# Source paper archive - Source title: DECAPPING PROTEIN 5 regulates the circadian clock in Arabidopsis - Author(s): Daniel Alejandro Careno, Soledad Perez Santangelo, Julieta Lisa Lisa Mateos, Luciana Bianchimano, Horacio Pallares, Dominique Jacques-Vuarambon, Micaela Sol Shalom, Abril San Martin, Maria Jose de Leone, Victor Daniel Aldas Bulos, Anthony Treichel, Martin Lewinski, Rodrigo Siqueira Reis, Dorothee Staiger, Ariel Alejandro Bazzini, Marcelo Javier Yanovsky - Source landing URL: https://www.biorxiv.org/content/10.64898/2026.10.30.763074v1 - PDF URL: https://www.biorxiv.org/content/10.64898/2026.10.30.763074v1.full.pdf - PDF SHA-256: 6013c8cb8101874ba858446502d88524dba8287d53006734cdb16521e2cb5be6 - 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/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 1 **DECAPPING PROTEIN 5 regulates the circadian clock in Arabidopsis** 2 Daniel Alejandro Careno [1] *, Soledad Perez Santangelo [2], Julieta Lisa Mateos [3,4,5], Luciana 3 Bianchimano [1], Horacio Pallarés [6], Dominique Jacques-Vuarambon [7], Micaela Sol 4 Shalom [1], Abril San Martín [1], María José de Leone [1], Victor Daniel Aldas Bulos [6], Anthony 5 Treichel [6], Martin Lewinski [5], Rodrigo Siqueira Reis [7], Dorothee Staiger [5], Ariel Alejandro 6 Bazzini [6], Marcelo Javier Yanovsky [1] 7 8 1 Fundación Instituto Leloir, Instituto de Investigaciones Bioquímicas de Buenos Aires 9 (CONICET), Buenos Aires, Argentina 10 2 School of Biological Sciences, University of Auckland, Auckland, New Zealand 11 3 Universidad de Buenos Aires, Facultad de Ciencias Exactas y Naturales, 12 Departamento de Fisiología, Biología Molecular y Celular, Buenos Aires, Argentina 13 4 Instituto de Fisiología, Biología Molecular y Neurociencias (CONICET), Buenos Aires, 14 Argentina 15 5 RNA Biology and Molecular Physiology, Faculty of Biology, Bielefeld University, 16 Bielefeld, Germany 17 6 Stowers Institute for Medical Research, Kansas City, United States 18 7 Institute of Plant Sciences, University of Bern, Bern, Switzerland 19 20 *Corresponding authors e-mails: [dcareno@leloir.org.ar, myanovsky@leloir.org.ar](mailto:dcareno@leloir.org.ar) 1 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 21 **Abstract** 22 The Arabidopsis transcriptome must be continuously reshaped to integrate endogenous 23 and environmental cues, including the circadian clock and ambient light. This 24 reprogramming relies on the coordinated synthesis and degradation of mRNAs, yet the 25 contribution of RNA decay factors to circadian regulation remains poorly characterized. 26 Here, we show that nearly all core clock genes are high-amplitude clock-controlled genes 27 (HA-CCGs). A meta-analysis of transcriptome-wide mRNA decay datasets revealed that 28 HA-CCGs have consistently shorter half-lives across multiple studies. We further identify 29 the mRNA decapping activator DECAPPING PROTEIN 5 (DCP5) as a key regulator of 30 circadian RNA stability. Using iCLIP, we show that DCP5 binds preferentially to HA-CCG 31 transcripts at AU-rich motifs in their 3′ UTRs, thereby modulating their stability. 32 Accordingly, the _dcp5_ mutant displays an extended circadian period at both physiological 33 and molecular levels. Our findings establish mRNA decay as a post-transcriptional 34 regulatory layer that fine-tunes the decay of multiple key transcripts within the circadian 35 system. 2 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 36 **Introduction (622 words)** 37 The biological clock is an endogenous mechanism that provides temporal information 38 even in the absence of external signals and is found in organisms as diverse as bacteria, 39 plants, and animals [1] . The clock generates biological oscillations known as circadian 40 rhythms that regulate many aspects of an organism's behavior, physiology, and 41 development. The central oscillator —a group of a few dozen genes— is a gene network 42 based on multiple interconnected feedback loops, in which a set of genes acts as mutually 43 regulating transcriptional repressors and activators, as well as controlling the expression 44 of thousands of other clock-controlled genes (CCG) [2] . 45 In Arabidopsis, the central oscillator genes _LATE ELONGATED HYPOCOTYL_ ( _LHY_ ) and 46 _CIRCADIAN CLOCK ASSOCIATED 1_ ( _CCA1_ ) reach their peak of expression at dawn 47 and repress the expression of the afternoon and evening genes [3–5] . Later in the morning, 48 the _NIGHT LIGHT-INDUCIBLE AND CLOCK-REGULATED GENE_ ( _LNK_ ) family interacts 49 with members of the _REVEILLE_ ( _RVE_ ) gene family and promotes the expression of 50 _PSEUDO-RESPONSE REGULATOR 5_ ( _PRR5_ ), _PRR1/TIME OF CAB EXPRESSION_ 51 ( _TOC1_ ) and _EARLY FLOWERING 4_ ( _ELF4_ ) [6–8] . Throughout the day, members of the _PRR_ 52 family, starting with _PRR9_ and ending with _TOC1_, reach their respective expression 53 peaks and sequentially suppress the expression of _LHY_ and _CCA1_ . In turn, _PRRs_ repress 54 one another [9,10] . After dusk, ELF3, ELF4, and LUX ARRHYTHMO (LUX) interact to form 55 the Evening Complex (EC). The EC maintains the repression of _PRR9_, _PRR7_, _PRR5_, 56 _GIGANTEA_ ( _GI_ ), and _LUX_ itself [11–15] . The repression of _PRRs_ indirectly activates the 57 expression of _CCA1_ and _LHY_, which, toward the end of the night, repress the EC, and 58 the cycle begins again. Most of these central oscillator genes encode transcription factors 3 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 59 and co-activators, thus, the gene circuit can explain circadian oscillations at the 60 transcription level. However, emerging evidence suggests that mRNA circadian 61 oscillations are also controlled in their decay phase [16] . 62 Two main pathways of mRNA degradation shape the stability of transcripts. These 63 pathways begin with the conversion of a polyadenylated mRNA to an oligoadenylated 64 mRNA through the shortening of the poly(A) tail. Then, mRNAs can be substrates of the 65 3'-5' degradation pathway and be degraded by the multi-subunit complex called the 66 exosome, or alternatively, the 5' m [7] G cap can be removed by the decapping complex, 67 and decapped transcripts can be degraded by the cytoplasmic 5'-3' exoribonuclease 68 XRN4 [17], the plant homolog of yeast and metazoan XRN1. 69 There is emerging evidence that shows the importance of RNA decay regulation in the 70 function of the clock. It has been reported that alternative splicing coupled to nonsense 71 mediated decay controls the expression levels of _TOC1_ and _ELF3_ [18,19] . Also, N6 72 methyladenosine (m6A) modification has been shown to tune the RNA half-life of _CCA1_ [20] . 73 Regarding specific components of the RNA decay machinery, it has been shown that 74 XRN4, and the members of the decay promoting complex LSM1-7 _SM-LIKE 1_ ( _LSM1_ ), 75 _LSM4_ and _LSM5_ are required for proper clock function [21–24] . 76 LSM1, LSM4 and XRN4 localize in a type of membraneless cytoplasmic organelle called 77 processing bodies (P-bodies), which provide a higher decay rate than in the rest of the 78 cytoplasm [25] . DECAPPING PROTEIN 5 (DCP5) is a P-body component required for 79 proper mRNA decay [26] . Here, we describe how DCP5 is necessary for the functioning of 80 the clock, reinforcing the idea that circadian oscillations are strongly regulated at the post 81 transcriptional level. 4 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 82 **Results** 83 _Clock-controlled transcripts have short half-lives_ 84 Circadian transcripts undergo pronounced changes in abundance over the course of a 85 day, leading us to hypothesize that they are subject to precise temporal coordination of 86 their transcription and mRNA degradation rates. Transcripts that oscillate with high 87 circadian amplitude should be less stable than those with constant expression: once peak 88 abundance is reached, the transcript must decay rapidly enough to reach its trough within 89 ~12 h. If certain transcripts, particularly those of the central oscillator, do not decay quickly 90 enough, the functioning of the clock would be compromised. An extreme case can be 91 seen when these genes are constitutively expressed using strong promoters. In these 92 cases, a loss of rhythmicity is often observed [27–30] . 93 According to Romanowski _et al._ (2020) [31], the Arabidopsis transcriptome contains 9,127 94 clock-controlled genes (CCGs), including both core clock genes and downstream output 95 genes. Their relative amplitudes, defined as the amplitude normalized to the trough value, 96 range from 0.2 to 167.3. Notably, most core oscillator genes display relative amplitudes 97 greater than 4 (Fig. 1a). We therefore used a relative amplitude of 4 as the threshold to 98 classify CCGs as high-amplitude (HA-CCGs, 621 genes) or low-amplitude (LA-CCGs, 99 8,506 genes) (Fig. 1b). 