A Complex Gene Regulatory Mechanism that Operates at the Nexus of Multiple RNA Processing Decisions

Abstract

Expression of crs1 pre-mRNA, encoding a meiotic cyclin, is blocked in actively growing fission yeast cells by a multifaceted mechanism. The most striking feature is that crs1 transcripts are continuously synthesized in vegetative cells, but are targeted for degradation rather than splicing and polyadenylation. Turnover of crs1 RNA requires the exosome, similar to previously described nuclear surveillance and silencing mechanisms, but does not involve a non-canonical poly(A) polymerase. Instead, crs1 transcripts are targeted for destruction by a factor previously implicated in turnover of meiotic RNAs in growing cells. Like exosome mutants, mmi1 mutants splice and polyadenylate vegetative crs1 transcripts. Two regulatory elements are located at the 3′ end of the crs1 gene, consistent with the increased accumulation of spliced RNA in polyadenylation factor mutants. This highly integrated regulatory strategy may ensure a rapid response to adverse conditions, thereby guaranteeing survival.

INTRODUCTION

Meiosis is a highly conserved cellular differentiation pathway in which one round of DNA synthesis is followed by two successive rounds of division, leading to the formation of haploid cells from diploid precursors¹. Although the meiotic cell cycle is similar among eukaryotes, the underlying regulatory mechanisms vary widely. In multicellular organisms, extrinsic cues from surrounding cells stimulate germ cells to enter the differentiation pathway, whereas in unicellular organisms such as the budding yeast Saccharomyces cerevisiae and the fission yeast Schizosaccharomyces pombe, entry into meiosis is triggered by nutrient deprivation. In either case, the decision to initiate sexual differentiation is tightly regulated to prevent execution of this specialized cell cycle under inappropriate conditions or in improper developmental contexts.

At the molecular level, meiosis involves a complex cascade of sequential changes in gene expression¹, which in S. pombe are due in part to an extensive program of meiosis-specific pre-mRNA splicing². In contrast to budding yeast, where the transcripts spliced exclusively during meiosis encode proteins involved in chromosome transactions^3,4, meiosis-specific splicing controls production of a wide variety of gene products in fission yeast^2,5. Moreover, at least for crs1 and rem1 RNAs, which encode meiotic cyclins, the block to splicing is biologically relevant, as over-expression in mitotic S. pombe cells is toxic^2,6.

Although regulated splicing in the form of intron retention precludes production of full-length meiotic proteins in actively growing cells of both yeasts, the underlying control mechanisms appear to be distinct. In S. cerevisiae, meiotically spliced RNAs contain intronic enhancer elements that promote splicing via binding to a KH domain RNA binding protein expressed only during meiosis⁴. In contrast, the three meiotically spliced S. pombe transcripts examined to date (mes1, crs1, and rem1) are regulated by non-intronic sequences outside the coding regions, which in both rem1 and mes1 reside upstream^2,6-8. These observations were reminiscent of the changes in alternative splicing patterns upon switching promoters in metazoans^2,8-10. However, the crs1 regulatory mechanism appeared to be distinct, as the flanking regions were unable to prevent splicing of heterologous introns in mitotically growing fission yeast cells, whereas the rem1 flanking regions sufficed to impose meiosis-specific splicing on an otherwise constitutively spliced transcript².

Here, we describe a series of experiments in S. pombe designed to illuminate the intricate molecular mechanism underlying crs1 regulation. As the regulatory sequences lie at the 3′ end of the gene, the mechanism clearly does not involve promoter-driven splicing. The most surprising finding was that increased RNA accumulation during sexual differentiation is not due to up-regulation of crs1 transcription, but rather mirrors alterations in nuclear RNA processing and turnover. Another unexpected feature of the crs1 control mechanism is coupling of polyadenylation and splicing, which was previously believed to occur only in mammals¹¹. We propose that the highly integrated crs1 regulatory strategy serves to “prime” the gene expression pump, allowing a rapid response to adverse conditions that ensures survival.

RESULTS

Accumulation of crs1 RNA does not reflect transcription

To follow up on our finding that critical splicing regulatory element(s) lie outside the crs1 coding region², we set out to determine if transcription of the gene changes during meiosis. To this end, we directly measured RNA synthesis using transcriptional run-on (TRO) assays¹² in the temperature-sensitive pat1–114 mutant, which undergoes ectopic meiosis at the restrictive temperature even as a haploid¹³. Remarkably, crs1 transcription was highest in vegetative cells (0 time point), followed by a steady decline to background levels 5 hr after meiotic induction (Fig. 1a, top). We conclude that the increased accumulation during meiosis must in fact reflect decreased turnover of the RNA. This inference is at odds with the central conclusion from genome-wide microarray and deep-sequencing analyses, which attributed changes in RNA levels between mitotic and meiotic fission yeast cells to transcriptional regulation^14-16.

Figure 1.

Transcription of the crs1 gene does not determine accumulation of the RNA. a, Transcriptional run-on (TRO) analysis of crs1 and intronless control genes over a meiotic time course using the F90 strain containing the pat1-114 mutation (see Supplementary Table 3 online, for the complete genotype), which undergoes ectopic meiosis at high temperature (34°C)²,¹². Cells were harvested at the times indicated after induction of meiosis and the rate of RNA synthesis measured as described previously¹³, with the modifications noted in Supplementary Information. Newly synthesized ³²P-UTP-labelled RNA was detected by hybridization to probes immobilized on nitrocellulose filters followed by PhosphorImager analysis. The decreased background in the 5-8 hr TRO assays reflects the lower overall incorporation during late meiosis (D. S. M. & J. A. W., in preparation), as equal amounts of RNA (by A₂₆₀) were applied to each filter. b, Quantitative real-time PCR (qPCR) analysis of crs1 RNA splicing over a meiotic time course. Total RNA was extracted from the F90 strain carrying a deletion of the crs1 gene² (see Supplementary Table 3 online for the complete genotype), and the levels of unspliced and spliced crs1 RNA at the times indicated after the temperature shift assayed as described in Supplementary Methods (online). The absolute amount of each species in aMol per μg RNA was determined by comparison to signals generated using known amounts of T7 transcripts. All assays were performed in triplicate and error bars indicating standard deviation are shown.

