Single molecule studies of DNA repair

Background: DNA repair

DNA is damaged continuously by agents that occur naturally within our cells as well as by exogenous factors such as high-energy radiation or alkylating agents during chemotherapy. If unrepaired, the resulting DNA base modifications and strand breaks can lead to mutations in transcribed genes and tumour development as well as cell death. DNA repair is hence essential for the maintenance of genomic stability and cellular viability. Cells have developed a variety of DNA repair mechanisms, which are each specialised to target specific damages in the genetic code. We are interested in resolving and understanding the remarkably different strategies realised by different DNA repair systems to find and recognise their target sites in DNA, and in the intricate interplay between different repair systems. Specifically, we currently work on damage recognition in base excision repair (BER), nucleotide excision repair (NER), and by the direct damage reversal DNA alkyltransferase (AGT).


Single molecule imaging by Atomic Force Microscopy (AFM) enables us to directly visualize and structurally characterize protein- and protein-DNA complexes at the level of the individual molecules. In addition, we use single molecule fluorescence microscopy coupled with an optical tweezers system to investigate dynamic processes on DNA in DNA lesion search and recognition. We further complement these studies with biochemical and biophysical ensemble techniques to address functional properties of the involved interactions.

Research foci:

1. Mechanical sensing in DNA lesion search and recognition by BER glycosylases.

Subtle chemical alterations of DNA bases such as oxidation or alkylation can cause transition mutations in the DNA and thus lead to genetic instability. These lesions are typically targeted and repaired by the base excision repair (BER) (Figure 1).

Figure 1

Figure 1: The BER pathway. Lesion recognition is the first step in BER and is carried out by a number of DNA glycosylases, which excise the lesion base and leave an abasic site in the DNA. Bifunctional glycosylases also host DNA lyase activity, while monofunctional glycosylases require the action of the apurinic/apyrimidinic endonuclease (APE1) to incise the DNA backbone at the abasic site. APE1 also processes the resulting DNA ends (dRP deoxyribose phosphate, P phosphate, PUA 3′-phosphor-α,β-unsaturated aldehyde). A DNA polymerase then re-fills the missing base in single nucleotide BER (SN-BER), or a strand displacing polymerase re-synthesizes a stretch of DNA bases in long patch BER (LP-BER). Finally, resulting DNA overhangs are processed by flap endonuclease FEN1 in LP-BER, and gaps between new and original bases are sealed by DNA ligase.

Glycosylases, the initiating enzymes of BER, employ a base flipping mechanism to identify their target inside a catalytic pocket. The specific approach for recognition of DNA flaws and damages, however, differs subtly between different glycosylases. There are a number of different glycosylases, 11 in humans, that each recognize and repair only one or only a few types of chemical base modifications. To enhance the efficiency of the lesion search process, glycosylases seem to exploit the mechanical properties at their specific target lesions in an initial lesion sensing step (Figure 2). [1,2]

Figure 2

Figure 2: AFM analyses of DNA bending by BER glycosylases. (A) Automated analysis of AFM imaging data for high throughput, accurate detection of DNA bend angles was developed in collaboration with the Heinze lab at the Virchow Center. The approach was applied to the structural characterization of protein-DNA complexes of a number of different glycosylases as well as their target lesions in the absence of protein. Protein peaks on the DNA are detected based on their enhanced heights, and the degree of DNA bending at these protein peaks is then measured automatically using tangent overlay. The insets show the skeletonisation of the DNA backbone for automated analysis (top) and an exemplary DNA bend angle distribution for the glycosylase MutY with the ∼15° and 50° bend angle species representing the search and lesion interrogation complex conformations, respectively (bottom). (B) Fluorescence resonance energy transfer (FRET) measurements and simulations corroborated the resulting bend angle distributions from automated AFM analyses. (C) These data support a model of initial lesion sensing by BER glycosylases that is based on the mechanical properties of their specific target lesions and the different energetic requirements for DNA bending at these lesion sites. [1]

