Uncoupling of 3 * -Phosphatase and 5 * -Kinase Functions in Budding Yeast OF SACCHAROMYCES CEREVISIAE DNA 3 -PHOSPHATASE

Polynucleotide kinase is a bifunctional enzyme containing both DNA 3 * -phosphatase and 5 * -kinase activities seemingly suited to the coupled repair of single-strand nicks in which the phosphate has remained with the 3 * -base. We show that the yeast Saccharomyces cerevisiae is able to repair transformed dephosphorylated linear plasmids by non-homologous end joining with considerable efficiency independently of the end-pro-cessing polymerase Pol4p. Homology searches and biochemical assays did not reveal a 5 * -kinase that would account for this repair, however. Instead, open reading frame YMR156C (here named TPP1 ) is shown to encode only a polynucleotide kinase-type 3 * -phosphatase. Tpp1p bears extensive similarity to the ancient L -2-halo- acid dehalogenase and DDDD phosphohydrolase super-families, but is specific for double-stranded DNA. It is present at high levels in cell extracts in a functional form and so does not represent a pseudogene. Moreover, the phosphatase-only nature of this gene is shared by Saccharomyces mikatae YMR156C and Arabidopsis thaliana K15M2.3. Repair of 3 * -phosphate and 5 * -hy-droxyl lesions is thus uncoupled in budding yeast as compared with metazoans. Repair of transformed dephosphorylated plasmids, and 5 * -hydroxyl blocking lesions

Oxidative damage to DNA can result from endogenously generated reactive oxygen species or from exposure to exogenous agents such as ionizing radiation or anticancer agents such as bleomycin and neocarzinostatin (reviewed in Ref. 1). Such damage and the enzymes involved in its repair frequently produce fragmentation of the deoxyribose sugar backbone, resulting in DNA strand breaks bearing abnormal structures at the 3Ј and 5Ј termini. These are termed "blocking lesions" because they prevent the reactions necessary to achieve final repair of the damaged strand, namely polymerization and ligation. Since both single-and double-strand lesions can occur with potential consequences that include replication failure and genomic rearrangement, the resolution of blocking lesions is of major importance in genome maintenance. Although many chemical forms are possible, important blocking lesions on 5Ј termini include hydroxyls and deoxyribose phosphates. As an example of the redundancies in end processing, deoxyribose phosphate moieties can be removed in short-patch base excision repair (BER) 1 by the lyase function of DNA polymerase ␤ (2, 3) or certain glycosylases (4,5) or in long-patch BER by flap excision and resynthesis (6). The extensive 5Ј-resection that occurs in the first steps of recombination also likely removes blocking lesions at double-strand breaks (7,8). Common blocking lesions on 3Ј termini include phosphates, ␣,␤-unsaturated aldehydes resulting from ␤-elimination reactions (9), and phosphoglycolate moieties that are the primary product of bleomycin action (10). The most potent 3Ј-processing enzymes are the apurinic-apyrimidinic endonucleases, which, in addition to cleaving strands at abasic sites, possess 3Ј-diesterase activities capable of removing most nucleotide fragments (11)(12)(13). It is again possible that 3Ј-lesions might be resolved by a more extensive degradation during recombination, e.g. by RecBCD (8).
Polynucleotide kinase (PNK) is best known due to the utility of the T4 enzyme (14) in molecular cloning, but it was demonstrated to exist in eukaryotes 30 years ago (15). Although not clearly indicated by their name, the PNK proteins studied to date bear two distinct catalytic activities, a 5Ј-kinase and a 3Ј-phosphatase (16,17). Although the precise biological role of eukaryotic PNK remains to be determined, it is clearly suited to directly reverse the two reciprocal blocking lesions that would result from a strand break with a misplaced phosphate, i.e. a 3Ј-phosphate and 5Ј-hydroxyl. Indeed, the preferred substrate of mammalian PNK is a DNA nick (18), and its enzyme activities are stimulated by interaction with the XRCC1 repair protein (19), strongly suggesting a role in BER/single-strand break repair.
As a continuation of our interest in delineating the mechanisms by which terminal damage is resolved during DNA double-strand break repair, we have been attempting to identify and characterize PNK from the yeast Saccharomyces cerevisiae, whose existence we inferred from the successful repair of transformed dephosphorylated linear plasmids. The recent cloning of human PNK/3Ј-phosphatase (hPNKP) (16,17) has greatly facilitated this by allowing homology searching against the yeast and other sequenced genomes. Surprisingly, we find that S. cerevisiae, at least one other Saccharomyces yeast, and species as distantly related as Arabidopsis thaliana contain a gene with homology to only the putative 3Ј-phosphatase por-tion of hPNKP. The S. cerevisiae protein, encoded by open reading frame (ORF) YMR156C, here named TPP1, shares many of the biochemical properties of the hPNKP 3Ј-phosphatase, but indeed is not a 5Ј-kinase. Despite the observed plasmid repair, structural comparisons and enzymatic assays failed to detect an unlinked 5Ј-kinase. Evolutionary models to explain these results are discussed in the context of alternative pathways for resolution of terminal damage during DNA repair.
