Cytoplasmic forces functionally reorganize nuclear condensates in oocytes | Panda Anku

Animals and oocyte collection

Mice

All animal studies were performed in accordance with the guidelines of the European Community and were approved by the French Ministry of Agriculture (authorization N°75–1170) and by the Direction Générale de la Recherche et de l’Innovation (DGRI; GMO agreement number DUO-5291). Mice were housed in the animal facility on a 12-h light/dark cycle, with an ambient temperature of 22–24° Celsius and humidity of 40–50%. Mice used in this study include female OF1 (Oncins France 1; Charles River Laboratories; 8 to 12 weeks old), female C57BL/6 (Charles River Laboratories; 10 to 14 weeks old), and Formin2 (ref. 53) (FMN2+/− and FMN2−/−; 8 to 16 weeks old). FMN2−/− males were crossed with FMN2+/− females to obtain both female Control FMN2+/− and Mutant FMN2−/− mice. At least two mice per experiment were used. Ovaries were extracted from mice as previously described54 into pre-warmed (37 °C) M2 + Bovine Serum Albumin (BSA; A3311, Sigma) medium supplemented with 1 µM Milrinone55, which maintains growing oocytes arrested in Prophase I. Ovarian follicles were punctured with surgical needles to release growing oocytes from antral follicles (end of oocyte growth56,57). Oocytes of different sizes were subsequently collected with a Stripper Micropipette (XLAB Solutions), washed, moved into dishes with fresh medium under mineral oil (M8410, Sigma), mechanically dissociated from follicular cells, and left to stabilize for an hour in the incubator at 37 °C before proceeding with experiments. Fully grown oocytes were washed 6 times with and transferred into fresh M2 + BSA medium without Milrinone to initiate oocyte in vitro maturation (IVM).

Drosophila melanogaster

The following stocks were used: v;; UASp–white-shRNA (BDSC #35573), v;; UASp–capu-shRNA (BDSC #32922), v; UASp–spir-shRNA (BDSC #43161), GFP-spirD (BDSC #24767), spirRP cn1 bw1/CyO (BDSC #5113), b1 pr1 spir2F cn1/CyO (BDSC #8723). BDSC corresponds to Bloomington stock center. Flies were maintained on a standard medium in 25 °C incubators on a 12-h light/dark cycle. The white-shRNA was used as a control for knock-down experiments since white is not expressed during oogenesis. We used the nos-GAL4-VP16 driver (BDSC #64188) to perform knockdowns or overexpression in the germline. Knockdowns were performed at 29 °C to increase the efficiency of the GAL4 driver. 3-day-old females were collected and dissected in oil (10 S, Voltalef, VWR) for live imaging or in PBS1X for fixed experiments and stage 9 egg chambers were selected based on the morphology of the follicular epithelium.

Plasmids, in vitro transcription of complementary RNA (cRNA), and oocyte microinjection

Histone 2B-RFP (pRN3-H2B-RFP) was used to visualize chromatin (gift from C. Tsurumi58). SRSF2-GFP (a.k.a. SC35; NM_011358) was used to visualize nuclear speckles and was purchased from OriGene Technologies (MG202528). FMN2 (pCS2-FMN2-Myc9) was used to rescue cytoplasmic stirring in FMN2−/− oocytes. Plasmids were linearized with appropriate restriction enzymes. The T3 mMessage mMachine (AM1384, Thermo Fisher), SP6 mMessage mMachine (AM1360, Thermo Fisher), and T7 mMessage mMachine (AM13344, Thermo Fisher) transcription kits were used to synthetize capped cRNAs and purified with the RNAeasy kit (Qiagen) as previously described59. SRSF2-GFP and FMN2 RNA were polyadenylated using the Poly(A) Tailing kit (AM1350, Thermo Fisher). RNA concentrations were measured using a NanoDrop 2000 (Thermo Scientific). cRNAs were centrifuged at 4 °C for 60 min at 25,000×g before co-microinjection of 600 ng/µl SRSF2-GFP and 125 ng/µl H2B-RFP cRNA or 900 ng/µl FMN2 into Prophase I oocytes in 37 °C M2 + BSA + Milrinone medium using an Eppendorf Femtojet microinjector. Oocytes were then incubated for 2 h for SRSF2-GFP/H2B-RP cRNA translation or 0 to 5 h for SRSF2-GFP/FMN2 or FMN2 cRNA translation before proceeding with experiments. Exogenous SRSF2-GFP expression profiles were comparable to endogenous nuclear speckles. Note that FMN2 is degraded at meiosis resumption and resynthesized during the first meiotic division48. A potential difference of FMN2 amount at the end of meiosis I in the two rescue conditions (0 and 5 h) prior to meiosis resumption thus cannot explain the extent of each rescue in terms of successful cell divisions. This is rather due to the capacity of FMN2 to rescue cytoplasmic stirring and, consequently, cytoplasmic remodeling, nucleus position, nuclear speckle dynamics, and splicing activity before meiosis resumption.

Pharmacological inhibitors

The following inhibitors were used: 1,6-Hexanediol29,60,61 (Hexanediol; 240117, Sigma-Aldrich) at 1 to 5%, Cytochalasin-D62 (CCD; PHZ1063, Thermo Fisher) at 1 µM, Nocodazole (Noco; M1404, Sigma-Aldrich) at 1 µM, Paclitaxel63 (Taxol; 580555, EMD Millipore) at 1 µM, Pladienolide B42,64 at 10 µM (CAS 445493-23-2, SC-391691, Santa Cruz Biotechnology), and Tubercidin41 at 10 µM (TO642, Sigma-Aldrich). Dimethyl sulfoxide (DMSO; D2650, Sigma) was used as a solvent for all inhibitors except Hexanediol before addition to M2 + BSA + Milrinone medium. Hexanediol crystals were directly dissolved in the medium at 37 °C before incubation with oocytes for 7 min (1% for Coilin and TDP-43) or 10 min (5% for nuclear speckles). Oocytes were washed three times in an inhibitor-supplemented medium and incubated for 60 min (CCD), 60 to 90 min (Noco, Taxol), or 120 to 240 min (Tubercidin, Pladienolide B) before proceeding. For droplet coalescence speed experiments, inhibitors were added 10 min before start of filming. For oocyte division experiments, oocytes were incubated in inhibitor-supplemented M2 + BSA + Milrinone medium for 2 or 4 h before washing out the inhibitors with M2 + BSA medium (without Milrinone) and proceeding with IVM in the incubator or the microscope chamber for up to 20 h.

