrast, in Nxt1 trans-heterozygote testes, 301 transcripts from short and intron-less genes were dramatically reduced. In Nxt1 trans-heterozygote 302 testes, addition of at least one intron to a down-regulated gene resulted in increased transcript 303 production [19]. We initially analysed the total intron length, number of introns and smallest/largest 304 intron. This revealed that, in direct contrast to the situation in testes, genes down-regulated in Nxt1 305 mutant carcass had more introns and a higher total intron length than non-differentially expressed 306 genes, while up-regulated genes had fewer introns and a lower total intron length than the genes 307 that were not differentially expressed (Mann-Whitney test p-value <0.05, except for 16-fold up 308 regulated (p value of 0.16)); (Table 1 and Figure 9). We extended the analysis by also looking at 309 the gene length and shortest/longest transcripts (Table 2). Again, a clear trend showed that down-310 regulated genes had longer median mRNA length and up-regulated genes had shorter median 311 mRNA length when compared to non-differentially expressed genes (Mann-Whitney p<0.05; Table  312 2). Finally, the genes down regulated in Nxt1 mutants produced more mRNA isoforms than those 313 up regulated or not differentially expressed (Table 2). qRT-PCR for 12 genes with long introns 314 confirmed the down-regulation seen in the RNA sequencing data ( Figure 10). 315 Long introns present a challenge to the spliceosome, and indeed some long introns are spliced as 316 several shorter fragments via recursive splicing [27]. Expression of most genes known to be 317 processed via recursive splicing was detected in our RNA-seq sample (112/116). Only 11 of these 318 were 8x or more down-regulated in the Nxt1 mutants (34 were 4x or more down-regulated; Figure  319 9). We concluded that Nxt1 is unlikely to be required for recursive splicing per se, and considered 320 other aspects of genes with long introns. 321 Genes with long introns are sources of circular RNAs, with many circular RNA exons being flanked 322 by long introns [10]. We therefore compared the genes down-regulated in Nxt1 muscles with lists of genes known to produce circRNA transcripts [13], and found a striking overlap. 466 of the 567 324 genes 4x or more down-regulated in Nxt1 muscles had at least one read consistent with a 325 circRNA. 199 of these genes were in the higher confidence set associated with at least 10 circRNA 326 reads. 57 of the 186 genes that were at least 8x down regulated in Nxt1 muscles were also in the 327 high confidence circRNA set. circRNAs are predominantly produced from genes that are known to 328 have neuronal functions and whose expression is higher in nervous system, even when the 329 analysis is performed on tissues outside the nervous system [13]. Consistent with this, embryonic 330 expression pattern enrichment analysis via FlyMine revealed that the genes down-regulated in 331 Nxt1 muscles are indeed significantly enriched for expression in nervous-system (such as ventral 332 nerve cord (p=1.13e-8), embryonic brain (p=3.06e-6)). We conclude that Nxt1 is important not only 333 for the export of mRNAs but also for normal expression of mRNAs from genes with long introns, 334 particularly those that produce circRNA products. 335 336 CircRNAs are reduced in Nxt1 mutants, but nascent transcripts are not 337 mRNAs and circRNAs are mutually exclusive products made from the same nascent transcript. If 338 Nxt1 and the RNA export pathway controls the ratio between mRNAs and circRNAs, down 339 regulation of Nxt1 would lead to an increase in circRNA products. Alternatively, Nxt1 could be 340 implicated in the stability of the nascent RNA to ensure normal levels of both products, in which 341 case we would expect a reduction in circRNA in target genes in the mutants. In a third model, Nxt1 342 could act after completion of splicing to stabilise just the mRNA; in this case the circRNA levels 343 would be expected not to change in the mutant compared to wild type. To determine which of 344 these models is most likely, we sequenced both total RNA (after depletion of the ribosomal RNAs) 345 and circRNA (the RNAse R resistant fraction) from larval carcass samples from wild type and Nxt1 346 transheterozygotes. The library preparation protocol did not include a selection for poly adenylated 347 transcripts, and thus allowed us to examine pre-mRNA, both spliced and unspliced, as well as 348 circRNA. 