Expression of DDX3X missense mutations in mouse primary neural progenitors through lentiviral delivery

Individuals carrying DDX3X missense mutations present with a spectrum of clinical and molecular phenotypes, encompassing neuroanatomical disruption, developmental disability, and altered biochemical activity. To model DDX3X syndrome in primary neural cells, we focused on four de novo DDX3X missense mutations, which cause diverse clinical presentations and molecular features, and are recurrently found in 4–8 individuals each, all females. This includes DDX3X^R376C and DDX3X^R488H, which are clinically mild and retain some helicase activity, and DDX3X^R326H and DDX3X^T532M, which are clinically severe and helicase-dead (Fig 1B and S1 Table) [9]. MRI of patients with clinically mild mutations show corpus callosum defects and decreased white matter volume (data available for 5/8 DDX3X^R376C patients, 4/5 DDX3X^R488H patients) [9,12,14]. Severe mutations are associated with partial agenesis of corpus callosum and reduced white matter volume, accompanied by PMG and, in some cases, microcephaly and developmental delay (data available for 4/4 DDX3X^R326H patients, 4/5 DDX3X^T532M patients) (Fig 1B and S1 Table) [9,14,21].

A challenge in studying RBPs and missense variants, including DDX3X, is that overexpression can cause cellular phenotypes such as apoptosis [5,7,35,36]. Thus, we developed a lentivirus strategy to introduce DDX3X missense variants into mouse neural progenitors without overexpression. We isolated primary neural progenitor cells (NPCs) and neurons from embryonic day (E)12.5 mouse cortices, pooling brains from independent embryos to create a homogenous cell population, which was then used for primary cell culture. These cells were transduced with lentivirus containing GFP-tagged wild-type (WT) or mutant human (Hs) DDX3X. Importantly, mouse and human DDX3X show 98.6% identity at the protein level. After a two-day period for lentiviral expression, cells were analyzed (Fig 1C). We optimized the lentivirus amount and timing for these experiments by monitoring levels of induced GFP-tagged HsDDX3X protein by western analysis (Fig 1D–1F). We identified experimental conditions in which total DDX3X protein levels were relatively equivalent to WT and not over-expressed, with the missense variants expressed at about 30% that of the endogenous protein (Fig 1E and 1F). Further, using fluorescence microscopy, we demonstrated DDX3X was transduced in virtually all cells (S1A and S1B Fig). Amongst transduced cells, there was some variation in GFP levels. However, this occurred equivalently across all DDX3X variants. These experiments establish an efficient strategy for evaluating the phenotypic impact of DDX3X mutations upon neural cells.

DDX3X missense mutations differentially impair neuron production and apoptosis

Previous studies showed that Ddx3x LoF disrupts progenitor cell cycle and its ability to produce neurons, however whether missense mutations also affect progenitors is unknown [34]. Using our paradigm for expression of DDX3X missense variants, we next assessed their functional impact upon cell fate. Towards this, we utilized a mouse line expressing Dcx::DsRed as a reporter for newborn neurons [37]. Lentiviral delivery of DDX3X was performed as in Fig 1C, followed by live imaging for 24 hours, and assessment of progeny, as previously [38] (Fig 2A). We first assessed the extent to which DDX3X missense mutations impact progenitors’ ability to divide. Progenitors expressing DDX3X^R376C, DDX3X^R488H, and DDX3X^T532M, each exhibited successful divisions 100% of the time. In contrast, DDX3X^R326H expression led 20% of progenitors to fail cell division (S1C and S1D Fig). Mitosis duration was not overtly different across the variants (S1E Fig). However, DDX3X^T532M and DDX3X^R488H displayed a higher rate of progenitor re-divisions (S1F and S1G Fig). This suggests that these mutations may affect progenitor cell cycle.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 2. DDX3X missense mutations impair neural cell fate and survival.

A) Schematic of the experimental timeline. Neural progenitor primary cells were isolated at E12.5 and DDX3X^WT and missense mutations were introduced through lentiviral delivery. After 2 days in vitro (2 DIV), live imaging was performed for 24hrs. B) Example of proliferative, asymmetric and neurogenic divisions. Arrows, neural progenitors (DCX::dsRed−): arrowheads, newborn neurons (DCX::dsRed+). C) Quantification of cell fate for proliferative symmetric (green), asymmetric (blue) and neurogenic asymmetric (red) divisions. D) Example of division with apoptotic progeny. E) Quantification of apoptotic progeny. F) Quantification of the percentage of progenitors that underwent apoptosis over total number of dsRed− cells. G) Quantification of the proportion of neuron that underwent apoptosis over total number of dsRed+ cells. (C, E) χ² analysis with post-hoc Bonferroni. n = 3 live-imaging sessions, n = 3 litters; no virus n = 545 divisions, DDX3X^WT n = 645 divisions, DDX3X^R326H n = 365 divisions, DDX3X^R376C n = 410 divisions, DDX3X^R488H n = 649 divisions, DDX3X^T532M n = 651 divisions. Scale bars: 10 μm. (F, G) Each dot represents a pooled primary culture. Two-way ANOVA. **p < 0.001; ****p < 0.0001; ns, not significant. Data are mean±SD. “S”: severe; “M”: mild.

