Introduction

The growth-defense trade-off, characterized by the simultaneous restriction of growth and activation of defense mechanisms in response to environmental threats, is observed in diverse organisms, from microbes to plants and mammalian cells1,2,3. This trade-off is often attributed to competition for shared resources along the growth and defense pathways4. For example, yeast cells prioritize the production of defense proteins by repressing transcripts related to cell division during acute stress5.

Although plants possess cell-autonomous checkpoints for stress-dependent control of cell division that are equivalent to those described in microorganisms and mammalian cell cultures6, the growth-defense trade-off results in a systemic growth arrest. Recent studies challenge the view that the trade-off is caused by simple ‘metabolic competition’ and show that it results from a programmed interaction between cellular signaling pathways7,8. In a first study, the trade-off triggered upon pathogen attack was eliminated by simultaneously enhancing the jasmonate-dependent defense response and reducing the activity of the phytochrome B (phyB)-dependent growth suppression pathway. These plants maintained their robust growth and defense capability against herbivorous insects7. In a second study, the stimulation of a specific branch of brassinosteroid signaling resulted in increased tolerance to drought stress without loss of growth8. However, phyB and brassinosteroids regulate multiple aspects of plant physiology. Indeed, loss of phyB activity also resulted in reduced photosynthetic capacity, but the decrease in energy supply was compensated by a reduction in leaf thickness, and brassinosteroids are also known to stimulate plant growth, so a more precise method of uncoupling growth inhibition from defense responses would be required. Moreover, even if the trade-off is not caused by competition for resources, a pivotal question remains unanswered: does the genetically-driven redistribution of resources confer an advantage under stress conditions, or is there an alternative benefit to the observed trade-off? To address this question, we have targeted the DELLA pathway to generate plants that specifically overcome the mechanism of growth inhibition by stress and studied their performance under water limitation. Studies in Arabidopsis thaliana have shown that DELLAs play an important role in limiting growth under stress9,10. By eliminating DELLA activity, inflorescence growth can occur even under stress conditions. However, DELLA proteins also promote oxidative stress tolerance9. To separate the role of DELLAs in growth regulation from defense responses, we have mutated the CDK-inhibitors (CKIs) of the KIP-RELATED PROTEIN (KRP) and SIAMESE-RELATED (SMR) families, whose activation by DELLA inhibits cell division11,12. In doing so, we have observed a severe increase in the accumulation of DNA damage linked to cell divisions that could potentially be transmitted to their progeny.

Results and discussion

Loss of SMR1 suppresses DELLA-mediated growth restriction

SMR1 emerged as the most relevant DELLA target in the context of inflorescence development. Mutation of SMR1 alone was sufficient to promote stem elongation in the presence of paclobutrazol (PAC), an inhibitor of gibberellin (GA) biosynthesis that leads to DELLA accumulation (Supplementary Fig. 1). Remarkably, neither the smr1 mutation nor mutations in other CKI genes showed enhanced shoot growth in PAC-free conditions. This DELLA-dependent effect of smr1 aligns with observations from previous studies, where similar effects were reported for other CKI mutations in different contexts11. A plausible explanation could involve functional redundancy among CKI genes, coupled with DELLA-dependent induction of specific CKI family members. The relevance of SMR1 was confirmed genetically, as only smr1 suppressed the dwarfism of plants expressing non-degradable versions of the DELLA proteins GAI13 (Fig. 1a; Supplementary Fig. 2a, b) and RGA14 (Supplementary Fig. 2c, d). It is worth noting that gai-1 smr1 plants developed significantly taller inflorescences and produced three times more reproductive biomass and seeds than plants with the dominant gai-1 allele (Fig. 1a, b; Supplementary Fig. 3a-c).

Fig. 1: DELLA proteins directly activate SMR1 transcription to restrict reproductive growth.
figure 1

a Representative inflorescences of five weeks-old WT, gai-1 and gai-1 smr1 plants. b, c Seed production (b) and number of EdU-positive cells in the SAM region (c) in WT, smr1, gai-1 and gai-1 smr1 plants. d Relative quantity (RQ) of the SMR1 transcript in dissected 5 mm inflorescence tips. e pSMR1:GUS reporter activity in representative WT and gai-1 inflorescence tips. Arrowheads indicate the upper limit of the stained region. f Relative quantity of SCL3 (positive control) and SMR1 transcripts in WT seedlings and in seedlings expressing a non-degradable GAI protein under the control of a heat-shock inducible promoter (HS:gai-1). Seedlings were exposed to 37 °C for 30’ (HS + ) or left under growth conditions (22 °C, HS-). g ChIP-qPCR showing the binding of the gai-1-3xYPet protein to genomic regions upstream the SMR1 coding sequence (CDS). A fragment of the SCL3 promoter was amplified as a positive control. WT seedlings were used as negative controls. In (b) and (c) the dots stand for individual observations [n = 12 in (b); n = 9, 9, 7, 7 (left to right) in (c)] and the horizontal lines for mean values. In (d), (f) and (g), bars represent the mean value of three independent biological replicates (represented by dots). In (b), (c), (f) and (g) letters indicate statistical groups obtained using Tukey’s post hoc test [p < 0.05 for (c); p < 0.01 for (b) and (g); p < 0.001 for (f)] following one-way ANOVA. In (d), p value is for the equality of means (two-tailed t test). Scale bars, 5 cm in (a) and 2 mm in (e). Source data are provided as a Source Data file.

