Shield strike shatter in DNA topology and nuclease interactions

Abstract

DNA degradation by nucleases is central to genome maintenance, immune defense, and the clearance of extracellular DNA, yet its execution at the single-molecule level remains poorly defined. Here, we use high-speed atomic force microscopy (HS-AFM) to directly visualize the real-time dynamics of DNA digestion by DNase I at nanometer resolution. DNase I repeatedly revisits structurally strained DNA regions before cleavage initiation, revealing a topology-sensitive mode of interaction that is inaccessible to ensemble biochemical approaches. Time-resolved imaging further uncovers a multistep degradation trajectory involving DNA scanning, localized engagement, strand rupture, and progressive filament disassembly, which we term STORM (Scan, Target, Occupy, Rupture, Mobilize) framework. In parallel, we show that protamine-induced DNA condensation into rod- and toroid-like architectures markedly suppresses enzymatic accessibility and stabilizes DNA against nuclease attack. Together, these findings establish a nanoscale structural and kinetic framework for how DNA is either degraded or protected under enzymatic stress, with implications for chromatin biology, innate immunity, autoimmune disease, and the design of nuclease-resistant gene delivery systems.

Researchers visualize how DNA topology governs nuclease interactions in real time. High-speed atomic force microscopy reveals dynamic enzyme scanning, DNA protection by protamine condensation, and topology-dependent cleavage, uncovering nanoscale principles of genome accessibility.

Introduction

Genome integrity is maintained within a highly organized nuclear environment, where the nuclear envelope and nuclear pore complexes (NPCs) define the interface between the genome and the cytoplasm, while chromatin architecture governs the physical organization and accessibility of DNA. DNA integrity is nevertheless continuously threatened by enzymatic degradation, oxidative stress, and mechanical damage¹. To safeguard the genome, cells rely on architectural proteins that condense and shield DNA from such threats. Nowhere is this principle more extreme than in spermatozoa, where histones are replaced by protamines (PRMs), the small, arginine-rich proteins that neutralize DNA charge and fold the paternal genome into one of the most compact and stable chromatin structures known². This unique condensation ensures genetic preservation during transit through hostile reproductive environments. Yet the precise molecular rules that govern how—and when—this compaction confers nuclease resistance remain poorly understood. Is protection solely a function of charge neutralization, or does it require specific three-dimensional geometries, such as rod-like bundles or toroidal loops? How rapidly must condensation occur to defend against active enzymatic threats?

Among nucleases, deoxyribonuclease I (DNase I) is a prototypical double-stranded DNA endonuclease that catalyzes phosphodiester bond cleavage, producing 5′-phosphate and 3′-hydroxyl termini^3–5. It operates under physiological pH and ionic conditions and plays essential roles in both immune homeostasis and clinical intervention. Reduced DNase I activity is a hallmark of autoimmune diseases such as systemic lupus erythematosus (SLE), in which persistent self-DNA drives immune activation^6,7. Conversely, therapeutic DNase I is widely used to reduce extracellular DNA (ecDNA)⁸ accumulation in cystic fibrosis, sepsis, and trauma⁹. This question takes on added urgency in the context of ecDNA, which serves as a potent damage-associated molecular pattern (DAMP) when released from apoptotic, necrotic, or NETotic cells. Once outside the nucleus, ecDNA binds to histones or other nuclear proteins and engages immune sensors such as TLR9 and cGAS–STING, triggering inflammatory responses, autoimmunity, and tissue damage. The persistence of ecDNA in circulation reflects a disrupted balance between DNA release and nuclease clearance, a pathological hallmark of diseases such as systemic lupus erythematosus, atherosclerosis, and various cancers¹⁰. Despite its clinical relevance, the nanoscopic molecular mechanism by which DNase I identifies, engages, and cleaves DNA, and how this activity is influenced by the physical state of the substrate, remains poorly defined.

Over the past three decades, techniques such as X-ray crystallography^4,11, transmission electron microscopy (TEM)^12,13, atomic force microscopy (AFM)^14–17, and fluorescence microscopy^18,19 have been applied to study DNase I–DNA interactions. While these methods have provided valuable structural insights, they are constrained by low temporal resolution, fixation-related artifacts, or indirect labeling strategies, making it challenging to capture the rapid, transient, and topology-sensitive steps of enzymatic engagement in real time²⁰. This lack of mechanistic clarity stems largely from the absence of tools capable of capturing the rapid, topology-sensitive steps of nuclease engagement in real time, leaving the structural basis of DNA susceptibility essentially unobserved.

A molecular-level understanding of these processes is not only fundamental to chromatin biology but also critical for developing therapeutic strategies to modulate inflammation, enhance immune clearance, or improve the stability of gene delivery vectors. To address this gap, we performed HS-AFM nanoimaging^21,22 to directly observe the dynamic interplay between DNA degradation by DNase I and DNA protection by protamine. HS-AFM enables rapid scanning of biological molecules within 100 milliseconds, combining nanometer spatial and sub-second temporal resolution²³. This is made possible by fast feedback control and ultra-short cantilevers, which allow precise imaging without disrupting the structural integrity of target molecules. Importantly, HS-AFM preserves native biomolecular function by eliminating the need for fixation, staining, or crystallization, making it ideal for capturing dynamic biological processes in near-physiological conditions^24,25. Leveraging this technique, we have made significant advances in biomolecular imaging, encompassing high-resolution structural analysis, real-time tracking of conformational changes, and direct observation of dynamic interactions between biomolecules and cellular structures^26–31, enabling real-time visualization of enzyme docking, strand scission, and substrate remodeling^32–34. This approach captures the full trajectory of DNA fate at the single-molecule level, providing a mechanistic framework for understanding nuclease action and architectural protection in real time.

In this study, we used HS-AFM to directly visualize, in real time, the full trajectory of DNA degradation by DNase I and its inhibition by protamine. We show that only rod-like and toroidal protamine DNA structures³³ resist enzymatic attack, whereas naked or loosely coiled DNA is rapidly cleaved at kinks, loops, and free ends. By reconstructing single-molecule degradation events, we reveal how DNA architecture governs enzymatic accessibility and outcome. These findings explain the exceptional stability of sperm chromatin, define structural thresholds for nuclease resistance, and establish HS-AFM as a powerful platform for evaluating DNA stability and nucleoprotein assembly in therapeutic and synthetic systems.

Results

Protamine-induced DNA condensation suppresses DNase I accessibility and cleavage dynamics

To investigate how DNA condensation modulates nuclease susceptibility, we performed real-time HS-AFM imaging of λ-DNA exposed to DNase I in the presence of protamine under near-physiological conditions. Protamine and DNA were incubated for ~10 min prior to deposition onto mica substrates for liquid-phase HS-AFM imaging. Under these conditions, DNA reproducibly condensed into two dominant structural populations: elongated rod-like condensates and toroidal condensates (Fig. 1a, d).