349 We first assessed the levels of circRNAs, focussing only on 336 structures identified in both 350 genotypes with high confidence (backspliced reads identified in at least 2/3 replicates). This 351 revealed that the majority of circRNAs are reduced in expression in Nxt1 compared to wild type 352 ( Figure 11A-D). For these genes, pre-mRNA levels are unchanged in between wild type and 353 mutants, while poly adenylated mRNA levels are reduced in mutants ( Figure 11). Similarly, we 354 found no difference between levels of spliced RNAs in when we assessed total RNA samples 355 ( Figure 11A-D). Thus, transcription of these genes, and splicing, is not affected in the mutants; 356 however accumulation of poly adenylated transcript is reduced. 357 Increased expression of abba rescues the muscle degeneration, but not semi-lethality 359 Having identified a globlal defect in mRNA and circRNA for many genes, we were interested in 360 understanding how this leads to the muscle degeneration phenotype. The RNA expression of 13 361 known muscle-specific or muscle-enriched genes was analysed in our original whole larva mRNA-362 seq data (Supplementary Table 2). Nine genes were more than 1.5-fold up regulated. abba (also 363 known as thin (tn)), the only gene significantly down regulated in this list, is an essential 364 TRIM/RBCC protein that maintains integrity of sarcomeric cytoarchitecture [28]. abba loss of 365 function mutations are lethal, with the larvae and pupae being long and thin and having muscle 366 degeneration [29]. abba is a large gene with several long introns, which also produces a circRNA 367 derived from circularisation of exon 7 ( Figure 11E). The encoded protein has a RING finger, B-Box 368 & CC-domain and several NHL repeats ( Figure 12A). qRT-PCR confirmed the reduction in abba 369 expression (n=30 larvae per sample; Figure 12). Levels of abba mRNA were highly variable 370 between individual stationary larvae ( Figure 12) but significantly down in 9 out of 10 larvae, when 371 compared to a pool of 10 WT larvae ( Figure 12B). abba has at least six isoforms with 13 exons in 372 its longest transcript and contains introns up to 4kb in length. We compared nascent (primers in 373 introns) versus spliced (primers in adjacent exons) abba transcript expression by Q-RT-PCR. Four 374 different regions were selected that would target as many isoforms as possible (Supplementary 375 Figure 3A). Consistent with our total RNA sequencing results, all regions throughout abba had 376 similar expression of nascent transcript in mutant compared to control animals, but the levels of 377 spliced transcript were reduced in mutants compared to control (Supplementary Figure 3B). Thus, 378 loss of Nxt1 does not impact on abba transcription, but Nxt1 is important for normal accumulation 379 of the processed mRNA. 380

381
To determine whether reduction in the expression of abba was implicated in the mutant phenotype, 382 we used a UAS-abba full length (fl) cDNA construct (kindly gifted by Hanh T. Nguyen; hereafter referred 383 to as UAS-abba) and expressed it under the muscle specific driver mef2-gal4 in the Nxt1 trans-384 heterozygote background. Staining of the muscles of these larvae with phalloidin, showed that 385 muscle degeneration was fully rescued in 12 out of 16 animals and partially in the other four 386 ( Figure 12C). qRT-PCR of individual stationary larvae showed again high variability between 387 individuals, but 8 out of 10 had abba expression equal to or exceeding that seen in wild type 388 ( Figure 12D). 389 This result is surprising given the large number of genes whose expression is altered in the mutant 390 larvae. We therefore examined the ability of these animals to survive to adulthood and found that 391 the lethality was not rescued (22%; N=187). These genetic rescue experiments indicate that the 392 muscle defect in Nxt1 mutant larvae is primarily due to the reduction in abba mRNA expression. 393 However, the pupal lethality of most Nxt1 transheterozygotes is not solely caused by the muscular degeneration, and reduced expression of other target genes is likely associated with reduction in 395 The defects in pupal morphology were caused by muscle degeneration during the third instar larval 413 stage. Second instar larvae had normal muscles pattern, morphology and mobility. This indicates 414 that the establishment of the larval musculature, which occurs during embryonic development, is 415 not affected by the reduction in Nxt1 in the hypomorphic allele. 