# Source paper archive - Source title: Complex temperature profiles recorded through antisense co-transcriptional sensitivity - Author(s): Xu, J., Menon, G., Guo, C., Montez, M., JIN, X., Yang, M., Sindalovskaya, M., Reeck, S., Duncan, S., Mateo Bonmati, E., Xu, C., Jang, G.-J., Zhang, Y., Zhai, J., Ding, Y., Dean, C. - Source landing URL: https://www.biorxiv.org/content/10.64898/2026.11.10.765395v1 - PDF URL: https://www.biorxiv.org/content/10.64898/2026.11.10.765395v1.full.pdf - PDF SHA-256: 1124d8a24083ae9a5d8cbe311e498d1ac7ef03915599a52ab454dc345dd48fa8 - 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/10.64898/2026.11.10.765395; this version posted November 12, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.10.765395) preprint (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in [perpetuity.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 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 1 **Complex temperature profiles recorded through antisense co-transcriptional sensitivity** 2 3 Jin Xu [1] -, Govind Menon [1,2] -, Chaocheng Guo [1] -, Miguel Montez [1], Xianhao Jin [1,3], Minglei Yang [1,4], 4 Maria Sindalovskaya [1], Svenja Reeck [1], Susan Duncan [1], Eduardo Mateo-Bonmatí [1,5], Congyao 5 Xu [1], Geng-Jen Jang [1,6], Yueying Zhang [1], Jixian Zhai [3], Yiliang Ding [1*] & Caroline Dean [1*] 6 7 1 Department of Cell and Developmental Biology, John Innes Centre, Norwich Research Park, 8 Norwich NR4 7UH, UK 9 2 Department of Computational and Systems Biology, John Innes Centre, Norwich Research Park, 10 Norwich NR4 7UH, UK 11 3 Shenzhen Key Laboratory of Plant Genetic Engineering and Molecular Design, Institute of Plant 12 and Food Science, Department of Biology, School of Life Sciences, Southern University of 13 Science and Technology, Shenzhen, China 14 4 Current address: Division of Life Sciences and Medicine, University of Science and Technology 15 of China, Hefei, China. 16 5 Current address: Centro de Biotecnología y Genómica de Plantas, Universidad Politécnica de 17 Madrid (UPM) - Instituto Nacional de Investigación y Tecnología Agraria y Alimentaria 18 (INIA)/CSIC, 28223 Pozuelo de Alarcón, Madrid, Spain 19 6 Current address: Department of Life Science, College of Life Science, National Taiwan 20 University, Taipei 106319, Taiwan 21 22 23 - These authors contributed equally to this work 24 [* Corresponding authors: caroline.dean@jic.ac.uk or yiliang.ding@jic.ac.uk](mailto:caroline.dean@jic.ac.uk) 25 26 27 **SUMMARY** 28 29 **Plants integrate widely fluctuating temperature over seasons to regulate their** 30 **development. In vernalization,** _**FLOWERING LOCUS C**_ **(** _**FLC**_ **)** **[1,2]** **is repressed by cold-** 31 **induced antisense (** _**COOLAIR**_ **) transcription and epigenetically silenced by Polycomb** 32 **(PRC2) epigenetic switching** **[3]** **. Here, we show how** _**COOLAIR**_ **co-transcriptional dynamics** 33 **are central to integrating different facets of the natural temperature profile over different** 34 **timescales. Constant cold alters** _**COOLAIR**_ **RNA polymerase II processivity, splicing and** 35 **polyadenylation site choice with distinct kinetics over days to weeks. In contrast,** 36 **fluctuations below freezing at the same mean temperature caused extensive premature** 37 **termination and chromatin retention of** _**COOLAIR**_ **. High-resolution chromatin contact** 38 **mapping showed** _**COOLAIR**_ **premature termination altered local chromatin topology. This** 39 **enhanced** _**FLC**_ **promoter-gene body contacts, coincident with H3K4me1 loss and reduced** 40 _**FLC**_ **transcription. Multiple features of** _**COOLAIR**_ **co-transcriptional sensitivity therefore** 41 **act as thermosensory antennae, integrating distinct facets of winter temperature through** 42 **local chromatin regulation, to robustly repress** _**FLC**_ **in natural conditions.** 43 44 1 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.10.765395; this version posted November 12, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.10.765395) 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 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 45 **INTRODUCTION** 46 Organisms use environmental cues to coordinate developmental transitions with favourable 47 seasonal conditions. For temperature cues, the wide fluctuations over hourly and daily timescales, 48 are often greater than the average changes over a season, so how gene regulatory systems 49 integrate different aspects of temperature variation remains poorly understood. Vernalization, the 50 acceleration of flowering by prolonged winter cold, provides a tractable system for investigating 51 this problem. 52 In _Arabidopsis_, vernalization involves progressive repression of _FLOWERING LOCUS C_ ( _FLC_ ), 53 a major floral repressor [1,2] . During cold, _FLC_ transcription declines while Polycomb Repressive 54 Complex 2 (PRC2) establishes epigenetic silencing through deposition of histone H3 lysine 27 55 trimethylation (H3K27me3) [4] . This silencing is maintained after plants return to warm conditions [5] . 56 In parallel to Polycomb-mediated regulation, antisense transcription across _FLC_ produces 57 noncoding RNAs collectively termed _COOLAIR_ that contribute to _FLC_ repression during cold [3,6-] 58 9 . The antisense-associated repression pathway can occur over fast and slow timescales, whereas 59 the PRC2-dependent silencing at _FLC_ is a low probability switching mechanism, and thus always 60 slow [3,10] . 61 The mechanism by which _COOLAIR_ represses _FLC_ has been investigated through mutations and 62 natural variants. These affect _COOLAIR_ splicing and polyadenylation, altering _FLC_ expression 63 and flowering [11-13] . Previous studies in warm-grown plants have also linked _COOLAIR_ regulation 64 to the local chromatin state at _FLC_ . _COOLAIR_ co-transcriptional processing and histone 65 demethylation feedback on each other to affect productive _FLC_ transcription and the 66 establishment of Polycomb silencing [11-16] . These local chromatin changes include enhanced 67 contact between the _FLC_ promoter and its 3’ region [3,17], and formation of _COOLAIR_ RNA 68 condensates over the _FLC_ locus, termed ‘clouds’ [18] . RNA structure of specific _COOLAIR_ 69 isoforms also affects chromatin association and _FLC_ expression [8] . 70 A serendipitous finding during Arabidopsis vernalization experiments in natural and 71 climatically distinct field sites, was a strong induction of _COOLAIR_ expression the first night 72 temperatures dipped below zero [19] . This was confirmed in chamber experiments comparing 73 fluctuating and constant cold conditions that had the same mean temperature. _COOLAIR_ hyper74 expression correlated with stronger _FLC_ repression [19], but not with increased H3K27me3 75 accumulation [3] . We have therefore continued to investigate how sustained cold and brief 76 fluctuations below freezing affect _COOLAIR_ and _FLC_ transcription. We developed targeted 77 FLEP-seq (tFLEP-seq), which uses long-read sequencing of chromatin-associated RNA to 78 resolve splicing and transcript 3’ ends, and combined it with Pol II profiling, RNA imaging, 79 genetic perturbations and chromatin-contact mapping. Our findings show how the sensitivity of 80 _COOLAIR_ transcriptional processing enables plants to integrate different facets of the natural 81 temperature profile over the prolonged timescales of winter. Coupling noncoding co82 transcriptional sensitivity to environmental variation may be a general mechanism for robust 83 gene regulation. 84 2 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.10.765395; this version posted November 12, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.10.765395) 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 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 85 **Results** 86 **Cold induces temporally distinct** _**COOLAIR**_ **co-transcriptional changes across the locus** 87 Gradual reduction in _FLC_ expression during cold exposure results from reduced productive 88 transcriptional initiation [20] . Counterintuitively, this is associated with increased co-transcriptional 89 _FLC_ splicing, which is influenced by antisense ( _COOLAIR_ ) transcription [20] . We therefore 90 investigated the transcriptional dynamics on the _COOLAIR_ strand using native elongating 91 transcript sequencing [21] (NET-seq) over the vernalization time course analysed previously for 92 _FLC_ transcription and splicing (Figure S1A, B). RNA polymerase II (Pol II) levels on the 93 _COOLAIR_ strand were higher in the cold (Figure S1A, C), and its spatial distribution revealed 94 unexpected features. On the _FLC_ strand, Pol II accumulated predominantly at the _FLC_ 3’ end [20], 95 as observed for many _Arabidopsis_ genes [22,23] . On the _COOLAIR_ strand, however, Pol II 96 accumulated at two sites, indicative of slowed elongation or stalling (Figure S1A). The first, at 97 the 3’ end of the proximal _COOLAIR_ isoforms, increased and broadened after one week of cold 98 (Figure S1A). The second, which we term the _COOLAIR_ stalling site (CSS), occurs two-thirds 99 into the _COOLAIR_ transcription unit within the large intron, at a site not previously linked to 100 polyadenylation (Figure S1A). The CSS lies ~350 bp from the H3K27me3 nucleation region and 101 close to the vernalization response element [24] . This is conserved in Brassicaceae _FLC_ [9] and 102 deletion of the CSS region affects _FLC_ regulation [24,25] . Cold reduced the proportion of Pol II 103 signal beyond the CSS, in the region corresponding to the _FLC_ promoter and _COOLAIR_ class II 104 polyadenylation sites (Figure S1D), indicative of cold-enhanced termination at the CSS. Thus, 105 cold induced qualitative changes in Pol II occupancy on the _COOLAIR_ strand that differed 106 markedly from those on the _FLC_ strand. 107 108 We reasoned that these temperature-sensitive transcriptional dynamics would be linked to 109 changes in co-transcriptional processing. However, the short-read (Illumina) NET-seq data could 110 not resolve differential splicing or polyadenylation. We therefore developed targeted FLEP-seq 111 (tFLEP-seq), a long-read approach built on full-length elongating and polyadenylated RNA 112 sequencing (FLEP-seq) [26] . Nanopore sequencing of chromatin-associated RNA gives low per113 locus coverage, which is especially limiting for low-abundance transcripts such as _FLC_ and 114 _COOLAIR_ . tFLEP-seq overcomes this with in-solution hybridization capture, allowing us to 115 quantify poly(A) and non-poly(A) chromatin-associated RNA, map their 3’ ends and resolve 116 splicing patterns across the locus throughout vernalization (Figure 1A). We enriched _FLC_, 117 _COOLAIR_ and 15 other loci [20] from chromatin-associated RNA across a vernalization time 118 course, achieving ~500-fold enrichment over genome-wide background (Figure S2A). The 119 previously described _COOLAIR_ class I, II and III poly(A) isoforms were clearly detected, and 120 the sequencing depth resolved poly(A) site clusters, termination windows and splicing efficiency 121 in greater detail than before. 122 123 Non-poly(A) reads in the sense direction showed that _FLC_ transcription decreased gradually with 124 increasing cold exposure. Their 3’ ends accumulated predominantly at the _FLC_ transcription 125 termination site (Figure 1B), with higher 3’ end signal over exons [27] (Figure 1B). In the antisense 126 direction, despite the high read depth, we did not detect _COLDAIR_ or _COLDWRAP_, two non127 poly(A) lncRNAs previously described using PCR [28,29] . The non-poly(A) _COOLAIR_ 3’ end 128 profile differed markedly from that of _FLC_, with signal accumulating progressively in different 129 regions of the locus over 4 weeks of cold, then returning to non-vernalized levels after 10 days 130 in the warm (Figure 1E and Figure S3). This progressive increase is consistent with the gradual 131 cold induction of CBFs [30] and NTL8 [31], transcription factors that promote _COOLAIR_ transcription. 132 Significant accumulation of non-poly(A) 3’ ends at the proximal _COOLAIR_ poly(A) site and at 133 the CSS supports Pol II stalling at these sites, consistent with the NET-seq data (Figure S3). 