Introduction
In budding yeast, most cytoplasmic mRNA turnover is initiated by deadenylation1. These messages are then either decapped and subject to 5' 
3' decay2, 3 or degraded by the cytoplasmic exosome4. Although these decay pathways are conserved5, 6, metazoans and fission yeast contain additional cytoplasmic RNA-processing enzymes; the roles of many of these have not yet been determined. Schizosaccharomyces pombe Cid1 is one such enzyme, a cytoplasmic member of a family of RNA nucleotidyl transferases7, 8.
Cid1, identified through its involvement in the S-M checkpoint7, is now known to be one of a subgroup of this family possessing either poly(U) polymerase (PUP) and/or terminal uridyl transferase (TUTase) activity9. This subgroup also includes the human enzymes U6 TUTase, Hs2 and Hs3 (refs. 9,10,11), although no member of this subgroup is present in budding yeast12.
Uridylation of mRNAs and noncoding RNAs has been described in fission yeast and metazoans. Known substrates include miRNA-directed cleavage products13 and replication-dependent histone mRNAs, which in metazoans contain a 3' stem-loop structure rather than a poly(A) tail, and decay of which is stimulated by oligouridylation14, 15, 16. In addition, we previously observed terminal uridyl residues on polyadenylated S. pombe act1 mRNA during S-phase arrest9. Although it seemed likely that Cid1-mediated uridylation was generally involved in mRNA metabolism, the effect of such a modification on polyadenylated messages was unclear.
Here we have used a circularized rapid amplification of cDNA ends (cRACE) technique to capture mRNA decay intermediates in S. pombe. Unexpectedly, in contrast to the situation in Saccharomyces cerevisiae, we find that decapped intermediates often contain substantial poly(A) tails, indicative of a novel deadenylation-independent decapping pathway for bulk mRNA in fission yeast. We also show that uridylation of polyadenylated mRNAs forms the basis for an additional 5' 3' decay pathway, probably conserved in higher eukaryotes, that elicits decapping and seems to be mediated by the Lsm1–7 complex. Uridylation and deadenylation have overlapping and distinct stimulatory effects on decapping of polyadenylated mRNAs.
Results
cRACE captures act1 mRNA degradation intermediates
To dissect RNA decay pathways in fission yeast, we used the cRACE technique6. A tail-independent method of capturing 3' and 5' ends, this procedure allows distinction between decapped and capped mRNAs (Fig. 1a). We first examined decapped and capped act1 transcripts from exponentially growing wild-type cells.
Figure 1: Decapping of mRNA can be independent of deadenylation.
(a) Overview of the cRACE procedures used to capture ends of decapped transcripts (gray) and mature transcripts (white) (see Online Methods). (b,c) The 5' (b) and 3' (c) ends of various types of act1 cRACE sequences are plotted as the distance (nucleotide position indicated on the horizontal axis) from the start codon. The open reading frame (ORF) is marked with a line. (d) Poly(A) tail lengths of decapped (black; 40 sequences) and capped (white; 20 sequences) act1 cRACE sequences were binned into groups of 10 nt. Tail lengths were then plotted as the percentage of adenylated species. (e) Box-and-whisker plots of poly(A) tail lengths found on six different decapped transcripts, act1, adh1, gar2, hcn1, pof9 and urg1 (n = 40, 16, 11, 12, 8 and 39, respectively). This plot depicts the quartiles of poly(A) tail length, with the whiskers representing the range of each data set and the boxplot demarcating the second and third quartiles, which are separated by the median.
Full size image (88 KB)We initially wished to determine whether those products isolated from decapped cRACE analysis were derived from decay intermediates. To do this, we compared the 5' ends isolated from capped and decapped mRNAs (Fig. 1b). The 5' ends of products from capped transcripts generally lay 57 nucleotides (nt) or 58 nt upstream from the start codon (Fig. 1b and Supplementary Fig. 1 online). These nucleotides presumably represent the major transcriptional start site for act1. In contrast, the 5' ends of decapped products were heterogeneous, always downstream from the major transcriptional start site and distributed significantly differently from the capped species (P < 0.0001, two-tailed Mann-Whitney test). Thus, we conclude that these products represent mRNAs that have been subject to decapping and subsequent partial 5' 3' decay in vivo.
We next compared the 3' ends of adenylated and non-adenylated transcripts (Fig. 1c). Similarly to previous observations17, we detected three cleavage and polyadenylation sites in the act1 gene: the first is approximately 1,190 nt downstream from the start codon (and 60 nt downstream from the stop codon); the second is 1,550 nt downstream from the start codon; and the third is almost 1,800 nucleotides downstream from the start codon. Whereas for the proximal and distal sites the exact site of cleavage and polyadenylation was heterogeneous, at the medial polyadenylation site cleavage occurred precisely at a single position.
