# Source paper archive - Source title: Thermopriming prewires orchestrated regulatory programs to enhance heat stress adaptability - Author(s): Nair, A. U., Vishwakarma, S., Kadumuri, R. V., Fritschi, F. B., Chavali, S., Allu, A. D. - Source landing URL: https://www.biorxiv.org/content/10.64898/2026.11.12.766148v1 - PDF URL: https://www.biorxiv.org/content/10.64898/2026.11.12.766148v1.full.pdf - PDF SHA-256: 00ade907f08d10fa76d90b320828b4835ebf31aefd7322dc8b6000b96e78af56 - Converter: pymupdf4llm 0.0.27 - Limits: Automated PDF text extraction; reading order, equations, tables and formatting may be incomplete or incorrect. Images are omitted, and no OCR was performed. Markup and links are neutralized for safe downloading. Consult the original PDF for the authoritative version. --- [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766148; this version posted November 13, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.12.766148) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 1 **Thermopriming prewires orchestrated regulatory programs to enhance heat** 2 **stress adaptability** 3 Akshay U Nair [1], Shubham Vishwakarma [1], Rajashekar Varma Kadumuri [1], Felix B. 4 Fritschi [2], Sreenivas Chavali [1,*], Annapurna Devi Allu [1,*] 5 1 Department of Biology, Indian Institute of Science Education and Research (IISER) 6 Tirupati, Tirupati- 517619, Andhra Pradesh, India. 7 2 Division of Plant Science and Technology, University of Missouri, Columbia, MO, 8 USA. 9 10 *Correspondence to: schavali@iisertirupati.ac.in; anu.allu@iisertirupati.ac.in 11 12 13 **Key words:** Thermopriming, Acquired thermotolerance, rice, transcriptional factors, 14 histone modifications, heat shock proteins Page 1 of 37 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766148; this version posted November 13, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.12.766148) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 15 **Summary** 16 - Plants can remember earlier stress exposures (priming), which improves tolerance 17 to later, more severe stress. How this stress memory is organised in crops remains 18 largely unresolved. 19 - We studied five-day-old rice seedlings subjected to severe heat stress with or 20 without thermopriming, combining phenotyping, time-resolved transcriptomic and 21 epigenomic profiling with systems biology approaches and targeted molecular 22 experiments. 23 - Compared with unprimed seedlings, primed seedlings showed a restrained, lower 24 entropy transcriptional response spanning diverse biological processes. This 25 response was driven by fewer transcription factors that modulate regulons for both 26 stress response and growth, together contributing to better adaptability to severe 27 stress. 28 - Primed seedlings maintained global H3K27me3 levels, which declined in unprimed 29 seedlings, linking retention rather than removal of this mark to the restrained 30 response. 31 - Priming led to the accumulation of memory chaperone transcripts, enabling a rapid 32 rise in protein abundance under severe stress. This supports better post 33 translational regulation, alongside the coordinated transcriptional response. 34 - Overall, thermopriming establishes pre-wired programmes that enhance 35 adaptability through regulatory order and speed, not amplitude, and repressive 36 mark retention, not removal, during early development. These findings have 37 implications for (i) understanding heat stress memory in addressing climate change 38 driven heat extremes, and (ii) engineering climate-resilient crops. 39 40 41 **Introduction** 42 Global warming poses a serious threat to crop productivity, as high temperatures 43 adversely affect key developmental stages, disrupt physiological processes and 44 ultimately result in substantial yield loss (Xu _et al._, 2020; Mthiyane _et al._, 2024; Singh 45 _et al._, 2024). To adapt to rapidly fluctuating and challenging environments, plants have 46 the ability to ‘remember’ prior stress encounters and use that ‘memory’ to better Page 2 of 37 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766148; this version posted November 13, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.12.766148) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 47 acclimate to future stresses. This adaptive strategy, known as ‘priming,’ provides 48 acquired tolerance for subsequent stresses (Lämke & Bäurle, 2017; Liu _et al._, 2022; 49 Nair _et al._, 2022). The magnitude, period and frequency of priming stimuli can 50 influence the duration and nature of stress memory, which can range from short- to 51 long-term somatic memory or transgenerational memory (Sani _et al._, 2013; Wang _et_ 52 _al._, 2016; Bäurle & Trindade, 2020). 53 Acquired thermotolerance, in which a sub-lethal heat stress serves as a priming cue, 54 enhances tolerance to subsequent heat stress (Charng _et al._, 2007; Lämke _et al._, 55 2016; Serrano _et al._, 2019). Individual molecular-level studies, predominantly in the 56 model plant _Arabidopsis thaliana_, show that (i) thermopriming enhances the 57 transcription of key heat shock factors ( _HSFs_ ) and other stress-responsive genes and 58 (ii) such HSFs collaborate with chromatin remodelers or histone-modifying enzymes 59 to alter chromatin accessibility and establish permissive epigenetic states that 60 contribute to heat stress adaptability (Brzezinka et al., 2016; Lämke _et al._, 2016; Liu 61 _et al._, 2018; Friedrich _et al._, 2021; Yamaguchi _et al._, 2021). Current models of 62 thermomemory, however, rest majorly on single memory genes, whose expression is 63 maintained or re-induced more strongly when stress recurs, with the help of HSFs and 64 chromatin modifiers such as REF6, an H3K27me3 demethylase (Brzezinka et al., 65 2016; Lämke _et al._, 2016; Yamaguchi & Ito, 2021; Yamaguchi _et al._, 2021). How 66 dynamic transcriptional reprogramming after priming shapes the response to repeated 67 stress, in terms of scale, coordination, and regulation, and which transcriptional and 68 epigenetic features drive it, is still largely unclear. This is particularly pertinent for rice 69 and other crops, where thermomemory is poorly characterised. 70 To address this, here we carried out extensive experiments and integrated time 71 resolved, genome-wide transcriptomic and epigenetic profiles in rice, a staple crop. 72 We show that thermopriming pre-wires a smaller, more ordered transcriptional 73 response to a subsequent triggering heat stress, characterised by fewer transcription 74 factors drawn from smaller paralog families, coordinated regulatory networks, 75 retention of H3K27me3, and higher levels of memory chaperone. 76 77 **Materials and Methods** 78 **Plant materials and growth conditions** Page 3 of 37 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766148; this version posted November 13, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.12.766148) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 79 Rice seedlings ( _Oryza sativa_ subsp. indica cv. Nagina22) were cultivated 80 hydroponically using 1X Hoagland nutrient solution, following the formulation of 81 Hoagland & Arnon (1950) with slight modifications. Before germination, seeds were 82 surface sterilised using 1% (v/v) sodium hypochlorite for 10 min and subsequently 83 rinsed several times with sterile distilled water. Around 15-20 sterilised seeds were 84 sown in each phyta-jar (transparent container with an aerated lid and valve). The 85 nutrient solution volume was adjusted between 40 and 80 mL per jar based on 86 seedling growth stage. Plants were grown under controlled conditions in a growth 87 chamber with a 14 h light/10 h dark cycle, maintaining day-time and night-time 88 temperatures of 28 °C and 26 °C, respectively. The relative humidity was set at 65 89 70%. 90 **Stress regime and physiological measurements** 91 Heat stress experiments were conducted on 5-day-old rice seedlings according to a 92 previously established protocol (Nair _et al._, 2026). Control plants were maintained 93 under optimal growth conditions at 28 °C. For priming, seedlings were exposed to a 94 sub-lethal heat treatment at 37 °C for 2 h, followed by a 2 h recovery at 28 °C. 95 Thereafter, both primed and unprimed seedlings were subjected to a lethal heat stress 96 (triggering treatment) at 45 °C for 4 h. Unprimed seedlings received only the triggering 97 stimulus without the priming step. Growth and physiological traits, including shoot 98 length, root length, and fresh weight, were measured on the 8 [th] day following heat 99 stress. Leaf surface temperature was recorded immediately after heat exposure using 100 a thermal infrared camera (FLIR A325sc), and data were analysed using FLIR RIRmax 101 software. 102 **Transcriptome analysis** 103 Shoot tissues from five-day-old rice seedlings were harvested at distinct time points: 104 immediately after priming, before the triggering stress, and immediately following 4 h 105 of triggering stress (45 °C) in primed and unprimed seedlings for three biological 106 replicates. Total RNA was extracted from 100 mg of tissue using the PureLink™ RNA 107 Mini Kit (Invitrogen), and RNA quality and integrity were assessed with the Agilent 108 2100 Bioanalyzer. Approximately 2 μg of high-quality RNA was used to prepare cDNA 109 libraries using the Illumina HiSeq2000 RNA Library Preparation Kit, followed by paired 110 end sequencing. Raw reads underwent quality control with FASTQC, and adapter Page 4 of 37 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766148; this version posted November 13, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.12.766148) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 111 trimming was performed using Trim Galore. Clean reads were aligned to the 112 _Oryza_sativa_ _aus_N22.assembly reference genome (from Rice Genome Hub) using 113 HISAT2 (Kim _et al._, 2019), and gene-level read counts were obtained with HTSeq 114 (Anders _et al._, 2015). Differential expression analysis was performed using DESeq2 115 (Love _et al.,_ 2014), and significantly upregulated (Log2FC ≥ 1.5) and downregulated 116 (Log2FC ≤ −1.5) genes were selected based on an adjusted P-value < 0.05. 