Identification of RNA helicases with unwinding activity on angiogenin-processed tRNAs

Abstract Stress-induced tRNA fragmentation upon environmental insult is a conserved cellular process catalysed by endonucleolytic activities targeting mature tRNAs. The resulting tRNA-derived small RNAs (tsRNAs) have been implicated in various biological processes that impact cell-to-cell signalling, cell survival as well as gene expression regulation during embryonic development. However, how endonuclease-targeted tRNAs give rise to individual and potentially biologically active tsRNAs remains poorly understood. Here, we report on the in vivo identification of proteins associated with stress-induced tsRNAs-containing protein complexes, which, together with a ‘tracer tRNA’ assay, were used to uncover enzymatic activities that can bind and process specific endonuclease-targeted tRNAs in vitro. Among those, we identified conserved ATP-dependent RNA helicases which can robustly separate tRNAs with endonuclease-mediated ‘nicks’ in their anticodon loops. These findings shed light on the existence of cellular pathways dedicated to producing individual tsRNAs after stress-induced tRNA hydrolysis, which adds to our understanding as to how tRNA fragmentation and the resulting tsRNAs might exert physiological impact.


INTRODUCTION
Besides their canonical function during protein translation, tRNAs r epr esent a sour ce of a heterogeneous class of small RNAs called tRNA-deri v ed small RNAs (tsR-N As).Currentl y, tsRN As are classified according to their parental tRNA isoacceptor identity, position of tRNA cleavage e v ents as well as their length ( 1 ).Specific tsRNA species, namely 5 and 3 tRNA halv es, hav e been repeatedly detected during cellular responses to stress such as starva tion ( 2 ), oxida tion, hypoxia and hypothermia ( 3 , 4 ) as well as after heat shock or irradiation ( 5 ).The RNase Afamily member Angiogenin (ANG) is the main nuclease in mammals responsible for fragmenting tRNAs during the str ess r esponse ( 3 , 6 ).It does so by targeting pyrimidinepurine dinucleotides in single-stranded regions of tRNAs, with pr efer ence for the anticodon (AC)-loop ( 7 ).Various tsRNA identities, including those that resulted also from hydrolysis e v ents in D-, variab le and T-loops) hav e been implicated in numerous biological processes, including prolifera tion, dif ferentia tion and translation (8)(9)(10)(11)(12)(13), control of transposon activity ( 14 ), intercellular communication ( 15 , 16 ), regulation of vertebrate embryogenesis ( 17 ) and transfer of extra-chromosomal information between generations ( 18 , 19 ).Howe v er, apart from a fe w e xceptions ( 11 , 13 , 20-22 ), how exactly individual tsRNAs act at the molecular le v el remains largely unclear.
Besides the existence of redundant (stress-induced) anticodon nucleases creating tRNA halves (23)(24)(25), and the likely existence of other RNases targeting the D-and Tloops, little is known about the mechanistic details of tsRN A bio genesis after tRN A hydrol ysis.Most notabl y, a major unresolved question relates to the molecular steps following tRN A hydrol ysis by endonucleases.Since tR-NAs are highly structured molecules, formed by numerous secondary and tertiary interactions, which are further r einfor ced by post-transcriptional nucleotide modifications ( 26 ), hydrolysis e v ents in any tRNA loop structure will not destabilise such 'nicked tRNAs' to an extent that would result in the complete separation of 5 from 3 tsRNAs.Howe v er, most reports hav e attributed biological impact to individual 5 and 3 tsRNAs (reviewed in ( 27 )).Furthermor e, many r eports on tsRNAs abundance, including small RNA sequencing experiments after stress exposure, injury or ANG ov ere xpression, re v ealed an asymmetric recovery of 5' and 3' tsRNAs from cells with particular 'jackpot' tsRNAs such as 5' tsRN A-Gl y GCC / CCC and 5' tsRNA-Glu CUC / UUC as overr epr esented in various data sets ( 21 , 23 , 28 , 29 ).While biases in cDNA libr ary prepar ation resulting from RNA modifications could be responsi-ble for these observations ( 30 ), experimental manipulation of biological systems with specific sequences mimicking 5' tsRNAs ( 5 , 8 , 11 , 12 , 21 , 28 , 31 ) or 3' tsRNAs ( 13 , 32-35 ) suggested impact of individual tsRNA moieties on specific cellular processes.If this holds true, this would r equir e the existence of cellular activities with the ability to separate 5 from 3 tsRNAs after endonucleolytic attack on parental tRNAs.Yet, the identity of such activities is currently unknown.
Here, we combined the biochemical enrichment and identification of tsRNAs-containing ribonucleoprotein particles (RNPs) from stressed cells with a versatile in vitro activity assay, which allows characterizing cellular activities that can act on 'nicked tRNAs' in vitro .Using these approaches, we identified various conserved ATP-dependent RNA helicases that can separate ANG-hydrolysed tRNAs.For one of them, DDX3X, we dissected tRNA substrate pr efer ence as well as tRNA binding and unwinding kinetics towards ANG-hydrol ysed tRN As and report that DDX3X function affects the stability of 3 tsRNAs in vivo .These findings implicate multiple RNA helicases in the biogenesis of individual tsRNAs and pave the way for further elucidating the molecular details of tsRN A bio genesis as a pr er equisite for understanding the biological function of these small RNAs.

RNA helicase knock-down using siRNAs in cell culture
HeLa cells were seeded in 6-well plates and transfected with control siRNAs or siRNAs targeting DDX3X or / and DDX5 (see oligonucleotide sequences) at a final concentration of 25 nM using DharmaFECT1 transfection reagent (Horizon Discovery) following the manufacturer's recommendations.After 72 h, cells were collected for cytoplasmic protein extract preparation or RNA extraction.

DDX3X knock-down by short-hairpin RNAs
HeLa or HEK293 cells were infected with 1:20 dilution of lentivirus preparations (produced by using a thirdgener ation lentivir al system in HEK293T cells) expressing a shRNA construct targeting human DDX3X (cloned into pLKO.1,see oligonucleotide sequences) in DMEM-F / G / P / S supplemented with 10 g / ml of polybrene transfection reagent (Sigma).After 24 h of infection, cells were transferred to selection media containing 300 g / ml of G418 for 72 h prior to experimental manipulation.

Rescue-of-function of DDX3X by transient transfection of pCMV-mDDX3X-HA
Mouse BMDF were seeded in 57 cm 2 dishes, and different plasmid masses encoding a CMV-dri v en HA-tagged murine DDX3X gene were transfected using poly-ethylenimine (PEI, Sigma), following the manufacturer's recommendations.After 24 h, BMDF were incubated in medium supplemented with 500 nM 4-hydro xy-tamo xifen (4-OHT, Sigma) for 72 h before protein extraction or experimental manipulation.

Str ess par adigm
For induction of tRNA fragmentation, cultured cells were incubated in media supplemented with 0.5 mM inorganic sodium meta-arsenite (As[III], Sigma) prior to experimental manipulation (2 h for size exclusion, or 1 h for stress granule core enrichment, or 1 h for northern blotting of total RNA).

tRNA isolation under denaturing conditions
Cells were washed once in 1 × PBS, scraped off the culture plate in 1 ml 1 × PBS, centrifuged for 2 min at 500 g at 4 • C and then resuspended in 200 l 1 × PBS. 1 ml of Trizol (Sigma) was immediately added followed by incubation for 10 min on ice.After adding 200 l of chloroform, samples were centrifuged at 21 000 g for 10 min at room temperature.The aqueous phase was re-extracted with an equal volume of chlorof orm, f ollowed by centrifugation at 21 000 g for 5 min at room temperature.The aqueous phase was precipitated using an equal volume of isopropanol and 1 l GlycoBlue (Ambion, 15 mg / ml) for 15 min at room temperature.RNA was collected by centrifugation at 21 000 g for 30 min at 4 • C. RNA pellets were washed once using 75% ethanol, followed by air-drying 5 min at room temperature.Total RNAs extracted from cultured cells were separated by urea-PAGE (6%) using 0.5 × TBE, followed by staining of RNA using SYBR-Gold (Invitro gen).RN As were visualised using a gel documentation system (Biorad), tRNAs were excised from PA gel and eluted in 3 volumes of gel extraction buffer (300 mM sodium acetate; 1 mM EDTA; 0.1% SDS).Eluted tRNAs wer e pr ecipitated using an equal volume of isopropanol and 1 l GlycoBlue (15 mg / ml) for 15 min at room temperatur e. tRNAs wer e r esuspended in water and the concentration was determined by NanoDrop (ThermoScientific).

tRNA isolation under native conditions
Cells were washed once in 1 × PBS supplemented with 5 mM MgCl 2 (PBS-Mg), scraped off the culture plate into 1 ml PBS-Mg, centrifuged for 2 min at 500 g at 4 • C and resuspended in 500 l of PBS-Mg.500 l of acidic phenol (Roth) was immediately added followed by incubation for 10 min on a rotating wheel.Samples were centrifuged at 16 000 g for 7 min at room temperature.The aqueous phase was re-extracted with an equal volume of chlorof orm, f ollowed by centrifuga tion a t 21 000 g for 5 min at room temperature.The aqueous phase was precipitated using an equal volume of isopropanol and 1 l GlycoBlue (15 mg / ml) for 15 min at room temperature.RNA was collected by centrifugation at 21 000 g for 30 min a t 4 • C .RNA pellets were washed once using 75% ethanol / 5mM MgCl 2 , followed by air-drying 5 min at room temperature.Total RNAs extracted from cultured cells under native conditions were separated by nati v e PAGE (6%) using 0.5 × TB supplemented with 5 mM MgCl 2 , followed by staining of RNA using SYBR-Gold (Invitro gen).RN As were visualised using a gel documentation system (Biorad), tRNAs were excised from PA gel and eluted in 3 volumes of gel extraction buffer containing 5 mM MgCl 2 .Eluted tRNAs were precipitated using an equal volume of isopropanol and 1 l GlycoBlue (15 mg / ml) for 15 min at room temperature.Eluted tR-NAs wer e r esuspended in 10 mM MgCl 2 and the concentration was determined by NanoDrop (ThermoScientific).

Purification of specific tRNA isoacceptors by RNA affinity capture
Total RNA extracted from 5 × 145 cm 2 dishes (HeLa cells) was re-suspended in 10 ml of IEX-buffer A (20 mM Tris; 10 mM KCl; 1.5 mM MgCl 2 ).For ion exchange chromatography , a HiT rap Q FF anion exchange chromato gra phy column (1 ml, Cytiva) was used on an ÄKTA-FPLC (Cytiva) at 4 • C and RNA was eluted in IEX-buffer B (20 mM Tris; 1 M NaCl; 10 mM KCl; 1.5 mM MgCl 2 ).Small RNAs eluting between 400-600 mM NaCl were collected and precipitated using isopropanol at -20 • C. The small RNA pool was resuspended in water, melted by incubation at 75 • C for 3 min, immediately placed on ice and 20 ml NHS-binding buffer (30 mM HEPES-KOH; 1.2 M NaCl; 10 mM MgCl 2 ) was added.For RNA affinity capture, 80 g of a 5' aminomodified DNA oligonucleotide complementary to the target tRNA (IDT) was covalently coupled to a HiTrap TM NHS-activated HP column (1 ml, Cytiva).tRNA binding to the NHS-column was performed by r ecir culation of the small RNA pool for 4 h minimum at 60 • C and thereafter washed using NHS-washing buffer (2.5 mM HEPES-KOH; 0.1 M NaCl; 10 mM MgCl 2 ).Bound RNAs were eluted by submerging the column in a water bath at 75 • C in NHS elution buffer (0.5 mM HEPES-KOH; 1 mM EDTA) followed by immediate precipitation in isopropanol at -20 • C .Precipita ted RNA w as re-suspended in w ater.Purified tRN A-Gl y GCC / CCC or tRN A-Lys UUU / CUU was further gel-purified (8% urea-PAGE in 0.5 × TBE) using RNA gel extraction buffer (0.3 M sodium acetate, pH 5.2; 0.1% SDS; 1 mM EDTA) and immediately precipitated in isopropanol a t -20 • C .

Pr epar ation of 5 end-labelled tRNAs
Total tRNAs were extracted from cultured cells by urea-PAGE (6%) as described above.One hundred (100) nanogr ams of extr acted full-length tRNAs (4 pmol) or affinity-purified tRN As (tRN A-Gl y GCC / CCC , tRN A-Lys UUU / CUU ) were de-phosphorylated using alkaline phosphatase (1 unit of FastAP, Thermo) for 20 min at 37 • C, followed by re-extraction with acidic phenol / chloroform / isoamyl alcohol (P / C / I, 25:24:1, Roth) and isopropanol precipitation.tRNAs were resuspended in water and 5 end-labelled using T4 PNK and 32 P-␥ -ATP at 37 • C for 60 min.Unincorporated 32 P-␥ -ATP was removed by desalting using spin columns (P-6DG pol yacrylamide, BioRad). 5 end-labelled tRN As were precipitated in 0.3 M sodium acetate (pH 5.2) and 1 l GlycoBlue (15 mg / ml) in an equal volume of isopropanol for 18 h at 4 • C. tRNA was collected by centrifugation at 21 000 g for 30 min at 4 • C. tRNA pellets were washed once using 75% ethanol, followed by air-drying 5 min at room temperatur e. tRNAs wer e r esuspended in water and stor ed in aliquots at -80 • C until use.