Immunostainings

Mouse oocytes

Prophase I oocytes were incubated for 2 to 5 min with a 0.4% Pronase (P5147, Sigma) solution in M2 + BSA + Milrinone to dissociate the zona pellucida washed six times with fresh M2 + BSA + Milrinone medium, and maintained in it for 90 additional minutes. For detection of new transcripts, washed oocytes were incubated with 5-Ethynyl Uridine65 (EU; 0.5 mM) from the Click-IT RNA Alexa Fluor-488 Imaging kit (C10329, Thermo Fisher) in M2 + BSA + Milrinone for 240 min. Before fixation, oocytes were washed three times in an M2 medium supplemented with PVP (P0930, Sigma) at 37 °C and placed on coverslips coated with gelatin and polylysine. Oocytes were then fixed without permeabilization in paraformaldehyde (PFA 4%; 18814, Polysciences or 15710, Electron Microscopy Sciences) at 30 °C for 30 min. After a PBS1X wash, oocytes were permeabilized and pre-blocked for 15 min with PBS1X with 0.5% Triton-X (93443, Sigma) and 3% BSA (A2153, Sigma) at room temperature before incubation with primary antibodies overnight at 4 °C followed by an hour incubation at room temperature with secondary antibodies. Primary and secondary antibodies were diluted in PBS1X with 0.2% Triton-X and 3% BSA. Cells were counterstained with DAPI (10 μg/ml; D9542, Sigma) for DNA observation and mounted in Prolong Gold antifade medium (P36941, Thermo Fisher) placed in 250 nm thick chambers (70366-12, Electron Microscopy Sciences) adhered to microscope slides (631–1554, VWR) to avoid oocyte squashing. EU revelation was performed before mounting according to the kit instructions. For F-actin visualization, oocytes were fixed as in (ref. 66) without Pronase treatment, labeled with Phalloidin conjugated with Alexa Fluor-488 (10 U/mL; A12379, Thermo Fisher), and mounted in Vectashield Antifade medium (H-1000, Vector Laboratories).

Drosophila melanogaster oocytes

Ovaries from 3-day-old females were dissected in PBS1X, fixed in PFA 4% for 20 min, permeabilized in PBT (0.2% Triton-X) for 30 min, left overnight with primary antibodies in PBT at 4 °C, washed three times 30 min in PBT, left with secondary antibody for 2 h at room temperature, washed three times 30 min in PBT and mounted in Citifluor medium (Electron Microscopy Sciences) for observation.

Antibodies

The following antibodies were used: rabbit anti-Coilin (1:2000; ab210785, Abcam), mouse IgG1 anti-Fibrillarin (1:60; ab4566, Abcam), mouse IgG1 anti-nuclear pore complex proteins MAb414 (1:1000; MMS-120P-100, Eurogentec), rabbit anti-NPAT (1:700; A302–772A, Bethyl Laboratories), mouse IgG1 anti-PSPC1(1:100; clone IL4, SAB4200503, Sigma-Aldrich), rabbit anti-pSF3b155 (phosphorylated at Thr313; 1:400; clone D8D8V, 25009, Cell Signaling), rabbit anti-SF3b155 (1:600; clone D7L5T, 14434, Cell Signaling), mouse IgG1 anti-SMN1 (1:100; Survival of Motor Neurons, clone 2B1, 05–1532, EMD Millipore), mouse IgG1 anti-SRSF2/SC35 (1:400; ab11826, Abcam), mouse IgG1 anti-TDP-43/TARDBP (1:1500; clone 3H8, MABN45, EMD Millipore), and species-specific Alexa Fluor secondary antibodies (1:400; Thermo Fisher). A recent study45 proposed that the main target of the monoclonal SRSF2/SC35 antibodies is SRRM2 instead of SRSF2/SC35. However, SRRM2 is a spliceosome-associated protein that sharply localizes to nuclear speckles45. We also confirmed that the monoclonal ab11826 antibody we used recognized SRS2-GFP+ speckles in fixed oocytes expressing SRSF2-GFP. The conclusions of our study are thus unaffected by the proposed SRSF2/SC35 antibody discrepancy45.

Microscopy

Mouse fixed and live imaging

Fixed and live mouse oocytes were examined with a Leica DMI6000B microscope equipped with a Plan-APO 40x/1.25 NA oil immersion objective, a motorized scanning deck and an incubation chamber (37 °C), a Retiga 3 CCD camera (QImaging, Burnaby) coupled to a Sutter filter wheel (Roper Scientific), and a Yokogawa CSU-X1-M1 spinning-disk. Images were acquired using Metamorph (Universal Imaging, version 7.7.9.0) with 500 nm z-steps for a total of 40–50 µm in case of fixed cells and 1000 nm z-steps for a total of 35–60 µm in case of live cells. Oocytes were placed in a 35 mm tissue culture dish with cover glass bottom (FluoroDish FD35–100; World Precision Instruments) for video microscopy. Time-lapse 3D images were acquired at different time intervals (∆t = 5 to 12 min) with an exposition time of 500 ms for SRSF2-GFP and 300 ms for H2B-RFP. For high temporal resolution videos, Metamorph stream acquisition mode was used to capture cytoplasmic random stirring (in bright-field) and rapid diffusive dynamics of nuclear SRSF2-GFP droplets or their surface fluctuations (491 nm excitation wavelength) with a ∆t of 0.5 s on a single z-plane focused on the nucleus or droplet center. For correlations between cytoplasmic stirring intensity and nuclear SRSF2-GFP droplet dynamics, oocytes were first filmed for 120 s in a bright-field and immediately followed by 120 s of filming with a 491 nm laser. For IVM experiments, oocytes were filmed in a bright-field with a ∆t of 3 min on a single z-plane focused on the nucleus for up to 20 h. Supplementary Movies 2–4 and 6 were processed in FIJI (smooth and mean filter functions) for visual purposes only.