Plasmid Transformation Assay-Plasmid pES26 has been previously described (23). Methods of plasmid preparation and transformation were exactly as described (22), but with the following addition. After BglII digestion, but before extraction and precipitation, an equal volume of 2ϫ calf intestinal alkaline phosphatase buffer (Roche Molecular Biochemicals) was added, followed by nothing (ligation-competent control plasmid) or 0.16 units of calf intestinal alkaline phosphatase/g of plasmid (dephosphorylated plasmid) and further incubation at 37°C for 30 min.
Multiple Sequence Alignment-BLASTP and Psi-BLAST homology searches were performed via the NCBI web server. Sequences included in the overall alignment were all hPNKP BLASTP matches from the non-redundant GenBank TM , expressed sequence tag, and sequence tagged sites data bases with E Ͻ 0.01, in addition to Trl1p, T4 PNK, and AcNPV-2. Expressed sequence tag and STS sequences were assembled into contigs and translated prior to inclusion in the alignment (details are available on request). Accession numbers for the sequences that are ostensibly complete but uncharacterized are as follows (see Table I): Mus musculus, AAF36487; Drosophila melanogaster (CG9601), AAF54229; At-1 (A. thaliana gene encoding a putative 3Ј-phosphatase), BAA97052; At-2 (A. thaliana gene encoding a putative 5Ј kinase), CAB81914; S. pombe (C23C11.04C), CAB11157; Spodoptera exigua NPV (ORF54), AAF33584; AcNPV-1 (Ac-HisP), AAA66663; and Ac-NPV-2 (A. californica PNK/ligase), AAA66716. In the case of Caenorhabditis elegans (accession number T21197), we extended the putative ORF F21D5.5 by appending both amino-and carboxyl-terminal exon translations that were not originally included. Alignments were performed with MACAW (25).
S. mikatae YMR156C Sequence-Primers were designed that corresponded to S. mikatae sequences (kindly provided by Dr. Mark Johnston) homologous to S. cerevisiae ORFs YMR154C and YMR157C (5Ј-TCCAGTTCAAAAGTAGGATTCC and 5Ј-TAGGTAAGGCCGACATCA-TC, respectively). These were used in a PCR with S. mikatae genomic DNA using the HF Advantage PCR kit (CLONTECH) according to the manufacturer's instructions. The resulting single ϳ5-kilobase pair amplified fragment was sequenced directly by the University of Michigan DNA Sequencing Core by walking from YMR157C through YMR156C and into YMR155W. The entire S. mikatae YMR156C coding sequence (accession number AF326782) was read without ambiguities, including a stop codon read clearly from two independent runs.
Extraction and Purification of Proteins from Yeast-Yeast strains expressing GST fusion proteins were isolated from the ORF array described and kindly provided (via Dr. Dennis Thiele) by Dr. Eric Phizicky and co-workers (26). Yeast cells from the YMR156C well were streaked to single colonies, and anti-GST (Santa Cruz Biotechnology) Western blotting was used to identify isolates expressing proteins whose size corresponded to GST-Tpp1p and non-recombinant GST. Purification was by glass bead lysis and salt extraction, followed by batch chromatography on glutathione-agarose (Amersham Pharmacia Biotech) as described (26), except that protein expression was induced by adding 0.1 mM CuSO 4 for 3 h prior to harvest. Typical final GST-Tpp1p dialysates derived from 50 ml of yeast culture contained 50 g/ml fusion protein in 600 l of 20 mM Tris-HCl (pH 7.5), 2 mM EDTA, 4 mM MgCl 2 , 1 mM dithiothreitol, 50 mM NaCl, and 50% (v/v) glycerol.
Crude whole-cell extracts of S. cerevisiae, S. mikatae, and S. pombe were all prepared by glass bead disruption of cells in ϳ1 cell pellet volume of 50 mM Tris-HCl (pH 7.5), 1 mM EDTA, 1 M NaCl, 10 mM MgCl 2 , 1 mM dithiothreitol, 10% glycerol, 2 g/ml aprotinin, 1 g/ml each leupeptin and pepstatin, and 1 mM phenylmethylsulfonyl fluoride, followed by centrifugation to remove cellular debris. Final extracts were diluted to 0.5 g/l protein.
