Damage recognition in nucleotide excision repair
Malta, E.
Citation
Malta, E. (2008, December 9). Damage recognition in nucleotide excision repair. Retrieved from https://hdl.handle.net/1887/13327
Version: Corrected Publisher’s Version
License: Licence agreement concerning inclusion of doctoral thesis in the Institutional Repository of the University of Leiden
Downloaded from: https://hdl.handle.net/1887/13327
Chapter 4
Functions of base flipping in E. coli nucleotide excision repair
Erik Malta, Carlo P. Verhagen, Geri F. Moolenaar, Dmitri V. Filippov, Gijs A. van der Marel and Nora Goosen
DNA repair, 2008, 7, 1647-1658
Abstract
UvrB is the main damage recognition protein in bacterial nucleotide excision repair and is capable of recognizing various structurally unrelated types of damage. Previously we have shown that upon binding of E.coli UvrB to damaged DNA two nucleotides become extrahelical: the nucleotide directly 3’ to the lesion and its base-pairing partner in the non- damaged strand. Here we demonstrate using a novel fluorescent 2-aminopurine-menthol modification that the position of the damaged nucleotide itself does not change upon UvrB binding. A co-crystal structure of B. caldotenax UvrB and DNA has revealed that one nucleotide is flipped out of the DNA helix into a pocket of the UvrB protein where it stacks on Phe249 [Truglio et al. Nat. Struct. Mol. Biol. 13 (2006) 360-364]. By mutating the equivalent of Phe249 (Tyr249) in the E.coli UvrB protein we show that on damaged DNA neither of the extrahelical nucleotides is inserted into this protein pocket. The mutant UvrB protein, however, resulted in an increased binding and incision of undamaged DNA showing that insertion of a base into the nucleotide-binding pocket is important for dissociation of UvrB from undamaged sites. Replacing the nucleotides in the non-damaged strand with a C3- linker revealed that the extruded base in the non-damaged strand is not directly involved in UvrB binding or UvrC-mediated incision, but that its displacement is needed to allow access for residues of UvrB or UvrC to the neighboring base, which is directly opposite the DNA damage. This interaction is shown to be essential for optimal 3’ incision by UvrC. After 3’
incision base flipping in the non-damaged DNA strand is lost, indicative for a conformational change needed to prepare the UvrB-DNA complex for 5’ incision.
Introduction
Nucleotide excision repair is a highly conserved DNA-repair mechanism that is characterized by the unique feature that it recognizes a vast array of unrelated types of damage.
If left unrepaired these damages might lead to mutations or cell death due to interference with vital cellular processes such as replication and transcription. Damage recognition and DNA incision in prokaryotes is a complex process that is facilitated by only three different proteins: UvrA, UvrB and UvrC (reviewed in Van Houten et al. 2005, Truglio et al. 2006a).
DNA repair by UvrABC takes place in a multistep, ATP-dependent reaction that starts with the formation of the UvrA2B2-complex in solution (Malta et al. 2007). This complex binds to the DNA and wraps the DNA around one of the UvrB subunits (Verhoeven et al. 2002a).
UvrA serves as the first damage recognition factor and is required for loading of UvrB onto a potential damaged site. Subsequently damage determinants of UvrB probe the DNA for the presence of a lesion. The crystal structures of the UvrB protein from Bacillus caldotenax (Theis et al. 1999) and Thermus thermophilus (Machius et al. 1999; Nakagawa et al. 1999) show a fold of the UvrB protein similar to that of the helicases NS3 (Kim et al. 1998), PcrA (Velankar et al. 1999) and Rep (Korolev et al. 1997). The UvrB protein consists of five
domains, which are referred to as 1a, 1b, 2, 3, and 4 (Fig. 8A). In addition UvrB has a flexible β-hairpin containing several highly conserved hydrophobic residues, which is not present in the aforementioned helicase family members. Mutational studies showed the importance of this β-hairpin and of its hydrophobic residues in damage recognition (Moolenaar et al.
2001; Skorvaga et al. 2002; Moolenaar et al. 2005). In particular residues Tyr92 and Tyr93 prevent stable binding of UvrB to undamaged DNA because they form a sterical hindrance when UvrB tries to bind to a non-damaged site (Moolenaar et al. 2001). Once a lesion is localized the preincision complex is formed consisting of a dimer of UvrB stably bound to the lesion with the DNA wrapped around one of the UvrB subunits (Verhoeven et al. 2001).
In this complex the C-terminal domain of UvrB is readily exposed and can interact with a homologous domain in UvrC (Malta et al. 2007). The resulting UvrBC-complex is capable of performing dual incision, first on the 3’ side and next on the 5’ side of the lesion (Verhoeven et al. 2000). In E. coli and several other bacterial species a second NER-specific nuclease is present, Cho, which can induce incision on the 3’ side of the lesion only (Moolenaar et al.
2002).
Previously we have shown by 2-aminopurine (2AP) fluorescence measurements that binding of UvrB to a cholesterol damage moves the base 3’ adjacent to the lesion into an extrahelical position where it is shielded from the solvent by residues of the UvrB protein (Malta et al. 2006). Also the opposite base in the non-damaged strand is flipped out of the DNA helix mediated by Tyr95, which is located at the bottom of the β-hairpin structure. This base however becomes solvent exposed. Evidence was presented that this conformational change in the non-damaged strand is important for efficient 3’ incision by UvrC (Malta et al. 2006).
Recently two crystal structures have been solved of the UvrB protein bound to two different DNA substrates containing a damaged nucleotide. Truglio et al (2006b) reported a structure of UvrB bound to a small DNA loop structure in which it was not the damage itself that was recognized by the UvrB protein but a single-strand double-strand DNA junction. This structure showed for the first time that one of the DNA strands threads behind the β-hairpin structure and that the other DNA strand passes in front of it. Whether it is the damaged DNA strand that is wedged between the β-hairpin structure and domain 1b of the UvrB protein could however not be determined in this study. The co-crystal structure also showed that one of the nucleotides that passes behind the β-hairpin is flipped into a hydrophobic binding pocket of the protein with residue Tyr96 of the β-hairpin filling up the vacated space in the DNA helix (Fig. 8B). This nucleotide binding pocket encompasses residues Phe249, Ile306 and Leu313 of which the phenylalanine stacks with the extruded nucleotide.
