The REC1 gene of Ustilago maydis involved in the cellular response to DNA damage encodes an exonuclease.

Mutation in the REC1 gene of Ustilago maydis is known to lead to a complex phenotype with alterations in DNA repair, recombination, mutagenesis, meiosis, and cell division. The predicted product of the REC1 gene is a polypeptide of 522 amino acid residues with a molecular mass of 56,866 daltons, with no overall sequence homology to any other known protein. The open reading frame of the REC1 gene placed by itself in a U. maydis expression vector was found to be sufficient to complement the rec1 mutant. Overexpression of REC1 in Escherichia coli gave rise to the anticipated 57-kDa product together with a 3'-->5' exonuclease activity. This activity was only present in cells overexpressing REC1 and its characteristics were distinguishable from the major bacterial nucleases, but it had certain enzymatic features in common with epsilon, the proofreading exonuclease subunit of E. coli DNA polymerase III holoenzyme. To facilitate isolation of the protein product from bacteria, the REC1 gene was overexpressed from a vector that fused a hexa-histidine-leader sequence onto the amino terminus, enabling the isolation of the HisREC1 product on an immobilized metal ion affinity column. The His-REC1 protein co-eluted with the novel exonuclease activity. Alignment of the amino acid sequence of the REC1 gene product with the conserved proofreading exonuclease motifs of DNA polymerases indicated significant homology.

The nucleotide sequence(s) reported in this paper has been submitted to the GenBank-IEMBL Data Bank with accession numberls) U01836.
* The use of the term "error-prone" indicates a phenomenon which the yeast paradigm have emerged from analyses of radiationsensitive mutants in other well-characterized but alternative systems such as Ustilago maydis. For instance, the radiationsensitive mutants recl, rec2, and uvsd isolated by Holliday (1965, 19671, which rank among the earliest examples of DNA repair mutants, appear to fall into the three broad categories defined in studies with S. cerevisiae. The mutant uvsd is defective in excision repair (Unrau, 19751, but proficient in recombination (Holliday, 1967). The rec2 mutant is quite sensitive to ionizing radiation, is defective in mitotic recombination, and is blocked in meiosis (Holliday, 1967), in line with an expected recombination defect, and thus comparable to the recombinational repair group. The reel mutant has an extremely complicated phenotype with certain hallmarks found only in the third error-prone repair group.
The recl mutant was originally isolated in a screen for mutants sensitive to ultraviolet light, but was also found to be sensitive to ionizing radiation and chemical alkylating agents (Holliday, 1965(Holliday, , 1967. It was discovered to have an elevated frequency of mitotic recombination, in which normal gene conversion is replaced by an aberrant form of crossing over which is strongly associated with chromosome breakage or loss (Holliday et al., 1976). The mutant is also blocked in radiationinduced allelic recombination. Homozygous diploids are unstable during mitotic growth and exhibit considerable lethal sectoring and striking variation in colony size and morphology.
Spontaneous mutation frequency is elevated, as evidenced by a 10-fold increase in the rate of reversion of particular auxotrophic markers, and an increase in forward mutation frequency, but no radiation-enhanced reversion nor induction of mutation is observed. Meiosis is aberrant with low viability of mitotic products and frequent production of diploids and aneuploids. This pleiotropic phenotype has no exact parallel among any member of the group of yeast mutants defective in error-prone repair. However, the disparate characteristics and especially the mutator phenotype are features giving clues that the RECl gene product might function in some aspect of an error-prone repair pathway.
Our interest in recombination and repair in U. maydis led us to develop a cloning and transformation system with the goals of isolating the genes involved, discovering their functions, and understanding the genetic consequences of their dysfunction. Through the efforts of our laboratory as well as others, a tractable molecular genetic system in U. maydis has been established. In previous studies by us (Tsukuda et al., 1989;Thelen et al., 1992) and independently by Holden et al. (19911, the isolation of genomic DNA fragments that complemented the DNA repair and recombination defects of the reel mutant was described and proof by gene disruption procedures was presented that these fragments contained the R E C l gene. The occurs during DNA damage-induced mutagenesis, and is not to be confused with the specific error-prone repair pathway of the SOS response in E. coli (see Friedberg et al. (1991)). sequence of a genomic fragment containing the RECl gene was reported (Holden et al., 1989a), but no insight into its function was obtained at that time after a search of the protein data bases (Holden et al., 1991). The aim of the present work was to overexpress and purify the RECl gene product so as to discover its biochemical activity and to gain insight into its function in the cellular response to DNA damage.

