A novel nuclear human poly(A) polymerase (PAP), PAP gamma.

Poly(A) polymerase (PAP) is present in multiple forms in mammalian cells and tissues. Here we show that the 90-kDa isoform is the product of the gene PAPOLG, which is distinct from the previously identified genes for poly(A) polymerases. The 90-kDa isoform is referred to as human PAP gamma (hsPAP gamma). hsPAP gamma shares 60% identity to human PAPII (hsPAPII) at the amino acid level. hsPAP gamma exhibits fundamental properties of a bona fide poly(A) polymerase, specificity for ATP, and cleavage and polyadenylation specificity factor/hexanucleotide-dependent polyadenylation activity. The catalytic parameters indicate similar catalytic efficiency to that of hsPAPII. Mutational analysis and sequence comparison revealed that hsPAP gamma and hsPAPII have similar organization of structural and functional domains. hsPAP gamma contains a U1A protein-interacting region in its C terminus, and PAP gamma activity can be inhibited, as hsPAPII, by the U1A protein. hsPAPgamma is restricted to the nucleus as revealed by in situ staining and by transfection experiments. Based on this and previous studies, it is obvious that multiple isoforms of PAP are generated by three distinct mechanisms: gene duplication, alternative RNA processing, and post-translational modification. The exclusive nuclear localization of hsPAP gamma establishes that multiple forms of PAP are unevenly distributed in the cell, implying specialized roles for the various isoforms.

mRNA poly(A) tail synthesis. PAP has been identified and cloned from several eukaryotic species, e.g. yeast, human, mouse, bovine, frog, and chicken (9 -19). Interestingly, multiple forms of PAP are present in cell lines and tissues of several species (9, 11, 13-16, 19 -21). In HeLa cell nuclear extracts, three isoforms, having apparent molecular masses of 90, 100, and 106 kDa, have been found (16). The molecular mechanisms for generating all these isoforms are still not completely understood. However, molecular cloning has established that at least five isoforms of full-length PAP can be generated by alternative RNA processing (15,17). 2 It is also known that phosphorylation contributes to the multiplicity of PAP (9,16,20). Recently it has been established that PAP and PAP-related genes are present in the human genome (11,13,14,22). Therefore, so far at least three distinct mechanisms can generate multiple isoforms of PAP: gene duplication, alternative RNA processing, and post-translational modification. These phenomena unexpectedly increase the diversity of PAP and raise questions about the functional significance of multiple PAPs in vivo.
It seems reasonable to hypothesize that different PAPs are responsible for different functions in vivo, because PAP participates in a whole set of different reactions, e.g. RNA cleavage at the poly(A) site and AAUAAA-dependent or -independent poly(A) tail synthesis (23, 24). Biochemical fractionation studies have indicated that different forms of PAP reside at different subcellular compartments (16,25). A testis-specific PAP has been identified, suggesting that some isoforms of PAP are restricted to certain developmental stages or tissues (11,13).
In this report we have molecularly cloned the human 90-kDa PAP isoform, previously identified in HeLa nuclear extracts. The 90-kDa isoform is encoded by a distinct locus recently identified as a PAP-related gene (22). The gene has been named PAPOLG, and its product is hsPAP␥. In a recent report (14) the same gene was implicated in monoadenylation of small RNAs. Here, we show that hsPAP␥ is a bona fide poly(A) polymerase harboring both nonspecific and CPSF/AAUAAA-dependent polyadenylation activity. hsPAP␥ is exclusively localized in the nucleus.

