tRNA 2’-O-methylation modulates small RNA silencing and life span in Drosophila

2’-O-methylation (Nm) represents one of the most common RNA modifications. Nm affects RNA structure and function with crucial roles in various RNA-mediated processes ranging from RNA silencing, translation, self versus non-self recognition to viral defense mechanisms. Here, we identify two novel Nm methyltransferases (Nm-MTases) in Drosophila melanogaster (CG7009 and CG5220) as functional orthologs of yeast TRM7 and human FTSJ1, respectively. Genetic knockout studies together with MALDI-TOF mass spectrometry and RiboMethSeq mapping revealed that CG7009 is responsible for methylating the wobble position in tRNAPhe, tRNATrp and tRNALeu, while subsequently, CG5220 methylates position C32 in the same tRNAs and targets also additional tRNAs. CG7009 or CG5220 mutant animals were viable and fertile but exhibited various phenotypes such as life span reduction, small RNA pathways dysfunction and increased sensitivity to RNA virus infections. Our results provide the first detailed characterization of two TRM7 family members in Drosophila and uncover a molecular link between enzymes catalysing Nm at specific tRNAs and small RNA-induced gene silencing pathways.


Introduction
The existence of RNA modifications has been known for over 50 years and many of the pioneering studies addressed the function of RNA modifications in abundantly expressed RNAs such as transfer RNAs (tRNAs) and ribosomal RNAs (rRNAs). tRNAs are the most heavily modified RNAs (1) . tRNAs are modified post-transcriptionally and the biosynthesis of modified nucleosides often requires different modification enzymes acting sequentially at distinct steps of tRNA maturation (2,3) . The complex mechanisms underlying the stepwise modification of tRNAs were largely deciphered in the yeast S. cerevisiae , as well as in studies conducted in prokaryotes and Archaea . More recently, some of these seminal findings in single-cell organisms were revisited using multi-cellular models including studies that aim at understanding how mutations in tRNA modification enzymes affect organismal development and disease etiology.
2'-O-methylation (Nm) is a common RNA modification. The addition of a methyl group to the 2' hydroxyl of the ribose moiety of a nucleoside creates Nm (reviewed in (4,5) ). Nm can occur at any nucleotide explaining the abundant nature of this modification. Nm residues are found at multiple and often highly conserved positions in tRNAs, rRNAs, and small nuclear RNAs (snRNAs) (6)(7)(8) . In eukaryotes, RNA modification reactions resulting in Nm on rRNAs and snRNAs are frequently catalyzed by evolutionarily conserved C/D-box small RNAs (SNORDs) involving guide ribonucleoprotein particles (RNPs) which contain the Nm-methylase fibrillarin.
Small nucleolar RNPs (snoRNPs) mediate the deposition of Nm at rRNAs while small Cajal bodies RNPs (scaRNPs) direct Nm-modification to snRNAs. In contrast, most of the Nm deposition occurring in eukaryotic tRNAs are mediated by stand-alone proteins without the need for guidance by small RNAs. However, recently it was reported that one snoRNA and one scaRNA can guide Nm deposition to tRNA Met in mammalian cells (9) . Importantly, Nm deposition was also reported at 3'-terminal nucleotides in small non-coding RNAs (sncRNAs) such as microRNAs (miRNAs) and small-interfering RNAs (siRNAs) in plants (10)(11)(12) , in Argonaute-2 (Ago2) loaded siRNAs and miRNAs in flies and in PIWI-interacting RNAs (piRNAs) in animals (13)(14)(15) . More recently Nm was also reported to be present in messenger RNA (mRNA) (16,17) .
The methyltransferase complex catalysing Nm formation in the ACL of mammalian and yeast tRNAs comprises the Nm-methyltransferases (Nm-MTases) FTSJ1 or TRM7, respectively.
Several studies have uncovered crucial roles for FTSJ1/TRM7 in normal and pathological conditions (reviewed in (4,5) ). While in S. cerevisiae , lack of TRM7 affected growth, FTSJ1 mutant mice showed impairment in their learning capacity as well as significantly reduced pain sensing (hypoalgesia) and altered gene expression profiles (40) . Similarly, in humans several mutations in FTSJ1 were shown to be causative of a neurodevelopmental disorder known as Non-Syndromic X-linked Intellectual Disability (NSXLID) (39,41,42) .
Importantly, expression of human FTSJ1 in yeast suppressed the severe growth defects observed in trm7Δ mutants, demonstrating that the TRM7 enzyme family and their RNA targets are highly conserved (35) .
While the molecular function of yeast and human Nm-MTases acting on specific tRNAs has been established, the molecular mechanisms causing the complexity of observed mutant phenotypes have not fully been elucidated. Importantly, a tractable multicellular model system that would allow studying Nm-MTase function systematically and thereby bridge the growth phenotypes observed in trm7 deficient yeast with the complex phenotypes observed in FTSJ1-mutant man has been lacking.
In this report, we show that, in contrast to yeast and humans, Drosophila melanogaster has evolved two Nm-MTase genes, CG5220 and CG7009 , whose genes products specialized their activity to respectively methylate positions 32 and 34 in the ACL of specific tRNAs. We demonstrate that the catalytic specificity of these Nm-MTases is dependent on the position 3 rather than the identity of the ACL nucleotides. Importantly, lack of these proteins reduced Drosophila life span and impaired various cellular pathways, which employ small RNAs to achieve post-transcriptional silencing. Hence, CG5220 and CG7009 mutant animals were more sensitive to RNA virus infections and showed dysfunctional control of transposable elements, suggesting a molecular link between Nm RNA modifications and small RNA gene silencing pathways in Drosophila .

