Co-phase separation of Y14 and RNA in vitro and its implication for DNA repair

Y14 contains evolutionarily conserved low-complexity domains rich in charged residues

All Y14 homologs have the same modular structure, comprising a central RRM flanked by low-complexity terminal domains (Fig. 1A). Both the amino- and carboxy-terminal domains are essentially unstructured except that the former contains a short α-helix and β-sheet (Fig. 1A; Fribourg et al. 2003; Shi and Xu 2003). Human Y14 is 100% identical to its mouse homolog, whereas it is only ∼64% identical to Drosophila Y14/Tsunagi and differs substantially from Caenorhabditis elegans Ce-Y14 (Fig. 1B). However, analysis of charge distribution in the terminal domains of human, Drosophila and Caenorhabditis Y14 homologs (Fig. 1B) using Classification of Intrinsically Disordered Ensemble Regions (http://157.245.85.131:8000/CIDER/) (Holehouse et al. 2017) revealed a conserved distribution of charged residues. The amino-terminal domain contains alternating stretches that are rich in either positively or negatively charged residues (Fig. 1C). This pattern of charge segregation is reminiscent of a set of proteins that have a high propensity to undergo LLPS, such as nucleolar nucleoplasmin 1 (NPM1), the stress-granule nucleator G3BP1, and the microtubule-associated protein tau (Mitrea et al. 2016, 2018; Boyko et al. 2019; Yang et al. 2020). Moreover, the carboxy-terminal region (ranging from 9 to 23 residues) of three Y14 homologs is highly positively charged. Therefore, these uniformly charged segments possibly contribute to electrostatic interactions intramolecularly and/or intermolecularly. Finally, the disorder-promoting score (Das and Pappu 2013) for each of the amino- and carboxy-terminal domains of human Y14 was calculated as 0.8 and 1.0, respectively (Supplemental Table S1), implying that they are relatively unstructured. The IDRs of Y14 do not have glutamine/asparagine-rich sequences that are typical of several disease-linked RBPs (Zbinden et al. 2020). Nevertheless, it was important to decipher whether Y14 undergoes LLPS and, if so, whether LLPS occurs in a manner similar to or distinct from that of disease-linked RBPs.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 1.

Charge distribution of the terminal domains of Y14. (A, upper panel) Domain organization of human Y14. (Lower panel) The structure of human Y14 was predicted by the AlphaFold Protein Structure network (https://github.com/deepmind/alphafold); the image was manipulated in PyMol Software (pymol.org). In all panels, acidic and basic residues are indicated in red and blue, respectively. RRM, RNA recognition motif; α0 and β0 represent the short α-helix and β-sheet, respectively. (B) Amino acid sequence alignment of human (Hs), Drosophila (Dm), and Caenorhabditis (Ce) Y14 homologs is shown according to a previous report (Fribourg et al. 2003). Conserved residues (identical or similar) in the amino-terminal domain are highlighted in gray. The two RS dipeptides in the carboxy-terminal domain of human Y14 are highlighted in yellow. (C) The amino- and carboxy-terminal domains of the three Y14 homologs in B were analyzed via Classification of Intrinsically Disordered Ensemble Regions (http://pappulab.wustl.edu/CIDER/analysis/). (NCPR) Net charge per residue distribution.

