P-bodies act as dynamic control hubs for RNA processing and storage

TL;DR

This study reveals that processing bodies (PBs) in yeast act as dynamic hubs that regulate mRNA storage and processing during nutrient stress.

Contribution

The study introduces a novel method to track mRNA localization and fate during glucose depletion, revealing distinct mRNA classes within PBs.

Findings

Preexisting mRNAs localize to PBs with varying kinetics during glucose starvation.

Most PB-localized mRNAs are stored temporarily and not degraded, challenging the traditional decay model.

Some transcripts accumulate in PBs over time, suggesting roles in post-stress adaptation.

Abstract

Processing bodies (PBs) are cytoplasmic granules that function in the cellular response to stress conditions by regulating mRNA metabolism. Initially, they were thought to represent sites of mRNA turnover, whereas more recent work points to a role in the storage of useful mRNAs. However, their exact intracellular role remains unclear. We used SH-linked alkylation for the metabolic sequencing of RNA (SH-linked alkylation for the metabolic sequencing of RNA) to study PB-localization and global mRNA fate during glucose depletion conditions that induce PB formation in yeast. This enabled us to differentiate newly synthesized and preexisting RNAs and to separately track mRNA synthesis and degradation. We show that preexisting mRNAs localize to PBs with differing kinetics with some transcripts localizing over the time course of glucose starvation and some transcripts localizing in a more dynamic manner. We identified a small number of transcripts that are enriched only transiently in PBs, consistent with the traditional view of PBs acting as sites for RNA decay. However, most transiently localized transcripts are not destabilized following glucose starvation, with PBs appearing to act as temporary storage sites for transcripts that later undergo alternative fates. For other transcripts, both their preexisting and newly made transcripts accumulate in PBs over the time course of glucose depletion, and we suggest that these transcripts are important for adaptation once the nutrient stress is relieved. Together, our data indicate a model where transcripts partition into different classes that behave differently following nutrient depletion with PBs acting as triage sites for mRNAs to direct their fate.

Figures

Full text

During periods of environmental change, cells must adapt their gene expression programs to optimize survival and maintain homeostasis. The regulation of RNA metabolism under such conditions involves various mechanisms that modulate transcription, translation, and RNA processing, enabling cells to reallocate their resources efficiently. Translational regulation enables cells to prioritize mRNAs necessary for survival while downregulating global translational activity (1). The localization of mRNAs within the cytoplasm can also influence their translation, storage, and degradation (2, 3). Ribonucleoprotein (RNP) granules play crucial roles in these cellular processes, particularly under conditions of stress, where they facilitate the regulation of gene expression, mRNA processing, and translational control.

RNP granules are complex cellular structures consisting of RNA molecules and associated protein components. They form through liquid-liquid phase separation where weak multivalent interactions between multidomain proteins and RNA form liquid-like droplets in cells (4). Prominent examples are RNA processing bodies (P-bodies, PBs) which are evolutionarily conserved, membrane-less compartments (5, 6, 7). PBs play important functional roles in cells by determining the fate of mRNAs, but their exact intracellular functions remain unclear. PBs were originally described as sites of mRNA decay, where mRNAs are decapped and degraded (8, 9), whereas other more recent studies have suggested that PBs act as storage sites where mRNAs are translationally repressed but not degraded (10).

The idea that PBs function in RNA decay comes from data indicating that PBs are enriched for proteins that function in mRNA turnover including, decapping, deadenylation, and exonucleolytic degradation (9, 11). Further studies have shown that inactivation of the Xrn1 exonuclease results in enlarged PBs and the accumulation of polyA RNAs which is thought to arise due to blocking mRNA decay after the decapping step (8, 9). Direct evidence for mRNA degradation has come from studies where the stabilization of a normally unstable mRNA by the introduction of a polyG tract into the mRNA 3′-UTR enabled the detection of mRNA degradation intermediates in PBs (8). In addition, fluorescent reporter constructs containing PP7 stem loops have been shown to localize to PBs where the signal decays over time suggesting that mRNA decay is occurring within PBs (12).

Evidence against a role in degradation has come from the finding that mRNA decay is unaffected in cells lacking detectable PBs suggesting that they are not essential for degradation, but this does not rule out a kinetic advantage for PB formation by acting as sites of accelerated mRNA decay for particular transcripts (10, 13). In addition, fluorescent biosensors used to visualize Xrn1-mediated mRNA decay of reporter mRNAs in HeLa cells were found not to be degraded when present in PBs (14). More recent studies examining the transcriptomes of PBs have suggested that PBs do not act as sites of mRNA decay but instead function in the storage of translationally repressed mRNAs. PBs have therefore been proposed to store selected mRNAs during stress conditions, such that these mRNAs can reenter the pool of mRNAs available for translation after exiting the condensate following stress recovery (15, 16).

Taken together, the existing data suggest that PBs could play roles in both mRNA decay and mRNA storage and that the biological context under which PBs form may be important. However, data identifying mRNA-specific fates in PBs is currently lacking. In this current study, we used SH-linked alkylation for the metabolic sequencing of RNA (SLAM-seq) (17) to study global mRNA metabolism during glucose depletion conditions. This strategy enabled us to differentiate preexisting mRNAs from newly synthesized mRNAs and to follow the kinetics of their localization to PBs formed in response to nutrient depletion conditions. Our data indicate that mRNA can be partitioned into different classes which behave differently following nutrient depletion, with PBs acting predominantly as sites for mRNA storage and triage, with a more minor transcript-specific role in mRNA decay. Intriguingly these data support a more nuanced role for PBs in the control of mRNA fate, such that PBs can have highly variable outcomes dependent upon the identity and presumably the RNA binding proteins (RBPs) bound to a specific mRNA.

