Diverse mechanisms for spliceosome-mediated 3′ end processing of telomerase RNA

The 3′ end of Schizosaccharomyces pombe telomerase RNA (SpTER1) is generated by spliceosomal cleavage, a reaction that corresponds to the first step of splicing. The observation that the spliceosome functions in 3′ end processing raised questions about the evolutionary origin and conservation of this mechanism. We now present data in support of spliceosomes generating 3′ ends of telomerase RNAs in other fungi. Strikingly, the mechanistic basis for restricting spliceosomal splicing to the first transesterification reaction differs substantially among species. Unlike S. pombe, two other fission yeasts rely on hyperstabilization of the U6 snRNA—5′ splice site interaction to impede the 2nd step of splicing. In contrast, a non-canonical 5′ splice site blocks the second transesterification reaction in Aspergillus species. These results demonstrate a conserved role for spliceosomes functioning in 3′ end processing. Divergent mechanisms of uncoupling the two steps of splicing argue for multiple origins of this pathway.

T elomerase is the ribonucleoprotein (RNP) complex that synthesizes telomeric repeats and maintains telomere length homeostasis. The catalytic core of telomerase is composed of a non-coding RNA (TER) that contains a template region and a protein subunit (TERT) that reiteratively reverse transcribes the template to add telomeric repeats onto chromosome ends 1 . Insufficient telomerase activity is causative in a number of degenerative syndromes where premature depletion of stem cell pools results in pleiotropic phenotypes, including nail dystrophy, aplastic anaemia and pulmonary fibrosis 2 . Conversely, telomerase is active in the majority of cancer cells and is required for their continued proliferation 3 . Telomerase inhibition may thus constitute a promising avenue in cancer treatment. While much has been learned over the past decade about the regulation of telomerase expression and activity, we know much less about the biogenesis of the RNP 4 .
The fission yeast Schizosaccharomyces pombe has served as a valuable model system for studying telomerase. The identification of human TERT and Pot1 proteins was aided by the prior characterization of the fission yeast homologues 5,6 , and analysis of telomerase biogenesis in S. pombe has elucidated an ordered sequence of events that culminate in the assembly of the functional RNP 7,8 . The extent to which these events are conserved in other species remains to be determined, but key aspects including cap hypermethylation and association with Sm proteins have been reported for human telomerase as well 9,10 . The S. pombe telomerase RNA subunit TER1 (refs 11,12) is transcribed as a precursor harbouring an intron immediately downstream of its mature 3 0 end 7 . Instead of the intron being removed by two consecutive transesterification reactions, the spliceosome only carries out the first cleavage reaction and releases the 5 0 exon to become the functional form of telomerase RNA.
In the first step of splicing, the 2 0 hydroxyl group of the branch point (BP) adenosine, located within the branch site (BS) sequence, attacks the sugar-phosphate bond at the 5 0 splice site (5 0 SS), forming a 2 0 -5 0 linkage and producing a free 5 0 exon and a branched species, the lariat intermediate. During intron excision, this first reaction is followed by a conformational rearrangement that brings the now exposed 3 0 hydroxyl of the upstream exon into close proximity with the 3 0 splice site (3 0 SS). A second transesterification reaction yields ligated messenger RNA (mRNA) and the lariat form of the intron. The release of the splicing intermediate as a functional product in the case of SpTER1 is triggered by an unusual combination of two intronic features-a long distance between the BP and 3 0 SS and a BS sequence that is fully complementary to the binding site in U2 small nuclear RNA (snRNA) 13 . These features attenuate the transition from first-to second-step conformation and result in the release of the splicing intermediates. Mechanistically, this process is akin to 'discard', a pathway that has been characterized through a series of elegant experiments in budding yeast where the DExD/H-box helicase Prp22 functions in the recovery of spliceosomes that have selected suboptimal 3 0 splice sites 14,15 . Up to this point, SpTER1 has been the only known example of the discard pathway generating a functional product.
