DHX9 resolves G-quadruplex condensation to prevent DNA double-strand breaks

doi:10.21203/rs.3.rs-2184718/v1

Download PDF

Article

DHX9 resolves G-quadruplex condensation to prevent DNA double-strand breaks

https://doi.org/10.21203/rs.3.rs-2184718/v1

This work is licensed under a CC BY 4.0 License

Version 1

posted

You are reading this latest preprint version

DNA G-quadruplexes (G4s) structures are abundantly present in mammalian genomes and correlated with genome instability. However, the mechanism by which G4s are timely resolved remains unknown. Here, we report that DHX9 functions as a resolvase to unwind G4s globally in activated B cells. DHX9-deficient B cells show gross DNA double-strand breaks at the accumulated G4 sites, which are clustered together and form liquid condensates. We demonstrate that DHX9 also undergoes phase separation and fuses with G4 condensates for the productive unwinding of G4s in an ATP-dependent manner. Physiologically, G4-accumulation-induced DNA breaks can promote immunoglobulin class-switch recombination for producing high-affinity antibodies. Surprisingly, the DHX9Y1189C mutant identified in Hashimoto’s thyroiditis patients shows compromised phase separation and G4 unwinding abilities, causing elevated DNA damage and abnormal antibody production. Our findings suggest a DHX9-dependent G4 condensation-resolving mechanism to prevent DNA damage in mammalian cells. Disrupting this homeostasis may induce autoimmune disorders and lymphoid malignancies.

Biological sciences/Molecular biology/Epigenetics

Biological sciences/Immunology/Immunogenetics

G-quadruplex

DHX9

Class switch recombination

DNA double-strand breaks

Phase separation

The mammalian genomes can generate various noncanonical structures during replication and transcription, such as G-quadruplexes (G4s), R-loops, and intercalated motifs (i-motifs)^1-4. G4s tended to form at G-rich single-stranded DNA (ssDNA) sequences through G-G Hoogsteen base pairing and were initially proposed during the study of immunoglobulin (Ig) switch region sequences in vitro⁵. DNA double-strand breaks (DSBs) at or around G-rich tandem repeats are prerequisites for productive class switch recombination (CSR) in activated B cells, leading to the replacement of the original production of IgM antibodies with that of IgG, IgA, or IgE antibodies^6-9. In addition to this physiological role in antibody diversity, approximately 70% of the aberrant chromosomal recombination sites in lymphoid malignancies were predicted to enrich with G4 motifs¹⁰. Moreover, computational predictions in various eukaryotic genomes showed that G4 motifs were correlated with genomic instability^11-14. Therefore, the timely unwinding of these stable G4 structures seems crucial for maintaining the integrity of the mammalian genomes.

G4-seq analysis of in vitro refolded ssDNA identified over 700,000 potential G4 sites in the human genome¹⁵. However, genome-wide in vivo mapping of G4 by ChIP-seq with a BG4 antibody only identified ~10,000 sites¹⁶. The small number of detected G4s may be due to the presence of nuclear helicases that can quickly resolve these structures. Although several eukaryotic helicases showed G4 unwinding activity in vitro^17-22, it remains unclear whether they could unwind G4 structures globally in living cells. DHX9 (DExH-box helicase 9) belongs to the DExH-box helicase family and contains a helicase core domain²³. An in vitro assay revealed that DHX9 could unwind multiple nucleic acid secondary structures, including R-loops and G4s^24,25. However, whether and how DHX9 modulates these noncanonical DNA structures, especially G4, within the nucleus is unknown.

As G4 structure is assumed to be prevalent at class switch regions, B cells emerge as an attractive model to study the physiological and pathological roles of G4 during antibody diversification. After activating B cells by an antigen, the activation-induced cytidine deaminase (AID) will be dramatically induced to create mutations at two specific regions of the antibody locus by deamination of cytosine bases to uracils in the ssDNAs^26-28. The resulting U: G mismatches are subsequently converted into DSBs via the base excision repair and mismatch repair pathways. These two DNA breaks are then joined together to make a new form of the antibody gene⁷. An in vitro study revealed that the G4 structure is the preferred AID substrate over ssDNA²⁹. To explain AID targeting specificity, we previously identified many AID-associated proteins by coimmunoprecipitation using a specific antibody³⁰; among the identified proteins, DHX9 attracted our attention because of its implication in autoimmune disease³¹.

In this study, we initially aimed to investigate the functional mechanism of DHX9 in antibody diversification. This investigation led to the unexpected finding that DHX9 knockout in activated B cells dramatically increases CSR efficiency by inducing DSBs at G4 accumulation sites, which are spatially close to each other and form G4 condensates. We further demonstrate that DHX9 can undergo phase separation within the nucleus and fuse with the G4 condensate to efficiently resolve these noncanonical DNA structures in an ATP-dependent manner. Such a mechanism is crucial for maintaining genome stability in B cells. Notably, the DHX9^Y1189C mutant identified in Hashimoto’s thyroiditis patients compromised its phase separation and G4 resolving abilities, leading to gross DNA damage and elevated CSR, which subsequently induce autoimmune disease.

DHX9 represses CSR in B Cells

To study the functional mechanisms of DHX9 in antibody diversification, we first generated a conditional knockout mouse model by inserting two loxP sites at intron 2 and intron 4 of DHX9 for Cre/loxP-mediated deletion (Fig. 1a). The ablation of exons 3 and 4 will introduce a stop codon in exon 6 and trigger the mRNA degradation by nonsense-mediated RNA decay. Next, we crossed the floxed mice with Aicda-Cre mice that express Cre recombinase under the control of the AID promoter. This crossing strategy allowed us to delete DHX9 in activated B cells specifically (Fig. 1a). Southern blotting and genotyping analysis confirmed the successful insertion of the loxP sites at the intronic regions of DHX9 (Extended Data Fig. 1a-b). After Cre-mediated recombination, genotyping with primers targeting upstream and downstream sequences of the loxP sites confirmed the complete deletion of exons 3 and 4 in germinal center B cells isolated from the Peyer's patch (PP) of DHX9 conditional knockout mice (cKO, DHX9^loxp/loxp/Aicda^Cre+/-; Fig. 1b). RT‒qPCR analysis of the DHX9 RNA levels in the isolated germinal center B cells from cKO mice revealed approximately 90% reduction compared to wild-type (WT) cells (Fig. 1c). Western blot further confirmed the absence of DHX9 protein in the cKO cells (Fig. 1d).

Similar to the WT littermates, DHX9 cKO mice were viable and fertile. After examining the serum antibody titer using blood collected from the retro-orbital plexus by ELISA, we unexpectedly found that the levels of IgA, IgG1, and IgG3 were increased by 40%-70% in unchallenged cKO mice compared to WT mice, whereas the IgM titers were not changed (Extended Data Fig. 1c). Consistently, the populations of surface IgG3⁺and IgA⁺ splenic memory B cells, splenic plasma cells, and PP B cells were also increased by at least 0.7-fold in DHX9 cKO mice compared to WT mice (Extended Data Fig. 1d-e). These results indicate that DHX9 knockout can cause elevated isotype switching in naïve mice. To determine the antibody-switching response to antigens, we challenged mice with a T-independent antigen LPS for 10 days. We measured the IgG3⁺ and IgA⁺ populations in activated B cells from the spleen or PP of WT and DHX9 cKO mice. We found that both IgG3⁺and IgA⁺ populations increased by at least 1.3-fold in DHX9 cKO mice compared with that in WT mice (Fig. 1e-f)., These results collectively reveal an unexpected role of DHX9 in repressing CSR in activated B cells.

DHX9-deficient B cells show accumulated G4s and DSBs

Noncanonical DNA structures such as R-loop and G4 have been proposed to be enriched at switched regions of the IgH locus for introducing DSBs³², which is required for joining the two break regions together to complete a CSR process (Fig. 2a). We speculate that DHX9 may repress CSR in activated B cells by modulating R-loops or G4 DNA structures. R-loops are nucleic acid structures composed of an RNA–DNA hybrid and a non-template single-stranded DNA, which can be monitored by immunofluorescence (IF) using the S9.6 antibody³³. Our IF staining showed that the S9.6 foci numbers and intensities were comparable in LPS-activated splenic B cells isolated from WT and cKO mice (Fig. 2b). The staining results were further confirmed by global mapping of R-loops with single-stranded DNA‒RNA immunoprecipitation sequencing³⁴ method (ssDRIP-seq; as shown in Extended Data Fig. 2a). Meta-analysis of the ssDRIP-seq data revealed that the R-loop levels were barely changed at the promoter and gene body regions in LPS-activated splenic B cells collected from WT and cKO mice (P > 0.1, Extended Data Fig. 2b). Next, we examined DNA G4 structures in LPS-activated splenic B cells. After removing RNA G4s by RNase A treatment, IF staining with a G4-specific BG4 antibody revealed that DHX9 ablation increased the number and intensity of G4 structures by at least 2-fold (Fig. 2c), indicating that DHX9 might resolve DNA G4 rather than R-loops to regulate CSR.

As DSBs are a prerequisite for CSR, we examined whether DHX9 ablation can induce DSBs. By monitoring the expression of several hallmark proteins of DSBs in DHX9-ablated B cells, including histone H2AX phosphorylation on serine 139 (gH2AX), P53 binding protein 1 (53BP1), and Rad51. We found that the numbers and density of gH2AX, 53BP1, and Rad51 foci were increased by at least 2-fold in DHX9-deficient B cells compared with WT cells upon LPS stimulation (Fig. 2d and Extended Data Fig. 2c-d). Western blot also showed the elevated expression of these DNA damage marks in DHX9 cKO cells (Fig. 2e). Moreover, we performed comet assay, a gel electrophoresis–based method for measuring DNA damage at the single-cell level, to examine DNA breaks upon DHX9 depletion. Compared to the positive control of H₂O₂treatment, the DNA tail moment length and proportion of DNA damage in DHX9-ablated B cells were increased to 2.1-fold and 3.3-fold, respectively (Fig. 2f-g). In contrast, the tail moment was barely detectable in the WT counterparts. These results demonstrate that DHX9 ablation induces G4 accumulation and gross DSBs in activated B cells.

G4 accumulation at enhancers and promoters in DHX9-deficient B cells

To determine the precise location of the accumulated G4s in DHX9-deficient B cells, we adapted a state-of-the-art method, G4 chromatin sequencing (ChIP-seq), to map the global landscape of DHX9-regulated G4 structures in the mouse genome¹⁶. In brief, LPS-activated B cells were first crosslinked with formaldehyde to fix DNA‒protein interactions, and chromatin was then extracted and sheared by sonication (Fig. 3a). After removing RNA G4s by RNase A treatment, the G4-specific BG4 antibody was applied to the soluble chromatin to specifically pulldown G4 DNA fragments. The G4 DNA fragments were subsequently purified and used to construct a high-throughput sequencing library with an ssDNA adaptable Accel-NGS 1S Plus DNA Library Kit (Fig. 3a).

