DELLA proteins positively regulate seed size in Arabidopsis

ABSTRACT Human and animal nutrition is mainly based on seeds. Seed size is a key factor affecting seed yield and has thus been one of the primary objectives of plant breeders since the domestication of crop plants. Seed size is coordinately regulated by signals of maternal and zygotic tissues that control the growth of the seed coat, endosperm and embryo. Here, we provide previously unreported evidence for the role of DELLA proteins, key repressors of gibberellin responses, in the maternal control of seed size. The gain-of-function della mutant gai-1 produces larger seeds as a result of an increase in the cell number in ovule integuments. This leads to an increase in ovule size and, in turn, to an increase in seed size. Moreover, DELLA activity promotes increased seed size by inducing the transcriptional activation of AINTEGUMENTA, a genetic factor that controls cell proliferation and organ growth, in the ovule integuments of gai-1. Overall, our results indicate that DELLA proteins are involved in the control of seed size and suggest that modulation of the DELLA-dependent pathway could be used to improve crop yield.

In this manuscript the authors provide evidences that DELLA proteins contribute to seed size control as a DELLA gain-of-function mutant produce larger seeds with increase cell number in ovule integuments. This activity may rely on AINTEGUMENTA as GAI promotes its transcriptional activation and that over-expression of ANT produces similar seed phenotype.
Overall I think the quality of the writing and the figures are very good. This is an interesting paper demonstrating a role for DELLA proteins and more generally GA in the control of seed size which is a valuable trait for crop improvement. The molecular mechanism proposed by the author relies on the trans-activation activity of GAI and the activation of the expression of the ANT transcription factor. This paper will be of interest for a wide readership.

Comments for the author
Minor comments: -The authors nicely show that gai-1 mutant produces larger seeds that the WT and that the gaiT6 loss-of-function mutant has smaller seeds. Other single della mutants also showed seed size phenotypes. As gid mutant and PBZ, an inhibitor of GA synthesis show an increase in seed size, the authors suggest that The greater increase in seed size following PBZ treatment could be due to the simultaneous stabilization of all DELLA proteins . I was wondering whether the authors have observed multiple combinaisons of DELLA mutants to precise the relative contribution of the different DELLA to this seed phenotype. Is seed size even smaller than gaiT6 in double or triple della mutant combinaison? or on the other hand, can we get seeds bigger than gai-1 in gai-1 rgl2∆17 double mutant for instance? - Figure 4: The authors claims that pCYCB1;2:Dbox-GFP expression is increased in gai-1 compare to the WT both in signal intensity and number of GFP positive cells. Can the authors provide some quantitative data on that point? -I understand that as ANT null mutant do not produce viable ovules or seeds, it is not possible to perform genetic assay to confirm that ANT and GAI act in the same pathway but I was wondering whether gaiT6 plants seed size is impacted by ANT over-expression? If the seed size of 35S::ANT and 35S::ANT gaiT6 are comparable, this could reinforce the idea that GAI positively regulates ANT. -In the discussion, the authors rule out the implication of BZR1 in this process based on the fact the bzr1-1D gain-of-function mutant has both an increased cell proliferation and cell expansion of the integuments cells compare to wild type. As BZR1 is controlling cell expansion through molecular complex with other regulators which is regulated by GA and DELLA, it possible that the control of cell expansion and cell division in the integument relies on two different molecular mechanisms (trans-activation versus sequestration). Even if the identification of the transcription factor interacting with GAI to allow expression of ANT is out of the scope of this paper for me, can the other discuss this possibility in the discussion of the paper. -line 267 CYCB1;1, CYCB1;2, and CYCB1;4 enhanced their expression in siliques… and line 296 ANT enhanced its expression in gai-1 siliques… I would prefer for instance, ANT expression level is increased in gai-1 siliques as, as I understand, ANT is not activating its own expression in gai-1. Please clarify this point for line 267 and 296.

Reviewer 2
Advance summary and potential significance to field Demonstration that DELLA proteins acting maternally influence seed size in Arabidopsis contributes to an understanding of the regulation of cell proliferation in the integuments and reinforces the importance of AINTEGUMENTA in the maternal control of seed deveolpment.