U. maydis Strains and
Plasmids-Cell cultures were grown on minimal or YEPS medium as described before (see Table I andHolloman (1989, 1990)). The RECl gene was previously isolated in this laboratory on a 6.9 kbp2 U. maydis genomic fragment that complemented all aspects of the recl phenotype (Tsukuda et al., 1989). The gene was then subcloned from pCM130 on a 2.7-kbp SmaI-Hind111 DNA fragment. The DNA sequence of this fragment was determined and an open reading frame of 1569 nucleotides was identified identical to that reported by Holden et al. (1989a). Two plasmids used in cloning portions of RECl were constructed by inserting fragments into the EcoRV site of pBluescriptII (Stratagene, La Jolla, CA): pCM224 contains the SmaI-Hind111 2.7-kbp fragment, and pCM251 contains the complete RECl open reading frame (OW) generated by the polymerase chain reaction (PCR), described below. For bacterial overexpression, the PCR-amplified ORF was removed from pCM251 as a 1.6-kbp NdeI fragment and inserted into the NdeI site of pET3c (obtained from F. W. Studier, Brookhaven National Laboratory), bringing it in-frame behind the T7+10 promoter, yielding pCM322. For overexpression with a hexahistidine-leader, the NdeI fragment containing the RECl ORF was inserted into the NdeI site of pET14b (Novagen, Madison, WI), yielding pCM391. The same RECl fragment was also inserted into the unique NdeI site of pCM317, a shuttle vector constructed by B. Rubin of this laboratory specifically for expression of genes in U. maydis (see Fig.   IA). The relevant features of pCM317 are that it was derived from pCM284, which is pBluescriptI1 with the U. maydis ARS (Tsukuda et al., 1988), and the HPH gene conferring resistance to hygromycin B (Wang et al., 1988). pCM317 was constructed by insertion into pCM284 of a 0.3-kbp fragment containing part of the U. maydis hsp7O promoter element (Holden et al., 1989b), generated by PCR to contain an NdeI cloning site for translational initiation during expression. The 1.6-kbp NdeI fragment from pCM251 containing the RECl ORF was then inserted into the NdeI site of pCM317 to yield pCM306.
Isolation of the RECl Open Reading Frame-A 1596-base pair sequence including the first ATG and extending to 30 nucleotides past the TAA termination codon of the RECl gene was amplified by PCR. The sequence amplified was contained within the cloned 2.7-kbp SmaI-HindIII fragment from pCM224. The two primers used in the amplification reaction were each designed to contain a NdeI site (underlined below) distal to the sequence overlapping either the 5'or 3"region of the RECl OW. Primer 1 was 5'-GCATCATATGCCGGCCGAGG- GAGCTT and was complementary to the first 19 nucleotide residues of the RECl ORF noncoding strand. Primer 2 was 5"GCATCATATGGC-GGCGATI'GCGCAAGGC and was complementary to nucleotides 1579-1596 of the coding strand. The reaction (100 pl) containing 10 m~ Tris-HC1, pH 8.3, 1.5 m~ MgCl,, 50 m~ KCl, 0.2 m~ each dNTP, 1 p each primer, 1 ng of pCM224, and 0.01% gelatin, was held at 95 "C for 5 min. Thermus aqwticus DNA polymerase (5 units, Perkin-Elmer Cetus) was added to begin the reaction and the temperature was then cycled 30 times from 72 "C (3 min) to 94 "C (1 min). Following amplification, the 1.6-kbp product was isolated after electrophoresis through a 1% agarose gel. Its termini were made flush by repair with the Klenow fragment of E. coli DNA polymerase I and then phosphorylated with T4 polynucleotide kinase and ATP. The linear product was then inserted into the EcoRV site of pBluescriptI1 to yield pCM251. The DNA sequence of the amplified RECl ORF was determined and was found to be identical to that of the RECl ORF in the genomic clone pCM224. Complementation of DNA Repair Deficiency-Transformed U. maydis recl strains to be tested after treatment with Cnitroquinoline-loxide were grown to lo7 celldml, 4-nitroquinoline-1-oxide added to 1 pglml, and further incubated with shaking at morn temperature. Aliquots (0.1 ml) were removed at timed intervals, washed twice in water, appropriately diluted and plated on selective medium. Survivors were counted as colonies visible after incubation for 3 days at 32 "C.
