Climatic temperature gradients constrain growth and fitness of plants. Plant species have preferred temperatures with daytime optima that vary substantially from 10 °C for the cool-preferring viola to 35 °C for the tropical kapok tree4. Temperatures substantially exceeding the optimum trigger acclimation responses that involve development and the heat-stress response (HSR)2. Several plant thermosensors that regulate temperature-induced developmental processes are known, such as Ca2+channels, transcriptional regulators and light receptors5,6.

In metazoans, fungi and plants, heat triggers the HSR through HSF1 transcription factors (TFs)1,7. After release from inhibitory molecular chaperones8, HSF1 induces expression of heat-shock proteins (HSPs) and initiates adaptation. In plants, the HSR involves several phytohormone pathways9,10,11,12,13. A. thaliana mutants that are deficient in biosynthesis or insensitive to the phytohormone abscisic acid (ABA) are compromised in basal and acquired thermotolerance14.

ABA is the key signal to regulate the water status of plants by adjusting transpiration and water resorption to the demand of atmospheric CO2 for photosynthesis15,16. Under water deficit, ABA levels increase and induce protective measures including stomatal closing, differential gene expression, downregulation of chlorophylls and biosynthesis of osmolytes such as proline. In natural environments, heat and drought occur often combined and are the cause of major crop losses17. Stimulating ABA signalling through ABA receptors can reduce transpiration while sustaining photosynthesis and growth, therefore offering an option for a more water-efficient agriculture18. However, reduced transpiration increases leaf temperature and may thereby acerbate thermal stress during a heat wave. On the other hand, ABA promotes thermotolerance, and enhanced ABA signalling might be beneficial for heat acclimation.

Identification of TWA1

To better understand the role of ABA in heat tolerance, A. thaliana lines that are hyper-responsive or insensitive towards ABA were examined for alterations in acquired thermotolerance induced by a priming heat exposure. Ectopic expression of ABA receptors RCAR6 and RCAR1 or deficiency in ABA co-receptors encoded by type 2C protein phosphatases (PP2Cs) results in ABA hypersensitivity18,19, while deficiency in multiple ABA receptors leads to ABA hyposensitivity20,21. Seedlings with such changes in the ABA response, including previously analysed ABA mutants14, were exposed to a priming period at 38 °C followed by a recovery phase and a subsequent variable period of heat stress at 45 °C (Extended Data Fig. 1a,b). Both, ABA-hypersensitive and ABA-insensitive lines were more susceptible to heat stress compared with the wild type (WT; Extended Data Fig. 1a,c,d). The ABA-hypersensitive lines had a heat-sensitive phenotype irrespective of light (Extended Data Fig. 1e). The findings argue for a dual role of ABA during heat stress, that is, initially promoting thermotolerance but eventually being counterproductive for the acclimation process. ABA receptors and PP2C co-receptors control fast and slow responses including ion channels and aquaporins by fast post-translational control and differential gene expression for long-term adjustments15,16. ABA hypersensitivity at the level of gene expression is conferred by deficiency in the C-terminal domain phosphatase-like (CPL) family members such as CPL322. We examined the thermotolerance of the cpl3-knockout mutant and four additional mutants identified in a screen for ABA hypersensitivity using an ABA-responsive luciferase (LUC) reporter line23 (Methods). All of these mutants were more heat sensitive than the WT line (Fig. 1a). One mutant line showed a strongly enhanced thermosensitivity with a half-maximum inhibitory period (IP50) of approximately 75 min at 45 °C, whereas the pp2c triple mutant and WT lines had IP50 values of 105 min and 155 min, respectively (Fig. 1b and Extended Data Fig. 1d). The gene locus was named THERMO-WITH ABA-RESPONSE 1 (TWA1). In the presence of exogenous ABA, the induction of ABA-responsive LUC reporters was enhanced in the twa1-1 background (Fig. 1c and Extended Data Fig. 2a–c), indicating hyperactivation of ABA signalling. Shifting the temperature from 17 °C to 27 °C induced a 3-fold increase in reporter expression in WT plants but caused a more than 100-fold increase in the twa1-1 seedlings (Fig. 1c). The twa1-1 mutant showed partial haploinsufficiency in a cross with the WT (Extended Data Fig. 2d) and the affected gene (At5g13590) was identified (Methods). The TWA1 locus encodes a predicted 130 kDa intrinsically disordered protein of unknown function with two potential ethylene-responsive-element-binding-factor-associated amphiphilic repression (EAR) motifs (LxLxL)24 (Fig. 1d). A point mutation in the twa1-1 allele terminates translation after 247 amino acids. The transfer DNA (T-DNA) knockout line twa1-2 with disrupted TWA1 in the first exon was heat sensitive, comparable to twa1-1 (Fig. 1b). Transfer of a genomic fragment encompassing the TWA1 gene complemented the thermosensitivity of both mutants (Fig. 1b). TWA1 is expressed throughout the plant25 (Extended Data Fig. 3a).

Fig. 1: The TWA1 locus.
figure 1

a, The A. thaliana twa1-1 mutant is conspicuously thermosensitive. Five-day-old seedlings of WT and mutants with ABA-hypersensitive gene regulation were exposed to 45 °C for 90 or 105 min or not after 90 min priming at 38 °C. Pictures were taken 5 days after heat stress. b, The acquired thermotolerance of WT, twa1-1 and twa1-2 plants, as well as mutants complemented with the TWA1 gene, that were assayed for root growth within 5 days after heat exposure as indicated. c, ABA-inducible pHB6:LUC expression in 5-day-old seedlings exposed to 1 µM ABA for 16 h at 17 °C or 27 °C. Left, photographs and light emission of WT and twa1-1 plants expressing the reporter. Right, reporter activity per seedling given in relative light units (RLU; n = 18). d, Prediction of intrinsically disordered regions ( with maximum score of 1, and a schematic of TWA1, showing the highly variable region (HVR, yellow) and two EAR domains (red) (top). TWA1 disruption in mutants is shown by arrowheads. Bottom, amino acid identity (red) of the HVR from A. thaliana (A.t.) and A. lyrata (A.l.). For b, data are mean ± s.d.; n = 6 per datapoint. Experimental and statistical significance details are provided in the Methods and as Source Data.

