Stress-induced nuclear condensation of NELF drives transcriptional downregulation

Abstract

In response to stress, human cells coordinately downregulate transcription and translation of housekeeping genes. To downregulate transcription, the negative elongation factor (NELF) is recruited to gene promoters impairing RNA polymerase II elongation. Here we report that NELF rapidly forms nuclear condensates upon stress in human cells. Condensate formation requires NELF dephosphorylation and SUMOylation induced by stress. The intrinsically disordered region (IDR) in NELFA is necessary for nuclear NELF condensation and can be functionally replaced by the IDR of FUS or EWSR1 protein. We find that biomolecular condensation facilitates enhanced recruitment of NELF to promoters upon stress to drive transcriptional downregulation. Importantly, NELF condensation is required for cellular viability under stressful conditions. We propose that stress-induced NELF condensates reported here are nuclear counterparts of cytosolic stress granules. These two stress-inducible condensates may drive the coordinated downregulation of transcription and translation, likely forming a critical node of the stress survival strategy.

Keywords: CDK9; RNA polymerase II; SUMO; heat shock; negative elongation factor (NELF); pausing; phase separation; proteostasis; transcriptional condensates; transcriptional stress response.

Graphical abstract

Figure 1

Stress induces NELF condensates in nuclei (A) Genome browser tracks showing ChIP-seq occupancy of RNA Pol II, NELFA, and NELFE at RPL14 and RPS3A loci in HEK293 cells. Vertical scale indicates normalized read density in RPM. Top, RNA Pol II occupancy in gene body regions shown at a scale different than for promoter region. NHS, no heat shock; HS, heat shock (43°C). The gene models are shown with exons as boxes. (B) Western blot analysis of indicated proteins after chromatin fractionation from HeLa cells exposed to heat shock (HS) compared to non-heat-shocked cells (NHS). Histone H3 was used for normalization. (C) Fluorescence microscopy images of HeLa cells showing NELFA-GFP, DAPI, and merged signals. NHS, no heat shock; HS, heat shock. Scale bar indicates 10 μm. (D) Fluorescence microscopy images of HeLa cells exposed to heat shock (HS) or recovered for 60 min following HS (Rec) showing NELFA-GFP signal. Scale bar indicates 10 μm. (E) Fluorescence microscopy images of HeLa cells expressing NELFA-mCherry fusion protein exposed to HS. Inset images are magnified versions of demarcated white squares. White arrowheads in the magnified images mark presumed droplet fusion events. Scale bar indicates 10 μm. (F) Fluorescence microscopy images of HeLa cells expressing NELFA-GFP and treated with vehicle or 1,6 hexanediol (1,6 hex). NHS, no heat shock; HS, heat shock. Scale bar indicates 10 μm. (G) Graph showing the number of NELFA-GFP condensates per nucleus in HeLa cells exposed to indicated conditions. NHS, no heat shock; HS, heat shock; 1,6 hex, 1,6 hexanediol. Each dot represents data for one nucleus. Asterisks denote p value of <0.0001 as calculated by the one-way ANOVA with Dunnet’s multiple comparison tests. See also Figures S1 and S2.

Figure 2

NELF complex undergoes liquid-liquid phase separation in vitro (A) Fluorescence microscopy images of purified recombinant NELF complex in buffer with variable NaCl concentration, as indicated. AF488, Alexa Fluor 488. Scale bar denotes 20 μm. (B) Fluorescence microscopy images of purified recombinant NELF complex at variable protein concentrations, as indicated. Scale bar denotes 20 μm. (C) Time series showing fusion of NELF droplets in vitro. t indicates time in seconds. Scale bar indicates 2 μm. (D) Relative quantification of fluorescence recovery kinetics of NELF droplets following partial droplet bleaching. The FRAP curve shows the mean and standard error (light gray) across three independent replicates and was fit to a double-exponential recovery curve (dark gray line). See also Figure S3.

Figure 3

NELF dephosphorylation is required for phase separation (A) Mean fold changes in phosphorylation levels of indicated NELFA residues in HEK293 cells in heat-shocked cells (HS) compared to non-heat-shocked cells (NHS), as detected by mass spectrometry. Error bars represent SD (n = 3 independent cell cultures). Red dotted line denotes the mean phosphorylation level in non-heat-shocked cells. (B) Fluorescence microscopy images of recombinant AF488-labeled NELF droplets that were incubated with P-TEFb containing either active WT CDK9 or a catalytically inactive CDK9 variant. NELF droplets were incubated with P-TEFb for 120 min. Scale bar indicates 5 μm. (C) Relative fluorescence recovery kinetics of dephosphorylated (gray) or P-TEFb-treated (orange) NELF following full droplet bleaching. The curve shows the mean and standard error (light gray or orange) across three experiments and was fit to a double-exponential recovery curve (dark gray or orange). (D) Representative fluorescence microscopy images of the full droplet bleaching experiment shown in (C). Scale bar indicates 2 μm. (E) Heat-shock-induced changes in CDK9-interaction scores of indicated proteins quantified by mass spectrometry in HEK293 cells. A value of 1 denoted by a dotted red line indicates no change in interaction with CDK9 upon heat shock. Error bars represent SD (n = 3 independent cell cultures). Asterisks denote p value of <0.0001 as calculated by one-way ANOVA with Dunnett’s multiple comparison tests. ns denotes non-significant. See also Figure S3.