Using the modified G4 ChIP-seq method, we mapped genome-wide G4 sites in LPS-activated B cells with or without DHX9 expression (Fig. 3a). Two highly reproducible biological replicates were generated for each condition (Extended Data Fig. 3a). After combining these replicates, we identified 6,181 and 9,361 high-confidence G4-containing genes in WT and cKO cells, respectively (Extended Data Fig. 3b). We noted that approximately 85% of the genes in WT cells overlapped with the genes in cKO cells, and 83.6% of them showed elevated G4 signals upon DHX9 deletion (Extended Data Fig. 3b and Supplementary Table 1). Importantly, G4 signals in the Sm and Sg3 regions of IgH loci were also increased by 90% compared to that in WT mice upon DHX9 ablation (Fig. 3b). Notably, most G4 peaks in the IgH loci were enriched at Em enhancer and 3’ regulatory region (3’ RR) super-enhancer (Fig. 3b).

We next explored the features of those G4 ChIP-seq peaks, and found that their average lengths were 278 nt (Extended Data Fig. 3c). Each G4 peak contained 4.4 putative G-quadruplex sequences, a computationally defined G4 motif (Extended Data Fig. 3d). In addition, the identified G4 sites seemed meaningful because: 1) The GC content in G4 ChIP-seq peaks was significantly higher than that in the randomly selected size-matched controls (40%), and the percentage plateaued at 60% at the center of G4 peaks in DHX9-deficient B cells (Extended Data Fig. 3e); 2) Our G4 ChIP-seq mapping also identified several well-known G4 sites, such as Cdipt, Pvt1, and Telo2 (Extended Data Fig. 3f); 3) We observed an overrepresented GC-rich consensus motif in both the WT and cKO peaks (Fig. 3c); and 4) As expected¹⁶, approximately 35% of the G4 peaks were enriched at promoter regions, but we unexpectedly found that approximately 44% of the G4 structures were located at enhancers (Fig. 3d). Collectively, these results demonstrate that G4s prefer to form at enhancer and promoter regulatory elements in B cells.

DHX9 resolves G4 structures in activated B cells

Consistent with our IF staining results (Fig. 2c), we observed a global increase in G4 signals at the transcription start site (TSS) of 73.8% of the genes upon DHX9 ablation (Fig. 3e). Specifically, in addition to the known G4 sites at the promoter regions of the Cdipt, Pvt1, and Telo2 genes, numerous newly identified sites, such as the Atxn2l, Jmjd5, Smg1, and Sh2b1 loci, exhibited accumulation of G4 structures at enhancer and promoter regions in DHX9-deficient B cells (Extended Data Fig. 3g). These results indicate that DHX9 may resolve G4 structures timely in activated B cells to prevent its accumulation.

To examine whether DHX9 showed any binding preference for G4 structures, we synthesized an Sm ssDNA fragment and annealed it into a linear or G4 structure in LiCl- or KCl-containing buffer, respectively. Circular dichroism spectroscopy showed that the Sm fragment formed a mixed conformation of trans- and cis-G4s (Extended Data Fig. 3h). A pulldown assay of B cell lysate using biotinylated Sm-G4 and Sm-linear probes revealed that DHX9 specifically bound G4 rather than linear structures (Fig. 3f and Extended Data Fig. 3i). This preference was similar to that of the positive control AID. Electrophoretic mobility shift assay (EMSA) further confirmed that DHX9 preferentially bound the Sm-G4 probe, with a K_D value of 66.1 nM, and showed little minimal binding to Sm-linear and Sm-mutant probes (Fig. 3g). These results demonstrate that G4s are the preferred substrates of DHX9.

In addition to DHX9, several helicase-core-domain-containing helicases also possess G4 resolving activity in vitro (Extended Data Fig. 4a), such as Wrn, Pif1, Fancj, Blm, and DHX36^35,36. However, only DHX9 has an RGG domain, a structural feature that can bind G4^23,37. In addition, using the reported RNA-seq data generated from naïve and activated B cells³⁸, we found that DHX9 is a top expressed helicase in both naïve and activated B cells (Extended Data Fig. 4b). Moreover, upon knocking down these helicases individually in CH12F3 mouse B cells with shRNAs (Extended Data Fig. 4c), the G4 staining foci numbers and intensities were only accumulated after depletion of Blm and DHX9 (Extended Data Fig. 4d). Because G4 foci numbers and intensities in DHX9-depleted cells were 1.5-fold higher than that in Blm-depleted cells (Extended Data Fig. 4d), we thus concluded that DHX9 is a major G4 resolvase in B cells.

G4 accumulation induces DSBs in B cells

The globally accumulated G4 structures and DNA DSBs in DHX9-deficient B cells prompted us to hypothesize that unresolved G4 structures may directly cause DNA damage. To test this hypothesis, we first used the G4 structure-specific small molecule pyridostatin (PDS)³⁹ to stabilize G4 in B cells and then examined the status of DNA damage. After treatment with PDS for 72 h, IF staining showed that G4 and γH2AX foci were massively accumulated in B cells (Fig. 4a, b). Moreover, Western blot analysis confirmed the elevated expression of the DSB markers γH2AX and 53BP1 (Fig. 4c). These results indicated that stabilizing G4s can induce DSBs in B cells. Next, we compared the G4 sites revealed by ChIP-seq with the DSB sites identified by Rad51 ChIP-seq in 53BP1^−/−Igκ AID B cells⁴⁰, in which the repair of AID-induced DSBs was switched from nonhomologous end joining to homologous recombination. Therefore, the binding of Rad51 could be used as an indicator of initial DSBs.

By examining the distribution of Rad51 ChIP-seq signals, we found that DSB sites were indeed preferentially located at G4 accumulation sites in WT and DHX9-deficient B cells (Fig. 4d). Among the 9,361 G4-containing genes identified in B cells by G4 ChIP-seq, 55% of them were overlapped with the Rad51 ChIP-seq targets (Extended Data Fig. 5a). Importantly, the overlapped Rad51 and G4 ChIP-seq targets tend to have significantly higher Rad51 ChIP-seq read densities than the nonoverlapping target genes (Extended Data Fig. 5b). Moreover, the strength of G4 ChIP-seq peaks was also positively correlated with the Rad51 ChIP-seq signals (Extended Data Fig. 5c). These data collectively suggest a potential causal relationship between accumulated G4s and DSBs.

Accumulated G4 sites are in spatial proximity

Having confirmed that DSBs tend to occur at accumulated G4 sites, we next explored the spatial proximity of these G4 sites within the nucleus. By examining circular chromosome conformation capture sequencing (4C-seq) data generated from activated splenic B cells⁴¹, we found that G4 structure accumulated sites in DHX9-deficient B cells usually overlapped with Rad51 ChIP-seq-identified DSBs and were spatially close to IgH locus (Fig. 4e, Supplementary Table 2). For example, at the c-MYC locus, which tends to translocate into the vicinity of the IgH locus in Burkitt's lymphomas^42,43, we observed extensive DSB signals and accumulated G4 peaks in promoter and gene body regions upon DHX9 depletion (Fig. 4f). Moreover, we found that 57% of the IgH locus proximally interacting genes determined by 4C-seq (P ≤ 0.001) had prominent G4 ChIP-seq peaks, whereas only 26% of the noninteracting genes had these signals (Fig. 4g), indicating that G4 structure containing genes tend to be clustered into proximity within the nucleus.

Next, we divided all the genes into three groups: genes containing Rad51 peaks that proximally interacted with the IgH locus (group 1), genes containing Rad51 targets but not interacting with the IgH locus (group 2), and other genes (group 3). We found that group 1 genes had a stronger G4 ChIP-seq signal than another two group genes (Extended Data Fig. 5d). For example, the proximal interactions between the IgH Sm and Pim1, Il4ra, CD79b, Arid5a, PPP1r16a, Pchsd2, Nsmce1, Srebf2, Mcl1, and Vars loci all showed accumulated G4 structures in DHX9-deficient B cells (Extended Data Fig. 5e). For validation, we determined the relative spatial distance between the IgH Sm and MYC or Pim1 loci. DNA fluorescence in situ hybridization (FISH) showed that the Sm and MYC loci were in close spatial proximity in LPS-activated splenic B cells. The Sm and Pim1 loci were also colocalized (Fig. 4h). Moreover, all three investigated loci overlapped with G4 staining foci in the nucleus (Fig. 4h). Collectively, these results demonstrate that sites of G4 accumulation in proximity induce DSBs and provide an environment conducive to chromosomal rearrangements.

DHX9 phase separates within the nucleus

As DHX9 immunostaining showed punctate and spherical foci (Fig. 2b), we next tested whether it could undergo liquid‒liquid phase separation (LLPS) within the nucleus for unwinding G4 DNA. PONDR algorithm prediction identified three intrinsically disordered regions (IDRs), a typical feature of proteins that drives phase separation⁴⁴, in the N-terminus, middle part, and C-terminus of DHX9 (Fig. 5a). We next purified and incubated the full-length DHX9 protein in a physiological buffer. The solutions were clear at 4 °C but became turbid after warming to 37 °C for as little as 10 min and then became clear again after cooling to 4 °C for 1 h (Fig. 5b). This warming-induced LLPS is similar to a lower critical solution temperature phase separation.

To characterize the properties of DHX9 phase separation, we purified GFP-DHX9 fusion proteins from transfected HEK293F cells. Time-lapse imaging showed that the small droplets gradually fused into larger droplets (Fig. 5c and Supplementary Movie 1), suggesting that the DHX9 phase is in a liquid-like state. The phase separation diagram showed that DHX9 failed to undergo phase separation when the concentrations of NaCl exceeded 300 mM (Extended Data Fig. 6a). A wetting assay performed on a glass slide further revealed the liquid spreading activity of DHX9 phase separation (Extended Data Fig. 6b and Supplementary Movie 2). Within the spherical droplet, DHX9 proteins were also freely diffused, as the photobleached area recovered almost completely within 10 min, as determined by time-lapse imaging (Fig. 5d). To test whether endogenous DHX9 protein can form droplets, we inserted a GFP or mScarlet cassette downstream of the ATG start codon to generate two stable cell lines that expressed either GFP-DHX9 or mScarlet-DHX9 at physiological levels (Extended Data Fig. 6c). Genotyping and Western blot analysis confirmed that the insertions were successful (Extended Data Fig. 6d, e). Confocal live-cell imaging revealed that both GFP-DHX9 and mScarlet-DHX9 formed punctate and spherical foci (Extended Data Fig. 6f), which recovered rapidly after photobleaching (Fig.5e). Collectively, these results demonstrate that DHX9 can undergo phase separation both in vitro and in vivo.

