DELLA Signaling Mediates Stress-Induced Cell Differentiation in Arabidopsis Leaves through Modulation of APC/C Activity 1[W]

Drought is responsible for considerable yield losses in agriculture due to its detrimental effects on growth. Drought responses have been extensively studied, but mostly on the level of complete plants or mature tissues. However, stress responses were shown to be highly tissue- and developmental stage-specific, and dividing tissues have developed unique mechanisms to respond to stress. Previously, we studied the effects of osmotic stress on dividing leaf cells in Arabidopsis ( Arabidopsis thaliana ), and found that stress causes early mitotic exit, in which cells end their mitotic division and start endoreduplication earlier. In this study we analyzed this phenomenon in more detail. Osmotic stress induces changes in gibberellin (GA) metabolism, resulting in stabilization of DELLAs, which are responsible for mitotic exit and earlier onset of endoreduplication. Consequently this response is absent in mutants with altered GA levels or DELLA activity. Mitotic exit and onset of endoreduplication do not correlate with an upregulation of known cell cycle inhibitors, but is a result of reduced levels of DEL1/E2FE and UVI4, both inhibitors of the developmental transition from mitosis to endoreduplication by modulating anaphase-promoting complex/cyclosome (APC/C) activity, which are downregulated rapidly following DELLA stabilization. This work fits into an emerging view of DELLAs as regulators of cell division by regulating the transition to endoreduplication and differentiation.


