Evidence That the DNA Endonuclease ARTEMIS also Has Intrinsic 5′-Exonuclease Activity*

Background: DNA-PKcs stimulates the endonucleolytic activities of ARTEMIS but not the 5′-exonuclease activity. Results: Point mutagenesis, small molecule inhibitors, and modulation of divalent cation concentrations all affect both the endonuclease and 5′-exonuclease activities in parallel. Conclusion: ARTEMIS has intrinsic 5′-exonuclease activity, in addition to its known endonuclease activity. Significance: ARTEMIS may use both its endo- and its 5′-exonuclease activity in NHEJ. ARTEMIS is a member of the metallo-β-lactamase protein family. ARTEMIS has endonuclease activity at DNA hairpins and at 5′- and 3′-DNA overhangs of duplex DNA, and this endonucleolytic activity is dependent upon DNA-PKcs. There has been uncertainty about whether ARTEMIS also has 5′-exonuclease activity on single-stranded DNA and 5′-overhangs, because this 5′-exonuclease is not dependent upon DNA-PKcs. Here, we show that the 5′-exonuclease and the endonuclease activities co-purify. Second, we show that a point mutant of ARTEMIS at a putative active site residue (H115A) markedly reduces both the endonuclease activity and the 5′-exonuclease activity. Third, divalent cation effects on the 5′-exonuclease and the endonuclease parallel one another. Fourth, both the endonuclease activity and 5′-exonuclease activity of ARTEMIS can be blocked in parallel by small molecule inhibitors, which do not block unrelated nucleases. We conclude that the 5′-exonuclease is intrinsic to ARTEMIS, making it relevant to the role of ARTEMIS in nonhomologous DNA end joining.

ARTEMIS is a member of the metallo-␤-lactamase superfamily of nucleases. The metallo-␤-lactamase family is characterized by a conserved metallo-␤-lactamase and ␤-CASP 2 domain and their ability to hydrolyze DNA or RNA. Two other mammalian members of the same family, SNM1A and SNM1B (Apollo), function in DNA interstrand cross-link repair (1) and protection of telomeres (2). ARTEMIS, also known as SNM1C, functions in the nonhomologous end joining (NHEJ) repair pathway. On duplex DNA, ARTEMIS has endonuclease activity on 5Ј-and 3Ј-DNA overhangs and at DNA hairpins, and these are all dependent on DNA-PKcs (3). In the context of V(D)J recombination, the ARTEMIS⅐DNA-PKcs complex is required for the opening of hairpins generated by the RAG complex (3). Patients lacking ARTEMIS suffer from severe combined immunodeficiency (4). NHEJ requires nucleases for resection of broken DNA ends, polymerase to fill in the gaps, and the XLF⅐XRCC4⅐DNA ligase IV complex for ligation (5). ARTEMIS is at least one of the nucleases participating in this process and is responsible for a subset of end processing during NHEJ in response to ionizing radiation (6). Patients or cells lacking ARTEMIS are deficient for repair caused by ionizing radiation (4,7) and for repair of breaks caused by topoisomerase II inhibitors (8).
In a biochemical system consisting of purified proteins of the NHEJ pathway, ARTEMIS, in complex with the serine/threonine kinase DNA-PKcs, removes 5Ј-and 3Ј-overhangs in order to join duplex DNA ends (9). ARTEMIS⅐DNA-PKcs is also able to open hairpins both in vivo and in vitro. ARTEMIS opens hairpins preferentially two nucleotides past the tip of a perfect hairpin (3). At a 5Ј-overhang, ARTEMIS⅐DNA-PKcs preferentially cuts directly at the double-stranded/single-stranded junction (3,10). At a 3Ј-overhang, ARTEMIS preferentially cuts 4 nucleotides into the single-stranded overhang from the double/ single-stranded DNA junction (11). All of the endonuclease activities of ARTEMIS on duplex DNA ends can be unified by a model invoking binding to the single-to double-stranded junction and occupying 4 nt along the single-stranded portion, followed by nicking preferentially 3Ј of those 4 nt (3).