416 Muscles are composed of tandem arrays of sarcomeres containing thick and thin filaments, 417 arranged in myofibrils. When a muscle twitches, the filaments slide past each other in response to 418 calcium release from the sarcoplasmic reticulum (SR) resulting in force generation. Our 419 examination of the muscle integrity in transheterozygote larvae revealed a variety of defects 420 including muscle atrophy (thinning) and loss of integrity (tearing and splitting). Frequently, we saw 421 loss of filamentous actin in the middle of the muscle while the ends had f-actin; very occasionally 422 we saw balls of muscles associated with loss of attachment or catastrophic failure of muscle 423 integrity and muscle severing. Even when the muscle shape was unaffected, we found a loss of 424 normal internal architecture with disruption of the sarcomeric arrays. Often the muscle was 425 present, but there was filamentous actin staining only towards the ends of the muscle and the 426 central region lacked the sarcomeric structure. It is likely that the reduction in larval mobility, the 427 axial ratio defect, the failure of spiracle eversion and the failure of air bubble movement are all a 428 direct result of the degeneration of muscles in the mutant larvae. All these processes require 429 efficient and coordinated muscle contractions, which are compromised by lack of Nxt1. The curved 430 shape of many mutant pupae is likely to be due to the variability in extent of muscle degeneration: if an individual has more damage on one side than the other, particularly of the longitudinal 432 muscles, there will be unequal contraction as the prepupa forms and one side of the body will 433 shorten more. 434 During the larval phase, the muscles undergo a ~50-fold increase in fibre size. This dramatic 435 growth occurs without addition of new cells, as the number of nuclei in the muscle syncytium 436 remains constant [3]. Increased DNA content is achieved via endoreduplication, and increased 437 muscle volume is driven by cell growth associated with new myofibril and sarcomere assembly [3]. 438 The majority of this growth occurs during the third instar larval stage, and depends on nutrient 439 supply and sensing of this via the Insulin/Akt/Tor pathway [32]. Deficits in muscle growth, for 440 example caused by reducing endoreduplication through over expression of cyclinE in muscle, have 441 non-autonomous effects on the growth of the whole larva [32]. We did not detect any difference in 442 the overall size of the third instar mutant larvae, indicating that the defects are not due to defects in 443 nutrient supply or sensing. When Nxt1 hypomorphic larvae are starved from the late 2 nd or early 3 rd 444 instar larval stage they remain alive and continue to move, but growth is blocked. This treatment 445 was sufficient to rescue the muscle degeneration that typically occurs during the 3 rd instar phase. 446 This indicates that the primary cause of the degeneration is defective (re)-organisation during 447 muscle growth, rather than damage caused by use of the muscles, although we cannot rule out 448 that the higher forces required for third instar larval movement are not also implicated. 449 450 Muscle degeneration can be rescued by increasing abba 451 We were surprised that a hypomorphic allele of a pleiotropic factor such as Nxt1 had such a 452 specific phenotype of muscle degeneration. We reasoned that, while it is likely to be important for 453 the normal expression of many genes, reduction in just one or a few crucial genes could underpin 454 the muscle defects. Our initial RNA-seq of whole larvae was a mixed sex population, and 455 unfortunately the majority of genes differentially expressed between samples were those with 456 testes-specific or testis-enriched expression. This is consistent with our previous findings that Nxt1 457 is critical for expression of genes regulated by tMAC in testes [19], but meant that the signal from 458 relatively mild expression changes from somatic tissues was less apparent. Nevertheless, one 459 gene, abba, with a known role in muscles stood out as being mildly down-regulated in stationary 460 larvae. The phenotype of abba mutants is strikingly similar to that we describe for Nxt1 461 hypomorphs, particularly with thinner muscles, long thin pupae and defects in both spiracle 462 eversion and air bubble migration [28,29]. Abba is a TRIM/RBCC protein involved in maintaining 463 the integrity of sarcomeric cytoarchitecture [28,29], and is the Drosophila melanogaster 464 homologue of human Trim32, defects in which cause limb girdle muscular dystrophy 2H [15]. 465 Trim32 is localised to the costamere, which overlies the Z-disk and ensures attachment of the 466 sarcomere to the overlying extracellular matrix via the dystrophin glycoprotein complex. 