The recent discovery of the Nrd1/Nab3 pathway mediating 3 0 end formation for telomerase RNA in budding yeast 16,17 has raised the question whether spliceosomal cleavage as a mechanism of 3 0 end processing is restricted to S. pombe. Here we show that telomerase RNAs from two other fission yeasts and from filamentous fungi contain introns that undergo spliceosomal cleavage. Interestingly, while the use of the first step of splicing to generate telomerase RNA 3 0 ends is conserved, the RNA elements that uncouple the two steps of splicing are entirely different.

Results
To examine whether the mechanism of telomerase RNA 3 0 end processing is conserved across fission yeast species, we identified the telomerase RNA subunits from S. cryophilus and S. octosporus based on synteny in the corresponding regions of the genomes followed by alignment with SpTER1. Pairwise alignments revealed substantial sequence divergence among the RNA subunits, consistent with previous reports that telomerase RNAs evolve rapidly [18][19][20] . It is important to note that the genus Schizosaccharomyces encompasses species that are as divergent as humans and amphioxus 21 . Even the two most closely related species S. cryophilus and S. octosporus are as distantly related as humans and dogs. The template region, which is generally more highly conserved among telomerase RNAs, differs between S. pombe and the other fission yeast species, with S. cryophilus and S. octosporus sequences containing a three-nucleotide insertion relative to S. pombe ( Supplementary Fig. 1). When this sequence was inserted into S. pombe TER1 and telomeres were cloned and sequenced, near-perfect GGGTTACTT repeats were added to chromosome ends rather than the heterogeneous G 2-6 TTACA 0-1 C 0-1 repeat generated by the endogenous template ( Supplementary Fig. 2). Examination of whole-genome sequencing data for S. cryophilus and S. octosporus by the tandem repeat finder identified the same perfect repeat as a probable candidate telomere sequence for both species 21 .
Probes for the presumed TER1 transcripts identified a prominent band migrating at B1,350 nucleotides in total RNA samples from S. cryophilus (Fig. 1a). Due to the size heterogeneity resulting from multiple polyadenylation sites and variable poly(A) tail length, it is not unusual for the telomerase RNA precursor to run as a smear on polyacrylamide gels. Two bands were observed for S. octosporus, reminiscent of the precursor and mature form described for S. pombe 12 . Mapping the 5 0 and 3 0 ends of these RNAs by circular reverse-transcription (RT)-PCR revealed further similarities as well as differences from S. pombe. The major forms of S. cryophilus and S. octosporus TER1 are 1,291 and 1,270 nucleotides in length (Supplementary Fig. 1) compared with 1,213 nucleotides for S. pombe 22 . Longer polyadenylated forms were also detected, as well as putative introns located downstream of the 3 0 end of the major forms. RT-PCR from RNA samples of the respective species confirmed the presence of unspliced and spliced forms, indicative of a conserved role for the spliceosome in telomerase RNA maturation (Fig. 1b).
Unlike S. pombe where the Sm-protein-binding site and a 5 0 splice site overlap by one nucleotide, 5 0 splice sites were found 14 (S. cryophilus) and 27 (S. octosporus) nucleotides downstream of the 3 0 ends of the mature forms, which fall within putative Sm-/LSm-binding sites (Fig. 1c). Detection of several clones that terminate at sites between the 5 0 SS and the Sm site suggest that spliceosomal cleavage is followed by exonucleolytic degradation until binding by the LSm2-8 complex stabilizes the 3 0 end, as shown for S. pombe 8 .
In light of the critical importance of a strong BS and long BP-3 0 SS distance for spliceosomal cleavage in S. pombe, we were surprised to find that the BS sequences did not correspond to the perfect complement of the binding sequence in U2 snRNA nor was the distance between BP and 3 0 SS unusually long (Fig. 1c). We therefore wondered whether 3 0 end formation in S. cryophilus and S. octosporus even involved the release of the 5 0 exon after the first cleavage reaction, or whether complete splicing is followed by exonucleolytic degradation.
