DNA Polymerase (cid:1) , a Novel DNA Repair Enzyme in Human Cells*

DNA polymerase lambda (pol (cid:1) ) is a novel family X DNA polymerase that has been suggested to play a role in meiotic recombination and DNA repair. The recent demonstration of an intrinsic 5 (cid:1) -deoxyribose-5-phos-phate lyase activity in pol (cid:1) supports a function of this enzyme in base excision repair. However, the biochemical properties of the polymerization activity of this enzyme are still largely unknown. We have cloned and purified human pol (cid:1) to homogeneity in a soluble and active form, and we present here a biochemical descrip-tion of its polymerization features. In support of a role in DNA repair, pol (cid:1) inserts nucleotides in a DNA tem-plate-dependent manner and is processive in small gaps containing a 5 (cid:1) -phosphate group. These properties, to-gether with its nucleotide insertion fidelity parameters and lack of proofreading activity, indicate that pol (cid:1) is a novel (cid:2) -like DNA polymerase. However, the high affinity of pol (cid:1)

All known DNA polymerases are believed to share a remarkably high functional and structural similarity (1,2). However, each of these enzymes possesses unique features that are crucial to cope with the different DNA transactions encountered while conducting DNA synthesis in vivo. Thus, in addition to the extensive DNA polymerization performed by replicative DNA polymerases, a large and growing number of enzymes have been found to be specialized in largely different and sometimes surprising types of synthesis (3)(4)(5)(6).
Among these enzymes is mammalian pol 1 ␤, a member of the family X DNA polymerases. Polymerase ␤ has been studied extensively and is now considered the paradigm of a repair DNA polymerase: it is ubiquitously expressed (7), and its biochemical features suggest that its in vivo role likely involves filling short gaps of DNA (8). Since its discovery, pol ␤ has been suggested to play a role in DNA synthesis associated with DNA repair processes in the nucleus of mammalian cells. Indeed, it is accepted today that its polymerization features, combined with its dRP lyase activity, make pol ␤ a crucial protein for base excision repair (9 -11).
In addition to pol ␤, all other enzymes belonging to family X (12) are small, monomeric enzymes that can be found in all realms of life, including archaea, eubacteria, eukaryotes, and viruses (13). The first family X enzyme to be identified in mammalian cells was terminal deoxynucleotidyl transferase, a template-independent enzyme that adds nucleotides at the junctions of rearranged Ig genes (14). In addition to terminal deoxynucleotidyl transferase and mammalian pol ␤, a large number of family X enzymes have been now characterized in different organisms, including yeast pol IV (15,16); pol /Trf 4 (17), recently renamed as pol (18); pol (19); and pol (20).
Polymerase is a nuclear enzyme that has been identified in mammals as well as different high eukaryotes including other vertebrates and plants such as Arabidopsis thaliana (20 -22). In addition to its 32% amino acid identity with pol ␤, sequence comparison and three-dimensional structure modeling predict that pol contains all four pol ␤ structural subdomains, named fingers, palm, thumb, and 8 kDa (20). However, unlike pol ␤, pol has a BRCT domain at its N terminus, which likely takes part in protein-protein or protein-DNA interactions (20 -22). Northern blot analysis reflects that murine and human pol mRNA is highly abundant in testis (20,22). Moreover, predominant expression of murine pol in pachytene spermatocytes has led to the hypothesis that it plays a role in DNA repair synthesis coupled to meiotic recombination (20). Besides having an intrinsic DNA polymerase activity, and the recent demonstration of an intrinsic dRP lyase activity (23), little is known about the biochemical features of pol . This is mainly because previous studies were limited by the lack of purified protein.
Here, we describe the purification of human pol and its basic in vitro polymerization features and discuss novel insights into its cellular function.

