Smg5 is required for multiple nonsense-mediated mRNA decay pathways in Drosophila

The nonsense-mediated mRNA decay (NMD) pathway is a cellular quality control and post-transcriptional gene regulatory mechanism and is essential for viability in most multicellular organisms. A complex of proteins has been identified to be required for NMD function to occur, however the individual contribution of each of these factors to the NMD process is not well understood. Central to the NMD process are two proteins Upf1 (SMG-2) and Upf2 (SMG-3), which are found in all eukaryotes and are absolutely required for NMD in all organisms in which it has been examined. The other known NMD factors, Smg1, Smg5, Smg6, and Smg7 are more variable in their presence in different orders of organisms, and are thought to have a more regulatory role. Here we present the first genetic analysis of the NMD factor Smg5 in Drosophila. Surprisingly, we find that unlike the other analyzed Smg genes in this organism, Smg5 is essential for NMD activity. We found this is due at least in part to a role for Smg5 in the activity of two separable NMD-target decay mechanisms: endonucleolytic cleavage and 5′-to-3′ exonucleolytic decay. Redundancy between these degradation pathways explains why some Drosophila NMD genes are not required for all NMD-pathway activity. We also found that while the NMD component Smg1 has only a minimal role in Drosophila NMD during normal conditions, it becomes essential when NMD activity is compromised by partial loss of Smg5 function. Our findings suggest that not all NMD complex components are required for NMD function at all times, but instead are utilized in a context dependent manner in vivo.


INTRODUCTION 47
Eukaryotic cells utilize a number of pathways to maintain error-free translation so 48 as to preserve the fidelity of protein function (Adjibade and Mazroui 2014). Nonsense-49 mediated mRNA decay (NMD) is one such pathway, which prevents the translation of 50 potentially harmful truncated proteins by recognizing and destroying mRNAs that contain 51 erroneous premature-termination codons (PTCs) (Celik et al. 2015). In addition to this 52 cellular quality control function, NMD degrades many endogenous wild-type mRNAs as 53 a mechanism of post-transcriptional gene regulation (Peccarelli and Kebaara 2014). 54 While the phenomenon of NMD has been well characterized for several decades, 55 the mechanisms initiating target recognition and degradation are still not well understood 56 and it remains unclear if all the factors required for NMD activity have even been 57 identified. Genes required for NMD were first found by genetic screens in yeast and C. 58 elegans, which led to the identification of seven proteins required for NMD (Hodgkin et 59 al. 1989; Leeds et al. 1991;1992;Cali et al. 1999). Three of these genes, Upf1, Upf2, 60 and Upf3, are present in every eukaryote examined, while the other four, Smg1, Smg5, 61 Smg6, and Smg7, have variable presence across species (Siwaszek et al. 2014). In the 62 absence of any one of these factors, PTC-containing mRNAs and endogenous targets are 63 not efficiently degraded and instead accumulate in the cell (Gatfield et al. 2003; 64 Rehwinkel et al. 2005). The molecular identities and biochemical characterization of the 65 individual NMD genes have revealed clues about their roles in the NMD pathway. Upf1 66 is an ATP-dependent RNA helicase, and this activity is required for NMD (Czaplinski et 67 al. 1995;Weng et al. 1996a;. Upf3 binds mRNAs both directly and through an 68 chromosome) strain (Venken et al. 2006). Expression levels of the modified SV40 162 3'UTR constructs were measured by mating transgenic males to y w FRT 19A  with Df(2L)BSC345 (Cook et al. 2012), which deletes the Smg5 locus, and lines that 178 failed to complement this deficiency for lethality or fluorescence enhancement were 179 balanced over CyO, P{Dfd:eYFP w + } (Le et al. 2006). 180 The screen for embryos with enhanced reporter expression is described in Förster 181 et al. (2010). Briefly, flies carrying NMD-sensitive UAS-GFP and UAS-Verm-mRFP 182 reporters expressed in the tracheal system were mutagenized with EMS and F2 lines were 183 established. F3s embryos were examined for reporter expression in tracheal cells. 184 185 RNA isolation and quantification: For qRT-PCR analyses, we collected five to ten 186 larvae from 0-4 h after the L2-L3 molt and froze them in liquid nitrogen. We isolated 187 total RNA using TRIzol reagent (Invitrogen) and Phase-Lock tubes (5-Prime), and the 188