2. Transcription regulation by the BER glycosylase hOGG1

Interestingly, the BER glycosylase hOGG1 (human oxoguanine glycosylase 1) has also been observed to function in the regulation of gene transcription under oxidative stress, in addition to its generic function in the repair of oxidative (oxoguanine) DNA lesions. We were recently able to resolve the molecular mechanism of recruitment of the oncogene transcription factor Myc by hOGG1. Using AFM in combination with biochemical approaches, we showed oxidation induced dimerization of hOGG1, enhanced interactions of the oxidized hOGG1 dimer with Myc (compared to monomeric hOGG1 under reducing conditions), inactivation of hOGG1 catalytic activity by Myc, and enhanced recruitment of Myc/Max to its E-box recognition sequence in proximity to oxidative lesions in DNA by hOGG1 specifically under oxidative stress conditions (Figure 3). [3]


Figure 3: Direct hOGG1-Myc interactions inhibit hOGG1 catalytic activity and recruit Myc to its promoters under oxidative stress. AFM analyses showed depletion of the base excision competent hOGG1 conformation in complex with Myc (yellow arrow, top left). Consistently, activity assays demonstrated complete lack of oxoguanine repair by hOGG1 in the presence of Myc (bottom left, black arrow indicates repair product). AFM analyses further revealed formation of a new conformation induced by hOGG1-Myc interactions (red arrow), which is characterized by ~20° DNA bending. Catalytic inactivation of hOGG1 by Myc likely leads to prolonged binding of hOGG1-Myc complexes to enable loading of Myc onto the DNA by hOGG1 at the lesion. Consistent with enhanced affinity between dimeric hOGG1 and Myc under oxidizing conditions, AFM studies (using quantum dot labeled Myc for specific identification in the complexes) demonstrated enhanced recruitment of Myc/Max to its E-box recognition sequence by hOGG1 specifically under oxidative stress (middle). The model on the right summarizes these findings. [3]

3. Conservation and divergence in NER lesion recognition.

In the highly conserved mechanism of NER, which targets a multitude of different DNA damages including UV irradiation damages, DNA target site recognition involves ATPases/helicases. In general, prokaryotic and eukaryotic nucleotide excision repair (NER) share the same mechanistic approach of excising an oligonucleotide containing the lesion from the DNA (Figure 4) [4-6]. They also display largely comparable target lesion specificities and repair efficiencies. However, to achieve the recognition and identification of their specific target lesions, eukaryotic and prokaryotic NER systems employ subtly different molecular approaches (Figure 5). [4,5]

NER mechanisms

Figure 4: The mechanisms of eukaryotic (A) and prokaryotic (B) NER. NER involves the interplay of many different proteins: XPA-G and p8, p34, p44, p52, and p62, as well as RPA for the eukaryotic and UvrA-D for the prokaryotic system (A=UvrA and B=UvrB in the figure). Two NER sub-pathways exist: transcription coupled NER is initiated by a stalled RNA polymerase while in global genome NER, initial lesion search and recognition is carried out by XPC/ERCC2 and UvrA in eu-and prokaryotes, respectively. The essential step of lesion verification is carried out by the NER helicases XPD and UvrB (marked by black ovals), leading to the subsequent recruitment of the endonucleases (XPF and XPG or UvrC) that excise the lesion containing stretch of ssDNA.

NER lesion recognition

Figure 5: NER lesion recognition by UvrB and XPD helicases. (A) AFM imaging of XPD/p44 or UvrB complexes on DNA substrates containing different types of NER target lesions at ∼30% of the DNA length show stalling at the lesions as enhanced binding at the lesion position (B). An unpaired DNA region (DNA bubble) was further introduced in the DNA substrates for loading of the helicases either 3′ or 5′ to the lesion. The known 5′-to-3′ polarity of these helicases also allowed us to distinguish between lesion recognition on the translocated versus on the non-translocated DNA strand. Interestingly, our data revealed that eukaryotic XPD and prokaryotic UvrB possess different strand preferences in the recognition of NER target lesions. (C) Specifically, a fluorescein adduct mimicking a bulky lesion was preferentially recognized by both XPD and UvrB on the translocated DNA strand (schematic in C, middle). The beta hairpin structure in UvrB that directly interacts with the lesion and the pore in XPD through which the translocated DNA strand is threaded are indicated. The UV product cyclobutane pyrimidine dimer (CPD), on the other hand, was preferentially recognized by UvrB also on the translocated strand, but by XPD on the opposite, non-translocated strand (schematic in C, bottom). In the complex, the lesion on the opposite strand comes in close proximity of an iron sulfur cluster in the XPD protein (indicated by the orange star), which may directly interact with the lesion for target site identification.

