Cell culture

HeLa and KB cells were used because they show prominent intranucleolar histone H2B-GFP fluorescence (S1C and S2 Figs) corresponding to the nucleosomal domains with the ultrastructural appearance of ICC. In the experiments dedicated to LM/EM distribution of UBF, fibrillarin, and histone H2B in living and fixed cells we applied transient transfection using human KB cells. KB cells contain well visible intra-nucleolar inclusions of ICC (S2 Fig) and large FCs associated with prominent DFC zones that facilitate LM discrimination of nucleolar sub-territories occupied by r-gene transcription and pre-rRNA processing machineries (S1A and S1B Fig). In addition, preliminary experiments revealed high stability of KB cells to damage provoked by the transfection procedure that yielded a large quantity of mono- and doubly-transfected cells.

Cultures were either mono-transfected with UBF-GFP or fibrillarin-GFP plasmids or co-transfected with H2B-GFP and UBF-dsRed plasmids. Transfected and non-transfected stock cultures were maintained in 40 ml flasks containing DMEM (Gibco, UK) supplemented with 10% calf serum and 1% penicillin/streptomycin mixture. Cells were reseeded in new medium 2–3 times per week, as soon as the monolayer became late preconfluent or early confluent. Tests for mycoplasma detection were performed monthly.

For inhibition of rRNA synthesis, cells were treated during 1–8 h by a low dose (0.05 μg/ml) of AMD (Sigma, USA) in two different conditions: 1) Working with a Bio-Rad MRC-1024ES Laser confocal microscope (LCM), we delivered the drug using a Bioptech (USA) perfusion system (for details see below). We noted that in this condition, the AMD concentration did not reach a high level immediately and thus, nucleolar remodeling developed slowly and took around 7 h to reach complete segregation. 2) Therefore, working with a Zeiss 710 NLO LCM and ILAS Spinning Disk confocal system (SDCS) we delivered AMD by direct addition to the culture medium for immediate contact of cells with the full concentration of drug; in these conditions nucleolar remodeling developed more rapidly so that we reached complete segregation after around 2–3 h. The same procedure was used for UBF fluorescence studies in fixed cells.

Transient transfection, observation and imaging of KB cells

A suspension of 5 x 10⁵ KB cells/ml was seeded onto the surface of Ø40 mm glass coverslips (Bioptech) placed in a Ø50 mm plastic Petri dish. Fugene-6 (Roche Diagnostics, Switzerland) was used to transfect cells with a cDNA construct coding for UBF-GFP, fibrillarin-GFP or histone H2B-GFP or in some experiments with two plasmids coding for H2B-GFP and UBF-dsRed. Coverslips were fixed in 4% paraformaldehyde (PAF, EMS, USA), mounted in Citifluor AF1 (Agar Scientific, UK) and stored at 4°C. Alternatively, part of the transfected and fixed cells was submitted to anti-GFP immunolabeling and TEM analysis. Preparations selected for LCM were examined and imaged on a Bio-Rad LCM (described above) combined with an Olympus IX70 fluorescence microscope using a Plan-Apochromat/x60/1.40/Oil objective (for details see S1–S3 Methods).

Conventional TEM analysis

Sub-confluent KB/UBF-GFP and HeLa/H2B-GFP expressing cells on Ø40 mm coverslips were rinsed (3 x 5 min) with 0.1M phosphate buffer (Biomérieux, France) and then fixed during 10 min with 4% PAF and 2.5% glutaraldehyde (both from EMS) in 0.1M phosphate buffer. After postfixation in 1% OsO4 (EMS) in the same buffer, the cells were rinsed with phosphate buffer and distilled water (MiliQ, Millipore, USA), scraped from coverslips, sedimented in 30% BSA (Sigma-Aldrich, USA) using 0.5 ml centrifuge tubes, and jellified after removing BSA by adding to the cell pellet a few drops of 25% glutaraldehyde. The pellets were cut into ~1mm³ pieces, washed in distilled water overnight, and then dehydrated in a series of acetone (Sigma, USA) in deionized water up to 100% acetone. Then cells were infiltrated in a mixture of 100% acetone-Embed 812 (EMS) 1:1 and 1:2 (overnight). The impregnation was continued with 2 changes of pure Embed 812. Finally, the samples were placed for polymerization at 60°C in BEAM capsules (#3, EMS) filled with a fresh portion of Embed 812. Silver-gold sections (~0.1 μm) cut using a Reichert-Young Ultracut ultramicrotome (Reichert, Austria) were contrasted with 5% aqueous uranyl acetate and lead citrate (both from Merck, USA) and then observed and imaged in a Hitachi H300 electron microscope at 70 kV.

Pre-embedding anti-GFP immuno-TEM labeling of UBF-GFP transfected KB cells

Ultrastructural localization of GFP-tagged proteins in fixed transfected KB cells was performed by using anti-GFP immuno-TEM, combining nanogold (Nanoprobes, USA) labeling with silver enhancement (HQ Silver Kit, Nanoprobes, USA) [57]. To detect GFP-positive sites in TEM we used the following antibodies: (i) anti-human GFP mouse monoclonal (Roche Diagnostics); (ii) goat anti-mouse (Jackson, USA); (iii) streptavidin-nanogold conjugate (Nanoprobes). Cells were harvested without osmication by meticulous scraping from the glass surface, then dehydrated and embedded in epoxy resin as described in the previous section. Ultrathin sections were picked up on 200 mesh copper grids (EMS) and stained with uranyl acetate and lead citrate and examined and imaged as described above (for details see S4 Method).

