Spliceosomal introns are ubiquitous non-coding RNAs that are typically destined for rapid debranching and degradation. Here we describe 34 excised introns in Saccharomyces cerevisiae that—despite being rapidly degraded in log-phase growth—accumulate as linear RNAs under either saturated-growth conditions or other stresses that cause prolonged inhibition of TORC1, which is a key integrator of growth signalling. Introns that become stabilized remain associated with components of the spliceosome and differ from other spliceosomal introns in having a short distance between their lariat branch point and 3′ splice site, which is necessary and sufficient for their stabilization. Deletion of these unusual introns is disadvantageous in saturated conditions and causes aberrantly high growth rates in yeast that are chronically challenged with the TORC1 inhibitor rapamycin. The reintroduction of native or engineered stable introns suppresses this aberrant rapamycin response. Thus, excised introns function within the TOR growth-signalling network of S. cerevisiae and, more generally, excised spliceosomal introns can have biological functions.
Spliceosomal introns are a defining feature of eukaryotic life; they are present in all known eukaryotic genomes and absent from all known non-eukaryotic genomes1,2. Introns and splicing can have important roles in eukaryotic gene evolution, mRNA expression and protein diversification3. Moreover, once excised from the primary transcript, introns can sometimes be further processed to generate non-coding RNAs, such as microRNAs or small nucleolar RNAs4. In principle, full-length excised introns could also be a fertile source of functional non-coding RNAs. However, excised intron lariats are debranched and degraded within seconds5,6,7,8,9, and thus are generally viewed as inert and ephemeral by-products of gene expression10.
Functions for individual introns have been examined most thoroughly in the intron-poor budding yeast S. cerevisae, which contains approximately 300 spliceosomal introns, with only a few annotated cases of alternative-splicing events11,12,13. For most introns tested, no growth phenotypes are detected upon their removal14,15,16, but for a few, phenotypes are observed. Some functions inferred from these phenotypes, such as an influence on gene expression17,18, manifest entirely during pre-mRNA production and processing, and thus before the introns exist as separate RNA molecules. Other functions, such as providing a source of small nucleolar RNAs, occur after splicing, although in these cases the flanking portions of the intron are still rapidly catabolized19. Thus, functional analyses in S. cerevisiae support the prevailing view that the collective fate of introns after splicing is solely to be debranched and at least partially degraded.
Although most yeast introns have now been individually assayed for function, relatively few experimental conditions have been explored. Most experiments assay cells during log-phase growth, which provides consistent measurements in a standardized system that is sensitized to detect differences in growth and metabolism. However, outside of the laboratory, yeast cells are unlikely to spend many consecutive generations rapidly dividing and more often face limiting nutrients or other stresses20. Accordingly, we set out to examine the gene regulation of S. cerevisiae outside of the context of log-phase growth.
Accumulation of excised, linear introns
We performed RNA sequencing (RNA-seq) on two S. cerevisiae samples: one taken from a culture in log-phase growth and the other from a saturated culture in which cell density was minimally increasing. For most intron-containing genes (exemplified by ACT1), very few intron-mapping RNA-seq reads were observed from either culture condition (Fig. 1a), as expected if introns were rapidly degraded after splicing. However, for a subset of genes (exemplified by ECM33), many intron-mapping reads were observed specifically from the saturated culture (Fig. 1b). Indeed, the higher density of reads for the ECM33 intron compared to that for its exons suggested that in this growth condition the intron accumulated to much greater levels (tenfold) than did its corresponding mature mRNA.
Read coverage dropped abruptly at the 5′ and 3′ boundaries of the ECM33 intron (Fig. 1b, Extended Data Fig. 1), which suggested accumulation of full-length intron. To examine whether our RNA-seq analysis was detecting the accumulation of excised introns of a defined size, we probed RNA blots for the inferred RNA species. Probes to the ECM33 intron detected a single major species migrating at the position expected for the full-length intron (Fig. 1c). Likewise, a probe for SAC6, another intron inferred by RNA-seq to accumulate in saturated culture (Fig. 1d), detected an RNA of the size expected for the full-length excised intron (Fig. 1e). These introns that accumulate in saturated cultures are hereafter called ‘stable introns’.
A known mechanism by which excised introns can be protected from degradation is by evading debranching and persisting as either lariat RNAs or circular derivatives in which the lariat tail is missing8,21. However, these nonlinear species have either branched RNA or a 2′–5′ phosphodiester linkage, which would impede reverse transcriptase7,22 and thereby cause RNA-seq reads to be depleted in the region of the branch point—a pattern that we did not observe in the RNA-seq profiles (Fig. 1b, Extended Data Fig. 1). To test further whether ECM33 intronic RNA accumulated as either a lariat RNA or its circular derivative, we collected RNA from yeast that lacked Dbr1 (the enzyme that is required to debranch intron lariats8) and compared the ECM33 intronic RNA that accumulated in the dbr1Δ strain with RNA that accumulated in wild-type saturated cultures. As expected, in dbr1Δ cultures the ECM33 intron was detected as two abundant species, which corresponded to the branched lariat and its circular derivative (Fig. 1f). Importantly, neither of these nonlinear species co-migrated with the linear intron identified in wild-type saturated cultures. These results confirmed that Dbr1 is necessary to form linear introns outside of log phase and showed that the previously known mode of decreased intron turnover cannot explain the accumulation of excised introns in a saturated culture.