467 The phenotypes of defective sarcomere structure, fraying and muscular atrophy are all consistent 468 with reduction in abba (costamere) function. The growth of muscles involves the generation of new 469 sarcomeric units, each requiring a new costamere and thus new Abba production and 470 incorporation. The gradual decline in muscle integrity seen during growth may be attributable to 471 reduction in Abba protein levels meaning that the stability of newly formed sarcomeres is reduced. 472 Indeed, in mouse, the expression of Trim32 is upregulated in muscles that are remodelling [33]. 473 Consistent with this model, increasing expression of abba using a cDNA construct driven with the 474 Gal4/UAS system is sufficient to rescue the muscle defects seen in Nxt1 mutants. Starvation from 475 late second or early third instar larval stage is also sufficient to rescue the muscle defect. In this 476 case, there is no muscle growth, so no new sarcomeric units need to be added, the existing 477 structures are maintained, and thus there is no muscle degeneration. The rescue by a cDNA also 478 indicates that it is the reduction in abba mRNA, rather than a reduction in abba circRNA that is 479 important for the muscle phenotype, as the cDNA construct cannot generate the circRNA. 480 RNAi constructs to disrupt Nxt1 function in muscles caused degeneration, again without pattern 481 defects even in second instar larvae, indicating that the precise level of Nxt1 function is critical. 482 Very high level, muscle specific, RNAi construct expression, driven from embryonic stages, 483 presumably reduces the level of active Nxt1 in muscles in embryos and early larval stages even 484 beyond that found in the hypomorphic allele combination. In this situation, there would be reduced 485 Abba both during embryonic muscle formation and at the early growth stages, and this would 486 impact on the muscle integrity even before the extensive third instar larval growth period. Driving 487 Nxt1 RNAi ubiquitously but at a lower level, phenocopies the Nxt1 transheterozygote situation, with 488 viability to third instar stage and muscle degeneration in these animals. Driving RNAi targeted 489 against other RNA export pathway factors ubiquitously, with arm-gal4, also phenocopies the Nxt1 490 transheterozygote situation. Importantly, these experiments, coupled with the partial rescue of the 491 muscle defects by expression of Nxt1, confirm that the phenotype is caused by reduction in Nxt1 492 function rather than by a different gene or by a neomorphic effect of the point mutation in Nxt1 Z2-493 0488 . Additionally, these experiments confirm that the role of Nxt1 in muscle maintenance is due to 494 its function in the RNA export pathway, rather than this being a moonlighting function for the 495 protein. Transcripts with many and large introns are more sensitive to the loss of Nxt1, which is consistent 501 genes with long introns being sensitive to the loss of the EJC [34]. In direct contrast, genes without 502 introns were particularly sensitive to loss of Nxt1 in testes [19]. RNA-seq data from larval 503 carcasses, which is highly enriched for larval muscles, allowed us to specifically examine the role of Nxt1 in transcript expression in a somatic tissue in which we have shown its function is critical. 505 This revealed a significant relationship between gene expression in Nxt1 mutants and intron length 506 for both down and up regulated genes. For down-regulated genes, the more dramatic the down-507 regulation, the longer the total intron length. Similarly, for up regulated genes, the more up 508 regulated the gene, the shorter the total intron length. No strong changes in EJC component 509 transcripts were found in either whole larvae and larval carcasses. Therefore, it is unlikely that 510 defects in EJC component expression levels are implicated in the phenotype. However clearly the 511 ratio between EJC, Nxf1 and Nxt1 will be affected when the level of active Nxt1 is reduced. During 512 primary transcript processing, the spliceosome removes the intron and then the EJC is recruited to 513 is less efficient on transcripts with long introns, particularly if these are being processed to make 519 circRNAs in addition to mRNAs, then these transcripts could be preferentially sensitive to the Nxt1 520 trans-heterozygotes where the availability of Nxt1 protein is limiting. 