Uridylation prevents 3′ trimming of oligoadenylated mRNAs

Materials

Arabidopsis thaliana plants, Columbia ecotype (Col-0), were grown on soil with 16 h light/8 h darkness cycles. The characterization of urt1-1 (Salk_087647C) and urt1-2 (WISCDSLOXHS208_08D) T-DNA insertion lines is shown in Supplementary Figure S1. xrn4-3 (SALK_014209) has been described previously (24). Sequences of primers are shown in Supplementary Table S1.

Subcellular localization of URT1

The coding sequence of URT1 (At2g45620), DCP1 (At1g08370) and RBP47 (At1g19130) was fused 5′ to GFP or RFP sequences using the Gateway® destination vectors pB7FWG2 or pH7RWG2, respectively. Biolistic transformation of tobacco BY2 cells and analysis by confocal microscopy were performed by standard methods.

Expression and purification of recombinant proteins

Recombinant His-tagged GST-URT1 or the catalytically inactive mutant GST-URT1^D491/3A were produced in BL21(DE3) cells grown at 17°C. Cells were disrupted by sonication in 20 mM MOPS, pH 7.5, 250 mM KCl, 15% (v/v) glycerol, 1 mM DTT, 0.1% (v/v) Tween 20 in presence of protease inhibitors (Roche). Recombinant proteins were purified by classical Immobilized Metal Affinity Chromatography on Ni-NTA resin followed by glutathion affinity chromatography. Purified proteins were dialysed against 20 mM MOPS, pH 7.5, 100 mM NaCl, 15% (v/v) glycerol, 0.1% (v/v) Tween 20. Aliquots were snap frozen in liquid nitrogen and stored at −80°C.

Activity assays

URT1 and Cid1 activities were compared by using identical amounts of arbitrary units defined by measuring trichloroacetic acid-precipitable incorporation of [α-³²P]-UTP into tRNAs and reaction conditions based on (25). In vitro assays shown in Figure 1A contained 100 nM of GST-URT1 or GST-URT1^D491/3A or 0.1 unit/μl of Cid1 (NEB), 25 mM potassium phosphate, pH 7, 5 mM MgCl₂, 5 mM DTT, 10% (v/v) glycerol, 0.1 μg/μl bovine serum albumin (BSA), 0.1 mM UTP, GTP, CTP or ATP and 5 nM of a 21-nucleotide synthetic RNA 5′-end labelled by T4 Polynucleotide Kinase (NEB) and [γ-³²P]-ATP. For experiments shown in Figure 1B, the reaction buffer was 20 mM MOPS, pH 7.5, 100 mM NaCl, 15% (v/v) glycerol, 0.1% (v/v) Tween 20, 1 mM MgCl₂, 0.5 µg/µl BSA, 1 mM UTP. Reaction products were separated by denaturing 17% (w/v) polyacrylamide gel electrophoresis before autoradiography. To determine the composition of the tails synthesized by URT1 in the presence of the four ribonucleotides, the latter reaction conditions were used except that the concentration for each ribonucleotide was 1 mM. Nucleotide composition was determined by 3′ Rapid Amplification of cDNA ends (RACE) polymerase chain reaction (PCR), cloning and sequencing as detailed below.

3′ RACE and circular reverse transcriptase-polymerase chain reaction (cRT-PCR) analyses

The 3′ RACE PCR protocol used to detect uridylated mRNAs is detailed in (26). Briefly, 5 µg of total RNA was purified with Nucleospin® RNA plant columns (Macherey Nagel), dephosphorylated with calf intestinal phosphatase (CIP, NEB) before ligation to the RNA primer (5′P-CUAGAUGAGACCGUCGACAUGAAUUC-3′NH₂) with T4 RNA ligase (Fermentas). This sequence was used as an anchor to initiate cDNA synthesis by SuperScript® III reverse transcriptase (Invitrogen) and the anchor rv1 primer (Supplementary Table S1). Two nested PCR amplification of 30 cycles were performed using anchor rv1 and rv2 primers in combination with gene-specific sense primers (Supplementary Table S1).

