Single-molecule Imaging of Genome Maintenance Proteins Encountering Specific DNA Sequences and Structures

Abstract

DNA repair pathways are tightly regulated processes that recognize specific hallmarks of DNA damage and coordinate lesion repair through discrete mechanisms, all within the context of a three-dimensional chromatin landscape. Dysregulation or malfunction of any one of the protein constituents in these pathways can contribute to aging and a variety of diseases. While the collective action of these many proteins is what drives DNA repair on the organismal scale, it is the interactions between individual proteins and DNA that facilitate each step of these pathways. In much the same way that ensemble biochemical techniques have characterized the various steps of DNA repair pathways, single-molecule imaging (SMI) approaches zoom in further, characterizing the individual protein-DNA interactions that compose each pathway step. SMI techniques offer the high resolving power needed to characterize the molecular structure and functional dynamics of individual biological interactions on the nanoscale. In this review, we highlight how our lab has used SMI techniques – traditional atomic force microscopy (AFM) imaging in air, high-speed AFM (HS-AFM) in liquids, and the DNA tightrope assay – over the past decade to study protein-nucleic acid interactions involved in DNA repair, mitochondrial DNA replication, and telomere maintenance. We discuss how DNA substrates containing specific DNA sequences or structures that emulate DNA repair intermediates or telomeres were generated and validated. For each highlighted project, we discuss novel findings made possible by the spatial and temporal resolution offered by these SMI techniques and unique DNA substrates.

2. Introduction

Cellular DNA damage repair involves a series of complex pathways each serving a function required to repair the lesion, including remodeling the surrounding chromatin, recognizing the damage site, and recruiting the appropriate downstream proteins for repair of the type of damage (1–6). The DNA repair research field has historically relied on biochemical assays and, more recently, time course or live-cell fluorescence imaging to study the protein-nucleic acid interactions that drive these pathways (7–12). While these techniques are plentiful and powerful, biochemistry assays generally share the limitation of characterizing ensemble molecular activities instead of individual interactions between biomolecules (13,14); consequently, these assays likely fail to capture important transient interactions. Further, while the advantages of live-cell imaging, including it being in cellulo and offering resolution typically on the hundreds of nanometers and approaching sub-nanometers with super-resolution imaging (15,16), are not to be overlooked, it can be challenging to generate a specific form of DNA damage in cellulo at defined locations and pinpoint direct interactions between a protein and DNA damage.

In vitro single-molecule imaging (SMI) approaches enable the study of the structure and dynamics of individual biomolecular interactions with nanometer resolution, often in real-time (17–27). The purpose of this review is to discuss how two SMI techniques, atomic force microscopy (AFM) and the DNA tightrope assay, can be used to study the molecular structure and dynamics of proteins of interest on DNA substrates emulating DNA repair or replication intermediates. The basic principles of these two approaches are first described, followed by a discussion of the generation and validation of select DNA substrates. Lastly, examples of how the high spatial and temporal resolution of SMI in combination with defined DNA substrates expanded our understanding of the cohesin complex, mitochondrial DNA replication, and telomere maintenance are highlighted. Additional information regarding sample preparation and instrumentation can be found in the referenced sources and previously published review papers (28–32).

3. AFM and the DNA tightrope assay: two complementary SMI techniques confer nanometer resolution

3.1. AFM reveals structure and dynamics of protein-nucleic acid interactions

AFM is a powerful SMI technique that had been used to study numerous DNA repair pathways, including nucleotide excision repair (33–35), mismatch repair (36–38), base excision repair (19,39–41), and homologous recombination (42,43). AFM imaging in air collects static images of a dry sample (28–30,44–46). The sample, containing proteins and nucleic acids diluted in a buffer, is typically deposited onto a freshly cleaved muscovite mica surface. The affinity between DNA molecules in the sample and the negatively charged mica surface can be supported by a variety of buffer chemistries or surface modification, including divalent magnesium ions and aminopropyl silatrane (APS) treatment. The sample deposition protocols require striking the careful balance of sufficiently adhering proteins and nucleic acids to the surface without interfering with their interactions and achieving equilibration on the surface as in an ideal two-dimensional solution (47,48). Once deposited, the sample is dried under a stream of nitrogen gas and typically imaged by oscillating an AFM cantilever through a vibration piezo at its first resonant frequency using the tapping (intermittent contact) mode (Fig. 1A) (44). Interactions between the cantilever and biomolecules on the surface cause deflection of the cantilever, damping the amplitude and modulating the phase of its oscillation. This change in oscillation pattern is monitored through a laser reflected off the back of the cantilever and directed at a four-quadrant photodetector (Fig. 1A). The associated sample height information for constructing a topographic image is obtained through a feedback system that maintains a constant set point for the amplitude of cantilever oscillation (49). Proteins on double stranded DNA (dsDNA) can be identified based on their heights and the AFM cantilever oscillation phase differentials: protein-DNA complexes are generally taller than DNA alone and display a lower phase (meaning they appear darker than nucleic acids in the phase channel of the image) (50). AFM can be used to study the oligomeric state of a protein (50,51), its binding affinity for sequences of interest (51–53), and its activities on various substrates, including but not limited to DNA looping (54–56), bridging (53,56), compaction (56,57), and unwinding (31,58).

Figure 1: Atomic force microscopy (AFM) and DNA tightrope assay.

(A) Diagram of AFM operation mechanism (29). (B) Schematic of Dual-Resonance-frequency-enhanced electrostatic force enhanced microscopy (DREEM) and example images of a nucleosome on dsDNA (76). (C) Diagram of the DNA tightrope assay. Gray spheres represent poly-L-lysine-treated silica beads (52). (D) Schematic of QD conjugation strategies. Left: Conjugation of His-tagged proteins to streptavidin-coated QDs through the biotinylated ^BT-tris-NTA linker. Right: Conjugation of epitope-tagged proteins to antibody-coated QDs using the “antibody sandwich” method (51,53).

While AFM imaging in air provides snapshots of protein-DNA complexes in action, imaging in liquids enables direct observation of sequential steps and conformational changes during biological processes (59–61). In addition, by reducing capillary forces that otherwise occur between the tip and a dry sample under ambient humidity, liquid imaging improves the spatial resolution of the images (62,63). Recent developments in enhancing the response speed of key devices, including the AFM closed feedback loop, small AFM cantilevers with high resonance frequencies, fast amplitude detectors, and high-speed scanners, have enabled high speed AFM (HS-AFM) imaging with up to ~50 frames per second (64–69). It is worth noting that for studying protein-DNA complexes, functionalization of the mica surface with the stable APS compound can enhance the reproducibility of time-lapse imaging in liquids (47).