Overexpression of the RECl Gene in E. coli-E. coli BLZl(DE3) [F hsd S gaNAint::lacUV5-T7genel immzl nin5 Sam71 transformed with a RECl containing plasmid was induced for expression with isopropyl-& D-thiogaladoside (IPTG) as described by Studier et al. (1990). A cell culture (from 5 to 1,000 ml) of BL21(DE3)/pCM322 was grown at 37 "C toAem of 0.6 and induced by addition of 1 m~ IPTG. After 3 h, cells were harvested by centrifugation, washed in a 0.05 volume of TNE (50 m~ Tris-HC1, pH 8.5, 100 m~ NaC1, 1 m~ EDTA), and stored at -70 "C. After thawing, cells were resuspended in 0.05 volume of TNE and lyfor 1 h. EDTA was then added to 10 m~ and Triton X-100 to 0.1% and sozyme was added to 1.0 mglml. The suspension was incubated on ice the suspension was held briefly at 37 "C until extremely viscous. The suspension was sonicated with 3 x 30-s bursts using an immersion probe (Branson Sonifier), then centrifuged for 15 min at 20,000 x g. Afterwards, all of the RECl product was contained in the precipitate, as demonstrated by protein staining on SDS-polyacrylamide gels (see for example, Fig. 2, lanes 2 and 3). The supernatant was therefore discarded, and the pellet was washed three times by centrifugation and resuspension (with the aid of a motor-driven teflon pestle) in 0.05 volume of buffer: first in TNE containing 2 M NaCl, then in TNE containing 2 M urea, and finally in 0.1 M Tris-HC1, pH 8.5. The washed precipitate obtained was stored at -20 "C until needed. E. coli strain BCM 503(DE3) ( e n d l thi-1 hsdR17 supE4f uvrD::Tn5/Aint::lacUV5-T7genel immZ1 nin5 Sam7) (constructed by B. Rubin of this laboratory) was used in a similar overexpressed preparation for the results presented in Table N: Exonuclease Assay-Standard reactions (0.2 ml) contained 50 m~ Tris acetate, pH 9.0, 10 m~ Mg2"acetate, 1 m~ dithiothreitol, 0.1 m~ EDTA, and substrate DNA. Exonuclease assays were carried out with 0.2 nmol (as total DNA nucleotide) of end-labeled DNA (see above), containing 0.2 pmol of either 32P-nucleotide as terminal label at a specific activity of 2 x lo6 cpdpmol, or 1 pmol of 36S-nucleotide as terminal label at a specific activity of 5 x lo4 cpdpmol. For general nuclease assays, 1 nmol (as total DNA nucleotide) of uniformly labeled SH-P22 DNA was used at a specific activity of 4 x lo4 cpdnmol. Reactions were started by the addition of 1-20 pl of enzyme, incubated at 37 "C for 10 min (or the time indicated), and terminated by the addition of ice-cold solutions of camer (300 pl, 0.5 mg/ml DNA (sonicated salmon sperm DNA), 25 n m EDTA) and 10% trichloroacetic acid (500 pl). The mixture was held on ice for 10 min, centrifuged at 10,000 x g for 10 min, and the supernatant was removed and mixed with 5 volumes of scintillant (Ecolume, ICN Biomedicals Inc.) for determination of radioactivity. Each experiment included at least three protein concentrations to determine the linear relationship between activity and protein. One unit of exonuclease activity is that amount of enzyme that converts 1 pmol of radiolabeled DNA to an acid-soluble form in 10 min under these conditions. Activity in protein fractions containing guanidine HCI were assayed by adding a 1-pl sample directly to the standard assay; under these conditions the RECl protein is soluble to at least 10 pg/ml, determined by measuring the radioactivity in protein labeled in vivo during overexpression (data not shown). Comparisons of purified E. coli exonucleases were performed using DNA polymerase I Klenow fragment (New England Biolabs, Inc.), exonuclease 111 (New England Biolabs, Inc.), exonuclease I (U. S. Biochemical Corp.), and A-exonuclease (Life Technologies Inc.). The e-subunit of DNA polymerase 111 holoenzyme (dnaQ gene product) was isolated from the overproducing E. coli strain MClOOO transformed with pRK248-cZts and pNS360 (provided by M. O'Donnell of this Department) as described by Scheuermann and Echols (1984).