Cloning of the Human pol cDNA
Cloning of the human POLL gene was initiated by the identification, in the public data base dbEST/GenBank, of a collection of ESTs that showed a high similarity with the mouse pol cDNA sequence, previously obtained in our laboratory (20). The several ESTs identified, with accession numbers AA742404, AA922738, AI091150, AA989195, W69567, AI123218, AI199486, AA576526, W69888, and T81701, corresponded to the 3Ј-untranslated region of murine pol cDNA, with the exception of EST H11886, which contained part of the coding region. Human pol cDNA was obtained from human placenta by PCR amplification. A first fragment (1419 bp long), spanning position 1107-2525 of the complete cDNA sequence (2678 nucleotides), was obtained by specific PCR using primers derived from the ESTs described above. The 3Ј-terminal segment, from 2526 to 2678, was deduced from the consensus of the ESTs. A second fragment (996 bp long) was obtained by semispecific PCR, using a sense primer derived from the murine sequence close to the initiation codon and an antisense primer derived from the first PCR fragment (1419 bp long), which contained the coding sequence corresponding to positions 380 -1375 of human pol cDNA. The cDNA sequence was completed by 5Ј-rapid amplification of cDNA ends, a 504-bp long fragment that contained the untranslated 5Ј-region, and the initial portion of the coding region.

Overproduction and Purification of Human pol Protein
The complete coding sequence corresponding to pol was cloned in the BamHI-EcoRI sites of the bacterial expression vector pRSET-B (Invitrogen), which allows the expression of recombinant proteins as fusions with a multifunctional leader peptide containing a hexahistidyl sequence for purification on Ni 2ϩ -affinity resins (24). Expression of pol was carried out in the Escherichia coli strain BL21-CodonPlus(DE3)-RIL (Stratagene), with extra copies of the argU, ileY, and leuW tRNA genes. Expression of pol protein was induced by the addition of 1 mM IPTG to 3 liters of log phase E. coli cells grown at 28°C in LB to an A 600 of 0.5. After 20 min of induction, rifampicin was added to a final concentration of 120 g/ml, and cells were incubated at 28°C for 2 h. Subsequently, the cultured cells were harvested, and the pelleted cells were weighted (7.5 g) and frozen (Ϫ20°C). Just before purification, which was carried out at 4°C, frozen cells were thawed and ground with 5.5-fold their weight of alumina (Sigma) for 20 min, in the presence (per g of cells) of 1 volume of buffer A (50 mM Tris, pH 7.5, 10% glycerol, 0.5 mM EDTA, 1 mM DTT) supplemented with 1 M NaCl. The cell lysates were collected with 4 volumes of buffer A and 1 M NaCl. Cell debris and alumina were discarded after a 5-min centrifugation at 1,500 ϫ g. Insoluble material was pelleted by a 20-min centrifugation at 12,000 ϫ g. DNA was precipitated with 0.3% polyethyleneimine (10% stock solution in water, pH 7.5) and sedimented by centrifugation for 10 min at 12,000 ϫ g. The supernatant was diluted to a final concentration of 250 mM NaCl with buffer A and precipitated with ammonium sulfate to 65% saturation to obtain a polyethyleneimine-free protein pellet. This pellet was resuspended in 100 ml of buffer A and 50 mM NaCl and loaded into a 3-ml PC column equilibrated previously in this buffer. After exhaustive washing with buffer A and 100 mM NaCl, pol was eluted with buffer A and 200 mM NaCl. The eluate containing pol was diluted with an equal volume of buffer B (50 mM Tris, pH 7.5, 10% glycerol) and loaded into a second PC column (1 ml), equilibrated previously in buffer A and 100 mM NaCl. The column was washed with buffer B supplemented with 100 mM NaCl, and the protein eluted with native binding buffer (20 mM phosphate buffer, pH 7.8, 500 mM NaCl; Invitrogen). The eluate was loaded into a Ni-NTA-agarose (Invitrogen) column equilibrated with native binding buffer. After several washing steps with native binding buffer containing increasing concentration of imidazole, pol was eluted at 400 mM imidazole. Polymerase -containing fractions were collected, 5-fold diluted with buffer A, bound to a third PC column (1 ml), and eluted with buffer A containing 500 mM NaCl. This fraction contains highly purified (Ͼ99%) human pol . Protein concentration was estimated by densitometry of Coomassie Blue-stained 10% SDS-polyacrylamide gels, using standards of known concentration. Under these conditions, the yield was 25 g of purified pol /g of E. coli cells. The final fraction, adjusted to 50% (v/v) glycerol and supplemented with 0.1 mg/ml BSA, was stored at Ϫ70°C. When indicated, the pure fraction was subjected to sedimentation in a 15-30% glycerol gradient. 100 g of protein was loaded onto a 5-ml glycerol gradient containing 20 mM Tris-HCl, pH 7.5, 0.2 M NaCl, 1 mM DTT, and 1 mM EDTA and centrifuged at 62,000 rpm (Beckman SW.50 rotor) for 26 h at 4°C. Individual 200-l fractions were collected from the top of the tube, examined in Coomassie Blue-stained 0.1% SDS-polyacrylamide gels, and tested for DNA polymerase activity on activated DNA.

