Mutations in a Peptidylprolyl-cis/trans-isomerase Gene Lead to a Defect in 3′-End Formation of a Pre-mRNA inSaccharomyces cerevisiae *

In a genetic screen aimed at the identification of trans-acting factors involved in mRNA 3′-end processing of budding yeast, we have previously isolated two temperature-sensitive mutants with an apparent defect in the 3′-end formation of a plasmid-derived pre-mRNA. Surprisingly, both mutants were rescued by the essential gene ESS1/PTF1 that encoded a putative peptidylprolyl-cis/trans-isomerase (PPIase) (Hani, J., Stumpf, G., and Domdey, H. (1995) FEBS Lett. 365, 198–202). Such enzymes, which catalyze thecis/trans-interconversion of peptide bonds N-terminal of prolines, are suggested to play a role in protein folding or trafficking. Here we report that Ptf1p shows PPIase activity in vitro, displaying an unusual substrate specificity for peptides with phosphorylated serine and threonine residues preceding proline. Both mutations were found to result in amino acid substitutions of highly conserved residues within the PPIase domain, causing a marked decrease in PPIase activity of the mutant enzymes. Our results are suggestive of a so far unknown involvement of a PPIase in mRNA 3′-end formation in Saccharomyces cerevisiae.

Despite intensive efforts to unravel the complex process of mRNA 3Ј-end formation in Saccharomyces cerevisiae, the list of participating factors still awaits its completion.
We have recently isolated a gene complementing the phenotype of two temperature-sensitive yeast mutants that were impaired in mRNA 3Ј-end formation. This gene, designated PTF1 (processing/termination factor 1; identical with the previously described ESS1 (1)), encodes a protein that, by virtue of sequence similarity, was identified as a peptidylprolyl-cis/ trans-isomerase (PPIase) 1 (2). PPIases are ubiquitous enzymes that catalyze the interconversion from cis to trans of peptide bonds preceding a proline and are thought to accelerate this often rate-limiting step in the folding of a number of proteins in vivo (3)(4)(5)(6).
PPIases are divided in three families, based on their sensitivities toward two clinically relevant immunosuppressants: the cyclosporin A-binding proteins (cyclophilins), the FK506binding proteins, and a third family, named after the Escherichia coli protein parvulin, which is not inhibited by either of the two drugs (for review see Refs. [3][4][5][6]. In addition, the members of each family are characterized by conserved but distinct amino acid domains. By this criterion, PTF1 was predicted to belong to the parvulin family of PPIases (2).
Although disruption of PPIase genes did not generally impair cell growth (7)(8), PTF1 was the first PPIase gene shown to be essential for cell viability (1). In fact, PTF1 is the only essential PPIase gene in S. cerevisiae as demonstrated more recently by the viability of a yeast mutant lacking the remaining 12 PPIases, members of the other two immunosuppressant binding families. (8). So far, the only other example of an essential PPIase is the recently discovered PIN1, a human protein, that is structurally and functionally related to Ptf1p (9 -10).
In this paper we describe the genetic screen that led to the isolation of PTF1 and the phenotypes of the two temperaturesensitive strains carrying mutations in this gene. Moreover, we demonstrate that Ptf1p displays PPIase activity in vitro and that this activity is drastically reduced in the two mutant PTF1 proteins isolated at nonpermissive temperatures. The intriguing observation that yeast cells harboring mutant PTF1 proteins are defective in the 3Ј-end formation of a plasmid-encoded pre-mRNA invites speculations on the possible role of a PPIase in mRNA 3Ј-end processing and/or transcription termination in budding yeast.
Plasmid Constructions-The screening plasmid pJH702CEN was constructed with plasmid pPZ[LEU2] (15) as basis vector in which the TRP1 gene was replaced by the LEU2 gene. In the single BamHI site of pPZ[LEU2] the following fragments were inserted: a 730-bp long BamHI/BglII fragment containing the ACT1 promoter plus flanking sequences, a modified ADH1 terminator without stop codons, and a 3073-bp BamHI fragment from pMC1871 (Pharmacia, Freiburg, Germany) containing the complete lacZ gene. The modified ADH1 terminator consisted of a 97-mer synthetic DNA fragment with bases exchanged at the three stop codons to provide an open reading frame through the whole DNA fragment (see Fig. 4).
The control plasmid pJH712CEN contained, instead of the modified ADH1 terminator sequence, a 28-bp synthetic DNA fragment connecting in frame the ACT1 sequence with the lacZ gene (Fig. 4).
For the generation of pGALPTF1, which contains the PTF1 gene under control of the GAL1 promoter, the EcoRI fragment of YIpH-GESS (1) carrying the ESS1 gene under GAL1 control was purified and inserted into pBluescript II KS Ϫ (16 -17). From this construct, a 60-ntlong fragment which still contained a part of the 5Ј non-transcribed region was deleted by site-specific mutagenesis according to Kunkel (18).