Source Data

Thermosensor function

The ABA-hypersensitive reporter expression in twa1 mutants (Fig. 1c and Extended Data Fig. 2) implied that TWA1 has a gene-regulatory role. Transient TWA1 expression in leaf protoplasts resulted in a dose-dependent downregulation of an ABA-responsive reporter (Fig. 2a). The inhibitory response required functional TWA1 and was moderate in WT protoplasts but more conspicuous in protoplasts of twa1-1 and twa1-2 plants. The TWA1-mediated inhibition occurred irrespective of exogenous ABA and was temperature dependent (Fig. 2b,c). There was no detectable inhibition by TWA1 at 20 °C, while increasing the incubation temperature from the standard 25 °C to 30 °C increased the repressive effect (Fig. 2b). Homologues of TWA1 are found in monocots and dicots (Extended Data Fig. 3b). The coding sequences of TWA1 homologues from the cold acclimated lyrate rock cress Arabidopsis lyrata and white mustard Sinapis alba, which naturally occurs in Mediterranean climates, were isolated and AlTWA1 and SaTWA1 were expressed in twa1-2 protoplasts (Fig. 2c). At 30 °C, they inhibited expression of the ABA-responsive reporter, consistent with a conserved function as orthologues. However, the temperature dependence differed with half-maximum inhibitory values (IT50) of approximately 20 °C, 26 °C and 30 °C for AlTWA1, TWA1 and SaTWA1, respectively. High variability between the A. thaliana and A. lyrata orthologues occurs in a stretch of 20 amino acids within an amino-terminal region of TWA1 named HVR (Fig. 1d). We reasoned that the HVR might contribute to thermal sensing. Exchanging the HVR for the AlTWA1 domain in TWA1 was sufficient to change the thermoresponse of the TWA1(AlHVR) chimera to mimic that of AlTWA1 (Fig. 2c). The truncated TWA1(ΔN) with deletion of amino acid residues 1–554 was inactive in inhibiting gene expression (Fig. 2c). The findings imply a thermosensory function of TWA1 involving the HVR. The presence of EAR motifs suggested that TWA1 might function as a transcriptional coregulator like the NOVEL INTERACTOR OF JAZ26. The search for TWA1-interacting proteins using the yeast-two-hybrid (Y2H) system and indexed open-reading-frame libraries27 resulted in the identification of the JAM2 TF (Extended Data Fig. 4). JAM2 is involved in signalling of the phytohormone jasmonate (JA) by antagonizing the JA-promoting action of the TF MYC2, both as a homo- and heterodimer28,29. The related TFs JAM1–AIB30, JAM3 and MYC2 did not interact with TWA1 in the yeast analysis (Extended Data Fig. 4). The specific JAM2 binding to TWA1 was corroborated in yeast using Förster resonance energy transfer–fluorescence lifetime imaging (FRET–FLIM). FRET analysis of fluorophores requires proximity in the nanometre range (<10 nm) and it reduces the fluorescence lifetime (FL) of the excited donor fluorophore. Changes in the fluorophore distance affect the FRET signal by a power of six31. We used the protein fluorophores mono enhanced GFP (GFP) and mCherry with GFP–TWA1 as donor and mCherry-tagged JAM2 or JAM3 as an acceptor. GFP–TWA1 is functional (Extended Data Fig. 5a) and showed a preferential nuclear localization in yeast (Fig. 2d,e). TWA1 abundance was not significantly affected by temperature (Extended Data Fig. 5b); however, temperature affected nuclear TWA1 crowding. At 30 °C and 35 °C, GFP–TWA1 accumulated in nuclear subdomains, which was not observed at 20 °C or for the truncated product of twa1-1 (twa1 product, Fig. 2d). mCherry-tagged JAM2 and JAM3 revealed a nuclear localization (Extended Data Fig. 6a). TWA1 physically interacted with JAM2 at 30 °C specifically within the nuclear subdomains, whereas JAM3 never detectably did (Extended Data Fig. 6b). JAM2 interaction was not observed with the twa1 product (Extended Data Fig. 6c) and not with TWA1 at 20 °C (Extended Data Fig. 6d) and therefore correlated with TWA1 aggregation in nuclear subdomains. In contrast to GFP–TWA1, GFP–TWA1(AlHVR) accumulated in nuclear subdomains already at 22 °C, consistent with a regulatory role of the HVR domain for TWA1 crowding (Fig. 2d). Intramolecular FRET–FLIM analysis of TWA1 in live yeast revealed proximity of the amino and carboxy termini at 17 °C (Fig. 2e). An increase in temperature moved the terminal domains of the sensor apart, as evidenced by the FL increase.