Full size image

Two observations confirmed that the differences in stem length were due to changes in cell division rates at the shoot apical meristem (SAM). First, gai-1 mutants had fewer cells labelled with 5-ethynyl-20-deoxyuridine (EdU) in the SAM region than the wild type (WT), which was restored by loss of SMR1 function (Fig. 1c; Supplementary Fig. 3d). And second, expression of the pCYCB1;2:GUS mitotic reporter was reduced in the apical zone of the inflorescence of gai-1, as previously reported11, but returned to WT levels by mutating SMR1 (Supplementary Fig. 3e, f). Importantly, the conclusion that SMR1 is a direct and functional target of DELLA activity is supported by the combination of three observations: the expansion of its expression domain in the inflorescence tips of the gai-1 mutant (Fig. 1d, e), the rapid induction of its expression following gai-1 transcriptional activation (Fig. 1f), and the significant binding of DELLA proteins GAI and RGA to SMR1 promoter regions, as shown by chromatin immunoprecipitation assays (Fig. 1g; Supplementary Fig. 4). Although stem elongation is still partially impaired in gai-1 smr1 plants, suggesting that additional factors mediate DELLA inhibition of reproductive growth, our results indicate that mutating SMR1 is an appropriate strategy to generate plants resistant to DELLA-mediated growth arrest, and that these plants can be used as a tool to test the hypothesis that growth inhibition is critical to guarantee optimal performance under abiotic stress.

Growth arrest is not required for optimal acclimation to water limitation

DELLA accumulation leads to increased tolerance to different types of abiotic stress, including drought15. Therefore, we investigated the physiological responses of WT, smr1, gai-1 and gai-1 smr1 plants to decreasing soil moisture. The correlation between leaf relative water content (LRWC, an indicator of plant water status) and soil water potential (SWP, ΨSoil, an indicator of soil moisture) was comparable between genotypes (Fig. 2a), demonstrating that all genotypes were being exposed to comparable stress intensities (see Materials and Methods for details). As expected, all plants responded to water limitation with a reduction in transpiration (Fig. 2b). However, transpiration rates normalized per rosette area were higher in gai-1 and gai-1 smr1 plants under stress (Fig. 2c), which could be explained by the higher stomatal density in gai-1 leaves16. Despite the higher water loss, the capacity of gai-1 plants to cope with water limitation in survival assays was better than that of WT plants (Fig. 2d, e). It is likely that this capacity is related to DELLA-enhanced production of antioxidants that buffer the toxic effect of reactive oxygen species (ROS) generated during extreme drought17,18,19. Supporting this idea, we observed an increased expression of genes involved in flavonol biosynthesis, accompanied by a higher accumulation of flavonols, in the rosette of gai-1 plants (Supplementary Fig. 5a-c). We also observed a reduction of oxidative damage in gai-1 root tips treated with methylviologen (Supplementary Fig. 5d-f).

Fig. 2: smr1 has no effect on the drought resistance of gai-1 mutants.
figure 2

Relative water content in rosette leaves (LRWC) (a), daily plant transpiration (b) and transpiration normalized to the rosette area (c) as a function of soil water potential (ψsoil) in WT, smr1, gai-1 and gai-1 smr1 plants. The dots denote individual plants (n ≥ 15) and the dashed lines represent the fitting functions of the data for each genotype. d, e Drought resistance of WT, smr1, gai-1 and gai-1 smr1. Four-weeks old plants were maintained at 15–20% LRWC for four days and survival rates were determined three days after rewatering. In (d), the dots represent independent assays (n = 10, 12, 15, 15 for WT; 10, 12, 13, 15 for smr1; 8, 12, 15, 15 for gai-1; 8, 16, 13, 15 for gai-1 smr1). The black horizontal lines represent the mean values. Letters indicate statistical groups obtained using Tukey’s post hoc test (p < 0.001) following one-way ANOVA. In (e), representative images after rewatering the plants are shown. In contrast to gai-1, the inflorescence of gai-1 smr1 plants had bolted and contained mature flowers and fruits by the end of the assay (see also Supplementary Fig. 6c). Scale bar, 10 cm. Source data are provided as a Source Data file.

Full size image

More importantly, the antioxidant protection and survival rate of gai-1 smr1 plants was the same as that of gai-1 plants (Fig. 2d, e; Supplementary Fig. 5), despite the differences in inflorescence growth (Supplementary Fig. 6a-c), and an equivalent behavior was observed when comparing rga-Δ17 smr1 and rga-Δ17 plants (Supplementary Fig. 6d). These data suggest that maintaining increased growth and biomass production does not compromise the proper acclimation and survival of plants exposed to drought stress.

Cell divisions aggravate genotoxicity

If the energy demands of the defense mechanisms do not justify the need to arrest growth, what is the reason for the conservation of this physiological response in plants over evolution? One possibility is that growth arrest is a tolerated consequence of a more valuable response: the DELLA-dependent stimulation of the response against oxidative stress9. However, this is unlikely because there are additional DELLA-independent mechanisms that promote stress-induced growth arrest20,21. A second possibility is that growth cessation due to cell cycle arrest is intrinsically beneficial under stress. In fact, DNA replication during the S phase of the cell cycle is particularly vulnerable to damage for several reasons. The formation of single-stranded DNA exposes it to oxidative stress and other damaging agents. Additionally, damaged bases can hinder the progression of the replication fork, further compromising the integrity of the genome22,23. In concert with this, roots incubated with zeocin, a radiomimetic drug that induces double-strand breaks in DNA, caused increased cell death that was more evident in the dividing cells of the meristem than in the elongation and differentiation zones (Fig. 3a; Supplementary Fig. 7; Supplementary Fig. 8), as previously reported22,24,25. This effect was further enhanced when cell divisions were stimulated by the presence of sucrose in the medium (Supplementary Fig. 7a-c; Supplementary Fig. 8b) or by della loss-of-function (Supplementary Fig. 7d-f; Supplementary Fig. 8c).

