Evidence for karyoplasmic homeostasis during endoreduplication and a ploidy-dependent increase in gene transcription during tomato fruit growth

Variation in cell ploidy levels, termed somatic endopolyploidy, is part of normal development and physiology in many eukaryotic organisms, ranging from insects to mammals and plants (Brodsky and Uryvaeva, 1977). As a widespread phenomenon, endopolyploidy can result from several mechanisms, such as the generation of multinucleate cells originating from acytokinetic mitosis, nuclear fusion, endomitosis or from endoreduplication. The last two mechanisms represent the most frequently encountered modes of polyploidisation in eukaryotes (Lee et al., 2010). During endomitosis, chromosomes double and condense, sister chromatids separate normally, but chromosomes return to the interphase state within a single nucleus, thus doubling the chromosome number. Although present in various phyla, endomitosis is rare in the plant kingdom (D'Amato, 1984). Rather, endoreduplication is the major source of endopolyploidy in higher plants, particularly in angiosperms in which it arises in most species and in many different tissues (Nagl, 1976; D'Amato, 1984). During endoreduplication, endonuclear DNA duplication occurs in the absence of any obvious condensation and decondensation steps, thus producing chromosomes with multivalent (2, 4, 8, 16, etc.) chromatids without any change in chromosome number (Joubès and Chevalier, 2000).

The adaptive value of endoreduplication during plant development is not fully understood, and various functional hypotheses have arisen in the literature. As an inducer of nuclear DNA amplification, endoreduplication has been proposed to provide a means to sustain growth under adverse environmental conditions, such as genotoxic stress (Hase et al., 2006; Adachi et al., 2011), saline stress (Ceccarelli et al., 2006), water deficit (Cookson et al., 2006) and low temperature (Barow, 2006).

The most obvious role for endoreduplication during plant development has been ascribed to the determination of cell size, as positive correlations between endoreduplication and cell size are classically observed in many plant species, organs and cell types (Chevalier et al., 2011). According to the ‘karyoplasmic ratio’ theory, successive endocycles influence the final size of the cell, which tends to adjust its cytoplasmic volume with respect to the nuclear DNA content (for a review, see Sugimoto-Shirasu and Roberts, 2003).

As far as morphogenesis is concerned, endoreduplication may impact not only the final size of the cell but also its differentiation. In Arabidopsis, four rounds of endoreduplication are required during the developmental programme of epidermal hair cells called trichomes in order to produce mature many-branched single cells with an average DNA content of 32C (Hülskamp et al., 1999), and a direct correlation between the ploidy level and the number of branches does occur: mutations that affect the number of endocycles lead to trichomes with supernumerary or reduced numbers of branches (Ishida et al., 2008). As an important determinant of cell identity, endoreduplication was recently demonstrated to be crucial for cell fate acquisition and maintenance, thus contributing to the formation of specialised cell type patterns (Bramsiepe et al., 2010; Roeder et al., 2010). Related to cell specialisation, numerous correlative examples from the animal and plant kingdoms have led to the popular proposal that when associated with highly metabolically active cells, endoreduplication would support an increase in transcriptional and metabolic activities (Lee et al., 2010). However, this proposal has never been clearly demonstrated.

Investigating endoreduplication in tomato (Solanum lycopersicum Mill.) is pertinent, not only because elevated levels of endopolyploidy occur in the course of fruit development when compared with other plant organs and species, but also because it does contribute to cell and fruit growth in a developmentally and genetically regulated manner (reviewed by Chevalier et al., 2011). Recently, we developed a fluorescent in situ hybridisation method that allowed us to establish a ploidy map of tomato pericarp at the mature green stage (Bourdon et al., 2011), thereby overcoming the difficulty of determining ploidy at the cellular level in a tissue context. To pave the way towards an understanding of the functional role of endoreduplication during tomato fruit development, we have performed a structural and molecular analysis of endoreduplicated nuclei in fruit pericarp and provide the first direct evidence for a ploidy-dependent increase in gene transcription.

Cherry tomato (Solanum lycopersicum Mill. cv Wva106) plants were grown in a greenhouse under a thermoperiod of 25°C/20°C and a photoperiod of 14/10 hours (day/night). Fruits were harvested at the mature green stage [29-32 days post-anthesis (dpa)].

Tomato pericarp nuclei were prepared, sorted by flow cytometry and fixed on microscope slides as described (Bourdon et al., 2011).

