**5. Mechanistic and structural aspects of the ICL formation in DNA duplexes and DNA-PNA heteroduplexes**

In comparison to inter-strand crosslinks (ICLs), the intra-strand DNA crosslinks have been better studied, mostly as part of locally multiply damage sites (MDS, clustered/tandem DNA lesions). Recently, Wang and co-workers (Lin *et al.*, 2010) found that the exposure of BrdU-substituted telomer G-quadruplex DNA to UVA light results in the formation of G[8- 5]U intra-strand crosslink (Fig. 6A). This finding presents evidence that free radical reactions, involving 5-uracil-yl-radical (•U) can also be a source of ICL, albeit not in completely dsDNA duplexes. Thus, Ding & Greenberg (2010) reported radiolytic formation of ICL in anaerobic solutions by BrdU-mimetic iodo-aryl nucleotides incorporated into synthetic dsDNA duplex. Similarly to BrdU-dsDNA irradiation, in these model systems

 

 

damaged nucleotides and/or direct 2'-deoxiribose oxidation yields AP (apurinic/apyrimidinic) abasic sites as a common DNA lesion; we also assessed the role of AP-sites in the PNA-DNA ICL formation using synthetic AP-containing oligodeoxyribonucletides. We found that presence of AP-sites at different positions of the DNA strand (3'- or 5'-end, and/or penultimate to the ends) results in ICL formation without radiation, but instead required addition of a strong reductant, *e.g.* NaCNBH3. The electrophoretic gel-mobility of thus formed ICL bands resembled that of -radiation generated ones. Therefore, we concluded that AP-sites on the DNA strand are the likely partners of the free NH2, or α- and ε-amino groups of Lys at the PNA ends in the formation

> **4 2 3**

of DNA-PNA crosslinks *via* a Schiff-base reaction, followed by imine reduction.

Fig. 5. Electrophoretic migration of bands representing the covalent PNA-DNA dimers (ICL). Aqueous solutions of 32P-DNA hybridized with various PNA were -irradiated (750 Gy) under N2 – atmosphere and in the presence of 25 mM EDTA. Samples differ only by the type of capping groups on PNA ends (R1 and R2, Fig. 3). Lanes from left to right: (A) NH2-K-PNA; (B) NH2-PNA; (C) Ac-PNA; (D) MeMor-K-PNA; (E) NH2-PNA-K; (F) Ac-PNA-K; and (G) MeMor-PNA-K; (1- 4): different crosslinked products involving the different amino groups. No ICL are formed with Ac-PNA (C) and MeMor-K-PNA (D). Adapted from

**5. Mechanistic and structural aspects of the ICL formation in DNA duplexes** 

In comparison to inter-strand crosslinks (ICLs), the intra-strand DNA crosslinks have been better studied, mostly as part of locally multiply damage sites (MDS, clustered/tandem DNA lesions). Recently, Wang and co-workers (Lin *et al.*, 2010) found that the exposure of BrdU-substituted telomer G-quadruplex DNA to UVA light results in the formation of G[8- 5]U intra-strand crosslink (Fig. 6A). This finding presents evidence that free radical reactions, involving 5-uracil-yl-radical (•U) can also be a source of ICL, albeit not in completely dsDNA duplexes. Thus, Ding & Greenberg (2010) reported radiolytic formation of ICL in anaerobic solutions by BrdU-mimetic iodo-aryl nucleotides incorporated into synthetic dsDNA duplex. Similarly to BrdU-dsDNA irradiation, in these model systems

**A B C D E F G** 

**1** 

 

 

Gantchev *et al.*, (2009).

**and DNA-PNA heteroduplexes** 

**2 3**

 **1** 

 **2**

stand breaks and alkali-labile lesions were also induced at the halogenated site and at the flanking nucleotides. The conclusions from this study highlight the importance of the local DNA structure (wobble *vs.* normal duplex) in terms of Watson-Crick pairing restrictions that likely prevent the 5-uracil-yl σ-radical (in contrast to the aryl radical) interaction with the opposite strand bases to produce ICL. The chemical structures of several DNA interstrand crosslinks have been reported only recently. Several groups have focused on the identification of ICL structures synthesized in model systems using light- and/or oxidationsensitive precursors (Hong *et al.*, 2006; Weng *et al.*, 2007; Peng *et al.*, 2008; Kim & Hong, 2008; Op de Beeck & Madder, 2011). A few structures have been definitively identified in natural conditions and, to our knowledge, only one in cellular DNA (Regulus *et al.*, 2007) (Fig. 6C). However, the authors in Regulus *et al.*, (2007) failed to present evidence that this crosslink is exclusively an inter-strand crosslink in cells. Mechanistically, the sole ICL-structure that is exclusively generated with the participation of a primary radiation-induced 5-(2' deoxyuridinyl) methyl free radical, a product of •OH-induced hydrogen-atom abstraction from thymine, is the T[5m-6n]A crosslink (Ding & Greenberg, 2007; Ding *et al.*, 2008) (Fig. 6B).