As controls, we performed TRO on two intronless genes. The pta1 gene, encoding the fission yeast orthologue of the polyadenylation factor symplekin¹⁷, displayed a transcription profile similar to crs1 (Fig. 1a, middle). However, in contrast to the spike in the crs1 microarray signal, pta1 RNA accumulation remained constant throughout meiosis¹⁴, a pattern typical of constitutive transcripts (including those encoding other core 3′ processing factors). The transcription profile of meu4, a meiotic gene of unknown function¹⁸, was distinct from both crs1 and pta1, with undetectable synthesis in mitotic cells and a dramatic peak during mid-meiosis (Fig. 1a, bottom).

To determine whether the discrepancy between synthesis and accumulation of crs1 RNA reflected the altered pattern of splicing observed previously², we used quantitative real-time PCR (qPCR) with primers designed to distinguish spliced transcripts from those that retained introns (see Supplementary Methods, online). Notably, the 74-fold difference in total steady-state crs1 RNA levels between vegetative cells and the peak observed during meiosis (∼3 hr post-induction) was attributable almost entirely to an increase in spliced crs1 RNA (Fig. 1b), consistent with processing, not transcription, determining increased expression during meiosis.

Polyadenylation is activated concurrently with splicing

In light of the unexpected discrepancy between the kinetics of transcription and splicing, we examined 3′ processing of crs1 RNA over a meiotic time course (Fig. 2a). Remarkably, vegetative transcripts (0 time point) were not only unspliced, but also lacked poly(A) tails. In meiotic cells, two polyadenylated crs1 species were detected; the larger RT-PCR product peaked at 3 hr, while the smaller one was approximately equal in intensity at 2-4 hr (Fig. 2a). Sequence analysis revealed that the distal crs1 3′ end, which predominates, is located within an element that closely resembles mammalian cleavage and polyadenylation signals^19,20, while the proximal, minor 3′ end is specified by an element that contains a critical deviation (AAUcAA) from the consensus (AAUAAA) mammalian hexamer²¹ (Fig. 2b). Intriguingly, the two polyadenylation signals are contiguous (Fig. 2b & c), and use of the wild-type proximal signal was increased by mutating the signature hexanucleotide to consensus (Supplementary Fig. 1 online). In contrast, 5′ RACE indicated that the transcription start site does not change between mitotic and meiotic cells (Fig. 2c).

Figure 2.

Polyadenylation of crs1 is activated concurrently with splicing during meiosis. a, Analysis of crs1 RNA polyadenylation over a meiotic time course. The presence of a poly(A) tail was assessed by RT-PCR using a 5′ primer within the terminal exon and oligo(dT) as the 3′ primer as shown in the schematic; the proximal (P) and distal (D) polyadenylation signals are indicated as gray and black boxes, respectively. Assays were performed on the same RNA samples analyzed in Fig. 1b and the products visualized negatively after ethidium bromide staining. The product marked with an asterisk gave no sequence with crs1-specific primers and is in fact derived from an intergenic region represented in the SPAC6G9 cosmid; the absence of this non-specific band during mid-meiosis most likely reflects preferential binding of the PCR primers to crs1 RNA. b, Comparison of the crs1 polyadenylation signals to the mammalian consensus. The distances between conserved elements within the mammalian consensus are indicated at the top¹⁹,²⁰. Within the crs1 polyadenylation signals, nucleotides that match the mammalian consensus are shown in uppercase and non-consensus/variable nucleotides are shown in lowercase. The crs1 sequences shown are numbered relative to the stop codon. Arrows mark the sites of crs1 RNA cleavage/polyadenylation determined by sequencing of both cloned cDNAs and RT-PCR products. c, Architecture of the crs1 locus. Promoter/termination regions are shown as lines, untranslated regions (UTRs) as open boxes, and coding exons as filled boxes. The proximal and distal polyadenylation signals are indicated as in Panel a. Arrowheads mark RNA termini determined by RACE, numbered relative to the start and stop codons as in Panel b. d, Analysis of crs1 RNA splicing over a meiotic time course. Splicing was assayed by RT-PCR on the same RNA preparations analyzed for polyadenylation in Panel a, using primers complementary to the terminal exons. SRP RNA served as an internal loading control. e, Concurrent assays of crs1 RNA splicing and polyadenylation over a meiotic time course. The coincidence of the two RNA processing reactions on individual crs1 transcripts was assessed as indicated on the diagram beneath the gel using the same 5′ primer as in Panel d and oligo(dT) as the 3′ primer. Assays employed the same RNA preparations analyzed in Panels a and d. The bands marked with asterisks migrate close to the predicted sizes of unspliced polyadenylated crs1 RNA but gave no sequence with crs1-specific primers.

To confirm that the kinetics of crs1 polyadenylation and splicing precisely coincide during meiosis, we assayed the same RNA preparations for intron removal (Fig. 2d). Like polyadenylation, splicing peaked at ∼3 hr post-meiotic induction, suggesting that the two RNA processing reactions might be coupled. This inference was confirmed by RT-PCR analysis with a 5′ primer complementary to the first exon and oligo(dT) as the 3′ primer, which demonstrated that individual crs1 transcripts had either undergone both RNA processing reactions or were neither spliced nor polyadenylated (Fig. 2e). In further support of coupling between 3′ processing and splicing, unspliced read-through transcripts, but not spliced read-through transcripts, were detected at early time points following meiotic induction (Supplementary Fig. 2 online).