Another mechanism that might protect stable introns from degradation in saturated cultures is incorporation into a ribonucleoprotein complex. Indeed, the ECM33 intron predominantly co-sedimented with complexes that were about the size of ribosomal subunits (35–50 Svedberg units, Extended Data Fig. 2a). To identify proteins that associate with the intron, we performed pull-down experiments from gradient fractions that contained the intron and used quantitative mass spectrometry to identify co-purifying proteins. For these experiments, we took advantage of the observation that tagged versions of the ECM33 intron excised from expression constructs retained the behaviour of the endogenous ECM33 intron (Extended Data Fig. 2b, c). Each of the top 10 proteins that consistently co-purified with MS2-tagged versions of the ECM33 intron (average enrichment, 4.9-fold) were spliceosomal proteins, the identities of which indicated that the excised and debranched ECM33 intron resided in a specific complex; when compared to known spliceosome complexes, this complex most closely resembled the intron–lariat spliceosome (ILS) complex23 (Extended Data Table 1).
Studies in log-phase extracts indicate that introns are debranched after spliceosome disassembly24. If stable introns were also debranched outside the spliceosome, they would then need to re-associate with spliceosome components as linear RNAs. A simpler model is one of continued association of stable introns with ILS components during their debranching; in this case, the relationship between ILS disassembly and debranching presumably varies depending on growth condition and/or intron identity. In either scenario, our findings indicated that, in saturated cultures, accumulating introns are bound to and presumably protected by a complex that resembles the ILS.
Defining features of stable introns
We performed a systematic search for all introns that undergo a switch in stability and accumulate as linear RNAs in saturated cultures. In this search, RNA-seq reads of each intron-containing gene were analysed for a preponderance of reads that mapped to introns—particularly those that mapped to edges of excised introns (consistent with a post-splicing intron) relative to those that mapped across splice sites and splice junctions (which represented signatures of intron retention and mature mRNA expression, respectively)25 (Extended Data Fig. 3a). Inspection of RNA-seq reads that mapped to 3′ edges of stable introns identified many that were extended by one or more untemplated adenosine residues (Extended Data Fig. 3b). Short 3′-terminal oligoadenylate tails are added by the TRAMP complex to mark nuclear RNAs for degradation26,27. Our finding that many stable-intron molecules had these tails suggested that these molecules might have been targeted for exosomal decay, yet had somehow escaped this decay. Regardless of their function, these untemplated adenosine residues provided an additional criterion for annotation of stable introns, which helped us to confidently identify another 28 introns that accumulated in saturated cultures (Extended Data Fig. 4, Extended Data Table 2).
We searched among the 30 stable introns for common features, which cellular machinery might use to differentiate them from the remaining introns that are rapidly degraded in saturated cultures. Compared to other introns, stable introns had similar strengths of canonical splicing motifs (Extended Data Fig. 5a), similar length distributions (Extended Data Fig. 5b), no common predicted structures or enriched sequence motifs (Extended Data Fig. 5c, d), and no enriched functional ontologies of their host genes. Of all the features we examined, the only one that differed was the distance between the lariat branch point and 3′ splice site (3′SS), which tended to be shorter for stable introns (Fig. 2a).
To investigate a potential role of branch-point position in influencing intron stability, we introduced mutations that changed endogenous branch point–3′SS distances and examined the effects on intron accumulation. Lengthening the short branch point–3′SS distance of the normally stable ECM33 intron from 25 to 45 nucleotides abrogated the accumulation of the full-length excised intron, which indicated that a short branch point–3′SS distance is required for the stability of this intron (Fig. 2b). Moreover, shortening the branch point–3′SS distance of the normally unstable ACT1 intron from 44 to 25 nucleotides conferred stability to this intron in saturated culture, which suggested that a short branch point–3′SS distance is not only necessary but also sufficient for stability (Fig. 2b).
The notion that a short branch point–3′SS distance is sufficient for stability suggested that introns that had short branch point–3′SS distances yet were not identified as stable introns (Fig. 2a) might not have been transcribed during the period at which stable introns are protected from degradation. To investigate this possibility, we used our expression construct to express introns that had not been identified as stable introns, choosing two introns with a short (20- and 25-nucleotide) and two with a long (37- and 44-nucleotide) branch point–3′SS distance. The two with short branch point–3′SS distances accumulated specifically in saturated cultures, whereas the two with long branch point–3′SS distances were unstable in both conditions (Extended Data Fig. 5e).
