CUCU Modification of mRNA Promotes Decapping and Transcript Degradation in Aspergillus nidulans

Abstract

In eukaryotes, mRNA decay is generally initiated by the shortening of the poly(A) tail mediated by the major deadenylase complex Ccr4-Caf1-Not. The deadenylated transcript is then rapidly degraded, primarily via the decapping-dependent pathway. Here we report that in Aspergillus nidulans both the Caf1 and Ccr4 orthologues are functionally distinct deadenylases in vivo: Caf1 is required for the regulated degradation of specific transcripts, and Ccr4 is responsible for basal degradation. Intriguingly disruption of the Ccr4-Caf1-Not complex leads to deadenylation-independent decapping. Additionally, decapping is correlated with a novel transcript modification, addition of a CUCU sequence. A member of the nucleotidyltransferase superfamily, CutA, is required for this modification, and its disruption leads to a reduced rate of decapping and subsequent transcript degradation. We propose that 3′ modification of adenylated mRNA, which is likely to represent a common eukaryotic process, primes the transcript for decapping and efficient degradation.

The primary stability determinants of eukaryotic mRNA are two cotranscriptional modifications, the 5′ cap and the 3′ poly(A) tail. Shortening of the poly(A) tail is generally the rate-limiting step, which triggers either transcript decapping and subsequent 5′-3′degradation (70) or exosome-dependent 3′-5′ decay (10, 45). Consequently deadenylation represents a critical control point in mRNA turnover. Several proteins possessing deadenylase activity have been identified, but their specific roles remain controversial (32). Studies of a variety of eukaryotes have identified distinct conserved enzyme complexes with deadenylase activity, including Ccr4-Caf1-Not (2, 66, 72) and PAN (8, 9, 54). Both these nuclease complexes are highly conserved between fungi, plants, and animals, unlike a third deadenylase, PARN/DAN (32, 36, 44), which has a very sporadic distribution.

The Ccr4-Caf1-Not complex contains two nucleases, Ccr4 and Caf1, and several accessory proteins, Not1p to Not5p (18, 19). The Ccr4-Caf1-Not complex provides the major deadenylase activity in vivo and is responsible for most cytoplasmic deadenylation of mRNA (54, 71, 72). Additionally, this complex has been implicated in other roles including the repression of transcription and protein modification (5, 18, 25).

The requirement for two distinct nucleases within the Ccr4-Caf1-Not complex is unexplained. In Saccharomyces cerevisiae Ccr4 is the predominant poly(A) nuclease (71, 72), and while Caf1 does not appear to be an active nuclease, it retains Ccr4 within the complex (13, 71) and is a target for specific RNA binding proteins (e.g., PUF5) which regulate deadenylation of specific transcripts (31). However, in a wide range of other organisms including plants, metazoans, and protists Caf1 appears to play a direct role in deadenylation (7, 40, 43, 60, 61, 66, 74). Where analyzed, Caf1 has been shown to interact directly with Ccr4, and this interaction is required to maintain Ccr4 within the multifunctional Ccr4-Caf1-Not complex (18, 19).

Shortening the mRNA poly(A) tail to around 15 residues triggers decapping (21, 24, 48), which facilitates very rapid degradation of the transcript by 5′ to 3′ exonucleolytic decay. Decapping is therefore a critical process that must be tightly controlled. However, the mechanism that triggers decapping in response to deadenylation is poorly defined. Recent studies on decapping factors have revealed that the Lsm1-7 complex, which enhances decapping, has a strong binding preference for oligoadenylated RNAs over nonadenylated or polyadenylated RNAs (15). Furthermore, in mammalian cells 3′ oligouridylation stimulates decapping of nonadenylated RNAs in an Lsm-dependent manner (50, 63). Uridylation of RNA has been shown to be catalyzed by members of the Cid family of noncanonical polynucleotide polymerases. In fission yeast, actin mRNA is uridylated in a Cid1-dependent manner upon S-phase arrest (56), and an independent study has recently shown that Cid1 and at least one other polynucleotide polymerase uridylates adenylated transcripts immediately prior to decapping (57). Hs2, the human orthologue of Cid1, also displays poly(U) polymerase activity in vitro (38), and two other related proteins are responsible for histone mRNA uridylation in human cells, TUTase 1 and 3 (50). Thus, modifications at the 3′ end of RNA are emerging as putative signals for mRNA fate (50, 62) and may represent a critical step in the decapping pathway.

