Ribosome profiling reveals post-translational signaling mechanisms drive the retrograde enhancement of presynaptic efficacy

Presynaptic efficacy can be modulated by retrograde control mechanisms, but the nature of these complex signaling systems remain obscure. We have developed and optimized a tissue specific ribosome profiling approach in Drosophila. We first demonstrate the ability of this technology to define genome-wide translational regulations. We then leverage this technology to test the relative contributions of transcriptional, translational, and post-translational mechanisms in the postsynaptic muscle that orchestrate the retrograde control of presynaptic function. Surprisingly, we find no changes in transcription or translation are necessary to enable retrograde homeostatic signaling. Rather, post-translational mechanisms appear to ultimately gate instructive retrograde communication. Finally, we find that a global increase in translation induces adaptive responses in both transcription and translation of protein chaperones and degradation factors to promote cellular proteostasis. Together, this demonstrates the power of ribosome profiling to define transcriptional, translational, and post-translational mechanisms driving retrograde signaling during adaptive plasticity. AUTHOR SUMMARY Recent advances in next-generation sequencing approaches have revolutionized our understanding of transcriptional expression in diverse systems. However, transcriptional expression alone does not necessarily report gene translation, the process of ultimate importance in understanding cellular function. To circumvent this limitation, biochemical tagging of ribosomes and isolation of ribosomally-associated mRNA has been developed. However, this approach, called TRAP, has been shown to lack quantitative resolution compared to a superior technology, ribosome profiling, which quantifies the number of ribosomes associated with each mRNA. Ribosome profiling typically requires large quantities of starting material, limiting progress in developing tissue-specific approaches. Here, we have developed the first tissue specific ribosome profiling system in Drosophila to reveal genome-wide changes in translation. We first demonstrate successful ribosome profiling from a specific tissue, muscle, with superior resolution compared to TRAP. We then use transcriptional and ribosome profiling to define transcriptional and translational adaptions necessary for synaptic signaling at the neuromuscular junction. Finally, we utilize ribosome profiling to demonstrate adaptive changes in cellular translation following cellular stress to muscle tissue. Together, this now enables the power of Drosophila genetics to be leveraged with translational profiling in specific tissues.


141
However, despite the potential of ribosome profiling, this approach has not been developed for 142 tissue-specific applications in Drosophila, nor brought to the study of retrograde homeostatic 143 signaling.

144
We have developed and optimized a streamlined system for ribosome profiling of 145 specific tissues in Drosophila. We first demonstrate the success of this approach in defining 146 translational regulation in the larval muscle, and reveal dynamics in translation that are distinct 147 from overall transcriptional expression. Next, we highlight the superior sensitivity of ribosome 148 profiling in reporting translational regulation over the conventional TRAP method. We go on to

186
To define genome-wide changes in mRNA transcription and translation in the 187 postsynaptic muscle that may be necessary for PHP signaling, we sought to purify RNA from 3 rd 188 instar larvae muscle in wild type, GluRIIA mutants, and Tor-OE (Fig. 1E). We then sought to 189 define mRNA expression through three methods: Transcriptional profiling, translational profiling 190 using translating ribosome affinity purification (TRAP), and ribosome profiling (Fig. 1F). First, 191 transcriptional profiling of total mRNA expression can be performed by isolating RNA from 192 dissected third instar muscle and prepared for RNA-seq through standard methods (Brown et  tissues. Our next objective was to optimize a tissue-specific ribosome profiling approach.

207
Optimization of a tissue-specific ribosome profiling approach in Drosophila 208 Ribosome profiling is a powerful approach for measuring genome-wide changes in mRNA 209 translation rates. However, high quantities of starting material is necessary to obtain sufficient 210 amounts of ribosome protected mRNA fragments for the subsequent processing steps involved 211 (Brar and Weissman, 2015). Since this approach has not been developed for Drosophila 212 tissues, we first engineered and optimized the processing steps necessary to enable highly 213 efficient affinity purification of ribosomes and ribosome protected mRNA fragments by 214 incorporating ribosome affinity purification into the ribosome profiling protocol.

