Mutational Analysis of the RecJ Exonuclease of Escherichia coli: Identification of Phosphoesterase Motifs

Abstract

The recJ gene, identified in Escherichia coli, encodes a Mg⁺²-dependent 5′-to-3′ exonuclease with high specificity for single-strand DNA. Genetic and biochemical experiments implicate RecJ exonuclease in homologous recombination, base excision, and methyl-directed mismatch repair. Genes encoding proteins with strong similarities to RecJ have been found in every eubacterial genome sequenced to date, with the exception of Mycoplasma and Mycobacterium tuberculosis. Multiple genes encoding proteins similar to RecJ are found in some eubacteria, including Bacillus and Helicobacter, and in the archaea. Among this divergent set of sequences, seven conserved motifs emerge. We demonstrate here that amino acids within six of these motifs are essential for both the biochemical and genetic functions of E. coli RecJ. These motifs may define interactions with Mg²⁺ ions or substrate DNA. A large family of proteins more distantly related to RecJ is present in archaea, eubacteria, and eukaryotes, including a hypothetical protein in the MgPa adhesin operon of Mycoplasma, a domain of putative polyA polymerases in Synechocystis and Aquifex, PRUNE of Drosophila, and an exopolyphosphatase (PPX1) of Saccharomyces cereviseae. Because these six RecJ motifs are shared between exonucleases and exopolyphosphatases, they may constitute an ancient phosphoesterase domain now found in all kingdoms of life.

The recJ gene of Escherichia coli encodes a Mg²⁺-dependent single-strand DNA (ssDNA) specific exonuclease with 5′-to-3′ polarity (19). The recJ gene was first identified by its effects on recBCD-independent recombination pathways; mutants deficient in both the RecBCD and RecJ nucleases show extreme recombination deficiency, sensitivity to UV irradiation, and partial cell inviability (18). Subsequently, RecJ was shown to affect the efficiency of RecBCD-mediated recombination as well (22, 26, 30). In addition, RecJ is one of three exonucleases that have been implicated in the excision reaction of the methyl-directed mismatch repair pathway responsible for removing polymerase errors (6). Removal of abasic deoxyribose phosphate residues in DNA, produced either spontaneously or by enzymatic removal of altered bases, may also be mediated by RecJ (7).

The ubiquity of RecJ-related proteins in eubacteria points to its ancient origin and important cellular function. Predicted proteins homologous to RecJ can be found in virtually all bacterial species for which we have genomic sequences, including eight major eubacterial divisions: Proteobacteria, Cyanobacteria, gram-positive bacteria, spirochetes, Thermotogales, radioresistant bacteria, Aquificales, and the chlamydiae. Within this diverged set of RecJ amino acid sequences, seven conserved motifs emerge. In this report, we present evidence that these conserved amino acids are essential for RecJ exonuclease activity in vitro and genetic activity in vivo. Based on similarity to motifs present in the 5′ exonucleases of Taq polymerase I and T4 RNase H, for which crystal structures have been determined (16, 24), these motifs within RecJ are likely to be sites of interaction with Mg²⁺ ions and phosphates of the substrate DNA.

Archaeal species also possess two or three proteins with reasonably strong similarity to RecJ; among these, six of the RecJ motifs are largely conserved. A larger group of other proteins in the database, including those from eubacterial, archaeal, and eukaryotic species, also possess these RecJ motifs but are otherwise distantly related to RecJ (2). These include, among others, a domain in polyA polymerase from Synechocystis and Aquifex, the 28-kDa protein of the MgPa adhesin operon in Mycoplasma genitalium, an exopolyphosphatase from yeast, and the PRUNE protein from Drosophila. These findings suggest that the conserved RecJ motifs constitute an ancient structure characteristic of the phosphoesterases now found in proteins from all three domains of life.

MATERIALS AND METHODS

Plasmid constructions.

Plasmid DNA was prepared by alkali lysis (3). Transformation of plasmid DNA into the appropriate strain backgrounds was performed by electroporation (8). Selection for plasmids employed Luria broth (23) containing 60 μg of kanamycin (Km) or 100 μg of ampicillin (Ap) per ml. Tetracycline (Tc) was used in some experiments at a concentration of 10 μg/ml to select for chromosomal Tn10 mutations.