Spliceosomal cleavage is conserved among fission yeasts. Replacement of the S. pombe TER1 intron with introns from protein-encoding genes, which lack the combination of features ARTICLE NATURE COMMUNICATIONS | DOI: 10.1038/ncomms7104 that promote release after the first step, results in complete splicing of TER1 (ref. 7). In contrast, when the introns of S. cryophilus and S. octosporus TER1 were introduced into SpTER1, the mature form of TER1 was readily produced (Fig. 1d). Similar to wild-type SpTER1, only a small amount of the spliced form was observed by RT-PCR (Fig. 2c). In summary, despite the absence of all features previously defined as critical for spliceosomal 3 0 end processing, the introns of S. cryophilus and S. octosporus TER1 predominantly undergo the first step of splicing followed by release of the 5 0 exon.
To understand why splicing fails to go to completion for these introns, we searched for features that distinguish the Sc and SoTER1 sequences from introns in protein-encoding genes.
Intron length, BP to 3 0 SS distance, BS and 3 0 SS sequence were all similar to the consensus derived from annotated introns in the respective species (data not shown). However, the 5 0 SS sequence in Sc and SoTER1 was found to be highly unusual (Fig. 1c). Only 3 of 5,702 introns in S. cryophilus and 1 out of 5,350 annotated introns in S. octosporus contain the 5 0 SS sequence GUCAGU indicative of negative selection ( Supplementary Fig. 3). This is likely significant in the context of TER1 introns undergoing the first, but not the second step of splicing. In budding yeast, studies of the dynamic interactions that occur between snRNAs and intronic sequences during splicing have shown that splicing efficiency is the outcome of kinetic competition between different conformations 23,24 . Among the mutations found to inhibit the second step of splicing were those that hyperstabilize the 5 0 SS:U6 snRNA interaction 25 .
To test whether the C at the þ 3 position was responsible for uncoupling the two steps of splicing in the context of TER1 (c) Schematic of the introns located downstream of the 3 0 ends of the mature forms in each of the three species. Distances between the Sm-binding site, 5 0 splice site, branch site and 3 0 splice site are given in nucleotides (nts). Designation of the Sm-/LSm-binding sites in S. cryophilus and S. octosporus is based on sequence similarity, not on experimental verification of Sm/LSm binding. To indicate this fact, the sites are labelled with a question mark. (d) RNaseH cleavage followed by northern blot visualizes the relative abundance of precursor, spliced and cleaved forms of SpTER1 containing the respective introns (colours as in c). An oligonucleotide probe against the snoRNA sn101 was used as a loading control. ARTICLE processing, we introduced a C3A mutation, thereby destabilizing the interaction with U6 snRNA (Fig. 2a). Following this singlenucleotide change, the 5 0 SS now matches the most common 5 0 splice site sequence in S. cryophilus and S. octosporus ( Supplementary Fig. 3). This single-nucleotide change had a profound effect on TER1 processing, with the spliced form now accumulating at the expense of the cleaved form (Fig. 2b,c). These observations indicate that the spliceosome has a conserved function in generating the 3 0 ends of telomerase RNA in diverse fission yeasts. Perhaps more surprisingly, the element that uncouples the two steps of splicing is entirely different from those in S. pombe TER1.
3 0 end processing of telomerase RNA in filamentous fungi.
To determine whether telomerase 3 0 end processing by the spliceosome is more widely conserved, we turned our attention to the recently identified TER1 sequences from Aspergilli where the presence of a non-canonical, putative 5 0 SS had been noted 26 .
The sequence AUA(A/C)GU, as well as putative branch and 3 0 splice sites are found downstream of a pyrimidine-rich motif that resembles an Sm-binding site (Fig. 3a). Mutation of the highly conserved first nucleotide in an intron from G to A blocks the second step of splicing for introns from budding yeast and humans [27][28][29] . However, no example of an AU intron functioning in 3 0 end processing had been reported thus far.