Isolation of the HisRECl Protein-E. coli BL21(DE3)/pCM391 cells (from 1-liter cultures) were induced and lysed as described above, and the insoluble precipitate containing the HisRECl protein was prepared. The washed precipitate was dissolved in 20 ml of Buffer A (6 M guanidine HCI, 10 lll~ Tris-HC1,O.l M sodium phosphate) at pH 8.0. The clear protein solution (Fraction I) was centrifuged to remove any minor insoluble material, and loaded onto a column (1 x 6.5 cm) containing nitrilotriacetic acid-agarose (NTA) which had been charged with Ni2+ and equilibrated with Buffer A according to the vendor's protocol (Qiagen, Inc.). The Ni2+-NTA column was developed by gravity (10 mVh) with a discontinuous pH gradient in Buffer A: first at pH 8.0 (50 ml), followed by pH 6.3 (25 ml), pH 5.9 (20 ml), and finally pH 4.5 (20 ml). Fractions of 1.5 ml were collected and assayed for exonuclease (see above) and fractions containing activity were combined (Fraction 11). Guanidine HCI was removed by dialysis of Fraction I1 for 12-18 h against one change of 0.5 M NaCl in Buffer Z (50 m~ Tris-HC1, pH 8.5, 20% glycerol, 5 nm dithiothreitol, 0.01% Triton X-loo), and then against two changes for 6 h each against Buffer Z containing 10 nm NaC1. About 90% of the protein was precipitated below a concentration of 0.5 M NaCl, and this was removed from the dialysate by centrifugation. The supernatant was applied to a Mono S HR 5/5 column (Pharmacia LKB Biotechnology Inc.) and the column developed in a 20-ml gradient of 0-400 m~ NaCl in Buffer Z. Fractions (1 ml) were collected and assayed for exonuclease activity. A single peak of exonuclease activity emerged at a salt concentration of 0.2 M. In order to estimate the size of the HisRECl protein, a 100-p1 sample from the peak Mono-S fraction was loaded on a 30-ml column of Superose 12 HR10/30 (Pharmacia) and the column developed with 30 ml of Buffer Z containing 50 m~ NaC1. Fractions (0.5 ml) were collected and assayed for exonuclease activity, and a single sharp peak of exonuclease activity was detected. Column fractions were analyzed for protein content with Coomassie Blue dye reagent (Bio-Rad), using the column buffer as blank. Samples containing guanidine HCl to be analyzed by SDS-gel electrophoresis were precipitated in 3 volumes of ethanol at -20 "C, a step necessary to remove guanidine HCl which precipitates SDS. The precipitates were washed with 75% ethanol before dissolving in SDS-sample buffer (1% SDS, 100 nm &thiothreitol, 50 m~ Tris-HC1, pH 6.8, 10% glycerol, 0.1% bromphenol blue).

RESULTS
The Functional Unit of the RECl Gene-In previous work from this laboratory a genomic DNA fragment that complemented the DNA repair and recombination phenotype of the recl mutant was isolated and established by one-step gene disruption to contain the RECl gene (Tsukuda et al., 1989).
After subcloning and sequencing, we found that the product of the RECl gene should be a polypeptide of 522 amino acid residues with a molecular mass of 56,866 daltons. The ORF of 1569 nucleotides was amplified by polymerase chain reaction, its complete DNA sequence rechecked, and then inserted in a U. muydis expression vector (Fig. lA) to test for biological activity. Following 4-nitroquinoline-1-oxide treatment of a recl strain transformed with this plasmid (Fig. B), the DNA repair defi- ( R E C l ) were transformed to hygromycin resistance with pCM306 which contains the PCR-amplified RECl ORF cloned into the NdeI site of the expression vector pCM317. ARer growth in YEPS containing 100 pg/ml hygromycin, cultures were treated with 4-nitroquinoline-1-oxide pCM306 (diamonds); UCM3/pCM306 (triangles). and cell survival was measured. UCM2UpCM317 (circles); UCM2U ciency was restored to the wild-type level providing independent proof that the isolated ORF contains the structural portion of the RECl gene.
Identification of an Exonucleolytic Activity in the RECl Gene Product-The RECl ORF was inserted behind the 410 bacteriophage T7 promoter in the PET vector system designed for overexpression of cloned genes in E. coli (Studier et al., 1990). After induction with IPTG, a protein with the predicted mass of the RECl gene product accumulated in the cells, as determined by SDS-gel electrophoresis (Fig. 2, lanes 1 and 2). The 57-kDa protein was specifically labeled when [35Slmethionine and rifampicin were added to the induced culture, as would be expected for a gene expressed in E. coli under the control of a T7 promoter. The overexpressed RECI protein was insoluble even when cell extracts were prepared in the presence of 6 M urea. Solubilization was achieved only by addition of 4 M guanidine HCl, 1% SDS, or 1% Sarkosyl. Because of the insoluble nature of the overproduced protein, most bacterial proteins in extracts were easily removed when the precipitate was washed in buffer containing urea (Fig. 2, lanes 3 and 4 ) .