Pr epar ation of 3 end-labelled tRNAs
Total tRNAs were extracted from cultured cells by urea-PAGE (6%) as described above.One hundred nanograms of extracted tRNA (about 4 pmol) were labelled using T4 RNA ligase 1 (NEB) and 5'-32 P-cytidine 3',5'-bisphosphate (pCp) at 37 • C for 60 min.Unincorporated pCp was removed by desalting using spin columns (P-6DG polyacrylamide, BioRad). 3 end-labelled tRNAs were precipitated in 0.3 M sodium acetate (pH 5.2) and 1 l GlycoBlue (15 mg / ml) in an equal volume of isopropanol for 18 h at 4 • C. tRNA was collected by centrifuga tion a t 21 000 g for 30 min a t 4 • C .RNA pellets were washed once using 75% ethanol, followed by air-drying 5 min at room temperature.tRNAs wer e r esuspended in wa ter, melted a t 75 • C for 3 min and immediately placed on ice.T4 PNK buffer A was added to 1 × (final), followed by incubation for 10 min at room temperature.Re-folded tRNAs were processed for ANG-mediated hydrolysis as described for 3'-tracer tRNAs.

Pr epar ation of 'tracer tRNAs' using recombinant ANG
For 3 -tracer tRNAs, 100 ng (4 pmol) of nati v e or refolded tRNA or specific affinity-purified tRNA wer e tr eated with 1 l of recombinant ANG (rANG, R&D, stock concentration 6 M in 1 × PBS containing 1 mg / l BSA) in T4 PNK buf fer A a t 37 • C for 30 min with intermittent mixing by pipetting.Spermine (Sigma) was added to a final concentration of 5 M, and the reaction was incubated at 37 • C for an additional 30 min.After rANG-mediated tRNA hydrolysis, 'nicked' tRNA duplexes were labelled using T4 PNK and 32 P-␥ -ATP at 37 • C for 60 min.3 -tracer tRNAs were r e-pr ecipitated in 1 × T4 PNK Buffer A supplemented with sodium acetate (pH 5.2) to a final concentration of 0.3 M final, 1 l GlycoBlue (15 mg / ml) in an equal volume of isopropanol for 18 h at 4 • C. 3 -tracer tRNAs were used for either RNA helicase activity assays or EMSA.
For 5 -tracer tRNAs, 5 labelled tRNAs (see above) were resuspended in water, melted at 75 • C for 3 min and immediately placed on ice.T4 PNK buffer A (1 l) was added, followed by incubation for 10 min at room temperature.Refolded tRNAs were incubated with rANG as described for 3 -tracer tRNAs.5 -tracer tRNAs were used for either RNA helicase activity assays or EMSA.

Cytoplasmic protein extraction
Cells were collected using trypsin-EDTA (0.05%, Thermo), washed with 1 × PBS, and incubated in a hypotonic buffer (30 mM HEPES-KOH, pH 7.4; 100 mM potassium acetate; 2 mM Mg(OAc) 2 ; 2 mM NaF; 5 mM DTT, 0.1% NP-40) supplemented with murine RNase inhibitor (5 U / 100 l (NEB)) and protease inhibitors (Roche) for 10 min on ice.Cells were broken by passage (20 ×) through a hypodermic needle (26 G).Intact cells were discarded after centrifugation at 300 g for 7 min at 4 • C, followed by removal of nuclei through centrifugation at 1000 g for 10 min at 4 • C. The concentration of the cytoplasmic protein extracts was measured using Bradford reagent (Biorad), followed by snap-freezing aliquots (10 g) in liquid nitrogen and storage at -80 • C until use.

Pr epar ation of HEK293 cells for biochemical fractionation
5 × 145 cm 2 dishes (HEK293 cells) at 80% confluency were acutely exposed to As[III] or left untreated (controls), followed by 2 washes using 10 ml of ice-cold 1 × PBS.Cells wer e cover ed with a thin film of ice-cold 1 × PBS and exposed to UV light (254 nm) with 400 mJ / cm 2 (Stratagene).Cells were scraped off the culture plate into 15 ml 1 × PBS, centrifuged for 3 min at 500 g a t 4 • C .Cell pellets were flashfrozen in liquid nitrogen and stored at -80 • C until further processing.
Cells were lysed by passage (20x) through a hypodermic needle (26 G), followed by centrifuga tion a t 21 000 x g for 5 min a t 4 • C .Soluble protein was saved, followed by reextraction of the cell pellet as described above.The resulting (pooled) soluble protein extract (2 ml) was filtered (45 m) and loaded onto a HiLoad ® 26 / 600 Super de x ® 200 pg SEC column pre-equilibrated in column running buffer (20 mM Tris, pH 7.4; 150 mM NaCl; 1 mM DTT) a t 4 • C .Gel filtration was performed on a ÄKTA Purifier FPLC system (Cytiva) for the duration of 1.1 column volume (CV) at a flow rate of 2.5 ml / min.Fractions (2 ml) were collected with a delay of 0.25 CV to avoid fractionation prior to void volume peak elution.750 l of each fraction was incubated with proteinase K (final concentration of 2 mg / ml) at 37 • C for 10 min under constant mixing (thermo-mixer).RNAs were extracted by adding an equal volume of acidic P / C / I and precipitated using an equal volume of isopropanol and 1 l GlycoBlue (15 mg / ml).RNA pellets were analysed by NB using DNA oligonucleotides for specific tRNAs.5 tsRNAs-containing SEC fractions were thawed on ice, pooled and subjected to anion exchange chromato gra phy (IEX) at 4 • C on a ÄKTA Purifier FPLC system (Cytiva) using HiTrap Q Fast Flow columns (Cytiva, 1 ml) with a flow rate of 1 ml / min in IEX buffer A (20 mM Tris, pH 7.4; 1.5 mM MgCl 2 ; 10 mM KCl).A linear gradient (40-60%) of IEX buffer B (20 mM Tris, pH 7.4; 1.5 mM MgCl 2 ; 10 mM KCl and 1M NaCl) was applied for the duration of 20 CV while fractions (750 l) were collected.Subsequently, RNA was extracted from 250 l of each IEX fraction followed by NB for specific tRNAs.

LC-MS / MS on 5 tsRNAs-containing RNPs
IEX fractions exhibiting a positive 5 tsRNA-Gl y GCC / CCC / tRN A-Gl y GCC / CCC signal ratio on NB were concentrated to an approximate volume of 50 l using a SpeedV ac V acuum Concentr ator on high temper ature settings (60 • C). 12.5 l of 4 × SDS-sample buffer was added to each fraction, followed by dena tura tion for 2 min at 90 • C. Fractions were separated by modified SDS-PAGE pr epar ed by overlaying equal volumes of stacking (4%) and resolving (6%) gels.Pre-stained protein marker (Thermo) was used to monitor sample migration through the stacking gel.Once samples reached the resolving gel, migration was allowed for a pproximatel y one centimetre, followed by fixation and staining of gels using the Colloidal Blue staining kit (Invitr ogen).Excised pr otein samples were processed for in-gel tryptic digestion and analysis b y L C-MS / MS a t the Max Perutz Labora tories (MPL) mass spectrometry facility.Specifically, gel sections were excised, cut up, and de-stained with acetonitrile (ACN) and 50 mM ammonium bicarbonate (ABC).Before each of the following reaction steps, gel pieces were washed with 50 mM ABC in 50% ACN and dehydrated in 100% ACN in order to facilitate the uptake of the solutions.Disulfide bridges wer e r educed in 10 mM dithiothr eitol in 50 mM ABC for 30 min a t 56 • C .Subsequently, free thiols were alkylated with 50 mM iodoacetamide in 50 mM ABC in the dark (30 min at RT).Proteins were digested with trypsin (Promega) in 50 mM ABC overnight at 37 • C. The reaction was stopped by adding 10 l of 10% (v / v) formic acid (FA) and peptides were extracted by sonication with 5% FA.Digests were concentrated and desalted using custommade C18 stage-tips.The purified peptides were separated on an Ultimate 3000 RSLC nano-flow chromato gra phy system (Thermo Fisher Scientific), using a pre-column for sample loading (PepMapAcclaim C18, 2 cm × 0.1 mm, 5 m) and a C18 analytical column (PepMapAcclaim C18, 50 cm × 0.75 mm, 2 m, both Dionex-Thermo Fisher Scientific), a ppl ying a linear gradient from 2 to 35% solvent B (B: 80% acetonitrile, 0.08% FA; solvent A: 0.1% FA) at a flow rate of 230 nl / min over 60 min.Eluting peptides were analysed on a Q Exacti v e HF Orbitrap mass spectrometer equipped with a Proxeon nanospray source (Thermo Fisher Scientific), operated in data-dependent mode.Survey scans were recorded in a scan range of 375-1500 m / z, at a resolution of 60 000 at 200 m / z and an AGC target value of 3E6.The eight most intense ions were selected with an isolation width of 1.6 Da, fragmented in the HCD cell at 27% collision energy and the spectra recorded at a target value of 1E5 and a resolution of 30 000.Peptides with a charge of +1 or > +6 were excluded from fragmenta tion, the peptide ma tch and exclude isotope featur es wer e enabled and selected pr ecursors wer e dynamically excluded from repeated sampling for 15 s.MaxQuant software package, version 1.6.0.16 was used to identify the proteins searching against the human uniprot database (2018.07UP000005640 9606 Homo sapiens all) plus the sequences of common contaminants with tryptic specificity allowing two missed cleavages.Oxidised methionine, acetylation of the protein N-terminus was set as variable, carbamidomethylation of cysteines as fixed modification.
Results were filtered at a false discovery rate of 1% at the protein and PSM le v el.

LC-MS / MS data analysis
Following computational MaxQuant analysis of individual LC-MS / MS runs, data was analysed using the following steps: (i) protein intensities of each sample (consisting of 4 fractions) were combined resulting in three values originating fr om contr ol or As [III] exposure experiments (sum of protein intensities), respecti v ely; (ii) since samples were pr epar ed, processed and measured a t dif ferent days, and to of f-set experimental varia tion such as protein concentration or machine performance, the sum of protein intensities was normalised within each experimental group (control and As[III] exposure) using the mean intensity of a set of three cytoskeletal proteins (sp|P60709|ACTB HUMAN Actin, sp|P68363|TBA1B HUMAN Tubulin ␣-1B chain, sp|P68371|TBB4B HUMAN Tubulin ␤-4B chain); (iii) resulting normalised protein intensities were log 2transf ormed; (iv) an y missing values were separately imputed using a normal distribution for each of the six intensity columns (three originating fr om contr ol or As [III] e xposure e xperiments, respecti v ely); (v) The mean intensity of the values originating from As[III] exposure versus control experiments were used as a measure of protein enrichment, including only those identities with at least two valid values in one of the experimental groups.

Enrichment of SG cores and co-immunoprecipitation from SG cores
HeLa cells were grown to 80% confluency, exposed to As[III] (0.5 mM for 60 min) and collected by trypsinization.Cells were washed once in 1 × PBS, pelleted and flash-frozen in liquid nitrogen.SG core enrichment was performed as described in ( 36 ).Final SG-enriched fractions were subjected to co-immunoprecipitation using antibodies (4 g) against eIF4A1 (Abcam, ab31217) or DDX3X (Bethyl Labs, A300-475A) coupled to 30 l (pr e-clear ed) Dynabeads ™ Protein A (Thermo).Immunopr ecipitations wer e perf ormed f or 2 h a t 4 • C .Aliquots were taken a t r epr esentati v e steps of the enrichment pr otocol, pr otein concentrations were determined using Bradford reagent, and equal mass of total proteins along with the content of the co-immunoprecipitations were analysed by western blotting.

Expression and purification of recombinant human DDX3X
The pETM41 e xpression v ectors harbouring a hexahistidine and MBP (maltose-binding protein) N-terminally fused to human DDX3X or DDX3X DQAD ( 37 ) were expressed in E. coli BL21(DE3) pLysS (Invitrogen) in LB medium supplemented with kanamycin (50 g / ml) and chloramphenicol (50 g / ml).Protein expression was induced at OD 600 = 0.6 using IPTG (1 mM) for 3 h at 37 • C. Bacteria were harvested by centrifugation at 4 • C (15 min at 3500 g).Cell pellets were resuspended in 20 ml His-Buffer A (50 mM Tris, pH 8; 10 mM imidazole; 500 mM KCl) supplemented with protease inhibitors (Roche).Cell disruption was performed by sonication on ice (5 cycles, 90 s work / 30 s pause).Lysates were cleared by centrifugation for 10 min using 21 000 g a t 4 • C .Lysa tes were loaded onto Ni-NTA His Trap High Performance columns (5 ml, Cytiva) that were pre-equilibrated in His-Buffer A. Columns were washed with Buffer A (10 CV) followed by elution of recombinant protein in a 0-100% gradient of His Buffer B (50 mM Tris, pH 8; 500 mM KCl; 1 M imidazole).The elution profile of His-MBP-DDX3X was evaluated using 10% SDS-PAGE and Coomassie-blue staining.Peak fractions (2 ml) containing His-MBP-DDX3X were subjected to gel filtration using a Super de x ® 200 10 / 300 column preequilibr ated in His-MBP-DDX3X stor a ge b uffer (22 mM Tris, pH 8; 330 mM KCl; 1.1 mM DTT) at a 0.5 ml / min flow rate for the duration equivalent to 1.1 CV.All purification steps were performed on the ÄKTA Purifier FPLC system (Cytiva).The elution profile of His-MBP-DDX3X (750 l fractions) was monitored by SDS-PAGE and Coomassieblue staining.His-MBP-DDX3X-containing fractions were supplemented with glycerol (final concentration of 10%), aliquots (20 l) were flash-frozen in liquid nitrogen and stored at -80 • C until usage.