3 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.10.765395; this version posted November 12, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.10.765395) 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 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 134 Moreover, _COOLAIR_ non-poly(A) fold changes exceeded those in the poly(A) fraction and in 135 NET-seq (Figure 1E). Whereas NET-seq reports Pol II occupancy, tFLEP-seq measures 136 chromatin-associated RNA more broadly; the larger non-poly(A) response therefore suggests 137 differential chromatin retention of _COOLAIR_ transcripts. 138 139 We next analysed _COOLAIR_ co-transcriptional changes in three windows across the locus: 140 proximal, gene body and distal. Over the constant-cold time course, _COOLAIR_ transcripts 141 accumulated in all three windows, although the early responses and polyadenylation patterns 142 differed between them (Figure 1B-D). Within one week of cold, poly(A) _COOLAIR_ increased in 143 the proximal and gene-body windows, whereas both poly(A) and non-poly(A) _COOLAIR_ 144 decreased in the distal window (Figure S2B and Figure 1F). This indicates cold-induced changes 145 in Pol II processivity and termination, consistent with the NET-seq data. Two weeks of cold 146 enhanced splicing of the first _COOLAIR_ intron (Figure 1G). Rapid polyadenylation changes and 147 slower splicing changes therefore differentially shift the proportions of _COOLAIR_ isoforms 148 known to repress _FLC_ . Despite the initial decrease in total distal poly(A) _COOLAIR_, distal class 149 II.ii _COOLAIR_, which represses _FLC_ by sequestering FRIGIDA in condensates [9], increased 150 rapidly in both absolute abundance and relative proportion. In contrast, distally spliced class II.i 151 _COOLAIR_, which directly influences _FLC_ transcription [8], and proximally spliced _COOLAIR_, 152 linked to transcription-coupled repression [15], increase more slowly. 153 154 **Co-transcriptional processing on the** _**COOLAIR**_ **strand captures signatures of fluctuating** 155 **temperatures** 156 _COOLAIR_ -mediated repression of _FLC_ transcription is particularly important under natural field 157 conditions, where dips below freezing super-induce _COOLAIR_ [19] . Given the co-transcriptional 158 sensitivity of _COOLAIR_ in constant cold, we analysed its response to a regime fluctuating to 159 below freezing (fluctuating strong, FS) [19], compared with constant cold at the same mean 160 temperature (5 °C) (Figure 2A). To determine whether FS alters _COOLAIR_ co-transcriptional 161 processing beyond its previously established transcriptional induction, we analysed chromatin162 associated RNA using tFLEP-seq. This revealed higher levels of both non-poly(A) and poly(A) 163 _COOLAIR_ in the proximal and gene-body windows than under constant cold. Most notably, both 164 non-poly(A) and poly(A) 3’ ends accumulated strongly around the CSS (Figure 2B-D and Figure 165 S4A). Poly(A) _COOLAIR_ transcripts terminating around the CSS, which we term class IV and 166 class V, made up only ~5-10% of chromatin-bound poly(A) _COOLAIR_ in constant cold but ~40% 167 under FS (Figure 2E). Class IV and V transcripts were enriched in the chromatin bound fraction 168 (Figure 2B, C) and, within the limits of read length, appeared unspliced (Figure 2B). Constant 169 cold enhanced both _COOLAIR_ expression and splicing efficiency, whereas fluctuations below 170 freezing super-induced _COOLAIR_ without further increasing splicing efficiency (Figure 2F). FS 171 therefore induces a distinct _COOLAIR_ co-transcriptional response, characterized by increased 172 termination around the CSS without a corresponding increase in splicing efficiency, rather than 173 simply amplifying the changes induced by constant cold. As predicted by modelling of field 174 data [32], FS did not enhance induction of _VIN3_ (Figure S4B), which encodes the Polycomb 175 accessory protein required for epigenetic silencing3. Consistent with previous evidence that FS 176 does not enhance H3K27me3 accumulation compared with constant cold [3], the additional 177 _COOLAIR_ -associated repression under FS is unlikely to act through enhanced H3K27me3 178 nucleation. 179 4 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.10.765395; this version posted November 12, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.10.765395) 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 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 180 **Cold-induced transcriptional changes enhance R-loop formation and** _**COOLAIR**_ **clouds** 181 The strong increase in chromatin-retained _COOLAIR_ under FS prompted us to investigate the 182 nature of this RNA. In warm conditions, _COOLAIR_ transcription forms an extensive RNA-DNA 183 hybrid over the proximal region that is associated with _FLC_ repression [33], and proximal 184 _COOLAIR_ is likely to be produced when this R-loop is resolved [34] . We used DRIPc (DNA-RNA 185 immunoprecipitation followed by cDNA conversion) to map R-loops strand-specifically. 186 Proximal R-loop levels increased in both constant cold and FS (Figure 3A), consistent with the 187 increase in chromatin-associated proximal _COOLAIR_ . We did not detect R-loops over the CSS, 188 suggesting that _COOLAIR_ retention at this site involves other mechanisms. 189 190 The increase in chromatin-retained _COOLAIR_ under FS detected by tFLEP-seq reflects an 191 average across many cells and _FLC_ copies. We therefore used single-molecule FISH (smFISH) 192 to determine whether this increase reflected greater accumulation at individual loci, a greater 193 fraction of cells with detectable _COOLAIR_ foci, or both. Previous smFISH showed that in 194 constant cold _COOLAIR_ forms large foci at the locus [18], termed ‘clouds’ by analogy with the _Xist_ 195 lncRNA that coats the inactive X chromosome [35] . Here we used _COOLAIR_ intronic probes that 196 also cover the CSS. As previously reported [24], _COOLAIR_ clouds were present in a small fraction 197 of cells under constant cold. Under FS, clouds increased in both size and intensity, indicating 198 greater _COOLAIR_ accumulation at loci with detectable clouds (Figure 3B, C). The number of 199 foci per nucleus and the fraction of cells with foci also increased (Figure 3D), consistent with 200 more frequent _COOLAIR_ transcription. Clouds form for _COOLAIR_ but not _FLC_ [18], and tFLEP201 seq detected more non-poly(A) _COOLAIR_ than non-poly(A) _FLC_ (Figure 1E). Together, these 202 observations suggest that the clouds comprise chromatin-retained non-poly(A) _COOLAIR_ at each 203 allele. Thus, constant cold and FS alter _COOLAIR_ co-transcriptional regulation (Pol II 204 processivity, termination and transcript release) through distinct modes at different sites within 205 the locus. 206 207 **Genetic perturbations distinguish contributions of** _**COOLAIR**_ **transcription and processing** 208 **to** _**FLC**_ **repression** 209 We next asked which _COOLAIR_ co-transcriptional changes are functionally important, using 210 mutants in which we had previously established a causal link between altered _COOLAIR_ 211 regulation and _FLC_ expression. NTL8 is a transcription factor that binds directly to the _COOLAIR_ 212 5’ region [19] . The dominant _ntl8-D3_ _FRI_ allele ectopically expresses high levels of _COOLAIR_, 213 resulting in _FLC_ hyper-repression [19] . tFLEP-seq confirmed high _COOLAIR_ expression in _ntl8-D3_ 214 _FRI_, which increased further after 2 weeks of constant cold (Figure 3E), accompanied by further 215 _FLC_ repression (Figure 3H). Non-poly(A) _COOLAIR_ increased significantly over the proximal 216 region and parts of the gene body but importantly not at the CSS (Figure 3E, Figure S4C). While 217 poly(A) _COOLAIR_ significantly increased over the proximal region, the proportion of class IV 218 and V poly(A) _COOLAIR_ did not increase in _ntl8-D3_ _FRI_ (Figure S4D). The increased proportion 219 of class IV and V polyadenylation under FS is therefore not simply a consequence of increased 220 _COOLAIR_ transcription. _COOLAIR_ clouds, which report gene-body chromatin-retained 221 transcripts, were enhanced in _ntl8-D3_ _FRI_ in the warm and further enhanced by constant cold 222 and FS (Figure 3F, G), with _FLC_ repression strongest under FS (Figure 3H). Together, these data 223 suggest that elevated _COOLAIR_ transcription is sufficient to enhance _FLC_ repression, whereas 224 additional FS-induced co-transcriptional changes, including those at the CSS, strengthen it 225 further. 226 5 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.10.765395; this version posted November 12, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.10.765395) 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 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 227 We had previously shown that a hypomorphic _smu1_ _FRI_ mutant disrupts cold-enhanced splicing 228 of the proximal intron and _COOLAIR_ proximal polyadenylation [12] . This effect was phenocopied 229 by cis mutation of the 3’ splice acceptor site, causally linking these _COOLAIR_ processing changes 230 to _FLC_ repression [12] . tFLEP-seq analysis of _smu1_ _FRI_ after 2 weeks of constant cold confirmed 231 reduced proximal intron splicing and proximal polyadenylation relative to Col _FRI_ (Figure S5A232 E), accompanied by impaired _FLC_ repression (Figure S5B). By contrast, under FS, _smu1_ _FRI_ 233 still showed reduced proximal splicing relative to Col _FRI_, to a similar extent as in constant cold 234 (Figure S5F), but proximal polyadenylation was not reduced relative to Col _FRI_, and _FLC_ 235 repression was retained (Figure S5G, I). Under FS, non-poly(A) _COOLAIR_ and poly(A) classes 236 IV and V also differed little between _smu1_ _FRI_ and Col _FRI_ (Figure S5H-J). These results 237 therefore show that reduced proximal splicing alone is insufficient to fully reduce _FLC_ repression 238 in FS, and that its functional effect depends on both the accompanying polyadenylation pattern 239 and the level of _COOLAIR_ transcription. 240 241 We then examined _aprf1-9_ _FRI_, which in the warm increases gene-body polyadenylation and 242 reduces distal polyadenylation of _COOLAIR_ [15], mimicking the cold-induced shift (Figure 1 and 243 2). APRF1 interacts with components of the phosphatase module of the CPF-like 3’ processing 244 complex that regulates _COOLAIR_ processing [15] . tFLEP-seq of _aprf1-9_ _FRI_ in the warm revealed 245 a polyadenylation shift consistent with previous data (Figure S6A). However, this shift does not 246 produce detectable _FLC_ repression in the warm [15], likely because _COOLAIR_ transcription remains 247 low. To probe the link between _COOLAIR_ polyadenylation and _FLC_ repression, we asked 248 whether the shift in _aprf1-9_ _FRI_ persists when cold induces _COOLAIR_, and whether it enhances 249 cold-induced _FLC_ repression. The polyadenylation shift persisted in constant cold (Figure S6A), 250 and _FLC_ repression was enhanced. Repression in _aprf1-9_ _FRI_ under constant cold was 251 comparable to that in wild type under FS (Figure 3I). This suggests that increased _COOLAIR_ 252 transcription in the cold strengthens the repressive effect of the polyadenylation shift by 253 increasing the frequency of processing and termination events at the favoured sites. Together 254 with the temperature-dependent _COOLAIR_ processing profiles, these results functionally 255 implicate polyadenylation and termination events associated with the CPF phosphatase module, 256 which occur predominantly in the proximal region in constant cold and extend into the gene body 257 under regimes fluctuating below freezing. 258 259 Together, these perturbations show that _COOLAIR_ -mediated _FLC_ repression depends on both the 260 level of _COOLAIR_ transcription and the accompanying pattern of co-transcriptional processing. 261 The functional consequences of altering a particular processing step therefore depend on the 262 broader transcriptional and processing response to the temperature regime. 