Enzyme Activity Assays-Oligonucleotides with and without 3Ј-phosphates were purchased from Operon Technologies, Inc. (see Figs. 5 and 7 for sequences). Oligonucleotides were 5Ј-end-labeled with [␥-32 P]ATP using 3Ј-phosphatase-free polynucleotide kinase (Roche Molecular Biochemicals). Final substrates were prepared by annealing labeled oligonucleotides to a 2-fold molar excess of the required unlabeled strands by heating to 90°C, followed by slow cooling. Standard assays of 3Ј-phosphatase activity contained 50 fmol of DNA substrate and 10 fmol of GST-Tpp1p or 1 g of crude cellular protein in a reaction volume of 10 l such that the final buffer was 50 mM Tris-HCl (pH 7.5), 100 mM NaCl, 10 mM MgCl 2 , 1 mM dithiothreitol, and 50 g/ml bovine serum albumin. Kinase assays were similar, except they used 1 pmol of GST-Tpp1p and 25 mM NaCl and also included 100 units of T4 DNA ligase (New England Biolabs Inc.), 1 mM ATP, and, where indicated, 5 units of T4 PNK (New England Biolabs Inc.) as an internal control. After incubation at 30°C for 10 min, formamide/EDTA loading buffer was added, and samples were electrophoresed on 7 M urea and 12% polyacrylamide gels, followed by autoradiography.

Repair of 5Ј-Hydroxyl Blocking Lesions during Yeast
NHEJ-To begin to determine the impact of 5Ј-hydroxyl blocking lesions on the repair of double-strand breaks, plasmid pES26 was digested in vitro with the restriction enzyme BglII and transformed in yeast cells, both with and without pretreatment with calf intestinal alkaline phosphatase. In this well established assay (22,23), recircularization by NHEJ (7,23,27) is required for plasmid stability and thus for expression of the plasmid URA3 marker gene. Because the BglII site resides in an essential region of the ADE2 marker gene also on this plasmid, transformation to Ade ϩ further requires that repair be precise (imprecise repair yields ade2 colonies that appear red instead of white). Dephosphorylated plasmid transformed into wild-type yeast showed only a 2-3-fold decrease in Ura ϩ colony recovery as compared with ligation-competent plasmid ( Fig. 1). Moreover, there was no increase in red colony recovery, indicating that joining remained precise. Transformation by dephosphorylated plasmids was decreased 54-fold in yeast deficient in the ligase required for NHEJ (dnl4-K282R) (23), verifying that the damaged ends are not routed into another pathway. As an additional control, it was verified that calf intestinal alkaline phosphatase-treated plasmids could not be religated in vitro by T4 DNA ligase unless further treated with T4 PNK, as indicated by gel electrophoresis and a Ͼ700-fold decrease in colony recovery following bacterial transformation (data not shown).
As shown in Fig. 1A, there are at least two ways that NHEJ of 5Ј-hydroxyl lesions might be achieved. In the first, the damaged 5Ј-nucleotide is nucleolytically removed and subsequently resynthesized on the opposite side of the break. We have previously observed that the yeast PolX family polymerase Pol4p can catalyze such base addition during NHEJ (22). pol4-D367E mutant yeast, which expresses catalytically inactive Pol4p (22), showed the same pattern as the wild type, however. Importantly, some polymerization-dependent NHEJ events show only a 2-fold defect in pol4 mutants (22), so it possible that another polymerase can substitute for Pol4p at dephosphorylated lesions as well. Nonetheless, these observa-tions were consistent with the alternative model in which the 5Ј-hydroxyl lesion is directly reversed by a PNK 5Ј-kinase. We sought to identify this in the experiments described below.
Conservation and Modular Evolution of PNK 5Ј-Kinase and 3Ј-Phosphatase Domains-BLAST searches against the nonredundant GenBank TM , expressed sequence tag, and sequence tagged sites data bases using hPNKP as the query revealed a set of 22 distinct PNK-like genes from 20 different species and viruses (Table I). Among these was the S. cerevisiae ORF YMR156C, but this weaker matche (E ϭ 0.006) corresponded to only a portion of the human protein and surprisingly lacked an apparent Walker A (i.e. P-loop) motif for ATP binding (28), which would be expected for a 5Ј-kinase. Absent as matches were three genes identified from literature searching that are known or believed to encode 5Ј-kinases: TRL1 tRNA ligase from S. cerevisiae (29,30), bacteriophage T4 PNK (14), and a putative PNK from AcNPV (31).