Fig. 1. Real-time HS-AFM visualization of protamine-mediated protection against DNase I digestion.

a–c Rod-like protamine–DNA assemblies resist nuclease-mediated fragmentation. Time-lapse HS-AFM imaging of λ-DNA condensed by protamine into elongated rod-like architectures following DNase I addition under near-physiological conditions. Scale bar, 100 nm. (a). Despite persistent enzyme association near the condensate surface, the overall architecture remained intact throughout continuous imaging without observable contour disruption or fragmentation. Schematic illustration of the proposed steric shielding effect imposed by protamine-mediated condensation (b). Corresponding 3D surface rendering highlights the compact nanoscale morphology of the rod-like assembly and surrounding DNase I particles (c). d–f Toroidal protamine–DNA condensates exhibit enhanced resistance to DNase I accessibility. HS-AFM time series showing toroidal λ-DNA–protamine condensates following DNase I addition. Scale bar, 60 nm. (d). The toroidal architecture remained morphologically preserved throughout prolonged imaging (>6 min) without detectable collapse or fragmentation. Schematic illustration showing restricted DNase I access to the densely packed toroidal interior (e). Corresponding 3D surface rendering further highlights the compact topology and elevated radial thickness of the condensate (f). g Cross-sectional height analysis distinguishes transient DNase I association from condensate disruption. Representative cross-sectional height profile extracted from the indicated region demonstrates nanoscale separation between DNase I particles and the protamine-condensed DNA structure. Localized height elevations corresponding to DNase I association are observed adjacent to the condensate surface, whereas the condensate core maintains a stable height profile without evidence of progressive erosion. Representative height profile measured along the blue arrow. Peak 1 corresponds to a transient DNase I particle, whereas peak 2 represents the protamine-condensed DNA rod, which remains structurally intact. h Time-resolved condensate-height measurements during DNase I exposure. Height measurements obtained from four independent positions on rod-like condensates revealed minimal fluctuation throughout continuous imaging. Orange, blue, red, and gray traces correspond to the color-coded positions above and show stable condensate height during DNase I exposure. No progressive thinning or cleavage-associated height reduction was observed, indicating that protamine-mediated compaction strongly suppresses productive DNase I fragmentation under the present imaging conditions. Source data are provided as a Source Data file.

To determine whether these architectures represent discrete structural states or a continuum of conformations, we performed quantitative analysis of protamine–DNA condensates collected from independent experiments (Supplementary Fig. 1a–c and Supplementary Movies 1, 2). Population analysis revealed a reproducible bimodal distribution in which rod-like (49.0%) and toroidal (37.8%) conformations constituted the dominant structural classes, whereas less compact coil-like intermediates (13.2%) occurred less frequently. These findings indicate that protamine-mediated DNA condensation preferentially stabilizes defined structural states rather than generating a broad stochastic continuum of morphologies.

Time-resolved HS-AFM imaging further captured dynamic structural transitions between these condensate states (Supplementary Fig. 1a, b and Supplementary Movies 1, 2). Rod-like intermediates progressively bent, compacted, and underwent end-to-end closure to form toroidal condensates, indicating that elongated conformations represent kinetically trapped intermediates along a common condensation pathway rather than independent structural species. These observations are consistent with the previously proposed CARD (Coil–Assembly–Rod–Doughnut) model of protamine-mediated DNA condensation and now provide direct single-molecule visualization of these transitions in real time^23,33.

We next examined how these condensed DNA architectures respond to nuclease exposure. Following addition of DNase I into the imaging chamber, individual enzyme particles dynamically localized around both rod-like and toroidal condensates (Fig. 1a,d and Supplementary Movies 3, 4). Notably, despite repeated enzyme association events, condensed DNA structures remained morphologically stable over the observation period.

Rod-like condensates displayed a relatively narrow radial geometry that permitted DNase I access from multiple directions (Fig. 1a–c). In contrast, toroidal condensates exhibited thicker and more compact ring-like organization, and DNase I particles rarely penetrated the central toroidal region (Fig. 1d–f). Three-dimensional surface reconstruction further highlighted the distinct topological accessibility of the two condensate geometries.

To distinguish direct condensate-associated interactions from adjacent enzyme localization, we analyzed cross-sectional height profiles extracted from HS-AFM trajectories (Fig. 1g). DNase I molecules associated with rod-like condensates produced continuous height transitions between the enzyme particle and condensate surface, consistent with direct physical interaction. By contrast, non-associated particles appeared as spatially separated peaks. Importantly, even during prolonged association events, condensate height and morphology remained stable, with no detectable contour erosion, collapse, or fragmentation.

Time-resolved height measurements acquired at multiple positions on toroidal condensates similarly revealed highly stable structural profiles over time (Fig. 1h). No progressive thinning, rupture, or localized collapse was observed during continuous imaging. In addition, tip-induced structural perturbation was minimal, as toroidal condensates retained their overall morphology throughout prolonged HS-AFM acquisition periods of up to ~6 min.

Together, these observations indicate that protamine condensation markedly suppresses productive DNase I-mediated degradation despite transient enzyme association. Importantly, HS-AFM imaging reveals that enzyme localization alone is insufficient to produce detectable cleavage under highly compacted DNA states, suggesting that condensate architecture strongly restricts access to cleavage-compatible DNA conformations.

To further distinguish geometric shielding from electrostatic inhibition, we next performed HS-AFM imaging of partially relaxed protamine–DNA architectures under conditions that promoted local structural loosening without complete condensate dissociation (Supplementary Fig. 2 and Supplementary Movie 5). Under these conditions, previously protected DNA structures progressively became susceptible to DNase I-mediated fragmentation.

Importantly, cleavage-associated structural disruption became detectable only after local loosening or partial opening of the condensed architecture was observed. DNase I did not immediately digest densely compact protamine–DNA condensates following transfer into the imaging conditions, indicating that restoration of enzymatic accessibility correlates primarily with local contour exposure and condensate relaxation rather than solution conditions alone. Notably, highly compact condensates that remained structurally preserved continued to exhibit prolonged resistance to nuclease-mediated degradation (Fig. 1 and Supplementary Movies 3–5).

Together, these observations suggest that although electrostatic interactions contribute to condensate stabilization, the dominant barrier to DNase I-mediated degradation arises from geometric and topological inaccessibility imposed by dense higher-order DNA compaction. HS-AFM imaging of partially relaxed protamine–DNA architectures (Supplementary Fig. 2) further revealed that DNase I susceptibility increased preferentially at locally exposed or structurally loosened DNA regions, whereas densely compact domains frequently remained resistant. These findings support a model in which electrostatic interactions primarily stabilize condensate architecture, while productive nuclease engagement depends strongly on local contour exposure and structural accessibility.

Collectively, these experiments demonstrate that protamine-mediated DNA condensation substantially suppresses nuclease-mediated degradation by limiting access to structurally permissive DNA conformations. More broadly, the results establish that DNA architecture itself critically influences enzymatic susceptibility and illustrate how real-time HS-AFM imaging can directly resolve the dynamic interplay between genome compaction and nuclease accessibility at the single-molecule level.

Substrate-dependent HS-AFM visualization reveals distinct DNase I fragmentation behavior and detectability-dependent signal decay

To investigate how surface confinement influences nuclease-mediated DNA degradation under HS-AFM conditions, we monitored real-time DNase I digestion of λ-DNA on three substrates with distinct DNA–surface interaction properties: Ni²⁺-functionalized mica, APTES-modified mica, and supported lipid bilayers (Fig. 2a–c and Supplementary Movies 6–11). These substrates provide different degrees of adsorption strength, filament mobility, and conformational confinement.

Fig. 2. Substrate-dependent DNase I fragmentation dynamics visualized by real-time HS-AFM.

a–c Time-lapse HS-AFM imaging of DNase I-mediated λ-DNA fragmentation on substrates with distinct DNA–surface interaction properties. a Ni²⁺-functionalized mica. DNA forms partially adsorbed filamentous networks that progressively fragment during imaging. Cleavage-associated contour disruption is accompanied by the redistribution and gradual reduction of detectable filamentous structures. Quantification of detectable DNA-covered area over time is shown at right. Scale bar, 200 nm. b APTES-functionalized mica. DNA exhibits stronger surface immobilization and reduced lateral mobility relative to Ni²⁺-supported substrates. DNase I exposure induces progressive contour fragmentation, with fragmentation products remaining spatially confined near the original adsorption regions. Quantification of detectable DNA-covered area over time is shown at right. Scale bar, 100 nm. c Supported lipid bilayer substrate. DNA displays increased lateral mobility and dynamic contour rearrangement on lipid-supported surfaces. DNase I exposure induces rapid contour disruption and progressive conversion of extended filaments into fragmented and weakly detectable punctate structures. Quantification of detectable DNA-covered area over time is shown at right. Scale bar, 200 nm. Color bars indicate height in nanometers. d–f Quantitative normalized corrected-height intensity (NCHI) analysis of substrate-dependent fragmentation dynamics. NCHI values extracted from independent HS-AFM trajectories were plotted over time and fitted using a one-phase exponential decay model under the present imaging conditions. Ni²⁺-supported substrates exhibited gradual apparent decay (d); APTES-supported substrates displayed variable fragmentation dynamics (e); Lipid-supported substrates exhibited the most rapid apparent decay and greatest trajectory heterogeneity (f); consistent with increased DNA mobility and reduced surface confinement respectively. Source data are provided as a Source Data file. Light-colored traces represent individual measurements, whereas red (NiCl₂), blue (APTES), and green (lipid) curves indicate exponential fits.

