Structural insights into the promutagenic bypass of the major cisplatin-induced DNA lesion.

The cisplatin-1,2-d(GpG) (Pt-GG) intrastrand crosslink is the predominant DNA lesion generated by cisplatin. Cisplatin has been shown to predominantly induce G to T mutations and Pt-GG permits significant misincorporation of dATP by human DNA polymerase β (polβ). In agreement, polβ overexpression, which is frequently observed in cancer cells, is linked to cisplatin resistance and a mutator phenotype. However, the structural basis for the misincorporation of dATP opposite Pt-GG is unknown. Here, we report the first structures of a DNA polymerase inaccurately bypassing Pt-GG. We solved two structures of polβ misincorporating dATP opposite the 5'-dG of Pt-GG in the presence of Mg2+ or Mn2+. The Mg2+-bound structure exhibits a sub-optimal conformation for catalysis, while the Mn2+-bound structure is in a catalytically more favorable semi-closed conformation. In both structures, dATP does not form a coplanar base pairing with Pt-GG. In the polβ active site, the syn-dATP opposite Pt-GG appears to be stabilized by protein templating and pi stacking interactions, which resembles the polβ-mediated dATP incorporation opposite an abasic site. Overall, our results suggest that the templating Pt-GG in the polβ active site behaves like an abasic site, promoting the insertion of dATP in a non-instructional manner.


Introduction
Cisplatin (cis-diamminedichloroplatinum(II)) is one of the most commonly employed anticancer drugs. Over the past four decades, cisplatin has been used to treat numerous cancers, including but not limited to ovarian, cervical, and breast cancers [1][2][3]. As a platinum-based chemotherapeutic agent, cisplatin exerts its cytotoxicity by targeting DNA to form mutagenic platinum-DNA (Pt-DNA) adducts. Upon cell entry, cisplatin is aquated to form a reactive platinum species. The resultant potent electrophile reacts with vulnerable nucleophilic N7 positions of purines to form platinum covalent linkages [4]. These Pt-DNA lesions generated by cisplatin include monoadducts, interstrand cross-links, and intrastrand cross-links, though the vast majority of cytotoxic lesions formed are cisplatin-1,2-d (GpG) (Pt-GG) intrastrand cross-links ( Figure 1) [1,5]. The Pt-adducts formed by cisplatin inhibit essential biological processes such as replication and transcription, promoting genomic instability, cell cycle arrest, and apoptosis.
Despite cisplatin's popularity and clinical efficacy, tumors often exhibit intrinsic resistance and notably, a significant fraction of tumors develop acquired resistance in response to treatment [6,7], particularly ovarian cancer [8]. The development of carboplatin provided a promising alternative for cisplatin-treated tumors [9]; though it exhibits reduced toxicity, this drug displays evidence of cross-resistance with cisplatin [7,10]. Common resistance mechanisms include increased nucleotide excision repair, loss of mismatch repair, and increased efflux of cisplatin, among others [7,8,11,12]. Interestingly, these resistance mechanisms can arise through cisplatin-induced mutations themselves, enabling cellular survival amidst cisplatin treatment [13][14][15]. Across various cell types and organisms, the mutation signature of cisplatin exhibits a high proportion of G:C to T:A transversion mutations, and the majority of cisplatin-induced mutations occur at typical sites of cisplatininduced intrastrand cross-links (GpG, GpA, or ApG dinucleotides) [16][17][18]. These prominent G-to-T transversions can be explained in part by the misincorporation of dATP opposite the 5′-dG or 3′-dG of Pt-GG by error-prone DNA polymerases.
Notably, Y-family DNA polymerase η (polη) and human X-family DNA polymerase β (polβ) have been observed to contribute to cellular resistance to cisplatin treatment through both the accurate and mutagenic bypasses of Pt-GG intrastrand cross-links [19][20][21][22][23][24]. In agreement, polβ, which is a crucial component of base excision repair (BER), has been shown to exhibit significant increases in expression at both the transcriptomic and proteomic levels in multiple tumor types. In particular, polβ has been quantified up to protein levels of 286-fold in breast cancer tissues, among many other reports [25][26][27][28][29][30]. The overexpression of polβ has been thought to enhance its interference with lagging strand replication [31] and increase its activities in nucleotide excision repair [29] and error-prone translesion DNA synthesis [32]. Corroborating studies report that knockdown or inhibition of polβ sensitizes cells to cisplatin, underscoring the significant role polβ plays in modulating cellular sensitivity to cisplatin [33,34]. Importantly, polβ overexpression has also been demonstrated to increase genomic instability and result in a mutator phenotype in some cancer cells due to elevated polβ-mediated error-prone translesion synthesis that increases spontaneous mutations, emphasizing the role of polβ in broadening the landscape of intratumor genetic heterogeneity upon which cisplatin therapeutic selection can act, promoting the emergence of resistant clones composing the tumor population [32,35,36].