Circular reverse transcriptase (cRT)-PCR analysis was used to characterize the 5′ and 3′ extremities of capped and uncapped LOM1 transcripts. To detect capped transcripts, 5 µg of total RNA was first treated with 5 units of CIP (NEB) to dephosphorylate uncapped RNAs and then with 2.5 units of tobacco acid pyrophosphatase (TAP) (Epicentre) to generate 5′ phosphorylated transcripts from capped mRNAs. The 5′ phosphorylated transcripts were then circularized by 20 units of T4 RNA ligase at 18°C for 16 h in a final volume of 200 µl. Uncapped transcripts were detected following mRNA circularization from total RNA that was neither treated with CIP nor with TAP. Negative controls were treated only with CIP because neither dephosphorylated nor capped mRNAs can be circularized by T4 RNA ligase.

3′ RACE and cRT PCR products were cloned in pGEM®-T Easy (Promega) before sequence analysis.

Polysome purification

Polysome purification was performed according to (27) except that 400 mg of tissue powder was loaded on 11 ml of 15–60% (w/v) sucrose gradients. A puromycin treatment abolished the typical UV polysome profiles and, as expected, dramatically reduced the RNA content in heavy fractions of the gradients, preventing reliable 3′ RACE analysis of puromycin controls.

Additional methods

Northern blot analyses, mRNA half-life measurements, microarray analysis and statistical analysis of microarray data are described in Supplementary Methods.

Nucleotide specificity of URT1

To characterize the nucleotidyltransferase activity of URT1, we produced recombinant proteins corresponding either to the wild-type protein or to a mutated version in which two conserved aspartate residues essential for catalysis of nucleotidyltransferases (D491 and D493 in URT1) were mutated to alanines. The nucleotidyltransferase activity of both proteins was tested using a 5′ [³²P]-labelled oligoribonucleotide substrate and conditions previously used to describe Cid1’s activity (25). No activity was observed for the mutant protein URT1^D491/3A, whereas URT1 showed a robust uridylation activity (Figure 1A). Under these assay conditions, URT1 synthesizes tails of >200 Us, while only 1–9 of the three other ribonucleotides are appended to the RNA substrate in 30 min. To further test that URT1 preferentially incorporates uridines, URT1 was incubated with the unlabelled oligoribonucleotide substrate and the four ribonucleotides ATP, GTP, CTP and UTP. The composition of tails synthesized by URT1 was then determined by sequence analysis of 12 cloned reaction products. URT1-synthesized tails contained 97.8% (947/967 nucleotides) uridines, confirming the strong uridine selectivity of URT1. In line with these in vitro results, the residues involved in uridine selectivity in Cid1 (D330, E333, H336) (28–30) are conserved in URT1 (D708, E711, H714). Interestingly, under the in vitro conditions used, URT1 better discriminates UTP from ATP as compared with Cid1 (Figure 1A).

Testing the stringency of URT1’s nucleotide specificity in vitro entailed the use of reaction conditions that maximize activity. As a result, large U-tails are polymerized, which unlikely reflect the in vivo situation. In fact, a restricted number of Us (1–10) are added to almost all RNA substrate molecules on short incubation times, indicating that the activity of URT1 is distributive for the first added nucleotides (Figure 1B). Following the addition of ∼10 uridines, URT1’s activity seems to become more processive. This switch from distributivity to processivity is not due to the short size of the initial oligoribonucleotide substrate (21 nt) because it was also observed for two RNA substrates of 31 nt (Supplementary Figure S2). The distributivity for the first added nucleotides appears therefore an inherent feature of URT1.

Taken together, these data show that Arabidopsis URT1 is a UTP:RNA uridylyltransferase with a marked preference for uridine polymerization and a distributive activity for the first added nucleotides. In line with this conclusion, the in vivo RNA substrates of URT1 are modified by the addition of a limited number of uridines as detailed below.