Expression and purification of recombinant murine eIF4A1
The pET3b expression vector harbouring Glutathione Str ansfer ase (GST) C-terminally fused to human eIF4A1 (eIF4A1-GST) ( 38 ) was expressed in E. coli BL21(DE3) pLysS (Invitrogen) as described above for DDX3X, except for 3 h at 30 • C. Cell pellets were resuspended in 10 ml of GST-Buffer A (20 mM Tris, pH 8; 500 mM NaCl; 1 mM DTT; 5% glycerol) supplemented with protease inhibitors (Roche).Cell disruption and clearing of lysate by centrifugation was performed as described for DDX3X.Lysates were loaded onto a GSTrap HiTrap Fast Flow (1 ml, Cytiva) pre-equilibra ted in GST-Buf fer A. The column w as w ashed with Buffer A (10 CV), followed by elution of recombinant protein in a 0-100% gradient of GST-Buffer B (GST-Buffer A supplemented with 10 mM reduced glutathione).The elution profile of GST-eIF4A1 was evaluated by 10% SDS-PAGE and Coomassie-blue staining.Peak fractions (2 ml) containing eIF4A1-GST were subjected to gel filtration on the ÄKTA Purifier FPLC system (Cytiva) using a Super de x ® 200 10 / 300 column pre-equilibrated in storage buffer (22 mM Tris, pH 8; 165 mM NaCl; 1.1 mM DTT) at a 0.5 ml / min flow rate for the duration equivalent to 1.1 CV.The elution profile of eIF4A1-GST (750 l fractions) was monitored by SDS-PAGE and Coomassie-blue staining.eIF4A1-GST-containing fractions were supplemented with glycerol (final concentration of 10%), aliquots (25 l) were flash-frozen in liquid nitrogen and stored at -80 • C until usage.

Expression and purification of recombinant human DDX5
MBP-DDX5-GST was expressed from a pMAL vector harbouring human DDX5 N-terminally fused to MBP and Cterminally fused to GST ( 39 ).MBP-DDX5-GST was expressed in E. coli BL21(DE3) pLysS (Invitrogen) after induction with 0.2 mM IPTG for 18 h at 16 • C, and purified as described for eIF4A1-GST.

Expression and purification of recombinant human DDX1
The pDEST17 expression vector harbouring DDX1 was transformed into E. coli BL21(DE3) pLysS (Invitrogen) and cells were grown in LB medium supplemented with ampicillin (50 g / ml) and chloramphenicol (50 g / ml).Protein expression was induced at OD600 = 0.6 using IPTG (1 mM) for 3 h a t 37 • C .Cells were harvested by centrifugation at 4 • C (15 min at 3500 g).Cell pellets wer e r esuspended in 25 ml His-Buffer A (20 mM Tris, pH 8; 500 mM NaCl; 10 mM imidazole; 2 mM DTT; 10% glycerol) supplemented with protease inhibitors (Roche).Cell disruption was performed by sonication on ice (5 cycles, 90 s work / 30 s pause).Lysates wer e clear ed by centrifuga tion for 10 min a t 21 000 g and 4 • C .Cleared lysa tes were loaded onto Ni-NTA His Trap High Performance columns (5 ml, Cytiva) that were preequilibra ted in His-Buf fer A. Columns were washed with Buffer A (10 CV) followed by elution of recombinant protein in a 0-100% gradient of His Buffer B (20 mM Tris, pH 8; 500 mM NaCl; 2 mM DTT; 1 M imidazole; 10% glycerol).The elution profile of His-DDX1 was evaluated using 10% SDS-PAGE and Coomassie-blue staining.Peak fractions (2 ml) were diluted in 20 ml ion exchange (IEX) buffer A (50 mM Tris, pH 8; 5 mM MgCl 2 ; 2 mM DTT) and loaded on a QFF Column (1 ml, Cytiva) that was pre-equilibrated in IEX Buffer A. Column was washed with Buffer A (10 CV) followed by elution of recombinant protein in IEX Q FF Buffer B (50 mM Tris, pH 8; 5 mM MgCl 2 ; 2 mM DTT; 1 M KCl).The elution profile of His-DDX1 was evaluated using 10% SDS-PAGE and Coomassie-blue staining.Peak fractions (2 ml) containing His-DDX1 were subjected to gel filtration using a Super de x ® 200 10 / 300 column preequilibrated in DDX1 storage buffer (20 mM Tris, pH 8; 200 mM KCl; 5 mM MgCl 2 ; 2 mM DTT) at a 0.5 ml / min flow rate for the duration equivalent to 1.1 CV.All purification steps were performed on the ÄKTA Purifier FPLC system (Cytiva).The elution profile of His-DDX1 (750 l fractions) was monitored by SDS-PAGE and Coomassieblue staining.His-DDX1-containing fractions were supplemented with glycerol (final concentration 10%), aliquots (25 l) were flash-frozen in liquid nitrogen and stored at -80 • C until use.

Expression and purification of recombinant D. melanogaster belle
The coding sequence of the D. melanogaster DDX3X homologue (belle, CG9748) was cloned by synthesis with an N-terminal MBP and a C-terminal His-tag (BioCat, Heidelberg) and expressed in E. coli BL21(DE3) pLysS (Invitrogen) in LB medium supplemented with ampicillin (50 g / ml) and chloramphenicol (50 g / ml).Protein expression was induced at OD600 = 0.6 using IPTG (0.5 mM) at 17 • C for 18 h.Bacteria were harvested by centrifugation (8 min at 3500 g).Cell pellets were resuspended in 10 ml His-buffer A (20 mM Tris, pH 8; 5 mM imidazole; 500 mM KCl; 1 mM DTT) supplemented with protease inhibitors (Roche).Cell disruption was performed by sonication on ice (5 cycles, 60 s work at 60% / 30 s pause).Lysates were cleared by centrifugation for 10 min using 21 000 g a t 4 • C .Cleared lysates were loaded onto Ni-NTA His Trap High Performance columns (5 ml, Cytiva) that were pre-equilibrated in His-buffer A. Columns were washed with buffer A followed by elution of recombinant protein in His-buffer B (20 mM Tris, pH 8; 500 mM imidazole; 500 mM KCl; 1 mM DTT).The elution profile of MBP-belle-His was evaluated using 10% SDS-PAGE and Coomassie-blue staining.Peak fractions (2 ml) containing MBP-belle-His were subjected to gel filtration using a Super de x ® 200 10 / 300 column pre-equilibrated in belle storage buffer (50 mM Tris, pH 8; 300 mM KCl; 1 mM DTT) at a 0.5 ml / min flow rate for the duration equivalent to 1.1 CV.All purification steps were performed on the ÄKTA Purifier FPLC system (Cytiva).The elution profile of MBP-belle-His (750 l fractions) was monitored by SDS-PAGE and Coomassieblue staining.Belle-containing fractions were supplemented with glycerol (final concentration of 10%), aliquots were flash-frozen in liquid nitrogen and stored at -80 • C until use.

AlkB demethylation activity assay
Purified AlkB wt and AlkB D135S were tested for demethylation activity on RNA oligonucleotides harbouring a 5end modification.In a reaction volume of 25 l, 50 or 250 pmol AlkB were incubated with 2 pmol of a synthetic RNA oligonucleotide carrying m 1 A at its 5 end (IBA Sciences) previously 32 P-labelled with T4 PNK (NEB) in demethylation buffer (50 mM HEPES-KOH, pH 8; 1 mM ␣-ketoglutaric acid; 2 mM sodium ascorbate; 75 M (NH 4 ) 2 Fe(SO 4 ) 2 ; 50 g / ml BSA) ( 30 ), 250 ng total RNA extracted from HAP1 cells, and 40 U murine RNase inhibitor (NEB) for 3 or 30 min at 25 • C. Similarly, 50 or 250 pmol of AlkB D135S were used to treat 2 pmol 32 P-labelled m 1 G-modified RNA oligonucleotide (kindly provided by R. Micura / Uni v ersity of Innsbruck) for 3 or 30 min.Aliquots were withdrawn at indicated time points, guanidine hydrochloride was added (final 166 mM), and samples were further processed for TLC as previously described ( 41 ).RNA integrity was verified by denaturing PAGE.

AlkB demethylation of purified tRNA
One microgram of PAGE-purified tRNA from HeLa cells was incubated with a combination of recombinant AlkB wt and AlkB D135S (250 pmol each) in demethylation buffer in a total volume of 25 l for a duration of 45 min.As control, purified tRNA was incubated in demethylation buffer without AlkB.400 l of stop buffer (300 mM sodium acetate, pH 5.0; 0.1% SDS; 1 mM EDTA) was added to demethylation reactions, followed by RNA extraction using acidic P / C / I and precipitation using isopropanol.tRNAs were resuspended in water and used for the preparation of 3 -tracer tRNAs.

Electro-mobility shift assays (EMSA)
Full-length tRNAs as well as tRNAs (full-length or affinityca ptured tRN As) that were subjected to rANG hydrolysis were tested for binding to MBP-DDX3X or eIF4A1-GST using EMSA. 32P-labelled tRNAs or synthetic tRNA-Lys UUU as well as 3'-tracer tRNAs were resuspended in DDX3X helicase buffer (40 mM Tris-Cl, pH 8; 50 mM NaCl; 0.5 mM MgCl 2 ; 2 mM DTT; 0.01% NP40) ( 37 ) or eIF4A1 binding buffer (25 mM Tris, pH 7.5; 1 mM DTT; 100 mM KCl; 5 mM MgCl 2 ) ( 45 ).MBP-DDX3X and GST-eIF4A1 aliquots were thawed on ice and dialysed into either helicase (DDX3X) or binding buffer (eIF4A1) using 0.025 m MCE nitrocellulose membranes (Millipor e).Incr easing amounts of recombinant proteins were pre-incubated with equimolar concentrations of AMP-PNP / MgCl 2 (2 mM) for 5 min at 37 • C. Binding r eactions wer e initiated by addition of either 32 P-labelled tRNAs or 3 -tracer tRNAs, followed by incubation for 30 min at 37 • C. Formed RNPs were exposed to UV light (254 nm) for 3 min (Stratagene).Equal volumes of 2x nati v e loading dye (25% sucrose, 5 mM MgCl 2 ) were added to binding reactions and RNPs were separated using nPAGE (6%) in 0.5 × TB supplemented with 5 mM MgCl 2 .Gels were exposed to storage phosphor screens (Cytiva) for ≤ 2 h and imaged using a Typhoon imager (Cytiva).For the quantification of DDX3X affinities to tRNA-Lys UUU / CUU , the percentages of detectable DDX3Xshifted tRNA-Lys UUU / CUU signal from triplicate EMSAs were determined using Biorad Image Lab, and values were plotted against the molar concentration of MBP-DDX3X.The dissociation constant ( K d ) was calculated by a ppl ying the equation for one site-specific binding curve using Prism 5.

RNA helicase activity assays ( 32 P-labelled 3 -tracer tRNAs)
For testing cytoplasmic protein extracts, 4 pmol of 3tracer tRNAs were resuspended in common RNA helicase buffer (10 mM Tris, pH 8.0; 50 mM KCl; 5 mM MgCl 2 ; 10 mM DTT).Aliquots of cytoplasmic protein extracts w ere thaw ed on ice and dial ysed against common RN A helicase buffer using 0.025 m MCE nitrocellulose membranes (Millipore).Protein concentrations were measured using Bradford and 0.5 to 10 g of dialysed cytoplasmic protein extract were used per RNA helicase reaction.For testing of recombinant proteins, 4-6.5 pmol of 3 -tracer tRNAs or 3 -tracer tRNAs-Gly GCC / CCC were resuspended in either DDX3X helicase buffer ( 37 ), eIF4A1 helicase buffer ( 45 ) or DDX5 helicase buffer ( 39 ) and placed on ice.MBP-DDX3X, MBP-DDX3X DQAD , eIF4A1-GST, MBP-DDX5-GST, His-DDX1 or GST-DHX36 136-988 aliquots w ere thaw ed on ice and dialysed against the respecti v e helicase buffers (canonical or non-canonical) using 0.025 m MCE nitrocellulose membranes (Millipor e).Befor e starting the RNA helicase reaction, protein extracts or recombinant proteins were pre-incubated either with equimolar amounts of ATP / MgCl 2 (2 mM) or AMP-PNP / MgCl 2 for 5 min a t 37 • C .RNA helicase activity was initiated by addition of 5 end-labelled tRNA duplexes, 3 -tracer tRNAs or 3 -tracer tRNAs-Gly GCC / CCC , followed by incuba tion a t 37 • C for the indica ted times.Reactions were stopped by placing reactions on ice and the addition of 2x nati v e loading dye (25% sucrose; 5 mM MgCl 2 ).Reactions were separated by nPAGE (10%) in 0.5 × TB supplemented with 5 mM MgCl 2 .Gels were exposed to storage phosphor screens for ≤ 2 h and imaged using a Typhoon imager (Cytiva).For time-course experiments, aliquots wer e r emoved a t indica ted times, quenched with 2x nati v e loading dye on ice and separated by nPAGE (6%) using 0.5 × TB supplemented with 5 mM MgCl 2 .The percentage of displaced 3 tsRNAs was calculated as follows: Frac SS = Intensity SS / (Intensity DS + Intensity SS ), ss = single-stranded, ds = double-stranded.To calculate unwinding rates for DDX3X reactions, time-course experiments were performed in three replicates and as technical duplicates.After averaging the fraction of duplexes (from technical duplica tes) tha t were not unwound by DDX3X activity, data was fitted to the integrated rate law for a homogeneous first-order reaction (Frac SS = A (1 -e −kt )) as described before ( 46 ), by a ppl ying the equation for one-phase exponential association using Prism 5. Statistical analysis was performed by a ppl ying an unpaired, two-tailed, t-test on replicate data r epr esenting separated 3' tsRNAs (in percent).For experiments using AlkB-treated tRNAs, signals originating from 3' tsRNA displacement at time point zero were subtracted from subsequent time points prior to statistical analysis.