Comments for the author
The manuscript of Gomez et al., examines the influence of DELLA proteins on seed size in Arabidopsis. The use of GA signalling gain-and loss of function mutants revealed a correlation between DELLA activity and seed size without modifying seed number or reserve composition. The authors show DELLA acts maternally to increase ovule size by increasing cell division in outer and inner integument layers and detail DELLA expression in developing ovule domains using reporter transgene imaging. They document increased expression of B-cyclins in gai-1 in support of the correlation between integument cell division and DELLA expression leading to increased integument cell proliferation. The authors implicate ANTEGUMENTA in mediating DELLA function noting the similarity of integument phenotypes between ANT overexpression and the gai-1 gain of function mutant as well as the enhanced transcription of ANT in gai-1 or by production of ANT after gai-1 induction.
The experimentation is convincing and the interpretation is appropriate. The evidence presented is novel, supporting DELLA proteins as determinants of Arabidopsis seed size by increasing integument cell number during ovule development. The involvement of ANT in DELLA regulation of seed size is plausible but additional experimentation to demonstrate regulation of ANT by GAI is necessary, perhaps by examining phenotypes of knock down alleles such as ant-T, screening for ANT promoter binding factors and rigorous GAI 2-hybrid screens with transcription factor libraries.
Comments: Figure 3G-H: I have impression that reporter expression stronger in II1 and OI1 whereas cell counts were higher for OI2 Fig2C. Please clarify. Figure 4G-H: Comparison between Ler and gai-1 fluorescence imaging is somewhat subjective. How many observations? Figure 6: Model is speculative, the essence of the study is the relation between DELLA and ANT whereas evidence for DELLA interactors/targets and possible co-regulation by GAI promoter binding by ANT not examined. Perhaps present a more general model including MEE45 and ARF2 regulation?

First revision
Author response to reviewers' comments

RESPONSE TO REVIEWERS:
In the revised manuscript, new text is indicated in blue, whereas eliminated text is mar ked by double strikethrough. Following, we provide our point-to-point response to reviewers:

Reviewer 1 Comments for the Author:
Minor comments: -The authors nicely show that gai-1 mutant produces larger seeds that the WT and that the gaiT6 loss-of-function mutant has smaller seeds. Other single della mutants also showed seed size phenotypes. As gid mutant and PBZ, an inhibitor of GA synthesis show an increase in seed size, the authors suggest that The greater increase in seed size following PBZ treatment could be due to the simultaneous stabilization of all DELLA proteins . I was wondering whether the authors have observed multiple combinaisons of DELLA mutants to precise the relative contribution of the different DELLA to this seed phenotype. Is seed size even smaller than gaiT6 in double or triple della mutant combinaison? or on the other hand, can we get seeds bigger than gai -1 in gai-1 rgl2∆17 double mutant for instance? Response: As nicely pointed by the reviewer, double and triple della mutants also show reduced seed size. We did not include this data in the original version as the single mutants already show a clear phenotype. Following the suggestion of the reviewer, we have included a new panel in Figure S2 (new S2A), where we show the reduced seed size of 3xdella and the 4xdella mutants. Regarding the gain-of-function mutants, we generated several double mutants, but in this case, the strong defects in plant development due to stable DELLA proteins hamper the determination of the seed phenotype. For example, the gai-1 rga 17 mutant (rga 17 from pRGA:GFP-rga 17 line) plant has extremely late flowering, and defects in flower and pistil development and fertility, gai-1 rgl2 17 (rgl2 17 from Gomez et al., 2019) is impaired in germination, whereas gai-1 rgl1 17 (rgl1∆17 Gomez et al., 2020) has strong defect in plant development and flowering. Besides, we also observed increased seed size in PBZ-treated Ler plants, where protein levels of all DELLAs should be increased ( Figure S2C,E). These are Ler plants treated upon bolting with 1 µM PZB, a non-saturated dose, so we avoid defects in germination, plant establishment, or flowering time observed in the double dominant mutants.
- Figure 4: The authors claims that pCYCB1;2:Dbox-GFP expression is increased in gai-1 compare to the WT both in signal intensity and number of GFP positive cells. Can the authors provide some quantitative data on that point? Response: All our confocal analyses are done using at least 3 batches of plants, grown at different times. From each, at least 5 inflorescences (one inflorescence by plant) are analyzed to look for specific developmental stages of ovule development. Overall at least 12-15 good quality images are obtained (a stack of images to detect all GFP-positive nucleus in each ovule, not in a single confocal plane). We have determined the number of GFP-positive nuclei from these images to quantify the increase in cyclin expression: gai-1 has a 70% increase in GFP-positive nuclei when compared to Ler (approximately 3 nuclei in Ler and 5 in gai-1). The new data is included in the text and in the figure legends.
-I understand that as ANT null mutant do not produce viable ovules or seeds, it is not possible to perform genetic assay to confirm that ANT and GAI act in the same pathway but I was wondering whether gaiT6 plants seed size is impacted by ANT over-expression? If the seed size of 35S::ANT and 35S::ANT gaiT6 are comparable, this could reinforce the idea that GAI positively regulates ANT. Response: Testing of 35S:ANT and 35S:ANT gaiT6 would be important. Unfortunately, we do not have this double mutant at this moment. The main reason is that 35S:ANT is in Col0, whereas gaiT6 is in Ler backgrounds. Therefore, generation of this mutant requires more than one year (initial cross and at least 3x backcrosses). We are generating this mutant to show the data in the context of the following paper where we plan to identify and describe the potential GAI interacting protein that is responsible of the direct regulation of ANT, as well as the identification of other GAI-targets during ovule development that would be involved in the increased ovule and seed size. We are currently embarked in the identification of GAI interacting proteins specifically in the ovule integuments by a TurboID approach (see full response to reviewer#2 below).
-In the discussion, the authors rule out the implication of BZR1 in this process based on the fact the bzr1 -1D gain-of-function mutant has both an increased cell proliferation and cell expansion of the integuments cells compare o wild type. As BZR1 is controlling cell expansion through molecular complex with other regulators which is regulated by GA and DELLA, it possible that the control of cell expansion and cell division in the integument relies on two different molecular mechanisms (trans-activation versus sequestration). Even if the identification of the transcription factor interacting with GAI to allow expression of ANT is out of the scope of this paper for me, can the other discuss this possibility in the discussion of the paper. Response: We agree that this idea is worth to be included in the discussion. We have proceeded accordingly and modified the discussion.
-line 267 CYCB1;1, CYCB1;2, and CYCB1;4 enhanced their expression in siliques… and line 296 ANT enhanced its expression in gai-1 siliques… I would prefer for instance, ANT expression level is increased in gai-1 siliques as, as I understand, ANT is not activating its own expression in gai-1. Please clarify this point for line 267 and 296. Response: Thanks for the tip, the mutants are always in italics through the text. We miss it here probably during handling of the text. Changes in the text are done as indicated.