Control imaging performed in the absence of DNase I showed no detectable contour disruption or filament fragmentation under any substrate condition (Supplementary Fig. 3 and Supplementary Movies 6-8), confirming that the observed structural changes primarily reflect nuclease-mediated processing rather than tip-induced damage.

Across all substrates, DNase I exposure induced progressive disruption of filamentous DNA structures, although the observable fragmentation behavior differed substantially depending on substrate chemistry and molecular confinement (Fig. 2a–c). On Ni²⁺-functionalized mica, DNA formed partially adsorbed filamentous networks that gradually fragmented over time. Cleavage-associated contour disruption frequently emerged near exposed termini, curved regions, or locally bent conformations (Supplementary Fig. 4g). Fragment redistribution remained relatively constrained under these moderately confined conditions.

In contrast, DNA adsorbed onto APTES-modified mica exhibited reduced large-scale mobility, with fragmentation products remaining spatially confined near the original adsorption regions (Fig. 2b and Supplementary Movie 10). DNA deposited onto supported lipid bilayers displayed markedly greater lateral mobility and dynamic contour rearrangement (Fig. 2c and Supplementary Movie 11). Under these conditions, DNase I exposure induced rapid contour disruption and progressive conversion of extended DNA filaments into highly dynamic punctate structures. Considerable trajectory-to-trajectory heterogeneity was observed on lipid-supported substrates.

To quantitatively compare substrate-dependent fragmentation behavior, we measured projected contour area, normalized corrected-height intensity (NCHI), and background-corrected integrated density over time (Fig. 2d–f). All substrates exhibited a progressive reduction in detectable DNA signal, although the apparent decay profiles differed depending on substrate confinement and molecular mobility.

Because DNase I cleavage fragments long DNA molecules into progressively shorter products without eliminating total DNA mass, we next examined whether signal reduction reflects fragmentation below the effective HS-AFM detectability limit. Calibration experiments using DNA fragments of defined lengths deposited at identical mass concentrations on APTES-supported substrates revealed strong length-dependent detectability under HS-AFM conditions (Supplementary Fig. 5 and Supplementary Movies 18–21). Whereas λ-DNA and 2.5 kbp fragments remained readily detectable, shorter fragments such as 300 bp DNA exhibited weak and discontinuous signals, and 35 bp oligonucleotides became largely indistinguishable from background noise.

Consistent with this interpretation, time-resolved HS-AFM imaging revealed progressive conversion of extended DNA filaments into fragmented nanoscale puncta and redistributed DNA material across all substrate systems (Supplementary Fig. 4 and Supplementary Movies 12–17).

Together, these observations indicate that the apparent degradation behavior observed by HS-AFM reflects an interplay between enzymatic fragmentation, local DNA topology, substrate-dependent confinement, and imaging detectability. More broadly, these findings highlight the ability of HS-AFM to directly visualize transient fragmentation behavior and nanoscale structural heterogeneity during real-time enzymatic DNA degradation.

Structural organization and nanoscale dynamic behavior of DNase I visualized by HS-AFM

To establish a structural and biophysical framework for interpreting DNase I behavior during real-time DNA degradation, we combined computational structural analysis with single-particle HS-AFM imaging under liquid conditions (Fig. 3 and Supplementary Movie 22).

Fig. 3. Structural, electrostatic, and nanoscale dynamic characterization of DNase I by modeling and HS-AFM.

a Schematic drawings of domain organization of bovine DNase I showing annotated DNA-binding (red), actin-binding (cyan), calcium binding-associated site (black), and disulfide bond (orange) regions across the 282-amino-acid sequence. b AlphaFold3 structural model of DNase I. The predicted structure reveals a folded globular architecture with spatial clustering of DNA-binding and catalytic regions. Functional regions are colored as in (a). c Predicted net charge of DNase I as a function of pH, indicating an overall negative charge at physiological pH (7.4; Z = −7.9). d, e Intrinsic disorder prediction of DNase I. Multiple algorithms predict a predominantly folded protein with minimal intrinsic disorder (IDR = 1.06%). Colored lines in (d) represent individual prediction algorithms, whereas (e) shows the consensus disorder score. f Simulated AFM topograph of DNase I generated from the AlphaFold3 structural model under tip-convolution conditions, predicting a nanoscale particle morphology with height features comparable to HS-AFM observations. g Representative HS-AFM image of an individual DNase I molecule. The experimentally observed particle in (h) exhibits a morphology and height range consistent with the simulated AFM projection. Scale bar, 10 nm. h Time-lapse HS-AFM imaging of a single DNase I molecule. Sequential images reveal low-amplitude nanoscale fluctuations and subtle morphological variability over time under liquid-phase imaging conditions. Scale bar, 10 nm. i Height distribution of DNase I particles measured by HS-AFM. Histogram of peak particle heights (n = 100) showing a dominant distribution centered at ~4–6 nm, consistent with folded DNase I particles imaged in liquid. j Representative single-particle height trajectories over time. Individual DNase I molecules exhibit non-periodic nanoscale height fluctuations during imaging, consistent with limited conformational or orientational variability of surface-associated particles rather than direct visualization of active-site rearrangements. Different colored traces represent independent DNase I molecules. Source data are provided as a Source Data file.

Domain mapping of bovine DNase I revealed spatial organization of DNA-binding, actin-binding, calcium-binding, and catalytic regions across the 282-amino-acid sequence (Fig. 3a). AlphaFold3 structural prediction further revealed a predominantly folded globular architecture with clustering of catalytic and DNA-interacting surfaces (Fig. 3b).

To further establish the structural basis underlying the observed DNase I interaction behaviors, we integrated domain annotation, structural modeling, docking analysis, and AFM topography simulation (Supplementary Fig. 6). Mapping of DNA-binding and catalytically associated regions onto the full-length DNase I structure revealed accessible substrate-engagement surfaces surrounding the catalytic cleft. Structural docking analyses further suggested that local DNA geometry and contour accessibility may influence productive enzyme engagement by modulating exposure of DNA-contacting interfaces.

Importantly, forward AFM simulations derived from the structural models closely resembled the experimentally observed particle morphology, supporting the interpretation that the HS-AFM particles correspond to intact folded DNase I molecules. Together, these analyses provide a structural framework consistent with the topology-associated interaction dynamics observed during HS-AFM imaging, while also emphasizing that catalytically productive conformations cannot be directly resolved from imaging or docking analyses alone.

Electrostatic analysis predicted an overall negative net charge at physiological pH (Z = −7.9 at pH 7.4), while retaining localized surface regions potentially compatible with transient DNA engagement (Fig. 3c).

To evaluate intrinsic structural flexibility, we next performed disorder-prediction analyses using multiple independent algorithms. DNase I exhibited uniformly low disorder propensity across most of the sequence, with limited predicted flexibility restricted primarily to short loop and terminal regions (Fig. 3d, e). The averaged disorder profile indicated an overall intrinsic disorder content of only ~1.06%, consistent with a predominantly folded enzyme architecture.

