Co-transcriptional gene regulation in plants

Abstract

The essence of life lies in the precise regulation of genetic information flow, namely, the central dogma, with gene transcription playing a pivotal role as the starting point. For a living cell, it is not only essential via transcription to synthesize the RNA but also to ensure its timely processing, packaging, and sorting, thereby determining its distinct fate, such as nuclear retention, export, translation, or degradation. Initially observed in yeast and animals, and more recently in plants, a large amount of evidence indicates that RNA processing and protein–RNA packaging occur largely concomitantly with transcription, a phenomenon known as co-transcriptional gene regulation. Increasing evidence suggests that this mechanism provides extensive regulatory potential for gene expression. It not only ensures timely RNA processing, thus determining the fate of RNA, but may also influence the transcription dynamics of RNA polymerase II (Pol II) and the chromatin environment. In this review, we highlight recent advances in understanding co-transcriptional gene regulation in the model plant Arabidopsis thaliana, focusing on Pol II dynamics post-initiation and their interplay with RNA-processing events such as capping, splicing, 3′ end processing, protein–RNA interactions, and RNA fate determination. By comparing these findings with progress in other model systems, we discuss the unique characteristics of co-transcriptional gene regulation in plants and its potential biological significance. Additionally, we introduce recent key discoveries at the FLOWERING LOCUS C (FLC) gene under warm conditions, which exemplify how co-transcriptional RNA processing influences the chromatin environment and leads to long-term regulatory impacts. Finally, we provide perspectives on yet-unanswered key questions related to co-transcriptional gene regulation in plants.

1. Introduction

Living systems uphold remarkable homeostasis despite the highly dynamic nature of the underlying molecular events. A striking feature of transcription is its intensely dynamic nature, continuously generating vast amounts of RNA. By the time these sequence-diverse RNAs are released from chromatin, they already appear to be “pre-packaged”, effectively prepared for their next destination, a process largely governed by co-transcriptional gene regulation [1,2]. In eukaryotes, precursor mRNAs (pre-mRNAs) must undergo extensive processing steps, such as 5′-end capping, splicing, 3′-end cleavage, and polyadenylation, to become mature mRNAs. These processing steps are tightly coupled to transcription [1]. The 5′ m⁷G cap is added through the recruitment of capping enzymes by the phosphorylated C-terminal domain (CTD) of RNA polymerase II (RNA Pol II), facilitated by its physical interaction with the core polymerase [3,4]. This process occurs as soon as the first 25–30 nucleotides stretch out from Pol II, and the resulting 5′ end cap not only protects the RNA from degradation by 5′–3′ RNA exonuclease but also serves as a platform for the recruitment of factors involved in subsequent RNA-processing steps. Pre-mRNA splicing largely occurs during transcription elongation and induces the most dramatic changes to pre-mRNA compared with other RNA-processing steps [3,5]. The removal of introns remodels the pre-mRNA into a significantly different and much shorter sequence. Importantly, splicing holds great regulatory potential, as different isoforms can be produced through alternative splicing [6,7] and RNA fate can be modulated through splicing-coupled RNA-binding protein delivery. The latter case is best exemplified by the splicing-dependent delivery of the exon junction complex (EJC), which is critical for nonsense-mediated decay (NMD) in the cytosol [8]. The elongation rate of Pol II can influence the choice of splice sites, and therefore affect the splicing outcome, a scenario known as “kinetic coupling” [9], whereas recognition of the intron 5′ splice site can prevent Pol II from premature termination when it transcribes through the intron, a mechanism known as “telescripting” [10,11]. This phenomenon indicates the intricate interplay between Pol II transcription and splicing. When Pol II transcribes through the 3′ end of a gene, the exposure of the polyadenylation signal in nascent RNA recruits the cleavage and polyadenylation specificity factor (CPSF, or cleavage and polyadenylation factor [CPF] in yeast), which is associated with Pol II and is responsible for endonuclease cleavage and subsequently polyadenylation [12,13]. These activities are coupled with Pol II slowing down and conformational changes, and eventually Pol II termination as a result of 5′–3′ exonuclease-mediated decay of the cleaved nascent RNA that is attached to Pol II [[14], [15], [16]].

The molecular details of co-transcriptional regulation in plants are gradually being uncovered, although significant gaps remain in both the depth and breadth of research compared to that in mammalian cells and yeast. Recent studies showed that many basic principles are shared among eukaryotes, including plants. For example, approximately 70 ​%–80 ​% of introns in Arabidopsis (Arabidopsis thaliana) are removed co-transcriptionally, similar to that observed in fruit fly (Drosophila melanogaster) and mammalian cells [5,17,18]. However, plant genomes exhibit unique characteristics: compared to mammalian genes, Arabidopsis genes generally have fewer and much shorter introns, and the intergenic distances are also much smaller (Table 1). Moreover, plants and animals display pronounced differences in alternative splicing. In plants, intron retention (IR) has been reported as the most frequent splicing pattern, whereas in animals, it is relatively rare [7,19]. The widespread IR suggests the existence of plant-specific mechanisms of RNA quality control. Furthermore, the short intergenic distance and unique gene structure in plants were linked to more efficient Pol II termination and effective co-transcriptional splicing (CTS) of intron-rich genes, respectively [32]. In addition, plant genomes encode a large number of so-called “plant-specific” proteins involved in transcription or RNA processing, for which clear homologs are missing or yet to be identified in animals [33]. The extent to which these proteins harbor plant-specific functions or represent functional counterparts of proteins in animals remains largely unknown. For example, FCA and FPA, two related RNA recognition motif–containing proteins involved in transcription termination and polyadenylation in plants, lack clear sequence-homologs in animals, albeit recent evidence suggests that Pcf11 may represent the functional counterpart of FCA in Drosophila melanogaster [[34], [35], [36], [37]]. Similarly, the N-terminal sequence of the RNA Pol II largest subunit NRPB1 shows only ∼50 ​% identity between plants and animals. In general, lots of basic principles of co-transcriptional gene regulation in plants remain unknown and necessitate in-depth future research.