4. Cooperative clusters in lesion search by the O6 alkylguanine DNA alkyltransferase AGT

The human O6-alkylguanine DNA alkyltransferase, AGT, recognizes and removes highly mutagenic and cytotoxic alkylation damages in our genetic material. AGT has also recently received particular interest as an inhibitor target to support chemotherapeutical efficiency, because AGT repair activity interferes with alkylation damage that is deliberately introduced into DNA to kill cancer cells.
Direct damage reversal by AGT involves search, recognition, identification, and removal of the alkyl lesion, which are all performed by the same protein and leave the DNA completely intact with the offending chemical group transferred to the protein. Using a combination of AFM imaging and analytical ultracentrifugation (AUC), we were able to confirm an atypical form of cooperative non-specific DNA binding, in which AGT forms protein clusters with a short, limiting length on DNA (Figure 6). [7-9]

These short length AGT clusters have been proposed to serve to enhance the speed and efficiency of DNA lesion search by AGT using a mechanism of facilitated diffusion based on preferential monomer addition at the 5’ end and dissociation from the 3’ end of clusters. Interestingly, however, single molecule fluorescence microscopy coupled with a dual trap optical tweezers system has shown rotational movement of AGT on DNA, but no enhancement of DNA translocation for AGT clusters compared to monomeric complexes (Figure 7A). Instead, these studies revealed preferential stabilization of clusters at an alkyl lesion in DNA, indicating a role of AGT clusters in lesion processing (Figure 7B). [10]

DNA lesion search by AGT

Figure 6: DNA lesion search by AGT. (A) Crystal structure of human AGT bound to an alkyl-lesion in DNA (pdb coordinates 1T38, protein shown in blue, DNA in red). The DNA binding helix-turn-helix motif is shown in darker blue. The alkylated base is flipped into the active site binding pocket of the protein where it is transferred onto a catalytic cysteine residue (C145 in human AGT). An arginine finger (R128) stabilizes the void left by the flipped base. (B) In contrast to the monomeric lesion-bound complex, AGT was proposed to form cooperative protein clusters that wrap helically around the DNA during its lesion search. In these clusters, cooperative contacts exist between each nth and (n+3)rd monomer. Figure based on (Adams et al. 2009 JMB). Top: side view of the DNA, bottom: view along the DNA helical axis. (C) AFM imaging was able to confirm this model of cooperative AGT clusters on non-specific DNA (i.e. during its lesion search). AFM images (top) showed clusters of AGT bound to non-damaged DNA (e.g. white arrow), with limiting cluster length of ∼7 monomers of AGT (bottom). Interestingly, the observed length limitation for AGT is in contrast to cooperative DNA binding in the McGhee-vanHippel model (bottom: grey zone), which predicts cluster lengths for unconstrained addition of monomers. For AGT, energetic costs of DNA untwisting upon addition of each monomer to the cluster may cancel energy gained from cooperative protein-protein interactions, stalling cluster growth and thus limiting cluster lengths to enhance the speed of cluster relocation and thus the efficiency of lesion search by AGT. Symbols: diamonds represent the originally measured cluster lengths, squares the cluster lengths corrected for AFM tip convolution (see also below); diamonds and squares span the range of true cluster lengths.