DNA Polymerization on Activated DNA and Poly(dA)⅐Oligo(dT)
The standard assay (25 l) contained 50 mM Tris-HCl, pH 7.5, 10 mM MgCl 2 , 1 mM DTT, 4% glycerol, 0.1 mg/ml BSA, 13.3 nM [␣-32 P] dTTP or dATP, and 625 ng of activated calf thymus DNA or 100 ng of poly(dA)⅐oligo(dT). When indicated, reaction components were added or omitted. Reactions were initiated by adding 3 nM pol or pol ␤ and incubated for 10 min at 37°C. After incubation, the reaction was stopped by adding 10 mM EDTA, and the samples were filtered through Sephadex G-50 spin columns. Polymerization activity was proportional to the amount of radioactivity present in the excluded volume, determined by counting Cerenkov radiation.

Kinetic Analysis of DNA Polymerization
Steady-state Conditions-DNA polymerization assays were performed as described above, using activated DNA and different dNTP concentrations. Measured enzyme velocity (fmol/min) was plotted as a function of dNTP concentration. The plotted data were fitted by a nonlinear regression curve to the Michaelis-Menten equation using KaleidaGraph software (Synergy Software, www.synergy.com). V max and K m(app) values were obtained from the fitted curves.
Single Turnover Analysis-Oligonucleotide P1 hybridized to oligonucleotide T6T was used as DNA substrate. Reactions (20 l) were performed as described above using 1.5 nM substrate and 20 nM pol or pol ␤. For each dNTP concentration, the amount of product formed with time was fit to a single exponential, where A is the amplitude of the exponential and k obs the exponential rate constant. The obtained single exponential rate constants were plotted as a function of substrate concentration and fit to a hyperbola, to derive K d and k pol values, the equilibrium dissociation constant for dNTP and intrinsic rate of insertion, respectively.

Short Gap Fidelity Assay
Gap filling reactions mixtures (20 l) contained 50 mM Tris-HCl, pH 7.5; 10 mM MgCl 2 ; 1 mM DTT; 0.1 mg/ml BSA; 4% glycerol; 0.5 mM each dATP, dTTP, dGTP, and dCTP; 1.6 nM gapped M13mp2 DNA; 50 nM pol ; and 400 units of T4 DNA ligase. After incubation at 37°C for 60 min, EDTA was added to 15 mM, and reaction products were resolved by electrophoresis in a neutral 0.8% agarose gel. Covalently closed, circular DNA products were electroeluted from gel slices, recovered by ethanol precipitation, introduced into E. coli MC1061 by electroporation, and plated. After scoring revertant and total plaques, the DNA of revertants was sequenced to define the sequence change responsible for the change of phenotype.