For Chromomycin A3 labelling experiments, three fruits were used to prepare slides containing 500 nuclei of each ploidy level (2C to 128C) sorted directly onto three-well microscope glass slides (Superfrost, CML, France) prepared with a cushion of 20 μl 500 mM sucrose, 22 mM MgCl₂, 15 mM sodium citrate, 10 mM MOPS pH 7, 2% formaldehyde and 50 μg/ml Chromomycin A3. Image acquisition was performed promptly after sorting using a Leica DMI6000 microscope with a 20× objective and excitation (Ex) and emission (Em) wavelengths as follow: Ex 450-490 nm, Em long-pass 515 nm.

For 3,3′-dihexyloxacarbocyanine iodide [DiOC₆(3)] labelling experiments, equatorial fragments of 30 dpa fruit pericarp were fixed under vacuum infiltration at 20 kPa for 1 hour at 20°C in 4% paraformaldehyde in 0.1 M phosphate buffer pH 7.2. Fixed tissues were left in the fixative solution overnight then washed for 10 minutes in 0.1 M phosphate buffer. A vibrating blade microtome was used to obtain 150 μm sections as described (Bourdon et al., 2011). Staining was performed for 10 minutes in a 1.75 μM DiOC₆(3) solution in 0.1 M phosphate buffer pH 7.2. For Rhodamine-123 labelling experiments (Emaus et al., 1986), handmade equatorial slices were prepared. Staining was performed for 1 minute in 0.5 μM Rhodamine-123, 0.1 M phosphate buffer pH 7.2 supplemented with 0.3 M sorbitol. For both labelling experiments the sections were washed for 1 minute in their respective buffers, and stained for 1 minute with 10 μg/ml DAPI. Sections were then mounted on microscope slides in the same respective buffer and immediately observed by confocal microscopy using the following wavelengths: for DiOC₆(3), Ex 488 nm and Em 505-555; for Rhodamine-123, Ex 496 nm and Em 500-560 nm.

Two anti-RNA polymerase II (RNA pol II) antibodies (S2P and NP) were used for dual immunolabelling of nuclear suspensions supplemented with 5% BSA and 0.3% Triton X-100 (to eliminate plastid fluorescence). Nuclei were washed in 1.5 ml tubes by centrifugation (800 g, 4 minutes) through this buffer supplemented with 0.5 M sucrose, onto a 2 M sucrose cushion. For each of 5000 nuclei classified for ploidy by simultaneous DAPI staining, the immunolabelling intensities of each antibody (S2P and NP; secondary antibodies labelled with Alexa Fluor 633 and Alexa Fluor 488, respectively) and their ratio (S2P/NP) was determined by flow cytometry. Fluorescent images of the nuclei were recorded as z-series on a confocal laser-scanning microscope (Leica, Scanhead TCS-SP2 on a DMRXA2) with a 63×/NA1.4 HCX PL APO objective, using the following wavelengths: Ex 488 nm, Em 507-594 nm. For image acquisition of DAPI/transmission pericarp slices, a Nikon DM5000 epifluorescence microscope equipped with a CCD CoolSNAP HQ2 (Photometrics) was used and the fluorescence was recovered with a 450-490 nm band pass filter under Ex 340-380 nm.

For FISH experiments, three fruits were used for preparing slides containing up to 2000 nuclei of a given ploidy level (2C to 128C). Nuclei were sorted on slides covered with a 10 μl 500 mM sucrose cushion, allowed to dry at room temperature and stored at −20°C.

DNA FISH was performed as described (Bourdon et al., 2011) using a 45S rDNA ribosomal probe corresponding to a 9 kb EcoRI fragment encoding the Triticum aestivum 45S rDNA repeat unit (18S-5.8S-26S genes and spacers) (Gerlach and Bedbrook, 1979).

For RNA FISH, antisense and sense (negative control) probes for the 5.8S sequence of the tomato 45S rDNA transcription unit were amplified by PCR using tomato (cv Microtom) genomic DNA and the set of primers described in supplementary material Table S1. Amplicons were purified on an agarose gel using the Wizard PCR Preps DNA Purification System (Promega). Digoxigenin (DIG)-labelled RNA probes were synthesized by in vitro transcription using T7 RNA polymerase and the SP6/T7 DIG RNA Labelling Kit (Roche Diagnostics) according to the manufacturer's instructions. RNA FISH was performed directly after nuclei isolation as described (Jolly et al., 1997). Nuclei were prepared on ice from ~2 mm thick sections of 30 dpa fruit pericarp by gentle chopping in 1 ml CyStain UV ploidy buffer (Partec). The suspension was filtered through a 30 μm nylon mesh directly onto a 400 μl 2 M sucrose cushion in the same buffer and centrifuged at 1200 g for 10 minutes at 4°C in order to pellet amyloplasts. Then, 150 μl of this suspension were recovered just over the cushion interface and gently homogenised by pipetting. Finally, 10 μl of this homogenate were spread on pasteurised glass microscope slides and allowed to dry for 10-15 minutes. The prepared nuclei were then subjected to fixation, permeabilisation and dehydration steps as described in protocol 3 of Jolly et al. (Jolly et al., 1997). Hybridisation conditions and post-hybridisation washes were as used for DNA FISH on isolated interphase nuclei as described (Bourdon et al., 2011). Signal detection of DIG-labelled probes was carried out with anti-DIG-fluorescein (Roche Diagnostics) for 1 hour in 4× SSC, 0.2% Tween 20 solution according to the manufacturer's instructions. The nuclei were finally mounted and counterstained in Vectashield (Vector Laboratories) containing 1.5 μg/ml DAPI. Three negative controls were performed to confirm the specificity of this protocol: (1) hybridisation with the T7 transcribed 5.8S sense probe described above; (2) a 1 hour treatment with RNase immediately after the nuclei isolation step; (3) hybridisation with a mock treatment (DEPC-treated water instead of the DIG-labelled probe).