polβ has been previously demonstrated to cause significant misincorporation of dATP opposite the 5′-dG of gapped DNA containing an oxaliplatin-GG adduct, a similar but bulkier Pt-DNA adduct than Pt-GG [37]. Kinetic assays revealed this misincorporation is 17fold higher in frequency across the major cisplatin-DNA adduct and showed that the Pt-GG adduct increases the frequency of dATP misinsertion relative to dTTP misinsertion, which is preferred on undamaged gapped DNA [37]. This further justifies the potential role of polβ as a key factor responsible for the mutagenic insertions that contribute to the prominent G-to-T transversions associated with the mutation signature of cisplatin. Therefore, the acquired resistance to cisplatin frequently observed at the cancer cell population-level may be, in part, due to the molecular-level consequences of the bipartite interplay between the cisplatininduced formation of Pt-GG intrastrand cross-links and the mutagenic insertion opposite the 5′-G of Pt-GG by polβ. To advance understandings of this consequential mutagenic insertion by polβ across the major cisplatin-DNA adduct, particularly in polβoverexpressing cancer cells, detailed structural characterizations are warranted.
We have previously reported crystal structures for the accurate nucleotide incorporation opposite Pt-GG, furthering understandings of polβ-mediated resistance to cisplatin [21]. However, the structural basis for the promutagenic catalysis across the major cisplatin-DNA adduct by polβ is not yet reported, despite the high yield of Pt-GG formation in cells during cisplatin treatment, the cisplatin-induced G-to-T mutation signature, and the significant dATP misinsertion previously observed across the 5′-dG. Herein, we report two crystal structures of polβ incorporating a non-hydrolyzable dATP analog opposite the 5′-dG of the templating Pt-GG cross-link, which reveal insights into the mutagenic bypass of cisplatin-DNA adducts by polβ.

Preparation of the DNA duplex used for X-ray crystallographic studies
Oligonucleotides used for preparation of the modified DNA duplex were purchased from Integrated DNA Technologies (Coralville, IA). The sequence of the template DNA strand used for cisplatin modification and crystallographic studies was 5′-CCCACGGCCCATCACC-3′ (where GG designates the site of formation of the Pt-GG intrastrand cross-link by cisplatin). The previously reported method was used to prepare the platinum-modified template [38]. Briefly, a reaction containing a 2 : 1 molar ratio of silver nitrate to cisplatin in water was incubated 18 h in the dark at 37°C to produce active aquated cisplatin. The resulting silver chloride was removed and active cisplatin was isolated through centrifugation of the reaction mixture at 13 000 g for 10 min and collection of the supernatant. To modify the template DNA strand, 1.2-fold molar excess of the active cisplatin was mixed with the unmodified template DNA in 10 mM sodium phosphate (pH 6.8) and incubated in the dark for 18 h at 37°C. The reaction mixture was immediately subjected to ion exchange chromatography (Mono Q 5/50 GL, GE Healthcare), and the Pt-GG-modified template DNA was purified by elution over a gradient of 0.1 to 1.0 M NaCl in 10 mM Tris (pH 8.0). The purified Pt-GG-modified template DNA was desalted with a Sep-Pak C18 cartridge (Waters). The modified DNA was dried under reduced pressure at 30°C (Eppendorf, Vacufuge plus) and resuspended in water. The template strand containing the Pt-GG intrastrand cross-link was annealed with upstream and downstream primers in a 1 : 1 : 1 molar ratio to yield a single nucleotide gap, as described previously [39]. The upstream primer sequence was 5′-GGTGATGGGC-3′, and the downstream primer sequence was 5′phosphate/GTGGG-3′.

