BLM helicase protein negatively regulates stress granule formation through unwinding RNA G-quadruplex structures

Abstract Bloom's syndrome (BLM) protein is a known nuclear helicase that is able to unwind DNA secondary structures such as G-quadruplexes (G4s). However, its role in the regulation of cytoplasmic processes that involve RNA G-quadruplexes (rG4s) has not been previously studied. Here, we demonstrate that BLM is recruited to stress granules (SGs), which are cytoplasmic biomolecular condensates composed of RNAs and RNA-binding proteins. BLM is enriched in SGs upon different stress conditions and in an rG4-dependent manner. Also, we show that BLM unwinds rG4s and acts as a negative regulator of SG formation. Altogether, our data expand the cellular activity of BLM and shed light on the function that helicases play in the dynamics of biomolecular condensates.


INTRODUCTION
DN A or RN A G-quadruplexes (dG4s or rG4s) are nucleic acid secondary structures that are formed in guanine (G)rich sequences (1)(2)(3).Thousands of such G4-forming sequences have been identified in the human genome (DNA) and transcriptome (RNA), on the basis of sequence predictions using bioinformatic tools and high-throughput sequencing-based methods ( 2 , 4-10 ).dG4s are highly polymorphic structures (par allel, anti-par allel, and hybrid G4s) that are involved in the control of replication, genome stability, and transcription ( 3 ).In contrast, rG4s primarily form parallel structures ( 1 , 11 ), and play regulatory roles in RN A splicing, mRN A stability, RN A tr ansport, tr anslation and stress response ( 1 , 6 , 11 , 12 ).
Complex biological processes involving G4s rely on the involvement of specific G4-binding proteins.The cellular localization of these proteins, their target specificity, binding affinity, and enzymatic activity contribute to defining their function.In this context, G4 helicases unwind dG4 and / or rG4 structures ( 13 , 14 ).Since stable dG4s could impair key biological processes, such as replication, transcription, translation and repair, these structural blocks must be unwound during these DNA transactions ( 11 ).
The connection between SGs and rG4s is further substantiated by the fact that se v eral rG4 binding or unwinding proteins have been found in SGs ( 6 , 15 , 25 , 26 ).Transfection of exogenous RNA with a pr efer ence to form rG4 promotes SG formation ( 15 ) and the rG4-helicase DHX36 remarkably affects SG formation ( 17 ).In addition, the SG core protein, RAS GTPase-activating binding protein 1 (G3BP1), is an rG4-binding protein and its interactions with rG4s have been suggested to contribute to SG formation ( 6 , 25 , 27 ).Together, these data imply that rG4s are key regulatory factors in SG biology.
Bloom's syndrome helicase (BLM) is one of the first human helicases reported to resolve dG4, requiring a 3 singlestranded overhang for enzyme loading ( 28 , 29 ).The gene encoding the BLM protein is mutated in patients with Bloom's syndrome, displaying high genomic instability and predisposition to cancers ( 30 , 31 ).In addition, we have previously reported that BLM resides in SGs, under sodium arsenate stress ( 32 ).Howe v er, BLM's regulatory r ole in the contr ol of rG4s and in the biology of SGs has not yet been studied.
Here, we demonstrate that BLM is enriched in SGs under se v eral stress conditions and also binds rG4s, thus expanding its reported functionality beyond dG4s.Moreover, we show that BLM is recruited to SGs in an rG4-dependent manner and negati v ely regula tes their forma tion.These observations provide new insights into the cellular regulation of the str ess r esponse and more broadly into the functions of G4 helicases in biomolecular condensates.

RN A / DN A oligos
Chemicall y synthesized DN A or RN A oligonucleotides (6FAM / Da bcyl-la belled or unla belled; Supplementary data 1, Datasheet S1) were from Sigma-Aldrich / MERCK or from Integrated DN A Technolo gies (IDT).We dissolved the oligonucleotides in RNase-free TEx1 buffer (10 mM Tris-HCl pH 7.5 and 1 mM EDTA) for the stock concentration of 100 uM and stored them at −80 • C in aliquots to avoid thaw-freeze cycles.

RN A / DN A G4 pr epar ation
We diluted the FAM-labelled oligonucleotides to desired concentration in a TE 1 × buffer with or without 150 mM DEPC-treated KCl or 150 mM DEPC-treated LiCl.Then, using a PCR machine RNAs folded to create secondary structures by heating to 90 • C for 5 min and then lowering the temperature to 25

Circular dichroism (CD) spectroscopy
We performed CD experiments at 25 • C using Chirascan ™plus ACD spectropolarimeter with a quartz cuvette with a 1 mm path length.We collected CD spectra from 360 to 210 nm.The bandwidth was 1 nm, and the response time was 1 s.CD spectra signal corrected to background (buffer only) and r epr esented the average of 3 runs.

Cloning, expression and purification of recombinant core BLM protein
A truncated BLM 636-1298 (cBLM, spanning the helicase, RQC and HRDC domains) was expressed in E. coli and purified as described previously ( 33 , 34 ) with the addition of a MonoS ion-exchange-and gel-filtration step.

Helicase activity in vitro assay
A mixture of 1 uM Da bcyl-la belled oligonucleotide (Dabcyl-S-rG4-VEGFA-U15) and 0.85 uM 6FAM-labelled oligonucleotide (F-rShort-6FAM) was pr epar ed in 20 mM Tris-HCl buffer (pH 7.5) containing 5 mM MgCl2, 1 mM KCl, 99 mM NaCl and RNase-free UPW.The 1.2fold excess of the da bcyl-la belled strand ensured complete hybridization of the 6FAM-labelled strand and maximal quenching of the fluorescent signal.The mixture was annealed by PCR program as follows: 90        C to end.The annealed samples were then stored at -80 • C in aliquots.A competitor oligon ucleotide (Trap oligo , 1 uM) was hea ted a t 90 • C for 5 min and then cooled on ice for 15 min before the reaction.Oligonucleotides described in Supplementary Data 1, Datasheet S1).Helicase reactions were performed based on Mendoza et al .( 35 ), with adaptations.The reactions were conducted in triplicates in 96-well plates (Nunc MicroWell 96-Well, black, Fla t-Bottom Micropla te, Thermo Fisher; Ca t.# 137103) a t 37 • C with the lid.Fluorescence was monitored using a microplate reader (Tecan Infinite 200 PRO).Each replicate consisted of a 50 ul solution containing 40 nM 6FAM-Dabcyl system (pre-annealed), varying amounts of cBLM protein, and 200 nM Trap oligonucleotide (unlabelled and complementary to the FAM-labelled strand).Subsequently, 5 ul of a 50 mM A TP / A TPgS solution (5 mM final concentration / reaction) was added to each well.The 96-well plate was stirred for 10 s, and fluorescence emission was recorded every 15 s at excitation / emission wavelengths of 492 nm / 525 nm, respecti v ely.