Results

PB-localized transcripts exhibit differential enrichment patterns

PBs are commonly induced following the imposition of nutrient depletion conditions such as glucose starvation. To determine fate of mRNAs during these conditions, we used a noninvasive metabolic labeling method to follow new mRNA synthesis and to examine the fate of preexisting mRNAs in cells. 4-Thiouracil (4TU) is a sulfur-substituted nucleotide precursor that can be used to biosynthetically label RNAs without affecting cellular growth or gene expression (12, 17, 18). 4TU is taken up by cells and converted into 4-ThioUTP enabling labeling of newly synthesized cellular RNAs and detection by deep sequencing. For our experiments, cells were grown to exponential phase before switching to media lacking glucose (Fig. 1A). At the same time, 4TU was added to the cultures to label newly synthesized transcripts and PBs were purified and RNAs extracted at 2.5, 5, 10, 20, and 40 min following glucose depletion using our established sedimentation and immuno-affinity based-approach (19). SLAM-seq was performed to follow the fate of the preexisting transcripts that localize to PBs during glucose depletion.Figure 1Preexisting mRNAs localize to PBs with different kinetics. A, schematic of PB purification and SLAM-seq methodology. Cells were grown to exponential phase before switching to media lacking glucose and containing 4TU. Following cross-linking using formaldehyde, whole cell extracts were prepared from Dcp1p-myc tagged strains. Clarified cell extracts were made using a gentle 1000g centrifugation step to remove cell debris and any unbroken cells. An initial centrifugation step (20,000g) was then used to enrich high molecular weight (HMW) complexes and PBs were isolated by immunoprecipitation of Dcp1. PBs were purified and RNAs extracted at 2.5, 5-, 10-, 20-, and 40-min following glucose depletion. B, transcripts localize to PBs following two different behavior patterns defined as PB_dynamic and PB_accumulated transcripts. PB-localized transcript abundances were regressed to either an asymmetric gaussian curve (PB_dynamic transcripts) or an exponential growth curve (PB_accumulated transcripts) which identified 1669 PB_dynamic transcripts and 442 PB_accumulated transcripts, respectively. C, example transcripts are shown that peak at 2.5 (RPC37), 5 (MUP3), 10 (TPI1), and 20 (ADK2) minutes glucose depletion or that show broad peaks (MET16 and RPS28B). Examples of PB_accumulated transcripts include SPG4 and SUE1. 4TU, 4-Thiouracil; PBs, processing bodies; SLAM-seq, SH-linked alkylation for the metabolic sequencing of RNA.

We found that preexisting transcripts localize to PBs follow two different behavior patterns (Fig. 1B). A first group of transcripts was enriched in PBs with a distinct peak of accumulation followed by a decrease in PB-localization, hereafter referred to as “PB_dynamic” transcripts. A second group of transcripts show increasing PB-localization over the time course of glucose depletion and is referred to as “PB_accumulated” transcripts. To further analyze these different classes of transcript, changes in PB-localized transcript abundance were regressed to appropriate mathematical models selecting the best fit representing either dynamic or accumulation behavior; an asymmetric gaussian curve (PB_dynamic transcripts), or an exponential growth curve (PB_accumulated transcripts). This identified 1669 PB_dynamic transcripts and 442 PB_accumulated transcripts, respectively (Fig. 1B and Table S1).

The peak of PB_dynamic transcript accumulation in PBs was found to vary over the time course of glucose depletion. Many PB_dynamic transcripts peaked within 10 min of glucose starvation, with smaller numbers of transcripts peaking at later time points; example transcripts are shown in Figure 1C that peak at 2.5 (RPC37), 5 (MUP3), 10 (TPI1), and 20 (ADK2) minutes following glucose depletion. Other transcripts showed broader peaks that spanned more than one time point. For example, MET16 and RPS28B showed a broad peak over the first 10 minutes of glucose depletion (Fig. 1C). RPS28B has previously been identified as a scaffolding mRNA in PBs (20). The PB_accumulated transcripts showed highest enrichments following 20- or 40-min glucose starvation consistent with sustained accumulation in PBs rather than the transient accumulation seen for PB_dynamic transcripts (Fig. 1C). Examples include SPG4 and SUE1 (Fig. 2B), which have previously been characterized as late-phase mRNAs that accumulate in PBs after an extended period of glucose starvation (21).Figure 2RNA presence in PBs is associated with key biophysical and functional characteristics. A, functional categorization of proteins encoded by mRNAs that localize to PBs with different kinetics. The top 10 GO slim categories are shown. Dots are scaled and colored according to the significance of enrichment (p-adjust), with the enrichment of proteins within each category shown as GeneRatio taken from the complete analysis included in Table S2. B, box plots are shown comparing the median polyA tail length for mRNAs enriched in the PB_dynamic or PB_accumulated datasets and for those not in PBs. C, comparison of abundance. D, comparison of RNA secondary structure (Gini index). E, comparison of RNA secondary structure (PARS score) in CDS. F, comparison of RNA secondary structure in 5′UTRs (PARS score). G, comparison of RNA secondary structure (PARS score) in 3′UTRs. All analyses are shown relative to data from the whole transcriptome as background control. Statistical significance was assessed using two-sided Mann–Whitney U tests (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001). CDS, coding sequence; GO, Gene Ontology; PARS, Parallel Analysis of RNA Structure; PBs, processing bodies.

The PB_dynamic transcripts were divided into early (2.5 and 5 min) and late (10 and 20 min) transcripts and examined for functional enrichment of Gene Ontology (GO) terms (Fig. 2A and Table S2). The early transcripts were enriched for broad GO categories related to translation including ribosome biogenesis and rRNA/tRNA metabolism including processing and maturation. The PB_dynamic transcripts that enriched at later time points included GO categories related to organic-acid metabolism including amino acid and fatty acid metabolism (Fig. 2A). The PB_accumulated transcripts were enriched for GO categories related to metabolism including carbohydrate metabolism, energy reserve metabolism and glycogen metabolism (Fig. 2A).

RNA presence in PBs is associated with key biophysical characteristics

We examined the PB-localized transcripts to see whether they share similar biophysical properties. Previous studies have provided a snapshot of the transcripts that localize to PBs at single time points following stress-induction. For example, the polyA tail lengths of the mRNAs that localize to PBs formed in response to glucose depletion were found to be shorter than the general transcriptome average (19). We similarly found that the average polyA tail length of PB_dynamic transcripts is significantly shorter when compared with mRNAs across the transcriptome (Fig. 2B). However, the average polyA tail length of the PB_accumulated transcripts is significantly longer than the transcriptome average. Shorter polyA tail sizes have been associated with abundant mRNAs that are well-translated mRNAs across eukaryotes (22). We found a correlation between polyA tail length and transcript abundance, since the PB_dynamic transcripts that had shorter polyA tail lengths, were more abundant, and the PB_accumulated transcripts that had longer polyA tails, were less abundant compared with the transcriptome average (Fig. 2C)

The mRNAs that localize to PBs formed following 10 minutes of glucose depletion have been reported to be significantly longer and more structured when compared with mRNAs across the transcriptome (19). This is similar to the RNAs that have been shown to phase separate in vitro, which tend to be longer structured RNAs that have a propensity to self-assemble into condensates (23, 24). To assess mRNA structure, we used a transcriptome-wide dataset that measured the propensity of mRNAs to form secondary structure using a dimethyl sulfate (DMS)-seq approach (25). The PB_accumulated transcripts had a higher Gini index score compared with the wider transcriptome, a metric indicating a higher propensity to contain structured regions (Fig. 2D). In contrast, transcripts in the PB_dynamic group were modestly less structured. As a second approach, we used a dataset that measured the probabilities of base pairing at single nucleotide resolution (26). The advantage of this approach is that it provides a measure of structure, the PARS score (parallel analysis of RNA structure, PARS), for each coding sequence (CDS), 5′- and 3′-UTR regions of mRNAs. This suggests that the increased structure of PB_accumulated transcripts is due to their 5′-UTR regions, whereas the decreased structure of PB_dynamic transcripts is due to their 3′UTRs regions (Fig. 2, E–G). We did not find any significant differences in 5′UTR or 3′UTR length but the CDS length was significantly longer for PB_accumulated transcripts (Fig. S1, A–C).