In the filamentous fungus Aspergillus nidulans, differential transcript degradation represents an integral component of the cellular response to nitrogen availability (11, 46, 47, 55). High intracellular glutamine (Gln) levels signal N sufficiency, leading to the rapid deadenylation and subsequent degradation of a specific subset of transcripts (11, 12, 46, 47). In order to identify which of the deadenylases are involved, we report here the systematic deletion of the genes encoding the three putative deadenylases in A. nidulans, pan2, ccr4, and caf1. We have shown that Ccr4 represents the major deadenylase activity, but Caf1 is required for accelerated degradation in response to the Gln signal. We have established that Caf1 is involved in preventing deadenylation-independent decapping. Cells lacking Caf1 possess decapped transcripts with abnormally long poly(A) tails of between 20 and 45 residues. Finally, we discovered that about 14% of the polyadenylated gdhA transcripts are modified by the addition of nucleotides with the consensus CUCU. In the wild type this modification is seen only on transcripts with short (∼15-nucleotide [nt]) poly(A) tails and occurs on 21% of natively decapped adenylated transcripts. A member of the nucleotidyltransferase superfamily, CutA, is required for CU modification, and its disruption leads to reduced rates of transcript degradation. We propose that CUCU modification is an integral step in transcript degradation, facilitating mRNA decapping.

MATERIALS AND METHODS

A. nidulans strains and genetic techniques.

A. nidulans strains carried markers in standard use (16, 17). Standard genetic techniques were used (16). Growth media were as described by Cove (22). Δpan2, Δccr4, Δcaf1, and ΔcutA strains were constructed by direct transformation of recombinant PCR constructs utilizing Aspergillus fumigatus pyrG (Δcaf1 and ΔcutA strains) or Neurospora crassa pyr4 (Δpan2 and Δccr4 strains) as the selectable marker, as described by Szewczyk et al. (64). The recipient strains were the pyrG89 riboB2 (Δpan2), sE51 pyrG89 wA3 fwA1 chaA1 pyroA4 nirA⁻ ΔnkuA::argB⁺ (Δccr4 and Δcaf1), and pyrG89 pabaB22 riboB2 ΔnkuA::argB⁺ (ΔcutA) strains.

Transcript analysis.

Growth of mycelia, RNA preparation, and quantitative Northern and RNase H analyses were as described previously (12, 47). Briefly, overnight cultures were incubated with ammonium as the N source, washed, and transferred to N-free media for 1 h prior to sampling. As described previously, proflavin was added to the cultures, to inhibit transcription, 10 min prior to the first sample being taken (23, 55). Analysis of bulk poly(A) tail length was conducted as described by Temme et al. (66) with minor modification. Two micrograms of total RNA was incubated in a 20-μl reaction volume with 10 μCi 3′-[α-³²P]dATP and 300 U yeast poly(A) polymerase (USB) for 1 h at 37°C. RNA was purified by ethanol precipitation and dissolved in 5 μl water. Labeled RNA (500,000 cpm) was digested in a 20-μl reaction volume with 1 μl RNase A/T1 (Ambion) in 50 mM Tris-HCl (pH 7.5), 100 mM KCl, and 10 mM MgCl₂ together with 20 μg yeast RNA (Ambion) for 30 min at 30°C. The digested RNA was ethanol precipitated and separated on 10% polyacrylamide-urea gels. Gels were analyzed on a Storm phosphorimager, and data were analyzed using ImageQuant software. Circularized reverse transcriptase PCR (cRT PCR) was conducted according to reference (50). First strand synthesis utilized the oligonucleotide gdhA_circ_P2 (TTGAAGCAGCAGAGAAGCAA), and PCR amplification utilized oligonucleotides gdhA_circ_P1 (AGTTCGCTACTCCTGCCAAG) and gdhA_circ_P2. The PCR products were cloned into pGEM-T Easy (Promega), and random clones were selected and sequenced. The nonnormal distributions of poly(A) tail lengths were analyzed using the nonparametic statistics Mann-Whitney test and Geni coefficient. The R-by-c chi-square test for trend was used to assess the CUCU motif (StatsDirect).

Bioinformatic analysis.