Site-directed mutants were constructed by two methods. RecJ-D81A and RecJ-D79A were constructed using the T7 In Vitro Mutagenesis Kit (United States Biochemical, Amersham); the remainder of the site-directed mutants were obtained using modifications (as suggested by New England Biolabs) to the Kunkel method (17). Single-strand plasmid template was obtained from the phagemid pSTL175 with the help of M13-K07 helper phage (28). pSTL175 contains the wild-type sequence of recJ fused to an optimal ribosome binding sequence derived from plasmid pRDK115 (20) in PstI-EcoRI sites of the high-copy vector pBluescript SK(−) (Stratagene). Primer sequences for site-directed mutagenesis are as follows: recJD79A, GGCGTCGAAATTACCGACCAC; recJD81A, GTCGGCGTCGGCATCACCGAC; recJD83A, GGTCGCGCCGGCGGCGTCGAA; recJD137A, AATACCGTTAGCCACCACGAC; recJH161A, TGGCAAATGGGCATCGGTAAC; recJH162A, GCCTGGCAAAGCGTGATCGGT; recJD236A, CGGCACGACGGCCGCCACTGT; recJR401A, CTGAATGGAGGCACCGGAACC; and recJH429A, CGCCATCGCAGCACCGCCAAA. Mutations were confirmed by DNA sequence analysis using a Sequenase version 2.0 DNA sequencing kit (Amersham) or a SequiTherm EXCEL DNA sequencing kit by Epicentre Technologies.

For biochemical and genetic analyses, the recJ mutant alleles were subcloned from the relevant high-copy pBluescript SK(−) plasmids into the low-copy vectors pWSK29, pWKS30, and pWKS130 (31) using restriction endonucleases ApaI and SacII. In the pWSK29 derivatives, the recJ genes are controlled by the T7 RNA polymerase promoter; in pWKS30 and pWKS130 derivatives, the genes are expressed from the lac promoter. pWSK29 and pWKS30 encode Ap resistance, and pWKS130 encodes Km resistance.

Mutation D160A was produced by random mutagenesis in a mutD mutator strain (STL913, F⁻ mutD5 lacZ trpA metE). The recJ⁺ gene was subcloned from plasmid pSTL175 into low-copy plasmid pWKS130 (31) with enzymes ApaI and SacII, producing pSTL245. After several cycles of mutagenic growth in STL913, plasmid DNA was harvested and transformed into RDK1656 (recJ284::Tn10 recB21 recC22 sbcA23 Rac⁺ derivative of AB1157). Candidate recJ mutants were identified by small colony morphology and UV sensitivity, as described below. The mutation D160A was identified by DNA sequence analysis in one of these isolates, and the recJD160A allele was subcloned into vectors pWKS30 and pWKS130.

Strain construction.

A deletion allele of recJ, recJΔ2060::Tn10-9, was constructed as follows. DNA from pRDK163, carrying recJ2003::Tn10-9, and pRDK161, carrying recJ2051::Tn10-9 (21), was cut with XhoI and AlwN, mixed, and ligated. A resultant plasmid, pSTL41, carries a deletion of recJ from nucleotides 124 to 1252, relative to the GTG initiation codon of recJ, which is marked with a single copy of Tn10-9, encoding Km resistance. By selection for Ap resistance, pSTL41 was transformed into strain RDK1445, a zgb-224::Tn10 serA6 derivative of AB1157. The zgb-224::Tn10 marker is tightly linked to recJ in transductional crosses (17a). Integrants of pSTL41 into the E. coli chromosome were selected by preparing a P1virA transducing lysate (23) on the pSTL41 transformant of RDK1445 and using it to transduce strain JC8679 (recB21 recC22 sbcA23 Rac⁺ derivative of AB1157) to Tc and Km resistance. The resulting strain, STL3801, carried an integration of the entire pSTL41 plasmid and was Ap resistant. Km^r Ap^s segregants that had deleted the plasmid but retained the recJΔ2060::Tn10-9 mutation were isolated, producing strain STL3803.

Nuclease activity and protein stability assays.