Optical tweezer setup

The custom-built optical tweezer system was generated by using a near-infrared fiber laser (1064 nm, 5 W, Ytterbium fiber laser, IPG Photonics, Oxford, Massachusetts) with a Nikon CFI Plan Apochromat Lambda, 100X, 1.45 NA, oil immersion objective (Tokyo, Japan) mounted on an inverted Nikon C1 Plus confocal microscope (Tokyo, Japan) as previously described in refs. 67,68. The setup is equipped with a temperature-controllable stage-top incubator (Tokai Hit STXG-WELSX, Gendoji-cho, Japan) that maintains cells at 37 °C in a humidified environment during experiments. Confocal images were acquired using a 488 nm solid-state excitation laser (Coherent, Santa Clara, California) with an ET525/50 bandpass filter (Chroma).

Positions of the trapped cytoplasmic vesicles were detected by recording the light pattern of the outgoing laser in the back focal plane of the condenser with a four-quadrant position-sensitive photodiode (QPD, Hamamatsu, Si PIN photodiode, Product No. S5981). The detected voltage values (V) were converted into displacements and forces by using the conversion value β (µm/V) and the trap stiffness (kappa) (pN/µm)69. Ideally, one would determine the conversion value and the trap stiffness for every trapped vesicle. However, the complexity and heterogeneity of the cellular environment and the occurrence of multiple vesicles being trapped during experiments make it very challenging to do so. We thus determined the value of β by using a polystyrene bead with a diameter of 1 µm to represent the vesicles by following well-established methods69. Briefly, given that the optical trap was stationary, we moved a bead that was stuck to the glass bottom of the cell culture dish across the optical trap laterally by using a piezo stage (Nano-LP100, MadCityLabs, Madison, Wisconsin) while recording the output voltage signals by the QPD at 1000 Hz. Within a specific regime where the voltage output is linearly proportional to the stage displacement, we calculated the conversion value β = 1/0.192 µm/V. The trap stiffness was determined by the Boltzmann statistics method69. Briefly, we recorded the position distribution of a trapped vesicle by the QPD and then fitted a parabola to the logarithm of the distribution. We obtained trap stiffness (kappa) = 12.8 pN/µm for both x and y directions assuming a symmetric trap. The advantage of the Boltzmann statistic method is that the drag coefficient of the trapped vesicle and the viscosity of the cytoplasm are not required.

To apply forces on the nuclear membranes, we used trapped cytoplasmic vesicles that were located close to the membranes. Given that the optical trap was stationary and the oocytes were immobile, settling on the bottom of the experiment dishes, we moved the piezo stage such that the membranes were moved towards the trapping center while the trapped vesicles were displaced away from the center. Due to the trapping forces of the optical tweezers, the vesicles were then translocated towards the trapping center where membranes were localized, thus pushing against the membranes and consequently exerting forces on the membranes. To calculate the forces (F) exerted by the vesicles onto the membranes, we used the displacement of the vesicle according to Eq. (1):

$$F=kappa cdot left(r-{r}_{o}right)$$

(1)

where ({r}_{o})and (r) are the initial and final vesicle positions, respectively.

Fluorescence recovery after photobleaching (FRAP)

For laser ablation experiments on SRSF2-GFP droplets, mouse oocytes were imaged at 37 °C with a Zeiss (Axio Observer.Z1/7) LSM 980 confocal microscope equipped with Airyscan 2, 2 PhotoMultiplier Tubes (PMT), a GaAsP spectral sensor, a Plan-APO 40x/1.3 oil immersion objective, a 488 nm 10 mW laser, and a motorized deck with a temperature-controlled chamber. Images were acquired using Zen 3.0 software. A total of 150 frames of 579 × 579 pixels (53 µm × 53 µm; 16-bit depth) were acquired on single z-planes with a bidirectional scan speed of 6 and 1.43 s intervals for a total of 214 s. Three frames preceded the bleaching (full 488 nm 10 mW laser power) of a fixed 8 µm × 2 µm region of SRSF2-GFP droplet-containing nucleoplasm. The bleached region size was fixed to be larger than the target droplet and thus included the neighboring dissolved SRSF2-GFP phase. Fluorescence recovery was imaged with 1.2% laser power.

Drosophila melanogaster fixed and live imaging

Drosophila egg chambers were imaged with an inverted spinning-disk confocal microscope (Roper/Nikon) equipped with an sCMOS camera, with a 20X/0.75 objective for live egg chambers and a 60X/1.4 oil immersion objective for fixed egg chambers. Images were acquired with Metamorph with 1000 nm z-steps for a total of 30 µm in the case of fixed cells. To monitor cytoplasmic stirring, autofluorescent yolk granules in live oocytes were captured on single z-planes every 7 s using a 405 nm laser.

Quantifications and image analyses

Data were obtained from at least three independent experiments unless stated otherwise. All images were analyzed on Fiji (Version 2.0.0c-rc-69/1.52t). All graphs and statistical analyses were generated using MS Excel (Version 16.16.27) and GraphPad Prism 9.

Cytoplasmic stirring

The cytoplasmic stirring intensity in mouse and Drosophila oocytes was determined by image correlation analyses using a previously published software9 and available on https://github.com/Carreau/OOCytes/tree/0.9. The software measures pixel changes between consecutive images. Raw time-lapse images (∆t = 0.5 s for mouse and ∆t = 7 s for Drosophila oocytes) were first realigned using the Fiji StackReg plugin. In mouse oocytes, bright-field image correlations were calculated in 3 (NSN) or 4 (SN) cytoplasmic regions of 324 µm2. In Drosophila oocytes, image correlations of autofluorescent yolk granules were calculated in 2 to 5 cytoplasmic regions of 441 µm2. Correlation values from different regions within a cell were averaged. For visual clarity purposes, final correlation values were transformed by subtracting the value of each timepoint from 1 to obtain an inverted exponential-like curve.

Mouse oocyte cytoplasmic vector maps were generated by the Spatiotemporal Image Correlation Spectroscopy70 (STICS) plugin previously implemented for detecting cytoplasmic flows in mouse oocytes71 and available on https://research.stowers.org/imagejplugins/. The maps show cytoplasmic flow velocity magnitude and direction. Bright-field time-lapse images (∆t=0.5 s) were converted to 32-bit format, realigned using the Fiji StackReg plugin, and masked specifically around the oocyte contour before launching the STICS map jru V2 plugin with box-checking of output velocities, movie mask use, and a time correlation shift of 3 frames.