The nearly undetectable level of partially spliced crs1 RNAs throughout meiosis (Fig. 2d) suggested that splicing of all four crs1 introns might be coordinately regulated. Consistent with a control mechanism extending throughout the gene, the predominant species detected when splicing was assayed at more frequent intervals was either full-length precursor or fully spliced RNA (Supplementary Fig. 3a online). Partially spliced products were also virtually undetectable when subsets of the crs1 introns were assayed (Supplementary Fig. 3b-i online). Taken together, these data support a regulatory mechanism that combines continuous transcription of the crs1 gene with suppression of RNA processing in vegetative cells and coupled activation of polyadenylation and splicing during meiosis.

crs1 transcripts are actively turned over in growing cells

As mature mRNAs are generally more stable than unprocessed precursors^22,23, it was conceivable that the increased accumulation of crs1 RNA during meiosis was due solely to relief of the block to splicing and polyadenylation. However, the dramatic difference between the transcription and accumulation profiles of crs1 RNA in mitotic cells (Fig. 1) suggested that an active turnover mechanism might be at work. Based on the increased levels of meiotic transcripts in vegetative cells carrying a deletion of the non-essential cid14 gene, encoding the fission yeast orthologue of the budding yeast Trf4/5 non-canonical RNA polymerase²⁴, we tested this strain for effects on processing and accumulation of crs1 RNA. Notably, crs1 transcripts remained largely unspliced and unpolyadenylated in the cid14Δ strain, as in wild-type cells (Fig. 3a). As Cid14/Trf4 functions in the context of the TRAMP (Trf4, Air2 and Mtr4 polyadenylation) complex to add short poly(A) tails to nuclear substrates destined for destruction^25-27, these results are consistent with our finding that crs1 RNA is not polyadenylated in actively growing S. pombe cells (Fig. 2a).

Figure 3.

The mechanism that blocks crs1 processing in vegetative cells requires the nuclear exosome in conjunction with a factor implicated in turnover of meiotic transcripts in proliferating cells. a, (Top panel) Splicing of crs1 pre-mRNA was assayed by RT-PCR as in Fig. 2d. The relevant genotype of each strain assayed is indicated above the corresponding gel lane; complete genotypes are provided in Supplementary Table 3 (online). All strains were grown at 30°C in EMM2 medium with required supplements⁶⁰. (Bottom panel) Polyadenylation of crs1 pre-mRNA was assayed by RT-PCR as in Fig. 2a, using the same RNA preparations as in Panel a. b,The mmi1-ts6 mutation leads to processing of crs1 RNA in proliferating cells. Mutant cells (see Supplementary Table 3 online, for the complete genotype) were grown to early log phase at the permissive temperature (25°C) and then shifted to the non-permissive temperature (36°C) for the indicated times. (Left panel) RT-PCR assays of splicing (top) and polyadenylation (bottom) were performed as in Fig. 2a & d. (Right panel) Splicing and polyadenylation assays on an isogenic wild-type strain subjected to the same growth regimen as the mutant strain. In the polyadenylation assays shown in both Panels a and b, an asterisk designates a product unrelated to crs1 (see Fig. 2a legend). c, qPCR analysis of spliced and unspliced crs1 in the same RNA preparations analyzed in Panel b.

The exosome, an assemblage of 3′-5′ exonucleases, has also been implicated in nuclear quality control mechanisms that affect multiple classes of RNA²⁵. To investigate a potential role for the exosome in determining the fate of crs1 RNA, we assayed splicing and polyadenylation in fission yeast cells harboring a deletion of the rrp6 gene. Although crs1 transcript accumulation was higher in this strain than in the wild type or cid14Δ mutant (Fig. 3a), the absence of Rrp6 also alleviated the block to both splicing and polyadenylation that normally occurs in vegetative cells (Fig. 3a). Further evidence that the nuclear exosome plays either a direct or indirect role in processing as well as turnover decisions for crs1 is provided by the increased splicing and RNA accumulation observed in the cold-sensitive dis3-54 mutant²⁸ (Supplementary Fig. 4 online), encoding the catalytically active subunit²⁹. Regulation of crs1 thus contrasts with heterochromatic silencing, where the nuclear exosome and Cid14 function together³⁰.

Mmi1 regulates crs1 RNA processing as well as turnover

In seeking to identify the mechanism by which crs1 RNA is targeted to the nuclear exosome, we noted that crs1 was among the dozen transcripts reported to yield stronger microarray signals at the non-permissive temperature in the mmi1 mutant, the only known trans-acting factor required for selective elimination of meiotic transcripts from vegetatively growing fission yeast cells³¹. To test whether Mmi1-mediated turnover and regulated processing of crs1 RNA might be manifestations of the same underlying control mechanism, we assayed both available temperature-sensitive mutants for relief of the block to splicing and polyadenylation in vegetative cells. Consistent with our hypothesis, crs1 RNA was fully processed at the non-permissive temperature in mmi1-ts6 (Fig. 3b, left panel) and mmi1-ts3 (Supplementary Fig. 5 online). The bypass of negative regulation was not due to elevated temperature alone, as the RNA was neither spliced nor polyadenylated in an isogenic wild-type strain subjected to the same growth regimen (Fig. 3b, right panel). Moreover, qPCR revealed that the increased accumulation of crs1 RNA in the mmi1-ts6 strain largely reflected an elevation of spliced RNA levels (Fig. 3c), similar to meiotic induction (Fig. 1b). The less dramatic difference (∼8.5- vs. 74-fold) may reflect residual Mmi1 activity in the conditional mutants and/or the existence of additional layers of regulation.