Figure 4

NELF SUMOylation drives nuclear condensate formation (A) Fluorescence microscopy images of HeLa NELFA-GFP cells treated with vehicle or CDK9 inhibitor 5,6-Dichloro-1-β-d-ribofuranosylbenzimidazole (DRB) under no heat shock conditions (NHS). Scale bar indicates 10 μm. (B) Fluorescence microscopy images of HeLa NELFA-GFP cells exposed to heat shock (HS) and treated with vehicle or ML-792. Scale bar indicates 10 μm. (C) Percentage of cells with NELFA condensates after heat shock in vehicle or ML-792 treated samples. Data quantification of (B). Error bars represent SD (n = 2 independent cell cultures). Asterisks denote p value <0.001 as calculated by two-tailed unpaired t test. (D) Fluorescence microscopy images of heat-shocked HeLa cells expressing NELFA-GFP. Cells were treated with siRNAs against SUMO E2 conjugating enzyme (UBC9), SUMO E3 ligases (PIAS1, PIAS4, and ZNF451), non-targeting siRNA (siNON), or left untreated. Scale bar indicates 10 μm. (E) Percentage of heat-shocked HeLa cells with NELFA condensates from experiment described in (D). Error bars represent SD (n = 2 independent replicates). Conditions compared to siNON. Asterisks denote p value of (^∗∗∗, 0.001; ^∗∗, 0.01) as calculated by one-way ANOVA with Dunnett’s multiple comparison tests. (F) Western blot analyses of NELF after performing in vitro SUMOylation reactions using NELFA (top) and NELFC/D antibodies (bottom). Either immunoprecipitated full-length E3 ligase ZNF451 or a recombinant N-terminal fragment were used at two different concentrations (STAR Methods). Reactions lacking ATP or ZNF451 serve as negative controls. (G) Western blot analysis of ZNF451 and NELFE is shown for indicated fractions from HeLa cells exposed to heat shock (HS) or no heat shock (NHS). SNRNP70 and RPB3 are used as loading controls. (H) Effect of SUMO E2 (UBC9) depletion on HS-induced changes in RNA Pol II ChIP-seq occupancy in gene body regions of top 250 expressed genes. Regions from transcription start site (TSS) +1.5 kb to transcription end site (TES) –0.5 kb are shown. p value from Wilcoxon test is indicated (n = 2 independent cell cultures). (I) Effect of SUMO E3 ligase ZNF451 depletion on HS-induced changes in nascent transcript counts as quantified by SLAM-seq in HeLa cells. Significantly downregulated genes in siNON cells as calculated by differential gene expression analysis based on the negative binomial distribution (DESeq2) are used. The red line indicates no change in expression. The violin plot depicts median, interquartile range, and 95% confidence interval with white dot, black bar, and thin black line, respectively. p value from Wilcoxon test is indicated (n = 3 independent cell cultures). See also Figures S3–S5.

Figure 5

NELF tentacles drive phase separation in vitro (A) Domain architecture of NELFA and NELFE subunits with corresponding disorder prediction using the software PONDR (Peng et al., 2006). pNLS, putative nuclear localization sequence; IDR, intrinsically disordered region; RD, Arg/Asp-rich domain; RRM, RNA recognition motif. (B) Phase separation assays with recombinant GFP-NELFA or GFP-NELFE tentacle fusion protein at low ionic strength (50 mM NaCl). Scale bar corresponds to 20 μm. (C) Phase separation assays with equimolar mixtures of recombinant GFP protein fused to either the NELFA or NELFE tentacle region at 50 mM NaCl. The concentration at which each protein was used is indicated. Scale bar corresponds to 20 μm. (D) Fluorescence microscopy images of concentration-dependent LLPS of purified recombinant double tentacle GFP fusion protein, which contains the NELFA and the NELFE tentacles at C and N terminus, respectively. Scale bar indicates 20 μm. (E) Phase separation assays with the double tentacle GFP protein containing NELFA and NELFE tentacles in the presence or absence of 10% 1,6 hexanediol (1,6 hex). Scale bar indicates 20 μm. (F) Phase separation assays with recombinant purified NELF complex that lacks either the NELFA tentacle (ΔNELFA tentacle) or the NELFE tentacle (ΔNELFE tentacle). Scale bar indicates 20 μm. (G) Scheme of the human RNA Pol II C-terminal domain (hCTD) with its 52 heptad repeats, which is composed of a yeast CTD-like proximal half (1–26 repeats) and a distal half (27–52 repeats). Sequence motifs of the CTD constructs used in the partitioning assays are shown below (Portz et al., 2017). (H) Partitioning of different CTD constructs relative to the NELF droplet phase visualized by fluorescence microscopy. Scale bars correspond to 10 μm. (I) Partitioning of carboxyfluorescein (FAM)-labeled CTD peptides relative to unlabeled NELF droplets visualized by fluorescence and brightfield microscopy. CTD peptides were either unphosphorylated (Unphos CTD) or phosphorylated at tyrosine-1 (Y1P), serine-2 (S2P), or serine-5 (S5P). For better visualization, the fluorescence images are displayed using a multicolor representation (color map representing the minimum [min] and maximum [max] intensity is shown). Scale bars correspond to 10 μm. (J) Quantification of partition coefficients of the different CTD peptides shown in (I). Individual dots (n = 12) represent coefficients calculated from different recorded images. Red lines correspond to the mean, and asterisks denote the p values as determined by the Kruskal-Wallis and Dunn’s multiple comparison tests (^∗, p < 0.05; ^∗∗, p < 0.01; ^∗∗∗, p < 0.001; ^∗∗∗∗, p < 0.0001). See also Figure S6.