Immunocytochemistry

All immunocytochemistry images spanned 40–50 µm (∆z = 0.5 µm) to include the entire ~30 µm wide nucleus and were examined in 3D. Condensate numbers and surfaces were quantified on z-projections covering the entire nucleus for mouse/Drosophila nuclear speckles, on z-projections covering the entire condensate signal for Coilin and TDP-43, and on single z-planes for the nucleolus. The maximal radius of nuclear speckles or nucleoli served to calculate their volume. In Control and mutant SN oocytes from the FMN2 mouse strain, 20 to 30% of cells do not present nuclear speckles or any EU incorporation. Total nucleoplasmic signal intensities of condensate markers, corresponding to the sum of condensed and dissolved phases, were quantified on z-projections covering the entire nucleus and normalized by cytoplasmic background signal intensity.

Total nucleoplasmic signal intensities of SF3b155 and pSF3b155 (pThr313) were quantified on z-projections covering the entire nucleus and normalized by cytoplasmic background signal intensity. The total signal intensity of pSF3b155 (pThr313) in nuclear speckle droplets was measured on z-projections covering single droplets, thus integrating the intensity within the droplet volume, and normalized by an equally sized nucleoplasmic signal of pSF3b155 (pThr313) in a region devoid of nuclear speckles and DNA. Nucleoplasmic pSF3b155 (pThr313) was predominantly composed of either droplet-associated or chromatin-associated signals; signal ratios of droplet-associated to chromatin-associated pSF3b155 (pThr313) were obtained from single z-sections. We documented that: cytoplasmic force intensification in oocytes drove an increase in nuclear speckle droplet mobility all over the nucleoplasm in absence of apparent chromatin association; an important transcriptional drop accompanied chromatin compaction in these final stages of oocyte growth; and that the large majority of active splicing signal by the end of growth occurred in nuclear speckle droplets, which are compartments proposed to be sites of spliceosome accumulation for post-transcriptional splicing completion33,34. We, therefore, speculate that, by the end of oocyte growth, splicing activity is predominantly post-transcriptional.

Cytoplasmic actin filament (stained with Phalloidin) density was quantified with the Fiji Tubeness plugin. The plugin calculates a score of how much local pixels represent a tube based on the eigenvalues of the Hessian matrix, with high-intensity pixels corresponding to tubular structures and thus, correspond to actin filaments. The plugin also applies a watershed algorithm to measure the average size of the signal void in between the filaments. We verified the expected inverse relation between actin filament density and signal void in our data. Twenty measurements of cytoplasmic regions of 10 µm × 20 µm were made per oocyte on different apico-basal z-planes to maximally cover the entire cytoplasm.

Oocyte volumes were calculated with measured radii values. Nuclear volumes were defined by the external boundaries of the dissolved-phase SRSF2/SC35 signal that occupies the entire nucleoplasm. Briefly, the nucleus was segmented from z-stacks with a Fiji plugin based on signal intensity thresholding of individual z-planes and reconstructed as a 3D-ellipsoid by calculating the minimal-volume enclosing ellipsoid of the thresholded pixels; center position and volume were then calculated from the fitted ellipsoid. The surface occupied by chromatin was quantified on z-projections covering the entire DAPI signal.

In summary, numerous parameters tested in fixed Control and FMN2-mutant oocytes were comparable. This includes cell growth (i.e., cell size, which is also a readout of transcript accumulation in oocytes), nucleus volume, nuclear lamin A/C and g-actin levels11, total nuclear protein levels of diverse nuclear condensate markers, the total volume of condensates per nucleus, and EU incorporation as a readout of nascent RNA transcription. This indicates that these parameters are not implicated in splicing changes observed in late-growth FMN2−/− oocytes.

Droplet tracking

Time-lapse images of oocytes expressing SRSF2-GFP were corrected for bleaching with the histogram matching method and realigned with the Fiji StackReg plugin before proceeding with the SRSF2-GFP droplet center tracking using the Fiji Manual Tracking plugin. Temporal Mean Square Displacements (MSD) were calculated from 20 s droplet trajectories. Curves were fitted with the Nelder-Mead method using R software to estimate the diffusion exponent alpha (α). As the diffusion was found to be anomalous (α < 1), we measured the “effective” diffusion coefficient to be able to compare the different conditions. The effective diffusion coefficient Deff was calculated from a linear fit on the 40 first points (20 s) of the temporal MSD curve and normalized by droplet size (Deff in μm2s−1 x (frac{3}{2}pi r) in μm). Although chromatin in fully-grown SN FMN2−/− oocytes is slightly less condensed than in Controls11 and may thus interfere with droplet diffusion, Taxol or Nocodazole treatments did not affect the observable chromatin condensation state when compared to Controls, thus eliminating the potential chromatin-based bias. In long timescale 3D-tracking, droplet trajectories were corrected for drift with the nucleus centroid trajectories, and displacement distances were squared in graphs. For SRSF2-GFP droplet number and size evolution in NSN oocytes, droplets were manually counted in 3D on every timepoint, and maximal droplet surfaces were measured on first and last timepoints. Chromatin (H2B-RFP) surface evolution was quantified in the same NSN cells on 40μm z-projections covering the whole signal and on every timepoint.

Droplet surface fluctuations

Droplet contour evolution in time was measured with a custom-built plugin Radioak for use in Fiji and available on https://github.com/gletort/ImageJFiles/tree/master/radioak. The plugin extracted the values of radii of a given selection for all angles around the selection center. Shape variation over time was measured by comparing the value of the radius relative to its average value for each angle. In Fig. 4a (upper right), contour fluctuation dynamics were represented visually with a color code: orange if the radius increased or decreased relative to the previous timepoint and white if stable when compared to a given threshold.

In all conditions, SRSF2-GFP droplets of comparable sizes (radius range from 2 to 2.7 μm) were selected. Time-lapse images spanning 15 s (∆t = 0.5 s) were cropped, realigned with the Fiji StackReg plugin, smoothened, and signal thresholded (dark background; default method). The generated binary droplet mask was then analyzed using the Analyze Particles option and the output was saved in zip format for Radiak. Radiak was subsequently launched with 1° angle increments from 0 to 360° (input of 360) and the radii values were extracted. For each angle, the mean radius “R” overall 30 timepoints was subtracted from the droplet radius “r” for each timepoint. The variance (r-R)2 in μm2 corresponds to the measure of droplet surface fluctuations that was plotted.