117 **Gene expression dispersion entropy** 118 The distribution of Log 2 fold change values for upregulated and downregulated 119 differentially expressed genes (DEGs) was visualised using box plots, with median 120 values indicated. Statistical differences between the two distributions were assessed 121 using the Wilcoxon rank-sum test. To quantify gene expression dispersion at a single 122 time point, entropy ( _H_ _stp_ ) was calculated based on transcriptome-wide variance using 123 the equation: 124 𝐻 !"# = ln(𝜎 [$] ) 125 The variance (σ [2] ) of the whole-transcriptome DE distribution is computed using the 126 “var” function in R (v3.6.2). Similar calculations were simulated for DEG TFs as well. 127 **Network entropy analysis** 128 To calculate the approximate von Neumann entropy of a directed graph, the following 129 equation is used 𝑑 & & 1 &)&" [⋅] 𝑑 &)&" 130 𝐻 % = 1 − [1] [1] 1 𝑑 & ['(] [,] |𝑉| [−] |𝑉| [$] 𝑑 & &)&" . 𝑑 '( (&,*)./ ! 131 For a graph G, 𝐸 % is the directed edge set and node set V. The edge commences at 132 node u and ends at node v. 𝑑 & ['(] is the in-degree of node u, 𝑑 & [)&"] is the out-degree of 133 node u. Shannon entropy was also calculated for the static directed network in After 134 triggering time point using the equation. 135 **Identification of transcription factors and TF-target interactions** 136 Transcription factors (TFs) among the differentially expressed genes (DEGs) were 137 identified by submitting their nucleotide sequences to the iTAK database (Zheng _et al._, 138 2016). To assess family-wise enrichment, the number of DEG TFs in each TF family 139 was normalised by the total number of known TFs in that family in rice. To predict Page 5 of 37 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766148; this version posted November 13, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.12.766148) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 140 regulatory interactions between TFs and their targets, we used the Regulation 141 Prediction tool from the PlantRegMap database (Tian _et al._, 2020). Promoter 142 sequences spanning 1 kb upstream of DEG transcription start sites were retrieved 143 from RAP-DB (Sakai _et al._, 2013). The database scans these promoter regions for 144 known TF binding motifs to infer putative TF-target interactions and identifies 145 overrepresented TFs likely regulating the DEGs. DEG TFs with enriched binding sites 146 in the 1 kb promoter regions were retained for further analysis. Predicted TF-target 147 pairs were then filtered based on co-expression analysis. Pearson correlation 148 coefficients were computed using FPKM values across biological replicates, 149 separately for control-primed and control-unprimed conditions. TF-target correlations 150 with FDR-adjusted P < 0.05 were considered significant. 151 **TF-Target network analyses and module detection** 152 TF-target interaction pairs identified under unprimed and primed conditions were used 153 to construct regulatory networks using Cytoscape v3.9.1. Network topology metrics, 154 including indegree and outdegree, were calculated using the Network Analysis tool in 155 Cytoscape. To identify functional clusters, TF–target networks were partitioned into 156 modules using the GLay community clustering algorithm implemented in the 157 clusterMaker2 plugin (Shannon _et al._, 2003; Morris _et al._, 2011). Gene Ontology (GO) 158 enrichment analysis for genes within each module was conducted using the GO-Slim 159 function of the PANTHER database (Thomas _et al._, 2022). The number of TFs and the 160 percentage of known heat stress-responsive genes in each module were determined 161 using STIFDB v.2 (Naika _et al._, 2013), a curated database of stress-responsive 162 transcription factors in plants. 163 **DEG-Paralog cluster analysis** 164 Differentially expressed transcription factors (DEG TFs) identified under unprimed and 165 primed conditions were queried in the Plant Ensembl database to retrieve their 166 corresponding paralogous genes. The dplyr package in R was used to compute the 167 size of each DEG TF's paralog cluster. Paralog clusters were categorised into five 168 size-based groups: 1, 2, 3, 4, and ≥5 members. Statistical significance of the 169 distribution differences between conditions was assessed using Fisher’s exact test. 170 **Gene Ontology and protein class enrichment analysis** Page 6 of 37 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766148; this version posted November 13, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.12.766148) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 171 Gene Ontology (GO) enrichment analyses were performed for the DEGs and TF-target 172 modules using ShinyGO (v0.76; Ge _et al._, 2020). To identify enriched protein classes, 173 the gene sets were further analysed using the PANTHER classification system 174 (https://pantherdb.org/). 175 **Total protein extraction** 176 Total protein was extracted from shoot tissue of 5-day-old rice seedlings using the 177 Pierce™ Plant Total Protein Extraction Kit (Cat. No. A44056). Shoot tissue frozen in 178 liquid nitrogen was ground to a fine powder, and 150 mg of the powdered tissue was 179 transferred to the provided cartridge. A volume of 100 μL of denaturing lysis buffer was 180 added, and the tissue was further homogenised using the plastic grinding rod (~60 181 strokes). The cartridge was capped and incubated at room temperature for 1–2 min. 182 Lysates were centrifuged at 16,000 x g for 5 min, and the supernatant containing 183 denatured total protein was collected into a microcentrifuge tube and stored at -80 °C 184 for further use. 185 **Histone protein extraction** 186 Histone proteins were extracted from shoot tissues of 5-day-old rice seedlings using 187 the EpiQuik™ Total Histone Extraction Kit (Cat. No. OP-0006-100). Approximately 188 200 mg of frozen shoot tissue was homogenised in 1 mL of 1X pre-lysis buffer by 189 vigorous vortexing, with intermittent cooling on ice. The homogenate was transferred 190 to a 2 mL microcentrifuge tube and centrifuged at 10,000 rpm for 1 min at 4 °C. The 191 supernatant was discarded, and the pellet was resuspended in 600 µL of lysis buffer, 192 followed by incubation on ice for 30 min. The lysate was centrifuged at 12,000 rpm for 193 5 min at 4 °C, and the clear supernatant was collected. A Balance-DTT buffer was 194 prepared by diluting DTT in Balance buffer at a 1:500 ratio, and 180 µL of this buffer 195 was added to the protein extract. The final histone protein samples were stored at 196 80 °C for further use. 197 **Immunoblot analysis** 198 Total protein extracts were used to quantify heat shock proteins (HSPs) at early and 199 later time points, while histone extracts were used to assess histone modifications at 200 corresponding time intervals. Western blotting was performed using the following 201 primary antibodies: anti-HSP101 (ab80121; Abcam), anti-HSP21 (AS08 285; 202 Agrisera), anti-Tubulin beta chain (AS10 680; Agrisera), anti-H3 (ab1791; Abcam), Page 7 of 37 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766148; this version posted November 13, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.12.766148) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 203 anti-H3K4me3 (ab8580; Abcam), anti-H3K27me3 (ab6002; Abcam), anti-H3K27ac 204 (ab4729; Abcam), and anti-H2A.Z (ab4174; Abcam). Signal detection was carried out 205 using the Pierce™ ECL Western Blotting HRP Substrate (Thermo Fisher Scientific). 206 Band intensities corresponding to specific proteins or histone modifications were 207 quantified using ImageJ software (Schneider _et al._, 2012). Each antibody-based assay 208 was performed with three independent biological replicates. 209 **Quantitative Real-Time PCR (qRT-PCR)** 210 Total RNA (2 μg) was treated with DNase using the Turbo DNase Kit (Thermo Fisher 211 Scientific) to remove any genomic DNA contamination. RNA concentration and purity 212 were assessed using a NanoDrop™ One/OneC Microvolume UV-Vis 213 Spectrophotometer (Thermo Scientific). First-strand cDNA synthesis was performed 214 using the RevertAid H Minus First Strand cDNA Synthesis Kit (Thermo Scientific). 215 Gene-specific primers for quantitative real-time PCR (qRT-PCR) were designed using 216 QUANTPRIME, a primer design tool optimised for high-throughput qPCR (Arvidsson 217 _et al._, 2008). Expression analysis of candidate genes was conducted using the SYBR 218 Green assay (POWER SYBR® GREEN PCR Master Mix) on BIO-RAD CFX384 and 219 CFX96 Real-Time PCR Systems with gene-specific forward and reverse primers. Melt 220 curve analysis was included to verify primer specificity and amplification efficiency. 221 The housekeeping gene _OsACTIN11_ was used as an internal reference for 222 normalising gene expression across samples. Relative expression levels were 223 calculated by converting C t values into Log₂ fold change (Log 2 FC) values. A complete 224 list of primers used in this study is provided in the supplementary table ( **Table S1.8** ). 225 **Chromatin Immunoprecipitation (ChIP) and library preparation** 226 For ChIP sequencing experiment five-day-old rice seedlings were used. Immediately 227 following 4 h of triggering stress (45 °C) in both primed and unprimed seedlings, about 228 2 g of shoot tissue was fixed in 1% formaldehyde. Nuclei isolation and chromatin 229 extraction were performed according to the protocols and extraction buffers described 230 by Song _et al._ (2016) and Ramirez-Prado _et al._ (2021). Chromatin was sheared using 231 a Sonics Vibra-Cell sonicator to obtain DNA fragments of 300-500 bp. Sonication 232 efficiency and fragment size distribution were verified using a TapeStation system 233 (Agilent). For each sample, 40 µL of Protein G Dynabeads (Thermo Fisher Scientific) 234 were washed with PBS-BSA and incubated overnight at 4 °C with 4 µg of anti Page 8 of 37 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766148; this version posted November 13, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.12.766148) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 235 H3K27me3 antibody (ab6002; Abcam). The sheared nuclear lysate was diluted four 236 fold and incubated overnight with the antibody-bound beads. The beads were then 237 collected on a magnetic stand and washed sequentially with low-salt, high-salt, LiCl 238 and TE buffers. Immunocomplexes were eluted using the elution buffer described by 239 Ramirez-Prado _et al._ (2021). Cross-links were reversed by incubation overnight at 65 240 °C, followed by treatment with RNase A and proteinase K. DNA was purified using the 241 QIAquick PCR Purification Kit (Qiagen). The immunoprecipitated DNA was used to 242 prepare ChIP-seq libraries with the Illumina-compatible NEXTflex Rapid DNA-Seq Kit 243 2.0 (Revvity), following the manufacturer’s instructions. 