APEX proximity labelling and pull-down
16 M Tetracycline-induced G3BP1-APEX or NES-APEX expressing U2OS cells were seeded per flask, in 180T flasks (90-100% confluency) with a medium which was supplemented with tetracycline for 24 h (50 ng / ml for each of the APEX baits) for inducing the APEX-bait gene expression.On the day of the experiment, the cells were incubated for 3 h with 1 uM QUMA-1, or an equivalent volume of DMSO prior to 150 uM Sodium arsenate (Sigma-Aldrich, 71287) stress for 2.5 h.Labelling activity was induced by supplementing Biotin-phenol (BP, 500 uM, Iris Biotech GmbH, LS-3500) for the last 30 min of the stress and then H 2 O 2 30% (1 mM, J.T.Baker 7722-84-1) was added for 1 min.APEX activity extinguished with quenching solution (sodium azide (10 mM, Mallinckrodt, 1953-57), sodium ascorbate (10 mM, Sigma-Aldrich, A7631) and Trolox (5 mM, Sigma-Aldrich, 238813) in PBS 1 × twice for 1 min each time.Then, the cells were washed two more times with PBS 1 ×, 1 min each time, and for the last wash with quencher solution as mentioned before.After the fifth wash, the cells were scraped in quencher solution, centrifuged at 800 × g for 10 min at 4 • C, pelleted and lysed in ice-cold RIPA lysis buffer supplemented with cOmplete Protease Inhibitor Cocktail (Roche, 4693116001) and PhosSTOP (Roche, 4906837001).Lysates were centrifuged at 14 000 × g for 10 min at 4 • C. Protein concentration was quantified with Bio-Rad Protein Assay Dye Reagent ('Bradford'; Bio-Rad, 500-0006).Streptavidin-coated magnetic beads (Pierce Streptavidin Magnetic Beads, Thermo-Fisher, 88816) were incubated for pulldown experiments with the ratio between extract: beads as 1 mg:200 ul, respecti v ely (for proteomics we used 500 ug of the extract with 100 ul beads per sample, in a complete volume of 500 ul) with rotation overnight at 4 • C.Then, the beads were washed twice with RIPA containing quencher solution, once with 1 M KCl solution, once with 0.1 M Na 2 CO 3 solution (po w der, Sigma-Aldrich, S7795), once with 2 M Urea solution (2 M urea; po w der, Sigma-Aldrich, U0631, and 10 mM Tris-HCl pH 8.0) and for proteomics, each wash was done for 3 min and the biotinylated proteins were transferred to on-bead digestion process by trypsinization.

Liquid chromatography and mass spectrometry
ULC / MS grade solvents were used for all chromatographic steps.Dry-digested samples were dissolved in 97:3% H 2 O / acetonitrile + 0.1% formic acid.Each sample was loaded and analyzed using split-less nano-Ultra Performance Liquid Chromato gra phy (10 kpsi nanoAcquity; Waters, Milford, MA, USA).The mobile phase was: A) H2O + 0.1% formic acid and B) acetonitrile + 0.1% formic acid.Desalting of the samples was performed online using a Symmetry C18 re v ersed-phase trapping column (180 um internal diameter, 20 mm length, 5 um particle size; Waters).The peptides were then separated using a T3 HSS nanocolumn (75 um internal diameter, 250 mm length, 1.8 um particle size; Wa ters) a t 0.35 ul / min.Peptides were eluted from the column into the mass spectrometer using the following gradient: 4 to 25%B in 155 min, 25 to 90%B in 5 min, maintained at 90% for 5 min and then back to initial conditions.
For MS, the nanoUPLC was coupled online through a nanoESI emitter (10 um tip; New Objective; Woburn, MA, USA) to a quadrupole orbitrap mass spectrometer (Q Exacti v e Plus, Thermo Scientific) using a FlexIon nanospray apparatus (Proxeon).
Data was acquired in data-dependent acquisition (DDA) mode, using a Top10 method.MS1 resolution was set to 70 000 (at 200 m / z ), a mass range of 375-1650 m / z , AGC of 3e6, and maximum injection time was set to 60 ms.MS2 resolution was set to 17500, quadrupole isolation 1.7 m / z , AGC of 1e5, dynamic exclusion of 45 s, and maximum injection time of 60 msec.

Raw proteomic data processing
The processing step was performed as described previously ( 32 ).In short, we processed the raw MS data by using MaxQuant version 1.6.2.6 ( 38 ), and a database search was done with the Andromeda search engine ( 39 , 40 ) by using the human Uniprot database.The data wer e filter ed with a threshold of 1% the false discovery rate (FDR) for both the peptide-spectrum matches and the protein le v els.The label-free quantification (LFQ) algorithm in MaxQuant ( 41 ) was utilized to compare experimental samples, except for the negati v e controls.Additional settings included variable modifications.The 'match between runs' option was enabled to transfer identification between separate LC-MS / MS runs based on their accurate mass and retention time after retention time alignment.