To examine the relationship between YMR156C and the other PNK-related genes in more detail, we performed an extensive multiple sequence alignment as shown in Figs. 2-4. Two sequence motifs, the Walker A box and the phosphotransferase motif DXDX(T/V) (32), were used as a means of unambiguously identifying the 5Ј-kinase and 3Ј-phosphatase catalytic cores, respectively, as suggested by previous alignments of smaller numbers of bifunctional PNK sequences (16,17). It was first apparent that genes may contain just the 3Ј-phosphatase domain, just the 5Ј-kinase domain, or both. This was not limited to S. cerevisiae. For example, A. thaliana contains two hypothetical genes, one encoding a 3Ј-phosphatase (designated for simplicity as At-1; see "Experimental Procedures" for accession numbers) and one encoding a 5Ј-kinase (At-2). These genes are on different chromosomes, and examination of the genomic sequence surrounding At-1 did not reveal any cryptic 5Ј-kinase domain exons. In addition, in some species, there is an apparent redundancy of function. For example, Dictyostelium discoideum has two distinct putative 5Ј-kinase genes. We note, however, that these genes correspond to two structurally evident 5Ј-kinase groupings (see below), so it is possible, if not likely, that 5Ј-kinases perform distinct tasks in the cell. Finally, in genes containing both domains, each of the two possible orders are observed.
Taken together, these observations reveal a tremendous modularity in the evolution of the PNK 3Ј-phosphatase and 5Ј-kinase catalytic domains. As a result, it is now possible to assign clear borders to these domains as independent units. The extensive overall conservation can be viewed as five distinct regions of homology to the hPNKP sequence.
Regions 1 and 2 correspond to the 3Ј-phosphatase domain (Fig. 3), best understood by comparison with the bacterial histidinol phosphatase domain, a homology that was revealed by a more sensitive iterative Psi-BLAST search. This domain catalyzes one step in histidine biosynthesis and is typified by one portion of the bifunctional Escherichia coli HisB (his7) protein. Most broadly, HisB-type histidinol phosphatases and hPNKP (16,17) are part of a large superfamily of proteins containing the L-2-halo-acid dehalogenase fold (33). These proteins catalyze a wide variety of hydrolytic reactions via a covalent substrate-enzyme intermediate. More restrictively, HisB is a member of the "DDDD" superfamily of phosphohydrolases, so named due to the presence of four invariant aspartates (34). We have labeled the overlapping amino acid motifs that define these domains as DDDD motifs 1-4 in Fig. 3. Their presence in all putative PNK-related 3Ј-phosphatases, including YMR156C, very strongly indicates that these proteins form a covalent bond with their substrate via the first aspartate of DDDD motif 1.
Two additional conserved non-DDDD motifs were evident in the putative eukaryotic 3Ј-phosphatases (called motifs A and B) (Fig. 3). Motif B, the more highly conserved, has the consensus sequence SX 2 DX 2 FAX 6 FXTPEX 2 F. It is entirely absent in HisB, but present in YMR156C, and likely serves in determining substrate recognition. Motifs A and B, as well as other characteristic amino acids in the DDDD motifs (Fig. 3), thus appear to provide a signature for identifying 3Ј-phosphatases. For example, AcNPV-1 has been called Ac-HisP because the presence of HisB motifs suggested that it was a histidinol phosphatase (see GenBank TM accession number AAA66663). In fact, this protein bears the more extended homology typical of 3Ј-phosphatases. Also, our alignment in motif 3, in contrast to that previously described (35), contains the high-scoring 3Ј-phosphatase sequence RX 5 MW in T4 PNK, even though this places a threonine at the first X instead of the nearly invariant halo-acid dehalogenase domain lysine.
Regions 3-5 correspond to the 5Ј-kinase domain (Fig. 4). Region 3, which includes the Walker A motif, is strikingly conserved at non-Walker positions from bacteriophage T4 to human, as previously described (16,17). This allowed T4 PNK, but not any S. cerevisiae kinases, to be returned by the Psi-BLAST search. Region 4 is primarily responsible for delineating what appears to be two subtypes of eukaryotic 5Ј-kinases. Most notably, the Dd-2 (D. discoideum gene) and At-2 genes possess the tetrapeptide DRCN, compared with the conserved DNTN sequence in the other eukaryotic kinases. The similarity of the former to the DRNN sequence in yeast Trl1p might suggest that these proteins are involved in metabolism of RNA FIG. 1. Efficient POL4-independent NHEJ of transformed plasmids bearing dephosphorylated BglII ends. A, the schematic shows potential pathways for processing of dephosphorylated BglII ends during NHEJ. Direct reversal of the 5Ј-hydroxyl by PNK or nucleotide removal and resynthesis through the combined action of a nuclease and polymerase (nuc ϩ pol) might each occur either before or after end annealing. B, after normalization to parallel transformations with uncut plasmid to correct for small differences in plasmid uptake, ligation-competent and dephosphorylated plasmid transformation efficiencies from wild-type (wt), dnl4-K282R (dnl4), and pol4-D367E (pol4) yeast strains were plotted relative to the wild type/non-calf intestinal alkaline phosphatase (CIP)-treated combination. White circles represent white (ADE2) colonies and therefore simple religation of the BglII ends, whereas black circles represent red (ade2) colonies. Results are from four independent transformation experiments performed on three different days (mean Ϯ S.D.). rather than DNA. Region 5 contains a very highly conserved EGF tripeptide specific to the eukaryotic proteins.