Fig. 3: Cell divisions overcome DELLA protection against genotoxic stress.
figure 3

a Representative confocal images of WT, krp4, gai-1 and gai krp4 root tips treated with 20 μg/ml zeocin for 24 h and stained with PI, which colours cell borders but penetrates dead cells. Seedlings were grown in sucrose-containing medium as indicated in Materials and Methods. The asterisks mark the position of the quiescent center and the arrowheads mark the RAM length. b, c Number of cells in the cortex layer (b) and quantification of cell death (c) in longitudinal sections through the center of the RAM of zeocin-treated seedlings. The dots represent individual values and the horizontal lines represent the mean value per genotype [n = 26, 26, 25, 25 (left to right) in (b); n = 26, 26, 25, 26 (left to right) in (c)]. d Orthogonal views of confocal image stacks of representative WT, smr1, gai-1 and gai-1 smr1 inflorescence apices treated with 100 μg/ml zeocin for 48 h and stained with PI. Dead cells were mainly detected in the L1-L3 cell layers of the SAM (marked with arrows). e Cell death volumes in the SAM region of zeocin-treated inflorescence tips. n = 12, 11, 17, 16 (left to right). Boxplots in raincloud plots show the median (center line), the first and third quartiles (lower and upper bounds, respectively), and the whiskers extend to the minimum and maximum values within the 1.5 times the inter-quartile range (IQR). In (b) and (c), letters denote statistical groups defined by Tukey’s post hoc test following one-way ANOVA [p < 0.05 for (b) and p < 0.01 for (c)]. In (e), p values correspond to one-tailed Mann-Whitney U tests. Scale bars, 50 μm. Source data are provided as a Source Data file.

Full size image

The necessity of DELLA for protection against genotoxic agents could be explained either by the prevention of DNA damage by reducing cell divisions or by DELLA-dependent promotion of the cellular DNA Damage Response (DDR) pathway26, for which there is currently no evidence apart from the observation that SUPRESSOR OF GAMMA RESPONSE 1 (SOG1), a major regulator of the DDR, and DELLAs share common transcriptional targets such as SMR111,27 (Supplementary Fig. 9). That the observed genotoxicity is related to cell division capacity was confirmed by analyzing zeocin-induced cell death in the roots of plants with mutant CKIs. Contrary to shoots, SMR1 is hardly relevant for cell divisions in roots (Supplementary Fig. 7i-l; Supplementary Fig. 8e-f), where KRP4 is a more important CKI28. Root meristems of krp4 and gai-1 krp4 mutants were larger than those of the corresponding controls (Fig. 3a, b), and the sensitivity to the cell-death inducing effect of zeocin was higher in gai-1 krp4 than in the gai-1 mutant (Fig. 3c; Supplementary Fig. 8a). Similarly, limiting cell divisions by the activity of DELLA in the SAM resulted in increased resistance to zeocin (Fig. 3d, e; Supplementary Fig. 10a-c), and this effect was reversed by smr1 (Fig. 3d, e; Supplementary Fig. 11), suggesting that protection of DELLA against genotoxic stress can be abolished by stimulating cell proliferation. The fact that ROS levels did not change significantly in SAM cells after treatment with zeocin (Supplementary Fig. 12) emphasizes that this protection must be linked to the control of cell divisions and not to an increased tolerance against oxidative stress. Interestingly, KRP4 and SMR1 act preferentially on G1/S and G2/M transitions, respectively29, but mutations in both genes cause the same effect on the tolerance to zeocin, indicating that impairing cell cycle progression is more critical than exerting it in one particular phase. Two observations provide additional evidence for the correlation between cell divisions and genotoxicity in the shoot apex. First, the analysis of a pCLV3:GFP(-ER) stem-cell niche reporter line showed that the early progeny of stem cells were more sensitive to the drug (Supplementary Fig. 10d). Second, the excessive proliferation of SAM cells in clavata1 (clv1) mutants with impaired stem cell homeostasis30 also resulted in larger regions of cell-death following zeocin exposure (Supplementary Fig. 10e-f). We conclude that limiting cell divisions protects against genotoxic stress in plant meristems.

Growth arrest under drought protects genome integrity

The first confirmation that cell divisions correlated with drought-induced damage came from the observation that the RAM of krp4 mutants accumulated a larger relative area of cell death compared to that of wild-type seedlings (Supplementary Fig. 13), so we decided to investigate the effect of drought in the SAM. It has been previously described that drought restricts overall plant growth and the SAM is arrested when the LRWC falls below 50%31,32. However, reproductive SAM activity is also modulated by a still unknown seed-derived signal in a process that leads to a global proliferative arrest (GPA)33,34. We have previously shown that GPA is delayed in both della gain- and loss-of-function mutants11, which share a reduction in seed number35. Thus, while mutants affected in DELLA are still a useful tool to alter meristem activity, the DELLA-dependent effect on fertility therefore overrides its direct role in the SAM11, and both gain- and loss-of-function mutants are expected to maintain a dividing SAM in fruit-bearing inflorescences under drought.