100 To test our hypothesis, the half-lives of CCGs were compared based on their amplitude 101 using previously published transcriptome-wide datasets reporting half-lives for 102 Arabidopsis mRNAs [32–35] . In all four datasets, the median half-life of HA-CCG is lower than 103 that of both LA-CCG and non-CCG, and in three cases out of the four these differences 104 are statistically significant (Fig. 1c-f). Furthermore, when CCGs were grouped into 5 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 105 amplitude deciles, half-lives trended downward with increasing amplitude 106 (Supplementary Fig. 1). These observations align with microarray data showing 107 enrichment of CCGs among fast-decay transcripts [36] . The consistently shorter half-lives 108 of HA-CCG mRNAs across independent datasets suggest that post-transcriptional control 109 of RNA stability plays a key role in shaping circadian oscillations. 6 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 110 111 **Figure 1.** **High-amplitude clock-controlled genes (CCGs) display shorter mRNA** 112 **half-lives than non-oscillating genes.** **a** Relative amplitude of the 9,127 CCGs identified 113 by Romanowski _et al._ (2020) [31] ranked in ascending order. Core oscillator genes are 114 labeled and highlighted in green. The orange dashed line indicates the threshold at 115 relative amplitude = 4, used to classify CCGs as high-amplitude (HA-CCGs) or low 116 amplitude (LA-CCGs). **b** Schematic illustrating that clock-controlled genes (CCGs) 117 classification in HA-CCGs (relative amplitude > 4), LA-CCGs (4 ≥ relative amplitude > 0) 118 and non-CCGs. **c-f** Distributions of mRNA half-lives for HA-CCGs, LA-CCGs, and non 119 CCGs across four independent genome-wide decay studies [32–35] . Statistical significance 7 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 120 was assessed by Kruskal–Wallis test followed by pairwise Wilcoxon rank-sum tests with 121 Benjamini–Hochberg (FDR) correction; *P < 0.05, **P < 0.01, ***P < 0.001; NS, not 122 significant. 8 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 123 _DCP5 is required for normal clock function_ 124 Previously, two components of the 5’-3’ RNA decay pathway, LSM1 and XRN4, have 125 been described as necessary for correct clock function [22,23] . These RNA decay factors 126 may be part of a mechanism that maintains CCG mRNA half-lives within a range 127 compatible with proper circadian oscillations, that is, the appropriate period, phase, and 128 amplitude. The 5’-3’ mRNA decay pathway requires removal of the 5’ m [7] G cap by the 129 decapping complex. In Arabidopsis, the core decapping complex includes DCP1, DCP2 130 (the component with enzymatic activity) and VARICOSE (VCS), and its assembly is 131 modulated by DECAPPING PROTEIN 5 (DCP5) [37,38] . While _dcp1_, _dcp2_ and _vcs_ are all 132 post-embryonic lethal mutants [39], a hypomorphic _dcp5_ mutant allele, _dcp5-1_, can grow 133 and complete its developmental life cycle [26] . We therefore studied clock function in the 134 _dcp5-1_ mutant and found a long circadian period phenotype for the rhythm of leaf 135 movement in _dcp5_ compared with wild-type plants (24.83 ± 0.44 h vs 27.57 ± 0.72 h) (Fig. 136 2a-b). Supporting the idea that the circadian phenotype observed in this mutant is due to 137 the mutation in _DCP5_, this phenotype was completely rescued when we complemented 138 _dcp5-1_ plants with a WT version of the _DCP5_ gene (Fig. 2b). We then analyzed circadian 139 rhythms in gene expression by evaluating rhythms in bioluminescence driven by the 140 _pTOC1:Luc_ and _pCCR2:Luc_ reporter genes, and we also observed a strong circadian 141 phenotype in _dcp5-1_ mutant plants. Indeed, the rhythm of expression of both reporters 142 was lengthened by ~2 h in _dcp5-1_ mutants compared with WT plants (24.18 ± 0.57 h vs 143 26.28 ± 0.34 h for _pTOC1:Luc_, and 23.65± 0.48 h vs 25.59 ± 0.50 h for _pCCR2:Luc_ ) (Fig. 144 2c-d and Supplementary Fig. 2). Together, these results demonstrate that DCP5 is 9 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 145 required for proper clock function and reinforce the view that mRNA decay constitutes a 146 key regulatory layer for circadian rhythms. 10 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 147 148 **Figure 2. DCP5 is required for proper clock function. a-b** Circadian rhythm of leaf 149 movement. **(a)** Leaf angle and **(b)** period length estimation of entrained under LD 150 conditions and then transferred to LL. WT (Col-0) (n=12), _dcp5-1_ mutant (n=4), _dcp5-1_ 151 _DCP5_ (n=4). **c-d** Luciferase reporter assay. **(c)** Expression of the night core-clock reporter 152 _pTOC1:Luc_ and **(d)** period length estimation. The expression was measured for 6 d in LL 153 after entrainment under LD conditions. WT (Col-0) (n=12), _dcp5-1_ mutant (n=12). The 154 length of leaf movement rhythms and bioluminescence was estimated by the fast Fourier 11 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 155 transform nonlinear least squares (FFT-NLLS) test. Data are mean ± s.e.m. ***P < 0.001 156 (One-way ANOVA and two-tailed Student’s t-test). 12 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 157 _DCP5 is a global regulator of RNA stability_ 158 To understand how DCP5 shapes the RNA decay landscape, we performed a cordycepin 159 transcriptional arrest assay followed by whole-transcriptome RNA sequencing. Wild-type 160 and _dcp5_ plants were treated with cordycepin, and samples were collected 0, 30, 60, and 161 90 min later. After sequencing, transcript abundances were used to estimate mRNA 162 decay rates in each genotype. RNA half-lives spanned a wide range, as previously 163 reported [32], with median values of 60 min and 67 min in wild-type and _dcp5_ plants, 164 respectively. Among the genes with the fastest decay (half-life < 10 min), three belonged 165 to the _SMALL AUXIN UP REGULATED_ ( _SAUR_ ) gene family (AT2G21210, AT4G38840, 166 and AT4G31320), which have previously been described as short-lived [40] . Of the 13,698 167 genes that passed the minimum expression threshold and met the model-fitting criteria, 168 3,206 were more stable in _dcp5_ plants, whereas 1,435 were more stable in wild-type 169 plants. Loss of DCP5 mainly affects genes with the shortest half-lives, with 80% of those 170 genes having a half-life shorter than 90 min (Fig. 3a). A Gene Ontology (GO) term 171 enrichment analysis revealed that categories related to mRNA translation, photosynthesis 172 and response to environmental stimuli, which are in part regulated by the biological clock, 173 were overrepresented among transcripts stabilized in the _dcp5_ mutant (Fig. 3b). 174 We found that the half-life tendencies of CCG mRNAs previously observed in other 175 datasets (Fig. 1c-f and Supplementary Fig. 1) were conserved in our dataset (Fig. 3c; 176 Supplementary Fig. 3). Consistent with the clock phenotype observed in the _dcp5_ mutant 177 (Fig. 2 and Supplementary Fig. 2), CCGs decayed more slowly in the mutant (Fig. 3c). 178 Among HA-CCGs, 280 showed no difference between genotypes and 26 were stabilized 179 in WT (Supplementary Fig. 4), while 111 had longer half-lives in _dcp5_ (Fig. 3d). Notably, 13 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 180 only a few core clock genes showed differences in their decay. This was also observed 181 in the _xrn4_ mutant [22], suggesting that the clock phenotypes observed in RNA decay 182 mutants may reflect coordinated changes in the decay of multiple clock-associated 183 transcripts rather than effects restricted to a small subset of core clock genes. 14 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 184 185 **Figure 3. DCP5 controls mRNA half-lives globally.** **a** Distribution of mRNA half-lives in 186 WT and _dcp5_ for transcripts stabilized in _dcp5_ (n = 3,206), stabilized in WT (n = 1,435) 187 and with unchanged stability (n = 9,070). The dashed lines mark the mean half-life in each 188 genotype. Genes with half-lives of 6,931.47 min (α = 0.0001 min ⁻ ¹, the lower bound of the 189 model) in both genotypes were excluded. **b** Gene Ontology (GO) term (biological process) 190 enrichment analysis of transcripts stabilized in the _dcp5_ mutant. Manually selected terms 191 are shown. X-axis position gives fold enrichment (observed/expected number of genes); 192 the dashed line marks no enrichment (fold = 1). Bubble color indicates significance (−log ₁₀ 193 adjusted P; one-sided hypergeometric test, Benjamini–Hochberg correction); the number 194 within each point is the count of genes from the list annotated to that term. **c** RNA half 195 lives distribution across high amplitude (HA-), low amplitude (LA-), and non- clock 196 controlled genes (CCGs) in WT and _dcp5_ plants. Within each group, WT and _dcp5_ half 15 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 197 lives were compared gene by gene using a two-sided Wilcoxon signed-rank test, with P 198 values adjusted across the three groups by the Benjamini–Hochberg method (*** adjusted 199 P < 0.001) **d** Decay profiles of HA-CCGs stabilized in _dcp5_ (n = 111). Heatmaps show 200 transcript abundance at 30, 60 and 90 min after cordycepin treatment, relative to t0, for 201 WT (left) and _dcp5_ (right). Rows are genes, in the same order in both panels, sorted by 202 WT half-life (shortest at the top). 