Chromatin retained MUSHER lncRNA integrates ABA and DOG1 signalling pathways to enhance Arabidopsis seeds dormancy

Abstract

Many plant lncRNAs regulate gene expression by binding to chromatin, but how they are retained at the target loci is unclear. We identify a chromatin-localized lncRNA - MUSHER, which activates two parallel regulatory pathways to increase Arabidopsis seed dormancy. MUSHER is upregulated in response to high temperatures, contributing to the induction of secondary dormancy. It promotes DOG1 expression by recruitment of the CPSF complex to enhance the proximal cleavage and polyadenylation at the DOG1 transcript. It also increases ABA sensitivity in seeds by activating PIR1 gene transcription. These genes, located on different chromosomes, are both bound by MUSHER, despite lacking sequence homology. The chromatin association of MUSHER enables the integration of the DOG1- and ABA pathways to adjust seed germination timing. Additionally, MUSHER and other lncRNAs interact with U1 snRNP, which is required for their chromatin localisation, revealing a role for U1 snRNP in plants.

Introduction

Multiple long non-coding RNA (lncRNA) are implicated in gene expression regulation in the nucleus by binding to chromatin¹. Well-known examples of chromatin-bound RNA include XIST spreading on the whole X-chromosome and involved in its inactivation in mammals^2,3, and MALAT1, tightly bound to chromatin and involved in gene expression and nuclear architecture regulation in humans⁴. There are also many reports of plant lncRNA functioning at the chromatin. This includes APOLO, involved in remodelling chromatin structure⁵, SVALKA proposed to act through induction of RNA polymerase II (Pol II) transcriptional collisions⁶ and PUPPIES, a lncRNA with a role in delaying Arabidopsis seed germination by controlling Pol II pausing at DOG1⁷. APOLO was shown to be retained at trans-target loci via the formation of R-loops⁵. Probably the best-described plant lncRNA COOLAIR, involved in flowering time regulation^8,9, has been shown to form a cloud around its locus¹⁰. Many lncRNAs functioning in the nucleus are involved in environmental sensing in plants. COOLAIR regulates vernalisation⁹, SVALKA is implicated in cold acclimation in seedlings⁶, ARTA enhances ABA response¹¹, 1GOD affects drought resistance in mature plants¹², PUPPIES regulates seed response to salt⁷ and HID1 promotes germination¹³. While an RNA structural element strongly influences COOLAIR chromatin association¹⁴, a broader mechanism for lncRNA retention at chromatin remains elusive.

Recently, a U1-based mechanism of human and mouse lncRNA retention at the chromatin was reported¹⁵. This is based on the physical binding of U1 to a large fraction of lncRNAs¹⁵. U1 is a highly conserved, small nuclear RNA (snRNA) that, together with conserved proteins including U1C and U1K70, forms the U1 snRNP¹⁶. U1 snRNP is primarily involved in the first step of splicing by recognising 5’ splice sites¹⁷. However, U1 is also engaged in several other chromatin-related activities: enhancement of Pol II promoters directionality¹⁸; boosting the speed of RNA polymerase II (Pol II)¹⁹; and, conserved in plants, suppression of premature cleavage and polyadenylation in introns, known as telescripting^20,21. Whether U1 snRNP plays a role in chromatin retention in plants is unknown.

Here, we leverage seed biology regulation to explore the functions of lncRNAs. Two key players in seed dormancy are plant hormone abscisic acid (ABA) and a plant-specific protein—DELAY OF GERMINATION 1 (DOG1). Both DOG1 and ABA inhibit seed germination by promoting dormancy, an ability to postpone germination despite favourable external conditions²². This allows seeds to ignore short spells of permissive conditions and align germination with seasonal cues. Seed dormancy is initially established when seeds develop on the mother plant²². Once this dormancy, known as primary, is released, exposure of imbibed seeds to unfavourable conditions can result in the establishment of secondary seed dormancy²³. In many plant species, including Arabidopsis, temperature is a key environmental factor required for the efficient establishment of secondary dormancy²⁴.

DOG1 protein has been suggested to act as part of a thermal-sensing mechanism inducing secondary dormancy²⁵, but the precise mechanism of how temperature is integrated into the DOG1 gene regulation during secondary dormancy establishment is not clear²². DOG1 expression is positively correlated with dormancy strength among Arabidopsis accessions, and it has been identified as a major QTL for seed dormancy and shown to have a significant association in genome-wide studies (GWAS)^26,27. DOG1 transcription is extensively regulated, including by two independent lncRNA. DOG1 expression is suppressed by an antisense transcript 1GOD, and activated in response to salt by PUPPIES, a lncRNA expressed in the DOG1 promoter^7,28. In addition, the selection of the DOG1 proximal polyA site results in the production of shDOG1 (short DOG1), the predominant and functional form of the DOG1 transcript, while the selection of the distal polyA site results in a presumably non-functional and much less abundant lgDOG1 (long DOG1) transcript^29,30,31. Multiple regulators of DOG1 proximal/distal polyA site selection have been described^30,32, but DOG1 alternative polyadenylation has so far not been linked to environmental responses.

DOG1 expression is, to a large extent, unaffected by ABA levels, and dog1 mutations only marginally affect ABA levels in seeds³³. Interestingly, seed sensitivity to ABA can be modified by both environmental factors and mutations in genes involved in the ABA signalling pathway, without affecting basal ABA levels³⁴. ABA is sensed in plants by ABA receptors PYR/PYLs, which once bound by ABA, bind to and inactivate several PP2C phosphatases, including PP2CA³⁵. One way plants can modify ABA sensitivity is by interfering with the ABA signal transduction pathway. For example, PP2CA interacting RING finger protein 1 (PIR1) and PIR2 have been shown to mediate PP2CA degradation to enhance abscisic acid response³⁶. Although dog1 mutants do not exhibit altered ABA sensitivity²⁷, DOG1 protein has been shown to interact with and inhibit the enzymatic activity of PP2CA in vitro^37,38. In summary, DOG1 and ABA pathways are closely intertwined at the protein level, but genetic and transcriptomic analyses indicate that DOG1 function independently from ABA^22,27,33.

Here, we describe MUSHER, a lncRNA that integrates ABA and DOG1 pathways on the transcriptional level by enhancing DOG1 and PIR1 expression. Through genetic analyses, we demonstrate that MUSHER acts as a positive regulator of dormancy. Our findings imply that MUSHER responds to increased temperature and is essential for high-temperature-induced secondary dormancy in Arabidopsis. At the molecular level, MUSHER promotes DOG1 mRNA proximal polyadenylation by recruiting the CPSF complex. Additionally, through RNA Pol II recruitment, MUSHER facilitates the expression of PIR1 in trans, thereby enforcing ABA sensitivity. Finally, we show that MUSHER, as well as COOLAIR and PUPPIES, interact with chromatin through U1 snRNP, suggesting the existence of a conserved mechanism of lncRNA-chromatin retention across eukaryotes.

Results

Chromatin-localised lncRNA, MUSHER, regulates DOG1 and seed dormancy

To identify lncRNA controlling DOG1, we performed RT-qPCR analysis with primers spanning the DOG1 locus in WT and hen2 mutant seeds (Fig. 1a,b). HEN2 is an RNA helicase that interacts with the exosome complex responsible for the nuclear 3’ to 5’ degradation of various ncRNAs³⁹, and its mutation provides a good background to identify unstable lncRNA species^40,41. We found a hen2-dependent signal near the DOG1 gene 3’end, pointing to transcription of a previously uncharacterised transcript targeted by the exosome in seeds (Fig. 1b). We did not detect increased signals in regions corresponding to PUPPIES and 1GOD transcripts^7,28, indicating that these DOG1 locus-originating lncRNAs are not targeted by the exosome in seeds. 5’ and 3’RACE-seq analyses (Fig. 1c, Supplementary Data 1) suggested the existence of an intergenic transcriptional unit between the SnRK3.6 and DOG1 genes, lacking protein-coding potential (Supplementary Data 2). Due to its location downstream of DOG1, we named this putative lncRNA MUSHER.

Fig. 1: MUSHER lncRNA acts as a positive regulator of seed dormancy and DOG1 expression.

a Schematic representation of the DOG1-MUSHER genomic region of chromosome 5 in A. thaliana. The direction of transcription is indicated with arrowheads. Grey rectangles represent coding exons and introns with thick black lines. The green arrow represents the newly annotated MUSHER transcriptional unit. Numbers along the x-axis on the top indicate genome coordinates (TAIR10). Triangles represent T-DNA insertions of two T-DNA lines (msh-1 and msh-2). The green vertical lines show the positions of multiple guides for msh_dCas9-1 and msh_dCas9-2 dCas9 lines. The green horizontal line corresponds to the region deleted using Cas9 (∆msh). b Primer walking RT-qPCR analysis across the DOG1 locus. The position of amplicons corresponds to the DOG1 locus representation above. Expression levels in the hen2 mutant were normalised to WT and UBC21. Data are presented as mean values of four biological replicates +/− SD. c Positions of MUSHER transcript ends found in WT seeds using 3’ and 5’ RACE-seq, coloured in dark and light green, respectively. d Expression profiles of DOG1 and MUSHER transcripts during seed maturation in the WT. Expression levels were normalised to UBC21 mRNA and 16 DAP. The data are shown as the average of four biological replicates for DOG1 and MUSHER 11DAP, five for MUSHER 16DAP, and six for MUSHER 8,14DAP ± SD. e DOG1 relative expression levels, musher mutants relative to WT in freshly harvested seeds, normalised to UBC21. *p < 0.05, two-tailed Student’s t-test. Data are the mean of five biological replicates, with error bars indicating the standard deviation. f Freshly harvested seeds of WT and different musher mutants were scored for germination. Germination was defined as radical protrusion and counted 4 d after sowing. Data are the mean of six biological replicates +/− SD (n = 6, *p < 0.05, two-tailed t-test). g DOG1 expression profile during silique development in WT and msh_dCas9-1 plants. Expression levels were normalised to UBC21 and shown at the indicated number of days after pollination (DAP); Data are presented as mean values of three biological replicates for WT and four for msh_dCas9-1 + /− SD. h Subcellular localisation of selected transcripts in maturing WT Arabidopsis seeds. Each transcript was tested in 5 biological replicates represented as separate bars. Black horizontal lines represent the mean of chromatin and nucleoplasmic fractions contribution of the total amount of indicated transcript.