244 **ChIP-Sequencing analysis** 245 Paired-end reads were trimmed with Trim Galore (v0.6.10; Phred < 20), aligned to the 246 N22 genome with BWA-MEM (v0.7.18), and duplicates were marked with Picard 247 (v3.3.0). H3K27me3 peaks were called for each replicate against its input using 248 MACS2 (v2.2.9.1; broad mode, broad cut-off 0.1, q < 0.05). Merged peaks (BEDTools 249 v2.31.1) present in both replicates were retained as high-confidence regions (16,525 250 primed; 9,377 unprimed). Genes whose gene body plus 2 kb upstream window 251 overlapped a high-confidence region were considered targets and classified as P-only, 252 UP-only, common or non-targets (36,140 genes). Signal was quantified as fragment 253 Counts Per Million (CPM) per window with a library-size-scaled pseudocount, 254 averaged across replicates and expressed as log 2 (primed/unprimed), used for ranking 255 only. 256 Targets were integrated with expression data ( **Table S1.1** ; DEGs: FDR < 0.05, 257 |log 2 FCh| ≥ 1.5). Primed log 2 FCh and DEG proportions of primed only (P-only) and 258 unprimed only (UP-only) genes were compared using two-sided Mann–Whitney U and 259 Fisher’s exact tests (Benjamini–Hochberg corrected; FDR). Paralog families were 260 defined as connected components of an all-against-all DIAMOND blastp search 261 (v2.2.6; e-value ≤ 1e−10, ≥ 40% identity, ≥ 50% coverage), and log 2 FCh of genes and 262 their paralogs was compared using Spearman’s correlation. Enrichment of GO, GO 263 slim, Pfam, InterPro, KEGG, Plant Reactome, MapMan, TF family and functional 264 categories (TFs, stress response, HSPs, HSFs) was assessed by one-sided 265 hypergeometric tests against annotated backgrounds (FDR < 0.05, ≥ 3 genes). Page 9 of 37 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766148; this version posted November 13, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.12.766148) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 266 Analyses were performed in Python (v3.9.21) and R (v4.3.3), and figures were 267 generated with ggplot2 (v3.5.2). 268 **Statistical analysis** 269 Phenotyping experiments were conducted with a minimum of three independent 270 biological replicates unless stated otherwise. For gene expression comparisons in 271 qRT-PCR assays, pairwise Student’s _t-_ tests were performed using R. Differences in 272 continuous and discrete variable distributions were assessed using the non 273 parametric Wilcoxon rank-sum test and Fisher’s exact test, respectively, with multiple 274 testing corrections applied using the false discovery rate (FDR) method. To evaluate 275 the enrichment of commonly differentially expressed genes (DEGs) between shoots 276 and roots, a permutation test was conducted with 10,000 randomisations. In each 277 iteration, the observed common DEGs were replaced with randomly selected 278 expressed genes to generate a null distribution of overlap counts. Z-scores were 279 computed based on the deviation of observed overlap from the mean of the 280 randomised overlaps, and P-values were estimated as the fraction of permutations 281 yielding an overlap equal to or greater than the observed value. All statistical analyses 282 were performed using R. All tests were two-sided unless stated otherwise, and FDR < 283 0.05 were considered significant. 284 285 **Results** 286 **Thermopriming** **enhances** **heat** **stress** **adaptability** **by** **streamlining** 287 **transcriptional responses** 288 To delineate how regulatory programmes are rewired during priming in crop plants, 289 we examined the impact of thermopriming on the heat stress response of rice 290 seedlings. We recently showed that exposure of naïve, unprimed five-day-old 291 seedlings to 45 °C for 4 h led to growth impairment compared to the control (untreated 292 seedlings) (Nair _et al._, 2026). To test the impact of thermopriming on the growth of rice 293 seedlings, we pre-exposed the seedlings to a sub-lethal heat stimulus at 37 °C for 2 h 294 (thermopriming), followed by a 2 h recovery period, and then subjected them to a 295 growth impairing heat stress of 45 °C for 4 h (triggering heat stress) ( **Figure 1a** ). 296 Measurements of shoot and root length and fresh weight (eight days post-triggering 297 stress) depict better growth in primed compared to unprimed seedlings ( **Figure 1b-d** ). Page 10 of 37 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766148; this version posted November 13, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.12.766148) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 298 These findings establish that thermopriming aids better adaptability to acute heat 299 stress during early seedling development in rice. 300 How does thermopriming confer such heat stress adaptability? To address this, we 301 profiled the transcriptomes of control, unprimed, and primed rice seedlings after 302 exposure to triggering heat stress (After triggering, AT). Surprisingly, we observed a 303 significantly lower number of differentially expressed genes (DEGs) in response to 304 stress in primed seedlings (3,957 DEGs; 2,080 upregulated and 1,877 downregulated) 305 compared to unprimed seedlings (4,748 DEGs; 2,742 upregulated and 2,006 306 downregulated; **Figure 1e-1f; Table S1.1** ) **.** Nevertheless, both conditions shared a 307 significant number of common DEGs (2,908 DEGs) ( **Figure 1g** ), alluding to an 308 overlapping basic framework of heat stress response. In addition to a reduction in the 309 number of DEGs, the magnitude of differential expression (represented by fold 310 change), specifically for the upregulated DEGs, was significantly lower in primed than 311 in the unprimed seedlings ( **Figure 1h** ). Estimation of dispersion entropy, which 312 assesses the extent of variability in the differential expression within each condition 313 (Zhu _et al._, 2020), revealed relatively lower entropy for upregulated DEGs in the 314 primed condition compared to those of the unprimed condition (H=4.16 vs 4.36; **Figure** 315 **1i** ). Notwithstanding a reduced number of DEGs, Gene Ontology biological processes 316 (GO-BP) enrichment analysis revealed a higher number of enriched GO-BPs for DEGs 317 upon priming compared to the unprimed state ( **Figure 1j; Table S1.2** ). Overall, 318 thermopriming appears to elicit a streamlined transcriptional response that engages 319 diverse stress-response processes, which may underlie improved adaptability of 320 primed seedlings. Page 11 of 37 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766148; this version posted November 13, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.12.766148) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) **Figure1** **a** 5d old seedling 28°C TS **b** **c** **d** Primed 9 7 5 3 1 2.6 2.2 1.8 1.4 1 Control Unprimed Primed Control (C) TS Unprimed (UP) (P) 20 16 12 8 4 **e** Unprimed Primed 2006 DEGs 2742 1877 DEGs 2080 Upregulated Downregulated **f** **h** **i** 100 75 50 25 0 10.0 7.5 5.0 2.5 0.0 Control Primed Unprimed **g** P = 2.2x10 [-16] Upregulated Downregulated Upregulated Downregulated Z-score = 97.09 P < 1.0x10 [-4] 0.020 0.015 0.010 0.005 0 |1046|U<br>niq<br>ueg<br>794 enes<br>0 384|Col3|Col4| |---|---|---|---| |0|794<br>|794<br>|794<br>| |1696|0|384|384| |0|1212|0|665| DEGS Non-DEGs 0.0 -2.5 -5.0 Unprimed Primed 625 675 725 775 825 3000 25 50 75 100 No. of genes **j** |Col1|Col2|Col3|Col4| |---|---|---|---| ||||| ||||| ||||| H = ln( σ [2] ) -n 0 n Log 2 FC Upregulated DEGs 4.5 4.0 3.5 3.0 Biological processes 321 Page 12 of 37 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766148; this version posted November 13, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.12.766148) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 322 323 **Figure 1. Thermopriming-induced transcriptional reprogramming confers heat** 324 **stress tolerance. (a)** Schematic representation of the experimental design. Five-day 325 old rice seedlings were either left untreated (control), exposed to a lethal triggering 326 heat stress (unprimed), or subjected to a mild pre-treatment at 37 °C for 2 h and a 327 recovery for 2 h (primed) before exposure to the triggering heat stress (45 °C for 4 h). 328 Representative images showing shoot and root growth differences across control, 329 primed, and unprimed seedlings eight days post-stress. Box plots indicating **(b)** shoot, 330 **(c)** root length and **(d)** total seedlings’ fresh weight ( _n =_ 15 seedlings for all 331 measurements; FW measurements were done by summation of the values for 15 332 seedlings together independently for three sets). _P_ values were calculated using a 333 two-tailed Student’s _t_ -test. **(e)** Distribution of differentially expressed genes (DEGs) 334 under unprimed and primed conditions following triggering heat stress. **(f)** Bar plot 335 showing the proportion of DEGs and non-DEGs among the total expressed genes. _P_ 336 values were calculated using Fisher's exact test. **(g)** Matrix plot showing the overlap 337 of DEGs between unprimed and primed. Permutation test for checking the enrichment 338 of DEGs overlap between unprimed and primed conditions. The grey histogram 339 represents the random expectation, and the red arrow indicates the actual 340 observation. **(h)** Boxplots showing the distribution of Log 2 fold change values of the 341 upregulated and downregulated DEGs. _P_ values were calculated using the Wilcoxon 342 rank sum test. **(i)** Line plot showing the dispersion entropy that was calculated based 343 on the variance of gene expression distribution of upregulated DEGs. **(j)** Dot plot 344 showing gene ontology (GO) enriched biological processes from DEGs in unprimed 345 and primed seedlings. Dot size reflects the fold enrichment. 