The recognition and UvrABC incision of the first interstrand cross-link discovered (2
) that results from a DNA radical produced as a result of oxidative stress (e.g. γ-radiolysis, hydroxyl radical) was examined (28
). The cross-linked product is incised less rapidly than a 2′-deoxyuridine derivatized at its C5-position by a fluoresceinylated alkyl chain, which is in a way a gold standard for UvrABC reactivity (36
). However, the ICL is cleaved with comparable efficiency to a tandem lesion containing thymine glycol and was a slightly poorer substrate than the isolated glycol (46
). In addition, the distribution of cleavage sites with respect to the position of the cross-linked nucleotides is very similar to the incision of other lesions by the UvrABC complex. For instance, the oligonucleotide fragment excised from damaged DNA is typically 11–13 nucleotides in length (9
). Indeed, the major incision sites in the top strands of 6
are between 11 and 14 nucleotides apart (). Incision on the 5′-side of the cross-link occurs most often at the 8th
phosphodiester from the damaged nucleotide. The majority of the ICLs are cleaved closer (3rd
phosphodiester) to the cross-link position on the 3′-side. These sites are comparable to those incised when UvrABC encounters a psoralen cross-link (37
The above data do not address the proclivity of the enzyme to cleave either strand. Previous studies demonstrated that the sequence surrounding the psoralen cross-link plays a large role in determining whether only the furan cross-linked strand is cleaved or both strands are by UvrABC (37
). Incision on both strands is discernible in any given experiment. We found that the strand containing the cross-linked dA (“Y”) is preferentially incised in the ICLs (6
), indicating that the orientation of this cross-link is more important in determining which strand is cleaved, than the surrounding DNA sequence. The observed strand selectivity is independent of whether the 5′- or 3′-terminus of a particular strand (e.g. the oligonucleotide containing the cross-linked dA) is radiolabeled, consistent with the coupled mechanism for strand scission by UvrABC. In contrast, the observed strand selectivity depends strongly on which oligonucleotide in the duplex is radiolabeled.
The preference is more readily apparent when the strand containing the cross-linked dA (“Y”) is radiolabeled ( and ). In these duplexes the dA containing strand is cleaved ~5–10-times more frequently than the one containing the cross-linked thymidine (“X”). More specifically, in 7 the labeled dA (“Y”) containing strand is incised almost 10-fold more frequently than that containing the modified thymidine, “X” (). However, the “Y” strand in 7 is incised only ~2-fold more frequently than the “X” strand () when the complement is radiolabeled. In 6, the preference for incising the labeled “Y” containing strand is less than in 7. When the “Y” strand is labeled it is incised ~5-times more frequently than the “X” containing complement (), and only ~20% more frequently when the “X” strand is radiolabeled ().
One possible explanation for the above observations is that UvrC incised an ICL molecule twice, resulting in double strand break formation. Double strand cleavage was unambiguously identified via native gel analysis of the UvrABC reactions (). DSBs are introduced into ~1 in 4 incised DNA molecules (25–29%). We did not expect double strand cleavage to result from dissociation after the first round of incision and reassociation of the protein, because UvrABC was not expected to bind the ternary complex created following the initial incision (48
). Indeed, UvrABC does not cleave independently synthesized preincised substrate containing ICL 2
). Instead, we propose that double strand cleavage results from two rounds of incision prior to dissociation (, ). If so, when the “X” strand is labeled () products II, III, and IV are indistinguishable by denaturing PAGE, and appear as though the oligonucleotide containing the cross-linked thymine, the less frequently incised strand, was cleaved. Only product I indicates incision of the “Y” containing strand. In contrast, when the “Y” strand is radiolabeled () products I, III, and IV reflect cleavage of this strand when the reaction is analyzed by gel electrophoresis, and a higher ratio of cleavage on the “Y” strand versus the “X” strand will be observed. This hypothesis explains the plots of incision as a function of time (, and See Supporting Information
) provided that UvrABC incision on both sides of the lesion in a single strand is coupled and that the following assumptions are valid. The first is that the incision of the original ICL is faster than the second round of cleavage (k1
). This is consistent with the observation that double strand breaks are the minor products. The other assumption is that incision of the “Y” containing strand is always faster than the “X” strand (k1
Repetitive incision of ICL containing radiolabeled “X” strand.
Repetitive incision of ICL containing radiolabeled “Y” strand.
Overall, when the “X” containing strand is radiolabeled (), incision of this less favored strand following cleavage of the complementary one (product I) to produce III reduces UvrABC’s observed preference for the dA (“Y”) containing strand, because the second incision prevents one from detecting cleavage on the “Y” containing strand by denaturing PAGE. In contrast, radiolabeling the preferentially incised “Y” strand () yields a larger observed preference for this strand because assuming that incision of this strand is still favored (k4 > k3) in the initially formed products (I and II), a second cleavage round increases the observed amount of “Y” incised (due to formation of IV) at the expense of “X” cleavage (attributable to II).
Quantification of the UvrABC incision in 7 () by nondenaturing PAGE corroborates this scenario. Analysis of cleaved 7 under denaturing conditions in which the “X” containing strand is radiolabeled revealed 19% cleavage of this strand, and 33% cleavage of the “Y” strand, which produces 10 (). When the sample was not denatured prior to loading, the native gel revealed 15% DSBs. Comparable analysis of 7 at the same concentration (2 nM) in which the “Y” containing strand was radiolabeled () indicated 47% incision of this strand, 5% cleavage of the unlabeled “X” containing strand (to yield the product comparable to 10), and 13% DSBs. It is gratifying that the differences (14%) in the amounts of “X” strand incision observed when it is labeled (19%) and unlabeled (5%) is approximately equal to the amount of DSBs observed. In addition, comparison of the differences (14%) in “Y” strand cleavage when this strand is labeled (47%) to unlabeled (33%) to the DSB level corroborates this observation.
Interstrand cross-link repair is vitally important because these lesions are absolute blocks to replication and transcription. Double strand breaks are even more significant because they are the most cytotoxic of lesions. Double strand cleavage could result from nucleotide excision repair of interstrand cross-links, but is extremely rare. For instance, double strand breaks are not formed when DNA cross-linked by N-methylmitomycin or psoralen are treated with UvrABC (14
). Recently, we described the first chemical characterization of double strand cleavage resulting from UvrABC repair of a cross-link (49
). However, in that instance the cross-link was adjacent to a nick. Double strand cleavage only required UvrABC to incise the strand opposite the nick. To our knowledge, UvrABC mediated double strand cleavage of 6
is the first example in which both strands of the interstrand cross-linked DNA are incised through dual incisions of phosphate diesters that flank the lesion on its 5′- and 3′-sides.