https://doi.org/10.1371/journal.pgen.1011555.g002

We next assessed if DDX3X missense mutations impact the type of divisions which progenitors undergo. Proliferative symmetric divisions were defined by production of two DsRed− cells, neurogenic asymmetric divisions by one DsRed− (progenitor) and one DsRed+ cell (neuron), and neurogenic symmetric divisions by two DsRed+ cells (Fig 2B). Progenitors expressing the clinically severe DDX3X mutations (DDX3X^R326H and DDX3X^T532M) exhibited significantly more proliferative divisions and fewer symmetric neurogenic divisions, relative to control cells (Fig 2C). In comparison, expression of clinically mild variants had only subtle impacts on cell fate, with no alterations observed with DDX3X^R488H and only slightly fewer asymmetric divisions with DDX3X^R376C (Fig 2C). Strikingly, all DDX3X missense mutations induced apoptosis in progeny which we identified by cell morphology as previously [38] (Fig 2D and 2E). These data demonstrate that DDX3X missense mutations differentially impact progenitors’ ability to produce viable neurons, with clinically severe mutations causing significant impairment of neurogenesis.

We also assessed cells that did not divide during the live imaging session to measure the extent to which DDX3X mutations influence survival of either progenitors (dsRed−) or neurons (dsRed+). Amongst all four mutations, only DDX3X^R376C expression led to a slight but significant increase in progenitor death compared to DDX3X^WT (Fig 2F). However, in neurons, expression of either clinically severe mutation, DDX3X^R326H and DDX3X^T532M, led to a striking threefold increase in apoptosis as compared to DDX3X^WT and clinically mild mutations (Fig 2G). These analyses indicate that DDX3X missense mutations influence cell viability, with a particularly profound impact of the clinically severe mutations upon neuronal survival.

DDX3X clinically severe missense mutations induce diverse transcriptomic changes in neural progenitors and neurons

We next investigated the underlying molecular mechanisms by which clinically distinct DDX3X missense mutations influence cell fate. Towards this, we conducted transcriptome analysis of pooled primary cells either without lentiviral transduction (control) or with lentiviral-mediated expression of DDX3X^WT or missense mutations (Fig 3B). Importantly, DDX3X^WT expressing cells did not display any differentially expressed transcripts compared to control (Fig 3D and S2 Table). Notably, in all conditions (WT and 4 mutations), there were no transcriptional changes in Ddx3x (total exogenous and endogenous) (S2A and S2C Fig). These data further confirm that our experimental setup did not induce Ddx3x overexpression.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 3. Transcriptome analysis reveals DDX3X missense mutations alter expression of DNA damage response genes.

A) Heatmap of differentially expressed (upregulated: orange, downregulated: blue, not significant: white) transcripts across the four missense mutations, organized by Gene Ontology categories. B) Schematic of experimental design: RNA was extracted from primary cultures with and without viruses and used for bulk RNA sequencing. C) Venn diagram of significant differentially regulated transcripts across all four mutations relative to control. D, E, F, G, H) Volcano plots of transcriptome datasets for DDX3X^WT and each indicated missense mutation. Each dot represents a transcript. Pink: transcript fold change >1.5 or <−1.5, p-value <0.05; grey: not significant; green: selected hits of biological significance. I, J) RNA levels of Atm (I) and Parp1 (J) measured by qPCR. RNA levels normalized of Gapdh. Each dot represents pooled primary culture. Two-way ANOVA. *p < 0.01; **p < 0.001. Data are mean ± SD. “S”: severe; “M”: mild.

https://doi.org/10.1371/journal.pgen.1011555.g003

We then examined the extent to which the global transcriptome was affected by DDX3X mutations (S2 Table). We identified extensive differentially expressed transcripts in cells expressing the severe mutations (DDX3X^R326H: 331 DDX3X^T532M: 259) (Fig 3A, 3C, 3E and 3H). In contrast, despite expression at similar levels, the mild mutations impacted far fewer transcripts (DDX3X^R376C: 94; DDX3X^R488H: 74 (Fig 3A, 3C, 3F and 3G). Notably, most of these differentially expressed transcripts were also present in the severe mutation datasets (Fig 3A and 3C). The DDX3X paralog on the Y chromosome, Ddx3y, was upregulated 3-fold in all missense mutation datasets relative to WT (S2B Fig). Notably, the number of reads mapped on the X and Y chromosome was consistent between experimental replicates (S2C Fig). Upregulation of Ddx3y has been previously observed in Ddx3x LoF brains, suggesting that it compensates for reduced Ddx3x levels in this model [34]. The upregulation of Ddx3y suggests it also could compensate for possible DDX3X sub-optimal function associated with the missense mutations [39]. Together, these transcriptome data reveal divergent impacts of missense mutations on the cellular transcriptome, with the most profound disruption evidenced with severe mutations.

Gene Ontology (GO) analysis of differentially regulated transcripts across all the DDX3X missense mutations revealed enrichment in categories associated with cellular death, cellular stress, P53 activation and DNA damage (Figs 3A and S2D and S2 Table). The transcripts associated with DNA damage were all upregulated, including several kinases involved in P53 activation (Atm, Atr, DNA-PK, and Parp1) [40] (Fig 3A, and 3E–3H). We independently validated the increased expression of Atm and Parp1 using qPCR of cells expressing DDX3X missense mutations (Fig 3I and 3J). These transcriptome findings highlight altered pathways associated with DNA damage response, including some which may be due to defective DNA replication. This is notable given our finding that DDX3X missense mutations caused extensive apoptosis (Fig 2).