Single Molecule Studies

Figure 7: Single molecule fluorescence studies of the human DNA alkyltransferase lesion search and repair mechanism. (A) Fluorescence kymographs of quantum dot labeled AGT translocating on DNA (top). Diffusional analyses (bottom) of kymographs showed a rapidly diffusing, short-lived species (light green) and a slowly diffusing, long-lived species (black) with diffusion constants consistent with rotational movement along the DNA minor groove in which AGT binds. Inset: Diffusion constants plotted over the fluorescence intensities of kymograph traces revealed no correlation indicating no enhancement of translocation speed by AGT clusters compared to monomers. (B) Positional analyses of AGT complexes bound to a DNA tether that contained an alkyl lesion at 30% of its length (dark green arrows) demonstrated lesion recognition by AGT. Importantly, higher fluorescence intensities of complexes located at the lesion position (compared to complexes bound elsewhere on the DNA) indicated preferential cluster formation at the lesion (bottom, dark green box). Clusters may serve to stabilize complexes at a lesion and further provide additional monomer subunits for recruitment of proteins, for instance from the replication machinery, to allow for rapid replication restart after successful repair of the highly mutagenic alkyl lesion by AGT. [10]

4. The link between alkyl lesion repair and the NER system by the alkyltransferase-like ATL protein

Alkyltransferase-like proteins (ATLs) form a protein family with high structural similarity to AGTs (Figure 8A). Single molecule AFM and fluorescence optical tweezers data show highly specific recognition of alkylguanine lesions in DNA and rapid scanning of the DNA by ATL in search of these target lesions (Figure 8). Like its alkyl lesion repair active homolog AGT, ATL forms clusters on DNA at high protein concentrations, which scan the DNA for lesions. While ATL itself is unable to process alkyl lesions, it recruits UvrA, the initiating enzyme of prokaryotic NER, to these lesions. ATL thus initiates NER of alkylguanine lesions that are otherwise not (or only poorly) recognized by the NER system (Figure 9). [11]

Alkyltransferase-like protein clusters

Figure 8: Alkyltransferase-like protein clusters scan DNA rapidly over long distances. (A) ATL is structurally highly similar to AGT, as seen in the overlay of the two DNA bound structures (yeast Atl1 in purple, pdb 3gyh; human AGTin cyan, pdb 1t38, DNA is shown in red). The lesion base is flipped into a binding pocket in AGT and ATL. ATL has been previously shown to be in a closed conformation when bound at a target lesion, with the binding site loop shifted towards the binding pocket (purple arrow), and an open conformation when in the apo form. (B) AFM analyses show different DNA bending by ATL at undamaged sites (grey arrows) and at an alkyl lesion (white arrows, bend angle distributions in (C) bottom), consistent with lesion scanning on undamaged DNA in the open conformation, and lesion interrogation in the closed conformation with the DNA more strongly bent. (C) Fluorescence optical tweezers data show ATL clusters (containing multiple monomers of ATL based on fluorescence intensities) on undamaged DNA in search of lesions, similar as seen for AGT (see above). Mean square displacement (MSD) analyses of the kymographs indicate fast but sub-diffusional sliding, suggesting random stalling of ATL to probe for lesions. The closed conformation of these lesion interrogation complexes is unstable at undamaged DNA sites and converts back to rapidly sliding, open complexes (arrow). Open and closed conformations are shown schematically with binding site loop position indicated. The inset shows two optically trapped beads connected by a DNA tether on which the translocating fluorescently labelled proteins can be observed.

ATL recruits NER

Figure 9: ATL recruits NER to alkyl-DNA lesions. (A) Overlay of fluorescence kymographs of UvrA (red fluorescence) and ATL (blue) show co-translocation of UvrA and ATL on DNA. (B) In our model, clusters of ATL (green) scan the DNA rapidly in search of lesions. These clusters can consist of either just ATL or they can contain UvrA (blue), which they thus transport to an alkyl lesion in the DNA. At the lesion, the clusters convert to monomeric ATL (with or without UvrA bound) due to enhanced conformational strain (DNA bending). The UvrA-ATL complex at the lesion then recruits the rest of the NER cascade (marked B and C for the NER helicase UvrB and endonuclease UvrC, respectively) for NER repair of the alkyl lesion.