Y14 undergoes LLPS in a pH-dependent manner

To investigate whether Y14 forms liquid-like droplets in vitro, we generated recombinant Y14 containing an amino-terminal S peptide (to promote expression in Escherichia coli) and a carboxy-terminal polyhistidine (His) tag (to enable affinity purification) (Fig. 2A; Supplemental Fig. 1A; Hsu et al. 2005). Analytic ultracentrifugation of purified recombinant Y14 at pH 7.9 revealed a single peak at the estimated molecular mass, reflecting its monomeric form (Supplemental Fig. 1B). This dual-tagged Y14 had an estimated isoelectric point (pI) value of 5.86 (Supplemental Table S2). We initially evaluated droplet formation of Y14 (20 µM) in buffers containing 50 mM KCl at pH values ranging from 6 to 7.5 using a differential interference contrast (DIC) microscope. Spherical droplets of Y14 formed after a 2-h incubation in pH 6.5 buffer (Fig. 2B). The number and size of Y14 droplets decreased gradually with increasing pH. On the other hand, Y14 droplets tended to form large amorphous assemblies at pH 6.0 (Fig. 2B). RNase treatment of recombinant Y14 did not affect droplet formation (Supplemental Fig. 1C). Next, we performed light scattering measurements at 400 nm to monitor the dynamics of droplet formation. As the optical density was recorded at 5-min intervals, we observed that Y14 underwent LLPS immediately when the pH changed from 7.9 to the range of 6.0 to 7.0, whereas no droplets formed at pH 7.5 (Fig. 2C, upper), consistent with the DIC observations. Y14 droplets precipitated over time with a rate that varied with pH. Faster precipitation at pH 6.0 coincided with larger droplets or aggregates of Y14 (Fig. 2C, upper). Evaluation during the first 25 min at 10-sec intervals revealed that the turbidity of 20 µM Y14 at pH 6.5 peaked after ∼5 min of incubation (Fig. 2C, bottom), indicating that pH alteration immediately triggered Y14 droplet formation. Y14 droplets that formed at pH 6.5 dissolved when the pH was adjusted to 8 and reformed after the pH returned to 6.5, and this cycle could be repeated (Fig. 2D). Reversible formation of Y14 droplets upon a pH change from 6.5 to 8 suggested a pH-dependent Y14 phase separation. Therefore, mild acidic conditions might promote Y14 droplet formation by reducing electrostatic repulsion. The observation that the phase behavior of Y14 could cycle with pH supported the idea that electrostatic interactions contribute to LLPS of Y14.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 2.

Y14 forms phase-separated droplets in a pH-dependent manner. (A) Recombinant Y14 containing the S peptide (S-tag) and hexahistidine (6 × His) tag. All recombinant Y14 variants contained these two tags. (B) In vitro LLPS of 20 µM recombinant Y14 was performed in buffers containing 50 mM KCl at different pH values for 2 h. DIC images show Y14 droplets (400 × 400 pixels; 1 pixel = 0.114942 µm); inset shows magnified images (100 × 100 pixels) of a selected area. Bar, 5 µm. The experiment was repeated more than five times (N > 5). (C) Turbidity assay showing the kinetics of LLPS of Y14 (20 µM) as in B. Turbidity at 400 nm (OD₄₀₀) was measured. Optical density was recorded at 5-min intervals for 3 h (upper). Dashed square: recording was additionally taken at 10-sec intervals for the first 25 min (lower). Mean and standard deviation were calculated for two to four replicates. (D) Y14 (20 µM) was initially incubated in a pH 6.5 buffer. The pH of the buffer was subsequently adjusted by sequential addition of KOH, HCl, and KOH again, as indicated. DIC images were acquired immediately or after a 2-h incubation, as indicated. Bar, 5 µm. N = 2.

Y14 droplets are dynamic and sensitive to salt concentration

Phase-separated droplets have several characteristics including fusion capacity, reversibility and dynamic exchange of components (Guo et al. 2021). The above result demonstrated that Y14 droplet formation could be reversed in response to a change in pH (Fig. 2D). Using DIC microscopy, we observed that two Y14 droplets could rapidly coalesce in 50 mM KCl/pH 6.5 buffer (Fig. 3A). Although Y14 underwent phase separation independent of molecular crowders, the presence of PEG4000 slightly increased the size of Y14 droplets (Supplemental Fig. 2A). Next, we examined the effect of KCl concentration on Y14 LLPS. The size of Y14 droplets increased with increasing KCl concentration up to 100 mM, although higher KCl concentrations inhibited droplet formation (Fig. 3B). Therefore, except for electrostatic interactions, Y14 LLPS may also involve other types of interaction forces. The notion that the aliphatic alcohol 1,6-hexanediol substantially disrupted Y14 droplets (Fig. 3C) implied the involvement of hydrophobic interactions. Conceivably, droplet size was proportional to Y14 concentration ranging from 5 to 50 µM in pH 6.5 buffer containing 50 mM KCl (Supplemental Fig. 2B). To measure the size of Y14 droplets, we conjugated purified Y14 with Alexa 488. Fluorescent Y14 signals overlapped well with the droplets, as observed by DIC and fluorescence microscopy (Supplemental Fig. 2C). Both coverage rate and droplet size gradually increased over time and peaked at ∼3 h and decreased thereafter (Fig. 3D for images and graph, and Supplemental Fig. 2D). For 20 µM Y14 at 3 h, the average droplet diameter was ∼0.9 µm. Together, the above results confirmed the liquid-like property of Y14 droplets, which were sensitive to pH and KCl concentration.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 3.

