Analysis of the Tomato Fruit Growth Response to Temperature and Plant Fruit Load in Relation to Cell Division, Cell Expansion and DNA Endoreduplication

Abstract

• Background and Aims To better understand the regulation of fruit growth in response to environmental factors, the effects of temperature and plant fruit load on cell number, cell size and DNA endoreduplication were analysed.

• Methods Plants were grown at 20/20 °C, 25/25 °C and 25/20 °C day/night temperatures, and inflorescences were pruned to two (‘2F’) or five (‘5F’) flowers.

• Key Results and Conclusions Despite a lower fruit growth rate at 20/20 °C, temperature did not affect final fruit size because of the compensation between cell number and size. The higher cell number at 20/20 °C (9·0 × 10⁶ against 7·9 × 10⁶ at 25/25 °C and 7·7 × 10⁶ at 25/20 °C) resulted from an extended period of cell division, and the smaller cell size was due to a shorter period of expansion rather than a lower expansion rate. By contrast, the lower fruit growth rate and size of 5F fruits compared with 2F fruits resulted from the slow down of cell expansion, whereas the number of cells was hardly affected in the proximal fruit. However, within the inflorescence the decreasing gradient of fruit size from proximal to distal fruits was due to a decrease in cell number with similar cell size. Fruit size variations within each treatment were always positively correlated to variations in cell number, but not in cell size. Negative correlations between cell size and cell number suggested that cells of tomato pericarp can be seen as a population of competing sinks. Mean ploidy was slightly delayed and reduced in 5F fruits compared with 2F fruits. It was highest at 25/25 °C and lowest at 25/20 °C. Treatments did not affect ploidy and cell size in similar ways, but within each treatment, positive correlations existed between mean ploidy and cell size, though significant only in the 2F-25/20 treatment.

INTRODUCTION

Much attention has been paid to the environmental influence on fruit growth in greenhouse tomato crops, and optimum temperature and light regime have been defined for fruit production (Pearce et al., 1993; Adams et al., 2001; Adams and Valdés, 2002). Since different processes are successively involved in the control of growth during fruit ageing, the sensitivity to environmental variations is expected to fluctuate during fruit development, as shown for temperature (De Koning, 1994; Adams et al., 2001). Moreover, compensation between the numerous components of growth may lead to underestimation of the fruit response. For instance, an increase of temperature effectively increases the maximum tomato growth rate, but it is compensated for by a shorter period of growth, so that fruit weight may not be significantly affected (Ho, 1996; Adams et al., 2001). To deepen our understanding and control of fruit growth in response to environmental fluctuations, it is necessary to describe better the individual processes involved in this response during fruit development.

Tomato is a fleshy fruit composed of different tissues: the epidermis, the pericarp (flesh) and the placenta, and locular tissue including seeds (pulp). Increase in fruit volume results from biophysical limitation by epidermal extensibility (Thompson, 2001) and from the development of pericarp tissue which generally accounts for more than two-thirds of the total fruit weight (Ho and Hewitt, 1986). Both division and expansion activity in pericarp tissue are determinant for tomato growth. Whereas epidermal cells divide throughout fruit development, cell division in the pericarp is limited to a short period of fruit development and is located in the external tissue around the vascular bundles and in the hypodermis. Once cell division ends, cell expansion becomes the dominant way to increase fruit size. In tomato, large endoreduplicated cells are located in the mesocarp (Bünger-Kibler and Bangerth, 1983).

