Genetic evidence for reconfiguration of DNA polymerase θ active site for error-free translesion synthesis in human cells

The action mechanisms revealed by the biochemical and structural analyses of replicative and translesion synthesis (TLS) DNA polymerases (Pols) are retained in their cellular roles. In this regard, DNA polymerase θ differs from other Pols in that whereas purified Polθ misincorporates an A opposite 1,N6-ethenodeoxyadenosine (ϵdA) using an abasic-like mode, Polθ performs predominantly error-free TLS in human cells. To test the hypothesis that Polθ adopts a different mechanism for replicating through ϵdA in human cells than in the purified Pol, here we analyze the effects of mutations in the two highly conserved tyrosine residues, Tyr-2387 and Tyr-2391, in the Polθ active site. Our findings that these residues are indispensable for TLS by the purified Pol but are not required in human cells, as well as other findings, provide strong evidence that the Polθ active site is reconfigured in human cells to stabilize ϵdA in the syn conformation for Hoogsteen base pairing with the correct nucleotide. The evidence that a DNA polymerase can configure its active site entirely differently in human cells than in the purified Pol establishes a new paradigm for DNA polymerase function.

Biochemical and structural studies with translesion synthesis (TLS) 2 DNA polymerases (Pols) have indicated a high degree of specificity in the types of DNA lesions they can replicate through (1). Thus, the ability to accommodate two template residues in its active site provides Pol the proficiency for replicating through the covalently linked cyclobutane pyrimidine dimer (CPD) (2)(3)(4)(5)(6). The adoption of a syn conformation by the purine template in the Pol active site for forming a Hoogsteen base pair with the incoming nucleotide enables it to insert nucleotides opposite DNA adducts which impair Watson-Crick (W-C) base pairing or impinge upon the DNA minor groove (7)(8)(9)(10). TLS studies in human cells have corroborated the roles and mechanisms inferred from biochemical and structural studies of Pol, Pol, and other TLS Pols (11)(12)(13)(14)(15).
In vitro studies of purified Pol, an A family Pol, have suggested that in contrast to Pol or Pol, it lacks the specificity for replicating through DNA lesions; and compared with TLS mediated by Pols with high specificity, Pol acts in a more error-prone manner (12,13). In human cells, for example, Pol functions in TLS opposite two very different types of lesions, CPDs and 1,N 6 -ethenodeoxyadenosine (⑀dA). TLS opposite CPDs occurs either by a Pol-dependent error-free pathway or by an alternative error-prone pathway in which Pol inserts a nucleotide opposite the 3Ј pyrimidine residue of a CPD from which Pol or Pol subsequently extend synthesis (13). TLS opposite the ⑀dA adduct, which is generated from interaction of DNA with aldehydes derived from lipid peroxidation (16, 17), and which impairs W-C base pairing, operates via two major pathways dependent upon Pol/Pol and Pol, respectively, in which the sequential action of Pol and Pol promotes errorfree TLS and Pol performs error-prone TLS (12). A third pathway dependent upon Rev1 polymerase activity makes a relatively minor contribution (12). Apart from these Pols, no other Pols such as Pol, Pol (12), or Pol are required for TLS opposite this adduct in human cells.
The ability of Pol to insert nucleotides opposite the ⑀dA adduct by Hoogsteen base pairing and the proficiency of Pol for extending synthesis from the nucleotide opposite ⑀dA explains the roles these Pols play in TLS through ⑀dA in human cells (9,12). Because Pol replicates DNA by utilizing classical W-C base pairing, ⑀dA would present a strong block unless the adduct adopts an extrahelical position in the Pol active site; hence, Pol replicates through ⑀dA using a mechanism similar to the one it uses for TLS through an abasic (AP) site. The observation that purified Pol replicates through both the ⑀dA and AP lesions by inserting an A is consistent with ⑀dA adopting an "AP" mode in the Pol active site (12). However, in striking contrast to the extremely error-prone TLS opposite ⑀dA by purified Pol, Pol-dependent TLS in human cells operates in a predominantly error-free manner wherein Pol incorporates over 90% T opposite ⑀dA (12). Such error-free TLS could occur in human cells only if the ⑀dA adduct adopts a syn confirmation in the Pol active site and forms a Hoogsteen base pair with the T residue.
To test the validity of the hypothesis that Pol adopts a different mechanism for TLS in human cells than in purified Pol, in this study, we analyze the effects of mutations in the two highly conserved tyrosine residues in the Pol active site on TLS opposite ⑀dA by purified Pol and on TLS in human and mouse cells. Our findings that these mutations affect TLS by purified Pol in a dramatically different way than they affect TLS in human and mouse cells strongly support the premise that the Pol active site is configured differently for TLS in human cells than in purified Pol.