Characterization of phase-separated droplets of Y14. (A) The LLPS assay was performed using 50 µM Y14 in a pH 6.5 buffer containing 50 mM KCl. DIC images show the time course of a fusion event of Y14 droplets (arrowhead). Bars, 5 µm. (B) Y14 (20 µM) was incubated in a pH 6.5 buffer containing different concentrations of KCl as indicated. Bar, 5 µm. N = 3. (C) Y14 (20 µM) was incubated in a pH 6.5 buffer containing 50 mM KCl in the absence or presence of 1,6-hexanediol. Bar, 5 µm. (D) A mixture of Alexa Fluor 488-conjugated Y14 and unlabeled Y14 at the molar ratio of 1:50 was incubated in a pH 6.5 buffer containing 50 mM KCl for different times. A representative fluorescent image (200 × 200 pixels) shows the droplets after 3-h of incubation. Bar, 5 µm. The coverage rate (the coverage area of fluorescent signals divided by the total area) was measured by the software MetaMorph. Mean size and relative coverage rate (time 0 was set to 1) of the droplets are shown in the dot graphs.

Y14 droplet formation is determined by multivalent electrostatic interactions

Next, we explored how the highly charged IDRs of Y14 contribute to its LLPS. The amino-terminal domain can be roughly divided into three subdomains, namely N1, N2, and N3, that are rich in negatively, positively and negatively charged residues, respectively (Fig. 4A). We generated and purified recombinant Y14 truncations, each of which lacked one or two subdomains (Fig. 4B). Removal of either positively charged domain (N2 or C) disrupted Y14 droplet formation at pH 6.5–7.5 and caused Y14 aggregation at pH 6, indicating that both domains are essential for LLPS of Y14 and cannot compensate each other (Fig. 4C, ΔN2 and ΔC). In contrast, deletion of either negatively charged N1 or N3 resulted in larger droplets at pH 6.5 (Fig. 4C, ΔN1 and ΔN3). Because these two truncations increased the pI value compared with wild-type Y14 (Supplemental Table S2), their droplets were detectable at higher pH values (7 and 7.5). We therefore concluded that the presence of oppositely charged regions was necessary for electrostatic interactions whereas one negatively charged region appeared to be sufficient for LLPS. Surprisingly, removal of both N1 and a large portion of N3, leaving very few negatively charged residues, did not disrupt droplet formation (Fig. 4C, ΔN1N3), and instead increased the droplet size at pH 7.5, in contrast to the full-length and the truncations lacking either N1 or N3 that developed larger droplets at pH 6.5 than higher pH (Fig. 4D). LLPS of ΔN1ΔN3 may involve other types of interaction forces (see Discussion).

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 4.

Charged regions of Y14 differentially influence LLPS. (A) The domain and subdomain organization of human Y14. Negatively and positively charged regions are indicated in red and blue, respectively. (B) SDS-PAGE gel showing purified recombinant full-length and truncated Y14 proteins. Protein bands were stained with ToolStart Blue Staining Reagent. ΔN2 showed a molecular weight larger than expected perhaps due to its high negative charge density in the amino terminus. (C) Diagrams showing domains of full-length and truncated Y14 proteins. Droplet formation for all Y14 versions (20 µM each) was performed at different pH values. Representative DIC images are shown. Bar, 5 µm. N > 3. (D) The droplet size of Y14 proteins as indicated was measured from the images of C. Mean and standard deviation were calculated for 50 droplets each. (E) DIC images showing droplet formation of wild-type Y14 or Y14-SD in buffer containing 50 mM KCl at the pH values indicated. Representative images (200 × 200 pixels) are shown. Bar, 5 µm. N = 3. Bar graph: the turbidity was measured as in E (mean ± SD; N = 2–4; [**] P < 0.01, [****] P < 0.0001).