Measuring this ncRNA level throughout seed development (Fig. 1d) revealed that MUSHER transcripts accumulate during seed maturation concomitant to DOG1 induction. To disrupt the MUSHER transcriptional unit, we used two T-DNA insertion lines, msh-1 and msh-2 (Fig. 1a), a Cas9 deletion mutant Δmsh, and two additional lines: msh_dCas9-1 and msh_dCas9-2 with catalytically inactive Cas9 (dCas9). dCas9 acts as a transcriptional repressor by binding to target DNA without inducing cleavage⁴². T-DNA insertion and dCas9 lines, but not the dCas9 controls targeting two intergenomic regions localised 100 kb from DOG1 locus showed reduced MUSHER levels in seeds (Supplementary Fig. 1a,b). Unexpectedly, Δmsh, despite deletion of mapped MUSHER start site, showed ectopic expression of an antisense transcript downstream of DOG1 (Supplementary Fig. 1c and d). Despite this, all tested musher mutants demonstrated reduced DOG1 expression in seeds and increased seed germination (Fig. 1e,f and Supplementary Fig. 1e). The impact of msh-1 and msh_dCas9-1 on DOG1 expression is observed throughout seed maturation (Fig. 1g and Supplementary Fig. 1f). Because of the observed MUSHER-like transcripts in Δmsh we decided to exclude this line from downstream analysis. We generated two MUSHER-complemented lines: msh-1 pMUSHER_1 and msh-1 pMUSHER_2 to determine whether expression of MUSHER under its endogenous promoter in the msh-1 background rescues its low-dormancy phenotype. Both lines exhibited higher DOG1 expression (Supplementary Fig. 1g) and stronger dormancy (Supplementary Fig. 1h,i) than msh-1 mutants, resembling wild-type (WT) plants, further supporting the role of MUSHER in regulating DOG1-dependent dormancy.

Since MUSHER and DOG1 are co-expressed during seed maturation (Fig. 1d), we questioned whether DOG1 expression can have a feedback effect on MUSHER transcription. We observed reduced MUSHER expression in dog1 insertional mutants, dog1-3 and dog1-4, but no effect in a DOG1 large deletion mutant generated using CRISPR/Cas9, dog1-Δ19 (Supplementary Fig. 1a). This suggested that while MUSHER regulates DOG1 expression, MUSHER is not dependent on DOG1 expression. We speculated that the downregulation of MUSHER expression in dog1 insertional mutants could result from local chromatin state changes caused by T-DNA insertion⁴³. Furthermore, MUSHER expression was unaffected by mutations in the downstream gene SnRK3.6 (Supplementary Fig. 1a), confirming that MUSHER is an independent transcriptional unit.

Next, we tested if MUSHER accumulates in the cytoplasm or chromatin, which could provide insight into its mode of action. To test MUSHER localisation, we used developing seeds, as this is when MUSHER and DOG1 are co-expressed at high levels, and DOG1 exerts its function in seed dormancy establishment. We isolated cytoplasmic, nucleoplasmic and chromatin fractions and performed RT-qPCR analysis. Mitochondrial and chloroplast-localised RNA, used as negative controls, are nearly exclusively found in the cytoplasmic fraction. Several tested nuclear-encoded mRNAs showed similar levels in cytoplasm, nucleoplasm and chromatin. U6, a spliceosome component, was detected mainly in chromatin and nucleoplasmic fractions. MUSHER, CHR1 - a transposon-derived RNA and U3.3, involved in co-transcriptional pre-rRNA modification, were predominantly detected in chromatin fraction (Fig. 1h).

In summary, our data showed that MUSHER is a lncRNA transcribed downstream of the DOG1 gene that increases DOG1 mRNA levels to delay the germination of freshly harvested Arabidopsis seeds. MUSHER’s chromatin association suggests that it may control DOG1 expression by binding to its chromatin.

MUSHER lncRNA is required for DOG1 regulation by CPL1, a Pol II phosphatase

We showed previously that RNA Polymerase II Carboxyl Terminal Domain (CTD) Phosphatase-Like1 (CPL1)-mediated suppression of DOG1 expression (Supplementary Fig. 2a) requires the 3’ end of the DOG1 locus but not the antisense 1GOD lncRNA expression from this region³². Consequently, the molecular mechanism of CPL1 control over DOG1 expression remains incompletely understood. Given the effect of MUSHER on DOG1 expression and dormancy, supported by the MUSHER-complemented lines (Supplementary Fig. 1g,h), we speculated that CPL1 could act through MUSHER regulation. Indeed, multiple cpl1 alleles showed elevated MUSHER transcript level (Supplementary Fig. 2b), suggesting that cpl1 may act as MUSHER lncRNA over-expressor.

To test the requirement of MUSHER for CPL1’s effect on DOG1, we constructed a full-length DOG1-luciferase (LUC) fusion with or without MUSHER (pDOG1::LUC-DOG1³⁰ and pDOG1::LUC-DOG1ΔMUSHER). Transient expression in Arabidopsis seedlings showed that MUSHER deletion suppressed DOG1 upregulation in the cpl1 plants (Supplementary Fig. 2c).

To cross-validate this, we checked the effect of CPL1 on MUSHER locus activity. We generated transgenic plants expressing an internal ribosomal entry site-luciferase gene (IRES-LUC) reporter cassette fused with MUSHER and crossed them to cpl1-7 mutant. The resulting pMUSHER::MUSHER-IRES-LUC (pMUSHER_1) showed stronger activity in cpl1-7 when compared to WT, indicating that MUSHER is a direct target of CPL1 (Supplementary Fig. 2d). Our data showed that MUSHER lncRNA is required for CPL1 to control DOG1 expression.

Interestingly, inactivation of CPL1 led to preferential expression of proximally over distally polyadenylated DOG1 mRNA isoform, known respectively as shDOG1 (short DOG1) and lgDOG1 (long DOG1) (Supplementary Fig. 2e). Conversely, MUSHER inactivation resulted in reduced shDOG1 isoform (Supplementary Fig. 2e). Thus, MUSHER and CPL1 showed antagonistic effects on the production of shDOG1, encoding functional DOG1 protein and complementing the dog1 mutant³⁰. Importantly, mutating the cpl1 gene in the msh background did not reverse lgDOG1 polyadenylation preference (Supplementary Fig. 2e), suggesting DOG1 polyA site selection is directly affected by MUSHER and not by CPL1 protein. Taken together, our data showed that CPL1 indirectly affects DOG1 polyadenylation site selection, by suppressing MUSHER expression and activity.

MUSHER promotes DOG1 proximal polyadenylation by recruiting the CPSF complex

We used 3’RNA-seq (Supplementary Fig. 2f) to confirm that shDOG1 is decreased in freshly harvested msh-1 seeds, while lgDOG1 remains unchanged when compared to WT (Fig. 2a). Calculation of 3’RNA-seq read density assigned to DOG1 isoforms revealed a significant reduction of shDOG1/lgDOG1 ratio in msh-1 (Fig. 2b). Subsequent RT-qPCR analysis confirmed that shDOG1/lgDOG1 ratio is significantly reduced in all tested musher mutants (Fig. 2c).

Fig. 2: MUSHER enhances DOG1 proximal polyA site selection.

a Snapshot of 3′RNA-seq reads for DOG1 locus in WT and msh-1 mutant (n = 4). Schematic diagram showing DOG1 gene: exons (grey rectangles), shDOG1 3’UTR region (light grey arrow), lgDOG1 3’UTR region (grey arrow), and MUSHER (green). The regions used to estimate the expression of mRNA isoforms are shown below chromosome coordinates. b Ratio of reads corresponding to short and long DOG1 mRNA isoforms calculated for regions marked on the schematic. Data are the mean of four biological replicates; error bars show standard deviations. c The ratio of shDOG1 and lgDOG1 isoforms in freshly harvested seeds of WT and musher mutants, normalised to UBC21 and relative to WT. Data are presented as mean values of four biological replicates +/− SD (*p < 0.05, two-sided t-test). d CPSF73 occupancy was measured by ChIP-qPCR in WT, msh-1 and msh_dCas9-1 mutants. Data are presented as mean values of three biological replicates +/− SD (*p < 0.05, two-sided t-test). e The ratio of shDOG1 and lgDOG1 isoforms in seeds of WT and msh-1, esp1-2 and double msh-1 esp1-2 mutants, normalised to UBC21 and relative to WT (mean of five biological replicates +/− SD, *p < 0.05, two-sided t-test). The ratio of shDOG1 and lgDOG1 isoforms in seeds of WT and msh-1, esp1-2 and double msh-1 esp1-2 mutants, normalised to UBC21 and relative to WT (mean of 5 biological replicates, *p < 0.05, two-sided t-test). f MUSHER relative expression level in seeds of WT and esp1-2, normalised to UBC21 and relative to WT (mean of five biological replicates +/-SD, ns – not significant, two-sided t-test). g Pol II occupancy was measured by ChIP-qPCR in WT and msh-1 mutant. Data are presented as mean values of three biological replicates +/− SD (*p < 0.05, two-sided t-test). h H3, i H3K4me3 and j H3K27me3 levels were measured by ChIP-qPCR in WT and msh-1 mutant. H3K4me3 and H3K27me3 values were normalised to H3 and to ACTIN7 (AT5G09810) gene body. For all ChIP-qPCR experiments presented in this figure: p-value from the two-tailed t-test, *p < 0.05, ns - not significant. Data are the mean of four biological replicates, with error bars indicating the standard deviation. H3, H3K4me3 and H3K27me3 (h–j) ChIP were done side by side and share NoAB control shown on panel h.

