Functional Expression of Escherichia coli Endonuclease IV in Apurinic Endonuclease-deficient Yeast*

Saccharomyces cerevisiae Apn1 and Escherichia coli endonuclease IV are homologous enzymes that initiate the repair of abasic (AP) sites or oxidative DNA strand breaks. Yeast lacking Apn1 ( apn1 (cid:50) ) are hypersensitive to simple alkylating agents (which produce many AP sites) and to oxidants and display an elevated spontaneous mutation rate due to endogenous damages. We ex- plored whether the prokaryotic repair enzyme could substitute for its yeast counterpart. Plasmid constructs were generated that expressed endonuclease IV at 1/20 to 10-fold the AP endonuclease activity of wild-type yeast; some of these plasmids expressed hybrid forms of endonuclease IV equipped with the C-terminal nuclear localization signal of Apn1. Although hybrid endonuclease IV-Apn1 (but not native endonuclease IV) was selec- tively localized to the yeast nucleus, expression of this chimeric protein at 25% of the normal Apn1 level did not restore alkylation or oxidant resistance to apn1 (cid:50) yeast, but it did partially counteract the mutator phenotype of apn1 (cid:50) yeast. Expression of either the hybrid protein or native endonuclease IV at (cid:59) 10 times wild-type Apn1 levels restored wild-type resistance to methyl methanesul- fonate and near-wild-type H 2 0 2 resistance. High level expression of native endonuclease IV also restored the normal spontaneous mutation rate to apn1 (cid:50)

Saccharomyces cerevisiae Apn1 and Escherichia coli endonuclease IV are homologous enzymes that initiate the repair of abasic (AP) sites or oxidative DNA strand breaks. Yeast lacking Apn1 (apn1 ؊ ) are hypersensitive to simple alkylating agents (which produce many AP sites) and to oxidants and display an elevated spontaneous mutation rate due to endogenous damages. We explored whether the prokaryotic repair enzyme could substitute for its yeast counterpart. Plasmid constructs were generated that expressed endonuclease IV at 1/20 to 10-fold the AP endonuclease activity of wild-type yeast; some of these plasmids expressed hybrid forms of endonuclease IV equipped with the C-terminal nuclear localization signal of Apn1. Although hybrid endonuclease IV-Apn1 (but not native endonuclease IV) was selectively localized to the yeast nucleus, expression of this chimeric protein at 25% of the normal Apn1 level did not restore alkylation or oxidant resistance to apn1 ؊ yeast, but it did partially counteract the mutator phenotype of apn1 ؊ yeast. Expression of either the hybrid protein or native endonuclease IV at ϳ10 times wild-type Apn1 levels restored wild-type resistance to methyl methanesulfonate and near-wild-type H 2 0 2 resistance. High level expression of native endonuclease IV also restored the normal spontaneous mutation rate to apn1 ؊ yeast. These data place limits on the amounts of AP endonuclease activity necessary for repair of DNA damages caused by both endogenous and environmental agents and point to a direct role of spontaneous AP sites as potentially mutagenic lesions.
Ionizing radiation, chemical oxidants, and aerobic metabolism produce toxic oxygen derivatives such as superoxide, hydrogen peroxide, and hydroxyl radical (von Sonntag, 1987;Imlay and Linn, 1988). These reactive oxygen species have deleterious effects, in part by producing oxidative DNA strand breaks bearing blocked 3Ј-termini and various types of oxidized abasic (AP) 1 sites (Hutchinson, 1985;von Sonntag, 1987;Demple et al., 1986). Nonoxidized AP sites are also formed by simple alkylating agents and spontaneous base hydrolysis (Lindahl, 1993). Cells must repair oxidative strand breaks to allow survival (Demple et al., 1986;Demple and Harrison, 1994) and AP sites for survival and to prevent mutagenic errors during DNA replication (Loeb and Preston, 1986).