Table 1.

Comparison of gene structures and splicing patterns among different species.

Species	Exon	Intron	Intergenic distance	Frequency of ASa	Pattern of AS	References
Arabidopsis thaliana	∼299 bp	∼166 bp	∼1.6 ​kb	>60 ​%	Intron-retention: 40–60 ​%	[7,20,21]
Oryza sativa	∼359 bp	∼488 bp	∼7.5 ​kb	>50 ​%	Intron-retention: 40–50 ​%	[7,21,22]
Physcomitrium patens	∼321 bp	∼289 bp	∼16.5 ​kb	∼50 ​%	Intron-retention: ∼50 ​%	[23,24]
Homo sapiens	∼248 bp	∼6.2 ​kb	∼42.1 ​kb	∼95 ​%	Exon skipping: >40 ​%	[[25], [26], [27]]
Drosophila melanogaster	∼486 bp	∼1.6 ​kb	∼3.5 ​kb	∼40 ​%	Intron-retention: ∼20 ​%	[28,29]
Saccharomyces cerevisiae	∼1.2 ​kb	∼303 bp	∼0.5 ​kb	Very rare	Not applicable	[30,31]

^aAS, alternative splicing.

This review focuses on recent advances in understanding co-transcriptional gene regulation in plants. We highlight recent discoveries in Arabidopsis related to co-transcriptional RNA splicing, RNA Pol II transcription elongation, RNA fate determination, and transcription-coupled RNA–protein interactions. We also compare and discuss these discoveries in light of progress made in other model systems. In addition, we briefly introduce the progress at the FLOWERING LOCUS C (FLC) gene, which serves as an excellent model to dissect the co-transcriptional gene regulation linked to the chromatin environment. Finally, we propose a few key questions that may warrant future investigation.

2. Co-transcriptional splicing is a dominant mode of pre-mRNA splicing in plants

Current evidence indicates that CTS is widespread in eukaryotes. The discovery of CTS dates back to the 1980s, when the development of Miller spread techniques enabled the direct visualization of nascent RNA attached to chromatin via electron microscopy. Using such techniques, studies in Drosophila demonstrated that splicing could be completed within approximately 3 ​min following Pol II transcription of the intron's 3′ end [38]. With the advent of next-generation sequencing (NGS) technologies over the past 15 years, extensive data have further shown the prevalence of CTS across diverse species. One of the earliest evidence for CTS in plants emerged from studies on the FLC, a MADS-box transcription factor that represses flowering in response to both endogenous and environmental cues. Genetic analysis over the last 20 years has identified numerous regulators of FLC, including many RNA-processing factors, Pol II elongation factors, and chromatin modifiers, establishing it as a classic model for studying co-transcriptional gene regulation [39,40]. Loss of the RNA-binding protein FCA led to the simultaneous upregulation of FLC transcription initiation and elongation rates, resulting in a distinctive pattern of nascent RNA expression across FLC intron 1: a relatively low fold change between Col-0 and fca-9 at the 5′ end, yet a high fold change at the 3′ end, a profile characteristic of rapid CTS combined with an accelerated elongation rate in fca-9 [41]. Further support came from single-molecule RNA fluorescence in situ hybridization (smFISH) analysis: wherever unspliced FLC signals were observed in the nucleus, spliced FLC signals were also co-localized within the nucleus, consolidating the idea that FLC splicing predominantly takes place co-transcriptionally [42,43]. In 2019, two research groups independently reported the genome-wide features of CTS in Arabidopsis via Illumina short-read sequencing of chromatin-bound RNA (CB-RNA-seq) [17,18]. Methodology-wise, these studies leveraged the principle that washing chromatin with 1 ​M urea retains Pol II while effectively removing other loosely associated proteins [41,44], thereby enabling a relatively straightforward and efficient enrichment of RNA still attached to the transcription complex. By comparing the proportions of introns and exons in nascent RNA at steady state, both studies arrived at a similar conclusion: in about 70–80 ​% of cases, introns in the Arabidopsis genome are removed co-transcriptionally [17,18]. Subsequent CB-RNA-seq data from soybeans (Glycine max) corroborated these observations [45]. Moreover, researchers employed direct immunoprecipitation of Pol II followed by sequencing of its bound RNA, referred to as native elongating transcript sequencing (NET-seq), and detected a clear signal at the last nucleotide of the 3′ ends of exons [[46], [47], [48]]. Such a signal most likely represents a splicing intermediate (an exon with its 3′ end being cleaved by the spliceosome) that remains attached to the Pol II elongation complex, providing additional evidence that CTS is a prevailing mode of splicing in plants.

Collectively, these findings confirm the prevalent nature of CTS in plants. The following section focuses on the kinetics of splicing, examining the rate at which this process occurs and the factors that influence it.

3. Co-transcriptional splicing kinetics in plants

It is somewhat unexpected that pre-mRNAs in plants, fruit flies, and mammals exhibit similar CTS efficiencies, given the marked differences in gene structure between plants and mammals (Table 1). Mammalian genes typically contain introns measuring several kilobases (kb) in length, whereas their exons are often only ∼0.2 ​kb. By contrast, plant and fruit fly genes generally feature exons and introns of comparable size (approximately 0.3 ​kb each). It is widely accepted that longer introns or longer overall gene lengths favor CTS, because Pol II requires more time to complete transcription, thereby providing a larger “co-transcriptional window” for splicing [9,49]. Indeed, in both plants and animals, splicing of multi-intron genes often follows a “first come, first served” model, whereby introns closer to the 5′ end (and thus transcribed earlier) have a higher likelihood of being spliced earlier [50,51]. Given this context, the observation that plant genes, whose overall lengths are often at least an order of magnitude smaller than those of mammals, nevertheless maintain similar CTS efficiencies suggests the existence of plant-specific regulatory aspects ensuring efficient CTS.

