DNA charge transport as a first step in coordinating the detection of lesions by repair proteins

We have previously demonstrated with BER proteins that the ability of a protein to perform DNA CT directly correlates to its redistribution in the vicinity of base lesions or mismatches that inhibit CT (23). We tested XPD redistribution promoted through DNA CT by preparing DNA strands containing a single C∶A mismatch, a modification that we know to inhibit CT (40), alongside DNA strands containing no mismatches. To distinguish the strands in microscopy experiments, we made mismatched strands > 1,200 base pairs longer than the matched strands. AFM was utilized to gather images of DNA and bound protein that could be further analyzed to examine the propensity of XPD to redistribute. ATP was not incorporated in AFM experiments, as XPD helicase activity should not be required for a DNA-mediated redistribution.

Long (3.8 kilobase) DNA duplexes containing a site-specific C∶A mismatch and short (2.2 and 1.6 kilobases) well matched duplexes of the same total sequence were constructed. To prepare these DNA samples, pUC19 plasmid was amplified with primers incorporating a 2′ O-methyl residue to yield two short duplexes containing 14-nucleotide single-strand overhangs (23, 41). Upon ligation of these two duplexes, 3.8 kbp strands were formed that contained only one C∶A mismatch in the middle of the strand. The ligation reaction was not taken to completion so as to have a mixture of well matched short and mismatched long strands for protein distribution studies. Although a C∶A mismatch effectively inhibits DNA CT (40), it is not a lesion that is preferentially bound by XPD. However, the 14-nucleotide overhangs generated with PCR are specific substrates for XPD helicase (K_d approximately 1 μM) (25, 42). We directly observed this preference in initial AFM experiments. Protein assignments were verified through analysis of their 3–4 nm heights in the images; without protein, features of this dimension were not observed and larger heights (> 7 nm) indicated salt precipitates or protein aggregation. Only clearly identifiable long or short strands and bound proteins were counted. XPD protein position was determined based on the distance of the protein from the end of the strand.

We examined distribution with and without blocking the ends of short strands, given that XPD has some preferential binding affinity for single-strand overhangs. Comparing raw protein position (middle vs. ends) on long and short strands, XPD exhibited a large preference for the ends (approximately 300 bp range) of short duplexes containing overhangs (> 50% of bound proteins). To block XPD from binding to the 14 bp segment, overhang complements were added to DNA/protein solutions in excess. We found, when we blocked the overhangs, no preferential binding to the short strands was observed (vide infra).

AFM images of DNA-bound protein complexes are represented in Fig. 2. XPD protein bound to matched DNA (Fig. 2 top) can be easily distinguished based on strand height profile of tapping mode images. Zoomed in images display clearly identifiable long and short strands with protein bound (Fig. 2 bottom). Images of DNA and protein complexes were acquired with a scan size of 2 × 2 μm² or 3 × 3 μm² at a scan rate of 3.05 Hz. Because AFM images vary with surface cleavage, sample wetness, deposition time, and volume, > 200 long or short strands were counted for at least three independent samples. The uncertainty was based on the total number of proteins counted. Importantly, we measured the relative binding affinity of mutants and wild type (WT) proteins, and in all of the samples described, the number of proteins bound per base pair remained constant (Table S1). Thus, any changes we see in distribution are not due to differences in binding affinities of proteins.

In order to probe the DNA CT properties of a redox-active DNA helicase, XPD and DNA complexes were examined with AFM. XPD shows redistribution onto long mismatched strands with a ratio of protein binding densities, r (long/short) of 1.54 ± 0.08 (Fig. 3). We observed 0.23 proteins per kilobase of long strand and 0.15 proteins per kilobase of short strand (see Table S1 for further details). If, instead, we examine XPD distribution when long and short strands are fully matched, we see a binding density ratio of 0.94 ± 0.05. Previously, for EndoIII we found r, for (long/short) was 1.6 ± 0.09 (22). We expect a ratio of 1 if there is an equal distribution of proteins on matched and mismatched strands (22). Thus, similar to EndoIII distribution, XPD, an NER protein with DNA-bound redox activity, redistributes to localize in the vicinity of a C∶A mismatch.