215
Although tissue-specific ribosome affinity purification strategies have been developed , we selected an alternative ribosomal protein from the 227 large and small subunits expected to have C terminals exposed on the ribosome surface. We 228 cloned the Drosophila homologs of these subunits, RpL3 and RpS13, in frame with a C-terminal 10 3xFlag tag and inserted this sequence into the pACU2 vector for high expression under UAS 230 control (Han et al., 2011). We then determined whether intact ribosomes could be isolated in 231 muscle tissue following expression of the tagged ribosomal subunit. We drove expression of 232 UAS-RpL3-Flag or UAS-RpS13-Flag with a muscle-specific Gal4 driver (BG57-Gal4) and 233 performed anti-Flag immunoprecipitations ( Fig. 2A). An array of specific bands were detected in 234 a Commassie stained gel from the RpL3-Flag and RpS13-Flag immunoprecipitations, but no 235 such bands were observed in lysates from wild type (Fig. 2B). Importantly, identical sized bands 236 were observed in immunoprecipitates from both RpL3-Flag and RpS13-Flag animals, matching 237 the expected distribution of ribosomal proteins (Anger et al., 2013). The RPL3-Flag 238 immunoprecipitation showed the same distribution as RpS13 but higher band intensity,

239
indicating higher purification efficiency, so we used RpL3-Flag transgenic animals for the 240 remaining experiments. In addition to ribosomal proteins, the other major constituent of intact 241 ribosomes is ribosomal RNA. Significant amounts of ribosomal RNA were detected in an 242 agarose gel from RpL3-Flag immunoprecipitates (Fig. 2C), providing additional independent 243 evidence that this affinity purification strategy was efficient at purifying intact ribosomes.

244
Next, we tested the ability of RpL3-Flag to functionally integrate into intact ribosomes.

245
We generated an RpL3-Flag transgene under control of the endogenous promotor (genomic-

251
Thus, biochemical tagging of RpL3 does not disrupt its localization or ability to functionally 252 integrate into endogenous ribosomes.

253
Finally, we developed and optimized a method to process the isolated ribosomes to 254 generate only ribosome protected mRNA fragments to be used for RNA-seq analysis. First, we 11 digested the tissue lysate with RNaseT1, an enzyme that cuts single stranded RNA at G 256 residues (Fig. 2D). Following digestion, we ran RNA on a high percentage PAGE gel, excising 257 the mRNA fragments protected from digestion by ribosome binding (30-45 nucleotides in length;

282
Although both translational (TRAP) and ribosome profiling approaches can report TE, ribosome 283 profiling should, in principle, exhibit superior sensitivity in revealing translational dynamics. We 284 therefore compared translational and ribosome profiling directly to test this prediction.

285
We compared translation efficiency by comparing TRAP and ribosome profiling to 286 transcriptional profiling in wild-type muscle. In particular, we tested whether differences were 287 apparent in the number of genes revealed to be translationally suppressed or enhanced relative 288 to transcriptional level through ribosome profiling compared to TRAP. We first analyzed the

296
Further, we subdivided all measured genes into three categories: high TE, medium TE, and low 297 TE. These groups were based on translation efficiency as measured by ribosome profiling or 298 TRAP, with high TE genes having a TE value >2, representing genes that are translationally 299 enhanced relative to transcriptional level; low TE genes having a TE value <0.5, representing 300 genes translationally suppressed relative to transcriptional level; and medium TE genes having 301 a TE between 0.5 and 2; representing genes not under strong translational regulation. This 302 division revealed a higher number of genes in the high and low TE groups detected by ribosome 303 profiling compared to TRAP (Fig. 3C). Together, these results are consistent with ribosome 304 profiling detecting more genes under translational regulation compared to TRAP.

305
We next investigated the genes under significant translational regulation (genes with 306 high TE or low TE), detected through either ribosome profiling or TRAP, to determine whether 13 differences exist in the amplitude of translational regulation. Specifically, genes were divided 308 into the three categories mentioned above based on the average translation efficiency 309 measured by ribosome profiling and TRAP. We then determined the TE value ratio from 310 ribosome profiling compared to TRAP within the three categories. A ratio above 0 (log2 311 transformed) in the high TE group indicates a more sensitive reporting of translation for 312 ribosomal profiling, while a ratio below 0 in the low TE group would also indicate superior 313 sensitivity for the ribosomal profiling approach. This investigation revealed an average ratio of 314 0.28 within the high TE group, -0.15 within the medium TE group, and -1.25 within the low TE 315 group (Fig. 3D). This analysis demonstrates that ribosome profiling is at least 22% more   (Table S1). To minimize genetic variation, the three genotypes were bred into an 328 isogenic background, and three replicate experiments were performed for each genotype (see 329 materials and methods). We first determined the total number of genes expressed in Drosophila 330 muscle, as assessed through both transcription and ribosome profiling. The fly genome is 331 predicted to encode 15,583 genes (NCBI genome release 5_48). We found 6,835 genes to be 332 expressed in wild-type larval muscle through transcriptional profiling, and a similar number 14 (6,656) through ribosome profiling (Fig. 4A), with ~90% of transcripts being shared between the 334 two lists (Table S2), indicating that the vast majority of transcribed genes are also translated.