A common pathway for ICL formation in dsDNA is the condensation reaction between aldehydes (*e.g.* in abasic DNA sites) and exocyclic amines of opposite bases. Under radiation abasic (AP-apurinic/apyrimidinic) sites are generated either *via* direct H-atom abstraction by •OH radicals from 2'-deoxyribose or after oxidative base damage followed by N-glycosidic bond-cleavage. Oxidation of each of the five positions in 2'-deoxyribose in DNA is possible, but under -radiation the best known reactions involve H-atom abstraction at C1'-, C2'- and C4'-positions. The 4'-keto abasic site formed after C4'-oxidation (C4'-AP) is generally known as the "native" abasic site (Chen & Stubbe 2004). Subsequent interactions of sugar radicals with oxygen and/or elimination reactions give a variety of closedcycle/open-chain aldehydic products, accompanied (or not) by DNA strand-cleavage (Dedon, 2008). One of the first reported, and structurally identified DNA-ICL generated by -radiation and/or selective 4'-position 2'-deoxyribose oxidation by bleomycin in model systems and in cells (Fig. 6C; Regulus *et al.*, 2007; Cadet *et al.*, 2010) is produced *via* electrophilic interaction between 2'-deoxypentos-4-ulose abasic site (opened C4'-AP) and N4-dC. Formation of this ICL is accompanied by a DNA strand break. In a series of works, Greenberg and collaborators studied the ICL formation with participation of oxidized native C4'-AP (Sczepanski *et al.*, 2008, 2009a) and 5′-(2-phosphoryl-1,4-dioxobutane, DOB) (Guan & Greenberg, 2009) sites. DOB is produced concomitantly with a single-strand break byDNAdamaging agents capable of abstracting an H-atom from the C5′-position. When oxidized C4'-AP was opposed by dA, a single crosslink formation occurred exclusively with an adjacent dA on the 5′-side. The crosslink formation was attributed to condensation of C4'- AP with the N6-amino group of dA and less favorably with N4-amino group of dC, but not with dG or dT. Interestingly, C4'-AP produced ICLs in which both strands are either intact or ICLs, where the C4-AP containing strand was cleaved (3′ to the AP-site), while DOB-ICLs were always accompanied by an adjacent to the AP-site single-strand break, and thus constituting a clustered type (MDS) lesion.

In contrast, following -radiation of BrdU-substituted wobble-DNA (scDNA) duplexes multiple crosslinked products were generated which impedes their chemical identification (Cecchini et al., 2005a; Dextraze et al., 2009). However, in DNA-PNA heteroduplexes, because the amino-groups attached to PNA are exogenous and could be omitted/varied,

DNA Radiosensitization: The Search for Repair Refractive

bubbles in scDNA, only (see below).

Lesions Including Double Strand Breaks and Interstrand Crosslinks 427

together with the synthetically positioned AP-sites on DNA the participation of these two entities in the ICL formation was positively identified (Gantchev *et al.*, 2009). Our data are consistent with a mechanism of ICL formation that involves formation of Schiff bases between PNA-amino functional groups and radiation-damage induced AP-sites on DNA. This type of covalent bonding is widely accepted to take place in the formation of covalent links between NH2-peptide (protein) groups and damaged (aldehydic) DNA sites, albeit in the presence of an exogenous reducing agent (Mazumder *et al.*, 1996). The new finding is that apart from the prerequisite •OH-mediated, or direct -damage of DNA (formation of AP-sites), -radiation also provides reducing equivalents to transform the initially formed Schiff base linkage into a more stable reduced bond (amine), *i.e.* to produce irreversible ICL (Fig. 6D). This presents a typical case of radiation-induced MDS, where the synergism of the interactions of •OH, e¯aq, and possibly even •H radicals on PNA-DNA results in ICL. The 3D-modeling (Gantchev *et al.*, 2009) confirms experimental data that open-chain C4'-AP at several DNA-strand end, or penultimate positions are structurally allowed to form covalent bonds with the ε- and α-amino groups of opposite Lys residues, or PNA NH2-terminal groups and in all cases although, intra-helical ICLs are solvent accessible (*e.g.* the transient Schiff bases are available for interaction with e¯aq). Importantly, if dsDNA duplexes are compared, the e¯aq and solvent accessibility holds for the open structures in BrdU-DNA