263 264 _**COOLAIR**_ **co-transcriptional processing is linked to local chromatin interactions** 265 Pol II transcription can shape fine-scale chromatin organization at individual genes [36,37], including 266 in plants [38,39], where such interactions change in response to cold [40] . Chromosome conformation 267 capture (3C) previously showed that cold disrupts a gene loop linking the _FLC_ transcription start 268 site (TSS) and transcription termination site (TTS), which is also close to the _COOLAIR_ TSS [17], 269 and that this loop is affected by _COOLAIR_ transcription [3] . We therefore asked whether changes in 270 local chromatin interactions could link _COOLAIR_ co-transcriptional changes, and the associated 271 redistribution of Pol II and chromatin-retained RNA, to _FLC_ repression. We performed Micro 272 Capture-C [41] with a whole locus enrichment step (Methods), enabling single-gene-resolution 273 mapping of chromatin interactions across a 10-kb region around _FLC_ in warm, constant cold and 274 FS conditions (Figure 4A, Figure S6B). In the warm, the normalized contact map (Figure 4B, 275 upper triangles) revealed two major local interaction domains. One spans the _FLC_ gene from 6 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.10.765395; this version posted November 12, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.10.765395) 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 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 276 TSS to TTS. The other, upstream, covers the _FLC_ promoter and extends to the TTS of the 277 neighbouring gene _UPSTREAM OF FLC_ ( _UFC_ ; AT5G10150). Micro-C-XL previously detected 278 both domains in Col-0, where _FLC_ is epigenetically silenced [39], as subdomains of a larger 20-kb 279 domain in which the _FLC_ gene body contacts its upstream region; in those data, the promoter 280 domain was separate from _UFC_ . Both cold treatments disrupted the _FLC_ gene-domain 281 boundaries through the emergence of new interactions (Figure 4B and Figure S6C). To account 282 for distance-dependent decay and highlight local interaction patterns, we computed 283 observed/expected (O/E) matrices for each condition (Figure 4B, lower triangles). These showed 284 enriched interactions within the _FLC_ gene domain, consistent with the gene loop previously 285 detected by 3C [17] . Both constant cold and FS reduced interactions within the gene domain, 286 correlating with significantly reduced _FLC_ transcription (Figure 4B and Figure S6D). 287 288 The O/E matrices also revealed prominent contacts between the promoter domain and _FLC_ gene 289 regions enriched for _COOLAIR_ poly(A) sites; promoter-contact O/E values were consistently 290 elevated at regions of preferential _COOLAIR_ polyadenylation (Figure 4C and Figure S6E). In the 291 warm, these contacts were concentrated over the proximal _COOLAIR_ poly(A) region. Constant 292 cold enhanced contacts with both proximal and gene-body regions, paralleling the cold-induced 293 shift in _COOLAIR_ polyadenylation. This correspondence was strongest under FS, when a 294 prominent promoter contact emerged over the CSS (Figure 4D). This is the site where Pol II 295 accumulated in the cold and where, under FS, non-poly(A) and class V poly(A) _COOLAIR_ 296 increased. Together, these observations support a model in which the spatial pattern of _COOLAIR_ 297 3’ processing and Pol II termination helps determine where promoter-gene contacts form. 298 Whereas termination-associated gene looping has been described previously [42-45], our data link 299 this mechanism to environmentally regulated antisense transcription and to dynamic 300 repositioning of contacts within a single gene. 301 302 To test this association genetically, we examined three independent perturbations. The first was 303 a _COOLAIR_ mutant ( _FLCΔCOOLAIR_ ) that disrupts cold induction of _COOLAIR_ 304 transcription [9,19,46] . The second was _ntl8-D3_ _FRI_, which constitutively increases _COOLAIR_ 305 transcription, and the third was _aprf1-9_ _FRI_, which enhances gene-body _COOLAIR_ 306 polyadenylation. For each, we quantified the proportion of all observed contacts that occurred 307 between the promoter domain and the _FLC_ gene. In the _COOLAIR_ mutant, this proportion 308 remained significantly lower than in Col _FRI_, even under FS (Figure 5A, C), consistent with the 309 negligible poly(A) _COOLAIR_ detected over the locus (Figure S7A). _ntl8-D3_ _FRI_ did not increase 310 the promoter-gene contact proportion relative to Col _FRI_ in either warm or cold, despite 311 significantly higher _COOLAIR_ transcription (Figure 5C and Figure S7B, C). Increased _COOLAIR_ 312 transcription alone is therefore insufficient to enhance promoter-gene interactions. By contrast, 313 _aprf1-9_ _FRI_ increased the promoter-gene contact proportion in both warm and cold, 314 corresponding to its increased gene-body _COOLAIR_ polyadenylation (Figure 5A, B, D). In _aprf1-_ 315 _9_ _FRI_, elevated promoter contacts again coincided with regions of preferential _COOLAIR_ 316 polyadenylation in both conditions (Figure 5E). This shows that genetically redistributing 317 _COOLAIR_ 3’ processing also redistributes promoter contacts. Together, these perturbations 318 indicate that promoter contacts track the spatial pattern of _COOLAIR_ 3’ processing rather than 319 _COOLAIR_ transcription level. 320 321 **Distinct temperature regimes induce different local chromatin modification patterns** 322 We then asked how these co-transcriptional and chromatin-interaction changes relate to 323 chromatin modifications and _FLC_ expression. We previously showed that Pol II termination 324 promotes a repressive chromatin state at _FLC_ through H3K4me1 demethylation [15,16,47], together 7 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.10.765395; this version posted November 12, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.10.765395) 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 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 325 with the activity of APRF1 and the phosphatase module of the CPF-like complex [15] . We measured 326 H3K4me1 across the locus under FS and compared it with constant-cold time-course data [3] . In 327 both regimes, H3K4me1 decreased over the proximal and gene-body regions (Figure 6A, B). 328 This is consistent with increased conversion to H3K4me3 in the proximal window, associated 329 with increased _COOLAIR_ transcription (Figure S8A). It is also consistent with reduced co330 transcriptional deposition of H3K4me1 in the gene body, owing to reduced sense transcription 331 (Figure S8B). Other chromatin changes were detectable after one week of constant cold, 332 including reduced H3 acetylation (Figure S8C) and repositioning of the +1 nucleosome [48] . In a 333 distinct genetic context, nucleosome repositioning at _FLC_ also accompanied _COOLAIR_ co334 transcriptional changes in _ntl8-D3_ in the absence of cold (Figure S9A, B). This supports a 335 mechanistic link between antisense co-transcriptional changes and local nucleosome dynamics, 336 as reported in yeast [49,50] . 337 338 A clear difference emerged over the distal _COOLAIR_ / _FLC_ promoter region, immediately 339 downstream of the CSS. After prolonged constant cold (4 and 6 weeks), H3K4me1 [3] (Figure 6A, 340 B) and H3K4me2 [51] (Figure S8A) increased in the distal region. This is consistent with reduced 341 conversion of H3K4me1 to H3K4me2, and of H3K4me2 to H3K4me3, as initiation of _FLC_ sense 342 transcription declines. Under FS, however, although _FLC_ repression was similar in magnitude to 343 that after 4 weeks of constant cold, distal H3K4me1 did not increase but instead decreased 344 slightly (Figure 6A, B). These changes were disrupted in the _COOLAIR_ -defective TEX1.0 line 345 (Figure 6C), indicating that they depend on _COOLAIR_ and supporting a role for antisense co346 transcriptional processing in shaping local chromatin state. The loss of H3K4me1 over the distal 347 _COOLAIR_ region and _FLC_ 5’ end under FS suggests active demethylation, linked to enhanced 348 _COOLAIR_ termination at the CSS and the associated changes in local chromatin topology. 349 **Discussion** 350 Understanding how organisms integrate complex, fluctuating environmental signals is a 351 fundamental problem in biology. We addressed this question by studying the cold-induced 352 repression of the _Arabidopsis_ floral repressor _FLC_ . Our key finding is that distinct _COOLAIR_ co353 transcriptional processes act as multiple temperature antennae operating over different 354 timescales, providing a mechanism to integrate different features of winter temperature. Beyond 355 changes in splicing and polyadenylation, temperature strongly affected Pol II stalling and 356 processivity, and the retention of both non-poly(A) and poly(A) transcripts on chromatin. Genetic 357 perturbations further show that _COOLAIR_ -mediated _FLC_ repression depends on both the level 358 of _COOLAIR_ transcription and the accompanying pattern of co-transcriptional processing. The 359 greater sensitivity of _COOLAIR_ transcription than _FLC_ sense transcription to temperature 360 fluctuations may reflect differences in the factors loaded onto Pol II at initiation [52] . 361 362 These changes in Pol II dynamics and _COOLAIR_ processing are associated with altered local 363 chromatin interactions. Genetic perturbations suggest that the genomic distribution of _FLC_ 364 promoter contacts is associated with the locations of preferential _COOLAIR_ 3’ processing. 365 Increased _COOLAIR_ transcription does not necessarily increase the proportion of promoter–gene 366 contacts, while contacts can be enriched even when _COOLAIR_ transcription remains low, as in 367 warm _aprf1-9_ . In that genotype, contact enrichment is not accompanied by detectable _FLC_ 368 repression, indicating that increased contacts alone are insufficient for repression. A possible 369 physical basis for these processing-associated contacts is interactions between termination370 associated Pol II complexes [53], with Pol II terminating at the CSS interacting with Pol II 371 terminating across the _FLC_ promoter domain. Consistent with this, FS generates non-poly(A) 372 _COOLAIR_ 3’ ends across the promoter domain, including near the _UFC_ TTS (Figure S6F). The 8 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.10.765395; this version posted November 12, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.10.765395) 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 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 373 stronger contacts under FS than in constant cold coincide with the pronounced redistribution of 374 _COOLAIR_ termination towards the CSS. Repeated temperature changes may also favour more 375 transient transcriptional states than those occurring during prolonged constant cold, potentially 376 contributing to the stronger contacts under FS. These altered interactions could in turn influence 377 the recruitment or activity of chromatin-modifying factors at _FLC_ . 378 379 The distinct _COOLAIR_ responses to constant cold and FS are accompanied by differences in 380 local chromatin modifications. The contrasting distal H3K4me1 responses despite comparable 381 _FLC_ repression suggest that similar transcriptional outcomes can accompany different local 382 chromatin states. Enhanced _COOLAIR_ termination near the CSS may contribute to H3K4me1 383 demethylation under FS, consistent with previous links between Pol II termination and 384 demethylation [15] . H3K4me1 demethylation is one possible component of a broader chromatin 385 response that may also involve other histone modifications and changes in chromatin 386 organization. Such flexibility in temperature-dependent _COOLAIR_ regulation enables robust 387 _FLC_ repression across winters with different temperature profiles (Figure 6D). Related 388 antisense co-transcriptional responses are likely to also occur at other environmentally 389 responsive genes. Genome-wide 3’ mRNA sequencing (Quant-Seq) revealed that fluctuations 390 below freezing were associated with changed polyadenylation patterns for cold-induced 391 antisense transcription across hundreds of genes (Figure S9C). Similar changes were also 392 observed for protein-coding gene transcription genome wide (Figure S9D, E). Co393 transcriptional plasticity may therefore be a widespread mechanism for temporally integrating 394 naturally fluctuating signals to achieve robust gene regulation. 395 396 9 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.10.765395; this version posted November 12, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.10.765395) 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 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 397 **References** 398 399 1. Michaels, S.D., and Amasino, R.M. (1999). _FLOWERING LOCUS C_ Encodes a Novel 400 MADS Domain Protein That Acts as a Repressor of Flowering. The Plant Cell _11_, 949. 401 10.1105/tpc.11.5.949. 