Fig. 2: The thermosensor function of TWA1.
figure 2

ac, TWA1 and TWA1 variants inhibit ABA-responsive pRD29B:LUC expression in leaf protoplasts. a, Protoplasts of WT, twa1-1 and twa1-2 plants were transfected with a TWA1or twa1-1 expression cassette (shown as µg DNA per 105 protoplasts), incubated for 16 h at 25 °C with 10 µM ABA. Relative light units were normalized to fluorescence units of a co-transfected control (RFU). The dashed line indicates expression of mutant twa1 product in twa1-2 protoplasts. b, TWA1 action at different exogenous ABA levels. c, The action of TWA1 and variants at different incubation temperatures with 0.3 µg effector DNA per 105 twa1-2 protoplasts. The variants included TWA1(ΔN), A. lyrata AlTWA1, mustard SaTWA1 and TWA1(AlHVR). d, Top, GFP–tagged TWA1 and twa1 product in yeast nuclei and temperature-dependent phase separation. Bottom, nuclear subdomain formation of TWA1(AlHVR) at 22 °C. Insets: 1.5-fold magnifications. Scale bar, 10 μm. See the ‘Statistics and reproducibility’ section of the Methods. e, Temperature-induced changes in the proximity of TWA1 amino- and carboxy-terminal domains as revealed by intramolecular FRET–FLIM analysis. Top, false-colour images of GFP FLs of mCherry–TWA1–GFP (TWA1) in yeast nuclei. Scale bar, 2 µm. Bottom, GFP FL of TWA1 and mCherry–GFP. n = 10–15 per datapoint. Statistical analysis was performed using Student’s t-tests; *****P = 10−5. For the box plots, the centre line shows the median, the box limits show the upper (25th percentile) and lower (75th percentile) quartiles, and the whiskers show 1.5× the interquartile range. f,g, The temperature dependence of JAM2 binding to TWA1 and variants including the twa1 product (f) and binding to orthologues (g) was analysed using a Y2H growth assay. The growth rate µ was calculated on the basis of cell density increase within 24 h. JAM2–JAM2 dimerization served as the control (Extended Data Fig. 7). For ac,f,g, n = 3 per datapoint. Statistical significance details are provided as Source Data.

Source Data

As protein–protein interaction in the Y2H32 system provides histidine-autotrophic growth, this read-out enabled examination of the temperature dependence of JAM2 binding to TWA1. At 20 °C, little to no growth occurred with the JAM2–TWA1 pair compared with the dimer-forming JAM2–JAM2 combination under the histidine-selective condition (Extended Data Fig. 7a). Increasing the incubation temperature to 30 °C resulted in histidine-autotrophic growth of JAM2–TWA1-expressing yeast, consistent with the physical interaction of both proteins at permissive temperatures. Thus, yeast growth rates at different temperatures provided a proxy for JAM2–TWA1 interaction, with JAM2 dimerization serving as a reference (Fig. 2f and Extended Data Fig. 7b). In the temperature interval of 20 °C to 25 °C, TWA1 switches from undetectable to close to maximum JAM2 interaction. JAM2 did not bind to the twa1 product, whereas it interacted with TWA1(ΔN), deleted for the first 554 amino acid residues, at all temperatures (Fig. 2f). In contrast to TWA1, AlTWA1 supported yeast growth at 20 °C, whereas SaTWA1 required temperatures of around 30 °C for optimal growth (Fig. 2g and Extended Data Fig. 7b). TWA1(AlHVR) and AlTWA1 were indistinguishable in this analysis (Fig. 2g), consistent with the results of the plant cell analysis of the thermal characteristics of both sensor proteins (Fig. 2c). The findings reveal a role of the amino-terminal HVR for the temperature-dependent regulation of JAM2 binding to the carboxy-terminal half of TWA1.

TWA1-mediated transcriptional regulation

The EAR domain provides an interaction surface for the co-repressors TPL and TPRs, which suppress transcription and interact with the Mediator complex24,33. Y2H analysis revealed binding of TPL/TPRs to TWA1 (Extended Data Fig. 4). The TWA1 inhibitory action on ABA-dependent gene expression required TPL/TPRs in A. thaliana protoplasts. TWA1 did not inhibit reporter expression in the triple tpltpr2tpr4 mutant (Fig. 3a). Expression of TPL or TPR2 recovered TWA1-mediated gene regulation in the triple mutant, confirming a function of the co-repressors in the TWA1-mediated downregulation (Fig. 3a). Similarly, protoplasts of the jam1jam2jam3 triple mutant were unresponsive to ectopic TWA1 expression unless JAM1 or JAM2 were co-expressed (Fig. 3b). JAMs and MYC2 are basic helix-loop-helix TFs that compete for binding to G-box cis-elements and related motifs28,29 that include the ABA-responsive regulatory element34. Chromatin immunoprecipitation revealed that JAM2 binds to the ABA-responsive RD29B promoter close to the transcription start site and this interaction depended on both ABA and an ABA-responsive regulatory element (Extended Data Fig. 8a,b).

Fig. 3: The molecular mechanism of TWA1-mediated transcriptional repression.
figure 3

a, TPL and TPR corepressors are required for TWA1-inhibited reporter expression. Protoplasts of tpltpr2tpr4 triple-mutant plants were transfected with TWA1, TPL and TPR2 effector DNA. b, JAM1 and JAM2 co-expression rescues the TWA1 response in protoplasts of jam1jam2jam3 plants. For a,b, ABA response analysis was performed as in Fig. 2b with 3 µg effector DNA per 105 protoplasts. c,d, Accumulation and binding of GFP–TWA1 to JAM2 and TPL in nuclear subdomains of N. benthamiana epidermal cells. GFP–TWA1 and mCherry-tagged JAM2 or TPL were transiently expressed in leaves exposed to 37 °C for 2 h or kept at 20 °C. c, GFP imaging in the nucleus (see the ‘Statistics and reproducibility’ section in the Methods). Scale bar, 5 μm; insets are 1.5-fold magnifications. d, FRET–FLIM analysis. Statistical analysis was performed using two-sided Mann–Whitney U-tests; ***P = 0.005. n = 20–30 cells. Box plots are as described in Fig. 2e. e, Simplified model of repressor complex formation. Temperature transitions from 20 °C to 30 °C induce conformational changes in TWA1 that allow JAM2 binding to the carboxy-terminal part of TWA1 and TPL binding to the amino-terminal domain through EAR motifs (red). The HVR (yellow) is integral to thermosensing. JAM2 interacts with G-box-related cis-elements as dimers. TPL is tetrameric and targets subunits of the Mediator complex33. Statistical significance details are provided as Source Data.