RNA helicase activity assays ('cold' 3 -tracer tRNAs)
For testing tsRNA integrity deri v ed from 3 -tsRNAs, total tRNA was incubated with rANG, PNK treatment was omitted, followed by precipitation using isopropanol in the presence of 5 mM MgCl 2 .About 4 pmol of 'cold' 3 -tracer tRNAs were resuspended in common RNA helicase buffer (10 mM Tris, pH 8.0; 50 mM KCl; 5 mM MgCl 2 ; 10 mM DTT).Aliquots of cytoplasmic protein extracts from different cell lines were thawed on ice and dialysed against common RNA helicase buffer using 0.025 m MCE nitrocellulose membranes (Millipore).Protein concentrations were measured using Bradford, and 5 g of dialysed cytoplas-mic protein extracts were used along with 400 fmol 'cold' 3 -tracer tRNAs per activity assay reaction for 30 min at 37 • C. Reactions were stopped by addition of Proteinase K to 200 g / ml, followed by incubation for another 10 min a t 37 • C .RNA was extracted from reactions using acidic phenol (Roth) followed by ethanol precipitation in 300 mM NaOAc, pH 5. Sample RNAs were processed for denaturing PAGE and NB as described above.

RNA helicase-mediated annealing assays
3 -tracer tRNAs were hea t-dena tured in DDX3X helicase buf fer ( 37 ) a t 75 • C for 2 min to gener ate single-str anded RNAs, and placed on ice.MBP-DDX3X was thawed on ice and dialysed against DDX3X helicase buffer.Annealing reactions were initiated by addition of denatured 3tracer tRNAs, followed by incubation at 37 • C. Aliquots wer e r emoved a t indica ted times and quenched with 2 × nati v e loading dye (25% sucrose, 5 mM MgCl 2 ) on ice.Reactions were separated by nPAGE (10%) in 0.5 × TB supplemented with 5 mM MgCl 2 .Gels were exposed to storage phosphor screens for ≤ 2 h and imaged using a Typhoon imager (Cytiva).

iCLIP analysis
Raw fastq sequences resulting from DDX3X iCLIP and ma tched RNA-Seq da ta were downloaded from the Gene Expression Omnibus (GEO) GSE70804 and subjected to a mapping pipeline (written in Snak emak e) tailored to the accurate quantification of ncRNA ( 47 ).Specifically, r eads wer e trimmed for adapters and sequence quality (PHRED > 30) using Trim Galore! ( http://www.bioinformatics.babraham.ac.uk/projects/trim galore/.) .Reads were then ma pped sequentiall y to ncRN As using Bo wtie allo wing a maximum of two mismatches ( 48 ).Mapping was first performed to r epr esentati v e snRNA and rRNA sequences downloaded from NCBI ( 49 ), followed by mapping to mature tRNA sequences (3 CCA and 5 G added), then to immature tRNA sequences (containing introns and flanking 50 nt), and last to the human mitochondrial chromosome.Remaining reads were mapped to the human genome.tRNA annotations were downloaded from UCSC table browser, which sources the annotations from GtRNAdb ( 50 ).A custom-made script was used to r esolve those r eads that m ulti-ma pped to individual tRN A gene sequences, since tRNA genes within the same isoacceptor do have high sequence similarities.Crucially, since this le v el of sequence similarity is not identical between tRNA isoacceptors, quantification may be biased towards particular tRNA isoacceptors with less sequence similarity between their genes.This mapping strategy is data-dri v en and aimed at creating groups of genes that commonly m ulti-ma p.Even though aimed at resolving tRNA isoacceptors, the strategy occasionally results in merging two or more tRNA isoacceptors.To do so, first, multi-mapping r eads wer e labelled with the tRNA isoacceptor type to which they mapped, followed by only keeping reads with the least mismatches.Consensus isoacceptor groups were then deri v ed based on the percentage of m ulti-ma pping reads the user wishes to retain from each sample.For this mapping, 90% was chosen.For deriving positional informa tion from iCLIP da ta, crosslink positions along a mature tRNA sequence were assigned based on the mode of read starts -1.In case of no mode, read starts were randomly chosen from possible positions indicated by m ulti-ma pping reads.To generate tRN A meta profiles, reads at crosslink positions from each group were summed and divided by all reads mapping to tRNAs in a gi v en sample.All plotting was produced in R, using data.table( https://dplyr.tidyverse.org/refer ence/dplyr-package.html), dpl yr ( https://dpl yr.tidyverse.org/ ) and ggplot2 ( 51 ).
Small RNA sequencing 32 P-labelled tRNA duplexes or RNAs migrating at the mass of 32 P-labelled 3 tsRNAs were excised from nati v e PA gels, and RNA was extracted from crushed gel pieces using RNA extraction buffer (300 mM sodium acetate, pH 5.0; 0.1% SDS; 1 mM EDTA) after overnight incubation at 4 • C.After precipitation of 32 P-labelled RNAs using isopropanol, cDNA libraries were prepared for small RNA sequencing using NEBNe xt multiple x small RNA Library Prep set 2 for Illumina (NEB), following the manufacturer's instructions.Following cDNA synthesis and omitting the 75 • C denaturation step, 32 P-labelled RN A-cDN A hybrids wer e tr eated with RNase H (25 U, NEB) for 30 min at 37 • C, followed by PCR amplification.PCR amplicons were separated from remaining 32 P-labelled RNA fragments by nPAGE in 0.5 × TB supplemented with 5 mM MgCl 2 , normalised for quantity and submitted for sequencing on an Illumina MiSeq (PE75 mode) or NovaSeq S2 (PE150 mode) or NextSeq550 (SR150 mode).

Analysis of small RNA sequencing data
Reads were trimmed for adapter sequences and low-quality bases with q -score < 30 were removed from the 3' end using the cutada pt software.Ma pping of trimmed reads to a colla psed tRN A annotation was performed as previousl y published ( 52 ).In short, collapsed human tRNA genomic space with added CCA sequences on the 3' end was used to create the respecti v e Bowtie2 inde x.After mapping with Bowtie2, the abundance of tRNA-deri v ed reads was extracted using Samtools idxstats, normalised to library sequencing depth, and the normalised tRNA abundance per library was plotted as a heat map using python Seaborn library.In order to calculate fold-changes, the normalised tRNA abundance origina ting from DDX3X-trea ted sRNA libraries were divided by the normalised tRNA abundance originating from hea t-dena tured sRNA libraries from the respecti v e replicate experiment.The obtained values were plotted as a heat map using python Seaborn library.

Production of 'tracer tRNAs' for the in vitro dissection of tsRNA biogenesis
Previous biochemical attempts aimed at identifying tsRNAinteracting proteins mostly utilised synthetic 5 and / or 3 tsRNAs as bait molecules ( 8 , 10 , 11 , 21 , 53 ) while approaches using nati v e tsRNAs remain the e xception ( 54 ).Furthermore, nati v e tRNAs substrates and assay conditions that would allow for the molecular dissection of the biogenesis of stress-induced tsRNAs have not been reported yet.
We reasoned that such an experimental system should be biochemicall y tractable e x vivo , open to genetic manipulation of the individual reactants and relati v ely easy to use when monitoring the processing of tRNAs into tsRNAs in vitro .Most importantly, as post-transcriptional modifications were shown to modulate tRNA cleavage e v ents ( 55 , 56 ), the experimental system should incorporate endo genousl y modified tRN As as substrates for the activity of anticodon nucleases and for activities binding to and acting on 'nicked' tRNAs.To produce endogenously modified and, importantl y, structured tRN As with a break in the ACloop as a mimic for stress-induced tRNA fragmentation, bulk tRNAs were purified from human cells, followed by 5 or 3 end-labelling using ␥ -32 P-ATP or 3 ,5 [5 -32 P]pCp, respecti v el y. 32 P-labelled tRN As were hea t-dena tur ed, r efolded in the presence of Mg 2+ -ions and incubated with recombinant ANG (rANG).Nati v e polyacrylamide gel electrophoresis (nPAGE) was used to monitor the integrity of rANG-processed tRNAs (Supplementary Figure S1A).The results showed that 32 P-labelled tRNAs could be maintained as duplexes after rANG-mediated processing since hea t dena tura tion prior to nPAGE resulted in faster migration of 5 tsRNA moieties r epr esenting only a fraction of 3 2 P-labelled RNAs (Supplementary Figure S1B).In contr ast, rANG incubation of 3 end-r adiolabelled tRNAs resulted in the loss of 3 2 P-label, indicating that rANG removed the non-templa ted CCA a t 3 ends of tRNAs (Supplementary Figure S1C), which confirmed previous observations ( 57 ), but excluded 3 end-labelling as an option for following the fate of 'nicked' tRNAs.ANG-mediated RNA hydrol ysis creates 3 RN A moieties containing a 5 hydroxyl w hile 5 RN A moieties contain a 2 , 3 -cyclic phosphate (cycP) at their 3 ends ( 9 , 58 ).Notably, ANG displays some tRNA substrate specificity ( 23 , 57 ).To label only those tR-NAs that were targeted by ANG, purified human tRNAs were exposed to rANG followed by 3 2 P-labelling at 5 hydr oxyl gr oups that were crea ted a t 3 tsRNA moieties after ANG-mediated hydrolysis (Figure 1 A).nPAGE re v ealed that those tRNAs containing rANG-mediated 'nicks' migrated as duplexes, which consisted of 3 2 P-labelled 3 tsR-NAs bound to non-labelled 5 tsRNAs (Figure 1 B).This indica ted tha t specifically labelling tsRNA moieties after rANG-processing allowed tracing the fate of ANGprocessed substrates (termed 3'-tracer tRNAs from here on), w hile leaving tRN As tha t were non-ANG substra tes invisible to detection by phospho-imaging.To determine the identity of 3'-tracer tRNAs, they were eluted from nPAGE, linker-ligated,re v erse-transcribed, and cDNAs corresponding to the lengths of tsRNAs (20-40 nucleotides) were sequenced (sRNAseq).Read mapping was performed to nonredundant tRNA sequence clusters ( 52 ) to avoid multimapping biases of short or post-transcriptionally modified tRNA-deri v ed reads.Quantification of the relati v e abundance of tRNA-deri v ed reads re v ealed that only a few tRNA isoacceptors, all of which were previously reported to be ANG-substrates ( 23 , 57 ), contributed substantially to 3'-tracer tRNAs (Supplementary Figure S1D).These combined results showed that duplex structures of ANGprocessed tRNAs can be maintained as 3'-tracer tRNAs in vitro , opening up the possibility for dissecting their molecular processing into individual tsRNA moieties through experimental manipulation.

Mammalian cell extracts contain activities which separate and degrade 3'-tracer tRNAs
Specific tsRNAs can be detected in the cytoplasm of acutely stressed cells ( 54 , 59 ), indica ting tha t stress-induced tsRNA biogenesis is likely a cytoplasmic process.To determine if 3'-tracer tRNAs can be used to re v eal cellular acti vities tha t af fect the structural integrity of rANG-processed tR-N As, 3 -tracer tRN As were incubated with cytoplasmic protein extracts from mammalian cell lines and analysed by nPAGE.This re v ealed a faster migration of labelled 3 tsR-NAs that was ATP-dependent, albeit with noticeable loss of label from 3 tsRN As, w hich was also cell line-dependent (Figur e 1 C).Notably, PNK tr eatment of RNAs containing 3 cycP will result in dephosphorylation of these moieties, hence 3 -tracer tRNAs produced in this way are likely to become accessible to 3 -5 exonuclease activities.To address if the change in migration of labelled 3 tsRNAs from 3 -tracer tRNAs could have been the result of degrading non-labelled 5 tsRNAs, mature human tRNAs were 3 2 Plabelled at 5 moieties befor e exposur e to rANG, incubation with cytoplasmic protein extracts and analysis by nPAGE.
The results showed an ATP-dependent mobility increase of labelled 5 moieties originating from 5 -tracer tRNAs, the migration pattern of which was comparable to those of hea t-dena tured 5 -tracer tRNAs (Supplementary Figure S1E).Furthermore, northern blotting (NB) of 'cold' (non-labelled) 3 -tracer tRNAs for both 5 and 3 tsRNA-Gly GCC / CCC after exposure to cytoplasmic protein extracts re v ealed either equal le v els of both 5 and 3 tsRNAs (HeLaderi v ed e xtracts), or lower le v els of 5 tsRNAs (extracts deri v ed from other cell lines (Figure 1 D, E)).In addition, analysis of the integrity of the 3 moieties in 3 -tracer tRNAs using denaturing PAGE, indicated a gradual loss of 3 2 P-label and also 3 -terminal nucleotides after exposure to cell extract, both of which was protein concentration-dependent (Supplementary Figure S1F) and indicated cellular activities which degraded 3 tsRNAs from their 3 ends.Of note, the le v els of label loss from 3 -tracer tRNAs v ersus maintenance of 3 -tracer tRNA integrity were dependent on the buffer composition into which cytoplasmic protein extracts wer e dialysed befor e the activity assays on 3 -tracer tRNAs.This observation suggested the presence of competing cellular activities acting on 3 -tracer tRNAs (degrading versus unwinding 5 from 3 tsRNAs), the balance of which was determined by the particular buffer composition (Supplementary Figure S1G).These results indicated that cytoplasmic protein extracts contained ATP-dependent activities with the ability of separating 5 tsRNAs from 3 tsRNAs while partially degrading both tsRNA moieties from their respecti v e 3 ends.