Mutating polyadenylation factors disrupts crs1 regulation

Based on the presence of two differentially utilized 3′ processing signals in crs1, we hypothesized that one or more polyadenylation factors might also function in the regulatory mechanism. An excellent candidate was Pfs2, a WD repeat-containing protein functionally related to mammalian CstF-50³², in which a temperature-sensitive mutation (pfs2-11) conferred a chromosome segregation defect similar to the one observed in mitotic cells over-producing spliced crs1 mRNA^2,33. Consistent with a role for Pfs2 in crs1 regulation, shifting pfs2-11 (Fig. 4a, left panel), but not an isogenic wild-type strain (Fig. 4a, right panel), to the non-permissive temperature led to splicing of crs1 RNA even in non-meiotic cells. Futhermore, real-time PCR indicated that crs1 RNA accumulation is highly elevated in this strain (Fig. 4b), comparable to the mmi1-ts6 strain (Fig. 3c).

Figure 4.

The pfs2-11 mutation leads to splicing and increased accumulation of crs1 RNA in proliferating cells. a, (Left panel) RT-PCR analysis of crs1 RNA splicing in the pfs2-11 polyadenylation factor mutant (see Supplementary Table 3 online for the complete genotype). (Right panel) Splicing assays on an isogenic wild-type strain. Both strains were grown to early log phase at the permissive temperature (26°C) and then shifted to the non-permissive temperature (36°C) for the indicated times. b, qPCR analysis of spliced and unspliced crs1 in the same RNA samples analyzed in Panels a & b. c, Transcripts from 3 hr pfs2-11 (36°C) time-point that extend beyond the polyadenylation sites in the crs1 gene were detected using a 5′ primer complementary to Exon 5 in combination with 3′ primers complementary to sequences at the indicated positions downstream from the stop codon. The product marked with an asterisk in lane 1 is described in the legend to Fig. 2a, while none of the other bands yielded sequence data with crs1-specific primers. In lanes 2-7, the product migrating at the predicted size for the primer pair employed in each lane is indicated by a dot.

The increased accumulation of crs1 RNA in the pfs2-11 strain was surprising in that 3′ processing mutants generally led to mRNA destabilization in budding yeast³⁴. Nevertheless, in vivo assays indicated that this mutation is defective for constitutive polyadenylation³³, consistent with our inability to detect a poly(A) tail on the crs1 RNA produced in this strain (Fig. 4c, lane 1). Instead, read-through transcripts indicative of polyadenyation/transcription termination failure³⁵ were observed, in most cases at a level comparable to the 3′ UTR control (Fig 4c, compare lane 2 with lanes 3-7). A second conditional mutant with a chromosome segregation defect similar to crs1 over-expression, dhp1-1³⁶, also led to splicing of crs1 RNA combined with read-through transcription (Supplementary Fig. 6 online). The dhp1 gene encodes the fission yeast orthologue of Rat1/XRN2, a 5′ to 3′ exonuclease implicated in coupling cleavage/polyadenylation to RNA polymerase II transcription termination in both budding yeast and mammals^37,38. Consistent with the similar effects of dhp1-1 and pfs2-11 on crs1 RNA processing and accumulation, recent data suggest an additional, earlier function for Rat1 in enhancing recruitment of 3′ processing factors³⁹.

The proximal polyadenylation signal plays a regulatory role

To map the element(s) responsible for inhibiting crs1 splicing during vegetative growth, we took advantage of our previous observation that intron retention was relieved upon expressing the coding region under heterologous control². To circumvent the deleterious effects of expressing crs1 from the nmt1 vector², a thiamine-responsive expression system, we inserted the coding region into an attenuated version, nmt81^40,41 (see Methods), with the net result that the native DNA extending ∼1kb upstream from the start codon and downstream from the stop codon was replaced with vector sequences (designated NN-NN in Fig. 5a). As a control, a plasmid-borne allele carrying the same amount of native crs1 flanking DNA was also constructed (designated CC-CC). Cells harboring this allele have a growth rate similar to wild type and are morphologically normal².

Figure 5.

The proximal (non-canonical) polyadenylation signal regulates splicing and accumulation of crs1 RNA. a, Schematic diagrams indicating the regions interchanged in nmt81-crs1 chimerae, which are demarcated by brackets and designated by numbers. b, RT-PCR splicing assays on nmt81-crs1 chimerae. The letters above each lane on the gel indicate the source of the segments flanking the ORF, which correspond to the numbers on the schematic, with a dash representing the coding region. Heterologous flanking segments were derived from a thiamine-responsive expression plasmid in which the inserted ORF is surrounded by the nmt81 transcriptional and translational control regions⁴⁰,⁴¹. Analyses were carried out in the crs1 deletion strain JLP970² to ensure that all products were derived from the plasmid-borne alleles. c, Splicing assays on crs1 polyadenylation signal mutants. Lane 1: wild-type control. Lane 2: NN-CN construct (parent for the remaining alleles). Lanes 3 and 4: either the proximal (non-canonical) or distal (canonical) 57 nt polyadenylation signal (Fig. 2b) was replaced with complementary sequences. d, qPCR analysis of spliced and unspliced crs1 in the same RNA samples analyzed in Panels a-c, plus two additional chimerae (9 & 10). The percentage of spliced crs1 RNA is indicated next to each allele designation.