To assess whether the experimentally observed HS-AFM particles were morphologically compatible with folded DNase I molecules, we generated AFM-compatible structural simulations from the AlphaFold3 model. Simulated topographs predicted compact nanoscale particles with height distributions and lateral morphology closely resembling experimentally observed HS-AFM images (Fig. 3f, g; n = 194 particles).

Single-particle HS-AFM imaging under liquid conditions further revealed that DNase I molecules exhibit measurable nanoscale fluctuations over time (Fig. 3h and Supplementary Movie 22). Height distribution analysis of individual particles (n = 100 particles) revealed a dominant peak centered at ~5.1 ± 1.0 nm (mean ± SD), consistent with folded DNase I molecules imaged under hydrated conditions (Fig. 3i). Time-resolved height trajectories demonstrated low-amplitude non-periodic fluctuations during imaging (Fig. 3j), suggesting limited conformational or orientational variability of surface-associated particles.

Importantly, HS-AFM does not directly resolve active-site rearrangements or catalytic conformational transitions. Nevertheless, the observed nanoscale fluctuations indicate that DNase I behaves as a dynamically adaptable nanoscale particle rather than a rigid static structure under liquid imaging conditions.

Together, these analyses establish that DNase I is visualized by HS-AFM as a predominantly folded yet dynamically fluctuating enzyme with structural features compatible with transient and repeated DNA engagement during nuclease-mediated degradation.

Quantitative HS-AFM analysis reveals recurrent and topology-dependent DNase I engagement dynamics

To investigate how DNase I dynamically engages DNA during degradation, we performed time-resolved HS-AFM imaging of λ-DNA under near-physiological conditions (Fig. 4 and Supplementary Movies 23, 24). Across independent trajectories, cleavage-associated fragmentation was not uniformly distributed along the DNA contour but instead frequently emerged near exposed termini, curved segments, and locally bent conformations.

Fig. 4. HS-AFM visualization of recurrent and topology-dependent DNase I engagement during DNA degradation.

a, b Time-lapse HS-AFM imaging of λ-DNA undergoing DNase I-mediated degradation under liquid-phase conditions. DNase I molecules appear as transient globular features (~4–5 nm apparent height) dynamically associating with DNA contours. Colored numbered circles indicate representative loci exhibiting repeated local enzyme engagement across multiple frames. Cleavage-associated fragmentation frequently emerges near curved, exposed, or locally vulnerable DNA regions, leading to progressive contour disruption and near-complete degradation by ~182 s in (a) and ~238 s in (b). Scale bar, 10 nm. c, d Time-resolved height traces extracted from the color- and number-matched regions indicated in (a, b), respectively. Transient height increases correspond to localized DNase I association events, whereas decreases toward baseline reflect enzyme dissociation, contour redistribution, or cleavage-associated fragmentation. Repeated transient height peaks at specific loci indicate recurrent local engagement rather than uniform stochastic sampling along the DNA contour. Scale bar, 20 nm. Source data are provided as a Source Data file.

DNase I particles appeared as transient globular features dynamically associating with DNA filaments during degradation (Fig. 4a, b). Individual enzyme molecules frequently localized to restricted contour regions over multiple frames, often exhibiting repeated engagement events prior to detectable contour disruption. Corresponding height-trace analyses revealed localized occupancy states, recurrent contour-associated interactions, and cleavage-associated redistribution dynamics over time (Fig. 4c, d).

To quantitatively evaluate whether these recurrent localization events exceeded stochastic expectation, we performed integrated rebinding analyses across multiple independent trajectories (Supplementary Figs. 7–9). Spatial reassignment analysis revealed that DNase I detections repeatedly clustered within restricted contour regions rather than distributing uniformly across the imaging field. For example, in the trajectory corresponding to Fig. 4a, 47 individual detections collapsed into 23 distinct interaction sites, yielding a rebinding frequency of 37.8% (Supplementary Fig. 7).

Randomized simulations demonstrated that experimentally observed revisiting frequencies substantially exceeded unconstrained two-dimensional stochastic expectation. However, contour-constrained simulations additionally revealed that one-dimensional DNA geometry itself strongly elevates the baseline probability of repeated encounters under surface-supported imaging conditions. These analyses therefore indicate that recurrent DNase I engagement reflects a combination of topology-associated localization and geometric confinement rather than deterministic positional memory.

We next mapped enzyme localization and cleavage events relative to contour position and local DNA topology (Supplementary Figs. 10 and 11). DNase I localization exhibited significant enrichment near exposed DNA termini and highly curved regions compared with randomized controls. These positional biases were reproducibly observed across independent trajectories, indicating that topology-associated localization represents a robust feature of DNase I behavior.

Dwell-time analysis further revealed enrichment of longer-lived DNase I interactions at loci exhibiting recurrent localization and cleavage-associated engagement (Supplementary Fig. 12). Extended dwell events (>2 consecutive frames) were preferentially observed near curved and cleavage-prone regions, consistent with transient local stabilization prior to fragmentation.

Time-resolved HS-AFM trajectories additionally revealed heterogeneous cleavage progression dynamics. In some trajectories, contour rupture was followed by rapid fragment separation, whereas in others, cleavage-associated discontinuities remained transiently localized prior to redistribution (Supplementary Movies 25 and 30-31). Rare trajectories also revealed transient multi-enzyme occupancy within localized DNA regions.

Our HS-AFM data show that DNase I localization and cleavage-associated rupture events are enriched at regions of local DNA deformation where the enzyme exhibits persistent and topology-associated interactions. These observations are consistent with a model in which local DNA geometry increases the probability of forming catalytically productive configurations. However, because HS-AFM does not resolve active-site geometry, this relationship should be interpreted as correlative rather than directly resolved at the molecular level. Together, these analyses support a model in which DNase I dynamically samples locally exposed DNA conformations through recurrent topology-associated engagement rather than uniform stochastic contour scanning. Based on these reproducible spatiotemporal behaviors, we therefore summarize these observed interaction patterns using a hypothetical STORM framework—Scan, Target, Occupy, Rupture, and Mobilize—as a probabilistic description of the dynamic states associated with DNA fragmentation (Fig. 5 and Supplementary Figs 4 and 5).

Fig. 5. Cross-nuclease visualization and conceptual interpretation of the proposed STORM framework for nuclease-mediated DNA degradation.

a Conceptual schematic summarizing the proposed STORM sequence—Scan, Target, Occupy, Rupture, and Mobilize—derived from recurrent interaction behaviors observed during HS-AFM imaging of nuclease-mediated DNA degradation. The schematic illustrates transient DNA sampling, localized engagement, cleavage-associated contour disruption, and subsequent fragment redistribution dynamics. b Representative HS-AFM trajectory of DNase I interacting with DNA during progressive degradation. Sequential frames illustrate repeated local enzyme engagement, contour disruption, and redistribution of fragmented DNA regions consistent with the proposed STORM sequence. Scale bar, 20 nm (n = 8 molecules). c Kymograph analysis extracted from representative DNase I trajectories showing the temporal progression of localized interaction states along a DNA contour region. Distinct phases corresponding to scanning, localized occupancy, cleavage-associated rupture, and post-cleavage redistribution are indicated. d Time-resolved height trace corresponding to the kymograph region shown in (c). Transient height increases reflect localized DNase I association events, whereas signal reduction corresponds to cleavage-associated contour disruption and fragment redistribution. Numbered peaks correspond to the interaction states annotated in (c). Source data are provided as a Source Data file. e Conceptual schematic illustrating the corresponding STORM-like interaction sequence observed for micrococcal nuclease (MNase), despite major differences in catalytic mechanism and substrate preference relative to DNase I. f Representative HS-AFM trajectory of MNase interacting with DNA during degradation. Similar interaction behaviors, including localized engagement, contour disruption, and fragment redistribution, were reproducibly observed during MNase-mediated degradation. Scale bar, 20 nm (n = 3 molecules). In the schematics, blue and purple particles represent DNase I and MNase, respectively, and DNA is shown as a red–blue double helix.