Human DNA polymerase β expression and purification
Human polβ (amino acids 1-335) containing an N-terminal polyhistidine tag and Factor Xa cleavage site was expressed in E. coli BL21 (DE3) cells. Cultures were allowed to grow in LB at 37°C until OD 600 reached 0.6. Cells were induced with 0.1 mM of isopropyl β-D-1thiogalactopyranoside at 28°C for 18 h. Cells were pelleted at 4°C and lysed in HisTrap column binding buffer A (50 mM sodium phosphate (pH 7.8), 500 mM NaCl, 10% glycerol) containing 0.25 mM PMSF, 0.25% Nonidet-P-40, 0.25% Triton X-100, and 1 mg/ml lysozyme. Cell lysate was subjected to sonication for 60 s and centrifuged at 15 000 g at 4°C for 20 min. Protein was purified on a HisTrap HP column (GE Healthcare) and eluted in 500 mM imidazole. Protein was further purified on a Heparin HiTrap column (GE Healthcare) in 50 mM Tris ( pH 7.4), 10% glycerol, and 5 mM β-Mercaptoethanol and eluted in a gradient of 0.1-1 M NaCl. polβ was cleaved from its N-terminal tag by the addition of Factor Xa protease (NEB). Protein was subsequently purified by size exclusion chromatography (Superdex75, GE Healthcare), flash frozen in liquid nitrogen and stored at −80°C in 15 mg/ml aliquots in 20 mM Tris (pH 7.5), 10% glycerol, 200 mM NaCl, and 1 mM βmercaptoethanol.

Crystallization of the polβ:Pt-GG•dATP ternary complexes
The hanging drop method was used for crystallography by mixing 1 μl of the polβ:Pt-GG complex solution with 1 μl of reservoir solution in the presence of 5 mM MgCl 2 or MnCl 2 and incoming non-hydrolyzable dATP* (dAMPNPP; Jena Bioscience). The crystals of the polβ:Pt-GG•dATP* complex were grown in 23% PEG 3350, 50 mM Tris (pH 6.5), and 0.35 M sodium acetate at 22°C for one week. Crystals were cryo-protected in reservoir buffer supplemented with 15% ethylene glycol and flash frozen in liquid nitrogen for data collection. Diffraction data sets for crystals of the ternary polβ:Pt-GG•dATP* complexes were collected at a wavelength of 0.97948 Å on beamline 5.0.3 at the Advanced Light Source (Berkeley, CA). Diffraction data for the polβ: Pt-GG•dATP* complexes was indexed and scaled using the program HKL2000. The data for each structure was further processed using CCP4 package programs. The structures were solved by molecular replacement using the ternary structure of the polβ:Pt-GG•dCTP* complex in the presence of Mg 2+ (PDB ID: 4TUR) or Mn 2+ (PDB ID: 4TUS) [21] as the search models. The location of the polβ:Pt-GG model with side chains added was optimized by several rounds of rigid body refinement. Manual model rebuilding was carried out with the Coot program into electron-density maps and refined using the CCP4 Refmac or PHENIX software suite. Crystallographic figures were prepared using Pymol. Statistics of data processing, data quality, and the refined models are summarized in Table 1.

Ternary structure of polβ incorporating dATP opposite 5′-dG of the Pt-GG intrastrand cross-link in the presence of Mg 2+
To understand the structural basis for the dATP misinsertion across the 5′-dG of the major cisplatin-DNA adduct by polβ [37] and to provide insights into the mutagenic bypass of this adduct, we solved a crystal structure of polβ inserting the non-hydrolyzable dATP analog dAMPNPP (dATP* hereafter) across the templating 5′-dG of the Pt-GG cross-link in the presence of Mg 2+ (Figure 2). The polβ:Pt-GG•dATP*-Mg 2+ ternary complex structure was refined to 2.3 Å resolution ( Table 1). The Pt-GG-modified DNA duplex used here contained a single-nucleotide gap opposite the 5′G of the Pt-GG lesion and was of identical sequence as the DNA duplexes used previously in order to facilitate direct comparisons with reported polβ:Pt-GG•dCTP*, polβ: GG•dCTP*, and polβ : GG structures [21] (Figure 2A). The nonhydrolyzable dAMPNPP serves as an excellent isosteric substitute for dATP for crystallization and structural analyses. While usage of dATP would result in hydrolysis and subsequent nucleotidyl transfer, substitution with dATP* prevents catalysis through the presence of a bridging NH group. Importantly, this analog enables coordination of the primer terminus 3′-OH with the catalytic metal ion and otherwise does not affect the active site conformation of polβ, which is demonstrated in previous structural analyses of the prechemistry state of the enzyme's active site [40][41][42][43][44].