Current evidence indicates that CTS occurs more rapidly than previously anticipated and is tightly coupled to Pol II elongation. Many studies in animals and yeast have attempted to quantify the speed of splicing. One popular approach involves sequencing-based measurements of the distance between the nascent RNA 3′ end and the nearest exon–exon junction. Interestingly, although various studies broadly converge on the conclusion that CTS is indeed prevalent, the exact values for this distance differ significantly among species, even within the same species when different methods are used. For example, short-read Illumina sequencing in budding yeast (Saccharomyces cerevisiae) suggested that once Pol II transcribes roughly 45 bp past the 3′ splice site, about half of splicing events have already occurred [52]. In mammals, PacBio sequencing of the nascent RNA has placed this distance at around 300 bp [53], whereas another Nanopore-based study reported a much longer distance of several kb [54]. Notably, Nanopore full-length cDNA sequencing in Arabidopsis yielded a value of ∼1.2 ​kb [50]. Technical progress and discussions on these findings have been reviewed in detail elsewhere [5]. Overall, current studies imply that splicing occurs soon after Pol II transcribes past the intron's 3′ end. It is worth noting that current estimates of splicing kinetics should be interpreted with caution due to several technical limitations. Short- and long-read sequencing differ not only in read length but also in library preparation protocols, which may introduce bias. For example, the accuracy of cDNA-based long-read sequencing depends on unbiased reverse transcription and subsequent PCR amplification across an RNA pool with widely varying transcript lengths, a technically challenging task. In addition, the relatively high error rate of early Nanopore sequencing data, compared to PacBio or Sanger sequencing data, poses difficulties in accurately determining the identity and position of individual nucleotides. This is particularly critical for analyses that infer splicing kinetics based on the distance between RNA 3′ ends and exon–exon junctions. For instance, due to the limitations of Nanopore data, one study discarded all reads whose cDNA 5′ ends fell within ±10 nucleotides of an intron 5′ end, treating them as potential splicing intermediates [50]. In addition, variations in Pol II elongation rates across different genes (which may differ in intron length, intron number, or other characteristics) and species further complicate interpretation. Future efforts incorporating various RNA sequencing methods (e.g., direct RNA sequencing), enhanced library preparation, and plant-specific analytical pipelines will be crucial to get a more accurate and comprehensive understanding of CTS kinetics in plants.

Nevertheless, splicing kinetics data highlight the tight coupling between the splicing reaction and Pol II transcriptional elongation. An elegant structural study from the Cramer group provided a vivid depiction of this process in mammals: the RBP2 subunit of Pol II interacts with the U1–70K of the U1 small nuclear ribonucleoprotein particle (U1 snRNP) complex [55]. Shortly after the 5′ splice site is transcribed, cleavage occurs, and Pol II subsequently “drags” the 3′ end of the excised exon and the 5′ end of the intron while elongating. This may facilitate the recognition of the branch point near the 3′ end of the intron and the formation of the intron lariat, followed by the completion of the splicing reaction. This finding provides a mechanistic explanation for CTS: Pol II elongation and the spliceosome are physically coupled, and Pol II elongation through introns facilitates intron recognition and splicing. Notably, plants can also possess long introns (e.g., longer than 500 bp), such as intron 1 of FLC (and many other MADS-box-containing genes [56]), the splicing of which indeed occurs soon after Pol II finishes transcribing it [41]. Considering the generally short introns in the Arabidopsis genome, it is possible that these long introns are evolutionarily selected to bear unique regulatory potential, a scenario that necessitates further studies.

4. Coupling between splicing and 3′ end pausing of Pol II in plants

As mentioned above, a typical feature of plant genes is that introns and exons have comparable and generally short lengths. Related to this feature, the efficiency of CTS in Arabidopsis and soybean is in good correlation with intron number, rather than just overall gene length [32,45] (Fig. 1A and B). Specifically, genes with a larger number of introns tend to exhibit higher CTS efficiency across all introns, indicating coordination of splicing among individual introns within the same gene. Similar phenomena have been reported in mammals, in which introns within the same gene often exhibit an “all or none” pattern of splicing [53,57,58]. A recent study in Arabidopsis further revealed that the high CTS efficiency observed in multi-intron genes is closely linked to Pol II pausing near the 3′ end of the gene [32] (Fig. 1A and B). By integrating NET-seq data with mathematical modeling, researchers found that the more introns a gene contains, the longer Pol II tends to pause at the 3′ end, resulting in more efficient CTS. For example, in genes with lower levels of 3′-end Pol II pausing, even genes harboring more introns do not exhibit a marked increase in CTS efficiency relative to genes with fewer introns. This suggests that Pol II pausing at the 3′ end facilitates efficient CTS in multi-intron genes. Moreover, when spliceosome assembly was chemically inhibited, Pol II pausing at the 3′ end was dramatically reduced, and the correlation between exon number and Pol II pausing was abolished, suggesting that splicing actively promotes 3′ end pausing. Notably, 3′-end Pol II pausing is an innate feature associated with Pol II termination. These observations imply that 3′-end Pol II pausing functions as an RNA-processing checkpoint, at which spliceosome assembly promotes 3′-end Pol II pausing, which in turn ensures that the RNA released from chromatin is effectively spliced [32].

Fig. 1.