In addition to its endonucleolytic action on the single-strand to double-stranded transitions (overhangs and hairpins of duplex DNA), ARTEMIS alone has a weaker single-stranded DNA endonuclease activity, such as on poly(dT) (12). Similar to the endonuclease activity of ARTEMIS on duplex DNA substrates, the action of ARTEMIS on entirely single-stranded DNA is stimulated by the addition of DNA-PKcs (12), which forms a complex with ARTEMIS that is stable to well over 0.5 M monovalent salt (3). Our model for ARTEMIS endonucleolytic action on single-stranded DNA (ssDNA) proposes binding by ARTEMIS at random locations along the ssDNA, followed by nicking, and these activities of ARTEMIS are increased by DNA-PKcs.
Purified ARTEMIS protein also appears to have 5Ј-exonucleaseactivity,whichisDNA-PKcs-independent(3).The5Ј-exonuclease activity co-fractionates with the known endonucleolytic activities of ARTEMIS (12). Exonucleases are abundant in cells and are common contaminants. As other 5Ј-exonucleases are purified away, the apparent level of 5Ј-exonuclease activity falls, but not to zero, and the question arises whether the residual level of 5Ј-exonuclease activity is intrinsic to ARTEMIS or a contaminating enzyme that co-fractionates with ARTEMIS.
Similar to other members of the metallo-␤-lactamase family, it is thought that ARTEMIS requires the binding of divalent cations to two co-catalytic portions of the active site for catalysis (13). Each catalytic site may hold one Mg 2ϩ , Mn 2ϩ , or Zn 2ϩ ion. In the metallo-␤-lactamase family, the ligands for the divalent cations are generally conserved and consist of a combination of histidine and aspartic acid residues (14). Our previous studies have demonstrated that upon mutation of these conserved residues, the endonuclease activity of ARTEMIS⅐DNA-PKcs can be markedly reduced. However, the 5Ј-exonuclease activity appeared to remain (13). This raised the possibility that the 5Ј-exonuclease activity and the endonuclease activity of ARTEMIS use two independent catalytic sites. However, an alternative explanation is that previous attempts at testing point mutants were carried out using 293T cells, a system known for high levels of background nucleases, and complete removal of contaminating 5Ј-exonucleases may not have been successful.
Here, we present a point mutant of ARTEMIS created using a baculovirus expression system, a system that more readily permits purification away from host cell 5Ј-exonucleases. This mutation markedly reduces both the endonuclease as well as the 5Ј-exonuclease activities of ARTEMIS. We have also identified small molecules that inhibit both the ARTEMIS endonuclease and 5Ј-exonuclease activities in parallel. These and other findings shed light on ARTEMIS nuclease activity and expand the range of substrates that ARTEMIS can act upon. We discuss these findings in the context of NHEJ.

EXPERIMENTAL PROCEDURES
Oligonucleotides and DNA Substrates-Oligonucleotides used in this study were synthesized by Operon Biotechnologies, Inc. (Huntsville, AL) and Integrated DNA Technologies, Inc. (San Diego, CA). We purified the oligonucleotides using 12 or 15% denaturing PAGE and determined the concentration spectrophotometrically. DNA substrate 5Ј end-labeling was done with [␥-32 P]ATP (3000 Ci/mmol) (PerkinElmer Life Sciences) and T4 polynucleotide kinase (New England Biolabs) according to the manufacturer's instructions. Substrates were incubated with [␥-32 P]ATP and T4 PNK for 30 min at 37°C. T4 PNK was subsequently inactivated by incubating samples at 72°C for 20 min. Unincorporated radioisotope was removed by using G-25 Sephadex (Amersham Biosciences, Inc.) spin column chromatography. For the hairpin substrate, YM164-labeled oligonucleotide was diluted in a buffer containing 10 mM Tris-HCl (pH 8.0), 1 mM EDTA (pH 8.0), and 100 mM NaCl, heated at 100°C for 5 min, allowed to cool to room temperature for 3 h, and then incubated at 4°C overnight. The sequences of the oligonucleotides used in this study are as follows: JG 169, 5Ј-TTT TTT TTT  TTT TTT TTT TTT TTT TTT TTT-3Ј;  Protein Expression and Purification-ARM14 recombinant baculovirus with a C-terminal His 8 tag was cloned into baculodirect vector and transfected into Sf21 according to protocol (Invitrogen). Virus was harvested after 48 h and used to infect Sf9 cells in log phase growth. Wild type ARTEMIS was purified as described previously (12). Harvested ARM14-infected Sf9 cells were resuspended in Ni-NTA binding buffer (50 mM NaH 2 PO 4 (pH 7.8), 0.5 M KCl, 2 mM ␤-mercaptoethanol, 10% glycerol, 0.1% Triton X-100, and 20 mM imidazole (pH 7.8)) supplemented with protease inhibitors. Cells were lysed by sonication and applied to Ni-NTA-agarose resin (Qiagen). Bound ARM14 protein was eluted off using 250 mM imidazole. Eluted protein was directly applied to DEAE-Sepharose resin (GE Healthcare). Unbound soluble fraction was collected and dialyzed against Superose 12 buffer (500 mM KCl, 25 mM Tris-HCl, 1 mM DTT). ARM14 fractions eluted from Superose 12 were frozen down and stored at Ϫ80°C. Expression and purification of DNA-PKcs from HeLa cells has been described previously (15). Circular Dichroism-All CD studies were conducted on Jasco-810 Spectro Polarimeter. A wavelength of 260 nm was used at the start and 190 nm for the end reading. A data pitch of 0.5 nm was used with continuous reading and 20 accumulations per scan.
