Transcription of long noncoding RNAs (lncRNAs) regulates local gene expression in eukaryotes. Many examples of how a single lncRNA controls the expression of an adjacent or nearby protein-coding gene have been described. Here we examine the regulation of a locus consisting of two contiguous lncRNAs and the master regulator for entry into yeast meiosis, IME1. We find that the cluster of two lncRNAs together with several transcription factors form a regulatory circuit by which IME1 controls its own promoter and thereby promotes its own expression. Inhibition or stimulation of this unusual feedback circuit affects timing and rate of IME1 accumulation, and hence the ability for cells to enter meiosis. Our data demonstrate that orchestrated transcription through two contiguous lncRNAs promotes local gene expression and determines a critical cell fate decision.
Expression of long noncoding RNAs (lncRNAs) influences local gene expression1,2,3,4,5,6,7. This process is conserved across species. A widespread mechanism through which lncRNAs modulate the expression of adjacent genes is by means of transcription-coupled chromatin changes2,7,8,9,10,11. Numerous examples of how a single lncRNA alters the expression of a nearby gene have been described1,2,6,12,13,14,15. Whether loci of two or more contiguous lncRNAs exist and how they regulate local gene expression remains unexplored.
Meiosis is central to gametogenesis during which a diploid cell gives rise to haploid gametes16. In Saccharomyces cerevisiae or budding yeast, the decision to enter meiosis is controlled by the master regulator transcription factor, IME117,18,19. Transcription of IME1 is tightly controlled by mating-type and nutrient signals19. In the presence of nitrogen and fermentable carbon sources, IME1 is repressed via PKA and TORC signaling pathways20. During nutrient starvation, however, expression of IME1 is induced in diploid cells and as a consequence cells enter meiosis.
Transcription of lncRNAs governs mating-type control of entry into meiosis in yeast2. In cells with a haploid mating type, transcription of the lncRNA IRT1, encompassing almost the complete IME1 promoter, represses IME1 expression via transcription-coupled chromatin changes2. In diploid cells, the transcriptional activator of IRT1, RME1, is repressed by the a1α2 repressor complex. This ensures that only cells expressing both mating types (MATa and MATα) can undergo meiosis. Interestingly, a large fraction of yeast natural isolates displays Rme1 expression in diploid cells suggesting that IRT1 may also be active in this cell type21,22,23. Previous work suggested that a second lncRNA is expressed further upstream in the IME1 promoter directly adjacent to IRT12,24. The purpose of this transcript is not understood.
Here we examine how a cluster consisting of two lncRNAs and IME1 controls entry into meiosis. We find that transcription of two contiguous lncRNAs facilitates a regulatory circuit through which IME1 promotes its own expression and meiotic entry. Our results demonstrate how a locus of contiguous lncRNAs can interact in a non-intuitive manner to define a positive feedback loop that drives the decision to enter an important cell differentiation program. The work broadens the spectrum by which transcription of lncRNAs controls local gene expression.
Two contiguous lncRNAs are expressed in the IME1 promoter
Previous work showed that in cells with a single mating-type IME1 expression repressed by transcription through the IME1 promoter of the lncRNA IRT1. A second lncRNA, named MUT1523 or IRT2, annotated upstream in the IME1 promoter directly adjacent to IRT1, and expressed in the same direction as IME1 and IRT1 was reported (Fig. 1a and Supplementary Fig. 1a)2,24. This transcript is about 400 bp and expressed in diploid cells during starvation. To examine whether IRT2 is detectable by conventional northern or reverse transcription (RT)-PCR methods, we first measured its expression pattern in diploid cells of strain backgrounds that undergo meiosis proficiently (SK1) and poorly (S288C)21. We used SK1 because cells from this strain background enter meiosis synchronously, which makes the use of population-based assays possible for the study of meiotic regulatory mechanisms. In SK1, IRT2 was detectable by northern blot in diploid cells exposed to sporulation medium (SPO), which induces cells to enter meiosis (Fig. 1b and Supplementary Fig. 1a–c). When we further examined the expression pattern in relation to the meiotic program, we found that IRT2 was expressed prior and during meiotic divisions (Supplementary Fig. 1b, c). In S288C, IRT2 expression was also clearly detected at 8 and 24 h in SPO by RT-PCR (Fig. 1c). We conclude that a second lncRNA, IRT2, is expressed in the IME1 promoter in diploid cells during meiotic entry.
IRT1 and IRT2 control meiotic entry
The expression of IME1 is critical for entry into meiosis in yeast17,18. Diploid cells that do not induce IME1 expression will not enter meiosis. Given that IRT1 and IRT2 are localized in the IME1 promoter, we next examined how their expression affects the propensity for cells to enter meiosis. We generated strains where RME1, the activator of IRT1, was deleted (rme1Δ) and where IRT2 was expressed from an inducible promoter (pCUP-IRT2, +Cu)2,25 (Fig. 1d). This set of mutants allowed us to dissect how IRT1 and IRT2 control entry into meiosis. In S288C, the fraction of cells that completed at least one meiotic division increased in the rme1Δ (35% compared to 21% for the control) suggesting that IRT1 is also active in diploid cells (Fig. 1e). Furthermore, when pCUP-IRT2 was expressed, a comparable increase in meiotic cells was observed (Fig. 1e). Conversely, in the absence of IRT2, (pCUP-IRT2, −Cu) the fraction of cells that underwent meiosis decreased (11% compared to 21% for the control). These data show that expression of IRT1 represses meiosis, whereas expression of IRT2 promotes meiosis.