Thus, the two defining features of stable introns are (1) a short branch point–3′SS distance; and (2) expression within a cellular context in which introns are stabilized. After being modified to satisfy these criteria, all tested non-stable introns became stable introns (Fig. 2b, Extended Data Fig. 5e), which indicated that no other feature of either the intron or the host gene is required to achieve stability.
Regulation of stable introns
We next monitored levels of the ECM33 stable intron and its host mRNA through 72 h of culture to determine when the switch in intron stability occurred (Fig. 3a). The ECM33 mRNA levels were highest during rapid growth, whereas its intron levels were low through 8 h of growth and then markedly increased as cells exited rapid growth (Fig. 3a). Intron abundance remained high for at least the next 62 h, even as levels of mature mRNA decreased 23-fold. Although the dynamics of the ECM33 intron and mRNA illustrated the different behaviours sometimes observed for stable introns and their host mRNAs, the large decrease in the levels of the mature mRNA in saturated culture was not observed for most mRNAs that hosted stable introns (Extended Data Fig. 5f).
The timing of stable-intron accumulation suggested that their formation might be linked to exit from rapid growth. With this in mind, we used rapamycin to inhibit the target of rapamycin complex 1 (TORC1), which is a broadly conserved master integrator of nutritional and other environmental signals28,29. In yeast, rapamycin-mediated TORC1 inhibition leads to repression of anabolic processes, stimulation of catabolic processes and greatly reduced growth28. When a log-phase culture was resuspended in fresh medium containing rapamycin, the ECM33 intron accumulated after 4 h (Fig. 3b). A repeat of this experiment using a strain with a rapamycin-resistant allele of TOR1 (TOR1-1) yielded no accelerated intron accumulation upon treatment with rapamycin (Fig. 3c). TOR1-1 cultures nonetheless accumulated stable introns upon reaching saturation (Fig. 3c), as expected in this strain that is known to be sensitive to endogenous TORC1 inhibition30. Thus, rapamycin-induced intron accumulation was specifically due to inhibition of TORC1. The inhibition of protein synthesis in a TORC1-independent manner did not cause stable-intron formation (Extended Data Fig. 6a). Moreover, stable introns accumulated to lower levels in the TORC1* strain, which contains six hyperactive alleles of the regulatory network31 (Extended Data Fig. 7a)—results that further validated the involvement of TORC1 signalling.
To extend this investigation to the transcriptome, we prepared RNA-seq libraries from yeast that were treated for 4 h with rapamycin. We also investigated stable-intron formation in a metabolically different saturated-growth scenario: a lawn of yeast grown over 3 days in an aerobic environment. When compared to saturated-liquid culture, both rapamycin-treated cells and cells from the lawn showed a similar accumulation of not only the ECM33 intron (Extended Data Fig. 7b) but also other stable introns (Extended Data Fig. 7c). In total, 34 introns (11% of total S. cerevisiae introns) were classified as stable introns in at least one of the three conditions (Extended Data Fig. 4, Extended Data Table 2).
We tested several genetic and environmental perturbations for their effects on stable-intron accumulation. Gain-of-function mutations in TAP42 or SCH9 (the two major effector branches of TORC1) were individually insufficient to override TORC1 repression and attenuate stable-intron formation (Extended Data Fig. 6b, c), and Sch9 activity was not required for stable-intron regulation (Extended Data Fig. 6d), which implicated involvement of another effector branch of TORC1. Although depletion of carbon, nitrogen and amino acids rapidly inhibit facets of TORC1 signalling28, these conditions did not induce premature stable-intron formation (Extended Data Fig. 6e). Likewise, direct perturbations of the general amino acid control pathway did not disrupt stable-intron regulation (Extended Data Fig. 6f, g). Thus, the role of TORC1 in stable-intron regulation appeared separable from its role in rapid response to nutrient deprivation.
As illustrated for the ECM33 intron (Fig. 3d), stable introns were not detected after treating with rapamycin for 1 h—a time period that is sufficient to phenocopy the TORC1 response to nutrient deprivation. This result was consistent with our observation that some conditions that are known to rapidly inhibit aspects of TORC1 signalling did not induce stable introns, and motivated an attempt to account for this delayed response to rapamycin. Few studies have explored prolonged TORC1 inhibition in yeast, and those that report on ≥2-h rapamycin treatments observe notably different TORC1-mediated responses compared to short-term rapamycin treatment or starvation32,33,34,35. Some of these studies33,34,35 show that prolonged TORC1 inhibition (≥2-h treatment with rapamycin or tunicamycin) phenocopies secretory stress more closely than nutrient stress. We found that the secretory stressors tunicamycin and dithiothreitol (DTT) induced stable-intron formation, that for both stressors this induction required extended treatment durations similar to that required for rapamycin, and that induction of stable introns by these canonical activators of the unfolded-protein response did not require IRE1, the essential sensor of the unfolded-protein response (Fig. 3e, f). These results indicated that TORC1-mediated stable-intron accumulation depends on connections between TORC1 and membrane-trafficking or secretory stress36,37, and not the more-often-considered connections between TORC1 and nutrient stress.