346 347 **Thermopriming-mediated transcriptional streamlining is potentially driven** 348 **through constrained functional redundancy and H3K27me3-mediated gating** 349 How is transcriptional streamlining achieved in primed seedlings? We posit that such 350 streamlining could be driven through a systematic regulation of transcription factors 351 (TFs) and/or epigenetic control. Indeed, primed plants showed fewer DETFs than 352 unprimed plants (214 vs 285) across a similar number of TF families ( **Figure 2a; Table** 353 **S1.3** ), with a lower magnitude of differential expression of upregulated DETFs ( **Figure** 354 **2b** ), and lower dispersion entropy among upregulated DETFs (H = 4.25 vs. 4.53; 355 **Figure 2c** ), compared to unprimed plants. Functional redundancies resulting from 356 paralogs arising through gene-duplication (genetic redundancy), tend to immensely Page 13 of 37 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766148; this version posted November 13, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.12.766148) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 357 contribute during acute stress response (Zou _et al._, 2009). Interestingly, we find 358 constrained genetic redundancies upon priming, wherein DETFs and DEGs in the 359 primed condition are limited to small or moderately sized paralog clusters as against 360 DETFs and DEGs in the unprimed condition, which tend to be preponderant in large 361 paralog clusters (≥4 membership; **Figure 2d; Supplemental Figure 1; Tables S1.4** 362 **and S1.5** ). Together, these point to a streamlined transcriptional response that could 363 reduce functional redundancies or conflicts (Flavell, 1994; Garber _et al._, 2024) and the 364 degree of gene-expression variability, thereby supporting a more efficient adaptive 365 response upon priming. 366 What are the potential drivers of priming-mediated transcriptional streamlining? Given 367 the shorter duration of priming-induced stress memory (2 h recovery) studied here, we 368 hypothesised that the swift systematic alterations in gene expression could be 369 primarily driven through reversible histone modifications that promote (active marks) 370 or inhibit gene expression (repressive marks) (Lämke & Bäurle, 2017; Bäurle & 371 Trindade, 2020; Ueda & Seki, 2020). In this regard, we probed for known histone 372 marks (active: H3K27ac, H3K4me3 and repressive: H3K27me3) and a thermo 373 responsive histone variant H2A.Z whose eviction facilitates gene expression during 374 heat stress (Kumar & Wigge, 2010), in control, primed control (exposed to only priming 375 stimulus), primed and unprimed seedlings upon heat stress. We did not observe a 376 difference in the total levels for the active marks H3K27ac, H3K4me3 and H2A.Z 377 across all the conditions analysed. However, the repressive mark H3K27me3 was 378 found to be significantly lower in unprimed seedlings compared to control or primed 379 seedlings ( **Figure 3a** ). Page 14 of 37 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766148; this version posted November 13, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.12.766148) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) **Figure 2** **a** **b** P = 2.1x10 [-3] 0 50 100 150 200 250 300 No. of DETFs Unprimed Primed 10 7.5 5 2.5 120 90 60 30 0 **c** -1 -2 -3 -4 -5 P = 3.9x10 [-4] H = ln( σ [2] ) Upregulated DETFs 0 **d** DETF paralog cluster membership with respect to DETF- A Ancestral gene 4.5 4.0 3.5 3.0 Paralog cluster NPT (Non-DE paralogous TF) Paralog cluster Cluster membership = 2 DETFs Non-DETFs NPT 1 2 3 4 ≥5 DETF paralog cluster membership NPT 1 2 3 4 380 381 **Figure 2. Thermopriming limits transcriptional regulatory redundancy by** 382 **maintaining H3K27me3-mediated silencing. (a)** The total number of DETFs in UP 383 (n = 285) and P (n = 214). _P_ values were calculated using Fisher's exact test. **(b)** 384 Boxplots showing the distribution of Log 2 fold change values of the upregulated and 385 downregulated DETFs. _P_ values were calculated using the Wilcoxon rank sum test. 386 **(c)** Dispersion entropy was calculated based on the variance of gene expression 387 distribution for DE-TFs. **(d)** DEG paralog cluster size distribution for DETFs. _P_ values 388 were calculated using Fisher's exact test. 389 390 To examine how H3K27me3 might gate transcription, we performed chromatin 391 immunoprecipitation followed by sequencing (ChIP-seq) on primed and unprimed 392 seedlings after triggering stress. Consistent with our immunoblots ( **Figure 3a** ), primed 393 seedlings carried about 15% more H3K27me3-marked genes (10,475 vs. 9,129; 394 **Figure 3b; Table S1.6** ), including more condition-specific genes (4,268 vs. 2,922; Page 15 of 37 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766148; this version posted November 13, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.12.766148) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 395 **Figure 3c** ). A large proportion of marked genes (6,207) was common between the two 396 conditions, prompting us to ask whether the distribution of the marks along these 397 genes is similar or different. Interestingly, unprimed seedlings showed elevated signal 398 across gene bodies, whereas primed seedlings showed a broader set of marked 399 genes with modest enrichment upstream of the transcription start site (TSS; **Figure** 400 **3d** ). Thus, priming appears to spread H3K27me3 marking to more loci, preferentially 401 at regulatory regions. 402 To relate H3K27me3 deposition to gene expression, we compared marked genes in 403 primed and unprimed seedlings. This comparison revealed that H3K27me3 marking 404 was associated with (i) significantly lower proportion of differentially expressed marked 405 genes, from ~24% (unprimed) to ~14% (primed) for uniquely marked genes and from 406 26.5% to 22.6% for commonly marked genes ( **Figure 3e** ); and (ii) a dampened the 407 magnitude of differential expression of uniquely marked DEGs and DETFs in the 408 primed seedlings ( **Figure 3f** ). This restraint was stronger for genes that were 409 previously reported to be stress-related (Oryzabase; 410 https://shigen.nig.ac.jp/rice/oryzabase/gene/advanced/list). About 14.8% of marked 411 stress-related genes were differentially expressed in primed seedlings compared with 412 31.4% in unprimed seedlings, and they made up a smaller share of marked DEGs 413 (18.6% vs 25.8%; **Figure 3g** ). Thus, H3K27me3 retention upon priming appears to 414 limit both the expression of the number of genes in response to triggering stress and 415 their magnitude, mirroring the streamlined transcriptional response ( **Figures 1e–i and** 416 **2** ). 417 We next asked whether H3K27me3 marking contributes to the constrained paralog 418 redundancy observed upon priming ( **Figure 2d; Supplemental Figure 1** ). About 10% 419 of differentially expressed paralogs unique to unprimed seedlings carried H3K27me3 420 uniquely in either primed or unprimed seedlings ( **Supplemental Figure 2** ). The 421 position of the mark differed between conditions: in primed seedlings, marks were 422 enriched in the upstream region (−2 kb of TSS; ~60%), followed by the gene body 423 (TSS to TES; ~27%), whereas in unprimed seedlings, most marks spanned both 424 regions (~60%), with the remainder in the gene body (~20%) or upstream region 425 (~20%) alone. 426 We then asked whether this differential marking could explain why these paralogs are 427 differentially expressed only in unprimed seedlings. About 60% of paralogs carrying Page 16 of 37 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766148; this version posted November 13, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.12.766148) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 428 upstream marks in primed seedlings, where they were not differentially expressed 429 upon priming, were upregulated in unprimed seedlings, where they lacked this mark. 430 In contrast, paralogs marked at other positions tended to be downregulated in both 431 conditions. Thus, upstream H3K27me3 retained upon priming may restrain the 432 expression of these paralogs, whereas in unprimed seedlings, where the mark is less 433 retained, other regulatory mechanisms may act alongside H3K27me3. 434 Collectively, these findings imply that constrained functional redundancy and 435 transcriptional repression aided by repressive histone mark H3K27me3-mediated 436 gating may jointly drive streamlining of gene expression in the primed seedlings. **(a)** Upregulated Downregulated UP P UP P **(b)** **(d)** DETFs 1250 1000 750 500 250 0 P = 9.9 x 10 [-5] NPG 1 2 3 4 ≥5 **(c)** Number of DETFs 2 4 6 8 10 12 >14 UP P UP P Primed Unprimed Unique DETFs in Unprimed Fraction of DETFs with DEG paralogs in non unique TF set bHLH C2H2 OsGRX3 4/7 OsSNAT2 OsIDD12 OsIDD13 14/20 ILI5 OsbHLH053 OsbHLH052 RAI1 OsbHLH084 OsbHLH055 BU1 OsbHLH054 OsbHLH030 OsbHLH080 OsbHLH039 OsbHLH046 OsBC1 OsbHLH004 DEG paralog cluster membership UP P UP P UP P UP P UP P UP P HSF ENAC1 ONAC039 ONAC063 ONAC121 ONAC045 NAC17 NAC 6/10 MYB MYB35 3/4 MYB77 MYB44 OsERF027 OsERF011 OsDREB1G OsERF101 OsERF094 OsERF005 6/6 WRKY bZIP bZIP23 2/6 OsbZIP78 ERF 5/7 6/6 HSFB1 HSFC2A OsHsfC1b HSFA2E HSFA2B HSFA4D OsWRKY15 OsWRKY14 OsWRKY68 WRKY28 WRKY42 OsWRKY97 437 438 **Supplemental Figure 1. Paralog clusters of heat stress-specific DEGs and key** 439 **transcription factors (TFs) in primed and unprimed conditions. (a)** Heatmap Page 17 of 37 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766148; this version posted November 13, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.12.766148) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 440 showing the distribution of differentially expressed TF (DETFs) families in unprimed 441 (UP) and primed (P) seedlings following triggering stress. **(b)** Distribution of 442 differentially expressed gene (DEG) paralog cluster sizes under primed and unprimed 443 conditions. _P_ values were calculated using Fisher's exact test. **(c)** Schematic outlining 444 the identification of uniquely upregulated transcription factors (TFs) in unprimed 445 seedlings that have DEG paralogs (left). Uniquely downregulated paralog TFs in 446 unprimed seedlings for the bHLH and C2H2 TF families (right). **(d)** Uniquely 447 upregulated