In budding yeast, the a1α2 repressor complex, expressed in diploid cells from the opposite mating-type loci (MATa and MATα), represses RME1 and accordingly IRT1 expression with variable degree among yeast isolates22,23. SK1 harbors the RME1(del-308A) allele relative to S288C, which creates an additional canonical a1α2 repressor binding site21. This causes efficient RME1 repression and thereby increases the ability of SK1 diploid cells to enter meiosis. Hence, rme1Δ had a minor effect on meiosis in SK1 (Fig. 1f). Induction of pCUP-IRT2 expression also marginally affected meiosis in SK1 cells suggesting that IRT1 is needed for IRT2 function. To examine the relation between IRT1 and IRT2 further, we mutated the a1α2 repressor binding sites in the RME1 promoter (RME1-H) in SK1 cells, which de-represses RME1 expression2,25 (Fig. 1d). Consequently, IRT1 expression was increased and the fraction of cells entering meiosis reduced (75% compared to 14% for the control) (Fig. 1f and Supplementary Fig. 2a). Cells harboring RME1-H completed meiosis with a reduced efficiency when IRT2 was repressed (pCUP-IRT2, −Cu) compared to cells harboring the wild-type IRT2 promoter (4% compared to 14% for the control) (Fig. 1f). Importantly, induction of IRT2 (pCUP-IRT2, +Cu) suppressed the meiotic phenotype caused by the RME1-H almost completely (70% compared to 14% for the control). It is worth noting that deleting IRT2 also disrupted the IRT1 promoter, and therefore was not suitable for studying IRT2 function on its own (Supplementary Fig. 2b). We conclude that inhibitory effect of IRT1 expression on meiosis can be suppressed by expressing IRT2.
IRT2 promotes IME1 expression by repressing IRT1
Having established that IRT2 suppresses the effect of IRT1 on meiotic entry, we next examined how IRT2 influences IRT1 and IME1 expression directly. Given that the IRT2 transcript covers two Rme1-binding sites directly upstream of IRT1, we hypothesized that IRT2 transcription interferes with the binding of Rme1 and thereby represses IRT1 (Fig. 1a). To test this directly, we expressed IRT2 from an inducible promoter (pCUP-IRT2) in cells that also express Rme1 (RME1-H) (Fig. 1d). We found that in the absence of IRT2 (pCUP-IRT2, −Cu), IRT1 levels were constant, whereas transcription of IRT2 strongly inhibited IRT1 accumulation (Fig. 2a). Rme1 almost completely disassociated from the IME1 promoter in the presence of IRT2 (Fig. 2b). As a consequence, IME1 expression was induced, cells underwent meiosis with faster kinetics, and more cells completed meiosis (100% versus 65%) (Fig. 2a, c). We further examined whether IRT2 controls IME1 expression in cis or in trans. We generated a heterozygous IME1 allele diploid strain with one copy harboring pCUP-IRT2 together with ime1Δ and the other copy expressing the wild-type IME1. Since expression of IRT2 (pCUP-IRT2, +Cu) did not improve the kinetics of meiosis compared to no expression IRT2, we conclude that IRT2 controls IME1 expression in cis (Supplementary Fig. 3a–d). Finally, we obtained evidence that IRT2 represses the expression of IRT1 directly and not via Ime1. Given that IRT2 expression leads to increased expression of IME1, it is possible that Ime1 itself is responsible for repression of IRT1. To test this, we measured IRT1 expression in cells harboring a 3′ end mutation in IME1 (ime1-t), which impairs Ime1 function, in the presence of transcription of IRT2 (pCUP-IRT2, +Cu). We found that IRT2 efficiently repressed IRT1 expression despite the presence of ime1-t (Supplementary Fig. 3e). In conclusion, IRT2 transcription represses IRT1, and thereby promotes IME1 expression and entry into meiosis.
Ime1 feeds back to its own promoter and activates IRT2
IRT2 expression occurs in diploid cells only during entry into meiosis indicating that its transcription is tightly regulated. To identify transcriptional regulators of IRT2 expression, we scanned the IRT2 promoter for sequence motifs. An Ume6 binding site (tgggtggcta) was identified about 90 bp upstream IRT2, which corresponds to −2314 bp of the IME1 AUG (Fig. 3a). Indeed, Ume6 directly binds to the IME1 promoter as measured by chromatin immunoprecipitation (ChIP) (Fig. 3b). Cells lacking Ume6 motif did not show binding Ume6. Further, the binding of Ume6 was independent of IME1 expression because in the ime1-t mutant Ume6 binding was not affected (Supplementary Fig. 4a). The sequence motif is also present in related Saccharomyces species (Fig. 3a). Ume6 normally functions as a transcriptional repressor, however, in the presence of Ime1, Ume6 turns into a transcriptional activator26,27. Importantly, the expression of genes required for meiosis is controlled by Ume626,28. To examine how Ume6 and Ime1 control IRT2, we measured IRT2 expression in cells containing a deletion of the Ume6 motif (pIME1-Δu6), ime1-t, or both modifications. In pIME1-Δu6 cells, baseline IRT2 expression was detected (Fig. 3c, compare lanes 1–3 with 4–6). No IRT2 expression was observed in ime1-t cells, while the expression of the double mutant was comparable to pIME1-Δu6 (Fig. 3c, compare lanes 7–9 with 10–12 and 4–6). If IRT2 expression depends on Ime1 expression levels then induction of IME1 leads to increased IRT2 transcription. To test this, we measured IRT2 expression in cells harboring IME1 fused to an inducible promoter (pCUP-IME1). No detectable expression of IRT2 was detected when IME1 was not induced (Fig. 3d). Conversely, induction of pCUP-IME1 (+Cu) was directly followed by increased IRT2 expression, which required the Ume6-binding site (Fig. 3d and Supplementary Fig. 4b). In conclusion, Ume6 represses IRT2, whereas Ume6 together with Ime1 activate IRT2 expression.
