Mutational analysis of a viral RNA element that counteracts rapid RNA decay by interaction with the polyadenylate tail

Abstract

We previously demonstrated that the Kaposi's sarcoma-associated herpesvirus polyadenylated nuclear RNA contains a 79-nt cis-acting element, the ENE, which allows intronless polyadenylated transcripts to accumulate to high nuclear levels by protecting them from rapid degradation. We proposed a model based on the predicted structure of the ENE in which a U-rich internal loop hybridizes with the 3′-polyadenylate (polyA) tail to sequester it from exonucleolytic attack. We have tested this model by mutational analysis of the ENE. Point mutations in the predicted U-rich internal loop and in the flanking stems abolish the ENE's ability to (i) interact with the polyA tail, (ii) inhibit deadenylation in vitro, and (iii) stabilize transcripts in vivo. In all but one case, compensatory mutations in the flanking stems restore ENE activities, demonstrating the importance of these stems and uncovering a unique role for the loop-proximal G-C base pair in the lower stem. Increasing the U content of the U-rich internal loop surprisingly decreases stability in vivo but does not affect deadenylation in vitro, comparable to the effects of deleting certain “unstructured” regions of the ENE. Taken together, our data support the formation of the proposed ENE secondary structure in vivo and argue that the specific ENE structure inhibits rapid RNA decay in cis by engaging in a limited set of base-pairing interactions with the polyA tail.

RNA decay rates control the steady-state levels of an mRNA transcript, so an understanding of the determinants of RNA stability is essential to understanding gene expression. Specific cis-acting RNA elements can increase stability or accelerate decay (1, 2). A well studied example of regulated decay occurs with transcripts that contain an AU-rich element (ARE), which normally confers a short half-life but can stabilize the mRNA in response to environmental cues. RNA decay pathways also serve “quality control” functions (3), selectively degrading aberrant transcripts to ensure the fidelity of gene expression. Examples include nonsense-mediated, nonstop, and no-go decay pathways, whereby mRNAs are rapidly and selectively degraded when the translating ribosome encounters a premature termination codon, no termination codon or becomes translationally stalled, respectively (4–8). Surveillance mechanisms are not unique to mRNAs; noncoding and intergenic transcripts are also subject to quality control pathways (9–15).

In many cases, the first and rate-limiting step of mRNA decay is removal of the 3′ polyadenylate (polyA) tail (1, 2). In both yeast and mammalian cells, mRNA deadenylation generally precedes decapping and subsequent 5′ to 3′ and 3′ to 5′ exonucleolytic decay. Indeed, substrates undergoing both ARE-mediated decay and nonsense-mediated decay show increased deadenylation rates in vivo (16–18). Moreover, mRNAs targeted for destabilization by microRNAs have increased deadenylation rates (19, 20).

The Kaposi's sarcoma-associated herpesvirus polyadenylated nuclear (PAN) RNA (nut-1, T1.1) accumulates to unusually high levels (≈2.5 × 10⁵) in the nucleus of lytically infected human cells (21–23). Its high expression requires a 79-nt cis-acting element (called the ENE) that is also sufficient to increase the nuclear abundance of an otherwise poorly expressed β-globin cDNA transcript (24, 25). We have shown that the ENE decreases transcript decay rates both in vitro and in vivo (25). Specifically, regression analysis of in vivo decay data revealed two pools of PAN RNA: one that decays rapidly (t_1/2 ≈ 15 min) and another that decays more slowly (t_1/2 ≈ 3 h). The ENE renders transcripts less likely to be in the pool of rapidly decaying molecules. Inhibition of deadenylation by the ENE can be observed in vitro, both in nuclear extract and in a purified recombinant deadenylation system. Moreover, the ENE appears to interact with the polyA tail both in vivo and in vitro. We proposed a model wherein the U-rich internal loop of the 79-nt ENE hybridizes with the polyA tail, effectively sequestering it from exonucleases (Fig. 1A).

Fig. 1.