Taken together, our data indicate that transcripts that localize with PB_dynamic kinetics tend to be more abundant transcripts with less structure and shorter polyA tails, whereas transcript that localize with PB_accumulated kinetics tend to be less abundant transcripts with more structure and longer polyA tails. We next considered the fate of these mRNAs in cells during glucose depletion conditions.

The stability of most mRNAs is increased in response to glucose depletion

Our analysis of PB-localized transcripts identified two mRNA classes that either localize to PBs with a transient peak of accumulation or show increasing accumulation over the time course of glucose depletion. PB_accumulated transcripts may be transcripts that are stored in PBs for subsequent release once the stress is relieved. The kinetics of PB_dynamic transcript localization may indicate that some mRNAs enter PBs to be degraded, or alternatively, mRNAs may enter and exit PBs in a more dynamic manner. To differentiate these possibilities, we examined total mRNA stability following glucose depletion. We reasoned that if transcripts are localized to PBs to facilitate degradation or storage, then this will be reflected by what happens to total intracellular transcript abundances following glucose depletion. For these experiments, cells were grown to exponential phase before switching to media in the presence or absence of glucose. Following 4TU labeling and total RNA extraction, SLAM-seq was performed to follow the turnover of preexisting transcripts during normal or glucose depletion conditions (Fig. S2A).

Analysis of the transcriptome during normal glucose replete conditions revealed a median mRNA half-life of 4.11 min (Fig. 3A), comparable to what has been reported previously for the yeast transcriptome using SLAM-seq technology (12). Following glucose depletion, there was a significant shift in mRNA half-lives and the median mRNA half-life increased to 12.39 min (Fig. 3A). Comparing the half-lives of mRNAs under normal and glucose depleted conditions revealed that most transcripts showed increased stability in response to glucose depletion with many fewer decreasing in stability (Fig. 3B). This is similar to a previous study reporting that most cellular mRNAs are stabilized following an abrupt shift from glucose to galactose media (27). We found that 3056 transcripts showed increased stability following glucose depletion, and the stability of a further 1023 transcripts was unaffected by glucose depletion (Fig. 3C and Table S1). In contrast, just 188 transcripts had decreased stability (Fig. 3C). For these 188 transcripts, degradation appeared to occur with differing kinetics, with some mRNAs showing a rapid decrease in abundance within 2.5 min and other transcripts decreasing in abundance at later time points. It should be noted that a limited number of decreasing transcripts unexpectedly increased in abundance at later time points—a so-called “rebound” (Fig. 3C). We ascribe this to a potential artifact arising from new synthesis of transcripts utilizing unlabeled uracil released from degraded mRNAs. In contrast, this effect is less pronounced in transcripts with increased stability which appear to display more consistent increases in transcript abundance (Fig. 3C).Figure 3The stability of most mRNAs is increased in response to glucose depletion. A, average mRNA half-lives were determined during normal glucose replete conditions (blue) and following glucose depletion (red). B, changes in mRNA half-lives were determined following glucose depletion. Transcripts that significantly increased in stability (blue) or decreased in stability (red) are highlighted. C, heatmaps showing 3056 transcripts that increased in stability, 1023 transcripts that were unaffected, and 188 transcripts that decreased in stability following glucose depletion. D, the top 10 GO categories are shown for mRNAs that decreased, increased, or were unaffected in stability by the switch to glucose starvation. Dots are scaled and colored according to the significance of enrichment (p-adjust), with the enrichment of proteins within each category shown as GeneRatio taken from the complete analysis included in Table S2. GO, Gene Ontology.

Analysis of biophysical characteristics revealed that the transcripts with decreased stability were enriched for mRNAs that initially had lower abundances (Fig. S2B), shorter polyA tail lengths (Fig. S2C), less secondary structure (Fig. S2D), and shorter CDS lengths (Fig. S2F). The transcripts with increased stability also had moderately lower abundances compared to the global RNA pool (Fig. S2B) but had longer polyA tail lengths (Fig. S2C) more secondary structure (Fig. S2D) and longer CDS lengths (Fig. S2E). We did not find any significant differences 5′ or 3′UTR length for the transcripts with increased or decreased stability (Fig. S2, F–G).

In response to glucose starvation, yeast cells shut down energy requiring processes and activate other metabolic pathways that enable a switch from fermentation to respiration using available carbon sources such as ethanol, amino acids and lipids (28, 29). Although relatively few transcripts were destabilized following glucose starvation, we found that they were enriched for transcripts that would be expected to be downregulated during glucose starvation conditions from their associated GO terms (Fig. 3D and Table S2). This included enzymes related to amino acid biosynthesis and carbohydrate metabolism including glycolysis, glycerol synthesis, and glycogen synthesis. Other destabilized transcripts encoded ribosomal proteins in agreement with the idea that the decay of these transcripts is increased in response to glucose depletion (30). Arguably, GO analysis of the transcripts with increased stability may be less informative since more than half the transcriptome appears to be stabilized following glucose starvation conditions. Indeed, these transcripts were enriched for broad categories related to cell division and chromosome organization, including DNA repair, chromatin remodeling and regulation of the cell cycle (Fig. 3D). For the transcripts where stability was unaffected by the switch to glucose starvation, we found that the enriched categories were mainly related to translation including ribosome biogenesis, rRNA processing, and RNP complex assembly (Fig. 3D).

