1. Physics of Living Systems
  2. Structural Biology and Molecular Biophysics
Download icon

Nucleolar dynamics and interactions with nucleoplasm in living cells

  1. Christina M Caragine
  2. Shannon C Haley
  3. Alexandra Zidovska  Is a corresponding author
  1. New York University, United States
Research Article
  • Cited 0
  • Views 1,097
  • Annotations
Cite this article as: eLife 2019;8:e47533 doi: 10.7554/eLife.47533

Abstract

Liquid-liquid phase separation (LLPS) has been recognized as one of the key cellular organizing principles and was shown to be responsible for formation of membrane-less organelles such as nucleoli. Although nucleoli were found to behave like liquid droplets, many ramifications of LLPS including nucleolar dynamics and interactions with the surrounding liquid remain to be revealed. Here, we study the motion of human nucleoli in vivo, while monitoring the shape of the nucleolus-nucleoplasm interface. We reveal two types of nucleolar pair dynamics: an unexpected correlated motion prior to coalescence and an independent motion otherwise. This surprising kinetics leads to a nucleolar volume distribution, p(V)V-1, unaccounted for by any current theory. Moreover, we find that nucleolus-nucleoplasm interface is maintained by ATP-dependent processes and susceptible to changes in chromatin transcription and packing. Our results extend and enrich the LLPS framework by showing the impact of the surrounding nucleoplasm on nucleoli in living cells.

eLife digest

The inside of a cell is very organized. Just as bodies contain internal organs, cells contain many different compartments, called ‘organelles’, each with its own specific role. Most organelles are surrounded by a membrane that keeps their contents separate from the cytoplasm, the water-based liquid inside the rest of the cell.

Some organelles, however, are not bounded by a membrane. Instead, they act like tiny drops of oil in water, retaining their structure because they have different physical properties from the fluid around them, a phenomenon called liquid-liquid phase separation.

One such organelle is the nucleolus, which sits inside the cell’s nucleus (a membrane-bound organelle containing all the genetic material of the cell in the form of DNA). The nucleolus’s job is to produce ribosomes, the cellular machines that, once transported out of the nucleus, will make proteins.

Human cells start with 10 small nucleoli in the nucleus, which fuse together until only one or two larger ones remain. Previous research showed that nucleoli form and persist thanks to liquid-liquid phase separation, and they behave like liquid droplets. Despite this, exactly how nucleoli interact with each other and with the fluid environment in the rest of the nucleus remained unknown. Caragine et al. set out to measure the behavior and interactions of nucleoli in living human cells.

Microscopy experiments using human cells grown in the laboratory allowed the positions, size and shape of nucleoli to be tracked over time. This also yielded detailed information about the smoothness of their surface. Mathematical analysis revealed that pairs of nucleoli normally moved independently of each other, unless they were about to fuse, when they invariably slowed down and coordinated their movements. In addition, altering the state of DNA in the surrounding nucleus also affected the nucleoli. For example, when DNA was less densely packed, nucleoli shrank and their surfaces became smoother.

These results build on our knowledge of how cells are organized by showing, for the first time, that the environment within the nucleus directly shapes the behavior of nucleoli. In the future, a better understanding of how cells maintain healthy nucleoli may help develop new treatments for human diseases such as cancer, which are characterized by problems with this organelle.

Introduction

The nucleolus is the largest structure present in the cell nucleus of eukaryotic cells. This membraneless organelle is a site of ribosomal biogenesis and plays a key role in cell cycle progression and stress response (Alberts et al., 2014; Montanaro et al., 2008; Boulon et al., 2010). Nucleoli are composed of RNA and proteins and embedded in the chromatin solution inside the nucleus. They form at specific parts of genome called nucleolar organizer regions (NORs) containing rDNA, which is transcribed inside the nucleolus (McClintock, 1934; Ritossa and Spiegelman, 1965; Wallace and Birnstiel, 1966). At the beginning of the cell cycle a small nucleolus forms at each NOR. These nucleoli later fuse into larger ones, while remaining connected to their NORs in somatic cells (Amenta, 1961; Sullivan et al., 2001).

The lack of a nucleolar membrane has long been intriguing biologists and physicists alike, questioning the physical nature of the nucleolus. Pioneering studies in frogs found that nucleoli in X. laevis oocytes behave like liquid droplets in vivo, as well as when reconstituted in vitro, and suggested that nucleoli form through liquid-liquid phase separation of the nucleolar components in the nucleoplasm (Brangwynne et al., 2011; Berry et al., 2015; Feric et al., 2016). The volume distribution of such nucleoli was in agreement with a diffusion-limited aggregation process with a constant influx of particles (Brangwynne et al., 2011). In addition, the size of nucleoli in the worm C. elegans embryos was found to be dependent on the concentration of nucleolar components in the nucleoplasm which is consistent with the liquid-like nature of the nucleolus (Weber and Brangwynne, 2015). The nucleolar subcompartments, that is the granular and the dense fibrillar components, were also suggested to form via liquid-liquid phase separation (Feric et al., 2016). Recent studies in the fly D. melanogaster suggest that while the nucleolar assembly follows liquid-liquid phase separation, active protein recruitment is also involved (Falahati and Wieschaus, 2017).

Recently, we have shown that human nucleoli also exhibit liquid-like behavior (Caragine et al., 2018). By analyzing the shape fluctuations of nucleolar surface and kinetics of the nucleolar fusion in human cells in vivo, we found nucleolar dynamics to be consistent with that of liquid droplets with very low surface tension γ ~ 10-6 Nm-1 surrounded by highly viscous nucleoplasm of viscosity η ~ 10Pa s (Caragine et al., 2018). Strikingly, it is the nucleoplasm viscosity that sets the time scale for the nucleolar coalescence providing resistance to the already very low surface tension that drives the process (Caragine et al., 2018). Correspondingly, nucleolar coalescence in human cells takes hours to complete (until the newly formed nucleolus rounds up, Figure 1A), while the neck connecting two coalescing nucleoli is discernable only for minutes after their initial touch (Figure 1B) and its radius r grows in time as r(t)t1/2 (Caragine et al., 2018). Such long coalescence times have been speculated not to interfere with the rDNA transcription inside the nucleoli (Caragine et al., 2018).

Nucleolar coalescence.

(A) Time lapse of a nucleus with fluorescently labeled chromatin (H2B-GFP) and two fusing nucleoli (NPM-mApple). Time points depict: pre-fusion (t = 0 min), with two distinct nucleoli, fusion in progress (t = 33 min), with a clearly visible neck connecting the two nucleoli, and post-fusion (t = 273 min), where a resultant nucleolus can be seen still rounding up. (B) Time series showing the growth of the neck connecting two coalescing nucleoli. At t = 0 s, both the fluorescently labeled chromatin (H2B-GFP) and coalescing nucleoli (NPM-mApple) are depicted, the later frames, 20–600 s, show the progress of the nucleolar coalescence (NPM-mApple). Parts of (B) adapted from Figure 3b in Caragine et al. (2018). Scale bar, 2 µm.

The nucleoplasm (chromatin solution) and its physical properties clearly contribute to the nucleolar physiology. Interestingly, while the nucleolar coalescence can be described by a theory of passive liquid droplets within a highly viscous passive fluid (Caragine et al., 2018; Paulsen et al., 2014), nucleoplasm is an active fluid. Specifically, chromatin dynamics was shown to be active, that is ATP-dependent, and coherent, that is exhibiting correlated displacements, over 3–5 µm in human cells (Zidovska et al., 2013). Thus, the measured γ and η are likely effective quantities (Caragine et al., 2018). Chromatin is known to localize as a denser heterochromatin at the nucleolar surface (Padeken and Heun, 2014), yet the nature of physical interactions between the nucleolar surface and the chromatin solution remains to be revealed (Németh and Längst, 2011; Bickmore and van Steensel, 2013; Towbin et al., 2013). Disruption and dysfunction of the nucleolus is implicated in a large number of human diseases, such as skeletal and neurodegenerative disorders, cardiovascular disease and cancer (Hannan et al., 2013; Núñez Villacís et al., 2018; Ruggero and Pandolfi, 2003; Derenzini et al., 2009). Thus, elucidating physical principles governing the nucleolus-nucleoplasm interface might contribute to our understanding of the nucleolus in health and disease.

In this work, we investigate the physical interactions between the nucleoli and the surrounding nucleoplasm by studying the structural features and dynamical behavior of the nucleoli. Specifically, to illuminate the kinetics of nucleolar assembly process, we examine the changes in the nucleolar size distribution with progressing cell cycle. In addition, we probe the physical nature of the nucleolar subcompartments, specifically, the granular components and the dense fibrillar components, and their contribution to the nucleolar liquid-like properties. To elucidate the role of nucleoplasm in nucleolar coalescence, we interrogate size, shape, position and alignment, as well as mobility inside the nucleus for both nucleoli that are about to fuse as well as those that do not fuse. To determine the role of active processes in maintaining the liquid-like nucleolus-nucleoplasm interface, we deplete ATP and further evaluate its structure and dynamics. Finally, we probe the contribution of specific cellular processes (such as cytoskeletal forces, transcriptional activity as well as protein synthesis) to maintaining the nucleolus-nucleoplasm interface by employing targeted biochemical perturbations.

Results

Nucleolar size distribution during the cell cycle

To address the kinetics of the nucleolar assembly process, we have evaluated the number and size of the nucleoli at different times during the cell cycle. After mitosis, human nuclei initially contain 10 nucleoli, which later fuse to form fewer larger ones (Savino et al., 2001). Thus, due to the changing nucleolar number, the likelihood of their coalescence is expected to vary with the cell cycle progression. First, we measure the nucleolar size distribution in an unsynchronized cell population, which contains cells at all cell cycle stages (Figure 2A). Then we obtain the specific nucleolar size distributions at different, well-defined times of the cell cycle by synchronizing the cell population and monitoring their nucleolar count and size with progressing cell cycle (Figure 2B–D). Specifically, we carry out our measurements 1.5 hr and 3 hr after mitosis as well as at the end of the cell cycle, at the G2/M check point (Figure 2B–D). At every time point, we collect data from the entire volume of the cell nucleus by taking a z-stack with focal planes 0.5 μm apart. Figure 2A–D shows micrographs of nuclei with fluorescently labeled chromatin (H2B-GFP) and nucleoli (NPM-DsRed) for all studied populations, respectively. Moreover, Figure 2, insets 1–4, shows an enlarged view of the boxed in nucleus from Figure 2A–D, respectively. While Figure 2, inset 1 depicts a nucleus from an unsynchronized cell population, Figure 2, insets 2–3 show the same nucleus with progressing time. Note, the presence of both small and large nucleoli early in the cell cycle (Figure 2, inset 2–3), with the large ones becoming more spherical between 1.5 hr and 3 hr after mitosis, while only large nucleoli are seen at the end of the cell cycle (Figure 2, inset 4).

Figure 2 with 1 supplement see all
Nucleolar size distribution as a function of cell cycle.

(A–D) Micrographs of HeLa cell nuclei with fluorescently labeled chromatin (H2B-GFP) and nucleoli (NPM-DsRed) under the following conditions: unsynchronized cells (A), synchronized cells 1.5 hr (B) and 3 hr (C) after mitosis, and cells arrested at G2/M checkpoint (D). (1–4) enlarged view of the boxed nuclei from (A–D). (E) Average nucleolar area AN as a function of number of nucleoli per nucleus NN for all conditions from (A–D). For unsynchronized cells, total number of nucleoli analyzed NN = 1331 in 228 nuclei, for t = 1.5 hr, NN = 275 in 42 nuclei, for t = 3 hr, NN = 257 in 51 nuclei, and for t = G2/M, NN = 497 in 124 nuclei. (F) Distributions of nucleolar volume VN and their fit to f(VN)VNα for all conditions from (A–D). For all fits, the goodness-of-fit, R2 > 0.98. Scale bar, 15 µm.