INTRODUCTION
Abiotic stresses, such as drought, have long been known to inhibit plant growth and thereby decrease crop productivity. This growth inhibition is an active response to stress, but little is currently known on how this is brought about (Skirycz and Inzé, 2010). Recently, it has become evident that tolerance to stress, which has been extensively studied in the past, relies on different mechanisms than growth inhibition by stress, and that specific experimental setups have to be developed to investigate stress-induced growth inhibition (Skirycz et al., 2011b).
Organ growth is driven by both cell proliferation and cell expansion, in the case of Arabidopsis (Arabidopsis thaliana) leaf growth occurring sequentially in time. Initially, all cells in the leaf primordium are dividing but, later in development, cell division ceases from the tip to the base of the leaf, resulting in a cell cycle arrest front moving across the leaf (Donnelly et al., 1999;Kazama et al., 2010;Andriankaja et al., 2012). This transition from cell proliferation to cell expansion is accompanied by a switch from the mitotic cell cycle to endoreduplication, during which the genome is replicated, but mitosis does not occur, leading to cells with higher ploidy levels (Beemster et al., 2005). The current view is that this developmental transition from mitosis to endocycle, or mitotic exit, is triggered by a decrease in mitotic (B-type) cyclin-dependent kinase (CDK) activity (De Veylder et al., 2007;Breuer et al., 2010). This can occur through upregulation of cell cycle inhibitors such as Kip-related protein 2 (KRP2; Verkest et al. (2005)) and SIAMESE (SIM), which has been shown to control endoreduplication in Arabidopsis trichomes (Churchman et al., 2006) and can interact with CDKB/cyclin complexes (Van Leene et al., 2010). Another major pathway is the control of anaphase-promoting complex/cyclosome (APC/C) activity through its activating CCS52A subunits, which positively regulate endoreduplication (Lammens et al., 2008) by targeting mitotic cyclins for destruction (Boudolf et al., 2009;Kasili et al., 2010). Activation of APC/C by CCS52A proteins is counteracted by UVI4 (Hase et al., 2006;Heyman et al., 2011;Iwata et al., 2011). Upstream, the atypical E2F-like protein DEL1/E2FE represses the expression of CCS52A genes, thereby delaying the developmental onset of endoreduplication (Lammens et al., 2008). Regulation of DEL1 transcription is less clear, although recently it was shown that the classical E2Fs E2Fb and E2Fc antagonistically regulate DEL1 transcription, thereby controlling endoreduplication in the hypocotyl in response to light (Berckmans et al., 2011), while UVI4 expression is also activated by E2Fa and E2Fb (Heyman et al., 2011).
Endoreduplication generally correlates with cell expansion and differentiation, but its physiological role is still under debate, and is likely tissue-and stimulus-specific (De Veylder et al., 2011). One intriguing hypothesis is that endoreduplication ensures cell fate maintenance by preventing dedifferentiation (Bramsiepe et al., 2010).
When plants are confronted with limited water availability, both cell proliferation and cell expansion are affected, leading to smaller leaves composed of less and smaller cells (Schuppler et al., 1998;Granier and Tardieu, 1999;Aguirrezabal et al., 2006;Skirycz et al., 2011a). Previously, we could show that when stress hits during the proliferation phase, cell division is first reversibly arrested in a post-transcriptional manner by ethylene signaling resulting in a reduction of CDKA activity (Skirycz et al., 2011a). CDKA is the main driver of the cell cycle, and is involved in both the G1-to-S and G2-to-M transitions (Inzé and De Veylder, 2006). When the stress persists, cells start the transition to cell expansion by exiting the mitotic cell cycle in favor of endoreduplication. Here, we show that this process is dependent on gibberellin signaling. Gibberellins (GAs) are a class of diterpenoid hormones which are involved in various processes throughout the plant life cycle, including seed germination, vegetative growth, bolting and flowering, and stress response (Achard et al., 2006;Sun, 2008). Regulation of GA levels occurs at both biosynthesis, through GA 20-oxidases (GA20OX) and GA 3-oxidases (GA3OX), and degradation, which is mainly catalyzed by GA 2-oxidases (GA2OX). All these enzymes occur in small families in Arabidopsis, and have tissue-specific expression patterns (Mitchum et al., 2006;Rieu et al., 2008a;Rieu et al., 2008b), allowing for a tight temporal and spatial control of GA levels. GA signaling occurs by binding of GA to its receptor, GID1, leading to the formation of a complex with DELLA proteins, transcriptional regulators that inhibit GA responses in the absence of GA. This results in recognition and degradation of DELLA proteins by an SCF complex with SLY1 or SNE/SLY2 as the F-box components (Dill et al., 2004;Fu et al., 2004;Ariizumi et al., 2011). DELLA activity is also regulated through non-proteolytic means by phosphorylation, although its effects on DELLA activity are still under debate (Sun, 2010), and by the action of the N-acetylglucosamine (GlcNAc)transferase SPY, which directly or indirectly activates DELLAs (Jacobsen and Olszewski, 1993;Silverstone et al., 2007).
DELLAs are thought to be responsible for all GA responses, which are very pleiotropic. As a result, DELLAs have the potential to induce very different transcriptomes, depending on the tissue, developmental stage and stimulus that is studied, with only a small core involved in feedback regulation being conserved (Ogawa et al., 2003;Cao et al., 2006;Nemhauser et al., 2006;Zentella et al., 2007;Achard et al., 2008b;Hou et al., 2008;Gonzalez et al., 2010). As DELLAs have no DNA-binding domain, they function by interacting with and thereby inhibiting a wide array of other transcription factors, such as PIF3, PIF4, SPT, JAZ1, and SCL3 (de Lucas et al., 2008;Feng et al., 2008;Hou et al., 2010;Josse et al., 2011;Zhang et al., 2011). DELLAs can also activate transcription through association with DNA as shown by ChIP analysis (Zentella et al., 2007), and this most likely occurs through interaction with yet unknown DNA-binding factors. Arabidopsis contains 5 DELLAs (RGA, GAI, RGL1, RGL2, and RGL3) that are to some extent functionally divergent mainly due to their expression patterns (Gallego-Bartolomé et al., 2010), with for instance RGA and GAI being the most important for regulation of vegetative growth (Dill et al., 2001).
Earlier reports also showed that various stresses can induce cell cycle inhibitors of the CDK inhibitor (ICK)/KRP or SIM family, and thereby arrest the cell cycle (Wang et al., 1998;Pettkó-Szandtner et al., 2006;De Veylder et al., 2007;Peres et al., 2007). Furthermore, DELLAs appear to be able to control cell proliferation rates by regulating KRP2, SIM, SMR1 and SMR2 transcription (Achard et al., 2009). However, our results suggest that while stressinduced mitotic exit of proliferating cells is fully DELLA-dependent, it is not a result of elevated cell cycle inhibitor levels, but mitotic exit is rather regulated by modulation of APC/C activity through DEL1 and UVI4.

