Has the impact of endoreduplication on cell size been overestimated?

To date, several publications have referred to the relationship between the level of endoreduplication and cell size in Arabidopsis, claiming that enhanced/suppressed endocycling by mutation and transgenes results in increased/decreased cell size in leaves, respectively. Most of these studies have quantified the increase in cell size and ploidy level; a correlation between the two variables has been proposed but it has not been confronted to a qualitative meta-analysis of the existing data. For example, the virtual endoreduplication profiles shown in Fig. 1 mimic data frequently seen in reports on changes in endoreduplication levels in Arabidopsis. Arabidopsis leaves are composed of a mixture of cells with different ploidy levels, and flow-cytometry analysis gives us a mosaic profile of endoreduplication (Fig. 1, right). Although a study on this topic did not consider the relationships among cell size, ploidy, and the decision to divide which is regulated in a stochastic manner (Kawade & Tsukaya, 2017). In this case, the proportions of 8C and 16C are greater in the mutant than in the wild type (WT); thus, the average cell size is expected to be larger in a mutant than in the WT. Indeed, most reports describe observed changes in cell size in similar mutants as the result of altered endoreduplication levels. But how much difference should be expected? Past studies have indicated that the volume of 4C cells is twice that of 2C cells, and the 8C volume is twice that of the 4C volume, and so on. This rule is quite rigid because cell volume is correlated strongly with the nuclear ploidy level in plants, animals, and fungi (Müntzing, 1936; Fankhauser, 1952; Storchová et al., 2006). Since Arabidopsis leaves are composed of a mixture of different ploidy levels, cell size is presented as the average of measurements. In the earlier case, the endoreduplication profile for the WT is (2C 25%; 4C 50%; 8C 25%), and its average cell volume is equal to 2.25-fold the 2C cell volume (1 × 0.25 + 2 × 0.50 + 4 × 0.25). However, the endoreduplication profile for the mutant in the earlier mentioned case is (2C 15%; 4C 50%; 8C 30%; 16C 5%), and thus its average cell volume is equal to 2.75-fold that of 2C cells. Therefore, the mutant cell volume is expected to be only ~22% larger than the WT cell volume, although the two endoreduplication profiles differ. In past studies, irrespective of much subtler changes in endoreduplication profile, many mutants and transgenics have been reported to have much larger cell sizes than the WT (often 50–200% larger, even in a projected-area base measurement, but not in volume). Is the attribution of all observed cell size changes to the slight difference in endoreduplication profiles warranted? In addition, use of the median, rather than mean, cell size value highlights the problem. Since the mutant differs only in the minor population of cells in the ploidy, the median ploidy value is the same for the WT and mutant: 4C (see references cited in this report and Fig. 1, green arrows). Thus, if changes in the observed cell size are due to altered ploidy levels, the median cell size should be the same in the WT and the mutant. If the median cell size in a mutant differs from that in the WT, the difference in cell size should be attributed to factors other than endoreduplication. However, no quantitative analysis from such a viewpoint has been carried out. To examine the rigidity of the relationship between changes in the endoreduplication profile and cell size, I conducted a quantitative analysis of this correlation under the following assumptions. Initially, I assumed that leaf cell volume would double following additional endoreduplication, as explained earlier. Cell volume generally doubles with the doubling of ploidy level in plants, animals, and fungi. In the 1920s and 1930s, cell volume was shown to be proportional to the ploidy level in analyses of Physcomitrium piriforme, Oenothera spp., and Nicotiana spp. (von Wettstein, 1928; Gates & Goodwin, 1930; Kostoff, 1938). Also, if we measure the sizes of mesophyll protoplast cells of diploid (2C) and tetraploid (4C). Columbia WT Arabidopsis leaves, the 4C/2C ratio of the diameters of mesophyll protoplasts is 1.24 (Supporting Information Fig. S1), indicating that the 4C/2C ratio of the volume of mesophyll protoplasts is 1.89 (1.243). This is also the case for animals, as shown in rainbow trout (Refstie, 1981). Storchová et al. (2006) proposed that genome instability in budding yeast induced by polyploidization is caused by mismatches of scale: cell volume doubles with the ploidy level, but the spindle length does not change (see Fig. 5a in Storchová et al., 2006). This mismatch of scale is also a widely accepted mechanism of cancer formation in mammals (Ceol et al., 2007). Fankhauser (1941) reported that the nuclei were 1.5-fold larger in triploid newts (Triturus viridescens) than in diploid individuals. In mouse embryonic stem cells, tetraploidization showed a 2.2-fold increase in cell volume relative to diploid lines (Imai et al., 2016). Thus, the proportional effect of the ploidy level on cell volume is a widely conserved phenomenon. Endoreduplication is one of the natural processes that causes polyploidization as well as a proportional increase in cell volume (Robinson et al., 2018). If endoreduplication causes cell size changes that are not proportional, these changes should be considered to be caused by the combination