paralog TFs in unprimed seedlings. Major TF families include HSF, bZIP, 448 MYB, NAC, ERF, and WRKY. 449 Page 18 of 37 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766148; this version posted November 13, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.12.766148) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) **Figure 3** **a** AT 1.0 **b** **c** AT 1.0 12500 ~15% C PC P UP H2A.Z H3K27ac H3K4me3 H3K27me3 H3 Control (C) |= P 1 .9x|Col2| |---|---| ||| |Col1|10475| |---|---| |9129|9129| Unprimed (UP) 10000 7500 5000 2500 0 0.75 0.5 0.25 0 Primed control (PC) After triggering stress time point (AT) Primed (P) **d** Unique marks **e** Common Distribution of Expressed Genes Unprimed n = 2,922 Primed n = 4,268 Common n = 6,207 Unique marks Common marks 0 -2kb TSS TES +2kb 2 1 0 2 1 0 2 1 P = 3.161x10 [-7] P = 1.34x10 [-4] 100 75 50 25 0 DEGs Non-DEGs 100 75 50 25 0 -2kb TSS TES +2kb -2kb TSS TES +2kb Primed Unprimed Primed Unprimed Primed Unprimed Primed Unprimed Count per million (CPM) CPM (top colour ≥ 3.67, 99 [th] percentile) 0 1 2 3 Expression distribution of marked genes **g** Distribution of stress-trait associated genes Expression distribution of stress-trait **f** Expression distribution of marked genes **g** associated genes DEGs DETFs Unique marks Common marks Unique marks Common marks 15 10 5 0 -5 -10 100 75 50 25 0 |Col1|Col2| |---|---| ||| Not Stress Stress P = 0.0134 P = 1.0 P = 0.0004 P = 0.003 0 100 75 50 25 0 100 75 50 25 |Col1|n=450vs|Col3|Col4| |---|---|---|---| ||||| ||||| ||||| ||||| ||||| |5|P = 0.03 n= 19 vs 29 DETFs|Col3| |---|---|---| |||| |||| |||| |||| |||| |Col1|Col2| |---|---| |-|-| |+|-| |+|+| |Col1|Col2| |---|---| |-|-| |+<br>-|-| |+<br>-|+| + 15 10 5 0 -5 -10 P = 0.013 n = 450 vs 397 100 75 50 25 0 Differentially Expressed Genes (DEGs) Differentially Expressed Transcription factors (DETFs) Non DEG Upregulated Downregulated 450 451 **Figure 3. Retention of H3K27me3 upon priming contributes to the restrained** 452 **transcriptional response of marked genes. (a)** Western blot analysis of histone 453 marks H3K27ac, H3K4me3, H3K27me3, and variant H2A.Z across control (C), primed 454 control (PC), primed (P) and unprimed (UP) conditions after the triggering stress (AT); 455 H3 serves as the loading control. Bar plot shows H3K27me3 signal normalised to H3 Page 19 of 37 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766148; this version posted November 13, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.12.766148) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 456 in unprimed and primed seedlings; the P value was calculated using a two-tailed 457 Student’s t-test. ‘*’ indicates P < 0.05 ( **b** ) Number of H3K27me3-marked genes 458 identified by ChIP-seq in primed and unprimed seedlings following triggering stress. 459 **(c)** Overlap of H3K27me3-marked genes between unprimed (UP) and primed (P) 460 seedlings. **(d)** Metagene profiles (top; mean CPM) and heatmaps (bottom) showing 461 H3K27me3 signal across gene bodies, from 2 kb upstream of the transcription start 462 site (TSS) to 2 kb downstream of the transcription end site (TES), for unprimed-only, 463 primed-only and commonly marked genes. **(e)** Bar plots showing the proportion of 464 expressed genes: differentially and non-differentially expressed genes (DEGs and 465 non-DEGs) in uniquely and commonly marked genes in primed and unprimed. P value 466 was calculated using the Fischer’s exact test. **(f)** Box plots showing the magnitude of 467 expression changes for DEGs and DETFs marked in primed and unprimed seedlings. 468 P values were calculated using the Wilcoxon rank sum test. **(g)** Bar plots showing the 469 proportion of marked stress-related genes among uniquely and commonly marked 470 genes (left; Fischer’s exact test), and those that were upregulated, downregulated or 471 not differentially expressed (right; chi-square test). **a** **b** -3 100 75 50 25 P = 2.73x10 100 75 50 25 0 |1|1| |---|---| |2|2| |3|3| |3|2| |3|3| |+|+|+| |---|---|---| |+|+|-| |-|-|-| |-|-|-| |+|+|+| |---|---|---| |+|-|-| |-|-|-| + Upregulated - Downregulated 1 2 3 Upstream Gene body Upstream and gene body Unprimed Primed 472 473 **Supplemental Figure 2.** **Effect of H3K27me3 marks on differently expressed** 474 **paralogs unique to unprimed.** **(** a **)** Bar plot shows the mark location of unprimed 475 uniquely differentially expressed paralogs in primed (n= 81) and unprimed conditions 476 (n= 83). P-value was calculated using the chi-square test. **(** b **)** Bar plot shows the 477 expression distribution of marked genes correlative to different mark locations in 478 primed and unprimed. 479 Page 20 of 37 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766148; this version posted November 13, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.12.766148) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 480 **Priming reduces regulatory network complexity while preserving growth-related** 481 **modules** 482 How does streamlined transcription aid in priming-mediated responses? To address 483 this, we re-constructed the DETF-target regulatory networks as previously described 484 by us (Nair _et al._, 2026) ( **Figure 4a** ). Briefly, putative target genes were identified by 485 examining DEG promoter regions for enrichment of binding sites corresponding to 486 DETFs. We then quantified the association between DETFs and their predicted targets 487 using Pearson correlation analysis of their expression profiles. Subsequently, 488 significant DETF-target pairs (with FDR<0.05) were used to reconstruct a high 489 confidence network of DEG targets potentially regulated by DETFs. Strikingly, the 490 primed TF-target network was compact, consisting of 32 TFs and 983 targets with 491 1,808 interactions compared to 6,223 interactions among 46 TFs and 2,008 targets of 492 the unprimed network. Not surprisingly, the number of targets regulated by each TF 493 and the number of TFs regulating each target were significantly lower in primed 494 seedlings compared to the unprimed seedlings ( **Figure 4b** ). Estimation of von 495 Neumann entropy and Shannon entropy for the TF-target networks further confirmed 496 a lower network entropy state in primed seedlings ( **Figure 4c** ). These observations 497 underscore reduced transcriptional complexity in the primed condition. 498 Next, we investigated the impact of reduced transcriptional complexity upon priming 499 on the network architecture. For this, we analysed the primed and unprimed TF-target 500 networks using the GLay community clustering algorithm (Su _et al._, 2010), which 501 organises the TF-target networks into interconnected but distinct modules, as shown 502 previously (Nair _et al._, 2026). Modularity in biological networks can contribute to 503 efficient processing of information, increase robustness, adaptability, and reduce 504 perturbations, collectively contributing to the fitness or survivability of organisms 505 (Lorenz _et al._, 2011; Huitzil & Huepe, 2024). Interestingly, we found a comparable 506 number of modulons between the primed (n=14) and the unprimed (n=12) conditions, 507 though the primed TF-target network was considerably smaller in size. Akin to the 508 network-level observations, modules of the primed TF-target network showed lower 509 transcriptional complexity compared to that of the unprimed network ( **Supplemental** 510 **Figure 3a and 3b** ). We next examined the enriched biological processes for each 511 module across the two conditions and found that the distribution of biological 512 processes across the modules and the number of modules regulating each biological Page 21 of 37 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766148; this version posted November 13, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.12.766148) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 513 process to be comparable ( **Figure 4d and 4e; Table S1.7** ). A closer look at the 514 biological processes revealed common biological processes across primed and 515 unprimed conditions, such as transport, transcription, response to stress, metabolism, 516 and hormone signalling ( **Figure 4d** ). In addition, primed seedlings showed enrichment 517 for distinct biological processes that regulate genetic information flow and contribute 518 to growth, such as DNA replication (module M9), DNA repair (M8), protein folding (M7), 519 cell wall organisation (M10), and cell cycle progression (M6 and M9). In contrast, the 520 biological processes unique to the unprimed condition included shoot development 521 and post-translational modification. These results suggest that priming preserves the 522 regulation of biogenesis and maintenance processes, which may contribute to the 523 better growth of primed seedlings ( **Figure 1b–d** ). Collectively, these findings indicate 524 that orchestrated regulatory response during priming extends beyond reduced 525 transcriptional complexity at the node-level. It also involves a well-defined modular 526 organisation, where specific TF-regulated modules control distinct biological 527 processes, which may enhance robustness and adaptability while sustaining growth. 528 Page 22 of 37 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766148; this version posted November 13, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.12.766148) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) **Figure 4** **a** **b** Unprimed Primed 300 200 100 0 **c** 1.0000 0.9995 0.9990 0.9985 Unprimed Primed 46 TFs 6223 interactions 2008 targets Community clustering Modules Mn 32 TFs 1808 interactions 983 targets 8 6 4 2 0 8.9 8.5 8.1 7.7 |Col1|Col2| |---|---| ||| **d** **e** Unprimed Primed M12 M14 6 6 P= 0.81 P= 0.58 5 4 3 2 1 0 5 4 3 2 1 0 |Col1|P=| |---|---| ||| ||| ||| ||| ||| BP- Biological processes Post-translational modification 529 530 **Figure 4. Thermopriming reprograms transcriptional regulatory networks to** 531 **enhance specificity and reduce redundancy. (a)** Network maps of inferred TF 532 target interactions in unprimed and primed seedlings. Nodes represent genes (red: 533 TFs; blue: targets), and edges denote significant regulatory associations. **(b)** TF-target 534 associations in the unprimed and primed network. _P_ values were calculated using the 535 Wilcoxon rank sum test. **(c)** Calculation of network entropy using von-Neumann and 536 Shannon’s entropy. **(d)** Circos plots illustrating the biological processes regulated by 537 unprimed and primed modules. **(e)** Quantitative analysis of the regulatory interactions 538 between modules and biological processes in both unprimed and primed conditions. 