Reviewer 2 Comments for the Author:
The experimentation is convincing and the interpretation is appropriate. The evidence presented is novel, supporting DELLA proteins as determinants of Arabidopsis seed size by increasing integument cell number during ovule development. The involvement of ANT in DELLA regulation of seed size is plausible but additional experimentation to demonstrate regulation of ANT by GAI is necessary, perhaps by examining phenotypes of knock down alleles such as ant-T, screening for ANT promoter binding factors and rigorous GAI 2-hybrid screens with transcription factor libraries. Response: We agree that it is crucial to show the direct relationship between GAI and ANT and look for the GA-interacting protein involved. Unfortunately, in our experimental conditions, we have not observed any ovule/seed phenotype in ant-T mutant plants. Therefore, we have not used this line to address the direct regulation of GAI-ANT. We have a list of transcriptional regulators that are putative RGA and GAI-interacting proteins, in a Y2H assay (https://doi.org/10.1007/978-1-4939-7871-7_9) using a platform with over 1200 transcriptional regulators of Arabidopsis. This list of genes lacks the specificity as we are looking for interactors in a specific tissue and developmental stage. Similar criticism can be done to the search for ANT promoter binding factors. Both GAI and ANT are being expressed in different tissues during plant development. To cope with this, we are currently carrying out a TurboID approach to identify GAIinteracting proteins in the integuments of developing ovules. For this we have generated several lines that express the fusion of GAI-TurboID (TurboID can add biotin to proteins situated in close proximity) under the control of the INO (INNER-NO-OUTER) promoter that directs the expression specifically to the integuments during ovule development, as well as all the control lines needed to reduce false-positives. Upon biotin treatment in inflorescences, only proteins that are very close to GAI in integuments will be biotinylated by TurboID. Later, these proteins can be isolated and identified by MS. We hope that this strategy will provide experimental evidence of the genetic components integrating the molecular network involved in the control of the GAImediated ovule and seed. Figure 3G-H: I have impression that reporter expression stronger in II1 and OI1 whereas cell counts were higher for OI2 Fig2C. Please clarify. Response: Indeed, it seems that GFP signal is stronger in II1 and oi1, and lower in ii2 or oi2 in Figure 3G-H. It must be noted that cell division is reduced at the developmental stage shown in 3G. Moreover, gai-1 expression is stronger in the oi2 at earlier developmental stages, such as that shown in Figure 3F, when cell division is very active. Finally, it is not known whether cells of the inner integument could be able to activate division in the outer integument, which would be required for coordinated growth of all cell layers. Figure 4G-H: Comparison between Ler and gai-1 fluorescence imaging is somewhat subjective. How many observations? Response: As indicated above, we have compared these genotypes from different (at least three) batches of plants, growth at different moments. From each set of plants, we usually obtain good quality confocal images from at least 5 inflorescences. The selected image is always representative of the whole set of images. In the case of the CYC expression, we have used these good-quality images to count the number of GFP-positive nucleus as suggested by reviewer#1. We indicate this in Methods and in the figure legend. Figure 6: Model is speculative, the essence of the study is the relation between DELLA and ANT whereas evidence for DELLA interactors/targets and possible co-regulation by GAI promoter binding by ANT not examined. Perhaps present a more general model including MEE45 and ARF2 regulation?