Fluorescent images of nuclei were recorded as described above using a confocal laser-scanning microscope and the following wavelengths: for 45S and 5S rDNA FISH (Texas Red), Ex 543 nm and Em 570-675 nm; for 5.8S rRNA FISH (fluorescein), Ex 488 nm and Em 500-550 nm; for DAPI, Ex 405 nm and Em 430-490 nm.

Tomato pericarp fragments were chopped in 3 ml Galbraith's extraction buffer (Galbraith et al., 2011) plus 0.1% Triton X-100 and 30 units RNasin Plus RNase inhibitor (Promega) and nuclei were sorted as described (Bourdon et al., 2011). Total nuclear RNA was extracted from 30,000 sorted nuclei using Trizol reagent (Invitrogen), purified using the RNeasy Mini Kit (Qiagen), and then treated with 0.06 units/μl DNA-free TURBO DNase (Ambion) to remove genomic DNA. The reverse transcription of total nuclear RNA was performed using I-Script (Bio-Rad) in the presence of random primers. The cDNA obtained was diluted 1/10 in distilled water. For real-time PCR, reaction mixes (20 μl) were prepared with 10 μl GoTaq qPCR Master Mix (Promega), 0.4 μM each primer and 5 μl diluted cDNA. To monitor gene amplification on genomic DNA, 100 sorted nuclei previously boiled for 5 minutes were added to the reaction mix instead of diluted cDNA. PCR amplifications were carried out using a Q-PCR iCycler (Bio-Rad) as: an initial denaturation step of 5 minutes at 95°C, followed by 40 cycles of 15 seconds at 95°C and 25 seconds at 58°C. Data were expressed as relative gene expression normalised to the Ct value obtained for the 32C sample. Primers used in qPCR and RT-PCR experiments are listed in supplementary material Table S2.

Pericarp slices were fixed for 1 hour in 4% formaldehyde in Tris buffered saline (TBS; 10 mM Tris pH 7.5, 154 mM NaCl) under vacuum infiltration at 20 kPa. Fixed tissues were then washed twice for 5 minutes each in TBS and 150 μm sections were obtained using a vibrating blade microtome. The sections were incubated in blocking solution (5% BSA in TBS) for 30 minutes at 37°C, and washed in TBS for 5 minutes. The polyclonal rabbit anti-RNA pol II S2P (phosphorylated on Ser2; Abcam 5095) or NP (non-phosphorylated; Abcam 817) primary antibodies were diluted 1:200 in TBS containing 0.1% Tween 20 and 5% BSA, and sections were incubated overnight at 4°C. Samples were washed three times for 10 minutes each in TBS and incubated overnight at 4°C with goat anti-rabbit IgG-Alexa 488 (Invitrogen) diluted 1:100 in TBS containing 0.1% Tween 20 and 5% BSA. After three washes in TBS for 5 minutes each, pericarp slices were mounted and counterstained as described above.

Pericarp fragments were fixed under partial vacuum (20 kPa) in 2.5% glutaraldehyde in 150 mM phosphate buffer pH 7.2 for 4-6 hours at 4°C. After three washes in phosphate buffer (successively 200 mM, 150 mM and 100 mM), the samples were post-fixed in 1% osmium tetroxide in 100 mM phosphate buffer for 3 hours at 4°C, rinsed with phosphate buffer (100 mM then 50 mM) and finally water. An additional fixation step was performed in 1% tannic acid (BDH) in water for 30 minutes at room temperature. Tissues were dehydrated at 4°C through an ethanol series and embedded in Epon according to Carde (Carde, 1987).

