U2 small nuclear RNA is a substrate for the CCA-adding enzyme (tRNA nucleotidyltransferase).

The CCA-adding enzyme builds and repairs the 3' terminus of tRNA. Approximately 65% of mature human U2 small nuclear RNA (snRNA) ends in 3'-terminal CCA, as do all mature tRNAs; the other 35% ends in 3' CC or possibly 3' C. The 3'-terminal A of U2 snRNA cannot be encoded because the 3' end of the U2 snRNA coding region is CC/CC, where the slash indicates the last encoded nucleotide. The first detectable U2 snRNA precursor contains 10-16 extra 3' nucleotides that are removed by one or more 3' exonucleases. Thus, if 3' exonuclease activity removes the encoded 3' CC during U2 snRNA maturation, as appears to be the case in vitro, the cell may need to build or rebuild the 3'-terminal A, CA, or CCA of U2 snRNA. We asked whether homologous and heterologous class I and class II CCA-adding enzymes could add 3'-terminal A, CA, or CCA to human U2 snRNA lacking 3'-terminal A, CA, or CCA. The naked U2 snRNAs were good substrates for the human CCA-adding enzyme but were inactive with the Escherichia coli enzyme; activity was also observed on native U2 snRNPs. We suggest that the 3' stem/loop of U2 snRNA resembles a tRNA minihelix, the smallest efficient substrate for class I and II CCA-adding enzymes, and that CCA addition to U2 snRNA may take place in vivo after snRNP assembly has begun.

The CCA-adding enzyme (ATP(CTP):tRNA nucleotidyltransferase) synthesizes and regenerates the 3Ј-terminal CCA sequence of tRNA by adding three consecutive nucleotides in the order of C, C, and A (1-3). The CCA-adding enzyme is the only polymerase known that can synthesize a defined sequence without using a nucleic acid template (4,5). The enzyme is essential in eukaryotes, archaea, and some eubacteria, where many or all tRNA genes lack encoded CCA (6,7). In these organisms, 3Ј trailer sequences are removed from tRNA precursors by nucleases that stop at the discriminator base (position 73), leaving the acceptor stem intact as a substrate for the CCA-adding enzyme (3,6,7). In Escherichia coli and presumably in other eubacteria in which all tRNA genes encode CCA, the CCA-adding enzyme is not essential but is advantageous (8) because it can repair CCA termini depleted by exonuclease activity (9,10). The ability of CCA-adding enzymes to recognize all cytoplasmic tRNAs regardless of amino acid acceptor specificity (11) and the ability of synthetic tDNAs to serve as substrates for the E. coli CCA-adding enzyme (5,12) suggest that recognition involves structural features common to all but some unusual organellar tRNAs (13).
U2 small nuclear RNA (snRNA) 1 is a small, highly conserved, nonpolyadenylated nuclear RNA that plays an essential role in mRNA splicing (14,15). In vertebrates, the genes for all snRNAs of the Sm class (snRNAs that bind Sm antigens) and some small nucleolar RNAs of the box CϩD class such as U3 (16) share a common transcriptional apparatus: transcription is driven by a specialized RNA polymerase II promoter, consisting of an enhancer-like distal sequence element and a TATA-like proximal sequence element that are spaced almost precisely one nucleosome apart (17) and appear to be brought together by a bound nucleosome (18,19). A highly conserved 3Ј end formation signal or "3Ј box" (20 -22) located some 6 -25 nucleotides beyond the snRNA coding region directs formation of the first detectable U snRNA precursors (U2ϩ10) containing 10 -16 extra nucleotides and ending just upstream of the 3Ј box. However, the 3Ј box functions only when transcription is driven by a specialized U snRNA promoter; mRNA promoters cannot substitute for the U snRNA promoter (17,23,24).
Although U2 (but not U1) transcription appears to continue past the 3Ј box (25), it remains to be seen whether the 3Ј box functions as a transcription termination signal, as observed for rRNA transcription by RNA polymerase I (26), or as an RNA processing signal, like the AAUAAA element of polyadenylated mRNAs (27)(28)(29) and the CAGA box of nonpolyadenylated replicative histone mRNAs (30). U2ϩ10 is subsequently exported to the cytoplasm, where the 3Ј tail is trimmed (31)(32)(33)(34) by endonuclease and exonuclease activities that could be related to the yeast exosome, a multinuclease complex involved in processing many cellular RNAs (35,36). Sm antigens then assemble onto the Sm binding site, and the resulting immature U2 snRNP is imported into the nucleus for final assembly into a mature U2 snRNP (35,(37)(38)(39). Addition of U2 snRNP-specific antigens may occur in Cajal bodies (also known as coiled bodies or CBs) (40 -42), whereas nucleotide modifications (methylation and pseudouridylation) occur in the nucleolus (43).