The repair of AP sites is initiated by class II AP endonucleases, which make incisions on the immediate 5Ј side of abasic sites (Warner et al., 1980;Demple and Harrison, 1994). Many class II AP endonucleases also possess 3Ј-repair diesterase activity that selectively removes deoxyribose fragments such as phosphoglycolate esters from the 3Ј-termini of oxidatively damaged DNA (Henner et al., 1983;Johnson and Demple, 1988ab;Levin et al., 1988;Ramotar et al., 1991b;Demple and Harrison, 1994). Both enzymatic activities, class II AP endonuclease and 3Ј-diesterase, yield free 3Ј-hydroxyl groups that support DNA repair synthesis. In Escherichia coli, exonuclease III and endonuclease IV are the main AP endonucleases that function in cellular resistance to oxidants and alkylating agents and in limiting spontaneous mutagenesis (Cunningham et al., 1986;Demple et al., 1986;Levin et al., 1988;Kunz et al. 1994). In Saccharomyces cerevisiae, these functions are effected by a single enzyme, Apn1 (Ramotar et al., 1991b;Kunz et al., 1994), which is homologous to E. coli endonuclease IV (Popoff et al., 1990). Related genes have been identified in Mycobacterium leprae (Honore et al., 1993), Mycoplasm genitalium (Fraser et al., 1995), and in the fission yeast Schizosaccharomyces pombe, 2 but their biological roles remain undetermined. No homolog of exonuclease III has yet been reported for S. cerevisiae, but exonuclease III-related enzymes have been found in mammalian cells Seki et al., 1991), Drosophila (Sander et al., 1991), plants (Babiychuk et al., 1994), and Gram-positive bacteria (Puyet et al., 1989).
The alkylation and oxidation resistance conferred in yeast by Apn1 is due, respectively, to efficient repair of alkylation-induced AP sites and blocked oxidative 3Ј-termini in DNA (Ramotar et al., 1991b). The mutator phenotype of Apn1-deficient (apn1-⌬1) strains seems largely due to the generation of AP sites by endogenous mutagens, with the extra mutations primarily single base pair substitutions dominated by a 60-fold increase in the rate of AT 3 CG transversions (Kunz et al., 1994).
Apn1 protein and endonuclease IV share 41% amino acid identity over a region encompassing 95% of the bacterial protein and the N-terminal 285 residues of the yeast protein (Popoff et al., 1990). In addition, Apn1 has a C-terminal segment of 82 residues including three clusters of basic amino acids (Popoff et al. (1990), see Fig. 1). The most C-terminal of these basic clusters of Apn1 is required to localize Apn1 efficiently to the yeast nucleus, but not for enzymatic activity (Ramotar et al., 1993).
In an earlier report, we demonstrated that Apn1 expressed in E. coli can substitute specifically for endonuclease IV in the repair of damaged DNA in vivo (Ramotar et al., 1991a). It thus appears that Apn1 and endonuclease IV share in vivo substrate specificity in DNA repair consistent with their homology and similar in vitro properties (Johnson and Demple, 1988b;Levin et al., 1988). We show here that native E. coli endonuclease IV can functionally substitute for Apn1 in yeast when expressed at high but not low levels and is not localized to any specific organelle. A derivative of endonuclease IV equipped with the distal two basic clusters of Apn1 accumulated in the yeast nucleus and lowered the apn1-⌬1 mutation rate even when expressed at a lower level.
Construction of plac-EndoIV and plac-EndoIV*-The plasmid pRPC124 (generously provided by Dr. R. Cunningham, Albany, NY) contains the entire coding region of the E. coli nfo gene and the flanking 5Ј-and 3Ј-untranslated DNA sequences (Cunningham et al., 1986). This plasmid was used as the template to amplify by polymerase chain reaction (PCR) (Sambrook et al., 1989) the sequence from Ϫ21 to ϩ937 (numbering with respect to the first A of the nfo start codon) using the primers DR1 (5Ј-GGGTTTAACAGGGGTACCCGCATGAAATAC-3Ј, start codon shown in bold) and DR2 (5Ј-TTTCGTTCGGCTGGATCC-GCGGGTTACGCC-3Ј) bearing the restriction sites (underlined) for KpnI and BamHI, respectively. This procedure yielded a 958-bp fragment containing the entire nfo coding region and ϳ150 base pairs (bp) of the 3Ј-untranslated region, which was digested with KpnI and BamHI and subcloned next to the lac promoter in the E. coli expression vector pKEN2 (provided by Dr. G. Verdine, Harvard University). In this construct, 808 bp of the PCR product was replaced by the corresponding fragment of nfo from pRPC124, thus generating plac-Endo IV. This strategy allowed us to confirm the nfo DNA sequence by determining only the remaining 150 bp derived from PCR. In a similar manner, plasmid plac-EndoIV* was constructed in order to accommodate the in-frame attachment of APN1 regions (see below), except that the PCR product was generated using the primers DR1 and DR3 (5Ј-TCATCT-TCAGGGATCCGCTTTTTCAGTTTG-3Ј; BamHI site underlined). Thus, EndoIV* lacks the last two amino acid residues of endonuclease IV and the 3Ј-untranslated region of the nfo gene.