Structural
Features of Human pol -The human pol cDNA was cloned on the basis of its sequence homology with murine pol (POLL gene (20); see "Experimental Procedures"). The resulting cDNA is 2678 bp in length, with a 371-bp 5Ј-untranslated region, a 1728-bp coding sequence, and a 579-bp 3Јuntranslated region, and it exhibits 84% nucleotide sequence identity with its murine ortholog. The human POLL gene encodes a protein of 575 amino acids with an overall 83% amino acid identity to its murine ortholog. Fig. 1 shows a multiple alignment of human pol with its mouse ortholog and human pol ␤. This alignment predicts several structural domains in pol : an N-terminal BRCT domain (light gray area) and a connecting serine/proline-rich domain, both absent in pol ␤, and a pol ␤-like core containing a dRP lyase (8 kDa; gray area) and a polymerization (31 kDa; dark gray area) domain. Interestingly, the similarity between human and murine pol is maximal (92% amino acid identity) in the pol ␤ core (residues 239 -573 of murine pol and residues 241-575 of human pol ), very high (84% amino acid identity) in the BRCT domain (residues 35-125 of murine pol and residues 36 -126 of human pol ), and lower (55% amino acid identity) in the serine/proline-rich domain (residues 126 -238 of murine pol and residues 127-240 of human pol ). Despite an overall lesser similarity, the conservation in this domain appears to be more restricted to serine, threonine, and proline residues, thus emphasizing the putative role of this region as a target for phosphorylation (20). The amino acid similarity among human pol and human pol ␤ (about 33% amino acid identity) was high enough to build a computer-generated structural model of the pol ␤ core of human pol , whose architecture and subdomain composition (fingers, palm, thumb, and 8 kDa) are fully compatible with the three-dimensional structure of pol ␤ (not shown). A closer inspection of the predicted C-terminal pol ␤ core indicates that most residues characteristic of pol X family enzymes, including critical residues involved in dNTP binding, metal binding, DNA binding, and catalysis in pol ␤, are also invariant or highly conserved in human pol .
Analysis of pol mRNA Expression-In agreement with the meiotic function proposed for pol (20), Northern blot analysis of human pol mRNA reflected high expression levels in testis (not shown and Ref. 22). However, a more sensitive reverse transcription-PCR technique detected pol mRNA in every human and murine tissue examined (not shown). We therefore performed a quantitative PCR analysis of pol expression in human tissues comparing its expression levels with those of pol ␤, an enzyme thought to have a general role in DNA repair partly because of its housekeeping expression. Relative expression levels of both pol and pol ␤ mRNA varied in the different tissues examined, particularly in testis and brain, where pol mRNA levels were clearly higher relative to those of pol ␤ (2and 4-fold, respectively). An opposite situation occurs in liver, where the lowest pol :pol ␤ ratio (1:10) was observed. Because pol expression is not only restricted to germinal cells, an additional role(s) of this enzyme in somatic cells must be considered. The observation that the relative expression of pol and pol ␤ varies in some tissues suggests that the cellular functions of these two enzymes may be nonredundant.
pol Has an Intrinsic DNA Polymerase Activity-Human pol was expressed in E. coli cells as a fusion protein containing a histidine tag at its N terminus (see "Experimental Procedures"). As shown in Fig. 2, upon IPTG induction, a new polypeptide was observed migrating at the expected position for pol (ϳ68 kDa) and making up to 1% of the total protein extract. About 25% of the overproduced pol remained soluble and was precipitated with 65% ammonium sulfate. Polymerase was further purified using PC and Ni-NTA chromatography. was precipitated with ammonium sulfate at 65% saturation (AS), and further purified by PC and Ni-NTA (Ni) chromatography, as described under "Experimental Procedures." The electrophoretic migration of a collection of molecular mass markers (MW) is shown on the left. B, cosedimentation of a DNA polymerase activity with the human pol 68-kDa polypeptide. Purified human pol was subjected to a glycerol gradient sedimentation (see "Experimental Procedures"). The inset shows the analysis of fractions 4 -26 in a Coomassie Blue-stained SDSpolyacrylamide gel. DNA polymerase activity of each fraction was assayed on activated DNA and is expressed as dAMP incorporation (in pmol). Quantitation of the human pol protein band was carried out by densitometry of the stained gel and is expressed as optical density (a.u., arbitrary units).