Approximately 65% of mature human U2 snRNA ends in 3Ј-terminal CCA, as do all mature tRNAs; the other 35% ends in 3Ј CC or possibly 3Ј C, as these are not easily distinguished experimentally (44,45). The 3Ј-terminal A of U2 snRNA must be added posttranscriptionally because the U2 snRNA coding region ends with CC/CC, where the slash indicates the last encoded nucleotide (45)(46)(47). If 3Ј exonucleolytic processing removes the encoded 3Ј CC during U2 snRNA maturation, as appears to be the case in vitro (33,34,38), then the cell may need to build or rebuild the 3Ј-terminal A, CA, or CCA of U2 snRNA. Using a variety of synthetic and natural U2 snRNAs as * The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
¶ To whom correspondence should be addressed. substrate, we show here that the human but not the E. coli CCA-adding enzyme can build and repair the 3Ј-terminal CCA sequence of human U2 snRNA. The human CCA-adding enzyme is also active on native U2 snRNPs. The yeast (48) and vertebrate CCA-adding enzymes (49,50) are known to be present in the nucleus as well as the cytoplasm, consistent with the possibility that the human CCA-adding enzyme may build the 3Ј end of immature U2 snRNPs in the cytoplasm or nucleus or repair the 3Ј end of mature nuclear U2 snRNPs.

EXPERIMENTAL PROCEDURES
In Vitro Transcription and Purification of U2 snRNA Substrates-To generate templates for in vitro transcription of U2 snRNAs lacking 3Ј-terminal A, CA, and CCA, we performed PCR using plasmid pU2-T7 as template (51) and appropriate combinations of 5Ј and 3Ј primers. The three different 5Ј primers (T7U2-1, T7U2-111, T7U2-147) each contained the T7 RNA polymerase promoter sequence (underlined below) just upstream from the U2 snRNA sequence; T7U2-1 generated a template for in vitro transcription of full-length U2 snRNA, T7U2-111 generated a template for in vitro transcription of 3Ј-terminal stem/loops III and IV, and T7U2-147 generated a template for in vitro transcription of 3Ј-terminal stem/loop IV. The six different 3Ј primers (U2-N, U2-NC, U2-NCC, U2-NCCA, U2-NCU, and U2-NCUA, where N is the discriminator base) each contained a FokI recognition site (underlined below) downstream of the U2 snRNA sequence; digestion with FokI generated a template with the appropriate end (indicated below by a slash) for runoff transcription. The in vitro transcription reactions were carried out for 3 h at 37°C in 40 mM Tris-HCl, pH 8.0 (20°C), 6 mM MgCl 2 , 10 mM dithiothreitol, 2 mM spermidine, 1 mM each nucleotide triphosphates, 40 units of T7 RNA polymerase (Roche), and 500 g/ml DNA template. U2 snRNA products were purified on denaturing 10%, 12%, or 15% polyacrylamide gels containing 8 M urea; the single U2 snRNA band was visualized by UV shadowing, excised, eluted, and concentrated by ethanol precipitation with glycogen carrier (Roche). Uniformly labeled U2 snRNAs were transcribed in the presence of 50 Ci of [␣- 32 ). B, CCA addition in vitro to full-length U2 snRNA substrates lacking 3Ј-terminal CCA, CA, or A (A 185 , A 185 C, and A 185 CC, where A 185 corresponds to the discriminator base); the natural m 3 2,2,7 Gppp cap is replaced by 5Ј pppG in the T7 transcript. Asterisks indicate the labeled nucleotide. Reactions were resolved by denaturing 6% PAGE; the length of the product is indicated in parentheses. zyme was expressed as described previously (13) with a slight modification. E. coli BL21-codonPlus(DE3)-RIL (Stratagene) was transformed by the overexpression plasmid (13), inoculated into LB broth at 37°C in the presence of 30 g/ml kanamycin and 34 g/ml chloramphenicol, and grown to an A 600 of 0.8. Isopropyl-1-thio-␤-D-galactopyranoside (0.1 mM) was added to induce expression, and growth continued at room temperature for 3 h. The bacterial pellet was lysed by sonication in buffer A containing 20 mM Tris-HCl, pH 7.8, 10 mM MgCl 2 , 500 mM KCl, 6 mM ␤-mercaptoethanol, 0.1 mM phenylmethylsulfonyl fluoride, and 5% glycerol. The cleared lysate was loaded onto a nickel-NTA resin (Qiagen) column, and the column was washed extensively with buffer A plus 20 mM imidazole. Histidine-tagged protein was eluted with buffer A plus 150 mM imidazole and dialyzed against buffer A containing 0.25 M KCl and 10% glycerol with or without 0.1 mM phenylmethylsulfonyl fluoride.