DDX3X clinically severe missense mutations induce DNA damage and cause aberrant R-loop formation

Given these transcriptomic changes, we postulated that DDX3X mutations may impair genomic stability. To test this, we first assessed if expression of DDX3X missense variants activates the DNA damage response (DDR) gene, P53. For this we quantified the fraction of GFP+P53+ cells. All missense mutations led to P53 activation in more than 80% of cells, while control and DDX3X^WT did not (Fig 4A and 4D). Similarly, all missense mutations exhibited a striking increase in pH2A.X+ cells (DNA double strand breaks), with the clinically severe mutations (DDX3X^R326H and DDX3X^T532M) showing damage in ~60% of cells (Fig 4B and 4E). These data are consistent with the transcriptomic signature of elevated DDR in cells expressing all mutations, but especially severe variants. DNA damage was highest in Tuj1+ neurons, especially in those expressing clinically severe mutations (Fig 4B and 4F). This demonstrates that neurons expressing severe mutations are especially prone to DNA breaks, consistent with the higher apoptosis of neurons seen with expression of these specific variants (Fig 2G).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 4. Clinically severe DDX3X missense mutations induce DNA damage in neurons and accumulation of R-loops in the cytoplasm.

A) Representative images of P53 staining (magenta), GFP-HsDDX3X (WT or indicated mutation, green), and Hoechst (blue) in primary cells. B) Representative staining of Tuj1 (neurons, red) and pH2A.X (double strand breaks, magenta), GFP-HsDDX3X (WT or indicated mutation, green), and Hoechst (blue) in primary cultures. C) Representative images of S9.6 (R-loops, red), GFP-hsDDX3X (WT or indicated mutation, green), and Hoechst (blue) staining in primary cells. D) Quantification of the fraction of GFP+P53+ cells showed in A. E, F) Quantification of the fraction of GFP+pH2A.X+ (E) the fraction of GFP+pH2A.X+TuJ1+ (F) cells shown in B. G) Quantification of the fraction of GFP+S9.6+ cells shown in C. Scale bars: 10 μm. Two-way ANOVA *p < 0.01; **p < 0.001; ***p < 0.0001; ****p < 0.0001; ns, not significant. Eat dot represents a pooled primary culture. Data are mean ± SD. “S”: severe; “M”: mild.

https://doi.org/10.1371/journal.pgen.1011555.g004

As DDX3X is an RNA binding protein, we postulated that DDX3X mutations may induce DNA damage by influencing RNA metabolism. Of note, DDX3X binds not only RNA:RNA duplex but also ssDNA and DNA:RNA hybrids (R-loops) in vitro [41,42]. R-loops are essential for DNA replication and RNA transcription, and defective resolution of these structures has been associated with increased DNA damage [43]. Thus, we hypothesized that DDX3X missense mutations may increase R-loops in neural cells, causing cytoplasmic accumulation as has been recently reported in immortalized cells [43–46]. DDX3X^R376C and DDX3X^R488H mutations did not affect the presence of R-loops (S9.6+) (Figs 4C and 4G and S2F). This is consistent with the hypomorphic impact of these mutations upon RNA helicase activity [9] as well as relatively mild induction of DNA damage (Fig 4E). In contrast, cells expressing either clinically severe mutation showed significant cytoplasmic accumulation of R-loops (Fig 4C and 4G). Low levels of S9.6 were detectable in the nucleus, supporting the notion that this signal is due to co-transcriptional R-loop formation (S2E Fig) [43]. Together, our findings reveal that expression of DDX3X missense mutations, but not DDX3X^WT, induces DNA damage and P53 activation. Clinically severe mutations cause especially high DNA damage, particularly in neurons, and this is associated with defective RNA metabolism.

DDX3X missense mutations have both shared and unique protein interactomes

To further understand how missense mutations impact RNA metabolism and cellular fate, we next examined their protein interactomes. We thus conducted proximity labeling to discover the protein interactome of DDX3X^WT and missense mutations within neural cells in an unbiased fashion. For this we used TurboID fused to either DDX3X^WT or missense protein variants. TurboID, a biotin ligase that biotinylates nearby proteins, is commonly used to investigate proteins in close proximity and presumed interactors [47] (Fig 5A). These were expressed in primary neural progenitors using lentivirus as in Fig 1. Biotinylated proteins were then affinity purified using streptavidin beads and subjected to quantitative mass spectrometry. Samples were prepared for TurboID alone (negative control), DDX3X^WT, and mutations shown in Fig 1B. Unique peptides identified for each sample were normalized to the negative control, resulting in more than 300 significantly enriched proteins for each condition (fold change > 1.5; p-value <0.05; S3 and S4 Tables).

Download:

• PNG
  larger image
• TIFF
  original image

Fig 5. Turbo-ID proximity labeling analysis reveals DDX3X mutations exhibit differential protein interactomes.

A) Schematic of experimental design. Negative control (unfused TurboID) or HsDDX3X-TurboID (WT or missense mutations) were delivered in primary cells. B) Scatter plot of top hits for each sub-category (common, mutations only, severe or mild only, WT-specific, R326H-specific, R376C-specific, R488H-specific, T532M-specific). C) Venn diagram of significant interactors across all conditions. D) Venn diagram of all the significant interactors across the missense mutations and not shared with the WT condition. E-N) Western blots of immunoprecipitations (IPs) for targets highlighted in (B) performed in N2A cells and quantifications. Unpaired two-tailed t-test. *p < 0.01; **p < 0.001; ***p < 0.0001; ****p < 0.0001. Data are mean ± SD. “S”: severe; “M”: mild.

https://doi.org/10.1371/journal.pgen.1011555.g005

Proximity labeling revealed both unique and shared protein interactomes across the different DDX3X variants. We identified 211 protein interactors shared between DDX3X^WT and all missense mutations (Figs 5B and 5C and S3A and S4 Table). Notably, these shared protein partners were primarily enriched for mRNA metabolism categories, including mRNA splicing (GO:0000389) and RNA processing (GO: 0006396). This demonstrates that DDX3X variants lacking helicase activity (DDX3X^R326H and DDX3X^T532M) retain the ability to interact with RNA metabolism machinery (S3A–S3E Fig). Importantly, we also discovered DDX3X interactors previously observed in other cell types and species [26,30,48–54] (S3A Fig). This reinforces the robustness and validity of our datasets. Amongst these interactors was DDX3X itself, with similar fold enrichment across all mutations. This suggests that all mutations preserve the ability to homodimerize (Fig 5B).