Figure 2E shows the distributions of average nucleolar area of nucleoli in one nucleus, AN, as a function of the nucleolar number in the given nucleus, NN, for the unsynchronized and synchronized cell populations at the studied time points. The distributions of nucleolar area, AN, for each time point are shown in Figure 2—figure supplement 1. For each nucleolus we measure its area in its respective focal plane within the collected z-stack. We find that as the cell cycle progresses, the number of nucleoli per nucleus decreases, while the average nucleolar area in the nucleus increases (Figure 2E). Interestingly, this trend persists beyond 3 hr into the cell cycle suggesting that the fusion of nucleoli is not limited to the first two hours of the cell cycle as previously hypothesized (Savino et al., 2001). To gain further mechanical insight into the nucleolar coalescence kinetics during the cell cycle, we have analyzed the nucleolar volume distribution for each time point (Figure 2F). We calculated nucleolar volume assuming a spherical shape, VN=4πr3/3, where r is the radius of a circle with the area equal to the nucleolar area, and using the least square method we fitted the nucleolar volume distribution P(VN) to a power law f(VN)VNα. Our data shows that P(VN) can be described by a power law with α ~ -1 for all cell populations, unsynchronized as well as synchronized at all studied time points. The confidence intervals for the fitting parameter α are listed in Figure 2F with the goodness-of-fit R2 > 0.98 for all fits. It is noteworthy, that such distribution is divergent, and so is its first moment, the mean, if integrated over all volumes (from 0 to ∞). However, the measured p(V) distribution does have finite bounds given by the physical cut-offs for the nucleolar size, the minimum and maximum that it can reach inside a cell nucleus.

Physical nature of nucleolar subcompartments

The human nucleolus behaves like liquid droplet (Caragine et al., 2018), yet the nucleolar fluid is complex, containing three distinct subcompartments; fibrillar center (FC), dense fibrillar component (DFC) and granular component (GC). They all play a different role in ribosome biogenesis and vary in protein composition: While FC contains polymerase I, DFC and GC are enriched in fibrillarin (FBL) and nucleophosmin (NPM), respectively (Boisvert et al., 2007). Moreover, they show a hierarchical organization, suggested to form via liquid-liquid phase separation (Feric et al., 2016), with FCs nested inside DFCs, which are embedded in GC.

To address the contributions of these subcompartments to the overall liquidity of the human nucleolus, we examine their physical properties. Figure 3A shows micrographs of three different nuclei with fluorescently labeled chromatin (H2B-GFP, green), GC (NPM-DsRed, red) and DFCs (FBL-mCerulean, blue). We obtain the nuclear and nucleolar contours from H2B-GFP and NPM-DsRed signal, respectively. By analyzing the FBL-mCerulean signal we procure the shape, size and number of DFCs inside a nucleolus. A visual inspection of our data reveals that DFCs appear to be close to spherical. To verify this observation, we measure the DFC eccentricity: First, we measure the length of the semi-major DFC axis a and the semi-minor DFC axis b (Materials and methods). Figure 3B displays the distributions of measured lengths of both a (red) and b (green), together with the Gaussian fits f(aDFC)e (aDFCaDFC)2/2σaDFC2 (red line) and f(bDFC)e (bDFCbDFC)2/2σbDFC2 (green line) of their respective distributions. From the Gaussian fits we obtain the following average values: aDFC= 210 ± 50 nm and bDFC= 180 ± 40 nm. Next, we evaluate the eccentricity e=a/b for each DFC and find that the DFC shape is indeed close to spherical with the average eccentricity e= 1.22 ± 0.17 (where e = 1 corresponds to a circle) and average area of ADFC= 0.13 ± 0.06 µm2, where ADFC=πab. The distributions of e and ADFC are shown in Figure 3—figure supplement 1. Overall, we identified 1279 DFCs over 114 nucleoli in 63 nuclei, and after the removal of the DFCs that were out of focus, we obtain measurements of a, b, e, and ADFC for 1035 DFCs.

Figure 3 with 1 supplement see all
Nucleolar internal structure.

(A) Micrographs of HeLa nuclei with fluorescently labeled chromatin (H2B-GFP, green), nucleolar granular component (NPM-DsRed, red) and nucleolar dense fibrillar component (DFC) (FBL-mCerulean, blue) and overlays of all three signals (green, red, blue) and red and blue signal. The insets in overlay images present an enlarged view of a nucleolus from the image. (B) Distributions measured for semi-major axis aDFC (red) and semi-minor axis bDFC (green) of single DFCs (NDFC = 1035). The solid red and green lines correspond to the Gaussian fits of distributions of aDFC and bDFC, respectively, with aDFC ≈ 210 nm and bDFC ≈ 180 nm. (C) Nucleolar area, AN, as a function of DFC number per nucleolus, NDFC, with a linear fit AN ≈ 0.92NDFC. We evaluated 1279 DFCs over 114 nucleoli in 63 nuclei. Scale bar, 5 µm.

Next, we evaluate the nucleolar area, AN, as a function of the DFC number, NDFC, inside the given nucleolus (Figure 3C). Our data reveals that AN grows linearly with NDFC, with a linear fit of AN = (0.92 ± 0.05)NDFC. This implies that upon nucleolar coalescence, which leads to larger AN, the new nucleolus contains the cumulative number of DFCs, indicating that DFCs do not fuse themselves. This is further corroborated by the volume distribution of DFCs, p(VDFC), which has a sharp peak at VDFC = 0.03 μm3 (Figure 3—figure supplement 1), indicating that DFCs are largely monodisperse. Moreover, this finding suggests that every DFC is associated with a GC domain of an area AGC ≈ 0.79 µm2. Since we found DFCs to exhibit a close to spherical shape, we can estimate the volume fraction of DFCs and GC phase in the human nucleolus, and find ΦDFC ≈ 0.1 and ΦGC ≈ 0.9, respectively.

Fusing and non-fusing nucleoli

Our recent study revealed that the timescale of the nucleolar coalescence is set by the high viscosity of the surrounding nucleoplasm (ηnp ~ 10Pa s) (Caragine et al., 2018). To elucidate the physical interactions of nucleolar droplets with the chromatin solution, we interrogate their size, shape, position and alignment inside the cell nucleus. Moreover, we compare these characteristics for nucleoli that fuse and the ones that do not fuse during our observation. For non-fusing nucleoli, we record time lapses for 60 min with a time step of 5 min and at every time step we collect a z-stack with focal planes 0.5 µm apart. By collecting a z-stack, we can monitor all nucleoli present in the given nucleus and obtain measurements in their respective focal planes. To capture a fusion of nucleoli, we observe a pair of nucleoli for 60–270 min with a time step of 5–15 min, and review at the end of the experiment if the fusion has occurred. In three cases, we were able to track three or four nucleoli simultaneously, the nucleoli closest together were then defined as pairs. In case of three nucleoli only one pair was analyzed, that is two closest nucleoli. In one case, a nucleolar pair fused while the measurement was being set up, and was therefore only analyzed in the post-fusion nucleolar population.

Figure 4A shows micrographs of a nucleus with fluorescently labeled chromatin (H2B-GFP) at t = 0 and 60 min, where the nucleoli correspond to the voids in the H2B-GFP signal and are highlighted by symbols (circle and triangle). In contrast, Figure 4D shows micrographs of nucleus with fluorescently labeled chromatin (H2B-GFP) at t = 0, 60 and 120 min, with fusion occurring shortly before t = 60 min. The nucleoli correspond to the voids in the H2B-GFP signal and are highlighted by triangle and cross before fusion and by circle during and after fusion.

Comparison of size, shape and nuclear positioning between fusing and nonfusing nucleoli.

(A) Micrographs of a nucleus with fluorescently labeled chromatin (H2B-GFP), where two void spaces (labeled by yellow triangle and yellow circle) correspond to two nucleoli that did not fuse between t = 0 and 60 min. (B) Schematics of measured variables. (C) Measured variables for nonfusing nucleoli: nucleolar area, AN, nucleolar eccentricity, e, shortest distance from the nucleolar centroid to the nuclear envelope, De, and the angle between the major nuclear and nucleolar axes, α (NN = 17, Ncell = 6). All characteristics are calculated in the nucleolar focal plane. (D) Micrographs of a nucleus with fluorescently labeled chromatin (H2B-GFP), where two void spaces (labeled by yellow triangle and yellow cross) correspond to two nucleoli before fusion at t = 0 min, while at t = 60 and 120 min they can be seen fusing (yellow circle). (E) Measured variables for fusing nucleoli: AN, e, De and α (NN = 12, Ncell = 7). The dashed red line at t = 0 min indicates fusion. All measurements are carried out in the nucleolar focal plane. Scale bar, 5 µm.

Next, we obtain contours for all nucleoli in their respective z-plane and measure their area, AN, by filling their contour. To evaluate the nucleolar shape we compute its eccentricity, e=a/b, with a and b being the semi-major and semi-minor axes of a fitted ellipse, respectively. For e = 1, the nucleolus is spherical, while for e > 1 the nucleolus has an elliptical shape. Further, we determine the shortest distance of the nucleolar centroid to the nuclear envelope, De, as well as the angle between the nuclear and nucleolar major axes, α, when fitted by an ellipse, respectively. Figure 4B provides an illustration of the measured parameters a, b, De and α. First, we evaluate these quantities for the non-fusing nucleoli (Figure 4C). We find that AN, e and De do not change appreciably, while α fluctuates significantly during the duration of the experiment. In fact, a constant area might indicate that there is no significant addition or removal of nucleolar material during this time. The eccentricity is rather low, often close to 1, making α susceptible to small fluctuations.

For comparison, Figure 4E shows the same quantities for the nucleoli that fused during the experiment. We aligned the different fusion events in time with fusion occurring at t = 0 min, as marked by the red dashed line. The pairs of fusing nucleoli are identified by symbols of the same color, with two nucleoli before fusion indicated by a triangle and cross, and the new nucleolus after fusion by a circle. Interestingly, AN upon fusion is about same as the summed area of both pre-fusion nucleoli. Since we do not observe any significant increase of AN after fusion, there is likely no significant material influx associated with nucleolar fusion. As expected, e decreases after fusion for all nucleoli, consistent with our prior finding of surface tension driving the fusion (Caragine et al., 2018). De does not show any leading trends for the position of the new nucleolus; some remain closer to a position of one of the pre-fusion nucleoli, whereas some move into an intermediate De of the two pre-fusion nucleoli. Interestingly, our α measurement shows that nucleoli after fusion are slowly moving into a parallel (α= 0°) or a perpendicular (α= 90°) alignment with the major nuclear axis.

Dynamics of fusing and non-fusing nucleoli

To further elucidate the nucleolus-nucleoplasm interactions, we investigate the dynamics of both non-fusing and fusing nucleoli. We track the nucleolar motion, by tracking its centroid in time, and obtain a trajectory for every nucleolus. By analyzing nucleolar trajectories, we can extract the nucleolar velocity, v, which we measure relative to the nuclear centroid. Further, we evaluate the radial velocity, vrad, which we define as the nucleolar velocity along the line connecting centroids of two nucleoli, with origin being the line center. vrad allows us to assess the motion of one nucleolus towards another, where a negative value of vrad corresponds to motion towards the other nucleolus. We also compute the angle, αv, between the nucleolar velocity v and the line connecting the two nucleoli, with αv ranging from 0° to 180°, informing if a nucleolus is traveling towards the other.

Figure 5A shows examples of trajectories for non-fusing nucleoli, with their temporal evolution color-coded (blue to red). An enlarged view of these trajectories and the areas they cover is depicted in Figure 5—figure supplement 1A–B. A distribution of nucleolar velocities v for non-fusing nucleoli is shown in Figure 5B, with a mean velocity v ≈ (0.49 ± 0.30) × 10-3 μms-1. Next, we look at the dynamic behavior of a nucleolar pair, and analyze their velocities with respect to each other. We plot the larger velocity, vmax, against the smaller velocity, vmin (Figure 5C). The scatter plot in Figure 5C shows a wide spread and no apparent correlation between vmax and vmin (Pearson correlation coefficient ρ= 0.41). Moreover, a distribution of vrad is centered around 0, thus not pointing towards any preferred direction of motion (Figure 5D). This observation is further corroborated, when we find vrad to fluctuate around 0 as a function of time (Figure 5E), as well as αv changing seemingly randomly with time (Figure 5F). By fitting a Gaussian curve to the vrad distribution in Figure 5D we obtain a variance σrad,nonfusing = 3 × 10-4  μms-1.

Figure 5 with 1 supplement see all
Comparison of dynamics between fusing and nonfusing nucleoli.