The CCA-adding enzyme genes of the other archaea (Methanobacterium thermoautotrophicum, Methoanococcus janaschii, and Pyrococcus furiosus) were amplified from genomic DNA by PCR. The wild type proteins are 454 (M. thermoautotrophicum), 449 (M. jannaschii), and 453 residues long (P. furiosus). The M. thermoautotrophicum gene was inserted between the NdeI and HindIII sites of pET22b(ϩ) (Novagen); the M. jannaschii and P. furiosus genes were inserted between the NdeI and XhoI sites. The initiator ATG of the M. thermoautotrophicum and P. furiosus enzymes was retained, but the M. jannaschii initiator GTG encoding valine was changed to ATG. All three enzymes have additional C-terminal residues including a hexahistidine tag: the M. thermoautotrophicum TGA stop codon was changed to AAG (lysine) to generate the C-terminal sequence LAAALEH 6 ; the M. jannaschii and P. furiosus stop codons (TGA and TAA, respectively) were changed to CTC (leucine) to generate the C-terminal sequence LEH 6 . These constructs were used to transform E. coli BL21(DE3) carrying an argU plasmid encoding a minor arginine tRNA. Transformants were inoculated into LB broth containing 100 g/ml ampicillin and 30 g/ml kanamycin, proteins were induced with 0.1 mM isopropyl-1-thio-␤-D-galactopyranoside, and growth continued at 37°C for 12 h. Histidine-tagged proteins were purified as described above, and stored at Ϫ20°C in buffer A containing 100 mM KCl and 50% glycerol.
Native U2 snRNPs-CCA addition to native U2 snRNPs was assayed by incubating 5 l of HeLa nuclear extract (Promega; 18.2 mg/ml protein) with the human CCA-adding enzyme (50 ng) for 15 min at 37°C in 100 mM glycine/NaOH (pH 9.0), 10 mM MgCl 2 , 1 mM dithiothreitol, 500 M CTP, 2 mM ATP, and 0.1 M [␣-32 P]CTP or [␣-32 P]ATP (3,000 Ci/mmol; Amersham Biosciences, Inc.) in 50-l total reaction volume. After phenol extraction and ethanol precipitation, RNAs were fractionated by denaturing 6% PAGE, and the gel was subjected to autoradiography. To reveal endogenous 3Ј exonucleases capable of acting on U2 snRNP, the nuclear extract was subjected to multiple freeze/ thaw cycles before assaying for CCA addition to U2 snRNP; for each cycle, extract was thawed on ice for 2 h and refrozen at Ϫ70°C for 2 h. To increase the fraction of U2 snRNPs with defective 3Ј ends, nuclear A, stem/loops III and IV of human U2 snRNA. B, CCA addition by the human enzyme to the 3Ј-terminal stem/loops III and IV of U2 snRNA. CCA addition to stem/loops III and IV of human U2 snRNA and to the B. subtilis tRNA Asp -D 73 CC control generates products of the same length. C, stem/loop IV of human U2 snRNA. D, CCA addition by the human enzyme to 3Ј-terminal stem/loop IV of U2 snRNA. Assays were performed as in A. E, the human CCA-adding enzyme repairs U2 snRNA faithfully. Uniformly labeled full-length U2 snRNA substrates with the mature 3Ј end or 3Ј ends lacking A, CA, or CCA were assayed in the presence of ATP alone (1 mM), CTP alone (200 M), or both ATP and CTP. Under these assay conditions, C addition is complete, and A addition is partial. CCA addition was faithful and, with these catalytic amounts of enzyme, no poly(C) polymerase activity (57,59) was observed even in the absence of ATP. extract was treated with venom 3Ј exonuclease as described previously (53). A 20-l reaction containing 100 mM Tris-HCl (pH 9.0), 100 mM NaCl, 15 mM MgCl 2 , 5 l of HeLa nuclear extract (18.2 mg/ml) and snake venom phosphodiesterase I (1.2 units/25 ml; Sigma) was incubated for 1 h at 20°C, and the enzyme was then inactivated by addition of 50 mM EDTA (54 -56).