The protein eluted from the Ni-NTA column as a homogeneous species as judged by SDS-PAGE and was devoid of nuclease contaminants, as tested in nuclease assays.
To analyze DNA polymerization activity in the purified pol fraction, we carried out different in vitro assays using either activated DNA or defined homopolymeric molecules as a substrate. As summarized in Table I, the purified pol fraction was able to catalyze dNMP incorporation in the presence of either Mg 2ϩ or Mn 2ϩ divalent metal ions.
As a control of specificity, we carried out a parallel purification from control E. coli cells (transformed with the pRSET plasmid). In this case, no polymerization activity was detectable in the final fractions (not shown). To ascertain further that the DNA polymerase activity present in the final purification fraction was intrinsic to pol , we demonstrated the cosedimentation in a glycerol gradient of the DNA polymerase activity with the 68 kDa pol protein peak, identified by Coomassie Blue staining after SDS-PAGE analysis of each gradient fraction (Fig. 2B).
In our standard assay, pol specific activity (30 fmol⅐min Ϫ1 ⅐pmol of enzyme Ϫ1 ) was 10 times higher compared with that of pol ␤ (3 fmol⅐min Ϫ1 ⅐pmol of enzyme Ϫ1 ) (Table I). Interestingly, replacing magnesium ions with manganese when using poly(dA)⅐oligo(dT) as a substrate stimulated pol and pol ␤ polymerization up to 20-fold and 30-fold, respectively. This substrate-specific effect could be explained if manganese ions facilitate the slippage capacity of both enzymes. As happens with pol ␤ and other pol ␤-like enzymes (15), ddNTPs inhibited polymerization by pol . The inhibition observed for pol is in agreement with the weak discrimination for the 3Ј-OH group of the incoming nucleotide displayed by family X DNA polymerases.
pol Lacks 3Ј 3 5Ј Exonuclease-The 3Ј 3 5Ј exonuclease active site of all proofreading eukaryotic and prokaryotic DNA polymerases is made up of three conserved amino acid motifs, named Exo I, Exo II, and Exo III (25). Similar motifs could not be identified in pol sequence, suggesting that, as other family X enzymes, pol has no proofreading activity. This prediction was confirmed by exonuclease assays (see "Experimental Procedures"), where purified pol failed to display any nucleolytic activity both on single stranded and template-primer sub-strates: a 30-fold excess of enzyme over substrate did not produce a detectable degradation (less than 0.1%) after 30 min at 37°C (not shown).
pol Is Template-dependent and Distributive on a Template-Primer-Although pol ␤ is the closest homolog of pol , the latter enzyme also shares significant sequence similarity with both terminal deoxynucleotidyl transferase and pol , two enzymes reported to display a template-independent DNA polymerization capacity. To assay whether pol is template-dependent or not, DNA molecules of defined sequence were used as substrates to assay DNA polymerization. Whereas pol was able to perform synthesis on a template-primer, it was unable to use single stranded DNA or blunt double stranded DNA (not shown), thus suggesting that pol is strictly a template-dependent enzyme.
Processivity is a common feature of DNA polymerases involved in extensive DNA synthesis (i.e. replicative polymerases), as a direct consequence of tight DNA binding and efficient nucleotide insertion. Conversely, DNA repair enzymes frequently display weaker DNA interactions and incorporate nucleotides more slowly and consequently synthesize DNA in a distributive mode. We used a template-primer and different enzyme:DNA ratios to study the processivity of human pol . On this template, pol was absolutely distributive both in the presence of Mg 2ϩ and Mn 2ϩ activating ions, because the length of the synthesized products was strongly dependent on the enzyme:DNA ratio (not shown). Although it can not be discarded that accessory proteins could modulate processivity, these results suggest that pol is not well suited to carry out long patch DNA synthesis in vivo.
