Stress-dependent miR-980 regulation of Rbfox1/A2bp1 promotes ribonucleoprotein granule formation and cell survival

Abstract

Upon stress, profound post-transcriptional adjustments of gene expression occur in spatially restricted, subcellular, membraneless compartments, or ribonucleoprotein (RNP) granules, which are formed by liquid phase separation of RNA-binding proteins with low complexity sequence domains (LCDs). Here, we show that Rbfox1 is an LCD-containing protein that aggregates into liquid droplets and amyloid-like fibers and promiscuously joins different nuclear and cytoplasmic RNP granules. Using Drosophila oogenesis as an in vivo system for stress response, we demonstrate a mechanism by which Rbfox1 promotes cell survival. The stress-dependent miRNA miR-980 acts to buffer Rbfox1 levels, since it targets only those Rbfox1 transcripts that contain extended 3'UTRs. Reduced miR-980 expression during stress leads to increased Rbfox1 levels, widespread formation of various RNP granules, and increased cell viability. We show that human RBFOX proteins also contain multiple LCDs and form membraneless compartments, suggesting that the RNP granule-linked control of cellular adaptive responses may contribute to a wide range of RBFOX-associated pathologies in humans.

Fig. 1

Rbfox1 levels depend on miR-980 and stress. a Schematics of ovaries and a single ovariole (lower panel), where the germline (blue) is surrounded by the somatic follicular epithelium cells, in which at st.7, Notch signaling-dependent mitotic-to-endocycle and at st.10b, endocycle-to-amplification transitions occur. These phases are distinguished by Cut (red), expression of which is repressed by activated Notch signaling. Upon stress, egg production is sharply terminated at the nutritional stress checkpoint, resulting in small ovaries. Images of sated and 10-day protein-starved Control and miR-980 ovaries. Starvation response is not as robust in miR-980 mutants, their ovaries are larger and contain later stage egg chambers. b miR-980 overexpression significantly reduces expression of luciferase reporters (P1 and P2), containing parts of the extended Rbfox1 3′UTR with different miR-980 binding sites. c miR-980 ovaries have significantly increased Rbfox1 mRNA levels. d Nutritional stress (2-day) influences miR-980 and Rbfox1 expression in ovaries: miR-980 is significantly downregulated, while Rbfox1 is elevated. e Rbfox1 levels in wild type (WT), miR-980, and Rbfox1-overexpressing (hsFlp; act > CD2 > Gal4 UAS-GFP/UAS-Rbfox1-RE) ovaries from sated and protein-deprived animals detected by western blotting. Colored arrows point to Rbfox1 isoforms that were used for Rbfox1 expression quantification shown in the bar graph. f Rbfox1 expression pattern is temporally and spatially dynamic. At mitotic stages (st.3–6), Rbfox1 is nuclear; at early endocycling (st.6–8), its levels are strongly decreased; at late endocycling (st.9–10a), nuclear Rbfox1 reappears. At amplification stages (st.11–14), Rbfox1 nuclear expression is strongly reduced, but the moderate cytosolic remains. g In miR-980^Ex1-2, Rbfox1 is elevated in comparison to Control egg chambers of the corresponding stages. h Assembly of Control and miR-980 ovarioles under normal and starvation stress conditions. Delayed termination of egg production in miR-980 mutants is rescued by Rbfox1 downregulation. i Quantification of ovarioles with no response to starvation. Introduction of one copy of Rbfox1 mutation to the miR-980 loss-of-function background rescues the phenotype. DAPI (green), Cut (blue), Rbfox1 (red, rabbit anti-Rbfox1). f, g maximum intensity projections of multiple Z-sections. Scale bars 5 µm. AVE ± SD are reported from triplicates, two-tailed Student’s t-test was applied for statistics in b–d, and Chi-square test in i. See Supplementary Tables 1–3. ** p ≤ 0.01, *** p ≤ 0.001