521 522 Circular RNAs are a relatively recently described class of RNAs that are produced as alternative 523 products from the same primary transcripts as mRNAs. Global analysis of circRNA abundance 524 reveals that many genes can encode circRNAs, but the propensity for a gene to produce a circular 525 transcript is related to the length of the introns, particularly those that flank the back-spliced 526 exon(s) [10]. Our RNA-seq analysis suggests that the mature mRNAs from genes that also 527 produce circRNAs is reduced in Nxt1 mutants. Additionally, we found that the circRNAs 528 themselves are reduced in the mutants. Mutation of exon junction complex components has also 529 been shown to have a preferential effect on mRNA production from genes with long introns [26]. At 530 least some genes down-regulated in the EJC knock down have been shown to have aberrant 531 splicing patterns in the absence of EJC [26]. In contrast, we found few defects in alternative 532 splicing in Nxt1 mutants. However, it is also interesting to note that 249/315 genes down-regulated 533 after knock down of the EJC were also on the list of genes that produce at least one circRNA, and 534 99 are on the higher confidence list of 10 or more circRNAs [13,26]. The RNA export pathway has 535 already been shown to be linked to 3' end processing and poly adenylation. TREX subunit THOC5 536 is recruited to target transcript 3'UTRs by poly adenylation specific factor 100 [38]. Ref association 537 with RNAs is promoted by the 5' cap, by the EJC, and, crucially, also in the 3' UTR by nuclear 538 polyA binding protein (PABPN1) [39]. In light of this, it is likely that the EJC and the export adapter 539 Nxt1 (and other factors in the RNA export pathway) are particularly important in processing of 540 transcripts that are alternatively spliced to produce mRNA and circular RNA outputs. We found that 541 pre-mRNA levels of these target genes are not reduced in the mutants, while mature, poly adenlyated mRNAs are, suggesting that the EJC and export factors are acting to ensure 543 stabilisation of the transcripts until completion of 3' end processing. 544 545 RNA regulation, and particularly RNA splicing has been linked to muscular degenerative diseases 546 [17]. Sequestration of the splicing regulator MBNL1 by RNA containing expanded CUG or CCUG 547 repeats reduces functional MBNL and results in aberant splicing of downstream targets in DM 548 patients [16]. The Drosophila orthologue of MBNL1, mbl, produces an abundant circRNA similarly 549 one of the many MBNL1 splice variants in humans is a circRNA [11]. mbl is expressed in both 550 muscles and the nervous system, and is one of the many circRNA-producing genes whose mRNA 551 is reduced (approximately 10x) in Nxt1 mutant larvae. The production of the mbl circRNA is 552 regulated by Mbl protein itself, in an autoregulatory feedback loop. In vertebrates, it appears that 553 MBNL1 does not regulate production of its own circRNA in neuronal cells, but the effect in muscle 554 has not been determined [18,40]. Our discovery that defects in an RNA export factor, Nxt1, can 555 generate a muscular dystrophy phenotype and is associated with differential expression of specific 556 mRNAs and circRNAs in Drosophila melanogaster suggests that investigation of the RNA export 557 pathway, and circRNAs, in context of the disease is warranted. 558 559

560
Drosophila culture, strains and genetics.

RNA extraction, cDNA synthesis and Quantitative PCR 578
Total RNA was extracted using Trizol (ThermoFisher Scientific) and then further cleaned up with 579 the RNeasy Mini Kit (Qiagen) according to manufacturer's protocol. The DNaseI step was 580 included. RNA was quantified with a nanodrop ND-1000 (ThermoFisher Scientific) and stored at -581 80 o C. For larval carcass RNA sequencing, 30 carcasses were used per replicate in triplicates. For 582 qRT-PCR, either a single carcass or a mix of up to 10 carcasses were used as detailed in the 583 results. cDNA was generated using 100ng total RNA and oligo dT primers with the Superscript III 584 kit (Invitrogen). The cDNA reaction was diluted to 60 or 120µl with dH 2 O, and 1µl of this cDNA was 585 further diluted with 7µl dH 2 O to use as a template in the qRT-PCR reactions. The larva was cleaned on the magnetic chamber by removing all internal organs carefully. All Ca 2+ 614 saline, HL-3 solution was removed with a P-100 pipette and fresh drops were added two more 615 times while cleaning the larval carcass. The larval carcass was fixated by adding fresh 4% 616 paraformaldehyde in PBS for 2 minutes. All the pins were removed from the larval carcass and the 617 sample was transferred to a glass well with 100µl 4% paraformaldehyde for another hour. The 618 fixative was removed, and the