Solvated electrons (e¯aq) are indispensible species for the formation of both strand breaks and interstrand crosslinks sensitized by BrdU in scDNA and crosslinks in DNA-PNA heteroduplexes. The e¯aq interaction rate with oxygen, k(e¯aq+ O2) = 2x1010 M-1s-1 is high, therefore hypoxic experimental conditions are important. The radiosensitization properties of BrdU are based on its ability to undergo dissociative electron transfer (ET) which is initiated by electron capture either from solution, or following excess ET from surrounding DNA bases (Fig. 1). The classical reducibility (electron affinity, EA) trend of nucleobases is BrU>U~T>C>A>G (Aflatooni *et al.*, 1998; Richardson *et al.*, 2004), with BrU being only ~ 40 mV easier to reduce than thymidine (Gaballah *et al.*, 2005). Using the approach described by Michaels & Hunt (1978) and quantitation of BrdU-mediated damage in mismatched duplexes, we calculated a value for k(e¯aq + BrdU-scDNA) of ~2x1010 M-1s-1. This value is particularly interesting in that the rate constant for BrdU interaction with e¯aq in mismatched, scDNA (single-stranded regions of the duplex) is practically the same as for the free base BrU in solution (Zimbrick *et al.*, 1969a) (*i.e.* essentially diffusion controlled) and, about two orders of magnitude higher than in normal dsDNA. Based on our results from the irradiation of solutions containing PNA-DNA hybrids, we calculated a rate constant for the formation of PNA-DNA crosslinks, assuming only interactions with hydrated electrons, equal to: k(e¯aq + PNA) ~ 5x109 M-1s-1, which is also high. While high rate interaction of e¯aq with PNA-DNA hetereoduplexes can be attributed to the lack of electrostatic repulsion (see Fig. 4 and legend), the increased rate of interaction with wobble scDNA is less obvious. We hypothesized that e−aq may have a restricted access to BrdU when incorporated in a normal DNA duplex, although the Br-atom is partially solvent-exposed in the major groove. To address this issue we applied molecular modeling and nanosecond scale molecular dynamics (MD) simulations, where the excess electron in solution was modeled as a localized e−(H2O)6 anionic cluster (Gantchev & Hunting, 2008, 2009). We compared the dynamics and interactions of e<sup>−</sup>aq with dsDNA containing a normal BrdU•dA pair in the center of the duplex (dsDNA) with that of a wobble DNA containing a single mismatched

Fig. 6. Chemical structures of known intra- and interstrand crosslinks as generated in model DNA systems, UV and ionizing radiation: (A) The structure of G[8-5]U intrastrand crosslink; the only known crosslink formed *via* direct addition to •U-radical (3). This pathway was not found for ICL formation in normal dsDNA, probably due to steric restrictions, however interstrand crosslinks generated with the participation of 5-uracil-yl radical are possible in wobble scDNA (Fig. 1); (B) The formation of T[5m-6n]A ICL initiated by •OH radical Habstraction from 5-CH3dT (4); the intermediate product is addition to N1 (5), which is further rearranged to the final product (6), (Ding *et al.*, 2008); (C) The ICL (11) formed *via* condensation reaction of exocyclic NH2-group of dC with β-elimination product (9) of an oxidized C4'-AP abasic site (7 and 8). This ICL is associated with a SSB formation, (Regulus *et al*., 2007) and; (D) Reaction of C4'-AP native abasic site (9) with L-Lys-capped PNA resulting in the formation of PNA-DNA interstrand crosslinks *via* Schiff base (12 and 13). Two free NH2-groups are equally reactive which can produce two ICLs of different structure (14 and 15). The abasic sites are formed by -radiation oxidation of DNA (*e.g.* by •OH radicals). The concomitant solvated electrons, eaq− are the essential reducing species required to convert Schiff bases in irreversible ICLs, (Gantchev *et al.*, 2009).

H2C

N N N N NH2

CH2

O

O

<sup>O</sup> <sup>O</sup> DNA DNA

Fig. 6. Chemical structures of known intra- and interstrand crosslinks as generated in model DNA systems, UV and ionizing radiation: (A) The structure of G[8-5]U intrastrand crosslink; the only known crosslink formed *via* direct addition to •U-radical (3). This pathway was not found for ICL formation in normal dsDNA, probably due to steric restrictions, however interstrand crosslinks generated with the participation of 5-uracil-yl radical are possible in wobble scDNA (Fig. 1); (B) The formation of T[5m-6n]A ICL initiated by •OH radical Habstraction from 5-CH3dT (4); the intermediate product is addition to N1 (5), which is further rearranged to the final product (6), (Ding *et al.*, 2008); (C) The ICL (11) formed *via* condensation reaction of exocyclic NH2-group of dC with β-elimination product (9) of an oxidized C4'-AP abasic site (7 and 8). This ICL is associated with a SSB formation, (Regulus *et al*., 2007) and; (D) Reaction of C4'-AP native abasic site (9) with L-Lys-capped PNA resulting in the formation of PNA-DNA interstrand crosslinks *via* Schiff base (12 and 13). Two free NH2-groups are equally reactive which can produce two ICLs of different structure (14 and 15). The abasic sites are formed by -radiation oxidation of DNA (*e.g.* by •OH radicals). The concomitant solvated electrons, eaq− are the essential reducing species