Biochemical Nuclease Assay-The in vitro DNA nuclease assay was performed in a volume of 10 l with a buffer composition of 25 mM Tris-HCl (pH 7.5), 10 mM KCl, 10 mM MgCl 2 , and 1 mM DTT. In the reaction, 50 nM single-stranded DNA or 20 nM hairpin DNA YM164 or SL23 was mixed with 50 nM ARM14 or 50 nM wild-type ARTEMIS. In specified cases, 10 mM MnCl 2 replaced 10 mM MgCl 2 . In specified cases, 50 nM DNA-PKcs was added with 0.5 mM ATP. DNase I and micrococcal nuclease were obtained from New England Biolabs and Sigma, respectively. Reactions were then incubated at 37°C for 1 h. After incubation, reactions were stopped and analyzed on a 12% denaturing PAGE.
Cells and Transfection-HEK293T cells were transfected by nucleofection using AMAXA nucleofector kit V (Lonza). Equal transfection efficiencies were confirmed by co-transfection of an enhanced green fluorescent protein expression plasmid followed by FACS analysis.
Immunoprecipitation and Immunoblotting-The immunoprecipitation was performed as described (3). Myc-tagged ARTEMIS was immunoprecipitated and detected with anti-Myc antibody (Invitrogen). SV40 large T antigen was used as loading control and was detected with anti-SV40 T-Ag antibody (Santa Cruz Biotechnology, Santa Cruz, CA). Endogenous DNA-PKcs was detected using anti-DNA-PKcs antibody (Abcam). RESULTS ARTEMIS Endonuclease and 5Ј-Exonuclease Activities Copurify with WT ARTEMIS Protein-ARTEMIS has a well documented set of endonuclease activities, all of which are stimulated by DNA-PKcs (3). The 5Ј-exonuclease activity of ARTEMIS co-fractionates with the ARTEMIS endonucleolytic activities (Fig. 1). The 5Ј-exonuclease activity is seen as the 1-nt product at the bottom of Fig. 1A. The 5Ј-exonuclease activity is not stimulated by DNA-PKcs (3), and for this reason, we wondered whether the 5Ј-exonuclease is intrinsic to ARTEMIS.
Expression and Purification of ARM14 Point Mutant-We created a point mutant of ARTEMIS by mutating the conserved histidine at amino acid 115 to an alanine ( Fig. 2A). The histidine at amino acid 115 is a putative binding site for divalent cations involved in the catalysis of DNA (13). This was subsequently cloned into a baculodirect vector (Invitrogen) to generate what we call the ARM14 baculovirus and mutant form of ARTEMIS protein. The ARM14 mutant and WT human ARTEMIS-His proteins were both expressed in Sf9 insect cells using a baculovirus expression system. Both the WT and ARM14 mutant proteins were purified using Ni-NTA, an anion exchange step, followed by a Superose 12 gel filtration column. Purified ARM14 mutant elutes as a relatively sharp peak off of the Superose 12 column and runs at 100 kDa on 8% SDS-PAGE (Fig. 2, B and D). The ARM14 mutant runs at the same position (ϳ100 kDa) as the WT ARTEMIS on 8% SDS-PAGE (Fig. 2C). To ensure that the ARM14 mutant was not denatured, we used circular dichroism (CD) on the WT ARTEMIS, on ARM14, and on completely heat-denatured ARTEMIS. The WT ARTEMIS and ARM14 had similar CD profiles (data not shown). Hence, purified ARM14 mutant has a conformation that is indistinguishable from WT ARTEMIS by CD.