The Pt-GG•dATP*-Mg 2+ complex adopts a 'semi-open' protein conformation that is similarly observed in the Pt-GG•dCTP*-Mg 2+ complex (PDB ID: 4TUR) ( Figure 2B-F) [21]. The semi-open protein conformation is characterized by a minor conformational change in the protein from the open conformation of the binary complex, specifically a 1.2-1.9 Å shift of the residues of the α-helix N towards the active site. Between the N2 of 3′-dG and N2 of 5′-dG, there is a large increase in distance to 9.5 Å ( Figure 2C), as compared with previously reported structures. Additionally, there is a substantial roll angle of 85.1° between the 3′-dG and 5′-dG of the Pt-DNA adduct ( Figure 2C), which is significantly larger than the 63.4° roll angle observed in the structure of the Pt-GG•dCTP* complex ( Figure 2F). This is mostly due to an approximate 20° opening shift in the 5′-dG between the structures.
Strikingly, there is no canonical Watson-Crick base pairing exhibited between the incoming nucleotide and 5′-dG ( Figure 3A,D). The incoming dATP appears to be in the mixture of the syn and anti conformations ( Figure 3D). The incoming nucleotide does not form a coplanar base pairing with the templating 5′-G of Pt-GG. A water-mediated hydrogen bond between N7 of dATP* and Asn279 facilitates stabilization of the syn-dATP ( Figure 3A). In addition, pi stacking interactions stabilize the binding of dATP* opposite the 5′-dG ( Figure 3B). In particular, the incoming dATP* engages in edge-to-face interactions with the 5′-dG and primer terminus dC. There are also pi stacking interactions observed between the incoming nucleotide and Tyr271. Key residues are within hydrogen bonding distance of the phosphate backbone of dATP* for stabilization, including Arg183, Ser180, and Gly189 ( Figure 3C). Within the active site, the distance between the 3′-OH of the primer terminus and the Pα of the incoming nucleotide is 3.3 Å, similar to the distance exhibited in the unmodified GG•dCTP* structure.
In the polβ:Pt-GG•dCTP*-Mg 2+ complex structure, the catalytic metal ion (metal A) is absent and only the nucleotide-binding metal ion (metal B) is bound [45,46]. Due to the absence of the catalytic metal ion, there is a lack of coordination of both Asp256 and the 3′-OH with Mg 2+ , which has been shown to be a key interaction that facilitates proton transfer from the 3′-OH to Asp256 during catalysis [47]. Instead, the 3′-OH forms a hydrogen bond with O2 of Pα (2.6 Å) of the incoming nucleotide. Moreover, two water molecules are positioned in the vacant spot typically occupied by the absent metal ion, which assist in coordination of the 3′-OH, dATP*, and nearby residues. These observations suggest that catalytic incorporation of dATP by polβ in the presence of Mg 2+ would be inefficient.
The overall structure of the Pt-GG•dATP*-Mg 2+ complex is highly similar to the overall structure of the published Pt-GG•dCTP*-Mg 2+ complex [21], with a root mean square deviation (r.m.s.d.) of 0.394 Å ( Figure 4A and Table 2). Both structures reveal a semi-open protein conformation, and the key active site residues are similarly positioned ( Figure 4B). Minor conformational differences are observed in the Lys280 and Arg283 residues, as well as the Pt-GG lesion. Lys280 is flipped towards the templating nucleotide and participates in its stabilization. Intriguingly, Arg283 adopts a different conformation for polβ and is positioned beneath the Pt-GG, rather than interacting with the O3′ of 5′-dG in Pt-GG•dCTP*-Mg 2+ . There are only minor differences in the DNA duplexes induced by the Pt-GG lesion between the correct and incorrect insertion structures, limited to the Pt-GG adduct and the dC 3′ to the lesion ( Figure 4C), suggesting that misincorporation of dATP does not perturb the conformation of the DNA further than what is already altered by the major cisplatin-DNA adduct during the correct insertion.