DNA G4s phase-separate in vitro

G4 staining in naïve and activated B cells consistently revealed dozens of spherical and punctate foci within the nucleus (Fig. 2c and 4h). This phenomenon raised an intriguing possibility that accumulated G4 DNA may also undergo LLPS transition to form these spherical droplets. To test this hypothesis, we generated different lengths of G4-forming ssDNA fragments from the Sm region that contained tandem (GAGCT)_nGGGGT repeats, where n can range between 1 and 7. An in vitro phase separation assay showed that the G4-containing fragments indeed condensed into liquid droplets under physiological pH and salt conditions; the droplet sizes and numbers were remarkably increased as the number of G4s in the fragment increased (Fig. 5f). As a control, a non-G4-forming fragment with similar size of 1,900 nt from the IgH locus failed to form liquid droplets (Extended Data Fig. 7a). These results highlight the importance of multivalent G4 structures in promoting phase separation.

To evaluate the properties of G4 DNA condensation, we labeled the 1,900 nt Sm and 200 nt Myc (another well-known G4 site) ssDNAs with Alexa Fluor 647 and Alexa Fluor 488, respectively. These labeled ssDNA fragments initially formed smaller liquid droplets in vitro but gradually fused into larger spherical droplets at a longer incubation time (Extended Data Fig. 7b). The G4 phase formed in vitro by Sm or Myc fragments could fuse with itself, indicating that these G4 condensates were in a liquid-like state (Extended Data Fig. 7c, Supplementary Movie 3 and 4). Moreover, the Sm-G4 phase quickly merged with the Myc-G4 phase within 10 seconds (Extended Data Fig. 7d), highlighting the liquidity state. A fluorescence recovery after photobleaching (FRAP) assay of the Sm phase, Myc phase, and Sm/Myc phase revealed that the Sm/Myc phase recovered more rapidly than the individual Sm and Myc phases (Extended Data Fig. 7e), indicating that the intermolecular interactions between different G4s can influence the molecular fluidity of the condensates.

Different G4 phases can fuse within the nucleus

Next, we evaluated the properties of G4 condensation in vivo. We transfected fluorescently labeled Sm- and Myc-G4 ssDNA into CH12F3 B cells and found that these G4-containing ssDNAs also formed spherical droplets within the nucleus and that Sm- and Myc-G4 condensates could fuse together (Extended Data Fig. 7f). Upon photobleaching, Sm-G4 condensates displayed rapid fluorescence recovery within 60 s (Extended Data Fig. 7g), indicating that DNA G4 condensates also exhibit dynamic liquid-like behavior within B cells. As IgH and Myc loci all contained extensive G4 structures, we next performed FISH with probes targeting G4 surrounding regions in IgH and Myc loci to examine their localization. We found that the IgH-G4 locus was indeed colocalized with the Myc-G4 locus in activated B cells (Extended Data Fig. 7h). These findings raised the intriguing possibility that many G4-containing genes may be located close to each other to form 20-30 phase-separated foci within the nucleus (Fig. 2c and 4h). These results suggest that the spatial proximity and accumulation of G4s may collectively promote the formation of G4 condensates in B cells.

DHX9 unwinds G4 DNA in a phase- and ATP-dependent manner

To understand how DHX9 dynamically binds G4, we mixed 40 mM GFP-DHX9 fusion proteins and 10 mM Cy5-labeled Sm DNA fragments in HEPES buffer in the absence of ATP and monitored the phase separation status for 0.5 h (Fig. 5g). Consistent with the preferential binding of DHX9 to G4 structures (Fig. 3g), GFP-DHX9 and Sm-G4 quickly formed fused protein‒DNA droplets, and importantly, those small droplets gradually fused into larger droplets with increased fluorescence intensity and larger equivalent diameter after longer incubation times (Fig. 5g and Supplementary Movie 5).

Having demonstrated that DHX9 or accumulated G4 forms a liquid-like phase within the nucleus, we next investigated how DHX9 accesses the DNA phase and unwinds G4 structures. First, we examined how DHX9 unwinds G4 DNA by adding ATP to the phase separation buffer. Surprisingly, the fused DNA‒protein droplets became irregularly shaped, and the G4 phase was gradually repulsed from the protein phase, which exhibited smaller but still spherical droplets (Fig. 5h). In contrast, when a nonhydrolyzable analog of ATP, adenylyl-imidodiphosphate (AMP-PNP), was added to the phase separation buffer, these phenomena were not observed (Fig. 5h). These results demonstrate that DHX9 may unwind G4 structures by fusing with the G4 DNA phase and then repelling the unwound ssDNA from the protein phase in an ATP-dependent manner.

Hashimoto’s thyroiditis-associated DHX9^Y1189C mutant fails to phase separate

Mutation or aberrant expression of DHX9 is correlated with systemic lupus erythematosus (SLE) and myelodysplastic syndrome (MDS)³¹. Aiming to identify recurrent mutations in DHX9 from 320 enrolled potential MDS patients, we noted 15 patients with a heterozygous mutation (c.306-308del) and two patients with a novel heterozygous mutation (p.Y1189C, both in women, 44/45 years old); these mutations were located in the IDR1 and IDR3 regions, respectively (Fig. 6a and Extended Data Fig. 8a). Notably, the same mutation was also found in the human dbSNP database (rs1370391733, Extended Data Fig. 8b). As IDR3 alone could drive phase separation and its deletion dramatically reduced the phase separation ability of DHX9 (Fig. 6b, c), we focused on the p.Y1189C mutation for further analysis.

In contrast to the patients in the healthy (WT) and c.306-308del cohorts, the patients with the p.Y1189C mutation exhibited significantly elevated serum titers of IgG3 and IgE (Fig. 6d), indicating more efficient CSR in B cells. Blood tests showed that the levels of anti-thyroglobulin antibodies (TGAb) and anti-thyroid peroxidase antibodies (TPOAb) in the patients with the p.Y1189C mutation were increased at least 12-fold compared with those in the healthy controls (WT, Extended Data Fig. 8c). The ratios of CD4 Th1/Th2 and CD8 Tc1/Tc2 cells in the patients were also increased by 2.6- or 3.5-fold, respectively (Extended Data Fig. 8c). These results collectively support the diagnosis of these two patients with Hashimoto’s thyroiditis, an autoimmune disorder in which the thyroid gland is progressively destroyed.

To examine whether the Y1189C mutation influences the phase separation ability of DHX9, we knocked Y1189C into the GFP-DHX9 allele in CH12F3 cells, with c.306-308del as a negative control (Extended Data Fig. 9a). Western blot analysis showed that neither of these mutations affected the expression of DHX9 (Extended Data Fig. 9b). However, we observed a dramatic reduction in the phase separation ability of the DHX9^Y1189C mutant compared with the c.306-308del and WT controls (Fig. 6e). In addition, the condensation driven by the Y1189C mutant exhibited a slower FRAP recovery rate (Extended Data Fig. 9c). Next, we examined whether the Y1189C mutant influences the formation of G4 or gH2AX foci. Upon stimulation of the CH12F3 knock-in cell lines with CIT (containing CD40L, IL-4, and TGFb) for 72 h, we found that the G4 and gH2AX foci numbers were increased by at least 2-fold in Y1189C mutant cells compared to WT cells (Fig. 6f). Notably, the intensity of the G4 signal was highly correlated with the intensity of the gH2AX signal. These results demonstrate that Y1189C compromises the phase separation ability of DHX9.

DHX9^Y1189C is a loss-of-function mutation

Having confirmed that the p.Y1189C mutation had the same effect as DHX9 ablation on DNA damage and G4 formation, we next investigated whether the Y1189C mutant can boost antibody production in CH12F3 cells. Again, the IgA concentration in the culture medium of Y1189C mutant cells was significantly higher than that in the culture medium of WT and c.306-308del control cells upon CIT stimulation (Extended Data Fig. 9d). Importantly, even without CIT stimulation, Y1189C mutant cells still showed elevated antibody concentrations (Extended Data Fig. 9d). This is reminiscent of the antibody titers in DHX9-ablated naïve B cells. These results demonstrate that the Y1189C mutant compromises the phase separation ability of DHX9 and therefore induces the accumulation of G4s and DSBs, which further promote the CSR occurrence and elevated antibody levels.

Similar to DHX9 cKO, the p.Y1189C mutation in CH12F3 cells and patients led to increased CSR and antibody production, indicating that the p.Y1189C mutation is a loss-of-function mutation. As the observed p.Y1189C mutation in two Hashimoto’s thyroiditis patients is heterozygous, we thus checked the phenotypes of heterozygous DHX9 cKO mice. The heterozygous mice exhibited significantly elevated serum levels of IgG3 and IgE (Extended Data Fig. 9e). Blood tests also showed that the levels of TGAb and TPOAb in the heterozygous DHX9 cKO mice were increased to 1.5-fold compared with those in WT (Extended Data Fig. 9f). Additionally, the ratios of CD4 Th1/Th2 and CD8 Tc1/Tc2 cells in heterozygous mice were also increased by 1.6- or 3.8-fold, respectively (Extended Data Fig. 9f). These results indicate that heterozygous DHX9 cKO mice have phenotypes similar to those of patients with the p.Y1189C mutation.

DHX9^Y1189C fails to unwind G4 in protein‒DNA fused condensations

We next explored how the Y1189C mutation causes G4 accumulation. We first examined whether the Y1189C mutation affects the structural integrity of the DHX9 protein. Circular dichroism spectroscopy analysis showed that the Y1189C mutant proteins have a different light absorption profile compared with WT proteins, indicating the secondary structure of mutant proteins might be changed for unknown reasons (Extended Data Fig. 10a). In addition, the droplets formed by Y1189C mutant proteins were 2-fold smaller than those formed by WT proteins (Extended Data Fig. 10b). Considering the cysteine amino acid introduced by the Y1189C mutation may generate new disulfide bonds to influence the phase separation ability of DHX9. We, therefore, examined the droplet behavior of DHX9 or Y1189C mutant proteins in the presence or absence of DTT. We found that the droplet size and numbers of WT and Y1189C mutant proteins were not influenced by DTT (Extended Data Fig. 10b). Thus excluding the possibility that altered phase separation behavior between WT and mutant proteins was due to the introduced additional cysteine.

Next, we examined whether the Y1189C mutation alters the phase fusion and repulsion behaviors of WT proteins with Sm-G4 in vitro (Fig. 5h). After ATP was added to the phase separation buffer, the spherical shape of the Y1189C-G4 phase became irregular after 10 min of incubation, and the Y1189C mutant still bound tightly to G4s (Fig. 6g). This phase behavior of the Y1189C mutant protein was dramatically different from that of the WT DHX9 protein, which gradually repelled the unwound Sm-G4 ssDNA from the protein‒DNA phase (Fig. 5h). These results suggest that the Y1189C mutant may disrupt the G4 DNA unwinding activity of DHX9.