402 2. Sheldon, C.C., Rouse, D.T., Finnegan, E.J., Peacock, W.J., and Dennis, E.S. (2000). 403 The molecular basis of vernalization: The central role of _FLOWERING LOCUS C_ 404 ( _FLC_ ). Proceedings of the National Academy of Sciences _97_, 3753-3758. 405 doi:10.1073/pnas.97.7.3753. 406 3. Nielsen, M., Menon, G., Zhao, Y., Mateo-Bonmati, E., Wolff, P., Zhou, S., Howard, 407 M., and Dean, C. (2024). _COOLAIR_ and PRC2 function in parallel to silence _FLC_ 408 during vernalization. Proceedings of the National Academy of Sciences _121_, 409 e2311474121. 10.1073/pnas.2311474121. 410 4. Gendall, A.R., Levy, Y.Y., Wilson, A., and Dean, C. (2001). The VERNALIZATION 2 411 Gene Mediates the Epigenetic Regulation of Vernalization in Arabidopsis. Cell _107_, 412 525-535. 10.1016/S0092-8674(01)00573-6. 413 5. Finnegan, E.J., and Dennis, E.S. (2007). Vernalization-Induced Trimethylation of 414 Histone H3 Lysine 27 at _FLC_ Is Not Maintained in Mitotically Quiescent Cells. 415 Current Biology _17_, 1978-1983. 10.1016/j.cub.2007.10.026. 416 6. Liu, F., Marquardt, S., Lister, C., Swiezewski, S., and Dean, C. (2010). Targeted 3’ 417 Processing of Antisense Transcripts Triggers _Arabidopsis FLC_ Chromatin Silencing. 418 Science _327_, 94-97. doi:10.1126/science.1180278. 419 7. Swiezewski, S., Liu, F., Magusin, A., and Dean, C. (2009). Cold-induced silencing by 420 long antisense transcripts of an Arabidopsis Polycomb target. Nature _462_, 799-802. 421 10.1038/nature08618. 422 8. Yang, M., Zhu, P., Cheema, J., Bloomer, R., Mikulski, P., Liu, Q., Zhang, Y., Dean, C., 423 and Ding, Y. (2022). _In vivo_ single-molecule analysis reveals _COOLAIR_ RNA 424 structural diversity. Nature _609_, 394-399. 10.1038/s41586-022-05135-9. 425 9. Zhu, P., Lister, C., and Dean, C. (2021). Cold-induced Arabidopsis FRIGIDA nuclear 426 condensates for _FLC_ repression. Nature _599_, 657-661. 10.1038/s41586-021-04062-5. 427 10. Berry, S., Hartley, M., Olsson, T.S.G., Dean, C., and Howard, M. (2015). Local 428 chromatin environment of a Polycomb target gene instructs its own epigenetic 429 inheritance. eLife _4_, e07205. 10.7554/eLife.07205. 430 11. Li, P., Tao, Z., and Dean, C. (2015). Phenotypic evolution through variation in splicing 431 of the noncoding RNA _COOLAIR_ . Genes & Development _29_, 696-701. 432 10.1101/gad.258814.115. 433 12. Long, X., Cai, Y., Wang, H., Liu, Y., Huang, X., Xuan, H., Li, W., Zhang, X., Zhang, 434 H., Fang, X., et al. (2024). Cotranscriptional splicing is required in the cold to produce 435 _COOLAIR_ isoforms that repress Arabidopsis _FLC_ . Proceedings of the National 436 Academy of Sciences _121_, e2407628121. doi:10.1073/pnas.2407628121. 437 13. Marquardt, S., Raitskin, O., Wu, Z., Liu, F., Sun, Q., and Dean, C. (2014). Functional 438 Consequences of Splicing of the Antisense Transcript _COOLAIR_ on _FLC_ 439 Transcription. Molecular Cell _54_, 156-165. 440 [https://doi.org/10.1016/j.molcel.2014.03.026.](https://doi.org/10.1016/j.molcel.2014.03.026) 441 14. Fang, X., Wu, Z., Raitskin, O., Webb, K., Voigt, P., Lu, T., Howard, M., and Dean, C. 442 (2020). The 3′ processing of antisense RNAs physically links to chromatin-based 443 transcriptional control. Proceedings of the National Academy of Sciences _117_, 15316444 15321. doi:10.1073/pnas.2007268117. 445 15. Mateo-Bonmatí, E., Montez, M., Maple, R., Fiedler, M., Fang, X., Saalbach, G., 446 Passmore, L.A., and Dean, C. (2024). A CPF-like phosphatase module links 10 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.10.765395; this version posted November 12, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.10.765395) 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 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 447 transcription termination to chromatin silencing. Molecular Cell _84_, 2272-2286.e2277. 448 10.1016/j.molcel.2024.05.016. 449 16. Menon, G., Mateo-Bonmati, E., Reeck, S., Maple, R., Wu, Z., Ietswaart, R., Dean, C., 450 and Howard, M. (2024). Proximal termination generates a transcriptional state that 451 determines the rate of establishment of Polycomb silencing. Molecular Cell _84_, 2255452 [2271.e2259. https://doi.org/10.1016/j.molcel.2024.05.014.](https://doi.org/10.1016/j.molcel.2024.05.014) 453 17. Crevillén, P., Sonmez, C., Wu, Z., and Dean, C. (2013). A gene loop containing the 454 floral repressor _FLC_ is disrupted in the early phase of vernalization. The EMBO 455 Journal _32_, 140-148. 10.1038/emboj.2012.324. 456 18. Rosa, S., Duncan, S., and Dean, C. (2016). Mutually exclusive sense–antisense 457 transcription at _FLC_ facilitates environmentally induced gene repression. Nature 458 Communications _7_, 13031. 10.1038/ncomms13031. 459 19. Zhao, Y., Zhu, P., Hepworth, J., Bloomer, R., Antoniou-Kourounioti, R.L., Doughty, J., 460 Heckmann, A., Xu, C., Yang, H., and Dean, C. (2021). Natural temperature 461 fluctuations promote _COOLAIR_ regulation of _FLC_ . Genes & Development _35_, 888462 898. 10.1101/gad.348362.121. 463 20. Maple, R., Menon, G., Duncan, S., Schulten, A., Sindalovskaya, M., Yang, H., 464 Bloomer, R., Howard, M., and Dean, C. (2026). Co-transcriptional splicing changes 465 combine with reduced productive transcription initiation for cold-induced repression 466 of _FLC_ . Nucleic Acids Research _54_ . 10.1093/nar/gkaf1462. 467 21. Churchman, L.S., and Weissman, J.S. (2011). Nascent transcript sequencing visualizes 468 transcription at nucleotide resolution. Nature _469_, 368-373. 10.1038/nature09652. 469 22. Kindgren, P., Ivanov, M., and Marquardt, S. (2020). Native elongation transcript 470 sequencing reveals temperature dependent dynamics of nascent RNAPII transcription 471 in Arabidopsis. Nucleic Acids Res _48_, 2332-2347. 10.1093/nar/gkz1189. 472 23. Zhu, J., Liu, M., Liu, X., and Dong, Z. (2018). RNA polymerase II activity revealed by 473 GRO-seq and pNET-seq in Arabidopsis. Nature Plants _4_, 1112-1123. 10.1038/s41477474 018-0280-0. 475 24. Sung, S., He, Y., Eshoo, T.W., Tamada, Y., Johnson, L., Nakahigashi, K., Goto, K., 476 Jacobsen, S.E., and Amasino, R.M. (2006). Epigenetic maintenance of the vernalized 477 state in _Arabidopsis thaliana_ requires LIKE HETEROCHROMATIN PROTEIN 1. 478 Nature Genetics _38_, 706-710. 10.1038/ng1795. 479 25. Sheldon, C.C., Conn, A.B., Dennis, E.S., and Peacock, W.J. (2002). Different 480 Regulatory Regions Are Required for the Vernalization-Induced Repression of 481 _FLOWERING LOCUS C_ and for the Epigenetic Maintenance of Repression. The Plant 482 Cell _14_, 2527-2537. 10.1105/tpc.004564. 483 26. Long, Y., Jia, J., Mo, W., Jin, X., and Zhai, J. (2021). FLEP-seq: simultaneous 484 detection of RNA polymerase II position, splicing status, polyadenylation site and 485 poly(A) tail length at genome-wide scale by single-molecule nascent RNA sequencing. 486 Nature Protocols _16_, 4355-4381. 10.1038/s41596-021-00581-7. 487 27. Mimoso, C.A., and Adelman, K. (2023). U1 snRNP increases RNA Pol II elongation 488 rate to enable synthesis of long genes. Molecular Cell _83_, 1264-1279.e1210. 489 [https://doi.org/10.1016/j.molcel.2023.03.002.](https://doi.org/10.1016/j.molcel.2023.03.002) 490 28. Heo, J.B., and Sung, S. (2011). Vernalization-Mediated Epigenetic Silencing by a 491 Long Intronic Noncoding RNA. Science _331_, 76-79. doi:10.1126/science.1197349. 492 29. Kim, D.-H., and Sung, S. (2017). Vernalization-Triggered Intragenic Chromatin Loop 493 Formation by Long Noncoding RNAs. Developmental Cell _40_, 302-312.e304. 494 10.1016/j.devcel.2016.12.021. 11 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.10.765395; this version posted November 12, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.10.765395) 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 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 495 30. Jeon, M., Jeong, G., Yang, Y., Luo, X., Jeong, D., Kyung, J., Hyun, Y., He, Y., and 496 Lee, I. (2023). Vernalization-triggered expression of the antisense transcript _COOLAIR_ 497 is mediated by CBF genes. eLife _12_, e84594. 10.7554/eLife.84594. 498 31. Zhao, Y., Antoniou-Kourounioti, R.L., Calder, G., Dean, C., and Howard, M. (2020). 499 Temperature-dependent growth contributes to long-term cold sensing. Nature _583_, 500 825-829. 10.1038/s41586-020-2485-4. 501 32. Antoniou-Kourounioti, R.L., Hepworth, J., Heckmann, A., Duncan, S., Qüesta, J., 502 Rosa, S., Säll, T., Holm, S., Dean, C., and Howard, M. (2018). Temperature Sensing Is 503 Distributed throughout the Regulatory Network that Controls _FLC_ Epigenetic 504 Silencing in Vernalization. Cell Syst _7_, 643-655.e649. 10.1016/j.cels.2018.10.011. 505 33. Sun, Q., Csorba, T., Skourti-Stathaki, K., Proudfoot, N.J., and Dean, C. (2013). R506 Loop Stabilization Represses Antisense Transcription at the _Arabidopsis FLC_ Locus. 507 Science _340_, 619-621. doi:10.1126/science.1234848. 508 34. Xu, C., Wu, Z., Duan, H.C., Fang, X., Jia, G., and Dean, C. (2021). R-loop resolution 509 promotes co-transcriptional chromatin silencing. Nat Commun _12_, 1790. 510 10.1038/s41467-021-22083-6. 511 35. Penny, G.D., Kay, G.F., Sheardown, S.A., Rastan, S., and Brockdorff, N. (1996). 512 Requirement for Xist in X chromosome inactivation. Nature _379_, 131-137. 513 10.1038/379131a0. 514 36. Hsieh, T.-H.S., Cattoglio, C., Slobodyanyuk, E., Hansen, A.S., Rando, O.J., Tjian, R., 515 and Darzacq, X. (2020). Resolving the 3D Landscape of Transcription-Linked 516 Mammalian Chromatin Folding. Molecular Cell _78_, 539-553.e538. 517 [https://doi.org/10.1016/j.molcel.2020.03.002.](https://doi.org/10.1016/j.molcel.2020.03.002) 518 37. Salari, H., Fourel, G., and Jost, D. (2024). Transcription regulates the spatio-temporal 519 dynamics of genes through micro-compartmentalization. Nature Communications _15_, 520 5393. 10.1038/s41467-024-49727-7. 521 38. Liu, C., Wang, C., Wang, G., Becker, C., Zaidem, M., and Weigel, D. (2016). Genome522 wide analysis of chromatin packing in _Arabidopsis thaliana_ at single-gene resolution. 523 Genome Research _26_, 1057-1068. 10.1101/gr.204032.116. 524 39. Sun, L., Zhou, J., Xu, X., Liu, Y., Ma, N., Liu, Y., Nie, W., Zou, L., Deng, X.W., and 525 He, H. (2024). Mapping nucleosome-resolution chromatin organization and enhancer526 promoter loops in plants using Micro-C-XL. Nature Communications _15_, 35. 527 10.1038/s41467-023-44347-z. 528 40. Zhang, Y., Dong, Q., Wang, Z., Liu, Q., Yu, H., Sun, W., Cheema, J., You, Q., Ding, 529 L., Cao, X., et al. (2024). A fine-scale Arabidopsis chromatin landscape reveals 530 chromatin conformation-associated transcriptional dynamics. Nature Communications 531 _15_, 3253. 10.1038/s41467-024-47678-7. 532 41. Hua, P., Badat, M., Hanssen, L.L.P., Hentges, L.D., Crump, N., Downes, D.J., 533 Jeziorska, D.M., Oudelaar, A.M., Schwessinger, R., Taylor, S., et al. (2021). Defining 534 genome architecture at base-pair resolution. Nature _595_, 125-129. 10.1038/s41586535 021-03639-4. 536 42. Ansari, A., and Hampsey, M. (2005). A role for the CPF 3′-end processing machinery 537 in RNAP II-dependent gene looping. Genes & Development _19_, 2969-2978. 538 10.1101/gad.1362305. 539 43. Lamas-Maceiras, M., Singh, B.N., Hampsey, M., and Freire-Picos, M.A. (2016). 540 Promoter-Terminator Gene Loops Affect Alternative 3′-End Processing in Yeast. 541 Journal of Biological Chemistry _291_, 8960-8968. 542 [https://doi.org/10.1074/jbc.M115.687491.](https://doi.org/10.1074/jbc.M115.687491) 12 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.10.765395; this version posted November 12, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.10.765395) 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 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 543 44. O'Sullivan, J.M., Tan-Wong, S.M., Morillon, A., Lee, B., Coles, J., Mellor, J., and 544 Proudfoot, N.J. (2004). Gene loops juxtapose promoters and terminators in yeast. 545 Nature Genetics _36_, 1014-1018. 10.1038/ng1411. 546 45. Tan-Wong, S.M., Zaugg, J.B., Camblong, J., Xu, Z., Zhang, D.W., Mischo, H.E., 547 Ansari, A.Z., Luscombe, N.M., Steinmetz, L.M., and Proudfoot, N.J. (2012). Gene 548 Loops Enhance Transcriptional Directionality. Science _338_, 671-675. 549 doi:10.1126/science.1224350. 550 46. Luo, X., Chen, T., Zeng, X., He, D., and He, Y. (2019). Feedback Regulation of _FLC_ 551 by FLOWERING LOCUS T (FT) and FD through a 5′ _FLC_ Promoter Region in 552 Arabidopsis. Molecular Plant _12_, 285-288. 553 [https://doi.org/10.1016/j.molp.2019.01.013.](https://doi.org/10.1016/j.molp.2019.01.013) 554 47. Oya, S., Takahashi, M., Takashima, K., Kakutani, T., and Inagaki, S. (2022). 