The insoluble RECl gene product remaining after extensive washing was dissolved in concentrated guanidine HCl and then processed through a renaturation regime involving the slow removal of denaturant by dialysis. A potent nucleolytic activity which released radiolabeled nucleotide preferentially from the 3'-end of DNA was detected in the renatured preparation of RECl protein (Fig. 3). Over an extended time uniformly labeled DNA could be converted almost completely to an acid-soluble form. The conditions for optimal activity were a low ionic strength buffer at pH 8.5-9.0 containing Mg2+ (Table 11). Activity was unaffected by ATP and d N T P s but was inhibited by dAMP. Salt at 100 mM reduced the activity nearly 80%. Little to no activity was observed when M e was replaced with Ca2+, Mn2+, Zn2+, or Co2+. A reducing agent was necessary for maintaining enzymatic stability over time during storage at 4 or -20 "C. Distinct Characteristics of the RECl-associated Exonuclease Activity-The activity observed in renatured preparations of RECl gene product was examined for properties that could distinguish i t from the major nucleases of E. coli and A-lysogens which might potentially contaminate the bacterial preparation.
Based on the published properties of well characterized E. coli nucleases, exonuclease V and exonuclease VI1 were ruled out as possible sources of contamination, since the former is dependent upon ATP (Wright et al., 1971) as a cofactor and the latter has no divalent cation requirement for activity (Chase and Richardson, 1974). We considered in more detail Mg2"requiring nucleases that are known to render DNA acid soluble. These include exonuclease I (Lehman and Nussbaum, 1964), the 3'45' proofreading activity of DNA polymerase I (Brutlag and Kornberg, 1972), exonuclease 111 (Richardson et al., 1964), and endonuclease I (Lehman et al., 1962). We also considered A-exonuclease (Carter and Radding, 1971), because the E. coli containing RECl was induced for expression and further processed a s described to obtain the washed precipitate fraction. Protein (approximately 3.5 mg) was solubilized in 2 ml of 4 M guanidine HCI in Buffer Z, diluted to 1 M guanidine HCI with buffer, then slowly renatured by dialysis against Buffer 2. The final volume of the dialysate was 7.5 ml containing about 0.34 mg of protein, representing a final yield of 1 W r .
Precipitated material was removed by centrifugation. Portions containing 90 ng each were incubated for activity in the standard reaction mixture with end-labeled linear duplex plasmid DNAs for the times indicated. Filled circles, 3'-Y%DNA substrate; open circles, 5'-R2P-DNA substrate.

TABLE I1
Characterization of nuclease activity in a n overexpressed RECl preparation The description of RECl sample preparation is given in the legend to Fig. 3. Standard incubation conditions were at 37 "C for 30 min in 25 nw "is acetate, pH 9.0, and 10 m Mg2'. Listed in the table are individual modifications to these conditions (e.g. replacement of M P with the other divalent cations) or test compounds which were added to the standard mixture (noted as '+").
total protein from the RECl sample and 1 nmol ( a s t o t a l DNA nucleo-Activity was measured in each reaction mixture using 0.9 pg of tide) heat denatured, uniformly labeled I5H1P22 DNA. For this experiment, the maximal activity was 100 pmol of nucleotide releaned in 30 min a t 37 "C. strain that we used for overexpression was a A-lysogen, but this possibility seemed unlikely since thepL-promoted transcription of the red genes, which occurs under conditions of DNA damage induction, would not be expected with IFTG induction.
When assayed under the same conditions i t was apparent that the exonuclease activity present after overexpression of RECl was different from several exonucleases. The RECZ-associated exonuclease showed only a 5-fold preference for singlestranded over duplex DNA as substrate (Table 111). unlike ex* The description of RECl sample preparation is given in the legend to Fig. 3. Purified E. coli exonucleases I and I11 and A-exonuclease were purchased.

DNA substrates. Activity was measured after 30 min in the standard
Ratio of nuclease activity on ss (heat-denatured) versus ds (native) reaction mixture containing uniformly labeled c3H1P22 DNA. The maximal activity resulted in the hydrolysis of about 25% of the substrate in each case. e Ratio of exonuclease activity on 3'-35Sversus 3'-32P-DNA substrates. Activity was measured after 10 min in the standard reaction mixture using end-labeled linear duplex plasmid DNA. The maximal activity resulted in the release of about 50% of the substrate in each case.
Ratio of exonuclease activity on 3'versus 5'-32P-DNA substrates. Activity was measured as described in footnote c. The maximal activity resulted in the release of about 20% of the substrate in each case.