For the overexpression of a mutated parvulin gene, the corresponding DNA was amplified by PCR from pSEP38 (19) with the primer MP1 (5Ј-GCA GGA TCC GAT GAC GAT GAC AAA GCA AAA ACA GCA GCA GCA C-3Ј) which contains an enterokinase cutting site and a BamHI restriction site and the primer MP3 (5Ј-CGG GCG AGC TCG GTA AAG CTA-3Ј) which contains a SacI restriction site. The resulting 532-bp fragment was cloned into pKSII Ϫ . Site-specific mutagenesis was carried out as described by Kunkel (18) with the oligonucleotide primers MP48 (5Ј-CAG GCA AAC GCG GCG ATG ATT TAG GTG AAT TCC-3Ј) and MP83 (5Ј-GCA CAC CCA GTT CTC ATA TCA CAT CAT TAA G-3Ј). The mutagenesis led to amino acid exchanges at position 48 from Gly to Asp (corresponding to the mutant Ptf1p-2) and at position 83 from Gly to Ser (corresponding to the mutant Ptf1p-5). The mutated DNA fragments were inserted into pQE30 which harbors a His 6 -tag coding sequence.
EMS Mutagenesis and Tetrad Analysis of Yeast Cells-EMS mutagenesis was performed as described by Lawrence (21). The production of diploid yeast cells, the sporulation analysis, and tetrad analysis were done according to Haber and Halvorson (22) and Sherman and Hicks (23).
RNA Analysis-Overnight cultures of YPM2 and of DH484 (200 ml), transformed with the selection plasmid pJH702CEN, were grown in selective medium to an A 600 ϭ 1 at 23°C. A 100-ml aliquot was removed for RNA preparation, and the remaining 100 ml were mixed with the same amount of medium prewarmed to 50°C. Incubation was continued at 37°C for 6 h when another 100 ml were removed for RNA preparation. The last aliquot of 100 ml was taken after another 13 h. Total RNA was prepared from the collected yeast cells by the hot phenol method as described by Köhrer and Domdey (24). Poly(A) ϩ RNA was isolated with an mRNA purification kit purchased from Amersham Pharmacia Biotech. Then all RNA aliquots were treated with RNasefree DNase to remove traces of remaining DNA.
For Northern blot analysis, 20 g of glyoxal-treated total RNA was separated on a 1.5% agarose gel and transferred to a Hybond N nylon membrane (Amersham Pharmacia Biotech).
Production of Recombinant Ptf1p-The coding region of PTF1 was amplified by PCR with the synthetic oligonucleotides JH34 (5Ј-AGG AAC ATA TGC CAT CTG ACG TAG CAT CG-3Ј) and GS2 (5Ј-AGG AAG GAT CCG AGG TGG AGA AGC AAA TGC C-3Ј) and inserted into the E. coli expression vector pET15b (25)(26) which includes the codons for an oligohistidine tag (Novagen). The expressed protein therefore contained an oligohistidine tag at its N terminus, followed by a thrombin recognition site. For the production of authentic Ptf1p without the oligohistidine tag, the E. coli expression vector pET11d was used (Novagen).
For the production of the mutated proteins Ptf1-2p (corresponding to the mutant Ptf1p from YPM2) and Ptf1-5p (corresponding to the mutant Ptf1p from YPM5), the PCR-amplified construct was modified by site-specific mutagenesis according to Kunkel (18) with the synthetic oligonucleotides JH37 (5Ј-TCT CCC GAA CCA GCC TAG GTC GTC GCC TCG CTT GTA TG-3Ј) for production of Ptf1-2p and JH38 (5Ј-ACC TAC CCG CTT GAT TAC ATG AAC ACT GCT TCC TGA TTC AAC-3Ј) for production of Ptf1-5p.
E. coli BL21(DE3) (Novagen) cells harboring the expression plasmids were grown at room temperature to an A 600 of 0.6 and then induced with 4 mM isopropyl-1-thio-b-D-galactopyranoside at 37°C for another 4 h.