Additionally, we performed a pulldown experiment to assess the ability of ARM14 protein to bind to endogenous DNA-PKcs. ARTEMIS or ARM14 expression plasmids were transfected into HEK293T cells. Myc-ARTEMIS and Myc-ARM14 were immunoprecipitated using anti-Myc antibody. We assayed to determine whether DNA-PKcs was also co-immunoprecipitated in both cases. The presence of DNA-PKcs was assayed using anti-DNA-PKcs antibody. We find that the ARM14 point mutant is able to bind to DNA-PKcs with indistinguishable affinity as WT ARTEMIS (Fig. 2F).
The ARM14 Mutant of ARTEMIS Does Not Possess Hairpin Opening, Endonuclease, or 5Ј-Exonuclease Activity-We tested the activity of ARM14 fractions collected from the Superose 12 column for nuclease action on a DNA hairpin and ssDNA substrates, specifically, a single-stranded poly(dT) substrate. The WT ARTEMIS endonuclease activity elutes as a symmetrical and defined peak. Unlike WT ARTEMIS protein, an equal amount of ARM14 mutant protein had no endonuclease activity on ssDNA (Fig. 2E, lanes 2-5). Importantly, no 5Ј-exonuclease activity was detected, as demonstrated by the lack of a band at the 1-nt position at the bottom of the gel (Fig. 3A, lane 2 versus lane 3). Therefore, the ARM14 mutant does not exhibit any endonuclease activity or 5Ј-exonuclease activity.
In some DNA DSBs, the 5Ј termini that are generated can be 5Ј-hydroxyl rather than 5Ј-phosphate. We tested for the ability of the intrinsic 5Ј-exonuclease activity of ARTEMIS to process a 3Ј-labeled substrate with a 5Ј-hydroxyl group. We labeled the substrate at the 3Ј end by using terminal deoxynucleotidyl transferase and dATP that has a radioactive phosphate at the ␣ position. To minimize addition of more than one radiolabeled mononucleotide, we include a 5-fold excess of ddTTP, which caps the 3Ј terminus. However, the final substrate is a mixture of lengths due to incorporation of 1, 2, or 3 (and smaller amounts of more) radiolabels, followed by one dideoxy mononucleotide. We tested the processitivity of the 5Ј-exonuclease activity of ARTEMIS using this 3Ј-labeled substrate at three time points. Unlike processive nucleases, ARTEMIS is able to generate intermediate products even at the 30-min time point (Fig. 3B, lane 4). Therefore, ARTEMIS acts nonprocessively and can process 5Ј-hydroxyl termini as shown here (Fig. 3B) and substrates with 5Ј-phosphate termini, as we have shown previously (see Fig. 3B in Ref. 12).
We compared the activity of ARM14 with WT ARTEMIS on a hairpin substrate. To do this, we designed an oligonucleotide DNA substrate that is able to fold back on to itself upon annealing to form a hairpin substrate (YM164). WT ARTEMIS protein shows weak hairpin opening, weak endonuclease activity, as well as weak 5Ј-exonuclease activity on this substrate in Mg 2ϩ -containing reactions that lack DNA-PKcs (Fig. 3C, lane 2). In contrast, the ARM14 protein generated no hairpin nicking, endonuclease products or 5Ј-exonuclease products even when we added DNA-PKcs to the reaction (Fig. 3C, lane 5). A, the activity of ARM14 mutant was assayed for activity and compared with that of WT ARTEMIS. 50 nM ARM14 mutant or WT ARTEMIS was incubated with 50 nM ssDNA poly(dT) substrate in a 10-l reaction. Reactions were incubated for 60 min at 37°C, stopped, and run on 12% denaturing PAGE. The 5Ј-exonuclease activity is seen as the 1-nt product at the bottom of the gel. B, the activity of WT ARTEMIS was assayed for activity on a 3Ј-labeled substrate to determine the processitivity of the 5Ј-exonuclease activity. 50 nM WT ARTEMIS was incubated with 50 nM of ssDNA poly(dT) substrate (3Ј-labeled by terminal deoxynucleotidyl transferase) in a 10-l reaction. Reactions were incubated for either 2, 5, or 30 min as indicated. Where compound SBI-713 is used, 20 M was added to the reaction. C, 50 nM hairpin DNA substrate (YM164) was incubated with 50 nM WT ARTEMIS or 50 nM ARM14 in a 10-l reaction. Reactions were incubated at 37°C for 60 min. In lanes where DNA-PKcs was added, 50 nM DNA-PKcs, 0.5 mM of ATP, and 0.25 M cold DNA YM8/9 were also added. Lane 1 contains only the hairpin substrate. Lane 2 contains ARTEMIS, which does have low dsDNA endonucleolytic activity without DNA-PKcs being present, but it has high 5Ј-exonucleolytic activity (1 nt product at bottom of gel). Lanes 4 and 5 are the ARM14 point mutant of ARTEMIS. Lane 3, WT ARTEMIS plus DNA-PKcs, which generates a hairpin opened product, and 5Ј-overhang cleavage products. Lane 6 is the control lane containing DNA-PKcs but no ARTEMIS.