The U-rich internal loop is important for ENE activity. (A Left) Proposed secondary structure for the PAN ENE and its interaction with the polyA tail. Positions (in nucleotides) are relative to the PAN start site (23). The structure shown is the lowest energy structure predicted by Mfold (26). The gray dashed line represents the PAN sequence 3′ of the ENE. (A Right) Mutations tested in the U-rich loop. (B) Northern blot analysis of PAN WT and indicated mutant RNAs (Top) and ORF50 mRNA (Middle; a cotransfected control). The bar graph in Bottom shows average values from five or six experiments relative to WT. Error bars show standard deviation. (C) Representative data assessing the effects of mutations on the ability of the ENE to interact with the polyA tail of transcripts generated in vivo. Schematic depicts the MS2-PAN RNA, showing the positions of complementary oligonucleotides used for RNase H cleavage with arrows. Northern blots show signal from the 5′ fragment (Upper) and that of the corresponding coselected 3′ fragment (Lower). Lanes 1–8 contain 10% of the input, and lanes 9–16 are 100% of the pellets. The 3′ doublet is likely due to alternate RNase H cleavage sites. Each construct was tested at least three times with similar results. (D) In vitro deadenylation assay. Each data point is the average of five or six experiments. Graphs showing values for the 15- and 30-min time points with standard deviations are in SI Fig. 5.

Here we have tested this model using mutational analyses. Alterations in the U-rich loop that are predicted to decrease base pairing between the ENE and the polyA tail, as well as those predicted to disrupt stem structures flanking the U-rich loop, abolish ENE function. In contrast, an unstructured region at the 3′ end of the ENE is largely dispensable for activity. Deletion of unstructured 5′ and central regions of the ENE, as well as mutations that alter the U-rich loop without decreasing its pairing potential with polyA, show intermediate effects that are more apparent in vivo than in vitro. These data confirm the existence of an essential U-rich internal loop and flanking stems in the ENE structure in vivo, as predicted by computer algorithms (26). They also suggest that other domains of the ENE are important for augmentation of ENE activity in vivo.

Results

The U-Rich Loop Is Essential for ENE Activity.

Our model predicts that the U-rich internal loop of the ENE structure enables base pairing to the polyA tail, thereby protecting it from deadenylases. To examine this postulate, we generated constructs with U to C mutations at positions 903 and 949 (in the middle of U₅ stretches) either alone or in combination (Fig. 1A). We also made constructs that increased the U content of the loop by changing the A at position 952 to U (A952U) or by inserting three U's (903+UUU). We examined the effects of these mutations on ENE activity by monitoring steady-state PAN RNA levels in vivo (Fig. 1B), by assessing interactions between the ENE and the polyA tail (Fig. 1C), and by testing the ability of the ENE to inhibit deadenylation in nuclear extract (Fig. 1D).

All alterations of the U-rich loop significantly reduce steady-state PAN RNA levels in transiently transfected HEK293 cells (Fig. 1B). Strikingly, single or double point mutations of U to C decrease PAN RNA abundance (≈3.2- to 4.8-fold; Fig. 1B, lanes 3–5) nearly as much as a deletion of the entire 79-nt ENE (Δ79; ≈7.1-fold; Fig. 1B, lane 2). Surprisingly, increasing the U content of the internal loop with the A952U and 903+UUU also decreases steady-state RNA levels, although to a lesser extent (≈1.7- and 2.3-fold, respectively; Fig. 1B, lanes 6 and 7). Thus, any perturbation of the U-rich loop is deleterious, arguing for its importance for ENE activity.

Next we examined the ENE's ability to interact with the polyA tail of its transcript using a previously described coselection procedure (25). We inserted six binding sites for bacteriophage MS2 coat protein near the 5′ end of PAN expression constructs. After transient expression in HEK293 cells, we gently lysed the cells, cleaved PAN RNA 3′ of the ENE by oligonucleotide-directed RNase H (Fig. 1C; NC289), precipitated the 5′ fragment using recombinant GST-MS2 protein, and asked whether the 3′ fragment was coselected. Subsequent to pull-down, an additional RNase H cleavage step (with NC244 and NC392) allowed detection of the 5′ (Fig. 1C Upper) and 3′ (Fig. 1C Lower) fragments on the same gel. As previously reported, the 3′ end of PAN is coselected with the 5′ fragment in the presence of the unmutated ENE (WT), but deletion of the ENE abrogates the interaction (Fig. 1C, compare lanes 9 and 10). As expected, precipitation of both the 5′ and 3′ fragments depends on the presence of the MS2-binding sites (Fig. 1C, lane 16). The U to C mutations abolish the interaction between the 3′ and 5′ ends of PAN RNA (Fig. 1C, lanes 11–13), but neither the A952U nor the 903+UUU mutations have a major impact using this assay.