Str ess-induced tsRNAs ar e associated with RNPs containing RNA helicases
Pre vious wor k aimed at identifying tsRNA-interacting proteins utilised e xclusi v el y e x-vivo RN A affinity a pproaches with specific tsRNAs as bait ( 8 , 10 , 11 , 21 , 53 , 54 ).Notably, the value of such approaches remains limited since protein binding to a respecti v e tsRNA species was usually forced by using exorbitant amounts of RNA bait in conjunction with ill-defined compositions of protein extracts.We reasoned that factors which bind and potentially process 'nicked' tRNAs should be in close physical proximity with stressinduced tsRNAs.To identify such factors in vivo , tsRNAcontaining RNPs were enriched by biochemical fractionation and identified by mass spectrometry.To do so, HEK293 cells were transiently exposed to inorganic sodium metaarsenite (As[III]), followed by UV-crosslinking and protein extraction under non-denaturing conditions (Figure 2 A).Western blotting for phosphorylated eIF2 ␣, a marker for the integrated str ess r esponse, verified the activation of str ess r esponse pathways upon As[III] exposur e (Supplementary Figure S2A).NB for tRN A-Gl y GCC / CCC , tRN A isoacceptors whose associated 5 tsRNAs have often been r eported as overr epr esented in various data sets ( 18 , 19 , 23 ), confirmed the partitioning of tsRNAs deri v ed from tRNA-Gly GCC / CCC to the soluble protein fraction with an apparent 5 / 3 asymmetry in tsRNA le v els towar ds the maintenance of 5 tsRNAs (Supplementary Figure S2B).Protein extr acts were fr actiona ted using size exclusion chroma tography (SEC, Figure 2 B) and RNPs co-migrating with 5 tsRNAs were traced by NB for various tRNA isoacceptors (Figure 2 C).SEC fractions containing 5 tsRNAs were pooled and further separated by ion exchange chromatography (IEX, Supplementary Figure S2C).This re v ealed that all 5 tsRNAs-containing IEX fractions also contained tRNAs, albeit at different r atios.IEX fr actions containing a positi v e 5 tsRN A / tRN A-Gl y GCC / CCC signal ratio on NB (Figure 2 D and Supplementary Figure S2D) were subjected to mass spectrometry.To identify proteins comigrating with full-length tRNAs, SEC and IEX fractionations were also performed on protein extracts from cells not exposed to As[III] (controls, Figure 2 D and Supplementary Figure S2D).Triplicate mass spectrometry analysis of control RNPs as well as RNPs containing more 5 tsRNAs than parental tRNAs resulted in the identification of a total of 395 proteins in at least two out of thr ee r eplicate experiments (Figur e 2 E and Supplementary Table S1).Among those, 267 proteins were enriched by a factor ≥ 2 over controls (Figure 2 F), suggesting these proteins were either migrating with 5 tsRNAs-containing RNPs, or increased their migration with full-length tRNAs upon As[III] exposure.GO annotation for molecular function re v ealed a statistically significant over-r epr esentation of 'RNA binding' in fractionated RNPs, supporting the notion that a combination of SEC and IEX allowed enriching for proteins with the potential to interact with tsR-NAs and / or tRNAs (Supplementary Figure S2E and Supplementary Table S2).Indeed, several proteins that had previously been connected to tRNA biology (TRMT10C, RTCA, EIF2S1, EEF1G, TUFM and EIF5A) co-migrated as enriched over controls upon As[III] exposure (Supplementary Figure S2F and Supplementary Table S1).Notably, 114 proteins, many of them with the potential to bind RNA (Supplementary Figure S2G and Supplementary Tab le S2), were e xclusi v el y co-migrating with 5 tsRN Ascontaining RNPs, as they were not detected in the corresponding fractions collected from cells grown under control conditions (Figure 2 F and Supplementary Table S1).This protein subset was manually curated to extract candidate proteins with the potential to affect tsRN A bio genesis or function.Among these, transla tion initia tion and  elongation factors (eIF3A / 3B / 3E and eEF1A / B2 / D / G), proteins involved in energy metabolism (ENOPH1, IDH1, GAPDHS , CS , G6PD), as well as se v eral enzymes involv ed in RNA processing (RNAseT2, eIF4A1 / A2, DDX5 / 17, DDX3X, DDX39A / B) were noticed (Supplementary Figure S2H, I and Supplementary Table S1).Especially the latter group of proteins suggested the physical proximity of stress-induced 5 tsRNAs with enzymatic activities that had the potential for separating parental tRNAs containing ANG-mediated 'nicks'.

Various RNA helicases colocalize with stress granules during the response to as[III]
Pr evious r eports associated specific str ess-induced 5 tsR-NAs with stress granule (SG) assembly ( 5 , 8 , 60 ).Furthermore, ANG colocalized with SG markers upon As[III] exposure ( 61 ) suggesting that stress-induced tRN A hydrol ysis and, potentially, the processing of 'nicked' tRNAs occurs in proximity to SG.Using two different antibodies against endogenous ANG confirmed its colocalization with G3BP1, an ubiquitous SG marker ( 62 ), in response to As[III] exposure (Supplementary Figure S3A, B).Notably, a subset of proteins tha t co-migra ted with 5 tsRN As / tRN Ascontaining RNPs had been previously linked to SG (Supplementary Figure S3C and Supplementary Table S3), supporting the notion that ANG-mediated tRN A hydrol ysis and processing might occur in proximity to SG.The co-migration of se v eral DEAD box RNA helicases with 5 tsRN As / tRN As-containing RNPs in vivo and the observa tion tha t cytoplasmic acti vities were ab le to separate 'nicked' tRNAs in vitro prompted us to investigate the subcellular localization of various RNA helicases during the str ess r esponse.Indir ect immunofluor escence detection of candidate RNA helicases previously connected to SG biology, as well as of those helicases co-migrating with 5 -tsRN As / tRN As-containing RNPs (Supplementary Figure S3D) re v ealed cytoplasmic, nuclear localization and combinations thereof under steady-state conditions (Figure 3 A).Notabl y, onl y a subset of these RNA helicases re-localized towar ds SG upon As[III] e xposur e (Figur e 3 B) suggesting functional relevance in or close to SG.Among those, eIF4A1 and DDX3X displayed absolute colocalization, while DDX1, DHX36 and DDX5 showed partial colocalization with G3BP1-positi v e SG.These data suggested physical proximity of ANG and specific RNA helicases at those subcellular sites that were previously connected to stress-induced tRNA hydrolysis and tsRNA biogenesis.

Specific RNA helicases co-precipitate in SG-enriched subcellular compartments
To biochemically validate the link between SG-associated RNA helicases and As[III]-induced tsRNAs, cofractionation experiments wer e r epeated and western blotting was performed on IEX fractions exhibiting positi v e 5 tsRN A / tRN A signal ratios on NB.The results confirmed the co-migration of eIF4A1 and DDX3X with tRN As / tsRN As-containing RNPs (Supplementary Figure S3E), while other RNA helicases were undetectable by western blotting.To directly determine which RNA helicases physically associated with SG, HeLa cells were exposed to As[III], followed by enrichment of SG cores ( 36 ) and immunoprecipitation of eIF4A1, an RNA helicase that is highly enriched in SG ( 63 ).Western blotting revealed the co-precipitation of eIF4A1 with the SG-marker G3BP1 and se v eral RNA helicases (DDX1, DDX3X, DHX36, Figure 3 C).Notably, reciprocal immunoprecipitation using antibodies against DDX3X re v ealed that SG core-enriched DDX3X can also form protein complexes with DDX5 and / or DDX1 without eIF4A1 and DHX36 (Figure 3 D).These combined results indicated proximity of various RNA with SGs and the products of stress-induced tRNA fr agmentation r aising the question whether one or more of these (candidate) RNA helicases would accept 'nicked' tRNAs as substrates.

Various RNA helicases display unwinding activity tow ar ds 3'tracer tRNAs
DDX5, DHX36 and DDX1 had been previously connected to tRN A biolo gy (64)(65)(66), w hile DDX3X and eIF4A have not been implicated in tRNA-related functions.To determine if these RNA helicases were able to catalyse the unwinding of 3 -tracer tRNAs, eIF4A1, DDX1, DDX3X and DDX5 were purified as recombinant fusion proteins (eIF4A1-GST, His-DDX1, MBP-DDX3X, MBP-DDX5-GST) w hile an N-terminall y truncated recombinant DHX36 (GST-DHX36 136-988 ) was commercially obtained.All recombinant proteins were tested for ATPase acti vity in pub lished reaction buffers (canonical buffers), w hich had previousl y supported their respecti v e acti vities on synthetic RNA substrates ( 37 , 39 , 42 , 44 , 45 ).All proteins hydrolysed ATP in the presence of RNA (Supplementary Figure S4A) indicating the purification of acti v e enzymes.Notably, MBP-DDX3X and GST-DHX36 136-988 hydrolysed some ATP in the absence of analyte RNA suggesting that these enzyme preparations contained traces of co-purified RNA or free phosphate moieties.Purified enzymes were incubated with 3 -tracer tRNAs in their respecti v e canonical buffers and subjected to fixed time point acti vity assays.nPAGE re v ealed that MBP-DDX3X and MBP-DDX5-GST separated 3 tsRNAs from a fraction of 3 -tracer tRNAs in a protein concentration-dependent fashion while equimolar amounts of eIF4A1-GST showed very weak activity (Figure 4 A).Notably, DDX3X activity was detectable without the addition of ATP suggesting either co-purification of A TP or A TP-independent activity on 3 -tracer tRNA.In contrast, equimolar amounts of neither His-DDX1 nor GST-DHX36 136-988 showed activity on 3 -tracer tRNAs (Figure 4 B, C).Notably, testing other buffer conditions (non-canonical buffers) re v ealed that, in addition to MBP-DDX3X and MBP-DDX5-GST, also eIF4A-GST displayed robust RNA helicase activity on 3 -tracer tRNAs (Figure 4 D).In contrast, His-DDX1 and GST-DHX36 136-988 remained inacti v e in all tested buffers (Supplementary Figure S4B, C).These results indicated that, under particular in vitro assay conditions, more than one DEAD box RNA helicase accepted tRNAs containing ANG-mediated 'nicks' as substrates resulting in the separation of 5 from 3 tsRNAs.

DDX3X interacts with specific tRNAs
DDX3X, DDX5 and eIF4A1 are RNA helicases involved in many cellular processes ( 67 , 68 ).Although none of these RNA helicases have been implicated in mammalian tRNA or tsRN A biolo gy, we focused onl y on DDX3X for further molecular characterization since this RNA helicase was detected by all employed enrichment approaches, and because DDX3X is thought to act on rather complex RNA structures in various RNA species (re vie wed in ( 68)).To valida te tha t DDX3X directl y binds to tRN A-deri v ed sequences, electrophoretic-mobility shift assays (EMSA) were performed using 5 end-labelled and refolded tRNAs.Using increasing enzyme molarities re v ealed that MBP-DDX3X shifted a portion of the tRNA pool, indicating MBP-DDX3X bound to specific tRNAs (Supplementary Figure S4D).In contrast, eIF4A1-GST, which also showed buffer composition-dependent activity on 3 -tracer tRNAs, did not bind to tRNAs in its canonical buffer and at protein concentrations equimolar to MBP-DDX3X (Supplementary Figure S4D).To determine if DDX3X interacted with tRNAs in vivo , published data obtained from individual nucleotide cross-linking and immunoprecipitation (iCLIP) ex-periments in HEK293 cells ectopically expressing DDX3X-FLAG ( 69 ) were re-analysed using a ne wly de v eloped noncoding RN A (ncRN A) 'aware' ma pping tool.This tool performs sequential read mapping to various categories of abundant repetiti v e ncRNAs resulting in more accurate quantifica tion of iCLIP-media ted protein binding to ncR-N As, including tRN As, especiall y since such reads are often underr epr esented when using standard mapping pipelines due to missing annotations and unresolved m ultima pping issues.The analysis re v ealed that between 10-20% of all iCLIP-deri v ed reads mapped to the human tRNA sequence space (Figure 5 A) suggesting that DDX3X interacted directl y with tRN As in vivo .Notabl y, As[III] exposur e r esulted in slightly increased iCLIP signals on tRNAs, which was not observed in cells over expr essing a DDX3X helicase domain mutant (DDX3X R534H , Figure 5   through 3 moieties (Figure 5 C).Importantly, the analysis also re v ealed that DDX3X binding to tRNAs in vivo e xhibited specificity towards particular tRNA isoacceptors (Figure 5 D).This specificity was not a function of tRNA expression since normalising the abundance of tRNA-deri v ed reads to their expression (based on matched total RNAseq datasets ( 69 )) re v ealed no correlation between DDX3X binding and tRNA expr ession, ther efor e suggesting binding specificity of DDX3X towards particular tRNAs (Figure 5 E).To validate the binding of DDX3X to specific tR-N As, tRN A-Lys UUU / CUU , w hich was among the top tRNAderi v ed iCLIP hits was affinity-purified from mammalian cells, 32 P-labelled at the 5 end, refolded and subjected to EMSA using MBP-DDX3X.The results showed that MBP-DDX3X bound purified endogenous tRNA-Lys UUU / CUU quantitati v ely and by forming a single RNA-protein complex (Figure 5 F and Supplementary Figure S4E) with a dissociation constant ( K d ) of 83.4 ± 47 nM (Figure 5 H and Supplementary Table S4).As controls, EMSA were performed using synthetic tRNA-Lys UUU (Figure 5 G), which showed that MBP-DDX3X formed various RNA-protein complexes with a fraction of synthetic tRNA-Lys UUU and a K d of 431 ± 295 nM (Figure 5 H and Supplementary Table S4).These results suggested that the binding of MBP-DDX3X to specific tRNAs was lik ely go verned by corr ect RNA structur es that ar e supported by particular RNA modifications, which were absent in synthetic tRNA sequences.