Conserved tyrosine residues in Pol fingers domain
The O-helix in the fingers domain is conserved among A-family Pols. Within the O-helix, the Tyr-2391 residue in human Pol is conserved among all of the eukaryotic, prokaryotic, and phage A-family DNA polymerases, whereas Tyr-2387 in human Pol is conserved in both of the eukaryotic A-family Pols Pol and Pol, but it is not conserved in Escherichia coli PolI, Taq polymerase, or T5 Pol (Fig. 1). The ternary crystal structures of human Pol have revealed that the Tyr-2387 residue contacts the ␤-phosphate of the incoming nucleotide, and Tyr-2391 lies beneath the template residue (18).

Indispensability of Tyr-2387 and Tyr-2391 for TLS through ⑀dA by purified Pol
To better understand the ability and mechanism of purified Pol for TLS through ⑀dA, we carried out in vitro DNA synthesis assays on DNA substrates that harbor a single ⑀dA lesion with the (residues 1708 -2590) WT Pol protein and the Y2387A and Y2391A mutant Pol proteins. For comparison, we also examined synthesis on DNA containing an AP site, in the form of a tetrahydrofuran (THF) moiety. The Pol(1708 -2590) protein affects TLS opposite the ⑀dA and AP site similarly as the full-length Pol (kindly provided by Richard Pomerantz).
We first performed assays with DNA substrates containing a running start primer, where DNA synthesis initiates 3 nt before the lesion ( Fig. 2A). We analyzed DNA synthesis by (residues 1708 -2590) WT Pol and the Y2387A and Y2391A mutant Pol proteins, each at three different protein concentrations (0.2, 1, and 10 nM) on undamaged DNA and on the ⑀dA and AP site-containing DNA substrates. The Y2387A mutation exhibited a strong deleterious effect on DNA polymerase activity of Pol. On undamaged DNA, DNA synthesis by 10 nM Pol Y2387A protein was about the same as that for 0.2 nM WT Pol, suggesting that it is at least ϳ50-fold less efficient for polymerase activity. Importantly, Y2387A Pol lacked the ability to incorporate a nucleotide opposite ⑀dA or opposite an AP site even at high protein concentrations ( Fig. 2A). On undamaged DNA, the Y2391A Pol protein exhibited a moderate decline in DNA polymerase activity, but not as severe as the Y2387A Pol. We estimate a reduction in catalytic efficiency of ϳ10 fold, based on the similar DNA synthesis by 1 nM WT Pol versus a 10 nM concentration of the Y2391A mutant. Even though Y2391A Pol is less efficient in DNA synthesis, it inserts a nucleotide opposite ⑀dA and the AP site. However, there is a complete lack of extension of synthesis opposite from either lesion ( Fig. 2A).
Next, we qualitatively assessed the fidelity of nucleotide incorporation opposite ⑀dA by including only a single nucleotide in the assays, rather than all four. For these assays on undamaged DNA, we used 10-fold more mutant protein than WT protein because DNA synthesis is reduced by the Y2387A and Y2391A mutations. On undamaged DNA, WT Pol incorporates T opposite A most efficiently, as do the Y2387A and Y2391A mutant proteins (Fig. 2B). In the presence of all four dNTPs, Y2387A also incorporates a C at about 20% compared with T. Y2391A Pol is also error-prone, as indicated by the number of doublets and altered DNA ladder as compared with WT Pol. Opposite ⑀dA, the WT protein can incorporate A or G, but an A is incorporated