16 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 203 _High-amplitude transcripts preferentially bind to DCP5_ 204 To identify the RNAs that are bound by DCP5 _in vivo_ and the specific binding nucleotide 205 sequence we performed individual-nucleotide resolution UV crosslinking and 206 immunoprecipitation (iCLIP). We collected samples of _dcp5_ seedlings expressing DCP5 207 YFP-HA at two time points to capture CCG transcripts across the circadian cycle (Fig. 208 4a). We detected 96,632 binding sites in 9,116 genes, with 67.7% of targets detected at 209 both time points (Fig. 4b). Although the number of transcripts that were present in both 210 datasets was high, some degree of time specificity was evident when we examined those 211 transcripts with expression peaks at CT12 and CT24 (Supplementary Figs. 5 and 6). A 212 similar high overlap but also time-of-day dependent targets was observed for the clock 213 regulated RNA-binding protein _At_ GRP7 at CT24, the circadian minimum, and at CT36, 214 the circadian maximum of its expression [41] . 215 Consistent with its role in mRNA decay, 97% of DCP5 targets were protein-coding, a 216 significant overrepresentation relative to the 84% of protein-coding genes in the 217 Arabidopsis genome (p < 0.001, hypergeometric test) (Fig. 4c). Binding sites were 218 preferentially located in the 3’UTR and only 8.5% (772 genes) of all target transcripts 219 lacked a binding site in the 3’UTR (Fig. 4d, Supplementary Fig. 7). 5′UTR sites peaked 220 ~2 nt upstream of the start codon, whereas 3′UTR sites peaked ~220 nt downstream of 221 the stop codon, typically near the 3′ end of the transcript (Fig. 4e, Supplementary Fig. 8). 222 A motif search within a 28 nt window on either side of each binding site revealed 223 enrichment for AU-rich sequences (Fig. 4f). Of all DCP5 targets, 85.8% contained at least 224 one of these motifs near the binding site (Fig. 4g). The binding sites usually overlapped 225 with these motifs (Fig. 4h). A transcriptome-wide search, across both DCP5-bound and 17 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 226 unbound transcripts, showed that these motifs were most densely distributed within 227 coding sequences (CDS) (Supplementary Fig. 9). However, DCP5 bound these motifs 228 preferentially when they were located in the 3′UTR (Supplementary Fig. 9), suggesting 229 that additional trans-acting factors are required to recruit DCP5. 230 The half-lives of DCP5 targets had a median value of 59 min in WT and 72 min in _dcp5_ 231 (paired Wilcoxon signed-rank test, P < 0.001), while the difference in the half-lives of non 232 targets was less pronounced, with a median of 61 min and 64 min in WT and _dcp5_ (paired 233 Wilcoxon signed-rank test, P < 0.001), respectively (Supplementary Fig. 10). 234 Notably, DCP5 bound the transcripts of most HA-CCGs (chi-square test, p < 0.001) (Fig. 235 5a). These targets included most core clock genes. Moreover, both the binding site 236 density and the PureCLIP scores, which reflect the confidence and strength of individual 237 binding sites, were higher in HA-CCGs than in LA-CCGs and non-CCGs (Fig. 5b, c). 238 These results suggest that DCP5 binds and likely regulates the stability of multiple clock 239 gene transcripts and many of their downstream circadian output genes. 18 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 240 241 **Figure 4. DCP5 binds preferentially protein-coding RNAs in their 3’UTR. a** 242 Experimental design. _dcp5-1_ plants expressing DCP5-YFP-HA were entrained under 12 243 h light/12 h dark cycles at 22 °C and transferred to continuous light. Samples were 244 collected at subjective dusk (CT12) and subjective dawn (CT24) for iCLIP. **b** Overlap 245 between genes with DCP5 binding sites at CT12 and CT24. **c** Number of DCP5 target 246 genes by biotype (CT12 and CT24 combined). Short non-coding includes snoRNA, tRNA, 247 snRNA, miRNA and rRNA. Long non-coding includes lncRNA and other ncRNA. **d** 19 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 248 Density of DCP5 binding sites in 5′ UTRs, CDSs and 3′ UTRs, as binding sites per 100 249 kb of the total length of each region. Both time points combined. **e** Distribution of distances 250 (nt) from binding sites to the start codon (5′ UTR and CDS sites) or to the stop codon 251 (CDS and 3′ UTR sites). Negative values are upstream of the respective codon. **f** 252 Sequence logos of the three top-scoring motifs found by SSMART in 64 nt windows 253 centered on binding sites (both time points combined). **g** UpSet plot of DCP5 target genes 254 grouped by which motifs occur in their binding sites. None: genes with no binding site 255 matching any of the three motifs. Intersections with fewer than 10 genes are not shown. 256 **h** Distribution of distances (nt) between each motif occurrence shown in (f) and the 257 nearest DCP5 binding site within 25 nt. Positive values mean the motif is downstream of 258 the binding site. (IUPAC code, W = A/U; M = A/C; Y = C/U). 20 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 259 260 **Figure 5. High-amplitude clock-controlled genes (HA-CCG) are overrepresented** 261 **among DCP5 targets.** **a** Proportion of HA-CCG, LA-CCG and non-CCGs that are 262 targeted by DCP5 (chi-square test, p < 0.001). CCGs and their relative amplitudes are 263 from Romanowski _et al_ . (2020). **b** Distribution of binding-site density (binding sites per kb 264 of the longest transcript) in DCP5 target genes, by CCG category. **c** PureCLIP scores of 265 DCP5 binding sites in HA-CCGs (n = 6,936 sites in 447 genes), LA-CCGs (n = 60,119 266 sites in 5,047 genes) and non-CCGs (n = 29,577 sites in 3,622 genes), CT12 and CT24 267 combined. Kruskal–Wallis test followed by Dunn’s test with Benjamini–Hochberg 268 correction; ***, adjusted P < 0.001. 21 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 269 _DCP5 regulates at least 25% of the transcriptome_ 270 To assess the impact of DCP5 on the transcriptome, we performed RNA-seq on wild-type 271 and _dcp5_ plants collected every 4 h over one day in free-running conditions after 272 entrainment under a 12 h light/12 h dark photoperiod (Fig. 6a). We then applied the 273 Jonckheere-Terpstra-Kendall (JTK) algorithm to identify genes undergoing circadian 274 oscillations, detecting 8,036 CCGs in WT (JTK BH.Q < 0.01). Of these, 5,399 genes 275 (67%) overlapped with CCGs identified by Romanowski _et al._ (2020) [31], with a relative 276 amplitude Pearson correlation of 0.83 (Fig. 6b, Supplementary Fig. 11a). Also, we 277 observed a distribution of relative amplitudes comparable to that previously observed in 278 the dataset from Romanowski _et al._ (2020) [31], with most core clock genes classified as 279 HA-CCGs (Supplementary Fig. 11b). 280 We also detected 5,711 CCGs in _dcp5_, including 4,065 genes (71%) that were also 281 detected as CCGs in our WT dataset (Fig. 6b). The Pearson correlation between relative 282 amplitudes in WT and _dcp5_ was 0.89, suggesting that DCP5 does not substantially alter 283 the relative amplitude pattern across genes (Supplementary Figure 12a). However, we 284 observed similar phase shifts in 62% of shared CCGs (Fig. 6c, Supplementary Fig. 12b). 285 These shifts are evident in the expression patterns of some core clock genes in _dcp5_ (Fig. 286 6d-g), which are consistent with the lengthened period observed in the leaf movement 287 and luciferase reporter assays (Fig. 2). 288 To further understand the effects of DCP5 on the transcriptome, we performed a 289 differential expression analysis. We defined differentially expressed genes (DEGs) 290 between WT and _dcp5_ as genes whose expression was significantly affected by genotype 291 across the time course, either at all time points or in a time-of-day-dependent manner 22 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 292 (likelihood-ratio test, adjusted P < 0.05). DEGs that showed a change of at least 1.5-fold 293 at one or more time points (adjusted P < 0.05) were classified as “robust”, and the 294 remainder as “soft” (See Methods). Thus, we identified 801 robust and 5,964 soft DEGs. 295 The subset of robust DEGs showed the highest proportions of DCP5 targets, transcript 296 stabilized in _dcp5_, and HA-CCGs, while the subset of soft DEGs showed larger 297 proportions in those categories than non-DEGs (Supplementary Fig. 13). Among robust 298 DEGs, only 27% (220 out of 801) showed an expression difference of |log2FoldChange| 299 - log2(1.5) at two or more time points. This suggests the DCP5 effect is time-dependent. 300 Notably, most core clock genes (17 of 27) were differentially expressed (4 soft and 13 301 robust DEGs). 302 Altogether, these findings suggest that loss of DCP5 does not produce large changes in 303 overall transcript levels but instead alters temporal expression patterns throughout the 304 day. We identified a subset of genes with altered circadian expression patterns that was 305 enriched for DCP5 targets. These HA-CCGs do not reach higher peak expression levels 306 in the _dcp5_ mutant but instead exhibit a shift in the time of peak expression, consistent 307 with a longer period, suggesting that DCP5 contributes to the decline in transcript 308 abundance after peak expression. 