Ultrathin sections (50 nm) were stained with 7% uranyl acetate in 50% ethanol and optionally with lead citrate. Digital images were acquired on a CM10 (FEI) transmission electron microscope with an AMT side-port camera (Elexience).

All measurements and counting were with Image-Pro Plus 6.2 software (Media Cybernetics), except for nuclear volumes which were measured with ImageJ software (NIH).

On TEM images, the nuclear radius was calculated as the mean average of Feret min and Feret max, being respectively the width and length of the smallest rectangle enclosing the nucleus. The nuclear perimeter was measured as the length of the nuclear envelope, including that of enclosed invaginations. Counting of mitochondria on TEM images was performed by tagging the mitochondria close to the nucleus, i.e. within ~1 μm of the nuclear envelope (the mean size of mitochondria in this tissue being 1 μm × 0.5 μm).

For counting mitochondria (stained with Rhodamine-123), the z-series were treated by blind deconvolution using AutoQuant X software (Roper Scientific). On each z-plane the mitochondria located within the DAPI-stained area or within ~1 μm of this area were manually tagged using Image-Pro Plus, the tags being cumulated for the whole z-series. This method ensures that mitochondria that are visible on two successive planes due to poor resolution in the z-axis (z-step of 1 μm) are not counted twice. The nuclear volumes were estimated after segmentation of each plane of the DAPI z-series using the Object Counter 3D plug-in of ImageJ.

The consequences of endopolyploidisation for the spatial organisation of chromosomes in tomato fruit cells was investigated by applying our recently described FISH technique (Bourdon et al., 2011) using flow-sorted nuclei at different ploidy levels and a 45S rDNA probe. Whatever the ploidy level, only two hybridisation signals were detected at the 45S rDNA locus, which is located on chromosome 2 in tomato (Jo et al., 2009), and their relative size correlated with the extent of ploidy (Fig. 1A). The use of Chromomycin A3, a DNA-specific dye that preferentially stains GC-rich regions, such as the nucleolar organising regions (NORs) located on chromosome 2 in the tomato genome (Arumuganathan et al., 1994; Chang et al., 2008), confirmed the 45S rDNA hybridisation profile (Fig. 1B).

Both experiments thus showed that chromosome number is unchanged during endopolyploidisation in tomato fruit tissues, demonstrating the occurrence of polytenic chromosomes as expected for endoreduplication.

Taking advantage of the ‘ploidy map’ across the pericarp of mature green tomato (Bourdon et al., 2011), we analysed nuclear morphology in the three most contrasting pericarp areas in terms of ploidy level: the outer epidermis (OE), which displays very low ploidy levels of 2C and 4C; the inner epidermis (IE), which displays mid-ploidy levels ranging from 8C to 32C; and the mid- to inner mesocarp (M), which displays high ploidy levels of 32C to 128C.

Using DiOC₆(3), a lipophilic fluorescent dye that labels intracellular membranes (Terasaki and Reese, 1992), and DAPI, a DNA-specific dye, to stain fruit tissues (Fig. 2A,C,E), small round nuclei were observed in OE (Fig. 2A,B), elongated and slightly grooved nuclei were found in IE (Fig. 2C,D) and large nuclei with numerous invaginations were found in M (Fig. 2E,F). These observations show a clear relationship between the complexity of nuclear shape and the ploidy level. This could be quantified on TEM pictures taking the nuclear area (A), the perimeter (P) corresponding to the length of the nuclear envelope, and the roundness factor (R) according to the formula R=P² × (4πA)⁻¹ (Fig. 3A). This factor is equal to 1 for a circle and increases for objects diverging from this shape. High values for R were found in pericarp zones of high ploidy levels (M), whereas low values were measured in OE (low ploidy levels) and intermediate values in IE (intermediate ploidy levels). The differences between the three pericarp areas were significant (P<0.001, non-parametric Kruskal-Wallis test), which strongly suggests that the extent of the grooves increases according to ploidy.