We next compared only DDX3X missense variant interactomes to each other. Amongst these, the clinically severe mutations DDX3X^R326H and DDX3X^T532M, showed the highest percentage of unique interactors (around 15–16% compared to 4–8% in the mild mutations) (Fig 5D). Strikingly these mutants had 40 protein interactors in common (Fig 5B and 5D). Gene ontology analysis showed that both DDX3X^R326H and DDX3X^T532M but not DDX3X^WT, interacted with proteins associated with noncoding post-transcriptional gene silencing (GO: 0016441) and regulation of protein localization (GO:0070201) (S4A–S4D Fig). While DDX3X^R488H did not present enough uniquely enriched peptides to perform gene ontology analysis, DDX3X^R376C displayed preferential interaction with proteins involved in DNA homeostasis (S4A–S4D Fig). The composition of the proximity labeling datasets suggests that while all DDX3X missense mutants interact with RNA metabolism proteins, clinically severe mutations acquire new mutation-specific interactors.

To validate the DDX3X protein interactors, we focused on novel DDX3X interactors not previously reported in other cell types. We prioritized interactors with the highest enrichment amongst DDX3X^WT and the four DDX3X missense variants (Fig 5B–5D). DDX3X^WT-specific interactor CFDP1 (Craniofacial Development Protein 1) is associated with cell cycle progression and the neurodevelopmental disorder Williams-Beuren syndrome [55]. Notably, the most enriched interactors of the missense mutants are associated with DNA damage (SMURF2 and RHOB, respectively DDX3X^R326H and DDX3X^T532M interactors) [56,57], stress granules and protein localization (HOOK2 and HSPBP1, respectively DDX3X^R376C and DDX3X^R488H interactors) [58–61]. These highlighted candidates were used to evaluate DDX3X variant-specific interactions in N2A cells.

We first validated interactions by co-expression of Flag-HsDDX3X and mScarlet-tagged interactors in N2A cells and performed co-immunoprecipitation (IP) using the Flag tag (Fig 5E–5N). To test for specificity, for each unique interaction we tested both Flag-DDX3X^WT and the missense mutations (Fig 5E–5N). DDX3X^WT, but not DDX3X^R326H, pulled down CFDP1 (Fig 5E and 5F). Similarly, DDX3X^R326H, DDX3X^R376C, DDX3X^R488H, and DDX3X^T532M mutant proteins each co-precipitated with their respective interactor, but not with WT protein (Fig 5G–5N). These data provide orthogonal validation of DDX3X variants’ unique physical interactions.

We also validated the TurboID results by measuring sub-cellular co-localization in N2A cells co-expressing the GFP-HsDDX3X variant with its mScarlet-tagged interactor. Consistent with previous studies, DDX3X was primarily cytoplasmic with some DDX3X+ punctae in the nucleus (S5C Fig) [9]. In cells expressing DDX3X^R326H, CFDP1 was primarily nuclear showing little overlap with DDX3X. In comparison, in the presence of DDX3X^WT, CFDP1 subcellular distribution was more cytoplasmic (S4E and S4F Fig). Co-expression of DDX3X^R326H and SMURF2 resulted in cytoplasmic sequestration of the latter, a phenotype previously observed for SMURF2 in cancer cells [62] (S4G and S4H Fig). Under pathogenic conditions, HOOK2 is involved in formation of aggresomes, a pericentrosomal accumulation of misfolded proteins [63]. Consistent with this, DDX3X^R376C and HOOK2 co-expression induced the formation of aggresomes [64] (S4I and S4J Fig). Moreover, while RHOB localized to the cell membrane in WT-expressing cells, co-expression with DDX3X^T532M led to a higher percentage of cells with cytoplasmic localization (S4K and S4L Fig). In contrast, DDX3X^R488H did not affect the localization of its top interactor (S4M and S4N Fig). Taken together, these data demonstrate that overexpression of DDX3X variants in neuronal cells can impact subcellular localization of protein interactors. Together this rich dataset highlights new binding partners which may inform the functional impact of DDX3X clinically severe and mild mutations.