We previously reported that mutations in pre-mRNA cleavage and polyadenylation factors result in reduced shDOG1 but not lgDOG1 expression³⁰, similar to MUSHER knockouts shown here. To confirm changes in DOG1 mRNA 3’end formation, we performed ChIP-qPCR analysis of the CPSF73, a subunit of the Cleavage Stimulatory Factor complex⁴⁴. We observed a localised decrease in CPSF73 occupancy at the DOG1 proximal but not distal termination site in msh-1 and msh_dCas9-1 mutants (Fig. 2d). Additionally, we performed a genetic analysis of MUSHER’s interaction with 3’end formation machinery, by crossing MUSHER mutants to esp-1. ESP encodes an auxiliary component of Cleavage Stimulatory Factor complex (CSTF64L). We used esp-1 and not cpsf73 mutants, because the latter do not produce seeds⁴⁵. Analysis of the msh-1 esp1-2 double mutant showed no additivity in DOG1 termination site selection defects suggesting that MUSHER controls DOG1 polyA site selection through CPSF complex (Fig. 2e). This was further supported by no change of MUSHER expression in the esp1 mutant (Fig. 2f). Our results implied that MUSHER increases DOG1 expression by enhancing CPSF complex recruitment to the proximal termination site. This is in agreement with the findings in cpl1-7 mutant, where MUSHER overexpression was accompanied by enhanced DOG1 proximal polyA site selection (Supplementary Fig. 2). Transcription termination can be coupled with Pol II occupancy changes⁴⁶. We, therefore, profiled Pol II occupancy along the DOG1 locus in the musher mutant. This revealed a significant decrease of Pol II level at the DOG1 proximal termination region but not at the DOG1 transcription start site (TSS) or the distal polyA site (Fig. 2g). This suggested that MUSHER-promoted proximal polyA site selection is not accompanied by changes in DOG1 transcription initiation but rather represents a result of MUSHER direct activity at shDOG1 gene 3’end.

The localised decrease of Pol II and CPSF complexes occupancy at the DOG1 proximal termination site indicates that MUSHER promotes the proximal termination site selection rather than controls transcription rates. To test this hypothesis, we took an orthogonal approach and analysed the intensity of DOG1 transcription using single-molecule RNA FISH (smFISH) in developing embryos⁷. We measured the brightness of nuclear foci identified as transcription sites, as this is generally considered to correlate with the number of RNA molecules present at transcription sites. We also counted the proportion of cells containing transcription sites, which reflects the frequency of promoter bursting⁷. We collected embryos from WT and msh-1 siliques at ten time points, corresponding to the experiments showing DOG1 and MUSHER expression during seed maturation (Fig. 1d). DOG1 smFISH showed a consistent and significant reduction in the number of both nuclear and cytoplasmic foci during seed development in msh-1 seeds when compared to WT (Supplementary Figs. 3a,b). We observed an increase in DOG1 transcript number in WT siliques between the 16^th and 21^st day, which was diminished in msh-1 (Supplementary Figs. 3a,b). Notably, the msh mutant showed no significant change in the intensity of nuclear dots identified as transcription sites (Supplementary Fig. 3c)⁷ and only minor changes in the frequency of cells with active DOG1 transcription (Supplementary Fig. 3d). We therefore conclude that, in agreement with Pol II ChIP (Fig. 2g), smFISH indicated no major changes in DOG1 transcription initiation in msh-1 seeds.

Our data suggest that MUSHER promotes DOG1 proximal termination by locally affecting cleavage and polyadenylation complexes and Pol II occupancy at the proximal transcription termination site without affecting DOG1 transcription levels. All modes of RNA Pol II activity, including transcription termination, are accompanied and modulated by histone posttranslational modifications. Some lncRNAs may recruit specific histone-modelling complexes, including PRC2 involved in H3K27me3 deposition^47,48. To address the effects of MUSHER on the DOG1 chromatin landscape, we analysed selected active (H3K4me3) or repressive (H3K27me3) chromatin histone marks along the DOG1 locus. Using ChIP-qPCR, we detected no significant difference in both tested histone modification levels in the msh-1 mutant throughout the DOG1 locus (Fig. 2h–j) compared to WT. This result showed that defects at the proximal transcription termination site (TTS) are unrelated to chromatin modifications and agree with no changes in Pol II recruitment for most of the DOG1 locus. In summary, we propose a model in which MUSHER lncRNA specifically enhances DOG1 proximal termination rather than Pol II recruitment.

MUSHER enhances secondary dormancy in response to high temperature

DOG1 is a known regulator of both primary and secondary dormancy^49,50. Based on the seed’s requirement for MUSHER to express DOG1 at high levels and fully develop primary dormancy (Fig. 1), we hypothesised that MUSHER may also be involved in secondary dormancy regulation. We used a well-established secondary dormancy induction protocol that consists of prolonged incubation in the darkness at high temperatures followed by germination analysis in permissive conditions⁴⁹. In agreement with defects in primary dormancy, msh-1 and msh_dCas9-1 showed weaker secondary seed dormancy, and cpl1-7, a mutant with high MUSHER expression, showed stronger secondary dormancy when compared to WT (Fig. 3a,b, Supplementary Fig. 4a,b). Consistent with these phenotypes, we observed the down- and up-regulation of the shDOG1 transcript in musher and cpl1-7 mutants, respectively (Fig. 3c,d).

Fig. 3: MUSHER activates DOG1 expression during secondary dormancy induction in response to high temperature.

a Seed germination of WT and different musher mutants after 7 days of secondary dormancy (SD) induction at 35 °C. Germination frequency was scored after 5 days. Data are presented as mean values of four biological replicates +/− SD. *p < 0.05 (t-test two-sided). b Seed germination of WT and cpl1-7 mutant seeds after 7 days of SD induction. Germination was scored after 5 days. Data are presented as mean values of five biological replicates +/− SD. *p < 0.05 (t-test two-sided). c The expression profile of DOG1 was analysed using RT-qPCR in seeds of WT, msh-1 and msh_dCas9-1 subjected to secondary dormancy induction. Data are presented as mean values of four biological replicates +/− SD. UBC21 mRNA was used as a reference gene. d The DOG1 relative expression level in seeds of WT and cpl1-7 mutant, subjected to SD induction. Data normalised to UBC21 and related to WT. Data are presented as mean values of four biological replicates +/− SD. p-value from the two-tailed t-test, *p < 0.05. e DOG1 and f MUSHER expression profile in after-ripened seeds subjected to secondary dormancy induction at 25 °C and 35 °C. Expression was normalised to UBC21 mRNA and dry seed levels. Data are presented as mean values of four biological replicates +/− SD (*p < 0.05, two-sided t-test). g Schematic representation of constructs introduced to transgenic plants. RT-qPCR analysis of luciferase mRNA levels in seeds subjected to secondary dormancy induction under temperatures indicated on the right. LUC expression level measured after 5 days of SD induction, normalised to UBC21 and related to 25 °C for each construct, Data are presented as mean values of four biological replicates +/− SD (*p < 0.05, two-sided t-test). h Long and short DOG1 expression analysis on the 5th day of secondary dormancy induction in WT at temperatures indicated. Data are presented as mean values of four biological replicates +/- SD (p-value from the two-tailed t-test, *p < 0.05). UBC21 mRNA was used as a reference gene.

Elevated temperature is a key environmental factor responsible for secondary dormancy establishment⁵¹. Consistent with previous studies^7,49,52,53, we showed that DOG1 expression increases in response to high temperatures during secondary dormancy establishment (Fig. 3e). At 35 °C, DOG1 expression dropped initially during imbibition but rose to levels observed in dry seeds on the 2^nd day of secondary dormancy induction and remained constant for the rest of the experiment (Fig. 3e). Likewise, high temperatures augmented MUSHER induction in imbibed seeds (Fig. 3f). MUSHER levels were similarly reduced by imbibition and recovered at 2^nd day, but, in contrast to DOG1, continued to rise during the entire experiment length without reaching the plateau. This suggests that MUSHER is more responsive to secondary dormancy-inducing conditions than DOG1 (Fig. 3f). Based on this observation, we speculated that MUSHER could be responsible for temperature sensing and the incorporation of this signal into DOG1 regulation.

Next, to analyse the temperature responsiveness of MUSHER during secondary dormancy induction, we took advantage of luciferase reporter lines (Supplementary Fig. 2c,d). We induced secondary dormancy at different temperatures (25, 28, and 35 °C) and measured the LUC activity for DOG1-LUC, MUSHER-LUC and the DOG1trunc-LUC, which is devoid of MUSHER region. We noted that MUSHER and DOG1 reporters showed similar gradual increases in expression in response to increasing temperatures (Fig. 3g). Interestingly, DOG1 reporter devoid of MUSHER region (pDOG1::DOG1-LUC truncated) showed no induction at 35 °C (Fig. 3g), indicating MUSHER is required for the high-temperature-driven DOG1 upregulation. The responsiveness of MUSHER to high temperatures, along with MUSHER requirement for high temperature-driven DOG1 upregulation, indicated that MUSHER conveys temperature signals during the induction of secondary dormancy. In agreement with these findings, we show the endogenous shDOG1 expression in wild-type seeds during secondary dormancy establishment is highly induced at 35 °C compared to 28 °C (Fig. 3h), with lgDOG1 transcript being only slightly affected. This is consistent with the concept of MUSHER acting through the polyA-site selection (Fig. 2).

