Alkyltransferase-like protein clusters scan DNA rapidly over long distances and recruit NER to alkyl-DNA lesions

Significance

We directly visualize cotranslocation on DNA by the alkyltransferase-like protein, ATL, and UvrA, the initiating enzyme of prokaryotic nucleotide excision repair (NER). Our findings offer an enhanced understanding of the recruitment of NER to nonnative NER target lesions. Furthermore, we elucidate the mechanism of DNA lesion search and recognition by ATL and its catalytically active homolog AGT. Our data support a paradigm for the transition from oligomeric lesion search complexes of ATL (and AGT) to the repair-initiating, monomeric protein complexes based on the dissociation of oligomeric structures due to stable, stronger DNA bending at a target lesion.

Abstract

Alkylation of guanine bases in DNA is detrimental to cells due to its high mutagenic and cytotoxic potential and is repaired by the alkyltransferase AGT. Additionally, alkyltransferase-like proteins (ATLs), which are structurally similar to AGTs, have been identified in many organisms. While ATLs are per se catalytically inactive, strong evidence has suggested that ATLs target alkyl lesions to the nucleotide excision repair system (NER). Using a combination of single-molecule and ensemble approaches, we show here recruitment of UvrA, the initiating enzyme of prokaryotic NER, to an alkyl lesion by ATL. We further characterize lesion recognition by ATL and directly visualize DNA lesion search by highly motile ATL and ATL–UvrA complexes on DNA at the molecular level. Based on the high similarity of ATLs and the DNA-interacting domain of AGTs, our results provide important insight in the lesion search mechanism, not only by ATL but also by AGT, thus opening opportunities for controlling the action of AGT for therapeutic benefit during chemotherapy.

Alkylation of DNA bases is highly mutagenic and cytotoxic. The importance of dedicated repair of alkyl lesions in DNA is underlined by the fact that they are the only type of DNA damage repaired by the highly efficient and noninvasive direct damage reversal machinery in humans (1, 2). In particular, guanine alkylation at the O⁶ position (O⁶-alkylguanine) is among the most lethal DNA lesions (reviewed in ref. 1) and is repaired by the direct damage reversal protein O⁶-alkylguanine-DNA alkyltransferase (AGT, also known as methylguanine methyltransferase, MGMT). AGT directly transfers the offending alkyl group from the base to a catalytic cysteine in the protein, resulting in intact DNA and alkylated AGT, which is consequently ubiquitinylated and degraded (3).

Catalytically inactive homologs of AGT, named alkyltransferase-like proteins, ATLs, have been identified in a wide spectrum of organisms ranging from prokaryotes to lower eukaryotes (4, 5). ATLs have strong structural similarity with the C-terminal domain of AGT, which harbors the DNA-binding and catalytic DNA repair activity of the protein. Consequently, DNA interactions by AGTs and ATLs appear very similar in crystal structures. Like AGT, structures of ATLs bound to alkylated DNA show the protein bound to the minor groove of DNA via a helix–turn–helix motif with the damaged base flipped into a binding pocket in the protein (SI Appendix, Fig. S1) (2, 6). However, ATLs also possess structural features and differences in amino acid sequence compared to AGTs that determine their distinctly different functional role. Notably, ATLs do not possess a catalytic cysteine residue (C145 in human AGT, hAGT) to accept the offending alkyl group from the damaged base (2, 6–8) and are thus catalytically inactive in DNA repair. Instead, ATLs form stable, DNA-distorting complexes at alkyl lesion sites that have been proposed to transform the relatively inconspicuous alkyl lesions into distinctly recognizable target sites of the nucleotide excision repair (NER) system (6). A conceptually similar marking of per se only mildly distorting lesions for recognition by NER has previously been described for cyclopyrimidine dimer ultraviolet (UV) lesions, which are targeted to NER by the UV-DDB (UV-damaged DNA binding) protein complex (9). A role of ATL in alkyl lesion repair by NER is supported by a marked epistatic relationship of ATL and NER proteins (6, 10), as well as by direct interactions between NER proteins and ATLs (but not AGTs) even cross-species, in line with the high degree of conservation of these proteins (6). A unique C-terminal extension in ATLs that is not found in AGTs supports DNA distortion by ATL and is a strong candidate for mediating protein interactions for the recruitment of NER components (6).

In light of the high similarity between ATLs and the DNA-interacting domain of AGTs, similar lesion search strategies may be expected. Indeed, cooperative protein–protein interactions upon DNA binding that play a role in DNA target site search by hAGT (11) have also been proposed for ATLs based on gel shift assays and sedimentation studies with Atl1 from Schizosaccharomyces pombe (6, 12). However, the N-terminal domain of hAGT, which mediates the long-range cooperative protein–protein interactions (between the n^th and n + third monomer unit) in DNA-bound complexes (13), is missing in ATLs. A conserved strategy of cooperative protein clusters in alkyl lesion search and recognition by AGTs and ATLs—despite differences in the precise structural mechanisms underlying such cooperativity—would hence be remarkable.

Here, we used single-molecule analyses by atomic force and fluorescence microscopy to investigate conserved mechanistic features between AGTs and ATLs despite marked structural and functional differences. In our studies we used ATL from Escherichia coli (eAtl), which will be referred to in the following simply as ATL. Our data reveal rapid dynamics of oligomeric ATL on DNA during lesion search and distinct structural changes in ATL complexes upon lesion recognition, which may also apply to AGT. Specific for ATL, we further demonstrate enhanced recognition of O⁶-methylguanine alkyl lesions by NER through direct interactions of E. coli ATL with UvrA, the initiating enzyme of prokaryotic NER, and directly show cotranslocation on DNA of ATL and UvrA during lesion search.

Results

Lesion Recognition by ATL.

Crystal structures of ATL–DNA complexes have shown single protein molecules bound to an alkyl lesion (6, 10, 14, 15). Our electromobility shift assay (EMSA) data and sedimentation coefficient distributions obtained by sedimentation velocity (SV) analytical ultracentrifugation (AUC) show monomeric ATL complexes exclusively on lesion-containing DNA at protein concentrations ≤1 μM (Fig. 1 A and B black arrow), consistent with previous studies (6). This low-molecular-weight protein–DNA complex species is absent in samples of ATL with nonspecific (undamaged) DNA substrate in EMSAs (Fig. 1 A, Left) and unstable in solution as demonstrated by the AUC sedimentation coefficient distribution (Fig. 1B). Consistently, atomic force microscopy (AFM) imaging experiments at 1 μM ATL concentration (Fig. 1C) show monomeric ATL (∼124 nm³ including DNA volume ∼63 nm³; Fig. 1D and SI Appendix, Fig. S2A) bound with high specificity to lesion sites (average specificity S_avg = 600 ± 252 from five independent experiments or S = 567 from a Gaussian fit to the pooled distribution; Fig. 1E).

Fig. 1.