Construction of Plasmids for Endonuclease IV Expression in Yeast-
Plasmids plac-EndoIV, plac-EndoIV*, plac-EndoIV 333 , and plac-En-doIV 297 were digested with KpnI and XhoI (sites located in the multiple cloning site of pKEN2) to release the DNA fragment encoding endonuclease IV and its derivatives containing Apn1 sequences. These fragments were subcloned directly next to the GAL1 promoter in the yeast expression vector pYES2.0 (Ramotar et al., 1993) to produce pGAL-EndoIV, pGAL-EndoIV*, pGAL-EndoIV 333 , and pGAL-EndoIV 297 . Like plac-EndoIV*, pGAL-EndoIV* also lacks the normal stop codon such that translation is expected to terminate within pYES2.0 sequences; this construct was predicted to encode endonuclease IV with an additional 55 amino acid residues at the C terminus.
Enzyme Assays and Activity Gels-Crude extracts were assayed for both AP endonuclease and 3Ј-diesterase activities as described by Levin et al. (1988) and Levin and Demple (1990). Activity gels were prepared as described by Bernelot-Moens and Demple (1989) as modified by Ramotar et al. (1991a).
Cellular Sensitivity Measurements-The sensitivity of yeast strains to hydrogen peroxide or to methyl methanesulfonate (MMS) was measured by standard survival curves using exponential phase cultures (Ramotar et al., 1991b) or by gradient plate assays (Cunningham et al., 1986;Ramotar et al., 1993). In the plates, the bottom layer (30 ml) contained 0.4 mmol (for E. coli) or 0.13 mmol (for yeast) of MMS. For tert-butylhydroperoxide, 3.9 mol was used in the bottom layer (30 ml) for gradient plates with E. coli.
Mutation Assay-Spontaneous mutation rates in yeast were determined using the fluctuation test described by von Borstel (1978). For each test, 24 independent 1-ml cultures were grown in 24-well plates (Costar) in SD medium (Sherman et al., 1983) with limiting concentrations of adenine (0.75 g/ml) for ade2-1 reversion or lysine (1.0 g/ml) for lys2-1 reversion. The initial cell density was typically ϳ4000 cells per ml and reached a density of ϳ1.0 ϫ 10 6 cells per ml when the adenine or lysine was exhausted. For induction of endonuclease IV, the SD medium contained 2% galactose instead of glucose. Calculation of the spontaneous mutation rates and experimental errors was also according to the formulae given by von Borstel (1978).
Immunological Methods-Rabbit polyclonal antibodies specific for endonuclease IV 3 were used at a dilution of 1:1000 in 10 mM Tris-HCl (pH 7.5), 150 mM NaCl, and 4% powdered milk (Gershoni and Palade, 1983). Ten ml of this solution was used to probe each nitrocellulose filter (8 ϫ 10 cm) blotted from SDS-polyacrylamide gels. Goat anti-rabbit IgG conjugated to horseradish peroxidase (BIO/CAN Scientific, Inc., Ontario, Canada) was used as the secondary antibody at a dilution of 1:5000. Immunoreactive polypeptides were detected using 4-chloro-1naphthol (Bio-Rad).
Indirect immunofluorescence to detect endonuclease IV expressed in yeast was performed according to Pringle et al. (1991) modified as follows. Cells were grown overnight in 1 ml of uracil omission medium (Sherman et al., 1983) containing 2% glucose; the following day, the cells were washed twice, each time with 1 ml of 20 mM potassium phosphate (pH 7.0), resuspended in 1 ml of uracil omission medium containing 2% (v/v) glycerol, and allowed to grow for 8 h at 30°C. Cells were washed as above, resuspended in 5 ml of uracil omission medium  (Popoff et al., 1990).