Results
A genome-wide RNAi screen identifies CG7009 as regulator of siRNA-mediated silencing pathways.
We previously developed and characterized a self-silencing genetic sensor ( automiG ) that combines the expression of GFP with miRNAs, miG-1 and miG-2, targeting GFP mRNA ( Figure 1A and (43) ). AutomiG self-silencing reports on the activity of canonical miRNA biogenesis factors such as Drosha and Dicer1 (Dcr1) and the function of RNA-induced silencing complex (RISC) factors such as Argonaute2 (Ago2) and Dicer2 ( Dcr2 ). Impairing the function of miRNA biogenesis or Ago2 silencing activity thus causes the de-repression of automiG self-silencing resulting in the expression of GFP (43) . To identify additional regulators of these two RNA silencing pathways, a genome-wide RNA interference screen was performed in in S2 cells expressing automiG resulted in increased GFP expression when compared to control constructs ( Figure 1B). In addition, a dual luciferase assay reporting on siRNA pathway activity in S2 cells (44,45) confirmed that Dcr2/Ago2-dependent silencing was affected in cells with down-regulated CG7009 expression (Supplementary Figure S1B).
In order to obtain insights into the impact of CG7009 loss on gene expression control  Figure S1C) suggesting that CG7009 may act upstream in the siRNA pathway by regulating Ago2 mRNA levels.

CG7009 encodes for a predicted Nm-MTase.
Amino acid (aa) sequence analysis suggested that the protein encoded by CG7009 in D.
melanogaster harbours a methyltransferase domain belonging to the conserved RlmE family and TRM7 subfamily of the class I-like SAM-binding methyltransferase superfamily (38) .
Sequence alignment of the putative CG7009 protein with the yeast Nm-MTase TRM7 showed 52 % aa sequence identity, including the conserved KDKE motif in the active site, and 66 % of aa coverage ( Figure 1C). FTSJ1 is the human ortholog of TRM7 (39) . CG7009 shares 51 % aa identity and 86 % aa coverage with FTSJ1 ( Figure 1C). Surprisingly, further sequence alignment of CG7009 protein with proteomes of different Drosophila species uncovered an additional gene, CG5220, whose annotated protein in Drosophila melanogaster displays 63 % aa sequence identity with CG7009 (Figure 1 C, D). Like CG7009, CG5220 was an uncharacterized gene with an amino acid composition that clearly showed an Nm-MTase signature ( Figure 1C).
Importantly, it was previously reported that overexpression of CG5220 was able to rescue the growth phenotype observed in trm7Δ mutant yeast (35) . Similarly to CG7009, CG5220 shows high similarity to TRM7 (48 % identity and 83 % coverage) as well as to FTSJ1 (58 % of identity and 82 % coverage, Figure 1C). These findings pointed to CG7009 and CG5220 as potential paralogs and conserved members of the TRM7 Nm-MTases family in Drosophila .

Mutations in CG7009 or CG5220 are viable and fertile.
To investigate the function of CG7009 and CG5220 during Drosophila development and to characterize the potential enzymatic activity of their gene products, we characterized existing mutations in CG7009 gene and generated CG5220 mutant flies. For CG7009 , one transposon insertion line ( CG7009 e02001 ) and two genomic deletion lines ( Def3340 and Def9487 ) were obtained and confirmed at the molecular level (Supplementary Figure S2A- Figure S2D, E). To address the function of CG5220, CRISPR/Cas9-mediated genome editing was used to create a CG5220 mutant allele ( CG5220 K>A ), which substituted a conserved lysine at position 28 in the predicted catalytic domain with alanine ( Figure 1C and Supplementary Figure S2F) Figure 2A ).

CG7009 and CG5220 contribute to efficient miRNA Ago2-mediated RNA silencing in vivo.
To address whether CG7009 affected small RNA silencing pathways in vivo , we expressed the automiW sensor, which is based on the knockdown of the white gene by means of white -targeting miRNAs in the developing eye (47) . Combining this sensor construct with RNAi-mediated knockdown of CG7009 or CG5220, we observed increased eye coloration when compared to controls ( Figure 2B). This observation indicated that Ago2-dependent silencing or miRNAs biogenesis of this reporter was affected non-redundantly when CG7009 or CG5220 were post-transcriptionally down-regulated. These results implicated both genes in proper functioning of either miRNA biogenesis and/or Ago2-dependent and small interfering RNA-mediated silencing in vivo .
siRNA-mediated RNA silencing is impaired in CG7009 and CG5220 mutant flies.
As small interfering RNA-mediated silencing is required for viral defence in Drosophila (48) , we tested whether viral defence was impaired in CG7009 or CG5220 K>A mutant flies. To this end, purified Drosophila C virus (DCV) was injected into the thorax of adult flies and the viral load was monitored by qRT-PCR 4 days after infection. The results of these experiments showed that flies lacking CG7009 or CG5220 function were significantly more sensitive to DCV infection when compared to control flies ( Figure 2C). The results indicated that CG7009 and 6 CG5220 K>A mutants failed to initiate or maintain a proper response to viral infection, which, together with the results of our genetic screen using automiG for miRNA biogenesis and Ago2-dependent silencing of gene expression ( Figure 1B and 2B) and the siRNA activity reporter assays (Supplementary Figure S1B), supported the notion that both gene products were required for efficient Ago2-dependent and small interfering RNA-mediated silencing activities in Drosophila .
piRNA-mediated RNA silencing is affected in CG7009 and CG5220 mutant flies .
During the characterization of the CG7009 mutants, we noticed that females ovaries were reduced in size when compared to BAC-rescued control flies ( Figure 2D). This ovarian phenotype was mild but similar phenotypes were previously described in several mutants for genes required in the piRNA pathway (49) . Although the genetic screen was specifically designed to identify genes involved in miRNA biogenesis or Ago2-mediated silencing pathways, we tested whether CG7009 and CG5220 function also affected transposable element (TE) silencing, especially through the somatic piRNA pathway. To this end, the activity of a somatic piRNA-mediated silencing reporter (50) was monitored in adult ovaries. This reporter faithfully recapitulates the expression of the retro-transposon gypsy in ovarian follicle cells, in which abundant somatic piRNAs are produced in defence against mobile elements (50) . Remarkably, piRNA-mediated silencing of this reporter was de-repressed upon somatic follicle cell-specific knockdown of CG5220 and CG7009 expression ( Figure 2 E-F). Furthermore, expression of endogenous gypsy mRNA was elevated in CG7009 mutants ( Figure 2F). Taken together, these results suggested that both genes are involved in TE silencing mediated by the somatic piRNA pathway in Drosophila .