539 _P_ values were calculated using the Wilcoxon rank sum test. Page 23 of 37 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766148; this version posted November 13, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.12.766148) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) % of heat stress responsive genes reported previously Downregulated Upregulated % of genes genes **b** 2.5% 5% 22.2% 3.5% 100 100 4.5% 100 100 100 540 541 **Supplemental Figure 3. Modular organisation of transcription factor-target** 542 **regulatory networks under heat stress. (a)** TF-target network modules derived from Page 24 of 37 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766148; this version posted November 13, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.12.766148) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 543 unprimed seedlings. **(b)** TF-target network modules derived from primed seedlings. 544 Modules represent groups of transcription factors (TFs) and their associated targets 545 inferred from gene expression data. Bar plots adjacent to each module indicate the 546 proportion of upregulated (black) and downregulated (grey) genes. Pie charts 547 represent the fraction of genes within each module previously implicated in heat stress 548 responses, based on the STIFDB database (Naika _et_ _al._, 2013; 549 https://caps.ncbs.res.in/stifdb2/). 550 551 **Priming-induced chaperone memory enables streamlined and effective heat** 552 **shock response** 553 Which priming-induced molecular changes contribute to the streamlined effective 554 transcriptional response after triggering stress? We hypothesised that the 555 transcriptional changes induced upon priming may contribute to the differential 556 response of the primed seedlings upon triggering heat stress. To unveil this, we 557 profiled the transcriptome changes immediately after exposure to priming (after 558 priming, AP) relative to respective control seedlings ( **Figure 5a** ). At AP, 200 genes 559 were differentially expressed (112 upregulated and 88 downregulated genes), which 560 showed enrichment for the biological processes ‘protein folding’ and ‘response to heat’ 561 ( **Figure 5b and 5c** ). Analysis of protein classes using protein domain-based 562 annotations for DEGs at AP revealed chaperones to be the most predominant class 563 ( **Figure 5d** ). These results suggest that the enhanced levels of chaperones after 564 priming could drive better heat stress response in primed seedlings after triggering 565 stress. 566 To test this, we investigated the expression profiles of the rice orthologs of a well 567 characterised Arabidopsis heat stress memory chaperone, AtHSP21 (rice ortholog: 568 OsHSP26; Sedaghatmehr _et al._, 2016) and the non-memory heat stress-responsive 569 chaperone, AtHSP101 (OsHSP101; Brzezinka et al., 2016; Lämke et al., 2016). As the 570 primed response to triggering stress relies on heat stress memory, we compared the 571 temporal expression dynamics of the heat stress memory chaperone with the non 572 memory but heat stress-responsive chaperone. At AP time point, both genes showed 573 induced expression ( **Figure 5e** ). At the before triggering (BT) time point (end of 574 recovery phase), _OsHSP26_ transcripts remained slightly elevated, whereas 575 _OsHSP101_ returned to levels seen in the control seedlings ( **Figure 5e** ). After triggering Page 25 of 37 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766148; this version posted November 13, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.12.766148) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 576 stress (AT), transcript levels of both chaperones were slightly lower in primed than in 577 unprimed seedlings ( **Figure 5e** ). We next asked whether the memory chaperone is 578 also retained at the protein level at BT. However, protein levels of both HSPs were 579 comparable between primed and control seedlings at BT ( **Figure 5f** ). In contrast, after 580 triggering stress, primed seedlings accumulated significantly more OsHSP26 protein 581 than unprimed seedlings, whereas OsHSP101 levels appeared similar ( **Figure 5f** ). 582 Thus, despite slightly lower transcript levels at AT, primed seedlings accumulate more 583 memory chaperone protein, suggesting that priming-mediated retention of OsHSP26 584 transcripts during recovery facilitates its enhanced accumulation upon triggering 585 stress. Notably, primed seedlings achieved higher memory chaperone protein levels 586 with lower transcriptional investment at AT, mirroring the reduced number and 587 magnitude of DEGs observed upon triggering stress ( **Figure 1e–j** ). Thus, priming 588 appears to pre-wire the protein maintenance machinery, allowing primed seedlings to 589 mount an effective heat shock response without the transcriptional burden seen in 590 unprimed seedlings, which may underlie their better growth after triggering stress 591 ( **Figure 1b–d** ). Page 26 of 37 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766148; this version posted November 13, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.12.766148) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) **Figure 5** After Priming DEGs **c** After Priming **a** **d** After Priming (AP) Before Triggering (BT) After Triggering (AT) After Priming **b** |Col1|88|Col3|Col4|Col5|112|Col7|Col8| |---|---|---|---|---|---|---|---| ||||||||| ||||||||| ||||||||| ||||||||| −10 −5 0 5 10 Log 2 FCh No. of genes Protein folding 4 9 14 Response to heat Response to salt stress Response to hydrogen peroxide Protein complex oligomerization Sucrose catabolic process AP BT AT Control Unprimed Primed AT 45°C 37°C 4h 60 40 20 0 25 20 15 10 5 AP BT |Col1|OsHSP26| |---|---| ||| ||| ||| ||| AP BT 8 6 4 2 0 HSP20 HSP70 Chlorophyll A-B binding protein DnaJ Alpha−amylase AAA_2 ClpB_D2−small Alpha−amyl_C2 Glycolytic BAG Glycosyl hydrolase family 100 No. of genes 2 4 10 OsHSP26 OsHSP101 alpha-tubulin **f** **e** Before Triggering kDa 25 100 50 AT AT After Triggering kDa 25 OsHSP26 0 50 |.9 x 10–3|Col2| |---|---| |P = 2|| ||| alpha-tubulin 100 OsHSP101 2.0 1.5 1 0.5 _OsHSP26_ 592 593 **Figure 5. Priming-induced chaperone memory enables a streamlined yet** 594 **effective heat stress response. (a)** Schematic of the experimental design. Sampling 595 time points include after priming (AP), before triggering (BT), and after triggering (AT). 596 **(b)** Volcano plot showing differentially expressed genes (DEGs) in primed plants at the 597 after-priming (AP) time point relative to the control. DEGs were identified with 598 thresholds of Log 2 fold change ≥1.5 or ≤ −1.5 and false discovery rate (FDR) <0.05. 599 DEGs are shown in brown. Numbers indicate the total count of significantly 600 downregulated (88) and upregulated (112) genes. **(c)** Bubble plot of enriched 601 biological process Gene Ontology (GO) terms among DEGs in primed plants after 602 priming. Bubble size corresponds to the number of genes assigned to each term, as 603 indicated in the legend. **(d)** Bubble plot showing enriched protein domain categories 604 among DEGs in primed plants after priming. Bubble size corresponds to the number 605 of genes, as indicated. **(e)** Relative transcript levels (Log₂FCh) of _OsHSP26 (left) and_ 606 _OsHSP101_ (right) across time points (AP, BT and AT) and conditions (primed and 607 unprimed). Data represent mean±S.E. **(f)** Immunoblot analysis of OsHSP26, 608 OsHSP101, and alpha-tubulin (loading control) protein levels before triggering (left; Page 27 of 37 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766148; this version posted November 13, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.12.766148) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 609 control versus primed) and after triggering (right; unprimed versus primed). Molecular 610 weights (kDa) are indicated. The bar graph (right) shows quantification of OsHSP26 611 signal normalised to alpha-tubulin in primed and unprimed plants after triggering. Data 612 represent mean±S.E. _P_ values (indicated by ‘*’) were calculated using two-tailed 613 Student’s _t_ -tests. 614 615 **Discussion** 616 Thermomemory models, thus far, are largely based on boosting/re-induction of 617 individual memory genes and their regulators, such as HSFs and chromatin modifiers, 618 upon repeated stress (Lämke _et al._, 2016; Friedrich _et al._, 2021; Yamaguchi _et al._, 619 2021; Nair _et al._, 2022). Here we take a new approach to priming-mediated stress 620 responses in a crop plant, investigating how transcriptional reprogramming after 621 priming shapes the response to repeated stress, by integrating global analysis of 622 different layers of genetic information flow in a time-resolved manner. The findings 623 presented here highlight that primed seedlings mount a streamlined transcriptional 624 response to triggering stress, indicating that memory both restrains and amplifies the 625 response. This restraint is associated with a distinct regulatory architecture: fewer 626 responding transcription factors (TFs) from smaller paralog families, more compact 627 TF–target networks and retention of the repressive histone mark ( **Figure 6** ), 628 suggesting a regulatory state pre-wired upon priming. 629 Page 28 of 37 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766148; this version posted November 13, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.12.766148) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 630 **Figure 6. Proposed model of how thermopriming restrains and prepares the** 631 **transcriptional response to triggering heat stress.