To explore whether growth arrest under drought stress reduces the negative impacts on the genome, we analyzed the inflorescence growth of plants under both control and drought conditions, focusing on ROS accumulation (as a potential cause) and cell death (as a consequence of DNA damage) in SAM cells. The stress-induced accumulation of ROS can damage diverse cellular structures, including the DNA36. Plants were watered normally until flowering, after which soil moisture was gradually reduced, lowering the LRWC to 40-50% over two weeks (Supplementary Fig. 14a). Water limitation restricted inflorescence growth across all genotypes studied and halted flower production in WT, smr1, and gai-1 smr1 plants (Supplementary Fig. 14b–d). In contrast, the SAM of dellaKO and gai-1 mutants remained active, as expected due to the delay in GPA caused by the reduced seed number35 (Fig. 1b), and continued to develop new flowers even at LRWC levels below 50% (Supplementary Fig. 14b). These results indicate that both developmental (seed) and environmental (stress) signals are required to induce SAM arrest. Importantly, the extended growth in dellaKO and gai-1 mutants under drought conditions resulted in higher ROS accumulation and the formation of cell death patches in the SAM region (Supplementary Fig. 15). ROS are key regulators of SAM homeostasis and may be influenced by seed signaling. For instance, hydrogen peroxide has been shown to promote stem cell differentiation37. Furthermore, dividing cells are closely associated with energy metabolism, which drives ROS production via mitochondrial activity38. Thus, SAM arrest not only protects DNA integrity in this context but also helps regulate ROS levels, which is crucial during stress.

To investigate whether cell death in the SAM region is linked to DNA damage, we analyzed shoot apices from WT and dellaKO plants (Fig. 4; Supplementary Fig. 16). As previously noted, the dellaKO SAM, characterized by faster cell-cycle progression (Supplementary Fig. 17), displayed larger and more frequent patches of cell death compared to WT under drought stress (Fig. 4b, c; Supplementary Fig. 16d; Supplementary Fig. 18). Furthermore, higher levels of DNA damage were detected in the more active dellaKO SAMs, as indicated by terminal deoxynucleotidyl transferase-mediated dUTP nick-end labelling (TUNEL)39 (Fig. 4d; Supplementary Fig. 16e). Although it remains unclear whether DNA breaks are the cause of cell death or a consequence of it, the larger extent of DNA-damaged cells relative to dead cells suggests that DNA damage may precede cell death in this context (Supplementary Fig. 19).

Fig. 4: SAM arrest protects DNA from drought stress.
figure 4

a Cumulative number of mature flowers and fruits in the main inflorescence of WT and dellaKO plants exposed to controlled water restriction after flowering (see Supplementary Fig. 16b for well-watered plants). The graph shows the mean (dots) and standard deviation (shading) of 15 plants per genotype. Asterisks indicate statistical differences, p < 0.01 (two-tailed t test). LRWC, relative water content in leaves (see Supplementary Fig. 16a for values). b PI staining to detect dead cells (marked with arrows) in the shoot apex of WT and dellaKO plants grown under control or drought conditions. Orthogonal views of confocal image stacks of representative shoot apices are shown. Cell death was mostly observed in the L1-L3 SAM cell layers of dellaKO mutants under drought stress. c Cell death quantification in the SAM region of WT and dellaKO plants grown under different watering regimes. Control, well-watered plants at a similar stage of development, determined by the number of mature flowers in the main inflorescence, as drought-stressed plants. Drought, plants with limited watering for the indicated number of days after flowering (daf). n = 3, 4, 3, 3, 6, 6 (left to right). Boxplots in raincloud plots show the median (center line), the first and third quartiles (lower and upper bounds, respectively), and the whiskers extend to the minimum and maximum values within the 1.5 times the inter-quartile range (IQR). Letters indicate statistical groups defined by Dunn’s post hoc test following multiple comparisons Kruskal-Wallis (p < 0.05). d TUNEL labelling to detect DNA breaks in the shoot apex of WT and dellaKO plants grown under well-watered (Control) or restricted watering (Drought) conditions for 20 days after flowering. Confocal maximum intensity projections are shown in the top panels (asterisks mark the SAM) and the corresponding longitudinal sections are shown in the bottom panels (dashed lines draw the SAM dome). TUNEL fluorescence was only detected in L1-L3 cell layers of the SAM of dellaKO mutants subjected to drought stress (white signal, marked with arrows). Scale bars, 50 μm. Source data are provided as a Source Data file.

Full size image

In summary, we have presented multiple lines of evidence demonstrating that growth and defense mechanisms are intricately linked to safeguard genome integrity in both root and shoot apical meristems (Fig. 5). However, it is crucial to recognize that the interplay between growth and stress protection is highly context-dependent. For instance, vegetative and reproductive SAMs respond differently to stress. Under water-limiting conditions, reproductive SAMs arrest growth, whereas severe stress in vegetative SAMs may sometimes trigger the drought escape response, promoting the transition to flowering in an effort to ensure reproductive success40,41. Similarly, the RAM of dark-grown seedlings exhibits a drought-stress-induced stimulation of cell division42, contrasting with the suppression of cell division observed in light-grown seedlings, as described here. Moreover, cell divisions in the reproductive SAM must integrate environmental signals with endogenous cues, such as age-dependent proliferative arrest, adding complexity to the mechanisms at play (Fig. 5). This multifaceted regulation makes it challenging to disentangle the specific molecular processes governing SAM behavior under stress.