There is no strong correlation between mRNA degradation and PB-localization

Having determined how preexisting transcripts localize to PBs, we next examined whether there is any correlation between PB localization and mRNA fate following glucose depletion. The kinetics of PB_dynamic transcript localization may indicate that some mRNAs enter PBs to be degraded. However, given that only 188 transcripts were found to decrease in abundance following glucose depletion, if PBs function in mRNA degradation then it must be for a relatively small number of transcripts. The PB_dynamic, PB_accumulated and nonlocalized transcripts (PB_Less) were partitioned into transcripts where abundance is increased, decreased, or remained the same following glucose depletion (Fig. 4A). We did not find any strong enrichment of mRNAs with decreased stability in PBs; 86 destabilized transcripts localized to PBs (64 PB_dynamic plus 22 PB_accumulated) compared with 92 transcripts that showed no PB-localization (Fig. 4A). The fact that transcripts that were unaffected or increased in stability were also found to localize to PBs with both PB_dynamic and PB_accumulated profiles also argues against a primary function for PBs in mRNA turnover (Fig. 4A).Figure 4PB localization does not correlate with mRNA degradation. A, partitioning of PB_dynamic, PB_accumulated, and nonlocalized transcripts (PB_Less) into transcripts where abundance is increased, decreased, or remains the same following glucose depletion. B, an lsm4ΔC edc3Δ mutant is defective in PB formation induced by glucose depletion. Representative epifluorescent microscopic images are shown for the wild-type and lsm4ΔC edc3Δ mutant expressing DCP1-GFP as a PB marker. The scale bar represents 4 μm. C and D, fate of PB_dynamic mRNAs in a mutant that cannot make PBs. Preexisting transcript levels and transcript localization to PBs is shown for destabilized (RPS28B, YPT31, and PYP1) and stabilized (TRS130, PRP38, and SPR6) transcripts. Total mRNA abundance was quantified using qPCR in the wild-type and lsm4ΔC edc3Δ mutant during glucose replete (+D) and following the shift to glucose starvation conditions (−D). E, box plots are shown comparing TEs determined from unstressed cells or cells following glucose depletion for mRNAs that display PB-dynamic kinetics and decrease, increase or are unaffected in stability following glucose depletion. F, box plots are shown comparing Pbp1-mRNA association as a marker for stress granule localization for mRNAs that display PB-dynamic kinetics and decrease, increase, or are unaffected in stability following glucose depletion. Statistical significance was assessed using two-sided Mann–Whitney U tests (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001). PBs, processing bodies; qPCR, quantitative real time PCR; TE, translational efficiency.

To further examine whether mRNA degradation correlates with PB formation, we examined total transcript levels for selected mRNAs in an lsm4ΔC edc3Δ mutant that is defective in PB formation (31). We first confirmed that the lsm4ΔC edc3Δ mutant abrogates PB formation in our yeast strain and under our experimental conditions (Fig. 4B). We chose three transcripts (RPS28B, YPT31, and PYP1) that localized with PB-dynamic kinetics and with decreased half-lives following glucose depletion, suggesting they are actively degraded (Fig. 4C). When quantitative real time PCR (qPCR) analysis was used to measure total transcript levels following glucose depletion, we found that these destabilized transcripts showed increased mRNA abundances in the lsm4ΔC edc3Δ mutant consistent with PBs being required for mRNA turnover (Fig. 4C). For comparison, we examined three transcripts (TRS130, PRP38, and SPR6) that displayed PB-dynamic kinetics and were stabilized following glucose depletion (Fig. 4D). In contrast, we found that these stabilized transcripts showed decreased mRNA abundances in the lsm4ΔC edc3Δ mutant consistent with PBs playing a protective, storage role following glucose starvation (Fig. 4D). Although we cannot rule out that the loss of mRNA decay factors such as Lsm4 and Edc3 might stabilize transcripts independently of PB formation, the decreased transcript levels for these more-stable transcripts is consistent with PBs functioning in stabilizing PB-dynamic transcripts during glucose depletion.

Given that PBs are thought to function in storing nontranslating mRNAs (32), we compared their translational activities and localization to stress granules (SGs) following glucose starvation. For translational activity, we used our previous translational efficiency (TE) data that has been obtained using ribosome profiling of mRNAs under normal glucose replete conditions and following 10 min glucose depletion (19). We found that for transcripts where stability was decreased or unaffected in response to glucose depletion, there was also a concomitant decrease in TE as might be expected for transcripts that are degraded or no longer required during the stress conditions (Fig. 4E). The stabilized transcripts were increased in TE (Fig. 4E) suggesting that some PB dynamic transcripts may exit PBs to return to the pool of actively translating mRNAs during glucose depletion conditions. This could act to fine-tune the temporal control of gene expression to time points in the stress recovery period when a protein is specifically required. An alternative fate for PB-localized transcripts is that they may exit PBs and localize to SGs. SG association was previously assessed using RIP-seq data obtained using Pbp1 as a canonical SG marker (19). Comparing our current data with this SG association data (19) revealed that the stabilized transcripts were enriched in Pbp1-association, whereas the destabilized and similar stability transcripts were not, suggesting that some of the transcripts that localize to PBs with dynamic kinetics may exit and localize to SGs. We anticipate that this may be for longer term storage compared with PBs, with the latter acting as dynamic, short term triage centers.

mRNA synthesis rates are strongly altered following glucose depletion

Having examined the multiple fates of preexisting mRNAs, we next turned our attention to the fate of newly synthesized transcripts. Glucose depletion is known to significantly affect gene expression via glucose repression/derepression signaling to activate pathways of alternative carbon source utilization and oxidative metabolism that likely facilitate adaptation to the starvation condition (33). We first used our SLAM-seq data to determine what happens to the synthesis of transcripts during the time course of glucose depletion (Fig. S2A). There was a significant shift with many transcripts showing altered synthesis rates following glucose depletion (Fig. 5A). For example, some transcripts such as the ENO1 enolase transcript (Fig. 5B) were present at low levels during glucose replete conditions, and their synthesis was induced in response to glucose depletion. This might be expected since the expression of the ENO1 mRNA is thought to be more readily associated with gluconeogenesis then its paralogue ENO2, which is primarily associated with glycolysis (34). Therefore, in the context of glucose depletion and preparation for growth on ethanol, induction of gluconeogenic enzymes would be anticipated. For other transcripts such as RPS28B, the rate of synthesis was strongly reduced in response to glucose starvation (Fig. 5C).Figure 5Alterations in mRNA synthesis rates following glucose depletion. A, changes in mRNA synthesis were determined following glucose depletion. Transcripts that significantly increased (blue) or decreased (red) in synthesis are highlighted. Example transcripts that increase in synthesis B. (ENO1) or decrease in synthesis C. (RPS28B) are shown during glucose replete (+D) and following the shift to glucose starvation conditions (−D). D, heatmaps showing 3376 transcripts that decrease, 564 transcripts that increase, and 1489 transcripts that show similar synthesis rates following glucose depletion. Box plots are shown comparing abundance (E) and polyA tail length (F) for transcripts that increase, decrease, or are unaffected in synthesis following glucose depletion. G, comparison of the enrichment of mRNAs that increase, decrease or are unaffected in synthesis with seven unique mRNA cohorts that were identified via differential engagement with the 43S complex, the closed-loop complex and the mRNA decay pathway (36) colored by the adjusted p value of enrichment. H, the top 10 GO categories are shown for mRNAs that decreased, increased, or were unaffected in synthesis following the switch to glucose starvation. Dots are scaled and colored according to the significance of enrichment (p-adjust), with the enrichment of proteins within each category shown as GeneRatio taken from the complete analysis included in Table S2. Statistical significance was assessed using two-sided Mann–Whitney U tests (∗p < 0.05, ∗∗p < 0.01, ∗∗∗p < 0.001). GO, Gene Ontology.