23 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 309 310 **Figure 6. The circadian transcriptome needs DCP5 for a correct time of peak** 311 **expression. a** Experimental design. WT (Col-0) and _dcp5-1_ seedlings were entrained 312 under 12 h light:12 h dark cycles at 22 °C and then moved to continuous light. Samples 313 were collected every 4 h from CT12 to CT32 (six time points; n = 2 biological replicates). 314 **b** Euler diagram showing how circadian clock-regulated genes (CCGs) overlap between 315 three sets: WT and _dcp5-1_ in this study, and WT in Romanowski _et al._ (2020). In this 316 study, CCGs were identified with JTK_CYCLE (BH-adjusted q < 0.01). **c** Peak phase of 317 each gene in WT versus _dcp5-1_, estimated with JTK_CYCLE. Only genes rhythmic in 318 both genotypes are shown (BH-adjusted q < 0.01; n = 4,065). **d-k** Expression of 24 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 319 representative clock genes **d** _CCA1_, **e** _LHY_, **f** _RVE8_, **g** _ELF3_, **h** _LNK2_, **i** _PRR9_, **j** _PRR5_ 320 and **k** _GRP7_ . Points show mean TPM ± s.e.m. (n = 2), and lines are loess fits. Gray and 321 white bars mark subjective night and subjective day. 25 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 322 **Discussion** 323 Here we show that the mRNA decapping activator DCP5 is a key component to maintain 324 a correct expression phase of transcripts in the circadian network in Arabidopsis. Using 325 iCLIP, we found that DCP5 binds roughly one-third of the transcriptome, with marked 326 enrichment for high-amplitude clock-controlled genes (CCGs). The _dcp5_ mutant 327 displayed long-period rhythms in leaf movement and clock-gene expression, together with 328 a global delay in the phase of peak expression of CCGs. DCP5 targets were more stable 329 in _dcp5_, consistent with a role for DCP5 in promoting mRNA decay. The number of 330 transcripts stabilized in _dcp5_ is smaller than that in other decay mutants such as _vcs_ or 331 _rh6/8/12_ [32,33] . This could be due to differences in experimental design and may also be 332 related to the lethal phenotypes associated with the loss of VCS or the combined loss of 333 RH6, RH8, and RH12 [33,39,42] . A small subset of targets instead had shorter half-lives in the 334 mutant, which could reflect an indirect effect whereby accumulated transcripts are 335 channeled into the RNA quality control small interfering RNA (rqc-siRNA) pathway [43,44] . 336 The long-period phenotype of _dcp5_ mirrors the clock defects of other RNA-turnover 337 mutants, including _xrn4_ and _lsm1_ [22,23] . The generally milder phenotypes of _xrn4_ compared 338 with _dcp5_ and _lsm1_ suggest a hierarchical organization in which DCP5 and LSM1 act 339 upstream of XRN4. By sequestering and translationally repressing target mRNAs, DCP5 340 and LSM1 may commit transcripts to translational repression even in the absence of 341 XRN4. In _dcp5_ and _lsm1_ mutants, by contrast, XRN4 would both lose this trigger and 342 degrade its substrates less efficiently, because they are no longer concentrated in 343 cytoplasmic granules. Similarly, the _dst1_ mutant, in which clock transcripts have longer 344 half-lives, also has a long period [45] . In contrast, loss of the cap-binding factor eIF4E, which 26 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 345 shields transcripts from degradation [46], shortens the circadian period [47] . Likewise, in 346 Drosophila, knockdown of the translation-promoting factor LARK shortens the period [48] . 347 Together, these observations support a model in which impaired decay and mRNA 348 accumulation lengthen the period (Fig. 7), whereas loss of mRNA stabilization 349 accelerates the clock. 350 Interestingly, DCP5 transcript levels do not oscillate, suggesting that _DCP5_ expression is 351 not under clock control. However, DCP5 interacts with many RNA-binding proteins whose 352 mRNAs do oscillate [31,49], raising the possibility that its activity or target specificity is 353 modulated rhythmically through its interactome. Also, it has been reported that DCP5 354 physically interacts with multiple poly(A) binding proteins (PAB2, PAB3, PAB4, PAB5 and 355 PAB8) [49], which is consistent with DCP5 preference for binding at the end of 3’ UTRs. 356 Besides its role in the circadian clock, DCP5 is involved in multiple physiological 357 processes such as photomorphogenesis [50], response to osmotic stress [51,52] and response 358 to hormones [53] . All these processes presumably require a re-programming of the 359 transcriptome, in which the DCP5 interactome could have a major role in selecting its 360 target transcripts. 361 More broadly, our results identify DCP5 as a post-transcriptional hub through which 362 control of mRNA stability shapes clock timing. A constitutively expressed decapping 363 activator whose activity is tuned by a rhythmic interactome could provide robust yet 364 flexible control over large sets of transcripts. An important open question is the extent to 365 which this regulation propagates to the translatome and proteome, particularly given that 366 a substantial fraction of rhythmic proteins arise from non-rhythmic transcripts [54] . 27 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 367 Addressing this question will be central to understanding how RNA turnover contributes 368 to rhythmic physiology in plants. 28 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 369 370 **Figure 7. Proposed role of RNA decay machinery in pacing circadian oscillations.** 371 In WT after the peak of expression is reached, mRNA enters a decreasing phase until it 372 reaches its minimum. When the decay machinery is impaired, such as in the _dcp5_ mutant, 373 the mRNA degradation is slower resulting in an extended circadian period. 29 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 374 **Methods** 375 _mRNA half-life meta-analysis_ 376 Half-lives were extracted from published Supplementary Tables. When needed, half-lives 377 were estimated from decay rate constant (alpha) as t 1/2 =ln(2)/alpha. In the case of Szabo 378 _et al._ (2020) compound model estimations were used. 379 _Plants and growth conditions_ 380 Plants were grown in pots containing a mixture of organic substrate, vermiculite, and 381 perlite [0.5:1:1 (vol/vol)] at the particular photoperiod and temperature conditions 382 indicated in each experiment. The mutant _dcp5-1_ (SALK_106536) was obtained from the 383 Arabidopsis Biological Resource Center (ABRC). 384 _Generation of dcp5 DCP5 transgenic plants_ 385 DCP5 CDS was cloned from WT cDNA and introduced in pDONR/Zeo vector (Invitrogen) 386 via the Gateway method. A binary vector containing this fragment was generated by LR 387 recombination with pEarlyGate 101, thus obtaining the DCP5-YFP-HA protein under the 388 control of the CaMV35S promoter. This plasmid was introduced into _Agrobacterium_ 389 _tumefaciens_ GV3101 strain and transformed into _dcp5-1_ plants using the floral dip method 390 55 . Transgenic lines were selected on Murashige and Skoog media supplemented with 15 391 mg/L PPT (Duchefa). 392 _Leaf movement assay_ 393 For leaf movement analysis, plants were entrained under a long-day photoperiod (16 h 394 light/8 h dark cycles, 80 μmol m ⁻ ² s ⁻ ¹), transferred to continuous white light, and the angle 395 of the first leaf pair was recorded every 1 h for 5–6 days using ImageJ (imagej.nih.gov/ij/). 30 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 396 Period estimates were calculated using BioDare2 (biodare2.ed.ac.uk) and analyzed by 397 FFT-NLLS. Statistical analysis was performed using a two-tailed Student's t-test or a one 398 way ANOVA, as appropriate. 399 _Luciferase reporter assays_ 400 Seedlings were grown on Murashige and Skoog medium with 0.8% agar supplemented 401 with 1% sucrose in a 96-well plate. One seed was placed per well, and the plate was kept 402 in darkness at 4 °C for 3 days. The seedlings were then entrained under a long-day cycle 403 (16 h light/8 h dark, 80 μmol m ⁻ ² s ⁻ ¹). After 7 days, luciferin was added to each plant, the 404 plate was transferred to constant light (50 μmol m ⁻ ² s ⁻ ¹) and temperature conditions (22 405 °C) and placed in a luminometer to measure the bioluminescence emitted by each 406 seedling every 2 h. After 5–6 days, data analysis was performed using Mikrowin 2000 407 software (version 4.29) (Mikrotek Laborsysteme). The luminometer used was the Centro 408 LB-960 microplate reader (Berthold Technologies). Period estimates were calculated 409 using BioDare2 (biodare2.ed.ac.uk) and analyzed by FFT-NLLS. Statistical analysis was 410 performed using a two-tailed Student's t-test. 411 _Hypocotyl elongation assays_ 412 Seeds were sown in pots, stratified for 2 days at 4 °C in darkness, and then grown in a 413 12 h light/ 12 h of dark photoperiod at 22 °C (80 μmol m ⁻ ² s ⁻ ¹). After germination, plants 414 were kept either in the same condition or transferred to the same photoperiod (80 μmol 415 m ⁻ ² s ⁻ ¹) with 28 °C days and 22 °C nights. Photographs of the plants were taken every 1 416 h and the hypocotyls were measured using ImageJ (imagej.nih.gov/ij/). 