Tight coupling between RNA Pol II pausing and co-transcriptional splicing efficiency. A Co-transcriptional splicing (CTS) efficiency is positively correlated with intron number. Transcripts with fewer introns exhibit lower CTS efficiency, whereas transcripts containing multiple introns demonstrate enhanced CTS. B RNA Pol II pausing at the 3′ end and its reciprocal effects on CTS efficiency. RNA Pol II occupancy profile showing a peak near the 3′ end of the gene. Splicing promotes this pausing, as the inhibition of spliceosome assembly reduces 3′ end pausing. Conversely, stronger 3′ end pausing enhances the correlation between exon number and CTS efficiency, indicating that 3′ end pausing likely facilitates effective CTS. C Intron-dependent CTS and its regulatory dynamics in RNA biogenesis. Intron-containing transcripts in the chromatin fraction include delayed splicing events and intermediate forms that have undergone partial processing. Fully spliced transcripts are enriched in the cytoplasm. TSS, transcription start site; TES, transcription end site; PAS, polyadenylation site.

Consistent with these observations, by ligating adapters to the 3′ ends of nascent RNAs and performing Nanopore cDNA sequencing, it was reported that certain introns can be spliced after the nascent RNA is polyadenylated, termed “post-transcriptionally (PTS) spliced introns” [50,59]. Intriguingly, these PTS introns were detected almost exclusively in the chromatin fraction, suggesting that many of them might represent intermediate splicing products awaiting completion of the splicing reaction on chromatin (Fig. 1C). The precise fraction of PTS introns that are strictly spliced post-transcriptionally (that is, after transcripts are released from chromatin upon transcription termination) and the biological implications of these introns, remain open questions. Nevertheless, some introns indeed have a delayed kinetics of splicing and may be related to IR events as previously reported, especially those targeted by Protein arginine methyltransferase 5 (PMRT5), a splicing regulator that likely functions at both the co- and post-transcriptional levels [50,60,61]. Collectively, current findings suggest that plants, despite having shorter genes than animals, have evolved a specialized mechanism in which Pol II pauses at the 3′ end to facilitate splicing, thereby achieving CTS efficiency comparable to that in animals, highlighting the unique regulatory strategies governing CTS in plants.

5. Pol II dynamics at the 5′ ends of genes as a key regulatory layer for transcription output

Studies over the past three decades have revealed that the regulatory mechanisms governing Pol II during its elongation are extraordinarily complex. In mammals, following transcription initiation, Pol II does not immediately enter active elongation; instead, it accumulates at approximately 20–60 nucleotides downstream of the transcription start site (TSS), referred to as promoter-proximal pausing [[62], [63], [64]] (Fig. 2A). Such a paused state is stabilized by DRB sensitivity-inducing factor (DSIF) and negative elongation factor (NELF), whereas the positive transcription elongation factor b (P-TEFb) kinase promotes the release of Pol II from pausing into active elongation by promoting the phosphorylation of elongation factor suppressor of Ty 5 (SPT5) and Ser2 of the Pol II CTD [65,66]. Notably, using improved sequencing technologies, the latest evidence suggests that up to 80 ​% of mammalian Pol II that initiate fail to enter active elongation; instead, they terminate prematurely in the promoter-proximal region [[67], [68], [69]]. The resulting transcripts are actively degraded and therefore rarely detected unless nuclear RNA decay complexes are inactivated. These findings parallel the stable unannotated transcripts and cryptic unstable transcripts observed in yeast, which are transcribed within the promoter-proximal regions [70,71]. Intriguingly, however, Pol II in yeast does not exhibit prominent promoter-proximal pausing comparable to that in mammals [72], implying that while promoter-proximal pausing may be species-specific, premature termination near the TSS could be broadly conserved across species [73]. Such premature termination may function as part of a quality control mechanism for transcription, serving to remove Pol II complexes that are not fully competent for elongation. Alternatively, it may be selectively utilized under certain conditions to fulfill particular roles, for instance, in shaping the local chromatin environment [74,75].

Fig. 2.

5′-end dynamics of RNA polymerase II and their impact on transcriptional output. A Contrasting 5′-end Pol II pausing and early termination in human and Arabidopsis. Human Pol II exhibits prominent promoter-proximal pausing, with ∼80 ​% of initiated Pol II undergoing premature termination. In contrast, plant (Arabidopsis) Pol II does not show marked accumulation in this narrow zone, although ∼14 ​% of expressed genes produce unstable short promoter-proximal RNAs. B U1 snRNP-mediated telescripting prevents premature polyadenylation in introns. U1 snRNP recognizes the 5′ splice site, promoting Pol II elongation and preventing premature polyadenylation, particularly at premature cleavage and polyadenylation sites within introns.

In plants, data from Pol II NET-seq and global run-on sequencing (GRO-seq) suggest that plant Pol II transcription more closely resembles that in yeast than that in mammals, as plant Pol II does not show a marked accumulation within a narrow promoter-proximal zone [46,47] (Fig. 2A). Nevertheless, both unphosphorylated CTD Pol II and Ser5-phosphorylated CTD Pol II tend to accumulate in the promoter-proximal region, whereas Ser2-phosphorylated Pol II is enriched at the 3′ end, indicating that, similar to animal Pol II, plant Pol II undergoes a CTD phosphorylation shift from unphosphorylated/Ser5P to Ser2P as it transitions from initiation to productive elongation [32,47]. Thus, plant Pol II likely also experiences promoter-proximal stalling akin to that observed in animals, although the exact stalling position may be less uniformly defined. Intriguingly, by using transcription isoform sequencing (TIF-seq), a technique that simultaneously captures RNA 5′- and 3′-end polyadenylation sites, it was revealed that approximately 14 ​% of expressed genes in Arabidopsis produce relatively unstable short promoter-proximal RNAs (sppRNAs) [76]. Many of these genes also exhibit an enrichment of unphosphorylated CTD Pol II at the 5′ end, indicating that a subset of plant Pol II that stalls in the promoter-proximal region also undergoes premature transcription termination. Thus, the regulatory control over Pol II fate near the TSS, shortly after initiation, appears to be a highly conserved mechanism across eukaryotes. To date, the fraction of Pol II undergoing promoter-proximal termination in plants and the biological consequence of such a regulatory strategy remain largely elusive.