In total, synthesis decreased for 3376 transcripts compared with just 564 transcripts that increased following the shift to glucose starvation conditions (Fig. 5D and Table S1), with a further 1489 transcripts unaffected (Fig. 5D). Previous studies have reported a reduction in total mRNA abundances during glucose starvation but not distinguished whether this effect arose from increased mRNA decay or decreased RNA synthesis (35). Our study directly addresses this question, demonstrating that the reduction in total mRNA abundance during glucose starvation conditions is driven predominantly by decreased RNA synthesis rather than increased RNA decay.

Comparing the properties of the transcripts that showed alterations in synthesis in response to glucose depletion conditions revealed that the upregulated transcripts as a whole were originally of lower abundance during normal glucose-replete conditions (Fig. 5E), and with longer polyA tail lengths compared with the downregulated transcripts (Fig. 5F). We have previously defined a set of seven distinct mRNA clusters, based on their interaction profiles with translation factors and RNA binding proteins (36). The transcripts with decreased synthesis rates in this study following glucose repression are significantly enriched in cluster 4 (Fig. 5G). Cluster 4 encompasses mRNAs that are heavily translated and responsible for producing a major part of the cellular proteome during normal nutrient-rich conditions. Taken together, these data indicate that while the synthesis of most transcripts remains the same or is decreased in response to glucose depletion, there are a significant number of transcripts which are normally present at low abundances during normal glucose replete conditions that are upregulated and synthesized at higher rates during glucose starvation conditions.

Glucose is the preferred energy source in yeast and acts to repress the expression of genes encoding products required for respiration and the metabolism of alternative carbon sources (37). GO analysis revealed that the transcripts that increased in synthesis following glucose starvation included several categories related to alternative carbon metabolism including respiration, glycogen metabolism, gluconeogenesis, carbohydrate transport, and fatty acid catabolism (Fig. 5H and Table S2). This is in agreement with the known targets of glucose-regulated transcription factors that are induced in response to glucose depletion (35). The downregulated transcripts included categories related to growth and processes that generally have a high energy demand, such as ribosome biogenesis, translation, the cell cycle, DNA replication, and transcription (Fig. 5H and Table S2).

Some newly synthesized mRNAs localize to PBs

We next turned our attention to the newly synthesized transcripts during glucose starvation conditions and examined whether they localize to PBs. Surprisingly, we found that 214 do localize to PBs following glucose starvation conditions (Fig. 6A and Table S1). It is unclear why cells would expend energy on making transcripts that localize to PBs during glucose starvation conditions. Most of these transcripts appeared to accumulate in PBs over the glucose depletion time course, similar to the preexisting transcripts that localized with PB_accumulated kinetics. Indeed, 144 of the 214 transcripts overlap with the PB_accumulated transcript set, with 24 overlapping with the PB_dynamic transcripts (Fig. 6B). When we examined for any enrichments for GO categories, we found that the newly made transcripts that localize to PBs are enriched for similar categories as for the PB_accumulated transcripts including carbohydrate metabolism, energy reserve metabolism and glycogen metabolism (Fig. 6C and Table S2).Figure 6Newly made mRNAs localize to PBs. A, heatmaps showing newly made transcript localization to PBs. B, Venn diagram showing comparison of newly made transcripts that localize to PBs with PB_dynamic, PB_accumulated and PB_less transcripts. C, The top 10 GO categories are shown for newly made mRNAs that localize to PBs. Dots are scaled and colored according to the significance of enrichment (p-adjust), with the enrichment of proteins within each category shown as GeneRatio taken from the complete analysis included in Table S2. Synthesis following glucose depletion (D), newly made transcript localization to PBs (E), preexisting transcript localization to PBs (F) and total mRNA abundance quantified using qPCR in the wild-type and lsm4ΔC edc3Δ mutant following the shift to glucose starvation conditions (G) is shown for four newly made transcripts (COX26, DDR2, AZF1, and GAC1) that localize to PBs. Data are shown during glucose replete (+D) or following the shift to glucose starvation conditions (−D). GO, Gene Ontology; PBs, processing bodies; qPCR, quantitative real time PCR.

To test whether PB localization serves a protective function storing transcripts during stress conditions, transcript levels were examined in the lsm4ΔC edc3Δ mutant that is defective in PB formation. For this analysis, we selected COX26, DDR2, AZF1, and GAC1 as examples of transcripts where increased synthesis in response to glucose starvation was observed (Fig. 6D). These newly made transcripts localized to PBs during glucose starvation conditions (Fig. 6E) and their corresponding preexisting transcripts localized to PBs with PB-accumulated kinetics (Fig. 6F). When qPCR analysis was used to measure total transcript levels following glucose depletion, we found that these transcripts increased in abundance in the wild-type strain, whereas mRNA abundances were strongly reduced in the lsm4ΔC edc3Δ mutant consistent with PBs playing a role in mRNA storage and stabilization. The yeast environmental stress response encompasses the coordinated transcriptional and translational adjustment of multiple transcripts in response to diverse stress conditions (38, 39). Many commonly induced transcripts are not required for the immediate response to specific stress conditions and one possibility is that sequestering these transcripts into PBs may provide a storage mechanism such that further energy is not expended translating these mRNAs before they are required.

Discussion

PBs are dynamic, heterogeneous condensates that function in posttranscriptional gene regulation. Various studies have suggested that PBs function as platforms for mRNA decay across eukaryotes (8, 11, 40), whereas other reports emphasized a storage/repression role where mRNAs are translationally repressed in PBs (10, 41). Hence, PBs do not appear to be solely sites of mRNA decay nor exclusively storage/repression compartments, so their exact intracellular role remains unresolved. In this current study we have taken a SLAM-seq approach to examine the fate of both preexisting and newly synthesized mRNAs during glucose starvation conditions that induce PB formation. We found that mRNAs localize to PBs with differing kinetics following nutrient depletion with some preexisting transcripts localizing to PBs over the time course of glucose starvation and some transcripts localizing in a more dynamic manner. This observation agrees with our previous work investigating different phases of mRNA localization to PBs using the MS2 system to study single mRNAs in live cells (21). By globally quantifying changes in transcript stability following glucose starvation, we reveal that PBs do not appear to function in bulk mRNA decay as relatively few transcripts are destabilized under the conditions that promote PB formation. Taken together, our data indicate that PBs play a more dynamic role in mRNA metabolism including, but not limited to, acting as sites for targeted mRNA decay, acting to store mRNAs which may shield them against degradation as well as sequestering them away from active translation and acting as intermediary sites where mRNAs are transited to other RNP granules such as SGs.