STORM-like interaction dynamics extend to mechanistically distinct nucleases

To determine whether the observed interaction dynamics are specific to DNase I or more broadly conserved, we extended our HS-AFM analyses to micrococcal nuclease (MNase)³⁵, an endonuclease with distinct catalytic properties and substrate preferences (Fig. 5e, f and Supplementary Movies 27, and 30, 31).

Despite these biochemical differences, MNase exhibited remarkably similar nanoscale interaction behaviors. Real-time HS-AFM imaging revealed transient contour sampling, localized engagement near exposed or curved DNA regions, repeated association events, cleavage-associated contour disruption, and subsequent fragment redistribution dynamics.

To further support the temporal organization of these interaction behaviors, we performed kymograph-based trajectory analysis together with time-resolved height profiling of representative nuclease–DNA interaction sites (Fig. 5c, d). These analyses revealed sequential transitions between transient scanning, localized occupancy, cleavage-associated rupture, and post-cleavage redistribution states along restricted DNA contour regions. Recurrent localized height increases corresponded to repeated nuclease association events prior to contour disruption, supporting the interpretation that nuclease engagement occurs through dynamically enriched local interaction states rather than uniform stochastic contour sampling.

To validate that the observed particles correspond to folded enzyme molecules, we integrated structural modeling, AFM simulation, and experimental HS-AFM characterization (Supplementary Fig. 13 and Supplementary Movie 26). Simulated AFM projections generated from the MNase crystal structure closely resembled experimentally observed particle morphology, and measured particle-height distributions were consistent with folded monomeric MNase under liquid imaging conditions.

Although individual trajectories remained heterogeneous in timing, revisiting frequency, and fragmentation behavior, DNase I and MNase reproducibly exhibited similar topology-associated interaction dynamics. These observations support the interpretation that STORM represents a probabilistic framework describing recurrent nuclease interaction states during DNA fragmentation rather than a nuclease-specific phenomenon (Figure 4, Supplementary Fig. 14 and Supplementary Movies 24, 28-31).

Collectively, the integration of recurrent engagement analysis, topology-dependent spatial mapping, dwell-time characterization, kymograph-based trajectory analysis, randomized simulations, and cross-nuclease validation establishes a quantitative nanoscale framework for interpreting dynamic nuclease–DNA interactions in real time under near-native conditions. We therefore interpret STORM as a topology-associated interaction framework rather than direct visualization of catalytic intermediates.

Discussion

The integrity of genetic information depends on a dynamic balance between enzymatic accessibility and higher-order structural protection³⁶. Although the biochemical consequences of nuclease-mediated degradation are well established, the nanoscale dynamics governing how nucleases locate, engage, and process DNA in real time have remained difficult to visualize directly. By combining HS-AFM nanoimaging with quantitative trajectory analysis, our study establishes a single-molecule framework linking DNA topology, structural accessibility, and nuclease-mediated fragmentation dynamics.

A central finding of this study is that protamine-mediated DNA condensation generates architectures that are highly resistant to DNase I digestion. Rod-like and toroidal condensates remained structurally preserved during prolonged HS-AFM imaging, even under nuclease-rich conditions (Fig. 1 and Supplementary Movies 1–3). Ionic-strength perturbation experiments further demonstrated that nuclease susceptibility was restored only after partial decondensation, indicating that protection arises predominantly from geometric inaccessibility rather than electrostatic inhibition alone. These observations are consistent with the previously proposed CARD (Coil–Assembly–Rod–Doughnut) model^23,33,37 and support the idea that densely packed DNA architectures suppress nuclease accessibility by limiting exposed termini, local curvature, and conformational flexibility.

In contrast, unprotected DNA exhibited progressive and highly heterogeneous fragmentation behavior that depended strongly on local topology and physical confinement. Across independent trajectories, cleavage-associated disruption frequently emerged near exposed termini, curved regions, and locally bent conformations. Dwell-time analyses further revealed prolonged nuclease residence at these loci relative to other DNA regions, suggesting that local DNA geometry strongly influences where nucleases preferentially engage and where fragmentation is most likely to occur.

Importantly, HS-AFM does not resolve active-site conformations or directly distinguish productive from nonproductive binding states. Accordingly, our data should not be interpreted as direct visualization of catalytic intermediates. Rather, the observed relationship between local DNA deformation, recurrent engagement, dwell-time stabilization, and subsequent fragmentation represents a robust spatiotemporal correlation consistent with established biochemical and structural models of nuclease catalysis. In this context, our observations do not redefine the chemistry of DNA cleavage but instead reveal how catalytically permissive DNA conformations may be dynamically accessed at the single-molecule level.

Our HS-AFM analyses reveal strong spatiotemporal coupling between local DNA topology, recurrent nuclease engagement, and cleavage-associated contour disruption. However, because HS-AFM does not directly visualize catalytic conformations or phosphodiester-bond chemistry, we cannot distinguish productive from nonproductive binding events at the molecular level. Accordingly, the proposed STORM framework should be interpreted as a probabilistic and mechanistically consistent interaction model inferred from reproducible spatiotemporal correlations rather than direct visualization of catalytic intermediates.

To summarize these observations, we introduce a hypothetical STORM framework—Scan, Target, Occupy, Rupture, and Mobilize—as a probabilistic description of the interaction states associated with nuclease-mediated DNA degradation (Fig. 5). Across independent trajectories, transient contour sampling, localized engagement, cleavage-associated rupture, and fragment redistribution emerged as reproducible spatiotemporal behaviors during degradation. STORM should therefore be interpreted as a conceptual framework inferred from recurrent localization patterns, dwell-time enrichment, and cleavage-associated redistribution rather than as direct visualization of catalytic progression or deterministic mechanistic states.

Our analyses further revealed that the interpretation of recurrent nuclease engagement depends strongly on the geometric assumptions of the underlying null model. While experimentally observed rebinding frequencies exceeded unconstrained two-dimensional stochastic expectation, contour-constrained simulations demonstrated that one-dimensional DNA geometry itself markedly elevates the baseline probability of repeated encounters under surface-supported imaging conditions. These findings indicate that recurrent engagement emerges from the combined influence of DNA topology, local geometric accessibility, and contour-constrained diffusion rather than strict positional memory.

Extension of the analyses to micrococcal nuclease (MNase) further supports the interpretation that STORM reflects a broader interaction framework rather than a DNase I-specific phenomenon. Despite distinct catalytic properties and substrate preferences, both nucleases exhibited similar localized engagement behaviors, including transient contour sampling, recurrent association, and cleavage-associated redistribution. Structural modeling and AFM simulation additionally supported the experimentally observed MNase particles correspond to folded enzyme structures compatible with DNA engagement.

More broadly, the present observations align with the emerging view that nucleases exhibit extensive mechanistic and structural diversity while nevertheless converging onto common physical principles of nucleic-acid engagement. Comprehensive comparative analyses of nuclease families have shown that distinct biological functions can be mediated by fundamentally different catalytic architectures, metal-ion dependencies, and tertiary folds, whereas similar catalytic strategies may operate across unrelated biological pathways³⁸. In this context, the similar nanoscale interaction behaviors observed here for DNase I^3,4 and MNase³⁵ suggest that recurrent sampling of structurally accessible DNA conformations may represent a shared higher-order physical feature of nuclease action that exists upstream of catalytic chemistry itself. Our HS-AFM analyses, therefore, complement biochemical and structural classifications of nucleases by providing a dynamic single-molecule perspective on how structurally distinct enzymes physically engage heterogeneous DNA substrates in real time.