Protein Purification-For the purification of the oligohistidinetagged fusion proteins, cells were harvested by centrifugation, resuspended in 2 mM Tris buffer, pH 8.0, and ruptured in an SLM Aminco French pressure cell. The cell lysate was stirred with 0.1% (v/v) Benzonase for 15 min at 4°C and centrifuged at 20,000 ϫ g for 30 min at 4°C. The supernatant was applied to a Ni-nitrilotriacetic acid column (1 ϫ 4 cm, Qiagen), equilibrated with 2 mM Tris buffer, pH 8.0. The column was washed with 100 ml of equilibration buffer to remove unbound protein. Bound protein was eluted by a linear gradient from 0 to 0.5 M imidazole in 60 ml of 2 mM Tris buffer, pH 8.0. The Ptf1pcontaining fractions were detected by Coomassie stained SDS-polyacrylamide gel electrophoresis and concentrated with a Filtron OME-GACELL, exclusion size 10,000 Da. Samples (1 ml) were applied to a Superdex 75 gel filtration column (1.6 ϫ 60 cm, Amersham Pharmacia Biotech), equilibrated with 10 mM HEPES buffer, pH 7.8, containing 150 mM KCl, 1.5 mM MgCl 2 , and 0.5 mM dithioerythritol. The flow rate was 0.8 ml/min. Fractions containing Ptf1p were pooled, and protein was dialyzed overnight against 3 liters of 10 mM HEPES buffer, pH 7.5, at 6°C. Ion exchange chromatography was performed for further protein purification using a Fractogel EMD SO 3 Ϫ 650(M) column (1 ϫ 6 cm) (Merck). The column was equilibrated with 10 mM HEPES buffer, pH 7.5. Protein was applied to the column at a flow rate of 1.5 ml/min, after which the column was washed with 100 ml of equilibration buffer. PPIase-containing fractions were obtained by running a linear gradient from 0 to 1 M KCl in 100 ml of 35 mM HEPES buffer, pH 7.5. The fractions were pooled and dialyzed for 3 h against 2 liters of 10 mM HEPES buffer, pH 7.5, and applied to a Fractogel TSK AF-Blue column (1 ϫ 6 cm) equilibrated with dialysis buffer. Homogenous Ptf1p was obtained by running a linear gradient from 0 to 2 M KCl in 60 ml of 10 mM HEPES buffer, pH 7.5.
For the purification of authentic heterologously expressed Ptf1p, the same procedure was used with the exception of the initial chromatographic step. The supernatant derived from ultracentrifugation was passed through a Fractogel EMD DEAE-650(M) column (2.5 ϫ 20), equilibrated with 2 mM Tris buffer, pH 8.0. Ptf1p passed unbound through the column and was applied to gel filtration as described above.
For the purification of the mutated recombinant parvulins, the harvested cells were resuspended in 2 mM Tris buffer, pH 8.0, containing a protease inhibitor mix as recommended by the manufacturer (Complete Protease Inhibitor Mixture Tablets, Boehringer Mannheim), lysed by French press, and ultracentrifuged as described for Ptf1p. The supernatant was applied to a Ni-nitrilotriacetic acid column, and His-tagged proteins were purified as described for Ptf1p. The eluted parvulincontaining fractions were dialyzed two times for 1 h against 3 liters of 10 mM HEPES buffer, pH 7.5, containing 1 mM dithiothreitol and the protease inhibitor mix. Ion exchange chromatography using a Fractogel EMD SO 3 Ϫ 650(M) column (1 ϫ 6 cm) (Merck) was carried out for further purification as described above. The recombinant proteins were eluted with a linear gradient from 0 to 2 M KCl in 10 mM HEPES buffer, pH 7.5.
Production of Polyclonal Antibodies Against Recombinant Ptf1p-Recombinant purified Ptf1p expressed in E. coli was used for production of polyclonal antibodies in rabbits. The immunization was done according to the instructions supplied with the Ribi Adjuvant System.
Western Blot Analysis-SDS-polyacrylamide gel electrophoresis was performed according to Laemmli (27). The protein was transferred onto nitrocellulose according to Haid and Suissa (28). Immunoreactions were carried out with the ECL Western blotting kit from Amersham Buchler.
PPIase Assay-For the proteolytic assay, measurements were carried out as described previously (29), using Suc-Ala-Xaa-Pro-Yaa-4NA (where NA is nitroanilide) as enzyme substrate. Xaa and Yaa denote variable aminoacyl residues at this position.
With respect to the prolyl bond, such peptide substrates exist in an equilibrium of about 5-20% cis-and 80 -95% trans-confomers. Commonly known endopeptidases like chymotrypsin, trypsin, or subtilisin split off the C-terminal 4NA residue only in the trans population of these proline-containing substrates. Thus, in the presence of sufficient amounts of protease in the reaction mixture, the trans population is rapidly cleaved, whereas the cis population remains intact (first rapid phase). The following slow isomerization reaction is accelerated by PPIases, resulting in the cleavable trans substrate (second, slow phase). The time course of 4-nitroaniline release was determined by monitoring the absorbance at 390 nm in a Hewlett-Packard 8452 diode array UV-visible spectrophotometer in 0.5-s intervals for a total of 4 min. Reported data are given as the mean value of three to five measurements.