Proteomic statistical analysis
Pr oteinGr oups output table was imported from MaxQuant to Perseus environment version 1.6.2.3 ( 42 ) and analyzed with Perseus and then with R version 4.0.5 ( 43 ).We excluded re v erse pr oteins, pr oteins identified only based on a modified peptide, and contaminants as quality control steps.Non-specific streptavidin-bead binding proteins were excluded by the following protocol: Intensity values were log 2 -transformed, and protein groups were filtered to retain only proteins with at least 2 valid values / group.Missing values were replaced by a constant low value ( 15 ).Student's t -test with S0 = 0.1 was performed with FDR P -value ≤0.05 and Fold Change (FC) > 0, for pairs of APEX-On and corresponding APEX-Off samples for each group of condition / treatment (DMSO / NES-APEX; DMSO / G3BP1-APEX; QUMA-1 / NES-APEX; QUMA-1 / G3BP1-APEX).Proteins that passed all QC filters were separated for each condition (DMSO or QUMA-1).Within each condition, the LFQ intensities of G3BP1-APEX samples wer e compar ed to the LFQ intensities of NES-APEX samples to characterize the stress-granule associated proteins under each condition, as follows: Data were filtered to retain only proteins with at least two LFQ values in at least 1 gr oup.Importantly, thr ough the analysis, one repeat from the G3BP1-APEX samples under DMSO treatment (sample 1) was excluded because of suboptimal correlation with the other samples from this group.Missing data were imputed by creating an artificial normal distribution with a downshift of 1.5 standard deviations and a width of 0.4 of the original ratio distribution.Student's ttest called enriched SG proteins with S0 = 0.1 and FDR P -value ≤0.05 and a minimum of two-fold enrichment of proteins in G3BP1-APEX samples versus NES-APEX (log 2 (SG-APEX -NES-APEX) > 1).Following that, after filtered valid values of at least 20% in total, we imputed the LFQ intensities by creating an artificial normal distribution with a downshift of 1.8 standard deviations and a width of 0.3, and the G3BP1-APEX values per each condition (DMSO or QUMA-1) were normalized by the mean of their corresponding NES-APEX values.Then, we categorized the normalized G3BP1-APEX for two groups: DMSO and QUMA-1.We compared these two conditions by student's t-test with FDR correction by using R as mentioned above.Enrichment of proteins in SGs for DMSO or QUMA-1 treatment was determined by FDR p-value ≤ 0.05 and a minimum of two-fold enrichment (for DMSO: lo g 2 (Q UMA-1 -DMSO) < -1; for QUMA-1: lo g 2 (Q UMA-1 -DMSO) > 1).Principal component analysis (PCA), Volcano plot and Heatmap for comparison between DMSO and QUMA-1 SG proteomes were generated by Perseus and R.

Mass spectrometry cBLM identification
Shifted G4 bands were cut from the gel and subjected to ingel tryptic digestion followed by a desalting step.The resulting peptides were subjected to nanoflow liquid chromatography (nanoAcquity) coupled to high resolution, high mass accuracy mass spectrometry (Q Exacti v e HF, discov ery mode).Raw data was processed using Proteome Discoverer version 2.4, and searched with SequestHT ( 44 ) and MS Amanda ( 45 ) against a protein r efer ence list containing the recombinant cBLM sequence that we provided, the E. coli K12 protein database was downloaded from Uniprot.org and an in-house list of 128 common lab contaminants.

Cell lysis and western blotting
For WB analysis of biotinylated proteins, the treatments and stress conditions as well as the APEX proximity labelling protocol were performed as described above.For BLM signal from pull-down of biotinylated proteins by APEX proximity labelling under DMSO or QUMA-1, we seeded 8-9 M tetracy cline-inducib le NES-APEX or G3BP1-APEX expressing U2OS cells with tetracycline in 80T flasks a day before the experiment.The wash steps were done except for the last steps, as follows: washes were done without prolonged incubation, and in addition to the wash steps above, the beads were washed twice more with RIPA b uffer a gain and the biotinylated proteins were eluted from the beads by boiling (95 • C) with 5 × sample buffer supplemented with 2 mM free biotin for 10 min.Based on protein quantification with Protein Assay Dye Reagent, we loaded the beads 360 ug with 40 ul beads per sample, before the washing steps.The supernatant of each solution was taken for loading on the gel after the beads were magnetized.
For general pattern of biotinylated proteins by APEX proximity labelling under DMSO or QUMA-1, we seeded 500 K from the cell lines above in 6-well plates a day before the e xperiment.Howe v er, w hole cell l ysate was obtained by RIPA lysis buffer without pull-down and wash steps with Streptavidin beads.Based on protein quantification as mentioned, the protein was loaded 50 ug of total protein per well.
For WB analysis of BLM signal under siRNA treatment, we seeded 100 K G3BP1-GFP U2OS cells in 6-well plates and were transfected with siRNA control or siRNA against BLM in triplicates, as described above (see 'siRNAs' section).Then, we lysed the cells with RIPA lysis buffer, and based on protein quantification, the protein was loaded with 50 ug of total protein per well.
After lysis and preparation steps of the samples, for all the WB experiments, we resolved the proteins by 10% SDS-PAGE at 100 V for 10 min and then 120 V up to 80 min.After gel electr ophoresis, pr oteins wer e transferr ed to nitrocellulose membranes (W ha tman; 10401383) a t 250 mA for 70 min.Membranes were blocked for 1 h at R.T. with 3% bovine albumin fraction V (MPBio; 160069) in PBS containing 0.05% Tween-20 (PBST) and for each experiment above, the procedure was done differently.
For WB analysis of BLM signal under siRNA treatment or in SGs after APEX proximity labelling and pull-down, we incubated the membranes with primary rabbit polyclonal antibody anti-BLM (1:500; a bcam, a b476) overnight at 4 • C with rocking in antibody solution (5% albumin, 0.02% sodium azide and fiv e drops of phenol red in 0.05% PBST).Specifically for WB analysis of BLM signal under siRNA treatment, we used also incubated the membranes with primary monoclonal mouse antibody anti-Tubulin (1:2000; Sigma-Aldrich, T9026) as a control overnight at 4 • C with rocking in antibody solution.Following primary antibodies incubation, membranes were washed three times for 5 min at R.T. with 0.05% PBST and were incubated for 1 hr at R.T. with horseradish peroxidase-conjugated speciesspecific secondary antibody.Specifically for WB analysis of biotinylated proteins, Streptavidin-HRP (1:1000; Sigma-Aldrich, Cat#RABHRP3) was used for 1 h at R.T in the dark.
For all the experiments we then washed the membranes three times for 5 min each in 0.05% PBST at R.T. and visualized them using EZ-ECL Chemiluminescence (Biological Industries, 20500-120) by ImageQuant LAS 4000 (GE Healthcare Life Sciences).Densitometric analysis was performed using Fiji softwar e (NIH) and r epr esentati v e bands ar e pr esented.