Database searches in the two major branches of the nucleotidyltransferase superfamily were conducted using PSI-BLAST (3) and query sequences of Schizosaccharomyces pombe Cid1 (class 1) and Bacillus stearothermophilus CCA-adding enzyme (class 2). Profiles were calculated after five iterations with default parameters, including an E-value cutoff of 0.005, and used to query proteins predicted from complete genomes. Genome sequences for A. nidulans (30), A. fumigatus (51), N. crassa (29), and Ustilago maydis were obtained from the Broad Institute website (http://www.broadinstitute.org/), while sequences for Homo sapiens, Arabidopsis thaliana, and Caenorhabditis elegans were obtained as species-specific subsets of Uniprot (52). Class 1 homologues retrieved from genomes were aligned against their PSI-BLAST profiles using CLUSTALW (14). Redundant sequences were removed using JALVIEW (76), which was also used to manipulate alignments. The resulting alignment was used for phylogenetic analysis using MEGA4 (37). The evolutionary history was inferred using the neighbor-joining method (59) using bootstrapping (1,000 replicates) to estimate reliability (26). The evolutionary distances were computed using the Poisson correction method (81) and were in the units of the number of amino acid substitutions per site. All positions containing alignment gaps and missing data were eliminated only in pairwise sequence comparisons. Default MEGA4 options for minimum evolution tree calculations were also tested. The salient tree features mentioned were present in both outputs, irrespective of calculation method (not shown).

RESULTS

Caf1 is required for regulated mRNA degradation.

We have shown previously that in A. nidulans the regulated degradation of specific transcripts in response to Gln correlates with their deadenylation rate (12, 47). In order to identify which of the deadenylases are involved, we systematically deleted the genes encoding the three putative deadenylases identified within the A. nidulans genome (30), orthologues of PAN2 (AN2065), CCR4 (AN3602), and CAF1 (AN7628). These have been named pan2, ccr4, and caf1, respectively. The A. nidulans genome does not encode a PARN orthologue. Deletion of pan2 did not alter the measured rates of transcript degradation (see Fig. S1 in the supplemental material), consistent with its putative role in mRNA maturation and poly(A) tail trimming (8, 9). However, deletion of either ccr4 or caf1 led to significantly altered transcript degradation profiles (Fig. 1), consistent with their encoding key components of the major deadenylase complex (54, 71, 72). From these data it is evident that disruption of ccr4 leads to a dramatic stabilization of transcripts, but, surprisingly, the rate of decay still increases significantly in the presence of Gln. Deletion of caf1 did not alter basal degradation (nitrogen-limiting conditions), consistent with Ccr4 being the principal deadenylase for general mRNA decay. Critically, the Gln response is greatly reduced in the Δcaf1 strain, with no significant difference in degradation rates between the different nitrogen regimens being observed. Deletion of either caf1 or ccr4 leads to a significantly reduced growth rate and loss of sporulation (data not shown). Our inability to obtain the double mutant may indicate that at least one of the two deadenylases is required for viability, which is consistent with the situation in other organisms (19). In summary, these data reveal that the two putative deadenylases have distinct roles, Ccr4 being required for general degradation and Caf1 mediating the regulated degradation of specific transcripts.

FIG. 1.

Northern analysis of areA transcript degradation in deadenylase mutants. (A) Northern blot analysis of areA (encoding a global transcription factor mediating nitrogen regulation), niaD (encoding nitrate reductase), and meaA (encoding a high-affinity ammonium transporter) was conducted over a 30-min time course, after transcription was inhibited, to monitor degradation rates under conditions of nitrogen starvation (−N) or nitrogen sufficiency (Gln). The wild type (WT) is compared with strains with ccr4 or caf1 deleted, as indicated. 18S rRNA was used as a loading control. Multiple Northern blots (≥3) were quantified, and the data are represented graphically (±standard deviations [SD]), with the y axis being a logarithmic plot of the percentage of transcript remaining (□, −N; •, Gln). (B) Extrapolated half-lives, derived by regression analysis. Between the two growth regimens there was a significant difference (t test, P < 0.05) in half-life for all three transcripts in the wild-type and Δccr4 strains. Conversely, in the Δcaf1 strain no significant difference was observed (areA, P = 0.216; niaD, P = 0.572; meaA, P = 0.865).

Disruption of either Caf1 or Ccr4 leads to deadenylation-independent decapping.

To monitor deadenylation of the areA transcript in vivo, we conducted RNase H Northern analysis (Fig. 2). The most striking observation is that deletion of caf1 appears to dramatically reduce deadenylation of the areA transcript. Although the level of areA mRNA decreases over the 30-min time course, little or no deadenylation is apparent in the absence of cycloheximide. Conversely, in the presence of cycloheximide, which leads to general stabilization of transcripts (6, 34, 47), deadenylation in response to the Gln signal is observed. This suggests that ablation of caf1 and probable disruption of the Ccr4-Caf1-Not complex (18, 19) circumvent the normal requirement for deadenylation to a short poly(A) tail length of ∼15 nucleotides prior to decapping (21, 24, 48). However, when transcript degradation is inhibited by cycloheximide, Gln-signaled deadenylation is observed, suggesting that Ccr4 or another deadenylase is actively recruited to these specific transcripts and retains some activity in the absence of Caf1. We have been unable to construct a Δccr4 Δcaf1 double deletion strain, preventing us from confirming that the residual deadenylation is Ccr4 dependent.