31 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 417 _Cordycepin time course experiment and mRNA decay analysis_ 418 Plants were grown in continuous light (LL, 50 μmol m ⁻ ² s ⁻ ¹) from germination, with no 419 entrainment. After 2 weeks, seedlings were transferred to flasks containing a buffer (1mM 420 1,4-piperazinediethanesulfonic acid, pH 6.25, 1mM sodium citrate, 1mM KCl and 15mM 421 sucrose), and after a 30-min incubation, cordycepin (0.6mM) was added, vacuum 422 infiltration was performed and the samples were taken at 0, 30, 60 and 90min. Total RNA 423 was extracted using TRIzol reagent (Invitrogen) and samples were stored for shipment in 424 GenTegra RNA tubes. Sequencing libraries were generated using Abclonal Fast RNA 425 seq Lib Prep Kit V2 (RK20306) with polyA capture by VAHTS mRNA Capture Beads 426 (Cat.No.N401-01/02) and sequenced on the Illumina NovaSeq X Plus platform, providing 427 150-bp paired-end reads (Novogene, Singapore). 428 The RNA sequencing reads were mapped to the Arabidopsis thaliana TAIR10 genome [56] 429 using STAR (version 2.7.11) [57] with the following parameters: --twopassMode Basic - 430 outFilterMultimapNmax 2 --outFilterType BySJout --outSJfilterReads Unique - 431 sjdbOverhang 149 --alignSJoverhangMin 6 --alignSJDBoverhangMin 3 --alignIntronMin 432 20 --alignIntronMax 5000. 433 Modeling of RNA decay was performed as previously reported [32,33] . Briefly, each 434 transcript's TPM expression was normalized to its genotype-specific mean T0 value and 435 then corrected by a decay factor estimated separately for each genotype × time point 436 condition. Decay factors were derived from reference genes selected as those that (i) 437 were among the 500 most highly expressed genes at T0, (ii) showed no significant 438 expression difference between genotypes at T0 (excluded if padj < 0.05 and |log2FC| ≥ 439 1), and (iii) showed monotonically increasing T0-normalized expression across all four 32 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 440 time points (T0 < T30 < T60 < T90) in both WT and _dcp5_ . Using those criteria, we found 441 and used for the decay factors 37 reference genes: AT3G09260, AT5G26000, 442 AT1G54410, AT5G35630, AT1G23130, AT4G04640, AT4G37930, AT5G42980, 443 AT5G25980, AT3G14420, AT1G23310, AT1G02780, AT1G62380, AT1G11860, 444 AT2G36530, AT1G20020, AT2G30950, AT1G78630, AT3G14415, AT1G43170, 445 AT5G44340, AT5G09660, AT4G02520, AT1G09690, AT1G09340, AT5G39570, 446 AT3G11940, AT5G30510, AT5G47210, AT1G67430, AT3G54050, AT3G63490, 447 AT5G02870, AT2G33450, AT3G62870, AT5G42270 and AT3G46780. 448 Decay modeling was performed on nuclear encoded genes with 1 TPM or more at T0 in 449 at least one genotype. Normalized decay trajectories were modeled per gene using the 450 RNAdecay Bioconductor package (version 1.26.0), with WT and _dcp5_ as the two 451 treatment groups and model selection by corrected Akaike information criterion (AICc). 452 Half-life was calculated as ln(2)/α from the best-fitting model per gene. Genes were 453 retained for downstream analysis only if the model fit passed a residual-variance quality 454 filter (σ² < 0.0625). 455 _iCLIP data generation_ 456 The iCLIP experiment was performed following the protocol of Lewinski _et al_ . (2024) [58] . 457 Plants expressing DCP5 fused to YFP, or to GFP as a negative control, were grown under 458 a 12:12 photoperiod (12 h light/12 h dark cycles, 80 μmol m ⁻ ² s ⁻ ¹). After two weeks, they 459 were transferred to continuous light. Three biological replicates of DCP5-YFP-HA and two 460 of GFP were collected per time point, at 12 and 24 hours after transfer. Samples were 461 treated with a UV light pulse (2,000 mJ/cm²) to covalently crosslink proteins to RNA and 462 were then immediately frozen in liquid nitrogen. Following cell lysis of the tissue, DCP5 33 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 463 YFP was immunoprecipitated with GFP-Trap agarose beads (Proteintech, cat. no. gta 464 20). Treatment with Turbo DNase (Invitrogen) and RNase I (Invitrogen) was carried out 465 to remove DNA in all samples and RNA in samples used for protein size reference. RNA 466 was labeled with γ-³²P, resolved by polyacrylamide gel electrophoresis, transferred to a 467 nitrocellulose membrane, and visualized by autoradiography. RNA of the appropriate size 468 was recovered from the nitrocellulose membrane and used to construct libraries for 469 sequencing. Sequencing was performed on the Illumina NextSeq 2000 platform at the 470 Institute of Molecular Biology in Mainz, Germany. 471 _iCLIP data analysis_ 472 Read processing and the identification of binding peaks were carried out following the 473 bioinformatic protocol of Lewinski _et al_ . (2024) [58] . Briefly, the sequencing reads from the 474 iCLIP experiment were sorted by biological replicate and condition, and the Illumina 475 adapter sequences were then removed using the Flexbar package [59] . The processed 476 reads were aligned to the TAIR10 genome with the AtRTD3 annotation using STAR [56,57] . 477 PCR duplicates were removed using umi_tools [60] . Binding peaks were determined using 478 PureCLIP [61] . The 2% of peaks with the worst global PureCLIP score were discarded, and 479 from the remaining peaks the 30% with the worst score within each gene were discarded. 480 Finally, the width of the binding sites was determined using the R package 481 BindingSiteFinder [62] . 482 For the classification genes were grouped according to its biotype in protein coding 483 (biotypes: protein coding, and non-translating CDS), long non-coding (biotypes: lncRNA 484 and ncRNA) and short non-coding (biotypes: miRNA, snoRNA, snRNA, rRNA and tRNA). 485 To estimate the mean distance to start and stop codons, we calculated the mean distance 34 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 486 across isoforms for each gene. For relativize binding sites to the length of the different 487 regions, we used the primary isoform according to Phytozome. 488 _Motif discovery_ 489 We searched for sequence–structure motifs enriched at DCP5 binding sites with 490 SSMART v1.2 [63] . Each binding site was extended from the center by 28 nt on either side, 491 giving 63 nt windows; windows were trimmed at transcript ends. Identical sequences 492 within each time point, which arise when a binding site maps to several isoforms, were 493 collapsed into one. This left 30,254 sequences for CT12 and 71,435 for CT24. 494 The secondary-structure context of each sequence was predicted with 495 RNAStructProfiling through the SSMART wrapper, which labels each nucleotide as paired 496 or unpaired. Sequence identifiers were tagged with their time point, and the two datasets 497 were combined into a single input of 101,689 sequences. SSMART was run with default 498 parameters. Motifs were ranked by SSMART score, and the top-scoring motifs were 499 reported. Position weight matrices were plotted with ggseqlogo. 500 _Time course sequencing experiment_ 501 For RNA sequencing, wild-type and _dcp5-1_ plants were grown under a 12:12 photoperiod 502 (12 h light/12 h dark cycles, 80 μmol m ⁻ ² s ⁻ ¹). After two weeks, the plants were transferred 503 to continuous light (22 °C), and two biological replicates were harvested at 12, 16, 20, 24, 504 28, and 32 h in continuous light, with half of the samples treated with a 4-h pulse at 28 °C 505 prior to harvest. Total RNA was extracted with TRIzol (Invitrogen), and libraries were 506 prepared with the mRNA Library Prep Kit (Watchmaker). The libraries were sequenced 507 on the Aviti platform (Element Biosciences) at the Stowers Institute for Medical Research. 35 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 508 _Time course sequencing experiment analysis_ 509 The RNA sequencing reads were mapped to the _Arabidopsis thaliana_ TAIR10 genome [56] 510 using STAR 2.7.11 [57] with the following parameters: --twopassMode Basic - 511 outFilterMultimapNmax 2 --outFilterType BySJout --outSJfilterReads Unique - 512 sjdbOverhang 149 --alignSJoverhangMin 6 --alignSJDBoverhangMin 3 --alignIntronMin 513 20 --alignIntronMax 5000. Raw counts were obtained with htseq-count. 514 _Transcriptome rhythmicity analysis_ 515 Rhythmic genes were identified separately for each genotype with JTK_CYCLE as 516 implemented in the meta3d function of MetaCycle (version 1.2.1) [64], using TPM values. 