Waters et al obtained a crystal structure of UvrB bound to a pentanucleotide containing a central fluorescein adduct (Waters et al. 2006). In this structure two nucleotides at the 3’ end of this pentanucleotide are located behind the β-hairpin and the adducted nucleotide is located at the entrance of the β-hairpin/domain 1b cavity (Fig. 8C,E). In this
structure, however, no base insertion into a protein pocket was observed.
In this paper we investigate the function of base flipping in damage specific binding.
We show that in the UvrB-DNA complex the base adjacent to the damage is not located in the pocket that was seen in the structure of Truglio et al (2006b). Instead we present evidence that the insertion of a nucleotide in this pocket plays a role in the dissociation of UvrB from undamaged DNA.
Materials and methods
Bacterial strains and plasmids
All strains used in this study (CS 5017 (∆5017 (∆uvrB), CS 5018 (∆uvrA, ∆uvrB), and CS 5639 (∆uvrB, ∆uvrC, ∆cho)) have been described (Moolenaar et al. 2005). The plasmids expressing the wildtype UvrB protein and the Y95A mutant have been described (Malta et al. 2006). The plasmid expressing the UvrB mutant (Y249A) was constructed using PCR- mediated site-directed mutagenesis, and the construct was verified by DNA sequencing. The plasmid used to produce additional UvrA in the cell (pNP120) is a pSC101 derivative and has been described before (Moolenaar et al. 2000a).
Proteins, chemicals and DNA substrates
The UvrA (Visse et al. 1992), UvrB (Moolenaar et al. 2001), UvrC (Visse et al.
1992) and Cho (Moolenaar et al. 2002) proteins were purified as described. Creatine kinase (CK), Creatine phosphate (CP) and ATP (containing <0.5% ADP) were obtained from Roche. The 50-mer DNA substrates containing a cholesterol modification (CholP) were obtained commercially (Eurogentec). The two different sequences that were used with the lesion flanked by three A residues (sA) or three T residues (sT) were described previously (Malta et al. 2006). The menthol modification connected to the N3-position of a thymine residue (Verhoeven et al. 2002b) was incorporated in the same sequences as the CholP modification. C3-linker modifications correspond to incorporation of a 1,3-propanediol moiety (Eurogentec) thereby removing both the base and ribose unit of a nucleotide while keeping the DNA backbone intact. Synthesis of the DNA modification containing a menthol group attached to the C6-position of a 2-aminopurine residue will be described elsewhere.
This modified nucleotide was also incorporated in the same sequence.
Incision Assay
The DNA substrates were 5’ labeled using polynucleotide kinase as described (Verhoeven et al. 2002b). The DNA substrates (0.2 nM) were incubated with 2.5 nM UvrA, 100 nM UvrB, and 25 nM UvrC/Cho in 20 μl Uvr-endo buffer (50 mM Tris-HCl pH 7.5, 10 mM MgCl2, 100 mM KCl, 0.1 μg/μl BSA, and 1 mM ATP). After an incubation of 30 minutes at 37 oC the reactions were terminated using 3 μl EDTA/SDS (0.33 M EDTA, 3.3%
SDS) and 2.4 μl glycogen (4 μg/μl) followed by ethanol precipitation. The incision products were visualized on a 15% denaturing gel. For the incision on supercoiled DNA unirradiated pNP228 (50 ng) and irradiated pUC18 (50 ng, irradiation at 300 J/m2) were used. DNA was incubated with 50 nM UvrA, 100 nM UvrB and 25 nM UvrC in 10 μl Uvr-endo buffer for 20 minutes at 37 oC. The incision reaction was stopped by addition of 3 μl gel loading buffer (10% Ficoll, 0.25% bromophenol blue, 0.25% xylene cyanol) containing 3.3% SDS and 20 mM EDTA. The samples were run on a 0.7% agarose gel in 1x Tris-borate buffer. DNA was visualized by ethidiumbromide staining.
Gel Retardation Assay
The 5’ terminally labeled DNA substrates (0.2 nM) were incubated with 1.25 nM UvrA and 100 nM UvrB in 10 μl Uvr-endo buffer for 10 minutes at 37 oC. The resulting samples were analyzed on a cooled 3.5% native polyacrylamide gel containing 1 mM of ATP and 10 mM MgCl2 in 1x Tris-borate/EDTA as described (Visse et al. 1992). To determine complex formation under the conditions used for the 2-aminopurine fluorescence measurements 0.2 nM of 5’ terminally labeled DNA substrate was mixed with 0.5 μM of the same unlabeled DNA substrate. The DNA was incubated in Uvr-endo buffer (without BSA) for 10 minutes at 37oC in the presence of the proteins (0.45 μM UvrA, 3.75 μM UvrB, and 0.17 μg/μl CK) and 20 mM CP. After incubation the samples were loaded on a 3.5% native polyacrylamide gel as described above.
2-Aminopurine fluorescence measurements
60 μl samples containing 0.5 μM DNA, 0.45 μM UvrA, 3.75 μM UvrB and 0.17 μg/
μl CK were incubated for 10 minutes at 37oC in Uvr-endo buffer (without BSA) containing 20 mM CP. Where indicated 300 mM acrylamide was included in the reaction mixture after incubation. After incubation spectra were obtained as described (Malta et al. 2006).
Filter binding assay
The samples for the filter binding assays (60 μl) were incubated using the same conditions as for the 2-aminopurine fluorescence measurements. After incubation the mixture was applied to a nitrocellulose filter, the incubation vial was rinsed with 100 μl Uvr-endo buffer (without BSA) and finally the filter was washed two times with 300 μl of the same buffer. Each sample was corrected for the amount of DNA bound to the filter in the absence of the NER proteins. Binding is indicated as the percentage of total input DNA retained on the filter.
Results
The conformation of the lesion does not change upon UvrB binding
We have previously shown by 2AP fluorescence measurements that binding of UvrB to a 50-mer dsDNA substrate containing a cholesterol lesion results in a DNA conformation in which the nucleotide 3’ to the lesion and its base-pairing partner in the non-damaged strand are both in an extrahelical position (Malta et al. 2006). However, this experimental setup did not enable us to investigate possible conformational changes of the damaged nucleotide itself. For this reason a novel DNA modification was synthesized consisting of a menthol group covalently attached to the C6 position of 2-aminopurine (Fig. 1A). This moiety enabled us to monitor potential flipping of the damaged nucleotide itself by looking at 2-aminopurine fluorescence. To exclude sequence specific effects on the fluorescence signal the modification was incorporated in two different DNA sequences, in which the lesion is flanked either by three adenines (sequence sA) or three thymines (sequence sT) on each side.