the most, and in the presence of all four dNTPs, only an A is incorporated, and Pol extends synthesis to the end of the template (Fig. 2B). At the same protein concentration, the Y2387A Pol protein is unable to incorporate a nucleotide opposite the ⑀dA or AP site. Opposite both the ⑀dA and AP lesions, nucleotide incorporation by Y2391A Pol is reduced compared with the WT protein; it primarily inserts a G, but an A is also inserted with a reduced proficiency. Also, as was seen in the running start assay ( Fig. 2A), Y2391A Pol is completely deficient in extending synthesis past ⑀dA or the AP site (Fig. 2B). Next, we examined the effects of Y2387A/Y2391A double mutation on DNA synthesis by Pol on undamaged and ⑀dAcontaining DNAs (Fig. 3). In contrast to the individual Pol Y2387A and Y2391A mutant proteins, the double Y2387A/ Y2391A mutant Pol is severely deficient in polymerase activity. When Pol Y2387A/Y2391A is assayed on the undamaged DNA substrate at a 5-fold molar excess of protein over DNA, the polymerase only incorporates 4 nt (Fig. 3, lane 8), whereas the Pol Y2387A single-mutant protein is able to synthesize up to ϳ17 nt in assays containing equimolar protein:DNA concentrations ( Fig. 2A, lane 7). Not surprisingly, on the ⑀dAand AP-containing DNA substrates, Pol Y2387A/Y2391A behaves Figure 2. DNA polymerase activity of (residues 1708 -2590) WT Pol, Y2387A Pol, or Y2391A Pol on undamaged, ⑀dA, or AP site-containing DNAs. A, increasing amounts of each protein were assayed with 10 nM template in the presence of 25 M each of dGTP, dATP, dTTP, and dCTP for 5 min using the standard DNA polymerase assay conditions given under "Experimental procedures." A diagrammatic representation of the DNA substrate is shown at the top, wherein the asterisk indicates the presence of an undamaged A, an ⑀dA, or an AP lesion. Increasing protein amounts are indicted by triangles, and the concentrations for each set were 0.2, 1.0, and 10.0 nM. The positions of the 29-mer primer and the 52-nt full extension products are shown on the right. The position of the template A, ⑀dA, or AP site, 4 nt 3Ј to the primer terminus, is indicated by the asterisk on the right. B, nucleotide incorporation by WT Pol, Y2387A Pol, or Y2391A Pol opposite an undamaged A, ⑀dA, or an AP site. Assays were performed using the standard DNA polymerase conditions and contained a 25 M concentration of either dCTP, dTTP, dGTP, or dATP, indicated by C, T, G, or A, or all four dNTPs combined, indicated by N. The protein concentration for each assay is given in parentheses, and all assays were carried out for 5 min, except those containing the Pol Y2387A mutant protein, which were carried out for 10 min. A diagrammatic representation of the DNA substrate is shown at the top, wherein the asterisk indicates the presence of an undamaged A, an ⑀dA, or an AP site. Positions of the primer and full-length product are shown on the right. The asterisk on the right indicates the position of the undamaged A, the ⑀dA, or the AP site.

Reconfiguration of DNA polymerase active site
similarly to Pol Y2387A, and no nucleotide incorporation is observed opposite either lesion. Thus, the reduced catalytic activity of the Pol Y2387A/Y2391A mutant appears to be an additive effect of each of the Tyr to Ala mutations.