EXPERIMENTAL PROCEDURES
Plasmid Constructs-Full-length hsPAP␥ and various deletion mutants were molecularly cloned by a standard RT-PCR procedure using the following strategy. A 371-amino acid N-terminal fragment of hs-PAP␥ was amplified by RT-PCR using HeLa total RNA and primers a and b, subcloned into pGEM-T vector (Promega Inc.), and further subcloned into the pET-32(a) vector (Novagen Inc.) between the NcoI and SacI sites. The resulting clone was called pPAP␥(H1-371), where H denotes the N-terminal tag and the numbers refer to amino acids in full-length hsPAP␥. C-terminal deletions of hsPAP␥ were cloned by inserting PCR products derived with primer pairs c-d, c-e, c-f, and c-g between the EcoRI and SacI sites of pPAP␥(H1-371) giving rise to pPAP␥(H1-493), pPAP␥(H1-506), pPAP␥(H1-575), and pPAP(H1-* This work was supported by the Swedish Strategic Research Foundation, the European Commission through its Training and Mobility of Researchers program, and funds from Uppsala University. 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: Tel.: 46-18-471-4908; Fax: 46-18-471-4862; E-mail: anders.virtanen@icm.uu.se. 1 The abbreviations used are: PAP, poly(A) polymerase; CPSF, cleavage and polyadenylation specificity factor; RT-PCR, reverse transcription-polymerase chain reaction; b-MEOH, ␤-mercaptoethanol; DTT, dithiothreitol; PBS, phosphate-buffered saline; PFA, paraformaldehyde; kb, kilobase(s); NLS, nuclear localization signal. 683), respectively. The C terminus of hsPAP␥ was identified and cloned into pGEM-T vector by 3Ј-RACE using the CLONTECH Smart Race cDNA amplification kit and gene-specific primers h and i according to the manufacturer. After identification of the C-terminal end, including the stop codon, primer pair c-j was used for generation of pPAP␥(H1-736) as described above. An N-terminal deletion mutant was amplified using primer pair k-j for deletion of the first 16 amino acids and the resulting clone was named pPAP␥(H17-736). The clone pPAP␥(H521-683) was generated as outlined above using primers l and g. The pET-32(a) vector introduces an N-terminal thioredoxin-tag which increases the expression of the soluble recombinant protein, a histidinetag and an S-tag enabling easy purification via affinity chromatography. Full-length hsPAP␥ and one C-terminal deletion mutant were also subcloned into the pCAL-c vector (Stratagene) between the NcoI and KpnI sites, using the same strategy as above (primer pairs a-j and a-g) giving rise to pPAP␥(1-736C) and pPAP␥(1-683C), respectively. Fulllength hsPAP␥ and C-terminal deletion mutants were also subcloned into the EGFP-C2 vector (CLONTECH) between the XhoI and KpnI sites using primer pairs m-b, m-e, and m-j. The resulting clones were named pPAP␥(EGFP1-371), pPAP␥(EGFP1-506), and pPAP(EG-FP1-736), respectively. Primers used were as follows: (a) 5Ј-CACCAT- Restriction sites for cloning are included in the primer sequences shown in boldface: NcoI in a; KpnI and SacI in b, d, e, f, and g and XhoI in m. A stop codon was introduced between the KpnI and SacI sites. The KpnI restriction site was introduced to enable cloning into the pCALc vector and adds two extra amino acids at the C terminus of all the pET-32a clones expressing hsPAP␥. The NcoI cloning site in primer a introduces a point mutation at the second amino acid in the sequence changing lysine to glutamate. All clones have been sequenced using the Big-Dye Terminator sequencing kit (Applied Biosystems). Expression and Purification of Recombinant Forms of PAP␥-Expression plasmids were used to transform BL21(DE3) pLysS Escherichia coli strains. One colony was used for inoculation of 50 -100 ml of TB medium in the presence of 50 g/ml carbenicillin and 34 g/ml chloramphenicol and grown overnight at 37°C without shaking. The 50-to 100-ml culture was inoculated into a final 0.5-to 1-liter culture in TB medium containing antibiotics. Bacteria were grown at 37°C (vigorous shaking) and were induced with 0.42 mM isopropyl-1-thio-␤-Dgalactopyranoside plus 0.524 mM MgCl 2 at A 600 ϳ 0.5-1.0. Cells were harvested by