HS-AFM has been broadly applied to study the dynamics of many protein complexes. Using HS-AFM, key protein conformational changes and novel activities have been uniquely revealed that traditional biochemistry techniques could only predict from static models. HS-AFM imaging of the glutamate transporter reconstituted in a lipid mixture revealed ligands are moved by largescale amplitude changes of key transport domains in the complex. This “elevator” mechanism was predicted by crystal structures as the mechanism by which neurotransmitters are removed from the synaptic cleft, but the dynamism-capturing abilities of HS-AFM enabled its validation (70). Similarly, the mechanism by which nucleoporins promote bidirectional macromolecule transit through nuclear pores was visualized in oocytes. This revealed a rapid and repeated lengthen-shorten transition of nucleoporins, which drives the movement of macromolecules through the pore (71). The conformational changes of KIF2 associated with ATP hydrolysis were elucidated with HS-AFM, furthering understanding of how this microtubule-destabilizing kinesin depolymerizes microtubules to facilitate intracellular transport and cellular movement (72). Particularly fascinating HS-AFM videos confirmed the ‘hand-over-hand’ movement of myosin V walking along actin fibrils, providing the first dynamic visual evidence of myosin V’s translocation mechanism (73). Part of the contended mechanism by which F₁-ATPase generates ATP was elucidated with HS-AFM, which showed the stator α₃β₃ ring cooperatively drives the power stroke of the complex independently of the central γ channel (74). Both the conformational changes and the three-dimensional diffusion of the CRISPR-Cas9-RNA complex along DNA were resolved with HS-AFM (75). These few examples out of a plethora highlight the versatility of HS-AFM and its unique ability to reveal the dynamic conformations and activities of protein complexes, both in vitro and in cellular membranes.

In traditional AFM images, DNA within protein-DNA complexes cannot be resolved, which prevents us from learning the DNA conformations in these complexes. Taking advantage of the electrostatic signals from DNA and proteins, we (Erie and Wang groups) developed Dual Resonance Frequency Enhanced Electrostatic Force Microscopy (DREEM) that simultaneously collects both topographic and electrostatic force gradient images (Fig. 1B). Through applying both a constant and modulated bias voltage to the cantilever, DREEM can resolve DNA within protein-DNA complexes due to charge neutralization where the protein and DNA interact (50,76). DREEM revealed dsDNA wrapping around nucleosomes and the path of dsDNA through mismatch repair proteins, establishing it as a powerful imaging technique to study DNA conformations in heterogeneous protein-DNA complexes (76).

3.2. DNA tightrope assay reveals protein binding dynamics

The DNA tightrope assay is a low-cost yet versatile fluorescence-based SMI method that enables a variety of dynamic protein-nucleic acid interactions to be studied (77). To form “DNA tightropes” in a flow chamber, DNA is stretched between poly-L-lysine-treated silica beads by hydrodynamic buffer flow using a syringe pump (Fig. 1C). Proteins labeled with either streptavidin or antibody-coated quantum dots (QDs) (Fig. 1D) are then flowed over the DNA tightropes. Fluorescence imaging of QD-labeled proteins is carried out using oblique angle Total Internal Reflection Fluorescence (TIRF) or highly inclined and laminated optical sheets (HILO) microscopy, which directs the laser slightly below the critical angle and enhances the signal-to-noise ratio compared to epifluorescence imaging (78). Tracking of QD-labeled proteins on DNA with high spatial (16 nm) and temporal resolutions (~50 ms/frame) allows us to study multiple parameters of protein activity. These include a protein’s diffusion constant (the rate at which a protein moves across DNA in µm²/s), alpha factor or diffusion exponent (characterizes a protein’s tendency to move in an unbiased fashion across DNA or to pause, suggesting affinity for a sequence), attachment lifetime (time bound to DNA from initial binding to dissociation), diffusion range (the distance over which a protein moves on the DNA tightrope), and binding position (79). Because DNA tightropes are elevated above the glass surface by the beads, DNA can be visualized in an extended form and protein movement on the DNA is not influenced by interactions with the surface and can be monitored without buffer flow (77).

For the DNA tightrope assay, we use either lambda DNA of ~ 48 kbp or ligated linear DNA substrates of ~20 to 40 kbp. The ligated DNA is prepared by restriction enzyme digestion of a plasmid containing a sequence of interest, followed by tandem end-to-end ligation. This protocol ensures that the DNA substrate is of sufficient length for stringing between the beads and the sequence of interest is positioned at regularly spaced intervals (Fig. 1C); Under the standard flow-stretching conditions (300 µl/min), the observed spacing of QD-labeled proteins on the tightropes is ~90% of what is expected for fully stretched DNA (79). Typically, when collecting data, DNA tightropes are not stained with dyes, such as YOYO1 or DAPI, as intercalating dyes have the potential to interfere with DNA binding by proteins. Rather, to confirm the presence of DNA tightropes, YOYO1 can be added to stain the DNA after collecting the protein-DNA binding data. Furthermore, the tension on DNA and the extent of DNA stretching can be adjusted using different buffer flow rates (79). Previously, we found that diffusion constants of TRF2 protein do not change significantly on DNA tightropes formed using 300 µl/min or 25 µl/min (stretching to ~88% DNA contour) (79). The optimal buffer flow rate that minimizes the impact of DNA tension on protein binding must be empirically determined for each individual protein. In addition, it is worth noting that DNA tightropes in stretched DNA form do not allow various DNA binding pathways, including segmental site transfer (80), DNA looping and compaction. To study DNA looping and compaction, we can anchor YOYO1-stained DNA at one end through biotin-streptavidin interactions and flow stretch the DNA, as demonstrated previously by numerous other groups (21,81,82).

4. SMI substrates: generation and validation of substrates emulating DNA repair and replication intermediates

4.1. Linear DNA with single-stranded “gap” region, single-stranded flap, and replication fork

Single-stranded DNA (ssDNA) is present ubiquitously in the nucleus, both as a structural component at the ends of chromosomes and as a transient intermediate during DNA replication (83), recombination, and repair (84–86). ssDNA plays numerous supporting roles in these processes, including helping to stimulate the reversal of stalled replication forks (83) and recruiting DNA repair proteins (86,87). While its formation is essential for these processes to occur, ssDNA is vulnerable to nuclease degradation, and thus it is critical for single-stranded binding (SSB) proteins to sequester ssDNA and protect it while also stimulating proteins that process it, including nucleases and helicases (84,88), in a highly coordinated fashion. Replication Protein A (RPA) is perhaps the most studied SSB in eukaryotes, as it rapidly coats freshly exposed ssDNA to prevent the formation of secondary structures that would prohibit subsequent recombination and repair (83,87).

With great diversity in ssDNA-binding domains (88) and a plethora of pathways in which ssDNA is an intermediate structure, novel single-stranded DNA binding proteins are discovered regularly. Given the multifaceted nature of SSB protein interaction networks, SMI approaches offer an expansive and detailed means of studying SSB-ssDNA interactions, enabling their inherently transient and highly dynamic characteristics to be captured (82,89–92).