Staining and microscopy
50 K G3BP1-GFP expressing U2OS cells, BJ fibroblasts or iPSC-deri v ed neurons seeded per well in 24-well plates on coverslips 24 hr prior to stress.After the stress induction, we fixed the cells with 4% PFA (Alfa Aesar, 43368) for 15 min at R.T. and washed them with RNase-free PBS 1X three times.Then, we treated the cells with 0.1% Triton-X for 15 min at R.T., blocked with CAS-Block reagent (ThermoFisher Scientific; 008120) for 10min at R.T., incubated with primary rabbit polyclonal anti-BLM antibody (1:100).Primary mouse monoclonal anti-G3BP1 antibody (1:200; Santa-cruz, sc-365338) incubated at cold room o.n.A day after, we washed the cells with RNasefr ee PBS thr ee times, 5 min each, and then incubated with secondary Cy5-conjugated anti-Rabbit antibody (1:200) or also Cy2-conjugated anti-Mouse antibody (1:200, for non U2OS cells) for 1 h at R.T. Plates kept in the dark, washed with RNase-free PBS three times, 5 min each, dried and mounted with DAPI (Fluoroshield with DAPI; Sigma-Aldrich; F6057).SG induction, performed with NaAsO2 (400 uM for 30 min, Sigma-Aldrich, 71287) or Thapsigargin (1 uM for 1 h, Sigma-Aldrich, T9033), Puromycin (200 ug / ml for 4 h, Invivogen, ANT-PR) or by heat shock for 90 min a t 43 • C .A similar procedure was done for BLM staining in SGs as a result of DMSO versus QUMA-1 treatments (U2OS cells).Specifically for this experiment, the cells were incubated with DMSO or QUMA-1 (1 uM) for 3 hr prior to sodium arsenate stress (150 uM, 2.5 h).
For QUMA-1 staining in fixed cells, we seeded 12 K G3BP1-GFP expressing U2OS cells per w ell (in 96-w ell plates) 24 h prior to transfection, incubated the cells with 1 uM siRNAs (final concentration) against Dhx36, Blm, or siControl for 4-16 hr.48 hr later, we fixed the cells with 4% PFA for 15 min and washed them with RNase-free PBS three times.Then, we incubated the cells with 2 uM QUMA-1 and Hoechst 33342 (1:8000; Sigma-Aldrich, B2261) for 10 min at 37 • C. We kept the plate in the dark from this point.Next, we washed the cells with RNase-free PBS three times, 5 min each.We acquired the fixed cells without (BLM stainings) or with (QUMA-1 staining) taking z-stacks via a Zeiss LSM900 laser scanning confocal microscopy system equipped with a Zeiss Axiovert microscope and using a 63 × 1.4 NA oil immersion lens.Similar steps after fixation as well as image processing and analysis were done also for RPE cells (wt versus BLM KO).
For APEX proximity labelling validation in fixed cells, 50 K tetracycline-induced G3BP1-APEX or NES-APEX expr essing U2OS cells wer e seeded 24 hr prior to stress.We incubated the cells with or without 400 uM sodium arsenate stress supplemented with or without 500 uM biotinphenol for 30 min, and then the APEX proximity labelling was induced by the presence of H 2 O 2 for 1 min.Next, the media was removed and the cells were washed three times with quencher solution (as mentioned above) and then fixed with 4% PFA for 15 min.After we washed them with PBS three times, the cells were treated with 0.1% Triton-X for 15 min at R.T., blocked with CAS-Block reagent for 10 min at R.T., and incubated with primary monoclonal anti-V5 tag (1:1000; ThermoFisher, R960-25), which r epr esents the AEPX-bait proteins and goat polyclonal anti-TIA1 (1:50; Santa cruz, sc-1751) in a cold room o.n.A day after, we washed the cells with PBS three times, 5 min each, and then incubated them for 1 h at R.T. in the dark with secondary Cy5-conjugated antigoa t antibod y, Cy2-conjuga ted anti-mouse antibod y and NeutrAvidin-TexasRed conjugate (ThermoFisher, A2665) to stain the biotinylated proteins (for all 1:200).We washed the cells with RNase-free PBS three times, 5 min each, dried them and mounted them on slides with DAPI.We acquired the fixed cells via a Zeiss LSM800 laser scanning confocal microscopy system equipped with a Zeiss Axiovert microscope, and using a 63 × 1.4 NA oil immersion lens.

Analysis of signal enrichment within stress granules
For analysis of the BLM enrichment signal, e v ery particular enrichment value was analyzed per single SG and was determined as the signal-to-background ratio of the BLM intensity (cy5; purple) in the SG (G3BP1-GFP; green) compared to the fixed surrounded cytoplasmic area of the same SG in the U2OS cells.The analysis was done with Fiji software.
QUMA-1 enrichment signal, analyzed per single SG as the signal-to-background ratio of the rG4 (QUMA-1; red) intensity in the volume of the SG (G3BP1-GFP (U2OS cells) or cy2, anti-G3BP1 (RPE cells); gr een) compar ed to the fixed surrounded cytoplasmic volume of the same SG.The analysis was done with the Arivis software.

Molecular cloning of mCherry-helicase o ver expr ession
We cloned DHX36 isoform 1 CDS or BLM cDNA into mCherry-containing pUltraHot vector by using a r estriction-fr ee (RF) procedur e with Q5 Hot start High-Fidelity DN A pol ymerase (NEB).The source of BLM cDNA was from a pTRIP-CMV-puro-2A-BLM plasmid (Addgene, plasmid #127641).The original plasmid for DHX36 CDS was a kind gift from Dr Daniel Benhalevy, Prof. Marcus Hafner and Prof. Katrin Paeschke.