centrifugation 3 h post-induction, and pellets were frozen at Ϫ70°C. Extracts for His-tagged PAP␥ were prepared by unthawing the cells on ice and lysing with buffer A containing 1 tablet of EDTAfree protease inhibitors), followed by sonication (three times 10 s), centrifugation 20 min at 39,000 ϫ g, and 0.45-m filtration. Extracts were mixed with 1 ml of Talon metal affinity resin (CLONTECH) equilibrated in buffer A, and proteins were bound batch-wise by 1-h rotation. The resin was washed with buffer A and subsequently washed with buffer B, and the proteins were eluted with buffer C. The eluate was loaded onto a HiTrap chelating column (Amersham Pharmacia Biotech) equilibrated with buffer D. The column was washed with buffer D and subsequently with buffer E. Proteins were eluted with buffer F. The eluate was loaded on a Heparin HiTrap column equilibrated in buffer G, washed with the same buffer and proteins were eluted with buffer H. Extracts for calmodulin-tagged PAP␥ where prepared as described above, but cells were lysed in buffer I containing 1 tablet of EDTA-free protease inhibitors. Extracts were mixed batchwise with 0.75 ml of calmodulin affinity resin (Stratagene) equilibrated in buffer I, and proteins were bound overnight. The resin was washed with buffer J followed by buffer K. Proteins were eluted with buffer L. Protease inhibitors 0.5 mM phenylmethylsulfonyl fluoride, 1.0 g/ml leupeptin, 1.0 g/ml pepstatin, and 1.0 g/ml aprotinin were added freshly to all buffer solutions, and all procedures were performed at 4°C.

Buffers Used for Purification of Recombinant Proteins-Buffer
Antibodies-Polyclonal antiserum specific for the C-terminal of hs-PAP␥ were generated by immunizing two rabbits using 0.45 mg/rabbit recombinant purified PAP␥(H521-683) polypeptide, followed by three boost injections with the same amount of antigen. The sera was named anti-PAP␥. Peptide antiserum specific for PAPII was purchased from Sigma-Genosys using a synthetic peptide (N-terminal: CKTSSTDLS-DIPA) corresponding to amino acids 715-726 of hsPAPII for immunization, and was named anti-PAPIIex22. Monoclonal antibodies were Y12 (26) and 20:14 (16).

SDS-Polyacrylamide Gel Electrophoresis and Western Blot
Analysis-HeLa nuclear extracts were purchased from the Computer Cell Culture Center. SDS-polyacrylamide gel electrophoresis was carried out according to a previous study (27) as was Western blotting (28). Detection was done by anti-mouse or anti-rabbit immunoglobulin, horseradish peroxidase-linked whole antibody (Amersham Pharmacia Biotech) diluted 1:1000, and ECL plus chemiluminescence reagent (Amersham Pharmacia Biotech).
Poly(A) Polymerase Assay Conditions-Nonspecific polyadenylation activity assays were carried out as described previously (21,29) with modifications optimizing the activity; the reaction mixture (25 l) contained: 100 mM Tris/HCl buffer, pH 8.6 (measured at room temperature), 40 mM KCl, 0.040 mM EDTA, 10% glycerol, 1 mM DTT, 9 units of RNasin (ribonuclease inhibitor), 0.1% Nonidet P-40, 0.5 mM MnCl 2 , 0.5 mg/ml bovine serum albumin, 0.5 mM cold ATP, 1.2 Ci of [␣-32 P]ATP (3000 Ci/mmol) and 2 M oligoA 15 (Dharmacon), and the reaction was performed for 20 min at 37°C. One unit of PAP is defined as the amount of enzyme needed for incorporation of 1 pmol of AMP per min. Reaction rate was measured in a linear range versus PAP concentrations (8 -23 nM) and time (10 -30 min). Kinetic parameters were determined using oligoA 15 in the concentration range 0.0125-2 M for the full-length hsPAP␥ and hsPAPII. In the case of kinetic estimations for the deletion mutants the same enzyme concentration was used, however, the primer concentration was in the 0.5-5 times K m range. Unthawed recombinant hsPAP␥ and hsPAPII were stabilized by addition of 0.05% Nonidet P-40, 20% glycerol, and 1 mM DTT for the time kept on ice. Reactions were stopped by precipitation of the insoluble polyadenylated product in acid conditions (5% trichloroacetic acid-1% sodium pyrophosphate) in glass fiber filters and washed three times with 5% trichloroacetic acid (30).