DDX3X displays unwinding activity on structured 3'-tracer tRNAs
To validate that DDX3X directly binds tRNAs containing ANG-mediated 'nicks', 3'-tracer tRNAs and MBP-DDX3X were incubated in its canonical buffer and analysed by EMSA.To minimise the release of RNA helicase substrate, the non-hydrol ysable ATP analo gue AMP-PNP was used.The results showed that MBP-DDX3X could directly associate with a fraction of 3'-tracer tRNAs (Supplementary Figure S5A).Notably, eIF4A1-GST, which also showed activity on 3 -tracer tRNAs (albeit in non-canonical buffers), did not stably bind 3 -tracer tRNAs in its canonical buffer (Supplementary Figure S5A).Furthermore, dena tura tion of 3'-tracer tRNAs prior to EMSA did not result in RNP formation indicating that MBP-DDX3X interacted with structured 3'-tracer tRNAs but not individual 3 tsRNAs (Supplementary Figure S5A).To address if DDX3X displayed ATP-dependent helicase activity towards 3'-tracer tRNAs, in vitro activity assays were performed in the presence or absence of ATP.The results showed MBP-DDX3X-mediated separation of 3 from 5 tsRNAs occurred also independently of additional ATP (Figur e 6 A) confirming pr evious observations.To r emove potentially contaminating ATP from the MBP-DDX3X pr eparation, MBP-DDX3X was pr e-incubated with AMP-PNP to displace ATP, w hich clearl y abro gated DDX3X activity on 3 -tracer tRNAs (Figure 6 A).In addition, an ATPase-deficient version of recombinant DDX3X (MBP-DDX3X DQAD ) displayed no activity on 3 -tracer tRNAs (Figure 6 A), confirming that MBP-DDX3X displaced 3 tsRNAs from 3'-tracer tRNAs in an ATP-dependent fash-ion.Replicate single-turnover time-course activity assays, performed under pre-steady state conditions, re v ealed that MBP-DDX3X was only acti v e on a fraction of 3'-tracer tR-NAs (Figure 6 B and Supplementary Figure S5B) with an observed 'unwinding' constant ( k unw ) of 0.4347 ± 0.04352 min −1 and an 'unwinding' amplitude of 49.82 ± 0.01% (Figure 6 C and Supplementary Table S4).Because 'unwinding' amplitudes of RNA helicases could be the result of concurrent RNA duplex separation and RNA annealing ( 70 ), MBP-DDX3X was also incubated with denatured 3'-tracer tRNAs in the absence of ATP.Analysis for potential annealing activity using single-turnover kinetic assa ys f ollowed by nPAGE indicated none to very weak MBP-DDX3Xmedia ted forma tion of RNA duplexes (Supplementary Figure S5C).These results indicated that MBP-DDX3X exhibited distincti v ely grea ter separa tion activity on 3'-tracer tR-NAs than annealing activity on denatured 3 -tracer tRNAs under these experimental conditions.

Evolutionary conservation of DDX3X activity on ANGprocessed tRNAs
DDX3X genes are evolutionary conserved ( 71 ).To determine if DDX3X activity on tRNAs containing ANG-mediated 'nicks' is conserved, belle, the Drosophila melanogaster homologue of human DDX3X ( 72 ) was purified as a fusion protein (MBP-belle) and tested for ATPase activity.MBP-belle hydrolysed ATP in the presence of RNA (Supplementary Figure S5D) indicating the purification of acti v e enzyme.MBP-belle was incubated with 3 -tracer tRN As, w hich were produced by incubating purified tRNA from Drosophila embryos with rANG, which had previously been shown to process tRNAs expressed by Drosophila cells ( 73 ).These 3 -tracer tRNAs were incubated with MBP-belle in different reaction buffers and MBP-belle activity was analysed by nPAGE.The results showed the separation of 3 tsRNAs from a fraction of 3 -tracer tRNAs under particular reaction buffer conditions by MBP-belle (Supplementary Figure S5E).Using these buffer conditions, replicate single-turnover time-course activity assays were performed at pre-steady state conditions resulting in a k unw of 0.934 ± 0.176 min −1 and an 'unwinding' amplitude of 0.36 ± 0.014% (Supplementary Figure S5F and Supplementary Table S4).These results confirmed that belle, the fruit fly homologue of mammalian DDX3X, was able to bind and separate 3 -tracer tRNAs in vitro .

DDX3X activity displays specificity tow ar ds particular ANG-substrate tRNAs
Since the activity of MBP-DDX3X affected only a fraction of 3'-tracer tRNAs, fixed-time point activity assays were perf ormed, f ollowed by gel purification of RNAs migra ting a t the le v el of displaced 3 tsRNAs.As controls, tsRNAs that had been displaced by dena tura tion wer e eluted (Figur e 6 D).sRNAseq was performed to r eveal the identities of tsRNAs that were produced by rANG followed by hea t dena tura tion (all potential RNA helicase substrates) versus those tsRNAs displaced by MBP-DDX3X activity only.Using the ranking of the most abundant tRNA-deri v ed reads after rANG incubation (Supplementary Figure S1D) as basis and plotting the relati v e abundance of tsRNA-deri v ed reads resulting from heat dena tura tion versus MBP-DDX3X-media ted activity on 3'tracer tRNAs suggested tha t MBP-DDX3X separa ted 3tracer tRNAs not strictly based on tRNA abundance in the 3 -tracer tRNA pool (Figure 6 E).Notably, comparing the relati v e abundance of tsRNA-deri v ed r eads r esulting from MBP-DDX3X-mediated activity on 3'-tracer tR-NAs with those obtained after dena tura tion of 3'-tracer tR-NAs showed that a positi v e fold-change was only detectable for a subset of tRNAs (tRNA-Lys CUU , tRNA-Val AA C / CA C , tRN A-Gl y GCC / CCC , Figure 6 F).These observations suggested that particular tRNA identities r epr esented pr eferential substrates for DDX3X-mediated activity on 3tracer tRNAs.To test this directly, tRNA-Gly GCC / CCC was affinity-purified from human cells, 5 -labelled or processed to obtain 3'-tracer tRN A-Gl y GCC / CCC , both of w hich were subjected to EMSA using MBP-DDX3X.The results showed that both tRN A-Gl y GCC / CCC as well as 3'tracer tRN A-Gl y GCC / CCC could be shifted to similarl y sized RNPs, which indica ted tha t MBP-DDX3X interacted with both tRN A-Gl y GCC / CCC preparations (Supplementary Figure S6A).Notab ly, single-turnov er time-course acti vity assa ys perf ormed under pre-steady state conditions re v ealed tha t MBP-DDX3X quantita ti v el y separated 3 tsRN As from 3'-tracer tRN A-Gl y GCC / CCC (Figur e 6 G).These r esults confirmed that MBP-DDX3X displayed specificity towards particular isoacceptors within the pool of 3'-tracer tRNAs, suggesting (unknown) denominators determining DDX3X activity.

Specific RNA modifications affect RNA structure but not DDX3X activity on 3 -tracer tRNAs
It is currently assumed that DEAD box RNA helicases reco gnize RN As through structural elements with poorly defined sequence context ( 74 ).In support of this notion, CLIP-based approaches have not produced testable consensus sequence motifs for DDX3X binding to RNAs, albeit complex secondary structures and high GC content have been suggested as targets of DDX3X binding and activity ( 69 , 75 , 76 ).To address if RNA modifications could affect DDX3X activity towards specific 3 -tracer tRNAs, including 3'-tracer tRN A-Gl y GCC / CCC , we aimed a t modula ting the RNA modification status of 3 -tracer tRNAs.As an unmodified tRNA sequence, a synthetic mimic of tRNA-Gly GCC harbouring no modifications was 32 P-labelled at the 5 -end, folded in the presence of Mg 2+ and subjected to rANG-mediated h ydrolysis.nPA GE analysis of this 5tracer tRNA sequence after dena tura tion re v ealed no 5 tsRNAs with higher mobility suggesting that rANG was unable to efficientl y hydrol yse synthetic tRN A-Gl y GCC sequences (Supplementary Figure S6B).This indicated that in the absence of RNA modifications the formation of the correct AC-loop structure, a prerequisite for ANG acti vity, was largely impaired.Ne xt, purified tRNAs from HAP1 cells harbouring a mutation in the TRMT10A gene were used for the production of 3 -tracer tRNAs.This gene encodes a guanosine-specific tRNA methyltr ansfer ase that targets position 9 (m 1 G9) of a subset of nuclear encoded tRNAs ( 77 ), which is potentially involved in suppressing the production of a specific 5 tsRNA in specific cell types ( 78 ).tRNAs from TRMT10A mutant cells were efficientl y hydrol ysed by rANG and could be maintained as structured 3 -tracter tRNAs during nPAGE (Supplementary Figure S6C).Replicate single-turnover time-course activity assays under pre-stead y sta te conditions were used to calculate k unw and 'unwinding' amplitudes for MBP-DDX3X activity on 3 -tracer tRNAs extracted from controls and TRMT10A mutant cells (Supplementary Figure S6D, E and Supplementary Table S4).The results indicated that the absence of m 1 G9 from a subset of cytoplasmic tR-NAs did not modulate the efficiency of MBP-DDX3X to separ ate 3 -tr acer tRNAs in vitr o in a sta tistically significant manner (p-value 0.7698, two-tailed and unpaired t-test).To r emove mor e than one RNA modification from individual 3 -tracer tRN As, tRN As extracted from HeLa cells were treated with recombinantly expressed AlkB (rAlkB), an enzyme with dealkylating activity for specific RNA modifications (i.e.m 1 A, m 1 G, m 2 2 G and m 3 C ( 30)), which can be found in tRNAs that are targeted by ANG ( 79 ).Two rAlkB forms (AlkB and AlkB D135S ) were activity-tested using synthetic RNA oligos carrying m 1 A or m 1 G at their 5 ends.The result confirmed robust demethylation activity of both oligos in a rAlkB concentration-dependent manner (Supplementary Figure S6F).Notably, a fraction of 3tracer tRNAs produced from AlkB-treated tRNAs was unstable during nPAGE indicating the r equir ement of particular RNA modifications for the structural integrity of some tRNAs containing ANG-mediated 'nicks' (Supplementary Figure S6G).Replicate single-turnover time-course activity assays under pre-steady state conditions were performed to test if the r emaining AlkB-tr eated and structurally maintained 3 -tracer tRNAs would be processed differently than controls by MBP-DDX3X.The results showed that AlkBtreated and structurally maintained 3 -tracer tRNAs were separated by MBP-DDX3X with similar unwinding constants and amplitudes as controls, the difference of which was not statistically significant ( P -value 0.3161, two-tailed and unpaired t-test, Supplementary Figure S6H, I and Supplementary Table S4) suggesting that AlkB-sensiti v e RNA modifica tions af fected likel y 3 tracer tRN A structure and, ther efor e, only indir ectly the activity of MBP-DDX3X on 3 tracer tRNAs.