Biological function of stable introns
The discovery of stable introns and their link to the TORC1 growth-signalling pathway brought to the fore the question of their function. To build strains to test for loss-of-function phenotypes, we used a CRISPR–Cas9 system38,39 that facilitated precise removal of introns without affecting exonic sequences. We reasoned that stable introns might contribute collectively to function, and therefore generated a quintuple mutant that lacked stable introns of five genes (ECM33, UBC4, HNT1, SAC6 and RFA2). Because introns from these genes normally accumulated to high levels, this strain, which we call EUHSR, had less than half the stable intron molecules as compared to the wild-type strain (Fig. 4a, Extended Data Fig. 8a).
To detect the consequences of reducing stable-intron abundance, we co-cultured the EUHSR and wild-type strains in a competitive-growth assay. The co-culture was grown for three days, diluted, grown for three days again, and so on, for a total of six cycles of growth. Time points taken just before dilution and after a day of regrowth revealed a saw-tooth pattern in which the wild-type strain had a significant advantage during the period of sustained saturated culture, whereas EUHSR had a significant advantage during the period that included re-entry to growth and exit from growth (Fig. 4b). These results indicated that stable-intron expression is beneficial in some physiological contexts and detrimental in others.
Because of the connection between TORC1 inhibition and stable-intron formation (Fig. 3b–f), we tested the effect of rapamycin on growth of mutant strains from a series of intron deletions that produced decreasing amounts of stable introns (Fig. 4a). In contrast to the co-culture experiment (Fig. 4b), seeding was with log-phase cultures, which helped to explain why all strains grew at equivalent rates in the absence of rapamycin (Fig. 4c). In the presence of rapamycin, growth was inhibited for all strains, but this growth inhibition was attenuated for strains with fewer stable introns (Fig. 4c). A notable correlation was observed between the estimated fraction of stable-intron molecules depleted from the transcriptome and attenuation of the rapamycin response (Fig. 4c, d). By contrast, the rapamycin response of control strains that lacked non-stable introns of highly expressed genes was indistinguishable from that of wild type (Fig. 4d, inset), which showed that the observed phenotype was specific to depletion of stable introns and not merely a consequence of decreased splicing load.
The attenuated response to TORC1 inhibition seemed to depend solely on the number of stable-intron molecules removed from the cell, and not on the identities of either the removed introns or their host genes (Fig. 4d). This finding suggested that the phenotype did not result from subtle changes in host-gene expression or other secondary effects of intron removal or strain construction (such as off-target effects of gene editing). To test the conclusion that the attenuated response to TORC1 inhibition depended solely on the aggregate number of stable-intron molecules produced, we examined the ability of ectopically expressed introns to rescue the phenotype. We first overexpressed the ECM33 stable intron or—as a control—its unstable derivative, which had a lengthened branch point–3′SS distance; these are referred to as ECM33-stable and ECM33-unstable, respectively. The stability of this overexpressed intron had no detectable effect on either wild-type or EUHSR growth in the absence of rapamycin, and had little effect on growth of the wild-type strain in rapamycin (Fig. 4e). Importantly, however, the stability of this ectopically expressed intron markedly influenced the growth of the quintuple mutant in the presence of rapamycin—nearly completely rescuing its defective response to rapamycin (Fig. 4e). Analogous results were observed when using stabilized and normal versions of the ACT1 intron (referred to as ACT1-stable and ACT1-unstable, respectively) (Fig. 4f). The ability of a stable version of the ACT1 intron (which does not exist naturally in S. cerevisiae) to rescue the quintuple-mutant phenotype confirmed that this phenotype was a primary consequence of reduced stable-intron abundance.
Taken together, our results show that stable introns function within the TORC1-mediated stress response in S. cerevisiae (Fig. 4g). TORC1 activity prevents stable-intron formation, as shown by the accumulation of these introns when cells undergo prolonged TORC1 inhibition (Fig. 3b–f), and stable introns inhibit growth of cells with decreased TORC1 activity, as shown by the increased growth observed in TORC1-inhibited strains with fewer stable introns (Fig. 4c–f). Thus, this double-negative regulation—in which TORC1 inhibits stable introns, which in turn inhibit growth—forms a previously unknown node of the TOR regulatory network, which works in concert with other TORC1-dependent and TORC1-independent pathways to control growth in S. cerevisiae (Fig. 4g).
The discovery of this node of the TOR regulatory network raises mechanistic questions, including that of how stable-intron accumulation might inhibit growth. Our results imply that the stoichiometry of stable introns relative to some cellular component underlies this function. One possibility is that stable introns sequester spliceosomes to reduce splicing activity. Supporting this idea, stable introns associate with spliceosome components (Extended Data Table 1) and S. cerevisiae pre-mRNAs compete for limited splicing machinery40. Moreover, in budding yeast sequestering spliceosomes might disproportionately affect mRNAs of ribosomal protein genes; because of their high expression levels and frequent possession of an intron, mRNAs of ribosomal protein genes are substrates for 90% of all splicing events in log-phase S. cerevisiae41. Although this percentage presumably decreases as cells exit log phase, stable introns could function as part of a larger regulatory network that uses inefficient splicing to help to keep ribosome production low during prolonged environmental and intracellular insults41,42.