To study ssDNA binding proteins, we generated a linear double-stranded DNA substrate containing a short single-stranded region (37 nt) referred to as the gap site. Treatment of a starting plasmid (originally generated in the Peggy Hsieh lab (93)) with a nicking enzyme generates a region with four closely-spaced nicks (Fig. 2A top panel). The gap site is exposed by adding an excess of DNA oligonucleotides complementary to the ssDNA between nicked sites (that dissociate from the duplex DNA during heating) (94). We quantify the gapping efficiency (typically ~90%) by restriction digestion at the gapped region. Subsequent restriction digestion of the gapped plasmid positions the single-stranded region at a defined location (23% from DNA end, Fig. 2A). Specific binding of mitochondrial single-stranded DNA binding protein (mtSSB) at the gap region further validates the substrate (Fig. 2A bottom panel). To exclude that the DNA sequence at the gap site does not contribute to protein binding preference, a control DNA substrate is generated by directly linearizing the plasmid (without gapping) (50,95).

Figure 2: Generation and validation of DNA substrates that emulate DNA repair and replication intermediates.

(A) Generation of linear gap DNA and its validation by mtSSB, which preferentially binds to the ssDNA (gap) region (50). (B) Generation of the DNA replication fork substrate and binding of cohesin SA2 at the fork junction (51). (C) Generation of the linear R-loop DNA through in vitro transcription and its validation using RNA:DNA hybrid-binding S9.6 antibody (52).

The protocol for generating the linear gap DNA can be modified to produce other substrates, including a “flap” or “fork” structure. By incubating the linear gap DNA with an oligo complementary at one end to the gap site, the non-complementary end of the oligo protrudes from the duplex once annealed to the gap site. This produces linear DNA with a single-stranded flap analogous to DNA repair intermediate structures (86). This single-stranded flap substrate can also be easily made into a linear double-stranded fork (analogous to a replication fork stalled when DNA damage is sustained during S-phase (96)), simply by incubating with an additional oligo complementary to the protruding flap (51).

4.2. DNA replication fork substrate

When DNA damage is sustained, replication pauses, leading to an accumulation of stalled replication forks (96–99). Various DNA repair proteins have been shown to bind stalled replication forks to protect them from nuclease degradation and promote their regression and restart (100–102). These factors are thought to be just the tip of the iceberg in a complex network of proteins involved in fork recognition and restart (100). To expand what is known about this protein network, in addition to the linear fork substrate described above, we also generated a circular DNA replication fork using a protocol established by the Jack Griffith group (Fig. 2B) (103). Briefly, the pGLGAP plasmid containing a 398-bp G-less cassette was nicked using a Nickase, followed by nick translation in the presence of dATP, dTTP, and dGTP using the Klenow fragment (exo-) to generate an ssDNA tail. To create a dsDNA tail, a primer was annealed to the ssDNA tail, followed by the extension using the Klenow fragment and purification. While we generated a 25 nt single-stranded gap region at the fork junction in our previous study (51), the length of the ssDNA gap region at the fork junction can be tuned by positioning the primer at different locations on the ssDNA tail. Typically, ~80% of the circular DNA displays a dsDNA tail, and the completion of the DNA replication fork substrate is validated using AFM imaging by measuring the length of the dsDNA tail (~129 nm) and DNA binding of cohesion SA2 protein, which binds at fork junctions (Fig. 2B bottom) (51).

4.3. R-loop DNA substrate

R-loops are RNA-DNA hybrid structures that form when a nascent RNA anneals to the template DNA strand, displacing the non-template strand as a ssDNA loop (104,105). R-loops appear to play diverse and important roles in normal cellular function, including regulating gene expression (105), telomere maintenance (104,106), and DNA double-strand break repair (104–106). However, aberrant R-loop formation is also associated with genome instability: their formation can be induced by DNA damage (106,107), and if not promptly resolved, their accumulation can be genotoxic (104–107). The multifaceted roles of R-loops in both DNA damage and repair make studying them complex, and understanding of how they are resolved remains limited, with an array of RNA:DNA helicases, RNases and DNA repair proteins that appear to contribute to their resolution through regulatory pathways that are not fully understood (106,107).

To study how DNA repair proteins function on R-loops, we generated a linear R-loop substrate by performing in vitro transcription on a plasmid originally created by the Chédin laboratory (108). This substrate contains the Airn gene cloned downstream of the T3 promotor (Fig. 2C top panel). A previous study established that Airn contains two R-loop forming sites (38% and 42% from one DNA end on our linearized DNA), with one overlapping with a G-stretch forming more stable R-loops. RNase treatment removes unpaired single-stranded RNA. Linearization by a restriction enzyme positions the R-loop regions at defined positions. Interestingly, AFM imaging revealed that the extruded non-template strand in the R-loop adopts distinct three-dimensional architectures, including ‘loop,’ ‘spur,’ and ‘blob’ conformations (108). These structures vary significantly in their sizes and shapes, and are distinguishable from dsDNA by their relative heights (Fig. 2C bottom panel) and volumes (108) using AFM. As the ‘spur’ and ‘loop’ conformations protrude outwards from the DNA backbone, there is uncertainty in determining their positions along the linear DNA. Thus, we focused our analysis on the ‘blob’ shape, which falls along the DNA backbone and is conveniently also the most abundant R-loop conformation (52). The presence and position of the R-loop can be further validated with an S9.6 antibody, which binds RNA-DNA hybrids and can be identified in AFM images based on its height above dsDNA (Fig. 2C bottom panel) (52,109). AFM imaging allows us to discern whether a protein preferentially binds to and is activated by R-loops based on its binding positions along the R-loop DNA (52,109). By measuring the maximum height of the R-loop from many molecules, we first established a threshold height for distinguishing the R-loop from surrounding dsDNA (0.89 nm, Fig. 2C bottom panel). We then can determine if a protein or complex of interest binds the R-loop position by measuring the height of that protein-R-loop complex. If bound to the R-loop position, the protein-R-loop complex should display a height significantly greater than R-loop alone.

5. Highlights: Novel findings from the AFM and the DNA tightrope assay

5.1. Mitochondrial DNA replication machinery

The mitochondrial genome is organized into many copies of circular, double-stranded DNA that encodes essential cellular respiration enzymes (110), as well as transfer and ribosomal RNAs required for translation within the mitochondria (110,111). Numerous heritable diseases are associated with the failure of the mitochondrial genome to be replicated and repaired properly (112). The mitochondrial replisome consists of the DNA polymerase γ holoenzyme, a single-stranded DNA binding protein (mtSSB), and a helicase known as Twinkle (50,95,113). mtSSB assists the mitochondrial replisome by suppressing secondary structures that form in unwound ssDNA (114) and stimulating the activities of DNA Polymerase γ (114,115) and Twinkle (116,117). When mtSSB is knocked down in cultured human cells, mtDNA synthesis decreases, leading to a reduction in mitochondrial genome copy number over time (118). Twinkle might also function in recombination-mediated double-stranded break (DSB) repair of mitochondrial DNA (119). The functions and activities of the E. coli replisome are well-characterized with significant contributions having been made by AFM and HS-AFM studies. Thanks to the ability to visualize the dynamics of protein-nucleic acid interactions, AFM has uniquely revealed that the RecG and PriA helicases of the E. coli replisome translocate away from the replication fork junction, in processes facilitated by E. coli SSB (120,121) and ATP (122,123), respectively. Additionally, AFM imaging has captured the collaborative binding and oligomerization of E. coli SSB tetramers on ssDNA (124,125). However, our understanding of how the mitochondrial genome is replicated and repaired, and how that process can be disrupted in human diseases, is limited. AFM and DREEM imaging offer the resolution needed to help address these gaps.