Mutation in CG7009 and CG5220 affects life span.
Although a size reduction of CG7009 , CG5220 double mutant adult flies could be observed (Figure 2A), no other obvious mutant phenotypes affecting flies morphology could be observed. Importantly, CG7009 e02001 , CG5220 K>A double mutant flies displayed reduced lifespan when compared to controls ( Figure 2G). Mutant flies lived, on average, about 25 days shorter than controls. Furthermore, homozygous CG7009 e02001 mutant flies as well as CG7009 e02001 , CG5220 K>A double mutants appeared sluggish and less active displaying general locomotion defects (not shown).

CG7009 and CG5220 are Nm-MTases acting on tRNAs.
To test whether CG7009 is an Nm-MTase, recombinant proteins were expressed and purified from E. coli . In vitro methylation assays using in vitro -synthesized Drosophila tRNA Phe did not reveal activity of recombinant CG7009 protein. In order to ascertain the predicted catalytic activities of CG7009 and CG5220, we analyzed the Nm methylation status of Drosophila tRNA Phe , which is a substrate of TRM7 in yeast, using control, CG7009 e02001 and CG5220 K>A mutants flies . We performed sequence-specific purification of tRNA Phe using biotinylated DNA oligonucleotides coupled to streptavidin matrices followed by RNase digestion and MALDI-TOF mass spectrometry. RNase A has a preference for hydrolysis at pyrimidine residues, while RNase T1 is strictly guanosine-specific. Because Nm at a given nucleotide position ( n ) protects the adjacent 3'-phosphodiester bond to the neighbouring nucleotide (position n+1 ) against nuclease attacks, various specific digestion products of Drosophila tRNA Phe can be expected as a result of RNase A or RNase T1 activities. In addition, according to the reported modification profile of Drosophila tRNA Phe (51,52) , which includes Nm at C 32 and G 34 , very specific RNA fragments were thus predicted ( Figure 3A and Supplementary Figure   S3).
First, we determined fragments that were obtained after RNase A hydrolysis of tRNA Phe , which should provide information on the Nm-modification status at C 32 . MALDI-TOF analysis revealed almost no fragment of 1327.2 Da (AGAC 32 p fragment) in control flies indicating that C 32 was modified with Nm thereby blocking RNase A activity at this position in tRNA Phe from wild type and rescue flies ( Figure 3A and Supplementary Figure S4A, B). This fragment increased significantly in CG5220 K>A mutants suggesting loss of protection from RNase A activity in animals without CG5220. Interestingly, the increase in RNase A-mediated tRNA Phe fragmentation observed in CG5220 K>A mutants could only be moderately observed when using tRNA Phe from CG7009 e02001 ,CG5220 K>A double mutant flies ( Figure 3A and Supplementary Figure   S4A) indicating that C 32 protection from RNase A is largely independent of CG7009. In support of this notion, the CG7009 e02001 mutation alone did not affect the RNAse A digestion profiles when compared to control ( Figure 3A and Supplementary Figure S4A) or BAC rescue CG7009 e02001 flies (Supplementary Figure S4B). These results indicated that CG5220, but not CG7009, harbors an activity that protects tRNA Phe at C 32 against RNase A digest, therefore making CG5220 the main candidate for an Nm-MTase at this position in Drosophila .
Next, we obtained RNase T1 digestion profiles to deduce the G 34 modification status of tRNA Phe in both control and mutant flies. MALDI-TOF analysis showed a ACm 32 UG 34 p fragment 8 (1318,1 Da) that could not be detected in wild type flies indicating that G 34 was modified with Nm thereby blocking RNase T1 activity at this position in wild type tRNA Phe ( Figure 3A, B). This fragment increased significantly in CG7009 e02001 mutants suggesting loss of protection from RNase T1 activity in animals without CG7009. The RNase T1 digestion profiles from controls and CG5220 K>A mutant flies were comparable ( Figure 3A, B and Supplementary Figure S4B) indicating that CG7009 but not CG5220 is implicated in protecting G 34  suggesting that CG5220 and CG7009 are the responsible Nm-MTase activities that modify C 32 and G 34 in tRNA Phe , respectively ( Figure 3A, B). Notably, Nm deposition at G 34 was not affected in CG5220 K>A mutants ( Figure 3A, B) indicating that CG5220-mediated modification at C 32 does not influence the deposition at neighboring G 34 . In contrast, loss of Nm at G 34 in CG7009 e02001 mutants moderately affected Nm C 32 deposition ( Figure 3A, B, 1304 Da). These results indicated that while Nm at C 32 was not a prerequisite for Nm at G 34 , Nm at G 34 was probably required for the correct formation of methylation at C 32 or for the stability of the methylated tRNA.
Collectively, these data demonstrated that genetic mutation of two candidate Nm-MTases in Drosophila resulted in the reciprocal loss of two conserved ACL modifications in tRNA Phe strongly suggesting that CG5220 and CG7009 are indeed functional methyltransferases responsible for the deposition of Nm at C 32 and G 34 of tRNA Phe , respectively. Interestingly, our results also suggest that Drosophila melanogaster , and most likely all other Drosophila species, evolved two distinct TRM7 family members to methylate the ACL on substrate tRNAs ( Figure   1D).