The specific polyadenylation activity was carried out as described previously (29,33) with modifications to normalize the differences in between the specific and nonspecific assays in this study. 32 P-Labeled and capped RNA substrates (L3(54), L3G(54)) were synthesized by in vitro transcription and purified as previously described (34). CPSF partially purified from calf thymus (35) and recombinant hsPAP␥ were used as specified in the figure legends. The reaction mixture (25 l) contained: 100 mM Tris/HCl buffer, pH 8.3 (measured at room temperature), 40 mM KCl, 0.040 mM EDTA, 9.6% glycerol, 1 mM DTT, 9 units of RNasin (ribonuclease inhibitor), 0.01% Nonidet P-40, 0.72 mM MgCl 2 , 1 mM ATP, 2.5% polyvinylalcohol, 20 mM creatinine phosphate, and the reaction was performed for 20 min at 30°C. The reaction was stopped in Proteinase K buffer, and the incubated RNA product was extracted and resolved in 10% polyacrylamide (acrylamide/bisacrylamide 19:1)-7 M urea.
Immunocytochemical Methods-HeLa cells were grown up to 50 -70% confluency on coverslips in the presence of Dulbecco's modified Eagle's medium supplemented with glutamine and 10% fetal calf serum (Life Technologies, Inc.). Coverslips were washed two times in PBS and fixed in 1% paraformaldehyde (PFA) in PBS (pH 7.3) for 3 min, extracted with 0.5% Triton X-100 in PBS for 15 min, and then post-fixed in 4% PFA in PBS for 10 min. For immunofluorescence staining the following antibodies were used: Primary antibodies; 20:14, anti-PAP␥, Y12, and anti-PAPIIex22 in dilutions 1:2, 1:40, 1:10, 1:20, respectively; the respective pre-immune serum was used in the cases of polyclonal antibodies at the same dilutions. Secondary antibodies; species-specific goat anti-mouse IgG coupled to biotin (Amersham Pharmacia) or to Alexa Fluor 488 (emission at green spectrum) (Molecular Probes), species-specific goat anti-rabbit IgG coupled to Alexa Fluor 594 (emission at red spectrum) (Molecular Probes). Fixed cells were incubated for 30 min with blocking reagent buffer (5% Eliza blocking reagent, Roche Molecular Biochemicals) at room temperature. Subsequently they were incubated for 30 min or 2 h with primary antibodies diluted in blocking reagent and washed 3 ϫ 5 min in PBS. Secondary antibodies were incubated for 30 min or 1 h and washed 3 ϫ 5 min in PBS. In the case where dual staining experiments were performed with monoclonal antibody 20:14 and anti-PAP␥ and a biotin labeled anti-mouse secondary antibody was used, the cells were washed 4 ϫ 5 min in PBS and incubated for 1 h with streptavidin coupled to Alexa Fluor 488 (Molecular Probes). All coverslips were mounted in Vectashield (Vector Laboratories) and shielded. Where polyclonal serum was used, a control pre-immune serum was used to subtract the background signal. Fluorescence microscopy was performed in an Axioplan 2 imaging fluorescence microscope, using a 100ϫ objective lens. Image analysis was done by the Axion vision.3 software.
Transfection Methods-HeLa cells, grown up to 50% confluency, were transfected using plasmids pPAP␥(EGFP1-371), pPAP␥(EGFP1-506), and pPAP␥(EGFP1-736), and left to grow for 10 or 24 h. The Superfect transfection reagent (Qiagen) was used, and conditions were optimized for ratio of DNA:transfection reagent, as suggested by the manufacturer. Cells were fixed in 4% PFA (in PBS, pH 7.3), shielded, and analyzed in a fluorescence microscope by excitation at 495 nm and emission at the green spectrum.