** A model showing the difference 632 in molecular regulation in the unprimed and primed seedlings in response to acute 633 heat stress. Unprimed seedlings mount a broader transcriptional response, with more 634 DEGs, a larger TF–target network and greater disorderliness, which is associated with 635 reduced growth after triggering stress ( **Figure 1b–d** ). By contrast, primed seedlings 636 mount a smaller, more ordered response, involving fewer TFs drawn from smaller 637 paralog families and a more compact network, and associated with retention of 638 H3K27me3 at more genes, which are less likely to respond to stress. Together with 639 greater accumulation of the memory chaperone OsHSP26 ( **Figure 5e–f** ), these 640 features are proposed to support heat stress tolerance with a smaller growth penalty. 641 642 Acquired thermotolerance involves sustained or enhanced re-induction of stress 643 responsive genes (Lämke _et al._, 2016; Friedrich _et al._, 2021). However, prolonged 644 stress responses may impair growth and are thought to be actively reset (Crisp _et al._, 645 2016; Sedaghatmehr & Balazadeh, 2024; Staacke _et al._, 2025). Interestingly, we find 646 that thermopriming shapes a lean transcriptional response, enabling a swift, 647 energetically efficient and robust heat stress response. Priming results in fewer 648 differentially expressed genes, with modest expression changes, lower entropy and a 649 compact TF–target network, together with improved growth after triggering stress. 650 Similar patterns have been observed in other systems. In bacteria, higher 651 transcriptional entropy predicts poor fitness under stress (Zhu _et al._, 2020), and 652 drought-tolerant _Brassica napus_ genotypes show a more restrained, targeted 653 response than sensitive ones (Huang _et al._, 2026). 654 Priming may be favoured where stress recurs predictably and is more economical than 655 repeatedly mounting a full defence response (van Hulten _et al._, 2006; Mitchell _et al._, 656 2009; Hilker _et al._, 2016). We find that priming induces the expression of TFs from 657 smaller paralog families, whereas the induction of expression of large, duplicated 658 families, as seen in unprimed seedlings, typically drives the acute stress response 659 (Hanada _et al._, 2008; Zou _et al._, 2009; Jiang _et al._, 2013). This suggests that priming 660 routes a restrained response through a core set of less redundant regulators. Further, 661 our findings point to a complementary role for retained H3K27me3, a repressive 662 histone mark, in restraining gene expression. Priming-mediated retention of Page 29 of 37 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766148; this version posted November 13, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.12.766148) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 663 H3K27me3 may therefore limit arbitrary gene activation, whereas its removal in 664 unprimed seedlings coincides with an expansive response. This is consistent with, 665 rather than contrary to, H3K27me3 removal at memory genes (Yamaguchi & Ito, 2021; 666 Yamaguchi _et al._, 2021), suggesting a complementary strategy in which demethylation 667 activates a select gene set and retention elsewhere tempers the wider heat stress 668 response. Although causality remains untested, the association of H3K27me3 669 retention and restrained transcriptional response offers a new view of epigenetic 670 regulation of plant resilience. Such restraints may therefore be beneficial where heat 671 episodes recur within hours to days. 672 What might drive the restrained transcriptional response upon priming, in contrast to 673 a broader response in unprimed seedlings? Disrupted proteostasis, a major problem 674 for cells under heat stress, could drive more stress-responsive genes to be expressed 675 at higher magnitude (Sugio _et al._, 2009; Masser _et al._, 2019). Priming may counter this 676 through transcriptional activation of protein quality control mechanisms, along with 677 retention of memory chaperone transcripts, such as OsHSP26, which together lead to 678 higher accumulation of chaperone proteins upon the triggering stress. This could 679 preserve proteostasis and keep proteins functional, protecting them from heat-induced 680 unfolding, misfolding and degradation. Such robust proteostasis might in turn facilitate 681 a more streamlined regulatory reprogramming in primed seedlings, with fewer 682 differentially expressed genes, less functional redundancy, and lower expression 683 magnitude. 684 Together, these observations point to a phenomenon in which thermopriming pre 685 wires the response to recurring heat by establishing a regulatory state that both 686 prepares and tailors the later response, both during priming and recovery ( **Figure 6** ). 687 Preparation is reflected in the chaperone-mediated preservation of proteostasis and 688 restraint is reflected by the expression of a limited set of stress-responsive regulators, 689 possibly aided by retention of the repressive histone mark H3K27me3. 690 Whether this pre-wired transcriptional reprogramming is conserved across species 691 remains to be determined. Other mechanisms, not examined here, may also contribute 692 to priming-mediated preparation and restraint, including negative feedback on HSF 693 activity by accumulating chaperones, promoter architecture, control of RNA 694 polymerase II activity at memory genes (Crawford _et al._, 2024) and post-transcriptional Page 30 of 37 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766148; this version posted November 13, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.12.766148) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 695 regulation such as mRNA turnover and alternative splicing (Crisp _et al._, 2016). Thus, 696 the coordinated, pre-wired regulatory networks that result from priming may reflect 697 efficient adaptation to recurrent heat in specific rice-growing environments and could 698 be mechanisms that have been selected for in heat-resilient crop varieties. 699 **Acknowledgements** 700 ADA thanks the USDA-FAS Norman E Borlaug Award for the financial support to 701 conduct the RNA sequencing experiments. ADA and FBF thank partial funding 702 provided by USDA-FAS project # FX22SR-10961R006. SV thanks the University 703 Grants Commission (UGC), Government of India, for the doctoral fellowship. We thank 704 IISER Tirupati core funding (ADA, SC, RVK, AUN), Ramalingaswami Re-entry 705 Fellowship (BT/RLF/Re-entry/05/2018) from Department of Biotechnology, 706 Government of India (SC), Core Research Grant (CRG/2023/004691) from 707 Anusandhan National Research Foundation, Government of India (SC). 708 709 **Competing interests** 710 All the authors declare that they have no competing interests. 711 712 **Author contributions** 713 ADA conceived the idea and supervised the work. AUN performed the phenotyping 714 and immunoblot, ChIP-seq experiments and RNA-seq data analysis. RVK and SV 715 performed the ChIP-seq analysis and relevant figure generation. AKS prepared the 716 initial draft. ADA and SC wrote and edited the manuscript. ADA, SC, AUN, SV were 717 involved in designing the computational analysis, generation of figures, interpreting 718 the data. FBF was involved in editing the manuscript. 719 720 **Data availability** 721 The raw transcriptome data are submitted to the Sequence Read Archive (SRA) of the 722 National Centre for Biotechnology Information (NCBI) and will be made available upon 723 the acceptance of the manuscript. 724 725 Page 31 of 37 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766148; this version posted November 13, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.12.766148) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 726 **References** 727 **Anders S, Pyl PT, Huber W. 2015.** HTSeq—a Python framework to work with high 728 throughput sequencing data. _Bioinformatics_ **31** : 166–169. 729 **Arvidsson S, Kwasniewski M, Riaño-Pachón DM, Mueller-Roeber B. 2008.** 730 QuantPrime – a flexible tool for reliable high-throughput primer design for quantitative 731 PCR. _BMC Bioinformatics_ **9** : 465. 732 **Bäurle I, Trindade I. 2020.** Chromatin regulation of somatic abiotic stress memory. 733 _Journal of Experimental Botany_ **71** : 5269–5279. 734 **Brzezinka K, Altmann S, Czesnick H, Nicolas P, Gorka M, Benke E, Kabelitz T,** 735 **Jähne F, Graf A, Kappel C et al. 2016.** _Arabidopsis_ FORGETTER1 mediates stress 736 induced chromatin memory through nucleosome remodeling. _eLife_ 5: e17061. 737 **Charng YY, Liu HC, Liu NY, Chi WT, Wang CN, Chang SH, Wang TT. 2007. A heat-** 738 **inducible transcription factor, HsfA2, is required for extension of acquired** 739 **thermotolerance in Arabidopsis. Plant Physiology 143: 251–262.** 740 **Crawford T, Siebler L, Sulkowska A, Nowack B, Jiang L, Pan Y, Lämke J, Kappel** 741 **C, Bäurle I. 2024.** The Mediator kinase module enhances polymerase activity to 742 regulate transcriptional memory after heat stress in _Arabidopsis_ . _The EMBO Journal_ 743 **43** : 437–461. 744 **Crisp PA, Ganguly D, Eichten SR, Borevitz JO, Pogson BJ. 2016.** Reconsidering 745 plant memory: Intersections between stress recovery, RNA turnover, and epigenetics. 746 _Science Advances_ **2** : e1501340. 747 **Flavell RB. 1994.** Inactivation of gene expression in plants as a consequence of 748 specific sequence duplication. _Proceedings of the National Academy of Sciences,_ 749 _USA_ **91** : 3490–3496. 750 **Friedrich T, Oberkofler V, Trindade I, Altmann S, Brzezinka K, Lämke J, Gorka M,** 751 **Kappel C, Sokolowska E, Skirycz A** _**et al.**_ **2021.** Heteromeric HSFA2/HSFA3 752 complexes drive transcriptional memory after heat stress in _Arabidopsis_ . _Nature_ 753 _Communications_ **12** : 3426. Page 32 of 37 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766148; this version posted November 13, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.12.766148) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 754 **Garber AI, Sano EB, Gallagher AL, Miller SR. 2024.** Duplicate gene expression and 755 possible mechanisms of paralog retention during bacterial genome expansion. 756 _Genome Biology and Evolution_ **16** : evae089. 757 **Ge SX, Jung D, Yao R. 2020.** ShinyGO: a graphical gene-set enrichment tool for 758 animals and plants. _Bioinformatics_ **36** : 2628–2629. 759 **Hanada K, Zou C, Lehti-Shiu MD, Shinozaki K, Shiu SH. 2008.** Importance of 760 lineage-specific expansion of plant tandem duplicates in the adaptive response to 761 environmental stimuli. _Plant Physiology_ **148** : 993–1003. 