DELLA Stabilization is First Observed 24 Hours after Stress Onset
Studying how osmotic stress affects proliferating cells in an organ requires a specific setup, as described previously (Skirycz et al., 2011a). In short, plants are grown on nylon meshes overlaying control medium until nine days after stratification (DAS), when all cells of the third true leaf are still dividing. The meshes are then transferred to either control medium or medium containing 25 mM mannitol, thus causing acute, yet mild stress to proliferating cells.
For all further analyses the third leaf is microdissected or cut from the plant at different time points following stress onset. Transcriptome analysis was performed on microdissected leaves harvested at different time points after stress onset (Skirycz et al., 2011a). In this experiment we found changes in GA metabolism genes after 24 hours, with the respective down-and upregulation of the biosynthesis gene GA3OX1 and the catabolism gene GA2OX6, which could be confirmed by quantitative PCR (qPCR) (Fig. 1A). Correspondingly, amongst genes differentially regulated in proliferating leaves 24 hours after transfer to mannitol, there was an overrepresentation of genes known to be regulated by DELLAs in response to biotic stress (flg22; Navarro et al., 2008) and salt stress (Achard et al., 2008b), providing further evidence for higher DELLA activity at this time point (Fig. 1B). The timing suggested a role for GAs and DELLAs in late stress-induced mitotic exit. Therefore we investigated the abundance of RGA by western blot on leaves of RGA::GFP-RGA plants. In accordance with the observed transcript changes, exposure of plants to osmotic stress caused a small but significant stabilization of GFP-RGA after 24 hours, and this persisted after 48 hours ( Fig 1C). In samples taken 3 and 12 hours after transfer no stabilization could be detected, confirming that DELLA accumulation is a relatively late event.