Tyr-2387 and Tyr-2391 are dispensable for Pol-mediated TLS through ⑀dA in human cells
Our findings, that the Y2387A and Y2391A mutations inactivate purified Pol's ability to replicate through the ⑀dA lesion and that there is a strong concordance in the pattern of TLS and nucleotide incorporation opposite the ⑀dA and AP lesions by the purified WT Pol and the Y2387A and Y2391A mutant Pol proteins (Fig. 2), have suggested that the Tyr-2387 and Tyr-2391 residues modulate TLS through ⑀dA, adopting an "AP" mode in the active site of purified Pol. Thereby, by predominantly inserting an A opposite the adduct, purified Pol conducts extremely error-prone TLS through ⑀dA. In human cells, however, Pol-mediated TLS through ⑀dA is largely error-free, as the correct nucleotide T is inserted in over 90% of TLS products (12). Because T insertion opposite ⑀dA could occur only if the adduct adopts a syn conformation and forms a Hoogsteen base pair with T(9), the Tyr-2387 and Tyr-2391 residues may play little or no role in TLS through ⑀dA in human cells because these residues effect the "AP" mutagenic mode of TLS through ⑀dA.
To determine the contribution of Tyr-2387 and Tyr-2391 residues to TLS in human cells, we analyzed the effects of Y2387A and Y2391A mutations in Pol(1708 -2590) on TLS opposite ⑀dA carried on the leading-strand template of a duplex plasmid in which bidirectional replication ensues from a replication origin (Fig. 4). As shown in Table 1, in WT HFs expressing genomic Pol, TLS opposite ⑀dA occurs with a frequency of ϳ25%. TLS is reduced to ϳ14% in Pol-depleted cells carrying the empty vector or carrying an siRNA-sensitive full-length WT Pol. TLS is restored to WT levels in Pol-depleted cells harboring siRNA-resistant WT Pol(1708 -2590). Thus, the effect of Pol(1708 -2590) on TLS opposite ⑀dA is the same as that of genomically expressed Pol. Importantly, in Pol-depleted cells expressing Y2387A or Y2391A mutant Pol (Fig.  5A), TLS occurs at WT levels ( Table 1). In the absence of Pol, TLS opposite ⑀dA is performed primarily by the Pol/Pol-dependent error-free pathway, which requires Rev1 as a scaffolding component, and by a relatively minor pathway, which requires Rev1 polymerase activity (12). Hence, in the absence of Rev1, both the Pol/Pol and Rev1 polymerase-dependent pathways become inactive, and only the Pol-dependent pathway remains functional, whereas in the absence Rev1 and Pol, all of the TLS pathways are inactivated (12). In HFs co-depleted for Rev1 and Pol where TLS would be abolished as indicated by the near absence of TLS in Rev1 Ϫ/Ϫ MEFs depleted for Pol or in Pol Ϫ/Ϫ MEFs depleted for Rev1 (12) (see Table 2), expression of siRNA-resistant WT Pol raises TLS to ϳ11%, and importantly, expression of siRNA-resistant Y2387A or Increasing amounts of each protein were assayed with 10 nM template in the presence of 25 M each of dGTP, dATP, dTTP, and dCTP for 5 min using the standard DNA polymerase assay conditions given under "Experimental procedures." A diagrammatic representation of the DNA substrate is shown at the top, wherein the asterisk indicates the presence of an undamaged A, an ⑀dA, or an AP site. Increasing protein amounts are indicted by triangles. The concentrations for each set were 0.2, 1, and 10 nM for the WT protein and 0.2, 1, 10, and 50 nM for the Y2387A/Y2391A mutant derivative. The positions of the primer and the 52-nt full extension products are shown on the right. The position of the template A, ⑀dA, or AP site, 4 nt 3Ј to the primer terminus, is indicated by the asterisk on the right.

Reconfiguration of DNA polymerase active site
Y2391A mutant Pol also raises TLS to WT Pol levels (Table  1). Thus, in contrast to their indispensability for TLS by purified Pol, the Y2387A or Y2391A mutations have no perceptible effect on TLS in human cells.
In biochemical assays, Y2387A mutant Pol is completely defective in TLS through ⑀dA, whereas Y2391A mutant Pol can insert a nucleotide opposite ⑀dA but fails to extend synthe-sis (Fig. 2). That raised the possibility that in human cells, pursuant to nucleotide insertion opposite the lesion site by Y2391A Pol, another polymerase extends synthesis. Because Pol is a proficient extender of synthesis from the nucleotide inserted opposite the ⑀dA lesion by Pol (9), and also opposite from a large variety of other distorting DNA lesions including the AP lesion (19,20), we determined whether such a Pol role could   Table 2). Expression of WT Pol raises the TLS level to ϳ21%, and expression of the Y2387A or Y2391A mutant Pol (Fig. 5B) also restores WT levels of TLS in Pol Ϫ/Ϫ MEFs (Table 2). In Pol Ϫ/Ϫ MEFs depleted for Rev1 and expressing either no Pol or catalytically inactive D570A/E571A Pol, TLS is almost completely abolished (ϳ1%), whereas expression of Y2387A or Y2391A mutant Pol raises TLS to the same level (ϳ9%) as expression of WT Pol (Table 2). Thus, both in HFs and MEFs, Y2387A and Y2391A mutations support TLS through ⑀dA to the same extent as does WT Pol.