(A) Trajectories of two nonfusing nucleoli color-coded by their temporal evolution (blue to red). The time step is 5 min. (B) Histogram of the velocity magnitude, v, for the nonfusing nucleoli (NN = 17, Ncell = 6). (C) Velocities for pairs of nonfusing nucleoli, where vmax and vmin is the larger and the smaller nucleolar velocity, respectively. (D) Histogram of radial velocity, vrad, for nonfusing nucleoli, with vrad calculated with respect to the midpoint distance between nucleoli. (E) vrad as a function of time for nonfusing nucleoli. (F) Velocity angle, αv, in polar coordinates as a function of time for nonfusing nucleoli. (G) Trajectories of pair of fusing nucleoli color-coded by their temporal evolution (blue to red). The pre-fusion nucleoli are visible at earlier times (blue to yellow), while the post-fusion nucleolus appears at later times (orange to red). The time step is 15 and 16 min. (H) Histogram of the v for the fusing nucleoli (NN = 12, Ncell = 7). (I) Velocities for pairs of fusing nucleoli, where vmax and vmin is the larger and the smaller nucleolar velocity, respectively, with a linear fit vmax 1.74vmin. (J) Histogram of vrad for fusing nucleoli. (K) vrad as a function of time for fusing nucleoli. (L) αv for fusing nucleoli as a function of time.

Next, we analyze the motion of pairs of fusing nucleoli in the same fashion. Examples of such trajectories, with their temporal evolution color-coded (blue to red), are shown in Figure 5G. An enlarged view of these trajectories and the areas they cover is depicted in Figure 5—figure supplement 1C–D. Figure 5H shows the distribution of the nucleolar velocities, v, for fusing nucleoli, with an average velocity v≈ (0.33 ± 0.26) × 10-3 μms-1, which is ∼ 30% smaller than for non-fusing nucleoli. This difference is statistically significant as corroborated by p-value of 4 × 10-4. Strikingly, when we review velocities of a fusing nucleolar pair and plot the larger velocity, vmax, against the smaller velocity, vmin (Figure 5I), we find a clear linear correlation between them, with the linear fit of vmax = (1.74 ± 0.20)vmin (Figure 5I, blue line) and a Pearson correlation coefficient ρ= 0.88. Moreover, we find that vmax/vmin ≈ 1.8 ± 0.6. The distribution of vrad for fusing nucleoli (Figure 5J) is still centered around 0, but is clearly narrower than for non-fusing nucleoli (Figure 5D) with a variance σrad,fusing = 1 × 10-4 μms-1 obtained by fitting a Gaussian curve to the vrad distribution in Figure 5J. σrad,fusing is about three times smaller than σrad,nonfusing. When reviewed over time, vrad exhibits much smaller fluctuations (Figure 5K) than in case of non-fusing nucleoli (Figure 5E). Lastly, Figure 5L shows αv of the pre-fusion nucleoli as a function of time, monitoring 40 min prior to fusion, which occurs at t = 0 min. Interestingly, the nucleoli seem not to follow any preferred direction, but instead follow a zig-zag motion while approaching each other to fuse.

Nucleolar response to ATP depletion

To investigate a possible role of active (ATP-dependent) processes in maintaining the nucleolus-nucleoplasm interface, we have examined nucleoli, specifically, their shape, surface roughness and possible fusion events, upon ATP depletion. The ATP was depleted using 2-deoxyglucose and trifluoromethoxy-carbonylcyanide phenylhydrazone (see Materials and methods). Figure 6A and B show micrographs of nuclei with fluorescently labeled chromatin (H2B-GFP, green) and nucleoli (NPM-DsRed, red) under physiological conditions (control) and upon ATP depletion, respectively. In addition, we review the z-projection of nuclear and nucleolar contours obtained in different planes of the respective z-stack (Figure 6A–B). We find that upon ATP depletion nucleoli do not exhibit spherical, but instead irregular shapes (Figure 6B). In fact, some of the larger irregularly shaped nucleoli might originate from nucleoli fusing at the time of ATP depletion, given the absence of nucleoli in the hour-glass shape, characteristic of nucleolar fusion under physiological conditions (Figure 1).

Nucleoli under physiological conditions and upon ATP-depletion.

(A–B) Micrographs of HeLa nuclei with fluorescently labeled chromatin (H2B-GFP, green) and nucleoli (NPM-DsRed, red), their color overlay and z-projections of nucleolar and nuclear contours: (A) under physiological conditions (control) and (B) after ATP depletion. (C) Distributions of the following nucleolar measurements under physiological conditions (NN= 648, NCell= 208) vs. upon ATP depletion (NN= 345, NCell= 127): nucleolar area normalized by nuclear area, AN/ANuc, nucleolar eccentricity, e, angle between the major nuclear and nucleolar axes, α, the shortest distance from the nucleolar centroid to the nuclear envelope normalized by the nuclear circumference, de, fraction of the nucleolar contour with negative curvature, fneg, and number of continuous regions of negative curvature along the nucleolar contour, Nneg. All measurements are carried out in the nucleolar focal plane. Scale bar, 5 µm.

To characterize the morphological changes of nucleoli upon ATP depletion, we define parameters describing their shape and compare against the control nucleoli. Specifically, after we obtain nuclear and nucleolar contours in their respective focal planes, we compute the following six parameters for every nucleolus: AN/ANuc (the nucleolar area AN normalized by the nuclear area ANuc), e (the eccentricity e=a/b, where a and b are the semi-major and semi-minor nucleolar axis, respectively), α (the angle between the nuclear and nucleolar major axes), de (the shortest distance from the nucleolar centroid to the nuclear envelope normalized by the nuclear circumference in the focal plane of the nucleus), fneg (the fraction of the nucleolar contour with negative curvature) and Nneg (the number of independent nucleolar contour regions with negative curvature), where the curvature corresponds to the in plane curvature of the nucleolar contour.

The comparison of these parameters for nucleoli under physiological conditions and upon ATP depletion is shown in Figure 6C. Table 1 provides a summary of means and standard deviations for measured distributions of AN/ANuc, e, α, de, fneg and Nneg. In addition, we evaluated the p-values for all measured physical quantities as well as the relative differences of their means with respect to control (Table 1). The relative difference (in %) of the means was calculated as 100 × [(μQμP)/μP], where μP is the mean of the probability distribution of the measured physical quantity under control conditions and μQ after the perturbation. Furthermore, we computed the skew of the measured distributions, which informs about their asymmetry, and the Kullback-Leibler divergence with respect to control (Table 1). The Kullback-Leibler divergence is a measure of the difference between two probability distributions. It is defined as P(i)ln(P(i)Q(i)), where P(i) and Q(i) are the two distributions. Here, P(i) corresponds to the probability distribution of the measured physical quantity under control conditions, Q(i) after the perturbation.

The most striking change that we observe upon the ATP depletion is the irregularity of the nucleolar shape. The dramatic increase in the nucleolar surface roughness upon ATP depletion is nicely captured by the growing amount of the nucleolar contour possessing negative curvature as quantified by fneg and Nneg, both showing increase of ∼ 20%. Remarkably, when compared to the control nucleoli, the ATP-depleted nucleoli not only show larger parts of their contour to posses negative curvature, but are also more likely to contain several more independent contour regions of negative curvature making them appear lobulated. In addition, the ATP-depleted nucleoli tend to localize further away from the nuclear envelope, as illustrated by de, than the control nucleoli. Surprisingly, there are no significant changes to the average nucleolar area, eccentricity and orientation as per AN/ANuc, e and α, respectively, upon ATP depletion.

Nucleolar response to biochemical perturbations

To probe contributions of specific cellular processes (such as cytoskeletal forces, transcriptional activity as well as protein synthesis) to maintaining the nucleolus-nucleoplasm interface, we employ targeted biochemical perturbations. Specifically, to inhibit cytoskeletal forces we treat the cells with blebbistatin, which is a myosin II inhibitor, latrunculin A, which prevents actin polymerization, and nocodazole, which is a microtubule polymerization blocker. To test the contributions of transcription-related processes we apply α-amanitin, which inhibits the RNA polymerase II activity, and flavopiridol, which blocks the positive transcription elongation factor P-TEFb. In addition, we probed the impact of the local chromatin packing state by applying trichostatin A, which prevents histone deacetylation, and thus leads to chromatin decondensation. Finally, since the nucleolus is the site of ribosome biogenesis and thus directly involved in cellular protein production, we explore the role of protein synthesis in maintaining the nucleolus-nucleoplasm interface. To do so, we use cycloheximide, a protein synthesis inhibitor, and evaluate its effect at two time points, t1= 30 min and t2= 6.5 hr, upon drug addition. We anticipate that at short timescales, we can observe a direct impact of protein synthesis inhibition on the nucleolar-nucleoplasm interface, while at longer timescales, we can investigate a possible feedback between protein synthesis inhibition and nucleolar size and shape.

Figure 7A shows micrographs of nuclei with fluorescently labeled chromatin (H2B-GFP, green) and nucleoli (NPM-DsRed, red) under the physiological conditions (control) and after treatment with blebbistatin, latrunculin A, nocodazole, α-amanitin, flavopiridol, trichostatin A and cycloheximide at t1 and t2. We also survey the z-projection of nuclear and nucleolar contours obtained in different planes of the respective z-stack (Figure 7—figure supplement 1). To examine morphological differences under the studied conditions, we evaluate the same six parameters used earlier: AN/ANuc, e, α, de, fneg, and Nneg and visualize their distributions as violin plots in Figure 7B. The red dot indicates the mean, while the solid and dashed red lines correspond to the median and quartiles, respectively. Table 1 provides a summary of means and standard deviations for measured distributions of AN/ANuc, e, α, de, fneg and Nneg. In addition, we evaluated the skew of measured distributions and computed p-values for all measured physical quantities, the relative differences of their means with respect to control, as well as the Kullback-Leibler divergence with respect to control (Table 1).

Figure 7 with 1 supplement see all
Nucleoli upon biochemical perturbations.

(A) Micrographs of HeLa nuclei with fluorescently labeled chromatin (H2B-GFP, green) and nucleoli (NPM-DsRed, red), and the overlay, under the following conditions: control, upon addition of blebbistatin, latrunculin A, nocodazole, α-amanitin, flavopiridol, trichostatin A, and cycloheximide (at t1= 30 min and t2= 6.5 hr). (B) Histograms of the following measurements for all conditions (width indicates probability): nucleolar area normalized by nuclear area, AN/ANuc, nucleolar eccentricity, e, angle between the major nuclear and nucleolar axes, α, the shortest distance from the nucleolar centroid to the nuclear envelope normalized by the nuclear circumference, de, fraction of the nucleolar contour with negative curvature, fneg, and number of continuous regions of negative curvature along the nucleolar contour, Nneg. All data collected in the nucleolar focal plane. Red dot, solid red line and dotted red lines indicate the mean, median and quartiles, respectively. Table 1 provides p-values for all measured data with respect to the control. The number of nucleoli and cells are as follows: control (NN= 648, NCell= 208), blebbistatin (NN= 399, NCell= 127), latrunculin A (NN= 307, NCell= 104), nocodazole (NN= 310, NCell= 106), α-amanitin (NN= 268, NCell= 95), flavopiridol (NN= 309, NCell= 105), trichostatin A (NN= 278, NCell= 95), and cycloheximide at t1= 30 min (NN= 291, NCell= 91), and cycloheximide at t2= 6.5 hr (NN= 294, NCell= 105). Scale bar, 5 µm.