FRAP

FRAP image sequences were realigned with Fiji StackReg and SRSF2-GFP droplets within the region of interest were selected for fluorescence recovery analysis. Recovery was measured as fluorescence intensity of photobleached droplet corrected for background and normalized by the droplet’s prebleach intensity. Recovery curves were rescaled from zero (bleach value) to one (100% recovery) and were fit to a simple exponential function (one-phase association) with Prism. Similar to (ref. 72), the recovery timescale (tau; τ) was extracted to calculate the apparent diffusion coefficients (Dapp) of droplets with a radius r as Dapp ~ r2/ τ. Mobile fractions were determined as fluorescence intensity recovery fractions at 120 s for NSN and at 60 s for SN oocytes, which correspond to Control curve plateaus.

Biophysical model

Here, we briefly present a theoretical framework that provides a physical basis for interpreting the observations reported in this study. In (ref. 9), we described the oocyte cytoplasm as a fluid actively driven out of equilibrium by actomyosin-based mechanical forces. This activity, whereby chemical energy was converted to random active forces, triggered the active diffusion of tracer particles (vesicles) in the cytoplasm. This could be quantified by the mean squared velocity <v2> obtained from PIV analyses and was found to be larger than classical thermal diffusion. We argued in (refs. 9, 11) that over diffusion timescales across the oocyte, the cytoplasmic activity could be interpreted as an effective temperature in the cytoplasm defined by Tc 2>. In (ref. 11), we proposed that cytoplasmic activity enhanced fluctuations of the nuclear membrane, which acted as a mechanical transducer, transmitting cytoplasmic active forces to the nuclear interior in the form of effective stirring of the nucleoplasm. This stirring resulted in the active diffusion of chromatin and the chromatin-embedded nucleolus11, and could again be interpreted as an effective temperature defined by Tn n2>, where vn is the velocity of intranuclear tracer particles.

Consistent with these earlier findings, we report here that cytoplasmic activity triggered the active diffusion of micron-scale nuclear liquid-like condensates, thereby significantly accelerating their coalescence dynamics. The following scaling argument supports this mechanism. As a readout of the cytoplasmic active forces that are effectively transduced to the nucleoplasm, we use the instantaneous velocity of the nuclear membrane, which can be accessed experimentally: vnm~ 0.3 µm/s in SN oocytes. We denote a the typical lateral extension of such active deformation of the nuclear membrane, which acts as a localized fluctuating source that stirs the nucleoplasm, and l~ 2.7 µm its mean amplitude, measured experimentally in SN oocytes. Assuming, in a first approximation, a viscous nucleoplasm, a single fluctuating active source at the nuclear membrane acts as a Stokeslet of random direction and intensity and induces a long-range fluctuating flow in the nucleoplasm such that <vst2(r)>~vnm2 a2/r2, where r denotes the distance from the source, and <.> the average over the fluctuating dynamics of the active source. Note that here ~ means proportionality with a dimensionless coefficient of order 1. We next consider that cytoplasmic activity effectively induces a uniform density c~1/a2 of independent active sources on the nuclear membrane, which is assumed to be spherical. Here independence of the sources assumes that correlations induced by tension and bending of the nuclear membrane are negligible, which holds for sufficient activity. Summing the contribution of all sources, the fluctuations of the nucleoplasmic flow at a given point at a distance r0=uR from the center of the nucleus (of radius R) can be obtained after standard algebra as in Eq. (2):

$$leftlangle {{{{{{rm{v}}}}}}}^{2}left(uright)rightrangle sim frac{{{{{{{rm{v}}}}}}}_{{{{{{rm{nm}}}}}}}^{2}}{u}{{{{{rm{ln}}}}}}frac{1+u}{1-u}$$

(2)

Averaging the next overall points in the nucleus shows finally that cytoplasmic activity, transduced by the nuclear membrane, causes a fluctuating flow within the nucleus such that <vnu2>~vnm2, where numerical prefactors of order 1 are omitted.

Importantly, the resulting flow spans the entire nucleus (independently of its radius), and has the same order of magnitude as the active sources localized at the nuclear membrane. The impact of such fluctuating flow on the dynamics of nuclear condensates can be quantified by the Peclet number Pn = vnu l/Dn, where Dn is the mean value of the diffusion coefficient of nuclear condensates. The above orders of magnitude, together with the measured Dn~0.16 µm2/s in SN oocytes, yields Pn~5, which is consistent with an effective diffusion of active origin. Of note, the same analysis applies to molecular transport in the nucleoplasm, taking instead of Dn the diffusion coefficient Dm of molecular markers (dissolved SRSF2-GFP), which we measured by FRAP in SN oocytes. Similarly, we found Dm ~0.3 µm2/s and thus Pn~2.5, consistent with an effective diffusion of active origin at the molecular scale in the nucleoplasm. Finally, this shows that cytoplasmic activity can be transduced within the nucleoplasm via a simple, passive (i.e., without further energy supply) physical mechanism, which enhances the diffusion of nuclear condensates and nucleoplasmic molecules.

Furthermore, we report that the cytoplasmic activity was also transduced down to molecular scales within condensates. Indeed, the liquid-like condensates displayed increased surface fluctuations (consistent with a large effective intranuclear temperature Tn), as well as a faster turn-over of their molecular content, which supposedly facilitated the biochemical processes hosted by the condensates, which in our case correspond to mRNA splicing. The instantaneous velocity of the droplet interface could be estimated from experimental data as vd~ 0.9 micron/s in SN oocytes. The fact that vd is of comparable order of magnitude to vnu suggests that droplet surface fluctuations are imposed by shear forces in the nucleoplasm, while capillary forces are negligible. In turn, such surface fluctuations drive flows within droplet condensates; the impact of such flows on molecular dynamics can be, as above, quantified by a Peclet number Pd = vd ld /Dd, where ld is the amplitude of fluctuations of the droplet interface and Dd the diffusion coefficient inferred from FRAP experiments in SN oocytes. Experimental values yield ld ~0.5 µm and Dd~0.3 µm2/s so that Pd~1.5, which is consistent with an effective diffusion of active origin at the molecular scale within nuclear liquid-like condensates.

Altogether, our observations suggest an energy cascade triggered by the cytoplasmic actomyosin activity, transduced by the nuclear membrane to the nucleoplasm causing active diffusion and coalescence of liquid-like condensates, and eventually transduced to the liquid-like condensate surface and interior, observed through condensate surface fluctuations and internal molecular turn-over. Based on this physical picture, it can be hypothesized that, due to energy dissipation, the contribution of active forces to fluctuations diminishes along this cascade from the cytoplasm to intranuclear liquid condensate interiors73. While we argued above that such energy cascades could, in principle, occur without extra energy inputs, additional, active mechanisms cannot be ruled out.