Of the 54 sequences obtained for decapped transcripts, 14 were not adenylated; similarly, 7 of 27 cRACE products obtained for capped transcripts did not contain untemplated adenyl residues. The 3' ends of these species were heterogeneous and did not coincide with those of polyadenylated species. Notably, all but two of the 3' ends of non-adenylated products mapped upstream from the most distal polyadenylation site. Thus, in the case of decapped transcripts, these data are consistent with concerted 5' 3' and 3' 5' decay. In the case of capped transcripts, these non-adenylated transcripts presumably arise either from ongoing transcription or from 3' 5' decay after deadenylation.
Deadenylation-independent decapping in S. pombe
Numerous decapped transcripts contained long poly(A) tails, which is unexpected given that deadenylation to oligo(A) tails is thought normally to precede decapping in budding yeast18. Notably, when we compared the poly(A) tail lengths of decapped and capped act1 RNAs, we found no significant difference between the two (P = 0.65; Fig. 1d). The average poly(A) tail length on decapped messages was 25 nt, and on capped messages it was 23 nt. This result is consistent with a previous report19 that found no correlation between poly(A) tail length and mRNA stability in fission yeast. In contrast, when we analyzed decapped S. cerevisiae act1 transcripts the average poly(A) tail length was 11 nt (Supplementary Fig. 2 online), consistent with previous analysis1. These tails were also significantly shorter than those observed on decapped S. pombe act1 mRNAs (P = 9 
10-6). Together, these data suggest that, in contrast to the situation in S. cerevisiae, where decapping of most messages requires deadenylation1, 18, in S. pombe there is a deadenylation-independent decay pathway for act1 mRNA.
We also carried out decapped cRACE analysis of five other transcripts, adh1, gar2, hcn1, pof9 and urg1, (Supplementary Fig. 3 online). These messages allowed us to test the generality of deadenylation-independent decapping. The distribution of poly(A) tail lengths differed between these six genes (Fig. 1e). The average poly(A) tail length on adh1 cRACE products was 16 nt, and on hcn1 products it was 12 nt. Comparison of poly(A) tail lengths of capped and decapped cRACE products suggests that hcn1 mRNA is mainly degraded by deadenylation-dependent pathways (Supplementary Fig. 4 online). On the other hand, the average length of poly(A) tails on decapped urg1 products was 23 nt, and more than one-third of polyadenylated, decapped urg1 products contained poly(A) tails longer than 30 nt. Similarly, we captured decapped pof9 products with tails as long as 54 nt. Given that poly(A) tails on average are 40 nt in S. pombe19, these data indicate that some mRNAs, such as act1, pof9 and urg1, are subject to deadenylation-independent decapping.
Decapped transcripts are often uridylated
We found one to two nontemplated 3' terminal uridyl residues on 25% of the polyadenylated, decapped act1 transcripts (Fig. 2a), suggesting a role for uridylation in a novel mRNA decay pathway. No terminal uridyl residues were observed on non-adenylated products. Similarly, we found terminal uridyl residues on 18% of decapped, adenylated urg1 messages, 25% of decapped, adenylated hcn1, adh1 and pof9 messages and 45% of decapped, adenylated gar2 messages (Fig. 2a). Uridylation thus seems to be a widespread mRNA modification in S. pombe. In contrast, we observed no terminal uridyl residues on 22 S. cerevisiae act1 cRACE products analyzed (P = 0.005; Supplementary Fig. 2b). This finding is consistent with the observation that budding yeast does not contain a Cid1 ortholog12, 20.
Figure 2: Decapped mRNAs are often uridylated.
(a) The percentage of decapped, adenylated cRACE products that contain terminal uridyl residues is shown for act1, adh1, gar2, hcn1, pof9 and urg1 (n = 10/40, 4/16, 5/11, 3/12, 2/8 and 7/39, respectively). (b) The poly(A) tail lengths of non-uridylated (black) and uridylated (white) decapped urg1 RNAs were binned into groups of 10 nt. Tail lengths were then plotted as the percentage of adenylated species. (c) The poly(A) tail lengths of all non-uridylated (black; 31 sequences) and uridylated (white; 95 sequences) decapped transcripts are compared. For each transcript, each tail length was normalized to the median of non-uridylated tail length to correct for intertranscript poly(A) tail length variability. These normalized lengths were then binned into groups and plotted as the percentage of adenylated species. (d) As in b, the poly(A) tail lengths of non-uridylated (black) and uridylated (white) decapped act1 RNAs were binned into groups of 10 nt. Tail lengths were then plotted as the percentage of adenylated species.