Figure 6

NELFA disordered region drives stress-induced nuclear NELF condensation, transcriptional downregulation, and cell survival upon stress (A) Fluorescence microscopy images of HeLa cells expressing GFP fused to either wild-type NELFA (WT) or NELFA containing 140-residue IDR deletion (ΔIDR). NHS, no heat shock; HS, heat shock. Scale bar indicates 10 μm. (B) ChIP-qPCR measuring NELFA occupancy at indicated gene promoters in HeLa cells stably expressing NELFA-GFP protein with either WT or ΔIDR NELFA. “No peak” primer set amplifies a genomic region not expected to bind NELFA and acts as a negative control. The y axis indicates the mean percent of immunoprecipitated DNA relative to starting input material. Error bars represent SD (n = 2, independent cell cultures). Asterisks denote p value (^∗∗∗, 0.001; ^∗∗, 0.01) as calculated by two-way ANOVA with Sidak’s multiple comparison tests. ns denotes non-significant. (C) Effect of deletion of NELFA IDR on HS-induced changes in nascent transcript counts as quantified by SLAM-seq in HeLa cells. Significantly downregulated genes in NELFA-WT cells upon heat shock as calculated by DESeq2 are used. The red line indicates no change in expression. The violin plot depicts median, interquartile range, and 95% confidence interval with white dot, black bar, and thin black line, respectively. p value from Wilcoxon test is indicated. (D) Normalized fraction of viable HeLa cells expressing NELFA-WT or -ΔIDR after exposure to heat shock (HS) for the indicated time and subsequent recovery for 24 h. Error bars represent SD (n = 6 independent cell cultures). Asterisks denote p value (^∗∗∗∗, 0.0001; ^∗∗∗, 0.001) as calculated by two-way ANOVA with Sidak’s multiple comparison tests. (E) Fluorescence microscopy images of HeLa cells expressing mCherry-tagged wild-type NELFA (WT) or NELFA-ΔIDR fused to FUS IDR (Replacement 1) or EWSR1 IDR (Replacement 2). NHS, no heat shock. Scale bar indicates 10 μm. (F) Percentage of cells with NELFA condensates under no heat shock conditions (NHS). Quantification of the experiment described in (A). Error bars represent SD (n = 2 independent cell cultures). Asterisks denote p value of 0.01 as calculated by one-way ANOVA with Dunnett’s multiple comparison tests. (G) RT-qPCR-based nascent-transcript quantification in non-heat-shocked HeLa cells expressing NELFA-WT, IDR replacement 1, or IDR replacement 2. The mean transcript level calculated from n = 4 independent cell cultures is shown for the indicated genes. Asterisks denote p value (^∗∗∗∗, <0.0001; ^∗∗, 0.01) as calculated by one-way ANOVA with Dunnett’s multiple comparison tests. (H) Normalized fraction of viable HeLa cells expressing NELFA-WT, IDR replacement 1, or IDR replacement 2 after exposure to heat shock (HS) for the indicated time and subsequent recovery for 24 h. Error bars represent SD (n = 6 independent cell cultures). Asterisks denote p value (^∗∗∗∗, <0.0001; ^∗∗∗, 0.001; ^∗∗, 0.01) as calculated by two-way ANOVA with Sidak’s multiple comparison tests. ns denotes non-significant. See also Figures S6 and S7.

Figure 7

Model for the regulation of NELF condensation upon stress such as heat shock The interaction between NELFA and NELFE tentacles of NELF complex are essential drivers of the NELF phase separation. The integration of two stress-dependent pathways—sequestration of P-TEFb complex and activation of SUMO E3 ligase ZNF451—results in NELF condensation in nuclei of stressed cells.