In contrast, the Pt-GG•dATP*-Mg 2+ complex varies considerably from the unmodified closed GG•dCTP* ternary complex (r.m.s.d. = 1.221 Å) ( Figure 5A) [21]. As expected, the Pt-GG adduct significantly alters the conformation of the DNA duplex ( Figure 5B). There are large variations in the base pairs upstream and downstream of the Pt-GG lesion. Moreover, the primer terminus is shifted further downstream by 1.6 Å. In the dATP* insertion structure, the minor groove recognition residues Asn279 and Arg283 do not participate in hydrogen-bonding interactions with the DNA ( Figure 5C). Due to the semiopen conformation of the α-helix N, Asn279 is positioned further away from the incoming nucleotide and is not within hydrogen-bonding distance, but rather participates in the watermediated hydrogen bonding interaction ( Figure 3A). Unlike the structure of the unmodified GG ternary complex, Arg283 does not interact with the minor groove edge of the templating base. The Tyr271 residue of Pt-GG•dATP*-Mg 2+ is positioned towards the imidazole ring of the incoming purine to participate in pi stacking interactions, rather than interacting with the primer terminus ( Figure 5D).
Overall, these molecular observations corroborate previous studies that highlight the sensitivity of polβ to bulky DNA adducts and its ability to adopt different conformations in response. Here, polβ recognizes the Pt-GG lesion and exhibits a semi-open conformation that appears to be sub-optimal for catalyzing the misin-sertion of dATP opposite the 5′-dG of Pt-GG.

Ternary structure of polβ incorporating dATP opposite 5′-dG of the Pt-GG intrastrand cross-link in the presence of Mn 2+
The Pt-GG«dATP*-Mg 2+ ternary complex structure most likely represents a ground state that must undergo a further metal-dependent open-to-closed protein conformational change for favorable catalysis [48]. In previously solved structures, we often observe the presence of one metal ion corresponds with an open protein conformation that is catalytically inefficient, whereas the binding of two metal ions may facilitate an open-to-closed conformational change to a catalytically competent state [21,46,49]. To gain further understandings of the polβ-mediated cisplatin-induced mutagenesis, we solved a crystal structure of polβ inserting dATP* across the templating 5′-dG of the Pt-GG adduct in the presence of the active site Mn 2+ . The Mn 2+ ion has been demonstrated to promote the formation of a closed conformation during incorrect insertion [42,46,50]. The Pt-GG•dATP*-Mn 2+ ternary complex structure was refined to 3.1 Å ( Figure 6A).
The Mn 2+ -bound ternary complex appears to be in a catalytically more favorable conformation than the Mg 2+ -bound ternary complex. The incoming dATP* is in a syn conformation and does not form a coplanar base pairing with the templating 5′G of Pt-GG ( Figure 6B,C). As in the Mg 2+ -bound structure, the incoming dATP* engages in stacking interactions with Tyr271 and the primer terminus dC. Unlike the Mg 2+ -bound structure, both the catalytic and nucleotide-binding metal ions are present in the Mn 2+ -bound structure, which is evident in the electron density map ( Figure 6B,D). The catalytic metal ion is coordinated with three catalytic carboxylates (Asp190, Asp192, Asp256), the primer terminus 3′-OH, and the phosphate oxygen. The α-helix N is in a semi-closed conformation rather than an open-conformation. The 3′-OH of primer terminus is 3.2 Å away from the Pα of dATP* and poised for in-line nucleophilic attack on the Pα ( Figure 6E).
The Pt-GG•dATP*-Mn 2+ ternary complex is significantly different from its Pt-GG•dATP*-Mg 2+ counterpart (r.m.s.d. = 1.000 Å; Figure 7A), largely due to the change in protein conformation and position of the nascent base pair. In the active site, there is a slight variation in the position of the dATP*, as well as a corresponding shift in the nucleotidebinding ion ( Figure 7B). There is a 2.1 Å upstream rise in the Pt atom, permitting a closer distance between the 3′-and 5′-dG of Pt-GG, which is also evident in the 3.8 Å rise in the phosphate and 2.8 Å rise in the O4′ of 5′-dG. Interestingly, this upstream shift of the Pt moiety and 5′-dG somewhat resembles the corresponding dG:dAMPCPP mismatch ternary structure (r.m.s.d. = 0.845 Å), where there is an upstream shift of the templating base from the coding template position, generating an abasic site intermediate ( Figure 7C) [43]. However, the presence of the cis-diammineplatinum moiety appears to prevent the full upstream shift that would be typically achieved in its absence ( Figure 7D). The primer terminus of the Pt-GG·dATP* complex does not rotate as significantly as that of the dG-dAMPCPP mismatch structure, enabling the formation of a much less staggered and more coplanar base pair with the 3′-dG. The combination of the constraints by the protein in the semi-closed conformation along with the bulkiness of the platinum moiety induce the compaction of the 3′-dG and 5′-dG of the Pt-GG intrastrand cross-link. Due to steric constraints of the intrastrand cross-link, this compaction is achieved through a tilting of the 5′-dG and an analogous adjustment in the dATP* to the syn conformation, which simultaneously supplies the aforementioned beneficial pi stacking interactions for the nucleotide's stabilization. The tilting of the 5′-dG positions the base partially out of the templating base pocket and places it closer in distance to the 3′-dG. This in contrast with the 5′-dG of the undamaged mismatch structure, which fully vacates the coding base region and forms a single hydrogen bond with N6 of the incoming dATP*.