Full size image (35 KB)To delineate this uridylation-dependent decay pathway, we first determined whether uridylation necessarily followed deadenylation by comparing the poly(A) tail lengths of uridylated and non-uridylated messages (Fig. 2b–d). Some long poly(A) tails did not have terminal uridyl residues, suggesting that the deadenylation-independent pathway described above is also uridylation-independent. Nevertheless, in several cases we found terminal uridyl residues on poly(A) tails longer than 30 nt, suggesting that uridylation does not necessarily follow deadenylation. Indeed, for urg1 products (Fig. 2b) there was no difference between the lengths of poly(A) tails on uridylated and non-uridylated products (P = 0.39, two-tailed Mann-Whitney test).
To address this question more thoroughly, we compared the poly(A) tail lengths on uridylated and non-uridylated products for all six genes analyzed here (Fig. 2c). To correct for the difference in poly(A) tail length distribution between these transcripts, we first normalized within each gene to the median length of non-uridylated products. When we compared these normalized lengths of non-uridylated and uridylated transcripts of these six genes, we found no significant difference (P = 0.21, two-tailed Mann-Whitney test). Nevertheless, it may be that individual transcripts, such as act1, are often degraded by the concerted action of uridylation and deadenylation (Fig. 2d). Taken together, these data suggest that uridylation and deadenylation can proceed both sequentially and in parallel.
Uridylation is Cid1 dependent
To investigate whether uridylation is Cid1 dependent, we performed act1 cRACE analysis on RNA from cid1
cells. Only 1 of 29 polyadenylated, decapped products (3.4%) contained a terminal uridyl residue (Fig. 3a), a significantly lower frequency than observed in wild-type cells (P = 0.007). We observed a tail of UUU on one non-adenylated, decapped transcript, out of a total of 45 products. We observed no terminal uridyl residues on capped messages. These data indicate that most mRNA uridylation is dependent on Cid1 and suggest the presence of a second, minor uridylating enzyme in S. pombe. It is interesting to note that we did not observe this low level of uridylation on budding yeast act1 products (Supplementary Fig. 2b), perhaps indicating that this second enzyme is another Cid1 family member not found in S. cerevisiae.
Figure 3: mRNA uridylation is Cid1-dependent, and impairment of decapping increases uridylation on decapped messages.
(a) The percentage of adenylated, decapped act1 sequences that contain (black) or lack (white) terminal uridyl residues is plotted for sequences isolated from wild-type (WT), cid1 and dcp1-ts cells (n = 40, 29 and 20, respectively). (b,c) Poly(A) tail lengths, binned into groups of 10 nt, of decapped (b) and capped (c) act1 sequences isolated from wild-type (black; n = 40 and 20, respectively) and cid1 (white; n = 29 and 36, respectively) cells are compared. (d,e) Poly(A) tail lengths, binned into groups of 10 nt, of decapped (d) and capped (e) act1 sequences isolated from wild-type (black; n = 40 and 20, respectively) and dcp1-ts (white; n = 29 and 18, respectively) cells are compared.
The poly(A) tails of decapped act1 mRNAs isolated from cid1 cells were significantly shorter than those from wild-type cells (P = 0.003; Fig. 3b). As there was no significant difference in the poly(A) tail length on capped transcripts (P = 0.59; Fig. 3c), the shorter tails observed on decapped act1 transcripts in cid1 cells are most likely due to increased deadenylation (see below), consistent with the notion that uridylation and deadenylation have redundant roles in stimulating mRNA decay in fission yeast.
Uridylation precedes decapping
The above analysis did not allow us to determine whether uridylation precedes or follows decapping. To test this, we performed two experiments. First, we used a primer complementary to a region 90 nt downstream from the start codon during the PCR step, thus amplifying transcripts that had not been subject to substantial 5' 3' decay (Supplementary Fig. 5 online). We found that there was no significant difference in uridylation between those products obtained with the upstream primer and those obtained with the downstream primer used previously, which is complementary to a region 650 nt beyond the start codon (P = 0.44). This result is consistent with uridylation preceding decapping.
In addition, we carried out act1 cRACE on RNA from dcp1-ts cells (Supplementary Fig. 6 online), which contain a temperature-sensitive activator of decapping21. Even at the permissive temperature, decapping was impaired in these cells, as decapped and capped act1 cRACE products contained significantly shorter poly(A) tails than those in wild-type cells (P = 9E-05 and P = 0.011, respectively; Fig. 3d,e), consistent with previous reports that, in cells lacking the decapping enzyme, deadenylated, capped mRNA accumulates2. Notably, 55% of the decapped act1 cRACE products from dcp1-ts cells contained terminal uridyl residues (Fig. 3a). This represents a significant increase in comparison to wild-type cells (P = 0.005). Moreover, this effect is specific to the decapping defect, as cells lacking Ski2, a cytoplasmic exosome component, did not show such an increase (Supplementary Fig. 5b).