URT1 uridylates oligoadenylated mRNAs

Expressing fluorescent URT1 fusion proteins revealed a diffuse cytosolic pattern with foci (Figure 2A). These foci were observed in all cells examined. URT1-GFP and URT1-RFP partially co-localized with known markers of processing bodies (P-Bodies) and stress granules, DCP1-RFP or RBP47-GFP, respectively (31,32), suggesting that URT1 can associate with cytosolic ribonucleoprotein (RNP) granules (Figure 2A). These results also indicate that URT1 targets cytosolic RNAs. To identify such targets, we investigated mRNA uridylation by nested 3′ RACE PCR experiments using gene-specific forward primers and reverse primers complementary to a 3′ ligated RNA anchor sequence as detailed in (26). Initial tests on mRNAs selected on different criteria such as short half-lives or misregulation in urt1 mutants revealed that any mRNA investigated is a substrate of URT1. Detailed features of mRNA uridylation are presented for five mRNAs: LOM1 (At2g45160), DREB2C (At2g40340), BAM3 (At4g20270), At2g21560 and At5g48250. Of note, DREB2C and BAM3 are among the shortest-lived mRNAs in Arabidopsis (33) and At2g21560 and At5g48250 were identified among genes misregulated in urt1 (Supplementary Table S2). mRNA uridylation was first assessed in WT plants. Cloning and sequence analysis of 3′ RACE products revealed that 14% (6/43) of LOM1, 43% (26/60) of DREB2C, 36% (23/64) of BAM3, 57% (33/58) of At2g21560 and 18% (19/105) of At5g48250 clones corresponded to uridylated mRNAs (Figure 2B). Uridine extensions were short in WT, with 1 or 2 uridines detected for 92% (98/107) of the clones corresponding to uridylated mRNAs (Figure 2C). Overall, uridylated transcripts had significant shorter poly(A) tails than non-uridylated ones (Figure 2D). The average size of uridylated oligo(A) tails for the 5 mRNAs analysed in WT was 14 (n = 106). This size is similar to the length recently reported for oligo(A) tails of uridylated CCR2 mRNAs (22) and is in the range for the known oligo(A) size of deadenylated mRNAs in eukaryotes. Taken together, these data indicate that oligoadenylated mRNAs can be uridylated in Arabidopsis.

Next, we analysed the uridylation status of the five candidate mRNAs in urt1 mutants. Interestingly, we did not detect uridylated transcripts for LOM1 (0/55), At2g21560 (0/55) and At5g48250 (0/103) and only 2/72 DREB2C clones and 1/107 BAM3 clones corresponded to uridylated transcripts in urt1 (Figure 2B). URT1 is therefore the nucleotidyltransferase activity that is mainly responsible for uridylating mRNAs in Arabidopsis.

mRNA uridylation was also determined in a xrn4 mutant because the Arabidopsis 5′ to 3′ cytosolic exoribonuclease XRN4 (34) is a reasonable candidate for degrading uridylated mRNAs by analogy with data reported for humans and S. pombe (19,20). As in WT, uridylated mRNAs were detected in xrn4 for the five candidate mRNAs (Figure 2B), and uridylated mRNAs had significant shorter oligo(A) tails than non-uridylated ones in xrn4 (Figure 2D). However, both composition and length of the U tails differ between WT and xrn4 (Figure 2C). Whereas most U tails of WT mRNAs are 1-2 Us in length, the majority of oligo(U) tails (68/122) in xrn4 were at least 3 nucleotide long and could reach up to 11 nucleotides (Figure 2C). In addition, the oligo(U)-rich tails in xrn4 included up to 10% As (37/381), whereas the short tails in WT were almost exclusively composed of Us (180/181 nucleotides). The fact that both length and composition of U-tails are modified in absence of XRN4 suggests that uridylated mRNAs are substrates of XRN4.