The development of grooves in polyploid nuclei may impact the exchange ability between the nucleus and cytoplasm. Postulating that nuclear exchange sites (mainly nuclear pore complexes) display the same density all over the nuclear envelope whatever the ploidy level, we hypothesized that the nuclear exchange ability might be represented by a simple ratio between nuclear envelope area and nuclear volume, assuming that the higher this ratio, the higher the nuclear exchange ability. As we performed this analysis on 2D TEM acquisitions, the nuclear envelope area/nuclear volume ratio was estimated as the nuclear P/A ratio and plotted against the mean nuclear radius (r) obtained by averaging the minimum and maximum diameter of each nucleus (Fig. 3B). For a circle, the P/A ratio decreased with r according to an inverse function (y=2/r), thus indicating a lower exchange ability with increasing size. For all measured nuclei, the P/A ratio was higher than for a circle of the same radius. Higher values for P/A may originate from an ellipsoid nuclear shape (especially nuclei from IE) or from an invaginated nuclear shape (especially within M), or from a combination of both. Most nuclei displayed a P/A ratio between 1.6 and 0.6, which corresponds to P/A ratios for circles of 1.3 and 3.3 μm radius, respectively. Even for the biggest nuclei from M (radius >20 μm), the P/A ratio was not lower than that of a 3.3 μm radius circle, indicating that the nuclei could maintain their exchange ability during endopolyploidisation.

Numerous mitochondria were observed in the close vicinity of polyploid nuclei after DiOC₆(3) staining or in TEM images (Fig. 2E,F). The use of Rhodamine-123, a fluorescent cationic dye used to label mitochondria in living cells (Emaus et al., 1986), confirmed these observations within M (Fig. 4C) and to a lesser extent within IE (Fig. 4B). Only a few mitochondria were seen close to the nuclei within OE (Fig. 4A). Moreover, many mitochondria co-localised with nuclear grooves (Fig. 4C). We then determined the number of mitochondria in proximity to nuclei within the three pericarp areas defined according to the ploidy map (Bourdon et al., 2011). From the z-series of confocal images, the mean number of mitochondria surrounding a nucleus (within 1 μm of the nuclear envelope) was ~14.5 in OE, 294 in M and 62 in IE (Fig. 4D). A similar relationship was obtained by counting mitochondria close to the nuclear envelope on TEM images (Fig. 4E), although these values were lower, being derived from single-plane sections of nuclei. Therefore, the number of mitochondria associated with the nuclei increased according to ploidy.

When the above data were expressed as the ratio of mitochondria per nuclear volume or area (Fig. 4F,G), no significant differences appeared between the different pericarp areas (P>0.08, Kruskal-Wallis test). Thus, the number of mitochondria relative to nuclear volume or unit area apparently remains constant whatever the ploidy level.

The developmental process of cell growth requires sustained protein synthesis and thus sustained ribosome biogenesis (Rudra and Warner, 2004). Ribosomal RNAs are transcribed and assembled into ribosomal subunits at the nucleolus (Hernandez-Verdun, 2006). To assess the transcriptional activity associated with the nucleolus, we performed an RNA FISH experiment using 5.8S rRNA as a probe (Fig. 5), which intimately hybridised to the nucleolus (Fig. 5A). The increase in the hybridisation signal area positively correlated with the nuclear area (Fig. 5B), which itself is correlated to ploidy level (Bourdon et al., 2011), thus indicating enhanced transcription for 5.8S rRNA in endoreduplicated nuclei.

To visualise transcriptional activity inside endoreduplicated nuclei from fruit pericarp, in situ immunocytolocalisation experiments were performed to detect the active form of RNA polymerase II (RNA pol II) using an antibody against the phosphorylated heptapeptide YSPTSPS repeat (on Ser2) of the RNA pol II C-terminal domain (CTD) (S2P antibody) (Gurtner et al., 2008) (Fig. 6). The topology of active RNA pol II was similar within the nucleus (Fig. 6C,F,I) whatever its position in the pericarp (Fig. 6A), i.e. whatever its ploidy level [according to the ploidy map reported by Bourdon et al. (Bourdon et al., 2011)]. Indeed, active RNA pol II was preferentially associated with less dense chromatin regions as demonstrated by attenuated DAPI staining (e.g. euchromatin; Fig. 6B,E,H) and was excluded from the nuclear periphery where DAPI staining was intense (e.g. heterochromatin; Fig. 6D,G,J).

A nuclear suspension was then prepared from mature green pericarp and immunolabelled with the S2P antibody together with an antibody raised against the unphosphorylated (NP) inactive form of RNA pol II. The fluorescence intensities resulting from each immunolabelling were simultaneously quantified for each ploidy class by cytometry. The S2P and NP immunolabellings were both positively correlated to ploidy levels (Fig. 7A), showing that the protein level for RNA pol II increased proportionally with ploidy. Analysis of double-immunolabelled nuclei confirmed that the distribution of S2P immunolabelling appears preferentially associated with euchromatin, as does that of NP (Fig. 7B-E). The labelling of inactive RNA pol II (NP) was highly punctate, possibly at promoter sequences (Fig. 7C), whereas that of the active form (S2P) was homogeneous (Fig. 7D).