~~_ _ _ _ nuclease I of E. coli (Lehman and Nussbaum, 1964). The activity also differed from the 3'+5' exonuclease activities of exonuclease I11 and DNA polymerase I in its relaxed specificity for the type of diester-linkage. These two bacterial enzymes are reduced more than 100-fold in activity when the linkage of the substrate DNA to the 3"terminal nucleotide was through a 3'-phosphorothioate ester linkage instead of the standard 3'phosphodiester linkage (Putney et al., 1981;Gupta et al., 1984). Under the conditions tested, the exonuclease in the RECl preparation exhibited a 60-fold preference for 3'4abeled substrate, unlike A-exonuclease (Carter and Radding, 19711, which preferentially degrades DNA in the 5'+3' direction (see also Fig. 3). The level of exonuclease activity was not reduced when the RECl gene product was overexpressed and prepared similarly from an e n d strain of E. coli (see Table IV), indicating the activity is distinct from E. coli endonuclease I.

Comparison of Exonuclease Activity in Overexpressed Pro-
teins-Further evidence that overexpression of the RECl gene was responsible for the 3'+5' exonuclease activity was obtained by a controlled comparison of several different protein extracts. In addition to the renatured preparation of RECl, samples containing the overexpressed products of HisRECl (REC1 with a hexa-histidine-leader, described below), and the E-subunit of E. coli DNA polymerase I11 were prepared in parallel and tested for activity using the 3'-thiophosphoryl substrate. Exonuclease activity was observed in RECl, HisRECl, and E preparations but was at the limit of detection in the control samples which were not expected to overproduce a protein ( Table IV). The activity measured in the HisRECl preparation was lower than that of RECl, but the characteristics of the activity were nearly identical. The E-subunit of DNA polymerase I11 holoenzyme remains the notable E. coli exonuclease with properties in line with the observations here. It has the same polarity and preference for single-stranded substrate (Scheuermann and Echols, 1984;Brenowitz et al., 1991). Furthermore, is also active on the thiophosphoryl end-labeled DNA (Griep et al., 1990;Table IV). In direct comparison with the RECl sample, the highest yield of activity after overexpression was obtained from the preparation of E . This protein was also insoluble after overexpression (Scheuermann and Echols, 19841, but following the denaturatiodrenaturation regime much more of the product was soluble than the products of RECl and its derivatives. However, the specific activity of the RECl sample was 2-fold higher than that of E in these prepa- a The description of the RECl preparation is given in the legend to Fig. 3, and all other samples were prepared in this way. The linear range of activity by RECl and Q was determined to be within 10 min for 10-50 ng of protein. This range of protein was then used from each sample to measure activity in the standard reaction mixture containing the 3'-35S-DNA substrate (or in the case of BCM503(DE3) samples, the 32P-DNA substrate was used). Values reported as "cn" were at the lower limit of detection in the standard activity assay. rations. Another difference was found with respect to cofactor requirements: Mn2+ stimulates the activity of overexpressed E by as much or slightly more than Mg2+ (Thompson, 1992) in contrast to that observed with the RECl-associated exonuclease (see Table 11). Furthermore, although all four dNMPs inhibit the activity of E , dAMP in particular appears to have much less of a n effect on E than it does on the activity in the RECl preparation (Scheuermann and Echols, 1984;Thompson, 1992; Table 11).
The comparison of the overexpressed proteins as well as the purified nucleases with the RECl sample thus provide several independent lines of evidence which strongly point to the conclusion that the exonuclease activity we observe is inherent in the RECl gene product.
Chromatographic Isolation and Size Estimation of the 3'+5' Exonuclease-By visual inspection of stained SDS gels it was estimated that greater than 10 times if not more of the overexpressed protein was observed when the RECl ORF was preceded by a 60-base pair leader sequence containing 6 consecutive histidine codons provided by the pET14b vector (Fig. 2,  lanes 2 and 5 ) . The reason for the increase in expression of the fusion product is not understood. It is possible that there was stimulation by the artificial leader sequence at the level of transcription or translation, or else that this construct was more stably maintained in the cells; nevertheless, the insolubility of the protein was apparently unaffected. Due to the utility of the histidine-leader sequence in afiinity chromatography, the resulting 60-kDa HisRECl product was considered more amenable for biochemical studies than the RECl product.