Fig. 1 Analysis of DNA substrates containing a menthol derivative of the fluorescent base 2-aminopurine. (A) Structure of the 2AP- menthol modification. Synthesis of this compound will be described elsewhere. (B) Gel-retardation analysis of protein-DNA complexes on DNA substrates containing a fluorescent 2AP-menthol modification (AP-M). 0.5 μM of 50-mer DNA substrate mixed with tracer of 5’ radioactively labeled DNA of the same substrate was incubated (10 min 37 oC) with 0.45 μM UvrA and 3.75 μM UvrB in the presence of the CP/
CK system. After incubation samples were analyzed on a 3.5% native polyacrylamide gel containing 1 mM ATP. The sequences of the two DNA substrates (sA and sT) were described previously (Malta et al. 2006), and are indicated. Also the positions of the co-migrating UvrA-DNA and UvrAB-DNA complexes (A(B)-DNA) and of the dimeric UvrB complex (B2-DNA) are indicated. (C) 2AP fluorescence spectra on substrates containing a 2AP-menthol modification in two different DNA sequences (sA and sT) as indicated. DNA (0.5 μM) was incubated for 10 minutes at 37 oC with (red line) or without (black line) 0.45 μM UvrA and 3.75 μM UvrB in the presence of the CP/CK system. After incubation the samples were transferred to a cuvette and emission spectra were recorded at 37 oC (excitation at 310 nm).
Using the conditions of the 2-aminopurine measurements UvrB-DNA complexes are formed on both DNA substrates (Fig. 1B) and these complexes give rise to efficient incision by UvrC (results not shown). Fluorescence of the two DNA substrates in the absence of proteins, 68 and 46 AU for sequence sA and sT respectively (Fig. 1C), is greatly reduced compared to the signal of the 2AP free in solution (234 AU) as was shown before (Malta
et al. 2006), indicating that the damaged base is inserted within the DNA helix. However, fluorescence of these DNA substrates is higher compared to an undamaged DNA substrate containing a 2AP residue, which was shown to be about 6 AU (Malta et al. 2006). This implies that the adducted base, due to the presence of the menthol modification, does not perfectly stack onto its neighbors. When both UvrA and UvrB are included in the reaction, allowing formation of UvrB-DNA complexes fluorescence remains unaltered (Fig. 1C). This clearly shows that the damaged nucleotide itself does not change position upon binding of UvrB.
Base flipping of the nucleotide in the undamaged strand is independent of the type of lesion In the previous study we made use of a DNA substrate containing a cholesterol moiety directly connected to the phosphate backbone, thereby missing both the sugar and base moieties. To investigate whether flipping also occurs on a substrate containing an adduct attached to the base of a nucleotide, fluorescence was measured on a DNA substrate containing a menthol modification connected to the N3-position of a thymine residue (Verhoeven et al.
2002b).
Fig. 2 Analysis of DNA substrates containing either a CholP lesion (Ch) or a menthol modification connected to the N3 position of a thymine residue (M). (A) DNA substrates (0.5 μM) containing either a menthol (M) or CholP (Ch) modification were incubated for 10 minutes at 37 oC with or without 0.45 μM UvrA and 3.75 μM UvrB in the presence of the CP/CK system. After incubation the samples were transferred to a cuvette and emission spectra were recorded at 37 oC (excitation at 310 nm). Contents of the samples used are indicated. In green is the spectrum recorded in the presence of acrylamide. All substrates contained either a 2AP 3’ adjacent to the lesion (left panel) or directly opposite this position in the non-damaged DNA strand (right panel). (B) Analysis of complex formation on 50-mer DNA substrates containing a CholP (Ch) or a menthol (M) modification in two DNA sequences (sA and sT). 0.5 μM of these 50-mer DNA substrates mixed with tracer of 5’ radioactively labeled DNA of the same substrate were incubated (10 min 37 oC) with 0.45 μM UvrA and 3.75 μM UvrB in the presence of the CP/CK system. After incubation samples were analyzed on a 3.5% native polyacrylamide gel containing 1 mM ATP. Positions of the co-migrating UvrA-DNA and UvrAB-DNA complexes (A(B)-DNA) and of the dimeric UvrB complex (B2-DNA) are indicated.
For the 2AP in the non-damaged DNA strand fluorescence in the presence of both UvrA and UvrB for the substrate containing the menthol lesion is identical to that of the cholesterol-containing DNA substrate (Fig. 2A, right graph). In addition, this signal can be
quenched by the addition of acrylamide to the same extent as was the case for the cholesterol containing DNA substrate (Malta et al. 2006). Since quenching by acrylamide can only occur for solvent-exposed residues this means that also for the substrate containing a menthol lesion the nucleotide in the non-damaged strand is extruded from the DNA. Also the amount of UvrB-DNA complexes under these conditions is equal (Fig. 2B, sequence sT). This shows that positioning of the nucleotide in the non-damaged strand is independent of the type of lesion to which UvrB is bound.
Flipping of the base adjacent to the lesion is dependent on the type of lesion
Next we tested fluorescence of a 2AP located directly 3’ to the menthol lesion.
Upon formation of the UvrB-DNA complex fluorescence increases compared to unbound dsDNA and ssDNA, indicating that also for the menthol lesion the 3’ nucleotide is moved in an extra-helical position (Fig. 2A, left graph). The increase in fluorescence signal for the menthol lesion however is significantly less compared to that of the cholesterol lesion. This cannot be ascribed to a quantitative difference in UvrB-DNA complexes, since binding of UvrB is shown to be the same for both DNA substrates (Fig. 2B, sequence sA). Apparently the position of the nucleotide 3’ to the lesion in the UvrB-DNA complex is different for both lesion types.
Acrylamide quenching experiments with the menthol-containing DNA substrate reveal, similar to what was found for the cholesterol-containing DNA substrate, that fluorescence is unaltered upon acrylamide addition (Fig. 2A). This shows that for both types of damage the nucleotide 3’ to the lesion is protected from the solution.