Subsets of PNK-related proteins showed additional homology outside of Regions 1-5 to each other or to non-PNK proteins, further extending the observed modularity. Of particular interest is a weak homology of At-1 to both poly(ADP-ribose) polymerase and DNA ligase III. Both of these proteins interact with XRCC1 and assist in BER/single-strand break repair (36), suggesting a similar role for At-1 and this family of proteins more generally. Indeed, hPNKP has recently been shown to interact structurally and functionally with XRCC1 (19).
S. mikatae YMR156C Lacks a Fused 5Ј-Kinase Domain-Based on the above, we hypothesized that YMR156C would be functional as a 3Ј-phosphatase despite lacking the 5Ј-kinase domain. Nonetheless, it was essential to exclude the alternative possibility that YMR156C is simply a rearrangement artifact of the sequenced S. cerevisiae strain S288C or perhaps of laboratory strains more generally. We therefore sequenced the  2. PNK multiple sequence alignment. This schematic depicts the alignment of a representative subset of PNK sequences. Tall and short boxes represent aligned and unaligned ("unlinked") regions, respectively; blank regions represent gaps; and diagonal lines indicate that sequences were deleted from the figure to conserve space. Gene/species designations are abbreviated as in Table I, with a hyphenated number corresponding to the gene number. Sequence Dd-1 is shown twice because the expressed sequence tags for this single cDNA could not be assembled into an unbroken contig. AcNPV-2 and T4 PNK are shown twice because the order of the 3Ј-phosphatase (3Ј Pase) and 5Ј-kinase domains are inverted relative to the other genes, and so a simple linear alignment is not possible. Homologies to non-PNK sequences are indicated above the relevant regions. PARP, poly(ADP-ribose) polymerase; LIG3, DNA ligase III; YOR258W, yeast hypothetical protein.
orthologous gene from the wild-type yeast species S. mikatae (strain 1815), a member of the S. cerevisiae sensu stricto group (20). The S. mikatae locus exhibited synteny with S. cerevisiae chromosome 13 that extended to the centromere-proximal and distal ORFs. The encoded 245-amino acid protein was 75% identical to S. cerevisiae YMR156C and also possessed the six DDDD and 3Ј-phosphatase motifs described above, with the non-conservative substitutions all clustered in the transitions between these motifs (data not shown). Most importantly, the S. mikatae ORF ended immediately after 3Ј-phosphatase motif  Table I, with a hyphenated number corresponding to the gene number.

FIG. 4. Structural motifs of the PNK 5-kinase domain.
Selected sequence regions from the complete alignment are shown that correspond to the conserved motifs of the 5Ј-kinase domain. The depiction is the same as in Fig. 3. Additionally, the Walker A box is indicated along with the PNK-type P-loop consensus sequence (h, hydrophobic; x, any amino acid). Asterisks mark amino acid positions that delineate the three apparent 5Ј-kinase subtypes, which are separated by horizontal lines. Gene/species designations are abbreviated as in Table I, with a hyphenated number corresponding to the gene number.
B, so this species also clearly lacks a fused kinase domain. It is thus highly likely that YMR156C evolved to support a specific function in budding yeast despite lacking the 5Ј-kinase domain.
YMR156C Encodes a Double-stranded DNA 3Ј-Phosphatase-Although YMR156C possesses all of the putative 3Ј-phosphatase motifs, it is the most divergent among the eukaryotic species as determined by calculation of phylogenetic distances (data not shown). As a result, alternative phosphatase-related functions could not be ruled out based on sequence alone. Indeed, prior to this work, the only described function for YMR156C was that forced overexpression reduced cellular sensitivity to ketoconazole, a non-phosphorylated antifungal agent that acts by inhibiting ergosterol biosynthesis (37). To directly assess its biochemical properties, YMR156C was overexpressed in yeast as a GST fusion protein and purified. As shown in Fig.  5B (lane 2), the purified GST-YMR156C fraction contained no other bands visible in Coomassie Blue-stained gels as candidate 5Ј-kinase partners. The protein catalyzed the efficient removal of a 3Ј-phosphate (Fig. 5C, lane 3) from a nicked oligonucleotide substrate (Fig. 5A; similar to Ref. 38). Importantly, there was no loss of the 5Ј-32 P label, thus demonstrating the 3Ј-specificity of the phosphatase. As a control, no 3Ј-phosphatase activity was detected when non-recombinant GST was purified in parallel and similarly tested (Fig. 5C, lane 4).