To evaluate this idea, we revisited our RNA-seq data from rapamycin-treated yeast with the goal of determining the aggregate level of stable introns relative to spliceosome components. This analysis showed that stable-intron molecules reached 40% of the level of U5 small nuclear RNA molecules (Extended Data Fig. 8b), and in principle, an essential splicing protein could be more limiting than this small nuclear RNA. Thus, stable-intron accumulation was within a regime in which depleting 55% of stable-intron molecules—as in our EUHSR strain—could substantially alter spliceosome availability. Furthermore, overexpressing a stable intron (ECM33) in the EUHSR strain grown in rapamycin resulted in more intron retention and less accumulation of mRNAs of ribosomal protein genes as compared with overexpression of a non-stable intron (ACT1) (Extended Data Fig. 8c, d). Future characterization of the stable-intron complex will provide more-detailed information on the cellular machinery that is sequestered and might also provide insight into other mechanistic questions, such as how stable introns are protected from degradation and how this protection is biochemically coupled to TORC1 inhibition.
The ease with which we were able to observe the accumulation of stable introns raises the question of why they were not detected earlier. The answer lies with two features of our analysis. First, we examined cells in saturated culture, whereas most analyses examine cells in log phase—a stage at which all excised introns are rapidly degraded. Second, we avoided mRNA poly(A) selection, whereas many analyses perform poly(A) selection before RNA-seq; this selection depletes excised introns, which lack poly(A) tails. In addition, we used an in-house RNA-seq protocol in which RNA was fragmented, and fragments of 27- to 40-nucleotides in size were isolated for sequencing; by contrast, most analyses use commercial RNA-seq kits that deplete RNAs that are shorter than a few hundred nucleotides in size—the size of excised yeast introns. In principle, studies that used splicing-isoform microarrays to assay intron retention during stress, including rapamycin and DTT treatment, might have detected stable introns40,43,44. However, those studies focus on measurements over the course of 40 min, with no measurements taken beyond 2 h of treatment, which might not have been enough time to detect stable introns under these conditions. Our findings on prolonged TORC1 inhibition strengthen the links between TORC1 signalling and secretory stress in yeast33,34,35, and show that short versus long durations of TORC1 inhibition can result in distinct effects on some downstream biological processes—including stable-intron regulation.
Although our ideas regarding the mechanism of stable-intron function require validation, one interesting aspect of this mechanism is that it operates regardless of stable-intron sequence or genomic origin. This observation provides an example in which the history of a non-coding RNA (that is, how the molecule is born) can be sufficient to impart a cellular function—independently of any primary-sequence considerations. Similar principles might apply to other non-coding RNAs, especially those that lack primary-sequence conservation.
Our results add to the known fates and functions of spliceosomal introns within eukaryotic biology. Although noteworthy examples of intron lariats and trimmed linear introns with lives after splicing have previously been reported10, the introns described here are unique in that they persist as excised, full-length and debranched RNAs. Moreover, their stability is regulated in response to environmental changes, and they collectively act as a type of functional non-coding RNA. Our focus has been on the contexts of TORC1 inhibition and saturated growth, but stable introns might also be induced and have similar functions in other conditions. Although intron-rich eukaryotes cannot as obviously leverage spliceosome availability to manipulate the production of specific subsets of proteins, some might use excised linear introns to perform biological functions—perhaps in environmental conditions that have not yet been extensively profiled. At this point, we know that excised linear introns accumulate and function in at least one eukaryotic lineage, and we would be surprised if this is the only one.
Yeast strains and genetic manipulations
S. cerevisiae strains used in this study are listed in Supplementary Table 1. With the exception of strains acquired for TOR1-1 (Fig. 3c) and TORC1* (Extended Data Fig. 7a) experiments, all strains were in the BY4741 background. A TAP42 heterozygous diploid knockout strain (Horizon Discovery) was transformed with plasmids encoding either wild-type (TAP42) or mutant (tap42-11) alleles before sporulation and tetrad dissection. Transformations were performed using standard methods. Intron deletions and endogenous branch-point manipulations were made using a CRISPR–Cas9 system adapted for use in S. cerevisiae38,39 and were confirmed by colony PCR and Sanger sequencing of relevant genomic loci. Clones with correct deletions were counter-selected for the Cas9 plasmid by plating onto 5-fluoroorotic acid (5-FOA) medium before their use in experiments. This process was iterated to generate strains with multiple deleted introns. Strains that underwent 5-FOA counter-selection were confirmed to not be petite by patching to YP-glycerol plates. his3Δ1 repair was performed by transforming BY4741 with linear HIS3 DNA, selecting for His+ transformants, and confirming repair of the endogenous locus by colony PCR and Sanger sequencing. Plasmids used in this study are listed in Supplementary Table 2. Oligonucleotides (IDT) used for guide RNAs and repair templates are listed in Supplementary Table 3.