Methylation specificity of both MTases depends on nucleotide position.
To obtain a comprehensive picture of the Nm specificity for CG7009 and CG5220 in vivo , we performed RiboMethSeq analysis on Drosophila tRNAs. RiboMethSeq allows RNA-wide Nm detection based on random RNA fragmentation by alkaline hydrolysis followed by library preparation and sequencing (53,54) . The presence or absence of Nm can be appreciated from characteristic coverage profiles of the 5′-/ 3′-ends of cDNAs. Since Nm residues protect the  Figure 3C). This demonstrated that, Nm was present at G 34 in Drosophila tRNA Phe as previously reported (51) and as shown by MALDI-TOF MS analysis ( Figure 3A).
Similarly, RiboMethSeq profile analysis of CG5220 K>A mutants indicated G 34 to be methylated (Supplementary Figure S5A). The presence of Nm at G 34 in CG5220 K>A mutant confirmed that Furthermore, RiboMethSeq analysis also identified other tRNAs potentially methylated by CG7009 and CG5220, some of which were already known as substrates of TRM7 orthologs in other species. For instance, alike tRNA Phe , we found CG7009-dependent methylation in tRNA Trp at position C 34 as well as CG5520-dependent Nm at position C 32 ( Figure 3D). Strikingly, the methylated nucleotide at position 34 in tRNA Trp of Drosophila is a cytosine, like in humans and in yeast (1,35,38,39) . Importantly the RiboMethSeq profiles clearly showed that CG7009 (and not CG5220) methylated this position ( Figure 3D) indicating that CG7009 can deposit Nm on G and C nucleotides. The same observation was made for CG7009-mediated methylation of C 34 in tRNA Leu(CAA) , in agreement with previous data showing that FTSJ1 was responsible for depositing Nm at f5C 34 /hm5C 34 in human tRNA Leu(CAA) (34) and Supplementary Figure S5B, C).
In addition, we identified previously unknown Nm-MTase substrate tRNAs. We found CG5220-dependent methylation of tRNA Gln and tRNA Glu at position C 32 ( Figure 3D and Supplementary Figure S5C). 2'-O-methylated C 32 in tRNA Glu(UUC) had been previously reported in Drosophila (1,55) . Interestingly, cytosine 32 was also reported to be 2'-O-methylated in human tRNA Gln by a yet unidentified enzyme (1) . Our data thus suggest that the human ortholog of CG5220, FTSJ1, may be the Nm-MTase responsible for the modification at this position.
Altogether, detailed RiboMethSeq analysis confirmed the MALDI-TOF MS results ( Figure   3A, B) demonstrating that CG5220 is specialized for depositing Nm at C 32 nucleotides while CG7009 is responsible for modifying the wobble position. Furthermore, the discovery of additional tRNA substrates (Supplementary Figure S5C) for both Nm-MTases suggested that their respective specificity is dependent on the position rather than on the nature of nucleotide (C, U or G).

CG33172 is part of the Nm-MTase complex.
The yeast TRM7 associates with two distinct proteins that are required for its catalytic activity (35,36) . Deposition of Nm at C 32 by TRM7 is supported by binding to TRM732 while the interaction with TRM734 is necessary for addition of Nm at position 34. THADA and WDR6 are the orthologs of TRM732 and TRM734 in humans, respectively, and the interactions with FTSJ1 are conserved (39) . In Drosophila , CG15618, also known as DmTHADA (56) , is the potential ortholog of TRM732 and THADA, while CG33172 is the putative orthologue of TRM734 and WDR6 (Supplementary Figure S6A). Importantly, CG33172, TRM734 and WDR6 are members of the WD40-repeat-containing domain superfamily that contains also the human protein WDR4, another tRNA-MTase cofactor involved, as FTSJ1, in neurodevelopmental disorders (57,58) .
The use of the automiW sensor combined with dsRNA-mediated knockdown of CG15618 and CG33172 in the Drosophila eye recapitulated the Ago2-mediated small RNA silencing failure observed in CG7009 and CG5220 mutants ( Figure 4A). Interestingly, dsRNA-mediated knockdown of CG33172 using the gypsy-lacZ sensor also recapitulated the somatic piRNA silencing failure observed in both CG7009 and CG5220 mutants ( Figure 2E), indicating genetic interactions between CG7009/ CG5220-mediated functions and these gene products.
In order to test for physical interactions between CG7009, CG15618 and CG33172, we cloned FLAG-tagged CG15618 and CG33172 with the aim of co-expressing these proteins along with GST::CG7009 in bacteria. While co-expression of FLAG::CG15618 was technically challenging due to the size of this protein (197 kDa), FLAG::CG33172 could be expressed and immunoprecipitated using anti-FLAG antibodies. The precipitate was tested for the presence of GST::CG7009 by using western blotting and anti-GST antibodies. The results showed that FLAG::CG33172 co-precipitated with GST-CG7009 but not GST alone suggesting a direct interaction between these two proteins ( Figure 4B). Collectively, these observations suggested the existence of an Nm-MTase complex containing CG7009 and at least one accessory protein, CG33172, which might be required for depositing Nm at position 34 on selected tRNAs.

Nm limits endonucleolytic cleavage of tRNA Phe .
We next addressed the mechanisms underlying the defects in the Ago2-mediated RNA silencing activity and life span observed in CG7009 mutant flies. It has been reported that loss of m 5 C from specific tRNAs resulted in increased tRNA fragmentation in Drosophila .
Interestingly, it was proposed that tRNA fragments could affect small RNA silencing pathways through binding to Dicer and Argonaute proteins thereby reducing their activity (59) . In addition, during the preparation of this manuscript, a study showed that Nm 34 protected tRNA Met(CAT) from endonucleolytic cleavage by stress-induced angiogenin in human cell (9) .
We  Figure 5A, B). Furthermore, we did not observe tRNA Phe fragmentation changes in CG5220 K>A mutants ( Figure 5B). Thus, these results indicated that Nm at position G 34 limits fragmentation of tRNA Phe suggesting that 3' terminal Cm 32 stabilizes the observed tRNA fragments in CG7009 mutants.

