DHX36 binding at G-rich sites in mRNA untranslated regions promotes translation

Translation efficiency can be affected by mRNA stability and secondary structures, including so-called G-quadruplex (G4) structures. The highly conserved and essential DEAH-box helicase DHX36/RHAU is able to resolve G4 structures on DNA and RNA in vitro, however a system-wide analysis of DHX36 targets and function is lacking. We globally mapped DHX36 occupancy in human cell lines and found that it preferentially binds to G-rich sequences in the coding sequences (CDS) and 5' and 3' untranslated regions (UTR) of more than 4,500 mRNAs. Functional analyses, including RNA sequencing, ribosome footprinting, and quantitative mass spectrometry revealed that DHX36 decreased target mRNA stability. However, target mRNA accumulation in DHX36 KO cells did not lead to a significant increase in ribosome footprints or protein output indicating that they were translationally incompetent. We hypothesize that DHX36 resolves G4 and other structures that interfere with efficient translation initiation.


38
All messenger RNAs (mRNAs) are coated with a dynamically changing repertoire of RNA 39 binding proteins (RBP) to form ribonucleoprotein particles (RNP) that mediate mRNA splicing, 40 editing, transport, turnover, and translation (Gerstberger et

121
While DHX36 has been described as both a DNA and RNA helicase in vitro, the major 122 source of nucleic acids in the cytosol is RNA. Thus, we tested whether DHX36 indeed interacted 123 with RNA, in particular mRNAs, in human cell lines. Cells were crosslinked using 254 nm UV-124 light in HEK293 cells followed by purification of polyadenylated RNA (Fig. 1D). We found that

154
RNPs of two biological replicates and generated small RNA cDNA libraries for next-generation 155 sequencing. We used the PARalyzer software (Corcoran et al., 2011) to determine clusters of 156 overlapping reads that contained T-to-C mutations diagnostic of the crosslinking event at higher 8 biological replicates from the FH-DHX36 and FH-DHX36 E335A PAR-CLIP exhibited high 159 correlation, with an R 2 of 0.79 and 0.93, respectively (Suppl. Fig. S1A

165
Our approach succeeded to capture 22 of 27 previously published targets of DHX36, 166 including TERC and PLAU, confirming the validity of our approach (Suppl . Table S1).

167
Consistent with their mainly cytoplasmic localization, 70% and 73% of FH-DHX36 and FH-168 DHX36 E335A binding sites, respectively, mapped to the exonic regions of more than 4,500 169 protein-coding genes with the rest found on intronic sequences or non-coding RNAs (Fig. 2C).

170
FH-DHX36 and FH-DHX36 E335A did not exhibit preference for the coding sequence or  Binding of FH-DHX36 and FH-DHX36 E335A to mRNAs showed no correlation to transcript 180 length or abundance as determined by RNA-seq in HEK293 cells (Suppl. Fig. S1C, D). This 181 suggested sequence-or structure-dependent determinants of FH-DHX36 binding, rather than 182 unspecific interactions. Therefore, we aimed to define the preferred RRE of FH-DHX36 and FH-183 DHX36 E335A. We first counted the occurrence of all possible 5-mer sequences in our high-9 confidence binding sites and calculated their Z-score over a background of shuffled sequences 185 of the same nucleotide composition. 5-mers that contained at least three guanines were 186 enriched in both, FH-DHX36 and FH-DHX36 E335A, PAR-CLIPs (Fig. 3A

187
S2A and Table S2) and in the FH-DHX36 E335A PAR-CLIP we additionally found the 188 enrichment of 5-mers that were A/U-rich ( Fig. 3C and Suppl. Table S2). We also used MEME 189 (Bailey et al., 2009) as an alternative in silico approach to define the RRE. Here, the most 190 significantly enriched RRE was also G-rich, and matched the motif for G4 formation (Todd et al.,191 2005) (Fig. 3D). Indeed, circular dichroism (CD) spectroscopy analysis confirmed that synthetic 192 oligonucleotide corresponding to an representative PAR-CLIP RRE showed the characteristic 193 spectrum of a parallel G4 structure, while sequence mutations exchanging G residues with A or 194 C resulted in a loss of the G4 signature (Suppl. Fig. S2B). Microscale thermophoresis 195 experiments confirmed that FH-DHX36 specifically bound the G4 forming oligonucleotides, but 196 not the mutant s (Fig. 3E).

197
Considering that this representative RRE was able to form a G4 structure we asked 198 whether the FH-DHX36 PAR-CLIP binding sites were contained in a previously reported dataset 199 that mapped G4 structures in vitro in polyadenylated RNA from HeLa cells in a transcriptome-200 wide manner (Kwok et al., 2016). Interestingly, 74% of the potential G4 sites found by Kwok et . Table S3). For the following analyses we focused on 220 DHX36 targets from the deeper FH-DHX36 E335A PAR-CLIP dataset (Fig. 4). Nevertheless, we 221 observed similar results using the highly overlapping FH-DHX36 PAR-CLIP data (Suppl.

265
Analogous to RNA abundance shown above, potentially G4 forming RNAs exhibited a 17% 266 decreased TE upon DHX36 KO (Fig. 6F). Taken together with our observation that >90% of 267 DHX36 were not co-sedimenting with translating ribosomes and thus was unlikely to influence 268 translation elongation, our data suggest that DHX36 increases the translational competence of

306
DHX36 binding sites transcriptome-wide allowed us to reconcile these binding models (Fig. 2).

307
DHX36 did indeed preferentially bind G-rich sequences, including sites in the 3' UTR of the first 308 described DHX36 target gene, PLAU. Nevertheless, we also observed an enrichment of A/U-309 rich sequences in DHX36 PAR-CLIPs using the catalytically inactive mutant DHX36 E335A ( Fig.   310 3). We think A/U-rich elements are genuine binding sites of DHX36 and may even serve as the 311 protein recruitment sites to mRNAs. However we suggest that DHX36 translocation from these 312 sites is rapid due to their less structured nature, which prevents efficient UV crosslinking to the 313 catalytically active helicase.

314
Most hypotheses on DHX36 function relate to its ability to bind and resolve G4

338
We speculate that most G4 resolving factors remain to be identified, considering that       Table. 390 Co-transfection of these plasmid together with pOG44 plasmid using Nanofectin resulted in

499
were averaged between ten accumulations with an instrument scanning speed of 10 nm/sec.

500
Oligodeoxynucleotides used in this experiment were purchased from Sigma-Aldrich.

637
Magnetic beads (Sigma) were added and incubated for 5 h on a head-over-tail wheel at 4°C.