Ime1 changes the local chromatin environment via IRT2
Having established that Ime1 promotes IRT2 transcription, we next investigated how the feedback loop regulates the local chromatin environment. It is well-established that transcription-coupled chromatin changes repress cryptic and local gene expression29. To examine chromatin changes in context of Ime1-mediated IRT2 transcription, we measured nucleosome positioning at the IRT2 region using micrococcal nuclease (MNase) and quantitative PCR during entry into meiosis in absence or presence of Ime1 and IRT2 expression. In the absence of Ime1 (pCUP-IME1, −Cu), the Rme1-binding sites were unprotected from the MNase digestion, which is indicative of an active promoter (Fig. 4a). In contrast, cells in which IME1 was induced (pCUP-IME1, +Cu), and therefore IRT2 was expressed, displayed stable nucleosomes around the Rme1-binding sites in the IME1 promoter (Fig. 4a). This result indicates that transcription of IRT2 establishes repressive chromatin to prevent Rme1 recruitment. Given that Rme1 promotes IRT1 transcription and therefore an open chromatin state near the Rme1-binding sites, we next examined how Rme1 expression affects the nucleosome positioning. Interestingly, stable nucleosomes were also detected, but less prominent, in RME1-H cells indeed suggesting that Rme1 and IRT2 have an opposing effect on nucleosome assembly in the upstream region of the IME1 promoter (Supplementary Fig. 5a, right panel). In line with this observation, we found that Rme1 binding was strongly reduced when IRT2 transcription was activated by Ime1 (Fig. 4b). As a consequence, IRT1 levels were reduced (Fig. 4c, compare lanes 3–9 with 10–16).
Transcription-coupled chromatin changes are often accompanied with post-translational modifications of histones that have an important function in establishing repressive chromatin29. Previous work showed that in haploid cells IRT1-mediated repression of IME1 requires Set2-dependent methylation of histone H3 lysine 36, and the histone deacetylase complex Set3C. Haploid cells lacking Set2 and Set3 fail to repress IME1 via IRT1, and undergo a lethal meiosis. In order to examine whether Set2 and Set3 also control IME1 expression in diploid cells, we measured IRT2, IRT1, and IME1 expression in set2Δ set3Δ cells harboring pRME1-H. As expected, IME1 expression was increased in set2Δ set3Δ cells because IRT1-mediated repression of IME1 is impaired in this mutant background (Fig. 4d). As a consequence, IRT2 expression was also elevated (Fig. 4d, compare lanes 7–11 with 18–22). Conversely, IRT1 expression decreased in the set2Δ set3Δ suggesting that IRT2-mediated repression of IRT1 does not require Set2 and Set3 (Fig. 4d, compare lanes 7–11 with 18–22). In line with this observation, histone H3 lysine 36 methylation was not detected at IRT2 while being transcribed (Supplementary Fig. 5b). IRT2-mediated repression of IRT1 was also not impaired in set2Δ set3Δ mutant cells when we expressed IRT2 from the CUP inducible promoter (pCUP-IRT2) (Supplementary Fig. 5c). We conclude that Set2 and Set3 regulate IRT1-mediated repression of IME1 also in diploid cells, but are not important for IRT2-mediated repression of IRT1. Taken together, these results show that Ime1-mediated activation of IRT2 changes the local chromatin environment.
Ime1 promotes its own expression
Our observation that Ime1-mediated activation of IRT2 represses IRT1 expression predicts that Ime1 is able to promote its own transcription. To test this directly, we expressed one copy of IME1 from an inducible promoter (pCUP-IME1), and inserted a superfolder green fluorescent protein (GFP) at the N-terminus (GFP-Ime1) of IME1, which allowed specific detection of IME1 expression from the endogenous wild-type locus. The GFP-Ime1 signal increased during starvation showing that the construct is representative for IME1 promoter activity (Supplementary Fig. 6a, b). When pCUP-IME1 was induced (+Cu) in cells with RME1-H, GFP-Ime1 levels increased more than fivefold compared to non-induced control cells, which required the Ume6-binding site (Fig. 4e and Supplementary Fig. 6c). Deleting the Ume6-binding site (pIME1-Δu6) caused an aberrant expression of IRT2 and an increased ability for cells to enter meiosis (Fig. 3c and Supplementary Fig. 7a–c), which explained the increase in GFP-Ime1 levels in pIME1-Δu6 cells compared to control cells (Fig. 4e). The results were comparable when we used cells harboring β-galactosidase IME1 promoter fusion instead of GFP-Ime1 (Fig. 4f). Thus, Ime1 promotes its own expression in trans.