Growth conditions and collecting
Yeast were grown at 30 °C on standard synthetic complete (SC) plate and liquid medium, except experiments involving TORC1* strains (Extended Data Fig. 7a), which were performed according to previously described protocols31, and tap42::KanMX strains (Extended Data Fig. 6b), which were performed at 25 °C owing to temperature sensitivity. SC–Ura and SC–Trp were used in some experiments to maintain selection for URA3- and TRP1-expressing plasmids. SC–His was used in experiments involving 3-aminotriazole. Rapamycin (LC Laboratories; 10 mM stock in DMSO) aliquots were stored at −80 °C, and diluted to 10 μM in water immediately before use. 3-aminotriazole (Sigma-Aldrich; 1 M stock in water), cycloheximide (Sigma-Aldrich; 100 mg ml−1 stock in DMSO), DTT (Thermo Scientific; 1 M stock in water), and tunicamycin (Millipore-Sigma; 10 mg ml−1 stock in DMSO) aliquots were stored at −20 °C. Cultures were seeded at OD600 nm of 0.2 from overnight cultures (typically OD600 nm ~6) for growth to log phase or saturation. Log-phase cultures were collected during early log phase, typically at OD600 nm of 0.5, reached 4–5 h after seeding. Unless otherwise indicated, saturated cultures were collected 18–20 h after seeding. In acute nutrient-depletion experiments, cultures were grown to mid-log phase (OD600 nm ~1), filtered, washed twice with water and resuspended in appropriate medium (SC–glucose, SC–ammonium sulfate, SC–leucine or SC–uracil). All cultures were rapidly collected by vacuum filtration and flash-frozen in liquid nitrogen as described46. Frozen pellets were mechanically lysed using a Sample Prep 6870 Freezer/Mill (Spex SamplePrep; 10 cycles of 2 min on, 2 min off at setting 10). Lysate powder was aliquoted and stored at −80 °C.
Wild-type and EUHSR strains were co-cultured in a series of batch cultures, which were performed in biological replicate. The experiment started with a 50-ml co-culture in a 250-ml baffled flask seeded with an equal amount of the two strains at a combined OD600 nm of 0.2. The co-culture was then allowed to grow to confluence for 3 days. At this point, it was diluted to OD600 nm of 0.2, and growth and dilution were repeated for a total of 6 cycles. Two 1-ml samples were taken at each cycle, one immediately before dilution and the other 24 h after dilution. Genomic DNA was collected using the “Bust ’n Grab” protocol47, and the relative abundance of wild type and EUHSR was measured by PCR across the ECM33 and UBC4 intron-deletion loci (primers listed in Supplementary Table 3). By comparing to a standard curve of PCR results that had been templated with defined ratios of independently isolated wild-type and EUHSR genomic DNA (ranging from 1:4 to 4:1), the fractional content of wild-type cells was determined at each time point.
Total RNA was extracted from lysate powder using TRI Reagent (Ambion) according to the manufacturer’s protocol. rRNA was depleted from 5 μg of total RNA using the Ribo-Zero Gold Yeast rRNA Removal Kit (Illumina) according to the manufacturer’s protocol. RNA-seq libraries were prepared as described48 and sequenced on the Illumina HiSeq platform using 40-bp single reads. A detailed protocol is available at http://bartellab.wi.mit.edu/protocols.html.
Reads were trimmed of adaptor sequence using cutadapt (v.1.8)49. Trimmed reads were aligned to the S. cerevisiae genome (R64-1-1, downloaded from www.yeastgenome.org) using STAR (v.2.4)50 with the parameters ‘–alignIntronMax 1000 –sjdbOverhang 31 –outSAMtype BAM SortedByCoordinate –quantMode GeneCounts’ and with ‘–sjdbGTFfile’ supplied with transcript annotations. Intron annotations were constructed from the UCSC Genome browser and the Ares laboratory yeast intron database (http://intron.ucsc.edu/yeast4.3/), with the Ares laboratory database11,12 providing branch-point annotations used in subsequent analyses. RNA-seq visualization was performed using IGV (v.2.3.57)51,52. If multiple biological replicates were sequenced, figures show representative RNA-seq profiles from one of the replicates.