Discussion
Surprisingly, while performing an RNAi genome-wide screen for modulators of the animal did not reveal other obvious mutant phenotypes with the notable exception of a reduction in size and weight, highlighting a potential role of these Nm-MTases in specific, but not general, translational control as previously reported for trm7 mutant yeast (38,60) .
Nm modifications in the ACL of specific tRNAs can affect translational efficiency and fidelity (61) . Consistently, Nm deposition in mRNA also affected translation through interference with tRNA decoding efficiency and thus could potentially rewire the genetic code (62,63) .
Interestingly, it was recently proposed that TRM7 methylates substrates different than tRNAs including mRNAs in yeast (16,17) suggesting that TRM7 family members can act as multi-substrate Nm-MTases, which modulate translation by methylating mRNA codons and tRNA anticodons. Importantly, loss of Nm at tRNA positions 32 and 34 in trm7 mutant yeast affected translation rates and, consecutively, cell growth (36,38) . Thus, the observed reduction in size and mass ( Figure 2A) in flies without TRM7 family members supports the hypothesis that CG5220 and CG7009 mutations affect translational efficiency in Drosophila .
Importantly, the lack of Nm at the wobble position in CG7009 mutants facilitated tRNA fragmentation. Transfer RNA fragmentation is a conserved response to various stress 13 conditions with functions ranging from protein synthesis, to apoptosis and the modulation of small non-coding RNA pathways (64)(65)(66)(67) . The influence of internal Nm modifications on tRNA stability has only been described very recently (9) . In support of the notion that Nm in tRNAs modulates their stability, we found that Nm 34 is protective against tRNA fragmentation in the ACL. However, in contrast to the Nm-mediated effect on tRNA cleavage in CG7009 mutants, the absence of Nm 32 in CG5220 mutants did not change tRNA fragmentation patterns. Since Hen1-mediated deposition of Nm at RNA 3'-termini (13,68) stabilizes small RNAs in various organisms, we propose that tRFs produced by ACL cleavage in CG7009 mutants may be stabilized through the existence of 3' terminal Nm 32 deposited by CG5220, explaining the abundance and apparent stability of tRFs in CG7009 single mutant in contrast to CG5220 or CG7009 , CG5220 double mutants. To the best of our knowledge this is the first report that involves a TRM7 family protein in tRNA fragmentation. On the other hand, tRNA fragments can associate with Dicer, Argonaute and Piwi proteins (59,(69)(70)(71) . One potential consequence of such interactions could be a reduction in the capacity of small RNA pathway components to process or bind to canonical RNA substrates. Indeed, tRF-mediated titration of proteins has been reported (72)(73)(74) . It remains to be tested whether also the newly identified Nm-MTase substrates ( i.e. , tRNA-Trp, Leu, Gln, Glu) accumulate tRFs in CG7009 and/or CG5220 mutant tissues, which would support the notion that loss of Nm in tRNAs is causative for decreased tRNA stability and thereby could also contribute to the failure of Ago2/Dcr2-dependent small RNA pathways.  (17) .
Our study thus strongly supports the emerging notion that an important biological impact of Nm-MTase activity is mobile element control affecting TEs and viruses. Importantly, our results in Drosophila also indicate that the molecular machinery necessary to deposit Nm in tRNAs and the associated physiological importance are conserved throughout evolution.
In summary, this study provides a comprehensive in vivo characterization of two Nm