Single-cell analyses of Ime1 feedback control
Population-based assays could mask changes in expression within subpopulations of cells. To circumvent this possible caveat, we combined time-lapse microscopy with microfluidics and monitored IME1 expression kinetics from a single locus (GFP-Ime1) in the absence or presence of the feedback control in single cells. The other IME1 locus expressed mCherry from an IME1 promoter lacking the IRT2 sequence (pIME1-Δirt2-mCherry), which disrupts both IRT1 and IRT2 transcription. The pIME1-Δirt2-mCherry served as an internal control and was used for normalizing the GFP-Ime1 signal (Fig. 5a). In addition, we mimicked RME1 expression of S288C by deleting only the proximal a1α2 repressor binding site in SK1 diploid cells (pRME1-Δaa1). Three parameters were measured from each single-cell GFP and mCherry trace: the time of IME1 induction (onset time), rate of accumulation, and maximum expression of IME1 (max intensity) (Fig. 5b and Supplementary Fig. 8). Cells containing pRME1-Δaa1 had a lower IME1 accumulation rate and lower max intensity compared to cells wild type for pRME1 confirming that increased IRT1 inhibits IME1 (Fig. 5c). Importantly, cells where IME1 was not able to feedback to its own promoter (pCUP-IRT2, −Cu) displayed a 60 min delay in onset of IME1, a twofold lower rate of accumulation, and a twofold lower max intensity. These data illustrate that the feedback loop consisting of Ime1, IRT2, and IRT1 is critical for the timing of Ime1 induction and the rate of Ime1 accumulation.
The Ime1 feedback loop is critical in yeasts expressing RME1
Our data demonstrate that Ime1 promotes its own expression by activating IRT2, which in turn represses IRT1 and thereby de-represses IME1. Thus, in order for the feedback circuit to function, it requires expression of IRT1 (Supplementary Fig. 2a). Indeed, the activator of IRT1, RME1, is expressed in a large proportion of diploid yeast isolates, including the S288C strain background21,22,23. To examine the physiological importance of the regulatory circuit consisting of Ime1, IRT2, and IRT1 more closely, we measured IME1 expression in S288C cells in the absence or presence of the feedback loop by modulating IRT2 expression. Since S288C cells enter meiosis asynchronously, we monitored IME1 expression using single-molecule RNA fluorescent in situ hybridization (smFISH) in single cells (Fig. 6a and Supplementary Fig. 9). When we repressed IRT2 expression (pCUP-IRT2, −Cu), the fraction of cells expressing high levels of IME1 (>40 copies per cell) decreased (2% compared to 13% for the wild type) (Fig. 6b). The fraction of cells harboring >40 IME1 transcripts per cell increased when IRT2 expression was elevated (pIME1-Δu6) (28% compared to 13% for the wild type) (Fig. 6b and Supplementary Fig. 7). The changes in the distribution of IME1 expression among single cells were also reflected in the fraction of cells that completed meiosis (Fig. 6c). In the absence of IRT2 expression (pCUP-IRT2, −Cu), the fraction of cells completing meiosis was >50% reduced compared to wild-type cells (Fig. 6c). Conversely, a higher fraction of cells completed meiosis when expression of IRT2 was increased (40% compared to 19% for the wild type). We conclude that the regulatory circuit consisting Ime1, IRT2, and IRT1 directs the cell fate decision whether or not to enter meiosis.
Model for IME1 regulation by noncoding transcription
Having established how transcription of IRT2 and IRT1 controls IME1 expression and entry into meiosis, we next formulated our findings in a mathematical model with the aim to examine the dynamic properties of the regulatory circuit (Fig. 7, Supplementary Fig. 10 and Supplementary Note 1). Specifically, our model accounts for the dynamics of the transcription rates of IRT1, IRT2, and IME1 as well as IME1 mRNA and Ime1 protein levels (Fig. 7b and Supplementary Fig. 10a). The starvation signal, which is the primary signal for activation of IRT1 and IME1 transcription, was considered to be either off (0) or on (1) (Fig. 7b)19,30. We note that this simple model is sufficient to reflect the experimental data. In the absence of IRT1 and IRT2 transcription, Ime1 levels accumulated faster and were higher compared to the wild type, in line with the observations from the single-cell analyses (compare Fig. 5c (middle and right panel) to Fig. 7b (middle and right panel)). Conversely, in the absence of IRT2 transcription, Ime1 levels were reduced compared to wild type. The model predicts a gradual increase in IME1 transcription rates during starvation, which is in line with our observation that Ime1 can stimulate its own expression via IRT2 and IRT1 (Fig. 7b, middle panel).
Finally, we used our model to simulate how IRT2 and IRT1 control the dynamics of Ime1 accumulation during different periods of starvation (Fig. 7c and Supplementary Fig. 10b). We find that prolonged starvation (5 h) lead to almost maximum Ime1 expression for the wild-type IME1 promoter, while a short period of starvation (1 h) resulted in low levels of Ime1 accumulation (Fig. 7c, compare left and right panels to Fig. 7b right panel). In the absence of IRT1 and IRT2, a short period of starvation (1 h) displayed comparable Ime1 levels to a simulation of the wild type starved for 5 h (Fig. 7c, compare yellow left panel to blue line right panel). This suggests that the IRT2-IRT1-Ime1 regulatory circuit restricts Ime1 expression in cells exposed for starvation for short time periods. Taken together, we propose that there are possibly two functions for the feedback loop: (1) to elevate the rate of IME1 transcription during prolonged starvation and (2) to prevent misexpression of Ime1 when the cells starved for a short period of time.