Analyses of the crystallographic thermal B-factors between the Mg 2+ and Mn 2+ ternary structures reveal that the dATP* and 5′-dG are most labile in the Mg 2+ complex, while the cisplatin moiety and 5′-dG are most labile in the Mn 2+ complex ( Figure 7E). The higher relative stability of the incoming dATP* in Pt-GG•dATP*-Mn 2+ may be due to the presence of the catalytic metal ion, which can increase the stability of the incoming nucleotide. The position of the purine ring of the dATP* in the Mg 2+ structure appears to be more flexible due to its lack of Watson-Crick base pairing with the 5′-dG, which also appears to be disordered. The primer terminus and one active-site metal of the Mg 2+ structure are moderately well-ordered, each with an average B-factor of 39 and 35 Å 2 , respectively. In addition, the average B-factor of the Pt moiety is 34 Å 2 , indicating it is well-ordered, as compared with the Mn 2+ structure. The average B-factors for the components of the Mn 2+bound active site structure are much higher, likely due to the low resolution of the structure. In addition, the semi-closed conformation of the α-helix N in the Mn 2+ structure may contribute to the high B-factor of the Pt atom through the constraints by the protein and resultant compaction of the 5′-dG towards the 3′-dG, as demonstrated by the shorter distance between the 3′-dG and 5′-dG than that of other Pt-GG structures ( Figure 2C,F and  6B).
Overall, the observation of the catalytic metal coordination, the semi-closed conformation, and syn-dATP in the Mn 2+ -bound structure suggest that the insertion of dATP opposite the Pt-GG adduct may occur without involving a coplanar base pairing.

The major cisplatin lesion behaves like an abasic site in polβ active site
Our data presented here reveal insights into cisplatin-induced mutagenesis by human DNA polymerase β. Our structures depict -within the confines of the polβ active site -how the Pt-GG intrastrand cross-link promotes misinsertion of dATP across the 5′-dG. In contrast with canonical Watson-Crick base pairing characteristics, the structures we present here are characterized by non-instructional interactions such as protein templating and pi stacking interactions with the DNA, which are similarly observed in the ternary structure of polβmediated misinsertion of dAMPCPP opposite an abasic site analog, tetrahydrofuran (THF) [44]. Specifically, in the catalytically competent Mn 2+ structure, we observe an altered conformation of the Arg283 side chain. While Arg283 typically interacts with the N3 of the templating 5′-dG and O4′ of the 3′-dG in the closed GG•dCTP* ternary structure of polβ, we observe the side chain of Arg283 only forms interactions with O3′ of 5′-dG and not with the nucleobase. Instead, Lys280 is positioned well to weakly interact with the N3 of 5′-dG within both the Mg 2+ and Mn 2+ structures ( Figure 7B). Typically, Lys280 stabilizes templating purines through van der Waals interactions [51]. Our observations indicate Lys280 plays a previously unobserved role in potentially forming a hydrogen bond to stabilize the templating 5′-dG. These weak interactions of the protein with the 5′-dG may allow for greater flexibility of the templating base. Furthermore, several pi stacking interactions occur and stabilize the purine moiety of the incoming nucleotide ( Figure 3B and 6C). The incoming dATP* forms edge-to-face interactions with the templating 5′-dG, primer terminus dC and Tyr271. These pi stacking interactions and hydrogen bonding interactions of dATP* with Tyr271 are similarly observed in the structure of polβ-mediated misinsertion of dAMPCPP opposite an abasic site analog [44]. Together, these protein templating and pi stacking interactions drive the inefficient mutagenic insertion of dATP.