A disadvantage of the protease-coupled assay is the requirement for high concentrations of helper proteases in order to obtain the two-phase reaction described above. This assay could therefore only be used for proteins that proved to be resistant toward attack from these enzymes, at least for the duration of the experiment. The intactness of each putative PPIase after the reaction was examined on Western blots. In the case of recombinant Ptf1p, thrombin and trypsin cleaved off the oligohistidine tag at the inserted thrombin cleavage site but did not further digest Ptf1p. Subtilisin did not attack the protein under the conditions of PPIase measurements, whereas chymotrypsin digested Ptf1p completely within 2 min of incubation (2). Thus, in the PPIase assay of Ptf1p, trypsin (0.08 and 0.02 mg/ml) was used for cleavage of the Lys-4NA and Arg-4NA bonds, respectively, and subtilisin (0.04 mg/ml), for cleaving the Phe-, Tyr-Leu-and Met-4NA bonds. Substrates were purchased from Bachem (Heidelberg) or synthesized according to Schutkowski et al. (30). Stock solutions of various substrates were prepared in dimethyl sulfoxide.
For inhibition studies, the Suc-Ala-Phe-Pro-Phe-4NA was used as substrate for parvulin and Suc-Ala-Ala-Pro-Arg-4NA as substrate for Ptf1p. FK506 was a gift from Fujisawa Pharmaceutical Co., Osaka. Stock solutions of the inhibitors were prepared in 50% ethanol. The incubation time was 5 min for FK506 and 15 min for cyclosporin A. Three independent experiments were performed.
Protease-free Assay-Because of the unexpected sensitivity of the mutant proteins Ptf1-2p and -5p toward proteases, PPIase activity was measured using a modification of the assay described by Kofron et al. (31) and others (32). 2 This assay exploits the difference in the absorption coefficients of cis and trans conformers of the substrates Suc-Ala Ala Pro-(NO 2 )Tyr-4-fluoranilide at 430 nm.

Identification and Characterization of Yeast Mutants Defective in Pre-mRNA 3Ј-End Formation-In an effort to identify
trans-acting factors participating in the 3Ј-end processing reaction of yeast pre-mRNAs, a genetic selection system was established, in which a defect in this reaction could be recognized by the appearance of blue-colored colonies. The system was based on a fusion construct, pJH702CEN, composed of a 3Ј-end truncated ACT1 gene joined to the lacZ gene by a minimum-sized ADH1 3Ј-end formation site in a centromere vector (Fig. 1). The ADH1-derived sequences had been modified such that an open reading frame was maintained throughout the originally noncoding 3Ј-region of the ADH1 gene and that translation of read-through transcripts would result in a lacZ fusion protein. Stable read-through transcripts were expected to occur only in plasmid-carrying cells in which the formation of mRNA 3Ј-ends was impaired.
Transformation of wild type yeast with pJH702CEN ( Fig. 1) yielded, as expected, white colonies on 5-bromo-4-chloro-3-indolyl b-D-galactopyranoside-containing medium, as the plasmid-derived transcripts were processed and polyadenylated at the ADH1 3Ј-end formation site. In contrast, a positive control construct (pJH712CEN, Fig. 1), in which the 3Ј-end truncated ACT1 gene was directly connected with the lacZ gene, gave rise to blue colonies on 5-bromo-4-chloro-3-indolyl b-D-galactopyranoside containing selective medium (data not shown).
After EMS mutagenesis of 1.9 million yeast cells containing the selection plasmid pJH702CEN, three viable mutants (yeast processing mutants YPM2, YPM3, and YPM5) that displayed the expected blue-color phenotype were isolated at 23°C. Two of them, YPM2 and YPM5, additionally showed temperaturesensitive growth at 30 and 37°C, respectively. The results of several control experiments demonstrated that the mutations leading to the specific phenotypes were chromosomal mutations (data not shown) as follows. (i) The plasmids isolated from the mutants and retransformed into wild type yeast led to a wild type phenotype, i.e. white colonies on 5-bromo-4-chloro-3indolyl b-D-galactopyranoside-containing plates. (ii) The DNA sequence of the ADH1 3Ј-end formation site and the beginning of the lacZ gene was not altered in the mutants. (iii) Mutants, which had been grown on non-selective medium until they had lost their plasmid, turned blue again after re-transformation with the original selection plasmid.
Shifting the YPM2-mutant cells from 23°C to the nonpermissive temperature of 37°C led to a significant decrease of the amount of poly(A) ϩ RNA within 2 h after the shift (Fig. 2).
To demonstrate that the blue phenotype of the mutants coincided with the presence of the chimeric ACT1-ADH1-lacZ read-through transcripts in the mutant cells, RT-PCR was performed with total RNA isolated from the mutant YPM2 and wild type yeast grown at 23 and 37°C, respectively. As a positive control, RNA from wild type yeast DH484, which had been transformed with the test plasmid, was isolated under identical conditions. Two sets of primer pairs were used as follows: the primers JH24 and JH25 amplified all ACT1-ADH1-lacZ-derived transcripts (Fig. 3A, product I), whereas the primers JH26 and JH27 were designed to amplify only those transcripts that had not been cleaved at the ADH1-derived polyadenylation site (Fig. 3A, product II). The radioactively labeled RT-PCR products, separated on agarose gels (Fig. 3B) were quantitated by scanning with a PhosphorImager. In wild type cells, read-through transcripts (Fig. 3A, product II) comprised only 2.5 and 1.5% of the total amount of plasmid-encoded ACT1 transcripts (Fig. 3A, product I) at 23°C and 37%, respectively. In contrast, 10-fold higher levels of read-through transcripts, i.e. 30%, were detected in the mutant cells at 37°C, with 13% already present at the permissive temperature of 23°C.