RESULTS 242
Isolation of Smg5 mutant alleles: To identify Drosophila Smg5 mutant alleles, we used 243 two different genetic screens. First, we performed an EMS-mutagenesis screen in mosaic 244 animals expressing an NMD-sensitive GFP reporter, a method similar to one we 245 previously used to recover Smg6 alleles (Frizzell et al. 2012). This reporter expresses 246 GFP from a pUAST construct (Brand and Perrimon 1993) bearing a UAS promoter and 247 an NMD-sensitive SV40 3'UTR (Metzstein and Krasnow 2006). We generated mosaics 248 using the da-GAL4 driver to ubiquitously express FLP-recombinase ( Figure 1A). 249 Individual homozygous mutant cells with defective NMD activity show increased 250 reporter expression and GFP fluorescence ( Figure 1A). The mosaic enhanced 251 fluorescence phenotype was easy to distinguish in late L3 larvae ( Figure 1B), and mosaic 252 animals remain viable and fertile, so even lethal alleles can be recovered from individual 253 mutants. An added benefit of this approach is that by mutagenizing animals that have an 254 FRT site located near the centromere on the left arm of the second chromosome (FRT 40A ), 255 we could specifically isolate mutations only on this chromosome arm. Since Smg5 is 256 located on the left arm of the second chromosome, mutations identified from the screen 257 would likely include Smg5 alleles. Using this approach, we screened 12,554 larvae and 258 identified three mutants with mosaic enhancement of GFP fluorescence ( Figure 1C). We 259 found each of these three mutants were homozygous lethal. We crossed each allele to a 260 deficiency that deletes Smg5 and found that all three failed to complement for lethality, 261 suggesting that they had mutations in Smg5. 262 Our second screen was of animals expressing a GFP::SV40 3'UTR reporter in the 263 embryonic tracheal system (Förster et al. 2010). This screen identified four mutants that 264 showed increased fluorescence ( Figure 1D, Supplemental Figure 1). All four of these 265 alleles failed to complement a Smg5 deficiency using increased fluorescence signal as an 266 assay (data not shown), indicating they contained mutations in Smg5. Gadd45 mRNA levels in early third instar larvae and found that Smg5 C391/G115 mutants 299 had a large increase in Gadd45 mRNA expression. In contrast, viable Smg5 Q454/G115 300 mutants showed a much smaller increase in Gadd45 levels ( Figure 2B). Increased 301 Gadd45 expression is a major factor contributing to the death of Upf1 and Upf2 mutants, 302 and loss of Gadd45 can suppress Upf1 and Upf2 mutant lethality (Nelson et al. 2016). 303 We found that loss of Gadd45 also suppresses the lethality of Smg5 C391/G115 mutants 304 (Figure 2A), indicating that these animals are dying due to a similar loss of NMD 305 function as Upf1 or Upf2 mutants. These results strongly suggest that Smg5 mutant 306 lethality is specifically due to a loss of NMD activity, and not due to loss of any NMD-307 independent Smg5 function. 308

Smg5 null mutants lack most, if not all, detectable NMD activity: To directly test if 310