The insoluble fusion protein was isolated from bacteria and processed exactly as in the RECl preparation (Fig. 2, lanes 5  and 6). After solubilization with guanidine HCl, the HisRECl protein was isolated using immobilized metal ion afiinity chromatography (reviewed by Porath (1992)). The denatured protein (Fraction I) bound stably to a Ni2+-NTA column and contaminating proteins which lacked the histidine-leader sequence were removed by washing sequentially with buffers at pH 8.0 and 6.3 (data not shown). A broad peak of exonuclease activity coinciding with the peak of HisRECl protein eluted with buffer at pH 5.9, giving rise to Fraction I1 ( Fig. 4A and Fig.   2, lane 7). At pH 5.9, the complex made with the immobilized Ni2+ through two or more consecutive histidines on the protein is disrupted. The total exonuclease activity measured in Fraction I1 was significantly higher (usually 10-fold) than that in the sample loaded, and consisted of about 10% of the total A, purification on a Ni2+-NTA affinity column. HisRECl protein was bound to the column in 6 M guanidine HCl at pH 8, washed further at pH 6.3, and then eluted at pH 5.9 as described (loading and wash steps are not shown). Exonuclease was assayed from 1 pl of each 1.5-ml fraction. The activity recovered in Fraction I1 was 50-fold over that in the sample loaded. Circles, activity; dots, protein. B, recovery of renatured exonuclease. Protein in Fraction I1 was renatured by dialysis and loaded onto a Mono-S cation exchange column, which was developed by fast protein liquid chromatography in a linear gradient of NaCl. Exonuclease was assayed from 2 pl of each 1-ml fraction, and indicated an overall recovery in Fraction I11 of 26%. C, native size estimation. A portion of Fraction I11 was analyzed by Superose-12 molecular sieve chromatography. Exonuclease was assayed from 10 pl of each 0.5-ml fraction, with 35% recovery from the sample loaded. Elution volumes of protein standards in a parallel run are indicated BSA, bovine serum albumin (66 m a ) , CA, carbonic anhydrase (30 m a ) , Cyt-c, cytochrome c (12 m a ) . The separate elution of E. coli exonuclease 111 (denoted as Xth) activity is also shown. protein of Fraction I. This 100-fold increase in specific activity can be accounted for by the removal of inhibitory contaminants which may also be responsible for the aggregation and precipitation of the overexpressed protein. The bulk of the HisRECl protein remained bound to the Ni2+-NTA column at pH 5.9 and was completely washed off at pH 4.5. Protein eluting at pH 4.5 is most likely in an aggregated form; this fraction of HisRECl was completely insoluble and was devoid of exonuclease activity. After renaturing the HisRECl protein in Fraction 11, about 10% remained soluble and was applied to a Mono-S cation exchange column. The exonuclease was isolated by fast protein liquid chromatography (Fig. 4B) and retained the characteristics of the renatured activity described in Table 11.
The native size of the exonuclease was determined by molecular sieve chromatography on a Superose-12 column (Fig.  4C). This value was estimated at 60 kDa which is in agreement with the molecular mass calculated from SDS gel analysis of the HisRECl protein, and was clearly resolved from the 28-kDa exonuclease I11 of E. coli. The size estimation indicates that the active species of HisRECl is monomeric under the chromatographic conditions. Sequence Homology Alignment with Other Exonucleases-In view of the novel exonuclease activity associated with the RECl gene product, a search for sequence similarity with other known exonucleases using the FASTA program in the GenBank data base was carried out. The search turned up no known exonucleases with overall homology. However, the segment of E. coli DNA polymerase I (residues 250460) containing the 3'+5' exonuclease active site was aligned with 209 residues of the NH2-terminal portion of RECl (residues 40-250) with 20% identity in the overlap. This similarity was then localized more specifically to residues in the three so-called "Exon motifs which are ordered within the domain of the 3'+5' exonuclease active site conserved in many other DNA polymerases (Fig. 5, A and   B; Bernad et al., 1989). Notable among these is the E. coli dnaQ gene product, the 3l-5' proofreading c-subunit of DNA polymerase I11 holoenzyme (Scheuermann and Echols, 1984). The comparison of E and RECl with the DNA polymerase I 3'+5' exonuclease active site region indicated the homology in the Exo motifs was higher in the three sequences than with any two of these sequences alone (Fig. 5A ). The three motifs in DNA polymerases were first identified by crystallographic studies on the Klenow fragment of E. coli DNA polymerase I, which revealed amino acid residues essential for metal ion binding, substrate orientation, and catalysis (Beese and Steitz, 1991;Freemont et al., 1988;Derbyshire et al., 1988Derbyshire et al., , 1991. This has been extended from the crystallographic model by sequence alignment to conserved residues in over 30 DNA polymerases (Bernad et al., 1989;Ito and Braithwaite, 1991;Momson et al., 1991), although many of these contain little homology in the NH2-terminal region apart from that observed in the three Exo motifs. That the conserved residues (indicated in Fig. 5 A ) are indeed essential for catalysis has been demonstrated by both genetic and biochemical tests following site-directed mutagenesis (Derbyshire et al., 1988(Derbyshire et al., , 1991. While these three motifs appear to be similar in their sequence composition and spatial arrangement in many DNA polymerases, their presence is not obvious in other well characterized 3'+5' exonucleases such as E. coli exonucleases I or 111. We were unable to make similar homologous alignments of the RECl gene product with these latter exonucleases.