A two-plasmid system was used to express RecJ proteins from the T7 promoter. The relevant pWSK29 mutant derivatives were introduced by transformation in strain STL327 (F⁻ recJ284::Tn10 sbcB15 endA Δ(xth-pncA) Δ(tet) gal thi) that harbors plasmid pGP1-2 (29) on which the T7 RNA polymerase is under control of the thermolabile cI857 repressor. Overexpression was induced by growth at high temperature as previously described (20), and rifampin (200 μg/ml) (Sigma) was added to inhibit E. coli RNA polymerase. Crude cell extracts were prepared by lysozyme treatment and were assayed for nuclease activity as previously described (20). Substrate DNA consisted of denatured bacteriophage T7 DNA which had been uniformly labeled with [³H]thymidine (12). The reaction buffer contained 20 mM Tris HCl (pH 8.5), 10 mM MgCl₂, 0.67 mM dithiothreitol, 1 mg of bovine serum albumin per ml, and 10 μg of substrate DNA per ml. One unit of nuclease activity corresponds to the production of 1 nmol of acid-soluble nucleotides in 20 min at 37°C. The protein concentration in the extracts was determined by the method of Bradford (5), using reagent from Bio-Rad and bovine serum albumin as the standard.

To measure the stability of mutant proteins, expression was induced as described above in the presence of rifampin. Cultures were prepared in minimal medium using 56/2 salts (32) and 0.4% glucose. Cellular proteins were labeled with [³⁵S]methionine (NEN Dupont) in a 5-min pulse as described previously (20). The cultures were split, and one was harvested immediately and the other was allowed to grow for an additional 10 h with the addition of an unlabeled methionine chase at a final concentration of 0.05%. Cells were then harvested and resuspended in electrophoresis sample buffer, and the proteins were resolved by polyacrylamide gel electrophoresis (3). Gels were dried, and the radioactivity in the band corresponding to RecJ was determined by phosphorimage analysis (Molecular Dynamics, Inc.) using ImageQuant software. The radioactivity in the RecJ band of the 10-h chased samples was compared to that of the immediate pulse to determine the stability of individual mutant proteins.

UV survival assays.

Genetic experiments for complementation and dominance employed recJ mutant constructs in the pWKS30 or pWKS130 vectors described above where recJ expression originates from the lac promoter. Plasmids were introduced by transformation into either STL3803 (recB21 recC22 sbcA23 recJΔ2060::Tn10-9) or JC8679 (recB21 recC22 sbcA23). Transformants were grown in Luria-Bertani agar plus Km and Ap (STL3803) or Luria-Bertani agar plus Km (JC8679). Early-exponential-phase cultures were split, and one half was allowed to grow in the original medium while the other half was grown in the same medium with 1 mM isopropyl-β-d-thiogalactopyranoside (IPTG) to induce expression of recJ from the lac promoter on the plasmid. After 2 h of growth, cultures were serially diluted in 56/2 salts, plated on 1.5% Luria-Bertani agar plus Km or Luria-Bertani agar plus Km and Ap. Plates were exposed to either 10 J of UV irradiation/m² or no irradiation. Plates were incubated overnight in the dark at 37°C. Surviving fractions were calculated as the number of CFU on UV-irradiated plates relative to the total number of CFU from the nonirradiated plate. Data reported are the averages of two to four experiments.

Database sequence analysis.

The PSI-BLAST program (1) was used to generate sequence alignments. The Geneman program (DNA Star, Inc.) was used to search for individual amino acid motifs. The sequence of the Chlamydophila caviae recJ gene was determined as described above from plasmid pRCH104 (15) kindly provided by Ru-Ching Hsia.

Nucleotide sequence accession number.

The sequence of the Chlamydophila caviae recJ gene was deposited in GenBank under accession no. AF058396.

RESULTS AND DISCUSSION

Alignment of eubacterial RecJs.

A multisequence alignment of the eubacterial RecJ-related proteins presently found the database is shown in Fig. 1. The PSI-BLAST program (1, 25) was used to generate initial alignments, although some portions were aligned manually. Not shown are sequences homologous to RecJ which are apparent in the incomplete genome sequences accessible through the National Center for Biotechnology Information (25): Actinobacillus actinomycetemcomitans, Bordetella pertussis, Campylobacter jejuni, Caulobacter crescentus, Chlorobium tepidum, Clostridium acetobutylicum, Deinococcus radiodurans, Enterococcus faecalis, Neisseria gonorrhoeae, Neisseria meningitidis, Pasteurella multocida, Porphyromonas gingivalis, Pseudomonas aeruginosa, Pseudomonas putida, Salmonella typhi, Shewanella putrefaciens, Staphylococcus aureus, Streptococcus pneumoniae, Streptococcus pyogenes, Thermotoga maritima, Thiobacillus ferrooxidans, Vibrio cholerae, and Yersinia pestis.

FIG. 1.