High alkyl-lesion specificity of ATL. (A) EMSA of DNA binding by increasing concentrations of ATL (50 nM DNA, from right to left: 0/100/250/500/750/1,000/2,000 nM ATL; left nonspecific [undamaged] DNA, right alkyl-lesion DNA). Arrows indicate free DNA (white), and monomeric (black) and oligomeric (gray) ATL complexes with DNA. (B) AUC sedimentation coefficient distributions of 500 nM and 2 μM ATL with 48-bp alkyl-lesion DNA (solid lines) and 2 μM ATL with nonspecific 48-bp DNA (dashed line). (C) AFM image of 1 μM ATL incubated with DNA containing an alkyl lesion at ∼30% of the DNA fragment length. Arrows indicate protein complexes on DNA (white: specific at ∼30%, gray: nonspecific). (D) Volume distribution of DNA-bound ATL complexes shows a maximum at ∼124 nm³ based on a Gaussian fit to the data (gray line). After subtraction of the contributing DNA volume of ∼63 nm³ (SI Appendix, Fig. S2A), this corresponds to ∼56 kDa, consistent with M_r = 57.5 kDa of the MBP-ATL fusion protein. (Inset) The volume-to-mass calibration for the employed AFM system (16). (E) Protein position distribution on the DNA substrate (pooled from five independent experiments, n_total = 603) plotted as the distance of protein peaks from the closer DNA end in units of total DNA length. The alkyl lesion position at ∼30% of the DNA length (black arrow) is recognized by ATL with a specificity S ∼600. (B–E) For ease of detection, for AUC and AFM experiments ATL N-terminally fused to MBP was used (total mass of 15.5 kDa + 42 kDa). EMSA with MBP-ATL showed results comparable to ATL without MBP (SI Appendix, Fig. S3), confirming that the MBP fusion does not affect ATL interactions. (F and G) DNA bend angles at complexes bound at the position of the alkyl lesion (at 31 ± 5%, F) and at nonspecific DNA sites (G) show distributions around ∼25° and 50° (example image shown in inset) for the lesion-bound complexes and ∼25° at nonspecific sites (example image in inset), as determined from Gaussian fits to the data (gray lines). In addition, a higher DNA bend angle can be discerned for both lesion-bound and nonspecific complexes (∼90°), which may be attributed to DNA superstructures or nonspecific contamination peaks on the DNA (examples shown in SI Appendix, Fig. S4).

Interestingly, our AFM-based volume analyses show no major differences between complexes bound at the alkyl lesion position (O⁶-methylguanine, in the following referred to only as “alkyl lesion” for simplicity) and those bound to nonspecific DNA sites (i.e., not at the target lesion). Monomeric volumes also at nonspecific DNA sites as seen by AFM may seem in apparent contrast to the absence of monomeric complexes on nonspecific DNA substrate in EMSAs. However, the strong enhancement of DNA binding at the lesion over binding to undamaged sites (Fig. 1E) indicates different stabilities of lesion-specific and nonspecific complexes, likely corresponding to different conformations of the two complex types. To characterize their conformational properties, we measured DNA bending by ATL bound at the alkyl lesion or bound on nonspecific DNA. At the alkyl lesion site, monomeric ATL–DNA complexes display an average DNA bend angle of ∼50° (Fig. 1F), consistent with DNA bending toward the major groove by ∼45° seen in crystal structures (6). In addition, a second state with DNA bending of ∼25° can be discerned in these specific complexes. Control measurements at alkyl lesion sites in the absence of protein showed no innate bending at the resolution of these experiments (bend angle of 0°; SI Appendix, Fig. S2B), consistent with previous reports (17). Intriguingly, in ATL complexes bound to the DNA at nonspecific sites, we exclusively observed the more gently bent ∼25° bend angle state (Fig. 1G). While the Gaussian fits provide only crude estimates of DNA bend angle states for the comparably low number of data points from three independent experiments, the absence of the ∼50° species in nonspecific complexes is clearly apparent. It is further important to note that intramolecular cross-linking of monomeric ATL complexes at nonspecific DNA sites was able to stabilize the ∼50° bent state observed in specific lesion-bound ATL complexes (SI Appendix, Fig. S4). These results are consistent with a conformational switch in the specific, monomeric, lesion-bound complex that leads to stronger protein–DNA contacts concomitant with enhanced complex stability and stronger DNA bending.

Cooperative DNA Binding by ATL.

In contrast to the monomeric species described above, we observed oligomeric complexes on lesion-containing as well as undamaged DNA at high protein concentrations (≥2 μM), as revealed by considerably larger shifts in EMSAs (Fig. 1A, gray arrow) and larger sedimentation coefficients in SV AUC (Fig. 2A). The EMSAs show a one-step increase from monomeric to higher-order complexes in the presence of an alkyl lesion and a one-step shift from unbound DNA to higher-order complexes for undamaged DNA, indicative of cooperativity (Fig. 1A), as also previously suggested (6). AUC experiments at high (20 μM) protein concentration and either alkyl lesion containing or undamaged DNA substrate revealed maximum sedimentation coefficients of ∼12 S consistent with a molecular mass of ∼300 kDa (Fig. 2A, black arrow). Although molecular weight estimates from SV AUC are not very precise due to contributions from protein shape to sedimentation velocities, these data are consistent with approximately five monomers of maltose binding protein (MBP)-tagged ATL in DNA-bound clusters (MBP-ATL relative molecular mass, M_r, ∼60 kDa). The binding site of ATL on DNA has been shown to cover ∼8 bp of DNA (6). The short DNA substrates (48 bp) employed in our AUC and EMSA studies hence provide a maximum of six binding sites for ATL. These studies therefore do not unambiguously allow conclusions on limiting ATL cluster sizes. In contrast, the DNA fragments used in our AFM experiments consist of ∼900 bp (alkyl lesion-containing DNA) or ∼2,000 bp (undamaged DNA). AFM analyses using these nonlimiting DNA substrate lengths hence allow us to investigate ATL cluster growth limitations. AFM imaging data on glutaraldehyde–cross-linked ATL–DNA samples at high protein concentrations (10 to 20 μM) show distinctly larger complex volumes on DNA (Fig. 2 B and C) than samples cross-linked at low protein concentration (1 μM). A multi-Gaussian fit to the data indicates limiting cluster sizes of ∼340 nm³ (arrow in Fig. 2C) for high and ∼124 nm³ for low ATL concentrations (including bound DNA volume; SI Appendix, Fig. S2C). After subtraction of the included DNA volume (∼63 nm³; SI Appendix, Fig. S2A) and applying the previously established calibration for our AFM setup (Fig. 1 D, Inset and ref. 16), this volume corresponds to ∼240-kDa protein molecular weight. Non–DNA-bound ATL, in contrast, shows predominantly monomeric volumes in these cross-linked samples at high protein concentrations (∼60 nm³; SI Appendix, Fig. S2D). Our AFM analyses hence indicate short ATL clusters on DNA, with limiting cluster size of approximately four monomers of ATL, in close agreement with our AUC-based estimate.

Fig. 2.

Cooperative ATL complexes with undamaged DNA at high protein concentration. (A) SV AUC sedimentation coefficient distributions for high ATL concentration (20 μM) with alkyl lesion containing or nonspecific DNA substrate show limiting cluster sizes with ∼12 S. ATL (at 20 μM) in the absence of DNA is shown for reference. (B) Representative AFM image of high ATL concentration (20 μM) incubated with undamaged DNA substrate. To avoid dissociation during dilution and deposition on mica surfaces, ATL clusters were cross-linked with glutaraldehyde prior to sample deposition. (C) AFM volume distribution of DNA-bound ATL at 20 μM. The arrow marks the cutoff at ∼340 nm³. (D) Manual model illustrating putative ATL–ATL interactions in a DNA-bound complex, based on the crystal structure of monomeric ATL from S. pombe (apo, PDB ID code 3gva), a binding site size of 8 bp per monomer of ATL, and a DNA bent by ∼25° (extracted from PDB ID code 4yex). The putative contact region between two ATL molecules is indicated by the dashed oval.

We visualized potential protein–protein interactions in cooperative ATL clusters on DNA by crude, manual modeling of two ATL monomers on DNA with a separation of 7 bp (Fig. 2D). The model is based on the crystal structure of apo ATL from S. pombe (Protein Data Bank [PDB] ID code 3gva), nonspecific DNA bending by ∼25° described above (∼25° bent DNA structure, extracted from PDB ID code 4yex) and the known DNA binding site size of ATL of 8 bp (6). Interestingly, our model shows the N terminus of one ATL molecule and the C-terminal loop extension of the neighboring one in close proximity to each other, with some side chains clashing; however, loop flexibility or slight alterations in the positions of the ATL molecules on a native DNA substrate may prevent this collision and thus result in a small, predominantly electrostatic interface between the two ATL molecules. We speculate that ATL molecules spaced by 7 bp along a gently bent DNA may indeed contact each other via an interface involving the N-terminal helix and C-terminal loop region of adjacent molecules. A more detailed structural understanding of potential ATL oligomerization on DNA and validation of the proposed protein–protein interface will require further studies, for example by molecular dynamics (MD) simulations.