In summary, our data are consistent with a model in which MUSHER lncRNA is induced during primary and secondary dormancy and activates DOG1 expression by promoting its proximal polyA site selection both during primary and secondary dormancy. In addition, we showed that during secondary dormancy induction, MUSHER lncRNA is required for DOG1 upregulation in response to high temperature.

MUSHER enhances the PIR1 gene expression independently of its role in DOG1 regulation

3’RNA-seq analysis revealed 166 differentially expressed genes, including DOG1, between WT and msh-1 freshly harvested seeds (Fig. 4a, Supplementary Data 3). Among the downregulated genes, PIR1 (PP2CA INTERACTING RING FINGER PROTEIN 1) showed the strongest decrease (Fig. 4a). PIR1 encodes an E3 ligase that enhances the abscisic acid (ABA) response and is highly expressed in maturing siliques³⁶. Inspection of 3’RNA-seq coverage plots showed a strong reduction at both PIR1 and an upstream localised AT2G35320 gene (Fig. 4b). RT-qPCR confirmed the strong downregulation of PIR1 expression in musher mutant, when tested in dry seeds (Fig. 4c), as well as in secondary dormancy-induced seeds (Fig. 4d), suggesting MUSHER activates PIR1 expression. Consistently, both MUSHER-complemented lines, msh-1 pMUSHER_1 and msh-1 pMUSHER_2, showed significantly higher expression of PIR1 than msh-1 (Supplementary Fig. 5c). Whereas cpl1-7 showed elevated PIR1 levels, along with MUSHER overexpression (Fig. 4d). Interestingly, PIR1 expression remained unchanged in the dog1 mutant. This indicates that PIR1 is activated by MUSHER independently from DOG1 (Fig. 4c,d, Supplementary Fig. 5a,b). ChIP-qPCR analysis revealed a significant decrease in Pol II occupancy throughout the entire PIR1 locus in msh, suggesting that in contrast to DOG1, MUSHER regulates PIR1 transcription levels rather than termination of PIR1 transcripts (Fig. 4e). ChIP-qPCR analysis along PIR1 locus showed a TSS-localised decrease in H3K4me3, slight decrease in H3 and no change in H3K27me3 levels in msh-1 (Fig. 4f–h), confirming the lower transcriptional activity of PIR1 promoter.

Fig. 4: MUSHER enhances PIR1 expression independently of its role in DOG1 regulation.

a Volcano plot showing differentially expressed genes for WT and msh-1 selected using DESeq2; absolute log2FC >  log2 (1.5), FDR  <  0.05, n = 4. DOG1 and PIR1 genes are highlighted. The number of downregulated (down) and upregulated (up) genes is provided on the plot. b Diagram of 3′RNA-seq reads for PIR1 locus in WT and msh-1 mutant. Schematic diagram showing PIR1 locus: exons (grey rectangles), 3’UTR region (grey arrow) c The PIR1 relative expression level in freshly harvested seeds (primary dormancy) of WT, msh-1 and dog1-4 mutants, normalised to UBC21 and relative to WT. Data are presented as mean values of four biological replicates +/− SD (p-value from the two-tailed t-test, *p < 0.05). *p < 0.05. d The PIR1 relative expression level in WT msh-1, dog1-4 and cpl1-7 mutant seeds, after 5 days of SD induction, normalised to UBC21 and relative to WT. Data are presented as mean values of four biological replicates +/− SD (p-value from the two-tailed t-test, *p < 0.05). e Pol II f H3 g H3K4me3 and h H3K27me3 occupancy was measured by ChIP-qPCR in WT, msh-1. (n = 4) H3K4me3 and H3K27me3 values were normalised to H3 and to ACTIN7 (AT5G09810); Pol II and H3 were normalised to ACTIN7 (AT5G09810). For all ChIP experiments *P < 0.05, ns - not significant (two-tailed t-test). Error bars represent the standard deviation of four biological replicates. H3, H3K4me3 and H3K27me3 (f–h) ChIP were done side by side and share NoAB control shown on panel h.

Our results showed that MUSHER activates PIR1 expression independently of its role in DOG1 gene regulation. In addition, our data suggest that MUSHER regulates transcriptional activity of the PIR1 promoter, whereas at DOG1, it specifically contributes to proximal termination site selection. Importantly, both DOG1 and PIR1 regulation by MUSHER is compatible with its chromatin association (Fig. 1).

MUSHER lncRNA activates PIR1 expression through PIR1-DOG1 chromatin interaction

We showed that MUSHER is chromatin-bound (Fig. 1h). To identify MUSHER binding sites, we conducted a Chromatin Isolation by MUSHER-RNA Purification followed by DNA-seq (ChIRP-seq) in seeds⁵⁴. This revealed 23 DNA regions bound by MUSHER lncRNA (Supplementary Data 4). The two most significantly enriched are PIR1 and DOG1 loci. Quantifying the ChIRP signal at DOG1 and PIR1 loci shows that target enrichment is greatly diminished in the msh-1 mutant (Fig. 5a,b). In addition, ChIRP using LacZ probes, a commonly used negative control, also failed to recover significant signals at both targets (Fig. 5a,b). Analysis of the ChIRP-seq data revealed that two replicates exhibited high correlation, confirming the reproducibility of the assay (Supplementary Fig. 6a,b). In agreement with our conclusion that MUSHER regulates PIR1 transcription initiation, MUSHER peaks centred around PIR1 TSS. On the DOG1 locus, MUSHER ChIRP-seq detected two strong peaks, one at a site where it is produced and one colocalising with the DOG1 exon2-intron2 region, just upstream of the proximal polyA site, the selection of which is promoted by MUSHER (Fig. 5a).

Fig. 5: MUSHER lncRNA integrates ABA and DOG1 pathways by interacting with PIR1 TSS and DOG1 proximal termination site.

Snapshot of ChIRP-seq in WT and msh-1 mutant using MUSHER-specific and LacZ-specific probes on a DOG1 and b PIR1 loci. Schematic diagram showing DOG1 and PIR1 locus: exons (grey rectangles), shDOG1 3’UTR region (light grey arrow), lgDOG1 3’UTR region (grey arrow), and MUSHER (green). The lines below the scheme mark regions used to design MUSHER probes. Data are the mean of two biological replicates. c Snapshot of ChIDP-seq in WT using PIR1-specific probes on DOG1 and d PIR1 loci. Data are the mean of two biological replicates. The lines below the scheme mark regions used to design PIR1 probes. e The germination rate of WT, dog1-3, at2g35320-1, pir1 and msh-1 mutant seeds on MS media supplemented with various ABA concentrations. Data are presented as mean values of four biological replicates for seeds germinated on ABA + /− SD and 2 biological replicates for mock. Asterix denotes a two-tailed Welch’s t-test comparing the wild type (WT) to each mutant yielded a p < 0.05.

Our data showed that the DOG1 proximal polyA site and PIR1 TSS are bound by MUSHER lncRNA. Together with the lack of PIR1 expression changes in the dog1 mutant (Fig. 4c and Supplementary Fig. 5b), this suggested that both DOG1 and PIR1 are direct and independent MUSHER targets. Of note, we could not find a common DNA motif or R-loop enrichment among MUSHER targets, indicating that MUSHER may not rely on a sequence-based mechanism for target specification. Whereas the DOG1 locus is localised next to MUSHER on chromosome 5, the PIR1 locus is localised on chromosome 2. This encouraged us to test the physical proximity of PIR1 and DOG1 locus chromatin.

To this end, we used PIR1 probes to fish out crosslinked DNA sequences followed by NGS. This assay is based on a similar principle as PiCh (Proteomics of isolated Chromatin segments)⁵⁵ but analyses DNA rather than proteins attached to a specific locus. This method was called chromatin-DNA precipitation assay (ChIDP). The use of glutaraldehyde combined with the covalent link between probes and beads allowed us to use more stringent conditions, resulting in a high signal-to-noise ratio (Fig. 5c,d, Supplementary Fig. 6a,c Supplementary Data 5). Examination of the ChIDP-seq results showed that two replicates were highly similar, verifying the assay’s consistency (Supplementary Fig. 6a). Surprisingly, we discovered that PIR1 DNA interacts with the proximal polyA site of DOG1 but not with MUSHER (Fig. 5c,d, Supplementary Fig. 6a,d). This suggests that MUSHER activity on PIR1 is mediated by long-range interaction established by the proximal polyA site of DOG1, which is occupied by MUSHER.

DOG1 and ABA are the two main regulators of dormancy in Arabidopsis⁵⁶. We confirmed the published results that pir1 mutants show decreased sensitivity to ABA-mediated germination inhibition (Fig. 5e)³⁶. Notably, a mutant in a gene directly upstream of PIR1 (at2g35320-1) showed no significant changes in ABA sensitivity compared to WT. In agreement with published results, dog1 mutant seed germination was suppressed by ABA in the same way as WT²⁷. Importantly, multiple musher alleles showed much weaker sensitivity to ABA-mediated inhibition of germination across all tested ABA concentrations, similar to the pir1 mutant (Fig. 5e, Supplementary Fig. 6e-g). Therefore, in contrast to the DOG1 protein, MUSHER lncRNA does modulate ABA responsiveness. Our data are consistent with a model where MUSHER lncRNA enhances the ABA sensitivity by activating PIR1 gene expression. MUSHER’s ability to activate DOG1 and PIR1 points to its role in integrating the DOG1 and ABA-dependent repression of seed germination.