Cell fate decisions rely on precise expression of master regulatory genes. Here we described how in budding yeast a cluster of two lncRNAs is critical for controlling the expression of the master regulator for entry into meiosis, IME1. We demonstrate that orchestrated transcription of the two lncRNAs, IRT2 and IRT1, define a feedback cascade that ensures timely induction of Ime1 and increases the rate of Ime1 accumulation (Fig. 7d). Our findings also illustrate how transcription of two lncRNAs transmits the signal of an upstream promoter element to the downstream gene. We propose that regulatory networks consisting of multiple contiguous lncRNAs can be used as a generic strategy for cells to control local gene expression dynamics.
Ime1 activates its own expression via transcription of the lncRNAs, IRT1 and IRT2. Previous work showed that transcription of IRT1 represses IME1 in haploid cells by establishing repressive chromatin in the IME1 promoter2. We found that IRT1 is also active in diploid cells, yet the purpose IRT1-mediated repression of IME1 in this cell type is not well understood. We propose that IRT1 slows down meiotic entry till cells are “ready” to undergo meiosis. When cells are starved for nutrients, transcription of IRT1 represses accumulation of IME1. As Ime1 levels increase, it is able to feed back to its own promoter. Together with Ume6, Ime1 activates transcription of IRT2, which in turn represses IRT1 and consequently more Ime1 is produced. It has been shown that Ime1 and Ume6 can physically interact with each other31,32. The Ume6-Ime1 interaction is required for the activation of early meiotic genes32. Thus, only in the presence of Ime1 and Ume6, IRT2 is transcribed and IRT1 is repressed. In cells where Ime1 is not able to feed back to its own promoter, the expression of IME1 is reduced and fewer cells enter the meiotic program. Thus, this regulatory switch ensures that Ime1 levels will increase only when cells are able to activate the expression of meiotic genes and progress into meiosis.
Why does Ime1 feed back to its own promoter via IRT2? Previous work showed that transcription of IRT1 represses a relatively large region (over 1.5 kb) of the IME1 promoter requiring the upstream Rme1-binding sites2. If Ime1 feeds back to its own promoter downstream of the Rme1-binding sites to directly activate its transcription, it would face interference by transcription of IRT1. Our data and modeling show that repression of IRT1 transcription facilitates full activation of the IME1 promoter. We propose that transcription of IRT2 is an effective way of repressing IRT1-mediated interference of the IME1 promoter, and thereby promotes Ime1 accumulation and entry into meiosis.
A widespread mechanism for regulating gene expression locally is by transcription-coupled chromatin changes2,3,6,9,10,15,29,33,34. Previous work showed that transcription of IRT1, and not the IRT1 RNA itself, is important for repression of the IME1 promoter in haploid cells2. We propose that IRT2 also uses a transcription-based mechanism for repressing IRT1. Like IRT1, transcription of IRT2 alters the local chromatin structure (Fig. 4a), and inhibits binding of a transcriptional activator to the IME1 promoter (Fig. 2b and Fig. 4b). In addition, we find that the repression of IRT1 by IRT2 works in cis only (Supplementary Fig. 3), suggesting that the IRT2 RNA itself does not contribute to the repression mechanism. It is possible, however, that the IRT1 and IRT2 nascent transcripts contribute to repression of IME1 and IRT1, respectively. For example, it was shown recently that histone methyltransferases, Set1 and Set2, interact with nascent mRNAs35,36. Interestingly, both enzymes are also important for IRT1-mediated repression of IME1, but not for repression of IRT1 by IRT2 transcription2. More work is needed to dissect roles of nascent RNAs and chromatin factors in facilitating the regulatory circuit consisting of Ime1, IRT2, and IRT1.
Our observations of the IME1 locus raise the question how often transcription of multiple lncRNAs influences the expression of nearby coding genes. In budding yeast, cryptic unstable transcripts (CUTs) and stable unannotated transcripts (SUTs) are widely expressed37,38. Throughout meiosis, CUTs and SUTs, also known as meiotic unannotated transcripts (MUTs), are also transcribed extensively24. The IRT1 and IRT2 transcripts were identified as a SUT and a MUT, respectively. Given that the yeast genome is gene-rich and CUTs are frequently transcribed in the opposite direction nearby of adjacent genes, it is not surprising that a high percentage of genes express at least one CUT or SUT within <400 bp distance of their coding sequence (Supplementary Fig. 11)37,38. Interestingly, we find that 8% of genes express more than two CUTs or SUTs within its proximity. Apart from IME1, the expression of at least two other genes is controlled by multiple lncRNAs, albeit by different mechanisms. At the FLO11 locus, two overlapping lncRNAs are expressed adjacent to the FLO11 gene, where they regulate a toggle switch for FLO11 expression39,40. In fission yeast, a cascade of overlapping lncRNAs is required for the activation of fbp1 expression5. Taken together, regulation of local gene expression by multiple lncRNAs is not limited to IME1 locus and may be widespread.
Like in yeast, in mammalian cells, lncRNAs are also expressed pervasively, and often nearby coding genes41,42,43. Several loci have been described where transcription of a single lncRNA regulates a nearby coding gene15,44,45. In addition, genome-wide approaches have also been used to identify loci where transcription of lncRNAs controls local gene expression4,46. Moreover, loci that express more than one lncRNAs also exits in mammalian cells. For example, the Myc oncogene expresses more than two lncRNAs upstream or downstream of its coding sequence47,48,49,50. While different functions have been attributed to the lncRNAs nearby the Myc gene, perhaps transcription of these lncRNAs influences Myc expression47,48. Thus, the basic features for regulating gene expression via transcription of multiple lncRNAs are present in mammalian cells. Dissecting the mechanisms of how local gene expression is controlled through transcription of lncRNAs in yeast may facilitate the understanding in higher eukaryotes.