High Mobility of ATL Clusters on DNA.

Using a correlative fluorescence-optical tweezers system (C-Trap; see Materials and Methods), we investigated dynamics of ATL complexes on nonspecific (undamaged) DNA (Fig. 3A). For detection, ATL was conjugated to quantum dots (QD with emission maximum at 605 nm, QD605) via an antibody sandwich linker (see Materials and Methods and SI Appendix, Fig. S5). To confirm that the QD–ATL conjugate was still functional, we used AFM analyses to demonstrate recognition of the alkyl lesion at ∼30% DNA length by QD–ATL (SI Appendix, Fig. S5). In fluorescence-optical tweezers experiments, samples of QD–ATL in the nanomolar range (80 nM) exhibited low fluorescence intensity traces on the DNA (51 ± 39 arbitrary units [a.u.], Fig. 3B gray bars and SI Appendix, Fig. S6), which we interpret as monomeric ATL. To achieve high (micromolar) ATL concentrations in the samples and still observe individual cluster traces on the DNA, QD-labeled ATL was mixed with an excess of unlabeled ATL; at a total ATL concentration of 2.25 μM, only one in eight ATL molecules was thus QD-labeled. In these measurements, we identified two dominant populations (Fig. 3B, black bars) with low (54 ± 28 a.u., ∼43% of complexes based on triple Gaussian fit) and medium fluorescence intensity (179 ± 42 a.u., ∼46%). The low-intensity species is consistent with the monomeric species in the nanomolar ATL sample (gray bars in Fig. 3B), which may serve as nucleation sites for cluster formation. The medium-intensity species indicate oligomeric ATL complexes on the DNA and are consistent with three or more ATL monomers per cluster. A minor population of complexes (∼11%) showed high fluorescence intensities (600 ± 52 au). Since only one in eight ATL monomers was fluorescently labeled in these samples, these signals are difficult to quantify in terms of stoichiometry; however, the data are overall consistent with our AFM volume analyses at high ATL concentrations (Fig. 2C). Although QDs are known to blink (frequent switching between fluorescent on and off states), changes observed in the fluorescence intensities of the traces may further indicate exchange of monomer subunits within the clusters (Fig. 3 A, Middle). Importantly, ATL clusters remained stably bound to the DNA for minutes and scanned the DNA over distances of several micrometers (Fig. 3 A and C). Interestingly, the ATL clusters display rapid (up to ∼16 μm/s) bidirectional diffusion over long distances as well as static binding, and are able to convert between these different modes (e.g., Fig. 3 A, Top, yellow arrow). Gaussian fitting and mean square displacement (MSD) analysis of kymographs (Materials and Methods) support multiple modes of DNA interactions, providing diffusion constants that span a large range (D ∼0.001 μm²/s to 2 μm²/s) and anomalous diffusive exponents (α factors) of 0.55 ± 0.17 (Fig. 3D), indicating smooth, unbiased, subdiffusive movement along the DNA. In addition, when a gentle flow force was introduced (e.g., Fig. 3 A, Bottom), slow long-distance sliding of ATL clusters along the DNA could be observed in the direction of the flow (intensities of 224 ± 190 a.u. with overall speeds of ∼0.05 to 8 μm/s over >12 μm, not included in quantitative analyses of Fig. 3 C and D). These data are consistent with a loosely bound conformation of ATL on undamaged DNA, both for monomeric ATL (as also indicated by the DNA bend angles observed in AFM) and oligomeric complexes that can scan undamaged DNA rapidly and over large distances.

Fig. 3.

High mobility of ATL clusters on DNA. (A) Kymographs of QD605 labeled ATL on nonspecific DNA tethers obtained by a combined fluorescence-optical tweezers setup. The beads and DNA tether are indicated schematically on the left and a fluorescent image of the protein complexes on the DNA stretched between the trapped beads is shown in the top right corner. The yellow arrow points at an example of mode change in a DNA-bound ATL cluster. (B) Intensity distributions of ATL complexes on the DNA at 80 nM (gray) and 2 μM (black) ATL. For 80 nM ATL (18 traces), only a single species with ∼50 a.u. intensity was observed. At 2 μM ATL (95 traces), the data show two dominant, approximately equally populated species with ∼50 a.u. and ∼200 a.u,, and a minor species with ∼600 a.u. (∼14%) intensities. (C) Distance covered by ATL clusters on the DNA plotted versus their respective intensities. Only kymographs obtained in the absence of flow were included (n = 29). (D) Anomalous diffusion exponents (alpha) plotted over diffusion constants. Only kymographs obtained in the absence of flow and with R² ≥ 0.84 in MSD fits were included (n = 25).

Recruitment of NER.

Cell survival analyses using different deletion strains indicated that ATL and NER proteins are parts of the same pathway for alkyl lesion repair (6). Comparison of crystal structures of ATL in its apo form and bound to DNA at an alkyl lesion site led to the hypothesis that a central loop (termed the binding site loop) of ATL shifts toward the binding site and away from the C-terminal loop upon DNA binding, thus exposing a binding site for UvrA (SI Appendix, Fig. S1B and ref. 6). There are three possible scenarios that each imply a subtly different mechanism of NER recruitment by ATL: 1) UvrA is recruited by ATL only once ATL is bound at an alkyl lesion site, 2) UvrA is recruited by nonspecifically DNA-bound ATL, or 3) UvrA and ATL interact with each other prior to binding to DNA. Scenario 1 requires ATL to be specifically bound at the target lesion with the alkylated base flipped into its binding pocket for UvrA recruitment. In contrast, scenarios 2 and 3 involve ATL and UvrA cotranslating on the DNA toward alkyl lesion sites. We aimed at distinguishing between these three scenarios, using EMSA, AUC, AFM, and fluorescence-optical tweezers analyses.

EMSA data and AUC sedimentation coefficient distributions show distinct complexes formed with DNA in the presence of ATL and UvrA (black arrow in Fig. 4A; ∼11 S species in AUC black line in Fig. 4B). These complexes clearly differ from DNA-bound complexes of the individual proteins (dark gray arrows in Fig. 4A and ∼5 S and 7 S peaks of dashed line in Fig. 4B for ATL, light gray arrows in Fig. 4A and ∼10 S peak of gray line in Fig. 4B for UvrA). Interestingly, the EMSA data showed ATL–UvrA complexes also on nonspecific DNA, in line with scenarios 2 and 3. In addition, biolayer interferometry (BLI) measurements yielded a moderately high affinity of ATL and UvrA in the absence of DNA with a dissociation constant K_d ∼130 nM (128 nM ± 20 nM; Fig. 4C and SI Appendix, Fig. S7A). High-affinity interactions (30 nM) have previously been shown between UvrA and ATL from Thermus thermophilus (18). However, the affinity displayed here by proteins from two different bacterial organisms (ATL from E. coli and UvrA from the thermophilic Bacillus caldotenax) is remarkable and consistent with highly conserved proteins that allow for interspecies interactions [as also previously described (6)]. Furthermore, BLI and thermophoresis measurements also support comparable affinities for UvrA and the MBP-tagged ATL protein (K_d of 198 nM ± 15 nM; SI Appendix, Fig. S7), thus confirming that the MBP tag on ATL (in AUC and AFM experiments) does not interfere with the interaction. Together with the EMSA and AUC data, these results argue against the requirement for ATL to be bound to an alkyl lesion to recruit UvrA from solution. Instead, a preformed ATL–UvrA complex may scan the DNA for lesions, poised to deliver UvrA to the lesion to initiate NER.