U1 snRNP mediates MUSHER’s association to chromatin

MUSHER physically binds to and regulates DOG1 and PIR1 gene expression (Figs. 1,2,5). Since most MUSHER transcripts are localised at the chromatin, it suggests that an active mechanism exists that sequesters MUSHER at the chromatin (Fig. 1h).

In addition to its well-known splicing function, U1 snRNP has been shown in mammals to bind various lncRNAs, facilitating their chromatin tethering¹⁵. Interestingly, we identified four putative U1 binding sites within the MUSHER sequence, even though we failed to detect MUSHER splicing (Supplementary Data 1,6 and Fig. 1c). Therefore, we analysed whether MUSHER interacts with U1 snRNP. Using probes specific for MUSHER, but not LacZ probes, we selectively recovered U1, but not U4 or U6 RNA using formaldehyde-assisted RNA ImmunoPrecipitation (RIP) (Fig. 6a). Supporting the specificity of the approach, U1 recovery was strongly diminished when we repeated the experiment using MUSHER probes in msh-1 lines (Fig. 6b). Next, we validated the MUSHER-U1 association by RNA Immunoprecipitation (RIP) using the U1 snRNP component U1-70K. We used a recently reported u1-70k complementing line expressing GFP tagged U1-70K driven by endogenous U1-70K promoter (pU1-70K::U1-70K-GFPu1-70k)²¹. RT-qPCR showed that U1-70K immunoprecipitation recovered spliced control RNA, failed to recover unspliced control mRNA, and, importantly, retrieved a significant amount of MUSHER lncRNA (Fig. 6c).

Fig. 6: U1 is required for chromatin tethering of MUSHER and other lncRNA.

a MUSHER or LacZ probes were used in formaldehyde-assisted RIP to enrich associated RNA. U1, U4 and U6 recovery was analysed using RT-qPCR. Data are the mean of four biological replicates +/− SD. b MUSHER probes were used to enrich MUSHER from WT and as a negative control in msh-1. U1 and U6 recovery was analysed using RT-qPCR. The data represent the average of four biological replicates +/− SD. c U1A protein was precipitated side by side with a no-antibodies (NoAB) control from plants expressing pU1-70K::U1-70K-GFP u1-70k, and co-precipitated RNA was analysed using RT-qPCR for spliced (AT5G45830) and unspliced (AT5G45820) RNA controls and MUSHER. The data are averaged from three biological replicates + /− SD. A two-tailed Welch’s t-test p-value was calculated compared to WT (*p < 0.05). d Subcellular localisation of mRNAs and lncRNAs in maturing Arabidopsis seeds of WT and u1c amiRNA line. Lines represent the mean of chromatin and nucleoplasmic fractions contribution of the total amount of indicated transcript. The blue-shaded colour refers to Col-0, while the grey-shaded colour corresponds to u1c. Each transcript was measured in 4 replicates represented as a separate bar. *p  <  0.01 from two-tailed Student t-test. e After ripened seeds of WT (WT), u1c amiRNA and u1k-70k mutants were induced into secondary dormancy, moved to permissive conditions and scored for germination ability. Germination (%) was scored as radical protrusion. Data are the mean of four biological replicates +/− SD (two-tailed t-test, *p < 0.05). f shDOG1 expression was analysed in WT, u1c amiRNA and u1-70k mutants seeds induced into secondary dormancy. Data are the mean of four biological replicates +/− SD. A two-tailed t-test p was calculated compared to WT (*p < 0.05). Normalised to UBC21 and relative to WT.

Having shown that MUSHER is bound by U1 but not U6 or U4 snRNP, we asked if this contributes to MUSHER chromatin association. To do this, we used a recently described artificial microRNA (amiRNA) line that shows significantly downregulated u1c levels²¹. Subcellular localisation showed that UBC21 and DOG1 mRNA are nearly equally distributed between cytoplasmic, nucleoplasm and chromatin-bound fractions, as shown also in Fig. 1 (Fig. 6d). In addition, neither UBC21 nor DOG1 mRNA distribution among the cellular fractions was altered in the u1c mutant (Fig. 6d). Importantly, MUSHER’s association to chromatin was strongly reduced in u1c mutant, suggesting U1 tethers MUSHER lncRNA at the chromatin. Next, we asked whether the U1 requirement for RNA retention at chromatin is a more general phenomenon in plants. We started by confirming that COOLAIR, shown to be expressed in seeds⁵⁷, is predominantly chromatin-bound, like MUSHER (Fig. 6d). In addition, unspliced PUPPIES isoform, a lncRNA involved in Pol II pausing regulation at DOG1 in seeds⁷, was also strongly enriched in chromatin fraction. Both PUPPIES and COOLAIR have multiple U1 binding sites (Supplementary Data 6) and we found that in contrast to the control mRNA, MUSHER and all other analysed lncRNAs showed reduced chromatin association and a corresponding increase in nucleoplasmic abundancy in the u1c mutant (Fig. 6d). We conclude that MUSHER binds U1 snRNP and that U1 facilitates MUSHER and other tested lncRNA-chromatin association and propose that the U1 role in chromatin retention of lncRNA is conserved between animals and plants.

Finally, we asked if MUSHER’s chromatin association is required for its function. Both u1c and another amiRNA mutant in the U1 snRNP component U1-70K, show pleiotropic phenotypes, including significant dwarfism and delayed development²¹. Thus, it would be difficult to distinguish between primary dormancy phenotype and developmental defects resulting in delayed germination. To circumvent this, we took advantage of the fact MUSHER also shows defects during secondary dormancy induction. To synchronise the physiological state of seed, they were matured on mother plants and stored until they lost primary dormancy and then tested for the induced secondary dormancy. u1c and u1-70k mutant seeds showed low secondary dormancy phenotype with nearly all seeds germinating compared to less than 50% of germinated seeds in the WT (Fig. 6e, Supplementary Fig. 7). The reduction in u1c and u1-70k secondary dormancy phenotype was accompanied by decreased shDOG1 expression (Fig. 6f). This agrees with a model where U1 is required for MUSHER’s association to chromatin and its function in seed dormancy induction. In summary, U1 is required to tether multiple lncRNAs at the chromatin, and in case of MUSHER, we showed this chromatin association is required for the lncRNA function.

Discussion

Our results indicate that MUSHER modulates seed dormancy through the transcriptional regulation of DOG1 and ABA sensitivity pathways (Fig. 7a). We show that MUSHER lncRNA is chromatin-bound and requires U1 snRNP for its retention and activity at chromatin (Fig. 7b).

Fig. 7: MUSHER regulates seed dormancy and lncRNA chromatin retention in Arabidopsis.

a MUSHER modulates seed dormancy through the transcriptional regulation of DOG1 and ABA pathways. Created in BioRender Halale manjunath, V. (2025) https://BioRender.com/djj6tmh. b U1 is required for chromatin retention of lncRNA, including MUSHER in Arabidopsis. Created in BioRender. Halale manjunath, V. (2025) https://BioRender.com/zss2438.

MUSHER-mediated regulation of DOG1 and PIR1 expression

Our analysis showed that MUSHER employs two different mechanisms to activate DOG1 and PIR1. MUSHER promotes DOG1 expression by enhancing CPSF complex recruitment to the proximal cleavage and polyadenylation site, resulting in enhanced production of a functional short DOG1 alternative polyadenylation isoform (Fig. 2). In agreement, the MUSHER effect on DOG1 is not accompanied by changes in Pol II recruitment or H3K4me3 levels at the DOG1 gene (Fig. 2 and Supplementary Fig. 3). In contrast, MUSHER activates PIR1 expression by enhancing Pol II recruitment and consequent H3K4me3 deposition at the PIR1 promoter (Fig. 4). Similar to HID1 lncRNA¹³, and differently then COOLAIR⁵⁸ and APOLO⁵, MUSHER promotes H3K4me3 deposition without affecting H3K27me3 repressive marks. Notably, HID1 represses transcription by blocking the binding of ARABIDOPSIS TRITHORAX-RELATED7 (ATXR7), an H3K4me3 methyltransferase, to nucleosomes, leading to increased H3K4me3 levels when HID1 is inactivated¹³. In contrast, MUSHER inactivation reduces H3K4me3 deposition at the PIR1 promoter, suggesting that MUSHER and HID1 regulate H3K4me3 through distinct molecular mechanisms. Of note, we did not detect PIR1 proximal polyadenylation either in our seed-specific 3’RNA-seq data (Fig. 4b) or in the published long read-based transcript mapping dataset³⁹. It is currently not clear why MUSHER-dependent DOG1 and PIR1 regulation are based on different modes of activity. It is possible that MUSHER acts in the same way on DOG1 and PIR1, but the local chromatin environment of the target or MUSHER site of binding with respect to the target transcription start site results in seemingly different mechanisms. For example, the recruitment of CPSF complex to the gene promoter may enhance transcription initiation or transition to efficient elongation. We, however, favour a model where MUSHER employs two different mechanisms to enhance DOG1 and PIR1 expression. This view is supported by the fact that DOG1 expression is reduced roughly by 50%, while PIR1 expression is undetectable in multiple msh mutants tested.

Chromatin precipitation using MUSHER RNA (ChIRP) followed by sequencing revealed only a very limited number of MUSHER targets, with DOG1 and PIR1 being the most significantly enriched. The high selectivity of MUSHER is somehow surprising, as many of the published plant lncRNAs show binding to multiple sites spread across the genome^5,59. PIR1 DNA chromatin-DNA precipitation assay (ChIDP) followed by sequencing showed that PIR1 DNA also has a very limited number of DNA targets, with, excluding PIR1 bait, DOG1 being the strongest. This suggests that high MUSHER selectivity could be guided by a spatial arrangement of DNA in seeds.