Previously, typical DNA substrates utilized to assess DNA unwinding by Twinkle in vitro contain only short duplex regions (~60 bp) (126). Thus, the dynamics and processivity of DNA unwinding by Twinkle were largely unknown. To fill this knowledge gap, we used AFM imaging in air to characterize the helicase activity of human Twinkle (hTwinkle) that generates ssDNA (31,95). We imaged hTwinkle on circular gap DNA in the presence of ATP and observed an accumulation of ssDNA with lower heights than dsDNA by approximately 200 pm (Fig. 3A). AFM imaging of hTwinkle-gap DNA samples deposited on mica surfaces after defined incubation times revealed that hTwinkle unwinds several hundred base pairs of duplex DNA at an average rate of ~240 bp/min.

Figure 3: Structure and dynamics of DNA binding and unwinding by Twinkle revealed by AFM imaging in air and HS-AFM imaging in liquids.

(A) Representative AFM image capturing hTwinkle’s helicase activity. ssDNA accumulates around hTwinkle as it unwinds duplex DNA (left panel). ssDNA (denoted by gray line) can be distinguished from dsDNA (denoted by blue line) by their height difference (~200 pm, right panel) (95). (B) Representative AFM images of hTwinkle in the presence of mtSSB on circular DNA with two gap sites (left panels), and quantification of hTwinkle’s DNA unwinding rate (bp/min) with increasing concentrations of mtSSB (right panels). hTwinkle partially unwinds one or both gap sites, or fully unwinds the duplex between the gap sites (95). (C) Time-lapse HS-AFM images showing LcTwinkle captures DNA through N-protrusion (denoted by green arrow) (129). Also see Video S1. (D) Time-lapse HS-AFM images showing LcTwinkle transiently binds to DNA through N-protrusion (denoted by green arrow), followed by the transfer to its central channel (denoted by pink arrow) (129). Also see Video S2. (E) Model of Twinkle DNA binding based on cryo-EM structure and AFM imaging of LcTwinkle-DNA (129).

While mtSSB is clearly established as an essential player in mitochondrial genome replication, full understanding of how it binds to DNA to complete these roles has been limited (50), compared to E. coli SSBs (117,120–122,124,127). We first used AFM imaging in air to establish the native conformation of mtSSB in the absence of DNA. Volumetric analysis of individual mtSSB complexes correlated with mtSSB forming tetramers, as assessed using the standard calibration curve detailed above (50). Knowing the oligomeric state of apo-mtSSB, we next applied AFM imaging in air as well as DREEM imaging to characterize the association of mtSSB with DNA. When incubated with single-stranded circular DNA, mtSSB binding density increases proportionally to its concentration. However, each bound mtSSB is clearly separate from adjacent ones on the DNA, indicating mtSSB does not bind ssDNA cooperatively, making it distinctly different from E. coli SSB (50). When incubated on the linear gap DNA, mtSSB preferentially binds the single-stranded gap site (Fig. 2A), while it is randomly distributed across control DNA. Volumetric analysis again indicated mtSSB binds the gap site as a tetramer (see (50)). Interestingly, mtSSB binding appeared to shorten the linear gap DNA by approximately 20 nm or 40 nucleotides, equating to the girth of one mtSSB tetramer. This raised the question if mtSSB wraps ssDNA around itself. To address this question, we probed the molecular structure of mtSSB bound at the gap position using DREEM imaging. Previously, we established that DREEM can reveal ssDNA wrapping around histone proteins (Fig. 4A) (128). Using DREEM, we observed DNA (visible as a thin light-colored band wrapped around dark structures) wrapping around individual mtSSB tetramers (the darker structures) with one turn at the position of the gap site (Fig. 4B), consistent with the measured shortening of the gap DNA.

Figure 4: DREEM reveals DNA paths in protein-DNA complexes.

(A) ssDNA wraps twice around histone proteins (128). (B) dsDNA wraps around mtSSB once (50). (C) DNA appears at the edge of large TRF2 complexes (208). (D) dsDNA wraps around the TRFH domain from TRF2 (171). Left panels: Topographical AFM images. Right panels: DREEM images. Inserts in A, B, and D: models of protein-DNA complexes based on DREEM images.

Beyond characterizing the structure of mtSSB bound to DNA, we also used AFM imaging in air to study how mtSSB affects the helicase activity of Twinkle. We co-incubated hTwinkle with increasing concentrations of mtSSB. By tracking the sample incubation time and measuring the length of ssDNA produced by hTwinkle at increasing concentrations of mtSSB, we discovered that the addition of mtSSB results in a ~5-fold stimulation of the apparent DNA-unwinding rate (Fig. 3B). mtSSB also increases the estimated translocation processivity from 1750 to >9000 bp before hTwinkle disassociation (95). AFM analyses indicated that mtSSB both enhances loading of Twinkle onto defined DNA substrates and stabilizes the unwound ssDNA product (95).

How Twinkle self-loads onto DNA without a loader protein was a long-standing question. While AFM imaging in air provides the resolution needed to observe Twinkle’s oligomeric state and helicase activity, it lacks the temporal resolution for capturing Twinkle’s conformational changes required for self-loading onto DNA. By tuning the 1-(3-aminopropyl)silatrane (APS) (47) concentration on a mica surface, we developed robust sample deposition conditions to achieve the intricate balance of keeping protein and DNA molecules partially anchored onto a surface while still being mobile (129). HS-AFM imaging in liquids (0.8 to 2.3 frames/s) enabled Twinkle’s DNA binding dynamics to be observed in real-time. For HS-AFM imaging, we used a Twinkle homolog that is similar to hTwinkle, Lates calcarifer (Lc) Twinkle, which is amendable to cryo-EM structure determination (129). At a distance (> 200 nm) from DNA, LcTwinkle switches between open and closed ring conformations, but exhibits limited conformational changes. Interestingly, when DNA is in close proximity (distance <20 nm), conformations of LcTwinkle change dramatically with small protrusions (~5 nm) directed towards the DNA (Fig. 3C, green arrow; Video S1) observed from all LcTwinkle molecules. LcTwinkle appears to use this protrusion to initially capture DNA, followed by the loading of DNA into its central channel in the open-ring form (Fig. 3D, pink arrow; Video S2) (129). Taken together, the results from HS-AFM and cryo-EM (129) revealed two separate DNA binding regions on LcTwinkle: the N-terminal domain (NTD) for initial transient DNA capture and the helicase central channel for DNA unwinding. Our results suggest that electrostatic interactions between the positive charges on the NTD and the negatively charged DNA drive the initial loading of Twinkle on DNA (Fig. 3E). We proposed that the flexible N-C linker between the NTD and CTD enables the NTD to protrude in the presence of DNA.