555 Transcription-coupled and epigenome-encoded mechanisms direct H3K4 methylation. 556 Nature Communications _13_, 4521. 10.1038/s41467-022-32165-8. 557 48. Montez, M., Zhu, D., Huertas, J., Maristany, M.J., Rutjens, B., Nielsen, M., 558 Collepardo-Guevara, R., and Dean, C. (2025). Cold-induced nucleosome dynamics 559 linked to silencing of Arabidopsis _FLC_ . Nature Communications _16_, 5550. 560 10.1038/s41467-025-60735-z. 561 49. Gill, J.K., Maffioletti, A., García-Molinero, V., Stutz, F., and Soudet, J. (2020). Fine 562 Chromatin-Driven Mechanism of Transcription Interference by Antisense Noncoding 563 Transcription. Cell Reports _31_, 107612. [https://doi.org/10.1016/j.celrep.2020.107612.](https://doi.org/10.1016/j.celrep.2020.107612) 564 50. Korber, P., and Barbaric, S. (2014). The yeast PHO5 promoter: from single locus to 565 systems biology of a paradigm for gene regulation through chromatin. Nucleic Acids 566 Research _42_, 10888-10902. 10.1093/nar/gku784. 567 51. Yang, H., Howard, M., and Dean, C. (2014). Antagonistic roles for H3K36me3 and 568 H3K27me3 in the cold-induced epigenetic switch at _Arabidopsis_ _FLC_ . Curr Biol _24_, 569 1793-1797. 10.1016/j.cub.2014.06.047. 570 52. Cacioppo, R., Gillis, A., Shlamovitz, I., Zeller, A., Castiblanco, D., Crisp, A., 571 Haworth, B., Arabiotorre, A., Abyaneh, P., Bao, Y., et al. (2024). CRL3 [ARMC5] ubiquitin 572 ligase and Integrator phosphatase form parallel mechanisms to control early stages of 573 RNA Pol II transcription. Molecular Cell _84_, 4808-4823.e4813. 574 10.1016/j.molcel.2024.11.024. 575 53. Carminati, M., Rodríguez-Molina, J.B., Manav, M.C., Bellini, D., and Passmore, L.A. 576 (2023). A direct interaction between CPF and RNA Pol II links RNA 3’ end processing 577 to transcription. Molecular Cell _83_, 4461-4478.e4413. 10.1016/j.molcel.2023.11.004. 578 579 13 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.10.765395; this version posted November 12, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.10.765395) 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 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 580 585 vernalized, 1W: one week cold, 2W: 2 weeks cold, 4W: four weeks cold, 4WT10: four weeks 586 cold and ten days post-cold. 587 14 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.10.765395; this version posted November 12, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.10.765395) 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 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 588 **B** . Distribution of non-poly(A) and poly(A) RNA sequencing reads across the _FLC_ locus over 589 the vernalization time course. The inset bar plot (top right) quantifies the number of reads of 590 _COOLAIR_ antisense transcripts and _FLC_ sense transcripts. The black boxes and lines indicate 591 gene models. 592 **C** . Poly(A) transcript changes in the three windows over the vernalization time course. See 593 methods for window definitions. Poly(A) read counts were normalized to the geometric mean of 594 _ACT7_ and _UBC21_ and then normalized to the NV (non-vernalized) condition to show changes in 595 transcript levels relative to NV. 596 **D** . Relative abundance (normalized to the geometric mean of _ACT7_ and _UBC21_ ) of poly(A) 597 transcripts detected by tFLEP-seq, in different parts of _COOLAIR_ locus in NV condition. 598 **E** . Normalized read counts of _FLC_ and _COOLAIR_ over the vernalization time course. Reads were 599 normalized to the geometric mean of _ACT7_ and _UBC21_ as done for the _FLC_ strand [20] . 600 **F** . Proportion of _COOLAIR_ non-poly(A) reads with 3’ ends in the distal window calculated 601 relative to the total _COOLAIR_ non-poly(A) reads across the locus, for each timepoint. 602 **G** . Overall ratio of spliced/unspliced reads in non-poly(A) transcripts from tFLEP-seq at 603 _COOLAIR_ strand over the time course. _n_ = 3 biological replicates at all timepoints. Error bars 604 indicate mean ± standard deviation in panels **C-G** . Statistical comparisons were performed using 605 Tukey’s HSD (ns indicates no significance; * indicates adjusted _p_ < 0.05; ** indicates adjusted 606 _p_ < 0.01). 607 15 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.10.765395; this version posted November 12, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.10.765395) 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 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 608 621 conditions. Read signal was normalized to the geometric mean of _ACT7_ and _UBC21_ . 622 **E** . Proportion of _COOLAIR_ poly(A) reads with 3’ ends in the proximal, gene body and distal 623 window calculated relative to the total _COOLAIR_ poly(A) reads across the locus. See Figure 1 624 for window definition. _n_ = 3 biological replicates at all timepoints. Statistical comparisons of the 16 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.10.765395; this version posted November 12, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.10.765395) 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 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 625 means were performed using Tukey’s HSD (ns indicates no significance; ** indicates adjusted _p_ 626 < 0.01). 627 **F** . Ratio of spliced/unspliced reads in non-poly(A) transcripts from tFLEP-seq at _COOLAIR_ 628 strand in constant cold and FS conditions. _n_ = 3 biological replicates at all timepoints. Error bars 629 indicate mean ± standard deviation in panels **E-F** . Statistical comparison of the means was 630 performed using a two-sided Welch’s t-test (** indicates _p_ < 0.01). 631 17 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.10.765395; this version posted November 12, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.10.765395) 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 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 632 640 with reduced DAPI staining are indicative of nucleoli. Scale bars = 5 µm. 641 **C** . _COOLAIR_ foci quantification in Col _FRI_ by smFISH. Dots indicate foci volume (μm [3] ) (left) 642 and normalized intensity (a.u.) (right). Constant cold samples (blue), FS samples (red). Error bars 643 indicate mean ± standard deviation. Number of cells in each sample indicated. In both cases, a 18 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.10.765395; this version posted November 12, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.10.765395) 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 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 644 one-sided Welch’s t-test was used to compare the means (* indicates _p_ < 0.05; *** indicates _p_ < 645 0.001). 646 **D** . _COOLAIR_ foci in cold conditions quantified in Col _FRI._ Left panel: Proportion of cells 647 showing _COOLAIR_ foci, measured by Whole-mount smFISH in root tips (see Methods). Right 648 panel: Number of _COOLAIR_ foci per nucleus measured Whole-mount smFISH in root tips. n 649 indicates the number of cells included in the quantification. Constant cold condition: blue; FS 650 condition: red. In both cases, a one-sided Welch’s t-test was used to compare the means (* 651 indicates _p_ < 0.05; *** indicates _p_ < 0.001). 652 **E** . Distribution of chromatin-bound non-poly(A) and chromatin-bound poly(A) reads of _ntl8-D3_ 653 _FRI_ across the _FLC_ locus in constant cold condition. The inset bar plot (top right) quantifies the 654 number of reads of _COOLAIR_ antisense transcripts and _FLC_ sense transcripts. 655 **F** . _COOLAIR_ foci per nucleus quantified in _ntl8-D3_ by Whole-mount smFISH in root tips. Non656 vernalized: green; Constant cold condition: blue; FS condition: red. One-sided Welch’s t-tests 657 were used to compare the means and _P_ -values were adjusted using a Bonferroni correction (* 658 indicates adjusted _p_ < 0.05; *** indicates adjusted _p_ < 0.001). 659 **G** . Comparing the proportion of _COOLAIR_ positive cells in Col _FRI_ and _ntl8-D3_ measured by 660 Whole-mount smFISH in root tips under different conditions. Col _FRI_ : blue; _ntl8-D3_ : red. 661 **H** . _FLC_ transcriptional output (normalized to the geometric mean of _PP2A_ and _UBC_ ) between 662 FS and constant cold conditions in Col _FRI_ and _ntl8_ - _D3_ _FRI_ mutant, using a two-sided Student’s 663 t-test (** indicates _p_ < 0.01; *** indicates _p_ < 0.001). 664 **I** . Comparison of changes in _FLC_ transcriptional output measured by qPCR between FS and 665 constant cold conditions in Col _FRI_ and _aprf1-9_ _FRI_ mutant. In each condition, log 2 FC for _FLC_ 666 spliced transcript in Col _FRI_ and _aprf1-9_ _FRI_ were compared using a two-sided Student’s t-test 667 (** indicates _p_ < 0.01; *** indicates _p_ < 0.001). 668 19 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.10.765395; this version posted November 12, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.10.765395) 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 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 669 681 transcripts are also shown as coverage tracks aligned to the contact map coordinates. Read signal 682 was normalized to the geometric mean of _ACT7_ and _UBC21_ . 683 **C** . Promoter-contact observed/expected (OE) values for regions with and without detected 684 _COOLAIR_ polyadenylation in NV, constant cold 2W and FS 2W conditions. Points represent 685 individual 200-bp bins; boxes show the median and interquartile range, with whiskers extending 20 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.10.765395; this version posted November 12, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.10.765395) 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 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 686 to 1.5 times the interquartile range. The grey line indicates OE = 1. AUC (area under the receiver 687 operating characteristic curve) gives the probability that a randomly chosen poly(A)-positive 688 region has higher OE than a randomly chosen poly(A)-negative region; _P_ values were obtained 689 by randomly reassigning the observed number of poly(A)-positive labels among the 200-bp bins 690 and recalculating the AUC. 691 **D** . Micro-Capture-C contact comparisons between the CSS and the 10-kb region under the 692 indicated conditions. Interaction fold change was calculated from observed/expected (OE) 693 matrices normalized for genomic distance. Red and blue indicate increase and decrease in 694 interactions, respectively, and the line width represents the magnitude of the fold change. 695 696 21 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.10.765395; this version posted November 12, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.10.765395) 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 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 697 712 conditions. 713 **D** . Proportion of interactions between the promoter region and the _FLC_ gene locus in Col _FRI_ 714 under NV, constant 2W, and FS conditions, and in _aprf1-9_ _FRI_ under NV and constant 2W 715 conditions. In **C** and **D**, statistical significance was assessed using two-side Students’s t-tests. _P_ 22 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.10.765395; this version posted November 12, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.10.765395) 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 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 716 values were adjusted using the Holm-Sidak approach (ns indicates no significance; * _p_ < 0.05; 717 ** _p_ < 0.01; *** _p_ < 0.001). 