FIG. 2.

RNase H analysis of areA poly(A) tail length in deadenylase mutants. RNase H Northern blot analysis was used to monitor the rate of the deadenylation over a 30-min time course for the areA transcript in the presence (Gln) or absence (−N) of glutamine with or without cycloheximide (CX) as indicated. Transcription was inhibited prior to the time course being initiated. The wild type (WT) is compared with strains with ccr4 or caf1 deleted, as indicated. A separate experiment comparing time zero samples for all three strains on the same gel confirmed that there is no major difference with respect to the maximal poly(A) tail length observed (data not shown). 18S rRNA is used as a loading control. The T₀ sample was treated with RNase H and oligo(dT) and utilized as a size marker for deadenylated transcripts (A₀). As noted previously, 3′ degradation proceeds beyond the poly(A) tail after cycloheximide treatment in the wild type (47); this is also the case for the Δcaf4 strain but not the Δcaf1 strain.

In the Δccr4 strain the rate of deadenylation is retarded, with a significant proportion of the transcripts retaining a long poly(A) tail throughout the time course, after Gln treatment. However, unlike what is found for the Δcaf1 strain, deadenylated transcripts are observed in the absence of cycloheximide, although the proportion does not increase over time. This is again consistent with a major proportion of transcripts being degraded independently of deadenylation. In the presence of cycloheximide, accelerated deadenylation in response to Gln is apparent and the proportion of deadenylated transcripts increases during the time course. In the Δccr4 strain 3′ degradation also proceeds beyond the fully deadenylated transcript (A₀), but not to the extent observed in the wild type. These data imply that, although deadenylation does take place in the Δccr4 strain, it is significantly reduced. The residual deadenylase activity may be that of Caf1 or another deadenylase.

To assess whether these observations represent general phenomena or are specific to the areA transcript, we undertook RNase H analysis of niaD mRNA, which encodes nitrate reductase, and essentially observed the same trends (data not shown). To confirm that this was not restricted to transcripts whose stability is differentially regulated by nitrogen status, we profiled the poly(A) tail length of bulk mRNA (Fig. 3). The two mutant strains and wild type produced distinct profiles. In particular, the Δcaf1 strain has a very low proportion of transcripts with short poly(A) tails, similar to the specific profile for areA (Fig. 2). This is consistent with deadenylation-independent decapping being a general consequence of caf1 ablation. The Δccr4 strain, as in the RNase H analysis, revealed two populations of adenylated transcripts. These were centered on 54 and 38 A residues and indicate that the efficiency of the remaining deadenylase activities varies in relation to poly(A) tail length.

FIG. 3.

Poly(A) tail profile. (A) The poly(A) tail length distribution for bulk mRNA was assayed by 3′ end labeling of total RNA. RNA samples were equivalent to the T₀ and T₃₀ Gln plus cycloheximide samples used in Fig. 2. The samples assayed are from the Δccr4 strain at T₀ (lane 1) and T₃₀ (lane 2), from the wild-type (WT) strain at T₀ (lane 3) and T₃₀ (lane 4), and from the Δcaf1 strain at T₀ (lane 5) and T₃₀ (lane 6). (B) The profile for each T₀ sample was obtained using a phosphorimager, and the data from triplicate experiments were combined prior to normalization. The standard errors are indicated. A DNA sequencing run was included on the gels to provide a size marker. As with RNase H analysis, distinct profiles were obtained; in particular, the Δcaf1 strain shows a relatively small proportion of transcripts with very short poly(A) tails.

Poly(A) tail length and modification.

RNase H and bulk poly(A) tail analysis provides indirect evidence for increased prevalence of deadenylation-independent decapping in both Δcaf1 and Δccr4 strains. In order to address this directly, we undertook a detailed analysis of the gdhA transcript. This is a highly expressed gene which encodes glutamate dehydrogenase, and its transcript is not subject to Gln-signaled degradation (12). To characterize the poly(A) tail distribution, we conducted cRT PCR, which involves RNA ligase-mediated transcript circularization. The ligation products are then subjected to a reverse transcriptase PCR, amplifying the region across the 5′-3′ boundary. The PCR products are then cloned and random samples sequenced in order to specifically define the length of the poly(A) tail. In order to distinguish decapped transcripts from the general mRNA pool, we utilized differential treatment with tobacco acid pyrophosphatase (TAP), which decaps mRNA (27). Ligation without TAP pretreatment will circularize only decapped transcripts, while ligation after TAP treatment will identify both capped and decapped transcripts. From this analysis (Fig. 4) the distribution of poly(A) tail length is consistent with RNase H and bulk poly(A) data; the Δcaf1 strain exhibits relatively long poly(A) tails, with no significant difference in distribution between decapped and total mRNA, the medians being A₃₁ and A₃₀, respectively (P = 0.89) (see Table S1 in the supplemental material). In the wild type the distributions were significantly different for these two treatments, with the naturally decapped sample exhibiting a poly(A) tail distribution with a median of 0, compared with A₂₀ for the TAP-treated sample (P < 0.0001) (see Table S1 in the supplemental material). The native decapped samples from the Δccr4 strain revealed the same broad distribution as the total RNA pool in the wild type (P < 0.0001) (see Table S1 in the supplemental material), but this distribution was different from that of native decapped wild-type RNA (P = 0.35). These data support the observation made for areA mRNA, utilizing RNase H analysis, that deletion of ccr4 or caf1 and the consequent disruption of the Ccr4-Caf1-Not complex promote deadenylation-independent decapping.