Yeast strains and plasmids
Yeast strains used in this paper were derived from the SK1 and the S288C strain backgrounds. The genotypes are listed in Supplementary Table 1. Gene or promoter deletions were generated using the one-step deletion protocol as described previously51. Strains harboring deletions in two or one a1α2-binding sites present in the RME1 promoter resulting in the RME1-H and pRME1-Δaa1 alleles, respectively, were described previously2. The ime1-t mutant harbors a truncation and a 3′ end mutation in IME1 (L325A), which was obtained through one-step integration of a NATmx cassette resulting in a deletion from 50 bp of the 3′-end of IME1 and two nucleotide changes within IME1. The strains harboring RME1 with three copies of V5 at C-terminus was described previously2. Plasmid p158 harboring pIME1-LacZ was previously described52. Plasmid p260 lacking part of the IRT2 sequence was generated by digestion of p158 with EcoRI and EcoNI, the ends were blunted and ligated together to generate pIME1-EIt-LacZ. The IME1 promoter fused to LacZ strains in Fig. 4e and Supplementary Figure 6 were generated by linearizing the plasmids, p158 and p260, with StuI and integrated at the URA3 locus. The IME1 with the N-terminal superfolder GFP (GFP) was generated using a seamless tagging approach as described previously53.
Growth and conditions
All experiments were performed at 30 °C in a shaker incubator at 300 r.p.m. Starvation-induced synchronous sporulation were previously described54. SK1 strains were grown till saturation for 24 h in YPD (1.0% (w/v) yeast extract, 2.0% (w/v) peptone, 2.0% (w/v) glucose, and supplemented with uracil (2.4 mg/l) and adenine (1.2 mg/l)), cells were then diluted at OD600 = 0.4 to pre-sporulation medium (BYTA) (1.0% (w/v) yeast extract, 2.0% (w/v) bacto tryptone, 1.0% (w/v) potassium acetate, 50 mM potassium phthalate) grown for about 16 h, subsequently centrifuged, washed with sterile miliQ water, centrifuged again, and re-suspended at OD600 = 1.8 in SPO (0.3% (w/v) potassium acetate and 0.02% (w/v) raffinose)). S288C strains were directly shifted from saturated YPD to SPO medium following the same procedure. For Fig. 1f and Supplementary Figure 7b, SK1 strains were induced to sporulate by shifting them directly from YPD to SPO following the same procedure.
Cells harboring CUP1 promoter fused to IRT2 or IME1 (pCUP-IRT2 or pCUP-IME1) were treated after 1 h in SPO with copper sulfate when cells were shifted from BYTA to SPO or treated after 2 h when cells were shifted from YPD to SPO. In Fig. 4e and Supplementary Figure 6, 5 μM of copper sulfate was used, whereas for other experiments using the SK1 strain background 25 μM was used. To induce pCUP-IRT2 in the S288C strain background, 50 μM was used.
For the LacZ IME1 promoter induction assay presented in Fig. 4f, cells were grown to saturation in YPD and diluted at OD600 = 0.05 overnight, then at around OD600 = 6–8 cells were washed once with sterile millQ water and re-suspended in SPO at OD600 = 1.8.
For the live-cell imaging presented in Fig. 5 and Supplementary Figure 8, cells were grown to log phase in synthetic complete media (SCD) and loaded into a microfluidic device (Cellasic, Y04C plate) where they were exposed to SCD for an additional 2 h, after which we exposed the cells to YNA (2% potassium acetate, 0.25% yeast extract) for 50 h at constant flow rate (0.6 psi) at 25 °C temperature55.
DAPI staining was used to monitor meiotic divisions throughout time courses. Cells were fixed in 80% (v/v) ethanol, pelleted by centrifugation, and re-suspended in 100 mM phosphate buffer (pH 7) with 1 μg/ml 4′,6-diamidino-2-phenylindole (DAPI). Cells were then sonicated for a few seconds and left in the dark at room temperature for at least 5 min. The proportion of cells containing one, two (meiosis I), three or four (meiosis II) DAPI masses were counted using a fluorescent microscope.
To quantify IRT1, IRT2, and IME1 RNA levels in S288C and SK1 strain backgrounds, we used a reverse transcription combined with quantitative PCR (RT-qPCR). Total RNAs were isolated by acid phenol–chloroform extraction, ethanol precipitated, treated with DNAse, and further column purified (Macherey-Nagel). About 1 μg of total RNAs were reverse-transcribed using random primers and Protoscript II (NEB), and single-stranded complementary DNAs were quantified by real-time PCR using Express SYBR green mix (Life Technologies) on a 7500 Fast Real-Time PCR system (Life Technologies). The signals were normalized to ACT1 mRNA levels. Oligo nucleotide sequences used for RT-qPCR experiments are displayed in Supplementary Table 2.