718 **E** . Promoter-contact observed/expected (OE) values for regions with and without detected 719 _COOLAIR_ polyadenylation in NV and constant cold conditions. Points represent individual 200720 bp bins; boxes show the median and interquartile range, with whiskers extending to 1.5 times the 721 interquartile range. The grey line indicates OE = 1. AUC gives the probability that a randomly 722 chosen poly(A)-positive region has higher OE than a randomly chosen poly(A)-negative region; 723 _P_ values were obtained by randomly reassigning the observed number of poly(A)-positive labels 724 among the 200-bp bins and recalculating the AUC. 725 726 23 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.10.765395; this version posted November 12, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.10.765395) 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 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 727 746 mediated _FLC_ repression responds to temperature changes over multiple timescales. Different 747 winter temperature regimes affect not only transcriptional initiation but also RNA Pol II 748 progression and RNA processing during _COOLAIR_ transcription. Early during cold exposure, 749 increased RNA Pol II stalling within the _COOLAIR_ gene body limits polymerase progression 24 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.10.765395; this version posted November 12, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.10.765395) 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 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 750 towards its 3’ end, near the _FLC_ transcription start site. With prolonged cold exposure, 751 proximal _COOLAIR_ transcripts progressively accumulate. Fluctuations below freezing (FS) 752 further enhance RNA Pol II stalling and promote premature termination within the gene body, 753 generating both polyadenylated and non-polyadenylated _COOLAIR_ 3’ ends. Regions of 754 preferential _COOLAIR_ 3’ processing are enriched for contacts with the _FLC_ promoter, and 755 genetic perturbations of processing alter the distribution of these contacts. Under FS, enhanced 756 promoter-gene-body contacts coincide with reduced H3K4me1 and _FLC_ transcription. 757 758 25 [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.10.765395; this version posted November 12, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.10.765395) 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 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 1 **Methods** 2 **Plant material and growth conditions** 3 All mutant and transgenic lines were in the _FRI_ _[sf2]_ background. _ntl8-D3 FRI_ _[1]_ _, aprf1-9_ _FRI_ [2], 4 _smu1-6_ _FRI_ [3] ( _smu1-7 FRI_ ) and _FLC_ _ΔCOOLAIR_ were described previously [1-4] . Seeds were surface5 sterilized and sown on Murashige and Skoog (MS) agar plates without glucose or sucrose 6 (unless otherwise stated) and kept at 4 °C in the dark for 3 days. Non-vernalized (NV) seedlings 7 were cultivated under warm conditions (16 h light/8 h dark at a constant 20 °C) for 10 days, 8 unless otherwise specified. For vernalization treatment, seedlings were initially grown under 9 the same warm conditions for 10 days and subsequently moved to cold conditions (8 h light/16 10 h dark at a constant 5 °C) for extended periods, such as 2 or 4 weeks. For cold-to-warm transfer 11 experiments, after 4 weeks of cold treatment, seedlings were moved to warm conditions (16 h 12 light, 8 h darkness with constant 20 °C) for 10 days. Three biological replicate samples were 13 taken at 9:00 after treatment. 14 15 Plants for fluctuating freezing experiments were generally grown in growth conditions as 16 described previously [1] . Seeds were sown onto soil and kept at 4 °C in the dark for 3 days. NV 17 plants were cultivated under warm conditions (16 h light/8 h dark at a constant 20 °C) for 10 18 days before plants were then moved to 8h light/16h dark growth chambers. For vernalization, 19 plants were treated with two different temperature regimes: constant 5°C and daily fluctuating 20 5°C (-1°C to 12°C). Three biological replicate samples were taken at 9:00 after 2 weeks 21 exposure to vernalization. 22 23 24 **RNA expression analysis** 25 Total RNA was isolated using the previously described hot phenol protocol [5] . Genomic DNA 26 was removed by treatment with TURBO DNase (Invitrogen) according to the manufacturer’s 27 instructions. First-strand cDNA was generated using SuperScript IV reverse transcriptase 28 (Invitrogen). Gene-specific primers were applied for reverse transcription of _COOLAIR_ and 29 _FLC_ . Quantitative PCR (qPCR) was conducted with SYBR Green I Master (Roche) on a 30 LightCycler 480 system (Roche). Ct values were normalized to the geometric mean of 31 _UBIQUITIN CARRIER PROTEIN 1_ ( _UBC_ ) and _SERINE/THREONINE PROTEIN_ 32 _PHOSPHATASE 2A_ ( _PP2A_ ). Primer sequences are provided in Supplementary Table 1. 33 34 35 **tFLEP-seq library preparation and sequencing** 36 Chromatin bound RNA extraction was performed following the protocol described [6] with minor 37 modifications. 2.5 g 10-day-old seedlings were harvested and grounded into a fine powder. 38 Suspend powder in 25 mL of Honda buffer (1.25% (wt/vol) Ficoll, 2.5% (wt/vol) Dextran T40, 39 0.44 M sucrose, 20 mM HEPES (pH 7.4), 10 mM MgCl 2, 0.5% (vol/vol) Triton X-100, 1 mM 40 DTT, 1 × protease inhibitor cocktail (Roche), and 100 ng/μL tRNA), filtered through two layers 41 of Miracloth, and centrifuged at 4 °C, 2000 × _g_ for 10 min. The nuclei pellets were resuspended 42 and washed in Honda buffer three times, or until the pellet appears colorless. The supernatant 43 was removed thoroughly and discarded. The nuclei pellets were weighed, and one volume of [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.10.765395; this version posted November 12, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.10.765395) 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 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 44 nuclei resuspension buffer (25 mM Tris-HCl (pH 7.5), 50% (vol/vol) glycerol, 0.5 mM EDTA 45 (pH 8.0), 1 mM DTT, 100 mM NaCl, 0.4 U/μL RNaseOUT and 100 ng/μL tRNA) was added 46 to each pellet according to the weight of the pellet (e.g., 100 mg = 100 μL). Two volumes of 47 washing buffer (25 mM Tris-HCl (pH 7.5), 300 mM NaCl, 1 M urea, 0.5 mM EDTA (pH 8.0), 48 1 mM DTT, 1% (vol/vol) Tween 20, 0.4 U/μL RNaseOUT and 200 ng/μL tRNA) was added, 49 and the pellets were washed by pipetting up and down seven to ten times. The samples were 50 centrifuged at 4 °C, 8,000 × _g_, for 1 min, the supernatant was removed and the pellets were 51 resuspended with one volume of nuclei resuspension buffer. One volume of washing buffer 52 was added, and the pellets were washed by pipetting up and down seven to ten times. The 53 samples were centrifuged at 4 °C, 8,000 × _g_, for 1 min, the supernatant was removed and add 54 1 ml TRIzol to extract RNA. 55 56 tFLEP-seq was performed following the protocol described [6] with modifications. rRNA 57 depletion was performed using riboPOOL probe (siTOOLs Biotech). The RNA was cleaned 58 and concentrated using the ZYMO RNA Clean & Concentrator-5 kit. 1 μL of 50 pmol 3’ 59 adapter (Universal miRNA cloning linker, NEB, S1315S) was added to the rRNA-depleted 60 RNA, followed by ligation using T4 RNA Ligase 2, truncated K227Q (NEB, M0373L). The 61 RNA was cleaned and concentrated. Reverse transcription was carried out using a 3’ cDNA 62 RT primer with the following sequence: 5’-phos/ACTTGCCTGTCGCTCTATCTTC-index63 ATTGATGGTGCCTACAG-3’. After first-strand cDNA synthesis, PCR amplification was 64 performed using the universal primer pair (Supplementary Table 2). Target enrichment was 65 conducted according to the manufacturer’s instructions using the myBaits Custom DNA 66 Targeted Sequencing Kits. The sequence of all baits is listed in Supplementary Table 3. 67 Enriched libraries were subsequently prepared for Nanopore sequencing. Library construction 68 was performed using the Ligation Sequencing Kit V14 for non-barcoded samples, or in 69 combination with the Native Barcoding Kit 24 V14 when barcoding was required. Sequencing 70 was carried out on R10 flow cell (Oxford Nanopore Technologies, FLO-MIN114) on a Minion 71 Mk1B or PromethION 2 Solo sequencer (Oxford Nanopore Technologies). Biological 72 replicates and correlation analysis of replicates for all samples are included in the 73 supplementary files. 74 75 76 **DRIPc-qPCR** 77 DRIPc-qPCR was performed following the protocol described [7] with modifications. 2.5 g 1078 day-old seedlings were harvested and grounded into a fine powder. The powder was suspended 79 in 25 mL of Honda buffer, filtered through two layers of Miracloth, and centrifuged at 3500 × 80 _g_ for 15 min. According to the weight of the nuclei pellets, 5 volumes of nuclear lysis buffer 81 (50 mM Tris-HCl (pH 7.5), 10 mM EDTA) were added into each sample. 3.5 µL of Proteinase 82 K (20 mg/ml, AM2546) were then added per 100µL of sample from step 1, add SDS to a final 83 concentration of 0.5% from 10% (w/v) SDS solution. The samples were incubated overnight 84 at 55°C with gentle rotation. Equal volume of phenol/chloroform was added to each sample, 85 mixed gently and centrifuged at 13,000 × g for 10 min. The supernatant was transferred into a 86 new tube, and the samples were precipitated with 1 volume of Isopropanol and 1/10 volume of 87 NaOAc. The tubes were gently inverted until DNA began to precipitate. Centrifugation was [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.10.765395; this version posted November 12, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.10.765395) 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 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 88 not used. DNA was spooled using pipette tips. The pellets were washed once with 70% EtOH 89 and dried. The DNA pellets were then dissolved in 200 µL H 2 O and mixed gently until 90 absolutely dissolved. 4 μg DNA were then used for restriction enzyme (HindIII, NdeI, XbaI) 91 digestion overnight, followed by phenol/chloroform extraction and DNA precipitation with 92 isopropanol. After restriction enzyme digestion and purification, aliquots of the products were 93 treated with RNase III (NEB, M0245S) or RNase H (NEB, M0297S), followed by phenol94 chloroform extraction and precipitation. 50 μL DNA (~2 μg DNA) was diluted in ChIP dilution 95 buffer (16.7 mM Tris-HCl (pH 7.5), 1.2 mM EDTA, 0.1% Triton X-100, 167 mM NaCl) to 495 96 μL, supplemented with protease inhibitors, and incubated with 5 μg S9.6 antibody (Kerafast, 97 ENH001). A 5 μL aliquot was collected as input control and stored at 4 °C. The remaining 98 sample was incubated overnight at 4 °C on a rotating platform. 50 μL Protein G agarose beads 99 were then added and the samples were incubated for a further 2-3 h at 4 °C with rotation. Beads 100 were collected by centrifugation (1000 × g, 1 min, 4 °C), washed five times with ChIP dilution 101 buffer (5 min each), and then washed twice with TE buffer (10 mM Tris-HCl (pH 7.5), 1 mM 102 EDTA, 0.1% Triton X-100) at room temperature. DNA-RNA hybrids were eluted by adding 103 200 μL Elution buffer (10 mM Tris-HCl (pH 7.5), 2 mM EDTA, 0.2% SDS, 100 ng/mL tRNA) 104 and 5 μL Proteinase K to each sample, including input controls, followed by incubation at 55 105 °C for 2 h. Beads were pelleted, and the supernatant was transferred to a new tube and mixed 106 with 300 μL H 2 O, followed by phenol/chloroform extraction and DNA precipitation with 107 isopropanol, pelleted DNA was resuspended in 40 μL H 2 O. 108 109 DNA/RNA hybrids (excluding input samples) were denatured at 95 °C for 2 min and 110 immediately left on ice for at least 2 min. DNA was then removed using the TURBO DNA111 free kit according to the manufacturer’s instructions. RNA was precipitated by adding 1 μL 112 glycoblue, 1/10 NaOAc, and 1 volume isopropanol, followed by overnight incubation at -20 113 °C. Samples were centrifuged at 12,000 × g for 15 min at 4 °C, and the RNA was washed once 114 with 70% ethanol and centrifuged again at 12,000 × g for 5 min at 4 °C. After air-drying, RNA 115 was resuspended in 40 μL H 2 O. A total of 10 μL RNA was used for reverse transcription 116 reactions, which were performed separately using pooled forward primers or pooled reverse 117 primers (Supplementary Table 4). COPIA and XCRR reverse primers were included as 118 endogenous controls. cDNA was diluted by adding 400 μL H 2 O. For input samples, 10 μL was 119 diluted with 410 μL H 2 O. The cDNA and input were diluted and analyzed by qPCR via 120 LightCycler480 II (Roche). 121 122 123 **ChIP-qPCR** 124 Chromatin immunoprecipitation (ChIP)-qPCR was performed as described previously [8,9] . ChIP 125 was performed with antibodies: α-H3 (Abcam, ab1791), α-H3K4me1 (Abcam, ab8895), α126 H3K4me2 (Millipore, 07-030), α-H3K4me3 (Millipore, 04-745), α-H3K9K14ac (Merck, #06127 599), α-H3K27ac (Abcam, ab4729), α-H3K36ac (Thermo Fisher, #39379), α-H3K36me3 128 (Abcam, ab9050). ChIP enrichment was quantified by qPCR with primers listed in 129 Supplementary Table 5. For histone, ChIP values were normalized to H3. For H3K36me3 and 130 H3K27me3, values were further normalized to enrichment at the positive control loci ACTIN 131 for H3K36me3 and STM for H3K27me3. [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.10.765395; this version posted November 12, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.10.765395) 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 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 132 133 134 **Micro Capture-C** 135 Micro-Capture-C was performed as previously described by Hamley et al., 2023 [10] with 136 modifications. 