Processive Gap Filling by Human pol -Although being a distributive polymerase on a template-primer substrate, pol ␤ is known to conduct processive DNA synthesis in small (Ͻ6) DNA gaps, believed to be its physiological substrate. Structural evidence (26) suggests that this is the consequence of the additional contacts that are established in a gapped substrate between the N-terminal 8-kDa domain of the enzyme and the DNA downstream to the gap. Among them, the capacity of the 8-kDa domain to bind a terminal 5Ј-phosphate group is particularly important for both processivity and binding of pol ␤ during gap filling synthesis (8). We therefore compared the processivity of pols and ␤ using a template-primer and a 5-nucleotide gap with or without a phosphate group at its 5Ј-side (Fig. 3A). As can be seen in Fig. 3B, the processivity of both enzymes was increased in the presence of a phosphate group at the 5Ј-side of the gap (right panels). For pol , this increase was strictly dependent on the presence of a phosphate group, whereas pol ␤ showed some increase in processivity (visible at longer times) even in its absence (middle panels). However, this phosphate-independent increase in processivity was only seen when the gap had been partially filled (by insertion of 3 nucleotides), suggesting the establishment of specific contacts between pol ␤ and the DNA requiring a defined gap size (1-2 nucleotides). Whereas pol ␤ displayed a significant strand displacement capacity and efficiently inserted one additional nucleotide after filling the gap, pol limited its synthesis to the 5 nucleotides of the gap. This imprecise gap filling by pol ␤, which has been already described (8), can be overcome in vitro by association with XRCC1 protein (27). Because pol appears to be intrinsically able to restrict its synthesis to the length of the gap, it would be a suitable candidate to participate in a XRCC1-independent gap filling synthesis (28).
Strand Displacement Coupled with Gap Filling Synthesis Is Enhanced by Mn 2ϩ Ions-The use of Mn 2ϩ ions as metal activators is known to affect both catalytic efficiency and fidelity of several DNA polymerases in vitro (29), including family X DNA polymerases such as pol (19) or pol ␤ (30). Surprisingly, as shown in Fig. 3C, manganese greatly stimulated strand displacement synthesis on gapped DNA, catalyzed both by pol ␤ and pol . For both enzymes, the extension products obtained on the gapped substrate were comparable with those obtained in the control (no gap) template-primer molecule, and only a slight increase in the proportion of ϩ5 and ϩ6 products revealed that the reaction was occurring in a gapped molecule. Interestingly, this stimulation of strand displacement-coupled DNA polymerization was observed in both the presence or absence of a phosphate group at the 5Ј-side of the gap, although the presence of this phosphate group slightly constrained the strand displacement capacity of both enzymes. pol Has a High Affinity for dNTPs-An early observation was that the rate of nucleotide incorporation by pol on activated DNA was not augmented by using dNTP concentrations over 1 M. Contrarily, a reduction in nucleotide incorporation was observed when higher amounts (100 M) of dNTPs were used (data not shown). These data suggested that pol was saturated at a low nucleotide concentration. Therefore, a steady-state kinetic analysis of dNMP incorporation on acti-vated DNA was performed in parallel for both pol ␤ and pol . The data obtained were fit to the Michaelis-Menten equation (Equation 1) and used to determine apparent K m values. Interestingly, as shown in Table II, pol displayed an K m(app) 15-fold lower than that of pol ␤ (0.50 Ϯ 0.07 and 7.60 Ϯ 0.70 M, respectively). This difference was investigated further using defined DNA molecules and single turnover conditions, where the enzyme concentration is higher than the concentration of DNA. Under these conditions, DNA binding differences between the two enzymes are minimized, and the kinetic constants reflect the equilibrium binding constant (K d ) and the intrinsic rate of insertion (k pol ). For each dNTP concentration, the amount of product formed with time was fit to a single exponential (Equation 2) to derive the observed polymerization rate (k obs ). The exponential rate constants (k obs ) were then plotted as a function of substrate concentration and fit to a hyperbola (Equation 3) to obtain K d and k pol . Single turnover analysis of dAMP incorporation clearly revealed that pol has a higher affinity for dAMP than pol ␤. As shown in Table II, a 37-fold difference in the K d was observed (0.145 Ϯ 0.010 M for pol and 5.384 Ϯ 0.500 M for pol ␤).
