Molecular switching in transcription through splicing and proline-isomerization regulates stress responses in plants

Abstract

The Arabidopsis thaliana DREB2A transcription factor interacts with the negative regulator RCD1 and the ACID domain of subunit 25 of the transcriptional co-regulator mediator (Med25) to integrate stress signals for gene expression, with elusive molecular interplay. Using biophysical and structural analyses together with high-throughput screening, we reveal a bivalent binding switch in DREB2A containing an ACID-binding motif (ABS) and the known RCD1-binding motif (RIM). The RIM is lacking in a stress-induced DREB2A splice variant with retained transcriptional activity. ABS and RIM bind to separate sites on Med25-ACID, and NMR analyses show a structurally heterogeneous complex deriving from a DREB2A-ABS proline residue populating cis- and trans-isomers with remote impact on the RIM. The cis-isomer stabilizes an α-helix, while the trans-isomer may introduce energetic frustration facilitating rapid exchange between activators and repressors. Thus, DREB2A uses a post-transcriptionally and post-translationally modulated switch for transcriptional regulation.

Fig. 1. Identification of the Med25-ACID binding region of DREB2A with a primary SLiM.

A Truncation of DREB2A and binding affinities determined using ITC. The graph shows disorder (DISOPRED) and helix propensity (Agadir) predictions of DREB2A. Red data points: AD score of 40-residue tiles as a function of the tile center position. Bottom heat map: residue specific AD scores calculated as the average score of all overlapping tiles. The higher AD score, the higher transcriptional activity. The hatched region is the minimal region needed to achieve maximum Med25 affinity. *Affinity determined using NMR. B Alignment of DREB2A orthologs from different plants species showing the position and consensus sequence of the ABS (red) and the RIM (blue). Source data are provided as a Source Data file.

Fig. 2. Characterization of the AtMed25-ACID domain structure using NMR, MD, and AlphaFold2.

A Secondary structure propensities from assigned chemical shifts calculated using MICS. Asterisks indicate residues for which only one or fewer ¹³C chemical shifts could be assigned. B AF2 structure prediction of Med25-ACID colored using the SSP scores in panel A. C Analysis of Med25-ACID fast timescale dynamics using NMR relaxation parameters, {¹H}−¹⁵N hetNOEs and MD simulation. Regions showing increased dynamics, indicated by lowered R₂/R₁ or lowered hetNOEs or by increased root-mean-square fluctuations (RMSF), are highlighted in gray. R₂/R₁ error bars represent propagated standard errors. HetNOE errors are standard deviations of three technical replicates. Dashed horizontal lines indicate average values of R₂/R₁ and hetNOEs for the folded part of the ACID domain. D Mapping of MD derived residue fluctuations onto the Med25-ACID structure. Source data are provided as a Source Data file.

Fig. 3. Mapping of DREB2A binding sites in Med25-ACID using NMR.

A ¹⁵N-HSQC spectra of ¹⁵N labeled Med25-ACID in complex with ABS (red), RIM (blue). and ABS-RIM (yellow). Zoom: residues affected differently depending on the DREB2A fragment. B ¹⁵N,H^N CSPs of Med25-ACID induced by ABS (top), RIM (middle), and ABS-RIM (bottom). Red and blue background: regions forming the ABS and the RIM binding sites, respectively. Bars are colored according to their residue color in panels D and E. *Residues highlighted in panel A. C Same as B, but tracing HSQC relative peak intensities as a function of DREB2A fragment concentration (dark to light). Molar ratios for the final intensity level (light color) are provided in the top right corner. D Mapping of ABS induced Med25-ACID CSPs on the AF2 structure. E Mapping of Med25-ACID CSPs induced by the bivalent ABS-RIM fragment. Blue spheres: RIM induced additional CSPs. D and E are colored according to significance levels of 0.1 (light) and 0.2 ppm (dark). Source data are provided as a Source Data file.

Fig. 4. Modeling of the Med25-ACID:DREB2A complex using AF2.

A Three structure models of Med25-ACID:DREB2A predicted using AF2 shown with the ABS-RIM induced Med25-ACID CSPs (green). The ABS (red) was predicted to form a single turn helix while the RIM (blue) was predicted to form a 10-residue helix. B Comparison of the packing of Tyr242 and Trp244 in model 1 (left) and 2 (right). C AF2 complex model 2 with the hydrophobic residues shown as spheres and the affinity effect of specific DREB2A variants shown. Source data are provided as a Source Data file.

Fig. 5. CEST and stopped-flow spectroscopy analyses of DREB2A showing two bound states.

A Example ¹⁵N-CEST profiles for DREB2A Gly243 in the trans and cis proline states, illustrating binding of both isomers, and Phe274 illustrating three-state behavior of sequentially distant residues. Shown profiles were recorded with (colored) and without (black) 5% unlabeled Med25-ACID at 25 °C. Vertical gray lines indicate the chemical shifts of detected populations, with the solid line marking the free state and dashed line(s) marking the bound state(s). B Kinetic model for analysis of the ¹⁵N-CEST data with derived affinities. Exchange between the cis and trans bound states is not considered. C DREB2A ¹⁵N CSPs extracted from the CEST experiment. The dashed line indicates an arbitrary significance level of 2.2 ppm. D Mapping of DREB2A ¹⁵N CSP on the AF2 complex model. Residues with significant CSPs are shown. E Stopped-flow fluorescence spectroscopy analysis of ABS (top) and ABS-RIM (bottom) to determine association (left) and dissociation (right) rate constants. Data points and error bars represent mean and standard deviation of 19 or more individually fitted traces assuming a normal distribution. Example averaged traces are shown for each analysis. Averaged traces and derived k_obs for all data points are shown in Supplementary Figs. 11 and 12. Source data are provided as a Source Data file.

Fig. 6. ¹³C-CEST and thermodynamic analyses show coupled binding and folding of DREB2A.

A ¹³C-CEST profiles of DREB2A with 5% Med25-ACID. B SSPs calculated using δ2D of DREB2A in the unbound and the Med25-ACID bound states. Bound state SSPs were predicted for chemical shifts extracted from the minimum and maximum perturbations using ¹⁵N, ¹³C^α, and ¹³C’. Unbound DREB2A chemical shifts were extracted from previously published data. *Data insufficient to predict SSP. C Model illustrating the difference in bound state secondary structure observed for the cis and trans proline states. D Thermodynamic characterization of the DREB2A interaction with Med25-ACID. (Top) temperature dependence of the thermodynamic parameters and derived parameters at 25 °C. Error bands for the linear fits are 95 % confidence intervals and error bars for the derived parameters obtained from the standard errors of the linear fits. The n number is the number of experiments for each fragment. (Bottom) R_th,ID based on the ID-adapted Spolar-Record method and alignment of fragment sequences. R_th,ID error bars were obtained using a Monte Carlo approach incorporating standard errors from the linear temperature dependence of the thermodynamic parameters. Source data are provided as a Source Data file.

Fig. 7. Stress dependent positive and negative regulation of DREB2A.

A Bivalent DREB2A interaction region, produced under non-stress conditions, forms a complex with Med25-ACID using both the ABS and the RIM, whereas the high-affinity RCD1-RST interaction is driven by the RIM. The ABS interaction with Med25 scaffolds the mediator complex while proline-dependent frustration of the RIM may promote binding of additional co-regulators. B Heat stress induced synthesis of a DREB2A splice variant lacking the RIM (and AD1) is unable to bind RCD1-RST but retains the ability to scaffold mediator through the ABS and activate transcription through AD2. Affinities are given at 25 °C while kinetic rate constants are given at 10 °C.