MUSHER as a Link Between ABA Sensitivity and DOG1

Interestingly, both MUSHER targets encode proteins, PIR1 and DOG1, that interact with PP2CA protein, a regulator of ABA signalling³⁵. PIR1 binding to PP2CA promotes its degradation, while DOG1 inhibits PP2CA phosphatase activity^35,37. Both dog1 and ABA synthesis deficient nced2/5/6/9 mutants fail to enter dormancy. Simultaneously, dog1 does not show major changes in ABA levels or sensitivity^27,60 and nced2/5/6/9 shows unaffected DOG1 induction kinetics during secondary dormancy establishment⁵³. This suggests that DOG1 and ABA synthesis pathways function independently in dormancy regulation. We demonstrate that MUSHER inactivation leads to decreased transcription of DOG1 and PIR1, which results in decreased ABA sensitivity. We propose that MUSHER lncRNA facilitates the concurrent activation of DOG1 and ABA pathways, integrating them into a shared regulatory network.

High temperatures are known to enhance DOG1 expression and ABA biosynthesis in seeds^51,61. We observed altered ABA sensitivity and reduced secondary dormancy induction in response to high temperature in msh mutants. Elevated temperatures contribute to two well-described responses: (1) thermoinhibition, a high-temperature-mediated inhibition of germination in light⁶² and (2) secondary dormancy induction in non-dormant imbibed seeds exposed to high temperatures in the darkness^31,49,51. We hypothesise that MUSHER-mediated integration of ABA and DOG1 pathways contributes not only to dormancy but also to other ABA/DOG1-related phenotypes (e.g., drought tolerance or thermoinhibition). We conclude that, similar to other long non-coding RNAs involved in temperature sensing, such as COOLAIR⁸, COLDWRAP ⁶³ and SVALKA⁶ in plants, and Hsrω⁶⁴ and Heat ⁶⁵ in animals, MUSHER also plays a role in temperature sensing in plants.

U1 role in chromatin retention of lncRNA in Arabidopsis

In contrast to mature mRNAs, lncRNAs are often enriched in the nucleus^1,15. Consistently, many lncRNA have been shown to regulate gene expression by binding to chromatin^1,5,6,39. The molecular mechanism of human lncRNA retention in the nucleus has only recently started to be explored^15,66,67, and remains largely unknown in plants. Mechanisms ensuring human lncRNA nuclear retention encompass the presence of specific sequences⁶⁸, including Alu transposable elements binding the HNRNPK complex⁶⁹. These protein-RNA interactions may contribute to liquid-liquid phase separation (LLPS) that can also be responsible for the nuclear retention of lncRNA. For example, LLPS by NEAT1 lncRNA results in the formation of nuclear paraspeckles⁷⁰ and XIST lncRNA was reported to undergo LLPS with different partners localised at inactive X chromosome⁷¹. However, nuclear retention may not be enough to guide lncRNA directly to chromatin. This has been proposed to be a result of direct RNA-DNA interaction, either in the form of R-loops or RNA:DNA triplex¹⁵. In addition, U1 snRNP, involved in 5’ splice site recognition, has recently been shown to be responsible for the chromatin binding of a large fraction of lncRNAs¹⁵. In contrast to Alu and other sequence elements identified in humans, U1 binding sites are well conserved in plant genomes.

Plant lncRNA retention at chromatin has not been explored, but RNA-seq analysis indicates that lncRNAs are enriched in the nucleus⁷². Also, like their human counterparts, plant lncRNAs often regulate gene expression by chromatin binding⁶, indicating that their chromatin retention may be actively maintained. Multiple plant lncRNAs have been described, but few of them have been reported to be expressed in seeds and to control gene expression by chromatin-related mechanisms. This includes COOLAIR regulating the FLC gene chromatin state^9,73, PUPPIES regulating Pol II pausing at the DOG1 gene⁷ and, described here, MUSHER promoting the DOG1 gene proximal cleavage and polyadenylation site selection. In seedlings, COOLAIR is retained at chromatin, presumably by its ability to form R-loops⁷⁴ and a cloud around its transcription site¹⁰. Similarly, APOLO lncRNA has been shown to form R-loops at its targets located in trans to its transcription site⁵. Here, we propose that U1-mediated chromatin binding of lncRNAs is a largely sequence-independent mechanism used by plants to retain lncRNA at chromatin. We showed that U1 snRNP is required for MUSHER, COOLAIR and PUPPIES to bind the chromatin efficiently. As we did not test the interplay between U1 binding and LLPS, we cannot exclude the possibility that U1 binding is required or otherwise interlinked with the phase separation of lncRNA at chromatin. R-loops have not been reported in seeds, and the one shown in seedlings at PIR1 locus⁷⁵ is unlikely to be formed by MUSHER since MUSHER is very lowly expressed in seedlings. Therefore, while we cannot exclude this possibility, MUSHER is unlikely to form R-loops. As concluded for humans, U1 lncRNA targets¹⁵ do not form direct R-loops with the target DNA. This suggests that in plants, R-loops probably do not contribute to U1-mediated lncRNA retention. The fact that the lncRNA tested here have different modes of action, implies that the U1 role in chromatin retention is not limited to a specific lncRNA activity. Altogether, our data strongly supports the idea that the U1 role in chromatin retention of lncRNA is evolutionarily conserved between animals and plants.

Methods

Plant material and growth conditions

WT used in this work is Columbia (Col-0) Arabidopsis thaliana accession. Plants were grown on soil in a greenhouse under long-day conditions (16 h light/8 h dark) at 22 °C/18 °C. All mutants and transgenic lines were created in Col-0 background and were homozygous for the indicated genotype. MUSHER lines msh-1(SM_29297) and msh-2(SM_29310) were obtained from the NASC centre. Other mutants used in this study have been described previously: cpl1-7⁷⁶, cpl1-8 (GK-165H09)³², cpl1-9 (GK-849H10)³², dog1-3 (SALK_000867)²⁷, dog1-4 (SM_3_20886)³⁰, fy-2 (SAIL_ 657D4)⁷⁷, esp1-2 (SALK_078793)⁷⁸, pir1 (SALK_078237)³⁶, amiR-u1-70k and amiR-u1-c²¹.

Germination tests

Germination assays were performed as described previously to estimate primary dormancy levels^7,30. Briefly, 100–200 freshly harvested seeds were sown on Petri plates with blue germination paper (HoffmanMfg) and a layer of water-soaked fabric underneath to keep moisture. The plates were sealed and placed in the Percival chamber under long-day conditions (16 h light/8 h dark) at 22 °C/18 °C. Photographic documentation was performed once a day, and seeds with root protrusion were counted as germinated.

To evaluate secondary dormancy induction levels, seeds were sown on Petri plates with blue germination paper (HoffmanMfg) and a layer of water-soaked fabric underneath to keep moisture. The plates were sealed and incubated in the darkness at varying temperatures (25 °C, 28 °C, 35 °C) for 7 days. Subsequently, plates were transferred to the Percival chamber under long-day conditions (16 h light/8 h dark) at 22 °C/18 °C.

An abscisic acid (ABA) sensitivity test (Fig. S5) was conducted using the genotypes Col-0, dCas9-control-1, msh-1, msh-2, msh_dCas9-1, msh_dCas9-2, pir1, dog1-4 and msh-1xpir1. Seeds were sown on blue paper soaked with 45 ml water supplemented with 0.1 μM ABA and stratified for 24 h before transferring to 22 °C (long day conditions). Differences in germination rates across genotypes were measured after 4 days and assessed using a Kruskal–Wallis test within each treatment, followed by Dunn’s post-hoc test for pairwise comparisons.

Cloning, plant transformation and CRISPR-dCas9 mutant generation

The generation of the dCas9 transgenic plants was performed according to Čermák et al.⁴². The specific sgRNAs to Arabidopsis TAIR10 genome were designed using the CRISPR-P 2.0 (http://crispr.hzau.edu.cn/cgi-bin/CRISPR2/CRISPR) to select the canonical NGG PAM motif. sgRNAs were cloned to the dCas9 vector by Golden Gate assembly (dCas9_control-1, dCas9_control-2, msh_dCas9-1 and msh_dCas9-2) according to the protocols 3B and 5⁴². Primers used for cloning are listed in Supplementary Data 7. All constructs were verified by Sanger sequencing. Each construct was transformed into Agrobacterium strain GV3101 and then transformed into Col-0 or an appropriate mutant background using the floral dip method. T1 transformant plants were identified by selecting the resistance to glufosinate. The seeds from the selected T1 plants were used for phenotypic analysis. At least two independent transgenic lines were tested and propagated. As a control for potential indirect effects from dCas9 expression, we also created two dCas9 lines with guides recognising different intergenic regions localised 100 kb from DOG1 locus: dCas9_control_1 and dCas9_control_2, which showed no influence on MUSHER transcription. To generate MUSHER-complemented lines, msh-1 pMUSHER_1 and msh-1 pMUSHER_2, we modified the pGWB635_LUC vector by introducing the MUSHER genomic sequence along with its promoter region (>4 kb). Although both constructs share the same MUSHER promoter start site in the Arabidopsis thaliana genome, the luciferase (LUC) reporter integration begins at position 18,589,360 for pMUSHER_1 and 18,589,685 for pMUSHER_2. Due to the shared BASTA resistance in msh-1 and the pGWB635_LUC plasmid, we crossed pMUSHER_1 in the Col-0 background with msh-1 plants to generate complemented lines, selecting double-homozygous individuals in subsequent generations.