Fig. 2

Rbfox1 is a pro-survival factor that regulates follicular epithelium differentiation. a In starved females, Rbfox1-overexpressing clones (Rbfox1 OE, GPF-positive, hsFlp; UAS-Rbfox1-RE/ + ; act > CD2 > Gal4 UAS-GFP) have higher cell density and decreased cell death. Note condensed DNA (separate channel, arrows) in non-clonal control cells, which die in response to starvation. b Abnormal autophagy measured by the significantly increased Atg8 intensity in Rbfox1 OE (n = 27) versus neighbouring control cells within the same egg chamber (n = 27); c Decreased apoptosis in Rbfox1 OE cells detected by the reduced Caspase3 intensity (Rbfox1 OE, n = 28, Control, n = 35). d Even without starvation, Rbfox1 overexpression reduces cell growth, illustrated by the significantly decreased Myc expression in Rbfox1 OE clones in comparison to the equal size of the non-clonal neighboring area within the same egg chamber (n = 17). e Rbfox1 OE cells express higher levels of the cell adhesion proteins Cadherin (DE-Cad, n = 10 clones) and β-Catenin (Arm, n = 5 clones), and form a multilayered epithelium (arrow, f). g, h Temporal regulation of the Notch-dependent transcription factor Cut is abnormal, showing that activation of Notch signaling required for proper follicle cell differentiation is slowed-down upon Rbfox1 overexpression (g). i When Rbfox1 is downregulated, Cut expression is also abnormal. h Bar graph and scheme show that timely repression or activation of Cut is delayed in the endocycling (st.6-10a) or amplifying (st.10b-14) mutant follicular epithelial cells, respectively. Cut expression during endocycling stages was detected in 20% and 80% of clones that have higher or lower Rbfox1 levels, respectively, in comparison to 0% in control non-clonal cells. At the same time, significantly higher percentages, 80% and 70% of clonal cells with overexpressed or downregulated Rbfox1, failed to turn on Cut expression during amplification stages (n = 19 and 43 clones for Rbfox1 OE and Rbfox1 KD, respectively). j Scheme represents miR-980/Rbfox1 function in response to nutritional stress. Upon stress, miR-980 levels are reduced, causing Rbfox1 upregulation and promoting cell survival. Images in a–g and i are maximum intensity projections of multiple Z-sections. Scale bars 5 µm. Developmental stages (st.) of egg chambers are marked. Relative antibody staining intensity (AVE ± AD) is presented as relative to control (b–e). Student’s t-test was applied for statistics in b–e and Chi square in h, *p ≤ 0.05; **p ≤ 0.01; ***p ≤ 0.001

Fig. 3

Upon stress Rbfox1 aggregates in higher-order subcellular structures in cells. a Western blot demonstrates the behavior of Rbfox1 protein in response to b-isox treatment. Rbfox1 precipitation is seen in both control and Rbfox1-overexpressing ovarian lysates at the 100 μM compound concentration. Note the relocation of Rbfox1 from supernatant to pellet. Asterisks show various Rbfox1 isoforms precipitated by b-isox. b Aggregation of Rbfox1 protein extracted from ovaries analyzed using SDD-AGE. As a negative control, protein extracts were boiled at 95 °C for 10 min. Note that in the starved control sample, some of the Rbfox1-positive aggregates remain stable at high temperatures. Rbfox1 forms SDS-resistant particles under all tested conditions. Starvation stress increases the levels of SDS-resistant particles. In miR-980 ovaries and during starvation, larger amounts of Rbfox1-positive SDS-resistant aggregates are detected when compared to well-fed controls. c Under control conditions, in non-transfected S2R + cells, Rbfox1 is found in small aggregates in the nucleus and cytoplasm (mottled pattern). Treatment by sodium arsenite, a stress factor promoting RNP granule formation, resulted in Rbfox1 accumulation in more defined speckles (granular pattern). Application of 1,6-hexanediol, which triggers the dissolution of non-amyloid RNP assemblies, caused Rbfox1 granule dissociation (dispersed pattern). d The frequencies of the observed Rbfox1 expression patterns under different conditions (n ≥ 50 cells for each condition, two biological replicates). Two-Way Tables and Chi-square test were applied for statistics, ***p ≤ 0.001. e Upon overexpression, Flag-tagged Rbfox1-PE protein is predominantly found in the nucleus (nuclear only pattern). Upon sodium arsenite treatment, Rbfox1 cytoplasmic granules appear, which are dissolved into “dispersed” speckles upon 1,6-hexanediol treatment. Colored squares in the upper corners in c and e correspond to the quantified phenotypes in d. f In Rbfox1-overexpressing S2R + cells, starvation (1 h) can result in the formation of distinct Rbfox1-positive fibers (yellow arrow). g, h in situ hybridization that detects poly(A) tails of mRNAs shows colocalization of RNA-containing RNP granules with Rbfox1 (arrows). g stressed S2R + cells expressing Flag-Rbfox1-PE and h stressed non-transfected S2R + cells. DAPI (magenta or blue). Specificity of Rbfox1 aggregation is confirmed by polyclonal guinea pig anti-Rbfox1(GP) and anti-Flag antibodies (green). c–d, f maximum intensity projections of multiple Z-sections, g, h single Z-sections. Scale bar=5 µm