larval carcass was washed twice with 100µl PBS-T (0.1% Triton X-619 100) for 5 minutes each. Two drops of Alexa Fluor TM 488 Phalloidin (ThermoFisher) was added to 620 1ml of PBS-T. 100µl Alexa Fluor TM 488 Phalloidin mix was added to the well to stain F-actin in 621 larval muscles for 1 hour. Alternatively, FITC phalloidin was used at a final concentration of 1 µg/ml 622 in PBS-T. Phalloidin solution was removed and the larval carcass was washed with PBS-T two 623 times for 5 minutes each. The larva carcass was put on a microscope slide and mounted in 85% 624 glycerol + 2.5% n-propyl gallate. Images were made on an Olympus BX 50 (Olympus) microscope 625 and Hamamatsu ORCA-05G digital camera. For a close-up of the larval muscle sarcomeres 626 pattern, images were taken with a Leica DM6000B upright microscope with HC PL Fluotar 627 20x/0.50 and HCX PL APO 40x/1.25 oil objectives. 628 629

Criteria for scoring muscle defects in larvae 630
The integrity of larva muscles was calculated as a percentage from a total of 8 hemisegments. The 631 hemisegments A2-A5 were in the abdominal area and were not damaged by any of the pins. Each 632 hemisegment contained 30 different muscles, so a total of 240 muscles were inspected. The 633 integrity of a muscle is compromised if the muscle is damaged in any way, such as being torn, thin, 634 loss of sarcomeric structure or missing. The muscle damage percentage is the total number of 635 damaged muscles divided by 240 and multiplied by 100. 636 637

Starvation of larvae 638
Larvae were fed up to 70, 71, 72 and 73 hours after egg laying (AEL) before removal from the 639 food. The larvae were transferred to a petri dish with a moist filter paper in it to prevent desiccation. 640 Larvae in the petri dish were checked at regular intervals to ensure the filter paper was moist and 641 keep track of the progress of the metamorphosis. Larvae deprived of food 70 hours AEL that 642 survived for four days were analysed for muscle integrity.  High rates of duplicated and multi-mapping reads were observed in all samples. Upon thorough 729 investigation, we concluded that those were not artificial. Duplication rates were due to a very high 730 abundance of a small number of genes, and was linked to multi-mappers (i.e., multi-mapping reads 731 were also duplicated). Multi-mapping was due to a small rRNA contamination, and highly 732 expressed small RNA (e.g. cuticle gene Cpr49Ac) or RNA with repeated conserved domains (e.g. 733 cuticle genes lcp1 and lcp2). Differential expression analyses were also run with and without 734 deduplication and led to the same conclusions. Multi-mappers were excluded from further analyses 735 but duplicated reads were kept. Paired-end reads were merged prior to analysis. CircRNAs were identified using PTESFinder v.1 747 [51] (parameters -s 65 -u) with alignments to the Drosophila genome (Dm6.23) and the UCSC 748 Refseq transcriptome. Only those circRNAs that were identified in at least two RNAseR-treated 749 biological replicates were included in further analyses. This is a stringent criterion given the high 750 variability between RNase R-treated replicates and the sequencing depth [52]. 336 circRNA 751 structures were identified in at least two replicates of both WT and Nxt1 samples and are included 752 in further analyses. 753 As a measure of circRNA abundance, we used the number of back-splice per million mapped 755 reads (BPM), defined as: # reads supporting each structure generated by PTESFinder / total # 756 mapped reads (junctions+canonical) *10^6. We averaged the two highest biological replicates. The 757 corresponding two mock (non RNAseR treated) replicates were used to quantify the spliced RNA 758 (but not necessarily poly adenylated) and pre-mRNA of the host transcript. Plots were generated in 759 R with the ggplot2 package.    Nxt1 trans-heterozygote pupae were smaller, but had normal morphology when food was withheld 831 from 70h AEL. Three food-deprived and three normal feeding animals are shown. B) Nxt1 trans-832 heterozygote 70h AEL were examined for muscle degeneration. Individuals that had not pupated 833 four days after no feeding still had intact muscles (N= 10).  genes, coloured to show total intron length. Genes with long introns are typically expressed at 844 lower levels in the mutant than in the wild type sample. B) Genes (red in panel A) whose total 845 intron length is >50kb, C) Genes (green in panel A) whose total intron length is between 5 and 15 846 kb, D) Genes (blue in panel A) whose total intron length is <5 kb. 847     min."speed" mean"speed" max."speed"