A sequence alignment with eubacterial RecJ-related sequences. Shown in parentheses are the number of amino acids not shown in the alignment. Shown in bold are residues conserved in at least 13 of the 14 eubacterial sequences. Motifs with invariant residues are numbered. Asterisks are placed above those residues mutated in this study. Sequences are listed in order of similarity to E. coli RecJ as determined by BLAST alignment (25). Eco, E. coli RecJ (PID2507106); Ech, Erwinia chrysanthemi (PID2507107); Hin, Haemophilus influenzae (PID1172895); Bsu1, B. subtilis YrvE (PID2635226); Rpr, Rickettsia prowazekii (PID3861077); Ctr, Chlamydia trachomatis (PID3328879); Ssp, Synechocystis sp. (PID2984349); Hpy1, H. pylori HP0348 (RecJ) (PID2313437); Cca, C. caviae (PID2126389); Aae, Aquifex aeolicus (PID2984349); Bbu, Borrelia burgdorferi (PID2688156); Tpa, Treponema pallidum (PID3323008); Bsu2, B. subtilis YorK (PID2634428); Hpy2, H. pylori HP0425 (PID2313533).

Excepting the Mycoplasma species and Mycobacterium tuberculosis, all the eubacterial genomes contain sequences whose products are closely related to E. coli RecJ. Bacillus subtilis has two RecJ homologs, YrvE and YorK. YorK appears to be encoded by a temperate bacteriophage, SPβc2. YrvE is more closely related to the proteobacterial forms than it is to YorK, but BLAST sequence alignment demonstrates that both proteins show strong similarity to E. coli RecJ (data not shown). Likewise, Helicobacter pylori has three RecJ-related sequences: one (HP0348) appears to be the true RecJ homolog, and two others, the products of duplicated gene segments (HP0425 and HP1410), are more weakly similar to the other eubacterial RecJs.

The amino acid sequence is quite diverged in this set (Fig. 1), and seven conserved motifs emerge, four of which contain six invariant aspartate residues. Five of these motifs have also been noted by others (2). These motifs are good candidates for interaction with Mg²⁺ ions, a cofactor required for RecJ-mediated hydrolysis of DNA (19). Motifs 1 and 2 of RecJ are somewhat similar to the ExoI and ExoII motifs of the PolI-Fen1 group of 5′ exonucleases (10, 11, 13, 27) (Fig. 2). The structures of three members of this group, Thermus aquaticus (Taq) polymerase, bacteriophage T4 RNase H, and Pyrococcus furiosus FEN-1, have been solved and shown to have similar features (14, 16, 24): the conserved carboxylates cluster at the bottom of a cleft and coordinate two (RNase H and FEN-1) or three (Taq Pol) metal ions. By analogy to the 3′ exonuclease of the Klenow fragment (E. coli PolI), the metal ions are believed to interact with the phosphate of the substrate, generate the attacking hydroxide ion, and facilitate product release (4).

FIG. 2.

Similarity of RecJ motifs 1 and 2 with ExoI and ExoII motifs as exhibited by members of the RAD2-FEN-1 family of 5′ exonucleases. Sequences include E. coli RecJ, T4 RNase H, Taq polymerase, T7 gene 6 protein, S. cerevisiae RAD2, human XP-G, and murine FEN-1 with respect to the published alignment of the RAD2-FEN-1 group (24).

Mutational analysis of E. coli RecJ.

Mutations within six of the conserved motifs were isolated either by random mutagenesis, selecting for loss of recJ genetic function in vivo, or by site-directed mutagenesis of plasmid-borne recJ. Tests were employed to measure the loss of RecJ function both biochemically and genetically. The mutant proteins were overexpressed from the T7 promoter, and exonuclease activity on ssDNA was assayed in crude extracts. The stability of mutant RecJ proteins was determined by [³⁵S]methionine pulse-chase experiments. In addition, the genetic function was tested by expressing the mutant RecJ proteins from the lac promoter on low-copy plasmids, both with and without IPTG induction, in various strain backgrounds. Complementation was tested by the UV survival of plasmid transformants of a strain carrying recJΔ2060::Tn10-9 on its chromosome, STL3803. Genetic dominance of the mutants was tested similarly by UV survival assays of plasmid transformants of a strain carrying recJ⁺ on its chromosome, JC8679.