Endoreduplication is an incomplete cell cycle that leads to the increase of nuclear DNA content (D'Amato, 1964; Galbraith et al., 1991), which in fruit pericarp reaches levels up to 256C (C is the haploid nuclei DNA content) in cherry tomatoes as well as in large-size fruit cultivars (Bergervoet et al., 1996; Joubès et al., 1999). Endoreduplication may be involved in the control of fruit growth since it was suggested to set the size limit of a cell (Traas et al., 1998). In Arabidopsis a positive relationship between endoreduplication and cell size was reported in epidermis cells (Melaragno et al., 1993). Among pea seed genotypes, a linear relationship was reported between endoreduplication in cotyledon cells and seed dry weight or mean cell volume (Lemontey et al., 2000). In tomato, there is little experimental evidence of any direct relationship between endoreduplication and fruit size (Bünger-Kibler and Bangerth, 1983; Bertin et al., 2003). However, it cannot be ruled out that endoreduplication is involved in the control of cell growth in response to environmental variations, since it has been rarely investigated.

Studies of fruit growth in response to environmental variations have been mainly focused on fruit expansion processes (Ehret and Ho, 1986; Pearce et al., 1993; Adams et al., 2001), though the very precocious control of final fruit size by cell division prior to anthesis has been well documented at the fruit level (Bohner and Bangerth, 1988; Ho, 1996) and also at the gene level (Frary et al., 2000). A decrease of the source : sink ratio on the plant diminishes final fruit size by reducing both cell number and cell size (Bohner and Bangerth, 1988; Bertin et al., 2003). Many studies have described the effects of temperature on cell cycle duration, cell division rate and cell expansion rate in root or shoot meristems and in leaves (Brown and Rickless, 1949; Lopez-Saez et al., 1966; Francis and Barlow, 1998; Granier et al., 2000; Tardieu and Granier, 2000), but more rarely in fruit, though temperature is the primary climatic factor affecting tomato fruit growth rate (Walker and Ho, 1977; Pearce et al., 1993; Peet et al., 1997; Willits and Peet, 1998).

In this study, the response of tomato fruit growth to temperature and plant fruit load was analysed, in relation to the cell number, cell size and nuclei DNA endoreduplication in fruit pericarp.

MATERIALS AND METHODS

Plant material and cultural conditions

The experiment was carried out in an 8·75-m² (21-m³) growth climatic chamber under controlled conditions. Seeds of tomato ‘Raïssa’ were sown in sand, and 12 homogenous plants were pricked out at a developmental stage of about four or five visible leaves, in 10-dm³ pots filled with a balanced oxygenated nutrient solution, whose composition was checked every week and readjusted when necessary. Sowing took place in the growth climatic chamber itself under climatic conditions similar to those monitored after planting. Metal halide lamps were used to provide artificial lighting. A 12-h photoperiod was applied with a photon flux of about 500 µmol m⁻² s⁻¹ PAR above the canopy. Air relative humidity was maintained around 70 %. From anthesis of the first truss, air was enriched to 800 µl CO₂ l⁻¹ during the light period. Flowers were pollinated as they opened with an electrical shaker, and all side shoots were removed as they appeared.

Experimental treatments

Three successive experiments were conducted under the same controlled conditions except the day/night air temperature regime which was successively set to 20/20 °C, 25/25 °C and 25/20 °C ± 0·5 °C. Temperature of the nutritive solution was monitored to 22 °C in all experiments. In each experiment, inflorescences were pruned to five flowers (‘5F’ treatments) on six plants and to two flowers on the other six (‘2F’ treatments). Pruning took place at 50 % anthesis of the truss. Plants were topped two leaves above the tenth truss.

Observations and measurements

The developmental stage of each individual flower bud was recorded twice a week from appearance to fruit set (about 5 mm diameter). Anthesis time was considered as full flower opening. As fruits developed, the equatorial diameter of the first and second fruits (F1 and F2) on plants pruned to two flowers per truss, and the diameter of the first, third and fifth fruits (F1, F3 and F5) on plants pruned to five flowers, were recorded once a week with a caliper square.