Table 1
Statistical characteristics of distributions for physical quantities evaluated for nucleoli upon biochemical perturbations (see Figures 67).
Mean ± standard deviation
NNucleoliNCellsAN/ANuc e α de fneg Nneg
Control6482080.052 ± 0.0401.30 ± 0.3247 ± 260.076 ± 0.0290.033 ± 0.0660.66 ± 1.37
ATP-depletion3451270.054 ± 0.0421.29 ± 0.2848 ± 260.082 ± 0.0280.038 ± 0.0640.82 ± 1.39
Blebbistatin3991270.052 ± 0.0411.29 ± 0.2846 ± 260.076 ± 0.0280.036 ± 0.0670.72 ± 1.38
Latrunculin A3071040.061 ± 0.0441.31 ± 0.3647 ± 250.071 ± 0.0280.030 ± 0.0630.56 ± 1.21
Nocodazole3101060.052 ± 0.0371.28 ± 0.2550 ± 280.074 ± 0.0290.032 ± 0.0610.67 ± 1.38
α-amanitin268950.060 ± 0.0471.34 ± 0.2944 ± 270.074 ± 0.0290.046 ± 0.0720.99 ± 1.71
Flavopiridol3091050.052 ± 0.0401.34 ± 0.3443 ± 270.074 ± 0.0280.048 ± 0.0800.92 ± 1.67
Trichostatin A278950.044 ± 0.0321.29 ± 0.3643 ± 280.073 ± 0.0260.025 ± 0.0630.53 ± 1.35
Cyclohex I291910.053 ± 0.0371.32 ± 0.2946 ± 260.076 ± 0.0250.038 ± 0.0690.76 ± 1.49
Cyclohex II2941050.052 ± 0.0391.35 ± 0.3847 ± 250.074 ± 0.0290.040 ± 0.0670.74 ± 1.33
p-values (with respect to control)Relative difference of mean [%]
AN/ANuc e α de fneg NnegAN/ANuc e α de fnegNneg
ATP-depletion0.3590.6740.5130.0090.2580.0934%−1%2%8%15%24%
Blebbistatin0.8930.5150.6940.9110.5820.4950%−1%−2%0%9%9%
Latrunculin A0.0020.7140.9400.0110.4640.24617%1%0%−7%−9%−15%
Nocodazole0.9850.3760.1690.3350.7340.9250%−2%6%−3%−3%2%
α-amanitin0.0110.0890.0900.2800.0170.00615%3%−6%−3%39%50%
Flavopiridol0.7890.0890.0590.2720.0060 .0180%3%−9%−3%45%39%
Trichostatin A0.0020.9060.0510.0990.0870.160−15%−1%−9%−4%−24%−20%
Cyclohex I0.6540.2570.6330.9320.3550.3592%2%−2%0%15%15%
Cyclohex II0.9840.0350.9820.2840.1630.3810%4%0%−3%21%12%
Kullback-Leibler divergence (with respect to control)Skew
AN/ANuc e α de fneg NnegAN/ANuc e α de fnegNneg
Control1.103.07−0.09−0.042.172.68
ATP-depletion0.0260.0110.0210.0320.0510.0181.012.23−0.19−0.161.801.87
Blebbistatin0.0230.0250.0310.0360.0220.0071.312.26−0.110.032.002.14
Latrunculin A0.0590.0290.0280.0400.0500.0150.793.01−0.15−0.072.062.36
Nocodazole0.0220.0140.0420.0250.0330.0100.871.84−0.280.071.902.55
α-amanitin0.0600.0430.0570.0320.0620.0360.981.950.080.081.612.26
Flavopiridol0.0380.0270.0440.0350.0530.0211.282.360.03−0.021.622.09
Trichostatin A0.0560.0460.0620.0590.0470.0340.853.150.17−0.252.953.15
Cyclohex I0.0410.0320.0430.0510.0430.0151.132.13−0.08−0.121.982.70
Cyclohex II0.0130.0260.0490.0210.0400.0140.984.19−0.080.011.662.14

A close inspection of the violin plots (Figure 7B) and their corresponding statistical characteristics (Table 1) reveals the following morphological changes upon cytoskeletal, chromatin and protein synthesis perturbations.

Interestingly, the cytoskeletal perturbations, which act on the cytoskeleton outside the cell nucleus, did not lead to any major changes in the nucleolar morphology except for the actin polymerization inhibitor latrunculin A, which led to an increase of AN/ANuc compared to the control nucleoli. This increase, however, is due to the rounding up of nuclei upon the latrunculin A treatment (Khatau et al., 2009; Burnette et al., 2014), which leads to a decrease in the nuclear area ANuc, and thus causes the apparent increase of AN/ANuc, while the measured nucleolar area AN remains comparable to the AN of control nucleoli. Similarly, the observed decrease in the distance of nucleoli from the nuclear envelope, de, is likely caused by a decrease of the observed nuclear area ANuc.

In contrast, chromatin perturbations such as transcription inhibitors α-amanitin, flavopiridol and histone deacetylase inhibitor trichostatin A led to visible changes in the nucleolar morphology as well as in the roughness of the nucleolus-nucleoplasm interface. Specifically, upon perturbing polymerase II activity using α-amanitin, we find that the nucleolar size AN/ANuc increases by ∼ 15% and nucleoli are on average more elliptical (e). Moreover, the roughness of the nucleolus surface strongly increases by ∼ 40–50% as measured by the amount of negative curvature (fneg and Nneg). Similarly, when we perturb the transcription elongation using flavopiridol, we observe ∼ 40–45% increase in the nucleolus surface roughness (fneg and Nneg) and nucleoli become on average more elliptical (e). Finally, when we block the histone deacetylation using trichostatin A, which leads to chromatin decondensation, we find that the nucleolar size AN/ANuc decreases by ∼ 15%, while their eccentricity (e) remains unchanged. Strikingly, the nucleolar surface roughness decreases by ∼ 20–25% (fneg and Nneg), in other words upon trichostatin A treatment it becomes smoother than in the control.

Finally, the protein synthesis inhibition using cycloheximide left the nucleolar size unchanged, while the nucleoli became on average more elliptical (e) at longer times (t2= 6.5 hr). Furthermore, protein synthesis inhibition led to a moderate ∼ 15% increase in the nucleolar surface roughness (fneg and Nneg) at both times (t1= 30 min and t2= 6.5 hr).

The orientation of the nucleoli within the nucleus and the nucleolar distance from the nuclear envelope is not significantly affected by any of the studied perturbations as illustrated by α and de, respectively.

Discussion

In this work, we study the nucleolus as the archetype of cellular organelles formed by liquid-liquid phase separation (LLPS) and monitor its size, shape and dynamics during its lifetime in human cells in vivo. We discover a rich phenomenology that grows the LLPS framework in new and unexpected ways: (i) We find that nucleoli exhibit anomalous dynamics and anomalous volume distribution during the cell cycle that defies any current theory and necessitates a new one. (ii) We uncover that the nucleolar fluid is a colloidal solution containing solid-like granules, the DFCs. (iii) We reveal that the surrounding nucleoplasm plays a key role in the LLPS of nucleoli that might have been previously overlooked and find that active (ATP-dependent) processes are involved in maintaining the nucleolus-nucleoplasm interface. Moreover, we identify specific biological processes participating in the nucleolus-nucleoplasm interactions.

Our findings show that the nucleolar volume distribution scales as P(V)V1 during the entire cell cycle. The scale-free nature of this distribution suggests that nucleoli of any size can coalesce, moreover, there is no preferred size that nucleoli need to reach before/after they coalesce. It also suggests, that nucleoli of different sizes follow the same coalescence kinetics (Caragine et al., 2018). Furthermore, the nucleolar volume distribution remains unchanged during the cell cycle, suggesting that the fusion of nucleoli is not limited to the first two hours of the cell cycle as previously hypothesized (Savino et al., 2001), but can occur at any time. Nucleolar coalescence occurs from the early stages in the cell cycle, where it is thought to be a part of the nucleolar assembly process (Savino et al., 2001). It is conceivable that at later times, the nucleolar coalescence might serve a different purpose as it is less likely to happen with decreasing nucleolar number.

Interestingly, a volume distribution P(V)V-1 was previously found also for liquid-like P-granules in the C. elegans oocyte (Hubstenberger et al., 2013). In contrast, the volume distribution of nucleoli in the X. laevis oocyte follows V1.5, which was shown to be consistent with diffusion-limited aggregation with constant influx of particles (Brangwynne et al., 2011). The kinetics of human nucleolar assembly likely differs from that of the frog oocyte due to numerous differences between these two systems. For example, the nucleolar count is much lower in human cells (∼ 100 times less than in frog oocyte), there is a dense actin network present in the frog oocyte nucleus (germinal vesicle), and human somatic nucleoli are connected to the chromatin fiber, thus, they cannot freely diffuse as it is in the case of nucleoli in the X. laevis oocyte (Gall et al., 2004; Brangwynne et al., 2011; Caragine et al., 2018; Berry et al., 2018). Further differences between the two systems include a large difference in the nuclear size (diameter ~ 1000 μm in frog oocytes, ∼ 10 µm in human cells) and the nucleolar size with volumes of 10–10µm3 in frog oocytes and 10-2–10µm3 in human cells (Gall et al., 2004; Brangwynne et al., 2011; Caragine et al., 2018).

The anomalous volume distribution of human nucleoli might likely be connected to their anomalous dynamics. Remarkably, our data suggest that one can predict if a pair of nucleoli is going to fuse by analyzing their motion. The differences in the dynamical behavior of non-fusing nucleoli and the ones in approach for fusion are stark. While the non-fusing ones appear to move randomly through the nucleoplasm, nucleoli that will fuse in the near future, move slower than non-fusing ones and show a linear correlation in their velocities (Figure 5I). Considering the nucleolar size and the fact that they are physically tethered to the chromatin fiber, their motion unavoidably leads to local spatial reorganization of chromatin. Alternatively, a local chromatin rearrangement could facilitate the nucleolar pre-fusion approach. In fact, in our earlier work we found that the velocities of the growth of the neck connecting two fusing nucleoli (Figure 1B) are intriguingly similar to the velocities measured for active chromatin motion (Caragine et al., 2018; Zidovska et al., 2013). Since nucleoli move in an active fluid (chromatin solution), we speculate that active processes might be involved in bringing them together to undergo fusion. To explore this hypothesis, future experiments and theories are needed to probe the nucleolar interactions with chromatin.

The complex nature of the nucleolar fluid might also contribute to the anomalous behavior of human nuceoli. Strikingly, we find that dense fibrillar components (DFCs) behave as monodisperse solid-like colloidal particles (granules) suspended in a liquid phase of granular component (GC). Our data shows that DFCs do not undergo aggregation, but remain of well-defined size and dimensions with a semi-major axis length of 210 ± 50 nm and semi-minor axis length of 180 ± 40 nm, as well as shape with aspect ratio of 1.22 ± 0.17 even upon nucleolar coalescence, which is consistent with solid-like particles. In contrast, the DFCs in frog oocytes were found to be polydisperse with diameter ∼ 2–5 µm, liquid-like with viscoelastic behavior (Brangwynne et al., 2011; Feric et al., 2016), and with their fusion being observed upon latrunculin A treatment (Feric et al., 2016). It is also noteworthy that one frog oocyte DFC can be larger than the entire human nucleolus. Furthermore, our data reveal that human nucleoli obey a volumetric ratio for GC and DFC content, with DFC volume fraction ∼ 0.1, which is significantly lower than in frog oocytes (∼ 0.25) (Feric et al., 2016). This suggests it is the rRNA-rich GC phase that provides the human nucleolus with its liquid-like properties.

To investigate the nucleolar interactions with the surrounding nucleoplasm, we have tested the impact of active (ATP-dependent) processes in general as well as specific biological processes such as cytoskeletal and transcriptional activity, chromatin packing state and protein synthesis. Our data suggest that nucleoli are closely dependent on ATP-dependent processes, losing their spherical shape upon ATP-depletion by exhibiting increased surface roughness (local deformations). In our earlier study (Caragine et al., 2018) we have shown that the surface roughness can serve as a readout of the nucleolar surface tension. Specifically, local nucleolar surface deformations, which may be driven thermally or by active processes, are opposed by the surface tension. Thus, the larger the surface roughness, the lower its surface tension. Hence, a possible interpretation of the increase in nucleolar surface roughness upon ATP-depletion is a reduction of the surface tension γ of the nucleolus-nucleoplasm interface. These findings are consistent with our earlier study, which found that γ under physiological conditions is an effective quantity, and is therefore, likely dependent on some of the ATP-dependent cellular processes (Caragine et al., 2018).

Our data show that the roughness (local deformations) of the nucleolus-nucleoplasm interface is highly sensitive to the transcriptional activity in the nucleus. Interestingly, the inhibition of transcriptional activity (such as polymerase II activity or mRNA elongation) in the nucleus leads to an increase of the relative nucleolar size and the nucleolus becomes more elongated (less spherical) with a number of local deformations leading to high surface roughness. However, upon blocking the histone deacetylases, which causes a visible chromatin decondensation (Tóth et al., 2004), we find not only a reduction in the relative nucleolar size, but also an increasingly smooth nucleolus-nucleoplasm interface. This suggests that the nucleolus-nucleoplasm interface is closely linked to the chromatin packing state as well as its transcriptional activity. Conversely, the perinucleolar chromatin is mostly heterochromatic, that is largely transcriptionally inactive, yet, its peculiar packing at the nucleolar surface might require active remodeling. Moreover, this is in agreement with our hypothesis that the surface tension γ is maintained by active processes and thus is an effective physical quantity. To elucidate the underlying physics, new theories accounting for the non-equilibrium nature of the nucleolus-nucleoplasm liquid interface need to be developed.

In contrast, we find that the cytoskeletal forces exerted on the nucleus from the cytoplasm, do not contribute to the local roughness of the nucleolus-nucleoplasm interface, nor do they impact the nucleolar size and shape. Interestingly, in frog oocytes the disruption of the dense nuclear actin network by latrunculin A facilitates nucleolar fusion, leading to an increase in the nucleolar size (Feric et al., 2016). Conversely, there is no filamentous actin network present in human cell nucleus.