O

<sup>O</sup> <sup>O</sup> DNA DNA

O O O

O <sup>O</sup> <sup>O</sup>

9

DNA

NH2

NH2

N

12

N

O NH2

PNA

eaq H eaq <sup>+</sup> <sup>H</sup><sup>+</sup>

PNA

14 15

[C4'-AP] ― NH2-PNA

<sup>O</sup> DNA

OH H H

O

HO

N HN O

4

O O

O

O

N H2C

OH

5

H2C

HN N N N NH

O

O

N N N NH

> O O

<sup>N</sup> PNA O

13

<sup>N</sup> PNA O

O O O

O

O

T[5m-6n]A ICL

OH 6 H

NH2 PNA

<sup>O</sup> H2N

<sup>O</sup> H2N

O <sup>O</sup> <sup>O</sup> DNA DNA

O <sup>O</sup> <sup>O</sup> DNA DNA

6

HO

N HN O

<sup>O</sup> <sup>N</sup>

<sup>H</sup> <sup>H</sup>

O O

HO

N HN O

O O

O

O

O

A B

NH O

N O O P O O-O

Br

1

NH O

N O O O P O O-O

O

C D

O O

N N NH2

O O

O

O

required to convert Schiff bases in irreversible ICLs, (Gantchev *et al.*, 2009).

10

O

HN N N O

H2N <sup>N</sup>

O O 0

UVA

HN N N O

H2N <sup>N</sup> O O 0

HOO O

> O O

N N O

N OH

O HO OH

O

HH

O O

11

O O O

[C4'-AP] ― dC

G[8-5]U Intrastrand CL

O

7 8 9

together with the synthetically positioned AP-sites on DNA the participation of these two entities in the ICL formation was positively identified (Gantchev *et al.*, 2009). Our data are consistent with a mechanism of ICL formation that involves formation of Schiff bases between PNA-amino functional groups and radiation-damage induced AP-sites on DNA. This type of covalent bonding is widely accepted to take place in the formation of covalent links between NH2-peptide (protein) groups and damaged (aldehydic) DNA sites, albeit in the presence of an exogenous reducing agent (Mazumder *et al.*, 1996). The new finding is that apart from the prerequisite •OH-mediated, or direct -damage of DNA (formation of AP-sites), -radiation also provides reducing equivalents to transform the initially formed Schiff base linkage into a more stable reduced bond (amine), *i.e.* to produce irreversible ICL (Fig. 6D). This presents a typical case of radiation-induced MDS, where the synergism of the interactions of •OH, e¯aq, and possibly even •H radicals on PNA-DNA results in ICL. The 3D-modeling (Gantchev *et al.*, 2009) confirms experimental data that open-chain C4'-AP at several DNA-strand end, or penultimate positions are structurally allowed to form covalent bonds with the ε- and α-amino groups of opposite Lys residues, or PNA NH2-terminal groups and in all cases although, intra-helical ICLs are solvent accessible (*e.g.* the transient Schiff bases are available for interaction with e¯aq). Importantly, if dsDNA duplexes are compared, the e¯aq and solvent accessibility holds for the open structures in BrdU-DNA bubbles in scDNA, only (see below).

Solvated electrons (e¯aq) are indispensible species for the formation of both strand breaks and interstrand crosslinks sensitized by BrdU in scDNA and crosslinks in DNA-PNA heteroduplexes. The e¯aq interaction rate with oxygen, k(e¯aq+ O2) = 2x1010 M-1s-1 is high, therefore hypoxic experimental conditions are important. The radiosensitization properties of BrdU are based on its ability to undergo dissociative electron transfer (ET) which is initiated by electron capture either from solution, or following excess ET from surrounding DNA bases (Fig. 1). The classical reducibility (electron affinity, EA) trend of nucleobases is BrU>U~T>C>A>G (Aflatooni *et al.*, 1998; Richardson *et al.*, 2004), with BrU being only ~ 40 mV easier to reduce than thymidine (Gaballah *et al.*, 2005). Using the approach described by Michaels & Hunt (1978) and quantitation of BrdU-mediated damage in mismatched duplexes, we calculated a value for k(e¯aq + BrdU-scDNA) of ~2x1010 M-1s-1. This value is particularly interesting in that the rate constant for BrdU interaction with e¯aq in mismatched, scDNA (single-stranded regions of the duplex) is practically the same as for the free base BrU in solution (Zimbrick *et al.*, 1969a) (*i.e.* essentially diffusion controlled) and, about two orders of magnitude higher than in normal dsDNA. Based on our results from the irradiation of solutions containing PNA-DNA hybrids, we calculated a rate constant for the formation of PNA-DNA crosslinks, assuming only interactions with hydrated electrons, equal to: k(e¯aq + PNA) ~ 5x109 M-1s-1, which is also high. While high rate interaction of e¯aq with PNA-DNA hetereoduplexes can be attributed to the lack of electrostatic repulsion (see Fig. 4 and legend), the increased rate of interaction with wobble scDNA is less obvious. We hypothesized that e−aq may have a restricted access to BrdU when incorporated in a normal DNA duplex, although the Br-atom is partially solvent-exposed in the major groove. To address this issue we applied molecular modeling and nanosecond scale molecular dynamics (MD) simulations, where the excess electron in solution was modeled as a localized e−(H2O)6 anionic cluster (Gantchev & Hunting, 2008, 2009). We compared the dynamics and interactions of e<sup>−</sup>aq with dsDNA containing a normal BrdU•dA pair in the center of the duplex (dsDNA) with that of a wobble DNA containing a single mismatched

DNA Radiosensitization: The Search for Repair Refractive

sensitized damages (DSB and ICL) in wobble scDNA.