Tyr-2387 is required for mutagenic TLS by Pol opposite ⑀dA in human cells
In human cells, Pol replicates through ⑀dA by incorporating the correct nucleotide T in over 90% of TLS products, and it also incorporates a C in ϳ5% or an A in ϳ3% of TLS products (12). Because Pol and Rev1 polymerase activity contribute to alternative error-prone TLS pathways (12), in Pol-depleted HFs carrying siRNA-sensitive full-length WT Pol, mutagenic TLS emanating from Rev1 polymerase action occurs at a frequency of ϳ11% (Table 3). Expression of siRNA-resistant WT Pol raises mutagenic TLS to ϳ15%, the increase in mutagenic TLS resulting from Pol contribution (Table 3). Importantly, expression of Y2387A Pol reduces mutagenic TLS to ϳ6% (Table 3). Because error-prone TLS by Rev1 polymerase action would remain in these cells, this reduction in mutagenic TLS could have come about if Pol's involvement in mutagenic TLS was inhibited by the Y2387A mutation. To confirm this possibility, we examined the frequency of mutagenic TLS in HFs co-depleted for Rev1 and Pol and expressing Y2387A mutant Pol (Table 3, last row). Our results that mutagenic TLS is abolished in these HFs concur with a role of Tyr-2387 in encumbering upon Pol the capacity for mutagenic TLS opposite ⑀dA.
Next, we verified these observations in Pol Ϫ/Ϫ MEFs. As shown in Table 4, mutagenic TLS in Pol Ϫ/Ϫ MEFs, which would accrue from a Rev1 polymerase role, occurs at a frequency of ϳ10%, and this frequency rises to ϳ15% in cells expressing WT Pol; by contrast, expression of Y2387A Pol in Pol Ϫ/Ϫ MEFs reduces mutagenic TLS to ϳ8%. Our results that in Rev1-depleted Pol Ϫ/Ϫ MEFs expressing WT Pol, mutagenic TLS occurs at ϳ7% (Table 4, fourth row from bottom) and that mutagenic TLS is abolished in Rev1-depleted Pol Ϫ/Ϫ MEFs expressing Y2387A Pol (Table 4, third row from bottom) add further confirmatory evidence that Tyr-2387 confers upon Pol the capability for mutagenic TLS in MEFs similar to that in HFs (Table 3).

Tyr-2391 affects suppression of mutagenic TLS by Pol opposite ⑀dA in human cells
In contrast to the effect of Y2387A mutation on the ablation of mutagenic TLS, the frequency of mutagenic TLS is elevated to ϳ36% in Pol-depleted HFs expressing Y2391A Pol (Table  3). In Pol Ϫ/Ϫ MEFs expressing Y2391A Pol, mutagenic TLS occurs at ϳ28% (Table 4). Because mutagenic TLS conferred by both Rev1 polymerase and Y2391A Pol would operate in these cells, we analyzed the frequency of mutagenic TLS in Pol Ϫ/Ϫ

Reconfiguration of DNA polymerase active site
MEFs depleted for Rev1 and expressing Y2391A Pol, because then only the contribution of Y2391A Pol would remain. We find that mutagenic TLS occurs at ϳ20% in these MEFs (Table  4, second row from bottom). This observation that Y2391A elevates Pol-mediated mutagenic TLS implies a role of Tyr-2391 in the suppression of mutagenic TLS.