Collectively, these data support the idea that URT1 uridylates mRNAs, that a deadenylation step precedes uridylation in Arabidopsis and that uridylated mRNAs are substrates of XRN4.

Uridylation marks uncapped, 5′ shortened LOM1 transcripts

Uridylation was reported to enhance decapping of mRNAs in S. pombe and A. nidulans and of non-polyadenylated histone mRNAs in humans (19–21). We therefore compared the proportion of uridylated transcripts between capped and uncapped transcripts by cRT-PCR using LOM1 as a model substrate. Capped LOM1 transcripts were detected in WT, urt1 and xrn4, whereas uncapped LOM1 mRNAs accumulated to detectable levels only in xrn4 (Figure 3A). This is in line with the idea that uncapped transcripts are degraded by XRN4. The 3′ extremities of both capped and uncapped transcripts in xrn4 mapped between positions +2074 to +2102 (Figure 3B). For the uncapped samples, 42 out of 70 clones were uridylated and only 2 out of 70 were not. For the rest of the clones (26/70), sequence analysis of cRT-PCR products cannot discriminate between uridines that could be either genomically encoded at the 5′ or added at the 3′ extremities. We favour the latter possibility because >90% of the uridylated LOM1 mRNAs had shortened poly(A) tails and half of them correspond to 5′ truncated transcripts (see below). Therefore up to 98% of the clones in the uncapped sample correspond to uridylated transcripts. By contrast, only 1 clone out of 73 corresponding to capped transcripts was uridylated. In addition, 95% (69/73) of 5′ extremities of capped transcripts map at or within 27-nt downstream of the known LOM1 5′-ends. By contrast, the 5′ extremities of 49% (34/70) of uncapped transcripts map within the coding sequence, indicating that a 5′ shortening occurs even in absence of the known cytosolic 5′ to 3′ exoribonuclease XRN4 (Figure 3B). Whether these 5′ shortened transcripts in xrn4 are generated by an endoribonuclease activity or by a cytosolic 5′ to 3′ exoribonuclease distinct from XRN4 is presently unknown. Interestingly, the uncapped, 5′ shortened transcripts that are detected in xrn4 mutants share the same 3′ extremities as capped transcripts. This indicates that even in xrn4 where 5′ to 3′ RNA degradation is compromised, uridylated LOM1 transcripts are still prone to 5′ to 3′ degradation rather than to 3′ attacks.

The cRT-PCR experiments also confirmed that uridylated LOM1 transcripts have a decreased poly(A) size as compared with non-uridylated capped transcripts (Figure 3C). In xrn4, the LOM1 poly(A) tail distribution had a median of 50 As for capped transcripts and 11 As for uridylated uncapped transcripts (P < 0.0001, Mann–Whitney test). The prevalence of uridylated transcripts in the uncapped fraction, the 5′ shortening of uridylated transcripts and the fact that uridylated transcripts have short poly(A) tails suggest that mRNA uridylation in Arabidopsis is linked to decapping and 5′ to 3′ degradation, as previously reported in humans, S. pombe and A. nidulans. However, it is difficult at present to estimate whether uridylation facilitates decapping in Arabidopsis because uridylation marks oligoadenylated mRNAs, and deadenylation alone can promote decapping. Importantly, none of the phenotypes linked with compromised decapping or 5′ to 3′ RNA degradation in Arabidopsis were observed in urt1 mutants, suggesting that decapping is not adversely compromised in absence of URT1. For instance, urt1 mutants do not show the ethylene-insensitive phenotype of xrn4 seedlings (35,36) (Supplementary Figure S3) and we did not observe abnormalities in embryo-to-seedling transition, a developmental step strongly affected by the loss of decapping activity (31,37).