To assess whether the Aspergillus sequences can function in telomerase 3 0 end processing, we replaced the TER1 intron and second exon with the corresponding sequence from A. niger (Fig. 3b). The mature, non-polyadenylated form of TER1 was readily detected by 3 0 end cloning from cells harbouring the A. niger sequence (Fig. 3c). Sequencing revealed that these RNAs terminate within the Sm-binding site (Fig. 3d). Reminiscent of wild-type S. pombe TER1, the product of the first step of splicing is by far the most abundant form detected by northern analysis (Fig. 4a). Consistent with the adenine at the first position of the intron blocking the second step, the spliced form is not observed by northern blot and is only barely detected by RT-PCR across the intron (Fig. 4b).
We next mutated the A at the first position of the intron to G to ascertain whether this non-canonical 5 0 SS was indeed responsible for blocking the second step of splicing. This singlenucleotide change resulted in a 410-fold reduction in the cleaved form (Fig. 4a). Spliced product was now the dominant form of the RNA detected by RT-PCR (Fig. 4b). When telomerase activity was examined in extracts prepared from cells harbouring TER1 with the A. niger intron or an A1G mutant   version, the former showed near-wild-type levels of activity (Fig. 4c, compares lanes 1 and 3). In contrast, telomerase activity was barely detectable for the A1G mutant (lane 4). Similarly, a 5 0 SS mutant in the S. pombe TER1 intron, which blocks the first step, reduced telomerase activity to below the threshold of detection (lane 2). Consistent with the results from in vitro telomerase assays, the wild-type A. niger sequence fully rescued the maintenance of wild-type telomere length, whereas cells harbouring the A-to-G point mutant had shorter telomeres (Fig. 4d). These results demonstrate that the non-canonical 5 0 SS found near the 3 0 end of telomerase RNA from Aspergillus species is necessary to trigger spliceosomal cleavage and can thus function in 3 0 end processing of telomerase RNA. The result further demonstrates that 3 0 end cleavage, but not the underlying mechanism, is critical for generating functional telomerase.
Context-dependent effects of variant 5 0 splice sites. Considering that adenosine in intron position 1 or cytosine in position 3 are exceedingly rare among annotated introns, we asked whether these deviations from a consensus 5 0 SS are by themselves sufficient to trigger spliceosomal cleavage. We had previously introduced the G1A mutation into the S. pombe TER1 intron and found that it blocked the first step of splicing 7 . We now introduced this mutation in the context of an intron from the protein-encoding tif212 gene, which also has the 5 0 SS sequence GUAUGU. As cleavage products of the mRNAs may be unstable, we performed these experiments in the context of TER1 exonic sequences, where we had previously demonstrated that LSm2-8 binding stabilizes the cleaved product 8 . The þ 1A mutation inhibited splicing (Fig. 5a) and polyadenylation was observed to occur at a variety of sites within the intron ( Supplementary  Fig. 4A), consistent with U1 binding inhibiting polyadenylation nearby 30 . Splicing was also inhibited by the þ 3C mutation, but most clones terminated downstream of the intron (Supplementary Fig. 4B). A few clones ending at the Smbinding site were observed, but based on the scattered distribution of 3 0 ends these are more likely to be the result of exonucleolytic degradation of the precursor than spliceosomal cleavage. We thus conclude that in the context of the tif212 intron, the A3C mutation inhibits spliceosome assembly, or the first step.
We next introduced three naturally occurring S. pombe introns with GUCAGU and one with a GUCUGU 5 0 SS sequence into TER1 to ascertain whether these introns are subject to spliceosomal cleavage. Similar to the endogenous TER1 intron, only small amounts of the spliced product were observed with each of these introns ( Supplementary Fig. 5A). Although the polyadenylated precursor was the dominant form based on northern blots, cleaved forms were indeed observed and for intron 5 from the alp41 gene amounted to 40% of the cleaved RNA seen with the native TER1 intron (Fig. 5b). Cloning and sequencing of 3 0 ends confirmed the presence of cleaved products terminating at the Sm-binding site for each intron (Supplementary Fig. 5B). In summary, both þ 3C and þ 1A promote spliceosomal cleavage in a context-dependent manner, but are by themselves insufficient to promote spliceosomal cleavage to the extent observed in telomerase RNA introns.