The cytometer was used to calculate the immunolabelling intensity ratio (S2P/NP) nucleus by nucleus. For each ploidy class, the coefficients of variation (CV) of this ratio are shown in Fig. 7A. In the 2C and 4C populations, this ratio had a high variance, whereas nuclei of 16C and above were exceptionally uniform, with the ratio having 0.2% CV. The relationship between the phosphorylated and non-phosphorylated states of RNA pol II must be weak in 2C and 4C nuclei, as if multiple nuclear states exist, but it is tight in endoreduplicated nuclei in favour of a uniform nuclear state (for reference, the CV of the DNA peaks typically remain unchanged throughout the ploidy scale).

To address gene expression during endoreduplication, we selected five single-copy protein-coding genes in tomato: RPB1, which encodes the large subunit of RNA pol II; the endocycle regulating genes CCS52A and WEE1 (Gonzalez et al., 2007; Mathieu-Rivet et al., 2010); and two mitosis-associated genes used as negative controls of gene expression in endoreduplicated nuclei, encoding CDKB2, a cyclin-dependent kinase specific to the G2-M transition (Chevalier, 2007), and the kinetochore MIS12 protein (Sato et al., 2005). qPCR on genomic DNA extracted from 4C, 16C and 32C sorted nuclei confirmed that the extent of all five gene duplications was linearly correlated with the degree of nuclear polyploidisation (supplementary material Fig. S1). Then, real-time qRT-PCR was performed on nuclear RNA extracted from an equal number (~30,000) of 4C, 16C and 32C nuclei to detect the selected five gene transcripts. For each gene, the quantity of transcripts detected in 32C nuclei was used as a reference (relative maximum of expression) for determining transcript levels in 4C and 16C nuclei. As shown in Fig. 8, the amount of RPB1 transcripts increased according to ploidy levels, in accordance with the observed increase in protein level (Fig. 7A). CCS52A and WEE1 displayed a similar behaviour for transcript accumulation, whereas CDKB2 and MIS12 transcripts were undetectable, as expected for mitosis-associated genes. Therefore, despite the correlation between DNA ploidy content and overall transcriptional activity noted above, there is still specificity at the gene transcription level to support physiological functions related to endoreduplication.

Recent advances in understanding of the molecular control and function of endoreduplication in higher plants points to its importance in various developmental processes as an essential part of growth (De Veylder et al., 2011). However, little is known about the related cytological aspects, such as the structure of endoreduplicated nuclei and chromosomes or the subcellular modifications associated to endoreduplication in a particular plant tissue and cellular context. As part of our contribution to deciphering the role of endoreduplication during plant development, we aimed here at addressing the influence of ploidy levels on nuclear architecture and on nuclear processes such as transcription.

As in most angiosperm species (Nagl, 1976; D'Amato, 1984; Joubès and Chevalier, 2000), endoreduplication per se is the main mode of cell endopolyploidisation in tomato fruit tissues (Fig. 1). The use of a highly tandemly repeated DNA sequence probe (2300 copies for 45S rDNA) (Chang et al., 2008) resulted in different hybridisation patterns in FISH assays when compared with non-repeated DNA sequences encoded by bacterial artificial chromosome (BAC) probes, i.e. the presence of two homogeneous large signals for 45S rDNA as compared with two ‘clouds’ of clustered hybridisation dots for BAC sequences (Bourdon et al., 2011). This difference can be explained by the fact that the 45S rDNA repeats encompass a much larger chromosomal sequence than any BAC probe (respectively, ~20 Mb versus 0.1 Mb on average) (Chang et al., 2008; Bourdon et al., 2011). In addition, untranscribed copies of 45S rDNA are often condensed and located just outside the nucleolus and may contribute to this large hybridisation signal (Heslop-Harrison and Schwarzacher, 2011). The use of a probe specific for the Arabidopsis 5S rDNA repeats resulted in hybridisation signals very similar to those for non-repeated DNA sequences (supplementary material Fig. S2) (Bourdon et al., 2011), being organised into two clusters displaying a number of hybridising dots (revealing the chromatids) that correlate with ploidy levels, thus confirming that endopolyploidisation in tomato produces polytenic chromosomes with multivalent sister chromatids. These data support the observation of Schubert et al. (Schubert et al., 2006) of a high degree of sister chromatid separation according to increased endopolyploid levels in Arabidopsis nuclei.