5.2. R-loops recruit cohesin SA1/SA2 and PARP1

Emerging evidence supports the notion that the three-dimensional (3D) structure of the interphase genome plays key roles in regulating DNA replication and repair (130). In addition to its role in maintaining sister chromatids cohesion during S-phase, the cohesin complex, assisted by NIPBL, is the primary orchestrator of the 3D genome organization (131–134). Cohesin contains SMC1, SMC3, RAD21, and mutually exclusive SA1 or SA2 (135). Differences in cohesin-SA1 and cohesin-SA2 relative abundances in human cells suggest these complexes might have different functions conferred by SA1 and SA2 (136).

While cohesin helps maintain the architecture of the entire genome, it is also recruited to DNA damage sites (137–142). Recent research shows cohesin performs chromatin loop extrusion at these damage sites, and this activity appears to promote the formation of DNA damage response foci (143) and subsequent DNA repair (140,144–146). However, the mechanisms by which cohesin recognizes various DNA damage sites and promotes their repair is not well understood. Additionally, research has shown that SA1 and SA2 are differentially essential for maintaining sister chromatids cohesion between telomeric and centromeric regions, respectively (136). How SA1 and SA2 might recognize these sequences and DNA damage sites and differentially respond to them has not been well studied. Using AFM and the DNA tightrope assay, we showed that SA1 and SA2 are DNA-binding proteins that independently recognize several hallmarks of DNA damage and display differential binding dynamics on DNA sequences of interest (51,53).

AFM imaging revealed both SA1 and SA2 directly bind dsDNA and target DNA ends in particular (51,53). However, SA1 and SA2 display distinct binding behaviors on DNA substrates containing telomeric or centromeric sequences. AFM imaging of SA1 on telomeric T270 DNA showed a higher percentage of SA1 binding events occurring at the telomeric sequence repeat region compared to that same region of control DNA (53). SA2, meanwhile, does not specifically bind to telomeric or centromeric sequences (51). This finding highlights a distinct benefit offered by AFM compared to ensemble biochemistry assays: while techniques like electrophoretic mobility shift assay (EMSA) and fluorescence anisotropy are powerful tools for gauging if a protein can bind a nucleic acid substrate, they offer no information on where binding occurs, such as at DNA ends versus internally. Furthermore, the DNA tightrope assay offered insight into how SA1 finds the telomeric region and revealed unique binding modes that SA1 and SA2 display. The majority of QD-labeled SA1 statically binds the T270 tightropes at regularly spaced intervals corresponding to the position of telomeric repeats for extensive time periods often exceeding the 2-minute recording window, suggesting a high-affinity interaction. Some SA1 proteins diffuse across the tightropes at ranges spanning several telomeric repeats and sometimes paused at telomeric regions. SA1 displays a much shorter dwell time and diffuses freely across tightropes containing genomic DNA and centromeric sequences, further supporting SA1’s preference for telomeric sequences (53). QD-labeled SA2 diffuses freely over DNA tightropes containing either genomic, centromeric or telomeric sequences, supporting our findings with AFM that SA2 shows no binding preference for these sequences in vitro (51).

While SA2 displays no clear sequence specificity for the aforementioned dsDNA substrates, it does preferentially bind to ssDNA (51). Using AFM imaging in air, we found that SA2 primarily binds at the single-stranded region of linear gap DNA, while it is randomly distributed across control dsDNA (51). AFM imaging revealed that DNA binding (dsDNA as well as ssDNA) stimulates SA2 to transition from a mostly monomeric form to a higher-order oligomeric state (51,52). QD-labeled SA2 displays one-dimensional search behavior, diffusing over DNA and binding statically to regularly spaced ssDNA gap regions. On control non-gapped DNA, SA2 is primarily diffusive. We also found using AFM imaging that SA2 preferentially binds to DNA flap and fork substrates, with most binding events occurring at the DNA junction on the DNA replication fork substrate (Fig. 2B bottom panel) (51). Taken together, these results show a clear preference by SA2 for intermediate structures during DNA repair, recombination, and replication. These findings led us to propose a novel model in which cohesin SA1 and SA2 serve as “anchor” proteins to localize cohesin at specific DNA sequences and structures. Recently, the cryo-EM structure of cohesin-NIPBL (147) and biochemical studies (64) further establish the interface between SA1 and DNA.

Given the increasingly recognized role RNA plays in DNA repair (148–152), we explored the behavior of SA1 and SA2 on various RNA and RNA:DNA hybrid substrates. AFM height analysis showed SA1 and SA2 oligomerize on ssRNA oligos, long ssRNA, and R-loop DNA (52). On the linear R-loop DNA, both SA1 and SA2 preferentially target the R-loop region (Fig. 5A) yet randomly distribute across control DNA. This was validated with the tightrope assay, which demonstrated both SA1 and SA2 bind at regularly spaced intervals corresponding to the positions of the tandemly ligated R-loops (Fig. 5B). A recent study confirmed that in cells SA1/SA2 localize to R-loops and are required for cohesin loading onto DNA in the absence of NIPBL. Furthermore, SA1 may play a key role in R-loop homeostasis. (153).

Figure 5: R-loops recruit SA1/SA2 and PARP1, and activate PARP1.

(A) Diagram of the linear R-loop DNA (top) and representative AFM image of SA2 on the R-loop DNA (bottom) (52). (B) Diagrams of the DNA substrates (top left), representative fluorescence images (middle left) and kymographs (bottom left) of QD-labeled SA2, and analysis of the spacing between QD-labeled SA2 on R-loop and control DNA (right) (52). The distribution of distances between adjacent QD-SA2 on R-loop DNA tightropes is consistent with SA2 specifically binds to R-loops (marked by the red lines). (C) Model and representative AFM image of PARP1 binding linear R-loop DNA (left panel) and analysis of positions of R-loops, PARP1 on R-loop DNA and control DNA in the absence of NAD⁺ (right panel) (109). (D) Model and representative AFM image showing PARP1 auto-PARylation on the R-loop DNA in the presence of NAD+. Arrow: PAR chain (109).