One key regulatory feature during early elongation is the presence of introns near the 5′ ends of genes, which can significantly influence Pol II behavior by affecting elongation dynamics and termination decisions. Introns generally exhibit sequence features similar to the 3′ ends of genes, possessing significantly higher AT content than exons, thereby potentially harboring polyadenylation signal–like features. In plants, poly(A)-tag sequencing (PAT-seq) analyses have shown that alternative polyadenylation signals are frequently present within introns [[77], [78], [79]]. In both mammals and plants, a mechanism termed “telescripting” has been reported, whereby recognition of the intron 5′ splice site by U1 snRNP prevents premature transcription termination and polyadenylation in the promoter-proximal region, particularly within introns [10,11,80,81] (Fig. 2B). Notably, knockdown of the U1 snRNP components U1–70K and U1–C in Arabidopsis led to a further downstream shift of Pol II accumulation at the 3′ end of the gene, indicating a tight link between the U1 snRNP (or the spliceosome) and elongating Pol II [82]. Thus, the function of U1 snRNP extends beyond merely inhibiting polyadenylation within the gene body or introns. Related to this scenario, a recent study in mammals offered an additional perspective on how U1 snRNP influences Pol II elongation, revealing that recognition of the 5′ splice site by U1 accelerates Pol II elongation through AT-rich intronic sequences, thereby minimizing premature termination [10]. Indeed, sequences with high AT content, such as introns, has been implicated in reduced RNA polymerase stability during transcription, and more rapid elongation might help Pol II transcribe through these regions successfully [[83], [84], [85]]. The details of Pol II dynamics during transcription through introns in plants have yet to be examined. In addition, for both plants and mammals, the mechanistic details of how U1 snRNP and intronic sequences regulate Pol II transcription elongation and termination remain to be clarified.

6. Co-transcriptional determination of RNA fate

As summarized above, transcription is a highly dynamic and tightly regulated process. It can give rise to a wide array of RNA transcripts, and the sorting and identification of these RNAs presents a major challenge to the cell. Accumulating evidence indicates that co-transcriptional RNA processing plays a pivotal role in determining RNA fate. The 5′ capping of RNA occurs shortly after 20–30 nucleotides have been transcribed, a process catalyzed sequentially by RNA guanylyltransferase and RNA guanine-7 methyltransferase (RNMT) [86]. The resulting m⁷G cap serves as a binding platform for a variety of proteins, such as cap-binding complexes and nuclear export factors, which contribute to mRNA stability, nuclear export, and even the regulation of transcription elongation [[87], [88], [89]]. Conversely, an uncapped 5′ end renders the RNA a substrate for nuclear exonucleases such as XRN2 and XRN3. In the cytoplasm, RNA decapping is a key initiating step during mRNA degradation [[90], [91], [92]]. Intriguingly, it was recently discovered that some eukaryotic RNAs possess a nicotinamide adenine dinucleotide (NAD) cap instead of an m⁷G cap at the 5′ end [[93], [94], [95]]. Arabidopsis DXO1, a decapping enzyme for NAD-capped RNAs, was identified as a component of the RNMT complex that catalyzes the addition of the m⁷G cap [[96], [97], [98]]. The dynamic balance between NAD and m⁷G caps and how this influences RNA fate in plants remains an intriguing and unresolved question (Fig. 3A).

Fig. 3.

Regulatory checkpoints that define RNA fate through co-transcriptional processing. A Non-canonical NAD capping and its interplay with canonical m⁷G in plants. The RNMT complex, together with DXO1, drives the transformation of NAD caps into m⁷G caps. This dynamic process governs RNA fate by balancing NAD and m⁷G capping mechanisms in plants. B Fate of intron-retaining transcripts. The fate of intron-retained RNAs is multifaceted, encompassing roles as transcription intermediates, targets for nuclear RNA degradation, substrates that can be translated into alternative proteins or directed to translation-coupled decay, or transcripts undergoing delayed splicing followed by translation. C Roles of the exon junction complex (EJC) in mRNA quality control. In normal transcripts, EJCs are removed during the pioneer round of translation. However, in the presence of a premature termination codon (PTC), EJCs downstream of the PTC are retained, triggering nonsense-mediated decay (NMD)-mediated degradation. It is possible that in plants, EJC and NMD mechanisms may differ from those in mammals, potentially including processes such as pre-translational degradation, RNA-induced silencing complex (RISC)-mediated cleavage, or post-translational EJC retention.

Splicing of introns represents another major co-transcriptional decision point in RNA fate determination. One striking feature of plant transcriptomes distinct from those of mammals is the prevalence of IR [99]. IR-containing transcripts almost inevitably harbor premature termination codons (PTC) within the retained intron. The biological significance and regulatory mechanisms of IR have long been a focus of plant research. Current evidence indicates that IR transcripts contribute little to the proteome [100,101], implying that they are either rapidly degraded prior to translation or eventually undergo splicing, representing intermediates in the RNA maturation pipeline (Fig. 3B). Indeed, a pioneering study employed the NMD mutants upf1 and upf3 and showed that approximately one-ninth of the ∼950 alternatively spliced transcripts studied were substrates of NMD, whereas many others were NMD-insensitive [27]. Subsequent studies further revealed that many IR transcripts are retained in the nucleus, suggesting that their ultimate fate is likely to be spliced and exported from the nucleus [50,102]. A recent study proposed that light regulates spliceosome activity and the nuclear retention of IR transcripts to modulate photomorphogenesis [103]. Generally, it remains elusive whether the nucleus-retained IR transcripts are also subject to rapid nuclear RNA degradation (distinct from cytoplasmic NMD), and under what environmental or developmental cues they could be further processed and exported.