Fig. 3 Analysis of the E. coli Tyr249A mutant. (A) Incision of DNA containing a CholP modification by UvrC (lane 1 and 2) or Cho (lane 3 and 4) in the presence of wtUvrB or mutant Y249A as indicated. The 5’
terminally labeled DNA substrates were incubated with 2.5 nM UvrA, 100 nM UvrB and 25 nM UvrC/
Cho at 37 oC for 30 minutes. The incision products were analyzed on 15% denaturing polyacrylamide gels and are indicated. (B) Fluorescence spectra comparing wildtype UvrB (black line) with UvrB Y249A (red line) on a substrate containing a cholesterol modification. The DNA (0.5 μM) was incubated for 10 minutes at 37
oC with 0.45 μM UvrA and 3.75 μM (mutant) UvrB in the presence of the CP/CK system. After incubation the samples were transferred to a cuvette and emission spectra were recorded at 37 oC (excitation at 310 nm). Substrates contained either a 2AP modification in the damaged (left panel) or non-damaged strand (right panel).
Tyr249 does not interact with the extrahelical nucleotides
The co-crystal structure of Truglio et al (2006b) showed that one nucleotide of the DNA strand that passes behind the β-hairpin is flipped into a pocket of the Bacillus caldotenax UvrB protein where it stacks on Phe249 (Fig. 8B, D). To investigate the importance of this pocket for the base flipping observed in the fluorescence measurements we mutated the E.
coli residue Tyr249 (analogous to the conserved hydrophobic residue Phe249) to alanine.
No significant difference in incision by either UvrC or Cho compared to the wildtype UvrB protein could be discerned (Fig. 3A), showing that Tyr249 is not essential for repair. The same result was also found for the corresponding F249A mutant in B. caldotenax (Skorvaga et al. 2004).
Next we tested the Y249A mutant for an effect on base flipping. The substrate containing the 2AP adjacent to the cholesterol lesion (sA) exhibits a higher fluorescence signal than wildtype UvrB (Fig. 3B, left graph). A quantitative analysis of the amount of complexes by filter binding reveals that complex formation of the Y249A mutant (73.1 ± 1.4%) is increased by ~30% as compared to the wildtype UvrB protein (54.7 ± 4.7%). This is similar to the extent of increase observed for the fluorescence signal. Apparently, mutating the conserved hydrophobic residue of the pocket of the UvrB protein somehow enhances DNA binding but does not seem to affect base flipping 3’ to the lesion.
For the substrate containing the 2AP in the non-damaged DNA strand DNA-binding is only slightly enhanced for the Y249A mutant (65.7 ± 1.5%) as compared to wildtype UvrB (61.2 ± 2.3%). It should be noted however that binding of wildtype UvrB to sT (61.2 ± 2.3%) is already higher than binding to sA (54.7 ± 4.7%). This became also clear from bandshift analysis (Fig. 2B). When we measured the fluorescence of Y249A on the DNA substrate with the 2AP in the non-damaged strand it was found that fluorescence is unaltered for the mutant UvrB as compared to wildtype (Fig. 3B, right graph).
Fig. 4 Analysis of the importance of the nucleotide directly 3’ to the CholP lesion. (A) Gel-retardation analysis of a substrate either with or without a C3-linker directly 3’ to the CholP modification. The 50-mer 5’ radioactively labeled DNA substrates were incubated with 1.25 μM UvrA and 100 nM UvrB wildtype (wt) or mutant Y249A (Y249) for 10 min at 37 oC. After incubation samples were analyzed on a 3.5% native polyacrylamide gel containing 1 mM ATP. The DNA substrates used are indicated above the panel where Δ indicates the CholP lesion and
• indicates the position of a C3-linker. Positions of the co-migrating UvrA-DNA and UvrAB-DNA complexes (A(B)-DNA) and of the UvrB complex (B-DNA) are indicated. (B) UvrABC or UvrABCho incision on a substrate without (lanes 1 and 2) or with (lanes 3 and 4) a C3-linker 3’
adjacent to the lesion. The 5’ terminally labeled DNA substrates were incubated with 2.5 nM UvrA, 100 nM UvrB and 25 nM UvrC/Cho at 37 oC for 30 minutes. The incision products were analyzed on 15%
denaturing polyacrylamide gels and are indicated.
To test the importance of the extrahelical nucleotide adjacent to the damage we constructed a DNA substrate lacking the nucleotide directly 3’ to the lesion by replacing this nucleotide by a C3-linker. Gel retardation analysis with this DNA substrate showed a significant reduction in the amount of UvrB-DNA complexes, both for wildtype UvrB and UvrB(Y249A) as compared to the intact 50-mer DNA substrate (Fig. 4A). Incision efficiencies on the other hand remain unaltered for both UvrC and Cho (Fig. 4B), showing that UvrB does bind to the substrate lacking the 3’ nucleotide but that the resulting UvrB- DNA complex is less stable. The same results were found when an abasic site was introduced 3’ to the lesion (results not shown). Apparently the presence of an extrahelical nucleotide 3’
to the lesion is not a prerequisite for damage recognition and subsequent incision, but only confers extra stability to the preincision complex.
In summary, we have observed that mutating Tyr249 to alanine results in an increase in UvrB-DNA complex formation whereas removal of the nucleotide directly 3’ to the lesion renders the complexes less stable. These results combined with the fluorescence experiments clearly show that in the complex of UvrB bound to damaged DNA the nucleotide 3’ to the lesion does not stack on Tyr249 and is therefore not expected to be inserted into the pocket that was observed in the co-crystal of Truglio et al (2006b).
Table 1 Lethality of the UvrB(Y249A) mutant
UvrBwt UvrBY249A
∆B ++ ±
∆AB ++ ++
∆BCCho ++ ++
∆B + pNP120 ++ -
∆BCCho + pNP120 ++ -
Plasmids expressing the wild type or mutant UvrB protein were introduced in different bacterial strains as indicated. pNP120 contains the uvrA gene on a pSC101 derivative. ++ normal size colonies; ± very small colonies;
- no colonies.