DELLAs Trigger Mitotic Exit
As timing of DELLA accumulation fitted well with the timing of mitotic exit and early onset of endoreduplication, we wanted to further explore this event and investigate whether DELLAs could be causative. To confirm that the earlier onset of endoreduplication was due to proliferating cells exiting mitosis and entering endoreduplication, we cut leaves in half at stages where the bottom half was still proliferating, while the top half had already begun expanding (Andriankaja et al., 2012), and measured ploidy levels of both halves. This confirmed that upon mannitol treatment, the main changes are found in the bottom half (Supplementary Figure 1). In control leaves, high levels of 4C mitotic nuclei sharply drop as the proliferation zone collapses and then go up again after a small pause as endoreduplication starts. In mannitol-treated leaves endoreduplication starts earlier, confirming that the changes we see in the endoreduplication index are due to proliferating cells ceasing proliferation. To investigate the effects of DELLAs on mitotic exit, we first exposed plants to paclobutrazol (PAC), a chemical that stabilizes DELLAs by inhibiting GA biosynthesis. Transfer of seedlings to PAC led to cell cycle arrest, as shown by reduced cell numbers ( Fig. 2A), and triggered early differentiation, manifested by weaker and patchy CYCB1;1::DBox-GUS staining ( Fig. 2B), which stains mitotic cells at the G2-to-M transition (Colón-Carmona et al., 1999) and earlier endoreduplication onset as demonstrated by ploidy measurements (Fig. 2C).
This chemical treatment was confirmed by mutant analysis. To avoid pleiotropic effects, we selected mutants with altered GA levels or DELLA activity that showed relatively limited growth phenotypes under normal conditions: q-ga2ox, a quintuple knock-out for five GA 2oxidases, resulting in higher GA levels (Rieu et al., 2008a); spy-3, a weak loss-of-function allele of the DELLA activator SPY (Jacobsen and Olszewski, 1993;Silverstone et al., 2007); the double DELLA loss-of-function mutant rga-28 gai-2; and ga3ox1-3, in which the major GA 3-oxidase responsible for GA biosynthesis in vegetative tissues is inactive, resulting in lower GA levels (Mitchum et al., 2006). q-ga2ox, spy-3 and rga-28 gai-2 mutants showed a similar cell number reduction by mannitol compared to wild type (WT) (Supplemental Fig.   S2, A-C). Cell number reductions were quite variable between experiments, which is why care was taken to always compare to a WT grown in the same experiment. Measurement of ploidy levels however showed that the early differentiation caused by mannitol was completely abolished in these mutants (Fig. 3, A-D). Surprisingly, while cell proliferation was more inhibited in the ga3ox1-3 mutant (Supplemental Fig. S2D), this mutant also lacked mannitol-induced early differentiation (Fig. 3E). Importantly, none of the mutants, except for rga-28 gai-2, exhibited altered ploidy levels under standard conditions (Supplemental Fig.   S3). Further confirmation came from 35S::gai-GR lines, which overexpress a non-degradable variant of the DELLA protein GAI fused to the rat glucocorticoid receptor, rendering GAI activity inducible by dexamethasone (DEX), which allows migration of the fusion protein to the nucleus where it can be active. Although also under control conditions both the complete plants and the third leaf are smaller than WT (Fig. 4A), the relative growth rate of the third leaf is similar during the developmental stages studied here (Fig. 4B). Upon induction with DEX, growth rates are drastically reduced (Fig. 4B), and a reduction in epidermal cell number is apparent (Fig. 4C). Importantly, early endoreduplication is observed, similar to WT plants treated with mannitol (Fig. 4D). This phenotype is also already apparent, albeit to a lesser extent, without DEX, either due to leakiness of the construct or a cytosolic function of GAI.

Inhibitors
As DELLAs were previously shown to control transcription of KRP2, SIM, SMR1, and SMR2 (Achard et al., 2009), these were obvious targets to explain the mitotic exit. However, in our previous microarray analysis on mannitol-treated proliferating leaves, none of the known cell cycle inhibitors were upregulated; some even showed downregulation, as was the general trend among cell cycle genes 24 hours after stress onset (Skirycz et al., 2011a). These microarray results were confirmed by qPCR for SIM and SMR1, showing that even 72 hours after stress onset, when mitotic exit had occurred, the expression remained low (Fig. 5A).
Similar to mannitol treatment, in 35S::gai-GR leaves negative cell cycle regulators were downregulated along with other cell cycle genes only at 12 hours after DEX induction, indicative of a portion of cells going into mitotic exit (Fig. 5B). It rather seems that DELLAs influence endoreduplication onset by modulating APC/C activity, as DEL1/E2FE and UVI4 were significantly downregulated already 4 hours after DEX induction, preceding downregulation of mitotic transcripts such as CYCB1;1 and CDKB1;1, which occurred only 12 hours after induction (Fig. 5B). The transcription dynamics of DEL1 were next analyzed in more detail following DEX induction, showing that DEL1 repression already started 1 hour after GAI activation, and reached a significant level 2 hours after induction, suggesting that this is a primary target of DELLA signaling (Fig. 5C). DEL1 and UVI4 were also downregulated 24 hours after stress onset, and this persisted at 72 hours (Fig. 5A).