Our results also demonstrate that enzymatic degradation is strongly influenced by the physical microenvironment. DNA mobility, adsorption strength, and surface confinement substantially altered apparent fragmentation behavior across Ni²⁺-mica, APTES-modified surfaces, and supported lipid bilayers. Calibration experiments further showed that progressive signal reduction during degradation largely reflects fragmentation below the effective HS-AFM detectability threshold rather than the disappearance of DNA mass itself. These findings underscore the importance of considering detectability limits and substrate-dependent confinement when interpreting real-time polymer degradation imaging experiments.

Finally, HS-AFM imaging combined with AlphaFold3 structural prediction and AFM simulation revealed that DNase I behaves as a compact yet dynamically adaptable nanoscale particle under liquid imaging conditions. Observed nanoscale height fluctuations and localized mobility suggest limited conformational flexibility that may facilitate repeated sampling of structurally heterogeneous DNA conformations during degradation. Together, our findings establish a quantitative nanoscale framework for understanding how higher-order DNA architecture, local topology, and structural accessibility regulate nuclease-mediated degradation in real time. More broadly, this study highlights the ability of HS-AFM to bridge static structural information with dynamic enzymatic behavior at the single-molecule level under near-native conditions, providing a general approach for resolving how DNA organization regulates molecular accessibility, enzymatic attack, and ultimately DNA persistence or clearance.

Methods

Materials

Linear double-stranded λ-DNA (48,502 bp) was purchased from Nippon Gene Co., Ltd. (Tokyo, Japan) and diluted in Milli-Q water prior to HS-AFM experiments. DNase I from bovine pancreas was obtained from Abnova (Cat. No. P5221; Lot No. 51J21494). Micrococcal nuclease (MNase) was purchased from Takara Bio Inc. (Cat. No. 2910 A). Synthetic 35-mer DNA oligonucleotides (GAGAGGGGCGACCGGTTTAAAAGTTTATCTCGCCG) were purchased from Eurofins Genomics. Additional DNA fragments, including ~300 bp and ~2500 bp DNA fragments, were prepared as previously described^32,33. Unless otherwise indicated, all experiments were performed using linear λ-DNA as the primary substrate for HS-AFM imaging and degradation analyses.

Computational characterization of DNase I structural properties

The net charge state of DNase I at physiological pH (7.4) was estimated using the ProtPi server (https://www.protpi.ch/), which calculates theoretical isoelectric points and protein charge distributions based on amino acid composition.

Intrinsic disorder propensity within the DNase I amino acid sequence was evaluated using six independent disorder-prediction algorithms, including IUPred (short and long modes), PONDR VL3, PONDR VLXT, PONDR VSL2, and PrDOS. Prediction outputs from the individual algorithms were combined to generate a composite disorder profile as previously described32. These analyses were used to assess the overall structural organization and potential flexible regions within DNase I relevant to dynamic enzyme–DNA interactions observed by HS-AFM.

Structural modeling, visualization, and AFM-compatible surface rendering

Structural visualization and molecular surface analyses were performed using PyMOL (Schrödinger, LLC). Crystal structures of Bos taurus (Bovine) DNase I (PDB: 2DNJ) and micrococcal nuclease (MNase; PDB: 1SNC) were used for structural representation, electrostatic interpretation, and AFM-compatible surface visualization. Structural figures shown in Supplementary Figs. 6 and 13 were generated using PyMOL with standard ribbon, surface, and electrostatic rendering modes. Putative DNA-binding interfaces and catalytic surface orientations were interpreted based on previously reported structural information together with qualitative docking geometry relative to DNA substrates. These analyses were used primarily to support structural compatibility between experimentally observed HS-AFM interaction behaviors and known nuclease architectures, rather than to assign atomically resolved catalytic intermediates.

For AFM-compatible structural comparison, molecular surface representations were oriented to approximate experimentally observed particle geometries and were compared qualitatively with HS-AFM topographic features and simulated AFM projections. Structural dimensions and surface accessibility were interpreted in the context of experimentally measured particle morphology and height distributions.

Structural modeling using AlphaFold3

Structural models of bovine DNase I were generated using the AlphaFold3 server (https://alphafoldserver.com/) based on the full-length amino acid sequence of DNase I. Predicted structures were exported in PDB format and used for structural visualization, surface analysis, AFM-compatible orientation, and qualitative comparison with experimentally observed HS-AFM particle morphology. Predicted structural models were interpreted in combination with previously reported crystallographic information and were used primarily to assess overall molecular architecture, domain organization, and potential surface accessibility relevant to DNA interaction. Structural representations and surface visualizations were generated using PyMOL (Schrödinger, LLC). Because the present study focuses on dynamic single-molecule imaging rather than atomically resolved catalytic-state assignment, structural models were used as qualitative interpretative frameworks rather than definitive representations of catalytically competent enzyme–DNA complexes.

High-speed AFM Imaging of protamine–DNA complex formation

Protamine (PRM) and λ-DNA (48.5 kbp) were prepared in imaging buffer containing 50 mM Tris–HCl, 150 mM NaCl, and 3 mM MgCl₂ (pH 7.4), unless otherwise indicated. All HS-AFM experiments were performed in liquid at room temperature. Freshly cleaved mica was used as the standard negatively charged substrate for adsorption-based imaging. For PRM–DNA complex formation, PRM and λ-DNA were premixed at defined charge ratios in imaging buffer and incubated on ice for 10 min prior to surface deposition. The mixtures were then applied onto bare mica and allowed to adsorb for 10 min, followed by gentle rinsing with imaging buffer to remove unbound material before imaging. To directly visualize DNA compaction dynamics, λ-DNA was first deposited onto either bare mica or positively charged supported lipid bilayer substrates. Supported lipid bilayers were prepared by sonication of DPPC/DPTAP/biotin-cap-DPPE mixtures (89:10:1, w/w; 0.2 mg/mL) in MgCl₂-containing buffer for 5 min using a bath sonicator, followed by deposition onto mica to generate positively charged supported membranes that reduce strong DNA–substrate interactions. After DNA adsorption and gentle rinsing to remove unbound molecules, PRM-containing imaging buffer was introduced during continuous HS-AFM acquisition to visualize real-time DNA condensation dynamics as previously described^32,33. All experiments were independently repeated at least five times to confirm the reproducibility of the observed structural and dynamic behaviors.

High-speed AFM imaging of DNase I activity

Real-time imaging of DNase I-mediated DNA degradation was performed using a custom-built high-speed atomic force microscope (HS-AFM) operated in tapping mode, as previously described. Imaging was conducted under liquid conditions at room temperature using ultra-short cantilevers, including BL-AC10DS-A2 (Olympus, Tokyo, Japan) and USC-F1.2-k0.15 or USC-F1.5-k0.6 (NanoWorld, Neuchâtel, Switzerland), with nominal spring constants of 0.1–0.6 N/m and resonance frequencies in liquid of ~1.2–1.5 MHz. Cantilever tips were sharpened by electron-beam deposition of amorphous carbon using a field-emission scanning electron microscope (ELS-7500; Elionix Inc., Tokyo, Japan). Cantilever deflection was detected using a 670 nm laser focused through a 20× long-working-distance objective lens onto a two-segment PIN photodiode. To minimize force-induced perturbation during imaging, the free oscillation amplitude (A_0) was typically maintained between 1.5 and 2.5 nm, and the set-point amplitude was adjusted to 80–90% of A_0. Unless otherwise indicated, DNase I imaging experiments were performed in buffer containing 10 mM Tris–HCl and 2 mM MgCl₂ (pH 7.4). Depending on the experiment, freshly cleaved mica, Ni²⁺-functionalized mica, APTES-modified mica, or supported lipid bilayers were used to modulate DNA–surface interactions. λ-DNA was deposited onto substrates, gently rinsed with imaging buffer to remove unbound molecules, and imaged continuously after introduction of DNase I into the liquid chamber.