Luciferase measurement

100 seeds placed on a microscope glass slide were squashed to release the embryo from the seed coat. After adding D-Luciferin (Beetle luciferin potassium salt, Promega) and 20 min of incubation in darkness, the light signal was measured using a NightSHADE camera (Berthold Technologies GmbH & Co.KG) with a 10-minute exposure.

RNA extraction

RNA extraction protocol was adapted from Kowalczyk et al.³². After grinding in liquid nitrogen, samples were resuspended in 500 μl of homogenisation buffer (100 mM Tris pH 8.5, 5 mM EDTA, 100 mM NaCl, 0.5% SDS, 1% β-mercaptoethanol) and centrifuged for 5 min at room temperature at 8000 g. The supernatant was transferred to the new tubes containing 250 μl of chloroform. After 15 min of shaking, 250 μl of phenol was added, and shaking was repeated, followed by centrifugation for 10 min at 10,000 g. The upper phase was transferred to new tubes, and an equal volume of phenol:chloroform:isoamyl alcohol was added. Shaking and centrifugation steps were repeated. The upper phase was mixed with 10% of the volume of 3 M sodium acetate (pH 5.2) and an equal volume of isopropanol. After centrifugation for 30 min at 20,000 g at 4 °C, the pellet was washed with 70% EtOH and resuspended in water.

RT–qPCR

Trace DNA in the RNA samples was removed using a TURBO DNA-free kit (Thermo Fisher Scientific) according to the manufacturer’s protocol. The DNA removal was confirmed by PCR using primers to the PP2A gene (AT1G13320). The RNA sample quality and quantity were checked by agarose gel electrophoresis and using a Nanodrop 2000 spectrophotometer. Reverse transcription using 2.5 µg of RNA was performed with a Superscript III cDNA Synthesis kit (Thermo Fisher Scientific) according to the manufacturer’s protocol. An equal mixture of oligo(dT) and hexamer primers was used for cDNA synthesis according to the manufacturer’s protocol. 5x diluted cDNA was used as a template in the qPCR reaction (LightCycler 480; Roche) with SYBR Green mix (Roche). The relative transcript level was determined with the 2 − ΔΔCt method²⁸ and normalised to mRNA of the UBC21 (AT5G25760) gene. The primers used are listed in Supplementary Data 7.

Chromatin ImmunoPrecipitation (ChIP)

ChIP was performed as described³² with the following modifications. 100 mg of fresh siliques (13–20 days after pollination) were ground in liquid nitrogen and resuspended in 20 ml of crosslink buffer (10 mM HEPES-KOH pH 7.4, 50 mM NaCl, 100 mM sucrose, 1% formaldehyde). After 10 min incubation on a rotator at 4 °C, the crosslinking agent was quenched by adding 1.25 ml of 2 M glycine. After another 5 min incubation on a rotator, samples were centrifuged at 4 °C and 2000 g for 15 min. The pellets were resuspended in 20 ml of cold Honda Buffer (20 mM HEPES-KOH pH 7.4, 0.44 M sucrose, 1.25% Ficoll, 2.5% Dextran T40, 10 mM MgCl₂, 5 mM DTT, 0.5% Triton X-100, 10 mM β-mercaptoethanol, 1 mM PMSF, 1x Complete protease inhibitors (Roche)) and filtered through Miracloth. Samples were centrifuged for 15 min at 2000 g at 4 °C. The pellet was resuspended in ChIP lysis buffer (50 mM HEPES-KOH pH 7.4, 10 mM EDTA, 1% SDS, 1x Complete protease inhibitors (Roche)) and sonicated using Bioruptor (Diagenode) to fragments of 300–500 bp. Samples were next diluted 10 times with ChIP Dilution Buffer (1.1% Triton X-100, 1 mM EDTA, 16.7 mM HEPES-KOH pH 7.4, 167 mM NaCl, 1 mM DTT, 1 mM PMSF, 1x Complete protease inhibitors (Roche)). Then antibodies against RNA polymerase II subunit B1 (Agrisera, AS11 1804), H3 (Abcam, ab1791), H3K4me3 (Millipore, 07-473), H3K27me3 (Abcam, ab6002) or CPSF73 (PhytoAB, PHY1043A) were added followed by addition of the mixture of Dynabeads Protein A and Protein G beads (Thermo Fisher Scientific). Next, samples were incubated overnight on a rotator at 4 °C. Two washing steps with low salt buffer (150 mM NaCl, 1 mM EDTA, 10 mM HEPES-KOH pH 7.4, 0.1% Triton X-100) and two washing steps using high salt buffer (500 m M NaCl, 1 mM EDTA, 10 mM HEPES-KOH pH 7.4, 0.1% Triton X-100) were performed. DNA was released from beads and reverse-crosslinked by incubation at 95 °C with 10% Chelex 100 suspension for 15 min. After Proteinase K treatment (55 °C) for 90 min and repeated incubation at 95 °C for 10 min, samples were centrifuged for 5 min at RT at 10,000 g. Recovered DNA was used in qPCR reaction with primers listed in Supplementary Data 7.

RNA ImmunoPrecipitation (RIP)

For U1-70K-GFP RIP, we used 100 mg of fresh siliques (13–20 days after pollination), which were ground in liquid nitrogen and resuspended in 20 ml of crosslink buffer (10 mM HEPES-KOH pH 7.4, 50 mM NaCl, 100 mM sucrose, 1% formaldehyde). After 10 min incubation on a rotator at 4 °C, the crosslinking agent was quenched by adding 1.25 ml 2 M glycine. After another 5 min incubation on a rotator, samples were centrifuged at 4 °C and 2,000 g for 15 min. The pellets were resuspended in 20 ml of cold Honda Buffer (20 mM HEPES-KOH pH 7.4, 0.44 M sucrose, 1.25% Ficoll, 2.5% Dextran T40, 10 mM MgCl₂, 5 mM DTT, 0.5% Triton X-100, 10 mM β-mercaptoethanol, 1 mM PMSF, 1x Complete protease inhibitors (Roche), and 5 U/ml murine RNase inhibitor (NEB)) and filtered through Miracloth. Samples were centrifuged for 15 min at 2000 g at 4 °C. The pellet was resuspended in Nuclear Lysis Buffer (50 mM HEPES-KOH pH 7.4, 10 mM MgCl₂, 1% Triton X-100, 1x Complete protease inhibitors (Roche), 20 U/ml murine RNase inhibitor (NEB)), followed by 20 min incubation at 37 °C with TURBO DNAse (Thermo Fisher Scientific). After adding SDS up to 1%, samples were sonicated using Bioruptor (Diagenode). Next, samples were diluted 10 times with RIP Dilution Buffer (0.5 mM EDTA, 10 mM HEPES-KOH pH 7.4, 150 mM NaCl, 20 U/ml murine RNase inhibitor (NEB)) and incubated with GFP Agarose beads (Chromotek) for 1 h on a rotator at 4 °C. After three washing steps with RIP Wash Buffer (150 mM NaCl, 1 mM EDTA, 10 mM HEPES-KOH pH 7.4, 0.05% NP-40, 20 U/ml murine RNase inhibitor (NEB)), RNA was eluted (150 mM NaCl, 1 mM EDTA, 10 mM HEPES-KOH pH 7.4, 1% SDS, 20 U/ml murine RNase inhibitor (NEB)) and reverse-crosslinked by incubation at 95 °C for 15 min. After 1 h Proteinase K treatment (55 °C), RNA was isolated with Trisure (Bioline) according to producer protocol. After cDNA synthesis with SuperScript III (Thermo Fisher Scientific), samples were used in qPCR reaction with appropriate primers listed in Supplementary Data 7.

Nuclei purification for ChIRP, RIRP and ChIDP

To uncover RNA-RNA (RNA Isolation by RNA Purification), RNA-DNA (Chromatin Isolation by RNA Purification) and DNA-DNA (Chromatin Isolation by DNA Purification Assay) interactions, we used 400 mg of fresh siliques (13–20 days after pollination), which were ground in liquid nitrogen and resuspended in 20 ml of crosslink buffer (10 mM HEPES-KOH pH 7.4, 50 mM NaCl, 100 mM sucrose, 1% formaldehyde (for RNA-RNA and RNA-DNA) or 0.5% glutaraldehyde (for ChIDP)). After 10 min incubation on a rotator at 4 °C, the crosslinking reaction was quenched by adding 1.25 ml 2 M glycine. After an additional 5 min incubation on a rotator, samples were centrifuged at 4 °C and 2,000 g for 15 min. The pellets were resuspended in 20 ml of cold Honda Buffer (20 mM HEPES-KOH pH 7.4, 0.44 M sucrose, 1.25% Ficoll, 2.5% Dextran T40, 10 mM MgCl₂, 5 mM DTT, 0.5% Triton X-100, 10 mM β-mercaptoethanol, 1 mM PMSF, 1x Complete protease inhibitors (Roche) and 5 U/ml murine RNase inhibitor (NEB)) and filtered through Miracloth. Subsequently, samples were centrifuged for 15 min at 2000 g at 4 °C. According to downstream assays, the nuclei pellets were resuspended in appropriate buffers (Element Zero Biolabs kit).

Chromatin isolation by RNA purification assay (ChIRP)

To enrich specific RNA molecules or chromatin fragments, we used MagIC beads with custom probes targeting those sequences (Element Zero Biolabs). Element Zero Biololabs designed probes are delivered with kits and remain confidential as protected intellectual property.