In order to test whether the observed increase in the amount of read-through transcripts from the selection plasmid in YPM2 was restricted to the special template, the transcripts of three other yeast genes, ACT1, CYC1, and YPT1, were examined on Northern blots. To facilitate the analysis, YPM2 cells were transformed with plasmids containing 3Ј-terminal fragments of these genes inserted between the ACT1 promoter and the terminator of either ACT1 or ADH1 (Fig. 4 (20)). As a control, ADH1 was included with either a complete or truncated version of its 3Ј-terminal region (Fig. 4). The latter was very similar to the one used in the mutant screen.
The plasmid constructs and positions of the oligonucleotide probes SH16 and SH18 are shown schematically in Fig. 4. The two probes enabled the distinction between transcripts ending in the inserted fragment and unprocessed products ending within the downstream adjacent ADH1 or ACT1 terminal region.
Additionally, both probes could also detect endogenous ACT1 mRNA. The corresponding signal was used as a standard for the amount of RNA isolated from YPM2 cells that differed only in the DNA constructs with which they had been transformed. Northern blot analysis was carried out with total RNA from cells that had been grown at either the permissive or the nonpermissive temperature (Fig. 5). The same blot was successively probed with SH16 (Fig. 5A) and SH18 (Fig. 5B).
Hybridization performed with SH16 and RNA that had been synthesized at 23°C resulted mostly in strong signals, corresponding well in size to transcripts that had initiated at the plasmid-derived ACT1 promoter and ended at the polyadenyl- ation sites within the respective, gene-specific 3Ј-end insertions (Fig. 5A, lanes 1-5, and legend). In each lane, the endogenous ACT1 mRNA appeared above those signals. As judged by this internal standard, the amount of total RNA in lanes 3 and 8 had apparently been underestimated, resulting in only weak signals for the YPT1 transcript at both temperatures.
Hybridization of RNA isolated from cells cultured at 37°C with the same probe led to generally less intense signals for each of the gene-specific transcripts and also for the endogenous ACT1 mRNA (Fig. 5A, lanes 6 -10). As the amount of total RNA loaded was kept the same, this result reflected again the significant reduction of poly(A) ϩ RNA in mutant cells grown at the non-permissive temperature (see Fig. 2).
Any read-through transcripts that extended into the second 3Ј-terminal insertion of each DNA construct was expected to hybridize also to SH18 (Fig. 4), as this probe, in addition to endogenous ACT1 mRNA, could only light up transcripts that had failed to be processed within the gene-specific 3Ј-end insertions. To facilitate the evaluation, the blot in Fig. 5B was drastically overexposed as exemplified by the increased intensity of the internal ACT1 standard.
The presence of a disproportionally strong signal at the position of the endogenous ACT1 mRNA in lane 8 of Fig. 5B clearly demonstrated that the YPT1 transcript, present in mutant cells grown at 37°C, was a genuine read-through transcript. A similarly clear result was obtained for the ADH1 transcript expressed from the DNA construct carrying the truncated version of the ADH1 3Ј-terminal element; a readthrough transcript was visible at 37°C (lane 9) and at 23°C (lane 4). Interestingly, the presence of the complete ADH1 3Ј-end formation site totally suppressed the formation of longer transcripts which seems also to be true for the ACT1 and CYC1 constructs. The shorter RNA species, detected in the overexposed autoradiogram of Fig. 5B, lanes 1 and 3, and 6 and 8, are presumably transcripts initiating at cryptic promoter sites downstream of the ACT1 promoter.
To establish whether the mutant phenotype could also morphologically be distinguished from wild type cells, mutant and wild type protoplasts were inspected under the microscope. This analysis revealed that protoplasts of the mutant YPM2, maintained at the non-permissive temperature for 10 h, had about double the diameter of wild type protoplasts (data not shown).
In addition, immunofluorescence microscopy with ␤-tubulin antibodies showed that the spindles in YPM2 cells were very large, compared with the ones seen in the wild type and occurred also in cells that did not show budding (data not shown). This observation indicates that the mutants were arrested in the late anaphase I.
Consistent with the observed lethal phenotype of PTF1 mutants, tetrad analysis of diploid YPM2/DBY874 yeast reproducibly yielded a 2:2 co-segregation of the expression of ␤-galactosidase together with the temperature sensitivity (data not shown).