Fig. 5: Model for the interplay between growth and defence pathways in root (RAM) and reproductive shoot (SAM) meristems.
figure 5

In plant meristems, cell divisions inherently expose DNA to increased risk from mutagenic agents, such as genotoxic compounds and reactive oxygen species (ROS). ROS levels naturally rise under environmental stress and as a by-product of the cellular metabolism, and play complex roles in regulating developmental processes including meristem activity38. Plants have developed distinct mechanisms to safeguard DNA integrity under stress, with strategies varying by organ. One DNA protective mechanism involves DELLA-mediated reduction of ROS levels in leaves and root meristems9. Another mechanism, as proposed in this study, reduces cell division rates, minimizing DNA exposure to mutagens. This reduction can occur via a DELLA-dependent pathway that upregulates cyclin-dependent kinase inhibitors (CKIs), or through DELLA-independent pathways, potentially involving the DNA damage-responsive transcription factor SOG1, which plays a role in the cellular response to stress-induced DNA lesions27. In the reproductive SAM, however, the DELLA influence is modulated by global proliferative arrest (GPA) signals originating from developing seeds. Consequently, fertile wild-type plants exhibit more robust drought-induced meristem arrest than the less fertile mutants with altered DELLA activity11,35, highlighting the role of reproductive signaling in fine-tuning meristem resilience and adaptation under stress.

Full size image

A reduction in growth rate has been shown to be critical for survival under stress in animal cells in culture43 and in yeast44, which is associated with a reallocation of resources towards functions of stress tolerance45. In contrast, we have shown that efficient defense can occur in plants even when this redistribution is limited, perhaps because plant growth is mainly governed by a small group of dividing stem cells at the meristems that give rise to vegetative and reproductive organs. Growth arrest specifically at the meristems appears to be a critical strategy for plant survival, as these cells are hypersensitive to environmental threats. In barley, drought conditions that minimally affect other organs inhibit the formation of new primordia by the SAM46. Likewise, exposure to low temperatures induces cell-cycle arrest in the SAM region and in the youngest leaf primordia, with minimal impact on DNA replication in older leaves47.

Additionally, in line with the role of the SAM in progeny generation, the inner layers constituting the germline are shielded from DNA damage by localized expression of genes involved in DDR and telomere maintenance in L2 and L3, but not in L148,49. The extensive DNA damage (Fig. 4d; Supplementary Fig. 16e; Supplementary Fig. 19) and cell death (Figs. 3d, e; 4b, c; Supplementary Fig. 10; Supplementary Fig. 15; Supplementary Fig. 16 d) observed in the SAM under stress is therefore expected to result in germline mutations. Stress has been shown to accelerate mutation rates in plants, which are efficiently transmitted to the progeny50,51. This suggests that while growth arrest under stress may be an evolutionarily ancient cell-autonomous response, in plants, it triggers a systemic arrest of the entire individual safeguarding not only its own genomic integrity, but potentially that of its offspring.

Methods

Plant material

The Arabidopsis thaliana Landsberg erecta (Ler) and Columbia (Col) accessions were used as wild-type controls. The following lines in the Ler background have been described elsewhere: dellaKO, gai-1, pRGA:GFP-rga-Δ17, pGAI:gai-1-3xYPet, smr1-1, krp2-3, krp4-2, clv1-4 and pCLV3:GFP(-ER)11,14,30,52,53,54,55,56,57. The following lines in the Col background were also described: pHS:gai-1, sim-1, smr2-1, krp1-1 and HyPer758,59,60,61,62. The sog1-1 mutant is in a hybrid Ler/Col background63. The double mutant lines gai-1 smr1-1, gai-1 krp2-3, gai-1 krp4-2 and pRGA:GFP-rga-Δ17 smr1-1 were generated by crossing. The pCYCB1;2:GUS reporter line, originally in the Col background64, was backcrossed three times with Ler before being introduced into gai-1 and gai-1 smr1-1 mutants. A detailed description of mutant and transgenic lines is given in Supplementary Data 1. Mutant combinations were identified by PCR with the primers listed in Supplementary Data 2.

Growth conditions

Arabidopsis seeds were sown on a mixture of sphagnum:perlite:vermiculite (2:1:1) or on 41 mm Jiffy peat plugs (Semillas Batlle S.A., Barcelona, Spain) for experiments that required tight control of watering. After stratification at 4 °C for 5–7 days, the plants were grown under controlled conditions of temperature (22 °C), photoperiod (16/8 h light/dark), and light intensity (100 μmol m−2 s−1). Transpiration and drought survival assays were performed at 23/18 °C day/night temperature, 8/16 h light/dark photoperiod, 100 μmol m−2 s−1 light intensity and 60–65% humidity. For the in vitro experiments, seeds were surface sterilized and sown on horizontal or vertical plates containing half-strength Murashige and Skoog Medium including Vitamins (½ MS, Duchefa) with 0.8% phytoagar, pH 5.7. The growth medium always included 1% sucrose, except where indicated. After stratification, the plates were placed in cabinets with a 16 h light photoperiod at 22 °C, unless otherwise stated.