Molecular
Cloning of Human PAP␥-Monoclonal antibodies NN:2 and 20:14 raised against hsPAPII recognize three isoforms of PAP: 90, 100, and 106 kDa in sizes (16). However, a polyclonal antibody raised against bovine PAPII recognizes only the two larger forms (24,36). A reason for this discrepancy could be that the monoclonal antibodies recognize a common epitope shared among all three isoforms of PAP, whereas the polyclonal antibody is directed against epitopes not present in the 90-kDa form. This experimental evidence suggests that the 90-kDa isoform of PAP has unique antigenic epitopes unrelated to the 100-and 106-kDa forms, implying that the 90-kDa isoform could be encoded by a separate gene.
To identify potential human PAP-related genes we regularly searched using the BLAST algorithm (37) high throughput (htgs) and non-redundant sequence data bases, while they were being released during the human genome sequencing project (38). During these searches we identified a PAP-related sequence in the human genomic clone (AC011245.6) located on chromosome 2. The same locus has recently been identified as a PAP-related gene and as a small RNA monoadenylating enzyme (14,22). Further data base searches revealed several overlapping expressed sequence tags. These results combined with 3Ј-RACE semi-nested RT-PCR allowed us to predict the sequence of an mRNA encoding a potential PAP. The novel human gene was named PAPOLG and its product hsPAP␥. The sequence information was used to molecularly clone cDNAs originating from HeLa cells by RT-PCR. A schematic drawing of the exon/intron organization of hsPAP␥ and comparison to the previously reported gene hsPAPII is shown in Fig. 1A. The deduced amino acid sequence of hsPAP␥ is presented in Fig. 1B and compared with the bovine and human PAPII. Structural and functional domains/motifs are also represented. A comparison using the ClustalX algorithmic approach (39) showed that hsPAP␥ has an overall identity of 67% at the nucleotide level and 60% identity at the amino acid level.
Organization of the hsPAP␥ Gene-The genes encoding hs-PAPII and hsPAP␥ span 62.5 and 37 kb of genomic sequences, respectively. They both contain 22 exons, and all splice sites obey the GT/AG rule (40). The topology and the sizes of exons 2-16 are shared between the two genes, implicating that they share a common ancestor and arose through gene duplication (41,42). Sequence comparison (Fig. 1A) revealed that the exons 1 of both genes were unrelated to each other; exons 2-16 were highly conserved both at the amino acid and nucleotide levels, having an overall identity of ϳ75% at both levels; exons 17-21 were less conserved in their sequences whereas exon 22 exhibits a high degree of identity both at the amino acid and nucleotide levels. The last half of exon 22 encodes a potential U1A protein-interacting region (see also below).
Structural Organization of hsPAP␥-An inspection of known structural and functional motifs/domains in hsPAPII revealed that several of those were conserved in hsPAP␥. These motifs/ domains included amino acids important for catalysis, recognition of the ATP substrate, and RNA binding (29,43) (Fig. 1B). The cyclin-recognition motif and four of the seven consensus and non-consensus phosphorylation sites that have been mapped for cyclin dependent kinases were conserved (44,45). Two nuclear localization signals (NLS) (46) were conserved between the two PAPs, whereas a third putative bipartite NLS was found in the C-terminal end of hsPAP␥. The sequence encompassing the U1A protein interaction region is highly conserved (14 out of 18 amino acids).
The 90-kDa Isoform Is the Product of the Novel hsPAP Gene-To raise an antiserum specific for hsPAP␥, we molecularly cloned the C-terminal region of hsPAP␥ spanning amino acids 521-683 into the pET32(a) vector. The recombinant polypeptide was expressed in E. coli, purified to homogeneity, and used to raise an hsPAP␥-specific antiserum, named anti-PAP␥. Fig. 2A shows that the obtained serum was specific for hsPAP␥, because it only recognized recombinant versions of hsPAP␥ and not hsPAPII. In these experiments C-terminally calmodulin-tagged recombinant proteins were used to exclude recognition of the N-terminal-located tags present in the polypeptide used for immunization. As predicted the monoclonal antibody (20:14) raised against hsPAPII recognized hsPAP␥ (Fig. 2B). An analysis of C-terminal deletion constructs of hsPAP␥ revealed that the epitope was located in the highly conserved N-terminal region of hsPAP␥ and hsPAP II (Figs. 1B and 2B) To investigate if the anti-PAP␥ serum recognized the 90-kDa isoform of HeLa cell PAP, we probed HeLa nuclear extracts. Fig. 2C shows that the serum exclusively recognized the 90-kDa species. An hsPAPII-specific polyclonal antiserum, named anti-PAPIIex22, was raised against a synthetic peptide of exon 22 (amino acids 715-726) of hsPAPII. An affinity-purified anti-PAPIIex22 recognized the 100-and 106-kDa mobility species and not the 90-kDa isoform (Fig. 2C). Thus, we conclude that the 90-kDa isoform corresponds to hsPAP␥, the product of the PAPOLG gene.