The structure of the Pt-GG•dATP*-Mn 2+ ternary complex superimposes most similarly to the structure of the abasic site insertion (r.m.s.d. = 0.763 Å) ( Figure 8A and Table 2). In both structures, the protein adopts a semi-closed conformation. The template strands both experience a shift upstream compared with the ternary structure of polβ correctly inserting dCTP across an unmodified templating dG. The bases of the Pt-GG adduct shift upstream ~2.5 Å in the presence of the dATP analog compared with the position of undamaged DNA upon binding the correct dCTP (not shown), albeit the template strand of the THF-dATP* complex is shifted upstream to a greater degree. Most notably, Tyr271 of both structures is pointed away from the primer terminus to form dual interactions with the incoming dATP analog ( Figure 8B). The hydroxyl group of Tyr271 is within weak hydrogen bonding distance to the N7 or N3 of dATP* of the Pt-GG or abasic site structure, respectively; the tyrosyl ring is poised to form pi stacking interactions with the purine moiety in both structures, assisting in the nucleotide's stabilization where base pairing is absent. While the incoming dATP* of the THF-dATP* complex is in the anti conformation, the dATP* bound across from the Pt-GG adduct benefits from a stabilizing edge-to-face interaction with the primer terminus when it adopts the syn conformation ( Figure 8A). This also in tandem allows for a weak hydrogen bond of N1 of dATP* with N2 of the 3′-dG for further stability ( Figure 8A).
In the case of dATP misincorporation opposite an abasic site by polβ, this ternary structure offers understandings of a protein templating basis, where the polymerase contributes hydrogen bonding interactions and geometric constraints to stabilize and favor insertion of the incoming dATP. Here, the structures of polβ misincorporating dATP opposite the Pt-GG intrastrand cross-link also demonstrate that protein templating, particularly an interaction with Lys280, in addition to pi stacking interactions of the purine with nearby bases and Tyr271, favor the mutagenic insertion. In addition, as a purine, adenine is better suited to form pi stacking interactions with its environment than a pyrimidine, which cannot contribute as significantly to stacking and hydrophobic interactions. In support of this notion, a previous report shows adenine exhibits a lower dipole moment than other bases, enabling it to contribute markedly to stacking interactions [52]. Overall, the empty space within the nascent base pair binding pocket is best filled with adenine due to its steric bulkiness as a purine and abilities to form pi stacking interactions. With a role for polβ in mutagenic replication past other DNA lesions [49,53,54], once again we are poised with evidence to appreciate the sensitivity of polβ in binding inaccurate nucleotides across bulky DNA lesions with minor alterations in protein conformation.

Conclusion
Here, we report the first structures of a DNA polymerase bypassing the major cisplatin-DNA lesion, specifically polβ-catalyzed misincorporation of dATP opposite the 5′-dG of the Pt-GG intrastrand cross-link. In the polβ active site, dATP does not form a coplanar base pairing with the 5′-dG of Pt-GG. The incoming dATP opposite the 5′-dG is accommodated via pi stacking interactions and protein templating. We also note similarities between the dATP misinsertion opposite Pt-GG lesion and the dATP misinsertion opposite an abasic site, largely due to the upstream shift in the template strand and the role of Tyr271 in stabilizing the incoming purine. Our results, together with the reported overexpression of polβ in tumors, provide insights into cisplatin-induced mutagenesis, particularly in cisplatin-treated polβ-overexpressing tumors, where this misinsertion may be inefficient yet more pervasive.

BER base excision repair
Pt-DNA platinum-DNA THF tetrahydrofuran  (A) DNA sequence used for crystallography. The downstream primer contains a 5′phosphate that is not shown. The site representing nucleotide insertion opposite the 5′-G is indicated with an arrow. (B) Overall structure of the Pt-GG•dATP*-Mg 2+ ternary complex (PDB ID: 6U2O). The protein is shown in white and its α-helix N is in red. The template strand is depicted in orange and the primer strands in yellow. The Pt-GG bases are shown in magenta, and the dC primer terminus is in green. The incoming dATP* is depicted in green.     2 R merge = Σ|I -|/ΣI, where I is the integrated intensity of a given reflection.