Transpiration and drought survival assays

To determine transpiration rates, plants were grown on Jiffy peat plugs covered with plastic shells (pots) to minimize substrate evaporation. Four weeks-old plants were randomly divided into four groups of four plants each and differentially watered for ten days to achieve a range of soil water potentials from well-watered conditions to plant wilting (0 to −12 MPa)65,66. The pots were then weighed at 24 h intervals to account for transpiration water loss in the plants. Projected rosette areas were determined to normalize whole-plant transpiration per mm2. To monitor the relative leaf water content (LRWC) of individual plants, fully developed rosette leaves were cut, weighed (fresh weight) and placed in 2 ml plastic tubes filled with tap water. After 24 h at 4 °C, the leaves were weighed again (saturation weight) and the LRWC was calculated as a quotient between fresh weight and saturation weight. Finally, rosettes were harvested to quantify flavonols and analyse gene expression, as described below.

For the survival assays, four weeks-old plants growing on peat plugs were watered to reach the same low soil water content (20% volumetric water content), which corresponds to a point close to a sharp drop in the soil water potential66. Once the soil water content of all plants was homogenised, watering was stopped. Four days after water withdrawal, the plants were watered again until saturation and checked for survival three days later.

Generation of a pSMR1:GUS reporter line

The pSMR1:GUS construct contained the β-glucuronidase gene (GUS) flanked by 3845 bp and 3715 bp of the 5’ and 3’ genomic regions between SMR1 and the flanking genes. The 5’ region was amplified from genomic DNA of the Ler ecotype with the primers SMR1 5’ F and SMR1 5’ R (Supplementary Data 2). This fragment was cloned into the pDONR221 vector using Gateway BP recombination (Thermo Fisher Scientific) to generate pENTR-5’gSMR1. The 3’ region was obtained from Ler genomic DNA in two independent reactions. A 2.3 kb fragment from the SMR1 STOP codon to a HindIII restriction site 2264 bp downstream (3’ fragment 1, f1) was amplified with the primers SMR1 3’ f1 F and SMR1 3’ f1 R (Supplementary Data 2). A 1.4 kb fragment from the HindIII site to the downstream gene (3’ fragment 2, f2) was amplified with the primers SMR1 3’ f2 F and SMR1 3’ f2 R (Supplementary Data 2). After A-tailing, f1 and f2 were cloned into the pGEM-T Easy vector (Promega) to generate pGEMT-3’gSMR1-f1 and pGEMT-3’gSMR1-f2, respectively. Both genomic fragments were fused at the HindIII/PstI sites of pGEMT-3’gSMR1-f1 to produce pGEMT-3’gSMR1. To generate the pSMR1:GUS reporter construct, the 5’ genomic region of pENTR-5’gSMR1 was cloned upstream of the GUS gene into the pMpGWB104 binary vector67 using Gateway LR recombination (Thermo Fisher Scientific). The 3’ genomic region of pGEMT-3’gSMR1 was cloned into the resulting vector as AscI/SacI, replacing the nopaline synthase terminator. The pSMR1:GUS plasmid was introduced into WT plants of the Ler ecotype by floral dip68. A representative homozygous line carrying the transgene at a single locus was selected for analysis.

GUS staining

Shoot apices at a similar developmental stage (with one to three open flowers) were pre-fixed in 90% cold acetone for 20 min, briefly washed with staining solution without X-Gluc and vacuum-infiltrated in staining solution for 10 min. For pSMR1:GUS samples, the staining solution contained 50 mM sodium phosphate buffer (pH 7.2), 2 mM potassium ferrocyanide, 2 mM potassium ferricyanide, 0.2% Triton X-100 and 1 mM 5-bromo-4-chloro-3-indolylb-D-glucuronic acid (X-Gluc). For pCYCB1;2:GUS samples, with weaker GUS activity, the staining solution did not contain ferro/ferricyanide and the X-Gluc concentration was increased to 2 mM. After vacuum infiltration, the apices were incubated at 37 °C for 18 (pSMR1:GUS) or 48 h (pCYCB1;2:GUS). The samples were de-stained with increasing concentrations of ethanol (30, 50 and 70%) and stored in 70% ethanol at 4 °C. Images of representative apices were obtained using a Leica DMS1000 digital microscope.

Quantification of flavonols

Quercetin and kaempferol content were measured in the rosettes of the WT, smr1, gai-1 and gai-1 smr1 plants used for the transpiration assays. Five to six individuals were analyzed for each genotype. Flavonols were extracted in 75% acetonitrile/water supplemented with 1 ppm genistein. To release the aglicones, the extract was mixed with an equal volume of HCl 2 N and incubated for 2 h at 80 °C. 1 μl was injected into an Orbitrap Exploris 120 mass spectrometer coupled to a Vanquish UHPLC system (Thermo Fisher Scientific, Waltham, MA, USA). Reverse-phase ultraperformance liquid chromatography was performed using an Acquity PREMIER BEH C18 UPLC column (2.1 × 150 mm, 1.7 μm particle size) (Waters Corp., Mildford, MA, USA). The mobile phase consisted of 0.1% formic acid in water (phase A), and 0.1% formic acid in acetonitrile (phase B). The solvent gradient program was as follows: 0.5% phase B over the first 2 min, 0.5–30% phase B over 25 min, 30–100% phase B over 13 min, 2 min at 100% B, return to the initial 0.5% phase B over 1 min, and conditioning at 0.5% B for 2 min. The flow rate was 0.4 ml/min and the column temperature was set at 40 °C. Ionisation was performed with heated electrospray ionization (H-ESI) in negative mode. A standard curve was generated with authentic standards, using genistein as the internal standard. The TraceFinder software (Thermo Scientific, Waltham, MA, USA) was used for data processing. The data were normalized to the mean of the WT samples.