Properties of hsPAP␥-To investigate if hsPAP␥ had polyadenylating activity, we used the nonspecific polyadenylation assay in the presence of Mn(II) and various nucleotide triphosphates. The assay was designed so that the amount of hsPAP␥ was provided in excess. Fig. 3A shows that recombinant PAP␥(H1-736) exhibited specificity for incorporation of ATP, whereas incorporation of UTP, GTP, CTP, and dATP were inefficient and stopped after the addition of one to two molecules of the respective nucleotide analogue.
PAPII acquires specificity for hexanucleotide containing mRNAs in the presence of cleavage and polyadenylation specificity factor (CPSF) (33). To investigate whether hsPAP␥ had this classical property, we performed specific polyadenylation assays in the presence of Mg(II) and partially purified CPSF from calf thymus (35). Fig. 3B shows that hsPAP␥ exhibited CPSF/AAUAAA-dependent activity, because the L3(54) RNA substrate was efficiently polyadenylated compared with the hexanucleotide mutated L3G(54) RNA substrate. Furthermore, hsPAP␥ did not exhibit any polyadenylation activity in the presence of Mg(II) when CPSF was omitted. A C-terminal deletion mutant PAP␥(H1-371) was inactive, as described for bovine PAPII (29). We conclude that hsPAP␥ exhibits the fundamental catalytic properties for a bona fide poly(A) polymerase, i.e. specific incorporation of ATP-and CPSF/AAUAAA-dependent polyadenylation activity. Kinetic Parameters and Mutational Analysis of hsPAP␥-To further characterize hsPAP␥ we determined the kinetic parameters (K m and V max ). Nonspecific polyadenylation activity was measured in the presence of Mn(II) and highly purified oli-goA 15 . The K m for hsPAP␥ was shown to be 0.051 M using a variety of calculation methods. The ratio V max /K m that represents the real efficiency of the reaction in terms of affinity to the primer and actual catalytic capacity was determined. The kinetic parameters for hsPAP␥ (Table I) are in a similar range as for recombinant hsPAPII and for the reported bovine PAPII (24,29). A calmodulin tag at the C terminus of hsPAP␥ and hsPAPII did not significantly alter the kinetic parameters (Table I). Thus, recombinant hsPAP␥ has similar kinetic properties as hsPAPII.
Inhibition of hsPAP␥ Activity by U1A-The inhibition of PAPII by two molecules of U1A protein is well documented (31,32). The inhibition requires the last 18 amino acids of PAPII and a region of U1A corresponding to amino acids 102-115. Because hsPAP␥ contains a putative U1A interaction motif, we tested whether it could be inhibited by U1A. Fig. 4 shows that hsPAP␥ was inhibited by an U1A di-peptide but not by the U1A mono-peptide. hsPAPII was inhibited to the same extent under these conditions (data not shown). In these experiments Cterminally tagged recombinant PAP␥(1-736C) was used, because even a loss of two to three amino acids from the C terminus abolishes the inhibition effect. 3 Thus, the C-terminally located U1A interaction motif is functional, and hsPAP␥ activity can be inhibited by U1A as previously reported for bovine PAPII.