We previously observed that the ENE decreases deadenylation rates in HeLa nuclear extract (25). To examine the anti-deadenylation activity of ENE mutants we adapted a quantitative deadenylation assay (27, 28) for use in nuclear extract. Substrates containing the 3′-most 327 nt of PAN (25) with a 15-nt polyA tail were uniformly radiolabeled at A residues and incubated in nuclear extract under deadenylation conditions. Liberation of AMP was monitored by using TLC [Fig. 1D and supporting information (SI) Fig. 5]. We observe that both the single and double U to C mutants, but not the A952U or the 903+UUU mutations, decrease the ENE's ability to inhibit deadenylation (Fig. 1D).

To summarize, decreasing the U content of the central loop significantly diminishes ENE activity in all three assays. In contrast, mutations predicted to increase the U content of the U-rich loop affect steady-state levels (albeit to a lesser degree) without affecting interactions with the polyA tail or decreasing deadenylation in nuclear extract, suggesting that a particular structure of the ENE and/or ENE-binding proteins contributes to its activity in vivo.

The Upper Stem of the Predicted Structure Is Essential for ENE Activity.

The predicted structure of the ENE contains a 6-bp upper stem closed by two C-G base pairs that form between positions 906 and 946 and positions 907 and 945 (Fig. 2A). To determine their importance for ENE activity, we mutated these bases to their complements (CC906GG and GG945CC) and then combined the mutations (CC906GG//GG945CC; CC “double arrow” GG in Fig. 2). We reasoned that if the function of these bases is simply to contribute to stem formation, the former mutations will abrogate ENE activity and the compensatory mutations should restore it.

Fig. 2.

The upper stem is essential for ENE activity, whereas other regions are partially dispensable. (A) Proposed ENE structure with mutations in the upper stem and positions of deletions indicated. (B–D) Steady-state (B), coselection (C), and in vitro deadenylation (D) analyses were as described in Fig. 1 B–D.

We examined these mutants using the same three assays as above (Fig. 2). Both the CC to GG and the GG to CC mutations significantly decrease steady-state PAN RNA levels, but combining the mutations restores PAN RNA to ≈80% of the abundance of WT (Fig. 2B, lanes 1–4). Similarly, each double mutation destroys interaction between the ENE and the polyA tail (Fig. 2C, lanes 11 and 12), whereas combining the mutations restores the pull-down (Fig. 2C, lane 13). Finally, CC906GG and GG945CC mutations both diminish the ability of the ENE to inhibit deadenylation in extract, but the deadenylation of a substrate carrying the compensatory mutations is indistinguishable from WT (Fig. 2D). Together, these data strongly support the conclusions that the C-G base pairs of the upper stem form both in vivo and in vitro and are essential for ENE activity.

5′ and Central Unstructured Regions of the ENE Contribute to Steady-State PAN RNA Levels But Do Not Affect Activity in Vitro.

We also tested other portions of the predicted ENE structure by deleting the 5′- and 3′-most 7 nt of the ENE, as well as a central 31-nt region (Δ888–894, Δ960–966, and Δ911–941, respectively) (Fig. 2A). Although it is depicted as a stem, the latter sequence can fold into several distinct structures with similar ΔG values (25). We replaced the sequence with a GCAA tetraloop to avoid disruption of the predicted 6-bp upper stem.

Steady-state analyses show that Δ911–941 and Δ888–894 decrease steady-state PAN levels by ≈2.8- and 2.2-fold, respectively, whereas the Δ960–966 deletion accumulates to ≈75% of WT (Fig. 2B, lanes 5–7). All three deletions retain interaction with the polyA tail (Fig. 2C, lanes 14–16) and inhibition of deadenylation in vitro (Fig. 2D). We conclude that nucleotides 960–966 are not a part of the core ENE. In contrast, although the Δ888–894 and Δ911–941 deletions have little or no effect on ENE–polyA tail interactions or on inhibition of deadenylation in vitro, they do decrease steady-state RNA levels, demonstrating a requirement for these regions in vivo.