Fruits F1 and F2 in the 2F treatments, and F1, F3 and F5 in the 5F treatments were sampled as the first truss ripened. The pericarp cell number was measured on all these fruits (except F2 in the 2F-25/20 °C treatment) on the first seven trusses. At this time the cell division activity had been completed in the first four trusses. The ploidy level of pericarp cells was measured on ovaries and fruits of different ages sampled on the first eight trusses at the first (F1) and fifth (F5) positions. The mean cell volume was estimated by dividing the pericarp volume (measured by water displacement) by the total number of pericarp cells. Previous measurements of cell area made in situ on pericarp slices (using the method described in Bertin et al., 2003) confirmed that this ratio is a good indicator of cell size, considering that the total intercellular space of tomato pericarp is relatively reduced (N. Bertin, unpubl. res.), and likely not to be affected by treatments.

The number of pericarp cells was measured after tissue dissociation according to a method adapted from that of Bünger-Kibler and Bangerth (1983). Details of the applied method are given in Bertin et al. (2002). The ploidy level was measured by means of a PARTEC flow cytometer (PARTEC Ploidy Analyser PA, GmbH, Germany), equipped with an HBO lamp for UV excitation (Bertin et al., 2003). Three replicate measurements were made in mature fruit, but only one measurement in flower buds and young ovaries because of the small quantity of material. Mean endoreduplication level was calculated as:

where n is the number of peaks of DNA content (max = 8) in the sample, C_i is the C value in the nuclei of peak n_i (C₁ = 2, C₂ = 4 ⋯ C₈ = 256), N_i is the number of nuclei in the peak n_i and N_tot is the number of nuclei in all peaks of the sample.

Statistical analysis

The effects of temperature or plant fruit load on cell number were analysed in interaction with the truss position (first four trusses) by two-way ANOVA (Jandel Scientific Sigmastat), and F-tests were used to determine the statistical significance. When significant effects were detected, a Tukey test was applied for all pairwise comparisons of mean responses.

Fruit growth and mean ploidy curves were fitted to three-parameter Gompertz and sigmoid functions, respectively. Parameter estimation was carried out using least squares method (Jandel Scientific SigmaPlot). The difference between two (or more) treatments was tested by comparing the sum of the residual sums of squares for the two (or more) individual fittings (ΣSS_i) with the residual sum of squares for the common fitting to pooled treatments (SS_c) considering that the statistic:

follows Fisher's law with (n − 1)k and (N_data − k) degrees of freedom. N_data is the total number of data, n is the number of individual regression and k is the number of fitted parameters for each regression (three for Gompertz and sigmoid functions).

Correlations between fruit weight, cell number, cell size and mean ploidy, were analysed on fruits older than 30 d after anthesis (daa), by running the Pearson product Moment test (Jandel Scientific Sigmastat). Correlation coefficients are reported in the legends to Figs 2, 4 and 6 and statistical significance of these correlations are given in the text.

Fig. 2.

Correlations between (A) cell size and cell number, (B) fruit fresh weight and cell number and (C) fruit fresh weight and cell size. Lines represent linear adjustments. Each point is an individual fruit older than 30 daa sampled at the first or second position (F1 and F2) in the first four trusses of plants grown at 20/20 °C [open circles and dotted line R = −0·17 (A), 0·73 (B), 0·53 (C)], 25/25 °C [grey circles and dashed line R = −0·66 (A), 0·87 (B), −0·25 (C)] or 25/20 °C [black circles and continuous line R = −0·60 (A), 0·86 (B), −0·12 (C)] day/night temperature.

Fig. 4.

Correlations between (A) cell size and cell number, (B) fruit fresh weight and cell number and (C) fruit fresh weight and cell size. Lines represent linear adjustments. Each point is an individual fruit older than 30 daa sampled at the first position (F1) in the first four trusses of 2F [circles and full line R = −0·60 (A), 0·86 (B), −0·12 (C)] and 5F [triangles and dotted line R = −0·56 (A), 0·87 (B), −0·12 (C)] plants grown at 25/20 °C day/night temperature regime.

Fig. 6.