Computational models and 3D-simulations

Agents

To simulate the diffusion of nuclear droplets, we adapted 3D agent-based simulations from our previous work12 implemented in C++; also see (refs. 74, 75) for analytical modeling. We chose an agent-based framework for its flexibility, allowing us to test and compare directly different scenarios (chromatin presence, chromatin compaction, nucleus shape…). We defined three different types of agents: nuclear speckle-like (SRSF2+) droplets, a single nucleolus of fixed size, and chromatin-like obstacles. To simplify the model, each agent was represented as a sphere determined by its center position and radius. Mobility analyses of the droplets and the nucleolus in the experiments revealed sub-diffusive motion inside the nucleus and, thus, obstacles were used to simulate molecular crowding in the nucleus. Nuclear obstacle amounts were based on experimental chromatin surface measurements. Potential droplet neo-nucleation and dissolution were excluded from simulations for simplicity. As observed experimentally, droplets could undergo collision-coalescence.

Confinement in the nucleus

All agents were confined inside a static spherical boundary representing the nuclear membrane. The confinement inside the nucleus was modeled as a repulsive force effective as soon as the agent was in contact with the membrane (see refs. 12, 76).

Agent diffusion

Each agent was presumed to diffuse randomly (Brownian motion) and their velocity was dampened by nucleoplasmic friction. Droplet mobility was assumed to obey Stokes’ law after verifying experimentally that the relation between SRSF2-GFP droplet velocity and their size was coherent with this law. In continuity with our previous work11, we modeled the effect of the cytoplasmic stirring activity on the agent diffusion coefficient inside the nucleus as in Eq. (3):

$$D={k}_{B}(T+alpha {T}_{a})/lambda={D}_{0}+alpha {D}_{a}$$

(3)

where Da is the normalized activity-induced diffusion and D0 is the diffusion without activity. The values of D0 and α were estimated by linear regression on the effective diffusion coefficients of the nucleolus relative to cytoplasmic stirring activity shown in the supplementary data of ref. 11.

Agent contacts

Contact between agents depended on their type.

-Obstacle or nucleolus with any agent type (droplet, nucleolus, obstacle): contact between two spheres created a hard-core repulsive force to avoid physical overlap, as in refs. 12, 77. Repulsion strength increased with sphere overlap, accounting for the limited compressibility of the biological objects78.

-Droplet with droplet: droplets coming into close proximity coalesced as soon as they collided, consistent with experimental SRSF2-GFP data. Due to this biological rapidity of SRSF2-GFP droplet coalescence, we rendered coalescence instantaneous in simulations. The coalescing spheres were then replaced by a single sphere with a volume corresponding to the sum of the two original sphere volumes in order to conserve the initial mass.

Agent motion

The overall motion of each agent was determined by the balance of all forces it experienced: its intrinsic motility, its contact with other agents, and its contact with the nuclear membrane12, as in Eq. (4):

$$vec{{v}_{i}}=frac{1}{{n}_{i}}left(mathop{sum}limits_{j}{vec{F}}_{a}left(i,jright)+{vec{F}}_{{ci}}+{vec{B}}_{i}right)$$

(4)

with ηi = 6πriγ as the friction coefficient opposing agent motion and calculated according to Stokes’ law, and where γ is the viscosity of the medium, Fa is the interaction between agents (hard-core repulsion or attraction), Fc is the force of confinement inside the nucleus, and Bi is the Brownian motion with a diffusion coefficient determined by the cytoplasmic activity.

Cytoplasmic activity in Controls ± pharmacological inhibitors and mutants

The main effect of the conditions tested in experiments (Control + Nocodazole, Control + Taxol, FMN2−/−) was assumed to be the shift in cytoplasmic forces transmitted into the nucleus. In simulations, only the value of the cytoplasmic activity was therefore tuned to mirror these experimental conditions. The values for SN oocytes were all based on experimental data11 with Controls being 1, Controls+Nocodazole being 1.9, and FMN2−/− oocytes being 0.2. The values for NSN oocytes were based on the measurements of NSN Control nuclear membrane fluctuations, done as previously11 and shown in Supplementary Fig. 1j, and with a value estimated at 0.55. The activities of the other biological conditions (inhibitors or mutant) for NSN oocytes were assumed to have a Control-like ratio of activity between SN and NSN oocytes corresponding to a factor of 1.8.

Long timescale simulations

In long timescale simulations, we added the possibility for obstacles to adhere compactly to the nucleolus similar to how chromatin condenses around the nucleolus in physiological conditions of oocyte growth. The obstacles were subsequently constrained to a fixed position relative to the nucleolus whose mobility they follow, and were only permitted small random fluctuations around this position. This allowed us to simulate three different nucleus configurations and test the effect of chromatin compaction on nuclear droplet dynamics. The following simulation series with different nuclear states and a wide range of starting point cytoplasmic stirring activities were performed:

-NSN-like state (simulation series 1): obstacles were widely spread in the nucleoplasm and the starting point intensity of cytoplasmic activity was maintained constant throughout the simulation.

-NSN-like to SN-like transition state (simulation series 2): simulations were first in the NSN-like state with all obstacles widely spread and with low agent mobility. Then at 12 h, 40% of obstacles closest to the nucleolus were attracted towards and adhered to the nucleolus, mimicking chromatin condensation that occurs as of the Trans-stage into the SN stage. The percentage of obstacles that adhere to the nucleolus was roughly estimated from experimental measurements of the occupied surface of SN chromatin relative to NSN chromatin. Also, cytoplasmic activity at 12 h was multiplied by 1.8 times to consider the spike of activity measured in experimental conditions during the NSN to SN transition. The transition estimate of 12 h was defined by the largest number of SRSF2 droplets quantified in the nucleus of a Control Trans-staged oocyte (Supplementary Fig. 4a). Due to assimilation of both chromatin condensation and cytoplasmic force intensification, we consider this simulation series with a starting point activity of 0.55 to be the most representative of nuclear droplet kinetics during the physiological end of oocyte growth (Control NSN to SN).

-SN-like state (simulation series 3): 40% of the obstacles adhere to the nucleolus while the remaining 60% remain spread out. Starting point intensity of cytoplasmic activity is maintained constant throughout the simulation.