Time-lapse observation and imaging of living KB cells

To trace the dynamics of fibrillarin and UBF redistribution in the course of nucleolar segregation during AMD treatment we performed long-term 4D studies using fibrillarin-GFP and UBF-GFP expressing cells growing on Ø40 mm coverslips, using a living cells observation system based on a Bioptechs FCS2 incubation chamber with a specimen and objective temperature control module and peristaltic pumps that enabled perfusion with fresh medium and delivery of AMD at 37±0.1°C. The flow of the pumps was adjusted to 10 ml/h. Images were collected using a Bio-Rad MRC-1024ES/Olympus IX70 LCM in fast scanning mode (1/2 sec for each 512x512 image) to diminish the exposure of cells to the laser beam. To demonstrate the general features of the behavior of components containing fibrillarin and UBF during nucleolar segregation we used simple visualization tools in the form of 2D movies (standard 2D+time mode) provided by ImageJ or QuickTime. For spatial reconstruction with subsequent dynamic modeling in the form of 4D movies we utilized a more sophisticated approach, Rev4D software developed by the CRESTIC group (University of Reims, Reims, France) to analyze and visualize the evolution of multiple topology-changing structures during their deformation, division, fusion, etc. [58, 59]. To demonstrate redistribution of nucleolar proteins in 4D the most significant data were extracted from time/z-series, then 3D reconstructed using Amira 5.5 software (TGS, Mercury Computer Systems, France) and finally displayed in chronological order as a gallery of consequential volume models (for details see S5 Method).

Imaging of histone H2B and UBF in living HeLa cells

To study the 4D dynamics of the ICC and its relationship with FCs during nucleolar segregation and capping, we used a CLEM approach combined with post-fixation immunolabeling of UBF. For time-lapse imaging a suspension of histone H2B-GFP tagged HeLa cells at 60–70% confluence was plated on Ø35 mm uncoated glass “MatTek” or “Ibidi μ-Dish-500” Petri dishes (MatTek, USA and Ibidi GmbH, Germany) with Ø14 mm and Ø21 mm wells as growth area. The bottom coverslip of these dishes is an etched finder grid that facilitates the search for regions/cells of interest (ROI/COI). After 24–48 h of incubation the cells were briefly rinsed 3x with PBS and incubated in fresh medium at 37°C for 2–3 h before experiments.

All procedures connected with confocal microscopy such as selection and marking of appropriate cells or groups, capture of control z-stacks as well as time-lapse imaging during AMD action, were conducted at 37°C in a CO₂-enriched atmosphere using a special microscope plate holder with a close-fitting lead. To outline a ROI, the culture was first previewed and recorded under low magnification on a Zeiss 710 NLO LCM and Spinning Disk ILAS2 confocal system (SDCS) (Roper Scientific, USA) coupled with a Zeiss Axioobserver inverted microscope with a 512x512 EMCCD camera (for details see S6 Method).

Postfixation imaging of HeLa cells after anti-UBF immunolabeling

Imaging of labelled UBF is widely used to identify undercondensed active or inactive rDNA genes folded into the structure of interphase FC and metaphase NORs [15, 60–62]. To perform UBF labeling, AMD treated H2B-GFP transfected cells were immediately fixed at the end of time-lapse imaging. To label UBF we used: (i) mouse anti-UBF/F9-fragment monoclonal antibodies (Santa Cruz Biotechnology, USA); (ii) biotinylated goat anti-mouse antibodies (Jackson, USA); (iii) streptavidin-Alexa568 (Invitrogen Molecular Probes, USA). After immunolabeling, COI previously imaged during time-lapse experiments were found by their position on the finder grid. After 3D imaging of immunostained UBF, cells were prepared for TEM and CLEM analysis (for details see S7 Method).

CLEM technique

Cells growing in MatTek dishes were fixed in 4% PAF, rinsed in PBS, and post-fixed in 2.5% glutaraldehyde and 1% OsO4 in 0.1M phosphate buffer (Biomérieux) and rinsed with PBS and distilled water. The cells were dehydrated through a series of ethanol-deionized water mixtures up to 100% ethanol, infiltrated in a mixture of 100% ethanol-Embed 812 (EMS) 1:1 and 1:2, and then with pure Embed 812, and polymerized in dishes at 60°C. Silver-gold serial ultrathin sections (~0.1μm) were picked up using coated one-hole copper slot grids (Pella, USA) or uncoated H7 Hexagonal Maxtaform grids (EMS), stained with uranyl acetate and lead citrate, and imaged by electron microscopy as described above (for details see S8 Method).

CM/TEM overlay of doubly-tagged HeLa cells

To merge identical ROI imaged by CM and EM within the same cell, a modified CTO technique was utilized [56]. For merging of the two images it is essential that the ultrathin and optical sections are at approximately the same depth and orientation of a given nucleus/nucleolus. Thus we used serial ultrathin sectioning; the block face, trimmed for ultramicrotomy, was adjusted relative to the knife so that every point of the surface containing the COI was at the same distance from the knife edge. Further sectioning provided sections beginning at the top surface of the block and penetrating the cell in the Z direction to a depth of several tens of μM. Thus the appropriate plane of interest containing nucleoli could be chosen during examination of serial sections in TEM and its depth was noted. Then the corresponding depth on LSM image stacks was calculated and appropriate planes were visualized using virtual serial slicing. Overlays were performed using Corel Draw software. Serial fluorescent and EM images were adjusted to the same magnification and nuclei/nucleoli, FCs, ICCs, and NVs were contoured in different colors on transparent plastic sheets using fine permanent markers, overlaid, and roughly aligned with the TEM image below and the confocal image above. The transparency of the confocal image was adjusted so that the TEM image was seen and the images were precisely aligned by adjusting their orientation and magnification.