The amplification of genomic DNA within endoreduplicated chromosomes is likely to be strictly conservative, as suggested from flow cytometry histograms of pericarp nuclei (Bourdon et al., 2011). So far, all the single-copy genes or chromosomal regions used as probes in qPCR and FISH experiments, such as 5S rDNA, 45S rDNA, RPB1, CCS52A, WEE1, LE_HBa0048M11 and LE_HBa0037G17 BAC clones (Bourdon et al., 2011), which are respectively located on chromosomes 1, 2, 8, 9, 7 and 12 (Tomato Genome Consortium, 2012; http://solgenomics.net/organism/Solanum_lycopersicum/genome), were found amplified exponentially according to ploidy (Fig. 1; supplementary material Figs S1, S2) (Bourdon et al., 2011). This appears to be a general rule for higher plant endoreduplication, with a notable exception being the developmentally regulated partial endoreduplication observed in orchids (Lepers-Andrzejewski et al., 2011).

Our previous analysis of the ploidy distribution across tomato fruit pericarp showed no evidence for any positive correlation between endoreduplication and cell fate, as a definite ploidy level cannot be assigned to a precise cell size or cell fate (Bourdon et al., 2011). Although the positive correlation between endoreduplication and cell size has been described in wild-type tomato fruits (Cheniclet et al., 2005; Nafati et al., 2011), we have shown that there is a broad variety of cell types from IE to OE according to a complex differentiation pattern during fruit development, including a large discrepancy in cell sizes with partially overlapping ploidy levels (Bourdon et al., 2011). These data and additional functional studies led us to propose that endoreduplication serves as a trigger for organ growth and provides the potential to support a range of cell sizes rather than any precise, definitive size (Mathieu-Rivet et al., 2010; Nafati et al., 2011).

The nuclear morphology of endoreduplicated cells in tomato pericarp is strongly affected by endopolyploidisation, with highly invaginated and grooved nuclear envelopes (Fig. 2). Such characteristics have already been reported for plant and animal nuclei (Fricker et al., 1997; Collings et al., 2000). Interestingly, the endoreduplicated and invaginated nuclei can maintain their exchange ability during endopolyploidisation (Fig. 3B). The exchange ability is related to the distance to be travelled by molecules from their site of synthesis within the nucleus to their export site on the envelope. For a circle the maximum distance between any point within the circle and the periphery is equal to its radius. The fact that nuclei larger than 20 μm in radius display an exchange ability (estimated by the P/A ratio) similar to that of nuclei (or circles) of 3.3 μm radius suggests that their complex shape minimises the distance between the inner volume and the envelope of the nucleus. In addition, the pronounced invaginations of endoreduplicated nuclei enable the maintenance of efficient nucleus-to-cytoplasm communication, despite the increase in nuclear volume. Indeed, numerous mitochondria were found to be localised in the grooves of the nuclear envelope (Fig. 4), i.e. at the site of cytoplasm-to-nucleus exchange. Such a direct proportionality between nuclear DNA content and the number of organelles – at least for plastids – has been reported (D'Amato, 1998). In addition, labelling with JC-1 dye to examine mitochondrial membrane potential revealed that mitochondria were largely active whatever the pericarp area considered (supplementary material Fig. S3). Therefore, the presence of a large population of functional mitochondria close to the envelope of endoreduplicated nuclei could be explained by an absolute need for ATP synthesis as an energy supply for nuclear biochemical processes. Thus, our observation at the cellular level supports the commonly inferred, but not experimentally demonstrated, association of endopolyploidy with metabolically highly active cells (Larkins et al., 2001; Sugimoto-Shirasu and Roberts, 2003).

The highly polyploid endosperm cells of maize kernels, which accumulate large amounts of starch and storage proteins (Sabelli and Larkins, 2009), and the nitrogen-fixing symbiotic root nodule cells in legumes (Cebolla et al., 1999) are quoted as metabolically specialised and highly endoreduplicated cells. Interestingly, an important remodelling of the transcriptome is tightly linked to endoreduplication in nitrogen-fixing nodules (Mergaert et al., 2003). As far as tomato fruit is concerned, only descriptive analyses for transcriptional and metabolic profiling in endoreduplicated tissues have been provided (Lemaire-Chamley et al., 2005; Mounet et al., 2009; Schauer et al., 2006; Steinhauser et al., 2010), without any direct link with the extent of endoreduplication.

