Cancer-associated DDX3X mutations drive stress granule assembly and impair global translation

DDX3X is a DEAD-box RNA helicase that has been implicated in multiple aspects of RNA metabolism including translation initiation and the assembly of stress granules (SGs). Recent genomic studies have reported recurrent DDX3X mutations in numerous tumors including medulloblastoma (MB), but the physiological impact of these mutations is poorly understood. Here we show that a consistent feature of MB-associated mutations is SG hyper-assembly and concomitant translation impairment. We used CLIP-seq to obtain a comprehensive assessment of DDX3X binding targets and ribosome profiling for high-resolution assessment of global translation. Surprisingly, mutant DDX3X expression caused broad inhibition of translation that impacted DDX3X targeted and non-targeted mRNAs alike. Assessment of translation efficiency with single-cell resolution revealed that SG hyper-assembly correlated precisely with impaired global translation. SG hyper-assembly and translation impairment driven by mutant DDX3X were rescued by a genetic approach that limited SG assembly and by deletion of the N-terminal low complexity domain within DDX3X. Thus, in addition to a primary defect at the level of translation initiation caused by DDX3X mutation, SG assembly itself contributes to global translation inhibition. This work provides mechanistic insights into the consequences of cancer-related DDX3X mutations, suggesting that globally reduced translation may provide a context-dependent survival advantage that must be considered as a possible contributor to tumorigenesis.


Supplemental figure legends:
Supplemental figure S1. High expression and punctated distribution of DDX3X in human cancer.
a. Histopathological assessment of DDX3X expression in a panel of pediatric medulloblastomas and normal brain tissues. Representative images obtained from normal brain cortex and cerebellum and 10 pediatric medulloblastomas.  a. Table summarizing the the raw reads and the reads uniquely  showed DDX3X relocalization to SG under stress condition when control siRNA oligos were used; whereas only ~30% of cells treated with G3BP1/G3BP2 siRNAs showed DDX3X in SG. (N=3, ***; P < 0.001, Student t test between sicontrol and siG3BP1/siG3BP2 treated samples).

Supplementary Table S1. List of peaks identified in the DDX3X CLIP-seq experiments
The table presents all the high confident peaks identified in the DDX3X

Supplementary Table S2. Gene Ontology analysis of DDX3X CLIP-seq targets
David Bioinformatic Resources 6.7 was used to perform gene ontology analyses. mRNAs encoding proteins involved in translation are enriched among the DDX3X mRNA targets.
Samples were denatured by boiling for 3 min in 4X NUPAGE LDL sample buffer (Life Technologies, cat. no. NP0007) supplemented with β-mercaptoethanol and were run in a 4-12% NUPAGE Bis-Tris gel (Invitrogen). Protein were transferred into a nitrocellulose membrane (Amersham, cat. no. RPN3032D) using a semi-dry system at 13 V for 1 h, and membranes were blocked with 5% milk in PBS-T for 30 min.

Pulse chase
Tet-On inducible HEK293T cells were treated with 1 µg mL −1 doxycycline for 24 h (pulse) and then washed with PBS three times. Cells maintained in normal media without doxycycline for the indicated time points (chase). Protein lysates were obtained and Western blot was performed against FLAG-DDX3X and β-actin. β-actin was used to normalyzed relative DDX3X levels in each sample, and the residual DDX3X was calculated by designing DDX3X levels in time point 0 min as 100%. Image J64 was used to quantify the relative levels of DDX3X.

Analysis of human medulloblastomas
Human tumor samples were obtained with informed consent through an institutional review board approved protocol at SJCRH. For immunofluorescence, tissue microarrays of human tumor samples were stained using standard protocol in association with anti-DDX3X antibody (Bethyl, cat. no. A300-475A, 1:200), and nuclei were counterstained with 4',6-diamidino-2-phenylindole (DAPI; Vector Labs). Images were captured using a LSM510 (Zeiss) confocal microscope with a 63X objective and Zeiss ZEN software. Protein expression and punctate intensity were measured using the ImageJ64 software (ImageJ 1.47v). All images were analyzed using the same threshold and the image particles were then analyzed. The average number of particles was used for the punctate intensity. The protein expression was calculated by measuring the intensity after setting a common threshold for all images.

CLIP-seq
HEK293T cells were cross-linked on ice with UV radiation (254 nm) once at 400 mJ per cm 2 followed by a second UV treatment using 200 mJ per cm 2 in the presence of cold PBS (5 ml per 10 cm dish). Cells were pelleted at 1,000 rpm at 4 °C and stored at -80 °C until further use. Cells lysis was performed in cold Nonidet P-40 buffer (50mM Tris-HCl, pH 7.5, 150mM NaCl, 0.5% Nonidet P-40, 50mM NaF, and 1 × proteinase inhibitor cocktail; Roche Diagnostics, cat. no. 11836145001) for 10 min on ice followed

RNA binding assay
After transfection with the desired plasmids for 32 h in 10 cm dish, HEK293T cells were UV-treated and collected as for CLIP-seq experiment. Cell lysis was done using