Discussion
The observation that the spliceosome has a role in 3 0 end processing of telomerase RNA in S. pombe came as a surprise at first. Subsequent mechanistic studies revealed that release of cleaved RNA after the first transesterification reaction requires a strong BS and a long distance between the BP and 3 0 SS 13 . Here we demonstrate that despite these features being absent from telomerase RNAs in other fission yeasts and filamentous fungi, the role of the spliceosome in telomerase RNA 3 0 end formation is nevertheless conserved. The evolutionary implication of these findings is that the ancestral telomerase RNA subunit must have contained an intron that was spliced during biogenesis. At least three independent mechanisms have emerged to convert splicing into 3 0 end processing by aborting the process after the first cleavage reaction.
It has long been appreciated that the spliceosome is a highly dynamic machine, which undergoes several compositional and conformational changes during the course of a splicing reaction. Comprehensive genetic and biochemical analyses of mutations in intron sequences and splicing factors, and in particular examination of compensatory effects, have resulted in the 'two-state model', which postulates that conformations supporting the two catalytic reactions are in kinetic competition with each other 23,31,32 . Numerous mutations in the RNA substrate, as well as in the protein and RNA components of the spliceosome, have been found to stabilize one conformation over the other thereby affecting the outcome of the reaction. These studies, combined with examination of telomerase RNA processing in different species, now suggest a common basis for how the spliceosome functions in 3 0 end processing: L a n e ARTICLE In S. pombe, the transition from the first to the pre-second-step conformation is slow due to the long distance between BP and 3 0 SS, and the perfect sequence complementarity between the BS and the BS-binding sequence in U2 snRNA (Fig. 5c). Consequently, the splicing intermediates are discarded via a kinetic proofreading mechanism involving the DExD/H-box helicases Prp22 and Prp43 (refs 13,15). Several other mutations have been described in the splicing literature that trigger discard by attenuating the transition to the pre-second-step conformation or by blocking second-step catalysis, including changes of conserved nucleotides at the 5 0 SS, BS and 3 0 SS. We now show that at least two of these sequence variants occur naturally in RNA substrates that undergo spliceosomal cleavage as part of telomerase RNA biogenesis.
In the fission yeasts S. cryophilus and S. octosporus, the cytosine at the third position of the highly unusual 5 0 SS sequence GUCAGU provides the opportunity for an additional G:C base pair with U6 snRNA, thereby hyperstabilizing the 5 0 SS:U6 interaction (Fig. 5c). Extensive mutational analysis in the context of the budding yeast actin intron demonstrated that a hyperstabilized 5 0 SS:U6 duplex becomes rate limiting for the transition between the first and second step 25 . In this case, GUCUGU inhibited the second step, whereas GUCAGU, which lacks the ability to base pair at position 4, did not. It appears that the fission yeast spliceosome is more sensitive with regards to 5 0 SS:U6 stability at least in the context of TER1 introns, since the GUCAGU sequence inhibits the second step for several introns examined here. We have previously shown that GUCUGU and GUCAGU both inhibit the first step when introduced into the S. pombe TER1 intron 13 , further illustrating that the effects of splice site sequence variations are highly context dependent.
In the case of Aspergillus sp., a non-canonical 5 0 splice site AU is critical for limiting the spliceosome to carrying out the first reaction followed by the release of the cleaved products. On the basis of covariation of the first and last nucleotide of introns and the reciprocal suppression of mutations at these positions, it has been proposed that a non-Watson-Crick interaction between the first and last guanosine is pivotal to position the 3 0 splice site for the second cleavage reaction 33 . The adenine at position 1 of the TER1 intron in Aspergillus blocks this interaction and promotes release of the products after the first cleavage reaction (Fig. 5c).