DDX3X missense mutations generate stress granules with specific cellular and physical characteristics

These localization studies indicate that DDX3X missense mutations can impact the sub-cellular distribution of DDX3X and their protein interactors. Notably, DDX3X is a component of cytoplasmic stress granules, and overexpression of DDX3X missense variants can induce cytoplasmic granules in neural and immortalized cells in the absence of stress [9,30]. This is relevant, given our finding that DDX3X missense mutations induce DNA damage (Fig 3), a well characterized source of cell stress [65]. Thus, we hypothesized that DDX3X missense mutations drive formation of stress granules in primary neural cells and this may be most prominent for those mutations associated with significant DDR. To test this, we used our lentivirus paradigm to introduce GFP-tagged DDX3X variants into primary neural cells. In the absence of external cell stress, we visualized localization of these variants (Figs 6A and S5B). Notably, DDX3X^WT induced spontaneous granule formation in less than 1% of transduced cells, with an average of 1 granule per cell (Fig 6B and 6C). The clinically severe mutations (DDX3X^R326H and DDX3X^T532M) showed granules in around 10% of transduced cells with a mean of 20 granules per cell. In contrast, the clinically mild mutations (DDX3X^R376C and DDX3X^R488H) showed granules in only about 5% of cells, with an average of less than 10 granules per cell (Fig 6B and 6C). Additionally, the clinically severe mutations formed granules preferentially in neurons compared to WT, while DDX3X^R376C showed the opposite trend, with more granules in progenitors (Figs 6D and 6E and S5A). These results demonstrate that DDX3X missense mutations have different propensity for granule formation, with the clinically severe variants exhibiting higher potency of formation, especially in neurons.

Download:

• PNG
  larger image
• TIFF
  original image

Fig 6. DDX3X missense mutations cause formation of stress granules in primary neural progenitors and neurons.

A) Examples of GFP-HsDDX3X (green) subcellular distribution and stress granule formation in primary cells. Arrows point to examples of individual granules. B) Quantification of percentage of GFP+ cells with granules (n = 150–400 cells across 6 litters). C) Quantification of the average number of granules per GFP+ cell (n = 40–100 cells across 6 litters) D) Quantification of percentage of Tuj1+ or SOX2+ cells with DDX3X granules (N= 74–272 cells across 6 litters) E) Example of Tuj1 (red), SOX2 (magenta), GFP (green) and Hoechst (blue) staining in primary cultures. Arrows indicated SOX2+ cells with granules, arrowheads point to Tuj1+ cells with granules. Scale bars: 10 μm. Two-way ANOVA **p < 0.001; ***p < 0.0001; ****p < 0.0001; ns, not significant. Data are mean ± SD. “S”: severe; “M”: mild.

https://doi.org/10.1371/journal.pgen.1011555.g006

We next characterized the nature of these granules, as cytoplasmic granules induced with extrinsic stressors have distinct protein composition depending on cell types and disease analyzed [26]. Towards this, we characterized DDX3X granules by immunostaining for RNA binding proteins which can be recruited to stress granules (FMRP and G3BP1) [9,66]. Notably, almost a third of DDX3X^R326H and DDX3X^T532M granules colocalized with FMRP while the percentage was almost double for the mild mutations (Fig 7A and 7C). In contrast, neither DDX3X^WT nor any of the missense mutations overlapped with G3BP1 (Fig 7B and 7D). These observations are consistent with overexpression of DDX3X variants in N2A cells [9].

Download:

• PNG
  larger image
• TIFF
  original image

Fig 7. DDX3X missense granules show differential protein composition and physical properties.

A) FMRP (red), GFP (green), Hoechst (blue) staining in primary cells. Inserts show overlap of FMRP and DDX3X in individual granules. B) G3BP1 (red), GFP (green), and Hoechst (blue) staining in primary cells. Inserts show overlap of FMRP and DDX3X in single granules. C) Quantification of FMRP+ DDX3X granules. D) Quantification of G3BP1+ DDX3X granules. E) Schematic depiction of FRAP experiments. Single DDX3X granules were photobleached using a laser and recovery of fluorescence was followed for 20 min. F) Quantification of fluorescence intensity of DDX3X granules during FRAP assay. G) Examples of GFP-HsDDX3X signal before, during and after photobleaching. Scale bars: 10 μm. Two-way ANOVA. *p < 0.01; **p < 0.001; ***p < 0.0001; ****p < 0.0001; ns, not significant. Data are mean ± SD. “S”: severe; “M”: mild.

https://doi.org/10.1371/journal.pgen.1011555.g007

To investigate the physical properties of these granules, we characterized DDX3X granule dynamics using Fluorescence Recovery After Photobleaching (FRAP) (Fig 7E). In this assay we did not observe clear segregation of granule dynamics by clinical severity. Instead, the mutations in different protein regions tended to show similar properties. Mutations in the N-terminal helicase domain (DDX3X^R326H and DDX3X^R376C) showed a partial recovery of granule fluorescence after photobleaching. These showed differential dynamics, with DDX3X^R376C exhibiting a ~60% recovery, and DDX3X^R326H showing only a ~40% recovery (Fig 7F and 7G). This indicates that while both N-terminal mutations display limited granule dynamics, DDX3X^R326H has a lower capability for cytoplasmic exchange. Mutations in the C-terminal DDX3X helicase domain (DDX3X^R488H and DDX3X^T532M), did not show any fluorescence recovery after photobleaching of their granules (Fig 7F and 7G), demonstrating that granules formed by these specific mutations are more solid-like. These results show that DDX3X C-terminal missense mutations induced granules which exhibit reduced molecular exchange with the cytoplasm. The differences observed in the dynamics and composition of these granules thus might contribute to the variable clinical severity associated with DDX3X mutations.

Ethics statement

All animal procedures were approved by the Duke Institutional Animal Care and Use Committee (IACUC) and performed in agreement with the ethical guidelines of the Division of Laboratory Animal Resources (DLAR) from Duke University. We used the previously described mouse line: Dcx::DsRed [37]. The following mouse strains were obtained from Charles Rivers: CD1 (strain 022). For embryo staging, plug dates were defined as embryonic day (E) 0.5 on the morning the plug was identified.