To test whether the Y1189C mutation influences DHX9 binding affinity to G4 structures, we performed a gel shift assay and found that the binding of the Y1189C mutant to Sm-G4 was not affected and that its binding constants were similar to those of the WT protein (Extended Data Fig. 10c). Notably, in the presence of ATP, the WT protein could efficiently unwind G4 structures into linear sequences, but the Y1189C mutant lost this ability, for unknown reasons (Fig. 6h). Next, we studied the helicase activity of DHX9 or different mutants in phase separation buffer using CD spectroscopy. Our results showed that the helicase activity of the Y1189C mutant appeared to be significantly reduced compared with WT proteins in the presence of ATP (Extended Data Fig. 10d). The behavior of the Y1189C mutant is similar to the K417R catalytic dead mutation that completely abolishes the ATP-dependent helicase activity of DHX9⁴⁵. Moreover, in the presence of AMP/PNP, an analogy of ATP, the WT, Y1189C, and K417R mutant proteins showed similar light absorption profiles (Extended Data Fig. 10d). These results indicated that the failure of the Y1189C mutant to suppress CSR in vivo is most likely due to its compromised abilities for phase separation and G4 unwinding. It will be interesting in the future to study how a single amino acid change influences the helicase activity of DHX9 using advanced structure tools. These results suggest that the Y1189C mutation may induce Hashimoto’s thyroiditis by activating uncontrolled CSR in the naïve state and overactivating CSR after antigen stimulation.

In this study, we showed that DHX9 functions as a novel resolvase to unwind G4 structures globally at enhancer and promoter regions within the nucleus. Mechanistically, we demonstrate that DHX9 undergoes phase separation via the IDR3 region during this process and then fuses with DNA G4 condensates formed between spatially proximal G4 sites to unwind those noncanonical structures in an ATP-dependent manner (Fig. 6i). This phase fusion and repulsion mechanism seems critical for maintaining genome stability, as DHX9 cKO in B cells leads to genome-wide accumulation of G4 sites and triggers the formation of much larger G4 condensates within the nucleus. These unresolved and accumulated G4 structures finally induce gross DSBs in B cells, causing elevated CSR (Fig. 6i). Moreover, such an active G4 unwinding mechanism may be critical for preventing various autoimmune diseases, including Hashimoto’s thyroiditis.

DHX9 is a multifunctional helicase that can unwind various DNA and RNA secondary structures in vitro, such as R-loops and G4s²⁴. It has been demonstrated that DHX9 can prevent R-loop accumulation in HeLa cells treated with the Top1 inhibitor camptothecin, which is known to induce R-loop accumulation and DNA damage⁴⁶. Unexpectedly, both immunofluorescence staining and genome-wide mapping showed that the R-loop level was barely changed in DHX9-ablated B cells. Although the apparent difference between the in vitro and in vivo results remains unknown, we speculate that some unknown RNA helicases in B cells may function as surrogates to resolve R-loops promptly in the absence of DHX9. In contrast, we found that G4 sites in DNA increased globally, as evidenced by immunofluorescence staining and genome-wide mapping with a specific antibody. Biochemical assays revealed that DHX9 preferentially bound G4 in vitro and could efficiently unwind G4 structures into linear sequences in the presence of ATP. These data collectively suggest that DHX9 is a novel G4 resolvase in B cells. DHX9 is widely expressed in various tissues and cell types, so we speculate that it may function as a general helicase to resolve harmful G4 structures.

Genome-wide ChIP-seq mapping in various cancer cell lines and patient samples has revealed that G4 sites are located predominantly in regulatory regions, such as promoters and 5’ UTRs^16,47. Our G4 ChIP-seq mapping in WT and DHX9-ablated B cells confirmed this preferential localization and revealed an unexpected prevalence of G4 sites at enhancers. As enhancers tend to form long-range loops to contact promoters to boost transcription and a single enhancer can interact with dozens or even hundreds of promoters^48,49, these proximally interacting and clustered regulatory regions increase the concentration of G4s for condensate formation. Timely resolving these G4 structures within the nucleus seems essential for maintaining genome stability. We found that long-term and unresolved G4 sites strongly correlated with DSB sites. Moreover, the top G4 sites identified by ChIP-seq in B cells tended to coincide with sites of genomic translocation to the IgH locus, where the first mammalian enhancer was identified and is located^50,51. As chromosomal translocation between IgH and non-IgH loci is usually associated with B cell lymphomas^43,52-55, we hypothesize that maintaining the optimal level of DHX9 in mature and activated B cells may be crucial for preventing B cell malignancies.

Protein and RNA phase separation have been extensively involved in many biological processes^56-59. Disrupting or promoting the phase behavior of these two kinds of macromolecules has been proven to be involved in various diseases^58,60. Unlike proteins and RNA, DNA binding with proteins such as HP1a and cGAS can promote the formation of phase-separated droplets^61,62. However, whether DNA itself can undergo phase separation within the nucleus remains unknown. In this study, we unexpectedly found that G4 sites in DNA undergo LLPS in B cells for the first time. Many factors, including the following, may collectively enable the formation of G4 condensates: (1) A single G4 peak typically contains 4-5 G4 motifs, while peaks at the IgH locus have 52 G4 motifs and 298 G4 repeats. Such repeat elements may provide multivalence to support phase separation. (2) The topological organization of genomic DNA within nuclei, such as topologically associating domains and loops, may also provide an environment conducive to placing G4 sites in proximity. (3) G4 DNA can form multiplanar structures via intermolecular and intramolecular G4 stacking^4,63, and we demonstrated that mixing Sm-G4 and Myc-G4 enable the more rapid formation of G4 condensates than either alone. G4 condensate formation seems beneficial for efficiently resolving these noncanonical structures at specific sites by recruiting the novel helicase DHX9 within the nucleus. However, this mechanism also provides an environment supporting DSB-induced proximal chromosomal translocation, which is a primary driver of B cell malignancies⁶⁴. Therefore, maintaining G4 at appropriate levels is crucial to prevent unwanted translocations.

Mammalian cells may encode various RNA and DNA helicases to resolve G4 sites and condensate in time. In this study, we demonstrated that DHX9 is a G4 resolvase. An appropriate expression level of DHX9 is critical for health. DHX9 has been reported to be an autoantigen in SLE patients³¹. Although the detailed mechanism of DHX9 in SLE occurrence remains unknown, determining whether G4 DNA is changed in patients with SLE will be interesting in the future. In addition, we identified a novel Y1189C mutation in the RGG box of DHX9 in Hashimoto’s thyroiditis patients. We further demonstrated that this mutation largely compromised the phase separation ability and G4 unwinding activity of DHX9, leading to G4 accumulation and gross DSBs. As the Y1189C conversion is not located in the helicase-associated domain, the mechanism that influences the global structure and helicase activity of DHX9 also needs future investigation.

In summary, DHX9-mediated unwinding of G4 non-canonical DNA structures functions as a novel regulatory mechanism to maintain genome stability in B cells. Although the DHX9 protein and G4 DNA phases fusion‒repulsion model was proposed, and this mechanism is active in activated B cells, it may also be present and extensively involved in controlling genome stability in other cell types. Our study highlights the physiological and pathological functions of G4s in DNA and sheds new light on treating autoimmune diseases and cancers by modulating the DHX9-G4 axis.