The Endonuclease and 5Ј-Exonuclease Activities of ARTEMIS Can Be Inhibited in Parallel-We further tested the ability of selected compounds to inhibit the hairpin opening, endonuclease, and 5Ј-exonuclease activities of ARTEMIS. Prior to the high throughput screen, a small screen showed that ampicillin, a ␤-lactamase antibiotic that inhibits transpeptidase, is also inhibitory, probably because ARTEMIS is a member of the ␤-lactamase family. SBI-713 and SBI-116 are two structurally distinct Ͻ500 Da small molecules (data not shown) that were identified using a high throughput screen of a 31,000 compound subset of the National Institutes of Health Molecular Libraries Small Molecule Repository collection (Pubchem Bio-Assay AID 686956).
Given that ARTEMIS is an endonuclease, we wondered whether ARTEMIS cutting at the 5Ј most nt could merely reflect a preferential cutting by ARTEMIS endonuclease activity. To distinguish this from its 5Ј-exonuclease activity, we utilized an ssDNA poly(dA) substrate. Single-stranded polypurine substrates are not endonucleolytically cut as efficiently by ARTEMIS, which prefers polypyrimidine single-stranded DNA substrates (9). From our previous work, we knew that the 5Јexonuclease activity of ARTEMIS is DNA-PKcs-independent (3). We tested the exonuclease activity of ARTEMIS in the presence and absence of DNA-PKcs (Fig. 4B, lane 2 versus lane 3). Consistent with our previous findings, we saw no difference in the exonuclease activity of purified WT ARTEMIS with or without DNA-PKcs. The addition of 20 M of the inhibitor SBI-713 was able to inhibit the 5Ј-exonuclease activity of ARTEMIS, as demonstrated by the lack of the 1-nt exonuclease product (Fig. 4B, lane 4). This further demonstrates the 5Ј-exonuclease activity of ARTEMIS. Furthermore, we tested the ability of the ARM14 point mutant to process the poly(dA) ssDNA. The ARM14 mutant does not have any apparent endonuclease or 5Ј-exonuclease activity on this substrate when compared with WT ARTEMIS (Fig. 4C, lane 3 versus lane 2), indicating that the point mutation markedly reduces both endonuclease and 5Ј-exonuclease activities.
We also tested the inhibitors in titrations to determine whether the 5Ј-exonuclease and the endonuclease of ARTEMIS are inhibited over the same concentration ranges (Fig. 5). We found that as the concentration of ampicillin or SBI-713 increases, there is a concomitant inhibition of both the endonuclease and the 5Ј-exonuclease activity of ARTEMIS. The level of inhibition of the 5Ј exo-and endonuclease activities is similar for both inhibitors. If the 5Ј-exonuclease were a contaminant, we would not expect both the ARTEMIS endonuclease and the contaminating 5Ј-exonuclease to be inhibited over the very same concentration ranges for each drug (Fig. 5).
We have also tested SBI-713 and ampicillin against other nucleases to assess specificity for inhibition of ARTEMIS. These included DNase I and micrococcal nuclease. SBI-713 does not inhibit either of these counterscreen nucleases (Fig.  4D).