Overexpression of DEL1 Does Not Completely Abolish Early Endoreduplication on Stress
If DEL1 downregulation is causative of the observed early mitotic exit following mannitol treatment, then overexpression of DEL1 is expected to counteract this. We however still observed an early onset of endoreduplication when DEL1-overexpressing plants were transferred to mannitol, although this was not as pronounced as for the WT (Fig. 6). Possibly this is due to functional redundancy with UVI4. Overexpression of UVI4 is most likely lethal however (Heyman et al., 2011), so this could not be tested.

DELLAs Control Mitotic Exit Induced by Osmotic Stress
We previously demonstrated that proliferating tissues respond to stress by exiting mitosis early in favor of endoreduplication (Skirycz et al., 2011a). This limits the number of cells that form the organ, and therefore contributes to inhibition of growth by stress. Here, we showed that mild osmotic stress causes changes in GA metabolism specifically in proliferating leaf cells. What controls these changes in GA metabolism is currently unknown. In the shoot apical meristem KNOX I proteins regulate gibberellin levels and thereby proliferation (Hay and Tsiantis, 2009), but KNOX I transcripts (STM, BP, KNAT2 and KNAT6) do not show changes in response to mannitol, making it unlikely that they are involved here (Skirycz et al., 2011a). Another possibility involves CBF-related proteins, which were shown to modulate GA levels in response to cold stress (Achard et al., 2008a), but these also did not transcriptionally respond to osmotic stress (Skirycz et al., 2011a). The changes in GA metabolism then result in DELLA stabilisation, which is responsible for mitotic exit.
Consequently, ectopically stabilizing DELLAs, either through PAC addition or DEX induction of gai-GR, resulted in cell cycle inhibition and mitotic exit. However, mutants with lower DELLA activity (spy-3, q-ga2ox, and rga-28 gai-2) did not show relief of cell cycle inhibition by osmotic stress. This suggests that while DELLAs are able to inhibit cell proliferation, they are not a determining factor in osmotic stress-induced inhibition of cell division, as lower DELLA activity does not release this inhibition. This is consistent with the model we previously proposed, in which ethylene arrests the cell cycle rapidly, posttranscriptionally and reversibly (Skirycz et al., 2011a), and this mechanism is still active in these mutants. Our mutant analysis however confirms that the mitotic exit and early onset of endoreduplication induced by osmotic stress is regulated by DELLAs. Interestingly, both the mutants with elevated as well as those with downregulated DELLA activity lacked the stressinduced earlier onset of cellular differentiation, indicating that this process is controlled by a fine balance of GA and DELLA activity. This may not be so uncommon in hormone effects, as for instance a study on the role of an ethylene responsive factor in flg22-mediated growth inhibition revealed that both mutation and overexpression had the same effect, leading to the conclusion that it is maintained at an optimal level, and any deviation tips the fragile signaling balance (Bethke et al., 2009). Similarly, root meristem size is regulated by a fine balance of brassinosteroid signaling, and modulation of this pathway in either direction leads to short root phenotypes (González-García et al., 2011). It is also interesting to note that perturbations of DELLA levels sufficient to disturb the stress-induced cell differentiation generally do not affect ploidy levels under normal conditions.

DELLAs Control Cell Proliferation by Inducing Differentiation
Our data fit into a network of emerging evidence that DELLAs are important for the control of differentiation and mitotic cell numbers, thereby adapting cell production rates and thus organ growth. Cell production rates are calculated by expressing the net gain of cells relative to the total number of cells in the organ, and therefore integrate the fraction of cells that are dividing and their average division rate or cell cycle duration, both of which could potentially be modulated to change the cellular output of the organ. In roots the average cell cycle duration is constant, and changes in cell production rates are mostly due to changes in Our findings point to GA/DELLA-mediated control of mitotic exit in Arabidopsis leaves as well, at least under stress conditions. Furthermore, our finding that PAC and DELLA stabilization induce endoreduplication is in contrast to earlier observations based on GA addition experiments and mutant analysis showing that GA promotes endoreduplication in for instance pea (Mohamed and Bopp, 1980), Arabidopsis hypocotyls (Gendreau et al., 1999;Saibo et al., 2003) and tomato fruits (Serrani et al., 2007). This again suggests that effects of hormones depend greatly on concentration, tissue or cell type, species, and environmental parameters.