Fig. 4.

UvrA recruitment. (A) EMSA analyses show UvrA–ATL complex formation (black arrow) on alkyl-lesion DNA (Left) as well as nonspecific DNA (Right). When present (+), UvrA was at 500 nM; ATL was applied at low (500 nM) or high (2 μM) concentration as indicated by the triangular symbol. A larger shift is seen between UvrA and ATL–UvrA complexes on DNA using MBP-tagged ATL, consistent with its larger molecular mass (SI Appendix, Fig. S3B). (B) AUC sedimentation coefficient distributions for incubations of alkyl-lesion DNA with ATL (2 μM, dashed line), UvrA (500 nM, gray), and ATL and UvrA together (black line). For better detection MBP tagged ATL was used in the experiments. (C) BLI association and dissociation curves for the UvrA–ATL interaction in the absence of DNA. UvrA was titrated (from 50 nM to 1 μM as indicated) to surface immobilized ATL (50 nM). Equilibrium Hill fits provided a K_d of 128 ± 20 (SI Appendix, Fig. S7). Similar K_d values were also obtained by microscale thermophoresis using fluorescently labeled ATL with nonlabeled UvrA or QD conjugated UvrA with nonlabeled MBP-ATL (K_ds of ∼200 nM, SI Appendix, Fig. S7). (D) Overlay of QD525-ATL (blue) and QD705-UvrA (red) kymographs demonstrate cotranslocation of UvrA and ATL on undamaged lambda phage DNA. Individual ATL and UvrA traces are shown below. Additional kymographs are presented in SI Appendix, Fig. S8. (E–G) AFM imaging analysis of UvrA (500 nM) and ATL (2 μM, MBP-ATL) incubated with DNA substrate containing an alkyl lesion at ∼30% of its length. (E) Exemplary AFM images. (F) Stacked graph representation of volumes of protein complexes bound to the alkyl-lesion position (black, n = 176) or nonspecific (strand-internal) DNA sites (gray, n = 277). Data were pooled from four independent experiments. The red box indicates complex sizes that are not observed in UvrA (Inset, n = 127) or ATL only samples (SI Appendix, Fig. S2E). (G) Distribution of protein binding positions from four independent experiments (n_total = 496 excluding end binding). (Inset) Specifically binding positions of large protein complexes >500 nm³ (n_total = 125) from the red box species in E. Position distributions of other volume species are presented in SI Appendix, Fig. S9.

We indeed observed UvrA and ATL cotranslocating on DNA in fluorescence-optical tweezers experiments with QD-labeled ATL and UvrA (QD525-ATL and QD705-UvrA, blue and red in Fig. 4D, respectively; see also SI Appendix, Fig. S8). The large separation in emission wavelengths of these QD species (525 nm and 705 nm for QD525 and QD705, respectively) allowed us to unambiguously identify the two different proteins ATL and UvrA in the kymograph traces. Under the conditions of our experiments, UvrA in the absence of ATL displayed little DNA binding, no mobility, and only short traces on the DNA (≤ ∼7 s; SI Appendix, Fig. S8), consistent with previous reports (19). The presence of ATL significantly enhanced UvrA motility (44% of traces displayed movement on DNA, n_total = 48) and lifetime on DNA (56% of traces >10 s, 35% of traces >100 s; SI Appendix, Fig. S8F). Interestingly, similar effects on UvrA motility on DNA have previously been observed for UvrA interactions with the NER helicase UvrB (19). ATL–UvrA complexes showed different modes of translocation, including rapid sliding on the DNA (up to ∼4 μm/s), similar to ATL alone. These data indicate that UvrA is likely not in contact with the DNA during the lesion search process, although diffusion constants were somewhat smaller than for ATL alone, in the range of 0.0006 to 0.7 μm²/s (n = 21, obtained from QD525-ATL and QD705-UvrA kymograph analyses). Preincubation of UvrA and ATL did not further enhance the formation of ATL–UvrA complexes on DNA, indicating that binding of UvrA likely occurs to already DNA-bound ATL (scenario 2 above). Interestingly, in the EMSA data we observe ATL–UvrA complexes on DNA only at an ATL concentration that allows for cooperative cluster formation (≥ ∼2 μM), while at lower ATL concentration (≤∼500 nM) UvrA and ATL are seen bound to the DNA individually (arrows in Fig. 4A). This may implicate oligomeric ATL (clusters) in NER recruitment to alkyl lesions. Consistently, our fluorescence-optical tweezers experiments also support oligomeric ATL in ATL–UvrA complexes. Fluorescence intensities for traces of QD525-ATL that overlay with QD705-UvrA traces (potential ATL–UvrA complexes) were ∼30 to 50 a.u., compared to ∼10 a.u. for weakest-intensity traces for QD525-ATL alone that we interpret as monomeric ATL (SI Appendix, Fig. S8E), as described above for QD605-ATL.

To directly characterize ATL–UvrA complexes on DNA at the molecular level we used AFM analysis (Fig. 4 E–G). We measured the volumes of peaks on the alkyl lesion-containing DNA (at ∼30% DNA length) for mixed samples of UvrA and ATL (500 nM and 2 μM, respectively). Samples of the individual proteins incubated with the DNA displayed volumes for ATL that were similar to those at 1 μM (100 to 200 nm³; SI Appendix, Fig. S2E) and for UvrA ∼400 nm³ (SI Appendix, Fig. S2F), consistent with dimeric UvrA bound to DNA, as previously reported (20). Incubations of both ATL and UvrA with the alkyl lesion DNA showed larger volumes (∼500 to 1,000 nm³; red box in Fig. 4F), supporting the presence of heteromeric ATL–UvrA complexes. Importantly, these larger complexes were bound both at the lesion position as well as at nonspecific DNA sites (black and gray bars in Fig. 4F, respectively), consistent with our EMSA and fluorescence-optical tweezers data. Inspection of the binding positions of protein complexes on the DNA substrate still revealed preferential binding to the alkyl lesion (arrow in Fig. 4G), especially by large species (>500 nm³; Fig. 4 G, Inset and see SI Appendix, Fig. S9 for other volume species). Gaussian analysis indicated enhanced positional flexibility (Gaussian width >9% DNA length for the peak at the lesion position for ATL–UvrA compared to 5% for ATL in the absence of UvrA), consistent with higher dynamics of ATL–UvrA complexes. However, the apparently reduced specificity for the lesion (S ∼450 for ATL–UvrA samples compared to S ∼600 for ATL alone) is likely due to enhanced background binding to nonspecific DNA by ATL–UvrA complexes and ATL at high concentration. In addition, UvrA alone is known to possess strong affinity for undamaged DNA (21), hence adding to nonspecific background binding. To investigate preferential localization of UvrA at the alkyl lesion in the absence and presence of ATL, we used QD-labeled UvrA. QDs produce distinct, high features in the topographical AFM images that allow for unambiguous identification of the labeled UvrA molecules within DNA-bound complexes. These experiments also support preferential binding at the lesion position by UvrA in the presence of ATL, while UvrA in the absence of ATL showed no recognition of the alkyl lesion (SI Appendix, Fig. S10). In summary, these experiments strongly support recruitment of UvrA to alkyl lesions by ATL that involves cotranslocation on DNA of preformed ATL–UvrA complexes and lesion recognition by ATL.