FIG. 4.

cRT PCR analysis of gdhA mRNA. (A) Poly(A) tail length was determined by cRT PCR and sequencing of RNA samples derived from the wild-type, Δcaf1, ΔcutA, and Δccr4 strains, with (+) or without (−) pretreatment with TAP to remove the 5′ cap structure. The distribution of the poly(A) tail lengths is displayed using a box plot, where the top and bottom of the box represent limits of the upper and lower quartiles, with the median being indicated by the horizontal line which lies within the box. The whiskers show the highest and lowest reading within 1.5 times the interquartile range. The outliers are indicated (□). Projected onto this is the distribution of clones that include CUCU-derived modifications (○). The sequences of specific modifications identified and their respective poly(A) tail lengths are shown. These data are derived from three separate experiments. The total numbers of transcripts analyzed are also given. Based on semiquantitative PCR, the wild-type TAP-untreated sample had <2% uncapped transcripts, compared to the treated sample (see Fig. S3 in the supplemental material). (B) Distribution of poly(A) tail lengths in the wild type (WT) and ΔcutA strains without pretreatment with TAP. The x axis shows those with a tail in A₈ groups.

Among the polyadenylated transcripts derived from the wild type, Δccr4, and Δcaf1 strains, a small proportion (∼14%) exhibited an additional element with the consensus CUCU (Fig. 4). This analysis cannot distinguish whether CUCU modification occurs at the 3′ or 5′ end, but it was observed only among transcripts that retained a significant poly(A) tail. In the wild type this averaged 15 A residues, with very little variation, correlating exactly with the expected point at which poly(A) tail shortening triggers decapping (21, 24, 48). For both the Δcaf1 and Δccr4 strains, in which deadenylation-independent decapping is prevalent, the poly(A) tail lengths of modified transcripts vary widely, ranging from A₂₄ to A₃₉ and A₁₁ to A₃₅, respectively. These data are consistent with the 3′ modification of adenylated transcripts serving as a signal for decapping.

Identification of candidate CUCU-adding enzymes.

The vast nucleotidyltransferase superfamily can be split into two major branches (79). The presence of poly(A) polymerases, RNA-editing 3′ terminal uridylyltransferases from Trypanosoma brucei, humans, and, as recently described (57), S. pombe, in class 1 suggested that this group likely contained the CUCU-adding enzyme. However, although archaeal CCA-adding tRNA-processing enzymes are found in class 1, the independently evolved bacterial/eukaryotic class of CCA-adding enzymes are found in class 2 (4). Thus, we could not rule out the possibility that the CUCU enzyme would also be located in class 2. This observation was particularly important since these tRNA-processing enzymes appear to be the only precedents in the superfamily of proteins which can switch nucleotide specificity.

Our search of the A. nidulans genome using a class 2 lb/in²-BLAST profile (see Materials and Methods) revealed a single homologue, AN10924. This is presently annotated as the product of an uncharacterized open reading frame (ORF), but reciprocal genome BLASTs showed a clear correspondence with annotated tRNA nucleotidyltransferases in S. cerevisiae (CCA1_YEAST) (1) and S. pombe (CCA1_SCHPO and CCA2_SCHPO). A similar search with the class 1 lb/in²-BLAST profile produced three hits, AN7748, AN0874, and AN5694 in descending order of significance (E values of 1 × 10⁻⁵⁹, 2 × 10⁻⁴⁶, and 2 × 10⁻⁴⁵, respectively). Of these, AN0874 could be straightforwardly discounted on the basis of clear matching, by reciprocal genome BLASTs and domain architectures, to poly(A) polymerases from S. cerevisiae (PAP_YEAST) (41) and S. pombe (PAP_SCHPO) (53). Similar analysis suggested that the closest S. pombe relatives to AN7748 were Cid13 and Cid11, while Cid14 was closest to AN5694. However, since the relationships were less clear, we carried out a more rigorous bootstrapped phylogenetic analysis, by several different algorithms, of all class 1 homologues that could be detected in a range of fungal and other genomes (see Materials and Methods).