The UvrB nucleotide-binding pocket prevents binding to undamaged DNA
During the construction of the Y249A mutant we noticed that cells expressing the mutant UvrB protein gave rise to much smaller colonies than cells expressing the wildtype
UvrB protein. To investigate this further we introduced the plasmid expressing the mutant protein in different genetic backgrounds (Table 1). In a strain lacking UvrA or the UvrC and Cho proteins the plasmid did not cause any problem in bacterial growth. Upon introduction of a plasmid expressing an elevated level of UvrA (pNP120) however, no bacterial growth could be observed at all. Even in a background lacking UvrC and Cho the mutant UvrB protein turned out to be lethal, most likely as a result of a too strong binding to undamaged DNA. UvrA-mediated binding of mutant Y249A to undamaged DNA also becomes apparent from incision of supercoiled DNA substrates (Fig. 5). In a mixture of UV-irradiated and unirradiated plasmid DNA the wild type UvrABC proteins mainly incise the damaged DNA (lane 2). In the presence of Y249A, however, also the undamaged plasmid is significantly incised (lane 3), again showing that the mutant UvrB has a higher affinity for undamaged DNA than the wild type protein. This phenotype of Y249A closely resembles that of the Y92A and Y93A mutants (Moolenaar et al. 2001, 2005) indicating that not only these aromatic residues of the hairpin but also the protein-pocket of UvrB is involved in preventing binding of the protein to undamaged DNA.
Fig. 5 Incision of (UV-irradiated) supercoiled DNA in the presence of (mutant) UvrABC. A mixture of UV-irradiated pUC18 (2700 bp, I) and unirradiated pNP228 (4800 bp, II) supercoiled DNA was incubated at 37
°C for 15 minutes with 50 nM UvrA, 25 nM UvrC with or without 100 nM wt UvrB or UvrB(Y249A) as indicated. The positions of the supercoiled (sc) and open circle (oc) forms of the plasmids are shown.
Function of flipping in the undamaged DNA strand in incision by UvrC
Previously we have shown that Tyr95 is required for base flipping in the undamaged DNA strand and is important for efficient incision by UvrC, but not by Cho (Malta et al.
2006). From this result it was postulated that the extrahelical nucleotide might function as a target for UvrC binding. To test this hypothesis we replaced the nucleotide from the non- damaged DNA strand that becomes extrahelical in the UvrB-DNA complex with a C3-linker.
Removal of this nucleotide did not have any influence on UvrB-DNA complex formation (Fig. 6A) or on incision by Cho (Fig. 6B, lanes 5 and 7). Surprisingly however, it also did not affect the UvrC-mediated incision (Fig. 6B, lanes 1 and 3). Moreover, the UvrC incisions of complexes formed on this substrate by UvrB(Y95A) or wt UvrB are now equal (Fig. 6B lanes 3 and 4) which means that residue Tyr95 is no longer required when the nucleotide in the non-damaged strand is missing. Apparently the extrahelical nucleotide in the non-damaged
DNA strand does not serve as a target for UvrC recruitment but flipping is required to create space inside the DNA helix. When in a similar way the nucleotide opposite the damage was removed by replacing it with a C3-linker incision by UvrC did dramatically decrease (Fig.
6C, lane 1). Also for this substrate the UvrB-DNA complex formation and Cho-mediated incision are unaffected (Fig. 6A, 6C lane 2). Taken together this means that for efficient UvrC incision to occur the nucleotide opposite the damage needs to be present whereas its adjacent neighbor has to be extruded from the helix by Tyr95. This strongly suggests that the Tyr95-mediated base flipping in the non-damaged DNA strand serves to enable residue(s) of UvrB or UvrC to stack on the nucleotide opposite the lesion. This stacking interaction in its turn stabilizes a specific conformation of the UvrBC-DNA complex required for 3’
incision.
Fig. 6 Analysis of the importance of the nucleotides opposite the CholP lesion. (A) Complex formation of the wild type UvrA and UvrB proteins on 50- mer substrates containing a cholesterol lesion (Δ) with or without a C3-linker (•) in the non- damaged strand. The 50-mer 5’ radioactively labeled DNA substrates were incubated with 1.25 μM UvrA and 100 nM UvrB for 10 min at 37 oC. After incubation samples were analyzed on a 3.5% native polyacrylamide gel containing 1 mM ATP. Positions of the co-migrating UvrA- DNA and UvrAB-DNA complexes (A(B)-DNA) and of the UvrB complex (B-DNA) are indicated.
(B) UvrABC (lanes 1-4) or UvrABCho (lanes 5-8) incision on substrates with or without a C3-linker 5’ adjacent to the nucleotide opposite the lesion as indicated with (•). The 5’ terminally labeled DNA substrates were incubated with 2.5 nM UvrA, 100 nM UvrB and 25 nM UvrC/Cho at 37 oC for 30 minutes. The incision products were analyzed on 15% denaturing polyacrylamide gels and are indicated. Wildtype UvrB (lanes 1, 3, 5, and 7) or UvrB mutant Y95A (2, 4, 6, and 8) are used as indicated. (C) UvrABC (lane 1) or UvrABCho (lane 2) incision on a substrate containing a C3-linker directly opposite the lesion in the non-damaged DNA strand. The linker is indicated as before. Same conditions as for (B) were used.
The nucleotide in the non-damaged DNA strand is no longer extruded after 3’ incision To see whether base flipping still occurs after 3’ incision has taken place we made use of a 3’ pre-nicked DNA substrate containing a 2AP residue in either the damaged (sA) or non-damaged (sT) DNA strand. Bandshift analysis on these DNA substrates showed that the UvrB-DNA complex is unstable on the sA substrate, while stable on the sT substrate (Fig. 7A, lanes 2 and 5). Apparently, for unknown reasons, the sequence context surrounding the lesion has a large effect on UvrB-DNA complex stability on these pre-nicked DNA substrates.
If instead of a 3’ nick- a 3’ gap-containing DNA substrate (which lacks the top
strand beyond the 3’ incision position, see Fig. 7A) was used it was found that the resulting UvrB-DNA complex is very stable on both DNA sequences (Fig. 7A, lanes 3 and 6). Like for the 3’ nick containing DNA substrates a 3’ gap mimics a situation in which 3’ incision has already taken place since very efficient damage-dependent 5’ incision has been shown to occur on such DNA substrates (Moolenaar et al. 2000b).