Identification of the PTF1 Locus and Characterization of Wild Type and Mutant
Genes-Complementation of the temperature sensitivity of both YPM2 and YPM5 with two different genomic libraries resulted in the cloning and isolation of the same genomic region and the identification of PTF1 (2). Its predicted gene product is characterized by a PPIase domain with homology to E. coli parvulin and a WW domain (W denotes the invariant tryptophans), which is assumed to mediate protein-protein interactions (34).
The mutant PTF1 gene locus that caused the temperature sensitivity was isolated by PCR amplification of genomic DNA from the mutants. The amplified products were either sequenced directly or after cloning in a Bluescript vector. In YPM2, a single point mutation led to a change from the glycine residue at position 127 to aspartic acid. The same was true for YPM5, where again a single nucleotide difference caused an amino acid change, this time from a glycine at position 163 to a serine. On the DNA level, both mutations resulted from transitions of a guanine to adenine. More importantly, both mutations were located within the highly conserved regions of PPIases (Fig. 6).
Coexpression of wild type Ptf1p in each mutant strain restored cell growth at non-permissive temperatures. Moreover, in contrast to results from a previous publication (1), overexpression of Ptf1p in wild type yeast did not lead to cell death (data not shown).
Western Blot Analysis and Mass Determination of Ptf1p-A polyclonal antiserum raised against denatured Ptf1p was obtained from rabbit by immunization with the recombinant Histagged protein that was described previously (2). Using this antiserum, one of two signals, corresponding to proteins of 70 (Fig. 7, lanes 4 and 6) and 23 kDa (Fig. 7, lanes 3 and 5), respectively, appeared on Western blots prepared with proteins contained in yeast whole cell extracts or fractions thereof. The appearance of either species depended on the method of extract preparation. In extracts, prepared with glass beads, a single species of 70 kDa was identified, which is in obvious contrast to the predicted molecular mass of Ptf1p (19,404 Da). This species survived even the most stringent denaturation protocol applied to the protein sample prior to gel electrophoresis. However, when the extract was prepared in the same manner as for the in vitro 3Ј-end processing reactions (35), the only signal visible matched the expected molecular mass of Ptf1p (about 23 kDa). Surprisingly, after fractionation of the same extract with 40% ammonium sulfate, the 23-kDa species was found in the supernatant, whereas the 70-kDa signal appeared in the pellet. So far, we can only hypothesize about the nature of the latter protein species. It might represent Ptf1p in an unusually stable multimeric state or in tight complex with another protein, possibly its target. A third possible explanation that the observed high molecular weight protein resulted from cross-reaction with the Ptf1p antiserum can be largely excluded, since in a computer search we did not detect any Ptf1p-related sequences within the complete yeast genome (36).
The molecular mass of the recombinant proteins was analyzed by electrospray mass spectrometry. The resulting values of 21,438 Da for the oligohistidine-tagged protein and 19,263 Da for the one without tag agree well with the predicted sizes and provide no indication for a post-translational modification in E. coli besides removal of the starting Met residue.
PPIase Activity of Recombinant Ptf1p-A protease-coupled assay (see Refs. 37 and 38 and "Experimental Procedures") was employed to determine the PPIase activity of wild type Ptf1p isolated from E. coli cells that expressed the recombinant, His-tagged protein. cis/trans-Isomerization of the proline bond in synthetic tetrapeptide substrates was measured spectroscopically, and activity was expressed as k cat /K M .
The values for PPIase activity of Ptf1p and its substrate specificity are summarized in Table I. Using the standard PPIase substrate Suc-Ala-Ala-Pro-Phe-4NA, only moderate enzymatic activity was detected (k cat /K M ϭ 5.9 ϫ 10 3 M Ϫ1 s Ϫ1 ). Using a series of peptides, which differed in the residues flanking a single proline, the highest specificity constant was obtained with Suc-Ala-Glu-Pro-Phe-4NA (k cat /K M ϭ 4.2 ϫ 10 6 M Ϫ1 s Ϫ1 ). This almost 1000-fold higher enzymatic activity of Ptf1p toward the latter substrate in which glutamic acid, a phosphorylated serine surrogate, preceded proline, led us to investigate peptides containing phosphorylated Ser, Thr, or Tyr residues in the equivalent position. As depicted in Table I, the enzyme showed the highest increase in activity over the standard value when a phosphorylated serine (k cat /K M ϭ 1.7 ϫ 10 7 M Ϫ1 s Ϫ1 ) or, to a lesser extent, a phosphorylated threonine (k cat /K M ϭ 2 ϫ Fig. 4 and that had been grown at either 23°C or shifted to 37°C. In each lane, 10 g of total RNA of YPM2 cells were separated. M, DNA length standard digested with HindIII. A, hybridization with oligonucleotide SH16 (see Fig. 4). B, hybridization with oligonucleotide SH18 (see Fig. 4).  Other PPIases like human FKPB12, human CYP18, trigger factor, and more importantly, E. coli parvulin showed a decrease rather than an increase of their specificity constants when tested with the phosphorylated substrates. 3 In order to rule out that the His-tag of the recombinant protein might have affected the activity and specificity of the enzyme, an authentic His-tag-free recombinant Ptf1p was expressed in E. coli and purified to homogeneity. The integrity of the protein was confirmed by high performance liquid chromatography and mass spectrometry. No differences were found between the enzymatic activities of oligohistidine-tagged and the authentic His-tag-free recombinant Ptf1p (data not shown).