Previous studies have identified the cohort of mRNAs that localize to PBs in mammalian systems and yeast cells following different stress conditions (10, 19, 42). We have now defined the kinetics of PB localization and identified two classes of transcripts that localize with transient (PB-dynamic) or continuing (PB_accumulated) kinetics. We hypothesized that the PB_dynamic transcripts are the prime candidates for mRNA degradation in PBs. However, subsequent analysis of global transcript stabilities showed a different picture, with relatively few mRNAs (<5%) decreasing in stability during the conditions that induce PB formation suggesting that PBs do not play a major role in global mRNA turnover. This is perhaps not surprising since it has been estimated that only approximately 10% of cytoplasmic mRNAs partition into PBs and SGs during stress, although this can be higher for some individual mRNAs (43, 44). It is possible that PBs act to degrade a limited set of transcripts, and we found that abrogating PB formation in a lsm4ΔC edc3Δ mutant resulted in increased mRNA abundance for some PB_dynamic transcripts that were normally destabilized. Furthermore, the PB_dynamic transcripts that were degraded included mRNAs encoding products that would not be expected to be required during glucose starvation conditions such as products involved in amino acid biosynthesis, glycolysis, glycerol synthesis, and tRNA aminoacylation.

Most of the mRNAs that dynamically localize to PBs are unaffected or increase in stability during the PB-inducing nutrient depletion conditions. There are several potential explanations for why transcripts might be transiently enriched in PBs, with PBs essentially acting as triage sites determining mRNA fate. RBPs have been extensively linked with RNP granule formation, facilitating the complex multivalent RNA-protein interactions that underpin PB assembly (6, 45, 46). One possibility, therefore, is that mRNAs bound to RBPs localize to PBs as “passenger” molecules before being released from their corresponding RBPs once in the condensate environment. These released mRNAs may be returned to the pool of actively translating mRNAs and we found that the PB_dynamic mRNAs that show increased stability following glucose starvation are enriched for mRNAs that show higher translational efficiencies during these conditions. These mRNAs may undergo short-term storage and are then required at specific points in the adaptation process. Alternatively, PB_dynamic transcripts may be transitioned to RNP granules such as SGs. SGs are primarily thought to function in the storage of translationally arrested mRNAs, and their formation has been linked with PB formation (47, 48). These mRNAs may be subject to longer term storage in SGs where they interact with translation factors and are hence primed for translation initiation once the stress is resolved. We found that the PB_dynamic mRNAs with increased stability following glucose starvation are enriched in SGs adding support to this idea.

The advantage of our SLAM-seq approach is that this approach enabled us to compare the fate of both preexisting and newly made transcripts separately. Interestingly, we found some transcripts where both the preexisting pool of transcripts, as well as the corresponding newly made transcripts, localize to PBs during glucose starvation conditions. Initially, it seems counter-intuitive to actively synthesize transcripts that are then localized to PBs. Glucose starvation has long been known to cause large-scale changes in gene expression (38, 39), while concomitantly causing a global inhibition of translation (49, 50). Localizing these mRNAs to PBs may therefore provide additional levels of regulation whereby mRNAs can be sequestered away from the translatome until such time as the stress is removed and active translation is required. Similarly, previous studies have identified certain mRNAs that have the ability to reenter the translational pool following relief from stress (35). We also found that PB_accumulated transcripts are enriched for transcripts with increased secondary structure within their 5′UTRs. Given that mRNA secondary structure content has been extensively linked with TE and regulation (51, 52), this aligns with the idea that these transcripts are subject to translational regulation. Furthermore, these mRNAs encode products required for carbohydrate and alternative carbon source metabolism that would be required for cells to adapt to growth conditions lacking glucose and any resumption of growth. These transcripts, may therefore, be synthesized and stored such that they are available following the initial response to glucose starvation where growth is slowed and gene expression is reprogrammed to adapt to the metabolism of alternative carbon sources.

It is unclear what factors may dictate why certain newly made transcripts localize to PBs. It is tempting to speculate that certain mRNP interactions that occur during their synthesis maybe important for directing newly synthesized mRNAs to PBs. For example, there are several precedents for cotranscriptionally added factors that subsequently control cytoplasmic localization and TE (53, 54). The addition of RBPs to newly made RNAs could therefore occur in the nucleus and act to control mRNA fate, in this case localization to PBs. In contrast to our findings made during glucose depletion conditions, a recent study found that newly synthesized mRNAs escape sequestration into condensates during high temperature conditions that induce SG formation (55). These transcripts are actively translated and the timing of their production is thought to prevent condensation. Similarly, the timing of new gene expression has been proposed to account for mRNAs that avoid translational repression during glucose depletion conditions (30). Our data suggest that this response is not rigid and there is also a subset of newly made transcripts that localize to PBs during glucose depletion that are not immediately translated.

Overall, our studies add to the idea that cells utilize PBs to manage their transcriptome under conditions such as stress where global translation of mRNAs is repressed. The mRNA content of PBs is known to vary depending on the stress condition used suggesting that PB localization acts in stress-specific adaptation mechanisms to respond to different stresses (19, 42). We found that most of the transcripts that localize to PBs show increased stability during the conditions that promote PB formation consistent with PBs actively sequestering these mRNAs during the adaptive response to the stress. This adaptive mechanism allows cells to alter their gene expression programs while triaging mRNAs in PBs before resuming normal protein synthesis once conditions improve. Our data also highlight the importance of PBs in posttranscriptional regulation and stress recovery. Rather than simply sequestering mRNAs that are no longer needed for translation, PBs appear to play more nuanced roles effectively allowing for a more strategic use of resources. The dynamic nature of PBs also allows them to disassemble once cellular homeostasis is achieved, enhancing their adaptability to changing stress levels.

Experimental procedures

Yeast strain and growth conditions

All Saccharomyces cerevisiae strains used were derivatives of W303-1A (MATa ura3-52 leu2-3 leu2-112 trp1-1 ade2-1 his3-11 can1-100). Dcp1-Myc tagged strains (19) and lsm4ΔC edc3Δ mutant strains (56) have been described previously. Cells were cultured in synthetic complete dextrose (SCD) media (2% w/v glucose, 0.67% (w/v) yeast nitrogen base with ammonium sulfate). SCD media were supplemented with standard amino acid concentrations (Formedium) except for uracil which was present at 10 mg/L for 4TU labeling. For glucose starvation conditions, cells were switched to SC media lacking glucose.