Source Data

The physical interaction of TWA1 with TPL and JAM2 was confirmed using FRET–FLIM in Nicotiana benthamiana. Transient expression of GFP-tagged TWA1 in epidermal cells showed a nuclear localization of TWA1 that accumulated in subdomains (Fig. 3c), similar to the yeast analysis (Fig. 2d). FRET–FLIM supported binding of TPL to TWA1 and corroborated the TWA1–JAM2 complex formation at 37 °C (Fig. 3d). Both interactions occurred specifically in these nuclear subdomains but not at 20 °C (Fig. 3d). The data support the model in which elevated temperatures induce intramolecular rearrangements of TWA1 that allow interaction with JAM2 and TPL for assembly of the transcriptional repressor complex (Fig. 3e) in nuclear subdomains.

Thermotolerance conferred by TWA1

In agreement with the functional interactions of TWA1 with TPL/TPR and JAM proteins, seedlings of the jam1jam2jam3 and tpltpr2tpr4 triple mutants showed a compromised acquired thermotolerance like twa1-1 and twa1-2 plants (Fig. 4a). Ectopic TWA1 expression under the viral 35S promoter resulted in 15- and 18-fold increased TWA1 transcript levels in the twa1-2 mutant compared with in the WT in two representative TWA1-overexpression (TWA1(oe)) lines (Extended Data Fig. 9a), which exhibited higher acquired (Fig. 4a and Extended Data Fig. 9b) and basal (Fig. 4b) thermotolerance. The thermotolerance conferred by TWA1 expression depended on JAMs, as shown by the thermosensitivity of TWA1(oe) lines in the jam1jam2jam3 genotype (Fig. 4a). A. thaliana seedlings that were exposed to 37 °C responded with chlorophyll loss that was enhanced in twa1 mutants and ameliorated by ectopic TWA1 expression compared with the WT (Extended Data Fig. 9c). The improved basal thermotolerance of the TWA1(oe) lines was also evident in the reduced ion leakage of 3-week-old plantlets that were exposed for 24 h to 37 °C (Fig. 4c). Notably, the TWA1(oe) lines cultivated at ambient conditions were indistinguishable from the WT in growth, photosynthesis and gas-exchange parameters (Fig. 4d and Extended Data Fig. 9d,e).

Fig. 4: Thermotolerance and heat shock transcripts are controlled by TWA1.
figure 4

a, Acquired thermotolerance depends on TPL/TPR-type corepressors and on JAM TFs with loss of thermotolerance conferred by TWA1 expression (TWA1(oe) lines) in jam1jam2jam3 (jam) plants. TWA1 was expressed under the viral 35S promoter in twa1-2 plants. tpr signifies the tpltpr2tpr4 genotype. Photographs were taken 1 day after heat stress (150 min, 45 °C) or without heat stress. Scale bar, 1 cm. b, The basal thermotolerance of 5-day-old seedlings. Root growth analysis as in Fig. 1b but without priming. c, Ion leakage of 3-week-old plantlets that were exposed to 37 °C for 24 h or kept at 22 °C. Ion leakage was expressed relative to maximum leakage by boiling. Data are mean ± s.d. n = 12 from three repetitions. Scale bar, 1 cm. d, Improved thermotolerance of TWA1(oe) lines without trade-offs in growth or changes in apparent photosynthetic parameters (Extended Data Fig. 9d,e). Photograph of plants grown for 38 days under short-day conditions at 22 °C and thermal images in false colours. Scale bars, 3 cm. e, Temperature- and TWA1-dependent induction of the HSFA2 promoter driving LUC expression in leaf protoplasts of different lines. Analysis was performed as in Fig. 3a. Expression of AlTWA1 and SaTWA1 in twa1-2 plants. Data are mean ± s.d. n = 3 per datapoint. f, Transcript changes in 5-day-old WT and twa1 seedlings grown at 20 °C and exposed to 37 °C for up to 20 min. The transcript level at t = 0 was set to 1. Data are mean ± s.d. n = 3, 10 seedlings per n. Statistical analysis details for b,c,e are provided as Source Data.

Source Data

The enhanced thermotolerance by TWA1 expression might be caused by a preinduced HSR due to constitutive upregulation of HSP transcripts35,36,37,38,39,40,41. We examined the possibility. In WT seedlings, a temperature shift to priming periods at 37 °C induced transcripts for small HSPs and the major ATP-dependent chaperone HSP70 as part of the HSR (Extended Data Fig. 10a). At ambient temperatures, the levels of these transcripts were similar to the WT in TWA1-deficient and TWA1(oe) lines, therefore not supporting preinduction of HSR transcripts in the TWA1(oe) lines (Extended Data Fig. 10b,c). A 4 h priming phase increased HSP transcript levels in the TWA1(oe) lines up to 30-fold higher for HSP26.5 compared with the WT, while  the increases were marginally reduced in both twa1 mutants (Extended Data Fig. 10b,c). Given the IT50 value of TWA1 around 26 °C, the induction of early HSR transcripts was monitored at lower temperatures. First, we used the activation of the HSFA2 promoter as a readout. The HSFA2 promoter is directly targeted by the master HSF1-class TFs, HSFA1s in plants, and the HSFA2 induction is HSFA1 dependent and sufficient for establishing heat tolerance36,37. In WT protoplasts, a shift from 20 °C to 25 °C increased HSFA2-promoter-driven LUC expression with optimal values at 30 °C and 35 °C (Fig. 4e).