517 Genes were kept if TPM > 1 in at least 12 of the 24 22 °C libraries. Because the time 518 series covers a single 20-h window, the six time points were concatenated to form a 519 second cycle (CT36–CT56), and JTK_CYCLE was run on the resulting 12-time point 520 series with the default period range (20–28 h). Genes with a Benjamini–Hochberg 521 adjusted q < 0.01 were called circadian clock-regulated genes (CCGs), and their phase 522 was taken as the JTK-adjusted phase. For each genotype, relative amplitude was 523 calculated from replicate-averaged TPM as (max − min) / (min + 1) across the six time 524 points. CCGs with relative amplitude > 4 were called high-amplitude CCGs. 525 _Differential expression analysis between genotypes_ 526 Differential expression between wild-type (WT) and _dcp5_ plants at 22 °C was analyzed 527 from raw gene counts using DESeq2 (version 1.46.0) [65] . Genes were retained for 528 downstream analyses if they had at least 10 raw counts in at least two samples, 529 corresponding to the size of the smallest experimental group. This filtering step was 36 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 530 performed once across all 48 libraries to ensure that the same set of genes was used in 531 all subsequent analyses. 532 To assess genotype-dependent expression, two nested likelihood-ratio tests (LRTs) were 533 performed. The genotype main effect was tested by comparing the full model ~ time + 534 genotype against the reduced model ~ time. The genotype-by-time interaction was tested 535 by comparing the full model ~ time + genotype + time:genotype against the reduced 536 model ~ time + genotype. For each LRT, dispersion estimates were obtained under the 537 corresponding full model. Genes with an adjusted P value < 0.05 in either LRT were 538 considered differentially expressed at the global level. 539 To identify large, time point-specific differences, WT and _dcp5_ were also compared 540 separately at each of the six circadian times (CT12–CT32). For this, the 22 °C samples 541 were fitted with a single factor combining genotype and time (~ group; 12 levels), and 542 each pairwise contrast (WT vs. _dcp5_ at a given CT) was tested with a Wald test. Fitting 543 all samples in one model let the six contrasts share size factors and dispersion estimates 544 from the full dataset. Log ₂ fold changes were shrunk with ashr (Stephens, 2017), and P 545 values were adjusted across genes with the Benjamini–Hochberg method within each 546 contrast. A difference at a given time point was considered large when adjusted P < 0.05 547 and |log ₂ FC| > log ₂ (1.5). 548 Genes were subsequently classified into three categories. Robust differentially expressed 549 genes (DEGs) were defined as genes with an adjusted P value < 0.05 in either LRT and 550 at least one time point with an adjusted P value < 0.05 and an absolute shrunken log ₂ 551 fold change greater than log ₂ (1.5) in the pairwise WT-versus- _dcp5_ contrast. Soft DEGs 552 were genes significant in either LRT (adjusted P < 0.05) but without any time point 37 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 553 meeting the per-time-point significance and fold-change criteria. Non-DEGs comprised 554 genes that were not significant in either LRT, including genes that met the per-time-point 555 criteria alone (n = 70). 556 Phase differences between genotypes were calculated only for genes that were CCGs in 557 both genotypes, using circular arithmetic on a 24-h period: Δφ = ((φdcp5 − φWT + 12) 558 mod 24) − 12. Antipodal values (−12 h) were set to +12 h. 559 _Identification of warm-temperature responsive genes_ 560 Warm-temperature responsive genes (WTRGs) were identified independently in wild type 561 and _dcp5_ using DESeq2. For each genotype, the analysis included 24 RNA-seq libraries, 562 corresponding to six circadian time points, two temperatures, and two biological replicates 563 per condition. Genes were classified as WTRGs if they showed a statistically significant 564 temperature effect across the time course and an absolute temperature-dependent 565 change greater than 1.5-fold at one or more time points. 566 The temperature effect was assessed using two nested likelihood ratio tests (LRTs). The 567 first LRT compared the model ~ time + temperature with the reduced model ~ time, 568 thereby identifying genes responding to temperature consistently across all time points. 569 The second LRT compared ~ time + temperature + time:temperature with ~ time + 570 temperature, thereby identifying genes for which the magnitude or direction of the 571 temperature response depended on time of day. For each LRT, dispersion was estimated 572 under the corresponding full model. Genes with a Benjamini–Hochberg-adjusted _P_ value 573 ( _padj_ ) < 0.05 in either LRT were retained for subsequent time-point-level testing. 38 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 574 Within each genotype, a model containing 12 levels corresponding to the temperature 575 by-time-point combinations was fitted. The 28 °C versus 22 °C contrast was then 576 computed at each of the six time points. Because the 28 °C treatment consisted of a 4 h 577 pulse that ended at harvest, each contrast represents the acute transcriptional response 578 to warm temperature during the 4 h preceding the indicated circadian time. A time point 579 was considered significant when _padj_ < 0.05 and the absolute _ashr_ -shrunk log2 fold 580 change exceeded log2(1.5). Positive log2 fold changes indicate higher expression at 28 581 °C than at 22 °C. 582 Genes satisfying both the gene-level LRT criterion and at least one significant time-point 583 level contrast were defined as WTRGs. This dual criterion excluded genes with 584 statistically significant but biologically small temperature effects, as well as genes whose 585 time-point-level response was not supported by the gene-level test. Time-point-level 586 differential-expression calls were used in downstream analyses only for genes classified 587 as WTRGs. 588 _GO term analysis_ 589 Gene Ontology (GO) enrichment of biological process terms was tested with 590 clusterProfiler (version 4.14.6), using annotations from org.At.tair.db (version 3.20.0). 591 Only terms annotated to at least 10 genes of the background set were tested. For each 592 gene set, enrichment was assessed with a one-sided hypergeometric test, and p-values 593 were adjusted for multiple testing with the Benjamini–Hochberg procedure. Terms with 594 an adjusted p-value below 0.05 were considered significantly enriched. Fold enrichment 595 was calculated as the proportion of genes in the set annotated to a term, divided by the 596 corresponding proportion in the background set. 39 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 597 Among genes whose decay model passed σ² < 0.0625, those with a lower decay rate in 598 _dcp5_ than in wild type were classified as stabilized in _dcp5_, and those with a higher decay 599 rate as stabilized in wild type. All 13,698 genes passing quality control served as 600 background. 40 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 601 **References** 602 1. Young, M. W. & Kay, S. A. Time zones: a comparative genetics of circadian clocks. 603 _Nature Reviews Genetics_ 2, 702–715 (2001). 604 2. Nakamichi, N. The transcriptional network in the arabidopsis circadian clock system. 605 _Genes_ 11, 1–13 (2020). 606 3. Kamioka, M. _et al._ Direct repression of evening genes by CIRCADIAN CLOCK 607 ASSOCIATED1 in the Arabidopsis circadian clock. _Plant Cell_ 28, 696–711 (2016). 608 4. Lu, S. X. _et al._ CCA1 and ELF3 interact in the control of hypocotyl length and 609 flowering time in arabidopsis. _Plant Physiology_ 158, 1079–1088 (2012). 610 5. Nagel, D. H. _et al._ Genome-wide identification of CCA1 targets uncovers an 611 expanded clock network in _Arabidopsis_ . _Proceedings of the National Academy of_ 612 _Sciences_ 112, (2015). 613 6. Pérez-García, P., Ma, Y., Yanovsky, M. J. & Mas, P. Time-dependent sequestration 614 of RVE8 by LNK proteins shapes the diurnal oscillation of anthocyanin biosynthesis. 615 _Proceedings of the National Academy of Sciences_ 112, 5249–5253 (2015). 616 7. Xie, Q. _et al._ LNK1 and LNK2 Are Transcriptional Coactivators in the _Arabidopsis_ 617 Circadian Oscillator. _The Plant Cell_ 26, 2843–2857 (2014). 618 8. Ma, Y., Gil, S., Grasser, K. D. & Mas, P. Targeted Recruitment of the Basal 619 Transcriptional Machinery by LNK Clock Components Controls the Circadian 620 Rhythms of Nascent RNAs in Arabidopsis. _The Plant Cell_ 30, 907–924 (2018). 621 9. Huang, W. _et al._ Mapping the core of the Arabidopsis circadian clock defines the 622 network structure of the oscillator. _Science_ 335, 75–79 (2012). 41 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 623 10. Nakamichi, N. _et al._ PSEUDO-RESPONSE ReGULATORS 9, 7, and 5 are 624 transcriptional repressors in the Arabidopsis circadian clock. _Plant Cell_ 22, 594–605 625 (2010). 626 11. Chow, B. Y., Helfer, A., Nusinow, D. A. & Kay, S. A. ELF3 recruitment to the 627 PRR9 promoter requires other Evening Complex members in the Arabidopsis 628 circadian clock. _Plant Signaling & Behavior_ 7, 170–173 (2012). 629 12. Dixon, L. E. _et al._ Temporal repression of core circadian genes is mediated 630 through EARLY FLOWERING 3 in Arabidopsis. _Current Biology_ 21, 120–125 (2011). 631 13. Helfer, A. _et al._ LUX ARRHYTHMO encodes a nighttime repressor of circadian 632 gene expression in the Arabidopsis core clock. _Current Biology_ 21, 126–133 (2011). 633 14. Herrero, E. _et al._ EARLY FLOWERING4 recruitment of EARLY FLOWERING3 in 634 the nucleus sustains the Arabidopsis circadian clock. _Plant Cell_ 24, 428–443 (2012). 