Untemplated nucleotides were found and catalogued by extracting reads from BAM files that uniquely mapped to the sense strand of an intron, but which also contained soft-clipped, non-mapping nucleotides at the 3′ end of the read. The mixture-of-isoforms (MISO, v.0.5.4)25 framework was used to quantify the relative expression of three potential isoforms for every intron: spliced (intron-degraded), spliced (intron-stable) and intron retained. MISO requires uniform read length as input. As such, the 3′ adaptor sequence was removed by trimming a constant 8 nt from every read. Based on the distribution of prior cutadapt-mediated trimming, this removed all 3′ adaptor sequence from >98% of reads. Retained-intron GFF events were constructed using ‘gff_make_annotation’ from rnaseqlib (http://rnaseqlib.readthedocs.io/en/clip/). Stable-intron GFF events were made by directly modifying the retained-intron GFF events to instead include ‘intron only’ as a potential outcome of splicing. Although reads that supported a stable-intron splicing event necessarily could also support a retained-intron splicing event, in practice, the greater abundance of most stable introns relative to flanking exons enabled MISO to identify stable introns as the predominant isoform for many introns in stable-intron-inducing conditions.
To be identified as a stable intron expressed in a given condition, accumulation of the intron had to exceed thresholds calibrated on experimentally validated cases. First, intron accumulation (transcripts per million, TPM) in the stable-intron-inducing condition had to be greater than half of its associated mRNA accumulation (TPM). Second, the intron accumulation (TPM) in the stable-intron-inducing condition had to be more than twice that of its accumulation (TPM) in log phase (a pseudocount of 0.01 TPM was added to all introns to account for introns with zero mapped reads). Third, the ratio of intron:mRNA accumulation in the stable-intron-inducing condition had to be greater than fourfold that of the intron:mRNA ratio in log phase. Imposing these thresholds eliminated many false positives, including some attributed to intronic small nucleolar RNA expression or constitutively poor splicing. Additionally, all introns containing small nucleolar RNA were manually inspected for evidence of full-length introns, which was not found in any case. For introns exceeding these accumulation thresholds, at least one of the following two additional criteria were required for annotation as a stable intron: (1) ≥2 terminal adenylylated reads mapping to the 3′ terminus of the intron, and/or (2) MISO-based support for preferential accumulation of the stable-intron isoform in the stable-intron-inducing condition (Bayes factor ≥30 when compared to log phase). These criteria identified conservative sets of stable introns expressed in a given condition, erring towards reducing false-positive identifications.
A search for motifs enriched in stable introns was performed using the MEME suite53 with remaining introns as the background set. k-mer frequencies were generated with the fasta-get-markov program from the MEME Suite. In addition, a search for enrichment of position-specific motifs was performed using kpLogo54 with remaining introns as the background set.
Growth curves were collected using Nunc Edge 2.0 96-well plates in a Multiskan GO Microplate Spectrophotometer. Wells were seeded at OD600 nm = 0.2 from early-log-phase cultures, and plates were incubated at 30 °C. Absorbance was read every 5 min with shaking on the ‘background’ setting, cycling between 1 min on and 1 min off. Every strain tested on each plate was run in technical triplicate. Single wells were censored if artefactual spikes in OD600 nm attributable to bubbles or condensation were observed. Strains were censored if more than one well was censored. Biological replicates for each strain included at least two independently derived transformants. A replicate of parental BY4741 was included on every plate as a control strain, and SC medium was included on every plate as a control condition for every strain being assayed. The growth rate was calculated from the log-linear portion of each growth curve. Growth curves were analysed with SkanIt (v.3.2).
Total RNA was extracted from lysate powder using TRI Reagent (Ambion) according to the manufacturer’s protocol. Ten micrograms of total RNA for each sample was resolved on a denaturing polyacrylamide gel and transferred onto a Hybond membrane (GE Healthcare) using a semi-dry transfer cell. Because ultraviolet crosslinking is biased against shorter RNAs, EDC (N-(3-dimethylaminopropyl)-N′-ethylcarbodiimide; Sigma-Aldrich) was used to chemically crosslink 5′ phosphates to the membrane55. Blots were hybridized to radio-labelled DNA probes. Probe oligonucleotides are listed in Supplementary Table 4. More details of this protocol are available at http://bartellab.wi.mit.edu/protocols.html. For experiments in Fig. 3a, RNAs were instead resolved on a glyoxal agarose gel and transferred overnight onto a Nytran SuPerCharge Turboblotter membrane (GE Healthcare). RNA-blot data were analysed with ImageQuant TL (v.184.108.40.206).
Crude lysates were prepared by resuspending an aliquot of thawed lysate powder (500–800 μl of loosely packed powder) in 1 ml of lysis buffer (10 mM Tris-HCl (pH 7.4), 5 mM MgCl2, 100 mM KCl, 1% Triton X-100, 1% sodium deoxycholate, 2 mM DTT, 20 U/ml SUPERase•In (Ambion), and cOmplete EDTA-free Protease Inhibitor Cocktail (Roche)). The lysates were placed on a rotator mixer at 4 °C for 5 min to allow for resuspension. Following brief vortexing, lysates were centrifuged at 1,300g for 10 min, and the supernatant loaded onto a 12.5-ml linear 10–30% (w/v) sucrose gradient (20 mM HEPES-KOH (pH 7.4), 5 mM MgCl2, 100 mM KCl, 2 mM DTT, and 20 U/ml SUPERase•In). Gradients were centrifuged in a pre-chilled SW-41 Ti rotor at 38,000 rpm for 4 h at 4 °C. Gradients were fractionated using a Piston Gradient Fractionator (Biocomp) in 1-ml fractions. For RNA analysis, total RNA was extracted from a portion of each fraction using TRI Reagent. RNA blots were performed as described for total RNA, except RNA loading was normalized by per cent of gradient fraction rather than by RNA concentration. For pull-downs, fractions were flash-frozen and stored at −80 °C.