Lastly, our findings reveal that nucleoli are only moderately sensitive to the protein synthesis inhibition at time scales from 30 min to 6.5 hr. We do not observe any change in their size, only a slight increase in the surface roughness. However, it is possible that to observe an effect on nucleoli from the lack of protein synthesis much larger times need to be explored.

In summary, we speculate that the interplay of the complex nature of the nucleolar fluid, the reduced mobility of nucleoli due to their chromatin tethering, as well as their interactions with the surrounding nucleoplasm, might impact the nucleolar assembly kinetics and lead to the observed anomalous nucleolar volume distribution (V-1).

In conclusion, nucleoplasm plays a major role in the life of nucleoli, the archetype of the liquid condensate formed by liquid-liquid phase separation in biology. Nucleoplasm, the fluid surrounding the nucleoli, is a complex polymeric solution containing chromatin. Chromatin fiber serves as the template for nucleolar formation and later forms a boundary at the nucleolus-nucleoplasm interface (McClintock, 1934; Ritossa and Spiegelman, 1965; Wallace and Birnstiel, 1966; Bickmore and van Steensel, 2013; Németh and Längst, 2011; Towbin et al., 2013). Strikingly, the DNA sequences at which nucleoli form (NORs) and the genes located at the nucleolar interface are by no means random (Németh and Längst, 2011). This likely impacts the 3D chromosomal organization in the nucleolar vicinity. Moreover, considering chromatin’s active nature (Zidovska et al., 2013) and the fact that nucleoli are tethered to it during their lifetime, we speculate that active positional fluctuations (or rearrangements) of chromatin could bring nucleoli together, facilitating fusion. While this hypothesis remains to be tested, it is consistent with our observations that nucleoli, which are in approach to fusion, exhibit different dynamics than non-fusing ones. It is also supported by previous studies of colloidal mixtures, where the presence of particles with an actively driven translational motion leads to phase separation of active and passive (i.e., thermally driven) components (Stenhammar et al., 2015). Similar behavior has been found for polymer mixtures containing active and passive polymers (Smrek and Kremer, 2017). We speculate that, with respect to its translational mobility, the nucleolus could be abstracted as a passive droplet (or colloid) immersed in an active polymer (chromatin). In such case, the active positional fluctuations of the polymer could cause demixing of the passive phase and thus effectively bring the passive colloids (nucleoli) together. That is, the active entities phase separate from the passive entities, enabling nucleolar coalescence. Such phase separation is distinct from the liquid-liquid phase separation by which the nucleoli are thought to form at the beginning of the cell cycle (Brangwynne et al., 2011; Berry et al., 2015; Feric et al., 2016).

The nucleolus plays a key role in cellular protein synthesis, thus any changes in nucleolar composition, structure or function can lead to cell abnormalities often connected with human diseases. For example, mutations in nucleolar proteins, which interact with RNA polymerase I, regulate rRNA transcription or participate in rRNA processing, are associated with cell cycle arrest and improper nucleolar assembly. This can lead to diseases such as skeletal and neurodegenerative disorders, cardiovascular disease and cancer (Hannan et al., 2013; Núñez Villacís et al., 2018; Ruggero and Pandolfi, 2003; Derenzini et al., 2009). Moreover, in many diseases such as cancer, Alzheimer’s and Parkinson’s disease, but also in aging, human nucleoli change their shape and size (Hannan et al., 2013; Tsai and Pederson, 2014; Núñez Villacís et al., 2018; Tiku and Antebi, 2018), making the nucleolus a potential valuable diagnostic marker. Hence, a mechanistic understanding of nucleolus, its material properties and physical interactions with the nucleoplasm, might illuminate nucleolus in health and disease, contributing to new paths for diagnosis and therapy.

Materials and methods

Key resources table
Reagent type (species) or resourceDesignationSource or referenceIdentifiersAdditional information
Cell line
(H. sapiens)
HeLaATCCATCC:CCL2
RRID:CVCL_0030
Stable H2B-GFP cell line
Transfected construct
(H. sapiens)
NPM-DsRedAddgene34553
RRID:Addgene_34553
Transfected construct
(H. sapiens)
mCerulean-Fibrillarin-7Addgene55368
RRID:Addgene_55368
Transfected construct
(H. sapiens)
NPM-mAppleCaragine et al., 2018n/aGenerated by Zidovska Lab – published in Caragine et al., 2018
Chemical compound, drugRO-3306Enzo Life SciencesALX-270–463
Chemical compound, drug2-Deoxy-D-Glucose (DOG)Millipore SigmaD8375
Chemical compound, drugTrifluoromethoxy-carbonylcyanide phenylhydrazone (FCCP)Millipore SigmaC2920
Chemical compound, drugLatrunculin AMillipore SigmaL5163
Chemical compound, drugBlebbistatinMillipore SigmaB0560
Chemical compound, drugNocodazoleMillipore SigmaM1404
Chemical compound, drugα-AmanitinSanta Cruz Biotechnologysc-202440
Chemical compound, drugCycloheximideSanta CruzBiotechnologysc-3508
Chemical compound, drugFlavopiridolSanta Cruz Biotechnologysc-202157
Chemical compound, drugTrichostatin AMillipore SigmaT8552
Software,
algorithm
MatlabMathWorks2017a, 2019a
Software, algorithmAdobe Illustrator, PhotoshopAdobe Inc.CC2018

Cell culture and cell transfection

Request a detailed protocol

The stable HeLa H2B-GFP cell line was cultured according to ATCC recommendations (CCL-2). Cells were cultured in a humidified, 5% CO2 (vol/vol) atmosphere in Gibco Dulbecco’s modified eagle medium (DMEM) supplemented with 10% FBS (vol/vol), 100 units/mL penicillin, 100 μg/mL streptomycin (Invitrogen) and 4.5 µg/mL Plasmocin Prophylactic (Invivogen). Cells were mycoplasma free, as determined by the Invivogen PlasmoTest (Invivogen). For H2B-GFP imaging, cells were plated onto 35 mm MatTek dishes with glass bottom no. 1.5 (MatTek) 24 hr before imaging. We performed four independent experiments. For concurrent H2B-GFP and NPM-DsRed (or NPM-mApple) imaging, cells were plated onto 35 mm MatTek dishes 48 hr before imaging and transiently transfected with NPM-DsRed (or NPM-mApple) 24 hr prior to the experiment. All transfections were carried out using Lipofectamine 2000 (Invitrogen) following the manufacturer’s protocol. When indicated, cells were synchronized using 10 µM RO-3306 (Enzo Life Sciences) and imaged before the drug was removed after 16 hr, as well as 3.5 and 5 hr after the drug removal. The synchronized and unsynchronized populations were evaluated in two distinct experiments. For concurrent imaging of H2B GFP, NPM-DsRed and mCerulean-Fibrillarin-7, cells were plated onto 35 mm MatTek dishes 48 hr before the experiment and transiently transfected with both NPM-DsRed and m-Cerulean-Fibrillarin-7 24 hr prior to the experiment. We performed six independent experiments, all of which were analyzed qualitatively and one quantitatively. NPM-DsRed and FBL-mCerulean (mCerulean3-Fibrillarin-7) were gifts from Mary Dasso (Addgene plasmid # 34553) (Yun et al., 2008) and from Michael Davidson (Addgene plasmid # 55368) (Markwardt et al., 2011), respectively. NPM-mApple plasmid was created as described earlier (Caragine et al., 2018). For experiments involving biochemical perturbations, cells were plated onto 35 mm MatTek dishes 72 hr in advance of the experiment, transiently transfected with NPM-DsRed 48 hr prior to the experiment, and replated onto 35 mm MatTek dishes 24 hr prior to the experiment. We performed three independent experiments for each perturbation and six for the control. All imaging experiments were performed in the Gibco CO2-independent media (Invitrogen) supplemented with L-Glutamine (Invitrogen) and with MatTek dish containing cells mounted on the microscope stage in a custom-built environmental chamber maintained at 37°C with 5% CO2 supplied throughout the experiment.

Biochemical perturbations

Request a detailed protocol

To deplete ATP, cells were treated with 6 mM 2-deoxyglucose (DOG) and 1 µM trifluoromethoxy-carbonylcyanide phenylhydrazone (FCCP) dissolved in CO2-independent medium supplemented with L-glutamine 2 hr before imaging. For cytoskeletal perturbations 10 µM latrunculin A, 10 µM blebbistatin or 10 µM nocodazole, respectively, in CO2-independent medium supplemented with L-glutamine were added to cells 30 min before imaging. For chromatin perturbations, 20 µg/mL α-amanitin (Santa Cruz Biotechnology), 5 µg/mL cycloheximide (Santa Cruz Biotechnology), 83 nM flavopiridol (Santa Cruz Biotechnology), or 624 nM trichostatin A (TSA), respectively, in CO2-independent medium supplemented with L-glutamine were added to cells 30 min, 30 min, 2 hr, and 24 hr before imaging, respectively. For cycloheximide, additional timepoint, 6.5 hr after drug additon, was evaluated. All chemicals were from Sigma Aldrich unless stated otherwise.

Microscopy and image acquisition

Request a detailed protocol

Cells were imaged with a Yokogawa CSU-X1 confocal head with an internal motorized high-speed emission filter wheel, Spectral Applied Research Borealis modification for increased light throughput and illumination homogeneity on a Nikon Ti-E inverted microscope equipped with an oil-immersion 100× Plan Apo NA 1.4 objective lens, an oil-immersion 40× Plan Fluor NA 1.3 objective lens, and the Perfect Focus system. The microscope was mounted on a vibration-isolation air table. The pixel size for the 100× and 40× objective was 0.065 µm and 0.1625 µm, respectively. H2B-GFP and NPM-DsRed (or NPM-mApple) fluorescence was excited with a 488 nm and a 561 nm solid-state laser, respectively. To image H2B-GFP and NPM-DsRed (or NPM-mApple) at the same time, we illuminated the sample simultaneously with both excitation wavelengths, 488 nm and 561 nm. The emission was separated by the W-View Gemini Image Splitter (Hamamatsu) using a dichroic mirror (Chroma Technology), followed by an ET525/30m and an ET630/75m emission filter (Chroma Technology). The two fluorescent signals were allocated to the two halves of the image sensor, producing two distinct images. The exposure time for each frame was 250 ms. For three color imaging, H2B-GFP, NPM-DsRed, and FBL-mCerulean were excited with 488 nm, 561 nm, and 405 nm solid state lasers, respectively, and fluorescence was collected with a 405/488/561/640 multiband-pass dichroic mirror (Semrock) and then an ET525/50m, ET600/50m and ET450/50m emission filter, respectively (Chroma Technology). The exposure time was 250 ms, 250 ms, and 1000 ms for H2B-GFP, NPM-DsRed, and FBL-mCerulean, respectively. Z-stacks were taken with a z axis step size of 500 nm, with the shutter closed in-between steps and an exposure time of 250 ms per plane. Images were obtained with a Hamamatsu ORCA-R2 cooled CCD camera controlled with MetaMorph 7 (Molecular Devices). The streams of 16-bit images were saved as multi-tiff stacks.

Image processing and data analysis

Request a detailed protocol

Images were converted to single-tiff images and analyzed with MatLab (The MathWorks). The nuclear and nucleolar contours were determined from the H2B-GFP and NPM-DsRed signal, respectively, using previously published procedures (Chu et al., 2017; Caragine et al., 2018).

The nucleolar velocity was determined as the displacement of the centroid of the filled nucleolar contour relative to the displacement of the centroid of the filled nuclear contour, divided by the elapsed time. For radial velocity calculations, we define the radial distance of each nucleolus as its distance from the midpoint of the linear distance between the centroids of two nucleoli. The radial distances for both nucleoli are measured relative to the midpoint of the linear distance between the nucleoli found in two time points, in order to exclude the movement of the other nucleolus in the calculation of nucleolar velocity. Finally, to calculate the radial velocity we divide the change in the radial distance by the time elapsed.

The nucleolar distance from the nuclear envelope was found by finding the minimum distance between the nucleolar centroid and the nuclear contour. The angle between the nucleolus and nucleus was determined by fitting both nucleus and nucleolus with an ellipse and measuring the angle between their respective major axes, for angles greater than 90°, its supplement was taken. The nuclear and nucleolar area were determined as the number of pixels filling its respective contours. Nuclear and nucleolar eccentricity was calculated as the ratio of the semi-major axis length and its semi-minor axis length, when fitted to an ellipse.