Epistatic interaction of Tyr-2391 with Tyr-2387 dampens Pol mutagenicity opposite ⑀dA in human cells
The abolition of mutagenic TLS by the Y2387A mutation and the enhancement of mutagenic TLS by the Y2391A mutation suggested that the Tyr-2387 and Tyr-2391 residues interact epistatically such that Tyr-2391 suppresses Tyr-2387 action in mutagenic TLS, and the observed frequency of ϳ6 -8% of mutagenic TLS by Pol is sustained by that interaction. To explore this possibility, we analyzed the effects of the Y2387A/ Y2391A double mutation on the frequency of TLS and its mutagenicity in Pol-depleted HFs and in Pol Ϫ/Ϫ MEFs. Surprisingly, despite the severe defect in DNA synthesis by the purified enzyme (Fig. 3)

Evidence for adoption of a different configuration by the Pol active site for TLS through ⑀dA in human cells
The observation that similar to that opposite an AP site, purified Pol predominantly inserts an A opposite ⑀dA has suggested that Pol replicates through ⑀dA using an "AP" mode wherein ⑀dA becomes extrahelical. In human cells, however, Pol replicates through ⑀dA by inserting the correct nucleotide T in over 90% of TLS products. Because ⑀dA lacks the W-C edge (Fig. 4A), a T could be inserted opposite ⑀dA only if the adduct adopts a syn conformation and forms a Hoogsteen base pair with the incoming T (Fig. 6). Hence, the Pol active site must adopt a different configuration for mediating TLS in human cells than that in purified Pol. Evidence from biochemical and genetic studies with mutations in the highly conserved Tyr-2387 and Tyr-2391 residues in the Pol active site validates this hypothesis.
In TLS assays with purified Pol, Y2387A mutant Pol lacks the capacity to insert a nucleotide opposite ⑀dA, whereas Y2391A mutant Pol primarily inserts a G and to a lesser extent an A (Fig. 2B), but it fails to extend synthesis further (Fig. 2). Similar to that seen with WT Pol, the pattern of TLS and of nucleotide incorporation by mutant Pol proteins opposite ⑀dA resembles that opposite the AP lesion (Fig. 2). The complete inhibition of TLS by the Y2387A mutation opposite ⑀dA and the AP lesion indicates that Tyr-2387 is indispensable for TLS opposite both the lesions, and the observation that Y2391A Pol predominantly inserts a G and less well an A opposite both of the lesions suggests that in the absence of functional Tyr- The number of colonies in which TLS occurred by insertion of a nucleotide other than T is shown in parenthesis. b The greater reduction in mutation frequency in rows 3 and 5 than in row 1 is because of the increase in TLS frequency that occurs in cells expressing the Y2387A or the Y2387A/Y2391A Pol (see Table I) and because of the ablation of Pol-mediated mutagenic TLS by the Y2387A mutation. c siR, siRNA-resistant.

Reconfiguration of DNA polymerase active site
2391, Tyr-2387 promotes the insertion of a G or an A but does not support extension.
In striking contrast to the indispensability of Tyr-2387 and Tyr-2391 for TLS by purified Pol, mutational inactivation of these residues has no perceptible effect on TLS opposite ⑀dA in HFs or MEFs; these mutations, however, affect the mutagenicity of TLS in HFs and MEFs. Mutational analyses of TLS products in WT HFs and in Rev1 Ϫ/Ϫ MEFs in a previous study (12) and in WT HFs and Pol Ϫ/Ϫ MEFs in this study show that whereas TLS mediated by WT Pol generates ϳ6 -8% of mutational TLS products in which a C (ϳ5%) or an A or G (ϳ1-3%) are incorporated opposite ⑀dA, the Y2387A mutation inhibits mutagenic TLS, and the Y2391A mutation increases the misincorporation of C, A, or G to ϳ20% (Table 4). These results taken together with the observation that mutagenic TLS is also inhibited by the Y2387A/Y2391A double mutation (Table 4) suggest that the observed level of mutagenic TLS by WT Pol (ϳ6 -8%) in HFs and MEFs is attained by a mechanism in which Tyr-2387 promotes the misincorporation of nucleotide opposite ⑀dA, whereas Tyr-2391 suppresses it.
The indispensability of Tyr-2387 and Tyr-2391 for TLS by purified Pol (Fig. 2) but not for TLS in HFs and MEFs (Tables 1 and 2) implies that for mediating TLS through the ⑀dA adduct, the roles of these highly conserved residues-important for DNA synthesis by the purified enzyme-are minimized in the Pol active site reconfigured for TLS through ⑀dA in human cells.