Owing to the position of the most abundant transcriptional start site of act1, it was not possible to determine definitively by cRACE analysis whether capped mRNA was uridylated (Supplementary Fig. 7 online). To address whether 3' uridylated, capped species accumulated in dcp1-ts cells, we used a 3' cRACE technique called hybrid-selection cRACE (HSC-RACE)22, which has been used previously9 to demonstrate uridylation of act1 messages (Fig. 4a,b and Supplementary Tables 1 and 2 online). Here transcripts are specifically selected with a biotinylated RNA probe on magnetic streptavidin beads. The 3' ends of the transcripts are liberated by oligonucleotide-directed RNase H cleavage and then circularized, amplified and sequenced as in cRACE. As RNase H cleavage precisely defines the 5' ends of the captured RNAs, this procedure allows definitive sequencing of 3' ends. Most of these products are derived from the abundant, capped mRNAs. Consistent with the above cRACE analysis, the poly(A) tails on act1 mRNAs were significantly shorter in dcp1-ts cells than in wild-type cells (P = 0.0002; Fig. 4c). Whereas uridylation was observed on 17% of act1 transcripts in wild-type cells, this increased to 78% in dcp1-ts cells (P = 0.0002; Fig. 4d). These data suggest that uridylation precedes decapping and, when decapping is impaired, uridylated, capped act1 intermediates accumulate.
Figure 4: Uridylation precedes decapping.
(a,b) HSC-RACE products of total act1 transcripts isolated from wild-type (WT) cells (a) and dcp1-ts cells (b). The final four codons and stop codon (marked by *) are shown, as well as the 3' UTR. White boxes denote sequences with pure poly(A) tails; gray boxes denote poly(A) tails with internal non-adenyl residues; black boxes denote poly(A) tails with terminal uridyl residue(s). (c) Poly(A) tail lengths, binned into groups of 10 nt, of HSC-RACE act1 species from wild-type (white bars) and dcp1-ts (black bars) cells. (d) Percentage of adenylated HSC-RACE act1 products that contain (black) or lack (white) terminal uridyl residues. (e) The percentage of capped hcn1 transcripts that contain (black) or lack (white) terminal uridyl residues is compared for RNA isolated from wild-type and dcp1-ts cells (n = 18 and 14, respectively).
Full size image (56 KB)To support this analysis, we also used cRACE analysis to compare capped hcn1 transcripts isolated from wild-type and dcp1-ts cells. No uridylated, capped hcn1 intermediates were identified in wild-type cells (Fig. 4e). Uridylation-mediated decapping may itself be regulated by additional elements within the 3' UTR, such that decapping of uridylated hcn1 transcripts occurs more quickly than for uridylated act1 mRNAs. Notably, when we examined capped hcn1 transcripts in dcp1-ts cells, 21% were now uridylated (P = 0.02; Fig. 4e). Thus, we conclude that uridylation occurs before decapping as part of a novel mRNA decay pathway.
urg1 transcripts are stabilized in cid1
cells
We next asked whether mRNA stability was affected by the absence of uridylation-dependent decapping. Determination of mRNA half-lives in other experimental systems is often based on the use of general transcriptional inhibitors, but inhibitors of this sort lacking various off-target effects have not yet been described in S. pombe19, 23. As an alternative approach, we therefore examined urg1 mRNA: when uracil is removed from the medium, transcription of urg1 stops, and the transcript undergoes rapid decay24. In addition, because urg1 is uridylated (Fig. 2a) and there are few off-target effects elicited by the removal of uracil24, this system appeared to be ideal to test the effect of uridylation on mRNA decay.
We grew cultures in minimal medium containing uracil to mid-exponential phase, and then we shifted the cells to medium lacking uracil to repress urg1 transcription. Samples were taken 0 min, 10 min, 20 min and 30 min after uracil washout. We quantified the amount of urg1 mRNA remaining by northern blotting, normalized to pik1 mRNA, which is unaffected by the removal of uracil24, and then determined the half-life of urg1 mRNA.
In wild-type cells, urg1 mRNA had a half-life of 9.8 
1.4 min (Fig. 5), consistent with previous microarray analysis24. Notably, in cells lacking Cid1, urg1 mRNA was stabilized, and the half-life of this transcript increased to 17.6 1.0 min, significantly longer than that in wild-type cells (P = 0.0007, Fig. 5b). We conclude that Cid1-mediated uridylation stimulates the decay of urg1 mRNA.
Figure 5: urg1 mRNA is more stable in cells lacking Cid1.