Comments:
Minor issues: Response: We agree that this figure is speculative as it shows the current working model of the possible molecular mechanism integrating the DELLA and ANT in the control of ovule and seed size. Following the suggestion, we have modified Figure 6 to include the direct regulation of the expression of ANT by MEE45 and ARF2.
156: substitute "…was associated with…" for "caused" Response: Changes in the text are done as indicated.

221: fig S4A-C Integument cell size, length increases number increases but size decreases?
Response: We understand that the current version of the figure S4 may be misleading. The panel A indicates the length of the whole oi2 cell layer, whereas panels B and C represent the average cell number and size, respectively, in the same oi2 cell layer. The whole cell layer increases, due mainly to strong increase in cell number, although these cells are slightly smaller. To avoid confusion, the figure legend and the Y-axe labels have been modified.

289: elaborate "indirect" GAI binding sites
Response: We have modified the text to clarify this issue. In the introduction we have already stated that DELLA proteins do not contain a DNA binding domain and hence they regulate transcription by direct binding to transcription factors (Vera-Sirera et al., 2015;Daviere and Achard, 2016). 345: insert "…potentially powerful..." 365: substitute "…associated with an…" for "…likely through…" Response: Changes in the text are done as indicated. 371-372 and 398: are these observations mutually compatible in the present study?
Response: We agree that both statements seem contradictory, but they are not. A delay in embryonic development has been observed in some (but not all) mutants that have larger seeds. We therefore wanted to check whether this was also the case in gai-1 and found that gai-1 embryonic development is not delayed compared to WT. This would be compatible with the fact that gai-1 seeds are larger due to increased expression of ANT, as ANT prolongs cell proliferation at very early stages of ovule development, before fertilization occurs. We would like to emphasize that the mature gai-1 ovule is already larger than WT ovule, which may explain that the embryo development is not delayed. 393: substitute "… the ANT-YPet protein level is greater in…" for "protein increases its levels" 818: suppress "Ratio" I am happy to tell you that your manuscript has been accepted for publication in Development, pending our standard ethics checks.

Advance summary and potential significance to field
In this manuscript the authors provide evidences that DELLA proteins contribute to seed size control as a DELLA gain-of-function mutant produce larger seeds with increase cell number in ovule integuments. This activity may rely on AINTEGUMENTA as GAI promotes its transcriptional activation and that over-expression of ANT produces similar seed phenotype.

Comments for the author
This last version of the manuscript together with the added data takes into account all of my comments therefore I believe it's ready for publication.

Reviewer 2
Advance summary and potential significance to field A demonstration of correlation between DELLA activity and seed size mediated through an increase in ovule size resulting from increased cell proliferation in outer and inner integument layers. Development of conceptual model implicating AINTEGUMENTA as a nexus for control of organ growth.

Comments for the author
The revised manuscript of Gomez et al., has addressed the minor issues raised in my initial review. There are now a few points to correct and elaborate in version 2. Lines 267-269: The statement "The number of GFP positive…respectively)" refers to which panels in Figure 4?