Phosphorylation suppresses Y14 LLPS

Y14 has two RS dipeptide repeats in the carboxy-terminal domain. Phosphorylation of Y14 at the RS repeats modulates its protein–protein interactions as well as its subcellular localization (Hsu et al. 2005; Chuang et al. 2013; Tatsuno and Ishigaki 2018). We generated a phosphomimetic mutant of Y14 containing two serine-to-aspartate (SD) substitutions at the RS dipeptides. Y14-SD formed droplets at pH 6.5, albeit slightly fewer than that of wild-type Y14, but droplets did not form at pH 7.0 (Fig. 4E, DIC). This observation was confirmed by the turbidity assay (Fig. 4E, bar graph). This result may reflect the effect of the lower pI value of Y14-SD compared with wild-type or, alternatively, disruption of the charge blockiness of the carboxy-terminal IDR (see Discussion). Nevertheless, this result suggested that phosphorylation of the RS dipeptides may inhibit LLPS of Y14.

Positively charged peptides in trans promote LLPS of truncated Y14

Deletion of the Arg-rich carboxy-terminal domain disrupted Y14 LLPS (Fig. 4C). Therefore, we wondered whether the synthetic carboxy-terminal peptide could in trans promote droplet formation of ΔC. The carboxy-terminal peptide of human Y14 (HsC23), although failed to form droplets by itself, induced LLPS of Y14ΔC when mixed at an equimolar concentration at pH 6.5 (Fig. 5A,B, HsC23). This result was reminiscent of a previous finding that the oppositely charged domains of tau—although individually having low propensity for LLPS—could form droplets after mixing (Boyko et al. 2019). Surprisingly, HsC23 also promoted LLPS of ΔN2, despite N2 containing more lysine residues than Arg residues (Fig. 5A). This observation indicated that positively charged residues in the N2 or C domain are important for LLPS. We noted that, in contrast to human Y14, Drosophila and Caenorhabditis Y14 had a shorter and Lys-rich carboxyl terminus (Fig. 5A). DmC14 could promote droplet formation of both ΔC and ΔN2, although the size of DmC14-induced droplets was smaller than those induced by HsC23 (Fig. 5C). CeC9 barely induced LLPS perhaps because it was too short. The amino-terminal IDR of the human papillomavirus E7 protein, lacking positively charged residues, had no effect on in trans LLPS induction (Fig. 5B, E7N). Our result emphasized the critical role of the positively charged N2 and C regions in LLPS; however, Arg and Lys residues appeared to be interchangeable, at least to some extent.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 5.

Carboxy-terminal peptides of Y14 in trans promote LLPS of truncated Y14. (A) Schematic diagram shows the sequence of N2 and C regions of human Y14 (upper), and the peptides derived from the carboxy-terminal domain of three Y14 homologs and the amino-terminal IDR of HPV E7 protein (E7N20). (B) The LLPS assay was performed with indicated peptide alone (20 µM) or in a mixture containing ΔN2 or ΔC (20 µM) in a 50 mM KCl pH 6.5 buffer. Representative DIC images are shown. Bar, 5 µm. N = 2. (C) The droplet size of truncated Y14/peptide mixtures was measured. Mean and standard deviation were calculated for 50 droplets each.