Fig. 7 Analysis of base flipping after 3’ incision. (A) Complex formation of the wild type UvrA and UvrB proteins on a 50-mer double stranded DNA substrate containing a cholesterol lesion (lanes 1 an 4) or the same substrates containing a 3’
nick (lanes 2 and 5) or a 3’ gap (lanes 3 and 6), studied for both sequences (sA and sT). 0.5 μM of the different DNA substrates mixed with tracer of 5’ radioactively labeled DNA of the same substrate were incubated (10 min 37 oC) with 0.45 μM UvrA and 3.75 μM UvrB in the presence of the CP/CK system. After incubation samples were analyzed on a 3.5%
native polyacrylamide gel containing 1 mM ATP. Positions of the co-migrating UvrA-DNA and UvrAB-DNA complexes (A(B)-DNA) and of the dimeric UvrB complex (B2-DNA) are indicated. (B) Fluorescence spectra comparing emission on a double-stranded DNA substrate (ds), a substrate containing a nick at the 3’ incision position (nick) or a substrate lacking the DNA of the damaged strand 3’ to this incision position (gap). The DNA (0.5 μM) was incubated for 10 minutes at 37 oC with or without 0.45 μM UvrA and 3.75 μM UvrB in the presence of the CP/CK system. After incubation the samples were transferred to a cuvette and emission spectra were recorded at 37 oC (excitation at 310 nm). Spectra were recorded of substrates containing a 2AP in the damaged (left panel) or non-damaged (right panel) DNA strand. (C) UvrABC incision on substrates containing a nick at the 3’ incision position of UvrC with or without a C3-linker in the non-damaged DNA strand. The 5’ terminally labeled DNA substrates were incubated with 2.5 nM UvrA, 100 nM UvrB and 25 nM UvrC/Cho at 37 oC for 30 minutes. The incision products were analyzed on 15% denaturing polyacrylamide gels and are indicated. The position of the C3- linker directly opposite the lesion (lane 2) or 5’ adjacent to this position (lane 3) is indicated with (•).
Fluorescence measurements on the 3’ nick DNA substrates containing a 2AP adjacent to the lesion (sA) show that upon addition of the proteins fluorescence is only slightly increased (Fig. 7B, left graph), which is in accordance with the low stability of the resulting UvrB-DNA complex. However, when UvrB is bound to the substrate containing a 3’ gap fluorescence is even slightly higher compared to that of complexes on the double strand DNA substrate. This is probably due to elevated complex formation on the gapped substrate (Fig. 7A, compare lanes 1 and 3). Addition of acrylamide again did not significantly alter the fluorescence signal indicating that the nucleotide 3’ to the lesion is still protected from solution by residues of UvrB (results not shown). In the non-damaged DNA strand
(sT) of the 3’ nick- or 3’ gap-containing DNA substrates on the other hand no increase in fluorescence upon addition of UvrA and UvrB is observed anymore (Fig. 7B, right graph).
Apparently, introduction of a nick 3’ to the damage results in loss of base flipping in the non- damaged DNA strand implying that it is only required for 3’ incision and not for 5’ incision.
Indeed removal of either of the two residues in the non-damaged DNA strand of a pre-nicked DNA substrate (i.e. the nucleotide opposite the lesion or the 5’ adjacent one) did not have a significant effect on 5’ incision by UvrC (Fig. 7C, lanes 1-3).
Fig. 8 Crystal structures of UvrB and UvrB-DNA complexes. (A) Crystal structure of the UvrB protein from Bacillus caldotenax (PDB entry 1D9Z). The different domains of the protein (domains 1a, 1b, 2, and 3) and the β-hairpin structure are indicated. The C-terminal domain (domain 4) is not visible. (B,D) Close-up views of the crystal structure of B. caldotenax UvrB bound to a small DNA hairpin with a 3’ overhang (PDB entry 2FDC). The positions of specific hairpin residues and Tyr249 are highlighted. C18 (yellow) corresponds to the nucleotide that is rotated behind the hairpin structure in a pocket of the UvrB protein. G17 (pink) is the postulated damage position.
(C,E) Close-up view of the crystal structure of Bacillus subtilis UvrB bound to a pentanucleotide ssDNA substrate
containing a central fluorescein adduct (PDB entry 2NMV). The positions of specific hairpin residues and Tyr249 are highlighted. T3 (pink) is the thymine residue containing the fluorescein lesion. T4 (yellow) is the residue at the 3’ side of the lesion, corresponding to the 2AP in the damaged strand of our constructs. T5 (green) is not completely visible since it is highly disordered. For further details see text.
Discussion
In this paper we have investigated the function of base flipping by UvrB in E.
coli nucleotide excision repair. Structural analysis of UvrB-DNA complexes indicated that upon binding of UvrB to double-stranded DNA the β-hairpin of the UvrB protein will insert between the two DNA strands (Truglio et al. 2006b; Waters et al. 2006). In the structure of UvrB bound to a DNA hairpin with a 3 nt 3’-overhang, the 3’-overhang threads behind the β-hairpin and the complementary DNA strand is predicted to pass in front of it (Fig. 8B).
However, since UvrB in this complex is not bound to a damaged site it is unclear which of the DNA strands represents the damaged one. Fluorescence measurements in both this and a previous study (Malta et al. 2006) have shown that Tyr95 is directly involved in the extrusion of a base from the non-damaged strand. Since this protein residue is solvent exposed in the co-crystal structure and is predicted to interact with the outer DNA strand (Fig. 8B) this shows that the non-damaged strand passes in front of the β-hairpin. This polarity is supported by the structure of Waters et al (2006), where it was shown that a single-stranded DNA substrate containing a fluorescein modification is partly located behind the β-hairpin (Fig.
8C). In this structure the adducted nucleotide is located at the entrance to the β-hairpin/domain 1b cavity and it was proposed that damage recognition is achieved by sterically excluding lesions from the β-hairpin/domain 1b interface (Waters et al. 2006). Additional support for the orientation of the damaged strand comes from DNaseI footprinting studies. Binding of UvrB to a damaged site results in the appearance of a clear, ATP-dependent hypersensitive site centered around the 11th phosphodiester bond 5’ to the lesion (Van Houten et al. 1987;
Moolenaar et al. 2005). Such a site is indicative of a local widening of the minor groove.
From structural similarity between domains 1a and 3 of UvrB with domains of helicases PcrA and NS3 it has been proposed that UvrB undergoes motions of these domains driven by ATP hydrolysis (Theis et al. 1999). If the damaged strand is located behind the β-hairpin the 5’ DNaseI hypersensitive site can indeed be aligned with the interface between domains 1a and 3.
In a previous study it has been shown that upon binding of UvrB to the damaged site the nucleotide directly 3’ to the lesion is flipped out of the DNA helix (Malta et al. 2006).