635 15. Mizuno, T. _et al._ Ambient temperature signal feeds into the circadian clock 636 transcriptional circuitry through the EC night-time repressor in arabidopsis thaliana. 637 _Plant and Cell Physiology_ 55, 958–976 (2014). 638 16. Mateos, J. L., de Leone, M. J., Torchio, J., Reichel, M. & Staiger, D. Beyond 639 Transcription: Fine-Tuning of Circadian Timekeeping by Post-Transcriptional 640 Regulation. _Genes_ 9, 616 (2018). 641 17. Zhang, X. & Guo, H. mRNA decay in plants : both quantity and quality matter. 642 _Current Opinion in Plant Biology_ 35, 138–144 (2017). 643 18. James, A. B. _et al._ Alternative Splicing Mediates Responses of the _Arabidopsis_ 644 Circadian Clock to Temperature Changes. _The Plant Cell_ 24, 961–981 (2012). 42 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 645 19. Kwon, Y. J., Park, M. J., Kim, S. G., Baldwin, I. T. & Park, C. M. Alternative 646 splicing and nonsense-mediated decay of circadian clock genes under environmental 647 stress conditions in Arabidopsis. _BMC Plant Biology_ 14, (2014). 648 20. Wang, X. _et al._ A photoregulatory mechanism of the circadian clock in 649 Arabidopsis. _Nature Plants_ 7, 1397–1408 (2021). 650 21. Perez-Santángelo, S. _et al._ Role for LSM genes in the regulation of circadian 651 rhythms. _Proceedings of the National Academy of Sciences_ 111, 15166–15171 652 (2014). 653 22. Careno, D. A., Perez Santangelo, S., Macknight, R. C. & Yanovsky, M. J. The 5ʹ 654 3ʹ mRNA Decay Pathway Modulates the Plant Circadian Network in Arabidopsis. 655 _Plant and Cell Physiology_ 63, 1709–1719 (2022). 656 23. Prasetyaningrum, P. _et al._ Inhibition of RNA degradation integrates the metabolic 657 signals induced by osmotic stress into the Arabidopsis circadian system. _Journal of_ 658 _Experimental Botany_ 74, 5805–5819 (2023). 659 24. Agrofoglio, Y. C. _et al._ Arginine methylation of SM-LIKE PROTEIN 4 660 antagonistically affects alternative splicing during Arabidopsis stress responses. _Plant_ 661 _Cell_ 36, 2219–2237 (2024). 662 25. Blake, L. A., Watkins, L., Liu, Y., Inoue, T. & Wu, B. A rapid inducible RNA decay 663 system reveals fast mRNA decay in P-bodies. _Nature Communications_ 15, (2024). 664 26. Xu, J. & Chua, N.-H. _Arabidopsis_ Decapping 5 Is Required for mRNA Decapping, 665 P-Body Formation, and Translational Repression during Postembryonic 666 Development. _The Plant Cell_ 21, 3270–3279 (2009). 43 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 667 27. Schaffer, R. _et al._ The late elongated hypocotyl mutation of Arabidopsis disrupts 668 circadian rhythms and the photoperiodic control of flowering. _Cell_ 93, 1219–29 669 (1998). 670 28. Wang, Z. Y. & Tobin, E. M. Constitutive expression of the CIRCADIAN CLOCK 671 ASSOCIATED 1 (CCA1) gene disrupts circadian rhythms and suppresses its own 672 expression. _Cell_ 93, 1207–17 (1998). 673 29. Alabadí, D. _et al._ Reciprocal regulation between TOC1 and LHY/CCA1 within the 674 Arabidopsis circadian clock. _Science (New York, N.Y.)_ 293, 880–3 (2001). 675 30. Gendron, J. M. _et al._ Arabidopsis circadian clock protein, TOC1, is a DNA 676 binding transcription factor. _Proceedings of the National Academy of Sciences of the_ 677 _United States of America_ 109, 3167–72 (2012). 678 31. Romanowski, A., Schlaen, R. G., Perez ‐ Santangelo, S., Mancini, E. & Yanovsky, 679 M. J. Global transcriptome analysis reveals circadian control of splicing events in 680 Arabidopsis thaliana. _The Plant Journal_ 103, 889–902 (2020). 681 32. Sorenson, R. S., Deshotel, M. J., Johnson, K., Adler, F. R. & Sieburth, L. E. 682 Arabidopsis mRNA decay landscape arises from specialized RNA decay substrates, 683 decapping-mediated feedback, and redundancy. _Proceedings of the National_ 684 _Academy of Sciences_ 115, E1485–E1494 (2018). 685 33. Chantarachot, T. _et al._ DHH1/DDX6-like RNA helicases maintain ephemeral half 686 lives of stress-response mRNAs. _Nature Plants_ 6, 675–685 (2020). 687 34. Carpentier, M. C. _et al._ Genome-wide analysis of mRNA decay in Arabidopsis 688 shoot and root reveals the importance of co-translational mRNA decay in the general 689 mRNA turnover. _Nucleic Acids Research_ 52, 7910–7924 (2024). 44 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 690 35. Szabo, E. X. _et al._ Metabolic Labeling of RNAs Uncovers Hidden Features and 691 Dynamics of the Arabidopsis Transcriptome. _Plant Cell_ 32, 871–887 (2020). 692 36. Gutierrez, R. A., Ewing, R. M., Cherry, J. M. & Green, P. J. Identification of 693 unstable transcripts in Arabidopsis by cDNA microarray analysis: Rapid decay is 694 associated with a group of touch- and specific clock-controlled genes. _Proceedings of_ 695 _the National Academy of Sciences_ 99, 11513–11518 (2002). 696 37. Iwasaki, S., Takeda, A., Motose, H. & Watanabe, Y. Characterization of 697 Arabidopsis decapping proteins AtDCP1 and AtDCP2, which are essential for post 698 embryonic development. _FEBS Letters_ 581, 2455–2459 (2007). 699 38. Xu, J., Yang, J.-Y., Niu, Q.-W. & Chua, N.-H. Arabidopsis DCP2, DCP1, and 700 VARICOSE Form a Decapping Complex Required for Postembryonic Development. 701 _The Plant Cell_ 18, 3386–3398 (2007). 702 39. Xu, J., Yang, J.-Y., Niu, Q.-W. & Chua, N.-H. Arabidopsis DCP2, DCP1, and 703 VARICOSE form a decapping complex required for postembryonic development. _The_ 704 _Plant cell_ 18, 3386–3398 (2006). 705 40. Narsai, R., Howell, K. A., Millar, A. H., Small, I. & Whelan, J. Genome-Wide 706 Analysis of mRNA Decay Rates and Their Determinants in Arabidopsis thaliana. 19, 707 3418–3436 (2007). 708 41. Meyer, K. _et al._ Adaptation of iCLIP to plants determines the binding landscape 709 of the clock-regulated RNA-binding protein AtGRP7. _Genome Biology_ 18, 204 (2017). 710 42. Goeres, D. C. _et al._ Components of the Arabidopsis mRNA Decapping Complex 711 Are Required for Early Seedling Development. _the Plant Cell Online_ 19, 1549–1564 712 (2007). 45 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 713 43. De Alba, A. E. M. _et al._ In plants, decapping prevents RDR6-dependent 714 production of small interfering RNAs from endogenous mRNAs. _Nucleic Acids_ 715 _Research_ 43, 2902–2913 (2015). 716 44. Hoffmann, G. _et al._ Arabidopsis RNA processing body components LSM1 and 717 DCP5 aid in the evasion of translational repression during Cauliflower mosaic virus 718 infection. _Plant Cell_ 34, 3128–3147 (2022). 719 45. Lidder, P., Gutiérrez, R. A., Salomé, P. A., McClung, C. R. & Green, P. J. 720 Circadian Control of Messenger RNA Stability. Association with a Sequence-Specific 721 Messenger RNA Decay Pathway. _Plant Physiology_ 138, 2374–2385 (2005). 722 46. Batool, A., Aashaq, S. & Andrabi, K. I. Eukaryotic initiation factor 4E (eIF4E): A 723 recap of the cap ‐ binding protein. _J of Cellular Biochemistry_ 120, 14201–14212 724 (2019). 725 47. Zhang, J. _et al._ The eukaryotic translation initiation factor eIF4E regulates 726 flowering and circadian rhythm in Arabidopsis. _Plant Journal_ 727 https://doi.org/10.1111/tpj.16975 (2024) doi:10.1111/tpj.16975. 728 48. Huang, Y., McNeil, G. P. & Jackson, F. R. Translational Regulation of the 729 DOUBLETIME/CKIδ/ε Kinase by LARK Contributes to Circadian Period Modulation. 730 _PLoS Genet_ 10, e1004536 (2014). 731 49. Chicois, C. _et al._ The UPF1 interactome reveals interaction networks between 732 RNA degradation and translation repression factors in Arabidopsis. _The Plant Journal_ 733 96, 119–132 (2018). 46 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 734 50. Jang, G.-J., Yang, J.-Y., Hsieh, H.-L. & Wu, S.-H. Processing bodies control the 735 selective translation for optimal development of Arabidopsis young seedlings. 736 _Proceedings of the National Academy of Sciences_ 116, 6451–6456 (2019). 737 51. Xu, J. & Chua, N.-H. Dehydration stress activates Arabidopsis MPK6 to signal 738 DCP1 phosphorylation. _The EMBO Journal_ 31, 1975–1984 (2012). 739 52. Wang, Z. _et al._ A cytoplasmic osmosensing mechanism mediated by molecular 740 crowding–sensitive DCP5. _Science_ 386, eadk9067 (2024). 741 53. Wawer, I. _et al._ mRNA Decapping and 5′-3′ Decay Contribute to the Regulation of 742 ABA Signaling in Arabidopsis thaliana. _Frontiers in Plant Science_ 9, 1–12 (2018). 743 54. Choudhary, M. K., Nomura, Y., Shi, H., Nakagami, H. & Somers, D. E. Circadian 744 profiling of the Arabidopsis proteome using 2D-DIGE. _Frontiers in Plant Science_ 7, 745 (2016). 746 55. Clough, S. J. & Bent, A. F. Floral dip: a simplified method for _Agrobacterium_ ‐ 747 mediated transformation of _Arabidopsis thaliana_ . _The Plant Journal_ 16, 735–743 748 (1998). 749 56. Lamesch, P. _et al._ The Arabidopsis Information Resource (TAIR): improved gene 750 annotation and new tools. _Nucleic Acids Research_ 40, D1202–D1210 (2012). 751 57. Dobin, A. _et al._ STAR: ultrafast universal RNA-seq aligner. _Bioinformatics_ 29, 15– 752 21 (2013). 753 58. Lewinski, M. _et al._ Mapping protein–RNA binding in plants with individual 754 nucleotide-resolution UV cross-linking and immunoprecipitation (plant iCLIP2). _Nature_ 755 _Protocols_ 19, 33–48 (2024). 