2.5 g of 10-day-old seedlings, as well as seedlings subjected to an additional 2 137 weeks vernalization treatment under constant 5 °C or daily FS conditions, were harvested and 138 crosslinked for 10 min in 1% (w/v) formaldehyde in PBS. Crosslinking was quenched with 139 glycine (0.125 M final concentration) for 5 min. Snap freeze with liquid nitrogen. Nuclei were 140 extracted from 2 g of ground frozen material with Honda buffe and washed with Honda buffer 141 three times. Subsequent steps were performed according to the protocol described by Hamley 142 et al., 2023. Libraries were pooled and subjected to target enrichment using the myBaits 143 Custom Kit under high-sensitivity hybridisation conditions. The sequence of all baits is listed 144 in Supplementary Table 6). 145 146 147 **Single Molecule RNA FISH (smFISH)** 148 The protocol described by Duncan et al., 2017 [11] was followed with minor modifications for 149 root squash imaging. Briefly, root tips were fixed for 30 min in 4% paraformaldehyde (PFA), 150 prepared from 16% EM-grade PFA (Electron Microscopy Sciences, 15710), and washed twice 151 in 1× PBS. Roots were squashed onto slides under a coverslip and permeabilized in 70% 152 ethanol (EtOH) for 4 h. Next, sequential 30 min incubations were carried out with 100% 153 methanol (MeOH), then 100% EtOH followed by a 5 min incubation in wash buffer (2× SSC, 154 Invitrogen, AM9763, containing 10% formamide (Sigma, F9037)). Slides were incubated 155 overnight at 37 °C in the dark with Stellaris probes [12] (Supplementary Table 7) conjugated with 156 Quasar 670 dye (LGC Biosearch Technologies) diluted to a final concentration of 150 nM in 157 hybridization solution (2× SSC, 10% dextran sulfate (Thermo Scientific, AC433240050), 10% 158 formamide). The following day, samples were rinsed twice and incubated for 30 min in wash 159 buffer. A second wash buffer containing 100 ng/mL DAPI (Thermo Scientific, 62248) was 160 applied and samples were incubated for 30 mins at 37 °C. Slides were then washed in 2× SSC 161 for 5 min and mounted in GLOX buffer (0.4% glucose, 10 mM Tris-HCl pH 8.0, 2× SSC) 162 supplemented with glucose oxidase and catalase (Sigma-Aldrich; 1 µL each per 100 µL buffer). 163 164 Whole mount smFISH was performed as described by Zhao et al. 2023 [13] with minor 165 modifications. Root tips were fixed as described above, then two brief 1 x PBS washes were 166 carried out, followed by tissue permeabilization in 75% EtOH (~16 h), 100% MeOH (2× 15 167 min), and 100% EtOH (2× 15 min), then a 1× PBS wash. The roots were cleared in ClearSee 168 (10% (w/v) xylitol, 15% (w/v) sodium deoxycholate, 25% (w/v) urea) [14] at 4 °C for 7 days then 169 embedded in an acrylamide-based hydrogel (30% acrylamide-bisacrylamide (29:1), 10% 170 ammonium persulfate, 5% TEMED). The roots were mounted on poly-L-lysine-coated slides 171 with 20 µL embedding solution and allowed to polymerize for 20-30 min prior to coverslip 172 removal. Hybridization and washes were performed as described above. Cell walls were 173 stained with 0.1% (v/v) SCRI Renaissance 2200, followed by washing and mounting in 20 µL 174 Vectashield (BioNordika, H-1000-10). [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.10.765395; this version posted November 12, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.10.765395) 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 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 175 176 177 **Image Acquisition** 178 Squashed-root imaging was performed using a Zeiss Elyra PS1 wide-field epifluorescence 179 microscope equipped with a 100× oil-immersion objective (NA 1.46) and a cooled EMCCD 180 camera (Andor iXon 897; 512 × 512 pixels, QE > 90%) operated in standard mode. Z-stacks 181 spanning the entire nucleus were acquired with a step size of 0.2 µm. DAPI was excited at 405 182 nm and detected at 420-480 nm, while Quasar 670-labelled probes were excited at 642 nm and 183 detected at 655-710 nm. 184 185 Whole-mount FISH imaging was performed using a Leica Stellaris 8 FALCON confocal 186 microscope equipped with an HC PL APO CS2 63× water-immersion objective (NA 1.20) in 187 photon-counting mode. Optical sections were acquired with a z-step of 0.3 µm at a scan speed 188 of 400 Hz, with the pinhole set to 2 Airy units. 189 190 Quasar 670-labelled probes were excited at 645 nm using a white light laser (80 MHz, 85% 191 laser power, 5% intensity), with emission collected at 660-730 nm (HyD X detector; line 192 accumulation 6; frame averaging 2). SCRI Renaissance 2200 was excited at 405 nm, and 193 emission was detected at 410-503 nm. 194 195 196 **Image Analysis** 197 _COOLAIR_ foci quantification was carried out using FIJI [15] . Volume and intensity were 198 determined from root squash smFISH images with foci first saved as individual regions of 199 Interest (ROIs). To determine volume, these ROIs were processed with a 3D Gaussian blur (σ 200 = 0.5 µm in x, y, z) and background-subtracted (rolling ball radius = 20). Thresholding (2124201 65535) was applied to generate masks, and volumes were quantified using the 3D Objects 202 Counter plugin (minimum voxel count = 15, maximum = 19,488). To determine foci intensity, 203 ROIs were blurred (σ = 0.5 µm), background-subtracted (rolling = 20), and summed along the 204 z-axis. Thresholding was applied to select foci, and intensity values were measured using ROI 205 Manager. Proportions of _COOLAIR_ positive and negative cells were determined from z slices 206 acquired around the mid-sections and epidermal regions of whole mount smFISH samples. 207 Foci per nucleus data were determined from manually selected nucleus ROIs from maximum 208 z-projections of these images using the Find Maxima command (prominence = 5), following 209 smoothing (Gaussian blur, σ = 0.5 µm), and background-subtraction (rolling = 20). 210 211 212 **MNase-seq** 213 MNase-seq was performed following the protocol described in Montez et al., 2025 [16] . _ntl8-D3_ 214 seedlings were cultivated under warm conditions. Three biological replicates were used. For 215 each replicate, 1.5 g of tissue was frozen in liquid nitrogen and ground into a powder. The 216 material was resuspended in 35 mL of Nuclei Isolation Buffer (20 mM HEPES-KOH pH 7.4, 217 0.44 M sucrose, 1.25% Ficoll, 2.5% Dextran T40, 20 mM KCl, 0.5 mM spermidine, 5 mM [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.10.765395; this version posted November 12, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.10.765395) 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 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 218 DTT, 1% Triton X-100, protease inhibitors, and 0.1 mM PMSF). After filtration through two 219 layers of Miracloth, the homogenate was centrifuged at 2000 × g for 15 min at 4 °C. 220 221 The nuclei pellet was washed once in 5 mL of isolation buffer and further purified using a 222 Percoll gradient composed of 75% and 40% layers. Following centrifugation at 7000 × g for 223 30 min at 4 °C, nuclei were collected from the interface between the two layers and washed 224 again in 6 mL of isolation buffer. Purified nuclei were resuspended in 1.5 mL MNase buffer 225 (20 mM Tris pH 8.0, 0.125 M sucrose, 5 mM NaCl, 40 mM KCl, 2 mM CaCl 2, protease 226 inhibitors, and 0.1 mM PMSF), transferred to a clean tube, and pelleted at 1500 × g for 10 min 227 at 4 °C. After removing the supernatant, the pellet mass was determined. A total of 65 mg 228 nuclei in 300 μL MNase buffer was digested with 10 U MNase for 30 min at 37 °C. Reactions 229 were terminated by adding EDTA, NaCl, SDS, and Proteinase K, followed by incubation at 230 65 °C for 1 h. DNA was purified by two phenol-chloroform extractions, one chloroform 231 extraction, and ethanol precipitation with glycoblue at -80 °C. The DNA pellet was washed 232 with 75% ethanol and resuspended in 30 μL water. RNA contamination was removed by RNase 233 A treatment at 37 °C for 20 min. Mono- and subnucleosomal DNA fragments (< 200 bp) were 234 enriched using AMPure XP beads while excluding fragments >300 bp. Sequencing libraries 235 were prepared using the NEBNext Ultra II DNA Library Prep Kit according to the 236 manufacturer’s protocol. After adaptor ligation, fragments > 200 bp were retained with 237 AMPure XP beads, and libraries were amplified by 8 cycles of Illumina indexing PCR. 238 239 Libraries were pooled and subjected to target enrichment using the myBaits Custom Kit under 240 high-sensitivity hybridisation conditions. The sequence of all baits is listed in Supplementary 241 Table 6. Enriched libraries were sequenced on an DNBseq G400 platform (BGI Tech) with 242 paired-end 150 bp reads. Data in Fig. S6 from FRI warm and cold has been re-plotted from 243 published data in Montez et al., 2025 [16] . 244 245 246 **Quant-seq** 247 For Quant-seq experiments, total RNA was isolated as described for qPCR and further purified 248 using the Qiagen RNeasy Miniprep Kit (74106). Sequencing libraries were prepared from 100 249 ng total RNA using the Quant-seq 3’ mRNA-Seq Library Prep Kit REV for Illumina 250 (015UG009V0271) according to the manufacturer’s instructions. Library quality was assessed 251 on a Fragment Analyzer using the HS-DNA assay, and concentrations were measured with a 252 Qubit dsDNA HS assay (Thermo Fisher). Indexed libraries were pooled and sequenced on an 253 Illumina NextSeq 2000 using a 100-cycle cartridge and the Lexogen Custom Sequencing 254 Primer (CSP). Library preparation and sequencing were performed by Lexogen GmbH. 255 256 257 **Data processing and visualization** 258 **NET-seq bioinformatics analysis** 259 The NET-seq data were obtained from previously published datasets [17] . To identify significant 260 pausing sites, a sliding window approach was applied to scan signal enrichment across the [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.10.765395; this version posted November 12, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.10.765395) 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 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 261 COOLAIR gene region (step size = 50 bp; window size = 250 bp). For each window, the total 262 signal within the window was defined as the observed value, while signals from the 263 immediately upstream and downstream regions of equal window size were used as background 264 controls. Statistical significance was assessed using a Poisson test and resulting _P_ -values were 265 adjusted for multiple testing using the Benjamini-Hochberg (BH) method. 266 267 268 **tFLEP bioinformatics analysis** 269 Processing of tFLEP-seq reads was performed using the pipeline available at 270 [https://github.com/ZhaiLab-SUSTech/FLEPSeq. Based on the presence or absence of poly(A)](https://github.com/ZhaiLab-SUSTech/FLEPSeq) 271 tails, reads were classified into non-poly(A) reads (representing elongating RNA) and poly(A) 272 reads (representing polyadenylated RNA). To estimate Pol II occupancy, the number of reads 273 corresponding to 3’ end positions was quantified and normalized against two housekeeping 274 genes ( _ACT7_ and _UBC21_ ). For visualization, signal profiles were generated using a sliding 275 window approach (step size = 20 bp; window size = 100 bp). To compare Pol II dynamics 276 between conditions, three regions associated with major poly(A) clusters induced under cold 277 treatment were defined. The proximal region was defined as ±300 bp surrounding the major 278 poly(A) site (3,173,953) of the proximal isoform (Class I), while the distal region was defined 279 similarly around the major poly(A) site (3,179,581) of the distal isoform (Class II). To avoid 280 interference from proximal signals, a non-overlapping region (3,174,500-3,179,000) was 281 defined as the gene body region. Detailed library quality metrics, including mapped rate, 282 enrichment rate, proportion of poly(A), and proportion of non-poly(A) reads, are provided in 283 Supplementary Table 8. 284 285 286 **Micro Capture-C sequencing bioinformatics analysis** 287 Low-quality reads and adapter sequences were removed using fastp [18] . The resulting clean reads 288 were aligned to the reference genome using BWA-MEM (https://github.com/lh3/bwa). PCR 289 duplicates and reads with low mapping quality (MAPQ < 40) were filtered using pairtools [19] . 