A total of 21 different experiments were performed. In each of them, a minimum of 10 cells were observed and different ROI were imaged. For example, although ROI shown on S9 Fig contains 9 interesting cells, only 4 cells from this series (marked by numbers) were used for CLEM. Altogether, twenty double-labeled cells belonging to different experimental series were fully analyzed using the CLEM approach. In order to display nucleolar changes accompanying the gradual decrease in number of FCs, we show the three most demonstrative cases that appeared after treatment with AMD for 1 h. These were: cells N°1 and N°2 containing pre-segregated nucleoli with several FCs from series shown on S9 and S14A–S14E Figs and one cell from another series, containing one ring-shaped nucleolus with one FC (S14F–S14I Fig). Additionally, two nucleoli were analyzed in cell N°2.

Nucleoli remain independent units upon inhibition of rRNA synthesis

Time-lapse imaging of cells containing fibrillarin-GFP was performed to study the 3D dynamics of nucleoli during inhibition of rRNA synthesis by AMD. Fibrillarin was chosen because: i) it is present at high concentration within the DFC where its rRNA methyl transferase activity is required for rRNA processing, and ii) it is also present but at lower concentration within the other nucleolar compartments and within the nucleoplasm. These characteristics allowed the simultaneous imaging of DFC, whole nucleoli, and nucleoplasm and the study of their 3D modifications during time series. Thus, 100 z-stacks containing 60 optical sections each were acquired every 5 min for 8 h. Projection of the sections in each z-stack was performed to create a movie (S1 Movie) showing the modifications arising in a representative cell, so that we were able to simultaneously visualize: i) the reorganization of DFC within each nucleolus, ii) all nucleoli, and iii) the contour of the nucleus. Using surface visualization we observed that each nucleolus, which was initially extended and irregularly shaped, became spherical during inhibition of rRNA synthesis (for example, nucleolus N°2 in Fig 1, S3 Fig and S1 and S2 Movies; see Fig 1 for numbering).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. 4D evolution of nucleolar volume and topography during AMD treatment in living KB cells.

One hundred Z-stacks were acquired during 8 h and images were generated by Rev4D software. (A–H) Gallery of 3D reconstructions displaying nuclear and nucleolar evolution and showing the whole nucleus (low threshold, transparent yellow), the nucleoli (medium threshold, transparent orange) and the DFC (high threshold, solid red) (same cell as in S8 Fig and S1–S5 Movies). (F–H) Trajectories of the center of mass of each nucleolus, demonstrating that their relative position is fixed, that there is no nucleolar fusion, and that all nucleoli rotate around nucleolus #2 (see traces of nucleoli 1 to 6 on H). The scale bars represent 10 μm in (A); 8 μm in (F). (I, J) Changes of volume and topography of six nucleoli within the same nucleus, shown in different colors. During inhibition, nucleoli decrease in size (I) and are positioned at a constant distance from each other (J).

https://doi.org/10.1371/journal.pone.0187977.g001

Using Rev4D software, we simultaneously visualized the 3D organization of DFC (high threshold, solid red), of nucleoli (medium threshold, transparent orange) and of whole nuclei (low threshold, transparent) yellow (Fig 1). To analyze the behavior of nucleoli, we identified their center of mass at each time-point and calculated their trajectories in space and time, their relative distance, and their volume. These analyses showed that during AMD treatment, nucleoli maintained their relative distance although they rotated by around 90° in the nucleus (the center of rotation is close to nucleolus N°2) and they lost half of their initial volume (S3–S5 Movies).

Altogether, these data demonstrate that inhibition of rRNA synthesis induces a strong internal reorganization of nucleoli (and of the DFC in different nucleoli) but never induces their fusion. Moreover, they show that nucleoli keep constant positions relative to one another during this process.

This analysis also showed that the re-organization of the DFC (S5 Movie) was a sequential process during which its cord-like structure was first broken into several fragments. Then, these fragments became spherical, came close together, and finally fused to constitute 1 to 3 caps at the border of each reorganized spherical nucleolus. As the DFC is the site of early steps of rRNA processing, we can hypothesize that its spatiotemporal behavior could be different from that of rDNA genes and of chromatin. Thus, in the following we investigated how active rDNA genes and condensed chromatin reorganize when rRNA transcription is inhibited by AMD.

Inhibition of rRNA synthesis induces enlargement of nucleolar structures containing UBF

As UBF is specifically associated with active undercondensed rDNA genes [17–19] we used confocal microscopy of UBF-GFP within fixed KB cells to analyze the behavior of active rDNA genes (S4 Fig). In control cells (S4A–S4C Fig), GFP-UBF was present in numerous nucleolar spheroidal foci ~ 0.3–0.5 μm in diameter, disposed as so-called “necklaces” by 3D visualization. After 1 h with AMD (S4D–S4F Fig), all nucleoli became smaller and ovoid. They contained several prominent UBF-GFP positive spheres ~ 0.5–1 μm in diameter located within areas of low phase contrast. After 2 h of inhibition (S4G–S4I Fig), UBF-GFP fluorescence was present within 1–3 cap-like structures ~ 1–2 μm in their larger axis located on the outer part of spheroidal nucleoli within large caps of low phase contrast.