Chromatin immunoprecipitation experiments were performed as described previously2. Cells were fixed in 1.0% w/v formaldehyde for 25 minutes at room temperature and quenched with 100 mM glycine. Cells were lysed in FA lysis buffer (50 mM HEPES–KOH, pH 7.5, 150 mM NaCl, 1 mM EDTA, 1% Triton X-100, 0.1% Na-deoxycholate, 0.1% SDS, and protease cocktail inhibitor used as recommended by the manufacturer (complete mini EDTA-free, Roche)) using beadbeater (BioSpec) and chromatin was sheared by sonication using a Bioruptor (Diagenode, 8 cycles of 30 s on/off). Extracts were incubated for 2 h with 15 μl of anti-V5 agarose beads (Sigma) or 20 μl of magnetic Prot A beads (Sigma) coupled with 0.5 μg of a polyclonal antibody raised against Histone H3 (Ab1791, Abcam) or histone H3 trimethyl lysine 36 (Ab9050, Abcam), washed twice with FA lysis buffer, twice with wash buffer 1 (FA lysis buffer containing 0.5 M NaCl), and twice with wash buffer 2 (10 mM Tris-HCl, pH 8.0, 0.25 M LiCl, 1 mM EDTA, 0.5% NP-40, 0.5% Na-deoxycholate). Subsequently, reverse cross-linking was done in 1% SDS-TE buffer (100 mM Tris pH 8.0, 10 mM EDTA, 1.0% v/v SDS) at 65 °C overnight. After 2 h of proteinase K treatment, samples were purified and DNA fragments were quantified by real-time PCR using SYBR green mix (Life Technologies) using primers described in Supplementary Table 2. Signals were normalized over the HMR locus, which showed no binding for Ume6 or Rme1. For Supplementary figure 5b, H3K36me3 ChIP signals were normalized to histone H3 ChIP.
Micrococcal nuclease digestion qPCR
To determine the chromatin structure at the IRT2 locus, mononucleosomes were extracted and purified using a MNase digestion protocol that was described previously56. Approximately, 160 OD600 units of cells were crosslinked for 25 min at 30 °C and 250 r.p.m. with formaldehyde (1% v/v). Reaction was quenched with glycine (125 mM). Subsequently, cells were re-suspended in 20 ml of buffer Z (1 M sorbitol, 50 mM Tris-HCl pH 7.4) plus β-mercaptoethanol (10 mM) and treated with 250 μg of T100 Zymolase for 60 min. Next, cells were re-suspended in 1 ml NP buffer (0.5 mM spermidine, 1 mM β-mercaptoethanol, 0.075% (w/v) tergitol solution-type NP-40 detergent, 50 mM NaCl, 10 mM Tris-HCl pH 7.4, 5 mM MgCl2, 1 mM CaCl2), vortexed for 10 s, and 100 μl of extract was treated with 0.2 μl of MNase (2 mg/ml, NEB) for 30 min at 37 °C, the reaction was quenched with EDTA (10 mM), and reverse crosslinked overnight in 1% SDS-TE (100 mM Tris pH 8.0, 10 mM EDTA, 1.0% v/v SDS). Samples were treated with RNase A, purified DNA fragments were separated by gel electrophoresis, and mononucleosome bands were gel purified. MNase-treated and input samples were quantified by qPCR on a 7500 FAST Real-Time PCR machine (Life Technologies) using SYBR green mix (Life Technologies). The scanning primer pairs covering the IRT2 locus and upstream region used for the analysis are available in Supplementary Table 2.
We adapted a northern blot protocol that was described previously57. In short, total RNA was extracted with acid phenol:chloroform:isoamyl alcohol (125:24:1) and precipitated in ethanol with 0.3 M sodium acetate. RNA samples were denatured in a glyoxal/DMSO mix (1 M deionized glyoxal, 50% v/v DMSO, 10 mM NaPi buffer pH 6.5–6.8) at 70 °C for 10 min. Denatured samples were mixed with loading buffer (10% v/v glycerol, 2 mM NaPi buffer pH 6.5–6.8, 0.4% w/v bromophenol blue) and separated on an agarose gel (1.1% w/v agarose, 0.01 M NaPi buffer) for at least 2 h at 80 V. RNAs were then transferred onto nylon membranes overnight by capillary transfer. The ribosomal RNA bands were visualized by methylene blue staining. The membranes were blocked for at least 3 h at 42 °C in hybridization buffer (1% w/v SDS, 50% v/v deionized formamide, 25% w/v dextran sulfate, 58 g/l NaCl, 200 mg/l herring sperm single strand DNA, 2 g/l BSA, 2 g/l polyvinyl-pyrolidone, 2 g/l ficoll, 1.7 g/l pyrophosphate, 50 mM Tris pH 7.5) before hybridization. The radioactive probes were synthesized using a Prime-It II Random Primer Labeling Kit (Agilent), a target-specific DNA template and dATP, [α-32P] (Perkin-Elmer). The oligo nucleotide sequences used to generate target-specific DNA template for amplifying the northern blot probes are displayed in Supplementary Table 2. A representative blot of at least two experimental repeats was used in all figures. The full size uncropped blots are displayed in Supplementary Figure 12.