The Lower Stem Is Essential for ENE Activity, and the Loop-Proximal G-C Pair May Contribute to Tertiary or Protein Interactions.

The ENE is predicted to contain a 6-bp lower stem closed by three G-C base pairs at positions 898–900 and 954–956 (Fig. 3A). To ask whether this duplex is important for ENE activity, we first mutated all three G's to C's (GGG898CCC) and vice versa (CCC954GGG) and then combined these changes to test for compensatory activity (GGG898CCC//CCC954GGG; abbreviated GGG “double-arrow” CCC in Fig. 3).

Fig. 3.

The lower stem is essential for ENE activity. (A) Proposed ENE structure with the lower stem mutations indicated. (B–D) Steady-state (B), coselection (C), and in vitro deadenylation (D) analyses were as described in Fig. 1 B–D.

Northern blot analysis shows that, although the GGG898CCC and the CCC954GGG mutations decrease PAN RNA abundance ≈5.5-fold (Fig. 3B, lanes 2 and 3), combining the mutations does not compensate (≈3.8-fold reduction) (Fig. 3B, lane 4). Individual G to C (898–900) or C to G (954–956) mutations at these positions also decrease steady-state levels (Fig. 3B, lanes 5–10). We then swapped each G-C base pair individually and found that the pairs farthest from the U-rich loop (G898C//C956G, G899C//C955G) can tolerate G-C to C-G conversion (Fig. 3B, lanes 11 and 12). In contrast, steady-state PAN RNA levels remain low (≈2.9-fold reduced) when the predicted G-C base pair at the top of the stem is changed to C-G (G900C//C954G) (Fig. 3B, lane 13).

Mutation of all three G's or C's in the lower stem or the complete reversal of the G-C to C-G base pairs significantly impairs the ENE–polyA tail interaction (Fig. 3C, lanes 13–15). Likewise, G898C and G899C single mutants do not support this interaction (Fig. 3C, lanes 19–20). Swapping G-C for C-G base pairs restores the interaction for the G898C//C956G and G899C//C955G mutants, but not for the G900C//C954G mutant (Fig. 3C, compare lanes 16–18). Thus, the ability of the ENE carrying mutations in the lower G-C stem to interact with its polyA tail correlates with the steady-state levels of these mutant RNAs.

Finally, we examined how lower stem alterations affect the inhibition of deadenylation (Fig. 3D). As observed for the other ENE activities, swapping G-C base pairs for C-G base pairs maintains WT activity for G898C//C956G and G899C//C955G but not for G900C//C954G. Single mutations in the lower stem also abrogate inhibition of deadenylation in nuclear extract (SI Fig. 5). Both G900C//C954G and GGG898CCC//CCC954GGG show an intermediate level of activity compared with WT and Δ79.

We conclude that the lower stem, or at least the G898-C956 and G899-C955 base pairs, forms in vivo and that pairing of these positions is essential for all ENE activities tested. In contrast, substituting a C-G base pair for G-C immediately adjacent to the U-rich loop does not support ENE activity, demonstrating that one or both of these residues contributes in some way other than simple base pairing, perhaps in base-specific tertiary interactions or in protein binding.

ENE Mutations Alter in vivo RNA Decay Profiles.

The ENE increases the stability of PAN RNA in vivo (25). We used a transcription pulse strategy (29) to determine the effect of the Δ79, U903C, and A952U mutations on PAN half-life in vivo (Fig. 4; representative data are in SI Fig. 6). Each of the mutant transcripts generated in a 3.5-h transcription pulse was less stable than WT in vivo, apparent from the difference in decay behavior at the earliest time points (Fig. 4A, 15 and 30 min); no statistically significant difference between the mutants was evident.

Fig. 4.