762 **Hilker M, Schwachtje J, Baier M, Balazadeh S, Bäurle I, Geiselhardt S, Hincha** 763 **DK, Kunze R, Mueller-Roeber B, Rillig MC** _**et al.**_ **2016.** Priming and memory of stress 764 responses in organisms lacking a nervous system. _Biological Reviews_ **91** : 1118–1133. 765 **Hoagland DR, Arnon DI. 1950.** The water-culture method for growing plants without 766 soil. Circular 347. Berkeley, CA, USA: California Agricultural Experiment Station. 767 **Huang W, Zhu W, Zhang Z, Yan T. 2026.** Integrated epigenomic and transcriptional 768 profiling reveals genotype-specific adaptive reprogramming to drought stress in 769 _Brassica napus_ . _BMC Plant Biology_ **26** : 625. 770 **Huitzil S, Huepe C. 2024.** Life's building blocks: the modular path to multiscale 771 complexity. _Frontiers in Systems Biology_ **4** : 1417800. 772 **Jiang SY, González JM, Ramachandran S. 2013.** Comparative genomic and 773 transcriptomic analysis of tandemly and segmentally duplicated genes in rice. _PLoS_ 774 _ONE_ **8** : e63551. 775 **Kim D, Paggi JM, Park C, Bennett C, Salzberg SL. 2019. Graph-based genome** 776 **alignment and genotyping with HISAT2 and HISAT-genotype. Nature** 777 **Biotechnology 37: 907–915.** 778 **Kumar SV, Wigge PA. 2010.** H2A.Z-containing nucleosomes mediate the 779 thermosensory response in _Arabidopsis_ . _Cell_ **140** : 136–147. 780 **Lämke J, Bäurle I. 2017.** Epigenetic and chromatin-based mechanisms in 781 environmental stress adaptation and stress memory in plants. _Genome Biology_ **18** : 782 124. Page 33 of 37 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766148; this version posted November 13, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.12.766148) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 783 **Lämke J, Brzezinka K, Altmann S, Bäurle I. 2016.** A hit-and-run heat shock factor 784 governs sustained histone methylation and transcriptional stress memory. _The EMBO_ 785 _Journal_ **35** : 162–175. 786 **Liu H, Able AJ, Able JA. 2022.** Priming crops for the future: rewiring stress memory. 787 _Trends in Plant Science_ **27** : 699–716. 788 **Liu HC, Lämke J, Lin SY, Hung MJ, Liu KM, Charng YY, Bäurle I. 2018.** Distinct 789 heat shock factors and chromatin modifications mediate the organ-autonomous 790 transcriptional memory of heat stress. _The Plant Journal_ **95** : 401–413. 791 **Lorenz DM, Jeng A, Deem MW. 2011.** The emergence of modularity in biological 792 systems. _Physics of Life Reviews_ **8** : 129–160. 793 **Love MI, Huber W, Anders S. 2014.** Moderated estimation of fold change and 794 dispersion for RNA-seq data with DESeq2. _Genome Biology_ **15** : 550. 795 **Masser AE, Kang W, Roy J, Kaimal JM, Quintana-Cordero J, Friedländer MR,** 796 **Andréasson C. 2019. Cytoplasmic protein misfolding titrates Hsp70 to activate** 797 **nuclear Hsf1. eLife 8: e47791.** 798 **Mitchell A, Romano GH, Groisman B, Yona A, Dekel E, Kupiec M, Dahan O, Pilpel** 799 **Y. 2009.** Adaptive prediction of environmental changes by microorganisms. _Nature_ 800 **460** : 220–224. 801 **Morris JH, Apeltsin L, Newman AM, Baumbach J, Wittkop T, Su G, Bader GD,** 802 **Ferrin TE. 2011.** clusterMaker: a multi-algorithm clustering plugin for Cytoscape. _BMC_ 803 _Bioinformatics_ **12** : 436. 804 **Mthiyane P, Aycan M, Mitsui T. 2024.** Strategic advancements in rice cultivation: 805 Combating heat stress through genetic innovation and sustainable practices- A review. 806 _Stresses_ **4** : 452–480. 807 **Naika M, Shameer K, Mathew OK, Gowda R, Sowdhamini R. 2013.** STIFDB2: an 808 updated version of plant stress-responsive transcription factor database with 809 additional stress signals, stress-responsive transcription factor binding sites and 810 stress-responsive genes in _Arabidopsis_ and rice. _Plant and Cell Physiology_ **54** : e8. Page 34 of 37 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766148; this version posted November 13, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.12.766148) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 811 **Nair AU, Bhukya DPN, Sunkar R, Chavali S, Allu AD. 2022.** Molecular basis of 812 priming-induced acquired tolerance to multiple abiotic stresses in plants. _Journal of_ 813 _Experimental Botany_ **73** : 3355–3371. 814 **Nair AU, Vishwakarma S, Guha T, Kadumuri RV, Fritschi FB, Chavali S, Allu AD.** 815 **2026.** Ethylene–jasmonate interplay involving ERF regulons modulates acute heat 816 stress response in rice seedlings. _The Plant Journal_ 126: e70850. 817 **Ramirez-Prado JS, Latrasse D, Benhamed M. 2021. Histone modification ChIP-** 818 **seq on Arabidopsis thaliana plantlets. Bio-protocol 11: e4211.** 819 **Sakai H, Lee SS, Tanaka T, Numa H, Kim J, Kawahara Y, Wakimoto H, Yang C,** 820 **Iwamoto M, Abe T et al. 2013. Rice Annotation Project Database (RAP-DB): An** 821 **Integrative and Interactive Database for Rice Genomics. Plant and Cell** 822 **Physiology 54: e6.** 823 **Sani E, Herzyk P, Perrella G, Colot V, Amtmann A. 2013.** Hyperosmotic priming of 824 _Arabidopsis_ seedlings establishes a long-term somatic memory accompanied by 825 specific changes of the epigenome. _Genome Biology_ **14** : R59. 826 **Schneider CA, Rasband WS, Eliceiri KW. 2012.** NIH Image to ImageJ: 25 years of 827 image analysis. _Nature Methods_ **9** : 671–675. 828 **Sedaghatmehr M, Balazadeh S. 2024.** Autophagy: a key player in the recovery of 829 plants from heat stress. _Journal of Experimental Botany_ **75** : 2246–2255. 830 **Sedaghatmehr M, Mueller-Roeber B, Balazadeh S. 2016.** The plastid 831 metalloprotease FtsH6 and small heat shock protein HSP21 jointly regulate 832 thermomemory in _Arabidopsis_ . _Nature Communications_ **7** : 12439. 833 **Serrano N, Ling Y, Bahieldin A, Mahfouz MM. 2019.** Thermopriming reprograms 834 metabolic homeostasis to confer heat tolerance. _Scientific Reports_ **9** : 181. 835 **Shannon P, Markiel A, Ozier O, Baliga NS, Wang JT, Ramage D, Amin N,** 836 **Schwikowski B, Ideker T. 2003.** Cytoscape: A Software Environment for Integrated 837 Models of Biomolecular Interaction Networks. _Genome Research_ **13** : 2498–2504. 838 **Singh S, Viswanath A, Chakraborty A, Narayanan N, Malipatil R, Jacob J, Mittal** 839 **S, Satyavathi TC, Thirunavukkarasu N. 2024.** Identification of key genes and Page 35 of 37 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766148; this version posted November 13, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.12.766148) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 840 molecular pathways regulating heat stress tolerance in pearl millet to sustain 841 productivity in challenging ecologies. _Frontiers in Plant Science_ **15** : 1443681. 842 **Song L, Koga Y, Ecker JR. 2016. Profiling of transcription factor binding events** 843 **by chromatin immunoprecipitation sequencing (ChIP-seq). Current Protocols in** 844 **Plant Biology 1: 293–306.** 845 **Staacke T, Mueller** - **Roeber B, Balazadeh S. 2025.** Stress resilience in plants: the 846 complex interplay between heat stress memory and resetting. _New Phytologist_ **245** : 847 2402–2421. 848 **Su G, Kuchinsky A, Morris JH, States DJ, Meng F. 2010.** GLay: community structure 849 analysis of biological networks. _Bioinformatics_ **26** : 3135–3137. 850 **Sugio A, Dreos R, Aparicio F, Maule AJ. 2009. The cytosolic protein response as** 851 **a subcomponent of the wider heat shock response in Arabidopsis. Plant Cell 21:** 852 **642–654.** 853 **Thomas PD, Ebert D, Muruganujan A, Mushayahama T, Albou LP, Mi H. 2022.** 854 PANTHER: Making genome-scale phylogenetics accessible to all. _Protein Science_ **31** : 855 8–22. 856 **Tian F, Yang DC, Meng YQ, Jin J, Gao G. 2020.** PlantRegMap: charting functional 857 regulatory maps in plants. _Nucleic Acids Research_ **48** : D1104–D1113. 858 **Ueda M, Seki M. 2020.** Histone modifications form epigenetic regulatory networks to 859 regulate abiotic stress response. _Plant Physiology_ **182** : 15–26. 860 **van Hulten M, Pelser M, van Loon LC, Pieterse CMJ, Ton J. 2006.** Costs and 861 benefits of priming for defense in _Arabidopsis_ . _Proceedings of the National Academy_ 862 _of Sciences, USA_ **103** : 5602–5607. 863 **Wang X, Xin C, Cai J, Zhou Q, Dai T, Cao W, Jiang D. 2016.** Heat priming induces 864 trans-generational tolerance to high temperature stress in wheat. _Frontiers in Plant_ 865 _Science_ **7** : 501. 866 **Xu J, Henry A, Sreenivasulu N. 2020.** Rice yield formation under high day and night 867 temperatures-A prerequisite to ensure future food security. _Plant, Cell & Environment_ 868 **43** : 1595–1608. Page 36 of 37 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.12.766148; this version posted November 13, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.12.766148) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity. It is made available under a CC-BY-NC-ND 4.0 International license.](http://creativecommons.org/licenses/by-nc-nd/4.0/) 869 **Yamaguchi N, Ito T. 2021.** JMJ histone demethylases balance H3K27me3 and 870 H3K4me3 levels at the _HSP21_ locus during heat acclimation in _Arabidopsis_ . 871 _Biomolecules_ **11** : 852. 872 **Yamaguchi N, Matsubara S, Yoshimizu K, Seki M, Hamada K, Kamitani M, Kurita** 873 **Y, Nomura Y, Nagashima K, Inagaki S** _**et al.**_ **2021.** H3K27me3 demethylases alter 874 _HSP22_ and _HSP17.6C_ expression in response to recurring heat in _Arabidopsis_ . _Nature_ 875 _Communications_ **12** : 3480. 876 **Zheng Y, Jiao C, Sun H, Rosli HG, Pombo MA, Zhang P, Banf M, Dai X, Martin GB,** 877 **Giovannoni JJ** _**et al.**_ **2016.** iTAK: a program for genome-wide prediction and 878 classification of plant transcription factors, transcriptional regulators, and protein 879 kinases. _Molecular Plant_ **9** : 1667–1670. 880 **Zhu Z, Surujon D, Ortiz-Marquez JC, Huo W, Isberg RR, Bento J, van Opijnen T.** 881 **2020.** Entropy of a bacterial stress response is a generalizable predictor for fitness 882 and antibiotic sensitivity. _Nature Communications_ **11** : 4365. 883 **Zou C, Lehti-Shiu MD, Thomashow M, Shiu SH. 2009.** Evolution of stress-regulated 884 gene expression in duplicate genes of _Arabidopsis thaliana_ . _PLoS Genetics_ **5** : 885 e1000581. Page 37 of 37