RNA modulates Y14 droplet formation

RNA acts as a scaffold for several membrane-less organelles and regulates the size of ribonucleoprotein assemblies in different cellular processes such as transcription and stress-granule formation (Fay and Anderson 2018; Maharana et al. 2018; Garcia-Jove Navarro et al. 2019; Henninger et al. 2021). Therefore, we evaluated whether RNA could co-condense with Y14 and thereby modulate Y14 droplet formation. Both DIC and fluorescence microscopy revealed that Cy5-labeled poly(U) 25-mer (U25) RNA colocalized with Y14 droplets (Fig. 6A, wild-type), whereas U25 RNA alone did not form droplets. We previously identified a single point mutation (W73V) of Y14 that substantially decreases its RNA binding capacity (Chuang et al. 2016). In the absence of RNA, Y14–W73V formed a greater number of slightly smaller droplets as compared with wild-type Y14 (Fig. 6A, W73V, no RNA). When Cy5–U25 was added to W73V, fluorescently labeled droplets were barely detected (Fig. 6A, +RNA). A similar result was observed using fluorescent poly(A) 25-mer (A25) RNA (Supplemental Fig. 3A). Therefore, the RNA binding activity of Y14 is critical for its recruitment of RNA to phase-separated condensates. Next, we titrated U25 in the phase separation assay. Consistent with previous findings (Maharana et al. 2018; Garcia-Jove Navarro et al. 2019; Henninger et al. 2021), low levels of U25 promoted phase separation of Y14, whereas high levels of U25 had an inhibitory effect, indicating that RNA influences Y14 LLPS in a concentration-dependent manner (Fig. 6B; Supplemental Fig. 3B). Yeast total RNA, however, inhibited Y14 droplet formation even at very low concentrations (Supplemental Fig. 3C). Considering that cellular RNAs are, on average, >100-fold longer than synthetic polymers, the excess negative charge of RNA may interfere with Y14 LLPS, consistent with previous observations (Maharana et al. 2018).

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 6.

RNA modulates LLPS of Y14. (A) Wild-type or W73V Y14 (20 µM) was mock-incubated (no RNA, DIC) or incubated with 200 nM Cy5-labeled U25 (DIC, Cy5, and merge) in a pH 6.5 buffer containing 50 mM KCl for 2 h. Cy5-labeled U25 (200 nM) alone is shown at right. Bar, 5 µm. Dot graph shows the size of droplets in the absence or presence of RNA (50 droplets were used for measurement). (B) Y14 (10 µM) was incubated with increasing concentrations of U25 RNA (a mixture of Cy5-labeled and unlabeled U25 RNA). Bar, 5 µm. Cy5-labeled U25 (1600 nM) alone is shown at right. Dot graph shows the relative coverage rate of Y14 in the presence of different concentrations of U25. Coverage rate at 25 nM U25 was set to 1. N = 3. (C) EMSA showing Y14 (wild-type or W73V, concentration as indicated) binding to ³²P-labeled HM1.4 (HOTAIRM1 632-882, 17 nM). (D) Wild-type or W73V Y14 (10 µM) was mixed with Cy5-labeled HM1.4 (concentration as indicated) for the LLPS assay. Bar, 5 µm. (E) Wild-type Y14 (10 µM) was mixed with Cy5-labeled HM1.4 (5 to 1600 nM) or full-length HOTAIRM1 (FL-HM1, 5 to 200 nM) for the LLPS assay. Representative images (DIC, fluorescence, merged) show Y14 LLPS in the presence of 25 nM RNA. Cy5-labeled RNAs (200 nM) alone are shown at right. Bar, 5 µm. Dot graphs show the relative coverage rate that was measured over the entire range of concentrations of HM1.4 and full-length HOTAIRM1 (FL-HM1). N = 3.

Y14 interacts with DNA repair factors via the lncRNA HOTAIRM1, of 1044 nt in length (Chuang et al. 2023). Therefore, we tested whether HOTAIRM1 could promote Y14 LLPS under certain circumstances. To determine how Y14 binds HOTAIRM1, we generated five HOTAIRM1 fragments, of which the size was suitable for electrophoretic mobility shift assay (EMSA). The result revealed that Y14 could bind all, albeit with a slightly higher affinity to the fragment of nucleotide 632–882 (namely HM1.4, Supplemental Fig. 3D). Y14–W73V did not bind HM1.4 (Fig. 6C). Cy5-labeled HM1.4 alone did not form droplets, whereas it co-phase separated with Y14 and promoted Y14 LLPS at lower concentrations (Fig. 6D; Supplemental Fig. 3E). Full-length HOTAIRM1 behaved similarly to HM1.4 (Fig. 6E, images). However, at higher concentrations, full-length HOTAIRM1 had a stronger effect than HM1.4 on suppressing Y14 LLPS (Fig. 6E, dot graphs). These results indicated that HOTAIRM1 not only co-condenses with Y14 but also has the capacity to modulate Y14 LLPS—possibly during the DNA repair process.