Phylogenetic analysis (see Fig. S2 in the supplemental material) confirmed our annotation of AN0874 as a poly(A) polymerase. Each fungal genome contains a single representative, the group of which clusters with human and C. elegans proteins and, more distantly, with several Arabidopsis sequences. All trees gave a monophyletic group comprising AN5694 with its orthologue in A. fumigatus (Afu7g04130) and the N. crassa protein NCUT_05553. This group clusters reliably with a clade including the experimentally characterized nuclear poly(A) RNA polymerases from S. cerevisiae (PAP2_YEAST) (73) and S. pombe (CID14_SCHPO) (77). AN5694 and the other proteins mentioned above are therefore likely to be nuclear poly(A) RNA polymerases.

Our analysis consistently supported a monophyletic origin for AN7748, Afu5g07790 from A. fumigatus, and both Cid11 and Cid13 from S. pombe. While Cid11 remains uncharacterized, apart from a known mixed cytosolic and nuclear subcellular distribution (42), Cid13 is involved in specific regulation of cytosolic suc22 mRNA (58). S. pombe Cid1 has recently been implicated in the uridylation of mRNA (57). Our phylogenetic analysis shows that this consistently groups with NCUT_05266 from N. crassa and U. maydis UM05013, but no A. nidulans protein. Having assigned likely functions to AN0874 and AN5694, and in view of a finding suggestive of the evolutionary grouping of AN7748 with Cid13, a known mRNA-modifying regulatory nucleotidyltransferase, we considered AN7748 as the favored candidate CUCU-adding enzyme.

CutA is required for CUCU modification.

Based on our bioinformatic analysis we deleted the AN7748 gene, which was subsequently designated cutA (CU nucleotidyltransferase). Morphologically the strain resembled the wild type, with no abnormal growth phenotype on hydroxyurea and/or caffeine (data not shown). This observation distinguishes the ΔcutA strain from Δcid1 and Δcid13 strains of S. pombe (58, 75). Northern analysis revealed significant retardation of transcript degradation, significantly increasing the half-lives of areA, meaA, and gdhA mRNA (Fig. 5). In order to establish whether cutA is required for CU modification of transcripts, we undertook cRT PCR. For the natively decapped gdhA transcripts, no modifications were observed among 55 adenylated transcripts derived from the ΔcutA strain, which is unique for the three genes in this study. Among 58 TAP-treated samples, which included capped and decapped transcripts, two showed 3′ modification, one UU and one CUUUUUUUUU, which are significantly different from the consensus modification observed in the wild type (P < 0.0003; chi-square) (see Table S1 in the supplemental material), suggesting that there is a distinct residual activity, possibly associated with AN5694.

FIG. 5.

Northern analysis of areA transcript degradation in the ΔcutA mutant. (A) Northern blot analyses of areA, gdhA, and meaA mRNA under conditions of nitrogen starvation (−N) or nitrogen sufficiency (Gln) were conducted as for Fig. 1. The wild type (WT) is compared with the strain with cutA deleted, as indicated. 18S rRNA was used as a loading control. Multiple Northern blots (≥3) were quantified for the wild type (dashed lines) incubated in the absence of nitrogen (○) or in the presence of Gln (•) or the ΔcutA strain (solid lines) incubated in the absence of nitrogen (□) or in the presence of Gln (▪), and the data were represented graphically (±SD), with the y axis being a logarithmic plot of the percentage of transcript remaining. (B) The extrapolated half-lives for areA, gdhA, and meaA transcripts, under both nitrogen regimens, are given to facilitate comparison of the three strains.

Deletion of cutA had a dramatic effect on the distribution of gdhA poly(A) tail lengths. Unlike what was found for the wild type, TAP treatment made no significant difference (P = 0.05; Gini coefficient) to the overall transcript profile, even though the median length of natively decapped transcripts was significantly shorter than that of the total gdhA transcript population (A₂ versus A₇; P < 0.0001; Mann-Whitney). This reflects a dramatic increase in the proportion of transcripts with very short (A₁ to A₁₅) poly(A) tails among the gdhA mRNA population (TAP treated) compared with the wild type (P = 0.0094 [Mann-Whitney]; P < 0.05 [Gini coefficient]) (Fig. 4A; see Table S1 in the supplemental material). Additionally, in the ΔcutA strain the proportion of natively decapped transcripts that were fully deadenylated (A₀) was significantly reduced (P = 0.0001) (Fig. 4B). These data imply that the deletion of cutA leads to both a delay in decapping, which normally occurs efficiently when the transcript has been deadenylated to A₁₅, and a reduction in the pace of the final stages of deadenylation (<A₁₅).