Time-lapse confocal microscopy shows a sequential gathering and fusion of nucleolar structures containing UBF during inhibition of rRNA synthesis

To understand exactly how structures containing UBF are reorganized during AMD treatment, we studied living cells by time-lapse confocal microscopy. Thus, 100 z-stacks containing 60 optical sections each were acquired every 5 min for 8 h. For each z-stack, a projection of optical sections and 3D visualization were performed (for example Fig 2 and S6 Movie). Before AMD treatment, UBF-GFP fluorescence appeared as numerous spherical entities ~0.3–0.5 μm in diameter disposed as a 3D chain-like structure within 4 nucleoli (Panels A1 and A6 in Fig 2). From 30 min to 2.5 h of AMD treatment, all fluorescent UBF spots came close to each other but maintained their number and size.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. 4D dynamics of UBF-GFP during inhibition of rRNA synthesis in living KB cells.

(A–D), gallery of fluorescence maximum intensity projections and 3D reconstructions displaying the dynamics of UBF-GFP during nucleolar segregation. Reorganization of UBF-GFP spots was based on their successive gathering, fusion, enlargement, and final grouping into two or three caps localized at the nucleolar periphery. Blue and red arrowheads identify two large nucleoli to demonstrate their rotation within the nucleus during the step-by-step reorganization of UBF-GFP spots and (E1-E4) their 3D visualization at times 0 and 8 h respectively. The scale bars represent 10 μm.

https://doi.org/10.1371/journal.pone.0187977.g002

Between 2.5 and 5 h, UBF-positive spots strongly reorganized within smaller and roundish nucleoli. Their number decreased due to their coalescence, whereas their individual size and fluorescence intensity increased markedly. Four to five hours after beginning AMD treatment, nucleoli contained a few large and brightly fluorescent aggregates which shifted toward the nucleolar periphery and fused to constitute two to three brightly fluorescent caps 0.7–3.5 μm in size. At 8 h, the four nucleoli contained UBF caps

To better analyze the step-by-step coalescence of UBF-GFP structures, we chose to zoom and orientate 5 of these structures during the period between 2 h 45 min and 7 h 30 min (Fig 3). Here, five UBF-positive spheroids (N°1—N°5) contained in a ROI (delimited by the blue dotted line) successively gathered to form a typical cap. Between 2 h 45 min and 7 h 30 min, spheroids N°1and N°2 fused to give structure A; the latter gathered with N°3 to give B and, finally, N°4 and N°5 gathered with B to give C. At the end of this process, C had a typical 3D cap structure which resulted from both fusion and gathering of the different spheroids and which was much more compact than the initial ROI.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. 4D visualization of UBF-GFP dynamics during inhibition of rRNA synthesis in the same living KB cells as presented on Fig 2.

(A–T) An example of 4D dynamics demonstrating successive fusion and gathering of five UBF positive spheroids (N°1 –N°5) contained in a ROI (delimited by the blue dotted line). Between 2 h 45 min and 7 h 30 min, we noticed the following steps: spheroids N°1and N°2 fused to give structure A; the latter gathered with N°3 to give B, and N°4 and N°5 finally gathered to B to give C. Note that the cap resulting from the fusion and gathering of the different spheroids is more compact than the initial region limited by the blue dotted line. The scale bar represents 2 μm.

https://doi.org/10.1371/journal.pone.0187977.g003

Altogether, as UBF-GFP is a marker of under-condensed rRNA genes, we conclude that during AMD treatment fusion of rRNA gene clusters is restricted to each nucleolus and never occurs between clusters located in different nucleoli. To further question if condensed nucleolar chromatin behaves in the same way, we investigated its 3D reorganization by confocal microscopy in fixed and in living cells.

The intra- and perinucleolar condensed chromatin network reorganizes strongly during AMD treatment

3D organization of the ICC network in control HeLa cells.

In fixed control KB and HeLa cells transiently or stably expressing histone H2B-GFP, we imaged a 3D network of nucleolar chromatin (Fig 4, S1C and S2 Figs) constituted of a shell of PCC in continuity with strands of ICC.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. ICC network inside the nucleolar volume of one control HeLa cell expressing histone H2B-GFP.

(A) X/Y plane section of the nucleus. (B–D) Three consecutive optical sections (sections 27 (S27), 29 (S29) and 30 (S30) passing through the nucleolus. The ICC network was clearly seen (red arrow on C). (E) Y/Z plane section of the nucleus. (F-I) Two consecutive Y/Z (F, G) and X/Z (H, I) plane sections at higher magnification. Red arrows indicate strands of ICC localized in the depth of the nucleolar volume. The scale bars represent 7 μm in (A); 4.5 μm in (B-D); 5.5 μm in (E); 4.5 μm in (F-I).

https://doi.org/10.1371/journal.pone.0187977.g004

In cells fixed at different times during the action of AMD, we observed that the ICC network became less and less extended compared to control cells (Fig 5A–5E). After 3 h of treatment (I), the ICC was seen as a small single nucleolar spheroid which was often in contact with the inner surface of the nucleolar cap (Fig 5F) or PCC.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 5. AMD-induced remodeling of ICC as revealed in fixed HeLa cells stably expressing H2B-GFP.