We also tested the possible redistribution of a CT-deficient mutant of SaXPD, L325V (L461 in human XPD). L325V, aligns with mutated residues in human XPD and Schizosaccharomyces pombe Rad 15 that are associated with TTD, XP, and XP/CS (25, 31, 33). It should be noted, however, that the structural scaffold required to form the TFIIH machinery is not disrupted in the L325V mutant. As evident in Fig. 3, the L325V XPD mutant exhibits an electrochemical signal that is less than half that of WT XPD, indicating it is deficient at performing DNA CT. Interestingly, L325V is 30 Å away from the [4Fe-4S] cluster, yet affects protein/DNA CT. Significantly, when we examine whether this electrochemically deficient mutant redistributes onto the mismatched strand in our AFM assay, we find that there is no preference for the mismatched strands; L325V exhibits a protein binding density ratio, r (long/short) of 1.14 ± 0.06 in mismatched samples, which is within error of the fully matched binding density ratio r (long/short) of 1.05 ± 0.07 (Fig. 3). Thus, as with the CT-deficient EndoIII mutant Y82A (22), we find a correlation between the inability of L325V to perform CT and its lack of redistribution onto mismatched strands. Even though the [4Fe-4S] cluster is intact, this mutant protein cannot cooperate with other proteins using DNA CT.

We have established that DNA-mediated CT occurs on a much faster time scale than protein diffusion (43, 44), and we have characterized several mutants of EndoIII, from the BER repair pathway that exhibit similar midpoint potentials of approximately 80( ± 30) mV vs. NHE yet differ in their ability to carry out DNA CT (14, 23, 26). We have not, however, yet provided evidence that different proteins can signal one another through DNA-mediated CT. AFM provides an opportunity to examine this issue by assaying protein mixtures.

To test this model, WT XPD and EndoIII, both proficient at DNA CT, were mixed in a 1∶1 equimolar ratio with DNA. Short duplex overhangs were blocked with excess complementary single strand. Fig. 4 shows representative images of fully matched DNA strands or mismatched DNA strands incubated overnight with XPD/EndoIII 1∶1 protein (WT^XPD/WT^EndoIII). XPD is twice as large as EndoIII (65 and 32 kDa respectively) (PDB IDs: 3CRV, 1P59). The average height of the protein in AFM studies can be estimated using the equation: R = 0.717(M)^1/3 where R is the radius of the protein globule in Å and M is the mass of the protein in Da (45). Thus, XPD is expected to have a height of 5.6 nm, which is within error of EndoIII (4.6 nm). While in general we cannot distinguish them, select images (Fig. 4 inset) are suggestive of both EndoIII and XPD binding. XPD (0.3 μM)/EndoIII (0.3 μM) 1∶1 protein mixture shows that XPD and EndoIII redistribute onto long mismatched strands, with a binding density ratio r (long/short) of 1.75 ± 0.13. Fully matched controls yield a binding density ratio r (long/short) of 1.02 ± 0.07. Importantly, controls were also performed with only XPD (0.3 μM) under the same conditions as mixed protein experiments. Consistent with XPD at a higher concentration (0.6 μM), XPD at half the concentration (0.3 μM) redistributes onto long mismatched strands (r = 1.38 ± 0.07) with no redistribution in fully matched samples (r = 1.08 ± 0.07). Significantly, mixing XPD and EndoIII protein results in a binding density ratio of 1.75, which is higher than that of each protein separately. This result suggests that these two proteins signal one another to localize in the vicinity of the lesion.

In order to explore cooperativity between XPD and EndoIII, we then replaced WT EndoIII with mutant Y82A EndoIII, a protein that binds to DNA but cannot perform DNA-mediated CT efficiently (22). In previous studies, Y82A did not redistribute in mismatched samples (r = 0.9 ± 0.1) (22). When XPD is mixed with Y82A 1∶1 (WT^XPD/Y82A^EndoIII), there is no redistribution onto mismatched strands (r = 0.98 ± 0.05). WT^XPD/Y82A^EndoIII matched controls are within error of mismatched results with a binding density ratio of 1.11 ± 0.07. The number of proteins bound per kilobasepair remains the same between WT^XPD/Y82A^EndoIII and WT^XPD/WT^EndoIII mixtures (0.13 proteins/kbp). When WT^XPD is titrated into Y82A^EndoIII at a ratio of 3∶1, increasing the probability that two proteins bound on the same strand are both electrochemically active, the proteins once again redistribute onto mismatched strands.