DISCUSSION
The principal conclusion from biochemical experiments with the RECl gene product isolated from E. coli after overexpression is that an exonuclease activity is inherent in the RECl protein. Evidence supporting this conclusion comes from sev-  ., 1991); the three amino acid sequences were aligned using the Genalign program and refined by eye using a minimum of spacing. Stars over the sequences refer to those residues in DNA polymerase I which are involved in metal ion binding and catalysis (see text). Chemically similar amino acids were considered as follows: A and G D, E, N, and Q; K, H, and R S and T; F, Y, I, L, V, C, and M. Numbering of residues is from the NH2 terminus in each case. Another possible alignment of the DXE motif within Ex01 occurs in residues 116118 in RECl, although the surrounding amino acid similarity is not as good as that shown here. B, order and spatial arrangement of the three exonuclease motifs. The polypeptide lengths of DNA polymerase I, c, and RECl (depicted to scale) are 928, 243, A d 522 resiiues, respectively. era1 independent lines of investigation. 1) A potent 3l-5' exonuclease activity with novel properties appears in preparations of the RECl gene product, but is not found in extracts from bacterial cells where the RECl gene has not been expressed. 2) The exonuclease activity was isolated after three column chromatography steps: in the first of these steps the activity was bound to a n immobilized Ni2+ column under denaturing conditions, and released at the pH predicted to disrupt this specific and high affinity complex with the histidine-leader on the fusion protein HisRECl. After a second column step the size of the exonuclease was estimated to be 60,000 daltons, corresponding to the molecular mass calculated from the deduced amino acid sequence, and to the molecular mass of the overexpressed fusion gene product as determined by SDS-gel electrophoresis.
3) The activity of the RECl preparation on duplex DNA, the ability to hydrolyze a phosphorothioate ester, the lack of ATP requirement, the requirement for M e , the polarity of DNA degradation, and the presence of the nuclease activity in an e n d strain are all observations that place this activity in a class apart from E. coli exonuclease I, exonuclease 111, exonuclease V, exonuclease VII, &exonuclease, DNA polymerase I, and endonuclease I which are abundant enzymes and the most likely sources of contamination by cellular nucleases. We cannot completely rule out contamination by the €-subunit of DNA polymerase I11 holoenzyme. However, we think this is unlikely since Q is present at only a few copies per cell. 4) The amino acid sequence alignments of the RECl gene product with those of E. coli DNA polymerase I and E , both in the conserved "Ex0 domain" landmark amino acids and in the order and spacing of the three segments constituting these motifs (Fig. 5), are consistent with the alignments of many DNA polymerases with real or predicted exonuclease proofreading activity. The simplest and most straightforward interpretation of these various observations is that the novel 3 ' 4 5 ' exonuclease activity appearing in cells overexpressing the RECl gene is a n intrinsic feature of the RECl protein.
The exonuclease activity identified in the RECl gene product provides a mechanistic framework for beginning to rationalize the complex phenotype of the recl mutant, which is characterized by radiation sensitivity, lethal sectoring, elevation in spontaneous mutation frequency, aberrant meiosis, elevated spontaneous allelic recombination, and an altered form of crossing over associated with a high rate of chromatid breakage and loss (Holliday et al., 1976). For instance, sensitivity to radiation could result from failure of a defective exonuclease to function in repair of DNA lesions. The mutator phenotype of the recl mutant could arise as a consequence of a dysfunctional exonuclease, as postulated for some mutants in dmQ (Echols et al., 1983;Takano et al., 1986;Schaaper, 1989). Other aspects of the phenotype such as lethal sectoring, immutability by ultraviolet irradiation, and crossing over associated chromatid breakage are less easy to rationalize in terms of loss of a n exonuclease function and could be indicative of some intricate role of RECl in controlling or interacting with other genes as postulated by Holliday (e.g, Holliday et al., 1976;Holliday, 1977). More biochemical and molecular genetic studies will be necessary to gain a better understanding of the in vivo functioning of the RECl gene product. It would come as no surprise to find interplay between the RECl gene product and other components important in error-prone DNA repair.