(A–E) Gradual transition of ICC from a network-like organization (red arrows) to coarse clumps during 1.5 h of AMD treatment. (A) After 15 min the ICC network was more prominent than in control cells. (B) After 30 min the ICC network began coarsening. (C) After 45 min gradual shrinkage of the ICC network and fusion of individual ICC areas into large clumps became obvious. (D, E) During the next 45 min, ICC areas transformed into single large spheroids, often in contact with the PCC shell. (F) Nucleolar capping after 3 h; only tiny ICC inclusions (red arrows) attached to the inner margin of the nucleolar cap were observed. The scale bars indicate 6 μm in (A, B); 5 μm in (C); 4.5 μm in (D); 4 μm in (E); 4.5 μm in (F).

https://doi.org/10.1371/journal.pone.0187977.g005

4D dynamics of ICC and PCC during AMD treatment.

To further understand the precise reorganization of nucleolar chromatin, we performed time-lapse imaging of living HeLa cells expressing histone H2B-GFP. The displacement of ICC inside the nucleolus during AMD treatment is shown in movies (S7 and S8 Movies) and a gallery displaying 2D images of the corresponding movie (Fig 6). Initially (start of experiment, Fig 6A) all nucleoli revealed fluorescence mostly dispersed in the form of a fine meshwork, consisting of intermittent thin filaments and small clumps that always contacted the PCC. Within the first 15 min drastic changes in the localization and morphology of chromatin were observed; the filamentous structure began to condense into coarse aggregates (15 min, Fig 6B) and we observed a progressive condensation and increase of thickness of the PCC shell. After 30 min, we observed a significant disorganization of the intranucleolar meshwork: the filamentous appearance of histone H2B-GFP labeling was replaced by large clumps (30 min and 45 min, Fig 6C and 6D). Simultaneously these large clumps shifted from the nucleolar interior towards the PCC. During this migration several clumps approached each other and fused before they coalesced with the PCC shell (45 min, 60 min and 90 min). Thus, the number of GFP labeled intranucleolar structures decreased gradually. Next, the ICC shifted completely to the periphery and appeared like PCC protrusions into the nucleolar body (105–180 min). Frequently, fluorescence of H2B-GFP was no longer visible in the nucleoli.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 6. The dynamics of histone H2B-GFP during inhibition of rRNA synthesis in living HeLa cells.

This gallery of optical sections was extracted from z-stacks of corresponding time series and displayed with 15–45 min intervals to fix the main stages of ICC evolution (corresponds to S8 Movie). The nuclei of three cells in the white rectangle on the initial image (“Start of experiment”) were enlarged to visualize the evolution of the ICC in more detail. In the course of AMD treatment we observed a gradual condensation of filamentous structures of the ICC into coarse clumps during 120 min with their migration from the nucleolar interior towards the PCC shell (165 min). During this movement several ICC clumps approached each other and fused just before the coalescence with the PCC. Note that at the end of the experiment (180 min) there was no more chromatin within the nucleolus. The scale bars represent 12 μm.

https://doi.org/10.1371/journal.pone.0187977.g006

The intra- and perinucleolar condensed chromatin network is always in contact with structures containing UBF during nucleolar reorganization

To analyze the 3D relationships between the nucleolar condensed chromatin and structures containing UBF (termed UBF spots hereafter) we imaged immunolabeled UBF and H2B-GFP (or UBFdsRed and H2B-GFP) in fixed cells. In 3D views of control cells (Fig 7; S5 Fig), ICC appeared as an intra-nucleolar network connecting UBF spots with the PCC shell. This clearly demonstrated the proximity of ICC and UBF spots and the spatial integration of the latter into the whole network of NAC.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 7. Spatial proximity of intranucleolar chromatin (ICC), perinucleolar chromatin (PCC), and UBF spots in fixed control cells.

(A) 3D reconstruction of the nucleus in a HeLa cell stably expressing histone H2B-GFP using volume rendering. A virtual cube inserted into the nuclear volume delineates the nucleolar territory (nl) extracted in order to analyze its interior. (B) The nucleolar volume reconstructed using simultaneous volume rendering and surface visualization of intranucleolar H2B-GFP for better visualization of ICC. After volume extraction, the nucleolar interior revealed a network of ICC interconnected with the PCC shell. (C, D) Relationships between ICC (green) and UBF spots (red) within the nucleolus of KB cells cotransfected with H2B-GFP and UBF-dsRed. These 3D views demonstrate, inside the nucleolus, the presence of the ICC network in contact with UBF spots. The scale bar represents 3.5 μm in (A, B); 5 μm in (C) and 3.5 μm in (D).

https://doi.org/10.1371/journal.pone.0187977.g007

This spatial interaction between UBF spots and histone H2B-GFP could be also demonstrated on 3D reconstructions of nucleoli of cells treated with AMD during 1 h (Fig 8). Moreover we also observed that ICC clumps, which were linked to UBF spots, were also in contact with a PCC shell and created a bridge between all these components. On the other hand, the same analysis performed on fully segregated nucleoli (2 h of AMD treatment) confirmed that there was no more ICC and that the convex side of the large UBF-positive caps was in contact with PCC.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 8. Interaction between UBF spots, ICC, and PCC within the pre-segregated nucleoli of HeLa and KB cells.