(a–c) Northern blots of urg1 transcript remaining at various time points after uracil washout (above) in wild-type (WT; a), cid1 (b) and ccr4 cells (c). urg1 mRNA levels were normalized to pik1 mRNA (below). The percentage of urg1 remaining at each time point (shown below the lower panel) was calculated by comparison to the normalized amount at 0 min. (d) The percentage of urg1 mRNA remaining after uracil washout is shown for three strains, WT (open circles), ccr4 (black squares) and cid1 (open triangles). At least three independent replicates were performed; error bars represent s.e.m. (e) The half-life of urg1 mRNA in different strains is shown. At least two independent replicates were performed for each strain. *P < 0.01; **P < 0.001. Error bars denote s.d.
The Ccr4–Caf1 deadenylase complex has been described as the major cytoplasmic deadenylase in budding yeast and mammalian cells25, 26, 27. In support of this view, deletion of ccr4 also resulted in an urg1 mRNA half-life longer than that in wild-type cells (21.7 5.6 min, P = 0.009; Fig. 5c,d). Notably, this did not differ significantly from that observed in cid1 cells (P = 0.14). These data suggest that, as in budding yeast26, Ccr4-mediated deadenylation stimulates mRNA decay in fission yeast.
The fission yeast genome encodes two additional presumptive deadenylases, namely the Pan2–Pan3 complex and PARN. The Pan2–Pan3 complex is widely conserved among eukaryotes. In budding yeast, this complex is thought to be involved in nuclear trimming of poly(A) tails and is not a major factor in mRNA decay26, 28, 29. In line with analogous observations in budding yeast26, in pan2 cells urg1 mRNA was not stabilized and had a half-life of 10.2 0.6 min (P = 0.40; Fig. 6). Although PARN has a role in specific mRNA decay pathways30, 31, this deadenylase is notably absent from budding yeast and Drosophila melanogaster32. Schizosaccharomyces pombe PARN does not seem to be a major enzyme in mRNA turnover, as the urg1 half-life in parn cells did not differ significantly from that in wild-type cells (half-life of 11.4 3.0 min, P = 0.24; Fig. 6d). Taken together, these data demonstrate that Cid1 and Ccr4, but neither the Pan2–Pan3 complex nor PARN, are important for mRNA decay, as judged by the increased stability of urg1 transcripts in the corresponding deletion strains. Thus, we suggest that deadenylation, mediated by Ccr4, and uridylation, mediated by Cid1, function in mRNA decay pathways in fission yeast.
Figure 6: Deadenylation and uridylation function as redundant pathways in mRNA decay.
(a) The percentage of decapped, adenylated urg1 sequences that contain (black) or lack (white) terminal uridyl residues is compared for RNA isolated from wild-type (WT), ccr4, cid1 cells, pan2 cells and pan3 cells (n = 39, 19, 19, 27 and 20, respectively). (b–d) Poly(A) tail lengths, binned into groups of 10 nt, of decapped urg1 mRNAs isolated from cid1 cells (b, white), pan cells (c, white) and ccr4 cells (d, white) compared to those products from wild-type cells (black).
In cells lacking both Cid1 and Ccr4, urg1 mRNA was again stabilized (half-life of 19.0 1.6 min, P = 0.0008; Supplementary Fig. 8 online), although this half-life did not differ significantly from that observed in either of the single-deletion strains. It seems likely that uridylation and deadenylation have distinct, though overlapping, influences on mRNA stability. Accordingly, other deadenylases, such as the Pan2–Pan3 complex or PARN, may be able to act redundantly with Cid1 and the Ccr4–Caf1 complex, as has been suggested in both budding yeast and human cells26, 27. Nevertheless, as urg1 mRNA is stabilized in cid1 cells, we conclude that uridylation forms the basis of a novel decay pathway for polyadenylated mRNA.
Uridylation and deadenylation are redundant pathways
We next hypothesized that, if uridylation and deadenylation act redundantly to lead to decapping and mRNA decay, then cells lacking uridylation might show increased deadenylation, and, conversely, cells lacking deadenylases might show increased uridylation. To address this, we performed cRACE analysis on decapped urg1 transcripts isolated from cells lacking Cid1, Ccr4, Pan2 or Pan3 (Fig. 6 and Supplementary Fig. 9 online). We analyzed cells grown in rich medium, which contains uracil, to ensure that sufficient urg1 mRNA was present to enable analysis.