Lesions Including Double Strand Breaks and Interstrand Crosslinks 429

In addition, the incorporated mismatched pairs (T^T or BU^T) alter the dynamics of the neighboring bases due to incomplete 5′-stacking. Together with the narrowing of the minor groove these phenomena bring the two strands closer which creates conditions for crossstrand (cs) stacking and single and multiple cross-strand (cs) H-bonding not only within the mismatched regions, but also encompassing penultimate nucleotides to create extended "zipper-like" motives (Špačková *et al*., 2000). The properties of single-mismatch scDNA duplexes, including the effect of the nearest sequence context (*e.g.* presence of T-tract DNA) have been discussed elsewhere (Gantchev *et al*., 2005). A schematic presentation of the most often formed cross-strand inter-base contacts is given in Fig. 8. The close presence of eaq¯, although causes dynamic instability and fluctuations around the mismatched BrdU^dT pair does not abolish, but in contrast, provokes additional frequent cs-H-bonding interactions (Gantchev & Hunting, 2009). All these findings are important in terms of facilitated chargetransfer along UV-, or -activated BrdU-scDNA. The intrahelical electron or hole transfer to BrdU and/or •U-yl radical are the next important factor that is largely expected to control the efficiency (and location) of the ensuing DNA damage; the formation of DSB and ICL. Indeed, recently a more effective electron transfer has been reported for mismatched duplexes than for fully complementary DNA (Ito *et al.*, 2009). Using a two electron acceptor DNA model system with incorporated BrdA, BrdG, BrdU and TT-dimer Fazio *et al.* (Fazio *et al.*, 2011) were able to estimate the absolute electron-hoping rates in DNA and have shown that the electron transfer is more efficient in 5' → 3' direction. As mentioned, in unsubstituted DNA pyrimidine rather than purine bases have been considered as trapping sites for excess electrons. This is illustrated by resonant free electron attachment experiments (Stokes *et al.*, 2007) which show that both thymine and cytosine form stable valence anions for low energy electrons, *i.e.* both thymine and cytosine possess positive adiabatic electron affinities. However, recently a stabile anionic state of adenine (A−) has been detected (Haranczyk *et al.*, 2007). Subsequently, this finding has been shown to have a pronounced effect in the ultrafast ET in DNA and on dissociative bond cleavage (Wang *et al.*, 2009), including ET to BrdU from A− acting as primary trap of radiolysis-generated pre-hydrated electrons (Wang *et al.*., 2010). These new developments in the field add to the existing puzzles of the precise determination of successive chain events leading to multiple BrdU-

Repair of interstrand crosslinks (ICLs) requires multiple strand incisions to separate the two covalently linked DNA strands. It is unclear how these incisions are generated. DNA double-strand breaks (DSBs) have been identified as intermediates in ICL repair, but eukaryotic enzymes responsible for producing these intermediates are not well known (Wang, 2007; Moldovan & D'Andrea, 2009a,b; D'Andrea & Grompe, 2003; Liu *et al.*, 2010; Hanada *et al.*, 2006). Ongoing research shows that in cell free model systems ICLs of different chemical structure exert different effects during repair and some may be difficult to repair. The repair refractive character of a particular ICL resulting from the C4'-AP abasic site and identified to occur as a clustered ICL-SSB lesion (Sczepanski *et al.*, 2008, 2009a) was recently demonstrated to give rise to even more toxic DSBs when subjected to NER (Sczepanski *et al.*., 2009b). Likewise, during UvrABC nucleotide excision repair of the welldefined T[5m-6n]A single-lesion crosslink imbedded in dsDNA (Fig. 6B, Ding *et al.*, 2008),

DNA packing into chromatin adds to the complexity of DNA damage recognition and removal, because the highly condensed chromatin is, in general, refractory to DNA repair (Hara *et al*., 2000; Thoma, 2005). In order to grant access to DNA repair machinery, the

DSB were produced in almost 30% of the excision events (Peng *et al.*, 2010).

BrdU^dT pair (scDNA), *i.e.* replacing dA with dT. Rather unexpectedly we found that the occupancy of the close-to-DNA space for scDNA and dsDNA at cut-off distance <5 Å was 0.7% vs. 1.6%, respectively (from a total of 4,000 MD configurations). However, the electron interacted with a larger number of individual bases in scDNA. For instance, in dsDNA, the electron moved closely toward only four nucleobases, all from the non-brominated DNA strand, while in scDNA eleven nucleobases from both strands were found to come within reach of eaq¯. The different clustering of the electron (occupation of close to DNA sites) in both duplexes is presented graphically in Fig. 7 (see legend for details). Notably, BrU incorporated in the central (sixth) position of both DNA duplexes, was approached by eaq¯ several times in scDNA only. Likewise only in scDNA, the eaq¯ preferentially occupied close sites and formed contacts with the two most perturbed thymidines (dT5 and dT7) flanking BrdU. At present, there is no explanation for the disparity of eaq¯ interactions with dsDNA *vs*. scDNA, other than the different dynamic structure of the isosteric DNA sequences under study, including the dynamics of structured water and Debay-Hückel layers (Gantchev & Hunting, 2008, 2009). The exposure of wobble-pair pyrimidine carbonyl groups into the DNA grooves results in excess solvation of the mismatched pairs (Sherer & Cramer, 2004).