Discussion

ATLs display high structural similarity with the DNA-binding and lesion-processing C-terminal domain of alkylguanine DNA alkyltransferases (AGTs). However, they do not possess any DNA lesion repair activity, as they lack the catalytic cysteine of AGTs. AGTs play an important role in the direct damage reversal of highly mutagenic and cytotoxic alkylated guanines. In particular, hAGT has been the focus of considerable research interest, and recent efforts have been dedicated to the development of hAGT inhibitors to counteract its repair of alkylation damage deliberately introduced during chemotherapy. In contrast to AGT, the catalytically inactive ATLs have been proposed to target alkyl lesions to the NER pathway (6, 10). Here, we present single-molecule structural and functional analyses of ATL interactions to investigate the mechanisms of target site search by ATL and AGT, based on structural similarities, and ATL-induced recruitment of NER to DNA alkyl lesions.

AGTs from human (hAGT) as well as bacterial species have previously been shown to form clusters on nonspecific DNA during their target lesion search (11, 13, 22). AGT clusters are stabilized by electrostatic interactions that mediate cooperativity between monomer subunits in a helical array around the DNA (13, 23). Similarly as for AGT, our AFM, AUC, and EMSA data demonstrate cooperative cluster formation for ATL at high (≥2 μM) protein concentrations on nonspecific DNA (during lesion search), consistent with previous EMSA and sedimentation data (6). In hAGT, cooperative interactions are mediated by the N-terminal domain (13) that is absent in ATLs. A strong candidate site for protein–protein interactions in ATL is the C-terminal extension that is not present in hAGT. Interestingly, also for AGT from Mycobacterium, cooperative interactions were suggested to be mediated via a C-terminal extension that is absent in hAGT (22). A conserved approach of cooperative protein clusters that involves distinct structural elements in different AGTs/ATLs indicates the importance of such a mechanism in lesion search by the proteins. In our model (Fig. 2D), the tip of the N-terminal helix of one DNA-bound ATL comes into close proximity to the C-terminal extension of its direct neighbor. The putative contact area is rather electrostatic, as also seen in hAGT cooperative clusters (13). To understand the nature of these interactions in detail, further structural and mutational analyses and/or MD simulations will be required.

The estimation of the stoichiometry of these ATL clusters by AFM volume analyses revealed limiting cluster sizes of approximately four monomer subunits of ATL per cluster. For hAGT, limitation in cluster sizes to about seven subunits has been explained by a balance between the energetic cost of DNA untwisting upon addition of each subunit to the DNA-bound cluster and the energy gained from cooperative protein–protein interactions (11). The limited cluster size is thought to enhance the efficiency of lesion search by AGT through a repositioning mechanism that is based on preferential binding to the 5′ and dissociation from the 3′ end of the cluster (2). Similar to AGT, the size of ATL clusters may be regulated by constraints within protein–DNA interactions leading to energetic costs that eventually overrule the energy gained from cooperative protein–protein interactions. Oligomerization of ATL likely involves weaker subunit interactions than AGT, as manifested by smaller cluster sizes for ATL and the need to employ cross-linking in our AFM depositions to stabilize ATL clusters [in contrast to AGT (11)].

In contrast to the oligomeric search complexes (clusters), our EMSA data show monomeric DNA-bound ATL in the presence of an O⁶-methylguanine alkyl lesion (Fig. 1), consistent with previous EMSA studies (6) and with crystallographic data (6) (SI Appendix, Fig. S1). In our AFM analyses, monomeric ATL bound at its target lesion shows an equilibrium between two conformational states characterized by mild (∼25°) and stronger DNA bending (∼50°), which we interpret as search complex (SC) and interrogation complex (IC) conformations, respectively, as similarly described for glycosylases that initiate base excision repair (BER). Crystal structures of lesion-bound ATL show the alkylated base flipped into a binding site pocket of the protein and the DNA bent by ∼45° (6), consistent with our IC conformation. Comparison of crystal structures of apo ATL and ATL bound to an alkyl lesion in DNA (SI Appendix, Fig. S1B) reveals a conformational switch from an open to a closed, compact state of the protein with the binding site loop shifting toward the binding site and away from the C-terminal loop in the lesion-bound complex (6). This closed conformation conceivably correlates with enhanced bending of the DNA (∼50°, IC) in our AFM images, while complexes that bend the DNA less strongly likely reflect the open conformation (SC).

Base flipping into a binding pocket where protein contacts stabilize the complex for slow nucleolytic attack is famously known for BER glycosylases. Interestingly, however, in contrast to glycosylases, which continuously interrogate the DNA for lesions and thus display an equilibrium between SC and IC also during their target search (24), ATL shows exclusively the SC conformation on nonspecific DNA. The conformational switch from the open to the closed conformation has been suggested to be triggered by DNA binding (6). However, the absence of the IC in nonspecific ATL complexes in our AFM experiments implies that switching from the open to the closed conformation occurs after initial lesion detection in the loosely clamped, open SC conformation. Interestingly, for AGT, MD simulations have shown that fluctuations in the binding site loop are a requirement for base flipping into the catalytic pocket (25). We speculate that base flipping is directly coupled with binding site loop displacement in ATL, which stabilizes the bent DNA conformation in the lesion-bound complex. At nonspecific DNA sites, the IC conformation appears to be less stable, likely due to missing interactions of the flipped base with residues in the lesion recognition pocket, and requires cross-linking to persist through the AFM sample deposition step (SI Appendix, Fig. S4). Search complexes (including clusters) of ATL thus likely adopt the more loosely DNA-bound, open SC conformation to enable highly mobile scanning of the DNA for target sites as seen in our fluorescence-optical tweezers experiments.

Interestingly, the high diffusion constants obtained from our fluorescence-optical tweezers studies for ATL clusters on DNA (up to ∼2 μm²/s, average D of 0.29 ± 0.42 μm²/s) are inconsistent with rotational sliding of clusters along DNA grooves [D_1D,lim ∼0.035 μm²/s for monomeric QD–ATL and yet smaller for oligomers (26); Materials and Methods]. Such high diffusivity is often associated with facilitated diffusion via hopping (microdissociation events), which is enhanced at high salt concentrations (27). In our experiments, we do not observe hopping events, where ATL clusters would be able to transverse an obstacle posed by a second DNA-bound ATL cluster or monomer. Furthermore, we find high D values at a relatively low salt concentration (25 mM sodium acetate). Our data hence support smooth bidirectional sliding in an open, loosely bound complex conformation, consistent with the small DNA bend angles (∼25°) observed at nonspecifically bound ATL clusters by AFM. It is tempting to hypothesize based on their high structural similarity and similar search cluster approach that AGT, like ATL, adopts a loosely bound, highly motile SC conformation during lesion search. Fast sliding on DNA in an open SC conformation has also previously been proposed for other proteins, for instance the human uracil DNA glycosylase (hUNG) and the transcription factor Egr-1 (27–29). In addition, high diffusivity of ATL clusters may be supported by a novel type of facilitated diffusion mechanism based on enhanced association at one end of the cluster and dissociation at the other, as proposed for AGT (2). Kymograph analyses from fluorescence-optical tweezers experiments further revealed alpha factors of ∼0.55 for ATL clusters, indicating subdiffusive motion, in which the protein clusters test for the presence of DNA lesions. Such lesion probing is likely based on mechanical lesion properties (17, 30), facilitated by the mild bending of the DNA in the search clusters. The more strongly bent DNA conformation, in which the protein attempts to flip a target base into its binding pocket is likely immobile and serves to interrogate the DNA lesion. Consistently, our fluorescence-optical tweezers data show frequent switching between a highly mobile mode and a paused complex (e.g., Fig. 3A). These data also indicate that this bent DNA complex conformation is unstable at undamaged DNA sites. Final anchoring of ATL at a lesion involves strong, stable DNA bending in the IC conformation, which may drive the dissociation of the clusters due to steric constraints. This model reconciles the oligomeric search complexes on nonspecific DNA with the monomeric complex at a DNA lesion, as observed for ATL and AGT.