As expected from the sequence alignment, repeated examination of the GST-YMR156C fusion protein did not reveal an associated 5Ј-kinase activity (see Fig. 7D, lane 3; and data not shown). Thus, we conclude that YMR156C acts exclusively as a 3Ј-phosphatase and have designated the gene TPP1 (three prime phosphatase-1). These results further indicate that there is not a 5Ј-kinase that is tightly associated and copurifies with Tpp1p. Given the sensitivity of the assays and amount of fusion protein used, it is not unreasonable to expect to have observed a copurifying activity even in the face of GST-Tpp1p overexpression. This is supported by our observation that an arraybased two-hybrid screen (39) did not identify any P-loop-containing Tpp1p-interacting proteins (data not shown).
Examination of the conditional dependence of GST-Tpp1p 3Ј-phosphatase activity revealed the following. Activity was maximal at 100 mM NaCl, whereas salt concentrations above 200 mM were inhibitory (Fig. 5D). The enzyme was active over a broad pH range (pH 6 -9; data not shown) and displayed an absolute requirement for a metal cofactor. Activity was optimal at 10 mM MgCl 2 (Fig. 5E), but the enzyme was similarly active in buffer containing 1 mM MnCl 2 (Fig. 5F). Other divalent cations, including Ca 2ϩ , Ni 2ϩ , Co 2ϩ , and Zn 2ϩ , could not support enzyme activity (data not shown). Utilization of different DNA substrates revealed that GST-Tpp1p removed 3Ј-phosphates from nicks and single-nucleotide gaps with equal efficiency (Fig. 6). Although active at blunt ends, the enzyme was inactive on single-stranded oligonucleotides, further demonstrating that it is a structure-specific DNA 3Ј-phosphatase. Thus, it is highly likely that Tpp1p acts in the repair of damaged DNA. At present, it is unclear how this pattern of biochemical activity might explain the observed suppression of ketoconazole resistance by TPP1.
Removal of 3Ј-Phosphates and 3Ј-Phosphoryl-terminated Nucleotides by Yeast Cell-free Extracts-Crude whole cell extracts were next tested to assess the constitutive levels of Tpp1p-dependent and -independent 3Ј-phosphatase activities in yeast cells. It has been previously demonstrated that Apn1p, the major apurinic-apyrimidinic endonuclease/3Ј-diesterase in S. cerevisiae, can remove 3Ј-phosphates (13). We therefore anticipated that Apn1p activity would substantially compete with Tpp1p in this experiment and so tested extracts prepared from wild-type as well as isogenic tpp1, apn1, and tpp1 apn1 yeast cells. Wild-type extract resulted in the formation of the expected 3Ј-dephosphorylated 22-mer oligonucleotide product as well as an unexpected 21-mer product corresponding to re- for 10 min at 30°C and analyzed on a sequencing gel, followed by autoradiography. Lanes 1 and 2 contain the corresponding 22-mer oligonucleotide synthesized without and with a 3Ј-phosphate, respectively. These marker lanes are repeated in D and Fig. 7A. D, salt dependence of GST-Tpp1p activity was assayed as described for C, except the NaCl concentration was varied as shown. E and F, divalent cation dependence of GST-Tpp1p activity was assayed as described for C, except the MgCl 2 and MnCl 2 concentrations were varied as shown. In F, the triangles indicate decreasing amounts of GST-Tpp1p (from left to right: 10, 4, 2, 1.3, and 1 fmol).
FIG. 6. Tpp1p is active on double-stranded DNA, but not single-stranded DNA. Oligonucleotides containing 3Ј-phosphates in different configurations were tested for their ability to act as substrates for varying amounts of GST-Tpp1p, as in the left panel of Fig. 5F. Substrates are schematized above the relevant lanes and represent 3Јphosphates at a nick, single-nucleotide gap, single-stranded DNA end, extended 5Ј-overhang, and double-stranded DNA blunt end (from left to right and from top to bottom). All were approximately equally efficient substrates, except single-stranded DNA, which was not dephosphorylated by even the highest amount of GST-Tpp1p. moval of the entire 3Ј-nucleotide at the nick (Fig. 7A, lane 3). Again unexpectedly, nucleotide removal was not observed with either of the extracts lacking Apn1p (i.e. apn1 and tpp1 apn1) ( lanes 5 and 6, respectively). These data suggest that Apn1p itself is able to remove both a 3Ј-phosphate and a nucleotide at a nick, a hypothesis that has been validated in a separate study. 2 Of primary interest here, 3Ј-phosphate removal, evidenced by disappearance of the substrate, proceeded efficiently in both the tpp1 and apn1 mutant extracts (lanes 4 and 5, respectively), but was completely absent in the tpp1 apn1 mutant extract (lane 6). Comparison of the apn1 and tpp1 apn1 extracts (lanes 5 and 6, respectively) specifically demonstrated that TPP1 contributes an abundant 3Ј-phosphatase activity.