Fig. 7. Superimposed snapshots from ns MD of regular (left) and mismatched (right) 11 mer DNA containing a Watson-Crick BrU6•A17 normal pair, or BrU^T17 wobble pair, respectively. The e¯aq is represented by a e−[H2O]6 anionic cluster. The shown dynamic states are selected by the rule, distance |e¯aq – nucleobase|< 5 Å. From the total of 59 states for normal DNA, there are no configurations where e¯aq is close to the central BrU6**•**A17 base pair. In contrast, in the wobble scDNA from the all 22 states obeying the same selection rule, ~ 65% show close approach to the wobble BrU6^T17 pair. The electron resides most often in the vicinity of the flanking T7 base and less frequently approaches BrU6 and T17. Hydration water and counterions are not shown; BrU•A and BrU^T are in the middle and labeled. Color code: Br-vdW sphere (green); DNA backbone (cyan); nucleotide atoms (standard color); e¯aq (yellow). For details see: Gantchev & Hunting, (2008, 2009).

BrdU^dT pair (scDNA), *i.e.* replacing dA with dT. Rather unexpectedly we found that the occupancy of the close-to-DNA space for scDNA and dsDNA at cut-off distance <5 Å was 0.7% vs. 1.6%, respectively (from a total of 4,000 MD configurations). However, the electron interacted with a larger number of individual bases in scDNA. For instance, in dsDNA, the electron moved closely toward only four nucleobases, all from the non-brominated DNA strand, while in scDNA eleven nucleobases from both strands were found to come within reach of eaq¯. The different clustering of the electron (occupation of close to DNA sites) in both duplexes is presented graphically in Fig. 7 (see legend for details). Notably, BrU incorporated in the central (sixth) position of both DNA duplexes, was approached by eaq¯ several times in scDNA only. Likewise only in scDNA, the eaq¯ preferentially occupied close sites and formed contacts with the two most perturbed thymidines (dT5 and dT7) flanking BrdU. At present, there is no explanation for the disparity of eaq¯ interactions with dsDNA *vs*. scDNA, other than the different dynamic structure of the isosteric DNA sequences under study, including the dynamics of structured water and Debay-Hückel layers (Gantchev & Hunting, 2008, 2009). The exposure of wobble-pair pyrimidine carbonyl groups into the DNA grooves results in excess solvation of the mismatched pairs (Sherer & Cramer, 2004).

Fig. 7. Superimposed snapshots from ns MD of regular (left) and mismatched (right) 11 mer DNA containing a Watson-Crick BrU6•A17 normal pair, or BrU^T17 wobble pair, respectively. The e¯aq is represented by a e−[H2O]6 anionic cluster. The shown dynamic states are selected by the rule, distance |e¯aq – nucleobase|< 5 Å. From the total of 59 states for normal DNA, there are no configurations where e¯aq is close to the central BrU6**•**A17 base pair. In contrast, in the wobble scDNA from the all 22 states obeying the same selection rule, ~ 65% show close approach to the wobble BrU6^T17 pair. The electron resides most often in the vicinity of the flanking T7 base and less frequently approaches BrU6 and T17. Hydration water and counterions are not shown; BrU•A and BrU^T are in the middle and labeled. Color code: Br-vdW sphere (green); DNA backbone (cyan); nucleotide atoms (standard color); e¯aq (yellow). For details see: Gantchev &

A B

Hunting, (2008, 2009).