DISCUSSION

From the A. nidulans genome sequence (30) we identified orthologues of all the major components involved in eukaryotic transcript degradation, consistent with conservation of their associated biological processes. We investigated potential roles for Pan2, Ccr4, and Caf1 in the regulated degradation of specific transcripts. From this we determined that Caf1 and Ccr4 play distinct roles in transcript deadenylation and degradation.

Ablation of ccr4 leads to general stabilization of transcripts (Fig. 1), consistent with its accepted role as a major deadenylase. However, Gln-signaled transcript degradation persists in the Δccr4 strain, albeit at a diminished rate, indicating that the regulated response is not primarily dependent on Ccr4. Both RNase H (Fig. 2) and bulk poly(A) tail analyses of the Δccr4 strain (Fig. 3) revealed reduced deadenylation, although a distinct proportion of the transcripts had short poly(A) tails. This bimodal distribution of poly(A) tail length demonstrates that the efficiencies of the remaining deadenylase activities vary, possibly in relation to factors such as the proximity of the 3′ end to the poly(A) binding protein.

Northern analysis shows that Gln-signaled transcript degradation is largely Caf1 dependent (Fig. 1). The orthologous proteins in both S. cerevisiae and Drosophila melanogaster have previously been identified as key regulatory targets within the Ccr4-Caf1-Not complex, facilitating the modulation of deadenylation at specific transcripts (31, 33, 35). Two key questions arise from these observations: whether this function is associated with Caf1 deadenylase activity and whether this role is critical for the cellular response to other regulatory signals.

In eukaryotes, deadenylation to a short poly(A) tail (∼15 nt) precedes decapping and subsequent, rapid 5′-3′ degradation for the majority of transcripts (20, 21, 70, 80). The cRT PCR analysis of gdhA mRNA indicates that this is the case for A. nidulans, with <3% of natively decapped transcripts having poly(A) tails of 16 or more A nucleotides in the wild type (Fig. 4 and data not shown). However, consistent with RNase H and bulk poly(A) tail analyses, cRT PCR revealed that in the Δccr4 strain >40% of transcripts had undergone deadenylation-independent decapping. Increased prevalence of deadenylation-independent decapping has also been observed in an S. pombe Δccr4 strain (57). The occurrence of deadenylation-independent decapping was even more apparent in the Δcaf1 strain (Fig. 2, 3, and 4). In particular, cRT PCR analysis of gdhA mRNA (Fig. 4; see the supplemental material) revealed that >90% were decapped prior to deadenylation to ≤15 A nucleotides. However, in the presence of cycloheximide, deadenylation is observed (Fig. 2), demonstrating that an active deadenylase is present. It is possible that cycloheximide in some way increases deadenylase activity, either activating or stabilizing an enzyme, but the most likely explanation is that in the absence of cycloheximide deadenylase activity is obscured by decapping and subsequent 5′-3′ degradation of the transcripts prior to poly(A) shortening. Importantly, from a review of published data for a range of other organisms, including S. pombe, Trypanosoma brucei, D. melanogaster, and Homo sapiens, it is apparent that deletion or knockdown of caf1 orthologues leads to a similar profile with a general loss of transcripts having short poly(A) tails (28, 61, 65, 66). These data suggest that a general consequence of disrupting either deadenylase within the Ccr4-Caf1-Not complex is that a large proportion of transcripts undergo deadenylation-independent decapping.

From our data, decapping of the gdhA transcript appears to be associated with CUCU modification, and we have identified that CutA is required for this novel catalytic activity. Deletion of cutA leads to a significant reduction of 3′ end modification, the residual activity being primarily polyuridylation. Most known nucleotidyltransferases catalyze the repetitive attachment of a single base, A or U, with the tRNA-modifying CCA-adding enzymes being an exception. Evidence suggests that this activity evolved twice (4), with only very distant homology between archaeal enzymes (class 1) and the eukaryotic/bacterial enzymes (class 2). Structural data are available for both types, but they shed light only on the mechanism of alternating nucleotide substrate specificity in the latter case (39). In contrast, nucleotide binding to the Archaeoglobus fulgidus enzyme is nonspecific, resulting in the hypothesis that the incoming tRNA substrate influences specificity (78). For this reason, sequence and structure analyses to explain the predominantly alternating substrate specificity of CutA were not attempted. One notable characteristic of CutA is the long C-terminal region of 650 residues, which is predicted to be almost entirely intrinsically disordered. It is known that intrinsic disorder is strongly correlated with protein-protein interactions (69), and this may be relevant to the control of CutA activity.