Stronger bending of alkyl lesion DNA by ATL [∼45° (6, 14) from crystal structures and ∼50° reported here] compared to AGT [15 to 30° in crystal structures (2, 7)] is supported by DNA interactions by the C- and N-terminal helices of ATL that are absent in AGT (SI Appendix, Fig. S1A and refs. 2, 6). It has previously been suggested that the high-affinity, strongly distorting binding of ATL to alkyl lesions may serve as a marker, so that the lesion is recognized and processed by the NER pathway that targets bulky DNA adducts (6, 31), reminiscent of UV-lesion marking for NER by the DDB protein complex (9). In principle, both UV and alkyl lesions can also be excised by NER in the absence of marking by DDB or ATL, albeit only very inefficiently (32, 33). The C-terminal loop of ATL, which is rendered highly accessible in the closed conformation of lesion-bound ATL (SI Appendix, Fig. S1B), has been suggested to interact with NER proteins to recruit them to the lesion (6). Intriguingly, our data reveal interactions between ATL and the initiating enzyme of prokaryotic NER, UvrA, at an alkyl lesion (closed conformation), but also on nonspecific DNA (open conformation), as well as in the absence of DNA (apo form, open conformation). As shown here for ATL lesion search clusters, open conformation (monomeric or oligomeric) ATL in complex with UvrA is able to translocate rapidly on nonspecific DNA to deliver UvrA to alkyl lesion sites.

We have visualized our findings in a summarizing model in Fig. 5. It is worth noting that, although in our model ATL is depicted as part of the complex with UvrA/B/C, this may not be true for the short-chain (methyl) alkyl-guanine lesion employed in our studies. A previous study showed weaker binding of ATL from S. pombe (Atl1) to methyl-guanine compared to larger alkyl-chain lesions, suggesting that ATL may be displaced from methyl-guanine lesions by NER proteins during global genome NER (10). Also, in “traditional” NER, ATP hydrolysis induces dissociation of UvrA from the complex upon UvrB interactions with a target lesion (34). It will be interesting to resolve the formation and dynamics of such ATL–UvrA and ATL–UvrA/B/C complexes in future studies.

Fig. 5.

Model of ATL function in NER activation. Our data suggest that ATL searches for alkyl lesions in an oligomeric assembly (based on apo ATL1 monomer, PDB ID code 3gva) that scans the DNA. Rapid, bidirectional sliding (black arrows) likely requires the protein to be in the open conformation (individual monomers in different shades of light green), consistent with only mild DNA bending observed in our AFM analyses (∼25°). Large diffusion coefficients of ATL clusters on DNA may be supported by preferential monomer association at the 5′ and dissociation from the 3′ end of search clusters (gray arrows), as previously proposed for AGT (2), the structurally similar, catalytically active homolog of ATL. Recruitment of the NER pathway for alkyl-lesion repair then involves one of two possible scenarios. (Left) At a lesion, conformational changes in ATL lead to the dissociation of the cluster and the formation of a stable, monomeric, strongly bent complex with the target base flipped into a binding pocket of ATL (dark green, based on PDB ID code 3gyh). As previously proposed (6), this complex then recruits UvrA, the initiating enzyme of prokaryotic NER (UvrA dimer blue, based on PDB ID code 2r6f). (Right) UvrA interacts either with DNA-bound ATL or with ATL prior to DNA binding (in the open conformation, light green) and the (oligomeric) ATL–UvrA complex then comigrates along the DNA in search of a target lesion. At a lesion, additional subunits dissociate due to conformational changes in the complex, as above, leaving a stably bound ATL–UvrA complex at the lesion. In a common final step, UvrA then initiates NER, in which the NER endonuclease UvrC (dark blue) excises a stretch of the DNA strand containing the alkyl lesion by 3′ and 5′ incisions (only the first, 3′ incision is shown, blue flash) (35). Although UvrA is shown here in complex with UvrB (light blue, based on PDB ID code 1d9x) and the NER endonuclease UvrC (dark blue), in “traditional” NER UvrA dissociates from the complex once UvrB forms the specific preincision complex at the lesion, which recruits UvrC. The exact order of events in ATL-induced NER alkyl lesion repair requires further investigation.

Finally, previous studies showed that expression of E. coli ATL protects mammalian cells against cytotoxicity of methyl-guanine lesions (36). However, protection was conveyed predominantly via shielding of alkyl-guanine–induced alkylG:T mispairs from futile rounds of DNA mismatch repair that result in DNA double strand break formation, as also shown in E. coli (36, 37). ATL did not initiate NER of methyl-guanine in mammalian cells in these studies (36). Consistently, in E. coli enhanced NER by ATL could only be observed for larger alkyl lesions and not for methyl-guanine (37, 38). In contrast, studies in S. pombe, which does not possess an active alkyltransferase, indicated a role of Atl1 in global genome NER of methyl-guanine (10). More stable binding of Atl1 to larger alkyl lesions directed these to repair by the transcription-coupled repair subpathway of NER (TC-NER) (10), which is initiated by stalling of RNA polymerase at the ATL-lesion complex. Interestingly, in future structural analyses of the eukaryotic NER complex, ATL together with larger alkyl-guanine lesions may provide an attractive experimental tool to block transcription and thus activate xeroderma pigmentosum group D protein in the RNA polymerase–TFIIH preinitiation complex for TC-NER (10, 39).

Materials and Methods

Reagents.

Agar and LB medium were obtained from Carl Roth AG. All other biochemicals were from Sigma-Aldrich.

Protein Preparation.

ATL is encoded in E. coli by the ybaZ gene. The ybaZ gene was cloned into the pETM41 vector, resulting in a (His)₆-tag followed by MBP on the N terminus of full-length eATL (MBP-ATL, total M_r 57.5 kDa, 15.5 kDa + 42 kDa). To exclude interference by the comparably large MBP tag, E. coli ATL was further cloned into the pCDF vector as a N-terminally His₆-tagged truncated construct (ATLΔN, amino acids 30 to 129, 11.0 kDa). This construct lacks the largely unstructured N-terminal region that is unique to E. coli ATL, not present in any of the crystallized orthologs from other bacteria or yeast (6, 31, 40), and likely not involved in the well-conserved DNA interactions. For optional specific labeling of the protein with organic fluorophores via maleimide conjugation chemistry, the N-terminal aspartic acid (D31) of this construct was mutated to cysteine. All primers used for cloning are listed in SI Appendix, Table S1. ATLΔN and MBP-ATL were expressed in E. coli BL21 cells and purified via Ni²⁺-NTA and amylose affinity chromatography, respectively, followed by size-exclusion chromatography with an SD75 16/600 column similar as previously described (41). Proteins were purified and stored in ATL buffer [20 mM Tris (pH 7.7 at 20 °C), 150 mM NaCl, 5 mM EDTA, 10% glycerol, and 1 mM Tris(2-carboxyethyl)phosphine]. ATLΔN and MBP-ATL were >95% pure as judged from Coommassie stained SDS polyacrylamide gel electrophoresis. Protein solutions were stored frozen at −80 °C. ATL and MBP-ATL concentrations were measured spectrophotometrically using ε₂₈₀ = 2,25 × 10⁴ M^-1cm⁻¹ (ATLΔN) and 9,35 × 10⁴ M^-1cm⁻¹ (MBP-ATL).

Because of its larger size and stronger absorption signals, MBP-ATL was used in AFM imaging and AUC experiments; MBP-ATL was also used in fluorescence-optical tweezers experiments; ATLΔN was used in EMSA studies. We confirmed that the MBP fusion does not affect ATL interactions by EMSA and BLI studies on both protein variants under identical conditions (SI Appendix, Figs. S3 and S7).