We investigated the uridylation of decapped urg1 transcripts in cid1 cells. Whereas in wild-type cells, as described above, 18% of urg1 cRACE products contained a terminal uridyl residue, none of the 19 decapped, adenylated products analyzed in cid1 cells contained a terminal uridyl residue (P = 0.03; Fig. 6a). In addition, as with decapped act1 products (Fig. 2), decapped urg1 messages from cid1 cells contained significantly shorter poly(A) tails than those found in wild-type cells (P = 0.001, one-tailed Mann-Whitney test; Fig. 6b). Thus, it seems likely that deadenylation can compensate when uridylation-mediated decapping is lacking.
We next characterized poly(A) tails on decapped transcripts from the deadenylase deletion strains. Although global poly(A) tail length increases in pan2 cells33, there was no significant difference in the length of the poly(A) tails observed on decapped urg1 cRACE products in pan2 cells compared with those from wild-type cells (P = 0.24, two-tailed Mann-Whitney test; Fig. 6b). These data suggest that an additional deadenylase has an overlapping function with Pan2; this deadenylase is likely to be the Ccr4–Caf1 complex, as has been described in budding yeast26.
In contrast, poly(A) tails on decapped urg1 cRACE products from ccr4 cells were significantly longer than those from wild-type cells (P = 0.002, one-tailed Mann-Whitney test; Fig. 6d). Some non-uridylated urg1 species contained poly(A) tails longer than 50 nt and indicate the presence of the deadenylation-independent decapping pathway described above (Fig. 1). Whereas in wild-type cells we observed a bimodal distribution for poly(A) tail length, the peak of shorter tails was lost in ccr4 cells (Fig. 6d), consistent with impaired deadenylation.
We analyzed uridylation in cells lacking components of deadenylase complexes (Fig. 6a). In contrast to wild-type cells, where 18% of decapped, adenylated urg1 products were uridylated, in ccr4 cells this figure rose to 53% (P = 0.003). Notably, uridylation also increased in pan2 and pan3 cells. We observed terminal uridyl residues on 44% of decapped, adenylated products from pan2 cells and on 50% from pan3 cells, a significant increase in comparison to wild-type cells (P = 0.007 and P = 0.005, respectively). We thus propose that uridylation and deadenylation function redundantly to stimulate mRNA decay.
Lsm1–7 mediates uridylation-dependent decapping
Previous reports have indicated that the Lsm1–7 complex not only enhances decapping in vivo but also is able to bind oligo(U) tracts in vitro to stimulate decapping34, 35. Indeed, previous analysis has shown that the presence of a single terminal uridyl residue was sufficient to stimulate decapping significantly35. We therefore wondered whether the Lsm1–7 complex might mediate uridylation-dependent decapping. We carried out act1 cRACE analysis on RNA isolated from lsm1 cells. As with dcp1-ts cells, poly(A) tails of capped and decapped act1 mRNAs from lsm1 cells were significantly shorter than those from wild-type cells (P = 0.0002 and P = 0.0004, respectively; Fig. 7a,b). This is consistent with previous observations in budding yeast that cells lacking Lsm1, similar to those lacking Dcp1, accumulate deadenylated, capped intermediates34.
Figure 7: Uridylation-mediated decapping requires Lsm1.
(a,b) Poly(A) tail lengths, binned into groups of 10 nt, of capped (a) and decapped (b) act1 sequences isolated from wild-type (WT, black; n = 20 and 40, respectively) and lsm1 (white; n = 19 and 22, respectively) cells are compared. (c) The percentage of decapped, adenylated act1 sequences that contain (black) or lack (white) terminal uridyl residues is compared for RNA isolated from WT and lsm1 cells.
Notably, in lsm1 cells, as with dcp1-ts cells, there was a significant accumulation of decapped, adenylated act1 messages that had been uridylated. Whereas in wild-type cells 25% were uridylated, in lsm1 cells 55% were uridylated (P = 0.01; Fig. 7c). There was no difference between uridylation of decapped act1 transcripts in lsm1 and dcp1-ts cells, however (P = 0.33). Analysis of capped transcripts similarly suggested that uridylated, capped intermediates accumulate in lsm1 cells as they do in dcp1-ts cells (Supplementary Fig. 7). Consistent with previous analysis, urg1 transcripts were stabilized in lsm1 cells, in which the half-life was 41.3 3.3 min (P = 0.0003; Supplementary Fig. 8). This half-life was significantly longer than that observed in cid1 or in ccr4 cells (P = 0.0005 and P = 0.006, respectively). Taken together, these data suggest that the Cid1-mediated mRNA decapping pathway involves the Lsm1 protein, and by extension the Lsm1–7 complex, downstream from the mRNA uridylation step. Moreover, as the Lsm1–7 complex is also known to mediate decapping after deadenylation34, we propose that this complex also acts downstream from Ccr4-mediated deadenylation and, thus, stimulates decapping after either uridylation or deadenylation.