Mechanism of Pol for replicating through ⑀dA in human cells
The indispensability of Tyr-2387 for mutagenic TLS through ⑀dA by purified Pol and the requirement of this residue for Pol-dependent mutagenic TLS in HFs and MEFs might suggest that mutagenic TLS in human and mouse cells operates by the same mechanism that the purified enzyme employs for replicating through ⑀dA, wherein ⑀dA adopts an "AP" mode. However, because a C is inserted opposite ⑀dA in mutagenic TLS in WT HFs and MEFs, and because a C could be inserted only if ⑀dA adopts a syn conformation and forms a Hoogsteen base pair with C in anti conformation (Fig. 6), Tyr-2387-mediated C insertion opposite ⑀dA would occur via this mechanism. The adoption of syn conformation for C incorporation modulated by the Tyr-2387 residue would suggest that the incorporation of an A or a G opposite ⑀dA by Tyr-2387 also occurs by Hoogsteen pairing between ⑀dA in syn conformation and an A or G in anti conformation (Fig. 6). Thus, the mechanism of Hoogsteen base pairing via which Tyr-2387 and Tyr-2391 coordinate the incorporation of C, A, or G opposite ⑀dA in HFs and MEFs would differ from the mechanism of adopting an AP-like mode that purified Pol employs for misincorporating A opposite ⑀dA. And importantly, the predominant incorporation of T opposite ⑀dA (ϳ92%) could only occur by the adoption of syn conformation by ⑀dA in the Pol active site (Fig. 6).

Possible mechanism for reconfiguration of the Pol active site for TLS through ⑀dA in human cells
The lack of requirement of the Tyr-2387 and Tyr-2391 residues for the predominant error-free TLS through ⑀dA in human and mouse cells and the proposal that even the mutagenic TLS that depends upon these residues would entail the adoption of a syn conformation by ⑀dA in the Pol active site can be rationalized if the Pol active site adopts a different configuration for TLS in human cells than in purified Pol. To explain the acquisition of a different configuration in the Pol active site, we posit that Pol functions in TLS in human cells as a component of a multiprotein ensemble and that protein-protein interactions and post-translational modifications in the components of this ensemble modulate the Pol active site such that it promotes rotation of ⑀dA into a syn conformation, allowing for Hoogsteen base pairing with the incoming nucleotide.
In the Pol active site reconfigured for conducting predominantly error-free TLS through ⑀dA in human cells, the roles of Tyr-2387 and Tyr-2391 residues become much less eminent, affecting only the mutagenic TLS. In the structure of purified Pol, Tyr-2387 participates in hydrogen-bonding to the ␤-phosphate of the incoming dNTP, and Tyr-2391 forms part of the active-site floor beneath the template residue (18). This explains the requirement of these residues for efficient DNA synthesis on undamaged DNA and for TLS through ⑀dA by the purified enzyme (Figs. 2 and 3). By contrast, the lack of their requirement for predominantly error-free TLS through ⑀dA in human cells implies that in the reconfigured Pol active site, these residues no longer affect the stabilization of the template or the incoming nucleotide for incorporation of the correct dNTP.
TLS Pols, such as , , , or Rev1 have a preformed active site adapted for replicating through specific types of DNA lesions. In these Pols, the action mechanisms stay the same for TLS in human cells as those indicated from biochemical and structural studies of the purified Pol, as, for example, in the role of Pol in TLS opposite CPDs and in the role of Pol in TLS opposite ⑀dA. Replicative DNA Pols also utilize similar action mechanisms in vitro and in vivo. Thus, among DNA Pols, Pol provides the first example where the action mechanism for TLS in human cells differs from the mechanism adopted by the purified enzyme.