4TU labeling and total RNA purification

Cells were grown to exponential phase in SCD media (A600 = 0.6–0.8), before being harvested, washed, and resuspended in prewarmed SC or SCD media. Newly synthesized RNA was labeled by treating cells with 2 mM 4TU (Sigma-Aldrich). Cultures were maintained at 30 °C with shaking and samples taken at 2.5, 5, 10, 20, and 40 min. At each time point, 50 ml of culture was pelleted, snap-frozen in liquid nitrogen, and stored at −80 °C. RNA was isolated using TRIzol (Life Technologies) as previously described (57). Pellets were disrupted with glass beads (Mini-Beadbeater, BioSpec; 4 × 20 s bursts, with 30 s cooling intervals), followed by phenol–chloroform extraction. RNA was precipitated with isopropanol and GlycoBlue (Invitrogen), washed with 75% ethanol, and dissolved in RNase-free water. RNA samples were treated with DNase I at 37 °C for 10 min and purified by phenol–chloroform extraction followed by ethanol precipitation. RNA concentration was measured using a Qubit RNA High Sensitivity Assay Kit and integrity assessed using a TapeStation 4200 (Agilent).

PB purification and RNA extraction

PBs were purified essentially as described previously (19). Yeast cells expressing Dcp1-Myc were labeled with 4TU as described above and crosslinked with 0.8% formaldehyde on ice. Cells were crosslinked for 1 h and then quenched with 125 mM glycine pH 7. Yeast cells were harvested by centrifugation and lysed in Blob100 buffer (20 mM Tris–HCl pH 8.0, 100 mM NaCl, 1 mM MgCl_2_, 0.5% NP-40, and 0.5 mM TCEP) supplemented with RNasin Plus (Promega) and EDTA-free protease inhibitors (Roche Diagnostics). Lysates were cleared by centrifugation at 1000g for 10 min, and high-molecular weight complexes pelleted at 20,000g for 10 min; the resulting pellet was used for immunoprecipitation (IP). Approximately 2 mg total protein was incubated with anti-Myc magnetic beads for 30 min at room temperature or overnight at 4 °C. Beads were sequentially washed four times in Blob500 (500 mM NaCl) followed by Blob100 buffer without detergent. Bound RNA was released from beads by incubating with Proteinase K at 55 °C for 15 min. Crosslinks were reversed at 70 °C for 40 min in Blob100 buffer modified with 0.5% SDS plus 1 mM EDTA and without RNasin/MgCl_2_ (19). RNA was purified using the RiboPure Yeast RNA Kit (Invitrogen). RNA samples were treated with DNase I at 37 °C for 10 min and purified by phenol–chloroform extraction followed by ethanol precipitation. RNA concentration and integrity was assessed as described above.

RNA alkylation and next generation sequencing

Incorporated 4TU residues were chemically alkylated using the SLAM-seq protocol (58). Briefly, 4 μg of RNA was incubated in 50 mM sodium phosphate buffer (pH 8.0) containing 30% dimethyl sulfoxide (DMSO) and 10 mM iodoacetamide (IAA) for 15 min at room temperature in the dark, quenched with 10 mM DTT, and ethanol-precipitated. Strand-specific, 3′-mRNA libraries were generated using the Lexogen QuantSeq 3′ mRNA-Seq V2 (FWD) kit with Unique Dual Indices, following the manufacturer’s instructions. Libraries (200 ng RNA input) were PCR-amplified (13–15 cycles), purified, quantified (Qubit dsDNA HS), and sized (Agilent TapeStation 4200). Indexed libraries were pooled equimolarly (10–20 nM final concentration) and sequenced on an Illumina NovaSeq 6000 at the University of Manchester Genomic Technologies Core Facility (GTCF). All conditions included three independent biological replicates for both total and IP RNA. Reads were trimmed, quality-filtered, and aligned to the S. cerevisiae W303 reference genome using NextGenMap (59). T→C conversion events were quantified using the SLAM-DUNK pipeline (60).

Data processing

Sequencing data were analyzed using a customized Nextflow implementation of the nf-core/slamseq pipeline for S. cerevisiae (https://github.com/Mattfeng414/slamseq). The W303 reference genome and 3′-UTR BED file (17) were used for alignment and quantification. Workflows were containerized with Docker to ensure reproducibility. Resulting count and conversion matrices were used for downstream modeling. An efficiency term was used to account for differences in 4TU incorporation during glucose depletion conditions. This was derived from approximately 100 transcripts identified in this study that were not expressed under normal glucose conditions and comparing their total RNA counts to 4TU-labeled RNA counts under glucose-depletion conditions. The resulting time-dependent labeling efficiencies were then applied as correction factors, assuming a comparable incorporation rate across transcripts. Under glucose-replete conditions, over 80% of transcripts were labeled within 10 min and no correction was necessary.

RNA half-life analysis

To estimate mRNA stability in total RNA, transcript decay was modeled under both glucose-replete (Glu+) and glucose-deplete (Glu-) conditions using a single-phase exponential decay equation:

$$[eqn]$$

Where a is the initial abundance and k the decay constant, from which half-life was derived as:

$$[eqn]$$

Fitting was performed using nonlinear least squares with nonnegativity constraints (a ≥ 0; k ≥ 0), starting from initial estimates a = p(t0) and k = 0.1. Poorly convergent fits were excluded—nonlinear regression failed to converge or if the resulting model had R^2^ < 0.4, negative or zero decay constants (k ≤ 0). Following decay modeling, genes with valid half-life estimates under both Glu+ and Glu-conditions, supported by three independent replicates, were retained. For each gene, we performed a single two-sample Welch’s t test comparing mean half-life between Glu- and Glu + conditions. The relative difference (rel_diff) was calculated as:

$$[eqn]$$

where HL_G0_ and HL_G2_ denote half-life estimates under glucose-replete (Glu+) and glucose-deplete (Glu-) conditions, respectively. Genes were classified as showing decreased stability (rel_diff > 0.3, p < 0.05) or increased stability (rel_diff > 0.8, p < 0.05). Genes were classified as showing similar stability if the per-condition variability (CV = SD/mean) was within the interquartile range and the rel_diff was ≤ 0.3 when HL_G2_ > HL_G0_ or ≤ 0.8 when HL_G0_ > HL_G2._

RNA synthesis analysis

To model new RNA synthesis, 4TU-labeled total RNA time-course data were fitted using a third-degree polynomial regression:

$$[eqn]$$

implemented via scikit-learn’s PolynomialFeatures and LinearRegression.

This flexible model accommodates nonmonotonic synthesis typical of stress-induced transcripts. Fitted curves were evaluated on 100 evenly spaced time points, and the area under the curve (AUC) calculated using Simpson’s rule (scipy.integrate.simps) as a measure of total RNA production across the assay window.