Correlations between mean ploidy of pericarp cells and fruit fresh weight (A and B), cell volume (C and D) and cell number (E and F). Lines represent linear adjustments. Each point is an individual fruit older than 30 daa sampled on the first four trusses at the F1 or F2 positions for the 2F treatments (circles) and at the F1 [open triangles and dotted line R = −0·29 (A), 0·22 (C), −0·29 (E)] or F5 [black triangles and dashed line R = 0·64 (A), 0·69 (C), −0·59 (E)] positions for the 5F treatment. Plants were grown at 20/20 °C [open circles and dotted line R = −0·25 (B), 0·007 (D), −0·33 (F)], 25/25 °C [grey circles and dashed line R = −0·58 (B), 0·54 (D), −0·72 (F)] or 25/20 °C [black circles and full line R = −0·11 (B), 0·66 (D), −0·43 (F) and triangles] day/night temperature.

RESULTS

Maximum plant fruit load was achieved on the 2F treatments at each of the three temperatures. On the contrary for 5F treatments, numerous fruit abortions occurred in the 25/25 °C treatment and to a lesser extent in the 20/20 °C treatment, where four to five fruits actually set on each truss. To avoid confusion due to different source–sink balances among treatments, the fruit load effect was analysed only on the 25/20 °C treatment, whereas the temperature effect was analysed only among the 2F treatments.

Effect of temperature on fruit growth in relation to cell number and cell size when carbon supply is non-limiting

At 20/20 °C, fruit growth (F1 + F2) was significantly decreased (P = 0·01) compared with 25/25 °C and 25/20 °C which were similar (Fig. 1A). Differences in the growth pattern did not consist of a large difference in final fruit size, but rather in the duration of growth and in the maximum growth rate.

Fig. 1.

(A) Fruit size and fruit growth rate measured on 2F plants grown at 20/20 °C (dashed line), 25/25 °C (dotted line) or 25/20 °C (continuous line) day/night temperature. Three-parameter Gompertz functions were fitted on fruit size measurements made on the first and second fruits (F1 and F2) of the first four trusses. Adjustment was made on pooled data from six plants (R² > 0·95) and vertical bars indicate the standard error calculated on adjustments made on individual plants. Daily fruit growth rates were obtained by derivative functions. (B) Change in cell volume during fruit ageing, estimated by dividing the pericarp volume by the total number of pericarp cells in 2F treatments. Each point is an individual fruit sampled at the first or second positions (F1 and F2) on the first eight trusses of four plants grown at 20/20 °C (open circles), 25/25 °C (grey circles) or 25/20 °C (black circles) day/night temperature.

The cell volume estimates on fruits of different ages and the total number of pericarp cells in the first four trusses of 2F plants are shown in Fig. 1B and Table 1 for the three temperature regimes. The onset of cell expansion was delayed by about 5 d at 20/20 °C compared with 25/25 °C and 25/20 °C (Fig. 1B), and similar rates of increase in cell volume (slopes of the linear part of the curves) at all temperatures led to lower final cell size at 20/20 °C. Since final fruit sizes were similar at the three temperature regimes (Fig. 1A), the lower cell size at 20/20 °C was compensated for by a higher number of cells. Indeed in the first four trusses the number of pericarp cells was higher at 20/20 °C (P = 0·09) without any significant interaction with the truss position (Table 1). On average on the first four trusses, the number of cells was 7·9 × 10⁶ and 7·7 × 10⁶ at 25/25 °C and 25/20 °C, respectively, against 9·0 × 10⁶ at 20/20 °C.

Table 1.