Fig. 4

Upon stress Rbfox1 aggregates in higher-order subcellular structures in ovaries. a, b Upon prolonged protein starvation (1–3 weeks) in Control (a) and miR-980 (b) ovaries, Rbfox1 aggregates in cytoplasmic and nuclear droplets of various size and shape. c, d Sometimes, Rbfox1 can form various phase-separated long fibers in both germline (c lower panel) and somatic follicle cells (d) of miR-980 mutants. e Similar higher order structures can be found in starved mitochondrial stress-linked parkin mutant ovaries. f, g Rbfox1-positive fibers (f) and round granules (g) formed in starved ovaries co-stain with Thioflavin T, a benzothiazole dye that exhibits enhanced fluorescence upon binding to amyloid fibrils. h–j FRAP experiments performed on ovaries of flies with GFP-tagged Rbfox1 protein (Rbfox1^CC00511/TM3). h Images of germline cell at different time points; prebleached (−1 s), postbleached (0 s), and after recovery (7 s). White box (ROI1) outlines an Rbfox1-positive granule subjected to photobleaching, yellow box (ROI2) outlines an Rbfox1-positive granule not subjected to photobleaching, and blue box (ROI3) outlines a background area that was not subjected to photobleaching (n > 50 measurments for each ROI). i Scatter plot for correlation analysis between the initial intensity, bleaching efficiency, mobile fraction, and recovery half-time parameters. j Fitted plots of the FRAP recovery curves show two distinct groups of Rbfox1-positive droplets that differ in their recovery half-times. Note that for FRAP experiments, due to lethality, heterozygous animals that have only one copy of GFP-tagged Rbfox1 protein were used (Rbfox1^CC00511/TM3). This means that only half of the Rbfox1 proteins could be subjected to photobleaching and that non-labelled Rbfox1 proteins would also participate in liquid droplet recovery. Thus, the half-time of fluorescence recovery estimated here is most likely two times slower than the actual recovery half-time. DAPI (nuclei, magenta, a–e); Thioflavin T (amyloid, magenta, f, g); Rbfox1 (green, a–g). White rectangle in f, g outlines the area presented as separate panels (right), arrows point to Rbfox1-positive subcellular aggregates. a, b—ovarioles; c, f, g—egg chamber showing germline and somatic cells with large and small nuclei, respectively; d, e—follicular epithelial cells. Images in a–e are maximum intensity projections of multiple Z-sections, and images in f–h are single Z-sections. Scale bars 5 µm

Fig. 5

Rbfox1 colocalizes with membraneless nuclear organelles associated with RNA metabolism. a In follicle cell nuclei, Rbfox1 is associated with chromatin and enriched in the granular component of the nucleolus. Rbfox1 upregulation in miR-980 mutants modifies nucleolar shape and size. b Quantifications of nucleolar type distribution in st.9-10 follicular epithelial cells (n = 273 cells for Control, 234 for miR-980^Ex1-2 and 92 for Rbfox1 OE). c Increased Rbfox1 expression modifies nucleolar shape and size and CB size and numbers. Rbfox1-overexpressing clone is outlined by yellow line (hsFlp; UAS-Rbfox1/ + ; act > CD2 > Gal4 UAS-GFP). d Most of the st.9-10 follicle cells contain one CB, while Rbfox1-overexpressing cells have 1 to > 10 smaller CBs per cell (n = 199 and 156 cells, respectively). e Rbfox1 association with CBs is enhanced by miR-980 loss and stress. f Percentages of Rbfox1-positive CBs are significantly increased in miR-980 mutants in comparison to Control (n = 82 and 256 CBs, respectively). This tendency becomes even more pronounced upon starvation stress (n = 598 and 262 CBs for miR-980 and Control, respectively). g In the germline, Rbfox1 colocalizes with CBs (blue arrowheads) and this colocalization is enhanced by miR-980 loss and stress. Rbfox1 can be also detected in the nuage of the germline nurse cells (white arrowheads). In starved miR-980 mutants, Rbfox1-positive fibers can be observed (yellow arrows, right panel). Rbfox1 (green); DAPI (blue); Fibrillarin (nucleolus, red, a–c); Coilin (CB, cyan, c, red, e, g). Follicular epithelial cells (a–c, e), germline nurse cells (f). Images in a, b, e–g - single Z-sections, c - maximum intensity projection of multiple Z-sections. Scale bars 5 µm. Two-Way Tables and Chi-square test were applied for statistics in b and f, ***p ≤ 0.001

Fig. 6

Rbfox1 colocalizes with P-bodies and stress granules in the cytoplasm. a In the germline, Rbfox1 co-stains with the P-body marker Pacman. Note that all Pacman-positive granules are also Rbfox1-positive, but not all Rbfox1 foci stain with Pacman. b, c Upon miR-980 loss or starvation, which leads to elevated Rbfox1 levels, the appearance of P-bodies increases. d Upon temperature stress, Rbfox1 associates with the stress granules marked with anti-Hsp70/Hsc70 in the germarium and later stage egg chambers (g. and st.6 shown in lower panels). e Stress granules are undetectable in non-stressed ovaries. f Schematic drawing depicts a summary of subcellular localization of Rbfox1 protein. Rbfox1 is associated with the major membraneless organelles that regulate RNA metabolism, such as nucleoli and Cajal bodies in the nucleus and P-bodies, stress granules and amyloid-like fibers in the cytoplasm. Rbfox1 (green); DAPI (nuclei, blue); Pacman (P-bodies, red); Hsp70/Hsc70 (stress granules, red). Egg chambers are shown in a–e. Developmental stages (st.) of egg chambers are marked, germarium (g.). Images in a–e are single Z-sections. Scale bars 5 µm