After curve-based synthesis estimates were obtained, genes were retained if each growth condition (Glu-, Glu+) had three replicates with valid AUCs. For each gene we then performed a two-sample t test comparing AUCs between Glu- and Glu + conditions. Cohorts were defined based on the direction of the mean AUC difference (increased or decreased synthesis) and statistical significance (p < 0.05). Genes were classified as showing similar synthesis if they showed no significant differences between conditions (p ≥ 0.05). To ensure that the lack of significance reflected genuine similarity rather than variability, only genes with a coefficient of variation (CV = SD/mean) of AUCs within the interquartile range were retained per condition.

P-body localization dynamics

For PB dynamic transcripts, normalized IP RNA profiles were fitted using an asymmetric Gaussian model with a baseline term for each biological replicate:

$$[eqn]$$

where A is the amplitude (extent of enrichment), x0 the time of maximal localization, σ the spread (allowing different widths for rise and fall phases), and b the baseline.

This model captures transient localization waves where transcripts are enriched in PBs in a transient manner. Nonlinear least-squares fitting (SciPy curve_fit) was performed on z-scored expression data across replicates, with parameter bounds ensuring biologically plausible solutions (e.g., A ≥ 0, σ > 0). The fitted parameters (A, x0, σ, b) and R^2^ values provided a compact description of localization kinetics for each transcript. The goodness-of-fit was quantified by the coefficient of determination (R^2^), and genes with R^2^ ≥ 0.4 were classified as PB dynamic genes.

For PB accumulated transcripts, normalized IP RNA profiles were fitted using nonlinear least squares (SciPy curve_fit) with two alternative three-parameter models. The first was an exponential growth model:

$$[eqn]$$

where a is the maximal enrichment amplitude, b the accumulation rate constant, and c the baseline offset, representing gradual association of transcripts with PBs over time. The second was a logistic growth model:

$$[eqn]$$

where a denotes the asymptotic maximum enrichment, b controls the slope (accumulation rate), and c indicates the inflection time (midaccumulation). Both models were fitted with bounded parameters (a≥0, b ≥ 0, c unbounded for the exponential; c ∈ [min(x), max(x)] for the logistic model). For each gene, the goodness of fit was assessed by the coefficient of determination (R^2^), and the model with the higher R^2^ was selected to represent its accumulation dynamics (R^2^ ≥ 0.4 were retained), providing quantitative measures of the rate and extent of IP RNA association over time.

Genes that did not meet either dynamic-curve or accumulated-curve criteria were further evaluated using linear regression. Those with a slope < 0.01 were assigned to the PB less-responsive cohort.

Filtering and classification of newly synthesized transcripts that localize to PBs

Genes showing accumulation of newly synthesized RNA in PBs (IP samples, Glu-) were defined through a multistep filtering process. First, a likelihood-ratio test (LRT) in DESeq2 was applied with the design (∼ batch + time point) and the reduced model (∼batch) to detect genes whose expression changed significantly over time, retaining those with adjusted p < 0.01. To ensure sufficient data support, genes were also required to have ≥ 100 raw counts in all three replicates across the time course. Next, the effect-size criterion was applied by comparing variance-stabilized (VST) expression levels at each time point with the baseline (time = 0 min). Genes were retained only if their maximum log_2_(VST) increase relative to baseline was ≥ 1.5, corresponding to an approximately 2.8-fold upregulation. To ensure that changes reflected genuine temporal dynamics, a variability filter was applied: genes were required to show a median absolute deviation (MAD) ≥ 0.9 across time points. Finally, to avoid artefactual enrichment due to constitutively high baseline expression, genes whose maximum CPM value occurred at T = 0 in any replicate were removed. The final PB-enriched gene set comprised the intersection of genes passing the time-course significance, raw-count support, effect-size, and variability filters, after excluding baseline-peaking transcripts.

Bioinformatic analyses

Heatmaps were generated to visualize temporal expression patterns of mRNAs. Average expression values (normalized counts) across biological replicates for each glucose condition (Glu + or Glu-) and time point (0, 2.5, 5, 10, 20, and 40 min) were computed. Expression values for each gene were then z-normalized across time to emphasize relative changes for each gene. Heatmaps were generated using the pheatmap R package with row clustering enabled and fixed color scaling as indicated in each heatmap. GO enrichment analyses were performed using clusterProfiler (61). Biological process ontology terms were tested using the enrichGO function, applying Benjamini–Hochberg correction for multiple testing and a significance threshold of false discovery rate-adjusted p < 0.05. The top 10 GO terms were selected based on their adjusted p values and ranked by their gene ratio on plots.

RNA structural properties were assessed using genome-wide PARS scores (51) and Gini indices (25). PARS profiles were normalized into 100 bins per transcript to account for gene length and averaged across groups, while Gini indices quantified transcript-wide folding inequality. Regional RNA structure analyses for the 5′UTR, CDS, and 3′UTR regions were based on genomic annotations from the Saccharomyces Genome Database. PolyA tail length data were derived from (62). RNA abundance data generated in this study were log_10_-transformed and compared across gene sets to assess differences in transcript levels. TE data were obtained from ribosome profiling data measured under Glu+ and Glu-conditions (19). SG association was obtained from data using Pbp1 as a canonical SG marker under glucose starvation conditions (19). All comparisons were performed in R using the Mann–Whitney U test. p values were adjusted for multiple testing using the Benjamini–Hochberg method, and significance was defined as false discovery rate-adjusted p (p < 0.05 ∗, p < 0.01 ∗∗, p < 0.001 ∗∗∗).

Quantitative RT-PCR analysis

qPCR data were processed to calculate relative transcript abundance across time points for wild-type and lsm4ΔC edc3Δ mutant strains. Raw cycle threshold (CT) values were normalized within each biological batch using 18S rRNA as the internal reference to obtain ΔCT values. For each target gene, ΔΔCT was calculated relative to the baseline sample (time point 0 min, replicate 1) from the same batch, and fold change was computed as 2^-ΔΔCT^. Fold change values were averaged across biological replicates, and the mean ± standard deviation was plotted over time for each gene and strain. PCR primer sequences used for qPCR validation are listed in Table S3.

Microscopy

PBs were visualized using a Leica DM5500 B microscope equipped with a Leica K5 cMOS camera. Confocal images were taken with a HCX PL APO 100x/1.40 – 0.70 OIL objective to visualize Dcp2-CFP (56) using Application Suite X (Leica). The fluorescent parameters for CFP were: excitation 436/20 nm, emission 480/40 nm. Images were taken as 0.20 μm Z-stacks, which were processed using ImageJ software (National Institutes of Health)

Data availability

The data used and analyzed in this study are available from the corresponding author upon request. RNA-seq data have been deposited at Gene Expression Omnibus (GEO) as GSE312543.

Supporting information

This article contains supporting information.

Conflict of interest

The authors declare that they have no conflicts of interest with the contents of this article.