Uncoupling of 3Ј-Phosphatase and 5Ј-Kinase Activities in Saccharomyces as Compared with Schizosaccharomyces Extracts-Finally, crude extracts of S. cerevisiae, S. mikatae, and S. pombe were prepared and assayed using nicked oligonucleotide substrates to compare the relative levels of 5Ј-kinase and 3Ј-phosphatase activities. As predicted from the above results, all extracts possessed a 3Ј-phosphatase activity, although its level was considerably lower in the S. pombe extract (Fig. 7B).
In contrast, but again predicted by the homology searches, only the S. pombe extract showed a detectable 5Ј-kinase (Fig. 7D,  lanes 4 -7) in an assay that depends on conversion of the nick to a ligatable form by 5Ј-phosphorylation (Fig. 7C) (similar to Ref. 38). Importantly, the appearance of the 47-mer product was not the result of displacement polymerization because it was dependent on addition of T4 DNA ligase (data not shown). In addition to the S. pombe 5Ј-kinase, these assay conditions also detected T4 PNK activity when it was added to the Saccharomyces extracts (Fig. 7D, lanes 9 -12). We conclude that, in contrast to their abundant 3Ј-phosphatase, Saccharomyces yeast cells do not constitutively express a detectable DNA 5Ј-kinase and likely lack one entirely, despite the initial observations with the plasmid transformation NHEJ assay. DISCUSSION Given the potentially disastrous consequences of persistently blocked DNA termini, it is not surprising that multiple mechanisms have evolved to deal with these lesions. Blocking lesions can be resolved by strand degradation mechanisms such that nucleotides must be resynthesized in excess of those that have been directly damaged. Alternatively, direct reversal can occur by removal of only the fragmented nucleotide to leave the next available 3Ј-hydroxyl or 5Ј-phosphate or by simple rephosphorylation of a 5Ј-hydroxyl. The recently described bifunctional hPNKP appears to have evolved to directly reverse both of the reciprocal lesions created by damage that results in a misplaced phosphate (16,17). The results presented here reinforce that this is not the only mechanism of dealing with such lesions, however, and that coupling is not obligatory.
Ultimately, it must be assumed that the PNK domains maintained through evolution reflect the biology of lesion repair in a given organism (Figs. 2-4 and Table I). Both 5Ј-kinase and 3Ј-phosphatase domains are present in metazoans and fission yeast in a linked fashion, indeed suggesting the potential for efficient reversal of coupled 3Ј-phosphate and 5Ј-hydroxyl lesions. In at least some plants, this gene linkage was lost or never established. Importantly, it is as yet uncertain whether all putative 5Ј-kinase genes listed in Table I will function in DNA metabolism (note especially At-2, the only 5Ј-kinase gene in A. thaliana), and so these plants may represent 3Ј-phosphatase-only species from the standpoint of DNA repair.
Budding yeast stands out as the clearest example of a eukaryotic branch that possesses only a 3Ј-phosphatase. Caution is indicated, however, given that this conclusion is based on essentially negative results, namely that no PNK 5Ј-kinase homolog is revealed by comparison searching; no 5Ј-kinase is detected in the same crude extracts that display abundant 3Ј-phosphatase activity (Fig. 7); and no apparent 5Ј-kinase has been identified as interacting with Tpp1p in copurification (Figs. 5 and 7) and two-hybrid approaches (data not shown). It is possible that another gene provides a cryptic unlinked 5Јkinase activity. For example, Trl1p possesses coupled 5Ј-kinase and ligase activities whose biological role is in tRNA processing (29,40). The inviability of trl1 mutants prevents a simple testing of the hypothesis that there is a functional overlap between tRNA processing and DNA repair, but this seems unlikely, especially since Trl1p-dependent DNA 5Ј-kinase activity was not detected in the crude extracts. An alternative search approach based on pattern matching identified two ORFs that contain the PNK-like P-loop consensus sequence (FYWVLIMC)GXP(GAS)XGKS(TSHY)(FYWVLIMC) (Fig. 4), but that have no clear function based on literature or homology searching (YFR007W and YOR262W). These ORFs have no similarity to hPNKP beyond the P-loop, however; and it is difficult to reconcile why evolution would have resulted in a clearly conserved 3Ј-phosphatase and yet a highly divergent 5Ј-kinase. Even if these or another ORF is ultimately shown to provide cryptic DNA 5Ј-kinase activity, our results nonetheless demonstrate that there has been both a biochemical and structural uncoupling of 5Ј-kinase and 3Ј-phosphatase functions in Saccharomyces yeast.