The induction did not occur in twa1 protoplasts and was enhanced in TWA1(oe) cells (Fig. 4e). A thermal response shift to lower and higher temperatures occurred in protoplasts from AlTWA1(oe1) and SaTWA1(oe1) plants expressing the lyrate and mustard orthologue in the twa1-2 genotype compared with protoplasts of the WT and TWA1(oe) plants (Fig. 4e). The TWA1 dependence of the response was corroborated by transcript analysis of seedlings that were exposed to a 1 h temperature shift from 20 °C to 25 °C (Extended Data Fig. 10d). Compared with the WT, the transcript increase for HSFA2, HSP21 and HSP26.5 was supressed in twa1-1, twa1-2 and SaTWA1(oe) seedlings, while it was enhanced in TWA1(oe) and AlTWA1-expressing seedlings (Extended Data Fig. 10d). At 35 °C, the transcript levels were comparable in SaTWA1(oe), TWA1(oe) and AlTWA1(oe) seedlings and consistently higher than in the WT (Extended Data Fig. 10e–g). To further elucidate the role of TWA1, WT and TWA1-deficient A. thaliana seedlings were exposed to 5 up to 20 minutes of heat stress (Fig. 4f). The transcript levels of HSFA2, HSP17.6, HSP21 and HSP26.5 were clearly induced in both WT lines but did not change in both twa1 mutants, in agreement with a positive regulatory function of the thermosensor for the upregulation of HSR transcripts.

In summary, our study reveals a thermosensor function of TWA1 that is required for heat tolerance and timely induction of HSFA2 and HSP transcripts in A. thaliana. The HSR is triggered by activation of key HSF1-type TFs. In A. thaliana, three related HSFA1s redundantly initiate the HSR38. The combined inactivity of two of those master TFs still allows HSFA2 induction and thermotolerance to heat-induced ion leakage39,41. Both responses are impaired in plants with TWA1 deficiency, supporting an action of the thermosensor upstream of HSFA1s or, alternatively, in parallel. Ectopic expression of constitutively active HSFA1 TFs39,40 and of TWA1 provide improved thermotolerance. Constitutive heat tolerance by HSFA1 is associated with preinduced HSR and growth penalty39, traits that are also conferred by ectopic HSFA2 and HSFA3 expression35,37. However, neither adverse physiological effect was observed in TWA1-mediated heat tolerance. Moreover, the induction of HSR transcripts within minutes required a functional TWA1. The findings point to an inducible protective function of TWA1 in early signalling in response to temperature rise.

Thermal activation of TWA1 involves the HVR and temperature-induced domain rearrangements that control the binding of JAM2 and repressor proteins. The role of TWA1 bears parallels to the thermosensor EARLY FLOWERING 3 (ELF3)6. ELF3 is a component of the evening complex, which functions as a transcriptional co-repressor in the control of the circadian clock and flowering time in A. thaliana. Thermally, TWA1 acts in the opposite manner to ELF3 by repressing gene expression at elevated temperatures, and it responds more sensitively to temperature changes. The TWA1 response has an apparent Q10 value of around 150, that is, a 10 °C difference can induce a 150-fold change in activity as deduced from growth rates in Y2H analyses (Methods). This responsiveness of TWA1 is in the range of the most thermosensitive proteins known, including members of the TRPA1 Ca2+channel of snake42 and mosquitos43 that are used for thermal prey localization. In eukaryotes, heat stress induces cytosolic and organellar stress granules pioneered in plants44 and the formation of nuclear subdomains that are reversible aggregates of proteins and RNA for cellular adaptation to thermal challenges8,45,46,47. Liquid–liquid phase separation (LLPS) has a prominent role in the formation of multilayered nuclear subdomains as sites of active gene expression48. After thermal activation, TWA1 accumulated in nuclear subdomains that are reminiscent of LLPS. These subdomains are the sites for TWA1 interaction with JAM2 and TPL, indicating the formation of active repressor complexes specifically in the presumed LLPS. Proteins with intrinsically disordered regions have a propensity for LLPS49 and include plant proteins with a prion-like domain50,51. Under heat stress, the intrinsically disordered region of the yeast HSF1 promotes condensate formation and multiple chromatin interactions for transcription within a single LLPS52. The temperature-sensing mechanism leading to HSF1 activation in yeast and mammals is still not clarified, while there is mounting evidence for the chaperone HSF1 titration model8.

TWA1-type sensors might be unique to plants. TWA1 sensed sub-heat temperature elevations between 20 °C and 30 °C and was required for the induction of early HSR transcripts. However, after a 4 h priming phase, HSP transcript levels were comparable between the WT and twa1 mutants, implying additional modes of transcript upregulation. In plants, heat-triggered increases in hydrogen peroxide53,54 and nitrogen oxide55 activate HSFA1s and allow for redox signal integration into the HSR. TWA1 possibly provides another node of signal integration. Plants frequently experience combined stress with the challenge to integrate conflicting single stress responses56. For example, the ABA-mediated proline increase under water deficit provides protective osmolytes, but high proline levels are toxic under heat57. TWA1 probably orchestrates the acclimation responses by integrating temperature with ABA and JA signalling58 that relay abiotic and biotic cues, respectively. In the future, using TWA1-type temperature sensors with different thermal characteristics through breeding and biotechnology will enable the adjustment of the acclimation responses of crops to a warming climate59. Moreover, TWA1 and its orthologues provide prototypes for engineering temperature switches in the emerging field of thermogenetics60.


Plant materials and growth conditions

Plants of A. thaliana Heynh. ecotype Col and La-er were grown in pots on a perlite/soil mixture at 22 °C under long-day conditions with 16 h light (150 µE m−2 s−1) unless otherwise stated. The plants were used for stable transformation, protoplast preparation and DNA extraction. A. thaliana seedlings were grown on agar plates with half-strength Murashige and Skoog agar medium for physiological assays as previously described61. For thermotolerance tests, growth conditions were as reported previously14. T-DNA knockout lines of the GABI-Kat collection62, including GK-476H03 (At5g13590; twa1-2) and SALK collection63 including SALK_143411 (At2g33540; designated cpl3-10) were obtained from The European Arabidopsis Stock Centre.