Large-field degradation assays were typically acquired over scan areas of ~500–1000 nm with frame intervals of ~1–3 s, whereas high-magnification single-enzyme tracking experiments were acquired over smaller scan areas of ~50–150 nm with frame intervals of ~0.3–1.0 s per frame. Pixel resolution and scan conditions were optimized according to experimental purpose. Control experiments performed in the absence of DNase I confirmed that λ-DNA filaments remained structurally stable during continuous HS-AFM imaging under the conditions used in this study, indicating that the fragmentation observed in enzyme-containing experiments primarily reflected nuclease-mediated degradation rather than tip-induced damage. All experiments were independently repeated at least five times using freshly prepared substrates and enzyme solutions. Representative trajectories shown in the figures were selected based on image stability, molecular clarity, and reproducibility across independent datasets.

Surface preparation, molecule binding, and imaging conditions

Three substrate systems with distinct DNA–surface interaction properties were used throughout this study: Ni²⁺-functionalized mica, APTES-modified mica, and supported lipid bilayers. These substrates were selected to modulate DNA adsorption strength, filament mobility, and local surface confinement during HS-AFM imaging.

For Ni²⁺-functionalized substrates, freshly cleaved mica was incubated with 1 mM NiCl₂ solution for 10 min prior to DNA deposition to promote moderate electrostatic tethering of negatively charged DNA while preserving partial contour flexibility. For APTES-modified substrates, mica surfaces were functionalized with 3-aminopropyltriethoxysilane (APTES) to generate positively charged surfaces that enhance DNA immobilization and reduce large-scale filament drift during imaging.

Supported lipid bilayers were prepared by sonication of DPPC/DPTAP/biotin-cap-DPPE mixtures (89:10:1, w/w; 0.2 mg/mL) in MgCl₂-containing buffer for 5 min, followed by deposition onto mica substrates to form positively charged supported membranes. This platform reduced excessive DNA–surface interactions and preserved higher filament mobility under liquid conditions.

Unless otherwise indicated, λ-DNA (48.5 kbp) was diluted to ~1.5–2.5 ng μL⁻¹ in imaging buffer containing 10 mM Tris–HCl and 2 mM MgCl₂ (pH 7.4–7.5). DNA samples were deposited onto the designated substrates for ~10 min, followed by gentle rinsing with imaging buffer to remove unbound molecules while preserving surface-associated DNA filaments.

For DNase I degradation experiments, DNA-coated substrates were equilibrated in imaging buffer prior to enzyme introduction. DNase I was diluted immediately before use and introduced directly into the HS-AFM liquid chamber during continuous scanning. Unless otherwise indicated, DNase I was used at ~100 ng μL⁻¹. Continuous HS-AFM imaging was then performed to capture DNA degradation dynamics, including transient enzyme association, contour fragmentation, and post-cleavage redistribution events.

Imaging was performed in tapping mode under liquid conditions using high-speed cantilevers. Depending on scan area and experimental purpose, frame acquisition rates typically ranged from ~1–3 frames s⁻¹, with higher temporal resolution (≥3 frames s⁻¹) used for rapid cleavage-event analyses. Control experiments performed in the absence of DNase I confirmed that continuous scanning alone did not induce detectable DNA fragmentation under the imaging conditions used.

For protamine-condensation protection assays, pre-condensed rod-like or toroidal DNA assemblies were imaged following DNase I addition under identical buffer conditions. Condensed DNA structures remained morphologically stable throughout prolonged imaging periods, whereas uncondensed DNA filaments underwent progressive degradation.

All experiments were independently repeated at least three times using freshly prepared substrates, DNA samples, and enzyme solutions to confirm reproducibility of the observed structural and dynamic behaviors.

Image processing and quantitative analysis

Raw HS-AFM image sequences were processed using ImageJ/Fiji (NIH, https://imagej.net). Individual frames were subjected to first-order plane subtraction and mild Gaussian filtering where necessary to improve signal-to-noise ratio while preserving overall molecular morphology. Drift correction was performed using stable substrate landmarks or fixed structural reference points when required. Identical processing parameters were applied consistently across all frames within the same dataset.

Height profiles, contour measurements, and particle-height analyses were extracted from processed HS-AFM images using ImageJ measurement tools. Sequential image stacks were exported as AVI movies and assembled using Adobe Creative Cloud software (Adobe Systems Inc.). Brightness and contrast adjustments were applied uniformly across entire image frames and did not alter the underlying quantitative analyses.

For quantitative degradation analyses, DNA-associated signals were identified using fixed threshold-based segmentation. Projected area, mean height, integrated height, and corrected integrated density values were measured frame-by-frame within defined regions of interest. To compensate for variation in initial DNA coverage between imaging fields, corrected integrated height values were normalized to the initial DNA-covered area of each trajectory. Time-dependent decay curves were analyzed using GraphPad Prism (GraphPad Software) and fitted using a one-phase exponential decay model.

To compensate for differences in initial DNA coverage between independent imaging fields, corrected integrated height values were normalized to the DNA-covered area measured in the first frame of each trajectory. The resulting parameter, termed normalized corrected integrated height per initial area, was calculated as:

$$\mathrm{Normalized}\mathrm{corrected}\mathrm{integrated}\mathrm{height}\mathrm{per}\mathrm{initial}\mathrm{area}=\frac{{\mathrm{Corrected}\mathrm{integrated}\mathrm{height}}_t/\mathrm{Initial}\mathrm{DNA}\mathrm{area}}{{\mathrm{Corrected}\mathrm{integrated}\mathrm{height}}_{t=0}/\mathrm{Initial}\mathrm{DNA}\mathrm{area}}$$

where t represents the imaging time point.

To quantify recurrent nuclease engagement, individual DNase I localization events were manually annotated from time-resolved HS-AFM trajectories based on transient globular particle appearance adjacent to DNA contours. Spatial coordinates were extracted using ImageJ/Fiji and analyzed using custom Python scripts. Localization events occurring within a predefined reassignment radius (~1.8–4.5 nm depending on dataset resolution) were operationally classified as belonging to the same interaction site. Non-consecutive revisiting of previously assigned sites was classified as recurrent engagement (“rebinding”).

Rebinding frequency was calculated as:

$$\mathrm{Rebinding}\mathrm{frequency}=\frac{\mathrm{Number}\mathrm{of}\mathrm{rebinding}\mathrm{events}}{\mathrm{Total}\mathrm{number}\mathrm{of}\mathrm{binding}\mathrm{episodes}}$$

where total binding episodes corresponded to temporally separated interaction events identified along the DNA contour.

To evaluate whether recurrent engagement could arise from stochastic sampling alone, three randomization models were implemented using custom Python-based simulations: unconstrained two-dimensional randomization, contour-length constrained randomization, and contour-preserved randomization. Simulations were repeated over 10,000 iterations per dataset, and experimentally observed rebinding frequencies were compared with simulated distributions. Individual DNase I localization events were analyzed independently, including recurrent events at previously visited DNA regions; thus, repeated end-distance or local-curvature values could arise when events mapped to the same or spatially equivalent contour positions. Randomized datasets were generated independently for each simulation. Because rebinding frequency was calculated from a finite number of binding episodes, simulated frequencies were discrete and identical values could occur across independent simulations.