Combined DDX3X and DDX5 depletion does not abrogate activities on 3'-tracer tRNAs in vitro
To address if a loss-of-function of DDX3X might affect tRN A and / or tsRN A le v els and / or integrity, especially during the response to As[III], we used siRNA-mediated knockdown (KD) of DDX3X and / or DDX5 in HeLa cells.Western blotting confirmed the downregulation of both enzymes albeit traces of DDX3X were still detectable after KD (Figure 7 A).Notably, probing for the phosphorylated form of eIF2 ␣ indicated the activation of stress responses after siRNA-mediated KD of either or both RNA helicases suggesting that impairing DDX3 or DDX5 function resulted in cellular stress.NB showed that reduced expression of DDX3X and / or DDX5 had no apparent effect on the stead y-sta te le v els of tRNAs before and after exposure to As[III] as determined by probing for tRNA-Gly GCC / CCC (Supplementary Figure S7A).An apparent increase in 5 and 3 tsRNA le v els after As[III]-induced stress was not specific to either DDX3X and / or DDX5 dose reduction since also control siRNAs caused similar changes in tsRNA signals (Supplementary Figure S7A).These observations suggested that transfection of siRNAs activated cellular pathways resulting in tRNA fragmentation that was independent of eIF2 ␣ as had been observed bef ore f or stress-induced tRNA fragmentation ( 31 , 80 ).Similar observations were made after KD of DDX3X in HeLa or HEK293 cells using virus-mediated expression of shorthairpin RN As (shRN As) (Supplementary Figure S7B, C), which confirmed that downregulation of DDX3X expression did neither af fect ma ture tRN A-Gl y GCC / CCC le v els nor the le v els of tsRNAs deri v ed from these tRNA isoacceptors while virus infection or short hairpin RNA expression increased tsRNA le v els.Furthermore, incubation of cytoplasmic protein extracts (after siRNA KD) with 3 -tracer tR-NAs still re v ealed separation of 3 tsRNAs indicating that either remaining traces of DDX3X, DDX5 or additional RNA helicases (i.e.eIF4A1) remained acti v e in KD e xtracts (Figure 7 B).These combined results indica ted tha t manipulation of RNA helicase le v els by siRNA-mediated KD did not allow deciphering the potential contribution of multifunctional and ubiquitous RNA helicases such as DDX3X or DDX5 to tRNA and tsRNA le v els or integrity, especially during the response to oxidati v e stress.

DDX3X depletion does not abrogate separation but affects the stability of 3'-tracer tRNAs in vitro
To remov e DDX3X acti vity quantitati v ely from cells, we used mouse bone marrow-deri v ed fibrob lasts (BMDF) containing a genetically engineered DDX3X locus, which allows recombina tion-media ted excision of the DDX3X locus from the genome through addition of 4-hydro xytamo xifen (4-OHT) to cultured cells ( 81 ).Deletion of the DDX3X locus r educed DDX3X expr ession to almost undetectable protein le v els by 72 h of continuous 4-OHT treatment (Figure 7 C).NB for tRN A-Gl y GCC / CCC before and after exposure to As[III] confirmed that loss of DDX3X function had no effect on tRNA or tsRNA le v els (Figure 7 D).To address if DDX3X depletion from cells affected the separation and degradation of 3 -tracer tRNAs in vitro , cytoplasmic protein extracts from DDX3X-depleted BMDFs were incubated with 3'-tracer tRNAs followed by nPAGE analysis.The results showed displacement of 3 tsRNAs from 3'-tracer tRNAs in DDX3X-depleted protein extracts, although higher protein concentrations were r equir ed than when using DDX3X-containing protein extracts (Supplementary Figure S7D), which confirmed redundancy through other activities such as DDX5 and / or eIF4A1 but indicated that DDX3X in cytoplasmic protein extr acts play ed a potentially major role in separating 3'tracer tRNAs in vitr o .Notably, 3 tsRNA-degrada tion, seen in particular buffer conditions, appeared to be DDX3Xdependent since increasing amounts of DDX3X-containing extr acts display ed a protein concentr ation-dependent loss of 3 tsRN As, w hile increasing amounts of DDX3Xdepleted cytoplasmic protein extracts maintained equal 3 tsRNA le v els (Figure 7 E).These observations suggested that DDX3X activity on 3 -tracer tRNAs was a pr er equisite for the access of 3 tsRNA-degrading activities in vitro .

DDX3X dose affects 3 tsRNA stability in vivo
To test for a role of DDX3X on the biogenesis and stability of stress-induced tsRNAs in vivo , BMDF (with or without deleted DDX3X locus) were exposed to As[III], followed by cell lysis, SEC fractionation and NB for tRN A-Gl y GCC / CCC .This re v ealed tha t fractiona ted RNPs from BMDF expressing endogenous DDX3X (controls) contained relati v ely fewer 3 tsRN As-Gl y GCC / CCC than 5 tsRN As-Gl y GCC / CCC in, w hich confirmed 5 / 3 tsRN A asymmetry in vivo (Figure 7 F).In contrast, RNPs extracted from DDX3X-depleted BMDF contained almost equal le v els of 5 and 3 tsRNAs-Gly GCC / CCC (Figure 7 F) indicating stabilization of 3 tsR-NAs in the absence of DDX3X.These observations suggested that DDX3X function could be r equir ed for the degradation of 3 -tsRNAs after ANG-mediated tRNA hydrol ysis.Notabl y, transfection of a CMV-promoter containing plasmid harbouring the coding sequence of DDX3X into DDX3X-depleted BMDF resulted in re-establishment of the apparent 5 / 3 tsRNA asymmetry (Supplementary Figure S7E, F and Figure 7 F).These combined results indicated a contributing role for DDX3X to 5 tsRNA biogenesis in vivo , specifically by promoting the instability of 3 tsRNAs after ANG-mediated hydrolysis and separation of 'nicked' tRNA isoacceptors.

DISCUSSION
tsRNAs have attracted incr eased inter est over the last decade, not only because of their proposed connection to various biological processes, but also due to their association with a multitude of patholo gical conditions, especiall y in humans (re vie wed in ( 27 )).Importantl y, the physiolo gical impact of tsRNAs has largely been attributed to individual tsRN A sequences, likel y through their ca pability of interacting with other RNAs and proteins ( 8 , 13 , 21 , 32 , 53 , 82 ).In addition, intermolecular interactions between tsRNAs through G-quadruplexes ( 66 ) or by homo-or heterotypic dimerization of specific 5 tsRNAs have been reported ( 83 ).If individual tsRNAs convey molecular function and physiological impact, then a major unresolved question relates to how one endonuclease-mediated tRN A hydrol ysis e v ent resulting in 'nicked' tRNAs would gi v e rise to individual tsRN As ca pable of interacting with other molecules.This is likely not a spontaneous process governed by thermodynamics and stochasticity but would r equir e cellular activities capable of separating and / or annealing tRNA-deri v ed sequences.Intriguingly, the molecular identity of such activities still remains unknown.Assuming that such activities e xist, we estab lished an in vitro assay capable of detecting enzymatic activities which can bind and process tR-NAs with endonuclease-mediated 'nicks' (tracer tRNAs).This so-called 'tracer tRNA' assay is experimentally versatile since it can be modulated at the le v el of substrate tRNAderi v ed sequences ( i.e. synthetic, endo genousl y modified, isoacceptor-purified, structured RNAs) and at the source of the catalytic activities to be tested ( i.e. recombinant proteins, protein extracts depleted of candidate activities).Using this assay in combination with protein lysates from various mammalian cell lines indicated the presence of enzyma tic activities tha t can separa te 'nicked' tRNAs, thereby producing individual 5 and 3 tsRNAs, as well as of nucleol ytic activities, w hich partiall y degraded 3 tsRN A moieties, particularly from their 3 ends.To identify these activities, we reasoned that they should be in close proximity to tRNAs, 'nicked' tRNAs and, potentially also tsRNAs, particularly during the stress response in vivo .By tracking RNPs after As[III]-induced tRNA fragmentation using biochemical fractionation in human cells, we identified many proteins that co-migrated with both tRNAs and 5 tsR-NAs during the acute stress response.Among these, various proteins stood out for having the potential to contribute to tsRN A bio genesis.Notabl y, we identified RNaseT2, but not ANG, as co-migrating with tsRNAs-containing RNPs during the response to oxidati v e str ess.RNaseT2 is r esponsible for tsRNA production in Sacchar om y ces cer evisiae ( 84 ), Tetrahymena ( 85 ) and Arabidopsis ( 86 ).Importantly, r ecent r eports showed that ANG-deficient cells still produced stress-inducible tsRNAs ( 23 , 24 ), suggesting the existence of redundant nuclease activities acting on tRNAs during the stress response.Howe v er, e v en though human RNaseT2 hydrolysed tRNAs in vitro ( 86 ), its activity on tRNAs in human cells in vivo remains to be experimentally tested.Furthermore, in light of the observed partial in vitr o degrada tion of 3 tracer tRNAs by unknown activities contained in cytoplasmic extracts, it remains to be tested if RNAseT2 could be one of the responsible enzymes me-in RNA helicase activities able to act on 3 -tracer tRNAs.Howe v er, 3 -tracer tRNA assays performed with protein extracts from DDX3X knockout cells re v ealed that DDX3X dose affected both the degree of tsRNA separation and, importantly, the stability of those 3 tsRNAs that became separ ated from 3 -tr acer tRNAs.This result indicated an interaction between DDX3X-mediated 3 -tracer tRNA separation and unknown RNA-degrading activities targeting the 3 moiety of 'tracer tRNAs' for either 3 to 5 (exo-) or endonucleol ytic degradation.Importantl y, this observation was corroborated by probing SEC fractions from cells containing DDX3X, cells with a deleted DDX3X locus and those in which DDX3X expression was reconstituted by plasmid transfection, for tsRNA abundance, which re v ealed that DDX3X dose affected 3 tsRNA le v els in vivo .This is surprising gi v en that probing tRNA integrity and tsRNA le v els in total RNA pools during the response to As[III] by NB had not re v ealed differences in 3 tsRNA le v els between cells with or without DDX3X.One explanation allowing to reconcile these observations is the timing of RNA extraction from the two sample ca tegories.W hile total RNA extraction from DDX3X KD and knockout cells using Trizol (category 1) immediately neutralizes cellular RNA degradation activities, production of protein lysates, their processing by gel filtration followed by RNA extraction from SEC fractions (category 2) is a ra ther elabora te process, which might have supported 3 tsRNA-degrading activities in time.Of note, RNAse-inhibitor cocktail was present in the protein extraction buffers but not in the large volumes of gel filtration and elution buffers, which might have supported re v ealing such acti vities.
Taken together, the presented findings show that tRNAs containing ANG-mediated 'nicks' can be substrates for specific ATP-dependent RNA helicases, resulting in the separation of 5 from 3 tsRNAs in vitro .DDX3X binding to tRNAs in vivo and the effects of DDX3X deletion on 3 tsRNA stability suggest that DDX3X plays a prominent role in contributing to the molecular processing of endonuclease-hydrol ysed tRN As into individual tsRNAs in vivo .The phenomenon of widespread tsRNA abundance in many species, the functional association of defined tsRNAs with specific cellular processes and the potential for specific tsRNAs to serve as biomarkers for human diseases, suggests that identifying the complete molecular machinery involved in producing tsRNAs is of great importance for our understanding as to how these small RNAs exert biological impact.

DA T A A V AILABILITY
Small RNA sequencing data has been deposited at NCBI GEO database under accession number GSE187021.
The mass spectrometry proteomics data have been deposited to the ProteomeXchange Consortium via the PRIDE ( 102 ) partner repository under accession number: PXD029869.The ncRNA-aware CLIP pipeline is available at the Zenodo repository (doi: 10.5281/zenodo.7524185).The DDX3X iCLIP data and matched RNA-Seq was previously published and is available from the Gene Expression Omnibus (GEO) GSE70804 ( 69 ).

SUPPLEMENT ARY DA T A
Supplementary Data are available at NAR Online.

1336Figure 1 .
Figure 1.Cytoplasmic protein extracts contain activities which separate and degrade 3'-tracer tRNAs.( A ) Schematic r epr esentation of enzymatic reactions for the production of 3 -tracer tRNAs. 32P-␥ -ATP-labelling of tRNAs that were hydrolysed by recombinant ANG.Position of the label ( 32 P) on 3' tsRNAs is marked as a green dot.rANG, recombinant Angiogenin; PNK, T4 polynucleotide kinase; cycP, 2 -3 cyclic phosphate; -OH in black, hydroxyl moiety as result of rANG activity; -OH in red, hydroxyl moieties as result of PNK activity.( B ) Representati v e nPAGE of 32 P-␥ -ATP-labelled 3 -tracer tRNAs before and after heat denaturation.Black arrowhead, duplex of 3 -tracer tRNAs; green arrowhead, 3 tsRNAs.( C ) 3 -tracer tRNAs (8 nM final) were incubated with cytoplasmic protein extracts (2 g) obtained from different cell lines in the presence or absence of 2 mM ATP.Reactions were separated by nPAGE and 3 tsRNA signals were collected by exposing PAA gels to phosphor-imaging plates for ≤ 2 h.Black arrow head, 3 -tracer tRN As; green arrow head, 3 tsRN As; grey arrow heads, signals from 3 tsRNA degradation; PE, protein extract.( D ) 'Cold' 3 -tracer tRNAs (8 nM final) were incubated with cytoplasmic protein extracts (2g) obtained from different cell lines in the presence or absence of 2 mM ATP.RNA was extracted from activity assays and separated by on denaturing PAGE followed by NB for tRN A-Gl y GCC / CCC using a probe against the 5 moiety (right panel).To re v eal the tRNA and 5 tsRNA content of cytoplasmic protein extracts, total RNAs extracted from 2 g protein extracts were probed in parallel (left panel).Top panels, NB; bottom panels, SYBR staining; black arrow head, tRN A-Gl y GCC / CCC in protein extracts; white arrowhead, combined signal from tRNAs (protein extracts) and 'cold' 3 -tracer tRNAs; red arrowhead, 5 tsRNAs.( E ) 'Cold' 3 -tracer tRNA assay as described in (D) but probed for the 3 moieties of tRNA-Gly GCC / CCC .Top panels, NB; bottom panels, SYBR staining; black arrow head, tRN A-Gl y GCC / CCC in protein extracts; w hite arrow head, combined signal from tRNAs (protein extracts) and 'cold' 3 -tracer tRNAs; green arrowhead, 3 tsRNAs.