Zeraati, M. et al. I-motif DNA structures are formed in the nuclei of human cells. Nat Chem 10, 631–637 (2018).
Varshney, D., Spiegel, J., Zyner, K., Tannahill, D. & Balasubramanian, S. The regulation and functions of DNA and RNA G-quadruplexes. Nature reviews. Molecular cell biology 21, 459–474 (2020).
Santos-Pereira, J. M. & Aguilera, A. R loops: new modulators of genome dynamics and function. Nature reviews. Genetics 16, 583–597 (2015).
Bochman, M. L., Paeschke, K. & Zakian, V. A. DNA secondary structures: stability and function of G-quadruplex structures. Nature reviews. Genetics 13, 770–780 (2012).
Sen, D. & Gilbert, W. Formation of parallel four-stranded complexes by guanine-rich motifs in DNA and its implications for meiosis. Nature 334, 364–366 (1988).
Min, I. M. et al. The Smu tandem repeat region is critical for Ig isotype switching in the absence of Msh2. Immunity 19, 515–524 (2003).
Saha, T., Sundaravinayagam, D. & Di Virgilio, M. Charting a DNA Repair Roadmap for Immunoglobulin Class Switch Recombination. Trends in biochemical sciences 46, 184–199 (2021).
Dunnick, W., Hertz, G. Z., Scappino, L. & Gritzmacher, C. DNA sequences at immunoglobulin switch region recombination sites. Nucleic acids research 21, 365–372 (1993).
Lee, C. G., Kondo, S. & Honjo, T. Frequent but biased class switch recombination in the S mu flanking regions. Current biology: CB 8, 227–230 (1998).
Katapadi, V. K., Nambiar, M. & Raghavan, S. C. Potential G-quadruplex formation at breakpoint regions of chromosomal translocations in cancer may explain their fragility. Genomics 100, 72–80 (2012).
Cheung, I., Schertzer, M., Rose, A. & Lansdorp, P. M. Disruption of dog-1 in Caenorhabditis elegans triggers deletions upstream of guanine-rich DNA. Nature genetics 31, 405–409 (2002).
Georgakopoulos-Soares, I., Morganella, S., Jain, N., Hemberg, M. & Nik-Zainal, S. Noncanonical secondary structures arising from non-B DNA motifs are determinants of mutagenesis. Genome research 28, 1264–1271 (2018).
Paeschke, K., Capra, J. A. & Zakian, V. A. DNA replication through G-quadruplex motifs is promoted by the Saccharomyces cerevisiae Pif1 DNA helicase. Cell 145, 678–691 (2011).
Lensing, S. V. et al. DSBCapture: in situ capture and sequencing of DNA breaks. Nature methods 13, 855–857 (2016).
Chambers, V. S. et al. High-throughput sequencing of DNA G-quadruplex structures in the human genome. Nature biotechnology 33, 877–881 (2015).
Hansel-Hertsch, R. et al. G-quadruplex structures mark human regulatory chromatin. Nature genetics 48, 1267–1272 (2016).
Sanders, C. M. Human Pif1 helicase is a G-quadruplex DNA-binding protein with G-quadruplex DNA-unwinding activity. Biochem J 430, 119–128 (2010).
Wu, Y., Shin-ya, K. & Brosh, R. M., Jr. FANCJ helicase defective in Fanconia anemia and breast cancer unwinds G-quadruplex DNA to defend genomic stability. Molecular and cellular biology 28, 4116–4128 (2008).
Guo, M. et al. A distinct triplex DNA unwinding activity of ChlR1 helicase. The Journal of biological chemistry 290, 5174–5189 (2015).
Vannier, J. B. et al. RTEL1 is a replisome-associated helicase that promotes telomere and genome-wide replication. Science 342, 239–242 (2013).
Sun, H., Karow, J. K., Hickson, I. D. & Maizels, N. The Bloom's syndrome helicase unwinds G4 DNA. The Journal of biological chemistry 273, 27587–27592 (1998).
Fry, M. & Loeb, L. A. Human werner syndrome DNA helicase unwinds tetrahelical structures of the fragile X syndrome repeat sequence d(CGG)n. The Journal of biological chemistry 274, 12797–12802 (1999).
Lee, T. & Pelletier, J. The biology of DHX9 and its potential as a therapeutic target. Oncotarget 7, 42716–42739 (2016).
Chakraborty, P. & Grosse, F. Human DHX9 helicase preferentially unwinds RNA-containing displacement loops (R-loops) and G-quadruplexes. DNA repair 10, 654–665 (2011).
Jain, A. et al. DHX9 helicase is involved in preventing genomic instability induced by alternatively structured DNA in human cells. Nucleic acids research 41, 10345–10357 (2013).
Di Noia, J. M. & Neuberger, M. S. Molecular mechanisms of antibody somatic hypermutation. Annual review of biochemistry 76, 1–22 (2007).
Casellas, R. et al. Mutations, kataegis and translocations in B cells: understanding AID promiscuous activity. Nature reviews. Immunology 16, 164–176 (2016).
Chaudhuri, J. & Alt, F. W. Class-switch recombination: interplay of transcription, DNA deamination and DNA repair. Nat Rev Immunol 4, 541–552 (2004).
Qiao, Q. et al. AID Recognizes Structured DNA for Class Switch Recombination. Mol Cell 67, 361–373 e364 (2017).
Chen, J. et al. The RNA-binding protein ROD1/PTBP3 cotranscriptionally defines AID-loading sites to mediate antibody class switch in mammalian genomes. Cell research 28, 981–995 (2018).
Vazquez-Del Mercado, M. et al. High prevalence of autoantibodies to RNA helicase A in Mexican patients with systemic lupus erythematosus. Arthritis research & therapy 12, R6 (2010).
Wiedemann, E. M., Peycheva, M. & Pavri, R. DNA Replication Origins in Immunoglobulin Switch Regions Regulate Class Switch Recombination in an R-Loop-Dependent Manner. Cell reports 17, 2927–2942 (2016).
Boguslawski, S. J. et al. Characterization of monoclonal antibody to DNA.RNA and its application to immunodetection of hybrids. Journal of immunological methods 89, 123–130 (1986).
Xu, W. et al. The R-loop is a common chromatin feature of the Arabidopsis genome. Nature plants 3, 704–714 (2017).
Estep, K. N., Butler, T. J., Ding, J. & Brosh, R. M. G4-Interacting DNA Helicases and Polymerases: Potential Therapeutic Targets. Curr Med Chem 26, 2881–2897 (2019).
Maizels, N. G4-associated human diseases. EMBO Rep 16, 910–922 (2015).
Huang, Z. L. et al. Identification of G-Quadruplex-Binding Protein from the Exploration of RGG Motif/G-Quadruplex Interactions. J Am Chem Soc 140, 17945–17955 (2018).
Aslam, M. A. et al. Histone methyltransferase DOT1L controls state-specific identity during B cell differentiation. EMBO Rep 22, e51184 (2021).
Muller, S., Kumari, S., Rodriguez, R. & Balasubramanian, S. Small-molecule-mediated G-quadruplex isolation from human cells. Nat Chem 2, 1095–1098 (2010).
Yamane, A. et al. RPA accumulation during class switch recombination represents 5'-3' DNA-end resection during the S-G2/M phase of the cell cycle. Cell Rep 3, 138–147 (2013).
Rocha, P. P. et al. Close proximity to Igh is a contributing factor to AID-mediated translocations. Molecular cell 47, 873–885 (2012).
Bemark, M. & Neuberger, M. S. The c-MYC allele that is translocated into the IgH locus undergoes constitutive hypermutation in a Burkitt's lymphoma line. Oncogene 19, 3404–3410 (2000).
Kuppers, R. & Dalla-Favera, R. Mechanisms of chromosomal translocations in B cell lymphomas. Oncogene 20, 5580–5594 (2001).
Xue, B., Dunbrack, R. L., Williams, R. W., Dunker, A. K. & Uversky, V. N. PONDR-FIT: a meta-predictor of intrinsically disordered amino acids. Biochimica et biophysica acta 1804, 996–1010 (2010).
Ng, Y. C. et al. A DNA-sensing-independent role of a nuclear RNA helicase, DHX9, in stimulation of NF-kappaB-mediated innate immunity against DNA virus infection. Nucleic Acids Res 46, 9011–9026 (2018).
Cristini, A., Groh, M., Kristiansen, M. S. & Gromak, N. RNA/DNA Hybrid Interactome Identifies DXH9 as a Molecular Player in Transcriptional Termination and R-Loop-Associated DNA Damage. Cell reports 23, 1891–1905 (2018).
Hansel-Hertsch, R. et al. Landscape of G-quadruplex DNA structural regions in breast cancer. Nature genetics 52, 878–883 (2020).
Cai, Z. et al. RIC-seq for global in situ profiling of RNA–RNA spatial interactions. Nature, 1–6 (2020).
Ye, R., Cao, C. & Xue, Y. Enhancer RNA: biogenesis, function, and regulation. Essays Biochem 64, 883–894 (2020).
Banerji, J., Olson, L. & Schaffner, W. A lymphocyte-specific cellular enhancer is located downstream of the joining region in immunoglobulin heavy chain genes. Cell 33, 729–740 (1983).
Gillies, S. D., Morrison, S. L., Oi, V. T. & Tonegawa, S. A tissue-specific transcription enhancer element is located in the major intron of a rearranged immunoglobulin heavy chain gene. Cell 33, 717–728 (1983).
Chiarle, R. et al. Genome-wide Translocation Sequencing Reveals Mechanisms of Chromosome Breaks and Rearrangements in B Cells. Cell 147, 107–119 (2011).
Klein, I. A. et al. Translocation-Capture Sequencing Reveals the Extent and Nature of Chromosomal Rearrangements in B Lymphocytes. Cell 147, 95–106 (2011).
Gostissa, M. et al. Long-range oncogenic activation of Igh-c-myc translocations by the Igh 3' regulatory region. Nature 462, 803–807 (2009).
Kuppers, R. Mechanisms of B-cell lymphoma pathogenesis. Nat Rev Cancer 5, 251–262 (2005).
Tauber, D. et al. Modulation of RNA Condensation by the DEAD-Box Protein eIF4A. Cell 180, 411–426 e416 (2020).
Hondele, M. et al. DEAD-box ATPases are global regulators of phase-separated organelles. Nature 573, 144–148 (2019).
Jain, A. & Vale, R. D. RNA phase transitions in repeat expansion disorders. Nature 546, 243–247 (2017).
Zhang, Y. et al. G-quadruplex structures trigger RNA phase separation. Nucleic Acids Res 47, 11746–11754 (2019).
Shin, Y. & Brangwynne, C. P. Liquid phase condensation in cell physiology and disease. Science 357 (2017).
Larson, A. G. et al. Liquid droplet formation by HP1α suggests a role for phase separation in heterochromatin. Nature 547, 236–240 (2017).
Du, M. & Chen, Z. J. DNA-induced liquid phase condensation of cGAS activates innate immune signaling. Science 361, 704–709 (2018).
Hansel-Hertsch, R., Di Antonio, M. & Balasubramanian, S. DNA G-quadruplexes in the human genome: detection, functions and therapeutic potential. Nat Rev Mol Cell Biol 18, 279–284 (2017).
Nussenzweig, A. & Nussenzweig, M. C. Origin of chromosomal translocations in lymphoid cancer. Cell 141, 27–38 (2010).
Biffi, G., Tannahill, D., McCafferty, J. & Balasubramanian, S. Quantitative visualization of DNA G-quadruplex structures in human cells. Nat Chem 5, 182–186 (2013).
Hänsel-Hertsch, R., Spiegel, J., Marsico, G., Tannahill, D. & Balasubramanian, S. Genome-wide mapping of endogenous G-quadruplex DNA structures by chromatin immunoprecipitation and high-throughput sequencing. Nature protocols 13, 551–564 (2018).

Acknowledgments

We thank Dr. Frank Grosse at Friedrich-Schiller University for kindly sharing the pFastBac-hDHX9 plasmid, Dr. Qianwen Sun at Tsinghua University for providing S9.6 cells and helping with ssDRIP-seq, Drs. Yan Teng and Yun Feng from the Biological Imaging Center for taking and analyzing confocal images, and Drs. Junying Jia and Shu Meng for technical assistance in FACS sorting. This work was supported by the National Natural Science Foundation of China (32025008, 91740201, 91940306, 81921003), the Strategic Priority Program of CAS (XDB37000000), the Ministry of Science and Technology of China (2017YFA0504400) to Y. X. and by the National Natural Science Foundation of China (31970611) to J. C.

Contributions

Y. X. and J. M. conceived the project. J. C. generated knock-in cell lines and performed genotyping, FACS, DNA FISH, immunofluorescence, CSR, ELISA, the comet assay, Western blot analysis, the G4 pulldown assay, the in vivo phase separation assay, and ssDRIP-seq with the help of M. L.; X. L. and W. T. created expression plasmids, purified DHX9 proteins, and performed EMSA and the in vitro phase separation assay; J. C. and S. L. performed the in vitro helicase assay; S. L. constructed G4 ChIP-seq libraries with the help of J. C.; Y. Z. and C. C. collected patient samples and performed genome analysis; H. Z. performed the bioinformatic analysis; Y. X. wrote the manuscript with the help of J. C.; X. L.; L. W. and C.C.

Data availability

G4 ChIP-seq and ssDRIP-seq data have been deposited at Gene Expression Omnibus under the accession number GSE179244.

Code availability

The script for analyzing G4 ChIP-seq data is available on GitHub (https://github.com/Zhaohailian/G4_ChIPseq_mapping).

Competing interests

The authors declare no competing financial interests.

DHX9 cKO mice generation

DHX9^loxp/loxpmice were generated at BIOCYTOGEN company by injecting Cas9 mRNA, sgRNAs, and pUC57 donor plasmid directly into zygotes. Aicda^Cre+/- mice were kindly provided by Dr. Baidong Hou at the Institute of Biophysics, CAS. DHX9^loxp/loxpmice and Aicda^Cre+/- mice were maintained in C57BL/6J genetic background and intercrossed to generate the mice with conditional deletion of DHX9 in the activated B cells. Genotyping was performed by PCR with loxP-specific primer pairs and confirmed by Sanger sequencing (see Supplementary Table 3). The PCR products of the WT allele were 3,126 bp, while the DHX9-ablation allele showed a 403 bp fragment. Six- to ten-week-old mice were used for splenic B cell isolation and antibody titer measurement. The DHX9^loxp/loxp^{, Aicda-Cre} mice were used for subsequent analysis, and their littermates with genome type of DHX9^loxp/loxp or DHX9^loxp/+ were used as controls. All the surgical procedures were approved and performed under the guidelines of the IBP Animal Care and Use Committee.