ADP-ribosyltransferases (ARTs), which transfer either one (MARylate) or multiple ADP-ribose units (PARylate) from NAD⁺ onto substrates, play essential roles in regulating DNA damage response (DDR) (154–159). ART family members, including PARP1, PARP2, and PARP3, are activated in the presence of damaged DNA. The resulting linear or branched synthesized PAR polymers are important for assembling and activating many DNA repair proteins at the damaged site (160–164). PARP1 is one of the earliest proteins to arrive at a damage site during the DNA damage response (165). Previous AFM studies support this, demonstrating PARP1’s binding preference for DNA damage sites in vitro (19,24,162,166). A growing body of evidence suggests PARP1, R-loops, and R-loop resolving proteins overlap in several pathways (109,167–169). However, the mechanisms by which PARP1 and R-loop resolving proteins interact at R-loops and how this contributes to R-loop resolution are not well understood.

AFM imaging in air revealed that PARP1 preferentially binds R-loops on linear dsDNA (Fig. 5C). Most PARP1 binding events without NAD⁺, occur at the position of the R-loop along dsDNA, yet are distributed randomly across control DNA (109). It is worth noting that the identification of PARP1-R-loop complexes on dsDNA was based on the significantly increased height compared to R-loop alone. In some cases, PARP1 can be identified binding to the “spur” R-loop structure, as seen in Fig. 5C. Interestingly, binding to the R-loop in the presence of NAD+ activated PARP1’s PARylation activity, which is visible on AFM as branches budding out from PARP1 bound at the R-loop (Fig. 5D). Additionally, AFM revealed that PARP1 can bridge two R-loop DNA molecules together at the position of the R-loops (109). These findings were validated using a PARP1 knockout cell line, in which it was found that R-loops accumulate in the absence of PARP1. These results suggest that activation of PARP1 at R-loops could serve to suppress R-loop formation or facilitate the resolution of R-loops (109).

5.3. Dynamics of TRF1 and TRF2 binding to telomeric DNA

Telomeres are nucleoprotein structures that assemble at the ends of linear chromosomes (170) and prevent chromosomal ends from erroneously initiating DNA damage checkpoints, recombination, and end-fusion (170–177). Telomeric sequences are bound by the protective shelterin complex, which contains six protein constituents (Fig. 6A), including Telomeric Repeat Binding Factors-1 and −2 (TRF1 and TRF2), TRF1-Interacting Nuclear Protein-2 (TIN2), Protection of Telomeres-1 (POT1), Telomere Protection Protein-1 (TPP1), and Repressor/Activator Protein-1 (RAP1) (178). TRF1 and TRF2 homodimerize and directly bind telomere DNA, which consists of 2–20 kb tandem arrays of the non-coding (TTAGGG)_N sequence that terminates with a 50–400 nucleotide-long G-rich 3’ overhang (179). Telomere DNA can assume a variety of higher-order structures that play important roles in regulating telomere maintenance, including the formation of G-quadruplex structures (180–183) in the 3’ overhang and T-loops (171,184,185). T-loop formation is dependent on TRF2 (186) and occurs through invasion of the G-rich 3’ overhang into the telomere duplex DNA, forming a lasso-like loop (Fig. 6A) (184), as has previously been captured with AFM (187). Through sequestration of the 3’ overhang, T-loop formation is an important means by which chromosomal ends are distinguished from DNA double-stranded breaks, thus helping prevent aberrant initiation of the DNA damage response (185,186). Additionally, although long thought to be transcriptionally silent, sub-telomeric DNA is transcribed to long noncoding Telomeric Repeat-containing RNA (TERRA) in mammalian cells, a significant portion of which remains associated with telomeres, either those from which it is transcribed (cis interacting) or with telomeres of other chromosomes (trans interacting) through poorly understood mechanisms (188,189). There, TERRA serves numerous important roles, including promoting homology-directed repair of telomeres (190) and telomere length maintenance (181), and serving as a binding scaffold for protein factors, such as TRF2 (191) and several DNA repair factors (192,193). Loss of TERRA has been shown to cause telomere structural aberrations (191) and promote the erosion of telomeric sequences (194,195), indicating TERRA plays essential roles in telomere maintenance.

Figure 6: DNA tightrope study of TRF1/TRF2 and the interaction between TRF1 and cohesin SA1.

(A) Diagram of shelterin complex and T-loop (32). (B) Schematic of telomere DNA substrate [54]. (C) Representative fluorescence images of QD-TRF1 and QD-TRF2 on T270 tightropes. Their regular spacing shows specific binding of TRF1 and TRF2 to telomere regions at even intervals (79). (D) Representative kymographs for QD-labeled TRF1 and TRF2 on T270 DNA tightropes (79). (E) Representative kymograph showing QD-SA1 pauses while moving across telomere sequences (53). (F and G) Representative tightrope kymographs (F) and analysis of diffusion range (G) of QD-TRF1 (green) and QD-SA1 (red) on telomeric T270 DNA and genomic DNA (G-DNA) (53).

Additionally, numerous DNA repair proteins play important roles in telomere maintenance (170,171,196,197). While understanding of the DNA binding activities of individual telomere proteins is expanding (180,198,199), many holes remain in our understanding of the interplay between telomere proteins and DNA repair pathways. In particular, it is not well understood how telomere-binding proteins assemble on telomeric sequences to promote higher-order structures, like T-loops, and telomere compaction that regulate DNA repair protein access at telomeres (170,200). The complex array of interactions between shelterin components, DNA repair proteins, telomeric DNA, and TERRA are extensive, making it difficult to tease apart the functional roles and collaborative activities of each of the different components in telomere maintenance. As such, while ensemble biochemistry assays have revealed essential elements of telomere biology (170,201), they are limited in their abilities to capture the molecular mechanisms by which shelterin proteins associate with telomeric sequences and how the shelterin complex regulates the formation and maintenance of higher-order telomere structures. Further, while cell-based assays are essential approaches for studying the downstream effects of protein knockouts (173,191,202,203), T-loop formation, and protein localization to telomeres (188,204), teasing apart the contributions from the different constituents of shelterin with these methods is challenging. Some shelterin proteins bind DNA collaboratively, so that eliminating one protein may simultaneously remove several of its binding partners from DNA. In comparison, AFM and the DNA tightrope assay enable the binding dynamics and structures of shelterin’s constituents on telomeric sequences to be directly (and individually) studied in vitro.

To study telomere binding proteins using SMI, we utilize a DNA substrate (T270) that was initially designed with 270 repeats of the telomeric sequence by the de Lange lab (205). We linearized it to position the telomeric region at a desired location (Fig. 6B). This substrate can be used directly for AFM imaging or can be ligated in tandem for the DNA tightrope assay.