ImageStream analysis
Le v els of QUMA-1 signal under knockdown of DHX36 and BLM.800 K G3BP1-GFP expressing U2OS cells were seeded.A day after, the cells were transfected with siRNAs against DHX36, BLM, or siControl as described above (see 'siRNAs' section).Next, we incubated the cells with or without 150 uM sodium arsenate stress for 2.5 h.We fixed the cells with 4% PFA for 15 min at R.T. and washed them three times with PBS 1 ×, 5 min each time.Then, the cells were incubated with 0.5 uM QUMA-1 and Hoechst (1:8000) for 15 min at 37 • C and were washed three times with PBS 1 ×, 5 min each.The cells were scraped and collected with PBS supplemented with 1% BSA.The cells were centrifuged gently (300 × g, 10 min at 4 • C) and were suspended by quick vortex in the volume of 20-50 ul of PBS 1 × with 1% BSA.Cells were imaged by an Imaging Flow Cytometer (ImageStreamX Mark II, AMNIS corp., Luminex, TX, USA).Data were acquired using a 40 × lens, and the lasers used were 405 nm (20 mW), 488 nm (100 mW), 561 nm (20 mW) and 785 nm (1 mW).Data were analyzed using the manufacturer's image analysis software IDEAS 6.3 (AMNIS corp.).Images were compensated for spectr al over lap using single-stained controls.Cells were selected for Hoechst positi v e cells by plotting the area of the DNA staining (AREA M07, in square microns) vs .the intensity of the DNA staining (Intensity MC Hoechst, arbitrary units).Cells positi v e for GFP e xpr ession wer e selected by plotting the intensity vs. Max Pixel (the value of the highintensity pixel) of the GFP channel (ch02).To eliminate outof-focus cells, cells were further gated using the Gradient RMS and contrast features (measures the sharpness quality of an image by detecting large changes of pixel values in the image).Single cells were selected by plotting again the area of Hoechst staining, versus the aspect ratio normalized for the intensity of the Hoechst staining (Aspect Ratio Intensity M07 Hoechst).Flat cells were further selected according to the intensity vs. Max Pixel values of the Hoechst staining.The normalized le v els of QUMA-1 were calculated by dividing the total intensity (Intensity MC QUMA) by the cell area of the bright-field image (Area M01).
For RPE cells (wt versus BLM KO), we performed the same fixation procedure as above, but we incubated the cells with 0.5 uM QUMA-1 3 hr prior to the fixation without any other treatment before fixation.After the fixation, we incubated the cells with Hoechst (1:8000) for 15 min at 37 • C and were washed three times with PBS 1X, 5 min each.Im-ageStream analysis and the settings were the same as above.
Quantification of SGs in mCherry-positi v e cells.400 K G3BP1-GFP expressing U2OS cells were seeded in 6-well plates.To generate mCherry ov ere xpressing cells, 24 h later, the cells were tr ansiently tr ansfected with pUltr aHot-mCherry, pUltr aHot-mCherry-DHX36 or pUltraHot-mCherry-BLM expression plasmids, as described above.Stress induction, fixation, wash, and cell collection steps were performed as mentioned for the first experiment above (without QUMA staining but with Hoechst staining (1:800) to dye the nuclei).Quantification of SGs was taken into account only in mCherry ov ere xpressing cells.The laser settings were the same as above, and mCherry was collected in channel 4. Cells positi v e for mCherry were selected by plotting the intensity versus Max Pixel of the mCherry channel.To identify cells with SGs, two truth populations were selected, and a classifier was created using the machine learning module in IDEAS 6.3.The percentage of cells with SGs within mCherry positi v e cells was quantified by plotting the granule classifier vs. the Max Pixel values of the GFP channel.
For experiments of SG formation under knock-down conditions, we transfected the G3BP1-GFP expressing U2OS cells with 1 uM siRNAs (final concentration) against Dhx36, Blm, or siControl, and incubated them for 4 h to on and then replaced the transfection medium to complete medium for 48 h.
For experiments of SG formation under over expr ession conditions, we transfected the G3BP1-GFP express-ing U2OS cells with pUltraHot-mCherry or pUltraHot-mCherry-DHX36 or with pUltraHot-mCherry-BLM by using jetOPTIMUS DNA transfection reagent (Polyplus; 101000051) and incubated them for 4 h and then replaced the transfection medium to complete medium.A day later, we induced ov ere xpression with tetracycline.
48 hr after the transfection (both siRNAs or ov ere xpression plasmids), we replaced the medium with a 150 uM NaAsO2-added medium and immediately took them to the microscope to monitor SG formation.We took SG li v e imaging by a PCO-Edge sCMOS camera controlled by V isV iew installed on a VisiScope Confocal Cell Explorer system (Yokogawa spinning disk scanning unit; CSU-W1) and an inverted Olympus microscope (60 × oil objective; excitation wavelength: GFP -488 nm).We analyzed SG and cell areas using surface features in Imaris software 9.5.1.

Real-time polymerase chain reaction (rt-PCR)
To validate the efficiency and the function of the siRNAs, we performed rtPCR on cDNA from G3BP1-GFP expressing U2OS cells after transfection with the siRNAs.Whole cell RNA extract was isolated from the cells by TRI reagent (Sigma-Aldrich, T9424) and RNA isolation kit (Direct-zol RNA miniprep; Zymo, R2051).Next, cDNA was generated from the extracted RNA by qScript cDNA synthesis kit (Quantabio, 95047).We performed a real-time PCR procedure using KAPA SYBR Fast qPCR kit Master Mix (2 ×) Prism ABI (Kapabiosystems, KK4604) and measured the amplification cycles per tested gene and control gene (housekeeping gene, Gapdh) compared to negati v e control samples by the StepOne Plus machine (ThermoFisher, 4376600).