Expression analyses by RT-qPCR

To study SMR1 expression, inflorescence tips (5 mm in length) were cut from plants with about three open flowers and floral buds beyond stage 12 were removed69. To induce DNA damage, the dissected inflorescence tips were incubated for 48 h in sterile plastic boxes with ½ MS medium (mock samples) or ½ MS medium supplemented with 100 μg/ml zeocin. The medium also contained 300 μg/ml carbenicillin to minimize bacterial growth. Total RNA was extracted from approximately 100 mg of ground tissue using the NucleoSpin RNA Plant Kit (Macherey-Nagel) and treated with DNase I on the column, according to the manufacturer’s instructions. cDNA was prepared from 1.5 μg of total RNA using the PrimeScript 1st strand cDNA Synthesis Kit (Takara). PCR reactions were performed with 1 μl of a 1:4 dilution of cDNA in ddH2O, SYBR Premix Ex Taq II (Takara) and the primers listed in Supplementary Data 2; using a 7500 Fast Real-Time PCR System and the associated 7500/7500 software (Applied Biosystems). Relative expression levels were calculated according to the 2−ΔΔCt method70, using the At1g13320 (PDF2) gene as a reference. All experiments included three biological replicates, each measured as the mean of three technical replicates. The biological replicates included five to eight inflorescence tips. For the heat shock assays with WT and pHS:gai-1 plants, the biological replicates contained approximately 50 seedlings grown on plates for 10 days. These assays included SCL3 (At1g50420) as a positive control for DELLA activity71.

The expression levels of CHS and FLS1 were examined in the same plant material used for the quantification of flavonols. Total RNA extraction, cDNA preparation and PCR were performed as described above. Gene expression was normalized to the geometric mean of the PDF2 and TIP41 (At4g34270) standards. Five to six plants (biological replicates) were analyzed in triplicate per genotype.

Chromatin immunoprecipitation (ChIP-qPCR)

1-week-old seedlings of pGAI:gai-1-3xYPet, pRGA:GFP-rga-Δ17 and non-transgenic Ler WT lines grown under continuous light were harvested. In vitro double cross-linking with ethylene glycol bis(succinimidyl succinate) (EGS) and formaldehyde, chromatin extraction, chromatin sonication, chromatin immunoprecipitation with an anti-GFP antibody (5 μg of ab290 per sample; Abcam), and DNA purification were performed as previously described55. ChIP-qPCR reactions were carried out in the same manner as qPCR analyses using the primers listed in Supplementary Data 2. In addition to the SMR1 genomic regions, a fragment of the SCL3 promoter was amplified as a positive control for DELLA binding72. Enrichments were determined as the ratio of immunoprecipitated DNA to input DNA.

Chemical treatments and confocal imaging of root tips

Four days-old seedlings grown in vertical plates under our standard conditions were transferred to new plates containing fresh ½ MS medium (mock samples) or ½ MS medium containing 20 μg/ml zeocin (Invitrogen) or 3 μM methyl viologen dichloride hydrate (MV, Sigma-Aldrich). Treatments were performed for 24 h under the same growth conditions, unless the zeocin plates and corresponding mocks were covered with foil. Next, roots were mounted on slides with 10 μg/ml propidium iodide solution (PI, Sigma-Aldrich) to simultaneously stain cell borders and cell death regions. Confocal longitudinal stacks through the center of the primary root were acquired using a Zeiss AxioObserver 780 confocal microscope with a Plan-Apochromat 20x/0.8 M27 objective. The excitation wavelength of the laser was 561 nm, and the PI fluorescence was recorded in the range of 600-656 nm. The laser power and digital gain were individually adjusted to maximise signal intensity without saturation. The areas of cell death were quantified with Fiji ImageJ in the region between the quiescent center of the root and the elongation zone (defined at the cortex cell layer). A brightness threshold of 160-255 was set to delimit the areas of cell death. Root tips of seedlings expressing the HyPer7 probe for H2O2 were sequentially excited at 405 nm and 488 nm to detect the reduced and oxidized forms, respectively, and fluorescence emission was collected at the 508-535 nm range in both cases. Imaging settings were kept constant for all samples.