A potential role for Y14 LLPS in repairing DNA damage

Our in vitro analysis revealed that the carboxy-terminal truncation and W73V mutation abrogated the LLPS capacity of Y14 to different extents. Next, we evaluated the effects of these two mutants on the cellular distribution and function of Y14. Using a fusion protein comprising enhanced green fluorescent protein (GFP) and wild-type or mutant Y14, we examined their intracellular localization and mobility. Confocal microscopy revealed that wild-type Y14 and two mutants (W73V and ΔN2) formed speckles in the nucleoplasm, which were excluded from the nucleoli. In contrast, Y14ΔC showed a less prominent speckled pattern, and in part distributed into the nucleoli (Fig. 7A). Fluorescence recovery after photobleaching (FRAP) analysis revealed that wild-type and mutant Y14 foci were dynamic to different extents (Fig. 7B). All three mutants showed higher mobility than the wild-type. In particular, the fluorescence intensity of ΔC fluctuated, suggesting that it was least constrained.

View larger version:

• In this window
• In a new window

• Download as PowerPoint Slide

FIGURE 7.

Co-phase separation of Y14 and RNA is essential for DNA damage repair. (A) U2OS cells were transfected with an expression vector encoding GFP-tagged Y14 (wild-type, W73V, ΔC, and ΔN2). Confocal microscopy images showing their subnuclear localization. Bar, 10 µm. (B) Graph showing the recovery kinetics of FRAP analysis of GFP-Y14 (as in A). Percentage represents normalized fluorescence intensity (to that of the first post-bleach value). Mean and standard-deviation values were calculated for 18–24 samples. (C) Diagram showing the design of NHEJ-mediated repair of DSBs. DSB was induced by transfection of cells with the Cas9/single guide RNA (sgHPRT) vector. Incorporation of the double-stranded oligonucleotide “Ins” into the DSB sites was measured by PCR using the indicated primers. (D) HeLa cells were transfected with Cas9/sgHPRT (all samples) without or with the indicated siRNA and siRNA-resistant FLAG-Y14 (wild-type or mutant). Genomic DNA was recovered for qPCR using the primer sets I/R and F/R; the latter was used for normalization. Bar graph shows relative PCR products (sg + Ins was set to 1), representing DSB repair efficiency (mean ± SD; N = 2–4; (***) P < 0.001; n.s., no significant difference). Immunoblotting was performed using anti-Y14, which detected endogenous Y14 and overexpressed FLAG-Y14, and anti-FLAG, which detected FLAG-Y14 (endogenous and FLAG-Y14ΔC at the same position on SDS-PAGE). GAPDH was used as the control.

Next, we evaluated whether the LLPS capacity of Y14 contributes to its activity in nonhomologous end joining (NHEJ)-mediated DSB repair. We took advantage of a CRISPR/Cas9-induced DNA cleavage system to measure DSB repair (Fig. 7C; Du et al. 2018). The Cas9 vector coexpressing a single guide RNA targeting HPRT, which encodes hypoxanthine guanine phospho-ribosyltransferase (Cas9/sgHPRT) was transfected into HeLa cells to generate a site-specific DSB in HPRT. Cotransfected double-stranded oligonucleotide (Ins) can be incorporated into HPRT via the NHEJ pathway (Du et al. 2018). We measured Ins incorporation by PCR using the primers complementary to the Ins sequence and HPRT. The level of PCR products represented DNA repair. The DNA repair efficiency was reduced by siRNA-mediated Y14 depletion by ∼50% (Fig. 7D, lane 3), and was restored by cotransfection of cells with siRNA-resistant wild-type Y14 (lane 4) but not by any Y14 mutant (lanes 5–7). This result supported that the LLPS potential of Y14 may contribute to its activity in DSB repair. Nevertheless, the observation that all three mutants, unlike the wild-type, failed to interact with the NHEJ factors and further accumulate on the chromatin after IR (Supplemental Fig. 4) provided a second explanation for their inability to repair DNA. The evaluation of Y14/RNA LLPS at DNA damage foci requires further investigation.

In conclusion, our results demonstrate that Y14 undergoes LLPS through multivalent electrostatic interactions mediated by its highly charged IDRs in vitro and that LLPS of Y14 may in part contribute to its function in DNA damage repair.