TRF1 and TRF2 homodimerize through their TRFH domain and recruit other shelterin components to telomeres (56,206). However, it was poorly understood how these proteins initially find telomeric sequences in the genome and subsequently recruit other shelterin and DNA repair proteins (170). A previous in cellulo study using fluorescence recovery after photobleaching suggested TRF1 and TRF2 bind telomeric sequences dynamically (207). We used the DNA tightrope assay to characterize the binding dynamics of individual TRF1 and TRF2 proteins on telomeric (Fig. 6B) and non-telomeric DNA tightropes. On T270 DNA tightropes, static TRF1 and TRF2 show spacing corresponding to the positions of telomere repeat sequences (Fig. 6C) (79). Interestingly, mobile TRF1 on telomeric regions displays a much narrower diffusion range than TRF2 (Fig. 6D, Video S3). Taken together, the DNA tightrope assay revealed that TRF1 and TRF2 use distinct search mechanisms to find telomere sequences, in which TRF1 binds DNA largely through 3D diffusion while TRF2 performs extensive 1D search on DNA through a sliding mechanism. We propose that individual TRF proteins use 1D sliding on DNA to find protein partners and assemble the shelterin complex, which in turn stabilizes the interactions on telomeric DNA (79).

It is worth noting that recent sequencing of the T270 plasmid using the Nanopore technology (Plasmidsaurus) revealed degeneracy in the telomeric sequences, which shows roughly 153 TTAGGG repeats and 126 TTGGGG repeats. While this new information should not affect the interpretation of results from previous studies using the T270 plasmid, it does raise the awareness that telomeric sequences are highly prone to mutations during replication inside bacteria. To compare across different studies, we need to monitor and report the telomeric sequences when using plasmid-based DNA substrates.

5.4. Higher-order telomere structures

To study how TRF2 promotes higher-order DNA structures, we first directly probed TRF2’s DNA compaction ability using DREEM imaging. We observed T270 DNA passing through small TRF2 complexes (208) and budding out from the edge of large TRF2 complexes (Fig. 4C), suggesting DNA is compacted within the complex. Additionally, DREEM imaging of TRF2’s homodimerization domain (TRFH) on telomeric DNA revealed that TRFH wraps DNA around itself (Fig. 5D) (171). These findings suggest that TRF2 condenses telomeric DNA by wrapping it around individual TRFH domains inside TRF2 and by corralling it into protein-free loops that protrude from multi-TRF2 complexes.

In cellulo, most TRF1 and TRF2 homodimers are complexed with TIN2 (209). While TIN2 itself cannot bind DNA, it is known as the linchpin of the shelterin complex as it bridges TRF1 to TRF2, stabilizing their interaction on telomere sequences and with other protein components of shelterin (Fig. 6A) (170,209–212). This stabilization of TRF1 and TRF2 on telomeres is essential to prevent triggering a DNA damage response (212). However, these very attributes that make TIN2 essential for telomere biology also complicate its study: because it interacts with several shelterin components, it is difficult to characterize all its functional roles with ensemble biochemistry approaches. Because TIN2 facilitates TRF1/TRF2 association with telomeres, the knockdown of TIN2 in cells also removes TRF1/TRF2 from DNA, confounding the interpretation of observed downstream effects.

Using both short (S) and long (L) TIN2 isomers, DNA tightrope studies showed that TRF2 is required for the recruitment of TIN2 to telomere sequences in vitro (213). We next sought to determine how TIN2 affects TRF2 DNA binding. Previous research using biochemical assays (214) and AFM (214,215) suggest that TRF1 and TRF2 compact short telomeric DNA substrates. One technical barrier in the single-molecule study of DNA compaction by the telomere system is the availability of long native telomeric DNA (>2 kb) with high purity. To overcome this technical barrier, we optimized the DNA target capture technology, the Region-Specific Extraction (RSE) system, for purifying longer telomeric DNA (Fig. 7A). We first digested genomic DNA purified from mouse liver tissues using a restriction enzyme cocktail, followed by the hybridization of a complementary capture primer to the 3’ overhang of the telomere DNA, extension of the primer using PCR reactions containing biotinylated-dNTP, and purification using streptavidin-coated magnetic beads (213). PCR-extended primers contain multiple biotinylated nucleotides that provide high stability and accessibility to streptavidin-coated magnetic beads. From 125 μg of mouse genomic DNA, the RSE protocol generated ~0.18 ug of purified mouse telomeric DNA with lengths equivalent to λ DNA (~48 kbp), and its purity was validated using a dot blot assay (213). We characterized the effect of TIN2 on TRF2-mediated DNA compaction using the surface-anchored and flow-stretched telomeric DNA purified using the RSE method. We measured the change in end-to-end length of YOYO1-labeled long mouse telomeric DNA as an indicator of DNA compaction and found that TIN2L and TIN2S isomers significantly enhanced TRF2-mediated telomeric DNA compaction (Fig. 7B&C) (213). This result was corroborated by observations from AFM imaging.

Figure 7: TIN2 enhances TRF2-mediated DNA compaction, bridging of single-stranded telomeric RNA to telomere sequences, and T-loop formation.

(A) The RSE method for purifying long telomeric DNA from mouse liver (213). (B and C) QD-TIN2 (purple) enhances TRF2-mediated compaction of YOYO1-labeled long telomere DNA (tDNA) purified using the RSE method. Black arrows mark the DNA anchoring proint. White arrows mark reversal in buffer flow direction to validate the absence of nonspecific interactions between DNA and surface (213). (D) Schematic of the DNA tightrope assay for monitoring the bridging of single-stranded telomeric RNA (TERRA) onto DNA tightropes. (E) example images (top) and kymographs (bottom) of QD-labeled telomeric RNA on DNA tightropes in the presence of TRF2 or TRF2-TIN2S (213). (F) Representative AFM images showing TRF2 and TRF2-TIN2L mediated T-loops on linear T270 DNA. A QD in the top image marks the non-telomeric end (213).

TRF2 directly binds to TERRA (191,216). To investigate if TIN2 promotes the recruitment of TERRA to telomeric DNA in trans, we used a modification of the tightrope assay. We labeled biotinylated ssDNA or ssRNA (TERRA) molecules containing twelve TTAGGG or UUAGGG repeats, respectively, with QDs (Fig. 7D) and introduced them to a flow cell containing ligated T270 DNA tightropes. Without proteins, no signal from telomeric ssDNA or ssRNA accumulated on the tightropes, indicating any bridging we might observe is protein-dependent. TRF2 successfully recruits TERRA molecules to the tightropes at spacing consistent with the distance between telomeric regions (Fig. 7E) (213); the QD signal remains at the tightropes for the extent of the 2-minute recording window, indicating the interaction is highly stable. In the presence of TIN2, the intensity of QD recruited to T270 tightropes increases significantly, indicating TIN2 induces cluster formation at telomere sequences by enhancing TRF2-mediated bridging of telomeric ssDNA and ssRNA. The bridging of telomeric ssRNA to telomeric duplex DNA mediated by TRF2 and TIN2 is consistent with a model in which TERRA serves as a scaffold for shelterin protein binding. These results also provide direct experimental evidence that TERRA expressed from one chromosome can act in trans at other chromosomal ends.