RESULTS
The Human CCA-adding Enzyme Can Add 3Ј-terminal A, CA, and CCA to U2 snRNA Substrates-To ask whether the human CCA-adding enzyme can build or repair the 3Ј-terminal CCA sequence of U2 snRNA, we generated three sets of U2 snRNA substrates (full-length U2 snRNA, 3Ј-terminal stem/ loops III and IV, and stem/loop IV only), each with three different 3Ј-terminal sequences (N, NC, and NCC, where N is the discriminator base). Using full-length U2 snRNA lacking CCA, CA, or A as substrates (Fig. 1A), we found that the human CCA-adding enzyme incorporates only CTP into U2 snRNA substrates lacking CCA or CA and only ATP into U2 snRNA substrates lacking A (Fig. 1B). Thus, the human CCA-adding enzyme could be responsible for adding 3Ј-terminal A, CA, and CCA to U2 snRNA processing intermediates or to mature U2 snRNA in need of repair.
To identify the minimum structure or sequence required for CCA addition by the human enzyme, we assayed 3Ј-terminal stem/loops III and IV of U2 snRNA ( Fig. 2A) and 3Ј-terminal stem/loop IV alone (Fig. 2C) as substrates. The 3Ј-terminal stem/ loops III and IV of U2 snRNA were a good substrate (Fig. 2B,  lanes 1-3), but equivalent substrates with mutant or mature 3Ј ends (A 185 CU, A 185 CUA, and A 185 CCA, where A 185 is the discriminator base) were not (Fig. 2B, lanes 4 -6). The 3Ј-terminal stem/loop IV of U2 snRNA alone was also a good substrate (Fig.  2, C and D), suggesting that the 3Ј end of U2 snRNA resembles a tRNA minihelix (57,58) and that the TC loop contributes marginally, if at all, to substrate recognition (12).
The Human CCA-adding Enzyme Repairs U2 snRNA Faithfully-To examine the fidelity of CCA addition by the human enzyme to U2 snRNA, we assayed the addition of ATP alone, CTP alone, or ATP and CTP together to uniformly labeled U2 snRNAs with incomplete or mature 3Ј ends (Fig. 2E). As expected, U2 snRNAs lacking CCA or CA are substrates for CTP addition; U2 snRNA lacking A is a substrate for ATP addition; U2 snRNAs lacking CCA or CA are substrates for both CTP and ATP addition; and neither CTP nor ATP is added to U2 snRNA with a mature CCA end. Importantly, the human CCAadding enzyme does not exhibit poly(C) polymerase activity on any of these 3Ј-terminally defective U2 snRNA substrates, even in the absence of ATP (Fig. 2E, middle panel), when the enzyme is assayed at catalytic levels (enzyme to substrate ratio, 1:250) instead of the more nearly stoichiometric levels (1:10 or 1:20) used by others (57, 59).

FIG. 2-continued
Heterologous CCA-adding Enzymes Can Add 3Ј-terminal CCA to U2 snRNA-We next asked whether heterologous CCAadding enzymes belonging to class I and class II could add CCA to human U2 snRNA substrates. We tested class I enzymes from archaea (M. jannaschii, M. thermoautotrophicum, P. furiosus, and S. shibatae) and class II enzymes from Grampositive (B. subtilis and B. stearothermophilus 2 ) and Gramnegative (E. coli) eubacteria, with the human class II enzyme as control. Interestingly, only one class I enzyme (M. thermoautotrophicum) and two class II enzymes (Homo sapiens and B. stearothermophilus) were fully active for CCA addition to U2 snRNA; the other enzymes were either very weakly active (S. shibatae, M. jannaschii, and P. furiosus) or inactive (E. coli and B. subtilis) (Fig. 3A). Although fully active on tRNA substrates (data not shown), the E. coli class II enzyme was also inactive on U2 snRNA substrates lacking CA or A (Fig. 3B).