Pol expression in yeast
The human Pol(1708 -2590) protein harboring the catalytically active C-terminal DNA polymerase domain was expressed as a fusion with glutathione S-transferase (GST) from plasmid pPOL507 as described (21). The Y2387A and Y2391A mutations were each generated by PCR using mutagenic oligomers and the Pol(1708 -2590) cDNA in pPOL523 as template. The mutant cDNAs were fully sequenced to confirm the presence of the mutations and were cloned into the expression vector, generating plasmids pJR65, pPOL665, and pBJ2333, which express GST-tagged Pol(1708 -2590) Y2387A, Pol(1708 -2590) Y2391A, and Pol(1708 -2590) Y2387A/Y2391A, respectively.
WT and mutant Pol(1708 -2590) expression plasmids were transformed into yeast strain YRP654, and the proteins were expressed and affinity-purified using GSH Sepharose as described (22). The GST fusion tag was removed from each Pol(1708 -2590) protein by treatment with prescission protease, leaving a 7-amino acid linker attached to the N terminus of Pol. Proteins were quantified by densitometry of Coomassiestained protein samples separated by 11% SDS-PAGE using ImageQuant software (GE Biotech).

DNA polymerase assays
The standard DNA polymerase assay (5 l) contained 25 mM Tris-HCl, pH 7.5, 5 mM MgCl 2 , 1 mM DTT, 10% glycerol, 0.1 mg/ml BSA, and DNA substrate. The DNA substrates consisted of a 32 P-5Ј-labeled DNA primer annealed to a 52-mer template with the sequence 5Ј-TTCGTATA ATGCCTAC ACT-[A]GAGT ACCGGAGC ATCGTCGT GACTGGGA AAAC-3Ј, in which [A] at position 20 indicates either an undamaged A, ⑀dA, or a THF moiety (AP site analog). The ⑀dAand THFcontaining templates were synthesized by the Midland Certified Reagent Company (Midland, TX) and were PAGE-purified. For running start assays, the 29-mer oligonucleotide primer 5Ј-GTTTTCCCAG TCACGACGAT GCTCCGGT-A-3Ј was annealed to each template. To assay nucleotide incorporation opposite A, ⑀dA, or the AP site, the 23-mer primer 5Ј-GTCACGACGATGCTCCGGTACTC-3Ј was used. Single dNTPs, dATP, dGTP, dTTP, dCTP, or all four dNTPs combined were included at concentrations indicated in the figure legends. Reactions were initiated by the addition of 1 l of DNA polymerase in 5ϫ reaction buffer (125 mM Tris-HCl, pH 7.5, 5 mM DTT, 0.5 mg/ml BSA) and carried out at 37°C for times indicated in the figure legends before termination by 6 volumes of 95% formamide loading buffer containing 0.06% xylene cyanol, 0.06% bromphenol blue. Reaction products were separated by 12 or 20% TBE, 8 M urea-PAGE. Gels were fixed in 10% methanol, 10% acetic acid for 10 min and dried, and products were visualized by phosphorimaging on a Typhoon FLA7000 (GE Biotech).

Construction of ⑀dA plasmid vectors and TLS assays
The in-frame target sequence of the lacZЈ gene containing the ⑀dA lesion is shown in Fig. 4. The detailed methods for construction of lesion-containing SV40 duplex plasmid, for TLS assays, and for mutational analysis of TLS products have been described previously (15,23).

Stable expression of WT Pol and mutant Pol in HFs or MEFs
DNAs encoding human WT Pol(1708 -2590) or the mutant (residues 1708 -2590) Y2387A, Y2391A, or Y2387A/Y2391A Pol, respectively, were cloned into vector pCMV7-3xFLAGzeo (Sigma). The resulting vectors were transfected into normal human fibroblast (GM637) cells or Pol Ϫ/Ϫ MEF cells by iMFectin transfection reagent (GenDEPOT). After a 24-h incubation, 0.5 g of Zeocin (GenDEPOT) were added to the culture medium. After 3 days of incubation, cells were washed with PBS buffer and were continuously cultured with the medium containing 250 ng of Zeocin for ϳ2 weeks. Protein expression and siRNA knockdowns were checked by Western blot analysis (Fig. 5) as described before (13).