Given the collaborative activities of TRF2 and TIN2 on telomeric DNA that we were able to observe using AFM imaging, we next sought to investigate if TIN2 modulates TRF2-mediated T-loop formation. Importantly, previous research has demonstrated that TRF2 alone is sufficient for T-loop formation (198), but has relied on TIN2-knockouts that are also known to destabilize TRF2 at telomeres. To characterize the effect of TIN2 on TRF2-dependent T-loop formation, we generated a model T-loop substrate by incubating linear T270 DNA with an exonuclease, producing a G-rich 3’ overhang (213). We attached a QD via a biotin-tagged DNA primer to the opposite non-telomeric side to distinguish the telomeric end from the non-telomeric one (Fig. 7F top panel). Using AFM imaging in air, we observed TRF2 alone promotes T-loop formation at the 3’ overhang in approximately 10% of DNA molecules (213). In the presence of TIN2, however, the percent of DNA molecules with T-loops increases in a TIN2 concentration-dependent manner (213). We observed large multi-protein complexes of TRF2 and TIN2 form at the base of the T-loops (Fig. 7F bottom panel), which grew larger with increasing TIN2 concentrations. No T-loop formation was observed on control T270 DNA not treated with an exonuclease, indicating this collaborative TRF2-TIN2-mediated activity requires a 3’ overhang.

Importantly, in a separate study of TRF1 and TIN2 using AFM imaging and the DNA tightrope assay, we observed that TIN2 also facilitates TRF1-mediated telomeric DNA compaction and DNA-DNA bridging (56). These observations support a model that TIN2 serves as an architectural protein at telomeres by promoting TRF1 and TRF2-mediated higher-order DNA structures, including DNA compaction, bridging of TERRA to duplex telomeric DNA, and T-loop formation.

5.5. The interplay between cohesin SA1 and telomere binding protein TRF1

Cohesion between the telomere regions of sister chromatids is an important prerequisite to genome stability, as it promotes proper chromosome segregation (217). However, the mechanism by which this occurs at telomeres was not well understood. Previous research has shown that cohesin’s SA1 subunit, TRF1, and TIN2 interact (203). Sister telomere association is largely dependent on SA1, and not the cohesin ring subunits (SMC1, SMC3, and RAD21) (218). Given the importance of SA1 for telomere cohesion and its interaction with shelterin proteins TRF1 and TIN2, we sought to determine how these proteins collaboratively maintain telomere DNA-DNA bridging, which is a key activity during sister telomere cohesion. Having discovered using AFM imaging in air that SA1 directly binds telomere repeat sequences in vitro (53), we used the tightrope assay to characterize how SA1 finds telomere sequences and if TRF1 influences this process. As anticipated, we observed QD-labeled SA1 binds to tightropes of ligated T270 DNA with spacing corresponding to the positions of telomeric sequences. Some QD-SA1 proteins bound telomere sequences statically, while others remain mobile, covering distances up to 8.55 μm along the tightropes (Fig. 6E&G), a distance that would include several telomere sequence repeats. A subpopulation of mobile QD-SA1 molecules alternates between fast and slow diffusion, a behavior known as bimodal diffusion; periods of slow diffusion correspond to the positions of telomere sequences (Fig. 6E and Video S4). This result suggests that SA1 alone diffuses along DNA containing telomeric and non-telomeric sequences and pauses as it passes over telomere sequences. We next repeated the tightrope assay with SA1 and TRF1 differentially conjugated to red and green quantum dots, respectively. Under these conditions, a subpopulation of QD bound to the tightropes was dual-colored, indicating QD-TRF1 and QD-SA1 are bound to the same telomere repeat sequence (Fig. 6F). When complexed with TRF1 in this way, a greater percentage of SA1 binding events are static than for QD-SA1 alone and bound with spacing corresponding to the telomere sequence spacing. Furthermore, the SA1-TRF1 complexes that remained mobile possess a much narrower diffusion range of approximately 0.5 μm (Fig. 6G); this confined distance is less than the length of one (TTAGGG)₂₇₀ region on the tightrope. On non-telomeric control DNA, most mobile SA1-TRF1 complexes are not confined to such a narrow diffusion range and display much higher diffusion constants and alpha factors than on telomeric DNA. In summary, evidence from the tightrope assay revealed that TRF1 interacts with SA1 to enhance stable SA1 association with telomere sequences. Furthermore, AFM imaging revealed that SA1 enhances TRF1-mediated telomeric DNA-DNA bridging (53). Similar to what we observed on telomeric DNA, we speculate a longer dwell time of SA1 at AT-rich sequences (in its recognition mode) could stabilize the cohesin complex for carrying out unique functions at these regions across the genome. We propose a model in which SA1 and SA2 are the “DNA sequence and structure guide” for the cohesin complex, which directs its loading at specific sequences/structures along the genome.

6. Outlook

Synergistic with high-resolution structure determination using cryo-EM, HS-AFM will continue to shed unprecedented insight into the dynamic of multi-protein DNA repair complexes. While AFM resolution is limited by the radius of the tip, localization AFM (LAFM) (219) and simulation AFM (220) further improve image resolution for biological samples. By applying localization image reconstruction algorithms, simulation AFM and automated fitting to experimental images, these analytical approaches thus reveal yet finer details of protein-DNA interactions. Another promising development in AFM technology is nanoendoscopy-AFM, which inserts a nanoprobe into living cells and images their interior while preserving cellular viability (221). Molecular resolution imaging of intracellular or even intranuclear dynamics of DNA repair proteins in live cells would contribute fascinating research to the DNA repair field. Furthermore, the recently developed Single-Molecule Analysis of DNA-binding proteins from Nuclear Extracts (SMADNE) promises to expand the capacity of the DNA tightrope assay to probing of native protein-DNA interactions in the context of physiologically-relevant nuclear proteins (222). Furthermore, laser-trap based SMADNE enabled studies of protein-DNA interactions under variable tension. For example, using SMADNE, it was demonstrated that increasing tension on nicked DNA leads to more binding events by PARP1 (222). With new generations of high-speed and high-sensitivity scientific CMOS cameras, brighter and more stable fluorophores will break the temporal resolution bottleneck of the fluorescence imaging-based DNA tightrope assay. In summary, SMI techniques, including AFM and the tightrope assay, continue to evolve and enhance spatial and temporal resolutions of protein-DNA complexes, which will drive the new frontiers in the DNA repair research field.

Highlights.

• High-speed AFM and the DNA tightrope assays provide novel findings regarding the structure and dynamics of protein-DNA interactions.

• Unique substrates containing specific DNA sequences and structures are generated to emulate DNA repair or replication intermediates.

• Higher-order telomeric DNA structures, such as DNA compaction and T-loop, are formed through multi-protein interaction networks.

• Twinkle helicase self-loads onto DNA through domain protrusion, and mitochondrial SSB wraps DNA.

• R-loops recruit cohesin and activate PARP1.