Statistical analysis
We performed statistics with Prism software 9.3.1 or with R (version 4.0.5)( 43 ).Most of the data were log 2 -transformed unless it is not written.Normal distribution was tested after this transformation.We used an unpaired t-test or Welch's test for pairwise comparisons.We analyzed multiple-group comparisons using one-way ANOVA with Dunnett's correction.For statistical analysis for proteomics see the specific section above.We used a r epeated-measur e two-way ANOVA test for helicase activity assay (with Tukey's correction) and for the analysis of li v e-imaging e xperiments.For the latter, we tested the normal distribution of the residuals of the data (by histograms) and used the Le v ene test to compare variances between the treatments within the da ta.Sta tistical tests wer e consider ed significant if adjusted p-values or FDR-corrected P -values ≤0.05.We show data as means ± SD (or ± SEM for helicase activity assay).

Figures' design
We placed and organized all the figures by using Adobe Illustrator software.We generated all the graphs by Prism or R (version 4.0.5)( 43 ) software.We generated the graphical abstract and Figures 4 A and 5 G by BioRender.com.

BLM is recruited to stress granules under a variety of stress conditions
BLM is often thought of as a nuclear DNA helicase ( 29 , 30 , 46 ).Howe v er, in a recent study, we characterized the composition of SGs as a result of sodium arsenate stress and found that BLM localizes in SGs ( 32 ).Since SG composition varies as a function of the stress type (47)(48)(49), we decided to test whether BLM is present in SGs as a result of other stressors; heat shock, puromycin, and thapsigargin.
Under basal growth conditions, without stress, BLM was mainly located in the nucleus.Howe v er, BLM was detected within cytoplasmic G3BP1-GFP expressing SGs under all the stress conditions tested in U2OS cells (Figure 1 A).The enrichment of BLM in SGs, relati v e to the surrounding cytoplasm, was in the range of ∼2to 4fold (average ×2.4 (thapsigargin), ×2.9 (sodium arsena te), ×3.0 (hea t shock) and ×4.2 (puromycin), Figure 1 B and Supplementary Data 1, Datasheet S2).In addition, colocalization between G3BP1-stained SGs and BLM was observed in fibroblasts and iPSC-derived neurons, under sodium arsenate stress (400 uM, 30 min, Supplementary Figure 1).This indicates that BLM is recruited to SGs, under broad types of cellular stress and in different cell lines.
For validation, in another EMSA experiment, the bands that correspond to the rG4s (dG4-cMyc-T15 or rG4-VEGFA-U15) bound to cBLM were extracted and analyzed by mass spectrometry, to re v eal the nature of the predominant proteins in these bands (Supplementary Figure 3B, C).Gel areas without cBLM in lanes with rG4s alone served as negati v e controls, and cBLM alone serv ed as a positi v e control.The main identified protein was the human BLM with peptide spectrum match score (#PSM) of ∼3-3000 times higher abundance than common laboratory contaminants and ∼100 times higher than negligible E.coli peptides (Thioredoxin-1, Supplementary Data 1, Datasheets S4).To substantia te tha t BLM is an rG4-binding protein, we next tested whether the binding of cBLM to rG4s is affected by the presence of the specific small molecule rG4-ligand Q UMA-1 ( 55 ), w hich has been shown to compete with rG4protein interactions ( 6 , 56 ).We found that QUMA-1 reduced cBLM binding to rG4s (Figure 2 C, D and Supplementary Data 1, Datasheet S5).The bound / free rG4 ratio (binding percentage) varied between different rG4-forming sequences, with the lowest binding pr efer ence of cBLM for rG4-NRAS.The competition of QUMA-1 decreased the binding capacity of cBLM to rG4-NRAS and rG4-BCL2 by a pproximatel y 50% ( P -value = 0.0034 or 0.0262, unpair ed t -test, r especti v el y), w hile it barel y affected cBLM binding to rG4-VEGFA-U15 (Figure 2 D).This might be related to cBLM's interactions with both VEGFA-U15 rG4 structure and its 3 single-stranded uridine sequence tail (U15), which is unlikely to bind QUMA-1.Together, this series of results confirm that BLM is both an dG4 and rG4 binding protein.