To analyze 3Ј-OH discrimination by pol further, we used a defined template-primer DNA molecule to compare dAMP and ddAMP incorporation. Polymerase was able to insert ddAMP efficiently. The same behavior was observed for pol ␤, in agree-   (15,31). Single turnover parameters (Table II) reflected that dAMP and ddAMP are inserted by both enzymes with a similar efficiency, suggesting that these proteins do not show a strong selection for the 3Ј-OH group of the nucleotide. Fidelity of pol -We have shown previously that pol is a DNA-dependent DNA polymerase with no proofreading activity. As a first analysis of the capacity of pol to catalyze faithful DNA synthesis, each of the four dNTPs was assayed individually as a substrate to be incorporated opposite the four possible templating bases, in the presence of Mg 2ϩ or Mn 2ϩ ions. Fig. 4 shows that in all cases, pol preferentially inserted the correct dNTP. Thus, pol performs DNA synthesis following the Watson-Crick base pairing rules. For a more quantitative analysis, we then assayed the fidelity of pol synthesis while filling a 5-nucleotide gap in a M13mp2 DNA. This DNA substrate produces a colorless M13 plaque phenotype because of a TGA stop codon in the lacZ ␣ complementation gene sequence within the gap. As described previously (32) base substitution errors that revert the nonsense codon are scored as blue plaque revertants among total copied and ligated products. Synthesis by pol to fill in the gap generated products with a lacZ reversion frequency of 9 ϫ 10 Ϫ4 , which is similar to the reversion frequency observed with pol ␤. Thus, although pol base substitution fidelity is lower than that of replicative polymerases, it is relatively high when compared with that of the polymerases in the recently identified pol Y family (33). DNA sequence analysis of pol -generated revertants revealed significant differences in the base substitution specificity of the two enzymes. Polymerase predominantly generates transition errors. Note that of the 55 pol -produced base substitutions (Table III) only 3 were transversion errors. In contrast pol ␤-generated base substitutions are more divided evenly between transitions and transversions. DISCUSSION Polymerase is a recently discovered enzyme belonging to the family X of DNA polymerases (20 -22). Amino acid sequence comparison among all members of this family reveals a common domain organization, which could imply a similar catalytic mechanism (34). However, despite these general similarities, family X includes heterodox polymerases such as terminal deoxynucleotidyl transferase, able to conduct templateindependent synthesis, and pol , also endowed with some degree of template independence and extremely unfaithful (19). Although pol is the closest relative of pol ␤ (33% amino acid identity), no evidence had been published to date regarding pol polymerization properties. Unlike its murine ortholog (20), human pol could be expressed in E. coli in a soluble and active form and purified to homogeneity. Biochemical analysis of human pol showed basic features very similar to pol ␤: 1) strict template dependence when using Mg 2ϩ as metal activator; 2) lack of an intrinsic 3Ј 3 5Ј exonuclease; 3) distributive on a template-primer but processive in short gaps having a phosphate group at its 5Јside; 4) low discrimination against ddNTPs; 5) preferential insertion of complementary dNMPs and similar base substitution fidelity. Moreover, as has been shown recently, pol has an intrinsic dRP lyase activity (23), an activity that is crucial for the base excision repair pathway. All of these similarities suggest that pol is a ␤-like enzyme that could play a role in DNA repair in vivo.