ENE mutations decrease transcript stability in vivo. (A) Regression analyses of data from 3.5-h transcription pulse experiments, with the mutations identified in the keys. Each point is the average of three to five experiments with error bars representing standard deviation. The data are fit to a two-component exponential decay curve. Rate constants and other decay parameters are given in SI Table 2. (B) Same as A but with an ≈19-h transcription pulse. (C) Graphs of data from in vivo decay analysis from 3.5-h transcription pulses, with the mutations indicated in the key. Each data point is the average of three to five experiments with error bars showing standard deviation. Graphs are linear splines, not regressions. (D) Same as C but with an ≈19-h transcription pulse.

This equivalent impairment of activity in vivo was somewhat surprising in the context of the data presented above. The A952U mutation decreases steady-state RNA levels, but significantly less than the U903C mutation. In addition, A952U shows no loss of interaction with the polyA tail or of deadenylation inhibition in vitro (Fig. 1). One possible explanation is that the A952U mutant has intermediate stability, but the experimental conditions do not allow us to discern this difference in vivo.

Our previous experiments demonstrated that longer pulse times enhance the difference between the decay profiles of WT PAN and ENE-lacking transcripts because of the accumulation of ENE-containing transcripts subject to a slower decay pathway (25) (Fig. 4B). After an ≈19-h pulse, the A952U transcripts are slightly less stable than WT, but clearly more stable than transcripts with either the Δ79 or U903C mutation. Thus, both the U903C and A952U mutations affect the decay profiles in vivo, but differentially. The stability of the U903C mutant transcripts closely resembles that of the complete ENE deletion when assayed after either a short or long pulse. In contrast, the A952U mutant behaves similarly to the ENE deletion transcript after short pulses but is significantly more stable than the ENE deletion mutant after long pulses.

Finally, we extended the in vivo decay analysis to other ENE mutations. We chose three mutants that, like U903C, have significantly reduced ENE activity in the assays described above (CC906GG, GGG898CCC, and GGG898CCC//CCC954GGG) and three that decrease steady-state levels but otherwise have minimal effect on ENE activities (903+UUU, Δ911–941, and Δ888–894). To expedite analysis, we collected data only for the initial time points subsequent to transcription shutoff. All six mutants exhibit significantly more rapid decay than WT when examined after a 3.5-h pulse (Fig. 4C). Indeed, none of the tested alterations are significantly more stable than Δ79 under these conditions. Similarly, after ≈19-h pulses, the decay of the CC906GG, GGG898CCC, and GGG898CCC//CCC954GGG transcripts resembles that of Δ79 (Fig. 4D). In contrast, the 903+UUU, Δ911–941, and Δ888–894 mutants are all more stable in vivo than the ENE deletion transcript but less stable than PAN WT.

In conclusion, our ENE mutations fall into two categories. Those that decrease the ENE's ability to interact with the polyA tail and to inhibit in vitro deadenylation have in vivo decay profiles similar to those of PAN transcripts completely lacking the ENE. In contrast, ENE mutations that reduce steady-state RNA levels but maintain the ENE's ability to interact with the polyA tail and to inhibit deadenylation in vitro are more stable than Δ79 transcripts but somewhat less stable than WT in vivo.

Discussion

Our previous studies had suggested that the ENE of the Kaposi's sarcoma-associated herpesvirus PAN RNA blocks an important mammalian deadenylation-dependent RNA decay pathway through interactions with the polyA tail in cis. Here we have examined the effects of mutations designed to alter predicted structural features of the ENE, including the U-rich internal loop, the upper and lower flanking stems, and the unstructured central, 5′ and 3′ regions. Mutated transcripts were assessed for their ability to (i) maintain high steady-state PAN RNA levels, (ii) promote interactions between the ENE and the polyA tail, (iii) inhibit deadenylation in nuclear extract, and (iv) increase the stability of PAN RNA in vivo. The results support the existence and significance of the ENE structure shown in Fig. 1 and identify additional features that enhance ENE activity in vivo.