RT-qPCR

Total RNA was extracted from WT or DHX9 cKO B cells with TRIzol reagent (Thermo Fisher, 15596018) and converted into cDNA with M-MLV Reverse Transcriptase (Promega, M1708) after depleting genomic DNA by RQ1 DNase I digestion. Quantitative PCR was performed on a StepOne Plus Real-Time PCR System (Applied Biosystems) with Hieff® qPCR SYBR Green Master mix (YEASEN, 11203ES08). GAPDH served as an internal control. All primers used in this study are shown in Supplementary Table 3.

Cell culture, immunization, and CSR

Splenic B cells were isolated from WT and DHX9 cKO mice by negative selection using the EasyStep mouse B cell isolation kit (Stem cell, 19854). The B220⁺ populations typically accounted for over 95% of the isolated cells. Primary B cells and CH12F3 cells were cultured in RPMI 1640 medium supplemented with 10% fetal bovine serum (FBS, Gibco, 16000044) and 50 µM 2-mecaptoethanol in 5% CO₂ at 37 °C. LPS (25 mg/ml) was used to activate ex vivo cultured WT or DHX9 cKO B cells. DHX9 cKO mice and their WT littermates were intraperitoneally injected with 5 mg LPS per mouse (Sigma-Aldrich, L3024). After immunization, spleens or Peyers’patches were collected on day 10 for FACS analysis. Memory B cells (B220⁺/CD38⁺/GL7^-/IgD^-/IgM^-) and plasma cells (B220^-/CD138⁺) were gated and sorted out using different antibodies. The IgA⁺ or IgG3⁺ B cells from non-immunized or LPS-immunized mice were quantified by FACS using specific antibodies. The combination of CD40L (0.5 mg/ml, Peprotech, 315-15), IL-4 (20 ng/ml), and TGF-β (20 ng/ml, Peprotech, 100-21C) was directly added to the CH12F3 culture medium and incubated for additional 72 h to induce CSR from IgM to IgA. CSR was examined as we previously described³⁰. The following antibodies were used for flow cytometry analysis: anti-B220 (Biolegend, 103224/103205; Invitrogen, 4339662), anti-CD19 (Biolegend, 115519), anti-GL7 (Biolegend, 144603) and anti-Fas (Biolegend, 152603), anti-IgG3 (Biolegend, 406803), anti-CD38 (Invitrogen, 2312470; Biolegend, 102711), anti-CD138 (Biolegend, 142505), anti-IgD (Biolegend, 405740/405716/405711; Invitrogen, 4346705), anti-IgM (Biolegend, 406505/406511), anti-IgA (Biolegend, 407003), anti-streptavidin PE (eBioscience, 12-4317-87), anti- streptavidin FITC (eBioscience, 11-4317-87) and anti-streptavidin e450 (Invitrogen, 2110549). All the FACS analyses were performed on Fortessa (BD Biosciences) and analyzed by FlowJo software (Tree Star).

Antibody quantification by ELISA

Mouse serum was collected from the orbital venous plexus of unchallenged WT and DHX9 cKO mice. The concentrations of Ig (IgA, IgM, IgG1, and IgG3) were measured using mouse Ig ELISA quantitation sets according to the manufacturer’s protocol (Bethyl Laboratories, E90-103, E90-101, E90-105, E90-111). The culture medium from DHX9^WT, DHX9^306-308del, and DHX9^Y1189C mutant CH12F3 cells was collected for quantifying the IgA titers with mouse IgA ELISA quantitation set (E90-103). For patient serum, the concentration of IgG3 and IgE were measured by unlabeled mouse anti-human IgG3 hinge (SouthernBiotech, HP6050) and human IgE ELISA quantification kits (Bethyl Laboratories, A80-108), respectively. The absorbance was measured on a microplate spectrophotometer (Thermo Fisher Scientific, 1510).

Lentiviral shRNA packaging and transduction

The pLKO.1 lentiviral shRNA clones (TRCN0000335501, TRCN0000071058, TRCN0000071019, TRCN0000086693, TRCN0000312950) were applied to knockdown Wrn, Pif1, Fancj, Blm, and DHX36, respectively. The sequences of DHX9 shRNA used in CH12F3 cells are shown in Supplementary Table 3. To produce lentiviral particles, 6 mg of shRNA plasmid and packaging plasmids were co-transfected into 293T cells with Lipofectamine 2000 reagent as we previously described. The lentiviral particles in the culture medium were collected twice after 48 h and 72 h of transfection. For transduction, 5 ml of lentiviral supernatant and 5 ml of fresh medium were gently mixed with 8 mg/ml of polybrene (Sigma-Aldrich, H9268) and then applied to CH12F3 cells for 24 h. The stably transduced CH12F3 cells were selected by puromycin (8 mg/ml) for an additional 3 days.

GFP and mScarlet knock-in and mutant cell lines

To generate GFP- and mScarlet-tagged DHX9 cell lines, the homology repair template consisted of GFP or mScarlet cDNA sequence flanked on either side by 1200 bp homology arms was amplified from genomic DNA using KOD Xtreme Hot Start polymerase (Novagen, 71975-3). The repair templates were subsequently cloned into pEGFP-N1 using the Not1-HF restriction enzyme (NEB, R3189L), while the sgRNA was inserted into PX330-mcherry using the Bbs1 enzyme (Thermo Fisher Scientific, FD1014). The 5 mg of PX330-mcherry-sgRNA plasmid and 5 mg of homology repair template pEGFP-N1 were transfected into 10⁶ CH12F3 cells using Electrotachometer (Lonza). The mCherry and GFP double-positive cells were sorted out after 24 h of transfection and expanded for one additional week. GFP- or mScarlet-positive cells were further sorted out and plated into a 96-well plate. Genotyping was performed using 2 × M5 Hiper plus Taq HiFi PCR mix (Mei5 Biotechnology, MF002-plus-01), and the PCR products were gel purified for Sanger sequencing. To further generate c.306-308del or p.Y1189C mutations, 10⁶ GFP-knockin CH12F3 cells were transfected with 5 mg PX330-mcherry-sgRNA and 5 mg repair template containing the relevant mutation. The correct knock-in clones were individually cultured and confirmed by sanger sequencing. The sequences of sgRNA and primers used for genotyping are shown in Supplementary Table 3.

Plasmid construction

The pFastBac-hDHX9 plasmid was kindly provided by Dr. Frank Grosse, while the pCMV6-mDHX9 plasmid was purchased from Origene (MR226503). The human DHX9 fragment was inserted into the pMamBac-GFP vector between XhoI and EcoRI restriction sites for protein purification. The pMamBac-DHX9^Y1189C-GFP, pMamBac-DHX9^K417R-GFP, and pMamBac-DHX9^ΔIDR3-GFP plasmids were generated by site-directed mutagenesis. The IDR1, IDR2, and IDR3 of DHX9 were inserted into the pSumo backbone or the pET-3C-GFP backbone between XhoI and EcoRI restriction sites. The pSANG10-3F-BG4 plasmid used for expressing BG4 scFv anti-G-quadruplex antibody was a gift from Shankar Balasubramanian (Addgene plasmid # 55756).

Immunofluorescence

Splenic B cells were treated with PDS (2 mM; MCE, 1781882) for 72 h to stabilize G4 structures for examining the effect of G4 accumulation on DNA damage. Splenic B cells or CH12F3 cells were fixed with 4% paraformaldehyde (PFA, Sigma-Aldrich, F8775) in PBS and then permeabilized with 0.3% Triton X-100 for 15 min. For DNA G4 staining, cells were first treated with 50 µg/ml of RNase A to eliminate potential RNA G4s. After washing three times with PBS, splenic B cells were blocked with 10% normal donkey serum in PBS for 1 h. The primary antibodies against G4, R-loop, FLAG (Sigma-Aldrich, F1804; ABclonal, AE004), DHX9 (Invitrogen, PA5-19542; Abcam, ab26271), 53BP1 (Santa Cruz, sc-515841), Rad51 (Santa Cruz, sc-398587), and gH2AX (Millipore, 05-636) were diluted in PBST and incubated with cells overnight at 4 °C. Cells were washed with PBST for three times and then incubated with donkey anti-mouse 488 (Life technology, A21202), donkey anti-rabbit 488 (Life technology, A21206), donkey anti-mouse 594 (Life technology, A21203), donkey anti-rabbit 594 (Life technology, A21207), and goat anti-mouse 647 (Life technology, A28181) for 1 h at RT and counterstained with DAPI for 10 min. The cells were spread on cover slides and imaged with confocal FV1200. The BG4 and S9.6 antibodies were purified as previously described^34,65.

DNA FISH and combined G4 staining

DNA FISH probes were prepared by PCR amplification of the genomic fragments covering the G4 peaks of IgH Sm, Myc, and Pim1. The purified PCR products were fluorescently labeled with Alexa Fluor 546 or 647 using a ULYSIS Nucleic Acid Labelling Kit (Thermo Fisher Scientific, U21652, U21660). The labeled DNA FISH probes need to be denatured at 95 °C for 10 min and snap-chill on ice before use. For DNA FISH, LPS-activated B cells were first fixed with 4% PFA for 15 min at RT and then permeabilized with 70% ethanol for 15 min. The permeabilized cells were subsequently denatured in 2 × SSC buffer containing 70% formamide for 10 min at 75 °C. The mixture of 1% salmon sperm DNA (Thermo Fisher Scientific, 15632011) and 1 mg of labeled probes were applied to the cells and hybridized overnight at 37 °C. On the next day, slides were washed 3 times with 2 × SSC buffer at 45 °C for 10 min and stained with DAPI (Sigma-Aldrich, F6057) for 10 min. After washing twice with 2 × SSC buffer, the slides were sealed with mounting medium (Vector Laboratories, H-1000) for visualization with the Olympus FV1200 microscope.

For combined DNA FISH and G4 immunofluorescent staining, the permeabilized cells were pre-treated with 50 µg/ml RNase A at 37 °C for 2 h and washed three times with PBS. After blocking with 10% normal donkey serum for 1 h, the BG4 antibody was 1:100 diluted with PBS buffer and directly added to the blocked cells overnight at 4 °C. Subsequently, the anti-flag and donkey anti-mouse 488 antibodies were applied to visualize G4. After washing three times with PBST, the cells were permeabilized with 0.7% Triton-X-100/0.1 M HCl for 10 min on ice and denatured with 1.9 M HCl for 30 min at RT. Cells were then hybridized with DNA probes overnight at 37 °C and washed three times with 2 × SSC for 20 min. The nuclei were finally counterstained with DAPI.