BLM unwinds rG4s in vitro and in cells and affects the enrichment of rG4s in SGs
Based on the abov e results, we e xamined the potential helicase activity of BLM on rG4s by an in vitro fluorescencebased unwinding assay ( 35 ).In this assay, a dabcyl quencher containing-rG4 oligonucleotide (S-rG4-VEGFA-U15) is hybridized with a 3 6FAM labelled short complementary oligonucleotide that has the potential to emit fluorescence onl y w hen unpaired from the quencher.cBLM led to the unwinding of the rG4-containing quencher, once ATP is added, triggering an increase of the FAM fluorescence.This fluor escence incr ease was found to be dependent on the cBLM concentration, reaching up to 40% of the unfolded rG4 at high cBLM concentrations (up to 16 uM, adjusted P -value = 0.0231, Two-way ANOVA repeated measure with Tukey's test; Figure 3 A and Supplementary Data1, Datasheet S6).
To demonstrate that BLM functions as an ATPdependent rG4 helicase, as pr eviously r eported for its G4 DNA helicase activity ( 29 , 57 ), we performed the unwinding assay with either ATP or non-hydrolyzable ATP analogue, A TPgammaS (A TPgS).We monitored a significant increase in unwinding cBLM activity in the presence of ATP as compar ed to ATPgS (Figur e 3 B and Supplementary Data 1, Datasheet S6).Ther efor e, BLM is an ATP-dependent rG4 helicase protein in addition to its known function as a dG4 helicase.
Next, we tested whether BLM serves as a helicase for endogenous rG4s.For this purpose, we knocked-down BLM in U2OS cells, using short interfering RNAs (siRNAs).Forty-eight hours after transfection, the le v els of BLM were decreased by a pproximatel y 50% (Supplementary Figure 4 and Supplementary Data 1, Datasheets S7 and S8).We then added QUMA-1 (0.5 uM, 3 hr).ImageStream analysis rev ealed higher le v els of endogenous rG4 after BLM knockdown (Figure 3 C and Supplementary Data 1, Datasheet S9) and similar results were obtained after knocking-down DHX36 expression, a known rG4 helicase ( 58 ).Orthogonall y, ImageStream anal ysis of Q UMA-1 intensities demonstra ted tha t BLM KO RPE cells display a higher QUMA-1 fluorescence, attributed to rG4 bona fide signal, as compared to wild-type RPE cells (Figure 3 D and Supplementary Data 1, Datasheet S10).This indicates that BLM indeed acts as an rG4 helicase in cells.
In addition, because rG4s accumulate and regulate SG formation ( 6 , 15 , 17 , 59 ), the knock-down of either BLM or DHX36 in U2OS cells resulted in a higher le v els of SGassociated rG4s, as compared to control (Figure 3 E and Supplementary Da ta 1, Da tasheet S11), supporting tha t BLM regulates rG4s le v els also within SGs.
Finally, we quantified the SG-enriched rG4 signal as a result of BLM KO in RPE cells (Figure 3 F).G3BP1-stained SGs contained a higher rG4 signal in the BLM KO RPE cells compared to wild-type RPE cells (Figure 3 F and Supplementary Da ta 1, Da tasheet S12).We note tha t the enrichment of the rG4 signal within SGs in RPE cells was generally lower than in the U2OS cells.Collecti v ely, this series of results confirmed that BLM regulates rG4 le v els in SGs.
BLM is recruited to SGs through rG4 binding under stress rG4 interactions contribute to SG formation ( 6 ).In this context, we further asked how rG4s recruit BLM to SGs.Under sodium arsenate-induced stress, we used APEX proximity labelling (60)(61)(62) to explore the impact of rG4 availability on the SG proteome.We induced G3BP1-APEX or nuclear export signal-APEX (NES-APEX) expression in U2OS cells (for labelling the SG proteome or cytoplasmic background, respecti v ely) that we pre viously de v eloped ((32), and Supplementary Figure 5).Cells were incubated with QUMA-1 (1 uM, 3 hr) to sequester rG4s or with DMSO (carrier; as control) and subjected to stress (150 uM sodium arsenate, 2.5 hr).APEX activity was induced with biotin phenol and hydrogen peroxide for biotin labelling of SG proteins.By western blot (WB) analysis, we confirmed the comparable protein pattern for each of the APEX baits (NES or G3BP1) either with QUMA-1 or DMSO (Supplementary Figure 6).Furthermore, when the APEX labelling was not activated, the background levels of detected proteins were negligible (Supplementary Figures 5 and 6).Therefore, QUMA-1 does not affect APEX labelling directly.
Follo wing the pull-do wn of biotinylated proteins, shotgun mass spectrometry identified proteins that are enriched or depleted in SGs, in response to the availability of cellular rG4s (Figure 4 A and Supplementary Figure 7).Overall, 472 SG-associated proteins were identified with good confidence (adjusted P -value < 0.05, FDR correction, log 2 FC > 0).Of these, 90 proteins were enriched and 20 depleted when rG4s were sequestered by QUMA-1 (adjusted P -value < 0.05, FDR correction, log 2 FC > 1).Clustering of experimental repeats between the proteomes af-fected by QUMA-1 versus control treatment was satisfying (Figure 4 B) and the differences in proteome composition in response to sequestration of rG4s by QUMA-1 were quantified (Figure 4 C, D and Supplementary Data 1, Datasheets S13 and S14).
BLM was relati v ely depleted fr om SG pr oteome in cells treated with QUMA-1, suggesting that rG4 le v els and / or their accessibility affect its recruitment (adjusted pvalue = 0.03; Figure 4 C, D).This observation was orthogonally v alidated b y a WB analysis of BLM, which was pulled down after G3BP1-APEX labelling under stress conditions (Figure 4 E and Supplementary Da ta 1, Da tasheet S15).In addition, a reduction in BLM abundance in SGs was observed by fluorescence microscopy in stained cells that wer e tr eated with Q UMA-1 or DMSO (carrier ; as control) prior to their fixation (Figure 4 F and Supplementary Data 1, Datasheet S16).In conclusion, BLM is a SG-associated rG4 helicase that is recruited to SGs in an rG4-dependent manner.

BLM inhibits SG formation
This wealth of data led us to postulate that BLM may regula te SG forma tion, as rG4s af fect SG forma tion ( 6 ), recruit BLM to SGs (Figure 4 ) and BLM likely unwinds rG4s (as demonstrated in vitro and in cells, Figure 3 ).To investiga te this, we manipula ted BLM le v els in U2OS cells.Li v ecell imaging re v ealed tha t SG forma tion was increased by knocking down BLM or DHX36 le v els, relati v e to control (Figure 5 A, B, E and Supplementary Data 1, Datasheet S17).Accor dingly, the ov ere xpression of mCherry-BLM or mCherry-DHX36 inhibited SG forma tion, rela ti v e to mCherry ov ere xpression control (Figure 5 C, D, F and Supplementary Data 1, Datasheet S18).eIF2-alpha phosphorylation was not induced by the ov ere xpression of mCherry, suggesting that it did not affect the cellular stress response by itself (Supplementary Figure 8, and Supplementary Data 1, Datasheet S19).
Ortho gonall y, we detected lower percentages of cells that harbor visible SG in cultures that express mCherry-BLM or mCherry-DHX36, relati v e to cultures that expressed mCherry alone (Figure 5 G, H and Supplementary Data 1, Datasheet S20).Altogether, we suggest a model whereby BLM is a SG-associated rG4 helicase tha t nega ti v ely regulates SG formation via unwinding of rG4s (Figure 5 I).