containing 2% galactose, and grown overnight at 30°C to permit induction of endonuclease IV. The cells were then fixed in 4.4% formaldehyde for 90 min at room temperature. After fixation, the cells were permeabilized by a 20-min incubation with zymolyase, washed twice with 1 M sorbitol, 10 mM Tris-HCl, pH 7.5, and resuspended in 50 l of the same buffer. Ten l of the resulting spheroblasts was placed on a polylysinecoated glass microscope slide and incubated overnight at room temperature with 20 l of anti-endonuclease IV antiserum at a dilution of 1:1000 in 137 mM NaCl, 2.6 mM KCl, 7.9 mM Na 2 HP0 4 , 1.4 mM KH 2 PO 4 , pH 7.3, 1 mg/ml bovine serum albumin (PBS-BSA; Pringle et al. (1991)). The cells were then washed for 30 min with 10 drops of PBS-BSA. RNase was added to 50 g/ml in a final volume of 20 l, and the incubation continued at room temperature for 30 min. Cells were then washed for 10 min with 5 drops of PBS-BSA, and the secondary antibody (fluorescein-conjugated "Affini-pure" donkey anti-rabbit IgG (Jackson ImmunoResearch Laboratories)) was added to the cells as 20 l of a 1:50 dilution, and the incubation continued for 2 h at room temperature. The cells were then washed for 1 h with 20 drops of PBS-BSA, and 20 l of 200 g/ml propidium iodide was added to each well for 10 min. Cells were then washed only twice, each time with 2 drops of PBS-BSA (excessive washing removed the propidium iodide and weakened the DNA staining). The results were documented using a Zeiss fluorescence microscope equipped with an Axiophot camera.

RESULTS
A plasmid was constructed to express E. coli endonuclease IV in yeast by placing the nfo gene behind the yeast galactoseinducible GAL1 promoter in the multicopy vector pYES2.0 (Ramotar et al., 1993). The resulting construct (pGAL-EndoIV) was transformed into the yeast strain FY86 and its apn-⌬1 derivative DRY377, and the transformants were examined for expression of AP endonuclease activity. AP endonuclease activity was not detectable in cell-free extracts of the apn1-⌬1 mutant harboring the vector pYES2.0, while the expected activity (ϳ60 units/mg) was seen in the APN1 ϩ strain for cells grown in either glucose or galactose (Table I). The plasmid pGAL-EndoIV directed the synthesis of AP endonuclease activity in the apn1-⌬1 strain at ϳ25% of the level for glucosegrown, wild-type yeast, with an induction of ϳ50-fold by galactose (Table I). Under the same conditions, plasmid pDR6 containing the APN1 ϩ gene under pGAL1 control (Ramotar et al., 1993) gave both basal and induced AP endonuclease levels in DR377 at ϳ6-fold higher than observed for pGAL-Endo IV (Table I). We do not know whether this difference reflects differences in the synthesis or the stability of endonuclease IV and Apn1 or a failure of the bacterial enzyme to be generated in the fully active form in yeast. A more remote possibility is that endonuclease IV is somehow poorly extracted from yeast cells under our conditions. The lower induced activity of endonuclease IV in yeast extracts was not due to the presence of an inhibitor, because the addition of purified endonuclease IV to apn1-⌬1 extracts did not alter the activity of the bacterial enzyme (data not shown).
The expression in yeast of active, full-length endonuclease IV was confirmed by activity gel analysis. In this approach, enzymatic cleavage of a synthetic 3Ј-[ 32 P]phosphoglycoaldehyde substrate releases the label, which then diffuses out of the gel to leave a clear band upon autoradiography (Bernelot-Moens and Demple, 1989;Ramotar et al., 1991a). Extracts prepared from the apn1-⌬1 strain failed to form a detectable band, while the presence of plasmid pGAL-EndoIV gave rise to a clear activity band (Fig. 2, lane 3) at the position expected for endonuclease IV (lane 7). Growth of this strain in galactose allowed the endonuclease IV activity band to be easily detected with 20-fold less protein (lane 3 versus 4). Samples of wild-type yeast (lane 1) or strains bearing plasmid pDR6 (lanes 5 and 6) produced bands at the position expected for Apn1 (lane 8), in proportion to the amount of AP endonuclease activity detected in the quantitative enzymatic assay (Table I).