290 Reproducibility and correlation between biological replicates were assessed using the 291 _hicCorrelate_ function in HiCExplorer [20] . Contact matrices were normalized by the total number 292 of valid read pairs mapped within the selected _FLC_ region and stored in _mcool_ format using 293 cooler [21] . The insulation score and boundaries within the FLC region were calculated at 1 kb 294 resolution using cooltools [22] . The observed/expected (O/E) matrix was calculated to correct for 295 the distance-dependent decay of interaction frequency. Visualization of chromatin interaction 296 patterns was performed using CoolBox [23] . 297 298 For the poly(A)-promoter contact analysis, the _FLC_ locus was divided into 200-bp bins 299 spanning the _FLC_ gene region from 3,172,000 to 3,179,400, including the _COOLAIR_ TSS 300 region. Bins were classified as poly(A)-positive when the summed tFLEP-seq _COOLAIR_ 301 poly(A) read count across the three replicates exceeded 2 reads and as poly(A)-negative 302 otherwise; promoter-contact OE values were then compared between the two classes using the 303 AUC, calculated from the Mann-Whitney U statistic, with significance assessed by randomly [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.10.765395; this version posted November 12, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.10.765395) 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 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 304 reassigning the observed number of poly(A)-positive labels among bins and recalculating the 305 AUC. 306 307 308 **Quant-seq analysis** 309 Raw reads were quality-trimmed, and adapter sequences were removed using fastp [18] . Clean 310 reads were aligned to the _Arabidopsis thaliana_ reference genome (TAIR10) using STAR 311 (https://github.com/alexdobin/STAR). The 3’ end sites were extracted from aligned reads by 312 taking the terminal genomic position of each read in a strand-specific manner (the 3’ end of 313 reads on the forward strand and the 5’ end of reads on the reverse strand). To examine antisense 314 poly(A) pattern changes induced by FS, we analysed a set of genes with antisense transcription 315 initiating close to their poly(A) sites, identified by plaNET-seq in a previous study [24] . For 316 identification of poly(A) site clusters, sites on the same strand were sorted by genomic 317 coordinate and a sliding-window algorithm was applied to merge adjacent sites within 24 bp 318 into clusters. The highest-read-count site within the cluster was set as cluster summit. The 319 clusters with a total read count below 5 were discarded. 320 321 322 **Data availability** 323 Raw sequencing data generated in this study is available under the SRA Bioproject accession 324 number PRJNA1558601. The NET-seq data used in this study is from a previous study [17], 325 available under the SRA Bioproject accession number PRJNA1144665. Images generated for 326 the smFISH analysis are available at the BioImage Archive under accession number S 327 BIAD3599. 328 329 330 **Code availability** 331 Code used for the analysis and visualization in this study is available via GitHub 332 (https://github.com/DingLab-RNAstructure/COOLAIR_co_transcription_analysis). 333 334 335 **Methods References** 336 337 1 Zhao, Y. _et al._ Natural temperature fluctuations promote _COOLAIR_ regulation of 338 _FLC_ . _Genes Dev_ **35** [, 888-898 (2021). https://doi.org/10.1101/gad.348362.121](https://doi.org/10.1101/gad.348362.121) 339 2 Mateo-Bonmatí, E. _et al._ A CPF-like phosphatase module links transcription 340 termination to chromatin silencing. _Molecular Cell_ **84**, 2272-2286.e2277 (2024). 341 [https://doi.org/10.1016/j.molcel.2024.05.016](https://doi.org/10.1016/j.molcel.2024.05.016) 342 3 Long, X. _et al._ Cotranscriptional splicing is required in the cold to produce COOLAIR 343 isoforms that repress Arabidopsis FLC. _Proc Natl Acad Sci U S A_ **121**, e2407628121 344 (2024). [https://doi.org/10.1073/pnas.2407628121](https://doi.org/10.1073/pnas.2407628121) 345 4 Luo, X., Chen, T., Zeng, X., He, D. & He, Y. Feedback Regulation of FLC by 346 FLOWERING LOCUS T (FT) and FD through a 5′ FLC Promoter Region in 347 Arabidopsis. _Molecular Plant_ **12**, 285-288 (2019). 348 [https://doi.org/https://doi.org/10.1016/j.molp.2019.01.013](https://doi.org/https://doi.org/10.1016/j.molp.2019.01.013) [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.10.765395; this version posted November 12, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.10.765395) 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 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 349 5 Box, M. S., Coustham, V., Dean, C. & Mylne, J. S. Protocol: A simple phenol-based 350 method for 96-well extraction of high quality RNA from Arabidopsis. _Plant Methods_ 351 **7** [, 7 (2011). https://doi.org/10.1186/1746-4811-7-7](https://doi.org/10.1186/1746-4811-7-7) 352 6 Long, Y., Jia, J., Mo, W., Jin, X. & Zhai, J. FLEP-seq: simultaneous detection of RNA 353 polymerase II position, splicing status, polyadenylation site and poly(A) tail length at 354 genome-wide scale by single-molecule nascent RNA sequencing. _Nature Protocols_ 355 **16**, 4355-4381 (2021). [https://doi.org/10.1038/s41596-021-00581-7](https://doi.org/10.1038/s41596-021-00581-7) 356 7 Sanz, L. A. & Chédin, F. High-resolution, strand-specific R-loop mapping via S9.6357 based DNA–RNA immunoprecipitation and high-throughput sequencing. _Nature_ 358 _Protocols_ **14** [, 1734-1755 (2019). https://doi.org/10.1038/s41596-019-0159-1](https://doi.org/10.1038/s41596-019-0159-1) 359 8 Franco-Echevarría, E. _et al._ Distinct accessory roles of Arabidopsis VEL proteins in 360 Polycomb silencing. _Genes & Development_ **37**, 801-817 (2023). 361 [https://doi.org/10.1101/gad.350814.123](https://doi.org/10.1101/gad.350814.123) 362 9 Schulten, A. _et al._ VEL-dependent polymerization maintains the chromatin 363 association of Polycomb proteins for the switch to epigenetic silencing. _Molecular_ 364 _Cell_ **85** [, 3321-3332.e3325 (2025). https://doi.org/10.1016/j.molcel.2025.08.002](https://doi.org/10.1016/j.molcel.2025.08.002) 365 10 Hamley, J. C., Li, H., Denny, N., Downes, D. & Davies, J. O. J. Determining 366 chromatin architecture with Micro Capture-C. _Nature Protocols_ **18**, 1687-1711 367 (2023). [https://doi.org/10.1038/s41596-023-00817-8](https://doi.org/10.1038/s41596-023-00817-8) 368 11 Duncan, S., Olsson, T. S. G., Hartley, M., Dean, C. & Rosa, S. Single Molecule RNA 369 FISH in Arabidopsis Root Cells. _Bio-protocol_ **7**, e2240 (2017). 370 [https://doi.org/10.21769/BioProtoc.2240](https://doi.org/10.21769/BioProtoc.2240) 371 12 Rosa, S., Duncan, S. & Dean, C. Mutually exclusive sense–antisense transcription at 372 _FLC_ facilitates environmentally induced gene repression. _Nature Communications_ **7**, 373 [13031 (2016). https://doi.org/10.1038/ncomms13031](https://doi.org/10.1038/ncomms13031) 374 13 Zhao, L., Fonseca, A., Meschichi, A., Sicard, A. & Rosa, S. Whole-mount smFISH 375 allows combining RNA and protein quantification at cellular and subcellular 376 resolution. _Nature Plants_ **9**, 1094-1102 (2023). [https://doi.org/10.1038/s41477-023-](https://doi.org/10.1038/s41477-023-01442-9) 377 [01442-9](https://doi.org/10.1038/s41477-023-01442-9) 378 14 Kurihara, D., Mizuta, Y., Sato, Y. & Higashiyama, T. ClearSee: a rapid optical 379 clearing reagent for whole-plant fluorescence imaging. _Development_ **142**, 4168-4179 380 (2015). [https://doi.org/10.1242/dev.127613](https://doi.org/10.1242/dev.127613) 381 15 Schindelin, J. _et al._ Fiji: an open-source platform for biological-image analysis. 382 _Nature Methods_ **9** [, 676-682 (2012). https://doi.org/10.1038/Nmeth.2019](https://doi.org/10.1038/Nmeth.2019) 383 16 Montez, M. _et al._ Cold-induced nucleosome dynamics linked to silencing of 384 Arabidopsis FLC. _Nature Communications_ **16**, 5550 (2025). 385 [https://doi.org/10.1038/s41467-025-60735-z](https://doi.org/10.1038/s41467-025-60735-z) 386 17 Maple, R. _et al._ Co-transcriptional splicing changes combine with reduced productive 387 transcription initiation for cold-induced repression of FLC. _Nucleic Acids Research_ **54** 388 (2026). [https://doi.org/10.1093/nar/gkaf1462](https://doi.org/10.1093/nar/gkaf1462) 389 18 Chen, S., Zhou, Y., Chen, Y. & Gu, J. fastp: an ultra-fast all-in-one FASTQ 390 preprocessor. _Bioinformatics_ **34**, i884-i890 (2018). 391 [https://doi.org/10.1093/bioinformatics/bty560](https://doi.org/10.1093/bioinformatics/bty560) 392 19 Open2C _et al._ Pairtools: From sequencing data to chromosome contacts. _PLOS_ 393 _Computational Biology_ **20**, e1012164 (2024). 394 [https://doi.org/10.1371/journal.pcbi.1012164](https://doi.org/10.1371/journal.pcbi.1012164) 395 20 Ramírez, F. _et al._ High-resolution TADs reveal DNA sequences underlying genome 396 organization in flies. _Nature Communications_ **9**, 189 (2018). 397 [https://doi.org/10.1038/s41467-017-02525-w](https://doi.org/10.1038/s41467-017-02525-w) [bioRxiv preprint doi: https://doi.org/10.64898/2026.11.10.765395; this version posted November 12, 2026. The copyright holder for this](https://doi.org/10.64898/2026.11.10.765395) 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 4.0 International license.](http://creativecommons.org/licenses/by/4.0/) 398 21 Abdennur, N. & Mirny, L. A. Cooler: scalable storage for Hi-C data and other 399 genomically labeled arrays. _Bioinformatics_ **36**, 311-316 (2019). 400 [https://doi.org/10.1093/bioinformatics/btz540](https://doi.org/10.1093/bioinformatics/btz540) 401 22 Open2C _et al._ Cooltools: Enabling high-resolution Hi-C analysis in Python. _PLOS_ 402 _Computational Biology_ **20**, e1012067 (2024). 403 [https://doi.org/10.1371/journal.pcbi.1012067](https://doi.org/10.1371/journal.pcbi.1012067) 404 23 Xu, W. _et al._ CoolBox: a flexible toolkit for visual analysis of genomics data. _BMC_ 405 _Bioinformatics_ **22** [, 489 (2021). https://doi.org/10.1186/s12859-021-04408-w](https://doi.org/10.1186/s12859-021-04408-w) 406 24 Meena, S. K. _et al._ Antisense transcription from stress-responsive transcription factors 407 fine-tunes the cold response in Arabidopsis. _The Plant Cell_ **36**, 3467-3482 (2024). 408 [https://doi.org/10.1093/plcell/koae160](https://doi.org/10.1093/plcell/koae160) 409 410 411 **Acknowledgements** We thank Dr. Qianli Dong for help with Micro Capture-C and Shuqin 412 Chen for excellent technical assistance, and current and former members of the Dean and Ding 413 labs for helpful discussions. We thank Prof. Lori Passmore, Prof. Chang Liu, and Prof. Jiawei 414 Wang for comments on the manuscript. We also thank the JIC Bioimaging facility and staff for 415 their contribution to this publication. 416 417 418 **Funding statement** This work was funded by European Research Council Advanced Grants 419 (EPISWITCH, 833254) and European Research Council (ERC) [selected by the ERC, funded 420 by BBSRC Horizon Europe Guarantee [EP/Y009886/1, VENTURE], Wellcome Trust 421 (210654/Z/18/Z), and a Royal Society Professorship (RP\R1\ RSRP\R\231006) to C.D., the 422 Royal Society Faraday Discovery Fellowships [FDF\S2\251014] to Y.D., Human Frontier 423 Science Program Fellowship [LT0029/2024-L] to J.X. and BBSRC Institute Strategic 424 Programmes (BB/J004588/1 and BB/P013511/1). Maria Sindalovskaya was supported by the 425 John Innes Foundation PhD programme. Svenja Reeck was funded by the HORIZON 426 EUROPE Marie Sklodowska-Curie Actions PEP-NET (Predictive Epigenetics: Fusing Theory 427 and Experiment) grant 813282. Open access publication was funded by the John Innes Centre. 428 429 430 **Author contributions** C.D. M.M. J.X, J.Z. & Y.D. conceived the study. J.X, C.G, M.S, E. M431 B., M.Y, X.J, S.R., S.D., Y.Z., C.X., G-J.J. performed the experiments and analysed the data. 432 G.M. analysed the data. C.D. obtained funding and C.D and Y.D supervised the work. J.X, C.G, 433 G.M., M.M., C.D contributed to writing the manuscript and preparing the figures. 434 435 436 **Declaration of interests** The authors declare no competing interests. 437 438 439 **Additional information** 440 **Supplementary information** Supplementary files presenting individual replicate analysis of 441 sequencing data and primer and probe information are provided with the manuscript.