5. The beauty of AFM and combination with Fluorescence Microscopy

AFM is highly synergistic to other structural techniques such as X-ray crystallography and kinetic studies of DNA repair processes. It is the only imaging platform, which allows the monitoring of protein dynamics without any labeling modification in physiologically relevant conditions. In contrast to crystallography, it does not require homogeneity of the sample molecules allowing it to resolve structural and stoichiometric heterogeneity in a sample. Conformational heterogeneity can, for instance, be an important indicator of intra- or inter-molecular flexibility. Importantly, AFM is one of very few techniques that allows the resolution and structural characterization of non-specifically bound protein-DNA complexes (e.g. DNA lesion search complexes). Although AFM does not achieve the high atomic resolution of crystallography, its spatial resolution at the molecular or sub-molecular level can nevertheless often resolve individual protein domains and provide valuable structural information on proteins and protein complexes. [12-14]

AFM topographic images contain 3D information. We can hence measure the volumes of sample particles in the images to estimate their molecular masses and, for instance, distinguish between different oligomeric states of a protein. [12-14]

It is important to bear in mind that AFM images constitute convolution images of the true sample topography and the AFM tip geometry. The contribution of the AFM tip to image features can, however, be calculated and subtracted post-experimentally when the approximate end diameter of the AFM tip is known. Using simple geometrical models to describe sample particles and AFM tip, we can extract the AFM tip dimensions from convoluted AFM images using DNA as a standard (Figure 10). The tip contribution to other sample particles can then be calculated and subtracted. [7,15]

AFM can also be combined with other techniques to provide orthogonal information on protein-protein and protein-DNA interactions. Combining AFM with super-resolution fluorescence microscopy (FIONA, fluorescence imaging with one nanometer accuracy) opens the possibility of pinpointing specific, fluorescently labeled proteins within heteromeric assemblies on DNA (Figure 11) [16].

Deconvolution of AFM images

Figure 10: Deconvolution of AFM images. Due to the small but finite dimensions of the AFM probe (A), AFM images are convolutions of the actual sample topography and the AFM tip geometry (B). Using the width of DNA (red arrows) as an internal standard (C), we can estimate the size of the AFM tip based on a simple Pythagoras (D) and then calculate and subtract contributions from the tip to image dimensions (e.g. to the dimensions of the protein cluster on DNA in (C, yellow arrow)).


Figure 11: FIONA-AFM of UvrA-UvrB-quantum dot (QD) complexes with UV-damaged DNA. (A) Registration of the raw fluorescence signals with AFM topography of the same sample area. Fluorescence is shown as red color (with light center for easier visualization of AFM features) overlaid on AFM topography. The scale bar in (A) is 2 μm. A higher resolution display of the 8 μm x 8 μm area in the red box in (A) is shown in (B). (B) Fluorescence signals are shown as FIONA signals in red, corresponding to area of localization probability of the fluorescence centers. The red box in (B) indicates the QD-protein-DNA complex shown in C and D as top view (C) and 3D representation (D). The scale bar in (D) corresponds to 30 nm. These zoom in figures demonstrate good FIONA-AFM overlay accuracy. Clusters of and closely co-localized QDs (which can result from the formation of protein complexes containing more than one molecule of UvrB) can be seen in the image (for example blue arrows in (A,B)). Such closely localized fluorescence sources can lead to distorted signals (see (A)) and large inaccuracies in image alignment and are therefore not included in the image registration process and not resolved in the FIONA-AFM image (B). Optimal FIONA-AFM image registration accuracy of (13.6 ± 8.3) nm was achieved for these data using 7 QD fiducial markers.


[1] DM Bangalore, HS Heil, CF Mehringer, L Hirsch, K Hemmen, KG Heinze, I Tessmer (2020) Automated AFM analysis of DNA bending reveals initial lesion sensing strategies of DNA glycosylases, Scientific Reports 10(1), 15484l, doi: 10.1038/s41598-020-72102-7.

[2] CN Buechner, A Maiti, AC Drohat, I Tessmer (2015) Lesion search and recognition by thymine DNA glycosylase revealed by single molecule imaging, Nucleic Acids Research 43(5): 2716-2729, doi: 10.1093/nar/gkv139.