Functional Complementation of Yeast apn1-⌬1 Mutants by E. coli Endonuclease IV-We tested whether endonuclease IV, which harbors both AP endonuclease and 3Ј-repair activities similar to Apn1, could functionally substitute for the yeast enzyme in vivo to provide cellular resistance to oxidative or alkylation damage. Under noninducing conditions (growth in glucose), the apn1-⌬1 strain DRY377 harboring pGAL-EndoIV remained hypersensitive to both MMS (Fig. 3A) and H 2 O 2 (Fig.  3C). Upon growth in galactose, however, wild-type resistance to MMS (Fig. 3B) and near-wild-type resistance to H 2 O 2 was provided by pGAL-EndoIV. The APN1 ϩ plasmid pDR6 gave wild-type MMS and H 2 O 2 resistance to the apn1-⌬1 strain even under noninducing conditions (Fig. 3, A and C), and this resistance was not further enhanced by the strong overproduction of Apn1 resulting from growth in galactose (Fig. 3, B and D). Thus, the normal, wild-type level of Apn1 is not limiting for the repair of either alkylation-induced AP sites or oxidative 3Ј-damages.
Yeast cells that lack Apn1 accumulate mutations at a high rate, as measured using two different ochre alleles, ade2-1 and lys2-1 (Ramotar et al., 1991b). Basal expression of endonuclease IV from pGAL-Endo IV in glucose failed to suppress this mutator phenotype (Table II). However, pGAL-EndoIV expressed in galactose-containing medium eliminated the excess spontaneous mutagenesis for both the ade2-1 and the lys2-1 alleles in strain DRY373 (Table II). It is noteworthy that the  substantial overproduction of either endonuclease IV (Table II) or of Apn1 from pDR6 (data not shown) did not significantly lower the spontaneous mutation rate below that of wild-type levels. However, the results clearly indicate that endonuclease IV can functionally replace Apn1 in repairing both the endogenously generated DNA lesions that potentiate spontaneous mutagenesis (Kunz et al., 1994), as well as lethal lesions produced by treatment with exogenous DNA damaging agents. Activity of Endonuclease IV-Apn1 Hybrid Proteins-Although pGAL-EndoIV expressed active endonuclease IV in yeast (Table I), this level of activity was insufficient at the basal level to substitute for Apn1 in providing resistance to the DNA-damaging agents MMS and H 2 O 2 in yeast (Fig. 3, A and  C). Since endonuclease IV bears no obvious nuclear localization signal, it seemed likely that basal expression of endonuclease IV did not yield sufficient enzyme in the yeast nucleus to effect DNA repair; this insufficiency is evidently overcome by the overproduction of endonuclease IV. Thus, if endonuclease IV were modified to achieve targeting to the yeast nucleus, basal expression of the bacterial enzyme might provide significant capacity for DNA repair.
Previously, we reported that basic cluster 1 of the Apn1 C terminus (Fig. 1) is essential for efficient targeting of the protein to the yeast nucleus (Ramotar et al., 1993). We tested whether basic cluster 1, or a combination of clusters 1 and 2, could act as nuclear transport signals for endonuclease IV. A series of plasmids was constructed (see "Materials and Methods") to express various endonuclease IV-Apn1 hybrid proteins (Fig. 1). The plasmid series plac-EndoIV, plac-EndoIV 333 , and plac-EndoIV 297 were designed for expression of these hybrid proteins in E. coli, under control of the lac promoter. When expressed even under noninducing conditions in the AP endonuclease-deficient, lacI ϩ strain BW528, all three plasmids directed the synthesis of substantial AP endonuclease activity (Table III). All three plasmids also conferred resistance in strain BW528 to both the alkylating agent MMS and the oxidant t-butyl hydroperoxide (Fig. 4). A slightly truncated (by two residues) endonuclease IV derivative, Endo IV*, produced during the construction of the chimeric proteins (see "Materials and Methods"), also retained functional activity in E. coli (Table III, Fig. 4).
For expression in yeast, EndoIV 333 and EndoIV 297 were positioned next to the GAL1 promoter to produce, respectively, pGAL-EndoIV 333 and pGAL-EndoIV 297 (see "Materials and Methods"). These plasmids were transformed into the yeast apn1-⌬1 mutant DRY377, and the expression of endonuclease IV derivatives was assessed by assaying AP endonuclease activity (Table I). Both pGAL-EndoIV 297 and pGAL-EndoIV 333 directed basal AP endonuclease expression at ϳ15-20% the level detected for wild-type yeast and galactose-induced levels ϳ10-fold higher than wild-type (Table I).