Here we show using a menthol modification covalently attached to a 2-aminopurine residue that the damaged base itself does not change position upon binding of UvrB but retains its original intra-helical conformation. Although these experiments could not determine the localization of the menthol adduct itself, its hydrophobic nature suggests that this group also remains buried inside the DNA double helix. Taken together we can conclude that flipping
in the damaged strand is limited to the non-damaged nucleotide 3’ to the lesion. Acrylamide quenching experiments revealed that this extrahelical nucleotide is not solvent-exposed, suggesting that it is situated in a pocket of the UvrB protein. Also in the structure of Truglio et al one of the nucleotides (C18) is rotated behind the β-hairpin and located in a pocket of UvrB where it stacks on Phe249 (Fig. 8B). To study if this extrahelical C18 (Fig. 8B) represents the flipped-out nucleotide that we observed by 2AP fluorescence we mutated the residue stacking on the extrahelical nucleotide (Tyr249 in E. coli). Mutation of this residue did however not alter the fluorescence signal of the 2AP as would be expected if the 2AP base and Tyr249 would interact. Furthermore, we show that removal of the nucleotide 3’ to the lesion severely lowers complex stability whereas mutation of Tyr249, which would stack with this nucleotide, results in more UvrB-DNA complexes. This makes it highly unlikely that our fluorescent 2AP is located inside the observed protein pocket. Our results are further supported by the structure of Waters et al (Fig. 8C) where none of the nucleotides can be seen to stack on Phe249 and where this pocket residue assumes a completely different position as compared to the cocrystal structure of Truglio et al (Fig. 8B). The reduction in complex stability upon removal of the nucleotide 3’ to the lesion is also supported by the structure of Waters et al since this nucleotide is seen to stack with Tyr96 (Fig. 8C), an interaction that is very likely to stabilize the UvrB-DNA complex .
In this paper we show that fluorescence of the 2AP in the damaged strand is dependent on the type of lesion. Possibly the nature of the lesion determines how far the damaged strand can be translocated behind the hairpin. The cholesterol lesion which lacks a nucleotide and contains a long flexible linker to which the cholesterol moiety is attached can be expected to translocate further behind the hairpin as compared to the menthol lesion, which is attached to a base. As a consequence the position of the adjacent 2AP with respect to Tyr96 and neighboring DNA bases might be somewhat different for both types of lesion, resulting in a differential quenching of the fluorescence signal by these residues.
The question that now arises is: what is then the relevance of the pocket present in the UvrB protein? We have seen that mutation of pocket residue Tyr249 results in increased binding of UvrB to undamaged DNA, indicating that insertion of a nucleotide into the protein pocket prevents binding to undamaged DNA. Previously we have shown that Tyr92 and Tyr93 prevent binding to undamaged DNA as well (Moolenaar et al. 2001, 2005). These residues form a sterical hindrance to the DNA when UvrB attempts to bind an undamaged site. In the structure of Truglio et al however such a sterical hindrance is not observed, but the nucleotide in the closest vicinity of these aromatic residues (A19) is solvent exposed because its base-pairing partner is absent (Fig. 8D). Possibly in duplex DNA base-pairing of adjacent bases at the 3’ side places A19 in such a position that it will clash with Tyr92 and/or Tyr93. Stacking of A18 inside the protein pocket would then fine-tune the positioning of the adjacent nucleotide with respect to Tyr92 and/or Tyr93, contributing to dissociation of UvrB from undamaged DNA. The clashing process may not directly lead to dissociation
of the complex but may result in the translocation of the DNA strand behind the hairpin by 1 nucleotide, possibly via triggering of the ATPase of UvrB. Translocation of the DNA behind the β-hairpin has also been proposed by Truglio et al (2006b). Multiple rounds of clashing and translocation would continue until topological constraints in the DNA result in the release of the DNA from UvrB. When clashing is reduced by the absence of the Tyr92/93 or Tyr249 residues translocation will be arrested or slowed down leading to an enhanced life-time of the protein on undamaged DNA, explaining the observed increased incision of undamaged DNA and the lethal phenotype of the mutant proteins. On damaged DNA the presence of the lesion might block translocation behind the hairpin in such a way that incorporation of the subsequent nucleotide into the pocket of UvrB is prevented. As a consequence the nucleotides at the 2nd and/or 3rd positions 3’ to the lesion will have more conformational freedom and under the influence of Tyr92 and Tyr93 they will be able to occupy a position where they will no longer clash with these aromatic residues. As a result the ATPase may no longer be induced leaving an arrested ATP-bound form of UvrB on the DNA, which is now prone to incision by UvrC. This model is supported by the occurrence of DEPC sensitive sites at the 2nd and 3rd nucleotide 3’ to the lesion when UvrB binds to a damaged site (Moolenaar et al.
2005). In the UvrB-DNA complex formed by a protein lacking Tyr92 and Tyr93 these DEPC sites disappear (Moolenaar et al. 2005) indicating that indeed one or both tyrosines alter the position of the nucleotides 2 or 3 positions downstream from the lesion.
In the UvrB-DNA complex formed on a substrate mimicking the DNA after 3’
incision the fluorescence signal of the 2AP directly 3’ to the lesion is not significantly altered. This not only means that the interactions with the damaged strand remain the same after 3’ incision, but also that in order to proceed with the 5’ incision no rearrangement of the damaged strand is required.
The Tyr95 mediated extrusion of the base in the non-damaged strand has previously been shown not to be important for damage-specific binding of UvrB, but for the subsequent step, 3’ incision by UvrC (Malta et al. 2006). Here we show that the extruded base does not form a target for UvrC binding, since removal of the nucleotide from the DNA substrate did not affect incision efficiency. Instead the neighboring base, which is directly opposite the damage appears to be essential for the incision event. Possibly a residue of either UvrB or UvrC needs to enter the DNA helix and interact with this nucleotide which is facilitated when the neighboring nucleotide is in an extrahelical position. This interaction in its turn stabilizes a specific conformation of the UvrBC-DNA complex required for 3’ incision. After 3’ incision, the UvrB-DNA complex undergoes a rearrangement resulting in a conformation where the nucleotide in the non-damaged DNA strand is no longer extruded. Most likely this rearrangement is a prerequisite for 5’ incision, thereby ensuring that 5’ incision can only occur after 3’ incision has taken place.
Acknowledgements
The authors thank Elizabeth Meulenbroek for technical assistance in constructing and purifying the Y249A mutant UvrB protein. This work was funded by the Leiden Institute of Chemistry and the Netherlands Organization for Scientific Research (NWO) by grant CW 700.52.706.