In addition, the incorporated mismatched pairs (T^T or BU^T) alter the dynamics of the neighboring bases due to incomplete 5′-stacking. Together with the narrowing of the minor groove these phenomena bring the two strands closer which creates conditions for crossstrand (cs) stacking and single and multiple cross-strand (cs) H-bonding not only within the mismatched regions, but also encompassing penultimate nucleotides to create extended "zipper-like" motives (Špačková *et al*., 2000). The properties of single-mismatch scDNA duplexes, including the effect of the nearest sequence context (*e.g.* presence of T-tract DNA) have been discussed elsewhere (Gantchev *et al*., 2005). A schematic presentation of the most often formed cross-strand inter-base contacts is given in Fig. 8. The close presence of eaq¯, although causes dynamic instability and fluctuations around the mismatched BrdU^dT pair does not abolish, but in contrast, provokes additional frequent cs-H-bonding interactions (Gantchev & Hunting, 2009). All these findings are important in terms of facilitated chargetransfer along UV-, or -activated BrdU-scDNA. The intrahelical electron or hole transfer to BrdU and/or •U-yl radical are the next important factor that is largely expected to control the efficiency (and location) of the ensuing DNA damage; the formation of DSB and ICL. Indeed, recently a more effective electron transfer has been reported for mismatched duplexes than for fully complementary DNA (Ito *et al.*, 2009). Using a two electron acceptor DNA model system with incorporated BrdA, BrdG, BrdU and TT-dimer Fazio *et al.* (Fazio *et al.*, 2011) were able to estimate the absolute electron-hoping rates in DNA and have shown that the electron transfer is more efficient in 5' → 3' direction. As mentioned, in unsubstituted DNA pyrimidine rather than purine bases have been considered as trapping sites for excess electrons. This is illustrated by resonant free electron attachment experiments (Stokes *et al.*, 2007) which show that both thymine and cytosine form stable valence anions for low energy electrons, *i.e.* both thymine and cytosine possess positive adiabatic electron affinities. However, recently a stabile anionic state of adenine (A−) has been detected (Haranczyk *et al.*, 2007). Subsequently, this finding has been shown to have a pronounced effect in the ultrafast ET in DNA and on dissociative bond cleavage (Wang *et al.*, 2009), including ET to BrdU from A− acting as primary trap of radiolysis-generated pre-hydrated electrons (Wang *et al.*., 2010). These new developments in the field add to the existing puzzles of the precise determination of successive chain events leading to multiple BrdUsensitized damages (DSB and ICL) in wobble scDNA.

Repair of interstrand crosslinks (ICLs) requires multiple strand incisions to separate the two covalently linked DNA strands. It is unclear how these incisions are generated. DNA double-strand breaks (DSBs) have been identified as intermediates in ICL repair, but eukaryotic enzymes responsible for producing these intermediates are not well known (Wang, 2007; Moldovan & D'Andrea, 2009a,b; D'Andrea & Grompe, 2003; Liu *et al.*, 2010; Hanada *et al.*, 2006). Ongoing research shows that in cell free model systems ICLs of different chemical structure exert different effects during repair and some may be difficult to repair. The repair refractive character of a particular ICL resulting from the C4'-AP abasic site and identified to occur as a clustered ICL-SSB lesion (Sczepanski *et al.*, 2008, 2009a) was recently demonstrated to give rise to even more toxic DSBs when subjected to NER (Sczepanski *et al.*., 2009b). Likewise, during UvrABC nucleotide excision repair of the welldefined T[5m-6n]A single-lesion crosslink imbedded in dsDNA (Fig. 6B, Ding *et al.*, 2008), DSB were produced in almost 30% of the excision events (Peng *et al.*, 2010).

DNA packing into chromatin adds to the complexity of DNA damage recognition and removal, because the highly condensed chromatin is, in general, refractory to DNA repair (Hara *et al*., 2000; Thoma, 2005). In order to grant access to DNA repair machinery, the

DNA Radiosensitization: The Search for Repair Refractive

as high as 0.46%, *i.e.* 6.5-fold above the background.

**6. Conclusion** 

**7. References** 

Lesions Including Double Strand Breaks and Interstrand Crosslinks 431

difficult to repair without a loss of DNA information, and/or would require major chromatin remodeling. Indeed, Faruqi *et al*. (1998) designed PNAs that bind to the *sup*FG1 mutation reporter gene and found that 8- or 10- b.p. PNA bound to this site induces mutations at frequencies in the range of 0.1%, well above the *in vivo* background. Later, the same group (Kim *et al*., 2006) demonstrated that a psoralen-*bis*-PNA conjugate directs the formation of a photoadduct at the 5'-TpA step of the PNA binding site to the same *sup*FG1 gene. In mammalian cells, the UV-generated PNA-targeted psoralen phtoadducts induced mutations

In conclusion, there is a need to better understand the parameters which control the formation and repair of complex DNA lesions, such as interstrand crosslinks. Such complex, repair refractive lesions may offer a means to selectively kill tumor cells by taking advantage

Aflatooni K., Gallup G. A., Burrow P.D. (1998) Electron Attachment Energies of the DNA

Cadet J., Douki T., Ravanat J.-L. (2010) Oxidatively generated base damage to cellular DNA.

Cai Y., Patel D.J., Geacintov N.E., Broyde S. (2007) Dynamics of benzo[a]pyren-derived

Cecchini S., Girouard S., Huels M.A., Sanche L., Hunting D.J. (2004) Single strand specific radiosensitization of DNA by Bromodeoxyuridine. *Radiation Research* 162, 604-615. Cecchini S., Girouard S., Huels M.A., Sanche L., Hunting D.J. (2005a) Interstrand crosslinks:

Cecchini S., Masson C., La Madeleine C., Huels M.A., Sanche L., Wagner J.R., Hunting D.J.