Figure. 2. CG7009 and CG5220 affect small RNA silencing pathways and life span.
A, Homozygous CG7009, CG5220 double mutant flies display reduced adult weight and size. Images of adult females and males CG7009, CG5220 homozygous double mutants (homo) compared to heterozygous double mutants (hetero). Below the images is indicated the average weight for flies in milligrams (mg) calculated for 3-day-old flies measured on precision balance. p -value<0,001 in a Student's T-test. B, CG7009 and CG5220 modulate the Ago2-dependent gene silencing in somatic tissues. The UAS> automiW construct is a sensor derived from automiG where two miRNAs were reprogrammed to target the white gene (47) . KD indicates eye-specific GMR-Gal4/UAS-RNAi-mediated inactivation of the respective genes ( white , CG7009 , CG5220 or Ago2 ). Canton-S was used as control for eye color. Darker eye coloration than Canton-S (top right) indicates that the sensor is failing to inactivate the white gene through Ago2-dependent silencing. Flies images were taken at the same age (5 days old). C, The siRNA-dependent viral defence is compromised in CG7009 and CG5220 mutants. RT-qPCR using Drosophila C Virus (DCV) specific primers three days after injection with DCV solution or solution free of DCV as control (not shown) in heterozygous CG7009 e02001 mutants (control) or homozygous CG7009 e02001 (CG7009) and CG5220 K>A (CG5220) mutants. Represented is the relative DCV expression to Rp49. Error bars represent the mean ± s.d (standard deviation) between biological replicates (n=2 replicates where n is a mix of 2 to 3 flies). D, CG7009 mutation is associated with ovarian size reduction. The images show representative examples of ovaries from a 4 days old fertilized female raised on fresh yeast of trans-heterozygous CG7009 e02001 / Def9487 mutant (Mut CG7009) and a genetic rescue ovaries of CG7009 (BAC)/ +; CG7009 e02001 / Def9487 mutant (Rescue CG7009); n>15 for each genotype. WT ovaries (not shown) were similar to rescued ovaries of CG7009 mutant ; Mut: mutant. E , CG7009 and CG5220 are involved in gypsy TE-repression in Drosophila ovaries. Gypsy::lacZ sensor is silenced in follicle cells using tj>Gal4-mediated expression of an UAS-RNAi line (KD) against the white gene (KD control, R; tj>Gal4/ +; Gypsy::lacZ/UAS-white -RNAi), no blue coloration = no β-Gal staining). Gypsy silencing is disrupted using piwi KD (positive control: blue coloration = positive β-Gal staining) and after KD against the indicated genes. The Gypsy::lacZ sensor is also de-repressed in null CG7009 homozygous mutant (KO). KD: knock down, KO: knock out. F, RT-qPCR using gyspy or lacZ specific primers on ovaries from flies KD for white (control) or CG7009 expressing the Gypsy::lacZ sensor (tj>Gal4/ +; Gypsy::lacZ/UAS-RNAi) as described in Figure 2E). Error bars represent the mean ± s.d (standard deviation) between three independent biological replicates. G, CG7009, CG5220 double KD flies display reduced life span. Survival curves of males double KD for CG5220 , CG7009 with (RU200) or without (RU0) RU486-mediated RNAi transgene induction. Constitutive expression (RU200) of CG5220 , CG7009 KD transgenes was induced by RU486 exposure (20 mg/ml during adulthood). The curves represent the average values of at least three biological replicates of 10 flies per experiment.  A , CG33172 and CG15618 modulate Ago2-dependent silencing in adults flies. CG33172 and CG15618 were knocked down by using UAS-RNAi (KD) lines and eye specific GMR-Gal4, wdriver (indicated as KD), as in Figure 2B. Canton-S (wild type w+) and Ago2 KD were used as controls. A darker eye coloration than Canton-expressing automiG lines (top right) indicates that the miRNAs of the sensor are failing to inactivate the white gene through Ago2-dependant silencing. B , CG33172 interacts in vitro with CG7009. Co-immunoprecipitation of tagged CG7009 and CG33172 recombinant proteins co-expressed in bacteria. Anti-GST WB on co-expressed GST::CG7009 and FLAG::CG33172 inputs and after FLAG-IP; Lower panel, Anti-GST WB reveals a GST "alone" signal in the co-expressed GST and FLAG::CG33172. Inputs correspond to 10% of 10 µg of protein eluates. The expected proteins sizes are GST (26 kDa) and GST::CG7009 (62 kDa). WB -Western blot; kDa -kilodaltons. / Def3340 or rescued mutants for CG7009 (rescue BAC) as indicated using a 5'-tRNA Phe(GAA) specific probe and a 5S rRNA probe as loading and transfer control. Mature tRNA Phe size is 73 nt (full length). 5'-tRNA Phe -derived tRNA fragments (5'-tRF Phe ) were detected at~35 nt (halves). The same experiment was performed on heat-shocked flies (1 hours at 37°C in incubator), RNAs were extracted after 5 hours at 25°C (indicated as HS, Heat Shock). nt: nucleotide. B, The same experiment as in Figure 5A above was performed on heat-shocked flies (1 hours at 37°C in incubator), RNAs were extracted after 5 hours of flies recovery at 25°C (indicated as HS, heat shock) with additional indicated genotypes. nt: nucleotide; dbl mutant: Double Mutant CG7009 e02001 , CG5220 K>A .  B, Agarose gel electrophoretic separation of the PCR reaction made on gDNA (genomic DNA) of adult flies CG7009 homozygous or heterozygous for the mutant allele CG7009 e02001 or Ctl -(no DNA) using CG7009-FW and CG7009-Rev primers. Expected sizes in base paired (Kbp) are indicated on the left. C, Reverse Transcription PCR (RT-PCR) on total RNA extracts from ovaries. Electrophoretic separation of the RT-PCR reaction made on total RNA of flies CG7009 homozygous or CG7009 heterozygous for the mutant allele CG7009 e02001 or no RT control (no reverse transcriptase in the RT reaction). The used primers for the PCR reaction are CG7009-middle Rev and CG7009-FW (expected product size 500nt) and tubulin primers (expected product size 150nt). D, Reverse Transcription qPCR (RT-qPCR) on total RNA extracts from adult females of the indicated genotypes. CG7009 e02001 heterozygous (Control), CG7009 e02001 homozygous mutant (Mut CG7009) and rescue CG7009 (BAC). E, Genotyping by PCR on genomic DNA of heterozygous Def9487 and Def3340 and CG7009 e02001 . BAC (rescue) / CyO ; CG7009 e02001 / Def9487 lines, w1118 and CG7009 e02001 homozygous lines. PCR on gDNA extracted from adult single flies with the indicated genotypes using primers CG7009-FW and CG7009-Rev. The band at 1148 bp corresponds to the WT CG7009 locus, the band at 7119 bp corresponds to the mutant allele CG7009 e02001 , containing the inserted PiggyBac transposon (Supplementary Figure S2A, B). BAC: Bacterial Artificial Chromosome containing the wild type CG7009 genomic region; CyO; TM3,Sb; TM6,Tb,Sb: balancer chromosomes; Kbp: Kilo base pairs; gDNA genomic DNA. F, Validation by sanger sequencing of the CRISPR/Cas9 mutants CG5220 K>A and double mutant CG5220 K>A , CG7009 e02001 recombination. Briefly, CG5220 PCR fragments were amplified by PCR from flies gDNA bearing the CG7009 e02001 allele (giving the [w+] phenotype) and the mutations CG5220 K>A . The corresponding simple mutant CG5220 K>A (heterozygous for CG5220 K>A ) was used as positive controls and flies characterized with no CG5220 mutation were used as negative controls ( CG7009 e02001 ). All sequencing experiments were performed on heteroallelic combinations over balanced chromosomes, explaining the double picks at the edited region. The results were obtained using 4Peaks. They correspond to a PCR products obtained using VIE0197/VIE0198 primers and sequenced with the primer VIE0198. The targeted nucleotides are indicated under the red lines.  Figure S4. A, MALDI TOF-MS spectrum of fragments resulting from RNase A digestion of tRNA Phe(GAA) originating from indicated genotypes. B, Top, MALDI TOF-MS spectrum of fragments resulting from RNase A digestion of tRNA Phe(GAA) originating from indicated genotypes (homozygous adult CG7009 e02001 mutants rescued with one CG7009 WT copy (BAC)). Bottom : MALDI TOF-MS spectrum of fragments resulting from RNase T1 digestion of tRNA Phe(GAA) originating from indicated genotypes (homozygous adult CG7009 e02001 mutants rescued with one CG7009 WT copy (BAC)). Relevant peaks are identified by their m/z values. Figure S5. A, related to Figure 3C. RiboMethSeq analysis of tRNA Phe(GAA) modification at positions Cm 32 and Gm 34 . Alkaline fragmentation-based RiboMethSeq protocol was performed on total RNAs extracted from whole flies homozygous mutant for CG5220 K>A , homozygous for CG7009 e02001 / Def3340 and homozygous double mutant CG5220 K>A , CG7009 e02001 as indicated. Normalized cleavage efficiency, calculated from cumulated 5'-end and 3'-end coverage, is shown for the ± 5 neighboring nucleotides. The positions of interest (Cm 32 and Gm 34 ) in tRNA Phe(GAA) are shown by red arrows. Protection against cleavage is indicated (+) protected, (-) not protected. Protection at Cm 32 in control flies is only moderate, indicating incomplete tRNA methylation (+low). Higher methylation levels at this positions are observed in CG7009 homozygous and Rescue CG7009 (BAC) in Figure 3C. B, RiboMethSeq was performed as described in Supplementary Figure S5A Figure S6. A, Pourcentage of amino acid (aa) identity between CG15618, human THADA and yeast TRM732, and between CG33172, human WDR6 and yeast TRM734 (RTT10). Alignment was performed using BLAST/ BLAT tool at www.ensembl.org .