The use of Mn 2ϩ ions to activate pol ␤ results in an increased reactivity at the catalytic site, lowering the fidelity of synthesis and even allowing template-independent reactions (30). Here we describe an additional effect of Mn 2ϩ ions, increasing the FIG. 5. Structural basis for nucleotide binding by human pol . The nucleotide binding pocket of human pol ␤ (A) is compared with the putative (modeled) nucleotide binding pocket of human pol (B). The incoming ddCTP (red) is shown, hydrogen bonded to its templating base (light yellow). Relevant residues are shown (ball and stick), and their position relative to the N terminus of the protein is indicated. One of the most striking differences between the pol ␤ and pol dNTP binding site is the nonconservation of pol ␤ residue Asp 276 (green), which plays a direct role in dNTP binding and selectivity (for details, see "Discussion"). A model structure of the whole ␤-core of human pol , with the exception of ␣-helix A, was generated with the program Swiss Model (www.expasy.ch/swissmod/swiss-model.html). The figure was made with Swiss PDB Viewer ((42) www.expasy.ch/spdbv/) and rendered with POV Ray (www.povray.org). Despite all their similarities, pol 's lack of an apurinic/ apyrimidinic lyase activity (23) or its different base substitution specificity compared with pol ␤ reveal that pol and pol ␤ are biochemically distinguishable. Moreover, an important difference between these two enzymes is related to the efficiency of dNTP binding and selection. Steady-state kinetic analysis reflected that the K m(app) (dNTP) of pol is an order of magnitude lower compared with that of pol ␤, and consistent with single turnover experiments pol has a 37-fold higher affinity for incoming nucleotides than pol ␤.
Most residues involved in dRP lyase catalysis and 5Јphosphate binding are conserved between pol ␤ and pol (23). Moreover, pol conserves most pol ␤ key residues related to polymerization catalysis including the three invariant aspartates that coordinate the divalent metal ions (Asp 427 , Asp 429 , and Asp 490 ), pivotal residues that ensure fidelity of catalysis by pol ␤ such as Arg 283 and Phe 272 (Arg 517 and Phe 506 in pol , respectively), and other invariant residues in family X polymerases (34). However, in agreement with their biochemical differences, a strong dissimilarity can be observed at the dNTP binding pocket of both enzymes. As illustrated in Fig. 5, pol conserves most pol ␤ residues that surround the incoming dNTP in the closed conformation of the ternary complex (26) but differs in some 8-kDa residues presumed to increase dNTP binding in a single nucleotide gap (i.e. pol ␤ residues Arg 40 and Lys 27 (36)). More interestingly, pol ␤ Asp 276 , known to make Van der Waals interactions with the base of the incoming nucleotide (11), is replaced by Ala 510 in pol . Besides forming the only electrostatic interaction between ␣-helix N and the 8-kDa domain of pol ␤ (through hydrogen bonding with Arg 40 (36)), Asp 276 is believed to restrict dNTP binding. Indeed, replacing Asp 276 with an uncharged residue (i.e. valine or glycine) results in an apparent increase in the nucleotide binding affinity (37). Moreover, a D276V mutant of pol ␤ has a 4-9-fold increased nucleotide binding affinity compared with that of the wild-type enzyme (36). Accordingly, pol , having an uncharged residue at this position (Ala 510 ), has a 37-fold higher nucleotide binding affinity than pol ␤, thus behaving similarly to the D276V mutant of pol ␤.
As shown here, it can be predicted that at a low dNTP concentration, pol would be a much more active enzyme than pol ␤. This biochemical difference suggests that both enzymes, instead of having a redundant function, could play a specific role related to the cellular concentration of nucleotide precursors for DNA synthesis. Thus, DNA repair could take advantage of a dual functional solution similar to that described for glucose phosphorylation, which involves both a low K m hexokinase and a high K m glucokinase (hexokinase IV (38)). Interestingly, pol mRNA expression is apparently cell cycle-dependent, being higher in quiescent and S to M phase cycling cells (22). Thus, pol could be specialized in DNA repair taking place during specific stages of the cell cycle, required in nonproliferative cell types or during differentiation processes. Such a specialized DNA repair function in pol could be perhaps related to its N-terminal BRCT domain. This domain, absent in pol ␤ but present in a large number of nuclear proteins, has been suggested to take part in different protein-protein and protein-DNA interactions (39 -41). Therefore, an important difference with pol ␤ could be the potential to establish (via its BRCT domain) distinct interactions that could regulate pol cellular function. Further work should be carried out to ascertain the specific pathway(s) recruiting this enzyme and its contribution to the maintenance of genetic stability.