of the pure ploidy effect with another, unknown, effect. In almost all reports examined here, cell size has been measured in terms of cell area from the paradermal view (projected area against the paradermal plane), and 1.3–2.3-fold increases in area have been reported to be associated with enhanced endoreduplication profiles. A 2-fold change in cell volume is equivalent to a ~1.59-fold change in cell area; in other words, a 2-fold change in cell area represents a 2.83-fold cell volume change. This assumption allowed me to compare data among publications in a precise manner to more effectively determine the relationship between cell size and endoreduplication. Calculation and estimation of cell volume were carried out as described in the text. S-phase cells in young leaf primordia were visualized following the method described by Yin & Tsukaya (2016). Cell size was measured using the ImageJ software (http://imagej.nih.gov/ij/). The mutant strain fugu2-1 has been described previously (Hisanaga et al., 2013). All data analyzed in this study are open. I chose 11 publications (De Veylder et al., 2001; Ferjani et al., 2007; Ramirez-Parra & Gutierrez, 2007; Kurepa et al., 2009; Sonoda et al., 2009; Baloban et al., 2013; Hur et al., 2015; Peng et al., 2015; Xu et al., 2016; Wang et al., 2017; Zhao et al., 2017) that contained data on the endoreduplication profile and cell size for leaves at the same stage, same node position, and under the same growth conditions. For each publication, I normalized the data to eliminate the influences of laboratory and culture conditions by converting cell size data for mutants/transgenics to relative values against the WT. Additionally, I calculated the expected change in cell size from the observed differences in endoreduplication profiles (ratios of 2C, 4C, 8C, 16C, and 32C). Because 2nC cells are assumed to be 2-fold larger in volume than nC cells, the average cell size was given as a unit volume of 2C cells, as calculated earlier for the examples shown in Fig. 1. For each publication, I calculated the expected cell size in terms of the relative change from the WT. Thus, for the example given in Fig. 1, the relative cell size of the mutant against the WT is 2.75/2.25 = ~1.22. All cell sizes were measured as areas on the paradermal plane, and calculated cell volumes were converted to cell areas as (relative volumes)2/3, based on the assumption that parenchymatous cells are approximately spherical, in particular under normal growth conditions (sun-type leaves develop rod-like palisade tissue cells under very high (> 150 μmol m−2 s−1) light irradiation, whereas shade-type leaves with sphere-like palisade tissue cells form under standard growth conditions; see Weston et al., 2000). For the given example, the cell-volume change is a ~1.14-fold increase. Finally, because the impact of the nuclear ploidy level on cell volume differs between epidermal and palisade cells (Katagiri et al., 2016; Kawade & Tsukaya, 2017), data sets were divided into those for epidermal and palisade cells. If the level of endoreduplication were the sole factor determining leaf cell size, then the observed change in cell size would perfectly match the calculated change in expected cell size. However, this was not the case, as shown in Fig. 2. In many cases, the observed changes in cell size were far larger than the expected changes (increases or decreases) in epidermal and palisade cells. These results raise doubt about whether cell size changes were caused by altered endoreduplication. To explain these large differences in cell size, an investigation of the effects of other factor(s) that might amplify the ploidy-dependent effects appears to be required. This large discrepancy can also be investigated more accurately by analyzing data distribution, e.g. by comparison of median values of cell size in the examined mutants/transgenics. Ferjani et al. (2007) examined endoreduplication levels and cell size in five lines, and all examined lines had the same median ploidy value (4C; see Fig. 8 in Ferjani et al., 2007). Therefore, if only the endoreduplication level can affect cell size, the median cell size in these five lines should be the same. I gained access to the raw data set of Ferjani et al. (2007) and found that median cell sizes differed markedly among lines (WT, 2812.5 μm2; fugu2, 5697.7 μm2; fugu5, 3628.2 μm2; an3-4, 5452.6 μm2; er-102, 4171.6 μm2; and KRP2o/x, 5344.3 μm2). These differences strongly indicate that not only the ploidy-dependent pathway via endoreduplication, but also another unknown factor, regulates cell size. The impact of endoreduplication-dependent ploidy on cell size increase is not plausibly amplified by endocycling because this impact was shown to be linear in a single-cell-level analysis of WT leaves (Katagiri et al., 2016). Alternatively, if increased endoreduplication were associated with decreased mitosis in a trade-off manner, mutants/transgenics with increased endocycling would produce fewer cells in mitosis, which is known to trigger compensating cell enlargement (Hisanaga et al., 2015). Since compensating cell enlargement is generally independent of the nuclear ploidy level (Ferjani et al., 2007), observed increases in cell size could be the results of a synergistic effect between endoreduplication and compensation. This hypothesis might explain the greater increase in cell size observed in some enhanced endoreduplication mutants/transgenics, but it cannot explain the greater decrease in cell size when endoreduplication