Drosophila Smg5 mutants have any residual NMD activity, we measured the relative 311 stability of PTC-containing mRNAs in Smg5 mutants. We found that Smg5 C391/G115 312 mutants fully stabilized the expression of the PTC-containing dHR78 3 mRNA relative to 313 the expression of wild-type dHR78 mRNA ( Figure 2C), indicating NMD-mediated 314 degradation of PTC-containing mRNA is absent. Since Smg6-mediated cleavage is a 315 known mechanism for degradation of NMD targets in Drosophila, we tested if Smg5 316 mutants still retain this endonuclease activity. NMD-target cleavage can be observed 317 through measuring the relative abundance of NMD-target mRNA fragments 5' to the stop 318 codon in relation to fragments 3' to the stop codon ( Figure 2D) in animals lacking the 319 only cytoplasmic 5'-to-3' exonuclease Xrn1, which is encoded by the gene pacman (pcm) 320 (Till et al. 1998). Null pcm mutants have no 5'-to-3' exonuclease activity (Waldron et al. 321 2015), and thus mRNAs cleaved by Smg6 near the stop codon show increased abundance 322 of the 3' cleavage fragment compared to the 5' fragment (Nelson et al. 2016). We found 323 this bias to be lost in the absence of Smg6 ( Figure 2E), confirming that it is caused by 324 Smg6 endonuclease activity. Interestingly, we found that the preferential stabilization of 325 the 3' fragment is also lost in double mutants of the null alleles of pcm and Smg5 ( Figure  326 2E), revealing that Smg5 is required for Smg6 endonuclease activity. These combined 327 results indicate that Drosophila Smg5 mutants lack any NMD activity. 328 As an additional gauge of NMD activity in Smg5 mutants, we directly measured 329 fluorescence levels of NMD-sensitive reporters in homozygous mutant embryos ( Figure  330 3). We found that homozygous Smg5 C391 embryos exhibited ~5-fold increase in 331 fluorescent signal compared to Smg5 + embryos, comparable to the increase in GFP 332 mRNA levels observed in the strongest previously measured NMD mutant, Upf2 25G 333 (Metzstein and Krasnow 2006  has been proposed to be a critical step in the NMD process, at least in part by recruiting 353 Smg6 to the NMD complex to initiate target degradation (Hug et al. 2016). 354 Dephosphorylation of Upf1 is thought to be mediated by Smg5, which interacts with the 355 PP2A phosphatase; this activity may be required for complex disassembly after target 356 degradation has been initiated (Ohnishi et al. 2003 suppress Smg5 mutant lethality. In contrast to this prediction, we found that Smg1; Smg5 361 double mutants were in fact no more viable than Smg5 mutants ( Figure 4A), and Smg1 362 mutants had no effect on the developmental delay or lethal stage of Smg5 mutants 363 (Supplemental Figure 2A, B). These findings suggest that a failure to dephosphorylate 364 Upf1 is not responsible for Smg5 mutant lethality; however, the lethality of Smg1; Smg5 365 double mutants may also be explained by unknown factors that phosphorylate Upf1 in the 366 absence of Smg1. 367 If failure to dephosphorylate Upf1 causes lethality in both Smg5 mutants and Smg1; 368 Smg5 double mutants, we would expect loss of Smg1 to have no effect on the viability of 369 hypomorphic Smg5 mutants, since these alleles are viable (Figure 2A), and so should 370 have sufficient Upf1-dephosphorylation. Surprisingly, we found that double mutants for a 371 Smg1 null allele and a hypomorphic Smg5 allele show significant lethality, even though 372 each mutation on its own is viable ( Figure 4A). This result reveals that Smg5 mutant 373 lethality is not due to failure to dephosphorylate Upf1, but instead is consistent with an 374 alternative proposed model that Upf1 phosphorylation by Smg1 is not required for NMD 375 under normal conditions, but serves to enhance Smg6 and Smg5 efficiency upon stress 376 conditions to reinforce NMD activity (Durand et al. 2016). In agreement with this model, 377 we found that the relative increase in Gadd45 expression upon loss of Smg1 is greater in 378 this Smg5 hypomorphic background than in animals with functioning Smg5 (Figure 4B), 379 indicating that Smg1 has a greater contribution to NMD activity when Smg5 functions 380 inefficiently. Importantly, loss of Smg1 has no greater impact on Gadd45 expression in 381 Smg5 