To obtain an accurate count of DFCs, we developed a feature-finding procedure. A mask created by the nucleolar contour was applied to the FBL-mCerulean image to remove the background signal. Using a local-maxima function, we found a large number of local maxima indicating possible features in the image, most of which correspond to noise. We manually selected DFCs from the local maxima that were found. Next, we manually fit each DFC with a circumscribed and an inscribed circle, measuring the semi-major DFC axis a and the semi-minor DFC axis b, respectively.

References

  1. 1
  2. 2
  3. 3
  4. 4
  5. 5
  6. 6
  7. 7
  8. 8
  9. 9
  10. 10
  11. 11
  12. 12
  13. 13
  14. 14
  15. 15
  16. 16
  17. 17
  18. 18
  19. 19
  20. 20
  21. 21
  22. 22
  23. 23
  24. 24
  25. 25
  26. 26
  27. 27
  28. 28
  29. 29
  30. 30
  31. 31
  32. 32
  33. 33
  34. 34
  35. 35
  36. 36
    Ribosomal cistrons and the nucleolar organizer
    1. H Wallace
    2. ML Birnstiel
    (1966)
    Biochimica Et Biophysica Acta (BBA) - Nucleic Acids and Protein Synthesis 114:296–310.
    https://doi.org/10.1016/0005-2787(66)90311-X
  37. 37
  38. 38
  39. 39

Decision letter

  1. Detlef Weigel
    Senior Editor; Max Planck Institute for Developmental Biology, Germany
  2. Arup K Chakraborty
    Reviewing Editor; Massachusetts Institute of Technology, United States
  3. Rohit V Pappu
    Reviewer; Washington University in St Louis, United States
  4. Pilong Li
    Reviewer

In the interests of transparency, eLife publishes the most substantive revision requests and the accompanying author responses.

Thank you for submitting your article "Nucleolar dynamics and interactions with nucleoplasm in living cells" for consideration by eLife. Your article has been reviewed by three peer reviewers, and the evaluation has been overseen by Arup Chakraborty as the Senior and Reviewing Editor. The following individuals involved in review of your submission have agreed to reveal their identity: Rohit V Pappu (Reviewer #1); Pilong Li (Reviewer #3).

The reviewers have discussed the reviews with one another and the Reviewing Editor has drafted this decision to help you prepare a revised submission.

While this paper builds on your past publication in Physical Review Letters in 2018, there are considerable new insights in this manuscript pertaining to the interplay between spontaneous vs. driven processes in regulating nucleolar shapes, sizes, and cell cycle dependent dynamics. However, it is important to highlight the new findings more clearly, and very importantly, to link the data reported in this manuscript to the conclusions drawn. At present, the conclusions are very general and speculative, rather than closely tied to the actual observations. Please find the three reviews appended below as it will help you address additional significant points.

Reviewer #1:

This work is an important and essential follow up to previous studies, one from this very group, on the dynamics of nucleoli. There is growing interest in understanding the interplay between spontaneous and driven (or what some folk refer to as passive vs. active) processes. Yes, facsimiles of nucleoli can be reconstituted in vitro, but the interplay between spontaneous vs. driven processes in regulating nucleolar shapes, sizes, and cell cycle dependent dynamics remains unclear. Here, the authors provide interesting insights based on quantitative imaging that have surprising ramifications. Overall, this is a really important contribution and should be published in eLife. Prior to publication there are some issues that need to be addressed and these are listed below.

1) Please clarify whether Figure 1 is a new image or adapted from Caragine et al., 2018.

2) In the fourth paragraph of the Introduction, the authors note that the nucleoplasm is an active fluid and that chromatin dynamics are active and coherent. It would help if the authors were to clarify what they mean by the terms active and coherent. These have specific meanings in the active matter field, and given the possible lack of familiarity that eLife readers might have with these concepts. In fact, I'd like to know precisely what the authors mean by active and coherent in this particular context.

3) The authors indicate that "the number of nucleoli per nucleus decreases…"; where is this shown in the data?

4) The fit of the p(V) raises some questions of importance: explanation or consideration. In order to be able to interpret the model that emerges from the fits, I propose the following: (1) fit the data to a proper power law of the form p(V) = f(V)*V^(α – 1) and fix α to be > 0; (2) then assess the value of α; (3) according to these fits, α should be equal to 2; (4) this would indicate that the distribution of nucleolar volumes follows a model where the moments V^m diverge for all values of m ≥ α – 1. This implies that estimates of moments, including the first moment viz., the average, will not converge no matter how many samples one extracts. This hypothesis should be tested. I suspect it will be true and if true, the implication is that the underlying features of nucleoli probably derive from Levy flight statistics and are incorrectly analyzed in terms of classical random walks, which lead to erroneous viewpoints such as effective diffusion coefficients and/or effective surface tensions. The underlying asymptotic scale free nature of volume distributions points to interesting properties that aren't explored here, but should perhaps be addressed.

5) Unless I missed it, the sphericity of the DFC wasn't quantitatively analyzed and yet the authors suggest that the DFC exhibits spherical shapes.

6) The trajectories shown in Figure 5A, C, and G of are difficult to parse. Inasmuch as these are trajectories, it might help to have some annotation so one can follow how the trajectories evolve.

7) I felt hard pressed to compare the violin plots to one another. There is a lot of really important information in these plots and the biological implications are rather significant. Therefore, I ask that the authors consider using something like a Kullback-Leibler divergence measure as a way to compare the prior and posterior distributions where the prior would be the default and the posterior would be the effect of the biological perturbation. Absent this, it is hard to judge if the inferences are qualitative or quantitative.

8) The data reveal some extraordinary insights, but interestingly the conclusions seem to either be more general or not about the data themselves. For example, the notion that there are two distinct behaviors for fusing vs. non-fusing nucleoli seems like a bit of a stretch from a quantitative standpoint. Absent a rigorous quantification, I am not convinced that the data do indeed reveal a strong dependence of the nucleolus-nucleoplasm interface on the chromatin transcriptional activity and the packing state. It is an attractive idea, but do the data really support such a strong assertion? I am not entirely sure. As noted in point 4 above, the power law distribution needs further analysis and explanation because this does imply very interesting physics that might even explain the apparent DLCA like behavior that the authors have observed. The authors appear to be invoking a direct connection to ideas from active matter, specifically motility induced phase separation; however, it is not clear that the data directly connect to these ideas. It is more than possible that I have missed the connection. Perhaps the authors would consider elaborating on how their data connect to ideas put forth by Cates and coworkers.

Reviewer #2:

The article by Caragine et al. reports the results of extensive live cell imaging of human HeLa cells expressing H2B-GFP as a chromatin marker and NPM-DsRed, and sometimes also FBL-mCerulean, as nucleolar markers. The senior author, Dr. Zidovska, has previously reported that the high viscosity of the nucleoplasm influences the fluid behavior of nucleoli [Caragine et al., 2018] and extends these studies in the current report to include analysis of nucleolar size and shape at different cell cycle stages, upon depletion of ATP, and after several different biochemical perturbations. The authors emphasize that nucleoli are tethered to and immersed within a fluid sea of chromatin, a perspective often overlook in studies of nucleoli. The results are generally consistent with our understanding of the fluid nature of nucleoli from past studies from others although these authors report the most detailed analysis of human nucleoli to date. The results in Figure 2 show that the size distribution of nucleoli follows a power law with an exponent of -1, in contrast to the exponent of -1.5 observed previously for frog oocyte nucleoli by Cliff Brangwynne. The authors note that differences in the numbers of nucleoli between the two systems may account for the difference in size distribution. Simultaneous observation of NPM1 in the GC and FBL in the DFC showed that DFCs behave like colloidal particles in the fluid GC and that they consistently occupy ~0.1 volume fraction of nucleoli (Figure 3). These results are consistent with findings reported by Cliff Brangwynne in 2016 [Feric, et al., 2016]. The authors compared the movements of nucleoli that, in some cases, lead to encounters and fusion (Figures 4 and 5). They state that the movements of non-fusing and fusing nucleoli are different, as follows: "While the non-fusing ones appear to move randomly through the nucleoplasm, nucleoli that will fuse in the near future, show a bias in motion towards each other." This is a truism; if two nucleoli don't move towards each other, they won't fuse. The authors speculate about the influence of chromatin dynamics on nucleolar fusion but did not perform any experiments to test their ideas. Next, the authors imaged cells after depletion of ATP and observed that nucleoli become irregularly shaped, which they suggested was due to decreased surface tension, although this suggestion was not tested through experiments. Finally, the authors subjected cells to chemical treatments that alter the cytoskeleton, chromatin or protein synthesis and characterized nucleolar number, size and shape using live imaging. In some cases, the treatments led to changes in nucleolar features but this very large dataset did not provide insights of fundamental significance regarding nucleolar structure and function and interactions with the surrounding chromatin. The Conclusions section of the paper presents a series of almost exclusively speculative statements about various aspects of interactions between fluid chromatin and nucleoli. Many of these are interesting hypotheses for testing in the future but are not appropriate as conclusion statements in a primary research article in eLife.

The article presents the results of a powerful live cell imaging platform capable of monitoring the structural and dynamic features of chromatin and nucleoli in live cells in real time. However, the results presented largely confirm past observations regarding the features of nucleoli and do not provide new insights. The results can serve as a starting point for future investigations of, for example, links between the state of surrounding chromatin and nucleolar function, which the authors are encouraged to undertake.

Reviewer #3:

The authors extensively studied the dynamics and morphology of individual nucleoli as well as relative dynamics between fusing or non-fusing nucleoli pairs in HeLa cells. These experiments are further extensions of their recent Physical Review Letters paper. However, I find their current effort compelling as they are inferring nucleoli-nucleoplasm interactions, an understudied important topic in the field of LLPS. The authors have established a number of quantitative measurements. Using these assays, they were able to evaluate the effects of ATP hydrolysis on dynamics of nucleoli and have found out some ATP-dependent cellular processes likely regulate dynamics of nucleoli. They also evaluated alterations to dynamics of nucleoli by perturbations to cytoskeleton dynamics, transcription, epigenetic regulation, or protein synthesis. For example, they have found out that alterations to chromatin compaction via epigenetic perturbation cause salient changes in nucleoli dynamics. While it is well-accepted that ATP-dependent processes can regulate dynamics biomolecular condensates, the observation of potential interplay between global epigenetics landscape and nucleoli regulation is rather intriguing. It is highly conceivable that scientists in the field can reveal a lot more nucleoli-nucleoplasm interactions and other biomolecular condensates-surrounding environment interactions by combining these quantitative measurements and genetic perturbations. For at least these reasons, I recommend the publication of this manuscript in eLife.

https://doi.org/10.7554/eLife.47533.sa1

Author response

While this paper builds on your past publication in PRL in 2018, there are considerable new insights in this manuscript pertaining to the interplay between spontaneous vs. driven processes in regulating nucleolar shapes, sizes, and cell cycle dependent dynamics. However, it is important to highlight the new findings more clearly, and very importantly, to link the data reported in this manuscript to the conclusions drawn. At present, the conclusions are very general and speculative, rather than closely tied to the actual observations. Please find the three reviews appended below as it will help you address additional significant points.

We thank the editor for the valuable feedback and the three reviews. In our original manuscript, we have included our data-related conclusions interspersed throughout the Results section and listed only general conclusions in the Conclusions section, which we now realize must have been confusing. Encouraged by the editor’s and reviewers’ comments, we have remedied this in the revised manuscript, by listing and explaining all our data-related conclusions in a new section Discussion and Conclusions. We have also introduced a new Table 1, which provides an overview of our quantitative results. We hope this will help to clarify many of the questions below.

Reviewer #1:

[…]

1) Please clarify whether Figure 1 is a new image or adapted from Caragine et al., 2018.

We thank the reviewer for pointing this out. Images in Figure 1A are new and were not previously published, images in Figure 1B are in part new and in part adapted from Caragine et al., 2018; specifically, only 2 panels from 14 panels in Figure 1B were previously shown and have been adapted for eLife. To account for this, in the revised manuscript we added an explanatory note into the Figure 1 legend.

2) In the fourth paragraph of the Introduction, the authors note that the nucleoplasm is an active fluid and that chromatin dynamics are active and coherent. It would help if the authors were to clarify what they mean by the terms active and coherent. These have specific meanings in the active matter field, and given the possible lack of familiarity that eLife readers might have with these concepts. In fact, I'd like to know precisely what the authors mean by active and coherent in this particular context.