Number of pericarp cells (×10⁶) measured in the first (F1), and third and fifth fruits (5F plants) of the first four trusses and mean cell number in these four trusses

	Truss 1	Truss 2	Truss 3	Truss 4	Mean of four trusses
2F-20/20 °C F1	9·65 ± 2·79	9·74 ± 1·30	9·53 ± 2·79	7·09 ± 1·58	9·00 ± 1·25
2F-25/25 °C F1	6·91 ± 1·85	7·73 ± 1·98	8·87 ± 0·52	7·91 ± 1·25	7·86 ± 0·79
2F-25/20 °C F1	6·43 ± 2·08	8·65 ± 1·24	7·78 ± 1·16	7·94 ± 0·61	7·70 ± 0·91
5F-25/20 °C F1	5·28 ± 1·34	10·05 ± 1·40	9·02 ± 1·75	9·75 ± 1·49	8·52 ± 2·49
5F-25/20 °C F3	6·18 ± 2·84	5·15 ± 1·78	5·34 ± 2·14	5·96 ± 2·47	5·66 ± 0·50
5F-25/20 °C F5	4·69 ± 2·03	3·45 ± 1·26	5·25 ± 0·94	5·02 ± 0·15	4·60 ± 0·91

Data are means of four plants ± 95 % confidence intervals.

To analyse the correlation between fruit size, cell number and cell size, fruits older than 30 daa, which were close to their final size were examined. A negative correlation between cell size and cell number confirmed the compensation between cell size and cell number among temperature treatments (Fig. 2A). The negative correlation also held within the 25/25 °C (P < 0·01) and 25/20 °C (P = 0·053) treatments, indicating that this compensation was not a specific response to temperature. Similarly the variations of fruit weight within each temperature treatment were positively correlated with the variations in cell number (P < 0·001 at the three temperature regimes), but not with those in cell size (Fig. 2B and C), except at 20/20 °C (P = 0·02) where the correlation was significant only due to one extreme point

Effects of plant fruit load on fruit growth in relation to cell number and cell size

Plant fruit load effects were analysed on F1 in the 25/20 °C treatment. Growth of F1 globally measured on the first four trusses was significantly (P < 0·05) reduced in the 5F treatment compared with the 2F treatment, due to the lower growth rate (Fig. 3A). In parallel a clear difference was observed between the two treatments in the increase of cell volume during fruit ageing (Fig. 3B). In contrast to the temperature effect (Fig. 1B), the lower cell volume in 5F fruits was more likely due to the decline of the cell expansion rate from about 25 daa.

Fig. 3.

(A) Fruit size and fruit growth rate of the first fruit (F1) of 2F (continuous line) and 5F (broken line) plants grown at 25/20 °C. Three-parameter Gompertz functions were fitted on fruit size measurements made on the first four trusses. Adjustment was made on pooled data from six plants (R² > 0·85) and vertical bars indicate the standard error calculated on adjustments made on individual plants. Daily fruit growth rates were obtained by derivative functions. (B) Evolution of cell volume during fruit ageing. Each point is an individual fruit sampled at the first position (F1) on the first eight trusses of four 2F (circles) and 5F (triangles) plants grown at 25/20 °C.

The number of cells in the pericarp of F1 was not significantly different between the 2F and 5F treatments (Table 1), but it significantly varied among trusses (P < 0·01), the first truss containing fewer cells than the following ones (significant only in the 5F treatment). On average on the first four trusses the number of pericarp cells was similar in the two treatments (8·5 × 10⁶ and 7·7 × 10⁶ cells for 5F and 2F, respectively).

Analysis of fruits older than 30 daa showed that the negative correlation between cell volume and cell number was still present within each treatment (P = 0·053 and 0·059 for 2F and 5F, respectively), but no compensation occurred across the treatments (Fig. 4A), as was observed for temperature (Fig. 2A). Similarly a significant positive correlation (P < 0·001) between fruit weight and cell number existed within each treatment (Fig. 4B), but no significant correlation between fruit weight and cell volume was observed, either within each treatment or across the treatments (Fig. 4C).