References
Kim, J.L., Morgenstern, K.A., Griffith, J.P., Dwyer, M.D., Thomson, J.A., Murcko, M.A., Lin, C. and Caron, P.R. (1998) Hepatitis C virus NS3 RNA helicase domain with a bound oligonucleotide: the crystal structure provides insights into the mode of unwinding. Structure, 6, 89-100.
Korolev, S., Hsieh, J., Gauss, G.H., Lohman, T.M. and Waksman, G. (1997) Major domain swiveling revealed by the crystal structures of complexes of E. coli Rep helicase bound to single-stranded DNA and ADP. Cell, 90, 635-647.
Machius, M., Henry, L., Palnitkar, M. and Deisenhofer, J. (1999) Crystal structure of the DNA nucleotide excision repair enzyme UvrB from Thermus thermophilus. Proc. Natl. Acad. Sci. USA, 96, 11717-11722.
Malta, E., Moolenaar, G.F. and Goosen, N. (2006) Base flipping in nucleotide excision repair. J. Biol. Chem., 281, 2184-2194.
Malta, E., Moolenaar, G.F. and Goosen, N. (2007) Dynamics of the UvrABC nucleotide excision repair proteins analyzed by fluorescence resonance energy transfer. Biochemistry, 46, 9080-9088.
Moolenaar, G.F., Moorman, C. and Goosen, N. (2000a) Role of the Escherichia coli nucleotide excision repair proteins in DNA replication. J. Bacteriol., 182, 5706-5714.
Moolenaar, G.F., Monaco, V., van der Marel, G.A., van Boom, J.H., Visse, R. and Goosen, N. (2000b) The effect of the DNA flanking the lesion on formation of the UvrB-DNA preincision complex. Mechanism for the UvrA-mediated loading of UvrB onto a DNA damaged site. J. Biol. Chem., 275, 8038-8043.8038-8043.
Moolenaar, G.F., Höglund, L. and Goosen, N. (2001) Clue to damage recognition by UvrB: residues in the beta- hairpin structure prevent binding to non-damaged DNA. EMBO J., 20, 6140-6149.
Moolenaar, G.F., van Rossum-Fikkert, S., van Kesteren, M. and Goosen, N. (2002) Cho, a second endonuclease involved in Escherichia coli nucleotide excision repair. Proc. Natl. Acad. Sci. USA, 99, 1467-1472.
Moolenaar, G.F., Schut, M. and Goosen, N. (2005) Binding of the UvrB dimer to non-damaged and damaged DNA: residues Y92 and Y93 influence the stability of both subunits. DNA repair, 4, 699-713.
Nakagawa, N., Sugahara, M., Masui, R., Kato, R., Fukuyama, K. and Kuramitsu, S. (1999) Crystal structure of Thermus thermophilus HB8 UvrB protein, a key enzyme of nucleotide excision repair. J. Biochem., 126, 986- 990.
Skorvaga, M., Theis, K., Mandavilli, B.S., Kisker, C. and Van Houten, B. (2002) The beta -hairpin motif of UvrB is essential for DNA binding, damage processing, and UvrC-mediated incisions. J. Biol. Chem., 277, 1553- 1559.
Skorvaga, M., DellaVecchia, M.J., Croteau, D.L., Theis, K., Truglio, J.J., Mandavilli, B.S., Kisker, C. and Van Houten, B. (2004) Identification of residues within UvrB that are important for efficient DNA binding and damage processing. J. Biol. Chem., 279, 51574-51580.
Theis, K., Chen, P.J., Skorvaga, M., Van Houten, B. and Kisker, C. (1999) Crystal structure of UvrB, a DNA helicase adapted for nucleotide excision repair. EMBO J., 18, 6899-6907.
Truglio, J.J., Croteau, D.L., Van Houten, B. and Kisker C. (2006a) Prokaryotic nucleotide excision repair: the UvrABC system. Chem. Rev., 106, 233-252.
Truglio, J.J., Karakas, E. Hau, B., Wang, H., DellaVecchia, M.J., Van Houten, B. and Kisker, C. (2006b)
Structural basis for DNA recognition and processing by UvrB. Nat. Struct. Mol. Biol., 13, 360-364.
Van Houten, B., Gamper, H., Sancar, A. and Hearst, J.E. (1987) DNase I footprint of ABC excinuclease.DNase I footprint of ABC excinuclease. J.
Biol. Chem., 262, 13180-13187.
Van Houten, B., Croteau, D.L., DellaVecchia, M.J., Wang, H. and Kisker, C. (2005) ‘Close-fitting sleeves’:
DNA damage recognition by the UvrABC nuclease system. Mutat. Res., 577, 92-117.
Velankar, S.S., Soultanas, P., Dillingham, M.S., Subramanya, H.S. and Wigley, D.B. (1999) Crystal structures of complexes of PcrA DNA helicase with a DNA substrate indicate an inchworm mechanism. Cell, 97, 75-84.
Verhoeven, E.E., van Kesteren, M., Moolenaar, G.F., Visse, R. and Goosen, N. (2000) Catalytic sites for 3’
and 5’ incision of Escherichia coli nucleotide excision repair are both located in UvrC. J. Biol. Chem., 275, 5120- 5123.
Verhoeven, E.E., Wyman, C., Moolenaar, G.F., Hoeijmakers, J.H. and Goosen, N. (2001) Architecture of nucleotide excision repair complexes: DNA is wrapped by UvrB before and after damage recognition. EMBO J., 20, 601-611.
Verhoeven, E.E., Wyman, C., Moolenaar, G.F. and Goosen, N. (2002a) The presence of two UvrB subunits in the UvrAB complex ensures damage detection in both DNA strands. EMBO J., 21, 4196-4205.
Verhoeven, E.E., van Kesteren, M., Turner, J.J., van der Marel, G.H., van Boom, J.H., Moolenaar, G.F. and Goosen, N. (2002b) The C-terminal region of Escherichia coli UvrC contributes to the flexibility of the UvrABC nucleotide excision repair system. Nucleic Acids Res., 30, 2492-2500.
Visse, R., de Ruijter, M., Moolenaar, G.F. and van de Putte, P. (1992) Analysis of UvrABC endonuclease reaction intermediates on cisplatin-damaged DNA using mobility shift gel electrophoresis. J. Biol. Chem., 267, 6736-6742.
Waters, T.R., Eryilmaz, J., Geddes, S. and Barrett, T.E. (2006) Damage detection by the UvrABC pathway:
crystal structure of UvrB bound to fluorescein-adducted DNA. FEBS Lett., 580, 6423-6427.