The JAM2/bHLH013 (At1g01260) gene was inactivated in the pHB6:LUC line using the CRISPR–Cas9 system64,65 and vector pDGE63, (Addgene plasmid, 79445). Two independent jam2 isolates contained a frameshift, which resulted in early termination of translation after amino acid 12 and we designated them as jam2-2. DNA-sequencing analysis revealed the absence of off-target mutations in the closely related genes JAM1/bHLH017 (At2g46510) and JAM3/bHLH003 (At4g16430). The triple mutant jam1-2jam2-2jam3-2 (jam) was generated by crossing jam2-2, jam1-2 (GK_285E09) and jam3-2 (GK_301G05)28 plants and selecting for the homozygous jam mutant in the offspring. Similarly, the tpl-8tpr2-2 tpr4-2 triple mutant (tpr) was obtained by crossing tpl-8 (SALK_036566) with SALK_112730 (tpr2-2) and SALK_002209 (designated tpr4-2) plants. The generation of RCAR1- and RCAR6-overexpressing lines has been reported previously61. Quadruple RCAR-knockout lines were obtained by combining the multiple PYR/PYL knockout line pyr1pyl1pyl2 pyl4pyl5pyl866, the rcar9 mutant67 and lines SALK_083621 (rcar1) and GK-012D02 (rcar13). The triple mutants abi1-2abi2-2 hab1-1 (pp2c_a) and abi1-2hab1-1pp2ca-1 (pp2c_b)19 were a gift from P. L. Rodriguez together with the multiple PYR/PYL knockout. Mutants were crossed to the ABA reporter lines pRD29B:LUC and pHB6:LUC23. Lines homozygous for reporter constructs and for mutant alleles were used throughout the experiments.

Mutant isolation and ABA exposure

Mutants with ABA-hypersensitive reporter activation were recovered from a screen of the maternal second generation (M2) of ethyl methanesulfonate-mutagenized A. thaliana seeds of the ABA reporter line pHB6:LUC23 in the Col-0 background. In brief, the M2 seedlings were grown on solidified half-strength MS medium for 5 days before transfer to medium supplemented with 3 µM (+)-cis-trans-ABA (Chemos; or 0.3 M mannitol for 24 h followed by life luciferase imaging with the intensity of light emission from plant organs of interest being measured using Simple PCI 6.6 software ( Among the mutants recovered, certain mutants were hypersensitive to exogenously applied ABA, while mutants with an unaltered response to ABA were considered to carry lesions in drought stress signalling upstream of the canonical ABA signalling pathway. About 115,000 seedlings were screened and 109 putatively hypersensitive mutants were selected. Candidates showing a hypersensitive reporter response were propagated and the progeny was re-examined. Finally, 24 mutants were confirmed, among which 16 were found to be ABA-hypersensitive and 8 were found to be affected in drought stress signalling upstream of ABA. We selected mutants with a robust phenotype for map-based cloning in combination with next-generation sequencing and identified twa1-1 together with mutants allelic to cpl1 and cpl3. The twa1-1 mutant was backcrossed to the pHB6:LUC reporter line four times.

Gene identification

The TWA1 locus has been identified by bulked segregant analysis68. In brief, approximately 50 homozygous mutant seedlings were pooled and processed for next-generation DNA-sequencing analysis to identify single-nucleotide polymorphisms (SNPs) compared with the reference genome Col-0. The analysis confined the location of the target gene within a 180 kb genomic fragment on chromosome 5. In this fragment, four SNPs were found within genes, two of them were synonymous. The non-synonymous mutations generated a premature TGA stop codon in At5g13590 at nucleotide 744 of the coding sequence and the other caused a conservative amino acid exchange in At5g13930. The identity of TWA1 (At5g13590) was confirmed by complementation of the ABA-hypersensitive phenotype by gene transfer of a 7 kb genomic fragment encompassing 1.1 kb of the promoter region, the structural gene and the terminator (a list of the primers and restriction sites used for cloning is shown in Supplementary Table 1).

Effector constructs and analysis of gene expression

Plant RNA extraction, cDNA synthesis and construction of plasmids for effector expression were performed as described previously61. In brief, cDNA was generated from mRNA isolated from leaves of A. thaliana Col-0, A. lyrata and S. alba. Total RNA was purified from leaves using the analytik jena-innuPREP Plant RNA Kit (Analytik Jena) and cDNA was synthesized using the RevertAid First Strand cDNA Synthesis Kit (Thermo Fisher Scientific). The coding sequences of effectors were integrated into a modified Bluescript vector with an expression cassette consisting of the 35S promoter, followed by the coding sequence and the NOS terminator61. A list of the primers and restriction sites used for cloning is provided in Supplementary Table 1. The 363 bp AvrII–HindIII DNA fragment of the AlTWA1 gene encompassing the HVR was exchanged with the AvrII–HindIII fragment of TWA1 to yield TWA1(AlHVR). Constructs used for FRET–FLIM analysis were cloned using the Golden-Gate system69. TWA1, twa1-1, JAM2 and JAM3 cDNAs were cloned into the level I vector (LI_BpiI) and subsequently into LII expression vectors (JAM2 and JAM3: LII_3-4_CEN_LEU with promoter pTDH3 and terminator tDH1, plus an N-terminal mCherry tag; TWA1and twa1: LII_1-2_CEN_LEU with pTDH3 and tDH1, plus an N-terminal GFP tag). For intramolecular FRET, TWA1 was cloned into LII_3-4_CEN_LEU with pTDH3 and tDH1, including an N-terminal mCherry tag and a C-terminal GFP (GFP used in all constructs was mono enhanced GFP) tag linked to TWA1 through a glycine–serine linker (mCherry: GGGGSGGGGSGGGGSG; GFP: GGGGGSGGGGSGGGGS). As a control, both fluorophores were fused with the same glycine–serine linkers and expressed using the LII_3-4_LEU expression cassette. The GFP-TWA1(AlHVR) construct was cloned into pGREG57470 using the TWA1(AlHVR) cassette (see above) and SalI. The reporter construct pHSFA2:LUC was generated by replacing the RD29B promoter in pRD29B:LUC with an amplified 2 kb fragment of the HSFA2 promoter region as described previously54. The correctness of all constructs was verified by DNA-sequencing analysis. Quantitative gene expression was monitored with BRYT Green Dye-based qPCR (GoTaq qPCR Master Mix kit, Promega) using the LightCycler 480 instrument (Roche) and gene-specific primers. TIP41L, UBC9 and UBI10 were used for normalization. A list of the primers used for quantitative PCR with reverse transcription is shown in Supplementary Table 2.