Natural U2 snRNA Is a Substrate for the Human CCA-adding Enzyme-Unlike in vitro transcripts, natural U2 snRNA is capped, 2Ј-O-methylated, and pseudouridylated (60,61). Although none of these modifications fall within the 3Јterminal stem/loops III and IV that serve as minimal substrates for the human CCA-adding enzyme, it was nonetheless important to establish that natural U2 snRNA, purified from HeLa nuclear extract, would function as substrate for the human CCA-adding enzyme. CTP addition to natural U2 snRNA was robust, but ATP addition was relatively weak (Fig. 4A, compare lanes 1). Weak ATP addition could mean that U2 snRNA is not a natural substrate for the human CCA-adding enzyme or that the distribution of U2 3Ј ends in this extract is different from that observed previously (45); alternatively, C addition may be less specific than A addition because U1 snRNA and 5 S rRNA also appear to accept one or more 3Ј-terminal Cs (Fig. 4A, lane  1). We therefore asked whether U2 snRNA lacking A might be a substrate for the human SRP (or 7SL RNA) adenylating enzyme; this enzyme is homologous to the mRNA poly(A) polymerase but is the product of a distinct gene (62). However, unlike the human CCA-adding enzyme (Fig. 4A, lanes 1), the SRP 7SL RNA adenylating enzyme failed to add ATP to natural U2 snRNA and instead generated a range of labeled products averaging over 400 nt (Fig. 4A, lanes 2). When the human CCA-adding enzyme (Fig. 4B, left panel) and SRP adenylating enzyme (Fig. 4B, right panel) were assayed on 77 homogeneous nucleotide substrates (3Ј-terminal stem/loops III and IV of U2 snRNA or a B. subtilis tRNA Asp tRNA substrate, both lacking 3Ј-terminal A), the CCA-adding enzyme added a single ATP to each substrate, but the SRP adenylation enzyme generated large polyadenylated products of Ͼ400 nt presumably containing poly(A) tails of Ͼ300 nt. These results suggest that human SRP 7SL RNA adenylating enzyme lacks substrate specificity, at least on naked RNAs, and can function as a processive poly(A) polymerase; it is thus unlikely to play any role in U2 snRNA metabolism. 2 The B. stearothermophilus CCA-adding enzyme was cloned using PCR primers corresponding to the most highly conserved regions of the class II enzymes (H. D. Cho and A. M. Weiner, manuscript in preparation).

FIG. 3. CCA addition to human U2 snRNA by heterologous enzymes.
A, heterologous class I and class II CCA-adding enzymes can add 3Ј-terminal CCA to a U2 snRNA substrate lacking CA. Asterisks indicate labeled nucleotide; assays were performed as described in the Fig. 2  were assayed using 3Ј-terminal stem/loops III and IV of U2 snRNA and a B. subtilis tRNA Asp substrate, both lacking 3Ј-terminal A. Assays were performed as described in the Fig. 1 legend, but products were resolved by denaturing 10% PAGE (A) or 12% PAGE (B).
To determine whether assembly of U2 snRNA into a ribonucleoprotein particle might inhibit or facilitate CCA addition by the human enzyme, we performed CCA addition assays in HeLa nuclear extract (Fig. 5). Consistent with the results for naked human U2 snRNA (Fig. 4A), CTP addition to endogenous U2 snRNPs was much more efficient than ATP addition (compare Fig. 5A, lane 2, with Fig. 5B); moreover, CTP stimulated ATP addition to tRNA and also, albeit much more weakly, to U2 snRNA (Fig. 5B). Although active on naked RNAs (Fig.  4B), the SRP 7SL adenylating enzyme did not appear to recognize U2 snRNP or any other RNP substrates in nuclear extract (Fig. 5A, lane 1). These results suggest that endogenous 3Ј exonucleases remove CA or CCA from mature U2 snRNA and tRNA and that the human CCA-adding enzyme can repair these species.