47 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 756 59. Roehr, J. T., Dieterich, C. & Reinert, K. Flexbar 3.0 – SIMD and multicore 757 parallelization. _Bioinformatics_ 33, 2941–2942 (2017). 758 60. Smith, T., Heger, A. & Sudbery, I. UMI-tools: modeling sequencing errors in 759 Unique Molecular Identifiers to improve quantification accuracy. _Genome Res._ 27, 760 491–499 (2017). 761 61. Krakau, S., Richard, H. & Marsico, A. PureCLIP: Capturing target-specific 762 protein-RNA interaction footprints from single-nucleotide CLIP-seq data. _Genome_ 763 _Biology_ 18, 1–17 (2017). 764 62. Busch, A., Brüggemann, M., Ebersberger, S. & Zarnack, K. iCLIP data analysis: 765 A complete pipeline from sequencing reads to RBP binding sites. _Methods_ 178, 49– 766 62 (2020). 767 63. Munteanu, A., Mukherjee, N. & Ohler, U. SSMART: sequence-structure motif 768 identification for RNA-binding proteins. _Bioinformatics_ 34, 3990–3998 (2018). 769 64. Wu, G., Anafi, R. C., Hughes, M. E., Kornacker, K. & Hogenesch, J. B. 770 MetaCycle: an integrated R package to evaluate periodicity in large scale data. 771 _Bioinformatics_ 32, 3351–3353 (2016). 772 65. Love, M. I., Huber, W. & Anders, S. Moderated estimation of fold change and 773 dispersion for RNA-seq data with DESeq2. _Genome Biology_ 15, 1–21 (2014). 774 48 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 775 **Acknowledgements** 776 Research was supported by the Consejo Nacional de Investigaciones Científicas y 777 Técnicas (CONICET), the Deutscher Akademischer Austauschdienst (DAAD), the 778 American Society of Biochemistry and Molecular Biology (ASBMB), the Agencia Nacional 779 de Promoción de la Investigación, el Desarrollo Tecnológico y la Innovación (Agencia 780 I+D+i), and the Deutsche Forschungsgemeinschaft (DFG, grant STA653/14-1). We 781 acknowledge the Institute of Molecular Biology Genomics Core Facility in Mainz, 782 Germany. 783 784 **Author contributions** 785 Conceptualization: D.A.C., S.P.S., R.S.R., D.S., A.A.B. and M.J.Y. Data curation: D.A.C., 786 D.J.V., V.D.A.B. Formal analysis: D.A.C., S.P.S., D.J.V., V.D.A.B. and M.L. Funding 787 acquisition: D.A.C., S.P.S., D.S., A.A.B. and M.J.Y. Investigation: D.A.C., S.P.S., J.L.M., 788 L.B., H.P., M.S.S., A.S.M. and M.J.d.L. Methodology: L.B., D.J.V., A.T. and M.L. 789 Visualization: D.A.C. and S.P.S. Writing – original draft: D.A.C and M.J.Y. Writing – review 790 & editing: D.A.C., S.P.S., J.L.M., R.S.R., D.S. and M.J.Y. 791 792 **Competing interests** 793 The authors declare no competing interests. 794 49 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 795 **Supplementary Figures** 796 797 798 **Supplementary Figure 1.** mRNA half-life decreases with circadian oscillation amplitude. 799 CCGs were grouped into deciles of relative amplitude in each dataset (decile 1, lowest 800 amplitude; decile 10, highest amplitude). Boxplots show the distribution of mRNA half 801 lives within each decile. The trendline shows a robust linear regression. 802 50 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 803 804 **Supplementary Figure 2. Luciferase oscillation of pCCR2:Luc.** **(a)** Expression of the 805 morning reporter pCCR2:Luc and **(b)** period length estimation. The expression was 806 measured for 7 d in LL after entrainment under LD conditions. WT (Col-0) (n=12), _dcp5-_ 807 _1_ mutant (n= 5). The period of bioluminescence rhythms was estimated by the fast 808 Fourier transform nonlinear least squares (FFT-NLLS) test. Error bars represent the 809 mean with range. ***P < 0.001 (Student’s t-test). 810 811 51 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 812 813 **Supplementary Figure 3.** **mRNA half-life decreases with circadian oscillation** 814 **amplitude in both WT and** _**dcp5**_ **plants.** Clock controlled genes (CCG) were grouped 815 into deciles of relative amplitude of its circadian expression (decile 1, lowest amplitude; 816 decile 10, highest amplitude). Boxplots show the distribution of mRNA half-lives within 817 each decile. The trendline and the slopes are the results of a robust linear regression. 818 Relative amplitude values were taken from Romanowski _et al_ . (2020). 52 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 819 820 821 **Supplementary Figure 4.** **Decay profiles of high amplitude clock-controlled genes** 822 **(HA-CCGs) not stabilized in** _**dcp5**_ **.** **a** HA-CCGs with equal decay rates in WT and _dcp5_ 823 (n = 280). **b** HA-CCGs more stable in WT than in _dcp5_ (n = 26). Heatmaps show transcript 824 abundance at 30, 60 and 90 min after cordycepin treatment, relative to t0. Rows are 825 genes, sorted by WT half-life (shortest at the top) and kept in the same order in both 826 genotypes. 53 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 827 828 **Supplementary Figure 5.** **Peak-expression phases of HA-CCGs bound by DCP5.** 829 Number of DCP5 binding sites at CT12 (left) and CT24 (right) in high-amplitude clock 830 controlled genes (HA-CCGs; relative amplitude > 4), grouped by phase of peak 831 expression (h). Phases and amplitudes are from Romanowski _et al_ . (2020); phase 24 is 832 plotted as 0. 833 54 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 834 835 **Supplementary Figure 6. DCP5 binding sites in the dusk- and dawn-phased clock** 836 **genes TOC1 and LHY.** DCP5 binding sites along TOC1 (AT5G61380) and LHY 837 (AT1G01060) at CT12 and CT24. Gene models (TAIR10) are shown below. The height 838 of each peak represents its PureCLIP score. 839 55 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 840 841 **Supplementary Figure 7.** **Most DCP5 binding sites are present in 3UTR.** **a** Number 842 of DCP5 binding sites in each transcript region (CT12 and CT24 combined). **b** UpSet plot 843 of DCP5 target genes grouped by the combination of regions that contain binding sites. 844 56 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 845 846 **Supplementary Figure 8. Metagene profiles of DCP5 binding across transcript** 847 **regions.** Mean coverage of CT24 binding sites along 5′ UTRs, CDSs and 3′ UTRs, each 848 scaled to 100 bins. 849 57 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 850 851 **Supplementary Figure 9. Distribution of bound and unbound motif occurrences** 852 **across transcript regions.** Fraction of all transcriptome occurrences of the three DCP5 853 motifs (WAWUW, WAMAYW and WNUAMW) found in each transcript region. 854 Occurrences within 25 nt of a DCP5 binding site predicted by SSMART as a target of the 855 same motif are classed as bound. (IUPAC code, W = A/U; M = A/C; Y = C/U). 58 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 856 857 **Supplementary Figure 10. Half-lives of DCP5 target and non-target transcripts in** 858 **wild-type and** _**dcp5**_ **plants.** mRNA half-lives of DCP5 target and non-target transcripts 859 in WT (Col-0) and _dcp5-1_ . 860 59 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 861 862 863 **Supplementary Figure 11. Comparison of CCG identified by Romanowski** _**et al.**_ 864 **(2020) and this study. a** Correlation between relative amplitude in clock-controlled genes 865 (CCG) identified by both Romanowski _et al._ (2020) and by our WT dataset. **b** Relative 866 amplitude of the 8,036 CCGs identified in our WT dataset ranked in descending order. 867 Core oscillator genes are labeled and highlighted in green. The red dashed line indicates 868 the threshold at relative amplitude = 4, used to classify CCGs as high-amplitude (HA 869 CCGs) or low-amplitude (LA-CCGs). 870 60 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 871 872 **Supplementary Figure 12. Comparison of CCG identified in wild-type and** _**dcp5**_ 873 **plants. a** Correlation between relative amplitude in clock-controlled genes (CCG) 874 identified in our WT and _dcp5_ in our datasets. **b** Correlation plot of phase of expression 875 in CCG found in both genotypes. 876 61 [bioRxiv preprint doi: https://doi.org/10.64898/2026.10.30.763074; this version posted November 2, 2026. The copyright holder for this preprint](https://doi.org/10.64898/2026.10.30.763074) (which was not certified by peer review) is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made [available under a CC-BY-NC 4.0 International license.](http://creativecommons.org/licenses/by-nc/4.0/) 877 878 **Supplementary Figure 13. Analysis of differentially expressed genes in** _**dcp5**_ 879 **mutant plants. a-d** Proportion of **(a)** DCP5 targets determined by iCLIP, **(b)** stabilized, 880 destabilized, and equally stable transcripts between WT and _dcp5_ determined in the 881 cordycepin timecourse assay, high amplitude (HA-), low amplitude (LA-) and non clock 882 controlled genes (CCGs) in wild type plants determined by **(c)** Romanowski _et al_ . (2020) 883 and **(d)** this work among different differential expressed genes (DEG) categories. Soft 884 DEGs have a FDR < 0.05 in the likelihood ratio test (LRT), while robust DEGs also have 885 at least one timepoint with |log2FoldChange( _dcp5_ /WT)| > log2(1.5). non-DEGs have and 886 FDR > 0.05 in the LRT. 62