In addition to the parental and wild type alleles, seven chimeric genes carrying different combinations of crs1 and nmt81 sequences in their promoter/5′UTR and 3′UTR/terminator regions were introduced into mitotically growing fission yeast cells and assayed for splicing (Fig. 5b). The recipient carried a deletion of the endogenous crs1 gene to ensure that all signals in the assays were from plasmid transcripts. Consistent with a role for the polyadenylation machinery in regulation (Fig. 4 and Supplementary Fig. 6 online), substituting the sequences downstream, but not upstream, from the crs1 ORF led to removal of all four introns in vegetative cells (Fig. 5b, lanes 3 & 4). More precise mapping revealed a critical role for the region encompassing the 3′ untranslated region (UTR) including both polyadenylation signals (lanes 5-8) in blocking accumulation of spliced crs1 RNA in vegetative cells.

Given the unusual arrangement of polyadenylation sites, we next tested whether either the proximal (non-canonical) or distal (canonical) signal (Fig. 2b) present in this region is relevant to the inhibitory mechanism. Each 57 nt segment was separately replaced with complementary sequences in the context of heterologous or homologous flanking DNA. When the proximal (non-canonical) signal was ablated in a construct in which only the coding region and the segment immediately downstream were derived from crs1 (NN-P’N), the inhibition of crs1 splicing normally observed in mitotic cells was almost completely relieved (Fig. 5c, lane 3). In contrast, crs1 RNA remained largely unspliced when the distal (canonical) polyadenylation signal was mutated (lane 4), as did the RNA produced from the wild type and parental alleles (lanes 1 and 2). Parallel but less dramatic results were obtained with a series of constructs in which the polyadenylation signals were mutated in the context of native crs1 flanking DNA (Supplementary Fig. 7 online).

Based on the indistinguishable kinetics of increased crs1 RNA accumulation and activation of splicing during meiosis (Fig. 2a & d), we asked whether substituting or mutating either polyadenylation signal increased the steady-state level of crs1 RNA in vegetative cells. Strikingly, the proximal (non-canonical) polyadenylation signal mutant (NN-P’N) produced considerably more RNA than either the wild type (CC-CC) or parental alleles (NN-CN; Fig. 5d, compare c3 with b1 & b8). Moreover, as during meiosis (Fig. 1b), the increase was due almost entirely to a higher level of spliced RNA. In contrast, ablating the distal (canonical) polyadenylation signal depressed the level of both spliced and unspliced RNA (Fig. 5d, compare b8 & c4). We conclude that the proximal (non-canonical) polyadenylation signal is important not only for preventing splicing, but also contributes to restricting accumulation of mature crs1 RNA. This conclusion is bolstered by quantitative comparisons of splicing and RNA accumulation between chimeras that differ in only a single segment (Supplementary Tables 1 and 2 online).

The proximal polyadenylation signal controls 3′ processing

Given the coupling of crs1 RNA splicing and polyadenylation during meiosis (Fig. 2e), we predicted that the non-canonical polyadenylation signal mutation (NN-P’N) would also lead to polyadenylation of plasmid-derived transcripts. Indeed, this allele produced readily detectable levels of RNA polyadenylated at the distal site (Fig. 6a, lane 3). In contrast, polyadenylated crs1 RNA was absent in the distal (canonical) poly(A) signal mutant (lane 4), similar to wild type (lane 1), while a very faint signal was detected in the parental allele (NN-CN) (lane 2). Parallel results were observed when splicing and polyadenylation were assayed concurrently (Fig. 6b), providing additional evidence for mechanistic coupling.

Figure 6.

The proximal (non-canonical) polyadenylation signal also regulates 3′ processing of crs1 RNA. a, Assays for the presence of a poly(A) tail in crs1 polyadenylation signal mutants. Polyadenylation was assayed as in Fig. 2a using the same RNA samples assayed for splicing in Fig. 5c. Sequence analysis indicated that the band visible in lane 3 arose via use of the distal (canonical) polyadenylation signal. The absence of the non-specific band observed in Fig. 2a may reflect production of higher levels of crs1 RNA from multi-copy plasmids vs. the single chromosomal copy. b, Concurrent assays of splicing and polyadenylation in crs1 poladenylation signal mutants. The primers employed were the same as in Fig. 2e. c, Assays of read-through transcription for the same mutants assayed in Panels a and b and a wild-type control. The products are of slightly different sizes (compare lane 1 with lanes 2 & 4) because, as indicated in the schematic, the 3′ primer complementary to crs1 sequences in the wild-type allele was slightly farther downstream from the 5′ primer than was the 3′ primer complementary to nmt81 sequences for the mutants. d, Assays to detect extended read-through transcripts for wild-type crs1. (Left panel) Plasmid-borne allele (CC-CC); (Right panel) Chromosomal allele. The 5′ primer was the same one used to assay polyadenylation while the 3′ primers were complementary to sequences within and downstream from the distal polyadenylation site. The distance from the stop codon to the end of the downstream primer is indicated above each lane of the gels and beneath the schematic diagram.

The absence of a signal in the polyadenylation assay indicated that wild-type crs1 transcripts present in actively growing cells either lack poly(A) tails altogether or possess short tails not detected under the conditions employed. To distinguish these possibilities, we tested the mutants for read-through transcription³⁵. Transcripts extending beyond the polyadenylation signals were observed in the proximal polyadenylation signal mutant (Fig. 6c, lane 3) but not in any of the other alleles (lanes 1, 2 and 4). Moreover, using a common 5′ primer and 3′ primers complementary to sequences at varying distances downstream, read-through transcripts extending over 500 nt beyond the distal polyadenylation site were detected from both the plasmid-borne allele and the endogenous crs1 locus (Fig. 6c). Thus, the crs1 transcripts present in vegetative cells not only lack poly(A) tails, but are apparently not cleaved or terminated either at the polyadenylation sites used during meiosis (Fig. 2b) or for a considerable distance downstream.