More efficient addition of CTP than ATP may reflect the 50-fold higher K m for ATP than for CTP (Fig. 6). Kinetic parameters for the human CCA-adding enzyme with both U2 snRNA and tRNA substrates were determined by measuring the addition of [␣-32 P]CTP to tRNA-DC and U2 snRNA lacking 3Ј-terminal CA in the presence of saturating ATP (Fig. 6A) and the addition of [␣-32 P]ATP to tRNA-DCC and U2 snRNA lacking 3Ј-terminal A (Fig. 6B). The observed Michaelis constants (K m ) were 10 M for CTP and 200 M for ATP with the tRNA-DCC substrate and 33 M for CTP and 1.5 mM for ATP with the U2 snRNA substrate. The substrate dependence of K m values for CTP and ATP could reflect a conformational change in the proteins (70) or collaborative templating by substrate and enzyme together (5).
In an effort to uncover endogenous 3Ј exonuclease activities that might act on U2 snRNP, we subjected HeLa nuclear extract to prolonged incubation at room temperature or to multiple freeze/thaw cycles; however, CCA addition to U2 snRNA was unaffected by these treatments. We also treated HeLa nuclear extract with snake venom 3Ј exonuclease to determine whether the 3Ј end of U2 snRNA is accessible to exonuclease attack, but CCA addition was unaffected by this treatment as well. These negative results (data not shown; see "Experimental Procedures") suggest that U2 snRNAs with incomplete CCA termini do not reflect degradation of mature U2 snRNPs but are generated by RNA processing or 3Ј exonuclease activity before addition of the Sm and/or AЈ and BЉ proteins to U2 snRNA.

DISCUSSION
All vertebrate U snRNAs belonging to the Sm class of U snRNPs and some small nucleolar RNAs such as U3 small nucleolar RNA apparently use the same 3Ј end formation apparatus: specialized U snRNA promoter elements drive transcription by a form of RNA polymerase II that allows recognition of a highly conserved 3Ј end formation signal (3Ј box) located 6 -25 nt downstream from the U snRNA coding (20). Whether the 3Ј box functions as a transcription termination signal or as an RNA processing signal is still unknown (25); however, the first detectable U snRNA precursors have 3Јterminal tails that extend nearly to the 3Ј box (22,31,32,35,36,71) and are trimmed by cytoplasmic nucleases that may be related to the yeast multinuclease complex known as the exosome (10,35,36,38).
The 3Ј-terminal sequence of most U snRNAs can be aligned with the corresponding genes, suggesting that nucleases are capable of generating the mature 3Ј end of most U snRNA precursors. However, ϳ65% of mature human U2 snRNA ends in 3Ј-terminal CCA; the other 35% ends in 3Ј CC or possibly 3Ј C (44,45). The 3Ј-terminal A of U2 snRNA cannot be encoded because the 3Ј end of the U2 snRNA gene is CC/CC, where the slash indicates the last encoded nucleotide (47). The 3Ј-terminal A is also unlikely to be an untemplated addition by the transcribing RNA polymerase (72) because transcription must continue for Ͼ10 bp beyond the 3Ј end to generate the U2ϩ10 family of precursors. We cannot exclude template-instructed misincorporation, as occurs in certain RNA viruses (73), but this has never been observed for a cellular RNA polymerase. Thus, the solitary 3Ј-terminal A of U2 snRNA is almost surely added posttranscriptionally.
A to U2 snRNA is the SRP 7SL adenylating enzyme (62); another is the CCA-adding enzyme (tRNA nucleotidyltransferase), which builds and repairs the 3Ј-terminal CCA of tRNA. In fact, the 3Ј end of U2 strongly resembles the minimal substrate for the CCA-adding enzyme: a tRNA minihelix (the "top half "of tRNA) in which the acceptor stem stacks on the TC stem/loop. Not only is U2 stem IV 12 bp, the optimal length for CCA addition by the E. coli and S. shibatae enzymes to a minihelix substrate (12), but the stem is followed by A 185 CCA OH , where A 185 is formally analogous to the discriminator base of tRNA, most frequently a purine (74). Thus, in principle, the CCA-adding enzyme might be capable of completely rebuilding the 3Ј-terminal CCA sequence of U2 snRNA if nucleases overprocessed the U2 snRNA precursor or other 3Ј exonucleases attacked the mature U2 snRNA.