[3] DM Bangalore and I Tessmer (2022) Direct hOGG1-Myc interactions inhibit hOGG1 catalytic activity and recruit Myc to its promoters under oxidative stress, Nucleic Acids Research 50(18): 10385-10398, doi: 10.1093/nar/gkac796.

[4] J Gross, N Wirth, I Tessmer (2017) Atomic force microscopy investigations of DNA lesion recognition in nucleotide excision repair, Journal of Visualized Experiments (123), e55501, doi: 10.3791/55501.

[5] N Wirth, J Gross, HM Roth, CN Buechner, C Kisker, I Tessmer (2016) Conservation and divergence in nucleotide excision repair lesion recognition, Journal of Biological Chemistry 291(36): 18932-18946, doi: 10.1074/jbc.M116.739425.

[6] CN Buechner, K Heil, G Michels, T Carell, C Kisker, I Tessmer (2014) Strand specific recognition of DNA damages by XPD provides insights into Nucleotide Excision Repair substrate versatility, Journal of Biological Chemistry 289(6): 3613-3624, doi: 10.1074/jbc.M113.523001.

[7] I Tessmer* **, M Melikishvili**, MG Fried*(2012) Cooperative cluster formation, DNA bending and base-flipping by O6-alkylguanine DNA alkyltransferase, Nucleic Acids Research 40(17): 8296-8308, doi: 10.1093/nar/gks574. (*joint correspondence, **equal contributions)

[8] I Tessmer* and MG Fried* (2015) Characterization of homogeneous, cooperative protein-DNA clusters by sedimentation equilibrium analytical ultracentrifugation and atomic force microscopy, in: Methods in Enzymology 562 “Analytical Ultracentrifugation” (ed. J Cole): 331-348, doi: 10.1016/bs.mie.2015.06.036. (*joint correspondence)

[9] I Tessmer* and MG Fried* (2014) Insight into the cooperative DNA binding of the O6-alkylguanine DNA alkyltransferase, special issue on “Single molecule approaches: watching DNA repair one molecule at a time”, DNA Repair 20: 14-22, doi: 10.1016/j.dnarep.2014.01.006. (*joint correspondence)

[10] N Rill, A Mukhortava, S Lorenz, I Tessmer (2020) Alkyltransferase-like protein clusters scan DNA rapidly over long distances and recruit NER to alkyl-DNA lesions, Proceedings of the National Academy of Sciences of the USA 117(17): 9318-9328, doi: 10.1073/pnas.1916860117.

[11] S Kono, A van den Berg, M Simonetta, A Mukhortava, EF Garman, I Tessmer (2022) Resolving the subtle details of human DNA alkyltransferase lesion search and repair mechanism by single-molecule studies, Proceedings of the National Academy of Sciences of the USA 119(11): e2116218119, doi: 10.1073/pnas.2116218119.

[12] CN Buechner and I Tessmer (2013) DNA substrate preparation for atomic force microscopy studies of protein-DNA interactions, Journal of Molecular Recognition 12: 605-17, doi: 10.1002/jmr.2311.

[13] DM Bangalore and I Tessmer (2018) Unique insight into protein-DNA interactions from single molecule atomic force microscopy, AIMS Biophysics 5(3):194-216, doi: 10.3934/biophy.2018.3.194.

[14] I Tessmer*, P Kaur, J Lin, H Wang (2013) Investigating bioconjugation by atomic force microscopy, Journal of Nanobiotechnology 11:25. doi: 10.1186/1477-3155-11-25. (* correspondence)

[15] AT Winzer, C Kraft, S Bhushan, V Stepanenko, I Tessmer (2012) Correcting for AFM tip induced topography convolutions in protein-DNA samples, Ultramicroscopy 121: 8-15, doi: 10.1016/j.ultramic.2012.07.002.

[16] DN Fronczek, C Quammen, H Wang, C Kisker, R Superfine, R Taylor, DA Erie, I Tessmer (2011) High accuracy FIONA-AFM hybrid imaging, Ultramicroscopy 111: 350-355, doi: 10.1016/j.ultramic.2011.01.020.