A second regulatory element resides in the terminal exon

We showed previously that, although the DNA flanking the crs1 coding region is necessary to block splicing of the native RNA during mitotic growth, it is not sufficient, as expression of otherwise constitutively spliced heterologous transcripts under crs1 control did not block their splicing in actively growing cells². Thus, the non-canonical polyadenylation signal is likely to function in collaboration with one or more elements that reside within the crs1 ORF. We initially focused our search for additional regulatory sequences on the crs1 introns, as each contains splicing signals that deviate from consensus (Supplementary Fig. 8a online); however, deletion analysis indicated that the non-canonical splice sites are not crucial for inhibition of splicing in vegetative cells (Supplementary Fig. 8b online).

Using coupled splicing/polyadenylation assays on chimeras between the crs1 and SPAC6F6.11c genes, the second regulatory element was shown to reside within an exon. SPAC6F6.11c encodes a constitutively processed transcript similar in overall architecture to crs1 that was previously shown to be spliced after replacement of its flanking regions with crs1 sequences². Substitution of crs1 Exons 2 through 5 and their preceding introns with SPAC6F6.11c sequences allowed RNA processing in mitotic cells (Fig. 7, lanes 2-5), while a control construct in which only the fourth intron was substituted produced no detectable fully processed RNA (lane 6), similar to wild-type crs1 (lane 1). Because all chimeras containing crs1 Exon 5 failed to yield product in this assay, we analyzed the effects of smaller substitutions near the 3′ end of the ORF. These experiments localized the second negative regulatory element to the penultimate 48 nt of the crs1 coding sequence (compare lane 7 to lanes 8-11).

Figure 7.

An element within the fifth exon coordinately inhibits crs1 splicing and polyadenylation. Chimerae carrying the exon/intron combinations indicated at the bottom were analyzed using a coupled splicing/polyadenylation assay as in Figs. 2e & 6b. Introns derived from crs1 point upwards and those derived from SPAC6F6.11c point downwards and are shown in bold. A bracket beneath constructs 1, 6 & 7 indicates the location of the inhibitory sequence in Exon 5.

The sequence of the Exon 5 element provides no obvious hints as to how it might function to regulate processing of crs1 transcripts (Supplementary Fig. 9a online). Nevertheless, assays of splicing alone also supported such a role, as all alleles lacking the native element produced spliced crs1 RNA, albeit in variable yields (Supplementary Fig. 9b online). The similar effect of substituting this 48 nt region of Exon 5 and mutating the first crs1 regulatory element that we discovered, the proximal (non-canonical) polyadenylation signal, prompted us to ask whether they might function collaboratively. Consistent with this hypothesis, increasing the distance between them compromised regulation (Supplementary Fig. 9c online).

Reversing polyadenylation signals compromises regulation

The presence of tandem polyadenylation signals in the crs1 gene provided an opportunity to test the effect of switching the order in which they are encountered during transcription by RNA polymerase II. Strikingly, the Switch mutant (NN-SwN) RNA is polyadenylated (Fig. 8a, lane 1) and fully spliced (Fig. 8b, lane 3), buttressing the case for coupling of these RNA processing events. Moreover, 3′ processing of Switch mutant transcripts was efficient as judged by the absence of read-through transcription (Fig. 8a, lane 3) indicative of polyadenylation/termination failure³⁵. The dramatic effect of the Switch mutation strongly implies that crs1 RNA is not committed to splicing until the polyadenylation signals have been traversed by RNA polymerase II. Curiously, however, qPCR revealed that despite being fully processed, the amount of Switch mutant RNA produced was quite low (74.6 aMol/μg RNA), comparable to wild-type crs1 or a mutant containing the native 3′ flanking region combined with nmt81 5′ DNA (Fig. 5d, lanes b1 & b8; Supplementary Table 2 online). A plausible interpretation is that the Exon 5 element targets crs1 RNA for elimination, and only transcripts that escape this fate are processed.

Figure 8.

Effect of reversing the order of the polyadenylation signals on crs1 RNA processing. The proximal (non-canonical) polyadenylation signal (P) was transplanted downstream from the distal (canonical) signal (D) as described in Supplementary Information (Switch mutant). Note that this manipulation increases the distance of the non-canonical polyadenylation signal from the Exon 5 element by 57 nt. a, Analysis of 3′ processing in the Switch (NN-SwN) mutant. Lane 1: the presence of a poly(A) tail was assayed as in Fig. 2a. Lane 2: crs1 RNA was detected independent of splicing or polyadenylation using a 3′ primer complementary to the 3′ UTR. Lane 3: read-through transcription was assayed by RT-PCR with the same 5′ primer used to assay polyadenylation and a 3′ primer complementary to the nmt sequence 68 nt downstream from the distal polyadenylation signal (see diagram). b, Analysis of splicing in the Switch mutant (NN-SwN) and controls are as denoted in Fig 5. Percent splicing for each allele determined from qPCR data is shown beneath the gel.

DISCUSSION

It is now widely appreciated that the nuclear events leading from synthesis of a primary transcript to export of a mature mRNA to the cytoplasm are carefully orchestrated through a series of dynamic interactions involving multiple machines within the gene expression factory^42,43. To date, the evidence for communication between the transcription apparatus and components of the splicing, polyadenylation and turnover machineries has come principally through biochemical and genetic studies of constitutively expressed genes. By analyzing a transcript that is subject to regulation, we have provided insight into how the integration of nuclear events impacts a key biological process, namely meiotic differentiation, and in the process made several unexpected observations.