(A-C) Consecutive optical sections (S.21-23) of nucleoli in HeLa cells stably expressing histone H2B-GFP. Note the prominent ICC clumps within the nucleolar interior after rRNA synthesis inhibition. These cells were immunolabeled for UBF at the end of time-lapse imaging (the corresponding nuclei are marked by red stars on A, D). (D-F) The same optical sections merged with 3D surface rendering of UBF spots (red). Note the direct contact between UBF spots and ICC inclusions. At the same time UBF spots that shifted to the nucleolar periphery became incorporated into the solid PCC shell. (G-I) 3D reconstructions of enlarged UBF spots extracted from the nucleolus of a KB cell doubly transfected with H2B-GFP and UBF-dsRed plasmids, fixed and imaged. A typical pre-segregated nucleolus containing five large UBF spots reconstructed using volume rendering (green) and surface visualization (red). Incorporation of peripherally-located UBF spots into the massive PCC shell as well as their contact with ICC can be readily recognized. Note the ICC clump that is in contact with two UBF spots. After extraction according to the virtual cubes in H, images were rotated at an appropriate angle to exemplify how UBF spots can be linked to each other by ICC. The scale bars represent 10 μm in (A); 5 μm in (G, H); 2 μm in (I).

https://doi.org/10.1371/journal.pone.0187977.g008

Altogether, this analysis of fixed cells shows the permanent contact between UBF spots and ICC clumps during nucleolar reorganization.

Ultrastructural imaging shows that UBF is always contained within FCs during inhibition of rRNA synthesis

To identify the nucleolar structures which contain UBF during the reorganization induced by AMD, we immunolocalized UBF-GFP molecules and imaged them by electron microscopy, using a pre-embedding technique [35]. In control cells (Fig 9), numerous silver/gold particles identifying UBF-GFP were localized in FCs (limited by a dotted line) lined by a discontinuous cord of DFC (delimited by a black line on Fig 9C). Very few particles were identified within the DFC, granular component (GC), nucleoplasm, or cytoplasm.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 9. Ultrastructural pre-embedding localization of UBF within the nucleolus of control KB cells.

(A) An abundant labeling (silver/gold particles) inside several fibrillar centers (fc) (outlined by a white dotted line). A few particles (black arrows) were visible within the nucleolar dense fibrillary component (dfc), nucleolar granular component (gc), nucleoplasm, or cytoplasm. (B, C) At high magnification, numerous particles were located within fibrillar centers (fc) (outlined by white dotted lines). The dfc is outlined by a black line on C. The scale bars represent 500 nm.

https://doi.org/10.1371/journal.pone.0187977.g009

After 1 h of treatment with AMD (Fig 10A), silver/gold particles identifying UBF-GFP were exclusively localized within FCs which were positioned close to each other and were frequently linked to vacuoles or interstices (arrows) containing small clumps of ICC. Contrary to control cells, the DFC partly surrounded the FC and was thicker; moreover the GC was more homogeneous and compact. After 2 h (Fig 10B), UBF-GFP labeling was found in cap-shaped FCs located on the periphery of ovoid and compact nucleoli. A direct contact between these FCs and large clumps of PCC was frequently observed.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 10. Ultrastructural pre-embedding localization of UBF within the nucleolus of KB cells treated with AMD for 1 h (A) or 2 h (B).

(A) Abundant labeling (silver/gold particles) inside two large ovoid fcs (outlined by a white dotted line). White arrows point to electron-transparent vacuoles (or interstices) containing small clumps of intranucleolar condensed chromatin (icc). The dfc is outlined by a black line. (B) Numerous silver/gold particles were located within three cap-shaped fcs (outlined by white dotted lines) positioned at the nucleolar periphery. Note the direct contact between large clumps of perinucleolar condensed chromatin (pcc; outlined by a dark dotted line) and two fcs. Very few particles were found on the dfc (outlined by a black line) and the gc. Scale bars represent 500 nm.

https://doi.org/10.1371/journal.pone.0187977.g010

Successive time-lapse confocal microscopy and ultrastructural imaging of the same cells shows that condensed intranucleolar chromatin and fibrillar centers are close together and are contained within nucleolar vacuoles

To analyze the ultrastructure of nucleolar components in which ICC clumps observed in living cells are localized during the pre-segregated stages of AMD action, we developed a correlative light and electron microscopy (CLEM) approach (Fig 11, S6 and S17 Figs). The different steps of this approach were:

1. step 1 (S6 and S7 Figs): a group of COI (red circle on S6A–S6F and S7 Figs) was imaged by time-lapse confocal microscopy to analyze the AMD-induced 3D reorganization of intra-nucleolar chromatin tagged with H2B-GFP;
2. step 2 (S8 Fig, S9 Movie): after 1 h of AMD treatment, cells were fixed and immunostained for UBF. The same COI were then imaged by confocal microscopy to perform a simultaneous 3D visualization of ICC and UBF;
3. step 3 (S9 Fig): the cells were processed for electron microscopy and the COI were identified and trimmed; these cells were then cut through numerous 0.1 μm serial ultrathin sections and their nuclei were imaged at high magnification (S10–S13 Figs);
4. step 4 (Fig 11, S14–S17 Figs): the contours of nucleolar components identified on the serial ultrathin sections were superposed to build 3D models of different nucleoli treated with AMD during 1 h. Confocal images of the same nucleoli were used to confirm the identification of ICC (GFP fluorescence) and of FCs (UBF fluorescence)

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 11. CLEM approach: Step 4.