null mutants than in animals with functional Smg5 (Figure 4B), indicating that the 382 compensatory Smg1 activity in Smg5 hypomorphs is the enhancement of Smg5 function, 383 rather than an increase in Smg5-independent decay activity. Together, these findings 384 indicate that the requirement of Smg5 for NMD activity is independent of Smg1, but that 385 Smg1 can enhance Smg5 activity when NMD function is compromised. Xrn1-mediated degradation of Gadd45 mRNA requires Smg5 activity. Importantly, 418 Gadd45 expression was measured using the qRT-PCR primer pair 5' to the Gadd45 stop 419 codon ( Figure 2D), implying that Xrn1 can only degrade these mRNAs after decapping. 420 These findings suggest that Smg5 may be involved in promoting the decapping of NMD 421 targets. 422

DISCUSSION 424
The degradation of both specific normal and many kinds of erroneous mRNAs by the 425 NMD pathway is a crucial gene regulatory mechanism and arose in the ancestors to all 426 eukaryotes. While many factors required for NMD have been biochemically 427 characterized, the individual contribution of each factor to the recognition and 428 degradation of NMD targets is not fully understood, especially in vivo in complex 429 organisms. Through our genetic analysis of Smg5 in Drosophila and by examination of 430 NMD-gene double mutants, we have found that NMD utilizes multiple mechanisms to 431 promote target degradation in vivo. One of our main findings is that Smg5 null mutants 432 have as severe defects as either Upf1 and Upf2 null mutants, indicating that Smg5 is a 433 critical factor for promoting NMD-target recognition and/or decay. In support of this 434 interpretation, we found that Smg5 is required for both Smg6-mediated endonucleolytic 435 cleavage of NMD targets and a separate, Smg6-independent, decay process that at least 436 partially requires Xrn1 5'-to-3' exonuclease activity. Our findings are surprising, given 437 that Smg5 has primarily been thought to promote NMD complex recycling, but with only 438 a secondary requirement to stimulate decay activity (Ohnishi et al. 2003). Instead, we 439 propose that Smg5 is a critical NMD factor necessary for at least two, independent NMD 440 degradation mechanisms. 441 In contrast to Smg5 having a critical role in target degradation, our data is less 442 supportive for a Smg5 function in NMD complex recycling. The phosphorylation of Upf1 mutants suggests that Smg6-indepdendent decay is sufficient to maintain most NMD 494 activity. It is also possible that the preference for which decay mechanism degrades NMD 495 targets may be different between individual NMD targets. Furthermore, the choice 496 between decay mechanisms may differ in tissue-specific or developmental contexts. It 497 will be important to parse the relative contribution of each decay pathway to the 498 degradation of NMD targets to understand the mechanism underlying the bias in decay. 499 Here we performed the first double mutant analysis of multiple NMD factors, 500 providing genetic evidence of the relative contribution of individual NMD genes. We 501 also characterized the first Drosophila Smg5 mutants, identifying that Smg5 is critical for 502 NMD function and viability, similar to Upf1 and Upf2, and providing the first genetic 503 evidence for an essential role of Smg5 function in a model system. Our findings suggest 504 that NMD utilizes multiple branched decay mechanisms to destroy its targets. All of these 505 pathways depend on Smg5, indicating that Smg5 plays more fundamental roles in NMD 506 than has previously been appreciated. More closely characterizing the molecular 507 mechanisms of Smg5 function in NMD may reveal novel key features of NMD activity 508 that have thus far escaped detection. 509 510 510 Acknowledgements 511

512
We thank x for critical comments on the manuscript. We are very grateful to 513 Sarah Newbury for providing fly stocks prior to publication.     Fluorescence intensity (a.u.)