Droplet dynamics were simulated for 20 h (simulation series 1 and 3) or for 40 h (simulation series 2) with longer-term dynamics (up to 100 h) predicted by a decreasing exponential fit calculated from the last 5 h of the simulations.

FMN2-mutant oocyte specificities of nuclear shape and chromatin compaction

In simulations of the main manuscript, we explicitly model FMN2−/−-like cytoplasmic stirring intensity with Control-like nucleus shapes and chromatin compaction to predict the time necessary to reach a four-droplet state, which corresponds to 59 ± 8 h. However, fully-grown SN FMN2−/− oocytes also present a less spherical nucleus shape due to microtubules nucleated from the microtubule organizing centers11. These SN oocytes also present a ~10% less compact chromatin than Controls11 in a nucleus of comparable volume. Chromatin compaction in mutant NSN and Trans oocytes is comparable to Controls. We, therefore, assessed the effect of these two supplementary differences (nuclear shape and chromatin decompaction) on droplet dynamics in additional 3D simulations. We found that nuclear shape does not affect droplet dynamics in the context of Control-like cytoplasmic activity, as the time necessary to reach the four-droplet state remained comparable (Control-like cytoplasmic activity, nucleus shape, and chromatin compaction t = 15 ± 8 h; Control-like cytoplasmic activity and chromatin compaction with FMN2−/−-like nucleus shape t = 18 ± 6 h). In contexts of FMN2−/−-like cytoplasmic activity, a 10% less condensed chromatin slightly increased the time necessary to reach 4 droplets, and a less spherical nuclear shape increased the time further (FMN2−/−-like cytoplasmic activity with Control-like nucleus shape and chromatin compaction t = 59 ± 8 h; FMN2−/−-like cytoplasmic activity and chromatin compaction with Control-like nucleus shape t = 87 ± 0.5 h; FMN2−/−-like cytoplasmic activity and nucleus shape with Control-like chromatin compaction t > 100 h). Merging all three FMN2−/−-like properties together also significantly increased the time necessary to reach the four-droplet state (FMN2−/−-like cytoplasmic activity, nucleus shape, and chromatin compaction t > 100 h). With the additional layers of complexity, these results provide more precise predictions of droplet dynamics in the most “realistic” version of an FMN2-mutant oocyte. Nevertheless, we excluded these data from the computational section of the main manuscript for simplicity and primary focus on the consequences of cytoplasmic forces on nuclear droplet dynamics with Control-like parameters of nucleus shape and chromatin compaction during oocyte growth.

Parameter values and sources

Agent/Parameter Value Source/Information

Nucleus

Radius 12 µm Experimental measurements

Nucleolus

Radius 4.5 µm ±0.2 Experimental measurements
Friction 1 Empirical to obtain diffusion coefficients similar to experimental measurements of nucleoli in Control NSN oocytes
Nucleus repulsion 30 Strength of confinement; empirical to avoid objects leaving the nucleus while allowing small deformation
Sphere repulsion 30 Strength of hard-core repulsion with other agents

Droplet

Initial radius 0.9 µm ± 0.01 Experimental measurements of nuclear speckles in NSN oocytes
Initial number 55 Experimental quantifications of nuclear speckles in NSN oocytes
Nucleus repulsion 30 Strength of confinement; empirical to avoid objects leaving the nucleus while allowing small deformation
Sphere repulsion 10 Hard-core repulsion with other agents; empirical, with values moderated to allow small deformation
Fusion speed Instantaneous Simplification based on experimental observations
Friction 0.55 Empirical to obtain diffusion coefficients similar to experimental measurements of speckles in Control NSN oocytes

Obstacle

Radius 0.85 µm ± 0.4 Empirical to confine nucleolus and speckle motion and to obtain similar diffusion coefficients and anomalous MSD exponents as in NSN experiments
Initial number 875 Empirical to confine nucleolus and speckle motion and to obtain similar diffusion coefficients and anomalous MSD exponents as in NSN experiments
Nucleus repulsion 30 Strength of confinement; empirical to avoid objects leaving the nucleus while allowing small deformation
Sphere repulsion 30 Hard-core repulsion with other agents; an empirical value equal to the nucleolus value
Friction 5 Empirical to have lower mobility when compared to other agents
Proportion contact 0 initial 0.4 after switch Evaluated from experimental NSN vs. SN chromatin area in experiments

Simulation

Time step 0.01 s For numerical stability
Transition time 12 h Time when simulation state switches from NSN-like to SN-like
Minimal mobility D0 0.01 Calculated by linear regression of nucleolus diffusion relative to cytoplasmic stirring11
Coefficient of mobility α 0.11 Calculated by linear regression of nucleolus diffusion relative to cytoplasmic stirring11

RNA-sequencing, bioinformatics, and RT-qPCR

RNA extraction and sequencing

Mouse oocyte RNA extraction and sequencing were performed previously11 and can be accessed on the Gene Expression Omnibus (accession number, GEO:GSE103718).

Exon usage analysis

For each sample, between 4 and 15 million reads were mapped to the mm10 reference genome using the splice-aware alignment program Hisat2 (ref. 79) (IUC Galaxy wrapper hisat2 v2.1.0). The differential exon usage expression analysis was based on DEXSeq80. Briefly, we used the DEXSeq prepare_annotation.py script from the Conda environment bioconductor-dexseq==1.28.1 to generate pseudo exons (exon-bins) by in silico segmentation of genes using the Ensembl mm10 gene annotation file Mus_musculus.GRCm38.83.chr.gtf. The relative abundance of each exon-bin was calculated with the DEXSeq-Count algorithm by counting all sequenced reads assigned to each exon-bin. Exon-bins related to multiple genes were excluded from further analysis. Finally, the differential usage of exon-bins between FMN2+/− and FMN2−/− conditions was done with DEXSeq (IUC Galaxy wrappers DEXSeq and DEXSeq-Count, v1.24.0.0). Differentially used exon-bins with a Benjamini–Hochberg adjusted p value (Padj) threshold of 0.05 were extracted into sheet 1 of Supplementary Data 1 and selected for further RT-qPCR and exon-bin size analyses. Exon-bin sizes in sheet 2 of Supplementary Data 1 were plotted relative to their over- or under- representation in FMN2−/− oocytes (Padj < 0.05). Exons from protein-coding transcripts for RT-qPCR validation were selected according to high fold changes (>2) and low Padj (<0.05).