Ectopic intron expression, affinity purification and mass spectrometry
Intron-overexpression constructs were based on SC–Ura-selectable pRSII416 (CEN) or pRSII426 (2μ) vectors56 (Extended Data Fig. 3c, Supplementary Table 2). Introns were inserted 49 nucleotides downstream of the URA3 start codon by Gibson assembly. Proper splicing was confirmed by RNA blot and cell viability, as cells unable to produce Ura3p through intron removal could not grow on SC–Ura medium.
Pull-down experiments were performed in the ECM33Δintron strain. Introns were purified using two MS2 hairpins (2 × MS2) inserted in various positions within the ECM33 intron (Supplementary Table 5). To minimize potential effects on splicing, the sequence within 60 nt of the 5′ splice site and 80 nt of the branch point was kept constant across all constructs. The 2 × MS2 hairpin sequence was based on CRISPR RNA scaffold designs57. 3×Flag-tagged MS2 coat protein with a C-terminal nuclear-localization signal (FLAG–MCP) was co-expressed from the constructs that also expressed the MS2-tagged introns.
We were unable to purify intact introns from the supernatant of saturated cultures owing to increased endogenous RNase activity in saturated cultures58. To circumvent this, sucrose-gradient fractions containing the intron of interest (typically fractions 6–9, determined by RNA blot) were pooled and used as the starting material for purification. Anti-Flag M2 Magnetic Beads (Sigma-Aldrich) (20 μl of packed-bead volume) were equilibrated and washed twice in 10 volumes of buffer 1 (100 mM KCl, 20 mM HEPES KOH (pH 7.9), 1% Triton X-100, 20 U/ml SUPERase•In, and cOmplete EDTA-free Protease Inhibitor Cocktail). Four hundred microlitres of the pooled fractions was added to the beads and incubated on a rotator mixer for 30 min at 4 °C. Remaining at 4 °C, the beads were washed twice in 10 volumes of buffer 1 and twice in 10 volumes of buffer 2 (200 mM KCl, 20 mM HEPES KOH (pH 7.9), 1% Triton X-100, 20 U/ml SUPERase•In, and cOmplete EDTA-free Protease Inhibitor Cocktail). Bound Flag–MCP was eluted from beads with 30 μl of 150 ng/μl 3×Flag peptide (Sigma-Aldrich) on a rotator mixer for 30 min at 4 °C. The eluents were precipitated with TCA, digested with trypsin and labelled with tandem mass tags to allow for quantitative comparisons between six total samples (three control and three test samples). Peptides were analysed by liquid chromatography–tandem mass spectrometry using an Orbitrap Elite (Thermo Fisher) coupled with a NanoAcquity UPLC system (Waters). Peptides were identified using SEQUEST and data analysed using PEAKS Studio (Bioinformatics Solutions).
Experimental design and reproducibility
No statistical methods were used to predetermine sample size. Growth-curve cultures were randomized by permutation of strain placement on 96-well plates across experiments. Edge effects were not significant within the time span that measurements were taken. The investigators were not blinded to allocation during experiments and outcome assessment. RNA-seq results for biological replicates correlated well (log-phase culture, R2 = 0.98 (mRNA, n = 5,652); saturated culture, R2 = 0.88 (mRNA, n = 5,762) and 0.78 (stable introns, n = 30)).
Further information on research design is available in the Nature Research Reporting Summary linked to this paper.
Sequencing data and the processed data for each gene and intron are available at the Gene Expression Omnibus (http://www.ncbi.nlm.nih.gov/geo) under accession number GSE121765. For gel source data, see Supplementary Fig. 1. All other data are available from the corresponding author upon reasonable request.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
We thank A. Amon, C. Burge, D. Pincus, D. Sabatini, P. Sharp and members of the Bartel and Fink laboratories for comments and discussion, G. Li, T. Pham and A. Symbor-Nagrabska for experimental assistance, A. Amon, R. Loewith, S. Schreiber and V. Vyas for reagents, and Whitehead core facilities for sequencing and mass spectrometry. This work was supported by NIH grants GM035010 (G.R.F.) and GM118135 (D.P.B.). D.P.B. is an investigator of the Howard Hughes Medical Institute.
Nature thanks T. Johnson, S. Montgomery and the other anonymous reviewer(s) for their contribution to the peer review of this work.
Extended data figures and tables
About this article