To simulate drought stress in vitro, four days-old seedlings grown on vertical plates were transferred to new plates with ½ MS medium (mock samples) or ½ MS supplemented with 25% poly(ethylene glycol) (Sigma-Aldrich) (PEG samples). The seedlings were grown under these conditions for a further six days. For DNA labelling with 5-ethynyl-2’-deoxyuridine (EdU), seedlings were incubated in liquid ½ MS supplemented with 10 μM EdU (Invitrogen) for 4 h. The seedlings were fixed in 4% (v/v) paraformaldehyde and 1% Triton X-100 in phosphate-buffered saline (PBS) (0.137 M NaCl, 0.05 M NaH2PO4, pH 7) for 1 h. Seedlings were then rinsed three times with PBS, incubated for 1 h in darkness in 100 mM Tris-HCl buffer pH 8.5 with 10 mM Alexa 488-azide (Invitrogen) (pre-staining solution), followed by a 30 min incubation in darkness in 100 mM Tris-HCl buffer pH 8.5, 10 mM Alexa 488-azide, 1 mM CuSO4 and 100 mM ascorbic acid (click-chemistry solution). The seedlings were rinsed three times with PBS before imaging. All steps were performed with gentle shaking at room temperature. Confocal stacks were obtained using a Leica Stellaris FALCON microscope with a HC PL APO CS2 20x/0.75 DRY objective. Excitation was performed at 498 nm with a white light laser and fluorescence was collected in the 510-530 nm range. PI-staining of PEG-treated root tips, confocal imaging and image analysis were performed as described above.

Chemical treatments and confocal imaging of inflorescence tips

For combined labelling with modified pseudo-Schiff-propidium iodide (mPS-PI) and EdU, shoot apices from plants with approximately three open flowers were cut and dissected to expose the SAM region. The dissected apices were cultured in sterile plastic boxes containing ½ MS medium for 22 h, transferred to new boxes containing ½ MS medium supplemented with 10 μM EdU (Invitrogen) and cultured for 2 h to achieve a recovery time of 24 h. Shoot apices were fixed, stained and processed for imaging as previously described73. Confocal z-stacks (0.5 μm step size) were acquired using a Leica Stellaris 8 FALCON confocal microscope with a HC PL APO CS2 40x/1.25 GLYC objective. Excitation was performed at 500 nm with a white light laser and the emission filters were set to 510-530 nm for EdU and 600–700 for PI. EdU-labelled cells in the SAM L1-L3 cell layers were counted manually using the Multi-point Tool of Fiji ImageJ.

To trigger DNA damage in shoot apices in vitro, dissected inflorescence tips from plants with approximately three open flowers were grown for 48 h in sterile plastic boxes containing ½ MS medium plus 300 μg/ml carbenicillin (mock) or supplemented with 100 μg/ml zeocin. To determine the presence of cell death, shoot apices were stained with 1 mg/ml PI for 5 min. Confocal z-stacks (0.5 μm) were imaged using a Leica Stellaris 8 confocal microscope with a Fluotar VISIR 25x/0.95 water dipping lense. Excitation was at 535 nm and PI fluorescence was recorded at 575–700 nm. For shoot apices expressing a pCLV3:GFP(-ER) reporter, excitation was set to 488 nm and an additional emission filter was set to 500–525 nm to collect GFP fluorescence. The laser power and digital gain were adjusted for each sample to achieve the highest signal intensity without saturation. 3D cell death regions in the SAM domain were selected using a brightness threshold of 220–255 and quantified with the Voxel Volume macro of Fiji72.

To detect cell death in the SAM region of plants exposed to drought stress, shoot apices from WT and dellaKO plants were dissected and immediately stained with PI. Image acquisition and analysis were performed as described above.

To monitor the accumulation of hydrogen peroxide (H2O2), inflorescence tips were immersed in staining buffer containing 25 μM DCFH-DA (Sigma-Aldrich), 10 mM K+-MES pH6.1, 50 mM KCl and 0.05% Silwett L-77. After a 2 min vacuum pulse, the samples were stained for 1 h in the dark at room temperature. Prior to imaging, the shoot apices were washed twice with staining buffer without DCFH-DA and carefully dissected to expose the SAM region. Confocal z-stacks (1 μm step size) were acquired using the Leica Stellaris 8 confocal microscope with the Fluotar VISIR 25x/0.95 water dipping lense. Excitation was at 480 nm and DCFH-DA fluorescence was recorded at the 515–550 nm range. Laser intensity and digital gain were kept constant for all images.

To check DNA fragmentation, TUNEL assays were performed using the In Situ Cell Death Detection Kit Fluorescein (Roche), following the manufacturer’s protocol with some modifications. Dissected inflorescence tips were fixed with 4% (v/v) paraformaldehyde and 0.1% Triton X-100 in PBS buffer under vacuum for 60 min. After washing with PBS, the samples were treated with a permeabilization solution [3% (v/v) Nonidet P40 (Sigma-Aldrich) and 10% DMSO in PBS] for 60 min at room temperature. After new washes with PBS, the samples were incubated with TUNEL reaction mixture for 1 h at 37 °C in a wet chamber. The shoot apices were washed again with PBS and water before imaging. Confocal z-stacks (1 μm step size) were imaged using a Zeiss AxioObserver 780 confocal microscope equipped with a C-Apochromat 40x/1.20 water dipping lense. Fluorescein was excited with a laser line of 488 nm, and emission was detected in the range of 515–565 nm. The imaging parameters were kept constant for all samples.

Statistics and reproducibility

For each experimental setup, relevant statistical information (group sizes, tests, parameters, etc.) is given in the corresponding figure caption. No statistical method was used to predetermine sample size. Sample sizes were selected based on field standards and our own experience in order to ensure reproducibility and obtain reliable results. No data were excluded from the analyses. For the transpiration assays, plants were randomly divided into four groups as indicated in the corresponding Methods section. To avoid position effects, the genotype distribution in growth chambers/cabinets was randomized for all experiments. The investigators were not blinded to allocation during experiments and outcome assessment.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.