We did not expect to find that the major crs1 3′ processing signal used during meiosis closely resembles cleavage and polyadenylation signals found in higher eukaryotes, which was unanticipated in light of earlier reports that polyadenylation in fission yeast is mediated by highly degenerate elements similar to those found in budding yeast^44,45. The apparent discrepancy may be specific to the particular genes examined, or to the emphasis on transcription termination rather than polyadenylation per se in the earlier studies^44,45. Consistent with our finding that use of the non-canonical signal is increased upon mutating it to the mammalian consensus, it has been reported that a plant polyadenylation signal containing a consensus hexanucleotide is recognized in fission yeast⁴⁶.

Even more surprising was our discovery that polyadenylation of crs1 is mechanistically linked to splicing, as coupling between these RNA processing reactions had heretofore been observed only in metazoans, and was believed to be absent from organisms that use intron definition for the initial pairing of splice sites¹¹. It will be interesting to learn whether splicing and polyadenylation are coordinated by a conserved mechanism, as only the 3′-most intron appears to be affected in mammals^47,48 whereas all four crs1 introns are activated for splicing concurrent with crs1 polyadenylation during meiosis (Fig. 2d) and also display parallel responses to cis- and trans-acting mutations (Figs. 3-8).

Most unprecedented was our finding that transcription of the crs1 gene actually decreases steadily as cells enter meiotic differentiation and is near background levels when RNA accumulation peaks (Fig. 1). This observation provided the first indication that the dramatic increase in accumulation of crs1 transcripts in meiotic cells is not merely a passive consequence of proper maturation but instead reflects an active mechanism for destroying crs1 RNA (and presumably other mmi1-sensitive transcripts³¹) in mitotic cells. The cis-acting signals recognized by Mmi1 in mei4, rec8, ssm4 and spo5, designated “determinants of selective removal” (DSR’s), bear no obvious resemblance to each other³¹ nor to the non-canonical polyadenylation signal or Exon 5 element in crs1 (Fig. 2b and Supplementary Fig. 9a online). However, both the crs1 regulatory elements and the majority of known DSR’s are located near the 3′ ends of the genes³¹. Conceivably, a conserved motif may emerge once the other elements are mapped more precisely; alternatively, it is possible that Mmi1 binds RNA in a sequence-independent manner. Our results significantly extend those reported earlier³¹ in that, although Mmi1 is a member of the YTH family of RNA binding proteins related to factors implicated in alternative splicing in mammals⁴⁹, it was not previously known to affect splicing or polyadenylation.

The 3′ location of the crs1 regulatory elements has two important implications. First, it distinguishes the regulatory mechanism from two previously studied fission yeast genes (rem1 and mes1), in which meiosis-specific splicing is enforced by sequences that lie upstream from the coding regions^7,8. While it remains to be determined whether rem1 and mes1 regulation are related to the well-established but as yet poorly understood ability of promoter identity to influence alternative splicing patterns in metazoans⁵⁰, the shared 5′ location is suggestive. Second, the downstream location of the crs1 control element implies that commitment of this transcript to splicing is delayed until its synthesis is virtually complete, which is remarkable in that the first intron is located over a kilobase away from the non-canonical polyadenylation signal and the Exon 5 element (see Fig 2c). One tantalizing possibility is that processing of crs1 RNA is promoted by juxtaposition of distant segments of the gene during meiosis, analogous to the looping observed during transcriptional induction in S. cerevisiae^51,52.

Fig. 9 shows a working model for the roles of both cis-acting signals and trans-acting factors in regulating the maturation and accumulation of crs1 RNA in wild-type mitotic and meiotic cells as well as the mmi1-ts mutant. In actively growing cells (left panel), we propose that Mmi1 nucleates formation of a complex on nascent crs1 RNA that includes the exosome, consistent with the similar molecular signatures (increased accumulation and processing of crs1 RNA) that we observe in the rrp6Δ and mmi1-ts mutants, as well as a two-hybrid Mmi1-Rrp6 interaction³¹. In that deletion of a TRAMP complex subunit has no discernible impact on crs1 RNA processing or accumulation, the most parsimonious explanation is that in the specialized turnover pathway that blocks production of meiotic mRNAs in mitotic S. pombe cells, Mmi1 plays a role analogous to TRAMP-mediated polyadenylation in other nuclear surveillance and silencing pathways^27,30.

Figure 9.

Model to explain the different processing fates of the crs1 RNA during vegetative growth and meiotic differentiation of wild-type cells and in the mmi1-ts mutants. The symbols representing the various factors implicated in the crs1 regulatory mechanism are indicated in the key beneath the diagrams. The question mark over the RNA polymerase II body indicates that it has not been directly implicated in the crs1 regulatory mechanism. As Pfs2 orthologues do not interact with RNA³² whereas orthologues of Dhp1 do³⁹, these factors are arranged accordingly in the diagram; alternatively or in addition, it is possible that other polyadenylation factors are involved in the crs1 regulatory mechanism. During meiosis and in the mmi1-ts mutants (middle and right panels), where the nascent RNA is not destroyed, the exosome is depicted as remaining associated with the elongating Pol II even though this has not been directly demonstrated.

Although we have not definitively proven that the crs1 regulatory mechanism operates co-transcriptionally, several lines of circumstantial evidence support this idea. By analogy to the known association of the exosome with RNA polymerase II in Drosophila⁵³, it seems plausible to suggest this mode of delivery in crs1. Co-transcriptional assembly of the complex that regulates turnover and processing is also consistent with the effect of reversing the order of the polyadenylation signals, as well as the involvement of Pfs2, a polyadenylation factor functionally related to mammalian CstF-50, which is known to bind the Pol II CTD⁴². The involvement of dhp1 in crs1 regulation (Supplementary Fig 6.) provides perhaps the most compelling evidence for a co-transcriptional mechanism, as its orthologues rat1/XRN2 link transcription termination with polyadenylation^37,38. Finally, several meiotic transcripts were spliced upon lowering the gene dosage of a fission yeast cyclophilin that modulates Pol II CTD phosphorylation⁵⁴, although crs1 was not among the RNAs examined.