Uridylation has no major impact on mRNA degradation rates

The genome-wide impact of URT1 on mRNA steady-state levels was determined by microarray analysis. One hundred sixty-four genes showed significant different expression levels between WT and urt1 (65 up, 99 down) (see Supplementary Table S2 for a list of misregulated genes and gene ontology analysis, and Supplementary Figure S4A for the validation of three misregulated genes in urt1). The number of differentially expressed genes is comparable with those observed for xrn4 (38,39). However, there is no significant overlap between urt1 and xrn4 data sets because only one gene (At5g49280) is misregulated both in urt1 and xrn4. To test a potential impact of URT1 on mRNA degradation rates, half-lives of five short-lived transcripts (i.e. half-lives ranging between 20 and 40 min) were determined following inhibition of transcription by cordycepin or actinomycin D. The degradation rates of the five mRNAs were not significantly affected regardless of their respective up-regulated, down-regulated or unchanged expression level in urt1 (Supplementary Figure S4B and C). Therefore, the variation of steady-state levels of these mRNAs in urt1 is rather due to indirect effects or multifactorial causes that remain to be identified. Although we cannot exclude that the mRNA half-lives determined following transcription inhibition by cordycepin or actinomycin D could not faithfully reproduce the actual in vivo half-lives, these data argue that URT1 has no major impact on mRNA degradation rates.

Oligoadenylated mRNAs are trimmed in urt1 mutants

A new role for mRNA uridylation in Arabidopsis was determined following the observation of a slight but systematic reduction in the size of the main 3′ RACE PCR products for 16 of the most up- or down-regulated mRNAs in urt1 as well as for the short-lived transcripts DREB2C and BAM3 or any other mRNAs investigated (Figures 4A and B, Supplementary Figure S1C and S5B). This size shift in urt1 is not only detected for all mRNAs tested but is also observed for the different developmental stages analysed: rosette leaves, seedlings and flowers (Figures 4A and B, Supplementary Figure S5B) showing the robustness of this molecular phenotype in urt1 mutants.

To understand the nature of the size shift, we cloned size-selected amplification products corresponding to the main band for DREB2C and BAM3. The distribution of poly(A) sites was similar between WT, urt1 and xrn4. However, the cloned poly(A) tails were shorter in urt1 (Mann–Whitney test P < 0.001) (Figure 4C). Enhanced deadenylation is therefore responsible for the size shift observed by 3′ RACE PCR in urt1. Because 3′ RACE PCR preferentially detects shorter over long poly(A) tails, we checked whether only oligo(A) tails rather than full-length poly(A) tails were shortened in urt1 mutants, a rationale hypothesis because URT1’s RNA substrates correspond to oligoadenylated mRNAs. To this end, we chose the At5g08040 mRNAs which, unlike DREB2C and BAM3, are expressed to sufficient levels to be detected by northern analysis and are well resolved on acrylamide gels owing to their short size of 371 nt [excluding poly(A) tails]. A northern analysis was performed using RNA samples from WT, urt1 and xrn4 flowers that were either treated with RNase H and oligo(dT) to remove poly(A) tails, or not treated with RNase H. This analysis revealed that similar long poly(A) tails of a maximal size of ∼100 nt are observed for At5g08040 mRNAs in the three genotypes (Supplementary Figure S5A). The slight size shift for RNA species with shorter tails is more difficult to observe on northern blots but is convincingly detected by 3′ RACE PCR in flowers, as well as in seedlings (Supplementary Figure S5B).