Fig. 7

Human RBFOX proteins also contain multiple LCDs. a Protein sequence alignment of RNA recognition motifs (RRM) of human and Drosophila Rbfox1 reveals high conservation. b Schematic view of Drosophila Rbfox1 and human Fox family proteins (RBFOX1, RBFOX2, and RBFOX3). Our analysis predicts that human Fox family proteins also contain multiple LCDs; however, in comparison to human orthologs, Drosophila Rbfox1 has a larger number of LCDs with clearly defined poly-glutamine stretches. c In human fibroblasts, similarly to Drosophila Rbfox1, human RBFOX1 protein precipitates with b-isox in a concentration-dependent manner. Western blot with anti-human RBFOX1 antibodies shows complete precipitation of 43 kDa RBFOX1 upon treatment with 100 μM b-isox. Note the protein is appearing in the pellet, while disappearing from the supernatant fraction

Fig. 8

Conserved biological function of human RBFOX1 as a component of stress-induced subcellular compartments. a In human fibroblasts, RBFOX1 is detected in small cytoplasmic and nuclear speckles in normal conditions. b Treatment by the stress-promoting chemical sodium arsenite induces formation of larger RBFOX1-positive granules. c Moreover, application of 1,6-hexanediol, which dissociates non-amyloid RNP assemblies, causes partial disaggregation of RBFOX1-positive stress granules. d The number of RBFOX1-positive foci increases upon stress. The mean size of Rbfox1-positive speckles significantly increases upon sodium arsenite treatment, while additional application of 1,6-hexanediol reduces their size to control-like values. e, f Late passages of human fibroblast cell cultures have large aggregations of RBFOX1 protein in the cytoplasm and in the nucleus. The yellow arrow points to the nucleolus marked with Fibrillarin in e, white arrowheads point to mitochondria marked by Mic60 in f. g In human neurons, RBFOX1 shows similar behavior; smaller RBFOX1-positive granules are observed in control. h Their size is evidently increased upon stress induction; note the large nuclear and cytoplasmic RBFOX1-positive assemblies. i, Some of these granules show liquid state RNP granule characteristics, as they can be partially disassembled by 1,6-hexanediol. Note that RBFOX1 protein is re-localized from the nucleus to the cytoplasm upon sodium arsenite treatment (g, h). Quantification showed that in control (g), 69.5% had nuclear and cytoplasmic RBFOX1 and 30.5% expressed only cytoplasmic RBFOX1 (n = 23), while among sodium arsenite-treated cells (h), 48.6% had nuclear and cytoplasmic RBFOX1 and 51.4% cytoplasmic (n = 37). a–c, e–f, white arrows indicate associations of RBFOX1 with mitochondria, white arrowheads show RBFOX1 accumulation in the perinuclear area of the cytoplasm, yellow arrow points to the nucleolus. Nucleus is outlined. Images in a–c, g–i are maximum intensity projections of multiple Z-sections, and in e, f are single Z-sections. Scale bar 5 µm. DAPI marks nuclei. d, Speckle sizes are measured using the particle analyzer tool in ImageJ software. Mean speckle area and standard error of the mean are plotted in the bar graph. AVE ± SD is shown, experiments were done at least in triplicate, Student’s t-test was applied for statistics, *p ≤ 0.05

Fig. 9

The miR-980/Rbfox1-mediated stress-responsive signaling cascade. Stress-dependent miR-980 regulates Rbfox1, amending cell survival. miR-980 can target only the portion of Rbfox1 mRNAs with the alternative extended 3′UTRs. This prevents the reduction of Rbfox1 levels below a certain threshold, which is detrimental for cellular homeostasis. Rbfox1 is a LCD-containing protein that phase-separates into liquid droplets and amyloid-like fibers in vivo and promiscuously joins nuclear and cytoplasmic RNP granules. Their assembly depends on Rbfox1 concentration and the presence of alternatively spliced isoforms. Since upon stress or miR-980 loss, the levels of Rbfox1 generally increase, this, via LCD-mediated interactions, stimulates aggregation of other LCD-containing proteins that have specific functions in the nuclear and cytoplasmic liquid organelles. Since Rbfox1 levels globally influence RNA granule dynamics, we propose that the cell survival and differentiation defects observed as a result of Rbfox1 up- or down-regulation could be caused by Rbfox1 function as a liquid droplet-assembling LCD-containing protein