UvrA from B. caldotenax was expressed and purified as described (19) from a plasmid, which encodes for a C-terminal avi-tag for biotinylation (SI Appendix). The plasmid was kindly provided by H. Wang, North Carolina State University, Raleigh, NC.

DNA Substrates.

All DNA sequences are listed in SI Appendix, Table S1.

For EMSA and AUC, synthetic 48-nt DNA oligonucleotide containing O⁶-alkylguanine at position 20 (alkG; SI Appendix, Table S1) was obtained from Gene Link and annealed with its complementary strand at equimolar concentrations. For fluorescence detection in EMSAs, the bottom strand in these substrates (bottom in SI Appendix, Table S1) was 5′-terminally labeled with an Alexa Fluor 647 (AF647) fluorophore (Sigma).

The same (5′phosphorylated, nonfluorescently labeled) O⁶-alkylguanine oligonucleotide (alkG) sequence was also used to prepare a long DNA substrate containing the alkyl lesion at a well-defined position (at 31% of total DNA substrate length) for AFM experiments, as previously reported (42) and as described in SI Appendix, Fig. S11A)

Fluorescent Labeling of Proteins.

For fluorescence-optical tweezers experiments, MBP-ATL was conjugated to fluorescent QDs with emission at 605 nm (for experiments on ATL alone) or at 525 nm (for experiments with ATL and UvrA) via an antibody sandwich linker (QD605 and QD525, respectively; Thermo Fisher Scientific; SI Appendix, Fig. S5), similar to what was described for other proteins (43). Details of the conjugation procedure are outlined in SI Appendix, Fig. S5.

The N-terminally introduced cysteine is the sole cysteine in ATLΔN and is easily accessible for fluorophore attachment, allowing for site-specific 1:1 labeling. For microscale thermophoresis (MST) the protein was incubated with a molar excess (10-fold) of NT647 maleimide (Ɛ_650nm = 195,000 M^-1cm⁻¹; NanoTemper). Details of the labeling procedure are described in SI Appendix.

UvrA was biotinylated and fluorescently labeled by conjugation to streptavidin-coated QDs with maximum emission at 525 nm (QD525, Thermo Fisher Scientific, for AFM topography and MST experiments) and 705 nm (QD705, ThermoFisher Scientific, for fluorescence-optical tweezers experiments). For details see SI Appendix.

EMSA.

EMSAs were performed using 6%, 8%, and 10% nondenaturing arylamide gels prepared with 29:1 bis-acrylamide:acrylamide ratio solution. Incubation of samples was carried out in incubation buffer (25 mM Hepes [pH 7.5 at 20 °C]), 25 mM sodium acetate, and 10 mM magnesium acetate) for 30 min at ambient temperature. To study cooperative DNA binding by ATL, 50 nM of either alkyl lesion containing or undamaged 48-bp DNA substrates were incubated with increasing concentrations of either MBP-ATL or ATLΔN (100 nM to 10 μM). To study ATL–UvrA interaction, 50 nM of lesion or undamaged DNA were incubated with 500 nM or 2 μM ATL (with and without MBP tag) and 200 nM or 500 nM UvrA. Samples (10 μL) were mixed with Orange G loading dye (3 μL) and loaded onto the gel. Gels were prerun for 10 min at 100 V before loading the incubated samples and then run at 100 V for 60 to 120 min depending on gel concentration. Alexa Fluor 647 (AF647) on the DNA substrate provided the fluororescence signal for detection using a PhosphoImager (Pharos; BioRad).

SV AUC.

To study ATL interactions with alkyl lesion containing and undamaged DNA, increasing concentrations of MBP-ATL (500 nM to 20 μM) were mixed with 200 nM DNA substrate (with or without an alkyl lesion) in ATL incubation buffer (discussed above). To analyze UvrA recruitment to alkyl lesions, alkyl lesion containing DNA substrate (200 nM) was mixed with either 2 μM MBP-ATL, 500 nM UvrA, or 2 μM MBP-ATL together with 500 nM UvrA in ATL incubation buffer. Sample volumes of 385 μL were loaded in standard double-sector charcoal-filled Epon centerpieces equipped with sapphire windows and run (versus a blank of incubation buffer) in a Beckman Optima XL-I analytical ultracentrifuge (Beckman Coulter) with an eight-hole An-50 Ti rotor at 40,000 rpm and 20 °C until completely sedimented. Data were collected via absorption detection at a wavelength of 280 nm in continuous mode at a step size of 0.003 cm. Data were analyzed using the NIH software SEDFIT to determine continuous distributions for solutions to the Lamm equation c(s), as previously described (44). Analysis was performed with regularization at confidence levels of 0.68 and floating frictional ratio, time-independent noise, baseline, and meniscus position, to rmsd values of < 0.014. All experiments at <10 μM MBP-ATL were carried out in duplicate providing reproducible sedimentation coefficients. Experiments at 10 μM and 20 μM MBP-ATL showed comparable saturating sedimentation coefficients with both alkyl lesion and nonspecific DNA substrate.

AFM Experiments.

ATL–DNA samples were incubated at low (1 μM) or high (10 to 20 μM) concentrations of ATL and DNA concentrations of 50 to 100 nM in ATL buffer (discussed above). To facilitate detection, MBP-ATL fusion protein was used in all AFM experiments (discussed above). The DNAs used in these experiments were linear 916-bp fragments containing an alkylated guanine at 31% of the DNA substrate length (alkG-DNA) or nonspecific DNA substrate (that did not contain any damage sites, ∼2,000 bp), as indicated (see SI Appendix, Fig. S11 for details). Mixed samples of ATL and UvrA were incubated at intermediate (2 μM) MBP-ATL and 500 nM UvrA concentrations with 100 nM alkyl lesion containing (916 bp) DNA substrate. All incubations were carried out for 30 min at ambient temperature in incubation buffer (EMSA). To avoid dissociation of multimeric, cooperative protein clusters on the DNA during dilution, samples were additionally cross-linked (where indicated) for 10 min in 0.1% glutaraldehyde at ambient temperature. For deposition, samples were diluted 50- to 200-fold (depending on protein concentration) in the same buffer. Immediately after dilution (≤15 s), 20-μL volumes were deposited on freshly cleaved mica (grade V; SPI Supplies), rinsed with purified deionized water, dried in a stream of nitrogen, and imaged using a Molecular Force Probe 3D-Bio AFM (Asylum Research). Data acquisition was carried out in oscillating mode using OMCL-AC240TS (Olympus) noncontact/tapping mode silicon probes with spring constants of ∼2 N/m and resonance frequencies of ∼70 kHz. Images were captured at scan speeds of 2.5 μm/s and pixel resolution of 1.95 nm per pixel. All experiments were carried out at least in triplicate.

Binding specificities (S) to the alkyl lesion at 31% of the alkG-DNA substrate were derived from the ratio of total occupancy for the specific (alkylG) site (A_spec, area under Gaussian fit) and that of nonspecific DNA sites (i.e., all other positions on the DNA, A_nsp, area of background from 2 to 50% DNA length) using Eq. 1 (45):

$$S=N\cdot \frac{A_{spec}}{A_{nsp}}+1.$$ [1]

N is the number of total available binding sites (number of base pairs, N = 916 bp − 2 × 2% at ends = 879 bp). DNA bend angles were determined as 180° − β, where β is the angle between line segments drawn on the DNA contour on both sides of a protein peak (or of the lesion position in control experiments in the absence of protein). AFM volumes of DNA bound protein peaks were determined from their measured average heights and base areas (as the volume of a deformed cylinder) using the software ImageSXM (S. Barrett, University of Liverpool, Liverpool, UK). Volume distributions were plotted and fitted by (single or multiple) Gaussians to obtain dominant peak volumes from the centers of the Gaussian curves using Origin. The AFM volumes of protein peaks can be translated into approximate molecular weights (45). For the AFM system employed here, the relationship between protein peak volume V and molecular weight MW is given by a previously established calibration curve (16):

$$\text{MW}=\left(V+5.9\right)/1.2.$$ [2]

For details on AFM data analyses see SI Appendix.