is decreased because compensating cell enlargement is observed only when mitosis is decreased, but the reverse has not been observed (Hisanaga et al., 2015). I hypothesized that this discrepancy is the result of a hidden assumption: that cell size at the mitotic 2C stage is equal in the mutant/transgenic and WT. That is, cell size is assumed to be the same immediately after mitosis, with differences in cell size emerging later during cell growth. The difference in median cell size mentioned earlier suggests that this assumption is incorrect. This estimation perfectly matches the observed change in cell size at the mature leaf stage (2.3-fold larger than the WT) (Ferjani et al., 2007). Thus, at least in KRP2 ox, the discrepancy between the endoreduplication-based expectation and the real cell size change can be fully accounted for by the difference in the default cell size at the mitotic 2C stage. How to explain the small phenotypes observed in some of the mutants/transgenics shown in Fig. 2? Fig. 2(a) shows that the pA12-A12-SRDX lines (Hur et al., 2015) are far smaller than expected due to the decreased ploidy profile. A12 is ATHB12, a member of a homeodomain-leucine zipper class I gene, and pA12-A12-SRDX lines are transgenics that were forced to suppress all ATHB12-related pathways (Hur et al., 2015). The expected cell size reduction for pA12-A12-SRDX due to decreased endoreduplication was calculated as 0.90-fold; however, the observed cell size difference was 0.38. Importantly, according to Fig. 3(b) of Hur et al. (2015), the cell size at the mitotic phase (6 days after sowing) was already much smaller in pA12-A12-SRDX than in the WT (P < 0.01). Cell number at this phase was < 1/10 of that at the mature stage (14 days after sowing), showing that pA12-A12-SRDX had much smaller mitotic 2C cells than did the WT. Thus, the observed small cell phenotype can again be explained by a synergistic effect between the small cell phenotype at the mitotic phase and the endoreduplication-dependent factor. Interestingly, A12 provided further evidence of the impact of default cell size at the 2C phase. In contrast to pA12-A12-SRDX, which lacks all A12-related activity, A12ox exhibited much larger cell sizes than expected from the endoreduplication profile (Fig. 2a). Furthermore, Fig. 3(b) of Hur et al. (2015) shows that the cell size was c. 1.74-fold larger 6 days after sowing (P < 0.01), at the mitotic stage. Since the change in cell size expected from the endoreduplication profile was only 1.12-fold (Fig. S1), most of the final difference in the A12ox cell size (1.62-fold larger than the WT) should be attributed to the default cell size at the mitotic 2C phase. Finally, I examined the cell size of fasciata1/fugu2-1 (fugu2) mutant leaves that had also showed discrepancy between the expected and observed changes (Fig. 2a). I measured the fugu2 cell size at the mitotic phase in early leaf primordia and compared it with that of the WT using EdU incorporation as a marker of the active cell proliferative zone. As shown in Fig. 3, the fugu2 mutant had much larger cells at the most actively proliferating 2C zone. Therefore, although the fugu2 mutant has been reported to exhibit enhanced cell expansion and endoreduplication (Hisanaga et al., 2013), the observed large cell phenotype should be attributed at least partly to the large default cell size at the 2C mitotic phase. Taken together, these results indicate that (1) in most cases, changes in endoreduplication cannot fully explain the observed cell size phenotype; and (2) cell size can differ due to mutations or transgenes, even at the actively mitotic 2C stage. The earlier point has been noted in several reports to date (De Veylder et al., 2001; Sizani et al., 2019), and it has even been suspected that 'cell growth can drive endocycle but the inverse is very unlikely' (Sizani et al., 2019). The present analysis seems to support this idea. To date, the latter point has been overlooked, and all cell size data have been discussed only from the perspective of ploidy-dependent cell enlargement. If we assume that endoreduplication is the only cause for cell size increase, cell volume would need to increase by a factor of four or five in each cycle. The regulators of endoreduplication examined in this metadata analysis would also affect other aspects of mitosis that may be more relevant to cell size increase. To understand the mechanisms of cell size regulation, we must analyze cell size not only in the post-mitotic phase, but also at mitosis. The author thanks the research community of this topic, which stimulated the discussion on this theme. Anonymous referee #2 kindly suggested that median values be used for discussion of the discrepancy between the observed ploidy level and cell size. Dr Ferjani Ali of Tokyo Gakugei University kindly supplied his raw data sets for the metadata analyses. This research was supported by a Grant-in-Aid for Scientific Research on Innovation Areas (#25113002). The author declares that there are no competing interests. Please note: Wiley Blackwell are not responsible for the content or functionality of any Supporting Information supplied by the authors. Any queries (other than missing material) should be directed to the New Phytologist Central Office. Please note: The publisher is not responsible for the content or functionality of any supporting information supplied by the authors. Any queries (other than missing content) should be directed to the corresponding author for the article.