Subcellular Localization of hsPAP␥-Biochemical fractionation studies suggested that the 90-kDa isoform of PAP was nuclear, whereas the 106-and100-kDa isoforms were both nuclear and cytoplasmic (16). To study whether hsPAP␥ localize to the nucleus we used antibodies 20:14 and anti-PAP␥ in a dual staining approach using indirect immunofluorescence techniques. Fig. 5E shows that the monoclonal antibody 20:14 gave a nuclear and weak cytoplasmic staining, as previously reported (36). In the same cell, native hsPAP␥ exhibited an exclusive nuclear distribution (compare panels E and F, Fig. 5). The polyclonal sera specific for exon 22 of hsPAPII type (anti-PAPIIex22) stained both cytoplasm and nucleus (Fig. 5C). Control antibody Y12, recognizing the Sm epitope of general nuclear splicing factors, showed an expected nuclear distribution (Fig. 5A). Intriguingly, in dual staining experiments using anti-PAP␥ and Y12, a high degree of co-localization was observed indicating that hsPAP␥ localizes close to structures enriched in general splicing factors (data not shown). Our data show that endogenous hsPAP␥ is exclusively nuclear, whereas hsPAPII is both nuclear and cytoplasmic.
The C-terminal Region of hsPAP␥ Is Important for Nuclear Localization-To identify regions important for guiding hs-PAP␥ to the nucleus, we used a transient expression approach with hsPAP␥/EGFP chimeric proteins. Two C-terminal deletion mutants were constructed: PAP␥(EGFP1-506) containing the putative NLS 1 region and full-length PAP␥(EGFP1-736) containing all three putative NLS 1, 2, and 3 regions. Fig. 5 (H-J) shows that PAP␥(EGFP1-736) was exclusively nuclear, whereas PAP␥(EGFP1-506) was both nuclear and cytoplasmic. 3 S. Gunderson, personal communication.  a If the deviation of the parameters was Ͻ40%, the average is indicated, otherwise the range is shown. The assays were carried out at least two to five times, using two independent recombinant protein preparations. PAP␥(EGFP1-736) and PAP␥(EGFP1-506) have predicted molecular masses of ϳ115 and 70 kDa, respectively. This makes both proteins too large to passively enter the nucleus (47). Thus, we conclude that NLS 1 can mediate partial nuclear import and that the entire C-terminal region (amino acids 506 -735) of hsPAP␥ is required for exclusive nuclear localization. We note that this region contains the putative NLS 2 and 3 elements.

DISCUSSION
Multiple Isoforms of PAP-Multiple forms of PAP are present in mammalian cell lines and tissues (11,13,14,16,20,21). In HeLa cell nuclear extracts three isoforms, having apparent molecular masses of 90, 100, and 106 kDa, have been found (16). In this study, we show that the 90-kDa isoform (hsPAP␥) is encoded by a distinct locus named PAPOLG. Similar sequences and exon topology of the hsPAP␥ and hsPAPII genes suggest that the two genes arose by gene duplication (Fig. 1A).
In summary, at least three mechanisms can encounter for multiple isoforms of PAP, gene duplication, alternative RNA processing, and post-translational modifications (11, 13, 15-17, 20 -22). These phenomena unexpectedly increase the diversity of PAP and provoke questions about the functional significance of multiple PAPs in vivo.
hsPAP␥ Is a Bona Fide PAP-Several lines of evidence suggest that hsPAP␥ is a bona fide PAP. Most importantly, hsPAP␥ is specific for the ATP substrate (Fig. 3A), shows hexanucleotide-dependent polyadenylation activity in the presence of CPSF (Fig. 3B), and has similar kinetic parameters as human/bovine PAPII (Table I and Refs. 24,29). Furthermore, the amino acid sequence of the region required for PAPII catalytic activity and CPSF/hexanucleotide-dependent polyadenylation activity is highly conserved in hsPAP␥, suggesting similar functional and structural properties. The resolved crystal structure of PAP (43,48) demonstrated that amino acids 365 to 513 of bovine PAP, folds into a compact globular domain topologically similar to the RNA binding domains of several RNA binding proteins. The same region of hsPAP␥ contains most likely a similar RNA binding domain and is, in analogy to bovine PAPII, important for CPSF/hexanucleotide-dependent activity (Tables I and II). hsPAP␥ has been implicated as an enzyme responsible for monoadenylation of small RNAs (14). However, the monoadenylating activity is, in contrast to the polyadenylation activity of PAPII (31,32) and hsPAP␥ (Fig. 4), not inhibited by U1A (49). The reason for this inability of inhibition is not known yet. Possibly, the U1A inhibitory effect does not occur unless multiple adenosine residues are incorporated by PAP. Another possibility is that an alternatively processed isoform of hs-PAP␥, lacking exon 22, is responsible for the more specialized monoadenylating function in vivo.