FIG. 5. Northern blot analysis of RNA transcripts from cells that had been transformed with one of the constructs shown in
PPIase Activity of Recombinant Ptf1p from the Yeast Mutants YPM2 and YPM5-As described above, one single, although distinct, amino acid change in Ptf1p led to temperature sensitivity of the yeast mutants YPM2 and YPM5 (Fig. 6). To test the PPIase activity of the mutant forms, the corresponding recombinant proteins were expressed in E. coli and isolated from cells grown at non-permissive temperature (37°C). The mutant proteins turned out to be very unstable and sensitive toward all proteases used for activity measurements. Therefore, activity was determined in a protease-free assay (see "Experimental Procedures" and Ref. 32). 2 The data in Table II demonstrate that the PPIase activity of both mutants was strongly reduced as compared with the wild type protein. Moreover, the degree of temperature sensitivity correlated directly with the relative decrease in enzymatic activities of the corresponding enzymes, i.e. Ptf1p from the mutant YPM2 (Ptf1-2p) which showed temperature sensitivity at 30°C was about half as active as Ptf1p from YPM5 (Ptf1-5p) which is still viable at 30°C.
To establish whether the mutations at the conserved Gly residues in Ptf1p affected also the activity of another PPIase within the same family, the equivalent Gly substitutions were introduced into E. coli parvulin, resulting in the parvulin mutants Gly 48 -Asp, corresponding to Ptf1-2p, and Gly 83 -Ser, comparable to Ptf1-5p. Like the Ptf1p mutants, both parvulin mutant proteins were found to be very unstable. Most notably, as shown in Table III, the presence of either mutation virtually abolished the enzymatic activity of parvulin (less than 0.1%).

DISCUSSION
The most curious aspect of PTF1, the first essential PPIase gene identified in S. cerevisiae, is the fact that it was discovered in a genetic screen aimed at the identification of factors involved in the mRNA 3Ј-end processing pathway in bakers' yeast (2). Moreover, PTF1 was the only gene found in this screen to complement the two independently isolated, EMSinduced temperature-sensitive mutants, YPM2 and YPM5, with an apparent defect in the 3Ј-end formation of a plasmidencoded pre-mRNA. Isolation and sequence determination of the PTF1 genes from both mutants revealed a single, albeit distinct, point mutation at the DNA level, each resulting in a change of highly conserved amino acid residues within the PPIase domain. These changes were accompanied by a drastic decrease in PPIase activity. Coexpression of wild type Ptf1p in   either mutant was sufficient to restore growth at non-permissive temperatures.
Our mutant screen was based on the detection of plasmidderived transcripts arising from lack of cleavage and polyadenylation at a truncated ADH1 terminator region, located between yeast actin sequences and the bacterial lacZ gene ( Fig.  1 and text). Still, with the notable exception of YPT1, no such read-through events were observed in the mutant strains, when the truncated ADH1 terminator region was replaced by its complete version, or by equivalent sequences from the ACT1 and CYC1 genes (Fig. 4).
A similar "diffuse" effect on mRNA 3Ј-end formation was reported by Russnak et al. (39). The ref2-1 mutants, which the authors identified, showed differences in mRNA 3Ј-end processing only for transcripts encoded by certain, artificially designed plasmids. By this criterion, the authors defined "weak" and "strong" polyadenylation sites, a definition which had been introduced earlier by Irniger et al. (40). Moreover it was suggested that impairment of processing efficiency should not occur without a negative effect on termination frequency in yeast cells, as in this organism, unlike in higher eucaryotes, transcription and polyadenylation are tightly coupled. In fact, it might even have been impossible to distinguish in our screen whether the primary defect had occurred in transcription termination or at the level of endonucleolytic cleavage and polyadenylation because of their expected common consequences. Any failure in transcription termination should also interfere with the transcription initiation of the gene following immediately downstream, as in yeast there is generally little space between polyadenylation sites and the transcription start site of the adjacent gene (41). Additionally, read-through transcripts should not be stable in either scenario, whether they fail to be processed at all or lack the 5Ј-cap, once they are cleaved at their respective poly(A) sites but transcription continues. The very first consequence would be a rapid decline in mRNA accumulation, which is what we observed in the YPM2 mutant at non-permissive temperatures as early as 2 h after the shift (Fig. 2, lane 6).