Amino Acid conservation and phylogenetic analysis
Sequence alignments and visualization were done in Kalign ( www.ebi.ac.uk/Tools/msa/kalign/ ) and Unipro UGENE 1.32.0. Percentage of amino acid (aa) identities and coverages between CG7009, CG5220, TRM7, and FTSJ1 proteins were determined on the Ensembl project website ( www.ensembl.org ). For phylogenetic analysis, protein alignments were performed using mafft v7.407 with default parameters (78) . Removal of positions with more than 50% of gaps was obtained by using trimal v1.4 (79) . Phylogenetic analysis was performed using raxml v8.2.12 (80) under the PROTGAMMALG model by combining a rapid bootstrap analysis (100 replicates) and search for the best ML tree (-f a option).

Total RNA Extraction for MALDI-TOF and RiboMethSeq analysis
3-5 days old females and males were homogenized on Precellys 24® tissue homogenizer (Bertin Technology) in 1 mL TRI-reagent (Sigma Aldrich). Total RNA from 20 mL of the fly lysates, was extracted with 8 mL of Chloroform and precipitated with 2/3 volumes of Isopropanol. The pellets were air dried and resuspended in RNase-free water.

Purification of tRNA Phe(GAA)
Total RNA preparations (~7 mg ) were supplemented with LiCl to a final concentration of 0.8 M and incubated overnight at 4°C to precipitate high-molecular mass molecules. The precipitate was eliminated by centrifugation and the supernatant was supplemented with two volumes of 100 % ethanol and incubated at -20°C for two hours to precipitate small RNAs. After centrifugation, pelleted small RNAs were washed twice in 70% ethanol and resuspended in one ml of RNase-free water. tRNAs were further purified using the NucleoBond RNA/DNA 400 kit (Macherey-Nagel) following manufacturer's instructions, except that the elution step was performed with 5 ml of 100 mM Tris-acetate (pH 6.3); 15 % ethanol and 600 mM KCl. Eluted tRNA were ethanol precipitated and resuspended in one ml of RNase-free water. Purification of tRNA Phe(GAA) was performed using a 5' biotinylated complementary oligonucleotide

MALDI-TOF analysis of digested tRNA Phe(GAA)
For mass spectrometry analysis, about 500 ng of tRNA Phe(GAA) were digested with 100 units of RNase T1 (Sigma) in a final volume of 10 µL at 37°C for 4 h. RNase T1 cleaves the phosphodiester bond between the 3'-guanylic residue and the 5'-OH residue of adjacent nucleotides and generates 3'-phosphate nucleosides. One microliter of digest was mixed with 9 µL HPA (40 mg/ml in water : acetonitrile 50:50) and 1 µL of the mixture was spotted on the MALDI plate and air-dried ("dried droplet" method). MALDI-TOF MS analyses were performed directly on the digestion products using an UltrafleXtreme spectrometer (Bruker Daltonique, France). Acquisitions were performed in positive ion mode.

RiboMethSeq
RiboMethSeq analysis of D.melanogaster tRNAs was performed as described in (81) . Briefly, tRNAs extracted from whole flies were fragmented in 50 mM bicarbonate buffer pH 9.2 for 15 min at 95°C. The reaction was stopped by ethanol precipitation. The pellet was washed with 80% ethanol and sizes of generated RNA fragments were assessed by capillary electrophoresis using a Small RNA chip on Bioanalyzer 2100 (Agilent, USA). RNA fragments were directly 3'-end dephosphorylated using 5 U of Antarctic Phosphatase (New England Biolabs, UK) for 30 min at 37°C. After inactivation of the phosphatase for 5 min at 70°C, RNA fragments were phosphorylated at the 5'-end using T4 PNK and 1 mM ATP for 1h at 37°C. End-repaired RNA fragments were then purified using RNeasy MinElute Cleanup kit (QIAGEN, Germany) according to the manufacturer's recommendations. RNA fragments were converted to library using NEBNext® Small RNA Library kit (ref#E7330S, New England Biolabs, UK or equivalent from Illumina, USA) following the manufacturer's instructions. DNA library quality was assessed using a High Sensitivity DNA chip on a Bioanalyzer 2100. Library sequencing was performed on Illumina HiSeq 1000 in a single read mode for 50 nt. Primary analysis of sequencing quality was done with RTA 2.12 software, to insure >Q30 quality score for >95% of obtained sequences.
Following SR50 sequencing run, demultiplexing was done with BclToFastq v2.4, reads not passing quality filter were removed. Raw reads after demultiplexing were trimmed with Trimmomatic v0.32 (82) . Alignment to the reference tDNA sequences was performed with bowtie 2 ver2.2.4 (83) in End-to-End mode. Uniquely mapped reads were extracted from *sam file by RNA ID and converted to *.bed format using bedtools v2.25.0 (84) . Positional counting of 5'-and 3'-ends of each read was performed with awk Unix command. Further treatment steps were done in R environment (v3.0.1). In brief, 5'-end and 3'-end counts were merged together by RNA position and used for calculation of ScoreMEAN (derived from MAX Score described previously), as well as Scores A and B (53) and MethScore (54) . Scores were calculated for 2 neighboring nucleotides. Profiles of RNA cleavage at selected (candidate and previously known) positions were extracted and visually inspected.

Northern blot
For Northern blot analysis of tRNA, 10 µg of total RNA from adults flies were resolved on a 15% urea-polyacrylamide gel (Biorad), transferred to Hybond-NX membrane (GE Healthcare) and EDC-crosslinked (Sigma Aldrich). The membranes were probed with 5′-32 P end-labeled DNA oligonucleotides using T4 polynucleotide kinase (Fermentas). Hybridization was performed overnight at 37°C in PerfectHyb Plus (Sigma) hybridization buffer. Probe sequences are available in the Primers and Probes section . More details on NB procedure are available in (43) .