Exit
Previously DELLAs were reported to have the potential to control cell production rates in Arabidopsis by upregulating KRP2, SIM, SMR1 and SMR2, which block cell cycle by inhibiting CDKA activity (Achard et al., 2009). Our data rather suggest that DELLA-induced mitotic exit rather relies on the modulation of APC/C activity through downregulation of DEL1 and UVI4. These regulators likely act together, as overexpression of DEL1 alone was not enough to completely eliminate the early onset of endoreduplication under stress.

DELLA Stabilization Resulting in Mitotic Exit is an Important Response to Stress
We present a model in which DELLA stabilization in proliferating cells, a relatively late event following the onset of osmotic stress, would drive the cells away from the mitotic cell cycle, and into endoreduplication. As an early onset of endoreduplication is a natural response to stress, it is tempting to speculate that it contributes to stress tolerance. This is consistent with the finding that leaf growth of del1-1 mutants, which exhibit an early onset of endoreduplication, was found to be less sensitive to water deficit, while DEL1-overexpressing plants were more sensitive (Cookson et al., 2006). This shows that careful dissection of responses of growing tissues to stress allows the modulation of growth inhibition, holding great promise for yield stabilization in the coming decades.
Plates were overlaid with a nylon mesh (Prosep, Zaventem, Belgium) of 20-µm pore size to avoid growth of roots into the medium. Depending on the experiment, 32 or 64 seeds were equally distributed on a 150-mm diameter plate. Mutant plants were grown together with their wild-type controls on the same plate.

Stress and Chemical Treatments
At 9 DAS, when the third leaf is fully proliferating, seedlings were transferred to plates containing control medium or medium supplemented with 25 mM mannitol (Sigma-Aldrich, St-Louis, MO, USA), 1 µM paclobutrazol (Sigma-Aldrich) or 5 µM dexamethasone (Sigma-Aldrich) by gently lifting the nylon mesh with forceps. All transfers were performed 3 h into the day.

Growth Analysis
Growth analysis was performed on the third true leaf harvested at different time points after transfer. After clearing with 70% ethanol, leaves were mounted in lactic acid on microscopic slides. For each experiment, 8 to 12 leaves were photographed with a binocular, and epidermal cells (40 to 300) were drawn for four representative leaves with a DMLB microscope (Leica) fitted with a drawing tubus and a differential interference contrast objective. Photographs of leaves and drawings were used to measure leaf area and cell size, respectively, using ImageJ v1.41o (NIH; http://rsb.info.nih.gov/ij/), and from these cell numbers were calculated. RGR was calculated as the slope of a linear trend line fitted to lntransformed leaf area data.

Expression Analysis
The third leaf was harvested from plants at the indicated time points after transfer. The

Western Blot
Total soluble protein was extracted from 64 to 128 leaves by adding extraction buffer (Van Leene et al., 2007) to ground samples, followed by two freeze-thaw steps and two centrifugation steps (20,817 g; 10 min; 4°C) whereby the supernatant was collected each time. 50 µg of total soluble protein was used for protein gel blot analysis with either primary rabbit anti-GFP antibodies (Santa Cruz, CA, USA) (diluted 1:200) and a secondary horseradish peroxidase-conjugated donkey anti-rabbit antibody (GE-Healthcare) (diluted 1:10,000).

Proteins were detected by chemiluminescence (Western Lightning Plus ECL, PerkinElmer
Life Sciences, Boston, MA, USA). Protein amounts were quantified with ImageJ v1.41o.
Cross-reacting bands were used for normalization. Control samples were arbitrarily set at 100%.

Supplementary Data
The following materials are available in the online version of this article. Figure S1. Endoreduplication index (EI) of bottom (proliferating) and top (expanding) half of leaves exposed to osmotic stress.