The expression in yeast of endonuclease IV and the endonuclease IV-Apn1 hybrid proteins was confirmed by immunoblot analysis using anti-endonuclease IV polyclonal antibodies. Extracts of strain DRY377/pGAL-EndoIV expressed a single immunoreactive polypeptide of the same size as purified endonuclease IV (Fig. 5; compare lanes 2 and 6), while pGAL-EndoIV 333 gave rise to two polypeptides of ϳ38 and ϳ35 kDa in size (Fig. 5, lane 4); the slower migrating polypeptide of ϳ38 kDa was of the size predicted for EndoIV 333 and accounted for most of the cross-reacting material. The cross-reacting polypeptide of ϳ35 kDa was likely a proteolytic fragment of EndoIV 333 that arose during extractions rather than an endogenous product (see below). Crude extracts of strain DRY377/pGAL-En-doIV 297 expressed a single polypeptide of ϳ31 kDa in nearly the same amount as native endonuclease IV (Fig. 5, compare lanes 5 and 2). Since cluster 1 contains only 12 residues, this small segment was not expected to alter the molecular size of EndoIV 297 significantly relative to native endonuclease IV. By the same token, immunoblotting could not establish that cluster 1 was retained in EndoIV 297 .
Intracellular Localization of Endonuclease IV and Derivatives-The intracellular localization of endonuclease IV and its derivatives was determined using endonuclease IV-specific polyclonal antibodies. Endonuclease IV encoded by plasmid pGAL-EndoIV was not detectable under our conditions by immunofluorescence in cells grown in glucose medium (data not  shown). After growth in galactose medium, indirect immunofluorescence staining of strain DRY377 (apn1-⌬1) bearing plasmid pGAL-EndoIV exhibited cross-reacting material distributed throughout the cells (Fig. 6B) and not specifically associated with the yeast nucleus (Fig. 6A). In contrast, strain DRY377/pGAL-EndoIV 333 grown in galactose contained crossreacting material that was preferentially localized to the yeast nucleus (Fig. 6D). The EndoIV 297 derivative was not strongly concentrated in the nucleus (Fig. 6, F versus D), which suggests that cluster 1 of Apn1 is not sufficient for nuclear localization in yeast, assuming that the cluster remains attached in vivo.
The combination of clusters 2 and 1 of Apn1 apparently does constitute an effective, transferable nuclear localization signal. A gradient plate assay was used to compare the resistance to MMS of the apn1-⌬1 strains expressing native endonuclease IV and its derivatives. The apn1-⌬1 strain bearing the vector pYES2.0 was hypersensitive to MMS compared to the APN1 ϩ strain in both glucose and galactose media (Fig. 7). In glucose medium, none of the endonuclease IV expression plasmids (pGAL-EndoIV, pGAL-EndoIV 333 , and pGAL-EndoIV 297 ) provided detectable MMS resistance to the apn1-⌬1 strain (Fig.  7A). Thus, even though EndoIV 333 enters the yeast nucleus, the basal expression of this protein (20% of wild-type yeast levels) was unable to correct the repair defect in the Apn1-deficient strain. In contrast, plasmid pDR6, with the APN1 gene under the control of the GAL1 promoter, provided apparently full resistance to the apn1-⌬1 mutant even in glucose medium (Fig.  7A, bar 6), in which Apn1 activity is present at a level 50% higher than in wild-type yeast (Table I). In galactose medium, all the plasmids (except the pYES2.0 vector) provided full MMS resistance to the apn1-⌬1 strain (Fig. 7B).
Although the basal expression of pGAL-EndoIV 333 did not affect cell survival in the face of high-level damage by MMS, an effect on the apn1-⌬1 mutator phenotype was observed. The uninduced level of the EndoIV 333 hybrid protein diminished the spontaneous mutator rate in Apn1-deficient cells Ͼ50%, while induced expression of this protein provided a wild-type mutation rate (Table IV). The pronounced antimutator effect of EndoIV 333 confirms that the hybrid protein is efficiently transported to the yeast nucleus, although some proportion of En-doIV 333 could remain in the cytoplasm. FIG. 7. Resistance to MMS conferred by pGAL-EndoIV and its derivatives in yeast apn1-⌬1 cells. Gradient plates were prepared and analyzed as described under "Materials and Methods"; "100% growth" corresponds to the full length of the tested gradient. Bars 1-6 in each panel represent the following yeast strains: 1, FY86 (APN1 ϩ )/ pYES2.0; 2, DRY377 (apn1-⌬1)/pYES2.0; 3, DRY377/pGAL-EndoIV; 4, DRY377/pGAL-EndoIV 333 ; 5, DRY377/pGAL-EndoIV 297 ; 6, DRY377/ pDR6 (with the APN1 gene under control of the GAL1 promoter). The results presented here demonstrate that E. coli endonuclease IV is actively expressed in yeast and can functionally substitute for its S. cerevisiae homolog Apn1. Both endonuclease IV and Apn1 were shown previously to repair damaged DNA containing AP sites and blocked 3Ј-termini (Levin et al., 1988;Johnson et al., 1988ab;Ramotar et al., 1991b). Thus, the observed substitution of yeast Apn1 by E. coli endonuclease IV is likely a direct effect of endonuclease IV acting in vivo to repair damaged chromosomal DNA in yeast cells. The crossspecies complementation by both endonuclease IV and Apn1 (Ramotar et al., 1991a) suggests that their dual enzymatic activities, AP endonuclease and 3Ј-diesterase, were selected and conserved during evolution. Moreover, the need for these enzymes to repair spontaneous DNA lesions strengthens the likelihood that similar functions will be present in more complex eukaryotic cells.