Chen J., Stubbe JA. (2004) Synthesis and Characterization of Oligonucleotides Containing a

Chen T., Cook G.P., Koppisch A.T., Greenberg M.M. (2000) Investigation of the Origin of the

D'Andrea A.D., Grompe M. (2003) The Fanconi anaemia/BRCA pathway. *Nature Reviews* 

Dedon P.C. (2008) The Chemical Toxycology of 2-Deoxyribose Oxidation in DNA. *Chemical* 

guanine DNA lesion in TGT and CGC sequence contexts: Enhanced mobility in TGT explains conformational heterogeneity, flexible bending, and greater susceptibility to nucleotide excision repair. *Journal of Molecular Biology* 374, 292-305.

a new type of gamma-ray damage in bromodeoxyuridine-substituted DNA.

(2005b) Interstrand cross-link induction by UV radiation in bromodeoxyuridinesubstituted DNA: Dependence on DNA conformation. *Biochemistry*, 44, 16957-

Sequence Selectivity for the 5-Halo-2'-deoxyuridine Sensitization of DNA to Damage by UV-Irradiation. *Journal of the American Chemical Society* 122, 3861-3866. Cutrona G., Carpaneto E.M., Ulivi M., Roncella S., Landt O., Ferrarini M., Boffa L.C. (2000)

Effects in live cells of a c-*myc* anti-gene PNA linked to a nuclear localization signal.

of either enhanced formation or reduced repair within the tumor environment.

Bases. *Journal of Physical Chemistry A*, 102, 6205–6207.

*Free Radical Biology & Medicine* 49, 9–21.

4'-Keto Abasic Site. *Biochemistry* 43, 5278–5286.

*Biochemistry* 44, 1932–1940.

*Nature Biotechnology* 18, 300-303.

*Research in Toxicology* 21, 206-219.

*Cancer* 3, 23-34.

16966.

Fig. 8. Nearest-neighbour sequence effects in wobble semi-complementary DNA (scDNA) as observed by MD (adapted from Gantchev *et al*. 2005). Schematic presentation of the frequent cross-strand (cs) inter-base contacts formed in the studied 11-mer DNA duplexes containing a single mismatch: T^T or T^BrU, incorporated in the central triplets: d(CTA)(TTG) (A), d(TTT)(ATA) (B) and d(TBrdUT)(ATA) (C): bold dashed lines (most frequently observed cs H-bons), dotted lines (less frequent cs H-bonds). Note that cs contacts in (C) coincide with those observed also in the presence of e¯aq (Gantchev & Hunting, 2009). These data underline the importance of the wobble DNA dynamic structure for both, interstrand ET and high-frequency opposite-strand atom encounter for the generation of (asymmetric) ICL (Fig. 1).

chromatin response to DNA damage involves activation of ATP-dependent chromatinremodeling complexes and histone post-translational modification pathways (Peterson & Côte, 2004; Nag & Smerdon, 2009; Méndez-Acuña *et al*., 2010). Again, DSBs recognition and repair in the context of chromatin rearrangement is better studied and understood at the expense of other DNA damages, such as ICLs. One crucial chromatin modification, the phosphorylation of the histone variant H2AX (γH2AX) is perhaps the best example of a histone modification in response to DSB induction in DNA (Van Attikum & Gasser, 2005). Despite the progress achieved in understanding of the repair of certain UV-induced DNA damages (intra-strand crosslinks), *e.g.* cyclobutane pyrimidine dimers (CPD) and 6-4 pyrimidine photoadducts (6-4 PP), or the acetylaminofluorene-guanine (AAF-G) covalent adduct, little is known about the effects of other bulky DNA lesions (*e.g.* ICLs) on the nucleosome structural dynamics and its interplay with the versatile NER pathway (Smerdon & Lieberman, 1978; Pehrson, 1995; Gaillard et al., 2003; Gospodinov & Herceg, 2011). There is a consensus that NER functionality depends primarily on the damage recognition step, which in turn depends on the degree of DNA helix distortion induced by a particular lesion (Cai *et al*., 2007). It has been hypothesized that structurally different interstrand crosslinks would affect chromatin remodeling and damage recognition in different ways, and some ICLs might retain their refractive character to recognition/repair, or at least will exert an altered repair efficacy. Thus, a recent *in vivo* study (Hlavin *et al.*, 2010) confirmed that the structure of synthetic interstrand crosslinks between mismatched bases affects the repair rate (in this case, transcription coupled NER). It can be further hypothesized that PNA-patches hybridized to DNA (*e.g.*, > 8-10 b.p.; PNA- invaded DNA strands, PNA-DNA triple helices, and/or DNA-PNA covalent adducts) would be difficult to repair without a loss of DNA information, and/or would require major chromatin remodeling. Indeed, Faruqi *et al*. (1998) designed PNAs that bind to the *sup*FG1 mutation reporter gene and found that 8- or 10- b.p. PNA bound to this site induces mutations at frequencies in the range of 0.1%, well above the *in vivo* background. Later, the same group (Kim *et al*., 2006) demonstrated that a psoralen-*bis*-PNA conjugate directs the formation of a photoadduct at the 5'-TpA step of the PNA binding site to the same *sup*FG1 gene. In mammalian cells, the UV-generated PNA-targeted psoralen phtoadducts induced mutations as high as 0.46%, *i.e.* 6.5-fold above the background.
