Robust organ size in Arabidopsis is primarily governed by cell growth rather than cell division patterns

Abstract

Organ sizes and shapes are highly reproducible, or robust, within a species and individuals. Arabidopsis thaliana sepals, which are the leaf-like organs that enclose flower buds, have consistent size and shape, indicating robust development. Cell growth is locally heterogeneous due to intrinsic and extrinsic noise. To achieve robust organ shape, fluctuations in cell growth must average to an even growth rate, which requires that fluctuations are uncorrelated or anti-correlated in time and space. Here, we live image and quantify the development of sepals with an increased or decreased number of cell divisions (lgo mutant and LGO overexpression, respectively), a mutant with altered cell growth variability (ftsh4), and double mutants combining these. Changes in the number of cell divisions do not change the overall growth pattern. By contrast, in ftsh4 mutants, cell growth accumulates in patches of over- and undergrowth owing to correlations that impair averaging, resulting in increased organ shape variability. Thus, we demonstrate in vivo that the number of cell divisions does not affect averaging of cell growth, preserving robust organ morphogenesis, whereas correlated growth fluctuations impair averaging.

Keywords: Arabidopsis thaliana; FTSH4; Endoreduplication; Giant cells; LGO; Morphogenesis; Sepal; Spatiotemporal averaging.

Fig. 1.

Organ size and shape is robust to changes in LGO expression, but not in ftsh4-5 and double mutants. (A-L) Stage 15 mature flowers (A-F) and stage 12 flower buds (G-L) of wild type (A,G), lgo-2 mutant (B,H), LGOoe (C,I; pML1::LGO), ftsh4-5 mutant (D,J), lgo-2 ftsh4-5 double mutant (E,K) and LGOoe ftsh4-5 (F,L). Scale bars: 1 mm. Red arrows point to smaller and irregularly shaped sepals. (M) Violin plots overlaid with dot plots of standard deviation of sepal area within one flower. n=104 (wild type), 100 (lgo-2), 104 (LGOoe), 108 (ftsh4-5), 116 (lgo-2 ftsh4-5), 108 (LGOoe ftsh4) sepals. Bars show the quartiles. Letters mark the groups that are not significantly different. (N-S) Contours of mature outer (abaxial) sepals normalized by size (red lines) and the average sepal shape (black line). n=18 (wild type), 23 (lgo-2), 22 (LGOoe), 24 (ftsh4-5), 27 (lgo-2 ftsh4-5), 22 (LGOoe ftsh4) sepals. (T) Violin plots overlaid with dot plots of variability of abaxial sepal shape (S_2 as described in Hong et al., 2016). n=77 (wild type), 66 (lgo-2), 78 (LGOoe), 85 (ftsh4-5), 96 (lgo-2 ftsh4-5), 74 (LGOoe ftsh4). Bars show the quartiles. Letters mark the groups that are not significantly different. Contours for inner (adaxial) and lateral sepals are available in Fig. S1.

Fig. 2.

The number of cell divisions is decreased by LGO overexpression (LGOoe) and in LGOoe ftsh4 mutants and is increased in lgo-2 and lgo-2 ftsh4-5 mutants. (A-F) Heat maps of number of daughter cells per lineage using lineage tracking from 0 h time point to 120 h time point that are projected onto the 120 h time point for wild type (A), lgo-2 (B), LGOoe (C), ftsh4-5 (D), lgo-2 ftsh4-4 (E) and LGOoe (F). The heat map scale is 1 to 15 daughter cells, where 1 indicates that no divisions have taken place because one cell gave rise to one cell at the final time point. Scale bars: 50 µm. Representative images from n=3 biological replicates. Additional replicates are available in Fig. S2A-F. (G) Probability distribution plots of the number of daughter cells per lineage over the 120 h time-lapse imaging. n≥134 cells per genotype. (H) The count of cells that do not divide over the 120 h time-lapse imaging for each sepal. n=3.

Fig. 3.

Cell sizes remain smaller in lgo-2 and lgo-2 ftsh4-5 and become progressively larger in LGOoe and LGOoe ftsh4. (A-G) Heat maps of cell area at each image time point for wild type (A), lgo-2 (B), LGOoe (C), ftsh4-5 (E), lgo-2 ftsh4-5 (F) and LGOoe ftsh4-5 (G). The heat map scale is 0-4000 µm². Scale bars: 50 µm. Representative images from n=3 biological replicates. Additional replicates and quantification of area variability are available in Fig. S3. (D) Distribution of cell areas at each time point. Statistical analysis (multidimensional scaling) available in Fig. S4.

Fig. 4.