Figure 2 .
Figure 2. Biochemical fractionation identifies proteins co-migrating with 5 tsRNAs after the stress response.( A ) Schematic representation of the workflow for the biochemical fractionation resulting in the enrichment and identification of tsRNA-containing RNPs.As[III], inorganic sodium meta-arsenite; UV, ultraviolet light trea tment; SEC , size exclusion chroma to gra phy; IEX, ion exchange chromato gra phy; RNPs, ribonucleoprotein particles.( B ) Representati v e size exclusion chromatogram for RNPs originating from cell extracts obtained from HEK293 cells after As[III] exposure (black line, UV trace; blue dashed line, injection of sample).RNA from e v ery second fraction of denoted SEC fractions (red marks) was extr acted, separ ated by urea-PAGE and stained with SYBR-Gold; mAU, milli absorbance units; black arrowhead, mature tRNAs; grey arrowhead, small RNAs including tsRNAs.( C ) NB of total RNA extracted from r epr esentative SEC fractions as shown in (B) using a probe against the 5 end of tRN A-Gl y GCC / CCC , tRN A-Glu CUC / UUC and tRN A-Val AAC .Black arrowheads, mature tRNAs; red arrowheads , 5 tsRNAs; asterisks , SEC fractions that were pooled and further subjected to ion-exchange chromato gra phy.( D ) Representati v e NB of total RNA extracted from fractions ( N = 12) that were obtained from subjecting 5 tsRNA-containing SEC fractions to ion-exchange chromato gra phy (IEX) using a probe against tRN A-Gl y GCC / CCC .Upper panel r epr esents a blot obtained after IEX on SEC fractions from cells exposed to sodium meta-arsenite (+ As[III]).Lower panel r epr esents blot obtained after IEX on SEC fractions from control cells.Dashed boxes mark fractions ( 5-8 ) that were analysed for protein content by mass spectrometry.Black arrow heads, mature tRN As; red arrow heads, 5 tsRN As. ( E ) Heat ma p r epr esenting mean peptide intensity values of proteins detected by mass spectr ometry acr oss replicate experiments (R1-3) on control and As[III]-exposed protein extracts.Peptide intensity values (y-axis) were sorted in a descending manner according to their log 2 -transformed enrichment upon As[III] exposure.( F ) Scatter plot of proteins identified in IEX fractions.Mean peptide intensities of identified proteins (y-axis) from r eplicate experiments wer e plotted against their log 2 -transformed enrichment after As[III] exposure.Blue dots depict peptides enriched more than 2-fold (log 2 > 1, r epr esented by vertical gr ey lines) in controls (log 2 < -1) and after As[III] exposure (log 2 > 1).Red dots r epr esent peptides specifically detected upon As[III] exposure.Grey dots represent peptides enriched less than 2-fold in both experimental conditions (-1 < log 2 < 1).

Figure 3 .Figure 4 .
Figure 3. Various RNA helicases re-localize to and co-precipitate in As[III]-induced SG. ( A ) Indirect immunofluorescence images (false-coloured and merged) of HeLa cells analysed for the subcellular localization of RNA helicases (magenta), which wer e r epeatedly identified as SG-associated and that co-migrated with 5 tsRNAs-containing RNPs (this work) and the ubiquitous SG marker, G3BP3 (cyan), during stead y-sta te conditions.( B ) Indirect immunofluorescence images of HeLa cells (false-coloured) analysed for the colocalization of various RNA helicases (magenta) with G3BP1 (cyan) after acute exposure to As[III] (0.3 mM for 1 h).( C ) Western blotting of different candidate helicases on samples obtained during the steps of SG core enrichment according to ( 36 ) and after immunoprecipitation of eIF4A1 from enriched SG cores from HeLa cells were either acutely exposed to As[III] (0.5 mM for 1 h, acute, -r ecovery) or wer e incubated in fr esh medium (+ r ecovery) after r emoval of As[III].( D ) Western blotting of differ ent candidate helicases on samples obtained during the steps of SG core enrichment according to ( 36 ) and after immunoprecipitation of DDX3X from enriched SG cores extracted from Hela cells acutely exposed to As[III] (0.5 mM for 1 h, acute).
B) suggesting a contribution of the helicase domain to DDX3X-tRNA interactions during the response to As[III].Mapping individual cDNA read start positions (indicati v e of re v erse transcription blocks by cross-links) to tRNA sequences showed cross-linking signatures at AC-and T-loop regions suggesting that DDX3X pr efer entiall y interacted with tRN As

Figure 5 .
Figure 5. DDX3X binds to specific tRNAs.( A ) Biotype distribution of DDX3X-deri v ed iCLIP reads that were re-analysed from ( 69 ).Columns represent duplicate iCLIP experiments using human DDX3X (wildtype, wt) or the catalytic DDX3X mutant (R534H) under stead y-sta te conditions and after As[III] exposure.( B ) Changes in DDX3X-mediated iCLIP-derived tRNA reads and in tRNA expression data obtained from duplicate experiments using DDX3X or DDX3X (R534H) during stead y-sta te conditions and after iAs exposure ( 69 ).Normalised abundance of tRNA-derived reads from iCLIP (left panel) or w hole RN A-sequencing experiments (right panel) are shown.( C ) Meta plot depicting positional informa tion of DDX3X-media ted iCLIP signa tures on tRNA sequences obtained from ( 69 ).The 5 nucleotide positions from individual tRNA-derived reads were determined using all human tRNA genes and plotted against their abundance (red).As a comparison, tRNA-deri v ed reads from total RNA-seq data are shown (b lue).Indi vidual lines r epr esent r eplicate experiments.Letters indicate the positions of D-, anticodon-(AC) and T-loops within tRNA sequences.( D ) Plot r epr esenting normalised abundance of DDX3X-deri v ed iCLIP reads obtained from each experiment described in ( 69 ) and after mapping deposited data to tRNA-deri v ed sequences.Indi vidual rows at each tRNA isoacceptor le v el r epr esent the proportion of tRNA-deri v ed reads among all ma pped tRN A sequences in replicate iCLIP experiments.( E ) Plot r epr esenting fold-changes (FC) of DDX3X-deri v ed iCLIP read abundance normalised to the RNA-seq read abundance of the corresponding tRNA isoacceptors in HEK293 cells.Individual rows at each tRNA isoacceptor le v el r epr esent the FC values of corresponding tRNA-deri v ed reads among all mapped tRNA sequences in replicate iCLIP experiments.( F ) Representati v e EMSA after combining increasing molarities of MBP-DDX3X and 5 end-labelled tRNA-Lys UUU / CUU (10 nM final).UV-crosslinked RNPs were separated using nPAGE.Black arrowhead, non-bound tRNAs; red arrow head, DDX3X-tRN A-Lys UUU / CUU complex es.( G ) Repr esentative EMSA after combining increasing molarities of MBP-DDX3X and synthetic 5 end-labelled tRNA-Lys UUU sequences (30 nM final).UV-crosslinked RNPs were separated using nPAGE.Black arrowhead, non-bound tRNAs; red arrow head, DDX3X-tRN A-Lys UUU complexes.( H ) Quantification of independent EMSA experiments ( n = 3) for the calculation of the equilibrium dissociation constant ( K d ) between MBP-DDX3X and endo genousl y purified and 5 end-labelled tRNA-Lys UUU / CUU as well as a synthetic 5 end-labelled tRNA-Lys UUU sequence.Line marks the fit of the mean values (percent of shifted tRNA-Lys signal) to the hyperbolic binding isotherm used to calculate the Kd value while error bars r epr esent standard deviations.

Figure 6 .
Figure 6.DDX3X separ ates 3 -tr acer tRNAs with specificity for particular tRNA isoacceptors ( A ) Representati v e RNA helicase acti vity assay (fixed time-point) using equal molarities of recombinant MBP-DDX3X or MBP-DDX3X DQAD (750 nM) and 3 -tracer tRNAs (60 nM final) in the presence of equimolar ATP / MgCl 2 or AMP-PNP / MgCl 2 (2 mM).Black arrowhead, 3 -tracer tRNAs; gr een arrowhead, 3 tsRNAs.( B ) Repr esentati v e RNA helicase activity assay using MBP-DDX3X (750 nM) and 3 -tracer tRNAs (20 nM final) in the presence of equimolar ATP / MgCl 2 (2 mM).Aliquots wer e r emoved from reactions at indicated time points and separated using nPAGE.Arrowheads, as described in (A).( C ) Quantification of triplicate time-course RNA helicase activity assays (see Supplemental Figure S5B) using MBP-DDX3X (750 nM) and 3 -tracer tRNAs (10 nM final) in the presence of equimolar ATP / MgCl 2 (2 mM) to deri v e K unw .Line marks the fit of the mean values to the integrated first-order rate equation, while error bars r epr esent standard devia tions.( D ) Representa tive fixed-time point RNA helicase assay using MBP-DDX3X (750 nM) on 3 -tracer tRNAs (60 nM final) in the presence of equimolar ATP / MgCl 2 (2 mM).3 -tracer tRNAs that were heat-denatured were loaded in parallel.RNAs migrating at the le v el of heat-denatured and MBP-DDX3X-displaced 3 tsRNAs (dashed and red squares, respecti v ely) were e xcised in technical duplicates from nPAGE, cloned and subjected to small RNA sequencing.Black arrowhead, 3 -tracer tRNAs; green arr owhead, 3 tsRNAs; white arr owhead, position where also non-labelled 5 tsRNAs will migra te. ( E ) Hea t maps r epr esenting r elati v e read abundance (normalised lo g 2 values) of known ANG tRN A substra tes.The da ta was obtained by mapping replica te sRNA-seq da ta (columns) origina ting from hea t-dena tured 3 -tracer tRNAs (left panel) and MPB-DDX3X-displaced 3 -tsRNAs (right panel) to colla psed tRN A clusters as described in( 52 ).P anels ar e ranked by r elati v e abundance of the respecti v e tRNA cluster in a descending manner.( F ) Heat maps r epr esenting fold-changes of the r elati v e read abundance (per tRN A cluster) after ma pping reads originating from MPB-DDX3X-displaced tsRN As divided by read abundance from hea t-dena tured 3 -tracer tRNAs (as described in (E)).Combined replicates (columns) ranked by relati v e abundance of the respecti v e tRNA cluster in a descending manner.Asteriks denote tRNA isoacceptors with positi v e fold change ra tios, indica ting pr efer ential activity of MBP-DDX3X on these tRN As, w hich is independent of the corresponding tracer tRNA abundance in the 3 -tracer tRNA pool.( G ) Representati v e time-course RNA helicase activity assay using MBP-DDX3X (750 nM) and 3 -tracer tRN A-Gl y GCC / CCC (10 nM final) in the presence of equimolar ATP / MgCl 2 (2 mM).Aliquots were removed from reactions at indicated time points and separated using nPAGE.Arrowheads, as described in (A).

Figure 7 .
Figure 7. DDX3X affects the stability of 3 -tracer tRNAs in vitro and in vivo.( A ) Western blotting for DDX3X, DDX5 and eIF2 ␣ (serine-51 phospho) before and 48 h after transfection with siRNA pools (non-targeting as controls versus targeting each RNA helicase alone or in combination).Antibodies against vinculin were used as a loading control.Lane marked with (*) points at background eIF2 ␣-P signal, while lanes marked with (**) indicate siRNAmediated induction of eIF2 ␣-P signal in RNA helicase knockdowns.( B ) 3 -tracer tRNAs (8 nM final) were incubated with cytoplasmic protein extracts (2 or 4 g) obtained from HeLa cells transfected with siRNAs targeting DDX3X or DDX3X in combination with DDX5.Reactions were separated by nPAGE and 3 tsRNA signals were collected as described above.Black arrowhead, 3 -tracer tRNAs; green arrowhead, 3 tsRNAs.( C ) Western blotting for DDX3X in total protein extract obtained from BMDF after 72 h of 4-hydro xytamo xifen (4-OHT)-mediated deletion of the DDX3X gene locus.␤actin was used as a loading control.( D ) NB using sequential probing of the same membrane against the 5 and 3 portions of tRN A-Gl y GCC / CCC on total RNA purified from BMDF after 72 h of 4-OHT-mediated deletion of the DDX3X gene locus before and after As[III] exposure (0.5 mM for 1 h).Black arrowheads, mature tRNAs; red arrowhead, 5 tsRNAs; green arrowhead, 3 tsRNAs.( E ) 3 -tracer tRNAs (8 nM final) were incubated with cytoplasmic protein extracts (5 or 10 g) obtained from BMDF before or after 4-OHT-mediated deletion of the DDX3X.Reactions were separated by nPAGE and 3 tsRNA signals were collected as described above.Black arrow head, 3 -tracer tRN As; green arrow head, 3 tsRN As; grey arrow head, signals from partial 3 tsRNA degradation.( F ) NB of total RNA purified from SEC fractions originating from As[III]-exposed BMDF before (upper panels), after 4-hydro xytamo xifen (4-OHT) mediated deletion of DDX3X (middle panels), and in the presence of ectopic expression of wild type DDX3X (pCMV-DDX3X, lower panels) using sequential probing of the same membrane.Probes for the 5 (left panels) and 3 (right panels) moieties of tRN A-Gl y GCC / CCC were used.Black arrowheads, mature tRNAs; red arrowhead, 5 tsRNAs; green arrowhead, 3 tsRNAs.