The ability of EndoIV 333 to diminish the spontaneous mutation rate in Apn1-deficient cells, even when expressed at only ϳ1000 molecules per cell (Table I), indicates that this effect occurs via repair of major substrate(s) of the enzyme. Since the high-level expression of either native endonuclease IV or En-doIV 333 restored normal spontaneous mutation rates to apn1-⌬1 strains, a mutagenic DNA damage that is a substrate for both the bacterial and the yeast enzyme is formed spontaneously in vivo. A major contributor is evidently the endogenous production of AP sites, since about half of the observed apn1-⌬1 mutator effect depends on a DNA glycosylase encoded by the yeast MAG1 gene (Xiao and Samson, 1993;Kunz et al., 1994). Oxidative DNA strand breaks could still constitute potentially mutagenic damages in vivo (Ramotar et al., 1991b), but such a hypothesis is undermined by the recently observed ability of the human Ape AP endonuclease to restore normal mutation rates in Apn1-deficient yeast (Wilson et al., 1995). Ape protein exhibits powerful AP endonuclease activity, but only weak 3Ј-repair diesterase .
Despite its active 3Ј-repair diesterase, endonuclease IV did not completely replace the function of Apn1 in providing resistance to the oxidant hydrogen peroxide, which produces a variety of DNA lesions including modified bases, oxidized abasic sites, and single strand breaks with blocked 3Ј-termini (von Sonntag, 1987). One explanation for the only partial restoration of H 2 O 2 resistance in yeast apn1-⌬ mutants expressing endonuclease IV is that this agent may produce lesions in vivo that are not efficiently repaired by endonuclease IV, but no such damage has yet been identified (Johnson and Demple, 1988b;Levin et al., 1988;Demple and Harrison, 1994). Recently, it has been shown that endonuclease IV cleaves the oxidative DNA lesion ␣-deoxyadenosine (Ide et al., 1994), which arises by abstraction at C1Ј of deoxyribose and subsequent hydrogen donation by compounds such as glutathione (von Sonntag, 1987). A number of other DNA repair enzymes, in-cluding E. coli exonuclease III, endonuclease III, endonuclease VIII, Fpg glycosylase, and T4 endonuclease V do not cleave at ␣-deoxyadenosine, but Apn1 was not tested (Ide et al., 1994). Other candidate oxidative lesions include deoxyribonolactone (1Ј-oxidized) and deoxypentos-4-ulose (4Ј-oxidized) residues, the latter actually cleaved preferentially by endonuclease IV (Hä ring et al., 1994). The activity of Apn1 against these oxidative abasic lesions is unknown. It is also possible that partial complementation of H 2 O 2 resistance in apn1-⌬1 yeast is due to an inability of the bacterial protein to gain access to all of the oxidative damages present in yeast chromatin or by a failure of endonuclease IV to mediate productive interactions with (unknown) accessory repair proteins in yeast.
E. coli endonuclease IV lacks a eukaryotic nuclear localization signal and hence was not concentrated in the yeast nucleus. In this study, we achieved nuclear targeting of endonuclease IV in yeast by attaching both basic clusters 2 and 1 of Apn1 to the bacterial protein; cluster 1 by itself was insufficient to mediate such localization. Together with the delocalizing effect of deleting only cluster 1 from Apn1 (Ramotar et al., 1993), this result indicates that nuclear transport of Apn1 is achieved by a bipartite nuclear localization signal, analogous to the native signals reported for a variety of nuclear proteins (Robbins et al., 1991).