To ask whether CCA-adding enzymes can add 3Ј-terminal A, CA, or CCA to U2 snRNA precursors, we used three sets of substrates (full-length U2 snRNA, 3Ј-terminal stem/loops III and IV, and stem/loop IV alone), each lacking 3Ј-terminal A, CA, or CCA. We assayed these substrates with CCA-adding enzymes of both classes from all three kingdoms (3): class I CCA-adding enzymes from archaea (S. shibatae, M. jannaschii, P. furiosus, and M. thermoautotrophicum), class II enzymes from Gram-positive (B. stearothermophilus and B. subtilis) and Gram-negative (E. coli) eubacteria, and the human class II enzyme. All substrates were active with the human class II CCA-adding enzyme, including 3Ј-terminal stem/loop IV alone (Figs. 1-3), and CCA addition was faithful (Fig. 2E); however, the E. coli class II CCA-adding enzyme was inactive on the U2 snRNA substrates but fully active on tRNA substrates, consistent with the idea that CCA addition to U2 snRNA is specific. In contrast to CCA-adding enzymes, the SRP 7SL RNA adenylating enzyme nonspecifically added long 3Ј-terminal tails of Ͼ400 nt to U2 and other RNA substrates (Fig. 4). Thus, the SRP RNA adenylating enzyme behaves like a poly(A) polymerase and is unlikely to play a role in U2 snRNA biosynthesis. These results suggest that the 3Ј stem/loop of U2 snRNA does in fact resemble a tRNA minihelix, the smallest efficient substrate for class I and II CCA-adding enzymes (12,57), and that CCA addition to U2 snRNA could take place in vivo.
We also found that the human CCA-adding enzyme is active on natural U2 snRNA purified from nuclear extract (Fig. 4) and on native U2 snRNPs (Fig. 5). In the mature 17S U2 snRNP, SF3a and SF3b splicing factors are bound to the 5Ј-terminal domain including stem/loops I, IIa, and IIb; Sm proteins are bound to the conserved Sm-binding sequence; and U2-specific proteins AЈ and BЉ are bound to stem/loop IV (62-64, 75, 76). However, assembly of Sm proteins onto U2ϩ10 precursors occurs before formation of the mature 3Ј end (34,38), and U2 snRNP particles are the substrates for U2 snRNA 3Ј processing (34,67,68). The activity of the human CCA-adding enzyme on native U2 snRNPs is consistent with the observation that 3Ј maturation of U2 snRNA occurs in vivo after snRNP assembly has begun.
There are ample precedents for the ability of CCA-adding enzymes to recognize tRNA-like structures. tRNA consists of two structurally and functionally independent domains: a "top half" or minihelix consisting of the acceptor stem stacked on the TC stem/loop, and a "bottom half" consisting of the DHU stem/loop stacked on the anticodon stem/loop (77,78). Not only does a tRNA minihelix serve as substrate for CCA-adding enzymes (12,57), but single-stranded RNA viruses often have 3Ј-terminal tRNA-like structures that serve as substrates for enzymes of tRNA metabolism including tRNA synthetases, the CCA-adding enzyme, and RNase P (11, 79 -81). Maize mitochondrial mRNAs also contain posttranscriptionally added 3Ј-terminal nucleotides that could be the product of a mitochondrial form of the CCA-adding enzyme (82).
Almost all U snRNAs have 3Ј-terminal stem/loops, presumably to confer stability on the RNAs (83), and some small RNAs may be stabilized by posttranscription adenylation (45,46). Although direct RNA sequencing data are scanty, the U2 snRNA 3Ј-terminal stem/loop appears to be followed by CCA only in mammals (human and rat); even frog U2 snRNA lacks CCA (Fig. 7). The U2 sequences of other organisms are either cDNA or DNA sequences. The cDNA sequences are all derived from poly(A)-tailed RNA, so that it is impossible to distinguish posttranscriptional 3Ј-terminal A addition from the poly(A) tail. The DNA sequences almost all have CϩA-rich 3Ј-flanking regions; whether this CϩA-rich region plays a role in 3Ј end formation remains to be seen (20,25,38), but it has already led to arbitrary assignment of U2 snRNA 3Ј ends based solely on DNA sequence information. Because U2 snRNA sequences are strongly conserved from yeasts to mammals, failure to conserve the 3Ј-terminal CCA suggests that this sequence is not required for U2 snRNA function; however, mammalian U2 snRNA is known to be very stable (84), and CCA addition could serve to protect U2 snRNA from 3Ј exonucleases. In fact, U2 might exploit the CCA-adding enzyme, which exists in nuclear, cytoplasmic, and mitochondrial forms (48), 3 as the functional homolog of telomerase, to replenish 3Ј-terminal nucleotide sequences that are lost by attrition (76,77).