DISCUSSION
In this study, we demonstrate that BLM, a known nuclear DNA G4 helicase, also localizes to cytosolic SGs, under a variety of stress conditions and in different cell types.In SGs, BLM binds to and unwinds endogenous RNA G4s.Furthermore, BLM was found to be recruited into SGs in an rG4-dependent manner and regulates their formation.
We propose that cellular BLM le v els alter SG formation via rG4 unwinding, which is in line with our recent observation that rG4-protein interactions and rG4 availability contribute to SG formation ( 6 ).This hypothesis is further substantiated by both the results collected thanks to the newly de v eloped in vitro cBLM helicase activity and previous cellular results showing that rG4 accumulation in SGs is prevented by DHX36, which then leads to SG reduction ( 17 ).
C in 5 • C intervals (from 95 • C to 50 • C and from 30 • C to 25 • C) or in 10 • C intervals (between 50 • C to 30 • C) as follows: 85 -70 • C for 5 min each, 65-50 • C for 15 min each, 40-30 • C for 30 min each and 25 • C for 2 hr.All samples were stored at 4 • C.

Figure 1 .Figure 2 .
Figure 1.BLM is a resident protein of stress granules.( A ) Confocal micro gra phs of BLM imm unostaining (Cy5, Purple), in U2OS cells under a variety of stressors and co-localization with stress granule marker G3BP1-GFP (green).Nuclei (DAPI, blue).×63 lens.Scale bar -20 um.Inset scale bar 2 um.Intensity profiles for SGs and BLM channels in r epr esentati v e SGs under different stress conditions using Fiji software.( B ) Box plot of BLM enrichment in SGs, which were quantified from micro gra phs of > 30 cells per treatment.Median (|) and mean (+).Analyzed using Fiji software.

Figure 3 .
Figure 3. BLM unwinds rG4s in vitro and in cultured cells.In-vitro unwinding assay monitored in real-time, as the relati v e emission of a 6FAM-labelled fluorescent short oligo (% unwound) unwound from a da bcyl la belled quencher oligo, VEGFA-rG4-U15, and in response to ( A ) cBLM concentrations (1, 4, 8, 16 uM) with ATP addition (+ATP) or ( B ) 8 uM cBLM in addition of ATP or a non hydrolysable analogue A TPgS (+A TP / +A TPgS).Data normalized to 0 μM cBLM and to the first time point, per condition.Average of two or five technical repeats in A or B, respecti v ely; Two-way ANOVA repeated measure with Tukey's test for multiple comparisons.Representati v e scatter plots of the endogenous rG4 signal gained by quantification of QUMA-1 staining in ( C ) U2OS cells with siRNA knock-down of Dhx36 or Blm.Stress induced by sodium arsenate, 150 uM, 2.5 hr.siControl non-targeting oligonucleotides served as controls, or ( D ) RPE wild-type versus BLM KO cells.Three experimental repeats; log 2 -transformed rG4 intensity in individual cell, normalized to the cell area.Horizontal line -mean.Representati v e box plots of endogenous rG4 signal, gained by QUMA-1 quantification in stress granules, relati v e to adjacent cytoplasm of ( E ) U2OS cells or ( F ) RPE wild-type versus BLM KO cells, under conditions identical to those in panels C and D. Three experimental repeats; Median (|) and mean (+) of log 2 -transformation of rG4 enriched signal in stress granules.One-way ANOVA with Dunnett's test (C, E), Two-tailed (D) or one-tailed (F) unpaired t -test.* P -value < 0.05, ** P -value < 0.01, *** P -value < 0.001, **** P -value < 0.0001.

Figure 4 .
Figure 4. BLM is recruited to SGs in an rG4-dependent manner.( A ) Diagram of the experimental design.( B ) Principal component analysis of the proteomic content of APEX-isolated stress granules under QUMA-1 treatment (1 uM, 3 hr) or control (carrier, DMSO).( C ) A volcano plot of APEX-isolated SG proteins, obtained under QUMA-1 treatment (orange), relati v e to DMSO control (blue).Y-axis -log 10 of P -value ( P -value < 0.05) and x-axis, log 2 values of fold-change.BLM is highlighted in r ed.Gr ey proteins were lacking statistical significance.( D ) Heatmap of unsupervised clustering of the final 472 SG-associa ted proteins tha t were enriched or depleted under experimental conditions in stress granules (upper), and bar plot r epr esents BLM's intensity under QUMA-1 or DMSO conditions (orange / blue, lower).FDR corrected P -value (* adjusted P -value < 0.05).( E ) APEX-isolated SG proteins blotted with anti-BLM antibody after QUMA-1 treatment normalized to DMSO control.BLM is r epr esented at ∼169 kDa.( F ) Repr esentati v e box plot of BLM enrichment in SGs.Quantification of immuno-stained confocal micrographs with median (|) and mean (+) of log 2 -transformation of BLM enriched signal in SGs, relati v e to adjacent cytoplasm of U2OS cells.Three experimental r epeats.Two-tailed unpair ed t -test, * P -value < 0.05 (E),*** P -value < 0.001 (F).

Figure 5 .
Figure 5. BLM negati v ely regulates SG formation.Quantification of the ratio of SG area, to the cell area, by li v e imaging of G3BP1-GFP in U2OS cells treated with 150 uM of sodium arsenate for 2.5 hr.siRNA knockdown of ( A ) Dhx36, or ( B ) Blm compared to siControl.Ov er-e xpression of ( C ) mCherry-DHX36, or ( D ) mCherry-BLM compar ed to mCherry-only over expr ession control.Four sites per well, 3-4 wells per condition.Three independent repeats.Two-way ANOVA repeated measures with FDR correction, **** P -value < 0.0001.( E ) Box plot quantification of data from (A, B) and ( F ) Box plot quantification of data from (C, D), 120 min after stress induction.Data normalized to control average, per repeat.One-way ANOVA with Dunnett's test, * P -v alue < 0.05, ** P -v alue < 0.01, *** P -v alue < 0.001.( G ) Bar plot of the percentage of stress granule-positi v e U2OS cells with ov ere xpression of mCherry-DHX36 or mCherry-BLM, compared mCherry only as a control.ImageStream study, one-way ANOVA with Dunett's test, * P -value < 0.05.( H ) Representati v e micro gra ph channels: G3BP1-GFP (gr een), mCherry (r ed), Hoechst 44432 (Blue).( I ) A model for the regulatory role of BLM in SG formation through the unwinding of rG4.BLM is recruited to SGs, where plausibly it performs its rG4 helicase activity.