Formally it is possible that the CUCU modification is added to the 5′ end of the transcript. However, in the Δcaf1 strain this modification was observed in both the naturally decapped and total RNA pools. Also the modification is not seen in the wild type on transcripts with very short poly(A) tails (<13 A nucleotides), consistent with its loss being concurrent with the final stages of poly(A) degradation. Additionally, based on phylogeny CutA acts at the 3′ end of the transcript. These observations suggest that the CUCU modification occurs at the 3′ ends of transcripts prior to decapping. As the CUCU element is apparently subject to degradation, the length of the initial sequence could not be established, but the longest we observed was CUCUCUC.

Cycloheximide treatment leads to general stabilization of transcripts. The basis of this is unproven but may relate to cycloheximide's well-characterized role as a translational inhibitor, which results in mRNA remaining bound to the polysomes (6). If stalled ribosomes inhibit degradation of the associated transcripts, one possible role for the CUCU modification could be to tag transcripts for transport away from ribosomes, to a location where they can be more efficiently decapped and degraded. However, in S. cerevisiae, which does not exhibit 3′ mRNA modification (57), it has recently been reported that for PGK1, a highly expressed and relatively stable transcript (49), a large proportion of degradation is cotranslational (34).

The CU-rich element may act by recruiting factors, such as the Lsm1-7 complex, which promote decapping. Consistent with this we have shown that deletion of cutA results in reduced rates of transcript degradation (Fig. 5), with the possibility that reduced efficiency of decapping limits degradation. Additionally, gdhA poly(A) tail length in the ΔcutA strain is significantly different from that in the wild type (Fig. 4; see Table S1 in the supplemental material), consistent with an accumulation of transcripts with short poly(A) tails, which in the wild type would be efficiently decapped and rapidly degraded. The suggested link between CU modification and decapping efficiency would be consistent with the observation that the Lsm complex has greater affinity for uridinylated RNA (15, 63, 67) and that S. pombe strains with lsm1 deleted accumulate transcripts with short poly(A) tails (57).

Why do CUCU modification and decapping occur prior to the poly(A) tail being shortened to ∼A₁₅ in the Δcaf1 and Δccr4 strains? In the wild type, the 3′ ends of transcripts are associated with the Ccr4-Caf1-Not complex. Ablation of caf1 or ccr4 significantly disrupts this complex; in particular, deletion of caf1 probably causes the dissociation of Ccr4 (13, 68, 71). This suggests that the Ccr4-Caf1-Not complex, although progressively deadenylating the mRNA in a controlled manner, also provides protection from premature modification and decapping, both of which are triggered at ∼A₁₅ in the wild type. As described above this also appears to be the case in a range of other organisms.

In the Δcaf1 strain a larger proportion of modified transcripts occur in the total mRNA population than in the decapped sample (24% versus 8%) (Fig. 4), which is not the case for the wild type. Additionally, if transcript modification inexorably leads to decapping, one would expect the mRNA to be relatively unstable in the Δcaf1 and Δccr4 strains, which is not the case. These observations suggest that in the wild type the combined effect of a short poly(A) tail and CUCU modification leads to efficient decapping while in the Δcaf1 strain the long poly(A) tail inhibits this process. This is consistent with the current model, which proposes that decapping is triggered by disruption of the closed circle mRNP structure due to dissociation of poly(A) binding protein from a short poly(A) tail when it is degraded below 15 A nucleotides (20). We therefore propose that CUCU modification, which occurs precisely at this point in the wild type, is integral to this process, promoting the recruitment of decapping factors (e.g., Lsm complex) and the progression into the final stages of mRNA degradation.

Since mRNA 3′ modification immediately prior to decapping has now been described for polyadenylated transcripts in S. pombe (57) and A. nidulans, as well as the nonadenylated human histone mRNAs (58), it may be a general phenomenon in eukaryotes, with S. cerevisiae being one prominent exception (57). If so, it immediately offers a further layer of control for the degradation of specific transcripts and a means of effecting global regulatory responses to signals such as stress. Notably, Cid1 and CutA are not particularly closely evolutionarily related (see Fig. S2 in the supplemental material), suggesting that, even among the fungi surveyed here, other as yet uncharacterized nucleotidyltransferases may have important roles in mRNA modification. The regulation of CutA and its wide range of homologues in other species will inevitably be of great interest and importance.