Protein sequence alignments

Homologues of TWA1 were identified through BLASTp searches against genomes on NCBI ( Amino acid sequences of TWA1 and representative proteins were aligned with the NCBI COBALT alignment tool71.

Prediction of intrinsically disordered proteins

Prediction of intrinsically disordered regions was performed using metapredict (

Protoplast assays and transgenic plants

Transient expression analysis in A. thaliana protoplasts, and expression cassettes for pRDB29:LUC, p35S:GUS and effectors have been described previously73. Primers for DNA amplification and restriction sites used for generation of effector constructs are listed (Supplementary Table 1). Protoplasts were isolated from leaves of Columbia WT accession (Col-0) or from mutant lines of the Col-0 background. In brief, 105 protoplasts were transfected with DNA of different expression cassettes including the ABA-responsive pRD29B:LUC reporter (5 µg), the p35S:GUS control reporter (3 µg) for expression normalization and the indicated amounts of various effector plasmids and incubated at 25 °C, unless otherwise stated, for 16 h before assessment of luciferase and glucuronidase activity. Ectopic expression of TWA1 and twa1 (mutant allele of twa1-1), AlTWA1 and SaTWA1 in A. thaliana plants was under the 35S promoter in the twa1-2 background. The cDNA expression cassettes of the TWA1 variants and orthologues were inserted as an AscI DNA fragment into the pGreenII 0179 vector74 modified with an AscI cloning site in the T-DNA region. Transgenic plants were generated by Agrobacterium tumefaciens-mediated gene transfer as described previously75.

Thermotolerance evaluation

Thermotolerance was analysed as reported previously14. In brief, 5-day-old light-grown seedlings (22 °C) were exposed to heat stress at 45 °C for 90 min to 180 min with 15 min intervals for assessment of basal thermotolerance. For analysis of acquired thermotolerance, seedlings had a preceding 90 min acclimation period at 38 °C followed by a 120 min recovery phase (22 °C) before heat stress at 45 °C. For root growth assays, seedlings were allowed to recover for 5 days under continuous light (22 °C) and the root extension after heat stress was determined.

Determination of chlorophyll content and of electrolyte leakage

For chlorophyll analysis, the seedlings were frozen in liquid nitrogen after their fresh weight had been determined, homogenized using the TissueLyser II (Qiagen) and extracted with methanol for determination of chlorophyll as reported previously76. Absorption of cleared extracts was recorded at 665 nm and 652 nm. In the electrolyte-leakage assay, leaves of 3-week-old A. thaliana plants were placed into Petri dishes containing 25 ml water and were incubated at 37 °C or 22 °C in the light (65 µmol m−2 s−1) for 24 h. Thereafter, single leaves were immersed in 1 ml of double-distilled water in Eppendorf tubes and incubated for 30 min with gentle shaking9. Conductivity was measured using a Conductivity Meter (Seven Easy Mettler Toledo, InLab 752-6 mm conductivity sensor). Electrolyte leakage was normalized to the conductivity after heating the samples to 99 °C for 10 min.

Chromatin immunoprecipitation experiments

GFP–JAM2 was expressed in A. thaliana Col-0 protoplasts that were incubated for 16 h in the absence or presence of ABA (10 µM). Protoplasts transfected with an empty vector served as the control. Protoplasts were collected by centrifugation (500g, 2 min) and the supernatant was discarded. Subsequent steps (nucleus isolation, shearing of chromatin, preclearing, immunoprecipitation, reverse cross-linking and DNA purification) were performed as described previously77. Anti-GFP antibodies were obtained from Santa Cruz Biotechnology (1:2,000 dilution was used for antibodies) and ChromoTek GFP-Trap Agarose was obtained from Proteintech. The enrichment of different DNA fragments encompassing the RD29B promoter in the antibody containing fraction (as compared to the control without antibody) was quantified using quantitative PCR from immunoprecipitated samples.

Photosynthetic parameters and thermoimaging

Gas-exchange measurements and thermoimaging were performed as described previously18. For quantifying net photosynthesis (An), transpiration (E) and stomatal conductance (gs) of the whole rosette, the GFS-3000 gas-exchange system was equipped with custom-built whole plant cuvettes (Heinz Walz). The analyses were conducted at 150 µmol m−2 s−1 PAR, 400 µmol mol−1 external CO2 and a water vapour deficit of 1.3 ± 0.1 kPa using the software of the instrument supplier. PAM imaging was performed by using a MAXI version of IMAGING PAM (Heinz Walz). The operation of the PAM imaging system was performed according to the manufacturer’s instructions. In brief, plants were dark-adapted for 30 min and then subjected to a saturating light pulse, and the maximum quantum efficiency of photosystem II (ϕPSIImax) was calculated from basic (F0) and maximum level of fluorescence (Fm). Actinic light was then applied (150 µmol m−2 s−1 PAR) and, after 1 h of illumination, a saturating light pulse was triggered to determine transient fluorescence (Ft) and maximal fluorescence (Fm′). The corresponding quantum efficiency of photosystem II (ϕPSII) was calculated as (Fm′ − Ft)/Fm′. The non-photochemical quenching (NPQ) was determined as the ratio of (Fm − Fm′)/Fm′. The images of NPQ are presented using the standard false colour code, with rescaled values (original values divided by 4) ranging from 0 to 1.