Isoform usage analysis

Transcript isoform abundance quantifications were obtained with RSEM81, which uses Bowtie2 (ref. 82) for alignment with a custom index built from the mm10 reference genome and the GTF file Mus_musculus.GRCm38.83.chr, and then computes the expression abundance for each isoform (RSEM, ARTbio Galaxy wrapper, Version 0.9.0). Isoform differential expression analysis between FMN2+/− and FMN2−/− conditions was performed for expressed genes (TPM >1) with more than one isoform using the R package IsoformSwitchAnalysisR83 (iSAR), which uses counts generated by RSEM and the DEXSeq algorithm to calculate the differential usage of isoforms. Supplementary Data 2 contains the lists of differentially used isoforms with an FDR-adjusted P-value (False Discovery Rate) threshold of 0.05, alternative splicing events per transcript, splicing coordinates, and isoform switch consequences. Differential isoforms generated by exon skipping from a single gene similarly expressed in Controls and Mutants were chosen for RT-qPCR validation of alternative splicing. Using this approach, we detected 1259 transcripts with altered splicing patterns in FMN2−/− oocytes (Supplementary Data 2), which probably start appearing by the end of growth as of the Trans-stage when splicing changes assessed by pSF3b155 become visible (Fig. 4d, e). This number of transcripts is coherent with studies reporting ~7000 to 11,000 expressed genes during the whole process of mouse oocyte growth84,85.

Enrichment and spatial correlations of SRSF1 and SRSF2 binding sites

SRSF1 and SRSF2 binding sites were obtained from ref. 38 for binding site enrichment analyses. The numbers of SRSF1 and SRSF2 binding sites in differentially used exons (DEXSeq) or transcripts (iSAR) were then compared to the number of SRSF1 and SRSF2 binding sites in all genes. Correlation tests of distance metrics between genomic features were implemented using the GenometriCorr package86, which computes spatial association of datasets to highlight potentially relevant relationships between them. We compared sites of differential exon usage (DEXSeq) or alternative splice sites (iSAR) with binding sites of SRSF1, SRSF2, MBNL3, and YY1(38,39,40). Two in silico controls were also generated (Prom50 and Term50), which correspond to the first and last 50 nucleotides of all RefSeqNCBI transcripts, respectively. All Jaccard measures were significant (P < 0.02) except for the overlap between differentially used exon sites and Prom50 sites. When necessary, datasets were converted to the Ensembl mm10 gene annotation using the IUC Galaxy wrapper crossmap.bed v0.5.2 + galaxy0 (http://crossmap.sourceforge.net).

Gene ontology and analyses of the translational status of transcripts

The list of genes affected by differential exon (DEXSeq list) or isoform usage (iSAR list) were merged for functional enrichment analyses and for comparative analyses with transcripts translated during the first meiotic division (Supplementary Data 4,5). Biological processes and Gene Ontologies were analyzed using Enrichr87,88 available through the web interface https://maayanlab.cloud/Enrichr/. The full lists of translated, activated (engaged in translation), and repressed (degraded) transcripts were kindly shared by Marco Conti and based on their previous work46. Their probe set GeneID’s were updated and converted using the Mouse Genome Informatics website (http://www.informatics.jax.org/batch) before comparison with the FMN2−/− list of genes shown in Supplementary Data 5.

RT-qPCR

Total cellular RNA (25 oocytes per sample) was extracted with the RNAqueous-Micro Total RNA Isolation Kit (AM1931; Thermo Fisher) following the manufacturer’s protocol and eluted into 20 μl of elution buffer before DNAase I treatment. cDNA was synthesized using the iScript Reverse Transcription Supermix (1708840; Bio-Rad) and the quantitative PCR was performed in triplicate with primer pairs (listed below) in a CFX96 Touch Real-Time PCR Detection System (Bio-Rad) followed by an analysis of the CFX Maestro Software (Bio-Rad). Samples were normalized to that of Rpl19 (Ribosomal protein L19) and Gapdh (Glyceraldehyde 3-phosphate dehydrogenase), and relative fold changes were calculated using the 2−∆∆C method. For DEXSeq exon usage validation with RT-qPCR, we designed the two primers against the same exon detected as differentially used by DEXSeq. The expression of the total transcript was verified (except for Kctd20) with primer pairs against junctions of constitutive exons, which were not classified as differentially used by DEXSeq. The general primer design strategy is illustrated in sheet 2 of Supplementary Data 1. For iSAR isoform usage validation with RT-qPCR, we designed isoform-specific primer pairs against exon 2–3 and exon 4 for the long isoform and exon 2 and exon 3–6 for the short isoform, as illustrated in sheet 5 of Supplementary Data 2.

The following primers were used:

Ercc5 (ENSMUSE00000812268) forward primer (5′to3′) F=TCTAAGGAGAGGAACTCAGGGG, reverse primer (5′to3′) R=TCTGCTAGATCATCACTGCTGC; Ercc5 (NM_011729.2) F=GCGTCCTTTATCCTAACGGGA, R=GCCAAATGCTAATATCCACGGC; Hnrnpdl (ENSMUSE00000825422) F=TCTATCTCTGGGGGTCGCAC, R= CTTTACGCTGGTACATGAAGTTGG; Hnrnpdl (XM_036165266.1) F=ATAGGTTCTGGGAAGTGCGA, R=TTGGTTCCAGTTTTGGCCCT; Kctd20 (ENSMUSE00000788494) F=AGATCAAGAGGAGACCTGGCG, R=ACATGGCGACTCTTTCCTTCC; Ncapg2 (ENSMUSE00000266378) F= TCATCCATGTCATCCGCCAC, R= TGGCATTCTCCTTCTCGCATT; Ncapg2 (NM_133762.4) F= GGACCTGATGCAGACTACGG, R=AGGGAGCCTTACAACCCCAG; Ogdh (ENSMUSE00001059524) F=TTAAGGCCATTGACAGCCTCC, R=TACAGGTGCAGAATAGCACCG; Ogdh (NM_001252283.1) F=CCCCTTTCCCTGAGTCGAAG, R=TGGTGACCCCTGACCTGATA; Exosc1 isoform 1 (NM_025644.4) F=AGAATGGCGCGGTTCCC, R=GGCAAACCGTGAGTTGATGC; Exosc1 isoform 2 (NM_001164561.1) F=TGAAGACCAGCGAGAATGGC, R=AATTTCTACCTTACAGGTGACGAC.

Leave a Comment