Whereas the number of cells in the pericarp of F1 was not significantly affected by the plant fruit load, the numbers of cells in the pericarp of F3 and F5 were significantly reduced (Table 1). A two-way ANOVA was performed to assess the truss and fruit effects on cell number of the first four trusses in the 5F-25/20 °C treatment. Except in the first truss a significant gradient of cell numbers existed, F1 containing more cells than other fruits of the truss. Differences among trusses were not significant and on the first four trusses the mean numbers of cells were 8·5 × 10⁶, 5·7 × 10⁶ and 4·6 × 10⁶ in F1, F3 and F5, respectively. Compared with F1, the final fruit size was reduced by 12 % and 19 % in F3 and F5, respectively, and the fruit growth curves significantly (P < 0·05) decreased from F1 to F5, due to a decreasing gradient in fruit growth rate. This could not be related to differences in estimated cell size which did not vary clearly among fruits of the same inflorescence.

Considering fruits older than 30 daa, a positive correlation existed between fruit weight and pericarp cell number, but the compensation relationship between cell number and cell size which was observed in F1 (Fig. 4A), did not hold for the third and fifth fruits of the truss.

Effect of temperature and plant fruit load on the ploidy level of pericarp cells and relationship with other fruit traits

The dynamic of mean ploidy during fruit ageing measured on the first eight trusses is presented in Fig. 5 for fruit load and temperature treatments. Ploidy of fruit pericarp was similar in F1 and F5 fruits of the 5F treatment (not shown) and these data were pooled. Mean ploidy was initially delayed in the 5F-25/20 °C treatment compared with the 2F-25/20 °C treatment, but final values were close to one another. Fitted curves were significantly different at the 5 % error level, but not at 1 %.

Fig. 5.

Mean C-value of pericarp cells measured in F1 and F2 for 2F treatments (circles) and in fruits F1 and F5 for 5F treatment (triangles). Each point is the mean of three measurements performed on an individual fruit sampled on the first eight trusses of plants grown at 20/20 °C (open circles), 25/25 °C (grey circles) or 25/20 °C (black circles and triangles) day/night temperature. Lines represent three-parameter sigmoid functions fitted on data from each treatment (R² > 0·84). On the right axis, are indicated the corresponding numbers of incomplete cell cycles.

Temperature significantly changed the dynamic of pericarp cell ploidy (P < 1 %). Mean ploidy was maximum at 25/25 °C, but surprisingly it was higher at 20/20 °C than at 25/20 °C. Other measurements made in the 5F treatmentsat 20/20 °C and 25/25 °C agreed with these patterns (not shown). Differences in DNA amount between 25/20 °C and 25/25 °C corresponded to less than one incomplete cycle.

Nuclear DNA content was measured on some of the fruits used for cell counting, so that correlative analysis between ploidy and other fruit traits could be performed on fruits older than 30 daa which had almost reached their final size (Fig. 6). In none of the treatments, was fruit fresh weight significantly correlated with the mean ploidy of pericarp cells and through the temperature treatments a negative correlation was even observed, though not significant (Fig. 6A and B). Within each treatment, positive correlations between cell size and mean ploidy (significant only in the 2F-25/20 °C treatment P = 0·04; Fig. 6C and D) and negative correlations between cell number and mean ploidy (significant only in the 2F-25/25 °C treatment P = 0·013; Fig. 6E and F) could be noted. These correlations held across treatments only comparing the 20/20 °C and 25/25 °C treatments, which was the only case where cell size and cell number really compensated for fruit weight. In that case only, the increase in cell size at 25/25 °C could be related to an increase in ploidy. The low ploidy level measured at 25/20 °C for both 2F and 5F treatments could not be related to any fruit traits. At this temperature the decrease in cell size in the 5F treatment was not associated with low ploidy levels.

DISCUSSION

In accordance with the literature (Pearce et al., 1993; Adams et al., 2001) the maximum fruit growth rate was achieved at 25/25 °C (Fig. 1A). Actual fruit temperature was likely to be higher than 25 °C, since air temperature was monitored instead of fruit temperature (Adams and Valdès, 2002). High temperature may affect fruit growth through indirect effects on plant development, maintenance respiration and assimilate availability. These indirect effects were avoided here by considering only the 2F plants in the analysis of the response to temperature.