The middle panel of Fig. 9 depicts the proposed molecular events in wild-type S. pombe cells undergoing sexual differentiation. While it remains to be determined whether Mmi1 influences the delivery, stabilization or activation of the exosome in mitotic cells, in meiotic cells it cannot perform this role due to sequestration by the Mei2 protein, a key regulator of sexual development in fission yeast^1,31. The functional inactivation of Mmi1 in differentiating cells renders it unavailable to bind nascent crs1 transcripts, thus allowing them to undergo normal 3′ maturation and splicing, presumably followed by export of the mature mRNA to the cytoplasm and translation to produce the Crs1 cyclin. In mmi1-ts mutant cells (right panel), we propose a similar sequence of events except that a mutation rather than sequestration prevents Mmi1 from associating with newly synthesized crs1 RNA. The molecular events in the Switch mutant are more speculative, but a plausible interpretation of the low yield of fully processed RNA is a bifurcating pathway in which destruction of the RNA competes with processing (Supplementary Fig. 10 online).

The precise role of Pfs2 is even more enigmatic, as the pfs2-11 mutation affects constitutive as well as regulated polyadenylation, and in the context of crs1 uncouples this RNA processing event from splicing. One possible explanation is that splicing in pfs2-11 (and dhp1-1/rat1, which behaves similarly) occurs post- rather than co-transcriptionally. The involvement of Pfs2 in crs1 regulation suggests a link to control mechanisms discovered in more complex eukaryotes, as putative orthologues of this protein (CstF-50 paralogues) have been implicated in the regulation of Arabidopsis flowering⁵⁵ and display increased expression during mouse spermatogenesis⁵⁶. A key difference, however, is that previous studies of regulated 3′ processing, including two (non-meiotic) examples in S. pombe^20,57-59 described switching between sites rather than the differential activation implied by our data (Fig. 2a). Based on these precedents, we fully expected a positive control mechanism in which a meiosis-specific factor would allow use of the proximal (non-canonical) polyadenylation signal; however, all of our data point instead to a negative mechanism that blocks the use of both sites in vegetative cells.

The complex, and to our knowledge unprecedented, mechanism that up-regulates crs1 gene expression independent of transcriptional induction raises an obvious question: why do vegetative fission yeast cells continuously synthesize crs1 RNA, only to destroy it? We propose that this regulatory strategy, in addition to preventing the deleterious production of a meiotic protein in growing cells², serves to “prime” the gene expression pump, ensuring a very rapid response to adverse conditions in which mating, meiosis and sporulation are the organism’s sole means of survival. Given that some elements of the crs1 regulatory mechanism have antecedents in other organisms, it is tempting to speculate that selective RNA turnover pathways may contribute more generally to changes in gene expression in response to environmental or developmental cues.

METHODS

S. pombe manipulations

See Supplementary Table 3 (online) for complete genotypes and sources for the strains used in this study. Fission yeast media and growth conditions were as described earlier². We induced ectopic meiosis in the pat1-114 mutant (Figs. 1 & 2) by shifting the fission yeast strain F90 to the non-permissive temperature (34°C) as described previously². Growth conditions for other strains are described in the figure legends.

Oligonucleotides

See Supplementary Table 4 (online) for sequences of all oligonucleotides used in this study; these are grouped according to the purpose for which they were used.

Mapping the crs1 termini

To determine the 5′ and 3′ ends of crs1 RNA, we performed RACE (Rapid Amplification of cDNA Ends) on RNA extracted from vegetatively growing or meiotic (4 h at 34°C) F90 cells. Products visible by ethidium bromide staining were cloned and sequenced by the facility at SUNY, Stony Brook.

Construction of chimeric and mutant plasmids

Due to the large number of constructs analyzed and their complexity, the relevant information is provided in Supplementary Methods online.

RNA processing assays

Processing of crs1 RNA was assayed by both standard and real-time (quantitative) PCR. Standard (semi-quantitative) RT-PCR splicing assays were performed as described earlier² using crs1-SAF (oligo #23) and crs1-SAR3 (oligo #68) as primers. Real-time PCR (qPCR) splicing assays were performed and the data analyzed as described in Supplementary Methods online. RT-PCR assays of polyadenylation were performed similarly except that, depending on the experimental question, two different 5′ primers (crs1-RC3′, oligo #28, or crs1-SAF, oligo #23) were used in combination with oligo(dT)₂₄ (oligo #69) as the 3′ primer. The 3′ UTR primer used to detect crs1 RNA independent of polyadenylation state was crs1-nmtTRNew (oligo #17). To assay for read-through transcription, the 3′ primers were crs1-CER (oligo #70) or nmt-SEQ B². All RNA processing assays were repeated 2-4 times using independent transformants in the case of plasmid-based experiments.

Transcription Run-On (TRO) analysis

S. pombe TRO analysis was performed essentially as described¹³ with the following modifications. Cells were permeabilized in 0.6% Sarcosyl (w/v) for 25 minutes with constant mixing on ice. Following this step, transcription was carried out at 30°C for 6 minutes. Equal numbers of micrograms of ³²P-UTP-labeled RNA from each time point was hybridized to filters on which 1.4 pmole of each denatured PCR fragment had been immobilized. The probes, which correspond to the 3′ ends of the coding region for each gene analyzed, were made by PCR using the following oligonucleotide primers (Supplementary Table 4 online): for crs1, crs1-nmtTR3New (oligo #19 and crs-SPAC6F-Ex5(N) (oligo #59); for pta1, Pta1-PF (oligo #83) and Pta1-PR (oligo #84); and for meu4, Meu4-DF (oligo #85) and Meu4-DR (oligo #86).