Mutational analysis showed that all six motifs are essential for RecJ function (Table 1). Mutations in the conserved aspartates in motifs 1, 3, and 4 (D79A, D81A, D83A, D160A, and D236A) produced a severe (at least 400-fold) reduction of ssDNA exonuclease activity assayed in crude extracts. (Other nucleases in E. coli contribute to nuclease activity in crude extracts; therefore, reduction in RecJ activity below this level cannot be ascertained.) These mutant genes, even when overexpressed, could not genetically complement the UV repair deficiency conferred by a recJ chromosomal mutation. This group of mutations also had a genetic dominant-negative effect: they produced a RecJ⁻ (UV-sensitive) phenotype even in strains carrying recJ⁺ on the chromosome. For D79A, D81A, D83A, and D160A, this dominance was more severe when expression from the plasmid was induced with IPTG, indicating a competitive effect between the wild-type and mutant forms of RecJ. In contrast, the D236A mutation was dominant even without the induction of the lac promoter. All of the mutant proteins also produced a dominant effect on genetic recombination as measured after conjugation (data not shown). None of these dominant mutants expressed a noticeably unstable protein, even after overnight pulse-chase experiments. These results and their observed genetic dominance suggest that no gross conformational changes result from these mutations.

TABLE 1.

Biochemical and genetic analysis of RecJ mutants

RecJ allele	Mutated motif	Nuclease activity (U/mg of protein)a	Protein stability (%) after 10-h chaseb	UV survival (%) at 10 J/m2 in:				Genetic effect
				STL3803 (recJ mutant)c		JC8679 (recJ+)d
				Without IPTG	With IPTG	Without IPTG	With IPTG
Vector		5.7		0.0027	0.0085	17	17
recJ+		7,100	42	36	20	18	20
recJD79A	1	16	22	0.0031	0.0016	7.8	1.1	Dominant
recJD81A	1	18	59	0.0035	0.0061	8.7	1.4	Dominant
recJD83A	1	14	44	0.0018	0.0011	11.0	1.2	Dominant
recJD137A	2	13	25	0.0014	0.0035	9.8	9.2	Recessive
recJD160A	3	7.5	52	0.0059	0.0030	12	1.8	Dominant
recJH161A	3	15	54	0.0088	0.0015	16	13	Recessive
recJH162A	3	19	33	0.0019	0.0033	20	10	Recessive
recJD236A	4	11	29	0.0069	0.0059	0.61	0.48	Dominant
recJR401A	6	18	23	0.0076	0.0016	22	16	Recessive
recJH429A	7	63	39	0.0038	0.085	18	22	Recessive

^aNuclease activity in crude extracts of various pWSK29-RecJ plasmid transformants (of STL327/pGP1-2) after induction of T7 promoter. Data are averages of two to four experiments. 

^bPercentage of RecJ protein retained after 10-h chase as described in Materials and Methods after T7 promoter induction of various pWSK29-RecJ derivatives in STL327/pGP1-2 strains. 

^cComplementation tests of pWKS30-RecJ derivatives, with and without induction of the lac promoter. Data are averages of two to four experiments. 

^dDominance tests of pWKS130-RecJ derivatives, with and without induction of the lac promoter. Data are averages of two to four experiments. 

Mutations in the invariant residues of motif 2 (D137A), motif 6 (R401A), or motif 7 (H429A) or in the histidines of motif 3 (H161A and H162A) produced at least a 100- to 500-fold decrease in exonuclease activity and a recessive mutant phenotype. The H429A mutation appeared genetically “leaky” in two respects: it showed a level of exonuclease activity above background and that induction of high expression with IPTG partially suppressed its mutant phenotype. These recessive mutant proteins were not detectably unstable in overnight pulse-chase experiments, indicating that they most likely have a specific defect in catalysis rather than gross conformational alterations.

The genetic dominance of the aspartate mutants in motifs 1, 3, and 4 suggested that these mutant proteins in some way interfere with the normal functioning of wild-type RecJ protein in vivo. Genetic dominance is common among mutants of multimeric proteins. However, purified RecJ exonuclease is believed to be a monomer, as judged by gel filtration and glycerol gradient sedimentation criteria (9, 20a). We cannot, however, rule out the possibility that a RecJ multimer has other, unknown, functions or that interactions with other unknown partners in vivo account for the dominant-negative phenotype. Another explanation for the observed dominance may be that these proteins bind substrate but do not degrade it, acting as inhibitors of the wild-type RecJ and potentially of other enzymes as well. We have observed that Mg²⁺ is not required for the binding of DNA oligonucleotides by RecJ protein in gel-shift experiments (21a). These dominant mutant proteins may have lost the ability to bind one or more Mg²⁺ ions, which would negate exonucleolytic activity but not DNA binding. According to this latter scenario, the recessive mutants may have lost their ability to bind DNA substrate and therefore cannot interfere with the wild-type RecJ protein. Further experiments with purified mutant proteins should clarify the molecular basis of this observed dominance.