Perhaps the most well-defined example of RNA processing marking RNA fate is the EJC, which is deposited upstream of exon–exon junctions in a splicing-dependent manner. In animals, the EJC not only marks the completion of splicing but also participates in diverse co- and post-transcriptional events, including RNA splicing, export, decay, and translation [104,105]. Most notably, the EJC plays a central role in translation-dependent NMD in mammals. In this canonical model, the ribosome during the pioneer round of translation removes the EJC as it scans along the transcript (Fig. 3C). However, when a stop codon is located upstream of a remaining EJC, this triggers the recruitment of NMD factors to form an SMG1–UPF1–eRF1–eRF3 complex (SURF complex) and subsequently leads to mRNA degradation [106,107]. In mammals, the EJC is deposited approximately 24 nucleotides upstream of exon–exon junctions [108]. In plants, however, the precise location of EJC deposition remains largely speculative. Interestingly, by RNA degradome sequencing, a study found the enrichment of sequencing footprints near canonical EJC binding sites, suggesting that EJC positioning in plants is similar to that in mammals [109]. Curiously, these EJC-like signals were found in degrading RNAs lacking a 5′ cap, which differs from the model in mammals, whose EJCs are removed during the pioneer round of translation and thus should not appear in de-capped RNAs. This raises the possibility that the binding behavior and functions of the EJC in plants may differ substantially from those in mammals, and further research is required to clarify this divergence.

The case of the EJC illustrates a broader principle: RNA processing may frequently be coupled with the binding of specific RNA-binding proteins, many of which reside in the nucleus and form complex regulatory networks. For example, the Arabidopsis genome encodes at least 19 serine/arginine-rich (SR) proteins, substantially more than in mammals. These SR proteins interact dynamically with the EJC and have been implicated in nearly every stage of the mRNA lifecycle, including transcription elongation, splicing, export, and degradation [108,[110], [111], [112]]. How these and other RNA-binding proteins bind to RNAs in an RNA-processing-dependent or -independent manner remains largely elusive. One example in plants is the heterogeneous nuclear ribonucleoprotein (hnRNP) A1-like protein 1 (HLP1), an RNA-binding protein involved in transcription termination and polyadenylation. A pioneering study employing crosslinking and immunoprecipitation (CLIP) followed by high-throughput sequencing (CLIP-seq) [[113], [114], [115]] showed that HLP1 binds directly and preferentially to the 3′ end region of pre-mRNAs across the transcriptome, illustrating tight coupling between its binding and polyadenylation [116]. Another example is RZ-1B/1C, a redundant pair of glycine-rich RRM-domain-containing proteins that co-fractionate with chromatin, implying transcription or RNA-processing-coupled RNA binding. Enhanced-CLIP sequencing (eCLIP-seq) [117] revealed that RZ-1C preferentially binds to exons, and the loss of RZ-1B and RZ-1C leads to the nuclear accumulation of IR transcripts, suggesting that RZ-1B/1C may promote either nuclear RNA decay or CTS [18]. GRP7, another glycine-rich RNA-binding protein, was shown via individual-nucleotide-resolution CLIP sequencing (iCLIP-seq) [118] to bind to RNA broadly and regulate splicing events across the transcriptome [119]. RS31, a plant-specific SR protein, was demonstrated by iCLIP to bind to CU-rich and CAGA RNA motifs, thereby regulating alternative splicing events, particularly IR and exitron splicing, and modulating the activities of other splicing regulators in Arabidopsis [120]. SKRP (serine/lysine/arginine-rich protein), an RNA-binding protein that is associated with the spliceosome, was shown by eCLIP-seq to bind to the 3′ ends of unspliced exons. Thus, RNA processing coupled and position-specific RNA binding could be one of the predominant modes of RNA binding in plants [121].

A revealing aspect of co-transcriptional protein–RNA interactions concerns the degradation of RNA within the nucleus. Two major nuclear RNA decay pathways have been described in eukaryotes: the 3′–5′ decay pathway mediated by the RNA exosome, and the 5′–3′ decay pathway catalyzed by exonucleases. In mammals, a critical regulator of exosome activity is the RNA helicase MTR4; its homologs in Arabidopsis are known as MTR4 and HEN2 [122]. The main nuclear 5′–3′ exonuclease in mammals, XRN2, has two homologs in plants: XRN2 and XRN3 [123]. Recent studies in mammalian cells have revealed dynamic competition between nuclear RNA processing and nuclear RNA decay [[124], [125], [126], [127]]. Remarkably, not only defective transcripts but also some properly processed RNAs can serve as substrates for nuclear RNA decay [128]. For instance, mutants defective in RNA degradation pathways often accumulate both aberrant and normally processed transcripts. This raises a long-standing question in the field: What determines whether an RNA molecule is exported from the nucleus or degraded? A recent study provided intriguing mechanistic insight into this question. The zinc finger–containing RNA-binding protein ZFC3H1 was shown to associate with nascent RNA in a positionally biased manner, binding preferentially near the first exon/intron co-transcriptionally [127]. Depending on transcript features such as intron number and/or RNA-processing status, ZFC3H1 appears to undergo conformational changes: for short and intron-poor RNAs, it facilitates recruitment of the RNA exosome, leading to nuclear RNA decay; in contrast, for long and intron-rich transcripts, ZFC3H1 protects the transcripts from exosome-mediated decay. This finding highlights the importance of specific RNA–protein interactions as key mediators determining RNA fate. Furthermore, it suggests that RNA fate is not pre-fixed, but rather can be dynamically reprogrammed depending on the associated RNA-binding proteins and the RNA-processing status. Taken together, these findings underscore the urgent need for more studies in plants, particularly studies focused on RNA–protein interactions in the context of RNA processing, to fully elucidate how co-transcriptional protein–RNA packaging governs RNA fate.