In the range of temperature investigated in this study, many authors reported no or negative effects of increasing temperature on final tomato size and attributed this to compensating effects on the rates of fruit growth and fruit development (De Koning, 1994; Ho, 1996; Adams et al., 2001). The present investigation further indicated that this compensation is also due to inverse and compensating effects of temperature on cell number and cell size (Fig. 2A). Between 20 and 25 °C, an increase of temperature promoted cell size, but slightly decreased the number of cells, so that final fruit size was hardly affected (Fig. 1A). The increase in cell number at 20/20 °C resulted from the extended period of cell division (Fig. 1B). Interestingly the different patterns of fruit growth at 20/20 °C (slow) and 25/25 °C (accelerated) resulted from differences in the duration of the cell division period and onset of cell expansion which mainly shortened the period of cell expansion at 20/20 °C, but did not strongly reduce the rate of cell expansion. This is in accordance with Adams et al. (2001) who hastened flower opening by heating flower buds at 25 °C. The shortening of the cell division period together with a reduction in the final number of cells at 25/25 °C, suggested that the proportion of cycling cells in the pericarp and its evolution during the division period were not affected by temperature between 20 and 25 °C. Indeed, considering that the time required for a cell to divide decreases in response to an increase of temperature with a minimum duration around 30 °C for many species (Francis and Barlow, 1988), a compensation between the cell cycle duration and the proportion of cycling cells did not occur, since fruits finally contained less cells at 25/25 °C than at 20/20 °C.

Fruits produced at 25/20 °C were close to those obtained at 25/25 °C with respect to fruit growth, and cell number and size. Tomato plants are known to integrate day/night temperatures in terms of fruit yield (Hurd and Graves, 1984; Peet et al., 1997) and thus the 25/20 °C treatment can be considered as a constant 22·5 °C temperature regime. However, cell number and cell size were similar in the 25/25 °C and 25/20 °C treatments, which is inconsistent with the absence of compensation between cell cycle duration and proportion of cycling cells, except if the cell cycle duration is already minimum at 22 °C in the tomato pericarp.

In contrast to what was observed in leaves of sunflower, tobacco and pea (Granier et al., 2000) cell division and tissue expansion did not have a common response to temperature in tomato, so that final fruit size and cell number were not correlated throughout the temperature treatments (Fig. 2B). But within each treatment, the variations in cell number were always tightly and positively correlated with the variations in fruit size, whereas the variations in cell size were not. Thus considering a given level of carbon and water supply for fruit growth, cell division plays a dominant role in the determination of intra-treatment variations of fruit size, which largely exceeded the inter-treatment variations.

Reduction of plant fruit load (2F against 5F) promoted the fruit growth rate and final fruit size of F1, by increasing cell expansion without any inverse compensating effects on cell number. However, the absence of effects of fruit load on cell number may be due to the fact that the comparison of 2F and 5F treatments concerned only F1. Indeed F3 and F5 contained far fewer cells than F1 (Table 1) and are much more sensitive to the competition for assimilates than F1 (Bangerth and Ho, 1984; Bohner and Bangerth, 1988; Bertin et al., 2003). For instance, plant defoliation induces a decrease in both cell number and cell size more important in distal than in proximal fruits (Bohner and Bangerth, 1988). As observed by these authors on control trusses with six fruits, the cell size did not vary among fruits within the same truss in the 5F treatment, so that the gradient in fruit size could be totally attributed to the gradient in cell number. This indicated that the competition for assimilates in a truss with six fruits, as currently conducted in a large-size tomato, is not high enough to affect cell size in distal fruits. It is suggested that the numerous cells in F1 were in competition for assimilates, whereas in F3 and F5 the low number of cells could be sufficiently supplied to grow as much as the numerous cells in F1. In that case, a negative correlation between cell number and cell size could be expected, which was true for F1 only. In F3 and F5, cells of the pericarp were likely not to be in competition because of their low number. It is interesting to note that for F1 and F2, the negative correlation between cell size and cell number held within each treatment (less at 20/20 °C). Thus cells of the same tissue can be seen as a population of sinks in competition for the supplied assimilates. Any treatment that affects the number of cells without limiting the global supply of assimilates, is expected to have inverse effects on cell number and cell size, and thus no effect on fruit size. This is what was observed among the temperature treatments. On the contrary, any treatment that affects the supply of assimilates to fruits, as does truss pruning, with or without any effects on cell number should also affect fruit size.