The model proposes that the U-rich internal loop would be essential for ENE activity because it drives interaction with the polyA tail by engaging in A-U base pairing (Fig. 1A). Point mutations predicted to disrupt either of the two hypothesized 5-bp A-U helical stretches significantly decrease ENE activities in all assays. In fact, in every assay, the single mutations are only slightly less deleterious than ENE deletion (Figs. 1 and 4). Two alterations predicted to increase the number of base pairs between the polyA tail and ENE (A952U and 903+UUU) might have been expected to improve ENE function. However, they negatively affected ENE activity in vivo (Fig. 4) while maintaining interactions with the polyA tail and inhibition of deadenylation in nuclear extract (Fig. 1). Therefore, the size or symmetry of the U-rich loop appears important for complete ENE activity in vivo, suggesting that these features promote RNA folding or recruit ENE-binding proteins that augment function. Indeed, proteins may be critical for ENE activity in vivo. The ENE can act in the absence of protein in vitro, but activity requires careful folding of substrates; in contrast, no folding step is necessary for activity in nuclear extract (25).

The predicted upper and lower stems of the ENE are also essential (Figs. 2–4). Disruption of the G-C base pairs in the upper stem abrogates all ENE activities. However, because nearly full rescue was achieved by replacing G-C with C-G pairs, we conclude that this stem is purely a secondary structure feature. Similar compensatory changes support the existence of the lower ENE stem, but surprisingly the loop-proximal pair (G900–C954) could not be exchanged without loss of ENE activity. One or both of these positions may be important for base-specific binding of protein(s) or for tertiary interactions, such as A-minor interactions, which are known to be sensitive to a switch in base pair composition (30). Unfortunately, the absence of PAN in related herpesviruses precludes obtaining phylogenetic evidence for the existence of the ENE stems. Our results demonstrate that both flanking stems are necessary for ENE activity, but we cannot distinguish whether they are important for defining the U-rich loop, for constricting it to the appropriate size, for protein binding, or for tertiary interactions.

We tested deletions in three regions of the ENE predicted not to participate in well defined secondary structure formation (Δ911–941, Δ888–894, and Δ960–966). Because the 3′-most 7 nt could be deleted without affecting the ability of the ENE to interact with the polyA tail, to inhibit deadenylation, or to maintain PAN RNA levels at ≈75% of those of WT, we conclude that the core ENE is 72 rather than 79 nt long. Deletion of both the 5′ and central regions decreases steady-state RNA levels but has little effect on the in vitro deadenylation and polyA tail interaction assays. Therefore, like the 903+UUU and A952U alterations, these regions may contribute to proper folding of the ENE in vivo or to protein binding.

The correlation we observe for the behavior of ENE mutants in in vitro deadenylation and the polyA tail coselection assays further supports our model that the ENE inhibits deadenylation in vivo by interacting with the polyA tail. In all cases where a mutant fails to inhibit deadenylation in extract and to interact with the polyA tail, steady-state RNA levels decrease >3-fold (Figs. 1–3) and the transcript is markedly destabilized in vivo (Fig. 4). Thus, loss of interaction with the polyA tail and loss of inhibition of deadenylation are good indicators of loss of ENE function in vivo.

Some of the mutations tested did not decrease the ability of the ENE to inhibit deadenylation in vitro or to interact with the polyA tail but still affected steady-state transcript levels (Figs. 1 and 2; A952U, 903+UUU, Δ888–894, and Δ911–941). One explanation for this discrepancy is that interaction of the ENE with the polyA tail is both necessary and sufficient for activity in vitro but not sufficient in vivo. Perhaps ongoing transcription, protein assembly, and 3′-end formation of the nascent transcript place more stringent requirements on ENE function. An alternative interpretation is that interaction with the polyA tail is indeed sufficient for mutant ENE activity in vivo, but that a smaller fraction of these mutant transcripts form the interaction in vivo. Thus, only the portion (smaller than with WT but larger than in the more severe ENE mutants) that manages to establish interactions with the polyA tail accumulates, such that formation of the ENE–polyA tail interaction becomes limiting, not the stabilizing effect of the interaction once formed. Because these mutants are as effective as WT in preventing in vitro deadenylation (Figs. 1 and 2), this model presupposes a difference between the in vivo assembly of an ENE–polyA tail interaction and that necessary for inhibition of deadenylation in vitro. Regardless, the mutant data, combined with our previous observations (25), indicate that interactions between the ENE and polyA tail are necessary for ENE-mediated escape from rapid decay.