COMET assay

Splenic B cells from WT and DHX9 cKO mice were ex vivo cultured and stimulated with LPS for 72 h. DNA damage was assessed with a Comet Assay Kit (R&D, 4250-050-K). Briefly, 10 µl of cell suspension was mixed with 50 µl of 0.5% low-melting agarose and spread onto a CometSlide (Trevigen) pre-coated with 100 µl of 1% normal-melting agarose. Cells were lysed by immersing the slides in a freshly prepared lysis solution (2.5 M NaCl, 100 mM ethylenediaminetetraacetic acid (EDTA), 10 mM Tris, 1% Triton X-100, 10% DMSO, pH 10.0) at 4 °C for 2 h. Subsequently, slides were washed three times in 1 × PBS and placed in a gel electrophoresis apparatus with freshly prepared electrophoresis buffer (1 mM EDTA, 300 mM NaOH, pH 13.0) for 25 min to allow DNA unwinding. Electrophoresis was performed at 20 V with a starting current of 300 mA for 20 min. Lastly, slides were neutralized with 0.4 M Tris (pH 7.5) three times and stained with 1 × SYBR Gold solution for visualization under a fluorescence microscope (Life Technologies). Splenic B cells isolated from WT mice were treated with 100 mM hydrogen peroxide (H₂O₂) for 20 min at 4 °C and served as a positive control. The tail moment was calculated by the length between the center of the head and the tail.

Western blot

Western blot was performed as previously described³⁰. The following antibodies were used: anti-DHX9 (Invitrogen, PA5-19542), anti-GAPDH (Cell Signaling Technology, 2118L), anti-Actin (Invitrogen, MA5-15739), anti-Wrn (Immunoway, YT7780), anti-Pif1 (Proteintech, 19006-1-AP), anti-Fancj (Proteintech, 24436-1-AP), anti-Blm (Bioss, bs-12872R), anti-DHX36 (Proteintech, 13159-1-AP), anti-ROD1 (Abnova, H00009991-M01), anti-AID (Life Technologies, 392500), anti-53BP1 (Santa Cruz, sc-515841), anti-Rad51 (Santa Cruz, sc-398587), anti-gH2AX (Millipore, 05-636), and anti-H2AX (Abcam, ab20669). Horseradish peroxidase (HRP)-coupled secondary antibodies (1:5,000) were purchased from Pierce (31460, 31430), and mouse anti-rabbit IgG mAb (Conformation Specific, HRP Conjugate, 5127S) was purchased from Cell Signaling Technology.

Protein purification

DHX9 and mutant proteins were expressed using an insect and mammalian expression system. Briefly, 1 mg of pMamBac-hDHX9, hDHX9^WT-GFP, hDHX9^Y1189C-GFP, hDHX9^ΔIDR3-GFP, or hDHX9^K417R-GFP plasmids were individually transfected into 10⁶ Sf9 cells and incubated at 27 °C for 72 h. The P1 baculovirus supernatant was collected by centrifugation at 500 g for 5 min, and 1-2% of the P1 baculovirus was directly added to 2 × 10⁶/ml Sf9 suspension cells and incubated at 27 °C on a shaker at 130 rpm for an additional 72 h. The supernatant was centrifuged at 500 g for 5 min to collect P2 baculovirus, which was used to generate high-titer P3 baculovirus. To express the recombinant proteins, 3 × 10⁶/ml HEK293F (Gibco, R79007) suspension cells were infected with 10-20% of P3 baculovirus for 12 h. The infected cells were treated with 10 mM sodium butyrate and incubated for an additional 60 h at 30 °C on a shaker at 130 rpm. The cells were collected by centrifugation at 4000 rpm with a Superspeed centrifuge (Thermo Fisher Scientific, LYNX6000) for 10 min at 4 °C. After resuspending the cell pellets in binding buffer (20 mM Tris-Cl pH 8.0, 500 mM NaCl, 25 mM imidazole) and lysed with an ultrasonic disruptor (Ningbo Scientz Biotechnology, SB-80). The lysates were centrifuged at 18000 rpm for 1 h at 4 °C, and the supernatant was used for purifying recombinant proteins with StrepTrap^TM HP (Cytiva, 10278182). After eliminating nucleic acids with a buffer (20 mM Tris-Cl pH 8.0, 2 M NaCl, 25 mM imidazole), the proteins were eluted with an elution buffer (20 mM Tris-Cl pH 8.0, 500 mM NaCl, 25 mM imidazole, and 10mM desulfurizing biotin), and purified with a Hiload^TM 16/600 Superdex^TM 200pg column (Cytiva, 28-9893-35) on an ÄKTA Pure (Cytiva). The tagged proteins were eluted with a buffer containing 20 mM Tris-Cl pH 8.0, 300 mM KCl, and 1 mM DTT. The PreScission protease was added to the eluted protein solution to remove his tag from purified proteins and incubated overnight at 4 °C. The His-Strep-GFP tag and His-tag were eliminated by HisTrap^TM HP (Cytiva, 10251115). Untagged proteins were acquired in flowthrough and purified with a Hiload^TM 16/600 Superdex^TM 200pg column on an ÄKTA Pure (Cytiva) using buffer (20 mM Tris-Cl pH 8.0, 500 mM NaCl, and 1 mM DTT).

For expressing IDR1, IDR2, and IDR3 truncated proteins, pSUMO-IDR1, pSUMO-IDR2, and pSUMO-IDR3 plasmids were individually transformed into BL21 (DE3) strain. The proteins were expressed and purified as we previously described³⁰. To remove the his-sumo tag on protein, Ulp1 protease was added to the eluted product and dialyzed with buffer (20 mM Tris-HCl pH 8.0, 500 mM NaCl) overnight at 4 °C. The dialyzed protein samples were purified using another Ni-NTA resin to remove the protease, His-SUMO tag, or uncleaved proteins. The proteins were concentrated by centrifugation using an ultrafiltration tube and purified with a pre-equilibrated HiLoad^TM 16/60 Superdex^TM75pg column (Cytiva, 17-1068-01). Finally, the IDR1, IDR2, and IDR3 proteins were eluted using a buffer containing 20 mM Tris-HCl, pH 8.0, 300 mM KCl, and 1 mM DTT. The purity of the proteins was confirmed by fractionating on 4-15% SDS-PAGE and stained with Coomassie blue G-250 (Urchem, 71011381).

In vitro phase separation assay

To monitor temperature-dependent DHX9 phase separation, 1.6 mM DHX9 proteins in 300 ml buffer P (20 mM Tris-HCl pH 8.0, 150 mM NaCl, 1 mM DTT) were first incubated at 4 °C for 20 min and then transferred to 37 °C for additional 10 min. Subsequently, the mixture was incubated at 4 °C for 60 min. Images were collected at different time points by the FV3000 microscope. The sample without DHX9 proteins served as a negative control. To examine the dynamics of liquid droplets, GFP-tagged DHX9, the truncates, or the mutant proteins were dissolved in buffer P (20 mM Tris-HCl pH 8.0, 150 mM NaCl, 1 mM DTT) at a final concentration of 40 mM and transferred to a glass slide for monitoring the phase behaviors at indicated time points. For examining the wetting properties of liquid droplets on the surface of a glass slide, GFP-tagged DHX9 dissolved in phase separation buffer was transferred to the glass slide and incubated at RT for 0.5 h. Then the images were collected at indicated time points. To evaluate the influence of salt ions on DHX9 droplets, different amounts of DHX9 proteins were added to the buffer containing 20 mM Tris-HCl pH 8.0, 1 mM DTT, and varying concentrations of NaCl. The tubes were incubated at RT for 10 min and then loaded onto the glass slide for monitoring droplet formation under different conditions.

To examine the influence of G4 DNA on DHX9 droplet, 40 mM of GFP-tagged DHX9 in buffer D (20 mM HEPES pH 7.5, 150 mM KCl, 1 mM DTT) were mixed with 10 mM of Cy5-labeled Sm-G4. After transferring to the glass slide and incubating at RT for 1 h, the phase behavior was recorded using the FV3000 microscope. To study the ability of DHX9 or Y1189C mutant on unwinding G4, 10 mM GFP-tagged DHX9 or Y1189C mutant were first mixed with 2.5 mM Cy5-labeled Sm-G4 in buffer D (20 mM HEPES, pH 7.5, 150 mM KCl, 1 mM DTT) containing 3 mM MgCl₂, followed by adding Ficoll-400 to a final concentration of 150 mg/ml (Solarbio, F8150), and ATP (Meilunbio, MB3157) or AMP-PNP (Sigma-Aldrich, A2647) to a final concentration of 3 mM. The mixtures were transferred to a glass slide and imaged at different time points by FV3000 microscopy.

In vitro DNA phase separation assay

Single-stranded Sm DNA fragments, including 1 × Sm, 2 × Sm, and 4 × Sm, were synthesized by BGI genomics. To generate longer G4-containing ssDNA fragments, we developed a PCR-based lambda exonuclease digestion approach. The forward and reverse primers were synthesized and 5’ end-labeled with biotin or phosphorylation, respectively. The G4-containing Sm (1000 and 1900 bp) and Myc (202 bp) regions were PCR amplified with the primer pairs, 2 × M-PCR OPTI Mix (Biotool, B45012), and 50 ng murine genomic DNA. A 1000 bp non-G repeat fragment from IgH was amplified to serve as a negative control. The PCR fragments were gel purified, and 5 mg of them were treated with 1 ml of lambda exonuclease (NEB, M0262S) for 3 h at 37 °C to digest the phosphorylated strand. The biotinylated forward strand DNA was purified by phenol: chloroform: isoamyl alcohol (25: 24: 1) extraction and ethanol precipitation. The precipitated ssDNA fragments were resuspended in HEPES buffer (20 mM HEPES pH 7.5, 100 mM KCl, 1 mM DTT) and folded into G4 structures. Circular Dichroism (CD) was used to determine the conformation of the G4 structures.

To test the effect of G4 length on droplet formation, 300 ng Sm-G4 with varying lengths were added to the buffer G (20 mM HEPES pH 7.5, 100 mM KCl, 1 mM DTT) containing 10% PEG8000 and incubated for 1 h at RT. The solution was transferred onto the glass slide, and images were collected by an FV3000 microscope. For monitoring the internal dynamics of G4 condensation, the 1900 nt Sm-G4 and 202 nt Myc-G4 fragments were labeled with streptavidin-647 (Invitrogen, S21374) or streptavidin-488 (Invitrogen, S11223), respectively. 300 ng Sm-G4-647 or Myc-G4-488 was individually added to the buffer G containing 10% PEG8000, and images were collected at indicated time points by the FV3000 microscope. To examine the fusion ability between different G4 molecules, 300 ng Sm-G4-647 and Myc-G4-488 were added to the buffer G containing 10% PEG8000 and incubated for 1 h at RT. Images were collected by an FV3000 microscope.

G4 and linear DNA purification

To generate G4 substrate, Sm and Myc ssDNA fragments were first dissolved in HEPES buffer (20 mM HEPES pH 7.5, 100 mM KCl, 1 mM DTT) and then denatured at 95 °C for 5 min, followed by slowly cooling down to RT. To make linear DNA, we denature both fragments at 95 °C for 10 min and immediately put the tube on ice to prevent G4 structure formation. The size exclusion column Superdex75 or 200 (GE Healthcare) was used to purify G4 and linear DNA fractions in an AKTA purifier. Circular dichroism (CD) spectra for the Sm-G4 and Sm-linear were recorded at RT on a Chirascan Plus at the wavelength range of 220-350 nm.