Within cells treated with AMD during 1 h, the contours of nucleolar components identified on serial ultrathin sections were superposed to build 3D models of four different nucleoli (nl). Confocal images of the same nucleoli were used to confirm the identification of ICC (GFP fluorescence) and of FC (UBF fluorescence). The color-code is: nuclear and nucleolar contours—black and light blue lines; FCs—red and brown lines; ICC—black shading; NVs—green lines; DFC—dark blue lines. (A) 3D reconstruction of the equatorial part of the nucleolus in cell N°1(serial ultrathin sections on S10 Fig and serial fluorescent images shown on S14D and S14E Fig). (B, C) Two pre-segregated nucleoli in cell N°2 (serial ultrathin sections on S11 and S12 Figs and serial fluorescent images shown on S14A–S14C Fig). (D) 3D reconstruction of a nucleus with a ring-shaped nucleolus with one FC (serial ultrathin sections on S13 Fig and serial fluorescent images shown on S14F–S14I Fig).

https://doi.org/10.1371/journal.pone.0187977.g011

Altogether, this overlay technique (CTO) allowed the demonstration that ICC clumps and FC are always close together during all the steps of nucleolar reorganization and are contained within a complex network of nucleolar vacuolar channels.

Are nucleoli “immobile” due to the fixed position of rDNA genes within NOR-bearing chromosomal territories?

It is known that active r-DNA genes occupy and share, together with mRNA genes, a central position in the nucleus [19, 78]. In the present study we showed that during segregation and capping, nucleoli do not change their initial relative positions within the nuclear volume, so that nucleolar relocation and fusion was never observed. In addition, FC and DFC from different nucleoli never gathered or fused, always staying within the limits of their own nucleolar territory. Consequently, each nucleolus may represent a separate compartment emerging due to the spatial interplay between r-chromatin loop(s) and nucleolar associated chromatin belonging to distinct NOR-bearing chromosomal territories that include ribosomal and non-ribosomal genomic regions. These findings imply that active and inactive nucleoli have a similar central location within the nucleus. This could result from the attachment of corresponding NOR-bearing chromosomal territories to the nuclear matrix that plays a key role in the compartmentalization and therefore in the function of the nucleus [79].

Concerted contraction of nucleolar associated chromatin puts nucleolar components in motion

The ability of FCs and DFCs to move and fuse within a crowded nucleolar volume was revealed by early EM studies on artificial inactivation of r-genes, including the first experiments where AMD was used as an inhibitor [1–3, 52–54]. However, the mechanisms underlying spatial displacement of r-genes and transcription sites, as well as the sites of pre-rRNA processing at molecular and structural levels, have not been elucidated so far. Our interpretation of the dynamics of intra-nucleolar inactivation involves the mobility of the whole nucleolar-associated chromatin system on the one hand, and its structural integrity with the FC and adjacent DFC (hereafter FC/DFC assembly) on the other. It is the simultaneous contraction of the ICC and PCC and perinucleolar dislocation of ICC that induce the motion of the FC/DFC assembly ending in the form of cap-shaped structures tightly adherent to the PCC rim.

Contrary to a relatively fixed position of rDNA genes in response to their selective inhibition, we observed clear intradomain dynamics. Therefore, together with strong rotational movement of the nucleolus we traced the intranucleolar dynamics of FCs, particularly, their initial (within 1 h) confluence into larger structures, and followed up their further migration from the center of the nucleolus to its periphery. The motion of individual FCs was apparently asynchronous.

As a rule, large FCs localized on the periphery of the nucleolus remained immobile for a long time until their final transformation into caps. At the same time, smaller FCs, more dispersed throughout the nucleolar volume, were continuously migrating. The structures resulting from the fusion of single small FCs gradually moved up to approach the larger “immobile” FCs, first coming closer and then merging with them.

ICC clumps also demonstrated similar trajectories and asynchrony, whereas large peripherally located ICC blocks bound to the PCC rim exhibited low mobility. Analysis of the movies shows that the disorganization of the intra-nucleolar chromatin network involved the gradual gathering of the filamentous structure of ICC into coarse aggregates and their shift from the nucleolar interior towards the PCC. Like FCs, in later steps of AMD action, ICC clumps gradually shifted to a perinucleolar location. However, peripherally located ICC clumps, before they completely merged with the PCC rim, were capable of keeping a long-lasting contact with the PCC and although relatively “immobile”, they still remained clearly visible as distinct intranucleolar structures. At the same time, new blocks from the inner regions of the nucleolus started to move closer to them because of further condensation of the unit ICC network. They either merged with larger ICC clumps or, having reached the nucleolar periphery, stayed “immobile” and visible within a certain period of time until incorporated into PCC. These live cell observations strongly support our proposition that gradual coalescence events are largely involved in the mechanisms of nucleolar segregation and capping.

Is intranucleolar dynamics of FCs restricted due to attachment to PCC via DJ heterochromatin?

The dynamics of FCs observed during time-lapse imaging can be explained by the progressive condensation of nucleolar-associated chromatin in response to inactivation of r-gene transcription. The physical bond between FCs and ICC may provoke their movement due to the dynamics of nucleolar-associated chromatin constituents. Thus within incompletely segregated nucleoli, FCs were always in contact with intranucleolar blocks of chromatin, which in turn were connected with PCC that forms a thick layer around an inactivated nucleolus. The position of peripherally shifted ICC blocks is reminiscent of that of flanking rDNA heterochromatin protrusions, DJ in particular [25]. This suggests that ICC may represent protrusions of PJ and DJ into nucleolar territory whereas PCC blocks surrounding the nucleolar territory may be constituted by centromeric and telomeric heterochromatin. Finally, the disappearance of ICC (observed on Figs 5 and 6) could result from its incorporation into PCC. After this step, FCs spread onto the surface of the residual nucleolus and we propose that they acquired a crescent-like shape due to their squeezing between the PCC rim and the spherical body of the residual nucleolus.