For MUSHER probes RIRP and ChIRP with MUSHER and lacZ probes, nuclei pellets were resuspended in 1.1 ml of MagIC Lysis Buffer FA (Element Zero Biolabs kit), supplemented with 1x Complete protease inhibitors (Roche) and 5 mM DTT. Next, samples were sonicated using a Bioruptor (Diagenode) and diluted two times with MagIC Lysate Dilution Buffer FA, with 1% of the volume saved as an input sample. MagIC beads were pre-cleared according to the producers’ protocol and added to the samples that were next incubated at 40 °C for 30 min with occasional shaking. This was followed by three washes using MagIC Wash Buffer I FA and nucleic acids elution by incubation at 92 °C for 5 min with 10 mM Tris-HCl pH 7.5. After Proteinase K treatment (55 °C, 90 min) and phenol:chloroform:isoamyl alcohol extraction, DNA and RNA were precipitated by adding 1 µl glycogen, 10% of the volume of 3 M LiCl, and an equal volume of isopropanol. Samples were centrifuged for 30 min at 4 °C at 20,000 g. Pellets were washed with cold 80% ethanol, resuspended in water, and samples were split in half. The presence of specific DNA fragments interacting directly or indirectly with MUSHER probes was validated by qPCR analysis. To identify RNAs enriched with MUSHER probes, half of the sample was incubated with TURBO DNAse (Thermo Fisher Scientific), extracted with phenol:chloroform: isoamyl alcohol as above, and used for cDNA synthesis and qPCR analysis. For ChIRP library preparation, DNA fragments were eluted from MagIC beads with RNase If (NEB). Libraries were prepared with the NEBNext® Ultra™ DNA Library Prep Kit for Illumina using NEBNext Multiplex Dual Index Primers. Paired-end sequencing was performed on the Illumina NovaSeq 6000.

Chromatin isolation by DNA purification assay (ChIDP)

For ChIDP, the nuclei pellet was resuspended in 1.1 ml of MagIC Lysis Buffer FA (Element Zero kit) supplemented with 1x Complete protease inhibitors (Roche) and 5 mM DTT. Samples were sonicated using Bioruptor (Diagenode), diluted with MagIC Lysate Dilution Buffer FA, and split into halves (treated with RNaseA at 37 °C for 20 min and non-treated). MagIC beads recognising the PIR1 locus were mixed with the lysates that were next incubated at 81 °C for 5 min with occasional shaking to denature DNA partially. Subsequently, samples were transferred to 40 °C for 20 min to enable binding of the probes. This was followed by six washing steps with the MagIC Lysate Wash buffer I and II, and DNA was eluted with 10 mM Tris-HCl pH 7.5 at 92 °C for 5 min. After adding an equal volume of MagIC Lysate Wash buffer I to eluted samples, proteinase K treatment (55 °C, 120 min) was performed. To extract DNA fragments, an equal volume of phenol:chloroform:isoamyl alcohol was added. The upper phase was mixed with 1 µl of glycogen, 10% of the volume of 3 M LiCl and an equal amount of isopropanol. After centrifugation for 30 min at 20000 g at 4 °C, the pellet was washed with 80% EtOH and resuspended in water. Libraries were prepared with the NEBNext® Ultra™ DNA Library Prep Kit for Illumina using NEBNext Multiplex Dual Index Primers. Paired-end sequencing was performed on the Illumina NovaSeq 6000.

ChIRP-seq and ChIDP-seq analysis

For ChIRP-seq and ChIDP-seq data analysis, Bowtie2 (v2.5.3)⁷⁹ was used to align the reads to the reference genome (TAIR10, with default parameters). To identify significantly enriched regions, MACS3 was used with the default parameters: bandwidth:300; model fold of 5:50; q-value cutoff: 0.05 (v3.0.0a6)⁸⁰. bamCoverage (v3.5.4)⁸¹ with CPM normalisation was used to create bigwig files.

5’RACE-seq

Analysis was performed as described previously⁷. Shortly, 500 ng of total RNA, after DNAse treatment, was used as the template for cDNA generation using SuperScript II (Thermo Fisher Scientific). The resulting cDNA was purified using AMPure XP magnetic beads (Beckman Coulter) and amplified in a series of three PCR reactions with Phusion polymerase and specific primers (1^st PCR: only TSO_n1, 2^nd PCR: TSO_n2 and MUSHER_5RACE, 3^rd PCR: Illumina indexing primers). The final PCR product was sequenced using Illumina MiSeq. For data analysis, STAR (v2.7.8a)⁸² was used to align the reads to the reference genome, followed by UMI-based PCR duplicates removal using UMI-tools (v1.1.0)⁸³. The position of the first 5′ nucleotide was extracted using bedtools (v2.30.0)⁸⁴.

3’RACE-seq

Analysis was performed as described previously⁷ and based on the earlier procedure⁸⁵. The ligation of the pre-adenylated adaptor to the 3’end of the RNA was performed using truncated T4 RNA Ligase 2 (NEB). RNA ligated with RA3_15N adaptor (containing UMI) was cleaned with AMPure XP magnetic beads and subjected to reverse transcription reaction with SuperScript III (Thermo Fisher Scientific). After three rounds of PCR with specific primers (1^st PCR: MUSHER_3RACE and RTPXT, 2^nd PCR: mXTf and mXTr,3^rd PCR: Illumina indexing primers), each followed with AMPure beads purification, prepared libraries were sequenced using Illumina MiSeq. For data analysis, STAR (v2.7.8a)⁸² was used to align the reads to the reference genome, followed by UMI-based PCR duplicates removal using UMI-tools (v1.1.0)⁸³. The position of the last 3′ nucleotide was extracted using bedtools (v2.30.0).

3’RNA-seq

Analysis was performed as described previously⁴⁹. Briefly, 0.5 µg of total RNA was reverse transcribed with UMI-containing oligo(dT) primers. Differently barcoded cDNAs were pooled together, and second-strand cDNA was synthesised using the nick translation method. Next, libraries were tagmented with Tn5 transposase, and the 3’cDNA fragments were amplified with Illumina-compatible primers⁴⁹. Paired-end sequencing was performed on the Illumina NextSeq 500. Alignment and read demultiplexing were performed using STAR (v2.7.8a)⁸² and BRBseqTools (v 1.6)⁸⁶, respectively. Differentially expressed genes were identified using DESeq2 with a cutoff of FDR < 0.05 and FC > 1.5.

Fractionation

Chromatin RNA extraction was performed using the described method⁷. 100 mg of siliques were ground in liquid nitrogen and resuspended in 20 ml of cold Honda Buffer (20 mM HEPES-KOH pH 7.4, 0.44 M sucrose, 1.25% Ficoll, 2.5% Dextran T40, 10 mM MgCl₂, 5 mM DTT, 0.5% Triton X−100, 10 mM β-mercaptoethanol, 1 mM PMSF, 1x Complete protease inhibitors (Roche), and 5 U/ml murine RNase inhibitor (NEB)). After rotation at 4 °C for 10 min and filtration through Miracloth, samples were centrifuged at 2000 g for 15 min at 4 °C. The nuclei pellet was washed twice with 5 ml of Honda buffer. The pellet was resuspended in 600 µl of Honda buffer and purified on a 40%-75% PercoII density gradient by centrifugation at 10,000 g for 30 min at 4 °C. Purified nuclei were collected from the Percoll interface, washed with Honda buffer and resuspended in 500 µl of cold glycerol buffer (20 mM Tris-HCl pH 8.0, 75 mM NaCl, 0.5 mM EDTA, 0.85 mM DTT, 50% glycerol, 1% Empigen, 10 mM β-mercaptoethanol, 0.125 mM PMSF, 1x Complete protease inhibitors (Roche), and 5 U/ml murine RNase inhibitor (NEB)). Samples were overlaid on urea lysis buffer (10 mM HEPES-KOH pH 7.4, 7.5 mM MgCl₂, 0.2 mM EDTA,300 mM NaCl, 1 M urea, 1% NP-40, 10 mM β-mercaptoethanol, 0.5 mM PMSF, 1x Complete protease inhibitors (Roche), and 5U/ml murine RNase inhibitor (NEB), vortexed for 2 s, and incubated on ice for 30 min, followed by centrifugation at 20,000 g for 2 min at 4 °C. The chromatin pellet was washed twice with 600 µl of urea lysis buffer for 30 min on a rotator at 4 °C. Chromatin pellet was used for RNA extraction and next for DNAse treatment as described above. The chromatin-attached RNA quality was tested by agarose gel electrophoresis and quantified with a Nanodrop 2000 spectrophotometer. cDNA synthesis was performed using a SuperScript III kit (Invitrogen).

smFISH

smFISH assay was performed as described previously⁷ using a manually dissected radicle from dry embryos. The Stellaris probes used for smFISH target the full shDOG1 sequence, including intron 1 and are labelled with Quasar670 fluorophore (Biosearch Technologies). The embryos were fixed, permeabilised and hybridised with the probes as described⁷. The smFISH signals were gathered using a widefield fluorescence microscope Olympus IX81 (Olympus) and a Hamamatsu Orca-R2 (C10600) CCD camera. The xCellence software (Olympus) was used for image acquisition. Cells were manually segmented using Napari⁸⁷ and the PartSeq plugin, which allows foci identification and classification based on subcellular location⁸⁸. Dots number and intensity were analysed using R scripts described earlier⁷.

Statistics & reproducibility

All statistical analyses were performed using RStudio or Excel. Sample sizes are mentioned in figure legends and/or displayed in plots as individual data points. Sample sizes were not pre-specified. Information regarding the statistical test applied to each data is mentioned in the corresponding figure legend or source data. Significant differences were accepted at p  <  0.05.

Declaration of generative AI and AI-assisted technologies in the writing process

While preparing this work, the authors used Grammarly to improve readability and language. After using this tool, the authors reviewed and edited the content as needed and take full responsibility for the publication’s content.

Reporting summary

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