Complementing WT^XPD/WT^EndoIII and WT^XPD/Y82A^EndoIII mixture studies, we also investigated signaling between EndoIII and XPD mutant L325V. Similar to EndoIII Y82A, XPD L325V cannot redistribute onto mismatched strands. WT^XPD/L325V^XPD 1∶1 mixtures were examined initially to determine whether L325V, deficient in DNA CT, could signal XPD. WT^XPD/L325V^XPD mixtures showed no redistribution (r = 0.88 ± 0.04). This result was within error of matched controls, with a binding density ratio revealing even a slight preference for short strands (r = 0.90 ± 0.05). L325V inhibits XPD from relocalizing in the vicinity of lesions, but does L325V have any influence on the search for damage by EndoIII BER protein? WT^EndoIII/L325V^XPD mixtures were examined and exhibited no difference in mismatched vs. matched samples, where r = 1.02 ± 0.05 and 0.94 ± 0.04 respectively. As with the WT^XPD/Y82A^EndoIII mixture, here we see that L325V binding alters EndoIII redistribution. Fig. 5 shows binding density ratio comparisons for WT^XPD/WT^EndoIII, WT^XPD/Y82A^EndoIII, WT^EndoIII/L325V^XPD, and WT^XPD/L325V^XPD 1∶1 protein mixtures.

In the case of the mixtures of active proteins and CT-deficient mutants, we might have expected binding density ratios to be equal to that of pure samples of active repair protein. However, we are testing these redistributions at protein loadings on the DNA strands of approximately 2 per strand. Thus if one of the proteins is CT defective, there is no protein to which the CT-proficient protein may signal. Certainly the striking differences seen between CT-active protein mixtures and those containing CT-deficient mutants support the idea that the proteins can carry out DNA CT to one another. Having established that this ability to redistribute in the vicinity of mismatches depends on the ability of the protein to carry out DNA CT (23), these results thus provide evidence for DNA CT as a means of interprotein signaling.

Four primers with the following sequences were synthesized using standard phosphoramidite chemistry:

The underlined bases highlight the location of a 2′-O-methyl residue (Glen Research). Primers were phosphorylated using 5 U PNK, 10% PNK buffer, 0.5 mM ATP for 5 h at 37 °C. Primers were purified using phenol-chloroform extraction followed by ethanol precipitation (54). After being dried in vacuo, primers were redissolved in 20 μL water and used in separate PCR reactions (41) using pUC19 as a template to generate two duplexes 1,610 bp and 2,157 bp (matched), each containing one 14-nucleotide single-strand overhang. Each 100 μL PCR reaction contained 50 pmol of each of two primers, 1X Taq buffer (100 mM Tris-HCl, 15 mM MgCl₂, 500 mM KCl, pH 8.3), 0.2 mM each dNTP, 1 ng plasmid template pUC19 and 3 U Taq polymerase (Roche). A typical step program for PCR was as follows: After incubation at 94 °C for 10 min, 34 cycles were performed as follows: 94 °C for 1 min, (52 °C for primers 1 + 2, 54 °C for primers 2 + 3, or 2 + 4), for 1 min, and 72 °C for 3 min. The PCR product was then suspended in 50 mM NaCl/ 5 mM phosphate buffer and quantitated. Separate duplexes were annealed at 65 °C for 8 min in 10 mM Tris buffer, then cooled to 20 °C during 2 h. A total of 15 units of T4 DNA ligase and 10% T4 ligase buffer were added (total reaction volume ∼20 μL) and incubated overnight at 16 °C, followed by deactivation for 10 min at 65 °C to yield the 3,767 bp (mismatched) long strand. We did not bring the ligation reaction to completion, so as to obtain a mixture of DNA samples that were equivalent other than the presence of the mismatch at the ligation site. The DNA duplexes (ligated and unligated) were then analyzed by 0.6% agarose gel electrophoresis. Single DNA strands complementary to short duplex overhangs were ordered from IDT: 5′-GACTGAACTCTGTAC-3′ Tm = 41.6 °C (1.6 kbp duplex overhang), 5′-GTACAGAGTTCAGTC-3′ Tm = 41.6 °C (2.2 kbp matched duplex overhang), and 5′-GTATAGAGTTCAGTC-3′ Tm = 37.5 °C (2.2 kbp mismatched duplex overhang). Single-strand DNA was purified using reversed-phased HPLC and verified with MALDI-TOF mass spectrometry.