A second molecular phenotype was observed in urt1. As shown for several experiments for BAM3 and DREB2C, additional 3′ RACE PCR products with reduced sizes were occasionally amplified in urt1 mutants (Figure 4B). To understand the nature of these heterogeneous patterns, whole PCR reactions from two to three independent experiments were cloned. The corresponding sequence analysis is presented for LOM1, DREB2C, BAM3, At2g21560 and At5g48250 (Figure 5). Cloned 3′ RACE PCR products were split into three categories: (i) polyadenylated transcripts, (ii) uridylated transcripts and (iii) non-polyadenylated or deadenylated transcripts. Transcripts were considered deadenylated when the poly(A) tail length was <10, a size inferior to the minimal poly(A) tail length required for the binding of a poly(A) binding protein (40). Plotting the clone numbers against the position of 3′-ends on the respective mRNAs shows that multiple polyadenylation sites are detected for the five genes (Figure 5A). This observation is coherent with the intrinsic 3′ heterogeneity described for mRNA 3′-ends in Arabidopsis (41). The overall distribution of polyadenylation sites appeared unchanged in WT, urt1 and xrn4 (Figure 5A). Of note, the distribution of uridylated ends in WT and xrn4 coincides perfectly with the positions of polyadenylation sites (Figure 5A). In fact, all uridylated mRNAs detected in the course of this study in WT correspond to adenylated mRNAs [with a strong bias toward oligo(A) tails]. Therefore, although we do not formally exclude the possible uridylation of truncated and/or non-polyadenylated mRNAs in WT, our data indicate that oligoadenylated mRNAs are the main substrates of URT1. Importantly, the number and the diversity of clones corresponding to deadenylated or non-polyadenylated (from 0 to 10 As) transcripts were increased in urt1 as compared with WT and xrn4 (Figure 5A). Because clones having identical 3′-ends could correspond either to sister amplification or reflect the amplification of preferential degradation intermediates, we considered both the number of clones and the number of distinct 3′-ends detected for non/de-adenylated mRNAs. Overall, both the total number of clones and the diversity of the 3′-ends were similar in WT and xrn4 but markedly increased in urt1 (Figures 5B and C, respectively). For LOM1 and DREB2C, the 3′ extremities of deadenylated or non-polyadenylated transcripts covered the whole range of the sequence investigated by 3′ RACE PCR with some 3′-ends mapping even within the CDS (Figure 5A). This analysis revealed that the occasional shorter PCR products detected by 3′ RACE correspond to 3′ truncated transcripts and that the occurrence of 3′ truncated transcripts is increased in urt1.

Taken together, these results show that the major phenotype observed in urt1 mutants corresponds to a 3′ shortening of oligoadenylated mRNAs. Because this shortening was systematically observed and uridylation was detected for any mRNA investigated, we propose that mRNA uridylation is a frequent process in Arabidopsis and that uridylation prevents complete deadenylation of mRNAs. This is further supported by the increased occurrence of 3′ truncated transcripts in urt1 mutants that likely results from 3′ exoribonucleolytic attacks in the mRNA body once the mRNA has been fully deadenylated (see ‘Discussion’ section).

Uridylation can prevent 3′ trimming of polysomal mRNAs

The occurrence of co-translational mRNA decay was reported in S. cerevisiae (42) and recently, uridylation was shown to tag polysomal mRNAs in another fungus, A. nidulans (22). In this species, uridylation tags oligoadenylated transcripts in a similar way as for Arabidopsis mRNAs (22). We therefore tested whether uridylation could prevent the occurrence of 3′ shortened mRNAs associated with polysomes. Polysomes were isolated from seedlings, and typical UV profiles were observed for WT, urt1 and xrn4 samples, indicating no major effects of the urt1 mutation on polysome assembly (Figure 6A). RNA extracted from polysomal fractions was used to determine 3′ RACE PCR profiles for four transcripts, and the characteristic 3′ shortening was observed in urt1 (Figure 6B). Two experiments are shown for BAM3 and At5g48250 to show that the 3′ shortening is systematically observed in urt1 even though the PCR profiles may vary slightly. In addition, shorter amplicons corresponding to 3′ truncated transcripts were occasionally detected in urt1 polysomal fractions as shown for BAM3. Cloning and sequence analysis of BAM3 and At5g48250 samples revealed the presence of uridylated mRNAs in polysomal fractions in WT and xrn4 and confirmed that both the number of clones corresponding to non- or de-adenylated transcripts and the number of distinct 3′-ends are increased in urt1 (Figure 7). These results indicate that uridylation can prevent 3′ trimming of mRNAs associated with polysomes.