Primary cultures

Primary cortical cultures were derived from E12.5 embryonic dorsal cortices, as previously described in [38]. In brief, tissue was prepared in a single cell suspension by trypsinization (7 min with 0.25% Trypsin-EDTA, Thermo Fisher Scientific, 25200056), and mechanical dissociation by pipetting p200 pipette. 12–16 (from 6–8 embryos) cortices were pooled together and isolated cells were plated in neural progenitor media (DMEM Thermo Fisher Scientific 11965092; B-27 supplement Life Technologies 12587010; N2 supplement Thermo Fisher Scientific 17502048; N-acetyl cystein Sigma A9165; bFGF R&D 3139-FbB-025/CF). Cells were plated 24-well glass-bottom dish (MatTek P24G-1.5-10-F) and given 3h to settle at 37°C in 5% CO₂ prior to lentiviral transduction. After 48hrs from lentiviral transduction, the cells were harvested for analysis.

Live imaging

Dcx::DsRed E12.5 cortices were used to derive primary culture as described above. 48hrs after viral transduction, the neural progenitor media was change. Images were captured every 20 min for 24 h using a 20× magnification on a Zeiss Axio Observer Z.1 equipped with a XL multi S1 incubation chamber, CO₂ module S, temperature module S, and humidity control. The cells were kept 37°C and 5% CO₂ for the duration of the live imaging. Mitosis duration, catastrophe and viability were identified by morphology (rounding of cells and visual identification of condensation of chromatin). Cells were considered neurons if expressing dsRed for at least 4 hours.

Plasmid constructs

All cloning PCR products were amplified by Q5 polymerase (NEB, M049L). For construction of vectors used for lentiviral assays, HsDDX3X^WT and missense mutations were amplified from previously generated plasmid [9] and inserted into the pUbc-EGFP (Addgene # 98916), EGFP or TurboID, by EcoRV enzyme sites. For generating mScarlet constructs, gene of interest were amplified from mouse cDNA and inserted in pCAGGS-mScarlet by BsrI enzyme site. HiFi DNA Assembly Master Mix (NEB, E2621L) was used to aligned cloned inserted and cut plasmids. Sequence was validated by Sanger sequencing.

Lentivirus packaging and transduction

HEK293T cultured in DMEM+10%FBS+1%P/S were used for lentivirus production. 26.75ug pMF58 containing EGFP-HsDDX3X or TurboID-HsDDX3X (DDX3X^WT or missense mutations), 20 µg Packaging plasmid (psPAX2) and 6.25 µg Envelope plasmid (pMD2.G) were transfect with PEI-MAX (Polysciences, 24765-100) when cells reached to 70–80% in 15 cm dish. 48hrs after transfection, the medium was collected and centrifuged at 5000 g × 10 min. The supernatant was then filtered through a 0.45 mm filter (VWR, 28143-352) into an ultracentrifuge tube (Beckman-Coulter, 357448). 4ml sterile 20% sucrose was added below the medium using a 5 ml serological pipette. The mixture was centrifuge at 19700 rpm for 2hrs at 4°C in a swinging bucket rotor (Beckman-Coulter, SW28). The supernatant was then removed, 100 µl of PBS was added, and the virus was resuspended overnight at 4°C with rocking. The resuspended virus was aliquoted, flash-freeze and stored at −80°C. To transduce cells, each lentivirus was titrated to achieve ~30% overexpression and added to the cells in the evening. After 16–18 hours, the media was changed. Nearly all cells were transduced by the following day (S1A and S1B Fig).

Immunofluorescence staining

Cells were fixed in 4% PFA for 10min at room temperature. Permeabilization was done with 1X PBS/0.25% TritonX-100 for 10min and blocked with 5% NGS/PBS for 30min at room temperature. Cells were incubated with primary antibodies overnight at 4°C, and secondary antibodies at room temperature for 1 hours (Alexa Fluor-conjugated, Thermo Fisher, 1:800). The following primary antibodies were used: DDX3X (Sigma, HPA001648, 1:500), SOX2 (Thermo Fisher, 14-9811-82, 1:1000), FMRP (Sigma, F4055, 1:200), G3BP1 (Proteintech, 13057-2-AP, 1:100), TUJ1 (Biolegend, 801202, 1:1000), pH2A.X (Cell Signaling, 9718, 1:200), P53 (Leica, CM5, 1:250), S9.6 (Kerafast, ENH001, 1:100). Slides were mounted with Vectashield (Vector Labs, H-1000-10).

Microscope image acquisition and processing

Images were captured using a Zeiss Axio Observer Z.1 equipped with an Apotome for optical sectioning at 20x (0.8 NA) (S1 Fig), 40x (1.4 NA) (Figs 4A and 4B, and 6E) and/or 63x (1.4 NA) (Fig 4C), alternatively Zeiss LSM 780 confocal microscope using 63x magnification (1.4 NA) (Figs 6A, 7A and 7B, S4E, S4G, S4I, S4K and S4M). Images were acquired using Zen Blue 2.6 (Zeiss Axio Observer Z.1) or Zen Black 2011 (Zeiss LSM 780). 10 fields of view were captured per experiment and all images for a giver experiment were captured with identical exposures. All images were captured as 16 bit and processed using ImageJ. Cells were manually counted (Fiji cell counter).