The reviewer raises a great point and we completely agree that the terms “active” and “coherent” referring to the chromatin dynamics should be explained. By “active”, we mean that the chromatin dynamics is ATP-dependent, i.e. it dissipates energy [Zidovska et al., 2013]. In the same study we found that chromatin dynamics is “coherent” over several seconds and microns, i.e. chromatin motion is correlated within micron-scale patches inside the cell nucleus [Zidovska et al., 2013]. In the revised manuscript, we have added an explanation for both terms in the fourth paragraph of the Introduction.

3) The authors indicate that "the number of nucleoli per nucleus decreases…"; where is this shown in the data?

We thank the reviewer for pointing this out. The fact that the number of nucleoli decreases with the progressing cell cycle is shown in Figure 2E, specifically, in the temporal evolution of the distributions measured at 1.5h (blue), 3h (yellow) and G2/M (green). The distribution becomes heavier at small nucleolar numbers (1-5) as the time progresses indicating that more nuclei possess smaller number of nucleoli.

4) The fit of the p(V) raises some questions of importance: explanation or consideration. In order to be able to interpret the model that emerges from the fits, I propose the following: (1) fit the data to a proper power law of the form p(V) = f(V)*V^(α – 1) and fix α to be > 0; (2) then assess the value of α; (3) according to these fits, α should be equal to 2; (4) this would indicate that the distribution of nucleolar volumes follows a model where the moments V^m diverge for all values of m ≥ α – 1. This implies that estimates of moments, including the first moment viz., the average, will not converge no matter how many samples one extracts. This hypothesis should be tested. I suspect it will be true and if true, the implication is that the underlying features of nucleoli probably derive from Levy flight statistics and are incorrectly analyzed in terms of classical random walks, which lead to erroneous viewpoints such as effective diffusion coefficients and/or effective surface tensions. The underlying asymptotic scale free nature of volume distributions points to interesting properties that aren't explored here, but should perhaps be addressed.

The reviewer raises a great point. The fact that the nucleolar volume distribution scales as p(V) ~ V-1 has very interesting consequences. Indeed, such distribution is divergent and so is its first moment, the mean. To examine this property of our p(V) distribution, we have performed the following statistical test: for each condition from Figure 2F, e.g. unsynchronized cells (NN = 1331), we have split our data into 3 random nonoverlapping subpopulations and evaluated their means. We find that the means are quite similar. This is not unexpected, as our p(V) distribution does have finite bounds, specifically, there are physical cut-offs as to what size a nucleolus can have: with maximum being the size of the cell nucleus and minimum being the molecular scale of nucleolar components, although in practice the minimum size that can be experimentally measured, is given by the optical resolution. Thus, in between those bounds, p(V) can be integrated and also possesses a mean. However, the scale-free nature of p(V) even in between those bounds suggests that there is no preferred length scale associated with the nucleolar size. This is very interesting as it suggests that nucleoli of any size can coalesce and there is no preferred size they need to reach before/after they coalesce.

In the revised manuscript we have expanded the discussion of the intriguing properties and consequences of the measured volume distribution in the Results section and in the Discussion and Conclusions section. To more fully investigate the mechanism underlying this phenomenon we propose to address it in a future study. We also hope that our experimental observations will encourage theorists to develop new theories describing this phenomenon.

5) Unless I missed it, the sphericity of the DFC wasn't quantitatively analyzed and yet the authors suggest that the DFC exhibits spherical shapes.

The reviewer is correct, in the original manuscript we have not measured the sphericity of DFCs, since a visual inspection of our data revealed that DFCs appear to be close to spherical. Thus, we fitted the observed DFCs with circles to extract their areas/radii. Encouraged by the reviewer’s comment, we have developed a new data analysis that would allow us to assess the eccentricity of each DFC. To do so, we have developed a feature finding algorithm to identify DFCs, which we then manually verify and fit each DFC with a circumscribed and an inscribed circle, measuring the semi-major DFC axis a and the semi-minor DFC axis b, respectively. This way, we can evaluate the DFC eccentricity e=a/b. Using this algorithm we identified 1279 DFCs over 114 nucleoli in 63 nuclei, and after the removal of the DFCs that were out of focus, we obtained measurements of a, b, e and area for 1035 DFCs.

Our results are shown in the revised Figure 3B, where we present the measurements of both the semi-major DFC axis a (red) and the semi-minor DFC axis b (green), together with the Gaussian fits of their respective distributions. From the Gaussian fits we obtain the following average values: aDFC = 210 ± 50 nm and bDFC = 180 ± 40 nm.

Next, we evaluate the eccentricity e=a/b for each DFC. The distribution of the measured eccentricities e is shown in Figure 3—figure supplement 1A. Our data shows, that the DFC shape is indeed very close to spherical with the mean e = 1.22 ± 0.17. Figure 3—figure supplement 1B shows the distribution of areas of single DFCs, where the DFC area is calculated as A = πab for every DFC. Figure 3—figure supplement 1C shows the distribution of volumes for the DFCs, where V=4πa3/3, providing the upper boundary on the volume estimate.

In the revised manuscript, we have inserted the description of the DFC eccentricity measurement in the Results section, and the Materials and Methods section. We have also updated Figure 3B and C with the plots of data using our new algorithm and revised Figure 3—figure supplement 1.

6) The trajectories shown in Figure 5A, C, and G of are difficult to parse. Inasmuch as these are trajectories, it might help to have some annotation so one can follow how the trajectories evolve.

We thank the reviewer for pointing this out. To remedy this, we have replaced the groups of trajectories originally shown in Figure 5A and G with a single set of trajectories giving an example of a non-fusing and fusing pair of nucleoli. This way the trajectory plots are much larger and easier to parse. In addition, we have color-coded these trajectories by their time progression (blue to red), visualizing their evolution in time. More examples of trajectories with color-coded time progression, for both non-fusing and fusing nucleoli, are presented in the Figure 5—figure supplement 1.

7) I felt hard pressed to compare the violin plots to one another. There is a lot of really important information in these plots and the biological implications are rather significant. Therefore, I ask that the authors consider using something like a Kullback-Leibler divergence measure as a way to compare the prior and posterior distributions where the prior would be the default and the posterior would be the effect of the biological perturbation. Absent this, it is hard to judge if the inferences are qualitative or quantitative.

We thank the reviewer for the helpful suggestion. We agree that a quantitative comparison of the distributions of the physical quantities would be very helpful. To that end we have expanded our analysis of the distribution statistics also to the comparative analysis. Thus, in addition to the means and standard deviations of all distributions and p-values with respect to the control (Table 1), in the revised manuscript, we have also computed the relative difference of means with respect to control, the skew of all distributions and the Kullback-Leibler divergence with respect to the control (Table 1). Such analysis allows us to compare the measured distributions against the control as well as to evaluate the statistical significance of their deviation from the control, showing that some of the observed changes are indeed significant. We present all statistical characteristics that we computed for the measured distributions in Table 1 in the revised manuscript and provide their definitions and discuss the specific significant changes in the Results section.

8) The data reveal some extraordinary insights, but interestingly the conclusions seem to either be more general or not about the data themselves.

We thank the reviewer for this helpful comment and apologize for the confusion. In our original manuscript, we have included our data-related conclusions interspersed throughout the Results section and listed only general conclusions in the Conclusions section, which we now realize must have been confusing. Encouraged by the reviewer’s comment, we have remedied this in the revised manuscript, by listing and explaining all our data-related conclusions in a new section Discussion and Conclusions. This might clarify many of the questions below.

For example, the notion that there are two distinct behaviors for fusing vs. non-fusing nucleoli seems like a bit of a stretch from a quantitative standpoint.

The reviewer raises a valid point. We based our claim that fusing and non-fusing nucleoli exhibit different dynamics on purely quantitative evidence analyzing the dynamic behavior of fusing and non-fusing nucleoli (Figure 5). As shown in Figure 5C and I, the velocity amplitudes of two fusing nucleoli are highly correlated with each other exhibiting strong linear dependence (Figure 5I, Pearson correlation coefficient ρ = 0.88), unlike the non-fusing nucleoli, whose velocity amplitudes are totally independent (Figure 5C, Pearson correlation coefficient ρ = 0.41). Further, the velocities measured for fusing nucleoli are 30% smaller than for non-fusing nucleoli (Figure 5B and H), with p-value of 4⋅10-4. Similarly, the distribution of radial velocities exhibit much larger variance for non-fusing nucleoli than for the fusing ones (Figure 5D and J), with the variance obtained by fitting a Gaussian curve to both distributions: σrad, fusing = 1⋅10-4 µms-1 for the fusing nucleoli and σrad, non-fusing = 3⋅10-4 µms-1 for the non-fusing nucleoli. All of these differences are statistically significant and hint at a possible process responsible for bringing the fusing nucleoli together.

In the revised manuscript we address this point in the Results and the new Discussion and Conclusions sections. We plan to investigate further the mechanism beyond this phenomenon in a future study.

Absent a rigorous quantification, I am not convinced that the data do indeed reveal a strong dependence of the nucleolus-nucleoplasm interface on the chromatin transcriptional activity and the packing state. It is an attractive idea, but do the data really support such a strong assertion? I am not entirely sure.

We thank the reviewer for raising this point and apologize for the confusion. Our data shows that the nucleolus-nucleoplasm interface is highly sensitive to chromatin transcriptional activity and packing state, the former evidenced by large (40-50%) increase in the roughness of the nucleolus-nucleoplasm interface (fneg, Nneg) upon α-amanitin and flavopiridol treatment (both transcription inhibitors) and (~15%) increase in nucleolar size upon α-amanitin treatment, the latter evidenced by a significant (~20%) decrease in the roughness of the nucleolus-nucleoplasm interface and (~15%) decrease in nucleolar size upon trichostatin A treatment (leads to chromatin decondensation). Specifically, α-amanitin is a polymerase II transcription inhibitor, flavopiridol is a transcription elongation inhibitor and trichostatin A is a histone deacetylase inhibitor leading to chromatin decondensation. The statistical significance of all the observed changes was verified by evaluating the p-values and as per reviewer’s suggestion we also computed the Kullback-Leibler divergences for measured distributions of physical quantities. We have also evaluated the skew and the relative changes of means measured for distributions of physical quantities upon biochemical perturbations with respect to the control.

To strengthen this point, in the revised manuscript, we have introduced Table 1 in Results section, which provides the summary of statistical characteristics for measured distributions of physical quantities upon biochemical perturbations such as mean, standard deviations, p-values (with respect to control), the relative differences of means (with respect to control), skew and the Kullback-Leibler divergence (with respect to control). Table 1 provides a direct overview of the observed quantitative changes. Furthermore, we also emphasize these results in the Results section and in the new Discussion and Conclusions section.

As noted in point 4 above, the power law distribution needs further analysis and explanation because this does imply very interesting physics that might even explain the apparent DLCA like behavior that the authors have observed.

See point 4 above.

The authors appear to be invoking a direct connection to ideas from active matter, specifically motility induced phase separation; however, it is not clear that the data directly connect to these ideas. It is more than possible that I have missed the connection. Perhaps the authors would consider elaborating on how their data connect to ideas put forth by Cates and coworkers.

We thank the reviewer for bringing up this point. In earlier studies [Stenhammer et al., 2015, Smrek and Kremer, 2017], an activity driven phase separation was observed in both polymer and colloidal systems consisting of a mixture of active and passive components. In both systems (polymers, colloids), the active entities phase separate from the passive ones. We speculate that in our case the nucleolus could be abstracted as a passive droplet (or colloid) immersed in the active polymer (chromatin). In such case, it is conceivable that the active positional fluctuations of the polymer (chromatin) could cause demixing of the passive phase and thus bring together the passive colloids (nucleoli), i.e. the active entities phase separate from the passive entities in the system. As the mechanism that brings two nucleoli remains mysterious, we speculate that the active motion of the chromatin could indeed contribute to bringing the nucleoli together. In the revised manuscript, we expand our discussion of this point in the new Discussion and Conclusions section.

Reviewer #2:

[…] The results in Figure 2 show that the size distribution of nucleoli follows a power law with an exponent of -1, in contrast to the exponent of -1.5 observed previously for frog oocyte nucleoli by Cliff Brangwynne. The authors note that differences in the numbers of nucleoli between the two systems may account for the difference in size distribution.