EMSA

25 nM of Sm-G4 labeled with ROX dye was mixed with increasing amounts of full-length DHX9 or Y1189C mutant proteins in binding buffer (20 mM HEPES pH 7.5, 100 mM KCl, 1 mM DTT, 5% glycerol) and incubated on ice for 30 min. The samples were then fractionated on a 5% polyacrylamide gel at 4 °C in TBE buffer (40 mM Tris-HCl, PH 8.3, 45 mM boric acid, 1 mM EDTA). The shift bands were visualized at the ROX channel using Typhoon 7000. Bound and free DNA were quantified using ImageJ software, and the binding curves were fitted by GraphPad Prism 8.0 software.

In vitro helicase assay

To examine the G4 unwinding activity of DHX9 and Y1189C/K417R mutants, 100 nM of the ROX-labeled Sm-G4 was dissolved in 10 ml of 20 mM Tris-HCl, pH 7.5, 3.5 mM MgCl₂, 3.5 mM ATP, 0.1 mg/ml BSA, 5 mM DTT and 10% (v/v) glycerol. The Sm linear sequence was used as a negative control. After adding the different amounts of DHX9 or Y1189C mutant proteins and incubating at 37 °C for 30 min, the reaction was terminated by snap cooling on ice, followed by the addition of 3.5 ml stop solution (50 mM EDTA, 2% (w/v) SDS, 0.1% (w/v) xylene cyanol, 0.1% (w/v) bromophenol blue, 40% (w/v) glycerol). The samples were fractionated on 5% native polyacrylamide gels and visualized with a ROX channel using a Typhoon 7000. Sm-G4 or Sm-linear DNA was quantified using ImageJ and GraphPad Prism 8.0.

To test the G4 unwinding activity of DHX9 or different mutants in phase state, Sm-G4 was dissolved in phase buffer containing 20 mM HEPES, 150 mM KCl, 1 mM MgCl₂, 1 mM DTT with 1 mM ATP or AMP/PNP. DHX9 or mutants were added to the mixture and incubated at 37 °C for 20 min. The G4 structure resolving ability of DHX9 or mutants was examined using CD spectroscopy.

G4 pulldown assay

The 110 nt 5’-biotinylated Sm linear sequence was folded into Sm-G4 in a folding buffer (20 mM HEPES pH 7.5, 100 mM KCl, 1 mM DTT) as previously described ²⁹. The MyOne streptavidin C1 beads (Life Technology, 65002) were washed three times with binding buffer (20 mM Tris-HCl, pH8.0, 500 mM KCl) and then coupled with the biotin-labeled Sm-G4 or Sm-linear for 1 h at RT, respectively. Splenic B cells were ex vivo cultured and stimulated with LPS for 72 h. The cells were lysed in protein lysis buffer (50 mM Tri–HCl pH 7.5, 150 mM NaCl, 5 mM EDTA, 5 mM EGTA, 1 % NP-40, protease inhibitor), and the extract was pre-cleared at 4 °C for 1 h with C1 beads. The pre-cleared lysate was diluted with an equal volume of 2 × binding buffer (20 mM Tris-HCl pH7.5, 2 mM EDTA, 200 mM KCl, 0.2 mM DTT, 10% glycerol, 0.02 mg/ml BSA) and incubated with Sm-DNA-coupled C1 beads for 1 h at 4 °C. After thoroughly washing with lysis buffer three times, the G4-binding proteins were eluted into 1 × LDS sample buffer (Invitrogen, NP0008) for western blot analysis.

ssDRIP-seq and G4 ChIP-seq

ssDRIP-seq was performed as previously described³⁴. G4 ChIP-seq was modified from a previously published protocol⁶⁶. Briefly, splenic B cells were crosslinked with 1% formaldehyde for 10 min at RT and stopped by adding 125 mM glycine for 10 min. After washing twice with ice-cold PBS, the cells were pellet down at 300 g for 5 min at 4 °C. The cell pellets were resuspended in 1 ml of ice-cold cyto lysis buffer (10 mM Tris-Cl, pH 8.0, 10 mM NaCl, 0.5% NP-40, 1 × protease inhibitor cocktail) and incubated on ice for 10 min with occasional inversion every 2 min. Pellet down cells at 2,300 g for 5 min at 4 °C. The supernatant was discarded, and the remaining nuclear pellet was resuspended in 500 ml of nuclear lysis buffer (1% SDS, 10 mM EDTA, 50 mM Tris-HCl, pH 8.1, 1 × protease inhibitor cocktail). The nuclear lysate was then sonicated seven times at the maximum setting by BRANSON SLPe Sonifier (output setting 4, 10 seconds per cycle). The soluble chromatin was collected by centrifugation at 14,000 rpm for 10 min. After quantification, approximately 1 mg of chromatin (2.5 ml) was transferred to a 1.5 ml Eppendorf LoBind microcentrifuge tube, followed by adding 1 ml of RNase A and 43.5 ml of blocking buffer (1% BSA, 25 mM HEPES, pH 7.5, 10.5 mM NaCl, 110 mM KCl, 1 mM MgCl₂) and incubated in a thermomixer at 37 °C for 20 min. Subsequently, 1 mg of BG4 antibody was added to the mixture and incubated by rotation at 1400 rpm in a thermomixer at 16 °C for 1 h. To pulldown BG4 bound chromatin fragments, 10 ml of pre-blocked anti-FLAG M2 magnetic beads (Sigma-Aldrich, M8823) was further applied to the tube and incubated at 16 °C for 1 h. The beads were washed three times with 200 ml of ice-cold wash buffer (100 mM KCl, 10 mM Tris-HCl, pH 7.4, 0.1% Tween-20), and then washed with the wash buffer for additional three times at 1400 rpm for 10 min at 37 °C. Placed the tube on a magnetic stand for 1 min and discarded the supernatant. Elute G4 DNA from the beads by adding 100 ml of TE buffer (10 mM Tris-HCl, pH 8.0, 1 mM EDTA) and 2 ml of proteinase K to the tube and incubated at 37 °C for 2 h at 1400 rpm, followed by de-crosslinking overnight at 65 °C. The G4 DNA in the supernatant was purified with a QIAquick PCR purification kit (Qiagen, 28104) and quantified by Qubit. The purified G4 DNA was converted into a library using the Accel-NGS 1S Plus DNA Library Kit (Swift Biosciences, 10024).

Imaging of DHX9 or G4 condensate in living cells

To monitor G4 condensate in vivo, CH12F3 cells were transfected with 10 pmol of Sm-G4 and Myc-G4 for 12 h, followed by counterstaining with Hoechst33342 for 10 min. The stained CH12F3 cells were transferred to a glass-bottom cell culture dish (NEST, 801002) for imaging by an FV1200 microscope. For examining the phase-separation ability of GFP- and mScarlet-tagged DHX9^306-308del, DHX9^Y1189C, or WT proteins in vivo, the corresponding CH12F3 cells were individually plated onto a confocal dish for 12 h, and the cells were stained with Hoechst33342 for imaging by FV1200 as described above.

FRAP

For in vitro FRAP experiments, DHX9 proteins or DNA G4 droplets were first formed at room temperature and transferred onto a glass slide. The selected regions were photobleached with 100% laser power by an FV3000 microscope at 488 nm and 647 nm. Images were acquired over a time course of 20 min. Fluorescence intensity in bleached regions was measured by Imaris and further analyzed by GraphPad Prism. For DHX9-related FRAP experiments within cells, the GFP/mScarlet-tagged or 306-308del/Y1189C mutant CH12F3 cells were cultured in a 96-well plate (PerkinElmer, 6055302) for 12 h. The CH12F3 cells were rinsed three times with PBS and stained with Hoechst33342 to visualize the nucleus. A single DHX9-GFP condensate was partially photobleached with 100% laser power by FV1200 on 488 nm and 594 nm. Live-cell images were recorded over a time course of 5 min. For G4 FRAP analysis, CH12F3 cells were transfected with Sm-G4 and cultured for 12 h in a 96-well plate. After counterstaining with Hoechst33342, G4 condensates were partially photobleached with 100% laser power by FV1200 on 594 nm. Cells were imaged over a time course of 5 min.

Patient samples collection

All the subjects enrolled for genetic analysis had written informed consent, and the protocol was approved by the Ethics Committee of Shanghai Jiao Tong University Affiliated Sixth People's Hospital. The studies are strictly in compliance with the Declaration of Helsinki. Bone marrow (BM) samples from subjects were obtained via aspiration, and the lymphocyte subsets were analyzed by FACS. The mononuclear cells were collected by density gradient centrifugation, and subsequently, the genomic DNA was extracted and sequenced by Kingmed Diagnostics. The blood samples from subjects were used for routine analysis, and the serum samples were used to detect antibody levels by ELISA.

Statistical analysis

All the experiments were independently repeated at least twice, and no inconsistent results were observed. GraphPad Prism 8.0 software was used to perform statistical analyses. Data are presented as the mean ± S.D. The box borders in the boxplots and violin plots represent the upper and lower quartiles (25th and 75th percentiles), and the centre line denotes the median. *p < 0.05, **p < 0.01, ****p < 0.0001, and p > 0.05 is considered as non-significant. All the data were reproducible, and details of replicates are described in the figure legends.

There is NO Competing Interest.

SupplementaryTable1.G4containinggenesidentifiedbyG4ChIPseq.xlsx
SupplementaryTable2.ListofIgHproximalinteractinggenesusedfortheanalysis.xlsx
SupplementaryTable3.Primersusedinthisstudy.xlsx
Supplementaryvideo1.ThefusionofGFPDHX9dropletsinvitrorelatedtoFigure5c.mp4
Supplementary video 1
Supplementaryvideo2.TheliquidspreadingactivityofGFPDHX9dropletsrelatedtoExtendedData6b.mp4
Supplementary video 2
Supplementaryvideo3.ThefusionofSuG4dropletsinvitrorelatedtoExtendedData7c.mp4
Supplementary video 3
Supplementaryvideo4.ThefusionofMycG4dropletsinvitrorelatedtoExtendedData7c.mp4
Supplementary video 4
Supplementaryvideo5.ThefusionofGFPDHX9andSuG4dropletsinvitrorelatedtoFigure5g.mp4
Supplementary video 5
SupplementaryMaterialsLegends.docx
ExtendedDataFiguresandLegends.docx

Download PDF

Version 1

posted

You are reading this latest preprint version

DHX9 resolves G-quadruplex condensation to prevent DNA double-strand breaks

Status:

Version 1

Abstract

Figures

Introduction

Results

Discussion

References

Declarations

Methods

Additional Declarations

Supplementary Files

Status:

Version 1