It must be emphasized that no eubacterial or archaebacterial proteins were uncovered in sequence comparisons or literature searches as containing either 3Ј-phosphatase or 5Ј-kinase domains, despite the large number of complete genomic sequences that are available. It is clear that a common DDDDtype phosphatase domain precursor was transmitted to both eukaryotes and bacteria, which, in bacteria, became histidinol phosphatase (34). In many bacteria, this was fused to create a bifunctional enzyme containing the additional activity imidazole-glycerophosphate dehydratase (e.g. E. coli HisB). Interestingly, the yeast histidinol phosphatase His2p evolved from a structurally distinct phosphatase precursor of the "PHP" type that is also found in DNA polymerases in all branches of life (41). A tremendous modularity is thus apparent in the evolution of phosphatase domains, with the ancient precursor types being variably adapted to fulfill critical cellular functions and, in some cases, fused to novel domains. In contrast, no clear bacterial counterpart to the 5Ј-kinase domain was made evident in the Psi-BLAST search, beyond the ubiquitous Walker A motif. This pattern strongly suggests that the PNK 3Ј-phosphatase domain is evolutionarily older, whereas the DNA 5Ј-kinase arose independently and was fused only after the divergence of budding yeasts (Fig. 8). An alternative hypothesis would be that the 5Ј-kinase and 3Ј-phosphatase domains were only uncommonly split and/or lost in some eukaryotic lineages from a common coupled PNK progenitor. Redundant mechanisms for 5Ј-blocking lesion resolution might have simply allowed the 5Ј-kinase domain to be lost without detriment as a result of a rare deletion.
As sequences are generated from still more budding yeasts as a result of comparative genomic efforts (42), it should be possible to determine whether the 5Ј-kinase domain is indeed absent from this lineage. At present, we favor this hypothesis in part because it is less cumbersome than a cycle of fusion and loss. Moreover, the alternative hypothesis presupposes that coupling of 5Ј-kinase and 3Ј-phosphatase functions is inherently advantageous in most species. In fact, the fragmentation of bases generates a great diversity of blocking lesions, of which coupled 3Ј-phosphates and 5Ј-hydroxyls are only a subset. For example, 3Ј-phosphate generation by Fpg entails sequential ␤and ␦-elimination reactions, leaving 3Ј-and 5Ј-phosphates at a single nucleotide gap (5). Also, cleavage of topoisomerase I-DNA covalent complexes by tyrosyl-DNA phosphodiesterase (Tdp1p in S. cerevisiae) is thought to occur after replication fork collapse caused by collision-induced double-strand breaks, and so the resulting 3Ј-phosphate and 5Ј-hydroxyl would no longer be physically linked in a single-strand nick (43). We suggest that the overall pattern of PNK conservation reflects this lack of coupling. The greatest question is whether domain linkage in fission yeast and metazoans underlies a true increase in nick repair efficiency, fulfills a biological role specific to cells that divide by fission, or was more simply a way of ensuring coinheritance once the domains each occupied a unique niche in nick repair.
In the case of 5Ј-blocking lesions, higher eukaryotes depend on DNA polymerase ␤ lyase activity to remove 5Ј-deoxyribose phosphate lesions in the rate-limiting step in BER (2,3). In contrast, the only yeast ␤-like polymerase, Pol4p, is not required for BER (44,45). Instead, polymerization function is provided primarily by DNA polymerase ␦ (45,46), presumably in conjunction with more extensive 5Ј-base loss in a long-patch repair. Similarly, budding yeast cells process double-strand breaks almost exclusively by recombination mechanisms that entail highly efficient 5Ј-degradation, whereas end-preserving NHEJ plays a far greater role in higher eukaryotes (7,27). We argue that this emphasis on more extensive 5Ј-degradation also applies to 5Ј-hydroxyl lesions in S. cerevisiae. As a result, the need for a 5Ј-kinase, just as for the 5Ј-deoxyribose-phosphate lyase, does not exist. The observed repair of dephosphorylated plasmids would thus occur by base removal and resynthesis, which, based on our results, must occur by a Pol4p-independent pathway of end processing in NHEJ. Because NHEJ entailing simple nucleotide gaps shows a minimal pol4 effect, this is perhaps not surprising (compare Fig. 1

and Ref. 22).
In the case of 3Ј-blocking lesions, further studies will be required to determine the specific role(s) that TPP1 has been maintained to fulfill. The observed specificity of this enzyme for 3Ј-phosphates present in double-stranded DNA strongly suggests a role in genome repair (Fig. 6). More specifically, the competitiveness of Tpp1p and the promiscuous 3Ј-diesterase Apn1p observed using cell-free extracts seems to foreshadow a complex interplay between multiple pathways of 3Ј-processing (Fig. 7). Results to be presented elsewhere indicate that several DNA repair pathways display functional overlap between Tpp1p and Apn1p. 2 In contrast, bacteria apparently require only the universally conserved 3Ј-diesterases (47). Why the seemingly redundant Tpp1p/Apn1p enzyme combination was established and maintained in budding yeast (and indeed, all eukaryotes characterized to date) while the PNK 5Ј-kinase was not will be important issues to resolve.