RNA extraction and RT-qPCR

Cells were centrifuged to make them precipitate and RNA was purified from Trizol (Thermo, 15596026) method. cDNA was prepared using iScript kit (Bio-Rad, 1708891) following manufacturer’s protocol. qPCR was performed using SYBR Green iTaq (Bio-Rad, 1725124). In at least three independent biological samples in a QuantStudio 3 machine (Applied System). Values were normalized to Gapdh as loading control. The following primers were used: Mus musculus Gapdh (Forward 5’-TGGATTTGGACGCATTGGTC-3’ and Reverse 5’-TTTGCACTGGTACGTGTTGAT-3’), Mus musculus Atm (Forward 5’-GCTTCCTCCCGAAATTCCTGT-3’ and Reverse 5’-CCTCTAAAGGGTCCCATTCGT-3’) and Mus Musculus Parp1 (Forward 5’-CTCTGTACTTTGAAAACCACCGT-3’ and Reverse 5’-GCTCAGTCGGACACCATGTA-3’).

Bulk RNA sequencing and analysis

Primary culture cells were precipitated and RNA was purified RNAeasy kit (Qiagen), each experimental condition has 3 biological replicates. cDNA libraries were prepared by Illumina TruSeq stranded mRNA kit. RNAseq libraries were sequenced on the NovaSeq (PE100) with 20M paired end reads. RNAseq libraries were sequenced to a depth of ~40 million total reads per sample. RNASeq data aligned by Star Salmon from FASTAq files, and normalized on “no virus” samples. Differential gene expression (DGE) analysis using DESeq2. DGE lists were defined using an FDR <0.05, with Log2(FC) ≥ 1.5 or ≤ −1.5. Overlapped and specific gene expression changes were defined by Biovinn (https://www.biovenn.nl/). GO analysis were performed on PANTHER [89] and categories were selected by FDR <0.05.

Proximity labeling with TurboID and mass spectrometry

Primary culture cells from E12.5 embryos from 3 CD1 mouse litters (per trial to obtain 1 biological replicate for each lentivirus) were dissected, mixed, and plated such that each condition had the same background primary cells prior to transduction. Cells were transduced with TurboID or TurboID-DDX3X (DDX3X^WT or missense mutations) for 48 hrs. 500uM biotin (Sigma) were added to the medium and incubate 45min at 37°C. Medium was removed and cells were washed with ice cold 1xDPBS 5 times on ice. Then, cells were lysate in 1x RIPA buffer (50 mM Tris-HCl pH = 7.4, 150 mM NaCl, 1% sodium deoxycholate, 2% Triton-X, 0.2% SDS) with protease inhibitors (Roche, 4693132001). Lysates were frozen at −80°C for 3min to further lyse the cells. Lysates were thawed and triturated 5 times with 25G needle, then, spun down at 15000 g for 10min at 4°C. The supernatants were frozen at −80°C. After three biological replicates for each condition were obtained, the supernatants were thawed and incubated with 50 µl of Streptavidin beads (Thermo Fisher, 65601) overnight at 4°C with end-over-end rotation. Beads were magnetized and washed with the following: 2× RIPA buffer, 2× 1M NaCl, 1× 2M urea, 2× 50 mM ammonium bicarbonate. Beads were eluted in 100 µl of elution buffer (25 mM tris, 50 mM NaCl, 10 mM DTT, 2% SDS, 5 mM free biotin) and boiled for 10 min at 98°C. 15 uL of input, supernatant, wash 1, and elution were saved to perform western blotting to confirm the IPs were successful. Elution was sent for mass spectroscopy at the Duke Proteomics and Metabolomics Core Facility. Samples were spiked with undigested bovine casein at a total of either 1 or 2 pmol as an internal quality control standard. Next, samples were reduced (10 mM dithiolthreitol already in samples) for 30 min at 80C, alkylated with 20 mM iodoacetamide for 30 min at room temperature, then supplemented with a final concentration of 1.2% phosphoric acid and 866 μL (depending on sample volume) of S-Trap (Protifi) binding buffer (90% MeOH/100mM TEAB). Proteins were trapped on the S-Trap micro cartridge, digested using 20 ng/μL sequencing grade trypsin (Promega) for 1 hr at 47C, and eluted using 50 mM TEAB, followed by 0.2% FA, and lastly using 50% ACN/0.2% FA. All samples were then lyophilized to dryness. Samples were resolubilized using 12 μL of 1% TFA/2% ACN with 12.5 fmol/μL yeast ADH.

Quantitative LC/MS/MS was performed on 3 μL (25% of total sample) using an MClass UPLC system (Waters Corp) coupled to a Thermo Orbitrap Fusion Lumos high resolution accurate mass tandem mass spectrometer (Thermo) equipped with a FAIMSPro device via a nanoelectrospray ionization source. Briefly, the sample was first trapped on a Symmetry C18 20 mm × 180 μm trapping column (5 μl/min at 99.9/0.1 v/v water/acetonitrile), after which the analytical separation was performed using a 1.8 μm Acquity HSS T3 C18 75 μm × 250 mm column (Waters Corp.) with a 90-min linear gradient of 5–30% acetonitrile with 0.1% formic acid at a flow rate of 400 nanoliters/minute (nL/min) with a column temperature of 55C. Data collection on the Fusion Lumos mass spectrometer was performed for three difference compensation voltages (−40v, −60v, −80v). Within each CV, a data-dependent acquisition (DDA) mode of acquisition with a r = 120,000 (@ m/z 200) full MS scan from m/z 375–1500 with a target AGC value of 4e5 ions was performed. MS/MS scans were acquired in the ion trap in Rapid mode with a target AGC value of 1e4 and max fill time of 35 ms. The total cycle time for each CV was 0.66s, with total cycle times of 2 sec between like full MS scans. A 20s dynamic exclusion was employed to increase depth of coverage. The total analysis cycle time for each injection was approximately 2 hours.