Control of decay through interactions with the polyA tail is unlikely unique to PAN RNA. Muhlrad and Parker (31) proposed a similar mechanism for the yeast EDC1 mRNA where a long U-rich (36/38 nt U) element present in the 3′ UTR appears both necessary and sufficient to block cytoplasmic deadenylation. In mammalian cells, a 28-nt stretch of U's does not substitute for ENE function in vivo (25), consistent with the structural requirements confirmed in this study. Instead, the ENE appears to fold into a stem-loop structure similar to that of an H/ACA small nucleolar RNA (32, 33), with the U-rich central loop acting as the guide sequence and the polyA tail mimicking the rRNA substrate. Similarities include the ≈10 nt of base pairing that is continuous on the rRNA target but disrupted on the small nucleolar RNA by a stem structure. Additionally, both small nucleolar RNA–rRNA and ENE–polyA interactions can be observed in the absence of protein but are likely to exist in a protein-bound state in vivo. The data reported here provide a foundation for future structural and functional studies of the ENE and will inform bioinformatic attempts to identify candidate cellular ENE-like elements.

Materials and Methods

Cell Culture, RNA Procedures, and Plasmid Construction.

Cell culture, media, and RNA procedures were as described (25). Oligonucleotide and PCR primer sequences are in SI Table 1. Plasmid construction is described in SI Materials and Methods.

GST-MS2 Coselection Assays.

These assays were performed as previously described (25). To equalize expression levels between samples, ≈1.5- to 3-fold more of the mutant plasmids were transfected. For Northern detection of the 5′ and 3′ fragments, kinased oligonucleotides NC313 and NC301 were used as probes.

In Vitro Deadenylation Assays.

Deadenylation reactions were performed in HeLa nuclear extract essentially as previously described (25, 34), but with some modifications. The m⁷GpppG capped substrates had a 15-nt polyA tail and were uniformly radiolabeled with [α-P³²]ATP to a specific activity of ≈50 Ci/mmol; ≈150 fmol of substrate was used in a 10-μl reaction. During the course of the reaction, 0.5-μl aliquots were removed at given times and spotted directly on Cellulose 300 polyethyleneimine TLC plates (Sorbent Technologies, Atlanta, GA). Before spotting, the TLC plates were prerun in distilled deionized water and dried. Samples were fractionated by using 0.75 M KH₂PO₄ (pH 3.5) as a solvent and dried; signal was detected by PhosphorImager analysis. To normalize experiments, the 60-min WT value was set to 1.

In vivo Decay Assays.

The experiments were modified from ref. 25. For each sample, 12 μg of plasmid was mixed with 2.8 ml of Opti-MEM (Invitrogen, Carlsbad, CA) and 42 μl of Lipofectamine 2000 (Invitrogen) and incubated at room temperature for 20 min. Twenty milliliters of medium (DMEM) with 10% Tet-system approved FBS (BD Biosciences, San Jose, CA) containing 1.15 × 10⁷ HeLa Tet-off cells was added to the transfection mix, and 2 ml was plated per well of a six-well culture dish. For 3.5-h pulse experiments, 2 ng/ml doxycycline was included in the media and, ≈18–20 h after transfection, cells were washed twice with 1× PBS and placed in media lacking doxycycline for 3.5 h before doxycycline addition to 50 ng/ml. For long pulse experiments, 50 ng/ml doxycycline was added 18–20 h after transfection. At appropriate time points, medium was aspirated and 1 ml of TRIzol (Invitrogen) was added to the cells. Data analysis was performed as in ref. 25 except that 7SK RNA served as a loading control (data not shown).

The values determined for the fraction of transcripts undergoing rapid decay are significantly higher than our previous work [≈38% of PAN RNA in the rapid decay pathway after a 3.5-h pulse (25) as compared with the 58.8% reported here]. Changes in the transfection procedure, in the Tet-responsive constructs, or in the HeLa Tet-off cells may be responsible for the discrepancy.

Abbreviations

PAN

polyadenylated nuclear

polyA

polyadenylate.