Using RNA FISH we demonstrated that the transcription of 5.8S rDNA within the nucleolus is enhanced during endoreduplication (Fig. 5). In addition, the 5.8S rRNA hybridisation signal that reflects the size of the nucleolus was positively correlated with nuclear size and consequently with ploidy levels. These data are consistent with our observation of NOR enlargement according to ploidy (Fig. 1B). Such a nucleolar enlargement has been reported in rodent trophoblasts (Zybina and Zybina, 1996) and in Drosophila (Pierce et al., 2004) and associated with increased transcriptional and translational activities in the context of endopolyploidisation. This feature has also been associated with the rate of proliferation in cancer cells (Derenzini et al., 2000). It is therefore tempting to draw a parallel with the increased rate of growth that is often associated with endoreduplication. We found that the cytoplasmic area of cells within the three most contrasting pericarp zones in terms of ploidy levels (OE, IE and M, where the most common nuclear classes are 2C, 16C and 64C, respectively) indeed increases with ploidy levels (supplementary material Fig. S4A), as does the estimated cytoplasmic volume (supplementary material Fig. S4B). This demonstrates the maintenance of the nuclear to cytoplasmic ratio to drive cell growth, consistent with the karyoplasmic ratio theory. Ribosomal RNA, protein synthesis and exchange must be proportionally enhanced to increase ribosome biogenesis and hence maintain homeostasis of cytoplasmic components essential for sustained cell growth.

Previous functional analyses of genes regulating the endocycle (Gonzalez et al., 2007; Mathieu-Rivet et al., 2010; Nafati et al., 2011) demonstrated the biological significance of endoreduplication as a morphogenetic factor for cell and fruit growth in tomato. In addition, enhancing endoreduplication and consequently fruit growth also strongly impacted nitrogen metabolism in developing fruits (Mathieu-Rivet et al., 2010), establishing a link between endoreduplication-induced growth and primary metabolism, as reported in maize endosperm (Leiva-Neto et al., 2004). Interestingly, a causal relationship between cell size, polyploidy and transcription in yeast has been demonstrated, suggesting that cells adjust their transcriptional capacities to their geometric properties, including cell size (Wu et al., 2010).

Here, we showed that transcription of the WEE1 and CCS52A genes that control endoreduplication (Gonzalez et al., 2007; Mathieu-Rivet et al., 2010) is increased on a per-nucleus basis according to ploidy (Fig. 8). This observation was also true for RPB1, which encodes a subunit of RNA pol II. The use of polyclonal antibodies targeting the YSPTSPS repeat of the RNA pol II CTD allowed us to discriminate between the active and inactive status of RNA pol II. The RNA pol II CTD repeat is differentially phosphorylated according to its position on a gene: when located on a promoter sequence, it is largely unphosphorylated (NP form), whereas the active form of RNA pol II requires phosphorylation at Ser2 and Ser5 (Hirose and Ohkuma, 2007). In addition, phosphorylation at Ser5 is mainly associated with the initiation of transcription (Hengartner et al., 1998; Komarnitsky et al., 2000), whereas phosphorylation at Ser2 (S2P form) is predominantly present during the elongation step of transcription (Ni et al., 2004). Endoreduplication is linked to the sustained transcription/translation and activity of RPB1 (Figs 6, 7). Furthermore, the results obtained with the S2P antibody suggest that transcriptional elongation is excluded from the perinuclear heterochromatic region and occurs within euchromatin in a homogeneous manner (Fig. 6). Subnuclear compartments were not evident, suggesting that the euchromatin regions in polyploid nuclei are apparently indiscriminately transcribed, as if all copies of amplified genes are equally functional in the tomato endoreduplicated genome. Moreover, the exceptionally tight CV observed for the S2P/NP ratio in polyploid nuclei (Fig. 7) reveals a uniformity of transcriptional activity when compared with that observed in 2C and 4C nuclei. Altogether, these results demonstrate a strong relationship between endoreduplication and transcription.

The role of endoreduplication in organ growth is still a matter of debate. An organismal regulation of growth in which organ growth itself acts as a ‘hub’ driving cell division and expansion, as well as endoreduplication, with compensation if one of the underlying cellular processes is limiting (John and Qi, 2008; Massonnet et al., 2011), is opposed to a cell-based (autonomous) mode of growth control that relies on the balance of cellular processes (division, expansion and endoreduplication) (Sugimoto-Shirasu and Roberts, 2003).

The results presented here provide quantitative evidence in favour of the karyoplasmic ratio theory. Different levels of specific structural and functional organisation in endoreduplicated cells were revealed, shedding light on the contribution of endoreduplication to the establishment of a highly structured cellular system that integrates multiple physiological functions, including transcription, and acts as a morphogenetic factor supporting cell growth during tomato fruit development. Our perspective is that karyoplasmic homeostasis ensures function not only by geometry, but also by cooperative adaptation.

Image acquisition and analysis were performed at the Plant Imaging Pole of the Bordeaux Imaging Center. Flow cytometry was performed at the Imagif Platform (IFR87 and ASTRE, CNRS, Gif-sur-Yvette) with Marie-Noëlle Soler, whom we thank for help with imaging. We thank Norbert Bollier for contribution to the results obtained on mitochondria.