In related studies, in 10 mM MgCl 2 reaction buffers, addition of ZnCl 2 above 10 M inhibits ARTEMIS endonuclease activity (Fig. 6). It has been shown previously in certain cases that excess zinc can inhibit the enzymatic activity of zinc-dependent enzymes. This is thought to be caused by excess zinc-monohydroxide binding to the active form of the zinc-dependent enzyme, thereby preventing the catalytic site from binding active zinc (17). In the case of ARTEMIS, the inhibition by excess zinc is observed for the endonuclease activity as well as the 5Ј-exonuclease activity over the same Zn 2ϩ concentration range (10 to 100 M) (Fig. 6). This is consistent with the 5Ј-exonuclease activity of ARTEMIS sharing the same catalytic site as the endonuclease activity of ARTEMIS.

DISCUSSION
Although a 5Ј-DNA exonuclease activity co-purifies with ARTEMIS ( Fig. 1), the 5Ј-exonuclease is not dependent on DNA-PKcs, whereas the ARTEMIS DNA endonuclease activities (5Ј-and 3Ј-overhang cleavage and hairpin opening) do depend on DNA-PKcs. For this reason, we did further studies described here to evaluate whether the 5Ј-exonuclease is intrinsic to ARTEMIS.
Here, we have provided several independent lines of evidence that the 5Ј-exonuclease is intrinsic to ARTEMIS. First, the 5Ј-exonuclease and the endonuclease activities co-fractionate.
Second, a point mutant of ARTEMIS is markedly reduced in 5Ј-exonuclease activity in the presence and absence of DNA-PKcs. A point mutant of ARTEMIS is very unlikely to affect any contaminating 5Ј-exonuclease. This alone is strong evidence of an intrinsic 5Ј-exonuclease.
Third, both the endonuclease and the 5Ј-exonuclease activity of ARTEMIS can be inhibited by small molecules that are not chelators. It is unlikely that any one inhibitor is able to inhibit both ARTEMIS endonuclease activity and an unidentified contaminating 5Ј-exonuclease. It is even more unlikely that these specific inhibitors would affect a contaminating 5Ј-exonuclease at the same concentrations as they affect ARTEMIS endonuclease activity.
Fourth, disruption of the catalytic sites by divalent cation replacement is known to alter the activity of enzymes in the metallo-␤-lactamase family of nucleases (18). We found that the addition of excess zinc (Ͼ10 M) to the reaction buffer decreases the endonuclease as well as the 5Ј-exonuclease activity of ARTEMIS. It seems unlikely that a contaminating 5Јexonuclease would also show this behavior, particularly at the same low (micromolar) concentrations of Zn 2ϩ .
Fifth, we explored the possibility that the 5Ј-exonuclease activity seen in our gels is actually the result of the ARTEMIS endonuclease activity acting on the 5Ј-terminal nucleotide. We created a poly(dA) ssDNA substrate that is not as readily endonucleolytically processed by the ARTEMIS endonuclease activity but conceivably might be processed by the 5Ј-exonuclease activity of ARTEMIS. Indeed, we do observe 5Ј-exonuclease activity when using this poly(dA) ssDNA substrate, indicating that the activity seen at the 5Ј end on a polypyrimidine substrate is attributable to 5Ј-exonuclease activity and not merely due to theendonucleaseactivityofARTEMIS.Importantly,this5Ј-exonuclease activity is completely inhibited by the SBI-713 compound (Fig. 4). From the evidence presented here, we conclude that the 5Ј-exonuclease activity seen in purified ARTEMIS is intrinsic to the ARTEMIS protein and not due to a co-purifying contaminating enzyme.
The metallo-␤-lactamase family of nucleases uses several conserved catalytic residues to coordinate divalent cations. The two most conserved residues within the catalytic site are histidine and aspartic acid (14). We generated the ARM14 point mutant, which changed a conserved histidine residue to an alanine residue. If both the endonuclease activity of ARTEMIS and 5Ј-exonuclease activity of ARTEMIS shared a common catalytic site, by disrupting the divalent ligand in the site there should be a corresponding decrease in both endonuclease as well as the 5Ј-exonuclease activity of ARTEMIS, and this is what we observe here.