Dynamic model of the nucleolus

Now, our aim is to formulate a dynamic model embracing the data about a possible structural and functional impact of nucleolar-associated chromatin and of the vacuolar component in the global organization and inactivation dynamics of the nucleolus. By merging fluorescence and EM images we reached the conclusion that r-chromatin of FCs, together with intra- and perinucleolar condensed chromatin, is part of a unit system structurally integrated into the network of nucleolar vacuoles. Importantly, during the selective inactivation of r-gene transcription this system reacts in a concerted manner, so that 3D reorganization of the vacuolar system proceeds in parallel with the rearrangement of the FC/NAC unit. Therefore we consider nucleolar vacuoles rather as an active nucleolar constituent, crucial for the spatial distribution of ICC and FCs. In this context we hypothesize that continuous ICC strands, emanating from the PCC shell, are interconnected with FCs through the network of united channels attributed on ultrathin sections to the vacuolar component.

The unit system described comprises the UBF-positive non-nucleosomal segments of r-chromatin, loosely folded at the transcription sites, as well as nucleosomal chromatin located between FCs and at the border with PCC. In contrast to nucleosomal chromatin, the relaxed state could be considered as ubiquitous, facilitating transcription and replication within FC. Meanwhile, nucleosomal intra- and peri-nucleolar condensed chromatin regions are physically connected with r-chromatin and may contain silenced r-genes as well as non-ribosomal genes. At least the portion of intra-nucleolar chromatin seems to be constructed from centromeric and telomeric heterochromatin segments linking respectively to the proximal and the distal rDNA flanking regions.

Thus, the functional organization of r-chromatin within the nucleolar space can exist due to the spatial interplay between rDNA and heterochromatic blocks emanating from peri-centromeric and peri-telomeric segments of the short chromosome arms. We can also propose that during nucleologenesis [80] the centromeric and telomeric heterochromatin blocks of several juxtaposed NOR-bearing chromosomal territories can interact, forming a PCC shell separating the nucleolar volume from the nucleoplasm. Then, it could be well assumed that the permanent repressive and heterochromatinized state of ICC or of the entire NAC may appear to provide the force ensuring the global integrity of the nucleolus.

Furthermore, peripherally-shifted ICC clumps directly linked to the surrounding PCC can be attributed to the peri-telomeric DJ regions, while ICC localized inside the nucleolar volume corresponds to peri-centromeric PJ chromatin. This sophisticated structure provides the structural support for spatial organization of the nucleolus.

The obvious implication of this model is the possibility of schematizing the pathway of nucleolar reorganization during inhibition of rRNA synthesis. During this inhibition the size of the nucleolus gradually decreases. Moreover, we propose that the density of compaction within intra-nucleolar chromatin increases with the inactivation of r-genes. Thus we can hypothesize that the chromatin of DJ origin positioned in the peripheral sites of intra-nucleolar condensed chromatin meshwork becomes more condensed, then shortened and shifted to PCC due to its contraction. During their migration towards PCC they also pull the connected FCs to the nucleolar periphery. Peri-centromeric intra-nucleolar condensed chromatin from the nucleolar interior, localized between FCs, becomes more condensed as well. As a result, the distance between FCs also decreases, so that FCs may juxtapose within the pre-segregated nucleoli. The juxtaposed FCs can be readily recognized as individual entities because they are still separated by ICC clumps. Very likely the physical/chemical state and therefore the dynamics of individual FC/DFC assemblies, enriched by UBF and pre-rRNA, can be similar to those of self-assembling soft active aggregates that are characterized by a high concentration of proteins and RNAs [81]. Hence, by using a targeted cryo-correlative nano-imaging approach, we recently demonstrated that DFC and FC of control cells contained around 70% and 80% of water, respectively [82]. In the same study, we also quantified the water content in cells treated with AMD for 3 h and found an increase of water content in all compartments of the nucleus and cytoplasm and, more particularly, that segregated DFC and FC (caps) contained around 80 and 90% of water, respectively. Thus, we propose that the increase of water content induced by AMD treatment could decrease the viscosity of nuclear compartments and facilitate the reorganization and fusion of chromatin clumps, of DFC and FCs as described for different types of protein- and RNA-rich NBs considered as condensed liquid-like droplets [29, 81]. Finally, ICC clumps inserted between FCs fused with the PCC shell, provoking coalescence of individual FCs and their gathering as nucleolar caps on the nucleolar periphery. This suggests that each nucleolar cap might be composed of several FCs juxtaposed on the margin of the nucleolar remnants.

Concluding remarks

In the present study we used several complementary cell imaging approaches to provide new insights into the spatial reorganization of r-genes upon AMD inactivation, relative to condensed chromatin with a canonical structure. Time-lapse confocal microscopy showed that: i) nucleoli remain independent units upon inhibition of rRNA synthesis, ii) in each nucleolus, structures containing UBF gather during a sequential process and are always in contact with the condensing network of intra and peri-nucleolar chromatin. Successive time-lapse confocal microscopy and ultrastructural imaging of the same cells demonstrate that all FCs contain UBF and are always in contact with condensing intra-nucleolar chromatin. Finally, the finding that FCs and intra-nucleolar chromatin are contained within large nucleolar vacuoles suggest that they are in continuity and belong to the same chromatin entity.