Confocal microscopy

Confocal microscopy and FRET–FLIM analyses were performed as described previously78. In brief, leaves of 5-week-old tobacco (N. benthamiana) were infiltrated with a suspension of A. tumefaciens GV3101 (MP90) for expression of the viral p19 protein and A. thaliana proteins. The bacteria contained binary level II plasmids79 for the expression of GFP-TWA1/twa1, and mCherry fusions with JAM2, JAM3 and TPL under the control of the viral 35S promoter in plant leaves. Infiltrated plants were incubated for 2 days at 20 °C and exposed for 2 h to 37 °C or 20 °C. For analysis of yeast, freshly transformed yeast cells of strain AH109 (MATa, obtained from J. Uhrig, Cologne) were cultivated overnight in 1 ml synthetic dextrose medium (SD) at different temperatures as indicated before confocal analysis. Confocal analysis was conducted using the Olympus FluoView 3000 inverse laser-scanning confocal microscope with the UPLSAPO 60XW 60×/NA 1.2/WD 0.28 water-immersion objective (Olympus). For imaging of the GFP and mCherry fluorophore, tissue samples were excited at 488 and 561 nm, respectively. Specific GFP fluorescence in the nucleus was calculated by subtracting the background fluorescence80. For FRET–FLIM data acquisition, the PicoQuant advanced FCS/FRET–FLIM/rapidFLIM upgrade kit (PicoQuant) was used. GFP was excited at 485 nm with a pulsed laser (pulse rate, 40 MHz; laser driver, PDL 828 SEPIA II; laser, LDH-D-C-485, PicoQuant), and fluorescence emission was collected using the Hybrid Photomultiplier Detector Assembly 40 (PicoQuant) and processed by the TimeHarp 260 PICO Time-Correlated Single Photon Counting module (resolution, 25 ps; PicoQuant). At least 250 photons per pixel were recorded for each analysed sample. Data were fitted to a bi-exponential decay function and convoluted using SymPhoTime 64 software (PicoQuant).

Yeast constructs and growth analyses

Yeast growth and Y2H assays were performed as reported previously67 unless otherwise stated using pGAD424 (GenBank: U07647) and pBRIDGE vectors for expression of A. thaliana proteins (Clontech). A screen for proteins interacting with TWA1 was performed with A. thaliana Y2H libraries27,81. Coding sequences of TWA1 and homologues were expressed as fusions with the GAL4-activation domain (AD), while JAM, MYC2, TPL and TPRs were fused to the GAL4-DNA binding domain. For analysis of yeast growth in liquid culture, precultures of three independently transformed yeast colonies per construct were used for inoculation of 1.5 ml SD medium containing 2% glucose supplemented with 20 mg l−1 uracil, 20 mg l−1 methionine and, for selection of plasmids and protein interaction, supplemented with 20 mg l−1 histidine, 60 mg l−1 leucine and 50 mg l−1 tryptophan as indicated. After growth overnight in a gyratory shaker at 200 rpm and 30 °C, 13.5 ml supplemented SD was added, and the suspension culture was further cultivated until the beginning of the exponential growth phase was reached at optical densities at 600 nm (OD600) of between 0.6 and 0.8. Subsequently, cells were sedimented (1,500g, 5 min), resuspend in selective SD and used to inoculate 20 ml fresh SD to a final OD600 of 0.020 for monitoring growth at different temperatures (200 rpm). The apparent growth rate µ was calculated in the first 24 h of culturing using the formula µ = (ln[OD24 h/OD0 h])/24 h. The growth rate in yeast was used to approximate the Q10 temperature coefficient of the TWA1 response as Q10 = [µT2/µT1]10/(T2 − T1) where T is temperature in Celsius and µT1 and µT2 are the rates at T1 and T2, respectively. The Q10 value of ~150 was estimated using the measured rates of 0.138 and 0.011 h−1 at 25 °C and 20 °C, respectively.

Statistics and reproducibility

Statistics were derived from data of biological replicates excluding technical replicates. The sample size was calculated using the G*Power v. software82 for unpaired two-sided t-tests, with n = 3 for protoplast and yeast experiments with a presumed effect size of 5, α and β of 0.05, and power (1 − β) = 0.95. The number of biological replicates was n = 6 for single datapoints in the seedling growth assay with a presumed effect size of 2.5, α and β of 0.05, and power (1 − β) = 0.95. Each experiment was repeated at least twice with similar results. Representative confocal images are shown in Figs. 2d and  3c. In yeast, approximately 85% of fluorescent cells showed the GFP–TWA1 subnuclear crowding at 30 °C, never observed with the twa1 product (Fig. 2d, n > 100). In tobacco epidermal cells, the nuclear subdomain accumulation of GFP–TWA1 was 75% at 37 °C (Fig. 3c, n > 50). The nuclear TWA1 crowding was consistent in independent experiments. Data were analysed using Mann–Whitney U-tests unless otherwise stated ( and, to calculate exact P values of <0.0001). One-way analysis of variance was performed using SPSS v.16.0 software for Windows. Student’s t-tests were performed using Excel 2016. Box plots were drawn using OriginPro 2020. Details on the statistical analyses are found in the Source Data.

Reporting summary

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