The localization of cell division and cell growth are not affected by number of cell divisions and are made patchier by ftsh4-5. (A-F) Heat maps of number of daughter cells per cell lineage over 24 h intervals for wild type (A), lgo-2 (B), LGOoe (C), ftsh4-5 (D), lgo-2 ftsh4-5 (E) and LGOoe ftsh4-5 (F). The lowest heat map value represents one cell per lineage, which means no division. The greatest heat map value represents four or more cells per lineage. Heat maps are projected onto the later time point. Representative images from n=3 biological replicates. Additional replicates are available in Fig. S2. (G-L) Heat maps of cell area growth over each 24 h interval that are projected onto the later time point for wild type (G), lgo-2 (H), LGOoe (I), ftsh4-5 (J), lgo-2 ftsh4-5 (K) and LGOoe ftsh4-5 (L). The heat map represents the change in ratio of cell area (cell area of later time point divided by cell area of earlier time point) and the scale is 1-3. Daughter cells that result from a division over a given time interval area outlined in white. Localization of fast growth is marked by red outlines, and is band-like in wild type, lgo-2 and LGOoe and patchy in ftsh4-5, lgo-2 ftsh4-5 and LGOoe ftsh4-5. Representative images from n=3 biological replicates. Additional replicates are available in Fig. S5. Distribution of cell area growth for each time interval available in Fig. S7. Statistical analysis (multidimensional scaling) available in Fig. S8. Distribution of cell area growth related to the number of cell divisions in the lineage is available in Fig. S6. Scale bars: 50 µm.

Fig. 5.

The basipetal gradient is preserved when division rate changes and in the ftsh4-5 background. Plots showing 24 h interval growth rates (measured as area doubling per 24 h) projected on the proximal-distal axis for a representative replicate of each genotype. The left side of the x-axis is proximal, and the right side of the x-axis is distal. A stereotypical basipetal growth gradient is observed across all genotypes, where growth starts near the distal tip (24-48 h) and then propagates proximally (48-72 and 72-96 h) and eventually slows down (96-120 h). The green line shows a fitted third order polynomial, which is used to define a reference spatial growth profile for each time interval. This reference is subtracted to obtain the growth rate fluctuations. The y-axis is log₂ so cells that double in area have a value of 1.

Fig. 6.

Subcellular differences in growth rate may facilitate the basipetal gradient in growth. (A,B) Giant cells are artificially subdivided into multiple cells and outlined in white. The heat maps of cell area growth over each 24 h interval are projected onto the later time point for wild type (A) and LGOoe (B). The heat map represents the change in ratio of cell area (cell area of later time point divided by cell area of earlier time point) and the scale is 1-3. Scale bars: 50 µm. Representative images from n=3 biological replicates. Additional replicates are available in Fig. S9.

Fig. 7.

Cell growth rate averages in wild type, LGOoe and lgo-2 but is patchy in ftsh4-5, lgo-2 ftsh4-5 and LGOoe ftsh4-5. (A-F) Heat maps from 24 h time point to 96 h time point for change in ratio of cell area (cell area of later time point divided by cell area of earlier time point) projected on the earlier time point. Growth accumulates into a band across the sepal in wild type (A), lgo-2 (B) and LGOoe (C), and accumulates in patches in ftsh4-5 (D), lgo-2 ftsh4-5 (E) and LGOoe ftsh4-5 (F). Three replicates are shown for each genotype. The heat map scale is 1-10. Scale bars: 20 µm. The principal directions of cell growth are overlaid on the heat maps as black lines that are oriented in the direction that each cell had the most growth and have a length that corresponds to the magnitude of the ratio of growth parallel to the principal direction of growth to growth perpendicular to the principal direction of growth. Red circles mark the region of the sepal with greater area growth.

Fig. 8.

Patchiness is increased in ftsh4-5, despite decreased temporal fluctuations in growth rate. (A-F) Heat maps show spatial fluctuations in cumulative growth over 3 days. Wild type (A), lgo-2 (B) and LGOoe (C) have fewer fluctuations, and ftsh4-5 (D), lgo-2 ftsh4-5 (E) and LGOoe ftsh4-5 appear patchy (F). Fluctuations are calculated by subtracting a fitted third order polynomial as a function of the proximal-distal position (Fig. 5). Scale bars: 40 µm. Explanation of sphere projections in Fig. S11. (G) Patchiness is calculated from the standard deviation in the 3-day growth fluctuations after averaging locally over nearest neighbors. A higher patchiness indicates that spatial fluctuations in cumulative growth rates are more correlated in space. The three replicates for each genotype were analyzed separately. (H) Panels show the 24 h growth fluctuations (growth rate with the contribution of growth from the basipetal gradient subtracted) of lineages on subsequent time intervals. Stronger correlations are visible in the genotypes that include the ftsh4-5 mutation, indicating that lineages have decreased fluctuations in growth rate between time intervals. Disk radius indicates the size of the lineage's parent cell. The 24 h spatial fluctuations in growth are in Fig. S11. (I) Temporal correlations of growth across subsequent 24 h intervals that are plotted in H. Data from the three pairs of time intervals and three replicates is pooled for each genotype to calculate the correlation (Fig. 6). Each growth rate pair is weighed with the size of the lineage parent cell on day 2 in the correlation calculation. Error bars show 95% confidence interval from bootstrap analysis (1000 resamplings).

Fig. 9.

Temporal correlation and patchiness are associated with the ftsh4-5 loss of developmental robustness. (A-D) The mean sepal size variability (A) (average standard deviation of area within a flower, plotted in Fig. 1M) and mean sepal shape variability (B) (plotted in Fig. 1T) are positively correlated with temporal correlations in cell growth rates (plotted in Fig. 8I). Mean sepal size variability (C) and mean sepal shape variability (D) are also positively correlated with patchiness (plotted in Fig. 8G). (E) Therefore, uncorrelated fluctuations in growth lead to reproducible or robust development of organ size and shape, whereas correlated fluctuations in growth lead to variable organ size and shape.