Following UPLC-MS/MS analyses, data were imported into Proteome Discoverer 2.5 (Thermo Scientific Inc.). In addition to quantitative signal extraction, the MS/MS data was searched against the SwissProt M. musculus database (downloaded in Nov 2019), the human DDX3X sequence, and a common contaminant/spiked protein database (bovine albumin, bovine casein, yeast ADH, etc.), and an equal number of reversed-sequence “decoys” for false discovery rate determination. Sequest (v 2.5, Thermo PD) was utilized to produce fragment ion spectra and to perform the database searches. Database search parameters included fixed modification on Cys (carbamidomethyl) and variable modification on Met (oxidation). Search tolerances were 2ppm precursor and 0.8Da product ion with full trypsin enzyme rules. Peptide Validator and Protein FDR Validator nodes in Proteome Discoverer were used to annotate the data at a maximum 1% protein false discovery rate based on q-value calculations. Note that peptide homology was addressed using razor rules in which a peptide matched to multiple different proteins was exclusively assigned to the protein has more identified peptides. Protein homology was addressed by grouping proteins that had the same set of peptides to account for their identification. A master protein within a group was assigned based on % coverage. Prior to imputation, a filter was applied such that a peptide was removed if it was not measured in at least 2 unique samples (50% of a single group). After that filter, any data missing values were imputed using the following rules; 1) if only one single signal was missing within the group of three, an average of the other two values was used or 2) if two out of three signals were missing within the group of three, a randomized intensity within the bottom 2% of the detectable signals was used. To summarize to the protein level, all peptides belonging to the same protein were summed into a single intensity. These protein levels were then subjected to a normalization in which the top and bottom 10 percent of the signals were excluded and the average of the remaining values was used to normalize across all samples. Each of the identified putative interactors of the DDX3X-fused conditions resulting values across the three biological values have been normalized over the negative control condition (TurboID). The normalized values have been used to calculate fold change and p-values (t-test). The significative hits (fold change >1.5, p-value <0.05) of the DDX3X^R326H, DDX3X^R376C, DDX3X^R488H and DDX3X^T532M have been also used to do normalization over DDX3X^WT to analyzed changes between WT protein and missense variations.

Western blot and co-immune precipitations

Cell cultures were lysed at 4°C in TNE buffer (50 mM Tris-HCl pH = 7.4, 137 mM NaCl, 0.1 mM EDTA) containing 0.5% Triton X-100, plus protease inhibitors (Roche). Protein lysates were incubated on ice for 10 min and clarified by centrifugation at 20,000 rcf for 5 min. For co-immunoprecipitations 1% octyl-beta-glucoside was added to the lysis buffer, and samples were incubated for 40 min on ice and clarified by manual spin for 15 sec at 4°C. 0.5–1 ml of each protein lysate was incubated with 50 µl anti-Flag M2 Magnetic beads (Sigma, M8823) overnight at 4°C with shaking. Beads were magnetized and after 5 washes with TNE buffer plus 0.5% NP-40 detergent and protein inhibitors, beads were resuspended in elution buffer (TNE buffer plus 1x laemmli buffer and 50 mM DTT) and boiled 5–10 min at 98°C. For Western blot, 20 ug of protein lysates, plus 1x laemmli buffer and 50 mM DTT, were boiled for 5 min at 98°C. SDS-PAGE was run using 4–20% Mini-PROTEAN TGX precast gels (Bio-Rad, 4568094) and run at 130V for 1 hrs. Blotting was done using the Trans-Blot Turbo Transfer System (Bio-Rad). Blots were all blocked in 5% milk in TBST (TBS buffer, 0.1% tween detergent). Primary antibodies were incubated overnight at 4°C shaking. The following day, they were rinsed 3 times 10 min in TBST buffer, incubated with secondary antibody in TBST buffer for 1 hr at room temperature and rinsed 3 times 10 min in TBST buffer. Western imaging was done using ECL Western blotting substrate (Thermo Scientific, 32106) supplemented occasionally with SuperSignal West Femto Maximum Sensitivity Substrate (Thermo Scientific, 34095). Blots were exposed in a Bio-Rad Gel Doc XR system. Quantification of Western Blot was performed using Bio-Rad Image Lab software. Primary antibodies used: DDX3X (Sigma, HPA001648, 1:1000), FLAG-HRP (Sigma, A8592, 1:5000), RFP (Rockland, 600-401-379, 1:500).

Fluorescence recovery after photobleaching (FRAP)

For FRAP of the live cells, cells expressing EGFP-HsDDX3X were cultured in 24well plate glass-bottomed dishes (Mattek, P24G-1.5-10-F). The FRAP assays were conducted using the bleaching module of the Zeiss LSM 780 confocal microscope using 63x magnification. The 488 nm laser was used to bleach the EGFP signal. Bleaching was focused on a circular region of interest (ROI) using 100% laser power. Time lapse images were collected every 30 sec starting 90sec before the bleaching for a total time of 20 min. A same-sized circular area away from the bleaching point was selected as an unbleached control. The fluorescence intensity was measured in Fiji.

Quantification and statistical analysis

Experimenters were blinded at the analysis stage and all analyses were performed blindly by 1 or more investigators. Sample sizes were based on previous experiments [9,34,38]. No data was excluded from analysis. Number of data points and statistical tests used for all the comparisons are indicated in the figure legends.