Many studies describe the endoreduplication dynamic during development of various species and plant organs, but it is still unclear if endoreduplication is involved in the control of cell growth (Sugimoto-Shirasu and Roberts, 2003). A link between cell size and average C-value was found in leaf epidermis (Melaragno et al., 1993), in seed (Lemontey et al., 2000) and in flowers (Kudo and Kimura, 2002; Lee et al., 2004). In yeast cells, the ploidy-regulation of the cell cycle progression could explain the control of endoreduplication on cell size (Galitski et al., 1999). The hypothesis that endoreduplication determines cell growth is further supported by the fact that endoreduplication precedes cell expansion (Traas et al., 1998). In tomato fruit endoreduplication starts very early during the most intensive period of cell division and stops more or less with the cessation of cell expansion (Fig. 5). Nevertheless, relationships observed in yeast or in leaf epidermis, may not exist at the level of fruit tissue such as tomato pericarp with millions of cells at different states, some still dividing, other highly endoreduplicated. A common point of all hypotheses raised and whatever the controlling factor, is that the earlier the end of mitotic activity, the higher the ploidy level, since endoreduplication starts as mitosis is blocked. In that case the increase of endoreduplication at 25/25 °C compared with 20/20 °C would result only from the earlier cessation of mitotic activity at 25/25 °C, which agrees with the delayed onset of cell expansion at 20/20 °C (Fig. 1). A significant effect of temperature on maize endosperm endoreduplication was also reported by Schweizer et al. (1995) to explain inter-seasonal variations. Why was endoreduplication minimum at 25/20 °C (Fig. 5)? If the cell division period was as long at 25/20 °C as at 25/25 °C, as suggested before, then endoreduplication was expected to be the same in both treatments. This may be an experimental artefact, but data were confirmed comparing 5F treatments at the three temperature regimes (not shown). An effect of day/night temperature fluctuations on endoreduplication may also be assumed, but it has never been reported. Temperature may also affect the rate of progression of nuclei from lower C-value to higher C-value and contribute to increasing mean ploidy of the tissue, but such control has never been investigated, although the rate of progression among C-values is not constant (Schweizer et al., 1995).

Absence of an effect of the competition among fruits within a truss or among trusses on endoreduplication has been reported already, even when cell size is affected (Bünger-Kinbler and Bangerth, 1983; Bertin et al., 2003). Traas et al. (1998) proposed that the increase in nuclear DNA content provides a given amount of DNA to support a given future increase in mass, but the exact final size is rather defined by the fruit environment, so that endoreduplication is expected to be more or less loosely correlated with a range of cell size, as observed in Fig. 6 in the present study. Thus rather than being involved in the control of cell growth, endoreduplication may initially determine the potential size of the cell by controlling the switch from complete to incomplete cell cycle, whereas actual cell size would mainly depend on carbon supply to individual cells. Figure 6C and D suggests that the potential cell size determined by DNA endoreduplication was not reached despite low competition (2F treatments) because cell size was the same at 25/25 °C and 25/20 with high and low ploidy levels, respectively. Thus, endoreduplication seems to be a poor indicator of actual cell size in tomato pericarp, since endoreduplication and cell size do not respond in similar ways to the fruit environment.