Discussion
Uridylation of mRNAs and noncoding RNAs, such as U6 snRNA11, has been described in fission yeast and metazoans, but until now the significance of this modification for bulk, polyadenylated messages has been unclear. The recent observation that uridylation of replication-dependent histone mRNAs stimulates their decapping15 provided a valuable clue, but left open the question of whether this was a decay pathway specific for these nonpolyadenylated messages. Metazoan replication-dependent histone mRNAs differ from other mRNAs in that they contain a 3' stem-loop structure rather than a poly(A) tail16, 36. This structure, bound by the stem-loop binding protein (SLBP), is responsible for the instability of these transcripts, which are degraded at the end of S phase16. Although decay of these messages is dependent on both SLBP and Upf1, an important nonsense-mediated decay factor14, 16, these proteins are not thought to be involved in general mRNA decay. Given such differences in these decay pathways, we wished to determine whether uridylation-stimulated decapping was histone specific or whether uridylation formed the basis of a hitherto undiscovered pathway for bulk mRNA decay.
Here we have identified two novel pathways for bulk mRNA decay in fission yeast, which act in parallel with the classical deadenylation-dependent pathway (Fig. 8). First, cytoplasmic uridylation, although absent from budding yeast (Supplementary Fig. 2), may be used in fission yeast to stimulate decapping and decay. A significant fraction of six functionally diverse transcripts, act1, adh1, gar2, hcn1, pof9 and urg1, were uridylated. It is important to note that the nature of this modification, typically one to two uridyl residues, precludes the use of genome-wide hybridization techniques, such as oligo(dA)-primed reverse transcription. Alternative strategies will be required to catalog all of the transcripts affected by uridylation in fission yeast.
In S. pombe, some transcripts are also subject to deadenylation- and uridylation-independent decapping. This stands in contrast to the situation in budding yeast, where decay is initiated by deadenylation to a tail shorter than 11 nt1 (Supplementary Fig. 2). This deadenylation-independent pathway in fission yeast may reflect similarities to mammalian systems: here poly(A) tails are generally longer than in S. pombe and decapping can occur after initial deadenylation shortens the poly(A) tail to 30–60 nt6, 27.
Uridylation is largely Cid1 dependent and precedes decapping (Fig. 8). Although we have captured rare non-adenylated products with (U)8 and (U)15 tails (O.S.R., unpublished data), the vast majority of uridylation on polyadenylated messages unexpectedly involves only one or two uridyl residues. It is formally possible that we did not observe long uridyl tails on polyadenylated messages because of an inability of RNA ligase to efficiently ligate poly(A)-oligo(U) hairpin structures. However, this seems unlikely for three reasons. First, we isolated several S. cerevisiae act1 cRACE products in which the cleavage site, GTA(U)5AA, was followed by an oligo(A) tail; this construct would also be able to form hairpin structures. Similarly, in our act1 HSC-RACE analysis on RNA from dcp1-ts cells, we were also able to isolate species with (A)8(U)4 and (A)10(U)4 tails. Taken together, these data suggest that the predominance of mono- and di-uridylation in wild-type cells does not result from any inability to ligate longer oligo(U) tails. Second, we also observed longer uridyl tails upon inhibition of decapping. Finally, no longer oligo(U) tails were identified when the RNA circularization step was performed at 65 °C with a thermostable ligase (data not shown). Thus, we suggest that, on polyadenylated messages in vivo, Cid1 may have distributive rather than processive polymerase activity. As we observed only one or two uridyl residues on the various decapped, polyadenylated transcripts analyzed in wild-type cells, longer oligo(U) tracts seem to be unnecessary for the mediation of decapping (see below). It is also possible that there is a highly efficient poly(U) exonuclease in S. pombe and that uridylation itself may be reversible.
Given that Dcp1 is required for viability in S. pombe21 whereas ski2 cells show few differences from wild-type cells (O.S.R., unpublished data), it seems that, as in budding yeast1, 5' 3' mRNA decay is the major degradative pathway. However, in contrast to what occurs in budding yeast, multiple redundant pathways can stimulate decapping. Notably, urg1 mRNA was stabilized in cells lacking Cid1 (Fig. 6). The magnitude of stabilization observed in our study is similar to that seen in S. cerevisiae ccr4 cells26.
In addition, urg1 mRNA was also stabilized to a similar extent in ccr4 cells, indicating that Ccr4–Caf1 is the major cytoplasmic deadenylase in S. pombe. In contrast, the roles of Pan2 and PARN, additional deadenylases, seem to be minor during bulk mRNA decay, as inactivation of either had no effect on urg1 mRNA stability (Fig. 6). It may be that the S. pombe Pan2–Pan3 complex has a role in nuclear trimming of poly(A) tails, as has been shown in budding yeast28, 29, rather than a role in mRNA decay, as is the case in higher eukaryotes27, 37, although this would not explain the observed increase in mRNA uridylation in pan2 and pan3 cells (Fig. 6a). Dissection of the specific roles of the Pan2–Pan3 complex and PARN is currently being pursued.