We have previously purified ARM14 mutant from 293T cells using a two-step purification. Although this point mutant is inactive for the hairpin opening and endonuclease activity on a hairpin substrate, this version of ARM14 appeared to still retain most of the 5Ј-exonuclease activity (13). We attribute the discrepancies in our findings with insect cell-purified ARM14 in the current study to our earlier study to the high levels of contaminating nucleases in mammalian cells that are difficult to remove during purification. The baculovirus system for ARTEMIS production here contains much lower levels of background contaminating nucleases, which can be subsequently removed in our three step purification.
Using a purification process consisting of phosphocellulose, Ni-NTA-agarose, and hydroxyapatite, one group separated the usual bulk of other cellular 5Ј-exonuclease activity away from the endonuclease activity of ARTEMIS (19). They concluded that the 5Ј-exonuclease activity was not intrinsic to ARTEMIS based on an assay that was much less sensitive than our assay. Even at their reduced sensitivity, we note that some 5Ј-exonuclease activity is still apparent in their assay gels, which they did not point out (19). For example, they tested their purified ARTEMIS on a single-stranded 5Ј radiolabeled DNA substrate to assay for 5Ј-exonuclease activity (Fig. 5B of Ref. 19). Although they could detect no 5Ј-exonuclease activity, their assay was insensitive to the extent that they also could not detect any of the single-stranded endonuclease activity of ARTEMIS (19). This was due in part to exposure intensities of their gels that were 10-to 20-fold lower than what we use to detect both single-stranded endo-and 5Ј-exonuclease activities (12). Despite the fact that their gels show apparent residual 5Ј-exonuclease activity at their darker radiolabel exposures (Fig. 5C, lanes 10 & 11 in ref. (19)), they nevertheless concluded that ARTEMIS has no intrinsic 5Ј-exonuclease activity (19). Reactions were incubated at 37°C for 30 min, stopped, and analyzed by 12% denaturing PAGE. Quantitation indicates that similarly low levels of both the endonuclease and the 5Ј-exonuclease activity (1-nt product) escape inhibition by ampicillin. B, addition of the compound SBI-713 inhibits Artemis on a single-stranded poly(dT) substrate. The ssDNA substrate (50 nM JG169) was incubated with 50 nM ARTEMIS in a 10-l reaction. SBI-713 was added in concentration(s) as indicated. Reactions were incubated at 37°C for 30 min, stopped, and analyzed by 12% denaturing PAGE. SBI-713 inhibition of both the 5Ј-exonuclease and the endonuclease is complete at 10 and 20 M. A small level of residual 5Ј-exonuclease and endonuclease persists even at the highest levels of ampicillin tested.
Other members of the metallo-␤-lactamase family possess nuclease activities similar to those of ARTEMIS. CPSF-73, an enzyme involved in the 3Ј end processing of histone pre-mRNA, has been shown to possess endonuclease activity as well as 5Ј-exonuclease activity on single-stranded RNA substrates (18,20). The nuclease activities of CPSF-73 can be abolished by the addition of zinc specific chelators. We find that ARTEMIS nuclease activities (both endo-and 5Ј-exonuclease) are abolished by addition of these same zinc-specific chelators (data not shown), suggesting that members of this family use the same divalent cations in the catalytic site (21). Biochemical studies have also demonstrated that human SMN1A and SMN1B proteins, also members of the metallo-␤-lactamase family and involved in inter-strand cross-link DNA repair, are capable of acting as a 5Ј-exonuclease on single-stranded and double-stranded DNA substrates (16), and this is its only biochemically defined enzymatic activity. Mutations of key cata-lytic residues within the metallo-␤-lactamase domain also abolish the 5Ј-3Ј-DNA processing ability of hSMN1a and hSMN1b. Therefore, 5Ј-exonuclease activity is intrinsic to many members of the metallo-␤-lactamase nuclease family, and the catalytic site involved during DNA processing resides in the metallo-␤-lactamase domain. Our current study clearly shows that ARTEMIS also has a similar 5Ј-exonuclease activity.
The 5Ј-exonuclease activity described here would permit ARTEMIS to act on 1-and 2-nt 5Ј-overhangs more efficiently because such short 5Ј-overhangs are not optimal substrates for action by ARTEMIS in an endonucleolytic manner (3). This would be of particular utility for repair of ionizing radiation damage, which might have short overhangs with either a 5Ј-phosphate or 5Ј-hydroxyl.