For spatial analyses, DNA contours were manually traced from HS-AFM images to reconstruct two-dimensional trajectories. Nuclease localization events and cleavage sites were projected onto reconstructed DNA contours to evaluate nearest-end preference, local curvature enrichment, and topology-associated localization patterns. Cleavage events were operationally defined as the first frame exhibiting detectable contour discontinuity accompanied by separation of daughter DNA ends.

Independent HS-AFM trajectories acquired from separate imaging fields, substrates, and enzyme additions were analyzed throughout the study to evaluate reproducibility of recurrent engagement, cleavage-site preference, dwell-time behavior, and fragmentation dynamics. Quantitative analyses were performed separately for individual trajectories prior to comparison across datasets.

Dwell time was defined as the number of consecutive frames during which a nuclease particle remained associated with the same reassigned interaction site. Residence durations were converted into approximate interaction times based on frame acquisition intervals and compared between target and non-target DNA regions. Because HS-AFM acquisition occurs at finite temporal and spatial resolution, dwell-time measurements were interpreted as operational descriptors of local interaction persistence rather than direct measurements of catalytic-state lifetimes.

Height-profile analyses shown in Fig. 1g and Fig. 1h were performed using the in-house UMEX Viewer software. Cross-sectional height profiles were extracted from protamine-condensed DNA structures and DNase I interaction sites to evaluate enzyme association and condensate structural stability during continuous HS-AFM imaging.

Quantification of substrate-dependent DNA degradation dynamics

For quantitative analysis of substrate-dependent DNA degradation kinetics shown in Fig. 2, HS-AFM image sequences were analyzed using ImageJ/Fiji (NIH). DNA-associated signals were identified using threshold-based segmentation applied consistently across all frames within each dataset. Projected DNA area, mean height, and integrated height were measured frame-by-frame within defined regions of interest. DNA signal thresholds were typically defined within the range of ~1.2–2.7 nm on mica-supported substrates. On supported lipid-bilayer substrates, the apparent DNA height shifted to ~4.7–5.9 nm due to the underlying lipid layer, and threshold ranges were adjusted accordingly for DNA quantification. Background-corrected integrated height values were calculated by subtracting substrate-associated background contributions from the total integrated signal. To compensate for variability in initial DNA adsorption between imaging fields, corrected integrated height values were normalized to the DNA-covered area measured in the first frame of each trajectory. Time-dependent decay curves were analyzed using GraphPad Prism (GraphPad Software) and fitted using a one-phase exponential decay model to estimate apparent degradation kinetics and relative signal decay across substrates and experimental conditions.

Time-resolved height analysis of recurrent nuclease engagement

For analyses shown in Fig. 4c, d regions of interest corresponding to recurrent DNase I interaction sites were manually defined from HS-AFM image sequences using ImageJ/Fiji. Local height values were extracted frame-by-frame using identical measurement parameters throughout each trajectory. DNA baseline signals typically exhibited apparent heights of ~2–3 nm, whereas transient DNase I association events produced localized height increases of ~4–5 nm. Repeated transient height elevations at non-consecutive time points were operationally classified as recurrent nuclease engagement events.

HS-AFM analysis of individual DNase I particle morphology

For the analysis shown in Fig. 3g, individual DNase I molecules were imaged under liquid conditions following low-concentration deposition onto freshly cleaved mica. HS-AFM image sequences were processed using first-order plane subtraction and mild Gaussian filtering. Representative particles were selected based on image stability and isolation from neighboring structures. Apparent particle morphology and dimensions were compared qualitatively with AFM-compatible structural simulations generated from AlphaFold3 and crystallographic models. Cross-sectional height profiles and lateral dimensions were extracted using ImageJ measurement tools.

Single-molecule visualization of DNase I interaction dynamics

For the analyses shown in Fig. 4a, time-resolved HS-AFM imaging was performed on λ-DNA substrates undergoing DNase I-mediated degradation under liquid conditions. Sequential image frames were analyzed to identify transient DNase I localization events, contour disruption, and cleavage-associated fragmentation dynamics. DNase I interaction events were operationally identified based on transient globular particles adjacent to DNA contours. Structurally vulnerable DNA regions, including exposed termini, curved segments, and bent conformations, were manually annotated from reconstructed DNA contours. Color-coded numbered circles shown in Fig. 4a correspond to representative recurrent interaction sites identified during frame-by-frame trajectory analysis. Cleavage events were operationally defined as the first detectable frame exhibiting contour discontinuity accompanied by measurable separation of newly generated DNA ends.

Quantification of cleavage-associated DNA-end separation dynamics

To evaluate whether contour discontinuities observed during HS-AFM imaging corresponded to localized cleavage events rather than global filament drift, cleavage-associated DNA-end separation dynamics were quantified from representative DNase I degradation trajectories. Newly generated DNA ends were manually tracked frame-by-frame using ImageJ/Fiji, and Euclidean distances between daughter DNA termini were measured following cleavage initiation. Time-to-resolve was defined as the elapsed time between the first detectable contour discontinuity and the first frame in which daughter DNA ends became spatially distinguishable. To distinguish localized rupture propagation from global filament motion, displacement of newly generated DNA ends was compared with movement of original pre-existing DNA termini within the same imaging interval. Quantitative analyses were performed independently for trajectories acquired on both Ni²⁺-functionalized mica and APTES-modified substrates.

Statistical analysis

Quantitative data were analyzed and visualized using GraphPad Prism³⁹ (GraphPad Software, CA, USA). The number of independent experiments, particles, trajectories, or measurements analyzed for each dataset is indicated in the corresponding figure legends. Unless otherwise specified, data are presented as mean ± standard deviation (SD) or as individual data points with summary statistics. Exact p values, statistical tests, sample sizes, and definitions of error bars are provided in the corresponding Source Data sheets. Simulation-based probabilities were calculated empirically from randomized datasets generated by custom Python scripts. Because single-molecule HS-AFM trajectories exhibit inherent stochastic variability, quantitative analyses were interpreted based on reproducible statistical enrichment patterns observed across independent datasets rather than strict deterministic event recurrence. Statistical significance was generally defined as p < 0.05.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.

Author contributions

R.W. conceived and designed the study. T.A. designed the HS-AFM (Washi / Tsubaki). J.Y. performed most of the HS-AFM experiments, data analysis, and manuscript revision. Y.Q. contributed to HS-AFM experiments and data analysis. K.L. contributed through discussion and interpretation of the data. All authors analyzed the data. R.W. wrote the manuscript. All authors read, discussed, and approved the final manuscript. R.W. supervised and administered the study.

Peer review

Peer review information

Nature Communications thanks the anonymous reviewers for their contribution to the peer review of this work. A peer review file is available.

Funding

This work was supported by The World Premier International Research Center Initiative (WPI), MEXT, Japan. This work was supported by WISE Program for Nano-Precision Medicine, Science, and Technology of Kanazawa University by MEXT (to J.Y.), JST, Grant Number JPMJFS2116 and JST SPRING, Grant Number JPMJSP2135 (to Y.Q.); MEXT/JSPS KAKENHI grant number 24K18449 (to K.L.) and 22H05537, 22H02209, 23H04278 and 24H01276, 25H02360, 26K01637, 26K22904 (to R.W.W) from MEXT Japan and by JST CREST grant number JPMJCR22E3 (to R.W.W), and by grants from the Hokuriku Bank grant, The Astellas foundation for research on metabolic disorders, Japan (to K.L.), the Takeda Science Foundation, Japan (to R.W.W) and the Shimadzu Science Foundation, Japan (to R.W.W).

Data availability

All data supporting the findings of this study are available within the Article, its Supplementary Information, and the accompanying Source Data file. Additional data and materials are available from the corresponding author upon request. Source data are provided with this paper.

Code availability

Custom Python scripts used for data analysis in this study are available at Zenodo under DOI: 10.5281/zenodo.21478288.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at https://doi.org/10.1038/s41467-026-77354-x.