The realization that several different sequence variations can attenuate or block the transition to the second-step conformation has hampered computational attempts to find other RNAs that undergo spliceosomal cleavage as a mechanism of 3 0 end maturation. It appears that processing of telomerase RNA by spliceosomal cleavage is under positive selection in diverse fungi, but the underlying mechanism is not. With evidence for three distinct means of attenuating the transition to the second step being critical for telomerase RNA 3 0 end processing in different species, it is tempting to speculate that additional mechanisms underlying 3 0 end processing by the spliceosome will be identified in due course. For example, a guanosine at the BP or an AG to AC mutation at a 3 0 SS have both been shown to block the second step of splicing 15,34 . It remains to be seen whether either of these sequence variations occurs naturally in the context of an RNA that undergoes spliceosomal cleavage as part of its normal biogenesis. The observation that introns are present beyond the 3 0 ends of telomerase RNAs in other fungi as well 26,35 suggests a more widely conserved function for the spliceosome in telomerase RNA processing at least in fungi. It will also be important to determine whether telomerase processing is highly unusual or the first example of a group of RNAs for which the spliceosome generates functional products via a single cleavage event. The observation that an intron from the protein-encoding alp41 gene is efficiently cleaved further supports that spliceosomal cleavage may function as a regulatory mechanism by competing with the completion of splicing, thereby reducing the amount of mature mRNA available for translation.

Methods
Yeast strains and constructs. Genotypes of the strains used in this study are listed in Supplementary Table 1. Constructs containing S. cryophilus, S. octosporus and A. niger sequences were generated by replacing the corresponding regions of SpTER1 in pJW10 (ref. 22) with synthetic DNA fragments. Derivatives of these constructs were generated by site-directed mutagenesis and subcloning. Plasmids were introduced into S. pombe strains PP138, PP399 and PP407 by electroporation, and transformants were selected on Edinburgh Minimal Media lacking uracil.
RNA analysis. RNA was isolated from cells grown to a density of 5 Â 10 6 cells ml -1 at 32°C. Cells were collected by centrifugation, washed twice in ddH 2 O, resuspended in ddH2O and quick-frozen by dripping the cell suspension into liquid nitrogen. Cells were lysed in a 6,850 Freezer mill (SPEX SamplePrep) using 8 cycles (2 min) at a rate of 10 per second with 2 min cooling time between cycles. The lysed cell powder was transferred directly into tubes containing 10 ml phenol/chloroform/isoamyl alcohol (25:24:1) and 10 ml sodium acetate (50 mM), 1% (w/v) sodium dodecyl sulphate preheated to 65°C. RNA was extracted five times with phenol/chloroform/isoamyl alcohol and once with chloroform/isoamyl alcohol. Total RNA was ethanol precipitated and resuspended in 50 mM sodium acetate (pH 5.2). For use in RT-PCR, the RNA was further purified using the RNeasy mini kit (Qiagen) following the manufacturer's instructions for DNase digestion and RNA cleanup. With the exception of Figs 1a and 5b, all samples used for northern blotting were subjected to RNaseH cleavage prior to gel electrophoresis to enhance resolution of precursor, spliced and cleaved forms. Briefly, 15 mg of DNase-treated RNA were combined with 600 pmol of each of the two DNA oligonucleotides (BLoli1043; 5 0 -AGGCAGAAGACTCACGTACACTGAC-3 0 and BLoli1275; 5 0 -CGGAAACGGAATTCAGCATGT-3 0 ) complementary to sites in the first and second exon, respectively. The mix was heated to 65°C in a heat block, and a second heat block at 75°C was placed on top of the tubes to reduce condensation. After 5 min, the heat-block sandwich was transferred onto a Styrofoam box to allow slow cooling to room temperature. After 45 min, RNaseH buffer New England Biolabs (NEB, final concentration 1 Â ) and RNaseH enzyme (5 units) were added and samples were incubated at 37°C for 30 min. RNaseH-treated samples were ethanol precipitated for at least 1 h at À 20°C. RNA pellets were recovered by centrifugation, dissolved in 1 Â formamide loading buffer and run on a polyacrylamide gel for northern analysis as described below.