We thank the reviewer for raising this point. We note that the kinetics of human nucleolar assembly likely differs from that of the frog oocyte due to the much lower nucleolar count in human cells (~100 times less than in frog oocyte), as well as the fact that human somatic nucleoli are connected to the chromatin fiber and therefore are not free to diffuse as it is in the case of nucleoli in the X. laevis oocyte [Gall et al., 2004, Brangwynne et al., 2011, Caragine et al., 2018, Berry et al., 2018]. Further differences between the two systems include a large difference in the nuclear size (~ 1 mm diameter in frog oocytes, ~ 10 µm diameter in human cells) and the nucleolar size with volumes of 10–103 µm3 in frog oocytes [Brangwynne et al., 2011] and 10-2–102 µm3 in human cells [current study]. Moreover, frog oocytes contain a dense actin network, which was also implicated in the maintenance of the nucleolar interactions [Feric et al., 2016]. Considering all of these differences, it is conceivable that the different nucleolar volume distributions observed for frog oocytes, p(V)~V-1.5, and human cells, p(V)~V-1, might hint at two distinct mechanisms at play in the two systems.

In the revised manuscript, we have extended our discussion of this point in the Discussion and Conclusions section.

Simultaneous observation of NPM1 in the GC and FBL in the DFC showed that DFCs behave like colloidal particles in the fluid GC and that they consistently occupy ~0.1 volume fraction of nucleoli (Figure 3). These results are consistent with findings reported by Cliff Brangwynne in 2016 [Feric, et al., 2016].

The reviewer raises a good point. While our DFC data is in general consistent with the earlier study by [Feric et al., 2016], it also shows striking differences between the two studied systems: frog oocytes in [Feric et al., 2016] and human cells in this work. Specifically, the DFCs in frog oocytes were found to behave as viscoelastic droplets of varying sizes in micrometer range, which can fuse [Feric et al., 2016]. In contrast, we find that DFCs in human cells behave as solid-like colloidal particles, which do not fuse. Moreover, these colloids are monodisperse granules of submicron size with very well defined dimensions: semi-major axis 210 ± 50 nm, semi-minor axis 180 ± 40 nm and eccentricity ~ 1.2 ± 0.17. Furthermore, the DFC volume fraction of nucleoli is only ~ 0.1 human cells, while it is ~ 0.25 in frog oocytes [Feric et al., 2016]. It is also noteworthy that one frog oocyte DFC can be larger than the entire human nucleolus. Thus, the overall differences between the frog oocyte and human cells are indeed significant.

In the revised manuscript we have addressed this point in the Discussion and Conclusions section. We have also inserted new measurements of the DFC semi-major and semi-minor axes (Figure 3B), as well as eccentricity (Figure 3—figure supplement 1A) in the Results section.

The authors compared the movements of nucleoli that, in some cases, lead to encounters and fusion (Figures 4 and 5). They state that the movements of non-fusing and fusing nucleoli are different, as follows: "While the non-fusing ones appear to move randomly through the nucleoplasm, nucleoli that will fuse in the near future, show a bias in motion towards each other." This is a truism; if two nucleoli don't move towards each other, they won't fuse.

We thank the reviewer for pointing this out and apologize for the confusion. What we intended to state was that two fusing nucleoli exhibit a linear correlation in the magnitudes of their velocities (Figure 5I). Conversely, the velocity magnitudes of non-fusing nucleoli are uncorrelated (Figure 5C). In addition, the velocity magnitudes, both absolute (Figure 5B and H) and radial (Figure 5D and J), of fusing nucleoli are on average smaller than those of non-fusing ones. All of these differences are statistically significant as verified by the Pearson coefficient and p-values. In the revised manuscript, we have rephrased the statement in the subsection “Dynamics of Fusing and Non-Fusing Nucleoli”.

The authors speculate about the influence of chromatin dynamics on nucleolar fusion but did not perform any experiments to test their ideas.

We thank the reviewer for bringing this up. This comment was meant to be speculative and thought-provoking following up on our data-related conclusions. We realize that it might have been confusing as in the original manuscript we had our data-related conclusions interspersed throughout the Results section. In the revised manuscript, we have summarized all our data-related discussion and conclusions in a new section Discussion and Conclusions.

Next, the authors imaged cells after depletion of ATP and observed that nucleoli become irregularly shaped, which they suggested was due to decreased surface tension, although this suggestion was not tested through experiments.

We thank the reviewer for raising this point. In our earlier study [Caragine et al., 2018] we have established a direct connection between the surface roughness and surface tension of the nucleoli. Specifically, local nucleolar surface deformations, which may be driven thermally or by active processes, are opposed by the surface tension. The surface tension drives the minimization of the surface of a liquid-liquid interface, i.e. a 3D liquid droplet strives to become a sphere. Thus, the larger the surface roughness, i.e. the less spherical the droplet, the lower its surface tension. Under this assumption, the effective surface tension of the nucleoli upon ATP depletion is lower than in the control, since we found that nucleoli exhibit rougher (more uneven) surface as quantified by the fraction of negative curvature fneg and the total number of the regions of negative curvature Nneg. These experiments and findings are summarized in Figure 6 and a newly introduced Table 1. While we are not measuring the absolute value of the surface tension, our experiments allow for relative comparison between the control and ATP-depleted nucleoli via e, fnegand Nneg. In the revised manuscript, we address this point in detail in the Discussion and Conclusions section.

Finally, the authors subjected cells to chemical treatments that alter the cytoskeleton, chromatin or protein synthesis and characterized nucleolar number, size and shape using live imaging. In some cases, the treatments led to changes in nucleolar features but this very large dataset did not provide insights of fundamental significance regarding nucleolar structure and function and interactions with the surrounding chromatin.

We thank the reviewer for raising this point and apologize for the confusion. The main focus of our large dataset was to interrogate the nucleoplasm-nucleolus interactions, specifically, involvement of certain biological processes. To do so, we have evaluated in great detail the shape and roughness of the nucleoplasm-nucleolus interface upon biochemical perturbations perturbing specific processes. Our data shows that the nucleolus-nucleoplasm interface is highly sensitive to chromatin transcriptional activity and packing state, the former evidenced by large (40-50%) increase in the roughness of the nucleolus-nucleoplasm interface (fneg, Nneg) upon α-amanitin and flavopiridol treatment (both transcription inhibitors) and (~15%) increase in nucleolar size upon α-amanitin treatment, the latter evidenced by a significant (~20%) decrease in the roughness of the nucleolus-nucleoplasm interface and (~15%) decrease in nucleolar size upon trichostatin A treatment (leads to chromatin decondensation). Specifically, α-amanitin is a polymerase II transcription inhibitor, flavopiridol is a transcription elongation inhibitor and trichostatin A is a histone deacetylase inhibitor leading to chromatin decondensation. The statistical significance of all these changes was verified by evaluating the p-values as well as the Kullback-Leibler divergences for measured distributions of physical quantities. Our findings suggest that the nucleolus-nucleoplasm interface is dependent on certain active processes, and thus extends our equilibrium liquid-liquid phase separation (LLPS) picture of the nucleolus.

Furthermore, we find that cytoskeletal perturbations do not lead to any significant changes in the nucleolar size as well as the roughness of the nucleolus-nucleoplasm interface (fneg, Nneg). Conversely, in frog oocytes latrunculin A facilitates nucleolar fusion as the dense nuclear actin network dissolves leading to an increase in the nucleolar size [Feric et al., 2016].

In the revised manuscript, we have introduced Table 1 in Results section, which provides the summary of statistical characteristics for measured distributions of physical quantities upon biochemical perturbations such as mean, standard deviations, p-values (with respect to control), skew and the Kullback-Leibler divergence (with respect to control). Table 1 provides a direct overview of the observed quantitative changes. Furthermore, we also emphasize these results in the Results section, and in the new Discussion and Conclusions section.

The Conclusions section of the paper presents a series of almost exclusively speculative statements about various aspects of interactions between fluid chromatin and nucleoli. Many of these are interesting hypotheses for testing in the future but are not appropriate as conclusion statements in a primary research article in eLife.

We thank the reviewer for this helpful comment and apologize for the confusion. In our original manuscript, we have included our data-related conclusions interspersed throughout the Results section and listed only general conclusions in the Conclusions section, which we now realize must have been confusing. Encouraged by the reviewer’s comment, we have remedied this in the revised manuscript, by listing and explaining all our data-related conclusions in a new section Discussion and Conclusions.

The article presents the results of a powerful live cell imaging platform capable of monitoring the structural and dynamic features of chromatin and nucleoli in live cells in real time. However, the results presented largely confirm past observations regarding the features of nucleoli and do not provide new insights. The results can serve as a starting point for future investigations of, for example, links between the state of surrounding chromatin and nucleolar function, which the authors are encouraged to undertake.

We thank the reviewer for feedback and valuable comments. In our original manuscript, we have included our data-related conclusions interspersed throughout the Results section and listed only general conclusions in the Conclusions section, which we now realize must have been confusing. Encouraged by the reviewer’s comments, we have remedied this in the revised manuscript, by listing and explaining all our data-related conclusions in a new section Discussion and Conclusions. This might clarify many of the questions above.

In this work, we study the nucleolus as the archetype of the liquid-liquid phase separation (LLPS) in biology and monitor its size, shape and dynamics during its entire lifetime. We discover a rich phenomenology that grows the LLPS framework in new and unexpected ways:

i) We find that nucleoli exhibit anomalous dynamics and anomalous volume distribution during the cell cycle that defies any current theory and necessitates a new one.

ii) We find that the nucleolar fluid in human cells is in fact a colloidal solution.

iii) We find that the surrounding nucleoplasm plays a key role in the LLPS of nucleoli that might have been previously overlooked.

iv) We identify specific biological processes participating in the nucleolus-nucleoplasm interactions.

https://doi.org/10.7554/eLife.47533.sa2

Article and author information

Author details

  1. Christina M Caragine

    Center for Soft Matter Research, Department of Physics, New York University, New York, United States
    Contribution
    Resources, Data curation, Software, Formal analysis, Investigation, Visualization, Writing—original draft, Writing—review and editing
    Competing interests
    No competing interests declared
  2. Shannon C Haley

    Center for Soft Matter Research, Department of Physics, New York University, New York, United States
    Contribution
    Resources, Software, Formal analysis, Investigation
    Competing interests
    No competing interests declared
  3. Alexandra Zidovska

    Center for Soft Matter Research, Department of Physics, New York University, New York, United States
    Contribution
    Conceptualization, Resources, Data curation, Software, Formal analysis, Supervision, Funding acquisition, Validation, Investigation, Visualization, Methodology, Writing—original draft, Project administration, Writing—review and editing
    For correspondence
    alexandra.zidovska@nyu.edu
    Competing interests
    No competing interests declared
    ORCID icon "This ORCID iD identifies the author of this article:" 0000-0002-4619-4424

Funding

National Institutes of Health (R00-GM104152)

  • Alexandra Zidovska

National Science Foundation (CAREER PHY-1554880)

  • Alexandra Zidovska

National Science Foundation (CMMI-1762506)

  • Alexandra Zidovska

The authors declare that the funders had no role in study design, data collection and interpretation, or the decision to submit the work for publication.

Acknowledgements

This research was supported by the National Institutes of Health Grant R00-GM104152 and by the National Science Foundation Grants CAREER PHY-1554880 and CMMI-1762506.

Senior Editor

  1. Detlef Weigel, Max Planck Institute for Developmental Biology, Germany

Reviewing Editor

  1. Arup K Chakraborty, Massachusetts Institute of Technology, United States

Reviewers

  1. Rohit V Pappu, Washington University in St Louis, United States
  2. Pilong Li

Publication history

  1. Received: April 9, 2019
  2. Accepted: October 13, 2019
  3. Version of Record published: November 26, 2019 (version 1)

Copyright

© 2019, Caragine et al.

This article is distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use and redistribution provided that the original author and source are credited.

Metrics

  • 1,097
    Page views
  • 212
    Downloads
  • 0
    Citations

Article citation count generated by polling the highest count across the following sources: Crossref, PubMed Central, Scopus.

Download links

A two-part list of links to download the article, or parts of the article, in various formats.

Downloads (link to download the article as PDF)

Download citations (links to download the citations from this article in formats compatible with various reference manager tools)

Open citations (links to open the citations from this article in various online reference manager services)

Further reading

    1. Neuroscience
    2. Physics of Living Systems
    Louis Kang, Michael R DeWeese
    Research Article Updated
    1. Cell Biology
    2. Physics of Living Systems
    Florian Thüroff et al.
    Research Article