335
We found no significant differences in the size of the transcriptome and translatome between 336 wild type, GluRIIA mutants, and Tor-OE (Table S2). We then compared the muscle 337 transcriptome to a published transcriptome from the central nervous system (CNS) of third-338 instar larvae (Brown et al., 2014). This comparison revealed dramatic differences in gene 339 expression between the two tissues ( Fig. 4B). In particular, we found several genes known to be 340 enriched in muscle, including myosin heavy chain, α actinin, and zasp52, to be significantly 341 transcribed and translated in muscle, as expected. In contrast, neural-specific genes such as 342 the active zone scaffold bruchpilot, the post-mitotic neuronal transcription factor elav, and the 343 microtubule associated protein tau, were highly expressed in the CNS but not detected in 344 muscle (Table S3 and Table S3).     regions differ between these transcripts (Fig. 5C), enabling us to distinguish expression 407 between these transcripts. We first confirmed a large increase in Tor coding region expression 408 through both transcriptional profiling (68-fold) and ribosome profiling (1200 fold) (Fig. 5C, black).

409
In contrast, analysis of the 5' and 3' UTRs revealed very little difference in endogenous Tor 410 expression in UAS-Tor compared to wild type (Fig. 5C, grey), while a dramatic increase in both 17 transcription (125-fold) and translation (1200-fold) was observed (Fig. 5C, red). Indeed, the

18
We first compared transcription and translation in GluRIIA mutants and Tor-OE relative 437 to wild type by plotting the measured expression values for each condition and determining the 438 coefficient of determination, r 2 . An r 2 value equal to 1 indicates no difference between the two 439 conditions, while a value of 0 implies all genes are differentially expressed. This analysis 440 revealed a high degree of similarity between wild type and GluRIIA mutants in both transcription 441 and translation, with r 2 values above 0.98 (Fig. 6A). In contrast, a slightly larger difference exists 442 in transcription between Tor-OE and wild type, with r 2 =0.920 (Fig. 6B). Although transcription 443 should not be directly affected by Tor-OE, this implies that perhaps an adaptation in 444 transcription was induced in the muscle in response to chronically elevated translation. Finally, 445 translational differences were the largest between Tor-OE and wild type, with r 2 =0.363 (Fig. 6B).  average of 1.6 fold change in translation compared to 1.09 for wild type (Fig. 6C). This shift is 466 significant when tested by Kolmogorov-Smirnov test (p<0.001) (Fig. 6D). We then performed

471
Although no specific translational targets were identified to significantly change in

472
GluRIIA mutants compared to wild type, we did identify 46 genes that exhibited significant 473 increases in translation efficiency in Tor-OE (fold change more than 2, p-value less than 0.05) 474 (Table S5). We characterized the expression of these genes in GluRIIA vs WT to determine 475 whether a trend was observed that may differentiate the translational adaptations that drive 476 retrograde PHP signaling in Tor-OE compared to the more general overall increase in protein 477 synthesis. We therefore generated a heatmap of these 46 genes in Tor-OE vs WT and 478 compared this to the same 46 genes in GluRIIA vs WT (Fig. 6E). This analysis revealed no 479 particular trend or correlation in GluRIIA among the 46 genes with increased translation 480 efficiency in Tor-OE (Fig. 6E). Together, these results suggest that retrograde signaling in the   Tor-OE (Fig. 6B). This suggested that adaptations in transcription, and perhaps also translation,     shock protein genes with significant expression in the muscle (Table S6)   Tor-OE (Fig. 7E). Interestingly, proteasome subunits were reported to be upregulated in cells 526 with increased Tor activity (Zhang et al., 2014). We also identified the RNA polymerase subunit 527 rpl1 and transcription factor myc to be upregulated following Tor-OE (Fig. 7F,G). These genes

577
This optimized ribosome profiling approach has illuminated genome-wide translational 578 dynamics in Drosophila muscle tissue and demonstrated two opposing protein production 579 strategies utilized in these cells: elevated transcriptional expression coupled with low translation 580 efficiency, which was apparent for genes encoding ribosomal subunits (Fig. 4F), and low 581 transcriptional expression coupled with high translation efficiency, which was observed for 582 genes encoding proteins belonging to diverse functional classes (Fig. 4G). These

729
The lysate-beads mixture was incubated at 4°C with rotation for 2 hours, then washed in buffer

813
All data are presented as mean +/-SEM. Student's t test was used to compare two groups. A 814 one-way ANOVA followed by a post-hoc Bonferroni's test was used to compare three or more 815 groups. All data was analyzed using Graphpad Prism or Microsoft Excel software, with varying 816 levels of significance assessed as p<0.05 (*), p<0.01 (**), p<0.001 (***), N.S.=not significant.

817
Statistical analysis on next generation sequencing data was described in the High-throughput 818 sequencing and data analysis section.