EndoIII and Y82A EndoIII were expressed from the pNTH10 expression vector and purified as described previously (22). EndoIII and Y82A were stored in 20 mM sodium phosphate pH 7.5, 100 mM NaCl, 20% glycerol, and 0.5 mM EDTA buffer. Protein concentrations were determined using the UV-visible absorbance of the [4Fe-4S] cluster (410 nm, ε = 17,000) (55). XPD was purified as previously described (34).

AFM experiments were performed using the protocol similar to that reported previously (23). Stock DNA solution contained 50–200 ng of total DNA (approximately 6 μM) composed of the mixture of ligated 3.8-kbp duplexes and the two unligated duplexes (1.6 and 2.2 kb) in 6 mM MgCl₂/Tris-EDTA buffer. The 1.6 kb overhang complement (60 μM) was added to the DNA solution to block the 14 bp single-strand overhangs generated by PCR. This sample was then incubated overnight at 4 °C. XPD protein was dialyzed against the protein buffer (20 mM phosphate, 100 mM NaCl, 1 mM EDTA, 5% glycerol, pH 7.5 and filtered prior to use) to remove residual DTT. The concentration of individual proteins were determined by UV/visible spectrophotometry (Beckman DU 7400) using ε = 17,000 M^-1 at 410 nM for the [4Fe-4S] cluster. After addition of excess 2.2 kb duplex complement (60 μM), XPD (0.6 μM) was added to the stock DNA solution. This protein/DNA solution was incubated at 4 °C overnight. Sample was then deposited (5–10 μL) onto a freshly cleaved mica surface for 1–2 min, rinsed with 2 mL of water, and dried under argon. Mutant XPD protein (L325V) was added to a stock solution of 50 ng DNA for a final protein concentration of 0.6 μM. Deposition conditions were identical to that for WT XPD- DNA samples after incubation at 4 °C overnight.

For mixed protein experiments (WT^XPD/WT^EndoIII, WT^XPD/Y82A^EndoIII, WT^EndoIII/L325V^XPD, and WT^XPD/L325V^XPD), XPD, EndoIII, L325V, or Y82A were added to the prepared DNA solution described previously at equimolar (1∶1) concentrations (0.3 μM each) prior to incubation at 4 °C overnight. Protein/DNA complexes were formed with DNA solution containing approximately 200 ng of the mixture of PCR products and overhang complements (approximately 6 μM DNA) in 6 mM MgCl₂ /Tris-EDTA buffer at 4 °C overnight. The reaction mixture was then deposited (5–10 μL) on the mica surface for 1–2 min, rinsed with 2 mL water and dried under argon.

Silicon AFM Probes purchased from Budget Sensors, with a spring constant of 3 N/m and a resonance frequency of 75 kHz, were used in a Digital Instruments Multimode SPM. Images were captured in air with scan areas of 2 × 2 μm² or 3 × 3 μm² in tapping mode, at an amplitude of 0.54–2.00 V and at a scan rate of 3.05 Hz. Scan rates of 3.05 Hz were used in order to obtain images of higher quality.

WSxM software was used to measure general DNA contour lengths and height profiles of the proteins as described previously (23, 56). For each dataset, images from at least three independent samples were analyzed, compared, and pooled (> 200 long or short strands). Distinguishable strands and protein positions were counted by hand. The binding density ratio, r, is defined as the ratio of the proteins bound on long strands divided by proteins bound on short strands. The ratio is normalized for length by dividing by 1.9 kbp, which is the average length of the short strands. Binding affinities were found by determining the number of proteins bound per kilobasepair strand. The uncertainty was determined through the total number of proteins observed.

Protein electrochemistry was performed as previously described (34). Briefly, individual proteins samples were dialyzed to remove residual DTT and quantified based on 410 nm absorbance. Protein was then added to a DNA modified electrode containing a 9 nucleotide 5′ single-strand overhang. Cyclic Voltammograms were then obtained using Ag/AgCl reference electrodes, Pt auxiliary electrode at 50 mV/s scan rate on a CH Instruments 620C electrochemical analyzer.