The multiple phenotypes of the mutant strains described before could be reversed by coexpression of a gene (PTF1 (2)) encoding a peptidylprolyl-cis/trans-isomerase. Whereas the previous identification of Ptf1p as a putative PPIase resulted from its sequence similarity with the E. coli protein parvulin (29), two eucaryotic counterparts of Ptf1p have been discovered in the meantime: DODO from Drosophila melanogaster (42) and PIN1 (9) from human cells. All three proteins are clearly distinct from the prokaryotic members of the Parvulin family of PPIases and have recently been suggested to be named products of the dodo gene family (43). In addition to their high degree of structural identity, these proteins are also functional homologues as demonstrated by the fact that the human PIN1, as well as the fly DODO, complemented the lethal phenotype of ESS1/PTF1 disrupted yeast cells (9,42). Moreover, intact PPIase activity of PIN1 was necessary for successful complementation. Yet, whereas PTF1 and PIN1 were shown to be essen-tial for growth of yeast and HeLa cells, respectively, total removal of the fly gene did not impair development of the mutant insects (42). Whether the differences in importance of these related genes in their natural host organisms reflect differences in their functions or in host cell requirements remains to be seen.
Unusual for PPIases, Ptf1p revealed a distinct substrate specificity with respect to the amino acid residues preceding the prolyl-peptide bond. The highest activity was achieved with peptides containing phosphorylated Ser/Thr moieties at this position. Most remarkably, the activity toward the optimal substrate (Ac-Ala-(PO 3 H 2 )Ser-Pro-Tyr-4NA) was enhanced up to 3000-fold over the value obtained with the standard substrate. This rather unique substrate specificity is also shared by the human PIN1 (9), very likely reflecting the nature of in vivo targets (discussed below).
In agreement with the putative function of Ptf1p as a PPIase in vivo, its mutated forms in YPM2 and YPM5 displayed significantly reduced activities in vitro compared with the wild type protein.
Taken together, our results clearly correlate a PPIase activity with efficient pre-mRNA 3Ј-end processing and/or transcription termination in S. cerevisiae. Thus, the most compelling question arising from our studies concerns the nature of the putative involvement of a PPIase in these processes. In the absence of more experimental clues, we speculate that Ptf1p interacts with components of the mRNA transcription complex, if only at the final stages of RNA transcription. Upon the appropriate trigger (or at the time of entry), Ptf1p might induce a conformational switch in either the accessory proteins or the polymerase itself, ultimately causing the dissociation of RNA polymerase II from the DNA template, i.e. transcription termination.
The discovery of a PPIase of the cyclophilin family, shown to interact with the C-terminal domain of mammalian polymerase II in a yeast two-hybrid system (44), seems to lend support to the postulated interaction of Ptf1p with the transcription machinery in yeast. More importantly, the former interaction depended on the presence of phosphoepitopes on the C-terminal domain, as extensive treatment of the yeast extracts with phosphatases resulted in complete loss of this interaction. As the C-terminal domain, which consists of multiple tandem repeats of a heptapeptide Tyr-Ser-Pro-Thr-Ser-Pro-Ser, is highly conserved between yeast and mammals, this domain represents an ideal candidate for a PPIase, with specificity toward phosphorylated serines N-terminal to prolines.
Moreover, two recent publications (45)(46), in which another cyclophilin, USA-CyP (a member of the cyclosporin A-binding PPIase family), was shown to be tightly associated with the spliceosomal (U4/U6.U5) tri-small nuclear ribonucleoprotein in HeLa cells, also support the proposed role of Ptf1p in the dissociation of the transcription machinery from the DNA template. In one of the papers (45), the authors discuss the intriguing possibility that the role of USA-CyP in the human spliceosome is not to act as a chaperone in the folding or assembly steps but rather to assist in the disassembly of spliceosomes. This view was spurred by the example of one of the best characterized cyclophilins, cyclophilin-A, for which there is convincing evidence that it exerts its function by binding specifically to the HIV-1 capsid protein (CA), destabilizing interactions between the CA molecules and thus facilitating the disassembly of the CA core (47)(48)(49). Nevertheless, as much as it is tempting to draw parallels from the above examples to the physiological function of Ptf1p, its putative role in RNA transcription (and/or termination) has first to await experimental confirmation. Finally, we should like to consider again PIN1, the functional and structural human homologue of yeast Ptf1p. When underexpressed, both proteins induced mitotic arrest of their respective host cells, suggesting a role in cell cycle regulation. In agreement with this notion, PIN1 has been found to be part of the nuclear speckle (33), a large protein complex, which contains several other mitosis-related proteins and, most notably, some splicing factors. Given the fact that PIN1 substitutes functionally for the putative pre-mRNA 3Ј-end processing factor ESS1/PTF1 in yeast cells, the presence of Pin1 in the nuclear speckle, the components of which are supposedly involved in cell cycle regulation, is rather intriguing. It may be that factors such as these serve as checkpoints for the integrity of the mRNA maturation process thus acting as a link between pre-mRNA 3Ј-end processing and cell cycle regulation.