Molecular Modeling.

A representative manual model of two ATL molecules on DNA (Fig. 2D) was generated based on the crystal structures of ATL from S. pombe bound to alkyl lesion-containing DNA (PDB ID code 3gyh) (6), apo ATL from S. pombe (PDB ID code 3gva) (6), and a 15-bp DNA fragment of ∼25° bend angle (extracted from PDB ID code 4yex) (46). We first superposed two copies of ATL–DNA complexes from 3gyh with undamaged DNA extracted from 4yex, such that the DNA molecules aligned around the lesion site in the DNA in 3gyh and the spacing between the two ATL molecules was 7 bp along the undamaged DNA. Subsequently, we replaced the structure of each of the two ATL molecules with that of apo ATL (by superposition of two copies of 3gva with the two molecules derived from 3gyh). All superpositions and the illustration were done in PyMOL Molecular Graphics System, Version 2.0 (Schrödinger LLC).

While we chose a DNA structure for modeling that was bent by approximately the same degree as measured in our AFM analyses (∼25°), the exact angle in nonspecifically bound ATL–DNA complexes may differ slightly. It is also worth noting that the DNA used for modeling lacks the slight widening of the minor groove introduced by ATL binding. Our model hence contains minor clashes that presumably result from the crude nature of the modeling procedure. Note that the model is only meant to visualize how ATL molecules may get into proximity and cross-talk with each other without making any further structural claims.

Fluorescence-Optical Tweezers.

Experiments were carried out on the commercially available C-Trap system integrating optical tweezers, confocal fluorescence microscopy, and microfluidics (LUMICKS B.V.). MBP-ATL was fluorescently labeled with QD605 or QD525 QDs (discussed above). In experiments with ATL only (no UvrA), the QD605-ATL conjugate was employed. QD525-ATL was used in experiments with UvrA and ATL to allow for unambiguous spectral separation of QD705 (emission at 705 nm) on UvrA and QD525 (emission at 525 nm) on ATL. To study the interactions between ATL (or ATL and UvrA) and DNA, bacteriophage λ-DNA (48,502 bp) was functionalized at either end with biotin as previously described (47) and was tethered between two 4.4-μm streptavidin-coated polystyrene beads (Spherotech Inc.) using a laminar flow cell (LUMICKS B.V.). In the laminar multichannel flow cell, the beads were put in channel 1, the DNA in channel 2, ATL buffer for measurements in channel 3, and the ATL protein sample in channel 4. In addition, in experiments with UvrA and ATL, UvrA was added either in a fifth channel, or mixed in channel 4 with ATL (concentrations are discussed below). After formation of a DNA tether between two beads, the tether was suspended in the protein channel for protein–DNA binding and subsequently transferred to the buffer channel for measurements. MBP-ATL was either applied at low (80 nM) or at high (2 μM) concentration. For experiments at 80 nM, the DNA tether was additionally preincubated with SYTOX Orange Nucleic Acid Stain (in the fifth channel) to verify the position of the DNA tether for kymograph recording. To allow for single-molecule resolution at high protein concentration (2 μM), only every eighth protein molecule was fluorescently labeled (discussed above). For experiments with UvrA, MBP/ATL at 2 μM and UvrA at 10 nM were applied in separate channels (channels 4 and 5). Alternatively, MBP/ATL and UvrA were premixed in channel 4 at 1 μM and 5 nM. Interactions between QD605-/MBP-ATL, or QD525/MBP-ATL and QD705/UvrA and the DNA were recorded in ATL incubation buffer (discussed above) containing 1× BSA. In experiments with UvrA, 1 mM ATP was included with the incubation buffer to enable potentially required conformational changes in UvrA. QD605, QD525, and QD705 fluorescence was excited by a 488-nm laser and emission detected by green filter (582/75nm), blue filter (512/25nm), and red filter (700/100nm), respectively. Emission signals were recorded using BlueLake software (LUMICKS B.V.). Lower intensities of QD525 (in the blue detection channel) compared to QD605 (in the green channel) are due to different intensities of the different QD species and the narrower detection bandwidth of the blue channel.

From the recorded kymographs, distances covered and local speeds of ATL on the DNA were measured directly in Fiji. Diffusion constants D and diffusive exponents α were determined by Gaussian tracing in ImageJ and MSD analysis using a custom routine in Labview software, as previously described (48). Based on previous analyses of hydrodynamic radii of QDs with similar surface functionalization [∼9 nm for QD-streptavidin (49)], we conservatively estimate the radii of the QD605-antibody conjugate used in our experiments as 9 nm. Together with the hydrodynamic radii of MBP [∼3.5 nm (50)] and ATL (∼1.8 nm based on sequence), this results in an effective hydrodynamic radius of R ∼ 9.2 nm for the QD605-MBP-ATL conjugate. For the viscosity of the buffer used in our experiments (ɳ = 1.0296 × 10⁻³ Ns/m², determined with Sednterp [T. Laue]), the BBX (Bagchi/Blainey/Xie) model (26) thus gives a diffusion limit for rotational sliding along the DNA double helix groove of ∼0.04 μm²/s for monomeric QD-MBP-ATL, and ∼0.01 μm²/s (R∼13.3 nm) and 0.008 μm²/s (R ∼ 14.6 nm) for trimeric and tetrameric complexes, respectively.

BLI.

BLI biosensors coated with anti-6xhistidine tag antibody (ForteBio) were loaded with his tagged ATL or MBP-ATL (both at 50 nM) and association and dissociation of increasing concentrations (10 to 1,000 nM) of UvrA were detected. Measurements were performed in ATL buffer (discussed above) containing 1× BSA using the Octet Red system (ForteBio) at ambient temperature at 1,000 rpm with incubation times of 2,000 s in UvrA solutions (for association) and 2,000 s in buffer (for dissociation). Experiments were performed in duplicate using proteins from two different purifications. Response curves displayed limitation by mass transport (slow dissociation compared to association) as well as biphasic association and dissociation behavior. Approximate K_ds were therefore determined by equilibrium analysis from the maximum signals at increasing UvrA concentrations using Hill fits.

Microscale Thermophoresis.

For MST measurements, ATLΔN (without MBP tag) was fluorescently labeled with NT647 (SI Appendix). NT647-labeled ATL was mixed at 100 nM with increasing concentrations of (nonlabeled) UvrA. In addition, QD-UvrA (5 nM) was mixed with increasing concentrations of MBP-ATL. The samples were incubated in ATL buffer containing 1× BSA for 30 min at ambient temperature and then filled into standard (for the QD-UvrA/MBP-ATL samples; NanoTemper) or hydrophobic capillaries (for the ATLΔN-NT647/UvrA samples; NanoTemper). Measurements were carried out in a Monolith NT.115 system (NanoTemper) at ambient temperature with MST laser powers of 40% and 60% and light-emitting diode (fluorescence excitation) powers of 40 to 80%. Thermophoresis measures changes in protein complex diffusion out of a laser focus due to protein–protein interactions. However, for both systems (fluorescently labeled ATL as well as QD-UvrA) the thermophoresis signal was dominated by a fluorescence increase at higher concentrations of the interaction partner. The fluorescence increase curves upon excitation with the red laser for ATL-NT647 and with the blue laser for QD-UvrA) was exploited to determine approximate K_d values of the UvrA-ATL interaction. Experiments were performed in duplicate.

Data Availability.

The data supporting the findings of this study are available within the main text and SI Appendix. In addition, all original AFM images and fluorescence kymograph data are available on the Open Science Framework at https://osf.io/673aq/ (51).