This study illustrates the usefulness of a large sequence set to identify catalytically important amino acid residues. The most pronounced genetic effects were obtained for D236A, a strongly dominant allele, and this residue was difficult to identify as conserved by many pairwise sequence comparisons. Motif 5 was likewise often not aligned in BLAST searches but was seen in certain position-iterated, PSI-BLAST alignments. Mutations in this motif are presently being tested for activity.

Multiple RecJ genes and a larger RecJ-related superfamily.

Helicobacter, Bacillus, and the archaeal genera possess more than one sequence related to RecJ. Two of the proteins in Helicobacter, HP0425 and HP1410, and their DNA sequences are virtually identical and are therefore the result of recent gene duplication. The archaeal forms are quite diverged from other forms within the same organism and therefore have resulted from more ancient duplication or by horizontal transfer from another species. One of the two Bacillus forms is encoded on a temperate bacteriophage. In other organisms, multiple genes could conceivably express enzymes with differential substrate specificities, reaction optima, or regulations. As RecJ may be involved in the repair of apurinic sites (7), those organisms that grow at low pH or high temperature may have a higher demand for RecJ activity because of elevated rates of spontaneous depurination.

Searches for individual motifs (motif 1, DXDG; motif 3, DHH; and motif 6, GGH) and position-specific iterated BLAST searches (1) with numerous RecJ-related proteins reveal a larger set of sequences more weakly related to RecJ. This set includes sequences from archaea, eubacteria, and eukaryotes. Many organisms harbor multiple members of this larger RecJ-related family; a total of seven sequences with RecJ motifs were found in Archaeoglobus, six each in Methanococcus and B. subtilis, four in Helicobacter, and three each in Synechocystis and Aquifex. We have recently demonstrated that one of the methanococcal recJ-related genes encodes a thermostable ssDNA exonuclease activity, similar to E. coli RecJ exonuclease in many of its properties (25a). This superfamily also includes the hypothetical 29-kDa protein of the MgPa adhesin operon (MG170) and numerous related proteins of Mycoplasma species. Two putative polyA polymerases from Aquifex and Synechocystis carry the RecJ motifs at their N termini, in domains distinct from their polymerase domains. The existence of this larger family of proteins has also been noted independently and described in detail by others (2).

Included in this larger group is the S. cerevisiae PPX1 protein, a cytoplasmic exopolyphosphatase that is believed to degrade storage polyphosphate (33, 34). Proteins in four other organisms (Archaeoglobus, Methanococcus, Streptococcus mutans, and Schizosaccharomyces pombe) have been designated exopolyphosphatases based on similarity to the S. cerevisiae protein. Polyphosphate is stored in many organisms, including bacteria and fungi, and some of these RecJ-related proteins may play a role in phosphate release. Other eukaryotic proteins, Drosophila melanogaster PRUNE protein and the products of two open reading frames from the protozoa Leishmania major, are also members of this subfamily. However, except for S. cerevisiae PPX1, E. coli RecJ, and the methanococcal RecJ-like protein mentioned above, whether individual members of this superfamily are polyphosphatases or DNA exonucleases has not been experimentally determined.

A common feature of the RecJ and PPX1 enzymes is phosphoesterase activity, and therefore these seven motifs may be predictive of phosphoesterases. Both RecJ and PPX1 activities are dependent on Mg²⁺ cofactor, and we presume that the motifs reflect a conserved architecture of Mg²⁺ coordination and phosphate interaction sites within the enzymes. The substrate range of the proteins within the PPX subfamily is not known but could conceivably include other phosphate esters including protein phosphoester bonds, DNA, RNA, or nucleotide phosphate esters. It is not known whether such proteins can hydrolyze multiple classes of substrates (tests for ssDNA exonuclease activity for PPX1 protein have not been reported); however, we have detected no polyphosphatase activity for purified RecJ protein (data not shown). The existence of similar motifs within enzymes with quite different biological substrates points out a potential danger of assigning biochemical activity to unknown proteins purely on the basis of sequence similarity.