The cRACE analysis of urg1 mRNAs in cid1 and ccr4 cells strongly suggests that uridylation and deadenylation act redundantly. Although we found evidence for the existence of a second, minor uridylating enzyme, Cid1 was responsible for most mRNA uridylation (Figs. 2 and 6), and uridylation was significantly decreased in cid1 cells. Consistent with redundancy of uridylation and deadenylation, the poly(A) tails of decapped act1 and urg1 mRNAs were significantly shorter in these cells than in wild-type cells (Figs. 2 and 6). This difference arises specifically during mRNA decay, as we observed no difference in poly(A) tail lengths on capped act1 messages from cid1 and wild-type cells. Conversely, in ccr4 cells, where deadenylation is impaired, uridylation of decapped, adenylated urg1 cRACE products significantly increased when compared to those from wild-type cells.
We suggest that decapping of uridylated mRNAs is mediated by Lsm1 (Fig. 7), and by extension the Lsm1-7 complex, which is a known enhancer of decapping34. Although we observed only one or two terminal uridyl residues in wild-type cells, we suggest that this modification may be recognized by the Lsm1–7 complex for two reasons. First, a recent study15 has shown that decapping of human replication-dependent histone mRNAs involves oligouridylation and is mediated by the Lsm1–7 complex. Second, an in vitro analysis of decapping supports this interpretation35. In this study, Song and Kiledjian demonstrated that even a single terminal uridyl residue was able to stimulate decapping in cellular extracts. These authors proposed that uridylation of act1 mRNA, which we had previously described9, might indeed trigger decapping mediated by the Lsm1–7 complex. Considering the cRACE analysis presented here (Fig. 7), we propose that the Lsm1–7 complex is responsible for recognizing even mono-uridylated transcripts and stimulating decapping in fission yeast.
The Lsm1–7 complex is known to localize to P-bodies38. These cytoplasmic foci are sites of RNA decay, but they have also been shown to store translationally repressed mRNAs32, 38, 39. Presumably, recruitment of the Lsm1–7 complex results in the localization of uridylated mRNAs to P-bodies. It may be that, during times of cellular stress and/or for some transcripts, P-body localization elicits translational repression, rather than mRNA decay.
It is likely that the Lsm1–7 complex is also involved in deadenylation-dependent decapping in fission yeast, as has been described in budding yeast34. Notably, S. pombe lsm1 cells, similar to dcp1-ts cells, show a slow-growth phenotype, whereas those lacking Cid1 or cytoplasmic deadenylases do not (O.S.R., unpublished data). Moreover, in lsm1 cells, urg1 mRNA was significantly stabilized in comparison with that in cid1 and ccr4 cells, consistent with the Lsm1–7 complex acting downstream of both uridylation and deadenylation. Although urg1 transcripts were not stabilized in cid1ccr4 cells, relative to those in either single deletion, this result may reflect overlapping deadenylase activities. Accordingly, it may be necessary to create various triple-deletion strains, such as ccr4cid1pan2, to recapitulate the growth defect and extended urg1 half-life seen in lsm1 cells, in which decapping after both deadenylation and uridylation is inhibited.
Budding yeast lacks mRNA uridylation (Supplementary Fig. 2), and RNA decay in this organism principally occurs through deadenylation-dependent pathways2, 18. Notably, however, given the observations of histone mRNA uridylation15, the described PUP/TUTase activities of the human Cid1-like enzymes9, 10, and observations presented here, uridylation is likely to have an important role in the decay of polyadenylated mRNA in higher eukaryotes.
Note: Supplementary information is available on the Nature Structural & Molecular Biology website.
or 
is inactivated
-like nucleotidyltransferase superfamily: identification of three new families, classification and evolutionary history
g of this RNA overnight with 1 unit of T4 RNA ligase (New England Biolabs) in a 400 
-32P] dCTP (Perkin Elmer) using the Rediprime II kit (Amersham), according to manufacturer's instructions. We hybridized probes to the membrane overnight at 60 °C in ExpressHyb solution (Clontech) and then washed them according to manufacturer's instructions. We exposed the membrane to a phosphor-screen (Molecular Dynamics), which we subsequently scanned using a FLA5000 Fuji phosphoimager. We analyzed images using AIDA software (Raytest GmbH), normalized the urg1 signal to the pik1 signal and then compared it to that at the 0 min time point. We calculated half-lives using a semi-log plot and calculating the line of best fit using Microsoft Excel.