7. Co-transcriptional RNA processing linking to the chromatin environment

Co-transcriptional RNA processing is not only crucial for determining the fate of RNA molecules but also plays an important role in regulating the chromatin environment of the gene locus via nascent RNA as an intermediate scaffold, allowing stable control of gene expression. A well-characterized example is the plant-specific de novo DNA methylation pathway known as RNA-directed DNA methylation (RdDM), which plays critical roles in repressing transposable elements and maintaining genome stability. RdDM is mediated by RNA polymerase IV (Pol IV) and RNA polymerase V (Pol V), two plant-specific, specialized forms of RNA polymerase [[129], [130], [131], [132], [133]]. Based on cryo-electron microscopy data, it is reported that Pol V likely exhibits very slow transcription elongation and frequent backtracking, which presumably leads to prolonged retention of its transcripts on chromatin [134]. These transcripts act as scaffolds to bind siRNAs and recruit DRM2, which catalyzes DNA methylation [[135], [136], [137], [138]]. In organisms lacking Pol V, similar siRNA-mediated gene silencing pathways, such as PIWI-mediated transposon silencing in the animal germline [75,[139], [140], [141]], utilize Pol II-mediated mechanisms analogous to RdDM [[142], [143], [144]]. These mechanisms reveal a seemingly paradoxical yet internally consistent principle: the transcription process itself is critical for maintaining transcriptional silencing via chromatin-based mechanisms.

The FLC gene represents a paradigm for dissecting the regulation of the chromatin environment mediated by Pol II transcripts and RNA processing. FLC encodes a MADS-box transcription factor that acts as a repressor of flowering. Detailed mechanisms of FLC regulation have been reviewed previously, so this review only focuses on the latest discoveries and is restricted to normal growth temperature [39,145,146]. The expression level of FLC during the seedling stage directly determines flowering time and the vernalization requirement. In summer annual Arabidopsis ecotypes such as Col-0, FLC is in an epigenetically repressed state, which is maintained through the cooperative activity of a group of RNA processing and transcription-related factors linked to histone methyltransferases, also known as the autonomous pathway [39,147]. Through genetic screens and protein interaction–based assays, various autonomous pathway components have been identified, which can be divided into four categories: A. RNA 3′ end processing regulators, including RRM domain–containing proteins FCA [148] and FPA [149]; FY, a homolog of yeast WD repeat-containing protein 33 (WDR33) that recognizes the polyadenylation signal [150]; FLOWERING LOCUS K (FLK) [151], a KH-domain containing RNA-binding protein; and cleavage stimulation factor 64 (Cstf64) and Cstf77, which serves as core components of the 3′ end processing machinery [152]. B. Pol II transcription elongation regulators, including the Pol II CTD Ser2P kinase CDKC; 2 (P-TEFb homolog) [153] and the CPF phosphatase regulated factors, ANTHESIS PROMOTING FACTOR 1 (APRF1), LUMINIDEPENDENS (LD), and TYPE ONE SERINE/THREONINE PROTEIN PHOSPHATASE 4 (TOPP4), which are involved in the dephosphorylation of Pol II and its elongation factor, making them critical for termination [74]. C. Splicing and RNA modification-related factors, such as the spliceosome component Prp8 [39,154], the m⁶A methyltransferase complex components FKBP12 INTERACTING PROTEIN 37KD (FIP37) and mRNA adenosine methylase (MTA) [155,156], and the R-loop stabilization factor Eoc1 [156,157]. D. Histone mark–modifying enzymes, including FLOWERING LOCUS D (FLD), an H3K4me1 demethylase [37,158]; SET DOMAIN GROUP 26 (SDG26), a catalytic activity uncertain protein that cooperates with FLD and LD [159,160]; SDG8, an H3K36me3 methyltransferase [161,162]; and components of polycomb repressive complex 2 (PRC2), which catalyzes H3K27me3 (Fig. 4).

Fig. 4.

Co-transcriptional RNA dynamics and chromatin regulation at the FLC locus. A Regulatory protein complexes that mediate co-transcriptional and chromatin control of FLC. Co-transcriptional regulation of COOLAIR by RNA splicing and modification-related proteins, Pol II transcription elongation regulators, and RNA 3′ end processing factors contributes to its proximal termination. This process facilitates the recruitment of chromatin modifiers, resulting in reduced H3K4me1 and enhanced H3K27me3 levels at the FLC locus, thereby establishing a repressive chromatin environment and silencing FLC transcription. B Close interplay between co-transcriptional RNA processing and chromatin changes at the FLC locus. Co-transcriptional regulation of COOLAIR integrates RNA-processing factors with chromatin modifications, contributing to the transcriptional repression of FLC.

The understanding of how these functionally diverse factors are integrated to effectively and stably repress FLC expression has been continuously evolving. In the wild-type Col-0 background, FLC is transcriptionally repressed, featuring low Pol II initiation and elongation rates, for not only the transcription of FLC sense strand but also the antisense strand, also known as COOLAIR [41]. The transcription of COOLAIR is particularly inefficient, the elongation of which presumably requires Pol II bearing Ser2P at its CTD and is associated with R-loop formation [156], m⁶A modification on nascent RNAs [163], and the usage of a proximal intron acceptor site [154]. These transcriptional features promote the binding of the nascent transcript by FCA. Simultaneously, through mechanisms including liquid–liquid phase separation, other RNA 3ʹ end processing factors are likely also recruited, ultimately leading to the termination and polyadenylation of COOLAIR at the promoter-proximal region [37,74,164]. This proximal termination or polyadenylation is coupled with the histone demethylase FLD, conferring low levels of H3K4me1 within the FLC gene body, while further promoting high levels of H3K27me3, likely indirectly by shutting down transcription [160,165]. Finally, the repressive chromatin environment ensures that both FLC and COOLAIR are transcribed at low levels at normal growth temperature. A long-standing question has been how COOLAIR transcription is linked to downstream histone modifications. Intriguingly, a recent study revealed that COOLAIR transcription termination requires a CPF-like module comprising APRF1, a homolog of yeast Swd2 and human WDR82. APRF1 interacts with TOPP4 (homologous to yeast Glc7/human PP1) and LD, which structurally resembles Ref2/PNUTS [74]. Importantly, LD directly interacts with FLD, suggesting a possible mechanism whereby during the slow elongation and termination of COOLAIR transcription, LD physically carries FLD, thereby modulating the chromatin landscape to achieve silencing [37,74,160]. This is in line with previous discoveries that FLD is preferentially localized to the 3ʹ ends of genes genome-wide and may function to limit intergenic transcriptional readthrough [165].