Single-molecule RNA fluorescent in situ hybridization
The smRNAFISH was performed as described previously2,58. In short, cells were fixed with 3% (w/v) formaldehyde overnight, treated with zymolyase and further fixed in 80% ethanol. Subsequently cells were hybridized for at least 16 h at 30 °C in a buffer (10% v/v formamide, 10% w/v dextran sulfate, 2 mM ribonucleoside vanadyl complexes, 0.02% w/v RNAse-free BSA, 1 mg/ml Escherichia coli transfer RNAs, 0.3 M sodium chloride, 0.03 M sodium citrate) containing fluorophore labeled probes directed to IME1 (AF594) and the internal control ACT1 (Cy5). Cells were pelleted after addition of 1 ml of wash buffer (10% formamide, 0.3 M sodium chloride, 0.03 M sodium citrate), re-suspended in 400 μl of Dapi buffer (5 μg/ml 4′,6-diamidino-2-phenylindole, 10% formamide, 0.3 M sodium chloride, 0.03 M sodium citrate) and incubated 30 min at 30 °C. Cells were then pelleted, re-suspended in 1 ml of wash buffer, incubated 30 min at 30 °C before being pelleted, and re-suspended in 200 μl of 0.3 M sodium chloride, 0.03 M sodium citrate. Cells were imaged using a ×100 oil objective, Numerical aperture (NA) 1.4, on a Nikon TI-E imaging system (Nikon). DIC, DAPI, AF594 (IME1), Cy5 (ACT1) images were collected every 0.3 micron (20 stacks) using an ORCA-FLASH 4.0 camera (Hamamatsu) and NIS-element software (Nikon). ImageJ software was used to make maximum intensity Z projections of the images59. Subsequently, StarSearch software (http://rajlab.seas.upenn.edu/StarSearch/launch.html, Raj laboratory, University of Pennsylvania) was used to quantify transcripts in single cells. Comparable thresholds were used to count RNA foci in single cells. Only cells positive for the internal control ACT1 were used for the analysis. At least a total n = 120 cells were counted for each experiment.
Liquid ortho-nitrophenyl-β-galactoside (ONPG) assay was performed as described previously2. In short, 1 ml of OD600 = 1.8 cells were pelleted and washed once with buffer Z (Phosphate buffer pH 7, KCl 10 mM, MgCl 1 mM) and cells were snap-frozen in liquid nitrogen. Cells were chemically disrupted using Y-PER buffer (Thermo Scientific), cells were incubated with ONPG (Sigma) (1 mg/ml in Z buffer plus 50 mM β-mercaptoethanol) till yellow coloring occurred. The reaction was quenched using sodium carbonate (1 mM) and cell debris were cleared by centrifugation. Absorption of each sample was measured at OD420 using a 96 wells plate reader. Miller units were calculated according to a standard formula: Miller unit = (signal from plate reader (OD420) × 1000)/(cell density (OD600) × time of incubation with ONPG (min)). The data from the experiments represents the SEM of at least two biological experiments.
Fluorescence GFP quantification
For Fig. 4e and Supplementary Figure 6, cells were fixed with 3% (w/v) formaldehyde for 10 min at room temperature, cells were washed once with phosphate-sorbitol buffer (0.1 M kPi (pH 7), 1.2 M sorbitol), re-suspended in phosphate-sorbitol buffer, and stored till imaging at 4 °C. Imaging was performed using a ×100 oil objective, NA 1.4, on a Nikon TI-E imaging system (Nikon), and 500 ms exposure time for the GFP channel. Images were collected using an ORCA-FLASH 4.0 camera (Hamamatsu) and NIS-element software (Nikon). ImageJ software was used for the quantification59. The fluorescence signals were determined by quantifying the whole-cell fluorescence, which was corrected for the background fluorescence signal and auto-fluorescence signal from wild-type cells. The signals represented in Fig. 4e and Supplementary Figure 6 are relative to the 1 h time point.
Cells were imaged using a Zeiss Observer Z1 microscope equipped with a motorized stage, automatic focusing hardware, and temperature control with a ×40 (Zeiss EC Plan-Neofluar ×40/1.3 Ph3 WD = 0.21 M27, oil) phase objective every 12 min. Exposure times were 250 ms for GFP and 120 ms for mCherry using the Zeiss Kolibri LED light system with the 470 nm and 540–580 LEDs at 25% intensity was used with the following filter sets for GFP: excitation: FF01-504/12-25, dichroic: FF518-Di01-25x36, emission: FF01-530/11-25 and for mCherry: excitation: FF01-563/9-25, dichroic: FF573-Di01-25x36, emission: FF01-598/25-25, (all filters from Semrock). Cells were segmented and tracked using a previously published algorithm60. Fluorescent concentrations were calculated by subtracting background signals from the segmented cells and dividing with the approximated cell volumes (cell area). Activation times and slopes were calculated using custom MATLAB software (available upon request). GFP-Ime1 onset time was determined relative to pIME1-Δirt2-mCherry onset that was used to define the time 0. The activation rates were calculated by fitting a line to the concentration curve after the onset and finding its slope.
Determining the proximity of CUTs and SUTs to coding genes
For Supplementary Figure 11, a data set from Wery et al.61 was used. For the analysis. we searched for genes that showed expression of CUTs or SUTs within 400 bp upstream or downstream of the coding sequence (±400 bp). An overlap of 1 bp or more between a gene ±400 bp and a CUTs or SUTs was considered a positive hit in our analysis.
The data generated in the current study are available from the corresponding author.
Publisher’s note: Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.
We are grateful to Michael Knop for the GFP tagging plasmid, and Frank Uhlmann, Paola Scaffidi, Jessica Greenwood, and the members of the CGR lab for their critical reading of the manuscript. This work was supported by CPRIT (RR150058) & the Welch foundation (I-1919-20170325) to A.D. and by the Francis Crick Institute to F.M. and F.J.v.W., which receives its core funding from Cancer Research UK (FC001203), the UK Medical Research Council (FC001203), and the Wellcome Trust (FC001203).