RNA immunoprecipitation (RIP) and RT-qPCR
For experiments using exogenous DDX3X variants, HEK293T were transfected with 5 µg pCDNA3.1 FLAG-tagged DDX3X plasmids in 10 cm culture plates for 32 h. Plates were washed once with 1× cold PBS and treated once with 400 mJ per cm 2 UV followed by a second UV treatment using 200 mJ per cm 2 in the presence of cold PBS (5 ml per 10 cm dish). Cells were collected by centrifugation for 5 min at 2000 r.p.m. at 4 °C, resuspended in 1 mL PBS and centrifuged again at maximum speed for 10 seconds. Pellet were stored at -80 °C until further use. Cells lysis was performed in cold Nonidet P-40 buffer for 10 min on ice followed by addition of a 1:50 volume of RQ1 DNase and incubation at 37 °C for 10 min at 1000 r.p.m. Protein cell extracts were centrifuged for 10 min at 14,000 r.p.m. at 4 °C. The supernatant was incubated at 4 °C for 3.5 h with 10 µg of anti-FLAG M2 antibody previously conjugated to 50 µl protein G Dynabeads. Beads were washed three times with 500 µl cold NP-40 lysis buffer, twice with 500 µl high salt buffer, twice with 500 µl PNK buffer. For RNA isolation, beads were treated with proteinase K as in CLIP-seq experiment. RNA extraction was done using Acid Phenol:Chloroform and ethanol precipitation. One-step RT-qPCR was performed using TaqMan RNA-to-Ct 1-Step kit according to the manufacture's instructions (Applied Byosystems, cat. no. 4392938) in costume designed Taqman array fast plates containing primer/probe sets against DDX3X targets (Applied Biosystems). RT-qPCR for input RNA was also done using 5% of each sample. For mRNA expression analyses, total RNA was extracted from HEK293T cells (previously transfected with FLAG-DDX3X constructs for 32 h or with control/DDX3X siRNA oligos for 48 h) using RNeasy kit according to manufacturer's instructions (Qiagen) followed by one-step RT-qPCR analysis as indicated above.
Quantitative PCR analyses were done using a 7900HT Fast Real Time PCR System (Applied Biosystems, SDS Software 2.2).

Ribosome half-transit time measurements
HEK293T cells (~ 50% confluent) were transfected in 15 cm dish with 15 µg plasmids for 24 h using Lipofectamine 2000. After transfection, cells were trypsinzed (TripLE Express, GIBCO) and washed with 5 mL 1× PBS. Cells were then incubated with 5 mL labeling media for 30 min (see above). 35 S-Met/Cys was then added to a final dilution of 50 µCi per mL. At the indicated time points, 800 µL media were removed and placed on a new tube containing 1 mL 1× cold PBS+100 µg ml −1 CHX to stop the reaction.
Cells were quickly spun down and washed once more with 1 mL 1× cold PBS+100 µg ml −1 CHX. Cells were pelleted, and lyzed in 500 µL polysome buffer for 20 min on ice.

CLIP-seq:
Mapping: Illumina single ended reads were trimmed of adapter using Cutadapt and any reads less than 21bp were excluded from downstream analysis. The resulting paired-end reads were aligned to four databases using Burrows-Wheeler Aligner (BWA 0.5.5) 2 : (i) human NCBI Build 37 reference sequence, (ii) RefSeq, (iii) a sequence file that represents all possible combinations of non-sequential pairs in RefSeq exons, and (iv) AceView flat file (UCSC), representing transcripts constructed from human expressed sequence tags (ESTs). After this initial mapping, final BAM files were produced by selecting the best alignment in the four databases. Duplicates were removed. Coverage: Effective coverage of the whole genome and whole exome was obtained by summarizing coverage of aligned bases with quality score ≥ 15 at each position of the reference genome (excluding sequencing gaps and ambiguous bases) using the Coverage module of Bambino 3 . Wig files were converted to BigWig files using the UCSC tool; wigToBigWig. Peak finding: The genome was split into 100Kb bins with 10Kb overlap and coverage in these windows assessed (using the UCSC tool bigWigSummary) for maximum coverage greater than or equal to the desired cut-off (10). If these criteria were met then this bin was further split into smaller and smaller bins to find the exact peak location. Once this window was small enough base pair by base pair coverage was assessed. Peaks with at least 10 bp long were then refined and merged if required. Peaks present in both the DDX3X and the control IgG CLIPseq were removed. Peaks were manually reviewed using a local mirror of the UCSC genome browser displaying bigWig coverage files as custom tracks to assess their validity.
We used the HOMER (Hypergeometric Optimization of Motif EnRichment) 4 tool "annotatePeaks.pl" to annotate the genomic ontology and gene ontology of each peak.
This provided details of the more gross features with peaks and also the possible enrichment of pathways.

RNA-seq:
Mapping: See CLIP-seq mapping. Expression Analysis: To compare the expression of genes which contained DDX3X peaks a simple count of the number of reads mapping to that gene was carried out. To compare the expression of genes independently of the CLIPSeq analysis we first obtained counts of the number of reads per gene and carried out FPKM normalization (taking into account gene length and total reads). We applied a log-start transformation to the normalized read counts to minimize the effects of small differences in read counts. Log2ratios were then computed and the average of each comparison calculated. These results were then plotted in a Magnitude Average plot (MA plot) in a manner similar to microarray analysis.
We also applied a method that utilized exon and exon-exon junction reads. Exon reads were once again normalized by FPKM and exon junctions by the total read count (RPM) and proportional difference values were used in both cases. High confidence results were defined by 10 reads in at least one class.

Ribosome profiling:
Ribosome profiling (Ribo-seq) data analysis was carried out using the same peak detection algorithm as the CLIP-Seq data analysis. Peak statistics based on location in the genome were also determined using the same methods as described in the CLIPseq section.