hsPAP␥ Is a Nuclear PAP-In this study we show that hs-PAP␥ (i.e. 90-kDa isoform) resides exclusively in the nucleus, whereas the 100-and 106-kDa isoforms of PAP are both nuclear and cytoplasmic (Fig. 5), in keeping with our previous cell fractionation studies (16). It has been reported that polyadenylation factors and a subset of PAP are concentrated at sites of RNA synthesis and associated with domains enriched in splicing factors (36,50). The antibody used in these studies was the monoclonal 20:14, which recognizes both hsPAP and hsPAP␥. It is tempting to speculate that the subset of PAP at sites of RNA synthesis and 3Ј-end processing is hsPAP␥, because we have observed a high degree of co-localization of hsPAP␥ with basal splicing factors (data not shown). These observations suggest that hsPAP␥ participates in the nuclear polyadenylation reaction. In support of this we have previously shown that a fraction enriched in hsPAP␥ is active both in pre-mRNA cleavage and poly(A) addition (51).
In Fig. 5 we show that the PAP␥/EGFP chimeric hsPAP␥ is imported into the nucleus. The molecular mass of this chimera is higher than the size limit for passive diffusion through the nuclear pore. This suggests an active transport mechanism (47,52). The region important for guiding hsPAP␥ to the nucleus must reside in the C-terminal region, because elimination of it (amino acids 507-736) disturbs the observed nuclear pattern of PAP␥/EGFP chimeric protein. It has been reported that the NLS 1 and 2 elements are important for efficiently directing bovine PAPI and PAPII to the nucleus using transfection experiments (46). In our experiments the NLS 1 element of hs-PAP␥ was needed for partial nuclear localization. A careful inspection of the C-terminal sequence of hsPAP␥ revealed a putative bipartite NLS, spanning amino acids 680 -714 (NLS 3, Fig. 1B). Two potential phosphorylation sites can be predicted in the C terminus of hsPAP␥, upstream and in close vicinity of the putative NLS 3. There is increasing amount of data suggesting that regulated phosphorylation is a mechanism that modulates recognition of NLSs by components of the nuclear import machinery (52). A detailed site-directed mutagenic analysis of hsPAP␥ combined with the fusion of NLS 3 to EGFP constructs would be informative to investigate the interesting possibility that phosphorylation may regulate subcellular distribution of hsPAP␥.
Phylogenetic Conservation of hsPAP␥-PAP␥ is a phylogenetically conserved vertebrate variant of PAP present already in the bony fish branch (Table III). The existence of a goldfish hsPAP␥ orthologue supports the hypothesis that gene duplication was an important event in the evolution of early vertebrates (41,42). In a newly duplicated gene, mutations are generally selectively neutral because of redundancy of genetic information (41,42). The rate between degenerative and advantageous mutations can be influenced in the gene's favor, if the probability of forming novel regulatory interactions with other genes that are evolving in parallel occurs. Only once a new function has been acquired, the duplicated paralogue will be retained in the population as an evolutionary change. The unique C-terminal region (amino acids 507-736) of hsPAP␥ could be implicated in directing a new function. It is evident that the evolutionary machinery has selected nucleotide sub-  b The comparison was done using the multiple sequence alignment mode of the ClustalX algorithmic approach. Percentage of identity compared to hsPAPII or hsPAP␥ is listed.
c The presence of a U1A-interacting region in the C-terminal end is indicated when present. stitutions that will create changes at the amino acid level (Fig.  1A). In this study we have shown that at least one of these selected functions is exclusive nuclear localization for hsPAP␥.