RNA Interference in S2R+ cells
Double-stranded RNAs (dsRNA) were synthesized by in vitro transcription (MEGAscript® T7 Kit, Ambion) of PCR products amplified from w 1118 genomic DNA using primers flanked by T7 promoters. Sequences of amplicon templates for dsRNA production are available from the Drosophila RNAi Screening Center ( http://www.flyrnai.org/cgi-bin/RNAi_gene_lookup_public.pl ).

S2R+ cell transfection
100 µL of cells at 10 6 cells/ml resuspended in Schneider's Drosophila medium (GIBCO-Invitrogen) were plated in 96-well plates. Cells were transfected with dsRNA or co-transfected with dsRNA and the corresponding sensor using Effectene (Qiagen) following the manufacturer's instructions. 30 minutes after transfection 50 µL Schneider's Drosophila medium (GIBCO-Invitrogen), completed with 10% heat-inactivated fetal calf serum, 100 U/ml penicillin and 100 mg/ml streptomycin were added. Cells were grown at 23°C without CO 2 . After 24 to 48 hours, CuSO 4 was added to a final concentration of 600 µM and GFP fluorescence was followed using an inverted epifluorescence basic microscope. For Ago2-mediated miRNA pathway involvement ( automiG ), cells were co-transfected with 0.1 µg of automiG -vector and 0.32 µg of dsRNA targeting either Ago2, CG7009 or Ftz, Dcr1, Dcr2, Drosha, Ago1. 48 hours later the automiG promoter was induced by adding CuSO 4 to a final concentration of 600 μM (more details in (43) ).
For the luciferase assay experiment, S2R+ cells were treated for 4 days with dsRNA inactivating specifically the indicated genes. Cells were co-transfected with two plasmids expressing the automiW Experiments with the automiW eyes sensor were performed as described in (47) . Eye images were acquired with an Axio-ApoTome (Zeiss) and ZEN2 software.

DCV injection
Flies with the following genotypes were subjected to intra-thoracic injection with the Drosophila Three to four days after the injection and Before death, 3 injected flies from each genotypes and conditions (+ or -DCV) were frozen at -20°C. 3 flies from each condition were then crushed with a pestle in TRI-Reagent (Sigma Aldrich) and total RNA was extracted as described above.
DNase digestion and RT-qPCR were carried out as described with DCV_FW and DCV_Rev specific primers and Rp49 for normalization ( Primers and Probes section ).

CRISPR/Cas9 -mediated genome editing and genotyping
Mutant alleles for CG5220 were generated using CRISPR/Cas9-mediated editing in Drosophila as previously described (85) . The CG5220 K>A allele was obtained using the gRNA (guide RNA)

RT-qPCR
Whole flies or dissected ovaries were crushed with a pestle in TRI-Reagent (Sigma Aldrich).
After DNase digestion of total RNA using the TURBO DNA-free™ Kit (Ambion), 500 ng were used in a reverse transcription reaction with Random Primers (Promega) and SuperScript® II Reverse Transcriptase (Invitrogen). The cDNA was used to perform qPCR on a CFX96 Touch™ Real-Time PCR Detection System (Bio Rad) using target-specific primers. Rp49 was used for normalization ( Primers and Probes section ). The analysis was performed using ∆∆ Ct, error bars represent SD of three biological replicates. Statistical analysis using a Student's T-test was performed and p -values were determined. purified over glutathione-coupled resin (Pharmacia) as previously described (87,88) . The same protocol was used for purification of pET-28a Flag fusion proteins. Bound peptides were eluted with 400 μg/ml Flag peptide (Sigma) in BC100 buffer for 20 min on ice.

In vitro interaction of GST-CG7009 and FLAG-CG33172
Briefly, GST-alone (control) or fusion proteins GST::G7009 (pGEX4T1-CG7009) and FLAG::CG33172 (pET28a-FLAG-CG33172) were co-expressed in C41 (86) bacteria and purified over Flag-coupled resin (Sigma). Bound proteins were washed 3 times in 500 mM KCl and eluted on Bio-spin disposable chromatography columns (Bio-Rad) with flag peptide as described in (88) . Western blot of the immunoprecipitated recombinant proteins was performed as described in the above section using anti-GST HRP (horseradish peroxidase) conjugate (1:10,000 Amersham GE Healthcare) for 60 minutes at room temperature under agitation. HRP was detected by enhanced chemiluminescent (ECL).

RNA-seq on S2R+ cells
Knock down for CG7009 in S2R+ cells was performed consecutively two times day 0 and 3.
Total RNAs were extracted at day 5 as described above. Libraries were prepared using the Illumina TruSeq Sequencing Kit (Illumina) by following the manufacturer's protocol for paired-end and directional sequencing on Illumina HiSeq 2500. The read length was 71 bp paired-end.

Computational analysis of RNAseq experiments
RNA-seq data were mapped against the Drosophila genome assembly BDGP6 (Ensembl release 79) using STAR52 (version 2.4.0). After mapping, bam files were filtered for secondary alignments using samtools (version 1.2). Reads on genes were counted using htseq-count (version 0.6.1p1). After read counting, differential expression analysis was performed between CG7009 KD and scramble KD conditions using DESeq2 (version 1.6.3) and filtered for a false discovery rate (FDR) < 5%. The sample Ctrl_3 was excluded as an outlier from the differential expression analysis.

Imaging
Ovaries and eyes images were acquired with a WILD M3Z (Leica) binocular combined with a Q IMAGING Color 12 bit (Q27959) camera and QCapture Pro software.

Weighing
Average weight for flies in milligrams (mg) was calculated for 3-days-old flies measured on precision balance (>= 0,001g) after dehydration of the bodies. n >=14, where each n is the mean value of 10 flies measured in a batch of 10. p -value<0,001 in a Student's T-test.

Data availability
RNA sequencing data will be deposited at GEO database upon publication of this manuscript.

Acknowledgements
We