Notably, a recent discovery highlights that FLC sense nascent transcripts also play a role in governing the chromatin environment at this locus [163]. The plant-specific protein NEEDED FOR RDR2-INDEPENDENT DNA METHYLATION (NERD) binds to FLC nascent transcripts and represses its transcription. NERD interacts with the m⁶A writer complex and the H3K36me3 methyltransferase SDG8, resulting in the addition of m⁶A at FLC sense transcripts and reduced H3K36me3 deposition. These results highlighted that as a united locus, FLC sense transcripts may also exert a non-coding function like COOLAIR, an intriguing direction that currently remains largely unexplored.

Intuitively, the mechanism at FLC parallels several different scenarios discovered in other model systems, such as Pol V-mediated RdDM in plants and PIWI-interacting RNA-mediated silencing in Drosophila, all of which highlight the crucial role of slow transcription elongation in establishing chromatin-based gene silencing. Notably, in Drosophila, Pcf11 encodes a transcription termination factor whose N-terminal domain mediates liquid–liquid phase separation, a property critical for its function in gene silencing [75]. Remarkably, a Pcf11 mutant lacking such phase separation capability can be fully complemented by introducing the C-terminal region of FCA, while much less effectively by introducing the LCD domain of FUS, suggesting that condensates of Pcf11 or FCA may possess a yet-unclear functional specificity, which warrants further investigation.

8. Future perspectives

Research over the last few decades has revealed tremendous details and exciting principles in co-transcriptional gene regulation in plants. The transcription process itself holds great regulatory potential, and our current understanding is likely still at the juvenile stage. Looking ahead, lots of exciting questions remain to be answered. Within the nucleus, there are hundreds of RNA-binding proteins, generally with weak sequence specificity. What are the rules underlying their interactions with nascent RNAs during transcription, and what are their surrounding microenvironments and interacting partners during RNA binding? Among the numerous RNA-binding proteins, it would be particularly informative to investigate those that carry additional domains related to chromatin modification, Pol II, or RNA modifications, as well as those exhibiting temporal or spatial expression specificity. Functional studies of these proteins will bring important insight into co-transcriptional regulation. In addition, large-scale mapping of RNA-binding proteins using high-resolution techniques such as CLIP-seq [113,114,117,118] is required to answer the above questions in plants, especially when combined with approaches to interfere with transcription elongation and RNA processing.

Another intriguing aspect of co-transcriptional gene regulation lies in the dynamic nature of RNA structure. How does RNA fold co-transcriptionally, and how does such RNA folding respond to internal (e.g., chromatin state and Pol II elongation rate) and environmental cues? How are these structures remodeled to influence various RNA-processing steps? A recent report on COOLAIR has already revealed the potential functional importance of its RNA secondary structure, a combination of multiple conformations, during the vernalization response [166]. In-depth investigations into RNA structure in plants, as well as the development and use of novel technologies for probing RNA conformation, including artificial intelligence and cryo-electron microscopy, will likely yield exciting new discoveries in the future. In addition, integrating CLIP-seq data with transcriptome-wide RNA structure mapping (e.g., SHAPE-eCLIP [167]) and genetic perturbations will help dissect how RNA-binding preferences are modulated by RNA conformation and the cellular context.

Beyond that, a fundamental question remains: What are the core mechanisms and logic that govern RNA fate in plants? More specifically, how do RNA processing and the transcription process itself influence nuclear RNA degradation? Intriguingly, intron-retained transcripts are widespread in plants, but what is their fate? Are they predominantly degraded, or are some selectively processed and exported under specific conditions? Dissecting the molecular determinants, such as specific RNA-binding proteins, nuclear surveillance factors, or RNA modifications that distinguish transcripts destined for decay from those undergoing maturation will be critical. In particular, uncovering how nuclear RNA quality control mechanisms interpret these features to modulate RNA stability would be important. In addition, future studies on RNA metabolism, especially those integrating nucleoside analog labeling and single-molecule RNA imaging to track RNA dynamics, will likely provide valuable insights into these questions.

Finally, the role of co-transcriptional processing in shaping the chromatin environment remains a compelling topic. For FLC regulation, a particularly intriguing question is as follows: What unique attributes of FLC make numerous general regulators converge at this locus, resulting in effective gene silencing? More broadly, under what circumstances does transcription, together with its coupled RNA-processing events, become sufficient to trigger gene silencing? Additionally, what is the precise relationship between RNA-processing factors such as FCA and histone-modifying enzymes such as FLD? Given that FLD is also transcription-coupled, why is the presence of RNA-processing factors like FCA still necessary to confer chromatin silencing? Answering these questions will help unravel the unique mechanisms behind co-transcriptional gene regulation in plants.

CRediT authorship contribution statement

Mengshi Wu: Writing – original draft, Visualization, Validation, Conceptualization. Danling Zhu: Writing – review & editing. Zhe Wu: Writing – review & editing, Writing – original draft, Supervision, Conceptualization.

Declaration of competing interest

The authors declare that they have no competing interests.

Data availability

No data was used for the research described in the article.